Efficient Deamination of 5-Methylcytidine and 5-Substituted Cytidine Residues in DNA by Human APOBEC3A Cytidine Deaminase

Deamination of 5-methylcytidine (5MeC) in DNA results in a G:T mismatch unlike cytidine (C) deamination which gives rise to a G:U pair. Deamination of C was generally considered to arise spontaneously. It is now clear that human APOBEC3A (A3A), a polynucleotide cytidine deaminase (PCD) with specificity for single stranded DNA, can extensively deaminate human nuclear DNA. It is shown here that A3A among all human PCDs can deaminate 5-methylcytidine in a variety of single stranded DNA substrates both in vitro and in transfected cells almost as efficiently as cytidine itself. This ability of A3A to accommodate 5-methyl moiety extends to other small and physiologically relevant substituted cytidine bases such as 5-hydroxy and 5-bromocytidine. As 5MeCpG deamination hotspots characterize many genes associated with cancer it is plausible that A3A is a major player in the onset of cancer.


Introduction
The human APOBEC3 (A3) locus encodes a seven gene cluster of PCDs [1]. While several clearly function as restriction factors for retroviruses and DNA viruses, their roles in the absence of infection are largely undefined [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17]. While transfected plasmid DNA can be hyperedited by A3A, A3C and A3H, A3A is by far the most efficient of the three enzymes [15,18]. Furthermore, human mitochondrial DNA (mtDNA) in the cytoplasm is susceptible to cytidine deamination, probably by one or more A3 PCDs [19]. Singularly, only A3A could deaminate human nuclear DNA (nuDNA). As observed levels of mt and nuDNA editing were dependent on uracil DNA glycosylase, it was suggested that A3 editing of human DNA was part of a novel DNA catabolic pathway [19]. 5-methylcytidine is the most common DNA modification and cytidine methylation is a major epigenetic mechanism of gene regulation and development. 5MeCpG mutation hotspots within cancer associated genes have long been described, the CpG motif being dictated by the host methylase while deamination is considered to be spontaneous [20][21][22][23]. Some reports suggest that 5MeC deamination is kinetically favoured over C deamination [24,25], while others have suggested that DNA repair of the T:G base pair with respect to a U:G pair contributes to the so called mutation hotspots [26][27][28].
5-methycytidine is not the only naturally occurring cytidine derivative in DNA. Activated neutrophils and eosinophils in particular generate high levels of hydrogen peroxide, HOCl and HOBr. Cytidine can be oxidized by N OH, N OCl, and N OBr free radicals to yield 5-hydroxy, 5-chloro and 5-bromocytidine derivatives [29][30][31]. Ultimately these bases have to be catabolyzed.
Generally the cytidine heterocycle is oxidized first by deamination to uracil, then barbituric acid derivatives and finally urea and malonic acid. As we have argued that A3A plays a physiological role in the catabolism of nuDNA [19], the question arose as to whether A3A could also oxidize 5-substituted cytidine bases in DNA. It is shown here that A3A is the only human PCD that edits 5MeC efficiently as well as 5-hydroxy and 5-bromo-derivatives.

Cell Culture and Transfections
Japanese quail embryonic fibroblast QT6 cells (ATCC CRL 1708) were maintained in Ham's medium supplemented with 100 units/ml penicillin, 2 mM glutamine, 5% tryptose phosphate, 1% chicken serum and 10% fetal calf serum. 6610 5 QT6 cells in 6 well-plates were transfected with functional A3 expression plasmids (2 mg), the A3A C101S catalytic mutant or pv (empty vector) and one day later, transfected with heat denatured T5MeCGA DNA (200 ng) using JetPrime (Polyplus Transfection, USA), total DNA was recovered at 48 hours post initial transfection and extracted using the MasterPure Complete DNA and RNA purification kit (Epicentre).

Results and Discussion
A totally 5MeC substituted 685 bp fragment of HIV DNA was made by PCR and co-transfected into QT6 cells along with the cytidine deaminase expression plasmid. QT6 cells do not produce an endogenous cytidine-editing background because the avian lineage does not encode A1 or A3 othologs [16,32,35,36]. After 24 hours, total DNA was recovered and amplification of an internal 485 bp fragment was made by standard PCR. Subsequently a nested 310 bp fragment was recovered by 3DPCR [33], a technique that allows recovery of AT rich DNA by exploiting its lower denaturation temperature (Td). As can be seen ,81.2uC represents the lowest Td of input DNA ( Figure 1A). Of the eight functional human PCDs only for A3A were 3DPCR products recovered at lower temperatures, down to 77.6uC. When sequenced these products proved to be hypermutated (mean 42%, range 30-66% methylcytidines edited, Figures 1B and 1C). There was a strong preference for editing in 59TpC dinucleotides on both strands, a trait of A3A, which was not altered by the 5methyl group ( Figure 1D) [19]. The negative findings for all the other active PCDs show that A3A deamination of traces of input normal DNA used for making the totally substituted 5MeC containing DNA was not responsible for the 3DPCR signals ( Figure 1A).
To assess whether A3A deaminates 5MeC and C with similar efficiency, a customized oligodeoxynucleotide referred to as CER was designed to explore editing in matched dinucleotide contexts (Figure 2A). This substrate was incubated with highly active myc-His6-tagged A3A purified from HEK293 cells [18]. Purified A3G from baculovirus infected insect cells was used as negative control [1,34]. DNA was recovered by 3DPCR. As can be seen from Figure 2B, DNA was recovered from the A3A-reaction down to a denaturation temperature of 77.3uC while the corresponding temperature for the A3G reaction was 82.2uC ( Figure 2B). Interestingly, A3A and A3G enzymes purified from E. coli and baculovirus infected cells (not shown) gave similar editing frequency and context analysis. This precludes the necessity for any cellular partner in the deamination reaction. To generate sitespecific editing frequencies DNA was cloned and sequenced from A) The CER target sequence is nested between two PCR primer targets (black arrows). Every 5MeC site has matched non-methylated equivalent to allow comparisons. B) 3DPCR recovery of A3A edited CER DNA. A3A-myc-His was purified from HEK293T cells as described [18]. The asterisks denote the PCR products cloned and sequenced. C) A selection of deaminated CER sequences with the number of mutant sequences shown to the right. D) Site specific cytidine and 5-methylcytidine deamination frequencies as a function of A3A concentration. doi:10.1371/journal.pone.0063461.g002 a number of reactions notably at 95uC and 84uC to avoid selection biases ( Figure 2B). Using this assay, both matched 5MeC and C sites were edited efficiently, with the latter appearing slightly more susceptible ( Figure 2C and 2D). However, as comparison of the ApC 6 and Ap m C 10 shows, occasionally 5MeC deamination could be as efficient as unmodified C ( Figure 2D).
As several TP53 CpG methylation sites are mutational hotspots in cancer (www.iarc.fr/p53/), an oligodeoxynucleotide corresponding to part of exon 8 was synthesized with 5MeC incorporated at two known sites of methylation in codons 273 and 282 ( Figure 3A). Following incubation with purified myc-His6-tagged A3A and recovery of products by standard PCR, both 5MeC and C were readily deaminated in a comparable manner ( Figure 3B and 3C), in agreement with the observations for the CER oligonucleotide. In terms of pathology, it is plausible that many CG-.TA mutations associated with cancer may be precipitated by PCD-catalyzed DNA cytidine deamination events hitherto attributed to spontaneous hydrolysis. Given that 5MeC is deaminated by A3A the singularity of 5MeCpG mutation hotspots in cancer probably has more to do with the relative efficiency of T:G mismatch repair compared to highly efficient U:G repair initiated by UNG [26][27][28].
These findings show that when over-expressed A3A is by far the most efficient human PCD at deaminating 5MeC DNA. They do not preclude some activity from other PCDs, notably AID, which is known to have relatively low catalytic activity [37,38]. Although 3DPCR failed to pick evidence of hyperediting, it is known that 3DPCR underestimates lightly edited DNA molecules. That baculovirus derived A3G could occasionally edit a 5MeC residue in the CER oligodeoxynucleotide suggests that A3A is singular in terms of 5MeC deamination efficiency, rather than having 5MeC deamination as a unique property ( Figure 2C).
As A3A is able to deaminate 5MeC, we explored its capacity to deaminate other 5-modified cytidine residues in ssDNA. Totally substituted DNA products were made by PCR using 5modified dCTP derivatives using Taq polymerase (Bioline, USA). We succeeded in making hydroxymethyl (5CH 2 OH-), hydroxy (5OH-), bromo (5Br-) and iodo (5I-) cytidine derivates. The same experimental procedure was adopted as described for Figure 1A. As can be seen in Figure 4A 3DPCR recovered DNA at or below the restrictive temperature of 80.2uC for the bromocytidine (79.6uC) and hydroxycytidine (80.2uC) derivatives, although compared to 5MeC, deamination was less efficient. Cloning and sequencing revealed that the bromo and hydroxycytidine adducts were indeed deaminated by A3A with deamination frequencies of between 36-57% and 2-26% per sequence. The dinucleotide context showed the invariant bias in favour of TpC, a trait for A3A ( Figure 4B). The volumes of these 5-moeities are: hydroxy 19 Å 3 , methyl 31 Å 3 , bromo 33 Å 3 , iodo 38 Å 3 and hydroxymethyl 41 Å 3 [39]. As the iodo and hydroxymethyl derivatives were not deaminated it is possible that the A3A binding pocket cannot accommodate cytidine derivatives larger than 5-bromocytidine [39,40].  As A3 editing is part of a DNA catabolic pathway [19], the problem of catabolizing modified DNA bases arises. 5MeC is found in nuclear and mitochondrial human DNA as well as bacterial DNA. Other recent reports have shown that A3A can deaminate 5MeC [41,42]. However, that A3A can also edit free radical oxidation DNA products expands its catabolic role. As 5chloro-dCTP is not commercially available, we were not able to explore the effect of A3A on a substrate akin to NOCl oxidized DNA. However, given that the volume of the 5-chloro moiety is 27 Å 3 , less than that of bromine (33 Å 3 ) [39], it is likely that A3A could deaminate 5-chlorocytidine in ssDNA.
Once again A3A emerges as one of the most singular of human PCDs -it alone is able to hyperedit nuDNA with a mutation frequency approaching 0.5 as well as deaminating 5-substituted cytidine in ssDNA. Although hypermutation is synonymous with DNA catabolism, a little A3A editing might be compatible with cell survival. The repair of deaminated 5-modified cytidine residues proceeds by mechanisms that are not as efficient as for the G:U pair which invariably is initiated by the highly efficient enzyme UNG [26][27][28]. As 5MeCpG deamination hotspots characterize many genes associated with cancer [43], it is plausible that A3A is a major player in the onset of cancer.

Author Contributions
Conceived and designed the experiments: RS JPV SWH. Performed the experiments: RS MMA. Analyzed the data: RS JPV SWH. Contributed reagents/materials/analysis tools: RS MMA. Wrote the paper: JPV SWH.