Low Barrier Hydrogen Bond Is Absent in the Catalytic Triads in the Ground State but Is Present in a Transition-state Complex in the Prolyl Oligopeptidase Family of Serine Proteases*

High frequency proton NMR spectra for two members of the prolyl oligopeptidase class of serine proteases, prolyl oligopeptidase and oligopeptidase B, showed that resonances corresponding to the active center histidine Nδ1H and Nε2H generally observed in this region, are absent in these enzymes. However, for both enzymes, as well as with the H652A and H652Q active center variants of oligopeptidase B, there are two resonances observed in this region that could be assigned to two protonated histidines with a noncatalytic function. The results indicate that these two histidines participate in strong hydrogen bonds. The absence of resonances pertinent to the active center histidine resonances suggests the absence of a low barrier hydrogen bond between the Asp and His in these two enzymes in their ground states. Addition of the peptide boronic acidt-butoxycarbonyl-(d)Val-Leu-(l)boroArg to oligopeptidase B resulted in potent, slow binding inhibition of the enzyme and the appearance of a new resonance at 15.8 ppm, whose chemical shift is appropriate for a tetrahedral boronate complex and a low barrier hydrogen bond. The results demonstrate important dissimilarities between the active centers of the prolyl oligopeptidase class of serine proteases and the pancreatic and subtilisin classes both in the ground state and in the transition-state analog complexes.

The hydrogen bond between the Asp and His residues in the catalytic triad of serine proteases has been presented as an example of the potential contribution of low barrier hydrogen bonds (LBHB) 1 to catalysis of enzymatic reactions (1)(2)(3). It was hypothesized by some that such an LBHB could lead to transition-state stabilization for enzymatic reactions by as much as 20 kcal/mol in favorable cases (4,5). Others have argued that one need not invoke the existence of an LBHB to understand enzymatic catalysis; in fact, such a strong hydrogen bond may be "anticatalytic" (6). In enzymes it has been difficult to obtain direct evidence pro or con of this LBHB hypothesis. Although there are many x-ray structures of serine peptidases in the literature, the level of precision of the analysis does not permit one to conclude either that there is such an LBHB at the active center or that such bonds are formed in transition-state-like structures. In solution, NMR and Fourier transform-IR offer possible methods for detection of such unusually strong hydrogen bonds (2,7). One of the experimental observations used to support the existence of such an LBHB (1) is the presence of a high frequency proton magnetic resonance (13-20 ppm) for several serine proteases (8 -18) that is attributed to the proton in a hydrogen bond between the Asp 102 ␤ carboxylate and the imidazole N ␦1 H of His 57 (chymotrypsin numbering) of the active center Asp..His..Ser triad. The resonance is present in the spectra of serine proteases of both the trypsin/chymotrypsin and of the subtilisin family, both in the absence and in the presence of a variety of inhibitors of several classes (8 -18). In the uncomplexed enzymes, the chemical shift of this broad resonance is pH-dependent and unusual (14 -15 ppm at pH values above the His 57 pK a , and 17.5-18.3 ppm at pH values below the pK a (see Refs. 13-16 for evidence on chymotrypsin, trypsin, and subtilisin)). These chemical shifts are significantly larger (more deshielded) than those of the corresponding resonance in model systems, indeed suggesting the presence of unusual hydrogen bonds (7) in both ionization states of the triad even in the ground states of these enzymes.
Herein are reported proton magnetic resonance experiments in the high frequency chemical shift range carried out on aqueous solutions of two members of the recently identified prolyl oligopeptidase family of serine proteases (19 -21): prolyl oligopeptidase (PO, formerly known as prolyl endopeptidase or post-proline cleaving enzyme) and oligopeptidase B (OpB, formerly known as protease II). PO and OpB are large proteins (80 kDa) with homologous amino acid sequences displaying the peptidase domain at the carboxyl terminus. The order of the catalytic serine and histidine in the sequence is reversed with respect to that of trypsin and subtilisin, indicating an unrelated three-dimensional structure (19,22). PO is a ubiquitous enzyme that cleaves the peptide bond at the carboxyl end of proline residues (20). OpB is found in Escherichia coli and is specific for Arg and Lys residues (21) as also found with trypsin. In contrast to the chymotrypsin/trypsin and subtilisin classes of serine proteases, PO and OpB do not give evidence for a strong hydrogen bond between the Asp and His residues at the active center in the absence of transition-state analogs, but such a resonance is evident in the complex of OpB with a peptide boronic acid. Most importantly, the spectra reveal the presence of strongly hydrogen-bonded non-catalytic histidines in both enzymes.

EXPERIMENTAL PROCEDURES
Proton Nuclear Magnetic Resonance-Proton nuclear magnetic resonance experiments were carried out in Wilmad 535-PP 5-mm tubes at 400 and 500 MHz on VARIAN VXR 400 -89 or VARIAN Inova 500 spectrometers operated at the temperatures indicated in the figure legends. Concentration of PO was 0.2-0.4 mM, and DSS was used as internal chemical shift reference. PO was stored with 3% sucrose in the lyophilized state. Prior to the NMR experiments, the lyophilized enzyme containing 20 -30 mg of PO was dissolved at 4°C in 1-2 ml of 20 mM potassium phosphate, pH 7.0, containing 1 mM EDTA, 1.0 mM dithiothreitol, and 0.01% (w/v) NaN 3 and chromatographed on a Sephadex G-25 (Sigma) column equilibrated with the same buffer. The eluate was concentrated to 0.3-0.7 ml in an Amicon Centricon 30 concentrator at 4°C. The lyophilized OpB (20 -50 mg) was dissolved in 1-2 ml of 20 mM phosphate buffer, pH 8.0, and chromatographed on a Sephadex G-25 column (50 ml bed volume) equilibrated with the same buffer. The eluate was concentrated as described for PO.
Prolyl Oligopeptidase-PO was isolated from pig muscle as described (20). The activity of PO was determined fluorimetrically with benzyloxycarbonyl-Gly-Pro-2-naphthylamide as substrate (20) by using a Jasco FP777 spectrofluorimeter. The excitation and emission wavelengths were, respectively, 340 and 410 nm.
Oligopeptidase B-OpB was expressed in E. coli JM83 and purified with slight modification as described by Kanatani et al. (23). The activity of the enzyme was measured spectrophotometrically with N-benzo-yl-L-Arg-p-nitroanilide-HCl (Sigma) as substrate at 410 nm (⑀ ϭ 8900 M Ϫ1 cm Ϫ1 , see Ref. 24). The substrate and the enzyme were both dissolved in 0.1 M Tris buffer, pH 8.0, containing 1 mM EDTA.
Preparation of Mutant Oligopeptidase B-The pSKOpB vector contained the wild-type oligopeptidase gene. To change the active site His 652 to Gln or Ala, the primers H652Q (GGACTCACCGCCAAGGC-CGGGCAAATCTGG) containing the BglI restriction enzyme site and H652A (GGAACTTCAAGGCGCCGGCGGCAAATCTGG) containing the EheI site were synthesized. The original NcoI site was eliminated in both primers. The desired mutations were introduced with one of the sense primers His652Gln or His652Ala and the antisense PROT3Ј primer (CGGTCGACGAACCGCGATCCGGGC), with a SalI restriction enzyme site underlined. The PCR-I product was synthesized in a mixture containing the pSKOpB vector as template, 100 nmol each of the two primers, 400 M dNTP, 0.02 units/ml of Pfu DNA polymerase (Stratagene) in 1 ϫ Pfu reaction buffer (Stratagene) with 30 cycles at 94, 55, and 72°C for 1, 2, and 3 min, respectively. The PCR product (200 base pairs) was identified on an agarose gel (1.7%), extracted, and purified on Sephaglas (Pharmacia Biotech Inc.).
The second round of PCR used the same program as in the previous paragraph, using the pSKOpB vector as template with the 200-base pair megaprimer and the PROT5Ј primer (CGAATTCATCCCCGGT-GAGTTTTGCCACC) containing the underlined EcoRI site (providing the second PCR product). Subcloning was carried out by double digestion of this second PCR product with EcoRI and SalI restriction enzymes and then ligation into the EcoRI/SalI sites of a pBluescript SK(ϩ) vector (Stratagene). The resulting pSKOpBH652Q and The mutation in plasmids pSKOpBH652Q and pSKOpBH652A was verified by digestion with BglI and EheI restriction enzymes, respectively. The mutations were also confirmed by DNA sequence analysis of the mutated plasmids prepared in E. coli DH5␣ cells, using a sequencing oligo primer 5Ј-GGATCCGCAATATTAACGAGTACATGAAA-3Ј and the PRISM TM Ready Reaction dideoxy terminator kit from Perkin-Elmer and an Applied Biosystems model 373 DNA sequencer. The sequences obtained showed that the only alterations in the mutated genes were those desired.
The mutated oligopeptidase B was expressed and purified as the wild-type enzyme in E. coli XL1-Blue, except that the enzymes were produced at 30°C rather than at 37°C.

Inhibition Studies of OpB by t-butoxycarbonyl-(D)Val-Leu-(L)boro-Arg-The
Boc-(D)Val-Leu-(L)boroArg-pinanediol-benzenesufonate 2 was synthesized according to published methods (25). All kinetic assays were carried out at 25°C. First, the K m was determined for N-benzoyl-Arg-p-nitroanilide-HCl as 0.24 Ϯ 0.06 mM, compared with 0.25 mM reported by Kanatani et al. (23), along with a k cat of 24.5 Ϯ 0.07 s Ϫ1 (the number quoted in Ref. 23 is low by a factor of 1000, perhaps the result of an incorrect dilution factor). The Boc-(D)Val-Leu-(L)boroArg-pinanediol-benzenesufonate (at 0.75 mg/ml) was incubated in 0.1 M Tris buffer, pH 8.0, containing 1 mM EDTA, to hydrolyze the pinanediol ester of the boronic acid. A solution of OpB (4.4 nM) was then incubated with the inhibitor solution for 1-2 h at 25°C, and then the activity was assayed for 3 min. Preincubation of the inhibitor with OpB produced a lag phase for p-nitroaniline release before the steady-state kinetic phase was reached. A Lineweaver-Burk plot of the data collected during the steady-state phase indicated the presence of competitive inhibition and a K i of 3.1 nM (Fig. 1A). Care must be exercised in the kinetic studies since OpB is subject to inhibition at substrate concentrations in excess of 0.5 mM. 3 When the inhibitor and substrate were premixed and the reaction was initiated by the addition of OpB, the progress curves for p-nitroaniline release showed that the magnitude of the absorbance plateau varied inversely with inhibitor concentration (Fig. 1B). These experiments strongly imply that the inhibitor is of the "slow-binding" type (25). appear to be pH-independent to pH 8.88 (not shown). Fig. 3 shows spectra at 10°C for OpB with pH-independent resonances at 16.5 and 17.8 ppm between pH 6.50 and 9.46 that are no longer detectable at pH 9.75. There are additional resonances observed at 13.5 and 14.3 ppm, the latter weaker than the former but quite evident in several spectra, especially at 2°C (not shown). Fig. 4 shows spectra of OpB (middle), with excess antipain (top, an aldehyde-type inhibitor as is benzyloxycarbonyl-Pro-Prolinal for PO), and for the H652A active center variant of OpB (bottom), all at pH 8.1-8.2. The much sharper resonances observed at 14.2 and 13.3 ppm in the presence of the peptide aldehyde inhibitor are not surprising since the inhibitor would tend to make the structure more rigid, perhaps also protecting the histidines in question from exchange with solvent, the principal mechanism responsible for the line width of such resonances. The chemical shifts of the resonances are superimposable within experimental error, as are those observed for the H652Q variant (not shown). One-dimensional difference nuclear Overhauser enhancement spectroscopy (NOESY) experiments (Fig. 5) were carried out on OpB at pH 7.94, showing that the resonance at 17.9 ppm (bottom) correlates with resonances at 13.6, 11 (broad), and 7.8 ppm while the 16.5 ppm resonance (top) correlates with resonances at 11 (broad), 9.6, and 8.4 ppm. It should be noted that the resonances are already broad due to the size of the enzyme, and the high frequency ones are further broadened by exchange (NHs attached to an aromatic imidazolium ring are being reported). Therefore, integration, especially in a difference spectrum, is of virtually no value, and the number of protons represented by each resonance is difficult to ascertain.  H652A (and H652Q) variant of OpB strongly suggests that this variant is correctly refolded as the presence of these resonances is clear evidence for the formation of the proper tertiary structure. At the same time, the variant enzyme is completely inactive (activity Ͻ0.01%), which confirms that the active site His has indeed been altered. (f) Nuclear Overhauser enhancement experiments provide strong evidence that the two resonances observed in OpB pertain to different histidines. We suggest that the 17.9-and 16.6-ppm resonances represent the N ␦1 H of two different histidines, whereas the NOEs imply that the resonance at 13.5 may correspond to the N ⑀2 H of the histidine with N ␦1 H at 17.9 ppm. This resonance at 13.5 ppm is visible in several spectra, as is a different one at 14.2 ppm (see Figs. 3 and 4, and especially Fig.   4, top spectrum, in the presence of antipain that apparently slows down the exchange rate of these two protons). The resonance at 14.2 ppm likely corresponds to an N ⑀2 H residing on the histidine with an N ␦1 H at 16.6 ppm. The resonances at lower frequency between 9.6 and 7.8 ppm could pertain to the C2 and C4 hydrogens of the imidazole rings and probably to some backbone NHs. These assignments are in accord with a recent study by Markley and Westler (12) on chymotrypsinogen and affirm that the two histidines are in their histidinium ionization state. (g) pH titration of the solution of OpB shows the pK a of the two histidines to be greater than 9.5, and those of PO to be greater than 8.9. For comparison, the pK a for His 57 in uncomplexed chymotrypsin is 7, and the chemical shift of the active center His 57 varies with pH (8 -10, 14). (h) The peptide boronic acid Boc-(D)Val-Leu-(L)boroArg turned out to be a very potent competitive slow binding inhibitor (K i ϭ 3 nM) of OpB, in whose presence the spectrum of OpB exhibits a single new resonance at 15.8 ppm. In the chymotrypsin/trypsin and subtilisin class, the serine-bound boronates exhibit two resonances pertinent to N ␦1 H and N ⑀2 H (14). The observation of only a single resonance with a chemical shift of 15.8 ppm pertinent to the complex is not only appropriate for a nitrogen-bound proton at histidine in complexes of serine proteases with peptide bo-  Fig. 2 was used. Backward linear prediction was used to improve the signal to noise, and especially the base line. ronic acids but the behavior is also consistent with the boron being bound to a His at N ⑀2 (in this case His 652 ), or to Ser and His concurrently, rather than to a serine at the active center (14,28) and the resonance being pertinent to N ␦1 H.

RESULTS AND DISCUSSION
Steady-state kinetic measurements of k cat /K m on both wildtype PO and OpB provide evidence for a pH optimum near 8.0, i.e. the pK a of the active site His is below this value, whereas that of the observed histidines in both PO and OpB is higher than 8.8. These results, in addition to the evidence showing the same resonances in the wild-type and H652A variants of OpB, lead to the conclusion that we are observing noncatalytic histidines rather than the one at the active center. While both hydrogens on the histidine ring nitrogens are found in strong hydrogen bonds, such bonds to N ␦1 H are evidently much stronger than those to N ⑀2 H. The elevated basicity of the observed histidines compared with that in chymotrypsin, trypsin, and subtilisin is also consistent with such strong hydrogen bonds and/or additional nearby negative charges.
The chemical shifts of protons at N ␦1 and N ⑀2 vary considerably among various complexes. The N ␦1 H is at 18 ppm for protonated His 57 and is at 15 ppm for neutral His 57 in uncomplexed chymotrypsin (8 -10, 14). It is between 16 and 17.2 ppm in the peptide boronic acid complexes of ␣-lytic protease (28), chymotrypsin (14), and subtilisin, 4 and it is at 18.7 ppm in the complex of chymotrypsin with peptidyltrifluoromethylketone (29,30). In both types of transition-state analog complexes, the His 57 is protonated, hence the LBHB is present in both ground state and transition-state complexes. Yet, these resonances are not even seen between pH 5.14 and 9.46 in spectra of OpB and PO, ruling out an LBHB at any pH in these enzymes at their active centers.
It is concluded that in PO and OpB there are very likely two non-catalytic histidines present, which are in a strongly hydrogen bonded environment, enabling us to observe resonances for both N ␦1 H and N ⑀2 H. According to a sequence alignment (23), there are six sets of conserved histidines in these enzymes at positions (first number is for PO and the one in parentheses is for OpB) 20 (15), 213 (204), 307 (289), 456 (436), 640 (616), and 680 (652). The last one (680 in PO and 652 in OpB) has been identified as the one at the active center. Further research will be required to determine which two of the five remaining conserved histidines are being observed in these experiments and are deduced to participate in strong hydrogen bonds.
The inability to observe the active center His resonances in spectra of OpB and PO strongly suggests that the Asp-His hydrogen bond is much weaker in these enzymes than in the chymotrypsin/subtilisin classes of serine peptidases. The absence of such a strong hydrogen bond between the active center Asp and His in OpB and PO suggests that: (a) there is no LBHB at this position in these two enzymes in the uncomplexed form or in the tetrahedral complexes formed with the aldehydebased so called "transition-state" analogs; (b) there is an LBHB in the tetrahedral complex formed with the peptide boronic acid, which installs a negative charge on the N ⑀2 side as well. One may also conclude (as we did in Ref. 13) that near neutral pH values, the negative charge at the N ⑀2 side creates a better transition-state analog in terms of electrostatic interactions (AspCOO Ϫ HisH ϩ SerX Ϫ ) than the aldehyde, which appears to form a neutral tetrahedral complex (AspCOO Ϫ HisH o SerX; compare results on OpB with antipain and the peptide boronic acid and the results on trypsin in the presence of leupeptin in Ref. 16). In the case of OpB, the boronate complex with the tetrahedral anionic character evidently induces a stronger hydrogen bond between the Asp and N ␦1 H at the active center. These results are consistent with previous reports on transition-state analogs that install a negative charge on the N ⑀2 side, such as complexes of peptide boronates (13)(14)(15)(16)28), peptidyltrifluoromethylketones (29 -30), as well as the monoisopropylphosphoryl Ser 195 derivatives of chymotrypsin, trypsin, and subtilisin (31), all of which uniformly raise the pK a (increase the basicity) of the active center histidine substantially. Similarly elevated pK a of the active center His in going to the transition state would be expected with the developing oxyanion, thereby reducing the ⌬pK a between the histidine and the serine, and with possible concomitant strengthening of the hydrogen bond between the Asp and N ␦1 H.
The results also confirm that such exceptionally strong hydrogen bonds, reflected by such unusually large chemical shifts, will be revealed in the structures of many enzymes. Additional examples are provided by some coenzyme-dependent enzymes: the pyridoxal phosphate-dependent enzymes that gave rise to several resonances with chemical shifts between 13-20 ppm, one of them assigned to the pyridinium N1-proton (32)(33)(34)(35); and the thiamin diphosphate-dependent pyruvate decarboxylase that exhibits a chemical shift of 17.2 ppm, a resonance as yet unassigned. 5 In summary, there are distinct differences found between the active centers of this new prolyl oligopeptidase class of serine peptidases and the well studied chymotrypsin/trypsin and subtilisin classes. (a) There is no LBHB in the ground state, but there are two non-catalytic histidines that appear to participate in LBHBs in the prolyl oligopeptidase class. (b) One can design potent, slow-binding peptide boronic acid-type inhibitors for both the prolyl oligopeptidase, and the chymotrypsin/ trypsin and subtilisin classes. But, while the former appears to form a boronate complex to His N ⑀2 (or simultaneously to His and Ser, a possibility that is difficult to differentiate by NMR methods) at the active center, the chymotrypsin/trypsin and subtilisin classes form a serine-bound boronate. It is concluded, therefore, that there are subtle novel features observed in both the ground state and in the transition-state type complexes in this prolyl oligopeptidase class of serine peptidases that have not been previously seen in any other serine proteases. 4 D. Bao and F. Jordan, unpublished data. 5 D. Zhang, F. Guo, and F. Jordan, unpublished observations. SCHEME 1. High frequency histidine proton chemical shifts in serine protease active centers.