Time-resolved Crystallography

Observation of Unstable Species in Enzyme-Catalyzed Transformations Using Protein Crystallography

Recent advances in rapid X-ray diffraction data collection methods, cryocrystallography, and other techniques have made it possible to visualize short-lived species in enzyme-catalyzed reactions directly at atomic resolution for a significant number of crystalline enzymes.  The wide range in reaction types, intermediate lifetimes, and crystal characteristics means that different methods must be employed for each case, but there are enough examples now of successful structure determinations of normally unstable species to suggest guidelines for future investigations, which we are actively pursuing. 

Protein structures determined by X-ray diffraction - or, more precisely, the electron densities from which those structures are derived - are time-and-space averages.  Spatial averaging is over all of the trillions of molecules in the crystal; temporal averaging is over the time required to collect the X-ray diffraction data.  Both effects have an impact on attempts to use the power of X-ray crystallography to visualize discrete steps in an enzyme-catalyzed reaction at atomic resolution.  For unambiguous interpretation of the experimentally-observed  electron density, that density must come from a single species at high occupancy (the noise level in an electron density map at typical resolutions is about 25% of its average value).  Averaging over all molecules in the crystal means that this condition will only be met if the catalytic reaction is synchronized throughout the entire crystal or if the experimental parameters have been chosen so that a discrete intermediate accumulates.  Ringe has shown that even if the reaction can be initiated synchronously a distribution of species will quickly develop, so the better strategy is to adjust the temperature, pH and other variables in the experiment so that a particular complex becomes metastable.  Time-averaging means that the lifetime of this complex must exceed the X-ray data collection time or else mixtures of species will form.  Even with synchrotron sources, high-resolution data collection with monochromatic radiation takes many minutes, so intermediates must be stabilized for at least that long .  Use of polychromatic synchrotron radiation (the Laue method) can reduce data collection times to milliseconds or less, but this technique imposes severe restrictions on crystal quality.  Many protein crystals that are perfectly suitable for monochromatic data collection turn out to be too mosaic for the Laue method.  Other complications include the serious radiation damage that the white X-ray beam induces, even in frozen crystals.  For these reasons, most recent successful time-resolved structures have been carried out with monochromatic X-rays, and the key issue in each case is the technique used to insure that the desired intermediate accumulated in the crystal to high occupancy for minutes or longer.

Over the past twenty years or so a variety of techniques have been developed, mostly in our lab, to enable normally unstable reaction intermediates to build up and be stabilized in protein  crystals. From these examples one can choose the method needed for any particular case.  Although the number of crystalline enzymes where all of the intermediates can be observed is probably small, it seems likely that at least one such species can be observed for a large number of them.  As our work has shown, looking at productive substrate complexes, as opposed to inhibitor complexes, can be very informative about catalytic mechanism.

Working at far from optimum catalytic conditions

In favorable situations the pH-dependence of an enzyme-catalyzed reaction can be such that the turnover time of a particular intermediate is very long at pH values several units away from the optimum.  For such enzymes it may be possible simply to incubate the crystal with substrate at this extreme pH: the intermediate will accumulate but its breakdown will be long compared with time required to collect the diffraction data.  It is often necessary to use low-temperatures as well as non-optimal pH to insure that the intermediate does not turn over during that time.


The most commonly-used method for stabilization of reaction intermediates in protein crystals is to generate the species at a temperature at which it forms relatively rapidly and then rapidly cool the crystal.  If the Arrhenius rate equation holds over all temperature ranges it can be used to calculate how cold the crystal needs to be for any desired reaction time, provided the activation enthalpy is known.  However, a “glass transition” exists in the dynamical properties of all proteins at around 200K; below this temperature the nonharmonic, collective motions are “frozen out”.  Since such motions appear to be essential for many, if not all, enzyme reactions to proceed from substrate to product , and also for specific substrate binding and product release, it is only necessary to flash-freeze the crystal to below this temperature in order to trap whatever species was generated, essentially indefinitely.  Most protein crystals can be flash-frozen to about 100K - in fact, such freezing is routinely employed to reduce radiation damage – so it is often not difficult to try this method of trapping intermediates.  Figure 1 shows an example from our work on elastase.

Of special value is the use of microspectrophotometry to characterize the states formed in the crystals before any X-ray diffraction studies are attempted.  Such careful characterization of the reaction under the conditions of structure determination is the best way to establish that a single species has been generated, to prove that it is on the pathway between reactants and products, and to provide independent evidence for its chemical identity. 

Rapid Data Collection Techniques

Often an enzyme-catalyzed reaction will have several discrete steps that one wishes to observe.  While stabilizing them separately by the techniques described here is usually the easiest way to obtain structures for them, use of very fast data collection techniques such as Laue diffraction or fast monochromatic X-ray data acquisition can, in some cases, allow direct observation of all the kinetically-significant species in an enzyme reaction cycle.  The first example of this was our work on cytochrome P450 (Figure 2) in 2000 (Schlichting et al., Science).  It involved combined use of many of the methods discussed here to spread the time-course of the reaction out such that time-resolved data collection was possible.

Exclusion of water

For hydrolytic enzymes a clever variation on the mutagenesis strategy would be to mutate water itself, to a species that is not capable of functioning as a nucleophile.  The observation of Klibanov and Ringe that many crystalline enzymes can be transferred to neat organic solvents without disordering opens up the possibility for such substitution.  Farber and coworkers were able to use g-chymotrypsin crystals transferred to hexane to trap an enzyme-product complex and a putative terahedral intermediate.  More recently, Schmitke, Stern and Klibanov have been able to observe the acyl-enzyme intermediate in the subtilisin Carlsberg-catalyzed hydrolysis of a trans-cinnamoylimidazole substrate by transferring the crystals to anhydrous acetonitrile.  This approach has great promise for many enzymes in view of the widespread role of water as a nucleophile in biology. 

Direct observation of Michaelis complexes under equilibrium turnover conditions 

The earliest successful strategy for the direct observation of an enzyme-substrate complex was that employed by us for triosephosphate isomerase.  It makes use of the fact that if an enzyme catalyzes a simple single-substrate/single-product equilibration, all one has to do to observe the Michaelis complex is to soak the crystal in mother liquor containing a substrate concentration significantly greater (about 10 x) than Km.  Substrate will bind ot the active site and be converted to product, but product is just the substrate for the back reaction, so the enzyme will settle to equilibrium and a crystal structure determination will show E-S, or E-P or a mixture of the two, whatever is the lowest free-energy species.  The most thorough exploitation of this strategy has been in our studies of xylose isomerase, where it has been used to show that the mechanism of the enzyme is not the previously-assumed proton transfer mechanism but rather a hydride transfer pathway. Any isomerase, racemase, mutase, or epimerase can be studied this way, and in some cases other enzymes can also make use of this method if conditions can be found that cause the reaction to be effectively single-substrate/single-product. 

Establishment of pseudo steady-state conditions 

As first pointed out by us in Farber et al., another method for trapping intermediates makes use of the fact that, under steady-state conditions, the species prior to the rate-determining transition state in an enzymic reaction will tend to accumulate.  The rationale is simple: if product can be removed and there is enough substrate around to maintain saturation, then the enzyme will be in a substrate-bound form on average.  In a protein crystallographic investigation, this can be done most easily by using a flow cell, which allows fresh substrate to be flowed into the crystal while any product that is formed is washed away.  The most dramatic example of the power of this method is in the recent and beautiful work of Anderson and Schlichting and associates on tryptophan synthase.  Wild-type enzyme crystals were used, and the beta site substrate serine was flowed in under pseudo-steady-state conditions (the inhibitor 5-fluoroindole propanol phosphate was used to block the alpha site).  The structure clearly showed aminoacrylate intermediate of the beta reaction forming a Schiff base to the active site PLP cofactor.  Comparison of this structure with previous ones has allowed the pathway for communication between the alpha and beta active sites in this bifunctional enzyme to be established. 

Direct binding of product 

Considering that product is often a good competitive inhibitor for many enzymes, it is surprising that more investigators have not utilized the simple but very effective strategy of just soaking the enzyme crystals in a concentration of product 10-fold or more higher than its Ki.  After all, many phosphate-utilizing enzymes are found to have phosphate ion bound tot he active site naturally when their structures are determined, suggesting that it should be easy to do this, and when it has been employed this approach has worked very well. 

Use of substrate analogs 

Finally, a word must be said about the use of inhibitors and substrate analogs.  By definition, such molecules cannot show all of the interactions and chemical transformations expected for the actual substrate, but when they function as mechanism-based inactivators they can sometimes be used to trap an intermediate or a close analog of one. Fluorinated substrates are often useful for this kind of study.   The danger in such approaches is that analogous compounds do not always bind analogously, so it is imperative that mechanistic conclusions obtained in this way (or any way, for that matter) be consistent with data obtained by other methods such as kinetics, chemical modification, and mutagenesis.


figure 1figure 2The preceeding examples illustrate the various tools that have become available for trapping normally unstable species in enzyme-catalyzed reactions.  Data collection times have become sufficiently rapid that it is now only necessary to find conditions that will allow the desired species to accumulate in the crystal lattice and be stable for seconds to minutes.  If the lifetime is seconds, then some specific method of rapidly triggering the reaction must be found: light and gas flow are the most commonly employed.  If the lifetime is longer, then simple diffusion of substrate into the crystal may be sufficient to initiate the reaction.  Flash-freezing is probably the most versatile method for stabilizing the desired intermediate, but any set of conditions that prevents its breakdown may be used.   Since these conditions may include the conformational restrictions that can be imposed by the packing of adjacent molecules in the crystal lattice, one is tempted to offer the following advice: if no other strategy appears possible, we recommend simply soaking substrate into crystals of the active enzyme and seeing what happens.  It isn’t the most careful approach, but it may be fruitful more often than expected.

time resolved figure 2