Type I Interferons Direct Gammaherpesvirus Host Colonization

Gamma-herpesviruses colonise lymphocytes. Murid Herpesvirus-4 (MuHV-4) infects B cells via epithelial to myeloid to lymphoid transfer. This indirect route entails exposure to host defences, and type I interferons (IFN-I) limit infection while viral evasion promotes it. To understand how IFN-I and its evasion both control infection outcomes, we used Mx1-cre mice to tag floxed viral genomes in IFN-I responding cells. Epithelial-derived MuHV-4 showed low IFN-I exposure, and neither disrupting viral evasion nor blocking IFN-I signalling markedly affected acute viral replication in the lungs. Maximising IFN-I induction with poly(I:C) increased virus tagging in lung macrophages, but the tagged virus spread poorly. Lymphoid-derived MuHV-4 showed contrastingly high IFN-I exposure. This occurred mainly in B cells. IFN-I induction increased tagging without reducing viral loads; disrupting viral evasion caused marked attenuation; and blocking IFN-I signalling opened up new lytic spread between macrophages. Thus, the impact of IFN-I on viral replication was strongly cell type-dependent: epithelial infection induced little response; IFN-I largely suppressed macrophage infection; and viral evasion allowed passage through B cells despite IFN-I responses. As a result, IFN-I and its evasion promoted a switch in infection from acutely lytic in myeloid cells to chronically latent in B cells. Murine cytomegalovirus also showed a capacity to pass through IFN-I-responding cells, arguing that this is a core feature of herpesvirus host colonization.


Introduction
The γ-herpesviruses persist in lymphocytes and cause lymphoid and epithelial cancers. MuHV-4, like Epstein-Barr virus (EBV) and the Kaposi's Sarcoma-associated Herpesvirus (KSHV), persists in B cells [1]. After epithelial entry, it reaches B cells in organized lymphoid tissue via dendritic cells (DC) [2]. It then spreads with remarkable precision from splenic marginal zone (MZ) macrophages to MZ B cells, follicular DC, then follicular B cells [3]. Glycoprotein conformation changes guide host colonization, with epithelial-derived virions infecting myeloid cells but not B cells, myeloid-derived virions infecting B cells, and B cell-derived virions infecting epithelial cells [4]. However epithelial-derived virions still infect epithelial cells better than myeloid cells, and myeloid-derived virions still infect epithelial and myeloid cells better than B cells. Therefore efficient B cell colonization must involve also a suppression of non-B cell infections.
Immune cell colonization makes host defences an important feature of the γ-herpesvirus infection landscape. Type 1 interferons (IFN-I) are a core vertebrate anti-viral defence [5]. Myriad stimuli induce IFN-I: classically double-stranded RNA (dsRNA), but also other nucleic acids in unusual forms or places, such as unmethylated and cytoplasmic DNA [6]. IFN-I secretion is triggered by phosphorylation of interferon regulatory factors (IRFs) 3 and 7. Signalling through the STAT-1/2-linked IFN-I receptor (IFNAR) then induces an anti-viral state in infected and surrounding cells via restricted protein synthesis, a reduced apoptosis threshold and immune effector recruitment. IFNα is produced mainly by myeloid cells, IFNβ by many cell types, and both in large amounts by plasmacytoid DC [7]. The multiplicity of induction pathways sensitive to different infection hallmarks ensures that essentially all viruses elicit an IFN-I response.
Qualitative questions also remain. ORF11, ORF36, ORF37, ORF54 and M2 all limit IFN-I signalling to infected cells, so ongoing viral gene expression should preclude a cellular IFN-I response; but in cells already exposed to IFN-I, viral signalling blocks should be ineffective and infection should be suppressed. In a uniform cell population such mutual inhibition would manifest as transient viral replication; but in vivo infection is more complicated, with MuHV-4 spreading between cell types and between anatomic sites [23]. Splenic B cell colonization involves at least 5 lytic cycles [3], so virus entry into IFN-I-exposed B cells must be common. For Herpes simplex virus, IFN-I does not prevent epithelial virus production but impaired responses promote pathological neuronal infection [24], suggesting that cell type is an important outcome determinant [25]. Most studies of MuHV-4 and IFN-I have averaged effects across whole organs, so how host response and viral evasion vary with cell type is unclear. To understand this better we tracked MuHV-4 replication in IFN-I responding cells. The results provide new insight into how a complex pathogen interacts with an ancient host defence, and how cell type-specific infection outcomes guide host colonization.

IFN-I induction reduces MuHV-4 replication in vitro
MuHV-4 replicates in a wide range of primary and transformed cells, suggesting that it either induces little IFN-I or evades its effects. Weak inhibition of fibroblast infection by MuHV-4-expressed IFNα [26] suggests effector evasion. However the expression kinetics of IFNα were unclear, and the cells infected (hamster fibroblasts or IFNAR -/fibroblasts transfected with IFNAR encoding DNA) may have responded poorly to murine IFNα. Myeloid cells play a central role in MuHV-4 host colonization [27], and RAW-264 monocyte-macrophages are welldescribed IFN-I producers and responders [28]. Therefore we tested in RAW-264 cells whether MuHV-4 induced IFN-I, and whether IFN-I affected viral replication (Fig 1). ELISA of cell supernatants for IFNβ showed only limited induction in infected cultures (Fig 1A), so we tested viral susceptibility to IFN-I signalling more stringently by treating the RAW-264 cells with poly(I:C), a well-characterised TLR3 ligand. IFNβ was then readily detectable regardless of whether virus was present, and viral replication was reduced 10-fold ( Fig 1B). However the infectivity of induced RAW-264 cell cultures still increased 100-fold between day(d)1 and d2 after MuHV-4 inoculation (Fig 1B), despite IFNβ being detected at both time points (Fig 1A), and continued to increase at d3. Therefore MuHV-4 both induced little IFNβ and partly resisted its effects.

IFN-I induction transiently reduces MuHV-4 replication in lungs
We tested next how IFN-I affected MuHV-4 replication in the lungs, where it infects alveolar macrophages (AMs) and type 1 alveolar epithelial cells (AEC1s) [29]. Again we wanted to know whether IFN-I was induced, and if induced whether MuHV-4 resisted its effects. IFN-I induction is generally rapid, for example peaking 12-24h after experimental respiratory virus infection [30], but IFNAR -/mice given MuHV-4 i.n. have normal virus titers in the lungs for at least 4d [16], implying poor IFN-I induction or poor efficacy. Bioassays have suggested poor induction [31], and consistent with this result, IFNβ was not detected by ELISA of lung washes at 1d post-infection ( Fig 1C). To test MuHV-4 resistance to IFN-I therefore, we induced it with poly(I:C) 6h before virus inoculation. This gave detectable IFNβ in lung washes (Fig 1C), and increased Mx1 transcription in lungs by d1, whereas infection took until d3 ( Fig 1D). Live imaging of luciferase + MuHV-4 ( Fig 1E) and plaque assays of infectious virus ( Fig 1F) showed IFN-I induction reducing lung infection at d3 but not at other time points. Although poly(I:C) also induces inflammatory cytokines, these seem not to affect acute MuHV-4 replication [32,33] and IFNAR blockade abrogated the effect of poly(I:C) (Fig 1G). This implied also that IFN-III induction by poly(I:C) [34] did not have a marked effect. Thus, as with RAW-264 cell infection in vitro, MuHV-4 in the lungs induced little IFN-I, and inducing IFN-I only modestly reduced viral replication.

MuHV-4 passes through IFN-I responding cells
To track how IFN-I responses and viral replication overlap in single cells, we infected Mx1-cre mice, in which cre is transcribed from an IFN-I-inducible Mx1 promoter [35], with MuHV-4 in which cre switches a fitness-neutral (S1 Fig) reporter construct from mCherry (red) to eGFP expression (green) (MHV-RG) [27] (Fig 2). By typing recovered viruses as mCherry + or eGFP + we could determine whether they had passed through a cre + cell; and as Mx1 induction is essentially specific to IFN-I [36][37][38][39], this would tell us whether they had passed through an IFN-I-responding cell. MuHV-4 infected Mx1-cre mice equivalently to non-transgenic littermates (S2 Fig). Viruses recovered from the lungs at d6 after i.n. inoculation under anesthesia a. RAW-264 monocyte/macrophages were infected or not with MuHV-4 (2 p.f.u. / cell) then cultured ± poly(I:C) as a positive control of IFN-I induction (50μg/ml). Cell supernatants pooled from triplicate cultures were assayed for IFNβ by ELISA 1 and 2d later. The dashed line shows the limit of assay sensitivity. A replicate experiment gave equivalent results. b. RAW264 cells were infected and cultured ± poly(I:C) to induce IFN-I as in a. Supernatants were assayed for infectious virus by plaque assay. Symbols show mean ± SEM of triplicate cultures. IFN-I induction significantly reduced virus titers from d1 onwards (p<0.03). c. Mice were given i.n. MuHV-4 (3x10 4 p.f.u.), with or without poly(I:C) inoculation (50μg i.p. and i.n., 6h before and at the time of virus inoculation) as a positive control of IFN-I induction. After 1d bronchial washes were assayed for IFNβ by ELISA. Each point shows 1 mouse. Bars show mean ± SEM. The dashed line shows the limit of assay sensitivity. Poly(I:C) significantly increased IFNβ production in infected mice (p<0.01). d. Mice were infected i.n. or not with MuHV-4 with or without poly(I:C) treatment as in c. 1 and 3d later Mx1 mRNA in lungs was quantitated by real time PCR and normalized by cellular nidogen-1 copy number. Numbers show the fold increase in Mx1 copy number of treated mice over the untreated controls (dashed horizontal line). Bars show mean ± SEM of 3 mice. Mx1 copies were significantly increased by poly-IC at d1 (p<0.05) but not at d3. e. Mice were infected i.n. with MHV-LUC, and treated or not with poly(I:C) as in c. showed little fluorochrome switching (Fig 2A); those recovered from mediastinal lymph nodes (MLN) at d10 were significantly more switched; and those recovered from spleens at d14 were more switched again. Upper respiratory tract inoculation similarly gave little virus switching in noses at d5; significantly more in superficial cervical LN (SCLN) at d5; and significantly more again in SCLN at d14. Therefore exposure to IFN-I was associated with lymphoid rather than epithelial infection.
As lymphoid infection follows epithelial infection, greater fluorochrome switching in spleens than in lungs could also have reflected time-dependent IFN-I induction. To analyse spleen infection without a preceding lung infection, we gave MHV-RG to Mx1-cre mice i.p., when it reaches splenic MZ macrophages directly-presumably via the thoracic duct and blood [3]. At d3 after i.p. inoculation, splenic virus was significantly more switched than that recovered from lungs at d3 after i.n. inoculation (Fig 2C). This result suggested that viral IFN-I f.u. in 5μl without anesthesia) to infect just the nose, then analysed for viral fluorochrome switching as in a. SCLN = superficial cervical lymph nodes. Viruses were recovered from noses by plaque assay and from SCLN by intact cell explant (infectious centre assay). c. Mice were given i.n. or i.p. MHV-RG, with or without poly(I:C) (50μg i.n. or i.p. 6h before and at the time of infection) to maximally induce IFN-I. Viruses recovered 3d later from lungs by plaque assay (i.n.) or from spleens by infectious centre assay (i.p.) were analysed for fluorochrome expression as in a. d. Mice were infected i. n. (lung) or i.p. (spleen) and given poly(I:C) or not as in c. Virus titers were determined 3d later by plaque assay. Circles show individuals, crosses show means. e. Mx1-cre mice were depleted or not of pDCs with mAb 120G8, then infected i.p. with MuHV-RG (10 5 p.f.u.). 5d later spleens were titered for total recoverable virus by explant of intact splenocytes onto BHK-21 cells. Infectious centres (ICs) were also typed as mCherry + (unswitched) or GFP + (switched). % eGFP + = % of total plaques that were switched. Bars show mean ± SEM, other symbols show individual mice. pDC depletion had no significant effect. exposure is intrinsically greater in the spleen. Poly(I:C), which is used widely to activate the Mx1-cre transgene, increased virus switching in each site, with again significantly more switching in spleens. It inhibited viral replication only in lungs ( Fig 2D). IFN-I responses should drive viral fluorochrome switching but inhibit replication; viral inhibition of IFN-I production or signalling should preserve replication but limit switching; high level switched virus recovery from spleens implied both IFN-I induction and resistance to its effects.
pDCs are important IFN-I producers in some settings [7]. Depleting them with mAb 120G8 (Fig 2E) had no significant effect on the switching or titer of i.p.-inoculated MHV-RG. The same treatment increased significantly the titer of footpad-inoculated MuHV-4 (S3 Fig), so it was functionally effective. Thus, the IFN-I response to spleen infection did not depend strongly on plasmacytoid DCs.

MuHV-4 resists IFN-I in B cells
The M3 promoter used to drive MuHV-4 fluorochrome expression is active in the lytic cycle [40]. It is transcribed in infected lungs and acutely infected lymphoid tissue [41,42], including flow cytometrically sorted lymphoid and myeloid sub-populations [43], and it has revealed AMs and AEC1 colonization in the lungs [29], and MZ macrophage and B cell colonization in the spleen [3]. Therefore it identifies not only any virus reactivating ex vivo (Fig 2), but also the cells known to be acutely infected in vivo (Fig 3). In infected lungs, >99% of fluorescent cells were unswitched (mCherry + eGFP -). Even when IFN-I was induced with poly(I:C), essentially all the cells infected at d4 were unswitched. Fig 3A shows examples of staining; Fig 3B shows pooled results. At d5, most fluorescent cells were unswitched AEC1s (mCherry + PDP + ), but of fluorescent AMs (CD68 + or CD169 + ) 50% were eGFP + . By d6, all fluorescent cells were again mCherry + . Therefore MuHV-4 initiated lytic gene expression in IFN-I responding AMs, but the subsequent loss of eGFP + AMs, lack of eGFP spread to AEC1s ( Fig 3B) and low switching of recovered virus (Fig 2) argued that these cells poorly supported new virion production. The lack of eGFP + AEC1s after poly(I:C) treatment suggested that these cells made little Mx1 response, although mCherry + cells might still produce eGFP + virions, as switching could occur after viral fluorochrome expression.
Virus tagging is most informative acutely, as the site of tagging is then clearer. Thus to track IFN-I exposure in spleens we gave mice MHV-RG i.p. for direct infection, as in Fig  To test more stringently the capacity of MuHV-4 to resist IFN-I in spleens, we induced it with poly(I:C) (Fig 3E and 3F). Now at d3 50% of fluorescent macrophages were eGFP + (switched), and at d3-4 >90% of fluorescent B cells were eGFP + . >80% of these were in the MZ and stained for IgM ( Fig 3G). As in lungs, the number of switched macrophages declined after d3, and the number of fluorescent B cells declined after d4, consistent with M3 transcription being silenced as splenic colonization shifts to latency in the white pulp (WP) [3]; but the extensive switching of viruses recovered from spleens at d14 after i.n. infection (Fig 2A and 2B) indicated that IFN-I-exposed genomes remained reactivation-competent.

MuHV-4 induces IFN-I responses in B cells but not AEC1s
The Mx1-cre transgene has been used extensively to delete floxed cellular genes via poly(I:C) injection. We reasoned that it could also reveal cellular exposure to IFN-I, via activation of a floxed cellular fluorochrome. To this end we crossed Mx1-cre with ROSA26-YFP mice, in which cre-dependent removal of a floxed translational stop activates YFP expression, and infected them with wild-type MuHV-4 (Fig 4). Infected lungs contained few YFP + cells, even after IFN-I induction with poly(I:C) ( Fig 4A). The YFP + cells were all AMs: none showed the characteristic cytoplasmic extensions of AEC1s and YFP failed to co-localize with PDP, despite AEC1s being the main site of lung infection [29]. Thus, viral fluorochrome switching in the lungs was limited both by poor IFN-I induction, and-as IFNβ was readily detected in lung washes after poly(I:C) treatment ( Fig 1C)-also by the lack of Mx1 response made by AEC1s. The Mx1 response of lung cells evident in Fig 1D was presumably made by AMs.
Infected spleens contained contrastingly large numbers of YFP + macrophages (F4/80 + , CD169 + ) and B cells (B220 + ) ( Fig 4B). However YFP expression in macrophages was not specific to infection, as it was widespread also in naive (and specific pathogen-free) Mx1-cre x ROSA26-YFP mice (Fig 4C). The original description of Mx1-cre mice showed little spontaneous cre expression [35]. However the marker used-floxed DNA polymerase inactivation-might have been subject to negative selection. Our evidence of constitutive Mx1 transcription in macrophages was consistent with the significant role IFN-I plays in normal myeloid cell differentiation [44]. YFP + B220 + B cells were contrastingly uncommon in naive Mx1-cre x ROSA26-YFP mice and increased >5-fold by infection. IgM + YFP + cell numbers also increased significantly after MuHV-4 infection or poly(I:C) injection (Fig 4D and 4E). Therefore while it was not possible to infer viral exposure to IFN-I in macrophages, cellular fluorochrome switching supported the idea that MuHV-4 is little exposed to IFN-I signalling in AEC1s and abundantly exposed in B cells.

Viral IFN-I evasion promotes splenic infection
B cell colonization despite viral exposure to IFN-I implied an important role for IFN-I evasion. The MuHV-4 ORF36 inhibits IRF3 signalling, and ORF36 -MuHV-4 delivered i.n. shows an IFNAR-dependent infection defect in lungs and spleens [9]. Because spleens are colonized down-stream of lungs, direct and indirect effects on spleen infection are hard to distinguish by i.n. inoculation. Therefore we compared ORF36 -MuHV-4 with wild-type also by i.p. inoculation, which reaches the spleen directly [3].
Our ORF36 -MuHV-4 showed a relatively minor defect in direct lung infection when given i.n. (Fig 5A), and a marked defect in direct spleen infection when given i.p. (Fig 5B). IFN-I induction with poly(I:C) increased the lung infection defect (Fig 5A). Thus in lungs, limited . 5d later lung sections were stained for viral eGFP / mCherry, and the cell markers podoplanin (PDP, AEC1s), CD68 (macrophages) and CD169 (AMs). Each image is representative of 5 mice. Arrows show example positive cells. Quantitation is shown in b. The cytoplasmic rather than nuclear distribution of eGFP is typical for the relatively mild fixation of periodate-lysine-1% formaldehyde. b. Lung sections of mice infected as in a (5 per group) were analysed at d4-6, counting eGFP + (green) and mCherry + (red) PDP + and CD68 + cells for 3 sections per mouse. Each section corresponds to approximately 10 fields of view as illustrated in a. Bars show mean ± SEM. At d5 CD68 + cells showed significantly more fluorochrome switching than PDP + cells (p<0.01). c. Mx1-cre mice were infected i.p. with MHV-RG (10 5 pfu), and eGFP + (green) and mCherry + (red) MZ macrophages (CD169 + ) and B cells (B220 + ) counted 4d later. Bars show mean ± SEM counts for 3 sections each of 3 mice per group. B220 + cells were significantly more switched than unswitched, whereas CD169 + were significantly more unswitched than switched (p<0.02). d. IFN-I induction normally allows ORF36 -MuHV-4 to achieve near wild-type titers; but in spleens, where IFN-I induction was greater, ORF36 was an important outcome determinant. Poor splenic infection by ORF36 -MuHV-4 was associated with a shift in viral antigen from MZ (CD169 + ) to red pulp (RP, F4/80 + ) macrophages (Fig 5C and 5D). Impaired virus transfer to the WP provided an explanation for the reduction in titer, as lymphoproliferation in the WP normally amplifies the viral load. Virus deposition in the RP was consistent with virions being carried by blood flow from the MZ when transfer to the WP was impaired.

IFN-I blockade increases MuHV-4 replication in macrophages
Although MuHV-4 colonizes IFN-I-responding mice, i.n. and i.p. inoculations are more pathogenic in IFNAR -/mutants [16,45], so IFN-I must normally exert some restriction on virus replication. Which cell types support the additional infection when IFN-I is lacking has been unclear. Blocking IFN-I signalling with an IFNAR-specific antibody increased acute lung infection <5-fold (Fig 6A), consistent with little IFN-I induction in this site (Fig 1); by contrast spleen increased infection >50-fold. Immunostaining revealed extensive viral lytic spread through the splenic MZ and RP, with increased viral antigen expression in cells morphologically typical of macrophages ( Fig 6B). Splenic macrophages are diverse and lack a single unifying marker. Viral antigens were evident in CD169 + (MZ macrophage) (Fig 6C and 6D), CD206 + (tissue resident, non-MZ macrophage) and F4/80 + (RP macrophage) populations ( Fig  6E), so IFNAR blockade made many macrophage subtypes more permissive for lytic infection. Viral staining in the WP remained low (Fig 6D). Thus in the absence of IFN-I, MZ to WP transfer became rate-limiting for virus spread, and most infection was diverted to the RP, again consistent with untransferred MZ virus following splenic blood flow.
We tracked latent infection in spleens by viral eGFP expression from a constitutive promoter (Fig 6F). Again blocking IFNAR increased infectious virus titers at d4. Total recoverable virus (infectious centre assay) increased similarly to infectious virus, consistent with most early spleen infection being lytic in MZ macrophages [46]. By d7 after i.p. challenge, when spleen infection is largely latent in B cells [46], virus titers remained elevated above controls in IFNAR-blocked mice; however the elevation was no greater than at d4 (Fig 6F), implying that IFN-I limited mainly MZ macrophage infection. Staining spleen sections at d7 (Fig 6G and  6H) showed more WP infection (eGFP + ) in IFNAR-blocked mice than in controls, so more . IFN-I was induced or not with poly(I:C) (50μg i.n. and i.p., 6h before and at the time of infection). Lungs were harvested at d3 and d5 and stained for YFP, CD68 and MuHV-4 antigens. Nuclei were stained with DAPI. White arrows show example YFP + CD68 + MuHV-4 + AMs; open arrows show YFP + CD68 + AMs; Grey-filled arrows show YFP -MuHV-4 + AEC1s. No AEC1s were YFP + . All YFP + cells were AMs (n>30). Images are representative of 5 sections from each of 3 mice per group. b. Mx1-cre x ROSA26-YFP mice were infected i.p. with MuHV-4 (10 5 p.f.u.). IFN-I was induced or not with poly(I:C) (50μg i.p., 6h before and at the time of infection). D5 spleen sections were stained for YFP and CD169 (MZ macrophages), F4/80 (RP macrophages) or B220 (B cells). Nuclei were stained with DAPI and the cells visualized by confocal microscopy. Arrows show example YFP + cells expressing the relevant cellular marker. For CD169 staining a dashed line shows the boundary between the B cell-dominated WP and the macrophage-dominated MZ, with CD169 + MZ macrophages lying adjacent to WP B cells. For B220 staining after virus + poly(I:C), separate channels are shown to make clear the extensive YFP expression in B cells. Images are representative of 5 sections from each of 3 mice per group. >50% of each macrophage population was YFP + , with no difference between poly-IC treated and untreated. In WP follicles, 10-50% of B220 + B cells were YFP + after poly(I:C) treatment and 5-20% were YFP + without poly(I:C) treatment. Quantitation for IgM + B cells is shown in e. c. Naïve C57BL/6 and Mx1-cre x ROSA26-YFP spleen sections were stained for YFP, B220 and F4/80. YFP expression was evident in >80% of F4/80 + macrophages and <5% of B220 + B cells. Dashed lines show the MZ demarcation between F4/80 + RP macrophages and B220 + WP B cells. lytic infection in MZ macrophages eventually fed through to more latent infection in WP B cells. However IFN-I blockade did not increase viral antigen + / viral eGFP + cell ratios in the WP at d7-in fact they were reduced-so the proportion of lytic infection in WP B cells did not increase. We conclude that splenic macrophage infection was strongly restricted by IFN-I, but that acute viral reactivation in WP B cells was largely IFN-I-independent.
IFN-I principally targets viral lytic infection. Therefore it possibly had least effect on B cell infection because this is mostly latent. MuHV-4 that cannot shut down lytic infection due to an additional promoter element inserted upstream of its ORF50 lytic switch gene (M50) replicates normally in mice for 3d, but is then progressively attenuated [47]. This virus cannot drive B cell proliferation because it is constitutively lytic, but it is attenuated also in the lungs, where AEC1s and alveolar macrophages are infected. That IFNAR deficiency does not significantly increase MuHV-4 lung infection for at least 4d [16] suggested that M50 virus attenuation might be due to impaired IFN-I evasion. To test this hypothesis we gave mice IFNAR blocking antibody i.p. then wildtype or M50 MuHV-4 i.n. (10 5 p.f.u.). Lung virus titers after 7d showed restoration of the M50 replication defect by IFN-I blockade (S4 Fig). Thus, a capacity to establish latency was important for IFN-I evasion.

Murine cytomegalovirus also passes through IFN-I-responding cells
MuHV-4 passage through IFN-I responding cells was surprising: we expected that either viral evasion would prevent an IFN-I response in infected cells, or IFN-I would prevent viral replication. To test whether it was a unique to MuHV-4, we infected Mx1-cre mice with murine cytomegalovirus that cre switches from eGFP to tdTomato expression (MCMV-GR). We induced IFN-I or not with poly(I:C), gave MCMV-GR i.p., and recovered infectious virus from livers and spleens by plaque assay at d3 (Fig 7A). As with MHV-RG, IFN-I induction with poly(I:C) increased MCMV-GR switching (Fig 7B) without reducing titers. Thus MCMV, which like MuHV-4 inhibits IFN-I induction and signalling [48], could also pass through IFN-I-responding cells. Spleen sections (Fig 7C and 7D) revealed eGFP -tdTomato + cells around lymphoid follicles, consistent with MCMV infecting IFN-I-responding MZ macrophages [49]. Liver sections of (Fig 7D and 7E) showed CD68 -tdTomato + cells with the morphology of hepatocytes. Infected cells were more switched than recovered virions, consistent with IFN-I exerting some restriction on virion production [50], but productive infection was clearly possible in IFN-I-responding cells.

Discussion
MuHV-4 provides an experimental window onto the γ-herpesviruses, whose colonization of lymphoid tissue directly confronts host immune defences. IFN-I attenuated macrophage infection, whereas B cell infection was protected by viral evasion. This cell type-dependent outcome explained how IFN-I and its evasion both control infection, with each dominating in a different setting (S5 Fig). Together they played a significant role in shifting the focus of viral tropism from macrophages acutely to B cells chronically.
In the lungs, AMs provide MuHV-4 with a gateway to AEC1s, which then support lytic replication [29]. AEC1s made no detectable Mx1 response to infection or to poly(I:C). Thus, although IFN-I responding AMs poorly supported virus spread, inducing or blocking IFN-I or disrupting its evasion all had modest effects because AEC1s still allowed virus replication. Splenic MZ macrophages provide a gateway to MZ B cells. Again IFN-I restricted virus spread Virus in lungs (i.n.) or spleens (i.p.) was titered by plaque assay after 4d. Circles show individuals, bars show means. The IFNAR blockade significantly increased virus titers in both sites, but the difference was significantly greater in spleens (p<0.01). b. Spleens of mice infected i.p. as in a were stained for MuHV-4 antigens. Nuclei were stained with DAPI. Dashed lines show approximate MZ / WP boundaries, based on CD169 staining of adjacent sections. IFNAR blockade increased MZ infection >100 fold and WP infection <5-fold. In the MZ >90% of viral antigen + cells had typical myeloid cell rather than B cell morphology. c. Co-staining of a representative anti-IFNAR-treated, i.p.-infected mouse after 4d shows extensive viral antigen expression in CD169 + splenic MZ macrophages (arrows). Quantitation is shown in d. d. Quantitation of MZ and WP viral antigen staining for spleen sections from 5 mice per group, treated or not with anti-IFNAR antibody and infected i.p. as in a. Bars show group means, other symbols show mean counts for 3 sections per mouse. IFNAR blockade significantly increased viral antigen expression in all populations (p<0.05), most markedly in CD169 + MZ macrophages (p<0.01). e. Spleen sections of mice infected as in b, showing additional co-localization of viral antigen with CD206 + and F4/80 + macrophages after IFNAR blockade (arrows). Again IFNAR blockade significantly increased MuHV-4 antigen expression in these populations relative to no antibody controls (p<0.01). f. Mice were given anti-IFNAR antibody or not (control) and 2d later infected i.p. with MHV-GFP. Spleens were titered for infectious virus by plaque assay after 4d and for total reactivatable virus by infectious centre (IC) assay after 4 and 7d. Circles show individuals, horizontal bars show means. g. Spleens of mice infected as in f were analysed at d7 for viral eGFP and antigen expression in CD169 + macrophages and B220 + B cells. Bars show mean ± SEM of 5 mice. IFNAR blockade significantly increased viral eGFP expression in all populations (p<0.05), most markedly in WP B cells (p<10 −4 ). h. Example images of mice infected as in f, stained for viral eGFP (green) and cell markers (red). Nuclei were stained with DAPI (blue). The dashed lines show approximate WP / MZ boundaries, based on CD169 staining. Arrows show example eGFP + cells. Quantitation is shown in g. An important role for the ORF36 IFN-I evasion gene in splenic infection was surprising, as its inhibition of IRF3 should limit IFN-I induction, and the abundant exposure of wild-type splenic virus to IFN-I implied that MuHV-4 evades its effector functions rather than relying solely on limiting induction. A possible explanation is that ORF36, as a lytic gene, operates more in macrophages than in B cells, as macrophages support lytic infection [43] and are important IFN-I producers, whereas B cells support a more tightly latent infection and function more prominently as IFN-I responders. IFN-I production by macrophages depends on positive feedback through IFNAR [51]. Thus ORF36, by limiting IFN-I production, should limit IFN-I signalling by infected macrophages both to themselves and to B cells. This would promote macrophage to B cell virus transfer: first, by promoting new virion production in macrophages; and second, by reducing the effector evasion required for those virions to productively enter B cells. The importance for splenic infection of IFNAR degradation by ORF54 [12] suggests that it may play a similar role. Plasmacytoid DC produce IFN-I independently of feedback through IFNAR [7], so their production would not be affected by ORF36 or ORF54. They (B220 + DC) are not an acute infection target in spleens [3]. However they contributed relatively little to virus switching or control, so other evasion mechanisms may limit their scope.
The recovery of fluorochrome-switched virions implied that viral IFN-I evasion also operates down-stream of IFN-I induction or signalling, that is downstream of ORF36. Switching was evident early after i.p. virus inoculation, when splenic infection is mostly lytic [46]. Established lytic infection would inhibit IFN-I signalling, so we envisage that most switching occurred when virions entered cells already making IFN-I responses. In this context the disassembly of IFN-I induced ND10 domains [52] may be an important evasion mechanism. This is a function of the MuHV-4 ORF75c tegument component [53,54] and is conserved in the homologous Kaposi's Sarcoma-associated Herpesvirus ORF75 [55] and Epstein-Barr virus BNRF1 [56]. Tagged viral genomes could then remain latent until IFN-I responses have subsided [14]. Therefore, pace cell differences in IFN-I response, macrophage infection may be more susceptible than B cell infection to inhibition by IFN-I because it is more lytic [43,46].
Together the IFN-I response and its evasion promoted B cell over macrophage infection, making IFN-I an important determinant of in vivo viral tropism. A corollary is that IFN-Ibased therapies are likely to have only a small window of efficacy-mainly reducing acute gamma-herpesvirus replication in myeloid cells. Downstream of this, virions entering IFN-Iexposed B cells would still establish a viable infection. IFN-I deficiency increases ex vivo MuHV-4 reactivation rates [14]. However the explant reactivation assay is complicated, relying on plaque formation in complex in vitro cultures that contain both virus-infected and immune cells. Reactivation rates are low, particularly from B cells [43], and may be affected by cell viability, in vitro antibody production and cytokines. The failure of IFNAR blockade to increase the viral antigen / eGFP staining ratio of d7 splenic WP B cells argued that here IFN-I is not a major regulator of in vivo lytic reactivation. The capacity of MCMV also to re-emerge acutely after Mx1-dependent tagging suggested that many herpesviruses can enter IFN-I responding cells and rapidly re-emerge. How far IFN-I can restrict MCMV was not explored, but clearly there are limits on what it alone is likely to achieve.

Virus assays
To titer infectious virus, culture-grown stocks or freeze-thawed organ homogenates were plated on BHK-21 (MuHV-4) or embryonic fibroblast (MCMV) monolayers [61]. To titer total reactivatable MuHV-4, organs were disrupted into single cell suspensions then plated on BHK-21 cells. The cells were cultured in complete medium for 3h, overlaid with complete medium plus 0.3% carboxymethylcellulose, cultured for 4d, then fixed with 1% formaldehyde and stained with 0.1% toluidine blue. To measure viral fluorochrome switching, plaque assays were performed at limiting dilution, with 16 replicate wells per dilution. After 4d wells were scored for green (eGFP) and red (mCherry or tdTomato) fluorescence to derive virus titers for each colour., with % switching = 100 x switched plaque titer / (switched plaque titer + unswitched plaque titer).