MicroRNA-deficient embryonic stem cells acquire a functional Interferon response

When mammalian cells detect a viral infection, they initiate a type-I Interferon (IFNs) response as part of their innate immune system. This antiviral mechanism is conserved in virtually all cell types, except for embryonic stem cells (ESCs) and oocytes which are intrinsically incapable of producing IFNs. Despite the importance of the IFN response to fight viral infections, the mechanisms regulating this pathway during pluripotency are still unknown. Here we show that, in the absence of miRNAs, ESCs acquire an active IFN response. Proteomic analysis identified MAVS, a central component of the IFN pathway, to be actively silenced by miRNAs and responsible for suppressing IFN expression in ESCs. Furthermore, we show that knocking out a single miRNA, miR-673, restores the antiviral response in ESCs through MAVS regulation. Our findings suggest that the interaction between miR-673 and MAVS acts as a switch to suppress the antiviral IFN during pluripotency and present genetic approaches to enhance their antiviral immunity.


Introduction
Type-I Interferons (IFN) are crucial cytokines of the innate antiviral response. Although

44
showing great variation, most mammalian cell types are capable of synthesizing type-I IFNs in 45 response to invading viruses and other pathogens. Once type-I IFNs are secreted, they activate the JAK-STAT pathway and production of Interferon-stimulated genes (ISGs) in both the infected and neighbouring cells to induce an antiviral state (Ivashkiv and Donlin, 2015). Two to that of plants and insects, is suggested to function as an alternative antiviral mechanism (Maillard et al., 2013). And in humans, ESCs intrinsically express high levels of a subgroup of antiviral signalling protein), an essential and central protein in the interferon response pathway.
NIH3T3 and BV-2 cells, which correlates with the ability of these cell lines to induce 117 Ifnb1mRNA expression (Figure 1C).

118
The ability of cells to express IFN in response to viruses or immunogenic nucleic acids is 119 assumed to be acquired during differentiation. To test this model, we in vitro differentiated 120 both ESC lines with retinoic acid and determined their ability to respond to poly(I:C). Briefly, 121 embryoid bodies were generated by a hanging droplet method for 48 hours before being 122 cultured in the presence of retinoic acid for 2 or 10 days. Samples from each of these time 123 points were analysed for expression of pluripotency and differentiation markers. The 124 pluripotency markers Nanog and Pou5f1 (Oct-4) showed a rapid decrease in mRNA expression 125 during differentiation in both the cell lines (Suppl. Figure 1A), whereas differentiation 126 markers Neurog2, Gata-6 and Gata-4 showed a gradual increase (Suppl. Figure 1B) 127 confirming successful differentiation of the ESCs. Next, we compared the ability of ESCs (day 128 0) and retinoic-acid differentiated cells after 10 days (day 10) to express Ifnb1 mRNA in 129 response to poly(I:C), and confirmed that differentiated cells acquired the ability to synthesize 130 Ifnb1 to similar levels to the positive control cell line, BV-2 ( Figure 1D).

132
Given the relevance of RNAi as an antiviral mechanism in mouse ESCs (Maillard et al., 2013), 133 we next asked if ESCs, in the absence of the central factor for RNAi, Dicer, would be more 134 susceptible to RNA viruses. Unexpectedly, Dicer -/-ESCs were more resistant to viruses 135 compared to their wild-type counterparts (previously named ESC2) (Figure 2A). Similar 136 results were obtained using the (-) ssRNA virus, Influenza A (IAV) ( Figure 2B). Importantly, 137 mammalian Dicer has a dual function, being essential for both siRNA and miRNA biogenesis.

138
To determine whether these differences in viral susceptibility were due to the activity of Dicer 139 on siRNA or miRNA production, we compared Dicer -/cells with ESCs lacking the essential nuclear factor for miRNA biogenesis, Dgcr8. The absence of Dgcr8 also decreased TMEV and IAV viral susceptibility, suggesting that miRNAs are responsible for suppressing the antiviral 142 response in ESCs (Figure 2A-B). Interestingly, Dgcr8 -/cells were more resistant to virus 143 infection than Dicer -/cells, which supports a dual function for Dicer by also acting as a direct 144 antiviral factor targeting viral transcripts for degradation by RNAi. To rule out the possibility 145 of morphological differences influencing viral susceptibility, we performed a virus binding and 146 entry assay which showed no differences (Suppl. Figure 2).

147
Even though ESCs lack an IFN response (Figure 1), we wondered whether the differential 148 resistance to viral infections were the result of abnormal IFN activation due to the absence of 149 miRNAs. To test this hypothesis, we transfected the dsRNA analogue, poly(I:C) and G3-YSD in Dgcr8 or Dicer deficient mESCs, and quantified Ifnb1 expression by RT-qPCR. ESCs

154
These results show there is a correlation between viral susceptibility and the ability of miRNA-155 deficient ESCs to express Ifnb1, and that miRNAs are responsible for silencing the IFN 156 response to dsRNA. To verify that the observed results are solely due to the absence of ( Figure 2E-F).

162
To understand where the IFN pathway is silenced in ESCs we blocked the interferon response 163 at defined points in the pathway and measured viral susceptibility. The inhibitor BX795 blocks TBK1/IKKε phosphorylation and consequently IRF3 transcriptional activity, whereas 165 BMS345541 is an inhibitor of the catalytic subunits of IKK and thus blocks Nf-κB-driven 166 transcription. Both transcription factors are essential for the expression of Ifnb1 and other pro-167 inflammatory cytokines and initiation of an antiviral response (Lawrence, 2009, Schafer et al., 168 1998. Both inhibitors increased viral susceptibility in wild type cells lines, however, the effect 169 was far greater in the knock out cell lines (Figure 3A and Suppl. Figure 4A), suggesting that 170 miRNAs regulate the interferon pathway upstream Ifnb1 transcription.

171
We next aimed to identify, in an unbiased manner, differentially expressed proteins involved 172 in viral susceptibility in the presence or absence of miRNAs. To this end, the total proteome of  (Scaduto and Grotyohann, 1999), confirmed 179 lower oxidative phosphorylation activity in the absence of miRNAs (Dgcr8 -/and Dicer -/-) 180 (Suppl. Figure 4B). A search for differentially expressed proteins involved in the IFN response 181 did not reveal any significant changes except for the Mitochondrial antiviral-signalling protein 182 (MAVS), which in contrast to many other mitochondria-related proteins, was upregulated in 183 the absence of miRNAs. This protein has a central role in the RLR-induced (Rig-I-like 184 receptors) interferon pathway, where activated MDA5 and RIG-I receptors translocate to the 185 mitochondria and bind MAVS to ultimately induce Ifnb1 expression (Kawai et al., 2005).

186
Western blot and qRT-PCR analysis confirmed that MAVS was the only factor consistently 187 expressed to higher levels in both miRNA-deficient cell lines, Dgcr8 -/and Dicer -/- (Figure 3C, lanes 2 and 5, and Suppl. Figure 4C), compared to a panel of other components of the same innate immune response pathway (Suppl. Figure 4D).
To confirm the involvement of miRNAs on MAVS expression, a dual luciferase assay system was used where the 3'UTRs of MAVS, MDA5 and RIG-I were fused to a luciferase reporter gene to compare luciferase activity in wild-type and knock-out ESCs. Only the MAVS 3'UTR  shown that human MAVS is regulated by miR-125a, miR-125b and miR-22 (Hsu et al., 2017, 209 Wan et al., 2016. However, only miR-125a-5p and miR-125b-5p have conserved binding sites  Figure 5B).

222
showed the largest effect on MAVS protein expression both when depleted and overexpressed,

223
we hypothesize that miR-673 is a crucial miRNA involved on MAVS regulation.

224
We further investigated the role of miR-673-5p in ESCs by creating stable knock-out cell lines 225 for miR-673 by CRISPR/Cas9. Three cell lines were selected based on the genomic deletion 226 and confirmed undetectable expression of miR-673-5p ( Figure 5D and Suppl. Figure 6A).

227
The absence of miR-673-5p was enough to observe an increase in MAVS expression both at 228 the mRNA and protein levels ( Figure 5E and Suppl. Figure 6B). In addition, we measured

234
Next, miR-673 deficient cell lines were tested for TMEV susceptibility, which showed a 235 consistent decrease in virus replication, similar to that observed in the absence of all miRNAs (Dgcr8 -/-), suggesting this miRNA is essential in regulating the innate antiviral response in ESCs ( Figure 5F).
the post-transcriptional regulation of MAVS expression by miR-673-5p.

Discussion
Several studies suggest that the pluripotent state of a cell is incompatible with an active 242 interferon response (Guo et al., 2015). Both mouse and human stem cells fail to synthesize 243 interferons in response to dsRNA (Wang et al., 2013, Chen, Yang andCarmichael, 2010),

244
implying that this characteristic is acquired during differentiation (D' Angelo et al., 2016).

245
Embryonic carcinoma cells, which are still pluripotent, also fail to produce interferons in 246 response to viral RNA mimics (Burke, Graham and Lehman, 1978). In agreement,

247
reprogramming of somatic cells to iPSCs (induced pluripotent stem cells) leads to a loss of interferon response, suggesting the presence of regulatory mechanisms able to switch this 249 antiviral pathway on or off between the differentiated and pluripotent states (Chen et al., 2012).

250
Another feature of pluripotent cells is their attenuated response to exogenous type-I interferons.

251
Mammalian pluripotent stem cells, iPSCs and embryonic carcinoma cells exhibit an attenuated 252 production of interferon-stimulated genes upon type-I IFN stimulation (Hong and Carmichael, 253 2013, Irudayam et al., 2015, Wang et al., 2014, Burke, Graham and Lehman, 1978. Why 254 these activities are supressed is still not understood, but it has been hypothesized that type-I 255 IFN stimulation could impair their self-renewal capacity, since these compounds are well-256 known antiproliferative agents and inducers of cell death (Bekisz et al., 2010). Indeed, type-I contrast to most somatic cell types that silence their expression (Yin, Zhou and Yuan, 2018).

274
These repetitive elements produce cytoplasmic RNA molecules as an intermediate for 275 mobilisation, which can be accidentally recognised as immunogenic or non-self RNAs, as it 276 has been previously shown for the human non-LTR retroelement Alu in the context of Aicardi-

281
Cells that are incapable of activating the RNA-mediated IFN response have developed 282 alternative antiviral defence pathways. The endonuclease Dicer can act as an antiviral factor in 283 mouse ESCs by generating antiviral siRNAs (Maillard et al. 2013). Detection of antiviral Dicer activity is facilitated in the absence of a competent IFN response, such as in the case of impaired (Maillard et al., 2016). These findings are supported by the observation that in IFNcompetent cells, the RNA sensor LGP2 acts as an inhibitor of Dicer cleavage activity on 288 dsRNA (van der Veen et al., 2018). However, Dicer activity has also been reported in other 289 cell lines, independently of their IFN-proficiency capacity (Li et al., 2016). Interestingly, when 290 we disrupt Dicer in ESCs, which inherently lack an IFN response and would theoretically 291 render these cells highly sensitive to viral infections, they become more resistant by acquiring 292 an active IFN response. All these results support the presence of extensive cross-talk between 293 the different antiviral strategies, and suggests that cells have developed mechanisms to 294 compensate for the loss of a specific antiviral pathway.

295
Our model shows that MAVS and miR-673 levels are the key factors regulating the IFN 296 response to dsRNAs during pluripotency. Accordingly, overexpressing MAVS or knocking-297 out this single miRNA in ESCs is enough to enhance their antiviral response. Interestingly,

298
this miRNA is only conserved in rodents, despite human ESCs also suppressing type-I IFNs 299 expression (Hong and Carmichael, 2013). This suggests that either other miRNAs regulate 300 MAVS expression in human ESCs, or alternative mechanisms operate to silence IFN.

301
Interestingly, human ESCs constitutively express a subset of Interferon stimulated genes to 302 protect them from viruses (Wu, et al., 2018), but whether miRNAs control the expression of 303 Ifnb1 or this subset of ISGs in this context remains an unexplored matter.
Previous findings also support a general role for DICER and miRNAs acting as negative 305 regulators of the IFN response in human and mouse models outside pluripotency 306 (Papadopoulou et al., 2012, Witteveldt, Ivens andMacias, 2018). In agreement, an indirect virus, showed that miRNAs are also relevant to control the expression of pro-inflammatory 309 cytokines during viral chronic infections, but not in the acute antiviral response (Aguado et al., 310 2015). However, the concept of miRNAs acting as direct antiviral factors is still controversial.

311
It is relevant to mention that some of the results leading to this conclusion have been primarily 312 generated in Dicer -/-HEK293T human cell line (Bogerd et al., 2014, Tsai et al., 2018, which 313 has an attenuated IFN response due to low PRRs expression (Rice et al., 2014, Witteveldt, 314 Ivens and Macias, 2018).

315
We have shown that overexpression of MAVS or silencing specific miRNAs in a transient or     Table S1.

393
For the miRNA mimics (miScript, Qiagen) a final concentration of 1 mM was transfected into cells using Dharmafect (Dharmacon), incubated for the desired period and further processed.

395
The same procedure was followed for the antagomirs (Dharmacon), but at a concentration of 396 100nM. All experiments were performed in 24-well format, with cells at approximately 80% 397 confluency.

399
Total RNA from cells was isolated using Tri reagent (Sigma-Aldrich) according to the 400 manufacturer's instructions. 0.5-1 µg RNA was subsequently reverse transcribed using M-time PCR machine (ThermoFisher) using GoTaq master mix (Promega). Data was analysed 403 using the StepOne software package. Oligonucleotides used are listed in Table S1.

405
Cells used for Western blot analysis were lysed in RIPA buffer (50 mM TRIS-HCl, pH 7.4,

448
Five µg of peptides were injected onto a C18 packed emitter and eluted over a gradient of 2%-were ionised at +2kV before data-dependent analysis on a Thermo Q-Exactive Plus. MS1 was

483
To create a cell line lacking mmu-mmiR-673-5p, the Alt-R® CRISPR-Cas9 System (IDT) was 484 used. Two different crRNAs were designed to target sequences within the pri-miRNA sequence 485 hairpin to induce structural changes disrupting processing by the Microprocessor and Dicer.

486
Cas9 protein and tracrRNAs were transfected with the Neon® Transfection System followed 487 by cell sorting to create single cell clones. Genomic DNA was purified and screened by PCR 488 followed by restriction site disruption analyses for the pri-miRNA sequence. Genomic DNA 489 of the pri-miRNA sequence of candidates was amplified using primers in Table S1, and cloned 490 into pGEMt-easy vector for sequencing.

492
Total RNA (100ng) was used to quantify mmu-mmiR-673-5p levels. RNA was first converted 493 to cDNA using miRCURY LNA RT kit (Qiagen). cDNA was diluted 1/25 for RT-qPCR using 494 miRCURY LNA SYBR Green kit and amplified using mmu-mmiR-673-5p specific primers (Qiagen) and U6 as a loading control. Quantitative PCR was carried out on a Roche LC480 light cycler and analysed using the second derivative method.

498
All processed Mass spectrometry data is provided as a Supplementary Excel File, including

499
LFQ intensity values for each protein detected in each of the samples. All raw data are available 500 from corresponding author upon request.

502
We thank our colleagues at the Institute of Immunology and Infection Research for advice and

510
The manuscript was co-written by all authors.