Multifaceted HIV-1 Vif interactions with Human E3 Ubiquitin Ligase and APOBEC3s

Abstract

APOBEC3 (A3) proteins are a family of host antiviral restriction factors that potently inhibit various retroviral infections, including human immunodeficiency virus (HIV)-1. To overcome this restriction, HIV-1 virion infectivity factor (Vif) recruits the cellular cofactor CBFβ to assist in targeting A3 proteins to a host E3 ligase complex for poly-ubiquitination and subsequent proteasomal degradation. Intervention of the Vif-A3 interactions could be a promising therapeutic strategy to facilitate A3-mediated suppression of HIV-1 in patients. In this structural snapshot, we review the structural features of the recently determined structure of human A3F in complex with HIV-1 Vif and its cofactor CBFβ, discuss insights into the molecular principles of Vif-A3 interplay during the arms race between the virus and host, and highlight the therapeutic implications.

Introduction

During the course of HIV infection, host cells employ a wide array of restriction factors to defend against the virus at different stages of the viral life cycle. APOBEC3 (A3) family of cytidine deaminases are among the most potent to restrict the virus during the early stage of HIV-1 infection. A3 enzymes from a virus-producing cell are incorporated into HIV virions and extensively hypermutate the viral genome during the next round of infection, by converting cytosine to uracil in the minus strand of the viral single stranded DNA (ssDNA) during reverse transcription [1–3]. A3 enzymes can also inhibit viral replication through deaminase-independent mechanisms by interfering with the functions of reverse transcriptase or integrase [4–12]. Humans express seven A3 proteins (A3A-A3H) with one or two structurally similar zinc-binding domains that are categorized into Z1, Z2, or Z3 phylogenies based on sequence[13]. Among them, A3C, A3D, A3F, A3G, and A3H have been found to inhibit HIV-1 replication, with the di-domain A3G and A3F enzymes being the most potent HIV-1 restrictors [14, 15]. The di-domain A3 enzymes normally have their C-terminal domain (CTD) catalytically active for deamination and their N-terminal domain (NTD) responsible for virion packaging, nucleic acid binding, enzyme processivity and oligomerization [16]. The catalytic domains of A3s all contain a well-conserved, zinc-coordinating H-X-E-X_23–28-P-C-X_2–4-C motif at the active site [13, 17–19], which preferentially selects different nucleotide motifs surrounding the sites of deamination [1, 20–24].

HIV-1 evades the restriction imposed by A3s through an accessory protein Vif, which hijacks the host ubiquitin-proteasome pathway to polyubiquitinate A3s for degradation [25–29]. To recruit the ubiquitin ligase, HIV-1 Vif mimics the substrate receptor component of the cellular Cullin-RING E3 ligase (CRL) machinery composed of the scaffold protein Cullin 5 (Cul5), the E2-activating RING finger protein Rbx2, and the adaptor proteins Elongin B (EloB) and Elongin C (EloC) that connect the substrate receptor to Cul5 (Fig. 1) [28]. This recruiting event also requires a non-canonical cellular cofactor core-binding factor beta (CBFβ) to stabilize Vif and its assembly with the E3 ligase [30–33]. Since the restriction by A3 is so potent, Vif expression is essential for HIV-1 infection in most cell types expressing A3 proteins. Disruption of the interaction between HIV-1 Vif and A3s would significantly impair HIV-1 infection. Therefore, the HIV-1 Vif-A3 interaction is an attractive target for potential therapeutic development.

Fig. 1. Models of Vif/CBFβ/A3F assemblies with or without CRL5.

(A) Superposition of the crystal structure of rA3G (PDB 6P40, blue) onto the cryo-EM structure of Vif/CBFβ/A3F_CTDternary complex (circled), where rA3G NTD is situated at a position similar to that of a neighboring A3F_CTD (denoted by A3F_CTD*, light red) observed in the cryo-EM structure, but with modest clashes with Vif. (B) Proposed model of Vif/CBFβ/A3F in complex with unneddylated CRL5 by overlaying the Vif/CBFβ/A3F_CTD ternary complex with Vif/CBFβ/EloB/EloC/Cul5_NTD(PDB 4N9F), Cul5_CTD/Rbx2 (PDB 3DPL), and Cul1 FL/Rbx1 (PDB 1LDK). (C) Putative model of Vif/CBFβ/A3F in complex with the neddylated CRL5 with E2 and ubiquitin (Ub) by further overlaying the neddylated CRL1 structure (PDB 6TTU). The corresponding schematics of the models are shown in the lower panels of (B) and (C). The structural figures were generated using PyMol [103].

Vif interactions with host cofactors and E3 ubiquitin ligase machinery

Lentiviral Vifs recruit various host cofactors

Vif proteins from different lentiviruses have a conserved function of recruiting CRL complexes from the cognate hosts to initiate A3 degradation, but do so with different non-canonical cofactors [34]. For instance, primate lentiviral Vif proteins recruit CBFβ, which endogenously forms a heterodimer with the transcription factor RUNX1 to control expression of genes implicated in many cellular processes, including the immune system [35]. However, CBFβ is dispensable for non-primate Vifs. Bovine immunodeficiency virus (BIV) Vif assembles a host CRL complex without a cofactor, while maedi-visna virus (MVV) Vif requires a different cofactor Cyclophilin A (CypA), a prolyl isomerase that normally plays key roles in regulating immune responses in cells, as well as in multiple aspects of the HIV-1 life cycle [36]. Both CBFβ and CypA exist and are nearly identical in primates and sheep, so it is intriguing that HIV-1 and MVV Vif proteins have evolved to select host cofactors differentially. As the host CRL complexes are conserved among different species, it is possible that the divergent selection of host cofactors is primarily driven by the differences in Vif proteins themselves and the A3 proteins they target.

Structural features of primate lentiviral Vif-containing CRL5 complexes

To date, the crystal structures of two primate lentiviral (HIV-1 and simian immunodeficiency virus (SIV_rcm)) Vif proteins in complex with various CRL5 assemblies have been determined, providing mechanistic insights into how primate lentiviral Vifs, facilitate by CBFβ, hijacks the host CRL5 complex [37, 38]. Both HIV-1 and SIV_rcm Vif proteins adopt a similar di-domain architecture, with a smaller α domain containing two loosely-packed α helices and a larger α/β domain composed of an antiparallel β sheet tightly packed with three α helices. One helix of the α domain interacts with Cul5, and the other helix binds EloC through a conserved BC-box mimicking that of the natural substrate receptors SOCS-box proteins of CRL5 [39]. An extensive interaction between CBFβ and the antiparallel β sheet of the Vif α/β domain shields a large hydrophobic interface on Vif from solvent and stabilizes Vif and its assembly with CRL5. While both Vif and RUNX1 bind to CBFβ at the similar site, Vif buries a larger surface area, indicating a tighter binding affinity of Vif to CBFβ [37]. This sequestration of CBFβ by Vif also disrupts the interactions of CBFβ with RUNX transcription factors that regulate genes involved in various immune responses, including A3 proteins [32, 40, 41], which is potentially beneficial for the viruses. Although the overall structures are very similar, detailed conformational differences are still observed in the two primate lentiviral Vif-associated CRL5 complexes [37, 38]. These differences include variations of surface charge characters and the lengths of secondary structural elements at the interfaces, leading to modest changes in the packing between the interacting partners, all of which potentially results in the modulation of recognition and degradation of A3 proteins in distinct species.

Vif-mediated CRL-catalyzed ubiquitination

The CRL-dependent ubiquitination is specifically activated by the ubiquitin-like protein, NEDD8 [42], which forms an isopeptide bond with a specific lysine residue in the C-terminal WHB domain of Cullins [43]. The molecular mechanisms of the CRL-catalyzed ubiquitination and NEDD8 activation remains unclear until the recently determined cryo-EM structure capturing substrate ubiquitination by the neddylated CRL1 in complex with a ubiquitin-conjugating E2 enzyme [44]. Neddylation of Cul1 allows the complex to adapt a conformation that is amenable to the ubiquitination of the substrate. NEDD8 acts as a pivot to position the ubiquitin-carrying E2 and the E2-activating Rbx1 at the Cul1 C-terminus to be adjacent to the substrate receptor and the associated substrate at the Cul1 N-terminus for ubiquitination to occur. The NEDD8 activation of the CRL5-catalyzed ubiquitination is expected to work in a similar manner. Neddylation of Cul5 is required for HIV Vif-mediated degradation of A3G [45]. Additionally, a recent work showed that a member of the RING-Between-RING (RBR) E3 ligase family, ARIH2, is essential for the CRL5-based HIV infection in primary CD4⁺ T cells [46]. It is proposed that by interacting with the Vif-associated and neddylated CRL5, ARIH2 primes the initial ubiquitination of both A3F and A3G for further polyubiquitination by the E2 enzyme. ARIH2 is also found to be required for the ubiquitination of other CRL5 substrates, indicating that the priming regulation by ARIH2 may represent a universal mechanism of the CRL5-mediated ubiquitination.

The molecular mechanisms of APOBEC3 sequestration by Vif

A3 proteins are notoriously difficult to purify in vitro, because A3s usually associate tightly with nucleic acids and form higher ordered oligomers and soluble aggregates. Some soluble constructs of A3F and A3G have been engineered for biochemical and structural studies [47–53], but the yield of a wild type protein purified from a recombinant expression system is still limited. The topology of either single- or di-domain A3s are highly conserved, composed of five β strands surrounded by six α helices. The two most potent HIV-1 restrictors, A3G and A3F, are di-domain and interact with Vif through distinct domains: NTD for A3G and CTD for A3F [14, 54–56]. In addition, within the respective Vif-binding domains of A3G and A3F, separate interfaces have been mapped for Vif interaction through extensive mutagenesis studies [57, 58]. Similarly, Vif is also expected to recruit A3G and A3F through disparate, partially overlapping regions [58, 59], indicating that Vif may rely on multiple mechanisms to target A3s for degradation.

An atomic structure of the Vif/A3 complex with CRL5 has eluded the field for nearly two decades, primarily due to the poor solution behavior and flexible binding of Vif and A3s. Recently, we succeeded in obtaining a high resolution cryo-EM structure [60] focusing on the sub-complex of Vif/CBFβ with a solubility-enhanced A3F CTD (A3F_CTD) [50]. A few specific construct design elements have improved the complex assembly to help achieve atomic resolution. First, we fused A3F_CTD to CBFβ through an engineered linker to enhance the binding between Vif and A3F, which surprisingly stabilized a weak tetrameric form of the ternary complex. Moreover, we removed the α domain of Vif along with the associated C-terminal regions of CBFβ, which protrude away from the molecular core without being involved in the A3F interaction. This truncation further stabilized the tetramer formation of the ternary complex, resulting in a more rigid four-fold D2 symmetry which substantially improved the cryo-EM reconstruction.

CBFβ-A3F_CTD interactions

A major finding from our Vif/CBFβ/A3F_CTD complex structure is the direct participation of CBFβ in A3F recruitment. CBFβ is known to act as a molecular chaperone to enable initial Vif assembly with CRL5. CBFβ has also been suggested to bind Vif through a region partially overlapping with the A3F-binding interface, and therefore need to be dislodged from Vif to allow the A3F recruitment [31, 58]. The ternary complex structure reveals that CBFβ and Vif together form a wedge-like platform to accommodate A3F_CTD with little conformational change. This observed novel CBFβ-A3F_CTD interface has been further validated through biochemical and virology studies, including a charge-swapped CBFβ E54K/A3F_CTDm R293D double mutation, which rescued the ternary complex formation in vitro, and restored the Vif-mediated A3F degradation and viral infectivity in vivo. Interestingly, the observed CBFβ interface exclusively binds A3F, but not A3G, indicating CBFβ either interacts with A3G through another interface, or does not participate in A3G recruitment at all. The roles of CBFβ in the Vif-mediated degradation of other A3 proteins warrant further investigation. The interactions between CBFβ and A3s are most likely modulated or stabilized by Vif, given the fact that the association of CBFβ and A3F was not detected in the absence of Vif [60].

Vif-A3F_CTD interactions

The major Vif/CBFβ-A3F_CTD binding interface within the ternary complex buries a relatively small surface area of 1004 Ų, explaining our observation of a relatively weak binding during complex reconstitution. This Vif-A3F_CTD interface involves multiple electrostatic and hydrophobic regions of the Vif α/β domain located on the opposite side of its Cul5/EloC binding interface. In addition to the major interface, two potential minor Vif-A3F_CTD interfaces are formed at the inter-ternary complex interfaces within the tetramer. One of the minor interfaces is located at the Vif α1 helix close to the major interface, while the other one is situated in the ⁵⁵VxIPLx_4–5L⁶⁴ motif of Vif located on the opposite side. Extensive biochemical and virological studies have validated the importance of the observed major interface [14, 59, 61–66], with some data also support the interactions at the minor interfaces [59, 62, 67–70], although no evidence pointing to the presence of homo-tetramers in cells. Our in vivo A3F degradation data also supports that A3F is recruited by Vif to the CLR5 machinery primarily through interactions at the major interface. Whether the observed minor Vif-A3F interactions are involved in the Vif antagonism of A3F awaits further studies.

Degradation-dependent mechanism of Vif antagonization of APOBEC3s

The advances in the structural understanding of various subcomplexes help piece together an overall picture of the A3 degradation machinery. In addition to the Vif-containing complexes described above, recent crystal structures of a full-length rhesus A3G (rA3G_FL, with various solubility-enhanced mutations) provided the first glimpse of di-domain A3 structures with various inter-domain interactions through a flexible linker [71]. When the rA3G_FL structure is superimposed onto that of A3F_CTD in the Vif complex, the rA3G NTD spatially collides with the Vif α1 helix (Fig. 1A). It has been reported that A3F_NTD-A3G_CTD chimeric proteins retain low-affinity binding to Vif, indicating potential Vif binding motifs in A3F_NTD, as A3G_CTD does not support the binding [55]. Therefore, the Vif α1 helix may be a potential site for interacting with A3F_NTD, whose orientation may be adjusted through the inter-domain elasticity during the interaction. Aligning the Vif/CBFβ portion of the Vif/CBFβ/A3F_CTD ternary complex structure with that of the Vif-containing CRL5 crystal structure [37] also shows no spatial conflict with the overlaid rA3G_FL and Vif-interacting CRL5 components (Fig. 1B). By further superimposing the structure of neddylated CRL1 in complex with the ubiquitin-conjugating E2 [44] onto CRL5, we obtain a preliminary structural model for the Vif-mediated ubiquitination of A3F by neddylated CRL5 (Fig. 1C). In this model, there is considerable spatial distance between A3F_CTD and the E2 enzyme, whose location is likely dictated by the large substrate recruited in the particular CRL1 complex structure [44]. However, conformational elasticity has been shown for the C-terminal WHB domain of Cullins and the RING domain of Rbx1/2 [44, 72]. Furthermore, ARIH2 may facilitate the juxtaposition of A3F_CTDand E2 for priming the ubiquitination. Once the first ubiquitin is transferred onto A3F_CTD, the geometry for subsequent polyubiquitination would be substantially less restricted due to the increased length, flexibility, and accessibility of the growing ubiquitin chain. Besides, the catalytic site of A3F_CTD is located away from the major interface, exposing to solvent for substrate contact, indicating the binding of Vif/CBFβ may not directly affect the A3F deamination activity, therefore this proposed model could represent the primary degradation-dependent mechanism of Vif antagonization of A3F (Fig. 2).

Fig. 2. Schematics of the Vif-mediated CRL5-dependent polyubiquitination cycle of A3F.

Upon A3F recruitment by Vif to CRL5 E3 ligase, neddylation of CRL5 by NEDD8 triggers conformational changes of Cul5 C-terminal domain and Rbx2. Ubiquitinated ARIH2 then interacts with the neddylated CRL5 and initializes the ubiquitination reaction by transferring the first ubiquitin to A3F. ARIH2 disassociates from CRL5 after priming the ubiquitination, followed by the association of a ubiquitin-carrying E2 enzyme to further catalyze the polyubiquitination reaction by transferring additional ubiquitin molecules to A3F in repeating cycles.

Degradation-independent mechanism of Vif antagonization of APOBEC3s

Vif can also counteract A3 restrictions through degradation-independent mechanisms. Vif does not appear to directly inhibit the A3F deamination activity by targeting the enzyme to the E3 ligase complex, as the catalytic site of A3F is still accessible in the observed major interaction mode. It was found that Vif can regulate the translation of A3G mRNA to lower A3G levels with an unknown mechanism [73–75]. Furthermore, blocking the Vif-mediated ubiquitination of A3C, A3F and A3G was shown to restore their cellular levels but without reinstating A3-mediated restriction of HIV-1 infection, as the restored A3s failed to package in viral like particles (VLPs) [76]. In addition, several enzymatic studies showed that Vif can either attenuate the processivity of A3G by disrupting its scanning on viral ssDNA [77], or inhibit the deaminase activity of A3G [78] potentially by competing for its binding to the ssDNA substrate. Our in vitro deamination study also indicated that, although not affected by Vif-CRL5 complex formation at low concentration, the A3F_CTD enzymatic activity could be inhibited by the Vif complex at high concentration [60]. The inhibition is not revoked by the Vif R15E mutation disrupting the Vif/A3F_CTD/CBFβ ternary complex, implying it is not mediated by A3F recruiting to the E3 ligase, consistent with our structural model that the catalytic site of A3F_CTD is solvent-exposed in the ternary complex. Instead, the inhibition may come from the Vif interaction at one of the minor interfaces observed in our tetramer, which potentially confers weak binding but blocks the A3F_CTDactive site at high concentrations. The exact mechanisms of the degradation-independent antagonization of A3s by Vif should be further investigated in vitro and in vivo.

Structural insights for Vif interactions with different APOBEC3s

The molecular details provided by our Vif/CBFβ/A3F_CTD complex structure allow for a detailed scrutiny of the existing mutational analysis probing the Vif-A3F interface. Many of the Vif residues previously anticipated to interact with A3F, e.g. ¹¹WQVD¹⁴, ⁷⁴TGER⁷⁷, ⁸⁴GVSIEW⁸⁹, and ⁹⁶TQx5ADx2I¹⁰⁷, are either buried in the molecular core or located at the Vif-CBFβ interface [59, 64, 66, 79, 80]. Mutations of these residues likely impair proper Vif folding or its stability maintained by CBFβ, rather than directly disrupting the Vif-A3F interaction. Furthermore, some A3F residues speculated to bind Vif [64, 65, 81] are actually interacting with CBFβ, all of which result in a much smaller Vif-A3F binding interface than predicted.

The A3-Vif interaction sites can be divided into three categories based on the prototypical A3s: A3C/D/F type, A3G type, and A3H type (Fig. 3). A3C, A3D_CTD and A3F_CTD have been predicted to share a similar binding interface to interact with Vif [14, 61, 62, 82], although local discrepancies still exist based on the mapping studies (Fig. 3). A3C residues C130/E133 are important for Vif interaction while the equivalent A3F_CTD residues D313/E316 are not [83]. A residue cluster in A3F_CTD (L291/A292/R293/E324) are found to be critical for Vif interaction, while the equivalent residues in A3C are dispensable for efficient binding [61, 62, 83]. Interestingly, our Vif/CBFβ/A3F_CTD complex structure shows that this A3F_CTD cluster contacts CBFβ rather than Vif, indicating the observed CBFβ interface may not be involved in A3C recruitment. Furthermore, a positively-charged patch on Vif (R17/E171/R173), which help position the Vif C-terminal loop for A3F interactions, is less critical for A3C degradation [62]. In contrast to the modest binding variations within the A3C/D/F type, A3G_NTD employs an entirely distinct region at the opposite edge of the α-helical face of the molecule to interact with Vif, and correspondingly, Vif also uses a different interface to interact with A3G_NTD [53, 84–89] (Fig. 3). Many of the Vif residues previously proposed for both A3F and A3G binding are found to play indirect structural roles, resulting in a more divergent A3F and A3G binding interfaces in Vif. In addition to these two types of interfaces, A3H haplotype II (hapII) has been mapped to interact with Vif through a third interface located between those of A3C/D/F-like and A3G [84, 90, 91] (Fig. 3). Accordingly, Vif is postulated to contact A3H hapII through a distinct interface located on the opposite side of that for A3C/D/F and A3G, partially overlapping with the A3G-interface [90, 92, 93]. These diverse interactions highlight that HIV-1 Vif is a versatile A3 antagonist, actively defending against various A3s through multifarious mechanisms.

Fig. 3. Structural analysis of different Vif/CBFβ-A3 interfaces.

Comparison of the distinct Vif/CBFβ-A3 interfaces (marked with ovals of different colors) for A3F_CTD (bottom left), A3C (bottom middle, PDB 3VOW), A3D_CTD (bottom right, homology model built from A3F_CTD (PDB 3WUS)), A3G_NTD (top left, homology model built from rhesus A3G_NTD (PDB 5K81)) and A3H hapII (top right, PDB 6BBO). Critical residues observed are highlighted in red at the CBFβ-A3F interface and in yellow at the Vif-A3 interfaces. The zinc ion at the ssDNA-binding or catalytic site is shown as a dark gray sphere. A summary of the diverse locations of the Vif/CBFβ-A3 interfaces is illustrated at top middle, with ovals of corresponding colors. The structural figures were generated using PyMol [103].

Therapeutic implications

Disrupting the Vif-mediated neutralization of A3 restrictions has the therapeutic potential to treat HIV-1 infections. Attempts at gene silencing [94, 95] and small molecule inhibitors directly targeting Vif [96] have been reported. However, as a viral protein which evolves rapidly, it may develop resistance to designed drugs, reducing the efficiency of HIV therapy. An alternative strategy is to suppress the host factors and/or their involvements in the Vif-mediated A3 degradation. It has been found that the pharmacological inhibition of NEDD8 or knockdown of its conjugating enzyme UBE2F or Rbx2 blocks HIV infectivity and restores A3G restriction [45]. A potent HIV-1 inhibitor ZBMA-1 is identified to potentially bind EloC, interrupting recruitment of Vif to the E3 ligase [97]. The small Vif-Cul5 interface could be another promising drug target, although Vif may still counteract A3 restrictions through degradation-independent mechanisms. Conversely, the extensive Vif-CBFβ interface is less feasible to be disrupted by a small drug molecule. There are also a few small molecule inhibitors reported to reduce viral infectivity by intervening Vif-A3G interaction [98–102], with unclear mechanism due to the lack of a Vif-A3 complex structure. However, a common problem for the strategies targeting host proteins is that it may also affect their normal cellular activities leading to cytotoxicity.

The molecular details elucidated by various Vif complexes may enable the design of inhibitors targeting unique host-viral interactions with minimal interference with cellular functions. For example, the novel CBFβ-A3F binding interface discovered in our Vif/CBFβ/A3F complex structure could open up a new avenue for anti-HIV therapeutic strategies. This interface is important for Vif-mediated degradation of A3F, with no conflict with the RUNX binding to CBFβ, and likely does not have an intrinsic cellular function. Another advantage of this interface is its small size, making it a druggable target to allow the design of small molecule compounds or peptides, which may effectively disrupt the CBFβ-A3F interactions without potential cytotoxic effect. In addition, the potential A3F/G/H overlapping binding regions on Vif could also be a promising drug target to inhibit Vif interactions with multiple A3 members, reducing the chance of drug-resistance. These highlight the importance of the structural details at the Vif/A3/E3 ligase interfaces in promoting the development of next generation of anti-HIV therapies.

Abbreviations:

APOBEC3

apolipoprotein-B mRNA-editing catalytic polypeptide-like 3

HIV

human immunodeficiency virus

MVV

maedi-visna virus

SIV

simian immunodeficiency virus

BIV

Bovine immunodeficiency virus

Vif

viral infectivity factor

CBFβ

Core binding factor beta

RUNX

runt-related transcription factor

ssDNA

single stranded DNA

NTD

N-terminal domain

CTD

C-terminal domain

CRL

Cullin-RING E3 ligase

Cul

cullin

EloB

Elongin B

EloC

Elongin C

CypA

Cyclophilin A

RBR

RING-Between-RING

ARIH2

Ariadne RBR E3 Ubiquitin Protein Ligase 2