SEARCH

SEARCH BY CITATION

Keywords:

  • GH10 xylanase;
  • glycosidase;
  • saturation transfer difference;
  • STD NMR ;
  • sugar binding

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

1H solution NMR spectroscopy is used synergistically with 3D crystallographic structures to map experimentally significant hydrophobic interactions upon substrate binding in solution under thermodynamic equilibrium. Using saturation transfer difference spectroscopy (STD NMR), a comparison is made between wild-type xylanase XT6 and its acid/base catalytic mutant E159Q – a non-active, single-heteroatom alteration that has been previously utilized to measure binding thermodynamics across a series of xylooligosaccharide–xylanase complexes [Zolotnitsky et al. (2004) Proc Natl Acad Sci USA 101, 11275–11280). In this study, performing STD NMR of one substrate screens binding interactions to two proteins, avoiding many disadvantages inherent to the technique and clearly revealing subtle changes in binding induced upon mutation of the catalytic Glu. To visualize and compare the binding epitopes of xylobiose–xylanase complexes, a ‘SASSY’ plot (saturation difference transfer spectroscopy) is used. Two extraordinarily strong, but previously unrecognized, non-covalent interactions with H2–5 of xylobiose were observed in the wild-type enzyme but not in the E159Q mutant. Based on the crystal structure, these interactions were assigned to tryptophan residues at the −1 subsite. The mutant selectively binds only the β–xylobiose anomer. The 1H solution NMR spectrum of a xylotriose–E159Q complex displays non-uniform broadening of the NMR signals. Differential broadening provides a unique subsite assignment tool based on structural knowledge of face-to-face stacking with a conserved tyrosine residue at the +1 subsite. The results obtained herein by substrate-observed NMR spectroscopy are discussed further in terms of methodological contributions and mechanistic understanding of substrate-binding adjustments upon a charge change in the E159Q construct.


Abbreviations
DSS

sodium 3-(trimethylsilyl)propane-1-sulfonate

GH

glycoside hydrolase

ITC

isothermal titration calorimetry

STD

saturation transfer difference

X1

xylose

X2

xylobiose

X3

xylotriose

XT6

GH10 endo-xylanase from Geobacillus stearothermophilus T-6

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Enzymes achieve rate accelerations as large as 20 orders of magnitude over uncatalyzed reactions [1]. This requires precise positioning of the bound substrate, with local structural constraints of the order of 0.1–1 Å [2]. Designs for therapeutic agents [3] or synthetic enzymes with tailored and/or multiple catalytic properties benefit from understanding and control of the underlying forces and energy landscapes. The goal of understanding biomolecular functionality, which depends on structure and also dynamics, requires additional information beyond currently available frozen structures [4] . Superimposition of energetic and dynamic insights onto high-resolution 3D snapshots, together with exposure of invisible states [5] and quantum mechanical calculations, synergistically reveals the subtleties by which enzymes function [2].

Glycosidic hydrolases (GH) are key enzymes in nature. Genes encoding GHs constitute approximately 1–3% of the genome in nearly every living organism [6]. Approximately two-thirds of the carbon in the biosphere is in the form of carbohydrates, mostly as cellulose and hemicelluloses. The spontaneous half-life of glycosidic bonds is estimated in the millions of years [7-9]. Over 130 000 glycosidases have been identified and classified into approximately 130 GH families [8, 10]. Glycoside hydrolases within the same family have similar 3D structures, homologously placed catalytic residues, and the same mechanism of hydrolysis, specifically retaining or inverting the anomeric configuration after cleavage [8].

The endo-1,4–β-xylanases (β–1,4–d-xylan xylanhydrolase; EC 3.2.1.8) are extremely well-studied retaining GHs (with molecular masses of approximately 40 kDa), with over 100 high-resolution crystal structures of enzymes and enzyme complexes available to date [10]. The structural scaffold of these enzymes, the (β/α)8 (triosephosphate isomerase, TIM) barrel (Fig. 1A), is one of the most abundant and stable protein folds known [8, 11]. Structural conservation and rigidity are demonstrated by the small RMSD (< 3 Å) between crystalline and solution structures [12, 13]. Furthermore, in-depth intra-family structural comparisons vary by approximately 1 Å (RMSD) upon superimposition of the Cα of the catalytic domains, despite significant differences in enzyme thermostabilities, and whether or not the xylanase was complexed with xylobiose or related compounds [14].

image

Figure 1. Xylopentaose bound to xylanase T–6 from Geobacillus stearothermophilus as determined by X–ray crystallography (PDB code 1R87) drawn using UCSF Chimera [48]. (A) View of the protein, showing a TIM or (β/α)8 barrel fold. The C–centered monoclinic crystal system (space group C2) structure (1R87) is represented by a gray ribbon [33]. The native protein (PDB code 1R85) is overlaid in red. The same native protein, crystallized in the primitive trigonal space group P3221 (PDB code 1HIZ), is superimposed in blue [55]. Scale bar = 10 Å. (b) Magnification showing the catalytic Glu residues (E159 and E265) in yellow. The xylopentaose carbon atoms are shown in black, oxygen atoms in red, and hydrogen atoms in white. The subsites are numbered and labeled according to convention [56], with +2 at the top and −3 at the bottom. (C) Ribbon and (D) surface representations showing binding of xylopentaose onto the surface of a deep cleft of the protein.

Download figure to PowerPoint

The mechanism by which these enzymes perform general acid catalysis has traditionally been inferred via biochemical studies involving synthetic transition-state analogs and catalytic mutants. However, recent high-resolution structural studies of enzyme–substrate/product complexes have combined NMR spectroscopy and X–ray crystallography to contribute new insights into mechanisms of catalysis [15-18]. Correct substrate positioning finely tuned through specific binding interactions with the enzyme is a prerequisite for hydrolysis of the glycosidic bond. To map hydrogen bonding networks [19] of enzyme–substrate complexes and generate a list of van der Waals contacts from protein structures are routine tasks. However, experimental ranking of the non-covalent enzyme–substrate interactions that determine successful binding may differ from nominal proximities. 1H solution NMR saturation transfer difference spectroscopy (STD NMR) may be used synergistically with structural data to recognize, map and interpret experimentally significant hydrophobic-binding interactions.

STD NMR [20, 21] is a technique that tags hydrogen atoms of the substrate interacting with the protein under conditions of thermodynamic equilibrium. It has been used for (a) group epitope mapping of non-labile hydrogen atoms [22], including anomeric specificities [23], (b) screening of substrate libraries to identify micromolar range binders [24, 25], and (c) as a probe for comparing variations in binding to a specific substrate between similar proteins [26-29] or different protein environments, such as a native versus a reintegrated membrane protein [30].

Well-known advantages of STD NMR include (a) the relative ease of 1H detection, (b) use of non-labeled proteins in relatively small quantities without an upper limit on protein molecular weight, and (c) the ability to highlight atom-by-atom differences in cross-relaxation efficiencies (non-covalent interactions) for non-labile hydrogen atoms [22, 24, 31]. Limitations of STD NMR arise from (a) a dependence on appropriate residence times of bound substrate within the protein active site (Doc. S1 and Table S1), (b) the loss of information when comparing atoms with non-similar NMR relaxation time constants at saturation times that exceed the shortest T1, and (c) the loss of atom-by-atom specificity where spectral overlap occurs. Here we give examples that are challenged by these disadvantages.

The wild-type GH10 endo-xylanase from Geobacillus stearothermophilus T–6 (XT6, molecular mass 43.8 kDa) is excreted by the organism into the extracellular environment to depolymerize the β–1,4-glycosidic bonds of xylan to short decorated oligosaccharides that are then transported into the cell by specific ABC sugar transporters [32]. The three-dimensional structure of XT6 in complex with xylopentaose (Fig. 1; PDB code 1R87) demonstrates substrate binding at a long, deep groove on the surface of the protein corresponding to the −3 to +2 subsites. Glycosidic bond cleavage between the −1 and +1 subsites is mediated by E159, the acid/base catalyst (general proton donor/acceptor), and E265, the catalytic nucleophile (Fig. 1B) [33, 34].

In this study, XT6 is compared with its single-heteroatom mutant, E159Q, in which the glutamic acid -COOH is changed to glutamine -CONH2 [19].We have succeeded in using STD NMR [21] to clearly identify subtle changes between WT XT6 and the non-catalytic mutant E195Q. We have observed differences in anomeric specificity, identified important residues for hydrophobic interactions between the protein and the bound substrate, and assigned subsite binding positions for the xylan rings. Using the results from this study in conjunction with the crystallographic structural information provides a framework through which the observed binding changes induced by the charge change in the mutant can be understood.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Xylose

STD NMR on solutions of xylose (X1) with either WT XT6 [35] or E159Q XT6 [36] did not show any saturation transfer from the enzyme to the monosaccharide. Therefore, xylose was subsequently used as an internal control for the xylobiose STD NMR measurements (Fig. 2A). As all the exchangeable protons (namely the OH groups) undergo solvent exchange with D2O, they are not detected in 1H NMR spectra.

image

Figure 2. High-resolution 1H and STD NMR spectra using a single substrate (xylobiose) to screen for binding differences between wild-type and E159Q XT6 (initial solutions had 40-fold molar excess of substrate). (A) Chemical shift assignment of the monosaccharide xylose. Chemical shift assignments for both X1 and X2 include the anomeric signifier, α or β, when referring to the reducing end of the sugar. In the disaccharide, only β–anomers are present at the non-reducing end. (B, C) Off-resonance STD NMR spectra of xylobiose shortly after addition of WT XT6 (B), and 2 days later (C), after hydrolysis of half the substrate has occurred, demonstrating residual catalytic activity. Well-resolved peaks corresponding to xylose, resulting from hydrolysis of the disaccharide by the xylanase, are indicated by asterisks. All xylose peaks are accounted for. (D) X2–WT XT6 STD NMR spectrum. The same saturation transfer difference spectrum is observed whether a low or high concentration of xylose is present. (E) X2–E159Q XT6 STD NMR spectra. Both STD NMR spectra shown have identical acquisition (5.0 s saturation time) and processing parameters. Note the absence of peaks arising from the α–anomer, indicated by the double dagger. (F) The subsite assignment and hydrogen atom numbering of the two xylose rings of xylobiose (PDB code 1R87) drawn using UCSF Chimera [48]. Hydroxyl hydrogen atoms are not shown as a reminder that they undergo chemical exchange with the deuterated solvent and therefore do not appear in the 1H NMR spectra. (G) Xylobiose (top trace) 1′ and 3′ peaks show no differences between α- or β–anomers until protein is added (bottom trace, with WT XT6).

Download figure to PowerPoint

Xylobiose

The xylobiose–WT XT6 (X2–WT XT6) 1H NMR spectrum is shown in Fig. 2B. The red asterisk at 3.9 p.p.m. marks the H–5βeq of X1. This well-resolved peak is an unambiguous indicator for the presence of xylose, and increases slowly as a function of time (Fig. 2C). Over the course of the NMR experiments, the wild-type enzyme slowly hydrolyzed X2 to two molecules of X1 (note the overlap of the xylose and xylobiose 1β peaks at 4.6 p.p.m.; complete spectral assignments are given in Experimental procedures. Further details and supporting spectra are given in Figs S1–S3). 1H NMR spectral changes were used to track the extent of hydrolysis over a period of approximately 10 days, the slope of the linear decrease in absolute intensity of the H–5βeq peak as a function of time yields a hydrolysis rate of 5.2 μm·h−1. This translates into a specific catalytic activity of 4 × 10−5 μm·min−1·mg−1. Eventually, xylobiose was completely hydrolyzed to xylose, after which no further changes to the sample occurred. An initial 40-fold molar excess of xylobiose over WT XT6 is sufficient to ensure the persistence of xylobiose in solution over several days, despite residual catalytic activity. This allows sufficient time to perform experiments with a variety of STD NMR saturation times and/or variable temperatures with adequate signal-to-noise ratios.

Different ratios of xylose to xylobiose in the NMR sample did not affect the resulting 1H STD NMR spectra; only peaks due to xylobiose appear (Fig. 2D) and the spectra are identical (data not shown). This emphasizes the usefulness of STD NMR as a spectral editing technique for enzyme-binders with appropriate residence times. Dissociation constants from 10−3 to 10−8 m have been estimated as suitable for STD NMR. Assuming a diffusion-controlled association rate (kon) of 108 s−1·m−1 [22, 24], this corresponds to a residence time (τres = 1/koff = Kd/kon) of 10 μs−1 s (Doc. S1 and Table S1). This also confirms the usefulness of X1 as an internal experimental control. The buffer peak at 3.7 p.p.m. (Fig. 2B,C) is also absent from the STD spectra (Fig. 2D,E), serving as an additional internal control for the STD NMR experiment.

Xylobiose 1H NMR spectra show an α : β anomeric ratio of 36 : 64 (Fig. S2). Binding of the α–anomer of X2 differs between the two proteins. Wild-type XT6 binds either anomer. E159Q XT6 shows no significant STD effect for the α–anomer of X2 (Fig. 2E; red double daggers indicate the missing signals of the α–anomer). Thus, changing a single heteroatom in the mutant E159Q xylanase, where the side-chain hydroxyl of glutamate is replaced by an amine functional group, induces anomeric selectivity for the β–anomer only. Comparison of binding of the β–anomer between the two proteins was considered next. Overall, the relevant peaks in the STD NMR spectra are apparently similar (Fig. 2D,E and Fig. S4). This suggests that the gross features of binding, such as the primary subsites occupied by the two rings of xylobiose, are conserved regardless of the E159Q mutation.

The second ring of X2, for both anomers, is locked into the β–anomeric configuration (Fig. 2F; the non-reducing end is indicated by numbering with a prime symbol). The data show differences in the second ring of the bound xylobiose α- and β–anomers. When X2 is free in solution, the 1′ resonance (4.45 p.p.m., doublet) and 3′ resonance (3.42 p.p.m., triplet) are insensitive to the anomeric configuration at the reducing end of the disaccharide (Fig. 2G). In the presence of XT6, the α- and β–anomers are shifted differently. This loss of degeneracy indicates that the anomers show different chemical shifts corresponding to their different chemical environments when bound (shielding). For X2–WT XT6, the spectra show that H–1′ of the α–anomer is more shielded than that of the β–anomer, and vice versa for the 3′ hydrogen atom. Assignment of the two anomers (Fig. 2E) is based on (a) the STD NMR of X2–E159Q XT6, which exclusively shows the β–anomer peaks, and (b) the known α : β anomeric ratio. These effects, although subtle, are clearly present.

To probe subtle binding differences between X2–XT6 complexes, STD NMR spectra were acquired at a short saturation time (0.5 s) suitable for epitope mapping [37]. The fractional STD effect, (I0 − Isat)/I0 [22], was computed for each peak and plotted with respect to the 1H NMR spectrum (Fig. 3). This graphical representation is referred to as a SASSY plot (saturation difference transfer spectroscopy). The T1 values for X2 are the same for both X2–XT6 complexes, but differ from atom to atom within xylobiose (Table S2). In the SASSY plot, the data points are color-coded to provide information at a glance regarding the relative T1 times for the free substrate: short (white symbol), medium (pale orange symbol) and long (black symbol). The strongest STD effect was found for the xylobiose axial and equatorial hydrogen atoms at 5β in the X2–WT XT6 solution. In the plot, fractional STD values for both wild-type and mutant are normalized to the two H2–5β of X2–WT XT6.

image

Figure 3. The fractional STD effect, (I0 − Isat)/I0, plotted as a function of the 1H chemical shift for 0.5 s saturation time. The fractional STD effect of Xylobiose in the presence of either WT XT6 or E159Q XT6 has been normalized to the same 5β hydrogen atoms of X2–WT XT6. T1 values are grouped into short (white symbols), medium (pale orange symbols) and long (black symbols) time constant values. This format for presenting the data, analogous to 2D NMR techniques, is referred to in the text as a SASSY (saturation difference transfer spectroscopy) plot.

Download figure to PowerPoint

For both xylobiose–XT6 samples, the initial molar excess of X2 was 40-fold. The actual effective molar excess of X2 to XT6 for both samples decreased, for different reasons, by roughly half. In the WT XT6 sample, half of the X2 was hydrolyzed to xylose by the enzyme's residual catalytic activity. In the E159Q mutant, binding only occurs with the β–anomer fraction. Thus, no additional amplification factor was calculated [22] and the data were compared directly.

The strongest STD effect (Fig. 3, 100%) is seen for the 5βeq and 5βax hydrogen atoms of X2 in the presence of WT XT6 (Fig. 2c and Fig. S4). Geminal couplings have the potential to attenuate the STD effect through strong intra-dipolar relaxation [38]. In addition, these nuclei also have short T1 values (0.6 s; Table S2), which efficiently counteract the saturation by quickly returning those nuclei to their thermal equilibrium conditions. As each of these effects only act to under-estimate the actual STD factor [37], the exceptionally strong STD signal for these two hydrogen atoms therefore reflects significant hydrophobic interactions between the substrate and the WT XT6 xylanase.

The next largest STD effects on X2 occur in the mid-range of the SASSY plot (STD effect approximately 50%). These medium-strength STDs occur at chemical shifts (3.74 and 3.53 p.p.m.) where the α- and β–anomer peaks overlap in the X2–WT XT6 STD spectrum. These unresolved peaks are not analyzed in further detail; the data arising from only the β–anomer of X2, both for wild-type and mutant XT6, are discussed below.

The strongest STD effects for X2–E159Q XT6 are at the lower end of the mid-range region. These include the 5βeq as well as the 1β and 1′ hydrogen atoms. In contrast to wild-type, the STD factor for the E159Q 5βax hydrogen atom is smaller than that for H–5βeq. Altogether, the spread of the STD effects across xylobiose with the mutant xylanase is smaller than that of the wild-type enzyme.

X2–XT6 STD data suffer from a T1 bias at 5 s, i.e. for H atoms with short T1 values, the extent of saturation may be reduced. As an additional pitfall, it is necessary to account for the inherently lower signal-to-noise ratio of difference spectra compared to their parent spectra. An intensity bias may be misconstrued as stronger STD effects; the peaks with the greatest signal intensity (more hydrogen atoms and fewer splittings) are the first to emerge above the noise. Therefore, special attention was given to acquiring difference spectra with an adequate signal-to-noise ratio across all peaks. Examples of T1 bias and intensity bias for X2–XT6 are described in Docs S2 and S3. STD data misinterpretation is easy if care is not taken. SASSY plots are useful for providing a compact and comprehensive visual representation to compare epitope maps of a single substrate bound to different proteins. The SASSY plots should be based on data acquired with a good signal-to-noise ratio and a relatively short saturation time, where the fractional STD effect is calculated for each peak and T1 values are color-coded.

The literature cites another solution for acquiring STD data unbiased by T1's while compensating for the lower signal-to-noise ratio at short saturation times by calculating the slope of the STD build-up curve at zero saturation time, STD(fit) = STDmax × ksat [39, 40]. On a peak-by-peak basis, the rate constant of the STD build-up curve (ksat) is calculated by non-linear least-squares data fitting [41]. STDmax represents the maximum STD value reached per peak at long saturation times. The product of these two values matched well with the fractional STDs at short saturation times for two systems. In the X2–XT6 systems studied here, the difference in the STD effect between 0.5 s saturation and the fit was < 10% on average (Table S2) [37].

Xylotriose

The 1H spectra of xylotriose (X3) (Fig. 4A) give rise to four sets of signals labeled as α + β for the first ring, prime (′) for the middle ring and double-prime (″) for the ring at the non-reducing end. Upon adding the wild-type enzyme to the X3 solution, complete hydrolysis takes place, consistent with its catalytic activity. Therefore, no NMR data may be acquired to monitor binding of X3 to the WT XT6.

image

Figure 4. Assignment of the xylotriose-binding subsites in E159Q XT6 based on differential broadening of the ligand 1H NMR signals in the presence of the protein. (A) Chemical shift assignment of xylotriose. α or β specifies the anomeric conformation adopted by the reducing end of the disaccharide. Hydrogen atoms of the second ring are indicated by primes (′) and those of the third ring are indicated by double primes (″). Both the second and third rings have only the β–anomeric configuration. (B) The off-resonance STD NMR spectrum of xylotriose. The asterisks indicate where substantial broadening may be seen. (C) STD NMR difference spectrum of xylotriose (5.0 s saturation time), showing the same β–anomeric preference for the mutant enzyme as observed with X2–E159Q XT6. The double dagger is used to indicate where the α–anomer peaks should appear. (D) Substrate hydrogen atom numbering and subsite binding assignment, drawn using UCSF Chimera [48] from the PDB file for 1R87. Hydroxyl hydrogen atoms are not shown as a reminder that they undergo chemical exchange with the deuterated solvent and therefore do not appear in 1H NMR spectra.

Download figure to PowerPoint

When E159Q XT6 is added to a solution of X3, non-uniform broadening of the substrate signals occurs (Fig. 4B). The resonance peak of H–1α remains narrow, and its STD effect is negligible (Fig. 4C). The disappearance of the buffer peak in the STD NMR spectrum serves as an internal control for the saturation transfer experiment. The same β–anomer binding specificity detected in the mutant for X2 is also observed with X3.

The most substantial broadening occurs at the 1β and 5β hydrogen atoms in the first ring of X3 (Fig. 4B, red asterisks). Peaks assigned to the center ring have intermediate broadening, and those to the third have the least. In this case, the line broadening is caused by chemical exchange between the two chemical environments, i.e. the free and bound states of the substrate. The extent of the broadening depends only on the chemical shift difference between the free and bound substrate resonances versus the exchange rate (k). On the assumption that k across the relatively short X3 molecule is uniform, the differential broadening must be due to shielding variations within the bound substrate. We rationalize the greatest broadening seen for the H–1β and H–5β peaks by examining the 3D structure of XT6 xylanase (PDB code 1R87): the xylose ring bound at the +1 subsite is parallel to the aromatic ring of Y203. Thus significant shielding is expected to be induced in these two hydrogens by the aromatic ring (Fig. 5). Each of the xylose rings experiences unique local chemical environments defined by the specific residues of their respective binding subsites, resulting in differential shielding. Thus, the observed 1H NMR spectral broadening patterns may be used for subsite assignment.

image

Figure 5. When the two PDB coordinate files 1R85 (apo XT6) and 1R87 (xylopentaose complex of XT6) are superimposed, the side chains overlap. An interesting exception occurs for two amino acids that appear in two distinct positions in the apo protein (K62/K62′ and Q102/Q102′), but only appear in the non-primed position in the complex. When the xylopentaose is bound, the ε–amino group of K62 points towards the middle of the xylose ring at the −2 subsite. The largest STD NMR effects are seen with the WT XT6 and the H–5 hydrogen atoms at the −1 binding subsite, which fit into a pocket created by two of three highly conserved Trp residues comprising the ‘Trp cage’. The +1 subsite is dominated by an interaction with Y203. The aromatic ring is oriented parallel to the xylose ring (xylose ring not shown). A cation–π interaction with K210 appears to assist in alignment of the ring.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Xylose: neither a substrate nor an inhibitor

No STD NMR effect of xylose with either wild-type or E159Q XT6 was observed. Thus, the residence time of xylose in the binding site is either too long or too short (including the possibility of non-binding) for these experiments to be appropriate. Varying the relative concentrations of xylose versus xylobiose in solution with WT XT6 (as a result of residual hydrolytic activity) did not affect the resulting STD of xylobiose. This indicates that xylose is not a competitive binder against xylobiose, ruling out the possibility that the residence time is too long. This is consistent with isothermal titration calorimetry (ITC) experiments, in which no binding between E159Q XT6 and xylose was detected [19]. The two methods combined indicate that xylose does not bind to xylanase under the conditions studied. Interestingly, X–ray crystallography of xylose with xylanase gave varying results. Family 10 xylanase from Streptomyces olivaceoviridis E–86 does not show xylose occupying the catalytic domain [42], but the native xylanase from Penicillium simplicissimum shows three molecules of xylose occupying subsites −3, −2 and positioned between subsites +1 and +2 [43]. The differences in xylose binding may arise as a result of intrinsic differences between the xylanase enzymes [44] or different sample conditions.

Xylobiose: enzyme subsite screening using a single substrate

In this study, xylobiose was used to investigate whether a non-active mutant of XT6 preserves wild-type binding. If substrate binding and subsequent catalysis are considered as independent events, interfering with catalysis may be accomplished without necessarily perturbing binding – either through substrate or protein modifications. Thus, a non-catalytic mutant of XT6 (E159Q) was engineered to systematically study the thermodynamics of substrate binding [19]. Here we show that 1H STD NMR is a sensitive method for comparing epitope maps of a single substrate binding to two different proteins, and even for probing fine differences in substrate-binding patterns (conformational energy landscapes). In other words, rather than screening a library of small molecules against a single protein, a single substrate is used to compare binding between multiple proteins.

We note that T1 bias is one of the major drawbacks of the STD technique in general. In this study, the peak-by-peak T1 values for the free xylobiose are the same in the presence of either tested protein. This single-substrate comparison avoids T1 complications for fixed experimental conditions (concentration, temperature and magnetic field). This method may be generalized to other systems as a way of avoiding T1 bias when making comparisons of the same hydrogen atom in different complexes while screening multiple proteins and/or binding sites using a single substrate [43].

X2: subsite binding assignments

Solution NMR captures a dynamic ensemble of binding occurrences. Free and bound substrate molecules exchange under thermodynamic equilibrium conditions. A single 1H resonance peak at the weighted mean chemical shift between the free and bound populations is observed for each hydrogen atom of the substrate (fast exchange conditions). In this study, the relatively low (20-fold) effective molar excess of substrate in the X2–XT6 NMR samples makes this system sensitive to the chemical shift of the bound xylobiose (as opposed to a higher molar excess, where the bound substrate's contribution to the detected mean chemical shift becomes ‘diluted’). The chemical shifts and line widths in the STD NMR spectra of X2 with wild-type or E159Q XT6 are the same. Thus, the chemical environments of the bound substrates must also be the same. This means that xylobiose binds across the same subsites for the two xylanases.

X–ray crystallography of X2–xylanase crystals shows binding across subsites −2 and −1, corroborated by ITC studies [14, 19, 42, 43, 45, 46]. The −2/−1 subsites together have been dubbed the substrate recognition area, as opposed to the +1/+2 subsites, which are referred to as the product release area [43]. In the current study, the substantial line broadening observed for the xylose ring at the reducing end of X3 (complexed with E159Q) is consistent with shielding by the proximate aromatic ring of Y203 in the +1 subsite (PDB code 1R87). The same shielding phenomenon occurs on binding of X2 at the +1 subsite, as its dissociation constant is similar to that of X3 (Kd values of 38 and 9.4 μm, respectively) at 30 °C [19]. The line widths for X2 are narrow. Therefore, the NMR spectra of X2–XT6 are most consistent with binding across the −2 and −1 subsites. Xylobiose is therefore considered to act as a substrate mimic as opposed to a catalytic product.

The primary binding being probed in this study occurs across the −2/−1 subsites. However, some binding across the −1/+1 subsites occurs as residual catalytic activity is observed. As a generalization, even under conditions where multiple subsites are occupied, the resulting epitope map, an averaged representation, is still meaningful when probing for differences detected on the same substrate.

X2: binding of wild-type versus E159Q XT6

The results of our study indicate that E159Q binds only the β–anomer of X2 and X3. Therefore, the previously determined Kd values are under-estimated [19]. In principle, 1H NMR spectroscopy may be used to follow binding of the two anomers as a function of concentration in order to determine and compare their Kd values [47]. However, the NMR clearly shows that WT XT6 does not have an anomeric preference for xylobiose, with each anomer binding under conditions of fast exchange.

The complementarity of the three techniques, X–ray crystallography, ITC and NMR spectroscopy, is exploited here. Use of atomic-level structural information, hydrogen bonding schemes, global binding energetics and ring stacking contributions, combined with the atom-by-atom binding information, provides a comprehensive combination to obtain mechanistic insights. While exercising caution based on the limitations of STD NMR with regard to predicting proximities (T1 biases, geminal couplings, non-uniform motions, differences in neighboring 1H densities), the comparison of strong, medium and weak STD effects is informative regarding relative substrate–protein hydrophobic interaction strengths for individual atoms. Using SASSY plots, the relative ‘weights’ of the hydrophobic interactions are readily visualized.

The non-redundancy of data collected using the various biophysical techniques is explored by comparing the interactions predicted based on the crystal structure with those measured by solution STD NMR. Using UCSF Chimera [48] and the PDB coordinates from the WT XT6–xylopentaose complex (PDB code 1R87), van der Waals contacts (≤ 3.1 Å) between each of the xylose ring hydrogen atoms and the protein residues of the five occupied subsites were determined (Table S3). In this crystal structure, subsite −1 clearly has the largest number of reported hydrophobic contacts. The most abundant and closest interactions are with H–1β, followed by H–5βeq. In comparison, the STD NMR data show that the most significant effects occur equally at H–5βeq and H–5βax. H–1β only displays a weak STD effect. In the mutant, the STD effects are much weaker than in the wild-type, with the strongest being in the medium range (Fig. 3). H–1β, H–5βeq and H–1′ all give equivalent medium-strength STD effects. In the crystal structure (PDB code 1R87), no van der Waals contacts for the 1′ hydrogen atom were listed (Table S3).

At subsite −1, the abundance of hydrogen bonding [19] and hydrophobic interactions with the protein promote the correct substrate alignment and residence time essential for catalysis. The only point of agreement between the 1R87 crystal structure and the STD NMR of both the wild-type and mutant X2–XT6 solutions is the tight interaction with H–5βeq. This hydrogen atom fits neatly into a pocket comprising W316 and W324 (Fig. 5; 1R87 numbering), which is part of a highly conserved tryptophan cage [42, 46]. Three contacts under 3 Å were identified between the 5β equatorial hydrogen at subsite −1 and XT6 residue W324. In comparison, no contacts of < 3 Å were found with W316. The homology sequence of family 10 xylanases reported previously [14] overlooked these essential and conserved tryptophans, W316 and W324, for XT6 (Gsx in their nomenclature, corresponding to TmxB W794 and W802 in their Fig. 3); hence an afasta file with the proper alignment is provided in Doc. S4.

The significant non-covalent interactions between the 5β hydrogen atoms of the xylose ring at the −1 subsite and the tryptophan pocket have gone unrecognized. The results draw attention to the importance of the tryptophan cage as a ‘hydrophobic lock’ to bring the substrate into productive alignment for catalysis. The crystal structural information provided by the X–ray data cannot measure the relative strengths of the hydrophobic interactions crucial for proper positioning of the bound substrate prior to hydrolysis; however, STD NMR provides individual atom contributions to the substrate–protein interactions under conditions of thermodynamic equilibrium. The STD NMR epitope mapping presented here is thus complementary and non-redundant to the structural studies. Viewed together, a more complete picture emerges of the key interactions at the −1 subsite for capturing and aligning that particular xylose ring.

When comparing the STD effects between wild-type and mutant, it is apparent that the interactions are not the same. Based on the observable differences, we hypothesize on the mechanistic effects of the mutation. To start with, K62 and Gln102, both proximal to the −2 subsite in the wild-type enzyme crystal structure, appear in two alternative positions in XT6 (PDB code 1R85). In the presence of bound xylopentaose (PDB code 1R87), only a single preferred position is seen (Fig. 5). Taken together with the information provided by the SASSY plot, it is possible to rationalize the anomeric selectivity exhibited by the mutant (E159Q) due to the charge change resulting from the Glu to Gln mutation. In the bound form of WT XT6, the positive charge on the Lys side chain points towards the center of the xylose ring at the −2 subsite. Perhaps these two positions reflect a functional modality analogous to a ratchet motion that can align and swing the ring into position. In this ‘locked’ position in bound xylopentaose–XT6, the K62 NH3+ is 8–9 Å away from the side chain of Glu195, the catalytic acid/base residue, and hence well within the range for electrostatic charge–charge interactions.

The natural molecular thermal motions of the protein complex in aqueous solution at room temperature under thermodynamic equilibrium demand from us not a static picture but a moving image. Taking the X2 interaction with wild-type XT6 as a reference for rigid binding, where the range of motions of the substrate within the binding subsites is minimal, STD NMR shows that the Trp cage at subsite −1 provides a particularly strong anchor. When the K62–E195 electrostatic interaction is lost via mutation, the critical protein residues responsible for xylose binding are allowed a greater freedom of motion, which increases the looseness of the bound substrate. This increased motion maintains the same mean chemical environment, but attenuates dipolar interactions underlying the saturation transfer measured by the STD. Although we cannot distinguish between increased degrees of freedom at one point that influence the whole or more complex global changes due to propagation and magnification of subtle effects, a measurably different substrate–protein interaction landscape is induced. STD NMR indicates that the H–5βax interaction with the Trp cage is weakened (loss of strong interaction energy), while subsite −1 interactions with the 1β hydrogen atom are strengthened (H–1β becomes more important to the total binding energy). This results in a xylanase with increased sensitivity to the H–1β energy contribution, giving rise to the observed anomeric preference. This prediction may be tested (and refined) by comparison with a high-resolution structure of substrate-free and complexed (xylopentaose) E159Q XT6.

Xylotriose: subsite assignment based on shielding

Given the fourfold higher Kd of X3 versus X2 [19] and the approximately 20-fold substrate excess, the relative bound population is higher for X3 than it was for X2. Line-broadening effects are therefore expected to be more pronounced. Broadening of the xylotriose peaks was also observed at a slightly lower magnetic field (500 MHz). The peaks narrowed in a predictable manner on increasing the molar excess of X3 [36]. Differential broadening across the 1H NMR spectrum is attributed to non-uniform changes in shielding of the bound X3. Taking these results together, non-specific binding may be ruled out. Based on identifying key chemical environments using UCSF Chimera [48] and the high-resolution crystal structure of the xylopentaose–XT6 complex (PDB code 1R87), the extent of broadening on a per atom basis is consistent with binding of X3 to subsites −2, −1 and +1 in solution. In particular, strong shielding of the sugar ring at the +1 subsite is expected due its face-to-face alignment with the aromatic Tyr ring of Y203; the alignment of this ring may be strengthened by a cation–π interaction with K210 (Fig. 5). Extensive broadening of the ring at the reducing end of the sugar is consistent with binding at the +1 subsite, in contrast to the disaccharide, for which the data support primary binding across the −2/−1 subsites.

Interestingly, the X–ray structure of a native Penicillium simplicissimum xylanase, for which xylotriose diffused into the crystal during a short soaking time, shows substrate binding across subsites −3, −2 and −1 [43]. Binding at these subsites, in contrast to what is deduced in the XT6 system by ITC and NMR, may reflect different biological systems or sample environments: crystalline versus thermodynamic equilibrium. The paucity of substrate–protein interactions at the −3 subsite results in narrow NMR lines, similar to the line widths seen for X2 binding across subsites −2 and −1.

Engineering hydrophobic interactions

Enzymatic bleaching of kraft pulp by the paper industry using wild-type or engineered xylanases calls for stability and catalysis under conditions of high temperature and alkaline pH [49]. The importance of aromatic clustering (hydrophobic interactions) for thermal stability has already been indicated for a xylanase hyperthermophile [14]. Highlighting important specific hydrophobic interactions, such as the conserved tryptophan cage, contributes to the process of rational design. The pH values for industrial use [49] are below the pKa for xylose, approximately 12 [50]. The amino acid that participate as donors and acceptors in the substrate hydrogen bonding network of XT6 [19] are mostly unaffected over the industrial pH values of interest (based on model pKa values, the actual pKa's due to local environments are unknown). Notable exceptions are H95 and H236, proximal to the catalytic glutamic acids E195 and E265 and are thus candidates for catalytically important roles in modulating local pKa values and influencing hydrolytic activity. Mutational studies and solution NMR may be useful for determining pKa values of specific residues. On the other hand, the hydrophobic interactions highlighted in this study are pH-independent. Thus, we suggest that addition of a Trp residue proximal to the xylose ring in subsites −2 and −3 may enhance binding at higher temperatures, irrespective of pH.

Experimental procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Protein preparation

The wild-type and E159Q XT6 enzymes were over-expressed in Escherichia coli and purified as previously described [19, 51]. Final protein solutions in 100 mm phosphate buffer were concentrated by ultra-filtration using Centricon Ultracel YM–10 membrane filters (Merck/Millipore, Billerica, MA, USA). Three rounds of filtration (final volume 0.5 mL, approximately 20 min) and resuspension (2 mL 100 mm phosphate buffer in D2O; Cambridge Isotope Laboratories Inc., Andover, MA, USA) were performed to reduce the residual 1H concentration of the aqueous solution to below 1% – sufficient to remove the need for any water suppression methods in the 1H NMR measurements. The final protein concentrations of the stock solutions were determined to be 9.6 mg·mL−1 for wild-type XT6 and 18.3 mg·mL−1 for mutant XT6 using an extinction coefficient of 78 090 cm−1·m−1 and measuring the absorbance at 280 nm using a NanoDrop ND–1000 spectrophotometer (NanoDrop, Wilmington, DE, USA).

Solution NMR sample preparations

Stock solutions of the powdered substrates, xylose (X1), xylobiose (X2) or xylotriose (X3), dissolved in D2O were prepared to final concentrations of 33, 41 and 41 mm, respectively. From the substrate stock solution, 30 μL was mixed with approximately 400 μL D2O (depending on the molecular mass of the substrate), to give a final substrate concentration of 2 mm in a standard high-quality 5 mm outer-diameter NMR tube, and 1H NMR spectra were acquired. Next, protein was added to the same NMR tube as the substrate to achieve a final concentration of 0.05 mm XT6 xylanase for further spectroscopic measurements. These initial solution concentrations contained a 40-fold molar excess of substrate over protein.

NMR spectroscopy and data analysis

High-resolution solution NMR experiments were performed on a Bruker AV 500 or Bruker AV–III 600 spectrometer using Topspin 2.1 (Bruker Corporation, Billerica, MA, USA). Bruker direct detection (BBO) probeheads operating at 1H frequencies of 500.13 and 600.55 MHz, for the two spectrometers respectively, were used with automatic tuning and matching, 2H–lock and z–gradients.

Chemical shifts were calibrated relative to the methyl peak of sodium 3–(trimethylsilyl)propane-1–sulfonate (DSS) (δ = 0.017 p.p.m.) [52]. A nominal temperature of 301 K placed the resonance peak from the residual water 1H signal between the α- and β–anomeric protons, thus avoiding any overlap with peaks of interest. NMR spectra were acquired without water suppression.

For STD NMR [24], free-induction decays were collected in pairs alternating after every acquisition (single-shot) between on (1.5 p.p.m.) and off (−33.3 p.p.m.) resonance in pseudo-2D experimental mode. All STD experiments were performed with a constant time of 5 s before the 90° read pulse (i.e. the first delay plus the saturation time totaled 5 s). Selective saturation of the protein was achieved using a train of Gaussian-shaped pulses, each of 50 ms length. The total saturation time, adjusted for the number of pulses in the train, varied from 0.5 to 5 s. The intensity of the Gaussian pulse corresponded to a B1 field of 81.3 Hz (the power level corresponding to a 3 ms 90° square pulse). Identical spectroscopic parameters were used for both X2–XT6 samples (wild-type and mutant), which, although not mandatory, simplifies direct comparisons of the binding epitope between the two systems.

A spin-lock pulse serving as a T filter, with a total length of 50 ms and a B1 field of 71.4 Hz (corresponding to a 35 μs 90° pulse), was used to remove the broad protein signals, resulting in flat baselines. Samples were run both with and without the spin-lock pulse, confirming that the filter did not alter the reproducibility of the degree of saturation transfer on the resulting difference spectrum, but achieved a cleaner-looking result.

Typically, a 16 p.p.m. spectral width with 64 k real points was used. Signal averaging for the STD spectra shown, with excellent signal-to-noise ratios, included four dummy shots followed by 1600 transients for the X2–XT6 samples (10.5 h total acquisition time) and 2592 transients for the broad lines of the larger trisaccharide of the X3–XT6 sample (17 h).

To process the data, the two serial free-induction decays in the single pseudo-2D file were split into two separate 1D files. The difference between the on and off resonance data was calculated in the time domain to obtain a difference free-induction decay. The data shown were processed with 0.5 Hz line broadening and exponential multiplication of the zero-filled free-induction decay. A subtraction artifact in the difference spectrum (a dispersion-phased residual water signal) was removed from the time domain using a Gaussian function of width 0.1 p.p.m. [53].

Transferred 1H saturation of the free substrate as a function of concentration was checked for various substrate–protein ratios (10–200-fold substrate molar excess). In theory, the best optimization method is to calculate the STD amplification factor [22], which grows as the substrate excess increases until reaching a maximum, after which increasing substrate excess no longer affects the degree of 1H signal saturation. In practice, the final 1 : 40 ratio used in these studies was arbitrary.

The longitudinal relaxation time constant (T1) of the free (excess) substrate was measured in the presence of protein using the standard inversion–recovery sequence with a relaxation delay of 10 s. Typically, a 25 p.p.m. spectral width and 64 k real points were acquired with four dummy shots and 32 transients.

Spectral assignments of xylose, xylobiose and xylotriose

1D and 2D spectra used for the assignments are shown in Figs S1–S3. NUMMRIT [54] within SpinWorks (K. Marat, University of Manitoba, Department of Chemistry, Winnipeg, Canada) was used to simulate the chemical shifts and higher-order J–couplings of H–3α, H–4α and H–5α of xylose (data not shown).

Xylose

1H NMR (500.13 MHZ, D2O, δDSS = 0 p.p.m.): 5.18 (d, J = 3.7 Hz, 1α), 4.56 (d, J = 7.9 Hz, 1β), 3.91 (dd, J = 5.6, −11.6 Hz, 5βeq), 3.68 (dd, J = 5.6, −11.5 Hz, 5αeq), 3.67 (dd, J = 10.7, −11.5 Hz, 5αax), 3.67 (ddd, J = 5.6, 8, 10.7 Hz, 4α), 3.64 (dd, J = 8, 9.5 Hz, 3α), 3.62 (ddd, J = 5.6, 9.4, 10.7 Hz, 4β), 3.51 (dd, J = 3.7, 9.5 Hz, 2α), 3.41 (t, J = 9.4 Hz, 3β), 3.31 (dd, J = 10.7, −11.6 Hz, 5βax), 3.21 (dd, J = 7.9, 9.4 Hz, 2β).

Xylobiose

1H NMR (500.13 MHZ, D2O, δDSS = 0 p.p.m.): 5.18 (d, J = 3.7 Hz, 1α), 4.58 (d, J = 7.9 Hz, 1β), 4.45 (d, J = 7.9 Hz, 1′), 4.05 (dd, J = 5.3, −11.7 Hz, 5βeq), 3.96 (dd, J = 5.5, −11.6 Hz, 5′eq), 3.80 (dd, J = 9, 14 Hz, 4α), 3.78 (ddd, J = 5.5, 9.2, 10.6 Hz, 4β), 3.74 (d, J = 5.5 Hz, 5eqα), 3.74 (d, J = −10.6 Hz, 5axα), 3.74 (dd, J = 9, 9.3 Hz, 3α), 3.62 (ddd, J = 5.5, 9.2, 10.6 Hz, 4′), 3.54 (t, J = 9.2 Hz, 3β), 3.54 (dd, J = 3.6, 9.3 Hz, 2α), 3.42 (dd, J = 9.2, 9.3 Hz, 3′), 3.37 (dd, J = 10.6, −11.9 Hz, 5βax), 3.30 (dd, J = 10.6, −11.6 Hz, 5′ax), 3.26 (dd, J = 7.9, 9.3 Hz, 2′), 3.25 (dd, J = 7.9, 9.2 Hz, 2β).

Xylotriose

The chemical similarity of the three rings led to overlapping peaks in the 1H spectrum. These are well separated through a combination of 1D selective TOCSY and NOE experiments and 2D J–resolved spectroscopy (Fig. S3). The higher-order splittings from the α–anomeric ring were not determined.

1H NMR (500.13 MHZ, D2O, δDSS = 0 p.p.m.): 5.18 (d, J = 3.6 Hz, 1α), 4.58 (d, J = 7.9 Hz, 1β), 4.47 (d, J = 7.8 Hz, 1′), 4.45 (d, J = 7.9 Hz, 1″), 4.10 (dd, J = 5.2, −11.6 Hz, 5′eq), 4.05 (dd, J = 5.4, −11.6 Hz, 5βeq), 3.96 (dd, J = 5.5, −11.6 Hz, 5″eq), 3.81 (m, 4α), 3.78 (ddd, J = 5.2, 9.3 Hz, 10.5 Hz, 4′), 3.77 (ddd, J = 5.4, 9.3 Hz, 10.5 Hz, 4β), 3.75 (m, 5α), 3.75 (dd, J = 3.6, 9 Hz, 3α), 3.62 (ddd, J = 5.5, 9.3 Hz, 10.5 Hz, 4″), 3.55 (t, J = 9.3 Hz, 3′), 3.54 (t, J = 9.3 Hz, 3β), 3.54 (dd, J = 3.6, 9.5 Hz, 2α), 3.42 (t, J = 9.3 Hz, 3″), 3.37 (dd, J = 10.5, −11.6 Hz, 5′ax), 3.37 (dd, J = 10.5, −11.6 Hz, 5βax), 3.30 (dd, J = 10.5, −11.6 Hz, 5″ax), 3.28 (dd, J = 7.4, 9.3 Hz, 2′), 3.25 (dd, J = 7.4, 9.3 Hz, 2″), 3.24 (dd, J = 7.4, 9.3 Hz, 2β).

Molecular graphics and computing

Images from PDB files were created using UCSF Chimera [48]. The corrected sequence alignment and afasta file were generated in UCSF Chimera [48] using the Multalign Viewer. Substrate–protein hydrophobic interactions were determined from the PDB coordinate file using the surface/binding analysis find clashes/contacts tool in UCSF Chimera to highlight van der Waals contacts. Images were also created using PyMOL version 1.5.0.4 (Schrödinger, LLC, https://www.shrodinger.com).

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank the Schulich Faculty of Chemistry NMR Facility, Dan Goldman and Orly Tabachnikov (Faculty of Biotechnology & Food Engineering, Technion) for sample preparations, and Vered Solomon (Institute of Chemistry, The Hebrew University of Jerusalem) for assistance with structural data analysis. This work was partially supported by Israel Science Foundation grants 500/10 and 152/11, the I–CORE Program of the Planning and Budgeting Committee, the Ministry of Environmental Protection, and the Grand Technion Energy Program, and comprises part of the Leona M. and Harry B. Helmsley Charitable Trust reports on alternative energy series of the Technion – Israel Institute of Technology and the Weizmann Institute of Science. Y.S. holds the Erwin and Rosl Pollak Chair in Biotechnology at the Technion – Israel Institute of Technology.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Results
  5. Discussion
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
febs12437-sup-0001-FigS1-S4_DocS1-S4_TableS1-S3.zipapplication/ZIP693K

Fig. S1. Spectral assignment of xylose.

Fig. S2. Spectral assignment of xylobiose.

Fig. S3. Spectral assignment of xylotriose.

Fig. S4. Substrate STD NMR results summarized.

Table S1. Residence times for STD NMR.

Table S2. STD epitopes (fit) and T1 tables.

Table S3. X–ray predictions of hydrophobic interactions.

Doc. S1. Residence times for STD NMR.

Doc. S2. Intensity bias on STD NMR data.

Doc. S3. T1 bias on STD NMR data.

Doc. S4. afasta homology file.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.