Hydrolysis of O-Acetyl-ADP-ribose Isomers by ADP-ribosylhydrolase 3*

O-Acetyl-ADP-ribose (OAADPr), produced by the Sir2-catalyzed NAD+-dependent histone/protein deacetylase reaction, regulates diverse biological processes. Interconversion between two OAADPr isomers with acetyl attached to the C-2″ and C-3″ hydroxyl of ADP-ribose (ADPr) is rapid. We reported earlier that ADP-ribosylhydrolase 3 (ARH3), one of three ARH proteins sharing structural similarities, hydrolyzed OAADPr to ADPr and acetate, and poly(ADPr) to ADPr monomers. ARH1 also hydrolyzed OAADPr and poly(ADPr) as well as ADP-ribose-arginine, with arginine in α-anomeric linkage to C-1″ of ADP-ribose. Because both ARH3- and ARH1-catalyzed reactions involve nucleophilic attacks at the C-1″ position, it was perplexing that the ARH3 catalytic site would cleave OAADPr at either the 2″- or 3″-position, and we postulated the existence of a third isomer, 1″-OAADPr, in equilibrium with 2″- and 3″-isomers. A third isomer, consistent with 1″-OAADPr, was identified at pH 9.0. Further, ARH3 OAADPr hydrolase activity was greater at pH 9.0 than at neutral pH where 3″-OAADPr predominated. Consistent with our hypothesis, IC50 values for ARH3 inhibition by 2″- and 3″-N-acetyl-ADPr analogs of OAADPr were significantly higher than that for ADPr. ARH1 also hydrolyzed OAADPr more rapidly at alkaline pH, but cleavage of ADP-ribose-arginine was faster at neutral pH than pH 9.0. ARH3-catalyzed hydrolysis of OAADPr in H218O resulted in incorporation of one 18O into ADP-ribose by mass spectrometric analysis, consistent with cleavage at the C-1″ position. Together, these data suggest that ARH family members, ARH1 and ARH3, catalyze hydrolysis of the 1″-O linkage in their structurally diverse substrates.

Mono-ADP-ribosylation is a post-translational modification, in which the ADP-ribose (ADPr) 5 moiety of NAD is transferred to an acceptor protein (1). This modification serves as the mechanism by which several bacterial toxins (e.g. Pseu-domonas exoenzyme S, cholera toxin, diphtheria toxin) exert their effects on mammalian cells (2,3). Mammalian cells also produce endogenous ADP-ribosyltransferases that catalyze reactions similar to the bacterial toxins, specifically, the ADPribosylation of arginine residues in proteins (4). In addition, mammalian cells possess hydrolases that cleave the ADPr-protein linkage, releasing ADPr and regenerating the unmodified protein (5,6). An ADP-ribosyl(arginine) hydrolase, termed ARH1, catalyzes in a stereospecific manner, hydrolysis of the ␣-linkage of arginine-ribose found in ADP-ribosyl(arginine)protein to ADPr and (arginine)-protein (7,8), consistent with the regulation of ADP-ribosyl(arginine)-protein levels by opposing activities of transferases and hydrolases, participating in an ADP-ribosylation cycle (4,9).
Three known members (ARH1-3) of the ARH family of proteins are similar in molecular size (ϳ39 kDa) and amino acid sequence (10). As noted above, ARH1 catalyzes the hydrolysis of ADP-ribose-arginine and also hydrolyzes ADP-ribose linkages to guanidine. The reaction is stereospecific, and only the ␣-anomer at the C-1Љ position of ADP-ribose-arginine is hydrolyzed (8). ARH1 also hydrolyzed the stereospecific C-1Љ-C-2Ј linkage in poly(ADP-ribose). The ARH1 reaction is inhibited by ADP-ribose, but not by phosphoribose, suggesting that the catalytic site recognizes the adenosine moiety (8). The fact that only the guanidine group of arginine is necessary for hydrolysis further supports a critical role for the ADP-ribose as opposed to the arginine in catalysis (8). ARH2, which has considerable structural similarity to ARH1, has no reported activity. In this regard, the vicinal aspartate residues that were critical for ARH1 activity are replaced by an aspartate-asparagine, which may explain the lack of activity.
Both ARH1 and ARH3 appear to catalyze stereospecific hydrolysis of ADP-ribose-arginine and poly(ADP-ribose), respectively. In both cases, the linkage subject to hydrolysis is at the C-1Љ position (Fig. 1). In addition, both the ARH1-and ARH3-catalyzed reactions are inhibited by ADP-ribose. For ARH1, ADP-ribose, but not phosphoribose, is an inhibitor, and ADP-ribose-arginine, but not phosphoribose-arginine, is a substrate, demonstrating that the adenosine moiety is important for substrate recognition. Therefore, the ribose linkage to arginine is insufficient for substrate recognition. We observed that ARH1 as well as ARH3 hydrolyzed O-acetyl-ADP-ribose.
We thought it unlikely that ARH3 and ARH1 would hydrolyze the 1Љ-linkage in poly(ADPr) and ADP-ribose-arginine, respectively, and also act on 2Љ-or 3Љ-linkage in OAADPr. Further, in our current studies, ADP-ribose was a potent inhibitor, but not 2Љ-, 3Љ-N-acetyl-ADP-ribose. In addition, ARH1 hydrolyzed the C-1Љ derivative of ADP-ribose(arginine) as well as OAADPr, suggesting that the C-1Љ position is the preferred site of cleavage. Here, we propose that, analogous to its poly(ADPr)glycohydrolase activity, ARH3 hydrolyzes 1Љ-OAADPr, rather than the 2Љ-or 3Љ-isomer, and that 1Љ-, 2Љ-, and 3Љ-OAADPr exist in equilibrium at neutral and basic pH values, thus generating 1Љ-OAADPr for hydrolysis by ARH3. These studies demonstrate that ARH1 and ARH3 show a similar preference for substrates and that hydrolysis proceeds with attack at the C-1Љ position. Prior studies that showed the presence of C-2Љ and C-3Љ O-acetyl-ADP-ribose are in agreement with our current findings that at neutral pH, the equilibrium favors these two isomers and not C-1Љ O-acetyl-ADP-ribose.
Identification of Isomers of O-Acetyl-ADP-ribose-For identification of OAADPr isomers, 1 M OAADPr in 50 mM potassium phosphate (pH 9.0), 10 mM MgCl 2 , and 5 mM DTT (total volume, 200 l) were incubated (5 min at room temperature), before separation of isomers using RP-HPLC and analysis by MALDI-TOF mass spectrometry as described previously (11). Briefly, before mass spectrometry, 10 mg of ␣-cyano-4-hydroxycinnamic acid (Bruker Daltonics, Billerica, MA) was mixed in 1 ml of water/acetonitrile (50:50, v/v) containing TFA. Then, 40 pmol of OAADPr in 1 l was added to 1 l of matrix solution (described above) on a stainless steel target plate and

O-Acetyl-ADP-ribose Isomers as Substrates for ARH3
air-dried. Negative ion MALDI-TOF mass spectra were acquired using the 4700 Proteomics analyzer (Applied Biosystems, Foster City, CA) operated in reflection mode.
Kinetics of ARH3 Hydrolysis-Rates of OAADPr hydrolysis by purified recombinant ARH3 (1.5 pmol) were determined using the indicated concentrations of OAADPr in 50 mM potassium phosphate (pH 7.0), 10 mM MgCl 2 , and 5 mM DTT (total volume, 200 l) incubated for 20 min at 30°C. ADPr was quantified by RP-HPLC; K m and V max were calculated from Lineweaver-Burk double reciprocal plots.

RESULTS
pH-dependent Hydrolysis of OAADPr by ARH3-We compared ARH3 hydrolysis of OAADPr (produced from [ 14 C]NAD and acetylated histone H3 peptide by human SIRT1 (see "Experimental Procedures") at pH 5.0, 7.0, and 9.0. At 5.0, RP-HPLC analysis revealed two peaks of OAADPr, consistent with 2Љ- (Fig. 2A, peak B) and 3Љ-OAADPr ( Fig. 2A, peak A) (20,21). At pH 5.0, OAADPr hydrolysis in the presence of ARH3 was not above background ( Fig. 2A). After incubation with ARH3 at pH 7.0, the remaining substrate contained not only 2Љ-and 3Љ-OAADPr, but also an additional isomer (Fig. 2B, peak C). In OAADPr hydrolase assays containing ARH3 at pH 7.0, 119  pmol of OAADPr, containing all three isomers, were hydrolyzed, with accumulation of 127 pmol of ADPr (Fig. 2B). At pH 9.0, ARH3 activity was greater than that at pH 7.0, and amounts of all three isomers were reduced, with the expected increase in ADPr (Fig. 2C). Peak C, not observed at pH 5.0, was present at low levels at pH 7.0 and higher abundance at pH 9.0 (Fig. 2, A-C). To assess nonenzymatic OAADPr hydrolysis and determine whether the greater activity at pH 9.0 was enzymatic or nonenzymatic, OAADPr was incubated at pH 9.0 and 7.0. At pH 9.0 (Fig. 2E), nonenzymatic OAADPr hydrolysis was clearly faster than at pH 7.0 (Fig. 2D), and all three peaks were reduced. After nonenzymatic OAADPr hydrolysis, ARH3 activity at pH 9.0 was still higher than at pH 5.0 or pH 7.0 (subtracted nonenzymatic hydrolysis of OAADPr was significant only at pH 9.0). The mass of 558 m/z nonenzymatic product (data not shown) was consistent with ADPr. As shown in Fig. 2A, there was no ARH3 activity at pH 5.0. Two possible causes of deactivation of ARH3 at pH 5.0 were (i) denaturation of ARH3 due to low pH, and (ii) a pH below the optimal range for OAADPr hydrolysis by ARH3. To assess denaturation, ARH3 was incubated at pH 5.0 for 30 min and then at pH 7.0 for 5 min, revealing dramatically greater hydrolase activity than that at pH 5.0 (Fig. 2F). Activity of ARH3 that had been incubated at pH 5.0 for 30 min followed by 5 min of incubation at pH 7.0 was the same as that measured at pH 7.0 (Fig. 2F), indicating that the lower activity at pH 5.0 was not due to irreversible ARH3 denaturation. Thus, ARH3 activity correlated with the relative abundance of peak C at different pH values, and all three isomers were decreased by ARH3 activity, consistent with their interconversion.
Interconversion of O-Acetyl-ADP-ribose Isomers-Because the activity and pH-dependence of ARH3 correlated with the relative abundance of peak C, we further characterized that material. Peak A (3Љ-OAADPr) predominated under pH 2.0 (Fig. 3A), whereas at pH 9.0 three peaks (A, B, and C) were present (Fig. 3B) even if isolated peak C was used as the starting material. Upon addition of 0.05% (v/v) TFA to a solution of OAADPr after being incubated at pH 9.0, peak C was completely converted to peaks A and B (Fig. 3C). This indicated that peak C was in equilibrium with peaks A and B, and, importantly, formation of peak C was reversible. The data are consistent with pH-dependent interconversion of the OAADPr isomers. The mass of each of the three peaks was 600 m/z (Fig. 4, A-C), consistent with peak C being an isomer of OAADPr, with acetate perhaps at the 1Љ-position.  Fig. 1C were subjected to MALDI-TOF mass spectrometry to reveal a peak at 600 m/z, consistent with the predicted mass of OAADPr, in each sample. The experiment was repeated three times with similar results.
acetyl-ADPr were ϳ60 and ϳ200-fold that of ADPr (Table 1 and Fig. 5). Neither 2Љ-nor 3Љ-N-acetyl-ADPr was a substrate for ARH3 (data not shown). If the preferred substrate of ARH3 is 1Љ-OAADPr, it is predicted that ADPr would be better accommodated within the active site than either 2Љ-or 3Љ-OAADPr. Therefore, inhibition data are consistent with ARH3 preferentially binding to and hydrolyzing 1Љ-OAADPr.
pH Dependence of Hydrolysis of O-Acetyl-ADP-ribose and ADP-ribose-arginine by ARH1-Given that the reactions catalyzed by ARH3 and ARH1 are stereospecific at the C-1Љ position and prefer the ␣-anomer of poly(ADP-ribose) (ARH3, ARH1) and ADP-ribose-arginine (ARH1), we hypothesized that ARH1 would also hydrolyze O-acetyl-ADP-ribose at the C-1Љ position, and we expected that for ARH1, as for ARH3, conditions favoring the abundance of the C-1Љ would enhance hydrolysis. In contrast, hydrolysis of ADP-ribose-arginine by ARH1 should not depend on the alkaline conditions that favor the C-1Љ form of O-acetyl-ADP-ribose. Indeed, effects of pH on OAADPr hydrolase activity of ARH1 and ARH3 were similar (Fig. 6). As the hydrolysis we saw with ARH1 was less than a single turnover, the observed pH dependence of OAADPr hydrolysis was not a steady-state phenomenon, whereas differences with ARH3 were at steady state, thus no direct comparisons of rate are possible.
ARH1 also hydrolyzes ADPr-arginine, which exists as both ␣and ␤-anomers of the C-1Љ-guanidine linkage. The ␣-anomer predominated at pH 5.0 and amounts of ␣and ␤-anomers were similar at pH 9.0, consistent with pH-dependent anomer-ization (Fig. 7A). ARH1 activity was minimal at pH 5.0, but in contrast to the findings with OAADPr, its hydrolysis of ADPrarginine at pH 7.0 appeared greater than that at pH 9.0 (Fig. 7).   Fig. 4 before RP-HPLC separation and quantification of ADPr. Blank data without enzyme were subtracted from each of these points. The experiment was repeated three times with similar results.

Substrate
IC 50 mM ADP-ribose 0.012 Ϯ 0.003 2ЉN-Acetyl-ADP-ribose Ն0.71 3Љ-N-Acetyl-ADP-ribose Ն2.3 At pH 7.0, anomerization was relatively slow, and preferential hydrolysis of the ␣-anomer by ARH1 was seen with preservation of ␤-anomer. At pH 9.0, anomerization was relatively rapid, with depletion of both ␣and ␤-anomers, because hydrolysis of ␣-anomer was accompanied by continuing anomerization. ARH1 activities at pH 7.0 and 9.0 were constant with time (Fig. 7B), and rates of ARH1-catalyzed hydrolysis correlated with the abundance of the ␣-anomer at the 1Љ-position for ADPr-arginine. Thus, as expected, hydrolysis of O-acetyl-ADP-ribose and ADP-ribose-arginine appears to be dependent in part on pH conditions that favor the substrate form of the molecule. 18 O Incorporation into ADP-ribose-To obtain evidence that ARH3 cleaved OAADPr at the C-1Љ position, OAADPr was hydrolyzed by ARH3 in H 2 18 O. The reaction product, ADPr, was analyzed by mass spectrometry to quantify 18 O incorporation into ADPr. The major ADPr product (85% of total 558.15 ϩ 560.15 m/z) of ARH3-catalyzed hydrolysis incorporated one 18 O (560.15 m/z) (Fig. 8A). Reactions with heat-inactivated ARH3 yielded nonenzymatically hydrolyzed ADPr (m/z ϭ 558.15) with no significant 18 O incorporation above natural abundance (Fig. 8B). As expected, a 30-min reaction of heatinactivated ARH3 displayed some nonenzymatic 18 O incorporation (14% of total 558.15 ϩ 560.15 m/z) (Fig. 8C) 18 O incorporation was measured by the relative peak heights at each m/z. As a control, ADPr was incubated with heat-denatured ARH3 in the same mixture for 0 min (B) to confirm the natural distribution of heavy isotope or 30 min (C) to assess nonenzymatic exchange of the 1Љ-hydroxyl at 30°C. Detection and measurement of 18

O-Acetyl-ADP-ribose Isomers as Substrates for ARH3
lyzes attack of [ 18 O]water at the 1Љ-carbon of OAADPr, with incorporation of 18 O into ADPr at the 1Љ-position. Since only the 1Љ-position is exchangeable, the 18 O must be at the 1Љ-position.

DISCUSSION
Herein, we found not only 2Љ-and 3Љ-OAADPr at pH 9.0, but also substantial quantities of a third peak, of mass (600 m/z) that was consistent with 1Љ-OAADPr. At acidic pH 5.0, 3Љ-OAADPr predominated, whereas at higher pH, all three forms of OAADPr were observed. Rapid interconversion of 2Љ-and 3Љ-OAADPr had been reported (21), and the amount of peak C at pH 9.0 was greater than that at pH 7.0, whereas peak C was absent at pH 5.0. All three peaks appeared to be in rapid equilibrium at pH 9.0. ARH3 hydrolysis of OAADPr at pH 9.0 was significantly faster than that at pH 7.0. Thus, ARH3 activity correlated with peak C concentration, although other factors related to protein stability and active site chemistry may also affect hydrolase activity at acidic or alkaline pH.
ARH3 activity was not inhibited by d-ribose 5-phosphate, AMP, ADP, or ␤-NAD (11). ADPr was an effective inhibitor, likely due to competition for substrate binding within the ARH3 catalytic active site (11). In contrast to ADPr (11), 2Љ-and 3Љ-N-acetyl-ADPr were poor ARH3 inhibitors. These data suggested that the 2Љ-and 3Љ-acetyl-ribose rings were, relative to ADPr, accommodated poorly in the ARH3 active site, as might be expected if the preferred substrate were 1Љ-OAADPr.
ARH3 and ARH1 molecules share substantial amino acid sequence similarity (10) and stereospecificity (6 -8). ARH1 catalyzes hydrolysis of the ␣-anomer of ADP-ribosyl-arginine at the ribose C-1Љ position (6 -8). ARH1 also cleaved OAADPr to produce ADPr but at Ͻ1% the rate of ARH3 (11). In the experiments reported here, ARH1 activity was pH-dependent; pH 7.0 was preferred for ADPr-arginine hydrolysis, whereas pH 9.0 was optimal for OAADPr hydrolysis. These data are consistent with the hypothesis that the apparent pH optimum for ARH1 is, in part, determined by substrate availability.
If ARH3 hydrolyzed OAADPr at C-1Љ, 18 O should be incorporated into the OAADPr product at the 1Љ-hydroxyl. Otherwise, 18 O attack occurred on a carbon atom of an acetyl group, resulting in no 18 O incorporation into ADPr (scheme in Fig. 8). Indeed, one 18 O was incorporated into ADPr (560.15 m/z) in contrast to the result of control experiments with inactive ARH3, which showed basal uncatalyzed 18 O incorporation. In addition, when ADPr (560.15 m/z) was acidified in natural abundance water, 18 O exchanged out rapidly, in agreement with previous research (27). These data are consistent with OAADPr hydrolysis catalyzed by ARH3 occurring at the C-1Љ position.
In summary, we report that three isomers of OAADPr are in pH-dependent equilibrium: under acidic conditions 3Љ-OAADPr is most abundant, 3Љ-and 2Љ-OAADPr predominate at neutral pH, and a third isomer that is consistent with 1Љ-OAADPr predominates in a basic milieu. ARH3 activity with OAADPr as substrate was optimal at pH 9.0, where the putative 1Љ-isomer is the greatest. ARH3 activity was not observed at pH 5.0, where the 3Љ-isomer predominates. Consistent with the conclusion that hydrolysis occurs at the 1Љ-position of OAADPr, 2Љ-and 3Љ-N-acetyl-ADPr were much less effective inhibitors of ARH3 activity than ADPr. Moreover, the alkaline pH optimum for OAADPr hydrolysis was similar for ARH1 and ARH3, but ARH1 cleavage of ADPr-arginine, which does not have 2Љ-and 3Љ-isomers, was optimal at neutral pH. Based on these data, we propose that ARH3 specifically hydrolyzes the 1Љ-OAADPr isomer, i.e. a bond similar to those hydrolyzed by ARH3 in poly(ADPr). As the OAADPr produced in the Sir2catalyzed NAD-dependent histone/protein deacetylase reaction is reported to participate in several biological processes, including formation of silencing complexes, ion channel gating, and energy metabolism (15)(16)(17)(18)(19), ARH3 may influence several signaling pathways through degradation of OAADPr.