Pericyte Control of Blood Flow across Microvascular Zones in the Central Nervous System

Abstract

The vast majority of the brain’s vascular length is composed of capillaries, where our understanding of blood flow control remains incomplete. This review synthesizes current knowledge on control of blood flow across microvascular zones, by addressing issues with nomenclature and drawing on new developments from in vivo optical imaging and single cell transcriptomics. Recent studies have highlighted important distinctions in mural cell morphology, gene expression, and contractile dynamics, which can explain observed differences in response to vasoactive mediators between arteriole, transitional and capillary zones. Smooth muscle cells of arterioles and ensheathing pericytes of the arteriole-capillary transitional zone control large-scale, rapid changes in blood flow. In contrast, capillary pericytes downstream of the transitional zone act on slower and smaller scales and are involved in establishing resting capillary tone and flow heterogeneity. Many unanswered questions remain, including the vasoactive mediators that activate the different pericyte types in vivo, the role of pericyte-endothelial communication in conducting signals from capillaries to arterioles, and how neurological disease affects these mechanisms.

Brain microvasculature

The brain is the most metabolically demanding organ, consuming 20% of the total body’s oxygen while weighing only 2% of its mass. The brain relies on a continuous supply of blood since it lacks a local energy reserve, and the blood is delivered by an astounding total vascular length of 400 miles in the human brain (1). While arteriole and venous networks are visible to the naked eye, these larger caliber vessels are dwarfed in number by dense microscopic capillaries serving as the blood’s vast distribution network.

Clinical imaging cannot yet resolve individual capillaries in vivo. However, capillary perfusion can be appreciated from digital subtraction angiography, where a contrast agent transiting the microvasculature creates a ubiquitous “blush” across the hemisphere (Figure 1A). This blush predominantly represents flow through capillaries. It is here that the blood fulfills its role to supply oxygen and remove waste from the brain. Strikingly, the contrast agent perfuses the collective 400 miles of vasculature in less than 5 seconds, highlighting the efficiency of this microcirculatory process. In various diseases that affect the cerebrovasculature, the transit of blood cells through the capillary bed becomes impaired, contributing to brain dysfunction and neurodegeneration (2; 3).

Figure 1. Capillary perfusion and structure in the human brain.

(A) Schematic of major extracranial and intracranial arteries perfusion the brain are shown on the left. Digital subtraction angiography images on the right show movement of contrast agent as it enters the cerebral arteries (2 sec post-injection in the right internal carotid artery), transits capillaries (5 sec), and then exits through the venules and veins (7 sec). The image at 5 seconds represents capillary “blush” as the contrast agent is visible, but the individual microvessels perfused are below the resolution limit of the imaging technique. (B) Image of pial vasculature on the surface of the human cerebral cortex (Adapted from (4)). (C) Micrograph of a tissue slice from human cortex, showing a cortical penetrating vessel (arteriolar or venular origin unclear) surrounded by the capillary bed (Adapted from (4)).

Elegant anatomical studies by Duvernoy used India ink injections to reveal the elaborate vascular networks of the human brain (4). He imaged and depicted the arterioles and venules of the brain surface (Figure 1B), and the underlying capillary networks (Figure 1C), which could be densely packed because each capillary measured only about 5 μm in diameter, i.e., one-tenth the diameter of a human hair. These studies made it clear that capillaries represent the great majority of the brain’s vascular length, and that their function was therefore most important to understand.

Early histological studies led many to hypothesize that capillaries were involved in cerebral blood flow dynamics such as functional hyperemia, a phenomenon in which neural activity leads to increased local blood flow. Functional hyperemia is achieved by communication between neurons and vessels in a process called neurovascular coupling (5). At the level of brain arterioles, neurons communicate directly or indirectly through astrocytes to control tone of arteriolar smooth muscle cells (SMCs). Similarly, the current theory is that neurons or astrocytes also communicate with mural cells of vessels downstream of arterioles, called pericytes, to control blood on a finer scale (6). Fundamentally, there is logic to this concept as autonomous control of capillary diameter by pericytes could lead to more precise delivery of blood, making the redistribution of blood cells and delivery of oxygen more efficient.

Pericytes are the ideal candidate for regulation of capillary diameter. They line the walls of all microvessels that span between arterioles and venules, and have similar abluminal position as SMCs, which are known to exert powerful influence over vessel diameter and flow. Pericytes are embedded within the basement membrane of the capillary wall and form communicatory junctions (gap and adherens junctions) with endothelial cells at sites where the basement membrane thins (7). Through communication with the endothelium, pericytes serve diverse functions in brain vascular health, and these functions have been surveyed by many recent reviews detailing classic pericyte roles in blood-brain barrier establishment, blood flow control, and angiogenesis (8–11). Novel roles in regulation of brain immune responses are also emerging (12).

We focus this review on the data surrounding the hypothesis that pericytes control cerebral blood flow in vivo. This hypothesis has proven challenging to address for over a century due to many factors, including (1) a dearth of tools to unequivocally identify and genetically target pericytes on capillaries, (2) inconsistent nomenclature of pericyte subtypes, (3) the correlative nature of observational in vivo studies, and (4) lack of spatial resolution to resolve small changes in capillary diameter. However, recent advances have overcome some of these challenges, yielding novel data from ex vivo and in vivo preparations. The goal is to synthesize these findings and navigate past misconceptions in the field to distill key concepts and next steps.

Historical aspects of pericytes in blood flow control

Pericytes have been proposed to modulate capillary blood flow since their discovery in the frog by Eberth in 1871 (13) and Rouget in 1873 (14). This potential role of pericytes later collided with research showing that capillary blood flow is dynamic. August Krogh discovered that capillaries in skeletal muscle become perfused in response to muscle activity, a discovery that culminated in his Nobel Prize; see Poole et al. for review of Krogh’s work (15). An excerpt from Krogh’s speech accepting the Nobel Prize for “the capillary motor regulating mechanism” aptly describes the field in 1920, and the field now, over 100 years later:

There must be a special capillary-motor system by which the bore of the capillaries can be regulated, but stating that fact raises a whole series of new questions: is variation in diameter of capillaries independent of the arteries, or does it follow from them? In what way can the capillaries be excited – chemical, electrical or mechanical? Are they under nervous control, and, if so, by which nerve? (16)

Based on Krogh’s work, there was consensus that capillary blood flow was heterogeneous and dynamic, but questions remained as to why and how. Pericytes emerged as answers to these questions, launching several epochs of pericyte research, each characterized by new methodological advancements and fueled by controversy.

One initial controversy asked whether endothelial cells or pericytes were responsible for the observed narrowing of capillaries in frog nictitating membrane or rodent mesentery. Many groups observed focal changes in capillary diameter, but there was disagreement as to whether these changes were correlated with the location of pericytes, identified by their abluminal position on the capillary wall (17). In moving from observation to manipulation, the next wave of experiments activated mesenteric pericytes by touching them or electrically stimulating them with a glass micropipette (18). Some groups observed localized contraction when stimulating pericytes, while others did not.

With the advent of transmission electron microscopy, several groups noted electron-dense bands of filaments in pericytes that abutted endothelial cells in the brain and retina, which in some cases stained for actin, non-muscle myosin and tropomyosin (17). Scanning electron microscopy confirmed and elaborated upon earlier drawings of mural cells, showing that mural cell morphologies in the brain and retina exist in a continuum, which serves as the basis of ongoing disagreements about how to refer to different vascular segments and their associated mural cells. As Sims said in 1986, “a continuum of cell morphologies will occur in these regions, so that discrepancies of cell names will inevitably arise.”

Cell culture studies next suggested that pericytes, and not endothelial cells, were contractile and responsive to known vasoactive agents (19; 20). However, the morphology of pericytes, and expression of some contractile proteins (21), changes drastically under culture conditions, making it difficult to translate these results into the in vivo functions of pericytes.

Thereafter, ex vivo studies in retina confirmed that pericytes are excitable cells that can modulate vessel diameter over the course of minutes in response to vasoactive molecules such as angiotensin II, endothelin, and ATP (22; 23). The modern era of pericyte and blood flow research was ushered in by landmark ex vivo studies showing that direct pericyte stimulation, neurotransmitters, and ischemia modulate vessel diameter at pericyte locations in the brain and retina (24). With this work, pericytes became implicated in functional hyperemia, and in states of pathologically reduced blood flow such as stroke, dementia, and diabetic retinopathy. Many studies then began to investigate pericyte function and dysfunction in new ways, with new technologies including in vivo two-photon microscopy.

Despite the diverse experimental approaches and accumulating evidence, it remained unknown if pericytes were an answer to August Krogh’s question: How do capillaries change their blood flow to meet metabolic demands of the tissue? However, the proliferation of new tools such as two photon imaging, mouse models for targeting pericytes, and optogenetics permitted detailed in vivo investigations of the role of pericytes, adding new insight into a century old question.

Vascular and cellular structure

While there continues to be inconsistent nomenclature in the field (see these reviews for nomenclature variations (11; 25)), there is sufficient information at present to distinguish mural cell types in different vascular zones. One goal of this review is to clarify these mural cell types as seen in rodent cerebral cortex, and to propose a nomenclature that suitably distinguishes each cell type and the vascular zone it occupies. These distinctions are key because, if carefully considered and reported, they will enable different laboratories to compare and contrast their results.

The preponderance of data indicates there are at least four types of mural cells, which define four major microvascular zones (to be discussed in further detail below): SMCs on arterioles, ensheathing pericytes on the arteriole-capillary transition (ACT), capillary pericytes (displaying mesh and thin-strand processes) on capillaries, and stellate SMCs on venules (Figure 2A). While we depict mesh pericytes on post-capillary venules, it is currently unclear if these cells represent a meaningfully distinct population. Our suggested nomenclature identifies mural cell and vessel types based on their morphology and location, and not their function. This is because ascribed functions may change as we learn more. For example, the terms “contractile” versus “non-contractile” pericytes becomes a source of confusion, because mounting data suggests that all pericyte types are contractile, but function on different time-scales (26–28).

Figure 2. Mural cell types across the microvascular network and their contractile abilities.

(A) Schematic showing different zones of the brain microvasculature, including pial arteriole, penetrating arteriole, arteriole-capillary transition (ACT), capillary, post-capillary venule, ascending venule, and pial venule. The pre-capillary sphincter is occasionally seen at the start of the ACT zone. The appearance and names of mural cells occupying these distinct zones, as seen from sparsely labeled (NG2-tdTomato or Myh11-tdTomato) mice, are shown. (B) Mural cell type and corresponding vascular branch order, with small molecule/immunohistochemical stains that aid in distinguishing the zones. (C) Dilatory and contractile features of mural cells differ across microvascular zones. SMCs and ensheathing pericytes are very contractile. Capillary pericytes are contractile, but function on a slower time-scale. This may be a result of differential expression in genes that encode proteins for contractile machinery or control the phosphorylation of myosins. Key receptors for agonists that engage mural cell contraction via ATP, thromboxanes, and endothelin-1 are broadly expressed in mural cells across vascular zones. The gradients are qualitative only and derived from observations of single cell transcriptomic data of Vanlandewijck et al. (62).

Naturally, these vascular zones can vary across different tissue beds. For example, the cortex has distinct vascular structure and physiology than the hippocampus (29). However, recent in vivo studies from brain have focused on the upper layers of cerebral cortex (mostly sensory cortex), providing a common framework for discussion. In this vascular network, branching order from the cortical penetrating arteriole (denoted zero order) has been the primary means for communicating the location of microvessels within the network (Figure 2A,B). Penetrating arterioles send small offshoots that form the entryway into the capillary bed; these offshoots are denoted 1^st order. Thereafter, branching order increases by one at each bifurcation. When assigning a branch order to all capillaries, the most numerous are branch orders 3 to 6 (30; 31), but branch orders up to 9^th order can be reliably traced during in vivo imaging (26).

Mural cells exhibit diverse morphologies as cortical penetrating arterioles transition to capillaries (32). Penetrating arterioles are covered in concentric ring-like SMCs that are rich in ɑ-smooth muscle actin (ɑ-SMA; encoded by the acta2 gene), a component of the actomyosin contractile machinery (Figure 2A)(33). The penetrating arterioles also have a layer of elastin that is labeled by Alexa 633 hydrazide (Figure 2B)(34). These attributes facilitate rapid and robust contraction and relaxation of the penetrating arteriole during blood flow regulation.

Offshoots from the penetrating arteriole, and loss of the elastin layer marks the junction between the penetrating arteriole and ACT zones (Figure 2A,B). In 20–50% of the cases, annular ɑ-SMA-positive sphincter cells are positioned at the mouth of the offshoot (35). These sphincters maintain a localized region of constriction to buffer the capillary bed from pressure fluctuations occurring upstream, and also dilate during functional hyperemia to promote blood flow to the capillary bed.

Just downstream of the sphincter, if present, the vessel if covered by ensheathing pericytes. Ensheathing pericytes express ɑ-SMA like SMCs but possess hybrid morphology of both SMCs and classic pericytes of the capillary bed. Their processes enwrap the entire endothelial tube like SMCs, yet they are elongated and exhibit protruding cell bodies like classic pericytes (33). Sidebar 1 here.

Side bar 1: Nomenclature clash.

Ensheathing pericytes have been referred to by a variety of names in the literature. They have been called pre-capillary SMCs, aaSMCs, which creates the view that they are identical to SMCs, or call simply pericytes, which conjures images of classic pericytes of capillaries. This is the core of the controversy in recent literature on pericytes (105). Adding more confusion, the nomenclature for the zone occupied by ensheathing pericytes has also varied (pre-capillary arteriole, post-arteriole transitional zone, or simply “capillary”). Here, we support the use of the term “arteriole-capillary transition”. This is because pericytes, by definition, cannot be located on arterioles, as there is where SMCs reside. Further, the all-encompassing term “capillary” for vessels 1^st order and beyond obscures functional differences between the transitional and true capillary zones.

As microvessels branch further, ɑ-SMA expression drops abruptly to low or undetectable levels (26; 33; 36; 37). This marks the junction between the ACT and the true “capillary” zone, and the shift from coverage by ensheathing pericyte to capillary pericytes with classic pericyte morphology. The ACT-capillary junction can occur anywhere between branch orders 1^st to 4^th, with offshoots of larger diameter at the 1^st order brain exhibiting ɑ-SMA expression out to higher branch orders (33).

Capillary pericytes are unlike ensheathing pericytes, in part, due to a stark difference in ɑ-SMA content. Morphologically, capillary pericytes extend thin, longer processes that incompletely cover the endothelium. Their processes can reach out to contact hundreds of micrometer of capillary length, unlike the short-range coverage of ensheathing pericytes (25; 33). As such, capillary pericyte processes may span multiple capillary segments, and their processes bifurcate at capillary junctions. Closer to the ACT and in the post-capillary venule zone, capillary pericyte processes take on a mesh-like appearance, though still incompletely covering the endothelium. However, in the mid-capillary regions that are the majority of the capillary length, pericytes exhibit thin-strand processes. Mesh and thin-strand pericytes are difficult to objectively distinguish, and represent a true continuum in morphologies (33).

The precise boundary between the ACT and capillary zones remains difficult to pinpoint in vivo unless labels for pericyte morphology, ɑ-SMA, or capillary pericyte-specific dye (Neurotrace 500/525), are used (36; 38). Basal vessel diameter alone is not a reliable means to delineate the vascular zones in vivo, as there is overlap in diameters of ACT vessels and capillaries (33). In perfusion-fixed tissues, a cut-off of 7 μm diameter has been used to categorize capillaries in mouse brain connectome data (31), which is reasonable based on studies that have considered diameter alongside ɑ-SMA and pericyte labeling (33). However, the contraction of ensheathing pericytes in vivo may in fact contract ACT vessels to become smaller than capillaries (36). Using only branch order, one can identify the capillary bed (>5^th order) with confidence, based on studies that have carefully charted the probability of ɑ-SMA expression as a function of branch order in mouse cortex (26; 33; 36; 37).

Ensheathing and capillary pericyte somata also vary in their locations, with somata at capillary junctions (junctional) or on intervening portions of the capillary (en passant)(32; 39). Junctional capillary pericytes are more prevalent. This may be a consequence of capillary network construction during early postnatal development, where the sprouting of new capillary branches occurs preferentially at pericyte somata (40).

In summary, we recommend the following nomenclature based on current data, and use it as a basis in discussions of mural cell physiology in vivo: SMCs on penetrating arterioles, ensheathing pericytes on the ACT, capillary pericytes on capillaries, and stellate SMCs on venules. Critically, this system avoids confusion between the two functionally distinct pericyte types (ensheathing and capillary) that exist downstream of the penetrating arteriole

Dynamics of vascular segments

Vascular smooth muscle cells

Smooth muscle cells surround brain arterioles, of which there are two primary types that serve cerebral cortex: pial arterioles and penetrating arterioles. Pial arterioles sit between the arachnoid and pia mater and are distant from cells of the parenchyma. As discussed above, penetrating arterioles are within the brain parenchyma and thus intimately in contact with astrocytic endfeet and the metabolites they release (41).

At baseline in awake rodents, the diameter of pial arterioles ranges between 8 to 80 μm (42; 43). Penetrating arterioles range from 6 to 40 μm, with an average ~10 μm (35; 44; 45), There are many factors governing basal diameter, such as proximity to the sourcing arteriole, blood gases, blood pressure, and vasoactive mediators released through local brain activity, such as NO from interneurons (46) and astrocyte derived signals (47). Pial and penetrating arterioles exhibit vasomotion, where lumen diameter oscillates by about 10% at low frequency (~0.1 Hz)(36; 48). This is partly driven by slow fluctuations in γ-band power from neural activity (48), and partly by intrinsic rhythmicity of smooth muscle cells (49).

Hypercapnia is a strong vasodilatory stimulus. From their baseline diameter, pial arterioles (50) and penetrating arterioles (51) have been reported to increase in diameter by ~20% and 15%, respectively, in response to several minutes of hypercapnia. During sensory stimulation in awake mice (vibrissal stimulation), pial arterioles can dilate up to a peak of ~20% in a few seconds (average of ~10%)(42). Similarly, penetrating arterioles can increase in diameter by ~10–15% within a few seconds, in response to sensory stimuli (Figure 2C, 3A)(52). Thus, SMCs provide robust and rapid control of pial and penetrating arteriole diameter in vivo.

Figure 3. Effect of vessel diameter change across microvascular zones.

(A) A large region of neuronal activation leads to production of diffusible vasodilatory signals and conductive hyperpolarization from the capillary bed. This leads to broad dilation of the ACT, pre-capillary sphincter, penetrating arteriole and pial arterioles, and widespread hyperemia. (B) Capillary pericytes provide tone and contribute to heterogeneity in capillary diameter and blood cell flux. During functional hyperemia, flux among capillaries homogenizes by increasing flow through initially low flux capillaries. This process improves extraction of oxygen by simply redistributing blood within the capillary network, in addition to greater blood supply from upstream arterioles. (C) Junctional pericytes at the ACT exhibit preferential dilation of specific daughter branches based on direction of arriving conductive signal. This allows finer tuned control of hyperemia within the territory of the penetrating arteriole.

Ensheathing pericytes

Ensheathing pericytes have emerged as important cells because they are very dynamic at rest and dilate rapidly during sensory stimulation (Figure 2C, 3A) (36; 53-55). The ACT zone on which they reside generally has a baseline diameter range from 2 to 10 μm, with an average of ~5 μm in vivo (26). Similar to SMCs on arterioles, ensheathing pericytes exhibit vasomotor oscillations, with diameter changes on the order of 10% in the awake mouse brain (36; 38).

From their baseline state, vessels in the ACT zone have the capacity to dilate by ~25% in response to hypercapnia in anesthetized mice (26; 56). During sensory stimulation of alpha-chloralose anesthetized mice, Hall et al. showed the first order branch within the ACT dilated by ~5%, taking ~15 s to reach peak diameter (54). Critically, this dilation was slightly faster than that of neighboring penetrating arteriole, revealing that ensheathing pericytes are first to respond in promoting regional blood flow increase during neurovascular coupling (Figure 2C, 3A). Follow up studies corroborated these findings and added that the pre-capillary sphincter was also highly responsive to whisker stimulation, both of which dilate slightly earlier than neighboring penetrating arterioles (35; 55). Similar observations were made in the olfactory bulb in response to a brief odor presentation (53).

Capillary pericytes

Having firmly established that ensheathing pericytes of the ACT robustly control blood flow during neurovascular coupling, we now direct our attention to data on capillary pericytes, which by our definition are pericytes downstream of ensheathing pericytes, with low to undetectable levels of ɑ-SMA expression. Capillary baseline diameters are very heterogeneous, ~1.5 – 8 μm, and these diameters are quite stable (26; 57). At baseline, there is no discernible vasomotion in capillaries, in contrast to pial/penetrating arterioles and the ACT (36). The heterogeneity in capillary diameter contributes to variance in blood cell flux among capillaries within a network under basal conditions (discussed further below)(26).

With the strong stimulus of hypercapnia, capillaries clearly dilate by ~5–10% over the course of several minutes (26; 50; 56; 58). This dilation may not simply be a passive effect of upstream pressure, which would affect all downstream capillaries similarly. Hypercapnia produces homogenization of flow among capillaries, such that smaller capillaries show disproportionately greater dilation and increases in blood cell flux than capillaries with large baseline diameter (50; 58). Further, capillaries constrict back to baseline after hypercapnia at a slower rate than upstream vessels, suggesting independent action by capillary pericytes (26).

During sensory stimulation, studies seem to be consistent in showing that capillaries exhibit small and slow dilations on the order of 1–2% from baseline (36; 37; 53; 59; 60). These diameter changes are very small, questioning whether they are truly resolvable with in vivo two photon microscopy. Yet, the consistency of this observation among laboratories suggests it is worth deeper investigation. Interestingly, capillary pericytes (thin-strand morphology) decrease their frequency of calcium transients in response to neural activity simultaneous with diameter changes, which could contribute to relaxation of contractile machinery (53). Indeed, a small dilation of capillaries was necessary to achieve a full functional hyperemic response according in modeling studies (53). Further in silico studies suggest that very small capillary dilations produces hemodynamic responses that are much more tightly localized near neural activity than scenarios where capillaries are rigid (61).

Thus, capillaries may exhibit small and delayed dilations during neurovascular coupling, and this change may be permissive for blood cell passage during neurovascular coupling (Figure 3B). However, the existing experimental data comes short of confirming an active role of capillary pericytes in local blood flow control during functional hyperemia. This is an important open question, as small dilations could exert a large effect when multiplied across many capillaries, particularly if they occur at high resistance points within the capillary network. Sidebar 2 here.

Side bar 2: Pericytes and autoregulation.

There is some evidence for autoregulation or compensatory dilation at the capillary level. Stenosis of upstream vessels in atherosclerosis (106) and mechanical carotid artery occlusion (107) models produces dilation of capillaries. This may correlate with improved collateral flow that has been observed over time in humans with stenosis of intracranial arteries (108).

Potential contractile machinery in capillary pericytes

Single cell transcriptomic studies indicate that capillary pericytes express receptors for vasoactive mediators (discussed further below), but also express some genes involved in SMC actomyosin contraction such as myosin (myh11, myl9, myh9), regulators of myosin phosphorylation state (mylk and ppp1r12a, rock1), and L-type voltage-gated calcium channels (cacna1c)(62)(Figure 2C). However, smooth muscle actin transcripts (acta1, acta2, actg2, actc1) expected to bind myosin proteins are low to undetectable in expression. Nevertheless, low ɑ-SMA content may still be sufficient for contractility via canonical actomyosin cross-bridging, which may explain the slower kinetics of capillary pericytes. Recent studies in retina showed that immunohistological detection of low ɑ-SMA may be enhanced by special fixation procedures, and blockade of ɑ-SMA expression limits contraction of retinal capillary pericytes (63; 64). It is also possible that weak actomyosin contraction is coupled to F-actin polymerization, known to promote constriction in SMCs through RhoA-Rho kinase signaling.

Conductive responses

Active neurons in the brain parenchyma receive increased blood flow from arterioles that may be located hundreds of micrometers away. Recent studies demonstrate that capillary endothelial cells detect neural activity and communicate the need for increased blood flow to arterioles upstream (Figure 3A). Longden et al. reported that extracellular K⁺, released as a byproduct of neural activity, leads to endothelial hyperpolarization through activation of inward rectifier potassium family, subtype 2 channels (Kir2)(65). Endothelial cells conduct this hyperpolarization upstream, from capillaries to the ACT zone, where electrical transfer through myoendothelial junctions leads to rapid vasodilation (Figure 3A,C, 4B).

Figure 4. Signaling and contractile machinery leading to different dynamics across vascular zones.

(A) Schematic depicting how vasoconstrictive agonists act on mural cells across microvascular zones, simplified to discuss arteriole (SMC), ACT (ensheathing pericyte), and capillary zones (capillary pericyte). The ion channel or G-protein coupled receptors for the agonists are expressed at varying levels in these mural cells (based on single cell transcriptomic data), and the qualifiers “low,” “med”, and “high” are appended as subscript to each protein to show general expression level. Intracellularly, the expression levels of actomyosin contractile machinery components also varies across mural cell types. The contractile capacity of each mural cell type is based on observations from pharmacological studies and in vivo cell-specific optogenetic manipulations (26). (B) Dilatory agonists and their receptors with likely pathways leading to relaxation of the mural cell. Endothelial conductive hyperpolarization from capillary-to-arteriole promotes relaxation of upstream mural cells via transmission of hyperpolarization through myo-endothelial junctions. The speed of dilation is, in part, driven by intravascular pressure at each vascular zone. Definitions: ɑ/βAR = ɑ or β adrenoreceptor, A2a = A2a purinergic receptor, AA = arachidonic acid, ATP = adenosine triphosphate, cAMP/PKA = cyclic AMP-protein kinase A, EP4 = prostaglandin E₂ receptor 4, ET-1 = endothelin-1, ET_AR = endothelin A receptor, K_Ca = calcium-activated K⁺ channel, K_ATP = ATP-sensitive K⁺ channel, Kir2 = inward rectifying K⁺ channel 2, MLCK = myosin light chain kinase, MLC = myosin light chain, MYPT1 = myosin light chain phosphatase, P2X = P2X purinoreceptor, P2Y = P2Y receptor, PGE2 = prostaglandin E₂, PGH2 = prostaglandin H₂, RhoA/ROCK = RhoA-Rho kinase, sGC-cGMP = soluble guanylyl cyclase-cyclic guanosine monophosphate, TR = thromboxane receptor, TXA2 = thromboxane A2, VGCC = voltage-gated calcium channel.

In a parallel pathway, neurons activate endothelial TRPA1 channels, which conducts hyperpolarization to upstream arterioles more slowly than Kir2-mediated signals (66). The slower kinetics are likely due to intermediate steps that involve TRPA1 activation of endothelial pannexin (Panx1) proteins, which release ATP that then activates calcium-permeable P2X receptors on neighboring endothelial cells. In these novel capillary-to-arteriole signaling mechanisms, the capillary bed functions as an extensive sensory web to detect and transmit neuronal signals during neurovascular coupling (Figure 4B).

Pericytes are known to express gap junction proteins (Cx37 and Cx45) and new evidence in retina suggests that pericytes utilize them to communicate with endothelial cells, astrocytes, and other pericytes (67–69). There is also a clear anatomic basis for pericyte-endothelial contact, as pericytes and endothelial cells form interdigitating ‘peg-and-socket’ junctions that have recently been demonstrated with 3-D electron microscopy techniques (70). Interestingly, there is a directional bias in gap junction-mediated dye transfer from capillary pericytes towards upstream arterioles, further suggesting a role of pericytes in communicating dilatory signals to upstream arterioles (71).

A recent study showed that ensheathing pericytes may preferentially gate flow down specific ACT branches, based on the direction of the incoming hyperpolarizing signal (39). By dilating just one branch at a time, ensheathing pericytes provide a more economical means of providing blood flow when compared to the level of the whole penetrating arteriole (Figure 3C).

Cause-and-effect studies of blood flow control by capillary pericytes

To clearly dissect the role of capillary pericytes, new approaches were needed for selective pericyte manipulation in vivo without affecting mural cells of upstream vessels. In recent years, three cause-and-effect approaches were implemented to study capillary pericytes: single cell optogenetics, single cell optical ablation, and genetic ablation using diphtheria toxin.

Hill and colleagues were the first to apply optogenetics to stimulate mural cells in vivo, focusing on the question of which vascular zones were contractile and could contribute to functional hyperemia (36). These studies expressed the H134R variant of ChR2 in all mural cells and used two-photon ChR2 excitation to stimulate mural cells of specific zones in a spatially restricted manner. They discovered that arteriolar smooth muscle cells contracted readily when stimulated. However, capillary pericytes produced no change in capillary diameter with similar stimulation over tens of seconds, leading to the conclusion that they were incapable of regulating local blood flow in capillaries.

Hartmann et al. reassessed the contractile ability of capillary pericytes using the same approach (26). They found that capillary pericytes were indeed contractile, but required longer and more intense stimulation, due to their slower dynamics. Their contraction led to reductions in blood cell velocity and flux. These findings are not at odds with the work of Hill et al. (36) and added further detail to the distinct kinetics of different vascular zones in cortex. Further, the slow dynamics of capillary pericytes were inhibited by the vasodilator fasudil, suggesting capillary pericytes use a form of contractile machinery similar to SMCs. Similar probing of upstream ɑ-SMA-positive ensheathing pericytes and SMCs led to rapid contractions, consistent with findings of Hill et al.; note that pre-capillary SMCs in the study by Hill et al. are synonymous with ensheathing pericytes. Additionally, a separate study by Nelson et al. expressed ChR2 in capillary pericytes with a cell-specific Cre driver (Pericyte-CreER), and also found slow contraction of pericytes in response to ChR2 activation (27).

Demonstrations that capillary pericytes could contract in vivo were complemented by studies showing that optical ablation of capillary pericytes led to vasodilation. Berthiaume et al. optically ablated capillary pericytes in the cortex of adult mice, and found that it led to consistent dilation of regions lacking pericyte contact (25). This dilation resulted in aberrantly high levels of blood cell flux in the uncovered capillary, consistent with a role in maintenance of basal vascular tone (26). Tone was regained once pericyte contact was re-established by the growth of neighboring pericytes. This effect was not restricted to laser-induced pericyte loss, as capillary dilation was seen in response to pericyte loss in a model of epileptic seizure (72). In this model, tone was regained with gradual pericyte remodeling and endothelial contact, and also after accelerated remodeling through exogenous administration of PDGF-BB. Collectively, these data provide strong evidence that capillary pericytes are necessary and sufficient for basal capillary tone in vivo.

Using a genetic strategy, Nikolakopoulou et al. examined cerebral blood flow changes in response to global ablation of capillary pericytes (73). Here, capillary pericytes were ablated en masse using Pericyte-CreER mice crossed with diphtheria toxin receptor expressing mice. Extensive pericyte loss was accompanied by marked decrease in cerebral blood flow, most likely due to cerebral edema and tissue swelling. However, whether contractile tone of capillary pericytes changed in vivo was difficult to ascertain with the scale of injury. Global pericyte ablations and optical single pericyte ablations lie at two ends of the spectrum, and provide insight into how different scales of pericyte loss can affect cerebral hemodynamics.

Capillary heterogeneity

As briefly discussed above, capillary flow is heterogeneous at rest, which provides reserve space for redistribution of blood in capillary networks during functional hyperemia (74). As blood flow increases, the flux among capillaries homogenizes (Figure 3B). This mechanism has been strongly supported by in vivo imaging studies, which have revealed that flux among capillaries homogenize with sensory evoked functional hyperemia (59; 75-78). In particular, low baseline flux capillaries rapidly increase in flux, while high baseline flux capillaries tend to decrease slightly in flux. This homogenization of flow is expected to improve the efficiency of oxygenation extraction from the blood, and this has been described theoretically (79) and shown empirically (80).

The homogenization of capillary blood flow implies an active mechanism whereby pericytes covering small diameter capillaries act differently than those on large diameter capillaries. This mechanisms may work in concert with increased compliance of red blood cells that carry more oxygen (37), with the end goal of promoting blood cell passage through the smallest diameter capillaries of the network during functional hyperemia.

Related to these findings from sensory cortex, recent in vivo imaging studies in retina show that some pericytes link neighboring capillaries together with tunneling nanotubes (69). These “bridging” pericytes selectively increase flow in one capillary while decreasing flow in others. The capillary diameter changes associated with these flow changes in retina are also small, averaging 1–2% from baseline, similar to that described in cortex.

Vasoactive mediators known to act on pericytes

While cause-and-effect studies define the capabilities of pericytes in vivo, they do not inform us about the physiological drivers that engage pericytes. An excellent review from Hariharan et al. explored the battery of ion channels and GPCR receptors expressed by pericytes using single cell transcriptomic data from mouse brain (81). This revealed numerous signaling mechanisms through which pericytes might interact with their environment. While many pathways remain to be clarified by rigorous experimental studies, some of the findings reinforce more than a decade of research identifying vasomediators of pericyte contraction and dilation, as reviewed in detail by Hamilton et al. (6). We begin by discussing vasoactive signals that predominately lead to vasoconstriction of pericytes (Figure 4A).

ATP

ATP is a purine released by many cell types and has well-established effects on the tone of arteries and arterioles, both constrictive and dilatory. Application of ATP to isolated whole retina induces the contraction of pericyte somata around capillaries (24; 82). Direct injection of ATP into the brain parenchyma leads also to potent vasoconstriction in the ACT zone, with slower and more prolonged responses occurring with higher branch orders (55). ATP injection in vivo also led to delayed vasodilation, possibly because of metabolism of ATP to adenosine (55). ATP may act directly on purinergic receptors expressed by pericytes, and transcriptomic data confirms the expression of P2X1 (p2rx1) and P2X4 (p2rx4) by some but not all sampled ensheathing and capillary pericytes (62). The activation of these receptors would cause flux of Na⁺ and Ca⁺⁺ into the cell, initiating cell depolarization and engagement of contractile machinery. The most immediate source of ATP to brain and retinal pericytes would be the adjacent astrocytes, and activity in retinal astrocytes is known to cause ATP release (83). Blockade of P2X1 in vivo blocks sensory evoked dilation consistent with ATP involvement in neurovascular coupling (84). Application of P2Y receptor agonist, UTP, can also cause contraction of pericytes (24). UTP reacts with glucose to produce UTP-glucose, which is an agonist for the P2Y14 receptor (p2ry14) that is highly expressed by brain capillary pericytes. Activation of P2Y14 may lead to vasoconstriction through decrease in cAMP production via G protein signaling. As discussed previously, TRPA1 activation in endothelial cells also causes ATP release, which may act on nearby pericytes and initiate local contraction or conductive responses (66).

Noradrenaline

The superfusion of noradrenaline on brain slices leads to robust contraction of pericytes (24). In vivo, noradrenaline levels in the forebrain are modulated by the locus coeruleus and its vasoconstrictive effect is important for coordination of blood distribution during neurovascular coupling (85). Locus coeruleus projections form varicosities around brain capillaries (in addition to arterioles) but do not appear to directly innervate pericytes (86). Further, adrenoceptor mRNA expression is very low in capillary pericytes based on single cell transcriptomic data (62; 81). This suggests that noradrenaline may act on other cell types such as astrocytes that then secondarily release constrictive signals to pericytes.

Arachidonic acid derivatives

The thromboxane receptor A2 gene (tbxa2r) is strongly expressed by most capillary and ensheathing pericytes, and various arachidonic acid metabolites are endogenous agonists of the receptor. Astrocytes secrete arachidonic acid that can be metabolized by cyclooxygenase 1 to produce prostaglandin H2 (PGH2), which can be further processed into thromboxane A2, a potent thromboxane receptor agonist that can cause pericyte contraction. The stable synthetic analog of the endoperoxide prostaglandin PGH2, U46619, effectively contracts pericytes and is commonly used to create basal capillary tone in brain slice studies (84). In vivo superfusion of U46619 onto the cerebral cortex causes slow contraction of capillary pericytes (28). In retina, application of U46619 leads to marked contraction of ensheathing pericytes but slow contraction of capillary pericytes. This slow contraction was blocked with inhibitors of actin polymerization (cytochalasin D and latrunculin B), again indicating a slow-acting contractile mechanism in capillary pericytes (39). Arachidonic acid can also be processed by the cytochrome P450 ω-hydroxylase pathway to produce 20-HETE, which is also a thromboxane receptor agonist that may modulate capillary pericyte diameter. Thus, broad signaling pathways initiated by arachidonic acid metabolism can potently contract capillary pericytes.

Endothelin-1

Both ensheathing pericytes and capillary pericytes express high levels of endothelin receptor A (ednra) and contract robustly in response to treatment with endothelin-1 (ET-1), a potent vasoconstrictive peptide expressed by endothelial cells and other cell types (22). A recent studied showed that the pre-capillary sphincter and ensheathing pericytes (1^st-3^rd order branches) were most responsive to intraparenchymal delivery of ET-1 (87). Endothelin receptor signaling results in opening of intracellular Ca⁺⁺ stores and Ca⁺⁺ influx through membrane channels, which activates the Ca⁺⁺-dependent contractile machinery in pericytes. Studies also suggest that ET-1 inhibits K_ATP channels, which would prevent their ability to hyperpolarize and promote dilation (82). Prolonged exposure to ET-1 leads to a potential pathological condition that diminishes gap junction connectivity between pericytes (82).

Pathways to vasodilation

Vasoactive signals leading to pericyte dilation have also been identified (Figure 4B). The major excitatory neurotransmitter glutamate causes dilation of capillaries when superfused onto brain slices after capillaries had been partially constricted by noradrenaline. A series of elegant pharmacological studies by Attwell and colleagues concluded that glutamate likely activates arachidonic acid metabolism in astrocytes or neurons to produce prostaglandin E2 that then activates EP4 receptors on pericytes to promote vasodilation (54; 84). Single cell transcriptomic data suggests moderate and heterogeneous expression of EP4 receptor mRNA (ptger4) among capillary pericytes. Intracellularly, EP4 signaling may promote vasodilation by increasing cAMP production and PKA activation, which in turn inhibits myosin light chain kinase activity. Further, inhibiting 14,15-EET signaling, which also acts via EP4 receptors, caused vasoconstriction and blunted functional hyperemia in vivo (88). Alternatively, glutamate-induced nitric oxide production inhibits 20-HETE production to promote dilation by removing a constrictive signal (84).

Adenosine may also cause pericyte relaxation by acting through A2a purine receptors to promoting adenosine triphosphate (ATP)-sensitive K+ channel (K_ATP) channel flux (89). The gene encoding the A2a purine receptor (adora2) is expressed more in capillary and ensheathing pericytes than in SMCs. Side bar 3.

Side bar 3: Comparing apples and oranges: Kinetics of dilation and constriction.

Dilation is the passive process of losing contractile tone, and the speed of dilation is therefore also driven by intravascular pressure experienced at each vascular zone. Pressure is highest at arteriole and ACT zones enabling fast dilation, but low in capillaries leading to slow dilation. Thus, speed of contraction does not infer dilation speed. Optogenetic studies that have selectively contracted different mural cells types in vivo give insight into their individual contractile kinetics (26) The ACT dilates rapidly and robustly in response to neurovascular coupling, but it contracts more weakly than upstream arterioles..

K_ATP channels.

A striking observation from single cell transcriptomic data is that the ion channel repertoire expressed by capillary pericytes is dominated by proteins for K_ATP channels. Both subunits for the channel, i.e., inward rectifier (Kir) subunit, Kir6.1 (kcnj8) and sulfonylurea receptor 2 (abcc9) protein, are highly expressed. The role of pericyte K_ATP channels remains to be clarified, but its contribution to membrane hyperpolarization might lead to local pericyte dilation. Consistent with this, recent work by Zambach et al. showed that the K_ATP channel opener, pinacidil, evoked vasodilation, with strongest responses at the pre-capillary sphincter and ACT (87). In contrast, blockade of K_ATP channels with PNU37883 had no effect on basal vessel diameter but did blunt vasodilation evoked by sensory stimulation. While these drugs exert their greatest effects at the ACT, they may also act at the capillary zone and influence conduction of hyperpolarizing signals. Endogenously, K_ATP channel conductance is dependent on intracellular ATP levels, with lower levels leading to channel opening. This raises the intriguing hypothesis that reduction in pericyte ATP content during brain activity increases K_ATP channel conductance, leading to pericyte relaxation (81).

Pericyte pathology and cerebral blood flow deficits in disease

Stroke

Pericyte loss and dysfunction is a mechanism of blood flow impairment in acute stages of stroke. Pericytes contract in rigor during exposure to ischemia (24) and clamp down onto the capillary lumen, limiting blood cell passage. This can initiate during ischemia and persist even after successful recanalization or de-occlusion of the artery (90). Pericytes die as a result of excitotoxic injury (54) and also injury caused by oxidative/nitrosative stress (90). Sustained contraction is also observed at the level of ACT, where ensheathing pericytes contract just prior to entry to the capillary bed (36). Spreading depolarization alone, which can occur in the acute phase of stroke, was also shown to cause lasting constrictions in ACT zone (91).

In the stroke penumbra, recent studies have shown increased incidence of capillary flow stalls, which contributes to impaired perfusion and oxygen delivery. While these flow stalls have been ascribed to adhesion of neutrophils, aberrant pericyte contractions may heighten capillary plugging (92; 93). Spontaneous deep intracerebral hemorrhage (ICH) is a devastating subtype of stroke involving leakage of blood into deep brain regions. A recent study using a mouse model of ICH revealed specific changes at the ACT, where ensheathing pericytes proliferate and hypermuscularize the vessel (94). This leads to buildup of upstream pressure from arterioles increasing likelihood of vascular rupture.

Dementia

In protracted diseases such as Alzheimer’s disease (AD) and vascular contributions to cognitive impairment and dementia (VCID), pericyte loss has been widely reported (95–97), and is predictive of cognitive decline in APOE4 carriers (98). Patients with AD have 25–50% less pericytes compared to age-matched controls (96). Pericyte pathology is also seen in CADASIL, a monogenic form of vascular dementia primarily affecting arterioles and capillaries (99). A series of elegant studies from Zlokovic and colleagues on APOE4 and MMP9-CypA signaling have linked pericyte pathology in AD to blood-brain barrier disruption (100; 101), which has been confirmed by both post-mortem histology and in vivo MRI (98). The gradual pericyte loss seen in AD likely perturbs brain capillary flow, but the basis of this disease mechanism is less well understood. Recently, Nortley et al. showed in rodent and human tissues that Aβ-induced reactive oxygen species promotes release of ET-1, which binds to endothelin receptor A on pericytes leading to aberrant sustained contraction (102). The death of pericytes and loss of endothelial coverage may lead to heterogeneous capillary dilation, altered capillary flow patterns and mal-distribution of blood flow and tissue oxygenation. Increased capillary flow heterogeneity would also raise the threshold for capillary flux homogenization, further impairing supply of oxygen and nutrients at times of increased metabolic demand. Consistent with this concept, MRI studies have reported altered capillary flow patterns in individuals with Alzheimer’s diseases (103).

Epilepsy

SMCs and pericytes are lost during seizure and replaced over time through mural cell remodeling. This leads to re-organization of mural cell coverage and changes in basal vascular tone, which also alters capillary flow patterns (72). Treatment with PDGF-BB, a signal for pericyte migration and proliferation improves mural cell coverage. In a separate study, seizure activity was shown to associate with local constriction of pericytes and tissue hypoxia in hippocampus (104).

Summary points

1. The gene expression profile, cellular composition and physiological roles of distinct microvascular zones in brain are an active area of research of central importance to cerebral blood flow regulation under basal and active states.

2. The ACT and capillary zones are both covered by pericytes, but have distinct physiological roles in blood flow control. Ensheathing pericytes of the ACT, and upstream pre-capillary sphincter, dynamically gate blood flow into the capillary bed, while the slow dynamics of capillary pericytes contribute to basal capillary heterogeneity and perhaps flux homogenization during functional hyperemia.

3. Cause-and-effect optogenetic manipulations were used in vivo to confirm that capillary pericytes are sufficient to reduce capillary diameter, but are slow in their kinetics. Optical ablation of capillary pericytes confirmed that they are essential for maintenance of basal capillary tone.

4. The slow kinetics of capillary pericytes may be explained by lower contractile protein expression. However, like SMCs and ensheathing pericytes, capillary pericytes express many of the receptors bound by vasoactive mediators, indicating that pathways to engage contractile tone in vivo are present.

5. Aberrant pericyte contraction is a common facet of neurological disease. Loss of pericyte coverage may have negative effects on cerebral blood flow by increasing capillary heterogeneity, resulting in maldistribution of blood flow and oxygen.

Future issues

1. Recent studies have clarified the contractile ability and kinetics of the two major pericyte subtypes. However, the in vivo signals that lead to the vasodynamic changes remain unclear. What are the molecular contributors to basal capillary tone, capillary heterogeneity, and neurovascular coupling?

2. How do pericytes communicate in vivo with other brain cells in the neurovascular unit to regulate processes such as conductive signaling and capillary flux homogenization during neurovascular coupling?

3. Inconsistent semantics continue to be a hindrance in studying mural cells and vascular zones. Here, we clarify the morphology, function, vascular location, and gene expression of these subtypes, and suggest nomenclature to distinguish mural cells and the zones they occupy in cerebral cortex. Importantly, vascular zones can differ in tissues other than cerebral cortex, warranting separate characterization.

4. A difference in contractile kinetics is apparent between mural cell types (26), and this is reflected in differing transcriptomes between SMCs, ensheathing pericytes and capillary pericytes (62). It is unknown which contractile proteins are important for capillary pericyte contraction, given their expression of smooth muscle actin (acta2) is low to undetectable.

5. How can we leverage our knowledge of pericytes to alleviate neurologic disease? For example, restoring pericyte coverage with exogenous PDGF-BB in a mouse model of epilepsy (72) reduced seizure burden, and preventing aberrant pericyte contractions by amyloid beta was achieved with C-type natriuretic peptide (102). These are promising strategies that can be applied to other diseases.