MicroRNA Cluster 302–367 Enhances Somatic Cell Reprogramming by Accelerating a Mesenchymal-to-Epithelial Transition

Abstract

MicroRNAs (miRNAs) are emerging critical regulators of cell function that frequently reside in clusters throughout the genome. They influence a myriad of cell functions, including the generation of induced pluripotent stem cells, also termed reprogramming. Here, we have successfully delivered entire miRNA clusters into reprogramming fibroblasts using retroviral vectors. This strategy avoids caveats associated with transient transfection of chemically synthesized miRNA mimics. Overexpression of 2 miRNA clusters, 106a–363 and in particular 302–367, allowed potent increases in induced pluripotent stem cell generation efficiency in mouse fibroblasts using 3 exogenous factors (Sox2, Klf4, and Oct4). Pathway analysis highlighted potential relevant effectors, including mesenchymal-to-epithelial transition, cell cycle, and epigenetic regulators. Further study showed that miRNA cluster 302–367 targeted TGFβ receptor 2, promoted increased E-cadherin expression, and accelerated mesenchymal-to-epithelial changes necessary for colony formation. Our work thus provides an interesting alternative for improving reprogramming using miRNAs and adds new evidence for the emerging relationship between pluripotency and the epithelial phenotype.

Introduction

Pluripotent and differentiated cell fates are determined at least in part by tissue-specific transcription factors that impose a concrete genetic program (1). In addition to coding RNAs, noncoding RNAs (2) are an integral part of the genetic programs that specify cell fate, regulating, for example, the expression of key cell-specific transcription factors (3) and chromatin stability (4) and therefore cell-specific properties. miRNAs⁴ are 21–23-nucleotide-long noncoding RNAs that, by inducing degradation and/or preventing translation of target mRNAs (2), modulate a plethora of cell functions, including those related to ESC self-renewal/differentiation (5) and cell cycle progression (6). In this context, it is not only expected that miRNAs can enhance reprogramming but also tempting to speculate that, in the right combination, they might be able to reset somatic cells into iPSCs without added factors. Blelloch and co-workers (7) observed that, in the original mixture devised by Takahashi and Yamanaka (1), c-Myc can be substituted by components of the miR-290 cluster or by miR-302d. Elimination of c-Myc is desirable because it reduces tumor formation but has a negative effect on reprogramming. However, use of chemically synthesized oligonucleotides involves repeated transfection, and this implies a transient effect, toxicity, and an inability to pool large numbers of miRNAs without reducing their concentration beyond an optimal threshold. The latter is a concern because many ESC-specific miRNAs reside clustered within the same genome locus, and one could argue that delivering some or all these miRNA clusters together should be optimal for producing iPSCs. In this regard, we demonstrate herein that stable overexpression of entire endogenous miRNA clusters can potently improve reprogramming and be an effective tool for mechanistic analysis.

EXPERIMENTAL PROCEDURES

RESULTS AND DISCUSSION

We chose 3 miRNA clusters with high relative expression in ESCs compared with somatic cells based on our recent work (9) and insights from other studies (10, 11). miR-200b–429, miR-106a–363, and miR-302–367 are hereafter referred to as clusters A, B, and C, respectively (Fig. 1A). Cluster A contains miR-200b/a and -429, which are associated with maintenance of an epithelial phenotype (12). Cluster B contains miR-106a, -18b, -20b, -19b-2, -92a-2, and -363. Cluster C contains miR-302b/c/a/d and -367. miR-106a and -20b and miR-302 members, as well as constituents of the miR-294 cluster, belong to the so-called ESC cycle-regulating group of miRNAs responsible for modulating the unique characteristics of the ESC cycle (6). miR-106a/b and several miRNA components in cluster C display similarity in the seed region that helps determine target specificity (2) (supplemental Fig. S1). Members of the miR-294 cluster share this similarity as well, suggesting that the 3 clusters may act in tandem to impose changes in the ESC program. First, we verified high expression of each individual endogenous miRNA of the 3 clusters in mouse ESCs compared with fibroblasts by qRT-PCR (supplemental Fig. S2A). We then amplified their genomic loci by PCR and separately cloned them into a retroviral vector commonly used to induce reprogramming (1, 8). We envisaged that a fragment containing not only the miRNA precursor sequence but also tens of flanking bases would result in effective cleavage of the large primary miRNA product into mature miRNAs. Increased expression of each individual component was validated by qRT-PCR after delivery into fibroblasts (supplemental Fig. S3). The -fold increase in miR-106a, -20b, 19b, -92a, -367, and -429 was less remarkable possibly due to higher basal expression in fibroblasts relative to miR-302 components (supplemental Fig. S2B). Next, we overexpressed the 3 miRNA clusters together with Sox2, Klf4, and Oct4, with or without c-Myc, using donor fibroblasts that bear a transgenic oct4 reporter driving GFP expression as an indication of full reprogramming (8, 9). Cluster B and, more remarkably, cluster C significantly enhanced the number of GFP⁺ colonies counted at day 11 post-infection with 4 factors or day 15 with 3 factors (Fig. 1, B and C), whereas cluster A did not have any effect. The improvement by clusters B and C was more evident with 3 factors than with 4, and in both cases, we observed strong synergy with vitamin C, which is known to improve reprogramming (Fig. 1C) (8). The increase in GFP⁺ colonies and the accelerated time course with 3 factors and cluster C is notable as, under normal experimental conditions, only a few colonies emerge as late as day 20. We also noticed that most (if not all) of the colonies produced with 3 factors and cluster B or C had ESC-like morphology and were GFP⁺ rather than representing incompletely reprogrammed cells. GFP⁺ colonies could be readily picked and expanded and were pluripotent using standard procedures, including the formation of chimeric mice with germ-line transmission (Fig. 1, D and E, and supplemental Fig. S4). These miRNAs had integrated into the iPSC genome, but their expression was mostly silenced, as with the exogenous transcription factors (Fig. 1, F and G). Therefore, stable overexpression of endogenous miRNA clusters can greatly improve the reprogramming of somatic cells.

Next, we tried to understand the differential effect of miRNA clusters B and C on 3 versus 4 factors. We hypothesized that the 2-factor combinations may mediate differential activation of endogenous miRNA clusters B and C. qRT-PCR showed indeed that miR-106a, -19b, -20b, and -92a were significantly up-regulated by 4 factors compared with 3 and by individual overexpression of c-Myc (Fig. 2A). In contrast, miR-302b/d were induced by 4 factors but not by c-Myc alone (Fig. 2A). The latter result suggests an indirect effect of the 4 factors or a single factor-mediated effect that needs prior opening of the chromatin through the combined action of all factors. Other components of clusters B and C did not show any noticeable change (data not shown). We then analyzed the relative contribution of each miRNA by overexpressing them individually or in combination, also using retroviral vectors and adding vitamin C to magnify differences. In the context of cluster B, only miR-106a and -20b could improve the 3-factor basal efficiency (Fig. 2B). On the other hand, the combination of miR-302b/c/a/d was almost as potent as intact cluster C, whereas miR-367 alone had no effect (Fig. 2B). The choice for this division is that all miR-302 components are rather similar, and thus, one would expect redundancy. The relative contributions of endogenous miR-106a and -20b in cluster B and the miR-302 family in cluster C were demonstrated by transient transfection of specific miRNA antagonists (antagomirs) (13) in fibroblasts transduced with 3 factors plus miRNA cluster B or C (Fig. 2C); these experiments were performed also with added vitamin C. These results show that c-Myc directly and indirectly up-regulated key miRNAs to overcome the low basal reprogramming efficiency of 3 factors compared with 4.

We also investigated how miRNA clusters B and C enhance reprogramming mechanistically. We performed DNA microarray analysis of a time course experiment in which 3 factors and clusters B and C or a control empty vector were overexpressed, and the analysis yielded a large number of differentially expressed genes (supplemental Table S2). These differentially expressed genes did not group into pathways with an obvious connection to reprogramming (supplemental Table S3). However, among those that changed at an early time point (day 4), especially in cluster C, we found genes involved in three relevant processes: the cell cycle, epigenetic modulation, and EMT-MET (e.g. the transcriptional repressor Zeb1, E-cadherin, and occludin) (supplemental Table S4). At later time points, we observed also an increase in Esrrb, Nanog, and UTF1 (supplemental Table S4), which have been implicated in reprogramming. The effect of clusters B and C on cell cycle-related genes is not surprising, as some of the constituent miRNAs belong to the ESC cycle-regulating group (6). We (9) and others (14) showed recently that the process of MET (driven mainly by Klf4) coupled to down-regulation of EMT-related genes and TGFβ signaling (mediated mainly by c-Myc) is instrumental for the reprogramming of fibroblasts into iPSCs, but the precise molecular mechanism is not well understood yet. We hypothesized that miRNA cluster C enhances reprogramming at least in part through a MET-related mechanism. First, we explored a putative link between cluster C and MET in silico and found that TGFβR2 is indeed a potential target mRNA of miR-302 members as predicted with the program TargetScan (2). TGFβR2 binds TGFβ cytokines and induces EMT through phosphorylation of TGFβR1 and activation of the Smad signaling pathway (9). Using a reporter-based assay, we also detected that miR-302d, as a representative component of cluster C, bound to and repressed the 3′-UTR of TGFβR2 (Fig. 3, A and B). This was alleviated by mutating the miRNA-binding site in the 3′-UTR (Fig. 3B). In addition, overexpression of cluster C down-regulated TGFβR2 protein in fibroblasts (Fig. 3C). TargetScan predicted also that miR-20b and -106a can target TGFβR2, but we did not observe protein down-regulation after overexpressing cluster B in MEFs (Fig. 3C). During the time course of reprogramming with 3 factors, cluster C could increase E-cadherin expression compared with cluster B or the control, as assessed by qRT-PCR and Western blotting (Fig. 3, D and E). Interestingly, cluster A could also increase E-cadherin mRNA and protein (Fig. 3, D and E). Immunofluorescence microscopy for E-cadherin and its cytoplasmic partner β-catenin validated that both clusters C and A induce MET (Fig. 3F). E-cadherin induction by clusters A and C was comparable at early time points, but later on, the increase was more marked with cluster C (Fig. 3, D and F). On the other hand, overexpression of a constitutively active form of TGFβR1 (more potently) or TGFβR2 (9) attenuated the accelerated MET changes observed with miRNA cluster C and 3 factors (data not shown) and reduced the number of GFP⁺ colonies (Fig. 3G). This demonstrates that cluster C enhances reprogramming at least in part by targeting TGFβR2. While our manuscript was in preparation, Li et al. (15) showed that miR-106a/b enhance mouse reprogramming through down-regulation of p21 and TGFβR2. The lack of an obvious effect of cluster B (containing miR-106a) in repressing TGFβR2 in MEFs and in inducing MET during reprogramming in our model may be explained by differences in the delivery method. miR-106a/b and -302 components display similarity in the seed region, but their sequences are not identical (supplemental Fig. S1), which might determine a different affinity for targeting TGFβR2. In this regard, oligonucleotide transfection can likely achieve higher single miRNA expression levels than retroviral delivery of an entire miRNA cluster, thus bypassing potential differences in relative specificity. The lack of effect on colony formation of miRNA cluster A is also puzzling and contrasts with the data of Samavarchi-Tehrani et al. (14). Interestingly, we observed that cluster A seems to improve MET at the start, but the final number of reprogrammed colonies remains unaffected. Samavarchi-Tehrani et al. performed transient transfection, whereas in our work, high expression levels remained constant. In this regard, a recent study by Wellner et al. demonstrated (16) that miR-200 family members target Klf4 and Sox2, and this could explain why miRNA cluster A accelerates the reprogramming up to a certain stage (MET) and then blunts it. Potential targets of cluster A also belong to (Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways not related to cluster C, and this could also derail the process (supplemental Fig. S5 and Table S5). Delivery of cluster A using an inducible system may be a useful way to further study these possibilities. Of note, we also observed increased iPSC clone formation with 3 factors and miR-302–367 in human fibroblasts (supplemental Fig. S6); the effect was not as dramatic as in mouse cells but raises the possibility of further optimization. Lin et al. (11, 17) used miR-302 constituents to reprogram human skin cancer cells and hair follicle cells into clones that display similarity to iPSCs. Nevertheless, in the latter study, it is unclear whether the target cells were fibroblasts or a stem cell population, and the colonies did not display human ESC-like morphology (17).

In summary, our work describes an alternative method to test miRNA combinations in nuclear reprogramming by exogenous factors and gives mechanistic insight into how the process works (supplemental Fig. S7). Currently, we are testing different miRNA combinations that include existing genomic clusters and artificial mixtures to define additional signaling pathways and to devise strategies that will improve transgene-free reprogramming. For example, the same principle applied recently to produce human iPSCs with modified mRNAs (18) may be applied to deliver miRNA clusters like those described here. Our finding that miR-302 represses TGFβR2 may also be relevant in other settings, e.g. cancer stem cells (19).

Supplementary Material

Acknowledgments

We thank Dr. Lingwen Zeng and Wenzhi He for assistance with the DNA microarrays.

REFERENCES

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.