Nef Homodimers Downregulate SERINC5 by AP-2-Mediated Endocytosis to Promote HIV-1 Infectivity

SERINC5 is a multi-pass intrinsic membrane protein that suppresses HIV-1 infectivity when incorporated into budding virions. The HIV-1 Nef viru-lence factor prevents viral incorporation of SERINC5 by triggering its downregulation from the producer cell membrane through an AP-2-de-pendent endolysosomal pathway. However, the mechanistic basis for SERINC5 downregulation by Nef remains elusive. Here we demonstrate that Nef homodimers are important for SERINC5 downregulation, trafficking to late endosomes, and exclusion from newly synthesized viral particles. Based on previous X-ray crystal structures, we mutated three conserved residues in the Nef dimer interface (L112, Y115, F121) and demonstrated attenuated homodimer formation in a cell-based fluorescence complementation assay. Point mutations at each po-sition reduced the infectivity of HIV-1 produced from transfected 293T cells, the Jurkat TAg T cell line, and donor mononuclear cells in a SERINC5-dependent

tion assays demonstrated that dimerization-defective Nef mutants retained interaction with both SERINC5 and AP-2. These results show that downregulation of SERINC5 and subsequent enhancement of viral infectivity requires Nef homodimers and support a mechanism by which the Nef dimer bridges SERINC5 to AP-2 for endocytosis. Pharmacological disruption of Nef homodimers may control HIV-1 infectivity and viral spread by enhancing virion incorporation of SERINC5.
The HIV-1 Nef accessory protein is expressed early in the HIV-1 life cycle where it promotes viral replication and AIDS progression (1). Sole expression of Nef within the CD4-positive cell compartment induces an AIDS-like syndrome in transgenic mice (2). Conversely, patients harboring Nef-defective strains of HIV-1 exhibit reduced viral loads and delayed progression to AIDS (3,4). Enhancement of lentiviral pathogenicity by Nef is conserved across species. Disruption of the nef open reading frame of simian immunodeficiency virus (SIV) significantly reduces viral replication and prevents the development of AIDS-like disease in SIV-infected macaques (5). These are just a few examples of the evidence supporting a central role for Nef in HIV-1 and SIV pathogenicity.
Nef is a small protein (molecular weight range of 27-34 kDa, depending on the viral subtype) that is associated with host cell membranes by vir-tue of N-terminal myristylation (6). Nef lacks inherent enzymatic activity, functioning instead through multiple protein-protein interactions that alter host cell signaling and protein trafficking networks involving as many as 70 host cell proteins (7). Nef selectively binds and activates members of the Src and Tec protein-tyrosine kinase families (8)(9)(10)(11). Disruption of Nef-dependent kinase activation through genetic or pharmacological means impairs Nef-mediated enhancement of HIV-1 infectivity and replication (10)(11)(12)(13)(14). In terms of protein trafficking, Nef drives the downregulation of cell surface receptors essential for immune recognition and viral entry, including MHC-I, CXCR4, CCR5, and CD4 (6,15,16). Receptor downregulation involves Nef interactions with endosomal trafficking proteins including the adaptor protein complexes 1 and 2 (AP-1 and AP-2) (17,18). Downregulation of CD4 requires simultaneous engagement of Nef with both the cytoplasmic tail of CD4 and a hemicomplex formed by the α and σ2 subunits of AP-2 (19). Nef•CD4•AP-2 complexes form clathrin-coated pits at the cell surface, leading to endocytosis and lysosomal degradation of internalized CD4 (20).
Nef-mediated endocytosis of the SERINC5 restriction factor, a process essential for enhancement of HIV-1 infectivity, is also mediated by the AP-2 trafficking pathway. In the absence of Nef, SERINC5 is present on the surface of HIV-1 infected cells and incorporated into the membrane of newly synthesized virions (21)(22)(23). Incorporation of SERINC5 disrupts viral fusion with host cells and delivery of the viral core through a cryptic, Env-dependent mechanism. Nef antagonizes SERINC5 in part by promoting AP-2-dependent downregulation of SERINC5 from the plasma membrane of infected cells, thereby preventing incorporation into budding virions (22). Following downregulation by Nef, internalized SERINC5 is targeted for degradation via the endolysosomal pathway (24).
While Nef uses distinct structural motifs to recognize diverse host cell partners, many Nef functions also require homodimer formation. Mutants of Nef that are defective for homodimerization are unable to induce host cell tyrosine kinase activation, even though they retain interaction with the kinase proteins (11). This observation is consistent with the formation of Nef•kinase dimer complexes necessary for kinase activation by trans-autophosphorylation. Dimerization-defective Nef mutants are also unable to downregulate CD4 from the host cell surface, and these mutations reduce HIV-1 replication efficiency in cell lines (25,26).
In this study, we investigated the role of Nef homodimerization in HIV-1 infectivity and restriction by SERINC5. Using a series of Nef mutants defective for homodimer formation in cells, we demonstrated that Nef dimers enhance infectivity of HIV-1 produced from T cell lines and donor PBMCs, as well as 293T cells co-transfected with SERINC5. These same Nef mutants were attenuated in their ability to exclude SERINC5 from newly synthesized virions. Impairment of Nef dimerization reduced downregulation of SERINC5 from the cell surface and prevented intracellular trafficking to Rab7+ late endosomes. Critically, cell-based fluorescence complementation assays revealed that these Nef mutants retained interaction with both SERINC5 and the AP-2 α and σ2 subunits, indicating that the reduced antagonism of SERINC5 was not due to loss of Nef association with either partner protein. Taken together, our data demonstrate that Nef antagonizes the SERINC5 restriction factor as a homodimer, which may serve as a bridge linking SERINC5 with AP-2 for endocytosis and degradation.

Structural basis of HIV-1 Nef homodimerization.
In this study, we explored the role of Nef homodimerization in the downregulation of SERINC5 and promotion of HIV-1 infectivity using a genetic approach involving Nef mutants defective for homodimer formation. X-ray crystallography of Nef in complex with the regulatory SH3 and SH2 domains of Src-family kinases has shown that Nef forms homodimers through a hydrophobic interface involving side chains of residues in the Nef αB helices and adjacent loops (27,28). An example is provided by the structure of HIV-1 Nef in complex with the SH3-SH2 regulatory region of the Src-family kinase, Hck (PDB 4U5W; a model of the Nef homodimer is shown in Figure 1A) (28). In this structure, three conserved residues from one Nef subunit (Leu112, Tyr115, Phe121) converge to create a 'hydrophobic cup' that clasps the side chain of Val70 from the opposing Nef monomer in a reciprocal fashion ( Figure 1B). A recent study has shown that recombinant Nef proteins in which Leu112 or Tyr115 is replaced with aspartate or Phe121 is substituted with alanine fail to form dimers in solution, HIV-1 Nef downregulates SERINC5 as a homodimer providing direct evidence that these residues are essential for dimer stabilization (11). Mutations at these positions attenuate multiple Nef functions, including host cell kinase activation, enhancement of HIV-1 replication, and receptor downregulation including CD4 (see Introduction).
While the role of Nef homodimers in SERINC5 antagonism has not been explored, the mechanistic similarity between Nef-mediated downregulation of CD4 and SERINC5 suggests a key role. Both processes involve clathrin-mediated endocytosis via the tetrameric trafficking adaptor protein AP-2, which interacts directly with Nef through a hemicomplex of its α and σ2 subunits (19,29). Alignment of the Nef 4U5W homodimer with the crystal structure of Nef bound to the AP-2 α/σ2 hemicomplex shows that the Nef homodimer interface residues do not contact AP-2 ( Figure 1C), suggesting that the dimer interface mutations described above will not affect AP-2 recruitment. These mutants were therefore chosen to explore the requirement for Nef dimers in SERINC5 regulation.
Though Nef is present in HIV-1 virions and expressed early within the viral life cycle (29), previous analyses have examined the impact of mutations on Nef self-association 40-48 hours posttransfection (25,26) which corresponds to the end of the HIV-1 life cycle (30). We therefore explored the kinetics of wild-type Nef homodimer formation and membrane association at 12-hour intervals using a cell-based bimolecular fluorescence complementation (BiFC) assay (31). In this approach, wild-type and mutant Nef coding sequences are fused to an N-terminal fragment of the YFP variant, Venus (Nef-VN), while complementary Nef constructs are fused to a C-terminal Venus fragment (Nef-VC). Co-expression of the Nef-VN and Nef-VC fusion proteins in the same cell results in protein-protein interaction between Nef molecules, leading to complementation of the Venus fluorophore and a bright fluorescent signal.
Cultures of 293T cells were transfected with Nef-VN and Nef-VC expression vectors (wildtype or the dimerization interface mutants L112A, L112D, Y115D, F121A) and fixed with paraformaldehyde 12, 24, 36, and 48 hours later. Cells were then immunostained to confirm Nef protein expression, allowing normalization of the resulting BiFC signal to Nef expression levels within each culture. Figure 2A shows representative confocal images of Nef BiFC (homodimer formation) and immunofluorescence (Nef protein expression) at the 48 hour endpoint of the experiment, while Nef BiFC to immunofluorescence intensity ratios across all four time points are plotted in Figure 2B. Little fluorescence complementation was observed with any of the Nef BiFC pairs at 12 hours, which likely reflects the low levels of Nef expression at this time point. After 24 hours, wild-type Nef self-association was readily observed and increased in a linear fashion over 48 hours. In contrast to wild-type Nef, the L112D, Y115D, and F121A mutants demonstrated statistically significant defects in Nef homodimerization across all time points ( Figure S1, supplemental material). The Y115D mutant showed the lowest ratios at each time point, with the L112D and F121A mutants exhibiting an intermediate phenotype. By contrast, the L112A mutation did not decrease the BiFC signal suggesting that substitution of leucine with alanine does not disrupt homodimer formation. Inspection of the BiFC and immunofluorescence images suggests that all four Nef mutants remained associated with the plasma membrane, although the Y115D mutant showed increased cytoplasmic localization.
Dimerization-defective Nef attenuates HIV-1 infectivity and impairs SERINC5 antagonism. To test the requirement for Nef homodimers in SERINC5 antagonism, we first validated a 293T producer cell system to assess virion incorporation of SERINC5 and impact on viral infectivity. For these studies, 293T cells were transfected with proviral plasmids encoding wild-type HIV-1 or a mutant that fails to express Nef (∆Nef) in the presence and absence of a SERINC5 expression vector. Newly synthesized virions were concentrated from the culture supernatant two days later and levels of virion-incorporated SERINC5 were analyzed via immunoblot. The infectivity of each viral supernatant was assessed in parallel using the TZM-bl reporter cell line (32). Expression of SERINC5 in 293T producer cells inhibited HIV-1 ∆Nef infectivity but had little impact on the infectivity of the wild-type virus, consistent with previous results ( Figure 3A) (21,22). Nef antagonism of SERINC5 within this system is consistent with levels of virion-incorporated SERINC5 observed by immunoblot, with wild-type virions exhibiting decreased levels of SERINC5 compared to the ∆Nef by guest on November 6, 2020 http://www.jbc.org/ Downloaded from HIV-1 Nef downregulates SERINC5 as a homodimer virions ( Figure 3B). Co-transfection of an expression plasmid encoding Nef along with SERINC5 and the ∆Nef proviral plasmid rescued infectivity ( Figure 3C) and promoted exclusion of SERINC5 from newly synthesized virions ( Figure 3D). These observations validate the use of the 293T producer cell system to explore the role of Nef homodimerization in SERINC5-mediated restriction of HIV-1 infectivity.
We next transfected 293T cells with HIV-1 proviral DNA bearing the Nef dimerization interface mutations L112A, L112D, Y115D, and F121A in the presence and absence of the SERINC5 expression vector. The infectivity of HIV-1 expressing the L112D, Y115D, and F121A Nef mutants was reduced in the presence of SERINC5 to the levels observed with the ∆Nef virus, while the L112A mutant exhibited a partial loss of function ( Figure  4A). This finding is consistent with BiFC results, showing that L112A has little impact on Nef homodimer formation in this system. Immunoblot analysis confirmed that each of the Nef mutants was expressed in transfected 293T cells ( Figure 4B).
In a complementary set of experiments, we explored the role of Nef dimerization in SERINC5 antagonism using the Jurkat JTAg producer cell system, which expresses a relatively high level of endogenous SERINC5 like that observed in normal CD4 T cells. An isogenic Jurkat JTag cell population in which both SERINC5 alleles are inactivated by CRISPR/Cas9 was studied in parallel (21). Jurkat JTAg wild-type and SERINC5-knockout cells were transfected with HIV-1 proviral constructs bearing the same Nef dimerization interface mutations, along with wild-type and ∆Nef HIV-1. The infectivity of the viruses expressing the L112D, Y115D, and F121A Nef mutants produced in wildtype Jurkat JTAg cells was reduced to that of the ∆Nef virus, consistent with the results in 293T producer cells co-expressing SERINC5. The infectivity of these Nef mutant viruses was restored when produced in the SERINC5-knockout cells, providing evidence of a SERINC5-dependent mechanism in this T cell line ( Figure 4C). Wild-type HIV-1 produced from either of the Jurkat producer cell lines displayed similar infectivity, consistent with the ability of wild-type Nef to antagonize SERINC5. Infectivity of the Nef-L112A mutant virus was also less affected by the presence or absence of SERINC5, consistent with the lack of an effect of this mutation on Nef self-association in the BiFC assay. Expression of wild-type and mutant forms of Nef in Jurkat producer cells was verified by immunoblot analysis ( Figure 4D).
We also addressed whether Nef dimerization is important for infectivity of HIV-1 produced during a spreading infection in primary host cell culture. SERINC5-mediated restriction has been previously observed in donor PBMCs, which also express high levels of endogenous SERINC5 (21,22). For this experiment, PBMCs were infected with equivalent titers of wild-type, ∆Nef, and Nef dimer mutant viruses (initially produced in 293T cells) and allowed to replicate in PBMCs for 4 days followed by infectivity assay in TZM-bl cells. As observed in the other two producer cell systems, viruses expressing the Nef L112D, Y115D, and F121A mutations showed reduced infectivity comparable to that of the ∆Nef virus following amplification in PBMCs, with higher infectivity for wildtype and Nef-L112A HIV-1 ( Figure 4E).
We next investigated the requirement for Nef homodimers in the exclusion of SERINC5 from newly synthesized virions. For these experiments, 293T cells were co-transfected with HIV-1 wildtype, ∆Nef, and Nef dimerization-defective mutant proviral DNA in the presence of the SERINC5 expression plasmids. Newly synthesized virions were isolated and incorporation of SERINC5 was assessed by immunoblotting for the C-terminal HA epitope tag ( Figure 5A). The amount of virus associated SERINC5 was then normalized to p24 Gag levels from three independent experiments ( Figure  5B). SERINC5 incorporation into virions mirrored the infectivity data, with the L112D, Y115D, F121A, and ∆Nef virions exhibiting significantly higher levels of SERINC5 than the wild-type and L112A viruses. These findings demonstrate that enhancement of HIV-1 infectivity requires Nef homodimers and involves a mechanism linked to antagonism of SERINC5 incorporation by Nef dimers at the producer cell level. Interestingly, HIV-1 expressing the Nef-L112D mutant showed significantly higher incorporation of SERINC5 than virus that fails to express Nef (ΔNef). As described in the following section, this effect may be due to accumulation of SERINC5 at the producer cell membrane by this mutant due to attenuated endocytosis.

Nef
homodimerization enhances SERINC5 internalization and endosomal trafficking. Previous studies have implicated Nef homodimers in CD4 downregulation (25,26), a process that requires the endocytic trafficking adaptor protein, AP-2. Nef-mediated antagonism of SERINC5 is also mediated by AP-2-dependent downregulation of SERINC5 from the cell membrane, resulting in endocytosis and lysosomal degradation (22,24). Here we demonstrate a clear increase in virion-associated SERINC5 with HIV-1 expressing dimerization-defective Nef ( Figure 5), suggesting that impaired exclusion of SERINC5 from virions expressing these Nef mutants was due to dysfunctional downregulation of SERINC5 from the producer cell surface. To address this issue, the Jurkat SERINC5-knockout cells were transfected with expression vectors for Nef (wild-type or dimerization-defective mutants) which also expresses GFP as a gating marker and for SERINC5 with a hemagglutinin (HA) epitope tag on one of the extracellular loops (33). Transfected cells were then immunostained with an anti-HA antibody to detect cell-surface SERINC5 expression in the Nef-transfected (GFP+) cell population via flow cytometry. Expression of wild-type Nef decreased cell-surface SERINC5 expression by nearly 90% relative to the GFP-only control population. In contrast to wildtype Nef, the dimerization-defective mutants L112D, Y1125D, and F121A all showed significantly impaired SERINC5 downregulation, ranging from 65% to 85% relative to the control ( Figure 6A and 6B). The Nef-L112A mutant was not affected, consistent with the lack of a dimerization defect observed with this mutant. These results demonstrate that Nef homodimers are important for efficient downregulation of SERINC5 from the cell surface.
Because of the mechanistic similarity between Nef-induced downregulation of SERINC5 and CD4, we also explored the impact of the Nef dimer interface mutations on downregulation of endogenous CD4 in the CEM-SS T cell line (34). CEM-SS cells were transfected with the same Nef/GFP dual expression plasmid used for the SERINC5 study, followed by immunostaining for endogenous cell-surface CD4 and flow cytometry. Expression of wild-type Nef reduced cell surface CD4 expression by more than 60%, while the L112D, Y115D, and F121A Nef mutants were completely defective for CD4 downmodulation ( Figure  7A and 7B). Cells expressing Nef-L112A exhibited an intermediate phenotype, with about a 30% reduction in cell-surface CD4 expression. While the L112A mutant does not show a decrease in homodimer formation by BiFC assay, alteration of the dimer interface by this mutation may impact the efficiency of CD4 downregulation. Overall, these results agree with those observed with SERINC5 and are consistent with the shared mechanism of action involving AP-2-dependent endocytosis.
Following Nef-mediated endocytosis, SERINC5 traffics through Rab7+ late endosomes on route to proteolytic destruction following lysosomal fusion (21,24). We therefore investigated whether mutations disrupting Nef dimerization also impact the intracellular trafficking of SERINC5 along this pathway. SERINC5-knockout Jurkat cells were transfected with SERINC5 fused to the mCherry fluorescent protein in combination with wild-type or dimerization-defective forms of Nef fused to YFP. Two days later, transfected cells were immunostained for the late endosomal marker, Rab7. The expression and subcellular localization of SERINC5-mCherry, Nef-YFP, and Rab7 were then analyzed via confocal microscopy. When expressed alone, SERINC5-mCherry localized almost entirely to the plasma membrane with very little colocalization with Rab7 (confocal images of representative cells are shown in Figure 8). Co-expression of wild-type Nef-YFP resulted in a dramatic shift in SERINC5-mCherry localization from the cell surface into the Rab7+ endosomal compartment, observed as the co-localization of SERINC5-mCherry, Nef-YFP, and Rab7 fluorescence within multiple intracellular puncta. In marked contrast, the Nef dimer interface mutants, L112D, Y115D, and F121A all showed reduced intracellular targeting of SERINC5-mCherry, with both proteins retaining co-localization to the plasma membrane. Very few intracellular Rab7+ endosomes were observed with these three Nef mutants, consistent with a defect in endocytosis and intracellular trafficking. On the other hand, the Nef-L112A mutant displayed an intermediate phenotype, co-localizing with SERINC5-mCherry at the plasma membrane and in Rab7+ endosomes.
To quantify the extent of co-localization, Pearson's correlations were calculated for co-localization of the SERINC5-mCherry and Rab7 fluorescence for a minimum of 18 individual cells (Figure 9A). In the absence of Nef, the average Pearson's correlation coefficient was approximately 0.1,

HIV-1 Nef downregulates SERINC5 as a homodimer
indicating little correlation in terms of subcellular localization and consistent with the lack of SERINC5 internalization in the absence of Nef. Co-expression of Nef resulted in a significant increase in the correlation coefficient (mean σ > 0.5), consistent with the observed trafficking of SERINC5 to late endosomes. The correlation coefficient was significantly reduced with all four dimerization-defective Nef mutants compared to wild-type Nef, consistent with the confocal images showing retention of SERINC5-mCherry at the plasma membrane. As a control, we also performed correlation analysis for the co-localization of Nef-YFP and SERINC5-mCherry fluorescence ( Figure  9B). Notably, no significant difference was observed in these co-localization correlations (mean σ ≈ 0.9 in each case), hinting that the Nef dimer mutations may not affect interaction of Nef with SERINC5. This possibility was explored directly using BiFC as described in the next section.

Disruption of Nef homodimerization does not impair interaction with SERINC5 or AP-2.
Previous work suggests that Nef and SERINC5 interact directly within cells, with Nefmediated antagonism of SERINC5 mediated by a large intracellular loop within SERINC5 (24,35). While an X-ray crystal structure of SERINC5 has been elucidated recently (36), electron density for this intracellular loop was not observed suggesting that it is unstructured. Furthermore, the specific region of Nef responsible for interaction with SERINC5 is not known. While data presented above support a role for Nef homodimers in SERINC5 downregulation, the possibility remains that the amino acids involved in Nef homodimer formation may also contribute to interaction with SERINC5. To test this possibility, we employed the BiFC assay to visualize and quantify direct interaction of Nef with SERINC5 in cells (24).
SERINC5-VN and each of the Nef-VC fusion proteins (wild-type and mutants) were co-expressed in 293T cells followed by immunostaining for Nef and SERINC5 (via the HA tag) to control for protein expression levels. As a negative control, we also included a mutant of Nef that cannot be myristoylated (Nef-G2A) and therefore exhibits impaired membrane localization and SERINC5 interaction (24). Co-expression of wild-type Nef-VC with SERINC5-VN resulted in a positive BiFC signal, supporting direct protein-protein interaction in cells ( Figure 10A). Immunofluorescence both for Nef and SERINC5 expression was also readily observed in the same cells and colocalizes with the BiFC signal ( Figure S2, supplemental material). All four Nef-VC dimer interface mutants also generated a BiFC signal when partnered with SERINC5-VN, while the Nef-G2A mutant resulted in a lower BiFC signal despite a similar level of expression as wild-type Nef based on immunofluorescence. To quantify these results, the fluorescence intensity for the BiFC signal of each Nef interaction with SERINC5 was normalized to Nef expression levels as determined by immunofluorescence, and the resulting fluorescence intensity ratios are plotted for at least 100 cells in Figure 10B. No significant difference in SERINC5 interaction was observed between wild-type Nef and the dimerization-defective mutants, demonstrating that disruption of the Nef homodimerization interface does not affect interaction with SERINC5. In contrast, the Nef-G2A mutant showed a significant decrease in SERINC5 interaction, consistent with the inability of this mutant to localize to the plasma membrane. These observations suggest that impaired antagonism of SERINC5 by the Nef dimer mutants is not due to a defect in SERINC5 interaction with Nef.
Downregulation of SERINC5 by Nef also requires interaction with AP-2 through a complex of its α and σ2 subunits (19), resulting in endocytosis and lysosomal degradation. Structural alignment described above shows that the dimer interface of Nef does not overlap with regions of AP-2 contact, which involve a distinct intracellular loop in the Nef core (see Figure 1B). To determine whether Nef dimer interface mutations affect AP-2 subunit interaction, we employed a similar BiFC approach in which each Nef protein was fused to the N-terminal fragment of Venus (Nef-VN) while the AP-2 α and σ2 subunits were fused to the Cterminal fragment (α-VC and σ2-VC). The AP-2 fusion proteins also included a V5 tag for immunofluorescence detection as described previously (25). Each AP-2-VC construct was co-expressed with the wild-type and mutant forms of Nef-VN in 293T cells, followed by immunofluorescence detection of protein expression and confocal microscopy. Wildtype Nef as well as the dimer interface mutants all exhibited membrane-localized BiFC signals when co-expressed with AP-2 α-VC or σ2-VC ( Figure  10A). We then calculated normalized BiFC inter-HIV-1 Nef downregulates SERINC5 as a homodimer action ratios for Nef with each AP-2 subunit as described above for SERINC5 ( Figures 10B). The Nef L112A, L112D and F121A mutants interacted with each AP-2 subunit to the same extent as wildtype Nef, while the Y115D mutant ratio was somewhat reduced despite membrane localization. The Nef-G2A mutant showed a significant loss of interaction with both AP-2 subunits, consistent with its inability to localize to the plasma membrane. Expression of each Nef and AP-2 protein was confirmed by immunofluorescence microscopy, and representative images are presented in Figures S3  and S4 (supplemental material). These experiments show that amino acids present in the Nef dimer interface, except for Tyr115, are not involved in AP-2 interaction at the cell membrane.

Discussion
Antagonism of the host cell restriction factor SERINC5 is a key mechanism by which Nef promotes HIV-1 infectivity. Nef triggers internalization of SERINC5 from the surface of infected cells, thereby preventing incorporation of SERINC5 into budding virions and subsequent impairment of viral entry (21,22). While the cellular mechanisms of Nef-mediated SERINC5 internalization and degradation have been established, little is known about the structural features of Nef required for SERINC5 engagement and association with AP-2 for downregulation. In this study, we assessed the function of dimerization defective Nef mutants across all stages of SERINC5 antagonism including viral infectivity, virion incorporation of SERINC5, downregulation of SERINC5 from the surface of producer cells, and the intracellular fate of SERINC5. We observed consistent impairment of all stages of SERINC5 antagonism with three distinct dimerization defective Nef mutants, demonstrating that Nef counteracts SERINC5 as a homodimer. Importantly, each dimerization-defective Nef mutant retained interaction with both SERINC5 and the AP-2 subunits required for internalization, supporting the idea that the Nef homodimer may act as a bridge to link SERINC5 to the AP-2 pathway as illustrated in Figure 11.
Our data show that Nef dimerization promotes the infectivity of newly synthesized HIV-1 particles produced by transfected 293T cells as well as a T cell line in a SERINC5-dependent manner. These results may provide a mechanistic explanation for previous studies in which similar mutations were shown to influence viral infectivity. For example, substitution of Nef amino acids Leu112 and Phe121 has been reported to negatively impact HIV-1 infectivity and replication (37), although the role of SERINC5 was not known at that time. These same residues have also been linked to Nef recruitment of the endocytic GTPase Dynamin-2 as well as the human peroxisomal thioesterase 8 (37,38), suggesting a role for Nef homodimers in these interactions as well.
We observed that Nef-defective HIV-1 (ΔNef virus) produced in 293T cells in the absence of SERINC5 co-expression was less infectious than wild-type HIV-1 ( Figure 4A). This finding suggests that 293T cells may express an endogenous Nefsensitive restriction factor unrelated to SERINC5. Wu et al. also reported the existence of an unidentified, Nef-sensitive restriction factor other than SERINC5 (39). In our study, this additional factor appears to be limited to 293T cells, as wild-type and ΔNef HIV-1 produced from SERINC5-knockout Jurkat T cells showed equivalent infectivity.
Nef dimerization enhances SERINC5 downregulation from the cell surface, which is mechanistically like Nef-dependent downregulation of CD4. Both SERINC5 and CD4 downregulation are mediated via AP-2 and clathrin-dependent endocytosis. Mutations that disrupt Nef myristoylation and membrane association (G2A) as well as the dileucine motif required for AP-2 recruitment (LL164,165AA) impair downregulation of both proteins (21,40,41). Nef dimerization-defective mutants impaired for downregulation of SERINC5 were also unable to downregulate CD4. Expression of wild-type Nef in Jurkat T cells led to the re-localization of SERINC5 from the cell membrane into intracellular Rab7+ compartments (24) a fate also shared by CD4 (42). Intracellular trafficking and Rab7 co-localization were also significantly disrupted by mutations in the Nef dimerization interface. These findings suggest that Nef homodimers may link both SERINC5 and CD4 to the AP-2 endocytic pathway through separate binding surfaces on each Nef monomer ( Figure 11).
BiFC assays presented here demonstrate direct Nef•SERINC5 interaction at the plasma membrane in cells, consistent with recent work (24,43). Importantly, mutations in residues essential for Nef homodimer formation do not impact interaction of Nef with SERINC5 in the BiFC assay, consistent HIV-1 Nef downregulates SERINC5 as a homodimer with a role for the Nef αB helix in homodimer formation as demonstrated by X-ray crystal structures (27,28) rather than interaction with signaling partners. While the mechanism of SERINC5 engagement by Nef remains enigmatic, recent studies have illuminated the regions of SERINC5 required for viral restriction and susceptibility to Nef antagonism (36). SERINC5 is an integral membrane protein with ten trans-membrane helices and four intracellular loops (36). Sensitivity of SERINC5 to Nef antagonism appears to map to the fourth intracellular loop, as deletion or replacement of this loop with the Nef-resistant SERINC2 counterpart renders SERINC5 immune to Nef downregulation (35). However, whether this loop is sole point of contact in the Nef•SERINC5 complex remains an important open question.
The SERINC5 binding site within Nef also remains elusive. A recent structural study of SIVmm Nef in complex with AP-2 and tetherin showed that mutation of a critical residue within the tetherin binding site (His196) also impairs antagonism of SERINC5, suggesting a shared binding mechanism (41). In HIV-1 Nef, the analogous residue (His166) is in the flexible internal loop adjacent to the dileucine motif essential for AP-2 σ2 engagement. This Nef region is separate from the αB helix involved in self-association ( Figure 1C), consistent with the idea that Nef binds SERINC5 as a dimer. Nevertheless, a direct role for this binding site in HIV-1 Nef SERINC5 association remains to be confirmed.
Though the binding site for SERINC5 within Nef remains to be identified, our BiFC data clearly show that conserved residues in the dimer interface are not required for SERINC5 engagement. Nef dimer interface mutants L112A, L112D, and F121A also retained interaction with the AP-2 α and σ2 subunits via BiFC, consistent with the distinct interfaces for homodimerization and AP-2 recruitment illustrated by structural alignment in Figure 1C. However, the Nef Y115D mutant demonstrated statistically lower association with both AP-2 subunits. While the Y115D mutation may impact Nef folding or overall stability, unimpaired interaction of this mutant with SERINC5 suggests otherwise. Another possibility is that Tyr115 may have an allosteric influence on AP-2 binding by Nef.
In summary, disruption of Nef dimerization attenuates SERINC5 antagonism without impairing recruitment of SERINC5 or AP-2 as required for internalization. While the protein dynamics controlling Nef-dependent SERINC5 downregulation remain a mystery, our data support a mechanism in which Nef homodimers promote Nef•SERINC5•AP-2 complex formation at some point prior to SERINC5 internalization. One Nef monomer within the dimer may contact SERINC5 while the other engages AP-2, with the Nef dimer interface bridging the two proteins to initiate clathrin-mediated endocytosis. Understanding the molecular determinants of SERINC5 antagonism by Nef may help to identify therapeutic strategies that promote SERINC5 virion incorporation and restriction of viral infectivity. Along these lines, docking studies of small molecule inhibitors of Nefmediated enhancement of viral infectivity have identified potential binding sites within the dimer interface (12,44,45). Disruption of the Nef homodimer by these compounds may explain their inhibitory effects on multiple Nef functions, including not only infectivity but also host cell kinase activation, receptor downregulation, and enhancement of viral replication. Thus, the homodimer interface represents an attractive target for new antiretroviral drug development targeting Nef.

Experimental procedures Cell Culture
The human embryonic kidney cell line 293T was purchased from the ATCC. TZM-bl reporter cells were obtained from the AIDS Research and Reference Reagent Program. Both cell lines were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (FBS; Gemini Bio-Products). Jurkat JTAg T cells (wild-type and SERINC5-knockout) were a kind gift from Dr. Massimo Pizzato, University of Trento, while the CEM-SS T cell line kindly provided by Dr. Mark Brockman, Simon Fraser University. Both T cell lines were cultured in RPMI 1640 medium supplemented with 10% FBS, L-glutamine, and 2 µg/mL puromycin (CEM-SS cells only). Donor PBMCs were obtained from Vitalant Pittsburgh and isolated by Ficoll gradient centrifugation and activated for 3 days with PHA (Sigma, # L1668) and IL-2 (Thermo Fisher, # CB-40043B) as described elsewhere (44).

HIV-1 replication and infectivity assays
Human 293T cells were transfected with proviral expression plasmids using Xtreme Gene 9 (Sigma-Aldrich) for 48 h prior to harvesting the viral supernatant.
Jurkat JTAg wild-type and SERINC5-knockout cells were transfected with proviral plasmids using the Amaxa Cell Line Nucleofector Kit V and the resulting viral supernatant was harvested 48 h later. PBMCs were infected with 150 pg/mL of HIV-1 (wild-type, ∆Nef, and Nef mutants) for 4 days prior to assay. HIV-1 titers in clarified supernatants from all producer cells were quantified using the PerkinElmer p24 Al-phaLISA Detection kit. Each viral supernatant (5,000 pg/mL p24) was then used to infect TZM-bl cells in 96-well plates (2.5 x 10 4 cells/well). After 48 h, cells were lysed in Promega luciferase buffer, transferred to a white 96-well plate, mixed with luciferase reagent, and measured using a Biotek Cytation 5 plate reader.

Flow cytometry
Jurkat SERINC5-knockout cells and CEM-SS cells were transfected with the pSELECT-GFPzeo expression vector encoding wild-type and mutant forms of Nef, pSELECTGFPzeo alone and/or pBJ-SERINC5-iHA as described above. Forty-eight h later, cells were washed in 1X Phosphate Buffered Saline (PBS) supplemented with 2% fetal bovine serum, and stained for 1 h at ambient temperature with APC-conjugated anti-HA.11 (to detect SERINC5-HA) or phycoerythrin-conjugated anti-human CD4 antibody (both antibodies diluted 1:200). After washing, cells were analyzed for GFP (to quantify Nef expression) and either HA (to quantify JTAg cell surface S5-iHA expression) or phycoerythrin (to quantify CEM-SS cell surface CD4 expression) using a BD Accuri Flow Cytometer. Data were analyzed via BD CSampler software and illustrated using the FlowJo software package.

BiFC and Confocal Microscopy
293T cells were plated onto 35 mm coverslip-bottom dishes (MatTek, # P35G-1.5-14-C) and allowed to attach overnight. Cells were then transfected with pcDNA3.1 expression vectors encoding complementary BiFC partners using Xtreme Gene 9. Forty-eight h later, cells were fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.2% Triton X-100 for 15 min and blocked with 2% BSA in PBS for 1 h. Cells were then stained overnight at 4 °C with primary antibodies directed at the V5 tag, the HA tag, or Nef (#153). Following washing, cells were stained for 1 h with secondary antibodies conjugated to Texas Red (1:1000), Pacific Blue (1:500), or Alexa Fluor 647 (1:1000). Fluorescent images were acquired with an Olympus Fluoview FV1000 confocal microscope using the 40x objective and x-y scan mode. Immunofluorescence and BiFC signal intensities were quantified via Image J and are represented as the ratio of BiFC to Nef immunofluorescence as reported previously (11).
Co-localization of SERINC5 with Nef and Rab7+ endosomes was assessed in Jurkat JTAg SERINC5-knockout cells following transfection with SERINC5-mCherry and/or Nef-YFP (wildtype and dimer mutants) expression vectors as described above. Coverslip-bottom 35 mm dishes (MatTek) were treated with a 0.01% poly-L-Lysine solution (Sigma) for 1 h at 37 °C, and left to air-dry for 30 min. Transfected cells (1.2 x 10 6 in 200 µL of RPMI) were added to the poly-L-lysine-treated coverslips for 30 min at 37 °C. Cells were then fixed with 4% paraformaldehyde for 10 min, permeabilized with 0.2% Triton X-100 for 15 min, and blocked with 2% BSA in PBS for 1 h at room temperature. Cells were probed with a mouse monoclonal Rab7 antibody (1:1000) overnight at 4 °C, and then stained with an anti-mouse Alexa Fluor 647-conjugated secondary antibody (1:1000). Cells were imaged using an Olympus Fluoview FV1000 confocal microscope with a 60x oil objective and xy scan mode. Pearson's correlation coefficients for co-localization were generated using the FV1000 Fluoview software.
Data availability: All data are contained within the manuscript and the associated supporting information file.

46.
Narute, P. S., and Smithgall, T. E. (2012) Nef alleles from all major HIV-1 clades activate Src-family kinases and enhance HIV-1 replication in an inhibitor-sensitive manner. PLoS. One 7, e32561 proteins are not shown for clarity). The Nef monomers (NefA and NefB) are rendered in dark and light blue, respectively. B) Zoomed and rotated view of the Nef dimerization interface in which conserved residues Leu112, Tyr115, and Phe121 (cyan, orange, and green, respectively) form a hydrophobic cup around Val70 (pink) from the opposing monomer. C) One Nef monomer from the homodimer shown in part A was aligned with Nef in complex with the AP-2 α/σ2 hemicomplex (PDB: 4NEE). The AP-2 α and σ2 subunits are shown as surfaces in gray and light orange, respectively. This alignment shows that the Nef homodimer interface is structurally distinct from the point of contact with AP-2, which involves the Nef internal loop (not shown). 293T cells were co-transfected with expression plasmids for wild-type and mutant Nef proteins fused to non-fluorescent complementary fragments of the YFP variant, Venus. Cells were fixed 12, 24, 36, and 48 hours post-transfection and immunostained for Nef expression prior to imaging via confocal microscopy. Nef homodimerization results in reconstitution of the Venus fluorophore. A) Representative images comparing self-association of wild-type and mutant Nef proteins (BiFC; green). Nef immunofluorescence is shown in red (Nef IF). B) Time course of Nef dimerization by BiFC assay. The mean fluorescence intensity ratios of the BiFC and Nef immunofluorescence signals were calculated from a minimum of 100 cells for each condition using ImageJ. The entire experiment was repeated in duplicate, and each data point represents the mean ratio ± SD. Complete data and statistical analysis for representative cell populations at all time points are shown in the Figure S1 (supplemental material). A) 293T cells were co-transfected with HIV-1 NL4-3 proviral plasmids competent (wild-type) or defective (∆Nef) for Nef expression along with 0, 25, or 50 ng of a SERINC5 expression vector. Newly released virions were harvested after 48 h, quantified via p24 AlphaLISA, and used to infect TZM-bl reporter cells. Luciferase activity was measured 48 h later and normalized to that observed with wild-type virus produced in the absence of SERINC5. Mean values ± SE are shown from three independent experiments. Significance was assessed via t-test relative to each control; *, p < 0.01; **, p < 0.001. B) Representative western blot depicting incorporation of SERINC5-HA into wild-type and ∆Nef virions. Virions were collected in parallel during the infectivity assays and purified via ultracentrifugation through a 20% sucrose cushion, lysed into sample buffer, subjected to SDS-PAGE, and immunoblotted using antibodies to p24 Gag and the HA tag on SERINC5. Band intensities were quantified using LI-COR infrared imaging and the SERINC5:p24 ratios are shown. C) Viral antagonism of SERINC5 is rescued following ectopic expression of Nef within producer cells. The ∆Nef virus was expressed in 293T cells in the presence and absence of SERINC5 and Nef expression plasmids. Data are presented as the fold enhancement of viral infectivity, defined as the ratio of infectivity between Nef-expressing and Nef-deficient producer cells. Results represent normalized mean values ± SE from three independent experiments. Significance was assessed via t-test; **, p < 0.001. D) Representative immunoblots displaying incorporation of SERINC5 into purified ∆Nef virions produced in the presence and absence of Nef. Band intensities were quantified using LI-COR infrared imaging and the SERINC5:p24 ratios are shown. Each bar represents the mean value ± SE. B) Representative immunoblots verifying expression of wildtype and dimerization-defective Nef mutants in lysates of transfected 293T cells. Actin blots are shown as a loading control. C) Wild-type and SERINC5-knockout Jurkat TAg T cells were transfected with wildtype, ∆Nef, and dimerization-defective Nef mutant proviruses. Infectivity of equivalent p24 inputs was analyzed 48 h later in TZM-bl reporter cells. Results from three independent experiments are shown normalized to the infectivity of wild-type virus produced in wild-type Jurkat cells, with each bar representing the mean value ± SE. D) Representative immunoblots verifying expression of wild-type and dimerizationdefective Nef mutants in lysates of transfected Jurkat cells. Actin blots are shown as a loading control. E) PBMCs from uninfected donors were activated with PHA and IL-2 for three days, then infected with 150 pg p24/mL of wild-type, ∆Nef, and Nef dimerization-defective mutant viruses for four days. Infectivity of equivalent p24 inputs of the resulting viral supernatants were analyzed 48 h later in TZM-bl reporter cells.
Results from three independent experiments are shown normalized to the infectivity of wild-type virus, with each bar representing the mean value ± SE. In panels A and C, statistical significance was assessed between the indicated pairs of values by t-test, while in panel E each comparison was made vs. the wildtype control (*, p < 0.05; ***, p < 0.001; **** p < 0.0001; ns, not significant). Newly synthesized virions were purified by ultracentrifugation through a 20% sucrose cushion, and SERINC5 incorporation was analyzed via immunoblotting along with p24 Gag as a control. A) Representative SERINC5 and p24 immunoblots. B) SERINC5 and p24 protein expression levels were quantified by LI-COR Odyssey image analysis, and each data point represents the SERINC5 to p24 expression ratio for three independent experiments. Mean values in each group are indicated by the horizontal bar ± SE. Statistical significance relative to wild-type (WT) was determined by Student's t-test (**, p ≤ 0.01; ****, p ≤ 0.0001; ns, not significant). Note that Nef-L112D also incorporated significantly more SERINC5 than ΔNef; ***, p ≤ 0.001.   http://www.jbc.org/ Figure 9: Disruption of Nef dimerization impairs internalization and trafficking of SERINC5-mCherry to Rab7+ endosomes. SERINC5-knockout Jurkat T cells were co-transfected with expression plasmids expressing SERINC5-mCherry together with wild-type or dimerization-defective Nef mutants fused to YFP and immunostained with an antibody to the late endosomal marker Rab7 as described in the legend to Figure 8. A) Single cell confocal images were subjected to Pearson's correlation coefficient analysis to quantify the extent of SERINC5-mCherry colocalization with Rab7. A minimum of 18-20 cells per condition were analyzed across three independent experiments, and the Pearson's correlation coefficient for each cell is shown with the median value indicated by the horizontal bar. Statistical significance among the groups was performed via one-way ANOVA with Dunnett's multiple comparisons test relative to the cell population expressing wild-type (WT) Nef (****, p ≤ 0.0001). B) Pearson's correlation analysis of SERINC5-mCherry co-localization with Nef-YFP was performed in the same groups of cells, with median correlation coefficients indicated by the horizontal bars. Statistical analysis via ordinary one-way ANOVA did not reveal significant differences between the groups.