In vitro deamination of cytosine to uracil in single-stranded DNA by apolipoprotein B editing complex catalytic subunit 1 (APOBEC1).

Apolipoprotein B-editing complex catalytic subunit 1 (APOBEC1) is the catalytic component of an RNA-editing complex that deaminates C6666 --> U in apolipoprotein B RNA in gastrointestinal tissue, thereby generating a premature stop codon. Whereas RNA is the physiological substrate of APOBEC1, recent experiments have strongly indicated that, when expressed in bacteria, APOBEC1 and some of its homologues can deaminate cytosine in DNA. Indeed, genetic evidence demonstrates that the physiological function of activation-induced deaminase, a B lymphocyte-specific APOBEC1 homologue, is to perform targeted deamination of cytosine within the immunoglobulin locus, thereby triggering antibody gene diversification. However, biochemical evidence of in vitro DNA deamination by members of the APOBEC family is still needed. Here, we show that deamination of cytosine to uracil in DNA can be achieved in vitro using partially purified APOBEC1 from extracts of transformed Escherichia coli. Thus, APOBEC1 can deaminate cytosine in both RNA and DNA. Strikingly, its activity on DNA is specific for single-stranded DNA and exhibits dependence on local sequence context.


Apolipoprotein B-editing complex catalytic subunit 1 (APOBEC1) is the catalytic component of an RNA-editing complex that deaminates C
3 U in apolipoprotein B RNA in gastrointestinal tissue, thereby generating a premature stop codon. Whereas RNA is the physiological substrate of APOBEC1, recent experiments have strongly indicated that, when expressed in bacteria, APOBEC1 and some of its homologues can deaminate cytosine in DNA. Indeed, genetic evidence demonstrates that the physiological function of activation-induced deaminase, a B lymphocyte-specific APO-BEC1 homologue, is to perform targeted deamination of cytosine within the immunoglobulin locus, thereby triggering antibody gene diversification. However, biochemical evidence of in vitro DNA deamination by members of the APOBEC family is still needed. Here, we show that deamination of cytosine to uracil in DNA can be achieved in vitro using partially purified APOBEC1 from extracts of transformed Escherichia coli. Thus, APOBEC1 can deaminate cytosine in both RNA and DNA. Strikingly, its activity on DNA is specific for single-stranded DNA and exhibits dependence on local sequence context.
Deamination of cytosine to uracil can occur in vivo at the level of nucleotide, in RNA and in DNA. Thus, enzyme-catalyzed deamination of cytidylate to uridylate and of dCTP to dUTP plays a key role in nucleotide metabolism, underpinning the pathway of de novo thymidylate synthesis (1). In the context of RNA, the major physiological example of programmed cytosine deamination in mammals is the tissue-specific editing of the transcript encoding apolipoprotein B. This editing is performed by a complex comprising APOBEC1, a zinc-binding protein showing homology to cytidine deaminases of bacteria and other organisms (2)(3)(4)(5). In the context of DNA, the low level deamination of cytosine to uracil, which takes place spontaneously (and which might be of relatively minor significance when it occurs with free nucleotides or in mRNA), can have major effects, contributing to genome mutation, cancer, and evolution (6). Indeed, such DNA deamination is the focus of a specific base-excision repair pathway.
Recent work has now provided strong genetic evidence that deamination of cytosine can also occur in DNA as part of a physiological program. Thus, it appears that deamination of dC 3 dU within the immunoglobulin loci is triggered by AID 1 (activation-induced deaminase), a B lymphocyte-specific member of the APOBEC family, where it initiates major pathways of antibody gene diversification (7)(8)(9)(10)(11)(12). Furthermore, several members of the APOBEC family (including APOBEC1 itself) when expressed in bacteria appear to cause mutation through deamination of cytosine in DNA (10,13). Quite apart from the homology of the APOBEC proteins to bacterial cytidine deaminase (4), the genetic evidence very strongly indicates that the APOBEC family members themselves are intimately involved in the DNA deamination reaction. Thus, the mutations generated by APOBEC family members are all transitions at dC:dG with the frequency and nature of the nucleotide substitutions affected by deficiency in uracil-DNA glycosylase. Furthermore, the local sequence context of the dC residue targeted for mutation differs between APOBEC family members (10,13). Nevertheless, there is a need for biochemical evidence that APO-BEC family members can trigger such deamination in vitro, not merely to confirm the predictions of the genetic analyses but also so as to facilitate investigation into the kinetics, substrate specificity, and characteristics of the deamination reaction. Here, we describe an important step in that direction and show that the substrate for the deamination is single-stranded DNA and affected by local sequence context.

EXPERIMENTAL PROCEDURES
Plasmids and Bacteria-The pTrc99-and pET-based expression vectors for rat APOBEC1 and its Glu 63 3 Ala mutant, for human APO-BEC2 and for Escherichia coli dCTP deaminase as well as the E. coli host strains have been described previously (10,13). The pTrc99-and pET-based vectors differ both in the nature of the promoter used (pTrc99 uses the trp/lac hybrid promoter whereas pET uses the T7 promoter) and in the length of heterologous peptide linked to the amino terminus of the recombinant protein (9 amino acid with pTrc99 but 34 amino acid with pET (10, 13)).
Oligodeoxyribonucleotides-The oligodeoxyribonucleotides used are listed in Table I.
Preparation of Recombinant APOBEC1-A 2-ml overnight culture of a fresh E. coli transformant grown in LB, 0.2% glucose, 50 g/ml carbenicillin was diluted into 300 ml of the same medium and grown at 37°C to an A 600 of 0.8. The culture was chilled on ice for 20 min and then incubated with aeration for 16 h at 16°C in the presence of inducer (1 mM isopropyl-1-thio-␤-D-galactopyranoside). Cells were harvested by centrifugation, washed, and resuspended in 20 ml H buffer (50 mM Tris, pH 7.4, 50 mM KCl, 5 mM EDTA, 1 mM DTT, and a protease inhibitor mixture (Roche Diagnostics)). Following sonication and ultracentrifugation (100,000 ϫ g for 45 min), the supernatant was passed through a 0.2-m filter and applied to a Sepharose fast-flow Mono Q column (Amersham Biosciences; 10 ml bed volume). After washing with 7 column volumes of buffer H, bound proteins were eluted in buffer H supplemented with increasing salt concentrations (from 50 to 1500 mM * This work was supported in part by a grant from the Arthritis Research Campaign. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ To whom correspondence may be addressed. Cl Ϫ ) collecting 15-ml fractions. Fractions and flow-through were concentrated 100-fold using VivaSpin concentrators (M r 10,000 cut-off) (VivaScience) and assayed. Samples eluting with 1000 -1500 mM salt were pooled and loaded in a volume of 0.5 ml onto a HighPrep Sephacryl S-200 high resolution 16/60 gel-filtration column (Amersham Biosciences) in buffer H. Fractions (1 ml) were collected and concentrated 20-fold before analysis.
TLC-based Deaminase Assay-Samples (2-4 l) were incubated at 37°C for 5 h in 20 l of buffer R (40 mM Tris, pH 8, 40 mM KCl, 50 mM NaCl, 5 mM EDTA, 1 mM DTT, 10% glycerol) containing 75,000 cpm of ␣-32 P-dC-labeled single-stranded DNA (prepared by a 3 min heating to 95°C of the products of asymmetric PCR amplification of the lacI region in pTrc99 performed using [␣-32 P]dCTP (3000 Ci/mmol)). Following phenol extraction and ethanol precipitation, the DNA was digested with Penicillium citrinum P1 nuclease (Sigma) overnight at 37°C (14), and the P1 digests then subjected to thin layer chromatography on polyethyleneimine-cellulose in either (i) 0.5 M LiCl at 4°C or (ii) at room temperature in 1 M CH 3 COOH until the buffer front had migrated 2.5 cm and then in 0.9 M CH 3 COOH:0.3 M LiCl (15). Products were detected using a phosphorimager. Chemical deamination of cytosine in DNA using bisulfite/hydroquinone was performed as described previously (16).
UDG-based Deaminase Assay-Samples (1-2 l) were incubated at 37°C for 2 h in 10 l of buffer R with 5Ј-biotinylated oligonucleotides that either were synthesized with fluorescein at their 3Ј-ends (3 pmol of oligonucleotide per reaction) or were 3Ј-labeled by ligation with [␣-32 P]dideoxyadenylate (100,000 cpm; 0.1 pmol) using terminal deoxynucleotidyltransferase. Reactions were terminated by heating to 90°C for 3 min and oligonucleotides purified on streptavidin magnetic beads (Dynal), washing at 72°C (except in Fig. 2A, where the streptavidin purification step was omitted). Deamination of cytosine in the oligonucleotides was monitored by incubating the bead-immobilized oligonucleotides at 37°C for 30 min with excess uracil-DNA glycosylase (0.5 unit of UDG; enzyme and buffer from New England Biolabs) and then bringing the sample to 0.15 M in NaOH and incubating for a further 30 min. The oligonucleotides were then subjected to electrophoresis on 15% PAGE-urea gels which were developed by either fluorescence detection or phosphorimager analysis.
Western Blotting-Western blot detection of APOBEC1 following SDS-PAGE of samples that had been diluted 20 -100-fold was performed using a goat-anti-APOBEC1 serum (Santa Cruz Biotechnology), developing with horseradish peroxidase-conjugated donkey anti-goat immunoglobulin antiserum (Binding Site, Birmingham, UK). Low range molecular weight markers were from Bio-Rad.

DNA Deamination Assay in Cell
Extracts-Since, of all the APOBEC family members tested, APOBEC1 displayed the most potent mutator activity in the E. coli mutation assay (13), we focused on APOBEC1-transformed E. coli in order to see whether we could detect DNA deamination activity in vitro using cell extracts. Initially, we tried using the UDG-based deaminase assay, working with an oligodeoxyribonucleotide substrate. However, no evidence of deamination was obtained using double-stranded oligonucleotide substrates whereas single-stranded oligonucleotides were rapidly degraded by both APOBEC1 and control extracts (data not shown).
We considered the possibility that the DNA deaminating activity might be specific for single-stranded substrates but that this activity might be masked by nonspecific nucleases. We therefore devised an assay that would be less sensitive to contaminating nucleases (Fig. 1A). The bacterial extracts were incubated with ␣-32 P-dC-labeled single-stranded DNA which was then purified, digested with nuclease P1, and subjected to thin layer chromatography to test for the presence of [␣-32 P]dUMP. Clear evidence of dC deamination in this assay was detected using extracts of E. coli expressing two different APOBEC1 constructs but not from control extracts or from extracts made from E. coli cells carrying plasmids expressing mutant APOBEC1, APOBEC2, or dCTP deaminase (none of which function as DNA mutators in the bacterial assay (13)) (Fig. 1B). The DNA deaminase activity was also evident in APOBEC1 transformants of a mutant E. coli deficient in both dcd-and cdd-encoded deaminases (Fig. 1B, iii). That the prod-  [␣-32 P]dCMP-labeled single-stranded DNA was incubated with the indicated extracts, purified, digested with P1 nuclease, and analyzed by TLC in one of two buffer systems. B, analysis by TLC in either the LiCl (i) or CH 3 COOH ϩ LiCl (panels ii and iii) buffer systems of the assay products of [␣-32 P]dCMP-labeled single-stranded DNA incubated with sonic extracts of E. coli transformants that carry plasmids directing the overexpression of APOBEC1, APOBEC2, a mutant APOBEC1 (harboring an Glu 63 3 Ala substitution) or dCTP deaminase. Controls are provided by extracts from E. coli transformed with vector only (Ϫ) as well as by substrate DNA that has been subjected to chemical deamination using bisulfite. The plasmid/host strain combination used for recombinant protein expression was pTrc99/ E. coli KL16 except where (as indicated) the pET vector was used (in which case the host strain was BL21DE3) or where activity was monitored using the E. coli SØ177 host (which is deficient in both dcd and cdd deaminases). The migration of dUMP, dCMP, and [ 32 P]inorganic phosphate (Pi) markers is indicated. The abundance of wild type and Glu 63 3 Ala mutant APOBEC1 polypeptides in extracts was monitored by Western blot (lower part of panel iii).
In Vitro DNA Deamination by APOBEC1 19584 uct of APOBEC1 action was indeed dUMP is indicated by the co-migration of the radioactive product with dUMP in two distinct buffer systems. These results encouraged us to fractionate the extracts of APOBEC1-transformed E. coli to see whether the DNA deamination activity could be sufficiently separated from nonspecific nucleases so as to be detectable using the oligonucleotide cleavage assay.
Partial Purification-Pilot experiments revealed that ionexchange chromatography could be used to obtain samples of APOBEC1 that contained diminished nonspecific nuclease activity. Thus, while only a proportion of the APOBEC1 polypeptide bound to the Mono Q column (around 10 -20% based on ECL quantitation of the Western blot assay), elution of this bound fraction with Ͼ0.8 M Cl Ϫ yielded a sample that displayed cytosine-DNA deamination activity (as monitored using the TLC-based assay) but containing diminished nonspecific nuclease activity in the UDG-based assay (Fig. 2A). These fractions were then concentrated and subjected to gel filtration (Fig. 2B). The major APOBEC1 peak eluted in fractions 7-9 (corresponding to an M r of 95-140,000) co-eluting with peak DNA deaminating activity. Indeed, with these fractions from the gel filtration column, DNA deamination could now readily be detected by the UDG-based assay using a single-stranded oligonucleotide substrate (although the peak fractions also contained activity that removed the 3Ј-label from the oligonucleotide). Mass spectrometric analysis of proteins in fraction 9 following SDS-PAGE revealed the recombinant APOBEC1 migrating at the position marked by the asterisked in Fig. 2B, i, although the majority of the bands derived from ribosomal proteins.
Characteristics of the DNA Deaminating Activity-The UDGbased deaminase assay was used to monitor the specificity and characteristics of the partially purified APOBEC1 (Fig. 3A). The bands yielding peptide sequences derived from APOBEC1 and ribosomal proteins L1, L2, L6, and L9 and S4 are indicated. M, molecular weight markers. ii, Western blotting for APOBEC1; iii, TLC-based; iv, UDG-based deaminase assays, which were performed on samples of the total clarified bacterial lysate (T) as well as on the eluate from the Mono Q. The UDG-based deaminase assay was performed using 3Ј-␣-32 P-labeled SPM274; note that some of the 3Ј-label is removed during the incubation. The percentage of label associated with the 26-base product of the deamination/cleavage (as opposed to 40-base input oligonucleotide) is indicated.
FIG. 3. Specificity of APOBEC1-mediated DNA deamination using the UDG-based assay. A, schematic representation of the UDG-based deaminase assay. 5Ј-Biotinylated (circle) oligonucleotides that were 3Ј-labeled (asterisk) with fluorescein or [␣-32 P]dideoxyadenylate were incubated with APOBEC1-containing (or control) samples prior to streptavidin purification, UDG treatment, and PAGE-urea analysis. B, partially purified APOBEC1 as well as the Glu 63 3 Ala mutant were tested for their ability to deaminate 3Ј-fluorescein-conjugated oligonucleotide SPM168 using the UDG-based assay. The fluorescence scan of the gel, including controls performed without UDG treatment or without APOBEC1, is shown with the positions of the expected products and size markers indicated. C, time course of SPM168 deamination by partially purified APOBEC1. D, inclusion of RNase A (1 g) or of tetrahydrouridine (THU; 20 nmol, 2 nmol, or 200 pmol) does not inhibit the activity of APOBEC1. E, deaminating activity is specific for a single-stranded substrate. The assay was perfomed using 3Ј-fluorscein-labeled oligonucleotide SPM168 in the presence of the indicated ratio of either oligonucleotide SPM171 (which is complementary to SPM168) or SPM201 (which is not). F, comparison of 3Јfluorescein-labeled oligonucleotides SPM168 (left three lanes) and SPM163 (right three lanes) as targets for deamination by 0.5, 1, and 2 l of APOBEC1, respectively. G, comparison of 3Ј-␣-32 P-labeled oligonucleotides SPM274, SPM275, and SPM276 as targets for deamination by 0.3, 0.6, 0.9, and 1.8 l of APOBEC1.

In Vitro DNA Deamination by APOBEC1 19585
Samples were incubated with a single-stranded oligodeoxyribonucleotide (with or without its complement), which contained internal dC residue(s) and that was 5Ј-biotinylated as well as 3Ј-labeled. After purification on streptavidin, the oligonucleotide was treated with UDG (plus alkali), resulting in site-specific cleavage if the oligonucleotide had been subjected to dC 3 dU deamination. Thus, deamination is read out by the appearance of the specific cleavage product following PAGEurea analysis. The partially purified wild type protein (but not the Glu 63 3 Ala mutant) showed clear activity on a single-stranded oligonucleotide with the cleavage being dependent on the subsequent incubation with UDG (Fig. 3, B and C). The deaminating activity was not inhibited by tetrahydrouridine (which inhibits cytidine deaminases (17)) or by RNase (Fig. 3D). Strikingly (and consistent with our inability to detect deamination on double-stranded oligonucleotide substrates using crude extracts of bacterial transformants (see above)), the activity was blocked if a complementary (but not if an irrelevant) oligonucleotide was titrated into the assay (Fig. 3E). Examination of the cleavage products generated in the UDG-based assay suggests that not all dC residues are equally susceptible to APO-BEC1-mediated deamination. While the data do not allow us yet to identify the ideal in vitro target for APOBEC1-mediated deamination, it is clear that in oligonucleotide SPM168 the third cytosine in the sequence TCCGCG is much less favored than the other two (Fig. 3, B-E). Similarly, evidence of specificity comes from comparing various related oligonucleotides as substrates, where all the data taken together point to deamination being especially disfavored when a purine is located immediately 5Ј of the cytosine (Fig. 3, F and G). DISCUSSION The results described here provide biochemical evidence that APOBEC1-mediated deamination of cytosine to uracil can occur on single-stranded DNA, is dependent on local sequence context, and is abolished by mutation of the APOBEC1 zinc coordination motif. Unlike AID (where genetic evidence indicates that the natural physiological substrate of deamination is DNA (11,12)), the major physiological substrate of APOBEC1 is clearly apolipoprotein B RNA (4,5). Nevertheless, the observation that misexpression of APOBEC1 in transgenic mice predisposes to cancer (18) suggests that APOBEC1-mediated DNA deamination could well be of pathological relevance.
While these results now provide biochemical evidence of DNA deamination in vitro by an APOBEC family member, the DNA deaminating activity that we see here is weak. Thus, given the abundance of APOBEC1 polypeptide in the peak fraction from the gel filtration column, it appears that, on average, each molecule of recombinant APOBEC1 is responsible for in the order of a single deamination event in a 10-min incubation in the UDG-based assay. The weakness of this activity could of course reflect that the assay is suboptimal and might be enhanced by additional components or that the recombinant APOBEC1 is improperly folded, lacking some posttranslational modification or unstable. However, crude calculations indicate that if the ϳ500 molecules of APOBEC1 expressed in each E. coli transformant displayed a DNA deamination activity of this order in vivo and if this were targeted randomly to all cytosine residues in the genome, then this could, in principle, be more than sufficient to account for the several 1000-fold enhanced mutation frequencies seen at the rpoB and other loci in UDG-deficient E. coli following 20 generations of growth (13). Similarly, somatic hypermutation of immunoglobulin variable genes by targeted AID-mediated dC deamination may involve a single and most probably less than 10 targeted dC deamination events in each B lymphocyte cell cycle.
The results reveal a clear sensitivity to the local sequence context of the APOBEC1-mediated dC deamination in the in vitro assay, most readily explained by a bias against a 5Јflanking purine residue. This would accord well with the in vivo data where a near-total restriction to mutation at dC residues with a 5Ј-flanking pyrimidine is seen at the rpoB locus (13) as well as a similar bias at another test locus in E. coli. 2 Curiously, however, these findings contrast with the fact that the target cytosine of APOBEC1 in apolipoprotein B RNA is flanked by a 5Ј-A residue (2,3). It will obviously be interesting to obtain more information about both the structural and kinetic features of the interaction and activity of recombinant APOBEC1 on different RNA and DNA substrates.
The in vitro assay also reveals that APOBEC1 deamination is targeted to single-stranded DNA and, indeed, was undetectable on double-stranded DNA. This specificity for singlestranded DNA is in accordance with the fact that the natural substrate of APOBEC1 is most likely single-stranded RNA (5), and presumably, the same active site in APOBEC1 is used for both types of polynucleotide. Furthermore, spontaneous deamination of cytosine is also much more rapid in single-(as opposed to double-) stranded DNA (19). If this in vitro preference of APOBEC1 for a single-stranded DNA substrate can be extrapolated to the in vivo situation as well as to other DNAmutating APOBEC family members, it could go some way to explaining why AID-mediated DNA changes (somatic hypermutation and class switch recombination) correlate with transcription of the DNA target gene. Clearly, extending on the assays described here and applying them to other APOBEC family members should give valuable insight into the mechanism and specificity of APOBEC family-mediated DNA deamination.