Hyperediting of human T-cell leukemia virus type 2 and simian T-cell leukemia virus type 3 by the dsRNA adenosine deaminase ADAR-1

Abstract

RNA editing mediated by adenosine deaminases acting on RNA (ADARs) converts adenosine (A) to inosine (I) residues in dsRNA templates. While ADAR-1-mediated editing was essentially described for RNA viruses, the present work addresses the issue for two δ-retroviruses, human T-cell leukemia virus type 2 and simian T-cell leukemia virus type 3 (HTLV-2 and STLV-3). We examined whether ADAR-1 could edit HTLV-2 and STLV-3 virus genomes in cell culture and in vivo. Using a highly sensitive PCR-based method, referred to as 3DI-PCR, we showed that ADAR-1 could hypermutate adenosine residues in HTLV-2. STLV-3 hypermutation was obtained without using 3DI-PCR, suggesting a higher mutation frequency for this virus. Detailed analysis of the dinucleotide editing context showed preferences for 5′ ArA and 5′ UrA. In conclusion, the present observations demonstrate that ADAR-1 massively edits HTLV-2 and STLV-3 retroviruses in vitro, but probably remains a rare phenomenon in vivo.

RNA editing mediated by adenosine deaminases acting on RNA (ADARs) has been shown to be one of the most prevalent post-transcriptional RNA modification mechanisms in higher eukaryotes (Bass, 2002; Samuel, 2001). These enzymes convert adenosine (A) to inosine (I) residues in dsRNA templates. Inosine is essentially recognized by the translational machinery as guanine (G), leading to proteins that are frequently non-functional (Li et al., 1991). Three ADAR genes are known. They are specific for dsRNA. While ADAR-1 and ADAR-2 are expressed in many tissues, ADAR-3 is only expressed in the nervous system (Bass, 1997, 2002; Chen et al., 2000; Melcher et al., 1996). ADAR-1 gene consists of 17 exons across a 30 kb sequence (George & Samuel, 1999; Wang et al., 1995). ADAR-1 transcription is initiated from multiple promoters, one being inducible by type I and II interferons (IFNs), while the others are constitutively active (George & Samuel, 1999; Liu et al., 1997). Interestingly, of the ADAR-1 gene transcripts i.e. ADAR-1L and -1S, only the former can be induced by IFN-α/β and γ, underlining its role in antiviral responses. ADAR-1 editing was initially described in the context of subacute sclerosing panencephalitis, a rare chronic degenerative disease that occurs several years after measles virus infection (Cattaneo et al., 1987, 1988; Patterson et al., 2001; Wong et al., 1989, 1991). ADAR-1 editing was originally confined to negative-stranded viruses such as measles virus, vesicular stomatitis virus (O’Hara et al., 1984), human parainfluenza virus (Murphy et al., 1991), lymphocytic choriomeningitis virus (Grande-Pérez et al., 2002), respiratory syncytial virus (Martínez et al., 1997; Rueda et al., 1994), influenza virus (Suspène et al., 2011; Tenoever et al., 2007) and Rift Valley fever virus (Suspène et al., 2008). Recently, measles and influenza virus genomes derived from inactivated seasonal influenza and live-attenuated measles vaccines were also shown to be edited by ADAR-1 (Suspène et al., 2011).

ADAR editing is not restricted to negative-stranded viruses since the hepatitis C virus (Taylor et al., 2005) genome was also found to be edited. Among retroviruses, A→G editing was first described for Rous-associated virus RAV-1 (Hajjar & Linial, 1995), avian leukosis virus (Felder et al., 1994) and more recently for human immunodeficiency virus-1 (HIV-1) (Doria et al., 2009; Phuphuakrat et al., 2008) although at a low frequency. The δ-retrovirus group includes four human T-cell leukemia viruses (HTLV-1−4), and their simian T-cell leukemia virus counterparts (STLV-1, -2 and -3) (Mahieux & Gessain, 2011; Slattery et al., 1999). STLV-1 is widely distributed in Asian and African non-human primates with STLV-3 being only found in African non-human primates (Mahieux & Gessain, 2011).

HTLVs- or STLVs-edited genomes have not been described so far. Since, these viruses mostly replicate through clonal expansion of the infected cell, they may be less prone to genetic editing (Wattel et al., 1995). We have recently developed a PCR-based method, referred to as 3DI-PCR, that allows selective amplification of ADAR-edited RNAs (Suspène et al., 2008). Here, we show that HTLV-2 and STLV-3 RNAs can be efficiently and massively edited by ADAR-1.

In the first series of experiments, 293T cells were transfected (PolyFect; Qiagen) for 48 h with 2 µg plasmid encoding the full-length HTLV-2 genome (pH 6neo) (Chen et al., 1983), in the presence or absence of 0.5 µg ADAR-1 expression plasmid. Total RNA was recovered using Trizol and cDNA was synthesized by using random primers. A fragment of the HTLV-2 pX gene (nt 6660–6937) was amplified by using the previously described 3DI-PCR technique (Suspène et al., 2008). pH 2Sout 5′-CATAACCAGTATTCCCTTATCAACCC-3′ and pH 2Rout 5′-TTCTGCAGGAGCGTGAGGAGCGGGAGC-3′ primers were used for the first PCR round, while pH 2Rin 5′-GCTATAATAGACCTGCTAGCTTCTGC-3′ and pH 2Sin 5′-CGGCGCAGAAAGGAGCGCCTGCGG-3′ primers were used for the second PCR round. 3DI-PCR products corresponding to extracts obtained only from cells that have been transfected with both HTLV-2 and the ADAR-1 plasmid were recovered at a PCR denaturation temperature as low as 64.8 °C. They were then cloned and sequenced.

In the first series of analyses, 55 extensive and monotonously A→G-edited HTLV-2 pX sequences were recovered (Fig. 1a). As a control, a Western blot analysis was performed in cells transfected or not with the ADAR-1 expression plasmid (Liu et al., 1997) (Fig. 1b). ADAR-1 enzyme was able to extensively deaminate HTLV-2 RNA (Fig. 1a, c). Of note, hyperedited sequences could not be recovered in the absence of exogenously expressed ADAR-1 (data not shown). The A→G editing frequency distribution per clone shows some lightly edited genome with 1–5 mutations (~20 %) and a majority of highly mutated genomes (i.e. >20 mutations ~73 % of all A nucleotides, Fig. 1d). The mean editing frequency was ~46 % (range 1.5–78 %) (Fig. 1d). The dinucleotide context associated with adenosine editing showed a clear preference for 5′ArA and 5′ UrA and an aversion for 5′ GrA and 5′ CrA (Fig. 2a, left panel), which is in agreement with the literature (Lehmann & Bass, 2000; Suspène et al., 2008, 2011). In contrast, we could not detect any obvious 3′ context (Fig. 2a, right panel).

Fig. 1.

HTLV-2 genome is susceptible to ADAR-1 editing. (a) Selection of ADAR-1 hypermutated HTLV-2 sequences. The HTLV-2 sequence is given with respect to the retroviral plus strand. Only differences among hypermutated sequences are shown. The number of substitutions per sequence is indicated to the right. # and % indicates the number of A→G transitions and the percentage of A targets edited to G, respectively. (b) 293T cells were transfected or not with an ADAR-1 expression plasmid. Twenty-four hours after transfection, cell extracts were processed for Western blot analysis using an anti-ADAR-1 (ab88574; Abcam) antibody. (c) Mutation matrices of HTLV-2 hyperedited sequences. The 19 U→C transitions are distributed among several sequences. No U→C hyperedited sequence was found. (d) Frequency distribution of A→G editing per clone [# and % as in (a)].

Fig. 2.

HTLV-2 sequence context among sites of ADAR-1 deamination. (a) Dinucleotide analysis in 5′ (left) and 3′ (right) of HTLV-2-edited genomes. Dots indicate the edited base. χ² analysis indicates dinucleotide frequencies that significantly deviate from the expected values (P<0.05, *). (b) Deamination frequencies across the HTLV-2 target sequence. Base-specific deamination frequencies were calculated among a collection of 55 sequences from all the reactions and are given as a function of local sequence context. The horizontal bar indicates the mean site deamination frequency, assuming no effect of the sequence context. Values on the top of each histogram represent the number of deaminated A among the 55 ADAR-1 HTLV-2-edited sequences. (c) Amplification of ADAR-1 cDNA obtained from 20 HTLV-2-infected individuals. The expected size of the amplified product is 185 bp. Mw, Molecular mass marker.

With a large number of clones sequenced, we defined site-specific editing frequency (Fig. 2b). Among the sixty-seven potential adenosine targets that are present in the HTLV-2 pX RNA template, we observed that some positions were highly refractory to mutations (see for example A₀₉, A₂₂ and A₃₈). On the other hand, others were strongly deaminated. As an example, A₅₁ was edited in 42 out of 55 clones sequenced (i.e. ~76 %). One residue (A₆₃) was totally refractory to ADAR-1 editing.

In order to demonstrate a possible effect of ADAR-1 editing in vivo, peripheral blood mononuclear cells were obtained from 20 HTLV-2-infected individuals and immediately frozen (Douceron et al., 2012). Total RNA was extracted and cDNA was subsequently obtained using random primers. Given the fact that tax is usually expressed in less than 50 % of all HTLV-infected individuals, we amplified an env region by PCR and 3DI-PCR. However, hyperedited sequences could not be recovered upon cloning and sequencing of the 20 different 3DI-PCR products, suggesting that ADAR-1 editing is a rare event in vivo in HTLV-2-infected individuals (data not shown). As a control, ADAR-1 expression was detected by RT-PCR among the 20 HTLV-2-infected individuals, demonstrating that all ex vivo samples contained detectable levels of the ADAR-1 transcript (Fig. 2c).

To determine whether ADAR-1 editing could occur for another δ-retrovirus, 293T cells were transfected (Polyfect; Qiagen) with an STLV-3 (PPAF-3) infectious molecular clone (Calattini et al., 2006; Chevalier et al., 2007) without overexpressing ADAR-1. Forty-eight hours post-transfection, total cellular RNA was recovered and cDNA was synthesized with random primers as described above. A fragment of the PPAF-3 pX mRNA was amplified by a nested PCR procedure. PCR products obtained at a 95 °C denaturation temperature were cloned and sequenced. Surprisingly, out of five clones sequenced, we recovered three ADAR-edited sequences (Fig. 3a). These sequences uniquely displayed A→G transitions (Fig. 3b), with a ~29 % mean adenosine substitution frequency per clone (range ~22–38 %). Once again and similar to the HTLV-2 results (Fig. 2), a significant preference for 5′ ArA and 5′ UrA contexts was observed (Fig. 3c). As ADAR-1 is constitutively expressed in 293T (Wang & Samuel, 2009), the majority of STLV-3-edited sequences can probably be ascribed to the ADAR-1 deaminase.

Fig. 3.

STLV-3 genome is susceptible to ADAR-1 editing. (a) The sequence is given with respect to the retroviral plus strand. Only differences among hypermutated sequences are shown. The number of substitutions per sequence is indicated to the right. # and % indicate the number of A→G transitions and the percentage of A→G edited, respectively. (b) Mutation matrices of STLV-3 hyperedited sequences. (c) Dinucleotide analysis in 5′ (left) and 3′ (right) of STLV-3-edited genomes. Dots indicate the edited base. χ² analysis indicates dinucleotide frequencies that significantly deviate from the expected values (P<0.05, *).

The present study shows that HTLV-2 and STLV-3, two primate retroviruses, can be massively edited by ADAR-1 in cell culture. For HTLV-2, the selective and sensitive 3DI-PCR method was necessary to recover ADAR-1-edited sequences. By contrast ADAR-1-edited STLV-3 sequences were recovered after conventional nested-PCR. Since 293T cells were used for both experiments and since HTLV-2 and STLV-3 sequences were cloned in the same backbone SV2neo vector, this differential sensitivity to ADAR-1 is likely to be related to the viral genome. HTLV-2 and STLV-3 have different genetic structure at the 3′ end of their genome, although none of their gene products are known to be IFN antagonists. Another variable could be the degree of secondary structure in the target sequence. Indeed, ADAR-1 editing occurs by flipping out the adenosine in a dsRNA structure. Local structural differences might therefore explain the differences between HTLV-2 and STLV-3 results. If ADAR-1 was packaged more efficiently into STLV-3 capsids, the viral genome would probably be more efficiently edited. In any case, the susceptibility of STLV-3 to restriction by ADAR-1 is striking. The present data do not exclude editing of viral mRNAs in the cytoplasm as opposed to editing of genomic RNA within the virion.

The contrasts between the ADAR-1 and APOBEC3G editing enzymes are remarkable. Indeed, both are induced by IFN-α and target HTLV-1 or HIV-1 retroviruses. While these two retroviruses infect the same CD4⁺ T-lymphocytes, ADAR-1 massively edits HTLV-2 sequences in vitro, albeit at low frequency. While recent work has shown that HIV-1 can be edited by ADAR-1, very few A→G mutations could be detected (Doria et al., 2009; Phuphuakrat et al., 2008). Of note, we also failed to detect massive editing of HIV-1 TAR or env RNA with our 3DI-PCR approach (data not shown). In contrast, in the absence of Vif, HIV-1 cDNA is massively edited by APOBEC3G and 3DPCR(Suspène et al., 2005) is not needed to recover these sequences. By contrast, HTLV-1 cDNA is susceptible to APOBEC3G editing, but sensitive 3DPCR is necessary to recover edited sequences.

In conclusion, the present observation demonstrates that ADAR-1 massively edits HTLV-2 and STLV-3 retroviruses in vitro, but probably remains a rare phenomenon in vivo.