Upon a genome-wide evaluation of the baseline miRNA expression levels in HapMap LCLs derived from CEU and YRI HapMap samples, we identified a set of 33 miRNAs (16% of all miRNAs reliably interrogated in our study), whose expression levels significantly differ in vitro between these 2 populations. To examine the functional significance of these differentially expressed miRNAs, we set out to comprehensively investigate the effect of these miRNAs on the transcriptome under the hypothesis that gene expression is an endophenotype or intermediate to more complex traits. We further identified a set of 29 mRNAs whose expression levels differed between CEU and YRI and were inversely correlated with the differentially expressed miRNAs. In addition, a high proportion (64%) of the differentially expressed miRNAs was found to correlate with cellular sensitivity to various chemotherapeutic agents (FDR < 0.05), for which population differences in drug sensitivity have been observed in vitro. All these suggest the biological and pharmacological significance of observed differences in miRNA expression.
Population differences have been observed in many human complex traits, including disease susceptibility,29,30
and gene expression.6–8
Genetic polymorphisms have undergone extensive evaluation for their potential role in these observed population differences. In recent years, the significance of epigenetic factors in regulating complex traits has been actively pursued. MiRNAs have been shown to influence gene expression variation in the human genome33
and may affect other biological processes through their effect on the transcriptome. In addition, inter-ethnic differences in miR-26a expression have been observed in prostate cancer cell lines derived from African American men compared with those derived from Caucasians.34
In our study, we did not observe a significant difference in miR-26a expression; however, miR-26b expression does show a potential inter-ethnic difference with p value 0.00078 in LCLs derived from CEU and YRI (Sup. Table 1
). However, to date, a comprehensive investigation of population differences in miRNA expression has not been done. Our results demonstrate that miRNA expression levels exhibit population differences. Furthermore, the differential miRNA expression may contribute to observed population differences in mRNA expression levels and possibly other complex human traits.
Our findings indicate that certain differentially expressed miRNAs consistently correlated with the same target mRNA in both CEU and YRI samples. For example, miR-30b, 30d and 30e expression were negatively correlated with the expression of ECE1
in both populations (FDR < 0.1). ECE1 protein is involved in proteolytic processing of endothelin precursors to biologically active peptides. The expression of this gene product has been linked to non-small cell lung cancer,35
and Alzheimer disease.37
When we limited our analysis to differentially expressed mRNAs, we identified several miRNA-mRNA pairs that were inversely correlated in both CEU and YRI and have previously been implicated in studies of disease susceptibility: miR-342-3p and ANKRD49
, miR-30b, d and e and ZNRF1
. For example, miR-342-3p has been shown to be upregulated in neurodegenerative disease.38,39 ANKRD49
was identified as one of the 4 non-small cell lung cancer invasion-associated gene signatures, which can independently predict prognosis regardless of age, gender and stage of disease.40
miR-30b was found to be down-regulated in squamous cell lung carcinoma compared with normal lung tissue.41
It is also downregulated in multiple myeloma,42
in invasive urothelial carcinoma of the bladder,43
in polycythemia vera (a colonal hematopoietic stem cell disorder) reticulocytes.44
On the other hand, amplification and overexpression of miR-30b was observed in medulloblastoma.45 ZNRF
proteins play a role in the establishment and maintenance of neuronal transmission and plasticity via their ubiquitin ligase activity.46
Overexpression of ZNRF1
, which has been identified as a crucial molecule in nerve regeneration, causes morphological changes such as neurite-like elongation. ZNRF1
mediates regulation of neuritogenesis via interaction with tubulin.47
One caveat of the study is that these are LCLs collected at different times and cultured for different periods of time. Because of the differences in cell line collection time between the CEU and YRI samples,48,49
miRNA differences could be a combined effect of genetic and non-genetic factors. In addition, culture conditions or batch to batch variation could influence the observed differences in gene expression between the two populations.50
Therefore, to reduce these variables, cell culture protocols were optimized and samples (CEU and YRI) were randomized when cultured and evaluated for miRNA expression in this study.
In addition to their role in disease susceptibility, miRNAs have also been described to directly or indirectly affect drug efficacy and toxicity. For example, miR-214 was found to induce cell survival and cisplatin resistance by directly targeting and downregulating the PTEN
gene in ovarian cancer.21
Studies have shown that a miRNA regulates the expression of CYP3A4
, a major phase I drug metabolism enzyme, by targeting its major transcription factor, PXR.17
Furthermore, miRNAs have been shown to alter resistance to cytotoxic anticancer therapy.20
Our study supports the general conclusion of these findings with respect to miRNA function; we found that a substantial proportion (64%) of the differentially expressed miRNAs correlated with cellular sensitivity to at least one of the three chemotherapeutic agents evaluated in this study in one of the HapMap populations. These findings strongly argue for the inclusion of miRNAs in comprehensive pharmacogenomic studies.
Finally, we investigated the question of whether the population differences in miRNA expression are influenced by genetic polymorphisms. Of the 2,770 SNPs showing high levels of population differentiation between the CEU and YRI populations (Fst > 0.25), nearly 10% predict the expression level (p < 10−5) of at least one of 33 differentially expressed miRNAs. Furthermore, four of these differentially expressed miRNAs (12%) have pre-miRNAs in known CNV regions. Our findings thus highlight the potential role of genetic variants in mediating the observed population differences in miRNA expression.