The binding to only the glycone subsites is nonproductive for catalysis, and yet this has also been observed for other family 11 xylanases in com-plex with b-d-xylotetraose [Wakarchuk WW
Trang 1to family 11 xylanases using electrospray ionization
Fourier transform ion cyclotron resonance mass
spectrometry and X-ray crystallography
Janne Ja¨nis1, Johanna Hakanpa¨a¨1, Nina Hakulinen1, Farid M Ibatullin2, Antuan Hoxha3,
Peter J Derrick3, Juha Rouvinen1and Pirjo Vainiotalo1
1 Department of Chemistry, University of Joensuu, Finland
2 Biophysics Division, Petersburg Nuclear Physics Institute, Gatchina, Russia
3 Department of Chemistry, Mass Spectrometry Institute, University of Warwick, Coventry, UK
Glycoside hydrolases [1] are ubiquitous enzymes
involved in biochemical degradation of cellulose and
hemicellulose, the main constituents of plant cell walls
They cleave the glycosidic linkages between pyranose
or furanose rings of disaccharides, oligosaccharides and polysaccharides Glycoside hydrolases can be clas-sified on the basis of their substrate specificity, mech-anism of action, or amino-acid sequence [2–5] To
Keywords
Fourier transform ion cyclotron resonance
(FT-ICR); noncovalent binding;
thioxylo-oligosaccharides; X-ray crystallography;
xylanases
Correspondence
P Vainiotalo, Department of Chemistry,
University of Joensuu, PO Box 111,
FI-80101 Joensuu, Finland
Fax: +358 13 2513360
Tel: +358 13 2513362
E-mail: pirjo.vainiotalo@joensuu.fi
(Received 4 January 2005, revised 1 March
2005, accepted 10 March 2005)
doi:10.1111/j.1742-4658.2005.04659.x
Noncovalent binding of thioxylo-oligosaccharide inhibitors, methyl 4-thio-a-xylobioside (S-Xyl2-Me), methyl 4,4II-dithio-a-xylotrioside (S-Xyl3-Me), methyl 4,4II,4III-trithio-a-xylotetroside (S-Xyl4-Me), and methyl 4,4II,
4III,4IV-tetrathio-a-xylopentoside (S-Xyl5-Me), to three family 11 endo-1,4-b-xylanases from Trichoderma reesei (TRX I and TRX II) and Chaetomium thermophilum (CTX) was characterized using electrospray ionization Fourier transform ion cyclotron resonance (FT-ICR) MS and X-ray crys-tallography Ultra-high mass-resolving power and mass accuracy inherent
to FT-ICR allowed mass measurements for noncovalent complexes to within |DM|average of 2 p.p.m The binding constants determined by MS titration experiments were in the range 104)103 M)1, decreasing in the series of S-Xyl5-Me‡ S-Xyl4-Me > S-Xyl3-Me In contrast, S-Xyl2-Me did not bind to any xylanase at the initial concentration of 5–200 lm, indi-cating increasing affinity with increasing number of xylopyranosyl units, with a minimum requirement of three The crystal structures of CTX– inhibitor complexes gave interesting insights into the binding Surprisingly, none of the inhibitors occupied any of the aglycone subsites of the active site The binding to only the glycone subsites is nonproductive for catalysis, and yet this has also been observed for other family 11 xylanases in com-plex with b-d-xylotetraose [Wakarchuk WW, Campbell RL, Sung WL, Davoodi J & Makoto Y (1994) Protein Sci 3, 465–475, and Sabini E, Wilson KS, Danielsen S, Schu¨lein M & Davies GJ (2001) Acta Crystallogr D57, 1344–1347] Therefore, the role of the aglycone subsites remains con-troversial despite their obvious contribution to catalysis
Abbreviations
CTX, catalytic domain of Chaetomium thermophilum xylanase; ESI, electrospray ionization; FT-ICR, Fourier transform ion cyclotron
resonance; GlcNAc, N-acetyl- D -glucosamine; Hex, hexose (Man ⁄ Gal); S-Xyl2-Me, methyl 4-thio-a-xylobioside; S-Xyl3-Me, methyl 4,4 II -dithio-a-xylotrioside; S-Xyl4-Me, methyl 4,4II,4III-trithio-a-xylotetroside; S-Xyl5-Me, methyl 4,4II,4III,4IV-tetrathio-a-xylopentoside; TRX I, Trichoderma reesei xylanase I; TRX II, Trichoderma reesei xylanase II; Xyl2, b- D -xylobiose; Xyl3, b- D -xylotriose; Xyl4, b- D -xylotetraose; Xyl5,
b- D -xylopentaose; Xyl6, b- D -xylohexaose.
Trang 2date, the sequence-based classification of glycoside
hydrolases comprises more than 90 families, further
categorized into 14 clans displaying the same structural
folds and catalytic machinery [5,6]
Xylan is the most abundant hemicellulose
compo-nent in plant cell walls, mainly constituted of anhydro
b-1,4-d-xylopyranose backbone Natural xylan usually
contains various substituents, such as
4-O-methyl-a-1,2-d-glucuronic acid, 4-O-methyl-a-1,2-d-glucuronic acid,
a-1,3-d-arabinofuranosyl, and 2-O⁄ 3-O-acetyl, depending on
the botanical origin Xylan accounts for 7–10% dry
weight of softwoods, 15–30% of hardwoods and up to
30% of annual graminaceous plants [7,8]
Endo-1,4-b-xylanases (EC 3.2.1.8) are O-glycoside
hydrolases that catalyze a random hydrolysis of
inter-nal b-1,4-glycosidic linkages of d-xylan by a
double-displacement mechanism, with a net retention of the
anomeric configuration [8,9] The reaction proceeds
through a covalent intermediate with oxocarbenium
ion-like transition states, utilizing two conserved
cata-lytic glutamate residues, a nucleophile and an acid⁄
base catalyst [10,11] Xylanases have been generally
classified in the glycoside hydrolase families 10 and 11,
but recently xylanases associated with the families 5, 7,
8 and 43 have also been reported [8,12–14] Figure 1
shows the proposed reaction scheme for a family 11
xylanase Xylanases have a number of industrially
important applications [15–21], such as roles in animal
feeding [16,17], pulp processing [18,19] and baking [20,21] In addition, their potential use in the biomass conversion to liquid fuel (i.e bioethanol) has gained considerable interest [15]
X-ray crystallography [22] has been extensively used
to dissect catalytic mechanisms for glycoside hydro-lases, particularly through the use of specific covalent
or noncovalent inhibitors [11,12,23] Elegant experi-mental approaches providing snapshots along an enzy-matic reaction co-ordinate have been presented, in which the crystal structures for each of the enzyme– substrate (Michaelis), covalent intermediate and prod-uct complexes have been determined and further kinet-ically analyzed [24,25] Both fluoro [24–32] and epoxyalkyl [33–37] glycosides have been successfully used to identify catalytic residues and gather informa-tion on the reacinforma-tion mechanisms, such as the itinerary
of the sugar ring conformations along the catalytic pathway [24,25,29,30] or the inversion of the roles of the catalytic glutamates [34] Substrate derivatives with
a fluorine atom at the 2-O-position of the xylose or glucose moiety (e.g 2-deoxy-2-fluoroglycosides) slow down the formation of the intermediates by inductively destabilizing the oxocarbenium transition states and eliminating an important hydrogen bond to the 2-O-position [24] Epoxyalkyl glycosides bind to the enzymes by forming a covalent bond to the putative nucleophile [34]
Fig 1 Proposed reaction scheme for a reta-ining family 11 xylanase, with b- D -xylopenta-ose as a model substrate Putative glycone (from )3 to )1) and aglycone (+1 and +2) subsites at the enzyme active site have been numbered as described in [77] The reaction proceeds by a nucleophilic attack of the catalytic glutamate on the anomeric car-bon of the xylopyranoside ring (at the )1 subsite) to produce a covalent glycosyl– enzyme intermediate via an oxocarbenium ion-like transition state At this point, the first product (b- D -xylobiose) is released The intermediate is then hydrolyzed via a second transition state to give the second product (b- D -xylotriose) and the free enzyme The proposed conformations for the xylopyra-nose ring in the )1 subsite have been indicated.
Trang 3In recent years, many noncovalent glycosidase
inhib-itors, e.g glyco-, xylo-, manno-, and galacto-configured
aza [38,39], imino [40–42] tetrahydropyridoazole ([23]
and references therein), and hydroximolactam [43]
sugar derivatives have been introduced Imino-sugar
inhibitors, for instance, are potential transition state
mimics by virtue of the protonated nitrogen atom that
highly resembles the oxocarbenium ion-like
transition-state structure [12] However, only a handful of the
above inhibitors possess considerable aglycone (i.e the
subsites that bind the ‘aglycone’ leaving group portion
of the substrate; Fig 1) specificity However,
inter-actions of the substrate with the aglycone subsites play
an important role in transition-state stabilization along
the catalytic pathway [28] Attention has been drawn
to thio-oligosaccharides (Fig 2) as being promising
noncovalent inhibitor candidates for structural biology
studies among glycoside hydrolases [44–50] Such
oligosaccharides, in which two or more carbohydrate
residues are incorporated via S-glycosidic linkage(s),
should conserve the global geometry of the natural
substrate while being hydrolytically inert [44] In fact,
changes only occur at the glycosidic bond The length
and angle for the S-glycosidic bond are 1.83 A˚ and
97, whereas the respective values for O-glycosides are
1.41 A˚ and 117, resulting in a difference of 0.35 A˚
between the adjacent sugar rings [49]
During the past 15 years, electrospray ionization
(ESI) MS [51] has become an increasingly important
analytical technique for the study of protein structure
and function Of particular interest is the use of
ESI-MS in the studies of noncovalent interactions [52–54],
as valuable parameters such as stoichiometry and
bind-ing constants can be determined A Fourier transform
ion cyclotron resonance (FT-ICR) MS [55,56] has
extended the possibilities of MS in protein analysis
because of its inherently ultra-high mass-resolving
power and mass accuracy For instance, protein–ligand
[57], protein–carbohydrate [58,59], protein–peptide
[60], and protein–RNA [61] interactions have been analyzed by ESI FT-ICR MS, providing valuable ther-modynamic and kinetic data
The filamentous fungus Trichoderma reesei is an effi-cient xylanase producer, expressing at least four xylan-ases, of which TRX I and TRX II are the most characterized [62,63] CTX is a thermostable xylanase expressed from Chaetomium thermophilum, another filamentous fungus [64] TRX I, TRX II and CTX, associated with family 11 of glycoside hydrolases, are folded into a single domain (b-jelly roll) structure comprising two parallel b-sheets and a single a-helix TRX II and CTX have been previously studied using ESI FT-ICR [64–66] The complex structures of TRX II with covalently attached epoxyalkyl xylosides have been obtained using X-ray crystallography [34– 37] In this paper, we report the characterization of the noncovalent binding of thioxylo-oligosaccharide inhibi-tors to TRX I, TRX II and CTX using high-field ESI FT-ICR MS and X-ray crystallography
Results and Discussion
ESI FT-ICR analyses Figure 3 presents typical 9.4 T ESI FT-ICR mass spec-tra of TRX I, TRX II and CTX in 10 mm ammonium acetate buffer (pH 6.8) The resolving power of
150 000 (defined as m ⁄ DmFWHM, where m is the ion mass andDmFWHMis the peak full width at half-maxi-mum) allowed isotopic distributions, a consequence of the contributions of heavier isotopes (primarily 13C and 15N), to be well resolved (Fig 3B, inset) Each peak represents an unresolved superimposition of sev-eral isotopic compositions of the same nominal mass (actually differing by a few mDa), except the first peak, which represents the monoisotopic mass (i.e all hydrogens are 1H, all carbons are 12C, all nitrogens are 14N, etc.) However, the monoisotopic peak is often undetectable for proteins > 10 kDa, except for isotope-depleted proteins [67] Hence, the molecular masses reported hereafter refer to the masses calcula-ted from the most abundant isotopic peaks (exp.) or the sequence-derived, most abundant elemental compo-sition of a protein (theor.) The charge state for the species at a given mass-to-charge (m⁄ z) ratio can be readily assigned, as the spacing between the adjacent isotopic peaks corresponds to a reciprocal of the charge, i.e z)1for the species [M + zH]z+, which then allows the accurate mass to be unequivocally deter-mined
All proteins exhibited narrow charge state distribu-tions of mainly four charge states, from 7+ to 10+,
Fig 2 Structures of thioxylo-oligosaccharide inhibitors From top to
bottom: S-Xyl5-Me, S-Xyl4-Me, S-Xyl3-Me, and S-Xyl2-Me.
Trang 4at m⁄ z 2000–3500 We have previously shown that in
these conditions, TRX II exists in a single
conforma-tion which represents the native protein structure
[65,66] On the basis of the mass spectra presented in
Fig 3, all proteins had a variable degree of
modifica-tion The first peaks at each charge state in the mass
spectrum of TRX I (Fig 3A) represent the native
protein (19 046.920 ± 0.011 Da exp., 19 046.939 Da
theor.), and the second peaks are due to a mass
incre-ment of 162 Da, consistent with the
post-translation-ally attached hexose (Hex, +162.053 Da), the form
(TRX IHex) comprising less than 5% The first peaks
in the mass spectrum of TRX II (Fig 3B) represent
the native protein (20 824.823 ± 0.008 Da exp.,
20 824.850 Da theor.) with the N-terminal glutamine
existing in its cyclized pyrrolidonecarboxylic acid ()17.027 Da) form The second peaks correspond to a mass increment of 203 Da, consistent with the previ-ously observed N-glycosylation by a single
N-acetyl-d-glucosamine (GlcNAc, +203.076 Da; TRX IIGlcNAc) [65,66], comprising 30% of the protein content Exactly the same modifications as in TRX II were determined in CTX (Fig 3C), although the glycosyl-ated protein form (CTXGlcNAc) was the major form (21 680.114 ± 0.021 Da exp., 21 680.289 Da theor.) comprising 90% of the total protein content The observed modifications of TRX II and CTX agree with our earlier reports [64–66] The mass data are summar-ized in Table 1
Close examination of the peaks at m⁄ z 2900–3300 in the ESI FT-ICR spectrum of TRX II (Fig 3B) also revealed the existence of noncovalent protein dimers The peak at m⁄ z 3200 represents [(TRX II)2+ 13H]13+ (41 649.665 ± 0.029 Da exp., 41 649.700 Da theor.), whereas the peak at m⁄ z 2975 is a compos-ite, in which [TRX II + 7H]7+ and [(TRX II)2+ 14H]14+ are overlapping each other (Fig 4) This results in a superimposition of the two isotopic distri-butions, one with the peak spacing of 0.143 (z ¼ 7) representing the monomer, and the other with 0.071 (z¼ 14) representing the dimer Also, a heterodimer, TRX II–TRX IIGlcNAc, was observed at m⁄ z 2990 (14+) and 3220 (13+) We have previously reported noncovalent dimerization of TRX II upon heat-induced conformational change [63] Also, TRX I had minor peaks in the mass spectra representing noncova-lent dimer (Fig 3A), but CTX did not dimerize under any solution conditions, regardless of the close struc-tural homology with TRX II Although the only signi-ficant difference between TRX II and CTX was the extent of N-glycosylation, this would not explain the absence of the dimeric form of CTX, as the hetero-dimer was present in the case of TRX II
Observation of noncovalent protein–inhibitor complexes
The formation of noncovalent protein–inhibitor com-plexes was readily observed by mixing appropriate aliquots of xylanase and thio-oligosaccharide solutions before direct analysis by MS Figure 5 represents the ESI FT-ICR mass spectra of 10 lm TRX II with
50 lm methyl 4,4II,4III,4IV-tetrathio-a-xylopentoside (S-Xyl5-Me), 50 lm methyl 4,4II,4III -trithio-a-xylote-troside (S-Xyl4-Me), and 50 lm methyl 4,4II -dithio-a-xylotrioside (S-Xyl3-Me) in 10 mm ammonium acetate buffer (pH 6.8) Comparison with the spectrum
in Fig 3B reveals the presence of the noncovalent 1 : 1
Fig 3 9.4 T ESI FT-ICR mass spectra of 5 l M TRX I (A), 10 l M
TRX II (B), and 10 l M CTX (C) in 10 m M ammonium acetate buffer
(pH 6.8) Numbers (n+) denote charge states (B) The inset shows
the isotopic distribution for the charge state 9+, with white spheres
representing the theoretical abundance distribution Glycosylated
protein forms (TRX IHex, TRX IIGlcNAcand CTXGlcNAc) at each charge
state have been denoted by # (see text for details).
Trang 5protein–inhibitor complexes on the basis of the new
peaks at the expected m⁄ z values The accurate mass
data for the complexes are summarized in Table 1
However, no complexes could be detected for methyl
4-thio-a-xylobioside (S-Xyl2-Me), at the initial
concen-tration of 5–200 lm, suggesting no interaction or
the dissociation constant being a high millimolar
concentration, unreachable with our mass spectro-meter Similar results for the thio-oligosaccharide bind-ing were observed for TRX I and CTX (Table 1) The glycosylation in TRX II did not have any influence on the binding as the same ratio of protein–inhibitor com-plex⁄ free protein was obtained at each charge state for both the glycosylated and nonglycosylated protein forms The same was observed for CTX Therefore, the subsequent analyses for the calculation of binding constants were made on the basis of the major pro-tein forms only (i.e nonglycosylated propro-tein form rep-resenting TRX II and glycosylated protein form representing CTX)
In addition to equimolar complexes, 1 : 2 and 1 : 3 protein–inhibitor complexes were typically present with higher initial inhibitor concentrations (Fig 5) Com-plexes with higher stoichiometric compositions are probably due to nonspecific aggregation on the electro-spray process [59] On the basis of the crystal struc-tures, TRX I, TRX II and CTX have each only a single binding site Therefore, the subsequent molecules apparently bind to the protein surface by weak electro-static forces, e.g hydrogen bonds
Determination of binding constants
MS titration experiments [68] were performed to deter-mine binding constants for protein–inhibitor complexes
as explained in Experimental Procedures The mass spectra were first background subtracted Assuming that the observed protein ion intensities reflect the true
Table 1 The most abundant isotopic masses for TRX I, TRX II and CTX xylanases and their noncovalent thioxylo-oligosaccharide inhibitor complexes The data were obtained using 9.4 T ESI FT-ICR MS n.b., No binding (the peaks at the expected m ⁄ z for any charge states of the CTX–S-Xyl2-Me complex were not detected with [S-Xyl2-Me] initial ¼ 5–200 l M ).
a Mean ± SD measured over the charge state distributions b Calculated from the elemental composition of a protein and an inhibitor, inclu-ding observed post-translational modifications.cCalculated from DM (p.p.m.) ¼ [(M exp –M theor ) ⁄ M exp ] · 10 6
.
Fig 4 Expansion of the 9.4 T ESI FT-ICR mass spectrum of 10 l M
TRX II in 10 m M ammonium acetate (pH 6.8) at m ⁄ z 2920–3060
showing the group of peaks representing monomers (charge state
7+) and noncovalent dimers (charge state 14+) of TRX II and
its glycosylated form (TRX IIGlcNAc) From left to right:
[TRX II + 7H] 7+ ⁄ [(TRX II) 2 + 14H] 14+ (the inset shows the
superim-posed isotopic distributions with the peak spacing of Dm ⁄ z ¼ 0.143
for the monomer and Dm ⁄ z ¼ 0.071 for the dimer),
[TRX II + TRX IIGlcNAc+ 14H] 14+ (the inset shows isotopic
distribu-tion with Dm ⁄ z ¼ 0.071 for the dimer), and [TRX II GlcNAc + 7H]7+
(no inset).
Trang 6protein concentrations, one can readily calculate the
free and bound protein concentrations [68]
Conse-quently, the free ligand concentration can also be
calculated Previously it has been shown that different
charge states can represent different binding patterns,
reflecting different conformations [60] The intensities
used for these calculations were therefore summed over
the charge state distributions Moreover, in FT-ICR
the signal intensity (i.e the induced image current)
increases linearly with the charge state [55] Hence, the
data were charge-normalized by dividing the intensity
of each signal by the respective z This should
reduce any bias introduced by a possible shift of
the charge state distribution caused by ligand binding
The dimeric protein forms were disregarded in the
determination of binding constants because of their low abundance and the fact that no peaks representing dimers were actually observed at ligand concentrations higher than 20 lm The MS titration curves for TRX I, TRX II and CTX are presented in Fig 6 On the basis of the data presented in Fig 6, even at the highest ligand concentrations the proteins were still far from saturation, indicating relatively low binding con-stants The nonspecific binding was also observed in the titration experiments; the relatively important two and three ligand binding indicated that the nonspecific binding constants were of the same order of magni-tude
For simplicity, we will consider here only the case of
a protein with a single specific binding site As will be explained in more detail in the next few paragraphs, this situation corresponds well to TRX I, TRX II and CTX xylanases In such cases, the binding constant,
Ka, can be expressed as
KaỬ ơPL
in which [PL] is the concentration of the proteinỜlig-and complex, proteinỜlig-and [P] proteinỜlig-and [L] are the concentrations of the free protein and free ligand (i.e inhibitor), respect-ively [68] Expressed in terms of r, defined as the num-ber of ligands bound to one protein molecule, Eqn (1) can be written as:
rỬ naKaơL
in which na Ử 1 r ⁄ [L] plotted vs r is called the Scat-chard plot and is a straight line with a slope of ỜKa In the case of nonspecific binding (in our case, PL2 and
PL3 complexes), the number of ligands bound to one protein molecule is proportional to the concentration
of free ligand Therefore, another term has to be implemented in Eqn (2) giving:
rỬ naKaơL
1ợ KaơLợ đKnspơLỡ
a
đ3ỡ
in which Knspis a binding constant for the nonspecific proteinỜligand complexes, with negative or positive co-operativity (the coefficient a) MS measurements readily provided the values of r as follows:
rđơLinitial;ơPinitialỡ Ử
P
đIPLợ 2IPL2ợ 3IPL3ỡ P
đIPợ IPLợ IPL 2ợ IPL 3ỡ đ4ỡ
in which IP and IPLn are the intensities of a free pro-tein and different propro-teinỜligand complexes summed over the charge states and the isotopic distributions The free ligand concentration was then:
Fig 5 9.4 T ESI FT-ICR mass spectra of 10 l M TRX II with 50 l M
S-Xyl5-Me (A), 50 l M S-Xyl4-Me (B) and 50 l M S-Xyl3-Me (C) in
10 m M ammonium acetate buffer (pH 6.8) Noncovalent proteinỜ
inhibitor complexes are indicated (A) The insets show the isotopic
distributions for [TRX II + S-Xyl5-Me + 9H]9+and [TRX II +
(S-Xyl5-Me)2+ 9H] 9+ Only the 9+ charge states have been denoted for
clarity.
Trang 7½L ¼ ½Linitial r½Pinitial ð5Þ
To obtain values for Ka and Knsp, nonlinear curve
fittings based on the Levenberg–Marquardt algorithm
were performed using Microcal origin 6.1 (Origin
Laboratory Corp., Northampton, MA, USA) Briefly,
starting from the given set of parameters, the sum of
squared residuals of Eqn (3) from the experimental
data points was minimized by performing a series of
iterations Figure 7 shows the binding isotherm, i.e r
as a function of the free ligand concentration and the
fit to Eqn (3) (solid line), obtained in the case of
TRX I and S-Xyl5-Me On the basis of the Ka
obtained, one can calculate the binding free energy
(for specific binding) at a given temperature from the
general expression DGbind¼ –RTlnKa The Ka, Knsp, a
values determined and DGbind values calculated are
presented in Table 2
All of the thioxylo-oligosaccharide complexes had
specific binding constants (Ka) within the range
103)104 m)1, whereas the nonspecific binding constants
(Knsp) were 102)103m)1 (Table 2) The a values
were 1 in most cases, suggesting no significant
co-operativity in the nonspecific binding For both
TRX I and TRX II, a decreasing trend in the spe-cific binding constants was observed as follows: S-Xyl5-Me > S-Xyl4-Me > S-Xyl3-Me >>> S-Xyl2-Me For CTX, a similar trend, S-Xyl5-Me S-Xyl4-Me>S-Xyl3-Me >>> S-Xyl2-Me, was observed In fact, none
Fig 7 Binding isotherm for 5 l M TRX I with S-Xyl5-Me ([L] initial ¼ 0–100 l M ) at 293 K For details of the calculations of r and [L], see text and Eqns (1–5).
Fig 6 MS titration curves for TRX I (5 l M ), TRX II (10 l M ) and CTX (10 l M ) with S-Xyl5-Me, S-Xyl4-Me and S-Xyl3-Me (initial concentrations
of 0–100 l M ) [PL n ] is the concentration for 1:n protein–inhibitor complex with n ¼ 1–3, calculated from the 9.4 T ESI FT-ICR intensity data [L] is the free inhibitor concentration.
Trang 8of the xylanases had detectable affinity for S-Xyl2-Me,
at the initial concentration of 5–200 lm This may be
because the Ka values for the S-Xyl2 complexes were
in the range 1–100 m)1, which is undetectable with our
instrument These observations clearly show that the
ligand binding is highly influenced by the number of
xylopyranosyl units, with a minimum requirement of
three
Protein crystallography
The final model of CTX contained 191 residues (on
the basis of the ESI FT-ICR data, CTX contained
196 amino-acid residues with an N-terminal
pyrroli-donecarboxylic acid and glycosylation by a single
GlcNAc; neither the last five C-terminal amino acids
nor the carbohydrate residue were visible in the
crys-tal structures of CTX or CTX–S-Xyl5-Me complex)
in both of the two molecules (A and B) of the
asymmetric unit, 450 water molecules, three sulfate
ions and two inhibitor molecules (S-Xyl5-Me) partly
attached to the active site (Fig 8) In the
electron-density maps, three xylopyranose rings of the
inhib-itor molecules could be observed in the active site of
both molecules (Fig 9) The xylopyranose rings, all
adopting a normal 4C1 ground-state conformation,
were observed only in the )1, )2 and )3 subsites
(for the nomenclature, see [69]), missing the point of
catalysis which occurs between subsites )1 and +1
Additional densities were observed in the glycone
ends of the inhibitor chains in both molecules, but
no more xylopyranose rings could be unambiguously fitted into those electron densities In the active site
of molecule B, two of the rings, 1 and 2, were packed between two tryptophan residues (Trp19 and Trp80), with sugar ring 3 located just outside the active-site, forming hydrogen bonds only with water molecules The hydrogen bonds formed between CTX and S-Xyl5-Me are listed in Table 3 and sche-matically represented for molecule B in Fig 9C Similar results were obtained for other inhibitors (data not presented), except for S-Xyl2-Me When crystals were soaked in the solution containing
S-Xyl2-Me and further in 2-methyl-2,4-pentanediol before measurement at 120 K, no electron density caused by the inhibitor was detected However, some residual density was observed, consistent with
2-methyl-2,4-Table 2 Thermodynamic parameters for thioxylo-oligosaccharide
inhibitor binding to TRX I, TRX II, and CTX xylanases The data
were obtained in 10 m M ammonium acetate (pH 6.8) at 293 K
using 9.4 T ESI FT-ICR MS n.b., No binding (the peaks at the
expected m ⁄ z for any charge state of CTX–S-Xyl2-Me complex
were not detected within [S-Xyl2-Me] initial ¼ 5–200 l M ).
Protein Inhibitor
K a · 10)3 ( M )1)
DG bind (kJÆmol)1)
K nsp · 10)3 ( M )1)a
a TRX I S-Xyl2-Me n.b.
TRX II S-Xyl2-Me n.b.
CTX S-Xyl2-Me n.b.
a
The fitting procedure did not provide error for K nsp probably
because of several minima reached on replicate runs b Too low to
be accurately determined.
Fig 8 Cartoon representation (A) and surface representation (B) of the crystal structure of CTX with S-Xyl5-Me The observed part of the inhibitor (three xylopyranose rings) is shown at the active site Carbon atoms of the inhibitor are coloured in purple, oxygen atoms
in red, and sulfur atoms in orange The figure was created with
PYMOL [86].
Trang 9pentanediol bound to the active site On the other
hand the electron density for S-Xyl2-Me was actually
detected in subsites)1 and )2 when the measurements
were performed at room temperature This suggests
that the cryo-protectant replaces the bound inhibitor
before the cryogenic measurement, which is consistent
with the observations of low binding affinity by MS
The overall conformations of molecules A and B
were quite alike (rmsd¼ 0.85 A˚ for 1495 atoms)
However, both molecules in the asymmetric unit
contained several residues with signs of multiple
conformations Most of these residues were located
on the surface of the enzyme, and only residues TyrB74 and GluB178 were fitted into two conforma-tions in the final model, as these conformaconforma-tions are relevant to ligand binding In addition, molecule B contained a loop region (residues 162–167) that was slightly disordered In the disordered region, the main chain of the protein was intact, but clear signs
of peptide flipping and several side-chain conforma-tions could be seen, and fitting of the residues to the electron density was challenging
The main differences were the two conformations
of the catalytic glutamate, Glu178, in molecule B but not in molecule A and the disordered loop region in molecule B that was unambiguous in molecule A Glu178 has one unambiguous conformation in mole-cule A, where it points towards the other catalytic glutamate, Glu87 This conformation is also present
in molecule B together with another conformation, in which Glu178 is bent away from the inhibitor mole-cule Movement of Glu178 forces the tyrosine resi-due, Tyr74, to adopt another conformation in molecule B The single position of Tyr74 in molecule
A is an intermediate of the two conformations found
in molecule B
Fig 9 Final 2Fo-Fcelectron density map of the active site of CTX
in molecule A (top) and molecule B (middle) Contour level is 1.0 r.
Water molecules are depicted as red spheres The figure was
cre-ated with PYMOL [86] Schematic representation of the interactions
of CTX with S-Xyl5-Me inhibitor (in molecule B) is shown at the
bottom.
Table 3 Hydrogen bonds formed between CTX xylanase and S-Xyl5-Me inhibitor.
Inhibitor atom a ⁄ side chain hydroxy group Residue ⁄ water Distance (A ˚ )
a A and B refer to the corresponding molecules in the asymmetric unit, and numbers refer to the xylopyranose rings at the corres-ponding glycone subsites, )1, )2 and )3 b Side-chain methoxy group at 1-O-position.
Trang 10A few differences were observed when comparing
the CTX–S-Xyl5-Me complex with the native structure
of CTX The space group and cell dimensions in the
complex structure were the same as for the native
pro-tein, and the native structure also contained two
mole-cules in the asymmetric unit In molecule B, the
catalytic glutamate of the native protein adopted a
sin-gle conformation, the one described as bent away from
the inhibitor in the complex structure An arginine
residue, Arg105, on the surface of the native protein
was in a more extended conformation in the native
protein compared with the CTX–S-Xyl5-Me complex
in both A and B molecules This is probably due to
the packing of the molecules, as Arg105 bends away to
make space for the inhibitor molecule of the adjacent
enzyme molecule
Evaluation of inhibitor binding and implications
for catalysis
Ultra-high mass-resolving power and high mass
accu-racy inherent to FT-ICR, as demonstrated here,
allowed unequivocal identification of the different
pro-tein forms as well as noncovalent propro-tein complexes
The binding constants for thioxylo-oligosaccharides
were determined using ‘classical’ titration experiments
Such analyses are feasible using ESI-MS intensity data
to represent the thermodynamic equilibrium of free
protein and protein–ligand complexes in solution [68]
However, nonspecific protein–carbohydrate complexes,
which can be even energetically preferred in the gas
phase [70], often arise during the electrospray process
A large amount of nonspecific binding complicates the
analysis because of its indefinable manner Here, we
distinguished between these two types of binding from
the crystal structures, given that only a single binding
site exists in each xylanase, capable of occupying only
one inhibitor molecule An equimolar titration,
recently described for the determination of protein–
carbohydrate association constants [59], might have
been a better approach as it diminishes the extent of
nonspecific binding
Unfortunately, there are no other reports on the
binding of xylo-oligosaccharides or
thioxylo-oligosac-charides to family 11 xylanases for which the binding
constants had been determined However, some
com-parison can be on the basis of
xylo-oligosaccharide-binding thermodynamics reported for isolated
carbohydrate-binding domains from Clostridum
ther-mocellum Xyn10B (X6b domain, family 10 [71]) and
Pseudomonas cellulosa Xyn10C (CBM15 domain,
family 15 [72]) xylanases, which display a similar
b-jelly roll fold, characteristic of family 11 xylanases
The affinity for xylo-oligosaccharide binding deter-mined using isothermal titration calorimetry of both domains decreased in the series b-d-xylohexaose (Xyl6) > b-d-xylopentaose (Xyl5) > b-d-xylotetraose (Xyl4) > b-d-xylotriose (Xyl3), with no detectable affinity for b-d-xylobiose (Xyl2), which is consistent with our results However, absolute affinity values for X6b were 10-fold higher than the values for CBM15 and the values reported here A similar trend was seen
in the Michaelis constants (KM) for Penicillium simplic-issium xylanase from family 10 vs the length of the oligosaccharide (KM¼ 1.4, 3.1, 5.1 and 7.9 mm for Xyl6, Xyl5, Xyl4, and Xyl3, respectively), with no hydrolysis occurring in the case of Xyl2 [73]
On the basis of the results, it remains controversial why such a correlation between the number of xylo-pyranose rings and the binding constants was observed for the thio-oligosaccharides given that none of the sugar rings occupied any of the aglycone subsites The electron densities for three sugar rings were observed only in the )1, )2 and )3 subsites Yet, the same has previously been observed with catalytically incompet-ent Bacillus circulans E172C [74] and Bacillus agarad-haerens E94A [75] mutant xylanases in complex with Xyl4 In the case of B circulans xylanase, only two carbohydrate residues were unequivocally fitted into the electron densities The authors suggested that either the enzyme had a small amount of residual activity (i.e being able to hydrolyze Xyl4), which sounds questionable from the mechanistic point of view (E172C mutation) and in view of the KMand kcat values reported in the same paper, or that the enzyme requires a larger substrate for tight binding In CTX, the main contribution to the binding in the glycone subsites together with multiple hydrogen bonds (Table 3) is apparently the hydrophobic interactions of the proximal sugar ring (A2, B2) with the conserved active-site tryptophan residue (Trp19) (Fig 8C) There were no clear electron densities for the sugar rings A4⁄ B4 and A5 ⁄ B5 (some partial disordered density was, in fact, visible for the sugar ring B4 outside of the active site) However, the third sugar ring (A3⁄ B3) makes hydrogen bonds only to the adjacent water molecules, with no direct contacts to the protein (Fig 9C), the same as with B agaradhaerens E94A xylanase [75] Therefore, the major contributions to the binding in the glycone side must come from the interac-tions within the subsites)1 and )2 However, the ther-modynamic data suggest that, although the importance
of the glycone subsites in substrate binding is more evident, some affinity must remain in the aglycone sub-sites to explain the results In fact, the results could be easily interpreted in view of the proposed model for