SERINC5 Potently Restricts Retrovirus Infection In Vivo

This study examined for the first time the in vivo function of the serine incorporator (SERINC) proteins during retrovirus infection. SERINC3 and SERINC5 (SERINC3/5) restrict a number of retroviruses, including human immunodeficiency virus 1 (HIV-1) and murine leukemia virus (MLV), by blocking their entry into cells. Nevertheless, HIV-1 and MLV encode factors, Nef and glycosylated Gag, respectively, that counteract SERINC3/5 in vitro. We recently developed SERINC3 and SERINC5 knockout mice to examine the in vivo function of these genes. We found that SERINC5 restriction is dependent on the absence of glycosylated Gag and the expression of a specific viral envelope glycoprotein. On the other hand, SERINC3 had no antiviral function. Our findings have implications for the development of therapeutics that target SERINC5 during retrovirus infection.

virions (3,4) and block the step of HIV envelope fusion and pore formation with the target cell membrane (7,8).
MLV is divided into different subgroups on the basis of host range (9). Ecotropic MLVs, such as Friend MLV (F-MLV), infect only mouse cells; amphotropic MLVs, such as 4070A, infect both human and mouse cells; and xenotropic MLVs infect nonmurine cells (10). Glycosylated Gag (glyco-Gag) is a viral protein expressed by both the ecotropic and amphotropic MLVs (11). Glyco-Gag expression is initiated from a CUG upstream of the polyprotein's AUG, is highly glycosylated (12)(13)(14)(15), and is cleaved into two fragments by an unknown protease (14,16). The N-terminal fragment is a transmembrane protein, and the C-terminal fragment is secreted from the infected cell (12,17,18). While glyco-Gag is considered to be dispensable in vitro, glyco-Gag-deficient viruses exhibit lower infectivity in vivo (19)(20)(21). Glyco-Gag is important in vivo, because it protects the virus from both the deleterious effects of mouse apolipoprotein B editing complex 3 (APOBEC3 [A3]), a potent antiretroviral factor, and induction of the type I interferon (IFN) response (11,(22)(23)(24). Glyco-Gag antagonizes the antiviral function of SERINC3 and SERINC5 in vitro by blocking the incorporation of SERINC proteins into the budding virions, leading to their lysosomal degradation (3,4,25,26). Whether SERINC5 restricts retrovirus infection in vivo in a glyco-Gag-dependent manner is currently unknown.
While much work has been performed to understand the role of SERINC proteins in retrovirus infection in vitro, nothing is known about the antiretroviral function of these proteins in vivo. A host factor that has an antiretroviral function in vitro does not necessarily mean that it can restrict retrovirus infection in vivo. For example, SAMHD1, a potent retroviral factor, did not affect retrovirus infection in vivo (27). Here, for the first time, we examine the antiretroviral effect of SERINC5 in vivo and show that mouse SERINC5 (mSERINC5) restricts MLV infection in vivo. Moreover, SERINC5 restriction in vivo is influenced not only by the presence of glyco-Gag but also by the virus envelope. SERINC5 had no effect on F-MLV infectivity even when glyco-Gag was mutated; however, it was only when we replaced the F-MLV envelope with the amphotropic MLV 4070A envelope that we found that SERINC5 restricted MLV infection in a glyco-Gagdependent manner. Finally, unlike human SERINC3, mouse SERINC3 has no antiretroviral function either in vivo or in vitro.

RESULTS
Murine SERINC3 and SERINC5 are expressed in murine leukocytes and are not induced by F-MLV infection. To elucidate whether the mouse SERINC genes are expressed in leukocytes, cells that are naturally infected by F-MLV in mice (23,28,29), we isolated peripheral blood mononuclear cells (PBMCs) from 3 adult and 3 neonate BL/6 mice (C57BL/6N-wild-type mice) and performed cell sorting for T cells, B cells, and dendritic cells (DCs). Transcription levels of SERINC genes were determined by real-time quantitative PCR (RT-qPCR). We observed that only mSERINC1 RNA, mSERINC3 RNA, and mSERINC5 RNA were detected in both adults and neonates, while mSERINC2 and mSERINC4 were absent from these cell populations ( Fig. 1A to C). To determine the effect of F-MLV infection on mSERINC3 and mSERINC5 transcription, we infected EL-4 cells (T lymphoblast cell line) and MutuDC1940 cells (a murine DC cell line [30]) with F-MLV wild type (F-MLV WT) (0.1 multiplicity of infection [MOI]) followed by RT-qPCR to determine mSERINC3 and mSERINC5 transcription levels over time. No changes in the transcription levels of mSERINC5 and mSERINC3 were observed during infection ( Fig. 1D and E). Thus, we concluded that F-MLV infection does not affect mSERINC3 and mSERINC5 transcription. To determine if type I interferons (IFNs) induce the expression of SERINC genes in vivo, we treated BL/6 mice with polyinosinic:polycytidiylic acid [poly(I·C)] to mimic type I IFN stimulation. No increases were detected in the expression levels of any of the SERINC genes in the spleens of the poly(I·C)-treated mice (see Fig. S1 in the supplemental material). Thus, mouse SERINC genes, similarly to human SERINC genes (3), are not upregulated by type I IFNs.
Murine SERINC3 and SERINC5 have no effect on ecotropic MLV infection in vitro. While the antiretroviral role of human SERINC3 and SERINC5 has been extensively examined (3)(4)(5), there is little information regarding the role of mSERINC3 and mSERINC5. To examine the effect of glyco-Gag on mSERINC3 and mSERINC5 during infection, we generated two F-MLV infectious clones with mutations in the glyco-Gag gene (Fig. 3A), including one with two stop codons at amino acids 32 and 55 of glyco-Gag (gGag Ϫ F-MLV); previous work showed that introducing stop codons at these residues affected only glyco-Gag translation and not Gag translation (5). For the second virus (gGag mut F-MLV), residues P31, Y36, L39, and R63, which are critical for the anti-SERINC5 function of glyco-Gag, were mutated to alanine (5,25,31). NIH 3T3 cells, which express very low levels of mouse SERINC3 and SERINC5 (Fig. S3), were infected at MOI of 0.1 with gGag Ϫ F-MLV, gGag mut F-MLV, and F-MLV WT, and virus yields were determined at various times postinfection by p30 core antigen enzyme-linked immu-nosorbent assay (p30-CA ELISA), while MLV DNA levels in the infected cells were measured by RT-qPCR. gGag Ϫ F-MLV and gGag mut F-MLV did not differ from the F-MLV WT in either p30 or MLV DNA levels ( Fig. 3B and C); thus, glyco-Gag is not essential for ecotropic MLV replication and infectivity in vitro (5,20,32,33). Glyco-Gag blocks human SERINC5 and SERINC3 incorporation into nascent virions (3,4,34). To determine whether the glyco-Gag mutant viruses that we generated would affect the incorporation of mSERINC3 and mSERINC5 into budding MLV particles, we cotransfected 293T cells with either F-MLV WT or the F-MLV constructs with mutations in the glyco-Gag gene and with either mSERINC3 or mSERINC5. The virus released was examined for mSERINC3 and mSERINC5 content by Western blotting. We found that the gGag Ϫ F-MLV and gGag mut F-MLV particles had higher levels of mSERINC3 and mSERINC5 incorporated into their virions than the F-MLV WT particles (Fig. 3D). Thus, we concluded that glyco-Gag blocks mSERINC3 and mSERINC5 incorporation into the nascent virions.
We next examined the effect of mSERINC3 and mSERINC5 on F-MLV infection. Previous work had shown that although hSERINC3 and hSERINC5 did not affect virus release from producer cells, they blocked MLV infectivity in a glyco-Gag-dependent manner in target cells (3-5, 25, 34). To determine whether mSERINC3 and mSERINC5 packaged in the MLV particles blocked MLV infection, we infected Mus dunni cells with equal amounts of F-MLV WT, gGag Ϫ F-MLV, and gGag mut F-MLV generated in the presence of either mSERINC3 or mSERINC5. DNA was isolated from the infected cells, and viral DNA levels were measured by RT-qPCR. Interestingly, F-MLV infectivity was unaffected regardless of the levels of mSERINC3 or mSERINC5 incorporated into the nascent virions (Fig. 3E), demonstrating that while mSERINC3 and mSERINC5 were excluded from F-MLV in a glyco-Gag-dependent manner, these two host factors had no effect on F-MLV infectivity.
Murine SERINC5 does not restrict ecotropic MLV infection in vivo. While mSERINC3 and mSERINC5 did not affect F-MLV infectivity when glyco-Gag was mu- The genomic fragment between the two gRNAs was excised using CRISPR/Cas9 technology, resulting in the abrogation of the SERINC5 gene. Locations of the SERINC5 primers used for RT-PCR analysis are indicated by the "F" and "R" purple arrows. (B) Fold change of SERINC5 expression in the SERINC5 Ϫ/Ϫ and SERINC5 ϩ/Ϫ mice relative to C57BL/6 mice. SERINC5 RNA levels in the SERINC5 Ϫ/Ϫ mice (n ϭ 5), SERINC5 ϩ/Ϫ mice (n ϭ 6), and C57BL/6 mice (n ϭ 6) were determined by RT-qPCR and normalized to ␤-actin. (C) PBMCs from 4 C57BL/6 mice and 6 SERINC5 Ϫ/Ϫ mice were stained with anti-CD45R/B220 (B cells), anti-CD3 (T cells), anti-F4/80 (Macrophages), and anti-CD11c (Dendritic cells) antibodies and subjected to FACS analysis. Values are presented as means Ϯ SD. There were no significant differences in the percentages of T cells, B cells, macrophages, and dendritic cells among the C57BL/6 and SERINC5 Ϫ/Ϫ mice. For the data shown in panel B, statistical significance was determined by one-way ANOVA and Tukey's test; for the data shown in panel C, statistical significance was determined by an unpaired t test, **, P Ͻ 0.01; ****, P Ͻ 0.0001. S5, SERINC5; BL/6, C57BL/6. tated, we speculated that we might observe something different in vivo. In vitro data do not necessarily reflect what happens in vivo. MLV infects a variety of cells during in vivo infections, and involvement of the immune system, due to its activation by MLV infection, might result in changes in the levels of host genes such as the SERINC5 gene that might affect the outcome of infection. To determine the role of SERINC5 in retrovirus infection in vivo, we infected the SERINC5 Ϫ/Ϫ mice we developed with F-MLV WT, gGag Ϫ F-MLV, and gGag mut F-MLV, and virus titers in their spleens were measured at 10 days postinfection (dpi). We found that the BL/6 (SERINC5 ϩ/ϩ ) mice and SERINC5 Ϫ/Ϫ mice infected with F-MLV WT had similar virus titers in their spleens at 10 dpi (Fig. 4). Interestingly, gGag Ϫ F-MLV and gGag mut F-MLV also replicated to similar levels in the spleens of the BL/6 and SERINC5 Ϫ/Ϫ mice, albeit both viruses grew at significantly lower levels than F-MLV WT (Fig. 4).
As a positive control for our in vivo experiments, we used APOBEC3 Ϫ/Ϫ (A3 Ϫ/Ϫ ) mice, where a glyco-Gag mutant MLV has been shown to replicate similarly to wild-type MLV in the spleens of the infected mice (11,22,24). We infected APOBEC3 Ϫ/Ϫ mice with F-MLV WT, gGag Ϫ F-MLV, and gGag mut F-MLV, and, similarly to what had been previously reported (11,22,24), the three viruses replicated to similar levels in the spleens of the infected APOBEC3 Ϫ/Ϫ mice (Fig. 4).
Thus, we concluded that in the context of the ecotropic F-MLV, SERINC5, unlike APOBEC3, exerts no in vivo antiviral function regardless of the presence of glyco-Gag.
The amphotropic MLV envelope renders MLV vulnerable to SERINC5 restriction in a glyco-Gag-dependent manner in vitro. It was previously shown that envelopes , and MLV DNA levels were measured by RT-qPCR. Viral DNA levels normalized to GAPDH were used to calculate the percentage (%) of relative infectivity with respect to F-MLV WT virus produced in the presence of empty vector. All results are presented as means Ϯ SD. Statistical significance was determined by unpaired (two-tailed) t test for data points at 5 dpi (B and C) and by one-way ANOVA and Tukey's test (E). ns, not significant. Results are shown for n ϭ 3 independent experiments in panels A to E; representative immunoblotting results are shown in panels A and D. mS3, mouse SERINC3; mS5, mouse SERINC5; E.V., empty vector; gGag, glyco-Gag.

SERINC5 Antiretroviral Role in Mice
® from different HIV subtypes were able to overcome SERINC5 restriction without preventing SERINC5 incorporation into the virions (35,36). Therefore, virus susceptibility to human SERINC3 and SERINC5 restriction is dependent not only on glyco-Gag but also on the viral envelope (2)(3)(4)(5)35). We thus hypothesized that the envelope of MLV might play a crucial role in mSERINC5-mediated restriction in vivo. The amphotropic MLV envelope, just like the xenotropic MLV envelope, rendered MLV pseudoviruses sensitive to SERINC5 restriction (5,33), and the glyco-Gag sequences of amphotropic MLV and ecotropic MLV are 90% identical (data not shown). We developed chimeric F-MLV constructs with the aforementioned mutations in glyco-Gag, but the envelope was replaced with that of amphotropic MLV strain 4070A (Fig. 5A)-gGag Ϫ F-MLV/ Ampho Env and gGag mut F-MLV/Ampho Env . As a control, we also replaced the envelope of F-MLV WT with that of amphotropic 4070A-F-MLV/Ampho Env (Fig. 5A). To verify similar levels of incorporation of the amphotropic MLV envelope glycoproteins into the nascent virions generated by the three chimeric constructs, we performed Western blotting on purified virions and probed for gp70 and p15E, the two subunits of the MLV envelope glycoprotein. We found that the three chimeric viruses had similar levels of amphotropic MLV envelope glycoproteins on their virions (Fig. 5A, right panel). To determine whether glyco-Gag still affected mSERINC3 and mSERINC5 incorporation into virions with the amphotropic envelope, we cotransfected 293T cells with the chimeric envelope viral constructs and with either mSERINC3 or mSERINC5. Western blotting of purified virions showed that mSERINC3 and mSERINC5 were incorporated in a glyco-Gag-dependent manner, whereas the virions where glyco-Gag was either absent or mutated (gGag Ϫ F-MLV/Ampho Env and gGag mut F-MLV/Ampho Env ) had higher mSERINC3 and mSERINC5 levels packaged than the virions with intact glyco-Gag (F-MLV/Ampho Env ) (Fig. 5B). Thus, we concluded that glyco-Gag blocks the incorporation of mSERINC3 and mSERINC5 into the viral particles independently of the virus envelope.
We then performed growth curve analyses of the different chimeric viruses that we had generated. NIH 3T3 cells were infected at MOI of 0.1, and virus levels were determined at various times postinfection by p30-CA ELISA and by RT-qPCR measuring MLV DNA levels in the infected cells. Remarkably, in contrast to what we saw with the viruses that had ecotropic MLV envelope, the glyco-Gag mutant chimeric viruses carrying the amphotropic MLV envelope (gGag Ϫ F-MLV/Ampho Env and gGag mut F-MLV/ Ampho Env ) replicated at considerably lower levels than the chimeric virus carrying an  (Fig. 5C and D). Therefore, due to the different growth kinetics of the three viruses (gGag Ϫ F-MLV/Ampho Env , gGag mut F-MLV/ Ampho Env , and F-MLV/Ampho Env ), we have abstained from making any direct comparisons among the different viruses. To examine the effect of mSERINC3 and mSERINC5 on the infectivity of the chimeric viruses, we infected Mus dunni cells with equal amounts of the three chimeric viruses (gGag Ϫ F-MLV/Ampho Env , gGag mut F-MLV/Ampho Env , and F-MLV/Ampho Env ) that were generated in the presence of either mSERINC3 or mSERINC5. We then measured MLV DNA levels by RT-qPCR and found that the glyco-Gag-expressing chimeric virus, F-MLV/Ampho Env , replicated at the same levels in infected Mus dunni cells regardless of the presence of mSERINC3 or mSERINC5 (Fig. 5E). On the other hand, the chimeric viruses in which glyco-Gag was either mutated or deleted (gGag mut F-MLV/Ampho Env and gGag Ϫ F-MLV/Ampho Env ) replicated very poorly in the presence of mSERINC5 ( Fig. 5F and G). Interestingly, incorporation of mSERINC3 into the chimeric virions, even when glyco-Gag was disrupted, had no effect on virus infectivity ( Fig. 5F and G). This is particularly interesting as the human homologue of mSERINC3, human SERINC3, has antiretroviral functions (3,4,37). Thus, we concluded that mSERINC5, similarly to human SERINC5, restricts retrovirus infection in vitro in a glyco-Gag-dependent and envelope-dependent manner whereas mSERINC3 has no antiretroviral function in vitro.
Murine SERINC5 restricts MLV infection in vivo in a glyco-Gag-dependent and envelope-dependent manner. The change in the susceptibility to mSERINC5 restriction seen upon replacing the envelope of the ecotropic F-MLV with that of the amphotropic MLV, 4070A, provides further proof that the envelope of the viral particles is critical for SERINC5 restriction. To examine if the viral envelope affects SERINC5 restriction in vivo, we used the chimeric viruses F-MLV/Ampho Env , gGag Ϫ F-MLV/ Ampho Env , and gGag mut F-MLV/Ampho Env to infect SERINC5 Ϫ/Ϫ and BL/6 mice. Titers in the spleens of the infected mice were measured 10 dpi. In the case of F-MLV/Ampho Env , the SERINC5 Ϫ/Ϫ and BL/6 mice had similar viral titers in their spleens, which is attributed to the fact that the virus has an intact glyco-Gag (Fig. 6A). Interestingly, in the case of the gGag Ϫ F-MLV/Ampho Env , infection was abrogated for all animals independently of genotype (SERINC5 Ϫ/Ϫ , APOBEC3 Ϫ/Ϫ , and BL/6 mice) (Fig. S4). This was quite surprising, as in the context of the ecotropic envelope, deletion of glyco-Gag only resulted in reduced virus titers in the spleens of the infected BL/6 and SERINC5 Ϫ/Ϫ mice ( Fig. 4) and not in abrogation of infection such as we saw with the amphotropicenvelope-containing chimeric virus. Remarkably, in the case of gGag mut F-MLV/Ampho Env , we observed that the virus titers in the spleens of the SERINC5 Ϫ/Ϫ mice were significantly higher than those observed in the BL/6 mice (Fig. 6B). This finding not only demonstrates that mSERINC5 restricts MLV infection in vivo but also that MLV susceptibility to SERINC5 inhibition is dependent on both the viral envelope and the presence of glyco-Gag.
Our in vitro data show that glyco-Gag mutant viral particles with the amphotropic envelope (gGag mut F-MLV/Ampho Env ) have decreased infectivity when mSERINC5 is packaged in the virions. The spleen is one of the major organs that becomes infected by MLV in vivo (11,(38)(39)(40)(41) Thus, viral particles purified from the spleens of infected BL/6 mice should have reduced infectivity compared to those derived from SERINC5 Ϫ/Ϫ infected mice. To test this, we purified virions from the spleens of BL/6 and SERINC5 Ϫ/Ϫ mice infected with either gGag mut F-MLV/Ampho Env or F-MLV/Ampho Env . We initially measured viral RNA levels in the splenic extracts and found that F-MLV/Ampho Env viral RNA levels were similar in BL/6 and SERINC5 Ϫ/Ϫ mice (Fig. 6C) whereas gGag mut F-MLV/ Ampho Env viral RNA levels were higher in the splenic extracts from the SERINC5 Ϫ/Ϫ mice than in those from the BL/6 mice (Fig. 6D), which reflects the higher infection levels of the former (Fig. 6B). Subsequently, we infected Mus dunni cells with equal amounts of gGag mut F-MLV/Ampho Env or F-MLV/Ampho Env viral particles isolated from BL/6 or SERINC5 Ϫ/Ϫ mice after normalizing to viral RNA levels. We measured MLV DNA levels by RT-qPCR and found no difference in the levels of infectivity of F-MLV/ Ampho Env virions isolated from either BL/6 or SERINC5 Ϫ/Ϫ mice (Fig. 6E). On the other (C and D) Newborn mice were infected with F-MLV/Ampho Env (C) or with gGag mut F-MLV/Ampho Env (D), and splenic extracts were prepared in 5 ml (for panel C) or 2 ml (for panel D) of RPMI media 16 days postinfection. Viral RNA copy numbers in the splenic extracts were determined using RT-qPCR. For panel C, n ϭ 4 C57BL/6 mice and n ϭ 8 SERINC5 Ϫ/Ϫ mice; for panel D, n ϭ 3 C57BL/6 mice and n ϭ 3 SERINC5 Ϫ/Ϫ mice. (E and F) Equal amounts of viral RNA copies (from the experiments described in the panel C and D legends) were used to infect Mus dunni cells. Genomic DNA was isolated 19 h postinfection, viral DNA levels were measured by RT-qPCR and normalized to GAPDH, and results are presented. All results are presented as means Ϯ SD. Statistical significance was determined by Mann-Whitney (two-tailed) test (A and B) and by unpaired t test (C to F). *, P Ͻ 0.05; **, P Ͻ 0.01; ****, P Ͻ 0.0001; ns, not significant. S5, SERINC5; BL/6, C57BL/6; SE, splenic extracts; ICs, infectious centers. SERINC5 Antiretroviral Role in Mice ® hand, gGag mut F-MLV/Ampho Env virions isolated from SERINC5 Ϫ/Ϫ mice were more infectious than those isolated from BL/6 mice (Fig. 6F). The absence of an antibody that could detect endogenous levels of mSERINC5 prevented us from examining mSERINC5 incorporation into the nascent virions in vivo. Taken together, these data suggest that mSERINC5 is packaged into virions isolated from the spleens of infected mice in a glyco-Gag-dependent manner and limits viral infectivity.

SERINC5 and APOBEC3 additively act to restrict retrovirus infection in vivo.
Mouse APOBEC3 potently restricts MLV infection in vivo by blocking reverse transcription (11,22,24,38,41,42). SERINC5 blocks virus-cell fusion (7,8). As mouse APOBEC3 and mSERINC5 target different steps of the retrovirus life cycle and both restrict MLV infection in vivo, we asked whether these two factors act additively to block retrovirus infection in vivo. To examine any additive effect between mouse APOBEC3 and mSER-INC5, we developed double knockout mice for these two genes, resulting in APOBEC3/ SERINC5 Ϫ/Ϫ mice. We infected APOBEC3/SERINC5 Ϫ/Ϫ mice with F-MLV/Ampho Env and gGag mut F-MLV/Ampho Env and harvested their spleens at 10 dpi to determine viral titers. F-MLV/Ampho Env replicated at similar levels in the spleens of APOBEC3 Ϫ/Ϫ and APOBEC3/SERINC5 Ϫ/Ϫ mice (Fig. 7A). In the case of gGag mut F-MLV/Ampho Env , the APOBEC3/SERINC5 Ϫ/Ϫ mice had higher viral titers in their spleens than either the APO-BEC3 Ϫ/Ϫ mice or the SERINC5 Ϫ/Ϫ mice (Fig. 7B). Therefore, we concluded that mouse APOBEC3 and mSERINC5 act additively to block retrovirus infection in vivo.
Murine SERINC3 has no effect on MLV restriction in vivo. SERINC3 is another member of the SERINC protein family that has antiretroviral functions (3,4). To determine whether mSERINC3 could exert an antiretroviral function in vivo, we acquired mice from the Wellcome Trust Sanger Institute Mouse Genetics Project with loxP sites flanking exon 3 of SERINC3. We crossed these mice with CMV-Cre-expressing mice (Cre Deleter; Taconic) to develop complete "Ϫ/Ϫ" mice for mSERINC3 (Fig. 8A), as deletion of exon 3 results in a truncated mSERINC3 RNA transcript with a frameshift mutation that leads to a premature stop codon. Because there is no antibody to detect endogenous mSERINC3 protein levels, we probed for mSERINC3 mRNA levels by RT-qPCR and found that the SERINC3 Ϫ/Ϫ mice had significantly lower levels of mSER-INC3 transcripts than the BL/6 mice (Fig. 8B). The SERINC3 Ϫ/Ϫ mice had no apparent physical or behavioral deficits and had leukocyte populations similar to those seen with the BL/6 mice (Fig. 8C). Subsequently, we infected SERINC3 Ϫ/Ϫ mice with either gGag mut F-MLV/Ampho Env or F-MLV/Ampho Env . At 10 dpi, the two viruses replicated the same in the spleens of either BL/6 or mSERINC3 Ϫ/Ϫ mice ( Fig. 8D and E). Thus, mSERINC3, unlike human SERINC3, has no effect on MLV infection in vivo.
MLV sensitivity to mSERINC5 is independent of the route of virus entry into target cells. Previous studies on the envelope of ecotropic MLVs such as Moloney MLV Newborn mice were infected with F-MLV/Ampho Env (A) and gGag mut F-MLV/Ampho Env (B). Virus titers in the spleens were measured 10 days postinfection. Each point represents the titer obtained from an individual mouse, and the mean for each group is shown by a horizontal line. Mice were derived from 2 to 6 litters each; the BL/6 and SERINC5 mice represent littermates. Numbers of mice used in each group are indicated on the x axis. All results are presented as means Ϯ SD. Virus titers for BL/6 and SERINC5 Ϫ/Ϫ mice shown in panels A and B for F-MLV/Ampho Env and gGag mut F-MLV/Ampho Env are duplicated from Fig. 6A and B, respectively. Statistical significance was determined by Mann-Whitney (two-tailed) test. *, P Ͻ 0.5; ***, P ϭ 0.0001; ****, P Ͻ 0.0001; ns, not significant. S5, SERINC5; BL/6, C57BL/6; A3, APOBEC3.

® (MoMLV) and F-MLV showed that virus internalization varies with the host cell (43-45).
Our studies showed that infection of NIH 3T3 and Mus dunni cells, both murine fibroblasts, was affected by the presence of ammonium chloride, a pH inhibitor (Fig. 9A), which, in agreement with previous reports, suggested that virus fusion takes place in a low-pH environment after receptor-mediated endocytosis (44)(45)(46). On the other hand, in accord with previous reports (44)(45)(46)(47)(48)(49), ecotropic MLV infection is pH independent in XC cells, a rat sarcoma cell line, as we did not observe any effect on virus infection when ammonium chloride was added to the cells (Fig. 9A). In the case of amphotropic MLV, fusion happens on the cell surface and occurs at extracellular neutral pH (45,48). However, another study found that entry of amphotropic MLV into NIH 3T3 cells was pH dependent (47). Similarly to that report, we found that treatment of NIH 3T3 cells and Mus dunni with ammonium chloride blocked both ecotropic and amphotropic MLV infection (Fig. 9A and B). To rule out whether the route of entry of ecotropic MLVs is responsible for the resistance of F-MLV to mSERINC5 restriction, we infected XC cells (where entry occurs by surface fusion and is pH independent) with equal amounts of F-MLV WT, gGag mut F-MLV, gGag mut F-MLV/Ampho Env , or F-MLV/ Ampho Env generated in the presence of either mSERINC5 or empty vector. XC cells were Total RNA was isolated from C57BL/6 (n ϭ 9) and SERINC3 Ϫ/Ϫ (n ϭ 9) mice, and SERINC3 levels were measured by RT-qPCR and normalized to GAPDH. SERINC3 transcript fold expression was determined relative to that in C57BL/6 mice. (C) PBMCs from 3 C57BL6 mice and 4 SERINC3 Ϫ/Ϫ mice were stained with anti-CD45R/B220 (B cells), anti-CD3 (T cells), anti-F4/80 (Macrophages), and anti-CD11c (Dendritic cells) and subjected to FACS analysis. Values are presented as means Ϯ SD. There were no significant differences in the percentages of T cells, B cells, macrophages, and dendritic cells among the C57BL/6 and SERINC3 Ϫ/Ϫ mice. (D and E) C57BL/6 and SERINC3 Ϫ/Ϫ newborn mice were infected with F-MLV/Ampho Env (D) and gGag mut F-MLV/Ampho Env (E). Virus titers in spleens were measured 10 days postinfection. Each point represents the titer obtained from an individual mouse, and the mean for each group is shown by a horizontal line. Mice were derived from 2 to 4 litters each. Numbers of mice used in each group are indicated on the x axis. All results are presented as means Ϯ SD. Statistical significance was determined by unpaired t test (B and C) and by Mann-Whitney (two-tailed) test (D and E). ****, P Ͻ 0.0001; ns, not significant. S3, SERINC3; BL/6, C57BL/6; ICs, infectious centers.

SERINC5 Antiretroviral Role in Mice
® infected with equal amounts of virus, DNA was isolated 5 h postinfection (hpi), and RT-qPCR was performed to measure MLV DNA levels. As expected, F-MLV WT and F-MLV/Ampho Env had similar MLV DNA levels in the infected XC cells independently of the presence of mSERINC5 ( Fig. 9C and D). In the case of gGag mut F-MLV, virus DNA levels were similar in XC cells regardless of the presence of mSERINC5 in the virions (Fig. 9C), similarly to what we observed when we infected Mus dunni cells (Fig. 3E). However, when we infected XC cells with gGag mut F-MLV/Ampho Env , we observed that viral DNA levels were reduced when the virus was produced in the presence of mSERINC5 (Fig. 9E). These results are similar to what we had observed before when we used Mus dunni cells (Fig. 5F). Thus, the findings described above suggest that, as previously hypothesized, the resistance of ecotropic MLV to mSERINC5 is independent of the route of virus entry.

DISCUSSION
In vitro studies have shown that SERINC5 can block retrovirus infection (3)(4)(5)25). However, many retroviruses have developed mechanisms that can counteract SERINC5mediated restriction. In the case of MLV, glyco-Gag inhibits SERINC5 restriction by removing SERINC5 from the plasma membrane and thus blocking its incorporation into the budding virions (25,31). While much work has been done to understand the in vitro antiretroviral functions of SERINC proteins (3-5, 25, 43), nothing is known about the antiretroviral function of this family of proteins in vivo. For many antiretroviral proteins, e.g., SAMHD1, their antiviral effect in vitro is not always recapitulated in vivo (27). Many times, in vitro experiments are done using cell types that are not naturally infected by retroviruses or SERINC proteins are expressed at levels significantly higher than physiological levels. Finally, natural MLV infection in mice involves a multitude of different cell types, something that cannot be easily examined in vitro. In order to examine whether SERINC3 and SERINC5 have antiretroviral properties in vivo, we generated SERINC5 Ϫ/Ϫ mice and confirmed the disruption of the gene by RT-qPCR, as there is no antibody available to detect endogenous mSERINC5 protein levels. SERINC5 Ϫ/Ϫ mice had no apparent defects and thus represent a useful model to examine the in vivo effect of mSERINC5 on retrovirus infection. Using wild-type and glyco-Gag mutant F-MLV, we saw no differences in viral titers in the spleens of either the SERINC5 Ϫ/Ϫ or the BL/6 infected mice. On the other hand, we observed glyco-Gagdependent exclusion of mSERINC3 and mSERINC5 from the budding virions. We speculated that the lack of any obvious antiviral effect may have resulted from the fact that the viruses used in these experiments had the ecotropic envelope, which makes them resistant to SERINC5 restriction (5,43). Previous reports from studies using pseudoviruses showed that the presence of amphotropic or xenotropic MLV envelope on the surface of virions rendered them susceptible to human SERINC5 restriction in vitro (5,33). To examine the effect of the envelope on mSERINC5 restriction in fully infectious virus clones, we generated chimeric F-MLV constructs in which the ecotropic envelope was replaced with that of amphotropic MLV. Using these chimeric viruses, we found that the glyco-Gag mutant MLV with an amphotropic envelope replicated at higher levels in the spleens of SERINC5 Ϫ/Ϫ mice than in those of BL/6 mice, demonstrating the critical role of the viral envelope on mSERINC5 restriction in vivo. This report provides evidence that mSERINC5 is an antiretroviral factor in vivo and that the retroviral envelope is important in determining in vivo sensitivity to mSERINC5 restriction. Of note is the striking difference in the glyco-Gag requirement for virus infectivity between viral particles with the amphotropic envelope and those with ecotropic envelope. While the former required glyco-Gag for infectivity both in vivo and in vitro, the latter were unaffected in vitro whereas there was only about a 10-fold drop in infectivity in vivo when glyco-Gag was mutated. It is not clear why the presence of different envelopes modulates the role of glyco-Gag in virus infectivity. The relationship between envelope and glyco-Gag needs to be elucidated. To circumvent the restriction represented by the absence of an antibody that detects endogenous mSERINC5, we compared the levels of infectivity between viral particles with the amphotropic envelope and mutated glyco-Gag isolated from the spleens of either SERINC5 Ϫ/Ϫ or BL/6 infected mice. We found that virions isolated from the SERINC5 Ϫ/Ϫ mice were more infectious than those derived from BL/6 mice, which suggests that mSERINC5 is incorporated into virions in vivo and blocks virus infectivity. Furthermore, it has been hypothesized in the field that the route of entry of a virus could affect SERINC5 restriction. Our data suggest that the route of entry of a virus does not affect susceptibility to SERINC5 restriction, as ecotropic MLV was resistant to SERINC5 restriction both in Mus dunni cells, where the virus enters in a pH-dependent manner, and in XC cells, where entry occurs independently of pH. Thus, another factor and not the route of entry of a virus affects the susceptibility of the retrovirus envelope to SERINC5.
Similarly to previously reported characteristics of mouse APOBEC3 (24,38,41,50), our findings show that mSERINC5 potently restricts MLV infection in vivo. SERINC5 and mouse APOBEC3 restrict retrovirus infection at different steps of the retroviral life cycle, SERINC5 during virus-cell fusion (7) and APOBEC3 during reverse transcription (42,50). Using SERINC5/APOBEC3 double knockout mice, we showed that they indeed act additively to block retrovirus infection. To our knowledge, this is the first time that two different antiretroviral factors have been shown to act in unison in vivo. However, how these two factors come together to additively restrict MLV infection in vivo needs further investigation.
We also examined the role of mSERINC3 in retrovirus infectivity in vivo and in vitro. Human SERINC3 has been shown to be antiretroviral, albeit it is not as potent as SERINC5 (3,4), while there are differing opinions on the antiretroviral role of mSERINC3 (25,51). Our in vitro and in vivo findings showed that mSERINC3 has no antiretroviral functions against MLV, although it is targeted for internalization by glyco-Gag in vitro. It is possible that the antiretroviral function of the human homologue of SERINC3 was acquired at a later point in the evolutionary scale, possibly due to selective pressure from a retrovirus that was absent in the mouse evolutionary history, or it is possible that the murine SERINC3 has viral targets different from those associated with the human homologue.

MATERIALS AND METHODS
Ethics statement. All mice were housed and bred according to the policies of the Institutional Animal Care and Use committee of the University at Buffalo. The experiments performed with mice in this study were approved by this committee (MIC 46038Y).
Generation of SERINC5 ؊/؊ mice. To generate the SERINC5 Ϫ/Ϫ mice, exon1 and exon 2 were targeted by 2 guide RNAs (gRNAs) (gRNA1 [5=-ATGTCTGCCCGGTGCTGTGCTGG] and gRNA2 [5=-GCCCTA AGTTCCGACAGTCTCGG]) by the use of IDT Alt-R crRNA/trRNA CRISPR/Cas9 technology. The gRNAs and recombinant Cas9 (RNPs) were electroporated into zygotes of C57BL/6N mice (Charles River Laboratories). DNA from the mice generated from the electroporations was sent to the Genome Engineering and iPSc Center at Washington University in St. Louis, where it were screened via targeted next-generation sequencing (NGS) for the presence of deletion or indels at the target site. One founder line was discovered to have a sizable 49-kb deletion that included part of exon 1 and exon 2 along with the intron between these two exons. As such, for the purpose of genotyping we used two sets of primers, SEREx1F and SEREx2R/SEREx1F and SERInt2R (see Table S1 in the supplemental material); the first set of primers amplifies the deleted allele (as the two primers are 49 kb apart in the BL/6 mice), and the second set of primers amplifies a region of about 250 bp.
Generation of SERINC3 ؊/؊ mice. SERINC3 Ϫ/Ϫ mice were derived from in vitro fertilization of C57BL/6N females with sperm from a single heterozygous male from the mutant mouse line C57BL/ 6NTac-SERINC3 tm1a(KOMP)Wtsi /WtsiFlmg constructed by the Wellcome Trust Sanger Institute Mouse Genetics Project (Sanger MGP). Cre-Deleter mice [C57BL/6NTac-Gt(ROSA)26Sor tm16(cre)Arte ], where the cre gene is expressed under the control of the Gt(ROSA)26Sor gene, were purchased from Taconic Biosciences and used to develop SERINC3 Ϫ/Ϫ mice. Deletion by cre recombinase in SERINC3 tm1a(KOMP)Wtsi / WtsiFlmg results in the removal of exon 3, leading to a frameshift mutation and a premature stop codon. The following primers were used for genotyping: Serinc3_35556_F and Serinc3_35556_R/SERINC3Ex3F and SERINC3Ex3R for detection of WT alleles and Serinc3_35556_F and CAS_R1_Term/DelF and DelR for detection of deleted alleles. The primers used for detection of cre were CreF and CreR (see Table S1).
Infectivity measurement of in vivo-derived virus. Two-day-old SERINC5 Ϫ/Ϫ and BL/6 mice were infected by i.p. injections with 4 ϫ 10 5 ICs of virus, and spleens were harvested 16 days postinfection. Spleen extracts were prepared as previously described (59) with some modifications. Briefly, spleens were mashed through a cell strainer and splenocytes were collected in a mixture of 5 ml (for F-MLV/Ampho Env virus) or 2 ml (for gGag mut F-MLV/Ampho Env virus) of RPMI media with 10% FBS and 100 mg/ml P/S. Cells were pelleted at 1,200 rpm for 10 min at 4°C. The supernatant was removed and centrifuged at 3,000 rpm for 10 min at 4°C and then filtered through a 0.45-m-pore-size syringe filter. RNA was isolated from 100 l of the supernatant using an RNeasy kit (Qiagen) per the manufacturer's recommendation. cDNA synthesis and RT-qPCR were performed as described above (see "SERINC3/5 transcription levels"). Normalized spleen extracts (based on RNA copies) were used to infect 0.3 ϫ 10 5 Mus dunni cells seeded in a 24-well plate. Cells were harvested 19 hpi, and the amount of viral DNA in the infected cells was determined by RT-qPCR, as described above (see "Target cell assays").
Ammonium chloride treatment. One day prior to infection, NIH 3T3, Mus dunni, or XC cells were seeded in a six-well plate (0.8 ϫ 10 5 cells/well). Cells were pretreated with 30 mM ammonium chloride (NH 4 Cl) for 30 min before infection was performed with F-MLV WT or F-MLV/Ampho Env viruses (MOI 0.5) (45). The same concentration of NH 4 Cl was maintained during the infection period and an additional 2 h, after which cells were maintained in fresh media without NH 4 Cl. Cells were harvested 19 hpi, and the amount of viral DNA in the infected cells was determined by RT-qPCR, as described above (see "Target cell assays").
Statistical analysis. Statistical analyses were performed using GraphPad Prism software version 8.2. The statistical tests used to determine significance are described in the figure legends. A difference was considered to be significant for P values of Ͻ0.05.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
We thank Susan Ross, Leonard "Pug" Evans, Heinrich Gottlinger, and Hans Acha-Orbea for kindly providing us with reagents that were used in this study. We also thank Ray Kelleher for help with the primary cell sorting and the staff members of the Gene Targeting and Transgenic Shared Resource at the Roswell Park Comprehensive Cancer Center for assisting in the derivation of the SERINC5 Ϫ/Ϫ and SERINC3 Ϫ/Ϫ mice. We thank the staff members of the Welcome Trust Sanger Institute Mouse Genetics Project (Sanger MGP) and its funders for providing the mutant mouse line [C57BL/6NTac-SERINC3 tm1a(KOMP)Wtsi /WtsiFlmg] and INFRAFRONTIER/EMMA (www.infrafrontier.eu). Finally, we thank Sabine Baxter at the Cell Center Services of the University of Pennsyl-vania for assisting us with our antibody purification needs. Sanger MGP funding information may be found at www.sanger.ac.uk/mouseportal and associated primary phenotypic information at www.mousephenotype.org.
The work was supported by National Institutes of Health grant R21-AI144147-01A1 and a Mathilde Krim Fellowship in Basic Biomedical Research Phase II, American Foundation for AIDS Research 109741-63-RKHF.