Aspirin’s effects of platelet gene expression
After quality analysis of the platelet RNA sequencing data there were
312 platelet mRNA datasets available for analysis from 73 unique
patients representing 14333 transcripts. In general, the level of
contamination by granulocytes was low (proportion of samples withPTPRC /ITGA2B count ratio ≥ 0.01 =6/312) and did not vary
by treatment (likelihood ratio test P-value = 0.11). We have previously
shown that platelet gene expression levels are highly correlated within
an individual between Visits 1 and Visit 4.[23] Therefore in all
subsequent analyses, these two visits were treated equivalently as
no-drug treatment visits and no adjustment for leukocyte contamination
was performed.
To test our primary hypothesis that aspirin exposure alters the platelet
mRNA transcriptome, we used a mixed-effects regression framework to
identify transcripts that changed with any aspirin exposure compared to
no drug treatment visits. In a univariate analysis using the false
discovery rate (FDR) q-value < 0.10 to account for multiple
testing we identified 208 transcripts that were differentially expressed
(43 down regulated, 165 upregulated, Figure 2, Supplemental Data File 1)
after aspirin exposure. To test for aspirin-dose response relationships
we used two complementary approaches: 1) contrasting the effects of 81mg
vs. 325mg/day and 2) correlating gene expression with salicylate
concentrations during aspirin treatment. We found no evidence for
differences on platelet gene expression for the set of 208 gene aspirin
differentially expressed genes and higher aspirin dose or salicylate
concentrations (enrichment p-values = 0.3 and 0.9, respectively).
We used banked platelet RNA from prior research cohorts[19] of
healthy adults and patients with diabetes (n= 91 across both cohorts)
exposed to 81mg/day or 325mg/day aspirin using a custom Nanostring assay
to validate the effects of aspirin. Due to limits on the number of
transcripts available in a Nanostring assay, we prioritized the list 208
transcripts based on mean expression and robustness to various starting
points for quantifying expression, and genes that exhibited positive
correlation (r > 0.25) between RNAseq and Nanostring to a
subset of 51 genes to test for validation. Among these 51 genes, we
found five that independently validated in the replication cohorts
(fixed effects meta-analysis p-value for aspirin effect <
0.05) as changing in response to aspirin exposure: EIF2S3, CHRNB1,
EPAS1, SLC9A3R2, and HLA-DRA . (Figure 2B)