Characterization of Endogenous SERINC5 Protein as Anti-HIV-1 Factor

SERINC5 is the long-searched-for antiviral factor that is counteracted by the HIV-1 accessory gene product Nef. Here, we engineered, via CRISPR/Cas9 technology, T-cell lines that express endogenous SERINC5 alleles tagged with a knocked-in HA epitope. This genetic modification enabled us to study basic properties of endogenous SERINC5 and to verify proposed mechanisms of HIV-1 Nef-mediated counteraction of SERINC5. Using this unique resource, we identified the susceptibility of endogenous SERINC5 protein to posttranslational modulation by type I IFNs and suggest uncoupling of Nef-mediated functional antagonism from SERINC5 exclusion from virions.

Expression of endogenous SERINC5 mRNA and SERINC5 protein. HA epitope tagging of endogenous SERINC5 genes did not modulate SERINC5 mRNA expression levels as assessed by quantitative real-time PCR (Q-RT-PCR) analysis using a primerprobe assay that amplifies a region common to all three known isoforms (19) (Fig. 2A). Furthermore, Northern blot experiments using a SERINC5-specific probe revealed the expression of isoform 1 in parental Jurkat T cells and in SERINC5(iHA/iHA) clones (Fig.  2B), suggesting that HA-encoding SERINC5 mRNAs equal SERINC5 mRNAs of parental Jurkat T cells in terms of abundance and splicing behavior. HA sequence knock-in enabled detection of endogenous SERINC5 protein by anti-HA immunoblotting of cell lysates (Fig. 2C). Parental Jurkat T cells were negative for HA, and lysates from SERINC5(iHA/iHA) and SERINC5(iHA/KO) clones displayed migration patterns that were identical to those of lysates obtained from Jurkat T cells and HEK293T cells transduced with SERINC5(iHA). SERINC5(iHA) presented as two protein species of 51 and 35 kDa, respectively. The 51-kDa, but not the 35-kDa, species was sensitive to peptide-Nglycosidase (PNGase) digestion (Fig. 2D), suggesting that it corresponds to the reported molecular weight of fully N-glycosylated protein (16) and that the 35-kDa species represents a nonglycosylated albeit still antiviral SERINC5 species (16). Endogenous SERINC5(iHA) was detectable at the surface of intact cells by flow cytometric analysis of anti-HA immunostaining ( Fig. 2E and F), with a trend toward higher surface levels in SERINC5(iHA/iHA) clones than in most SERINC5(iHA/KO) clones. Immunofluorescence microscopy demonstrated presence of endogenous SERINC5(iHA) at the plasma membrane and colocalization with cholera toxin-stainable ganglioside GM1, suggesting an association with lipid rafts (Fig. 2G). Colocalization with the early endosomal marker EEA1 was only rarely observed (Fig. 2H).
Type I interferon modulates cell surface expression of endogenous SERINC5 protein in the absence of modulation of mRNA and protein quantities. In line with previously reported work (3,4), treatment of SERINC5(iHA/iHA) clones with individual interferon alpha (IFN-␣) subtypes failed to consistently upregulate SERINC5 mRNA expression, while IFIT1 mRNA expression was induced by 40-to 44,000-fold by all active IFN-␣ subtypes, confirming the effectiveness of the IFN-␣ treatment (Fig. 3A). Of note, individual IFN-␣ subtypes have been reported to substantially vary in their antiviral potencies (20), providing a rationale for testing them individually and side by side. IFN-␣1, which has a comparably low affinity for IFN-alpha/beta receptor (IFNAR) (21), was largely inactive under these experimental conditions. Additionally, no consistent changes in total levels of cell-associated SERINC5(iHA) protein were detected upon IFN-␣ treatment, arguing against a potential impact of IFN-␣ on SERINC5 protein synthesis and overall stability. ISG15 protein levels were upregulated 4-to 23-fold by individual active IFN-␣ subtypes (Fig. 3B). Interestingly, amounts of SERINC5 at the surface of SERINC5(iHA/iHA) T cells increased up to 2-fold upon IFN-␣ treatment ( Fig. 3C and D). IFN-␣-induced enhancement of SERINC5(iHA) surface levels was entirely abrogated in the context of cotreatment with the Jak/STAT inhibitor ruxolitinib. As a reference, within identical samples, the induction of intracellular MXA/B expression, which is strictly IFN dependent (22), was up to 3.6-fold (Fig. 3C). Immunofluorescence microscopy of SERINC5(iHA) revealed HA positivity at the surface of both mock-treated and IFN-␣-treated cells (Fig. 3E). Together, these data suggest that the subcellular localization of SERINC5 is regulated by type I IFNs in the absence of modulation of mRNA and protein quantities.
HIV-1 Nef-mediated enhancement of particle infectivity may occur in the absence of exclusion of endogenous SERINC5 from virions. Key concepts of SERINC5's antiviral mode of action and of its counteraction by HIV-1 Nef were established mainly in heterologous expression systems. Here, we aimed at testing these working models in the context of endogenously expressed SERINC5    4A). Infected parental cells and the SERINC5(iHA/iHA) and SERINC5 (iHA/KO) clones produced wild-type HIV-1 particles of similar infectivities and shared the requirement of proviral Nef expression for the generation of fully infectious particles (Fig. 4B). As expected, CRISPR/Cas9mediated knockout of both SERINC5 alleles fully rescued the ability of Jurkat T cells to generate infectious HIV-1 Δnef virions (Fig. 4B). Interestingly, the reduction of HIV-1 Env-dependent, T-20-sensitive fusogenicity (Fig. 4C, left) that accompanied the SERINC5-mediated reduction of particle infectivity (Fig. 4C, right) was relatively modest, if even detectable, indicating that additional, postfusion restrictions might be exerted by SERINC5. In addition, the fusion capabilities of HIV-1 wild-type and HIV-1 Δnef virions were very similar, suggesting that functional counteraction of SERINC5 by Nef does not occur at the level of virus-cell fusion. Immunoblotting of sucrose cushion-purified virus followed by quantitative infrared-based imaging revealed an association of the 35-kDa, but not of the 51-kDa, protein species of endogenous SERINC5(iHA) protein in HIV-1 Δnef particles derived from all Jurkat HA knock-in clones (Fig. 4D). In contrast, virions derived from Jurkat T cells and HEK293T cells that had been retrovirally transduced with SERINC5(iHA) presented exclusively and predominantly the 51-kDa species, respectively ( Fig. 4D), in accordance with previously reported results (4,16). Surprisingly, proviral Nef expression appeared to not reduce SERINC5 virion incorporation in all HA knock-in clones ( Fig. 4D and E). Virion-associated levels of SERINC5(iHA) were significantly reduced by 50 to 60% in wild-type HIV-1 particles (compared to HIV-1 Δnef particles) derived from the SERINC5(iHA/iHA) clones P1E8 and P2G3, the SERINC5(iHA/KO) clones P1A2 and P1A8, and Jurkat T cells transduced with SERINC5(iHA). In contrast, levels of SERINC5(iHA) were unaffected or even enhanced by Nef in virions derived from the SERINC5(iHA/iHA) clone P2B1 and the SERINC5(iHA/KO) clones P1B6, P1B12, P1C1, and P3A1 ( Fig. 4D and E). The fact that proviral HIV-1 Nef neither modulated SERINC5(iHA) incorporation into virions generated by SERINC5(iHA)-transduced HEK293T cells nor counteracted the SERINC5-imposed restriction at the infectivity level (WT average, 100% [standard error of the mean {SEM}, 26.8%]; Δnef average, 95.6% [SEM, 20.5%] [not significant]) is probably due to the saturating expression levels of SERINC5. Importantly, supernatants from uninfected or pVSV-G-transfected SERINC5(iHA/iHA) clones did not display HA positivity (Fig. 4D), arguing against an accumulation of SERINC5(iHA) in extracellular vesicles that may have cosedimented in our virus particle preparations (23). These results suggest that endogenous SERINC5 protein, at least the lowmolecular-weight form, which has been suggested to exert antiviral activity (16), is incorporated into HIV-1 Δnef particles at detectable levels. However, Nef expression in virus-producing cells may not necessarily decrease endogenous SERINC5 protein association into virions despite exerting a strong antagonistic activity at the functional level. HIV-1 Nef modulates subcellular localization and trafficking of endogenous SERINC5 protein. We next dissected the impact of HIV-1 Nef on endogenous SERINC5 protein surface localization in virus-producing cells. Flow cytometric analysis of infected SERINC5(iHA/iHA) and SERINC5(iHA/KO) clones demonstrated the susceptibility of the endogenous protein to downregulation by HIV-1 Nef (Fig. 5A and B). The magnitude of reduction from the plasma membrane ranged between 25 and 47% in individual clones  (Fig. 5B). Furthermore, HIV-1 Nef increased the rate of internalization of endogenous surface SERINC5 in a kinetic endocytosis assay in all SERINC5(iHA/iHA) clones (Fig. 5C).
Infection of cells at a high multiplicity of infection (MOI), however, failed to result in detectable quantities in both SERINC5(iHA) protein species (Fig. 5D and E), arguing against Nef-induced degradation of the antiviral factor. These results establish that proviral HIV-1 Nef modulates the cell surface expression and the rate of internalization of endogenous SERINC5 protein in the absence of an alteration of steady-state protein levels.

DISCUSSION
Functional studies in genetically modified T cells have shed light on the importance of SERINC5 in HIV-1 restriction and the ability of Nef to antagonize this antiviral factor (3,4). However, due to the current unavailability of an anti-SERINC5 antibody of sufficient sensitivity and specificity, we lack a good understanding of endogenous SERINC5 protein expression and subcellular localization. Concepts of its antiviral mode of action and of its counteraction by viral antagonists were derived, to a large extent, from experiments in which SERINC5 was ectopically expressed. However, knowledge gained through the heterologous expression of cellular factors does not necessarily reflect key aspects of the endogenous protein. This can be attributed to the absence of a natural gene expression context and/or to nonphysiological expression levels. Previous studies reported aberrant effects resulting from SERINC5 overexpression and stressed the importance of expressing SERINC5 at low levels using plasmids from which SERINC5 expression is driven by a low-activity promoter (4) in order to attain physiological expression levels.
Our work establishes predominant cell surface expression of endogenous SERINC5 with a high degree of association with lipid rafts. Steady-state intracellular SERINC5 expression was detectable but did not display consistent colocalization with the early endosomal marker EEA1, suggesting that the natural recycling pathway of SERINC5 does not involve early endosomes. Functional inactivation of one SERINC5 allele, resulting in SERINC5 expression exclusively from the second, remaining allele, did not significantly alter SERINC5 expression levels and, thus, did not modulate the antiviral capacity of Jurkat T cells, suggesting that the loss of one functional allele is compensated for by the other, functional, allele. Interestingly, type I IFN increased the abundance of surface SERINC5 in an entirely Jak/STAT inhibitor-sensitive manner without augmenting mRNA and whole-cell-associated protein quantities. This suggests that type I IFN treatment induces a relocalization of intracellular SERINC5 to the plasma membrane and/or stabilizes cell surface SERINC5 by impairing or retarding its endocytosis or recycling. Although the magnitude of enhancement of surface SERINC5 levels by type I IFN was lower than that of the induction of MXA/B protein expression, it may suffice to limit the efficiency of Nef-mediated antagonism. However, due to the multitude of IFN-stimulated antiviral genes that are induced by IFN treatment and that decrease the efficiency of several steps of the HIV-1 replication cycle, including tetherinmediated retention of virus particles (24,25) and 90K-mediated reduction of particle infectivity (26), experiments aiming at testing of this hypothesis have been inconclusive    so far. To our knowledge, this is the first example of an antiviral factor whose subcellular localization is modulated at the posttranslational level by type I IFNs, and future studies deciphering the IFN-induced interactome of endogenous SERINC5 are required to delineate the underlying mechanism. Previous work suggested that SERINC5's antiviral activity is largely driven by its association with virions, thereby modifying them in a manner that results in inefficient Env-mediated membrane fusion with target cells. However, the SERINC5-imposed reduction of Env-mediated fusion of BlaM-Vpr-containing viruses to target cells was relatively low (up to 3-fold) compared to the reduction of HIV-1 infectivity (up to 50-fold in the parallel infectivity assay). This observation has already been made in heterologous assays in which SERINC5-encoding plasmids were expressed at doses anticipated to approach physiological expression levels (3,4) and is consistent with the possibility of additional, postfusion restrictions exerted by SERINC5.
Exclusion of SERINC5 from virions is supposed to be a consequence of its Nefmediated downregulation from the cell surface (3,4) and potentially its degradation in intracellular compartments (17). Indeed, endogenous SERINC3 has been shown to be excluded from virion incorporation in a Nef-dependent manner in a mass spectrometrybased approach (3). Here, we demonstrate that SERINC5 is susceptible to Nef-mediated downregulation from the cell surface and internalization in all tested clones. Nefmediated reduction of surface levels of endogenous SERINC5 in infected T cells was statistically significant but Ͻ2-fold, which again appears to be relatively mild when set in relation to the much more pronounced, up to Ͼ250-fold enhancement of infectivity of released viruses. While targeting to lysosomal degradation by Nef has been suggested by others using heterologous expression systems (17), steady-state levels of endogenous SERINC5 remained unaltered in infected cells, irrespective of the Nef expression status. Furthermore, we demonstrate HA positivity associated with HIV-1 Δnef virions, suggesting viral incorporation of SERINC5, as expected. HA positivity in virions was associated with a band of 35 kDa, and we failed to detect the highmolecular-weight SERINC5 species in virus preparations. In contrast, the exclusive and predominant incorporation of the 51-kDa species in virions generated by Jurkat T cells and HEK293T cells expressing transduced SERINC5(iHA) demonstrates that our assay displays sufficient sensitivity to detect this SERINC5 species in general and reproduces previously reported findings (4,16). Along this line, SERINC5 clearly presented as two distinct species in cell lysates in our assay. Therefore, it is conceivable that the virus-associated HA signal derived from Jurkat SERINC5(iHA knock-in) clones represents impartially glycosylated SERINC5 species (16). In the context of heterologous expression, it has been suggested that a glycosylated species of SERINC5 with a molecular weight of 55 kDa is specifically incorporated into virions, whereas a low-molecularweight form of SERINC5 of Յ40 kDa predominates in cell lysates and corresponds to a nonglycosylated protein (16). The same study showed that N-glycosylation of SERINC5 at residue N294 stabilized its steady-state levels and prevented otherwise rapid targeting for proteasomal degradation. Importantly, the nonglycosylated, low-molecularweight form of SERINC5 maintained its antiviral capacity and its susceptibility to Nef-mediated counteraction. Whether SERINC5, when expressed from its endogenous promoter and under physiological conditions, is susceptible to N-linked glycosylation with efficiencies and kinetics similar to those of heterologously expressed SERINC5 remains an interesting question whose answer might help to reconcile the apparently divergent results obtained in heterologous and endogenous expression systems in the future.
While proviral Nef expression was uniformly required for particle infectivity rescue in all SERINC5-expressing clones, it reduced virus-associated HA positivity in particles in only four out of nine clones. Clone-specific properties at the genetic or transcriptional level that specifically modulate the susceptibility of SERINC5 to Nef-mediated exclusion from virions may cause this heterogeneity. Indeed, transcriptome sequencing (RNAseq) analysis revealed a set of genes that are differentially expressed in the two phenotypic groups of clones. Regardless of the underlying reason, the apparent dispensability of Nef-mediated exclusion of endogenous SERINC5 from virions for infectivity rescue provides evidence for the presence of additional counteractive mechanisms of Nef, directed against virus-associated pools of SERINC5, as has been postulated by another group (10).
Together, data from this study establish CRISPR/Cas9-assisted epitope tagging of endogenous alleles of SERINC5 as a useful technology that enabled us to investigate key aspects of SERINC5 antiviral restriction and HIV-1 Nef-mediated antagonism. A similar approach in other cell types in the future might reveal the extent to which our findings can be extrapolated in more physiologically relevant, primary T cells and macrophages. Future studies using this resource may help to advance our understanding of both SERINC5 restriction and viral counteraction and its physiological function.

MATERIALS AND METHODS
Cell lines. HEK293T cells and Jurkat T cells were purchased from the ATCC and cultured as recommended. TZM-bl cells were obtained from the NIH AIDS Reagent Program. For the generation of SERINC5(KO/KO) cell lines, Jurkat T cells were electroporated with 100 g/ml of a plasmid expressing Cas9-2A-enhanced GFP (EGFP) and U6-driven chimeric guide RNA (SERINC5 target sequence GCTGAGGGAC TGCCGAATCC[TGG]) using the Neon transfection system (Thermo Fisher Scientific, Darmstadt, Germany) at 1,500 V for 30 ms, with 1 pulse. EGFP-positive cells were sorted on a BD FACSAria cell sorter and clonally expanded. Individual clones were genotyped by Sanger sequencing (SeqLab, Göttingen, Germany) of the PCR-amplified genomic locus (forward primer TGCTGTGTTGACCAGGCTAA and reverse primer GGCATT GGATCCTGGAAAGC). Individual alleles were deduced using the Poly Peak Parser tool (27) and, where applicable, allocated based on peak strength. SERINC5(iHA/iHA) and SERINC5(iHA/KO) Jurkat clones were generated by coelectroporation of 1.8 M Cas9 RNPs (Alt-R Cas9 protein, trans-activating crRNA (tracrRNA), and CRISPR-RNA (crRNA) [target sequence, TTCAAGTTCTAGATGAACAT{GGG}]; IDT, Leuven, Belgium) and 5 M single-stranded DNA (ssDNA) repair oligonucleotide (ACTTTGTTTTTTCTTTTCAAGTTC TAGATGAATACCCATACGATGTTCCAGATTACGCTCATGGGAAAAATGTTACAATCTGTGTGCCTG [the HA tag is underlined]) using the Neon transfection system (1,500 V for 10 ms, with 3 pulses). Single-cell clones were screened by PCR (for PCR 1, for the presence of the HA tag, HA forward primer TACCCATACGAT GTTCCAGATTA and HA reverse primer AGTTCACGCTCTTCGCCTTT; for PCR 2, for insert size, forward primer CTTCTGTGCGTTACAACTGGCC and reverse primer TAGTCACCAAGTTTTCATCTCTGTACAGG), followed by Tris-borate-EDTA (TBE)-PAGE (7.5%), and genotyped by Sanger sequencing of the PCRamplified genomic locus (forward primer TGGCACTGAGCTGGAATCTG and reverse primer AGTTCACGCT CTTCGCCTTT).
Quantitative RT-PCR. Total RNA extraction from cells and DNase treatment were performed with a Maxwell LEV simplyRNA purification kit (Promega), followed by cDNA synthesis (NEB, Invitrogen). Quantification of relative SERINC5 and IFIT1 mRNA levels was performed with the 7500 Fast real-time PCR system (Applied Biosystems) using TaqMan PCR technology with premade primer-probe kits (Applied Biosystems). Relative mRNA levels were determined using the ΔΔC T method, with human RNASEP mRNA (Applied Biosystems) as an internal reference. Each sample was analyzed in triplicates. Data analysis was performed using Applied Biosystems 7500 Fast system software.
Northern blotting. RNA extraction, gel electrophoresis, blotting, and detection with a radiolabeled probe were performed as described previously (29) but with the following adjustments. Ten micrograms of total RNA for each cell line was loaded onto the gel. The probes were labeled with the DecaLabel DNA labeling kit (Thermo Fisher Scientific), and the membrane was exposed to Amersham Hyperfilm MP (GE Healthcare) for 8 days (in the case of the SERINC5-specific probe) or for 8 h (glyceraldehyde-3-phosphate dehydrogenase [GAPDH]-specific probe) at Ϫ80°C. The SERINC5-specific probe of 870 bp was prepared by digestion of a plasmid bearing the SERINC5 cDNA with XbaI and NotI. For the loading control, a GAPDH plasmid (a gift from K. Habers, Heinrich-Pette-Institut, Hamburg, Germany) was digested by EcoRI, generating a 1.3-kbp GAPDH-specific probe.
CRISPR/Cas9-Assisted Analysis of SERINC5 Journal of Virology Data presentation and statistical analysis. If not otherwise stated, bars and symbols show the arithmetic means from the indicated number of repetitions. Error bars indicate standard deviations (SD) from one representative experiment out of at least three or SEM from the indicated number of individual experiments. Significance values were calculated using the 2-tailed Student t test and are indicated in the figures (*, P Ͻ 0.05; **, P Ͻ 0.01; n.s., not significant).