Structural perspectives on HIV‐1 Vif and APOBEC3 restriction factor interactions

Abstract

Human immunodeficiency virus (HIV) is a retroviral pathogen that targets human immune cells such as CD4⁺ T cells, macrophages, and dendritic cells. The human apolipoprotein B mRNA‐editing catalytic polypeptide 3 (APOBEC3 or A3) cytidine deaminases are a key class of intrinsic restriction factors that inhibit replication of HIV. When HIV‐1 enters the cell, the immune system responds by inducing the activation of the A3 family proteins, which convert cytosines to uracils in single‐stranded DNA replication intermediates, neutralizing the virus. HIV counteracts this intrinsic immune response by encoding a protein termed viral infectivity factor (Vif). Vif targets A3 to an E3 ubiquitin ligase complex for poly‐ubiquitination and proteasomal degradation. Vif is unique in that it can recognize and counteract multiple A3 restriction factor substrates. Structural biology studies have provided significant insights into the overall architectures and functions of Vif and A3 proteins; however, a structure of the Vif‐A3 complex has remained elusive. In this review, we summarize and reanalyze experimental data from recent structural, biochemical, and functional studies to provide key perspectives on the residues involved in Vif‐A3 protein–protein interactions.

1. INTRODUCTION

Human immunodeficiency virus (HIV) is the biological agent that causes acquired immunodeficiency syndrome (AIDS) in humans. HIV belongs to the genus Lentivirus within the family Retroviridae. On the basis of general characteristics and differences in its evolutionary origins and viral proteins, HIV is classified into two types: HIV‐11 and HIV‐2.2 HIV‐1 infections are prevalent globally and almost invariably result in AIDS‐related symptoms in patients.3 In contrast, infections with HIV‐2 occur mainly in West Africa, and infections progress slowly to AIDS.

HIV‐1 and HIV‐2 are enveloped retroviruses. Each virus particle contains two identical single‐stranded, positive‐sense RNA copies (∼10 kb). The genome consists of at least nine genes (gag, pol, env, tat, rev, vif, vpr, vpu, and nef). The gag, pol, and env genes encode structural proteins and replication enzymes of the virus. The tat and rev gene products are essential regulatory proteins involved in reverse transcription and viral mRNA nuclear export, respectively. Vif, vpr, vpu, nef, and vpx genes encode accessory proteins with critical roles in HIV pathogenesis, immune evasion, and immune modulation. In this review, we will primarily focus on the HIV accessory protein viral infectivity factor (Vif).

Vif targets a family of host immune factors known as APOBEC3 (or A3) mutator enzymes. A3 family proteins convert cytosines to uracils in single‐stranded DNA replication intermediates. Vif counteracts this intrinsic immune response by targeting A3 proteins for poly‐ubiquitination and proteasomal degradation. In this article, we provide an in‐depth review of the structural, biophysical, and biochemical data on Vif. Using this functional data, we address a current and highly debated series of questions (also previously raised by Desimmie et al.4) as to how one relatively small Vif protein can encode all the required structural features to directly bind to and neutralize multiple A3 protein targets.

2. VIF IS A KEY PROTEIN IN HIV INFECTION

HIV‐1 Vif was described in 1986 as a ∼23‐kDa, highly‐basic polypeptide expressed from a short open reading frame overlapping the 3′ terminus of the pol gene locus.5 Early functional assays erroneously classified Vif as an accessory protein as its function appeared to be unnecessary for viral infection of host cells.5, 6 About a year later, two studies were published that showed that even though Vif defects did not affect the numbers of virions produced, they did result in drastically curtailed viral infectivity.7, 8 Concurrently, the sequence of HIV‐2 became available, and a distinct, yet equivalent, vif gene was identified.9 Despite low sequence identity between HIV‐1 and HIV‐2 Vif proteins (<25%), both are encoded in similar genomic locations (between pol and env genes), have similar molar masses (∼23 kDa), and have similarly high pI values (~10).10, 11 As observed for HIV‐1, Vif defects impaired HIV‐2 infectivity.12 These data collectively established Vif as a viral infectivity factor––hence its name.

Functional Vif is critical for production of fully infective progeny virions; however, this phenotype is cell specific.13, 14, 15, 16 Virions lacking a functional vif gene (Δvif) can infect and produce progeny virions in only the permissive target cells (such as SupT1 or 293T cell lines) but fail to infect nonpermissive cells (such as HUT78 or CEM cell lines or human primary T cells). It was hypothesized that the nonpermissive cell lines harbor a host restriction factor that, in the absence of Vif, inhibits HIV replication.15

In 2002, Sheehy and colleagues identified CEM15 as the elusive host restriction factor.17 CEM15 is selectively expressed in nonpermissive T cells, efficiently counteracted by HIV virions harboring a vif gene, and when expressed exogenously in permissive cell lines recapitulates the nonpermissive phenotype.17 Due to sequence similarity to APOBEC1 (an already identified nucleic‐acid cytidine deaminase involved in apolipoprotein B metabolism) CEM15 was renamed apolipoprotein B mRNA‐editing catalytic polypeptide‐like 3 family, enzyme G (APOBEC3G or A3G).17, 18 Given the known cytidine deaminase function of A3 family members, it was speculated that the A3G member of the family restricts HIV infection by mutating its genome. In fact, functional and structural characterization of A3G soon revealed that the protein has cytidine deamination activity that results in deleterious mutations in the HIV negative‐sense DNA intermediates during reverse transcription.19, 20, 21, 22 It is now established that members of the A3 family result in the nonpermissive phenotype of T cells and are the very factors targeted and neutralized by the Vif protein (reviewed in Olson et al.23).

3. VIF IS A VIRAL E3 LIGASE SUBSTRATE RECEPTOR

HIV‐1 Vif‐mediated neutralization of A3 proteins is well‐characterized. Vif acts as a substrate receptor for A3s in a Cullin–RING E3 ubiquitin ligase complex. Cullin–RING E3 ligase complexes are the largest subfamily of RING domain ligases. These complexes target cellular proteins for ubiquitin‐mediated degradation through the 26S proteasome.24 The Vif‐adapted E3 machinery is comprised of the scaffold protein cullin5 (Cul5), elonginB‐elonginC (EloBC), and E2‐linking RING‐box protein 2 (Rbx2)25, 26 (Figure 1a). In contrast to E3 ligase complexes isolated from cells, recombinant preparation of Vif‐EloBC‐Cul5 resulted in samples with poor solubility27 as well as decreased affinities for Cul5 and EloBC.28 In 2011, core‐binding factor‐β (CBFβ) was identified as a noncanonical protein component present in the Vif‐adapted ligase complex.25, 26 CBFβ facilitates Vif biosynthesis, chaperones its folding, and promotes the assembly of the Vif‐adapted ligase complex that targets A3 substrates.29, 30

Figure 1.

Structural organization of the HIV‐1 Vif E3 ubiquitin ligase complex. (a) Schematic of the Vif E3 ubiquitin ligase complex. (b) Ribbon diagram of the crystal structure of the Vif_VCBCC complex (PDB: http://firstglance.jmol.org/fg.htm?mol=4N9F). Shown in the inset boxes are zoomed views of the Vif interfaces with (i) CBFβ, (ii) EloC, and (iii) Cul5

A crystal structure of the Vif‐Cul5‐EloB‐EloC‐CBFβ (Vif_VCBCC) pentameric complex was determined in 2014.28 This structure, which revealed structural details of Vif regions and their interplay with the protein components of the ligase complex, currently provides the only available experimental model of a complete Vif protein (Figure 1b). Two functionally distinct domains of Vif mediate its function as the structural hub of the complex: an N‐terminal A3 substrate‐binding domain (termed the α/β domain) and a C‐terminal adaptor domain (the α domain) that interfaces with the E3 ligase machinery. The Vif α/β domain forms a central five‐stranded antiparallel β‐sheet (strands β2‐β6) that is flanked on its convex side by three helices (α1, α2, and α5). The α domain contains two α‐helices and harbors the BC‐box motif and a conserved H‐(Xaa)₅‐C‐(Xaa)_17–18‐C‐(Xaa)_3–5‐H (HCCH) motif. The HCCH contains His and Cys residues that tetrahedrally coordinate a single Zn²⁺ ion with high affinity.31, 32, 33 Evidence suggests that Vif interacts with four components of the ligase complex through various interfaces, as discussed in the sections below.

3.1. CBFβ interface

CBFβ forms hydrophobic interactions with Vif that occlude a surface area of 2,242 Å2 on Vif. The Vif β1 strand (Residues 5–12) forms a hybrid antiparallel β‐sheet with the CBFβ β3 strand (residue 64–70; Figure 1b). Mutagenesis studies demonstrated that Vif lacking Residues 1–7 does not interact with CBFβ (<5% wild‐type levels), suggesting the Vif β1 strand is critical for CBFβ engagement.34 A patch of highly charged C‐terminal CBFβ residues (¹⁴⁸ RRTREFEDRD ¹⁵⁶) is in the vicinity of the Vif α domain and HCCH motif. Despite the predominantly charged nature of these CBFβ residues (8 out of 10 residues are basic or acidic), the interface with Vif does not appear to be stabilized by ionic interactions between the two proteins. It was previously demonstrated that CBFβ_1‐140, which lacks this charged region, interacts with Vif with affinity similar to full‐length CBFβ.35 Further, studies using recombinant proteins showed that CBFβ_1‐140 efficiently binds Vif in pull‐down and size‐exclusion chromatography experiments,34 and it supports Cul5 and EloBC binding by Vif.35 In light of these experiments, it has been concluded that the contacts between CBFβ Residues 148–156 and Vif are not critical. The CBFβ‐Vif interaction likely requires only the formation of the β1_Vif‐β3_CBFβ hybrid β‐sheet (Figure 1b, inset i).

3.2. EloC interface

Vif interacts with EloC through its BC‐box motif (Residues 145–155) located in the α4. The Vif BC‐box motif was identified through sequence analyses and is similar in structure and function to BC‐box motifs in SOCS‐box and VHL proteins.36 The Vif BC‐box adopts a helical conformation in the crystal structure of the Vif_VCBCC complex28 and contains a conserved ¹⁴⁵SLQ¹⁴⁷ motif unique to lentiviral Vif BC‐box sequences.37 The Vif BC‐box motif (Residues 145–155) and α3 helix (Residues 119–125) contact the EloC α3 helix (Residues 66–84) and α4 helix (Residues 99–111; Figure 1b, inset ii).

3.3. EloB interface

The interaction of Vif with EloB is not as well characterized as that with EloC. In the Vif_VCBCC complex crystal structure there are no interactions between the EloB and Vif polypeptide chains (Figure 1b). However, multiple reports have shown the importance of Vif‐EloB interactions for Vif function.38, 39 In a series of biophysical and cell‐based experiments, Bergeron et al. demonstrated that EloB C‐terminal Residues 101–104 and the Vif PPLP motif (Residues 161–164) mediate interactions that are critical to A3G degradation.39 Specifically, the disruption of the PPLP motif by deletion or site‐directed mutagenesis is sufficient to abrogate Vif function, whereas the deletion of 16 residues C‐terminal to the PPLP motif is inconsequential. NMR titration experiments revealed a PPLP‐specific effect on observed ¹⁵N‐HSQC peak intensity values for EloB C‐terminal residues D101, V102, M103, and K104, suggesting a direct interaction between these residues of EloB and Vif.39 Additionally, an NMR solution structure of the Vif_139‐176‐EloBC complex revealed that the EloB C‐terminal Residues 102–106 and the PPLP region in Vif are in the vicinity of each other.40 Curiously, the Vif_VCBCC complex crystal structure28 shows a different PPLP conformation. In the crystal structure, the PPLP motif is distant from the EloB C‐terminal residues.

3.4. Cul5 interface

Vif interacts with Cul5 through the HCCH motif in the α domain. The α3 helix within the HCCH motif (Residues R121, I124, L125, and R127) mediates direct interactions with Cul5 (Figure 1b, inset iii). Mechanistic studies have revealed that, similar to zinc‐binding regions of zinc‐finger proteins, the HCCH motif undergoes a conformational change upon Zn²⁺ coordination33 to become competent for Cul5 binding.31, 41 In the Vif_VCBCC complex crystal structure, the loops between the α2‐α3 and α3‐α4 helices contain the conserved Zn²⁺ coordinating histidine and cysteine residues (H108, C114, C133, and H139). The binding of Zn²⁺ likely stabilizes these loops to position the α3 helix to interact with Cul5.

In addition to the direct Vif‐Cul5 interaction, Vif binds Cul5 indirectly through EloC. The area of EloC occluded by Cul5 binding (684 Ų) is significantly more extensive than that covered on Vif upon interaction with Cul5 (182 Ų). Cell‐based mutagenesis studies of Cul5 mutants unable to bind EloC demonstrated drastically reduced binding to the Vif complex.37 In addition, in vitro studies using purified EloBC and Cul5 proteins revealed negligible binding except in the presence of a BC‐box‐containing protein.42 In isolation, EloBC or Vif minimally coordinate Cul5 but together they synergistically bind Cul5 with low nanomolar affinity.43 In this respect, Vif resembles human SOCS‐like proteins, which also interact with EloBC and Cul5.42 In summary, the Vif‐Cul5 interaction requires Zn²⁺‐induced stabilization of the HCCH motif and contributions from EloC residues.

4. VIF SUBSTRATES: A3 PROTEINS

Seven A3‐encoding genes are found in the human genome (A3A, A3B, A3C, A3D, A3F, A3G, and A3H).18 A3 enzymes catalyze the C‐to‐U deamination of single‐stranded retroviral DNA intermediates resulting in mutations that can inactivate the viral genome23 (Figure 2a). With the exceptions of A3A and A3B, A3s have various degrees of anti‐HIV‐1 activity.44, 45 Vif counteracts A3C,44, 46, 47, 48, 49 A3D,50 A3F,51, 52, 53 A3G,54, 55, 56, 57, 58 and A3H59, 60 by targeting these proteins for degradation.

Figure 2.

Structural organization of A3 proteins. (a) In A3 proteins, an H/C‐A/V‐E‐X_24‐30‐P‐C‐X_2‐4‐C motif catalyzes C‐to‐U deamination of the single‐stranded DNA. (b) A3 proteins contain either one (A3A, A3C, and A3H) or two (A3B, A3D, A3F, and A3G) cytidine deaminase domains that are classified into one of three phylogenetically distinct Z domains (Z1, Z2, and Z3). The asterisks indicate the A3 proteins that have anti‐HIV activity. (c) The crystal structure of the A3F C‐terminal domain (PDB: http://firstglance.jmol.org/fg.htm?mol=3WUS) is typical of the A3 cytidine deaminase fold. Beta‐strands are numbered from 1 to 5 and alpha‐helices from α1–α6. (d) Crystal structures of A3 proteins bound to DNA (PDB: http://firstglance.jmol.org/fg.htm?mol=5SWW) and RNA (PDB: http://firstglance.jmol.org/fg.htm?mol=6BBO). (e) List of all experimentally determined structures of unliganded and liganded A3 proteins deposited in the PDB (as of September 2019). PDB codes of available NMR or crystallographic structures are listed in brackets

The human genes encoding the A3s are all located on chromosome 22 and evolved from a primordial locus in bony fish after gene duplication events that occurred about 500 million years ago.61, 62 The primordial A3‐like locus expressed a protein product with a single cytidine deaminase domain. In humans and other primates, some of the gene duplication events resulted in genes that express protein products with two interconnected domains (i.e., A3B, A3D, A3F, and A3G loci). In all double‐domain A3s, one of the two deaminase domains is enzymatically inactive18 and likely serves nonenzymatic functions.

Each cytidine deaminase domain contains a single zinc‐coordinating deaminase motif (H/C)‐(A/V)‐E‐(X_24–30)‐(P‐C‐X_2–4‐C); the E residue is involved in the catalytic process. Each deaminase domain also contains aromatic residues involved in single‐stranded DNA substrate‐binding.18, 63 Systematic sequence analysis of human and nonhuman A3s has resulted in the classification of cytidine deaminase domains into three zinc (Z) classes, some involving C‐terminal domains (CTDs) and others N‐terminal domains (NTDs): Z1 (A3A, A3B‐CTD, and A3G‐CTD), Z2 (A3B‐NTD, A3C, A3D‐NTD, A3D‐CTD, A3F‐NTD, A3F‐CTD, and A3G‐NTD), and Z3 (A3H)62 (Figure 2b).

There are currently more than 45 structures of A3 cytidine deaminase domains in the Protein Data Bank (PDB). Structures of the Vif‐binding A3 domains from A3C,64 A3F‐CTD,65, 66, 67, 68 A3G‐NTD,69, 70 and A3H71, 72, 73, 74 are available. Inspection of these A3 structures reveals a high degree of structural similarity in overall topology (Figure 2c), congruent with evolution from a common ancestor. Nevertheless, careful investigation has revealed subtle functional, biochemical, and structural differences between the A3 proteins.23 For instance, different A3 proteins prefer distinct nucleotide contexts around their target cytosine residues: In this respect, while A3G targets the HIV genome primarily at the sites of cytosines preceded by a 5′ cytosine residue (i.e., 5′‐CC), A3B, A3C, A3D, A3F, and A3H require a 5′ thymine (i.e., 5′‐TC).75 Curiously, A3A, which has the highest catalytic activity among the A3 enzymes,76 targets cytosines within both 5′‐CC and 5′‐TC contexts.77

Since the discovery of A3 proteins in 2002, their studies have predominantly revolved around their role in mutating DNA substrates (e.g., viral DNA intermediates). However, recent research is now beginning to unravel A3‐mediated functions on RNA. It is now shown that A3‐mediated RNA editing is not only critical to viral restriction,78 but also involved in general innate immunity pathways.79 Given the versatile roles of RNAs in cellular function, RNA editing by A3 proteins will likely provide exciting new areas of inquiry. Currently, structures of A3 proteins in DNA‐ or RNA‐liganded states are described (Figure 2d). However, while much published work on unliganded and nucleic‐acid liganded A3 structures is available (summarized in Figure 2e), no report of a Vif‐bound A3 structure yet exists.

5. VIF‐A3 INTERACTION INTERFACE: MUTAGENESIS APPROACHES

Attempts to characterize the Vif‐A3 interface have relied mostly on mutagenesis‐based experiments. These experiments have identified regions involved in A3 recognition of Vif. Putative residues involved in Vif‐A3 binding are summarized in Tables 1 and 2. In the sections below, we discuss the binding sites for Vif on A3 proteins and the sites of Vif that interact with A3s.

Table 1.

Putative APOBEC3 motifs involved in Vif interactions

APOBEC3	A3 residues	References
A3C/D/F‐like
A3C	106E‐to‐H111 region 72LxxFC76, 79ILS81, 86Y,129P, 141E	Smith et al. (2010)48 Kitamura et al. (2012)64
A3D CTD	302E‐to‐H307 region 268LxxFC272, 275ILS277, 282Y, 302EF303, 307H, 337E	Smith et al. (2010)48 Kitamura et al. (2012)64
A3F CTD	289E‐to‐H294 324E 283C‐to‐T300 region 255LxxFC259, 262ILS264, 269Y 286E, 290F, 294H, 316E, 319R, 320S	Smith et al. (2010)48 Albin et al. (2010)94 Russell et al. (2009)95 Kitamura et al. (2012)64 Siu et al. (2013)65
A3G‐like
A3G NTD	128D       128DPD130 126F‐to‐Q132 region 32T, 99K, 128DP129 19Y, 26IL27, 34W, 58VY59, 124Y, 126F, 127W	Bogerd et al. (2004)96; Mangeat et al. (2004)97; Schröfelbauer et al. (2004)98; Xu et al. (2004)99 Huthoff et al. (2007)100 Russell et al. (2009)95 Lavens et al. (2010)101 Kouno et al. (2015)69
A3H‐like
A3H Haplotype II	121D/E   86S, 90W, 93VDxIK97, 100D, 129S 86SS87, 90W, 94D, 125L, 129S	Zhen et al. (2010)102; Ooms et al. (2013)103 Nakashima et al. (2017)104 Ooms et al. (2017)86

Table 2.

Putative Vif motifs involved in A3‐binding

Vif region	APOBEC3 Paralog	Vif residues	References
α/β domain	A3C	14DR15, 79W	Zhang et al. (2008)46
	A3D	14DR15, 79W	Zhang et al. (2008)46
	A3F	11WQVDRMR17 74TGERxW79 80H	Russell et al. (2007)105 He et al. (2008)106 Letko et al. (2015)84
	A3G	40YRHHY44 22K, 26K, 30Y 45E, 52S, 70W	Russell et al. (2007)105 Chen et al. (2009)107 Letko et al. (2015)84
	A3H	39F, 48H   63K 60GDAK63	Binka et al. (2012)108; Ooms et al. (2013)103   Ooms et al. (2013).109 Refsland et al. (2014)110
	A3F/A3G	21WKSLVK26 69YWxL72 a 55VxIPLx4‐5L64 a 81LGxGxxIxW89 a 96TQx(D/E)PxxADxLI107 a	Yamashita et al. (2008)111; Dang et al. (2009)112 He et al. (2008)106; Pery et al. (2009)113 He et al. (2008)106 Dang et al. (2010)114 Dang et al. (2010)115
α domain	A3F	171EDRW174	Dang et al. (2010)114
	A3G	161PPLP164	Donahue et al. (2008)116

^aBased on the crystal structure of the Vif_VCBCC complex, these motifs are either mostly inaccessible for direct interactions with A3 proteins or folded toward the core of the Vif polypeptide.

5.1. Vif binding site on A3 proteins

Mutagenesis studies have shown that residues important for Vif binding involve primarily three helices on A3 proteins: α2, α3, and α4 (Figure 3). Interestingly, structural comparisons performed by Aydin et al. indicate that different A3 proteins use different structural features to bind Vif.80

Figure 3.

Residues on A3 proteins involved in Vif binding identified in mutagenesis studies. Structures of A3C, A3D, A3F, A3G, and A3H. Vif‐binding residues identified in mutagenesis studies are indicated. The PDB models used to generate the figures were A3C (PDB: http://firstglance.jmol.org/fg.htm?mol=3VOW), A3D (homology model using 3VOW as template), A3F (PDB: http://firstglance.jmol.org/fg.htm?mol=3WUS), A3G (homology model using 5K81 as template), and A3H (PDB: http://firstglance.jmol.org/fg.htm?mol=6B0B). Based on the clustering of residues, there appears to be three distinct classes of Vif‐binding sites present in: (a) A3C/A3D/A3F, (b) A3G, and (c) A3H. Ribbon diagrams, surface and surface electrostatic potentials for all three classes are shown. A3C, A3D, and A3F appear to bind HIV‐1 Vif in a similar location involving the α2, α3, and, α4 helices. In addition, the negatively charged electrostatic properties of this region are common among A3F, A3D, and A3F. In contrast, the A3G Vif‐binding sites require residues from the β4‐α4 loop, whereas the A3H interaction with Vif involves the α3 and α4 helices. Overall, the Vif binding sites on A3G and A3H are more neutral than are the strongly negative regions in A3C, A3D, and A3F implicated in Vif binding. For clarity, the Vif‐binding surfaces are outlined on the electrostatic representations of each A3 protein. Electrostatic properties are shown from −10 to 10 kcal/e

A3C, A3D, and A3F share a common mode of Vif binding. The residues involved are located on the α2, α3 helices, and the α4 helix. Structure‐guided mutagenesis revealed 10 hydrophobic and negatively charged A3C residues that are important for Vif binding in these proteins (L72, F75, C76, I79, L80, S81, Y86, E106, F107, and H111).64 Subsequent mutagenesis studies of A3F suggested that the conserved hydrophobic residues may also play a role in protein stability and that a set of negatively charged residues (E289, E316, and E324) located on the α3 and α4 helices are also important for Vif binding.65, 81 Inspection of the electrostatic surfaces of A3C, A3D, and A3F reveal strong conservation of this negatively charged patch (Figure 3a).

Vif interactions with A3G and A3H require distinct surface features that appear to be indicative of two additional Vif‐binding classes. The negatively charged patch observed in A3C, A3D, and A3F is not conserved in these A3s, as A3G and A3H have more neutral charges at the equivalent surfaces. Moreover, mutations to residues in A3G that are equivalent to those in the Vif‐binding sites of A3C, A3D, and A3F did not alter its phenotype.64 In A3G, residues involved in Vif binding map to the α4 helix and the β4‐α4 loop (Y19, ²⁶IL²⁷, T32, W34, ⁵⁸VY⁵⁹, K99, and ¹²⁴YYFWDPD¹³⁰; Figure 3b, Table 1). In A3H, residues from the α3 and α4 helices (S86, W90, ⁹³VDFIK⁹⁷, D100, and E121) are proposed to bind Vif (Figure 3c). Taken together, these results strongly suggest that there are three classes of Vif‐binding surfaces.

Although there are three distinct A3 Vif‐binding sites, there are commonalities in the general physiochemical determinant of A3‐Vif recognition. All A3 proteins require negatively charged and hydrophobic residues (Table 1). Specifically, the presence of a negatively charged residue in the α4 helix is important for A3F, A3G, and A3H Vif binding. The presence of ionic and hydrophobic interactions is consistent with the surface properties of Vif: Vif is strongly positively charged and has hydrophobic patches.

5.2. A3‐binding site on Vif

The putative A3‐binding motifs map primarily to the N‐terminal α/β domain on Vif and appear to cluster to three distinct regions depending on the A3 paralog (Table 2, Figure 4). The A3F‐binding region appears to be on the opposite face of Vif compared to the locations of A3G and A3H binding. A3F maps to Vif residues 11–17, 74–80, and 171–174. In contrast, A3G binds to a site involving K22, K26, Y30, ⁴⁰YRHHY⁴⁴, S52, W70, and ¹⁶¹PPLP¹⁶⁴. A3H binds to yet a different site that involves F39, H48, and ⁶⁰GDAK⁶³. A single study has explored A3C and A3D binding residues on Vif.47 The residues identified (D14, R15, and W79) overlap with those that contact A3F. It is therefore likely that A3 determinants recognized by Vif belong to three distinct classes: the A3C/A3D/A3F, A3G, and the A3H class.

Figure 4.

Residues on HIV‐1 Vif involved in A3 binding identified by mutagenesis. (a) Putative A3 binding residues identified from mutagenesis studies indicated on the structure of the HIV‐1 Vif polypeptide from the Vif_VCBCC complex (PDB:http://firstglance.jmol.org/fg.htm?mol=4N9F). Molecular surface and surface electrostatic representations are also shown. All representations are in the same orientation shown in Figure 1b. Like the A3 proteins, Vif appears to display three disparate A3 binding regions exemplified by interactions with A3F, A3G, and A3H. Each A3‐binding surface identified by mutagenesis is outlined on the electrostatic representation of Vif. (b) Schematic summarizing the regions and residues on Vif identified by mutagenesis studies involved in binding A3F, A3G, and A3H

Vif A3‐binding classes are mostly distinct and only a few residues are shared between the sites. However, all Vif A3‐binding sites commonly harbor at least a single positively charged patch that is neighbored by hydrophobic features (Figure 4a, electrostatic map representation). The general biophysical properties of Vif motifs (positive and hydrophobic features) are compatible with those of the A3 motifs (negative and hydrophobic). This compatibility implies the involvement of ionic and hydrophobic forces in A3‐Vif protein–protein interactions. The ionic forces likely determine the specificity of binding to A3 substrates, and the hydrophobic forces strengthen the interaction.

Despite the general congruence between the physiochemical properties of A3 and Vif binding motifs, certain confounding observations have been made. For example, some A3F‐binding residues (present in β1 and α1) are also involved in CBFβ binding, suggesting that CBFβ may mask the A3F‐binding site on Vif.82 In support of this, there is a report that CBFβ must be displaced prior to formation of the A3F‐Vif complex.83 In addition, certain A3G‐binding residues (i.e., the PPLP motif) are known to be involved in Cul5 interactions.37

6. VIF‐A3 INTERACTION INTERFACE: HYBRID APPROACHES

Given the absence of a high‐resolution Vif‐A3 structure and the inherent limitations of mutagenesis only approaches, recent studies have employed a hybrid approach that utilizes orthogonal computational methods to propose residues of Vif that directly contact A3F, A3G, and A3H interactions.84, 85, 86 Mutagenesis experiments provided an initial subset of Vif‐A3 amino acid pairs deemed, with high confidence, to interact directly. These amino acid pairs were then used as structural constraints for the computational approach. Additional amino acid pairs were also identified in viral evolution experiments in cell culture.

The hybrid methods have identified pairs of directly interacting residues in Vif‐A3F, Vif‐A3G, and Vif‐A3H (Figure 5). In Vif‐A3G, three sets of residues are important: R15_Vif‐D128_A3G, G82_Vif‐D130_A3G, and N19_Vif‐K22_Vif‐Y125_A3G.84 For Vif‐A3F, R15_Vif‐E289_A3F, G71_Vif‐E324_A3F, and the W79_Vif‐α2/α3 pocket on A3F are critical for binding.85 In Vif‐A3H, five sets of direct interactions were identified.86 Remarkably, although the Vif‐A3F, Vif‐A3G, and Vif‐A3H sites localize to different regions on both Vif and A3 proteins, their physiochemical properties overlap. All Vif‐A3 interactions involve an electrostatic pair with a positively charged residue on Vif and a negatively charged residue on A3. Vif‐A3F and Vif‐A3G interactions involve a common arginine (R15) on Vif and a glutamate or aspartate on A3F or A3G, respectively. In addition, a glycine residue in the Vif β4‐β5 loop and a glutamate or aspartate residue on A3F or A3G, respectively, mediate direct interactions. Mutagenesis studies identified E45 on Vif as critical for A3G binding.84 The hybrid models of Vif‐A3G and Vif‐A3H interactions suggest that E45 belongs to a stretch of six amino acids ⁴⁰YRHHYE⁴⁵ that are involved in direct binding of Vif to both A3G and A3H proteins.86 Residues Y40, H42, and Y44 appear to mediate direct contacts with both A3G and A3H. A contact with R41 is unique to A3H, and H43 directly contacts A3G.

Figure 5.

Identification of Vif‐A3 interacting residues using hybrid approaches. (a) Schematic summarizing the regions and residues identified by hybrid approaches involved in Vif binding of A3F, A3G, and A3H. Lollipops represent sites of direct Vif‐A3 contacts. Amino acids labeled in black represent Vif residues. (b) Tables of interacting amino acid pairs and the corresponding molecular surfaces of HIV‐1 Vif and A3F, A3G, or A3H proteins. The book icon illustrates how the Vif‐A3 interface is split open to show the Vif and A3‐binding footprints

The divergence of structural determinants involved in various Vif‐A3 interactions is intriguing, especially since HIV‐1 and its human host have only co‐evolved for less than 100 years.87, 88, 89 Richards et al. have postulated that Vif‐A3 co‐evolution can be explained by a “wobble” model.85 Based on this model, a strong interaction between a single ancestral A3 protein and a single lentiviral Vif may have been disrupted or attenuated following zoonosis or A3 gene duplication. This interaction is restored when compensatory mutations arise at, around, or on the periphery of the ancestral Vif‐A3 interaction site. This suggests that the Vif‐A3 interface is likely not extensive and only a few critical contacts are needed to form the interaction. Given the size limitations to the lentiviral genome, a single Vif may have evolved to utilize multiple existing protein surfaces to counteract the various A3 paralogs without the need for genetic expansion. The wobble model may exemplify a common rapid evolutionary process inherent to many pathogen‐host interactions.85

7. VIF‐A3 INTERACTION INTERFACE: NEW PERSPECTIVES

The Vif‐A3F, Vif‐A3G, and Vif‐A3H hybrid models greatly enhanced our structural understanding of Vif‐A3 recognition. These studies provided an elegant compromise necessary due to as yet unmet challenge of preparing homogeneous recombinant Vif‐A3 protein complexes. In this section, we analyze Vif residues identified as important for A3 binding by mutagenesis and hybrid methods and compare their evolutionary rates to provide additional insights into Vif‐A3 interaction dynamics.

7.1. Vif residues involved in A3 binding have different patterns of evolution

HIV has a high mutation rate in vivo (∼4 × 10⁻³ per base per cell).90 This has enabled the virus to rapidly evolve into genetically diverse populations. This genetic diversity supports the evolution of the virus, which is under constant selective pressures from endogenous (i.e., host innate and adaptive immunity responses) and exogenous (e.g., therapeutic interventions) selective forces.

The wobble model proposed by Richards et al.85 provides an evolutionary paradigm that explains the biochemical observation that Vif‐A3 interactions map to multiple and mostly distinct surfaces on both Vif and A3 proteins. Although a step‐wise evolution of the Vif‐A3 interaction interface is proposed by this model, it is not clear what the ancestral A3 target looked like nor which Vif amino acids are the ancestral A3‐binding determinants. The wobble model proposes that some of the Vif amino acid positions had to vary in order to better bind new A3 substrates.

To evaluate the variability of Vif residues involved in A3‐binding, we mapped the evolutionary rates of each amino acid position along the Vif polypeptide. Evolutionary rates were calculated on the ConSurf2016 server91 based on about 3,500 HIV‐1 Vif sequences. This method uses empirical Bayesian inference to provide position‐specific evolutionary rates that reflect the variability of amino acid sites in proteins.92, 93 The estimated rates take into account the phylogenetic relationships between the sequences aligned and the stochastic natures of amino acid replacements.91

As expected, the Vif determinants involved in binding CBFβ (β1 strand), Cul5 (zinc‐coordinating HCCH motif), and EloC (BC‐box motif) all map to sites of strong conservation (Figure 6a). In contrast, the residues involved in A3 binding display mixed patterns of conservation (i.e., conserved, intermediate, or variable; Figure 6a,b). All Vif residues involved in A3F binding are conserved. For A3G binding, the majority of Vif residues are conserved, with the exception of four amino acids at positions 19, 22, 30, and 41. In contrast, a minority of Vif residues involved in A3H binding (5 out of 17) are conserved. These observed patterns of conservation emphasize that A3F‐, A3G‐, and A3H‐binding Vif surfaces are not only biochemically distinct, but are also evolutionarily disparate entities. Since A3F‐binding Vif residues are invariably conserved (i.e., least evolving), it is tempting to speculate that these residues may have, in fact, been the ancestral site of protein–protein interactions with the A3 substrate. As we know, the Vif residues involved in binding A3F, A3C, and A3D are highly similar (Figure 3a). Therefore, the ancestral Vif‐A3 interaction may have involved the structural determinants that mediate interactions of Vif with A3F, A3C, and A3D.

Figure 6.

New perspectives on the Vif‐A3 interface. (a) Summary of evolutionary conservation scores of each Vif residue calculated based on ∼3,500 HIV‐1 Vif sequences. Vif amino acid positions are classified into conserved (Scores 9–7), intermediate (Scores 6–4), and variable (Scores 3–1) groups. A3F‐, A3G‐, and A3H‐binding sites are shown in green, orange, and pink, respectively. Lollipops represent those sites predicted to be involved in direct contacts by hybrid methods. (b) Representation of surface residue conservation calculated using ConSurf for Vif and for the interaction sites on Vif with A3F, A3G, and A3H. Regions colored in burgundy display strong conservation and are considered to be slowly evolving; cyan indicates high variability, rapidly evolving residues. (c) Physiochemical analysis of A3‐binding amino acid positions that demonstrate variable conservation patterns. Amino acid positions 19, 22, 30, 37, 48, and 92 that demonstrate both evolutionary and physiochemical variability are mapped on the Vif structure

7.2. Evolutionarily variable A3 binding residues have nonconserved physiochemical properties

Sequence‐based conservation analyses are far from deterministic. We therefore evaluated the observed conservation patterns from ConSurf in the context of the physiochemical properties of the various residues observed among the HIV‐1 isolates used in the analysis. Eleven of the evolutionarily variable Vif residues are involved in binding either A3G or A3H, but not A3F (Figure 6(a)]. Positions 39, 41, 61, 63, and 93 are physiochemically conserved, as they demonstrate semi‐conservative substitutions in more than 98% of the aligned sequences (Figure 6c). The remaining six, at positions 19, 22, 30, 37, 48, and 92, are neither evolutionarily nor physiochemically conserved, yet they are critical to Vif‐A3G and Vif‐A3H interactions and Vif‐mediated A3 degradation. The simplest explanation for the lack of conservation among these functionally critical Vif positions is that these amino acids could potentially mediate interactions with A3 substrates through sequence‐independent means. For example, main chain hydrogen bonds or van der Waals interactions to Vif do not require specific side‐chain atom interactions. The lack of conservation of residues involved in Vif‐A3G and Vif‐A3H binding is surprising. Structures of Vif‐A3G and Vif‐A3H co‐complexes will be needed to understand why the interface residues are so variable.

8. CONCLUDING REMARKS

Since the discovery of the critical role of Vif in targeting and degrading A3 innate immune factors in the early 2000s, efforts have been underway to identify the Vif and A3 amino acids involved in direct interactions in order to develop inhibitors that disrupt the Vif‐A3 interface. Disruption of the Vif‐A3 interaction would, in principle, prevent the degradation of A3, thus enhancing its antiviral activity. Such small‐molecule inhibitors will need to disrupt the Vif‐A3 interface without interfering with A3 enzymatic activity or A3 packaging into nascent virions. Given that there appears to be at least three distinct Vif‐A3 binding modes, it will be difficult to design a single inhibitor that will universally block all Vif‐A3 interactions. Our structural analyses indicated that the Vif surface involved in binding A3F, A3C, and A3D is highly conserved, but it does not contain deep surface pockets or grooves capable of binding a small‐molecule inhibitor. In contrast, the concave side of the Vif β sheet, which is the site of A3H binding, and the area between the α1 helix and the convex side of Vif β sheet, which is the site of A3G binding, harbor deep surface features that may prove amenable to structure‐guided drug design. An inhibitor targeting a single Vif‐A3 interaction site may still provide potent therapeutic benefits, especially in combination with already existing drugs administered as part of highly active antiretroviral therapies.

AUTHOR CONTRIBUTIONS

F.C.A. and J.E.L. wrote the manuscript.

COMPETING FINANCIAL INTERESTS

The authors declare no competing financial interests.

Azimi FC, Lee JE. Structural perspectives on HIV‐1 Vif and APOBEC3 restriction factor interactions. Protein Science. 2020;29:391–406. 10.1002/pro.3729