Deamidation of HPr, a phosphocarrier protein of the phosphoenolpyruvate:sugar phosphotransferase system, involves asparagine 38 (HPr-1) and asparagine 12 (HPr-2) in isoaspartyl acid formation.

Histidine-containing protein, HPr, of the phosphoenolpyruvate:sugar phosphotransferase system in Escherichia coli, when incubated at elevated temperatures forms many species of protein. The two major species are HPr-1 and HPr-2, which have been shown to lack one or two amides, respectively (Anderson, B., Weigel, N., Kundig, W., and Roseman, S. (1971) J. Biol. Chem. 246, 7023-7033). The formation of HPr-1 and HPr-2 is shown to be pH-dependent and does not occur readily below pH 6. Investigation of the identities and properties of the two residues that deamidate involved creation of site-directed mutants at the 6 glutamine and 2 asparagine residues of HPr; description of their deamidation species by isoelectric focusing; determination of their relative antibody binding properties; assay of their phosphoacceptor and phosphodonor activities; characterization of tryptic and V8-protease peptides; obtaining two-dimensional nuclear magnetic resonance spectra of HPr, HPr-1, and several mutants. It was determined that the sequential deamidation of Asn-38 and Asn-12 yields HPr-1 and HPr-2. Both residues exist as Asn-Gly pairs, and both deamidations probably form isoaspartyl acid. HPr from Bacillus subtilis and Staphylococcus carnosus which also have Asn-Gly at residues 38 and 39 form HPr-1 species presumably by deamidation. HPr from Streptococcus faecalis which does not have Asn-38 does not form a HPr-1 species. The E. coli mutant HPrs, N12D and Q51E, residues that may be involved in the active site, had impaired phosphohydrolysis properties and decreased phosphoenolpyruvate:sugar phosphotransferase system activity.

Histidine-containing protein, HPr, of the phosphoen-o1pyruvate:sugar phosphotransferase system in Escherichia coli, when incubated at elevated temperatures forms many species of protein. The two major species are HPr-1 and HPr-2, which have been shown to lack one or two amides, respectively (Anderson, B., Weigel, N., Kundig, W., and Roseman, S . (1971) J. BioZ. Chem. 246,[7023][7024][7025][7026][7027][7028][7029][7030][7031][7032][7033]. The formation of HPr-1 and HPr-2 is shown to be pH-dependent and does not occur readily below pH 6. Investigation of the identities and properties of the two residues that deamidate involved creation of site-directed mutants at the 6 glutamine and 2 asparagine residues of HPr; description of their deamidation species by isoelectric focusing; determination of their relative antibody binding properties; assay of their phosphoacceptor and phosphodonor activities; characterization of tryptic and V8-protease peptides; obtaining two-dimensional nuclear magnetic resonance spectra of HPr, HPr-1, and several mutants. It was determined that the sequential deamidation of Asn-38 and Asn-12 yields HPr-1 and HPr-2. Both residues exist as Asn-Gly pairs, and both deamidations probably form isoaspartyl acid. HPr from Bacillus subtilis and Staphylococcus carnosus which also have Asn-Gly at residues 38 and 39 form HPr-1 species presumably by deamidation. HPr from Streptococcus faecalis which does not have Asn-38 does not form a HPr-1 species. The E. coli mutant HPrs, N12D and QSlE, residues that may be involved in the active site, had impaired phosphohydrolysis properties and decreased phosphoeno1pyruvate:sugar phosphotransferase system activity. The histidine-containing phosphocarrier protein, HPr, of the phosphoeno1pyruvate:sugar phosphotransferase system (PTS)' is well described for a number of bacterial species. Tertiary structures of HPr from ~s c h e r i c h~a coli (Klevit and Waygood, 1986;El-Kabbani et al., 1987;Hammen et al., 1991;van Nuland et al., 1992), Bacillus subtilis (Wittekind et al., 1990(Wittekind et al., ,1992Herzberg et aE., 1992), Staphylococcus aurew (Kalbitzer et al., 1991), and Streptococcus faecalis (Jia et ai., 1993) have been reported. HPr has been considered a stable protein (Kundig et al., 1964), yet upon heating or storage, HPr from E. coli is susceptible to two major deamidation events to yield HPr-1 and HPr-2 (Anderson et al., 1971;Waygood et al., 1985). It was reported that both HPr-1 and HPr-2 had impaired activity in a sugar phospho~lation assay, and careful amino acid analysis indicated deamidation caused these two species (Anderson et al., 1971). However, Waygood et al. (1985) showed that HPr-1 retained full activity while HPr-2 had impaired activity with both enzyme I and enzyme IIMt'.
In addition, HPr-2 has impaired phosphohy~olysis properties in that the phosphohistidine, His(P)-15, is more stable. Neither HPr-1 nor HPr-2 had altered binding to several HPrspecific monoclonal antibodies , which suggests that conformational changes caused by deamidation are local (Sharma et al., 1991).
Deamidation in proteins is a widespread occurrence; generally a non-enzymatic event, but one for which there may be a repair system, protein carboxymethyl transferase. Deamidation is suggested as a potential biological clock in aging; a signal that initiates protein degradation; an important contributing factor in protein denaturation; and both enzymatic deamidation and carboxylmethylation are important features in signal transduction in chemotaxis in E. coli (see reviews, Clarke (1985) and Wright (1991a)). The most rapidly occurring deamidation events occur where asparaginyl-glycine pairs are found (Wright, 1991b), and this deamidation involves a @-aspartyl shift mechanism with a number of potential products including D-and L-isomer forms of aspartic and isoaspartic acid. The major species produced are the L-isomers of The abbreviations used are: PTS, phosphoeno1pyruvate:sugar phosphotransferase system; HPr, histidine-containing phosphocarrier protein of the PTS; MES, 2-(N-morphi1ino)ethanesulfonic acid; DQF-COSY, two-dimensional double-quantum filtered coherence spectroscopy; TOCSY, two-dimension^ total coherence spectroscopy; NOESY, two-dimensional nuclear Overhauser enhancement spectroscopy; Mtl, mannitol; HPLC, high performance liquid chromatography.
__ aspartic and isoaspartic acid in an approximate 1:3 ratio (Bornstein and Balian, 1970;Geiger and Clark, 1987). This mechanism is not the only mechanism that leads to deamidation of asparagines (Kossiakoff, 1988;Wright, 1991a). In general, deamidation of asparaginyl residues is increased by alkali pH, phosphate and arsenate anions, and inhibited by secondary and tertiary structure. Glutamine deamidations are not as well characterized (Robinson and Rudd, 1974;Wright, 1991a).
In HPr, there are two asparaginyl-glycine pairs, Asn-12-Gly-13 and Asn-38-Gly-39; the former is located in a loop, which contains the active site; the latter is part of a @-turn structure (Klevit and Waygood, 1986;El-Kabbani et aE., 1987). There are also 6 glutamine residues: Gln-3, . When HPr is boiled in phosphate buffer at pH 7.5, conditions which lead to rapid deamidation, there are many species of HPr formed with PI values that are more acidic. The major species formed are HPr-1 and HPr-2, and at least for HPr-1 the species appears identical whether formed under the above conditions or during long term storage (Waygood et al., 1985). HPr-1 was studied by ~o -~m e n s i o n a l NMR, and it was concluded that there were two glutamine deamidations, Q57E and Q71E, in this deamidated species (Klevit et al., 1988). However, the spectral evidence was not conclusive, and other approaches appeared necessary to confirm this conclusion, and to identify the events leading to HPr-2. During the course of this work, corrections were made to some of the NMR spectral a s s i~m e n t s (Hammen et ai., 1991;van Nuland et al., 1992), one of the major consequences of which was a reassessment of the HPr-1 spectra, and the conclusions derived from such spectra.
There are several reasons for investigating the deamidation of HPr. During HPr purification, storage, and manipulation, the deamidated product, HPr-1, is readily formed and it is prudent to know the location of such an event in order to assess its effect on HPr function and structure. The change that causes HPr-2 formation affects both phosphohy~olysis and kinetic functions of HPr (Waygood et al., 1985), and thus identification may offer some insight into the mechanism of the active site. HPr offers another opportunity to study deamidation in a protein in some detail. Lastly, HPr is proposed to be involved in chemotaxis that is dependent upon the PTS (Lengeler and Vogler, 1989;Gfibl et al., 1990), and glutaminyl deamidation and glutamyl methyl transfer are activities of CheB and CheR proteins that are involved with other bacterial chemotactic events (Stock et al., 1989). Deamidated HPr could be a species of HPr in uiuo to provide the proposed linkage between the PTS and the chemotactic response.
In this report, using a variety of approaches, it will be shown that two asparagine residues, Asn-38 and Asn-12, are sequentially deamidated to form HPr-1 and HPr-2 and that isoaspartyl is formed in both cases.

EXPERIMENTAL PROCEDURES
Materials-Enzymes used in the manipulation of plasmid DNA were obtained from either Pharmacia LKB Biotechnology Inc. or New England Biolabs. Ampholytes were obtained from Pharmacia. Radioactive chemicals were obtained from Du Pont-New England Nuclear. Q-Sepharose was obtained from Pharmacia. Acetonitrile was HPLC grade from Merck. Guanidine HC1 was enzyme grade from Schwarz-Mann Biochemical. Trypsin (L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated) and endoproteinase GluC (S. aureus strain V8-protease) were obtained from Worthington. D20, 99.96 atom % was obtained from Sigma.
Protein Preparations-HPr from E. coli was obtained initially as described by Waygood and Steeves (1980), and subsequently HPr and HPr mutants were obtained from cloned sources as described (Sharma et al., 1991;Anderson et al., 1991). HPr-1 and HPr-2 were derived as described by Waygood et al. (1985), and an improved separation was achieved using Q-Sepharose in place of DEAE-cellulose. HPr, Q4K, was obtained from Salmonella typhimurium strain SB3093 (Beneski et at., 1982) as described by Waygood et a t . (1987). B. subtilis HPr was obtained as described by Reizer et al. (1989); S. faecalis HPr was obtained as described by Kalbitzer et al. (1982) a n d S t a p h y~c~c u s carnosus HPr was obtained as described by Eisermann et al. (1991).
Enzyme I was obtained as described by Anderson et al. (1991). HPrspecific monoclonal antibodies were obtained as described by Waygood et a l . (1987). P h o s p h~n o l p~v a t e carboxykinase was a gift from Dr. Hughes Goldie (University of Saskatchewan).
Demonstration of Deamidation Using Isoelectric Focusing Gels-HPr and mutant HPrs were incubated in different buffers: pH 3 and 5, citrate-phosphate buffer; pH 6.0, MES; pH 7.5, potassium phosphate, Tris-HC1, and HEPES buffer; pH 8.0, potassium phosphate buffer; pH 9.0 and 9.5 borate-boric acid buffer; pH 10.0, borate-NaOH buffer. All buffers were adjusted at room temperature. The procedure used and the isoelectric focusing was as described by Waygood et al. (1985) Peptide Analysis-Samples of various HPrs (1 mg) were incubated with 0.1 mg of the proteolytic enzymes in 0.2 ml of 50 mM ammonium bicarbonate, pH 8.0. For the V8 protease incubations, 1 mM EDTA was included. Incubations were for 1.5, 3, 6, and 20 h. After incubation, the samples were brought to about pH 6.5 with the addition of HC1, and stored frozen (-20 "C). Separation of the peptides was performed on a reverse-phase column (Serva RP8, 10 mm, 4.6 mm X 250 mm) using a linear gradient at 1 ml/min. The gradient was solvent A (0.1% trifluoroacetic acid in water) 100 to 60% in 60 min; the solvent B was 60% acetonitrile in water. The effluent was monitored at 230 nm, and samples were collected manually. Samples applied were derived from approximately 1-2 nmol of protein.
Amino Acid Analysis-Amino acid analysis was performed on 200-400 pmol of peptide using the methodology Heindrikson and Meredith (1984) by Dr. Meyer at the Amino Acid Analysis Facility, University of Ruhr-Bochum.
Two-dimensional NMR Spectroscopy-Purified proteins were dialyzed against 5 mM potassium phosphate buffer, pH 6.5, containing 0.01 mM EDTA. The solutions were lyophilized and dissolved in 90% H20/10% D20 (or in 99.96% atom D20) to give samples that were -4 mM in protein, 50 mM in buffer, and 0.1 mM in EDTA. Twodimensional NMR spectra were acquired on a Bruker AM-500 spectrometer at 30 "C. They were obtained in the pure-phase absorption mode using time proportional phase incrementation (Marion and Wiithrich, 1983). Data were processed using the FELIX 1.0 (Hare Research, Woodinville, WA) on a Silicon Graphics 4D-35 workstation. Data sets were 600 by 2,000 complex points, and were zero-filled in each dimension to produce signal matrices that were 2,000 points X 2,000 points. A chemical shift of 4.8 ppm, relative to tetramethylsilane, was assigned to the solvent signal, to be used as a reference. DQF-COSY spectra were acquired using composite pulses to give signal filtering (Mueller et al., 1986). The MLEV-17 mixing sequence ( Bax and Davis, 1985) was used for DzO TOCSY spectra with mixing of 70 ms. Clean TOCSY spectra (Greisinger et al., 1988) were acquired for HPr and HPr-1 with mixing times of 46 ms. The NOESY spectrum of HPr-1 was acquired using a 32-step phase cycle and a mixing time of 100 ms. Scalar contributions to the signal were suppressed by randomization of the mixing time.
Other Methods-Assays for HPr were performed using the general methods described by Waygood et al. (1979) and Anderson et al. (1991). Antibody binding studies were carried out using the competition assay described by Waygood et al. (1987). The production of [32P]HPr was as described by Anderson et al. (1991). Phosphohydrolysis was carried out as described by Waygood et al. (1985). The protein concentration of HPr was measured using the following FIG. 1. pH and temperature dependence of HPr deamidation. HPr (0.1 mg/ml) was heated at various temperatures and pH values in a thin-walled glass test tube. Samples (0.05 ml) were taken at the times (minutes) indicated, cooled, and applied to an isoelectric focusing gel. The gels contained pH 3-10, 3-5, and 5-7 ampholytes (1:2:2) and were run for 2 h at 8 watts (voltage limiting at 1000 V). The 0.01 M buffers used were: pH 6, MES; pH 7.5, potassium phosphate; pH 10, borate. The gels shown had the cathode on top. H indicates a sample of HPr, which in some cases contains HPr-1 as indicated. methods: the Lowry method (Lowry et al., 1951) using bovine serum albumin and wild-type HPr as standards; the lactate dehydrogenase depletion assay (Waygood et al., 1979); the spectrophotometric determination, -A225 = 144 pg/ml/OD (Waddell, 1956).

RESULTS
Deamidation of HPr-Deamidation reactions in proteins occur more readily at alkali pH and are accelerated by the presence of anions such as phosphate (Robinson and Rudd, 1974;Wright, 1991a). Deamidation of HPr was assessed by heating HPr in buffers of different pH values and resolving the samples on isoelectric focusing gels2 ( Fig. 1). HPr at pH 3.0 and pH 5.0 did not deamidate to any significant extent for the time period studied, 2 h at 80 "C (results not shown). Even at pH 6.0 deamidation was not observed, but for increasing pH values, the rate of deamidation increased ( Fig. 1, B, D, and E ) . The rate of deamidation in phosphate buffer was somewhat faster than in HEPES and Tris-HC1 buffers at the same concentration and pH (results not shown). At all pH values, except pH 10 ( Fig. ID), the total amount of HPr observed on the isoelectric focusing gels (Figs. 1-3) remained approximately constant with time. At pH 10, the amount of HPr found on the gel declined with time indicating that alkalidependent chain cleavage may occur at the two Asn-Gly pairs found in HPr (see Wright (1991a)). The rate of deamidation at 60 "C ( Fig. IA) was clearly slower than would have been anticipated by comparison to the 80 and 100 "C rates (Fig. 1, B and C). Also at 60 "C, the formation of HPr-2 was not observed. This suggests that some conformational change or partial unfolding enhances the rate of deamidation. When The isoelectic focusing gels presented in this report were acquired over a number of years. The absolute distances between bands varies from gel to gel, but the relative positions and patterns are reproducible. The standard preparations H and H1 are often not the identical preparation used to test a parameter, and thus differences in the amount of HPr-1 are apparent. deamidation occurred, the major species formed were HPr-1 and HPr-2, however, as has previously been noted (Waygood et al., 1985) there were other species formed which are particularly evident following boiling (Fig. lA). Lower temperatures appeared to produced fewer observable species (Fig. 1, B and C). However, deamidation during long term storage of frozen solutions of HPr in 10 mM Tris-HC1 buffer, pH 7.5, resulted in many species of HPr (Waygood et al., 1985).
Deamidation Properties of HPr Mutants-A series of mutants were created for all 6 glutamine and 2 asparagine residues. The initial mutagenesis was carried out at residues Gln-57 and Gln-71 because of the identification of these residues as the sites of deamidation (Klevit et al., 1988). Both these residues were proposed to deamidate in HPr-1, and because mutation of these residues did not completely reproduce the HPr-1 properties, other residues were mutated. The rationale for mutant construction was as follows. If the amide residue is not involved in the formation of either HPr-1 or HPr-2, then a neutral change in the residue (e.g. to serine or threonine) should not alter the formation of HPr-1 and HPr-2, nor change the PI values of HPr and its deamidated species. A change to an acidic residue would allow the formation of HPr-1 or HPr-2, but would lead to major deamidation species with PI values lower than those observed for HPr-1 and HPr-2. If the residue was involved in the formation of HPr-1, then the neutral change would limit the subsequent formation of the major deamidated species to only one. A change to an acidic residue would mimic either HPr-1 and allow HPr-2 formation, or be equivalent to the deamidation that leads to HPr-2, and allow the formation of a species with a PI equivalent to HPr-2 following the deamidation at the HPr-1 residue(s). In addition, the appearance of doublet bands on the isoelectric focusing gels, which are characteristic of HPr-2 preparations was closely followed. These doublets (examples seen in Figs. 1 and 2) appear to be characteristic of isoaspartyl and aspartyl residue formation. There are several well characterized deamidations in calbindin and cholera toxin, where isoelectric focusing gels show doublet bands with approximately the 1:3 or 1:4 ratio (Chazin et al., 1989;Spangler and Westbrook, 1989).
The results of heating experiments to induce deamidation in preparations of Q57E, Q71E, and the double mutant Q57E+Q71E are shown in Fig. 2. Both Q57E and Q71E had PI values that were intermediate between HPr and HPr-1, and the double mutant had a PI that was very close to HPr-1 ( Fig. 2A). These results appeared to confirm the identification of HPr-1 (Klevit et at., 1988). However, neither Q57E nor Q71E when heated yielded major species with PI values equivalent to HPr-1 or HPr-2 (Fig. 2, B and C). The double mutation, Q57E+Q71E, which has a PI close to HPr-1, when heated gave a doublet band with a PI similar to that of HPr-2, but also a doublet band with a PI lower than that of HPr-2 (Fig. 2B). All three mutant HPrs gave deamidation species that were consistent with the proposal that none of these species are equivalent to HPr-1, and they simply led to major deamidated species with lower PI values, i.e. HPr-1 and HPr-2 formation was occurring in a more acidic species of HPr.
Other mutant HPrs were examined with respect to their PI values and their deamidation properties. Some of the results are presented in Fig. 3 and other results have been presented by Sharma (1992). The following neutral mutations, Q3S, Q4S, and Q51T, gave HPrs with PI values and deamidation patterns identical to HPr. In addition, Q3K and Q4K, which have higher PI values than that of HPr, gave deamidation patterns which resembled HPr, but with species for which the PI values were shifted to higher values. This confirms earlier reports that HPr-1 and HPr-2 generation in Q4K was similar to wild type (Beneski et al., 1982;Waygood et al., 1987). Q21T has a PI that is higher than that of HPr, for reasons that are not understood (Sharma, 1992). Q3E, Q21E, NlZD, and N38D all had PI values essentially identical to HPr-1 (Fig. 3). The surprising result was that all four mutants yielded only one other major deamidated species which in each case had a PI equivalent to HPr-2 (Fig. 3) and did not display the many species observed when HPr was similarly treated. The double mutation, N12D+N38D, which was produced after the identification of both deamidation sites, had a PI similar to HPr-2 and did not produce other deamidation products These results provided evidence against the involvement of Gln-4, Gln-51, Gln-57, and Gln-71 in the formation of HPr-1 and HPr-2. The results with Q3S suggested that Gln-3 was not involved, but the Q3E mutant behaved in a manner indistinguishable from QZlE, N12D, and N38D. The mutant Q4E was not obtained.
Antibody Binding-Three monoclonal antibodies for HPr, Je142, Je144, and Je1323, have been extensively investigated with respect to the effects of residue replacement in HPr upon antibody binding Sharma et al., 1991;Sharma, 1992). The deamidations that yield HPr-1 and HPr-2 do not affect the binding of these three antibodies . Competition binding experiments revealed that, with the exception of Q71E, all the mutants of HPr with acidic residues, Q3E, Q21E, Q51E, Q57E, NED, and N38D, did not alter the binding of all three antibodies. Q71E gave a 100-fold decrease in the binding to Je142 (Sharma et al., 1991) as did the double mutant Q57E+Q71E. Similarly, the neutral mutants, Q3S, Q4S, Q21T, and Q51T, had no effects upon binding. However, Q3K and Q4K have already been shown to alter Je142 and Je144 binding, respectively (Sharma et al., 1991). These results provided further evidence against the involvement of Gln-71 in the deamidation events.
The possibility that HPr-1, which is generated from deamidation of folded HPr, could have differences in local conformations from HPr that folds from a mutant polypeptide chain was tested by denaturing HPr-1 in 6 M guanidine HCl at 37 "C, dialyzing out the guanidine HCl, and testing the renatured HPr-1 for changes in antibody binding. No changes in antibody binding were detected.
Phosphohydrolysis-HPr-1 has phosphohydrolysis properties that are essentially identical to HPr, while HPr-2 has impaired phosphohydrolysis properties (Waygood et al., 1985). The mutant HPrs, Q21E, Q71E, and N38D, had phosphohydrolysis properties identical to wild-type HPr and HPr-1 (results not shown). Q57E and the double mutant Q57E+Q71E had lower maximal phosphohydrolysis rates ( k = 0.09 versus 0.12 min" for wild-type HPr). Q51E and N12D had phosphohydrolysis rates that were different from either HPr and HPr-1 or HPr-2 (Fig. 4). The principal effect of Q51E was a change in the apparent pK, from about 7.8 for wild type to about 7.4. However, the effect of the aspartic acid at residue 12 was phosphohydrolysis properties that had the most similarity to HPr-2. As the other results above exclude Gln-51 from being involved in the deamidation, the phosphohydrolysis results suggested that deamidation of Asn-12 is involved in the conversion of HPr-1 to HPr-2.

FIG. 2. Deamidation of mutant
Conditions for the isoelectric focusing gels were as described in Fig. 1 (1-24) a Peptides were analysed for all three incubation times used in the tryptic hydrolysis, for HPr, HPr-1, HPr-2 and for selected peptides of all mutant HPrs. * Peptides not found residues 24-28 and 41-45. et al., 1985). The amino acid determinations did not distinguish amide residues from acidic residues. changes were observed for peptide T4, which eluted as a broad peak (Fig. 5), and contained Asn-38-Gly-39. Except for HPr-2, the double mutant Q57E+Q71E, and N12D, similar properties were observed for the elution of peptides from all the HPr proteins. The double mutant Q57E+Q71E resulted in peptide T8, which contains both residues 57 and 7l;eluting approximately 2 min later, thus distinguishing the double mutant from either HPr-1 or HPr-2. HPr-2 had peptides T5 and T7 present immediately after 1.5-h digestion, and no peptide T6 was found (Fig. 5B). When HPr and HPr-2 tryptic digests were mixed, the HPr peptide T6 was found between the HPr-2 peptides T5 and T7 (Fig. 50). HPr-2 peptides T5 and T7 amino acid analysis gave the same composition as shown in Table I. The mutant N12D gave only peptide T7, and neither peptide T5 nor T6 were detected in any of the peptide preparations. The N12D peptides when mixed with HPr-2 peptides resulted in the superimposing of the peptides T7 from both sources (results not shown). The relative abundance of peptide T5 and T7 in HPr-2 was 1:3 which is typical of the isoaspartyl shift reaction. Presumably, peptide T5 contains an isoaspartyl residue. These results identify the site of deamidation responsible for the conversion of HPr-1 to HPr-2 as asparagine 12.

The sequences have been determined by both protein sequencing (Weigel et al., 1982; Powers and Roseman 1984) and DNA sequencing (DeReuse
The conversion of Asn-12 into the two deamidated isomers was much more rapid in the peptides than in HPr. In order to confirm that the conditions (other than the protease) used for proteolysis did not lead to Asn-12 deamidation, HPr was preincubated in 50 mM ammonium bicarbonate, pH 8.0, at 37 "C for 24 h (the tryptic digest conditions), prior to the addition of trypsin. The time-dependent appearance of peptides T5, T6, and T7 was indistinguishable from that observed for HPr peptides derived from freshly prepared lyophilized HPr. The 1.5-h digest yielded an elution profile identical to that seen in Fig. 5A.
The digestion of HPrs with V8 protease should generate different peptides as the result of the introduction of acidic residues in place of the amide residues. The peptides that were separated are shown in Table 11. Again HPr and HPr-1 had identical peptides, and peptide V7 which contained Asn-12 converted to peptides V6 and V8 in a manner essentially identical to that described for the tryptic peptides T5, T6, and T7. Similarly, HPr-2 gave peptides V6 and V8, and the mutant N12D gave only peptide V8. Peptide V9 ends in the sequence -EGEDE, residues 66-70, and there was no indication that this acidic sequence was further digested by the V8protease. Mutant HPrs Q51E and Q57E gave the appropriate peptides, while Q71E resulted in peptide V9 with an amino acid composition that suggested the sequence -EGEDEE at the end of the peptide, residues 66-71. Because this peptide has seven rather than 6 Glx residues, it was not possible to be completely certain about the quantitation. Moreover, the peptides 71-75 or 72-75 eluted in conjunction with a variable amount of the dipeptide, Leu-84-Glu-85 (Table 11), which made quantitation difficult. Mutant HPrs, Q3E and Q21E, did not produce additional sites for V8 protease digestion under the conditions used. These results provided further evidence against the involvement of Gln-51, Gln-57, and Gln-71 in the process of deamidation, but still left the roles of Gln-3, Gln-4, and in particular Gln-21 in question. The role  of Asn-12 was clearly established in HPr-2 formation, and the potential role of Asn-38 was not clarified.
Two-dimensional NMR Spectroscopy-The assignment of chemical shifts for HPr-1 was carried out using DQF-COSY, TOCSY, and NOESY spectra. A pair of methylene protons, with chemical shifts at 2.47 and 3.01 ppm, were observed that had not been found in the HPr spectra. Analysis of the DQF-COSY and TOCSY spectra placed these residues an AMXY spin system. Sequential assignment, using the NOESY spectrum, revealed cross-peaks connecting the amide proton (8.70 ppm) of this spin system with the B protons of Ser-37 and the a-methylene protons of Gly-39 (Fig. 6). The absence of NOES correlating the a proton (4.77 ppm) and the amide proton with the amide proton of Gly-39 are consistent with an isoaspartyl residue (Chazin et al., 1989), which includes the methylene group as part of the backbone. Both of the latter interactions are observed in the NOESY spectrum of HPr (Hammen et al., 1991).
The initial identification of the deamidated residues in HPr-1 (Klevit et al., 1988) was based on chemical shift assignments that have since been revised (Hammen et al., 1991;van Nuland et al., 1992). DQF-COSY spectra were obtained for HPr-1, Q57E, Q71E, Q57E+Q71E, NUD, and N38D. The spectra for NUD, Q57E, Q71E, and the double mutant Q57E+Q71E gave changes in chemical shift for protons in residues near the site of each mutation, but not at all similar to the HPr-1 spectra (results not shown).
Comparison of the DQF-COSY spectra of HPr-1 and wildtype HPr clearly showed that the chemical shift changes are for residues 38, residues near 38 in sequence, and residues near 38 in the tertiary structure (Fig. 7). The DQF-COSY spectra of HPr-1 and N38D were not identical (Fig. 8). Aside from obvious differences in the position of cross-peaks assigned to residue 38, there are significant differences in the position of cross-peaks associated with residues 39-41. Additional differences in the spectra were observed, mainly involving residues 51-54 which are close to residue 38 in the structure. In general, the spectrum of N38D compares more favorably with that of wild-type HPr than with HPr-1.
Low intensity cross-peaks were found in the spectra of HPr-1, corresponding to the positions of the cross-peaks for the aspartate residue created in N38D. To obtain values for the relative amounts of aspartic acid and isoaspartic acid present in HPr-1, the DQF-COSY, and TOCSY cross-peaks connecting the methylene protons of these residues were integrated. The intensities of these cross-peaks should be insensitive to local conformation, and therefore provide an unbiased measure of the mixture proportion. The integration showed that the HPr-1 samples contained about 6:l isoasparticaspartic acid.
PTSActivi~-P~viously it has been found that HPr-1 has no impairment to its ability to act as a substrate for enzyme I or enzyme IIMan, and had a small change in Km for enzyme IIMt'. HPr-2 however was impaired for enzyme I and enzyme IIMt', but not enzyme IIMm (Waygood et al., 1985). All the mutants that were produced were screened for activity using an enzyme IIMa" assay (Sharma et al., 1991), and the results indicated that only N12D and Q51E had impaired activity. Q51E, Q57E, Q71E, NUD, N38D, and the double mutant N12D+N38D have been more thoroughly investigated. The kinetic parameters for enzyme I and various enzymes 11 are given in Table 111. The N38D mutant resulted in normal kinetic responses, as would be expected for a deamidation leading to HPr-1. The N12D mutant did not give the same kinetic response as found for HPr-2 in which about 2535% is in the N12D form. However, the double mutation N12D+N38D was significantly impaired in the enzyme IIMt' which is similar to HPr-2. The impairment of HPr-2 in its interaction with enzyme I appears to be much more dependent upon the formation of a isoaspartyl at residue 12. Previously it had been reported that N12D was impaired in an enzyme IIM"" screening assay (Sharma et al., 1991), which was not found in the more careful kinetic analysis (Table 111). The discrepancy probably resulted from insufficient enzyme I being provided in the screening assay.
While all the above results indicate that Gln-51 is not involved in the formation of either HPr-1 nor HPr-2, its kinetic parameters were described as it is located near the active site. The deamidation of Gln-51 has the most effect of any of the deamidation mutants on PTS activity. All four measured activities were impaired to some degree (Table 111).
Deamidation in HPrs from Other Species-HPr from either B. subtilis or Staphy~occus spp. have Asn-38-Gly-39 sequences (Gonzy-Treboul et al., 1989;Eisermann et al., 1991), and both of these form a major species, similar to HPr-1, upon heating (Fig. 9). HPr from S. f~c a l i s which does not have this sequence (Deutscher et al., 1986) does not form any other species under these conditions (results not shown). It should be noted that the HPr-1 species generated in B. subtilis HPr and S. carnosus gave a doublet band on the isoelectric focusing gel which indicates isoaspartyf formation.

6ISCUSSION
The results in this paper identify the deamidation of residues Asn-38 and Asn-12 as being responsible for the formation of HPr-1 and HPr-2, respectively. The evidence is consistent with the proposal that HPr-2 derives from HPr-1, although no direct evidence for the identification of Asn-38 deamidation in HPr-2 has been obtained. Isolated HPr-1 yields HPr-2 upon heating, and HPr-1 is always accumulated before HPr-2 appearance (Figs. 1-3). The identification of these two asparagine residues involved the elimination of all other amide residues. The identification of the involvement of Asn-12 in the formation of HPr-2 was obtained from several experiments because HPr-2 has significant differences in properties from HPr. The identification of Asn-38 involvement in the formation of HPr-1 came from the elimination of other residues, and the altered NMR spectra. D e a m i~t i o n of proteins is wide-spread, and for the most part appears to affect asparagine residues (Wright, 1991a). The phenomenon is associated with protein aging and denaturation, and possibly is one of the more common spontaneous changes that occurs in protein preparations. There are various mechanism by which it can occur, but the fastest deamidations occur at Asn-Gly sequences by the mechanism shown in Fig. 10 (Bornstein and Balian, 1970;Clarke, 1985;Wright, 1991aWright, , 1991b. This deamidation is well studied in peptides, and the principle products are both D-and L-aspartic acid and D-and L-isoaspartyl acid in a 1:3 proportion (Geiger and Clarke, 1987). Several reports indicate that this proportionality is also found in other small proteins that deamidate at Asn-Gly sequences (Chazin et al., 1989;Spangler and Westbrook, 1989). Deamidation of glutamine residues as a spontaneous event is not as well characterized, but enzymatic deamidation of glutamines and methylation of glutamates are processes in chemotaxis in E. coli (Clarke, 1985;Stock et al., 1989). The PTS is involved in chemotaxis, and it has been proposed that HPr may have a key role in the interaction between the PTS and chemotaxis (Lengeler and Vogel, 1989;Griibl et al., 1990). The rates of asparagine deamidation in HPr observed in vitro suggest that HPr-1 could be present in vivo at about 1% concentration, Le. about 1 ~L M (Mattoo and Waygood, 1983). Assessment of the possible role of deamidated HPr in chemotaxis, although it is an asparagine deamidation, will be conducted using a gene replacement method.
The deamidation of E. coli HPr has been described since the first characterization (Anderson et al., 1971), but has not been described in HPrs from other species. It is now clear that HPrs from B. subtilis and S. carnosus with the conserved  al., 1991) also form an HPr-1 species, presumably by deamidation (Fig. 9). S. faecalis HPr, which lacks this sequence (Deutscher et al., 1986) does not form an HPr-1 species, and none of thege HPrs from Gram-positive bacteria form the HPr-2 species because Asn-12 is not conserved.
In the purification of E. coli HPr and derived mutants, variable amounts of the HPr-1 species are always generated.
The principal cause appears to be the exposure during the purification of HPr to buffers with pH values X . 0 . Clearly maintenance of reduced temperatures is an important factor in reducing the amount and nature of deamidation (Fig. 1). In general, the formation of HPr-2 is not observed unless HPr is treated harshly as was the case for early purification methods (Anderson et al., 1971). HPr-1 species can be relatively easily separated from HPr by the chromatography steps used in HPr purification: Q-Sepharose at alkali pH, or S-Sepharose at acidic pH. The results in this paper suggest that the latter is the safer method. The stability of HPr at acid pH, in respect to deamidation, has been observed in E. coli HPr crystals.
The protein in crystals that take months to grow at pH 3.7 and 15 "C (El-Kabbani et al., 1987) have been examined by isoelectric focusing, and the amount of HPr-1 detected was typical of that found in preparations at the beginning of crystallization?
We had previously reported that the deamidation of both Gln-57 and Gln-71 was responsible for the formation of HPr-1. These identifications were made by comparing NMR COSY spectra (Klevit et al., 1988)   All gave 100% Vmu. The HPrs were treated as described in Fig. 3. The gel used for S. carmsus HPr had pH 3-10 and pH 3.5-5 ampholytes (1:4 ratio) because of its lower PI. The control samples H, were E. coli HPr containing amounts of HPr and HPr-1. HI was isolated HPr-1.

Staphylocococcus carnosus
have since been corrected (Hammen et dl., 1991;van Nuland et al., 1992). The NMR spectral data presented in Figs. 6-8 leave no doubt that it is deamidation of Asn-38 that is responsible for HPr-1 formation. By chance, the double mutation Q57E+Q71E had several properties that were very similar to HPr-1, but it was the differences that led to the investigation of other amide residues. This double mutation had an almost identical PI to that of HPr-1, but other single glutamine to glutamate or asparagine to aspartate mutations yielded species with the same PI as that of HPr-1 (Fig. 3). There were small differences in activity and phosphohydrolysis properties between HPr-1 and the Q57E and Q71E mutants, but not enough to distinguish. However, the Q71E mutation gave a significant change in antibody binding (Sharma et al., 1991), which was different from HPr-1. Moreover, neither Q57E nor Q71E formed a HPr-2 species at the correct PI (Fig. 2).
Within the same small protein molecule, HPr, there are 2 Asn-Gly residues that deamidate to give isoaspartyl residues, but the behavior of these pairs is different. We now know that Asn-12 is in a type I &turn, and that Asn-Gly are residues 3 and 4 of the turn (Jia, 1992). This residue does not deami- ..
date readily in HPr under normal conditions, and it is more susceptible under conditions in which the tertiary structure is destroyed as in the peptide, or under conditions which are leading to denaturation as in the case of elevated temperatures (Fig. 1). Wearne and Creighton (1989) showed a similar behavior for Asn-67 in RNase that exists in a type I11 P-turn with Asn-Gly at positions 3 and 4. The mobility of the glycine residue at position 4 is constrained by the hydrogen bond of the turn, and presumably inhibits the cyclization reaction (Fig. 10). That the deamidation of Asn-12-Gly-13 occurs more rapidly in a less structured polypeptide is consistent with the proposed mechanism of isoaspartyl residue formation (Bornstein and Balian, 1970;Clarke, 1985;Geiger and Clarke, 1987;Wright, 1991a). Asn-38-Gly-39, which more easily yields an isoaspartyl residue, is in a type 111' @-turn structure in E. coli HPr (Jia, 1992). The two-dimensional NMR results suggest a type I' &turn for E. coli HPr and B. subtilis HPr (Hammen et al., 1991;Wittekind et al., 1992); but the distinction is probably not real due to overlapping bounds of the 6 and CP angles used to classify these turns. The Asn-Gly occupy positions 2 and 3 of the turn, and are unencumbered by hydrogen bonds. This Asn-Gly pair in HPrs from various species deamidates under mild conditions in the native protein to yield HPr-1 and appears to more readily deamidate at elevated temperatures (Fig. 1). No indication was found that the peptides containing this pair were more susceptible. The /3-turn with Asn-38-Gly-39 is in a conformation that causes it to be very solvent exposed, and away from the body of the structure in both E. coli and B. subtilis HPrs (Herzberg et al., 1992;Wittekind et al., 1992;Jia, 1992). However, the amide exchange rates with solvent suggest that the turn is more exposed in E. coli than B. subtilis HPr (Wittekind et al., 1992).
Two of the deamidation mutants, N12D and Q51E, and HPr-2 have alterations in their ability to carry out phosphoryltransfer. The structural information that is available for HPrs from different species shows that the overall folding patterns are similar, but the active site descriptions have significant differences. A number of reports indicate the involvement of residues 12 and 51, but with differences in the details. The E. coli HPr structures were the first reported (Klevit and Waygood, 1986;El-Kabbani et al., 1987), and both have now been modified (Hammen et al., 1991;van Nuland et al., 1992;Jia, 1992). A major difference between the two early E. coli HPr tertiary structures has been resolved with a new 2A resolution structure (Jia, 1992). It is, however, not clear what roles residues 12 and 51 have in the active site of E. coli HPr. In the HPrs from S. faecalis and B. subtilis, residue 12 is threonine and serine, respectively, and residue 51 is a methionine, and both residues are implicated in interactions with the active site His-15 (Herzberg et al., 1991;Jia, 1992;Jia et al., 1993). Unfortunately, the structural information on the 3. subtilis HPr has inadvertently been carried out on a M51V mutant (Wittekind et al., 1992).
In E. coli HPr, we have now investigated by site-directed mutagenesis, all the residues that so far have been implicated in hydrogen bonding to His-15 in any HPr structure. The acidic replacements of Asn-12 and Gin-51 and the deletion of Glu-85 (Anderson et al., 1991) have all resulted in changed phosphohy~olysis properties, but only modest changes to the kinetic parameters investigated. The lack of extent of these kinetic changes is particularly evident when these results are compared to the effects of alterations to Arg-17 which produce 100-1000-fold reductions in PTS activity (Anderson et al., 1993). Some potential insensitivity in the kinetic measurements have been discussed in respect to the Arg-17 mutants (Anderson et at., 1993), and those comments apply to the assessment of these deamidation mutants in respect to active site mechanism roles.