Dual regulation of miR-375 and CREM genes in pancreatic beta cells

ABSTRACT MicroRNA-375 (miR-375) is upregulated in the islets of some diabetics and is correlated with poor outcome. Previous work in our laboratory showed that cyclic adenosine monophosphate (cAMP) reduces miR-375 expression and could provide a way to restore normal miR-375 levels, however the transcription repression mechanism is unknown. Using a chromatin immunoprecipitation assay we show that cAMP response element modulator (CREM) binds to the miR-375 promoter 3-fold above background and we find that CREM represses transcription from the miR-375 promoter 1.8-fold. While investigating miR-375 target genes we discovered that several microRNA:mRNA target prediction algorithms listed human CREM as a target gene of miR-375. The predicted binding site is conserved in primates but not in other species. We found that indeed miR-375 binds to the predicted site on human CREM and represses translation of a green fluorescent protein reporter gene by 30%. These findings suggest a primate-specific double-negative feedback loop, a mechanism that would keep these important β-cell regulators in check.


Introduction
MicroRNA-375 (miR-375) is one of the most abundant miRNAs found in pancreatic islets, and is important for islet cell development and physiology. 1 In zebrafish, a model for vertebrate development, targeted disruption of miR-375 by morpholinos severely disrupts islet development. 2 In particular, miR-375 appears to control the ratio of β-cell to α-cells in the developing islets as miR-375 knockout mice have reduced numbers of βcells and a corresponding increase in α-cells. 3 In βcell line cultures, miR-375 inhibits insulin secretion and β-cell replication, in part by inhibiting the protein synthesis of the mRNAs for myotrophin (Mtpn) 4 and phosphoinositide 3-kinasedependent-kinase (Pdk1). 5 MiR-375 may play a role in the pathogenesis of type 2 diabetes mellitus in humans as well. In a small cohort of patients, Zhao et al. 6 discovered that miR-375 expression in the pancreas was increased approximately 4-fold in diabetic patients compared with non-diabetic control individuals. While additional patient studies need to be done, it suggests that the miR-375 gene can be misregulated in the pathogenic state. Intriguingly, diabetic patients exhibit decreased β-cell mass and increased α-cell mass due in part to dedifferentiation of βcells. 7 In patients, the gain of miR-375 expression correlates with the decrease in β-cell to α-cell ratio. 6 It is therefore at least a possibility that miR-375 contributes to the pathogenicity of diabetes by decreasing insulin secretion in β-cells, 4 and by decreasing insulin levels through the reduction in β-cell numbers. 5,6 We and others 5,[8][9][10][11] have made progress in the study of miR-375 gene regulation in order to determine its role in healthy and diabetic individuals. We initially identified the miR-375 promoter in independent genome-wide screens for binding sites for the transcription factors NeuroD1 and Pdx1. 10 Avnit-Sagi et al. 8 identified a 768 bp region upstream from miR-375 that directs its expression to pancreatic islets, and identified a TATA box and start site of transcription. Interestingly, they identified a 316 bp repression domain between the transcription start site and the miR-375 sequence. Work by El-Ouaamari et al. 5 showed that elevated glucose could repress miR-375 expression, and we demonstrated that cAMP could repress as well. 11 Therefore it is likely that the factor or factors responsible for repression will bind to the repression domain.
Cyclic-AMP-mediated transcriptional repression can occur via the protein CREM. 12 CREM is in the CREB family of bZIP transcription factors and binds to the CRE sequence and is alternatively spliced. Depending upon the inclusion of a glutamine-rich activation domain, some CREM splice variants activate and some repress transcription. 13,14 In addition, usage of an alternate intronic promoter generates an additional repressor called inducible-cAMPelevated repressor (ICER), 15 itself being activated by the cAMP -protein kinase A (PKA) axis. In βcells there have been several repressing CREM and ICER splice variants identified 16,17 which can repress gene transcription by recruiting histone deacetylase 1 to the promoter. 18 We have proposed that one way in which cAMP enhances β-cell function over the long term is through repression of miR-375 through the cAMP -PKA pathway. 11 In this model, cAMP agonists such as exendin-4 might enhance β-cell function in part through keeping miR-375 levels in check. To complete this model, however, it is essential to identify the factor or factors responsible for transcriptional repression of miR-375. Here we identify CREM as a regulator of miR-375 expression. Surprisingly, CREM mRNA itself is a target of miR-375, which suggests the presence of a double-negative feedback loop that keeps the expression of these two important regulators of βcell function in check. Moreover, the CREM -miR-375 interaction is primate-specific and therefore may account for primate-specific aspects of βcell regulation.

INS-1 cells express CREM repressors only
We initially suspected that CREM or ICER may repress the transcription of miR-375 because they are well-documented repressors activated by cAMP, 13 and they control genes involved in β-cell function in the normal and pathogenic states. 16,19 We wanted to perform a chromatin immunoprecipitation (ChIP) assay to test whether CREM was bound to the miR-375 promoter, but currently there are no antibodies that distinguish between the activating and repressing isoforms. We began by identifying the CREM exons that were expressed in rat insulinoma INS-1 832/13 cells (hereafter called INS-1) by using reverse transcription coupled with quantitative real-time PCR (qRT-PCR) (Figure 1(a)). Using exon-specific primers, we found that the majority of CREM exons were expressed in these cells (Figure 1(b)). A notable exception, however, was the critical activation exon G, the second of two glutamine-rich exons, and the key exon responsible for activation. 20 The absence of this exon in CREM mRNA is strong evidence that there are only repressing isoforms present, thus any CREM protein detected by ChIP assay would be a repressing isoform.
Through a combination of PCR and DNA sequencing using intron-spanning primers, we identified several transcript variants of CREM and ICER in INS-1 cells (data not shown). However, using an antibody against the essential DNA binding domain of CREM and ICER, we could only detect two isoforms that were translated into protein ( Figure 2(a)). An 18 kD isoform was induced by forskolin treatment, and matches to the predicted size of either ICERIγ or IIγ (Figure 2(b)). A 20 kD isoform appears to be the most abundant isoform expressed, though its identity is unknown at this time ( Figure 2(a)). Based on its predicted size and our DNA sequencing results (data not shown), it could be the rat ortholog of human CREMΦ2β, expressed from human CREM transcript variant 14. This variant contains the regulatory phosphorylation domains as well as DNA binding domain II (Figure 2(b)). In an attempt to identify the CREM isoforms expressed, we tested two additional antibodies, but neither could immunoprecipitate CREM (data not shown).

CREM binds to the miR-375 promoter
Having observed repressing CREM isoforms expressed in INS-1 cells, we next used the CREM antibody in a ChIP assay to determine if CREM binds to the miR-375 promoter sequence. As shown in Figure 3, CREM binding was enriched at the miR-375 promoter 3.0-fold compared to an anti-Flag antibody control (p = .046). In this assay, the c-Fos promoter was used as a positive control and a sequence 2 kb upstream from the miR-375 promoter was used as a negative control. Additional experiments showed that cAMP stimulation did not alter CREM occupancy on the promoter (data not shown). This is the situation for CREB, which can bind to promoters constitutively, but becomes active only upon cAMP stimulation, 21 and argues against ICER being the repressor.
The proximal miR-375 promoter contains two conserved regions, labeled 1 and 2 in Figure 4(a). Both domains 1 and 2 or just domain 2 were used in luciferase reporter assays to localize the domain of  CREM binding. Figure 4(b) shows that CREM binds to domain 2 and represses transcription 1.8-fold (p = 1.2 x 10 -5 ). A small interfering RNA (siRNA) against CREM relieves this repression, demonstrating the specificity of the reaction (Figure 4(c-d)). HEK 293 T cells were used in transfection assays shown in the figure because they do not express CREM or miR-375, though the same results were observed in INS-1 cells (data not shown).

miR-375 binds specifically to human CREM mRNA
While researching potential targets of miR-375, we discovered that human CREM was a predicted target identified by several independent algorithms, including TargetScan, 22 miRanda, 23 and DIANA-microT-CDS 24,25 ( Figure 5(a)). However, no algorithm predicted rat or mouse CREM to be a miR-375 target gene, as rodent CREM lacks a complementary sequence to miR-375ʹs seed region ( Figure 5(b)). For example, TargetScan identified an exact match between human CREM and nucleotides 2-8 of hsa-miR-375 (7mer-m8 site), but classified the interaction as poorly conserved. 22 To test whether miR-375 might regulate human CREM specifically, we cloned the miR-375 miRNA recognition element (MRE) from human CREM (hCREM) at the 3' end of the green fluorescent protein (GFP) gene. We also cloned the homologous sequence from rat CREM (rCREM), although the miR-375 binding site is not conserved. Our positive control was a perfectly complementary match to miR-375 (anti-sense, AS-mir375) and negative control was a scrambled sequence (SCR-mir375) predicted not to be regulated by any miRNA in βcells. We co-transfected either miR-375 or a C. elegans miRNA control predicted to not target any mammalian mRNA. We used HEK 293 T cells because they lack endogenous miR-375. Results showed that the miRNA control sequence did not bind to any of our GFP reporter genes ( Figure 6(a)), while miR-375 bound specifically to the hCREM MRE and repressed GFP expression by 30% (p = .0138) compared to the rCREM sequence ( Figure 6(b), compare columns 3 and 4). This supports the miRNA target prediction algorithm results that the miR-375:CREM interaction is human-specific.

Discussion
Previous work in our laboratory demonstrated that miR-375 is transcriptionally repressed by cAMP signaling through PKA. 11 Here we build on that previous work by showing that the cAMPregulated transcription repressor CREM binds to the miR-375 promoter ( Figure 3) and represses transcription (Figure 4(b,d)). While no cAMP response element was identified in the miR-375 promoter, there is a conserved AP1 site, 5'-TGAGTCA-3', in domain 2 of the miR-375 promoter which may provide a binding site for CREM. [26][27][28] Using miRNA target prediction algorithms we found that CREM is a target gene of miR-375, but specifically in humans and other primates ( Figure 5) and subsequently showed that miR-375 can bind specifically to the human CREM MRE (Figure 6). Consistent with our results for miR-375, several studies in β-cells have shown that other genes are down-regulated in response to cAMP. 16,19,29 In one study, several genes necessary for insulin secretion were repressed by hyperglycemic conditions in a PKAdependent manner. 19 It was hypothesized that this mechanism contributed to β-cell failure in type 2 diabetes.
Reciprocal regulation by miRNAs and transcriptional repressors is a recurrent theme in mammalian cells. 30 These double-negative feedback loops play a variety of roles in cells, for example by reinforcing cell fate decisions, 31,32 by synchronizing biological oscillators, 33,34 by dampening protein fluctuations, 35 and by increasing transcriptional response times. 36 As shown in Figure 7, our model for a double-negative feedback loop predicts that miR-375 is activated by factors such as Pdx1 and NeuroD1, 10 and then it subsequently represses CREM gene expression. If CREM activity is triggered by cAMP signaling, then miR-375 would be accordingly repressed. This double-negative feedback loop may switch the system between two states, either miR-375 'on' or CREM 'on,' depending on the relative strength of the activating inputs. This may have important consequences for β-cells, as CREM isoforms have been shown to repress genes involved in insulin production and secretion. 17,19 Intriguingly, the miR-375 MRE is present only in the CREM transcript of primates, suggesting a fundamental difference in CREM regulation in primates compared with other species. Due to the sequence conservation of the miR-375 promoter in primates and rodents, we hypothesize that CREM represses miR-375 transcription in a cAMP-dependent fashion in both, but that the feedback loop has evolved only in the primate lineage. This finding supports the model that miRNA tend to be more evolutionarily conserved than their target sequences. 37 Indeed, miR-375 is conserved perfectly in all mammals analyzed, yet the CREM MRE is perfectly conserved only in primates. Thus our study, along with others, 38,39 highlights a limitation of using rodents in β-cell research.

Cell culture
Rat

Statistics
Samples were analyzed by two-tailed, paired Student t-tests. Averages were plotted in Microsoft Excel with error bars representing ± 1 standard deviation.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the National Institute of Diabetes and Digestive and Kidney diseases [2R15DK088281-02A1].