Genome-wide association study of platelet aggregation in African Americans

Lower coronary artery disease survival rates have been observed in African Americans
even after clinical, demographic and socioeconomic variables are considered 16], 17], suggesting there are undiscovered factors accounting for this racial difference.
We and others have shown there is a strong genetic component to coronary artery disease
and platelet reactivity 11], 18]–20], but there is a paucity of data about responsible genetic mechanisms. Thus, the African
American participants in the GeneSTAR and PGAP studies represent unique and valuable
resources for discovering novel genetic variants associated with platelet aggregation,
a central process in acute coronary syndromes. We used the large GeneSTAR African
American cohort as the GWAS discovery sample, and separate PGAP and GeneSTAR European
Americans as replication cohorts. The major findings were: 1) identification of 3
replicated variants associated with platelet aggregation in the discovery sample,
2) rs12041331 in PEAR1 (previously reported genetic variant in European Americans and African Americans),
was associated with collagen, epinephrine and ADP aggregation in both the discovery
and replication cohorts, 3) rs11202221 in BMPR1A was associated with ADP aggregation in both the discovery and the African American
replication cohort, and 4) rs6566765 was associated with ADP aggregation in the African
American discovery and European American replication cohort. Since African Americans
are under-represented in most clinical studies of coronary artery disease (CAD), it
will be important to consider these established and novel genetic variants in this
racial group.

Despite modest sample sizes of the discovery and replication cohorts, we were able
to identify and validate three loci associated with platelet aggregation. The finding
of significant loci in modestly sized cohorts is not typical of most GWAS studies.
We had large effect sizes for these loci, probably because platelet aggregation is
more physiologically defined than most clinical phenotypes and represents a biological
process, not a disease outcome. The large effect sizes are likely due to the relatively
large percentage of heritability explained by the discovered loci in African Americans
(from 10 % with ADP 2uM to 17 % with epinephrine). With a larger sample size, we probably
would have discovered additional genetic variants with smaller effect sizes.

Assessment of ex vivo platelet function is labor intensive and very few cohorts have
been generated with this phenotype; fewer still have substantial numbers of African
Americans. In general, compared to the PGAP cohort, GeneSTAR participants had a higher
incidence of CAD risk factors. Nevertheless, the minor allele frequencies (MAF) for
each of the African American cohorts in this report were remarkably similar (Table 2). The GeneSTAR and PGAP studies also utilized the same platelet agonists and same
genotyping platform, features of the study design that support the validity of our
analyses. The replicated variants were not associated with fibrinogen levels in any
cohort (Additional file 1: Table S1).

The rs12041331 SNP in PEAR1 has previously been associated with epinephrine-and ADP-mediated platelet aggregation
in European Americans and African Americans 13], 15]. PEAR1 encodes a 1037 amino acid platelet cell surface receptor and, upon activation, an
intracellular tyrosine residue is phosphorylated, followed by degranulation, amplification
of the glycoprotein IIb/IIIa pathway and sustained platelet aggregation, most likely
through the PI3K/Akt pathway 21], 22]. We now extend the association to include collagen-mediated platelet aggregation,
findings confirmed in an independent group of African Americans. The variant identified
by rs12041331 has been shown to regulate expression of PEAR1 protein in a dose-dependent
fashion 15]. Taken together, these data suggest that PEAR1 levels are important effectors of
platelet aggregation in both African and European Americans.

Of the five novel variants associated with platelet aggregation in the discovery sample,
2 were confirmed in at least one of our replication cohorts. Three novel variants
did not replicate, suggesting they might be false positives. The most intriguing replicated
variant was rs11202221 in BMPR1A; the G-allele of this SNP was associated with decreased ADP-induced platelet aggregation.
This is of interest because BMPR1A has been implicated in vascular calcification as
well as in the development of atherosclerosis 23], and platelets play a role in pathogenesis of the latter. BMPR1A encompasses 168 kb in the 10q23.2 region and encodes a 532 amino acid long single-pass
cell surface receptor. This receptor belongs to the BMP receptor family of the transforming
growth factor-beta (TGF-?) receptor superfamily and is expressed widely in various
tissues. On ligand binding, BMPR1A activates intracellular signaling pathways, commonly
leading to altered gene expression. Although BMPR1A has not been identified in platelets,
several reports indicate it is expressed in megakaryocytes 24], 25], the bone marrow progenitor cell that produces platelets. Thus, the variants in BMPR1A that are associated with platelet aggregation could alter BMPR1A expression and/or
function in megakaryocytes, which in turn could alter gene expression in signaling
molecules mediating ADP-induced platelet aggregation. Alternatively, these BMPR1A SNPs could be in LD with other causative SNPs in either protein-coding or non-protein
coding genes. Among the protein-coding genes near BMPR1A, transcripts of MMRN2, GLUD1, WAPL, and PAPSS2 are present in platelets, however, only GLUD1 (glutamate dehydrogenase 1) and PAPSS2 (3?-phosphoadenosine 5?-phosphosulfate synthase 2) protein-products are present in
platelets 26]–28]. There are no microRNAs or lincRNAs within 500 kb of rs11202221.

It is intriguing that rs11202221 in BMPR1A did not replicate in European Americans. It is unlikely that this SNP is a false
positive because 1) it replicated in PGAP African Americans, 2) the effect of the
minor allele on platelet aggregation was the same direction in the two African American
populations, and 3) there were 5 other SNPs in BMPR1A that showed association (p??10^?5) with ADP-induced platelet aggregation in both African American populations. LD patterns
differ dramatically between European (CEU) and African (YRI) populations represented
in the 1000 Genomes Project (see Additional file 1 Fig. S3). There is a 53 Kb block of LD including rs11202221 in the CEU reference
population where variants are high in frequency (G allele frequency ~20 % at rs11202221,
and MAF of most SNPs in the block are ~20 %). In contrast, no LD block in the YRI
reference population includes rs11202221 and the allele frequency of the G allele
is considerably lower (4 %). Given the tagging approach employed herein with the GWAS
array, we speculate that rs11202221 tags the true ‘causal’ variant which itself may
be low in frequency. Under this hypothesis, the ‘causal’ variant would consequently
be tagged with higher correlation in the African American population in contrast to
the European American population and therefore yield significant results in the African
Americans but not European Americans. Future work will involve a targeted resequencing
in the full dataset to fully examine all sequence-identified variants in this region
and specifically test this hypothesis.

The second novel association with ADP-mediated aggregation is of an intergenic variant
at 18q22.3 locus with ADP-mediated platelet aggregation. The 500 kb region on either
side of this variant contains a few genes, but the transcript of only one, FBXO15
(which encodes F-box protein 15), has been reported at low levels in platelets. F-box
proteins are important in substrate recognition by certain ubiquitin protein ligase
complexes and thus are important in regulating protein degradation 29], 30]. While it is possible that this variant may affect protein degradation in platelets,
the role of FBXO15 in platelet aggregation remains unexplored.

Different agonists and agonist concentrations were utilized to generate more refined
platelet signaling pathway phenotypes. Collagen activates platelets via glycoprotein
VI, which signals via an immunoreceptor tyrosine activation motif 31]. ADP induces platelet activation through the P2Y₁₂ and P2Y₁ G protein coupled receptors (GPCRs), while epinephrine activates platelets via a
different GPCR, the ?2A-adrenergic receptor 32]. Each of these GPCRs activates different G protein families, which in turn activate
different sets of signaling molecules. Eventually, these different proximal signaling
pathways converge to a final common pathway resulting in integrin ?IIIb?3 activation
and platelet aggregation. Because SNPs in PEAR1 were associated with platelet aggregation
induced by all three of these physiologic agonists, our data suggest a potential role
for PEAR1 in a shared signaling pathway downstream of receptor-proximal signaling.
Furthermore, ADP at low concentrations (e.g., 2 ?M) induces rapid and reversible platelet
aggregation through G_q-coupled P2Y₁, whereas high ADP concentrations (e.g., 10 ?M) induce G_i-coupled P2Y₁₂ inhibition of adenyl cyclase and complete aggregation 33]. Thus, our findings that different loci were associated with different concentrations
of ADP-induced platelet aggregation support hypotheses whereby ALDH1L1-AS2, SUFU and BMPR1A regulate primarily platelet G_q signaling and SDK2 regulates primarily G_i signaling in platelets.