Variation of HIV-1 Mutation Spectra among Cell Types

Abstract

The high rates of mutation, recombination, and replication drive HIV-1 diversity. In this study, we investigated how cell type affects viral mutation rate and mutation spectra. In studying four different cell types, no differences in mutation rate were observed, but intriguingly cell type differences impacted HIV-1 mutation spectra. This is the first description of significant differences in HIV-1 mutation spectra observed in different cell types in the absence of changes in the viral mutation rate.

TEXT

Human immunodeficiency virus type 1 (HIV-1) infection remains a serious threat to global public health, with over 34 million people infected worldwide (http://www.unaids.org). In the absence of an effective vaccine, antiretroviral drug therapy remains the primary means for preventing transmission and disease progression, as well as new infections (1). The error-prone nature of HIV-1 reverse transcriptase (RT) remains a key determinant in the generation of mutations during HIV-1 replication. HIV-1 RT plays a prominent role in the high genetic diversity and evolution of retroviruses, which is driven by high rates of viral replication, mutation, and recombination (2). High genetic diversity allows HIV-1 to evade the immune system and become resistant to drug therapy. This high mutation rate and the high rate of virus evolution impact virus transmission and disease progression, as well as promotes the emergence of antiviral drug resistance (2).

Cellular proteins can exploit the propensity of HIV-1 to rapidly mutate by enhancing viral mutagenesis to the point where the virus is unable to replicate with enough fidelity to remain infectious (3). The apolipoprotein B mRNA-editing, enzyme-catalytic, polypeptide-like 3 (APOBEC3) family of proteins, for instance, act as cytidine deaminases during reverse transcription, causing G-to-A hypermutation in Vif-deficient HIV-1 (4, 5). APOBEC3G (A3G) has been particularly well characterized for its ability to induce cytosine deamination during the minus-strand DNA synthesis step of reverse transcription, resulting in G-to-A mutations in the plus-strand viral DNA (6, 7). The editing activities of APOBEC3G result in G-to-A hypermutation (and lethal mutagenesis), although sublethal mutagenesis has been demonstrated (8). APOBEC3D, APOBEC3F, and APOBEC3H have also been demonstrated to restrict and cause G-to-A hypermutation of Vif-deficient HIV-1 (9).

Previous studies have provided evidence that HIV-1 genetic variation is impacted by RNA polymerase II transcription errors as well as minus-strand and plus-strand mutations that arise during HIV-1 reverse transcription (10–12). Deoxynucleoside triphosphate (dNTP) levels have been shown to have direct effects on RT fidelity (13, 14). Macrophages have been shown to have low dNTP levels, and this decreases the efficiency of viral DNA synthesis and increases the likelihood of mutations occurring during HIV-1 reverse transcription (15). A previous study investigated the differences in the HIV-1 mutation rate and mutation spectra observed between HIV-1 replication in HeLa cells versus that of CEM-A cells (16). No significant differences were observed in the viral mutation rates between these two cells. While the number of mutants that were characterized was small and prohibited a statistical analysis of potential differences in the rates of mutation for specific mutation types, the general trends suggested that there was no significant difference in the rates of base pair substitution mutations, frameshift mutations, and deletion or deletion-with-insertion mutations (16).

In the present study, we sought to further investigate the role of how cell type affects the mutation rate of HIV-1 as well as the virus mutation spectra. For analysis of mutation rate and mutation spectra, CEM-GFP (green fluorescent protein-expressing T lymphoblast cell line; NIH AIDS Research and Reference Reagent Program via J. Corbeil), U373-MAGI-CXCR4_CEM cells (glioblastoma cell line; via M. Emerman through the NIH AIDS Research and Reference Reagent Program), 293T cells (human embryonic kidney cells; American Type Culture Collection [ATCC]) and SupT1 cells (T lymphoblast cell line; ATCC) were transduced with an HIV-1 vector, pHIG (8). Mutant detection was determined by detection of cells that had a mouse heat-stable antigen-negative (HSA⁻) GFP⁺ phenotype. The calculation of mutant frequency was determined by dividing the number of infected cells harboring a mutated provirus phenotype (i.e., HSA⁻ GFP⁺) by the total number of cells infected (i.e., HSA⁻ GFP⁺ and HSA⁺ GFP⁺). Using this analysis, it was determined that there was no significant difference (P > 0.05, Student's t test) in the frequencies of mutants recovered from these cell types in parallel analyses (Table 1). A similar lack of observed difference in mutation rates had previously been made in an analysis conducted comparing HeLa cells and CEM-A cells (16).

Table 1.

Mutant frequency of HIV-1 among selected cell lines^a

Cell line	Mutant frequency
CEM-GFP	0.11 ± 0.04
U373-MAGI	0.12 ± 0.03
293T	0.11 ± 0.04
SupT1	0.11 ± 0.01

^aTwo million 293T cells were transfected with 10 μg of pHIG and 1 μg of pVSV-G using the calcium phosphate method as previously described (17). The viral supernatants were collected 48 h posttransfection, filtered, and used to infect permissive target cells (1.4 × 10⁶). Twenty-four hours postinfection, the cells were harvested and prepared for flow cytometry as described previously (8). Infected cells, typically 20 to 40% of the total for each target cell line, were analyzed by flow cytometry for expression of two marker genes, coding for HSA and GFP. Mutant frequencies were calculated from the cell population phenotypes identified by flow cytometry using the formula (HSA⁻ GFP⁺)/[(HSA⁻ GFP⁺) + (HSA⁺ GFP⁺)]. For each target cell line, ∼7 × 10⁵ to 9 × 10⁵ cells were identified as infected cells (i.e., HSA⁻ GFP⁺ and HSA⁺ GFP⁺ cell populations) and ∼8 × 10⁴ cells were identified as infected cells with a mutation in the reporter gene (i.e., HSA⁻ GFP⁺).

To determine if the HIV-1 mutation spectra were influenced by cell type, proviral HSA mutation target sequences were analyzed from the four cell lines after parallel virus infections. Interestingly, a significant difference was noted when comparing T-to-C transition mutations between CEM-GFP and SupT1 cells (P < 0.05, Fisher's exact test) and G-to-A transition mutations between CEM-GFP and 293T (P < 0.01), CEM-GFP and SupT1 (P < 0.01), and U373-MAGI and 293T (P < 0.05) (Table 2) cells. No other significant differences (P > 0.05) between any of the other mutation types were detected in these analyses. The relatively high level of A-to-G mutations observed in all 4 cell lines could be due to either ADAR-1 or ADAR-2 activity (18–20). When sequences that possessed multiple G-to-A mutations in the proviral HSA mutation target gene sequence were removed from the analysis, no significant differences (P > 0.05) in G-to-A mutations among the 4 cell types analyzed were observed (data not shown). This indicates that the frequency of G-to-A transition mutations was significantly influenced by the proviral sequences recovered that contained multiple G-to-A mutations in the target gene sequence.

Table 2.

Mutation spectra in the HSA mutation target gene of HIV-1 proviral sequences^a

Mutation	% of cells with mutation
	CEM-GFPb	U373-MAGIc	293Td	SupT1e
G-to-A	29	26	17	17
A-to-G	27	29	29	29
T-to-C	27	29	30	36
C-to-T	10	10	14	11

^aCells were transduced with an HIV-1 vector (HIG) pseudotyped with vesicular stomatitis virus protein G (VSVG), and the HSA reporter gene sequence from the provirus in infected cells was PCR amplified and sequenced. The percentage of mutations for each mutation type compared to the total mutations identified is indicated.

^bTotal no. of sequences, 238; total no. of mutations, 365.

^cTotal no. of sequences, 225; total no. of mutations, 329.

^dTotal no. of sequences, 154; total no. of mutations, 199.

^eTotal no. of sequences, 173; total no. of mutations, 219.

Proviral sequences with multiple G-to-A mutations in the target gene sequence were recovered from CEM-GFP, U373-MAGI, and 293T cells (Table 3). There were significant differences (Fisher's exact test) in the frequencies of sequences with multiple G-to-A mutations recovered between CEM-GFP and 293T cells (P = 0.03), CEM-GFP and SupT1 cells (P = 0.0005), and U373-MAGI and SupT1 cells (P = 0.006). The mutational load for G-to-A mutations per proviral sequence from each of the cell lines was analyzed (Fig. 1A). Although the average numbers of G-to-A mutations per sequence were not statistically different between CEM-GFP, U373-MAGI, and 293T cells, there was a trend toward a higher degree of G-to-A mutations per sequence in CEM-GFP and U373-MAGI cells relative to that of 293T cells. The locations of G-to-A mutations analyzed indicated that certain G residues were hot spots for G-to-A mutations (Fig. 1B). Of these mutations, 73% of the G-to-A mutations occurred at GA dinucleotides, while 17% occurred at GG dinucleotides, 5% occurred at GT dinucleotides, and 5% occurred at GC dinucleotides.

Table 3.

Proportion of HIV-1 proviral sequences recovered per infected cell line analyzed possessing multiple G-to-A mutations in the HSA mutation target gene^a

Target cells	No. (%) of mutant sequences with multiple G-to-A mutations/total mutants analyzed
CEM-GFP	14/238 (5.9)
U373-MAGI	10/225 (4.4)
293T	2/154 (1.3)
SupT1	0/173 (0)

^aCells were transduced with an HIV-1 vector (HIG) pseudotyped with vesicular stomatitis virus protein G (VSVG), and the HSA reporter gene sequence from the provirus in infected cells was PCR amplified and sequenced. The number of proviral sequences having HSA reporter gene sequences with multiple G-to-A mutations from each cell type was divided by the total number of proviral sequences identified with mutations.

Fig 1.

G-to-A mutational load and mutation location in HIV-1 proviruses with multiple G-to-A mutations. (A) G-to-A mutational load in HIV-1 proviruses from CEM-GFP, U373-MAGI, and 293T cells. Each provirus with a mutant HSA sequence containing with multiple G-to-A mutations is indicated by a black circle (CEM-GFP), black square (MAGI), or black triangle (293T). The average G-to-A mutational load and standard deviation are indicated. (B) Location of G-to-A mutations in recovered proviruses harboring multiple G-to-A mutations. The red, green, and blue circles above G residues indicate the locations of G-to-A mutations (one circle per mutation identified) in proviruses recovered from CEM-GFP, U373-MAGI, and 293T, respectively. The start and stop codons of the HSA gene are identified by black rectangular boxes. (C) Quantitative RT-PCR was performed to determine the relative levels of APOBEC3 mRNA expression among the cell lines under investigation (i.e., CEM-GFP, SupT1, U373-MAGI, and 293T). The asterisk indicates that the APOBEC3C level from 293T cells was set to 1, and all other values are relative to this measurement. The mRNA expression levels were normalized to TATA-binding protein (TBP) mRNA levels. For these analyses, data with a difference in efficiencies between the APOBEC3 standard and TBP standard were ≤10% and R² ≥0.98. Threshold cycle (C_T) values were determined using the regression method, and data were analyzed by E^−ΔCT. Experiments were conducted in triplicate with the standard deviation indicated.

The observation that G-to-A mutations preferentially occurred at GA dinucleotides suggested that these mutations may be due to the expression of APOBEC3 proteins. In order to test whether there was a correlation between the frequency of recovering proviral sequences with multiple G-to-A mutations and that of APOBEC3 gene expression, mRNA expression levels of the APOBEC3 proteins were analyzed in each of the cell lines under study. The study of mRNA expression levels was done as a surrogate for assessment of APOBEC3 protein levels, given that antibodies that can readily differentiate the family of proteins are not currently available. A3A, A3B, A3C, A3D/E, A3F, A3G, A3H, and TATA-binding protein expression vectors were used as control standards (21). Quantitative reverse transcriptase PCR (qRT-PCR) analysis was done on each cell line for A3A, A3B, A3C, A3D/E, A3F, A3G, and A3H (Fig. 1C). A3C mRNA expression levels suggested a possible correlation with the prevalence of proviral sequences harboring multiple G-to-A mutations from each cell line (i.e., CEM-GFP > U373-MAGI > 293T > SupT1). However, analysis using a Pearson correlation coefficient test did not support this correlation (P = 0.052). The A3G mRNA expression also appeared to have an expression pattern that appeared to correlate to the proviral sequences with multiple G-to-A mutations from each cell line. In particular, significant differences were observed in A3C expression between CEM-GFP and SupT1 (P = 0.0015, Student's t test), CEM-GFP and 293T (P = 0.0213), U373-MAGI and 293T (P = 0.0286), U373-MAGI and SupT1 (P < 0.001), and 293T and SupT1 (P < 0.001) cells. However, the observed dinucleotide specificity (i.e., GA) was not consistent with the dinucleotide specificity of A3G (i.e., GG). All of the other APOBEC3s either lacked mRNA expression in one of the cell lines where multiple G-to-A mutations were observed, or there was no significant difference in mRNA levels for a particular A3 between the 4 cell lines analyzed. It is interesting to note that while both CEM-GFP and Sup T1 cells are T lymphoblast cell lines, these findings indicate that distinct virus mutation spectra were observed. Taken together, the origins of the multiple G-to-A mutations in recovered proviral sequences could not be directly attributed to A3 gene expression in the 4 target cell types.

To date, the nature of how cell type influences the mutation rate and mutation spectra of HIV-1 has not been extensively studied. The discovery of the APOBEC3 proteins provided clear evidence of the potential for the host cell to extensively edit and mutate the HIV-1 genome, which may lead to mutations that can shape HIV-1 evolution (8). In this study, we have investigated several cell lines that are commonly used to study HIV-1 replication in cell culture. Using an HIV-1 vector to study the mutation rate and mutation spectra, we observed that cell type did not influence the viral mutation rate, as had been observed in an analysis of HIV-1 mutation rates between HeLa and CEM-A cells (16). Importantly, we observed for the first time distinct differences in HIV-1 mutation spectra in parallel analyses of 4 different target cell types. In particular, there was a significant difference in the frequency of G-to-A transition mutations observed between CEM-GFP and 293T cells, U373-MAGI and 293T cells, and between CEM-GFP and SupT1 cells. A significant difference in T-to-C transition mutations was also observed between CEM-GFP and SupT1. Interestingly, analysis of the proviral sequences from CEM-GFP and U373-MAGI cells led to the characterization of sequences harboring multiple G-to-A mutations in the reporter gene sequence, ranging from 2 to 13 G-to-A mutations. Furthermore, there were significant differences in frequencies between CEM-GFP and 293T cells, CEM-GFP and SupT1 cells, and U373-MAGI and SupT1 cells. The proviral sequences with multiple G-to-A mutations occurred mainly at GA dinucleotides (i.e., 73%) and GG dinucleotides (i.e., 17%).

Analysis of APOBEC3 mRNA expression levels did not allow for a correlation between the expression level of any one particular APOBEC3 and the observed sequences with multiple G-to-A mutations. There have been previous reports implicating an APOBEC3 protein as having a target cell effect in generating G-to-A mutations (22, 23). Koning et al. hypothesized that A3A was responsible for G-to-A editing of HIV-1 cDNA in macrophages (23), while Bourara et al. hypothesized that A3C could mutate HIV-1 viral DNA in the target cell (22). In both studies, the G-to-A mutations were likely sublethal and did not result in G-to-A hypermutation. It is possible that in these previous studies, G-to-A hypermutation did occur but was at a low frequency and was not detected.

Our studies here are distinct in that mutation spectrum differences involving both sequences containing multiple G-to-A mutations in the mutation target gene sequence (which is indicative of G-to-A hypermutation), and differences in the frequencies of other mutation types were observed. While the origins of these mutations were not determined, differences in cellular protein levels that could edit HIV-1 DNA or differences in the fidelity of HIV-1 reverse transcriptase and/or cellular RNA polymerase II are likely responsible for the changes in mutation spectra observed in this study. As indicated earlier, differences in nucleotide pool levels in various cell types can influence the likelihood of RT-mediated mutation (15). Further studies to determine the molecular basis for these differences in mutation spectra would be of particular interest, as well as studies that would extend these studies to human primary cells (e.g., primary T cells and macrophages). Additional studies to investigate the ability of APOBEC3 proteins are warranted and would be enhanced with antibody reagents that allow for specific detection of each APOBEC3 protein as well as specific small interfering RNAs (siRNAs) for mRNA depletion studies.