dUTP/TTP in MDM and MC is high but variable with blood donors. In previous studies dUTP/dTTP ratios in the range 20:1 to 60:1 were reported [4,12,24], while the ratio was essentially zero in dividing cells due to the expression of dUTPase in replicating cells [4]. These different estimates of intracellular dUTP/dTTP may arise from the combined challenges of measuring low levels of dNTPs present in non-dividing cells, the inherent difficulty in accurately determining the ratio when one of the nucleotides is present at an extremely low level, intrinsic differences in dNTP pools in different donors, or differences in culture conditions or measurement methods.
We first extended the dTTP and dUTP measurements to freshly isolated monocytes with the expectation that the levels in freshly isolated cells would most closely match in vivo levels. A single nucleotide polymerase extension (SNE) assay was used where a 22 mer DNA primer-template containing a single adenine overhang on the template strand was extended by the dUTP or dTTP present in cell extracts (Fig. 2a)[3]. The assay takes advantage of the enzyme dUTPase to remove dUTP in the extract and allow measurement of dTTP in isolation (see methods). For these assessments, the [dTTP + dUTP] present in MC and MDM extracts were compared with the Hap1 dividing cell line (Fig. 2b and Supplemental Figure S1). We found similarly low levels of [dTTP + dUTP] in freshly isolated MC extracts as compared to MDM that had been differentiated over seven days in the presence of M-CSF (1.1 to 2.5 pmol dU(T)TP/million cells). For comparison, the combined [dTTP + dUTP] level in the Hap1 cell line is 25 to 70-fold higher (~70 pmol dU(T)TP/million cells). Depleting dUTP in these dNTP extracts by the addition of dUTPase allowed calculation of dUTP/dTTP in each cell type (Fig. 2c). Although freshly isolated MC and differentiated MDM showed similar elevated dUTP/dTTP ratios in the range 1:4 to 1:1 depending on the donor, the ratio was essentially zero for the HAP1 cells. Based on these and previous studies, we conclude that significantly elevated dUTP/dTTP is an intrinsic aspect of MC and MDM metabolism that differs from dividing cells.
Although a range of dUTP/dTTP values have been reported for MDM [4,24], all of the measurements are consistent with low dNTP pools and significant levels of dUTP for both MC and MDM. The current measurements indicate that uracilated viral cDNA will be produced during infection of both MC and MDM and that the UBER capacity of these cells could impact the outcome of infection. However, unlike previous studies where dUTP/dTTP > 20, the more or less balanced ratio (~1) indicates that viral dUMP residues have the potential for being replaced by dTMP after multiple repair cycles (i.e. each replacement attempt has a 50:50 chance of replacing dUMP with dTMP, but eventually all dUMP would be repaired).
To evaluate whether HIV RTase can discriminate between dUTP and dTTP, and therefore bias dUMP incorporation away from the level expected from the dUTP/dTTP ratio, we measured the RTase activity in vitro using both dUTP and dTTP as substrates (Fig. 2d). Using a low dNTP concentration that approximated that calculated for non-dividing cells, we were unable to detect any selectivity of RTase for either nucleotide. Thus, the relative amounts of dUMP and dTMP in HIV reverse transcripts should reflect the cellular dUTP/dTTP.
MC and MDM have different uracil base excision repair (UBER) activities. Given that RTase shows no discrimination between dUTP and dTTP, the observed ratios of dUTP/dTTP of 0.25 to 1 for MDM and MC indicate that 99.9 % of HIV DNA products would have at least one dUMP incorporation for every turn of the DNA helix assuming a random sequence containing 50% A/T base pairs. These putative densely spaced uracils would be subject to excision by the UBER pathway. Accordingly, we were interested in the relative levels of seven enzymes involved in deoxyuridine metabolism in MC and MDM as compared to the Hap1 dividing cell line (Fig. 3). These enzymes included uracil DNA glycosylase (hUNG), AP endonuclease 1 (APE1), DNA polymerase β (pol β), ligase IIIa (the ligase isoform expressed in non-dividing cells)[25], the dNTPase sterile alpha motif histidine-aspartate domain protein 1 (SAMHD1), and the DNA cytidine deaminases APOBEC3A (A3A) and APOBEC3G (A3G). The western blots and activity measurements for extracts collected from uninfected cells and MC and MDM that were infected with the CCR5 tropic HIV-1BaL viral strain revealed the following general trends (Fig. 3a-3i and summary graphs in Fig. 3m-3s). First, MC and MDM show similarly low levels of expression of the first enzyme in the UBER pathway (hUNG), which are about 25 to 50-fold lower than the Hap1 reference line. These low hUNG activity levels, which could result from either the mitochondrial (hUNG1) or nuclear isoforms (UNG2), indicate that viral uracils may not be efficiently excised once viral DNA enters the nuclear compartment. Surprisingly, the activity of the next UBER enzyme, APE1, is highly expressed in both MDM and Hap1 cells, indicating that the rate-limiting step for repair of uracil in non-dividing cells is likely the initial excision event by hUNG2. The next two enzymes in the pathway, pol β and lig IIIα, are almost undetectable in MC but much more prevalent in MDM, although not to the same level as the Hap1 line. This important distinction between MC and MDM suggests that uracilated viruses generated by direct infection of MC cannot be repaired until the MC differentiate into MDM. As expected, SAMHD1 dNTPase is highly expressed in both MC and MDM, but not Hap1 cells, consistent with the different dNTP pool levels present in these cell types (Fig. 2b). Both deaminase enzymes, A3G and A3A, are highly expressed in MC, but not MDM or Hap1 cells. Despite the high expression levels of A3G and A3A, the high dUMP content of viral DNA in MC is not derived from intrinsic A3A or A3G activity—at least not with HIV-1BaL virus that encodes viral infectivity factor (vif)(see below). None of the expression data were noticeably affected by whether the extracts were prepared before or after BaL virus infection (Fig. 3). We performed additional quantitative RT-qPCR measurements of mRNA expression levels of these genes which were fully consistent with the above trends (Supplemental Figure S2).
HIV Infection of MC and MDM at early and late stages of differentiation. To explore the origins and possible fate of infected circulating monocytes observed in HIV patients on ART, we performed HIV infections using freshly isolated, undifferentiated MC rather than fully differentiated MDM. We began this exploration by measuring the levels of dUMP present in proviral DNA under three distinct conditions (i) MC that were infected with HIV-1BaL immediately after isolation and then maintained as undifferentiated monocytes for 7 days by culturing under nonadherent conditions in the absence of M-CSF, (ii) MDM infected at the monocyte stage and then allowed to differentiate in the presence of M-CSF (Early Infection, MDMEI) and, (iii) MDM that were infected after seven days of differentiation in M-CSF (Late Infection, MDMLI) (Fig. 4a). For comparison, we infected activated T cells using the same virus stock and MOI. Proviral dUMP levels were determined using the alu-gag Ex-qPCR experiment after isolating genomic DNA at 1, 3, 7, 14- and 28-days post infection (dpi). To prevent multiple round infection, 10 µM enfuvirtide (T-20) was added one day after the initial infection (Fig. 4a). The Ex-qPCR analysis determines the fraction of integrated HIV viruses in a DNA sample that contain one or more dUMP residues on each DNA strand of a DNA amplicon that contains a 650 bp region of the 5’ LTR and a 700 bp portion of the gag gene [4,26]. For MC, which were cultured using non-adherent conditions in the absence of M-CSF to maintain a monocyte-like phenotype (Supplemental Fig. S3), no proviral DNA was detected at 1 dpi, but by day three nearly 100% of the proviral copies contained dUMP, which persisted until the end of the experiment (7 dpi)(Fig. 4b). Due to limitations in the length of time MC can be maintained using nonadherent conditions, this experiment could not be continued beyond seven days. For MDMEI, which were infected immediately after isolation of MC and then cultured using adherent conditions in the presence of M-CSF to immediately begin their differentiation into MDM, over 50% of the provirus contained dUMP at 1 dpi, which increased to almost 100% at 3 dpi. Unlike MC grown under non-adherent conditions in the absence of M-CSF, the copies of HIV in MDMEI that contained dUMP decreased by 7 dpi (Frac U ~ 75%) and even further at 14 dpi (Frac U ~ 20%). For MDMLI, which were infected after complete differentiation, the proviral copies that contained dUMP were initially lower than MC or MDMEI (Frac U ~ 15% at 1 dpi and 60% at 7 dpi), but ended up at the same level as MDMEI at 14 dpi (Frac U ~ 20%). We attribute the lower Frac U values of differentiated MDM at early times after infection to the greater rate of reverse transcription in a small sub-population of permissive MDM (<10%) which have higher dNTP pools and low dUTP [4]. As the post-infection time increases, the slower replicating uracilated DNA products in the major MDM population increasingly contributes to the bulk measurement. The control T cells showed no viral-associated dUMP at any time during infection, consistent with our previous finding (4).
We then investigated the kinetics for appearance of early and late reverse transcripts (ERT, LRT) and proviral DNA using the three infection conditions. The kinetics for forming ERT and LRT products followed the trend MDMLI > MDMEI >> MC (Fig. 4c, d), with ERT and LRT copy numbers for MDMEI and MC about 2 and 6-fold lower at 7 dpi as compared to MDMLI. A similar trend was observed for the proviral copy numbers (Fig. 4e). The control infections of activated T cells showed both faster reverse transcription kinetics (~70% complete in ~1 day, Fig. 4c, d, e) and a greater number of integrated proviruses.
dUMP in proviral DNA does not arise from cytidine deaminase activity in MC. Given the presence of APOBEC enzymes in MC, and to a lesser extent MDM as judged by immunoblotting (Fig. 3a, g), we wanted to confirm that most of the dUMP in HIV proviral DNA was derived from incorporation of dUMP by RTase. One established way to test this is to add high levels of thymidine (dThyd) to the cell culture media prior to infection to increase the levels of intracellular TTP and then look for a reduction in the number of viral DNA products that contain dUMP [4]. We infected both MC and fully differentiated MDM in the presence of 5 mM dThyd and measured the copies of uracilated proviruses at 7 dpi using the Alu-gag Ex-qPCR method (Fig. 5). For both cell types we observed a 4-8-fold reduction in uracilated viral copies, indicating a predominant role for dUMP incorporation. A further reduction in uracilated viral copies is not expected due to the minor population of highly permissive MDM that do not contain high dUTP levels [4,27]. A minor role for APOBEC DNA cytidine deamination is confirmed by our proviral DNA sequencing results reported below, which show a low frequency of GàA transition mutations on the proviral (+) strand DNA at known APOBEC hotspots.
Effect of viral dUMP on proviral DNA and extracellular RNA sequences. Although dUMP incorporation is not expected to introduce mutations because its Watson-Crick hydrogen bond donor and acceptor groups are identical with thymidine, dUMP/A base pairs have reduced thermodynamic stability in the context of B DNA [28], increased base pair dynamics [29], and dUMP has an increased propensity to form dUMP/G wobble mismatches relative to thymine due to its larger keto-enol tautomerization constant. The reduction in duplex stability arising from dUMP has been attributed to the lower electron density of the pyrimidine ring system of uracil relative to thymidine, which weakens base stacking [29]. Thus, we were interested if any of these potential effects of dUMP incorporation could be detected in cDNA sequences produced from reverse transcriptase or viral RNA genomes produced by RNA pol II transcription.
To investigate proviral DNA sequences produced by reverse transcription in both MC and fully differentiated MDM, we used HIV-1BaL virus to infect freshly isolated MC cultured under non-adherent conditions and also MDM after 7 days of culturing. Both cell types were cultured for seven days after infection and total DNA was isolated and diluted to the single copy level. Single
proviral clones were then amplified using ES7 and ES8 env primers (see Supplemental Table S1). Wells that were positive for HIV DNA clones by qPCR were reamplified using primers that generated a 592 bp amplicon covering the V3 and V4 regions of env, followed by sequencing using the Sanger method. The viral sequences were compared with the lab reference sequence to determine the mutation frequencies and types. For proviral sequences isolated from MC and MDM, the viral mutation frequency (~1.5 x 10-3) and mutational spectrum were similar (Table 1, Fig. 6a, b). For both cell types, 26 to 38% of the isolated proviral clones contained substitution mutations, with the majority appearing as transition mutations (~70-80%), and the remaining being more unusual transversion mutations (20% to 30%). Two (+) strand GàA mutations detected in both MDM and MC infections might be attributed to enzymatic cytosine deamination on the viral (-) strand cDNA based on the sequence preferences of A3A or A3G (Supplementary Table S2). However, for MDM we cannot exclude that these apparent enzyme derived mutations arose from chance misincorporation because four GàA mutations occurred on the viral (-) strand cDNA (corresponding to CàT on the positive strand). These mutations in MDM cannot be attributed to APOBEC activity because of their sequence context and occurrence on the (-) strand (Supplementary Table S2). The two (+) strand GàA mutations observed in the MC derived samples could arise from the high A3A activity in these cells. Finally, 40% and 55% of the proviral mutations in MDM and MC led to codon changes and could therefore affect viral fitness (Supplementary Table S2, S3).
Since the mutational spectrum for integrated virus was indistinguishable for MC and MDM, we chose to selectively sequence extracellular viral RNA produced from HIV-1BaL-infected MDM (Fig. 6c). The viral RNA sequences showed a slightly elevated mutation frequency compared the proviral DNA (1.7 x10-3) and similar percentages of transition and transversion mutations. In addition, about 65% of the viral RNA mutations led to codon changes (Supplementary Table S5). We recently reported a mutational frequency of 0.6 x 10-3 for RNA pol II transcription using linear uracilated DNA templates that were transfected into human UBER deficient Hap1 cells [9]. In this study, the 2.3-fold higher mutation frequency arising from reverse transcription obscures any additional viral RNA mutations that might arise from RNA pol II transcription.
Selective sequencing of dUMP-depleted DNA fraction. We were interested in whether a different mutational spectrum might result if the fraction of proviral DNA that contained dUMP was subtracted from the total population of proviral DNA before single-molecule amplification and sequencing. To address this question, we performed UNG digestion on the total DNA extracted from infected MDM to remove all of the dUMP-containing copies prior to PCR amplification and then repeated the limiting dilution steps and clonal sequencing (Fig. 7a). Although the dUMP-depleted DNA showed a modest 2-fold reduction in the mutation frequency (0.8 x 10-3) (Table 1, Supplemental Table S4), this reduction was not highly significant (p-value = 0.16). Despite the insignificant effects on the mutation frequency, the mutation spectrum for the dUMP-subtracted DNA fraction consisted almost entirely of transition mutations, with only a single transversion mutation after sequencing 22,000 bases (Fig. 7a). We calculate that there is only a 1% probability that a single transversion mutation would have been randomly observed in the dUMP-depleted sample based on Poisson statistics and the measured transversion frequency in the dUMP-rich DNA sample. Thus, transversion mutations appear to be correlated with the DNA fraction that contains dUMP.
To explore why DNA clones containing dUMP showed a higher frequency of transversion mutations, we examined the average [A + T(U)] content for 7 mer sequences centered on these transversion sites to explore whether the mutations might be correlated with the density of dUMP incorporation on either the (+) or (-) strand. For comparison, we determined the average [A + T(U)] content of the 7 mer sequences surrounding the transition mutation sites in the dUMP-depleted and total DNA samples (Supplemental Tables S2 and S4). This comparison showed a statistically higher average frequency of [A + T(U)] near the transversion sites (µ = 0.80) as compared to the transition mutation sites (dUMP-depleted DNA, µ = 0.58; total DNA, µ = 0.55) (Fig. 7b). The differences in the mean values are significant with p values < 0.003. Possible mechanistic implications of this result are discussed below.
Relative infectivity of HIV-1BaL produced from MC, MDMEI and MDMLI. We determined the relative efficacy by which MC, MDM and activated T cells produce virus particles and extracellular vRNA by normalizing the viral output by the average number of proviruses present in each infected cell type (Fig. 8a, b). Using p24 or extracellular viral RNA levels as the measure for output, there were only modest differences for the three myeloid cell infections at 7 dpi, although MDM that were infected after complete differentiation (MDMLI) showed a 2-fold greater yield than MC or MDMEI. For comparison, activated T cells infected under identical conditions showed a 5-fold higher viral output than MC (Fig. 8a, 8b), indicating an intrinsically higher efficiency than any of the infected myeloid cells.
Finally, we investigated the relative infectivity of HIV-1BaL produced from infected MC, MDMEI, MDMLI and activated T cells by collecting culture supernatants at 7 dpi and using equivalent copies of virus to infect MOLT-4/CCR5 target cells (Fig. 9a). Seven days after infection of the target cells, total cellular DNA was extracted, and LRT copies were quantified using qPCR (Fig. 9b) and p24 levels were measured in the culture media (Fig. 9c). Although equal p24 units were used in each of the infections, the target cell LRT copies were about 4-fold greater for infections initiated with viral supernatants obtained from MDMLI as compared to MC, while MDMEI showed an intermediate level of LRT copies. In addition, supernatant p24 levels were 6-fold greater for infections initiated with MDMLI. These differences indicate that the fitness of viral particles produced from MC is lower than MDM despite the similar efficiency of production from integrated viruses (Fig. 8). If we define an infection cycle that begins with viral infection of a MC or MDM and ends with the infection of a new target cell, we calculate an overall 50-fold lower efficiency for MC as compared to a fully differentiated MDM. This comparison between the two cell types is based on the product of the relative efficiencies for virus integration (Fig. 4), viral particle production from integrated virus (Fig. 8) and the productive infection of new target cells (Fig. 9). Using the same calculation, MC are 300-fold less efficient at producing infective virus as compared to activated T cells. It is not clear at this time how much viral dUMP levels or UBER contributes to the lower infection efficiency in myeloid cells.
Expression of exogenous hUNG2 before HIV infection depletes uracilated HIV DNA products. Because of the vanishingly low levels of UBER pathway enzymes detected in MDMs, and especially the very low hUNG2 activity observed in both monocytes and MDMs, we hypothesized that low hUNG2 levels might limit the restrictive effects of the UBER pathway. To test this hypothesis, we over expressed full length human UNG (hUNG2) in MDMs using a doxycycline inducible lentiviral transduction system. The inducible system allowed us to test whether hUNG2 expression had a greater effect before HIV infection, or alternatively, after HIV had integrated into the MDM genomic DNA. Control experiments demonstrated the presence of high hUNG2 activity in cell extracts prepared from transduced MDM at days 1 and 3 post induction, but no activity in the absence of doxycycline induction (Supplemental Fig. S5).
We first tested the effect of inducing hUNG2 expression prior to infection with VSVG pseudo-typed HIVpNL4-3 virus particles capable of only a single round infection. In this experiment, fully differentiated MDM were first transduced with the hUNG2 expressing lentivirus at an MOI of 5 (0.1 pg p24 antigen/cell) and expression of hUNG was induced using 1µg/ml of doxycycline three days after transduction, followed by infection with HIVNL4-3 one day later (MOI = 0.5) (Fig. 10a). We followed the HIV provirus copy number and the fraction of total provirus that contained dUMP using alu-gag Ex-qPCR. These measured outcomes were compared with those of an uninduced control infection (blue bars, Fig. 10a), as well as an infection with HIVpNL4-3 in the absence of any prior lentiviral transduction (green bars, Fig. 10a). With pre-infection induction of hUNG2 expression (red bars, Fig. 10a), there was a 50% decrease in the provirus copy number between 1- and 7-days post HIVpNL4-3 infection and the fraction of proviruses containing dUMP was at the limit of detection for the Ex-qPCR method (< 0.2). In contrast, the no induction and HIVNL4-3 only control infections showed a stable or slightly increasing proviral copy number over the same time period and the fraction of proviruses containing dUMP increased from ~0.4 to 0.8. As discussed above and previously[4,27], no more than a 50% loss of provirus is expected when hUNG2 is overexpressed because the MDM exist as a mixed population where only about 40-50% of the total proviruses contain dUMP. However, the essentially complete absence of viral dUMP at one to seven days post infection as measured by alu-gag Ex-qPCR clearly indicates cellular hUNG2 excised the uracils prior to viral integration. These results indicate that when hUNG2 is abundant in the target cell prior to HIV infection, uracilated viruses can be efficiently destroyed before integration.
In a second experimental protocol, hUNG2 induction was delayed until 7 days after HIVNL4-3 infection to allow most of the HIV cDNA to integrate (Fig. 10b). Under this infection scenario we saw a smaller decrease in HIV proviral DNA between 8- and 14-days post-infection (one to seven days post-induction) as compared to the pre-infection induction of hUNG2 shown in Figure 10a. In addition, a higher fraction of proviruses contained dUMP and these proviruses slowly disappeared over one to seven days post-induction. In fact, the rate of decrease of dUMP-containing proviruses was not significantly different for the hUNG2-induced condition and the uninduced controls (Fig. 10b). This indicates that the rate-limiting step for post-integration excision of uracils does not involve hUNG2 and may instead involve remodeling of chromatin.