The pathogenesis of cerebral small vessel disease and vascular cognitive impairment

Abstract

Cerebral small vessel disease (cSVD) is the broadly accepted term nowadays to designate a heterogeneous group of diseases caused by in situ damages of small brain vessels commonly related to aging, hypertension or genetic factors. Cardinal neuroimaging features include small (< 20 mm) infarcts or lacunes, cerebral microbleeds, white matter hyperintensities, enlarged perivascular spaces and brain atrophy. Overall, cSVD represents one of the major problems facing global society today, causing a quarter of all ischemic strokes, the vast majority of spontaneous hemorrhages, and accounting for 20% or more of all dementias. Yet, mechanisms of cSVD are still incompletely understood, and we have no effective proven treatments other than risk factor modification. Recently, major progress in understanding the underlying disease mechanisms has occurred thanks to novel approaches including advanced molecular, genetic and imaging tools. Here, we provide a comprehensive and critical appraisal of the biggest advances in our understanding of how cSVD affects the structure and function of small brain vessels, causes brain lesions and alters cognition. To set the stage, we begin by reviewing the molecular anatomy and physiology of healthy small brain vessels and report on the milestones from the medical literature, starting in the 1850s, that have laid the foundation for the “modern” definition of cSVD. We conclude by discussing the framework for clinical interventions that will emerge from these novel insights. We also highlight the outstanding questions to address and challenges to tackle to move the field forward.

Graphical Abstract

1. Introduction

The brain is the basis for our extraordinary motor and cognitive abilities. It is considered as the most complex organ of our body with almost one hundred billion neurons, 60 trillion of neuronal connections and one hundred billion glial cells (1, 2). The brain consumes up to 20% of total body energy budget despite being only 2% of the body mass; however, it has limited fuel reserves (3). Importantly, brain health relies on healthy brain vessels.

The brain is supplied by two main pairs of extracranial arteries that run along the neck, enter the skull and become connected at the base of the brain by a ring of anastomosing arteries. From this ring arise arteries (pial/leptomeningeal arteries) that run along the surface of the brain within the subarachnoid space, divide progressively into smaller arteries before penetrating into the brain in a centripetal fashion (4). This is in contrast to most other organs that receive their blood supply from arteries arriving in their hilum and branching towards the periphery. Penetrating arterioles further divide into smaller vessels that feed about 400 miles of capillaries where energy substrates (oxygen and glucose) are mainly delivered to the parenchyma (5). Pial arteries and penetrating arterioles are initially separated from the brain parenchyma by a fluid-filled perivascular compartment, referred to as the perivascular space. This space progressively collapses as arterioles penetrate deeper into the brain (6). At the level of capillaries, vascular cells are entirely covered by processes of glial cells and occasionally neurons (5).

There is a remarkable symbiotic relationship between brain cells and cerebral blood vessels that is involved in all facets of normal brain function. Arteries and arterioles adapt their diameter in the face of arterial blood pressure (BP) fluctuations (autoregulation) to protect, to a certain extent, the brain from deleterious decreases or increases in the cerebral blood flow (CBF) (7). There is a sophisticated and often redundant ensemble of mechanisms, which engages the entire cerebrovascular tree (from pial arteries to capillaries), to couple neural activity to blood supply of energy substrates at the right time and place (neurovascular coupling) (8). Unlike the vasculature in other organs, blood vessels form a tight protective barrier between the blood and the brain (blood brain barrier) (9). Finally, brain vessels from arteries to veins, participate in the process of cleaning the brain of its waste products and regulating the movement and amount of brain extracellular fluids (10). Hence, any damage to brain vessels, from pial arteries to capillaries and veins may have major consequences leading to stroke and cognitive impairment. The deep brain regions are more susceptible to brain vessel injury, likely because of the peculiar anatomy of the cerebrovasculature.

Cerebral small vessel disease: definition and classification

Cerebral small vessel disease (cSVD) is the umbrella term used to designate a heterogeneous group of chronic cerebrovascular diseases related to in situ pathology of brain vessels (11). cSVD can be caused by multiple pathologies which can be broadly classified into two categories; amyloid and non-amyloid cSVD (Table 1). Cerebral amyloid angiopathy (CAA), is a specific form of cSVD especially prevalent in the elderly, which is characterized by the deposition of ß-amyloid in the wall of leptomeningeal vessels and cortical vessels and most commonly presents with lobar intracerebral hemorrhage (ICH) (12). Non-amyloid cSVD comprises a group of pathologies commonly related to aging, hypertension, or genetic factors, although rarely caused by other conditions including radiotherapy and vasculitis (Table 1). Non-amyloid cSVD, thereafter referred to as simply “cSVD”, is the focus of this review, whereas CAA, which has been the subject of recent excellent reviews, is not discussed further (12, 13).

Table 1.

Classification of cSVD

Category	Disease	Details
Cerebral Amyloid angiopathy (CAA)	Sporadic CAA
	Familial CAA
	Iatrogenic CAA	Recent description of cases following neurosurgery operations
Non-amyloid cSVD	Sporadic	Major risk factor is hypertension so sometimes called “Hypertensive” cSVD
	Familial	CADASIL is the most common
	Inflammatory	Include systemic lupus erythematous and other connective tissue disorders; primary angiitis (vasculitis)
	Post-radiation	Usually given for intracranial tumors

The two core pathological features of cSVD include small (less than 15–20 mm diameter) infarcts in the deep brain regions (also known as lacunes or lacunar infarcts) and widespread white matter (WM) lesions (known as leukoaraïosis or vascular leukoencephalopathy due to their low density on computed tomography (CT), or as white matter hyperintensities (WMH) due to their high density on T2-weighted magnetic resonance imaging (MRI) (Figure 1). Other typical lesions on MRI include enlarged perivascular spaces, formerly named “état criblé”, cerebral microbleeds, intracerebral hemorrhages and brain atrophy (14).

Figure 1. Two core pathological features of cSVD.

A- Lacune (blue arrow) visible as a round subcortical, fluid filled cavity on FLAIR-MRI sequence (a T2-weighted sequence in which the free water is suppressed, note signal from the cerebrospinal fluid in the ventricles is low signal). B-C-Widespread white matter lesions (orange arrows) visible as low-density lesions on computed tomography (B) and hyperintense lesions on FLAIR-MRI sequence (C). Used with permission from Hugh Markus (doi: 10.1177/17474930241279888) under CC-BY 4.0 license.

The link between cSVD, stroke and vascular cognitive impairment

cSVD is considered as one of the major health problems affecting society today. It manifests predominantly as stroke and vascular cognitive impairment. Stroke is defined by the sudden loss of focal neurological function due to ischemia or hemorrhage in the corresponding part of the brain (15). cSVD accounts for a quarter of all strokes. It causes lacunar stroke which comprises ~30% of all ischemic strokes, manifesting as small artery occlusion (the 3^rd subtype of ischemic stroke according to the TOAST classification), and accounts for the vast majority of spontaneous intracerebral hemorrhage (ICH) (16) (Table 2).

Table 2:

Classification of strokes

Category	Details	Subtypes or main causes
Ischemic stroke	~80% of all strokes (17)	The 5 subtypes recognized by the Trial of Org 10172 in Acute Stroke Treatment (TOAST) (15, 18) are: 1- Cardioembolism 2- Large-artery atherosclerosis involving aortic arch, large extracranial arteries (carotid, vertebral or basilar artery) or large intracranial arteries leading to artery-to-artery embolism 3- Small-vessel occlusion 4- Stroke of other determined etiology 5- Stroke of undermined etiology
Intracerebral Hemorrhage	~20% of all strokes, but the most devastating form of stroke (19)	The 4 main causes (19) include: 1- CAA or non-amyloid cSVD (~80% of all ICH) 2- Vascular malformations 3- Cerebral sinus venous thrombosis 4- Bleeding into a brain tumor

Vascular cognitive impairment refers to the entire spectrum of cerebrovascular diseases contributing to cognitive impairment ranging from mild cognitive impairment to dementia (20–22) (Table 3). cSVD is now widely recognized as the most common pathology underlying vascular cognitive impairment and vascular dementia, which accounts for about 20% of all dementias (16). cSVD causes subcortical ischemic vascular dementia, the 2^nd subtype of vascular dementia according to the Vascular Impairment of Cognition Classification Consensus Study (20). However, the impact of cSVD on the overall dementia is much greater. As well as contributing to the 20% of dementia cases classified as vascular dementia, the presence of cSVD greatly increases the chance that neurodegenerative pathologies, such as Alzheimer’s disease, will result in clinical dementia during life. For example in the NUN study only half of those individuals with Alzheimer’s disease pathology at post-mortem had experienced clinical dementia during life in the absence of contribution of cerebrovascular disease, but this increased to over 90% in those with additional cSVD pathology (23). This interaction is important on a population basis as most patients with dementia have more than one underlying pathology (24). Therefore, cSVD is likely to contribute to over half of all dementia cases.

Table 3:

Four major subtypes of vascular dementia

Subtypes	Details
Multi-infarct dementia	relates to the involvement of multiple large cortical infarcts
Subcortical ischemic vascular dementia	relates to the contribution of brain lesions primarily located subcortically
Post-stroke dementia	defined as dementia that begins after, but within 6 months, of stroke that does not recover
mixed dementia	include all dementia due to mixed (neurodegenerative and vascular) pathologies

The burden of cSVD is even higher if we consider silent cSVD (cSVD-related lesions detected incidentally on MRI scans). Silent cSVD is extremely common as we age, affecting most people above 80 (25), and has been shown to be a risk factor for lesser degrees of age-related cognitive decline in non-demented individuals, which can nevertheless be disabling. cSVD affects all societies but the burden in lower- and middle-income countries is even greater (26). Despite this enormous health burden, there are as yet no mechanism-based treatments for these devastating diseases and therapeutic options are currently limited to preventative measures targeting modifiable risk factors. The main reason is that their pathogenesis is complex and far from being understood.

Objectives of the review

The aim of this review is to provide a detailed description of the conceptual meaning of cSVD and an appraisal of the current understanding of the complex pathogenesis of these diseases. We shall first describe the anatomy and molecular composition of brain vessels and how novel molecular and imaging tools during the past decade have revolutionized our understanding of the multifaceted functions of brain vessels. Next, we will review the medical literature, starting in the 1850s, and report on the milestones that have laid the foundation for the “modern” definition of cSVD. We shall then describe the main clinical and neuroimaging features of cSVD, highlighting the importance of brain MRI to pinpoint the diversity, the extent and pathological underpinning of brain lesions. Then, we will review the major advances in our understanding of the modifiable risk factors and genetic landscape of cSVD, recently transformed by large-scale biomedical databases, international collaborative networks and the affordability of high-throughput genotyping and sequencing. We next discuss animal models used for the study of of cSVD. cSVD results from functional or structural alterations of brain vessels and each microvascular compartment can be differentially affected by pathological processes underlying cSVD. Therefore, we shall review the structural, mechanical and functional alterations of brain vessels associated with sporadic or monogenic cSVD, the mechanisms of vascular pathology and brain lesions in individual forms of cSVD, highlighting shared mechanisms among them, challenging some well-rooted views as well as discussing novel avenues. To achieve this, we will combine complementary information gained from studies in patients and clinically relevant mouse models of cSVD. Then, we shall consider how cSVD causes cognitive impairment, a key and ultimate feature of these diseases, that can also arise early in the course of the disease. A growing body of literature suggests that age-related cognitive decline is driven by an interplay between neurodegenerative processes and small vessel pathologies. However, we will not be looking into this important issue, because we deem that this topic should deserve a full review. Finally, we conclude by reviewing the available therapeutics and discussing frameworks for clinical studies, highlighting the many challenges posed by these diseases. We conclude by discussing the many challenges and outstanding questions to address to move the field forward.

2. Anatomy and physiology of brain vessels

2.1. Anatomy of brain vessels

The brain is enveloped by meninges that are composed of 3 layers or maters, comprising from the inner to the outer, the pia, the arachnoid and the dura, which is adherent to the skull (27). The brain is vascularized by the two internal carotid arteries (anterior circulation), and the two vertebral arteries that join intracranially to give rise to the basilar artery (posterior circulation). The anterior and posterior circulations are connected by a ring of anastomosing arteries called the Circle of Willis that is situated at the base of the brain (Figure 2A). From these extracranial arteries and the Circle of Willis arise major arteries which divide into progressively smaller arteries that run along the surface of the brain within the subarachnoid space of the meninges (these vessels are often referred to pial or leptomeningeal arteries) where they form numerous anastomoses (4). Pial arteries dive into the brain at almost right angles to supply various regions of the brain in a centripetal fashion. Pial arteries and arterioles are richly innervated by sympathetic, parasympathetic and trigeminal nerves (extrinsic innervation) (4).

Figure 2. Blood supply of of the brain.

A- Schematic of the main arteries of the Circle of Willis supplying the brain. B- Different types of blood supply according to brain areas: (1) cortical layers are supplied by superficial, intermediate, deep and terminal branches of intracortical arteries, (2) the deep white matter is supplied by terminal branches of the long medullary arteries, (3) the U-fibers are supplied by the earliest branches of medullary arteries and the terminal branches of the longest cortical arteries and (4) the thalamus and basal ganglia are supplied by perforating arteries that originate from parent arteries at the base of the brain.

The cortex is vascularized by a dense network of short, intermediate and long cortical arteries, according to their degree of cortical penetration, that run perpendicular to the brain surface and give off many branches forming inverted elm trees (Figure 2B, area 1) (28). In contrast to pial arteries, there are no anastomoses between cortical arteries which make them a bottleneck in the perfusion of the cortex (29). The deep WM, which is present in human but not in rodent, is supplied by long medullary arteries that penetrate the cortex straight with almost no branching and then run through the WM towards an angle of the lateral ventricle; medullary arteries give off 5–10 long side branches and further divide into smaller arteries (Figure 2B-area 2) (30, 31). Unlike the deep WM, the subcortical U-fibers (short association fibers representing connections between adjacent gyri of the brain), which are usually spared in cSVD, have two sources of supply: the earliest branches of medullary arteries and the terminal branches of the longest cortical arteries that run along the subcortical U-fibers (Figure 2B-area 3) (32). Like the deep WM, the internal capsula, the thalamus and basal ganglia have a single source of blood supply consisting in long perforating arteries that originate from parent arteries at the base of the brain at an acute angle (Figure 2B-area 4) (31). The pattern of arterial distribution of the cerebellum is almost similar to the cortex (33). In the cortex, arteries/arterioles (smaller arteries) are encased by one sheet of pia matter, which reflects from the surface of the brain and separates the vessel from the parenchyma, whereas in the WM and basal ganglia, there are 2 sheets of pia matter; veins are surrounded by an incomplete layer of pia mater (34).

Blood flows from arteries, arterioles, capillaries to veins. Blood from the cerebral cortex and subcortical WM is routed towards superficial veins that run through the cortex outward to join the pial veins (superficial venous system) whereas blood from the deep brain regions (deep WM, 3^rd ventricle, basal ganglia, thalamus) drain into the internal cerebral vein or the basal vein, which heads towards the great cerebral vein of Galen (deep venous system) (35). Both superficial and deep venous systems drain into the cranial venous sinuses, which are sandwiched between the two layers of the dura mater. From here, blood exits the brain by emptying into the jugular veins (35). Originally described in the 1600s by Paolo Mascagni, the meningeal lymphatics were recently functionally characterized in humans (36) and rodents (37). Meningeal lymphatic vessels are located in the dura matter where they align with dural blood vessels; they exit the cranium via the foramina together with the venous sinuses or through the cribriform plate (under the olfactory bulb) and drain into the cervical lymph nodes (36–37).

Perivascular spaces (PVS), often referred to as Virchow-Robin spaces, are fluid-filled passageways surrounding brain blood vessels (6). PVS are delineated inside by the basement membrane of mural cells and outside by the glia limitans, formed by astrocyte end feet (Figure 3) (38). Despite their initial description over 170 years ago, the anatomy of PVS is still incompletely understood. The common view is that PVS surround penetrating arteries and arterioles, disappear as arterioles morph into capillaries, and reappear around veins (34). However, this statement comes from histological analyses of postmortem fixed tissues in which PVS can be artificially shrunk or conversely enlarged and thus should be taken with a grain of salt. What is indisputable is the presence of PVS surrounding pial arteries in the subarachnoid space that can be detected after cerebrospinal fluid (CSF) tracer injection by 2-photon imaging in rodents and by MRI in humans (39–41). Presence of PVS around pial veins is inconstant in rodents (40) and debated in humans (39, 42). What is less clear is what happens with vessels inside the brain. In rodents, two-photon imaging or histological analysis after CSF tracer injection shows that PVS diminish along the penetrating arteries, but how far is unclear (43, 44). In humans, a few PVS can be seen in young healthy individuals by MRI in the basal ganglia and in the centrum semiovale. PVS in the centrum semiovale are located a few millimeters beneath the cortex, running centripetally towards the lateral ventricle. 7T-MRI suggests that PVS correlate spatially with arteries but does not find evidence of PVS around veins (6).

Figure 3.

Schematic diagram of the perivascular space.

The microvasculature connectome has been characterized in whole brains of adult mice with sub-micrometer resolution (45–47). Penetrating veins are more numerous that penetrating arterioles, by a factor of 2.6 in the cortex. However, in humans, arterioles outnumber venules by a factor of ~3 to 5 (47, 48). Capillaries form a rich interconnected network that represent the vast majority (> 95%) of the total cerebrovascular length. Capillary length density (expressed in m/mm3) varies across brain regions by a factor of about 3, with the WM having the lowest density; noting that the capillary density is not correlated with the neuronal density (45). Regions with high capillary length density have relatively more short capillary branches and vice versa. The distance from any point in the parenchyma to the lumen of the nearest vessel ranges from 10 μm (in regions of high length density of capillaries) to 20 μm (in regions of low length density of capillaries, such as the WM) (46). On average, a capillary is closest from an artery (~5 branch point) than from a vein (~7 branch point) (45).

2.2. Cellular and molecular composition of brain vessels

2.2.1. Endothelial cells

Endothelial cells form a single cell layer that lines all blood vessels (arteries, capillaries and veins). Single-cell RNA sequencing of murine brain vascular cells has confirmed the preferential expression of certain genes in arterial (Vegfc, Bmx, Efnb2), capillary (Mfsd2a, Tfrc) and venous (Nr2f2, Vwf) endothelial cells (49). However, endothelial cells display a gradual change, called zonation, rather than an abrupt change in gene expression along the arterio-capillary-venous axis, indicating that endothelial subtypes are a continuum of phenotypes rather than distinct subtypes. Single-cell transcriptome analyses have also provided new insights into endothelial specialization. As an example, transcription factors are overrepresented on the arterial side, whereas blood brain barrier-associated transporters dominate in capillaries and veins (49). Comparison of the single-cell transcriptomes between human and mouse brain endothelial cells revealed large divergences (50). However, the zonal functional organization is conserved between mice and humans, although the sets of zonally distributed genes are divergent with only ~10% of genes conserved between species (50–52)

2.2.2. Mural cells

Mural cells exhibit changes in cell morphology, protein markers and gene expression profiles along the arteriovenous axis (Figure 4). Arteries and arterioles are covered by mural cells called arterial and arteriolar smooth muscle cells (SMCs) respectively; these SMCs have a ring-shaped morphology and are separated from the endothelial cells by a well-defined internal elastic lamina which is made of a thick layer of elastin. As small arterioles branch to form an extensive capillary network, they develop a distinguishable arteriole-capillary transition (ACT) zone/segment that lacks an elastic lamina and is surrounded by contractile mural cells (alternatively referred to as transitional or hybrid cells, ensheathing pericytes, contractile pericytes or precapillary SMCs) that possess more irregular ensheathing processes and a rounded nucleus (53–57). Mural cells covering arteries, arterioles and the ACT zone are characterized by high expression levels of contractile machinery-related proteins, like smooth muscle α-actin (encoded by ACTA2) or smooth muscle myosin heavy chain (encoded by MYH11), which gradually decrease from the artery to the ACT zone (54). A novel structure, known as the precapillary sphincter, has been identified in the cortex at the transition between some arterioles and this ACT zone (58). Sphincters are encircled by elastin and contractile SMCs. At the level of the capillary network, the endothelial tube is covered by thin-strand pericytes, which are nestled into the endothelial basement membrane. Pericytes share membranous interdigitations called “peg-and-socket” interactions with endothelial cells that likely facilitate endothelial-pericyte communication (59). Thin-strand pericytes express no or extremely low levels of contractile SMC markers but high levels of proteins like the pore-forming subunit Kir6.1 (encoded by KCNJ8) and the regulatory subunit sulfonylurea receptor SUR 2 (encoded by ABCC9) of the ATP-sensitive potassium (K_ATP) channel, ascribed to pericyte markers (60). Veins are covered by sparse mural cells with a stellar shape, usually called venous SMCs. As mural cells transition from pericytes to venous SMCs, expression levels of pericyte markers gradually decrease and low levels of contractile SMC markers reappear (49, 54).

Figure 4.

Simplified depiction of the four main vascular compartments including (1) the artery/arteriole, (2) the arteriole-capillary transition (ACT) zone, (3) the capillary (4) the venule/vein and their associated mural and perivascular cells; a sphincter may exist at the transition between the arteriole and the ACT zone. Mural cells form two distinct phenotypic continua including: (1) arterial SMCs (aSMC) and transitional cells with a gradual decrease in the expression of contractile markers from aSMCs to transitional cells and (2) pericytes and venous SMCs (vSMC) with a gradual loss of pericyte markers and the acquisition of low levels of SMC contractile markers. IEL, internal elastic lamina; PVFB, perivascular fibroblast; PVM, perivascular macrophage.

At the transcriptomic level, mural cells exhibit a zonation pattern that differs from the single arteriovenous continuum of endothelial cells. Specifically, hierarchical clustering for gene expression separates mural cells into two distinct phenotypic continua. One subclass includes pericytes in a continuum with venous SMCs through the gradual loss of pericyte markers and the acquisition of low levels of SMC markers; a second subclass includes arteriolar SMCs in continuum with arterial SMCs through the progressive acquisition of SMC contractile markers (49). It is yet unclear in which of these two subclasses mural cells of the ACT zone distribute. However, the prediction is that these cells should be closer to arteriolar SMC based on their expression of contractile markers. Differences between mice and humans include the presence in humans of two distinct subtypes of pericytes, the so-called T-pericytes, which are enriched for small-molecule transmembrane transporters, and M-pericytes, enriched for extracellular matrix (ECM) genes (50). Also present in humans is a cell type not previously identified in the mouse called “fibromyocytes”, identified on the basis their low expression of contractile SMC genes and high expression of genes commonly expressed in fibroblast and macrophage (52).

2.2.3. Perivascular cells

The two major resident cells of PVS are perivascular macrophages (PVMs) and perivascular fibroblasts (PVFBs) (Figure 4). PVMs are elongated cells that represent a population of resident macrophages that can be distinguished from microglial cells by their expression of MRC1 (mannose receptor C-type 1), also known as CD206, and LYVE1 (lymphatic vessel endothelial hyaluronan receptor-1) (61). PVFBs are fibroblast-like cells distinguishable from SMCs and pericytes by their expression of collagen 1 and absence of smooth muscle α-actin and Chondroitin Sulfate Proteoglycan 4 expression (49). Recent cell-fate–mapping experiments and in vivo imaging in the mouse have shown that PVMs and PVFBs develop concurrently. Both cell types originate from the meninges postnatally and cover vessels after SMCs and astrocytic endfeet are in place (61, 62). Single-cell transcriptome analyses provide a molecular definition of PVFBs and PVMs. The PVFB transcriptome is enriched for genes encoding matrisome proteins (the ensemble of proteins that constitute the extracellular and extracellular matrix-associated proteins) (49, 63, 64). PVMs actually represent at least two different populations characterized by distinct immunologic signatures. Major histocompatibility complex II positive (MHCII⁺) PVMs are enriched for genes encoding proteins involved in antigen processing and presentation, cell-adhesion molecules and the Toll-like receptor signaling pathway, whereas LYVE1⁺ PVMs express high levels of scavenger markers and genes associated with lysosome activity and endocytosis (65, 66).

2.3. The multifaceted functions of healthy brain vessels

Small brain vessels perform multifaceted roles, most of which are unique to the brain. Autoregulation maintains cerebral blood flow (CBF) relatively stable in the face of moment-to-moment fluctuations in arterial BP. Neurovascular coupling – the ensemble of mechanisms that mediate activity-dependent increases in blood perfusion (functional hyperemia) – ensures an appropriate delivery of nutrients and oxygen (O₂) in response to changes in local neural activity. The blood brain barrier (BBB) limits the movement of ions, amino-acids, molecules and cells in and out of the brain. The glymphatic system and the intraarterial drainage pathway are posited to play a critical role in clearing wastes from the brain parenchyma and regulating brain fluid movements.

2.3.1. Static and dynamic autoregulation of CBF

Autoregulation relates to the response of cerebral vessels to changes in perfusion pressure, which, under physiological conditions, is mostly determined by arterial BP (67). Within a range of BP, usually between 60–150 mmHg, autoregulation minimizes variations in CBF (68) and protects the brain from ischemia during drops in BP and from hemorrhage during surge in BP (4). Autoregulation is assessed by measuring the relationship between BP and CBF. The terms static and dynamic autoregulation refer to the experimental conditions and timescales under which BP and CBF are measured (Figure 5). Static autoregulation means that BP and CBF have reached a steady-state over a long-time interval (69). Dynamic autoregulation describes changes in CBF in respond to rapid (over seconds) changes in BP that occur spontaneously or during for example postural changes or physical activity (69). Therefore, static and dynamic autoregulation can be considered as the two ends of the autoregulation spectrum (7).

Figure 5.

Graphs illustrating the concepts of static (left) and dynamic autoregulation (right).

Autoregulation operates by modulating vessel diameter and resistance, the latter of which is inversely proportional to the fourth power of the vessel radius (70). Hence, a small change in vessel diameter will have a huge impact on vessel resistance (7). SMCs are the primary sensors of changes in BP and the primary effector cells that drive intravascular pressure-dependent changes in diameter (myogenic response), increasing or decreasing vessel diameter in response to BP decreases and increases respectively (7, 71). The myogenic response is initiated by the rise of intravascular pressure, which causes depolarization (an increase in membrane potential) of SMCs that in turn activates voltage-dependent Ca²⁺ channels, resulting in the entry of calcium (Ca²⁺) and subsequent vasoconstriction; this calcium influx via L-type calcium channels is essential for SMC contraction (72). Other positive contributors to pressure-induced constriction expressed in SMCs include the transient receptor potential TRPM4 and TRPC6 channels (73). In contrast, large-conductance, Ca²⁺-sensitive potassium (K⁺) (BK_Ca) channels and voltage-gated potassium (K_V) channels act as negative modulators of myogenic constriction by hyperpolarizing SMCs (74). Moreover, release of nitric oxide (NO), a soluble gas produced by endothelial cells, and also endothelial-dependent hyperpolarization can negatively modulate myogenic tone by acting directly on SMCs (75). Autoregulation takes place predominantly in pial and parenchymal arteries (7); noting that ex vivo experiments using pressurized isolated brain arteries have shown that parenchymal arteries develop more myogenic tone than pial arteries, because the former are more depolarized than the later at lower pressures (76). Capillaries, venules and veins have minor or even no contribution (7).

2.3.2. Neurovascular coupling

Recent years have witnessed significant improvement in our understanding of the molecular and cellular mechanisms of neurovascular coupling, largely owing to the continuing development of genetic and pharmacological strategies for manipulating signaling pathways, improved genetically encoded Ca²⁺-indicator models, and advances in in vivo two-photon imaging (77). Neurovascular coupling involves the coordinated crosstalk between excitatory and inhibitory neurons, astrocytes, endothelial and mural cells. The main driver of CBF increase is the increased activity of neurons that can release vasoactive molecules such as ions (K⁺ or H⁺), NO, adenosine, arachidonic acid metabolites, glutamate or neuropeptides (78–80). Notably, neuronal activity can control both basal arterial diameter and evoked arterial dilation (81). Arterial SMCs, mural cells of the ACT zone and to a lesser extent thin strand pericytes are the effector cells involved in CBF increase (55, 82, 83).

There are multiple redundant pathways and molecules involved in linking neural activity to vessel dilation. Neurons can signal directly to blood vessels to initiate the vascular response. Previous studies have shown that the synaptic release of glutamate can activate N-methyl-D-aspartate receptors on neurons, resulting in the raise of Ca²⁺, causing either the activation of neuronal NO synthase to release NO or the activation of phospholipase A2 to produce cyclooxygenase-2-dependent prostaglandin E2 from arachidonic acid, that then act on SMCs to dilate artery (80, 84). Recently, a new and distinct paradigm has emerged in which endothelial cells of the vast capillary network within the brain are envisioned to act as a sensory web capable of detecting increases in neuronal activity and sending rapid signals that dilate upstream arteries (Figure 6). Specifically, extracellular potassium and NO are viewed as the most important neurovascular coupling mediators (78). In the first instance, endothelial cells sense neural activity-derived K⁺ through the inward-rectifier K⁺ channel, Kir2.1, which is activated by modest elevations in extracellular K⁺ (produced during each action potential), resulting in endothelial cell hyperpolarization. Because Kir2.1 channels are also activated by hyperpolarization, this hyperpolarizing (electrical) signal regenerates, rapidly propagating retrogradely from cell to cell through the capillary network via gap junctions, ultimately reaching upstream arterioles and pial arteries. There, the signal is passed to SMCs through myoendothelial projections, dilating arteries/arterioles and increasing blood flow to the site of signal initiation (85). This hyperpolarization may also be passed to the upstream ACT zone, potentially modulating the flow of blood within the capillary bed (53). Intimately linked with this electrical signaling mechanism is a highly localized Ca²⁺ signaling mechanism in endothelial cells, driven by neuronal activity, that leads to the generation and release of NO, causing dilation of the ACT zone. This neural activity-derived Ca²⁺ signaling mechanism may enable fine control of blood flow to small clusters of neurons (86). More distally located, thin-strand pericytes have also been reported to regulate capillary blood flow, but with slower kinetics than arteriolar SMCs and mural cells of the ACT zone (83). More recently, an elegant study using three-dimensional reconstruction of correlative light electron microscopy demonstrated that glutamatergic axons could form functional neural-SMC junctions in cortical penetrating arteries of mice and primates. Remarkably, using both ex vivo and in vivo experiments, the authors demonstrated that activation of a single glutamatergic axon could elicit the dilation of the contacted arteriole (87).

Figure 6. Schematic of neurovascular coupling (NVC) involving capillary-to-arteriole transduction.

NVC starts with the increase in local neural activity that leads to the hyperpolarization of capillary endothelial cells via the activation of Kir2.1 inward rectifier potassium channels in capillary endothelial cells. This hyperpolarizing signal is quickly propagated to upstream arterioles/ arteries, resulting in retrograde vasodilation (gray arrows). This can be associated with a local calcium signaling mechanism leading to the release of nitric oxide (NO), causing the dilation of the arteriole-capillary transition zone (purple arrows).

The importance of astrocytes in neurovascular coupling has remained a matter of debate. However, recent works suggest that astrocytes modulate rather than mediate CBF increase (88, 89). Astrocytes can respond to neural activity by a rise in the concentration of intracellular Ca²⁺ and the release of vasoactive molecules that can dilate arterioles (90). Ex vivo brain slices experiments and in vivo experiments from Mishra and colleagues suggest that capillary dilation evoked by neuronal activity is dependent on the rise of astrocytic calcium mediated by the activation of astrocytic purinergic (P2X1) receptors by neuron-derived ATP; this results in the synthesis of arachidonic acid and the production of prostaglandin E2 via cyclooxygenase 1 (COX-1) by astrocytes that then dilate capillaries by acting on E-type prostanoid receptor 4 expressed by pericytes (84); noting here that capillaries rather resembled to offshoots of penetrating arterioles and thus to ACT zones. A recent comprehensive in vivo study performed in fully awake mice provides evidence that astrocyte Ca²⁺ does not initiate or mediate neurovascular coupling when neuronal activation is short (< 5 seconds). Instead the authors discovered that the fundamental role of astrocyte Ca²⁺ is to amplify the late phase of arteriole dilation in the context of prolonged neuronal activation (91).

Besides the importance of capillary endothelial cells as sensors of neuronal activity, the Gu’s lab extended these findings to identify another essential role of arterial endothelial cells in neurovascular coupling (92). This lab first showed that caveolae, which are membrane invaginations, are abundant in endothelial cells of arteries, unlike capillary endothelial cells. Using cell-type specific genetic manipulations, they further demonstrated that the elimination of caveolae in arterial endothelial cells, but not in neighboring SMCs, strongly impaired neurovascular coupling (92). Mechanistically, it has been proposed that caveolae may cluster ions channels, such as Kir2.1, or receptors implicated in neurovascular coupling (92).

Neurovascular coupling remains a hot topic of research, and one of the key questions is what purpose does it serve. The common view, mentioned in every text book and paper, is that functional hyperemia is a mechanism for supplying the metabolic needs of active neurons, because of the high energy demands of the brain and its very limited ability to store energy (93). However, in some cases, the increase in CBF evoked by neuronal activity is higher than what is required, whereas in other cases neurovascular coupling seems absent or inverted (94). This has raised the possibility that it may serve additional purposes, such as removing metabolic wastes through a vascular route, homogenizing flow in the capillary network, preventing capillary stalls by leucocytes, regulating brain temperature, facilitating cerebrospinal fluid movement, or stabilizing the vascular network (94).

2.3.3. The blood brain barrier

The BBB is present throughout the arterio-venous axis, but capillaries are thought to provide the greatest surface area for exchange and transport. In actuality, the BBB consists of a series of barriers that includes from luminal to abluminal side: (1) a single layer of endothelial cells, (2) the endothelial basement membrane, (3) pericytes (embedded in the endothelial basement membrane) at the capillary level or SMCs (with their own basement membrane) at the artery or venous level, and (4) astrocytes with capillary-encasing endfeet, which collectively constitute the neurovascular unit (9) (Figure 7). A notable feature of the highly specialized endothelial cells that form the BBB is the glycocalyx – a negatively charged polysaccharide structure that projects into the lumen and acts as a barrier for the passive transport of large (>40 kDa) but not small (<1 kDa) molecules (95). Endothelial cells are sealed to each other by tight junctions, which prevent free paracellular transport of molecules; they also express an abundant repertoire of specific influx transporters (many of which are ATPases and ATP-binding pumps) and efflux pumps that drive the active transport of specific solutes and metabolites in or out of the brain, respectively (9). Brain endothelial cells, in contrast to most peripheral endothelial cells, are enriched for the lipid transporter MFSD2a, which actively inhibits the rate of caveolin-mediated transcellular trafficking (transcytosis) (96, 97). Several studies have demonstrated that, by inhibiting transcytosis, pericytes play an essential role in the development of the BBB (98–100). Pericytes also appear to be required for BBB maintenance, although a significant reduction in pericyte coverage (>50%) seems to be required to induce BBB leakage (101).

Figure 7. Schematic illustrating the cellular and molecular properties of the blood brain barrier (BBB).

The BBB is formed from the luminal to the abluminal side by a continuous layer of endothelial cells (EC), pericytes (at the level of capillary), basement membranes (BM) and the endfeet of astrocytes. EC exhibit unique barrier properties including: (1) the EC glycocalyx, which prevents the passive transport of large molecules, (2) the presence of tight junctions between EC, which prevents free paracellular transport of molecules, the expression of specific influx transporters (3) and efflux pumps (4), which drive the active transport of specific solutes and metabolites in or out of the brain, respectively, and (5) the expression of MFSD2a, which inhibits the rate of caveolae-mediated transcytosis by controlling the plasma membrane omega-3 fatty acid docosahexaenoic acid (DHA) lipid composition.

2.3.4. Brain waste drainage

The glymphatic (a contraction of glial-lymphatic) system, (re)discovered a decade ago by the Nedergaard group, is a unidirectional fluid-clearance pathway thought to primarily serve the function of non-selectively clearing metabolic waste from the brain interstitial space (102). This process starts with the flow of CSF along PVS surrounding arteries and consists of three main steps: (1) the influx of CSF into the brain across the endfeet processes of astrocytes, a process which is mediated by the perivascular polarization of aquaporin-4 water channels, (2) the subsequent exchange of CSF with the interstitial fluid (ISF); from there the CSF-ISF mix, carrying solutes and waste products, flows through the extracellular spaces between cells and (3) the efflux of fluids out the brain in the PVS surrounding the draining veins (Figure 8). The glymphatic fluids drain out of the cranium to the cervical lymph nodes and the peripheral blood circulation through multiple pathways: (i) along the olfactory nerves through the cribriform plate or along the other cranial nerves, (ii) into the lymphatic vessels of the dura mater; the lymphatic vessels located around the large venous sinuses and bridging veins coming from the brain that cross the arachnoid membrane and drain into the sinuses have been recently identified in humans and mice and (iii) possibly via the arachnoid granulations/villi, which are small invaginations of the arachnoid membrane protruding into or possibly around the dural venous sinuses (10, 37, 42, 103).

Figure 8. Schematic depiction of the glymphatic and intramural periarterial drainage systems.

The glymphatic system involves the CSF influx along the periarterial spaces and its entry into the brain supported by aquaporin 4 (AQP4) channels expression on the astrocytic endfeet (1), the subsequent mix with the ISF and flow through the extracellular spaces (2) and the efflux of fluid and wastes along perivenous spaces (3). The intramural periarterial drainage system drains ISF and solutes out of the brain along basement membranes of SMCs of arterioles and arteries towards the lymph nodes. As, astrocytes; Neu, neuron.

First characterized in rodents, the glymphatic system has been demonstrated to operate in human, as exemplified by a study employing serial T1-weighted MRI scans performed during the 24 hours after lumbar-level intrathecal injection of gadobutrol, a small MRI contrast agent, used as a CSF tracer. This study showed that glymphatic enhancement of gadobutrol propagation occurred much more slowly in humans than in rodents (104). By focusing on the early propagation of the tracer within the subarachnoid space, the same group provided evidence for the existence in human of a space around the major arteries, delineated by a semi-permeable membrane, that facilitates the anterograde transport of solutes inside the brain (39).

The glymphatic system is modulated by the sleep-wake cycle, being mostly active during sleep in both rodents and humans (105–107), although this concept has been recently challenged (108). Neurons have been recently identified as a key player in this system. Specifically, synchronization of neuronal activity, as observed during sleep, and consequently of neuronal pumps, generates large-amplitude, rhythmic ionic waves in the ISF that facilitates the flow of CSF-ISF fluid within the brain parenchyma and its clearance outside the brain (109). Heart beat-driven artery distensions (arterial pulsatility), that reflect changes in arterial diameter caused by cardiac impulse waves, are also key physiological drivers of CSF anterograde pumping in the periarterial spaces (110). Other drivers of the glymphatic system include changes in arterial diameter caused by autoregulation or neurovascular coupling, low-frequency arteriolar dilatations produced by contraction of SMCs (vasomotion), and the respiratory cycle through changes in venous pressure (10, 44). Recent work has further implicated PVMs as important regulators of CSF flow dynamics through their involvement in arterial motion and remodeling of the vascular ECM (66).

The intramural periarterial drainage (IPAD) system, has been proposed as an additional drainage pathway. The IPAD model posits that solutes and wastes flow within the basement membrane of capillaries and arterial SMCs, in the opposite direction to blood flow, towards the subarachnoid space and ultimately cervical lymph nodes. Accordingly, fluids and wastes move out the brain centrifugally (Figure 8) (111). It has been suggested that fluid transport is driven by arterial vasomotion, i.e. SMC-initiated spontaneous vasoconstrictions and -dilations that happen at low (~0.1 Hz) frequencies (112).

Both the glymphatic and the IPAD systems have stirred debates and remain an area of intensive research. With respect to the glymphatic system, the influx of CSF through the astrocyte endfeet in an aquaporin-4 dependent manner (113) is puzzling because it is difficult to conceive how a water selective channel can facilitate the influx or efflux of solute and tracers (114). Second, the maximal size of particles that can physiologically be cleared through this pathway is unclear. Third, it has been postulated that solutes and fluids move into the brain from arteries to veins by a combination of convection and diffusion (115). However, evidence for a pressure gradient across arterial PVS to venous PVS is lacking. Fourth, although bulk flow has been shown at the level of the brain surface and pial vessels (110), in vivo data on perivascular flow along penetrating vessels are lacking. Moreover, the exit routes of solutes and fluids are still incompletely understood. On the other hand, there are also potential caveats with the IPAD system. First, analysis of the distribution of CSF tracers in perfusion fixed tissues is poised with artefacts. Indeed, in perfusion fixed tissues with aldehydes, PVS shrink and injected tracers are artefactually displaced into the surrounding SMC layer and the basement membrane (40, 110). Second, mathematical modeling suggests that the resistance within the basement membrane is far too high to allow solutes flow within this compartment (116). Finally, the flow of solutes and wastes in the opposite direction of blood flow seems counterintuitive. The CSF-ISF “mixing hypothesis” has been proposed as an alternative model for solute exchange between the CSF and brain ISF (114, 117, 118). According to this model, there is no bulk flow along the penetrating vessels and CSF does not recirculate. Instead, wastes egress by diffusion from the ISF to the arterial PVS and then towards the pial surface, vasomotion and arterial pulsations facilitate CSF-ISF mixing in the PVS of penetrating arteries, thereby enhancing ISF–CSF exchange and the diffusion gradient for waste clearance is created by CSF bulk flow along the large pial arteries on the brain surface (114, 117, 118).

3. Historical perspective on cerebral small vessel disease

The fact that cSVD appears in the text books for about 30 years does not mean that cSVD is a new disease. Actually, pathological lesions and clinical manifestations typical of cSVD can be traced back to the middle of the 19^th century. At that time, there were neither CT nor MRI scans (CT was introduced in the mid-1970s (119), and MRI in the early 1980s (120)). Moreover, the concept of hypertension was not yet known. The first apparatus combining a sphygmomanometer and a stethoscope to measure non-invasively both systolic and diastolic BP was developed by Nikolaï Sergueïevitch Korotkov in 1905 (121) and the importance of measuring arterial BP was realized only in the 1930s (122). However, this was the golden age of anatomo-clinical correlative studies, that were the starting point in establishing correlations between the topography of stroke and the resulting clinical signs, and, in understanding the relationship between brain vascularization and brain lesions.

The history of cSVD has not been written in a linear manner but rather as the assembly of a puzzle, that we tried to briefly touch on in this section. Here, we have to pay tribute to insightful and visionary physicians, Maxime Durand-Fardel (1815–1899), Amédée Dechambre (1812–1896), Pierre Marie (1853–1940), Alois Alzheimer (1864–1915), Otto Binswanger (1852–1929) and Charles Miller Fisher (1913–2012), to name a few, who pioneered the field of cSVD. For those who are familiar with the recent literature on cSVD, the wealth and quality of these anatomo-clinical studies are really impressive.

3.1. On « état criblé », lacune, lacunar infarct, « état lacunaire » and the lacunar syndrome

The first description of “état criblé” of the brain is attributed to the French physician Maxime Durand-Fardel (1815–1899) that we can read in his paper entitled “Mémoire sur une altération particulière de la substance cérébrale” published in 1842 (123) and in his practical treatise on the Diseases of the Elderly published 12 years later (124). Durand-Fardel reported that, upon sectioning of the brain, the hemispheric WM was riddled with many small (pin-sized) rounded holes, with well-delineated edges, surrounded by normal appearing brain tissue, that he called “état criblé”. These holes always contained a vessel and Durand-Fardel interpreted them as an enlargement of the normal “channels” located around vessels, also called perivascular spaces. These holes could be also detected in the basal ganglia, where their size could reach a few millimeters. He hypothesized that this “état criblé” resulted from mechanical compression of the brain parenchyma by repeated dilation of vessels.

The use of the word “lacune” can be traced back to 1838 in a paper entitled “Mémoire sur la curabilité du ramollissement cerebral” from the French physician Amédée Dechambre (1812–1896) (125). Dechambre reported the clinicopathological observation of 10 patients who suffered from hemiplegia. He used the word “lacune” to designate small round cavities of variable size, without membrane, more or less filled with “milky fluid”, that he attributed to the natural scarring of small areas of brain softening (a former name for stroke). Five years later, Max Durand-Fardel, employed the word lacune in the third chapter of his book entitled “Traité du ramollissement du cerveau” (126). Herein, he reported the presence of several small lacunes in the striatum of a patient (case n° 78) aged 77 who presented with cognitive decline, gait disturbance and dysarthria for an undetermined period of time and died after sudden neurological deterioration. It is worth mentioning that Durand-Fardel made the distinction between these lacunes and the “état criblé” of the hemispheric WM, that were both present in his patient.

In 1901, the French Neurologist Pierre Marie (1853–1940) established the concept of lacunar infarct, that he clearly distinguished from the “état criblé” previously reported by Durand-Fardell. In his paper entitled “Des foyers lacunaires de désintégration” Marie provided a detailed macroscopic description of lacunes and associated clinical symptoms (127). Marie, then Head of a Neurology department at Hospital Bicêtre, reported the clinicopathological observation of 50 patients with hemiplegia, and he stressed that hemiplegia in the elderly was more often caused by cerebral lacunes than by cerebral hemorrhage or large cortical softening. He made in this seminal paper some masterful observations, which are summarized herein. Lacunes were predominantly located in the lenticulostriate nucleus, the thalamus (formerly called “optic layers”), the caudate nucleus and the pons; their number ranged from 1 to > 10, and their size from 2 to 20 mm (Figure 9). Lacunes corresponded to the softening of a small brain area, that further progressed to a small cavity surrounded by a sclerotic area. Small vessels within the basal ganglia and thalamus were permeable, but their wall was thickened and pale; similar vascular lesions could be also observed in larger arteries. He found that some lacunes containing patent vessels were caused by an enlargement of the perivascular space which destroyed the adjacent brain tissue, a process he called “vaginalite destructive”. Moreover, he reported that lacunes were typically associated with an atrophy of the WM, an état criblé, an atrophy of the anterior part of the brain, a dilation of ventricles and occasionally a thinning of the corpus callosum. Remarkably, Marie mentioned that lacunes were associated with the presence of deep ICH in about one third of patients and that ICH was in these cases likely the cause of the death (Figure 9). Clinical symptoms, as reported by Marie, included recurrent acute events (such as hemiplegia, dysarthria or gait disorder) that often had a good spontaneous resolution. These manifestations occurred at a mean age of ~60 years, and progressed towards the appearance of a pseudobulbar palsy (a syndrome characterized by slurred speech, dysarthria, dysphagia and emotional lability, which is caused by bilateral lesions of corticobulbar pyramidal tracts), a step-by-step walk (“marche à petits pas”), cognitive decline, dementia and premature death. Marie coined the name “état lacunaire” for this pathological condition and concluded that it was likely caused by “senility with arteriosclerosis”, i.e., an intrinsic disease of brain vessels favored by aging, in which vessels may become “clogged” to cause lacunes or may rupture to cause ICH. It is worth remembering that measurement of arterial BP was not routinely available at that time; indeed, the sphygmomanometer has been discovered only in the early 1900s. Additional clinicopathological studies were made by Ferrand (128) and Catola (129), two clinical fellows of Marie, who confirmed and expanded the study of their mentor. Hence, Durand-Fardel, Dechambre and Marie had identified two important pieces of the cSVD puzzle: (1) the concept of lacune/lacunar infarct/ “état lacunaire” and (2) the concept of enlarged perivascular space, according to the modern terminology of cSVD.

Figure 9. The concept of “Etat lacunaire” according to Pierre Marie in 1901.

Horizontal section of a right brain hemisphere showing the coexistence of lacunes, appearing as small black holes, in the lenticular nucleus and a thalamic intracerebral hemorrhage (dotted line). (1) frontal lobe, (2) caudate nucleus, (3) lenticular nucleus, (4) thalamus, (5) occipital horn. Reproduced from Marie (127) (public domain).

Interest for lacunes waned until 1965 when the Canadian Neurologist Charles Miller Fisher (1913–2012) performed a systematic analysis of post-mortem brains from 1,042 consecutive adults, in which he identified 376 lacunes in 114 individuals (130). The vast majority of brains were from individuals aged 50 and over. Fisher described lacunes as irregular trabeculated cavities surrounded by well-developed gliosis and he assumed that lacunes represented old healed ischemic infarcts. Like Marie, Fisher found that lacunes ranged in size from 0.5 to 17 mm, although the majority were smaller than 5 mm; they were exclusively located in deep brain regions, involving in descending order the lenticular nucleus, pons, thalamus, caudate nucleus and capsule-corona radiata region and were associated with atherosclerosis of cerebral vessels (Figure 10). Likewise, Fisher reported that lacunes were associated with deep ICH in ~35% of patients, and with superficial cortical or cerebellar infarcts in ~25% of patients. Contrary to Marie, Fisher did not find clear evidence of a prior stroke or neurological deficits in about two third of patients, suggesting that a number of these lacunes may have been clinically silent, although he acknowledged that the clinical records were probably incomplete. Importantly, Fisher reported for the first time a strong association between the presence of lacunes and a documented past history of chronic hypertension, an observation which has been subsequently confirmed by numerous studies. Here, in addition to confirm the concept of lacune, Fisher added three new pieces to the puzzle: (1) the presence of deep ICH, (2) the concept of “silent” cSVD and (3) the strong association with chronic hypertension.

Figure 10. Macroscopic characteristics of lacunes according to the study of 1,042 postmortem brains by Fisher in 1965.

A- Coronal brain section showing several lacunes in the lenticular nucleus (black arrow) and one lacune in the head of the caudate nucleus (white arrow). B- Horizontal section of the brainstem through the pons showing a lacune in the anterior part of the pons (black arrow). C-D Analysis of the location (C) and dimension (D) of 204 lacunes showing that lacunes are predominantly located in the lenticular nucleus and have a dimension comprised between 1 and 10 mm. The bar graphs are based on data from (130). A and B are from (130) and used with permission from Wolters Kluwer Health.

Thereafter, the terminology “lacunar infarct” has been used to designate a small infarct (< 15–20 mm) lying in the deep non cortical regions of the cerebrum or brainstem; the term implies that the infarct results from the occlusion of penetrating branches of cerebral arteries but not of large arteries. When causing symptoms, “lacunar syndrome” referred to the clinical manifestations associated with a lacunar infarct (131). As with many new ideas, the concept of lacunar infarct as a specific stroke entity related to an intrinsic small vessel disorder has been strongly challenged. Some claimed “the fallacy of the lacune hypothesis”, arguing that “a lacune is a stroke, just a small one” and “the word lacune should be used to mean a small stroke” (132). However, time has proven them wrong since lacune, lacunar infarcts or lacunar stroke are nowadays widely recognized as key features of cSVD.

3.2. On white matter rarefaction or atrophy, Binswanger’s disease, subcortical arteriosclerotic encephalopathy, leukoencephalopathy, leuko-araïosis and white matter hyperintensities

The first description of WM rarefaction can be traced back to 1854 in the Traité “Clinique et Pratique des Maladies des Vieillards » by Durand-Fardel who reported the examination of postmortem brains from elderly people, that he named “interstitial atrophy of the brain” (124). Specifically, Durand-Fardel reported foci of “rarefied white matter”, which could be unique or multiple, in the cerebral hemispheres of individuals without prior neurological manifestations and in the absence of any other brain lesions, although he noticed that similar lesions could be observed around genuine brain hemorrhages. He clearly differentiated these WM lesions from brain softening (infarct) and “état criblé”, but remained intrigued by their cause.

At the end of 19^th century and the beginning of 20^th century, the Swiss psychiatrist and pathologist Otto Binswanger (1852–1929), and the German neuropsychiatrist and neuropathologist Alois Alzheimer (1864–1915) reported several cases of WM atrophy associated with dementia; both inferred that this WM atrophy had a vascular origin (133–135). Although Alois Alzheimer is famously known for his seminal work on the dementia bearing his name, he also deserves credit for his work on vascular dementia. At that time, dementia was poorly defined, encompassing a heterogeneous group of neuropsychiatric disorders. The main challenges were to differentiate those disorders from neurosyphilis, that was then known as “General paralysis of the insane”, and to separate them into distinct entities. In 1894, Otto Binswanger reported clinico-pathological cases of dementia, which he believed were distinct from neurosyphilis, that he further classified in 2 distinct entities: (1) the Encephalitis Subcorticalis Chronica Progressa (ESCP) named after him “Binswanger’s disease” and (2) Arteriosclerotic Brain Degeneration (ABD). Pathological features of ESCP included severe WM atrophy, most pronounced in the temporo-occipital lobes, severe ventricle enlargement, especially posteriorly, contrasting with minimal cortical atrophy, that were associated with marked “atheromatosis” of the cerebral arteries (133, 134). Clinical features were characterized by a slowly progressive intellectual decline, although with fluctuating focal neurological deficits, with an age of onset ~50 years of age and the death of the patient ~10 years later. He proposed that WM atrophy resulted from a chronic “malnutrition” of the brain due to advanced arteriosclerosis. The second entity (ABD) was characterized by brain atrophy, with a marked reduction in the brain weight, a diffuse discoloration of the WM and the cortex, a slight thinning of the cortex, ventricle enlargement, an “état criblé”, most prominent in the basal ganglia and internal capsule, and the presence of “miliary apoplexies” (likely corresponding to multiple small infarcts). There was a widespread arteriosclerosis of the large and small brain vessels, including small cortical and WM arteries showing atrophic or fatty degenerative changes, with occasionally lumen narrowing. Arteriosclerosis also affected other organs, especially the heart and kidney. Clinical features of ABD included cognitive decline, which had a more remittent course than ESCP cases, and focal deficits. Between 1894 and 1902, Alois Alzheimer provided a separate description of ESCP and ABD, broadly consistent with Binswanger’s report, and included a microscopy description of cerebrovascular lesions. With respect to ESCP, which he mentioned to have seen 3 cases, Alzheimer emphasized the severe hemispheric WM atrophy, which spared the U-fibers (short association myelin fibers beneath the cortex which connect adjacent gyri of the brain), the absence of focal softening typical of infarctions and the severe arteriosclerosis of the deep WM vessels (135). In addition to ESCP and ABD, Alzheimer reported an additional form of dementia of vascular origin, characterized by large, wedge-shaped cortical infarcts that he attributed to large-vessel arteriosclerosis and severe narrowing of large cerebral arteries (135). Unfortunately, no illustration was available, neither from Binswanger nor from Alzheimer’s case reports. Through a modern lens, the different entities described by Binswanger and Alzheimer can be regarded as the phenotypic spectrum of modern vascular dementia, with at one end, subcortical ischemic vascular dementia with atrophy of the deep WM in the absence or presence of multiple deep infarcts, consistent with cSVD, and at the opposite end, multi-infarct dementia, with multiple large cortical infarcts. Hence, with their work, Alois Alzheimer and Otto Binswanger added three new major pieces to the puzzle: (1) presence of WM lesions, (2) association with dementia, and (3) a possible link with diseased brain vessels.

In 1962, Olszewski reported 2 novel cases and on the basis of a thorough review of the literature during the past 60 years, he challenged the “Binswanger’s disease entity” as defined by a pure and posteriorly-localized WM atrophy. Instead, he stressed the common presence of lacunes, cortical infarcts and diffuse area of demyelination (136). Yet, subcortical arteriosclerotic encephalopathy/Binswanger’s disease was still diagnosed only at postmortem examination (137). In 1979, Rosenberg and colleagues described the use of the CT scan in making the diagnosis of Binswanger’s disease antemortem, by identifying extensive WM lesions in a patient having chronic hypertension and progressive dementia and confirming the diagnosis by postmortem examination of the brain (138). In 1985, Kinkel and colleagues used the newly developed MRI to study the brains of 23 elderly patients (54–86 years, mean, 72 years), with CT findings of periventricular hypodensities (leukoencephalopathy) of unknown origin (139) (Figure 11). These patients had been referred for CT scan because of non-specific complaints, dementia or acute stroke. In the 1980s, MRI operated at a very low field strength of 0.15 Tesla, and thus at very low sensitivity, compared to 3T now in routine clinical practice, and 7T currently used in research. On MRI, the leukoencephalopathy appeared dark on T1-weighted imaged and bright on T2- weighted imaged, hence the name WM hyperintensities (WMH) in the modern literature. Lacunes were detected in 50% patients and enlarged ventricles in 30%. Clinical features were highly variable but it is remarkable that 35% had neither stroke nor dementia and that 78% were hypertensive. The leukoencephalopathy ranged from mild to severe, but was more severe in patients with stroke or dementia. Interestingly, postmortem examination of the brain of one deceased patient was consistent with a subcortical arteriosclerotic encephalopathy.

Figure 11. Antemortem detection of white matter lesions of different severity by MRI and CT scan in 1985.

MRI (0.15 Tesla) (top) and CT scan (bottom) from three patients with white matter lesions of mild (left), moderate (center) and severe (right) degrees. Reproduced from (139). Used with permission from American Medical Association.

With the increasing use of CT scan or MRI in clinical practice, the 1980s were marked by a surge in the number of patients with WMH on MRI or hypodensities on CT scan diagnosed with Binswanger’s disease. Considering this “epidemic” of Binswanger’s disease improbable and because no definite clinical deficit nor pathological changes had been linked with these WM neuroimaging changes, Hachinski and colleagues proposed in 1987 to use the term “leuko-araiosis” (the root “leuko” signifying “white” and the adjective “araiosis” meaning “rarefied” in Greek) to designate WM changes detected on CT scan or MRI (140).

In 1989, Charles Miller Fisher, the same who performed the seminal work on lacunes, produced an authoritative review on a series of ~50 pathologically confirmed cases with Binswanger’s disease (141). Fisher confirmed the selective involvement of the hemispheric WM, with multiple foci showing every degree of destruction ranging from spongy looseness to typical lacunes, whereas the U-fibers and the WM of the brainstem were usually spared. He found that the penetrating arteries of the WM were often thickened and hyalinized, although being rarely stenosed or occluded. Importantly, he noticed that lacunar infarcts and dilation of perivascular spaces (“état criblé”) were almost constant features. Moreover, he confirmed the strong association of clinical and pathological features of Binswanger’s disease with aging and hypertension, which was present in at least 87% of patients. In this review, Fisher challenged the prevalent idea at that time that Binswanger’s disease simply resulted from an arterioslerotic obstruction of the long medullary arteries of the WM, leading to distal territory ischemic damage in the WM and that the disease was only driven by hypertension (141). Here, the major contribution of Fisher has been to assemble all major pieces of the puzzle in a single entity leading to the modern definition of cSVD.

3.3. On familial forms of Binswanger’s disease, multi-infarct dementia or vascular leukoencephalopathy

Another major breakthrough in the field has been the realization that some forms of cSVD may have a genetic underpinning. By the late 1970s and 80s, unusual forms of Binswanger’s disease or vascular leukoencephalopathy, remarkable by a younger age of onset, a familial occurrence and no hypertension, had been described. Below is a brief description of the landmark cases that led to the identification of the most frequent monogenic cSVD and their causative genes. A detailed description of all monogenic cSVDs including their clinical manifestations, pathological features and genetic defects with extensive references is provided in section 5.4.1.

The first family, reported in 1976, included a Japanese sibling of 3 brothers and sisters who developed, in their third decade, progressive dementia associated with spastic gait, incontinence of urine and a pseudobulbar palsy syndrome, and leading to death in about 10 years (142). The other remarkable features were that these patients belonged to a consanguineous family, had normal BP throughout the course of the disease, and exhibited extraneurological manifestations including lumbago and alopecia. Neuropathological features were characterized by extensive WM lesions of the cerebral hemispheres, sparing the U-fibers, multiple foci of softening in the WM, basal ganglia, thalamus and pons and a severe arteriopathy affecting small leptomeningeal and intracerebral arteries, consisting in intimal proliferation, splitting of the internal elastic lamina, hyaline degeneration of the vessel wall, narrowing and occasional occlusion of the lumen (142).

One year later, Sourander and Walinder reported a Swedish family under the heading of “Hereditary Multi-Infarct Dementia” in which both males and females, from 3 successive generations, suffered from recurrent strokes, gait disturbances and progressive cognitive decline, starting in the 4^th decade and leading to premature death 10–15 years later (143). Brain pathology was characterized by multiple lacunes, most notably in the pons, and also in the WM and deep brain regions associated with widespread ischemic changes in the 3^rd layer of the cortex. Vascular changes were again most prominent in small leptomeningeal and intracerebral arteries and were characterized by hyaline degeneration of the vessel wall, with occasional fibrinoid necrosis, splitting of the internal elastic lamina, thickening of the vessel wall and narrowing of the lumen with vessel thrombosis and occlusion (144). It is worth mentioning that according to the new classification of vascular dementia (Table 3), “multi-infarct dementia” is a misnomer, since infarcts were small and located in deep brain regions, and the term would now be used to describe multiple large cortical infarcts. The same year, an English family was reported by Stevens and colleagues under the heading of “Familial vascular leukoencephalopathy” (145) and, a decade later, a Finnish family was reported by Sonninen and Savontaus under the heading of “Hereditary Multi-Infarct Dementia” (146). In all these three families, segregation of the disease was consistent with an autosomal dominant pattern of inheritance, clinical manifestations and pathological features were roughly comparable between the 3 families, except for a lower age-of-onset in the Swedish family (145, 146).

The 1990s marked a turning point with the identification of over 50 novel families sharing the same clinical and neuroimaging features, transmitted with an autosomal mode of inheritance (147–151). Clinical manifestations were characterized by recurrent strokes, usually starting in mid-adulthood, mood disorders, stepwise cognitive decline leading to dementia, and premature death at around 60–70 years of age. Brain MRI displayed in all affected individuals widespread WMH and often small deep infarcts (149, 152). Importantly, systematic brain MRI analysis of all family members, including asymptomatic ones, enabled Tournier-Lasserve and colleagues to make the landmark discovery that WMH were already present in asymptomatic individuals as early as 30 years of age (147). Pathological examination of post-mortem brain showed diffuse myelin loss, but sparing the U-fibers, multiple lacunar infarcts in the white and deep gray matter, dilated perivascular spaces, and a severe vasculopathy affecting small leptomeningeal and intracerebral arteries characterized by hyaline degeneration and fibrosis of the arterial wall, especially in long medullary arteries of the WM (153). In this report, Baudrimont and colleagues emphasized that vascular lesions were strikingly different from atheromatous lesions (153). Another landmark discovery in these families was the identification of a unique electron microscopy signature of the underlying vasculopathy, which consisted of the presence of granular osmiophilic deposits located within the basement membrane of vascular cells not only in brain vessels but also throughout small vessels of the body (153, 154). In 1993, the disease locus was mapped to chromosome 19, the acronym CADASIL, for Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy, was coined to designate this novel hereditary disease, and subsequently NOTCH3 was identified as the mutated gene in 1996 (155, 156).

It turned out that affected members of the English family reported in 1977 carried typical CADASIL-associated NOTCH3 mutation. Members of the Japanese family reported in 1976 proved to be affected by a distinct disease, now called CARASIL (Cerebral Autosomal Recessive Arteriopathy with Subcortical Infarcts and Leukoencephalopathy), in reference to clinical similarities with CADASIL but to a recessive mode of inheritance. In 2009, Hara and colleagues discovered that CARASIL was caused by homozygous mutations in the HtrA serine protease 1 (HTRA1) gene (157). The Swedish family was recently shown to be affected by the pontine autosomal dominant microangiopathy with leukoencephalopathy (PADMAL) syndrome caused by a dominant mutation in the gene encoding for the alpha 1 chain of collagen IV (COL4A1) (158).

3.4. On cerebral small vessel diseases nowadays

The term Binswanger’s disease is rarely used today, and cSVD is nowadays the broadly accepted term to designate what was formerly called “état lacunaire”, Binswanger’s disease, subcortical vascular leukoencephalopathy, and all other entities described above that have in common to be caused by in situ pathology of small leptomeningeal or brain vessels. Thanks to advances in neuroimaging, MRI criteria have overtaken neuropathological criteria for the diagnosis of cSVD, and the diagnosis can be made in alive patients and even in asymptomatic individuals (see section 4 below). In 2013, a group of experts published Standards for Reporting Vascular Changes on Neuroimaging (STRIVE-1) – an attempt to harmonize terminology and definitions of key MRI features associated with cSVDs (14). These neuroimaging consensus standards, which are now widely adopted in clinical and research settings, have been recently updated to include smaller lesions, such as cortical cerebral microinfarcts and incidental DWI-positive lesions (159), which are better detected using ultra-high-field MRI or diffusion-weighted imaging (DWI). It is now recognized that the full spectrum of cSVD manifestations ranges from brain lesions incidentally detected on brain CT scan or MRI, especially in individuals 50–55 years or older with no overt clinical symptoms – known as covert cSVD – to disability and dementia (160).

Many studies have amply demonstrated that aging and hypertension are the two major risk factors for cSVD, although genetic and environmental factors are increasingly recognized as important contributors (see section 5 below). Monogenic forms of cSVD are considered to account for about 2% of cSVD, and a dozen or so different genetic abnormalities have been identified (see section 5.4.1). The identification of genes mutated in cSVD have allowed the development of diagnostic tests that are crucial not only for early detection of the disease, but also for deciphering cSVDs and understanding the natural history of disease manifestations. Moreover, gene identification offered an unprecedented opportunity to generate predictive animal models. In addition to monogenic forms of cSVD, thanks to large-scale biomedical databases, international collaborative networks and the affordability of high-throughput genotyping and sequencing, common variants in more than 50 genes have been shown to increase the risk of sporadic cSVD which is believed to have a polygenic basis (see section 5.4.2). In addition to confirming the importance of hypertension as the primary modifiable risk factor, these studies have unraveled a remarkable continuum between sporadic and monogenic cSVD (section 5.4.3), hence strengthening the relevance of genetic cSVDs as models to study the pathogenesis of cSVDs.

4. Clinical and neuroimaging manifestations of cSVD

4.1. Clinical manifestations

The classical clinical feature of cSVD is lacunar stroke. This results from small deep infarcts occurring in the WM or deep grey matter nuclei. Whether such an infarct produces symptoms, and the type of symptoms produced, depends on its location and in particular whether it disrupts WM tracts underlying eloquent functions. There are over 20 lacunar syndromes but the most common are pure motor hemiparesis, sensorimotor stroke, pure sensory stroke, ataxic hemiparesis and dysarthria-clumsy hand syndrome (161). Pure motor stroke, most commonly due to an infarct in the posterior limb of the internal capsule, accounts for about a half of all lacunar strokes; a key distinguishing feature of it is that two or more body parts are affected; for example, face and arm, or arm and leg, or face arm and leg. Pure motor stroke affecting one body part such as the arm or hand, is more commonly due to small embolic infarcts in the cortex (162). In 15–20% of patients with lacunar stroke early neurological deterioration (END) occurs over the first few hours or days, and is associated with worse outcome (163, 164). A recent meta-analysis found female sex, hypertension, diabetes, and smoking, were associated with END (165). A related variant is the capsular warning syndrome, first described by Donnan in 1993, as a pattern of transient ischemic attacks characterized by recurrent episodes of unilateral transient motor and/or sensory symptoms affecting two or more regions (face and upper and lower limbs) without cortical signs (aphasia, apraxia, and agnosia) within 24 h (166). In some cases this progressed to cerebral infarction, and when this occurred it was usually a lacunar infarct and involved a single penetrating vessel (166).

Cognitive impairment is a second key feature of cSVD (167). It is a most common cause of vascular dementia, but lesser degrees of cognitive impairment which can nevertheless be disabling are even more common (168). A characteristic pattern of cognitive impairment is found with prominent involvement of executive functions and slowing of processing speed (22, 169). Impairment of working memory is common early in the disease, but episodic memory involvement is less frequent although it can occur later in the disease. This pattern of cognitive involvement is sometimes referred to as a “subcortical” cognitive syndrome, and is similar to that seen in diseases affecting the frontal lobes, where it is called the “frontal lobe” syndrome. A consequence of this is that screening tools designed to detect Alzheimer’s disease, which rely on disruption of episodic memory and orientation, are less sensitive to vascular cognitive impairment due to cSVD particularly in its earlier stages (170). Its detection requires screening tools sensitive to executive function and impaired processing speed, and a number of these have been devised (170, 171). Using these it has been shown that some degree of cognitive impairment may be present in as many as 40% of individuals with lacunar stroke (167). In the Secondary Prevention of Small Subcortical Strokes (SPS3) trial, in the 1,636 participants in whom cognitive testing was available, 47% were classified as having mild cognitive impairment and in many of these there was only minimal or no physical disability. The SPS3 investigators concluded that cognitive dysfunction in lacunar stroke patients may commonly be overlooked in clinical practice, but may be as important as motor and sensory sequelae, being more prevalent than physical disability defined by Rankin score ≥ 2 (33%), and present in 41% of patients with no significant disability. In addition to the typical “subcortical profile” found in cSVD they found many cases of amnestic mild cognitive impairment raising the possibility of co-existing Alzheimer’s pathology contributing to the deficit (172).

There is increasing interest in other non-motor symptoms of cSVD. An increased incidence of both depression (173), and apathy (a lack of goal-directed activity and motivation compared to previous behavior) (174), have been described in cSVD. Apathy can be a particularly disabling symptom particularly for other family members and carers (175) and is not helped by conventional antidepressants (176). Other clinical features include gait disturbance (177) and urinary incontinence (178).

4.2. Neuroimaging

The cardinal MRI features of SVD are lacunar infarcts, lacunes, WMH, cerebral microbleeds (CMB) (Figures 12 and 13), enlarged perivascular spaces (PVS) (Figure 14), and cerebral atrophy (14, 159). Established lacunar infarcts appear as low-density on T1 or FLAIR MRI sequences, reflecting tissue loss, and are usually defined as being less than 15 to 20 mm maximum diameter to distinguish them from infarcts with a likely embolic cause. In the acute setting they are best seen on diffusion weighted imaging (DWI) sequences where they appear as high signal on DWI (Figure 12A), or low signal on apparent diffusion coefficient (ADC) sequences. DWI positive lesions are visible after the acute event within hours, and can stay DWI positive for a few weeks. Longitudinal studies have shown that not all DWI positive lesions progress to classical established lacunar infarcts (179). While some do, others may progress to WMH, while others are no longer visible on later MRI. MRI studies showed that 50–80% of acute small deep infarcts cavitate to evolve into lacunes (180–182). Lacunes of presumed “vascular origin” are defined on MRI as a round or ovoid, subcortical, fluid-filled cavity, of between 3mm and about 15mm in diameter, consistent with a previous acute small deep brain infarct or hemorrhage in the territory of a perforating arteriole (Figure 1A) (14). WMH are best seen as high signal lesions in the periventricular and deep WM regions on T2-weighted MRI (Figure 12B) or FLAIR (Figure 12C) (159). They correspond to low-density lesions on CT, but are seen much more clearly on MRI. CMB are seen as areas of signal dropout (low-density) on gradient echo or susceptibility weighted imaging (SWI) sequences (Figure 12D). They correspond to small areas of hemosiderin, presumably deposited following leakage of red cells from the small vessels. Due to a blooming effect, the size of the hemosiderin deposits is considerably smaller than the CMB visible on MRI. Their distribution can be useful in distinguishing CAA from cSVD primarily due to hypertension. In the former they occur predominantly in the cortical region, whereas in the latter they occur predominantly in the subcortical regions (Figure 13) (183). Enlarged PVS are another common feature in cSVD (Figure 14).

Figure 12. Features of cSVD on MRI in a patient presenting with lacunar stroke.

A- Acute lacunar infarct visible as high signal on diffusion weighted imaging (MRI). B- Multiple white matter hyperintensities (WMH) are visible on T2-weighted imaging. C- The WMH are more clearly visible on FLAIR- a T2 sequence in which the free water is suppressed (note signal from the cerebrospinal fluid in the ventricles is now low signal not high signal). D- Cerebral microbleeds visible as hypodense foci in the subcortical structures on susceptibility weighted MRI - one arrowed in the right thalamus. Reproduced from (16). Used with permission under CC-BY 4.0 license.

Figure 13. Microbleeds and intracerebral hemorrhage on susceptibility weighted MRI in a patient with no risk factor.

Multiple cerebral microbleeds (arrows) are visible in the subcortical regions. An old intracerebral hemorrhage (arrowhead) can be seen in the left basal ganglia as a triangular hypodense lesion. Reproduced from (doi: 10.1177/17474930241279888). Used with permission under CC-BY 4.0 license.

Figure 14. Enlarged perivascular spaces (PVS) visible on T2-weighted MRI.

On the axial image multiple high signal lesions due to PVS can be seen as circular lesions if cut in cross section (A), or longitudinal lesions if cut along their length (B). They can be better seen on the closes up as linear structures passing from the deep white matter to the cortex (C). Reproduced from (doi: 10.1177/17474930241279888). Used with permission under CC-BY 4.0 license.

Brain atrophy is also seen in cSVD, particularly in the later stages. It has been shown this can occur secondary to lacunar infarcts with retrograde degeneration along WM tracks (184), and is associated with WM tract damage (185). More recently it has been suggested that cerebral microinfarcts may also contribute to atrophy. It was thought that cSVD was an exclusively subcortical disease, but recent studies have shown that cerebral microinfarcts are a common feature. These are small infarcts that were initially only seen using the high-resolution of 7 MRI, as well as being visualized neuropathologically, but have recently been shown to be detectable on 3T MRI (186).

5. Main determinants of cSVD

5.1. Aging

Age is the strongest risk factor for cSVD. The incidence of lacunar stroke rises exponentially with age (187), while age accounts for 16% of the variance in WMH (188). This association persists after controlling for cardiovascular risk factors, the prevalence of which also rises with age. Similar exponential rises with age are seen for CMB, and enlarged PVS. WMH are uncommon before age 30, but their prevalence rises rapidly from middle age, and they are detectable in up to 90% of individuals by age 65 (25, 189). This raises the question as to whether some degree of cSVD should be seen as a ubiquitous feature of ageing, with its severity influenced by a variety of modifiable and non-modifiable risk factors.

5.2. Sex differences

A recent systematic review and meta-analysis of 123 studies including 36,910 participants examined sex differences in cSVD (190). It found that cSVD was more common, and more severe, in males and this difference was particularly marked in hospital-based compared to community-based studies. This difference was also observed in more severe cSVD, particularly cases presenting with stroke. The pattern was consistent between recent (2015–2020) and older (1989–2014) studies, and across different world regions. The authors suggest that the more severe cSVD in males, could indicate differences in risk factor exposures, susceptibility to cSVD, or adherence to risk factor interventions. However methodological issues including differences in study recruitment could also contribute to these differences. For example, in stroke studies females are usually older and have more disability at stroke onset (191). Furthermore, women with stroke may present with non-traditional symptoms like altered mental status which could be overlooked or misdiagnosed (192), and are important since atypical or neuropsychiatric symptoms are increasingly recognized to associate with cSVD. A more recent study of MRI cSVD burden in over 40,000 general population individuals found sex accounts for only 0.6% of the variance in WMH lesion volume (188), making it a less important risk factor for cSVD compared to age and cardiovascular risk factors.

5.3. Modifiable risk factors

Numerous studies have looked at the relationship of common modifiable vascular risks factors to lacunar stroke risk, both in comparison to stroke free controls and to other stroke subtypes. Interpretation of these is complicated because stroke subtyping has often been performed based on diagnosis of a clinical lacunar syndrome, combined with CT brain imaging. The later will often not show lacunar infarcts, particularly in the first 24 hours. Validation studies have shown that when more rigorous subtyping is performed as many as half of all patients classified as lacunar stroke in this way, do not have cSVD (193). There are many fewer studies where MRI, which allows more accurate diagnosis of lacunar infarction, was performed. An alternative approach is to study risk factor associations with MRI markers of cSVD in large cohorts of individuals, often population based. This allows accurate phenotyping of the features of cSVD but provides information on the more chronic features of cSVD, rather than risk factors for acute infarction.

A complementary approach, particularly to investigate whether risk factors are causally related to cSVD, is Mendelian Randomization. Mendelian randomization (MR) is an analytical method that uses genetic variants as instrumental variables for risk factors (194). It is increasingly used because it can overcome a major limitation of evidence from observational studies namely unmeasured confounding. Exposures can be any factor robustly associated with genetic variation in individuals. For example, it has been suggested that C-reactive protein (CRP), a marker of inflammation, is causally related to cardiovascular disease. However, an alternative interpretation is that the cardiovascular disease itself is causing inflammation, which therefore results in a rise in CRP. By selecting a genetic variant associated with increased CRP levels, which we can assume has been exposing the individual to higher CRP levels throughout their life, we can overcome this confounding. In this example it failed to confirm a causal relationship (195). Two sample MR using two different study samples to estimate the instrument-risk factor and instrument-outcome associations to estimate a causal effect of the risk factor on the outcome, has been most frequently used in studies in cSVD. This can be useful when the risk factor or outcome, or both, are expensive to measure and provides an opportunity to substantially increase the statistical power, by incorporating data from multiple sources, including large consortia (194).

A meta-analysis of data from 12 articles involving 6944 lacunes found that hypertension (odds ratio [OR] 3.16, 95% confidence interval [CI] 2.22–4.49), diabetes (OR 2.15, 95% CI 1.59–2.90), hyperlipidemia (OR 1.64, 95% CI 1.11–2.40), and smoking (OR 1.47, 95% CI 1.15–1.89) were significantly related to the risk of lacunes (196). MR studies support the associations with hypertension, smoking, and diabetes but failed to replicate any association with LDL cholesterol (197). Results from a large MR study using data from a well phenotyped lacunar stroke cohort are presented in Figure 13. The lack of association with LDL cholesterol has been replicated in a number of studies including those investigating MRI markers of cSVD. Interestingly in MR studies genetic predisposition to a higher HDL cholesterol level was associated with lower risk of lacunar stroke and lower WMH volume. HDL-C raising genetic variants in the gene locus of the target of Cholesteryl ester transfer protein (CETP) inhibitors were associated with lower risk of small vessel stroke and lower WMH volume, but a higher risk of ICH, suggesting that HDL-C raising strategies could be considered for the prevention of ischemic cSVD, although the net benefit of such an approach would need to be tested in a randomized controlled trial (198).

Risk factor profiles have also been compared between lacunar strokes and other stroke subtypes. Jackson and colleagues pooled individual data on 2875 patients with first-ever ischemic stroke from 5 prospective stroke registers. They found a lower prevalence of cardioembolic source (adjusted OR, 0.33; 95% CI, 0.24 to 0.46), ipsilateral carotid stenosis (OR, 0.21; 95% CI, 0.14 to 0.30), and ischemic heart disease (OR, 0.75; 95% CI, 0.58 to 0.97) in lacunar compared with non-lacunar patients, but no difference for hypertension, diabetes, or any other risk factor studied (199). However, only about 25% of cases had MRI, although an analysis confined to patients with a visible relevant infarct on brain imaging produced similar results. In contrast, other studies have reported hypertension is a stronger risk factor for lacunar stroke (200). One potential confounder is whether all patients with lacunar infarcts are included, or only those with the lacunar stroke subtype based on pathophysiological subtyping, such as the TOAST classification (18). The latter approach will classify patients with cardioembolic sources as cardioembolic stroke, and those with large artery stenosis as large artery disease, even if they have a lacunar infarct. Using this latter approach, in a well phenotyped group of patients of which many subtyped using MRI, hypertension was commoner in lacunar stroke than large artery stroke (OR 3.43 (2.32 to 5.07); p<0.001) whereas hypercholesterolemia (OR 0.34 (0.24 to 0.48); p<0.001), smoking (OR 0.63 (0.44 to 0.91); p = 0.012), myocardial infarction (OR 0.35 (0.20 to 0.59); p<0.001) and peripheral vascular disease (OR 0.32 (0.20 to 0.50); p<0.001) were commoner in large artery stroke (201). This would suggest that atherosclerotic risk factors are less important for lacunar stroke than for large artery stroke, while hypertension is more important.

Based on detailed pathological studies by Fisher, it has been suggested that there may be two types of cSVD (202), and that these can be differentiated on brain imaging (203). The first involves atheroma at the origins or proximal portions of the larger (200–800 μm diameter) perforating arteries. This is associated with single or a few larger lacunar infarcts without leukoaraïosis on CT or WMH on MRI (isolated lacunar stroke subtypes). The second involves a diffuse arteriopathy of the smaller perforating arteries, 40–200 μm in diameter, resulting in multiple smaller lacunar infarcts with leukoaraïosis on CT or confluent WMH on MRI (leukoaraïosis subtype). Lacunar stroke patients classified in this way were found to have differing risk factor profiles. Hypercholesterolemia (OR 0.45 (0.28 to 0.74); p = 0.002), diabetes (OR 0.42 (0.21 to 0.84); p = 0.014) and myocardial infarction (OR 0.18 (0.06 to 0.52); p = 0.001) were more common with the isolated lacunar stroke subtype. In contrast age (OR 1.11 (1.09 to 1.14); p<0.001) and hypertension (OR 3.32 (1.56 to 7.07); p = 0.002) were more common with the leukoaraïosis subtype (201).

Many studies investigating risk factors for MRI markers of cSVD have been published and recently large databases such UK Biobank have provided sufficient power to obtain robust estimates. The largest study to date is from UK Biobank and analyzed data from 41,626 individuals of mean age was 55 years, 52.8% of whom were female. Common cardiovascular risk factors combined accounted for ≈15% of the variation in WMH volume, with age accounting for an additional 16%. However, a large portion of the variance, and well over 60%, remains unexplained. The contribution of individual risk factors to WMH risk is shown in Table 4 (188). The strongest risk factors were blood pressure and hypertension, with the blood pressure parameters together accounting for ≈10.5% of total variance in WMH. As age increased, the amount of shared variance between most of the individual risk factors and WMH volume decreased. This is consistent with previous data indicating that the contribution of vascular risk factors to cerebrovascular disease and dementia risk declines with age (204), emphasizing the need to target modifiable risk factors in midlife. Early life factors including intellectual quotient, years of education and socioeconomic status are also predictors of WMH later in life (205). The importance of hypertension is consistent with the results of other smaller population based studies (206). Longitudinal studies have also established a relationship between hypertension and progression of WMH over time (207). MR studies support the role of blood pressure, smoking, type 2 diabetes, waist/hip ratio, body mass index and alcohol use as causal risk factors for WMH, but found no association with total cholesterol or LDL (208).

Table 4.

The amount of variance in white matter hyperintensity (WMH) volume explained by different cardiovascular risk factors.

Risk factor	WMH volume
	Correlation coefficient	% of variance
Hypertension	0.21	4.4
Systolic blood pressure	0.21	4.4
Diastolic blood pressure	0.13	1.7
Diabetes	0.08	0.6
HbA1c	0.12	1.4
Waist circumference	0.14	2.0
Body mass index	0.09	0.8
Hyperlipidaemia	0.07	0.5
Triglycerides	0.07	0.5
Age	0.37	13.7
Sex	0.08	0.6

The percentage of shared variance, calculated by multiplying the squared correlation coefficients by 100.

HbA1c, glycated hemoglobin. Adapted from (188)

The percentage of shared variance, calculated by multiplying the squared correlation coefficients by 100. HbA1c, glycated hemoglobin. The table is based on data from (188).

5.4. Genetic determinants

There has been increasing recent interest in the role of genetic risk factors in cSVD. Traditionally these have been classified as single-gene disorders causing familial (monogenic) cSVD, or as genetic variants which increase the risk of sporadic disease, often via interacting with conventional risk factors (polygenic/multifactorial cSVD). In this section, we first review the current knowledge on monogenic cSVD with an emphasis on the three most common genes (NOTCH3, HTRA1 and COL4A1/A2 genes) that are involved. We then provide an overview on genetic variants associated with cSVD-related lacunar stroke or MRI features. We then discuss the remarkable continuum between monogenic and multifactorial cSVD.

5.4.1. Monogenic cSVD

Inherited forms of cSVD are thought to account for about 2% of the total cSVD burden (209). Red flags of a possible hereditary cSVD include a family history of stroke or dementia, an early onset of symptoms before the age of 60, characteristic MRI appearances, presence of extra neurological manifestations and a negative workup for other causes of disease manifestations (210, 211). To date, 15 distinct Mendelian cSVDs, caused by mutations in 13 distinct genes, have been identified although other uncharacterized hereditary cSVDs are likely to remain. The main clinical, neuroimaging, pathological and molecular features of these monogenic cSVDs are summarized in Tables 5–7 respectively. Notably, mutations in NOTCH3, HTRA1 or COL4A1/COL4A2 account for the majority of the burden of adult-onset monogenic cSVDs (212) and will be further discussed below.

Table 5:

Clinical, neuroimaging and molecular features of molecularly defined monogenic cSVDs

Disease	Mutated gene (year of identification)	Mode of inheritance	Frequency	Typical mutations and consequences	Main clinical and neuroimaging features	Extracerebral features
CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy)	NOTCH3 (1996)(156)	AD	Minimal prevalence: 2–6/100 000 (259, 260) > 1,000 pedigrees Most common cause of inherited cSVD (up to 80% of diagnosed genetic cSVD)(238)	Heterozygous missense mutations leading to an odd number of Cysteine residues in one of the 34 EGFr of the extracellular domain of NOTCH3, resulting in the abnormal accumulation of Notch3ECD in granular osmiophilic material (GOM) (151, 215)	- Age at onset: usually 50–60 y (range 30–80 y) (migraine with aura occurs earlier) - Classical clinical features of a progressive and diffuse ischemic cSVD with recurrent ischemic stroke or transient ischemic attack, mood disturbances, cognitive decline ranging from executive dysfunction up to dementia, motor disability and gait disturbance (213) - Attacks of migraine with aura (50–75 %) (261, 262) - Possible spontaneous ICH (~2% in caucasian and 17% in Asian) (263) - Classical features of a progressive and diffuse ischemic cSVD with symmetrical and extensive WMH, lacunes, MCB, enlarged PVS and cerebral atrophy.	Accumulation of Notch3ECD and Granular Osmiophilic Material (GOM) in peripheral vessels (154, 264)
DADA2 (deficiency of adenosine deaminase 2 syndrome)	ADA2 (2014) (265)	AR	> 250 reported pedigrees (266)	Biallelic missense mutations leading to ADA2 loss of activity (265)	- Mean age at onset: 5–7 y (0–60 y) - Ischemic strokes and subtle neurocognitive disorders, possible hemorrhagic strokes - MRI features: lacunar infarcts in deep brain region, rare WMH (267, 268) - Symptoms can be significantly improved using Tumor necrosis factor-α inhibitors or hematopoietic stem cell transplantation (269).	- Fever and rashes - Systemic vasculopathy (livedo reticularis, polyarteritis nodosa, gastrointestinal, kidney and liver manifestations) - Hematological manifestations and mild immunodeficiency (268)
Hereditary SVD with osteoporotic feature	ARHGEF15 (2023) (270)	AD	4 reported pedigrees (270)	Heterozygous missense or nonsense mutations leading to reduced ARHGEF15 activity	- Age of onset: 40–50 y - Classical clinical and MRI features of a progressive and diffuse ischemic cSVD (see above CADASIL) (270)	Early onset severe osteoporosis and bone fractures (270)
COL4A1/2 hemorrhagic microangiopathy	COL4A1 (2005) (254, 256) COL4A2 (2012) (271)	AD	> 250 reported pedigrees (266)	Missense heterozygous mutations affecting highly conserved glycine residues in the Gly-X-Y motifs of the triple-helical domain and rare heterozygous nonsense or frameshift mutations leading to misfolding and reduced secretion of COL4A1/2 heterotrimer (249, 250)	- Age of onset highly variable from prenatal period to adulthood - Extreme phenotypic variability of clinical features within members of the same family - Spontaneous ICH - Ischemic stroke possible (< 20%) - Seizures (255) - MRI features: WMH, deep micro or macrobleeds, possible porencephalic cavities, schizencephaly or polymicrogyria (255) - Intracranial cerebral berry aneurysms - Can present with spontaneous miscarriage and neonatal abnormalities including porencephaly probably secondary to ICH occurring in utero or at childhood.	Inconstant, and mostly reported with COL4A1 mutations (HANAC syndrome) (272) - Eyes: tortuosity of retinal arteries and retinal hemorrhages, congenital or early-onset cataract or glaucoma, developmental defects of the anterior segment - Kidney: cysts, proteinuria, hematuria, or moderate renal failure - Muscle cramps
PADMAL (Pontine Autosomal Dominant Microangiopathy with Leukoencephalopathy)	COL4A1 (2016) (158, 257)	AD	< 10 reported pedigrees (266)	Heterozygous mutations located in the 3′UTR of COL4A1 leading to the upregulation of COL4A1/2 (257)	- Age at onset between 30 and 45 y - Classical features of a progressive ischemic cSVD (see above CADASIL) but with early gait instability and motor difficulties - Classical features of a progressive and diffuse ischemic cSVD, but with multiple lacunes in the pons (158, 257)	None
CARASAL (Cathepsin-A-related arteriopathy with Strokes and Leukoencephalopathy)	CTSA (2016) (273)	AD	< 10 reported pedigrees (266)	Heterozygous mutation possibly leading to CTSA accumulation (273)	- Age at onset between 30 and 50 y - Possible severe hypertension (273) - Classical clinical and MRI features of a progressive ischemic cSVD (see above CADASIL) (273, 274)	Possible dry eyes and mouth, muscle cramps (273)
Fabry disease (Anderson-Fabry disease)	GLA (alpha- galactosidase) (1970) (275)	X-linked	Prevalence of 1 in 5,500–200,000 (UK biobank) (276)	Missense or non-sense mutations or deletions/insertions in GLA resulting in deficient α-galactosidase A enzyme activity causing pathological accumulation of glycosphingolipid in lysosomes of various cells	- Age of onset: ~40 y - Increased risk (x 5 to 12) of ischemic stroke of large and small vessel subtypes - WMH and possible lacunar infarcts - Tortuosity and dilation of large intracranial arteries (277, 278)	- Age of onset: 5–6 y (males); ~9y (females) - Childhood neuropathic pain and hypohydrosis - Gastrointestinal manifestations - Angiokeratoma - Corneal dystrophy, hearing loss - Hypertension - Renal and cardiac dysfunction (279) that can be ameliorated by enzyme replacement therapy
CARASIL (Cerebral Autosomal Reccessive Arteriopathy with Subcortical Infarcts and Leukoencephalopathy)	HTRA1 (2009) (157)	AR	< 50 reported pedigrees (266)	Biallelic missense mutations involving the serine protease domain or nonsense, splice-site or small indels leading to a loss of HTRA1 protease activity (157)	- Age at onset before 30–40 y. - Classical clinical and MRI features of a progressive and diffuse ischemic cSVD (see above CADASIL) (280)	Frequent - Premature alopecia - Premature lumbar or cervical spondylosis with disk herniation (<40 y) (280)
HTRA1-related autosomal dominant cSVD	HTRA1 (2015) (237)	AD	< 100 reported pedigrees but many unreported; second most common form of cSVD after CADASIL (266)	Heterozygous missense, nonsense or frameshift mutations leading to reduced HTRA1 protease activity (237, 238, 245)	- Age of onset >50 y - Classical clinical and MRI features of progressive ischemic cSVD (see above CADASIL) (246)	Rare - Late onset disc arthrosis or spondylosis - Premature alopecia (mostly reported in Asian patients)
Early amnestic syndrome of the hippocampal type and leukoencephalopathy	LAMB1 (Laminin subunit beta 1) (2021) (281)	AD	6 reported pedigrees (281,282)	Heterozygous insertion, deletion, duplication or splice site mutation in LAMB1 resulting in end-truncated LAMB1	- Age of onset : ~ 60 y - Progressive amnestic syndrome of the hippocampal type not associated with hippocampal atrophy, with mild executive dysfunction - Acute episodes of unspecified neurologic symptoms - Diffuse WMH already detectable at 30 y. without lacunar infarcts (281, 282)	None
NIT-1 related cSVD	NIT1 (nitrilase 1) (2024) (283)	AR	Estimated prevalence ~1 in 450,000 6 reported pedigrees (283, 284)	Bi-allelic missense or truncating variants in NIT1 leading to a loss of function of NIT1 (283, 284)	- Age of onset : ~40–50 y. - Progressive movement disorders, gait disturbances, dysarthria and cognitive decline - Ischemic stroke, fatal deep ICH - Massively enlarged PVS - Variable burden of WMH and lacunes (283)
NOTCH1-related microangiopathy	NOTCH1 (2022) (285, 286)	AD	8 reported probands (285, 286)	De novo heterozygous missense mutations in the heterodimerization domain of NOTCH1 predicted to increase NOTCH1 receptor activity (285)	- Age of onset: neonatal to childhood - Slowly progressive motor problems leading to wheelchair dependency, developmental delay, mood disorders, cognitive deficits - CT and MRI features: WMH, numerous deep brain vascular calcifications and mild brain atrophy (285, 286)	None
Early onset autosomal recessive NOTCH3-related cSVD	NOTCH3 (2015)	AR	< 10 reported pedigrees	Biallelic nonsense or frameshift mutations predicted to abrogate NOTCH3 expression and signaling (233–235)	- Early onset: neonatal to childhood, - Severe phenotype with clinical and MRI features of progressive ischemic cSVD (see above) (233–235) - Possible arterial aneurysms (234)	- Livedo reticularis (50% patients) (233, 234)
Leukoencephalopathy with calcifications and cysts (LCC) (Labrune syndrome)	SNORD118 (2016) (287)	AR	~100 reported pedigrees (288, 289)	Biallelic rare variants predicted to be null or hypomorphic mutations resulting in strongly reduced SNORD118 activity (289)	- Age of onset: highly variable (1 month – 71 y) - Seizures, developmental delay, progressive neurological deficits, spontaneous ICH, cognitive impairment, mood disturbance, intracranial hypertension (288–290) - CT and MRI features: asymmetric cerebral or cerebellar WMH, multifocal cerebral or cerebellar calcifications and progressive formation of variable size brain cysts that can simulate a tumor.	None
RVCL-S (Retinal vasculopathy with Cerebral Leukoencephalopathy and Systemic manifestations)	TREX1 (2007) (291)	AD	< 50 reported pedigrees (266)	Heterozygous frameshift or nonsense mutations leading to a C-terminus truncated TREX1 protein (292, 293)	- Age at onset between 30 and 50 y - Acute and subacute focal neurological deficits, migraine, psychiatric symptoms, cognitive decline leading to dementia - CT & MRI features: WMH, “pseudo-tumor” lesions, calcifications (292, 293)	- Raynaud’s syndrome - Retinopathy leading to blindness - Kidney and liver dysfunction (292, 293)

Table 7:

Main characteristic of monogenic cSVD genes

Gene	Protein product	Expression pattern in the central nervous system	Physiological role in the central nervous system
NOTCH3	NOTCH3 (2321 aa), single-pass transmembrane receptor. Belongs to the highly conserved NOTCH receptor family. Synthesized as a single-chain precursor, cleaved into the transgolgi network as it traffics to the plasma membrane where it is expressed as a heterodimer composed of a large extracellular domain (Notch3ECD) and a transmembrane-intracellular domain (Notch3TMIC) (215, 309)	Strongly restricted to vascular SMCs and cerebroretinal pericytes (54, 215)	- Required for small artery integrity in the brain and retina (310, 311) - Involved in arterial differentiation and maturation of SMCs (310)
ADA2	ADA2 (Adenosine deaminase 2) (511aa). Dimeric and secreted enzyme that, similarly to its ADA1 isoform, catalyzes the deamination of adenosine and 2’- deoxyadenosine into inosine and deoxyinosine respectively, but affinity of ADA2 for adenosine is lower than that of ADA1 (312)	Highly expressed in immune cells, particularly in myeloid cells (312)	May regulate monocyte phenotype, neutrophil extracellular trap formation, and modulate innate immunity (312)
ARHGEF15	ARHGEF15 (Rho guanine nucleotide exchange factor 15) (841 aa). Catalyzes the exchange of bound GDP for GTP, leading to the formation of active GTP-bound Rho GTPases	Endothelial cells (313, 314) (http://betsholtzlab.org/VascularSingleCells/database.html)	Rho guanine nucleotide exchange factor involved in the regulation of actin cytoskeleton (270). Possibly involved in vessel stabilization during development (313, 314).
COL4A1/A2	COL4A1 (1669), Collagen alpha-1(IV) chain; COL4A2 (1712 aa), Collagen alpha-2(IV) chain. Each chain contains three major structural domains: the 7S, the triple helical (collagenous) and the globular non collagenous 1 (NC1) domains. Two α1 and one α2 chains assemble into heterotrimers within the endoplasmic reticulum before being secreted to the basement membrane (247, 315)	Synthesized by endothelial cells, SMCs, pericytes and fibroblasts (http://betsholtzlab.org/VascularSingleCells/database.html)	- Major structural component of vascular basement membrane (247, 249)
CTSA	CTSA (Cathepsin A) (480 aa) belongs to the family of serine proteases. Synthesized as a single-chain precursor converted into a catalytically active heterodimer consisting of 20- and 32-kDa subunits stabilized by a disulfide bond (316).	Ubiquitous expression. May have a higher expression in microglia/macrophage (http://betsholtzlab.org/VascularSingleCells/database.html)	- Has a protective function (form a multi-enzyme complex with β-galactosidase and α–neuraminidase to protect these two enzymes against lysosomal degradation) and a catalytic function to hydrolyze bioactive peptides (317). - Recessive CTSA mutations cause the lysosomal storage disease galactosialidosis due to deficiency of β-galactosidase and neuraminidase-1(318).
GLA	α-galactosidase A enzyme (429 amino acid precursor processed to a 370 aa glycoprotein functioning as a homodimer). Catalyzes the hydrolysis of substrates possessing terminal α-galactosidic residues and participates in their degradation in the lysosome (319).	Expression in a variety of cell types, including endothelial, smooth muscle, nerve cells, renal cells (podocytes, tubular cells, glomerular endothelial, mesangial, and interstitial cells), cardiac (cardiomyocytes and fibroblasts) and nerve cells (279).	Rate-limiting enzyme in the lysosomal metabolism of glycosphingolipids. Lack of α-galactosidase A leads to the progressive accumulation of glycosphingolipids, mainly globotriaosylceramide, and its deacylated form (319, 320).
HTRA1	HTRA1 (480 aa), secreted PDZ-serine protease which belongs to the HTRA protein family whose members function as chaperones and serine proteases. Exists in a dynamic equilibrium between monomers and trimers (243, 244). The sensor domain of loop 3 (L3) and the activation domain of loop D (LD) play essential roles in trimer-mediated activation of the neighboring HTRA1 (241, 243)	Synthesized by endothelial cells of large arteries and astrocytes (239)	Cleaves many extracellular proteins including latent transforming growth factor beta binding protein 1 (LTBP1) and may positively regulate the TGF-B pathway (240, 242)
LAMB1	LAMB1 (1786 aa). Laminins are cross-shaped heterotrimeric glycoproteins, composed of an α, β, and γ chain that assemble into the endoplasmic reticulum to be secreted in basement membranes where they interact with other extracellular matrix proteins and integrins. LAMB1 can associate with the γ1 and all 5 α subunits (321).	Synthesized mainly by fibroblasts including leptomeningeal and perivascular fibroblasts and to a lesser extent by smooth muscle and endothelial cells (http://betsholtzlab.org/VascularSingleCells/database.html)	Laminins are major components of basement membranes in the meninges and vessels where they play a key role in the integrity and stability of the basement membranes and adjacent extracellular matrix. They are also involved in cell–cell interactions, signaling pathways, cell adhesion, migration, and differentiation (63, 321, 322). Biallelic pathogenic variants in LAMB1 are associated with autosomal recessive lissencephaly (OMIM 615191).
NIT1	NIT1 (327 aa). Member of the highly conserved nitrilase enzyme protein family found in animals, plant, fungi and many bacteria (323).	Ubiquitous expression with higher expression level in arteriolar smooth muscle cells (http://betsholtzlab.ora/VascularSingleCells/database.html).	One of its function is to act as a metabolite “repair enzyme” by hydrolyzing the deaminated form of the common intracellular antioxidant glutathione, an undesired side reaction of numerous transaminases (324).
NOTCH1	NOTCH1 (2555 aa), single-pass transmembrane receptor. Belongs to the highly conserved NOTCH receptor family. Synthesized as a single-chain precursor, cleaved into the Trans Golgi network as it traffics to the plasma membrane where it is expressed as a heterodimer composed of a large extracellular domain (Notch1ECD) and a transmembrane-intracellular domain (Notch1TMIC) (309)	Endothelial cells, SMCs, pericytes, ependymal, subependymal cells and astrocytes (325) (http://betsholtzlab.org/VascularSingleCells/database.html)	Important regulator of arterial endothelial cell specification and quiescence and vein remodeling during development. Acts as an endothelial mechanosensor to maintain junctional stability and endothelial cell quiescence and suppress inflammation in adult arteries (326).
SNORD118	Small Nucleolar RNA, C/D Box 118 (SNORD118) is a non-coding RNA gene which belongs to the evolutionary conserved family of Box C/D snoRNAs vital for ribosomal maturation.		Promotes cleavage of ribosomal RNA that produces the mature 5.8S and 28S rRNAs of the large ribosomal subunit (327)
TREX1	TREX1 (314 aa), most abundant endogenous 3′–5′ DNA exonuclease with a high affinity for single strand DNA, functions as a homodimer. The N-terminus of TREX1 is important for its exonuclease activity; the C-terminal domain is necessary for anchoring TREX1 to the endoplasmic reticulum and the activity of the oligosaccharyltransferase complex (328)	Ubiquitous expression. May have a higher expression in microglia/macrophage (http://betsholtzlab.org/VascularSingleCells/database.html)	- Degrade endogenous DNA and prevent type I Interferon (IFN) activation (329) - Recessive loss-of-function TREX1 mutations cause the Aicardi-Goutières syndrome, an early-onset encephalopathy characterized by basal ganglia calcification, white matter abnormalities and an overproduction of IFN (329)

NOTCH3-related cSVD.

Highly stereotyped dominant mutations in NOTCH3 cause CADASIL, the most frequent monogenic cSVD that shares many clinical and MRI features with sporadic cSVD, except for an earlier age of onset (213), although the distribution of WMH particularly the involvement of the anterior temporal pole is a useful differentiator from sporadic cSVD (214). NOTCH3 encodes a single-pass transmembrane receptor, predominantly expressed in mural cells—SMCs and pericytes— of small blood vessels, that comprises a large extracellular domain (Notch3^ECD) containing 34 epidermal growth factor repeats (EGFr), non-covalently linked to a transmembrane intracellular domain (Notch3^TMIC) (215). Molecular analysis of large cohorts of CADASL patients has repeatedly shown that CADASIL-associated NOTCH3 mutations result in a typical numerical alteration of cysteine residues within Notch3^ECD, be they missense, splice site, insertion, deletion or intronic mutations (151, 216–220). The clinical presentation and severity can be highly variable between and within CADASIL families (213). Modifiers of disease severity include both the position of NOTCH3 cysteine altering variant, with variants located in EGFr 1–6, 8, 11 and 26 being associated with a more severe phenotype than variants located in other EGFr (221–224), and cardiovascular risks factors, particularly smoking and hypertension, being associated with earlier onset of stroke (225–227). In addition to having a specific molecular signature, CADASIL has two related pathological hallmarks, namely aggregates of Notch3^ECD and deposits of granular osmiophilic material (GOM) that are present around mural cells in brain vessels, and also in small peripheral vessels of every organ. Specifically, cysteine altering variants lead to the disruption of the conserved disulfide bond pattern within individual EGFr resulting in the abnormal multimerization and subsequent aggregation of mutant and wildtype Notch3^ECD molecules in GOM deposits (153, 154, 215, 228–231). The relationship between NOTCH3 receptor activity and Notch3^ECD accumulation on the one hand, and CADASIL vascular pathology on the other, has remained elusive for more than 2 decades. Using several Notch3-related mouse models, combined with histopathological, imaging, molecular and genetic approaches, a recent study provided strong experimental evidence that Notch3^ECD aggregates are a key driver of arterial pathology in CADASIL through a mechanism unrelated to the activity of NOTCH3 receptor (231). Consistent with the findings of this experimental study is a clinical study showing that EGFr risk category is positively associated with Notch3^ECD aggregation load (224, 232). Interestingly, recessive loss-of-function mutations in NOTCH3 are associated with a rare and very severe form of cSVD, that differs from CADASIL by its childhood-onset, the possible association with cutaneous manifestations and the absence of GOM deposits (233–236).

HTRA1-related cSVD.

Pathogenic mutations in HTRA1 can manifest either as a rare recessive cSVD, CARASIL, caused by biallelic mutations or a more frequent autosomal dominant cSVD caused by heterozygous missense, frameshift or non-sense mutations (157, 237, 238), and often referred to as CADASIL2. A distinctive feature of the autosomal recessive form of HTRA1-related cSVDs is additional clinical features of premature alopecia and spondylosis deformans. HTRA1, which is mostly synthesized in the brain by endothelial cells of leptomeningeal arteries and astrocytes, encodes for a member of the highly conserved high-temperature requirement A (HtrA) serine peptidase family that comprises from the N to the C-terminus, an insulin-like growth factor–binding protein domain, a Kazal-type serine protease inhibitor domain, a trypsin-like serine protease domain, and a PDZ domain (239). HTRA1 is secreted into the ECM and cleaves a plethora of secreted and ECM-associated substrates, including components of the Transforming Growth Factor ß (TGF-ß) signaling pathway (240–242). Full enzymatic activity of HTRA1 requires trimerization of individual protomers, whereas the N-terminal IGFBP and Kazal domains and the C-terminal PDZ domain are dispensable. Two possible modes of regulation of HTRA1 activation have been proposed: one model posits that the protease domain of unliganded (ie, without bound substrate) HTRA1 exists in a dynamic and reversible equilibrium between an active and an inactive conformation and that the substrate samples the catalytically active conformation, resulting in its proteolysis. A second model suggests that substrate binding to the active site is sufficient to stimulate catalytic activity (243, 244). Importantly, both recessive and dominant pathogenic mutations in HTRA1 result in a strong reduction of HTRA1 activity through a haploinsufficient, loss-of-function or dominant negative mechanism; noting that patients with recessive mutations in HTRA1 are reported to have a more severe phenotype than those with heterozygous mutations, likely because HTRA1 activity is reduced to a greater extent (237, 245, 246). HTRA1-related cSVDs show significant overlap with CADASIL in clinical, MRI (including involvement of the anterior temporal poles), and histopathological features (Tables 5–6). Remarkably, HTRA1 protein was recently shown to be a component of Notch3^ECD-containing aggregates. Moreover, the proteome of brain vessels from CADASIL patients showed a substantial overlap with the proteome of mice lacking HTRA1, suggesting common mechanisms for these 2 unrelated monogenic cSVDs (240)

Table 6:

Main pathological features of brain vessels in patients with monogenic cSVD

Disease (mutated gene)	Main features of cerebral vessel changes
CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy) (NOTCH3)	Affects leptomeningeal, medium and small arteries and veins throughout the brain. Hyalinosis of the media with loss of SMC (uncomplicated hyalinosis is characterized by almost complete degeneration of SMCs with concentric accumulation of extracellular matrix components). Fibrosis and thickening of the adventitia. Severity of arterial lesions varies across brain region and ranges from very severe in the WM leading to lumen narrowing, severe in the basal ganglia, to less severe lesions in cortical arteries (294–297). Possible collagenosis of periventricular veins (298). Reduced capillary density in affected WM (299). Features specific to CADASIL: accumulation of Notch3ECD around mural cells of arteries, arterioles, capillaries and veins throughout the brain. Granular osmiophilic material (GOM), containing Notch3ECD and other extracellular matrix proteins, in the basement membrane of all mural cells (154, 215, 229, 230).
DADA2 (deficiency of adenosine deaminase 2 syndrome) (ADA2)	Multiple foci of petechial hemorrhages around small-size vessels in the white matter (265). Documented in peripheral tissues and considered to be similar to vascular lesions of Polyarteritis nodosa. Fibrinoid necrosis of small arteries with inflammatory infiltrates in active lesions. Possible thrombotic occlusions. In advanced lesions, intimal hyperplasia and diffuse fibrotic changes within the vessel wall (267, 268).
Hereditary SVD with osteoporotic feature (ARHGEF15)	-
COL4A1/2 hemorrhagic microangiopathy (COL4A1/A2)	In fetuses, supra- and infratentorial multifocal hemorrhagic lesions in the vicinity of enlarged small vessels having a discontinuous wall (300). Brain tissue collected from surgery for temporo-occipital dysplasia of a 22 month-old patient showing an aberrant expression of α-smooth muscle actin expression in capillary-like vessels (255) Irregular thickening, folding, and fragmentation of vascular basement membrane (301)
PADMAL (Pontine Autosomal Dominant Microangiopathy with Leukoencephalopathy) (COL4A1)	Affects small arteries and arterioles throughout the brain including the leptomeninges. Concentric thickening of the vessel wall and narrowing of the lumen. Subendothelial fibrous proliferation or smudgy, hyaline degeneration of the intima. Reduplication and sometimes fragmentation of the internal elastic lamellae. Hyalinization and fibrosis of the media. Changes consistent with organized thrombi (144, 297). Increased vascular expression of Collagen type IV (297, 302).
CARASAL (Cathepsin-A-related arteriopathy with Strokes and Leukoencephalopathy) (CTSA)	Affects the small distal arterioles (or possibly the arteriole-capillary transition (ACT) zone) throughout the WM, basal nuclei, and subependymal regions. Prominent vessel wall fibrous thickening, which is remarkably asymmetric, and loss of SMCs with near-total occlusion of the lumen (273).
Fabry disease (Anderson-Fabry disease) (GLA)	Lysosomal inclusions or lipid deposits that are seen in almost all cell types including both endothelial and SMCs appearing as lamellar inclusions with concentric and zebra-like patterns by electron microscopy. Blood vessels may be thickened with characteristic arteriosclerotic changes (303).
CARASIL (Cerebral Autosomal Recessive Arteriopathy with Subcortical Infarcts and Leukoencephalopathy) (HTRA1)	Affects meningeal, medium and small arteries throughout the brain. Fibrous thickening and increased number of myointimal cells in the intima. Multilayered and fragmented internal elastica lamina, hyalinization of the tunica media with severe loss of arterial SMCs. Sclerotic changes are mild and infrequent; most of the arteries are enlarged rather than exhibiting luminal stenosis. Possible arterial occlusion. Possible atherosclerosis of large extracranial and intracranial arteries (246, 280, 304)
HTRA1-related autosomal dominant cSVD (HTRA1)	Histopathologic features of small vessels quite similar, but apparently less severe, to those of CARASIL (246, 305).
Early amnestic syndrome of the hippocampal type and leukoencephalopathy (LAMB1)	-
NIT-1 related cSVD (NIT1)	Perforating arteries and arterioles of the cerebrum, basal ganglia, and brain stem have strongly abnormal vessel wall morphology with a media thickened by abundant hyalin and fibrin and large amorphic deposits. Deposits stain negative for NOTCH3 and amyloid but positive for Periodic acid-Schiff and Toluidine Blue and are electron dense by electron microscopy. NIT1 staining is strongly reduced (283).
NOTCH1-related microangiopathy (NOTCH1)	Affects arteries and arterioles of the deep cerebral WM, basal ganglia and cerebellar WM. Leptomeningeal vessels seem spared. Normal intima. Calcifications of the media and narrowing or occlusion of the lumen. Noncalcified arteriolar walls are also thickened and hyalinized (285).
Early onset autosomal recessive NOTCH3-related cSVD (NOTCH3)	-
Leukoencephalopathy with calcifications and cysts (LCC) (Labrune syndrome) (SNORD118)	Can affect small and large calibre arteries or veins (306). A spectrum of vascular changes ranging from mild mural hyalinization to severe fibrosis and total occlusion of the vascular lumen by thrombi, or fibrinoid degeneration. Presence of clusters of abnormal tortuous small vessels, with irregular and possibly calcified wall, resembling vascular malformations (290, 306, 307).
RVCL-S (Retinal vasculopathy with Cerebral Leukoencephalopathy and Systemic manifestations) (TREX1)	Affects medium and small calibre parenchymal arteries as well as veins, whereas leptomeningeal and cortical gray matter vessels are typically spared. Hyalinization or fibrinoid necrosis in the media, adventitial fibrosis and luminal narrowing. Occasional vascular thrombosis. Multilaminated basement membrane, oedematous endothelial cells, SMC and pericyte degeneration (292, 308).

COL4A1/A2-related cSVD.

Two different classes of pathogenic mutations have been identified in COL4A1 and COL4A2 genes, producing very different clinical presentations and radically different effects on collagen type IV expression. COL4A1 and COL4A2 genes are located in tandem on chromosome 13 and encode α1 and α2 chains of collagen type IV, which is the major component of basement membranes (BM). Each chain contains a large central triple-helical collagenous domain, a C-terminal non-collagenous (NC1) domain, and an amino-terminal 7S domain. Two alpha 1 chains and one alpha 2 chain assemble, through their 7S and NC1 domains, into a heterotrimer within the endoplasmic reticulum before being secreted to BM (247). Dominant mutations in the coding region of COL4A1 or COL4A2 cause a monogenic form of cSVD which is characterized by spontaneous deep ICH that can be associated with a large spectrum of renal, ocular, and skeletal anomalies (248). MRI frequently shows confluent WMH, lacunar infarcts and multiple CMB. The most frequent mutations are glycine substitutions in the conserved Gly-X-Y motifs within the triple-helical collagenous domain that impair the folding and secretion of collagen IV into the BM, resulting in structural changes in BM (249, 250). Penetrance of this hemorrhagic cSVD is incomplete, with less than 15% of mutation carriers developing ICH over the course of their lifetimes (251). Moreover, there is a wide variability in the severity of clinical manifestations between and within families. Indeed, ICH can occur at every age from the perinatal period (in utero or after birth) to adulthood, and the severity of cerebral bleeding can range from asymptomatic CMB detected by MRI to life-threatening macroscopic ICH (252, 253). Perinatal hemorrhage predisposes to porencephalic cyst cavity (254). It can also present with childhood seizures, especially in association with porencephaly (255). In Col4a1 mutant mice, spontaneous delivery has been associated with an increased risk of ICH, a consequence prevented by surgical delivery (256), and a similar association has been reported in human, leading to the recommendation that Caesarian section should be adopted for delivery of infants harboring COL4A1/2 mutations (211). The most conspicuous factor accounting for this clinical variability is the position of the mutation, with mutations located in the C terminal part of the collagenous domain being associated with a severe hemorrhagic phenotype. Also, mutations in COL4A2 seem to be associated with a lower penetrance and a less severe phenotype than COL4A1 mutations (249). In contrast, mutations within the 3’-untranslated region of COL4A1 causes a distinct cSVD named “PADMAL” (pontine autosomal dominant microangiopathy with leukoencephalopathy) and characterized by ischemic lesions which commonly affect the pons. These mutations disrupt the binding site of the microRNA, miR-29, resulting in increased expression of COL4A1 (257).

Cysteine altering variants in NOTCH3 are by far the most frequent pathogenic mutations detected in cSVD patients. In a series of 3,853 consecutive unrelated adult patients with genetically suspected cSVD, 413 (10.7%) carried a typical cysteine altering variant in NOTCH3, 63 (1.6%) a mutation in COL4A1 or COL4A2, 50 (1.3%) a heterozygous mutation in HTRA1, a single one harbored a bi-allelic loss-of-function mutation in HTRA1 and 10 (0.26%) in other cSVD genes (238). Moreover, whole exome sequencing in a consecutive cohort of 309 Chinese adult patients suspected of genetic leukoencephalopathy (ie, patients with progressive neurological symptoms, cerebral WM lesions and a negative workup) found 16.2% had CADASIL mutations, 3.6% had HTRA1 mutations and 2% had COL4A1/A2 mutations (258). Also, analysis of a large cohort of 970 consecutive patients with young-onset (≤ 70 years) apparently “sporadic” lacunar stroke showed that the frequency of CADASIL-causing variants was 1.2% (95% confidence interval [CI] 0.6%–2.1%) and that this frequency increased to 3.7% in patients aged ≤60 years with confluent WMH (209).

5.4.2. Polygenic/multifactorial cSVD

The last five years have seen major advances into the understanding of polygenic contributions to cSVD, with more than 50 genetic loci now identified for the different cSVD phenotypes (197, 208, 330). Studies have largely relied on the genome wide association study (GWAS) approach and applied it to both lacunar (small vessel) stroke, and to the MRI phenotypes of cSVD. Although epidemiological studies suggested lacunar stroke had a significant genetic component (331), early GWAS studies identified few genetic loci. For example, in the METASTROKE case control study of ischemic stroke and its subtypes, although 32 loci were identified, only one was specific to lacunar stroke (332), and this had already been reported (333). A major factor underlying this initial lack of success, which was in contrast to success in other stroke subtypes, appears to be inadequately rigorous phenotyping of the lacunar stroke subtype. Heritability estimates obtained from the GWAS data for lacunar stroke were greatly increased when studies were performed in patients with MRI confirmed lacunar stroke (334). Using such a more rigorous phenotyping approach in 7338 cases, 5 loci for lacunar stroke were identified (ICA1L-WDR12-CARF-NBEAL1, ULK4, SPI1-SLC39A13-PSMC3-RAPSN, ZCCHC14, ZBTB14-EPB41L3), and a further 7 identified on multitrait analysis with WMH (SLC25A44-PMF1-BGLAP, LOX-ZNF474-LOC100505841, FOXF2-FOXQ1, VTA1-GPR126, SH3PXD2A, HTRA1-ARMS2, COL4A2)(197). The multitrait analysis of genome-wide association summary statistics (MTAG) approach involves joint analysis of multiple (genetically correlated) traits, which increases statistical power to detect genetic associations for each trait (335). The majority of loci identified so far have been in populations of European ancestry and GWAS studies in other ethnic groups are needed. A related stroke phenotype is ICH; cSVD can present with deep (subcortical) ICH. There have been smaller GWAS studies in this area but genetic loci are being identified (336). A recent review concluded that by combining univariate and multivariate findings, and taking into account the fact that three loci are shared between ICH and lacunar stroke, a total of 17 independent risk loci for cSVD-related stroke have been identified. At nearly all of these loci, the lead risk variants were common variants with a minor allele frequency >5% (330).

The other cSVD phenotype widely studied is MRI markers which includes WMH, diffusion tensor imaging (DTI) metrics of WM ultrastructure, small deep infarcts, CMB, and enlarged PVS. These studies have taken advantage of large population-based cohorts, including more recently UK Biobank, and have identified over 40 loci for cSVD (330, 337). Most studies have quantified WMH, or DTI markers (208), but a more recent study of enlarged PVS identified 24 loci (337). There have been fewer studies of CMB largely because large data sets are not available due to the need for visual identification on MRI scans; a recent meta-analysis in 25,862 participants of whom 3556 had CMB, suggested significant heritability for this phenotype but only identified one gene at genome wide significance, the Apolipoprotein E locus (338). Many loci are shared between lacunar stroke and MRI markers of cSVD demonstrating shared pathophysiology across the different phenotypes of cSVD, but others are not (339). This may reflect a lack of statistical power to detect genome wide associations particularly for lacunar stroke where the sample sizes are smaller, but could also reflect different processes being involved in the chronic manifestations of cSVD, compared with the acute onset ischemia occurring in lacunar stroke. GWAS data also supports heterogeneity within lacunar stroke, with different genetic architecture reported for cases of single isolated lacunar stroke, compared with multiple smaller lacunes with confluent WMH (334). This is consistent with early pathological reports from Miller Fisher reporting that the pathology associated with the former is atherosclerosis, and with the latter diffuse small arterial damage referred to as lipohyalinosis (334).

A number of the loci associated with cSVD are also associated with conventional cardiovascular risk factors, particularly BP, but 25 of the first 52 cSVD risk loci were found not to be shared with any vascular risk factor, suggesting novel disease mechanisms acting independent of conventional risk factors (330). The mechanisms linking these variants to cSVD are not fully understood, but a number of biological processes have been implicated. Disruption of the ECM and matrisome is emerging as a key disease mechanism in monogenic cSVD (322), and the finding that a number of the genetic loci associated with cSVD encode matrisome related proteins suggests that this may also be an important mechanism for sporadic cSVD. Not only are common variants in some matrisome related genes causing monogenic cSVD associated with sporadic cSVD (HTRA1 and COL4A2) (208), but a number of other matrisome associated proteins have been associated with sporadic cSVD including LAMC1, EFEMP1, VWA2, VCAN, LOX, SH3PXD2A, NID2, HBEGF, AGRG6, FBN2, ADAMTSL4) (197, 330). Other processes implicated by the GWAS studies of sporadic cSVD include disruption of myelination (LPAR1, ULK4, AGRG6, VCAN), membrane transport (KCNK2, SLC13A3, SLC202A), vascular development (FOXF2, WNT7A, LPAR1) and the BBB (TRIM47, FOXF2, LAMC1, WNT7A) (330, 337).

Increasingly GWAS data is being analyzed in combination with gene expression, proteomic and metabolomic data to better understand the underlying disease pathways, and identify potential treatment approaches (340). Metabolomic analyses have reported a number of associations with cSVD (341, 342). For example, decreased levels of multiple glycerophospholipids and sphingolipids were associated with increased WMH volume, lower mean diffusivity on DTI, greater brain atrophy, and impaired cognition. Higher levels of creatine were associated with increased lacune count, higher WMH and impaired cognition (341). Further work is required to integrate these findings with those from GWAS and other ‘omic’ technologies. Transcriptome-wide association studies integrate GWAS with expression quantitative trait loci (eQTL) datasets allowing associations between tissue-specific gene expression levels and a given phenotype to be identified. Colocalization analyses can then be used to test whether the same causal variant identified on GWAS with the disease of interest is also associated with expression levels of a gene in a given tissue (343). Such approaches identified association of genetically elevated expression of SLC25A44, ULK4, ICA1L, FAM117B, CARF, NBEAL1, with lacunar stroke (197). By combining analysis of GWAS and proteomic data from lacunar stroke, three proteins (ICA1L, CAND2, ALDH2) felt to be causally related to disease pathogenesis were identified (344).

MR is being increasingly applied to gain further insights into pathophysiology and treatment approaches. This technique uses genetic variants as instrumental variables to approximate the effect of an exposure to seek evidence for a causal association with stroke (345). Numerous MR studies have been published on cSVD and have confirmed the importance of conventional cardiovascular risk factors such as blood pressure (see section 5.3 on modifiable risk factors). Novel, potentially treatable risk factors have also been identified. For example, epidemiological associations had been reported between elevated serum homocysteine and lacunar stroke (346), but whether the association is causal, or secondary to confounding, for example by impaired renal function reducing homocysteine clearance, has been debated. MR studies found genetically predicted homocysteine was associated with lacunar stroke, but was not associated with large artery or cardioembolic stroke, supporting a causal link (347). Homocysteine levels can be reduced by folic acid and B vitamin supplementation. The large VITATOPS (The VITAmins TO Prevent Stroke) trial in patients with stroke or transient ischemic attack, found no benefit of this approach in patients with ischaemic stroke as a whole. However the MR data would suggest any effect would be limited to lacunar stroke, and consistent with this, a secondary analysis of VITATOPS showed a marginally significant treatment effect for small vessel stroke, but not other stroke subtypes (348). A MRI sub study embedded in the same trial showed that homocysteine-lowering therapy significantly reduced the progression of WMH on MRI (349). In 20,702 adults with hypertension but without a history of stroke or myocardial infarction folic acid supplementation reduced stroke risk (350).

This is an example of genomics-driven drug discovery. It has been estimated that genetic support of drug effects could double the success rate of clinical trials (351), and pharmaceutical companies are increasingly integrating genomics-based strategies into drug discovery pipelines. MR techniques can be applied using genetic variants as instrumental variables which approximate the effect of exposure to a particular drug. A study using genetic instruments mimicking the effects of different antihypertensive agents found calcium channel blockers, but not β-blockers, were associated with a lower risk of lacunar stroke (352). Genetic data can also be used to identify opportunities for drug repurposing by identifying enrichment of gene sets from GWAS data in known drugs (353). This technique has shown significant enrichment of PVS genes in targets of drugs validated or under investigation for vascular and cognitive disorders (for example, telavancin and davunetide) (337). Proteins are the targets of most medicines and are located more distally in the causal chain from gene to disease. GWAS of plasma protein levels have identified genetic variants that are associated with proteins, referred to as protein quantitative trait loci (pQTLs) (354). This offers the opportunity to test the causal effect of potential drug targets on the human disease phenome using MR (355). Such approaches are currently being applied to stroke, and have identified potential drug targets although there is limited data on lacunar stroke, partly due to the smaller number of cases with the lacunar subtype available reducing power compared to analyses in all ischemic stroke (356).

5.4.3. Blurring of boundaries between monogenic and polygenic cSVD

Traditionally, monogenic and polygenic forms of cSVD have been thought to be distinct, with the monogenic forms being rare and cardiovascular risk factors only influencing polygenic cSVD. However recent data is challenging these assumptions (212). On the one hand, while mutations in both HTRA1 and COL4A2 cause monogenic cSVD, common variants in the same genes are risk factors for sporadic cSVD. Specifically, a common intronic variant in HTRA1 was shown to be associated with a combination of MRI markers of extreme cSVD (including the presence of lacunes and extensive WMH) in population-based cohorts (357), and lacunar stroke in patient-based cohorts (197). Likewise, common variants near the COL4A1-COL4A2 locus were found to be associated with WMH burden (208, 358), non-lobar ICH (359), and lacunar stroke (197).

On the other hand, variants in NOTCH3, COL4A1/A2 and HTRA1, identical to those that cause monogenic cSVD, have been found to be much more common than expected in the general population, and some of these may contribute to the risk of apparent “sporadic” cSVD in the general population (212). The prevalence of CADASIL in the United Kingdom has been estimated to be 2–6 per 100,000 (259, 260). However, analysis of sequencing data from normal populations has reported that typical cysteine changing NOTCH3 mutations are much more common in the general population, with a frequency ranging from one in 1,000 in European populations, to up to one in 100 in certain Far Eastern populations such as Taiwan (360, 361). Despite apparently not causing classical CADASIL these mutations are nevertheless associated with an increased risk of stroke and vascular dementia in the general population (362). Why some cysteine mutations result in classical CADASIL, while others are asymptomatic or merely increase the risk of sporadic stroke, is incompletely understood. Mutations in the proximal EGFr (1–6) are more common in CADASIL clinic cohorts, while more distal mutations (EGFr 7–34) are more common in population cohorts (224). Proximal mutations are associated with an earlier age of onset of stroke (363) and more vascular Notch3^ECD accumulation (232). Conventional cardiovascular factors also appear to be a modifying factor; individuals with cysteine mutations in the general population were more likely to develop stroke if they had a higher Framingham cardiovascular risk score (362). This parallels data from CADASIL families demonstrating that cardiovascular risk factors influence the phenotype in CADASIL families. For example smoking was associated with an approximately 10 years earlier age at stroke onset (364), while hypertension doubled the risk of stroke (225, 227) and higher BP over the normal range was associated with an increased rate of brain volume loss (365). However, mutation site and cardiovascular risk factors fail to account for much of the phenotypic variation. It is thought that other genetic modifiers, outside the NOTCH3 gene, are important. This is supported by family studies within CADASIL families; the extent of T2 WMH lesion volume was highly heritable (heritability 0.634), independent of the site of the NOTCH3 mutation (366). Studies are currently underway to try to identify these genetic modifiers. Similar findings are emerging for other monogenic forms of cSVD. Mutations in both HTRA1 and COL4A1/2 are also much more common in the general population (1 in 832 and 1 in 1,353 respectively in the UK Biobank) than is the prevalence of familial forms of the diseases (362). HTRA1 mutations in UK Biobank were associated with an increased risk of stroke and vascular dementia, while COL4A1/2 mutations increased risk of both stroke and ICH. As for NOTCH3 mutations, a higher Framingham risk score was associated with increased stroke risk in these with HTRA1 population mutations. The much greater frequency of monogenic mutations in the general population than expected from the disease prevalence has now been reported for many single gene disorders across many organ systems (367, 368) and is resulting in a paradigm shift in the way we consider the boundaries between monogenic and polygenic stroke.

Taken together the above data demonstrate that the distinctions between monogenic and sporadic cSVD are less strict than previously thought. Rare variants may not only cause typical familial disease but may also contribute to the burden of apparent sporadic cSVD- for example in the UK Biobank study by Cho and colleagues such rare NOTCH3 variants accounted for 0.21% of all strokes and 0.77% of all vascular dementias in the general population (362). Typical familial disease is influenced by interaction with environmental risk factors such as hypertension and smoking, in a similar manner to sporadic cSVD. This emphasizes that both monogenic cSVD and sporadic cSVD have shared mechanisms, and therefore insights from monogenic forms of the disease are likely to have wider implications for sporadic cSVD.

6. Experimental animal models for the study of cSVD

Animal models are instrumental in exploring the sequence of pathologic events, elucidating the cellular, molecular and physiological mechanisms, identifying and validating biomarkers and testing therapeutic strategies. Because cSVD encompasses a heterogeneous group of diseases, it is expected that no single model will recapitulate all features of cSVD. A useful model should recapitulate the vascular pathology or brain lesions. With respect to the species of the animal model, mouse and to a certain extent rat have many advantages over large animals including a short duration of gestation (20–24 days), a high reproduction rate (6 to 8 pups per litter), lower operating costs, an abundance of genetic resources (https://www.informatics.jax.org/) and the possibility to “easily” perform genome editing for mechanistic studies. However, one limitation is the lower ratio of WM to gray matter volume in rodents compared to the human brain (369). In addition, the small size of the mouse brain can result in insufficient spatial resolution for MRI studies, although this is becoming less an issue with technical advances in this field including high field strength scanner (370). On the other hand, large animals such as nonhuman primates or pigs have gyrencephalic brains and abundant WM that make them closer to humans (369). Moreover, they have more physiological and hemodynamic similarities to humans compared with rodents. Below, we review the three main types of models that are used for the study of cSVDs and discuss their pros and cons.

6.1. Surgical animal models

6.1.1. Models of diffuse white matter lesions

Since their original description by Alzheimer and Binswanger, WM lesions in cSVD have been thought to result from chronic hypoxic hypoperfusion (see section 3.2). Below, we briefly review the main approaches that have been used to mimic diffuse WM lesions related to hypoperfusion.

A first approach consisted in the permanent occlusion of 2 or 3 major arteries (carotid or vertebral arteries) feeding the brain (371). Rats with bilateral occlusion of common carotid arteries develop extensive changes in the hemispheric WM bundles characterized by WM rarefaction and glial (astrocytes and microglia) activation (372). However, one major limitation of this model is the presence of severe damage in the visual pathway, precluding its use for behavioral tests. A similar approach has been applied in non-human primates by submitting adult baboons (7–12 years) to permanent occlusion of both the internal carotid arteries and the left vertebral artery (three-vessel occlusion) for 1 month (373). Similar to rodents, baboons show severe WM changes at 14 days after surgery (373).

A second approach consisted in stenosing rather than occluding these large extracranial arteries. In a seminal paper published in 2004, Shibata and colleagues developed the bilateral common carotid artery stenosis (BCAS) mouse model, in which stenosis was induced by twining micro-coils (0.18 mm inner diameter, made of piano wire) around the two common carotid arteries (374). Thereafter, several variants of this model have been reported using mice of different strains, different ages or micro-coils of smaller or larger diameters (375). In summary, in the BCAS model, the CBF is reduced to 60–70% of the baseline at 3 days in the cortex, and probably more in the WM and deep brain regions, and recovers to ~80% at one month likely because of the development of collateral circulation. Mice develop two-weeks after surgery demyelination and glial activation, with relative preservation of the visual pathway, and defects of spatial working memory (375). The cortex and the hippocampus are usually spared after one month of hypoperfusion but changes in these areas can be detected after prolonged (several months) hypoperfusion (376). Another variant of the BCAS model is the ACAS (asymmetric common carotid artery surgery) model that consists in the implantation of an ameroid constrictor to the right common carotid artery and the placement of a micro-coil to the contralateral left common carotid artery; the ameroid constrictor results in the gradual occlusion of the right carotid artery over 1 month and the microcoil induces an immediate 50% stenosis of the left carotid artery (377). In this model, CBF drop is slower (over 1 week) but much more pronounced, especially on the right hemisphere (ameroid side), in comparison to the BCAS model. In addition to WM lesions, ACAS mice develop cortical and subcortical infarcts on the right hemisphere (ameroid side), which are likely related to the more pronounced hypoperfusion (20–30% of the baseline) (377).

The pros and cons of diffuse WM lesions obtained by surgery.

These models can be useful to study the pathophysiological consequences of hypoperfusion, which might be one aspect of cSVD, and test therapeutic strategies (375). However, there are several important limitations to consider. First, hypoperfusion is induced by acting on major extracranial vessels (carotid or vertebral arteries) whereas cSVD affect primarily intracranial vessels. Second, these are models of subacute hypoperfusion, that appear over a few days, in previously normal vessels whereas hypoperfusion in cSVD patients occurs over months or years in the context of diseased vessels. Third, these models usually employ young and healthy animals, without any vascular risk factor.

6.1.2. Models of lacunar infarcts and microinfarcts

A first method, developed in rats and mice, involves the injection of microemboli, such as microspheres (378, 379), fibrin clots (379) or cholesterol crystals (379–381), through the internal or common carotid artery to cause microvessel occlusions. This method produces multiple small infarcts throughout the cortex, subcortical tissue and hippocampus (378, 381). Wang and colleagues studied the natural history of lesions in a mouse model of cholesterol crystals emboli and found the presence of two different classes of lesions: (1) conventional ischemic lesions bounded by a glial scar, with rapid neuronal loss and cavitation; these were the less frequent lesions and (2) incomplete lesions, the most frequent ones, with widespread gliosis and delayed neuronal loss and demyelination (381). Using two-photon imaging and mouse models of microembolization with fibrin clots or cholesterol crystals, the Grutzendler’s lab showed that a high proportion of microemboli failed to be eliminated within 48h but instead were extravasated into the perivascular space, by a process named angiophagy, allowing vessel recanalization and blood flow reestablishment within the next days (379, 382).

A second method, developed in the rat, involves the photochemical injury of the common carotid artery by the combination of laser irradiation of the artery and injection of a photosensitizer dye to cause a non-occlusive thrombus in the carotid artery producing distal platelet emboli in the brain. This method produces several small infarcts (0.7–2 mm) in the cortex, thalamus, basal ganglia and hippocampus (383–385). A refined variant of this method is the occlusion of a single penetrating arteriole by targeted photothrombosis that also uses a circulating photosensitizer but targets an individual penetrating artery in the cortex by the laser light of a two-photon microscope. This produces a highly localized infarct in the cortex, with a columnar shape, over a course of a week (386).

A third method involves the stereotactic injection into the brain parenchyma of endothelin-1, a potent vasoconstrictor peptide. Endothelin-1 injection induces a transient reduction (~60%) in CBF causing ischemic lesions (387). This method has been applied to different species such as the rat (388), pig (389) or macaque (390) and in various brain regions, such as the motor cortex, the striatum (391), the internal capsule (392) or the periventricular WM (393). Controversial results have been reported in the mouse with some reports showing ischemic lesions (394) whereas others reported no lesion (395). Variants of this method uses L-N5-(1-iminoethyl)ornitine (L-NIO), a potent selective inhibitor of endothelial nitric oxide synthase, instead of endothelin-1 (396) or endothelin-1 in combination with Nω-nitro-L-arginine methyl ester hydrochloride (L-NAME), another nitric oxide synthase inhibitor (397). In general, vasoconstrictors produce focal infarcts, especially when injected in the WM, that are characterized by glial and neuronal degeneration, usually without affecting the BBB, associated with sensory or motor deficits (392, 394) However, there are several limitations with the endothelin-1 models. A first one is that it induces the constriction of multiple vessels at the same time and thus the infarct can be bigger than a lacunar infarct. Second, in addition to vascular cells, endothelin-1 targets neurons and glial cells.

A last method consists in the selective occlusion of the anterior choroidal artery in minipigs which produces small infarcts in the internal capsule (398) However the surgery is complex, there are much less resources for minipigs than for mice and maintenance of minipigs is much more expansive.

The pros and cons of lacunar infarct models.

These models are valuable for understanding the progression of ischemic infarcts and their impact on motor deficits or cognition, for investigating recovery and repair in chronic stages and for evaluating therapeutic strategies. However, none of these surgical animal models accurately recapitulates one core feature of cSVD, namely the occurrence of brain lesions because of chronic damages of small brain vessels. Another drawback of these models is the mortality rate resulting from the surgical intervention (387).

6.2. Hypertensive models

As mentioned above (section 5.3), chronic hypertension is a major risk factor of cSVD. Many models of hypertension have been developed either by selective breeding, surgical and/or pharmacological manipulation or targeted genetic manipulation (399) A selection of the most significant models is discussed below.

The aortic coarctation in rhesus monkey has been used in a limited number of studies (400). This evokes a variable BP elevation ranging from ~150 to 250 mmHg (systolic BP). Monkeys develop complete and incomplete microinfarcts (< 1 mm), in highly variable number, in grey and WM that are associated with impaired short-term memory and executive functions, but no overt WM lesions, (401).

The renovascular hypertensive rat model, obtained by surgical constriction of the 2 renal arteries (two-kidney, two-clip model), develop sustained hypertension (180–220 mmHg, systolic BP) by ~20 weeks post-surgery (402, 403). Compared to nonhypertensive sham rats, hypertensive rats develop infarcts in the cerebral cortex and adjacent WM, hemorrhages and WM lesions with reactive gliosis in the corpus callosum. Small arteries or arterioles exhibit hyaline degeneration, fibrinoid necrosis and lumen narrowing; infarcts have been attributed to thrombotic occlusion (402–404). Although promising, one strong limitation of this model is the mortality rate associated with the surgery which can reach 35% (403).

The spontaneously hypertensive rat (SHR) has been developed in Japan in 1963 by selective cross breeding of Wistar Kyoto rats. The spontaneously hypertensive rat stroke-prone (SHRSP) has been developed from a substrain of the SHR line for very severe hypertension and stroke phenotype (405). The SHRSP is a widely used model, for a while. Increase in BP in SHRSP starts at 4 weeks and reaches a plateau at > 220–250 mmHg (systolic BP) by ~5 months in males and ~6–7 months in females. Brain pathology includes WM lesions by 20 weeks of age, small (2 mm³) and large (up to an entire hemisphere) infarcts as well as microhemorrhages in cortical and subcortical areas in ~80% of rats by 30 weeks of age, although with a wide range of severity (400, 405, 406). Some reports mention the vacuolization of the brain tissue which may progress to the formation of necrotic cysts and the presence of multifocal BBB leakages, especially in rats with BP above 220 mmHg with a close spatial relationship between brain lesions and the site of BBB rupture (407–409). In contrast, other reports fail to detect brain lesion in the absence of overt BBB leakage (405). Vascular lesions consist in the thickening of the wall of pial arteries whereas small intracerebral arteries, especially those close to brain lesions, may exhibit hyaline degeneration, SMC loss, fibrinoid necrosis and thrombosis (400, 405). Anti-hypertensive therapies improve survival, brain and vascular lesions (400). One big problem with the SHRSP model is the high variability in brain pathology ranging from no brain lesion to premature mortality by 30 weeks (387). For this reason, many studies have employed SHRSP rats fed with a high salt diet, a regimen which strongly accelerates the onset and severity of BP increase and brain lesions, and increases the incidence as well as the mortality rate of rats (410). However, this has caused a lot of confusion in the SHRSP literature. Remarkably, Herisson and colleagues recently provided strong experimental evidence that SHRSP fed with high-salt diet do not recapitulate cSVD but an acute syndrome in human called posterior reversible encephalopathy syndrome (PRES). PRES associates various neurological manifestations (e.g., encephalopathy, seizures) with a typical brain edema and often develops in the setting of malignant hypertension (411). Specifically, SHRSP on high-salt diet (also called permissive diet) develop severe BBB disruption, white matter vacuolization, microbleeds and a thrombotic microangiopathy, that are reversible upon switching to regular diet and implementing antihypertensive treatment (411). This is an example of hypertensive encephalopathy occurring as a result of a sudden, sustained rise in BP sufficient to produce a 3 to 4-fold increase in CBF, known as CBF autoregulatory breakthrough, that causes vasogenic oedema (4).

In the SHR rat, BP is normal at birth but gradually increases to ~ 180–200 mmHg (systolic BP) between 7 and 15 weeks of age (412). Compared to the SHRSP rats, SHR rats have no predisposition for stroke and no sensitivity to the high salt diet (411). Compared to the parent strain (Wistar Kyoto), SHR have a lower body weight irrespective of age. MRI in SHR shows progressively smaller total intracranial volume and enlarged lateral ventricles that are unrelated to hypertension (413, 414). SHR exhibit also a reduction of WM volume (413, 415) but it is unclear whether this is related to hypertension. For all these reasons, SHR is not considered as a good cSVD model.

Angiotensin II (Ang II) dependent models.

The renin-angiotensin-aldosterone system (RAAS) plays a key role in normal water and sodium homeostasis and RAAS activation is an important contributor in human primary hypertension (416). The R+/A+ double transgenic mouse, which overexpresses human renin and human angiotensinogen resulting in chronic overproduction of Ang II, develops a moderate hypertension (~150 mmHg, systolic BP) (417). Hypertensive mice do not develop overt brain lesions unless they are challenged with L-NAME and/or high salt diet, which aggravates hypertension (up to ~220 mmHg, systolic BP) and leads to multiple microhemorrhages in the cortex, deep brain region and brainstem and a few ischemic lesions in the cortex and the brainstem (418). Unchallenged transgenic mice exhibit a remodeling and hypertrophy of small pial arteries (419). A second model consists in infusing Ang II, via a minipump surgically implanted subcutaneously, during ~4 weeks (416). This is one of the most widely used model of hypertension in rodents and especially in the mouse. Different doses (slow, intermediate or high-pressor doses) of Ang II can be infused to analyze hypertension of different severities. Four-weeks infusion of Ang II increases BP within the first 24 hours and causes remodeling and hypertrophy of cerebral arteries within 2 weeks (described in detail in section 7) but no overt brain lesions (416). Only very high doses of Ang II inducing malignant hypertension produces cognitive deficits characterized by impaired learning ability and spatial memory as well as increased anxiety (420). A third interesting model consists in the administration of mineralocorticoids in combination with a high-salt diet (desoxycorticosterone (DOCA)-salt model); a unilateral nephrectomy is often included to further increase hypertension. In the DOCA-salt mouse model with unilateral nephrectomy, BP gradually increases over 2 weeks to ~140–150 mmHg (mean BP) (421). The DOCA-salt model is relevant to human hypertension as it involves an hypervolemia, an activation of the sympathetic nervous system and the suppression of the peripheral RAAS system, and thus a low renin hypertension, which is observed in ~32 % of the hypertensive population (422). Importantly, this model is associated with an activated renin-angiotensin system in the brain (423). Hence, this model can be ascribed as a two “hits” hypertensive model with one hit being the increased arterial pressure and the second hit, the activation of the brain RAAS (424).

The Schlager BPH/2 (Blood Pressure High) mouse has been developed by selective cross breeding to produce spontaneously hypertensive mice (425). BPH/2 mice are a model of moderate lifelong hypertension (~140mmHg, systolic BP) (419) with increased BP as early as 6 weeks of age (Schlager and Sides, 1997). Several lines of evidence indicate that BPH/2 mice have a neurogenic form and non–renin-dependent mode of hypertension (425). Compared to control mice (BPN/3, Blood Pressure Normal), BPH/2 have a lower weight, which does not seem to be related to hypertension, a mild cardiac hypertrophy, but no overt brain lesions (425). Deficit in spatial learning and spatial short-term memory is detected in middle aged BPH/2 mice (426). Pial arterioles of BPH/2 mice exhibit a hypertrophy but no remodeling in contrast to mice with an activation of the RAAS system (419).

The pros and cons of hypertensive models.

The use of hypertensive models is clinically relevant given the fact that longstanding hypertension is a major risk factor of sporadic cSVD. However, all these models are not equal and each have limitations. The surgical models (aortic coarctation and two-kidney, two-clip model) have technical (mortality rate, difficulty to implement) and ethical issues (use of non-human primates). The high variability of the cerebrovascular phenotype developed by SHRSP is an important limitation. Moreover, because of the similarity of brain phenotypes between SHRSP on high-salt diet and SHRSP fed with a “normal diet” but with BP above 220 mmHg, it is reasonable to speculate that SHRSP may be of uncertain relevance to cSVD and be rather a model of malignant hypertension. The Ang II infusion is a short-term hypertension model. No brain lesions have been reported in the BPH/2 model but its analysis may have been not optimal. The control strain (BPN/3) of BPH/2 does not differ only by BP levels but also by genetics (https://www.jax.org/strain/003004 versus strain/003004). In summary, the Ang II infusion, the DOCA-salt and the BPH/2 mouse models, which are complementary models in terms of mechanism and duration of hypertension, are easy to implement and represent a valuable asset to examine the impact of hypertension on the structure and function of brain vessels.

6.3. Genetically based models

To date, animal models have been generated for 11 out 15 monogenic cSVDs, including the 3 most frequent ones (CADASIL, HTRA1-related cSVD and COL4A1/2 hemorrhagic microangiopathy), yet with varying degrees of success. These models and their main manifestations are summarized in Table 8.

Table 8:

Animal models of monogenic cSVD

Disease	Mutated gene	Animal models	Main manifestations reported in the literature
CADASIL (Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy)	NOTCH3	Transgenic mice expressing NOTCH3 under the control of a SMC (SM22α) promoter: TgSM22α-humNotch3R90C (438); TgSM22α-humNotch3C428S (439).	Notch3ECD accumulation and GOM deposits (438, 439) Cerebrovascular dysfunction (TgSM22α-humNotch3R90C) (440) Increased susceptibility to cortical spreading depression, the phenomenon underlying migraine with aura (TgSM22α-humNotch3R90C)
		Conditional mice: ROSA26Sortm1(NOTCH3*R1031C); ROSA26Sortm1(NOTCH3*C455R) bred to SM22α-Cre mice (441). Expression of the mutated human NOTCH3 is inserted into the Gt(ROSA)26Sor (ROSA26) locus, blocked by an upstream loxP-flanked STOP sequence. When bred to SM22α-Cre mice, which express the Cre recombinase SMCs, the STOP sequence is deleted, resulting in NOTCH3 expression in SMCs.	GOM deposits (441)
		Transgenic mice overexpressing NOTCH3 under the control of the rat Notch3 promoter TgNotch3R169C (433) or the human NOTCH3 promoter TgNOTCH3R182C (432)	Notch3ECD accumulation and GOM deposits (432, 433) White matter lesions (TgNotch3R169C) (433, 435, 436) SMC loss (TgNotch3R169C on the C57BL/6 background) (231) Cerebrovascular dysfunction (TgNotch3R169C on the FVB background) (433, 434)
		Targeted mutation (K-in) into the endogenous Notch3 murine gene: Notch3R142C/+ (442); Notch3R170C/+ (443).	Notch3ECD accumulation and GOM deposits (Notch3R170C/+) (443, 444) Late arterial SMC loss (Notch3R170C/+) (231)
DADA2 (deficiency of adenosine deaminase 2 syndrome)	ADA2	No mouse model (the mouse lacks an ADA2 ortholog) Transient knockdown of cecr1b (the zebrafish paralogue of ADA2) using morpholino in zebrafish (cerc1b-MOs) (265) Targeted knockout of cerc1b in zebrafish (cerc1b-LoF) (445)	Intracranial hemorrhages and neutropenia (cerc1b-MOs) (265) Defective hematopoiesis and increased inflammation (cerc1b-LoF) (445)
Hereditary SVD with osteoporotic feature	ARHGEF15	Targeted mutation (K-in) into the endogenous Argef15 gene (Arhgef15-e(V368M)1) (270)	Increased rate of mortality, behavioral phenotypes and severe osteoporosis (270)
COL4A1/2 hemorrhagic microangiopathy	COL4A1 or COL4A2	Chemical mutagenesis (Ethylnitroso-Urea): Col4a1 G1064D/+ (SVC); Col4a1 G627W/+ (Bru) (429); Col4a1+/Δex41 (254); Col4a1 G394V/+; Col4a1 G646D/+; Col4a1 G658D/+; Col4a1 G912V/+; Col4a1 G1038S/+; Col4a1 G1180D/+; Col4a1 G1344D/+ (427).	Reduced viability; lower blood pressure; perinatal hemorrhages, porencephalic cyst cavities, microbleeds, deep intracerebral hemorrhages; eye, muscle and kidney defects with incomplete penetrance and variable expressivity (54, 250, 254, 427–431, 446) Cerebral and peripheral vascular dysfunction (54, 447–450) Arterial SMC loss and basement membrane defects (54, 250, 254, 429)
		Targeted mutation (K-in) into the endogenous Col4a1 murine gene: Col4a1G498V/+ (430)
		Conditional targeted mutation: Col4a1Flex41 bred to Rosa26-CreER mice (ubiquitous expression), Tie2-Cre mice (endothelial cells); Pdgfrb-Cre mice (mural cells) or Gfap-Cre mice (astrocytes) resulting in the expression of a mutant Col4a1 gene deleted of its exon 41 in the indicated cells (427).
PADMAL (Pontine Autosomal Dominant Microangiopathy with Leukoencephalopathy)	COL4A1	None reported yet.
CARASAL (Cathepsin-A-related arteriopathy with Strokes and Leukoencephalopathy)	CTSA	None reported yet.
Fabry disease (Anderson-Fabry disease)	GLA (alpha-galactosidase)	Targeted knockout of the Gla gene in the mouse (Gla−/−) (437)	Characteristic lipid deposits detected by electron microscopy; no overt clinical manifestations (437) except an age-dependent and distinct deficits of the sensory system (451).
CARASIL (Cerebral Autosomal Reccessive Arteriopathy with Subcortical Infarcts and Leukoencephalopathy)	HTRA1	Targeted knockout of the Htra1 gene in the mouse (Htra1−/−) (239)	Structural and functional abnormalities of pial arteries and brain capillaries with blood brain barrier leakage. No overt brain lesions, nor clinical manifestations (239).
HTRA1-related autosomal dominant cSVD	HTRA1	Targeted knockout of the Htra1 gene in the mouse (see above)
Early amnestic syndrome of the hippocampal type and leukoencephalopathy	LAMB1	Spontaneous non-sense mutation of the Lamb1 (lamb1t) leading to a premature stop codon, resulting in the deletion of the last 57 amino acids (452), a mutation very similar to the one described in patients.	Intermittent dystonic hindlimb movements and postures when awake, and hyperextension when asleep. Phenotype reported prior to the report of LAMB1-related cSVD (452).
NIT-1 related cSVD	NIT1	None reported yet.
Notch1-related microangiopathy	NOTCH1	None reported yet.
Early onset autosomal recessive NOTCH3-related cSVD	NOTCH3	Targeted knockout of the Notch3 gene in the mouse (Notch3#x2212/−) (453).	Prominent defects of vascular SMCs, especially in cerebral and retinal vessels with cerebrovascular dysfunction (54, 310, 311,454).
Leukoencephalopathy with calcifications and cysts (LCC) (Labrune syndrome)	SNORD118	Targeted inactivation of the U8-3 gene in zebrafish (455).	U8-3 zebrafish mutant is embryonic lethal with impaired development of the central nervous system and defective and disorganized vasculature of the trunk (455). Homozygous Snord118 KO mice in neurons dye by ~1 month of age. Lower body weight and smaller size and severe brain growth defects with decreased proliferation and increased death of neural progenitor cells (456).
		Conditional targeted mutation: Snord118flox/flox mice mated with neural-specific EMX1-Cre to induce deletion of Snord118 in neurons of the neocortex and hippocampus, and in the glial cells of the pallium (456).
RVCL-S (Retinal vasculopathy with Cerebral Leukoencephalopathy and Systemic manifestations)	TREX1	- Targeted mutation (K-in) into the endogenous Trex1 murine gene by which a frameshift mutation was introduced at amino acid 235 of TREX1 resulting in a truncated protein as observed in patients (Trex1fs235) (457). - TREX1-V235fs and TREX1-D272fs mice were generated by knocking in the murine trex1 gene a human TREX1 cDNA with the V235fs or the D272Vfs mutation (458).	Homozygous Trex1fs235 mice exhibit a reduced lifespan (457). Homozygous TREX1-V235fs and TREX1-D272fs mice show no major signs of retinal, cerebral or renal disease but exhibit striking elevations of autoantibodies in the serum (458). DNA damages but no overt phenotype in unchallenged mutant mice (459).
		Conditional mice: Expression of human TREX1 with the V235fs is inserted into the Gt(ROSA)26Sor (ROSA26) locus, blocked by an upstream loxP-flanked STOP sequence. When mated with Lys-Cre mice, which express the Cre recombinase in the myeloid cell lineage, the STOP sequence is deleted, resulting in mutant TREX1 expression in myeloid cells at various levels (459).

Mouse models of the COL4A1/2 hemorrhagic microangiopathy tick all the boxes in terms of clinical relevance and translatability. Indeed, Col4a1 mutant mice faithfully recapitulate the full spectrum of the human disease. Specifically, mutant mice develop, like patients, perinatal hemorrhages, porencephalic cavities, microbleeds or deep ICH, as well as kidney, eye or muscle defects (54, 250, 254, 427–431). Importantly, both the penetrance and expressivity of these manifestations are, as in patients, variable. In particular, using an allelic series of Col4a1 mutations, Jeanne and colleagues nicely demonstrated that the severity of the hemorrhagic phenotype was modulated by the position of the mutation, environmental factors (such as exercise) and the genetic background, findings which perfectly align with observations in patients (427).

In CADASIL, all types of genetic manipulations have been tried in the mouse, including transgenic models overexpressing the mutant NOTCH3 under the control of a SMC promoter or of the entire Notch3 promoter, conditional models or targeted mutation in the endogenous murine Notch3 gene (Table 8). In general, all these models (except the Notch3^R142C/+ mice which may have not been optimally assessed) develop the characteristic vascular GOM deposits or Notch3^ECD accumulation (Table 8), and the higher is the overexpression of mutant NOTCH3, the greater is the burden of the accumulation (231, 432). However, the TgNotch3^R169C model is the only one that exhibits both age-dependent structural and functional alterations of brain vessels as well as hemispheric WM lesions (231, 433–436). There are at least two possible explanations for this. First, the TgNotch3^R169C model has been the most extensively studied model and second, mutant mice overexpress the mutant protein about 4-fold over the endogenous NOTCH3, in contrast to the conditional and knock-in mice that express endogenous levels of NOTCH3 (433).

Results are more mixed for other cSVDs (Table 8). As a few examples, none of the Trex1 mutant mice show any of the major signs of retinal, cerebral or renal disease characteristic of the RVCL-S disease (Table 8). Gla knockout mice, considered as the gold standard model of the Fabry disease, exhibit only the characteristic lipid deposits (437). The HtrA1 knockout mice show some structural and functional abnormalities of brain vessels but no overt brain lesions, nor clinical manifestations of CARASIL (239). Also, despite the presence of major structural and functional abnormalities of brain vessels in Notch3KO mice neither overt brain lesions nor clinical manifestations have been reported in these mice, a finding which is in striking contrast with the extreme severity of the human disease (236).

One recurrent theme for these genetically based models is that phenotypic analysis may not have been optimal. Indeed, the vast majority of these models has not been examined thoroughly for the presence of WM lesions or lacunar infarcts using high resolution structural, diffusion MRI or appropriate histological techniques. Along this line, it is of particular interest that brain arterial pathology in CADASIL mouse models has been documented only very recently thanks to the use of thick (200- μm) brain sections that enables the imaging of long arterial segments unlike thin (< 10 μm) sections obtained with conventional histological techniques that under sample the vascular network (231). It is also worth mentioning that arterial pathology is present in TgNotch3^R169C mice crossed on the C57BL/6 background but absent when these mice are crossed on the FVB/N background, although Notch3^ECD accumulation remains the same (231).

The pros and then cons of genetic models.

With respect to the pros, firstly, these are clinically relevant models produced by a single mutation in a single gene that can be ascribed as “clean” models in contrast to all surgical models. Moreover, mutant and control lines differ only by the presence/absence of the mutation. In addition, the genetic background can be easily controlled. Secondly, when mutant animals display vascular changes or brain lesions, they provide a unique and unprecedented opportunity to elucidate the pathophysiological mechanisms, at the cellular, molecular and functional levels (see section 7). On the other hand (cons), these are models of “rare” cSVD in which vascular changes or brain lesions may have mechanisms distinct from the more common sporadic cSVD, although the recent finding that the boundaries between monogenic and polygenic cSVD are becoming increasingly blurred (see section 5.3.3) suggests the opposite. Secondly, except for the COL4A1/2 hemorrhagic microangiopathy, these models rarely recapitulate the full spectrum of the human disease. However, as discussed above, their phenotypic analysis may have been not optimal. All in all, the balance tips towards these genetically engineered animals as being relevant models for the study of cSVD.

7. Structural and functional alterations of small brain vessels in cSVD

In this section, we review the structural, mechanical and functional alterations of brain vessels associated with sporadic or monogenic cSVD. We also consider studies assessing the impact of aging and hypertension, the two major risk factors of cSVD.

7.1. Cellular and molecular features of vascular pathology

7.1.1. Arteries and arterioles

The perforating arteries affected by cSVD can now be visualized in humans using ultra-high-field imaging at 7T (16); however, in vivo assessment of the morphology of these vessels and pathological changes that accompany cSVD still remains out of reach using this technique. Therefore, neuropathological studies remain the reference standard. The most informative studies are those that have cut serial sections to reconstruct cerebral arteries from the cortical surface to the deep brain regions. In landmark studies published 50 years ago, Fisher, the same who pioneered the concept of lacune, performed serial sectioning of some 60 lacunar infarcts with several thousand 8–10 μm thick sections to reconstruct the vascular tree feeding the infarcted tissue, from the lacune up to the brain surface. He made the fundamental observation that the vast majority of lacunes resulted from the severe arterial stenosis or occlusion of penetrating branches from large cerebral arteries (middle cerebral, posterior cerebral or basilar arteries) supplying the infarcted territory (202, 460, 461). Specifically, Fisher reported two different vascular lesions: (1) segmental arterial disorganization and (2) microatheroma as the causal lesions of small (< 7mm) and larger (8–20 mmm) lacunes respectively. Segmental arterial disorganization (also known as lipohyalinosis, fibrinoid necrosis or angionecrosis with aneurysm formation) involved a limited portion (hence the term segmental) of small arteries of 40 to 200 μm diameter that could be focally enlarged; pathological features included the destruction of muscular and elastic components of the arterial wall that was replaced by collagen, infiltration of the wall by foam cells (fatty macrophages), red blood cells and sometimes the presence of fibrinoid material suggestive of an acute cellular destruction (202) (Figure 16). Microatheroma (or “lipid macrophage” plaque), ascribed to atherosclerosis, involved larger penetrating arteries of 300 to 900 μm diameter and was characterized by SMC loss that was replaced by a conglomerate of fat-filled macrophages leading to the severe stenosis or occlusion by a superimposed microthrombus (131, 460). These two types of lesions are nowadays rarely reported and there are two potential reasons that may explain this. First, such serial sectioning approach is no longer used in postmortem studies of the brain. Second, until the 1960s, there was almost no BP medication and the brains examined by Fisher were from patients with untreated longstanding hypertension. Therefore, it is plausible that lesions described by Fisher lie at the very severe end of the spectrum of cSVD pathology.

Figure 16. The arterial lesion underlying small lacunes according to CM Fisher.

Schematic drawn by CM Fisher showing the relation of a lacune in the pons to arterial lesion. Reproduced from (202). Used with permission from Springer-Verlag.

A serial sectioning approach has been also used to study cerebral medullary arteries of autopsy material (frontal lobes) from 7 middle-age (52–66 yo) and 5 aged (73–89 yo) patients with Binswanger’s disease in comparison with age-matched controls. About 700–1000 serial slices (5-μm thick) were cut to reconstruct medullary arteries from the penetrating site at the cortical surface to the distal level in the deep WM (462, 463) (Figure 17). The first prominent change was fibrosis of the intima, with or without atheroma, often in association with fibrin or plasma protein exudation; fibrosis was observed mainly in the intracortical and subcortical arterial segments. The second most noticeable abnormality, common to all patients, was the continuous or discontinuous loss of medial SMC, which was seen predominantly in the cortical and subcortical segments rather than in the distal segments in the deep WM. Most arteries with SMC loss presented irregular dilations and fibrosis of the media. The total extent of SMC loss was more severe in aged patients. A third change was fibrosis of the adventitia of intracerebral arteries, this was essentially restricted to the level of the WM and rare at the cortical level. The adventitia, which is the outermost layer of a vessel, is mainly composed of collagen fibers, rare elastic fibers and nerves, and both macrophages and fibroblasts. Whereas it is well-developed at the level of pial arteries, it is extremely thin in intracerebral arteries and nerves are absent (38, 464, 465). In addition, there was a significant thickening of the media of medullary arteries. Noting that severe stenosis or occlusion of these arteries were only occasionally detected. Eleven medullary arteries of the frontal lobe from a single CADASIL patient aged 75 years have been similarly reconstructed (466). Histopathological analysis showed essentially the same changes than those observed in Binswanger’s patients except that the loss of SMC was much more pronounced and more extended, affecting all arteries along almost their entire length from the cortical surface up to the distal end. Moreover, all pial arteries exhibited incomplete or complete loss of medial SMCs, which was more pronounced in small (< 100 μm) arteries, and often associated with segmental intimal fibrosis (466).

FIGURE 17. Reconstruction of 7 cerebral medullary arteries from patients with Binswanger’s encephalopathy.

Brain tissues were sectioned and examined by histology and cerebral medullary arteries were reconstructed from serial sections from the pial surface to the deep white matter. Reproduced from (462). Used with permission from Springer-verlag Berlin Heidelberg.

In the common neuropathological practice, postmortem brains undergo a much more limited sampling and lesions are assessed on vessel cross-sections. Brain arteriolosclerosis is regarded as the pathological hallmark of non-amyloid cSVD (11). According to the vascular cognitive impairment neuropathology guidelines (VCING), developed in 2016 by a group of neuropathologists, brain arteriolosclerosis is defined by the hyaline (glassy-looking acellular appearance) thickening of walls of vessels < 150 μm in diameter, not associated with lipid containing cells replacing the tunica media and an absence of intramural inflammation, amyloid or fibrinoid necrosis (467, 468). A semi-quantitative severity scale describes 4 categories of arteriolosclerosis including: (0) none, (1) mild, defined by mild thickening of the vessel media and mild fibrosis, (2) moderate, defined by partial loss of medial SMC and moderate hyaline fibrosis and (3) severe, defined by complete loss of medial SMC, severe hyaline fibrosis and lumen stenosis (468). According to this classification, loss of SMC is a key feature in the scoring of brain arteriolosclerosis severity. On the one hand, this classification offers the advantage to better categorize and standardize the reporting of arterial pathology in cSVDs. On the other hand, its inter-rater reliability assessment is moderate (467). In addition, the threshold of 150 μm for vessel diameter is questionable because arteriolosclerosis can be also observed in larger arteries (469). Moreover, it does not fully capture the complexity and diversity of vascular lesions as observed by serial sectioning nor does it take into account that severity of lesions can vary from one brain region to another. The “sclerotic index”, which is based on measurements of arteriolar wall thickness and luminal narrowing, is used by some groups to grade quantitatively the severity of brain arteriolosclerosis. Arteriolosclerosis is highly prevalent in autopsy specimens from individuals over 70 years of age, and its severity is significantly associated with the odds of WM degeneration, lacunes and subcortical microinfarcts (470–472). Hyalinosis of the media with prominent loss of SMCs and fibrosis of the arterial wall are common changes observed in hereditary cSVDs, although lesions and particularly SMC loss is usually much more severe in hereditary compared with sporadic cSVD (Table 6) (473).

Assessment of vascular pathologies in human brains remain mostly semi-quantitative and continues to be plagued by high variability and low confidence (474). Moreover, because of the use of 2D imaging and thin sections, these assessments often under-sample the vascular network and lack in-depth analyses of vascular cells. Finally, an inherent limitation of human studies, which are mostly performed post-mortem, is that they only monitor the end stage of the disease. Animal models of cSVD offer the opportunity to determine the spatiotemporal pathologies of vascular elements. Lesions of cerebral arteries, characterized by degeneration and loss of arterial SMCs, have been reported in mouse models of early onset autosomal recessive NOTCH3-related cSVD (Notch3KO), CADASIL (TgNotch3^R169C or Notch3^R170C/R170C when backcrossed on the C57BL/6 background) and COL4A1/2 hemorrhagic microangiopathy (Col4a1^G498V/+ or Col4a1^G1064D/+) (54, 231, 250, 310, 311). Interestingly, analysis of CADASIL mice showed that SMC defects were focal and segmental and equally affected pial, cortical and subcortical arteries (Figure 18) (231). In Col4a1 mutant mice, pathological analysis has shown that ICHs originate from arteries with reduced SMC coverage (250). In CADASIL and Notch3KO mice, arterial lesions occur prior to the appearance of lacunar infarcts and worsen with age. In CARASIL mouse model (Htra1^−/−), SMC defects have not been reported, but pial arteries exhibit a thickening of the subendothelial layer by the deposition of fibronectin, fibulin-5 and LTBP-4, a duplication/fragmentation of the elastica lamina, and an age-dependent progressive reduction in the expression of SMC contractile genes (239).

Figure 18. Focal and segmental loss of arterial SMCs in brain arteries of CADASIL (TgNotch3^R169C) mice.

A- Imaging of a 200- μm thick brain coronal section stained for α-SMA and perlecan enables the visualization of long arterial segments. B- Brain arteries stained for α-SMA and perlecan from a wildtype (WT) mouse aged 12 months and CADASIL (TgN3^R169C) mice aged 6 or 12 months show SMC defects characterized by discontinuous or notched α-SMA staining (yellow dotted lines) as well as gaps in α-SMA staining (brackets) in CADASIL mice. Scale bars: 20 μm (B). Reproduced from (231). Used with permission under CC-BY 4.0 license.

The retina is a developmental extension of the brain and its vascular network shows many similarities with the cerebral vasculature. Moreover, the superficial retinal vascular network is nearly planar on flat-mounted retina preparations and highly stereotyped, enabling robust quantification at cellular resolution in the mouse (54). Remarkably, the superficial retinal vascular network of mice completely lacking NOTCH3, mice with COL4A1 glycine-altering variants or Cysteine NOTCH3 variants exhibit arterial lesions strikingly similar to those observed in the cerebrovasculature (Figure 19) (54, 231, 250, 311, 475). Importantly, aging in wildtype mice is associated with similar progressive focal loss and degeneration of arterial SMCs (476). In summary, these histopathological analyses highlight the loss of SMCs in arteries as a common denominator in both sporadic and genetic cSVDs. Moreover, they indicate that the retina is a powerful model to assess vascular pathology in animal models of cSVD.

Figure 19. Focal and segmental loss of arterial SMCs in retinal arteries of CADASIL (TgNotch3^R169C) mice.

(A) Imaging of flat-mounted retina stained for α-SMA and perlecan enables the visualization of the entire vascular network. (B) Representative pictures of retinal arteries from WT and TgN3^R169C mice stained for α-SMA, perlecan, and MEF2C showing that arteries in mutant mice exhibit a discontinuous SMC coating at 6 months of age onwards (dotted yellow lines) that worsens with age, ultimately causing focal SMC gaps (brackets) and SMC loss. Scale bar: 10 μm (B). Reproduced from (231). Used with permission under CC-BY 4.0 license.

Although, we still have an incomplete understanding of the molecular mechanism of arterial SMC lesions in cSVD, recent studies have provided some clues. First, arterial lesions are seemingly much more severe in the retina than in the brain. Although, this might reflect an increased sensitivity of detection owing to the planar architecture of the retina preparation, which allows the entire arterial network to be analyzed at unprecedent high resolution, this might also be related to the fact that the retinal arterial network is subjected to higher arterial pressure (~80 mmHg) than that in the brain (<50 mmHg) (54). Hence, arterial lesions might be favored by a high transmural pressure. Second, transcriptomic analysis at the single cell level suggests that age-related arterial SMC loss might be attributable to a decline in NOTCH3 signaling (477). Molecular studies in Col4a1 mutant mice have recently shown an excess of transforming growth factor (TGF)-ß signaling in the brain of mutant mice. Notably, genetic reduction of TGFß1, an important TGFß ligand isoform, in Col4a1 mutant mice, prevented the loss of arterial SMC and reduced the burden of ICH, suggesting a causal-relationship between increased TGFß activity and arterial SMC loss in this disease (475). The mechanism by which COL4A1 mutations affect TGFß signaling is not yet fully understood but previous works have shown that components of the extracellular matrix play a role in the regulation of TGFß signaling (478). In CADASIL, recent work provided experimental evidence that Notch3^ECD aggregates contain both mutant and wildtype Notch3^ECD molecules and that arterial SMC loss is mostly driven by pathological accumulation of Notch3^ECD and not by a defective signaling of NOTCH3 receptor (231). Intriguingly, HTRA1, the protein product of the gene mutated in CARASIL, was found to abnormally accumulate in brain arteries from CADASIL patients and to colocalize with Notch3^ECD deposits. Moreover, proteomic analysis of brain vessels from CADASIL patients showed an increased levels of several HTRA1 substrates (240). These latter findings suggest the possibility that HTRA1, which is sequestered in Notch3^ECD deposits, progressively loses its activity to contribute to arterial pathology in CADASIL, but this awaits experimental confirmation.

7.1.2. ACT zone

Characterization of ACT zone pathology in cSVD is still rudimentary, owing to the fact that this vascular compartment has been recognized only recently (54). Nevertheless, recent works have revealed that the nature of lesions could differ between ischemic and hemorrhagic forms of cSVD, even though arteries and arterioles showed a loss of SMCs. In Col4a1 mutant mice (Col4a1^G498V/+ or Col4a1^G1064D/+), which recapitulate the COL4A1/2 hemorrhagic microangiopathy, the ACT zone is hypermuscularized, with an increased number of mural cells expressing a higher content of contractile proteins, and the ACT zone has more tone, likely as a consequence of an increased Notch3 activity. Interestingly, pathological analysis of postmortem brains from patients with sporadic hemorrhagic cSVDs similarly showed an excessive muscularization of the ACT zone. In contrast, Notch3KO mice, a model of ischemic cSVD, exhibit a loss of mural cells in the ACT zone (Figure 20) (54). This finding which remains to be confirmed in additional cSVD models, suggests that changes in the properties or density of mural cells in this segment could influence the ischemic versus hemorrhagic presentation of cSVDs.

Figure 20. Comparable loss of arterial SMCs but opposite changes of mural cells on the arteriole-capillary transition (ACT) zone between a hemorrhagic and an ischemic form of cSVD.

Schematic representation of brain vessels in the COL4A1/2 cSVD which is characterized by recurrent intracerebral hemorrhages (A) and in the NOTCH3 null-driven cSVD, which is characterized by ischemic infarcts (B). Arteries exhibit loss of SMCs in both the COL4A1/2 and the NOTCH3 disease. In contrast, the ACT zone shows an increased number of mural cells with higher contractile protein content in the COL4A1/2 cSVD but a loss of mural cells in the NOTCH3 cSVD.

7.1.3. Capillaries and veins

In postmortem human brains, aging is associated with a ~50% reduction in the density of pericytes in the cortex. In patients with vascular dementia, the density of capillary pericytes is preserved in the frontal cortex compared to age-matched controls but reduced in the frontal deep white matter, a region most frequently afflicted by cSVD (479, 480).

Rodents exhibit a small reduction in capillary length and pericyte number with aging in the deep cortical layers and in the WM (481, 482). The picture is more nuanced in mouse models of monogenic cSVD with a reduction in pericyte coverage in Htra1KO mice but a preserved pericyte density and/or coverage in Notch3KO mice, Col4a1 mutant mice and CADASIL mice (250, 299, 311).

Cerebral veins have been seldomly analyzed. This may be partly due to the difficulty in differentiating a vein from a highly disorganized arteriole with SMC loss using routine stain. A few studies have reported a thickening of veins (collagenosis) in the WM of cSVD patients (471, 472).

7.2. Arterial mechanics

The pulsatility index (the difference between the peak systolic flow and the minimum diastolic flow velocity, divided by the mean velocity), is thought to reflect arterial stiffness. It can be assessed in humans using ultra-high-field (7T) quantitative flow MRI (483). Two case-control studies showed an increased pulsatility index in arteries of the WM and the basal ganglia in patients with sporadic cSVD and in CADASIL patients (484, 485).

Arterial mechanics can be also assessed in experimental models. It is commonly studied at the level of pial arteries/arterioles, by assessing the pressure-diameter relationship in pressurized vessels in which myogenic tone is blocked (486). In general hypertension is associated with an increased stiffness of larger pial arteries but an increased distensibility in smaller arterioles that are quickly reversible with BP normalization (486, 487). Similar to the effect of hypertension, large pial arteries become stiffer with normal aging (488). Stiffening has been attributed to an increased deposition of fibrillar collagen and a reduction or fragmentation of elastin. Inward remodeling, which is characterized by a reduction in luminal diameter, is a unique feature of angiotensin II-induced hypertension that develops slowly and does not recover after BP normalization (486). Importantly, as mentioned above, a small reduction in lumen diameter can have a profound reduction in CBF, based on the fourth-power relationship between vessel radius and flow (7). Strikingly, pial arteries of CADASIL and CARASIL mice exhibit an early reduction in distensibility although in the context of normal BP. Moreover, stiffness is associated with an inward remodeling in CADASIL mice (239, 489).

7.3. Brain vessel dysfunction

In this section, we review how the distinct functions of brain vessels are disrupted in cSVD patients and rodent models. We start with the vasoreactivity to hypercapnia, a common challenge used in clinics to assess the ability of cerebral vessels to dilate, and continue with the four main functions of brain vessels (CBF autoregulation, neurovascular coupling, BBB and brain waste drainage). We review the timeline linking these functional alterations to brain lesions and clinical manifestations that are of great value for our understanding of the pathobiology. Using information gained from experimental studies, we discuss mechanisms by which cSVD risk factors (aging, hypertension and genetic mutations) disrupt these functions.

7.3.1. Vasoreactivity to hypercapnia

Cerebral vessels are highly sensitive to changes in PaC0₂. An increase in PaC0₂ (hypercapnia) produces vasodilation, an effect which is predominantly mediated by the extracellular acidosis resulting from hypercapnia relaxing SMCs (490–492). In clinics, hypercapnia is commonly used to assess the capacity of cerebral vessels to dilate by measuring CBF increases in response to breathing CO₂. Vasoreactivity to CO₂ has been investigated in large cross-sectional cohort studies including sporadic cSVD patients but no healthy controls. A reduced CO₂ response in the WM or subcortical gray matter is associated with a higher burden of WMHs, lacunes, MCB, enlarged PVS, brain atrophy and cognitive decline (493, 494). In CADASIL patients, CO₂ reactivity is reduced in WMH compared to normal appearing WM, although CO₂ reactivity is comparable between patients and controls in the grey matter and in the normal appearing WM (484). A longitudinal study conducted in patients with age-related WMHs showed that regions of normal-appearing WM at baseline that had progressed to WMHs one year later had a lower response to hypercapnia compared with normal-appearing WM (495).

Although hypercapnia primarily dilates arteries and arterioles (496), a 2–5 minutes hypercapnic challenge, as is performed in clinics, is likely to be associated with a passive dilation of the entire vascular bed including capillaries and veins. Therefore, a reduced CO₂ response may result either from a reduced capacity of the vascular bed to dilate or from a reduction in the volume of the vascular bed, without any assumption on the identity of the affected vascular compartment(s).

7.3.2. CBF autoregulation

A few studies, including a recent case-control study involving 113 cSVD patients and 83 controls, have shown that cSVD was associated with bilateral alteration of dynamic autoregulation and that the degree of impairment was positively associated with the burden of cSVD MRI markers (497). The common assumption is that chronic hypertension impairs CBF autoregulation (static or dynamic) with a right shift of lower and upper levels towards higher pressures. However, in a recent review of studies conducted during the last 40 years, involving > 700 hypertensive patients (from young to old ages), Classeen and colleagues did not find evidence of impaired autoregulation by chronic hypertension (7). Yet, in subjects aged over 70 years of age with hypertension, but without symptomatic cSVD, intensive BP lowering resulted in increased CBF, suggesting hypertension in older people shifts the autoregulatory CBF curve rightward and downward and is reversible with BP lowering (498). This picture was replicated in the CBF sub-study of the Systolic Blood Pressure Intervention Trial (SPRINT) randomized controlled trial comparing intensive with standard BP lowering in hypertensive individuals without symptomatic cSVD (499). On the other hand, in the PRESERVE clinical trial in patients with moderate to severe symptomatic cSVD (defined by symptomatic lacunar infarcts and confluent WM lesions), a similar BP lowering regimen did not reduce CBF over 3 months of treatment. This suggests that once cSVD had developed, autoregulation could not be reset, but was nevertheless sufficiently preserved to prevent any further significant CBF drop with reduction of BP (500).

Myogenic tone, the SMC response predominantly implicated in CBF autoregulation, has been assessed in pial or parenchymal arteries in hypertensive mice and genetic models of cSVD. In young hypertensive mice, myogenic tone of pial arteries is increased and the autoregulation curve is right-shifted towards higher BP. In hypertensive BPH/2 mice, increased myogenic tone results from a diminished activity of the large-conductance Ca²⁺-activated K⁺ channels, which are key vasodilatory ion channels of cerebral arterial SMCs (501). However, the increase in myogenic tone is lost in aged hypertensive mice (502). In Notch3KO, CADASIL (TgNotch3^R169C) and Col4a1 mutant (G1344D or G394V) mice, myogenic tone is reduced (76, 448, 449, 503). Myogenic tone reduction is associated with an extreme narrowing of the autoregulated range in Notch3KO mice and a rightward-shift of the lower limit of CBF autoregulation towards higher BP in CADASIL mice (310, 433). In Notch3KO mice, altered myogenic responses result from a defective contractile apparatus (503). In Col4a1 mutant mice, myogenic tone is lost because of a reduced activity of TRPM4 channel activity in arterial SMCs, however through a mechanism which differs according to the mutation. Col4a1 mutations impair the folding of collagen IV and its trafficking through the endoplasmic/sarcoplasmic reticulum, and secretion of collagen IV to the basement membranes is reduced. In Col4a1^+/G394V mice, there is an hyperactivity of TGF-β signaling pathways, possibly as a consequence of basement membrane abnormalities, that stimulates phosphoinositide 3-kinase (PI3K) to deplete phospholipid phosphatidylinositol 4,5 bisphosphate (PIP2), an activator of TRPM4 channels (448). In Col4a1^+/G1344D mice, there is a disruption of sarcoplasmic reticulum Ca²⁺ signaling, likely because of an endoplasmic/sarcoplasmic reticulum stress, that blunts activity of large-conductance Ca²⁺-activated K⁺ channels and transient receptor potential melastatin 4 (TRPM4) cation channels. Accordingly, treating Col4a1^+/G1344D mice using the chemical chaperone, 4-phenylbutyrate to restore the trafficking of misfolded proteins, restores myogenic tone (449). In CADASIL mice, myogenic tone is lost because of an increased activity of voltage gated K⁺ (K_V) channels in arterial SMCs. This defect results from an excess of TIMP3 (tissue inhibitor of metalloproteinase 3) protein, which abnormally accumulates in Notch3^ECD deposits, blunting the activity of the ADAM17/HB-EGF/(ErbB1/ErbB4) pathway, and resulting in an increased density of K_V channels (435, 504). In summary, since myogenic response is essentially inherent to arterial SMCs, these experimental studies point to an early dysfunction of SMCs in cSVD. Myogenic responses are important for the development of basal tone upon which other mechanisms exert vasoconstrictor and vasodilator influences (505), and thus play an important role in the regulation of CBF, which goes largely beyond the adaptation of CBF to changes in BP (autoregulation). Therefore, additional experimental studies are needed to better understand how defective myogenic responses may affect brain perfusion and function globally, and more importantly, in deep brain regions which are those predominantly affected in cSVD.

7.3.3. Neurovascular coupling

It has been proposed that CO₂ generated by neuronal metabolism contributes to neurovascular coupling (506). Consequently, neurovascular coupling is sometimes confused with vasoreactivity to CO₂. However, a recent experimental study provided strong evidence that hypercapnic challenge interrogates vascular compartments and regulations that are distinct from those elicited during neurovascular coupling (496). There are few studies on neurovascular coupling in cSVD patients and only those using a visual or motor stimulus and functional MRI to measure CBF changes will be reviewed herein, since those employing cognitive tasks add another layer of complexity to interpret the data (507). Two independent groups, using different imaging modalities and paradigms of visual stimulation, showed significant alterations in neurovascular coupling in CADASIL patients compared to age-matched healthy controls. One study showed a reduction in the amplitude of the hemodynamic response whereas the second one showed a time-shifted decrease in the hemodynamic response (484, 508). There are many potential confounders to take into account when interpreting neurovascular coupling studies in human. Among the most important are the intake of medications and the presence of brain lesions that could affect blood flow responses by reducing the neural response. By simultaneously recording CBF and neural activity, using cortical electroencephalogram, Huneau and colleagues showed that the altered blood flow response in CADASIL patients was not related to a reduction in the neural response but instead reflected a decrease in neurovascular coupling efficiency originating in the vascular bed (508).

Neurovascular coupling has been extensively studied in rodents upon aging or chronic hypertension and in mouse models of genetic cSVD. All these studies have similarly pointed to neurovascular coupling deficiencies (Figure 21). By simultaneous recording of vascular parameters and neural activity using two-photon imaging, Cai et al recently provided convincing evidence that neurovascular coupling deficiency upon aging was caused by an age-dependent decrease in the ability of mural cells to relax rather than by reduced neuronal activity (509). The reduced vasodilator responses were explained in part by an increased stiffness and a reduction in mural cell coverage, although the number of mural cells was preserved (Figure 21B) (509). This was associated with a decline in vessel density especially in proximity to the penetrating artery (509). Impaired neurovascular coupling has been also reported in AngII-induced hypertension and BPH/2 hypertensive mouse models (426, 510) as well as in models of monogenic cSVD including CADASIL (TgNotch3^R169C mice on the FVB/N background) (434) and collagen type IV-cSVD (Col4a1^G394V/+ mice) (447); noting that there was overt SMC pathology in these genetic models. Two different mechanisms have been described in the hypertensive models. One mechanism involves the activation of angiotensin type 1 receptors in PVMs leading to the production of reactive oxygen species, which impairs endothelium-dependent responses by reducing bioavailability of NO (Figure 21C) (426). A second mechanism involves defective capillary-to-arteriole signaling caused by a diminished activity of the capillary endothelial cell inward-rectifier K+ channel, Kir2.1 (510). Remarkably, neurovascular coupling deficit in Col4a1 and CADASIL mutant mice also results from altered capillary-to-arteriole signaling as a consequence of diminished capillary endothelial cell Kir2.1 channel activity (447, 511). Whereas the fundamental defect underlying this channelopathy – depletion of the minor membrane PIP2, a key activator of the Kir2.1 channel – is similar in Col4a1 and CADASIL mutant mice, the intrinsic mechanisms differ. In CADASIL, the defect is attributable to the pathological accumulation of TIMP3 in the endothelial-pericyte–shared basement membrane, which drives a decrease in the synthesis of PIP2 in capillary endothelial cells (Figure 21D) (229, 434, 504, 511, 512). In collagen IV disease, the defect is due to overactivity of TGF-β receptors, possibly as a result of collagen IV misfolding in the basement membrane, and the consequent increase in the activity of phosphatidylinositol-3-kinase (PI3K), the enzyme that converts PIP2 to PIP3, which leads to a reduction in PIP2 bioavailability (Figure 21E) (447). Remarkably, in these two monogenic cSVDs, there is a mechanistic link between a change in a component of vascular basement membranes – increased TIMP3 in CADASIL and reduced COL4A1/A2 in collagen IV disease – and pathogenic alterations of the same endothelial ion channel.

Figure 21. Distinct and shared cellular and molecular mechanisms involved in neurovascular coupling defects in sporadic and genetic cSVDs.

During aging (B), there is an increased arterial stiffness and a reduction in arterial SMC coverage crippling arterial dilation. In angiotensin II-dependent hypertension, one mechanism (C) involves an increased production of reactive oxygen species (ROS) by perivascular macrophages (PVM) that reduces the bioavailability of nitric oxide (NO) leading to a reduction in vasodilation. In CADASIL (D), TIMP3 accumulation in the pericyte-endothelial cell basement membranes causes a decrease in ATP-dependent synthesis of PIP2 leading to a reduction in Kir2.1 activity and crippling the endothelial cell (EC) retrograde hyperpolarization. In the COL4A1/2 cSVD (E), NVC is also caused by a reduction in Kir2.1 activity but this arises from a reduced PIP2 availability as a consequence of abnormal basement membranes (dashed grey line below SMCs).

Unlike human studies, studies using model animals enable rescue experiments to assess causal relationships between functional deficits and disease manifestations. Restoration of functional hyperemia in hypertensive mice by depletion of PVMs and in Col4a1 mutant mice by chronic inhibition of PI3K improves memory deficits (426, 447). Although these findings support the hypothesis that a chronic reduction in neurovascular coupling could account for cognitive deficits, further studies are needed to substantiate this relationship and rule out possible confounding effects of specific experimental maneuvers (e.g., PVM depletion or chronic inhibition of PI3K), which may have additional effects on the brain or brain vessels. On the other hand, restoration of functional hyperemia in CADASIL mice by passive immunization against NOTCH3 did not ameliorate WM lesions (435).

It should be stressed that the vast majority of these experimental studies have been performed in anesthetized mice and were restricted to the somatosensory cortex. Anesthesia has major effects on brain and cardiovascular physiological processes. Recent in vivo imaging of neurovascular coupling in awake mice found that aging caused delays in neurovascular responses rather than a reduced amplitude of responses (482). Therefore, it will be important to determine how changes in neurovascular coupling in the disease models highlighted above manifest under more physiological conditions, such as in freely moving awake mice, and in other brain regions, including deep regions.

7.3.4. Blood brain barrier

Historically, BBB integrity has been initially examined on postmortem brains by immunohistochemical staining for plasma proteins, like albumin, fibrinogen or immunoglobulins. It was found that plasma extravasation was common in human aged brains but did not seem to be associated with cSVD lesions (513, 514). However, such pathological studies are poorly quantitative and can be performed only at the end-stage of the disease. An increase in the CSF/serum albumin ratio in patients with cSVD has been interpreted as an indication of BBB dysfunction. However, this ratio rather assesses functionality of the blood-to-CSF barrier, which is functionally related to the choroid plexus and is influenced by the flow rate of the CSF (515). There are at least two MRI techniques that can quantitatively assess BBB leakage non-invasively in humans. BBB leakage associated with cSVD is expected to be more subtle than the gross disruption associated with brain tumors or large infarcts. The most widely used technique is dynamic contrast-enhanced MRI (DCE-MRI) that employs a small (~550 Da) gadolinium-based contrast agent injected intravenously. If the BBB is disrupted, the contrast agent extravasates from the blood into the brain and accumulates in the extracellular extravascular space, causing a change in the MRI signal intensity (Figure 22).

Figure 22. Imaging blood brain barrier leakage in human using dynamic contrast enhanced MRI.

Hotspots of increased leakage are shown in red. White matter hyperintensities are shown in green. The image on the right shows a close up of hotspots in the vicinity of the posterior horn of the right lateral ventricle.

DCE-MRI can quantify the BBB leakage rate, the spatial distribution of the leaking brain and the BBB hotspot volume fraction, using appropriate pharmacokinetic modeling (516). Prolonged imaging over a period of 20–30 minute is used to allow detection of subtle BBB leakage (517). Another promising MRI approach measures the BBB permeability to water using arterial spin labeling (ASL). This method does not require the injection of contrast agents, instead the arterial blood water is “labeled” non-invasively by radiofrequency pulses before it enters the brain. In this method, appropriate modeling quantifies the transfer rate of labeled water into the brain, which reflects the combination of passive transport of water across the BBB and the active transport of water from the perivascular space to the interstitial space through the aquaporin-4 channels. This later method should be more sensitive than DCE-MRI because water is smaller than gadolinium-based contrast agent (18 Da versus 550 Da) but interpretation of the output measure can be more complex because it reflects the combination of 2 different transports of water from the blood to the brain extravascular and extracellular spaces (518, 519).

An increasing number of studies points towards BBB dysfunction in cSVD patients. First, case-control cross-sectional studies, using DCE-MRI, have shown a widespread BBB disruption, characterized by an increased rate of leakage, higher leakage volumes or the presence of hotspot leakages, within both normal-appearing WM and WMHs, in patients with cSVD relative to age-matched controls (Figure 22) (517, 520–523). However, there are variations in the findings between studies. As an example, Zhang et al found that the rate of leakage was unchanged in patients and that higher leakage volumes were observed mainly in the range of low leakage rates, in the order of noise ratio, suggesting that BBB leakage in these cSVD patients was subtle (521). In contrast, Walsh et al reported an increased rate of leakage and hotspots of leakage in their patients (522). Whether these variations result from differences in populations studied or methodologies is unclear. Moreover, DCE-MRI and ASL-MRI gave different results in CADASIL patients. Whereas two studies did not find evidence of increased BBB permeability to gadolinium-based contrast agent (522, 524), two different groups reported diffuse alteration in the water exchange rate across the BBB, using ASL-MRI (524–526). This may result from a lack of sensitivity in the first study or a selective increase in the permeability to water in CADASIL patients. Second, cohort studies (ie studies in which all participants are patients) have shown an association between increased BBB permeability and the burden of cSVD lesions (527–529). Interestingly, hotspot BBB leakage could be detected at the edge of lacunes, MCB, WMH or within WMH (530). Third, a link between BBB leakage at baseline and the loss of microstructural integrity over time has been found in longitudinal studies (531, 532). However, it is important to bear in mind that in vivo assessment of BBB permeability is plagued by a number of confounding factors which may have a large effect especially when dealing with subtle BBB leakage. Among these are changes in blood flow rates or the surface of vascular area (that can both decreases with aging and the progression of disease) which may alter interpretation of tracer kinetic results. Another important factor, which is often ignored is the presence of recent microinfarcts (<1mm), which are undetectable by MRI, that could be the cause of BBB leakage spots and not the consequence. Indeed, postmortem studies have shown that subcortical microinfarcts, which can be less than 100 μm in size, are extremely frequent in CADASIL patients (299).

Studies in rodents have provided contrasting results. BBB integrity in the mouse was investigated in a longitudinal, intravital, two-photon microscopy study, which demonstrated increased permeability to exogenous, small (3 kDa) and large (40 kDa) tracers in aged (24-month-old) mice compared with young (5-month old) mice (533). It has also been reported that AngII-induced hypertension in the mouse enhances BBB permeability by reducing endothelial tight junctions and increasing transcytosis, mainly in arterioles and venules. The mechanism underlying this enhanced BBB permeability primarily involves cooperative interactions of Ang II type-1 receptor-expressing endothelial cells with PVMs (534). In contrast, the BBB was reported to be preserved in the deoxycorticosterone acetate salt hypertensive model, in which circulating AngII is suppressed (535). Moreover, neither small (~1 kDa) nor large (70 kDa) exogenous tracers injected intravenously appeared to leak into the brain of adult Col4a1 mutant, outside regions with ICH, CADASIL or CARASIL mice, despite a decrease in pericyte coverage in the latter model, suggesting that the BBB is globally preserved in these mouse models of genetic cSVDs (239, 250, 299). It should be noted, however, that permeability to water has not yet been assessed in these models.

7.3.5. Brain waste drainage

PVS are central in the glymphatic system and drainage of waste materials from the brain (536). An increase in the number of PVS, especially in the basal ganglia and the WM, is strongly associated with cSVD (537). Moreover, pathological analysis of postmortem brain tissues from CADASIL patients suggest that WMHs in the temporal poles, a key feature of CADASIL, are related to the presence of numerous enlarged PVS (538). Hence, it has been suggested, but not yet proven, that dilation of PVS, reflects a dysfunction of the glymphatic system in cSVD.

Assessment of the glymphatic function in human remains in an early phase of development. To date, the gold-standard approach relies upon the injection of a small contrast agent intrathecally (via a lumbar puncture), followed by serial T1-weighted MRIs to monitor qualitatively and quantitatively the entry of the contrast agent into the brain and its clearance out of the cranium over a period of a few hours up to 2 days (539). There are several important points to consider with this approach. First, this is an invasive approach and there are large inter-individual differences in the CSF to blood clearance of the tracer (540). Second, entry of the tracer from the periarterial spaces and CSF-ISF exchange occur mostly in the cortex and to some extent in the immediate subcortical WM but is almost negligible in the deep WM. Third, after intrathecal injection, a significant % of the tracer (up to 75%) is resorbed at the level of the spinal canal and thus does not reach the brain (541). In theory, the clearance phase of a tracer could be investigated after an intravenous rather than an intrathecal injection, however, interpretation of the measured signal remains yet extremely challenging to model because of the multiple entry points of the intravenously injected tracer (541). To date, there are no human studies in cSVD using intrathecally injected tracer. Alternative methods aim at measuring the flow of CSF or ISF non-invasively using DTI or ASL-MRI. One of this approach claims to isolate and measure water diffusivity within PVS running in parallel to deep medullary veins in the deep WM and has developed an index (the DTI-ALPS index) which is supposed to represent the CSF efflux function in the deep WM (542).

A growing number of studies have assessed the ALPS-index in patients with sporadic or monogenic cSVD in case-control or cohort studies and have reported an association between this index and the presence and severity of cSVD MRI markers and cognition (543–546). However, the validity of the DTI-ALPS index as a measure of glymphatic function remains uncertain. One concern is that any association with cSVD and its clinical features may merely reflect an association with DTI metrics such as mean diffusivity from which the DTI-ALPS is derived; and we know that these conventional metrics are abnormal in cSVD and associate strongly with impaired cognition (541, 547). A recent study in cSVD examined this and showed that controlling for mean diffusivity weakened but did not abolish, associations between DTI-ALPS and future dementia risk in cSVD (548).

In contrast, there are several available techniques to assess the glymphatic system in rodents, ex vivo or in vivo, from the microscopic to the mesoscopic level (10). These methods rely on the injection of tracers (fluorescent tracers, microbeads or contrast agents) into the cisterna magna or directly into the brain, to assess glymphatic influx and efflux respectively, and involve the use of two-photon microscopy, transcranial macroscopic epifluorescence imaging, MRI, PET, SPECT imaging or histological methods (10).

The glymphatic system has been investigated in the context of aging, acute or chronic hypertension in rodents. In the mouse, postmortem histological analyses and in vivo 2-photon microscopy imaging have revealed a reduction in paravascular penetration of tracers into the brain in middle-age (10–12 mo) and old (18–24 mo) compared to young (2–3 mo) mice, indicative of a reduction in the glymphatic influx. In addition, the clearance of intra-parenchymally injected tracers was similarly reduced in middle age and old mice, indicative of a reduction in the glymphatic efflux (549, 550). At least three potential factors responsible for the age-related decline in glymphatic function have been identified. First, perivascular pumping of CSF is likely reduced as a consequence of impaired cerebral arterial pulsatility with aging. Second, aquaporin-4 expression is normally highly polarized to astrocytic endfoot processes; this polarization of aquaporin-4 is lost in old mice especially around penetrating arterioles (549). Third, in old mice there is a decrease in the diameter and coverage of lymphatic vessels in the dura mater. Importantly, both CSF influx and efflux can be ameliorated by treating aged mice with vascular endothelial growth factor C, which increases lymphatic vessel diameter (550). Using fluorescent microparticles injected into the cisterna magna, and in vivo particle tracking by high-speed two-photon microscopy, Mestre et al, showed that acute hypertension, induced by the injection of angiotensin II, strongly reduced CSF flow speed by 40% through the PVS. This reduction was attributed to the increased backflow likely caused by hypertension-driven changes in the arterial wall motion (110). Effect of chronic hypertension on the glymphatic system was investigated in spontaneously hypertensive rats (SHR). Dynamic contrast-enhanced MRI was used to image the glymphatic transport of paramagnetic contrast agent injected into the cisterna in SHR. Quantitative analysis showed a reduction in both glymphatic influx and efflux in SHR compared with normotensive Wistar Kyoto rat rats (551). However, it is difficult to draw definite conclusions about the relationship between the alteration of glymphatic function and hypertension in this rat strain because SHR have normal-pressure hydrocephalus unrelated to hypertension that may affect CSF flow dynamics (414). Interestingly, the same technique, although with a different computational modeling approach, was applied to the spontaneously hypertensive stroke prone rat, another hypertensive rat strain which does not have hydrocephalus, and similarly showed decreased glymphatic influx. Further analyses suggest that an altered aquaporin-4 polarization may account for this defect (413). Among mouse models of monogenic cSVDs, only Notch3 deficient mice have been examined yet. Blood clearance of a fluorescent tracer injected into the cisterna magna was strongly reduced in Notch3KO mice compared to age-matched wildtype mice, indicating an alteration of the glymphatic system, although it is unclear whether it affects the influx or the efflux (477). There are at least two factors potentially responsible for this defect, first, a strongly reduced contractility of cerebral arteries, which is expected to reduce the CSF pumping into the brain and second a strong reduction in the number of PVMs (61, 310, 503). In summary, additional studies of the glymphatic clearance in both patients with cSVD and experimental models are worth considering.

8. How does cSVD causes brain lesions?

Despite the wealth of information gained from structural and functional studies of brain vessels in both patients and experimental models during the past years, we are still left with outstanding issues. In the following section, we discuss potential mechanisms of lacunar infarcts, WM lesions and hemorrhages and elaborate on which vascular /perivascular cell types are involved.

8.1. Ischemic infarcts and white matter hyperintensities

8.1.1. Ischemic infarcts

The landscape of infarcts observed in cSVD goes beyond the classical 3– 20 mm lacunar infarcts described by Marie in 1901 and Fisher in 1965. Studies during the past 15 years have revealed that it also includes cerebral microinfarcts (CMI), which have reported sizes between 50 μm and a few mm and thus are not visible to the naked eyes on gross pathology (186, 552, 553). CMI are detected by microscopic examination of post-mortem brain sections, which is the reference method but which largely underestimates the number of CMI since tissue sampling examines only a tiny fraction of the overall brain (554, 555). Although most CMIs remain invisible on MRI, larger CMIs (0.5–4 mm) can be detected using structural MRI, most often in the cortex because CMI are difficult to distinguish from other lesions in the WM (186). CMIs were first described on high field 7T MRI, but have more recently been described on 3T MRI and even 1.5T (186). Small asymptomatic infarcts can also be detected as hyperintense lesions on diffusion-weighted imaging (DWI) throughout the brain (i.e. incidental small DWI-positive lesions or ISDPLs) but DWI, only captures acute lesions since the signal typically disappears within 2–3 weeks (159). It is estimated that MRI detects less than 1% of CMI (159, 552). The first key point about CMI is that they are in very large number in cSVD brains, likely in the range of hundreds to thousands, and thus certainly outnumber the number of grossly visible lacunar infarcts (553), and with a high yearly incidence (556). Second, CMIs are present not only in deep brain regions, like the classical lacunar infarcts, but also in the cortex, indicating that cSVD pathological processes are more diffuse than initially thought (557, 558).

The seminal histopathological study of lacunes by Fisher demonstrated that lacunar infarcts can result from severe stenosis or occlusion of penetrating arteries/arterioles supplying the infarcted territory (202). Accordingly, the size of the infarct was thought to be related to the location (proximal versus distal) of the occlusion of the penetrating artery. Additional studies are consistent with an arterial origin of infarcts in cSVD. First, severe arterial stenosis or occlusion have been reported, although only occasionally, in monogenic cSVDs (Table 6). Secondly, a 3D reconstruction of 88 lacunes from 57 CADASIL patients using 3D-MRI showed that lacunes tend to align with the orientation of perforating arteries (559). Thirdly, utilizing the high resolution of 7T MRI the individual perforating artery supplying a lacunar infarct could be identified (560). Fourthly, a recent study combining histopathological and MRI analysis of single thick human brain slices showed that the location of CMI within the cortex was compatible with the layer-specific arterial supply of the human cerebral cortex (28, 561). Consistent with these human data, experimental occlusion in the rat of a single penetrating arteriole near the cortical surface produced a highly localized cylindrical region of infarcted tissue, the volume of which was correlated with the red blood cells volume flux passing through this arteriole prior to the occlusion (386). Remarkably, occlusion of a single penetrating venule generated a microinfarction similar to the one caused by the occlusion of a single penetrating arteriole. This later observation raises the interesting possibility that infarcts in cSVD may also have a venous origin (386).

It is well-established that thromboembolism plays a major role in large-artery stroke and cardioembolic stroke. Yet, whether in situ/local thrombosis initiates cSVD infarcts remains an outstanding unresolved question. First, local thrombosis is rarely reported in neuropathologic assessment of brain cSVDs, including the severe monogenic forms. This could mean thrombosis does not play an important role in lacunar stroke, but it is also possible that, contrary to large-artery stroke or cardioembolic stroke, a lacunar stroke rarely leads to early death of the patient and thus neuropathological studies of cSVD brains are usually performed several months or years after acute ischemic events by which time spontaneous resorption of the thrombus has occurred. Secondly, the efficacy of thrombolytic therapies in the acute treatment of lacunar stroke remains controversial as well as the efficacy of antiplatelets antiplatelet agents in secondary prevention (see section 10.1). Thirdly, a recent MR study found a highly significant association between genetic predisposition to venous thrombosis and both large-artery stroke and cardioembolic stroke but, notably, no association with lacunar ischemic stroke (562), suggesting thrombotic factors are not important for lacunar stroke. However, one limitation of this later study is that it did not capture all aspects of the thrombotic process and in particular local thrombosis of diseased small vessels. For example, it is well-known that, at site of vessel lesion, endothelial cells shift from an anti-coagulant to a procoagulant/prothrombotic phenotype. Here, platelets adhere to the exposed extracellular matrix, followed by platelet-platelet interactions to form a clot (563). Also, in diseased arterial or venous walls, increased collagen, which is the most potent activator of platelet adhesion and aggregation, might also promote thrombosis (564).

A second mechanism, which could occur with or without local thrombosis to cause infarcts, is altered hemodynamics. First, CBF decreases by 0.5% per year in humans (7). Moreover, many studies using a wide variety of methods to measure perfusion, including positron emission tomography (PET) (565), Xenon computed tomography (566) and MRI (567) have reported reduced CBF in cSVD patients compared to age matched controls, although proving this plays a causal role in disease pathogenesis, rather than being secondary to tissue damage with reduced demand resulting in reduced perfusion secondary to neurovascular coupling is more difficult to prove. Studies have uniformly showed CBF reduction in the subcortical structures including the WM, while there have been differing reports as to whether cortical CBF is also reduced (565, 567, 568), perhaps suggesting that this may only occur later in the disease course. In a cross-sectional study in patients with lacunar stroke and confluent WMH, CBF was reduced in the periventricular WM, not only within WMH but also in normal appearing WM, consistent with hypoperfusion playing a role earlier in the disease course (569). However there is limited longitudinal data examining whether reduced CBF precedes development of WMH (568). Determining whether there is reduced CBF at the time of lacunar infarction is challenging due to the spatial resolution required to detect small regions of hypoperfusion which is at the limit of CBF imaging techniques used in human. However, 52% to 76% of patients with single subcortical infarct had cerebral perfusion deficits within 24 h of symptom onset as shown using an MRI (570). Using DCE-MRI, patients with hypoperfusion patterns had the highest rate of early neurological deterioration (571). Moreover, widespread reduction in CBF has been documented in cSVD models, including Notch3KO, Htra1KO and CADASIL mice (239, 433, 477). It has been proposed that hypoperfusion could impede the washout of thrombus/emboli, especially in low perfusion territories (572). Furthermore, in vivo two-photon imaging in the mouse subjected to cerebral hypoperfusion induced surgically by bilateral common carotid stenosis showed that cerebral hypoperfusion was associated with increased rolling and adhesion of leukocytes in capillaries and pial vessels with a more prominent effect in venules (573). Hence, leukocyte plugging, favored by in situ vessel damages, could be exacerbated by cerebral hypoperfusion. Secondly, there are a few case reports of bilateral internal border-zone infarctions in CADASIL patients, in the absence of carotid stenosis or detectable embolic source; a number of these have been in patients with concurrent COVID-19 which perhaps contributes to an additional prothrombotic state (574–576). External (cortical) or internal (subcortical) border zone infarcts are ischemic lesions that occur at characteristic locations at the junction between two main arterial territories. The most frequent internal border-zone infarcts are located at the lenticulostriate–middle cerebral artery border zone, which is supplied by the distal branches of perforating lenticulostriate arteries and deep medullary arteries from the middle cerebral artery. On brain MRI, they have a rosary like pattern, parallel to the lateral ventricle in the centrum semiovale or corona radiata. Bilateral internal border-zone infarcts are most commonly caused by hemodynamic impairment, i.e. reduced CBF or reduced vasodilatory capacity. The current view is that decreased perfusion in these border-zone vascular territories, which can be provoked by hypotension or altered CBF autoregulation, make them vulnerable to infarction (577). Importantly, the notion that some of the border-zone infarctions that occurred in CADASIL patients were clearly associated with low BP or hypovolemia supports the possibility of a hemodynamic component (298, 578, 579).

Another putative mechanism of ischemic infarct involves a lowered threshold of ischemia. Studies of middle cerebral artery occlusion in animal have shown that the threshold of CBF below which brain tissue is at risk of infarction is around 40% of normal, and the threshold of CBF below which the tissue is irreversibly damaged is around 20% (580). However, in the context of cSVD, the threshold that may lead to impaired neuronal function and loss/dysfunction of more vulnerable cell types (specific neuronal subtypes, pericytes, oligodendrocytes) could be well above 40%. For example, degenerative changes in the WM and the cortex can be seen in the bilateral common carotid artery stenosis hypoperfusion models in which CBF (measured by ASL- MRI) typically decreases to around 50% from baseline after surgery and then gradually recovers to ~70% (375). A recent experimental study revealed that occlusion of the middle cerebral artery caused larger infarcts in two distinct CADASIL mouse models compared to control mice because mutant mice had a higher viability threshold; in other words, mutant brain tissue required a much higher residual CBF to survive. Further studies suggest that this lower threshold of ischemia is caused by an enhanced susceptibility to spreading depolarization, which is believed to exacerbate neuronal injury through prolonged ionic breakdown, as a consequence of altered extracellular potassium homeostasis (581).

8.1.2. White matter hyperintensities

Observational studies in patients with sporadic or genetic cSVD or community-dwelling participants have shown that the total volume of WMH increases over time (582, 583). DTI, which measures diffusion properties of water molecules in tissues, can detect microstructural alterations of WM tracts before these become visible on conventional structural MRI (584). The typical pattern in patients with cSVD is a reduction in fractional anisotropy (FA, representing the degree of directionality of water molecule diffusion) and an increase in mean diffusivity (MD, representing the magnitude of water diffusion in all directions) (Croall et al. 2017). DTI changes are detected in tissue surrounding WMH, an observation which has led to the concept of a WMH penumbra to define WM tissue at risk of turning into a more severe lesions, by analogy with the penumbra around a large artery infarct core (585). Longitudinal studies combining conventional MRI with DTI have revealed that impaired microstructural integrity precedes conversion into WMH, declines continuously over time (586), and predicts future conversion to dementia (587).

cSVD driven WMH are heterogeneous and have different pathological substrates (588). Neuropathological studies, aided by post-mortem MRI, in patients or cSVD models, suggest at least three different patterns (Figure 23). Smooth periventricular WMHs appearing as “caps” around the ventricular horns or periventricular lining or halos are extremely frequent in elderly individuals and likely relate to the disruption of the ependymal lining with subependymal gliosis and enlargement of the extracellular space. Punctate WMHs most likely correspond to enlarged PVS with or without associated demyelination. More extensive WMHs are associated with loss of myelin, axons and oligodendrocytes of various severity, and gliosis (588, 589). However, extensive WMHs in the temporal lobe in CADASIL patients have been associated with numerous enlarged PVS surrounded by demyelination and myelin degradation (538). High-resolution electron microscopy in a CADASIL mouse model showed that early WM lesions affect first and foremost the myelin sheath, suggestive of a primary myelin injury driven by a hypoxic/ischemic mechanism (436). Furthermore, it is important to bear in mind that lacunes and microinfarcts are frequent in the WM and that the vast majority of microinfarcts escapes detection by MRI (299).

Figure 23: Schematic representation of white matter hyperintensities (WMH).

WMH (pseudo colored in blue) are divided into deep and periventricular WMH (right side) and exhibit different patterns (left side).

Mechanisms of WMH are likely multifactorial. WMH are classically divided into deep WMH (separate from the ventricles) and periventricular WMH (contiguous with the margins of each lateral ventricle) (Figure 23). This subclassification may reflect differences in the severity of associated cognitive impairment (590) and microstructural changes (591). A GWAS analysis showed that deep and periventricular WMH have both different and shared genetic underpinnings, suggesting both shared and distinct mechanisms (592). Studies correlating pathology with MRI appearances suggested that in deep WMH, the main pathological changes were demyelination, axonal cleavage, and gliosis, while in periventricular WMH the predominant changes were the destruction of the ventricular ependyma and interstitial oedema (593, 594).

The two main proposed mechanisms of WMH, which are not mutually exclusive, are hypoperfusion and compromised regulation of brain fluids due to BBB leakage or glymphatic dysfunction. The assumption that WM lesions may be caused by cerebral hypoperfusion is as old as their description and stems from the following clinical observations. Affected WM is supplied by the distal end of long medullary arteries whereas U-fibers, which have a dual blood supply (see section 2.1), are relatively preserved. In large cohorts of CADASIL patients, and in elderly participants from the Austrian Stroke Prevention Study, analysis of lesion prevalence maps across different disease stages showed that WM lesions expand from the periventricular to subcortical regions, confirming that the predilection site for WMH is the vascular end zone (595). Neuropathological studies have shown that pathological changes associated with cSVD transform long medullary arteries to “earthen pipes”, likely hampering their ability to dilate. Moreover, other studies showed that the capillary density was reduced in the WM (see section 7.1). Both defects are predicted to compromise resting CBF and adaptive CBF responses. Many clinical studies, including a recent metanalysis of 34 studies and 2,180 participants, have reported a reduction of CBF in the WM of cSVD patients (596). Moreover, decreased reactivity to hypercapnia has been documented in the WM of cSVD patients at both the cross-sectional and longitudinal levels (495). Also, preclinical studies are consistent with this “hypoperfusion” hypothesis. As mentioned above, Notch3KO, CADASIL and Htra1KO mice exhibit a widespread hypoperfusion (239, 433, 477) and chronic cerebral hypoperfusion in rodents results in WM lesions (375). However, whether hypoperfusion is a cause or consequence of WM lesions remains controversial since the few available longitudinal studies have yielded contradictory results. What is crucially missing are clinical studies measuring in the same patients CBF and the O₂ extraction fraction (OEF). Indeed, low CBF and low OEF indicate matched low O₂ supply and demand whereas low CBF and elevated OEF are signatures of hypoxia/ischemia (597). Importantly, such studies should focus on homogeneous group of patients, such as patients with genetically-defined cSVD, to overcome the confounding issues of associated co-morbidities or other neurodegenerative processes often present in patients with sporadic cSVD that may also be associated with hypoperfusion. At the preclinical level, rodents have little WM and the corpus callosum, which is the biggest WM tract, is more akin to the U-fibers in terms of anatomy and blood supply. Moreover, the deep WM still remains technically poorly accessible to in vivo microscopy imaging (598). Nevertheless, longitudinal studies in genetic models of cSVD assessing simultaneously WM integrity at the whole brain level, using for example DTI, as well as resting CBF and adaptive responses to hypercapnia, using for example ASL-MRI, would be extremely informative. In addition, the development of preclinical ultrahigh field (17.2 Tesla) MRI makes it possible now to image the mouse brain at an unprecedented microscopic resolution (599).

An increasing body of evidence also implicates BBB dysfunction in the pathogenesis of sporadic SVD (see section 7.3.4). This hypothesis, fueled initially by neuropathological studies showing the extravasation of plasma proteins in the brain parenchyma in cSVD (600), has been invigorated during these last years by the ability to detect in vivo, non-invasively, changes consistent with an increased BBB permeability to small tracers in patients with different manifestations of cSVD, in WMH and the normal appearing WM (NAWM) (516, 520, 527). Using DCE-MRI, both widespread regions with subtle increased permeability and hotspots of leakage (522, 601) have been described. Moreover, using ASL-MRI, changes consistent with reduced BBB water exchange rate have been observed in multiple brain regions, in CADASIL and CARASIL patients (524) (see section 7.3.4). Deposition of fibrinogen and other plasma proteins in the parenchyma has deleterious effects that can activate microglia, cause neuronal or axonal loss and demyelination (602). Moreover, increased BBB permeability or reduced BBB water exchange could increase interstitial fluid and cause WM edema. Extracellular free water can be quantified non-invasively using DTI in combination with mathematical modeling of several compartments (603). Changes consistent with an increased extracellular free water have been reported in the WM and the NAWM in patients with sporadic cSVD and in CADASIL patients (604–606). In addition, extensive WMHs in certain CADASIL patients have been associated with an increased brain volume suggestive of a global increase in water content (607, 608). Yet, most studies in human of BBB dysfunction in cSVD have been cross-sectional and therefore cannot exclude a non-causal association; to date there are few longitudinal studies determining whether increased BBB permeability predicts disease progression, and whether hotspots of BBB leakage progress to WMH. One study (43 patients) showed a link between BBB leakage at baseline and the loss of microstructural integrity over 2 years in the perilesional zones around WMHs (531). Results of the INflammation and Small Vessel Disease (INSVD) study, a prospective observational multicenter longitudinal study, combining clinical and cognitive assessment, advanced neuroimaging markers of SVD and BBB permeability, performed in 200 individuals with symptomatic cSVD at baseline and 2-year follow-up, will provide data on whether BBB permeability precedes WMH and predicts cSVD progression. (609).

Glymphatic dysfunction is another emerging mechanism that could cause WM lesions by impairing brain fluid regulation or brain wastes removal. As mentioned in section 7.3.5, preliminary evidence of glymphatic dysfunction has been shown only in mouse models of cSVD. What is missing is the availability of a non-invasive approach, which has been validated against the reference MRI methodology using intrathecal injection of contrast agent (see section 7.3.5) that can be applied for cross-sectional and longitudinal studies in human.

Neuroinflammation has also been indicated in the pathogenesis of WMH, suggested by data from SHRSP with unilateral carotid artery occlusion and fed with the Japanese permissive diet (610), and neuropathological studies showing elevation of inflammatory markers close to diseased arteries (611). Circulating blood biomarkers of both inflammation and endothelial activation are elevated in cSVD, and Inter Cellular Adhesion Molecule-1 was an independent predictor of WMH progression over 3 and 6 years follow periods (612). In patients with cSVD circulating monocytes showed evidence of reprogramming toward a long-term proinflammatory phenotype, which has been termed trained immunity, characterized by increased cytokine production capacity and a pro-inflammatory transcriptional signature (613). Cytokine production capacity by monocytes was associated with WMH progression (614). However, while there is considerable evidence implicating systemic inflammation in cSVD, studies examining whether there is evidence of central nervous inflammation in human is limited. This reflects the challenges of measuring brain inflammation in vivo. Neuroinflammation can been measured using PET imaging of radioligands for the 18 kDa translocator protein (TSPO), whose expression is upregulated in activated microglia. Using the TSPO radiotracer ¹¹C-PK11195, increased ligand binding has been demonstrated in cohorts with lacunar stroke and WMH (522), and also in a mild cognitive impairment cohort in which they correlated with WMH (615). In addition to a global increase in binding, focal hotspots of increased ¹¹C-PK11195 binding could be identified (522). However, whether this increased brain inflammation is casual or a consequence of tissue damage remains to be determined. There has only been one longitudinal study to date, which showed that WM tissue destined to develop into new WMH over the subsequent year was associated with both lower neuroinflammation (¹¹C-PK11195 ligand binding), and WM ultrastructural damage on DTI, at baseline (616) This suggests that this tissue is already damaged 1 year prior to lesion formation. It could reflect that neuroinflammation plays a role earlier in WMH development, or that it is not causally related. Further longitudinal studies with longer follow-up are required.

A pathophysiological cascade linking BBB disruption and inflammation has been proposed in the SHRSP and unilateral carotid artery occlusion model of WM ischemia (610). Chronic inflammation resulted in WM hypoperfusion and hypoxia, which led to an increase in hypoxia inducible factor 1 subunit alpha inducing an inflammatory response, with release of matrix metalloproteinases (MMPs) that disrupted the extracellular matrix of the vascular endothelium leading to opening of the BBB. MMP9 was hypothesized to play a key role, and minocycline which inhibits MMP9 significantly reduced lesion size, improved cerebral blood flow, improved performance on the Morris water maze, and prolonged survival (610). This hypothesis implies that increases in BBB permeability and microglial activation may be related. Although hotspots of both BBB permeability and ¹¹C-PK11195 increased ligand binding have been demonstrated in patients with lacunar stroke and confluent WMH, the two did not colocalize (522). The recent MINERVA trial in the same patient group found minocycline, given for 3 months, was not associated with any reduction in either BBB permeability or ¹¹C-PK11195 ligand binding (617).

Another proposed mechanism identified in the SHRSP model involves a defective myelination caused by a blockage of oligodendrocyte differentiation driven by endothelial cell dysfunction (618). However, evidence of such a mechanism in more relevant cSVD models is lacking.

An important question is whether lacunar infarcts and WM lesions have common or distinct mechanisms. Two studies have examined the spatial relationship between incident (newly developed) lacunes and WMH, one in a cohort of 276 CADASIL patients (104 incident lacunes in 64 patients) (595), and the second one in a cohort of 503 sporadic cSVD patients (43 incident lacunes in 43 patients) (619). Both studies showed that a minority of lacunes (5%) developed within preexisting WMH, suggesting that cavitation of WMH is an uncommon mechanism of lacune formation. Importantly, the majority of lacunes (90%) in CADASIL patients developed at the edge of WMH, an observed distribution much higher than the simulated distribution taking into account the high burden of WMH in these patients. Moreover, most of these lacunes developed proximal to WMH with respect to the anatomical course of perforating arteries (595). Notably, a close proximity between prevalent lacunes and WMH was also detected in both CADASIL patients and a large cohort of 588 community-dwelling elderly subjects (595). However, results contrasted with those in sporadic cSVD patients, in whom less than 50% of lacunes occurred at the border of WMH (619). Although it could be argued that lacunar infarcts in CADASIL and sporadic cSVD have different mechanisms, a major strength with studies in genetically defined cSVD, such as CADASIL, is that lacunar infarcts can be definitely attributed to cSVD, which is not always the case for lacunar infarcts in sporadic cSVD, because of associated comorbidities. Nevertheless, what these data suggest is that the mechanism of lacunes and WMH are intimately connected, with one mechanism of infarct possibly involving an in-situ thrombosis of a stenosed/diseased vessel in a region of hypoperfused tissue.

8.1.3. Progression and regression of brain lesions

Recent data has suggested cSVD is more dynamic than previously thought. It has been shown that “asymptomatic” DWI positive lesions are much more frequent than clinical infarcts (557, 620, 621). In the RUN DMC - InTENse study, in which patients with cSVD were regularly scanned, the median monthly incidence of DWI+ lesions was 4.4% (557). All DWI+ lesions were silent, in that none of the patients experienced a clinical event between the respective visits. DWI+ lesions were distributed throughout the brain, with 32 (82%) being supratentorial; these were located in the WM (n = 5), subcortical gray matter (n = 3), cortex (n = 22), and the cortical gray–WM junction (n = 2). All 7 infratentorial DWI+ lesions were located in the cerebellar cortex. Follow-up scans were available for 92% of DWI+ lesions. Interestingly, they had a diverse evolution with two evolving into a WMH, one into a lacune ≥3mm, three into a small cavity <3mm, and 3 into a microbleed. Twenty-five DWI+ lesions disappeared or almost vanished on follow-up FLAIR and T1-MRI, despite having shown a FLAIR and/or T1 signal change in the acute phase. The majority of lesions that disappeared were located in the cortex. It is possible that such asymptomatic lesions contribute to the neuropathologically identified CMI seen at post-mortem, as well as brain atrophy. A more recent follow-up study from the same cohort, reported DWI positive lesions were associated with more rapid cognitive decline (622).

Individual WMH may also regress (623). This has been shown for individual lesions, but more recently it has been suggested that total WMH can also reduce as well as increase (624, 625). WMH regression was reported in 37% of patients with minor stroke (626). This suggests cSVD may be dynamic, but caution should be used in interpreting the data. Both regression and progression would be detected if lesions volumes were stable, but there was a measurement error in WMH lesion volume estimation. A more rigorous analytic approach, designed to reduce any such measurement error, suggested the actual rate of regression (using a definition of 0.25 cc to define regression) was much lower at 0.2 to 14.3% over periods of a few years in three cohorts with symptomatic cSVD (627). Further studies are required using careful measurement of WMH volume changes to determine how common WMH lesion volume reduction is, and how it relates to both risk factor treatment and cognitive impairment.

8.2. Spontaneous intracerebral hemorrhage

In his seminal paper « Des foyers lacunaires de désintégration », Marie mentioned the presence of ICH in about one third of his patients (127). Non amyloid cSVD-driven ICH are predominantly located in deep brain regions including the thalamus, basal ganglia and brainstem. The most common risk factors include age, hypertension and alcohol consumption (628). The importance of hypertension in ICH is supported by several very large randomized controlled trials that have demonstrated that BP lowering significantly reduces the risk of ICH in both primary and secondary prevention, a benefit which is much stronger than on ischemic stroke (629–631).

Takebayashi and Kaneto have examined lenticulostriate and thalamic arteries from 36 hypertensive patients with spontaneous deep ICH, collected during surgery or post-mortem evaluation of ICH (632). They identified 48 ruptured arteries in 21 patients and observed that, at the site of rupture, arteries were abnormally dilated. Using electron microscopy, they found that 46 out of the 48 ruptured arteries exhibited prominent changes of the media characterized by severe degeneration and loss of SMCs and breakage of the elastic lamina; they further reported that. degenerative changes were restricted to the middle and distal portions of the penetrating arteries (632). By applying his famous serial sectioning approach to post-mortem brain tissue with a recent thalamic hemorrhage, Fisher identified the initial source of bleeding and made an almost similar observation. He reported that the hemorrhage arose through a 1-mm opening of a distal branch (180- μm) of a deep penetrating artery which, at the site of rupture, was abnormally dilated (600 μm) over 5 mm and had a thin wall without any residual SMCs or elastic lamina, whereas the parent artery appeared almost normal except for increased thickness (633). More recently, immunohistochemical and ultrastructural analysis of Col4a1^G498V/+ mutant mice, a genetic model of non-hypertensive spontaneous deep ICH, similarly established that ICH originates from penetrating arteries that harbor focal degeneration and loss of SMCs (250). One factor which could explain the anatomical preference of ICH in deep brain regions is that these regions are supplied by short branches of major arteries of the circle of Willis or from the basilar artery (31). Computational modeling suggests that, owing to their proximity to the circle of Willis, these penetrating arteries are exposed to higher BP compared to other arteries like the cortical or medullary arteries (634). The prediction is that when damaged these arteries have a higher risk to rupture.

However, the finding that deep ICH results from focal degenerative changes of the arterial media raises several intriguing questions. How can hypertension be a major risk factor for two seemingly opposite manifestations, namely ischemic infarcts and hemorrhages? On the other hand, CADASIL is characterized by prominent degenerative arterial changes; thus why is ICH a rare manifestation in CADASIL, occurring in less than 2% of symptomatic cases? (263). The same holds true for other genetic cSVDs, like CARASIL and early onset autosomal recessive NOTCH3-related cSVD. Simultaneous analysis of Col4a1 cSVD hemorrhagic model and the Notch3KO ischemic model of cSVD provided evidence for two mutually-reinforcing defects of the brain microvasculature as a new mechanism for ICH (Figure 20). As mentioned in section 7.1, in Col4a1 mutant mice, arteries exhibit focal and segmental loss of SMCs, whereas the ACT zones, show an hypermuscularization characterized by an increased number of mural cells and a higher level of contractile proteins. Conversely, in Notch3KO mice, a model of cSVD which does not bleed, both arteries and ACT zones exhibit a loss of mural cells (54). These observations raised the possibility that hypermuscularization of the ACT zone could locally decrease vessel diameter, increase vascular resistance and raise the intravascular pressure in the upstream proximate artery, to favor its rupture at the site of SMC loss. Further experiments supported this prediction. First, functional studies demonstrated that hypermuscularization increased the contractility and myogenic tone of the ACT zone and computational modeling suggested that hypermuscularization of the ACT crippled the pressure drop along the arterio-ACT axis and thus resulted in an abnormally higher intravascular pressure in the proximate feeding artery. Secondly, molecular studies revealed that hypermuscularization of the ACT zone in Col4a1 mutant mice was driven by an increased activity of Notch3; genetic reduction of Notch3 in Col4a1 mutant mice attenuated hypermuscularization of the ACT zone, without affecting arterial SMC loss, and importantly, prevented the occurrence of ICH in Col4a1 mutant mice (54). It is anticipated that it is in the penetrating arteries which are already exposed to a high blood pressure, and thus in those supplying the deep brain regions, that the hypermuscularization should have the most deleterious effect (54).

The important question which comes next is whether such a mechanism, identified in a genetic model, happens in patients with sporadic ICH. In an immunohistochemical study of post-mortem brains from seven hypertensive patients with Binswanger’s disease performed more than 2 decades ago, Wang and Olsson concluded that “the smooth muscle cells of the media respond in different ways depending on the size of the affected vessel”. Indeed, they observed that larger brain arteries showed loss of smooth muscle actin immunostaining whereas many terminal arterioles, i.e. very likely ACT zones, presented a marked actin immunostaining possibly indicating hypertrophy of smooth muscle cells (635). More recently, an immunohistopathological analysis of human post-mortem brain tissues from 7 patients with spontaneous deep ICH and aged-matched controls revealed, in addition to segmental SMC degeneration in arteries, a hyperplasia/ hypertrophy of mural cells in the ACT zone (54). Together, these data suggest that the mechanism of ICH identified in Col4a1 mutant mice also operates in human sporadic deep ICH and further raise the interesting possibility that changes in the properties or density of mural cells in the ACT zone could influence the hemorrhagic versus ischemic presentation of cSVDs (Figure 20) (212).

Another important question pertains to the hematoma expansion. According to neuroimaging and pathological studies, the hematoma may grow as a consequence of secondary mechanical shearing and bleeding of adjacent damaged vessels (636, 637), a hypothesis proposed by Fisher in 1971 under the name of the “avalanche model” (638).

8.3. Cerebral microbleeds

CMBs likely occur due to the rupture of small arterioles and/or capillaries. In human CMBs, the diameter of ruptured vessels has been estimated to be less than 200 μm, with many bleeds occurring at the arteriole and capillary levels (639). CMBs are visible on gradient echo or susceptibility weighted MRI sequences as small (2–10mm) oval hypodense regions (black spots) (Figures 12 and 13), and correlative studies have shown they correspond pathologically to hemosiderin deposition. MRI is known to overestimate the size of CMB (the “blooming effect”), with MRI diameter being on average more than 150% of pathological lesions (640). Nevertheless, conventional MRI is likely to underestimate the total CMB burden. Hemosiderin remains present for a long time at the location of a previous bleeding, and therefore CMBs on MRI may have been formed many years previously. MRI-guided neuropathological examinations have confirmed that the majority of CMBs corresponds to recent or old microhemorrhages, although other lesions, such as microaneurysms, fibrinoid necrosis and hemorrhagic microinfarcts are thought to account for a small subset of MRI-detected CMB (641, 642).

The cellular and molecular mechanisms underlying microvascular fragility associated with the development of CMBs are not well understood, but a number of potential mechanisms have been suggested (643). Disruption of the extracellular matrix network is thought to play a critical factor in the development of CMBs. In experimental animals injection of extracellular matrix-degrading enzymes [e.g., collagenases] resulted in the development of ICH (644). In humans, patients with genetic disorders affecting components of the extracellular matrix often present with ICH, and it is of note that COL4A1/2 mutations have a particularly high number of CMBs compared with other monogenic forms of cSVD (248). MMP activation has been suggested to provide a common mechanism of CMBs associated with multiple diverse pathophysiological conditions including both CAA and cSVDs (643). Recently it has been suggested that brain endothelial erythrophagocytosis (the process by which injured or aged erythrocytes are ingested by endothelial cells), perhaps triggered by oxidative stress, resulting in passage of hemoglobin across the brain endothelial monolayer with unaltered monolayer integrity, may play a role in the pathogenesis of some CMB (645). CMBs in patients with hypertension have different distribution patterns from those with CAA, with CMBs predominantly located in deep and infratentorial regions (Figure 13). It has been proposed that in deep and infratentorial regions, penetrating branches arise directly from the posterior cerebral artery and the middle cerebral artery and because of their branching pattern, the segment of proximal vessels, whose resistance protects the microcirculation, is shorter than in other brain regions (646). These anatomical properties would render these vascular supply areas more vulnerable to sudden changes in BP. It is likely that many of the mechanisms predisposing to ICH described above in section 8.3 also predispose to CMB.

8.4. Which cell types are involved?

If there was a polling on this question, a dysfunction of endothelial cells (EC) would come first as the key initiator and driver of cSVD (647). EC occupy a strategic position within the neurovascular unit, with blood on their luminal side and mural /brain cells on their luminal side. They form a highly interconnected ~400 miles network which is the cornerstone of key functions of brain vessels including the BBB integrity and the regulation of resting CBF and adaptive CBF responses through their influence on vascular structure, mechanics and tone and their implication in neurovascular coupling (see section 2.3). Moreover, EC control blood coagulation (563) and may have a trophic support on glial cells (618). Aging has a profound effect of the transcriptome and proteome of brain ECs. Specifically, aging is associated with an upregulation of transcripts of the innate immunity and oxidative stress response pathways (648), an upregulation of proteins implicated in protein degradation and a decline in proteins implicated in vesicle-mediated transport (649). Moreover, GWAS studies have shown an enrichment of EC genes in loci associated with WMHs (358). The existence of BBB dysfunction as well as diminished resting CBF and impairment of neurovascular coupling in cSVD patients is highly suggestive of EC dysfunction (see section 7.3). Direct evidence of EC dysfunction comes from experimental studies that have shown enhanced BBB permeability and decreased EC-dependent vessel dilation upon aging and in models of AngII-driven hypertension, and diminished activity of the Kir2.1 EC channel activity as the mechanism of impaired neurovascular coupling in distinct mouse models of sporadic and genetic cSVD (see section 7.3.3). As discussed above, BBB leakage or defective regulation of CBF are potentially involved in cSVD-driven brain lesions. An additional piece of evidence comes from the observation that mice deficient for the enzyme which produces NO, the major mediator of EC-dependent vasodilation (eNOS), develop some key features of cSVD, including age-related cerebral hypoperfusion, BBB leakage, thrombotic cerebral microinfarcts, WM lesions and memory defects (650, 651). It is worth mentioning that NO has pleiotropic effects on platelets and the peripheral vasculature and that deleterious effects of eNOS deficiency in the mouse go beyond the brain vessels with systemic (652) and pulmonary (653) hypertension, enhanced atherosclerosis (654) and thrombotic kidney microangiopathy (655). In human, loss of function mutations in eNOS or the alpha1 subunit of the soluble guanylate cyclase (sGC), which is the major NO receptor in vessel, are associated with Moyamoya disease, a macroangiopathy characterized by a progressive stenosis of the intracranial internal carotid arteries and their proximal branches (656). Yet, genetic evidence that NOS3 (the gene encoding eNOS) is associated with features of cSVD is weak and limited to a GWAS study showing an association between a gene variant in the NOS3 gene and periventricular WMH (592).

The importance of EC should not disguise the fact that a dysfunction or a change in the number of mural cells including SMCs, pericytes and mural cells of the ACT zone, are definitely involved in cSVD. Arterial SMCs and mural cells of the ACT zone are the contractile cells enabling the dilation or constriction of arteries and as such are important for the regulation of both resting CBF and adaptive CBF responses and for the glymphatic function; thin-strand pericytes are involved in the regulation of CBF and BBB (see section 2.3). In support of a direct contribution of mural cells in cSVD, is the finding that mutations in the NOTCH3 receptor, the expression of which is highly restricted to mural cells and which is critically required for the arterial differentiation of SMC, their maintenance, survival and function (310), is sufficient to cause cSVD in human and mice (see section 5.4.1). Secondly, loss and degeneration of arterial SMCs are common features in sporadic and genetic cSVDs, and not just an end-stage lesions, as evidenced by pathological studies in mouse models. Remarkably, the most aggressive forms of cSVD are associated with very severe loss of arterial SMCs (see section 7.1.1). It is notable that a dysfunction of arterial SMCs, characterized by a change in the myogenic tone of cerebral arteries, has been described in several experimental cSVD models prior to SMC loss (see section 7.3). As mentioned above, loss of arterial SMCs and hyperplasia or hypertrophy of mural cells of the ACT zone are critically involved in the occurrence of hemorrhage, and analysis of distinct cSVD models suggest that the density of mural cells in the ACT zone may influence the ischemic versus hemorrhagic phenotype of cSVD (212). Thirdly, a reduction in the density of pericytes or pericyte coverage of brain capillaries has been observed during aging, and in patients or mice with cSVD, although inconsistently (see section 7.1.3). Whereas a significant reduction in pericyte coverage (>50%) seems to be required to induce BBB leakage (101), a focal loss of pericyte can impair brain capillary flow and architecture by inducing aberrant capillary dilation in the uncovered capillary branch, and, sometimes, blood flow stalls and capillary regression in capillary branches neighboring dilations. This results in increased blood flow heterogeneity that may compromise proper allocation of O₂ to brain parenchyma (657).

Yet, it is unclear, except in Notch3 mutant, whether mural cell degeneration occurs cell autonomously or is secondary to EC dysfunction. It has been proposed that arterial SMC degeneration, as typically observed in arteriolosclerosis, could be driven by the leakage of blood proteins into the vessel wall, possibly as a result of the disruption of EC tight junctions or increased EC transcytosis (658). Moreover, EC communicate with SMCs through diffusible gas, changes in the membrane potential or ligand/receptors pairs. As a few examples, NO produced by EC, diffuses to the neighboring SMCs where it activates the soluble guanylate cyclase to dilate the vessel (659), hyperpolarizing signal in EC during neurovascular coupling is passed to SMC through myoendothelial projections (85) and JAGGED1 transmembrane ligand expressed by EC activates NOCTH3 receptors expressed by SMCs (660). Thus, durable EC dysfunction could have a strong impact on the integrity and function of SMCs. Conversely, SMCs loss may affect EC health and function, a hypothesis which remains to be investigated. With respect to pericytes, experimental studies using pericyte-deficient mice generated by crippling PDGFB- PDGFR-ß (EC-pericyte) signaling have demonstrated that the loss of pericyte has dramatic effect on EC morphology and function characterized by a venous-shifted molecular pattern of capillary EC, abnormal and irregular patterns of EC tight and adherens junctions and increased EC transcytosis culminating in diffuse and hotspot BBB leakage (661).

Resident cells of PVS, i.e., PVFBs and PVMs, must be also considered as serious candidate cells. Experimental studies have begun to unravel the diversity of brain associated macrophages (662) and the role of PVMs in the brain (663). In physiological conditions, PVMs regulate CSF flow dynamics and the deposition of extracellular matrix protein deposition around vessels (66). In AngII-driven mouse model of hypertension, PVMs can be the source of deleterious reactive oxygen species having a deleterious impact on BBB integrity (534) and neurovascular coupling (426) (see section 7.3). Single cell transcriptomic studies have begun to unravel the molecular diversity of brain fibroblasts located in the meninges, choroid plexus or within the perivascular space (49, 50, 52, 63). It is well-established that fibroblasts are a major source of extracellular matrix proteins. Yet, whether and how PVFBs contribute to vessel fibrosis in cSVDs is unknown. Hence, more efforts must be invested in better understanding the role of PVBFs and PVMs in both physiological and cSVD conditions.

9. How does cSVD causes cognitive impairment?

MRI has provided important insights into the mechanisms underlying cognitive impairment in cSVD. A large number of studies have examined relationships between conventional MRI markers and cognitive impairment and other non-motor symptoms in cSVD. Because the different MRI cSVD markers are highly correlated any association needs to be shown to be independent of the other cSVD markers. Longitudinal studies in diverse populations have consistently demonstrated that increasing WMH volume predicts cognitive decline and dementia (664). In a meta-analysis containing mainly healthy community cohorts, both lacunar infarcts and WMH predicted future dementia risk; there was less data on CMB, but a non-significant trend was seen, while data was insufficient to examine associations with PVS (665). In cohorts of patients with symptomatic sporadic cSVD independent associations have been replicated between the number of lacunar infarcts and presence of cognitive impairment (666), while in CADASIL patients, lacune count was found to be the strongest predictor of cognitive impairment (667). Associations between WMH severity and cognition in patients with symptomatic SVD have been more varied; some have shown strong associations (668), while in others correlations have been weak or absent (169), perhaps because in many of these studies all patients have had advanced WMH, reducing the power to detect associations. Results have been more variable for CMB, but recent larger studies have reported associations between the presence and number of CMB with impaired cognition, particularly executive function and information processing speed, after controlling for other MRI SVD markers (668, 669). Most studies have found brain atrophy to be an independent predictor of cognitive impairment (670). Many studies have examined associations between enlarged PVS and impaired cognition but with conflicting results with some showing associations (671), but others not (672), a meta-analysis in 2019 found no association but heterogeneity in studies (673). More data from large high quality studies, ideally using quantitative scores rather than visual rating scales, and controlling for potential confounders, is required (674). Therefore, in conclusion, convincing data shows associations between lacunar infarcts, WMH and brain atrophy with cognition, limited data suggests an association with CMB, while the currently available data on PVS is inconclusive.

Lesion location, not just severity, also appears to be important. Voxel-based lesion–symptom mapping studies have shown that damage to key subcortical-frontal connections such as the anterior thalamic radiation and forceps minor by both WMH and lacunes is associated with low processing speed and executive dysfunctions. WMH burden in these strategic WM tracts appears to be more important in explaining variance in cognitive functioning than global WMH volume (675). Thalamic lacunes are more strongly associated with impaired information processing speed (676). In a large multicenter sample of 2950 subjects, infarcts in the left frontotemporal lobes, left thalamus, and right parietal lobe were strongly associated with post-stroke cognitive impairment (677).

Advanced MRI techniques are providing additional information on the type of WM damage and the mechanisms of cognitive impairment (483). Magnetization transfer imaging ratio provides an estimate of myelination. A correlation with myelin integrity has been validated for magnetization transfer ratio (MTR) in post-mortem studies in multiple sclerosis (678). A reduction in MTR has been shown in both sporadic cSVD (679) and CADASIL (680), both in WMH and also in normal appearing WM and normal appearing grey matter, and the degree of the reduction correlated with the extent of cognitive impairment, although DTI metrics correlated more strongly than MTR with cognition (679).

DTI provides measures of WM microstructural damage, by quantifying water movement. The two most widely derived metrics are mean diffusivity which provides a measure of the extent of water movement in all directions, and anisotropy which provides an estimate of the directionality of diffusion and therefore of WM tract integrity. Numerous studies, the first in 1999, have shown increased diffusivity and reduced anisotropy in cSVD, both sporadic (681) and genetic (682), with abnormalities not only within WMH but also in the WM appearing normal on conventional imaging (483, 584). This emphasizes the diffuse nature of the cSVD disease process in the WM. Multiple cross-sectional and longitudinal studies have shown strong associations between DTI metrics and clinical deficits in cSVD (683). Importantly prospective studies have shown DTI predicts future dementia risk, independent of other MRI markers (587). Interestingly DTI metrics within the normal appearing WM predict cognitive impairment and dementia better than do DTI abnormalities within the WMH, implying that diffuse disruption of WM structure plays a key role in the mechanism of cognitive impairment. This led to the hypothesis that WM tract disruption is a key mechanism in cognitive impairment due to cSVD, both in normal ageing (684) and in patients with symptomatic cSVD (685). Support for this hypothesis has now come from tractography studies and network analysis.

Structural network analysis using DTI images and tractography to construct WM pathways and by combining this with atlas-based brain parcellations, creates brain networks which can be analyzed using graph-theoretical metrics (483). Structural network integrity is disrupted in cSVD, and the degree of disruption correlated cross-sectionally with cognition independent of other MRI markers (686). The degree of disruption has also been shown to predict future dementia risk (687). Importantly, the effect of conventional markers of cSVD (lacunes, WMH, diffuse WM damage on DTI and CMBs) on cognition is mediated by the degree of network disruption (686). This provides strong evidence that network disruption, dependent on WM tract integrity, is the key mechanism underlying cognitive impairment and dementia in cSVD. Interestingly the effect of brain atrophy on cognition was not fully mediated by network disruption (686), suggesting other mechanisms may also contribute, possibly including cortical neuronal loss secondary to both cortical microinfarcts and also retrograde cortical degeneration. Network analysis can also be performed using blood-oxygen-level-dependent imaging (BOLD) imaging; using this technique disruption has also been reported in cSVD but caution needs to be used in interpretation because the reproducibility of functional networks has been shown to be very poor in cSVD (688, 689), possibly due to effect of the vascular pathology on neurovascular coupling on which the BOLD signal depends. Recent data suggest that network disruption, due to white matter tract damage, is not only a key mechanism underlying cognitive impairment but is also an important mechanism underlying other non-motor symptoms including apathy (690) and gait abnormalities (691).

10. Treatment – current state and future directions

10.1. Primary and secondary prevention in cSVD

Despite its public health importance there are limited treatments for cSVD. When considering treatments these can be categorized into treatments to prevent disease or delay disease progression, including risk factor modification, and symptomatic treatments for patients with established disease including the associated vascular cognitive impairment.

For lacunar stroke, secondary prevention strategies have been mostly inferred from studies of ischemic stroke in general, the majority of which did not specifically examine efficacy in lacunar stroke (692). This approach has a major limitation because stroke subtyping in these studies, if perform at all, has been based on diagnosis of cSVD based on a clinical lacunar syndrome, and often use of the Oxford Community Stroke Project (OSCP) Classification. Validation studies have shown that as many as half of all patients classified as lacunar stroke in this way, do not have cSVD, but another stroke mechanisms, on more rigorous subtyping (193).

Only one large definitive phase 3 trial has been performed in well subtyped lacunar stroke patients. The SPS3 trial (Secondary Prevention of Small Subcortical Strokes) recruited 3020 patients with recent symptomatic lacunar infarcts all confirmed on MRI, and showed that long-term dual antiplatelet therapy with clopidogrel and aspirin did not reduce the risk of recurrent stroke but did significantly increase the risk of bleeding and death (693). Based on this, current guidance recommends single antiplatelet agent use with either aspirin or clopidogrel in long term secondary prevention of cSVD after lacunar stroke. Whether single antiplatelets have any effect at all in cSVD is not certain. A systematic review reported single antiplatelet therapy is better than no therapy for preventing new stroke (ischemic and hemorrhagic combined; risk ratio, 0.77 [95% CI, 0.62–0.97]) or new ischemic stroke (risk ratio, 0.48 [95% CI, 0.30–0.78]) in patients with recent lacunar stroke, but cSVD subtyping within these studies was not optimal. It is interesting that MR studies have shown no association between altered coagulation and lacunar stroke, in contrast to significant associations with thromboembolic stroke secondary to cardioembolism or large artery atherosclerosis, questioning whether antiplatelet agents have any efficacy in cSVD (see section 8.1) (562).

The one risk factor for which robust trial data is available is treatment of hypertension, particularly highlighting the benefit of more intensive BP treatment regimens. SPS3 also included a BP arm and showed that more intensive BP lowering (target systolic, <130 mm Hg) reduced recurrent strokes. This reduction did not reach statistical significance for total stroke (ie, ischemic and hemorrhagic combined; hazard ratio, 0.81 [95% CI, 0.64–1.03]) but did for hemorrhagic stroke alone (hazard ratio, 0.37 [95% CI, 0.15–0.95]) (631). A benefit from intensive BP regimens is supported from the SPRINT trial which examined primary prevention. The overall SPRINT trial showed intensive BP lowering to target of less than 120 mm Hg systolic, versus a target of less than 140 mm Hg, was associated with a 25% reduction in fatal and nonfatal major cardiovascular events and death from any cause (694). Embedded in the trials were imaging and cognition sub studies. These showed that intensive BP lowering was associated with both reduced WHM progression on MRI (695), and a reduction in cognitive endpoints (696). The mean systolic BP was 121.6 mm Hg in the intensive treatment group and 134.8 mm Hg in the standard treatment group, resulting in a mean between-group difference of 13.3 mm Hg. Intensive BP control significantly reduced the combined rate of mild cognitive impairment or probable dementia (20.2 vs 24.1 cases per 1000 person-years, HR, 0.85). During the extended follow-up visits, the between-group BP difference was further reduced to 6.4 mm Hg, attributable to a further increase in the mean systolic BP in the intensive treatment group to 129.2 mm Hg. This seminal study has shown that BP lowering for a period of about 3 years can reduce progression of radiologically determined cSVD, and reduce the risk of future mild cognitive impairment/dementia by about 15%. Importantly it has led to the realization that people who have been thought to have cSVD with a ‘normal’ BP (historically 140-150mmHg) have a BP which is sufficient to cause future cognitive impairment. The SPRINT results, taken together with several other studies, both longitudinal observations and clinical trials, have led to a redefinition of hypertension to include systolic BPs of between 130 and 139 (697). Longitudinal studies have also demonstrated that mid-life hypertension at age 50 is a much stronger risk factor for dementia than late life (at age 70) hypertension, suggesting that for optimal prevention of cSVD and dementia BP should be treated from mid-life (204). Thus, the potential impact of treating mid-life hypertension, considering 50% of men and 44 % of women aged 45–54 years are hypertensive when defined using the 130mmHg cut-off (697), are enormous.

Therefore, BP control, ideally to the more intensive levels seen in SPS3 and SPRINT, should be implemented in people with clinical or MRI manifestations of cSVD. This is consistent with the strong epidemiological and MR data implicating BP as the most important risk factor for cSVD. There has been concern about BP reduction in patients with severe cSVD, in whom cerebral autoregulation may be impaired (698). However, the recent PRESERVE trial, in patients with lacunar stroke and confluent WMH, showed no reduction in CBF (500) or any increase in WM damage quantified on DTI, in patients randomized to a systolic BP of 125 mmHg, compared with 140 mmHg (699).

No trials similar to SPS3 have been performed to assess the effects of smoking cessation, diabetes management, statins, and lifestyle interventions on lacunar stroke. This has led to guidelines suggesting management should be inferred from the findings of large trials including all stroke subtypes (700). However, the guidelines have highlighted the need for more studies in well-phenotyped cohorts with the different manifestations of cSVD. Therefore, until we have more data available, treatment of risk factors is a key component of management, and as outlined in section 5.3, BP, diabetes, smoking and obesity appear strongly associated with cSVD and should be strictly controlled. The data from hypertension outlined above implies tight control of all risk factors is likely to be important, and that exposure to risk factors is important not only in symptomatic disease but also earlier in life if we are to prevent cSVD and its complications.

10.2. Symptomatic treatment

No treatments for the complications of cSVD have been proven in high quality trials. Anti-cholinesterase therapy has a modest benefit in Alzheimer’s disease and it has been suggested there may be a secondary cholinergic deficit in cSVD (701). A number of randomized clinical trials (RCTs) have examined cholinesterase inhibitors in vascular dementia and shown mixed results (702). Interpretation is complicated in that precise subtyping of dementia is difficult and in older adults mixed pathologies are common, so the interpretation of data in a ‘vascular’ dementia review needs to be mindful of this (702). A recent Cochrane review which performed network meta-analysis of RCTs using cholinesterase inhibitors found varying quality evidence that donepezil and galantamine may improve cognition compared to placebo, but the effect may not be sufficiently large to be clinically important (703).

One method of examining efficacy in a cohort with cSVD, unconfounded by co-existent neurodegenerative changes which become highly prevalent in an elderly population, is to perform RCTs in the younger onset form of cSVD, CADASIL. A European double blind multicenter RCT examined donezepil versus placebo in 168 participants with CADASIL and found no significant difference in the primary cognitive endpoint of vascular dementia assessment scale cognitive subscale at 18 weeks (704). There were small but significant improvements in executive function, but these had no impact on instrumental activities of daily living. This suggested that even though there may be a small biological effect, treatment had no clinically meaningful effect.

Most patients with dementia have mixed pathologies and utility of cholinesterase inhibitors has been reported for mixed dementia (Alzheimer’s disease plus vascular dementia), although with a lower degree of evidence than in pure Alzheimer’s disease (705).

10.3. Evaluating new therapies: challenges and pathways

Despite its enormous health importance there is no specific treatments for cSVD beyond risk factor modification. A number of factors account for this, including a lack of investment in vascular dementia in comparison with Alzheimer’s dementia, and a failure of stroke studies to specifically look at the patients with well phenotyped lacunar subtype (692). Another major factor has been a limited understanding of the underlying disease mechanism, but the recent advances described earlier now offer great opportunities for developing new treatment approaches. However, bringing new therapies into human presents considerable challenges, some of which are specific to cSVD. In this section, we describe ways to optimize evaluation of novel therapies, and how to most efficiently take new treatments through from early-stage studies to large definitive phase 3 studies.

10.3.1. Methods to evaluate the disease processes

Many of the disease processes believed to play a role in the pathogenesis of cSVD can be evaluated using brain imaging. A list of some techniques and processes they evaluate are shown in Table 9. One potentially important mechanism implicated is hypoperfusion, and a variety of hemodynamic variables, such as CBF, cerebrovascular reactivity, and cerebral autoregulation can be imaged. A number of imaging techniques can be used for measuring cerebral perfusion, with ASL being the most popular due its non-invasive nature (706). Cerebral reactivity can be estimated by measuring CBF, or a flow surrogate, in response to changes in inspired carbon dioxide or the carbonic anhydrase inhibitor acetazolamide (707). Dynamic cerebral autoregulation can be assessed by measuring blood flow changes following induced or spontaneous fluctuation in BP (708). It is important to recognize the limitations of each technique when designing experiments, including the reproducibility of each measure. A further important consideration is that some of these markers rely on imaging techniques which may themselves be affected by cSVD. For example, BOLD imaging is widely used to provide an estimate of vascular responses. It relies on the detection of deoxygenated hemoglobin which is paramagnetic, whereas oxygenated hemoglobin is not, and therefore the former causes local dephasing of protons, and reduce the returned signal. It has been widely used to measure functional activity in the brain in response to specific cognitive functions. However, it is possible that altered neurovascular coupling, which may occur in cSVD, could alter the BOLD responses and therefore make the technique less reliable in this patient group. This hypothesis is supported by studies using BOLD imaging to derive functional brain networks. Metrics from such networks were reproducible within normal individuals, but had low reproducibility in patients with cSVD (688). If such techniques are inappropriately used a lack of treatment response could reflect a lack of sensitivity of the technique, rather than a true lack of therapeutic efficacy.

Table 9.

Imaging techniques available to assess therapeutic efficacy in phase 2 trials.

Category	Technique	Details	Comments
Cerebral blood flow	MRI	– Arterial spin labelling (ASL) – Dynamic susceptibility contrast (DSC) MRI	ASL is non-invasive DSC-MRI requires intravenous contrast injection
	Positron emission tomography (PET)
	CT	Xenon-CT
		Perfusion CT	Usually measures relative not absolute changes
	Transcranial Doppler ultrasound (TCD)	Measures middle cerebral artery (MCA) velocity not flow, so only estimates flow if MCA diameter stays constant.	Estimates CBF in total MCA territory; cannot measure CBF in white matter regions
Cerebrovascular reactivity (711)	Measures CBF changes in response to vasodilatory stimulus of CO2 or acetazolamide	Multiple techniques can be used to estimate the CBF change – MRI ASL – BOLD MRI as a surrogate of CBF change – TCD to measure MCA velocity.	BOLD is not a direct measure of CBF, because it relies on neurovascular coupling which can be impaired in cSVD
Dynamic cerebral autoregulation (711)	CBF changes measured after induced or spontaneous fluctuations in blood pressure	TCD often used to estimate changes in CBF due to its high temporal resolution	Blood pressure changes are induced by deflating inflated cuffs on legs, or by monitoring spontaneous fluctuations in BP
Blood brain barrier permeability	Dynamic contrast-enhanced (DCE) MRI	Prolonged imaging after bolus of MRI contrast agent such as gadolinium. Leakage of contrast agent detected as change in T1 signal.	Most widely applied technique to cSVD More data on reproducibility and temporal variability required before use as a surrogate marker to assess therapies
	Water exchange MRI techniques (712)	ASL water exchange techniques study the kinetics of tagged arterial water as it passes through the vascular tree and into tissue.	Does not require contrast injection Few studies in cSVD to date
“Neuroinflammation”	PET with Translocator protein (TSPO) (18kDa protein) tracers	11C-PK11195 is the most widely used tracer	Believed to bind to activated microglia, but may also bind to other cell types

Another emerging pathological process implicated in cSVD is increased BBB permeability. Conventionally, this is measured by comparing the ratio of proteins, such as albumin, within the cerebrovascular fluid and blood. Such proteins are increased in the CSF if there is BBB permeability and this has been demonstrated in cSVD (709). However, MRI can also provide a measure of BBB permeability; it has the advantage that it not only avoids lumbar puncture but also can provide information on the spatial distribution of increased BBB permeability. DCE-MRI is used to model the movement of gadolinium contrast across the BBB into the brain, expressed as the influx constant Ki. As the BBB leakage seen in cSVD is low grade, and much less than that seen for example in an acute stroke, image acquisition needs to be extended to a period of 20–30 minutes. Increased BBB permeability has been shown in patients with both lacunar infarcts and WMH, and the severity of leakage correlates with the severity of radiological cSVD (516, 520, 527). This could represent a useful marker to test the efficacy of agents which may increase BBB integrity. However, one recent trial of minocycline, while reporting no effect on the drug on BBB permeability, found a very poor correlation between BBB measurements in the same individual at baseline and three months (617). More information on the temporal course of the increase in BBB permeability in cSVD, and the reproducibility of the technique, is required prior to its widespread use as surrogate disease marker for clinical trials.

Another disease process possibly implicated in cSVD is neuroinflammation. This can be imaged using positron emission tomography (PET) and tracers such as ¹¹C-PK11195, which binds to the 18kDa translocator protein (TSPO), with higher TSPO levels thought to indicate higher microglial activation (710). Increased ¹¹C-PK11195 ligand binding has been shown in patients with cSVD, both across the brain and in hot spots within the WM (522, 615). Such appearances are stable over a three months follow-up (617), but whether increased inflammation is casually related to disease progression or is a secondary phenomenon is uncertain.

10.3.2. Surrogate markers to assess treatment efficacy

Interventional trials in cSVD using clinical end points such as dementia or recurrent stroke require very large sample sizes and/or long trial durations owing to the low incidence of clinical end points. This has led to increasing interest in the use of surrogate end points as outcome measures to evaluate therapies in phase 2 trials prior to larger phase 3 trials (713). Surrogate endpoints proposed for trials in cSVD include blood markers, cognitive markers and brain imaging markers. It is important that any surrogate markers do indeed reflect the disease process and provide a valid measure of efficacy. Specific criteria such as those defined by Prentice (714) and the US Food and Drug Administration (715) have been developed to assess the suitability of markers as surrogate markers. These include sensitivity of the marker to change over time, correlation with clinical end points, and importantly that a treatment effect on the biomarker (eg, in a phase 2 trial) predicts a clinical benefit (eg, as assessed in a larger trial with clinical end points).

Blood biomarkers.

Blood based biomarkers are attractive as surrogate outcome markers as they can be easily collected in trials at low costs as part of routine assessments, and at a much lower cost than MRI. They can also be stored for central analysis on trial completion, thus greatly reducing inter site variability in multicenter trials (713). Many circulating biomarkers markers have been identified in cSVD, mostly in cross-sectional studies comparing cSVD cases with controls, or in populations correlating markers with MRI changes of cSVD. These include markers of endothelial dysfunction, neuronal injury, and BBB dysfunction. However, as yet, few have been shown to predict future dementia and stroke risk, and none to fulfil the requirements of a surrogate marker as outlined above. Consistent with this, the recent FINESSE criteria concluded that “no circulating blood marker has been demonstrated to be a valid outcome measure for clinical trials.” However with current advances in multiomics, promising makers may well emerge in the near future (713).

Cognitive function.

Cognition is an attractive marker to assess therapeutic efficacy. However, while currently used cognitive tests are very sensitive to the presence of vascular cognitive impairment in patients with cSVD, they have much lower sensitivity to change over time, particularly over follow-up durations of 2 to 3 years which might be used in a typical clinical trial. A number of prospective longitudinal cohort studies have found it difficult to detect change in cognition in cSVD over these time periods, with follow-up of up to 5 years or more required to detect a marked cognitive decline (676, 716). In the SPS3 trial in 3020 individuals with MRI-confirmed lacunar stroke, no change in cognition was detected over a 2- to 3-year period (717). Data from these longitudinal studies has been used to estimate sample size calculations for a clinical trial in patients with cSVD, and has suggested that thousands of individuals would be required to detect treatment effects based on the sensitivity of current cognitive tests to change (716). A number of factors may account for the low sensitivity; practice effects can last for a year or longer, and will reduce sensitivity to change (713). It is possible that computerized testing, and using parallel versions of the same test, may improve sensitivity, but this requires validation.

MRI.

In contrast to circulating biomarkers, and cognitive function, MRI surrogate markers offer considerable promise, and are being increasingly used in clinical trials in cSVD cohorts. A number of MRI markers have been proposed including WMH volume, lacunes, brain volume, and WM ultrastructure measured using DTI. WMH (695, 718), DTI (587, 716, 719), and brain volume (670) have been shown to be sensitive to change during follow-up periods of 2 to 3 years. Power calculations suggest that their use, depending on the effect size of the intervention, could reduce sample sizes to around 200–300 (676, 716, 718). Although the number of lacunes correlates strongly with cognition, the low frequency of incident lacunes means lacune count requires much larger sample sizes than the other MRI markers (676). There is strong evidence that WMH, brain volume, and DTI marker predict future dementia risk (587), but as yet much less data as to from clinical trials demonstrating that an intervention has the same effect on the surrogate end point as on the clinical outcome. However, the recent SPRINT-MIND (Systolic Blood Pressure Intervention Trial–Memory and Cognition in Decreased Hypertension) study demonstrated that intensive antihypertensive therapy reduced risk of the combined end point of dementia and mild cognitive impairment (696), and at the same time reduced WMH progression (695).

11. Summary and future directions

cSVD represents one of the major problems facing global society today. Perhaps as many as half of us may suffer some consequences from the disease, whether it be major complications such as ischemic stroke, ICH, or dementia, or one of the many more subtle consequences such as mild age-related cognitive impairment, gait disturbances, or psychiatric symptoms. Despite the importance of cSVD, we have no effective proven treatments other than risk factor modification. Why is this? An important factor is major lack of funding. The cost of vascular dementia has been estimated to be higher that of Alzheimer’s disease (720), yet vascular dementia research has long been underfunded compared to Alzheimer’s disease, which itself is underfunded compared with other diseases of similar impact (721). However, this has recently changed both with the realization of the importance of cSVD and its contribution to vascular dementia, and the appreciation that cSVD plays an important role in exacerbating the burden of neurodegenerative diseases such as Alzheimer’s disease. Hence, funding for Vascular Cognitive Impairment Dementia research in the United States is actually bundled together with Alzheimer’s disease research (722). Nevertheless, considerable extra focus and funding is required if we are to make progress.

Another factor has been neglect of this area by pharmaceutical companies. Recently they have shown intense interest in Alzheimer’s disease with a number of exciting new therapies, primarily targeting amyloid itself, which have been recently evaluated in phase 3 trials. However, there have been virtually no large pharmaceutical trials in cSVD and vascular dementia. This is partly related to a lack of promising novel therapies, but an important contributing factor has been a lack of well-defined trial methodology including robust endpoints accepted by the regulatory bodies. This has been addressed by the recent FINESSE guidelines (713), and hopefully these will encourage and facilitate randomized controlled trials in this area.

However, perhaps the most important reason for a lack of new therapeutic approaches, has been limited understanding of the underlying pathophysiology resulting in few potential therapeutic approaches other than risk factor modification. Recently, as described in this review, major progress in understanding the diversity and complexity of the underlying disease mechanisms has occurred thanks to novel approaches including advanced molecular, genetic, functional and imaging tools used in human or experimental models. Regardless of the initial environmental or genetic cause, dysfunction of small brain vessels occurs early, compromising the regulation of CBF, integrity of the BBB or brain fluid transport, to cause WM lesions and, likely in conjunction with segmental vascular pathology and in situ thrombosis, lacunar infarcts. All cells in the brain vasculature, i.e. endothelial cells, mural cells as well as resident cells of the perivascular space (PVM and PVFB) contribute to the pathophysiological changes, although the involved vascular compartments, i.e. artery, arteriole, capillary transition zone or veins and the nature of defects, i.e. dysfunction, loss or hyperplasia, differ between different forms of cSVDs and probably over the time. Matrisome disruption, dysfunction of ion channels and loss of arterial SMCs have been identified as key underlying mechanism occurring across sporadic and different genetic forms of the disease. Yet, there are still many challenges ahead for preclinical studies and outstanding questions to address to move the field forward. Among the most significant ones are the following:

1. Translation of genetic discoveries arising from genome-wide association studies into disease mechanisms. This includes the identification of the causal genes, cell type(s) and mechanisms by which risk variants exert their effect.

2. Development of innovative animal models of “sporadic cSVD”. One possibility could be to generate rodents expressing a combination of selected genetic risk variants of cSVD and to challenge them with longstanding, non-malignant hypertension.

3. Better and deeper phenotyping of clinically relevant cSVD animal models. Well-powered longitudinal or cross-sectional cohorts across different ages, with all data (positive and negative) being reported, should be given a higher priority (723). These studies should include: i) an in-depth analysis of the entire brain vasculature comprising structural and molecular analysis of the different microvascular compartments using modern histopathological approaches like 3D-imaging and single cell RNA sequencing, ii) an analysis of resting CBF across all brain regions and of the main brain vessel functions, promoting approaches conducted in awake animals, iii) an assessment of the integrity of the parenchyma across all brain regions using high resolution structural and diffusion MRI that most closely aligns with human studies and iv) an evaluation of executive functions (working memory, processing speed, attention, cognitive flexibility, planning), instead of conventional behavioral tests that have limited translational value; the touchscreen technology can perform high-level cognitive assessment cognition in rodents, with a high degree of automation and standardization, in a way that is directly relevant to humans (724, 725). Machine learning algorithms, a constantly evolving field, is a promising approach to tackle the common bottleneck associated with the processing, analysis and integration of all these morphological, molecular, functional and behavioral data.

4. Identification of molecular mechanisms of mural cell defects in cSVDs. Identifying shared mechanisms between distinct cSVDs could have a huge impact. Another major challenge is a better understanding of the mechanistic chains linking structural and functional changes of brain vessels to brain lesions, especially WM lesions and lacunar infarcts. Here, animal models which closely recapitulate human diseases will be critical.

5. Implementation of preclinical trials. A paradigmatic shift in the way these trials are conducted is needed with respect to the clinical relevance of the animal model, the robustness of the working hypothesis, well-defined endpoints with translational value, sample size calculation and all other parameters to ensure rigor and reproducibility. This should include implementation of lessons learnt from the STAIR criteria for acute stroke (726) including animal blinding and randomization, testing across different models and in multiple laboratories in a multicenter design, as well as testing therapies in the presence of normal aging and co-morbidities, paralleling the methodology applied to human clinical trials (723).

With respect to human studies, novel structural and functional MRI imaging approaches offer promise that defects in small brain vessel function and abnormalities in the WM microstructure can be detected at an earlier stage. These new insights have identified a number of novel therapeutic avenues as highlighted in this review, and thanks to ongoing and upcoming studies new molecular targets will come soon, which will now need following up, with the aim of translating new therapies into human for this devastating disease.

Figure 15. Associations of conventional risk factors with lacunar stroke derived from Mendelian randomisation analyses.

Estimates are represented as odd ratios per genetically proxied increase in each risk factor. Reproduced from (197). Used with permission under CC-BY 4.0 license.

Clinical Highlights.

• Cerebral small vessel disease (cSVD) is a major health problem causing lacunar stroke, intracerebral hemorrhage and dementia.

• Characteristic appearances are visible on MRI and include white matter hyperintensities, which are present in many middle aged and older individuals, and predict future stroke and dementia risk.

• cSVD is classified into two categories; Cerebral amyloid angiopathy (CAA), characterized by the ß-amyloid deposition in the wall of small cerebral vessels and most commonly presenting with intracerebral hemorrhage, and non-amyloid cSVD comprising a group of common pathologies related to aging, hypertension, or genetic factors. This review covers non-amyloid cSVD.

• Cardiovascular risk factors, particularly hypertension, increase the risk of cSVD but fail to account for much of the risk, some of which is explained by genetic factors.

• There are a number of monogenic forms of cSVD, the most common of which is CADASIL caused by NOTCH3 mutations. Rare variants not only cause typical familial diseases but also contribute to the burden of apparent sporadic cSVD, emphasizing that both monogenic cSVD and sporadic cSVD have shared mechanisms.

• Using information gained from clinical and experimental studies, we discuss how cSVD risk factors (aging, hypertension and genetic mutations) affect the structure and function of small brain vessels and the potential mechanisms by which cSVD causes white matter lesions, lacunar stroke, hemorrhage and dementia.

• Current treatment approaches are limited to risk factor modification, particularly blood pressure lowering to a systolic blood pressure of 120–130mmHg, which has been associated with reduced white matter hyperintensity progression, and a lower risk of stroke and of dementia.

• Clinical trials in well-phenotyped cases of cSVD are required to better refine treatment approaches. We discuss methodology for testing therapies, including the use of MRI surrogate markers to assess treatment efficacy.