CRISPR-Induced Expression of N-Terminally Truncated Dicer in Mouse Cells

RNA interference (RNAi) designates sequence-specific mRNA degradation mediated by small RNAs generated from long double-stranded RNA (dsRNA) by RNase III Dicer. RNAi appears inactive in mammalian cells except for mouse oocytes, where high RNAi activity exists because of an N-terminally truncated Dicer isoform, denoted DicerO. DicerO processes dsRNA into small RNAs more efficiently than the full-length Dicer expressed in somatic cells. DicerO is expressed from an oocyte-specific promoter of retrotransposon origin, which is silenced in other cell types. In this work, we evaluated CRISPR-based strategies for epigenetic targeting of the endogenous Dicer gene to restore DicerO expression and, consequently, RNAi. We show that reactivation of DicerO expression can be achieved in mouse embryonic stem cells, but it is not sufficient to establish a robust canonical RNAi response.


Introduction
Canonical RNA interference (RNAi) has been defined as sequence-specific RNA degradation induced by long double-stranded RNA (dsRNA) [1]. RNAi is initiated by processing long dsRNA by RNase III Dicer into~22 nt small interfering RNAs (siRNAs), which guide recognition and endonucleolytic cleavage of complementary mRNA molecules (reviewed in [2]). The mammalian canonical endogenous RNAi pathway is generally weak, if active at all (reviewed in [3]). Although mammalian genomes encode protein factors necessary and sufficient for reconstituting canonical RNAi in vitro [4] or in the yeast [5,6], these protein factors primarily support a gene-regulating microRNA pathway where small RNAs are produced from genome-encoded small hairpin precursors and typically guide translational repression coupled with transcript destabilization (reviewed in [7]). Furthermore, the primary mammalian mechanism responding to dsRNA in somatic cells is not RNAi but a sequence-independent interferon pathway (reviewed in [8]). The interferon pathway is an essential component of mammalian innate immunity and one of the factors impeding RNAi [9][10][11][12]. Another key factor limiting mammalian RNAi is low Dicer activity, which can be enhanced by truncating the Dicer's N-terminal helicase domain [9,[12][13][14]. Mouse oocytes, the only known mammalian cell type where RNAi is highly active and functionally important, express a unique, naturally N-terminally truncated Dicer isoform, denoted Dicer O ( Figure 1A) [14]. This truncated isoform arose upon intronic insertion of a long terminal repeat (LTR) from the MTC retrotransposon subfamily, which provides an oocyte-specific promoter and the first exon of Dicer O ( Figure 1B). This isoform is expressed only in mouse oocytes, and it was not observed in transcriptomes of somatic cells [14]. Previous studies showed that functional canonical RNAi could be restored in mammalian somatic cells under specific conditions [9,10,12,14], but it was unclear whether Dicer O reexpression from the endogenous Dicer locus could achieve such an effect as well. Accordingly, we examined whether Dicer O expression could be induced from the endogenous Dicer locus in mouse cells and whether Dicer O reactivation would be sufficient to restore robust endogenous canonical RNAi. The alternative promoter and the first exon reside in MTC LTR insertion between exons 6 and 7, and its structure and evolutionary history have been described elsewhere [15]. (C) Native chromatin immunoprecipitation analysis of H3K9 modifications in the MTC Dicer O promoter region in NIH 3T3 cells. Relative enrichment analyzed by two primer pairs is depicted as log 2 -fold enrichment calculated to IgG.
As Dicer O expression is restricted to oocytes and transcription factor(s) controlling it are unknown, we opted for an artificially designed transcriptional activator. Advances in the development of guided nucleases (reviewed in [16]) brought an opportunity to modify these sequence-specific enzymes as platforms for modulating mammalian transcription by fusing them with transcription activation or repression domains [17,18]. Among the guided nucleases, the clustered regularly interspaced short palindromic repeat (CRISPR) Cas9 nuclease gained major popularity because of simple programming of its sequence specificity with a single-guide RNA (sgRNA), which mediates recognition of a specific DNA sequence [19]. RNA-guided nuclease Cas9 was, therefore, converted to a sequence-specific DNA-binding platform, "deactivated Cas9" (dCas9), by inactivation of its two nuclease domains [19]. dCas9 can be fused to a trans-activation domain, such as VP64, to function as a transcriptional activator [20,21]. Further development brought more complex systems where the basic dCas9-VP64 module was extended by modifying the sgRNA structure such that sgRNA loops protruding from the ribonucleoprotein complex could carry sequences serving as binding platforms for additional transcription modulating proteins [22,23].
To test whether dCas9-mediated transcriptional activation could induce Dicer O expression from the Dicer gene in mouse embryonic stem cells and fibroblasts, we examined several different dCas9-transcriptional activation designs and obtained the best Dicer O activation with a system consisting of dCas9-VP64 enhanced with an MS2-p65-HSF1 module originally developed by Konermann et al. [23]. In this system, sgRNA loops protruding from the ribonucleoprotein complex carry a minimal hairpin aptamer, which is bound by the MS2 protein that can be fused with an additional factor enhancing transcription [22,23]. We report that we successfully induced robust Dicer O protein expression and observed limited RNAi activity in mouse embryonic stem cells, but not in NIH 3T3 fibroblasts. Our results suggest that induction of RNAi in mouse cells may be possible by inducing Dicer O expression, but achieving robust RNAi activity requires further optimization of Dicer O induction, presumably combined with a strategy reducing the inhibitory effects of innate immunity factors.

Plasmids
Catalytically deactivated Cas9 (dCas) was constructed from a pSpCas9n D10A plasmid (PX460; Addgene #48873) by mutating His to Ala at position 840 using Q5 site-directed mutagenesis kit (NEB) according to the manufacturer's instructions. dCas9 D10A/H480A was transferred into the pPuro backbone containing the puromycin resistance gene. VP64 was PCR amplified from a pTALE-VP64-EGFP plasmid [14] and inserted into pPuro_dCas9 together with a flexible linker.
An overview of different construct combinations used in the five characterized versions of the transcription-acting systems is provided in Table 1.

qPCR Analyses
Total RNA was isolated using RNeasy MINI kit (Qiagen), and 1 µg amount was used for the reverse transcription (RT) reaction using RevertAid premium (Fermentas) in a 30 µL volume. A 0.5 µL aliquot was used per qPCR reaction. Maxima SYBR Green qPCR master mix (Thermo Scientific) was ge C T values of the technical replicates were normalized toused for qPCR. qPCR was performed in technical triplicates for each biological sample. Avera housekeeping genes Hprt and B2MG using the ∆∆C T method [28]. Primers used for qPCR were as follows: Pou5f1:

Results and Discussion
As mentioned above, Dicer O is not expressed in other mouse cells than oocytes. Using chromatin immunoprecipitation, we examined histone modifications of the Dicer O promoter in NIH 3T3 cells to test whether Dicer O silencing involves more reversible facultative heterochromatin or more stable constitutive heterochromatin, which may be more difficult to reverse into euchromatin. Chromatin immunoprecipitation showed the presence of facultative heterochromatin mark H3K9me2, but not the constitutive heterochromatin mark H3K9me3 at the MTC LTR insertion controlling Dicer O expression ( Figure 1C). Regarding DNA methylation, the MTC LTR promoter is a CpG poor promoter where the presence of DNA methylation may not be a significant silencing factor [29].
To develop transcriptional reactivation of Dicer O in somatic cells, we tested several different designs built on the CRISPR dCas9 system recruiting transcriptional transactivator(s) into the Dicer O promoter locus. We started with a basic transcriptional activating system where dCas9 is fused with the VP64 transcriptional transactivator [20,21] and is guided by unmodified sgRNA (Figure 2A).
For initial testing of the CRISPR-driven transcriptional reactivation, we produced stable NIH 3T3 cell lines expressing dCas9-VP64 or dCas9-VP64-EGFP [14] transcriptional activators. This allowed us to reduce the number of transfected plasmids and ensure that all cells would have a constant amount of dCas9-VP64 expression. Cells stably expressing dCas9-VP64 were then co-transfected with an expression vector for sgRNA to target dCas9-VP64 into the promoter of interest cloned into a firefly luciferase reporter and an SV40 promoter-driven Renilla luciferase reporter allowing for normalization of the luciferase activity ( Figure 2B). Two different firefly reporters were produced to examine the efficiency of the dCas9 activation system, Pou5f1 (Oct-4) promoter and MTC-derived promoter for Dicer O ; for each promoter, we designed and tested several sgRNAs. Experiments with the Pou5f1 promoter-reporter and a set of sgRNAs have shown that the system can induce reporter expression ( Figure 2C). However, individual sgRNAs had just marginal effects. At the same time, strong transcriptional activation was observed when several sgRNAs were combined ( Figure 2C). Because the NIH 3T3 line expressing the dCas-VP64-EGFP fusion yielded an order of magnitude lower transcriptional activation estimated by the luciferase assay, we omitted the dCas-VP64-EGFP fusion from subsequent experiments and continued with dCas-VP64 (denoted version 1). When testing MTC promoter activation, we also observed that individual sgRNAs had a much lower effect than some of their combinations. The most powerful combination was observed with a mixture of all four tested sgRNAs (2-5) that could bind the promoter cloned into the firefly luciferase reporter ( Figure 2D). When examining transcriptional activation of the endogenous MTC promoter of Dicer O in 3T3 cells, the system yielded up to 90-fold transcriptional activation of Dicer O over the background level ( Figure 2E). Of note is that qPCR analysis of the endogenous promoter and luciferase reporter gave results differing for some sgRNAs and their combinations (e.g., sgRNA 4 vs. sgRNA 5 in Figure 2D,E). This may reflect different accessibility of the plasmid reporter and the genomic locus.
To test whether we could further enhance transcriptional activation effects, we employed a published sgRNA modification with MS2 binding sites, which allows binding of exposed sgRNA loops to the MS2 domain fused with additional transcriptional activators [23]. We tested three variants of the MS2 system: (1) an MS2-p65-HSF1 fusion construct with enhanced dimerization MS2 variant (MS2 dlFG [24]), which was encoded in a single construct together with dCas9-VP64 via T2A self-cleavage peptide as described previously [23]. We denoted this system as of version 2 ( Figure 3A); (2) modification of the MS2 component by using a covalently linked MS2-dimer (dimMS2 dlFG -p65-HSF1 [25]) in a single construct with dCas9-VP64, denoted version 3; and (3) use of separate constructs expressing dCas9-VP64 and dimMS2 dlFG -p65-HSF1, denoted version 4 ( Figure 3B). For each of the versions, we produced stable 3T3 lines and examined Pou5f1 and MTC reporter expression as described previously ( Figure 2B). A detailed overview of the variants of dCas9-VP64 systems is provided in Table 1. Expression of dCas9 and enhancing factors in stable cell lines is provided in Appendix A.
Pou5f1 reporters ( Figure 3C) revealed the best enhancing effects of version 4 of the transcriptional activation system; otherwise, the effects were comparable to the Cas9-VP64 version without MS2-based enhancement ( Figure 2C). Importantly, transcriptional activation of the endogenous MTC promoter of Dicer O in 3T3 cells with version 4 was much stronger than with version 3 ( Figure 3D). When we used transient transfection and compared version 1 and version 4, version 4 was clearly superior to version 1, suggesting that modified sgRNA and dimMS2 dlFG -p65-HSF1 indeed have an enhancing effect ( Figure 3E). In the same experiment, we also examined alternative MS2-based enhancement employing dimMS2 dlFG -p300 acetyltransferase fusion [22], denoted version 5, but version 4 using separate constructs expressing dCas9-VP64 and dimMS2 dlFG -p65-HSF1 was clearly superior ( Figure 3E). We thus selected version 4 for analysis of the production of the Dicer O protein.
To analyze Dicer O production's activation, we used stable cell lines expressing version 4 ( Figure 4A) and transient transfection of sgRNA-expressing constructs ( Figure 4B). When deploying the system into embryonic stem cells (ESCs), we were able to detect robust transcriptional activation and production of Dicer O protein, whose abundance in the lysate of transiently transfected cells was comparable to the endogenous Dicer protein ( Figure 4C). Although we achieved clear transcriptional activation of Dicer O in NIH 3T3 cells, the Dicer O transcript levels were approximately a hundredfold less induced than in ESCs, and the truncated protein could not be detected by Western blotting ( Figure 4D). Because of this result, we tested the induction of RNAi activity in ESCs.
To test the RNAi activity ( Figure 4E), we co-transfected ESCs stably expressing version 4 with a combination of four sgRNAs that yielded the most of the Dicer O protein and our established RNAi assay system described in detail previously [12]. Briefly, the RNAi assay system has three plasmid components: (1) a targeted Renilla luciferase reporter with a cognate Mos sequence in the 3 UTR, (2) a non-targeted firefly luciferase reporter, and (3) a plasmid expressing long dsRNA with the Mos sequence in the form of a long hairpin (MosIR). As a control for nonspecific dsRNA effects, we used unrelated dsRNA-expressing plasmids (Lin28IR and Elavl2IR) instead of MosIR. As negative control to MosIR, we used a MosMos construct, where Mos sequences are oriented head-to-tail. Hence the plasmid has the same sequence as MosIR but does not produce dsRNA. Western blot analysis again showed good induction of Dicer O expression and a marked reduction of the targeted Renilla reporter in the presence of MosIR ( Figure 4E). The average Renilla reporter reduction was 30% (p-value > 0.05), which was comparable to our previous experiments with cells that have an intact protein kinase R gene encoding one of the dsRNA-sensing components of the interferon response [12,14]. To sum up, we tested five different versions of the dCas9-VP64 transcriptional activation system, out of which we selected dCas9-VP64 combined with sgRNA carrying MS2-binding sites and a dimMS2 dlFG -p65-HSF1 fusion protein. Our work shows the extent of optimization needed to induce robust transcriptional activation (even if only in ESCs). It underscores the importance of setting up a good testing system for multiple sgRNAs and their combinations. We opted for producing stable lines expressing dCas9 and the enhancing factor. These cell lines represent one of the valuable contributions of this work as they can be used to program dCas9 with sgRNAs to influence gene expression in ESCs and 3T3 cells.
Using this three-component system and a combination of four sgRNAs targeting the MTC promoter of Dicer O isoform, we achieved robust Dicer O protein expression in ESCs comparable to full-length Dicer expression. In fact, Dicer O protein expression in cells transfected with sgRNAs likely exceeded that of endogenous full-length Dicer in those cells because in Western blots in Figure 4C,E, the Dicer O signal comes from fewer cells than the full-length Dicer protein detected above it. However, despite the level of Dicer O expression, the RNAi activity assessed by a reporter assay was minor, if any, suggesting that the sole expression of Dicer O is not sufficient to bring robust RNAi. On the other hand, this observation is consistent with other data showing that RNAi activity is low in the presence of specific innate immunity factors [9][10][11][12].  [14]. The graph depicts the relative activity of a Renilla luciferase reporter carrying a cognate Mos sequence as described previously [12]. Briefly, the Renilla-Mos reporter is co-transfected with a non-targeted firefly luciferase reporter and a reporter expressing dsRNA from an inverted repeat (IR). pBS and MosMos are controls not expressing dsRNA. All error bars = standard deviation.
Unfortunately, we did not succeed in inducing a similar expression of Dicer O in NIH 3T3 cells. This result seems to be caused by much lower transcriptional activation than in ESCs and could be influenced by different chromatin organization in ESCs and NIH 3T3 cells, which are derived from fibroblasts [30]. ESCs have a relatively open chromatin structure supporting higher transcription and reduced heterochromatin signature [31,32]. Thus, ESCs may better respond to transcriptional reactivation of Dicer O than NIH 3T3 cells. The system with histone acetyltransferase p300 (version 5) did not have an additive effect on top of VP64 ( Figure 3E), suggesting that targeting acetylation to the locus does not sufficiently enhance transcriptional activation of the MTC promoter. Perhaps a histone lysine demethylase targeted to the locus that would reduce H3K9me2 could make the transcriptional activation of Dicer O more efficient.
In conclusion, we showed that reactivation of Dicer O via a dCas9 system is not a viable strategy to induce a robust canonical RNAi pathway in cultured mouse cells. While rather negative, our results are useful for understanding the functional limits of the endogenous RNAi and long dsRNA processing capacity furnished by the expression of Dicer O from its endogenous locus. Although this time we did not observe robust RNAi, it is likely that more sensitive methods, such as RNA sequencing of small RNAs, would reveal the impact of expressed truncated Dicer on the long dsRNA metabolism. In our previous work in 3T3 and ESCs, we observed a robust increase in siRNA biogenesis with ectopic expression of Dicer O but rarely robust RNAi [12]. Our work offers another tool for studying the consequences of Dicer O expression in mouse cells and investigating whether low-activity RNAi would have any measurable impact on cell physiology, especially dsRNA metabolism, miRNA pathway, the sensitivity of innate immunity pathways to dsRNA, and viral resistance.