A Low-Barrier Hydrogen Bond in the Catalytic Triad of Serine Proteases? Theory Versus Experiment

See allHide authors and affiliations

Science  07 Nov 1997:
Vol. 278, Issue 5340, pp. 1128-1132
DOI: 10.1126/science.278.5340.1128


Cleland and Kreevoy recently advanced the idea that a special type of hydrogen bond (H-bond), termed a low-barrier hydrogen bond (LBHB), may account for the “missing” transition state stabilization underlying the catalytic power of many enzymes, and Frey et al. have proposed that the H-bond between aspartic acid 102 and histidine 57 in the catalytic triad of serine proteases is an example of a catalytically important LBHB. Experimental facts are here considered regarding the aspartic acid–histidine andcis–urocanic H-bonds that are inconsistent with fundamental tenets of the LBHB hypothesis. The inconsistencies between theory and experiment in these paradigm systems cast doubt on the existence of LBHBs, as currently defined, within enzyme active sites.

The H-bond inherently involves the sharing of hydrogen atoms to varying extents with other atoms (1). This sharing is often depicted as a chemical equilibration or resonance hybridization of structures such as 1 and 2 (Eq. 1). Proton sharing can also be depicted as a lengthening of the A-H bond of the donor, 1, as if the proton were in an intermediate stage of transfer to B (2). In conventional H-bonds the H atom is associated more with one heteroatom than the other.

LBHBs are distinguished from conventional H-bonds by equal proton sharing between the heteroatoms. LBHBs can be double-welled, depicted as 1 and 2 contributing equally to the system; or single-welled, depicted as a single structure with the proton residing at a point equidistant between A and B, as in3 (Eq. 1).

LBHBs were first observed in the gas phase (1,3), where evidence for their existence and strength is persuasive. Schowen first proposed that LBHBs may exist within the protected interior of proteins (4). Promotion of this idea by Cleland and Kreevoy (5) and Frey et al.(6) has led to its substantial (7), though not universal (8), acceptance. Physicochemical parameters used to identify LBHBs include (i) extreme low-field 1H nuclear magnetic resonance (NMR) chemical shifts [δ > 16 parts per million (ppm)]; (ii) deuterium isotope effects on low-field 1H resonances; (iii) low (<1.0) isotopic fractionation factors; and (iv) deuterium isotope effects on infrared and Raman frequencies (1)—the most unambiguous of which is claimed to be (i) (6). The proposal for an LBHB between Asp102 and His57 is based on criteria (i) and (ii) in studies of several serine proteases, and on the perception thatcis–urocanic acid mimics the Asp-His H-bond only in nonaqueous solvents.

The LBHB hypothesis states that LBHB formation requires the absence of H-bonding solvents, notably water, and requires matched pK a values (where K a is the acid constant) for the heteroatoms involved (6). Here we show experimental facts regarding the Asp-His andcis–urocanic acid systems that are inconsistent with these and other key tenets of the LBHB hypothesis as currently framed, namely: (i) the H-bonded (Nδ 1-H) proton is not equally shared between the carboxylate of Asp10 2 and Nδ1 of His57, but is localized on Nδ 1; (ii) the Asp102- His57 H-bond is not sequestered from water, but is surrounded by freely diffusable water—even in the presence of tightly bound inhibitors; (iii)cis–urocanic acid mimics the Asp102- His57 H-bond not only in organic solvents, but also in water; (iv) the pK a values of the carboxylic acid and imidazolium groups of cis–urocanic acid are not matched under conditions where cis–urocanic acid exhibits putative LBHB behavior, but differ by ∼4 pK aunits; and (v) the bond energies of the Asp-His andcis–urocanic acid H-bonds are not >10 kcal/mol, but are ∼5 kcal/mol.

A key element of the proposal of Frey et al. (6) is that the Asp-His H-bond is low-barrier—that is, equally shared between Asp102 and His57—only when the imidazole ring becomes protonated at Nɛ 2 (+ charged, 4) (Fig.1A). Were this not the case, the LBHB could be of no advantage to catalysis for reasons Frey et al. have outlined. The Nδ 1-H protons in α-lytic protease, chymotrypsin, and other serine proteases resonate at 17 to 18 ppm for protonated His57 and at 13.8 to 15 ppm for neutral His57 and thus conform to this requirement. In15N-labeled α-lytic protease, the Nδ 1–H-bond exhibits spin-coupling constants (1 J NH) of 80 ± 4 Hz and 90 ± 1 Hz for the protonated and neutral His57, respectively (9). Imidazole N-H spin-coupling constants are typically 90 to 98 Hz. Because 1 J NH spin-coupling is a through-bond phenomenon, it is a direct measure of covalent bond character. These 1 J NH values show that the proton is essentially 100% localized on Nδ 1 in neutral His57 and at least 85% on Nδ 1 when His57 becomes protonated and engaged in the putative LBHB. N-H spin-coupling is also observed in enzyme-inhibitor complexes, such as that between α-lytic protease and the potent transition-state analog inhibitor MeOSuc-Ala-Ala-Pro-boroVal (inhibition constantK i = 6.7 × 10 9). His57 becomes protonated in such complexes, as expected for transition-state or tetrahedral intermediate complexes formed with real substrates (10). The15Nδ 1-H proton resonance found at 16.1 ppm in the complex (Fig. 1B) exhibits1 J NH = 92 ± 1 Hz—again showing that the proton remains localized on Nδ 1. Halkides and co-workers report a1 J NH value of 81 Hz for aBacillus lentus subtilisin-Z-Leu-Leu-Phe-trifluoromethyl ketone (Z-LLF-CF3) complex, indicating ∼86% proton localization on Nδ 1 of the active site histidine (11).

Figure 1

(A) 15N and Nδ1-H proton chemical shifts, spin-coupling constants, and pK a value for low- and high-pH forms of resting α-lytic protease. (B) His57N-H protons of an α-lytic protease transition-state analog inhibitor complex exchanging with water. “Hard 1-1” 1H NMR spectra (500 MHz) of uniformly 15N-labeled α-lytic protease (25) complexed with MeOSuc-Ala-Ala-Pro-boroVal in aqueous solution (pH 4.5) at various temperatures. N-H proton lifetimes, τ, were determined from lineshapes by the slow-exchange approximation, τ = 1/(πΔW), where ΔW is excess linewidth attributable to exchange (baseline value of 35 Hz for both peaks obtained from the 5°C spectrum). This leads to lifetimes of Nδ1-H and Nɛ2-H protons, respectively: >42 and >13 ms at 5°C; 42 and 4.5 ms at 25°C; 16 and 1.7 ms at 35°C; and 1.9 and <1.7 ms at 55°C. (C) Active site histidine N-H protons of a subtilisin transition-state analog inhibitor complex exchanging more rapidly with water than N-H protons of four other histidine residues. 1H NMR spectra (500 MHz) of 15N histidine–labeled subtilisin BPN′ (26) complexed with Ac-Ala-Ala-Pro-boroPhe in aqueous solution (pH 5.5) at 25°C. The jump-return spectrum (top) showing1H resonances of five of the six subtilisin BPN′ histidines (18). The doublet at 16.9 ppm (1 J NH = 92 Hz) and the very broad peak at ∼15.4 ppm are assigned to the active site histidine (His64) Nδ1 and Nɛ2 protons, respectively. Among the resonances from 10.0 to 12.2 ppm are four additional histidine N-H protons, known to be pH-independent. In additional spectra, systematic reduction in data size down to 2048 data points, recycle delay down to 0.020 s, and repetition times down to 45 ms led to no appreciable loss of signal or noise of the histidine N-H protons (16). The equivalent of a “progressive saturation” T 1 determination, this places an upper limit on the lifetimes of all histidine N-H protons at <∼15 ms. The 15N-edited spin-echo difference, jump-return spectrum (SED + jump-return, middle) with 15N decoupling is simplified, revealing the five 1H resonances of subtilisin BPN′ histidine residues. The SED spectrum with water pre-saturation (SED + pre-sat, bottom) shows that reduction of the active site histidine Nδ1-H proton resonance at 16.9 ppm is substantially greater than that of the upfield four histidine N-H protons.

What magnitude of1 J NH should be expected for Nδ 1-H of His57 if the Asp-His H-bond were low-barrier? The hydrogen bifluoride ion (FHF), widely regarded as the best example of a solution phase, single-well system, exhibits 1 J F H ≈ 120 Hz in dipolar, aprotic solvents (δ = 16.4 ppm). This coupling is about one-fourth of that observed for HF (476 Hz) (12). By analogy with FHF, we might expect an N-H coupling of ∼24 Hz for Nδ1-H of His57 if the H-bond were single-welled. For a double-well system,1 J NH = 47 Hz would be expected from averaging the couplings (0 Hz and 94 Hz) for the two species associated with the potential minima. The observed Nδ1-H spin couplings in serine proteases indicate that the Nδ 1-H proton can be no more than 15 ± 8% delocalized when His57 is protonated, whether in the resting enzyme or in transition state–like enzyme-inhibitor complexes.

15N chemical shifts have been shown to be remarkably informative as to the location of imidazole N-H protons (13). In α-lytic protease, Nδ1 and Nɛ2 resonate at 199.4 and 138.0 ppm, respectively, when His57 is neutral, and at 191.6 and 204.0 ppm, respectively, when His57 is protonated (Fig. 1A). Pyrrole-like (>N-H), pyridine-like (>N:), and protonated, pyrrole-like (+>N-H) 15N atoms typically resonate at 211, 128, and 202 ± 2 ppm, respectively, in aqueous solution (14). H-bonding perturbs these shifts by up to 10 ppm—downfield for >N-H atoms acting as proton donors to carboxylate groups, and upfield for >N: acceptors (13). For α-lytic protease at pH > 9, His5715N shifts have been shown to reflect the exclusive presence of the Nδ1-H tautomer, 5, with both ring N atoms engaged in H-bonding; the H-bond to Asp102 moves Nδ1 downfield ∼10 ppm, whereas the H-bond from Ser195 moves Nɛ2 upfield ∼10 ppm from 15N shift values expected for the pure Nδ1-H tautomer. At pH < 5, where the imidazole ring of His57 is protonated, 4, the H-bond to Asp102 is reflected in the ∼10 ppm downfield displacement of the Nδ1resonance from ∼202 to 191.6 ppm. Because H-bonding induces chemical shift changes in the same direction, but lower in magnitude, as those induced by protonation or deprotonation, H-bonding effects can be interpreted in terms of partial proton transfer. Thus, the ∼10 ppm displacement of Nδ1, relative to that for full deprotonation, represents ∼14 ± 4% delocalization of the Nδ1-H proton (15), a value in good agreement with that determined from 1JNH.

The LBHB hypothesis requires the serine protease catalytic triad to be sequestered from solvent. Solvent accessibility of N-H groups is often gauged by measuring the rate at which deuterium replaces protons upon dissolution of the protein in D2O. The Nδ1-H proton, however, becomes fully deuterated before an NMR spectrum can be recorded in α-lytic protease and other serine proteases, placing the exchange half-life at <∼1.0 min (16). Longitudinal nuclear relaxation time (T 1) measurements on resting α-lytic protease indicate an Nδ1-H proton exchange lifetime of <9 ms (16). Transverse nuclear relaxation time (T 2) measurements (lineshape analysis) on chymotrypsinogen A indicate an Nδ1-H proton exchange lifetime in the millisecond range at 1° to 19°C (17). Such rapid exchange demonstrates the His57 Nδ1-H proton is highly accessible to solvent in resting serine proteases.

However, it could be argued that occupancy of the active site by substrates or inhibitors might block solvent access sufficiently for LBHB formation. MeOSuc-Ala-Ala-Pro-boroVal, a potent transition-state analog inhibitor of α-lytic protease, occupies subsites S1 to S4, as would a specific substrate. In addition to the Asp-His H-bond (δ = 16.1 ppm), a second putative LBHB (δ = 16.5 ppm) is formed between Nɛ 2-H and the boronyl group in this complex (10). The inhibitor does reduce solvent access, as demonstrated by the narrower linewidth of the Nδ1-H proton signal relative to that of resting enzyme (Fig. 1B). Nevertheless, lineshape analysis indicates that the Nδ1-H and Nɛ 2-H proton lifetimes are still on the order of milliseconds. In contrast, the lifetime of the bound inhibitor, with a K i value of 6.7 × 10 9, is on the order of minutes. Thus, even in this tightly bound inhibitor complex, both imidazole N-H protons of His57 are readily accessible to solvent.

The subtilisin BPN′–Ac-Ala-Ala-Pro-boroPhe (K i = 1 × 10 10) complex exhibits features similar to those of the α-lytic protease–MeOSuc-Ala-Ala-Pro-boroVal complex, except that exchange of the Nɛ 2-H proton signal with water is more rapid, because its resonance is barely observable at 25°C (Fig. 1C, top) (18). Five of subtilisin's six histidines, the catalytic His and four others, give rise to low-field 1H signals, an observation aided by15N spectral editing and 15N decoupling (Fig.1C, middle). Upon selective irradiation (presaturation) of water (Fig.1C, bottom), the active site histidine N-H proton resonance essentially disappears, whereas the other four remain visible. Thus, of the five observable histidine residues, the catalytic histidine is the most solvent accessible—even in a complex with a tight-binding transition-state analog inhibitor.

The perception that cis–urocanic acid, 6 (Fig.2), forms an LBHB similar to that of the Asp-His diad only in organic solvents is based on reports that a low-field proton signal is observed in these solvents (17.5 ppm in dimethylsulfoxide, 17.2 ppm in acetonitrile, and 16.9 ppm in acetone), but not in water (6). However, the ability ofcis–urocanic acid to mimic the Asp-His H-bond is not confined to organic solvents (Fig. 2). The 15N chemical shift behavior of cis–urocanic acid in 100% water at room temperature (19) closely resembles that of His57in α-lytic protease (20) over the entire range of pH stability for the enzyme (∼4 to 10). Particularly from pH 4 to 5, the15N chemical shifts of cis–urocanic acid show the same asymmetry exhibited by His57 of α-lytic protease and attributed to the H-bond with Asp102—namely, that Nδ1 resonates ∼10 ppm downfield from Nɛ 2 and from chemical shifts characteristic of +>N-H–type nitrogens.

Figure 2

15N chemical shifts versus pH ofcis–urocanic acid, 6, are similar to those of His57 of α-lytic protease, both in aqueous solution.15N shifts for cis–urocanic acid (solid line) are given over the pH range 2 to 10, and those of α-lytic protease (dotted line) over the pH range 4 to 10. [Reprinted with permission of Roberts et al. (19). Copyright American Chemical Society (1982)]

The above results, showing that aqueous cis–urocanic acid is as strongly H-bonded as the Asp-His diad, predict that a low-field Nδ1-H proton resonance should be present and suggest that the inability to detect it is due to rapid exchange with solvent. Upon cooling an 85% acetone-d 6, 15% H2O solution of cis–urocanic acid at pH 4.4 to −48°C, a low-field signal indeed appears, which sharpens upon cooling to −58°C (Fig. 3A). Furthermore, this signal moves from 15.0 ppm at pH 10.0 to 18.5 ppm at pH 4.6 (Fig. 3B), similar to the Asp-His proton in serine proteases (Fig.4A).

Figure 3

Low temperatures required forcis–urocanic acid to exhibit a low-field Nδ1-H proton resonance in water-containing solution. Low-field 400-MHz 1H NMR spectra are shown for ∼0.2 Mcis–urocanic acid versus pH and temperature in various acetone-d 6, water cryosolvent mixtures.cis-Urocanic acid was produced by photoisomerization of thetrans isomer as described (27). pH was measured by immersion of a combination glass electrode and calomel reference (Cole-Parmer, Vernon Hills, IL), standardized with National Bureau of Standards buffers at 25°C. (A) 1H NMR spectra of cis–urocanic acid in an 85% acetone-d 6, 15% H2O cryosolvent (pH 4.4) shows the temperature-dependent linewidth of the Nδ1-H proton resonance at 18.5 ppm. Upon sample cooling to −48°C, the Nδ1-H proton is observable in the aqueous-based solution. With further sample cooling (−58°C) the low-field resonance linewidth decreases, but its chemical shift remains at 18.5 ppm. (B) pH dependence of the low-field resonance chemical shift in cis–urocanic acid in 90% acetone-d 6, 10% H2O. The Nδ1-H proton signal moves from 15.0 ppm at pH 10.0 to 18.5 ppm at pH 4.6, and back to 15.2 ppm at pH 2.4. 1H spectra were recorded at various temperatures from −43° to −60°C and showed no temperature dependence of chemical shift when referenced to sodium 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) as an internal standard.

Figure 4

1H and13C chemical shifts of cis–urocanic acid versus pH, yielding stepwise pKa values of a diprotic acid. (A) 1H NMR chemical shifts for the Nδ1-H and ring protons versus pH in 90% acetone-d6, 10% H2O recorded at low temperatures ranging from −43° to −60°C at 400 MHz. Nonlinear least-squares regression analysis (14) yielded values of pK1 = 2.9 and pK2 = 7.2 for all three protons. The solid lines represent various cis–urocanic acid protons, and the dotted line (top) represents the Nδ1-H proton of chymotrypsin (28). (B) 13C NMR chemical shifts versus pH for the carboxylate and ring carbons ofcis–urocanic acid (•) and the carboxylate carbon oftrans–urocanic acid (○) in 90% acetone-d6, 10% H2O, recorded at 25°C and 100.6 MHz. Nonlinear least-squares analysis of these curves also yielded values of pK1 = 2.9 and pK2 = 7.2. (C) Schematic diagram ofcis–urocanic acid at low, middle, and high pH values.

The LBHB hypothesis requires the carboxylic acid and imidazolium groups of the Asp-His diad or of cis–urocanic acid to have equal proton affinities or pK a values. There is now ample evidence that catalytic histidines in serine proteases have pK a values of ∼7.0 (20,21). Direct pK a determination for Asp102, however, has proven elusive. In the absence of this measurement, it could be argued that the pK avalue assigned solely to His57 does not reflect the microscopic pK a of the imidazole group, but rather that of the Asp-His diad unit, within which Asp102and His57 would have equal pK avalues. As a model of the Asp-His diad, cis–urocanic acid affords an opportunity to address the question of the pK a value of the carboxylate. That a transition between imidazole and imidazolium ion occurs with a pK a of ∼7 is strongly supported by15N results (Fig. 2) because of the large, characteristic chemical-shift differences (Δδ) (13, 20). A separate ionization of the carboxylate with a pK a of ∼2.9 can be seen in these curves as the secondary inflection in which Nδ 1 moves from a position characteristic of an H-bonded (191 ppm) to a non–H bonded (∼201 ppm) +>N-H–type nitrogen as the pH is lowered.

Titration inflections of the imidazole ring protons can be followed under conditions where the Nδ 1-H proton is visible (Fig. 4A). The Δδ's of the Cɛ 1-H and Cδ2-H proton resonances associated with pK a = 7.2 are of the same magnitude and direction as that observed for histidine in aqueous solution at 25°C and is consistent with ionization of the imidazole ring (22). These resonances then move further downfield at the pH 2.9 transition but remain in the range characteristic of protonated imidazole rings. The Δδ of the Nδ 1-H proton of cis–urocanic acid closely parallels that of chymotrypsin (dotted line) over the pH 7.2 transition. The second ionization at pH 2.9 moves the Nδ 1-H proton of cis–urocanic acid 3.3 ppm upfield, effectively reversing the initial downfield movement at pK a = 7.2 that moved this resonance to a position qualifying it as an LBHB by the existing framework.

The CO, Cγ, Cɛ1, and Cδ 2 13C resonances ofcis–urocanic acid in the acetone-water cosolvent at 25°C respond to the same two pK a values (7.2 and 2.9) observed in the 1H and 15N spectra (Fig. 4B). The behavior of the 13CO signal was of special interest, because it might have been expected to provide direct information on the protonation state of the carboxylate group. This 13C resonance undergoes a relatively large chemical shift displacement at pK a = 7.2 and a much smaller one at pK a = 2.9. The direction and magnitude of the displacement (3.5 ppm, moving from 176.5 to 173.0 ppm as the group with pK a = 7.2 is protonated) is consistent with displacements observed previously for protonation of carboxylate groups (23). Although this could be construed as support for the LBHB hypothesis, Fig. 4B shows that the 13CO signal oftrans–urocanic acid, which cannot form an intramolecular H-bond, exhibits the same displacement at the imidazole pK a. The best explanation for the13CO results is transmission of the imidazole protonation effect through the extended conjugated π electron system to the carboxylate group.

All of the NMR results on cis–urocanic acid are consistent with the scheme in Fig. 4C. The extreme downfield chemical shift of the1H resonance in 8 must be attributed to a strengthening of the H-bond as a result of the coulombic interactions between oppositely charged donor (imidazolium ion) and acceptor (carboxylate) groups—an effect that disappears in 7(pH < 2.9) and in 9 (pH > 7.2), where either functional group is neutral. This scheme suggests that a second low-field signal should appear at pH < 2.9, owing to the presence of the protonated carboxylic acid. The spectrum ofcis–urocanic acid at pH 1.8 and −70°C in 100% acetone–d 6 does indeed reveal a second low-field signal at 16.0 ppm (16).

The differences in pK 1 and pK 2 between cis– (2.9 and 7.2) andtrans– (3.7 and 6.2) urocanic acids [ΔpK a of −0.8 and +1.0, respectively, in the acetone-water cosolvent; −0.7 and +0.9 in aqueous media (19)] can be attributed directly to the stabilization effect of the intramolecular H-bond (Fig. 4B). The spreading of pK 1 and pK 2 is the result of the H-bond stabilizing the conjugate acid (His+) in one equilibrium and the conjugate base (COO) in the other. The average ΔpK a (0.85 pH units in absolute value) is similar to the ΔpK a of +0.8 units between His57 of α-lytic protease and free histidine (20). This corresponds to about 1.2 kcal/mol of stabilization energy at 25°C. Together with the tautomeric equilibrium constant K T [>25:1 (14)] for the stabilization of the Nδ 1-H tautomer in cis–urocanic acid and in His57 (19), the overall energy of the H-bond is estimated to be ∼5 kcal/mol, which is in good agreement with that estimated from Asp102 mutagenesis experments (24), but far less than the 10 to 20 kcal/mol of the LBHB hypothesis.


View Abstract

Navigate This Article