To express two A3 transcripts on a single-cell basis, we used an expression plasmid, pVIVO2 (Invivogen), with two transcription units in a single vector. Constructs contained one HA-tagged A3 or one V5-tagged A3 or both. All V5-tagged A3s (A3F and A3G) were cloned using an XbaI site in MCS1. The cDNA for A3F and A3G were amplified from pcDNA constructs and a previously established cloning strategy was used to obtain V5-tagged versions of these cDNA in pVIVO2 (22, 26, 28, 29). The HA-tagged A3D was subcloned from a pcDNA construct into the MCS2 of pVIVO2 using NheI (30). The following reagents were obtained through the NIH HIV Reagent Program, Division of AIDS, NIAID, NIH: pcDNA3.1-APOBEC3G-HA Expressing Human APOBEC3G with C-Terminal Triple HA Tag, ARP-9952, contributed by Dr. Warner C. Greene and pcDNA3.1-APOBEC3DE-V5-6xHIS, ARP-11433, contributed by Dr. Yong-Hui Zheng.

Whole blood was collected from HIV- donors and Primary Blood Mononuclear Cells (PBMCs) were isolated. Ethics approval was obtained by the University of Saskatchewan Biomedical Research Ethics Board (Ethics Number 14-62). PBMCs were isolated using SepMate tubes (Stemcell Technologies), following the Ficoll-Paque method. Primary cells were cultured in RPMI supplemented with 10% Fetal Bovine Serum, Penicillin/Streptomycin, 5 mM HEPES, non-essential amino acids, sodium pyruvate, 50µM β-mercaptoethanol and stimulated to proliferate with 50 U/ml recombinant interleukin-2 (Sigma) and 10 µg/ml phytohemagglutinin (Fisher Scientific) for 72 h. The RNA preparation, cDNA synthesis and qPCR were carried out according to Refsland et al., except that a SYBR Green based qPCR method rather than the probe and primer method was used. The primers and qPCR cycle used for A3D, A3F, A3G, A3H and TBP were as reported in Refsland et al. (31).

To produce virus particles, 1 × 10⁵ 293T cells per well of a 12-well plate were co-transfected with 500 ng of HIV-1 LAI ΔEnv ΔVif (HIV -Vif) or HIV-1 LAI ΔEnv (HIV +Vif), 180 ng of VSV-G, and pVIVO2 plasmid (empty or A3 expressing) at 0 ng, 25 ng and 50 ng (A3F-V5, A3D-HA), or 50 ng and 100 ng (A3F-V5/A3D-HA). GeneJuice (Novagen/EMD Millipore) transfection reagent was used according to the manufacturer’s instructions. Forty-eight hours post-transfection virus containing supernatants were harvested and filtered through a 0.45 µm polyvinylidene difluoride syringe filter. Filtered virus was used for both infection of TZM-bl (HeLa CD4 + CCR5 + LRT lacZ cells) reporter cells and concentration and lysis for use in immunoblots. Virus producing cells were collected for use in immunoblots. To measure the infectivity, 1 × 10⁴ TZM-bl cells per well of a 96-well plate were infected with a dilution series of filtered virus in presence of 8 µg/mL of polybrene. Then, 48 h post-infection, the infectivity was measured through colorimetric detection using β-galactosidase assay reagent (Pierce). The infectivity of each virus was compared to the No A3 condition that was set to 100%.

Immunoblotting was conducted to measure the A3 encapsidation in virions and steady state protein levels in cell lysates. Rabbit anti-HA (1:4000, Sigma) and Mouse anti-V5 (1:1000, Sigma) were used for detection of HA-tagged A3D and V5-tagged A3F and A3G respectively. Rabbit anti-α-tubulin (1:5000, Sigma) and Mouse anti-p24 (1:1000, Cat#ARP-3537, NIH AIDS Reagent Program) were used as the loading controls for cell lysates and viral lysates, respectively. Secondary detection was performed using Licor IRDye antibodies produced in goat (1;10000, IRDye 680 labeled anti-Rabbit, and IRDye 800 labeled anti-Mouse). Blots were scanned via LICOR CLx. For quantification, Image Studio was used to detect the pixel intensity of the experimental and loading control bands. Each sample was normalized to its own loading control before comparison to other lanes. Cropped blots are shown in the manuscript for clarity. Original blot images are in Supplementary File 1 .

Relative levels of integrated proviral DNA was determined as previously described (32–34). In brief, 1 × 10⁵ 293T cells per well of a 12-well plate were infected by spinoculation (1 h at 800×g) in the presence of polybrene (8 µg/mL) with HIV produced from the single-cycle replication assays. DNA was extracted after 24 h using DNazol according to the manufacturer’s instructions (Invitrogen). The DNA was then treated with DpnI and integrated pro-viral DNA was amplified using HIV-specific primers with Alu-specific primers to ensure only integrated and not episomal HIV proviral DNA was amplified (32–34). This PCR product was diluted and used for qPCR to determine the amount of integrated proviral DNA relative to RNaseP (32–34).

For proviral sequencing, 1 × 10⁵ 293T cells per well of a 12-well plate were infected by spinoculation (1 h at 800×g) in the presence of polybrene (8 µg/mL) with HIV produced from the single-cycle replication assays. DNA was extracted after 48 h using DNazol according to manufacturer’s instructions (Invitrogen). The PCR amplification of a 582 bp region of the polymerase (pol) region of HIV and treatment of DNA with DpnI was carried out as previously described (32, 35). The PCR product was cloned using CloneJET PCR cloning kit (Thermo Scientific) and DNA sent for sequencing. Sequences were analyzed with Clustal Omega (36) and Hypermut (37). The number of clones analyzed for each condition was: No A3: 5; A3F: 6; A3D: 5; A3F/A3D: 7.

The 293T cells (2.5 × 10⁶ per T75 flask) were transfected with 2 µg of DNA. The transfections consisted of 1µg of pVIVO2 A3D-HA expression plasmid and 1µg of A3F-V5, or A3G-V5 expression plasmid. GeneJuice (Novagen/EMD Millipore) transfection reagent was used according to the manufacturer’s instructions. After 40 h of transfection, cells were washed with cold PBS, lysed with cold immunoprecipitation buffer (50 mM Tris-Cl [pH 7.4], 1% Nonident-P40, 0.1% sodium deoxycholate, 150 mM NaCl, 10% glycerol, complete, EDTA-free Protease Inhibitor Cocktail (Roche)) and clarified by centrifugation at 4°C. Clarified supernatants were split in two and treated or not treated with RNase A (0.1 mg/mL, Roche) and polyclonal anti-HA rabbit antibody (4 µL, Sigma, St. Louis, MO, USA) at 4°C during incubation with the lysate, and protein A/G agarose (Santa Cruz Biotechnology) for 90 min. The mock condition had no anti-HA antibody. The agarose resin was then washed with immunoprecipitation buffer three times before resuspending in 2× Laemmli buffer and then resolving on SDS-PAGE and transferring to nitrocellulose for immunoblotting. The blot then was probed with anti-V5 mouse antibody (1:1000, Sigma) for detecting HA-immunoprecipitated lysates. For detecting the expression of A3s in cell lysate, the nitrocellulose membrane was probed with anti-HA mouse (1:4000, Sigma), anti-V5 mouse (1:1000, Sigma) and anti-α-tubulin rabbit (1:5000, Sigma). Secondary detection was performed using Licor IRDye antibodies produced in goat (1:10000, IRDye 680 labeled anti-Rabbit, and IRDye 800 labeled anti-Mouse).

The DNA sequences for the HIV 5’UTR (1-497 nt in HIV genome) and a portion of the protease gene (2282-2402 in HIV genome) from the HIV-1 clone 93th253.3 (GenBank accession number U51189) were cloned in the pSP72 plasmid downstream of a T7 promotor using EcoRI and BglII restriction sites. Purified pSP72 plasmid was used as the template to produce 5’UTR and protease RNA in vitro according to the instructions for the mMessage mMachine T7 Ultra Kit (Life Technologies), but with the following modifications. The 5’UTR reaction was incubated overnight. The protease reaction was incubated for 4 h. To enhance the downstream reactions, the optional step of 3′ poly-A-tail was conducted, but used half the recommended E-PAP enzyme and a 30 min incubation. The RNA was then ethanol-precipitated and resuspended in nuclease-free water. For biotinylation, 5 µM RNA, 1 mM ATP, 1% (v/v) DMSO, 5 µM pCp-Biotin (cytidine 5′phosphate 3’-(6-aminohexyl), MJS BioLynx), 100 units of T4 RNA ligase 1 (NEB), and 1x NEB reaction buffer, were incubated at 37°C for 60 min. To remove free pCp-Biotin, the reaction was applied to a Micro Bio-Spin 6 size exclusion column (Bio-Rad) and the flow through collected.

The 293T cells (2.5 × 10⁶ per T75 flask) were transfected with 1 µg of pVIVO2 A3D-HA, A3G-V5, A3F-V5/A3D-HA, or A3G-V5/A3D-HA or 0.5 µg of A3F-V5 expression plasmid. GeneJuice (Novagen/EMD Millipore) transfection reagent was used according to the manufacturer’s instructions. After 40 h of transfection, cells were washed with cold PBS and lysed with cold immunoprecipitation buffer. S-protein agarose (EMD Millipore) was incubated with 400 µg of cell lysate and 500 µg biotinylated RNA for 90 min at 4°C. The S-protein agarose was washed three times with PBS before elution in 1× Laemmli buffer, resolution by SDS-PAGE, and transferring to nitrocellulose for immunoblotting. The blot then was probed with anti-V5 mouse antibody (1:1000, Sigma) and anti-HA rabbit antibody (1:5000, Sigma). For detecting the expression of A3s in cell lysate, the nitrocellulose membrane was probed with anti-HA rabbit, anti-V5 mouse and anti-α-tubulin rabbit (1:5000, Sigma). Secondary detection was performed using Licor IRDye antibodies produced in goat (1:10000, IRDye 680 labeled anti-Rabbit, and IRDye 800 labeled anti-Mouse).

To determine if A3D could contribute to restriction of HIV, we first determined the levels of A3D mRNA relative to A3F, A3G, and A3H in PBMCs of HIV- donors ( Figure 1 ). Using TBP mRNA to normalize the qPCR data, we observed that the A3 transcript levels between donors varied by an order of magnitude. Additionally, from each donor, the A3 transcript levels varied widely. Donor 10 showed similar levels of A3D and A3G mRNA, but had more than 100-fold less of A3F and A3H mRNA ( Figure 1A ). This was similar to Donor 17, except that A3D mRNA was approximately 2-fold more than A3G mRNA ( Figure 1B ). Donor 18 showed the lowest level of A3D mRNA ( Figure 1C ). Overall, A3F mRNA was always lower than A3G mRNA, similar to what was reported by Refsland et al. (31). However, in the donors tested we found that A3D mRNA levels can be as high as A3G mRNA, which has not been reported previously. Refsland et al. determined that there was good correlation between mRNA and protein levels for A3G and A3F, but for other A3s suitable native antibodies precluded this type of analysis (31). Although A3D catalytic activity has previously been reported to be low (38, 39), it may restrict HIV by deamination independent mechanisms when these high mRNA expression level occur, assuming this correlates with protein expression as has been found for A3G and A3F (31). In addition, hetero-oligomerization of A3D with highly related family member A3F may increase HIV restriction activity, similar to what has been reported for A3F and A3G (27, 35). That A3D mRNA levels were higher or similar to A3F mRNA suggest that the proteins when co-expressed could potentially influence each other’s restriction activity.

We examined the ability of A3D to become encapsidated into and restrict HIV -Vif and +Vif virions ( Figure 2 ). We compared the A3D restriction ability to A3F, which is similar to A3D in domain structure (40). We used two plasmid transfection amounts to observe a dose dependent effect and ensure we were not saturating the experimental system with overexpression of A3 enzymes ( Figures 2C, D ).We also used a vector with two MCS (pVIVO2) to express A3D-HA and A3F-V5 within the same cell (see Materials and methods). To be able to detect A3F and A3D that resolve by SDS-PAGE to the same apparent molecular weight, we used distinct tags (V5 or HA). These tags enable comparisons to be made between conditions for A3F only and A3D only, but not to directly compare A3F to A3D because the immunogenicity of the V5 and HA tags may differ.

Both A3F and A3D alone could restrict HIV -Vif, but A3D was approximately 2-fold more efficient at the low and high plasmid transfection amounts ( Figure 2A ). When A3D and A3F were co-expressed, we observed restriction of HIV -Vif, but the restriction amount was only 1.5-fold greater (low) or equal to (high) A3D alone ( Figure 2A ). To investigate why there was no cooperative effect, we used immunoblotting to determine the A3D and A3F protein levels in virus producing cells and virions. Using different tags, we could detect A3F-V5 and A3D-HA when expressed together. We observed that the expression levels in cells for each, A3F or A3D, were similar when they were expressed alone or together ( Figure 2C , Cells). However, the immunoblot showed that encapsidation of A3F and A3D was different ( Figure 2C , Virus). In particular, when A3D and A3F were co-expressed, we could not detect any A3F encapsidation by immunoblotting and the A3D encapsidation levels were similar to when A3D was expressed alone ( Figure 2C , Virus). These data suggest a competitive encapsidation process, where A3D excludes A3F from encapsidation into virions.

We also examined if this occurred during an HIV +Vif infection. The A3F and A3D alone could not efficiently restrict HIV +Vif, consistent with each A3 being sensitive to Vif-mediated degradation and only detectable on immunoblots at the high transfection level ( Figures 2B, D ). Accordingly, when A3D and A3F were each expressed alone we could detect encapsidation at the higher transfection level ( Figure 2D ). However, when A3D and A3F were co-expressed, both A3s were less sensitive to Vif-mediated degradation and could be detected in cell lysates at low and high transfection amounts ( Figure 2D ). In particular, there was 23-fold more A3D in cells when co-expressed with A3F, than in its absence ( Figure 2D , Cells, High). This correlated with 1.5-fold more encapsidation of A3D in virions in the when co-expressed with A3F, than in its absence ( Figure 2D , Virus, High). This potentially explains how A3D was able to partially restrict virus replication in the presence of A3F, but not its absence ( Figure 2B , High). Notably, we could not detect A3F in virions when co-expressed with A3D, despite more 3-fold more A3F in cell lysates when expressed with A3D compared to A3F expression alone ( Figure 2D , High). However, we cannot exclude that A3F becomes encapsidated into virions, but at a level below the detection limit of the immunoblotting. Overall, these data support the HIV -Vif results and further suggest that A3D competes with A3F for encapsidation into virions, even in the presence of Vif. This hypothesis would require that A3D bound more tightly to the HIV genomic RNA than A3F.

Deamination-independent mechanisms of A3s typically rely on RNA binding (34). A3 enzymes can bind HIV genomic RNA and inhibit the progression of reverse transcriptase, which results in less proviral DNA integration. We observed that A3D inhibited proviral integration 13.5-fold more than A3F ( Figure 3A ). The A3F/A3D co-expression inhibited proviral integration similar to A3D alone ( Figure 3A ). To exclude another mechanism of inhibiting proviral DNA integration through host-mediated degradation of proviral genomes after excision of uracils, we also determined the level of A3-induced mutations in a 582 bp region of the pol gene ( Figure 3B ). The A3-induced mutations are analyzed in the (+) DNA where uracils cause G→A mutations, primarily in a GA→AA context for A3F and A3D has a more broad mutation potential that includes GA→AA, GG→AG, and GC→AC (12, 41). Consistent with past studies, we observed that A3D was barely able to cause mutations above the reverse transcriptase background (no A3) ( Figure 3B , Supplementary Figure 1 ). In contrast, A3F was able to cause 26 GA→AA mutations per kb ( Figure 3B , Supplementary Figure 2 ). The A3F/A3D condition had approximately 10-fold fewer mutations than A3F alone, consistent with a decrease in, but not complete exclusion of A3F from virion encapsidation ( Figures 2C , 3B ; Supplementary Figure 3 ). Collectively, these data suggest that uracil-based degradation of proviral DNA and RNA binding cause the A3F-induced decrease of proviral DNA integration, whereas A3D decrease of proviral DNA integration is through only RNA binding ( Figures 3A, B ). This is consistent with a previous study that identified human A3D having lowered catalytic activity due to a Y320C change from chimpanzee A3D (17). These data support that A3D binds RNA more tightly than A3F and this could be a reason why A3D competes A3F off HIV RNA during encapsidation.

To determine if in addition or alternatively A3D directly interacts with A3F we co-expressed A3D and A3F and used co-immunoprecipitation (co-IP) in the absence or presence of RNAse A to determine if they interacted through a protein-protein interaction, such as hetero-oligomerization, or through an RNA intermediate. The co-IP showed that A3F most strongly immunoprecipitated when RNAse A was not added, demonstrating that their interaction is primarily through an RNA intermediate ( Figure 4A ). To determine if this was a common interaction mechanism, we also tested A3G. During the co-IP with A3D, A3G was more sensitive to the RNAse A treatment than A3F, and we observed no co-IP of A3G with A3D unless the cellular RNA was intact ( Figure 4B ). These data demonstrate that the interaction of A3D with A3F and A3G is RNA mediated.

Since the co-IP is performed in the presence of an abundance of cellular RNAs with different sequence specificities, we performed an additional experiment to directly assess if A3D is competitively excluding A3F from HIV RNA during encapsidation into virions. A3F and A3G have been shown to bind multiple diverse RNAs during encapsidation into HIV, including the HIV genomic RNA (42, 43). We used in vitro transcription to produce two regions of the HIV genomic RNA, the HIV 5’UTR (497 nt) and protease (120 nt), and labeled the RNAs with biotin to enable immobilization onto streptavidin-agarose resin. Then we added cell lysates that had A3D-HA, A3F-V5, A3G-V5, A3D-HA and A3F-V5, or A3D-HA and A3G-V5 ( Figure 5 ). This streptavidin capture experiment showed that both A3D and A3F alone bound to the HIV 5’UTR RNA, but when the A3D and A3F were present together, the A3D maintained binding to the HIV 5’UTR RNA and A3F was prevented from binding ( Figure 5 ). Similar results were found with A3G, however, A3G bound less well to the HIV 5’UTR than A3F and A3D alone ( Figure 5 ). To determine if this was specific to the HIV 5’UTR RNA, we also tested a portion of the HIV protease ( Figure 5 ) We observed similar results, although A3D was not able to completely exclude A3F and A3G from binding ( Figure 5 ). These data in combination with the co-IP of A3D and A3F as an RNA-mediated complex from cell lysates ( Figure 4A ) support the hypothesis that A3D excludes A3F from encapsidation into virions by competitive inhibition of A3F RNA binding.