Natural Polymorphisms and Oligomerization of Human APOBEC3H Contribute to Single-stranded DNA Scanning Ability*

Background: Human APOBEC3H is a mutation-inducing restriction factor for HIV-1 that exists as multiple haplotypes. Results: Two APOBEC3H haplotypes were biochemically characterized. Conclusion: An amino acid polymorphism between APOBEC3H haplotypes II and V alters their processivity, but not mutagenic ability. Dimerization is required for efficient processivity. Significance: APOBEC3H has redundant processive mechanisms enabling haplotypes to efficiently restrict HIV-1 but by distinct mechanisms. APOBEC3H is a deoxycytidine deaminase that can restrict the replication of HIV-1 in the absence of the viral protein Vif that induces APOBEC3H degradation in cells. APOBEC3H exists in humans as seven haplotypes (I–VII) with different cellular stabilities. Of the three stable APOBEC3H haplotypes (II, V, and VII), haplotypes II and V occur most frequently in the population. Despite APOBEC3H being a bona fide restriction factor, there has been no comparative biochemical characterization of APOBEC3H haplotypes. We characterized the ssDNA scanning mechanisms that haplotypes II and V use to search their ssDNA substrate for cytosine-containing deamination motifs. APOBEC3H haplotype II was able to processively deaminate multiple cytosines in a single enzyme-substrate encounter by using sliding, jumping, and intersegmental transfer movements. In contrast, APOBEC3H haplotype V exhibited diminished sliding and intersegmental transfer abilities but was able to jump along ssDNA. Due to an Asp or Glu at amino acid 178 differentiating these APOBEC3H haplotypes, the data indicated that this amino acid on helix 6 contributes to processivity. The diminished processivity of APOBEC3H haplotype V did not result in a reduced efficiency to restrict HIV-1 replication in single-cycle infectivity assays, suggesting a redundancy in the contributions of jumping and intersegmental transfer to mutagenic efficiency. Optimal processivity on ssDNA also required dimerization of APOBEC3H through the β2 strands. The findings support a model in which jumping can compensate for deficiencies in intersegmental transfer and suggest that APOBEC3H haplotypes II and V induce HIV-1 mutagenesis efficiently but by different mechanisms.

The human APOBEC3 (A3) 2 family contains seven deoxycytidine deaminases that act as cellular restriction factors for a number of retrotransposons, retroviruses, and DNA viruses (1,2). Four of the human A3 enzymes, A3D, A3F, A3G, and A3H, are able to restrict the replication of HIV-1 (referred to as HIV) by becoming encapsidated into budding virions and deaminating cytosines to uracils on the nascent (Ϫ)-DNA synthesized by HIV reverse transcriptase (3,4). The uracil lesions result in transition mutations upon using the (Ϫ)-DNA as a template to synthesize the (ϩ)-DNA (5)(6)(7). This results in decreases in HIV infectivity that are dependent on the number of mutations and the effect of the mutations on the viral proteins. Each deaminase has a preferred sequence context for deamination, but the minimal dinucleotides for recognition are 5ЈCC for A3G and 5ЈTC for A3D, A3F, and A3H (8 -13).
The number of mutations that an A3 can introduce into the HIV proviral genome is mediated by the level of encapsidation into the viral capsid and the biochemical properties of the A3 (2,14). The primary factor mediating A3 restriction ability is the HIV Vif (viral infectivity factor) protein (15). Vif acts as the substrate receptor of a CRL5 (Cullin RING ligase-5) E3 ligase complex to induce polyubiquitination and degradation of A3s relevant to HIV restriction (16 -22). This greatly decreases the encapsidation levels of A3s into HIV virions but does not completely block A3 virion encapsidation because G to A mutations in a sequence context indicating the action of A3 enzymes are present in HIV genomes recovered from infected individuals (23). In the absence of Vif and despite equal encapsidation of some A3s into virions, there are differences in their efficiency of HIV restriction (e.g. A3G and A3F) (24,25). These data demonstrate that there are biochemical differences in the enzymes that influence their activities as restriction factors (25). We have determined that the mechanisms that A3G and A3F use to scan the ssDNA substrate in search of the deamination motifs affect the number of induced mutations (25,26). Synthesis of the proviral dsDNA is a dynamic process concurrently involv-ing (Ϫ)-DNA synthesis, RNA degradation, and synthesis of (ϩ)-DNA that blocks A3 access to its single-stranded (Ϫ)-DNA substrate (8,27). As a result, the enzymes require an efficient search process to locate the deamination motifs (2). A3 enzymes studied thus far have been found to scan ssDNA by facilitated diffusion (25, 28 -30). Facilitated diffusion can involve one-dimensional sliding movements or three-dimensional jumping/hopping or intersegmental transfer movements on ssDNA (31,32). The most efficient search on ssDNA would involve a combination of one-and three-dimensional scanning movements (2,26,33). The one-dimensional sliding motions allow for an in depth local search (i.e. Ͻ20 nt) of the substrate for the deamination motif (2,34). The three-dimensional jumping/hopping or intersegmental transfer movements mediate larger translocations on the DNA substrate (2,34). Each type of movement provides a superficial scanning mechanism by itself, but together the enzyme is able to transition between an in depth search (sliding) for the catalytic sequence to random relocation (jumping/hopping or intersegmental transfer) on the DNA substrate to restart the search process (25,26). Jumping/hopping are differentiated from intersegmental transfer by the type of association the enzyme has with the DNA. Jumping/ hopping involves microscopic dissociation and reassociation of the enzyme with the same DNA without diffusion into the bulk solution. Intersegment transfer involves an enzyme with two DNA binding domains engaging in a doubly bound state with the DNA substrate before moving to one of the bound sites (32). Facilitated diffusion movements enable enzyme processivity, where an enzyme can catalyze multiple reactions before dissociating from the substrate and diffusing back into the bulk solution (2).
Of the A3 enzymes most effective against HIV, A3H is of special interest because it occurs with the highest number of haplotypes in the A3 family (13,35,36). A3H exists as at least seven haplotypes in humans, with three of the haplotypes being stable forms of A3H that can suppress HIV ( Table 1). The A3H haplotypes that are unstable and inactive against HIV contain two single nucleotide polymorphisms that occur in different combinations with other polymorphisms. Unstable A3H haplotypes (haplotypes I, III, IV, and VI) carry either a Gly-105, ⌬Asn-15, or both in combination with other possible polymorphisms at positions 18, 121, and 178 (35,36) (Table 1). Unstable A3H haplotype I occurs at the highest population frequency (0.308 -0.526) (35,36). An Arg-105 and Asn-15 are required for a stable A3H in cells (35). The three stable haplotypes all have an Asn-15, Arg-18, and Arg-105 and vary at the 121 and 178 positions (35,36) (Table 1). A3H haplotype II has been the most widely studied and has a DD sequence at positions 121 and 178 (37)(38)(39). A3H haplotype V has a DE sequence at positions 121 and 178 (13,36). A3H haplotype II and V comprise the majority of the stable A3H alleles (haplotype II, frequency of 0.061-0.265; haplotype V, frequency of 0.054 -0.202) (35,36). A3H haplotype VII has been identified only from one individual (frequency of 0.009) and has a KE sequence at positions 121 and 178 (36). The 121 position determines the sensitivity to Vif, where an Asp-121 is sensitive and a Lys-121 is not sensitive to Vif-mediated degradation (13,40,41). There is currently no known function of the polymorphism of Asp/Glu at position 178 (41).
Despite the several unique aspects of A3H in comparison with the other A3 enzymes, there has been no in depth biochemical characterization of the most frequently occurring stable A3H haplotypes (II and V). This type of characterization is necessary to fully understand the determinants required for HIV restriction because A3H has several distinctive characteristics in the A3 family. A3s can be divided into two families based on the number of zinc-coordinating domains (Z-domains) (42). A3B, A3D, A3F, and A3G are deaminases with two Z-domains, whereas A3A, A3C, and A3H are deaminases with a single Z-domain. Notably, A3H is the only single Z-domain A3 that can restrict the replication of HIV (3). In addition, A3s have been classified into three categories based on the phylogenetic analysis of their Z-domains (Z1, Z2, and Z3), and A3H is the only Z3-type domain A3 enzyme (42). Part of the reason why A3H is able to restrict HIV is that it is the only single Z-domain A3 that oligomerizes in solution and on cellular RNA, which is a prerequisite for virion encapsidation (43,44). However, there are no oligomerization models for single Z-domain A3 enzymes that have been verified as contributing to cellular function.
We undertook a biochemical characterization of A3H haplotypes II and V in order to determine whether there are shared biochemical characteristics between single Z-domain and double Z-domain A3 enzymes for restriction of HIV. The data demonstrate that A3H is a processive enzyme. A3H haplotype II uses sliding, jumping, and intersegmental transfer to locate deamination motifs on ssDNA. The Glu-178 polymorphism of A3H haplotype V causes diminished sliding and intersegmental transfer but does not affect jumping. This difference in processivity did not affect HIV restriction ability. Further, we identified that the dimerization interface in A3H is mediated by the pairing of ␤2 strands from two monomers, and tetramerization can occur through the pairing of loop 7 from two dimers, as predicted could occur from a crystal structure of APOBEC2 (A2) with a 40-amino acid N-terminal deletion (45). The ␤2-␤2 strand dimerization of A3H is required for efficient ssDNA scanning. The results demonstrate that A3H dimerization and Asp-178 is required for sliding and intersegmental transfer movements on ssDNA. The data combine to support a model in which A3H haplotypes II and V both induce HIV mutagenesis efficiently but by different mechanisms and suggest that jumping can compensate for deficiencies in intersegmental transfer.

Experimental Procedures
Cloning and Site-directed Mutagenesis-An A3H haplotype I clone (NCBI accession number BC069023) was obtained from Open Biosystems, and site-directed mutagenesis was used make A3H haplotype II (G105R/K121D/E178D) and haplotype V (G105R/K121D). A GST-A3H construct was subcloned from a pAcG2T vector (BD Biosciences) into a pFAST-bac1 vector (Life Technologies, Inc.) using EcoRI and NotI restriction sites. Mutants of A3H haplotype II (R44A/Y46A, Y111A/Y112A, D178K, and D178A) were made using the A3H haplotype II construct for site-directed mutagenesis (QuikChange site-directed mutagenesis protocol, Stratagene). The cDNA for A3H haplotypes II and V was also cloned into pcDNA3.1 with an HA tag. For generation of RNA in vitro, the HIV 5Ј-UTR (nucleotides 1-497) and human Alu sequence (46) were cloned into pSP72 vector (Promega) using XhoI and HindIII sites under the control of T7 promoter (5Ј-UTR) or SP6 promoter (Alu). All constructed plasmids were verified by DNA sequencing. Primers were obtained from Integrated DNA Technologies and are listed in supplemental Table S1.
Protein Expression and Purification-The pFASTbac1-GST-A3H vectors were used to produce recombinant baculovirus according to the protocol for the Bac-to-Bac system (Life Technologies). Recombinant GST-A3H baculovirus were then used to infect Sf9 cells at a multiplicity of infection of 20, and cells were harvested after 72 h. Cells were lysed, and clarified lysates were incubated with glutathione-Sepharose 4B resin (GE Healthcare) at 4°C and subjected to a series of salt washes as described previously (47). For all A3H haplotypes and mutants except for the A3H haplotype II R44A/Y46A, D178K, and D178A mutants, on-column cleavage from the GST tag with thrombin (GE Healthcare) was performed at 21°C for 18 h in thrombin digestion buffer (20 mM HEPES, pH 7.5, 150 mM NaCl, 10% glycerol, and 1 mM DTT). Enzymes were assessed to be 95% pure by SDS-PAGE. Due to the lower expression of the A3H haplotype II R44A/Y46A, D178K, and D178A mutants, the enzyme was eluted with the GST tag in elution buffer (100 mM Tris, pH 8.8, 150 mM NaCl, 10% (v/v) glycerol, and 50 mM reduced glutathione). A wild type GST-haplotype II was purified in parallel with the mutant for comparative analyses. Eluted GST-A3H was dialyzed against 100 mM Tris, pH 7.5, 250 mM NaCl, 10% glycerol, and 1 mM DTT overnight at 4°C. For size exclusion chromatography with the A3H haplotype II R44A/ Y46A mutant, the GST tag was not cleaved to increase protein stability. For deamination reactions with the A3H haplotype II R44A/Y46A, D178K, and D178A mutants, the GST tag was cleaved in solution with thrombin (GE Healthcare).
Single-cycle Infectivity Assay-VSV-G pseudotyped HIV pNL4-3 ⌬Vif viruses were produced by transfecting 3 ϫ 10 5 293T cells (ATCC CRL-3216) per well in a 6-well plate with GeneJuice (Novagen). Specifically, transfections used 500 ng of pHIV⌬Vif, which expresses an enhanced GFP reporter gene (48) and 200 ng of pMDG (49,50), which expresses the VSV-G protein, in the presence or absence of A3H haplotype II or A3H haplotype V in pcDNA3.1. Empty pcDNA3.1 was used to achieve equivalent amounts of transfected DNA. A titration of A3H expression vector was used that ranged from 25 to 200 ng. The amino acid sequences of the A3s were identical to those used in biochemical assays. Twenty-four hours after the transfection, the medium was replaced. Virus-containing supernatants were collected 48 h after the medium change and filtered through 0.45-m syringe filters. Virus was quantified by a p24 enzyme-linked immunosorbent assay (QuickTiter lentivirus titer kit (Cell Biolabs Inc.) or HIV-1 p24 ELISA kit (XpressBio)). Target 293T cells were infected at a multiplicity of infection of 0.5 by spinoculation at 800 ϫ g for 1 h in the presence of 8 g/ml Polybrene. Infection levels in 293T cells were determined by flow cytometry by detecting enhanced GFP fluorescence at 40 h postinfection, and data were normalized to HIV⌬vif infections in the absence of A3 enzymes.
Immunoblotting-To detect A3H protein levels, HA-tagged versions of the A3H enzymes used for the single-cycle infectivity assays were constructed in pcDNA3.1. For detection of A3H in cell lysates (50 g of total protein) and virions (50 ng of p24), anti-rabbit HA (Sigma) was used. Loading controls for cell lysates (␣-tubulin, Sigma) and virions (p24, catalog no. 3537, National Institutes of Health AIDS Reagent Program) were detected using mouse monoclonal antibodies. Tagged proteins and loading controls were detected simultaneously on the same immunoblot by using the LI-COR/Odyssey system (IRDye 680labeled goat anti-rabbit and IRDye 800-labeled goat anti-mouse secondary antibodies). Antibodies were used at a dilution of 1:1000.
Sequencing of Integrated Proviral DNA-Infected 293T cells were harvested after 40 h, and the DNA was extracted using DNAzol reagent (Life Technologies). DNA was treated with DpnI (New England Biolabs) for 1 h at 37°C to remove possible contaminating plasmid DNA, and the protease (prot) (nt 2280 -2631) sequences were amplified by PCR using Q5 polymerase (New England Biolabs). Primers were obtained from Integrated DNA Technologies and are listed in supplemental Table S1. PCR products were purified and cloned with the Zero Blunt TOPO PCR cloning kit (Life Technologies) or New England Biolabs PCR cloning kit. DNA was sequenced with kit-specific primers and carried out at the National Research Council of Canada (Saskatoon, Canada) or Eurofins Genomics (Huntsville, AL).
In Vitro Deamination Assay-All ssDNA substrates were obtained from Tri-Link Biotechnologies and are listed in supplemental Table S1. All reactions were carried out under singlehit conditions (i.e. Ͻ15% substrate usage) to ensure that deaminations on each ssDNA were catalyzed by a single enzyme (51). Under these conditions, a processivity factor can be determined by comparing the total amount of deaminations occurring at two sites on the same ssDNA with a calculated theoretical value of deaminations at these two sites if the deamination events were uncorrelated (not processive) (28). In order to achieve Ͻ15% substrate usage, the ssDNA substrate containing two 5Ј-CTC motifs (100 nM) was incubated with 50 nM A3H for 1.5 to 20 min at 37°C in deamination buffer containing 50 mM Tris, pH 7.5, 40 mM KCl, 10 mM MgCl 2 , and 1 mM DTT. The reaction time was varied according to the specific activity of the enzyme on each ssDNA to achieve ϳ10% substrate usage. DNA substrates with a dsDNA or RNA/DNA hybrid were formed by heat annealing in 50 mM Tris, pH 7.5, and 50 mM NaCl. Reactions were started by the addition of the ssDNA substrate. To test whether intersegmental transfer was occurring, the A3H/ ssDNA ratio (1:2) was kept constant, but increased amounts of enzyme and substrate were titrated into the reaction (ssDNA, 100 -700 nM; A3H, 50 -350 nM). A3H-catalyzed deaminations were detected by treating the ssDNA with uracil DNA glycosylase (New England Biolabs) and heating under alkaline conditions before resolving the fluorescein-labeled ssDNA on 10, 16, or 20% (v/v) denaturing polyacrylamide gels, depending on the sizes of the ssDNA fragments. Gel photos were obtained using a Typhoon Trio multipurpose scanner (GE Healthcare), and integrated gel band intensities were analyzed using ImageQuant (GE Healthcare) as described previously (28).
Size Exclusion Chromatography-The oligomerization states of A3H haplotype II, haplotype V, and haplotype II Y112A/ Y113A were determined by loading 150 or 300 g of the purified enzymes onto a Superdex 200 10/300 Increase column (GE Healthcare). The running buffer contained 20 mM Tris, pH 8.0, 300 mM NaCl, 10% (v/v) glycerol, and 1 mM DTT. The oligomerization states of GST-A3H haplotype II and GST-A3H haplotype II R44A/Y46A were examined by loading 10 g of the purified enzymes onto a size exclusion chromatography column. The column was prepared by pouring a 10-ml Superdex 200 (GE Healthcare) resin bed contained in a column with a 0.5-cm diameter and 16-cm height. The running buffer contained 20 mM Tris, pH 8.0, 300 mM NaCl, 10% (v/v) glycerol, and 5 mM DTT. The Bio-Rad gel filtration standard set was used to generate a calibration curve from which the apparent molecular masses and oligomerization states of the enzymes were determined.
Steady-state Rotational Anisotropy-The apparent K d values of A3G, A3H haplotype II, and A3H haplotype V for ssDNA, dsDNA, RNA/DNA hybrid, or RNA were determined using steady state fluorescence depolarization (rotational anisotropy). The ssDNA substrate was the 118-nt ssDNA used for deamination assays and contained an internal fluorescein-dT (supplemental Table S1). The dsDNA and RNA/DNA hybrid were composed of 20 nt that matched the hybrid regions used to block A3H sliding in deamination assays (supplemental Table S1).
Reactions (50 l) were conducted in deamination buffer and contained 50 nM fluorescein-labeled DNA or RNA and increasing amounts of A3 (A3G, 0 -1000 nM; A3H haplotypes II and V, 0 -1500 nM). A QuantaMaster QM-4 spectrofluorometer (Photon Technology International) with a dual emission channel was used to collect data and calculate anisotropy. Measurements were performed at 21°C. Samples were excited with vertically polarized light at 495 nm (6-nm band pass), and vertical and horizontal emissions were measured at 520 nm (6-nm band pass). The K d was obtained by fitting to a rectangular hyperbola or sigmoidal curve equation using SigmaPlot version 11.2 software.

A3H Haplotype II and V Processively Scan ssDNA-
The ssDNA scanning mechanism(s) of A3H haplotypes II and V were determined using an in vitro deamination assay that measures the processivity of purified A3H. Processivity is defined as the ability of an enzyme to deaminate more than one cytosine on an ssDNA substrate in a single enzyme-substrate encounter (28). The processivity was measured using different synthetic ssDNA substrates containing two deamination motifs (5ЈCTC, where the underlined C is preferentially deaminated) spaced various distances apart. This strategy allows the dissection between one-dimensional sliding and three-dimensional trans-location movements because it has been previously shown that substrates with closely spaced motifs (i.e. Ͻ20 nt) are processively deaminated through one-dimensional sliding motions (25,26). Processive deamination of more distantly spaced motifs requires a three-dimensional translocation mechanism, such as jumping or intersegmental transfer (2,25,26). All deamination reactions were performed under "single-hit" conditions (Ͻ15% substrate usage) to ensure that any given ssDNA substrate was acted upon by at most one enzyme during the reaction (51). Under these conditions, a processivity factor can be calculated to determine the likelihood that an enzyme would undergo processive deaminations in comparison with a nonprocessive deamination (see "Experimental Procedures" and Ref. 28).
On ssDNA substrates with deamination motifs separated by 5 and 14 nt, A3H haplotype II had processivity factors of 2.1 and 3.6, respectively (Fig. 1, A and B). Thus, depending on the distance between the deamination motifs, A3H haplotype II is ϳ2-4 times more likely to catalyze a processive deamination of the two deamination motifs than a single nonprocessive deamination of one of the deamination motifs. Due to the short distance between the deamination motifs, the data show that A3H haplotype II can processively scan the ssDNA by sliding. The amino acid sequences of A3H haplotype II and A3H haplotype V differ only at the amino acid at position 178, where A3H haplotype II has an Asp and A3H haplotype V has a Glu (13, 36) ( Table 1). Because the amino acid at position 178 is located on predicted helix 6, which mediates sliding ability in A3G and A3F (25,26), we tested A3H haplotype V on ssDNA substrates with closely spaced deamination motifs to determine whether the sliding ability was compromised. On the substrate with deamination motifs separated by 5 nt, the processivity factor was 1 (Fig. 1A). Because the processivity factor is a ratio, the processivity factor of 1 means that A3H haplotype V is essentially not processive on this substrate (see "Experimental Procedures"), in contrast to A3H haplotype II (Fig. 1A). On the substrate with cytosines separated by 14 nt, A3H haplotype V is processive but has a processivity factor ϳ2-fold less than A3H haplotype II (compare processivity factors of 3.6 and 2.1 (Fig.  1B)). The data indicated that in A3H, predicted helix 6 mediates sliding and that the Asp/Glu-178 polymorphism between A3H haplotype II and haplotype V caused a difference in the ssDNA sliding ability.
We also tested whether other processive mechanisms were affected by the polymorphism at amino acid position 178. On the ssDNA substrate with cytosines separated by 63 nt, A3H haplotype V had a processivity factor of 3.8, which is ϳ1.5-fold less processive as compared with A3H haplotype II on the same substrate (compare processivity factors of 5.2 and 3.8 (Fig. 1C)). That both A3H haplotypes were processive on a substrate with distantly spaced deamination motifs suggests that A3H uses a three-dimensional translocation mechanism to scan ssDNA. To confirm this, we annealed a 20-nt complementary DNA between the two deamination motifs (Fig. 1D, schematic). Because A3H haplotype II does not bind to this dsDNA region ( Table 2, no binding detected), it serves as a block to inhibit the sliding motion of the enzyme (28). A3H haplotype II can still processively deaminate the cytosines on this substrate (Fig. 1D, processivity factor of 2.2), which confirms that A3H haplotype II uses a three-dimensional translocation mechanism to overcome obstacles while scanning ssDNA. The ϳ2.5-fold decrease in processivity in comparison with the completely ssDNA (compare processivity factors; Fig. 1, C and D) is due to attempts to slide over the dsDNA region that instead result in dissociation of the enzyme into the bulk solution (25). A3H haplotype V bound to dsDNA with an affinity 10-fold less than ssDNA, suggesting that it would also not be able to slide over FIGURE 1. Analysis of A3H haplotype II and V processivity and ssDNA scanning. Processivity of A3H haplotype II (Hap II) and haplotype V (Hap V) was tested on ssDNA substrates that contained fluorescein-labeled deoxythymidine (yellow star) between two 5Ј-CTC deamination motifs separated by different distances. A, deamination of a 60-nt ssDNA substrate with deamination motifs spaced 5 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 42-and 23-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 5-nt labeled fragment. B, deamination of a 69-nt ssDNA substrate with deamination motifs spaced 14 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 51-and 32-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 14-nt labeled fragment. C, deamination of a 118-nt ssDNA substrate with deaminated cytosines spaced 63 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 100-and 81-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 63-nt labeled fragment. D, deamination of the same substrate described for C but with a 20-nt DNA annealed between the two deamination motifs. E, deamination of the same substrate described for C, but with a 20-nt RNA annealed between the two deamination motifs. A-E, the A3H/ssDNA ratio was 1:2. The measurements of enzyme processivity (processivity factor) and the S.D. are shown below the gels. All values are calculated from at least three independent experiments.  ). Similar to A3H haplotype II, the dsDNA region caused a decrease in the processivity of A3H haplotype V, but the enzyme was still processive (Fig. 1D, processivity factor of 1.7). Interestingly, we found that A3H haplotypes II and V are able to bind RNA/DNA hybrids with an apparent dissociation constant that is equivalent to that of ssDNA (  )). These data predict that A3H should be able to slide over RNA/DNA hybrids. To test this prediction, we annealed a 20-nt complementary RNA between the two deamination motifs (Fig. 1E, schematic). The enzyme processivity was not significantly different on this substrate in comparison with its analogous ssDNA for both A3H haplotypes II and V ( Fig. 1, C and E, processivity factors). This suggests that during HIV reverse transcription, A3H haplotypes II and V could slide on RNA/ DNA hybrid regions.
To determine the three-dimensional translocation mechanism of A3H haplotypes II and V, we tested whether the enzyme could move by intersegmental transfer. This mechanism requires that an enzyme with two DNA binding domains transfer to distal sites through a "doubly bound" intermediate state (32). A3H haplotype II has been shown to form dimers in solution (43), indicating that it could use intersegmental transfer. To observe intersegmental transfer, we increased the enzyme and substrate concentrations in the reaction but kept their ratios constant (Fig. 2). By crowding the reaction in this manner, the enzyme will become more likely to translocate to a different substrate than to translocate within a single ssDNA as the concentration of the reaction components increases (52). Thus, if intersegmental transfer is occurring, this would lead to an apparent decrease in the processivity factor. Intersegmental transfer should also result in an increase in the apparent reaction rate because increasing the cycling of the enzyme also increases the rate of searching for deamination motifs (52). The processivity factor of A3H haplotype II decreased ϳ2-fold with increasing reaction components ( Fig. 2A, processivity factor decreases from 5.2 to 3.0), and this was accompanied by a ϳ2-fold increase in the rate of deamination ( Fig. 2A, rate increases from 0.35 to 0.60%/min), demonstrating that A3H haplotype II can use intersegmental transfer to scan ssDNA. However, at the highest level of reaction components, the processivity plateaued and did not decrease to 1, which would otherwise indicate a complete loss of apparent processivity due to the reaction conditions ( Fig. 2A, processivity factor of 3.0). This indicates that A3H haplotype II may also use jumping to translocate on ssDNA. Processive ssDNA scanning mediated by jumping is insensitive to the crowding of the reaction components because during this movement, the enzyme remains within the negatively charged region of the ssDNA and out of the bulk solution (2). Altogether, the results indicate that A3H haplotype II uses sliding and both intersegmental transfer and jumping as a three-dimensional translocation mechanism to scan ssDNA for its preferred deamination motif. We next examined whether A3H haplotype V was capable of intersegmental transfer. A3H haplotype V did have a 1.5-fold decrease in apparent processivity, similar to A3H haplotype II, but there was no increase in the reaction rate observed (Fig. 2, compare A  and B). Thus, the data indicated that although A3H haplotype V could transfer to other ssDNA substrates, there was no corresponding increase in cycling between the ssDNAs, suggesting that the Glu-178 polymorphism also diminished intersegmental transfer ability. The processivity factor from the intersegmental transfer assay at the highest concentration of reaction components where the decrease in processivity plateaued was greater than 1, suggesting that A3H haplotype V was able to processively jump on ssDNA (Fig. 2B).
Of note, the identity of the amino acid at position 178 also affected the binding of dsDNA and RNA. Although A3H haplotypes II and V did not differ in their ssDNA binding ability as determined by their apparent dissociation constants (K d ) (   NOVEMBER (13,35,37,39,(61)(62)(63). In agreement with others, our data indicated that in a single cycle of replication, A3H haplotype II and V suppress HIV similarly over a range of A3H expression levels (36,39,64,65) (Fig. 3, A and B). Despite the 10 -20-fold differences in RNA binding affinities between A3H haplotypes II and V (Table 3), their virion encapsidation was comparable (Fig. 3B). It has been shown that A3G binds cellular or viral RNA indiscriminately to ensure encapsidation into virions in a nucleocapsid-specific manner (66), and our data suggest that the same encapsidation mechanism applies to A3H.

A3H Polymorphisms and Oligomerization Impact Processivity
It has been previously established for A3G and A3F that an efficient processive DNA scanning mechanism involving both one-dimensional sliding and three-dimensional translocation is required for inducing high numbers of deoxycytidine deaminations in HIV (Ϫ)-DNA (25). Because we observed that A3H haplotypes II and V were encapsidated into virions at similar levels to each other, and there were no significant differences in suppression of HIV infectivity (Fig. 3, A and B), the data indicate that the diminished processivity of A3H haplotype V in comparison with A3H haplotype II (Fig. 1) is inconsequential to A3H-mediated HIV restriction. This may be because A3H haplotype V did not have a complete loss of sliding ability and still maintained jumping as a three-dimensional translocation mechanism (Fig. 1). Further, the A3H haplotype II data suggest that the three-dimensional translocation mechanisms are redundant because A3H haplotype II did not restrict HIV replication more than A3H haplotype V over a range of cellular expression levels (Fig. 3, A and B). We sequenced proviral DNA from the transfection conditions that used 100 ng of A3H expression plasmid and found no significant differences in the spacing of the A3H haplotype II and V mutations, confirming that the diminished processivity of A3H haplotype V did not affect its ability to induce mutations in HIV proviral DNA (Fig.  3C). Processivity can be inferred from the mutational spectra because clustered mutations are indicative of sliding, and distantly spaced mutations are indicative of jumping or intersegmental transfer (25,26,67). Because there were similar distantly spaced mutations for both A3H haplotypes, the data further indicate that intersegmental transfer and jumping are redundant three-dimensional processive mechanisms (Fig. 3C). Both A3H haplotypes II and V also induced closely spaced mutations, indicative of sliding, in agreement with the processivity data from ssDNA oligonucleotides (Figs. 3C and 1 (A  and B)). The ϳ2-fold lower processivity of A3H haplotype V in comparison with A3H haplotype II for closely spaced deamination motifs on ssDNA oligonucleotide substrates (Fig. 1, A and  B) did not appear to affect its ability to induce closely spaced mutations during proviral DNA synthesis (Fig. 3C). At most, there was only a 2-fold difference in the G to A mutation frequency of A3H haplotypes II and V, and this was observed only at the 100-ng A3H expression plasmid condition ( Fig. 3A and Table 4).
Optimal Sliding and Intersegmental Transfer Require an Asp-178 -The A3H haplotypes II and V differ by only an Asp/ Glu at amino acid position 178. These amino acids have differences in side chain rigidity and pK a that could account for the effect on ssDNA scanning ability. However, we investigated whether a charge reversal or removal mutation could more drastically alter the processivity of A3H. The sliding ability of a D178K mutant was tested using an ssDNA substrate that had deamination motifs separated by 5 nt (Fig. 4A). Similar to A3H haplotype V, the A3H D178K mutant was not processive on this substrate, indicating a deficiency in sliding ability (compare Figs. 4A and 1A, processivity factors of 1). The D178K mutant was processive on the ssDNA substrate with deamination motifs separated by 63 nt, indicating that it could use a processive three-dimensional translocation mechanism (Fig. 4B). In the assay that tests intersegmental transfer ability, the D178K mutant processivity decreased 1.6-fold with increasing concentration of enzyme and substrate, but there was no increase in the reaction rate (Fig. 4B). Thus, the D178K mutant was similar to A3H haplotype V and could transfer to other ssDNA substrates but was unable to cycle efficiently (compare Figs. 2B and 4B). Altogether, the data indicated that rather than the amino acid charge, the Asp at amino acid position 178 was specifically important for sliding and intersegmental transfer. However, a charged amino acid is important at amino acid position 178 for enzyme stability because a D178A mutant was less stable than A3H haplotype II and had processivity factors of 1.7 or less on all ssDNA substrates tested (data not shown). This suggests that a salt bridge may be formed from the amino acid position at 178 to stabilize the enzyme and influence interactions with ssDNA.
The Dimer Interface of A3H Is Mediated by Predicted ␤2-␤2 Strand Interactions-The finding that an amino acid change on A3H predicted helix 6 affected sliding motions was consistent with published data (25, 26), but a role for helix 6 in interseg-

A3H Polymorphisms and Oligomerization Impact Processivity
mental transfer was unexpected. The current understanding of intersegmental transfer is that two DNA binding domains are required (32,34). Additional determinants for this ssDNA scanning mechanism have not been previously identified. However, there appear to be additional determinants required for A3 enzymes because those with two Z-domains in a single polypeptide (i.e. A3F and A3G) are unable to readily undergo intersegmental transfer under our assay conditions (25). Based on intersegmental transfer requiring two DNA binding domains (32), the data indicated that because we could not completely abolish intersegmental transfer ability by making mutations at the amino acid at position 178, the A3H was remaining in its oligomeric state, and we had identified an amino acid determinant required for intersegmental transfer.
To fully characterize the determinants for intersegmental transfer in A3H, we characterized the oligomeric interfaces of A3H haplotype II in order to design mutations that would result in an A3H haplotype II monomer to determine the effect on intersegmental transfer. Using size exclusion chromatography, we observed that A3H haplotype II and V both primarily form dimers in solution (apparent molecular mass of 44 kDa; Fig. 5, A and B). We also observed minor monomer (apparent molecular mass of 24 kDa) and tetramer (apparent molecular mass of 98 kDa) peaks (Fig. 5,  A and B). A dimer and tetramer of A3H are of equivalent molecular mass to a monomer and dimer of A3G, respectively (46). The formation of A3H tetramers was not concentration-dependent because the chromatograms for 150 and 300 g of enzyme showed that the tetramer fraction remained at only 20% (haplotype II) or 25% (haplotype V) of the total population (Fig. 5, A and inset graph). These data confirmed that Asp-178 was not responsible for mediating dimerization, and instead the amino acid at position 178 is a specific determinant that contributes to intersegmental transfer ability.
Because helix 6 did not mediate oligomerization for A3H (Fig. 5, A and B) as may have occurred based on A3C crystal structure contacts (68), we performed our structure-guided mutagenesis using another single Z-domain oligomerization model, the A2 crystal structure, and surrogate mutations made in activation-induced cytidine deaminase (45,69). The 40amino acid N terminus-deleted A2 that was crystallized was found to dimerize through two monomers pairing their ␤2 strands. Through sequence alignment of A2 and A3H and biochemical data from activation-induced cytidine deaminase (45,69), we generated an A3H mutant with amino acid substitutions R44A/Y46A located at this predicted oligomerization interface along the ␤2 strand (Fig. 5C) and determined the oligomerization state using size exclusion chromatography.
The A3H haplotype II predicted ␤2 strand mutant R44A/ Y46A was less stable than A3H haplotype II as evidenced by ϳ10-fold less recombinant protein being produced from the FIGURE 3. A3H haplotypes II and V demonstrate similar HIV⌬Vif restriction efficiencies. A, HIV ⌬Vif infectivity was measured by enhanced GFP expression in 293T cells infected with HIV⌬Vif that was produced in the absence or presence of untagged A3H haplotype II (Hap II) or A3H haplotype V (Hap V). A titration of A3H plasmid was used (25, 50, 100, or 200 ng). Results normalized to the no A3 condition are shown with the S.D. (error bars) calculated from at least three independent experiments. B, the level of HA-tagged A3H haplotype II and A3H haplotype V expressed in cells and encapsidated into HIV⌬Vif virions was determined using immunoblotting. The A3 enzymes were detected through the HA tag. The loading control for cell lysates was ␣-tubulin (␣-tub), and the loading control for virions was p24. A representative blot from three independent experiments is shown. C, proviral DNA sequencing data from a subset of clones (from 100 ng of A3H plasmid transfection) showing the sequence context of mutations induced by haplotype II or V in the protease region. Asterisks denote homology. Guanine (G) to adenine (A) mutations are shown in boldface type.

TABLE 4 Analysis of A3-induced mutagenesis of prot DNA from integrated HIVrVif
The proviral DNA analyzed was obtained from the transfection condition with 100 ng of A3H expression plasmid.  NOVEMBER 6, 2015 • VOLUME 290 • NUMBER 45

A3H Polymorphisms and Oligomerization Impact Processivity
Sf9 cell system (data not shown). Thus, the GST tag was not cleaved to extend the stability of the mutant for size exclusion chromatography because the R44A/Y46A mutant was prone to precipitation after extended incubation at 4°C in solution without GST. Only 10 g of GST-A3H wild type and mutant were applied to the column due to the lower yield of GST-A3H haplotype II R44A/Y46A mutant during purification, and the chromatograms were generated by manually quantifying integrated gel band intensities in each fraction from an SDS-polyacrylamide gel. Further, the GST-tagged wild type and mutants were analyzed on a 10-ml G200 Superdex column, in contrast to untagged A3H that was analyzed using a 25-ml G200 Superdex column (see "Experimental Procedures"). To ensure that the GST tag did not confound our interpretation of the results, the GST-tagged R44A/Y46A mutant was compared with GSTtagged A3H haplotype II during the size exclusion chromatography. Due to the 22-kDa GST tag, the predicted GST-A3H monomer molecular mass was 46 kDa. Consistent with ␤2-␤2 strand dimerization, the GST-A3H haplotype II R44A/Y46A mutant eluted predominantly as a monomer (apparent molecular mass of 37 kDa; Fig. 5, D-F). Although GST alone forms a dimer (apparent molecular mass of 51 kDa; Fig. 5G), the GST appears to be unable to dimerize when attached to the A3H protein (Fig. 5D); otherwise, we would be unable to resolve GST-A3H haplotype II R44A/Y46A at a molecular mass of 37 kDa. GST-A3H haplotype II resolved as a dimer (apparent molecular mass of 77 kDa; Fig. 5, D-F), consistent with the untagged A3H haplotype II data (Fig. 5A). There is a clear peak shift of the GST-A3H haplotype II and GST-A3H haplotype II R44A/Y46A in the SDS-PAGE analysis of the fractions, suggesting that GST-A3H haplotype II R44A/Y46A fully disrupted dimerization (Fig. 5, D and F). Nonetheless, in an attempt to examine the extent of the ␤2 strand contacts and whether GST-A3H haplotype II R44A/Y46A mutant fully disrupted dimerization, we also made a GST-A3H haplotype II P42S mutant, but this mutant was not stably produced from the Sf9 cell system. These data support the conclusion that A3H dimerization is mediated by multiple contacts along the predicted ␤2 strand and involves at least amino acid residues 44 RGY 46 . Further, these data demonstrated that the A3H haplotype II dimer was stable over a 15-fold concentration range (Fig. 5, compare A and D).
According to the A2 single Z-domain oligomerization model, a ␤2-␤2 mediated dimer can tetramerize via hydrophobic amino acids on loop 7 (45). A3H haplotype II formed tetramers in solution (Fig. 5A). To investigate the amino acids involved in A3H tetramerization, we made an A3H haplotype II loop 7 double mutant, Y112A/Y113A (Fig. 6A). During size exclusion chromatography of 150 g of A3H haplotype II Y112A/Y113A, predominantly dimers were formed (Fig. 6B, apparent molecular mass of 39 kDa), similar to wild type A3H haplotype II (Fig. 5A). In contrast to wild type A3H haplotype II, the Y112A/Y113A mutant formed fewer tetramers (Fig. 6B, peak denoted by T). At 300 g of A3H haplotype II Y112A/ Y113A, the mutant had no increase in tetramer formation, suggesting that Tyr-112 and Tyr-113 primarily contribute to tetramerization (Fig. 6B, inset graph). However, we could not test this hypothesis further because another loop 7 mutant, A3H haplotype II H114A/W115A, could not be stably produced from Sf9 cells, in agreement with the low cellular stability observed by others (36,38). Altogether, the data indicate that A3H haplotype II forms dimers through ␤2 strand interactions and tetramerizes through loop 7.
In solution, the tetramers were a small proportion of the total A3H (Fig. 5A). However, A3H haplotype II bound ssDNA cooperatively (Fig. 6C, Hill coefficient of 1.8), suggesting that tetramerization was promoted by A3H dimers binding ssDNA. In support of this hypothesis, the A3H haplotype II Y112A/ Y113A mutant bound ssDNA with an apparent K d similar to that of A3H haplotype II, but the saturation curve fit a noncooperative rectangular hyperbola (Fig. 6C). Together, these data indicate the formation of tetramers on ssDNA. A3H haplotype II Y112A/Y113A was not enzymatically active on several substrates with different deamination motifs (e.g. 5Ј-ATC, 5Ј-TTC, 5Ј-CTC, 5Ј-GTC, 5Ј-CCC, and 5Ј-AAC, in agreement with a previous study (36,38) and data not shown). Loop 7 is known to Processivity of A3H D178K was tested on ssDNA substrates that contained fluorescein-labeled deoxythymidine (yellow star) between two 5Ј-CTC deamination motifs separated by different distances. A, deamination of a 60-nt ssDNA substrate with deamination motifs spaced 5 nt apart. Single deaminations of the 5ЈC and 3ЈC are detected as the appearance of labeled 42-and 23-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 5-nt labeled fragment. B, deamination of a 118-nt ssDNA substrate with deaminated cytosines spaced 63 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 100-and 81-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 63-nt labeled fragment. For D178K, the D178K/ssDNA ratio of 1:2 was kept constant, but the total reaction components were increased as 50 nM/100 nM, 100 nM/200 nM, 150 nM/300 nM, 200 nM/400 nM, 250 nM/500 nM. These reaction conditions enabled the investigation of whether A3H haplotypes II and V could translocate on ssDNA by intersegment transfer. The measurements of enzyme processivity (processivity factor), S.D., and rate of deamination (percent/min) for each individual reaction condition are shown below the gel. All values are calculated from at least three independent experiments.
interact with ssDNA and mediate deamination motif preference in DNA binding models for multiple APOBEC deaminases (49, 70 -72). Because the A3H haplotype II Y112A/Y113A mutant bound ssDNA with an apparent K d similar to that of A3H haplotype II, the data indicated that the loop 7 mutant was probably inactive due to an altered conformation of ssDNA in the active site (Fig. 6C).
A3H haplotype II ␤2 Strand Mutant Exhibits Compromised Sliding and Intersegmental Transfer-We tested A3H haplotype II R44A/Y46A to determine whether this ␤2 strand mutant FIGURE 5. ␤2-␤2 strand amino acids mediate A3H dimerization. Size exclusion chromatography profiles of A3H haplotype II (Hap II) and A3H haplotype V (Hap V) obtained from a 25-ml G200 Superdex Increase column (A) were used to calculate the oligomerization states of the enzymes from a standard calibration curve (B). Analysis demonstrated that both A3H haplotype II and A3H haplotype V were able to form monomers (M), dimers (D), and tetramers (T) in solution. Both 150 and 300 g (inset graph) of enzyme were resolved to investigate whether A3H tetramer formation was concentration-dependent. According to the calibration curve, the apparent molecular masses of peak fractions for A3H haplotype II were 24 kDa (monomer), 44 kDa (dimer), and 94 kDa (tetramer), and for A3H haplotype V, they were 23 kDa (monomer), 44 kDa (dimer), and 102 kDa (tetramer). C, sequence alignment of A3H haplotype II and A2 ␤2 strand amino acid sequences. Amino acids that were mutated are indicated in red. The size exclusion chromatography profiles of GST-A3H haplotype II (GST-HapII) and GST-A3H haplotype II R44A/Y46A (GST-R44A/Y46A) obtained from a 10-ml G200 Superdex column (D) were used to calculate the oligomerization states of the enzymes from a standard calibration curve (E). When 10 g of enzyme was loaded onto the size exclusion column, GST-A3H haplotype II formed dimers in solution (apparent molecular mass of 77 kDa in peak fraction). This is in contrast to GST-A3H haplotype II R44A/Y46A, which formed monomers (apparent molecular mass of 37 kDa in peak fraction). F, the chromatograms from the 10-ml Sephadex 200 column (D) were constructed by analyzing the integrated gel band intensities of the protein in each fraction after resolution by SDS-PAGE. The gels show the peak fractions of GST-A3H haplotype II and GST-A3H haplotype II R44A/Y46A with start and end volumes corresponding to the fractions that were resolved by SDS-PAGE. G, the Sephadex 200 column (10-ml bed volume) demonstrates that GST is a dimer. The gel shows the peak fractions of GST with start and end volumes corresponding to the fractions that were resolved by SDS-PAGE. However, GST-tagged A3H does not dimerize through the GST tag based on data from GST-A3H haplotype II R44A/Y46A (D). AU, absorbance units. NOVEMBER 6, 2015 • VOLUME 290 • NUMBER 45 was still able to processively scan ssDNA. Interestingly, the A3H haplotype II R44A/Y46A mutant bound ssDNA with a cooperative binding curve (Fig. 7A, apparent K d of 0.21 M; Hill coefficient of 1.6), suggesting that the mutant can dimerize through loop 7 amino acids on ssDNA and does not remain a monomer as in solution (Fig. 5D). Because the Y112A/Y113A mutant was not catalytically active (data not shown), we were unable to test the deamination ability of an A3H R44A/Y46A/ Y112A/Y113A mutant. Nonetheless, testing the A3H haplotype II R44A/Y46A would reveal the function of the ␤2-␤2 strand dimer on catalytic activity and ssDNA scanning. We found that the A3H haplotype II R44A/Y46A could still slide to processively deaminate multiple cytosines but not as efficiently as A3H haplotype II. On an ssDNA substrate with cytosine separated by 5 nt, the A3H haplotype II R44A/Y46A had a processivity factor of 1, demonstrating that it was essentially not able to processively deaminate closely spaced cytosines (Fig.  7B), in contrast to A3H haplotype II, which had a processivity factor of 2.1 on this substrate (Fig. 1A). When the deamination motifs were spaced further apart, but still within a distance that requires sliding for processive deamination, the A3H haplotype II R44A/Y46A mutant was able to processively deaminate the two motifs (Fig. 7C, deamination motifs 14 nt apart, processivity factor of 2.1), but the sliding was not as efficient as A3H haplotype II (Fig. 1B, processivity factor of 3.6). These data indicated that although the A3H haplotype II R44A/Y46A mutant had a helix 6 that was able to mediate sliding, the sliding movement was compromised in the absence of the ␤2-␤2 strand dimer. The ␤2 strand dimerization may form a ssDNA binding groove that promotes processive sliding in combination with helix 6.

A3H Polymorphisms and Oligomerization Impact Processivity
To examine whether A3H haplotype II R44A/Y46A was still capable of undergoing intersegmental transfer, we used the assay that was used for A3H haplotypes II and V (Fig. 2). The processivity of A3H haplotype II R44A/Y46A did show a 2-fold decrease in the apparent processivity with an increased concentration of the reaction components (Fig. 7D), similar to A3H haplotype II ( Fig. 2A). However, the rate of the reaction did not increase as the reaction components increased (compare Figs. 2A and 7D). These data indicate that the A3H haplotype II R44A/Y46A could transfer to other ssDNA substrates, probably due to the dimerization mediated by loop 7 (Fig. 7A, cooperative binding curve), but did not readily cycle through the ssDNA substrates (Fig. 7D, no increase in reaction rate). This may be due to the loop 7 dimer being less stable than the ␤2 strand dimer, which is supported by the lack of a concentration dependence of tetramer formation in solution (Fig. 5A). However, the data suggested that A3H haplotype II R44A/Y46A mutant was able to undergo three-dimensional translocations because the processivity on the ssDNA substrate with the deamination motifs separated by 63 nt was similar to that of A3H haplotype II (compare processivity factors of 5.2 (Fig. 1C) and 4.6 ( Fig. 7D, first reaction lane)). The three-dimensional translocation ability of A3H haplotype II R44A/Y46A was confirmed by demonstrating that the mutant retained processivity in the presence of a dsDNA region between the deamination motifs (Fig. 7E).

Discussion
Human A3H is the most polymorphic human A3 enzyme with seven different haplotypes having different cellular stabilities (13,35,36). In comparison with other A3s, stable A3H FIGURE 6. Loop 7 amino acids Tyr-112 and Tyr-113 mediate A3H tetramerization. A, sequence alignment of A3H haplotype II (Hap II) and A2 loop 7 amino acid sequences. Amino acids that were mutated are indicated in red. B, the size exclusion chromatography profile of A3H haplotype II Y112A/ Y113A (Y112A/Y113A, 150 g) was used to calculate the oligomerization state of the enzyme from a standard calibration curve (Fig. 5B). Haplotype II Y112A/Y113A formed dimers (apparent molecular mass of 39 kDa in peak fraction). Overlay, haplotype II wild type (dashed line) is shown from Fig. 5A for comparison. A3H haplotype II Y112A/Y113A has less of a tetramer peak than haplotype II wild type. Inset graph, A3H haplotype II Y112A/Y113A did not resolve with a tetramer peak even when 300 g of enzyme was resolved on the size exclusion column. C, the apparent K d of A3H haplotype II and A3H haplotype II Y112A/Y113A from a 118-nt ssDNA was analyzed by steady-state rotational anisotropy. The A3H haplotype II bound ssDNA with a best least squares fit to a sigmoidal binding curve (Hill coefficient of 1.8) and an apparent K d of 0.72 Ϯ 0.14 M. The A3H haplotype II Y112A/Y113A mutant bound ssDNA with a best least squares fit to a rectangular hyperbola and an apparent haplotypes are most able to have a restrictive effect on the progression of an HIV infection in the early stages (37,39,61). This is due to A3H not being sensitive to all of the HIV Vif variants, unlike A3G and A3F (64,73,74). Because the majority of humans of non-African descent carry A3H alleles that code for an unstable form of A3H, not all HIV Vif variants are adapted to FIGURE 7. Analysis of A3H haplotype II ␤2 strand mutant R44A/Y46A ssDNA binding and scanning. A, A3H haplotype II (Hap II) R44A/Y46A ssDNA binding was analyzed by steady-state rotational anisotropy. The apparent dissociation constant (K d ) of A3H haplotype II R44A/Y46A from the 118-nt ssDNA was 0.21 Ϯ 0.04 M. The binding curve had a best least squares fit to a sigmoidal binding curve (Hill coefficient 1.6), demonstrating that the mutant could still oligomerize on ssDNA. Error bars, S.D. B-E, processivity of A3H haplotype II R44A/Y46A (R44A/Y46A) was tested on ssDNA substrates that contained fluorescein-labeled deoxythymidine (yellow star) between two 5Ј-CTC deamination motifs separated by different distances. B, deamination of a 60-nt ssDNA substrate with deamination motifs spaced 5 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 42-and 23-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 5-nt labeled fragment. C, deamination of a 69-nt ssDNA substrate with deamination motifs spaced 14 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 51-and 32-nt fragments, respectively; double deaminations of both C residues on the same molecule results in a 14-nt labeled fragment. D, deamination of a 118-nt ssDNA substrate with deamination motifs spaced 63 nt apart. Single deaminations of the 5Ј-C and 3Ј-C are detected as the appearance of labeled 100-and 81-nt fragments, respectively; double deamination of both C residues on the same molecule results in a 63-nt labeled fragment. For this assay, the R44A/Y46A/ssDNA ratio of 1:2 was kept constant, but the total reaction components were increased as 50 nM/100 nM, 100 nM/200 nM, 150 nM/300 nM, 200 nM/400 nM, 250 nM/500 nM. These reaction conditions enabled the investigation of whether A3H haplotype II R44A/Y46A could translocate on ssDNA by intersegment transfer. The reaction rate of deamination (percent/min) for each individual reaction condition is shown below the gel. E, deamination of the same substrate described for D but with a 20-nt DNA annealed between the two cytosines. For reactions performed on substrates described for B, C, and E, the A3H/ssDNA ratio was 1:2. The measurements of enzyme processivity (processivity factor) and the S.D. are shown below the gels. All values are calculated from at least three independent experiments. A white line between lanes in gel images indicates that an intervening lane(s) on the same gel was cropped for visualization. induce degradation of A3H if they are circulating in a geographical location where the majority of the population carries unstable A3H alleles (35)(36)(37)39). Thus, if an HIV strain from a human with inactive A3H alleles infects a human with a stable A3H allele, the Vif will have to adapt to degrade the stable A3H (37,39). It has been found that stable A3H can suppress HIV replication and delay time to treatment early in infection (37). Thus, stable A3H alleles could potentially act as an HIV infection barrier (39).
Despite the recognized importance of A3H in the A3 repertoire that can suppress HIV replication, there has been no biochemical characterization of the ssDNA scanning mechanisms of A3H, a biochemical comparison of stable haplotypes, or identification of the residues mediating oligomerization. Because oligomerization is required for virion encapsidation and enables A3H to be the only single Z-domain A3 capable of restricting HIV replication (3,44), understanding the interfaces where oligomerization occurs is important for realizing the determinants for HIV restriction. Thus, we focused our study on the two stable haplotypes, II and V, which occur with the highest frequency in humans (35,36). Despite differing by only a single amino acid, we found biochemical differences in haplotypes II and V. The more rigid side chain and lower isoelectric point of the A3H haplotype II Asp-178 in comparison with the A3H haplotype V Glu-178 appeared to affect the ssDNA scanning mechanisms and the binding affinity for dsDNA and RNA. Our study of A3H has established the basis of processivity for a single Z-domain A3 enzyme, that redundant processive mechanisms are used by A3H, a function of the A3H amino acid 178 polymorphism, and that helix 6 influences dsDNA and RNA binding affinity.
In general, an enzyme that moves processively on DNA using intersegmental transfer must have at least two DNA binding domains to mediate the doubly bound intermediate state where the enzyme simultaneously binds two segments of DNA before transferring to one of the bound segments (32,34). However, our data show that intersegmental transfer ability can be mediated by A3H at different levels. Both A3H haplotype V and A3H haplotype II R44A/Y46A have altered intersegmental transfer ability but for different reasons (Figs. 2B and 7D). The A3H haplotype II and V data (Fig. 2) demonstrate that Asp-178 on predicted helix 6 promotes enzyme cycling during intersegmental transfer. The absence of Asp-178, either by mutating A3H to a Lys-178 or studying A3H haplotype V that has a Glu-178 inhibited the increased reaction rate associated with intersegmental transfer despite the presence of two DNA binding domains (compare Figs. 2 (A and B), 4B, and 5A). This finding is consistent with the results of Mitra et al. (38), who found that helix 6 was important for A3H activity and ssDNA binding, although they did not test processivity. The polymorphism at amino acid position 178 has no other known function (41). The A3H haplotype II R44A/Y46A mutant demonstrated that the dimer interface is also important for promoting intersegmental transfer. Despite cooperative binding of A3H haplotype II R44A/Y46A to ssDNA, signifying the formation of oligomers, probably through loop 7, the mutant was unable to cycle efficiently after transfers to another DNA molecule (Fig. 7, A and  D). This may be due to either the loop 7-mediated dimer being less stable than a ␤2 strand-mediated dimer or the ␤2 strand dimer interface creating an ssDNA binding groove along the enzyme that promotes interaction of the ssDNA with helix 6. Intersegmental transfer ability is unique to A3H among the A3 enzymes characterized thus far (25). Although A3G has been characterized to undergo intersegmental transfer in one report (30), this has not been confirmed by others (25,75,76). Further, testing A3G for intersegmental transfer movement under the same reaction conditions as A3H showed that the processivity of A3G did not decrease with increasing concentration of reaction components (data not shown), demonstrating that A3G uses primarily jumping as a three-dimensional translocation mechanism (28). It is not known why A3 enzymes, such as A3G or A3F, do not readily undergo intersegmental transfer because there are multiple DNA binding domains within a monomeric unit (25,26,42). Based on our results with A3H, this may be due to differences in the ssDNA binding properties of helix 6. Similar to A3G and A3F, A3H also uses helix 6 to mediate sliding movements (25,26). Jumping movements of A3G are mediated by loop 7 (26); however, we could not test this with A3H because the loop 7 mutant was not catalytically active, in agreement with previous studies (36, 38) (data not shown).
A3H haplotype II was found to use several processive mechanisms to scan ssDNA. A3H haplotype II could slide on ssDNA, slide over RNA/DNA hybrids, and translocate by intersegmental transfer and jumping (Figs. 1 and 2A). Although A3H haplotype V exhibited diminished sliding and intersegmental transfer ability (Figs. 1 (A and B) and 2B), the enzyme was able to jump (Fig. 1, C and D) and restricted HIV replication similarly to A3H haplotype II (Fig. 3). Thus, the data indicate that for A3H, the three-dimensional translocation mechanisms of intersegmental transfer and jumping are redundant. A3H haplotype V could processively slide on ssDNA to deaminate cytosines that were 14 nt apart but not 5 nt apart (Fig. 1, A and B). However, A3H haplotype V-mediated mutagenesis of HIV was not adversely affected, despite a lack of a redundant mechanism for searching for closely spaced cytosines (Fig. 3). These data suggest that deficiencies in sliding are not as detrimental to the efficiency of the search for deamination motifs as deficiencies in jumping. This is supported by A3G reaction conditions and A3G mutants that removed either sliding or jumping ability (26,67). For A3G, larger decreases in mutagenic efficiency result from losses in jumping rather than sliding (26,67). Altogether, the A3H data are consistent with our previous work, which found that a high A3 mutagenic potential during HIV reverse transcription requires the enzyme to use both a one-dimensional and three-dimensional translocation mechanism (25). Our data are also in agreement with the findings of others who have shown A3H haplotype V to be as restrictive as A3H haplotype II in primary cells during a spreading infection and in single-cycle infectivity assays (36,39,64,65).
Our results indicate that A3H oligomerizes similarly to the model proposed for the N-terminal 40 amino acids-deleted A2 crystal structure (45). The A3H dimer is highly stable, and even at very low concentrations of A3H, the dimer interface was maintained (Fig. 5D, 10 g). The A2 model of oligomerization where a single Z-domain enzyme dimerizes through the ␤2 strands and tetramerizes through loop 7 has been questioned because this structure only forms in the N-terminal deleted version of A2, and human A2 in vitro activity has yet to be demonstrated (44,45,77). The A3H data demonstrate that the A2 model of oligomerization is a valid model for single Z-domain enzymes. A caveat of our data is that we could not produce a ␤2 strand mutant that was as stable as wild type A3H haplotype II. Thus, we required a GST tag to stabilize the ␤2 strand mutant. It is known that GST alone forms dimers (Fig. 5G), but our data demonstrate that GST-A3H does not form dimers through the GST tag, or we would be unable to resolve A3H haplotype II R44A/Y46A as a monomer (Fig. 5D). These data suggest that the A3H can block GST dimerization in our GST-A3H fusion construct. Nonetheless, this model of oligomerization does not appear to be conserved across all Z-type domains because multiple crystal structures of single Z-domain enzymes or C-terminal domains of double Z-domain enzymes have not oligomerized according to the A2 model (68, 78 -82). Because A3H is the only Z3-domain in humans, it may be the only A3 to oligomerize in this manner (42). A3H tetramers form primarily during DNA binding and not in solution. This is supported by the data for the A3H haplotype II ␤2 strand mutant R44A/Y46A that could still bind ssDNA cooperatively (Fig. 7A) and the A3H haplotype II loop 7 mutant Y112A/Y113A that could not (Fig.  6C). These data are also consistent with size exclusion chromatography profiles of A3H haplotypes II and V, where tetramers were a minority of the population in comparison with dimers (Fig. 5A). Mitra et al. (38) have investigated the role of loop 7 in the A3H-ssDNA interactions required for catalytic activity. Deamination reactions conducted with A3H haplotype II and mutants expressed from 293T cells demonstrated that Y112A and W115A mutants lose deamination activity (38). Interestingly, W115F rescued the deamination activity by 30%, suggesting that aromatic residues on loop 7 are involved in nucleic acid interactions that are required for catalysis through base stacking (38), in agreement with data from activation-induced cytidine deaminase (83). However, this does not affect the apparent K d from ssDNA (Fig. 6C).
The A3H dimer organization may provide an advantage by decreasing the Vif-mediated degradation efficiency and thus promoting the HIV restriction ability of A3H. We previously showed that A3G dimerization is disrupted by Vif (43). This may contribute to efficient Vif-induced polyubiquitination. Vif can disrupt A3G dimerization because the primary residue with which Vif interacts on A3G is Asp-128, which is adjacent to the primary dimerization amino acids, Phe-126 and Trp-127 (18, 43, 46, 84 -86). In contrast, when A3H haplotype II interacted with Vif, there was a structural change induced by Vif, but this did not appear to be a complete disruption of dimerization (43). Our data on the A3H dimer interface suggest that this is because A3H dimerizes through the ␤2 strands (Fig. 5D) and Vif interacts with A3H on a separate structure, predicted helix 4 (87,88). Using quantitative immunoblotting, it was shown that Vif HXB2 is 4-fold less efficient at degrading A3H than A3G. Further, both Vif NL4-3 and Vif HXB2 could induce polyubiquitination of A3H with Lys-63 and Lys-48 linkages, in contrast to A3G that was polyubiquitinated with only Lys-48 linkages (43). Thus, the A3H ␤2 strand dimer may decrease the effectiveness of Vif and promote alternate ubiquitination modifications.
Although during viral infections, HIV Vif can effectively suppress both A3G and A3H to enable viral replication (74), the higher resistance of A3H to Vif may provide an advantage during the early stages of infection and during Vif adaptation (37,39).