CTX-M-14 microcrystals offer perfect conditions for mix-and-diffuse experiments

The TapeDrive system26,27,37 was applied to collect serial diffraction data at the beamline P11, PETRA III/DESY, to explore the kinetics and structural intermediates of ligand binding to the -lactamase CTX-M-14. As a result, protein structures with delay times of 5010,000ms and a resolution range of 1.402.04 were obtained. For this purpose, CTX-M-14 microcrystals were mixed with boric acid to initiate the binding process, and diffraction data were collected after distinct pre-set delay times. To monitor the formation of a diester, microcrystals were pre-soaked with boric acid and subsequently mixed with glycerol and again diffraction data were collected after distinct delay times. The obtained data can reveal the time evolution of populations and, as for all mix-and-diffuse serial crystallography data collections, can represent multiple states in one structure. The delay times, the corresponding PDB entries, the obtained diffraction quality and model refinement statistics are summarized in Supplementary Tables1 and 2. In our own unpublished experiments, macro-crystals of CTX-M-14 were soaked with boric acid and diffraction data were collected by conventional rotation crystallography at cryo-conditions. Glycerol was used as a cryo-protectant, and thus the cyclic glycerol boric acid diester (GBE) in the active site described here has been observed (PDB code 8r7m). However, a time-resolved analysis of the processes seemed very intriguing due to the two sequential reactions. To observe the reactions via time-resolved crystallography applying the available TapeDrive setup of CFEL at PETRA III, DESY, the reaction rates needed to be decreased. In terms of pilot investigations, we observed that kcat is reduced approximately twofold at pH 4.5 compared to pH 7.4. Therefore, the relatively low pH 4.5 applied for the crystallization conditions supported the optimization of time-resolved diffraction data collection of CTX-M-14, although it does not correspond to the physiological pH value. The asymmetric unit of CTX-M-14 crystals contains one monomer with the active site region solvent accessible. The Matthews coefficient of the crystals is 2.153/Da, corresponding to a solvent content of 43%. The solvent channels in the crystal lattice allow rapid diffusion of low molecular weight ligands to the active site, scoring the CTX-M-14 crystals to be ideal for time-resolved serial crystallographic investigations, applying the TapeDrive mixing approach27. Furthermore, the small but excellent diffracting crystals of CTX-M-14 with dimensions of 1115m have a relatively small ligand distribution period within the crystal lattice and due to short diffusion times exhibit sharper delay time points compared to larger crystals38. As a reference, the TapeDrive was also used to collect serial data of the native CTX-M-14 crystals. The occupancies of boric acid and its glycerol diester in the active site obtained after different mixing time points were refined and compared to discuss the stepwise rearrangements in the active site in detail below. To avoid correlation of occupancies and B-factors during refinement, care was taken that between datasets of adjacent time points the individual B-factors of the respective ligands did not differ more than the Wilson B-factors and the average B-factors (Supplementary Fig.2).

CTX-M-14 has a crucial anion-binding site (Fig.1) close to the active site residues that is occupied by the carboxylate of -lactam substrates39. In the native enzyme, this site is occupied by a tetrahedral anion, such as a phosphate (PDB code 4ua640) or a sulfate (PDB code 7q0z13), as in the structures we refined (Figs.1 and 2). In a structure of CTX-M-14 in complex with ixazomib/bortezomib (PDB code 7q11/7q0y13), the inhibitor does not directly occupy the anion-binding site, but still displaces the bulky tetrahedral anion, which is replaced by a smaller chloride to balance the charge13.

a Cartoon plot of the CTX-M-14 -lactamase from Klebsiella pneumoniae and close-up views of the active site in surface representation of b the native constitution with a sulfate ion (SO4-A magenta; SO4-B yellow) in the anion-binding site, c with bound boric acid (BAB, pink) and a sulfate ion, and d with bound glycerol boric acid diester (GBE, pale green). BAB (c) and GBE (d) complex structures are shown with mixing delay times of 10s, respectively.

Stick and cartoon representation (left) as well as 2D-LigPlot+ representation (right) of the active-site amino acid residues highlighting the hydrogen bond network in the native form (a), with bound boric acid (10s, BAB) to Ser70 (b) and with the bound glycerol boric acid diester (10s, GBE) (c). Each equilibrium state is displayed individually without overlapping with the initial states. BAB and GBE oxygen atoms are labeled in red. Potential hydrogen bond distances () are indicated by dashed lines.

The rotationally disordered sulfate occupies two slightly displaced alternative positions in the native enzyme (SO4-A and SO4-B, S to S distance of 0.4, Fig.1). The alternative sulfates (A/B) are coordinated via hydrogen bonds with the side chains of Ser70 (3.1/3.4), Thr235 (3.1/3.3) and Ser237 (3.4/2.8) as well as the main chain nitrogen of Ser237 (3.1/3.2) (Fig.2a). SO4-B forms additional hydrogen bonds with side chains of Ser130 (2.9) and Lys234 (3.2) (Fig.2a). During boric acid binding, the sulfate is reoriented in such a way that it is more distant to the Ser70 side chain and coordinated by hydrogen bonds with the side chain hydroxyl groups of Ser130 (3.1), Thr235 (3.0) and Ser237 (2.9) (Figs.2b and3). In addition, the boric acid O2 (2.6) can act as a hydrogen bond donor for the sulfate. The esterification with glycerol finally displaces the sulfate ion, since the equivalent O2 of the cyclic diester cannot act as a hydrogen bond donor anymore and due to steric competition (Fig.2c). In the electron density maps substantially reduced density is observed at this site (Fig.4). After complete formation of the cyclic diester with boric acid and glycerol, a water molecule (OW357) occupies the position of the anion-binding site. Unlike the sulfate ion, the OW357 can act as a hydrogen bond donor and forms a hydrogen bond with O2 of GBE (2.7). The water molecule OW357 is further stabilized by a hydrogen bond with the hydroxyl group of Thr235 (2.8), as well as a weak hydrogen bond with the main chain carbonyl of Thr235 (3.5).

Polder electron density (contoured at 5 , green mesh) of the active site Ser70, the sulfate ions and bound boric acid are shown at different delay time points after mixing microcrystals with boric acid. The 1h soak structure was obtained with the TapeDrive after microcrystals have been soaked in boric acid for 1h and shows that almost no further increase in electron density is observed after 10s. BAB and GBE oxygen atoms are labeled in red. Potential hydrogen bond distances () are indicated by dashed lines.

Polder electron densities (contoured at 5 , green mesh) of the active site region of CTX-M-14. Microcrystals pre-soaked with boric acid and mixed with glycerol prior to serial diffraction data collection applying the TapeDrive setup at beamline P11, PETRA III/DESY, observing time-resolved the ester bond
formation between glycerol and the Ser70 borate ester. The sulfate anion present in the native conformation is displaced upon binding of GBE and finally replaced by solvent water OW357. BAB and GBE oxygen atoms are labeled in red. Potential hydrogen bond distances () are indicated by dashed lines.

A direct comparison with recently approved inhibitors such as relebactam (PDB code 6qw841) and avibactam (PDB code 6gth42) also shows that utilization of the anion-binding site supports the complex formation. These complexes are stabilized by hydrogen bonds and consequently, the affinity and overall activity of these inhibitors are increased. Accordingly, the sulfonate groups of these new diazabicyclooctane inhibitors occupy the anion-binding site discussed here (Supplementary Fig.3)13. In addition, vaborbactam (PDB code 6v7h43) and taniborbactam (PDB code 6sp612) are bound and coordinated in the active site in a similar way. The carboxylate appendage of their oxaborine or benzooxaborine moieties also occupies the anion-binding site10,12,43. Thus, for inhibition of SBLs, it is evident that the anion-binding site of the native enzyme is occupied by the inhibitor, supporting enhanced binding if inhibitors feature a suitable moiety that can bind in this region (Supplementary Fig.3). This anion-binding site represents a very important structural feature of -lactamases, to be considered in future drug development investigations. In this context, our data are unique, as we show via time-resolved crystallography the time course of the displacement of a sulfate ion from this particularly important binding site.

In addition, the oxyanion hole is utilized by a number of inhibitors forming hydrogen bonds with Ser70 NH and Ser237 NH (see Supplementary Fig.4). Furthermore, these structural features are also used in the binding modes of -lactam substrates such as ceftazidime (Supplementary Fig.4h), as well as in multiple other -lactamases.

The obtained refined time-resolved crystal structures provided insight into the molecular kinetics of the binding of boric acid (Fig.3 and Supplementary Fig.5). Starting from the native CTX-M-14 structure, the above-mentioned sulfate and some water molecules (notably OW174, OW352, OW353 and the catalytic OW10) are present in the active site well-defined in the electron density maps. At a delay time of 50ms after mixing the microcrystals with boric acid initially, no additional electron density for the boric acid was observed. Meanwhile, the electron density of the sulfate ion has already changed, indicating a slight shift between the two alternative locations. Initially, in the native enzyme, the alternative positioned sulfate ions refined to occupancies of 47% and 44% for SO4-A and SO4-B, respectively. These change in the 50ms structure to occupancies of 54% for the SO4-A and 41% for the SO4-B position also indicate that initially the position closer to the Ser70 is preferred before the boric acid will covalently bind to Ser70 OG. After a delay time of 80ms, a weak electron density for the bound boric acid (BAB) was observed in the calculated polder map, with a corresponding occupancy of 35%. At the same time, the sulfate ion in the SO4-B position was reoriented by slight translation and rotation so that an oxygen atom has a distance of 2.6 to the O2 hydroxyl group of BAB (Fig.3). The evaluation of the electron density maps revealed that the hydroxyl groups of BAB occupy approximately the same positions as previously occupied by an oxygen of SO4-A and the two water molecules OW352 and OW353. The calculated occupancy for BAB (Fig.5a, Supplementary Table3) and the corresponding electron density increased with longer delay times after mixing, resulting in a well-defined electron density for BAB in the calculated polder map after only 250ms delay time. At this delay time, the occupancy of BAB is already 49%, whereas the occupancy of SO4-A has dropped to 33%. In the further time course investigated, the occupancy of BAB increases only slightly. After a delay time of 10s, it reaches the maximum occupancy of 53%. Even soaking the CTX-M-14 microcrystals in boric acid for 1h could only increase the occupancy to 57%. This indicates that under these conditions the equilibrium of the BAB formation has been reached.

Plots of BAB (a) and GBE (b) with the refined occupancy values obtained in the context of the respective delay times (no linear display), after mixing with boric acid (BA) or glycerol (GOL). The occupancy of BAB increases with prolonged delay time after mixing with boric acid. Subsequent mixing with glycerol causes the BAB occupancy to decrease again, as it is esterified to GBE. The total boron content continues to increase along mixing with glycerol.

Boric acid binds to the active site of CTX-M-14 (Fig.1c) forming an ester with the Ser70 OG. The hydrogen bonding interactions that stabilize the tetrahedral transition state analog during initial binding include the oxyanion hole (Ser70 NH and Ser237 NH). Similar to the binding mechanism of substrates, the nucleophilic attack of Ser70 OG can be supported via activation of the OG by the general base Lys7344. The unprotonated Lys73 side chain can assist in the nucleophilic attack by acting as a general base thereby accepting the proton from the Ser70 OG when the tetrahedral intermediate is formed. A corresponding proposed reaction pathway is shown in Fig.6. Similar to the carboxylate of the acylenzyme intermediate, one hydroxyl group of boric acid (O1) forms hydrogen bonds with the main chain nitrogen atoms of Ser70 (2.8) and Ser237 (2.8), constituting the oxyanion hole (Fig.2b). In contrast to bortezomib and ixazomib, the remaining two hydroxyl groups of BAB do not form hydrogen bonds with Asn170 and Glu16613 (Supplementary Fig.4). In fact, the boric acid is shifted rather in the opposite direction in the anion-binding site, forcing a reorientation of the sulfate ion from the position of SO4-A to the position of SO4-B (Fig.2b), to prevent too close atomic contacts. The boric acid is further stabilized in this position via hydrogen bond interactions of the BAB hydroxyl group (O2) with the hydroxyl group of Ser130 (3.0) and the sulfate ion (SO4-B, 2.6). The third BAB hydroxyl group (O3) forms a hydrogen bond with the water molecule OW10 (2.8). In all observed time steps OW10 remains well-defined in the same position. This water molecule is well-known as the catalytic water molecule mandatory for the deacylating step in -lactam hydrolysis45, initiated by nucleophilic attack on the carbonyl carbon atom of the acylenzyme complex to hydrolyze the acyl bond. It forms hydrogen bonds with the side chains of Ser70 (2.6), Glu166 (2.6), Asn170 (2.5) and BAB (O3, 2.8) (Fig.7). All these intermolecular interactions ensure that BAB is very well coordinated, e.g. a rotational disorder around the Ser70 borate ester linkage is not observed.

Hydrogen bonds are displayed as dashed lines.

The active site of the (a) bound boric acid and (b) glycerol boric acid diester is shown at the 10s delay time point. OW10 is hydrogen bond donor and acceptor to the boric ester of Ser70 (2.6/2.8). The tetrahedral hydrogen bonding pattern of OW10 is completed by Glu166 (2.6) and Asn170 (2.5). Hydrogen bonds of OW10 to GBE are longer than to BAB (2.8/3.2) while the hydrogen bonding pattern with Glu166 (2.5) and Asn170 (2.6) remains similar. The boron atom is positioned at a distance of 3.0 (BAB,
10s) or 3.4 (GBE, 10s) from the catalytic water OW10. Thus, the catalytic water could perform a nucleophilic attack on the boron atom, leading to the reversible hydrolysis of the boric acid serine ester linkage in BAB and GBE. Potential hydrogen bond distances () are indicated by dashed lines.

After monitoring time-resolved structure and dynamics of boric acid binding in the active site of CTX-M-14, we have further investigated the esterification process of boric acid with glycerol. For this purpose, the TapeDrive setup was used again to mix glycerol with CTX-M-14 microcrystals complexed with boric acid beforehand. We defined the delay time 0ms as the starting condition where no glycerol was added, corresponding to the last time point (1h soak) of the serial data collection with boric acid, considering that CTX-M-14 microcrystals were saturated with boric acid (Fig.4 and Supplementary Fig.6). At this defined time point, the occupancy of BAB was refined to 57%. The first change in the electron density of the polder map appears already at the 50ms mixing/delay point. In the region of the BAB hydroxyl groups extending electron density was observed indicating the formation of a glycerol diester. The obtained electron densities allowed the insertion and refinement of a glycerol boric acid diester (GBE), resulting in a GBE occupancy of 26%, while the BAB occupancy remained almost the same with 55%. This indicated also that the formation of the GBE increases the total occupancy of bound ligand in the active site to 81%. The electron density of the sulfate decreased for SO4-A to zero, as the newly formed glycerol diester occupies this position. The alternatively positioned SO4-B fits into the active site together with the BAB and is therefore still present with the same occupancy as the BAB. The observed electron densities at the 80 and 100ms delay times showed only a slight increase for GBE occupancy. A sharp increase in the corresponding GBE occupancy to 51% was observed and refined at the 150ms time point, while in parallel the BAB occupancy dropped to 35% (Fig.5b, Supplementary Table3). By this time, all atoms of GBE are covered with the calculated polder electron density. At the 750ms time point, the entire GBE was well-fitted and covered in the calculated electron density map with a resulting occupancy of 65%. Consequently, since the GBE can no longer act as a hydrogen bond donor for SO4-B due to the lack of hydrogen atoms at the position O2. The sulfate ion is finally completely replaced by a water molecule, OW357, which is accompanied by an increasing GBE and a decreasing BAB occupancy. This correlates with reduced electron density in the SO4 site. The O3 of GBE can also no longer interact with OW10 as a hydrogen bond donor, but only as a hydrogen bond acceptor. GBE approached a refined occupancy of 67% after only 10s delay time, while BAB occupancy dropped to 21%. However, it is interesting to note that the overall occupancy of the ligands (BAB, GBE) bound to Ser70 increased with the observed increase in electron density obtained and refined for the cyclic diester. Thus, the total occupancy of the binding site and region increased from 57%, obtained for soaking only with boric acid, up to 88% when further mixing with glycerol up to a delay time of 10s. The stepwise blocking of the active site by boric acid and the subsequent glycerol diester formation is shown in Fig.4.

Boric and boronic acids have a propensity to form esters with polyalcohols, resulting in the formation of five- or six-membered rings46,47,48. The observed five-membered scaffold of GBE is reminiscent of the autoinducer-2. This borate diester was first observed in complex with the sensor protein LuxP of the marine bioluminescent bacterium Vibrio harveyi49. The triol glycerol can alternatively form both ring systems, with the formation of a six-membered ring being energetically preferred over the five-membered ring, as shown in a computational study46. The investigation of peptidomimetic-boronic acid inhibitors for flaviviral proteases revealed both, a five-membered ring formation of the boric acid moiety and glycerol in the active site for the West-Nile virus NS2BNS3 protease and a six-membered ring formation for the Zika virus NS2BNS3 protease47,48. Despite the high similarity of these enzymes, both ring formations were observed, clearly showing the influence of the individual active site, resulting in a preference due to steric constraints47,48. In the CTX-M-14 active site, glycerol forms a five-membered cyclic diester with two of the three hydroxyl groups (O2, O3) of boric acid that is bound to the active site Ser70 (Fig.2c). A corresponding proposed reaction pathway is shown in Fig.8. The remaining hydroxyl group (O1) of the boric acid maintains the stabilizing hydrogen bonds with the main chain nitrogen atoms of Ser70 (2.9) and Ser237 (3.0) in the oxyanion hole (Fig.2c). During the esterification the sulfate ion in the anion-binding site is finally replaced by a water molecule (OW357) that forms alternative hydrogen bonds with the cyclic diester O1 (2.7) and the hydroxyl group of Thr235 (2.8) (Fig.2c). The other oxygen of the cyclic diester O3 forms a hydrogen bond with OW10 (3.2), which itself is strongly coordinated by Ser70 (2.8), Glu166 (2.5) and Asn170 (2.7). The remaining free hydroxyl group of GBE (O4) forms an additional hydrogen bond with the amide side chain of Asn132 (3.0) and weak hydrogen bonds with amide side chains of Asn104 (3.5) and Asn170 (3.5) (Fig.2c). In that conformation all oxygen atoms of the GBE are coordinated via hydrogen bonds either directly with the enzyme or via a water molecule. This is probably also the reason for the preference of the five-membered over the six-membered cyclic diester in the CTX-M-14 active site. In a six-membered ring, the free hydroxyl group could not form hydrogen bonds with Asn132 because it would be located in the center of the molecule. In fact, there would probably be no side chain for possible hydrogen bond interactions with the free hydroxyl group in that orientation as it would point out of the active site. Thus, the formation of a hydrogen bond of the free hydroxyl group of GBE with Asn132 is probably the determining factor, explaining our observation of only five-membered cyclic diester formation in all obtained GBE structures.

Hydrogen bonds are displayed as dashed lines.

The central carbon atom of the glycerol diester with boric acid becomes a stereo center with S-configuration. Also, the boron atom of GBE is a stereo center with S-configuration. Both stereocenters are observed without any racemic disorder. This is probably an indication for the specific active site environment of the -lactamase. For example, proteinase K has weak specific substrate preferences and glycerol forms a simple monoester with the boric acid bound to the active site serine (PDB code 2id850). Obviously, the stepwise formation of a monoester and diester is much too fast to be observed with our experimental setup.

As expected, the covalent binding of boric acid and the boric acid diester to the catalytic Ser70 in the active site of CTX-M-14 -lactamase resulted also in the inhibition of the enzyme17,18. Boric acid remains in the active site of the -lactamase in the crystal lattice with an occupancy of 57% even after prolonged soaking. Consequently, it can be concluded that the boric acid diester does not dissociate over time and therefore inhibits the enzyme (in the crystal lattice) for a certain period if the solvent conditions are unchanged. To quantify the effect of the observed occupation of the active site,
enzymatic activity assays applying a photometric determination of the 50% inhibitory concentration (IC50) values were performed. Moderate IC50 values of 2.90.4mM for boric acid and 3.10.4mM for the combination of boric acid with glycerol were determined (Supplementary Fig.7). Interestingly, the IC50 values are quite similar even though the crystallographic data showed a higher occupancy of the GBE in the crystal lattice, which would imply a higher inhibition. Compounds that are considered as inhibitors usually have substantially lower IC50 values, therefore the boric acid and the glycerol diester at this point cannot be considered as effective -lactamase inhibitors. This is in line with the observed incomplete occupancy of the boric acid and its glycerol diester in the crystal structures and the potentially reversible binding of boric acid. The organization of the active site in the endpoint complexes may also indicate that reversible mechanism for the dissociation of the inhibitor. The boron atom is positioned at a distance of 3.0 (BAB, 10s) or 3.4 (GBE, 10s) from the catalytic water OW10 (Fig.7). Thus, the catalytic water is well positioned to perform a nucleophilic attack on the boron atom, leading to the reversible release of boric acid or the GBE. Reversible inhibitors have the advantage of not being depleted or modified by their target, thereby enabling their capacity to inhibit several enzymes during their lifetime. Our data highlight the potential of boric acid derivatives in medicinal chemistry.

Here is the original post:
Time-resolved crystallography of boric acid binding to the active site serine of the -lactamase CTX-M-14 and ... - Nature.com