|Home | About | Journals | Submit | Contact Us | Français|
Individual microRNAs (miRNAs) can target hundreds of messenger RNAs forming networks of presumably cooperating genes. To test this presumption, we functionally screened miRNAs and their targets in the context of de-differentiation of mouse fibroblasts to induced pluripotent stem cells (iPSCs). Along with the miR-302/miR-294 family, the miR-181 family arose as a novel enhancer of the initiation phase of reprogramming. Endogenous miR-181 miRNAs were transiently elevated with introduction of Oct4, Sox2, and Klf4 (OSK), and their inhibition diminished iPSC colony formation. We tested the functional contribution of 114 individual targets of the two families, revealing twenty-five genes that normally suppress initiation. Co-inhibition of targets cooperatively promoted both the frequency and kinetics of OSK reprogramming. These data establish two of the largest functionally defined networks of miRNA-mRNA interactions, elucidating novel relationships among genes that act together to suppress early stages of reprogramming.
MicroRNAs (miRNAs) are endogenous non-coding RNAs that regulate the translation of target genes. Together with Argonaute proteins, miRNAs form the RNA induced silencing complex (RISC) which suppresses messenger RNAs (mRNAs) by both inhibiting translation and accelerating degradation1. Targeting is largely determined by complementation of a 6-8 nucleotide seed sequence with the 3′ untranslated region (UTR) of the mRNA2. Single miRNAs can down-regulate hundreds of transcripts simultaneously3–7. Interestingly, the functional significance of this extensive parallel co-inhibition of gene networks, while the subject of much speculation, remains largely unexplored experimentally. Functional studies generally focus on the regulation a small number of target genes known to be involved in the biological process of interest8. Frequently, knockdown of these individual targets recapitulates the phenotype of over-expressing the miRNA itself, leading to the “dominant target” hypothesis. However, as a miRNA’s many mRNA targets were presumably evolutionarily selected to be co-regulated, a systematic dissection of these targets is likely to uncover a network of genes that function together rather than alone.
The evolutionary history of miRNAs suggests they play major roles in promoting specific cell fates in complex organisms9,10. We set out to functionally characterize the miRNA-mRNA interactions that promote the pluripotent cell fate using the assay of directed de-differentiation, also called reprogramming. During reprogramming, somatic cells are de-differentiated into induced pluripotent stem cells (iPSCs) via over-expression of defined transcription factors11. Reprogramming consists of two phases: initiation and maturation12,13. We chose to map functional miRNA-mRNA interactions in this system for three reasons. First, although several studies have dissected various aspects of the maturation phase, little is known about the early initiation phase where most reprogramming events are aborted12–14. Second, at least one family of miRNAs the embryonic stem cell enriched cell-cycle regulating (ESCC)-miRNAs, including miR-302 and miR-294 - potently regulates this transition15–23. Indeed, ESCC-miRNAs alone, or in combination with other miRNAs, have been shown to drive reprogramming in the absence of other reprogramming factors18,19,23. Thus, miRNA-mRNA interactions during reprogramming should offer insight into the mechanisms of this transition. Finally, we hypothesized that through mapping functional miRNA-mRNA interactions, we would identify networks of cooperating genes that could be manipulated with combinations of small molecules to enhance this transition.
We screened 570 chemically synthesized mature mouse miRNAs (mimics) for their ability to promote Oct4 (Pou5f1), Sox2 & Klf4 (OSK)-induced reprogramming of mouse embryonic fibroblasts (MEFs). Although many combinations of reprogramming factors now exist, OSK-reprogramming offered two distinct advantages. First, OSK is the most frequently reported core set of required reprogramming factors, with cMyc being both dispensable and possessing transformative properties24,25. Second, OSK reprograms with consistent but low efficiency, resulting in a sensitive assay for identification of barriers to this transition. Indeed, miR-302’s reprogramming enhancing properties were first discovered using this strategy15. We transfected individual wells of OSK-infected MEFs possessing an Oct4-GFP transgene with mimic on days 1 and 7 post-infection26 (Fig. 1a). The mimics functioned for 6 days post-transfection as determined by a GFP-based miRNA-activity reporter (Supplementary Fig. 1). Therefore, serially transfected mimics should function throughout reprogramming. We compared the number of day 16 Oct4-GFP+ colonies in each mimic-containing well to 16 mock-transfected wells using strictly standardized mean difference (SSMD), a statistical parameter measuring both magnitude and confidence of effect size27. Sixteen mimics enhanced the frequency of Oct4-GFP+ colony formation in biological duplicate screens (Fig. 1b and Supplementary Table 1a). OSK-mimic induced colonies were morphologically similar to mouse embryonic stem cells (mESCs) and expressed comparable levels of endogenous Oct4, Sox2, Klf4, Rex1, SSEA1 and NANOG (Supplementary Fig. 2a–b). Oct4-GFP+ colonies also silenced the exogenous retroviruses, indicating an advanced stage of reprogramming (Supplementary Fig. 2c).
Several of the miRNAs mimics that enhanced reprogramming shared a common seed sequence (Supplementary Table 1b). The most represented was the ESCC-miRNA seed sequence, validating the sensitivity of the screen (Fig. 1c). Indeed, even miR-467d, which contains a slightly diverged ESCC-miRNA seed sequence, enhanced OSK-reprogramming, consistent with previous reports that shifted or degenerate ESCC-miRNA seed sequences enhance reprogramming22,28. The second most enriched seed sequence was from the miR-181 family, not previously associated with reprogramming. Validation experiments confirmed the ability of the miR-181 family to enhance iPSC colony formation (Fig. 1c). OSK with miR-181 generated fully reprogrammed iPSCs with normal karyotypes, which contributed to all germ layers when injected into E3.5 blastocysts, including the germ line (Supplementary Fig. 2d–f). This screen confirmed the role of the ESCC-family of miRNAs as potent enhancers of reprogramming, and unveiled a similar ability for the miR-181 family.
The ESCC-miRNAs are expressed in pluripotent stem cells29,30. Similar to other pluripotency factors, such as Oct4 or Sox2, their ectopic over-expression can drive reprogramming, although endogenous activation occurs late in the transition14,15,23,31. In contrast, neither MEFs nor pluripotent stem cells express high levels of miR-18130. Further, unlike the ESCC-miRNAs, which block mESC differentiation, expression of miR-181 destabilizes mESCs32,33. Interestingly, the promoter of the miR-181c&d locus is bound by OCT4, SOX2 and NANOG31. Previously reported miRNA profiling suggested dynamic regulation of the miR-181 family during OSK+cMyc-reprogramming. In one dataset, miR-181c and miR-181d are activated, but miR-181a and miR-181b suppressed, as MEFs transitioned to iPSCs31. In a second report, miR-181a was activated then subsequently silenced in iPSCs14. We measured miR-181 family expression during the course of OSK-reprogramming. Reverse transcription quantitative polymerase chain reaction (RT-qPCR) demonstrated an early induction of these miRNAs by OSK, which persisted throughout reprogramming, followed by silencing in iPSCs34 (Fig. 2a). To determine the robustness and timing of endogenous miRNA activity, we generated GFP-based miRNA activity reporters for the miR-181 and ESCC-miRNA families as well as the let-7 family, which suppress reprogramming32 (Supplementary Fig. 1). Notably, let-7 miRNAs are highly expressed in MEFs and not silenced until late in reprogramming while ESCC-miRNAs are absent in MEFs and not activated until late in reprogramming14,15,30–32. Consistent with the timing of their expression, the ESCC-miR reporter was active (miRNAs low), and the let-7 reporter silenced (miRNAs high), during early reprogramming (Fig. 2b). In contrast, the miR-181-activity reporter was silenced shortly after OSK introduction, reaching maximum suppression as early as 4 days, consistent with OSK-induced miR-181 expression (Fig. 2b). Inhibition of miR-181 with transfected inhibitors on days 1 and 5 post-OSK infection reduced the number of day 16 Oct4-GFP+ colonies by 50-60%, showing that OSK functions, at least partially, through activation of endogenous miR-181 (Fig. 2c).
Previous studies have established at least two distinct phases of OSK+cMyc reprogramming12,13. Completion of the initiation phase is marked by a successful mesenchymal-to-epithelial transition (MET), but is otherwise poorly understood. Cells then enter the maturation phase, which has been characterized as a serial activation of the pluripotency hierarchy of transcription factors13. In OSK-reprogramming, a subpopulation of MEFs entered the maturation phase around day 8, as marked by the down-regulation of Slug and the activation of Cdh1 and Dnmt3l12 (Fig. 3a–c). The early activation and subsequent silencing of miR-181 suggests it functions during initiation. Similarly, the observation that ectopic introduction of ESCC-miRNAs alone can induce reprogramming, indicates that these miRNAs can independently initiate reprogramming23. To further evaluate when within reprogramming these miRNAs have their greatest effect, we conducted a time-course of single transfections. Populations of reprogramming cells are highly heterogeneous, and most cells do not complete initiation13,14. Therefore, transient miRNA mimics transfected on day 1 affect MEFs in initiation, whereas those transfected on day 9 affect a mixed population of cells both in initiation and in maturation (Fig. 3d). Introduction of miR-294 (an ESCC miRNA) and miR-181 at day 1 showed greater ability to enhance colony formation, as compared to introduction at later time points (Fig. 3e), suggesting that both miRNA families largely promote reprogramming during the initiation phase. Two further lines of evidence are consistent with this conclusion. First, we separated day 8 OSK-infected MEFs into initiation (CDH1-) and post-initiation (CDH1+) populations using flow cytometry, and transfected each population with mimic. miR-294, and miR-181 enhanced the number of day 20 Oct4-GFP+ colonies in the CDH1-, but not the CDH1+ populations (Fig. 3f). Second, we profiled gene expression on day 3 post-OSK-infection, 48 hours after the addition of miR-294 or miR-18135,36. Of the 3411 genes expressed significantly higher in iPSCs compared to MEFs, only 230 (6.7%) were up-regulated 3 days after addition of OSK and control mimic (Fig. 3g). Addition of miR-294 and miR-181 increased the number of up-regulated iPSC-specific genes to 15.2% and 8.5%, respectively. Similarly, we found 3754 genes with lower levels of expression in iPSCs as compared to MEFs, of which 372 (9.9%) were down-regulated by OSK and control mimic. Addition of miR-294 and miR-181 increased the set of iPSC-specific down-regulated genes to 15.4% and 10.7%, respectively. Together these data show that ectopic miR-294 and miR-181 promote iPSC production by regulating early reprogramming, and shift the transcriptional profile closer to that of fully-reprogrammed iPSCs as early as day 3, well before activation of the earliest maturation markers. Interestingly, the effects of miR-294 and miR-181 were not synergistic, suggesting that these miRNA families with different seed sequences, functionally converged down-stream (Fig 3h).
We next sought to dissect the mechanisms of reprogramming-enhancing miRNAs by knocking down individual targets. Previous attempts at defining the mechanism of the ESCC-miRNAs focused on a small number of targets selected based on expected roles in reprogramming16,17,21,23. To take an unbiased approach, we generated a database of predicted targets for miR-294 and miR-181, based upon relationships of inverse expression, but not considering knowledge of function. We consolidated genes that were previously verified to be significantly down-regulated on the protein or mRNA level by over-expression of the miRNAs in various cell types3,32,37–41. We then retained genes that contained a predicted miRNA binding site. For down-regulated genes originally identified in human cells, we required this binding site to be conserved between human and mouse and have a high ranking context score (Targetscan, context score <−0.25)42. Finally, we required the genes to be expressed in MEFs, reprogramming MEFs, iPSCs or ESCs32,43. This process resulted in sets of 1079 and 58 genes for miR-294 and miR-181, respectively. Small interfering RNA (siRNA) pools (Dharmacon) were designed against all of the miR-181 targets, the 5% most down-regulated miR-294 targets (56 genes), and 54 random genes, with no over-lapping genes (Supplementary Table 2). We transfected MEFs one day after OSK-infection. At day 16 post-infection, 10 of the 56 miR-294 targets and 12 of the 58 miR-181 targets demonstrated significant increases in Oct4-GFP+ colony formation relative to four independent non-targeting siRNA control pools (p-value <0.01 and SSMD >2 over three independent experiments) (Fig. 4a and Supplementary Fig. 3). In contrast, only 3 of the random pools of siRNAs demonstrated similar effects. We verified siRNA knockdown by RT-qPCR (Fig. 4b). To rule out off-target effects of siRNAs, we tested independent pools targeting distinct gene regions. The independent pools showed highly effective knockdown of corresponding genes (Fig. 4b). Of the ten miR-294 targets identified in the original screen, eight (Cdkn1a, Zfp148, Hivep2, Ddhd1, Dpysl2, Pten, Cfl2 and 9530068E07Rik) confirmed using the independent siRNA pools (Fig. 4c). Similarly, eight of the original twelve miR-181 targets (Bptf, Lin7c, Cpsf6, Nr2c2, Bclaf1, Nol8, Igfbp2, and Marcks) verified. In contrast, knockdown of only one of the randomly selected genes consistently enhanced reprogramming, revealing a strong enrichment for genes that influence reprogramming among predicted targets of miR-294 and miR-181.
Cdkn1a is an established miR-294 target44. To determine whether the other identified genes were direct targets, we cloned gene-specific 3′UTRs containing miRNA binding sites into luciferase reporter constructs (Supplementary Fig. 4). We also generated constructs containing mutated binding sites, and assayed both reporters for mimic-induced luciferase repression (Fig. 4d). With the exception of Lin7c, the expected miRNAs inhibited translation of every wildtype, but not mutant, construct, indicative of direct miRNA targeting. We confirmed the suppression of the targets by the miRNAs in the context of OSK-reprogramming by RT-qPCR and Western blots. For most targets, RT-qPCR showed decreased mRNA levels on day 3 of reprogramming, 48 hours post transfection of the miRNA (Fig. 4e). Of the remaining five genes, antibodies to DPYSL2 and PTEN were available and showed diminished protein levels following miRNA introduction by Western blot (Fig. 4f). Further, inhibition of miR-181 during reprogramming caused a reciprocal upregulation of Nr2c2 and Marcks by day 3 post-infection, demonstrating that these genes are targeted by OSK-activated endogenous miR-181 as well (Supplementary Fig. 5). Similar inhibition of the ESCC-miRNAs did not upregulate expression of targets, consistent with the lack of endogenous ESCC-miRNA activity during this time window (Supplementary Fig. 5). Together, these experiments identified seventeen miRNA-regulated genes that are barriers to reprogramming.
During the course of these experiments, we noted a consistent difference between miR-294 and miR-181-enhanced OSK-reprogramming. Whereas the day 16 Oct4-GFP+ colonies in miR-181 conditions were generally the same size as with control mimic, the miR-294 conditions yielded significantly larger colonies (Fig. 5a). This divergence of phenotype was also observed with siRNAs against the miRNA targets; that is, miR-294-targeted genes increased both area and number of colonies while miR-181-targeted genes generally increased number (Fig. 5b). Further analysis of the screen data revealed siRNAs against six additional miR-294 targets and three additional miR-181 targets that increased only colony area, but not number (Fig. 5b & Fig. 4d). The increase in colony number is consistent with an increase in the number of successful initiation events. In contrast, we hypothesized that colony size could reflect either the kinetics of reprogramming or the proliferation rate of reprogramming cells. To test these two possibilities, we transfected both established iPSCs lines and OSK-reprogramming MEFs with miR-294 or miR-181, and measured colony growth rate. Neither mimic significantly altered partial or established iPSC colony growth rate (Fig. 5 c&d). In contrast, in both contexts, colony size was highly correlated with onset of colony appearance, supporting an interpretation of colony area as a surrogate measurement for the kinetics of reprogramming (Fig. 5e&f). These data demonstrate that the frequency and rate of reprogramming initiation events are separable processes that can be independently altered by these miRNAs and their targets.
We next asked whether multiple miRNA-mRNA functional interactions could cooperate to further influence colony number and/or area during reprogramming. We screened siRNAs against targets of individual miRNA for cooperative functionality by co-transfection of all pair-wise combinations on day 1 of OSK-reprogramming (Fig. 6a). For both families, between 16 and 40% of potential relationships were cooperative between two co-targeted siRNAs, but not between targeted siRNA and control siRNA (Fig. 6a&b). In contrast, we detected very few disruptive relationships. Together, these data show that miR-294 and miR-181 act to enhance reprogramming through a network of cooperating miRNA-mRNA interactions.
As miR-294 and miR-181 did not show significant cooperation with each other (Fig. 3h), and shared no identified overlapping targets, we reasoned that two miRNA families might functionally converge on common signaling pathways or cellular processes. As our lists of functional target genes were too small to conduct pathway/cellular process enrichment analyses, we included high scoring computationally predicted targets (Targetscan, context score <−0.25)42. Among the top signaling pathways and processes predicted to be targeted by both miRNAs were Cadherin, Wnt, p53 and TGF-Beta signaling, as well as apoptosis and cell cycle regulation, each of which have been demonstrated to regulate reprogramming45–48 (Supplementary Fig. 6a&b and Fig. 7a). We further identified additional pathways and processes, including several that influence Akt signaling. To evaluate whether the miRNAs regulate the predicted downstream pathways in this biological context, we tested the influence of miR-294 or miR-181 on Akt, Wnt and TGF-Beta signaling during early reprogramming. MiR-294, but not miR-181, increased the ratio of IGF-activated phospho-AKT to total AKT on day 3 of reprogramming (Fig. 7b). Both miRNAs activated Wnt signaling during reprogramming as measured by TopFlash activity and nuclear localization of B-catenin49 (Fig. 7c–d). Similarly, both miRNAs regulated TGF-Beta signaling as measured by decreased endogenous phosphorylated-SMAD2 during OSK reprogramming (Fig. 7e). These data demonstrate that during reprogramming initiation the ESCC and miR-181 families converge on TGF-Beta signaling inhibition and Wnt signaling activation, and the ESCC-miRNAs additionally activate Akt-signaling.
The above data suggests that alternative means of manipulating the miRNA targeted genes or pathways, specifically during reprogramming initiation, could increase the overall efficiency of iPSC production. Prkaa1, Ddhd1, Cfl2, Pfn2, and Erap1 are interesting miR-294 targets as they demonstrate that directed manipulation of the metabolic circuit, cytoskeleton, and endoplasmic reticulum can actively enhance reprogramming. Therefore, we supplemented early OSK-reprogramming with small molecule inhibitors to Prkaa1 (Compound C) and Erap1 (Bestatin), and found that both also enhanced production of IPSC colonies (Supplementary Fig. 6c). We next focused on the identified signaling pathways. To manipulate Akt signaling, we expressed a tamoxifen-inducible active AKT (M+Akt:ER), or an inactive mutant (M−Akt:ER)50. Strikingly, the active AKT enhanced colony formation, specifically when tamoxifen was administered during reprogramming initiation (Supplementary Fig. 6d). These data corroborate our observation that siRNA against Pten, which inhibits Akt activity (Fig. 7b) also enhances reprogramming (Fig. 4c), consistent with two recent reports51,52. Likewise, recombinant WNT3A and a small molecule TGFBRI inhibitor (Tgfbr Inh), both known enhancers of reprogramming, functioned during the initiation phase (Supplementary Fig. 6e&f). Interestingly, whereas Akt and Wnt activation both exclusively functioned during the initiation phase, TGF-Beta inhibition functioned equally at both time-points. To test combinatorial effects of the pathways, we added activated M+Akt:ER, WNT3A, and Tgfbr Inh on days 2–8 of OSK-reprogramming. Increased Wnt and Akt signaling together did not further enhance colony formation suggesting redundant or converging roles of these pathways (Fig. 7f). Conversely, TGF-Beta signaling inhibition cooperated with both activated Wnt and Akt signaling (Fig. 7f). These data show that although miR-294 and miR-181 have independent targets that enhance reprogramming initiation, but converge on a subset of signaling pathways.
Together, this study identifies two miRNA families, 25 miRNA-mRNA interactions, three miRNA coordinated pathways and two small molecules that regulate the initiation phase of reprogramming, and can be used to manipulate distinct processes during this transition. Ectopic introduction of the ESCC family is a well-established enhancer of reprogramming, although the endogenous loci expressing ESCC-miRNAs are only activated late in the transition15–23. Here we uncover miR-181 as a novel enhancer to reprogramming and show that in contrast to the ESCC miRNAS, this family is activated shortly after the introduction of OSK and is not highly expressed in the final iPSC state. This transient expression is important for the reprogramming process, as knockdown of endogenous miR-181 suppressed iPSC formation. Endogenous miR-181 functions in part through the suppression of Nr2c2 and Marcks as the transcript levels of these targets were elevated following miR-181 knockdown and siRNAs to these targets enhanced iPSC formation. Regardless of the timing of endogenous OSK-induced expression, the ectopic introduction of both families suppressed many targets and had the greatest effect when added early in reprogramming. Overall, we find that ectopically introduced miRNAs remove multiple barriers that inhibit the initiation phase of OSK-reprogramming. It should be noted that although extensive, our methods were not comprehensive, and other functional targets of ectopic or endogenous ESCC or miR-181 family miRNAs during reprogramming likely remain to be uncovered.
The multiple functional targets we uncover as barriers can be grouped into various cellular processes. Among these processes, cell cycle/senescence and apoptosis are previously identified barriers20,53. Here, we also identify cellular metabolism, membrane trafficking, and actin dynamics as additional barriers. Previous studies identified a shift in AMPK-regulated metabolic state between the starting fibroblast population and the final iPSC state, that, if blocked, inhibited the transition54. However, it was unclear whether inducing this shift would further aid in accelerating this transition. Our data, both using siRNAs and a small molecule inhibitor strongly support this conclusion. Determining how membrane trafficking influences the transition will be interesting endeavors for future studies. It has been shown that regulated membrane trafficking of collagen IV plays a critical role in maintaining the embryonic stem cell state without influencing the fibroblast state55 We propose this example is just the tip of the iceberg in terms of the interconnection between membrane trafficking and the switch in cell state. Similarly, actin and cytoskeletal dynamics in regulating cell state is likely to be an important and expansive area of research. For example, our finding that Cfl2 and Pfn2 are barriers to de-differentiation suggests that inhibition of monomeric actin or the promotion of filamentous actin plays a critical role in the fibroblast to iPSC transition.
Our results show that many, if not most, of the barriers to de-differentiation are during the initiation phase of reprogramming. Our analysis following iPSC colonies over time enabled the measurement of two distinct types of barriers to reprogramming: one influencing the number of successful initiation events and the other the rate at which they occur. This separated the phenotypic consequences of target knockdown into three classes. One class predominantly reduces the total number of successful reprogramming events, while having little effect on the size of colonies, likely reflecting no overall change in the kinetics of the assay. Among the genes found in this set are Cfl2, Bptf, Lin7c, Cpsf6, Nr2c2, Bclaf1, Nol8, Igf2bp2, and Marcks. Another set of genes suppresses the kinetics of colony formation while having a much smaller effect on number. This set includes Pfn2, Erap1, Ankrd52, Prkaa1, Lats2, Zbtb41, Foxk1, Metap1, and Atm. Finally, there were a set of genes affecting both frequency and kinetics including Cdkn1a, Zfp148, Hivep2, Ddhd1, Dpysl2, Pten, and 9530068E07RIK. The cellular basis for these different outcomes remains to be determined.
Importantly, our data show that at least in the context in reprogramming, there is no “dominant” target underlying an ectopically introduced miRNA’s ability to promote cell state transitions. Focusing on a large subset of targets of both the ESCC and miR-181 miRNA mimics, we find roughly twenty percent can in part explain each miRNAs’ mechanism. Most published studies have focused on individual targets, often suggesting a dominant target underlies the effect of the miRNA though it is established that an individual miRNA suppresses many targets simultaneously3–6. The “dominant target” model is based on the observation that knockdown or knockout of an individual target often completely recapitulates a miRNA over-expression phenotype. Indeed, we find many of the individual targets we tested largely recapitulate the capacity of the corresponding miRNAs to enhance reprogramming of fibroblasts to iPSCs. However, we were also able to see cooperative effects when targets were suppressed in pair-wise combinations. Therefore, that knockdown of many targets can have effects close to that of the miRNA likely reflects both a combination of redundant functions among targets as well as experimental artifact. In particular, experimentally induced knockdown is much greater than the suppression caused by the miRNA (typically less than a 50% diminishment of the target protein3,4. While our studies are largely focused on ectopic introduction of miRNAs during an induced transition, endogenous miRNAs are also known to have multiple targets3. Therefore, the combinatorial effect of multiple cooperating targets, rather than a few dominant targets, is unlikely to be unique to in vitro reprogramming, but instead relevant to most instances of miRNA regulation.
Ectopic introduction of miRNAs can have remarkable impacts on cell state transitions such as fibroblast dedifferentiation to iPSCs as well as the transdifferentiation of fibroblasts to neurons or cardiomyocytes56–58. However, genetic deletion of miRNAs largely have no dramatic effects in vivo under homeostatic conditions59. Therefore, it has been proposed that miRNAs are generally not required to establish or maintain cell states, but rather stabilize cell states against random noise and environmental perturbations, through inhibition of stochastic and aberrant gene expression60. Interestingly, reprogramming initiation has been characterized by its highly stochastic gene expression13. Consistent with the robustness model for miRNA function, our expression data show that the ectopically introduced miRNAs function to “focus” this early stochastic expression of genes toward patterns more similar to the iPSC profile, presumably by removing molecular barriers that would otherwise divert reprogramming cells away from the path to pluripotency (Fig. 7g). It will be important to determine if miRNAs will function similarly in other cell state transitions.
MEF generation was conducted as previously described26. In brief, either rosa26-Bgal;Oct4-GFP or Oct4-GFP embryos were harvested on E13.5. Heads and visceral tissue were removed. Remaining tissue was disassociated with trypsin and physical disruption and plated (P0) in MEF media (high glucose (H-21) DMEM, 10%FBS, non-essential amino acids, L-glutamine, Penn/Strep, 55uM beta-mercaptoethanol). MEFs were expanded to P3 and frozen.
HEK293T cells grown to approximately 70% confluence were transfected with pCL-Eco and pMXs- or pWZL-expression plasmids at a ratio of 1:2 following the Fugene 6 manufacture’s protocol. At 24 hours, media was replaced with fresh MEF media. At 48 hours, supernatant was harvested, filtered (0.45uM) and frozen at -80 degrees. Virus preparations were only thawed once before use.
HEK293T cells grown to approximately 70% confluence were transfected with pMDL, pRSV, pVSVG and pSIN-expression plasmids at a ratio of 1:1:1:2 following the Fugene 6 manufacture’s protocol. Cells were left for 48 hours, then harvested as above.
Oct4-GFP MEFs (P5) were plated onto gelatin coated Whatman Clear View or Greiner uClear black-walled 96-well imaging plates at 900 cells / well. The next day, 50ul of each retrovirus-containing supernatant with 4ug/mL polybrene was added. Day 1 post infection, virus was replaced with fresh MEF media. Thereafter, media was changed every other day, with ES+FBS media (15%FBS, non-essential amino acids, L-glutamine, Penn/Strep, 55uM beta-mercaptoethanol and Lif) days 2 to 6 post-infection and ES+KSR media [Knock-out DMEM (Invitrogen), 15% Knock-out Serum Replacement (Invitrogen), non-essential amino acids, L-glutamine, Pen/Strep, 55uM beta-mercaptoethanol and Lif] days 6 to 16 post-infection. Supplements were added at indicated final concentrations: Tamoxifen (Sigma, 1nM), recombinant Wnt3a (R&D Biosystems, 50ng/mL), E-616452 (BioVision, TgfbR inhibitor, “RepSox”, 1uM), Compound C (Sigma, AMPK inhibitor, 400pg/mL), Bestatin (Sigma, 250nM). Oct4-GFP expression and colony formation was assessed on days indicated, usually day 16 post-infection. High throughput imaging and high content analysis were conducted with the InCell Analyzer 2000 imaging station and software suit (GE). For screens, colony counts and measurements were automated. Independent experiments are defined as independent MEF lots infected with independent virus preparations. To validate pluripotency, day 16 iPSC colonies were disassociated with trypsin and plated onto irradiated MEF feeder layers (P1) and expanded. Passage 3 colonies were harvested for RT-qPCR and fixed for immunohistochemistry. Passage 5 colonies were injected into blastocysts.
Karyotyping and blastocyst injections to assay for chimeric contribution were performed as previously described15,26. Blastocysts were obtained from E2.5 super-ovulated and fertilized C57BL/6 females (Taconic). Blastocysts were washed in M2 media (Specialty Media) and grown in KSOM media (Specialty Media) for 16h. 16h after blastocyst collection, 10–15 iPS cells were injected into cultured blastocysts, which were then transplanted into the uteri of E2.5 pseudo-pregnant Swiss-Webster females (Taconic). For analysis of tissue contribution, embryos were collected on E13, and stained for B-gal activity. For analysis of germ line contribution, embryos were collected on E13 and gonads were isolated and imaged under fluorescence. 80% of implanted blastocysts demonstrated high-grade chimeric contribution of iPS lines.
MicroRNA mimics (MIRIDIAN), siRNA pools (On-TargetPlus and siGenome), and ESCC family inhibitors (MIRIDIAN Hairpin Inhibitors) were generous gifts from Dharmacon. Transfections followed the Dharmafect manufacturer’s protocol. DMEM containing 1uM RNA and DMEM containing 6:1000 (v/v) Dharmafect 1 were pre-incubated at room temperture for 5min, then mixed 1:1. After 20min of room temperature incubation, transfection mixture was added to fresh media on cells for a final RNA concentration of 100nM. For miRNA family inhibition experiments, where available, full-family LNA were used (Exiqon, miRCURY LNA, miR-181). Otherwise, cocktails of equimolar individual miRNA inhibitors were used (Dharmacon, MIRIDIAN Hairpin Inhibitors, ESCCs). The ESCC inhibitor cocktail included inhibitors of miR-302a-d, miR-291-3p, miR-294 and miR-295.
On day 10 of reprogramming (see above), OSK-infected MEFs were treated with 5mg/mL collagenase type I (Gibco) for two consecutive 10minute incubations at 37 degrees, and then scraped. Digestion was quenched and cells washed in PBS containing 2% FBS. Cells were resuspended in PBS+2% FBS containing primary Cdh1 antibody (1:50 of 2mg/mL stock, E-cadherin monoclonal ECCD-2, Invitrogen #131900) at a maximum of 5 million cells / mL and incubated on ice for 30minutes. After washing in PBS+2%FBS, cells were treated with secondary antibody (1:200, Alexafluor 633, Invitrogen) for another 30minutes, washed again, and resuspended in 300uL PBS containing SytoxBlue (Invitrogen). Cells were sorted as live singlets into Cdh1+ and Cdh1- populations into ESC media containing 50% FBS. Of these, cells were plated at 1000 cells / well onto irradiated MEF feeder layers in standard ES+FBS media (see above). Cells were transfected the next day, then switched to ES+KSR media (see above) the next.
Total RNA was collected using either Trizol (manufacture’s protocol) or RNeasy spin columns (Qiagen, manufacture’s protocol). For mRNA amplification, RNA (1-5ug) was treated with DNAse I (Invitrogen) and reverse transcribed using the Superscriptase III kit (Invitrogen, manufacture’s protocol) with polyT primers. Total cDNA was diluted 1:5 and 1uL per reaction was amplified using gene specific primer sets (500nM) and Power SYBR Green PCR master mix (ABI). Endogenous and exogenous Oct4, Sox2, and Klf4 primers were previously described15. New primer sets are listed in Supplementary Table 3. Specificity of all primer sets was verified through analysis of disassociation curves in experimental, no RT, and water only samples. For miRNAs, qRT-qPCR was performed by polyadenylating the miRNAs and using a modified polyT RT primer as previously described34.
Cells were fixed for 15 minutes in 4% PFA, washed in PBT (PBS + 0.1% Triton x-100), incubated for one hour at room temperature with blocking buffer (PBT+1% goat serum+2% BSA), then incubated overnight at 4 degrees in primary antibody in blocking buffer as follows: Nanog 1:50 (Abcam ab21603), SSEA1 1:100 (Univ of Iowa MC-480), Ecad 1:120 (BD Transduction Laboratories 610181), beta-Catenin 1:100 (Cell Signaling 9587). For Nanog antibody, cells were also fixed with methanol at -20 degrees C for 5 min, prior to block. Cells were then washed in PBT, incubated for one hour at room temperature in secondary antibody in blocking buffer (Alexa Fluor 1:1000 Invitrogen), washed in PBT with Hoechst 33342 1:10000 (Invitrogen), and stored in PBS before imaging.
For small scale experiments performed in three or more independent experiments p-values were calculated using a student’s t-Test.
For large-scale siRNA screens, strictly standardized mean difference (SSMD) was calculated to compare single experimental wells to either i) sets of four matched scrambled siRNA transfected wells (Fig 4a and Fig 5), ii) sets of sixteen matched mock transfection wells (Fig 1b) or iii) pair-wise sets of individual siRNA with control siRNA (Fig 6) as described previously27.
For microarrays, total RNA from three experiments was analyzed using MouseRef-8 v2.0 Expression BeadChips through the UCLA Neuroscience Genomics Core. Data were quantile normalized using BeadArray, and statistically significant changes in gene expression between sets (p<0.05) were determined using Limma35,36.
Lists of genes significantly down-regulated by either miR-294 or miR-181 were obtained from previous publications. Specifically, for miR-294, microarrays were used to measure mRNA down-regulation upon addition of miR-294 to DGCR8−/− mESCs32. For miR-181, SILAC analysis was used to measure protein down-regulation upon addition of miR-181 to HeLa cells3. In both cases, authors’ cut-offs for significant down regulation were used. To these lists, known miR-294 family or miR-181 family targets were added. Genes were then required to have miR-294 or miR-181 binding sites in mouse, and to be expressed during the course of MEF to iPSC reprogramming43.
All experiments were performed using the Dual-Luciferase Reporter Assay System (Promega) on a dual-injecting SpectraMax L (Molecular Devices) luminometer according to the manufacturer’s protocol. Ratios of Renilla luciferase readings to firefly luciferase readings were averaged for each experiment. Replicates performed on separate days were mean centered with the readings from the individual days.
B-catenin reporter assay: Topflash reporter plasmid was obtained from Addgene (plasmid 12456)49. Mouse embryonic fibroblasts were cultured in Oct4, Sox2 and Klf4 reprogramming conditions described above. 24h post retroviral infection, cells were transfected with miRIDIAN miRNA mimics (Dharmacon) using Dharmafect1 (Dharmacon) as described above. 72h post retroviral infection, cells were transfected with TOPFlash reporter plasmid (final concentration 1ng/μl) and TK-renilla transfection control plasmid (Promega) (final concentration 0.33ng/μl) using Promega Fugene6 transfection reagent according to manufacturer’s protocol. Recombinant murine Wnt3a (R&D biosystems) was added to the transfection mix at a final concentration of 25ng/ml in ESC media. The cells were lysed 24h after TOPFlash transfection/Wnt3a stimulation, and the luciferase assay was performed.
Target verification reporter assay: 3′UTRs of indicated genes were amplified from the mouse genomic DNA cells using the Zero Blunt TOPO (Invitrogen) vector and subcloned into psiCHECK -2 vector (Promega) using the Cold Fusion Cloning Kit (System Biosciences). 3′UTR seed sequences were mutated using the Quickchange Lightning kit (Agilent). For transfection, 8,000 miRNA-deficient Dgcr8 / mouse ESCs were plated in ESC media onto a 96-well plate pretreated with 0.2% gelatin. The subsequent day, the cells were transfected with miRIDIAN miRNA mimics (Dharmacon) using Dharmafect1 (Dharmacon) at the manufacturer’s recommended concentration of 100 nM. Simultaneously, 200 ng of the psiCHECK-2 construct was transfected into the ESCs using Fugene6 (Roche) transfection reagent according to the manufacturer’s protocol. Transfection of each construct was performed in triplicate in each assay. The cells were lysed 24h after transfection, and the luciferase assay was performed.
MEFs were cultured in Oct4 Sox2 Klf4 reprogramming conditions as described above. 24h post retroviral infection, cells were transfected with miRIDIAN miRNA mimics (Dharmacon) with Dharmafect1 (Dharmacon) as described above. 72h post infection, cells were either serum starved (high glucose (H-21) DMEM, 0.5% FBS, non-essential amino acids, L-glutamine, Penn/Strep, 55uM beta-mercaptoethanol) or media was changed to regular ESC media. For some assays, 16hrs after serum starvation / media change, serum starved cells were stimulated with IGF1 protein (Abcam) for five minutes at a concentration of 6nM in serum starvation media. Lysates were collected in lysis buffer (25 mM Tris-HCl, pH 7.9, 150 mM NaCl, 0.1% Nonidet P-40, 0.1 mM EDTA, 10% Glycerol, 1mM DTT) containing 1× protease inhibitor cocktail (Roche) and 1xPhosSTOP Phosphatase Inhibitor Cocktail (Roche). Lysates were incubated at 4°C for 10 min rocking then collected by scraping. After three snap freeze-thaw cycles, lysate was spun at 4°C and approximately 20,000g in a table-top centrifuge. Protein was quantified using a Bio-Rad protein assay (Bio-Rad). Five micrograms of protein was resolved on a 10% SDS PAGE gel. Proteins were transferred to Immobilon-FL (Millipore) and processed for immunodetection. Blots were scanned on a Licor Odyssey Scanner (Licor). Antibodies were diluted as follows: GAPDH 1:5,000 (Santa Cruz, sc-25778), Phospho-Akt (Ser473) 1:2000 (Cell Signaling, #4060), Phospho-Akt (Thr308) 1:1000 (Cell Signaling, #2965), Akt (pan) 1:1000 (Cell Signaling, #2920), PTEN 1:2000 (Cell Signaling, #9552), Dpysl2/Crmp2 1:1000 (Cell Signaling, #9393) Phospho-Smad2 (Ser465/467) 1:1000 (Cell Signaling, #3108), Smad2 1:1000 (Cell Signaling, #3103). Secondary infrared-dye antibodies from Licor were used at 1:25,000. Images were quantified using Odyssey Software. Original images of blots used in this study can be found in Supplementary Figure 7.
The miR-302 sponge consists of complementary sequences to mature miR-302b miRNA with mismatches corresponding to basepairs 9–12 of the mature miRNA. miR-302b sponge sequence corresponding to basepairs 9–11 of the mature miRNA sequence were designed to be identical and a basepair corresponding to 12 was removed from the sponge. The intentional mismatches and deleted basepair in the sponge sequence were designed to induce a bulge in the basepairing between the mature miRNA and the sponge sequence to prevent endonucleolytic cleavage such as those occurring from exact basepairing siRNAs. The sponge sequence is CTACTAAAACACCTAGCACTTA. This sequence was repeated seven times with random 8 bp sequences between each repeated sponge site. The 7X miR-302b sponge fragment was cloned downstream of GFP in the pSIN construct using MluI and NsiI restriction sites.
NIH 3T3 fibroblasts were infected with GFP-302-sponge-puro lentivirus supernatant with 4ug/mL polybrene. After 24h, media was replaced by MEF media. After 48h, cells were split to 40% confluency and puromycin (1μg/ml) was added to this and subsequent media changes. After 10 days, foci of puromycin resistant fibroblast colonies became visible. Cells were grown to high confluency and frozen for subsequent experiments. GFP-302-sponge stably expressing fibroblasts were plated at a confluency of 300,000 cells per 6-well dish in MEF media and puromycin (1μg/ml). The subsequent day, the cells were transfected with miRIDIAN miRNA mimics (Dharmacon) with Dharmafect1 (Dharmacon) at the manufacturer’s recommended concentration of 100 nM. For 10 days following transfection, GFP expression was assessed using FITC-Intensity measurement by flow cytometry (LSRII BD) and fluorescence microscopy. Cells were kept at constant confluency by 1:3 split every 24h.
All animal experiments described in this article were approved by the Institutional Animal Care and Use Committee of the University of California San Francisco.
We would like to thank C. Belair, M. Cook, R. Krishnakumar, M. La Russa, M. Shveygert and other members of the Blelloch lab for critical reading of the manuscript. We would like to thank A. Amiet (Dharmacon Thermo Scientific) for providing miRNA and siRNA libraries and M. McMahon (University of California, San Francisco, San Francisco, California, USA) for AKT expression constructs. We would like to thank H. Zhang for assistance with the high content analysis, J. Paquette, R. Bell, and A. Diaz for advice concerning our statistical methods, A. Shenoy for bioinformatic assistance, and M. Kissner for flow cytometric assistance. This work was supported by funds to R.B from the National Institutes of Health (R01:GM101180), the Leona M. and Harry B. Helmsley Charitable Trust (09PG-T1D002) and the California Institute of Regenerative Medicine (RN2-00906-1). R.L.J. is supported by the National Science Foundation (NSF) graduate research fellowship.
ACCESSION CODES: Pending
Author ContributionsR.L.J. contributed to Figs 1, ,2,2, ,3,3, 4a–e, ,5,5, ,6,6, 7a,f–g and Supplementary Figs 2-6. T.G. contributed to Figs 4d,f, 7b–e and Supplementary Figs 1, 2, 4 and 7. R.J.P. contributed to Figs 2b and Supplementary Figs 1 and 2. R.B. and R.L.J. conceived the experiments, analyzed the data, and wrote the manuscript.