Mycobacterial OtsA structures reveal an allosteric site regulated by 2-oxoglutarate and 2-phosphoglycerate and the mechanism for substrate preference

Trehalose is an essential disaccharide for mycobacteria and a key constituent of several cell wall glycolipids with fundamental roles in pathogenesis. Mycobacteria possess two pathways for trehalose biosynthesis. However, only the OtsAB pathway was found to be essential in M. tuberculosis, with marked growth and virulence defects of OtsA mutants and strict essentiality of OtsB2. Herein, we report here the first mycobacterial OtsA structures from M. thermoresistibile in both apo and ligand-bound forms. Structural information reveals three key residues in the mechanism of substrate preference that were further confirmed by site-directed mutagenesis. Additionally, we identify 2-oxoglutarate and 2-phosphoglycerate as allosteric regulators of OtsA. The structural analysis in this work strongly contributed to define the mechanisms for feedback inhibition, show different conformational states of the enzyme and map a new allosteric site.


Introduction
Trehalose is a non-reducing disaccharide, formed by α-(1-1) linked glucoses, with a wide distribution in nature and present in all three domains of life (1). This remarkably widespread sugar performs multiple roles in a wide variety of organisms and it can also fulfil different roles within the same organism. Trehalose has been considered a compatible solute, conferring protection to proteins, DNA, membranes and whole cells from thermal shock, osmotic shock, freezing, ionizing radiation, oxidative stress and desiccation (1)(2)(3)(4)(5)(6)(7)(8)(9). This disaccharide, which can further function as a carbon and energy reserve molecule (1,(10)(11)(12), was also recently related to pathogenicity of Pseudomonas aeruginosa in plants (13). It additionally plays fundamental signaling roles in plants in the phosphorylated form (trehalose-6-phosphate) where it regulates sucrose metabolism and flowering (1,(14)(15)(16) and in yeast where it regulates gluconeogenesis and glycolysis (17,18). Trehalose was further shown to be an autophagy inducer, both in plants and in mammals, with potential biotechnological and clinical implications (19,20).
In mycobacteria, trehalose is also an essential component of mycolic acids, which are major protagonists in Mycobacterium tuberculosis pathogenesis (1,21), the causative agent of the widespread infectious disease tuberculosis. In these organisms, trehalose was further identified as a key signaling molecule of cell-envelope stress playing a role as an activator of the iniBAC operon, which is induced when mycobacteria are exposed to the first line drug isoniazid (22).
All mycobacterial species, with a few exceptions, possess two pathways to synthesize trehalose, the OtsAB and the TreYZ pathway (1). The OtsAB pathway is the most widely distributed trehalose biosynthesis pathway, present in bacteria, archaea and eukaryotes (23,24). In this pathway, which is conserved and essential in mycobacteria, trehalose is synthesized in a two-step process involving OtsA and OtsB2 enzymes. M. tuberculosis mutants showed that the OtsAB pathway was not only the dominant route for trehalose biosynthesis in this pathogen but also required for growth both in vitro and in a mouse infection model, with marked growth and virulence defects of OtsA mutants (25) and strict essentiality of OtsB2 due to the toxic effect of trehalose-6-phosphate (T6P) accumulation (25,26).
OtsA, a glycosyltransferase that belongs to the GT20 family of the CAZY classification (www.cazy.org), uses the α anomer of glucose-6-phosphate (G6P) as acceptor and NDPglucose as the donor to synthesize T6P, with net retention of the anomeric configuration of the donor substrate (27, 28). Interestingly, OtsAs from different organisms show different donor preferences as shown by kinetic studies and X-ray crystal structures, but the reasons behind these preferences are poorly understood (28)(29)(30)(31)(32)(33). Recently, different roles were identified for OtsA beyond its enzymatic activity. OtsA was reported to act as an osmotic stress sensor and morphogenetic protein that can regulate the switch to myceloid growth in Arthrobacter sp. strain A3, a pleomorphic soil dwelling actinobacteria (34).
In this work we have purified, crystallized and solved the structure of Mycobacterium thermoresistibile OtsA (MtrOtsA). To gain further insight into mycobacterial OtsA, we have obtained structures with substrates, product and pathway product that offer insight for mechanisms of substrate preference and feedback inhibition. We further performed structure-guided point mutations of key residues and characterized the mutants showing that three mutations are enough to change substrate preference. Importantly with these structures, we have also identified a new allosteric site and novel allosteric regulators of this enzyme that link glycolytic and TCA cycle metabolites to the regulation of "de novo" trehalose synthesis.

Overall structure
MtrOtsA is composed of two Rossmann-fold domains with a deep catalytic site at their interface, in an arrangement typical of GT-B glycosyltransferases as described previously for other organisms (28)(29)(30). The apo protein crystallized in the I4122 space group, with one protomer per asymmetric unit, and diffracted to ~1.8 Å resolution. The N-terminal domain is formed by a core of 7 parallel strands flanked on both sides by an antiparallel strand and surrounded by 8 α-helixes, one of which is composed of the final C-terminal residues (Fig. 1A). The C-terminal domain contains 6 parallel stands associated with 9 αhelixes with the last one undergoing a kink and extending to the N-terminal domain ( Fig   1A). Analysis of the B-factor distribution shows that the N-terminal domain has the highest values for atomic temperature factors, suggesting that this domain is more dynamic, which is consistent with the large movements of α-1 during catalytic activity.
MtrOtsA forms a tetramer in solution, as previously observed for M. tuberculosis OtsA.
This tetrameric form is also observed in all the crystal structures reported herein ( Fig 1B).
The amino acids involved in the tetramer interfaces are not conserved beyond mycobacteria and related species (Fig S1), suggesting that OtsA might have a different molecular assembly in different species. Indeed in E. coli both tetrameric and dimeric forms were reported (28)

and in Streptomyces venezuelae, Aspergillus fumigatus and
Candida albicans only the dimeric form was observed (29,30). However, in mycobacteria and closely related organisms the tetramer interfaces are highly conserved, suggesting a tetrameric assembly of OtsA in all of these organisms (Supplementary Fig   S1).

Catalytic site
OtsA, is a glycosyltransferase that uses the α anomer of G6P as acceptor and NDPglucose as the donor to synthesize T6P. The MtrOtsA catalytic site is located between the two Rossmann-fold domains in a large and deep cavity. Structures with donor substrates were determined by soaking the apo form crystals with ADP-glucose and GDP-glucose ( Fig 2). The donor substrates interact primarily with the C-terminal domain through the side chains of the highly conserved Arg286, Lys 291, Asp385, Glu393 and with the side chain of Arg365 (Fig 2), which is conserved in all mycobacteria and closely related species but less conserved outside this group ( Fig S1). Backbone interactions with the absolutely conserved residues Gly386, Met387 and Leu389 amine groups are also observed. N-terminal domain interactions with the donor substrate are only observed for His168 when the active site is in an open conformation (Fig 2). To obtain a structure with the acceptor substrate, we co-crystallized OtsA in the presence of 5 mM ADP and G6P.
As G6P binds to the protein in the presence of a donor substrate or its nucleotide, the active site adopts to a closed conformation ( Fig 3A). The OtsA:ADP:G6P ternary complex crystallized in the P6222 space group, with one protomer per asymmetric unit, and the crystals diffracted up to ~1.7 Å resolution. In this closed conformation new interactions with the donor substrate are formed with the side chain of Thr42 and the Gly39 peptide-NH function (Fig 3). Due to the absolute conserved nature of Gly39 and the contact it forms with the glucose-bound phosphate, this residue is highly likely to be mechanistically involved in the catalytic activity as proposed before (28).
The acceptor substrate interacts with the highly conserved residues Arg18, Tyr90, Asp144 and Gln146 of the N-terminal domain and with only a single residue of the C-terminal domain, Arg324 (Fig 3B). The residues involved in acceptor substrate interaction are highly conserved in all species ( Fig S1). As described for E. coli OtsA (28), the OtsA:ADP:G6P ternary structure shows the catalytic site in a closed conformation that substantially differs from the apo form, with α-helix 1 and the Arg35-Gly39 loop region moving up to ~12 Å (Fig 3A).

Properties of M. thermoresistibile OtsA.
MtrOtsA uses ADP-glucose, UDP-glucose and GDP-glucose as glucose donors with decreasing efficiency (Table 1) and G6P as the only acceptor, in accordance to what was reported before for M. tuberculosis OtsA (33,35). The kinetic parameters for MtrOtsA were obtained and are reported in (Table 1) (Table 1). For UDPglucose, the enzyme showed a ~7 fold higher Km of 1.7 ± 0.1 mM than the one obtained for ADP-glucose ( Table 1).
The preference of ADP-glucose over the other glucose donors was further confirmed by isothermal titration calorimetry (ITC). The binding affinity of substrate donors could only be determined for ADP-glucose in the tested conditions with an observable Kd of 27.17 µM ± 2.66 µM ( Fig S3). Other glucose donors analysed (GDP-glucose and UDP-glucose) had no observable heat of binding and the same was observed for G6P. The lack of heat of binding for GDP-and UDP-glucose is most likely due to their reduced affinity. For G6P case, it reflects a necessity of previous binding of the donor substrate.

Donor substrate preference is largely mediated by three residues
OtsAs from different organisms have been shown to have different donor substrate preferences, while being capable of using several nucleotide donors (30,33,36,37), which is reflected in lower conservation of the donor substrate interacting residues ( Fig   S1). Preference for ADP-glucose as the donor substrate in MtrOtsA is conferred by interactions with the deeply buried adenine moiety ( Fig 4A). The carbonyl groups of Leu319 and Arg361 interact with the primary amine, and the amide group of Val363 interacts with N1 of the adenine moiety ( Fig 4A). A highly coordinated water is also interacting with the primary amine of the adenine moiety and the carbonyl groups of Ala320, Leu359 and Arg361. The guanine moiety of GDP-glucose cannot occupy the same deeply buried pocket as the adenine because its primary amine group would sterically clash with Val363 carbonyl group ( Fig 4B). It thus binds to OtsA more weakly than ADP-glucose explaining the substrate preference and consequently the lack of observable heat of binding in the tested ITC conditions. A binary complex structure with OtsA and UDP-glucose was also obtained but the electron density was only observed for glucose and the two phosphates (not shown) indicating a reduced preference for this substrate as confirmed by enzymatic and biophysical data. Comparing mycobacterial OtsAs with E. coli, we hypothesised that the ADP preference was mediated by the substitution of an isoleucine (Ile295 in E. coli) for a leucine (Leu319 in M. thermoresistibile) that allows the primary amine of the adenine moiety to occupy a buried position interacting with the carbonyl group of Leu319 ( Fig 4C) that no other nucleotide activated donor can occupy. However, given the residue differences between the two enzymes it is likely that other residues could also play a role. To improve the selection of residues to mutate we employed a computational approach by using a mCSM-lig (38), a software developed by our group that predicts the effects of mutations on binding affinity.
The software predicted several that residues within 4.5 Å of the adenine moiety have a destabilizing effect on the interaction with the ligand if mutated to the E. coli OtsA equivalent (Table S2). Combining this information with structural analysis we selected three mutations predicted to be destabilizing; Val363Phe, Leu319Ile and Glu367Leu, the latter being a long distance mutation that we hypothesised would have a strong effect on ligand binding.
To confirm our hypothesis, we generated several MtrOtsA mutants (Leu319Ile, Val363Phe, Leu319Ile-Glu367Leu and a triple mutant) and we obtained kinetic properties for all of them ( Table 1). As we predicted the mutant Leu319Ile showed a ~3 fold increase e in the Km for ADP-glucose whereas, Km was similar GDP-glucose and UDP-glucose when compared to the wild type. Even though GDP-glucose now showed the best Km among the three different donors, the catalytic efficiency and turnover rate was still higher for ADP-glucose (Table 1). A complete reversal of the donor substrate preference for UDP-glucose was obtained only with a combination of the three mutations (Table 1). The triple mutant showed a ~5 fold increase in ADP-glucose Km and a reduction for the UDP-glucose Km by more than 3 fold ( Table 1). The catalytic efficiency was also completely reversed for ADP-and UDP-glucose between the wild-type and the triple mutant (Table 1). Although Leu319Ile showed the largest contribution for ADP-glucose Km increase, Val363Phe and Glu367Leu were also determinant for the reversal to UDP-Glucose preference (Table 1). This can be explained by the fact that the phenylalanine substitution at position 363 forces the loop between β-14 and α-14 to move towards the nucleotide binding site (Fig 4C), thus establishing stronger interactions with a pyrimidine nucleotide and clashing with purine ones. This effect is also observed for both ADP-and GDP-glucose, with a ~2.5 fold increase in Km when compared to the wild type. The leucine substitution at position 367 further helps in this move since the hydrogen bonds between the side chains of Arg266 and Glu367 are no longer present and α-14 is repelled from Arg266 due to the Leu side chain (Fig 4C).

Feedback inhibition
We soaked both trehalose and T6P into apo OtsA crystals and solved structures with both ligands (Fig 4D and 4E). In both structures OtsA active site assumes an open conformation and two compounds superpose almost perfectly at active site, recapitulating all of the interactions for the donor glucose and also interacting with His199 and Arg286, the latter a residue that interacts with the phosphates of the donor substrate (Fig 4D and   4E). The structure with T6P further shows the phosphate group of the product occupying a similar position to the phosphate group of G6P, interacting with the side chains of Tyr90, Gln146, Arg324 and Tyr145 a residue that does not interact with G6P ( Fig 4E).
Interestingly the buffer present in the crystallization condition (CHES) is observed in both trehalose and trehalose-6-phosphate bound structures, occupying the site of the nucleotide donor with the sulphate group binding in a phosphate site and interacting with hydroxyl groups from both glucose units of trehalose (Fig 4D and 4E).
After obtaining these structures, it was expected to observe some degree of feedback inhibition of OtsA by both trehalose and T6P. Both trehalose-6-phosphate and trehalose inhibited the enzyme, but the effect of trehalose was more pronounced with ~40% inhibition at 10 mM while for trehalose-6-phosphate a maximum inhibition of ~25% was reached at the same concentration (Fig S4). These results show that under physiological conditions OtsA is regulated by trehalose, the final product of the pathway which is highly abundant inside mycobacterial cells, but not by T6P which is known to be toxic for cells at high concentrations (26).

Identification of an allosteric site in M. thermoresistible OtsA
Trehalose is a multipurpose molecule that is abundant in mycobacteria. It can function not only as an energy reserve and as structural component of cell wall glycolipids, but can also lead to the synthesis of glycogen through the TreS-Mak-GlgE pathway and be directly synthesised from glycogen. It is therefore likely that "de novo" trehalose synthesis is under strong regulation.
In all the crystallization conditions that contained CHES, we could observe this compound in the crystal structures occupying a pocket formed by the contact interface of 2 OtsA protomers in the tetrameric assembly ( Fig 5A). The CHES sulphate group directly interacts with the side chains of Arg384 of protomer A and Arg213 of protomer B, forming hydrogen bonds with both, but also hydrophobic hydrogen-π interactions with Phe410 of protomer A (Fig 5B). The residues composing this site are completely conserved in mycobacterial and closely related species that are likely to arbour OtsA tetramers but not on others species known to have other oligomeric forms (Fig S1).
Arg384 sits in a loop that forms extensive contacts with the donor substrate through Asp385, Gly386, Met387 and Leu389 (Fig 2). Ligands interacting with the side chain of Arg384 could therefore have an impact on the activity of the enzyme making this site a prime candidate for allosteric regulation.
While CHES had no discernible impact on OtsA activity in the tested concentrations, it was reported before that F6P acted as an allosteric regulator of M. tuberculosis OtsA (35), and the same was reported for yeast OtsAB complex (39). We have tested MtrOtsA in the same conditions as those reported by Diez and colleagues, and could not detect any effect of F6P for MtrOtsA. Although we did not observe any effect of F6P on enzyme activity for MtrOtsAs, we nevertheless attempted to soak this compound in both apo form and OtsA:ADP:G6P ternary complex crystals, and also co-crystallize it in the presence and absence of ADP and G6P, all in CHES free conditions. None of these conditions provided a structure in which F6P occupied an allosteric site. We could however observe F6P when co-crystallized together with ADP, bound to the acceptor substrate site and occupying the G6P site, with OtsA presenting the active site in a closed conformation (Fig S5).
From the several compounds tested only 2OG and 2PG showed clear inhibition of OtsA activity with IC50 of 1.8 mM and 2.3 mM respectively ( Table 2) and (Fig 5C). To assess whether they could be acting allosterically on MtrOtsA and binding to the site identified by CHES we performed two mutations on the arginines that interacted with CHES (Arg213Glu and Arg384Glu) and tested the two mutants activity in the presence of 2OG and 2PG. The results show that the Arg213Glu mutation abolished the strong response of the protein to 2OG while there was a ~5 fold increase in 2PG IC50 to 12 mM. For the Arg384Glu mutant 2OG becomes unexpectedly becomes a strong activator with an EC50 of 0.7 mM while 2PG has no effect ( Table 2).

Discussion
Trehalose, an essential disaccharide in mycobacteria, is one of the critical components of the cell wall. The OtsA-B pathway is an essential source of trehalose for these organisms with knock-outs of this pathway being highly defective or non-viable (25,26). Here we reveal the first mycobacterial OtsA structure. The overall structure of M. thermoresistibile OtsA is similar to the reported structures of E. coli, S. venezuelae and Pseudoburkholderia xenovorans OtsAs despite relatively low sequence identity (33 and 35% respectively).
Remarkably, OtsAs have also a highly similar fold to the pseudo-glycosyltransferases VldE and ValL involved in validamycin A synthesis, a potent antifungal agent (42,43).
The structures obtained in this work allow us to explain how mycobacterial OtsA preference for ADP-glucose, is conferred by Leu319 which creates a cavity to accommodate the primary amine of the adenine moiety allowing it to interact with Leu319 and Arg361carbonyl groups. It is further mediated by Val363 carbonyl group that does not allow GDP-glucose to occupy the same position and Glu367. Mutating these three residues to the E. coli OtsA equivalents completely changed the preference to UDPglucose, the preferred E. coli OtsA donor substrate.
Trehalose is not only abundant inside the mycobacterial cell but also involved in a cycle that links it to the synthesis and degradation of glycogen (44). We have shown that OtsA is feedback inhibited by trehalose but not by T6P even though the structures obtained show that both compounds form extensive interactions with OtsA. However, given that these two structures were obtained by soaking apo form crystals with high concentrations of trehalose and T6P, and that CHES is also observed at the active site interacting with both compounds, it is possible that these two structures do not entirely reflect their natural interactions with the enzyme.
Further regulation of players in these pathways has been reported for GlgE, an enzyme that links trehalose recycling to glycogen synthesis, and that is regulated by phosphorylation (45). OtsA regulation by phosphorylation and methylation has also been observed for yeast trehalose-6-phosphate synthase complex (46,47). However, unlike  Fig S1).
We could nevertheless observe in many of our structures CHES occupying a site between two protomers that due to the interactions formed with Arg384 could affect the activity of the enzyme (Fig 5C). The reported interaction of OtsA with enolase (40) prompted us to test several glycolytic intermediates in our search for allosteric regulators, and we further included the metabolic regulators cAMP and 2OG. From the tested compounds only 2PG, the substrate of enolase, and 2OG had a regulatory effect of MtrOtsA activity within relevant physiological concentrations (48,49). Mutation of the allosteric-site residues Arg213 and Arg384 completely altered the effect of these allosteric inhibitors ( Table 2), demonstrating that the compounds indeed act through the allosteric site. The Arg384 mutation that transforms 2OG an allosteric inhibitor in an activator of OtsA hints at the possibility of allosteric activators of OtsA, however that remains to be confirmed.
The reported interaction between enolase and the fact that its substrate 2PG acts as an allosteric regulator of OtsA shows an interplay between these two enzymes that would be interesting to explore further on the enolase side. Similarly the allosteric regulation by 2OG a molecule that sits at the interface of carbon and nitrogen metabolism and that has been shown to regulate many different pathways (41) demands further work.
The results of this work are a significant step forward in understanding the regulation of "de novo" trehalose synthesis in mycobacteria the structural reasons behind substrate preference, and provide new important insight into this enzyme. were also constructed by site-specific mutagenesis, sequenced and transformed into E.

Structure solution and refinement
Diffraction data were processed and reduced using MOSFLM (50) and Aimless (51) from the CCP4 suite (52), or autoPROC from Global Phasing Limited (53). The apo form crystalized in I4122 space group with one protomer per asymmetric unit. ADP-G6P-OtsA ternary complex crystallized in P6222 space group again with one protomer per asymmetric unit.
Initial phases were determined with PHASER (54) from PHENIX software package (55) using the structure of E. coli OtsA (PDB entry 1UQU) (32) as a search model. Model building was done with Coot (56) and refinement was performed in PHENIX (55).

Prediction of mutations in ligand affinity
To provide an insight on which residues to mutate, we used mCSM-lig a software that predicts the effect of mutations in ligand affinity (38) on the X-ray crystal structure of OtsA with ADP-glucose (5JIO).

Enzymatic Assays
Formation of T6P was assessed by a continuous colorimetric assay that followed the release of NDP by measuring the oxidation of NADH at 340nm, in the presence of pyruvate kinase and lactate dehydrogenase. All reagents used were obtained from Sigma-       OtsA activity. Values are in (mM) and 95% confidence intervals are given in brackets.