Characterization of the Interaction of Full-Length HIV-1 Vif Protein with its Key Regulator CBFβ and CRL5 E3 Ubiquitin Ligase Components

Human immunodeficiency virus-1 (HIV-1) viral infectivity factor (Vif) is essential for viral replication because of its ability to eliminate the host's antiviral response to HIV-1 that is mediated by the APOBEC3 family of cellular cytidine deaminases. Vif targets these proteins, including APOBEC3G, for polyubiquitination and subsequent proteasome-mediated degradation via the formation of a Cullin5-ElonginB/C-based E3 ubiquitin ligase. Determining how the cellular components of this E3 ligase complex interact with Vif is critical to the intelligent design of new antiviral drugs. However, structural studies of Vif, both alone and in complex with cellular partners, have been hampered by an inability to express soluble full-length Vif protein. Here we demonstrate that a newly identified host regulator of Vif, core-binding factor-beta (CBFβ), interacts directly with Vif, including various isoforms and a truncated form of this regulator. In addition, carboxyl-terminal truncations of Vif lacking the BC-box and cullin box motifs were sufficient for CBFβ interaction. Furthermore, association of Vif with CBFβ, alone or in combination with Elongin B/C (EloB/C), greatly increased the solubility of full-length Vif. Finally, a stable complex containing Vif-CBFβ-EloB/C was purified in large quantity and shown to bind purified Cullin5 (Cul5). This efficient strategy for purifying Vif-Cul5-CBFβ-EloB/C complexes will facilitate future structural and biochemical studies of Vif function and may provide the basis for useful screening approaches for identifying novel anti-HIV drug candidates.

However, structural and functional analyses of full-length Vif continue to be limited by difficulty in obtaining suitable quantities of soluble full-length Vif protein [65,66,67,68]. In an attempt to overcome this limitation, a denaturing/refolding method has been developed for purifying soluble recombinant Vif [65,68,69,70]. Although this approach produced large quantities of full-length protein, the protein formed high molecular weight aggregates in solution [65,68,69,70]. Vif's tendency to aggregate and become insoluble has limited its structural characterization and functional analysis [63].
Co-expression of binding partners has been shown to improve the solubility and stability of various proteins [71]. Here, we report that co-expression of Vif with EloB/C and CBFb, a newly identified regulator of HIV-1 Vif function [72,73,74,75], can greatly improve the solubility of full-length Vif. We also demonstrate that C-terminal truncated Vif mutants of up to 140 amino acids can still interact with CBFb. Purified amino-terminal domain of Cul5 (residues 1-393) readily interacts with this complex. Vif-CBFb-EloB/C-Cul5 complexes purified by our strategy were not prone to aggregate and can therefore facilitate future structural and biochemical studies of Vif function.

Cloning, expression, and purification
Full-length Vif192 in the pET21 vector was a gift from Drs. Rahul M. Kohli and James T. Stivers. Truncated Vif176 and Vif140 were cloned into pET21 vector. Elongin B and Elongin C (residues 17 to 112) in the pACYC-Duet plasmid were a gift from Alex Bullock (University of Oxford, Oxford, United Kingdom). Mouse CBFb (residues 1-187) cDNA were a gift from Nancy A. Speck (University of Pennsylvania). CBFb isoform 1 (residues 1-187) from mouse, CBFb isoform 2 (residues 1-182) and truncated CBFb (residues 1-140) from human were cloned into pRSF-Duet. For expression, the plasmids were transformed into Escherichia coli BL21(DE3) cells. The constructs used in this study are summarized in Fig. 1. The proteins were over-expressed overnight at 16uC by induction with 0.1 mM isopropyl-D-thiogalactopyranoside (IPTG). Harvested cells were lysed in 20 mM Tris-HCl, pH 8.0, with 150 mM NaCl and then clarified by sonication and centrifugation at 13,000 g for 30 min. For solubility analysis, the supernatant was removed and the pellet resuspended to the original volume. For nickel affinity purification, the supernatant was transferred to Ni-NTA beads (Invitrogen), and the flowthrough was loaded onto Ni-NTA beads for two more passages. After washing with 20 mM Tris-HCl, pH 8.0, with 150 mM NaCl and 40 mM imidazole, the protein complex was eluted with 20 mM Tris-HCl, pH 8.0, with 150 mM NaCl and 400 mM imidazole. Gel filtration and anion exchange were utilized to remove trace contamination. Cul5-NTD (residues 1 to 393 with two point mutations, V341R and L345D) in the pGEX-6p-1 vector was expressed in E. coli BL21(DE3) cells overnight at 16uC by induction with 0.1 mM IPTG. Harvested cells were lysed by sonication in 20 mM Tris-HCl, pH 8.0, with 150 mM NaCl, then clarified by centrifugation at 13,000 g for 30 min. The supernatant was transferred to glutathione-Sepharose 4B beads (GE Healthcare) for glutathione S-transferase (GST) affinity chromatography. The GST tag was then removed using Prescission protease. Gel filtration chromatography was utilized for further purification.

Pull-down analysis of the Vif-CBFb interaction
For pull-down experiments analyzing the interactions between Vif and CBFb, supernatant was incubated on Ni-NTA agarose for 30 min at 4uC. After incubation, the reaction mixtures were washed 10 times with 1 ml lysis buffer. The samples were then analyzed by SDS-PAGE and visualized with Coomassie staining or by immunblotting with specific antibodies.

CBFb co-expression improves the solubility of Vif
To identify strategies that could result in the expression of large quantities of soluble full-length Vif recombinant proteins, we constructed various prokaryotic expression vectors for HIV-1 Vif and its co-factors ( Fig. 1). Recombinant Vif protein (residues 1-192) was efficiently expressed in E. coli BL21(DE3) but remained predominantly insoluble as indicated by Coomassie staining (Fig. 2A, lanes 1-3). The Vif protein was also identified by immunoblotting using a Vif-specific antibody (Fig. 2B, lanes 1-3). Co-expression with EloB/C improved the solubility of Vif, but only to a limited extent ( Fig. 2A and B, lanes 4-6). When Vif was co-expressed with CBFb140-His (residues 1-140 of CBFb with six histidine residues at the N-terminus), the solubility of Vif improved significantly ( Fig. 2A and B, lanes 7-9). Approximately 67% of the total Vif protein became soluble in the presence of CBFb140-His (Fig. 2C). Expressing CBFb and EloB/C together further enhanced the solubility of Vif ( Fig. 2A and B, lanes 10-12). When Vif was co-expressed with CBFb and EloB/C, .90% of the Vif proteins became soluble (Fig. 2C).

CBFb interacts with Vif
The ability of CBFb140-His to increase the solubility of Vif suggests that there is an interaction between Vif and CBFb140-His. To determine whether Vif and CBFb could interact directly, we attempted to co-precipitate Vif with CBFb140-His and found   that Vif in the soluble fraction could be efficiently pulled down by the CBFb140-His on a nickel column (Fig. 3B, lane 2). The presence of Vif and CBFb140-His in the soluble input fraction and the co-precipitated samples was confirmed by immunoblotting using a Vif-or CBFb-specific antibody (Fig. 3B).
There are two major CBFb isoforms that are highly conserved in mammals ( [76,77]: Isoform1 has 182 amino acids, while isoform 2 has a 187 amino acid sequence that is generated by alternative splicing. The two isoforms differ in the last 22 amino acids (Fig. 3A). Human and mouse CBFb differ by two amino acids (42 A/T and 117 Q/H). Next, we asked whether the natural isoforms of CBFb could interact with Vif and found that an interaction did indeed occur between HIV-1 Vif and isoform 1 CBFb182 (Fig. 3C) as well as isoform 2 CBFb187 (Fig. 3D) in co-precipitation experiments. To our knowledge, this is the first reported evidence of a direct interaction between HIV-1 Vif and various forms of CBFb, in vitro. Our data also indicate that amino acids 1-140 of CBFb are sufficient for HIV-1 Vif binding.
Purified Vif-CBFb-EloB/C proteins form a stable monomeric complex Soluble Vif and CBFb140 complexes were purified by nickel affinity chromatography and analyzed by gel filtration using a Superdex200 10/300 GL size exclusion column. Gel filtration analysis (Fig. 4A) suggested that Vif and CBFb140 formed a large aggregated complex of approximately 1000 kDa. Protein analysis by Coomassie staining of the peak fraction after separation by SDS-PAGE suggested a 1:1 ratio of Vif:CBFb140 (Fig. 4B). Full length or truncated CBFb were monomeric in solution [78]. This observation supports previous findings that Vif directly interacts with CBFb [72]. Gel filtration analysis of purified Vif-CBFb140-EloB/C revealed that the complex formed a homogeneous complex of ,65-75 kDa (Fig. 4C). Protein analysis by Coomassie staining of the peak fraction indicated a 1:1:1:1 ratio of Vif:CBFb140:EloB:EloC (Fig. 4D) or Vif:CBFb187:EloB:EloC (Fig. 4E). The calculated molecular weight of the monomeric Vif-CBFb140-EloB/C complex (,65 kDa) was in close agreement with our gel filtration results (,75 kDa) suggesting that Vif-CBFb-EloB/C complex is a monomeric complex in solution.
Previous studies have suggested that HIV-1 Vif can bind RNA [79,80,81,82]. We found that the Vif-CBFb140-EloB/C complexes were resistant to RNase treatment (Fig. 5A). Purified Vif-CBFb140-EloB/C complexes were untreated or treated with 40 mg/ml of RNase A and 20 U/ml RNase T1 at 37uC for 4 h. After buffer exchange, the treated samples were purified using nickel columns. RNase treatment did not affect the co-purification of Vif, EloB, and EloC with CBFb140-His (Fig. 5A, lane 2) when compared to the untreated sample (lane 1). These data suggest that the Vif-CBFb-EloB/C complexes are not RNA-dependent. The OD280/260 ratio in the peak fraction of the Vif-CBFb140 -EloB/ C complexes (Fig. 4C) also argued against the presence of RNA.

Interaction of CBFb with Vif truncation mutants
We next asked which region of Vif was required for the interaction between Vif and CBFb. Two truncated Vif mutants spanning residues 1-176 and 1-140 were constructed and co-expressed with CBFb140-His. Truncated Vif in the soluble fractions was analyzed by co-precipitation with CBFb140-His using nickel beads. SDS-PAGE and Coomassie staining indicated that both truncated Vif176 (Fig. 6A) and Vif140 (Fig. 6D) coprecipitated with CBFB140-His; this finding was confirmed by immunoblotting with a Vif-or CBFb-specific antibody (Fig. 6A  and D).The pulldown fractions were further analyzed by size exclusion. Both Vif176-CBFb140 (Fig. 6B) and Vif140-CBFb140 (Fig. 6E) formed large aggregates. Peak fractions were analyzed by SDS-PAGE followed by Coomassie staining. Both Vif176-CBFb140 (Fig. 6C) and Vif140-CBFb140 (Fig. 6F) showed a 1:1 ratio of Vif:CBFb. These results suggested that N-terminal residues 1-140 of HIV-1 Vif are sufficient for CBFb binding.

Vif-CBFb-EloB/C forms a complex with Cul5
Because binding to Cul5 is essential for Vif-mediated ubiquitination and degradation of target proteins such as A3G and A3F, we next determined whether these purified Vif-CBFb140-EloB/C complexes could interact with Cul5. Vif-CBFb140-EloB/C complexes and Cul5 NTD were purified separately (Fig. 7). The purified Vif-CBFb-EloB/C complexes were mixed with purified The purified complex (2 mg/ml, 100 ml) was incubated with 2 ml RNase Mix (RNase A/T1 Mix, Catalog EN055, Fermentas) at 37uC for 4 h according to the manufacturer's protocol, followed by buffer exchange to remove the EDTA. The complex then was analyzed by His-tag affinity pull-down. (B) The Vif-CBFb complex is not stable. Purified Vif-CBFb complexes were concentrated to 4 mg/ml and, after clarification at 13,000 g for 10 min, the supernatants were stored at 4uC (Input). Samples were then removed at different times (0 h, 6 h, 24 h), and after clarification at 13,000 g for 10 min, the supernatants (S) were removed and the pellets (P) resuspended to the original volume and checked by SDS-PAGE. (C) Purified Vif complexes were concentrated to 5 mg/ml (Input) and stored at 4uC overnight (,16 h). The supernatants (S) were removed after clarification at 13,000 g for 10 min, and the pellets (P) were resuspended to the original volume, then checked by SDS-PAGE. doi:10.1371/journal.pone.0033495.g005 Cul5 protein and subsequently analyzed by gel filtration. As compared to Vif-CBFb140-EloB/C (blue line) and Cul5 (cyan line), the mixture (red line) had an earlier elution peak (Fig. 8A). This result suggested that Vif-CBFb140-EloB/C may form a complex with Cul5. SDS-PAGE analysis of the peak fractions suggested that Cul5 and Vif-CBFb140-EloB/C formed a complex (Fig. 8B, upper panel). Molecular weight analysis by gel filtration (Fig. 8B, lower panel) indicated that the molecular size of the Vif-CBFb140-EloB/C-Cul5 complex was approximately 135 kDa, equal to the sum of Cul5 (,62 kDa) and Vif-CBFb140-EloB/C (,75 kDa). Further analysis using affinity pull-down via Histagged CBFb confirmed the formation of Cul5-Vif-CBFb140-EloB/C complexes (Fig. 8C). These Vif-CBFb140-EloB/C-Cul5 complexes were stable at 4uC over 16 h (Fig. 5C, lanes 7-9). The interaction between Cul5 and Vif-CBFb-EloB/C suggests that Vif-CBFb-EloB/C may be a functional complex, in vivo.

Discussion
Human CBFb has recently been identified as a critical regulator of HIV-1 Vif function [72,73,74,75]. In the present study, we demonstrate that this host regulator directly interacts with Vif alone and in complex with E3 ligase components, in vitro. CBFb is the non-DNA-binding subunit of a heterodimeric transcription factor, including RUNX family proteins [83,84]. CBFb regulates the folding and DNA-binding activity of RUNX partners, which play important roles in the development and differentiation of diverse cell types, including T lymphocytes and macrophages [83,84]. We have recently reported that CBFb is critical for Vifinduced A3G polyubiquitination and degradation [72]. Further clarification of the Vif-CBFb-EloB/C-Cul5 interaction and complex assembly would provide key insights into how Vif recruits these E3 ligase components to degrade A3G/A3F.
Co-expression of HIV-1 Vif with CBFb in the absence of all other human factors increased Vif solubility in E. coli. Soluble Vif could be co-precipitated with both His-tagged full length or truncated CBFb (Fig. 3C, D, and E) In the absence of binding partners, previous research has suggested full length Vif appears to be unstructured and poorly soluble, in vitro [85]. Recently, Wolfe et al. were able to obtain soluble C-terminal domain fragments of Vif in complex with EloB/C and Cul5 [63]. Attempts at characterizing full length Vif in complex with EloB/C and Cul5 were unsuccessful, suggesting that the N-terminus was responsible for Vif's poor solubility, in the absence of N-terminal binding partners. We have shown that CBFb binds the N-terminal region of Vif, specifically requiring hydrophobic interactions at amino acids W21 and W38 [72]. We hypothesize that the exposure of the N-terminal hydrophobic surface may contribute to Vif insolubilty when expressed alone. In vivo, CBFb appears to be necessary for Vif-Cul5 binding, though CBFb does not bind Cul5 directly [72,73]. Thus, a possible role for CBFb would be to stabilize Vif structure and promote the assembly of the Vif-Cul5 E3 ubiquitin ligase complex.
Vif and CBFb co-fractionated in gel filtration analyses and appeared as a 1:1 ratio complex. Isoforms 1 and 2 as well as a truncated form (amino acids 1-140) of CBFb all interacted with HIV-1 Vif. Thus, most, if not all, of the Vif binding activity is preserved within the first 140 amino acids of CBFb. Of note, Cterminal truncation of CBFb up to amino acids 1-135 have been reported to bind and act in complex with RUNX family proteins [86]. In addition, we have confirmed that CBFb binds to at least the first 140 amino acids of HIV-1 Vif. Thus, the known proteinbinding domains in Vif, including the EloB/C binding BC-box, the cullin box containing the PPLP motif, are not essential for the Vif-CBFb interaction. Vif forms homo-oligomers, and the PPLP motif has been suggested to be required for oligomerization [63,70,87,88,89,90]. Since Vif140 still forms oligomers with CBFb140, CBFb182, and CBFb187, our results suggest that regions in Vif in addition to PPLP may also participate in Vif oligomerization. This conclusion is consistent with the recent finding that the PPLP motif is not sufficient for Vif multimerization [64].
Biophysical and structural information for Vif has been limited as a result of its insolubility and strong tendency to oligomerize into high molecular weight aggregates. Of note, previous biochemical studies have employed full-length Vif protein obtained by the denaturing/refolding method [90] or have used truncated tagged protein [63]. Interestingly, when CBFb and EloB/C were present, even untagged full-length Vif could be purified as a stable and soluble complex.
Association of Vif with CBFb alone, and especially in combination with EloB/C, greatly increases the solubility of fulllength Vif. We have shown that a stable complex containing Vif-CBFb140-EloB/C can be purified in large quantities. This complex appeared to contain one subunit of each protein and did not dissociate upon RNase treatment. More importantly, the Vif-CBFb140-EloB/C complexes we produced could interact with purified Cul5 and form stable Vif-CBFb140-EloB/C-Cul5 complexes. This successful purification of monomeric Vif-E3 ligase complexes in high purity will greatly facilitate biochemical studies, structural determination, and functional analyses in this field.
Because CBFb is a unique regulator of Vif's ability to hijack the cellular CRL5 E3 ligase, disrupting interactions within the Vif-CBFb140-EloB/C-Cul5 complex represents an exciting drug strategy for targeting HIV-1. Inhibitors that prevent complex formation would be potential candidates for HIV-1 suppression, and purification of these Vif complexes in homogeneous form would provide the basis for screens to identify and evaluate inhibitor candidates. Thus, our strategy for purifying Vif-Cul5- CBFb-EloB/C complexes may lead to useful screening approaches for identifying novel anti-HIV drug candidates.