A Smoking-Associated miRNA-mRNA Co-Expression Network

Adverse effects of cigarette smoking on our health is well documented. On the other hand, we do not fully understand the mechanisms by which these adverse effects occur. In this issue of Circulation Cardiovascular Genetics, Willinger, Larson, Levy, and colleagues¹ identified a smoking-related microRNA (miRNA) -messenger RNA (mRNA) co-expressed gene network in the whole blood of previous and current smokers. They performed the study as part of the extensive phenotypic characterization performed in the participants of the Framingham Heart Study². miRNAs, which are small non-coding RNAs that regulate gene expression by binding to their target messenger RNAs (mRNAs) causing translational repression or target degradation³, have been recognized for their role in cardiovascular pathology^4–6. The results of the current study add another dimension to previous studies by pointing to dysregulated inflammatory responses in leukocytes by miRNA-specific gene regulatory networks in cigarette smokers. The authors arrived at this conclusion via a series of well-designed experiments. They used total RNA isolated from peripheral whole blood obtained from 5,023 non-fasting participants of the Framingham Study. These participants were part of the Offspring and Third Generation cohorts. The mean age of the participants was 55 years old, and 54% of them were female. 10% of them were current smokers, 41% were former smokers, and 48% had never smoked. In this study, former smokers were defined as those participants who had not smoked a cigarette in the past year before data collection. As expected, current smokers had increased systemic inflammation and airflow obstruction.

The authors first measured the expression level of 754 microRNAs in a subset of their samples using a high throughput RT-qPCR platform employing Taqman miRNA assays. This number represents approximately 30% of all known human miRNAs⁷. After extensive quality control, they determined that they could reliably measure the expression level of 289 miRNAs in their samples. Using linear regression, they identified six miRNAs which were associated with smoking status; only one was positively correlated (Figure 1). Five of these miRNAs were associated with current smokers, whereas one was associated with former smokers compared to those who never smoked. These miRNAs could be biomarkers of smoking or smoking-related complications of pulmonary function.

Figure 1.

Smoking-related microRNAs in the Framingham cohort. miRNA expression was measured in the peripheral whole blood isolated from 5,023 participants of the Framingham Heart Study. − and + represent negative and positive, associations with smoking, inflammation, and pulmonary function, respectively. n.s. represents non-significant results.

The authors then tested the association of six smoking-related miRNAs for inflammatory markers in the serum and measures of pulmonary function derived from spirometry tests. miR-1180, which had lower expression in current smokers, was negatively associated with C-reactive protein (CRP) levels in the serum, and positively associated with the prevalence of airway obstruction. Unexpectedly, miR-1285-3p, which also had lower expression in current smokers, was positively associated with serum interleukin-6 levels, a pro-inflammatory cytokine. It is possible that the miR-1285-3p association is due to a false positive finding as many statistical tests were performed in this large-scale study. If true, however, additional studies will be needed to identify the discrepancy between the miRNA associations with CRP and IL-6.

The authors further examined the co-expression pattern of mRNAs with the six smoking-related miRNAs. They measured the expression level of 17,873 genes using Affymetrix Human Exon 1.0 ST microarrays. More than 2,000 genes were correlated with these miRNAs, however, only 116 of the miRNA-mRNA pairs in which the mRNA was the predicted or validated target of the correlated miRNA, suggesting that the co-expression analyses captured the downstream impact of perturbed miRNAs. Since a single miRNA can potentially target hundreds of transcripts, even minor differences in miRNA abundance in a population can have significant impact; therefore, it will be important to understand the primary targets of dysregulated miRNAs. The majority of the enriched Gene Ontology pathways for the co-expressed genes were in inflammatory pathways, pointing to genes involved in monocyte differentiation, viral response, and T cell homeostasis. Overexpression of two of the miRNAs, miR-1180 and miR-1285-3p, in lung epithelial cells resulted in changes in the release of granulocyte macrophage colony-stimulating factor (GM-CSF), chemokine C-X-C motif ligand 5 (CXCL5), and chemokine C-X-C motif ligand 6 (CXCL6) under basal and tumor necrosis factor alpha-stimulated cells. Interestingly, IL-6 production did not change.

Smoking-related systemic inflammation leads to changes in the cellular composition of the blood, with prominent influences on the proportion of leukocytes; therefore, it is plausible that the miRNA differences may be a consequence of this change. To mitigate this, the authors first estimated the counts of multiple blood cell types using the whole genome gene expression data and then used these counts as covariates in their linear regression model to remove the bias. However, this is an imperfect method since imputed CBC counts had a prediction accuracy ranging from 0.25 to 0.89 depending on the cell type⁸. In the future, how smoking leads to changes in miRNA expression which then regulate their primary target genes in subsets of leukocytes will provide a mechanistic understanding of the molecular impact of cigarette smoking.

Overall, this comprehensive study identified a set of smoking-associated miRNAs some of which were also correlated with inflammatory markers and pulmonary function in a well-characterized cohort. These miRNAs may be biomarkers of exposure to cigarette smoke or may act mechanistically to regulate target gene expression. Two previous studies of subcutaneous adipose tissue miRNA expression in moderately sized cohorts^{9, 10} found very little overlap of their results; therefore, validation of the results of the current study in an independent cohort will be needed to establish the role of these miRNAs as biomarkers. This will certainly be challenging as various technical difficulties in reliably measuring miRNAs using different methods, such as RT-qPCR, microarrays, or small RNA sequencing, have been extensively documented^{11, 12}. Further mechanistic studies and mediation analysis will also determine whether the differences in miRNA levels are the cause or consequence of inflammation or smoking related diseases. Some of the authors of the current study previously published the results of their analyses on the genetic loci which are associated with the abundance of miRNAs in the same study participants¹³. Based on those results, it appears that genetic variants do not regulate the expression levels of six smoking-related miRNAs, suggesting that environmental factors lead to differences in miRNA expression among smokers and non-smokers.