Quantitative NMR Studies of High Molecular Weight Proteins ... · Some structural implications of the chemical shift perturbations upon ... and an energetic challenge for weak bases

Quantitative NMR Studies of High Molecular WeightProteins: Application to Domain Orientation andLigand Binding in the 723 Residue Enzyme MalateSynthase G

Vitali Tugarinov and Lewis E. Kay*

Protein Engineering NetworkCentres of Excellenceand the Departmentsof Medical GeneticsBiochemistry and ChemistryUniversity of Toronto1 King’s College CircleToronto, Ont.Canada M5S 1A8

A high-resolution multidimensional NMR study of ligand-binding toEscherichia coli malate synthase G (MSG), a 723-residue monomericenzyme (81.4 kDa), is presented. MSG catalyzes the condensation ofglyoxylate with an acetyl group of acetyl-CoA, producing malate, an inter-mediate in the citric-acid cycle. We show that despite the size of theprotein, important structural and dynamic information about themolecule can be obtained on a per-residue basis. 15N–1HN residual dipo-lar couplings and carbonyl chemical shift changes upon alignment in Pf1phage establish that there are no significant domain reorientations in themolecule upon ligand binding, in contrast to what was anticipated onthe basis of both the X-ray structure of the glyoxylate-bound form of theenzyme and structural studies of a related set of proteins. The chemicalshift changes of 1HN, 15N and 13CO nuclei upon binding of pyruvate, aglyoxylate-mimicking inhibitor, and acetyl-CoA have been mapped ontothe three-dimensional structure of the molecule. Binding constants ofpyruvate, glyoxylate, and acetyl-CoA (in the presence of pyruvate) havebeen measured, along with the kinetic parameters for glyoxylate andpyruvate binding. The on-rates of pyruvate and glyoxalate binding,,1.2 £ 106 M21 s21 and ,2.7 £ 106 M21 s21, respectively, are significantlylower than what is anticipated from a simple diffusion-controlled process.Some structural implications of the chemical shift perturbations uponbinding and the estimated ligand on-rates are discussed.

q 2003 Elsevier Science Ltd. All rights reserved

Keywords: ligand binding; malate synthase G; dipolar couplings; proteindomains; multi-dimensional NMR*Corresponding author

Introduction

Malate synthase G (MSG) from Escherichia coli, a723-residue monomeric enzyme1 (81.4 kDa), cata-lyzes the Claisen condensation of glyoxylate withan acetyl group of acetyl-CoA, as shown in Figure1(a), producing malate, an intermediate in thecitric-acid cycle. The reaction involves the inver-sion of configuration at the methyl group ofacetyl-CoA, as with other Claisen enzymes suchas citrate synthase.2,3 The enzymatic abstraction ofa proton from the a-methyl group of the acetyl

CoA thio-ester, which is the rate-determining step,has long been considered to present both a kineticand an energetic challenge for weak bases typicallyavailable in proteins3 and has led to an interest instructural studies of this enzyme. Recently thecrystal structure of MSG complexed with mag-nesium and glyoxylate was solved at 2.0 Aresolution.4 MSG is a multi-domain enzyme,comprised of four main domains as illustratedin Figure 1(b). The centrally located core of themolecule is based on a highly stable b8/a8 barrelfold (residues 117–134, 263–295, 334–550). It issupported on one side by an N-terminal a-helicalclasp (residues 3–88), which is linked to thefirst strand of the barrel by a long extended loop(residues 89–116). On the opposite side of the bar-rel is an a/b domain comprising residues 135–262and a loop (residues 296–333) extending from themolecular core. The C-terminal end of the enzyme

0022-2836/03/$ - see front matter q 2003 Elsevier Science Ltd. All rights reserved

E-mail address of the corresponding author:[email protected]

Abbreviations used: MSG, malate synthase G; CSA,chemical shift anisotropy; HNCO, 1H, 15N, 13CO 3D NMRcorrelation; TROSY, Transverse Relaxation OptimizedSpectroscopy.

doi:10.1016/S0022-2836(03)00238-9 J. Mol. Biol. (2003) 327, 1121–1133

(589–723) consists of a five-helix plug connected tothe barrel by a flexible loop-helix-turn-helix-loopmotif (residues 551–588). In addition, severalsmaller loops protrude from the globular core ofthe molecule. The active site of the enzyme islocated in a cleft at the interface between theC-terminal plug and loops at the C-terminal endsof several of the b-strands of the core barrel. Itwas suggested based on the crystallographic datathat the C-terminal plug may be mobile and thatits position relative to the rest of the enzyme maybe different in apo (open) and substrate-bound(closed) states of the molecule.4 Such mobilitywould then facilitate opening of the active sitecleft for substrate binding and product release.Earlier circular dichroism studies of substratebinding to both yeast and maize malate synthasesindicated that significant conformational changesoccur in the enzyme upon ligand binding.5,6 Pro-teolysis studies with the maize enzyme supportthe notion of a flexible linkage between core andC-terminal domains that is rigidified upon acetyl-CoA binding.6 Low-angle X-ray scattering studieson trimeric malate synthase from baker’s yeasthave shown that there is a decrease in the radiusof gyration upon substrate binding.7

Domain reorientations are likely to occur in anumber of other enzymes that are either struc-turally or mechanistically similar to MSG. Pyruvatekinase8 (PK) is a close structural analog of MSG.4 InPK the C-terminal extension is located opposite theactive site but a large insertion within the sequenceof the TIM-barrel core folds over the binding siteand plays the role of the C-terminal plug in MSG.

X-ray studies of PK with magnesium, potassiumand L-phospholactate9 establish that the proteinis a tetramer with two tetramers per asymmetricunit. The inserted domain is observed in differentpositions relative to the TIM-barrel core in thedifferent copies of the protein, with the trans-formation from the most “open” to the most“closed” copy given by a rotation of greater than208. This leads to significant differences in thedegree of closure of the active site cleft betweenmolecules.9 Citrate synthase is a TCA cycle enzymethat catalyzes a very similar reaction to MSG butwith a completely different fold.4,10 Dimeric citratesynthase undergoes a large conformational change,in which the C-terminal domain rotates from themain body of the dimer by approximately 188 toopen the active site for substrate entrance andproduct release.10

It is clear that in order to obtain a completedescription of the catalytic mechanism of E. coliMSG it is of considerable importance to establishhow the orientation of the domains in the proteinchange and what intra-domain structural pertur-bations occur in response to ligand binding (onlythe glyoxylate-bound structure of E. coli MSG iscurrently available). NMR is a particularly power-ful probe of ligand-induced structural changes inmacromolecules and related studies on a largenumber of diverse systems have been publishedover a period of many years. However, appli-cations involving MSG are complicated by boththe molecular weight (82 kDa) and the number ofresidues in the protein (723) and it was not clear apriori whether data of sufficient quality could be

Figure 1. (a) Reaction catalyzed by malate synthase G. The reaction proceeds via a Claisen condensation of glyoxyl-ate with the acetyl group of acetyl-CoA producing malate and CoA. (b) Ribbon diagram of MSG illustrating the fourmain domains of the molecule. The MSG coordinates were taken from the Protein Data Bank, accession code 1d8c4.

1122 NMR Studies of the 723-Residue Malate Synthase

obtained to allow us to address in a quantitativemanner issues of domain orientation, as well askinetic aspects of ligand binding. The demon-stration that, in fact, such information can beobtained in this case is one of the major results ofthis study.

Building on the virtually complete backbonechemical shift assignments of apo-MSG obtainedfrom four-dimensional NMR spectroscopy11 wepresent here the results of a domain orientationstudy of MSG in the apo-state. Specifically, weshow that the relative orientation of domains isconsistent with the inter-domain structure of theglyoxylate-bound form of the protein observed byX-ray crystallography.4 The absence of domainrearrangements upon ligand binding is distinctfrom what has been observed in versions of theprotein isolated from other species5 – 7 and in othervery related molecules.9,10 However, chemical shiftchanges upon pyruvate and acetyl-CoA bindingto MSG and the kinetic parameters of glyoxylateand pyruvate binding extracted from line-shapesimulations indicate that some minor structuralperturbations are likely to take place upon forma-tion of enzyme–substrate complexes.

Results and Discussion

Domain orientation in apo-MSG from residualdipolar couplings

The development of dilute liquid crystallinemedia for weak alignment of macromolecules12 – 16

along with sensitive methods for the measurementof dipolar couplings resulting from alignment17 – 20

have impacted significantly on the types of struc-tural problems that can be addressed by NMRspectroscopy. One particularly important appli-cation involves measurement of orientationalrestraints to align domains of a molecule withrespect to one another.21,22 We have recently com-pleted a dipolar coupling study of the liganddependence of domain orientation in maltosebinding protein, a 42 kDa single polypeptide chaincomprised of two domains.22,23 Notably, we findthat in a complex with the cyclic hepta-saccharideb-cyclodextrin,22 the relative orientation of thedomains differs between solution and crystal con-formations by 108, while for other complexesand in the apo-form, the inter-domain structuresobtained from NMR and X-ray analyses are verysimilar.23

The 15N– 1HN dipolar coupling (1DNH) is by farthe largest of those that can be measured in uni-formly deuterated proteins and we have thereforechosen to obtain these values as a probe of domainorientation in MSG. However, the measurement of1DNH in high molecular weight systems such asMSG is complicated by the fact that one of the two15N– 1HN multiplet components, referred to inwhat follows as the anti-TROSY component, issignificantly broadened due to the constructive

addition of dipolar and chemical shift anisotropy(CSA) fields.24 Since the distance between the com-ponents must be obtained in any measurement ofdipolar couplings, the accuracy of 1DNH values islimited by the broadening/intensity of the anti-TROSY cross-peak. With this in mind, Yang et al.18

and Bax and co-workers19 have developed TROSY-based HNCO experiments for measurement of1DNH values (Figure 2(a)). 1DNH is obtained byrecording a pair of spectra, the first of which is asimple reference TROSY-HNCO in which theobserved correlation is the red peak in Figure 2(a).In a second experiment correlations at the positionindicated by the blue peak are recorded,18,19 sothat the separation between cross-peaks in the setof two experiments is given by (1JNH þ 1DNH)/2 Hzand the value of 1DNH is readily obtained oncethe scalar coupling, 1JNH, is known (see below).In principle, it is possible to modify the secondexperiment so that correlations at any positionbetween the red (TROSY) and green (anti-TROSY)peaks in the Figure can be obtained. At first glanceit might appear, therefore, that positioning cross-peaks as far from the red correlation as possible inthe second experiment would be desirable to mini-mize the resultant error in 1DNH values. However,because the effective relaxation rate of signalincreases as the position of the cross-peak shiftstowards the green correlation, a compromise mustbe made between a large separation of componentsin the two experiments on the one hand and suf-ficient sensitivity to measure the correlationsaccurately in the first place. We have confirmedexperimentally that the choice of scaling used inthe present set of experiments is optimal for MSG.

In Figure 2(a) we have also included average 15Ntransverse relaxation rates for the TROSY (,65 ms)and anti-TROSY (,10 ms) multiplet componentsmeasured on a deuterated sample of MSG, asdescribed previously.11 In addition, the average T2

value for in-phase 15N magnetization (recordedwith 1H decoupling to inter-convert TROSY andanti-TROSY components) is shown. These valuesprovide a qualitative indication of the dramaticincrease in relaxation rates that can be expectedfor non-TROSY magnetization in a protein the sizeof MSG. A more quantitative estimate, taking intoaccount the specific details of the pulse sequenceand the experimental parameters employed,shows that the effective relaxation time of 15Nmagnetization that gives rise to the “blue corre-lations” in Figure 2 is approximately 30 ms.

Figure 2(b) shows cross-peaks obtained in refer-ence HNCO (red) and coupling-refocused HNCO(blue) spectra for Gly414 and Ala541 recorded ona sample of MSG dissolved in ,12 mg/ml Pf1phage. It is noteworthy that the up-field peaks inFigure 2(b) are significantly broader than thedownfield (reference TROSY) peaks, as expectedfrom the above discussion. The large size of theprotein, coupled with the fact that the sampleused was relatively dilute (0.7 mM) and thatthe measured coupling values are halved in the

NMR Studies of the 723-Residue Malate Synthase 1123

procedure that we have employed, does signifi-cantly increase the errors relative to what is gener-ally observed for applications involving smallermolecules. For example, the standard error of themeasured 1JNH values estimated on the basis ofduplicate experiments is 2.0 Hz, in excellent agree-ment with the standard deviation of the couplingvalues, 1.99 Hz. Of note, contributions to 1JNH frommagnetic susceptibility anisotropy, estimated bycalculating a susceptibility tensor from the orien-tation of aromatic rings and peptide bonds in themolecule,25 – 27 range from þ1.1 Hz to 20.9 Hz(800 MHz), with a standard deviation of 0.48 Hz.The standard error of the (1JNH þ 1DNH) measure-ments in the aligned phase as estimated fromduplicate experiments is 2.7 Hz, reflecting thesomewhat lower quality of NMR spectra obtainedfor MSG aligned in Pf1 phage.13

A total of 415 1DNH couplings, ranging between240 Hz and þ35 Hz have been obtained by(i) subtracting the experimentally determined 1JNH

values from the measured 1JNH þ 1DNH values, asis commonly done,12 and (ii) by subtracting anaverage 1JNH value of 293.9 Hz from 1JNH þ 1DNH.Residues with significantly higher than average15N T1r values,11 or correlations that were par-

tially overlapped were excluded from analysis. Inaddition, a small number (22) of 1DNH values wereexcluded because differences in dipolar couplingslarger than 4.5 Hz were obtained from repeatmeasurements. Since the relative orientation ofdifferent domains in the apo-form of MSG wasnot known a priori, the magnitude and orientationof alignment tensors were calculated separatelyfor each individual domain assuming that theintra-domain structure in solution is the same asin the glyoxylate-bound form determined by X-raystudies.4 The alignment tensor parameters, Aa andR, and the Euler angles, {a, b, g} that define theorientation of each alignment frame with respectto the PDB frame of the molecule are shown inFigure 3(a)–(d), along with the correlation betweenmeasured dipolar couplings and those calculatedon the basis of the domain structures. In this Figurewe have used 1DNH values obtained assuming auniform 1JNH value of 293.9 Hz (see above),although very similar results are generated usingmeasured scalar couplings. The agreementbetween calculated and experimental couplingvalues is very good, especially considering theerrors in the measurement. Compatibility betweenthe dipolar coupling set and the X-ray structure

Figure 2. (a) A schematic illustration of the positions of the TROSY (red), decoupled (blue) and anti-TROSY (green)15N multiplet components. In order to measure 1JNH or 1JNH þ 1DNH values two spectra are recorded, producing corre-lations of the “red” and “blue” types. Average 15N T2 relaxation times for each component are indicated; the 15N T2

value of the anti-TROSY component was estimated from the measured relaxation rates of 1H-decoupled and TROSYcomponents. (b) Correlations for G414 and A541. Note that the upfield and downfield cross-peaks are derived fromseparate spectra, as discussed in Results and Discussion, and have been combined here for ease of visualization. Thesplittings are equal to (1JNH þ 1DNH)/2.


of glyoxylate-bound MSG was established in aquantitative manner by calculating quality factors(Q):

Q ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiXi¼1;N

ðDmeasi 2 Dcalc

i Þ2

Xi¼1;N

ðDmeasi Þ2

vuuuuut ð1Þ

as defined by Ottiger & Bax.15 Quality factors of0.32, 0.27, 0.25, and 0.24 were obtained for theMSG core, the a-helical N-terminal clasp, the a/bdomain, and the C-terminal plug, respectively.The magnitudes of the alignment tensor are wellreproduced from one domain to another. Theslightly lower (absolute) value of the axial com-ponent of the alignment tensor (Aa) and the higherrhombicity (R) in the N-terminal clasp may be theresult of a skewed sampling of N–H bond vectororientations in this part of the protein sincethis domain consists primarily of a single longa-helix.4 Very similar values of the three Eulerangles were obtained for all the domains of MSGindicating that, contrary to the predictions basedon earlier studies with E. coli MSG4 and relatedproteins,7,9,10 the domain orientations in theapo-form are not significantly different from thosefound in the glyoxylate-bound form. Thus, thegross features of the active site within the crevicebetween C-terminal and core domains are pre-formed in the apo-state and subtle structuralchanges likely involving side-chains dominate

substrate binding. In principle, a translational dis-placement of the C-terminal plug would not bereflected in the values of 1DNH and therefore cannotbe ruled out. However, such translational move-ments would be very unlikely in the case ofMSG where the contact surface area between themolecular core and the C-terminal plug is close to2000 A2, since it would lead to the disruption ofnumerous inter-domain interactions.4

Carbonyl chemical shift changesupon alignment

Changes in chemical shifts upon alignment canalso provide valuable structural information28 – 30

and have been used to orient domains of proteins31

and in the refinement of protein32 and nucleic acidstructures.33 The carbonyl chemical shift is a par-ticularly good probe of structure29 since the shiftanisotropy of the carbonyl nucleus is large34 andbecause shifts can be measured accurately usingTROSY-based HNCO spectra.35,36 In this case it isadvantageous to record spectra at fields of600 MHz or lower, because the relaxation of thecarbonyl spin is dominated by CSA, which scaleswith the square of the magnetic field. For example,at 800 MHz the calculated contribution to therelaxation rate of a carbonyl spin in MSG (37 nscorrelation time at 37 8C) from CSA is 93 s21 (corre-sponding to a T2 of 10.8 ms), which compromisesthe accuracy with which such chemical shifts canbe measured. A total of 320 carbonyl shifts were

Figure 3. Experimental versus calculated 1DNH values for the (a) core domain (136 values), (b) N-terminal a-clasp (55values), (c) a/b domain (89 values), and (d) C-terminal plug (73 values). The values of Euler angles and the magni-tudes of the alignment parameters, Aa, R obtained for each domain are indicated. Errors in {Aa, R, a, b, g} of{0.03 £ 1023, 0.02, 1.18, 0.58, 1.48}, {0.09 £ 023, 0.13, 6.58, 1.58, 3.98}, {0.04 £ 1023, 0.03, 1.28, 0.68, 2.38} and {0.04 £ 1023,0.04, 1.48, 0.78, 2.48} are obtained for the core, N-terminal, a/b and C-terminal domains, respectively. Experimental ver-sus calculated carbonyl shifts upon alignment of MSG in Pf1 phage for (e) the whole molecule and for (f) residues ofthe C-terminal plug. The alignment tensor obtained from the full set of 1DNH data was used for the calculation of13CO shifts.


obtained for MSG from 3D TROSY-HNCO spectrarecorded at 600 MHz. Figure 3(e) shows theexcellent agreement between all experimentallymeasured values and those calculated from X-raycoordinates of the MSG–glyoxylate complex4

using the orientation and magnitude of the align-ment tensor obtained from the full set of 15N– 1HNdipolar couplings. The correlation of shiftsobtained for the C-terminal plug domain of MSGexclusively is shown in Figure 3(f), and again verygood agreement is observed. Of note, the level ofagreement between predicted and experimentallyobserved carbonyl shift changes upon align-ment is only slightly worse than that obtainedfor measured and predicted 15N–1HN dipolar

couplings (i.e. similar Q factors) even though theabsolute values of the shift differences are signifi-cantly less than peak line widths in the carbonyldimension of TROSY-HNCO spectra.

The carbonyl shift changes are fully consistentwith the lack of domain rearrangement uponligand binding indicated by the dipolar couplingdata. Very recently, Remington and co-workershave obtained preliminary structures of apoand pyruvate/acetyl-CoA-bound forms of MSG(personal communication) that show that there areno major structural differences between any formsof the protein, in agreement with the results of thepresent study.

Ligand-induced chemical shift changes in MSG

In order to probe possible minor structuralchanges that occur with ligand binding we havemeasured chemical shift changes that accompanythe titration of apo-MSG with pyruvate. Pyruvateserves as an inhibitor that competes with the natu-ral substrate glyoxylate in binding to the activesite of the protein, and can be used in place ofglyoxylate to form a stable ternary (MSG-pyru-vate/acetyl-CoA) complex (note that glyoxylate isconverted to malate in an MSG glyoxylate/acetyl-CoA complex). Since 2D 1HN– 15N correlationmaps of MSG are not sufficiently resolved, 3DTROSY-HNCO spectra were used to follow 3Dtrajectories of the resonances experiencing chemi-cal shift changes upon pyruvate binding. Notably,the chemical shift changes were predominantlysmall and localized to several stretches in theMSG sequence, so that unambiguous assignmentsof nearly all those amides assigned in the apo-form were possible (98.6%). After the enzyme wassaturated with pyruvate the sample was titratedwith acetyl-CoA, and the assignment procedurewas repeated. A total of 95.5% of all the amidesassigned in the apo-form were available for analy-sis in the pyruvate/acetyl-CoA-bound form.Unfortunately, several residues belonging to theactive site could not be identified since they wereunassigned in the apo-form of the protein.11 Figure4(a) summarizes the chemical shift changes uponpyruvate (black) and pyruvate/acetyl-CoA (red)binding. Relative to changes in chemical shiftsobserved in a ligand binding study involvingmaltose-binding protein,23 the shift changes notedupon pyruvate binding to MSG are quite small,with maximal changes in the 1HN, 15N, and 13COdimensions of 0.74 (E630), 3.2 (H641), and 1.3 ppm(Q116), respectively. Notably, the most significantchanges in 15N and 1HN chemical shifts upon pyru-vate binding occur in the I615-H641 stretch. Mostof this region in MSG is part of the b-hairpin loopin the C-terminal domain that interfaces with theTIM core barrel.4 The observed shifts indicate thatsome minor structural changes in this regionaccompany the binding of pyruvate.

Despite the aromatic moiety of acetyl-CoA, itsbinding results in shift perturbations that are on

Figure 4. (a) 1HN, 15N and 13CO chemical shift changesin MSG upon binding of pyruvate (black) and acetyl-CoA after saturation with pyruvate (red) versus residuenumber. The residues of MSG with atoms closer than8 A to any of the heavy atoms of the bound glyoxylatein the MSG crystal structure4 are shown in black abovethe plot. (b) Combined chemical shift changes37 uponbinding of pyruvate color-coded onto the 3D structureof MSG (PDB accession code 1d8c4). The shift changesshown in red exceed 0.2 ppm and those shown in yelloware .0.04 ppm and ,0.1 ppm. A gradient of color isused to indicate shift changes between 0.1 ppm and0.2 ppm. The bound glyoxylate molecule is shown ingreen. The Figure was prepared with the programMOLMOL.56


average 2.0–2.5 times smaller than what is notedupon pyruvate binding. These shifts are, ingeneral, confined to the same regions in the proteinsequence that are observed for pyruvate binding,with the addition of significant changes in the300–310 flexible loop in the acetyl-CoA-boundform. In addition, a 2D 1HN– 15N correlation-basedtitration of apo-MSG with glyoxylate was carriedout to get (limited) information about changesthat accompany glyoxylate binding. Of theapproximately 200 correlations in the 2D spectrathat were sufficiently well resolved to be quanti-fied, 32 peaks showed measurable shift changesthat were, in general, somewhat larger in absolutevalue than those occurring upon pyruvate binding.All shift changes were in the same direction asnoted for pyruvate with the exception of theamide 15N and 1HN shifts of W509. The amideproton and nitrogen atoms of W509 are 9.7 A and9.5 A away from the aldehyde carbon of glyoxylatein the MSG crystal structure,4 respectively, and aretherefore likely to be in proximity to the methylgroup of pyruvate in the pyruvate-MSG complex.

Figure 4(b) displays the combined 15N and 13Cbackbone chemical shift changes23,37 (see Materialsand Methods) that occur upon pyruvate bindingcolor-coded onto the crystal structure of glyoxyl-ate-bound MSG.4 Though the ligand-inducedchanges are clearly spatially concentrated aroundthe glyoxylate-binding site, several residues withsignificant shift changes are almost 20 A awayfrom the glyoxylate molecule. For example, thechemical shifts in the loop spanning residues89–115 linking the N-terminal clasp with themolecular core consistently changed in all of thetitration series. Only the side-chain atoms of V118and the carboxyl group of E116 within this loopare within 10 A of the bound glyoxylate. Thesechanges may be indicative of minor conforma-tional changes in the interface between themolecular core and the N-terminal a-helix that arenecessary to accommodate the substrate in theactive site. No significant chemical shift changeswere observed for the residues linking the core ofthe molecule with the C-terminal plug (551–589).In sum, the titration data provide evidence thatmajor structural rearrangements do not accompanyligand binding in MSG, consistent with the resultsof the dipolar coupling study and carbonyl shiftchanges that occur upon alignment describedabove.

Relaxation and hydrogen exchangemeasurements support the absence ofsignificant structural changes uponligand binding

It was shown in our previous study that apo-MSG tumbles as a single spherical particle witha correlation time of 37 ns at 37 8C.11 We havesupplemented the 15N relaxation study of apo-MSG described previously by recording 15N T1r

values for the protein in the pyruvate and

pyruvate/acetyl-CoA-bound states and thesevalues along with those for the free form of theenzyme are plotted as a function of residuenumber in Figure 5(a). The same subsets of 165well resolved peaks in 2D 1HN– 15N correlationmaps of MSG were selected for each state of theprotein for the comparison. The relaxation timeswere found to be homogeneous and small(,20 ms) throughout the MSG sequence for allforms of the protein, except for several residuesin flexible loops, including the loop comprisingresidues 143–160 in the a/b domain, the 300–310flexible stretch that is missing from the X-raycoordinates of MSG,4 and the linker between themolecular core and the C-terminal plug. Of note,the 300–310 loop appears to be more rigid in thepyruvate/acetyl-CoA-bound state, reflected in thelower 15N T1r values in relation to the other formsof the protein.

The 15N spin relaxation experiments describedabove probe dynamics on a ps-ns time scale. Inorder to establish whether there are slower pro-cesses that might be present we have alsomeasured hydrogen exchange with solvent using3D TROSY-HNCO versions of the CLEANEX-HSQC experiment.38 The exchange rates of 28amides could be quantified accurately (ratesgreater than ,3 s21) and are shown in Figure 5(b)for the apo-state of MSG. In general, regions withhigh exchange rates correlate well with thosehaving elevated 15N transverse relaxation times.Of note, a very similar exchange profile is obtainedfor the pyruvate bound state of the protein.

Kinetic parameters of ligand binding to MSGand structural implications

The binding of both pyruvate and glyoxylate toMSG occurs in the intermediate exchange regimeon the NMR chemical shift time scale for manyresidues in the protein and it is therefore possibleto quantify the kinetic parameters describingligand binding in a straightforward manner. Figure6(a) summarizes the results of a titration of apo-MSG (black) with pyruvate (fully bound is red),focusing on the well resolved region of the1HN–15N HSQC-TROSY spectrum of the protein.In many cases considerable line-broadening isapparent at intermediate ligand concentrations.Equilibrium dissociation constants (Kd) of theglyoxylate, pyruvate and pyruvate/acetylCoA-MSG complexes were determined by least-squaresfitting of the chemical shift changes accompanyingligand binding as a function of the total ligandconcentration as described in Materials andMethods. The best-fit curves in both 15N and 1HNdimensions for V620 titrated with pyruvate andglyoxylate are shown in Figure 6(b) and (c),respectively. Estimated Kd values averaged over 20well resolved peaks with significant shifts in either1HN or 15N dimensions are 1.02(^0.15) mM and600(^70) mM for pyruvate and glyoxylate, respec-tively. Notably, glyoxylate, the physiological


substrate of MSG, binds more strongly than pyru-vate. The Kd value for acetyl-CoA binding to thepyruvate-saturated form of MSG could not bedetermined with the same degree of accuracy asthe other ligands because of generally smallershift changes and was estimated as 270(^120) mM.

The exchange kinetics of binding have beendetermined for both pyruvate and glyoxylateassuming that the binding reaction can bedescribed in its simplest form as:

P þ L$kon

koff

PL ð2Þ

where P, L and PL denote free protein, free ligandand complex, respectively. NMR line broadeningis a function of kon[L] and koff and is, therefore, anexcellent probe of exchange kinetics.39 A numberof different methods have been used in the past tomeasure binding kinetics including relaxationdispersion spectroscopy39 and line-shape analy-

sis.40 – 42 Here we have used the latter approach bycurve-fitting the line-shapes derived from 2D1HN– 15N TROSY-HSQC spectra recorded at dif-ferent ligand concentrations. Since the dissociationconstants (Kd ¼ koff/kon) are already known a seriesof N spectra belonging to the same titration set canbe fit to a common on-rate, kon, and a set of N con-stants that scale each of the spectra independently.These constants account for the fact that in anygiven fitting only a single dimension of a 2D spec-trum is considered; any change in line width inthe second (unfitted) dimension that occurs uponligand binding also influences the intensity ofpeaks.42 Second, the efficiency of magnetizationtransfer during constant-time periods in the 2Dexperiment will be a function of the relaxationproperties of both 1HN and 15N magnetization,which will change with added ligand. Figure 6(d)and (e) shows the two best examples of line-shape simulations from titration data for glyoxyl-ate (1HN dimension of I482) and pyruvate (15N

Figure 5. (a) 15N T1r relaxationtimes versus the MSG sequence forthe same subset of 165 wellresolved residues in 2D TROSY-HSQC spectra of MSG in the apo(black), pyruvate-bound (red) andpyruvate/acetyl-CoA-bound(green) protein states. (b) Solventexchange rates for the amides ofapo-MSG versus residue number.The inset shows the build-up ofpeak intensities for residues E583and A589 as a function of mixingtime. Regions corresponding to thedifferent domains of MSG are colorcoded above the plots.


dimension of V620), respectively. It is noteworthythat even in the cases where the chemical shiftchanges in only one dimension (e.g. the 15N chemi-cal shift of I482 does not change upon titration withglyoxylate) severe signal attenuation may occur atintermediate ligand concentrations because of themuch lower efficiency of INEPT magnetizationtransfers in the TROSY-HSQC spectra (see above).

The on-rates for pyruvate derived from line-shape fitting of five peaks in both dimensionsvary between 0.8 £ 106 M21 s21 and 1.5 £106 M21 s21, while those for glyoxylate, derivedfrom fitting four peaks in both dimensions, areslightly higher, 1.5 £ 106 –3.8 £ 106 M21 s21. Theseon-rates are about two orders of magnitude lowerthan bimolecular diffusion-controlled rates (,108 –109 M21 s21) and are similar to values reported formany enzymes and their substrates43 (106 –108 M21 s21). The entry rate of these ligands maybe limited by the fact that they are polar and thatbinding is preceded by desolvation of the polarcarboxylic group. Alternatively, there may be acti-vation barriers associated with ligand burial insidethe protein that are due to the polarity of the ligandor that reflect subtle structural rearrangements thatoccur upon binding. The latter interpretation isconsistent with the large number of small chemicalshift changes that are observed upon ligand bind-ing. Of interest, the on-rates reported here are simi-lar to those measured for the binding of benzenederivatives to the Leu99Ala cavity mutant of T4lysozyme. It was suggested that in the lysozyme

case the magnitude of these values may reflectmore issues of charge burial/desolvation ratherthan steric constraints that must first be removedby dynamics.44

Conclusions

A major result of this study is that quantitative,site-specific structural and dynamic informationcan be obtained on systems as large as MSG. Thepresent work has exploited recent developmentsin NMR methodology including the measurementof orientational restraints in partially alignedsamples,45 and the preservation of magnetizationusing the TROSY principle.24 The latter is quiteessential for applications involving molecules thesize of MSG.

Based on dipolar couplings and chemical shiftchanges upon alignment it is clear that the inter-domain structures of the apo and glyoxylate-bound forms of MSG are very similar and that thechanges expected on the basis of earlier work4 donot occur. However, chemical shift changes uponpyruvate and acetyl-CoA binding to MSG and thekinetic parameters of glyoxylate and pyruvatebinding extracted from line-shape simulationsderived from titration data indicate that someminor structural adjustments are likely to takeplace upon formation of complexes. It is antici-pated that multi-dimensional NMR will play anincreasingly important role in the characterization

Figure 6. (a) Overlay of a selectedregion of 800 MHz 1HN–15NTROSY-HSQC spectra of [15N, 2H]-labeled MSG recorded as a functionof pyruvate concentration. Arrowsindicate the directions of peakmovement. Resonances with signifi-cant chemical shift changes arelabeled with residue numbers.(b) and (c) Kd determination usingchemical shift changes of V620upon titration with (b) pyruvateand (c) glyoxylate. The insets showthe best-fits from changes in 1HNshifts. (d) and (e) Examples of line-shape simulations for (d) I482 (1HNdimension) in titration withglyoxylate and (e) V620 (15N dimen-sion) in titration with pyruvate.Experimental line-shapes are shownin red and the results of the simu-lations are shown in blue.


of the structural and dynamic features of large pro-teins at a level of quantification that until recentlywas reserved for only small molecular weightsystems.

Materials and Methods

NMR samples

U-[15N,13C,2H] and U-[15N,2H]-labeled samples of MSGwere obtained by overexpression from cultures of E. coliBL21(DE3)pLysS cells transformed with the plasmidpMSG-B encoding all the residues of MSG with (i) thesubstitution of Ser2 with an Ala, (ii) the addition of anN-terminal Met and (iii) the addition of an eight-residue C-terminal hexa-histidine tag (LEHHHHHH) asdescribed in detail.11 For the preparation of 13C-labeledsamples [13C,1H]glucose (Cambridge Isotope Labora-tories, Andover, MA) was used as the sole carbon source.After initial purification on a nickel affinity column(Ni-NTA Agarose, Qiagen) the protein was fullydenatured and refolded in vitro in order to fully proto-nate the amide positions of those residues buried in thecore of the molecule (and therefore deuterated aftergrowths in the 2H2O-based bacterial medium). Thedetails of the in vitro refolding protocol and subsequentpurification steps have been described.11 The sodiumsalts of pyruvic and glyoxylic acids and acetyl-CoenzymeA used in the titrations described above werepurchased from Sigma (Ontario, Canada) and were notfurther purified.

All NMR samples were between 0.3 mM and 0.8 mMin protein dissolved in 20 mM sodium phosphate buffer(pH 7.1), 5 mM DTT, 20 mM MgCl2, 0.05% NaN3,0.1 mg/ml Pefabloc and 8% 2H2O. Alignment of proteinwas achieved using a sample with Pf1 phage13,46 addedto a concentration of ,12 mg/ml (residual 2H watersplitting of 9.8 Hz), which provided 15N–1HN dipolarcouplings between 240 Hz and þ35 Hz.

NMR spectroscopy

All NMR experiments were performed on four-channel Varian Inova spectrometers operating at 37 8Cand equipped with pulsed-field gradient triple reso-nance probes.

1DNH values were measured for apo-MSG from 3DTROSY-HNCO-based spectra18,19 recorded at 800 MHz.Two spectra were recorded in an interleaved manner,including a regular TROSY-HNCO and a TROSY-HNCOmodified by including a 1H inversion pulse at time t/2prior to the end of the final 15N constant-time evolutionperiod, where t is the 15N acquisition time.18,19 The sepa-ration of cross-peaks in the pair of spectra is givenby (1JNH þ 1DNH)/2 Hz. Data sets comprised of28 £ 54 £ 863 complex points in each of t1(

13CO), t2(15N),

t3(1HN), corresponding to acquisition times of 13.4 ms,

23.4 ms and 68 ms were recorded, with eight (TROSY-HNCO) and 12 (TROSY-HNCO þ 1H inversion pulse)scans per FID (1.7 second relaxation delay) for a totalacquisition time of approximately 60 hours per pair ofexperiments. Spectra were analyzed using PIPP/CAPP47

software with uncertainties in 1JNH and 1DNH valuesobtained from the pair-wise r.m.s.d of coupling data inrepeat experiments performed in both isotropic andaligned media (standard error in each of the two repeatexperiments ¼ 1/

p2 of the pair-wise r.m.s.d). 3D

TROSY-HNCO35,36,48 data sets were used to quantify thechanges in carbonyl chemical shifts upon alignment, asdescribed by Choy et al.31 Data sets were recorded at600 MHz and comprised 48 £ 54 £ 612 complex pointsin t1,t2,t3 (acquisition times of 31.2 ms, 31.2 ms, 64 ms int1,t2,t3). Eight scans were obtained per FID with a relax-ation delay of 1.9 seconds to give a net measuring timeof 44 hours for each data set. Selection of the TROSYcomponent in all the experiments was achieved pas-sively by relying exclusively on the efficient relaxationof the anti-TROSY component during the relatively longtransfer delays in the pulse schemes.35,49 All NMRspectra were processed using the suite of programsprovided in NMRPipe/NMRDraw software.50

1HN, 15N and 13CO chemical shift changes in MSGresulting from either pyruvate or acetyl-CoA bindingwere quantified by addition of aliquots of ligands andfollowing peak trajectories in 3D TROSY-HNCO spectrarecorded at 800 MHz. Pyruvate was added to a 0.35 mMU-[15N,13C,2H]-labeled sample of MSG to give finalligand concentrations of 0.35, 0.72, 0.90, 1.90, 3.90, 12.2and 20.5 mM; acetyl-CoA was subsequently added tothe same sample to concentrations of 0.3, 0.7, 1.7 and4.0 mM. Glyoxylate was added to a 0.50 mM U-[15N,2H]-labeled protein sample to give ligand concentrationsof 0.5, 1.5, 2.0, 5.0 and 20.0 mM and the titration wasmonitored using a series of 2D TROSY-HSQC spectra.All spectra were analyzed using NMRView software51 inconjunction with auxiliary tcl/tk scripts written in-house. The combined chemical shift change of a particu-lar residue upon ligand binding was calculated as:

Ddtot ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðwNDdNÞ

2 þ ðwCODdCOÞ2

qð3Þ

where wi denotes the weight factor of nucleus i. Theweight factors were determined from the ratio of theaverage standard deviations of the 1HN shifts andthe chemical shifts of nucleus type i as observed for the20 common amino acids in proteins using the Bio-MagResBank chemical shift database†: wHN ¼ 1, wN ¼0.154, and wCO ¼ 0.341.23,37 1HN shifts were not takeninto account in the calculation since proton shifts aremore sensitive to neighboring effects than 13C and 15Nchemical shifts. The value of Ddtot principally reflects,therefore, changes in backbone f/c dihedral angles.

NMR experiments for the measurement of 15N T1r at800 MHz 1H frequency made use of TROSY versions ofpulse sequences described elsewhere.52 The spin lockpower used in the 15N T1r experiments was 1.6 kHz. T1r

values were determined by non-linear least-squaresfitting of the experimental data to monoexponentialdecay functions, A expð2t=T1rÞ; using the programModelXY50 or MATLAB v.6 (MathWorks Inc., MA).

Amide hydrogen exchange rates with solvent weremeasured using a 3D TROSY-HNCO version of theCLEANEX experiment38 with exchange times of 22, 44and 66 ms. Spectra were recorded at 800 MHz with18 £ 44 £ 818 complex points in t1(

13CO),t2(15N),t3(

1HN)(acquisition times of 8.5 ms, 19.0 ms, 64.0 ms in t1,t2,t3).A total of 16 scans were obtained per FID with a relax-ation delay of 1.5 seconds to give a net measuring timeof 21 hours per spectrum. Reference intensities weremeasured from a 3D TROSY-HNCO spectrum recordedwith a ten second relaxation delay, four scans and allother parameters as above. Exchange rate constantswere extracted from a least-squares fit of normalized

† http://www.bmrb.wisc.edu


http://www.bmrb.wisc.edu

peak intensities to a magnetization build-up equationdescribed elsewhere.38

Analysis of dipolar couplings and 13CO chemicalshift changes upon alignment

Alignment parameters, Aa, R, and the three Eulerangles {a, b, g}, were obtained by optimizing the agree-ment between the experimental dipolar couplings andthe couplings predicted from the glyoxylate-boundMSG X-ray structure4 using in-house softwareRDCA.22,31 Predicted values are calculated from the coor-dinates of the 3D structure of MSG4 using the followingrelationship:12,25

Dij ¼ D0ijAaS ð3 cos2u2 1Þ þ

3

2R sin2u cos 2f

� �ð4Þ

where D0ij ¼ 2ð1=2pÞðm0=4pÞgigj�kr23

ij l is the dipolarinteraction constant, rij is the distance between nuclei iand j, gi is the gyromagnetic ratio of nucleus i, S is theorder parameter that reflects averaging due to fast localdynamics, and Aa and R are the axial and rhombiccomponents of the alignment tensor, respectively. Thepolar angles {u, f} describe the orientation of the dipolarvector with respect to the alignment frame, while theEuler angles {a, b, g} give the orientation of the principalaxes of the alignment tensor with respect to the PDBframe. Errors in the alignment parameters were esti-mated on the basis of jackknife analyses where 20% ofthe data was removed from each of 100 iterations andby a Monte Carlo approach, using a standard error indipolar coupling values of 3.2 Hz. The larger of the twosets of errors is reported for {Aa, R, a, b, g} in the legendto Figure 3.

The change in chemical shift for a 13CO spin uponalignment was calculated from the followingrelationship:25,29

Dd ¼ 1=3X

i¼x;y;z

Xj¼x;y;z

Ajj cos2uijdii ð5Þ

where uij is the angle between the ii principal axis of thetraceless CSA tensor (with principal components dii)and the jj principal axis of the traceless molecular align-ment tensor (with principal values Ajj). The followingvalues were used for the 13CO CSA tensor components:dxx ¼ 73.33 ppm, dyy ¼ 7.33 ppm and dzz ¼ 280.66 ppmwith the z axis orthogonal to the peptide plane and thex axis forming an angle of 408 with respect to the CO–Nbond (see Figure 1 of Choy et al.31); these values werenot altered during calculation of the r.m.s.d betweencalculated and experimental 13CO chemical shifts. Inaddition to the CSA dependent chemical shift offsetthere is an additional constant offset that derives fromthe fact that the frequency at which the spectrometer islocked changes with alignment. This is easily taken intoaccount from the 2H2O splitting in the aligned sample,as described.31

Binding constants and line-shape analysis

Dissociation constants for the pyruvate–MSG andglyoxylate–MSG complexes were obtained from least-squares fitting of chemical shift changes as a function oftotal ligand concentration according to the relation:53

di ¼b 2

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffib2 2 4 £ a £ c

p

2að6Þ

with a ¼ ðKa=dbÞ £ ½Pt�; b ¼ 1 þ Kað½Lti� þ ½Pt�Þ; andc ¼ dbKa½Lti�, where di is the absolute change in chemicalshift for each titration point, [Lti] is the total ligand con-centration at each titration point, [Pt] is the total proteinconcentration, Ka ¼ 1=Kd is the binding constant, and db

is the chemical shift of the resonance in question in thecomplex. Kd and db were used as fitting parameters inthis analysis.

Line-shape simulations of MSG spectra at differentligand concentration were performed separately in 1HNand 15N dimensions using software written for MATLABv.6 (MathWorks Inc., MS). Analytical expressions forline-shapes in NMR spectra of exchanging systemscommonly used to simulate two-site exchange54,55 wereused, with a pseudo-first-order on-rate, kon[L]. The lineintensities in a given dimension at each titration pointwere scaled using adjustable coefficients42 to accountfor (i) possible line broadening in the other dimensionand (ii) ligand-dependent efficiencies of INEPT-basedmagnetization transfers in the pulse sequence, asdescribed in Results and Discussion. In each simulationthe fitted parameters included kon and the adjustablecoefficients (equal to the number of simultaneously fittedtitration points). The intrinsic transverse relaxation timeswere not used as free fitting parameters and were esti-mated as 1=ðp £ LWÞ; where LW is the peak line width(Hz) at half-height in a given dimension of the 2DTROSY-HSQC spectrum of the ligand-free protein.

Acknowledgements

This work was supported by a grant from theCanadian Institutes of Health Research to L.E.K.We thank Professor S. J. Remington (University ofOregon) for the gift of the plasmid used for MSGexpression and for a number of valuable dis-cussions during the course of this work, ProfessorCheryl Arrowsmith (Ontario Cancer Institute) forkindly allowing us to carry out the proteinexpression and purification in her laboratory, DrNico Tjandra (NIH, Bethesda, MD) for the programto calculate dipolar couplings resulting from align-ment due to magnetic susceptibility anisotropy,and Drs Oscar Millet, Martin Tollinger and DmitryKorzhnev (University of Toronto) for assistanceand helpful discussions. V.T. is a recipient of aHuman Frontiers Postdoctoral Fellowship. L.E.K.holds a Canada Research Chair in Biochemistry.

References

1. Molina, I., Pellicer, M. T., Badia, J., Aguilar, J. &Baldoma, L. (1994). Molecular characterization ofEscherichia coli malate synthase G. Differentiationwith the malate synthase A isoenzyme. Eur.J. Biochem. 224, 541–548.

2. Cornforth, J. W., Redmond, J. W., Eggerer, H., Buckel,W. & Gutschow, C. (1969). Asymmetric methylgroups, and the mechanism of malate synthase.Nature, 221, 1212–1213.

3. Luthy, J., Retey, J. & Arigoni, D. (1969). Preparationand detection of chiral methyl groups. Nature, 221,1213–1215.


4. Howard, B. R., Endrizzi, J. A. & Remington, S. J.(2000). Crystal structure of Escherichia coli malatesynthase G complexed with magnesium and glyoxyl-ate at 2.0 A resolution: mechanistic implications. Bio-chemistry, 39, 3156–3168.

5. Schmid, G., Durchschlag, H., Biedermann, G.,Eggerer, H. & Jaenicke, R. (1974). Molecular structureof malate synthase and structural changes uponligand binding to the enzyme. Biochem. Biophys. Res.Commun. 58, 419–426.

6. Beeckmans, S., Khan, A. S., Kanarek, L. & VanDriessche, E. (1994). Ligand binding to maize (Zeamays) malate synthase: a structural study. Biochem.J. 303, 413–421.

7. Zipper, P. & Durchschlag, H. (1977). Small-angleX-ray studies on malate synthase from baker’syeast. Biochem. Biophys. Res. Commun. 75, 394–400.

8. Larsen, T. M., Laughlin, L. T., Holden, H. M.,Rayment, I. & Reed, G. H. (1994). Structure of rabbitmuscle pyruvate kinase complexed with Mn2 þ ,K þ , and pyruvate. Biochemistry, 33, 6301–6309.

9. Larsen, T. M., Laughlin, L. T., Holden, H. M.,Rayment, I. & Reed, G. H. (1997). Ligand-induceddomain movement in pyruvate kinase: structure ofthe enzyme from rabbit muscle with Mg2 þ , K þ ,and L-phospholactate at 2.7 A resolution. Arch. Bio-chem. Biophys. 345, 199–206.

10. Remington, S. J., Wiegand, G. & Huber, R. (1982).Crystallographic refinement and atomic models oftwo different forms of citrate synthase at 2.7 and1.7 A resolution. J. Mol. Biol. 158, 111–152.

11. Tugarinov, V., Muhandiram, R., Ayed, A. & Kay, L. E.(2002). Four-dimensional NMR spectroscopy of a723-residue protein: chemical shift assignments andsecondary structure of malate synthase G. J. Am.Chem. Soc. 124, 10025–10035.

12. Tjandra, N. & Bax, A. (1997). Direct measurement ofdistances and angles in biomolecules by NMR in adilute liquid crystalline medium. Science, 278,1111–1114.

13. Hansen, M. R., Mueller, L. & Pardi, A. (1998).Tunable alignment of macromolecules by fila-mentous phage yields dipolar coupling interactions.Nature Struct. Biol. 5, 1065–1074.

14. Clore, G. M., Starich, M. R. & Gronenborn, A. M.(1998). Measurement of residual dipolar couplingsof macromolecules aligned in the nematic phase ofa colloidal suspension of rod-shaped viruses. J. Am.Chem. Soc. 120, 10571–10572.

15. Ottiger, M. & Bax, A. (1999). Bicelle-based liquidcrystals for NMR-measurement of dipolar couplingsat acidic and basic pH values. J. Biomol. NMR, 13,187–191.

16. Ruckert, M. & Otting, G. (2000). Alignment of bio-logical macromolecules in novel nonionic liquidcrystalline media for NMR experiments. J. Am.Chem. Soc. 122, 7793–7797.

17. Ottiger, M., Delaglio, F. & Bax, A. (1998). Measure-ment of J and dipolar couplings from simplifiedtwo-dimensional NMR spectra. J. Magn. Reson. 131,373–378.

18. Yang, D., Venters, R. A., Mueller, G. A., Choy, W. Y. &Kay, L. E. (1999). TROSY-based HNCO pulsesequences for the measurement of 1HN–15N,15N–13CO, 1HN–13CO, 13CO–13Ca dipolar couplingsin 15N, 13C, 2H-labeled proteins. J. Biomol. NMR, 14,333–343.

19. Kontaxis, G., Clore, G. M. & Bax, A. (2000). Evalua-tion of cross-correlation effects and measurement of

one-bond couplings in proteins with short transverserelaxation times. J. Magn. Reson. 143, 184–196.

20. Permi, P. (2000). A set of HNCO-based experimentsfor measurement of residual dipolar couplings in15N, 13C, (2H)-labeled proteins. J. Biomol. NMR, 17,43–54.

21. Fischer, M. W., Losonczi, J. A., Weaver, J. L. & Preste-gard, J. H. (1999). Domain orientation and dynamicsin multidomain proteins from residual dipolarcouplings. Biochemistry, 38, 9013–9022.

22. Skrynnikov, N. R., Goto, N. K., Yang, D., Choy, W. Y.,Tolman, J. R., Mueller, G. A. & Kay, L. E. (2000).Orienting domains in proteins using dipolarcouplings measured by liquid-state NMR: differ-ences in solution and crystal forms of maltodextrinbinding protein loaded with b-cyclodextrin. J. Mol.Biol. 295, 1265–1273.

23. Evenas, J., Tugarinov, V., Skrynnikov, N. R., Goto,N. K., Muhandiram, R. & Kay, L. E. (2001). Ligand-induced structural changes to maltodextrin-bindingprotein as studied by solution NMR spectroscopy.J. Mol. Biol. 309, 961–974.

24. Pervushin, K., Riek, R., Wider, G. & Wuthrich, K.(1997). Attenuated T2 relaxation by mutual cancella-tion of dipole–dipole coupling and chemical shiftanisotropy indicates an avenue to NMR structuresof very large biological macromolecules in solution.Proc. Natl Acad. Sci. USA, 94, 12366–12371.

25. Bastiaan, E. W., MacLean, C., van Zijl, P. C. M. &Bothner-By, A. A. (1987). High resolution NMR ofliquids and gases: effects of magnetic-field-inducedmolecular alignment. Annu. Rep. NMR Spectrosc. 9,3635–3677.

26. Tolman, J. R., Flanagan, J. M., Kennedy, M. A. &Prestegard, J. H. (1995). Nuclear magnetic dipoleinteractions in field-oriented proteins: informationfor structure determination in solution. Proc. NatlAcad. Sci. 92, 9279–9283.

27. Tjandra, N., Grzesiek, S. & Bax, A. (1996). Magneticfield dependence of nitrogen-proton J splittings in15N enriched human ubiquitin resulting from relaxa-tion interference and residual dipolar coupling.J. Am. Chem. Soc. 118, 6264–6272.

28. Ottiger, M., Tjandra, N. & Bax, A. (1997). Magneticfield dependent amide 15N chemical shifts in aprotein–DNA complex resulting from magneticordering in solution. J. Am. Chem. Soc. 119,9825–9830.

29. Cornilescu, G., Marquardt, J., Ottiger, M. & Bax, A.(1998). Validation of protein structure from aniso-tropic carbonyl chemical shifts in a dilute liquidcrystalline phase. J. Am. Chem. Soc. 120, 6836–6837.

30. Boyd, J. & Redfield, C. (1999). Characterization of15N chemical shift anisotropy from orientation-dependent changes to 15N chemical shifts in dilutebicelle solution. J. Am. Chem. Soc. 121, 7441–7442.

31. Choy, W. Y., Tollinger, M., Mueller, G. A. & Kay, L. E.(2001). Direct structure refinement of high molecularweight proteins against residual dipolar couplingsand carbonyl chemical shift changes upon align-ment: an application to maltose binding protein.J. Biomol. NMR, 21, 31–40.

32. Lipsitz, R. S. & Tjandra, N. (2001). Carbonyl CSArestraints from solution NMR for protein structurerefinement. J. Am. Chem. Soc. 123, 11065–11066.

33. Wu, Z., Tjandra, N. & Bax, A. (2001). 31P chemicalshift anisotropy as an aid in determining nucleicacid structure in liquid crystals. J. Am. Chem. Soc.123, 3617–3618.


34. Teng, Q., Iqbal, M. & Cross, T. A. (1992). Determi-nation of the 13C chemical shift and 14N electricfield gradient tensor orientations with respect to themolecular frame in a polypeptide. J. Am. Chem. Soc.114, 5312–5321.

35. Yang, D. & Kay, L. E. (1999). Improved 1HN-detectedtriple resonance TROSY-based experiments. J. Biomol.NMR, 13, 3–10.

36. Salzmann, M., Pervushin, K., Wider, G., Senn, H. &Wuthrich, K. (1998). TROSY in triple-resonanceexperiments: new perspectives for sequential NMRassignment of large proteins. Proc. Natl Acad. Sci.USA, 95, 13585–13590.

37. Mulder, F. A. A., Schipper, D., Bott, R. & Boelens, R.(1999). Altered flexibility in the substrate-bindingsite of related natve and engineered high-alkalineBacillus subtilisins. J. Mol. Biol. 292, 111–123.

38. Hwang, T. L., van Zijl, P. C. & Mori, S. (1998).Accurate quantitation of water–amide protonexchange rates using the phase-modulated CLEANchemical EXchange (CLEANEX-PM) approach witha Fast-HSQC (FHSQC) detection scheme. J. Biomol.NMR, 11, 221–226.

39. Palmer, A. G., Kroenke, C. D. & Loria, J. P. (2001).NMR methods for quantifying microsecond-to-milli-second motions in biological macromolecules.Methods Enzymol. 339, 204–238.

40. Sandstrom, J. (1982). Dynamic NMR Spectroscopy,Academic Press, New York.

41. Hensmann, M., Booker, G. W., Panayotou, G., Boyd,J., Linacre, J., Waterfield, M. & Campbell, I. D.(1994). Phosphopeptide binding to the N-terminalSH2 domain of the p85 alpha subunit of PI 30-kinase:a heteronuclear NMR study. Protein Sci. 3,1020–1030.

42. Gunther, U. L. & Schaffhausen, B. (2002). NMRKIN:simulating lineshapes from two-dimensional spectraof proteins upon ligand binding. J. Biomol. NMR, 22,201–209.

43. Fersht, A. (1985). Enzyme Structure and Mechanism,2nd edit., Freeman, Co, New York.

44. Feher, V. A., Baldwin, E. P. & Dahlquist, F. W. (1996).Access of ligands to cavities within the core of aprotein is rapid. Nature Struct. Biol. 3, 516–521.

45. Prestegard, J. H. (1998). New techniques in structuralNMR—anisotropic interactions. Nature Struct. Biol.NMR supplement, 5, 517–522.

46. Zweckstetter, M. & Bax, A. (2001). Characterizationof molecular alignment in aqueous suspensions ofPf1 bacteriophage. J. Biomol. NMR, 20, 365–377.

47. Garrett, D. S., Powers, R., Gronenborn, A. M. &Clore, G. M. (1991). A common sense approach topeak picking in two-, three-, and four-dimensionalspectra using automatic computer analysis of con-tour diagrams. J. Magn. Reson. 95, 214–220.

48. Yang, D. & Kay, L. E. (1999). Improved lineshape andsensitivity in the HNCO-family of triple-resonanceexperiments. J. Biomol. NMR, 14, 273–276.

49. Yang, D. & Kay, L. E. (1999). TROSY triple resonancefour-dimensional NMR spectroscopy of a 46 nstumbling protein. J. Am. Chem. Soc. 121, 2571–2575.

50. Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G.,Pfeifer, J. & Bax, A. (1995). NMRPipe: a multidimen-sional spectral processing system based on UNIXpipes. J. Biomol. NMR, 6, 277–293.

51. Johnson, B. A. & Blevins, R. A. (1994). NMRView:a computer program for the visualization and analy-sis of NMR data. J. Biomol. NMR, 4, 603–614.

52. Farrow, N. A., Muhandiram, R., Singer, A. U., Pascal,S. M., Kay, C. M., Gish, G. et al. (1994). Backbonedynamics of a free and phosphopeptide-complexedsrc homology 2 domain studied by 15N NMR relaxa-tion. Biochemistry, 33, 5984–6003.

53. Johnson, P. E., Tomme, P., Joshi, M. D. & McIntosh,L. P. (1996). Interaction of soluble cellooligosacchar-ides with the N-terminal cellulose-binding domainof Cellulomonas fimi CenC 2. NMR and ultra-violet absorption spectroscopy. Biochemistry, 35,13895–13906.

54. Huang, G. S. & Oas, T. G. (1995). Submillisecondfolding of monomeric g repressor. Proc. Natl Acad.Sci. USA, 92, 6878–6882.

55. Burton, R. E., Busby, R. S. & Oas, T. G. (1998).ALASKA: a mathematica package for two-statekinetic analysis of protein folding reactions. J. Biomol.NMR, 11, 355–359.

56. Koradi, R., Billeter, M. & Wuthrich, K. (1996). MOL-MOL: a program for display and analysis of macro-molecular structures. J. Mol. Graph. 14, 51–55.

Edited by P. Wright

(Received 5 November 2002; received in revised form 31 January 2003; accepted 31 January 2003)


Documents

Quantitative NMR Studies of High Molecular Weight Proteins ... · Some structural implications of the chemical shift perturbations upon ... and an energetic challenge for weak bases