APOBEC3G Generates Nonsense Mutations in Human T-Cell Leukemia Virus Type 1 Proviral Genomes In Vivo ^{▿ †}

Cell lines and clinical samples.

MT-2 and MT-4 are HTLV-1-transformed cell lines. MT-1 and TL-Om1 were derived from leukemia cells. Jurkat cells were negative for HTLV-1. These cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM l-glutamine at 37°C in a 5% CO₂ atmosphere. Clinical samples were collected from 60 ATL patients and 10 HTLV-1 carriers. Genomic DNA was extracted from peripheral blood mononuclear cells using standard phenol-chloroform methods. This study was conducted according to the principles regarding human experimentation expressed in the Declaration of Helsinki. The study was approved by the Institutional Review Board of Kyoto University (G204). All patients provided written informed consent for the collection of samples and subsequent analysis.

Sequencing the complete provirus genomes in ATL cells.

The copy number of HTLV-1 provirus in cells from ATL patients was determined by inverse PCR (5), and ATL samples with one complete provirus were selected for direct sequencing. The complete provirus genome was amplified as two halves (fragments I and II) from genomic DNA samples derived from ATL patients. Fragment I was amplified using primers 5′-TGACAATGACCATGAGCCCCAAATATCC-3′ and 5′-CGGCTATTAAGACCAGGAAG-3′. PCR conditions were as follows: an initial step of 5 min at 94°C; 30 cycles, with 1 cycle consisting of 30 s at 94°C, 30 s at 64°C, and 5 min at 72°C; and a final step of 10 min at 72°C. Fragment II was amplified using primers 5′-AGAAACAAGCTCAGAAGCTA-3′ and 5′-TGTACTAAATTTCTCTCCTGAGAGTGC-3′. PCR conditions were as follows: an initial step of 5 min at 94°C; 30 cycles, with 1 cycle consisting of 30 s at 94°C, 30 s at 60°C, and 5 min at 72°C; and a final step of 10 min at 72°C. PCR products were then subjected to nested PCR to amplify I-1 and I-2 from fragment I and II-1 and II-2 from fragment II. The primers for I-1 were 5′-TGACAATGACCATGAGCCCCAAATATCC-3′ and 5′-GAGCTTAAAGTGATCTTGG-3′. The primers for I-2 were 5′-TTCCGATAGCCTTGTTCTCA-3′ and 5′-CGGCTATTAAGACCAGGAAG-3′. The primers for II-1 were 5′-TGGTATTATTTCAAGCTTCC-3′ and 5′-AAATGCAGGAGTTGGGGATT-3′. The primers for II-2 were 5′-AACTGTCTAGTATAGCCATC-3′ and 5′-TGTACTAAATTTCTCTCCTGAGAGTGC-3′. PCR conditions for nested PCR were as follows: an initial step of 5 min at 94°C; 28 cycles, with 1 cycle consisting of 30 s at 94°C, 30 s at X°C, and 2.5 min at 72°C; and a final step of 72°C for 10 min. X was 60°C for I-1 and II-2 and 58°C for I-2 and II-1. The final PCR products were gel purified, and the sequence was determined by direct sequencing.

Sequencing the pol and tax regions in HTLV-1 carriers.

The proviral load was determined as previously reported (32). Herculase II fusion DNA polymerase (Stratagene, La Jolla, CA), a Pfu-based DNA polymerase, was used to amplify pol and tax regions from genomic DNA derived from HTLV-1 carriers. The primers used to amplify the pol region were 5′-TACACCTTGCAATCCTATGG-3′ and 5′-GCTAGGCTTGCCTAGATGGG-3′. PCR conditions were as follows: an initial step of 5 min at 95°C; 30 cycles, with 1 cycle consisting of 20 s at 95°C, 20 s at 60°C, and 50 s at 72°C; and a final step of 50 s at 72°C. The primers used to amplify the tax region were 5′-GGTCCTCCGGGCATGACACA-3′ and 5′-TCTCCACGCTTTTATAGACT-3′. PCR conditions were as follows: an initial step of 5 min at 95°C; 28 cycles, with 1 cycle consisting of 20 s at 95°C, 20 s at 68°C, and 50 s at 72°C; and a final step of 5 min at 72°C. The PCR products were purified and cloned, and at least 20 clones were sequenced. DNA polymerase fidelity was studied by using a clone carrying an amplified tax region as the template instead of genomic DNA. The tax region was amplified using the same PCR conditions described above and cloned, and 25 clones were sequenced.

Sequencing the HTLV-1 genome in nontumor cells derived from ATL samples.

ATL patient-derived tumor cells are considered a mixture of a major population of monoclonal expanded tumor cells and a minor population of HTLV-1-infected nontumor cells. To amplify the HTLV-1 genome from infected nontumor cells, ATL samples carrying defective HTLV-1 genome based on direct sequencing were selected. Target regions were amplified by setting one of the primers at the deleted region. Sequences were determined by direct sequencing.

Mutation analysis.

Nucleotide sequences of all ATL samples were aligned using GENETYX-MAC version 13.0 software. Mutations were determined after filtering for (i) polymorphisms, defined as a nucleotide substitution occurring in >5 cases; and (ii) linkers, defined as >7 common nucleotide substitutions occurring in >2 cases. With these criteria, the nucleotide sequence from an ATL case, ATL-25, showed the fewest mutations (3 single-nucleotide substitutions with no deletions or insertions) and was selected as a standard sequence (AB513134). The nucleotide positions of identified mutations were numbered relative to this standard sequence.

For carrier samples, the sequence of a given region from all clones amplified from a carrier was aligned and compared with the standard sequence described above with mutated bases corrected. When a nucleotide substitution occurred in all clones, it was considered a polymorphism and excluded from the mutation repertoire.

cDNA synthesis and quantitative real-time RT-PCR.

Total RNA was extracted from transfectants using Trizol reagent (Invitrogen, Carlsbad, CA). RNA was treated with DNase I (Invitrogen) to eliminate the genomic DNA. cDNA was synthesized from 1 μg of total RNA with the Superscript preamplification system (Invitrogen) according to the manufacturer's protocol. cDNA product was quantified by real-time reverse transcription-PCR (RT-PCR) with power SYBR green PCR master mix and 7900HT fast real-time PCR system (both from Applied Biosystems) according to the manufacturer's instructions. The specific primers for the hA3G gene were 5′-CGCGTGCCACCATGAAGATC-3′ (forward) and 5′-TGTGGGTGGATCCATCGAGT-3′ (reverse). The specific primers for the human AID gene were 5′-TTCACCGCGCGCCTCTACTT-3′ (forward) and 5′-GCTGTCTGGAGAGACGAACT-3′ (reverse). Target cDNA was normalized to the amount of endogenous mRNA of ACTB. The primers used for ACTB were 5′-AGGCCAACCGCGAGAAGATG-3′ (forward) and 5′-CCAGAGGCGTACAGGGATAG-3′ (reverse). PCRs were carried out in triplicate. Data were analyzed by the comparative threshold cycle (C_T) method according to protocols of the manufacturer (Applied Biosystems).

In vitro editing of the HTLV-1 genome by hA3G.

QT6 quail cells were transfected with an infectious molecular clone of HTLV-1 (pX1MT-M) (30) in the presence or absence of the hA3G expression plasmid. Twenty-four hours posttransfection, HTLV-1-producing QT6 cells were treated with 200 μg/ml of mitomycin C at 37°C for 30 min, washed well, and then cocultivated 1:1 with fresh QT6 cells (13). Seventy-two hours later, the cells were collected, and genomic DNA was extracted. hA3G-mediated DNA hypermutation was detected using three-dimensional (3D) PCR that allows differential amplification of G-to-A hypermutants using a Taq DNA polymerase as previously described (25). The first round of PCR was performed with a denaturation temperature of 94°C using the primer pair 5′-CTGCAGATACAAAGTTAACC-3′ (forward) and 5′-TGGAGGAAGGAGGGTGGAAT-3′ (reverse). PCR conditions were as follows: an initial step of 5 min at 94°C; 35 cycles, with 1 cycle consisting of 30 s at 94°C, 30 s at 58°C, and 30 at 72°C; and a final extension step of 10 min at 72°C. Then, 0.5 μl of the first-round reaction product was used as the template for the second round of PCR using the following primers: 5′-CCATGCTTATTATCAGCCCA-3′ (forward) and 5′-GTTGGGGGTTGTATGAGTGA-3′ (reverse). The PCR program was the same as for the first round of reaction except that the denaturation gradient ranged from 94 to 82°C. PCR products were purified and cloned, and up to 20 clones were sequenced.

Nucleotide sequence accession number.

The nucleotide sequence of ATL-25, selected as the standard sequence, has been submitted to the DDBJ (DNA Data Bank of Japan) database under accession number AB513134.

HTLV-1 proviral sequence changes in ATL.

Previous studies showed that Tax expression was frequently lost by genetic changes in the tax gene, by DNA methylation of the 5′LTR, and by deletion of the 5′LTR (29). Similarly, although not well studied thus far, the expression of other viral proteins might also be lost through genetic changes in the provirus sequence. To analyze the genetic changes of the HTLV-1 provirus in ATL cells, we determined the entire sequence of the provirus (9,034 bp) in 60 ATL cases. For this study, we chose ATL cases with HTLV-1 proviruses containing both LTRs. As shown in Table 1, we detected genetic changes (nonsense mutations, deletions, and insertions) in the coding regions of viral genes in 28 of 60 ATL cases (46.7%). Deletions in the proviruses were detected in 27 cases (45%), while insertions were found in 10 cases (16.7%) (see Table S1 in the supplemental material). These deletions and insertions resulted in the loss or truncation of protein(s) in 18 cases (Table 1; also see Table S1 in the supplemental material).

Nonsense mutations, deletions, and insertions in the viral genes were found in all viral genes except the HBZ gene (Table 1). Since the p30 and p13 genes use an identical coding frame, the same mutations generate nonsense changes in both p30 and p13 (Fig. 1). The frequency of genetic changes in the tax gene was 16.7% (10/60) (Table 1), which was higher than in previous studies (7, 41). Of a total of 29 nonsense mutations, 27 preferentially accumulated in tryptophan codons (27/29 [93.1%]) (see Table S2 in the supplemental material). Since the tryptophan codon is TGG, a G-to-A mutation generates either a TGA or a TAG stop codon. In one nonsense mutation, the codon for arginine (CGA) was converted to the TGA stop codon by a C-to-T mutation. Interestingly, nonsense mutations tended to accumulate in the same cases (Table 1).

It was difficult to discriminate between mutations and polymorphisms in HTLV-1 provirus from ATL patients, since the proviral sequences in nonleukemic cells could not be analyzed in most ATL cases without an internal deletion of provirus. Some mutations might be characteristic of a subgroup of proviruses (polymorphism). Since nonsense mutations at the tax gene (nucleotide position 7469) were detected in 5 ATL cases, we tentatively classified base substitutions as polymorphisms when more than 5 identical base substitutions were observed in different cases. On the basis of this criterion, we observed 591 mutations in all ATL cases (see Table S3 in the supplemental material). Among them, the G-to-A mutation was the most frequent, contributing to 28.8% of the total mutations. Other frequently detected mutations were the C-to-T (23.7%), A-to-G (17.1%), and T-to-C (12.9%) mutations. Frequent G-to-A and C-to-T mutations strongly suggested the activity of deamination enzymes in the generation of these mutations.

Mutations in leukemic and nonleukemic cells in ATL cases.

In ATL patients, nonleukemic HTLV-1-infected cells coexist with leukemic cells. To analyze the proviral sequences of nonleukemic cells, we amplified proviral sequences using primers within regions deleted in leukemic cells to avoid amplification of provirus from leukemic cells. Nonsense mutations that were observed in leukemic cells were not detected in nonleukemic cells (Fig. 2), indicating that nonsense mutations generated in the provirus of leukemic cells were selected for. We compared the entire sequence of the HTLV-1 provirus in both leukemic cells and nonleukemic cells from five ATL cases. As shown in Table 2, only 37% of base substitutions were considered real mutations, since the nucleotide differed between leukemic cells and nonleukemic cells. However, 71% of G-to-A transitions were actually mutations, while 73% of other base substitutions were polymorphisms (Table 2), indicating that G-to-A mutations are predominant in ATL cells.

Association of G-to-A mutations with human APOBEC3G.

A high frequency of G-to-A mutations suggests the role of deaminase(s) in generating these mutations. Deamination enzymes are known to have individually preferred target sequences for deamination (44). Sequences surrounding the nonsense mutations generated by the G-to-A mutation in HTLV-1 proviruses showed a predominance of the 5′-GG dinucleotide context (target underlined) (Fig. 3 A). Among trinucleotides containing GG dinucleotides, CGG, TGG, and GGG were preferred (Fig. 3A); these are consistent with the target sequences of human APOBEC3G (hA3G) (48). The AGG sequence was also targeted in HTLV-1, unlike HIV-1. When we checked tetrameric sequences that contain a central GG dinucleotide, CGGG and TGGG were the preferred targets in the HTLV-1 provirus, similar to those reported for HIV-1 (Fig. 3B). As in HIV-1, a C at the +2 position (NGGC [N indicates T or C]) was not favored in HTLV-1. These findings suggest that the observed G-to-A mutations were generated by hA3G. Another deaminase, human APOBEC3F (hA3F), which is largely coexpressed with hA3G, is also reported to target single-stranded minus-strand DNA (46). Unlike hA3G, which favors the 5′-GG dinucleotide, the consensus target sequence of hA3F is 5′-GA (21). G-to-A mutations at GA sites contributed 13.4% of all G-to-A mutations (Fig. 3A), suggesting that hA3F might also play a role in these mutations in the HTLV-1 provirus.

Since hA3G targets the minus strand during reverse transcription, nonsense mutations generated by G-to-A mutations are not found in the HBZ gene. A G-to-A mutation in the proviral sense strand caused a nonsense mutation in the p12 gene (TGG to TGA), whereas the same mutation generated a missense mutation in the HBZ gene (CCA to TCA) (Fig. 1). G-to-A mutations occurred most frequently at GG sites (target underlined) with a T or C at their 5′ terminus (Fig. 3A). We therefore counted the number of TGG or CGG sites within individual HTLV-1 regions (Table 3). As NGGC (N indicates T or C) was not favored by hA3G (Fig. 3B), it was not included. Among HTLV-1 regions, the plus-strand sequence of the HBZ gene had the fewest TGG and CGG sites, even when the number of sites was normalized by the size of the region. This indicates that mutations in the HBZ gene-coding region caused by hA3G are also rare.

Mutations in the HTLV-1 proviral sequences from carriers.

Frequent detection of G-to-A mutations in HTLV-1 proviral sequences in ATL cases suggests that these nonsense mutations potentially occur in both HTLV-1 carriers and ATL patients. To explore this possibility, we next studied the proviral sequences of the pol, env, p12, p13/p30, p27, tax, and HBZ genes in asymptomatic HTLV-1 carriers. The proviral sequences in asymptomatic carriers were amplified using a high-fidelity DNA polymerase with proofreading activity, and after subcloning, the sequences were determined for at least 20 subclones. As seen in ATL cases, the mutation frequency differed drastically among individual carriers (representative data are shown in Tables S4 and S5 in the supplemental material), and the distribution of base substitutions among clones varied within individual carriers. G-to-A mutations accounted for 81% and 72% of all mutations in the pol and tax genes, respectively (Table 4 and 5, respectively). A high frequency of G-to-A mutations was also observed in other viral genes (data not shown). Analyses of the sequence context of G-to-A mutations showed a predominance of GG sequences, which is consistent with the finding that hA3G deaminates the viral genome during reverse transcription (48). In order to compare the frequencies of nonsense mutations in carriers versus ATL cases, the occurrence of nonsense mutations was further analyzed. Consistent with the findings for ATL cases, nonsense mutations were detected in the pol, env, p30/p13, p27, and tax genes, but not in the p12 and HBZ genes (Table 6). As in ATL cases, nonsense mutations in carriers were most frequently observed in the tax and pol genes. This result suggests that there is no bias for mutations in specific viral genes in ATL cases compared with carriers. These results suggest that nonsense mutations in provirus are not generated during oncogenesis but are present in the carrier state.

In addition to G-to-A mutations, C-to-T mutations were detected in the pol and tax genes in carrier cells as in ATL cells (see Table S2 in the supplemental material); they were also detected in other viral regions (data not shown). Comparative analyses of leukemic cell and nonleukemic cells showed that 79% of C-to-T base substitutions were polymorphisms (Table 2). However, after G-to-T mutations, C-to-T mutations were still the most frequent mutations in ATL patients and HTLV-1 carriers.

Expression of the hA3G gene and activation-induced deaminase gene.

In order to analyze the correlation between G-to-A mutation and hA3G expression, we first studied the mRNA level of hA3G using a real-time PCR assay. The hA3G gene was expressed in normal T cells, and the level of expression of hA3G moderately decreased in ATL cells. In ATL samples, there was no obvious correlation between the level of expression of the hA3G gene and the number of G-to-A mutations in the proviral genome.

Although only 21% of C-to-T base substitutions were mutations (Table 2), C-to-T mutations were the second most frequent mutation (see Table S2 in the supplemental material). Activation-induced deaminase (AID) was found to be a key factor for the switch recombination of immunoglobulin (Ig), and it is also implicated in hypermutations of Ig genes (15). AID deaminates cytosine to uracil. Aberrant expression of AID driven by NF-κB activation is found in gastric cancers and hepatomas, and AID expression has been invoked to explain induced mutations observed in cancer cells (28). Therefore, we studied AID expression in 7 ATL cases with different proviral mutation frequencies. As shown in Fig. 4, the AID gene transcription increased in ATL-15. In this case, we found six C-to-T mutations in the provirus, and the sequence context showed more AID preferred sequences in C-to-T mutated sites compared to other cases without AID activation (data not shown). Thus, AID might also play a role in the deamination of the HTLV-1 genome in at least some ATL cases. Since AID is thought to target the double-stranded DNA of provirus, it likely generates C-to-T mutations in the plus strand of the provirus. However, C-to-T base substitutions contained many polymorphisms, as shown in Table 2. Therefore, correlation between C-to-T mutations and AID expression remains to be studied.

In vitro editing of the HTLV-1 genome by hA3G.

Previous reports have suggested that HTLV-1 is relatively resistant to the antiviral effect of hA3G (4), consistent with our finding that the frequency of G-to-A mutations throughout the HTLV-1 proviral genome was low. In order to provide direct evidence that G-to-A mutations observed in the proviruses are the result of hA3G-mediated genome editing, we studied the editing effect of hA3G on the HTLV-1 genome in vitro. HTLV-1 viruses were generated in the presence of exogenous hA3G to allow packaging into budding viral particles, a step that is required for exerting editing activity of the cytidine deaminase. After a round of virus infection, the HTLV-1 DNA was analyzed using a highly sensitive PCR-based protocol, referred to as 3D PCR, that is capable of amplifying G-to-A hypermutated genomes as described in Materials and Methods. Differential amplification of a tax region from cells infected with viral particles produced in the presence of hA3G was achieved when the denaturation PCR temperature was lowered to between 84 and 86°C (data not shown). All sequences of differentially amplified PCR clones exhibited extensive G-to-A hypermutation, and the number of G-to-A transitions ranged from 8 (13%) to 29 (48%) per clone (Fig. 5). In ATL cases and in the same region, G-to-A mutations at four nucleotide positions resulted in premature stop codons in the p30 and tax genes (Fig. 1). The occurrence of such mutations was observed in vitro when viruses were produced together with hA3G. G-to-A mutations at two sites where nonsense mutations were more frequently detected in ATL cases were also frequent among 3D PCR clones. As expected, a preference for GG dinucleotide context in hA3G-mediated G-to-A editing was observed (Fig. 5C). Our data suggested that hA3G is indeed able to actively deaminate the HTLV-1 genome during reverse transcription although with low editing frequency, leading to G-to-A mutations in the plus strand of the provirus.