There are two major outcomes of this study. First, we have characterized atrial miR-133a species and find an extraordinary diversity of 5′ and 3′ miR-133a-3p and miR-133a-5p isomiRs, with two major 3p isomiRs and one major 5p isomiR in both human and mouse. These isomiRs have minimally-overlapping suites of predicted mRNA targets and functions, which suggests that they have distinct biological roles. Second, we report a novel MIR133A2 variant that alters strand abundance during miRNA processing and results in accumulation of miR-133a-5p. This variant is one of only a very few functional miRNA gene variants reported to date.
Despite the enormous interest in miRNAs as master regulators of gene expression, the role of genetic variation is incompletely understood. Single nucleotide polymorphisms (SNPs) in predicted miRNA binding sites on target mRNAs occur frequently and some of these have been associated with phenotypic traits
[22–25]. In contrast, SNPs in the miRNA genes themselves are far less common. Several SNPs in pri-miRNA and pre-miRNA have been functional evaluated and shown to result in defective processing and reduced levels of mature miRNA. SNPs are rarely seen (<1%) in the seed regions that are crucially required for target recognition
. These observations have provided an argument for strong selective constraint and suggest that seed variants would have substantial effects. The discovery of point mutations in the seed region of miR-96 in two families with nonsyndromic progressive hearing loss provided the first example of human Mendelian disease associated with a miRNA gene variant
. Functional SNPs can occur in mature miRNAs outside the seed region, as recently demonstrated by the finding of a rare variant at nt 17 of miR-499
. Evaluation of the effects of this variant in transgenic mice showed that it protected against the cardiomyopathy that developed with overexpression of the wildtype form of miR-499. Here we describe a MIR133A2 variant located in the mature miRNA outside the seed region that changes the way that the miRNA duplex is processed.
MiRNA processing is conventionally assumed to yield only one functional mature miRNA with the passenger strand being degraded and hence, functionally irrelevant. It has recently been appreciated however, that passenger forms are not always selected for degradation, rather, that both 3p and 5p strands of the miRNA duplex may be simultaneously present and that their ratio can be regulated in a tissue- or development-specific manner, suggesting that both strands have inherent functional activity
. Here we find that both (mature) miR-133a-3p and (passenger) miR-133a-5p are present in the atrium in humans and mice, with miR-133a-5p normally representing <1% of all miR-133a species. The main effect of the 79T > C MIR133A2 variant is to alter the relative ratio of miR-133a-3p and 5p strands.
Current models posit that the strand of the miRNA duplex with weaker base-pairing at the 5′ end is preferentially incorporated into RISC
. If the conventional murine miR-133a-5p sequence as annotated by miRBase (v18) represented the predominant isomiR in the normal human heart, then the nt corresponding to position 79 would lie just outside the base-paired region of the processed miRNA duplex. However, our deep sequencing data clearly show that the −1 isomiR is the most abundant miR-133a-5p species in the human atrium. Consequently, the 79T > C variant lies within the duplex and would directly prevent base-pairing and weaken thermostability at this site, favoring incorporation of miR-133a-5p into RISC. Additionally the lack of base pairing at position 79C is likely to enlarge the bulge in the stem-loop structure, a feature known to influence positional cleavage by Dicer and Drosha and result in increased variability of 5′ isomiRs
[8, 29]. The canonical miRNA biosynthesis pathway predicts 3p and 5p miRNAs of 22 nt in length, but deep sequencing studies have revealed a surprising extent of isomiR diversity in most tissues and species
[5–8]. These isomiRs may include 5′ or 3′ cleavage variations, and non-templated additions or trimming at the 3′ end. It has been found that the relative abundance and types of isomiRs in specific tissues can vary over time and with disease states, suggesting that they have distinctive biological functions
[5, 8, 30].
It is conventionally assumed that the seed regions are the major determinants of target specificity, and thus it is plausible that 5′ isomiR variation can affect mRNA target selection. For example, our group has recently demonstrated that the two most abundant miR-133a isomiRs in murine atrial HL-1 cells have different targeting properties
. Given that each isomiR has hundreds of mRNA targets, dynamic fluxes in isomiR characteristics may give rise to an unexpected complexity of miRNA:mRNA interactions and provide a mechanism for intricate regulation of gene expression in specific tissues. To assess the potential impact of the 79T > C MIR133A2 variant, we searched for differences in miR-133a-3p and miR-133a-5p predicted mRNA targets using the respective human and murine seed regions and a set of genes known to be differentially regulated in miR-133a knockout mice
. These results showed that only a minority of predicted miR-133a targets were shared, and that most were unique to either 3p or 5p forms. With the 79T > C MIR133A2 variant, no changes in miR-133a-3p target gene expression would be expected. In contrast, the relative increase in miR-133a-5p could have a relatively greater impact and give rise to selective repression of the 5p suite of targets. It is notable that at least 25% of the lowest-abundance miR-133a-5p targets include mRNAs involved in regulation of transcription, signaling and membrane transport. Further studies are required to determine whether changes in miR-133a-5p directly alter levels of these critical molecules and have biologically-significant functional effects.
The relationship, if any, between the 79T > C MIR133A2 variant and AF in the Family KB proband is questionable. Although DNA samples were unavailable from the deceased parents, it seems most likely that the 79 T > C substitution is a de novo sequence change which was uniquely present in one family member. A causal gene mutation that can account for the familial pattern of AF in this kindred has yet to be identified. The presence of the MIR133A2 variant in the Family KB proband could foreseeably have a modifying effect by altering gene expression profiles in the atrium. A number of factors in addition to genetic variation need to be considered in this individual, including the impact of AF per se and of co-existent risk factors for AF. Altered expression of miR-133 itself has been observed in cardiac tissues from patients with AF
[18, 19], and conditions that predispose to AF, such as atrial dilation, ventricular hypertrophy, and myocardial ischemia
[12, 31]. It is notable that individual II-5 had both hypertension and valvular heart disease which independently increase AF risk. A limitation of this study is the lack of atrial tissue samples to document gene expression patterns in 79T and 79C family members. The collective effects of the multiple variables present in individual II-5 on atrial gene expression and the electrical and structural properties of the atrial wall are difficult to predict and could promote or protect against AF.