Abstract
The African American (AA) population displays a 1.6 to 3-fold higher incidence of thrombosis and stroke mortality compared to European Americans (EA). Current anti-platelet therapies target the ADP-mediated signaling pathway, which displays significant pharmacogenetic variation for platelet reactivity. The focus of this study was to define underlying population differences in platelet function in an effort to identify novel molecular targets for future anti-platelet therapy. We performed deep coverage RNA-Seq to compare gene expression levels in platelets derived from a cohort of healthy volunteers defined by ancestry determination. We identified >13,000 expressed platelet genes of which 480 were significantly differentially expressed genes (DEGs) between AAs and EAs. DEGs encoding proteins known or predicted to modulate platelet aggregation, morphology or platelet count were up-regulated in AA platelets. Numerous G-protein coupled receptors (GPCRs), ion channels, and pro-inflammatory cytokines not previously associated with platelet function were likewise differentially expressed. Many of the signaling proteins represent potential pharmacologic targets of intervention. Notably, we confirmed the differential expression of cytokines IL32 and PROK2 in an independent cohort by qRT-PCR, and provide functional validation of the opposing actions of these two cytokines on collagen-induced AA platelet aggregation. Using GTEx whole blood data, we identified 516 eQTLs with Fst values >0.25, suggesting that population-differentiated alleles may contribute to differences in gene expression. This study identifies gene expression differences at the population level that may affect platelet function and serve as potential biomarkers to identify CVD risk. Additionally, our analysis uncovers candidate novel druggable targets for future anti-platelet therapies.
INTRODUCTION
Cardiovascular disease (CVD) is a complex disorder with multiple genetic and non-genetic factors contributing to individual risk. It is well-documented that AA are at a higher risk of developing CVD than individuals of EA ancestry (1–3). Risk of CVD is multi-factorial and AAs disproportionately carry more risk factors than EAs (4). It is likely that the complex interplay between both modifiable (i.e., socioeconomic) and unique non-modifiable (i.e., genetic) factors contribute to AA increased risk of CVD.
Platelets play a central role in the formation of arterial thromboses which can result in myocardial infarction and stroke. Platelet dysregulation is a component of inflammatory disorders, complications from infection (i.e., COVID-19) and thrombotic disease (5, 6). As a result, differential platelet regulation potentially contributes to the disparate CVD risk observed between EAs and AAs. Consistent with this, several studies have identified key differences in platelet function between EAs and AAs (7, 8). Platelets isolated from AAs exhibit a higher degree of PAR4 receptor-induced aggregation and intracellular calcium mobilization (9–11). Genome-wide association studies have identified alleles (i.e. BMPR1A) associated with platelet aggregation in AAs but not EAs (12). Moreover, current anti-platelet therapies are almost singularly focused on antagonizing the central ADP-P2Y12 pathway, and there is significant pharmacogenetic variation in therapeutic response to some P2Y12 antagonists (i.e., clopidogrel) in the AA population (13).
Given the disparities in CVD that exist between individuals of African and European ancestry, as well as the need for novel anti-platelet target discovery, we sought to identify key global differences in platelet gene expression between these two populations. Our study identified several key signaling pathways impacted by DEGs in AA platelets that could directly influence platelet function. Specifically, we have functionally characterized two cytokine genes for their novel roles in regulating platelet aggregation, representing new avenues of exploration for future targeted anti-platelet therapies.
MATERIAL & METHODS
Volunteer Recruitment.
The study was approved by both The George Washington University (GWU) and Northwestern University (NU) Internal Review Boards (IRB# 021747). Whole blood (~50 mL) was procured from healthy volunteers over 18 years-old, self-identifying as either AA or EA. All volunteers were interviewed on their medical history. Health information such as BMI, blood pressure, serum lipids and insulin resistance was not collected as part of the recruitment process. The Discovery Cohort consisted of 44 volunteers (AA = 21; EA = 23) for RNA-Seq analysis of platelets. The Validation Cohort consisted of platelets from 22 additional volunteers (AA = 12; EA = 10), which were used for RT-PCR validation of select genes/transcripts identified in the Discovery Cohort. See Supplemental Material and Methods for additional demographic information, medical history questionnaire and exclusion criteria.
Platelet purification and RNA isolation.
Platelets were isolated from whole blood samples through magnetic leukocyte depletion with CD45 microbeads (Miltenyi Biotec) as previously described with minor modification (14). RNA from leukocyte-depleted platelets was isolated using TRIzol extraction (Invitrogen) with resulting RNA Integrity Number (RIN) scores > 7.
RNA-Seq, DEG determination, and gene ontology/pathway analysis.
Paired-end libraries were constructed using NEBNext Ultra Directional RNA kit, and sequenced according to the Illumina protocol for the HiSeq2000 platform, and sequenced to a targeted depth of >100 million paired-end reads (150 nucleotides length). Raw sequence data were processed using Illumina’s RTA and CASAVA pipeline software. Sequencing datasets were quality assessed using fastqc (v. 0.11.9). The sequencing reads were trimmed to remove adaptor-related sequences and reads with ≥3 consecutive bases of aggregate quality score <20 using Trimmomatic (v. 0.38, (15)). The high-quality reads were quantified at transcript level against the latest version of the human reference transcriptome (Ensembl release 101, Homo_sapiens.GRCh38.cdna.all.fa), using Salmon (v.1.2.0, (16)). The transcript-level quantifications were aggregated at the gene level and subjected to DEG analysis using DeSeq2 (v.1.28.1, (17)). Unless otherwise noted, a false discovery rate (FDR) of 10% (q<0.1) was considered significant.
Gene ontology and pathway analyses of the platelet transcriptome was performed using Ingenuity Pathway Analysis (IPA) computational tool (Qiagen). A Fisher’s exact test with an Yekutieli and Benjamini (18) adjusted (adj) P value <0.05 was used to define significant over-representation of DEGs in ontologies and pathways.
Genotyping and ancestry estimation.
DNA was extracted from each blood sample using Gentra Puregene Blood kit (Qiagen). All DNA samples were bar coded. SNP were genotyped using Illumina Multi-Ethnic Genotyping array (MEGA) at the University of Chicago Functional Genomics Core using standard protocols. Genotyping outputs were created by Genome Studio using 0.15 GenCall score cutoff.
Standard QC procedures were conducted such as sex check by PLINK (19) to identify individuals with discordant sex information, identity-by-descent with a cutoff score of 0.25 to identify duplicated or related individuals. SNP QC consisted of removal of SNPs on the sex and mitochondrial chromosome, A/T or C/G SNPs, SNPs with missing rate > 5% or failed Hardy-Weinberg equilibrium (HWE) tests (p<0.00001).
We used the 1000 Genomes Project samples from AFR, East Asian (EAS), and European (CEU) ancestry as well as AA subjects enrolled in the ACCOuNT study (20) as the reference populations. Using Genome-wide Complex Trait Analysis (GCTA, version 1.24.4) (20) we conducted principal component analysis (PCA) for the Discovery Cohort. The first two PCs were mapped using R (Figure 1) to visualize the relationship between these reference populations and the Discovery Cohort.
Figure 1. Ancestry determination.
The Discovery Cohort of healthy volunteers (N = 42, Red) are shown along with the 1000 Genomes CEU (European – Blue), East Asian (Yellow), and African (YRI – Pink) populations and an AA cohort (grey). The AA healthy volunteer (N = 21) lie in the cluster between HapMap CEU and YRI as expected. The EA healthy volunteer (N = 23) cluster with the CEU reference population as expected.
eQTL analysis.
Genotype data from GTEx v7 comprising 635 individuals of AAs and European ancestry (EAs) was obtained and used for cis-eQTL mapping and Fst estimation as previously described in Zhong et al. (21).
The GTEx genotype data was then merged with 1000 Genomes phase 3, and principal-component analysis (PCA) and QC was performed using PLINK. SNPs were phased using SHAPEIT2 (22) and imputed with IMPUTE2 (23) using all reference populations from 1000 Genome phase 3. Fst estimations of differences in allele frequency between CEU and YRI were calculated using GCTA (Table S1). We used a stringent Fst threshold cutoff at 0.25 to differentiate AAs from EAs.
For cis-eQTL mapping, GTEx v7 whole-blood gene expression data (n = 308 EA and 48 AA) was downloaded. Individuals of east Asian ancestry were excluded. This dataset corresponded to 19,432 expressed autosomal genes. Gene-expression values were log2-transformed, identified using the HG Focus annotation, and mapped to gene positions from GENCODE release 19. Probabilistic estimation of expression residuals (PEER) variables were calculated using previously described method (24). A hierarchical correction method involving the Benjamini and Yekutieli (BY) method to adjust p-values for all association tests for each gene was performed (18), followed by pooling the minimum BY-adjusted p-value of every tested gene and correcting the pooled minimum p-values using Benjamini-Hochberg correction (25). Sex, sequencing platform, the first three genomic PCs, and 35 PEER variables were used as covariates, consistent with the latest GTEx analysis protocols. After regressing out covariates, Matrix eQTL was then used to map cis-eQTLs from each gene expression trait (26). We then extracted the whole blood eQTLs corresponding to the DEGs and hemostasis pathways to validate and elucidate the regulatory networks identified in platelet aggregation in AAs.
Platelet aggregation assay.
Platelet rich plasma (PRP) was collected from whole blood via centrifugation at 170 × g for 15 minutes. Platelets were stimulated using collagen (2 μg/ml), ADP (10 μM), epinephrine (10 μM), and thrombin (0.5 units/ml) for 10 minutes. Aggregation was performed using light transmission aggregometry on a Model 490 4+ Four Channel Optical Aggregometer (Chrono-log). Maximum aggregation, rate of aggregation, and time until start of aggregation were all recorded following stimulation.
To test the effect of neutralizing antibodies, PRP was incubated with an anti-IL32 (Abcam, 5 ug/mL), anti-PROK2 (Invitrogen, 5 ug/mL), or a rabbit IgG isotype control antibody (Abcam, 5 ug/mL) for five minutes. Aggregation was then stimulated with collagen (2 μg/ml).
Washed platelets were isolated from whole blood samples as previously described (27), and aggregation assays were performed by incubating with 400 ug/mL fibrinogen and either IL32 (500 nM, Sigma) or PROK2 (10 nM, Sigma) prior to stimulation with collagen (2 μg/ml).
Quantitative RT-PCR (qRT-PCR) validation.
Platelet total RNA from a separate cohort of healthy volunteers was reverse transcribed with random hexamer primers per manufacturer’s protocol (ThermoFisher Scientific). Quantitative PCR was performed on the resulting cDNA using the QuantStudio Real-Time PCR System (ThermoFisher Scientific), gene-specific primers and SYBR Green PCR Master Mix (Applied Biosystems).
See Supplemental Material and Methods for additional methodological details.
RESULTS
Overview Platelet Transcriptome Landscape.
Twenty-one AA and 23 EA healthy volunteers were recruited for platelet RNA-Seq analysis. There was no significant difference between AA (11 males, 9 females) and EA (9 males, 14 females) volunteers with respect to sex (p=0.4, Fisher’s Exact Test). From the PC visualization of genotyped DNA from healthy volunteers, we observed that the AA subjects clustered with the AA reference population. Likewise, the EA subjects clustered with the European (CEU) refence population. (Figure 1).
We initially combined both AA and EA RNA-Seq datasets in order to evaluate the overall platelet transcriptome landscape. Accordingly, the total number of expressed genes was 13,415 based on filtering for a minimal expression value of 0.5 TPM per gene and ≥ 50% expression across 44 samples (Table S2). The majority of encoded proteins were annotated by IPA (based on a nucleated cell nomenclature scheme) as residing in the cytoplasm (38%), followed by nucleus (26%), plasma membrane (12%) and extracellular space (5%) (Figure S1A).
Differential Gene Expression in AA and EA Platelets.
A comparison of the AA versus EA platelet transcriptome data revealed a total of 480 differentially expressed genes at an FDR q-value ≤ 0.10 (Figure S2, Table S3). Of these 480 differentially expressed genes representing ~4% of the platelet transcriptome, 324 genes were up-regulated and 156 genes were down-regulated in AA platelets relative to EA platelets. The median log2(AA/EA) ratio values for up- and down-regulated genes ranged from +1.11 (2.16-fold up-regulated) to −1.08 (2.12-fold down-regulated), respectively. The majority of proteins encoded by the differentially expressed genes were annotated by IPA as residing in the cytoplasm (29%), followed by nucleus (16%), plasma membrane (19%) and extracellular space (9%) (Figure S1B). This represents a conspicuous and significant redistribution of encoded proteins away from the cytoplasmic/nuclear categories and towards the plasma membrane/extracellular space categories when comparing the overall platelet transcriptome to the differentially expressed transcriptome (Figure 2A; p=0.022, Fisher’s Exact Test). In terms of broad functional entities of the proteins encoded by the DEGs, the majority were enzymes (11%), followed by transcription regulators (6%), transporters (5%), kinases (5%), transmembrane receptors not including GPCRs (4%), peptidases (3%), phosphatases (2%), GPCRs (2%), ion channels (1%) and cytokines (1%) (Figure S1B). Noteworthy was the near doubling of transmembrane receptors, GPCRs plus ion channels from 4% to 7% when comparing the overall platelet transcriptome to the differentially expressed transcriptome, respectively (Figure 2B).
Figure 2. Redistribution pattern of encoded proteins in the overall platelet transcriptome compared to differentially expressed platelet transcriptome.
Redistribution of (A) cellular location and (B) cellular function categories of encoded proteins in the overall platelet transcriptome (AA and EA populations combined) versus the differentially expressed platelet transcriptome (AA compared to EA population). Cellular location and function categories based on IPA analysis.
DEGs of known function were significantly enriched in a number of canonical signaling pathways as defined by IPA (Figure 3A), such as primary immunodeficiency signaling, B cell receptor signaling, B cell development, lupus B cell signaling, NFAT regulation of immune response, communication innate and adaptive immune cells, and phospholipase C signaling. Particularly noteworthy, the activity of several signaling pathways, including B cell receptor signaling, lupus B cell signaling, and phospholipase C signaling were predicted by IPA to be more highly activated in AA platelets relative to their EA counterpart (Figure 3A). A complete listing of the significant canonical signaling pathways and the differentially expressed genes that makeup these pathways is available in Table S4.
Figure 3. Depicted are the (A) canonical pathways and (B) representative biological processes significantly enriched with differentially expressed genes in a comparison of AA versus EA platelets.
The bar graph indicates canonical pathway with corrected P value. Closed circles indicate the number of genes undergoing differential expression in the canonical pathways or biological processes, respectively. Stippled bar indicates IPA-predicted activation of the associated canonical pathway or biological process in AA platelets relative to EA platelets. Fold-enrichment values from top (primary immunodeficiency signaling) to bottom (phospholipase C signaling) in panel A was 4.49, 1.97, 3.48, 1.43, 1.47, 1.55, 2.05, 1.20, respectively; fold-enrichment values from top (migration of cells) to bottom (degranulation of cells) in panel B was 1.67, 4.14, 1.69, 1.57, 4.44, 1.42, 3.48, 1.94, 1.21, 1.24, respectively. See Table S1 for list of genes and eQTLs in each category.
In terms of IPA-defined disease and biological processes (Figure 3B), a significant over-representation of DEGs was observed in migration of cells, mobilization of Ca2+, chronic inflammation disorder, ion homeostasis of cells, cell viability/survival, flux of Ca2+, inflammation of organ, hemostasis and bleeding, and degranulation of cells. Among these particular biological processes, IPA analysis predicted greater activation of cell viability/survival in AA platelets relative to EA platelets. A complete listing of the significant disease and biological processes, and the DEGs that make up these processes is available in Table S4.
We were particularly struck by the over-representation of DEGs in the canonical signaling and biological process categories of phospholipase C (PLC) signaling (with predicted pathway activation), mobilization of Ca2+ and flux of Ca2+, portending increased platelet activity and/or propensity of platelet aggregation in the AA population. Our RNA-Seq analysis identified numerous GPCRs, ligand-gated ion channels, intracellular signaling and cytokine genes (e.g., interleukin 32 (IL32) and prokineticin2 (PROK2)) that were significantly over-expressed in AA relative to EA platelets (Figure 4; Table S3). The encoded receptors, ion channels and intracellular signaling proteins are predicted to facilitate inside-out signaling, shape change, granule secretion, and aggregation in platelets, whereas the encoded cytokine proteins are associated with pro-inflammatory and/or pro-thrombotic states (Figure 4).
Figure 4. Schematic representation of significant differentially expressed genes in platelet aggregation pathway of AAs.
Red is representative of up-regulated genes in AA compared to EA. Blue is representative of down-regulated genes.
After comparison of the GTEx whole blood eQTLs, we found 197 unique eGenes (genes with at least one nearby SNP associated to gene expression) from the 480 DEGs identified. This resulted in 1938 eQTLs. Of these eQTLs, 513 had Fst measures over 0.25 representing SNPs with large allele frequency differences between global populations (Table S1).
Aggregation of AA and EA Platelets.
Ex-vivo platelet aggregation testing was performed on PRP from healthy AA (n=12) and EA (n=22) volunteers using light transmission aggregometry. Aggregation was tested in response to collagen (2 μg/ml), ADP (10 μM), epinephrine (10 μM), and thrombin (0.5 unit/ml), which activates platelets via the GP6 (GPVI), P2Y12, ADRA2A and the protease-activated receptors (PARs), respectively (Figure 4) (28). AA platelets exhibited significantly greater aggregation compared to EA platelets in response collagen stimulation (P=0.030), whereas differences in aggregatory ability was not observed with either ADP or thrombin (Figure 5A). Treatment with epinephrine trended towards higher total aggregation by AA platelets but missed significance (P=0.12). In addition to total aggregation, the rate of aggregation was significantly greater (P=0.042) in AA platelets compared to EA platelets following treatment with collagen (Figure 5B).
Figure 5. Comparison of agonist-stimulated platelet aggregation in the AA and EA populations.
Platelet-rich plasma from AA or EA individuals was stimulated with collagen at 2 μg/ml, ADP at 10 μM, epinephrine at 10 μM, or thrombin at 0.5 unit/ml. (A) Total platelet aggregation and (B) aggregation rate were greater in AA versus EA in response to collagen (*p<0.05).
Differential Effects of IL32 on Aggregation of AA Versus EA Platelets.
Given the importance of cytokines in platelet activity and function (29), the large number of cytokine genes identified by RNA-Seq as over-expressed in AA versus EA platelets and the availability of cytokine-neutralizing antibodies for functional analysis, we investigated the effects of IL32 and PROK2 on collagen-stimulated platelet aggregation in PRP. IL32 and PROK2 were two of the most differentially expressed cytokine genes, as well as being present multiple times in the over-represented categories of ‘Migration of cells,’ ‘Inflammation of organ’ and ‘Mobilization of Ca2+’ (Figure 2B, Table S3). Prior to initiation of these functional experiments, we performed a strict Bonferroni multiple testing correction (P<0.0002) for all cytokine genes in the human genome (30). Of the 11 cytokine genes defined as significant by FDR (Table S3), 10 remained significant following Bonferroni correction including IL32 and PROK2. Next, IL32 and PROK2 over-expression in AA platelets was validated in an independent cohort of healthy volunteers via qRT-PCR (Figure S3). Subsequently, one of the healthy AA volunteers from the RNA-Seq analysis was brought back for an additional 3 separate blood draws in order to perform 3 independent platelet aggregation assays. As demonstrated in Figure 6A, 2 μg/ml collagen-stimulated platelet aggregation and rate of aggregation were significantly augmented by the addition of a neutralizing antibody to IL32 (5 μg/ml final concentration) compared to collagen stimulation in the presence of IgG control (5 μg/ml final concentration). There was no difference in lag time for platelet aggregation following collagen stimulation in the presence or absence of neutralizing antibody (Figure 6A). Interestingly, 2 μg/ml collagen-stimulated platelet aggregation and rate of aggregation were significantly impaired by the addition of a neutralizing antibody to PROK2 (5 μg/ml final concentration) compared to collagen stimulation in the presence of 5 μg/ml IgG control (Figure 6B). These findings, reported for the first time, suggests that IL32 (up-regulated in AA platelets) may have anti-platelet aggregation activity, while PROK2 (up-regulated in AA platelets) exhibits pro-platelet aggregation activity. To confirm this hypothesis, we performed a reciprocal aggregation experiment where washed EA platelets were stimulated with 2 μg/ml collagen in the presence and absence of IL32 or PROK2. A healthy EA volunteer from the RNA-Seq analysis was brought back for an additional 3 separate blood draws in order to perform 3 independent washed platelet aggregation assays. As shown in Figure 6C, 2 μg/ml collagen-stimulated platelet aggregation and rate of aggregation were significantly impaired by the addition of 50 nM IL32. By contrast, 2 μg/ml collagen-stimulated platelet aggregation was not affected by the addition of PROK2 at a concentration of 10 nM (Figure 6D), representing ~10 to 50-fold the EC50 of this cytokine in in vitro assays (31, 32). Moreover, the addition of PROK2 alone to washed platelets did not lead to appreciable aggregation (Figure 6D).
Figure 6. Effect of neutralizing cytokine antibodies on collagen-stimulated AA platelet aggregation and cytokines on collagen-stimulated EA platelet aggregation.
Total aggregation (left panel), aggregation rate (middle panel), and lag time of platelet activation (right panel) following collagen-stimulation of AA platelets incubated with IgG control, (A) anti-IL32 or (B) anti-PROK2 antibody. Total aggregation (left panel), aggregation rate (middle panel), and lag time of platelet activation (right panel) following collagen-stimulation of EA platelets incubated without or with (C) IL32 or (D) PROK2 cytokine. Results are presented as mean ± SEM (n = 3, *p<0.05). Calculated P values for % aggregation and % aggregation/minute in Panel A were 0.009 and 0.021, Panel B were 0.016 and 0.019, and Panel C were 0.015 and 0.045, respectively.
DISCUSSION
RNA-Seq predicts enhanced PLC-gamma signaling and collagen-induced hyper-aggregation in AA platelets.
Previous studies have demonstrated that genetic variation significantly impacts platelet function (8, 12). To date, platelet reactivity to ADP and epinephrine has been demonstrated to exhibit moderate to strong heritability in both AA and EA individuals. However, only collagen-induced platelet aggregation has been reported to be heritable in AA platelets compared to their European ancestry counterparts (8). Our work builds on these findings by demonstrating that both magnitude and rate of platelet aggregation were significantly greater in AA versus EA platelets in response to collagen but not ADP, epinephrine and thrombin. This indicates that differential sensitivity to collagen stimulation might be one factor that distinguishes platelet function between AA and EA individuals.
Our finding that only collagen stimulation leads to higher levels of platelet aggregation in the AA versus EA population was in fact computationally anticipated. IPA analysis of the RNA-Seq results predicted activation of the PLC signaling based on the direction of differentially expressed genes in this pathway. Moreover, over-expression of B lymphoid tyrosine kinase (BLK), protein tyrosine phosphatase non-receptor type 6 (PTPN6) and CD36 in AA platelets suggested preferential enhanced signaling of PLC-gamma over PLC-beta (see Figure 4). BLK is a Src family tyrosine kinase has been demonstrated to activate PLC-gamma via phosphorylation of the adaptor protein B-cell adaptor protein with ankyrin repeats (BANK1) (33). PTPN6 is a known positive regulator of Src family tyrosine kinases and deficiencies in PTPN6 signaling has been linked to decreased phosphorylation and activation of PLC-gamma (34). All of this is in agreement with our platelet aggregation results as PLC-gamma is activated by collagen stimulation of GP6, while PLC-beta is activated by ADP, epinephrine or thrombin stimulation of GPCRs via Gq.
Up-regulated cytokines in AA platelets.
Comparison of AA and EA platelet transcriptomes revealed a number of up-regulated cytokines in AA platelets, including IL9, IL32, PROK2, CCL3L1, CCL4L2 and LGALS1. The participation of these up-regulated cytokines in AA hemostasis has not been fully explored and hence we tested the role, if any, of IL32 and PROK2 on collagen-stimulated aggregation of AA platelets. Interestingly, PROK2 and IL32 exhibited opposing pro-aggregatory and anti-aggregatory effects in antibody neutralization experiments, respectively. We then preformed reciprocal experiments by stimulating washed EA platelets with these two cytokines. IL32 inhibited collagen-stimulated platelet aggregation, while PROK2 had no effect on collagen-stimulated platelet aggregation and PROK2 alone did not stimulate platelet aggregation. These findings suggest that PROK2 by itself is insufficient to stimulate aggregation of washed platelets. It is well established that the cellular effects of a specific cytokine can require the presence of an additional cytokine or milieu of cytokines in a cell context-specific manner (35).
Additional AA up-regulated genes portending hyperaggregability.
Our study also found that phosphatidylcholine transfer protein (PCTP) expression was significantly higher in AA versus EA platelets (Table S3, Figure 4). This finding is consistent with previous work that also reported higher PCTP expression in AA platelets (10). Given that PCTP is associated with greater degree of platelet aggregation and increased risk of myocardial infarction, this represents another possible factor that contributes towards the increase risk of thrombotic disease in AAs (36).
Our eQTL analysis on whole blood identified numerous eQTLs for the DEGs. One gene of particular interest was beta-tubulin 2A (TUBB2A; Figure 4), which had several eQTLs with very high Fst. Four eQTLs (rs6934343, rs28661335, rs7747681, rs9503444) exhibited Fst values of over 0.6 with the major allele in African populations associated with increased expression of TUBB2A. Fst is a measure of population differences in allele frequency which may be driven by natural selection. Previous studies have found that TUBB2A platelet gene expression as well as SNPs regulating TUBB2A gene expression has been associated with paclitaxel and vincristine induced neurotoxicity (37, 38). TUBB2A whole blood expression was also was significantly upregulated in acute myocardial infarction in a Chinese patient cohort (39). Taken together, TUBB2A may play a role in arterial thrombus formation. Another gene of interest was gamma glutamyl transferase 1 (GGT1). We identified 7 high Fst eQTLs (rs3966287, rs34297729, rs17003991, rs6003892, rs8136927, rs6003894, rs5996618) ranging from 0.4 to 0.7 with the major allele in African populations associated with higher GGT1 expression. Serum GGT activity has been associated with CVD progression including stent restenosis, myocardial infarction and stroke (40, 41). It has been posited that a major source for serum GGT levels may by platelets (42), representing a potential novel target for pharmacologic intervention of CVD.
Differentially expressed AA genes encoding pharmacologically targetable signaling proteins.
Of interest is the large number of differentially expressed AA platelet genes encoding GPCRs, ligand-gated ion channels, intracellular kinases, cytokines and secreted factors. Our study has generated a comprehensive list of pharmacologically- and biologically (i.e., neutralizing antibodies)-targetable signaling molecules for potential repurposing of drugs with the aim of treating thrombotic events (see Table 1). Examples include LIMK2 inhibited by CRT0105446, TUBB2A by docetaxel and colchicine, nicotinic acetylcholine receptor (CHRNA7) by succinylcholine, LGALS1 by OTX008, FGFR1 by dovitinib, HDAC11 by belinostat, NDUFA6 by hydroquinone derivatives, and cytokines by neutralizing antibodies.
TABLE 1:
AA differentially expressed platelet genes encoding pharmacologically targetable signaling proteins
| Gene | Direction of change in AA platelets based on RNA-Seq | Agonist/Activator | Antagonist/Inhibitor | Approved for use in |
|---|---|---|---|---|
| BLK | Up | osimertinib | metastatic EGFRm NSCLC | |
| BTK | Down | ibrutinib | chronic lymphocytic leukemia, small lymphocytic lymphoma | |
| CD19 | Up | blinatumomab | minimal residual disease in acute lymphoblastic leukemia | |
| CD22 | Up | inotuzumab | B-cell acute lymphoblastic leukemia | |
| CD79B | Up | polatuzumab vedotin | relapsed or refractory diffuse large B-cell lymphoma | |
| CHRNA7 | Up | acetylcholine | mecamylamine, tubocurarine | mecamylamine is used to treat hypertension. Tubocurarine is a neuromuscular blocker. |
| FGFR1 | Down | sorafenib | advanced renal cell carcinoma | |
| GABRE | Up | benzodiazepine | generalized anxiety disorder, insomnia, seizures, social phobia, and panic disorder | |
| GABRR2 | Up | benzodiazepine | generalized anxiety disorder, insomnia, seizures, social phobia, and panic disorder | |
| HDAC11 | Up | belinostat | relapsed or refractory peripheral T-cell lymphoma | |
| ITGAV | Down | cilengitide | sarcoma, gliomas, lymphoma, leukemia, and lung Cancer (investigational) | |
| LGALS1 | Up | OTX008 | solid tumors (investigational) | |
| LIMK2 | Up | encorafenib | unresectable or metastatic melanoma with a BRAF V600E or V600K mutation | |
| MAPK11 | Up | regorafenib, Pirfenidone | regorafenib is approved for metastatic colorectal cancer, advanced gastrointestinal stromal tumors, and hepatocellular carcinoma; pirfenidone is approved for idiopathic pulmonary fibrosis | |
| PTPN6 | Up | tiludronate, SC-43 | tiludronate is approved for the treatment of Paget’s disease of bone | |
| TRPM2 | Up | econazole, miconazole | econazole is approved in the treatment of tinea pedis, tinea cruris, and tinea corporis; miconazole is approved to treat mucosal yeast infections | |
| TUBB2A | Up | Colchicine* Docetaxel* |
docetaxel is approved for breast, ovarian, and non-small cell lung cancer; colchicine is approved to treat gout |
Inhibit microtubule function
Study limitations.
There are limitations to this analysis that should be considered when generalizing these results. First, we relied upon volunteer responses, rather than medical records/chart reviews, to screen for inclusion/exclusion criteria (e.g. free of known coagulopathies, CVD). Additional health information such as preexisting co-morbidities, BMI, blood pressure, serum lipids and insulin resistance was not ascertained in our study. The sample size of our Discovery and Validation cohorts was relatively small given the high volume of blood required to extract sufficient platelets to produce reliable DEG analysis. This resulted in the potential for moderate DEGs between populations to be missed. The eQTLs used to calculate Fst values were mapped in whole blood and thus may not fully represent the regulatory landscape of platelet genes. Nonetheless, a majority of eQTLs have been shown to regulate gene expression across tissue types (43). Our previous work has shown that eQTLs found unique in African-ancestry populations have a higher Fst than those found across populations (44). Future studies are needed to definitively determine whether high Fst SNPs drive the observed gene expression differences. While genetic regulatory variants are known to play a role in gene expression differences, other untested factors may also be contributing, including epigenetic differences between populations, environmental and lifestyle differences, and co-morbid conditions. It remains to be determined if manipulation of IL32 and PROK2 cytokine signaling represents an effective strategy to modulate platelet function given the broad toxicities associated with such manipulations. It is also important to note that both thrombotic and hemorrhagic diseases involve a complex interplay among many cellular factors including platelets, immune cells and endothelial cells (45). This investigation only focused on one aspect of this complex process by providing a snapshot of DEGs in AA and EA platelets. Regardless, this is the first RNA-Seq study that also couples ancestry analysis to identify significant population differences in platelet gene expression and signaling, a tissue with critical importance to CVD.
Supplementary Material
Figure S3. Quantitative RT-PCR validation of the IL32 and PROK2 genes in a separate cohort of volunteers. Quantitative RT-PCR (qRT-PCR) results from a separate cohort of AA = 8 and EA = 8 are presented in the left panel and corresponding RNA-Seq results in the right panel. Results are presented as mean ± SEM (*p<0.05).
Figure S2. Differentially expressed platelet genes in the AA and EA populations. Heat map of 480 significant differentially expressed genes in platelets of AA individuals compared to EA individuals. Rows represent genes, columns represent healthy volunteers. Scale bar below the heat map is given on a log2 scale.
Figure S1. Platelet transcriptome landscape. Cellular location and cellular function categories of encoded proteins in the (A) overall platelet transcriptome (AA and EA populations combined), and (B) differentially expressed platelet transcriptome of the AA population compared to EA population. Cellular location and function categories based on IPA analysis.
STUDY HIGHLIGHTS.
What is the current knowledge on the topic?
Disparities in CVD exist between African and European-Americans. While platelets play a key role in thrombotic disease, little is known on how AAs and EAs differ in platelet gene expression.
What question did this study address?
This study focused on differentiating the platelet gene expression landscape between healthy AAs and EAs.
What does this study add to our knowledge?
The results suggest significant differences in platelet gene expression may portend increased platelet activity in AA individuals. Moreover, eQTL analysis suggests that many of the DEGs are related to sequence variation between these two populations.
How might this change clinical pharmacology or translational science?
Interestingly, many of the DEGs represent potentially novel targets for future anti-platelet therapy. Notably, we provide functional validation of the opposing actions of two cytokines, IL32 and PROK2, on collagen-induced platelet aggregation in AAs.
ACKNOWLEDGEMENTS
This work was supported by grants from the National Institutes of Health (U54MD010723 and R01CA204806). The authors wish to thank the University of Maryland Institute for Genome Sciences for performing the RNA-Seq.
Footnotes
The authors declared no competing interests for this work.
REFERENCES
- (1).Thomas KL, Honeycutt E, Shaw LK & Peterson ED Racial differences in long-term survival among patients with coronary artery disease. Am Heart J 160, 744–51 (2010). [DOI] [PubMed] [Google Scholar]
- (2).Clark LT et al. Coronary heart disease in African Americans. Heart Dis 3, 97–108 (2001). [DOI] [PubMed] [Google Scholar]
- (3).Centers for Disease, C. & Prevention. Disparities in premature deaths from heart disease--50 States and the District of Columbia, 2001. MMWR Morb Mortal Wkly Rep 53, 121–5 (2004). [PubMed] [Google Scholar]
- (4).Collins SD et al. Does black ethnicity influence the development of stent thrombosis in the drug-eluting stent era? Circulation 122, 1085–90 (2010). [DOI] [PubMed] [Google Scholar]
- (5).Koupenova M Potential role of platelets in COVID-19: Implications for thrombosis. Res Pract Thromb Haemost 4, 737–40 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (6).Mezger M et al. Platelets and Immune Responses During Thromboinflammation. Front Immunol 10, 1731 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (7).Kimura M, Cho JH, Lasker N & Aviv A Differences in platelet calcium regulation between African Americans and Caucasians: implications for the predisposition of African Americans to essential hypertension. J Hypertens 12, 199–207 (1994). [PubMed] [Google Scholar]
- (8).Bray PF et al. Heritability of platelet function in families with premature coronary artery disease. J Thromb Haemost 5, 1617–23 (2007). [DOI] [PubMed] [Google Scholar]
- (9).Tourdot BE, Conaway S, Niisuke K, Edelstein LC, Bray PF & Holinstat M Mechanism of race-dependent platelet activation through the protease-activated receptor-4 and Gq signaling axis. Arterioscler Thromb Vasc Biol 34, 2644–50 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (10).Edelstein LC et al. Racial differences in human platelet PAR4 reactivity reflect expression of PCTP and miR-376c. Nat Med 19, 1609–16 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (11).Kong X, Simon LM, Holinstat M, Shaw CA, Bray PF & Edelstein LC Identification of a functional genetic variant driving racially dimorphic platelet gene expression of the thrombin receptor regulator, PCTP. Thromb Haemost 117, 962–70 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (12).Qayyum R et al. Genome-wide association study of platelet aggregation in African Americans. BMC Genet 16, 58 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (13).Pendyala LK et al. Racial disparity with on-treatment platelet reactivity in patients undergoing percutaneous coronary intervention. Am Heart J 166, 266–72 (2013). [DOI] [PubMed] [Google Scholar]
- (14).Nagalla S et al. Platelet microRNA-mRNA coexpression profiles correlate with platelet reactivity. Blood 117, 5189–97 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (15).Bolger AM, Lohse M & Usadel B Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–20 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (16).Patro R, Duggal G, Love MI, Irizarry RA & Kingsford C Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods 14, 417–9 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (17).Love MI, Huber W & Anders S Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (18).Yekutieli D & Benjamini Y Resampling-based false discovery rate controlling multiple test procedures for correlated test statistics. J Stat Plan Inference 82, 171–96 (1999). [Google Scholar]
- (19).Purcell S et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 81, 559–75 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (20).Friedman PN et al. The ACCOuNT Consortium: A Model for the Discovery, Translation, and Implementation of Precision Medicine in African Americans. Clin Transl Sci 12, 209–17 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (21).Zhong Y, Perera MA & Gamazon ER On Using Local Ancestry to Characterize the Genetic Architecture of Human Traits: Genetic Regulation of Gene Expression in Multiethnic or Admixed Populations. Am J Hum Genet 104, 1097–115 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (22).Delaneau O, Coulonges C & Zagury JF Shape-IT: new rapid and accurate algorithm for haplotype inference. BMC Bioinformatics 9, 540 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (23).Howie BN, Donnelly P & Marchini J A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet 5, e1000529 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (24).Stegle O, Parts L, Durbin R & Winn J A Bayesian framework to account for complex non-genetic factors in gene expression levels greatly increases power in eQTL studies. PLoS Comput Biol 6, e1000770 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (25).Benjamini Y & Hochberg Y Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statist Soc 57, 289–300 (1995). [Google Scholar]
- (26).Shabalin AA Matrix eQTL: ultra fast eQTL analysis via large matrix operations. Bioinformatics 28, 1353–8 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (27).Babinska A, Markell MS, Salifu MO, Akoad M, Ehrlich YH & Kornecki E Enhancement of human platelet aggregation and secretion induced by rapamycin. Nephrol Dial Transplant 13, 3153–9 (1998). [DOI] [PubMed] [Google Scholar]
- (28).Stalker TJ, Newman DK, Ma P, Wannemacher KM & Brass LF Platelet signaling. Handb Exp Pharmacol, 59–85 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (29).Cognasse F et al. Platelet Inflammatory Response to Stress. Front Immunol 10, 1478 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (30).Carrasco Pro S et al. Global landscape of mouse and human cytokine transcriptional regulation. Nucleic Acids Res 46, 9321–37 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (31).Mok J, Park TS, Kim S, Kim D, Choi CS & Park J Prokineticin receptor 1 ameliorates insulin resistance in skeletal muscle. FASEB J, (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (32).Qiu CY, Liu YQ, Qiu F, Wu J, Zhou QY & Hu WP Prokineticin 2 potentiates acid-sensing ion channel activity in rat dorsal root ganglion neurons. J Neuroinflammation 9, 108 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (33).Bernal-Quirós M, Wu Y-Y, Alarcón-Riquelme ME & Castillejo-López C BANK1 and BLK Act through Phospholipase C Gamma 2 in B-Cell Signaling. PLoS One 8, e59842 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (34).Mazharian A et al. Megakaryocyte-specific deletion of the protein-tyrosine phosphatases Shp1 and Shp2 causes abnormal megakaryocyte development, platelet production, and function. Blood 121, 4205–20 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (35).Bartee E & McFadden G Cytokine synergy: an underappreciated contributor to innate anti-viral immunity. Cytokine 63, 237–40 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (36).Mao G et al. Transcription Factor RUNX1 Regulates Platelet PCTP (Phosphatidylcholine Transfer Protein): Implications for Cardiovascular Events: Differential Effects of RUNX1 Variants. Circulation 136, 927–39 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (37).Leandro-Garcia LJ et al. Regulatory polymorphisms in beta-tubulin IIa are associated with paclitaxel-induced peripheral neuropathy. Clin Cancer Res 18, 4441–8 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (38).Martin-Guerrero I et al. Variants in vincristine pharmacodynamic genes involved in neurotoxicity at induction phase in the therapy of pediatric acute lymphoblastic leukemia. Pharmacogenomics J 19, 564–9 (2019). [DOI] [PubMed] [Google Scholar]
- (39).Su J, Gao C, Wang R, Xiao C & Yang M Genes associated with inflammation and the cell cycle may serve as biomarkers for the diagnosis and prognosis of acute myocardial infarction in a Chinese population. Mol Med Rep 18, 1311–22 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (40).Ruttmann E et al. Gamma-glutamyltransferase as a risk factor for cardiovascular disease mortality: an epidemiological investigation in a cohort of 163,944 Austrian adults. Circulation 112, 2130–7 (2005). [DOI] [PubMed] [Google Scholar]
- (41).Ulus T et al. Serum gamma-glutamyl transferase activity: a new marker for stent restenosis? Atherosclerosis 195, 348–53 (2007). [DOI] [PubMed] [Google Scholar]
- (42).Gurdol F, Nwose OM & Mikhailidis DP Gamma-glutamyl Transferase Activity in Human Platelets: Quantification of Activity, Isoenzyme Characterization and Potential Clinical Relevance. Platelets 6, 200–3 (1995). [DOI] [PubMed] [Google Scholar]
- (43).Consortium GT et al. Genetic effects on gene expression across human tissues. Nature 550, 204–13 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (44).Zhong Y, De T, Alarcon C, Park CS, Lec B & Perera MA Discovery of novel hepatocyte eQTLs in African Americans. PLoS Genet 16, e1008662 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- (45).Esmon CT The interactions between inflammation and coagulation. Br J Haematol 131, 417–30 (2005). [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S3. Quantitative RT-PCR validation of the IL32 and PROK2 genes in a separate cohort of volunteers. Quantitative RT-PCR (qRT-PCR) results from a separate cohort of AA = 8 and EA = 8 are presented in the left panel and corresponding RNA-Seq results in the right panel. Results are presented as mean ± SEM (*p<0.05).
Figure S2. Differentially expressed platelet genes in the AA and EA populations. Heat map of 480 significant differentially expressed genes in platelets of AA individuals compared to EA individuals. Rows represent genes, columns represent healthy volunteers. Scale bar below the heat map is given on a log2 scale.
Figure S1. Platelet transcriptome landscape. Cellular location and cellular function categories of encoded proteins in the (A) overall platelet transcriptome (AA and EA populations combined), and (B) differentially expressed platelet transcriptome of the AA population compared to EA population. Cellular location and function categories based on IPA analysis.