Identification of a 1-deoxy-D-xylulose-5-phosphate synthase (DXS) mutant with improved crystallographic properties

In this report


Introduction
In 2015, the world health organization (WHO) published a global action plan on antimicrobial resistance (AMR) [1].One of the five main objectives in the plan is an increase of research and development to fight AMR.In addition to improvements in how available antibiotics can be used, the development of innovative drugs is an essential strategy to address emerging resistance.The 2019 WHO report on antibacterial agents in clinical development defines the required innovation of a drug by the absence of crossresistance, a new compound or target class or a new mode of inhibition.The authors estimate that in the next five years, eleven new antibiotics could be approved, but only one might be innovative and active against resistant Gram-negative bacteria -highlighting that the need for innovative antibiotics is as urgent as ever [2].
The targets of antibiotics currently on the market have been mainly involved in mechanisms essential for the proliferation of pathogens, such as protein and cell-wall biosynthesis or DNA/RNA replication and repair [3].Nowadays, in the search for innovative antibiotics, more and more unconventional targets are explored.One source for unconventional targets is the methylerythritolphosphate (MEP)-pathway.Before the discovery of this pathway in 1993, its products, isopentenyl diphosphate and dimethylallyl diphosphate, were thought to be exclusively accessible via the mevalonate pathway [4e6].With the discovery of the MEP pathway, an alternative biosynthetic route to these universal building blocks for all isoprenoids was found.Many bacteria and the chloroplasts of plants rely on this pathway, whereas humans and most Eukarya use the mevalonate pathway [5e7].This distinction amongst species makes the MEP pathway very attractive for the development of new drugs [8,9].As a result, the MEP pathway has been the target of several projects to develop new antibiotics [10e12].
The first enzyme of the MEP pathway, 1-deoxy-D-xylulose-5phosphate synthase (DXS), catalyzes the rate-limiting formation of 1-deoxy-d-xylulose 5-phosphate (DOXP) [13].Compared with all other downstream intermediates, DOXP offers the additional benefit of being also the starting material for the biosynthesis of thiamine diphosphate (vitamin B1) and pyridoxal 5-phosphate (vitamin B6) [14,15].As DXS is also involved in multiple essential biosynthetic routes, we selected the enzyme DXS as a strategic branch point and therefore as a particularly interesting target for our own drug development [16,17].
Modern hit-identification strategies benefit greatly from structural knowledge of the enzymatic target, and drug optimization is accelerated by detailed knowledge of the atomic interactions between the compounds of interest with its target.However, while DXS has long been a target of interest, there is a relative paucity of high resolution structural information on this target: the first two crystal structures from organisms Escherichia coli and Deinococcus radiodurans were published in 2007, at highest resolution of 2.4 Å [18].Notably, Xiang et al. reported that the DXS protein of E. coli was only crystallized successfully after a fungal contamination led to partial proteolysis of the enzyme [18,19].Due to the potential of the MEP pathway for the development of new antibiotics, the demand for additional structural information of DXS homologues is high.This is illustrated by reports on the computation of homology models of pathogenic organisms, such as Mycobacterium tuberculosis and Plasmodium falciparum [20,21], and the use of orthogonal methods to gain structural insights, such as H/D exchange MS [22].In 2019, two DXS crystal structures of D. radiodurans were solved to higher resolution (1.95 Å) by Drennan and coworkers [23].These structures give further insight on the catalytic steps of DXS, but crystals must be grown under anoxic conditions, limiting the functional states of the enzyme that can be structurally characterized.
To support our own structure-based drug-design projects, we developed a truncated DXS construct that crystallizes readily under aerobic conditions and diffracts to a resolution of 2.10 Å.The truncated loop has a very low evolutionary conservation and we show that its removal had no influence on enzymatic activity.Due to the low conservation of the identified region, a similar approach should also be applicable to homologues of DXS, enabling the determination of crystal structures from pathogenic organisms in the future.

Protein expression and purification
The truncated DXS gene was obtained commercially, cloned into the pETM-11 expression vector and transformed into Escherichia coli BL21 (DE3).drDXS and DdrDXS were expressed and purified as described by Xiang et al. with minor modifications [18].After IMAC and Ion-exchange chromatography, the DdrDXS-containing fractions were pooled and cleavage of the His-tag was performed by TEV-protease digestion at 4 C overnight.Removal of the tag and protease was achieved by reversed IMAC chromatography.After concentration by ultrafiltration using a VivaSpin ultrafiltration device with a molecular weight cut-off of 30 kDa, the final gel filtration chromatography of DdrDXS was performed in 20 mM Tris-HCl (pH 7.5) 150 mM NaCl, 10 mM DTT.

Crystallization
The protein was concentrated in the presence of 50 mM ThDP to 23 mg/mL by ultrafiltration.The sample was centrifuged at 14,000 rpm before setting up drops.Initial screening was performed using commercial screens at RT in 96-well, SDMRC2 sittingdrop plates.For optimization, crystals were grown at RT using 2 mL hanging-drops, 1:1 mixture of protein and mother liquor.Crystals were obtained after 48 h with 0.2 M calcium acetate hydrate, 0.1 M Tris pH 8.5 and 20% PEG 4000 as the precipitant.

Data collection, processing and refinement
Diffraction data were collected at beamlines P11 and P13, DESY, Hamburg.Data reduction and scaling was performed using XDS [24].Molecular replacement with Molrep was used for phasing, using 2o1x as a search model [25].The model was further refined during several rounds of iterative manual model building and refinement in the CCP4 suite using coot and refmac [26e28].Refinement statistics are shown in Table S1.

Kinetic measurements
The DXS activity was analyzed at RT as previously reported, with minor modifications [29,30].Assay volume was 60 mL, to enable the use of 384-well plates (Greiner BioOne), buffer was 200 mM HEPES, pH 8.0.Data analysis was performed with the enzyme kinetics module of Origin pro 2019.

Thermal-shift assay (TSA)
Analysis was performed using an ABI StepOneplus RT-PCR instrument using white 96-well plates.A continuous heating rate of 1 C/min from 20 C to 95 C was used.Sample volume was 25 mL, consisting of 20 mL TSA buffer (20 mM Tris-HCl, pH 8.0; 300 mM NaCl, 5 mM MgCl 2 ), 2.5 mL protein solution and 2.5 mL dye (Sypro Orange, 5000 x in DMSO, Sigma-Aldrich).Optimal concentrations were experimentally determined, 1 mg/mL of protein, 50 x Sypro Orange in TSA buffer yielded the best signal-to-noise ratio.

LC-MS measurements
The protocol for LC-MS measurements is provided, together with its results, in the SI.

Truncation strategy used to design a crystallizing protein construct of DXS
While reproducing the protein crystals of D. radiodurans DXS (drDXS), we have observed a partial proteolysis of the 67 kDa protein into fragments of 20 and 40 kDa size, as previously reported for E. coli and P. aeruginosa DXS (Fig. S1) [18,31].This prompted us to use the technique of limited proteolysis to optimize the DXS protein.This method is based on the observation that well-structured domains of a protein are protected against proteolytic digestion [32,33].Analysis of such a partial digestion can lead to reengineered proteins that contain the protected, well-folded domains and have often more suitable properties for protein crystallography.
In the case of DXS, while a digestion site is not identified in the published crystal structures, amino acids 199e244 show no electron density in the DXS structures 6ouw, 2o1x and 2o1s and only fragmented, partial density for 6ouv.As a result, we hypothesized that the 20 kDa fragment corresponds to amino acids 1e199 and the 40 kDa fragment to amino acids 240e629.The degraded protein sample was analyzed by LC-MS to identify the exact cleavage site.A ~20 kDa fragment could not be observed, but we could observe a mixture of three different species with masses of 42,890, 42,690 and 42,489 Da, respectively (Fig. S2, S3).The observed masses correspond well to the calculated masses of the drDXS protein fragments with the amino acids 232, 234 or 236 to 629, respectively (sequence following UniProt-ID: Q9RUB5).Taken together, we concluded that the loop of amino acids 199e236 of drDXS is particularly sensitive to proteases, but we cannot exclude the possibility of autocatalytic cleavage of this loop.
To answer the question if the flexible loop is a speciesindependent property of the DXS enzyme, we analyzed the sequence conservation of all 498 deposited and manually annotated bacterial DXS genes of the Uniprot database [34].A simplified image of the calculated multiple sequence alignment (MSA) is shown in Fig. 1 (full MSA in SI).While the sequence homology of all 498 DXS proteins is 62.6% overall, the digested loop (200e232) displays a lower homology of 41.2%.We found that the loop also has a high variability in length, ranging from 5 to 58 amino acids in Myxococcus xanthus and Kocuria rhizophila, respectively.The mean loop length is with 45 amino acids similar to that of the 43 amino acids of D. radiodurans.Such variability is an indication that this loop is not essential for the catalytic reaction.
Based on the LC-MS results, the MSA and the lack of density between amino acids 200e240 in 2o1x, we designed a construct that replaces amino acids 201e243 with six glycine residues.This linker was designed to be long enough to bridge the gap of 11.7 Å between the two amino acid chains, but short enough to avoid introduction of multiple linker conformations.We expected that these modifications yield an optimized protein (DdrDXS), with properties more suitable for crystallization (sequence in SI).

Biophysical characterization
Purified DdrDXS protein was analyzed by LC-MS.The sample eluted as a single peak with a mass of 63,382.95Da, which is in good agreement with the calculated mass for DdrDXS of 63,382.11Da (Fig. S4).LC-MS and SDS-PAGE analysis showed the intact protein, even after a week incubation at RT, confirming the desired improvement in stability of DdrDXS (Fig. S1, S4).
To analyze if the truncation affects the activity of DdrDXS, we determined the enzyme kinetics for both substrates [31].The results are summarized in Table 1 and shown in Fig. S6, S7.The truncated drDXS enzyme retains its catalytic activity.It shows, however, slightly lower affinities for both substrates and a reduced turnover number.
To further investigate the effects of the truncation, the melting points (T m ) were determined using a thermal shift assay (TSA) [35], in which any increase of the T m is a sign of improved protein stability.This is often used to screen for optimal buffer conditions or analyze the effect of mutations [36].With a T m of 55.2 C, the truncated enzyme shows a nearly identical value to that of the native enzyme, which has a T m of 55.0 C (Fig. S8), indicating that our loop truncation had no significant effect on protein stability.
Crystallization screening of DdrDXS yielded several conditions, with the best crystals diffracting to a resolution of 2.1 Å.The protein structure obtained is deposited in the PDB with the code 6xxg and the collection and refinement statistics are reported in Table S1.

Effects of the truncation on protein folding
The truncated protein is catalytically active and no major changes in its properties could be identified using biophysical characterization methods.Since the DdrDXS protein yields welldiffracting protein crystals, we were also able to analyze the effect of the truncation by comparison of the obtained X-ray structure with the wild-type enzyme.
To compare the structures, the C a -RMSD of residues 8e183, 253e288 and 322e627 between truncated and the wild-type structures were calculated, and a superposition colored by individual C a -RMSD is shown in Fig. 2 [26,37].The RMSD on C-alpha position is 0.459 Å, 0.476 Å and 0.328 Å with 2o1x, 6ouv and 6ouw, respectively.These values show that the majority of the structure is unaffected by the truncation.While comparing the structures, we could also identify two regions that are present in a novel conformation: an a-helix (residues 186e200) and a b-hairpin motif (residues 303e320; Fig. 2).
The b-hairpin motif (residues 303e320) was described recently as part of a so-called "spoon".This motif undergoes structural rearrangements during the catalytic cycle of the DXS enzyme upon pyruvate binding.In our structure, the b-sheets of chain A adopt a conformation similar to the reported "bent spoon"-motif, while the equivalent residues of chain B are disordered.However, the "bent spoon" of our structure is distinct from than reported by Drennan and coworkers in 6ouw [23].With this observation, the presented structure further contributes to the understanding of conformational changes in this region during catalysis.
The a-helix formed by residues 186e200 is directly adjacent to the truncated amino acids 201e243.It can also be observed in the wild-type structure, but starting at residue 193.The C-terminal extension of the helix alters its orientation and moves residues 184e188 away from the ThDP binding site.The amino acids folding away from the active site do not take part in the catalyzed reaction, but form the hydrophobic surface of the ThDP binding site.
The same conformational change of this a-helix can also be observed in the recently published structure 6ouw.In this structure, the amino acids from position 184 are also part of the a-helix and point away from the active site.This similar folding to the truncated structure shows that this conformational change is not caused by the truncation.Jordan and coworkers have also identified the amino acids 183e199 as flexible and solvent-exposed using H/D exchange MS experiments.They show that this part of the protein adopts two distinct states, driven by substrate binding [22].It seems that the truncation facilitates the formation of a stable ahelix.

Crystal contacts and packing
We could observe that the residues 186e200 of the a-helix form new lattice contacts.The a-helix of chain A is at a distance of 8 Å and parallel to an a-helix formed by the amino acids 28e46 of an adjacent DXS protein.This proximity enables salt bridges between Asn195 and Arg199 and the Glu35 of the neighboring protein in the crystal lattice with a distance of 2.7 Å and 4.5 Å, respectively (Fig. S9).In future constructs, these interactions could be strengthened by introduction of more charged amino acids.
Comparing the packing, we could identify two different forms of DXS protein crystals.The previously determined structures 2o1x and 6ouv both have a Matthew's coefficient (V M ) of 2.75 Å 3 /Da and a solvent content of 55%.The truncated structure has a V M of 2.27 Å 3 /Da and a solvent content of 45%, similar to the DXS structure of 6ouw.As shown in Fig. 3, these two proteins adopt a different orientation in the crystal lattice and have a tighter packing, reducing the unit cell parameters.In the tightly packed structure 6ouw, the residues 307e319 are in the "bent-spoon" conformation, in the truncated structure a similar ß-hairpin, but at a location 14 Å distant to the "bent spoon" motif, can be seen.Both structures have the previously described a-helixes of the "fork" motif (amino acids 292e306) in a disordered state with no observable electron density.
A higher density in protein crystals correlates with an increased resolution and is a desirable feature for future crystal structures in complex with ligands [38,39].Using the truncated protein and the addition of pyruvate during crystallization seems to push the protein into the "bent spoon" conformation and might be used in the future to obtain better protein crystals.With a sample size of only four protein crystals, this hypothesis will need further evaluation as more DXS structures become available.

Conclusion
In our efforts to create a DXS enzyme for crystallographic studies with improved stability, we identified the part of the enzyme that is most susceptible to degradation.The identified loop is comprised of the residues 201e243, which show a high evolutionary variability in both length and sequence through all known bacterial homologues.We designed and expressed a mutant protein lacking the identified loop.The truncation showed only a small effect on the enzymatic activity and no effect on the biophysical properties of DXS, but a substantial improvement in the crystallographic properties of the protein.The DdrDXS protein has an improved tendency to form protein crystals under aerobic conditions, and diffract to a better resolution than previously published aerobic crystals of the wild-type protein.
Comparison of the obtained crystal structure with published structures showed no effect of the truncation on the remaining protein.Only two regions of the enzyme, previously known to be flexible, were identified in a different conformation and give additional evidence for the recently reported "spoon"/"fork" motif at the active site, proposed by Drennan and coworkers [23].
As we could find this loop in all species, we expect that the transfer of the identified site of truncation to other bacterial

Fig. 1 .
Fig. 1.MSA of DXS enzymes, including Deinococcus radiodurans (1, DEIRA) and Mycobacterium tuberculosis (2, MYCTU).For the sake of clarity, the image only shows three sequences.The identity, shown as bar graph above the sequences, was calculated using all 498 aligned sequences.Between amino acids 200 and 240, a highly variable region can be observed.

Fig. 2 .Fig. 3 .
Fig. 2. Superimposition of the drDXS structures with the code 2o1x (gray) and 6xxg (colored by C a -RMSD).Color coding: blue e low RMSD to red e high RMSD.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Table 1
Kinetic comparison of the native and truncated drDXS enzyme.