The pK _a values of the catalytic residues in the retaining glycoside hydrolase T26H mutant of T4 lysozyme

Abstract

T4 phage lysozyme (T4L) is an enzyme that cleaves bacterial cell wall peptidoglycan. Remarkably, the single substitution of the active site Thr26 to a His (T26H) converts T4L from an inverting to a retaining glycoside hydrolase with transglycosylase activity. It has been proposed that T26H–T4L follows a double displacement mechanism with His26 serving as a nucleophile to form a covalent glycosyl‐enzyme intermediate (Kuroki et al., PNAS 1999; 96:8949–8954). To gain further insights into this or alternative mechanisms, we used NMR spectroscopy to measure the acid dissociation constants (pK _a values) and/or define the ionization states of the Asp, Glu, His, and Arg residues in the T4L mutant. Most notably, the pK _a value of the putative nucleophile His26 is 6.8 ± 0.1, whereas that of the general acid Glu11 is 4.7 ± 0.1. If the proposed mechanism holds true, then T26H–T4L follows a reverse protonation pathway in which only a minor population of the free enzyme is in its catalytically competent ionization state with His26 deprotonated and Glu11 protonated. Our studies also confirm that all arginines in T26H–T4L, including the active site Arg145, are positively charged under neutral pH conditions.

Brief statement

The replacement of a single amino acid changes T4 lysozyme from an inverting to a retaining glycoside hydrolase. Using NMR spectroscopy, we measured the pK _a values of the ionizable residues in the active site of this mutant enzyme. Along with previously reported data, these results provide important constraints for understanding the catalytic mechanisms by which the wild‐type and mutant form of T4 lysozyme cleave bacterial peptidoglycan.

Abbreviations

DAP

diaminopimelic acid

GH

glycoside hydrolase

NAG

N‐acetylglucosamine

NAM

N‐acetylmuramic acid

pH*

the pH meter reading uncorrected for isotope effects

pK_a

negative logarithm of the acid dissociation constant

rmsd

root‐mean‐square deviation

T4L*

cysteine‐free (C54T and C97A) variant of T4 lysozyme

Introduction

Lysozymes are glycoside hydrolases that cleave the β‐1,4‐glycosidic linkage between N‐acetylmuramic acid (NAM) and N‐acetylglucosamine (NAG) of peptidoglycan, a major component of the bacterial cell wall. These enzymes are classified into at least five families by the Carbohydrate Active Enzyme (CAZy) database based on their amino acid sequences.1, 2 Members of a family typically utilize a similar hydrolytic mechanism that leads to either retention or inversion of anomeric carbon stereochemistry at the site of hydrolysis.3, 4

The lysozyme from Enterobacteria phage T4 (T4L) is a member of the GH24 family. Although T4L is an extremely well‐characterized model system for investigating protein folding and dynamics,5 detailed enzymology studies have proven difficult as it acts on a complex glycopeptide substrate6 for which analogs amenable to rigorous kinetic measurements are not readily available. In 1995, Matthews and co‐workers reported that T4L is an inverting glycoside hydrolase.7 This conclusion was based in part on the observation that hydrolysis of a peptidic tetrasaccharide by T4L yielded a disaccharide product with the opposite anomeric stereochemistry to that generated by hen egg white lysozyme, a classic GH22 retaining hydrolase.8 Accordingly, it was proposed that T4L catalyzes a single‐displacement inverting reaction in which a nucleophilic water, activated by the general base Asp20 (with a reported9, 10 pK _a value of 3.6), attacks the β‐1,4‐linked anomeric C1 carbon of the peptidic NAM, while the general acid Glu11 (pK _a 5.4) concomitantly donates a proton to the leaving aglycone, NAG.

In the X‐ray crystallographic structures of wild‐type T4L, a candidate nucleophilic water is hydrogen bonded between the sidechains of Asp20 and Thr26. As part of an earlier attempt to engineer metal binding sites into T4L, the Matthews group introduced a Glu at position 26. Serendipitously, the T26E mutant was isolated from E. coli with Glu26 covalently bound to a peptidic NAG‐NAM disaccharide via an α‐linkage to the anomeric carbon of a distorted NAM moiety.11 This stable species appeared to be analogous to the glycosyl‐enzyme intermediate formed in the first step of a double‐displacement retaining mechanism. Subsequently, it was found that the T26H mutant is catalytically active, yet produces a hydrolysis product with the opposite anomeric stereochemistry to that yielded by wild‐type T4L.7 Remarkably, a single amino acid substitution at position 26 changes T4L from an inverting to a retaining glycoside hydrolase. Furthermore, unlike the wild‐type enzyme, T26H–T4L also functions as a transglycosylase.12

Based on mutational and X‐ray crystallographic studies, a two‐step double‐displacement mechanism was proposed for T26H–T4L [Fig. 1(A)].7, 12 In the glycosylation step, His26 is postulated to be a nucleophile, forming an α‐linked glycosyl‐enzyme intermediate. This is facilitated by Glu11 serving as a general acid to protonate the leaving aglycone. In the subsequent deglycosylation step, Glu11 functions as a general base to activate either a water (hydrolysis) or carbohydrate (transglycosylation) for nucleophilic displacement of His26 and release of the glycone with a net retention of β‐anomeric stereochemistry.

Figure 1.

(A) Proposed double‐displacement retaining mechanism for the hydrolysis reaction catalyzed by T26H–T4L.7, 12 (R₁ and R₂ correspond to NAG; R₃ to the peptide‐linked lactyl substituent of NAM.) A minor population of the enzyme is depicted with the hypothesized nucleophile His26 and the general acid/base Glu11 in their catalytically competent protonation states. (B) An alternative mechanism involving substrate‐assisted catalysis. Asp20 is shown as a general base, but could also electrostatically stabilize a charged oxazolinium intermediate without transfer of the N‐acetyl nitrogen‐bonded proton.

An important aspect of understanding the proposed mechanism, or any alternative, for T26H–T4L lies with defining the protonation states of catalytic residues along the reaction pathway. To this end, Kuroki et al.13 used room temperature neutron diffraction to characterize a sample of perdeuterated T26H–T4L*, crystallized in D₂O buffer (pD 7.0). In the reported 2.09 Å resolution structure (5XPF.pdb), His26 is predominantly in its neutral state and the carboxyl of Glu11 partially deuterated. This is consistent with the scheme of Figure 1(A). However, Asp20 and nearby Arg145 were also modeled to be neutral. This is perplexing since there are no obvious physicochemical reasons why these active site residues should have such highly perturbed pK _a values and unusual sidechain deuteration states at pD ~ 7. Indeed, in the 2.2 Å resolution neutron crystallographic structure of the perdeuterated wild‐type enzyme at cryogenic temperatures (pD 6 – 7, 5VNQ.pdb), all ionizable residues are in their charged forms.14 Inspired by these reports, we used NMR spectroscopy to investigate the protonation states and, when possible, determine the pK _a values of the His, Asp, Glu, and Arg residues in T26H–T4L. These measurements provide key insights into the catalytic mechanism and electrostatic properties of this mutant retaining glycoside hydrolase.

Results

T26H–T4L* yielded high‐quality NMR spectra, and thus assignment of the ¹H, ¹³C, and ¹⁵N signals from its mainchain and selected sidechain nuclei was straightforward using standard heteronuclear correlation experiments (Fig. S1). A comparison of the ¹⁵N‐HSQC spectrum of this mutant with that of pseudo‐wild‐type T4L* revealed moderate ¹H^N and ¹⁵N chemical shift differences for many corresponding amides (Fig. S2). When mapped onto the structure of lysozyme, these amides cluster in the N‐terminal lobe around position 26. The crystallographic co‐ordinates of T4L* (1CX6.pdb) and T26H–T4L* (1QT8.pdb) superimpose closely (0.28 Å rmsd for all atoms), and thus these chemical shift differences likely arise from small conformational changes due to the amino acid substitution, combined with electric field and ring current effects from the charged, aromatic imidazolium sidechain of histidine. These two co‐ordinate files were used to provide the inter‐residue distances cited below.

Histidine pK _a and tautomer state determination

The signals from the two histidines in uniformly ¹³C/¹⁵N‐labeled T26H–T4L* were monitored via ¹³C‐HSQC spectra as the protein was titrated between pH 5 and 11 [Figs. 2(A) and S3]. Fitting the pH‐dependent chemical shifts of the ring ¹³C^ε1, ¹H^ε1, ¹³C^δ2, and ¹H^δ2 reporter nuclei yielded average pK _a values of 6.8 ± 0.1 for His26 and 8.9 ± 0.1 for His31 [Fig. 2(B) and Table 1]. Furthermore, with highly diagnostic ¹³C^δ2 chemical shifts ~ 127 ppm, the neutral forms of both histidines can be confidently assigned to the less common N^δ1H tautomer.17, 18

Figure 2.

(A) Superimposed ¹⁵N‐decoupled ¹³C‐HSQC spectra of uniformly ¹³C/¹⁵N‐labeled T26H–T4L* showing the ¹H^ε1–¹³C^ε1 signals of His26 and His31 as the protein was titrated between pH 5 and 11 at 25°C (100 mM KCl, 30 mM potassium phosphate, 5% D₂O). Spectra for the ¹H^δ2–¹³C^δ2 signals are provided in Figure S3. (B) Fit titration curves for the ¹³C^ε1, ¹H^ε1, ¹³C^δ2, and ¹H^δ2 nuclei yielded average pK _a values of 6.8 ± 0.1 for His26 and 8.9 ± 0.1 for His31 (Table 1). The biphasic titration curve for the ¹³C^ε1 of His31 has a second fit pK _a ~ 6.3 for a small chemical shift change of −0.2 ppm with increasing sample pH that may be attributed to conformational or electrostatic perturbations16 upon deprotonation of nearby (~ 8 Å) His26. The structures of the imidazolium and neutral imidazole N^δ1H tautomer sidechains are shown.

Table 1.

pK_a values of T26H–T4L* and wild‐type T4L

Residue	T26H–T4L*a	T4Lb
His26	6.8 ± 0.1
His31	8.9 ± 0.1	9.1
Asp10	< 0.8c	< 0.5
Asp20	~ 1.5d	3.6
Asp47	3.1 ± 0.1	3.0
Asp61	3.6 ± 0.1	3.6
Asp70	< 0.8c	< 0.5
Asp72	3.3 ± 0.1	3.5
Asp89	4.0 ± 0.1	4.0
Asp92	2.7 ± 0.1	2.5
Asp127	3.5 ± 0.1	3.5
Asp159	3.4 ± 0.1	3.5
Glu5	3.7 ± 0.1
Glu11	4.7 ± 0.1	5.4
Glu22	4.3 ± 0.1
Glu45	3.8 ± 0.1
Glu62	2.9 ± 0.1	2.8
Glu64	4.4 ± 0.1
Glu108	4.2 ± 0.1
Glu128	4.4 ± 0.1
Arg95	> 12e
Arg145	> 11e
Arg148	> 11e

^aMeasured herein at 25°C with T26H–T4L* samples initially in 100 mM KCl and 30 mM potassium phosphate.

^bMeasured from the pH‐dependent His ¹H^ε1 and Asp/Glu ¹³C^γ/δ chemical shifts of amide deuterated or selectively sidechain labeled wild‐type T4L, respectively, at 10°C and initially in 100 mM KCl and 10 mM phosphate, as reported by Anderson et al.10, 15 The assignments were obtained via single‐site amino acid substitutions.

^cEstimated upper limit (pK _a < pH − 1) as no ¹³C^γ chemical shift changes due to protonation were seen at the titration endpoint of pH 1.8 [Figs. 3(A,C)].

^dFit with a restrained chemical shift change of 3.4 ppm upon deprotonation [Fig. 3(C)].

^eEstimated lower limit (pK _a > pH + 1) as no ¹⁵N^ε chemical shift changes due to deprotonation were seen up to pH 9.9 for Arg145 and Arg148, and pH 10.7 for Arg95 (Fig. 5).

These results are consistent with the previously reported pK _a value of 9.1 for His31 in wild‐type T4L,15 and the X‐ray crystallographic structure of the protein with this histidine donating a hydrogen bond from its protonated ring N^δ1 to the sidechain carboxylate of Asp70 (pK _a < 0.8, see below). Similarly, the structure of T26H–T4L* shows that His26 donates a hydrogen bond from its protonated N^δ1 to the buried mainchain carbonyl oxygen of Tyr24, leaving its N^ε2 solvent exposed to aid catalysis. Presumably, both hydrogen bonds are maintained when His26 and His31 are N^ε2‐deprotonated under alkaline conditions.

The pK _a value of an ionizable functional group in a folded protein reflects the pH‐dependent free energy difference between states with that group in its neutral versus charged forms, and hence a complex interplay of entropic and enthalpic terms, including desolvation, hydrogen bonding, and electrostatic interactions with surrounding charged and polar moieties. Dissecting the relative contributions of these terms is challenging experimentally and computationally.19 Nevertheless, it is well accepted that the highly elevated pK _a value of His31, compared to that of ~ 6.5 for a histidine in a random coil polypeptide,18 arises in large part from its formation of a very favorable salt‐bridge (i.e. a hydrogen‐bonded ion pair) with ionized Asp70.15 This conclusion is supported by a comparison of the pH‐dependent stability against thermal denaturation of the wild‐type protein versus variants with H31N and/or D70N substitutions.15 In contrast, the relatively unperturbed pK _a value of His26 may reflect a balance of unfavorable desolvation of the partially buried imidazolium sidechain and favorable electrostatic interactions with nearby Glu11 (~ 6 Å) and Asp20 (~ 3.5 Å). Based on their pK _a values presented below, these latter residues are both predominantly negatively charged around the midpoint pH 6.8 for the deprotonation of positively charged His26.

Aspartic acid pK_a determination

A sample of T26H–T4L* selectively labeled with ¹³C^γ‐aspartic acid was titrated between pH 1.8 and 7.8. The pH‐dependent chemical shifts of the ¹³C^γ nuclei were monitored via 1D ¹³C‐NMR spectra [Fig. 3(A)]. The Asp ¹³C^γ and Glu ¹³C^δ are part of the carboxylic acid functional group and thus generally reliable reporter nuclei for pK _a measurements by NMR spectroscopy.18 Of the 10 Asp residues, seven gave well‐defined titration curves with downfield chemical shift changes of ~ 3.4 ppm upon increasing sample pH [Fig. 3(C)]. Such spectral changes are diagnostic of sidechain deprotonation. Fitting the titration curves yielded the pK _a values summarized in Table 1. These values match well those reported previously and discussed in detail for the wild‐type protein.10

Figure 3.

Stacked ¹³C‐NMR spectra of T26H–T4L* selectively labeled with (A) ¹³C^γ‐Asp and (B) ¹³C^δ‐Glu, recorded as a function of sample pH at 25°C (initially 100 mM KCl, 30 mM potassium phosphate, 5% D₂O). Due to metabolic interconversion, glutamine (Q) sidechains were also labeled in the latter sample. The titration curves were fit to obtain the (C) Asp and (D) Glu pK _a values reported in Table 1. The biphasic titration curve for Asp20 followed one pK _a ~ 1.5 with a restrained ¹³C^γ chemical shift change of 3.4 ppm upon increasing sample pH that is attributed to its own ionization and a second small upfield chemical shift change of −0.5 ppm with pK _a ~ 6.6 that likely reflects structural or electrostatic perturbations due to the deprotonation of nearby (~ 3.5 Å) His26. Similarly, the titration curve for Glu11 followed a predominant ¹³C^δ chemical shift change of 2.7 ppm with pK _a 4.7 assigned to its own ionization and a smaller downfield change of 0.5 ppm with pK _a ~ 7 that also likely reflects the titration of His26 (~ 6 Å).

In contrast, the ¹³C^γ chemical shifts of Asp10 and Asp70 did not change markedly over the pH range studied, and that of Asp20 appeared to follow a titration with pK _a < 2. However, in the absence of a complete titration curve, it is difficult to infer the ionization state of an Asp residue based on chemical shifts alone. For example, Asp10 has a ¹³C^γ chemical shift indicative of a protonated carboxylic acid, whereas that of Asp70 corresponds to a deprotonated carboxylate. An alternative approach for distinguishing a carboxylic acid from a carboxylate lies with detecting the presence or absence, respectively, of a 2‐bond isotope shift of ~ 0.23 ppm for the ¹³C^γ due to a proton versus deuteron on the adjacent oxygen.20, 21, 22 As shown in Figure 4(A), none of the 10 Asp residues exhibited a ¹³C^γ chemical shift change of this magnitude when T26H–T4L* was transferred from H₂O at pH 5.8 to D₂O at pH* 5.7. This confirms that all are ionized under these conditions. Furthermore, since no NMR spectroscopic evidence for their protonation was seen even at pH 1.8 in ¹³C‐NMR monitored titrations, Asp10 and Asp70 are strong acids with pK _a values < 0.8 (i.e., from the Henderson–Hasselbalch equation for a simple acid dissociation equilibrium, at pH = pK _a + 1, the population ratio of conjugate base to acid is 0.1, which should be detectable). The same conclusions were previously reported for wild‐type T4L (Table 1), as evidenced by the lack of any NMR‐detectable titrations for Asp10 and Asp70 down to pH 1.5, combined with a careful study of the pH‐dependent stability of the protein under acidic conditions.10, 15 These highly reduced pK _a values, compared to that of ~ 4 for an aspartic acid in a random coil polypeptide,18 can be rationalized at least in part by the salt‐bridges between Asp10–Arg148 and Asp70–His31 observed by X‐ray crystallography.

Figure 4.

Selected regions of the 2D H^β/γ ₂(C^β/γ)C^γ/δ spectra of uniformly ¹⁵N/¹³C‐labeled T26H–T4L* in H₂O (blue; pH 5.8) and D₂O (red; pH* 5.7) NMR sample buffer. The spectra show correlations between (A) the H^ββ’ and ¹³C^γ of Asp and (B) the H^γγ’ and ¹³C^δ of Glu. The vertical insets are the 1D ¹³C‐NMR spectra of protein selectively labeled with (A) ¹³C^γ‐Asp or (B) ¹³C^δ‐Glu at pH 5.9 (taken from Fig. 3). The scale bar indicates the expected 0.23 ppm isotope shift for a neutral carboxylic acid (¹³COOH vs. ¹³COOD). The lack of any significant isotope shifts (< 0.04 ppm in magnitude) demonstrates that the Asp and Glu residues in T26H–T4L* are predominantly ionized under these conditions. Two exceptions are the small ¹³C^δ shift of 0.08 ppm exhibited by Glu11 (pK _a 4.7) that is attributed to a change in its fractional ionization due to the slightly different pH/pH* conditions, and the slight “reversed” isotope shift of Asp70 ¹³C^γ (−0.07 ppm) that may possibly result from protonation versus deuteration of its hydrogen bonded partner His31.22, 23 Also seen in (A) are ¹⁵N‐coupled signals from several Gln sidechains, which show expected20, 24 deuterium isotope shifts (¹³CONH₂ vs. ¹³COND₂) of ~ 0.12 ppm. Signals from the Asn sidechains, with similar isotope shifts, are outside of the presented ¹³C chemical shift windows.

Most important for this study, we conclude that Asp20 has a pK _a ~ 1.5 in T26H–T4L* (obtained by fitting the partial titration curve with a restrained chemical shift change of 3.4 ppm due to deprotonation). This is substantially lower than the pK _a of 3.6 reported for this residue in the wild‐type protein10 and may at least in part be due to the presence of a positively charged histidine versus neutral threonine at the nearby (~ 3.5 Å) position 26. Parenthetically, in wild‐type T4L, Asp20 shows an anomalously small ¹³C^γ chemical shift change over the pH range examined.10 Although this limits the confidence that can be placed on assigning the fit pK _a value of 3.6 to the deprotonation of this catalytic residue, no other titrations were detected for the ¹³C^γ of Asp20 between pH 1.5 and 5.5.

Glutamic acid pK_a determination

A sample of T26H–T4L* selectively labeled with δ‐¹³C glutamic acid was also titrated between pH 1.8 and 9.3 [Fig. 3(B,D)]. In ¹³C‐NMR spectra, all eight Glu residues showed clear titration curves with downfield ¹³C^γ chemical shift changes of ~ 3.8 ppm diagnostic of deprotonation with increasing sample pH [Fig. 3(B,D)]. Sidechain ionization was confirmed by the absence of a significant deuterium isotope shift for each residue in H₂O at pH 5.8 versus D₂O at pH* 5.7 [Fig. 4(B)]. Fitting the titration curves yielded the pK _a values summarized in Table 1. These results also generally match those reported previously for wild‐type T4L,10 with Glu62 having the lowest pK _a value of 2.9, likely due in part to its close proximity to Arg52 (3.2 Å).

Glu11 has a pK _a of 4.7 in T26H–T4L*. This is only marginally higher than the pK _a of ~ 4.4 for an unperturbed glutamic acid in a random coil polypeptide,18 yet lower than that of 5.4 measured for the wild‐type protein.10 Favorable electrostatic interactions with nearby (~ 6 Å) His26 in the T26H mutant may cause this pK _a reduction. However, as noted by Andersen et al.,10 the elevated pK _a value of Glu11 in wild‐type T4L is somewhat unexpected given its hydrogen bonding with Arg145, albeit offset by charge repulsion from Asp10 (~ 7 Å) and Asp20 (~ 7 Å). These authors suggested that conformational dynamics of T4L in solution, including a hinge bending motion between its N‐ and C‐lobes, may alter the solvent accessibility and electrostatic environment of active site residues relative to those seen in static crystal structures, thus precluding a simple rationalization of pK _a values.

All arginines are N ^ε protonated at neutral pH

We also sought to determine the protonation states of the arginines in T26H–T4L*. Sidechain ¹H^ε–¹⁵N^ε signals from the 13 arginines in both the wild‐type and mutant protein were readily detected and assigned in ¹⁵N‐HSQC spectra recorded at pH 5.9 (Fig. 5). This unequivocally demonstrates that under these conditions, all arginines, including Arg145, in both proteins are protonated at the N^ε position. However, it is possible that an arginine could be neutral due to the loss of a terminal H^η proton. Therefore, we also assigned the arginine ¹³C^ζ and ¹⁵N^η signals of T26H–T4L* at pH 6.5 (Fig. S4). Although caution must be exercised in inferring charge states from chemical shifts, all 13 had ¹⁵N^ε, ¹³C^ζ, and ¹⁵N^η signals highly indicative of a fully protonated guanidinium sidechain (e.g., 84.8 ppm, 159.5 ppm, and 75.0 and ~ 71 ppm, respectively, for Arg145).18, 25 Furthermore, the ¹³C and ¹⁵N NMR signals of corresponding arginines in the T26H–T4L* mutant and in the wild‐type protein are very similar.26, 27, 28 Of these, several, including Arg145, in T4L have been unambiguously demonstrated to be fully protonated at pH 5.5, and hence positively charged, on the basis of ¹H^η–¹⁵N^η scalar coupling patterns.29

Figure 5.

¹⁵N‐HSQC spectra showing the arginine guanidinium ¹H^ε‐¹⁵N^ε signals from uniformly ¹⁵N/¹³C‐labeled (A) wild‐type T4L* and (B) T26H–T4L* at pH 5.9. The data unambiguously demonstrate that all 13 arginines, including Arg145, are N^ε‐protonated in both proteins. Superimposition of these spectra (Fig. S2) reveals small ¹H^ε–¹⁵N^ε chemical shift differences between several corresponding arginines of the wild‐type and mutant protein attributable to the T26H substitution and variations in exact sample conditions. (C–E) With increasing pH, the ¹H^ε–¹⁵N^ε signals from T26H–T4L* disappeared due to base‐catalyzed hydrogen exchange. Only well‐protected arginines were detected under alkaline conditions. The lack of any significant pH‐dependent ¹⁵N^ε chemical shift changes shows that the detected Arg95, Arg145, and Arg148 are positively charged under these conditions and hence have high pK _a values (Table 1). The small downfield shift in the ¹H^ε signal of Arg148 followed an apparent pK _a of 7.2. There are no ionizable groups proximal to Arg148 with a pK _a value in this range and thus we speculate that this chemical shift change reflects altered pH‐dependent interactions with buffer phosphate (pK _a2 ~ 7).

We attempted to measure the pK _a values of the arginines in T26H–T4L* using 2D H^δ ₂(C^δ)N^ε‐type experiments.30 However, this was not successful as spectral crowding precluded the confident measurement of pH‐dependent ¹⁵N^ε chemical shifts. In principle, these reporter shifts, as well as those of the ¹³C^ζ nuclei, could be obtained from a N^ε/η–C^ζ correlation experiment.29 This was not pursued due to the low sensitivity of such a ¹³C‐detected approach and signal overlap in the resulting 2D spectra (Fig. S4). Regardless, the unperturbed pK _a value of an arginine in a random coil polypeptide is ~ 13.8,18, 25 and it is unlikely that lysozyme would remain folded, or even chemically intact, under the harsh alkaline conditions required to measure complete titration curves.

Alternatively, we simply recorded ¹⁵N‐HSQC spectra of T26H–T4L* as a function of sample pH value [Fig. 5(C–E)]. At pH 9.9, the ¹H^ε–¹⁵N^ε signals of all but three highly protected arginines (Arg95, Arg145, and Arg148) disappeared due to rapid base‐catalyzed hydrogen exchange with water.31 At pH 10.7, only Arg95 was observed. This protection, which matches the trends in arginine exchange rate constants recently measured for wild‐type T4L*,32 can be attributed at least in part to hydrogen bonds between Arg95 with the mainchain oxygen of Phe153, Arg145 with the sidechain oxygen of Asn101, and Arg148 with the carboxylate of Asp10. Importantly, no significant changes in the ¹⁵N^ε chemical shifts of these three residues were observed over the pH range examined. Since the ¹⁵N^ε signal of an arginine moves downfield by ~ 6 ppm upon deprotonation,18, 25 this implies that the pK _a values of Arg145 and Arg148 must be > 11, and that of Arg95 > 12 (Table 1).

Discussion

Using NMR spectroscopy, we measured the pK _a values of the Asp, Glu, and His residues in T26H–T4L* and confirmed that all arginines are N^ε protonated and most certainly charged under neutral pH conditions. Along with previously reported data for T4L, these results provide important constraints for understanding the catalytic mechanisms of the wild‐type and mutant lysozymes.

Inverting mechanism of wild‐type T4L

Wild‐type T4L is an inverting glycoside hydrolase. A reasonable single‐displacement mechanism11 involves Asp20 acting as a general base to activate a nucleophilic water that is also hydrogen bonded to Thr26. This water carries out a nucleophilic substitution, with inversion of stereochemistry, at the anomeric carbon of a peptidic NAM. Concomitantly, Glu11 serves as a proton donor to facilitate the departure of the NAG aglycone.

This mechanism leads to the prediction that T4L should display a bell‐shaped activity versus pH profile, with a pH optimum of ~ 4.5 where Asp20 (pK _a 3.6) is predominantly deprotonated to act as a general base and Glu11 (pK _a 5.4) predominantly protonated to function as a general acid. In this context (and below), “predicted activity” refers to k_cat/K_m, the second order rate constant for the reaction of free enzyme and substrate, and neglects likely contributions of pH‐dependent interactions between the charged substrate peptide moiety, lAla‐dGlu‐DAP‐lAla, with an extended binding site on the protein surface.11 The first‐order rate constant for turnover of the Michaelis complex, k _cat, will depend upon the pK _a values of catalytic residues with the bound substrate, which have yet to be measured. Unfortunately, these fundamental catalytic parameters have not been determined for T4L due to the lack of a well‐defined substrate amenable to detailed kinetic analyses. Although it has been reported that the lytic activity of T4L on chloroform‐treated Escherichia coli cells is maximum at pH ~ 7.3, with sensitivity to ionic strength, amines, and divalent cations,33, 34 numerous factors complicate the mechanistic interpretation of these results.

Retaining mechanism of T26H–T4L with His26 as a nucleophile

Remarkably, the single substitution of Thr26 with a histidine converts T4L from an inverting to a retaining glycoside hydrolase with transglycosylase activity. Based on the classical Koshland double‐displacement mechanism, a plausible pathway [Fig. 1(A)] was proposed for T26H–T4L that involves nucleophilic attack of His26 on the peptidic NAM moiety to form an α‐linked glycosyl‐enzyme intermediate.7, 11, 12 This glycosylation step is facilitated though protonation of the leaving aglycone by the general acid Glu11. In the subsequent deglycosylation step, this same residue serves as a general base to activate either a nucleophilic water (hydrolysis) or carbohydrate (transglycosylation) for His26 displacement and glycone release with overall retention of β‐anomeric stereochemistry. The alternating role of Glu11 as a general acid and base likely results from changes in its environment, and hence pK _a values, along the reaction pathway.35 In contrast to the wild‐type enzyme, Asp20 in T26H–T4L would not play any obvious direct catalytic role except perhaps to facilitate substrate binding by hydrogen bonding to the nitrogen of the N‐acetyl group of NAM. Evidence supporting this mechanism includes the observation of a stable peptidic NAG‐NAM moiety α‐linked to Glu26 in T26E–T4L,11 as well as the structural and enzymatic characterization of an extensive series of lysozymes with mutations at residues 11, 20, and 26.7, 12 In particular, the N^ε2 of His26 is approximately positioned for nucleophilic attack on a docked substrate, and the T26Q mutant is inactive.

Accepting this proposed mechanism leads to the prediction that the pH‐dependent activity (k _cat/K _m) of T26H–T4L will follow a bell‐shaped profile with a maximum at pH ~ 5.8. However, the pK _a value of the general acid Glu11 (4.7) is lower than that of the postulated nucleophile His26 (6.8). Thus, at this predicted pH optimum, only a small fraction of the protein (~ 1%) will be in the catalytically competent ionization state with Glu11 protonated and His26 deprotonated. Such a “reverse protonation mechanism” is well established (albeit underappreciated) for many enzymes and requires that the low population of suitably ionized residues be offset by their high catalytic efficiency.36 It may not seem intuitive that the pH‐dependence of k _cat/K _m for a reverse protonation mechanism would reflect the population‐averaged macroscopic pK _a values of the catalytic residues measured by NMR spectroscopy.16 That is, due to electrostatic interactions, the microscopic pK _a values of proximal Glu11 and His26 will likely differ depending on whether the other is neutral or charged. However, at the predicted pH optimum of ~ 5.8, the major population of T4L with His31 protonated and Glu11 deprotonated will be in a pH‐independent equilibrium with the minor population having the catalytically competent deprotonated His31 and protonated Glu11.

Alternative mechanism for T26H–T4L

Although in principle histidine can serve as a nucleophile for retaining glycoside hydrolases, it is striking that none to date has ever been demonstrated or proposed to play this role in any such enzyme other than T26H–T4L. This prompts the question whether T26H–T4L could follow an alternative catalytic mechanism. One possibility, exemplified by GH20 β‐hexosaminidases,37 involves the nucleophilic participation of the N‐acetyl group of NAM [Fig. 1(B)]. Such participation could become feasible if the T26H mutation results in interactions that are favorable to the requisite alignment of the side chain amide moiety. A particularly attractive feature of this hypothesized mechanism is that we know it is inherently possible for the N‐acetyl moiety to adopt the correct geometry for nucleophilic attack since it does so in other enzymes. In contrast, we do not know with certainty that this is the case for the nitrogen of His26.

In this hypothesized substrate‐assisted catalysis mechanism, Asp20 might act as a general base to abstract a proton from the N‐acetyl nitrogen and assist the formation of a bicyclic oxazoline intermediate. Glu11 would serve as a general acid for aglycone departure and then as a general base for hydrolysis or transglycosylation, with net retention of β‐anomeric stereochemistry. The predicted pH optimum in k_cat/K_m for such a mechanism would be ~ 3.1 with Asp20 (pK _a ~ 1.5) predominantly deprotonated and Glu11 (pK _a 4.7) predominantly protonated in the free enzyme. However, with such a low pK _a value, Asp20 would be a poor general base and might rather remain negatively charged to stabilize a positively charged oxazolinium intermediate without proton transfer.38 It is also worth considering that, in the Michaelis complex with bound substrate, the pK _a values of these catalytic residues could be better tuned for efficient catalysis.

In contrast to the previously proposed mechanism with a histidine‐linked glycosyl‐enzyme intermediate, in this alternative substrate‐assisted pathway Asp20 serves a direct catalytic role whereas that of His26 is not obviously defined. This is conformationally plausible as Asp20 is hydrogen bonded to the N‐acetyl group of NAM in the X‐ray crystallographic structure of glycosylated T26E–T4L,11 yet less consistent with the observations that the T4L–T26Q mutant is inactive and the D20C/T26H double mutant is still active as a retaining glycoside hydrolase.12 However, wild‐type T4L is tolerant to substitutions at position 20, including D20C for which cysteine could serve as a general base.34 Furthermore, the very fact that substituting Thr26 with His changes the stereochemical outcome of peptidogycan hydrolysis suggests an active site plasticity that tips the balance between different catalytic mechanisms. Resolving these mechanisms will require more detailed enzymatic studies with well‐defined substrates.

Arg145 is N ^ε protonated with a high pK _a value

Using NMR spectroscopy, we showed conclusively that the N^ε of all arginines, including Arg145, in both wild‐type T4L* and T26H–T4L* are protonated under typical experimental conditions. Also, from chemical shift arguments, the sidechain guanidinium group of this active site residue in the mutant protein is most certainly ionized with a pK _a > 11. Although largely buried in the X‐ray crystal structures of both proteins, stabilization of the delocalized positive charge of Arg145 may result from extensive hydrogen bonding between the terminal N^η1 and N^η2 with the carboxylate of Glu11 and the N^η2 and N^ε with the sidechain oxygen of Asn101. Considerable debate has centered on the possibility of a deprotonated arginine sidechain in a protein under physiological conditions.25, 39 With an intrinsic pK _a value of nearly 14, a substantial energetic penalty would be paid to accommodate a neutral guanidino group. This penalty is on the order of global protein folding energetics, and thus any such species is unlikely to exist except perhaps in a minor population along an enzymatic pathway in which arginine functions as a general base.40

Ionization states determined from neutron crystallography

Neutron crystallography is a well‐established approach for identifying hydrogens in molecules, including proteins. Indeed, the ionization states of the Asp, Glu, His, and Arg residues found herein by NMR spectroscopy are consistent with those determined in a recently published joint neutron and high‐resolution X‐ray diffraction study of perdeuterated T4L* at pD 6–7 (5VNQ.pdb and 5VNR.pdb).14 Thus it somewhat perplexing why in a similar joint diffraction study of perdeuterated T26H–T4L* at pD 7, Arg145 was reported to be N^ε deprotonated, Asp20 carboxyl protonated, Glu11 partially (~ 60%) carboxyl protonated, and His26 deprotonated at N^ε2 (5XPE.pdb and 5XPF.).13 As we have demonstrated for this protein in solution, Asp20 has a pK _a ~ 1.5 and Arg145 is N^ε protonated with an inferred pK _a > 11. Thus, both are effectively completely charged under physiologically relevant pH conditions. Also, at pH 7, Glu11 (pKa 4.7) and His 31 (pKa 6.8) should be ~ 1% and ~ 40% protonated, respectively. There is no obvious physicochemical reason as to why the substitution of Thr26 to His would dramatically alter the pK _a values of these active site residues in T26H–T4L* versus wild‐type T4L*, or why T26H–T4L* would differ between the solution and crystalline states. Although crystal packing or protein and solvent deuteration could explain the smaller discrepancies for Glu11 and His31, it is difficult to imagine how this could lead to such energetically unfavorable changes for Asp20 and Arg145. Unexplained disagreements between results obtained by NMR spectroscopy and neutron diffraction can also be found in the case of photoactive yellow protein.29, 41 This speaks to the challenges in defining protein protonation states and pK _a values, and the need to use complementary methods to obtain a comprehensive understanding of the pH‐dependent ionization states of biological macromolecules.

Materials and Methods

Protein expression and purification

The clones encoding T4L* (plasmid 18111) and T26H‐T4L* (plasmid 18251) were obtained from AddGene. T4L* is a cysteine‐free pseudo‐wild‐type lysozyme with the C54T and C97A substitutions. Protein expression and purification followed previously described protocols.42 Briefly, ¹³C/¹⁵N‐ or ¹⁵N‐labeled lysozyme was expressed in transformed prototrophic E. coli HMS174 cells. An LB seed culture was used to inoculate M9 media containing 3 g/L 99% ¹³C₆‐glucose and/or 1 g/L 98% ¹⁵NH₄Cl to an initial OD₆₀₀ of ~ 0.1. Expression was induced with 1 mM IPTG at OD₆₀₀ ~ 0.6, and the cells were harvested after ~ 4 h of growth at 37°C by centrifugation.

Samples of T26H–T4L* selectively labeled with ¹³C^γ‐aspartic acid or ¹³C^δ‐glutamic acid were expressed in auxotrophic E. coli strains EA1 (aspC, tyrB, asnAB::Tn5) and DL39 (aspC, ilvE, tyrB, avtA::Tn5), respectively.43 The asnAB mutations prevent the metabolic conversion of aspartic acid to asparagine. Unfortunately, a viable strain with a glnA genotype to prevent the conversion of glutamic acid to glutamine was not available. Overnight seed cultures in LB were used to inoculate minimal media containing 500 mg/L ¹³C^γ‐d/l‐aspartic acid (Cambridge Isotopes) or 800 mg/L ¹³C^δ‐d/l‐glutamic acid (Cambridge Isotopes), along with all other unlabeled amino acids.10, 44, 45 Cells were grown to an OD₆₀₀ of ~ 0.6, induced with 1 mM IPTG, and harvested by centrifugation after growth for 4 h at 37°C.

Cells were lysed by homogenization, loaded onto an SP‐Sepharose cation exchange column (GE Healthcare) in 50 mM sodium phosphate, 2 mM EDTA, pH 6.5, and eluted using a linear gradient of NaCl from 0 to 1 M. Samples were further purified by passage through a Superdex 75 size exclusion column in NMR buffer (100 mM KCl, 30 mM potassium phosphate, pH 5.5), and concentrated to ~ 1.5 mM using a 3 kDa MWCO Amicon ultracentrifugal filter. Finally, ~ 5% D₂O was added as a lock solvent. Protein concentrations were determined using ultraviolet absorbance (predicted ε₂₈₀ = 25,440 M⁻¹ cm⁻¹).

NMR spectroscopy

NMR spectra were recorded at 25°C with Bruker Avance III 500, 600, and 850 MHz spectrometers equipped with the xyz‐gradient TCI cryoprobes. Data were processed with NMRpipe46 and analyzed in NMRFAM‐SPARKY47 and Topspin. Standard heteronuclear scalar correlation experiments, including 3D HNCACB, CBCACONH, H(CCO)‐TOCSY‐NH, and C(CCO)‐TOCSY‐NH,48 were automatically interpreted using PINE49 and verified manually to assign relevant backbone and sidechain ¹H, ¹³C, and ¹⁵N signals. The sidechain carboxyl signals of Asp/Glu were assigned from a 3D H^β/γ ₂C^β/γCO^γ/δ experiment, which was based on a modified HCACO experiment.50 Carboxyl deuterium isotope shifts were measured from 2D H^β/γ ₂(C^β/γ)CO^γ/δ spectra of uniformly ¹⁵N/¹³C‐labeled T26H–T4L* in H₂O NMR sample buffer and then after lyophilization and resuspension in an equal volume of D₂O. The initial sample pH was 5.8, and the final sample pH* (uncorrected for isotope effects) was 5.7. Histidine imidazole signals were readily identified in conventional and constant time ¹³C‐HSQC spectra.51, 52 Arginine ¹⁵N^ε signals were assigned from an H^εN^εC^δ spectrum, recorded using a modified HNCA experiment,53 ¹³C^ζ signals from an H^εN^εC^ζ spectrum, recorded using a modified HNCO experiment,53 and ¹⁵N^η signals from a cross polarization ¹³C‐detected N^ε/η‐C^ζ correlation experiment.26, 29 The modifications involved setting the transmitter frequencies for ¹⁵N^ε (84.5 ppm), ¹³C^δ (43.5 ppm), and ¹³C^ζ (160 ppm), and lengthening the ¹H^ε–¹⁵N^ε de‐phasing/re‐phasing delays to 2.8 ms along with a 1.67 msec ¹⁵N^ε‐selective 180° pulse.

pK_a determination of His, Asp, and Glu

NMR‐monitored pH titrations were recorded at 25°C on a Bruker Avance III 850 MHz spectrometer. The titrations were carried out by transferring the sample between the NMR tube and a 1.5 mL Eppendorf tube and adding small aliquots of NMR buffer with ~ 0.1 M HCl or NaOH. The average pH meter reading, measured before and after spectra acquisition with a ThermoFisher Orion 9110DJWP electrode at room temperature (~ 20°C), was taken as the sample pH value. Histidines in ¹³C/¹⁵N‐labeled T26H–T4L* were monitored between pH 5.0 and 10.7 via ¹³C‐HSQC spectra. The sidechain carboxyl signals of Asp and Glu in selectively labeled protein samples were monitored between pH 1.8 to 7.8 and 1.8 to 9.8, respectively, via direct detection 1D ¹³C‐NMR with ¹H decoupling during acquisition. The pH‐dependent chemical shifts of individual ¹³C and ¹H nuclei were fit with GraphPad Prism to the Henderson–Hasselbalch equation for one or macroscopic ionization equilibria to obtain corresponding pK _a values.16 In the case of each His, results from four reporter nuclei were averaged to obtain the reported pK _a values with standard deviations. In the case of Asp and Glu, errors were estimated ± 0.1 as limited by the accuracy in measuring sample pH value.

Conflict of interests

The authors declare no conflicts of interest with respect to the publication of this article.

Supporting information

Appendix S1: Supplementary material