Genome-Wide Identification of Targets and Function of Individual MicroRNAs in Mouse Embryonic Stem Cells

Mouse Embryonic Stem (ES) cells express a unique set of microRNAs (miRNAs), the miR-290-295 cluster. To elucidate the role of these miRNAs and how they integrate into the ES cell regulatory network requires identification of their direct regulatory targets. The difficulty, however, arises from the limited complementarity of metazoan miRNAs to their targets, with the interaction requiring as few as six nucleotides of the miRNA seed sequence. To identify miR-294 targets, we used Dicer1-null ES cells, which lack all endogenous mature miRNAs, and introduced just miR-294 into these ES cells. We then employed two approaches to discover miR-294 targets in mouse ES cells: transcriptome profiling using microarrays and a biochemical approach to isolate mRNA targets associated with the Argonaute2 (Ago2) protein of the RISC (RNA Induced Silencing Complex) effector, followed by RNA–sequencing. In the absence of Dicer1, the RISC complexes are largely devoid of mature miRNAs and should therefore contain only transfected miR-294 and its base-paired targets. Our data suggest that miR-294 may promote pluripotency by regulating a subset of c-Myc target genes and upregulating pluripotency-associated genes such as Lin28.


Introduction
Embryonic stem cells, which are derived from the inner cell mass of the blastocyst, hold great clinical promise because of their unique capacity to both self-renew and differentiate into potentially any cell type. Understanding the molecular controls of pluripotency is key to realising their therapeutic potential. While the general importance of small RNAs in gene regulation has been recognised in plants and animals, much remains to be understood about the specific role of small RNAs in ES cells. miRNAs and small interfering RNAs (siRNAs) are a class of small (<20-25 nucleotide) non-coding RNAs that direct sequence-specific post-transcriptional repression of target mRNAs. Mature miRNAs and siRNAs are generated from double stranded RNA (dsRNA) precursors by the RNase III enzyme Dicer [1,2]. The mature small RNA is then incorporated into a protein of the Argonaute family [3,4]. This RNA-protein complex forms the core of the effector complex referred to as the RNA-induced silencing complex (RISC). Within the RISC, the small RNA acts a guide to direct Argonaute proteins to complementary target transcripts to elicit the cleavage, degradation or translational repression of their targets depending on their degree of complementarity [5].
Several studies implicate miRNAs in the control of early embryonic development and maintenance of the pluripotent stem cell state. Disruption of the single Dicer1 gene in mice leads to early embryonic lethality around E7.5 [6]. Dicer1 mutant embryos have greatly reduced expression of Oct-4 in the epiblast, implying a lack of pluripotent cells, and it is not possible to derive ES cells from Dicer1-null blastocysts. However, conditional deletion of Dicer1 from established ES cells results in an impaired capacity to differentiate, as well as a profound initial proliferation defect that is overcome with time [7,8]. Moreover, large scale cloning and sequencing efforts have revealed a subset of miRNAs that are unique to ES cells [9][10][11]. The miR-290-295 cluster ( Figure 1A) consists of 6 miRNAs that share a similar 59 region from nucleotides 2-8, known as the 'seed' sequence, which is thought to be the primary specificity determinant for target recognition in most miRNAs [12]. The miR-290-295 cluster accounts for the majority of all miRNAs expressed in undifferentiated ES cells but decreases after ES cells differentiate [9]. Recent evidence suggests that there are functional differences between miRNAs from the miR-290-295 cluster. Only miR-291-3p, miR-294 and miR-295 can promote the G1-S transition of the cell cycle and the induction of pluripotency [13,14]. Furthermore, miR-293 expression and seed sequence differs markedly from the other members of this family, indicating the need to re-examine previous inferences based on whole miR-290-295 overexpression studies [15][16][17].
To identify targets and study the function of individual miRNAs in ES cells, we used Dicer1-null ES cells in which we replaced endogenous Ago2 with a transgenic Ago2-myc (Dicer1 D/ D +Ago2-myc) ( Figure 1B). In this Dicer1-null background, there are no mature miRNAs [11], which enables the reintroduction of any single miRNA at a time. To investigate the role of miR-294 in ES cells, we profiled the transcriptome of Dicer1-null ES cells transfected with miR-294. In addition, we used a biochemical approach that exploits the direct interaction of the mature miRNA to its mRNA target within the Ago protein of the RISC complex [18]. Immunoprecipitation of Ago2-myc from Dicer1null ES transfected with miR-294, followed by RNA-sequencing of associated RNAs should lead to identification of direct miR-294 targets. Our strategy overcomes the problem of immunoprecipitating RISC complexes that contain many different miRNAs with their corresponding targets [19][20][21]. This approach is applicable to the study of any single miRNA or combination of miRNAs, potentially in any cell type. We describe here the results of the microarray and RNA-Sequencing analyses, and how miR-294 integrates into the ES cell regulatory network.

Results
Transcriptome analysis of miR-294-transfected Dicer1-null ES cells First, we established Dicer1-null ES cells that lack all mature miRNAs ( Figure S1B, S1C), into which we can reintroduce one specific miRNA of interest at a time. Previous studies have shown that the miR-290-295 cluster might have an important role in the pluripotency of ES cells [14,16,17] amongst which miR-294 is one of the most abundantly expressed miRNAs ( Figure S1A) [11]. We decided to investigate the potential role of miR-294 in pluripotency.
Metazoan miRNAs typically bind to partially complementary sites in the 39 untranslated region (39 UTRs) of their target mRNAs to direct translational repression or mRNA destabilisation [22][23][24]. Overexpression of a miRNA can affect hundreds of mRNAs and messages that are downregulated tend to have significant enrichment of sequences complementary to the corresponding seed of the miRNA [16,17,24]. To identify the global effects of miR-294 in ES cells, we analysed the transcriptomes of Dicer1 D/D +Ago2-myc ES cells transfected with miR-294 compared to Dicer1 D/D +Ago2-myc ES cells transfected with a cel-239b control miRNA, which has minimal sequence identity to mouse miRNAs. The transcriptomes were profiled using Illumina microarrays (GEO accession code: GSE20048). The relative signal intensities of each of the 15735 probes across the six samples were plotted as a heat map ( Figure 1C), which revealed global downregulation and upregulation of probes in each sample transfected with miR-294. To select differentially expressed genes, probes were ranked by the log odds ratio (Bstatistic) of each probe showing differential expression, and plotted against the log fold change in a 'volcano' plot ( Figure 1D). A Bstatistic .5 (and corresponding to an adjusted p-value ,0.0001) was selected as a cut-off for differential expression ( Figure 1C). Using this conservative cut-off gives, with very high confidence, 162 upregulated (Table S1) and 248 downregulated differentially expressed genes upon miR-294 transfection.
To determine whether the downregulated transcripts contain miR-294 seed matches, we used the k-mer composition analysis tool Sylamer [25] to search for overrepresentation of sequence motifs in the 39 UTRs. Each gene on the array (for which a 39 UTR sequence was annotated) was ranked from most downregulated to most upregulated according to log fold change. If enrichment of the seed sequence in the 39 UTRs correlates with the ranking of genes according to their fold change, then one would expect to observe a sharp peak in overrepresentation of the miRNA seed for the top-ranked genes. The resulting enrichment analysis plot shows that a strong signal is evident for 6-mer words corresponding to the seed region of miR-294 ( Figure 1E), peaking approximately at gene 1000 in the ranked list. Conversely, this sequence is depleted in upregulated genes. The maximum of the first enrichment peak can be chosen as a threshold, above which, genes can be considered candidate targets if they contain the appropriate seed sequences matches. This yielded a more relaxed list of 487 predicted targets, which have at least one 6-mer present in their 39 UTR (Table S2). This Sylamer list contains all of the conservative list of predicted targets (127 out of 248 have at least one 6-mer in their 39 UTR) selected using a B-statistic .5 (Table S2). In summary, there was overrepresentation of the miR-294 seed sequence in the most downregulated genes, indicating that many of the observed gene expression changes are likely to be consequences of miR-294 expression.

RNA-IP and RNA-sequencing of Ago2-myc-bound mRNAs
To discover direct targets of miR-294, we used a biochemical approach to isolate mRNAs targeted by miRNAs within the Ago2 protein of the RISC. To facilitate the immunoprecipitation of Ago2, we replaced endogenous Ago2 with a myc-epitope tagged Ago2. We also constructed a catalytically-inactive mutant Ago2myc (Ago2-myc-MUT) by point-mutating residues Q633R and H634A in the PIWI domain to capture the subset of target mRNAs that would otherwise be sliced by Ago2 [26,27]. Point mutations Q633R and H634A were previously shown to abolish mouse Ago2 cleavage-activity without affecting siRNA binding [27]. First, we confirmed the specificity of Ago2-myc immunoprecipitation (IP) ( Figure S2A), and that immunoprecipitation of Ago2-myc from Ago2 flox/flox ;Dicer1 flox/flox ES cells does retain

Author Summary
Stem cells in plants and animals contain many small RNAs, which help to regulate differentiation into diverse cell types. Mutation in a gene necessary for the maturation of small RNAs in plants causes the stem cells (called meristem cells) to remain in an indeterminate, overproliferating state. Similarly in worms, a small RNA called lin-4 miRNA prevents ''stem cell-like cells'' appearing at inappropriate times. Thus, it is important to determine the precise functions of key individual small RNAs in embryonic stem cells. To address this, we created embryonic stem cells lacking all miRNAs into which we introduced a single miRNA. We discovered that a single miRNA could affect the expression of many genes in stem cells, which in turn regulate key properties of stem cells. These together help establish an intricate network of gene regulation in stem cells that defines their properties. Our findings are of broad interest because different miRNAs have critical functions in diverse cell types in developing embryos. It is important to understand the function of these molecules also because misregulation of miRNA function underlies some human diseases, including cancers. miR-294 Targets in Mouse ES Cells miRNAs ( Figure S2B). We then assessed the enrichment of known targets using Dicer1-null ES cells transfected with miR-294. We found enrichment of Cdkn1a, a known miR-294 target [13], in immunoprecipitated RNA from Dicer1 D/D +Ago2-myc ES cells transfected with miR-294, compared to cel-239b ( Figure S2C). Following RNA-immunoprecipitation of miR-294-programmed Ago2-myc, RNA from the INPUT (total RNA) and IP ( Figure  S3A) were subjected to library preparation and sequenced by SOLiD (GEO accession code: GSE20199).
A comparison of the global gene expression changes detected by microarray and by RNA-Sequencing of the INPUT revealed a similar trend ( Figure 2A). Furthermore, there was overrepresentation of the miR-294 seed sequence in the most downregulated genes of the sequenced INPUT ( Figure 2B), indicating that miR-294 was functional and that the transfection efficiency was sufficiently high ( Figure S3B) to bring about gene expression changes. An examination of the IP vs. INPUT ratios for each sample revealed an overall greater dynamic range of enrichment for samples transfected with miR-294 compared to cel-239b ( Figure S4). The pattern was similar for the catalytically-inactive Ago2-myc. This result was expected since miRNA-directed slicing of the target mRNA is thought to occur rarely in metazoans [12]. However, the overall extra enrichment could not be accounted for by the overrepresentation of miR-294 seed sequence matches in the 39 UTRs of enriched genes ( Figure 3A). There is a surprising general tendency for hexamers with a higher GC content to be overrepresented in the 39 UTRs of enriched transcripts ( Figure 3A and 3B). This effect is also observed in samples transfected with the control cel-239b ( Figure 3C). However, the GC effect is markedly stronger in the samples transfected with miR-294.
To ascertain whether this extra IP enrichment contains miR-294 targets, the correlation between our data and computational miR-294 target predictions was tested. The fraction of genes ranked by p-value (based on the likelihood of enrichment) that have a TargetScan prediction was calculated. For samples transfected with miR-294, approximately one-tenth (0.1) are predicted targets by TargetScan for the top 1000-most enriched genes ( Figure 4A). For samples transfected with cel-239b, the cumulative fraction is considerably lower (0.04-0.06) ( Figure 4A). Thus, there is a correspondence between the degree of observed enrichment and the TargetScan computational target predictions. In contrast, there was no correlation with miRanda [28] target predictions ( Figure 4B).
To have a genome-wide benchmark for identifying miR-294 targets, we compared enriched genes in the IP (selected using a p-value ,0.003 cut-off) with our microarray-predicted targets. There is a statistically significant correlation between IP enrichment and the microarray-predicted targets ( Figure 4C and 4D). The trend is similar to the correlation between IP enrichment and TargetScan predictions ( Figure 4A). Thus, miR-294 targets are enriched in the IP but there is also a non-seed-match-specific binding effect that correlates with hexamer GC content.
In summary, a substantial proportion of enriched genes with miR-294 seed matches are not, however, detected in the microarray target predictions. These genes could be direct targets whose transcript levels remain unchanged by miR-294 overexpression and are therefore, not detected by microarray analysis. Alternatively, the basis of this extra enrichment could represent novel targets with non-canonical seed matches, non-specific associations, or technical limitations.

miR-294 may promote pluripotency through regulation of shared c-Myc targets and upregulation of Lin28
To gain an insight into the biological role of miR-294, we performed a Gene Ontology (GO) analysis on the upregulated and downregulated genes (with at least one 6-mer in their 39 UTR) selected from the microarray (B-statistic .5). The top ten terms included enrichment for genes involved in cell cycle (regulation of the G1-S transition), and development and transcription ( Figure 5A). The miR-290-295 cluster has been described as a 'Trojan horse' inside ES cells to bring about differentiation [29]. If this were indeed the case, then one would expect to find an enrichment of differentiation-associated terms in the upregulated genes. This is not the case. Instead, the majority of terms associated with differentiation were more enriched in the downregulated genes. In addition, Lin28 was upregulated by miR-294 transfection into Dicer1 D/D +Ago2-myc ES cells ( Figure 5C). Lin28 is considered to be important for stem cell maintenance by blocking the processing of let-7 [30,31], a critical miRNA involved in differentiation [32]. Furthermore, Lin28, in conjunction with Nanog, Oct-4 and Sox2 can reprogram human fibroblasts into pluripotent cells [33]. No pluripotency genes were detected amongst the downregulated genes. This is consistent with a potential role of miR-294 in the maintenance of the pluripotent state because co-expression of miRNAs that directly target the pluripotency factors would be detrimental to ES cells.
In keeping with cell proliferation as the top-ranking functional GO category ( Figure 5B), miR-294 has previously been reported to be able to substitute for c-Myc, but not Oct-4, Sox2 or Klf4, in the reprogramming of mouse embryonic fibroblasts (MEFs) into induced Pluripotent Stem (iPS) cells [14]. The authors posited that miR-294 acts as a downstream target of c-Myc, which binds to the promoter region of mir-290-295. To test the alternative (and not exclusive) possibility that miR-294 may also substitute for c-Myc by regulating shared targets, a GeneGo (GeneGo Inc) network was generated using the downregulated and upregulated genes (with a B-statistic .5) from the microarray as an input list. The resulting networks revealed enrichment for a subset of targets of the c-Myc network ( Figure 6). c-Myc was a central node with direct regulatory connections to target genes, but it was not itself a target of miR-294. However, many c-Myc target genes were downregulated directly or upregulated indirectly by miR-294. This suggests that miR-294 and c-Myc regulate an overlapping set of target genes, which is consistent with their comparable roles during iPS induction [14]. Thus, from the GeneGo network, miR-294 appears to act synergistically with c-Myc on a subset of targets. However, miR-294 may also repress a subset of targets that are induced by c-Myc. Since the reprogramming process is not identical when miR-294 is substituted for c-Myc [14], this might perhaps explain some of the observed differences between them during the derivation of iPS cells. This network motif, in which c-Myc both activates a target, and inhibits it via miR-294, is described as ''incoherent feedforward'' regulation [34]. Our network also includes a previously hypothesised incoherent feedforward loop involving Oct-4, miR-294 and Lefty1/Lefty2 [35]. The functional significance of the incoherent feedforward loop has not been fully delineated but it can accelerate the response following a stimulus by decreasing the target concentration, fine tune steady state levels and buffer against perturbation [34]. In conclusion, miR-294 may promote pluripotency through regulating a shared subset of c-Myc target genes rather than simply being a downstream effector of c-Myc, and through the upregulation of pluripotency-associated genes, such as Lin28.

Discussion
Our findings show that miR-294 regulates a subset of c-Myc target genes, and upregulates Lin28. We did not observe substantial overlap between Oct-4, Nanog or Sox2 regulated genes and miR-294-regulated transcripts. Thus, the effects on c-Myc and Lin28 are distinct from the other core pluripotency factors, consistent with previous data that proposed functional differences between targets of Myc and that of the other core pluripotency factors [36][37][38]. Furthermore, our conclusions are supported by a recent study [39]. However, we did not detect upregulation of c-Myc mRNA by miR-294 in our microarray results, but rather Mycl1, which belongs to the Myc family of transcription factors. This difference could be a result of the different microarray platforms used for the transcriptome analyses.
Indeed, we identified other networks that were enriched for miR-294-regulated transcripts, which included Sp1, Esr1, Hnf4alpha, p53 and Stat3. The integration of miRNAs into transcriptional networks offers versatility to the regulatory outputs: it enables miRNAs to affect the dynamical properties of transcription factor targets, modulate the strength of transcription factors, and increase the robustness of transcription factor networks. With respect to c-Myc, this model may provide insights on how the initial proliferation defects of Dicer1-null ES cells can be compensated with time [7,8], and indicates a resetting of the transcriptional network in the absence of miRNAs. This is the first study to use Dicer1-null ES cells combined with a biochemical approach to identify targets of a single miRNA at a time. A surprising finding was that there was a GC bias in hexamer composition in the 39 UTRs of enriched transcripts. This effect was much greater for samples transfected with miR-294 than with cel-239b, implying that there is a biological basis for this association. Such transcripts with no predicted canonical seed matches may reflect binding to miRNAs that don't follow the rules of seed-driven target recognition or non-specific associations. Recently, a role for seed sequence-independent regulation has emerged for miR-328, which acts through its C-rich region to titrate a translational inhibitor, hnRNP E2 protein, from its target mRNA [40]. However, the possibility of non-canonical targets would require careful verification.
A pertinent issue for the biochemical approach is the ability of miRNAs to induce mRNA destabilisation independently of Ago2catalysed slicing, as this could hamper the isolation of mRNAs that are being degraded. One possible way to address this in future work is to explore different Ago2 mutants that cannot interact with factors that bring about mRNA destabilisation, such as GW182 [41]. The use of such a mutant might improve the likelihood of retaining intact mRNAs and increase the efficiency of the pull down.
Despite a large discrepancy between miRNAs found in human and mouse ES cells, there are homologs of the miR-290-295 cluster in human ES cells (miR-371, miR-372 and miR-373) [10], alluding to an important role of these miRNAs in the pluripotency of mammalian ES cells. Given the expression of miR-294 in primordial germ cells [42] and its function in pluripotency, it will be interesting to determine the effects of miR-290-295 conditional deletion upon the germline.

Plasmids
The Ago2 coding sequence was amplified from 129Sv/Ev mouse ES cell cDNA by PCR using primers: Ago2-59 59-AGA ATT CAT GTA CTC GGG AGC CGG CCC CGT TCT T-39; Ago2-39 59-ATG CGG CCG CTC ACA GAT CCT CTT CTG AGA TGA GTT TTT GTT CAG CAA AGT ACA TGG TGC GCA G-39 to contain a carboxy-terminal myc-epitope tag. Q633R and H634A point-mutations were introduced by site-directed mutagenesis using the QuiKChange kit (Stratagene). For the mammalian expression constructs, wild-type and point-mutated Ago2-myc in pCAGIG were digested with Sal1-Not1 to include the CAG promoter upstream of the Ago2-myc coding sequence and ligated into pEGFP-1 (Clontech), which was pre-digested with Sal1-Not1 to remove the EGFP coding sequence.

Transfection
Transfections were performed using the Mouse ES Cell Nucleofection Kit (Lonza), and program A23 of Nucleofactor I apparatus (Lonza), as specified by the manufacturer's instructions. Transfections of miRNA mimics mmu-miR-294 and cel-239b (Dharmacon) was performed as described in [17]. Briefly, for microarray analysis, approximately 4610 6 Dicer1-null ES cells were resuspended in 90 ml mouse ES cell Nucleofection Solution. 300 pmol of miR-294 and 1 mg pEGFP-1 (Clontech) plasmid (which served as a control for transfection efficiency) were diluted in 10 ml Nucleofection Solution, mixed with the cells and transferred to a Lonza cuvette. For RNA-IP experiments, cells were transfected with 300 pmol of miR-294 or cel-239b control

miRNA-and RNA-immunoprecipitation
miRNA-and RNA-immunoprecipitation was performed in native conditions as described [45][46][47]. 2.5610 7 cells were pelleted and resuspended in 220 ml ice-cold Polysomal Lysis Buffer [100 mM KCl, 5 mM MgCl 2 , 10 mM HEPES (pH 7.0; Gibco), 0.5% NP40, 1 mM DTT, 200 U Recombinant RNasin Ribonuclease Inhibitor (Promega), 200 U SUPERaseNIn RNase inhibitor, and Protease Inhibitor Cocktail]. The lysate was passed through a 27G needle five times, incubated on ice for 30 min at 4uC and then transferred to -80uC to promote lysis. The lysate was then thawed on ice, centrifuged (15 min, maximum speed, 4uC), and the supernatant was transferred to a new tube. 400 U DNase (Roche) was added and incubated on 30 min at 4uC. 20 ml of the lysate was saved for the INPUT at -80uC. The remaining lysate was diluted in 800 ml NT2 buffer [50 mM Tris-Hcl (pH 7.4), 150 mM NaCl, 1 mM MgCl 2 , 0.05% NP40, Protease Inhibitor Cocktail, 200 U Recombinant RNasin Ribonuclease Inhibitor] and pre-cleared for 2 hr at 4uC with 20 ml Dynabeads-Protein A (Invitrogen), which had been pre-blocked in 0.5% BSA +1 mg/ml yeast tRNA (Ambion). The supernatant was incubated with 20 ml anti-myc (Cell Signaling Technology) antibody O/N at 4uC and 200 U Recombinant RNasin Ribonuclease Inhibitor. The next day, the RNA/ antibody complex was precipitated by addition of 50 ml Dynabeads-Protein A for 2 hr at 4uC. The beads were washed with 1 ml NT2 buffer four times and then resuspended in 100 ml NT2 buffer and transferred to a fresh tube. 80 ml of NT2 buffer was added to 20 ml of the stored INPUT. RNA was extracted with 1 ml Trizol according to the manufacturer's protocol and RNA was resuspended in 20 ml H 2 O. To amplify mRNA 20 cycles for sequencing, 1 ml of IP (10-40 ng/ml) and 1 ml of INPUT (diluted to 0.5-2.5 ng/ml) were used. For sequencing of mRNAs, the single cell method was modified according to [48]. For 220-plex microRNA expression profiling, 220 miRNAs were reverse transcribed, amplified and then analysed by Q-PCR, as described [49,50].

Real-time quantitative PCR
Quantitative PCR was performed on an ABI PRISM 7000 sequence detection system (Applied Biosystems). For SYBR Green fluorescent nucleic acid dye, primers were designed in order to achieve product lengths between 50 and 100 bp. TaqMan probes were used for detection of mature miRNAs by Q-PCR. The reaction conditions were: 95uC 10 min, 406 (95uC 15 s, 60uC 1 min). Data was normalised to HPRT, or INPUT RNA amount in the case of RNA-and miRNA-immunoprecipitation experiments, using the DDC T method [51].

Microarray
Feeder-free Ago2 D/D ;Dicer1 D/D +Ago2-myc ES cells were transfected in triplicate with 1 mg of pEGFP-1 and 300 pmol of miR-294 mimic or C. elegans cel-239b control miRNA and harvested 24 hr later. Cells were sorted for GFP expression and total RNA was isolated using the RNeasy Mini kit (Qiagen). 1 mg of total RNA from each triplicate (six samples in total) was sent to Cambridge Genomic Services, Cambridge University for sample processing and hybridisation to MouseWG-6 v2.0 Illumina microarrays. Cambridge Genomic Services performed the quality control analysis, gene selection, and data normalisation. Microarray data analysis was carried out using the R language with Bioconductor packages [52] and custom-written code. GO and network analysis were performed using the online software MetaCore from GeneGo Inc. Microarray data were deposited in the Gene Expression Omnibus (GEO) database (Accession Code: GSE20048).

RNA-Sequencing
The amplified cDNAs from the IP and INPUT were subjected to SOLiD library preparation by ABI. The cDNAs were then sequenced by ABI's next-generation sequencing SOLiD system. RNA-Sequencing reads were mapped to RefSeq transcripts using ABI's pipeline, as described [48]. To visualise the data, we typically plotted the fraction of reads for a given gene which came from the IP library: geneX reads in IP geneX reads in IPzgeneX reads in Input Differential representation was measured as the probability that geneX reads in library1 was drawn from a binomial distribution Figure 4. Comparison of genes enriched in the IP with computational and microarray predicted targets. (A,B) Correlation between enriched genes in the IP and computational target predictions. The x-axes represent fraction of genes ranked by increasing p-value for the likelihood of enrichment in the IP. The y-axes represent the fraction of those genes, which have a (A) TargetScan prediction (B) miRanda prediction. Red indicates Ago2-myc samples transfected with miR-294; green indicates Ago2-myc-MUT samples transfected with miR-294; blue indicates samples transfected with cel-239b control miRNA. The black line shows the expected outcome if enrichment and target predictions are independent from one another. (C) Correlation between enriched genes in the IP and microarray target predictions (selected from the Sylamer-based cut-off). The x-axis represents the fraction of genes ranked by p-value for the likelihood of enrichment in the IP. The threshold for selecting enrichment was a p-value ,0.003 from the binomial model. (D) Venn diagram overlap between Sylamer microarray target predictions and enriched genes in the IP. 8578 represents the number of genes which were considered (both microarray and RNA-IP data was available) but which were neither depleted on the array nor enriched in the IP. doi:10.1371/journal.pgen.1001163.g004 Figure 5. Functional enrichment of genes differentially expressed upon miR-294 transfection. Genes identified with differential expression from the microarray analysis were assigned functions based on Gene Ontology (GO) annotations using the GeneGo tool (GeneGo Inc). Functional ontologies are represented by (A) GeneGo canonical pathway maps (B) GO processes. The top ten terms are shown for downregulated genes (B-statistic .5) with at least one 6-mer site in its 39 UTR (orange) and upregulated genes (B-statistic .5) (blue). (C) Western blot of Lin28 expression in Dicer1 D/D +Ago2-myc ES cells transfected with cel-239b control miRNA or miR-294. doi:10.1371/journal.pgen.1001163.g005