Genetics of Vascular Cognitive Impairment

Vascular dementia (VAD) is the second most common cause of dementia after Alzheimer’s and accounts at least 20% of dementia cases. Lesser degrees of cognitive decline due to cerebrovascular disease are even more common, and the umbrella term Vascular Cognitive Impairment (VCI) has been introduced to cover all degrees of cognitive impairment due to cerebrovascular disease, from mild impairment through to dementia.(1)

In the same way that stroke is a heterogeneous disease caused by multiple different pathological processes, VCI also results from many different vascular pathologies. In principle any disease process causing cerebral ischaemia or haemorrhage can cause VCI. Neuroimaging and pathological studies have allowed division of VCI mechanisms into a number of broad groups including multiple infarcts, single strategic infarcts, intracerebral haemorrhage, hypoperfusion, and cerebral small vessel disease.(1) It is now recognised that by far the most common pathology underlying VCI is cerebral small vessel disease (SVD)(1), which is most readily seen on MRI as white matter hyperintensities and lacunar infarcts although other features which may contribute to cognitive decline include atrophy and cerebral microbleeds (CMB). In addition many of the pathologies causing SVD also predispose to intracerebral haemorrhage which itself can contribute to cognitive decline. The most popular hypothesis is that these diverse subcortical pathologies lead to damage to white matter pathways, which disrupts complex distributed cortical–subcortical circuits underlying cognitive processes such as executive function and processing speed. (2) Cerebral amyloid angiopathy (CAA) also affects the cerebral small vessels, and causes intracerebral haemorrhage and has been associated with VCI and dementia. (1)

How Could Genetic Predisposition Increase VCI Risk?

Although perhaps over-simplistic, a useful distinction can be made between genetic risk factors which increase the risk of the underlying pathologies causing VCI (such as small vessel disease), and risk factors which alter the way in which the brain responds to these vascular insults. Latter factors could include not only neuronal responses to brain injury and mechanisms associated with plasticity, but also factors altering cognitive brain reserve. For example it has been shown that a greater number of years in full time education is associated with a reduced dementia risk and this has been thought of as increased cognitive reserve, while genetic variants associated with education attainment have ben associated with a reduced risk of dementia.(3) Currently almost all data on VCI genetic risk has looked at the underlying stroke and cerebrovascular disease mechanisms. The fact that all monogenic forms of VAD result from gene mutations which alter arterial function, predominantly small cerebral arteries, suggests that this may be the dominant genetic mechanism leading to VCI.

Are Genetic Risk factors Important for VCI?

There is little data from twin and other epidemiological studies estimating the heritability of VAD with only a small twin study, which was too underpowered to be conclusive(4). However increasing evidence from both epidemiological, and genetic studies suggests that the pathological processes underlying VAD have significant heritability. According to genome wide association study (GWAS) heritability estimates of 20-40% for large artery, cardioembolic and small vessel stroke.(5,6) Furthermore the radiological features of SVD, the major pathology underlying VCI, have been shown to have high heritability. Twin and family studies estimate the heritability of WMH lesion volume to be between 50-80%,(7,8)

Perhaps the strongest evidence that genetic factors can cause VAD comes from the range of monogenic conditions which result in small artery stroke and VAD. (table 1) Such monogenic diseases tend to cause VCI and VAD in middle age. In this article we will first review monogenic conditions predisposing to VAD, and then cover recent insights into pathogenesis of apparently sporadic VCI.

Table 1. Monogenic forms of SVD which can also cause VCI.

AD= Autosomal dominant; AR= Autosomal recessive

Disease	Gene(s)	Inheritance	Gene function(s)	Mutations	Disease mechanism	Major clinical features
CADASIL (Cerebral Autosomal Dominant Arteriopathy with Sub-cortical Infarcts and Leukoencephalopathy)	NOTCH3	AD	Notch3 transmembrane receptor has roles in vascular smooth muscle cell remodelling and angiogenesis,	Cysteine-changing mutations in epidermal growth factor-like repeat region (EGFr) in exons 2 – 24	Accumulation of NOTCH3 ectodomain cleaved from mutant protein in extracellular spaces of small vessels.	Migraine with aura Subcortical lacunar infarcts VCI Depression Encephalopathy WMH and CMB on MRI
CARASIL (Cerebral autosomal recessive arteriopathy with subcortical infarcts and leukoencephalopathy)	HTRA1	AR	High temperature requirement serine protease A1 (HtrA1) switches off transforming growth factor β pathway	Missense, nonsense, and splice site mutations	Decreased protease activity Impaired activation of wild-type HtrA1 trimer subunits Inhibition of HtrA1 trimer formation and stabilisation	Subcortical lacunar infarcts VCI Non-neurological features – alopecia, spondylosis
Autosomal dominant HTRA1-related CSVD	HTRA1	AD	As for CARASIL	As for CARASIL	As for CARASIL	Subcortical lacunar infarcts and WMH VCI Encephalopathy Non-neurological features less common than for AR CARASIL
COL4-related SVD	COL4A1 COL4A2	AD	COL4A1/A2 encode α1 and α2 collagen chains, which are the most abundant components of the extracellular matrix	Missense mutations - most affect glycine residue in highly conserved Gly-X-Y repeat regions	Disrupted α1 or α2 chains prevent formation of collagen helix, resulting in basement membrane abnormalities	Porencephaly Infantile hemiparesis Intracerebral haemorrhage Lacunar stroke VCI CMB and WMH Ocular involvement incl cateracts Nephropathy
FOXC1/PITX2-related SVD	FOXC1 PITX2	AD	Forkhead box transcription factor C1 (Foxc1) has roles in blood vessel development	Deletions or duplications of 6p25 Mutations in Foxc1	FOXC1 interacts with PITX2 FOXC1 involved in pericyte and endothelial cell proliferation Impaired blood brain barrier function	Axenfeld-Rieger anomaly Cerebellar malformations Hydrocephalus Periventricular heterotopia WMH
Retinal Vasculopathy with Cerebral Leukodystrophy and Systemic Manifestations (RVCL-S)	TREX1	AD	TREX1 encodes DNase III (Three prime repair exonuclease), which has roles in DNA repair	Frameshift mutations in C-terminus	Impaired cellular localization of DNase III in endoplasmic reticulum	Retinal vasculopathy lacunar infarcts, WMH, pseudotumours Migraine VCI Seizures Multi-organ involvement: Raynaud’s, cirrhosis, renal dysfunction, osteonecrosis

Monogenic Forms of SVD

The most common familial form of SVD is Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy (CADASIL), caused by mutations in the NOTCH3 gene, a single pass transmembrane protein belonging to an evolutionarily conserved NOTCH receptor family. (9) Typical features include migraine, usually with aura, occurring in the twenties or thirties, lacunar strokes occurring in middle age, and VCI which may progress to VAD. (10) Other common features include an encephalopathy which can be the presenting feature in 10% of individuals, apathy, and depression.(10) MRI findings (figure 1) share many features with sporadic SVD including lacunar infarcts, confluent WMH, enlarged perivascular spaces, CMB, and atrophy. However the spatial pattern of WMH, particularly the involvement of the anterior temporal pole, is a useful diagnostic pointer.(11) Although originally thought to occur predominantly in the individuals without stroke risk factors, it is now well recognised that vascular risk factors, particularly smoking and hypertension, interact with the NOTCH3 mutation to cause earlier onset disease.(10,12) For example CADASIL patients who smoke on average have stroke ten years earlier (10). Therefore careful control of cardiovascular risk factors is important.

Fig 1. Typical Brain imaging appearances in CADASIL.

(A) Brain FLAIR MRI showing confluent white matter hyperintensities and lacunar infarcts. (B) Brain FLAIR MRI showing WMH involve the anterior temporal pole; this is a characteristic location for CADASIL. (C) Gradient echo MRI showing a microbleeds in the left thalamus.

An autosomal recessive form of familial SVD was first described in consanguineous Japanese and Chinese families and the underlying gene was identified as HtrA Serine Peptidase 1(HTRA1). (13) Such individuals also suffer multiple lacunar strokes, which could lead to VCI, although other systemic features occur including alopecia and low back pain. A heterozygous form of the disease due to a single HTRA1 mutation has recently been recognised (14) and this appears to be the second most common mutation underlying familial SVD after NOTCH3 in the UK. (Hugh Markus, unpublished data). It is associated with a similar phenotype to autosomal recessive CARASIL, with lacunar stoke, confluent WMH and sometimes encephalopathy, and VCI. However the age of presentation appears to be later and systemic features such as alopecia and back pain are less frequent.(14)

Mutations in Collagen Type IV Alpha 1 Chain (COL4A1) or COL4A2 affecting basal membrane integrity can cause familial SVD.(15) Both intracerebral haemorrhage, as well as lacunar stroke and WMH, are seen. The disease can also present with porencephaly in childhood as well as eye and retinal involvement. CMB are particularly frequent on MRI, reflecting the increased prevalence of intracerebral haemorrhage.(16) A number of other rarer single gene disorders have also been associated with SVD including mutations in TREX1(Three prime repair exonuclease 1), and FOXC1(Forkhead Box C1) (17) (see table 1).

Mutations causing cerebral amyloid angiopathy (CAA) can also be associated with VCI. Monogenic forms are caused by amyloid precursor protein (APP) mutations. Mutations at codon 693 cause amino acid substitutions at position 22 of amyloid-beta.(18) There is hereditary cerebral hemorrhage with amyloidosis—Dutch type (HCHWA-D) in which a Glu to Gln exchange leads to CAA with recurrent lobar hemorrhages and white matter damage.(19) At the same position a Glu to Lys exchange was been described in Italian families with multiple strokes (19) and a Glu to Gly exchange is characteristic for the Arctic type early-onset Alzheimer disease (AD) which was reported in Swedish families.(20) The Flemish mutation affects codon 692.(21) Codon 694 is typically affected in CAA of the Iowa type.(22) It is of note that familial AD cases caused by mutations in presenilin 1 or presenilin 2 genes also have severe CAA particularly if they occur behind codon 200 (23).

“Sporadic” Vascular Cognitive Impairment

Relatively few studies have investigated genetic risk factors for “sporadic” VCI itself as an outcome: many more have looked for risk factors underlying the contributing pathological processes such as all stroke, small artery stroke, and radiological markers of SVD.

Over the last two decades GWAS has transformed our understanding of the genetics of complex diseases including stroke. A major advantage of GWAS is that, by the use of hundreds of thousands of markers spanning the whole genome, entirely novel associations can be detected and no prior hypothesis is required. However the multiple statistical comparisons made mean that very large sample sizes, thousands or tens of thousands, are required, and such datasets are currently scarce in VCI. To date, only two very small GWASs have been reported with VAD as the outcome measure. A retrospective study in Korea (84 patients; 200 controls) detected no associations but was very underpowered;(24) GWAS studies of complex diseases tend to require sample sizes in the thousands to detect significant associations. A prospective study in the Netherlands (67 patients; 5700 controls) identified an association with the variant rs12007229 near the androgen receptor gene on the X chromosome: this was replicated in the German population of 221 cases and 213 controls.(25) However replication in a much larger sample size is required.

An alternative but less powerful approach is the candidate gene approach. Here a variant in a gene suspected of being associated with the disease in question is identified, and a study performed to determine whether the variant is more common in disease cases compared with controls. However candidate gene studies in many complex diseases including stroke have produced results which have failed to be replicated in GWAS studies involving many thousands of individuals, suggesting the underlying methodology was often not robust. Particular issues have included small sample sizes which we now in the GWAS era realise were too small, poor phenotyping, and publication bias with selective reporting of positive studies, and multiple testing of different polymorphisms without correction for multiple testing.

A number of meta-analyses have reviewed candidate gene studies in VCI, but caution in their interpretation is required as they cannot fully overcome poor methodology, including publication bias, in the contributing studies. Two meta-analyses assessed SNP associations with VCI and VAD. In one increased risk for VCI was found for apolipoprotein E(APOE ε4) and methylenetetrahydrofolate reductase(MTHFR) polymorphisms.(26) The second study (27) included 69 studies with 4462 cases and 11583 controls and identified APOE ε2/ε3/ε4, MTHFR C677T, Paraoxonase 1(PON1) L55M, transforming growth factor- β1 (TGF-β1) +29C/T, and tumor necrosis factor- α (TNF-α) -850C/T with VAD. A third meta-analysis focused on apolipoprotein E and identified 29 studies including 1763 VAD cases and 4534 controls.(28) Overall, apolipoprotein E and MTHFR polymorphisms are the most consistently found SNPs to be associated with VCI. Analysis of APOE ε2 revealed a significant protective effect against VAD in only one of the meta-analyses.(28) APOE has also been reported to be related to MRI correlates of SVD. Including WMH(29).

A potential issue in interpreting results from studies with VAD as the clinical endpoint relates to diagnosis. Until recently international definitions for VAD required impairment of episodic memory, which can often be a later feature of VAD due to SVD, but is more commonly seen early in the course of Alzheimer’s disease.(1) This, combined with the well-recognised observation that many cases of dementia in the elderly have mixed pathology including both vascular and AD pathology, mean that determining whether associations are specifically with VAD or could be due to genetic risk factors contributing to associated AD pathology can be difficult. More progress is likely to be made with studies in which accurate phenotyping is performed with neuroimaging to look at the type and extent of vascular changes, as well as the updated dementia classifications(1) which allow VAD cases to be included if cognitive deficits occur in a variety of different domains including executive function.

Genetic Influences for Sporadic Stroke

In contrast to the little GWAS data on VCI there is now increasing data on stroke, which is likely to be also relevant for VCI. A series of international collaborative GWAS studies in ischaemic stroke (WTTCC2(30), METASTROKE(31), MEGASTROKE(32)) have identified 32 loci associated with ischaemic stroke. An initial striking was that specific genes predisposed to individual stroke subtypes(31), emphasising that different ischaemic stroke subtypes have different genetic architecture. The recent MEGASTROKE analysis in 67162 cases and 454450 controls also identified a number of genes predisposing to conventional cardiovascular risk factors known to increase stroke risk.(32) Interestingly genetic variants predisposing to thrombosis have been associated with an increased risk of thromboembolic stroke secondary to large artery disease and cardioembolism, but not of small artery stroke, suggesting that thrombotic factors may be less important for SVD.(32,33)

Any of the genes identified as predisposing towards ischaemic stroke are likely to increase the risk of VCI, as all stroke subtypes can lead to cognitive impairment. However the most import in terms of contribution to VCI are likely to be genes predisposing to SVD because this is the major pathology underlying VCI. Unlike large artery and cardioembolic stroke for which many GWAS loci have been identified, there has been only one GWAS association with small vessel stroke. In 4203 small vessel strokes and 50728 controls an association was found with a locus on chromosome 16q24.2.(34) The same locus was associated with WMH.(34)

The small number of GWAS loci identified in SVD is likely to represent methodological issues. Many cohorts have used loose definitions for lacunar stroke including a lacunar syndrome with a normal CT scan, and it has been shown that up to 50% of such cases may not have SVD.(35) Accurate phenotyping with MRI increases the yield in genetic studies.(5) Current international collaborations are collecting large cohorts of MRI phenotyped SVD cases for GWAS.

Increasing evidence has highlighted the importance of endothelial dysfunction and blood brain barrier disruption in SVD.(36;37) Epidemiological studies have shown an association between homocysteine levels and ischaemic stroke, and its has been reported this may be particularly with SVD.(38) However determining whether this association is causal or confounded by other factors is challenging.(39) Genetic techniques can be useful in investigating causality using the technique called Mendelian randomisation. Here a genetic variant associated with increased level of the blood marker in question is identified and if it is associated with disease this provides evidence that the association is indeed causal, because the genetic variant has exposed the individual to increased levels of the marker in question throughout their life. Therefore the association of the MTHFR polymorphism, which increases homocysteine levels, suggests that homocysteine may be a causal risk factor in SVD.(40) It has been suggested this acts via affects on small vessel endothelial function.(38)

A number of GWAS studies have been performed in patients with MRI markers of SVD, particularly WMH.(41,42), and many loci associated with WMH, as well as a loci associated with white matter ultrastructural damage identified on diffusion tensor imaging.(43)

Genetics of sporadic CAA

Most cases of CAA are sporadic. It is present in approximately 10-40% of elderly brains and 80% of patients with AD.(44) For CAA-related intracerebral hemorrhage the estimated heritability is 70%.(45). Yet, ApoE ε4 and ε2 alleles are the only genetic risk factors that have been robustly associated with risk of developing sporadic CAA.(46) ApoE ε4 increases the risk for CAA and ApoE ε2 was shown to be a risk factor for ICH in Ab-CAA patients.(47). Although the exact reasons for the association of CAA with ApoE ε4 are not fully determined, it is possible that ApoE ε4 is associated with fibrillogenesis of Ab within brain tissue and in perivascular drainage pathways.(48) Also, ApoE ε4 binds to LRP-1 and interacts with soluble and aggregated Ab both in vitro and in vivo, influencing its conformation and clearance.(49) The AD–related locus, complement component receptor 1 (CR1), was reported to influence CAA-ICH incidence and recurrence while hypertension loci play no significant role in CAA-related intracerebral hemorrhage.(45)

The association of CAA with cognition has been debated. Several studies suggest a link between CAA and dementia (50), but there have also been clinico-pathological studies like the Honolulu Asia Aging Study (51) in which CAA was associated with cognition only in the presence of AD, questioning the degree to which CAA makes a separate contribution to dementia. In the Rush Memory and Aging Project and the Religious Orders Study (52) it was shown, however, that CAA affects cognitive outcomes over and above the main causes of dementia and that CAA was associated with a faster rate of decline in cognition. Hemorrhagic and ischemic brain abnormalities are likely to contribute to CAA-related cognitive decline. These include recurrent intracerebral hemorrhages, CMB, superficial cortical siderosis, convexal subarachnoid hemorrhage, brain infarcts, white matter lesions and microstructural tissue damage. The figure gives examples of CAA-related brain abnormalities on MRI.(figure 2)

Fig 2. CAA-related brain imaging findings.

(A) Axial computed tomography shows two intracerebral lobar haemorrhages in the same patients. Recurrent lobar bleeding occurred within 1 months. (B) Axial gradient echo (GRE) image shows multiple cerebral cortical microhaemorrhages. (C) Axial GRE sequence demonstrates cortical superficial siderosis characterized by curvilinear signal hypointensity following the gyral cortical surface. (D) Axial fluid attenuation inversion recovery (left) and GRE (right) images show non-traumatic convexity subarachnoid haemorrhage.

What are the Implications of Genetics for Vascular Dementia Care?

In the small group of patients with monogenic conditions underlying VCI, diagnosis can be important both in guiding management and prognosis. In the past screening for these disorders has been time consuming, and not always widely available with each individual disorders requiring Sanger sequencing of that individual gene, but with the advent of next generation sequencing (NGS) arrays it is now possible to screen for all possible SVD genes in one assay. Diagnosis also allows the possibility of presymptomatic and prenatal testing within families.

In the much more common multifactorial VAD, in which genetic influences are likely to be polygenic (ie. resulting from multiple genes each contributing a small increase in risk likely interacting with environmental risk factors), there is hope that genetic risk profiling may allow prediction of a group at high risk in whom particular preventative or therapeutic strategies could be performed. This is beginning to look promising in other complex diseases in which a significant genetic component has been identified, for example coronary artery disease.(53) However the limited information we currently have on the genetics of VAD makes such risk prediction impossible at the present time.

Perhaps more importantly future better understanding of the genetics of VAD may allow us to gain new insights into the underlying therapeutic processes and therefore develop better targeted treatment approaches.

Insights into the pathogenesis of SVD and vascular dementia from genetic studies

Recent studies in monogenic SVD are providing exciting insights into underlying disease processes underlying both SVD and VCI.(17) They have provided evidence for involvement of key extracellular matrix (ECM) or ‘matrisome’ proteins in pathogenesis.(17,54) The ECM comprises the non-cellular component of tissues made up of water, proteins and polysaccharides which provide scaffolding for cellular components by producing fibrous proteins such as collagen, elastin, and laminin. It is also biochemically active providing signals which contribute to tissue function and homeostasis. The matrisome is defined as the ensemble of nearly 300 proteins which make up the ECM (core matrisome), or are associated with the ECM (matrisome-associated proteins), and these have been characterised by proteomic and bioinformatic methods.(17)

Evidence is emerging suggesting monogenic types of SVD disrupt matrisome function. This has been best illustrated in CADASIL by an elegant series of experiments. Post-mortem studies in CADASIL report aggregation of key matrisome proteins.(55) Studies in a transgenic CADASIL mouse led to the hypothesis that NOTCH3 mutations result in aggregation of aberrant portions of the extra-cellular NOTCH3 protein,(56) which itself induces a cascade of aggregation of other matrisome proteins including tissue inhibitor of metalloproteinase3 (TIMP3) which promotes sequestration of other matrisome proteins including vitronectin.(57) Such a cascade can result in a potassium channelopathy and impaired cerebrovascular reactivity which itself could lead to cerebral ischaemia. (57)

One of the matrisome proteins identified in NOTCH3 extracellular protein aggregates in CADASIL has also been identified as a key molecule in CARASIL. Latent TGFB-binding protein1 (LTB-1) which was found to co-aggregate with NOTCH3 extracellular in CADASIL, was identified to be a target of the HTRA1 protease in a study in mouse and patient tissues.(58) Other mutations such as COL4A1/2 lead to direct disruption of ECM components, in this case the collagen scaffold.(54)

Taken together this evidence suggests that different mutations underlying monogenic SVD may causes SVD shared molecular pathways leading to matrisome dysfunction. The finding that common genetic variants in both COL4A1/2 (59) and HTRA1 (32) also increase risk of sporadic SVD and WMH suggests that similar molecular processes also play a role in sporadic SVD, and therefore may be of much larger importance to SVD and VCD. Excitingly this has led to potential targets, and intervening in the process has been shown to delay the pathological cascade in a transgenic CADASIL model.(57)

Conclusions and the future

The predominant importance of SVD in causing VCI is now well recognised, and an increasing number of monogenic forms of SVD and VCI are being identified. New NGS techniques are transforming our ability to make this diagnosis. Insights provided by experimental models of monogenic SVD, as well as from studies in man and models of sporadic SVD, are transforming our understating of SVD and VCI. They are emphasising the importance of disruption of the matrisome, as well as the role of endothelial dysfunction and blood-brain barrier dysfunction in the pathogenesis of SVD. These insights are presenting new treatment possibilities. Whether additional genetic factors determine whether a person with stroke or SVD, will develop VCI remains largely unknown, and understanding this will require large scale GWAS and other studies, similar to those which have made such progress in the understanding the genetic basis of stroke and other complex diseases.