Persistent RNA virus infection is short-lived at the single-cell level but leaves transcriptomic footprints

Viruses are commonly classified according to the fate of the infected host cell. Infections with cytolytic viruses eventually culminate in cell death. In contrast, noncytopathic/noncytolytic viruses such as hepatitis B and C virus (HBV, HCV) in humans and lymphocytic choriomeningitis virus (LCMV) in mice have developed strategies to avoid such a fatal outcome, supposedly with the aim of facilitating long-term persistence in their respective hosts. Members of the herpesvirus family as well as HBV and human papillomaviruses switch to a minimal viral gene expression program referred to as latent infection, maintaining their genomes as episomal DNA (Lieberman, 2016). In contrast, RNA viruses such as HCV or LCMV are thought to rely on continuously ongoing viral gene expression and RNA replication, thus bona fide “chronic infection” at the single-cell level. While the level of gene expression by these viruses may be tolerated by infected cells with only minimal detrimental effects, the antiviral immune response can also decide the fate of a virus-infected cell. CD8 T cells, for example, can eliminate virus-infected cells by cytotoxic mechanisms such as perforin and granzymes. Alternatively, CD8 T cells have been suggested to cleanse virus-infected cells by noncytolytic mechanisms, which prominently include IFN signaling (Binder and Griffin, 2001; Burdeinick-Kerr et al., 2009; Burdeinick-Kerr and Griffin, 2005; Guidotti et al., 1994; Guidotti et al., 1999a; Guidotti et al., 1996a; Guidotti et al., 1996b; Hausmann et al., 2005; Moseman et al., 2020; Oldstone et al., 1986; Parra et al., 1999; Patterson et al., 2002; Thimme et al., 2003; Tishon et al., 1993; Tishon et al., 1995). The latter mechanisms seem particularly important for irreplaceable cells such as neurons of the central nervous system (Binder and Griffin, 2001; Burdeinick-Kerr et al., 2009; Burdeinick-Kerr and Griffin, 2005; Griffin, 2010; Hausmann et al., 2005; Moseman et al., 2020; Oldstone et al., 1986; Parra et al., 1999; Patterson et al., 2002; Tishon et al., 1993) but have also been reported to play a prominent role in the clearance of persistently infected hepatocytes (Guidotti et al., 1994; Guidotti et al., 1999a; Guidotti et al., 1996a; Guidotti et al., 1996b; Guidotti et al., 1999b; Thimme et al., 2003). Interestingly, the survival of formerly virus-infected cells has even been reported for prototypic cytopathic viruses such as influenza A and B viruses (Chambers et al., 2019; Dumm et al., 2019; Hamilton et al., 2016; Heaton et al., 2014; Reuther et al., 2015). The underlying mechanisms, however, remain imperfectly understood. Also, these latter studies could not formally differentiate viral elimination after a period of active viral replication from abortive infection events. Finally, potential long-term consequences of a transient viral infection on surviving cells and their progeny remain to be investigated.

For several decades, LCMV infection of mice, its natural host, has served as the prototypic model to study systemic persistent RNA virus infection, the impact of viral persistence on immunity, and the immunological mechanism implied in viral clearance (Hotchin, 1962; Traub, 1936; Volkert and Lundstedt, 1968; Zinkernagel, 2002). LCMV clearance in adult mice depends chiefly on antiviral CD8 T cells, with a significant contribution by antiviral antibodies when the course of the infection is protracted (Bergthaler et al., 2009; Fung-Leung et al., 1991). Neonatal infection, however, results in the negative selection of the antiviral CD8 T cell repertoire and lifelong persistence (Pircher et al., 1989; Traub, 1936). This coexistence of virus and host in a “carrier” state resembles in many aspects perinatally acquired human HBV infection (Guidotti and Chisari, 2001). In light of the noncytolytic nature of LCMV infection, it is widely assumed that in immunologically tolerant virus carriers, infected cells promote permanent viral gene expression and genome replication for their entire natural life span. However, long-standing observations on fluctuating viral titers in the supernatant of persistently infected cell cultures and a decrease of viral protein and RNA over time contrasted with this concept and stimulated the hypothesis that LCMV infection in cell culture is cyclic and transient (Hotchin, 1973a; Hotchin, 1973b; Hotchin, 1974; King et al., 2018; Lehmann-Grube, 1967; Oldstone and Buchmeier, 1982; Weber et al., 1983). Still, technical limitations have for the longest time precluded the fate mapping of LCMV-infected cells, notably in persistently infected mice.

Here, we have exploited a Cre recombinase (Cre) encoding LCMV (Steinbach et al., 2019) in conjunction with gene-targeted fate-reporter mouse models and cell lines to revisit the dogma of life-long viral persistence at the single-cell level. We found that even in highly immune-deficient mice, LCMV cannot efficiently persist at the single-cell level but relies on continuous spread in order to maintain the infection. Clearance at the single-cell level was observed in the absence of adaptive immunity and independently of type I or type II IFN signaling, and was accompanied by profound transcriptome alterations.

To elucidate the fate of LCMV-infected cells in various experimental systems, we relied on genetically engineered cell fate reporter models. STOP_flox-RFP mice, for example, carry an RFP transgene in the ubiquitously active ROSA26 locus, preceded by a loxP-flanked transcription termination sequence. Cre-mediated loxP site recombination upon infection by a Cre-expressing LCMV (artLCMV_Cre) allows for cellular fate mapping based on the irreversible activation of RFP expression in the infected cell and its progeny. artLCMV_Cre encodes for Cre in a way that allows neither its immediate translation nor direct Cre mRNA transcription (Fig. S1, A and C). Rather, Cre expression and reporter gene trans-activation depend on full-length viral genome replication followed by mRNA transcription from the resulting antigenomic template. Experiments conducted in a Cre-reporting A549 cell line confirmed that viral particles delivering Cre failed to trans-activate fluorescent reporter activity unless the viral polymerase replicated the incoming viral genome into a viral antigenome and from this template transcribed viral Cre mRNA (Fig. S1, B and D). These studies validated artLCMV_Cre-induced fluorescence as a readout of current or past active viral genome replication, thus excluding abortive infection by replication-incompetent, defective viral particles as a potential confounder in our cellular fate mapping analyses.

To determine the fate of LCMV-infected primary hepatocytes in culture, we prepared single-cell suspensions from the liver of RFP fate reporter mice (STOP_flox-RFP) and infected them with artLCMV_Cre (Fig. 1 A and Fig. S1 H). 2 d later, the majority of hepatocytes expressed the viral nucleoprotein (NP) but had not yet accumulated detectable levels of RFP, a stage of infection subsequently referred to as “early infected.” By day 6 to day 8, virtually the entire culture expressed NP and had accumulated clearly detectable levels of RFP, a stage of infection herein referred to as “late infected.” Intriguingly, when following the culture for 13 or even 16 d, a substantial proportion of hepatocytes (∼20–60%) no longer expressed detectable levels of NP yet were clearly RFP-positive. This pattern (NP^–RFP⁺) suggested that the respective cells had cleared artLCMV_Cre infection in a noncytolytic, hepatocyte-intrinsic manner. An inherent limitation of this artLCMV_Cre-based fate mapping approach consists, however, in the inability to discriminate between formerly infected, virus-cleared cells and the progeny of such cells. For simplicity, both types of cells will be referred to as “cleared” throughout this study. Similar to the above findings with artLCMV_Cre, the percentage of NP-expressing (infected) primary hepatocytes increased from day 2 to day 6 after WT LCMV infection but declined continuously at later time points (Fig. S1 G).

While our primary hepatocyte cultures remained viable for 16 d (Fig. S1 H), transcriptional changes upon long-term ex vivo culture have been reported and may reflect cellular de-differentiation (Clayton and Darnell, 1983), with a currently unknown impact on viral persistence. Neonatal infection of mice with WT LCMV results in an immunologically tolerant carrier state (Cole et al., 1972; Hotchin, 1962; Traub, 1936). Here, we infected neonatal STOP_flox-RFP mice with artLCMV_Cre to test whether infected hepatocytes can undergo noncytolytic clearance in living animals. As expected, artLCMV_Cre established a systemic carrier state, which was evident in high levels of viral RNA in liver, lung, spleen, and brain at 3 wk of age, and infectious virus was detectable in blood (viremia), spleen, and lung for up to 20 wk (Fig. 1, B and C; and Fig. S2, A–D). In remarkable contrast, infection of adult mice with artLCMV_Cre did not result in viremia, and viral loads were only transiently detected in spleen and liver on day 4 after infection but were suppressed to undetectable levels by day 8 (Fig. S2, E–H). Immunohistochemical analyses demonstrated that artLCMV_Cre replication in the liver of adult infected animals was restricted to Kupffer cells, the first target of LCMV infection in this organ (Bergthaler et al., 2007), and was cleared before significant spread into hepatocytes occurred (Fig. S2 I). This limitation precluded the study of artLCMV_Cre clearance from hepatocytes in adult infected mice such that all further in vivo studies were conducted in neonatally infected animals. We isolated hepatocytes for flow cytometric analysis at week 2 after neonatal infection and found that ∼40% of all hepatocytes were actively infected as judged by NP expression (Fig. 1 D and Fig. S1 E). Of note, half of these cells expressed NP but had not yet accumulated detectable levels of RFP, and thus were early infected, while the other half were late-infected cells expressing both NP and RFP. Analogous to the above findings in cell culture, another 20% of hepatocytes in carrier mice reported Cre-induced RFP but were negative for the NP, indicating that these cells were cleared, i.e., had been infected by artLCMV_Cre but were no longer actively infected at the time point of analysis. By week 5 after neonatal infection, about half of the liver consisted of cleared hepatocytes, and such cells still made up for >40% at weeks 8 and 12 (Fig. 1 D). The proportion of actively infected cells (sum of early infected and late infected), however, decreased considerably over time and by week 12 after neonatal infection accounted for <1% of all hepatocytes. In a complementary approach, we analyzed noncytolytic artLCMV_Cre clearance on histological liver sections of enhanced YFP (eYFP) fate reporter mice (STOP_flox-eYFP). Immunostaining for viral NP revealed that infection of the liver emanated from numerous foci of infected cells, which were present on day 3 and spread through the entire organ by week 2, at which point a majority of hepatocytes expressed NP (Fig. 1 E; and Fig. S1, F and J). In contrast, at week 4 and similarly at week 12 after neonatal infection, only a comparably low number of infected (NP-positive) hepatocytes was loosely scattered throughout the tissue. On day 3 and at week 2, Cre-induced eYFP expression matched the abundance of NP-positive cells, whereas at week 4 and at week 12, eYFP⁺ cells outnumbered NP⁺ cells by far. This finding was in line with our flow cytometry–based analysis (Fig. 1 E, compare Fig. 1 D), indicating that a substantial proportion of hepatocytes had undergone noncytolytic viral clearance. Likewise, we detected noncytolytically cleared cells in lung, kidney, and ileum of STOP_flox-RFP mice at week 12 after infection (Fig. S1 I). Next, we investigated whether hepatocytes judged as cleared based on the lack of detectable viral NP were thoroughly cleansed from viral genomes. We performed single molecule in situ hybridization (Calabrese and Wieland, 2017) to detect artLCMV_Cre genomic RNA on liver sections from artLCMV_Cre-infected STOP_flox-RFP mice (Fig. 1 F and Fig. S1 K). A large proportion of cells, mostly hepatocytes, contained viral genomic RNA at week 3 after infection, whereas only very few randomly scattered cells still harbored LCMV RNA at week 14. The latter finding contrasted, again, with the vast number of cells that expressed recombination-induced RFP (Fig. 1 F). To corroborate these conclusions, we performed single-cell RT-qPCR (quantitative RT-PCR) on RFP-positive hepatocytes. Whereas most of the cells collected at week 3 after neonatal infection contained artLCMV_Cre RNA, the cells recovered at week 14 were virus-free (Fig. 1 G). Next, we investigated whether the population of cleared hepatocytes persisting at late time points after neonatal infection (compare Fig. 1 D) represented individual cellular infection events or rather clonally expanding populations of formerly infected cells. For this we exploited R26R-Confetti mice (Livet et al., 2007), which upon Cre-recombination stochastically express one out of four fluorescent proteins (Fig. S3 A). At week 12 after neonatal infection, these four fluorescent labels were randomly interspersed throughout the liver of neonatally infected animals (Fig. S3 B), indicating that even clusters of Cre-recombined hepatocytes originated from several distinct infection and recombination events. Taken together, these findings argued against a prominent clonal expansion of formerly infected cells and demonstrated that a substantial proportion of formerly infected hepatocytes had undergone noncytolytic viral clearance and populated the liver of carrier mice.

The frequency of actively artLCMV_Cre-infected hepatocytes decreased from about week 4 after neonatal infection onwards (Fig. 1, D–F), and viremia in STOP_flox-RFP mice also declined considerably (Fig. 2 A). This partial virus control coincided with the maturation of the murine adaptive immune system (Siegrist, 2001), prompting us to investigate a potential role of adaptive immunity in noncytolytic virus clearance from hepatocytes of STOP_flox-RFP mice. Unlike immunocompetent STOP_flox-RFP mice, Rag1-deficient STOP_flox-RFP mice (Rag1 × STOP_flox-RFP), devoid of T and B cells, remained highly viremic throughout the 12-wk observation period (Fig. 2 A), and a substantial proportion (∼25%) of the animals’ hepatocytes evidenced active viral replication at 12 wk of age (Fig. 2 C). This indicated that systemic control of congenital artLCMV_Cre infection relied on adaptive immune mechanisms. Even in the absence of T and/or B cell–mediated virus control, however, the liver of Rag1 × STOP_flox-RFP mice harbored ∼40% noncytolytically cleared (NP^–RFP⁺) hepatocytes (Fig. 2 C and Fig. S3 C). While the availability of antiviral T and B cells in WT mice is likely to influence the relative contribution of noncytolytic clearance to virus control in tissues, the present findings demonstrated that noncytolytic clearance can occur independently of the adaptive immune system. Type I (α/β) IFNs as well as T cell–secreted type II IFN (IFN-γ) have been described as key effector pathways in noncytolytic hepatocyte clearance, while type III IFN (IFN-λ) cannot be sensed by murine hepatocytes (Guidotti et al., 1996a; Guidotti et al., 1996b; Hermant et al., 2014). To explore a potential role of type I and/or type II IFN signaling in noncytolytic LCMV_Cre clearance, and to also address these pathways’ potential redundancy with adaptive immunity, we studied STOP_flox-RFP mice devoid of T and B cells (RAG1-deficient) and, additionally, devoid of functional type I and type II IFN receptors (AGRag [type I IFN receptor–, type II IFN receptor–, and RAG-1–deficient] × STOP_flox-RFP). Analogous to Rag1 × STOP_flox-RFP mice, neonatally artLCMV_Cre-infected AGRag × STOP_flox-RFP mice remained highly viremic throughout the observation period of 12 wk. Intriguingly, even in these highly immunodeficient mice, 40% of all hepatocytes were NP^–RFP⁺, thus noncytolytically cleared. These findings demonstrated that noncytolytic artLCMV_Cre clearance from hepatocytes can occur in the absence of T cells and B cells even when, in addition, type I and type II IFN sensing is disabled (Fig. 2, B and C).

To investigate whether noncytolytic clearance occurs also in immortalized cell lines, we infected Cre-reporting, human lung–derived A549 cells (A549_floxGFP) with artLCMV_Cre. After 3 d, the culture consisted mostly of NP-expressing and Cre-recombined (GFP⁺) late-infected cells. Over the subsequent days, the NP signal faded continuously until at around day 11, the vast majority of cells (∼97%), albeit almost uniformly expressing GFP, had become NP negative. This indicated viral clearance after a period of active viral replication and Cre expression. By day 14, however, the virus reemerged, and most cells were again NP⁺, resulting in a wave-like pattern of NP expression over time (Fig. 3 A; and Fig. S4, A and C). Irrespective of potential confounders such as cell culture splitting and accompanying culture medium exchange, this observation lent further support to the hypothesis that LCMV does not persist at the single-cell level but relies on continuous spread or, notably in the case of cell cultures, reinfection of formerly infected cells or their progeny. To test this postulate, we infected A549_floxGFP cells with a viral glycoprotein (GP)–deficient but GP protein pseudotyped, Cre-encoding LCMV (LCMVΔGP_Cre). Such viral particles can enter cells to amplify and express their genetic information but they cannot spread in the culture since they are unable produce infectious progeny particles (Flatz et al., 2010). NP expression peaked 1 d after infection (at multiplicity of infection [MOI] of ∼0.5), and from day 3 onwards, the culture contained ∼60% of GFP-positive Cre-reporting cells (Fig. 3 B and Fig. S4 D). NP levels, however, declined sharply from d 1 onwards and reached detection limits of flow cytometry by day 7. Importantly and in contrast to artLCMV_Cre infection (Fig. 3 A and Fig. S4 C), NP remained undetectable throughout day 28, indicating that artLCMV_Cre persistence in cell culture relies on continuous (GP dependent) reinfection cycles of formerly infected and transiently cleared cells and/or infection of their progeny. Accordingly, intracellular viral RNA in LCMVΔGP_Cre-infected A549_floxGFP cells reached an early peak, then declined rapidly. It fell below detection limits of TaqMan RT-PCR by day 42 after infection and did not reemerge until day 57, suggesting that LCMVΔGP_Cre had been eliminated from the culture because it failed to propagate (Fig. 3 C). In line with the observation that the proportion of formerly LCMVΔGP_Cre-infected and therefore Cre-recombined (GFP⁺) cells remained constant from day 3 until day 28 (Fig. 3 B and Fig. S4 D), these GFP⁺ cells exhibited unimpaired viability and divided at a rate that was indistinguishable from virus-naive cells in the same culture (Fig. S4, F–H). The above studies in AGRag × STOP_flox-RFP mice had indicated that type I IFN signaling was dispensable for noncytolytic viral clearance. To corroborate these conclusions and extend them to cell culture, we performed LCMVΔGP_Cre infection experiments in Cre-reporting Vero cells (Vero_floxGFP), a cell type known to be type I IFN–deficient (Desmyter et al., 1968; Mosca and Pitha, 1986; Fig. 3 D and Fig. S4 E). Viral NP expression peaked on day 1 but subsided to below detection limits by day 9 after infection, indicating that type I IFN–dependent mechanisms were not accountable for clearance from cell culture.

The wave-like pattern of viral clearance and subsequent reinfection in artLCMV_Cre-infected cultures (Fig. 3 A and Fig. S4 C) suggested that viral infection and clearance was accompanied by a refractory period, i.e., a time window during which infected or recently cleared cells were insusceptible to superinfection or reinfection, respectively. To test this hypothesis, we infected A549_floxGFP cultures with a LCMVΔGP_Cre virus carrying a Flag-tagged NP (LCMVΔGP_Cre/Flag-NP). Upon infection at MOI of ∼0.5, we differentiated three subsets of cells in the cultures. Besides (i) a population of uninfected, nonrecombined cells (Flag-NP^–GFP^–; Fig. 3 E and Fig. S4 B), the viral Flag-tag allowed us to differentiate Cre-recombined (GFP⁺) cells, which were either (ii) actively infected, nucleoprotein-expressing (Flag-NP⁺GFP⁺) cells or (iii) formerly infected cells that had cleared LCMVΔGP_Cre/Flag-NP-infection (Flag-NP^–GFP⁺; Fig. 3 E). At various time points after LCMVΔGP_Cre/Flag-NP “primary” infection, we exposed the cultures to “secondary” infection with a GP-deficient LCMV devoid of a Flag-tag but encoding for the cell surface–expressed Thy1.1 marker (LCMVΔGP_Thy1.1). The percentage of Thy1.1-expressing cells among the three aforementioned populations (i, ii, and iii) allowed us to determine the respective cells’ susceptibility to infection or superinfection, respectively. As expected, the proportion of Flag-NP⁺ cells declined continuously from day 1 onwards (data not shown). Independently thereof, nucleoprotein-expressing cells (ii) were, on average, four times less permissive to LCMV infection than not previously infected cells (i) or cells that had cleared LCMVΔGP (iii; Fig. 3 E). It remains unknown whether this result reflects only partial refractoriness during infection or, more likely, that the iii subset of cells consisted of some completely refractory cells and other partially or fully permissive cells, perhaps as a function of time elapsed after viral clearance. Still, this result was in line with the concept that active viral replication rendered cells refractory to superinfection (Damonte et al., 1983; Welsh and Pfau, 1972), a phenomenon also referred to as “superinfection exclusion” (Ellenberg et al., 2004) that likely accounted for the wave-like pattern of artLCMV_Cre infection in cell culture (Fig. 3 A). Two additional points are of note. First, “refractoriness” to LCMV infection was observed for cell populations growing in the midst of fully susceptible cells, suggesting that refractoriness was cell-intrinsic and not transmissible from one cell to the next. Second, Cre-recombined cells undergoing noncytolytic clearance reacquired, eventually, unimpaired susceptibility to LCMV (re)infection.

The viral matrix protein Z has been shown to inhibit viral RNA transcription and replication when present at high intracellular concentrations, supposedly by favoring viral genome packaging into budding particles (Cornu and de la Torre, 2001; Kranzusch and Whelan, 2011). It was therefore tempting to speculate that the viral Z protein mediated noncytolytic viral clearance. To test this hypothesis, we generated an LCMV-expressing Cre in lieu of the Z gene (LCMVΔZ_Cre). LCMVΔZ_Cre was strictly cell-associated and spread only slowly through A549_floxGFP cultures (Fig. 4 A). By day 45, only ∼50% of the cells had encountered the virus as evident by Cre-induced GFP expression. Intriguingly, though, more than half of these Cre-recombined cells contained undetectable levels of viral NP, indicating that noncytolytic clearance had occurred in the absence of the viral Z protein. By day 108, the vast majority of cells had been infected (GFP⁺) but had cleared the virus (NP^–). To investigate whether the viral Z protein is accountable for superinfection exclusion as previously postulated (Cornu et al., 2004), we generated a Z-deficient, Cre-encoding LCMV with a flag-tagged NP (LCMVΔZ_Cre/Flag-NP). We tested whether A549_floxGFP cells, when infected with LCMVΔZ_Cre/Flag-NP, remained permissive to LCMVΔGP_Thy1.1. On the contrary, Thy1.1 levels in NP-Flag–positive cells (LCMVΔZ_Cre/Flag-NP-infected, early or late infected) were substantially lower than in NP-Flag–negative cells (naive or cleared), indicating that Z protein expression was not or at least not fully accountable for superinfection exclusion (Fig. 4 B). Again, cells that had cleared the infection were as susceptible to infection as naive cells from the same culture.

Cell-intrinsic elimination of artLCMV_Cre with a refractory period and resistance to superinfection suggested changes in cellular homeostasis. Hence, we set out to investigate whether formerly infected but cleared cells exhibited transcriptomic alterations. To this end, we collected hepatocytes from 14-wk-old STOP_flox-RFP mice that had been neonatally infected with artLCMV_Cre (“carriers”). Hepatocytes were sorted according to RFP expression, and total cellular RNA of each population was subject to bulk mRNA sequencing (mRNA-Seq). Importantly, the proportion of actively infected hepatocytes was <1% at 14 wk after infection (Fig. 1 D and data not shown), such that RFP-negative hepatocytes consisted virtually exclusively of uninfected cells while RFP-positive hepatocytes were almost exclusively formerly infected, cleared cells. To account for potential transcriptome alterations as a result of RFP expression, not only hepatocytes from uninfected STOP_flox-RFP mice but also RFP-reporting hepatocytes from uninfected mice with liver-specific RFP expression (STOP_flox-RFP × Alb-Cre) were included as controls. Strikingly, the transcriptomes of noncytolytically cleared hepatocytes from six individual animals clustered together and were clearly differentiated from both, the RFP-negative hepatocytes of the same animals as well as from hepatocytes of the aforementioned uninfected control groups (Fig. 5 A). In noncytolytically cleared hepatocytes, 1,272 genes were significantly up-regulated as compared with uninfected cells from the same animals (Fig. 5 B). The transcriptome of the latter cells was more similar to hepatocytes of uninfected STOP_flox-RFP mice (271 differentially expressed genes; Fig. 5, A and B). Hepatocytes of uninfected STOP_flox-RFP mice differed only marginally from those of STOP_flox-RFP × Alb-Cre mice, excluding RFP expression as an important source of transcriptomic alterations. Altogether, this indicated that most of the transcriptome alterations in noncytolytically cleared hepatocytes had occurred in a cell-intrinsic manner rather than in a bystander fashion. Virus-cleared hepatocytes exhibited a prominent enrichment of gene sets and pathways related to (1) immune responses, (2) cell metabolism, and (3) proliferation and extracellular matrix remodeling (Fig. 5 C). Up-regulation of the latter may indicate a response to cellular stress and/or damage triggered as a consequence of the viral infection. In contrast, pathways involved in metabolic functions were negatively enriched. To identify transcription factors (TFs) supposedly driving these gene expression alterations, we employed a regulon analysis (Garcia-Alonso et al., 2019; Holland et al., 2020) focused on high-confidence TF–target gene relationships and limited to TFs detectable in our mRNA-Seq dataset. This analysis yielded seven factors, three of which were up-regulated in noncytolytically cleared hepatocytes: Spi1/PU.1, Erg, and Tal1 (Fig. 5 D and Fig. S5 A). Spi1/PU.1, our top hit, is a key regulator of pro-fibrotic gene expression (Wohlfahrt et al., 2019) and was up-regulated 13-fold. In contrast, Erg expression is known to counteract liver fibrogenesis (Dufton et al., 2017) and itself is a target of Tal1 (Thoms et al., 2011; Fig. 5 D). We performed TaqMan RT-PCR to validate genes of interest, which mRNA-Seq had suggested were differentially expressed in noncytolytically cleared hepatocytes (RFP⁺) and uninfected (RFP⁻) hepatocytes of carrier mice (Fig. S5 B). Finally, we investigated whether the transcriptomic footprint of LCMV clearance reflected, at least in part, a universal hepatocytic response to transient RNA virus infection. We compared our mRNA-Seq data to published genome-wide ChIPmentation-based chromatin immunoprecipitation DNA-sequencing datasets on human liver biopsies from HCV patients, either chronically infected or cleared by direct-acting antiviral (DAA) treatment (Fig. 5 E; and Fig. S5, C and D; Boldanova et al., 2017; Hamdane et al., 2019). Strikingly, for 82% of the genes, which we found up-regulated in formerly LCMV-infected mouse hepatocytes (Log₂[FC]>1), the human orthologs in DAA-cured subjects were characterized by H3K27ac modifications (Log₂[FC]>0; Fig. 5 E and Table S1). Analogous results were also obtained for a second independent cohort of DAA-cured HCV patients as well as for chronically HCV-infected patients (Fig. S5, C and D; and Table S1), and the relatedness of transcriptomic signatures was also validated by comparison to human RNA sequencing data (Fig. S5, E and F). These observations indicated that the transcriptomic footprint of cleared infection exhibited a substantial degree of conservation across different mammalian species and distantly related RNA viruses.

This study redefines LCMV persistence as a dynamic process, which depends on continuous viral spread; meanwhile, individual infected cells undergo cell-intrinsic noncytolytic clearance.

These conclusions are compatible with observations from other infections suggesting the viral persistence strategy described herein is not unique to the arenavirus family. Antibody-mediated elimination of HCV infection in human liver chimeric mice, for example, has demonstrated that HCV needs to continuously infect new hepatocytes to sustain a chronic infection (de Jong et al., 2014). Analogously, the entry inhibitor Hepcludex can reduce intracellular loads of hepatitis D virus and HBV genomes (Allweiss et al., 2018; Bogomolov et al., 2016). Henceforth, the demonstration of noncytolytic clearance at the single-cell level, a process occurring in a cell-intrinsic manner even when adaptive immunity and IFN signaling are disabled, provides a new mechanistic incentive for the use of entry inhibitors in the treatment of persistent RNA virus infections.

One consequence of this new concept of RNA virus persistence consists of permissive organs such as the liver harboring a substantial proportion of formerly virus-infected cells. Our transcriptome data indicate that cellular homeostasis remains perturbed beyond the viruses’ elimination. It will be of importance to investigate the longevity of this transcriptomic footprint at the single-cell level, and to determine whether repeated reinfection cycles lead to its intensification. Importantly, these alterations in murine livers, which we show are similar to those of HCV-cured patients, are restricted to formerly LCMV-infected hepatocytes and differentiate them from virus-naive bystander cells. By extrapolation to the human setting, where cellular fate mapping of viral infection is impossible, these findings could mean that a substantial proportion of the transcriptomic alterations in actively or formerly HCV-infected livers is due to a hepatocyte-intrinsic imprint of viral infection, independent of immune cell-mediated inflammation or fibrosis. While LCMV has not been linked to carcinogenesis, it is tempting to speculate that the persistent transcriptome alterations imprinted by HCV, the oncogenesis of which remains ill-defined (Hoshida et al., 2013; Nakagawa et al., 2016), may be decisive for the consequences of exposure to other carcinogenic agents such as aflatoxin (Chu et al., 2018; Palliyaguru and Wu, 2013). The elevated incidence of hepatocellular carcinoma in HCV-cured individuals is compatible with this hypothesis (El-Serag et al., 2016; Hamdane et al., 2019).

For several decades, persistent LCMV infection of mice has served as a primary model to study viral clearance by adoptive cell therapy (Ahmed et al., 1987; Gegin and Lehmann-Grube, 1992; Lehmann-Grube et al., 1985; Oldstone et al., 1986; Planz et al., 1997; Tishon et al., 1993; Varho et al., 1981; Volkert, 1963). The present findings have obvious implications also on the interpretation of these earlier studies. The observation that transfer of immune spleen cells resulted in the rapid clearance of most organs including the liver but much delayed virus elimination from neurons (Oldstone et al., 1986) was, subsequently, interpreted to reflect differential level of T cell immunosurveillance owing to low levels of MHC class I on neurons (Joly et al., 1991; Tishon et al., 1993). While this interpretation remains viable in principle, the present findings open up the alternative possibility that cell-intrinsic elimination of LCMV from neurons takes longer than from hepatocytes, and that adoptive cell therapy merely prevented or slowed down the rate at which these cells were reinfected. The observation that effective elimination of persistent LCMV infection by adoptive spleen cell therapy required virus-specific B cells as a source of antiviral antibodies (Planz et al., 1997) supports this alternative hypothesis.

Another important concept, which studies in LCMV carriers have contributed to, consists of the noncytolytic, cytokine-mediated clearance of virus-infected hepatocytes (Guidotti et al., 1999a; Tishon et al., 1995). The present findings, albeit not providing any evidence to the contrary, demonstrate that virus clearance can also occur in a cell-intrinsic manner, independently of either type I IFN or type II IFN signaling. Future studies should investigate the relative contribution of cytokine-mediated virus clearance at the level of individual hepatocytes versus indirect effects on systemic viral loads, which are predicted to impact the reinfection rate of hepatocytes and thereby may also reduce the viral burden in the liver.

The molecular mechanisms underlying cell-intrinsic noncytolytic clearance of LCMV infection requires further investigation. The processes accountable in immortalized cell lines must obviously rely on either autonomous cell-intrinsic and autocrine mechanisms or potentially involves paracrine signaling, while immune cell effects are excluded. Among the candidate soluble factors that may contribute to clearance in mice, if any, lymphotoxins should be mentioned. They can be produced by several cell types (Ware, 2005) and have a prominent role in the pathogenesis of HCV-induced hepatocellular carcinoma (Browning and French, 2002; Haybaeck et al., 2009) as well as in the cell-intrinsic clearance of HBV (Lucifora et al., 2014). A potential contribution of innate lymphoid cells such as natural killer cells should also be investigated, albeit natural killer cells of AGRAG mice are immature and dysfunctional (Boyman et al., 2004) and thus are likely dispensable for noncytolytic LCMV clearance.

The present investigations on hepatocyte clearance in mice relied entirely on the neonatal infection model. Antiviral T cell tolerance, one of its immunological hallmarks (Pircher et al., 1989), but possibly also the use of a transgenic and therefore attenuated reporter virus (Kallert et al., 2017), may have impacted the relative contribution of noncytolytic and cytolytic clearance mechanisms to virus control in tissues. Of note, however, congenital infection by vertical transmission represents the natural habitat of LCMV in the wild. Accordingly, LCMV has evolved to persist in a T cell–tolerant host, and the viruses’ susceptibility to cell-intrinsic clearance mechanisms may help assuring virus–host balance and long-term healthy coexistence. It seems plausible that a similar pathogen–host relationship predominates when HCV infection progresses to chronicity and antiviral T cells get functionally exhausted (Thimme, 2021).

We provide evidence that the well-known wave-like infection dynamics of LCMV-infected cell cultures (Hotchin, 1973a; Hotchin, 1973b; Hotchin, 1974; King et al., 2018; Weber et al., 1983) are due, at least in part, to noncytolytic clearance and superinfection exclusion. Using Z protein–deficient LCMV (Zaza et al., 2018), we find that the Z protein, albeit a negative regulator of viral transcription and genome replication (Cornu and de la Torre, 2001; Kranzusch and Whelan, 2011), is dispensable for superinfection exclusion of arenavirus-infected cells. It therefore seems unlikely that Z protein in its putative role as a viral molecular clock (Kranzusch and Whelan, 2011) extinguishes LCMV infection in cells. This postulate assumes, however, that cell-intrinsic noncytolytic clearance of LCMV and superinfection exclusion reflect two manifestations of the same underlying molecular events. While we consider this to be highly likely, we acknowledge that this assumption remains uncertain in the absence of a detailed molecular understanding.

Unraveling the mechanisms that underlie cell-intrinsic noncytolytic elimination of persistent RNA virus infection may reveal an Achilles’ heel of these infections. The demonstration herein of cell-intrinsic noncytolytic RNA virus clearance, therefore, furthers our understanding of complications arising from such infection while simultaneously opening up new avenues for their treatment.

BHK-21 cells were obtained from the European Collection of Authenticated Cell Cultures (clone 13; 85011433) and cultured in DMEM (Sigma-Aldrich) complemented with 10% FCS, 1 M Hepes, 100 mM sodium pyruvate, and 1× Tryptose Phosphate Broth (50× mix; Gibco). BHK-21 cells expressing the LCMV GP and the LCMV NP respectively, have been described previously (Darbre et al., 2015; Flatz et al., 2010) and were cultured in the presence of 2 µg/ml puromycin. LCMV GP expressing HEK293T cells (293T-GP) were described previously (Flatz et al., 2010) and were cultured in DMEM containing 10% FCS and 2 µg/ml puromycin. NIH/3T3 cells were obtained from the American Type Culture Collection (CRL-1658) and cultured in DMEM containing 10% FCS. A549_floxGFP and Vero_floxGFP cells were kindly provided by Martin Schwemmle, University of Freiburg, Freiburg, Germany. These cells were generated by retroviral transduction as described (Reuther et al., 2015) using the pMSCV-loxp-dsRed-loxp-eGFP-Puro-WPRE plasmid (Addgene plasmid 32702; donated by Hans Clevers, Hubrecht institute, Utrecht, Netherlands) and were cultured in the same medium as 293T-GP cells. For long-term culture of A549_floxGFP and Vero_floxGFP cells, cells were split three times a week at a 1:3 ratio.

Animals were euthanized with CO₂ followed by immediate perfusion of the liver via the inferior vena cava with 20 ml HBSS (GIBCO) supplemented with 0.5 mM EGTA (Sigma-Aldrich). The liver was digested with 20 ml prewarmed (40°C) digestion medium (L15 medium [GIBCO] containing 40 mg/l Liberase [Roche]) using a peristaltic pump (5 ml/min). Subsequent to the isolation of the organ, the liver capsule was dissected in a Petri dish containing 10 ml digestion medium. Hepatocytes were passed through a 70-µm cell strainer and centrifuged (50 g for 10 min at 4°C) and washed twice with 30 ml William’s E medium (GIBCO) containing 1% penicillin-streptomycin and 10% FCS. For ex vivo infection, cells were plated in 6-well plates (0.8 × 10e6 cells/well) coated with rat tail collagen I (Corning). 3 h after plating, cells were washed twice with PBS and the medium was exchanged for William’s E medium containing 1% penicillin-streptomycin and 0.5% FCS. For long-term culture, medium was exchanged three times a week.

The genetic engineering and rescue from cDNA of artLCMV, LCMVΔGP, and LCMVΔNP viruses and vectors were performed as described (Darbre et al., 2015; Flatz et al., 2006; Flatz et al., 2010; Kallert et al., 2017). The artLCMV and LCMVΔGP viruses used in this study are based on the Armstrong Clone 13 (Cl13) backbone (Flatz et al., 2006), but encode (artLCMV) or are pseudotyped with the GP of the LCMV strain WE. Sequences encoding Cre, GFP, dTomato, and Thy1.1 were PCR-amplified from plasmids containing the respective open reading frames. To generate LCMV encoding an N-terminally Flag(DYKDDDDK)-tagged NP, a primer comprising the Flag-tag sequence was used to amplify the NP gene. For the generation of LCMVΔZ_Cre, the Z-encoding sequence of the pI-L-Cl13(−) plasmid (Flatz et al., 2006) was exchanged for a synthetic, Cre-encoding sequence resulting in pI-L_Cre_IGR_L (Genscript). For the rescue of Z-deficient viruses, 0.1 µg of a Z-encoding pCAGGs expression plasmid (pC-Z) was added to the rescue mix as described (Zaza et al., 2018). 6 d after transfection of BHK-21 cells with pI-L_Cre_IGR_L and combination with the other LCMV rescue plasmids, cells were co-cultured with puromycin-resistant A549_floxGFP cells at a 1:1 ratio representing the start of the experiment depicted in Fig. 4 A. 3 d later, BHK-21 cells were eliminated from the co-culture by the addition of 2 µg/ml puromycin to the medium.

Virus stock and blood titers of artLCMV and LCMVΔGP were determined by immunofocus assay on NIH/3T3 cells and 293T-GP cells, respectively, using a rat-anti-LCMV NP antibody (VL-4) for detection of virus-infected cells as described (Battegay et al., 1991). NP-deficient viruses were quantified on BHK-21 cells expressing the LCMV NP using a mouse anti–LCMV GP antibody (83.4; Eschli et al., 2007) as reported previously (Darbre et al., 2015). STOP_flox-RFP mice and STOP_flox-RFP–derived hepatocytes were infected with an artLCMV encoding Cre in cis with the viral NP on the first S segment and a GFP together with the viral GP on the second S segment. For infection of STOP_flox-eYFP mice, A549_floxGFP cells, and Vero_floxGFP cells, an artLCMV encoding a red fluorescent tomato protein on the second, GP-encoding S segment was used. For infection of R26R-Confetti mice, we used a virus encoding a Cre transgene on both S segments.

Live/dead staining was performed with the Zombie UV Fixable viability kit (BioLegend, used for hepatocytes) or the Zombie Yellow Fixable viability kit (BioLegend, used for A549_floxGFP cells in Fig. S4 G). To determine the proliferation rate of A549_floxGFP cells in Fig. S4 F, cells were stained with the CellTrace Violet Cell Proliferation Kit (Thermo Fisher Scientific). Antibodies against Thy1.1 (Ox-7, BV421-conjugated; BioLegend) and the Flag-tag (L5, APC-conjugated; BioLegend) were used at a 1:100 dilution. The rat-anti-LCMV NP antibody (VL-4, hybridoma supernatant) was conjugated to Alexa Fluor 647 using the Alexa Fluor 647 Antibody Labeling Kit (Invitrogen) and was diluted 1:1,000. Cells were stained for 30 min at 4°C. For intracellular staining, cells were fixed for 15 min at room temperature (PBS/2% paraformaldehyde [PFA]) and permeabilized for 5 min in permeabilization buffer (FACS buffer/0.1% saponin). Labeled cells were measured on an LSRFortessa (Becton Dickinson) flow cytometer. For the isolation of nucleic acids, cells were sorted on a FACSAria III (Becton Dickinson).

Liver tissue was fixed in 4% PFA and embedded in paraffin. Subsequent to inactivation of endogenous peroxidases with PBS/3% hydrogen peroxide, tissue sections were blocked with PBS/10% FCS. For immunohistochemical staining of LCMV NP and eYFP, samples were incubated overnight (4°C) with rabbit-anti LCMV NP serum (diluted 1:10,000) and rabbit anti-GFP antibody (diluted 1:200; D5.1; Cell Signaling), respectively. Bound primary antibodies were incubated with the ImmPRESS HRP Goat Anti-Rabbit IgG Polymer Detection Kit (Vector Laboratories) for 30 min before treatment with 3,3′-diaminobenzidine (Dako). For immunofluorescent staining of eYFP, sections were treated as described above, and bound HRP was visualized with the TSA plus Cyanin 3 kit (Perkin Elmer). For fluorescent costaining of LCMV-NP, sections were treated with Fab mouse IgG Block (Jackson ImmunoResearch) and incubated with rat anti–LCMV NP serum (diluted 1:1,000) for 1 h at room temperature. Bound primary antibody was stained with Alexa Fluor 647–labeled donkey anti-rat antibody (diluted 1:200; AB_2340694; Jackson ImmunoResearch). Immunofluorescent staining of RFP was performed with a rabbit anti-RFP antibody (diluted 1:400; ab124754; Abcam) followed by incubation with a HRP goat anti-rabbit IgG polymer (K4003; Dako). Bound HRP was visualized with the Opal 570 reagent (Akoya Biosciences). For fluorescent costaining of GFP derivates, sections were incubated with a chicken anti-GFP antibody (diluted 1:200; CGFP-45ALY-Z; ICL). Bound GFP antibody was visualized with an A647-labeled goat anti-chicken antibody (diluted 1:200; A21449; Life). Hemalum (Merck) and DAPI (Invitrogen) were used to counterstain nuclei in brightfield and fluorescence microscopy, respectively. Immunohistochemical staining for RFP was performed with the Ventana Discovery Ultra (Roche Diagnostics [Suisse] SA) automated slide stainer. In brief, tissue sections were deparaffinized and rehydrated. Antigen was retrieved by heat in Cell Conditioning buffer 1 (CC1; 950–124; Ventana) at 95°C for 40 min. Primary antibody (preadsorbed anti-RFP; 600–401-379; 1:100 dilution; Rockland Antibodies and Assays) was manually applied and incubated for 1 h at 37°C. After washing, the HRP-polymer secondary antibody (anti-rabbit polymer HRP; 414141F; Nichirei Histofine Simple Stain MAX PO) was incubated for 1 h at 37°C. Detection was performed with the Ventana DISCOVERY ChromoMap DAB (760–159; Ventana) detection kit. Afterward, the slides were counterstained with hematoxylin II, followed by bluing reagent (respectively, 790–2208 and 760–2037; Ventana). Sections were then dehydrated, cleared, and mounted with permanent mounting. Slides stained for LCMV NP, GFP, and RFP (fluorescent) were scanned using a MIRAX Midi slide scanner (ZEISS) or a Panoramic Flash (3D Histech) at 200× magnification. Slides stained immunohistochemically for RFP were scanned with the Hamamatsu NanoZoomer S60.

Detection of viral RNA by in situ hybridization was performed on liver sections fixed in 4% PFA using the ViewRNA ISH Tissue Assay (Thermo Fisher Scientific) as previously described (Calabrese and Wieland, 2017) with the following changes in the protocol: probe-set dilution, 1:30; hybridization time, 2.5 h; preamplification and amplification time, 40 min; and label probe-AP dilution, 1:500. For specific detection of NP coding, LCMV S-segment RNA, we used a custom-made type I probe set (Thermo Fisher Scientific) binding to the negatively oriented LCMV NP gene (GenBank: KY514256.1; nucleotides 1640..3316). Stained slides were acquired in brightfield and in fluorescence (tetramethylrhodamine-isothiocyanate filter) using a slide scanner Hamamatsu NanoZoomer S60. The raw image files were converted in 8-bit grayscale files and overlaid to obtain the merged composite images using ImageJ.

For total cellular RNA isolation from organs and cultured cells, samples were collected in TRI Reagent (Sigma-Aldrich). Organ samples were subsequently homogenized using the TissueLyser II (Qiagen). For bulk analysis of isolated hepatocytes, cells were sorted directly into TRI Reagent LS (Sigma-Aldrich). RNA was extracted using the Direct-zol RNA MicroPrep kit (Zymo Research). NP-coding, LCMV RNA was quantified as previously described (Pinschewer et al., 2010). Applied Biosystems TaqMan kits (Thermo Fisher Scientific) were used to determine mRNA levels of PDGFRα (Mm00440701_m1), PDGFRβ (Mm00435546_m1), PDGFβ (Mm00440677_m1), Col1a1 (Mm00801666_g1), Col1a2 (Mm00483888_m1), Col4a1 (Mm01210125_m1), Loxl2 (Mm00804740_m1), H2-Ab1 (Mn00439216_m1), HGF (Mm01135184_m1), TLR4 (Mm00445273_m13), OAS3 (Mm00460944_m1), TGFβ1 (Mm01178820_m1), CTGF (Mm01192933_g1), GAPDH (NM_008084.2), Spi1 (Mm00488140_m1), Erg (Mm01214244_m1), and Tal1 (Mm01187033_m1). RT-qPCR was performed with Superscript III One Step Platinum Taq (Thermo Fisher Scientific). Samples with cycle threshold (Ct) values >40 were considered negative. For quantitative single-cell RT-qPCR, we used the CellsDirect One-Step qRT-PCR Kit (Thermo Fisher Scientific). Single hepatocytes were sorted into 5 µl of lysis solution. After lysis, we added 5 µl of diethyl pyrocarbonate–treated, nuclease-free water (Thermo Fisher Scientific). Levels of NP-encoding LCMV RNA were determined from nine tenths of the lysate. Expression levels were normalized to that of GAPDH, which we measured in the remaining one tenth of the sample. Samples with Ct_GAPDH >40 were excluded, and samples with Ct_NP >35 were considered negative.

mRNA-Seq libraries were prepared from RNA isolated from 50,000 sorted hepatocytes using the TruSeq kit (Illumina). For sequencing, an Illumina NextSeq 500 was used performing two 38-bp paired-end runs. Reads were aligned to the mouse mm10 genome using STAR (version 2.7.0c) and setting options for multi-mapping reads to “–outFilterMultimapNmax 10–outSAMmultNmax 1” (Dobin et al., 2013). Gene expression was quantified with featureCounts (Subread version 1.6.4) and extra options “-O–read2pos 5 –M –s 2 –p -B” and relied on gene annotation from Ensemble (version 94; Liao et al., 2014). All further analysis was performed within the R/Bioconductor framework (R version 3.5.0/Bioconductor version 3.8). Genes were filtered for a log-counts per million reads>0 in at least two samples (n = 12,251 genes). Differential expression analysis was performed with the Bioconductor package edgeR and using two different designs (McCarthy et al., 2012). A within-subject design was chosen to contrast RFP⁺ versus RFP⁻ cells in infected animals. A second design compared RFP⁺ cells from infected and naive animals, as well as RFP⁺ versus RFP⁻ cells from naive animals. This between-subjects design also accounted for the sex of the animals. The edgeR functions glmQLFit and glmQLFTest were used to calculate log-fold-changes and false discovery rate (FDR)–adjusted P values for all contrasts of interest. A function camera from the edgeR package was used to evaluate differential regulation of gene sets from MSigDb (version 6.0). The ComplexHeatmap package was used to draw heatmaps (Gu et al., 2016). The TF target gene resource for mouse from the R package dorothea was used to evaluate differential TF activity across RFP⁺ versus RFP⁻ cells in infected animals (Garcia-Alonso et al., 2019). Here, the camera function from edgeR was used to test TF target gene sets with at least 10 members and confidence levels A and B for differential regulation. The evaluation only included expressed TFs, according to the above criteria. Comparison to altered gene expression in the context of HCV liver infection relied on a supplementary gene list published by Hamdane et al. (2019). The list was translated to mouse orthologs, to the extent they are defined, using ensemble biomart and subsetted to the expressed genes in our study. This gave rise to 1,015 (HCV infection and DAA cure in patients: 2,193 genes) expressed mouse orthologs for which logFCs could be directly compared. The raw mRNA-Seq datasets of Boldanova et al. (2017) (accession no. GSE84346; HCV-infected + control samples) and Hamdane et al. (2019) (DAA-cured and control) were downloaded from GEO and mapped with STAR (version 2.7.0c) to the human hg38 analysis set genome. Genes were quantified with featureCounts (Subread version 1.6.4), and differential expression analysis was performed with the Bioconductor package edgeR. Ensembl biomart was used to map human genes to mouse orthologs. Mouse orthologs that were up-regulated with a logFC >1 were tested for enrichment among the mouse RFP⁺ versus RFP⁻ regulated genes using the cameraPR function of the limma package.

STOP_flox-RFP mice (B6.CG-GT[ROSA]26Sortm1Hjf; on C57BL/6 background; Luche et al., 2007), STOP_flox-eYFP mice (B6.129X1-Gt[ROSA]26Sortm1[EYFP]Cos/J, backcrossed to C57BL/6 and >98% pure by single nucleotide polymorphism typing; Srinivas et al., 2001) and R26R-Confetti mice (Gt[ROSA]26Sortm1[CAG-Brainbow2.1]Cle/J; backcrossed to C57BL/6 for at least three generations; Livet et al., 2007) have been described. RAG-1 × STOP_fl-RFP mice were generated by crossing STOP_flox-RFP mice to B6.RAG-1 mice (Rag-1^−/−; backcrossed to C57BL/6 for at least 10 generations and 99% genetically pure by single nucleotide polymorphism typing; Mombaerts et al., 1992). AGRAG × STOP_flox-RFP mice resulted from crossing RAG1 × STOP_flox-RFP mice to B6.AGRAG mice (type I IFN receptor^−/−, type II IFN receptor^−/−, Rag-1^−/−, each founding line backcrossed to C57BL/6 for at least 10 generations or generated on C57BL/6 background, respectively; Grob et al., 1999). For neonatal infection, male and female mice were injected intracranially within 24 h after birth with a total volume of 25 µl containing 2.5 × 10⁴ focus forming units of the respective virus. Uninfected litters were used as noncarrier controls. For adult infection, male and female mice were infected i.v. with 7 × 10⁵ focus forming units of artLCMV_Cre. Animal experiments were performed at the University of Basel in accordance with the Swiss law for animal protection and with permission by the Veterinäramt Basel-Stadt.

Error bars represent mean ± SEM or mean ± SD as reported in the figure legends. To compare two groups, two-tailed Student’s t test was used as indicated in the figure legends (unpaired or paired). Two-way ANOVA with Bonferroni’s post-test was used for multiple comparisons. Statistical significance is indicated in the figures as ‡, P < 0.1; *, P < 0.05; **, P < 0.01; GraphPad Prism8 was used for statistical analysis.

Fig. S1 reports the evaluation of artLCMVCre as a tool to induce Cre recombination exclusively in cells replicating the viral genome, the analysis of LCMV-infected mouse hepatocytes in vivo and in cell culture, and the occurrence of cleared cells in lung, kidney, and ileum. Fig. S2 displays the analysis of viral replication in mice infected neonatally or at adult age with artLCMVCre. Fig. S3 shows the clonality fate mapping of virus-cleared mouse hepatocytes in vivo and the visualization of noncytolytically cleared hepatocytes of RAG-1–deficient mice. Fig. S4 depicts the gating strategy to detect GFP, Thy1.1, and NP-Flag–expressing cells as well as the characterization of cultured cells cleared from infection. Fig. S5 provides a validation of differentially expressed genes and comparison of LCMV transcriptomic footprint to HCV-cured and -infected patients. Table S1 shows a comparison of H3K27ac changes observed in HCV-infected or DAA-cured patients with changes of mRNA levels between RFP-positive and RFP-negative hepatocytes of mice neonatally infected with artLCMV_Cre.

The data reported in this study are deposited in Zenodo open data repository (CERN), available at https://doi.org/10.5281/zenodo.5119267. Oligonucleotide primer sequences are available from the authors upon request. RNA sequencing data are deposited in GEO (accession no. GSE157431).

We thank Martin Schwemmle (University of Freiburg, Freiburg, Germany) for providing reagents and valuable comments on the manuscript. We are grateful to Stefan Wieland and Matthias Matter for helpful discussion. We are grateful to Karsten Stauffer for outstanding animal handling and care, as well as to Ingrid Wagner, Karim Hammad, and Cynthia Saadi for excellent tissue processing and histology.

This work was supported by the Swiss National Science Foundation (grant no. 310030_173132 to D.D. Pinschewer). D. Merkler is supported by the Swiss National Science Foundation (grant nos. 310030B_201271 and 310030_185321) and the European Research Council (grant no. 865026).

Author contributions: P. Reuther, K. Martin, D. Merkler, and D.D. Pinschewer contributed to experimental conception and design. P. Reuther, K. Martin, M. Ciancaglini, and D. Calabrese performed experiments. P. Reuther, K. Martin, D. Calabrese, M. Kreutzfeldt, F. Geier, and D. Merkler analyzed and/or interpreted the data. D.D. Pinschewer and P. Reuther wrote the manuscript.