The Three Mycobacterium tuberculosis Antigen 85 Isoforms Have Unique Substrates and Activities Determined by Non-active Site Regions*

Background: The existence of three antigen 85 isoforms in Mycobacterium tuberculosis has been proposed to be due to immune evasion. Results: Mutating divergent amino acids from outside the enzyme active site changes enzyme activity and substrate preference. Conclusion: These enzymes have unique substrate preferences and are regulated by a second carbohydrate-binding site. Significance: Understanding the enzymatic function of these enzymes may provide novel chemotherapeutic strategies. The three isoforms of antigen 85 (A, B, and C) are the most abundant secreted mycobacterial proteins and catalyze transesterification reactions that synthesize mycolated arabinogalactan, trehalose monomycolate (TMM), and trehalose dimycolate (TDM), important constituents of the outermost layer of the cellular envelope of Mycobacterium tuberculosis. These three enzymes are nearly identical at the active site and have therefore been postulated to exist to evade host immunity. Distal to the active site is a second putative carbohydrate-binding site of lower homology. Mutagenesis of the three isoforms at this second site affected both substrate selectivity and overall catalytic activity in vitro. Using synthetic and natural substrates, we show that these three enzymes exhibit unique selectivity; antigen 85A more efficiently mycolates TMM to form TDM, whereas C (and to a lesser extent B) has a higher rate of activity using free trehalose to form TMM. This difference in substrate selectivity extends to the hexasaccharide fragment of cell wall arabinan. Mutation of secondary site residues from the most active isoform (C) into those present in A or B partially interconverts this substrate selectivity. These experiments in combination with molecular dynamics simulations reveal that differences in the N-terminal helix α9, the adjacent Pro216–Phe228 loop, and helix α5 are the likely cause of changes in activity and substrate selectivity. These differences explain the existence of three isoforms and will allow for future work in developing inhibitors.

the Ag85 enzymes have shown only weak inhibition of mycobacterial growth at relatively high concentrations (6-azidotrehalose with a minimum inhibitory concentration of 200 g/ml (13) and ebselen with a minimum inhibitory concentration of 20 g/ml (14)).
The Ag85 enzymes have also been of interest from an immunologic perspective because they elicit a potent immune response (15) and bind fibronectin (16). Due to their immunogenicity, the Ag85A and -B isoforms are also components of some new anti-M. tuberculosis vaccine candidates (17). Because of their virtual identity in the immediate region of the active site, the three Ag85 isoforms (A, B, and C) have been proposed to play a role in evasion of the host immune response because their expression levels differ under specific environmental conditions (18,19). However, recent evidence from whole genome sequencing has revealed no evidence that antigens (including all three Ag85 proteins) show any more genetic variation than essential proteins, leading to the conclusion that immune recognition has been positively selected (20) and casting doubt on a rationale for the three isoforms involving immune evasion. A plausible explanation for the existence of three Ag85 isoforms in M. tuberculosis and an understanding of their functions within cell wall synthesis and M. tuberculosis survival are currently lacking.
The apparent functional redundancy of the three closely related Ag85 enzymes poses an additional challenge for drug development, because the biological function of the individual enzymes and the degree of cross-talk between the isoforms remain unclear. Although individual knock-out strains for each isoform have been generated (21,22), the triple knock-out ag85ABC in M. tuberculosis has not been reported, presumably because this is lethal. Because of the apparent ability of the Ag85 enzymes to complement each other, at least as assessed by their ability to continue to grow in vitro (23,24), individual mutants have thus far failed to fully dissect the functional roles of the Ag85s in cell envelope biogenesis. Work in Corynebacterium glutamicum has suggested that the Ag85 isoforms in Corynebacteria are not fully functionally redundant; mutants of individual corynebacterial Ag85 isoforms cannot be complemented through expression or overexpression of the M. tuberculosis isoforms (25). In M. tuberculosis, deletion of Ag85C resulted in a depletion of cell wall-linked mycolates, whereas ag85A and ag85B mutants show normal cell wall-associated mycolate levels. In enzyme assays measuring TMM formation from radiolabeled TDM and trehalose, Ag85A and particularly Ag85B, are markedly less active than Ag85C, although the enzymes have highly conserved sequences with nearly 100% identity of active site amino acids (13,26) (Fig. 2). Multiple structures for Ag85A, -B, and -C have been determined using x-ray crystallography and show that these three proteins are virtually identical (backbone root mean square deviation (RMSD) of ϳ0.9 -1.3 Å) with the single exception of the Ag85C structure bound to a covalent inhibitor (Protein Data Bank (PDB) entry 1DQY), which features a significant straightening of helix ␣9 and a disruption of the hydrogen bonding network of the catalytic triad. There are several regions outside of the active site that show some sequence divergence, including one region formed by two helices adjacent to the active site (␣5 and ␣9), where additional ligands have been found in co-crystals. Regulation of lipase and esterase catalytic activity by helix and lid flexibility near an active site is well documented (27). These non-conserved residues and the movement of helix ␣9 adjacent to the active site suggested a plausible mechanistic explanation for the observed differences in Ag85A, -B, and -C catalytic rates and suggested that there may be subtle substrate preferences. Here, we report that these secondary site amino acids mediate FIGURE 1. Activities of antigen 85. Antigen 85 catalyzes a transesterification reaction between an acyl donor (typically TMM) and an acceptor that can be free trehalose, another molecule of TMM, or the terminal arabinofuranose residues covalently anchored to the cell wall heteropolymer. The three common substrates are shown at the top, and the products of the enzymatic reaction are shown below each. Because they are highly insoluble, both cell wall polymer and TDM are likely to be terminal sinks for mycolic acids.
the flexibility of helix ␣9 and the configuration of the catalytic triad. The regulation conferred by these residues explains the previously observed differences in catalytic activity between Ag85A, -B, and -C and also supports a previously unreported substrate selectivity for trehalose, arabinan, and trehalose monomycolate among the Ag85 isoforms.

MATERIALS AND METHODS
Protein Expression and Mutagenesis-Plasmids were obtained from the Colorado State University tuberculosis research contract HHSN266200400091c as Rv3804/Ag85A Rv1886c/ Ag85B Rv0129c/Ag85C in pET15b (Ag85A) or pET23b (Ag85B/C) vectors containing a hexahistidine tag and were expressed in BL21 cells with overnight isopropyl 1-thio-␤-Dgalactopyranoside induction (500 M) in LB medium. Cells were harvested by centrifugation and lysed by sonication, and the protein was purified from the clarified lysate by affinity purification on Talon TM resin. Proteins were concentrated and buffer-exchanged into pH 7.2 phosphate buffer (20 mM) or pH 7.2 triethanolamine buffer (20 mM). Purity was evaluated by SDS-PAGE. Mutagenesis was conducted using a Stratagene QuikChange Lightning kit according to the manufacturer's instructions.
Circular Dichroism (CD)-Circular dichroism was measured in a Jasco J-715 spectropolarimeter, scanning four times from 260 to 190 nm at 50 nm/min, time constant ϭ 1 s, bandwidth ϭ 1 nm, slit width ϭ 500 m. Protein solutions in 10 mM phosphate buffer (protein concentrations, 100 g/ml) were held in 1-mm path length quartz cuvettes. Because of uncertainty about the protein concentrations, secondary structures were estimated as described previously (28).
Molecular Docking-Docking studies were performed using AutoDockTools version 1.5.4 (29) and AutoDock Vina (30). The coordinate .pdbqt file for Ag85A was prepared from PDB entry 1SFR by adding polar hydrogens and Kollman charges using AutoDockTools version 1.5.4. A grid box of 42 ϫ 34 ϫ 34 Å centered on the active site of Ag85A was determined. Energyminimized .pdb coordinates for the Ara5 ligand were prepared with ChemBio3D Ultra version 12.0. Gasteiger charges were added, non-polar hydrogens were merged, and rotatable bonds were set using AutoDockTools version 1.5.4 to generate a flexible coordinate .pdbqt file. The flexible ligand coordinates were docked into the Ag85a coordinates using AutoDock Vina employing a grid box consisting of 19,575 points. The resulting docking poses were visualized and overlaid with PyMOL version 0.99rc6.
Purification of [ 14 C]TDM and [ 14 C]TMM-Mid-log-phase H37Rv M. tuberculosis (A 650 ϭ 0.2, 1 liter in 7H9 medium) were labeled for 3 days with [ 14 C]acetate (2 Ci/ml; American Radiolabeled Chemicals), upon which cells were harvested by centrifugation (3000 rpm, 5 min) into 50-ml Falcon tubes and extracted into 2 ml of 2:1 chloroform/methanol for each 50-ml tube of cells for a total of 40 ml of organic extract. The organic layer was removed and dried, and the residue was resuspended in 2 ml of 2:1 chloroform/methanol. Then the entire sample was spotted onto two silica TLC plates, which were developed for 1 h using 80:20:2 chloroform/methanol/water, exposed to phosphor storage plates, which were scanned for radioactivity. A full-sized image of the plate was printed on a clear, plastic transparency film and used to identify the bands of interest (TDM, TMM, solvent front, and baseline), which were then scraped off. The silica was rinsed several times with 1 ml of 2:1 chloroform/methanol, the solvent was removed to a fresh vial, and the sample was concentrated. Trace silica was removed by passing through a small glass wool plug, and the final sample was resuspended in 2:1 chloroform/methanol to a final concentration of 10 Ci/ml for TMM and 50 Ci/ml for TDM.
TDM/TMM Assays-50 l of [ 14 C]TMM (100 Ci/l, ϳ200 M) stock solution in 2:1 chloroform/methanol was added to each reaction tube (Eppendorf, 1.5 ml), and all of the solvent was removed under a stream of argon. The remaining solid was resuspended by vortexing in 50 l of protein solution (20 M in 20 mM phosphate buffer, pH 7.2), and to this was added either 50 l of additional buffer or 50 l of alternate substrate. Trehalose and 6-azido-trehalose (26,31) were tested at a final working concentration of 200 M. Ara6 was tested at a final concentration of 200 M to 1 mM. The reactions were mixed with vortexing and allowed to incubate overnight at 37°C with shaking. Upon completion, the aqueous phase was extracted into 400 l of 2:1 chloroform/methanol, transferred to a new Eppendorf tube, dried under a stream of argon, resuspended in solvent (2:1 chloroform/methanol), and spotted. TLC plates were developed for 1 h (80:20:2 chloroform/methanol/water) and exposed to a phosphor storage plate, which was scanned for radioactivity.
Synthesis Synthesis of Ara6 -Synthesis of Ara6 was conducted from the intermediate pentasaccharide that has been reported previously (Scheme 1, compound 2 has been described previously (32)). Reactions were carried out in oven-dried glassware. All reagents used were purchased from commercial sources and were used without further purification unless noted. Reaction solvents were purified by successive passage through columns of aluminum oxide and copper under nitrogen. Unless stated otherwise, all reactions were carried out at room temperature under a positive pressure of argon and were monitored by TLC on Silica Gel 60 F254 (0.25 mm; Merck). Spots were detected under UV light or by charring with acidified p-anisaldehyde solution in EtOH. Unless otherwise indicated, all column chro-matography was performed on silica gel (40 -60 M). The ratio between silica gel and crude product ranged from 100 to 50:1 (w/w). 1 H NMR spectra were recorded at 500 MHz, and chemical shifts were referenced to TMS (0.0, CDCl 3 ). 1 H data were reported as if they were first order. 13 C NMR (attached proton test) spectra were recorded at 125 MHz, and 13 C chemical shifts were referenced to internal CDCl 3 (77.23, CDCl 3 ). Organic solutions were concentrated under vacuum at Ͻ40°C. (5). To a solution of alcohol 2 (0.238 g, 0.13 mmol) and trichloroacetonitrile (0.15 ml, 1.4 mmol) in CH 2 Cl 2 (5 ml) at 0°C was added 1,8-diazabicyclo [5.4.0]undec-7-ene (1 drop). The reaction mixture was stirred at 0°C for 30 min and then warmed to room temperature over 30 min. The solvent was then removed under vacuum, and a solution of dry hexane/toluene (2:3, 10 ml) was added. After being stirred for 5 min, this solution was quickly filtered through a short column of silica gel and Na 2 SO 4 (1:1). The resulting solution was then concentrated to yield the trichloroacetimidate derivative 3 (183 mg), which was used for glycosylation without any further purification. Alternatively, the syrupy residue obtained after the initial solvent evaporation following the reaction could be filtered quickly through a silica gel column (3:2 hexanes/EtOAc with 0.15% Et 3 N). The fractions containing the trichloroacetimidate derivative were concentrated, dried under high vacuum, and used immediately without any further purification. The trichloroacetimidate derivative 3 in CH 2 Cl 2 (3 ml) was added to a solution of alcohol 4 (0.07 g, 0.18 mmol) in CH 2 Cl 2 (4 ml) containing 4-Å molecular sieves (0.07 g; stirred already for about 30 min) at Ϫ20°C. A solution of trimethylsilyl trifluoromethanesulfonate (1.2 l, 0.0065 mmol) in CH 2 Cl 2 (50 l) was added dropwise over a period of 5 min. The reaction mixture was then warmed to Ϫ9°C over a period of 25 min and quenched by the addition of Et 3 N (3 drops). The solution was diluted with CH 2 Cl 2 (15 ml) and filtered. The filtrate was concentrated to a syrup that was purified by column chromatography (2:1 hexanes/EtOAc) to afford 5 (163 mg, 81%) as a thick syrup: . To a solution of 5 (163 mg, 0.075 mmol) in CH 2 Cl 2 -CH 3 OH (2:1 (v/v), 6 ml) at room temperature was added a solution of NaOCH 3 (20 mg, 0.37 mmol) in methanol. The reaction mixture was stirred for 36 h with the occasional addition of methanol (3 ml ϫ 4), and was neutralized with the careful addition of Amberlyst-15 (H ϩ ) cation exchange resin. The solution was filtered and concentrated to give syrup that was dissolved in distilled water (10 ml). The aqueous phase was repeatedly washed with EtOAc (6 ml ϫ 2), CH 2 Cl 2 (8 ml ϫ 2), and the separated aqueous phase was lyophilized to give 1 (55 mg, 89%) as a fluffy solid. The compound was fully characterized, and the spectral data obtained were in complete agreement with the data published earlier (33).
Molecular Dynamics-Crystal structures of the M. tuberculosis Ag85A (PDB entry 1SFR), Ag85B (PDB entry 1F0N), and Ag85C (PDB entry 1DQZ) were mutated or left unchanged and explicitly solvated with TIP3P water molecules and Na ϩ and Cl Ϫ counterions using the VMD program (34). Molecular dynamics simulations were performed under isobaric-isothermal conditions with periodic boundary conditions using the NAMD (35) program (version 2.8) on the Biowulf Linux cluster at the National Institutes of Health (Bethesda, MD). Electrostatic interactions were calculated using the particle-mesh Ewald summation. The CHARMM27 (36) force field was used with CHARMM atom types and charges. Prior to the start of a simulation, an energy minimization was performed using a conjugate gradient method, followed by slow warming to 310 K in 10-K increments with each increment equilibrating over 5 ps. Once the 310 K target temperature was reached, each system was equilibrated for an additional 50 ps. Production runs were conducted at 310 K for 100 ns with data collected at 1-ns intervals. For all simulations, a 1-fs integration time step was used along with a 12 Å non-bonded cut-off. Langevin dynamics were used to maintain temperature, and a modified Nosé-Hoover Langevin piston was used to control pressure.
Molecular Dynamics Analysis-Structures taken from molecular dynamics simulations at 1-ns intervals were aligned to each other, and an average structure was obtained using the UCSF Chimera package (37). Pairwise RMSDs of all C-␣ atoms were measured, and average RMSD values were added to the B-factor column of the averaged structure. The B-factor putty option was used in the PyMOL program (38) to generate tube widths and colors that correspond to the RMSD values.

Analysis of Structural Differences among Published Crystal
Structures of the Three Ag85 Proteins-The co-crystal structure of Ag85C bound to diethyl phosphate, a serine-reactive, covalent inhibitor, is the most divergent of the highly related Ag85 structures published to date (PDB entry 1DQY) (Fig. 2). Amino acids that differ between the Ag85 isoforms cluster to several regions, including at a secondary ligand binding site adjacent to the active site that has been serendipitously found occupied by trehalose, S-octyl-glucose, and glycerol in various structures (39 -41) ( Fig. 2A). Additionally, several amino acids located on helix ␣9 and the adjacent helix ␣5 are not conserved between Ag85A, -B, and -C but are highly conserved across related mycobacterial species ( 1DQY (Ag85C with Ser 124 bound to diethyphosphate) shows a significant straightening of helix ␣9 (39, 40, 42) (Fig. 2). This helix and the adjacent loop (blue) move substantially upon covalent modification, straightening the helix and folding the loop (cyan) into the active site, which in turn causes the catalytic glutamate (integral to helix ␣9) and the catalytic histidine to move away from the active site, disrupting the hydrogen bonding within the catalytic triad (Fig. 2B).
Mutagenesis of Isoform-specific Secondary Site Residues Modulates Acyltransferase Activity Levels-We theorized that the cluster of non-conserved residues that abut the secondary binding site could be partially responsible for differences in enzyme activity. Because Ag85B has been reported as the least active isoform (10 -20-fold less activity) and Ag85C has been reported as the most active in acyltransferase assays, our initial mutagenesis efforts focused on understanding the effect of converting this region of Ag85C into that found in Ag85B and vice versa. We produced wild type and chimeric proteins PDB entries 1F0N, 1DQY, and 1DQZ are shown. A, the difference in the position of helix ␣9 (apo-structure 1DQZ (blue) and diethylphosphate acylated 1DQY (cyan)), adjacent Pro 216 -Phe 228 loop (blue/cyan), and key amino acids from Ag85C (Trp 157 , Trp 158 , and Thr 231 ; yellow) and from Ag85A/B (Met 159 , Phe 232 , and Val 233 ; green), highlighting the differences between Ag85C and Ag85A/B at a secondary site that has been found with bound trehalose (1F0N, orange carbon, red oxygen atoms) or octylthioglucose (not shown). The catalytic triad is highlighted in magenta. B, the movement of helix ␣9 in A results in a disruption of the hydrogen bonding of the catalytic triad (dotted lines). The hydrogen-bonded triad in Ag85C structure 1DQZ is shown in magenta, and the disrupted triad in 1DQY is shown in cyan. C, sequence alignment of Ag85A, -B, and -C indicates the highly conserved nature of the enzymes. The box indicates sequence corresponding to helix ␣9 and the Pro 216 -Phe 228 loop, and the catalytic triad is indicated with arrows. Stars indicate the amino acids highlighted in A. Cyan, homology with Ag85A; magenta, homology with or unique to Ag85B; white, amino acids that are unique to Ag85C. Sequences are shown without signal peptides, and sequence numbering has been made consistent with that used previously (39 -41).
as His-tagged fusions and purified them by immobilized metal affinity chromatography. For all mutant and chimeric proteins, no significant difference in production level in Escherichia coli was observed, and all mutants were checked for proper folding using circular dichroism (CD) (data not shown).
Site-directed mutagenesis of Ag85C was directed initially at the two tryptophan residues at positions 157 and 158. Activity was assessed using a homogeneous, soluble in vitro enzyme assay (26), where trehalose and TDH are mixed with Ag85 and trehalose monohexanoate formation was monitored by MS. Mutating Trp 157 to alanine in Ag85C caused a doubling of activity, whereas mutation of Trp 158 to alanine resulted in a 10-fold reduction in activity to levels comparable with that observed in wild type Ag85B (Fig. 3A). Replacing both Trp 157 and Trp 158 with alanine also resulted in a relatively inactive enzyme. The Ag85C W157M/W158G double mutant (the wild type residues of Ag85A/B) was slightly more active than the double alanine mutant but still not very active. Further mutation of Leu 230 to alanine or phenylalanine in the double W157A/W158A mutant also had little effect on catalytic activity; nor did adding the L230F to the W157M/W158G to restore the full Ag85A/B MGF sequence in this region alter the low level activity observed with wild type Ag85B.
In contrast, mutation of these three residues on helix ␣9 and helix ␣5 in either Ag85A or -B to the corresponding residues occurring in Ag85C resulted in a 2-10-fold enhancement in activity (Fig. 3, B and C). In both Ag85A and Ag85B, introduc- tion of all three mutations simultaneously (M159W, G160W, and F232L (WWL)) resulted in a further enhancement of activity to levels nearly identical to Ag85C. Thus, these mutations in the secondary site of Ag85, notably positions Trp 157 , Trp 158 , and Leu 230 (Met 159 , Gly 160 , and Phe 232 in Ag85A/B), contribute directly to controlling the catalytic activity of the antigen 85 enzymes.
Excess Trehalose Induces Loss of Activity in Ag85C-Although moderate throughput, homogeneous, and quantitative, the Ag85 MS assay is limited by the short, synthetic nature of the lipid substrates attached to trehalose. To more fully understand the effects of these mutations on the activity of the Ag85 isoforms, we utilized an additional Ag85 enzyme assay, where [ 14 C]TMM (purified from M. tuberculosis grown in the presence of [ 14 C]acetic acid) was used as the mycolyl donor in in vitro reactions. This assay is similar to previously reported assays (13,43) in which purified TMM was used in enzyme reactions with [ 14 C]trehalose, but in this case, the mycolate is carrying the label and not the disaccharide. Using lipid-labeled [ 14 C]TMM allowed us to monitor the transmycolation of two TMMs to form TDM and free trehalose independent of the initial mycolation of trehalose to generate [ 14 C]TMM. When Ag85A, -B, or -C was added to a suspension of [ 14 C]TMM, [ 14 C]TDM was formed, and this could be monitored by TLC (Fig. 3D). Reactions were allowed to proceed overnight to evaluate the formation of product at thermodynamic equilibrium. Although steady state kinetics could not be determined due to poor solubility of [ 14 C]TMM and the low concentrations of actual substrate used, Ag85A, -B, and -C were all able to catalyze the formation of [ 14 C]TDM with Ag85C, producing very slightly less TDM than the other two isoforms. However, when 500 M trehalose was added to the reaction mixture, Ag85C converted less than half of the TMM substrate in the reaction mixture to TDM product (as quantified by phosphorimaging). In contrast, Ag85A and Ag85B maintained the same level of [ 14 C]TDM formation despite the presence of trehalose in the reaction mixture. This result suggests the possibility that occupancy of the second binding site of Ag85C with trehalose substantially reduces the catalytic activity of the enzyme. In contrast, Ag85A and -B were relatively insensitive to the presence of trehalose. We also incubated purified [ 14 C]TDM directly with Ag85C in the presence of excess unlabeled trehalose and observed negligible conversion of TDM to TMM (data not shown). Substituting in the three Ag85C residues (WWL) to Ag85B did slightly diminish the ability of this enzyme to convert [ 14 C]TMM to [ 14 C]TDM, but this effect was not augmented by the addition of trehalose. Likewise, the triple mutant of Ag85C bearing the wild type residues of Ag85B (MGF) had intermediate activity that was only somewhat affected by the addition of trehalose.
Ag85C Shows Higher Tolerance for the Altered Substrate 6-Azido-trehalose-We also explored using 6-azido-trehalose, which cannot undergo a second acylation reaction, as a surrogate for trehalose to detect transmycolyation reactions between TMM and trehalose. Incubation of 6-azido-trehalose and [ 14 C]TMM with Ag85C formed nearly equivalent amounts of TDM and 6-azido-trehalose-6Ј-monomycolate (6-Az-TMM) (Fig. 3E). In contrast, Ag85A continued to produce TDM effi-ciently and generated only a small amount of monomycolyated 6-azido-trehalose. Ag85B demonstrated an intermediate activity, forming TDM preferentially but still forming an appreciable quantity of 6-azido-trehalose-6Ј-monomycolate. Similar to the results with trehalose, substituting the three residues that differ between Ag85C and Ag85A/B did not recapitulate this differential activity toward 6-azido-trehalose (Fig. 3, D and E).
Molecular Dynamics Simulations Suggest Additional Determinants of Substrate Specificity-Because the Trp 157 , Trp 158 , and Leu 230 mutants failed to fully explain the observed differences in substrate specificity toward trehalose, we turned to molecular dynamics (MD) simulations to further understand the dissimilarities between the Ag85 isoforms.
MD simulations for all three of the Ag85 isoforms revealed that the N-terminal portion of helix ␣9, the Pro 216 -Phe 228 loop, and helix ␣5, all of which comprise the ligand entry site, are capable of significant conformational changes for all isoforms (Fig. 4). However, Ag85B and particularly Ag85A demonstrated greater overall helical motion for both ␣9 and ␣5. Simulations of the Ag85C Trp 157 , Trp 158 , and Leu 230 and the corresponding Ag85A/B M 159 Gly 160 and Phe 232 mutants revealed large changes in helical flexibility when these secondary site residues are mutated, consistent with the different conformations observed from x-ray crystallography experiments (Fig. 2).
Additionally, MD simulations revealed that another amino acid in the secondary site (Thr 231 in Ag85C, Val 233 in Ag85A/B) was probably involved in determining helix flexibility and that mutagenesis at this residue might further explain the observed differences. The bend in helix ␣9 occurs at Thr 231 for Ag85C and at Val 233 in Ag85A/B. In silico mutation of position 231 in Ag85C was followed by energy minimization and a 100-ns molecular dynamics simulation at 310 K. The mutant in which Thr 231 was replaced with alanine revealed a dramatic reduction in flexibility of this region and very stable short distances between the catalytic serine and histidine residues of the active site (Fig. 4). Substitution at this position with the Ag85A residue, T231V, modestly increased the flexibility and motion of helix ␣9.
Mutation of Val 233 in Ag85A appears to also dramatically alter helix flexibility and triad stability. In contrast with the stabilizing affect that the alanine mutation had in Ag85C, the Ag85A V233A mutant resulted in increased helical motion and large mean distances among residues of the catalytic triad. Substitution at this position with the Ag85C residue, V233T, dramatically decreased overall motion of both helices, conferring Ag85C-like motion on Ag85A.
Mutations at the Base of Helix ␣9 Affect Substrate Selectivity-To test our molecular dynamics predictions that single mutations at the hinge of helix ␣9 could partially interconvert the Ag85 isoforms, we generated a series of mutants in both Ag85A and C and tested them for changes in activity in our Ag85 assays. Mutations to Ala, Thr/Val, Leu, and Ile were conducted in both Ag85A and -C, resulting in dramatic changes in enzyme activity and substrate selectivity (Fig. 5).
For the T231A mutant of Ag85C, which had been predicted to have much less flexibility in this region and a much tighter active site, the overall rate of acylation was double that of the wild type enzyme, but even more strikingly, the rate of hydrol-ysis was 5 times higher than that of the wild type enzyme. The Ag85A V233A mutant also shows a moderate increase in acyltransferase activity and a 2-fold increase in hydrolase activity, suggesting that possible changes in helix ␣9 flexibility may be linked to acyl enzyme stability or accessibility of the active site to water. As expected in Ag85C, mutation of the bulkier valine, leucine, and isoleucine all afford a 2-3-fold decrease in relative rates of acylation and hydrolysis. Likewise, incorporation of the bulky isoleucine residue into Ag85A also resulted in a decrease in the rate of acylation and hydrolysis. In the MS assay, mutation to leucine results in a large drop in activity for Ag85C and a modest increase in acyl transfer and large increase in hydrolysis for Ag85A.
In the 6-azido-trehalose [ 14 C]TMM enzyme assays, the Ag85C T231V and T231L and T231I mutations all yielded a decrease in the production of 6-azido-TMM (Fig. 5C), supporting a model in which a change in Ag85C helix flexibility caused by the bulkier amino acid decreases the mycolation of free trehalose. The T231L and T231I mutations also induce an increase in TDM production, consistent with a shift toward Ag85A-like activity. In contrast, the Ag85C T231V mutation leads to retention of TMM but no dramatic increase in TDM production, which may reflect an intermediate phenotype. In Ag85A, the V233A and V233T mutations double the formation of TMM-6-azide and modestly decrease the formation of TDM, again consistent with a model in which inclusion of the Ag85C-like residues at this position leads to increased mycolation of free trehalose. For both Ag85A and -C, single mutations at the helix ␣9 bend dramatically change both the enzyme activity and substrate selectivity, which suggests that the degree of bend of helix ␣9 explains the difference in Ag85A, -B, and -C enzyme activities.
Potential Differences in Ag85 Isoform-specific Mycolation of the M. tuberculosis Cell Wall-One biologically plausible reason for the observed relationship of flexibility and activity may be related to constraints imposed by mycolation of the terminal arabinose units of the insoluble heteropolymeric arabinogalactan, an important second function of the Ag85 enzymes (Fig. 1). Ara6 (Fig. 6), the hexasaccharide terminus of the arabinogalactan polymer (44), was synthesized and assayed as a mycolyl acceptor in Ag85 assays. Docking studies with the related Ara5 suggested that this molecule should bind into an extended active site on Ag85 (Fig. 6, A and B), with the free 5-OH pointing toward the helix ␣9 groove. The already acylated mycolate in TMM or Ara6MM should bind along this hydrophobic helix. Thus, the flexibility of the helix and steric bulk of the amino acids would be critical for regulating binding and possibly the release of the substrate. Upon reaction with Ag85 and [ 14 C]TMM, a new mycolate-containing lipid species (asterisk) was observed (Fig. 6C). In the absence of Ara6, this species was not formed, although several minor, comigrating species were. To determine whether this species was possibly Ara6MM, 6-azido-[ 14 C]TMM was purified and used as a donor to mycolate the Ara6. Under the 6-azido-[ 14 C]TMM reaction conditions, exclusively this new solvent-font lipid species (asterisk) was generated, and in the absence of Ara6, no new species was observed. When 6-azido-[ 14 C]TMM is used as a mycolate source and no TDM is generated, all three Ag85 isoforms appear to generate equal proportions of the new lipid (Fig. 6C,  three right-hand lanes). We found that the new lipid species could be resolved from all other species by exposing the TLC plate to two rounds of chromatography; first, separation of TDM and TMM was achieved on two-thirds of the plate (80: 20:2 chloroform/methanol/water), and then the solvent front lipids were allowed to further separate on the front one-third of the plate (95:5 chloroform/methanol (Fig. 6, D and E). We postulated that the addition of high concentrations of competing trehalose (1 and 10 mM) to the reaction mixture might increase the mycolation of Ara6 by reducing the formation of the dead end TDM; however, in practice, additional trehalose did not have this effect (data not shown), although it did increase the retention of TMM in the reaction mixture. Ag85A produced more of the new lipid, both in the presence and absence of trehalose as a competitive substrate in the reaction mixture ( Fig. 6) (data not shown). Due to the previously documented (45,46) arabinan release upon treatment with the first line antimycobacterial drug EMB, we postulated that EMB treatment of M. tuberculosis might release mycolated fragments of arabinan of similar size to the novel species observed when Ara6 was subjected to the mycolation assay. As hypothesized, EMB treatment caused the production of a novel lipid species (double asterisk), which showed approximately the same apparent Rf as the lipid generated in the from the Ag85 mycolation assay in the presence of Ara6 (asterisk) (Fig. 6, F and G).

DISCUSSION
Previously reported differences in relative Ag85 catalytic activity toward trehalose have always stood in contrast to the extreme similarity of these proteins (13,26). Ag85C has been reported to have a higher k cat for the mycolation of free trehalose relative to Ag85B, and here we report that, in contrast with mycolation of trehalose, both Ag85A and -B show an increased capacity to mycolate TMM to TDM relative to Ag85C. Consis- tent with this finding, knockouts of ag85A have been shown to have decreased TDM levels (47).
The high level of sequence conservation among these orthologs caused us to focus on the few areas of divergence starting with amino acids along helix ␣9, and nearby helix ␣5, adjacent to the active site. Mutation of the three non-conserved residues in this region in Ag85A, -B, or -C (Trp 157 , Trp 158 , and Leu 230 in Ag85C and Met 159 , Gly 160 , and Phe 232 in both A and B isoforms) to alanine or their corresponding residues from the each other altered the rates of acyl transfer and acyl hydrolysis with shorter TDH substrate but did not affect the degree of trehalose mycolation from TMM. The presence of two tryptophans in the secondary site of Ag85C (Trp 157 and Trp 158 ) and absence in Ag85A/B suggests a carbohydrate binding role by -CH 2 interactions (48), which may in part explain the greater propensity for Ag85C to transesterify free trehalose when compared with Ag85A/B. However, mutation of these residues alone does not alter substrate specificity. Mutations of the helix ␣9 hinge do alter the substrate specificity, indicating that both carbohydrate binding at the secondary site and the related flexibility of helix ␣9 are required to explain the differences in rates and substrate selectivity of the Ag85 enzymes. Mutation of a single amino acid, Thr 231 (Ag85C) or Val 233 (Ag85A), partially interconverts enzyme substrate specificity. Introduction of bulkier Ag85A-like residues at Thr 231 (Val, Leu, or Ile) into Ag85C dramatically decreases mycolation of trehalose and 6-azido-trehalose. In contrast, introduction of the less bulky Ag85C-like residues into Ag85A at Val 233 (Ala or Thr) affords a FIGURE 6. The Ag85 substrate specificity may extend toward fragments of the mycobacterial cell wall, suggesting that the observed differences in enzyme activity may have implications in the partitioning of exported mycolates between TDM and the arabinan. A, Ara5 is docked in the active site of Ag85A (PDB entry 1SFR), revealing an extended active site that accommodates the bulk of the larger sugar. Residues are colored as in Fig. 1A. B, structures of docked Ara5 and assayed Ara6 are shown. My, mycolate attachment site. C, Ag85 TMM assays in the presence of Ara6 (500 M) and competing trehalose (200 M) reveal the formation of a new lipid species (*). Ag85A is most efficient at producing this new species from TMM. The new lipid (*) was the only significant product from the transesterification from TMM-6-azide to Ara6. TLC was developed in 80:20:2 chloroform/methanol/water. D, Ag85A-catalyzed formation of the new lipid species (*) in the absence (i) and presence (ii) of Ara6. E, the same TLC shown in D developed a second time in 95:5 chloroform/methanol to separate the solvent front lipid species, as indicated by the arrow. F, H37Rv M. tuberculosis treated with EMB at 10 times the minimum inhibitory concentration shows the accumulation of a new lipid species (**). TLC was developed in 100:10:0.1 chloroform/methanol/water. SQ109, which blocks the export of TMM, abolishes the formation of this species (**). G, the new lipid (**) in the EMB-treated 24-h sample from F has an Rf comparable with that of the product of the Ag85-catalyzed mycolation of Ara6 (*). TLC was developed twice first in 80:20:2, chloroform/methanol/water and then in 95:5 chloroform/methanol. Mycolation of Ara6 could not be increased further with the addition of higher concentrations of trehalose.
2-fold increase in mycolation of 6-azido trehalose. Mutation of Val 233 and Thr 231 also affects the relative rates of acryl transfer and acyl hydrolase activity, as measured by the mycolation of free trehalose from trehalose dihexanoate. Molecular dynamics links these key non-conserved residues to the flexibility of helices ␣9 and ␣5 and the catalytic triad position. Mutations that alter the flexibility of these helices dramatically alter the substrate selectivity, partially converting Ag85A to -C and vice versa. These results suggest that the secondary site in Ag85 is critical for the regulation of enzyme activity and substrate selectivity.
These results also allow us to propose a possible mechanism by which Ag85 activity is regulated by this secondary site. First, Ag85 binds to TMM at the active site, and transfer of the mycolyl chain from TMM to the enzyme occurs, forming the acyl enzyme intermediate. This transmycolation causes both a destabilization of the catalytic triad hydrogen bonding network and a concurrent straightening of helix ␣9 and folding of the Pro 216 -Phe 226 loop over the active site. The acyl enzyme is protected from hydrolysis by the disrupted triad and loop folded over the active site. Although the Ag85 enzymes show acyl hydrolase activity toward shorter trehalose esters, we observe negligible hydrolase activity toward TMM and TDM, suggesting that the acyl enzyme is very resistant to hydrolysis. When the acyl enzyme encounters either trehalose or TMM, these molecules bind at the secondary site, and this binding event is relayed to the active site by the reformation of the bend in ␣9 and restoration of the triad hydrogen bonding, mediated by Thr 231 /Val 233 . Once bound, the trehalose or TMM can move into the active site from the secondary site and be transesterified, producing either TMM or TDM. Any mutations that alter the flexibility of helix ␣9 or the residues in the secondary carbohydrate binding pocket affect the substrate specificity and relative activity of Ag85. Although some polysaccharide processing enzymes have been shown to contain secondary carbohydrate binding sites, to our knowledge, this is the first mechanistic example of how carbohydrate binding at a second site can directly control enzyme activity and substrate specificity (49).
The biological relevance of the existence of three Ag85 isoforms and their differences in relative catalytic activities remains to be fully understood. During the synthesis of the M. tuberculosis cell wall, the transmycolation reaction can occur from TMM to a second unit of TMM, generating TDM, or from either TMM or TDM to free trehalose or to the cell wall arabinogalactan polymer. Both TDM formation and anchoring mycolates to the cell wall constitute structurally important end points that probably represent mycolate sinks; although reversible in theory, the high lipophilicity and insolubility of these molecules probably means that mycolate flow terminates with their formation. The combined evidence suggests that one plausible scenario is that the primary physiologic function of Ag85C, which has the least flexibility in helices ␣5 and ␣9, is mycolyation of TMM to form TDM. Additionally, Ag85C may play a role in the initial mycolation of trehalose 6-phosphate from putative mycolyl carrier species, such as mycolylphospholipid (50), to generate TMM within the cell. We speculate that the flexibility in these two helices may be important for inter-action with the insoluble cell envelope and therefore that Ag85A and -B may be involved in mycolyation of the terminal arabinose units of the mature cell envelope. Alternatively, the observed difference in enzyme activity may reflect a means to partition TMM to both the cell wall and TDM, whereby Ag85C introduces the initial mycolate into each Ara6 motif, and subsequently Ag85A and -B are responsible for the addition of the remaining mycolates.
In order to more fully understand the differences in Ag85catalyzed mycolation of the mycobacterial cell wall, we used a short arabinan substrate in TMM-mediated mycolation assays. In the presence of Ara6 and TMM, Ag85A catalyzed the formation of a single new lipid species, putatively mycolated Ara6, with Ag85B and -C producing less of the new lipid. Although we had expected to observe formation of multiple mycolation states of the Ara6, the appearance of a single band suggests either processivity in the mycolation reaction or the addition of only a single mycolic acid. Such processivity is not unprecedented, because ϳ75% of the reducing termini of the M. tuberculosis cell wall are fully mycolated, whereas the rest are completely unmycolyated (44). Although we have been unable to fully characterize this new lipid species, despite considerable effort, lipid extracts generated from H37Rv M. tuberculosis, treated with EMB, a known inhibitor of arabinan synthesis that releases arabinan fragments, contain a product with a similar mobility on TLC (5,45,46,51). Product inhibition by TDM may reduce arabinan mycolation, particularly by Ag85C, as suggested by equal formation of the new lipid in the absence of TDM by all isoforms when 6-azido-[ 14 C]TMM is used as a mycolate source.
To date, few Ag85 inhibitors have been reported, and none have demonstrated inhibitory properties in vivo. The discovery that Ag85A and -B will generate large quantities of TDM even in the presence of high concentrations of competing trehalose may in part explain this paucity of inhibitors, because most groups have focused on the synthesis of trehalose analogues in attempts to generate Ag85 inhibitors. The ability of relatively subtle single amino acid changes at the secondary site of Ag85 to dramatically alter the catalytic activity of the enzymes suggests that, in addition to the active site, this portion of the enzyme may offer an alternative ligandable site that perturbs catalytic activity. Dissection of the individual roles of the Ag85 enzymes in M. tuberculosis remains challenged by the lack of the double and triple ag85abc mutants. Because the active sites of the Ag85 isoforms are nearly 100% conserved, this non-conserved site may also provide an additional handle for production of isoform-selective inhibitors and further characterization of individual isoform activity. These results with the Ara6 terminus suggest that Ag85A is particularly important for the synthesis of fully mycolated cell wall, but it remains to be seen whether these results have implications for cell wall turnover, TMM synthesis, or other putative functions of the Ag85 enzymes. Here we have provided the first molecular explanation of the differences in Ag85A, -B, and -C activity and propose that these differences in activity may serve important roles in the efficient synthesis and modulation of the production of the arabinogalactanmycolate polymer and the trehalose dimycolate.