Co-regulation of intragenic microRNA miR-153 and its host gene Ia-2 β: identification of miR-153 target genes with functions related to IA-2β in pancreas and brain

Genomic localisation of many mammalian miRNA transcription units has been determined previously [22]. Using these data and the University of California Santa Cruz (UCSC) genome browser (http://​genome.​ucsc.​edu/​cgi-bin/​hgTracks, accessed Sept 2012), we determined that two copies of the miR-153 small non-coding miRNAs exist in the human genome. In humans, miR-153-1 is localised at a highly conserved region in intron 19 of IA-2 on chromosome 2 (chr2:219862589-219882387) (Fig. 1a). miR-153-2 is localised at a highly conserved intronic region between exon 19 and 20 of IA-2β on chromosome 7 (chr7:157059789-157059875) (Fig. 1a). Next, we used ClustalW multiple alignment software (www.​genome.​jp/​tools/​clustalw/​) to determine whether miR-153-1 and miR-153-2 are evolutionarily conserved (Fig. 1b). We found that although miR-153-2 is evolutionarily highly conserved, miR-153-1 is more divergent. In particular, in rodents (mouse, rat), polymorphisms are localised within the seed sequence of miR-153-1. This makes it very unlikely that functional miR-153 is generated from the miR-153-1 transcription unit in rodents. This was validated by performing a predicted hybridisation analysis using the RNAhybrid algorithm [23] and SNCA 3′ UTR as a confirmed miR-153 target [24] (ESM Fig. 1). These data indicate that, based on the strong evolutionary conservation of miR-153-2 within the intron of the IA-2β gene, miR-153-2 and IA-2β might be functionally connected and co-regulated.

To investigate whether expression of miR-153-2 and IA-2β are co-regulated, we determined whether the IA-2β mRNA and the pri-miR-153 transcript are generated from the same or separate heterogeneous nuclear (hn)RNA transcripts. We therefore first analysed the expression of mir153 in brain material of wild-type, AKO, BKO and DKO mice. AKO and BKO mice were generated previously, resulting in a complete loss of Ia-2 and Ia-2β mRNA expression and protein production [10, 12]. The knockout strategy does not target the intron in which miR-153-2 is located. Quantitative PCR measurements showed that miR-153 expression was reduced in BKO and DKO, but not in wild-type or AKO mouse brain material after normalisation to three independent small non-coding RNA controls (miR-16 [also known as Gde1], RNU19 [also known as Snora74a], Sno135; only results for the latter are shown) (Fig. 2a). To rule out the possibility that the observed remaining miR-153 was generated by cross-hybridisation of the oligonucleotides with the miR-153-1 sequences in the Ia-2 locus, we analysed expression of the pri-miR-153-2 product, which only results in the amplification of the miR-153-2, but not of the miR-153-1 product. We showed that pri-miR-153 expression was reduced to the same extent as that of mature miR-153 in BKO and DKO, but not in wild-type and AKO mouse brain RNA (Fig. 2b). Interestingly, pri-miR-153 expression was reduced to almost undetectable levels in isolated DKO islets compared with wild-type (Fig. 2c).

Next, we analysed mirR-153 expression in mouse MIN6B cells and isolated mouse pancreatic islets after glucose stimulation (Fig. 2d,e). We detected a significant increase in the expression of miR-153, Ia-2 and Ia-2β after glucose stimulation of MIN6B cells at 25 mmol/l glucose compared with 5 mmol/l glucose (Fig. 2d). Finally, increasing glucose levels from 3.3 to 16.7 mmol/l dramatically increased miR-153, Ia-2 and Ia-2β expression in cultured isolated mouse islets (Fig. 2e). These studies show that miR-153, Ia-2 and Ia-2β expression is increased after glucose stimulation in MIN6B cells and isolated mouse pancreatic islets.

These results also show that whereas miR-153 expression was almost completely lost in DKO islets (Fig. 2c), miR-153 levels in DKO brain were still detectable at about 50% of the levels found in wild-type mouse brain (Fig. 2b). This might indicate that in addition to miR-153 being generated from a transcript and transcription start site (TSS) shared with the host gene Ia-2β, an alternative tissue-specific transcript and TSS for miR-153 might exist independently of its host gene Ia-2β. Moreover, the targeted Ia-2β allele was generated by deleting its promoter region [10]. Thus, loss of the Ia-2β promoter rules out the possibility that any remaining miR-153 expression is generated from shared regulatory elements in the promoter region of its host gene, Ia-2β. We performed an in silico analysis of putative miR-153-2 TSSs in humans, using a miRStart database search [25]. A putative miR153-2 TSS was identified at 38,909 bp from the start of the miR-153 precursor (ESM Fig. 2). Further experimental evidence, outside the scope of the current study, would be required to determine whether this is a conserved and tissue-specific miR-153 TSS.

Many potential miR-153 targets can be determined by various miRNA target-prediction software programs (data not shown). However, only a few miR-153 targets have been experimentally validated [24, 26‐31]. Although it is generally accepted that miRNA–mRNA co-expression leads to inhibition of target protein levels [15], several reports indicate that this correlation is less apparent in postnatal brain, where miRNA expression can be associated with the presence rather than absence of target protein expression [32, 33]. Furthermore, it has previously been shown that intronic miRNAs can regulate the expression of their target genes, which can play a role in the same functional regulatory pathway as the host gene in which the miRNA is located [34, 35]. Based on these findings and our observation that miR-153-2 and IA-2β expression are co-regulated, we can narrow down the list of candidate miR-153 targets based on its co-regulation with IA-2β expression. Predicted targets of miR-153 and potential protein regulatory networks regulated by miR-153 have previously been identified [24]. Using bioinformatics analysis (see Methods), we set out to refine this list of potential miR-153 targets as defined by correlation with IA-2β expression patterns (brain and pancreas) and IA-2β function (i.e. synaptic vesicle and dense core vesicle [DCV] release). From this analysis we extracted target genes that are highly associated with IA-2β expression and function (ESM Table 1). Next, we used IPA software to determine gene–function relationships (Table 1) and potential interactions (Fig. 3a) between IA-2β, miR-153 and the selected human candidate miR-153 target genes. We also showed that the predicted target genes and miR-153 are expressed in pancreas and brain as determined by quantitative PCR (Fig. 3b,c). Our data support a significant association of miR-153 and its target genes with vesicle exocytosis [24, 36‐38]. Based on these findings, the identified candidates are very likely to be regulated by miR-153.

Based on our data showing that miR-153 expression is significantly reduced in mouse pancreatic islets (Fig. 2c) and to a lesser extent in brains from DKO mice (Fig. 2b), we hypothesised that the expression of predicted target genes would be increased in DKO islets and brains. To test this, we examined the mRNA expression of predicted miR-153 targets in DKO islets and BKO mouse brains (Fig. 4a,b). In islets we found a significant increase in Pclo, Snap25 and Vamp2 expression (Fig. 4a), whereas only Bsn expression was significantly increased in DKO mouse brain (Fig. 4b). Although most other targets followed a trend towards increased expression, they did not show statistically significant changes. Next, based on our data showing that miR-153 expression was induced after glucose stimulation in MIN6B cells (Fig. 2d), we hypothesised that the expression of predicted target genes would be reduced after glucose stimulation if they are directly regulated by miR-153 action. To analyse this possibility, we used quantitative PCR to investigate target gene expression in glucose-stimulated MIN6B cells (Fig. 4c). Here we found that Bsn expression was significantly reduced (Fig. 4c), whereas most other targets, although showing a trend towards reduced expression, did not change significantly. Subsequently, to determine whether overexpression of miR-153 in mouse MIN6B or in human neuroblastoma SHSY5Y cells leads to reduced target gene expression, we transfected these cells either with miR-153 pre-miR precursor or a scrambled control precursor (SCR) (both at 50 nmol/l final concentration) and analysed target gene expression at 48 h after transfection. miR-153 expression was analysed by quantitative PCR analysis to determine the extent of miR-153 overexpression (ESM Fig. 3). We found significantly reduced expression of Bsn and Syt4 in mouse MIN6B cells (Fig. 4d), and significantly reduced BSN, SNAP25, SNCA and SYT4 expression in human SH-SY5Y cells (Fig. 4e). Next, we asked whether miR-153 can regulate endogenous protein levels of these predicted targets in SH-SY5Y cells. We used protein extracts from the same transfected cells used for the quantitative PCR analysis (Fig. 4e). We analysed the abundance of bassoon, SNAP25 and SNCA in SH-SY5Y cells, and showed that miR-153 overabundance downregulated bassoon abundance by 83%, SNAP25 by 43% and SNCA by 78% compared with scrambled control transfected cells (Fig. 4f). Protein levels of parkin and the housekeeping gene β-actin, which do not have any predicted miR153 seed sequences, remained unaffected by miR-153 overabundance.

Finally, to determine whether the predicted targets can be directly regulated by miR-153, we selected SNCA, bassoon and piccolo for further analysis. SNCA expression has been shown to be regulated by miR-153 [24], and both SNCA and IA-2β localise to secretory granules in beta cells, as well as to synaptic vesicles in neurons [13, 39]. We confirmed that SNCA expression is regulated by miR-153 in vitro (Fig. 4e,f). Regulation of bassoon and piccolo levels by miR-153 has not been investigated. The human SNCA, BSN and PCLO 3′ UTRs (wild-type) or 3′ UTR with a mutated miR-153 seed sequences (seed mutant) were cloned after the stop codon of Firefly luciferase under control of the human thymidine kinase promoter in the pGL3 expression construct (ESM Fig. 4) and co-transfected in SH-SY5Y cells with miR-153 or a negative control SCR (pre-miR precursor; 50 nmol/l; Ambion [Ghent, Belgium]). Co-transfection of miR-153 and luciferase constructs produced 35%, 80% and 40% reductions of Firefly luciferase activity in luciferase reporter constructs containing the 3′ UTRs of SNCA, BSN, and PCLO, respectively (Fig. 5). Mutation of the miR-153 seed sequence (seed mutant) abolished the inhibitory effect of miR-153 on luciferase activity in luciferase reporters of SNCA,and BSN. The human PCLO 3′ UTR contains two miR-153-binding sites (ESM Fig. 4). Only mutation of the B seed sequence abolished the inhibitory effect of miR-153 on the PCLO 3′ UTR luciferase reporter. These data show that miR-153 directly interacts with predicted seed sequences in SNCA, BSN and PCLO 3′ UTRs to inhibit translation of the luciferase reporter protein levels.

These data indicate that changes in miR-153 abundance in various settings can result in significant changes of predicted target gene expression.