The HIV accessory protein Nef is a major determinant of viral pathogenesis that facilitates viral particle release, prevents viral antigen presentation and increases infectivity of new virus particles. These functions of Nef involve its ability to remove specific host proteins from the surface of infected cells, including the CD4 receptor. Nef binds to the adaptor protein 2 (AP-2) and CD4 in clathrin-coated pits, forcing CD4 internalization and its subsequent targeting to lysosomes. Herein, we report that this lysosomal targeting requires a variant of AP-1 containing isoform 2 of γ-adaptin (AP1G2, hereafter γ2). Depletion of the γ2 or μ1A (AP1M1) subunits of AP-1, but not of γ1 (AP1G1), precludes Nef-mediated lysosomal degradation of CD4. In γ2-depleted cells, CD4 internalized by Nef accumulates in early endosomes and this alleviates CD4 removal from the cell surface. Depletion of γ2 also hinders EGFR–EGF-complex targeting to lysosomes, an effect that is not observed upon γ1 depletion. Taken together, our data provide evidence that the presence of γ1 or γ2 subunits delineates two distinct variants of AP-1 complexes, with different functions in protein sorting.
CD4 is a type I transmembrane glycoprotein that, in T cells, functions as a co-receptor, together with the T cell antigen receptor, in the recognition of antigenic peptides bound to major histocompatibility complex (MHC)-II molecules on antigen-presenting cells. Human cells that express CD4, such as T cells and cells of the monocyte and macrophage lineage, are the main targets for human immunodeficiency virus (HIV) infection. Binding of the HIV envelope glycoprotein (ENV) complex to cell surface CD4 is the initial step in virus entry, and is crucial for infection. During the early stages of infection, one of the most abundantly produced HIV proteins is Nef, an accessory protein with no enzymatic activity. Nef is a 27–35-kDa myristoylated protein encoded by the HIV-1, HIV-2 and the simian immunodeficiency virus (SIV) genomes, which is crucial for sustaining robust viral production and promoting disease progression to full-blown AIDS (Deacon et al., 1995; Gorry et al., 2007; Gulizia et al., 1997; Kestier et al., 1991; Learmont et al., 1999; Oelrichs et al., 1998).
The first function ascribed to Nef was the ability to dramatically reduce cell surface levels of CD4 (Garcia and Miller, 1991; Guy et al., 1987; reviewed by Pereira and daSilva, 2016). Nef-mediated downregulation of CD4 is important to counteract the detrimental effects of high CD4 expression in HIV-infected cells, such as to prevent superinfection (Benson et al., 1993) and to facilitate the release of viral progeny (Cortés et al., 2002; Lama et al., 1999). CD4 downregulation is a highly conserved function of Nef (Mariani and Skowronski, 1993) that is not restricted to natural HIV host cells, but is also active in other animal cell types (Chaudhuri et al., 2007). Nef expression does not alter the levels of CD4 synthesis or the kinetics by which newly synthesized CD4 molecules reach the plasma membrane (Rhee and Marsh, 1994). Instead, Nef accelerates the endocytosis of CD4 and targets the internalized molecules to the lumen of lysosomes for degradation (Aiken et al., 1994; Rhee and Marsh, 1994). To accomplish this, Nef directly binds the cytosolic tail of CD4 (Grzesiek et al., 1996; Preusser et al., 2001; Rossi et al., 1996) and the endocytic adaptor protein 2 (AP-2) (Chaudhuri et al., 2007, 2009; Doray et al., 2007; Lindwasser et al., 2008). This activity promotes the cooperative assembly of a tripartite Nef–AP-2–CD4 complex (Chaudhuri et al., 2009; Ren et al., 2014) that induces the endocytosis of CD4 mediated by clathrin-coated vesicles (CCVs) (Burtey et al., 2007; Greenberg et al., 1997).
CD4 molecules that have been internalized by Nef do not efficiently enter a retrieval pathway from early endosomes back to the plasma membrane, but are instead delivered to late endosomes or multivesicular bodies (MVBs) en route to lysosomes (daSilva et al., 2009; Schaefer et al., 2008). Delivery of transmembrane cargo to the intraluminal vesicles (ILVs) of MVBs, in most cases, requires the recognition of sorting signals by the endosomal sorting complexes required for transport (ESCRTs). Transmembrane proteins that undergo ILV sorting are typically modified by the covalent attachment of monoubiquitin to their cytosolic tail, which serves as a sorting signal for ESCRTs (Shields and Piper, 2011). Although independent of CD4 ubiquitylation, Nef-mediated depletion of CD4 requires functional ESCRTs (daSilva et al., 2009). Alix (also known as PDCD6IP), an ESCRT-accessory protein, that directly binds Nef (Amorim et al., 2014; Costa et al., 2006; Jesus da Costa et al., 2009), was recently shown to be required for Nef-mediated targeting of CD4 to lysosomes (Amorim et al., 2014). However, depletion of Alix or of Tsg101 (an ESCRT-I component), which inhibits lysosomal targeting of CD4 by Nef, does not impair the Nef-mediated reduction of CD4 surface levels (Amorim et al., 2014; daSilva et al., 2009). This observation suggests that in addition to mediating MVB sorting through Alix, Nef might also promote endosomal retention of CD4 earlier in the pathway to MVBs.
The adaptor proteins (APs) comprise a family of five heterotetrameric complexes in mammals (Bonifacino, 2014; Robinson, 2015). In addition to AP-2, formed by α, β2, μ2 and σ2 subunits (Table S1), two other AP complexes, AP-1 (γ1, β1, μ1 and σ1) and AP-3 (δ, β3, μ3 and σ3), are binding partners of Nef (Bresnahan et al., 1998; Chaudhuri et al., 2007; Coleman et al., 2005; Doray et al., 2007; Greenberg et al., 1998; Janvier et al., 2003a,b; Mattera et al., 2011). However, the role of the Nef interaction with AP-1 or AP-3 in CD4 downregulation is poorly understood. AP-1 facilitates CCV-mediated trafficking between the trans-Golgi network (TGN) and endosomes, whereas AP-3 participates in the trafficking to lysosomes and lysosome-related organelles, possibly through clathrin-dependent and clathrin-independent pathways (Robinson, 2015). In addition, it is possible that functional variants of the canonical AP-1, AP-2 and AP-3 complexes exist, given that some of their subunits occur as multiple isoforms encoded by different genes (Table S1). These isoforms include two γ (γ1 and γ2), two μ1 (μ1A and μ1B) and three σ1 (σ1A, σ1B and σ1C) subunits for AP-1; two α subunits for AP-2; and two μ3 (μ3A and μ3B), two β3 (β3A and β3B) and two σ3 (σ3A and σ3B) subunits for AP-3 (Boehm and Bonifacino, 2002). Nef binds to APs through a conserved ‘dileucine-based’ (D/E)xxxL(L/I) consensus motif in the viral protein (Craig et al., 1998; Greenberg et al., 1998), which is recognized by hemi-complexes comprising the γ1–σ1, α–σ2 and δ–σ3 subunits of the respective AP (Chaudhuri et al., 2007; Coleman et al., 2005; Doray et al., 2007; Janvier et al., 2003a,b). Subsequently, it has been shown that γ2 is also capable of mediating a σ1-dependent interaction with Nef (Mattera et al., 2011). However, the capacity of γ2 to act as a subunit of a functional variant of the canonical (γ1-containing) AP-1 complex is currently unknown.
To better understand the mechanism by which Nef induces the sustained cellular depletion of CD4, we examined the functional role of the Nef interaction with AP-1. Here, we show that endogenous γ1 and γ2 isoforms present partially distinct subcellular distributions, and efficient targeting of CD4 to lysosomes by Nef requires γ2, but not γ1. Depletion of γ2 compromises CD4 degradation and partially restores the surface levels of CD4 in cells expressing Nef. Similarly, μ1A depletion leads to the impairment of Nef-mediated CD4 degradation, which is not observed upon γ1-depletion. Our results also suggest that γ2 expression is essential for efficient lysosomal targeting of EGFR–EGF complexes, whereas γ1 expression is dispensable in this process. We propose that γ2 functions as a connector, promoting the incorporation of Nef-associated CD4 molecules into the MVB pathway. Moreover, these findings support the notion that γ1 and γ2 play distinct functions in protein trafficking, and provide evidence for the existence of functionally distinct AP-1 complexes, depending on the γ-subunit isoform composition.
Binding of Nef to AP-1 is mediated by the γ1–σ1A and γ2–σ1A hemi-complexes
Considering the evidence that Nef binds two variants of γ–σ1A hemi-complexes of AP-1, comprising either the γ1 or the γ2-isoforms of the γ-subunit (Mattera et al., 2011), we asked whether these isoforms could play distinct roles in Nef mediated downregulation of CD4. As has been previously shown by yeast three-hybrid (Y3H) assays, the interaction of Nef with γ1–σ1A requires a conserved (D/E)xxxL(L/I) signal in Nef (i.e the ENTSLL165 motif in HIV-1 NL4-3 Nef), and a conserved basic residue in γ1 at position 15 (Arg15) (Chaudhuri et al., 2007; Mattera et al., 2011) (Fig. 1A). The requirement of a basic residue at a similar position has also been shown in the interactions of Nef with the α–σ2 (α Arg21) or the δ–σ3 (δ Arg26) hemi-complexes of AP-2 and AP-3, respectively (Mattera et al., 2011). However, the sequence requirements for the interaction of Nef with the γ2–σ1A hemi-complex have not been described. Given that the structural basis of the interaction of Nef with α Arg21 has been previously elucidated (PDB ID 4NEE) (Ren et al., 2014), we evaluated, by homology modeling, whether a similar interaction was predicted between Nef and each of the γ subunits. Similar to the salt bridge interactions of Nef Glu160 with α Arg21 and σ2 Arg15 (Ren et al., 2014), the homology models predicted that Nef Glu160 could participate in salt bridge interactions with either γ1 Arg15 and σ1A Arg15 (Fig. 1B), or with γ2 Arg16 and σ1A Arg15 (Fig. 1C). The homology models predicted additional Nef residues participating in interactions with either of the γ–σ1A hemi-complexes (Fig. S1A,B). In addition, Nef Arg178 could be accommodated in a concave pocket predicted for each of the γ–σ1A hemi-complexes (Fig. S1C,D), stabilizing the interactions in a similar manner to the α–σ2 hemi-complex (Ren et al., 2014).
Our Y3H analysis confirmed that Nef binds to both the γ1–σ1A and γ2–σ1A hemi-complexes (Fig. 1D,E). Although Nef might interact more weakly with γ2–σ1A than with γ1–σ1A, as previously observed by Mattera et al. (2011), the interaction was dependent on the ENTSLL165 motif (Fig. 1D) and required σ1A coexpression (data not shown). This possibly weaker interaction of Nef with γ2–σ1A correlates with the fewer number of hydrogen-bond and salt bridge interactions predicted by the homology model (Fig. S1A,B). Next, the importance of γ2 Arg16 in the interaction with Nef was analyzed. Similar to upon the substitution of γ1 Arg15, we found that the substitution of γ2 Arg16 greatly reduced the binding to Nef (Fig. 1D,E). Taken together, our observations suggest that both γ1 and γ2 are able to interact with Nef-WT, but not with Nef-LL/AA (where L164 and L165 are mutated into alanine residues), and that Arg15 and Arg16 in γ1 and γ2, respectively, contribute to stabilize these interactions.
Nef requires γ2, but not γ1, for efficient targeting of CD4 to lysosomes
To comparatively assess the function of the Nef interaction with AP-1 through γ1 or γ2, we tested, by performing RNA interference (RNAi) studies, whether the specific depletion of either γ1 or γ2 affected Nef-induced downregulation of CD4 in HeLa cells. Knockdown (KD) of γ1 or γ2, mediated by small interfering RNA (siRNA), resulted in 77.9% (±4.7, mean±s.e.m., n=4) and 71.9% (±2.3, n=4) reduction in their expression levels, respectively (Fig. 2A). Moreover, siRNA directed to either γ1 or γ2 did not affect the expression levels of the other isoform (Fig. 2A). Immunoblot analysis confirmed that the expression of Nef decreased the total levels of CD4 (Fig. 2A,B). Nef-LL/AA or Nef-L/A (L165 mutated to an alanine residue) mutants were used as negative controls. These mutants did not interact with γ1–σ1, α–σ2, δ–σ3 (Chaudhuri et al., 2007; Mattera et al., 2011) or γ2–σ1 (Fig. 1D) hemi-complexes, and failed to reduce the total cell levels of CD4 (Fig. 2A,B), as previously shown (daSilva et al., 2009). The KD of γ1 did not significantly alter CD4 depletion by Nef-WT (Fig. 2A,B). In contrast, γ2 KD increased the steady state levels of CD4 in control conditions (i.e. in Nef-L/A-expressing cells) and impaired Nef-induced depletion of CD4 (Fig. 2A,B). Treatment of cells with an alternative γ2 siRNA sequence (herein termed γ2 siRNA #2), which produced a 61.7% (±3.8) reduction in γ2 expression (Fig. S2A,B) also impaired Nef-mediated CD4 depletion (Fig. S2C,D). Moreover, the degradation of CD4 induced by two other HIV-1 Nef alleles (variants NA7 and 248), and by the SIV Nef variant SIVmac293, previously shown to downregulate CD4 (Chaudhuri et al., 2007), was also inhibited in γ2 KD cells (Fig. S2E). Taken together, these results indicate that γ2 is required for the process of Nef-mediated targeting of CD4 to lysosomes, and also for normal CD4 turnover.
AP-1 is thought to participate in the transport of mannose-6-phosphate receptors (MPRs), which mediate sorting of acid hydrolases to lysosomes (reviewed by Ghosh et al., 2003). To rule out that the impairment of Nef-mediated degradation of CD4 in γ2 KD cells was due to a block in the delivery of acid hydrolases to lysosomes, we monitored lysosomal processing and secretion of the acid hydrolase cathepsin D (catD) by pulse-chase analysis. Cells were metabolically labeled with [35S]methionine-cysteine for 20 min, and after a 5-h chase, catD species were isolated by immunoprecipitation from both cell extracts and culture medium and resolved by SDS-PAGE and fluorography (Fig. S3A). In control cells, the precursor catD (pro-catD) was cleaved to generate either an intermediate (int-catD) or a mature lysosomal form (m-catD), with only trace amounts of pro-catD being secreted into the medium over the period of 5 h (Fig. S3). Depletion of γ1 led to the increased secretion of pro-catD into the medium and a consequent reduction in the amount of intracellular m-catD (Fig. S3), indicating a partial impairment of transport to lysosomes, as previously observed by Hirst et al., (2009). In γ2 KD cells, the ratio of intracellular m-catD and secreted pro-catD was similar to in control cells (Fig. S3A). We further analyzed the effect of γ2 KD in the subcellular distribution of catD by immunofluorescence confocal microscopy. In control cells, catD exhibited its characteristic localization at vesicular structures scattered throughout the cytoplasm (Fig. S3B). Although γ1 KD led to a greatly reduced staining for catD, the staining and distribution pattern for catD in γ2 KD cells were similar to in cells treated with control siRNA (Fig. S3B,C). These data suggest that deficient lysosomal delivery of hydrolases does not account for the reduced CD4 degradation in γ2 KD cells. Instead, γ2 is likely to play a more direct role in targeting Nef-internalized CD4 to lysosomes. Moreover, these results show that although γ1 function is impaired in our γ1 KD cells, this impairment is not sufficient to preclude CD4 degradation induced by Nef (Fig. 2A,B). In fact, there is evidence that the Golgi-localized, γ-ear containing, ARF-binding proteins (GGAs) help to mediate sorting of lysosomal hydrolases and could compensate for the lack of γ1 (Hirst et al., 2009).
Knockdown of γ2 alleviates surface depletion of CD4 by Nef
To investigate the differential requirements of γ1 and γ2 in the downregulation of CD4 by Nef, the levels of surface CD4 in cells expressing Nef were analyzed. Quantitative analyses by flow cytometry showed that the levels of CD4 on the surface of HeLa cells expressing Nef-WT were reduced by 88.8% (±4.8; mean±s.e.m.), when compared to control Nef-LL/AA cells (Fig. 2C,D). A smaller, 33.8% (±3.5), reduction in surface levels of CD4 was observed in γ2 KD cells upon expression of Nef-WT (Fig. 2C,D), indicating a partial impairment in Nef activity. Interestingly, the surface levels of CD4 were increased ∼twofold in control (Nef-LL/AA) cells depleted of γ2, confirming our previous observation that γ2 is required for normal CD4 turnover (Fig. 2A,B). As expected, depletion of γ1 did not affect CD4 downregulation by Nef (Fig. 2C,D). Taken together, the results shown in Fig. 2 indicate that γ2 is required for efficient targeting of CD4 to lysosomes by Nef. In contrast, γ1 expression is dispensable for CD4 downregulation by Nef, as previously reported (Leonard et al., 2011).
Additional evidence that γ2 is required for efficient Nef-mediated targeting of CD4 to lysosomes was obtained by immunofluorescence microscopy analysis. As expected, Nef caused a dramatic redistribution of CD4 from the plasma membrane (Fig. 3A–C) to cytoplasmic vesicular structures, which were mostly localized at a juxtanuclear region (Fig. 3D–F). Although γ2 presented a more dispersed localization in control-transfected cells (Fig. 3A–C), in cells expressing Nef some of the CD4-positive structures also contained γ2 (Fig. 3D–G; Table S2). In contrast, the pattern of CD4 redistribution mediated by Nef was severely modified by the depletion of γ2, and CD4 was detected at puncta distributed throughout the cytoplasm (Fig. 3H–M). A similar effect was observed when the γ2 siRNA #2 was used to deplete γ2 (Fig. S4). In γ2-depleted cells, expression of Nef led to an increase (from ∼27% to ∼70%) in colocalization of CD4 with EEA1, a peripheral membrane protein recruited to early endosomes (Fig. 4A–F,M; Table S2), suggesting that CD4 accumulates in pre-lysosomal compartments. Moreover, EEA1-positive endosomes appeared slightly enlarged and more dispersed in γ2 KD cells.
Next, we asked whether this accumulation in early endosomes could result in a reduced detection of CD4 in lysosomes of Nef cells. To this end, we treated cells with the fluorescent dye LysoTracker, a lysosomal marker that specifically accumulates within acidic compartments (Lemieux et al., 2004). Acidic vesicles are clearly detected in control and γ2 KD cells (Fig. 4G–L), indicating that γ2 depletion does not block vesicular acidification. Although, downregulated CD4 was occasionally detected in the periphery of these acidic vesicles in γ2 KD cells (Fig. 4J–L, see inset), we were unable to observe considerable differences in the amount of CD4 staining that overlaped with LysoTracker in γ2 KD cells expressing Nef, when compared to cells treated with control siRNA (Fig. 4M; Table S2). It is possible that proteolytic degradation prevents accumulation of CD4 within lysosomes and undermines detection by immunofluorescence microscopy (daSilva et al., 2009). Taken together, these results further support the notion that γ2 is required for efficient downregulation of CD4 by Nef.
γ2 is part of an AP-1 complex variant involved in Nef-mediated targeting of CD4 to lysosomes
The differential effects of γ1 and γ2 depletion on CD4 downregulation by Nef prompted the investigation of whether the role of γ2 in Nef-mediated CD4 trafficking is dependent on another AP-1 complex subunit. To this end, we reduced the expression of the μ1A subunit of AP-1 in HeLa cells with siRNA (a 91.1±6.2% depletion was achieved; mean±s.e.m., n=4; Fig. 5A). Interestingly, depletion of μ1A led to a strong reduction in the cellular levels of both γ1 and γ2 (77.9±4.7% and 71.9±2.3% depletion of γ1 and γ2, respectively; Fig. 5A). This suggests that depletion of μ1A compromises the stability of both γ subunit isoforms, and that they might mostly occur as complexes with μ1A in cells. Next, we carried out analyses, in μ1A KD cells, of either the total levels of CD4, as assessed by immunoblotting (Fig. 5A,B), or the cell surface levels of CD4, as assessed by FACS (Fig. 5C,D). Depletion of μ1A caused a partial block in the Nef-mediated reduction of total levels of CD4 (Fig. 5A,B). However, this inhibitory effect was not as strong as that observed upon γ2 depletion (compare with Fig. 2A,B), perhaps due to pleiotropic effects resulting from indirect co-depletion of γ1 (see Discussion). In contrast, μ1A depletion did not alter the reduction in the surface levels of CD4 mediated by Nef (Fig. 5C,D). Nonetheless, these results support the notion that γ2 is expressed as a complex with μ1A (Mattera et al., 2011), and that μ1A expression is also required for efficient degradation of CD4 induced by Nef. To investigate whether interaction of Nef with the AP-1 complex during CD4 downregulation occurs in subcellular sites where γ2 is present, we used bimolecular fluorescence complementation (BiFC). This technique has been broadly used to study Nef dimerization (Poe and Smithgall, 2009; Poe et al., 2014) and the interaction of Nef with host cell proteins (Amorim et al., 2014; Dikeakos et al., 2012; Dirk et al., 2015). To this end, we generated constructs encoding Nef and μ1A fused to the N- or C-terminal halves of the Venus fluorescent protein, respectively (Nef–VNt and μ1A–VCt). Importantly, it has been demonstrated that μ1A C-terminally tagged with epitopes or fluorescent proteins, assembles into the AP-1 complex and displays the expected subcellular localization (Guo et al., 2013).
HeLa cells co-transfected with CD4, Nef–VNt and μ1A–VCt plasmids showed a punctate pattern of Venus fluorescence distribution, mostly localized in the juxtanuclear region (Fig. 5E). Approximately 57% of the γ2 labeling overlapped with the BiFC signal (Fig. 5E–K; Table S2), and ∼73% of the structures that presented signal for both BiFC and CD4 also stained for γ2 (Fig. 5L). Moreover, no Venus fluorescence was detected when the Nef-LL/AA–VNt and μ1A–VCt constructs were co-transfected (result not shown), implying that BiFC was dependent on the ENTSLL165 motif. Taken together, the results indicate that a variant of AP-1 comprising γ2, but not γ1, is required for efficient targeting of CD4 to lysosomes by Nef.
Impairment of the ESCRT pathway leads to colocalization of γ2 and CD4 in enlarged endosomes upon Nef expression
To gain insights into the role of γ2 in the downregulation of CD4 by Nef, we sought to determine the subcellular localization of γ2. Initially, we directly compared the subcellular distribution of endogenous γ1 and γ2. Immunofluorescence microscopy showed that although a considerable fraction (∼34%) of γ2 colocalized with γ1, most of the γ2-positive structures were clearly separated from the γ1 labeling (Fig. 6A–D; Table S2). This differential pattern suggests that γ1 and γ2 are recruited to distinct subcellular structures, and that they might mediate protein transport from different locations. Further colocalization analyses revealed a partial colocalization of γ2 with endosomal markers. Approximately 26% of γ2 labeling localized to EEA1-positive structures (Fig. 6E–G,D; Table S2). A greater proportion of γ2 staining (∼42%) colocalized with transferrin receptor (TfR) (Fig. 6H–J,D; Table S2), a transmembrane protein that cycles between the plasma membrane and the endosomal system, and at steady state, is mostly localized in early and recycling endosomes (Grant and Donaldson, 2009). Moreover, ∼17% of the γ2 signal showed colocalization with that of CD63, a late endosome and lysosome protein (Fig. 6K–M,D; Table S2). In contrast, less than 0.05% of the γ2 signal overlapped with that of cytochrome c, a mitochondrial protein, serving as reference for absence of colocalization (Fig. 6N–P,D; Table S2). These findings indicate that a portion of γ2 is recruited to membranes of endosomal compartments. In fact, it has been shown that γ2 interacts with components of the ESCRT machinery and might play a role in protein sorting to late endosomes (Döring et al., 2010). Because Nef-mediated degradation of CD4 is ESCRT-dependent and requires CD4 targeting to the MVB pathway (daSilva et al., 2009), we hypothesized that γ2 could facilitate such targeting. To investigate this possibility, we overexpressed GFP-tagged HRS (also known as HGS), which is a component of the ESCRT-0 complex that initiates recruitment of the ESCRT machinery in early endosomes (Williams and Urbé, 2007). Overexpression of HRS negatively perturbs the ESCRT pathway, promoting accumulation of ESCRT-I in early endosomes that become enlarged (Bache et al., 2003; Bishop et al., 2002). Overexpression of cytosolic GFP did not seem to alter the distribution of γ2 and CD4 downregulated by Nef (Fig. 7A–D). However, in HRS–GFP-expressing cells, CD4 internalized by Nef was detected in enlarged endosomes (Fig. 7E–L). Under these conditions, a large proportion of γ2 signal (∼75%) was detected in structures that were positive for both CD4 and HRS–GFP (Fig. 7M). This result provides additional support to the notion that γ2 could be recruited to endosomal compartments to facilitate the efficient targeting of CD4 to the MVB pathway by Nef.
Requirement of γ2 for receptor downregulation through the canonical MVB pathway
The results shown above indicated a requirement for γ2 in Nef-mediated targeting of CD4 to lysosomes. We have previously shown that this process involves ESCRT-dependent targeting of CD4 to the MVB pathway (Amorim et al., 2014; daSilva et al., 2009). Signaling receptors such as the EGF receptor (EGFR) are also targeted to the MVB pathway after they undergo signal-mediated endocytosis (Felder et al., 1990). To explore whether γ2 has a role in this receptor downregulation pathway, we performed uptake experiments with fluorescent EGF (Alexa-Fluor-488–EGF). This approach allows monitoring the kinetics of lysosomal delivery of internalized EGF. HeLa cells treated with control siRNA or siRNA for γ1 or γ2, were serum starved and incubated with Alexa-Fluor-488–EGF for 5 min and chased for 30 or 120 min at 37°C. In control cells (Fig. 8A,C,E) and also in γ1-depleted cells (Fig. 8B,E), Alexa-Fluor-488–EGF labeling was drastically reduced upon a 120-min incubation, indicating delivery to an acidic compartment, in which EGF is degraded. Notably, in γ2-depleted cells a substantial amount of fluorescent EGF remained detectable after 120 min of incubation (Fig. 8D,E). This data indicates that γ2 is required for efficient degradation of internalized EGF, a process that normally depends on delivery of EGFR to MVBs. To further confirm the participation of γ2 in this pathway, we assayed for EGF-induced degradation of EGFR in control and γ2 KD HeLa cells. We found that γ2-depletion resulted in a significant reduction in the kinetics of EGFR degradation upon EGF stimulation (Fig. 8F,G). Taken together, the results suggest that γ2 expression is required for efficient delivery of EGFR–EGF complexes to the MBV pathway, and could also facilitate targeting of CD4 to this pathway by Nef. In this model, Nef interacts with AP-1 through the γ2–σ1 hemi-complex to accelerate targeting of CD4 to lysosomes, in the same manner as it interacts with the α–σ2 hemi-complex of AP-2 to promote CD4 endocytosis.
Nef interacts with AP-1 through either the γ1–σ1 or the γ2–σ1 hemi-complexes (Mattera et al., 2011; Fig. 1). However, prior to our study, the functional relevance of Nef–AP-1 interaction in the downregulation of CD4 was unknown. In the present study, we confirmed the interaction of Nef with γ2–σ1, and show that this binding requires a conserved D/ExxxL(L/I) motif in Nef (Fig. 1). The results of this study also demonstrate that efficient CD4 downregulation by Nef requires the function of γ2, whereas γ1 is dispensable (Fig. 2; Fig. S2).
Upon Nef- and AP-2-mediated internalization, CD4 is delivered to early endosomes and finally to lysosomes for degradation. A major event in this process involves targeting of CD4 to the ILVs of MVBs (daSilva et al., 2009; Schaefer et al., 2008) in a CD4-ubiquitylation-independent manner, which depends on the ESCRT machinery (daSilva et al., 2009) and is mediated by the direct interaction of Nef with Alix (Amorim et al., 2014). Although KD of Alix or Tsg101 (ESCRT-I) blocks CD4 degradation induced by Nef, it does not alleviate depletion of cell surface CD4 (Amorim et al., 2014; daSilva et al., 2009). This suggests that Nef- and Alix-mediated targeting of CD4 to ILVs occurs after CD4 molecules are already segregated away from endosomal domains that are engaged in recycling to the plasma membrane. Our findings that γ2 KD leads to the accumulation of CD4 in early endosomes and impairs cell surface depletion of CD4 mediated by Nef (Figs 2 and 4), indicate that γ2 activity is required in a step prior to sorting of CD4 into the degradative MVB pathway. Consistent with the role of γ2 in targeting internalized CD4 to late endosomes, most of the endogenous γ2 staining is located in early and recycling endosomes (Fig. 6; Table S2), and CD4 mostly colocalizes with γ2 in cells expressing Nef (Fig. 3). Moreover, upon artificially stabilizing ESCRT-I in early endosomes by overexpression of HRS (Bache et al., 2003; Bishop et al., 2002), internalized CD4 accumulates in enlarged HRS–GFP-positive endosomes, where colocalization with γ2 is prominent (Fig. 7).
Our results also demonstrate that depletion of γ2, but not γ1, slows down the degradation of EGF–EGFR complexes (Fig. 8), the prototype cargo of the canonical MVB pathway. Indeed, a role in endosome to lysosome maturation or trafficking has been proposed for γ2 by other authors (Döring et al., 2010; Rost et al., 2008). In particular, it has been demonstrated that γ2 associates with subunits of ESCRT-I and -III, and that γ2 depletion enhances HIV particle release, likely by inhibiting lysosome-mediated clearance of HIV Gag protein (Döring et al., 2010). We conclude that γ2 is an important factor in the targeting of endocytosed proteins to the degradative MVB pathway.
AP-1 is potentially the most diverse AP complex with two γ (γ1 and γ2), two μ1 (μ1A and μ1B) and three σ (σ1A, σ1B and σ1C) isoforms expressed in mammals (Boehm and Bonifacino, 2001), which could hypothetically give rise to twelve complexes. With the exception of μ1B, which is specifically expressed in epithelial cells (Ohno et al., 1999), all the other isoforms are ubiquitously expressed in different cell types. Both μ1A and μ1B can assemble in AP-1 complexes (Fölsch et al., 2001; Guo et al., 2013), and this enables AP-1 variants to mediate sorting of a broader range of cargos (Bonifacino, 2014; Guo et al., 2013). Our data indicate that the γ isoforms also endow AP-1 complexes with different properties. This notion is supported by previous reports showing that knockout of the γ1 gene in mice causes early embryonic lethality (Zizioli et al., 1999), and KD of γ1 is sufficient to cause missorting of specific proteins (Farías et al., 2012; Hirst et al., 2012). In addition, γ1–σ1 and γ2–σ1 hemi-complexes display different binding affinities for D/ExxxL(L/I) sorting signals, depending on the context of the signal (Mattera et al., 2011), suggesting distinct cargo recognition specificities.
Determinants within the γ and β1 subunits are thought to define the membrane localization of AP-1 (Austin et al., 2002; Heldwein et al., 2004; Ren et al., 2013; Wang et al., 2003). Therefore, we speculate that γ1 and γ2 may mediate recruitment of AP-1 variants to distinct locations within cells. Corroborating this notion, the direct comparison of endogenous γ1 and γ2 subcellular distribution has shown that these proteins are recruited mostly to distinct intracellular membrane domains (Fig. 6; Table S2). The demonstration that γ2 can interact with the other three AP-1 subunits (Mattera et al., 2011; Takatsu et al., 1998) indicates that γ2 is capable of forming an variant AP-1 complex. However, so far, no direct evidence has distinguished whether γ2 exerts its functions as a monomer or as part of an AP-1 complex. In the present study, the stability of both γ-subunit isoforms was compromised in the absence of μ1A (Fig. 5), indicating that γ1 and γ2 are mostly expressed as part of an AP-1 complex. In fact, it has been previously demonstrated that in the absence of one of the subunits, the respective AP complex becomes unstable and disassembles, resulting in decreased levels of the other subunits (DaSilva et al., 2016; Meyer et al., 2000; Motley et al., 2003; Peden et al., 2002). Moreover, as for γ1, interaction of Nef with γ2 absolutely requires co-expression of the σ1 subunit (Mattera et al., 2011; and our results, data not shown). Finally, we show that depletion of μ1A partially mimics γ2 depletion for the impairment of Nef-mediated degradation of CD4 (Figs 2 and 5). However, the inhibitory effect of μ1A KD was weaker when compared to γ2 KD (Figs 2 and 5). This weakened effect could be the consequence of co-depletion of AP-1 complexes comprising γ1, as the absence of both AP-1 variants might favor pathways mediated by other adaptor proteins that utilize common co-factors, such as AP-2. Specifically, it has been previously proposed that AP-2 depletion enhances AP-1-dependent pathways (Lubben et al., 2007). Therefore, γ2 is most likely part of an AP-1 complex variant involved in Nef-mediated targeting of CD4 to lysosomes, a notion that is further confirmed by BiFC assays showing that γ2 labeling greatly overlaps with signals for Nef interaction with AP-1 (μ1A) in CD4-positive structures (Fig. 5L–E).
In summary, our results demonstrate that efficient targeting of CD4 for degradation in lysosomes by Nef requires the Nef interaction partner γ2-adaptin. Our results also provide evidence that γ2 is a subunit of a new AP-1 variant, which functions in an endocytic pathway leading to late endosomes and lysosomes, and is functionally distinct from the canonical AP-1 complex containing the γ1 subunit. Because Nef plays a decisive role in HIV-1 pathogenesis, through the downregulation of host surface proteins (Pereira and daSilva, 2016; Sugden et al., 2016), it is important to test the role of γ2 in the downregulation of other Nef targets.
MATERIALS AND METHODS
Homology modeling and structural analysis
Molecular models of Nef bound to either the γ1–σ1A or the γ2–σ1A (AP-1) hemi-complexes were generated by homology modeling with YASARA Structure's homology modeling module (Krieger et al., 2009). As a model template, we used Nef bound to the α and σ2 subunits of AP-2 (PDB ID 4NEE) (Ren et al., 2014). The amino acid sequence of the HIV-1 isolate HXB2 Nef protein (UniProtKB ID P04601) and the human amino acid sequence of γ1 (UniProtKB ID O43747), γ2 (UniProtKB ID O75843) and σ1A (UniProtKB ID P61966) were used. The resulting refined models were analyzed on the PDBePISA server (http://www.ebi.ac.uk/pdbe/pisa/pistart.html) (Krissinel, 2010) for the identification of interfacing residues, as well as of residues forming hydrogen bonds, disulfide bonds or salt bridges. Structural figures were prepared with MacPymol (https://www.pymol.org/).
The mammalian expression vectors: pCMV.CD4, pCIneo.Nef-WT, pCIneo.Nef-LL/AA (encoding HIV-1 NL4-3 Nef-WT and the dileucine mutant) and the pIRES2.eGFP based vectors (encoding HIV-1 NL4-3, NA7, 248 or SIVmac239 Nef alleles) were as described by Chaudhuri et al. (2007). For pHRS-EGFP, the HRS cDNA was amplified from pCIneo3xHA-HRS and cloned into a pEGFP-C vector (Clontech, Palo Alto, CA). pGADT7 (Clontech) encoding either human γ2 or mouse γ1, and the pBrigde (Clontech) encoding NL4-3 Nef and rat σ1A (pBridge.Nef-Wt.σ1A), were as described previously (Mattera et al., 2011). The Nef-WT cDNA in pBridge.Nef-Wt.σ1A was replaced with the Nef-LL/AA cDNA from pCIneo.Nef-LL/AA to generate the pBridge.Nef-LL/AA.σ1A plasmid. For BiFC experiments, we used the pNef-VNt plasmid encoding NL4-3 Nef fused to the N-terminal half of Venus protein (VNt, residues 1 to 158) (Amorim et al., 2014); and the pμ1A-VCt plasmid, generated as follows: the mouse μ1A open reading frame (ORF) was amplified from pCIneoμ1A-HA (Guo et al., 2013) and used to replace the leucine zipper (LZ) sequence in the pcDNA3.1/Zeo-LZ-VCt (MacDonald et al., 2006). This resulted in the μ1A C-terminal appendage of a spacer sequence (GGGGSGGGGSSG), followed by the VCt sequence (residues 159 to 239 of Venus protein).
Yeast three-hybrid analyses
Y3H assays were performed as previously described (Mattera et al., 2011). Briefly, Saccharomyces cerevisiae cells (HF7c strain) were transformed with pairs of pBridge and pGADT7 vectors. Double transformants were selected on dropout agar plates lacking Leu, Trp, and Met, but containing His (+His). For colony growth assays, pooled colonies from each transformation were resuspended in water, counted using an hemocytometer, and tenfold serial dilutions from 104 cells/5 µl were prepared, and 5 µl of each cell suspension were dotted on: (1) +His plates (loading control); and (2) plates lacking leucine, tryptophan, methionine and histidine (–His/–Met), with or without 0.2 mM 3-amino-1,2,4-triazole (3-AT), a competitive inhibitor of the His3 protein.
Cell culture, transfections and RNA interference
HeLa CCL-2 cells (American Type Culture Collection, Manassas, VA), grown as previously described (Amorim et al., 2014), were transiently transfected with the indicated plasmids using Lipofectamine 2000 (Thermo Scientific). The siRNA were purchased from Dharmacon (Lafayette, CO) as nucleotide duplexes with 3′dTdT overhangs, designed to target human γ1 (5′-GGAAGAGCCUAUUCAGGUA-3′), γ2 (5′-AAACCCUGCUUUGCUGUUAA-3′) (Rost et al., 2006) and μ1A (5′-GGCAUCAAGUAUCGGAAGA-3′) (Janvier and Bonifacino, 2005). An alternative siRNA targeting γ2 (γ2 siRNA#2; 5′-CACAGAACACAGCAUAUCU-3′), with a 3′dTdT overhang, was purchased from Bioneer Inc (Alameda, CA). The MISSION siRNA Universal Negative Control (SIC001, Sigma-Aldrich) was used in control experiments. HeLa cells were subjected to two rounds of siRNA transfection with a 48 h interval, using the Oligofectamine reagent (Thermo Scientific). At 4 days after the first siRNA transfection, cells were transfected with DNA plasmids and analyzed on day 5.
Monoclonal antibodies to human CD4 (S3.5; dilution 1:100), used for immunofluorescence and FACS analysis, were from Thermo Scientific. Monoclonal antibodies to β-actin (C4; sc-47778, Santa Cruz Biotechnology) and γ1 (610386; clone 88, adaptin-γ; BD Biosciences, San Jose, CA), at 1:1000 dilution, and rabbit polyclonal antibodies to human CD4 (H370; Santa Cruz Biotechnology, CA), γ2-adaptin (HPA004106; Sigma-Aldrich), μ1A-adaptin (AB111135; Abcam, Cambridge, MA), cathepsin D (cat. no. 219361; Calbiochem, San Diego, CA), at 1:1000 dilution, and EGFR (1:5000, AB52894; Abcam), were used for immunoblot experiments. The rabbit antiserum to HIV-1 Nef, used for immunoblot experiments (1:2000 dilution), was obtained from the NIH AIDS Research and Reference Reagent Program (Shugars et al., 1993). Mouse monoclonal antibodies to human EEA1 (14/EEA1), CD63 (H5C6), γ1-adaptin (610386; clone 88, BD Biosciences), transferrin receptor (H68.4, Thermo Scientific) and cytochrome c (SC-13560, Santa Cruz Biotechnology); and rabbit polyclonal antibody to γ2-adaptin, were used for immunofluorescence experiments at 1:100 dilution. Horseradish-peroxidase-conjugated donkey anti-mouse immunoglobulin G (IgG) and donkey anti-rabbit IgG were obtained from GE Healthcare. Secondary antibodies conjugated to Alexa fluorophores were purchased from Thermo Scientific.
To assay for surface CD4, unfixed cells were incubated at 4°C with APC (allophycocyanin)-conjugated anti-CD4 and prepared for FACS analysis. Corresponding isotype control IgGs, conjugated to APC (Thermo Scientific) were used to control for non-specific antibody labeling. FACS data were acquired using the levels of APC fluorescence in cells expressing GFP from a FACSDiva flow cytometer (BD Biosciences). We used the FloJo software (Tree Star, Ashland, OR) for data analyses.
LysoTracker staining and immunofluorescence microscopy
To label acid vesicles, HeLa cells grown overnight on glass coverslips were rinsed in phosphate-buffered saline (PBS) and incubated with 75 nM of LysoTracker DND-99 (Life Technologies, Molecular Probes) in Dulbecco's modified Eagle's medium (DMEM) for 30 min, as previously described (Brito de Souza et al., 2014). For immunofluorescence, cells were then fixed for 15 min at room temperature with 4% (w/v) paraformaldehyde (PFA) in PBS. PFA-fixed cells were permeabilized with 0.01% (w/v) saponin in blocking solution [0.2% (w/v) pork skin gelatin in PBS] for 15 min at 37°C, and labeled as previously described. Cells were imaged on a Zeiss confocal laser-scanning microscope (LSM) 780 (Zeiss, Jena, Germany). Post-acquisition image processing and colocalization analysis were as previously described (Amorim et al., 2014).
SDS-PAGE and immunoblot analysis
Total cell lysates were prepared and equalized for total protein concentration, as previously described (Amorim et al., 2014). Protein homogenates were mixed with sample buffer [SDS 4%, Tris-HCl 160 mM (pH 6.8), glycerol 20%, DTT 100 mM and 0.005% Bromphenol Blue] and boiled. The proteins were resolved by SDS-PAGE under reducing conditions, and electro-transferred to nitrocellulose membranes (Millipore, Bedford, MA). Blots were probed with primary antibodies and appropriate HRP-conjugated secondary antibodies. Proteins in the blots were visualized by enhanced chemiluminescence using ECL (GE Healthcare).
Metabolic labeling and immunoprecipitation of cathepsin D
Metabolic labeling and immunoprecipitation were performed as previously described (Bonifacino, 2016) with minor changes. In brief, HeLa cells were treated with siRNA as described above. Cells were rinsed with PBS and incubated in pre-label medium [DMEM, –Cys –Met (Sigma-Aldrich)] for 30 min at 37°C. Cells were pulse-labeled for 20 min at 37°C using 0.1 mCi/ml [35S]methionine-cysteine (Express Protein Label; Perkin Elmer-Cetus, Boston, MA), rinsed and chased for 5 h at 37°C in regular culture medium supplemented with excess methionine (0.06 mg/ml) and cysteine (0.1 mg/ml). After 5 h, the chase medium was collected and cells were rinsed twice with ice-cold PBS and lysed in 50 mM Tris-HCl (pH 7.4), 300 mM NaCl, 5 mM EDTA), 0.5% (v/v) Triton X-100, supplemented with protease inhibitor cocktail (Sigma-Aldrich). Both cell extracts and media were used to immunoprecipitate cathepsin D, which was analyzed by SDS-PAGE and fluorography using a Pharos FX Plus Molecular Imager (Biorad).
EGF uptake and degradation assay
For immunofluorescence experiments, cells were incubated with Opti-MEM culture medium for 1 h at 37°C. Cells were then stimulated with Alexa-Fluor-488–EGF (Thermo Scientific) (100 ng/ml in Opti-MEM) for 5 min at 37°C, washed with complete DMEM, followed by Alexa-Fluor-488–EGF chasing for 30 or 120 min at 37°C. The cells were fixed after each time point with 4% (w/v) PFA in PBS for 15 min at room temperature. For immunoblot analysis of EGFR degradation, starved cells were pre-incubated for 30 min with 20 μg/ml cycloheximide (Sigma-Aldrich) in Opti-MEM. The cells were then incubated with 50 ng/ml of unlabeled EGF (Thermo Scientific) in Opti-MEM, supplemented with cycloheximide, at 4°C for 30 min. Cells were washed twice with ice-cold PBS, and further incubated with 0.2 mg/ml of EZ-Link Sulfo-NHS-LC-Biotin (Thermo Scientific) in ice-cold PBS for 30 min at 4°C, washed and further incubated with 50 mM NH4Cl for 15 min. Cells were incubated with complete DMEM at 37°C for the times indicated in Fig. 8. Cell lysates were centrifuged at 20,000 g for 10 min at 4°C, and equal amounts of total protein in supernatants were incubated with pre-washed 30 μl of Neutravidin Plus UltraLink Resin (Thermo Scientific) for 3 h on a wheel at 4°C. Following washing with ice-cold lysis buffer, proteins were eluted from the beads with sample buffer and analyzed by immunoblotting.
Data were plotted and analyzed using GraphPad Prism 5.0 software. For colocalization analyses, statistical significance was calculated by using a two-tailed paired t-test. One-way ANOVA, followed by a post-hoc multiple Bonferroni's test was used for all the other statistical significance calculations. The P values are as labeled as follows: *P<0.05; **P<0.005; ***P<0.0005; ns, not significant. Differences were considered statistically significant at P<0.05.
We thank J. Bonifacino, K. Strebel, J. Miller and the NIH AIDS Research and Reference Reagent Program for the kind donation of reagents; A.N. de Carvalho, R.R.C. Rosales, D. Ferraz and M.H.S. Martins for excellent technical assistance. We are grateful to R.O. de Castro for assistance with constructs.
L.A.T., E.M.L.d.S and L.L.P.d. designed the research; L.A.T., E.M.L.d.S., M.E.d.S.-J., Y.C.J. and J.V.d.C. performed the experiments; G.A.M. performed tridimensional structure analysis; L.A.T., E.M.L.d.S., M.E.d.S.-J., Y.C.J., G.A.M. and L.L.P.d. analyzed the data; L.A.T., E.M.L.d.S., G.A.M. and L.L.P.d. wrote the paper.
The present study was supported by the Fundação de Amparo à Pesquisa do Estado de São Paulo (São Paulo Research Foundation; FAPESP) grants (2014/25812-0 and 2014/02438-6) and Fundação de Apoio ao Ensino, Pesquisa e Assistência do Hospital das Clínicas da Faculdade de Medicina de Ribeirão Preto da Universidade de São Paulo (FAEPA) grants to L.L.P.d.; and a Dirección de Investigación, Universidad Austral de Chile (DID-UACH) grant to G.A.M. The study was also supported by FAPESP-EMU 2009/54014-7 for the Multi-user Laboratory of Multi-photon Microscopy. L.A.T. and M.E.d.S.-J. were supported with Masters fellowships; and E.M.L.d.S. and J.V.d.C. with doctoral fellowships by FAPESP. Y.C.J. was supported with a Masters fellowship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES; Brazilian Ministry of Education).
The authors declare no competing or financial interests.