Authentic Enzyme Intermediates Captured “on-the-fly” by Mix-and-Inject Serial Crystallography

Ever since the first structure of an enzyme was solved, the discovery of the mechanism and dynamics of reactions catalyzed by biomolecules has been the key goal for the understanding of the molecular processes that drive life on earth at the atomic scale. Despite a large number of successful methods for trapping reaction intermediates, the direct observation of an ongoing reaction at runtime has been possible only in rare and exceptional cases. Here, we demonstrate a general method for capturing enzyme catalysis ‘in action’ by ‘mix-and-inject serial crystallography’. Specifically, we follow the catalytic reaction of the Mycobacterium tuberculosis β-lactamase with the 3rd generation antibiotic ceftriaxone by time-resolved serial femtosecond crystallography. The results reveal, in near atomic detail, antibiotic cleavage and inactivation on the millisecond to second time scales including the crossover from transition state kinetics to steady-state kinetics.

Synopsis. An enzymatically catalyzed reaction is initiated by diffusion based mixing of substrate and followed at runtime by TR-SFX at an XFEL.
Ever since the first structure of an enzyme was solved, the discovery of the mechanism and dynamics of reactions catalyzed by biomolecules has been the key goal for the understanding of the molecular processes that drive life on earth at the atomic scale. Despite a large number of successful methods for trapping reaction intermediates, the direct observation of an ongoing reaction at runtime has been possible only in rare and exceptional cases. Here, we demonstrate a general method for capturing enzyme catalysis 'in action' by 'mix-and-inject serial crystallography'. Specifically, we follow the catalytic reaction of the Mycobacterium tuberculosis -lactamase with the 3 rd generation antibiotic ceftriaxone by time-resolved serial femtosecond crystallography. The results reveal, in near atomic detail, antibiotic cleavage and inactivation on the millisecond to second time scales including the crossover from transition state kinetics to steady-state kinetics.
Observing the catalytic action of a biomolecule in atomic detail has been the dream of structural biologists since the first structure of an enzyme was solved (2,3). By exploiting X-ray radiation from powerful synchrotron sources, time-resolved crystallographic methods were developed (4) with the goal to achieve a complete description of a reaction in real time (5,6). However, X-ray damage and the need for large single crystals made time-resolved crystallography very challenging. The advent of X-ray Free Electron Lasers (XFELs) enabled time resolved serial femtosecond (fs) crystallography, where X-ray damage is outrun by ultrashort fs X-ray pulses (7,8). This approach made it possible to follow and describe cyclic (reversible) reactions that can be triggered by light. Examples include pioneering studies that investigate the photocycle in the photactive yellow protein (9,10), myoglobin (11), or photosystem II (12)(13)(14)(15). However, structural investigations on one-pathway (irreversible) enzymatic reactions present additional difficulties, because diffusion of substrate(s) and products in and out of the crystals limit the accessible reaction times. Standard crystallography can be used to track reaction intermediates of slow reaction by flash freezing but is unable to study enzymatic reactions at room temperature in real time. The largest problem, however, is to start a reaction in large sized crystals; initiation by diffusion is far slower than the typical millisecond turnover times of enzymes. It was proposed that one can trigger enzymatic reactions by light by soaking inactive (caged) substrates (16) into the crystals, which can be activated for example by a laser pulse. The first proof of concept for TR-Laue crystallography triggered by a caged substrate was achieved in 1990 (17). While this method has great potential, its application has so far been limited due to significant experimental challenges. Only a few time-resolved experiments have been reported where highly reactive, caged substrates are readily available (17)(18)(19), or the reactions are slow to allow the use of conventional, monochromatic methods (20, 21). It is therefore highly desirable to develop new methods that open the field of time-resolved crystallography to the study of biomolecular reactions at room temperature with the native enzyme and its natural substrate(s). Structural studies at XFELs provide a breakthrough.
The XFEL intensity is high enough to generate a diffraction pattern from an exposure to a single fs X-ray pulse even from micron and submicron sized crystals. These tiny crystals allow for fast diffusion times which are not rate limiting (22-27). The nanocrystals are mixed "on the fly" and injected into the XFEL beam, a method we call "mix-and-inject serial crystallography" (MISC) (24, 26). In MISC, crystals react with their native substrate(s) at ambient temperature until they are intercepted (probed) the by a single Xray pulse, which destroys them only after diffraction data has been recorded. The pulses are short enough to essentially outrun radiation damage by means of the "diffractionbefore-diffraction" principle (28-30). Optimized injectors have been recently developed (31, 32) for MISC experiments with the potential to provide submillisecond time resolution (33). The microcrystals tolerate even larger conformational changes leading to unit cell or even space group changes (12,27).
Here, we apply MISC to the study of a very important public-health problem: antibiotic resistance of bacteria. Specifically, we have obtained time-resolved crystallographic data on the binding and cleavage of the third-generation antibiotic ceftriaxone (CEF) in microcrystals of the enzyme β-lactamase from M. tuberculosis (BlaC). In these experiments carried out at the Linac Coherent Light Source (LCLS) BlaC micro-crystals are mixed with CEF on the fly, and the cleavage and thereby inactivation of the antibiotics by -lactamase is followed in runtime. BlaC is a broad-spectrum βlactamase which provides tuberculosis with resistance to all classes of β-lactam antibiotics. BlaC chemistry has rendered the frontline arsenal of antibacterial agents ineffective against this deadly disease, creating a global public health crisis. More generally, our approach is applicable to broad classes of important enzymes with the potential to fundamentally alter our understanding of the molecular basis of biomolecular reactions vital to the design of novel drugs.
Beginning with the famous discovery of penicillin, -lactam antibiotics were widely used to eliminate deadly infectious diseases (34). More compounds with diverse chemical composition were found through the years (35), the most prominent of them are most likely the cephalosporins. The chemical structure of CEF is shown in Scheme 1. Unlike the penicillins which feature a 5-membered thiazolidine ring, in the chephalosporins a 6membered dihydrothiazine ring is fused to the lactam ring (1). However, resistance against these antibiotics was observed shortly after their widespread use, and is now rampant. -lactamases open the -lactam ring rendering the antibiotic inactive. BlaC from M. tuberculosis, an Ambler class A -lactamase (36), uses a conserved serine to attack the -lactam ring (scheme 1, blue arrow) thus inactivating the antibiotics. Because of the great medical challenge that BlaC causes for the fight against infectious diseases the process of catalysis has been studied by conventional biochemcial methods in detail leading to the hypothesis of a three step model of the cleavage process: The first step is the formation of the enzyme-substrate complex (1), and it has been proposed that the enzyme may use active site interactions to orient the β-lactam carbonyl-carbon near the Ser-70 nucleophile (37, 38). The next step proposed along the reaction coordinate is the opening of the β-lactam ring subsequently or concurrently with formation of the covalently bound active site acyl-intermediate (3). For cephalosporins there is evidence that during the enzymatic reaction a release group (denoted R2 in Scheme 1) is split off (39). In the third step, the open-ring β-lactam is hydrolyzed and leaves the enzyme. Various rates have been reported for this step of the catalytic reaction across different classes of βlactams, followed by product release (37). Obtaining time-resolved data on BlaC chemistry holds the potential to directly visualize substrate chemical intermediates and the accompanying active site interactions, for wide-ranging implications for all classes of β-lactams. Ultimately, knowledge of the physical processes by which BlaC is able to bind and catalyze the breakdown of β-lactams, will directly impact rational drug design against deadly human diseases.
Our previous results showed that CEF can diffuse into the crystals and binds to the active site of the tetrameric -lactamase (26). These first studies showed that the catalytic reaction is heterogeneous as the reactivity is specific to individual subunits in the -lactamase tetramer. Only subunits B and D bind and process CEF, while subunits A and C do not directly contribute to catalysis. However, this first proof of concept study was limited to a single time point about 2 s after reaction initiation. Multiple time points that cover the reaction are required. Here we present time series from 30 ms to 2 s after mixing with substrate in two crystal forms, shards and needles, which allow us to discover the conformational changes and to characterize the kinetics of this important class of enzymes directly from the X-ray data.
One of the critical questions in MISC concerns whether the enzyme in the crystals is still catalytically active and whether the reaction is limited by constraints of crystal packing or the solvent/precipitant used for crystallization. We have therefore crystallized BlaC in two different crystal forms. With phosphate as precipitant, the BlaC crystallizes in a shard-shaped crystal form with a tetramer in the asymmetric unit (Fig. 1a) (Fig. 2 a,b,c). This is followed by the attack of Ser-70 which opens the -lactam ring. At the same time,the release group is split off, which leads to the formation of a covalently bound shorter ligand denoted CFO. There are subtle differences between the results from the two crystal forms, and even between subunits (Fig. 2), confirming previous preliminary observations (26). In both crystal forms, at 100 ms a substantial fraction (~70%, see also Tab. S2) of CEF molecules are still intact.
A minor fraction (~30%) has an open -lactam ring (Fig. 2d,e,f). The open species CFO can be identified more clearly at 500 ms, where it dominates (70%) the electron density. This confirms, for the first time on a structural basis, previous predictions from biochemical results for other cephalosporin species (39). The red arrow in Fig. 2g indicates that the double bond  (Scheme 1) has reacted to an alcohol in subunit B. This behavior is not seen in subunit D, nor in the needle form of the crystals. At 2 s, the structures reveal the steady state of the enzyme, where the binding sites are occupied mainly by the full length CEF with minor contribution from CFO (< 20%) in the shards.
In the shards crystal from, subunits A and C do not directly participate in catalysis, at least not in the first 2 s. In the needles it appears that the reaction proceeds similarly to that observed in subunit D in the shards. However, substrate occupancy is lower compared to that in the shards, with substoichiometric occupation ranging from 20 % -40 %. The reason for this might be that the enzyme concentration in the needles is very high, 30 mmol/L (~ 940 mg/mL). In the shards it is only 16 mmol/L (~ 510 mg/mL). To reach full occupation in the needles, obviously at least 30 mmol/L of CEF (one CEF molecule per asymmetric unit) is initially required, which needs to be delivered by diffusion from the solution to the side of the crystal. While the outside CEF concentration is on the order of 200 mmol/L in both experiments, the ratio of CEF to enzyme varies in the shards and needles crystals forms. Fig. S3 shows how the solvent volume that contains CEF surrounding the BlaC molecules in the crystals varies. Where it is on the order of 65% for the shards it is substantially lower (38%) in the needles. Fig. S3 also shows that there are substantial differences in the solvent channel sizes in the two crystal forms. Both may significantly impact diffusion of substrate into the crystals. However, CEF is a slow binder (see discussion in the SM). The reaction initiation does not critically depend on the diffusion time of the substrate (Fig. 3). Accordingly, the reaction dynamics of the catalytic reaction in the needles and the shards crystal forms are similar.
An additional CEF molecule (CEF stack ) can be identified near the catalytic clefts of subunits B and D, each, in the shards crystal form (Fig. 1a,b,c). This molecule stacks to the CEF species that occupy the active sites on all time scales. CEF stack is non-covalently attached to Arg-126 and Tyr-127 of the subunits A or C, which are adjacent to the active catalytic clefts of subunits B or D, respectively (more details are listed in the SM). This way CEF stack is pre-oriented, and can rapidly access the active site after CFO has been hydrolyzed and left the enzyme. Since stacking is not observed in the monomeric needle crystal form, it might be argued that it represents a non-physiological, nonspecifically bound substrate that occurs only in BlaC dimers and that dimers exist only in crystals.
Previous studies showed that BlaC can crystallize in monomeric form (38) as in our needles. Others (1) report crystal forms with a tetramer (dimer of dimers) in the asymmetric unit as in our shards. Dynamic light scattering ( The interesting question of the physiological oligomeric state of the BlaC warrants further investigation. The binding of the additional CEF molecule could be an important mechanism to steer the substrate towards, and orient it with respect to, the active site. It appears that at the very high concentrations of CEF applied here, stacking is not required for effective catalysis, as the kinetics in the monomeric needles, where stacking does not occur, is similar to that in the tetrameric shard crystal form. However, when only small CEF concentrations are present, stacking might well be essential to recruit antibiotic substrate molecules to promote effective BlaC function. One of the major questions addressed here is whether the structural data obtained by MISC can be interpreted in accordance with previous investigations on BlaC catalysis. We then compared the simulations to our MISC experiment. Since only 4 time delays are available, the parameters in the mechanism can not be determined fully quantitatively, but the simulations reproduce the experimental observations. After initial formation of the enzyme-substrate (ES) complex represented by a non-covalently bound full length CEF, an enzyme-product species (EP) with a covalently bound CFO has its peak concentration at 500 ms. It has been previously suggested (38) that the hydrolytic cleavage of an acyl-adduct from Ser-70 (hydrolysis of species 4 in scheme 1) should be, if any, the rate limiting process of BlaC catalysis. Then the EP species should be the dominant species in the steady state (2 s). However, this is not the case as the ES complex with the non-covalently bound, full length CEF is prevalent (> 70%) in our MISC data at 2 s. This can be explained by the simulation, if the nucleophilic attack of Ser-70 on species (2) in scheme 1 is inhibited, or slowed down. Since high product concentrations of > 10 mmol/L are reached already after one catalytic cycle due to the very high enzyme and substrate concentrations, product inhibition is plausible. Its structural mechanism, however, remains unknown (see also the SM). After an initial burst over the first second, the nucleophilic attack by Ser-70 becomes the rate-limiting process, and the ES complex accumulates in the steady state as observed in our X-ray data.
Our results show that structural characterization of enzymatically catalyzed reactions on the millisecond time scale is now possible. With more conventional X-ray sources, radiation damage prevents the collection of even a single diffraction pattern (43) from these small crystals. These difficulties are circumvented by the ultra-short, brilliant hard X-ray pulses available at XFELs. MISC (26) can now be employed to investigate a large number of non-cyclic (single pass) reactions in proteins and enzymes, some of which are of immense biological importance, and might be, in addition, important targets for structure based drug design. With MHz X-ray pulse rates expected at LCLS-II and the European XFEL, multiple, finely spaced time delays may be collected rapidly to allow for a comprehensive description of the biomolecular reaction in terms of structure and kinetics. In the future, with further advances and more XFELs worldwide, enzymology might become predominantly structure based.

Acknowledgements.
This work was supported the NSF-STC "BioXFEL" through award STC-1231306, and in         Accordingly, two different strategies were followed to analyze the two types of data.   sufficiently to be detected. Since the ES complex reappears in our MISC data in the steady state on times > 1 s, product inhibition of rate coefficient k2 was assumed by lowering k2 to zero: = (1 − ), with Pn the concentration of the released product P divided by 1 mmol/L. Since the concentration of enzyme in the crystal is very high, even only one catalytic cycle produces on the order of 10 mmol/L product (Fig. 4, red line) which may strongly inhibit the reaction. This might be the first evidence for product inhibition of the BlaC reaction which awaits further investigations which are outside the scope of this paper.