Susceptibility to Zika virus in a Collaborative Cross mouse strain is induced by Irf3 deficiency in vitro but requires other variants in vivo

Zika virus (ZIKV) is a Flavivirus responsible for recent epidemics in Pacific Islands and in the Americas. In humans, the consequences of ZIKV infection range from asymptomatic infection to severe neurological disease such as Guillain-Barré syndrome or fetal neurodevelopmental defects, suggesting, among other factors, the influence of host genetic variants. We previously reported similar diverse outcomes of ZIKV infection in mice of the Collaborative Cross (CC), a collection of inbred strains with large genetic diversity. CC071/TauUnc (CC071) was the most susceptible CC strain with severe symptoms and lethality. Notably, CC071 has been recently reported to be also susceptible to other flaviviruses including dengue virus, Powassan virus, West Nile virus, and to Rift Valley fever virus. To identify the genetic origin of this broad susceptibility, we investigated ZIKV replication in mouse embryonic fibroblasts (MEFs) from CC071 and two resistant strains. CC071 showed uncontrolled ZIKV replication associated with delayed induction of type-I interferons (IFN-I). Genetic analysis identified a mutation in the Irf3 gene specific to the CC071 strain which prevents the protein phosphorylation required to activate interferon beta transcription. We demonstrated that this mutation induces the same defective IFN-I response and uncontrolled viral replication in MEFs as an Irf3 knock-out allele. By contrast, we also showed that Irf3 deficiency did not induce the high plasma viral load and clinical severity observed in CC071 mice and that susceptibility alleles at other genes, not associated with the IFN-I response, are required. Our results provide new insight into the in vitro and in vivo roles of Irf3, and into the genetic complexity of host responses to flaviviruses.


Introduction
Zika virus (ZIKV) is a mosquito-borne virus of the Flaviviridae family identified in 1947 in Uganda.The first noticeable human outbreaks occurred in Micronesia in 2007 and in French Polynesia and New Caledonia in 2013-2014.In 2015-2016, ZIKV caused an epidemic in Brazil which rapidly spread across the Americas and the Caribbean.To date, 89 countries have reported evidence of mosquito-transmitted Zika virus infection (https://www.who.int/healthtopics/zika-virus-disease#tab=tab_1).
While most people infected with ZIKV remain asymptomatic, some develop non-specific symptoms including rash, fever, conjunctivitis, muscle and joint pain, malaise and headache.Neurological complications have been described in adults such as Guillain-Barre ´syndrome [1] and encephalitis [2].Infection of pregnant women was associated with congenital Zika syndrome in the fetus, which can lead to neurodevelopmental deficiencies, brain malformation [3] or in some cases to fetal loss [4].
Many factors may contribute to this variable severity, including the viral strain, the infection route and dose, and the host genetic background [5,6].Indeed, mouse and human studies have shown that host genes influence flaviviral infections' outcomes [7].While human genetic studies are hampered by the variability of these multiple factors, they can be controlled in mouse models which have proven very valuable to identify susceptibility variants [8,9].Relevant ZIKV infection models have been developed in mice either using Ifnar1 knock-out (KO) mice in which the IFNAR receptor to IFN-I has been inactivated [10,11], or by blocking this receptor using a monoclonal antibody targeting the IFNAR1 receptor subunit (MAR1-5A3 [12]).
We have previously explored the role of mouse natural genetic variants on ZIKV susceptibility in the Collaborative Cross (CC), a panel of recombinant inbred mice encompassing a genetic diversity similar to that of the human population and capturing approximately 90% of the mouse natural genetic variants [13,14].We reported that the CC genetic diversity enabled large variations in the clinical severity of ZIKV disease, plasma viral load and intensity of brain pathology, comparable to those observed in human cases [15].We specifically identified CC071/TauUnc (CC071) mice as very susceptible, with high mortality and high peak plasma

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo viral load.Notably, the CC071 strain has been recently reported as also susceptible to other flaviviruses including dengue virus [15], Powassan virus [16], West Nile virus (WNV) [15], and to Rift Valley fever virus [17], emphasizing its value to decipher genetic factors controlling host responses to RNA viruses.
We previously demonstrated that genetic background influenced ZIKV replication in CC mouse embryonic fibroblasts (MEFs).Here, we investigated the mechanisms driving high viral replication in CC071 MEFs.We found that, compared with the more resistant C57BL/6J (B6) and CC001/Unc (CC001) strains, ZIKV-infected CC071 MEFs displayed a delayed expression of the IFN-I response genes.Genetic and functional analyses identified a strain-specific variant in the Irf3 gene, the first transcription factor involved in interferon (IFN) β expression, which mimics the effects of a null allele in vitro and fully explains the delayed IFN-I expression and uncontrolled viral replication.By contrast, we showed, from the in vivo comparison of CC071, Irf3-deficient and backcross mice, that the CC071 Irf3 mutation is not sufficient to explain the high susceptibility of CC071 mice to ZIKV infection and that other genes, not associated with the IFN-I response, are involved.These findings provide new insights into the roles of Irf3 in viral diseases and exemplify how the study of CC strains allows deciphering the role of host genes in viral pathogenesis.

CC071 MEFs show defective control of viral replication and delayed IFN-I expression, but normal response to IFN-I stimulation
We previously reported that, unlike CC001 MEFs, CC071 MEFs produced increasing quantities of viral particles during the first 72 hours post-infection (hpi) [15].Here, we confirmed and expanded this observation by infecting B6, CC001 and CC071 MEFs and by quantifying viral particles by FFA.After ZIKV infection, CC071 MEFs displayed high and increasing viral titers between 24 and 72 hpi, while CC001 and B6 MEFs showed stable and lower titers (Fig 1A).
To investigate the origin of the defective control of viral replication in CC071 MEFs, we measured the expression level of the Ifnb1 gene coding for IFNβ in ZIKV-infected CC071, CC001 and B6 MEFs.Ifnb1 expression is induced very rapidly after virus detection by sensors and triggers the innate antiviral response which is essential for limiting viral replication.In CC001 and B6 MEFs, Ifnb1 expression was significantly induced at 24 hpi and remained stable and high until at least 72 hpi (Fig 1B).In contrast, its expression in CC071 MEFs was low at 24 hpi and reached the level of CC001 only at 72 hpi.Similar results were obtained for Ifna4 which encodes one of the IFNα proteins (S1 Fig) .Notably, Ifnb1 expression in CC071 MEFs at 72 hpi was significantly higher than in B6 MEFs, showing that CC071 MEFs were delayed but not intrinsically hampered in their ability to induce strong IFN-I expression.
To test whether this defective induction of Ifnb1 expression was specific to ZIKV infection, MEFs were then transfected with the influenza A virus-derived 3-phosphate-hairpin-RNA (3p-hpRNA), an agonist of the RIG-I/MDA5-MAVS pathway, or treated with polyinosinepolycytidylic acid (poly (I:C)), that activates both Toll-like receptor (TLR) 3 (TLR3) and the RIG-I/MDA5-MAVS pathway [18].Here again, CC071 MEFs showed a delayed expression of Ifnb1 after both stimulations by comparison with B6 and CC001 MEFs (S2 Fig) , indicating that the defect in IFN-I genes expression in CC071 MEFs was not specific to ZIKV infection.This result suggested a defect in the molecular cascade between cellular sensors of pathogenassociated molecular patterns (PAMP) and the Ifnb1 gene transcription machinery.
To evaluate the capacity of CC071 MEFs to respond to IFN-I, they were treated with recombinant IFNα.The expression of IFN-stimulated genes (ISGs) such as Ifitm3 was induced with

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo

CC071's delayed Ifnb1 expression is strain-specific
To gain insight into the mechanisms responsible for defective Ifnb1 induction, we investigated the expression levels of genes involved in Ifnb1 expression on CC071, B6 and CC001 MEFs at 16, 24 and 32 hpi.Mock-infected MEFs were analyzed at 24 hours as controls.Expression levels were measured by RNA sequencing (RNAseq) which provided a comprehensive analysis of transcriptomic changes.In CC001 MEFs, the expression of many genes rapidly increased after infection (160 at 16h, 821 at 24h and 971 at 32h; log2 fold-change > 1, FDR = 0.05), reflecting a robust innate antiviral response (S3A Fig) .A similar pattern was observed in B6 MEFs.By contrast, the expression of only 38 genes was increased in CC071 MEFs at 32hpi (34 of which were also activated in CC001), consistent with the delayed induction of Ifnb1 expression.Among the genes that are involved in the pathway between PAMP sensors and Ifnb1 transcription, ISGs such as Tlr3, Ddx58 (coding for RIG-I sensor) or Irf7 were not induced upon infection in CC071, while constitutively expressed genes such as Mavs, Ticam1 (coding for the TRIF adaptor), Traf3 or Irf3, showed comparable levels of expression in the three strains (S3B Fig) .Therefore, this analysis did not provide new clues for identifying the gene responsible for the defect observed in CC071.
We then leveraged the genetic architecture of the CC which genomes are patchworks of haplotypes inherited from the eight founder strains [19].Although CC071 was the only strain with severe ZIKV disease, we hypothesized that, if the delayed activation of Ifnb1 resulting in uncontrolled viral replication observed in CC071 MEFs was due to an allele at one of the genes involved in the Ifnb1 induction pathway inherited from a parental strain, ZIKV-infected MEFs of CC strains carrying the same allele would present similarly high viral titers.We therefore derived MEFs from each CC strain available to us carrying the same ancestral haplotype as CC071 at one of the 13 genes of the pathway (Fig 2A).Upon ZIKV infection, none of these CC MEFs showed viral titer kinetics resembling that observed in CC071 MEFs (Fig 2B).These results suggested two alternative hypotheses.Either the delayed Ifnb1 activation involved two members of the pathway with a CC071-specific allelic combination leading to a non-functional interaction, or CC071 was carrying a strain-specific allele at one of these genes, resulting from a CC071-specific mutation that probably arose on an ancestral haplotype during the CC071 inbreeding.However, the sequencing of one male of each CC strain reported in 2017 [20] did not identify such "private" variants with high predicted impact in CC071 for any of these genes.Whatever the molecular mechanism, our results indicated that it was specific to CC071.

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo suggested the insertion of a repeated sequence between exons 6 and 7. Nevertheless, the functional consequence of this mutation was confirmed by Western blot using a specific C-terminal IRF3 antibody which showed that full-length IRF3 protein was absent in CC071 MEFs (Fig 5C).Moreover, immunofluorescence using an antibody directed against phosphorylated IRF3 detected a positive signal in the nucleus of many ZIKV-infected CC001 and B6 MEFs, but not

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo in CC071 MEFs (Fig 5D).Altogether, these results show that CC071 carries a mutation in Irf3 that prevents IRF3 phosphorylation which is required to induce Ifnb1 expression.Whether the altered mRNA sequence prevents the production of the protein or alters its activity, this mutation results in the loss of IRF3 transcriptional function.

CC071 Irf3 mutation is responsible for uncontrolled viral replication in MEFs
To test if the delayed Ifnb1 expression resulting in uncontrolled viral replication in CC071 MEFs was caused exclusively by the Irf3 mutation, we performed a quantitative complementation test by producing compound heterozygous MEFs carrying a knockout Irf3 allele (Irf3 KO ) and a CC071 allele (Irf3 71 ).These Irf3 KO/71 MEFs were compared with CC071 and B6-Irf3 KO/ KO MEFs, and with heterozygous MEFs carrying a B6 wildtype allele and either an Irf3 KO  ), B6-Irf3 KO/KO and Irf3 KO/71 MEFs carrying two defective alleles at Irf3 showed similar results (Fig 6B and 6C).These data demonstrate that, since the Irf3 KO did not complement the Irf3 71 allele, the Irf3 mutation in CC071 contributes to the defects observed in ZIKV-infected MEFs.Moreover, since the data obtained on CC071 and on Irf3 KO/71 MEFs were identical, we conclude that the CC071 Irf3 mutation is sufficient to induce the defects observed in CC071 MEFs.

CC071 Irf3 mutation is not sufficient to explain susceptibility in vivo
We investigated whether this Irf3 mutation was also responsible for the high susceptibility to ZIKV of CC071 mice.We first evaluated its effects in a context allowing IFN-I response.To this aim, we compared B6, B6-Irf3 KO and CC071 mice infected without prior MAR1-5A3 treatment, which results in moderate and short-lasting plasma viral loads without clinical signs, as we previously reported [15].Similar viral loads were observed in B6 and B6-Irf3 KO mice at all days p.i. (dpi), while CC071 mice showed significantly higher viral loads at days 1 and 2 p.i. (Fig 7A).This result indicates that, unlike in MEFs, the Irf3 null mutation does not induce elevated viral replication in vivo (as measured by plasma viral load) and suggests that CC071 mice carry susceptibility alleles at other loci.
We then blocked IFN-I response by pre-treating mice with MAR1-5A3 and compared peak plasma viral load and clinical signs in ZIKV-infected B6, B6-Irf3 KO and CC071 mice.While no differences were observed between B6 and B6-Irf3 KO mice (moderate plasma viral loads without clinical signs), as expected if the IFN-I response is neutralized, CC071 mice developed significantly higher viral loads at day 2 p.i. and signs of disease (ruffled fur, hunched posture and body weight loss) around 7 days p.i. (Fig 7B).This contrast could result either from functional differences between the Irf3 KO and Irf3 71 alleles or from CC071 alleles at other genes not linked with the IFN-I response.F1 mice between CC071 and B6-Irf3 KO mice gave similar results as B6-Irf3 KO mice, suggesting that CC071 susceptibility alleles were recessive.Therefore, we generated a small cohort of (CC071 x B6-Irf3 KO) x CC071 backcross mice and genotyped them for the Irf3 gene by PCR using primers that amplify intron 6 of the Irf3 KO allele, but not of the Irf3 71 allele.(Figs 5A and S5).We found no differences in plasma viral load and clinical scores between Irf3 KO/71 and Irf3 71/71 backcross mice, indicating that the two alleles are functionally equivalent in vivo.Moreover, Irf3 KO/71 backcross mice were more severely affected than Irf3 KO/71 F1 mice, indicating the strong effects of recessive alleles from CC071 which likely explain the variability of plasma viral load and clinical signs among backcross mice.Altogether, our results demonstrate that the CC071 mutation is a loss-of-function (LOF) variant which does not confer susceptibility to ZIKV disease in the absence of other recessive alleles carried by CC071 at genes not associated with the IFN-I response.

Discussion
An increasing number of publications have reported how the genetic diversity provided by the CC translates into an extended range of resistant and susceptible phenotypes in various infectious models.Since our first description of the high susceptibility of CC071 mice to ZIKV https://doi.org/10.1371/journal.ppat.1011446.g006

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo infection, with disease severity and peak plasma viral load almost as high as those of B6-Ifnar1 KO mice, and higher clinical signs and mortality than 129-Ifnar1 KO mice [15], several groups have reported similar observations for other viruses.High viral load was found in CC071 mice after dengue infection [15].Lethality after infection with West Nile virus [15] or Powassan virus [16] were also described, although there is likely contribution from the defective Oas1b allele that CC071 has inherited from 129S1/SvImJ.CC071 was also one of the most susceptible CC strains to Rift Valley Fever virus infection [17] and to hepacivirus with long-term viral persistence [21].Because of its unique genetic background, this strain will be increasingly useful for infectious diseases studies, which underlines the importance of deciphering the mechanisms of its susceptibility to each pathogen.
In this study, we leveraged from our previous observation that CC071 MEFs failed to control ZIKV replication by comparison with resistant strains.Although they cannot recapitulate the complex interactions between multiple pathways, cell types and tissues of a whole organism, MEFs are a convenient cellular model which can be easily derived from any mouse genetic background.We identified a delayed activation of the Ifnb1 gene in CC071 MEFs, resulting in delayed stimulation of ISGs.We then used a combination of genetic approaches to find the causative gene defect.The observation that MEFs were normally responsive to IFN-I stimulation was consistent with the higher susceptibility of mice treated by the MAR1-5A3 antibody compared with untreated mice which did not develop symptoms and showed much lower plasma viral load [15].These results pointed at the pathway from PAMP sensors to Ifnb1 transcription factors.Transcriptional analysis did not identify reduced expression of non-ISGs of this pathway.CC strains' genomic structure allows searching for haplotypes inherited from the same founder in CC strains showing similar phenotypes.However, we did not identify such haplotypes for genes of the Ifnb1 induction pathway, suggesting that the defects observed in CC071 MEFs were strain specific.It is finally the analysis of MEFs derived from backcross embryos that established a monogenic inheritance, and genetic linkage unambiguously pointed at the causative Irf3 gene.The MEF experimental model was particularly appropriate since we could derive cell lines from every backcross embryo.Our RNAseq data showed an RNA splicing defect in the Irf3 gene in CC071, which functional consequences could be validated in vitro.The formal proof that the CC071 Irf3 mutation was necessary and sufficient to cause Ifnb1 delayed activation and uncontrolled viral replication came from a quantitative complementation test in which the Irf3 KO allele was combined either with a CC071 or a B6 (functional) allele.Altogether, these results show that the Irf3 defect in CC071 abrogates its transcriptional activity.Unfortunately, none of the antibodies recognizing the N-terminal protein domain that we tested worked in our hands in western blot.Therefore, we could not establish whether the Irf3 mRNA detected by RNAseq is translated into an abnormal protein or not.However, the in vivo observation that Irf3 KO/71 and Irf3 71/71 backcross mice displayed the same phenotypes upon ZIKV infection shows that the two alleles are functionally equivalent in this context.
Holm correction for multiple testing (* p < 0.05).(B) B6, B6-Irf3 KO, CC071, (CC071 x Irf3 KO) F1 and (CC071 x Irf3 KO) x CC071 BC mice were infected IP with 10 7 FFUs of ZIKV after IP injection of 2 mg of MAR1-5A3 IFNARblocking monoclonal antibody 24 h before infection, and monitored for 7 days.The graph shows plasma viral loads quantified at day 2 p.i. by RT-qPCR.The two lines of the group legend (X-axis) indicate the genetic background of the mice and their genotype at the Irf3 locus, respectively.BC mice are separated into two groups depending on the genotype at the Irf3 gene: homozygous for the CC071 mutant allele (Irf3 71/71 ), or heterozygous for the CC071 and the KO alleles (Irf3 KO/71 ).Irf3 genotyping results are presented in S5 Fig.Each dot represents one mouse.Groups were compared by ANOVA followed by Tukey HSD (* p < 0.05, ** p < 0.01, *** p < 0.001).Below the graph is shown the distribution of clinical scores at day 7 p.i. in the same groups of mice as above (0, no symptoms; 1, slight hunched posture; 2, ruffled fur, hunched posture and/or mild ataxia; 3: prostration, ataxia, partial paralysis).https://doi.org/10.1371/journal.ppat.1011446.g007

PLOS PATHOGENS
Irf3 maps to a chromosomal region that CC071 inherited from CAST/EiJ, like the resistant CC001 strain (and CC046/Unc, not available to us).The CC071-specific mutation may have arisen during the inbreeding generations leading to this strain.Alternatively, the mutation might have been segregating in the CAST/EiJ founders and transmitted by chance only to the CC071 breeding funnel.Since all CAST/EiJ founders used in CC crosses originated from The Jackson Laboratory colony, genotyping past and present breeders of this colony could not only unravel the origin of the mutation but also allow eliminating this mutation to avoid unwanted effects.
Notably, this is not the first example of a CC-strain-specific mutation.We [22] and others [23] previously reported the extreme susceptibility of CC042/GeniUnc (CC042) to Salmonella Typhimurium and to Mycobacterium tuberculosis, respectively, as a consequence of a de novo 15-nucleotide deletion in the Itgal1 gene.In the case of Salmonella Typhimurium, CC042 was standing out, with bacterial loads up to 1000 times higher than other CC and the susceptible B6 strains.In our previous study on ZIKV, CC071 was the most susceptible strain, but its peak viral load was just the highest in a continuous distribution of values.Additionally, this is not the first example of a Irf3 spontaneous variant identified in mice modulating susceptibility to bacterial [24] or viral infections [25].
Irf3 is an important transcription factor involved in the innate immune response.It is constitutively expressed and, at rest, inactive IRF3 is present in the cytoplasm.Upon viral entry (or other stimuli that activate TLRs such as TLR3 and 4 or RIG-I-like receptors), signal transduction leads to the phosphorylation of IRF3, resulting in its dimerization and translocation to the nucleus where it binds to the Ifnb1 promoter [26].This mechanism leads to the very fast production of IFNβ which is secreted by the cell and triggers the immediate response to viral infection through the activation of ISGs with diverse antiviral functions [27].In addition to its transcriptional activity, IRF3 has been shown to counter viral infection by two additional mechanisms.IRF3 induces apoptosis through the RLR-induced IRF3-mediated pathway of apoptosis (RIPA) in infected cells [28] and inhibits NF-κB-mediated inflammation through the repression of IRF3-mediated NF-κB activity (RIKA) [29].These newly described roles of IRF3 depend on other protein domains than those involved in its transcriptional activity.Whether these activities are also abrogated by the CC071 Irf3 mutation remains to be established.
In mice, studies that used Irf3-deficient cellular models have consistently reported decreased IFN-I production and/or increased viral replication.For example, after infection with WNV, viral replication was increased in B6-Irf3 KO bone marrow macrophages (BMMs) and moderately in primary neurons [30].Higher viral replication was observed in herpes simplex virus 1 (HSV-1)-infected B6-Irf3 KO bone marrow-derived dendritic cells (BMDCs) and BMMs, and IFNβ was reduced in BMDCs supernatants [31].Similarly, Ifnb1 expression was reduced in HSV-1-infected [32] and CHIKV-infected [33] B6-Irf3 KO MEFs.In line with these findings, our study provides, to our knowledge, the first evidence for a role of Irf3 in the infection of murine cells by ZIKV.Moreover, we recently reported that primary cultured neurons (PCNs) show a delayed activation of the Ifnb1 expression upon ZIKV exposure compared with MEFs, and that this delay is even longer in PCNs derived from CC071 compared with CC001 [34].The identification of the Irf3 mutation in CC071 provides an explanation for this observation, although it would require confirmation in B6-Irf3 KO PCNs.
In contrast, in vivo studies have reported inconsistent consequences of Irf3 deficiency between viral infections.Here, we did not observe any differences in clinical signs nor in plasma viral loads between B6 and B6-Irf3 KO mice (pre-treated or not with the MAR1-5A3 antibody), consistently with a previous study that reported neither mortality, nor body weight loss after ZIKV infection [10].Likewise, B6-Irf3 KO mice have been reported to show no mortality and low virus load in the circulation following dengue virus [35] and CHIKV infection [33].Contrastingly, WNV infection was lethal in all B6-Irf3 KO mice with increased viral burdens in peripheral and central nervous system tissues, while 65% of infected WT mice survived [30].Differences were also associated with the route of infection.For example, all WT and B6-Irf3 KO mice survived after intravenous inoculation with HSV-1 [32], while intranasal inoculation led to 90% mortality in B6-Irf3 KO but only 30% in WT mice [36].In humans, LOF mutations in Irf3 have been associated with increased susceptibility to diseases caused by WNV [37,38], HSV-1 [39,40] and more recently by SARS-CoV-2 [41].Discrepancies between the consequences of Irf3 deficiency in in vitro and in vivo experiments likely reflect differences between a single cell type model with an intrinsic capacity to mount a more or less IRF3-dependent IFN-I response, and a multicellular organism in which the paracrine effect of IFN-I produced by the most reactive cells induces protection of other, less reactive, cell types.In fact, we have recently reported such cooperation between neurons and microglia cells [34].
Our in vivo data, showing high plasma viral loads and clinical signs in all CC071 and half of backcross mice after MAR1-5A3 treatment, demonstrated that the high susceptibility of CC071 mice requires the contribution of other recessive alleles.Since their effect was observed in mice which IFNAR receptor had been blocked, we conclude that their mode of action is not dependent on an intact IFN-I response.Moreover, the comparison between the results of the complementation test on MEFs, with CC071 and Irf3 KO/71 yielding very similar data, and the in vivo observation that CC071 showed higher viremia than Irf3 KO/71 F1 mice, suggests that these other susceptibility alleles may not act at the cell level, at least during the first 72 hpi, but rather at a more integrated or systemic level.Whether or not these alleles require Irf3 deficiency to induce the susceptibility observed in the CC071 mice has yet to be determined.While our limited understanding of the severe disease developing in CC071 ZIKV-infected mice does not point at potential mechanisms, identifying these alleles will allow disentangling the genetic factors controlling ZIKV and other viruses' pathogenesis.

ZIKA virus
The FG15 Asian Zika virus (ZIKV) strain, isolated from a patient during ZIKV outbreak in French Guiana in December 2015, was obtained from the Virology Laboratory of the Institut Pasteur of French Guiana.Viral stock (passage 5) was prepared from supernatant of infected C6/36 cells, clarified by centrifugation at 800g and titrated on Vero cells by focus-forming assay.

Mouse infection
All infection experiments were performed in a biosafety level 3 animal facility and mice were kept in isolators.Six-to 10-week-old, male or female mice were injected intraperitoneally with 10 7 PFU of ZIKV FG15.In some in vivo experiments, mice received an IP injection of 2 mg of MAR1-5A3 anti-IFNAR antibody (Euromedex, Cat#BX-BE0241) one day prior infection.Mouse numbers are indicated in figure legends.Both males and females were used since no differences between sexes were detected in our previous and present experiments.Clinical signs and body weight loss were recorded for up to seven days post infection.Blood samples were collected on EDTA from the retromandibular vein for plasma viral load assessment.Quantification of ZIKV viral copies by RT-qPCR was previously described [15].

MEFs isolation
Pregnant females were euthanized at day 13.5-15.5 of gestation.For B6, CC001, CC071 and crosses used for genetic mapping and the complementation test, MEFs were isolated from individual fetuses to obtain biological replicates.For other CC strains, MEFs were derived from individual or pooled fetuses.Fetus bodies were chopped and digested with trypsin (Gibco Cat#25300054), then cultured at 37˚C and 5% CO 2 in complete medium (DMEM Gibco Cat# 31966047, 10% fetal bovine serum PAA Laboratories Cat#A15-101, 1% penicillin/ streptomycin Sigma Cat#P4333).MEFs were used until passage 2. For the backcross experiment, MEF lines were isolated from 51 backcross fetuses.Heads were used to prepare DNA for whole-genome genotyping.

MEFs infection
MEFs were seeded at 5x10 4 cells per well in 24-well plates the day before infection.They were exposed to ZIKV FG15 strain at a MOI of 5 for 2 hours after which the inoculum was replaced with fresh complete medium and MEFs were incubated for up to 72 hours.For kinetics studies, different wells were used for each time point.Backcross MEFs were infected in 6 infection experiments, each of which included one CC001 and one CC071 MEF lines.

MEFs IFNα stimulation
MEFs were seeded at 5x10 4 cells per well in 24-well plates one day before stimulation and treated with 300 IU/mL IFNα (Miltenyi Biotec Cat#130-093-131) and incubated for up to 24 hours.

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo 4-12% Bis-Tris gels (Invitrogen Cat#NP0323BOX) in MOPS buffer (Invitrogen Cat#NP0001) and transferred to nitrocellulose membranes (Bio-Rad Cat#1620112) in a 25mM Tris 200mM glycine 20% ethanol buffer.Blots were blocked in 5% milk in TBS-T (0.1% Tween20 in Tris Base Sodium), incubated with primary antibodies diluted in 3% milk in TBST overnight at 4˚C, and incubated with secondary antibodies diluted in 3% milk in TBST for 90 min.Blots were revealed with ECL substrate (Thermo Scientific Cat#32132) and imaged with X-ray films.Primary and secondary antibodies are indicated in S2 Table.

Statistical analysis
Statistical analyses were performed using R version 4.1.0[43].Viral titers, gene expression and genome copies were log-transformed for graphs and statistical tests.One way ANOVA followed by Tukey HSD were used for testing multiple comparisons.For in vivo studies without MAR1-5A3 treatment, non-parametric Kruskal-Wallis followed by Wilcoxon tests with Holm correction for multiple testing were used to handle values below the limit of detection.Linear discriminant analysis (LDA) was conducted using the MASS package [44].The LDA was trained on the phenotypes of the two parental CC001 and CC071 strains from each infection batch.LDA coefficients were applied to backcross mice for assignment to "CC071-like" or "CC001-like" groups.

Genetic analysis
Raw genotypes were curated using the stuart package [45].QTL mapping was performed using R/qtl [46].LDA prediction was used as a binary trait.Statistical significance thresholds were computed by data permutation (n = 1000).95% confidence interval was estimated using the Bayesian method.

RNA sequencing
MEF RNA was prepared as described in the main text.RNA integrity and quantification were assessed using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies, CA, USA).One microgram of high-quality RNA samples (RIN > 9.2) representing biological triplicates were submitted to Novogene for RNA-sequencing (Novogene Beijing, China).Poly-A selected RNA was used for paired-end library preparation and transcriptome sequencing.Sequencing libraries were generated using NEBNext1 UltraTM RNA Library Prep Kit for Illumina1 (NEB, USA) following manufacturer's instructions.The library preparations were sequenced on an Illumina platform and paired-end reads were generated.The RNA-seq analysis was performed with Sequana 0.9.8 [47].In particular, we used the RNA-seq pipeline https://github.com/sequana/sequana_rnaseq)built on top of Snakemake 6.1.1 [48].Briefly, reads were trimmed from adapters using Cutadapt 2.7 then mapped to the Mus musculus genome assembly GCA_000001635.8 from NCBI using STAR 2.7.3a [49].FeatureCounts 1.6.4[50] was used to produce the count matrix, assigning reads to features using corresponding annotation v92 from NCBI with strand-specificity information.Quality control statistics were summarized using MultiQC 1.6 [51].Clustering of transcriptomic profiles were controlled using a Principal Component Analysis (PCA).Differential expression testing was conducted using DESeq2 library 1.24.0 [52] scripts indicating the significance (Benjamini-Hochberg adjusted p-values, false discovery rate FDR < 0.05) and the effect size (fold-change) for each comparison.Splicing analysis of the Irf3 gene was performed using Majiq 2.4 [53] with default parameters to investigate alternative transcripts between genotypes.

PLOS PATHOGENS
Irf3 deficiency induces susceptibility to ZIKV in vitro, not in vivo

Fig 1 .
Fig 1. CC071 MEFs fail to control viral replication, with delayed Ifnb1 expression but normal response to IFN-I.(A, B) MEFs derived from B6 (black circles), CC001 (blue triangles) and CC071 (red squares) were infected with ZIKV at a MOI of 5 and analyzed 24, 48 and 72 hpi.(A) Viral titer in supernatants was quantified by FFA.(B) Ifnb1 expression was normalized to Tbp reference gene.Data are mean +/-sem from 3 to 4 biological replicates per strain (MEFs derived from individual embryos).(C) MEFs were stimulated with recombinant IFNα.Ifitm3 relative expression normalized to Tbp reference gene is shown as an example of ISG.Data are mean +/-sem from 2 to 3

Fig 2 .
Fig 2. Haplotype analysis fails to identify a gene from the Ifnb1 induction pathway associated with uncontrolled viral replication.(A) Identification of CC strains which carry the same ancestral haplotype as CC071 at the genes involved in the pathway leading to Ifnb1 expression.Colored boxes indicate matched haplotypes between CC071 and other CC strains.Letters and colors designate the eight CC founder strains.A: A/J (yellow); B: C57BL/6J; C: 129S1/ SvImJ (pink); D: NOD/ShiLtJ; E: NZO/HILtJ (light blue); F: CAST/EiJ (green); G: PWK/PhJ (red); H: WSB/EiJ (purple).Doubled letters (eg AA) indicate homozygous genotypes.Heterozygous genotypes are indicated by the two corresponding letters (eg AD).(B) Kinetics of viral titer in MEFs from CC071 and the 8 CC strains shown in (A).Experimental conditions were as in Fig 1A n: number of technical replicates for each strain.Data are mean +/-sem from the technical replicates.Letters show statistical significance between CC071 and other strains.(ANOVA followed by Tukey HSD, a: p < 0.05, b: p < 0.01, c: p < 0.001).https://doi.org/10.1371/journal.ppat.1011446.g002

Fig 3 .
Fig 3. Backcross MEFs display either a CC001-like or a CC071-like phenotype.MEFs derived from CC001 (blue triangles), CC071 (red squares) and backcross (gray circles) embryos were infected with ZIKV at a MOI of 5. (A) Viral titer and (B) Ifnb1 expression in ZIKVinfected MEFs from 9 backcross embryos.Experimental conditions were as in Fig 1A and 1B.Red and blue curves show the results for CC071 and CC001 MEFs, respectively, from the same infection experiment.(C) Results of LDA on the backcross MEFs.LDA coefficients were calculated from Ifnb1 expression data in CC001 and CC071 infected MEFs, and applied to backcross MEFs.The graph shows the probability of each BC MEF to belong to the "CC001-like" group, resulting in two distinct populations shown in blue ("CC001-like") and in red ("CC071-like").n: number of BC MEFs assigned to each group.https://doi.org/10.1371/journal.ppat.1011446.g003

Fig 5 .
Fig 5. CC071 Irf3 mRNAs show abnormal splicing, resulting in defective IRF3 protein.(A) Schematic representation of the exons of the Irf3 gene with the number of reads spanning successive exons in the CC001 and CC071 RNAseq data (one sample of each strain).The red box between CC071's exons 6 and 7 depicts a novel exon resulting from abnormal splicing.(B) Schematic representation of the IRF3 protein structural domains (exon 1 is untranslated).Exon 8 encodes the serine rich region containing the phosphorylation sites for IRF3 activation.(C) Western blot using an anti-C-terminal IRF3 antibody from mock-infected and ZIKV-infected B6, CC001 and CC071 MEFs at 2 hpi, showing the absence of full-length IRF3 in CC071 MEFs.Vinculin was used as a loading control.(D) Immunofluorescence using an anti-phosphorylated IRF3 (pIRF3, green) in ZIKV-infected B6, CC001 and CC071 MEFs at 24 hpi, showing the absence of pIRF3 in the nucleus of CC071 MEFs upon infection.Red-labeled 4G2 antibody labels ZIKV-infected cells.Cell nuclear DNA labeled by Hoechst (blue).Quantification of the number of infected and pIRF3 positive cells is presented in the table.Proportions were established on 420, 428 and 551 cells for CC001, CC071 and B6, respectively.https://doi.org/10.1371/journal.ppat.1011446.g005 or an Irf3 71 allele (Irf3 +/KO or Irf3 +/71 , respectively, Fig 6A).While Irf3 +/KO and Irf3 +/71 ZIKVinfected MEFs showed the same pattern as CC001 or B6 MEFs (rapid induction of Ifnb1 expression and controlled viral replication, see Fig 1A for comparison), CC071 (Irf3 71/71

Fig 7 .
Fig 7. Irf3 LOF is not sufficient to explain CC071 susceptibility to ZIKV infection in vivo.(A) B6, B6-Irf3 KO and CC071 mice were infected IP with 10 7 FFUs of ZIKV and monitored for 7 days, without prior IP injection of MAR1-5A3 IFNAR-blocking monoclonal antibody.Plasma viral load was quantified at days 1 to 3 p.i. by RT-qPCR.The two lines of the group legend (X-axis) indicate the genetic background of the mice and their genotype at the Irf3 locus, respectively.Each dot represents one mouse.Groups were compared by Kruskal-Wallis followed by Wilcoxon test with