insight

nature structural biology • volume 5 number 10 • october 1998 891

1Division of Basic Sciences, Program in Structural Biology, Fred Hutchinson Cancer Research Center A3-023, 1100 Fairview Ave. N. Seattle Washington 98109, USA.2Department of Chemistry, University of California, Berkeley, California 94720, USA. 3Department of Molecular and Cell Biology, Stanley Hall, University of California,Berkeley, California 94720, USA.

Correspondence should be addressed to BLS. email: bstoddar@fred.fhcrc.org

The term ‘time-resolved’ crystallography usually refers mostspecifically to an experiment in which diffraction data are col-lected rapidly at a specific time point during a single reactionevent in the crystal. If reaction initiation is rapid and uniformthroughout the crystal, and if there is a discrete rate-limiting stepin the reaction pathway, a specific intermediate may transientlyaccumulate throughout most of the crystal active sites, and poly-chromatic Laue diffraction can be used to collect data during thelifetime of that species1–4. Using the most powerful synchrotronsources, useful exposure times currently may be as short as 120picoseconds5–7.

Rapid Laue studies have recently been reported on structuralintermediates formed during two separate photoreactionprocesses: CO-myoglobin photolysis and the photocycle of abacterial phototaxis protein (PYP)5,7,8. Apart from the inherentadvantage of having the photoexcitable molecule bound to theprotein at virtually 100% occupancy prior to reaction initiation,these studies were greatly facilitated by efficient triggering kinet-ics and by well characterized, separable spectroscopic signals foreach reaction intermediate9–12. This allowed nanosecond timeresolution of early events in these reactions. These studiesdemonstrated that polychromatic X-ray data may be collectedand processed to give very complete, highly redundant data sets,and that these data can be used to resolve extremely short-livedspecies. Additional time-resolved studies on both systems usinglow temperature trapping strategies have also been describedthat provide many additional details on the mechanism of thesereactions9,13–15.

Compared to time-resolved studies of photoreactions, mostenzyme catalysts offer several additional challenges that conspireagainst the accumulation of homogenous intermediate popula-tions and their visualization using single-turnover techniques.First, the photochemical rate and efficiency of many syntheticcaging groups are significantly lower than those of natural pro-tein-bound chromophores16,17. This is due in part to slow darkreactions that follow the actual photolytic event that can greatly

diminish the synchronicity of a turnover cycle throughout thecrystal. Second, while many cyclic or reversible photoreactionprocesses can be repeatedly or continuously ‘pumped’ by a lightsource (allowing the crystallographer to take multiple exposuresfrom a single crystal), similar experiments on enzyme catalysts,using irreversible release of photocaged substrates, usuallyinvolve single exposures for each crystal before and after thereaction initiation. This necessitates the use of multiple crystalsto produce complete data sets, and long enough exposure timesto ensure sufficient diffraction signal. Therefore, most singleturnover experiments on enzyme catalysts using the Lauemethod have targeted slow systems, with intermediate half-livesof seconds to minutes18–22. To date, millisecond studies of rate-limited enzyme species using single-turnover strategies have notbeen reported.

The enzyme isocitrate dehydrogenase provides an example ofthese challenges for time-resolved studies. This enzyme catalyzesthe oxidative decarboxylation of isocitrate to α-ketoglutarateand carbon dioxide, through formation of an oxalosuccinate(OSA) intermediate, and exhibits a catalytic mechanism withseveral rapid steps prior to the rate-limiting dissociation of prod-ucts (Fig. 1). The enzyme is fully reactive in the crystal andexhibits a kinetic mechanism similar to that in solution, asshown in kinetic experiments using a variety of substrate andstable intermediate species23. Under typical conditions (pH 7.5,22 oC) the enzyme turnover rate (60–70 sec-1) corresponds to ahalf-life for the bound product complex of approximately 10 ms,which can be extended to 40–50 ms by lowering either the tem-perature to 4 °C or the pH to 6.5. Initial structures of IDH apo-and phospho-enzyme, along with two stable binary substratecomplexes, have been solved by traditional monochromatic dif-fraction methods24–27. In order to accumulate the initial ES com-plex and the subsequent OSA intermediate, site-directedmutants have been used to impose specific kinetic barriers alongthe reaction coordinate which resulted in half-lives for thosespecies on the order of seconds. These mutants allowed steady-

Millisecond Laue structures of anenzyme–product complex using photocagedsubstrate analogsBarry L. Stoddard1,Bruce E. Cohen2, Michael Brubaker3, Andrew D. Mesecar3 and Daniel E. Koshland, Jr.3

The structure of a rate-limited product complex formed during a single initial round of turnover by isocitratedehydrogenase has been determined. Photolytic liberation of either caged substrate or caged cofactor and LaueX-ray data collection were used to visualize the complex, which has a minimum half-life of approximately 10milliseconds. The experiment was conducted with three different photoreactive compounds, each possessing aunique mechanism leading to the formation of the enzyme–substrate (ES) complex. Photoreaction efficiency andsubsequent substrate affinities and binding rates in the crystal are critical parameters for these experiments. Thestructure suggests that CO2 dissociation is a rapid event that may help drive product formation, and that smallconformational changes may contribute to slow product release.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

insight

892 nature structural biology • volume 5 number 10 • october 1998

state accumulation of the kinetically trapped species in the crys-tal to be combined with Laue data collection28 and moleculardynamics simulations29 to solve the structures of these interme-diates. More recently, cryo-trapping methods have beenemployed to study the factors that contribute to an efficienthydride transfer reaction30. This method also relied on trappingES intermediates with greatly extended half-lives.

Similar physical or chemical trapping strategies are not easilyexploited to visualize the enzyme product complex, because noreadily available techniques have been identified that significant-ly increase the lifetime of that species. Specifically, no proteinmutations are known that specifically reduce the rate of productrelease. Similarly, conducting steady-state reactions in the crystalat lower temperatures would require non-viscous cryobuffersthat are not well characterized kinetically, and further reductionsin pH would alter, if not change, the rate-limiting step of thereaction (A.D.M., unpublished results). We therefore have cho-sen a strategy that relies on a single-turnover Laue experimentwith wild type enzyme at 4 °C. In order to conduct this experi-ment, we have designed and implemented a series of experimen-tal photocaging strategies and used them in conjunction withLaue diffraction. Comparison of the relatively effectiveness ofeach strategy should be of general applicability for time-resolvedstudies on other enzyme systems.

Photoinitiation strategies and time-resolved studies To examine the effect of different photo-initiation strategies,three separate caged substrate molecules were synthesized (Fig. 2and Table 1)31,32 and used in these studies in independent experi-ments. The isocitrate substrate was modified to prevent binding

prior to photolysis, while the NADP cofactor was modified toeither allow binding but prevent hydride transfer, or to preventbinding altogether. The extent of photolysis and the fraction ofcaged substrate released was confirmed for each experiment bydissolution and HPLC analysis of irradiated crystals as describedin the Methods. Absorbance measurements of intact crystalsafter photolysis also indicated the formation of 10–20 mMNADPH in irradiated crystals, confirming the rapid photore-lease of free substrate from the caging group, followed by bind-ing and dehydration, the first catalytic step after formation of theES complex. However, the absorbance signal from NADPH doesnot allow us to easily discriminate between the next three species(5, 6 or 7) on the reaction pathway of the enzyme (Fig. 1), pre-venting us from determining the relative contribution of eachintermediate in the resulting structures by a method indepen-dent of the actual X-ray diffraction analysis. Non-uniformphotolysis, low occupancy binding, and/or asynchronous reac-tion progression through the crystal all may result in mixtures ofseveral species contributing to the resulting data and densitymaps. On the basis of the observed formation of NADPH in thecrystal after photolysis and the previously determined kineticprofile of the reaction28,33, we modeled each difference map as acomplex of α-ketoglutarate and NADPH, and performed aquantitative analysis of the agreement between the experimentaldata and that complex during the subsequent refinement. Thiswas done independently for each experiment to assess the effi-ciency of the triggering step and the quality of the resultingstructure (Table 2). The following parameters were quantifiedfor each experiment. (i) The signal strength in initial Fo - Fc andin subsequent simulated annealing (SA) omit difference maps

Fig. 1 a, General kinetic mechanism of isocitrate dehdyrogenase. Thechemical structures of isocitrate, oxalosuccinate, and α-ketoglutarate areindicated below the appropriate intermediates in the reaction. Theenzyme proceeds through a multistep reaction pathway, ultimately con-verting isocitrate and NAD(P) to α-ketoglutarate, carbon dioxide, andNAD(P)H. Random binding of substrate and cofactor (2,3) leads to for-mation of the initial ordered ES Michaelis complex (4). The nicotinamidering in this complex is structurally ordered primarily through electrostat-ic and van der Waals contacts with the bound isocitrate. Three subse-quent, sequential reaction steps proceed rapidly: reductivedehydrogenation to form oxalosuccinate and NADPH (5), decarboxyla-tion to form oxalosuccinate (6) and dissociation of CO2 to produce therate-limiting product complex (7). Intermediates 5, 6 and 7 all contributeto the absorbance signal from reduced NADPH cofactor. b, Overview ofthe IDH substrate-binding site.

a

b

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

insight

nature structural biology • volume 5 number 10 • october 1998 893

was measured and averaged over all modeled ligand atoms. Thestandard deviation was also calculated for this value over all mod-eled atoms for the bound ligand. (ii) The occupancy and thermalscattering B-factors were refined as single overall values for eachbound ligand. Experimental protocols that produced ambiguousdensity for the modeled product complex were observed to displaylow occupancies and elevated B-factors for the refined ligands. (iii)The maximum absolute value of residual difference density weremeasured for a volume covering the refined ligand atomic posi-tions and including an additional envelope extending 3 Å fromeach atom. For a properly modeled, well refined complex, onewould expect that this value should be quite low (approaching 1σcontour levels) and would be sensitive to either missing or poorlyoccupied atomic positions, or to the presence of additional electrondensity from alternate species, such as the initial isocitrate or oxalo-succinate species.

These parameters were found to be highly correlated, with spe-cific data sets displaying high- or low-quality values for each term(Table 2). As expected, the behavior of the bound α-ketoglutarateligand was most sensitive during modeling and refinement to vari-ations in experimental triggering strategies, because this species isthe primary site of chemical and structural homogeneity asturnover proceeds.

In order to maintain a high transmittance through the crystal,concentrations of caged substrates were limited to 50 mM in crystalsoaks and thin crystals were used (50–100 microns). Photolysis wasachieved using a light source partially monochromatized to 450nm, where the extinction coefficients of the photoreactive com-pounds are less than 0.1 mM–1 cm–1. Using previously establishedguidelines for light-triggered reaction initiation12,16, it is estimatedthat these conditions lead to at least 80% transmittance duringirradiation. All three caged compounds used for these experiments(Table 1) were subsequently shown to generate 10–20 mM releasedsubstrate in the crystal as a result of a 0.5 ms irradiation pulse. Therate of release during photolysis ranges from 10 s–1 (NPE-isocitrateat pH 8.0) to 13,000 s–1 (DMNPE-NADP at pH 7.0). After a 2 mslag, a single 10 ms X-ray exposure was taken. Multiple crystals wereused for each experiment as described in the Methods.

Once the conditions that support uniform photolysis and high

yield of photo-released substrate are established in the crystal,the remaining critical experimental questions are the actual rateof substrate photo-release, and the resulting substrate on-ratesand affinities in the crystal. Provided that the binding constant ofthe photoreleased substrate is in the physiological (micromolar)range in the crystal, diffusion and binding should be sufficientlyfast to support a uniform initial turnover cycle, based on bimol-ecular association rate constants of 105–106 M–1 s–1 and reactantconcentrations exceeding 10 mM. Not surprisingly, DMNPE-NADP (the most rapidly released caged compound) producedthe most cleanly interpretable and well-refined structure in theseexperiments (Table 2, Fig. 3). This compound allows prebindingof free isocitrate in the dark, and releases with a very fast decayrate of 13,000 s–1. The released NADP molecule binds with closeto full affinity in the presence of high sulfate, unlike the productof caged isocitrate, described below. The electron densitystrength averaged over all ligand atoms was greater than 5σ con-tour levels, and the final refined product complex gave refinedoccupancies and B-factors of 67% and 28.9 Å2 for α-ketoglu-tarate, and 72%, 26.1 Å2 for adenosyl-ribose (NADPH) respec-tively. The maximum absolute residue difference density for therefined substrate ligands are lowest for this structure (Table 2).

In contrast, the use of NPE-isocitrate at pH 6.5 and 8.0 consis-tently resulted in density maps that displayed a weak signal andpoor refinement statistics when the product complex was mod-eled (Table 2, Fig. 3). Upon further kinetic analysis of substratereactivity in the crystal, it was shown that caged isocitrate fails toefficiently induce formation of an ES complex for two differentreasons. At pH 8, the photolytic release rate is approximately10 s–1, which is no faster than the turnover rate of the enzyme at4 °C. At pH 6.5, the rate of photolysis is improved (~200 s–1), butat this pH in high sulfate crystal buffer the measured Km of isoci-trate increases from ~1 mM to >10 mM, which is approximatelythe concentration of isocitrate in the crystal produced by thephotoreaction. Weak binding degrades the experimental dataand the resulting experiment.

Results from the third caged molecule (CNB-NADP) demon-strate the use of a caging strategy that permits binding of sub-strate and cofactor prior to photolysis, but prevents formation of

a b c

Fig. 2 Chemical structures of the three caged substrate analogs used in this study. a, NPE-isocitrate (1-(2-nitrophenyl)ethyl-1-hydroxy-1,2-dicarboxy-3-propanecarboxylate). b, DMNPE-NADP (P2'-[1-(4,5-dimethoxy-2-ntrophenyl)-ethyl] NADP). c, CNB-NADP (N-(α-carboxy-2-nitrobenzyl)-NADP). Thephotoreleased blocking groups are outlined by dashed borders.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

insight

a productive ES complex due to steric interference. This method(‘catalytic’ caging, as opposed to ‘affinity’ caging)31 greatlyreduces the need for an extremely rapid phototrigger, by block-ing the nicotinamide ring of the cofactor. This modificationallows pre-binding of both the caged compound and isocitrateprior to photolysis, but prevents hydride transfer between thereactive groups. It was previously demonstrated for this systemthat the rate of hydride transfer is strongly dependent on theaverage approach distance and angle between cofactor and sub-strate30. Upon release of the blocking group (with a kphotol that isslightly pH independent, ranging from 30–40 s–1), rearrange-ment of the nicotinamide and ribose rings to an optimal orienta-tion for hydride transfer is estimated to require considerably lessthan 0.1 ms, based on published rates of bond rotation and com-putational simulations (B.L.S., unpublished results). The result-ing difference density maps from this compound, while not asclear as the very efficient DMNPE-NADP compound, are stillinterpretable (particularly at pH 6.5) and lead to refined ligandpositions that are well behaved (Table 2). One advantage of thissystem is that difference maps between dark and photolyzedcomplexes are reasonably clean, because all ligands are bound inthe dark prior to photolysis, maximizing the isomorphismbetween the structures and the data sets.

Structure of the product complexAs described above, the clearest difference maps, taken from theexperiment using DMNPE-NADP at pH 7.5, correspond tobound α-ketoglutarate and magnesium, and the adenosyl-riboseportion of the bound NADPH cofactor. The nicotinamide ringand phosphate backbone of NADPH are not visible. Mutagenicand cofactor perturbation studies on IDH indicate that the pri-mary energetic binding determinants for NADP(H) are between

894 nature structural biology • volume 5 number 10 • october 1998

the enzyme and atoms on the adenosyl-ribose groups, especiallythe 2' phosphate atoms, and that mutation of many of theseinteractions acts to reduce cofactor binding affinity and often toincrease the overall reaction rate (A.D.M., unpublished results).This is consistent both with the observed structure and with theformation of a rate-limiting product complex. One importantfeature of the electron maps is the absence of density at all stages of building and refinement for the labile β-carboxyl group of the substrate that is eliminated in the secondcatalytic step of the reaction. Based on these results, dissociationof this CO2 from the enzyme active site appears to be a rapid stepleading to the bound complex of α-ketoglutarate and NADPH.In contrast, previous steady-state Laue studies using a kineticmutant (K230M), which is known to catalyze decarboxylationvery inefficiently28, produced difference maps with clear positivedensity for the covalently bound β-carboxyl. Taken together, thisindicates that Laue studies can reliably discriminate between cat-alytic species that differ structurally by the presence or absence ofsmall functional groups. The rapid loss of CO2 from the IDHactive site may provide a thermodynamic driving force in thecrystal towards the product complex, even in this experimentalcase where substrate concentrations are rapidly depleted as aresult of a single turnover event. In solution, the generation ofCO2 may also drive the overall equilibrium of the reaction.

A small number of localized differences in the active site sidechains and backbone positions, and along the adjoining areas ofthe dimer interface are observed for the product complex. Thesemovements are not observed in the structures of IDH in binarycomplexes with isocitrate or with NADP alone25,27 or in ternaryES or EI complexes with isocitrate or oxalosuccinate28. Thisresult appears to be consistent with the rate-limit observed forproduct release, which may be coupled to small conformational

a

b

Fig. 3 Fo - Fc difference maps after initialprotein refinement against mergedLaue data, displayed over the sub-strate/product binding site. a, densitymap contoured at 1.2σ contour levelsfor caged isocitrate-triggered experi-ment at pH 6.5. The density for thebound substrate is not distinguishableabove background, due to poor bind-ing after photolysis as described in thetext. b, density map contoured at 3.5σcontour levels for DMNPE-NADP-trig-gered experiment at pH 7.5. This trig-ger is both a rapid phototrigger, andallows pre-binding of the isocitrate sub-strate prior to photolysis. The releasedNADP cofactor binds at full affinityunder crystallization buffer conditions,yielding strong signal in density maps.The initial density map and subsequentdifference maps show the absence ofthe eliminated β-carboxyl.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

insight

movements that provide a significant energy barrier toturnover. This result is also consistent with the observation thatmultiple turnover events in IDH crystals, or direct soaking ofproduct molecules into apo-enzyme crystals, cause significantdeterioration in diffraction strength28,34. The largest structuraldifference observed in the active site of IDH between the com-plexes of isocitrate and α-ketoglutarate involves Tyr 160, one oftwo residues involved in an interaction with the labile carboxylgroup of isocitrate which is lost upon conversion to α-ketoglu-tarate (Fig. 4). Upon hydride transfer, decarboxylation and for-mation of α-ketoglutarate, this tyrosine side chain and itsassociated backbone move by ~1 Å and approaches the non-bonded side-chain oxygen of Asp 307. This movement may beinvolved in the dissociation of α-ketoglutarate from theenzyme, by altering the affinity of the active site for the boundmetal–α-ketoglutarate complex. This interaction may alsoargue against Asp 307 acting as a catalytic base during hydridetransfer, a role that could be filled by a nearby bound watermolecule.

These studies indicate that a large number of experimentalvariables act together to produce a successful or a failed rapidtime-resolved experiment. The majority of these factors arerelated to the concentration of a given reactant generated by thephotoreaction, and the subsequent efficiency of ES complex for-mation. For this system, the primary obstacle to generating highconcentrations of released substrate or cofactor is the low quan-tum yield of their caged precursors (φ = 0.1–0.3), which areabout 1/2 to 1/5 of usually desirable values. However, two sepa-rate factors combine to offset this problem. First, the enzymeconcentration in the crystal lattice (which has a very high sol-vent content) is ~5 mM, which is quite low. Because the pho-tocaged molecules can be used at concentrations of up to 50mM in the crystal without interfering with uniform photolysis,even a final photoreaction yield of only 20% can yield a free sub-strate concentration that is approximately twice the concentra-tion of enzyme active sites, and (for NADP) ~1,000-fold abovethe Michaelis binding constant for the released molecule. Incomparison, light-triggered Laue studies using crystal systemswith low solvent content such as PYP (where the protein con-centration in the crystal lattice is >60 mM) are in this sensemore challenging to trigger, particularly when the chromophoreis covalently bound to an active site atom and therefore must be100% photolyzed in order to approach complete occupancy ofsubsequent photoreactive intermediates. Second, IDH crystalsare remarkably robust to irradiation using intense, short expo-sures from powerful excitation sources, allowing a large numberof photons to be directed on the sample during a 0.5 ms expo-sure. This serendipitous experimental factor also compensatesfor the low photoefficiency of these compounds. Finally, theobvious lesson from these studies is that, in the absence of extra-ordinarily rapid and efficient phototriggers, one should choose

nature structural biology • volume 5 number 10 • october 1998 895

to cage and release the reactive component that displays themost rapid binding kinetics and highest affinity under crystal-lization conditions (including, in the most ideal case, a catalyti-cally caged trigger with ‘perfect’ binding kinetics), whilepre-binding unmodified molecules that would perform poorlyas reaction triggers.

In summary, the use of phototriggered single turnover experi-ments can be used to visualize discrete catalytic states with half-lives in the millisecond range and perhaps less. When conductingsuch experiments, attention must be directed to several experi-mental considerations. Many subtle (and easily unrecognized)factors can undermine such an experiment, such as conflictsbetween the necessary limitations on concentrations of pho-totriggering agents in the crystal and the reduced substrate bind-ing affinities that are often seen under crystallization bufferconditions. Additionally, when such studies are performed onenzymes with multiple intermediates that cannot be easilyresolved spectroscopically, special care must be taken in modelbuilding and refinement. However, it is clear that there is a large(and growing) number of caged compounds now available forroutine use in such experiments (including nucleotides, metals,amino acids and a variety of other substrates containing primaryamines, hydroxyls, and free carboxyls)16,17,35, and that methods ofrapid data collection and/or flash-freezing may be used to collectdata of high quality for such studies.

MethodsOverproduction, purification and crystallization of wild type IDHwere conducted as described24. Caged compounds were synthesizedand purified as described31,32. The crystals were grown and main-tained in 30% saturated ammonium sulfate, and are identical tothose used in the determination of prior wild type complex struc-tures: space group P43212 with unit cell dimensions of a = b =105.1 Å and c = 150.3 Å (refined during data reduction). Thin crys-tals (0.5 mm × 0.5 mm × 50–100 microns) were soaked for one hourat pH 6.5 or 7.5 (100 mM Tris, 20 mM NaCl) in 50 mM substrate (isoc-itrate), cofactor (NADP) and magnesium. In each separate experi-ment one of these components (isocitrate or NADP) was included asa caged precursor, as shown in Table 1, while the others wereunmodified. Although higher concentrations of substrates (specifi-cally isocitrate) are desirable to promote efficient binding at highsulfate concentrations, in this system the photocaged triggers arelimited to no more than 50 mM to avoid overly strong absorbance,which would lead to concentration and thermal gradients acrossthe width of the crystal during photolysis.

All data were collected at beamline X26-C at NSLS (BrookhavenNational Laboratory)36. After mounting in glass capillaries undersafe light, crystals were cooled to 4 oC under a strong flow of chilledair, subjected to a 10 ms X-ray exposure in the dark, then excitedwith a single 0.5 ms pulse from a 1,000 W xenon arc lamp37, fol-lowed by a 2 ms lag and a second 10 ms polychromatic X-ray expo-sure. The total transient heating of the sample under theseconditions was measured to be less than 4 oC. The approximateusable X-ray spectrum was 0.5–1.6 Å. The light pulse and X-ray fastshutter were coordinated by an external control system using the

Fig. 4 Superimposed structures ofproduct complex (blue) and proteinside chains from the initial refine-ment model (isocitrate–NADP+ com-plex, red). Residues from monomer1 within 3.5 Å of the bound α-ketoglutarate are shown. Thelargest difference between the twostructures is seen at Tyr 160 asdescribed in the text.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

insight

flash output as a gate for electronic triggering of shutter rotation.The excitation pulse was focused to a 2 mm diameter cross sectionat the crystal and directed orthogonal to the plane of the thinnestcrystal edge. The pulse was partially monochromatized using a pri-mary bandpass filter (Oriel #58852; λ cutoff >550 nm) and a secondbroadband interference filter (Oriel #57530; 70 nm bandwidth, λmax

450 nm, peak transmittance 75%). The spectral energy density (irra-diated energy profile) of the excitation pulse was verified asdescribed37. The total photolytic energy onto the sample during theflash was estimated to be ~5 mJoules. For every crystal used, thepercent conversion of caged precursor to final product was verifiedby HPLC analysis of dissolved specimens after completion of theexperiment. Similar analyses on crystals that were X-rayed withoutphotolytic triggering were conducted as a control. Concentrationsof released substrates generated in the crystal by flash photolysiswere 10–22 mM, corresponding to 20–50% photoconversion of

896 nature structural biology • volume 5 number 10 • october 1998

starting material and well in excess of the enzyme concentration(5 mM) in the crystal lattice. No detectable thermal release wasmeasured in the absence of the light pulse for any crystal. The effectof pH and sulfate concentrations on Michaelis constants (Km) havebeen determined previously for isocitrate and NADP (B.L.S., unpub-lished results). Laue diffraction intensities were collected on a Fujiimaging plate system as described28. Data sets consisted of singleimages from six separate crystals separated by ~5° each, throughrange of 30°. Crystal morphology allows consistent pre-alignment ofspecimens to maximize data completeness and redundancy. A darkdata set was collected for each crystal in addition to the photolyzeddata sets; therefore final merged ‘dark’ and ‘light’ data sets(Table 2) collected under each experimental condition are derivedfrom multiple pairs of matched intensities in order to minimize sys-tematic errors between data sets. No significant radiation damagewas detected between pairs of exposures from individual crystals.

Table 1 Chemical structures and photolytic characterization of caged substrate and cofactor analogs

Compound kphotol (product release) t1/2 φ Dark binding?Caged isocitrate 234 s–1 (pH 6.0) 3 ms 0.3 No

(1-(2-nitrophenyl)ethyl-1-hydroxy-1,2-dicarboxy-3-propanecarboxylate) 10 s–1 (pH 8.0) 77 ms 0.3 No(NPE-isocitrate)

Affinity caged NADP 13,000 s–1 (pH 7.0) 0.05 ms 0.2 NoP2’-[1-(4,5-dimethoxy-2-ntrophenyl)-ethyl] NADP (DMNPE-NADP)

Catalytically caged NADP 42 s-1 (pH 6.0) 16 ms ~0.1 YesN-(α-carboxy-2-nitrobenzyl)-NADP(CNB-NADP) 30 s-1 (pH 7.0) 23 ms 0.1 Yes

Table 2 Crystallographic data collection, reduction and refinement (photoexcited data)

Caged substrate trigger NPE-Iso NPE-Iso DMNPE-NADP CNB-NADP CNB-NADPpH 6.5 8.0 7.5 6.5 7.5Resolution (Å) 2.45 2.5 2.5 2.6 2.45Internal merging and scaling

Singles:Reflections measured 395,801 351,749 360,784 314,349 365,521Unique reflections 23,434 22,074 21,751 20,316 22,898Rmerge (on ❘ F❘ ) 0.095 0.091 0.102 0.084 0.093Singles and multiples combined:Total final reflections 29,589 27,490 27,430 25,332 29,622Completeness (%, overall) 89.1 92.0 90.8 93.1 89.2Completeness (100–4 Å) 92.2 93.9 90.5 89.3 91.6Completeness (Highest 0.1 Å shell) 84.1 81.9 83.8 85.1 88.3Redundancy 12.3 13.1 12.9 11.9 13.7Average I/σ(I) 14.2 19.1 17.2 15.3 16.1Rdark/photoexcited 5.3 6.2 6.4 7.1 5.9

RefinementRcryst 21.6 23.2 22.2 23.1 23.0Rfree 27.7 28.4 27.1 28.5 29.6R.m.s.d. bond lengths (Å) 0.018 0.016 0.017 0.015 0.017R.m.s.d. bond angles (°) 2.7 2.4 2.8 2.5 2.8Avg. B-factor (protein) 22.5 21.9 24.0 23.6 23.9

Product ligand statistics1

Occupancy 0.21, 0.28 0.25, 0.29 0.67, 0.72 0.61, 0.69 0.43, 0.50B-factor (Å2) 54.2, 48.0 48.5, 49.1 28.9, 26.1 31.9, 28.6 47.2, 50.1Average SA omit map density

peak height (σ) 3.0, 3.9 3.8, 4.2 5.1, 5.3 4.8, 4.3 3.1, 3.6Maximum residual difference density (σ) 2.3, 1.4 2.6, 1.3 0.9,1.0 1.6, 1.3 1.5, 1.8

1Data statistics shown for merged data after photoexcitation (see the Methods) only. Statistics for dark data sets, collected prior to photolysis fromsame crystals, are similar. All data sets merged using exposures from six crystals as described in the Methods. Refinement values for bound productligands are from final refinements, shown for α-ketoglutarate and adenosyl/ribose respectively. All B-factors were restrained to single groupedmain-chain and side-chain values for each residue, while occupancies of bound product ligands were restrained to a single value for each molecule.Rcryst, Rfree = Σj ❘(Fo(j) - Fc(j))❘/Σj ❘ (Fo(j))❘; Rmerge = Σj ❘(F1(j) - F2(j))❘/Σj ❘ (F(j)).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

insight

Data were independently reduced and scaled with both the CCP4‘Daresbury’ Laue package38,39 and by the University of Chicago‘LaueView’ software package38,40. Laue reflections stimulated byseveral energies (harmonic overlaps) were successfully deconvolut-ed using LaueView and included in the final data sets, whichimproved the completeness at low resolution and the overall redun-dancy (Table 2). The overall integration and merging statistics wereotherwise similar between the data processing packages. The crys-tals diffract strongly to 2.5 Å resolution (I/σ(I) between 5.0 and 7.0,completeness greater than 80% from 2.6–2.5 Å), but the signal tonoise ratio decreases dramatically past at higher resolutions.

All data sets were refined against the photolyzed Laue data to2.5 Å resolution using X-PLOR41. This resolution limit represents theextent of diffraction of IDH crystals at 4 oC. Rfree was monitored atall stages of model building and refinement42. After initial refine-ments using protein coordinates only (starting model PDB number3ICD)24, product molecules (α-ketoglutarate and NADPH) werebuilt into Fo - Fc difference density calculated from simulatedannealing (SA) maps and the refinements were continued. For allrefinements, no density was visible for the nicotinamide ring or itsadjoining ribose, consistent with the release of those moieties froman ordered complex upon reduction of the cofactor27,28. For the

nature structural biology • volume 5 number 10 • october 1998 897

remainder of the refinement, the following four criteria were mon-itored to assess the quality of the time-resolved data: refined occu-pancies and thermal B-factors for the modeled ligands(adenosyl-ribose of NADPH and α-ketoglutarate/Mg2+), strength ofelectron density signal at the modeled atomic positions in SA omitmaps, and electron density gradients as a function of distance fromrefined ligand atom positions.

Coordinates. The coordinates have been deposited in the ProteinData Bank (accession code is 1b15).

AcknowledgmentsWe thank W. Scott, J. Bolduc, D. Dyer, M. Holmes, R. Strong, and K. Zhang foradvice and help with X-ray crystallography, R.M. Sweet, P. Singer, and G. Shea forassistance and technical support at NSLS Beamline X-26C (Brookhaven NationalLaboratories), D. Ringe and G. Petsko for extended use of their Xe flashlamp, and K. Moffat for extremely helpful advice and criticism at all stages of our time-resolvedstudies. B.L.S. is funded for this project by the NIH, DEK by the NSF and DOE.

Received 4 June, 1998; accepted 3 August, 1998.

1. Moffat, K. Laue Diffraction. Meth. Enz. 277, 433–447 (1997).2. Stoddard, B. L. Caught in a chemical trap. Nature Struct. Biol. 3, 907–909 (1996).3. Stoddard, B. L. Intermediate trapping and Laue X-ray diffraction: potential for

enzyme mechanism, dynamics, and inhibitor screening. Pharmacol. Thera. 70,215–256 (1996).

4. Moffat, K. & Ren, Z. Synchrotron radiation applications to macromolecularcrystallography. Curr. Opin. Struct. Biol. 7, 689–696 (1997).

5. Perman, B., et al. Energy transduction on the nanosecond time scale: earlystructural events in a xanthopsin photocycle. Science 279, 1946–1950 (1998).

6. Szebenyi, D. M. E. et al. Quantitative analysis of Laue diffraction patternsrecorded with a 120 picosecond exposure from an X-ray undulator. Trans. Am.Crystall. 24, 167–172 (1988).

7. Srajer, V. et al. Photolysis of the carbon monoxide complex of myoglobin:nanosecond time-resolved crystallography. Science 274, 1726–1729 (1996).

8. Genick, U. K. et al. Structure of a protein photocycle intermediate bymillisecond time-resolved crystallography. Science 275, 1471–1475 (1997).

9. Teng, T. Y., Srajer, V. & Moffat, K. Initial trajectory of carbon monoxide afterphotodissociation from myoglobin at cryogenic temperatures. Biochemistry 36,12087–12100 (1997).

10. Farber, G. K. Laue crystallography: Lights! Camera! Action! Curr. Biol. 7,R352–R354 (1997).

11. Imamoto, Y., Kataoka, M. & Tokunaga, F. Photoreaction cycle of photoactiveyellow protein from Ectothiorhodospira halophila studied by low-temperaturespectroscopy. Biochemistry 35, 14047–14053 (1996).

12. Ng, K., Getzoff, E. D. & Moffat, K. Optical studies of a bacterial photoreceptorprotein, photoactive yellow protein, in single crystals. Biochemistry 34, 879–889(1995).

13. Teng, T.-Y., Srajer, V. & Moffat, K. Photolysis-induced structural changes in singlecrystals of carbonmonoxy myoglobin at 40 K. Nature Struct. Biol. 1, 701–705(1994).

14. Genick, U. K., Soltis, S. M., Kuhn, P., Canestrelli, I. L. & Getzoff, E. D. Structure at0.85 Å resolution of an early protein photocycle intermediate. Nature 392,206–209 (1998).

15. Schlichting, I., Berendzen, J., Phillips, G. N. P. Jr. & Sweet, R. M. Crystal structureof photolysed carbonmonoxy-myoglobin. Nature 371, 808–812 (1994).

16. Schlichting, I. & Goody, R. S. Triggering methods in crystallographic enzymekinetics. Meth. Enz. 277, 467–490 (1997).

17. Corrie, J. E. T., Katayama, Y., Reid, G. P., Anson, M. & Trentham, D. R. Thedevelopment and application of photosensitive caged compounds to aid time-resolved structure determination of macromolecules. Phil. Trans. R. Soc. Lond. A340, 233–244 (1992).

18. Stoddard, B. L., Koenigs, P., Porter, N., Petratos, K., Petsko, G. A. & Ringe, D.Observation of the light-triggered binding of pyrone to chymotrypsin by LaueX-ray crystallography. Proc. Natl. Acad. Sci. USA 88, 5503–5507 (1991).

19. Schlichting, I. et al. Time-resolved X-ray crystallographic study of theconformational change in Ha-Ras p21 protein on GTP hydrolysis. Nature 345,309–315 (1990).

20. Singer, P. T., Smalas, A., Carty, R. P., Mangel, W. F. & Sweet, R. M. The hydrolyticwater molecule in trypsin, revealed by time-resolved Laue crystallography.Science 259, 669–673 (1993).

21. Duke, E. M. H., Wakatsuki, W., Hadfield, A. & Johnson, L. N. Laue andmonochromatic diffraction studies on catalysis in phosphorylase b crystals. Prot.Sci. 3, 1178–1196 (1994).

22. Hajdu, J. et al. Catalysis in the crystal: synchrotron radiation studies withglycogen phosphorylase b. EMBO J. 6, 539–546 (1987).

23. Stoddard, B. L. & Farber, G. K. Direct measurement of reactivity in the proteincrystal by steady-state kinetic studies. Structure 3, 991–996 (1995).

24. Hurley, J. H. et al. Structure of a bacterial enzyme regulated byphosphorylation, isocitrate dehdyrogenase. Proc. Natl. Acad. Sci. USA 86,8635–8639 (1989).

25. Hurley, J. H., Dean, A. M., Sohl, J. L., Koshland, D. E. & Stroud, R. M. Regulationof an enzyme by phosphorylation at the active site. Science 249, 1012–1016(1990).

26. Hurley, J. H., Dean, A. M., Thorsness, P. E., Koshland, K. E. & Stroud, R. M.Regulation of isocitrate dehydrogenase by phosphorylation involves no long-range conformational change in the free enzyme. J. Biol. Chem. 265, 3599–3602(1990).

27. Hurley, J. H., Dean, A. M., Koshland, D. E. & Stroud, R. M. Catalytic mechanism ofNADP-dependent isocitrate dehydrogenase: implications from the structures ofmagnesium-isocitrate and NADP complexes. Biochemistry 30, 8671–8678 (1991).

28. Bolduc, J. M. et al. Mutagenesis and Laue structures of enzyme intermediates:isocitrate dehydrogenase. Science 268, 1312–1318 (1995).

29. Stoddard, B. L., Dean, A. & Bash, P. A. Combining Laue diffraction and moleculardynamics to study enzyme intermediates. Nature Struct. Biol. 3, 590–595 (1996).

30. Mesecar, A. D., Stoddard, B. L. & D. E. Koshland, J. Orbital steering in thecatalytic power of enzymes: small structural changes with large catalyticconsequences. Science 277, 202–206 (1997).

31. Cohen, B. E., Stoddard, B. L. & D. E. Koshland, J. Caged NADP and NAD.Synthesis and characterization of functionally distinct caged compounds.Biochemistry 36, 9035–9044 (1997).

32. Brubaker, M. J., Dyer, D. H., Stoddard, B. & D. E. Koshland, J. Synthesis, kinetics,and structural studies of a photolabile caged isocitrate: a catalytic trigger forisocitrate dehydrogenase. Biochemistry 35, 2854–2864 (1996).

33. Dean, A. M. & Koshland, D. E. Kinetic mechanism of Escherichia coli isocitratedehydrogenase. Biochemistry 32, 9302–9309 (1993).

34. Stoddard, B. L. & D. E. Koshland, J. Structure of isocitrate dehydrogenase withα-ketoglutarate at 2.7 Å resolution: conformational chances induced bydecarboxylation of isocitrate. Biochemistry 32, 9317–9322 (1993).

35. Ottl, J., Gabriel, D. & Marriott, G. Preparation and Photoactivation Of CagedFluorophores and Caged Proteins Using a New Class Of Heterobifunctional,Photocleavable Cross-Linking Reagents. Bioconjugate Chem. 9, 143-151 (1998).

36. Getzoff, E. D., et al. Laue diffraction protein crystallography at the NationalSynchrotron Light Source. Nuclear Instr. Meth. Phys. Res. B 79, 249–255 (1993).

37. Rapp, G. & Guth, K. A low cost high intensity flash device for photolysisexperiments. Pflugers Arch (European J. of Phys.) 411, 200–203 (1988).

38. Clifton, I. J., Duke, E. M. H., Wakatsuki, S. & Ren, Z. Evaluation of Lauediffraction patterns. Meth. Enz. 277, 448–467 (1997).

39. Helliwell, J. R. et al. The recording and analysis of synchrotron X-radiation Lauediffraction photographs. J. Appl. Crystallogr. 22, 483–497 (1989).

40. Ren, Z., Ng, K., Borgstahl, G. E. O., Getzoff, E. D. & Moffat, K. Quantitativeanalysis of time-resolved Laue diffraction patterns. J. Appl. Crystallogr. 29,246–260 (1996).

41. Brünger, A. X-PLOR version 3.1: a system for X-ray crystallography and NMR(Yale Universuty Press, New Haven, Connecticut; 1992).

42. Brünger, A. Asessment of phase accuracy by cross validation: the free R value.Methods and applications. Acta Crystallogr. D 49, 24–36 (1993).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y