Dicer, a new regulator of pluripotency exit and LINE‐1 elements in mouse embryonic stem cells

A gene regulation network orchestrates processes ensuring the maintenance of cellular identity and genome integrity. Small RNAs generated by the RNAse III DICER have emerged as central players in this network. Moreover, deletion of Dicer in mice leads to early embryonic lethality. To better understand the underlying mechanisms leading to this phenotype, we generated Dicer‐deficient mouse embryonic stem cells (mESCs). Their detailed characterization revealed an impaired differentiation potential, and incapacity to exit from the pluripotency state. We also observed a strong accumulation of LINE‐1 (L1s) transcripts, which was translated at protein level and led to an increased L1s retrotransposition. Our findings reveal Dicer as a new essential player that sustains mESCs self‐renewal and genome integrity by controlling L1s regulation.

A gene regulation network orchestrates processes ensuring the maintenance of cellular identity and genome integrity. Small RNAs generated by the RNAse III DICER have emerged as central players in this network. Moreover, deletion of Dicer in mice leads to early embryonic lethality. To better understand the underlying mechanisms leading to this phenotype, we generated Dicer-deficient mouse embryonic stem cells (mESCs). Their detailed characterization revealed an impaired differentiation potential, and incapacity to exit from the pluripotency state. We also observed a strong accumulation of LINE-1 (L1s) transcripts, which was translated at protein level and led to an increased L1s retrotransposition. Our findings reveal Dicer as a new essential player that sustains mESCs self-renewal and genome integrity by controlling L1s regulation.
Since its discovery in 2001 [1], extensive studies revealed DICER as a key player of RNA interference (RNAi) processes. Indeed, this RNase III protein is essential for microRNAs (miRNAs) and small-interfering RNAs (siRNAs) biogenesis [2][3][4]. These eukaryotic small RNAs are central players in many biological processes by mediating gene silencing at transcriptional or post-transcriptional levels [5]. They are also essential actors of early mammalian development as key regulators of cell cycle and proliferation [6]. Moreover, particular miRNAs are also involved in embryonic stem cell fate regulation by promoting selfrenewal and differentiation [7][8][9][10][11]. The disruption of the Dicer gene leads to early embryonic lethality at the implantation stage, emphasizing its critical role during mouse early development [12,13]. Besides, RNAi pathways can act as defense mechanisms against endogenous and exogenous factors like transposable elements (TEs) and viruses [14][15][16]. In mammals, first evidence of TEs regulation by RNAi was reported in Dicer-depleted preimplantation mouse embryos, where specific subclasses of TEs were upregulated [14]. Nevertheless, the exact mechanisms by which RNAi players could act on TEs and the consequences of this regulation during early development remain unclear.
To better understand the functions of Dicer during early mammalian development, we used mouse embryonic stem cells (mESCs) as a model system. Derived from the inner cell mass of mouse blastocyst, mESCs present two substantial advantages: first, they can be maintained in a pluripotent state or conversely be differentiated into the three germ layers depending on the culture conditions [17]. Thus, making them a suitable model to study mouse embryonic developmental stages in vitro, otherwise difficult to assess in vivo. Second, TEs are not submitted to their major regulatory mechanisms at the blastocyst stage. Both, the DNA methylation and the PIWI-interacting RNA (piRNA) silencing taking place in somatic and germ cells, respectively, are absent at this stage [18,19]-suggesting the existence of alternative regulatory pathways. Therefore, mESCs represent a relevant model to study TEs regulation during mouse early development as well.
Long INterspersed Element-1 (LINE-1 or L1s), long terminal repeat (LTR), and short interspersed nuclear element (SINE) are the three main subgroups composing the retrotransposons family, which are the major class of TEs represented in mammalian genomes [20][21][22][23]. L1s are the most abundant TEs in human and mouse genomes (21% and 17%, respectively) [21,22]. They belong to the autonomous retrotransposon category, as they code for the machinery necessary for the RNA intermediate production, its reverse transcription, and integration into a new genomic location [24]. Although the large majority of L1s are inactive [25], it is estimated that around 3000 full-length L1s have maintained their ability to retrotranspose in the mouse genome [26][27][28]. Active full-length L1s, via their retrotransposition ability, can act as mutagens by inserting into exons, or induce aberrant splicing or exon skipping by inserting into introns [29]. Therefore, they can deeply influence the genome, in beneficial and detrimental ways [30], and need to be tightly controlled.
In order to investigate the roles of DICER during mouse early development, we generated new Dicer knockout mESCs mimicking previously described Dicer Cre-loxP mutants [31,32]. Their detailed characterization highlighted their inability to differentiate and revealed for the first time their incapacity to exit from the pluripotent state and a factual reinforcement of their pluripotency network. Additionally, transcriptome analysis of wild-type (WT) and Dicer_KO mESCs unveiled an up-regulation of LINE-1 transcripts. This increase of L1s mRNAs was translated at the protein level and led to an augmentation of their retrotransposition rate. Taken together, our experiments highlight critical roles of Dicer in the regulation of the pluripotency network and the control of LINE-1 elements in mESCs.

Materials and methods
Culture and in vitro differentiation of mESCs E14TG2a (ATCC CRL-1821) line has been used as WT mESCs. Cell culture and embryoid body (EB) differentiation assays were performed as described in [33]. Unless otherwise specified, mESCs were routinely cultured in serum + LIF condition.

Generation of Dicer_KO mESCs using CRISPR/ Cas9
Dicer_KO mESCs were generated from E14TG2a mESCs using a paired CRISPR/Cas9 strategy as described in [34]. Specific CRISPR/Cas9 sgRNAs have been generated using the E-CRISPR software [35] or chosen from an established library [36] and cloned into the plasmid pX330-U6-Chimeric_BB-CBh-hSpCas9 [37] using the BbsI restriction site. mESCs were single cell sorted 48 h after transfection. All the primers used for the CRISPR/Cas9 are described in Table S1. All newly generated plasmids are described in Table S2. All designs are based on the latest mouse genome assembly (GRCm38/mm10) provided by the UCSC Genome browser http://genome.ucsc.edu/.

Genomic DNA extraction and PCR
Genomic DNA was extracted from 1.10 6 mESCs using Roti Ò Phenol/Choloroform/Isoamyl Alcohol. Each PCR reaction has been performed using 50 ng of genomic DNA. Genotyping PCR primers sequences are described in Table S1.

RT-qPCR analysis
RT-qPCR analysis was performed as described in [33]. All the primers used for the RT-qPCR assays are described in Table S1.

Immunoblotting analysis and antibodies
Immunoblotting analysis was performed as described in [33]. All the antibodies used for the immunoblot assays are described in Table S3. In the case of subsequent reprobing, polyvinylidene difluoride membranes were reactivated into methanol, and then stripped with successive 0.2 M NaOH washes. Finally, membranes were blocked during 1 h at room temperature using a 5% milk solution, before reprobing with a second primary antibody.

Low molecular weight northern analysis
Low molecular weight northern analysis were performed as described in [38] using 10 lg of total RNA extracted from 1.10 6 mESCs pellets using TRizol Ò Reagent. Membranes were EDC cross-linked. For subsequent reprobing, membranes were stripped with boiling 0.1% SDS. All the DNA oligonucleotides complementary to miRNAs and U6 small RNA, used for the probes generation, are listed in Table S1. Hyridization Buffer at 42°C. All washes were performed in SSC 2X, SDS 0.1%. The radiolabeled L1_probe for the detection of full-length L1 transcripts was produced by random-priming of a PCR product generated from E14TG2a mESCs genomic DNA using specific primers [39] described in Table S1.

RNA sequencing
Total cellular RNA was extracted from 1.10 6 mESCs pellets using TRizol Ò Reagent. The quality of isolated RNA was determined with a Bioanalyzer 2100 (Agilent, Santa Clara, CA, USA). Up to 2 lg of polyA purified RNA was used for the library preparation, done with the TruSeq Pairedend stranded RNA Preparation Kit (Illumina, San Diego, CA, USA). The library preparation and sequencing (Illumina HiSeq 2000) were performed at the Functional Genomics Center Zurich (FGCZ). Paired end sequencing generated about 2 9 60 millions of reads per library. Reads from RNA sequencing were first preprocessed by trimmomatic (v0.32) [40] to remove low-quality ends and adapters using default settings. Reads were aligned to the mouse genome mm10 by STAR (v2.4.2a) [41] allowing for two mismatches and up to 3000 multiple-hits. FeatureCounts (v1.4.5-p1) [42] was used to count reads for genes (Ensembl GRCm38.78), ignoring reads on overlapping region and the plot was generated using GGPLOT2 (v1.0.1) [43]. TETOOLKIT (v1.5) [44] was used to count reads for repeat elements, accounting for multiple-hit reads and RPKM were calculated by using edgeR [45]. Complete RNA sequencing data of WT and Dicer_KO mESCs are available on the NCBI GEO database (GEO: GSE78971 for WT and GEO: GSE78973 for Dicer_KO).

Proliferation assay
Cells were plated in 96-well plate at a density of 15 000 cellÁcm À2 and proliferation was assessed every day during 4 days using the CellTiter-Glo Ò Luminescent Cell Viability Assay.

Cell cycle analysis
Cell cycle analysis was performed as described in [46].

Apoptotic cell population analysis
Apoptotic cell population analysis was performed as described in [47] (direct DNA staining in PI hypotonic solution and subsequent analysis by FACS).

Exit from pluripotency assay
Cells were plated in six-well plate at a density of 4500 cellsÁcm À2 and cultured in 2i medium (N 2 B 27 (Cellartis) complemented with 50 UÁmL À1 of penicillin and 0.05 mgÁmL À1 of streptomycin) containing or not the following inhibitors cocktail: PD032591 at 1 lM final concentration, CHIR99021 at 3 lM final concentration, and 1000 UÁmL À1 of leukemia inhibitory factor (LIF). The alkaline phosphatase (AP) staining was performed using the Leukocyte Alkaline Phosphatase kit (Sigma, St. Louis, MO, USA). For the clonal AP quantification, entire six-well plates used for AP staining assays were first scanned to capture the total plate area in a single image. Images were then processed using the ImageJ software [48]. The number of AP-positive colonies was calculated on threshold intensity (default parameters) of inverted regions that were user-selected (full wellidentical areas between conditions) using the Analyze Particles tools (default parameters).

Immunostaining
Cells were washed once with PBS1X, incubated 10 min at 37°C with 4% paraformaldehyde solution for fixation and then incubated 15 min on ice in a 90% methanol solution for permeabilization. Next, cells were incubated 1 h at room temperature with the primary and secondary antibody, successively. Between incubation steps, cells were washed once with PBS1X. Antibodies used for the immunostaining assays are described in Table S3. Cells were analyzed by flow cytometry using selective gating to exclude the doublets of cells.

Retrotransposition assay
Cells were plated at a density of 20 000 cellsÁcm À2 per well 24 h before transfection with 0.5 lg of plasmid DNA using Lipofectamine Ò 2000 reagent according to the manufacturer's instructions. Antibiotic selection started 24 h after transfection using puromycin-containing medium (1 lgÁlL À1 ) and maintained during the entire assay. Every week, cells were trypsinized and replated at a density of 5500 cellsÁcm À2 into a new gelatin-coated six-well tissue culture plate and the remaining cells were used for subsequent FACS analysis. In total, WT cells have been passaged six times and Dicer_KO mESCs four times (due to their proliferation defects). Cells were analyzed by FACS using selective gating excluding doublets of cells (Fig. 5E). The gating for EGFP-positive and EGFP-negative cells was determined by analyzing cells transfected with: a plasmid coding EGFP (positive control) and a puromycin-resistance gene and a plasmid coding only a puromycin-resistance gene (negative control), respectively (Fig. 5B). A final gate of 3.10 4 events per sample was acquired.

Generation and validation of Dicer_KO mESCs
We first generated two independent Dicer1 knockout (Dicer_KO) mESC lines using the CRISPR/Cas9 technology [37,49,50]. We opted for the paired CRISPR/ Cas9 approach [34,51] and generated two independent genomic deletion events D23 and D13 (Fig. 1A). Independent mESC clones were isolated and genomic deletions were confirmed by PCR (Fig. 1B) [34]. Immunobloting analysis validated the absence of DICER protein in both mutant mESC lines (Fig. 1C). The nonfunctionality of the Dicer knockouts was confirmed with the absence of two endogenous mature miR-NAs: miR-16 and miR-295 (Fig. 1D) [38,52,53]. The accumulation of miR-16 precursors (pre-miRNA) in both Dicer_KO mESCs proved the functionality of the microprocessor complex (DROSHA and DGCR8; Fig. 1D). Furthermore, immunoblotting revealed no differences in the expression of the other RNAi pathway proteins: DROSHA, DGCR8, and AGO1, between Dicer_KO mutants and WT mESCs (Fig. 1E) [2]. However, we observed dramatically reduced AGO2 levels in both Dicer_KO mutants, consistent with the lack of mature miRNAs leading to the destabilization of the AGO2 protein [54]. Finally, the analysis of the RNA sequencing data confirmed the loss of Dicer mRNA in both Dicer_KO mutants and profound changes in the transcriptome with 879 genes differentially expressed ( Fig. 1F and Table S4), involved in many biological pathways (Fig. 1G). Most of the genes differentially expressed were as expected up-regulated (80%), due to the essential role of DICER in post-transcriptional gene silencing mechanisms mediated by miRNAs ( Fig. 1F and Table S4). Taken together, these experiments     validate the successful generation of two new independent Dicer knockout lines. Next, we characterized our Dicer_KO mESCs and evaluated their proliferation rate. After 3 days, both Dicer mutants showed a strongly impaired proliferation (twofold) compared to WT mESCs. The proliferation defect was exacerbated after 4 days (threefold), confirming the delay (Fig. 2A). The cell cycle distribution analysis revealed an accumulation in G1-phase in both Dicer mutants, suggesting an impaired G1/S transition as the direct cause of the proliferation defect (Fig. 2B). Indeed, many miRNAs regulate the entry and G1-S-phase transition [55], making this observation consistent with the lack of miRNAs of Dicer_KO mESCs. Interestingly, Dicer_KO mESCs also showed a twofold increase of the apoptotic cells population compared to WT (Fig. 2C). Importantly, it has been also documented that numerous miRNAs are involved in apoptosis regulation [56]. In conclusion, newly generated Dicer_KO mESCs proliferate much slower due to a G1-phase arrest and an increased apoptosis rate.
Previously characterized Dicer Cre-loxP mutant mESCs failed to contribute to the embryo development when injected into WT blastocyst and could not differentiate in vitro [31]. To understand the molecular mechanisms leading to this differentiation defect, we first tested the ability of our mutants to form EBs in vitro. When cultured in suspension in the absence of cytokine LIF, mESCs form cell aggregates known as EBs, differentiating toward the three germ layers [17]. After 10 days of EBs differentiation (Day 10), WT mESCs produced fully developed EBs, while Dicer_KO mutants formed cells aggregates without morphological evidence of differentiation (Fig. 2D). RT-qPCR analysis performed at Day 0 and Day 10 with specific primers for the differentiation markers, Pax6, Nestin, Fgf5 (ectoderm; Fig. 2E); Dab2, Gata6, Gata4 (endoderm; Fig. 2F); and Fgf8, Brachyury, Actc1 (mesoderm; Fig. 2G), revealed that Dicer_KO mutants failed to differentiate to any of the three germ layers (Fig. 2E-G). Additionally, we assessed the expression of the pluripotency markers, Oct4 (Pou5f1), Sox2, and Nanog (OSN), at the mRNA and protein levels. These transcription factors constitute the core of the stem cell pluripotency network and are strongly expressed in undifferentiated mESCs and silenced during the differentiation process [57,58]. RT-qPCR analysis revealed a strong decrease of OSN mRNAs in WT mESCs after 10 days of differentiation (Fig. 2H). However, Dicer_KO mESCs presented an abundant accumulation of those mRNAs even after 10 days of differentiation (Fig. 2H). Immunoblotting analysis showed similar protein levels of these transcription factors in Dicer mutants and WT mESCs at Day 0 (Fig. 2I). More importantly, OCT4, NANOG, and SOX2 proteins were still expressed at Day 10 in both Dicer mutants, whereas no or very weak signals were observed in WT mESCs (Fig. 2I). These results confirm that Dicer is indeed necessary for the differentiation of mESCs.
Dicer is essential to exit the pluripotent state of mESCs For their commitment to differentiation, mESCs have to exit self-renewal state, repress the pluripotency network and initiate specific cellular lineage programs [59]. The high expression of the pluripotency core proteins observed at Day10 of EB differentiation (Fig. 2H) pointed toward a failure of our mutants to suppress the pluripotency network and to exit the pluripotent state. To test this hypothesis, we performed an exit from pluripotency assay [60][61][62]. Both Dicer_KO and WT mESCs were cultured during 3 days in a chemically defined medium (2i medium), containing selective GSK3b and MEK 1/2 inhibitors and LIF, to enhance viability of mESCs and to increase maintenance of pluripotency [63]. Subsequently, the cells were cultured for 4 days in a differentiation-permissive medium (2i medium without inhibitors and LIF) and afterwards, the 2i medium was restored for three more days before AP staining was performed (Fig. 3A). Only pluripotent stem cells can survive and express AP in 2i medium. To account for the strong proliferation defect of the Dicer_KO lines, we extended the permissive culture of the original protocol [62] from 3 to 4 days. After the exit from pluripotency assay, WT mESCs did not form colonies resembling embryonic stem cells and were AP negative, indicating that these cells committed to differentiation properly (Fig. 3B). In contrast, both Dicer mutants formed distinct AP-positive colonies (Fig. 3B), demonstrating for the first time that Dicer_KO mESCs were still able to proliferate in 2i medium after 4 days in permissive medium. The quantification of the total cell population revealed a strong increase (20-fold) of AP-positive colonies for both Dicer mutants compared to WT mESCs (Fig. 3B). This result indicates that Dicer_KO mESCs retained their self-renewal potential and remained undifferentiated in permissive conditions. Therefore, Dicer_KO mESCs failed to exit from the pluripotent state or presented a strong delay for their commitment.
In order to investigate the stemness status of our mutants, we assessed the expression of pluripotency and stem cell factors in different culture conditions [64,65]. OCT4/NANOG coimmunostaining flow cytometry analysis revealed that Dicer_KO mESCs cultured in serum + LIF condition presented a significant enrichment of cells coexpressing the pluripotent factors compared to WT mESCs (Fig. 3C). Furthermore, Dicer_KO mESCs presented similar coexpression levels when cultured in serum + LIF or 2i condition (Fig. 3D), thus indicating a reinforced pluripotency network compared to WT mESCs [65,66]. Additionally, similar enrichments were observed for the coexpression of two other pluripotent markers STELLA and SSEA-1 (Fig. 3E) [67,68]. Altogether, these observations reveal that Dicer depletion leads to a strengthening of the pluripotency network.

LINE-1 elements are strongly up-regulated in Dicer_KO mESCs
Interestingly, the analysis of the Dicer mutant transcriptomes revealed a significant accumulation of two particular TEs subclasses transcripts: L1s and LTR, compared to their WT counterparts (Fig. 4A). These observations are consistent with earlier reports showing the accumulation of transcripts from these two specific retrotransposon subgroups after Dicer knockout or knockdown during mouse early development [14,31]. However, we observed no difference in the expression of the SINE subclass (Fig. 4A). These observations were confirmed in our Dicer mutants by RT-qPCR (Fig. 4B). We focused our interest on the L1s subclass because they are the most abundant TEs in the mouse genome, and decided to monitor L1s in our system at mRNA and protein levels (Fig. 4C-E) [33]. RT-qPCR performed with primers designed in the ORF2 (L1_ORF2; Fig. 4C and Table S1) showed an eightfold increase of L1s mRNA accumulation in both Dicer mutant compared to WT mESCs (Fig. 4D). Using qPCR primers specific for each L1s subfamily (L1_Tf, L1_Gf and L1_A; Fig. 4C and Table S1), we were able to observe an accumulation of all L1s subtypes in Dicer mutant mESCs compare to WT cells (sixfold for the L1_Tf subfamily, fourfold for both, L1_Gf and L1_A subfamilies; Fig. 4D). Additionally, L1_ORF1 protein (derived from active murine L1s) was also strongly up-regulated in Dicer mutant mESCs (Fig. 4E). We concluded that in the absence of DICER, all L1s subclasses are up-regulated at mRNA and protein levels.

DICER restricts LINE-1 retrotransposition in mESCs
To investigate if the increased expression of L1s could result in an augmentation of their retrotransposition rate, we first performed high molecular weight northern blotting to monitor full-length L1s transcripts, which constitute retrotransposition-competent intermediates. We observed a strong accumulation of L1s fulllength transcripts in Dicer_KO mESCs (Fig. 5A). Next, we performed an EGFP-based retrotransposition assay in mESCs using the L1 RP -EGFP transgene [69][70][71][72]. This construct has been previously used to track embryonic L1s retrotransposition events in mice in vivo [73]. The transgene is composed of a L1 RP element fused to an EGFP gene ( Fig. 5B-C). The EGFP reporter gene is expressed only if the L1 RP element completes a full retrotransposition cycle and therefore, assessment of EGFP expression allows the evaluation of the L1 transgene retrotranposition rate (Fig. 5C). The proportion of GFP-positive cells observed after the L1 RP -EGFP transgene transfection is expected to be representative of the number of L1s retrotransposition events, and can be used to compare L1s retrotransposition capacity between mESC lines. As a negative control, we used the L1 JM111 -EGFP transgene, a mutated version of the L1 RP -EGFP transgene, that is unable to retrotranspose ( Fig. 5B-C) [73]. We transfected both Dicer mutants and WT mESCs with the L1 RP -EGFP and the L1 JM111 -EGFP constructs and measured EGFP expression after 3 and 6 weeks by FACS analysis (Fig. 5D-E). No differences between the mESC lines were detected after 3 weeks (Figs 5F and 6). Importantly, after 6 weeks, WT mESCs transfected with the intact construction (pL1 RP ) or with the mutated one (pL1 JM111 ) presented similar low levels of GFP-positive cells, indicating very low retrotransposition activity. However, both Dicer mutants transfected with the L1 RP vector showed a significant increase (twofold) of GFP-positive cells compared to their corresponding negative control and to WT mESCs (Figs 5F and 6). We hypothesize that the long period needed is probably due to the high cell mortality observed after transfection and during selection of the Dicer mutants. Moreover, the proliferation defect limited the number of cells available for the FACS analysis, thus leading to a possible underestimation of the retrotransposition events in our Dicer mutant cells [74,75]. Therefore, we concluded that in the absence of Dicer, mESCs accumulate full-length L1s transcripts and are more permissive to the L1s retrotransposition, demonstrating that Dicer is indeed involved in the regulation of L1s retrotransposition in mESCs.

Conclusion
In this study, we successfully generated and characterized two new independent Dicer1 knockout (Dicer_KO) mESC lines using the CRISPR/Cas9 technology, demonstrating the effectiveness of the paired strategy. This approach allowed us to produce deletions resulting in a complete ablation of the DICER protein function, mimicking the previously generated Dicer Cre-loxP mutants [31,32,76]. We further demonstrated that Dicer_KO mESCs are unable to exit from the pluripotency state and presented a factual reinforcement of the pluripotency network. Therefore, future studies involving the role of Dicer in stem cell biology should focus on cellular networks involved in pluripotency exit, an early step of mESCs commitment, rather than in the later stages of the differentiation process [62]. Interestingly, the transcriptome analysis of these mutants revealed a strong accumulation of transcripts from the L1 TE subclasses. We experimentally validated their up-regulation at mRNA as well as at protein levels. Moreover, we assessed the L1 retrotransposition activity in our Dicer_KO and WT mESCs and observed increased retrotransposition events in our mutant cells. However, we did not observe a direct correlation between L1s transcripts abundance and retrotransposition activity in our Dicer mutants, as it has been previously reported in human cell lines [77,78]. As retrotransposition events affect only 1% of our mutant cells population (Fig. 5F), it is unlikely that the increased retrotransposition rate is the cause of the exit from pluripotency failure. Nevertheless, the consequential accumulations of L1s transcripts and proteins observed might participate in this inability. For example, the activation of surveillance pathways or quality control mechanisms might prevent cellular differentiation in the presence of increased L1s activity, in order to avert genome instability [79,80]. Importantly, cell survival is intrinsically linked to genome instability [81]. Therefore, the increased apoptosis rate observed in our mutants might grant to Dicer a potential role in genome integrity maintenance and further support this hypothesis. Interestingly, possible effects of TEs overexpression in mESCs deserve further investigations, as they would imply that a tight monitoring of L1s (and LTR) is essential for normal mammalian development process, due to their essential role in genome integrity. Finally, as miR-NAs play a role in the regulation of the transcriptional network controlling pluripotency in mESCs [7,11], it is therefore possible that DICER is required in the exit from pluripotency process through its role in the biogenesis of miRNAs.
Nevertheless, our work, together with other studies performed in human cultured cells, indicates a role of DICER as a player in L1s regulation [77,82,83]. How DICER controls L1s still remains unclear and further investigations are needed. Recently, a study performed in human cells indicated that a particular microRNA, miR-128, was involved in the direct regulation of L1s transcripts [83]. Nevertheless, miR-128 is not expressed in mESCs (data not shown) and therefore cannot explain the regulation of L1s by DICER.
Among the other models proposed, one involves bidirectional transcription of L1 promoters and the potential to generate double-stranded RNA precursors. These are suitable substrates for DICER resulting in the production of endogenous siRNA, which can trigger repression of the corresponding homologous L1s sequences [1,[84][85][86]. Indeed, several studies reported the presence of active sense and antisense transcription Northern blot analysis using WT and Dicer_KO mESCs total RNA extract probed with a specific L1_probe. Full-length L1s transcripts are indicated with an arrow. Ethidium bromide staining before transfer was used to confirm equal loading. 28S RNA is shown as a loading control. (B) Description of the different plasmids used for the L1 EGFP-based retrotransposition assay. (C) Schematic representation of the L1 EGFP-transgene and its retrotransposition (adapted from [73]). The L1-EGFP transgene (pL1 RP ) consists of a human L1 RP element driven by the mouse RNA pol II promoter in addition to its endogenous 5 0 UTR. This element is coupled to an EGFP gene directed in the antisense orientation and interrupted by the mouse c-globin intron in the same transcriptional orientation as the L1. Therefore, when the L1-EGFP transgene transcript is processed, the mouse c-globin intron is spliced out and the EGFP gene can be expressed after reverse transcription and integration into the genomic DNA. In the case of retrotransposition events, mESCs will express EGFP. In the negative control (pL1 JM111 ), the L1 RP element has been replaced by the L1 JM111 element. The L1 JM111 element is a nonfunctional L1 transgene consisting in a human L1 mutated in ORF1 (*) [70], abrogating its retrotransposition activity. (D) Retrotransposition assay experiment design and time line in mESCs. (E) Flow cytometry gating strategy for the analysis of GFPpositive cells in (F). We first selected the mESC population and subsequently excluded the doublets in both dimensions. The data from the first triplicate of D13 mESCs transfected with the pL1 RP (week 6) plasmid were used to represent the gating strategy. The gating for EGFPpositive and EGFP-negative cells was determined by analyzing cells transfected with: a plasmid coding EGFP and a puromycin-resistance gene and a plasmid coding only a puromycin-resistance gene respectively described in (  from human and murine active L1s [77,87,88]. This model is also supported by the ability of mESCs to produce Dicer-dependent siRNAs [89], and the identification of a population in mESCs of sense and antisense small RNAs mapping to the 5 0 UTR of active L1_Tf elements [90,91]. To further explore the implication of Dicer is this regulation, it would be interesting to monitor L1s expression between the different mutants of the RNAi pathways. These mutants must be generated in the same genetic background to allow their comparison as the TEs composition differs depending on the mouse strains [92].
In conclusion, our results explain the previously observed impaired differentiation process of Dicer_KO mESCs and reveal that DICER is essential for the exit from pluripotency of mESCs and the regulation of L1 elements.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Table S1. Primers list. Table S2. Newly generated plasmids. Table S3. Antibodies list. Table S4. Differentially expressed genes in Dicer_KO mESCs.