N-Acetylglucosaminidases from CAZy Family GH3 Are Really Glycoside Phosphorylases, Thereby Explaining Their Use of Histidine as an Acid/Base Catalyst in Place of Glutamic Acid

Background: GH3 β-N-acetylglucosaminidases atypically employ a His-Asp dyad as a catalytic acid base.

Results: Enzymes from this GH3 subfamily are phosphorylases rather than hydrolases.

Conclusion: Replacement of the Glu acid/base by His avoids Coulombic repulsion with phosphate.

Significance: These are the first anomeric stereochemistry-retaining β-glycoside phosphorylases to be found.

Abstract

CAZy glycoside hydrolase family GH3 consists primarily of stereochemistry-retaining β-glucosidases but also contains a subfamily of β-N-acetylglucosaminidases. Enzymes from this subfamily were recently shown to use a histidine residue within a His-Asp dyad contained in a signature sequence as their catalytic acid/base residue. Reasons for their use of His rather than the Glu or Asp found in other glycosidases were not apparent. Through studies on a representative member, the Nag3 β-N-acetylglucosaminidase from Cellulomonas fimi, we now show that these enzymes act preferentially as glycoside phosphorylases. Their need to accommodate an anionic nucleophile within the enzyme active site explains why histidine is used as an acid/base catalyst in place of the anionic glutamate seen in other GH3 family members. Kinetic and mechanistic studies reveal that these enzymes also employ a double-displacement mechanism involving a covalent glycosyl-enzyme intermediate, which was directly detected by mass spectrometry. Phosphate has no effect on the rates of formation of the glycosyl-enzyme intermediate, but it accelerates turnover of the N-acetylglucosaminyl-enzyme intermediate ∼3-fold, while accelerating turnover of the glucosyl-enzyme intermediate several hundredfold. These represent the first reported examples of retaining β-glycoside phosphorylases, and the first instance of free β-GlcNAc-1-phosphate in a biological context.

Introduction

The glycoside hydrolases from CAZy family GH3 display an unusual diversity in structure, specificity, and biological role and are widely distributed in bacteria, fungi, and plants (see the CAZypedia website for a particularly informative discussion on GH3). This sequence-defined family contains β-glucosidases, N-acetylglucosaminidases (often called NagZs), β-xylosidases, and α-arabinofuranosidases. The principal initially identified role of GH3 glycosidases was in biomass degradation, but other roles in pathogen defense and plant and bacterial cell wall remodeling have become apparent. A subset of these enzymes plays a role in the recycling of peptidoglycan by removing the nonreducing end N-acetylglucosamine from the disaccharide product released by lytic transglycosylases and muramidases (GlcNAc-anhydro-MurNAc³-peptide and its hydrated version) in Gram-negative and Gram-positive bacteria, respectively (1). The resultant (anhydro)-MurNAc peptide is an activator of β-lactamase production in some Gram-negative bacteria, rendering the GH3 N-acetylglucosaminidases a possible therapeutic target (2–4). Accordingly, considerable attention has been devoted to the generation of potent, selective inhibitors for this group of enzymes.

Glycosidases from GH3 are retaining enzymes and cleave their substrates in an acid/base-catalyzed two-step double-displacement mechanism involving a covalent glycosyl-enzyme intermediate in which a fully conserved aspartic acid functions as the catalytic nucleophile. Detailed structural and mechanistic studies on several GH3 glycosidases have confirmed that both the β-glucosidases and the N-acetylglucosaminidases follow this mechanism (5–13). This contrasts sharply with the β-hexosaminidases from CAZy family GH20, which lack an enzymatic nucleophile and instead use the acetamide of the substrate in that role, forming an oxazoline or oxazolinium ion intermediate (14, 15). GH20 hexosaminidases also differ from those of GH3 in having both N-acetylglucosaminidase and N-acetylgalactosaminidase activity, whereas no N-acetylgalactosaminidase activity is seen in GH3 enzymes.

Although the catalytic nucleophile is fully conserved within the GH3 enzymes, such is not the case for the acid/base catalyst. In some two-domain enzymes, such as that from barley (9), the nucleophile is located in the active site within the N-terminal (β/α)₈-barrel, and the acid base protrudes into the active site from the C-terminal domain. By contrast, most of the GH3 N-acetylglucosaminidases are single domain enzymes for which, until recently, neither sequence alignments nor structures suggested an obvious candidate for the acid/base catalyst. However, the structure of NagZ from the Gram-positive organism Bacillus subtilis identified a histidine residue on a flexible loop, in conjunction with a nearby aspartic acid, as a candidate for this role (7). Importantly the loop containing this Asp-His dyad was highly conserved and indeed had been identified previously as a signature sequence for N-acetylglucosaminidases within GH3 and is proposed to play a role in binding of the acetamide moiety (10).

This proposal of a histidine as the acid/base was somewhat surprising because in essentially all other glycosidases this role has been played by a glutamate or aspartate, presumably in large part to generate a more anionic environment upon proton transfer that will help stabilize the oxocarbenium ion-like transition state (16, 17). However, as was noted previously (7, 18), a similar replacement of the anionic Glu or Asp with a neutral residue has been seen in several other groups of glycosidases, most notably the sialidases of GH33 and GH34 and the myrosinases of GH1. This substitution was attributed to the fact that the substrates themselves bear a negative charge; thus replacement of an anionic side chain with a neutral residue minimizes Coulombic repulsion. In the case of GH33 and GH34 sialidases, the substrate has a carboxylate group at C-1 and the catalytic nucleophile is a tyrosine residue, whereas in GH1 myrosinases the acid/base glutamate is replaced by glutamine to accommodate the anionic aglycone of the substrate (19–21). The reason for substitution of the acid/base Glu in GH3 N-acetylglucosaminidases by histidine is not obvious. One suggestion was that this would allow turnover of N-acetyl-4-O-[2-(acetylamino)-2-deoxy-β-d-glucopyranosyl]-muramic acid substrates from which the peptide chain has been hydrolyzed, because these bear an anionic lactyl moiety that could be involved in substrate-assisted catalysis (7, 18). However, subsequent studies rendered this explanation unlikely (18).

Further consideration on possible reasons for this Asp-His substitution brought us to the idea that perhaps the negative charge being accommodated was not in the carbohydrate substrate but in the attacking species. The most probable such anionic nucleophile in a cellular environment would be phosphate. If so, then perhaps these enzymes function preferentially as glycoside phosphorylases rather than as glycoside hydrolases. Glycoside phosphorylases are categorized in both GT and GH families within CAZy (22). The only GT-based phosphorylases to date are the α-retaining glycogen phosphorylases (GT35) and trehalose phosphorylases (GT4). Most other phosphorylases are inverting enzymes with major families being the β-inverting cellobiose/cellodextrin and chitobiose phosphorylases (GH94) and the α-inverting GH65 phosphorylases such as maltose phosphorylase, along with a few others in GH112 and GH130 (22–24). The only retaining GH phosphorylases described are GH13 phosphorylases such as sucrose phosphorylase and the recently uncovered maltose-1-phosphate maltosyltransferase, both GH13 α-cleaving/forming enzymes (25, 26). Interestingly, no retaining β-glycoside phosphorylases have been reported to date. Therefore, should these GH3 hexosaminidases prove to function as phosphorylases through an analogous retaining mechanism to that of the GH3 hydrolases, they would represent a new category of phosphorylases.

EXPERIMENTAL PROCEDURES

Materials and Reagents

Substrates used in this study were either purchased from suppliers or donated by members of this laboratory. p-Nitrophenyl N-acetyl-β-d-glucosaminide (pNPGlcNAc) was purchased from Calbiochem; p-nitrophenyl β-d-glucopyranoside (pNPGlc) was purchased from Sigma. 2,4-Dinitrophenyl β-d-glucopyranoside (DNPGlc) and 2,4-dinitrophenyl 2-deoxy-2-fluoro-β-d-glucopyranoside (DNP2FGlc) were generously donated by H. Chen and F. Liu, respectively. Concentration/removal of Nag3 was done using Amicon® Ultra-0.5-ml centrifugal filters (M_r 30,000) (Millipore, Darmstadt, Germany). TLC was done using TLC Silica Gel 60 F₂₅₄ TLC plates (EMD Millipore Corp., Billerica, MA). TLC stain composition was as follows: anisaldehyde reagent (92.5% ethanol, 4% H₂SO₄, 1.5% acetic acid, 2% p-anisaldehyde), and molybdate reagent (2.5% w/v ammonium molybdate, 1% w/v ceric ammonium sulfate, and 10% H₂SO₄).

General Kinetic Methods

Cellulomonas fimi Nag3 was expressed and purified, as reported previously (8), substituting HEPES buffer (50 mm HEPES, pH 7.0, 300 mm NaCl, 5 mm MgSO₄, 10% glycerol, 1 mm DTT) where sodium phosphate buffer was previously used. The enzymatic reaction rates with the chromogenic substrates were determined by monitoring the change in absorbance at 400 nm. Kinetic measurements were performed in matched 1-cm path length quartz cuvettes using a Varian Cary 300 Bio UV-visible spectrophotometer with an automatic cell changer and temperature controller at 25 °C in HEPES buffer. Reactions contained potassium phosphate, pH 7.0, where specified. Hydrolysis/phosphorolysis rates were calculated by measuring absorbance changes as a function of time and converting these to concentration with the following extinction coefficients: 7280 m⁻¹ cm⁻¹ (pNP) and 12,460 m⁻¹ cm⁻¹ (2,4DNP) at 25 °C, pH 7.0. Nonlinear regression was performed using GraphPad Prism version 6.0 or GraFit.

Pre-steady State Kinetics

The pre-steady state phase of DNPGlc binding to Nag3 was captured by performing the experiment in a spectrophotometer cuvette cooled to 12 °C and continuously monitoring absorbance at 400 nm. An assay buffer containing HEPES buffer and DNPGlc was preincubated in a quartz cuvette at 12 °C. Once the baseline was established, 40 μl of Nag3 (20 mg/ml) was added to 160 μl of assay buffer to give final concentrations of 4 mg/ml Nag3 (66 μm) and 250 μm DNPGlc. At 0.5 min, after steady state was observed, phosphate was added to a final concentration of 50 mm.

¹H NMR Analysis

10 mm substrate (pNPGlcNAc, pNPGlc, or DNPGlc) was incubated with 6 mg/ml Nag3 and 0, 5, 25 or 100 mm phosphate in HEPES buffer (with no glycerol) for 2 h (500-μl reaction volume). The incubation temperature was increased to 37 °C (compared with 25 °C used for kinetic assays) to help maximize substrate turnover. After the incubation time, Nag3 was removed with Amicon centrifugal filters, and the filtrate was freeze-dried using a SpeedVac (Savant SV 100) concentrator and dissolved in D₂O (500 μl). ¹H spectra were recorded on a Bruker 400 MHz Avance with inverse probehead. 16 scans employing presaturation water suppression (Bruker standard pulse program “zgcppr”) were recorded per sample.

β-GlcNAc-1-P and β-Glc-1-P Isolation and Characterization

β-Sugar 1-phosphates were prepared by combining 10 mm pNPGlc or pNPGlcNAc with 1 mg/ml Nag3 and 50 mm phosphate in HEPES buffer (10-ml reaction volume). After incubation for 4 h at 25 °C, the samples were freeze-dried using a SpeedVac (Savant SV 100) concentrator. To remove the HEPES buffer, the dried sample was dissolved in H₂O and the pH was adjusted to 5. A short column was filled with Dowex 50WX8 ion exchange resin (5 g), and the resin was washed with deionized water (100 ml). The sample was applied, and fractions of 10 ml were taken that were immediately neutralized (NaOH) after collection. The collected fractions were concentrated (Speedvac), pooled, and applied in H₂O (∼500 μl) to a Biogel P2 column (60 × 1.5 cm, eluent, H₂O, flow rate 0.5 ml/min, fraction size 3.5 ml). Positive fractions were identified by TLC, pooled, concentrated (Speedvac), and analyzed by NMR. Spectra were recorded in D₂O and referenced externally to dioxane in D₂O (¹H, 3.75 ppm and ¹³C, 67.4 ppm) or H₃PO₄ in D₂O (³¹P, 0 ppm). High resolution mass spectra were obtained on a Micromass LCT time-of-flight (TOF) mass spectrometer equipped with an electrospray ion source.

GlcNAc-1-P

¹H(400 MHz, D₂O): δ 4.93 (app t, 1 H, J_1,P = J_1,2 = 8.3 Hz, H-1), 3.91 (dd, 1 H, J_6a,6b = 12.3 Hz, J_6a,5 = 2.1 Hz, H-6a), 3.72 (dd, 1 H, J_2,1 = 8.3 Hz, J_2,3 = 10.4 Hz, H-2), 3.69 (dd, 1 H, J_6b,6a = 12.3 Hz, J_6b,6a = 6.6 Hz, H-6b), 3.54 (dd, 1 H, J_3,2 = 10.4 Hz, J_3,4 = 8.8 Hz, H-3), 3.49 (ddd, 1 H, J_5,4 = 9.8 Hz, J_5,6a = 2.1 Hz, J_5,6b = 6.7 Hz, H-6b), 3.38 (dd, 1 H, J_4,3 = 9.2 Hz, J_4,5 = 9.5 Hz, H-4), 2.05 (s, 3 H, NHAc). ¹³C (100 MHz, D₂O), δ 175.88 (NHAc), 96.48 (d, J_1,P = 5.1 Hz, C-1), 77.14 (C-5), 74.83 (C-3), 70.91 (C-4), 61.96 (C-6), 57.32 (d, J_2,P = 7.7 Hz, C-2), 23.20 (NHAc). For ³¹P (161.9 MHz, D₂O) δ 0.88 HRMS (m/z); TOF-MS ES [MH]⁻ calculated for C₈H₁₅NO₉P, 300.0484, and found, 300.0486.

Glc-1-P

¹H (400 MHz, D₂O): δ 4.83 (app t, J_1,P = J_1,2 7.7 Hz, H-1), 3.85 (dd, 1 H, J_6a,6b = 12.3 Hz, J_6a,5 = 2.1 Hz, H-6a), 3.61 (dd, 1 H, J_6b,6a = 12.3 Hz, J_6b,6a = 6.7 Hz, H-6b), 3.46 (t, 1 H, J_3,2 = J_3,4 = 9.2 Hz, H-3), 3.43 (m, 1 H, H-5), 3.28 (t, 1 H, J_4,3 = J_4,5 = 9.4 Hz, H-4), 3.24 (dd, 1 H, J_2,3 = 9.3 Hz, J_2,1 = 7.9 Hz, H-2). ¹³C (100 MHz, D₂O): δ 97.92 (d, J_1,P = 5 Hz, C-1), 77.09 (C-5), 76.18 (C-3), 74.93 (d, J_2,P = 7.0 Hz, C-2), 70.56 (C-4), 61.81 (C-6). ³¹P (161.9 MHz, D₂O) δ 1.35. HRMS (m/z): TOF-MS ES [MH]⁻ calculated for C₆H₁₂NO₉P, 259.0219, and found, 259.0216.

TLC Analysis

10 mm substrate (pNPGlcNAc, pNPGlc, or DNPGlc) was incubated with 20 mg/ml Nag3 and 0, 1, 5, 10, 25, 50, or 100 mm phosphate in HEPES buffer (with no glycerol) at 37 °C for 15 min (20 μl reaction volume). The reaction mixtures were placed on ice before being spotted on silica TLC plates. The plates were eluted with a mobile phase of 1-butanol, methanol, ammonium hydroxide, and water in a 5:4:4:1 ratio, respectively. Once dried, the pNPGlcNAc plate was stained with anisaldehyde reagent, and the pNPGlc and DNPGlc plates were stained with molybdate reagent and heated until the product signals became visible. Anisaldehyde reagent was used for the pNPGlcNAc plate to better visualize GlcNAc formation.

Electrospray Mass Spectrometry

1 mg/ml Nag3 (in 20 mm Tris, pH 7.0, 50 mm NaCl, 2% glycerol, 5 mm MgSO₄, and 1 mm DTT) was incubated with 1 mm DNPGlc or 1 mm DNPGlc and 50 mm phosphate for 1 h at room temperature (50-μl reaction volume). Samples were analyzed using methods described previously (10, 27).

Reactivation with Phosphate

Nag3 was inactivated by incubating 2 mg/ml active enzyme with 75 mm of the inactivator DNP2FGlc at 25 °C for 5 h in HEPES buffer (50 μl). Excess DNP2FGlc was diluted to ∼2 μm by dilution with HEPES buffer and re-concentrating with Amicon centrifugal filters. Reactivation was monitored by incubating samples of the inactive Nag3–2FGlc complex, (0.2 mg/ml) in HEPES buffer with 0, 10, 25, 50, 100, or 200 mm phosphate at 25 °C (50 μl reaction volume) and transferring 5-μl aliquots at the indicated time points from the reactivation reaction to an assay solution (195 ml) containing HEPES buffer, 0.5 mm DNPGlc, and 20 mm phosphate. Turnover rates were calculated as described above.

RESULTS AND DISCUSSION

The enzyme chosen to test this hypothesis was the GH3 β-glucosaminidase Nag3 from C. fimi, which our laboratory had characterized previously, and was shown to have dual β-glucosaminidase/β-glucosidase activity and to share the conserved β-glucosaminidase signature sequence containing the Asp-His dyad noted earlier (8, 10). All our previous studies were performed in phosphate buffer, so unwittingly phosphorolysis had been monitored rather than hydrolysis. Nag3 was expressed and purified as described previously (8) and exchanged into 50 mm HEPES buffer, pH 7.0. It was then assayed at fixed concentrations of three different glycoside substrates at each of a series of phosphate concentrations. Two glucoside substrates of quite different aglycone leaving group ability (pNP, pK_a = 7, and 2,4DNP, pK_a = 4) were chosen, along with one glucosaminide bearing the same pNP leaving group. As seen in Fig. 1, a strong dependence of rate on phosphate concentrations was observed, with rates increasing rapidly with rising phosphate at low concentrations and inhibition being seen at higher concentrations. This is particularly true for the glucoside substrates, where rate increases of over 100-fold are seen for aryl glucosides upon addition of phosphate. Rates of pNPGlcNAc cleavage in the absence of phosphate were comparable with those for pNPGlc, but the activity increases seen upon phosphate addition to pNPGlcNAc were much more modest.

FIGURE 1.

Phosphate activation of substrate cleavage by C. fimi Nag3 in 50 mm HEPES buffer, pH 7.0, at 25 °C. Michaelis-Menten and Lineweaver-Burk (right panel) plots of initial rates of hydrolysis/phosphorolysis of A pNPGlcNAc: ○, 1 mm; ●, 2 mm; □, 5 mm; ■, 10 mm; and ▵, 13 mm. B, pNPGlc: ○, 5 mm; ●, 10 mm; □, 25 mm; ■, 35 mm; and ▵, 50 mm. C, DNPGlc: ○, 0.063 mm; ●, 0.175 mm; □, 0.25 mm; ■, 0.5 mm; and ▵, 1 mm. The rate data were fit using the Michaelis-Menten equation incorporating substrate inhibition: V₀ = V_max·[S]/(K_m + [S]·(1 + [S]/K_i)). Lineweaver-Burk plots were fit to the equation for the ping-pong mechanism: 1/V₀ = (K_a/V_max)·(1/[A]) + (1/V_max)·(1 + K_b/[B]). Error bars are shown on data points, and where bars are not apparent, the errors were smaller than the symbol size used.

Formation of Sugar Phosphate Products

To ensure that these effects have their origins in phosphate acting as a co-substrate and not simply as an allosteric activator, the reactions were monitored by TLC and NMR, and the products were thereby identified. As can be seen in Fig. 2, reaction of all three substrates in the presence of increasing phosphate concentrations results in a change in the product formed from the free sugar (glucose or N-acetylglucosamine) to a product with a very similar TLC mobility to that of α-glucose 1-phosphate. This finding is consistent with the enzyme acting as a glycoside phosphorylase within or indeed below the physiologically relevant phosphate concentration range (20–30 mm) (28). Confirmation of the formation of a sugar 1-phosphate in each case and determination of the anomeric stereochemistry of that sugar phosphate was achieved by isolating the sugar phosphate and analyzing it by mass spectrometry and NMR. As can be seen in Fig. 2, the products are the β-anomers as follows: β-glucose 1-phosphate and β-N-acetylglucosamine 1-phosphate, with anomeric stereochemistries being confirmed by the large J_1,2 couplings of 7.9 and 8.3 Hz seen for Glc-1-P and GlcNAc-1-P, respectively. These results therefore confirm that Nag3 functions as a phosphorylase in each case and furthermore that it is a β-retaining phosphorylase, the first such enzyme identified. Indeed, not only is this the first reported example of a retaining β-glycoside phosphorylase, but also, to our knowledge, this is the first report of the occurrence of β-GlcNAc-1-phosphate itself in a biochemical system. The β-GlcNAc-1-phosphate linkage occurs within the undecaprenyl-phosphate donor sugars involved in the assembly of lipoteichoic acids, but β-GlcNAc-1-phosphate itself is not involved in their biosynthesis (29). Although we had initially assumed it would be an extremely labile derivative, thus hard to isolate due to the potential for facile oxazoline formation, it proved to be reasonably stable, as indeed reported by Martin and co-workers (30) for their synthetic material.

FIGURE 2.

Top panel, ¹H NMR spectra for Nag3 and 10 mm pNPGlcNAc, pNPGlc, and DNPGlc with 0, 5, 25, and 100 mm phosphate. Insets show ×4 amplification of the α-Glc/α-GlcNAc peak regions. Unfortunately the β-GlcNAc and β-Glc anomeric proton falls right under the large peaks from residual HOD, thus only the peaks from α-anomers are shown. Bottom panel, TLC analysis of Nag3 hydrolysis and phosphorolysis of 10 mm pNPGlcNAc, pNPGlc, and DNPGlc with 0, 1, 5, 10, 25, 50, and 100 mm phosphate. The aryl glycoside starting materials (SM) are labeled.

Determination of Kinetic Parameters

Presumably, this β-retaining glycoside phosphorylase also follows a double displacement mechanism like that of the β-retaining glycosidases, wherein phosphate serves as the ultimate nucleophile in place of water, attacking the glycosyl-enzyme intermediate as shown in Scheme 1. Ιn order to confirm this, kinetic parameters for the phosphorolysis and hydrolysis reactions were measured for each substrate, and these kinetic parameters are presented in Tables 1 and 2. As can be seen, values of both k_cat and K_m for pNPGlcNAc increase with phosphate concentration in a saturable fashion, according to an apparent K_m value for phosphate of 17 mm (Table 1). This behavior is consistent with phosphate increasing the deglycosylation rate constant (k₃) up to a limiting value of ∼0.2 s⁻¹. Values of k_cat/K_m remain constant, consistent with the fact that k_cat/K_m reflects the first irreversible step, which in this case will be the formation of the glycosyl-enzyme intermediate with concomitant loss of the phenolate. This behavior is much more extreme for the glucosides, with k_cat and K_m values also increasing with phosphate concentration but with little sign of apparent saturation within reasonable substrate concentrations. Again k_cat/K_m values remain roughly constant. Notably, the K_m value for DNPGlc in the absence of phosphate is remarkably low, at ∼3 μm, whereas the k_cat value is also very low at 0.04 s⁻¹. This is the expected behavior for cleavage of a substrate containing an excellent leaving group by an enzyme following such a double-displacement mechanism, for which the K_m is given by Equation 1,

SCHEME 1.

Mechanistic scheme for a retaining β-glycoside phosphorylase-glycosidase. For Nag3, k₂ has a similar value for 4-nitrophenyl glycosides where Y = OH and Y = NAc. Values of k_3W = 0.05 s⁻¹ are found for both Y = OH and NAc. Values of k_3P = 0.2 s⁻¹are found for Y = NAc, whereas k_3P is greater than 100 s⁻¹ for Y = OH.

TABLE 1.

Apparent kinetic parameters for phosphate activation of substrate cleavage by Nag3

Apparent values of K_m and K_i were calculated by fitting the rate data shown in Fig. 1 to the Michaelis-Menten equation incorporating substrate inhibition as follows: V₀ = V_max·[S]/(K_m + [S]·(1 + [S]/K_i)). K_a and K_b values were calculated by fitting the rate data to the equation for a ping-pong kinetic mechanism: V₀ = (V_max·[A]·[B])/(K_a·[B] + K_b·[A] + [A]·[B]).

Substrate	Km	Ki	Kb (NP-sugar)	Ka (phosphate)
	mm	mm	mm	mm
pNPGlcNAc	17 ± 6	400 ± 200	4.7 ± 1.0	6.0 ± 2.4
pNPGlc	0.3 ± 0.05	210 ± 30	64 ± 5	0.7 ± 0.06
DNPGlc	3 ± 0.3	320 ± 50	0.43 ± 0.03	0.8 ± 0.05

TABLE 2.

Kinetic parameters for the reaction of Nag3 with pNPGlcNAc, pNPGlc, and DNPGlc

The reactions were carried out in 50 mm HEPES buffer, pH 7.0, and the indicated concentration of potassium phosphate, pH 7.0, at 25 °C. The molar extinction coefficient (m⁻¹ cm⁻¹) at 400 nm was 7280 for pNP and 12,460 for 2,4 DNP. Parameters were calculated using the following: V₀ = [E]·k_cat·[S]/(K_m + [S]). Values of k_cat/K_m for pNPGlc were measured at concentrations of phosphate from 5 to 50 mm and ranged between 0.034 and 0.037 s⁻¹ mm⁻¹: high K_m values precluded measurement of individual parameters for k_cat and K_m.

Substrate	[Pi]	Km	kcat	kcat/Km
	mm	mm	s−1	s−1 mm−1
pNPGlcNAc	0	1.6 ± 0.2	0.05 ± 0.002	0.032
	10	2.7 ± 0.2	0.07 ± 0.002	0.027
	25	3.6 ± 0.2	0.09 ± 0.001	0.026
	50	4.3 ± 0.3	0.10 ± 0.003	0.023
	100	5.7 ± 0.5	0.15 ± 0.006	0.027
	200	7.3 ± 0.9	0.16 ± 0.010	0.022
pNPGlc	0	2.0 ± 0.5	0.04 ± 0.002	0.020
	0.1	8.9 ± 0.6	0.43 ± 0.01	0.048
	0.5	45 ± 6	2.0 ± 0.1	0.044
	1	149 ± 18	6.4 ± 0.6	0.043
	2.5	190 ± 72	8.5 ± 2.7	0.045
DNPGlc	0	0.003 ± 0.001	0.04 ± 0.001	13
	0.1	0.09 ± 0.02	3.1 ± 0.1	32
	0.5	0.20 ± 0.01	8.2 ± 0.2	40
	1	0.23 ± 0.02	10 ± 0.3	45
	2	0.29 ± 0.03	13 ± 0.5	45
	5	0.28 ± 0.02	14 ± 0.4	49
	10	0.34 ± 0.03	16 ± 0.7	48
	25	0.42 ± 0.04	19 ± 0.9	45
	50	0.55 ± 0.06	23 ± 1.3	42

In this scenario, as k₃ decreases relative to k₂, the K_m value will drop. Because 2,4DNP (pK_a = 4) is a much better leaving group than pNP (pK_a = 7), it is reasonable that k₂ for DNPGlc will be less than that for pNPGlc. Because it seems that k₃ is already the rate-limiting step, based upon essentially identical values of k_cat for pNPGlc and DNPGlc, a very low K_m value is expected for DNPGlc.

Interestingly, as noted earlier, the k_cat for pNPGlcNAc also has a similar value; thus the presence of the acetamide has little effect on the hydrolytic deglycosylation step. However, this is clearly not the case for the phosphorolytic deglycosylation step, where k₃ for phosphorolysis of the N-acetylglucosaminyl-enzyme intermediate has a value of 0.2 s⁻¹, only 4-fold higher than that for hydrolysis, whereas k₃^P for phosphorolysis of the glucosyl-enzyme intermediate is well in excess of 100 s⁻¹, thus much more than 500-fold higher than that for hydrolysis.

Reciprocal plots of the data from which kinetic parameters for each substrate in Table 2 are derived yields the three plots shown in the right-hand column of Fig. 1. These plots, with their parallel lines, are diagnostic of a ping-pong mechanism. The presence of substrate inhibition by phosphate leads to the deviations observed, which are particularly apparent at low pNP-glycoside concentrations. This kinetic behavior is what is predicted for a mechanism involving a covalent glycosyl-enzyme intermediate but is not seen for glycosidases due to the fact that the concentration of the second substrate, water, cannot realistically be altered.

Observation of Covalent Intermediate Accumulation

This interpretation of these kinetic parameters would mean that the glycosyl-enzyme intermediate has a half-life of ∼17 s under these conditions, suggesting that a pre-steady state “burst” should be observable in a standard UV-visible spectrometer, especially if the experiment is performed at a lower temperature to slow deglycosylation further. DNPGlc was chosen as the substrate to test this because it has the best leaving group (2,4-dinitrophenolate) with the highest extinction coefficient. Incubation of high concentrations of Nag3 with DNPGlc in a spectrophotometer following the absorbance at 400 nm produced a clear burst phase, followed by a steady state, as shown in Fig. 3. The magnitude of this initial burst, once converted to molar terms using the extinction coefficient for 2,4DNP corresponded perfectly with the enzyme concentration employed. Subsequent addition of phosphate (after ∼0.5 min) resulted in a very steep increase in rate due to the acceleration of the deglycosylation step through phosphorolysis of the glycosyl-enzyme intermediate. Final leveling off of the absorbance increase was due to substrate depletion.

FIGURE 3.

Pre-steady state burst of DNP released by Nag3 upon incubation at 12 °C with 250 μm DNPGlc, followed by activation with 50 mm phosphate. Solid arrow indicates addition of 33 μm Nag3. Dotted arrow indicates addition of 50 mm phosphate.

Further confirmation of the accumulation of a glucosyl-enzyme intermediate was obtained by use of electrospray ionization mass spectrometry to directly observe this intermediate, as shown in Fig. 4. The mass of the enzyme incubated in the presence of DNPGlc (Fig. 4B) is clearly seen to be greater than that of the free enzyme (Fig. 4A) by the mass of a glucosyl moiety, confirming the steady state accumulation of a covalent glucosyl-enzyme intermediate. Interestingly, inclusion of phosphate in this reaction mixture leads to the spectrum shown in Fig. 4C, where a mixture of free enzyme and glucosyl-enzyme intermediate are seen, consistent with the stimulation of turnover via phosphorolysis, leading to a lower steady state concentration of glycosyl enzyme.

FIGURE 4.

Electrospray mass spectra of Nag3 (A), Nag3 + 1 mm DNPGlc (B) and Nag3 + 1 mm DNP-Glc + 50 mm phosphate (C) incubated at room temperature for 1 h. The mass shift (162 Da) of the peak of the glycosyl-enzyme (calculated mass = 61,107.9 Da; observed mass = 61,109.0 Da) shown in B compared with the peak of the enzyme only (calculated mass = 60,944.9 Da; observed mass = 60,947.0 Da) shown in A is consistent, within error, to a glucosyl residue (163 Da) covalently bound to Nag3.

Is Phosphate the Natural Nucleophile?

One possible concern might be that the reaction with phosphate is just a reflection of a general “anion rescue” analogous to that seen with designed acid/base mutants of retaining glycosidases, where azide or acetate have proven to be the most effective (16, 31). This possibility was tested by measuring rates at saturating DNPGlc, for which deglycosylation is rate-limiting, in the presence of a variety of anions. As shown in Fig. 5, phosphate is indeed the best reactivator, by a wide margin. By contrast, phosphate has never been seen to function effectively in anion rescue of glycosidase mutants. Phosphate would therefore seem to be the natural substrate for this enzyme; thus Nag3 should be re-classified as a phosphorylase.

FIGURE 5.

Seven anionic salts were tested as activators of DNPGlc cleavage by Nag3. 50 mm d-glucuronic acid, sodium acetate, sodium citrate, sodium formate, potassium sulfate, sodium azide, or potassium phosphate were incubated with 0.1 μm Nag3 in 50 mm HEPES buffer containing 500 μm DNPGlc for 5 min at 25 °C.

Further insights into the role and importance of the phosphate reaction were obtained by trapping the reaction intermediate as the 2-fluoroglucosyl enzyme species by reacting the enzyme with DNP2FGlc and then measuring reactivation of the purified species in the presence of phosphate. As can be seen in Fig. 6, phosphate indeed stimulates the turnover of this species in a concentration-dependent manner. However, the instability of Nag3, along with the tendency of phosphate to stabilize it, unfortunately makes the extraction of definitive kinetic parameters difficult. Nonetheless, it is clear that even the lowest phosphate concentrations studied lead to substantial reactivation.

FIGURE 6.

Inactive 2FGlc-Nag3 was reactivated by incubation with increasing concentrations of phosphate. Samples were incubated at 25 °C in the following: 0 mm (●); 25 mm (♢); 50 mm (□); 100 mm (▵); and 200 mm (○) phosphate. Aliquots were assayed in HEPES buffer containing 500 μm DNPGlc and 20 mm phosphate.

Nag3 as a Phosphorylase and Its Role

As is seen most clearly in Fig. 2, once the phosphate concentration approaches 10 mm, the sugar phosphate is the almost exclusive product, with essentially no hydrolysis observed, whether the sugar donor is a glucoside or an N-acetylglucosaminide. Importantly, it should be noted that even though artificial aryl glycosides were employed to detect and characterize this behavior, exactly the same product partitioning would be expected whatever the nature of the donor glucoside or N-acetylglucosaminide, because the glycosyl enzymes formed are identical in each case. Thus, even though the identity of the natural substrate for Nag3 is not clear despite considerable investigative efforts (8) repeated again here, its cleavage will occur via phosphorolysis under normal physiological conditions.

The question now becomes the following. Why have these enzymes evolved to function as phosphorylases, and what is the fate of the β-phosphate products so formed? A clue to this, at least for C. fimi, comes from the presence within the same operon of a gene for a β-phosphoglucomutase. This enzyme would convert the β-phosphate into the corresponding sugar 6-phosphate, which would then feed directly into metabolism.

Structural/Mechanistic Implications

Ordinarily, glycoside hydrolases do not also function as phosphorylases, presumably in large part because the presence of the catalytic Glu and/or Asp residues provides Coulombic screening against binding of anions. This would be particularly important within the cellular milieu because otherwise the resident glycosidases would be constantly degrading key metabolic intermediates. Indeed, this is likely one of the reasons why carboxylic amino acids have assumed these roles in most glycosidases. It would appear that this subset of GH3 glycosidases has broadened their repertoire to allow phosphate into the active site by replacing the acid/base catalytic glutamate with a histidine residue, itself activated by an adjacent, but presumably sufficiently electrostatically remote, aspartate residue. This strategy of charge-avoidance has been seen previously in, for example, some sialyltransferases, where the catalytic base is also a histidine residue to avoid electrostatic interactions with the anionic substrate CMP-sialic acid, as shown by structural and NMR studies (32). Even more relevant are the inverting GH65 α-glycoside phosphorylases, which are closely structurally homologous to the inverting glucoamylases of GH15. In this case the phosphate moiety is bound in the GH65 phosphorylase at a location entirely equivalent to that of the general base, Glu400, of a corresponding GH15 glucoamylase (33).

The kinetic parameters measured reveal that phosphate has no noticeable effect on the glycosylation rate constant, as reflected in k_cat/K_m. Only an inhibitory effect is seen, and this only at high concentrations as seen in Fig. 1. This would be consistent with phosphate needing to compete for binding in the +1 site but only doing so weakly. Comparison of k_cat/K_m values for pNPGlcNAc and pNPGlc, which have identical leaving groups, reveals that the presence of OH or NAc at C-2 makes no real difference to the glycosylation rate constant. This was also the case for the hydrolytic deglycosylation rate constants, where both have values of 0.05 s⁻¹. However, the finding that phosphate only increases the deglycosylation rate constant for the N-acetylglucosaminyl enzyme some 4-fold, while it accelerates the breakdown of the glucosyl enzyme many hundredfold, demands some explanation. Perhaps the most logical explanation is one based on the notion that the conserved loop bearing the His-Asp dyad also forms the binding site for the acetamide moiety. Indeed, it was on this basis that this sequence was first identified (10). Thus tight ordering of the otherwise highly flexible loop by binding to the acetamide likely pulls the His-Asp dyad into close proximity to the substrate to provide some base catalysis of the hydrolytic reaction for GlcNAc substrates. Possibly this close contact sterically inhibits the approach of the phosphate to some degree, although phosphorolysis is still the preferred outcome. In the case of the glucosyl enzyme, the loop is not as tightly ordered, and as a consequence the phosphate can approach much more freely, and thus higher reaction rates are attained.

The discovery that Nag3, and by association other hexosaminidases from GH3, acts as a phosphorylase rather than a hydrolase provides a simple explanation for the use of a His-Asp dyad as an acid/base catalyst rather than the standard Glu or Asp residues by this sub-group of GH3 enzymes. The approach of an anionic phosphate to an active site bearing a deprotonated carboxylate base directly adjacent to the anomeric center of the glycosyl enzyme would be greatly disfavored, but it is avoided in this way. Nag3 therefore represents the first reported example of a retaining β-glycoside phosphorylase, and indeed to our knowledge, this represents the first report of the formation of free β-GlcNAc-1-phosphate in a biological system, although chemical syntheses have been reported. The presence of a β-phosphoglucomutase within the same operon as Nag3 in C. fimi suggests that the β-1-phosphates so produced are directly converted to their 6-phosphates for further metabolic conversion. These results also have implications for the design of inhibitors for NagZ enzymes as potential anti-bacterial agents; it may well prove fruitful to include phosphate moieties or surrogates within the structures to be tested.