Genetically proxied HTRA1 protease activity and circulating levels independently predict risk of ischemic stroke and coronary artery disease

HTRA1 has emerged as a major risk gene for stroke and cerebral small vessel disease with both rare and common variants contributing to disease risk. However, the precise mechanisms mediating this risk remain largely unknown as does the full spectrum of phenotypes associated with genetic variation in HTRA1 in the general population. Using a family-history informed approach, we first show that rare variants in HTRA1 are linked to ischemic stroke in 425,338 European individuals from the UK Biobank with replication in 143,149 individuals from the Biobank Japan. Integrating data from biochemical experiments on 76 mutations occurring in the UK Biobank, we next show that rare variants causing loss of protease function in vitro associate with ischemic stroke, coronary artery disease, and skeletal traits. In addition, a common causal variant (rs2672592) modulating circulating HTRA1 mRNA and protein levels enhances the risk of ischemic stroke, small vessel stroke, and coronary artery disease while lowering the risk of migraine and age-related macular dystrophy in GWAS and UK Biobank data from > 2,000,000 individuals. There was no evidence of an interaction between genetically proxied HTRA1 activity and levels. Our findings demonstrate a central role of HTRA1 for human disease including stroke and coronary artery disease and identify two independent mechanisms that might qualify as targets for future therapeutic interventions.

Additional Declarations: Yes there is potential Competing Interest.C.D.A. has received sponsored research support from Bayer AG, and has consulted for ApoPharma, unrelated to the content of this manuscript.All other authors declare no competing interests.

INTRODUCTION
Stroke and coronary artery disease (CAD) are the leading causes of death worldwide with an increasing burden particularly in low-income countries 1,2 .Recent genome-wide association studies (GWAS) have identi ed multiple risk loci for stroke and CAD and demonstrated that genetic information can be harnessed to de ne drug targets 3,4,5 .
High-Temperature Requirement A serine peptidase 1 (HTRA1), a known regulator of the extracellular matrix (ECM), has emerged as a risk gene for various phenotypes including stroke 3,4 and cerebral small vessel disease (cSVD) 6,7,8,9 .However, the mechanisms mediating this risk remain poorly understood.
Rare mutations in HTRA1 cause a Mendelian syndrome of small vessel stroke further manifesting with lumbago 6, 7, 8, 10,11,12 .Using data from the UK Biobank (UKB), we recently showed that rare missense variants targeting the HTRA1 protease domain are associated with radiological hallmarks of cSVD 13 .Whether such variants associate with stroke or other cardiovascular traits in the general population remains unclear.Aside from rare mutations, common variants at HTRA1 have been shown to associate with disease risk in GWAS for stroke 3,4 , CAD 5 , migraine 14 , and age-related macular dystrophy 15 , implying that the consequences of such variants reach beyond brain microvessels.However, the full spectrum of phenotypes associated with genetic variation in HTRA1 and the precise mechanisms mediating disease risk remain elusive.
Biochemical studies on HTRA1 mutations involved in Mendelian stroke have demonstrated a loss of protease activity 6, 7, 8, 10, 16, 17 .However, there have been no systematic efforts to study the functional consequences of rare variants in the general population and their relationship with disease.Also, it remains unknown whether the associations between common single nucleotide polymorphisms (SNPs) at HTRA1 and risk for stroke, CAD, and migraine relate to a shared set of causal variants.A third unresolved question relates to the interaction between rare and common variants at HTRA1 with respect to disease risk.
To address these questions, we initially performed rare-variant aggregation tests for ischemic stroke in the UKB 18 using a classical case-control design and a family history proxy phenotype 19 with replication in the Biobank Japan 20 .To assess whether HTRA1 protease activity mediates the observed association with ischemic stroke we determined the consequences of 78 UKB protease domain variants on enzymatic activity in vitro.We next performed phenome-wide association studies (PheWAS) with genetically proxied HTRA1 activity as the exposure and uncovered the phenotypic spectrum of HTRA1 mutation carriers which included stroke-related, CAD-related, and additional traits.Using GWAS data, we performed causal variant analysis on multiple traits and analyzed their association with HTRA1 mRNA and protein levels.As a last step we explored the interactions between genetically proxied HTRA1 activity and levels on cardiovascular risk.

RESULTS
To comprehensively assess the effects of rare and common genetic variation at HTRA1 on human disease we chose a four-pronged approach comprising (1) rare exonic variant analyses, including in vitro protease activity assays, (2) common variant analyses, including GWAS, colocalization, eQTL and pQTL analyzes, (3) assessment of their association with disease, including phenome-wide association studies (PheWAS) and (4) rare and common variant interaction analysis (Fig. 1).

Rare exonic variants in HTRA1 associate with ischemic stroke in the general population
To identify putative associations between rare genetic variation in HTRA1 and ischemic stroke risk in the general population, we used the UKB 450K WES dataset and performed whole-exome rare variant burden analyses using regenie 21 .Given the known involvement of HTRA1 in Mendelian stroke 6,7,8,10,11 and the increase in association power using a liability threshold model, that is conditional on case-control (CC) status and family history (LT-FH) 19 , we employed both a classical CC and LT-FH phenotype.
Ischemic stroke related to rare missense variants in HTRA1 is linked to loss of protease activity Given the molecular function of HTRA1 as a serine protease, we next explored the consequences of rare missense HTRA1 variants on enzymatic activity as a putative mechanism underlying ischemic stroke risk.We focused on variants within the protease domain (amino acids 157-480) as this domain is a prime target for mutations in HTRA1-related familial cSVD 6,7,8,10,11 .We prioritized mutations that affected residues conserved across all four members of the human HTRA family and selected 78 variants in 834 mutation carriers (all heterozygous except one compound heterozygote carrier) for biochemical assessment (Fig. 1, left panel).Speci cally, we measured protease activity in HEK-293E cells overexpressing HTRA1 wild type (wt), an active site mutation (S328A), or the selected UKB mutations (Suppl.Figure 2).
To obtain insights into structure-function relationships, we next mapped the position of the 76 variants with available protease activity data onto the structure of the HTRA1 trimer (PDB code 3TJO).We found the predicted trimer interface (Fig. 3a) to be the preferential site for variants showing residual activity < 0.5 (41% vs 10% of those with residual activity ≥ 0.5, p = 0.003) (Fig. 3b and c and Fig. 2, lower panel), underscoring the importance of the protomer-protomer interface for HTRA1 activity.
We next correlated the effects of the 76 studied variants on enzymatic activity with their effect sizes for the LT-FH stroke phenotype and found a signi cant negative correlation (r=-0.35,p = 0.0041) (Suppl. Figure 5).Expanding on our previous results 13 , we also found the reduction of protease activity to correlate with the effect sizes for log white matter hyperintensity (WMH) volume (N = 36,681 individuals; N variants = 31; N carriers = 109; r=-0.69,p = 0.0093) (Suppl.Figure 6).We further proxied the HTRA1 activity for all UKB individual based on their genotype and the in vitro biochemical measurements.Generalized additive model analysis con rmed that the LT-FH stroke phenotype and logWMH volume increased with decreased HTRA1 activity (LT-FH: p = 3.28E − 5 after correction for sex, age at baseline and 10 PCs; logWMH volume p = 4.79E − 9 after correction for sex, age at baseline and 10 PCs) (Suppl.Figures 7 and 8).Collectively, these data imply that the increase in ischemic stroke risk conferred by rare missense variants in HTRA1 relates to a loss of enzymatic activity.
PheWAS reveals association between genetically proxied loss of HTRA1 protease function and vascular as well as skeletal traits in the UKB Using the experimentally determined consequences of the 76 variants on HTRA1 activity as an exposure, we next performed a phenome-wide association analysis (PheWAS) to explore associations with all disease outcomes in the UKB.We leveraged DeepPheWAS 22 to assign case/control status of 1,463 standardized Phecodes for binary disease phenotypes to each individual.For each Phecode, we performed logistic regression using sex, age and 10 PCs as covariates.We deemed associations at an FDR level of 5% to be statistically signi cant.In addition to neurovascular disease-related Phecodes (P292.1:aphasia/speech disturbance -a frequent manifestation of stroke 23 , TIA 24 , and migraine with aura 25 ; P433.3: cerebral ischemia; P433: cerebrovascular disease; P340.1: migraine with aura), we found lumbago-and spondylosis-related Phecodes (P724: other unspeci ed back disorders; P724.9: other and unspeci ed disorders of back) to be associated with reduced HTRA1 activity (Fig. 4 and Table 1).Of note, next to stroke and WMHs, lumbago and spondylosis are among the most common manifestations in Mendelian disorders related to HTRA1 mutations 17,26 , suggesting that our approach accurately identi ed key phenotypic traits.As a major novel nding, we further identi ed a signi cant association between reduced protease activity and a range of CAD-related Phecodes including P411 (ischemic heart disease), P411.2 (myocardial infarction), P411.8 (other chronic ischemic heart disease, unspeci ed) and P411.4 (coronary atherosclerosis).CAD traits have so far not been considered part of the phenotypic spectrum of familial disorders linked to HTRA1 mutations.The directionality of effects was identical to stroke risk in that rare variants reducing HTRA1 activity were associated with an increased CAD burden.The effect size was strongest for P292.1 (aphasia/speech disturbance: OR = 0.84, CI95 [0.79-0.89]per 10% increase in HTRA1 activity) and overall effect sizes seemed stronger for cerebrovascular compared to CAD.The only phenotype for which a lower HTRA1 activity was associated with a lower risk of was P155 (Cancer of liver and intrahepatic bile duct; OR = 1.67,CI95 [1.37-2.03]per 10% increase in HTRA1 activity).
Collectively, these ndings show that the consequences of rare loss-of-function variants in HTRA1 extend beyond known phenotypes and include neurovascular, cardiovascular, and skeletal traits.
A common causal variant affects HTRA1 mRNA and protein levels and is associated with ischemic stroke and coronary artery disease Previous GWAS found common variants in HTRA1 to be associated with any ischemic stroke, small vessel stroke, CAD, and migraine 3,4,5,14,27 .To investigate the possibility that a single variant or limited set of variants is responsible for these associations with multiple traits, we performed colocalization tests using the latest GWAS summary statistics for ischemic stroke, small vessel stroke, lacunar stroke, CAD, and migraine, within a window of ± 150kb from HTRA1 using Hyprcoloc 28 .A common variant (rs2672592, G/T, allele frequency of the minor T allele = 36%) in the rst intron of HTRA1 was the single causal variant for all tested phenotypes (posterior probability of colocalization = 0.91).The major G allele was associated with increased risk for ischemic stroke, small vessel stroke, lacunar stroke and CAD while being associated with a lower risk for migraine (Fig. 5).Of note, rs2672592 is a known eQTL and pQTL for HTRA1 in peripheral blood mononuclear cells and plasma, with the G allele being associated with lower mRNA and protein levels 29,30 .The association between rs2672592 and the above disease and expression phenotypes was further con rmed using Primo 31 (posterior probability = 0.99, FDR = 0.03), making this SNP the prime candidate to be the common causal variant at this locus and to in uence disease risk via modulation of both HTRA1 mRNA and protein levels (Suppl Fig. 9).
We further performed a phenome-wide association study (PheWAS) using imputed allele dosage of rs2672592 as an exposure and sex, age and 10 PCs as covariates.We con rmed associations with cardiovascular disease (CVD), including both stroke and CAD-related Phecodes in the UKB (Table 2).Speci cally, we found coronary atherosclerosis (P411.4) and occlusion and stenosis of precerebral arteries (P433.1) to be among the top signals.Similar to the effect of rare loss-of-protease-function variants, stronger effects were found for cerebrovascular disorder (P433.Genetically proxied HTRA1 activity and levels independently predict risk of stroke and coronary artery disease To scrutinize the relationship between genetically proxied HTRA1 activity and levels in mediating disease risk, we next performed logistic regression analyses.In an interaction model with sex, age and 10 PCs as covariates there was no indication for an interaction between HTRA1 activity and HTRA1 mRNA/protein levels with regard to the association with stroke or CAD (interaction p-value > 0.5 for both phenotypes tested).These results show that the effects of HTRA1 activity as determined by rare damaging missense variants within the protease domain and HTRA1 mRNA and protein levels as proxied by rs2672592 on cardiovascular risk are independent.
Finally, to explore the combined effect of HTRA1 activity and levels on population health we combined ischemic stroke and CAD into a CVD phenotype and strati ed patients into disease risk categories as predicted by HTRA1 activity and levels.We found individuals with reduced HTRA1 activity predicted by a rare protease domain variant (N = 349) in combination with a GG genotype at rs2672592 to have an OR of 1.77 [1.34-2.34] for developing CVD compared to the reference group (TT genotype and no damaging rare variant; N = 56,424) (Fig. 6).

DISCUSSION
Our ndings highlight the importance of HTRA1 for human disease and identify two independent mechanisms by which rare and common variants at HTRA1 affect cardiovascular risk.We show that rare variants in HTRA1 are linked to ischemic stroke in the general population and that variants causing loss of protease activity associate with ischemic stroke, CAD, and skeletal traits.We further show that a common causal variant (rs2672592) modulating circulating HTRA1 mRNA and protein levels enhances the risk of ischemic stroke, small vessel stroke, and CAD while lowering the risk of migraine and agerelated macular dystrophy.Importantly, there was no evidence of an interaction between the effects of genetically proxied HTRA1 activity and levels on CVD risk suggesting that future strategies addressing either mechanism could be used to ne-tune the phenotypic consequences of therapeutic interventions.
Cardiovascular risk factors have repeatedly been reported in HTRA1-related familial stroke 17 .We found the association of rare damaging missense and loss-of-function variants in HTRA1 with ischemic stroke to be independent of traditional risk factors.Recent work has shown that loss of HTRA1 function results in profound changes of the vascular matrisome 16,32,33 .HTRA1 cleaves multiple constituents of the ECM including latent TGFβ binding protein (LTBP) 1 and 4, bronectin (FN), vitronectin (VN) and clusterin (CLU) 16,32,33,34,35,36 .Consequently, disease-related loss of HTRA1 function leads to accumulation of these and other matrisomal proteins in the brain vasculature 7,32,33 .Alterations of the ECM take center stage in cSVD 37,38 as further evidenced by the causal role of mutations in COL4A1 39,40 , COL4A2 41,42 , and LAMB1 43 .Extending these observations, the association between rare variants in HTRA1 and risk of CAD identi ed here suggests a mechanistic overlap between small vessel and large artery disease that involves multiple proteins of the ECM 26,44 .
So far, the loss of HTRA1 activity and its relationship with disease have been regarded as a binary trait.
Our results demonstrate the importance of considering HTRA1 activity as a continuous phenotype.For one, we show that naturally occurring mutations in HTRA1 result in variable degrees of enzymatic activity covering the full range from complete loss of function to increased activity, as exempli ed by L364F.We further demonstrate a gradual increase in the risk of ischemic stroke, CAD, and skeletal traits with larger reductions in genetically proxied protease activity.Whether residual protease activity also in uences the penetrance of mutations and clinical severity remains to be determined.Mapping HTRA1 variants to the HTRA1 structure identi ed the trimer interface as a hotspot for the most deleterious variants with respect to enzymatic function and cardiovascular risk.This could open perspectives for variant severity prediction via computational approaches.
Reduced HTRA1 protein levels should result in reduced protease activity.Accordingly, the phenotypes associated with genetically proxied HTRA1 protease activity and levels exhibit overlap with respect to cardiovascular risk.Of note, however, we found no association between low HTRA1 activity and risk for retinal disorders, and between low HTRA1 levels and skeletal traits.Also, genetically proxied lower activity was linked to higher risk for migraine with aura, whereas genetically proxied lower levels protected against migraine.Several explanations come in mind: First, in addition to their effects on protease activity, some missense variants might interfere with mRNA or protein levels e.g., by in uencing their stability or protein folding.Second, HTRA1 exhibits functions related to protein quality control that are unrelated to protease activity 45 : HTRA1 disintegrates Tau and AB brils in a PDZ-dependent but proteaseindependent manner which is then followed by proteolytic degradation.Whether a similar dual mechanism also applies to other proteins is currently unknown.Third, the observed divergence on phenotype associations could be explained by tissue-or cell type-speci c effects.For instance, in the brain, HTRA1 is predominantly produced by astrocytes, followed by arterial endothelial cells, whereas Müller and retinal pigment epithelial cells are the predominantly expressing cell types in the retina 46,47 .
Also, the effects of rs2672592 and the repertoire of HTRA1 substrates 33,35,36,48,49 might vary between cell types.
Our ndings open perspectives for risk strati cation through mutational screening in combination with measurements of protein function and levels.Our ndings further have implications for future therapeutic strategies.In principle, cardiovascular risk could be lowered either by restoring HTRA1 activity or by increasing its levels.However, and as shown by our PheWAS, raising HTRA1 levels is predicted to result in an increased risk for AMD and other retinal disorders.Conversely, raising HTRA1 activity is predicted to increase the risk of liver and bile duct cancer.Together, this highlights the need to develop organ or celltype speci c therapeutics.Intravitreal administration of an inhibitory HTRA1 antibody is currently being tested in a clinical trial for geographic atrophy secondary to AMD 35,50 and there are various strategies to speci cally target individual organs including brain and heart 51 or cell types such as macrophages 52 .
Our study has limitations.Although we assessed the functional consequences of 76 missense protease domain variants on enzymatic activity, there remain numerous other variants that warrant scrutiny, particularly additional missense variants in the predicted trimer interface, or targeting the signal, IGFBP7 or PDZ domain, as well as predicted LoF variants.Also, we used HEK-293E cells and LTBP1 as the sole cell line and substrate in our biochemical assays, while HTRA1 protease activity could vary depending on the cell-type or substrate.Moreover, the statistical power of PheWAS remains low for certain Phecodes, and we thus might have missed associated phenotypes, emphasizing the need for larger sample sizes.
In conclusion, we demonstrate associations between rare and common genetic variation in HTRA1 and cardiovascular outcomes in the general population.We identify two independent mechanisms by which these variants affect cardiovascular risk.Our ndings imply, that future clinical applications would need to consider both HTRA1 activity and levels.They further demonstrate a mechanistic overlap between cerebral small vessel and large artery disease and a pivotal role of the ECM.

UK Biobank
Our primary dataset consists of ~450,000 individuals from Biobank (UKB) with available WES data and linked health records.We only included unrelated (KING coe cient < 0.177 corresponding up to 2 nd degree) individuals of European origin.To assign European ancestry, we used ashpca2 53 and constructed 10 principal components (PCs) for all UKB participants and participants in 1000 Genomes v3 data using ~20,000 SNPs as input.With the resulting PCs, we trained a random forest classi er based on the gold standard 1000G v3 ethnicities.We combined samples from CEU, TSI, GBR, and IBS into a European clade.Each UKB individual was classi ed using the random forest classi er and those having a probability of >95% of being European were put in the European ancestry.Descriptive statistics of the studied individuals are provided in Suppl.Table 2.

LT-FH phenotype
The UKB holds information on paternal, maternal or sibling stroke (coded in UKB elds 20107, 20110 and 20111, respectively) and the number of siblings (UKB elds 1883 and 1873).We used this information to construct the LT-FH phenotype as previously described 19 .Speci cally, we used information on proband stroke, paternal stroke, maternal stroke, sibling stroke and number of siblings (with a maximum of 5 considered) to construct the continuous LT-FH phenotype.We used a heritability of 0.17 as proposed in Variant selection: Focusing on missense variants targeting the HTRA1 protease domain (amino acids 157-370), UKB variants were selected for biochemical assessment based on residue conservation in human HTRAs according to Clustal Omega 69 .Speci cally, all variants strictly conserved in human HTRA1-4 (n= 48), 29 out of 30 variants conserved in 3 human HTRAs (1 variant not included in the HTRA1 trimer structure and thus not selected), and 2 out of 42 variants conserved in ≤ 2 human HTRAs including R227W, the most frequent exonic mutation in the UKB, were selected for analysis (Figure 1).
Protease activity measurements: Eukaryotic expression plasmids encoding either full-length human HTRA1 (aa 1-480) fused to a C-terminal Myc-(His) 6 tag or the N-terminal region of human LTBP1s (aa 1-689) fused to a C-terminal V5-(His) 6 tag have been previously described 13,16 .Mutagenesis was conducted using the QuickChange Lightning Site-Directed Mutagenesis kit (Agilent Technologies).
Secretomes containing the protein of interest were collected and centrifuged for 10 min at 400 g to remove debris.Where indicated, cells were lysed in 50 mM Tris, 150 mM NaCl, 1% (v/v) triton X-100, 1% (w/v) sodium deoxycholate, 0.1% (w/v) sodium dodecyl sulphate, pH 7.4 for 20 min at 4°C and samples centrifuged 20 min at 16,000 g.Secretomes from non-transfected cells and from cells overexpressing wild type (wt) HTRA1 or the inactive variant S328A were included in each run as controls.
LTBP1 and HTRA1 signals were quanti ed using ImageJ.LTBP1 processing was evaluated as the ratio cleaved / intact LTBP1 and HTRA1 protease activity was estimated as the ratio LTBP1 processing / HTRA1 levels.The activity of wt HTRA1 was set to 1. Activity thresholds for follow-up analyses were arbitrarily set as follows: (i) activity estimates <0.25, (ii) >0.25 to <0.5, (iii) >0.5 to <1.0 and (iv) >1.

Correlation of effects
We constructed Pearson correlation coe cients between (i) effect sizes from single-variant analyses of logWMH (white matter hyperintensity) volume in the UKB vs. effect sizes constructed from the average HTRA1 activity ratio and (ii) effect sizes from single-variant analyses of LT-FH the UKB vs. effect sizes constructed from the average HTRA1 activity.All analyses were performed using R 4.3.0.

Predicting individual HTRA1 activity
To assess the association of HTRA1 activity with cardiovascular endpoints, we genetically proxied HTRA1 activity for each individual in the UKB.For this, we set HTRA1 activity for individuals without a rare HTRA1 protease domain mutation to 100%.For mutation carriers we multiplied the measured average HTRA1 activity from the in-vitro assays with 100%.This activity estimate was used as an exposure in our analysis.We used Generalized Additive Models (GAM) to show association between this exposure and the LT-FH phenotype and logWMH volume, respectively.

PheWAS
To explore the association of predicted HTRA1 activity with the full range of phenotypes encoded in the UKB, we used DeepPheWAS to assign participants to standardized Phecodes. 22We used all ICD10 codes (main position, secondary position, death records) from the UKB.We excluded Phecodes with <100 cases and Phecodes that are male-or female-speci c.Individuals were assigned a case status if >1 ICD10 code mapped to the respective Phecode.Individuals meeting the pre-speci ed exclusion criteria were removed from the analysis, otherwise the individual was assigned a control status.We used logistic regression with age, sex and 10 PCs as covariate to test rare variant carrier status (0/1) against the phenotype of interest.Results with p<0.05 were corrected with Firth's correction.We considered results with FDR<5% as statistically signi cant.

Colocalization of common variants
We obtained summary statistics for any ischemic stroke and small vessel stroke from the GIGASTROKE consortium 4 , for coronary artery disease (CAD) from the CARDIoGRAMplusC4D Consortium 5 , for lacunar stroke from the UK DNA Lacunar Stroke studies 1 and 2 and from collaborators within the International Stroke Genetics Consortium 27 , and for migraine through the International Headache Genetics Consortium (IHGC) 14 .We used an interval of +-150kb from HTRA1 and included all SNPs that were available for analysis for all phenotypes.We used Hyprcoloc 28 and synchronized effect sizes and standard errors to perform colocalization and report the resulting posterior probability of colocalization.
Blood eQTL and pQTL data were obtained from the eQTLgen consortium 29 and deCODE 30 , respectively.
We used Primo 31 to perform pleiotropy analyses between eQTL, pQTL and disease phenotype data and report posterior probability of pleiotropy.We used an interval of +-150kb from rs2672592 and included all SNPs that were available for analysis for all phenotypes.

Strati cation of individuals into disease risk classes
Using imputed genotype dosage of rs2672592 and rare HTRA1 protease domain carrier status, we used logistic regression with age, sex and 10 PCs as covariates.We set the lowest risk category (rs2672592 TT and no rare variant) as the reference category and all odds ratios are reported compared to this category.PheWAS of imputed HTRA1 protease activity reveals associations with neurovascular, skeletal and CADrelated traits.PheWAS results using predicted HTRA1 protease activity for each individual as the dependent variable.The x-axis shows the Odds Ratio for a 10% increase in HTRA1 activity, the y-axis displays a log-transformed p-value.The FDR cutoff of 0.05 is presented as a dashed line.

Display items
A common variant in HTRA1 shows colocalization with ve neurovascular phenotypes.For each of the phenotypes (any stroke; coronary artery disease; small vessel stroke; lacunar stroke; migraine) the HTRA1 gene ±150kb is depicted on the x-axis.The y-axis depicts the -log10 p-value of the respective GWAS.The dashed lines depict genome-wide signi cance (p=5E -8 ).We used Hyprcoloc to compute the posterior probability of colocalization (PIP) for each SNP in the genomic interval.The highest PIP was estimated for rs2672592.For this SNP, we depict the p-value for each of the ve phenotypes.

Figures
Figures were prepared with R version 4.3.0,ggplot2, Pymol, Adobe Illustrator, Microsoft Power Point and BioRender.com.

Figures Figure 1
Figures

Figure 3 The
Figure 3

Table 1
PheWAS results for genetically proxied HTRA1 protease activity in the UK Biobank.Odds Ratios are presented for a 10% increase in HTRA1 protease activity.CI95: 95% con dence interval; FDR: false discovery rate.

Table 2
Of interest, associations with Phecodes relating to age-related macular degeneration (AMD) and other retinal disorders showed opposite directionality compared to CVD-related phenotypes in that risk was higher in carriers of the minor T allele.Collectively, these ndings show a common causal variant affecting HTRA1 mRNA and protein levels is associated with ischemic stroke and CAD.PheWAS results for allelic dosage of the rs2672592 major G allele in the UK Biobank.CI95: 95% con dence interval; FDR: false discovery rate.
Terao C, et al.Population-speci c reference panel improves imputation quality and enhances locus discovery and ne-mapping.).Research Square (2023).