Thiolation-enhanced substrate recognition by D-alanyl carrier protein ligase DltA from Bacillus cereus

D-alanylation of the lipoteichoic acid on Gram-positive cell wall is dependent on dlt gene-encoded proteins DltA, DltB, DltC and DltD. The D-alanyl carrier protein ligase DltA, as a remote homolog of acyl-(coenzyme A) (CoA) synthetase, cycles through two active conformations for the catalysis of adenylation and subsequent thiolation of D-alanine (D-Ala). The crystal structure of DltA in the absence of any substrate was observed to have a noticeably more disordered pocket for ATP which would explain why DltA has relatively low affinity for ATP in the absence of any D-alanyl carrier. We have previously enabled the thiolation of D-alanine in the presence of CoA as the mimic of D-alanyl carrier protein DltC which carries a 4’-phosphopantetheine group on a serine residue. Here we show that the resulting Michaelis constants in the presence of saturating CoA for both ATP and D-alanine were reduced more than 10 fold as compared to the values obtained in the absence of CoA. The presence of CoA also made DltA ~100-fold more selective on D-alanine over L-alanine. The CoA-enhanced substrate recognition further implies that the ATP and D-alanine substrates of the adenylation reaction are incorporated when the DltA enzyme cycles through its thiolation conformation.


Introduction
The cell surface of most Gram-positive bacteria contains wall teichoic acid and lipoteichoic acid with a poly-alditol phosphate backbone. The remaining hydroxyls of the alditol moiety are ubiquitously modified by D-alanyl esterification or glycosylation [1][2][3] . A dlt operon, which typically codes for DltA, DltB, DltC and DltD proteins, is required for the D-alanylation of lipoteichoic acids 4 . D-alanylation brings in positively charged amino-groups and partially neutralizes the net negative charges of the phosphate groups on the lipoteichoic acid backbone. The reduction of D-alanyl content on the cell surface has been found to be associated with increased autolysis 3,5 and susceptibility to host defense peptides and other antibiotics 6,7 . Impaired D-alanylation of lipoteichoic acid also reduces the ability of bacteria to colonize any surface 8 and form antibiotics-resistant biofilms 9 . Therefore, the dlt-encoded proteins required for the lipoteichoic acid D-alanylation pathway could serve as novel targets for fighting emerging infectious diseases caused by Gram-positive pathogens 10  The D-alanyl carrier protein ligase DltA (formed by ~500 amino acid residues) 4 closely resembles (~30% sequence identity) the adenylation domains (also called adenosine monophosphate (AMP)-forming domains) found in bacterial non-ribosomal peptide synthetases 12 . Its remote homologs include the acyl-and aryl-(coenzyme A) CoA synthetases and firefly luciferases 13 . As shown in Figure 1, DltA catalyzes the ATP-driven adenylation of D-alanine and the subsequent transfer of the activated D-alanyl group to the thiol group of 4'-phosphopantetheine, which is covalently bound to a serine side chain of D-alanyl carrier protein DltC (~80 amino acid residues) 4 . As previously shown for such one-protein-twoenzymes homologs 14,15 , crystal structures of two DltA proteins have also been observed in adenylation and thiolation conformations, respectively 16,17 (Figure 1). The thiolation conformation of B. subtilis DltA (BsDltA) 18 resembles previously determined structures of a bacterial acetate-CoA synthetase (ACS) and 4-chlorobenzoate-CoA ligase crystallized in their respective thiolation state 15,19 . On the other hand, the adenylation conformation of B. cereus DltA (BcDltA) 17 resembles those observed in the crystal structures of the firefly luciferase PheA (phenylalanine-activating domain of the first module of the Bacillus brevis gramicidin S synthetase I), DhbE (2,3-dihydroxybenzoate activation domain from B. subtilis), a yeast acetyl-CoA synthetase and 4-chlorobenzoate-CoA ligase 13,14,20-22 . The two conformations captured in the crystal structures of the two closely related DltA proteins (56% sequence identity) are related by a ~146º rotation around a hinge residue Asp-399 in BcDltA (Asp-398 in BsDltA), which is a invariable residue equivalent to Asp-517 in Salmonella typhimurium acetyl-CoA synthetase 23 . The region surrounding this hinge residue (residues Arg-397 to Glu-413 of BcDltA green in Figure 1B) contains an important triphosphateinteracting residue Arg-397 24 and a β-hairpin which serves as part of the pantetheine channel as observed in the bacterial acetyl-CoA synthetase 23 .
We previously studied the adenylation reaction of BcDltA in the absence of any D-alanyl carrier to enable the thiolation reaction 17 . The resulting specificity constants (k cat /K M ) for D-and L-alanine differed by merely ~3-fold, which could not explain the fact that teichoic acid is overwhelmingly modified by D-but not L-alanine 25 . As suggested by a study on a spectrum of DltA homologs which have residual aminoacyl-CoA synthetase activity 42 , CoA has been confirmed to be a substitute for the D-alanyl carrier protein DltC as the thiolation substrate of BcDltA 24 . Here we report further biochemical analysis of BcDltA. Noticeable differences were observed in Michaelis constants K M and turnover rates k cat . The presence of CoA, hence the enabled adenylation and thiolation cycle, enhanced the enzyme's apparent affinities to its cognitive substrate ATP and D-alanine by approximately an order of magnitude. Since BcDltA is a slow enzyme with turnover rate less than 1 s -1 , much slower than bacterial Acyl-CoA synthetases (~10 2 s -1 ) 23 and 4-chlorobenzoate-CoA ligase (~10 s -1 ) 26 , the observed Michaelis constants should closely approximate the corresponding dissociation constants, and therefore provide some insight into the stability of short-lived enzyme-substrate complexes. We also determined the structure of BcDltA in the absence of any substrate. This structure is noticeably more disordered than previously reported DltA structures 16,17,24 , which may explain the enzyme's lower affinity to ATP in the absence of the other two substrates. Interestingly, CoA-enhanced affinities to ATP and D-alanine imply that the thiolation substrate CoAbound BcDltA has higher affinity to both adenylation substrates as compared to CoA-free BcDltA.

Methods
Cloning, protein preparation and crystallization All reagents were from VWR unless specified otherwise. The wildtype as well as the C269A mutant of DltA from B. cereus was cloned for over-expression of BcDltA as described previously 17 . The pET28-BcDltA construct carries an Ala-1 mutation at the N-terminus and eight extra residues at the C-terminus (LEHHHHHH). The soluble fraction of BcDltA was purified by nickel-affinity chromatography followed by gel filtration. BcDltA was concentrated to ~20 mg/mL by ultra-filtration.
Crystallization and structure determination The concentrated BcDltA protein was crystallized using the hanging drop crystallization method at a room temperature of 21°C. The optimal well solution for crystallization contained 0.1 M MgCl 2 , 0.5 M KCl, 16% polyethylene glycol (PEG) 3,350 (Sigma-Aldrich) and 0.05 M Hepes-NaOH buffer at pH 7.2. Each drop was composed of 1 μL of protein and 1 μL of well solution. The plate-shaped crystals grew to a maximal size of 0.4 mm × 0.3 mm × 0.05 mm in three days. Crystals were gradually transferred to stabilizing solutions composed of the crystallization well solution supplemented with 8%, 16% and 24% glycerol, soaked for 1 minute, then flashcooled to -173.15°C in a nitrogen stream generated by an Oxford CryoSystems device. A total of 400 0.4-degree oscillation images were acquired and processed using a Brukers Proteum-R system as already described 27 . The previously solved BcDltA model (PDB code 3DHV) 17 was used as the starting model for two-domain rigid body refinement followed by positional refinement using Crystallography & NMR System (CNS) 28 . This resulting model was subjected to ten cycles of rebuilding and refinement using Arp/Warp 29 . The rebuilt model was iteratively rebuilt using XtalView 30 and then refined using CNS 28 . The final model had 90.8% of the residues in the most favored regions on a Ramachandran plot. Val-301 with clear electron density and Asp-336 with blurry electron density were the only two residues found in the disfavored region. Statistics of the diffraction data, refinement and geometry are listed in Table 1. The molecular figures were generated using Molscript 31 and rendered using Raster3D 32 . The coordinates and structure factors have been deposited in the Protein Data Bank 33-35 (entry code 4PZP).

Tryptophan fluorescence measurement
The intrinsic tryptophan fluorescence of 1.0 ml 0.4 uM BcDltA solution with 0 to 2 mM ATP was acquired at a room temperature of 21°C using a PerkinElmer LS-55 fluorescence spectrometer. The excitation wavelength was 305 nm and a fluorescent emission in the

Thiol quantification assay
The free thiol group of CoA was quantified as described previously 24 by a dye solution composed of 1 mM 5,5'-dithio-bis (2nitrobenzoic acid) (DTBN), or Ellman's reagent (Sigma-Aldrich) and 50 mM Tris-EDTA solution at pH 8.0. Absorption at a wavelength of 412 nm was used to quantify the concentration of free thiol group. The reaction rate was derived by the rate of thiol depletion. The correlation between initial reaction rate and substrate concentration was also fitted with Michaelis-Menten equation using the Prism software.

Results
Crystal structure of substrate-free BcDltA The same DltA protein from B. cereus with a C-terminal hexahistidyl fusion tag used in our previous crystallographic studies on BcDltA 17,24 was crystallized in the absence of ATP, D-alanine or CoA. One crystal diffracted to 1.9 Å resolution and belonged to space group P2 1 (Table 1), the same space group as in previously reported crystals of BcDltA in complex with D-alanine adenylate 17 and with ATP 24 . Despite having 8 Å shorter crystallographic a axis, 5 Å shorter b axis and 5° smaller β angle than the previously reported crystal of DltA/D-alanine adenylate complex, the structure was successfully solved by rigid-body refinement using the previously determined BcDltA structure 17 (PDB code 3DHV) as the starting model. The 504-residue BcDltA structure can be divided into two domains ( Figure 2): an N-terminal major domain from the N-terminus to Asp-399, and a C-terminal minor domain from residue 400 to the C-terminus. The disposition of the two domains in the substrate-free BcDltA structure remains similar to that of the starting model ( Figure 2A). The electron density map indicated several disordered regions (Ser-153 to Pro-159, Pro-363 to Glu-367, Arg-397 to Glu-413, Lys-433 to Tyr-440) with the corresponding regions in the starting model highlighted in magenta in Figure 2B. The first disordered region is part of a highly conserved P-loop (Thr-152 to Lys-160) found in homologous AMP-forming proteins 37-39 . Due to its similar amino acid composition (glycine, serine, threonine and lysine) to that of P-loop or Walker A motif found in ATPases and GTPases 40 , this loop has long been thought to catalyze the adenylation reaction. In the crystal structure of human medium-chain acyl-CoA synthetase in complex with ATP, this loop intimately interacts with the βand γ-phosphates of the ATP substrate 41 . Functional K M and k cat of BcDltA in the presence and absence of D-alanyl carrier CoA In our previous study, we have verified that CoA can mimic D-alanyl carrier protein DltC 24 , as also discovered for DltA homologs 42 . In that study, we have observed that the reaction rate is increased by nearly an order of magnitude by the presence of saturating concentration of CoA, which is explained by the faster release of the thiolation product rather than by release of the adenylation intermediate. In order to get a comprehensive understanding of the effects by CoA as the DltC mimic, we further studied the enzymatic properties of BcDltA in the presence of a saturating concentration (5 mM) of ATP or D-alanine, and in the absence of CoA or in the presence of a saturating concentration (5 mM) of CoA. The reaction rates derived from the pyrophosphate accumulation assay and the thiol depletion assay were similar ( Figure 4). The thiolation assay was noticeably noisier than the pyrophosphate assay and we therefore limit the discussion to K M and k cat values derived from the pyrophosphate assay. Somewhat unexpectedly, BcDltA showed much higher apparent affinity, or decreased K M value, towards ATP (0.46 mM to 0.01 mM) and D-alanine (1.1 mM to 0.03 mM) in the presence of 5 mM CoA (Figure 3 and Figure 4, Table 2). On the contrary, the apparent affinity towards L-alanine decreased in the presence of CoA, with K M increased from 14.4 mM to 109 mM ( Figure 4 and Table 2).
Relaxed D-alanine preference by the C269A BcDltA mutant protein The side-chain of Cys-269 sits at the bottom of the D-alaninebinding pocket which may make VDW clash with the methyl sidechain of L-Alanine 16,17 . We also studied the effect of CoA on D-and L-alanine preference of the C269A mutant of BcDltA ( Figure 4 and Table 2). In the presence and absence of CoA, the C269A protein showed relaxed preference for D-alanine over L-alanine. As observed for the wild-type protein, CoA also enhances the D-alanine preference

Pyrophosphate assay (5 mM D-Ala, -CoA) (this work)
Wild-type (ATP) 0.46 ± 0.05 1.46 ± 0.04 3.2 ± 0.04 also implies that the previously observed adenylation and thiolation conformations are intrinsically unstable unless stabilized by the interaction with one or more substrates. This structural feature of BcDltA likely explains the relatively low sub-millimolar affinity for ATP, since part of the stabilizing BcDltA-ATP interactions would be used to compensate the cost of establishing the adenylation conformation of the protein.
BcDltA is a very slow enzyme. Unless the substrate dissociation step happens to be extremely slow as well, we could approximate the observed K M values to the K D values of corresponding BcDltAsubstrate intermediates, and therefore enable reasoning in the context of structural stability of such intermediates. As such, we reasoned that the approximately one order of magnitude difference in K M values for the adenylation substrates ATP and D-alanine observed in the presence and absence of the thiolation substrate may imply the existence of a quadruple intermediate of the DltA enzyme in complex with all three substrates, which may be markedly different from a ternary intermediate of BcDltA with the two adenylation substrates.
We then resorted to three-dimensional model building so as to answer the question on which of the adenylation and thiolation conformations may be compatible with binding to all three substrates. The BcDltA/D-alanine-adenylate complex (PDB entry 3DHV) 17 was chosen as the adenylation conformation and as the reference set of atomic coordinates. The adenylated intermediate was dissected to generate the D-alanine model. The N-terminal domain of the BcDltA/ATP complex (PDB entry 3FCE) 24 , which is also in the adenylation conformation, was superposed on the reference set to orient the ATP substrate. The N-terminal domain of the BsDltA/ AMP complex (PDB entry 3E7W) was also superposed on the reference set to derive the re-oriented thiolation conformation. Mainchain atoms equivalent to those interacting with AMP in BcDltA (270-272, 292-299) in the quadruple complex of acetyl-CoA synthetase in its thiolation conformation (PDB entry 2P2F) 23 were superimposed on the reference set to orient the CoA model. In the adenylation conformation ( Figure 5A, the pantetheine channel is apparently blocked by the main-chain atoms immediately preceding the catalytic Lys-492 of BcDltA (Lys-491 of BsDltA). Although we could not completely rule out the possibility that an allosteric site for CoA exists, no such site has ever been observed for this superfamily of enzymes. Therefore, a quadruple complex in the adenylation conformation is unlikely. The thiolation conformation ( Figure 5B), on the other hand, appears to be compatible with binding ATP so long as Arg-408 of BcDltA (Arg-407 of BsDltA) adopts another rotamer. It has been previously observed that the homologous acetyl-CoA synthetase in its thiolation conformation binds AMP, acetate and CoA 23 , and that BsDltA in its thiolation conformation binds AMP and appears to have a well-formed D-alaninebinding pocket 16 . These structural evidences seem to suggest that a quadruple intermediate may form with the enzyme in its thiolation conformation but not in the adenylation conformation. Since the thiolation conformation has an AMP-binding site and its N-terminal domain, which provides most of the ATP-interacting residues such as the P-loop and Arg-397, remains essentially identical to that that in the adenylation conformation, it is not surprising that only one arginine side-chain is required to adopt another rotamer to accommodate ATP. This arginine residue (Arg-408 of BcDltA) forms a salt bridge with the divalent cation-anchoring side-chain of Glu-298 in BcDltA (Glu-297 in BsDltA) 24 , which appears to be an important structural feature to modulate the conformational change 16 . Another arginine residue (Arg-397 of BcDltA, Arg-396 of BsDltA) also seems to facilitate the conformational change. It forms a salt bridge with the hinge aspartate residue in the thiolation conformation. In the adenylation conformation, this arginine sidechain adopts a more extended rotamer and forms a salt bridge with the β-phosphate group of ATP ( Figure 5A). ATP binding would result in re-orientation of both arginine residues and disruption of two salt bridges, thus mobilizing the thiolation conformation.
Structural basis for CoA-enhanced affinity for ATP The majority of ATP-binding elements lie in the N-terminal domain which remains similar in both adenylation and thiolation conformations. The most significant structural feature in the C-terminal domain for binding ATP is the catalytic Lys-492 of BcDltA (Lys-491 of BsDltA) in the adenylation conformation. In the thiolation conformation, the catalytic residue is replaced by Lys-403 (Lys-402 of BsDltA) ( Figure 5). In addition, the thiolation conformation is more ordered than the substrate-free conformation. The favourable DltA/ATP interactions may no longer be used to compensate the energetic cost of stabilizing the disordered hinge region. Therefore the affinity for ATP by the CoA-bound BcDltA in its thiolation conformation should be higher than substrate-free BcDltA.
Structural basis for CoA-enhanced enantiomer selectivity for D-alanine In the adenylation conformation of BcDltA/D-alanine adenylate, the shortest distance from Phe-196 side-chain to methyl group of D-alanine is 4.15 Å. In the modelled quadruple complex in the thiolation conformation, the Phe-196 side-chain does not contact D-alanine but the thiol group of CoA lies in closer proximity of D-alanine side-chain (3.34 Å), as expected from the pending thiolation reaction between the thiol group of CoA and the carbonyl group of D-alanine-adenylate. On the other hand, the tighter alanine-binding pocket may exert stronger VDW repulsion toward L-alanine, thus lowering further the affinity for the wrong enantiomer. Interestingly, Cys-269 also has a thiol group. The removal of this group in the C269A mutant of BcDltA reduces the affinity for D-alanine and increases the affinity for L-alanine by approximately one order of magnitude, which is almost exactly the opposite to the effect of introducing the thiol group of CoA in the alanine-binding pocket. Since sulphur atoms are larger and more inducible than the second-period elements carbon and oxygen, it is not surprising that the VDW interaction involving a thiol group makes a significant impact on the stability of enzyme/substrate complex.
Hypothesized enzymatic cycle of DltA Intracellular concentration of D-alanine 43 generally exceeds the K M value in the presence of saturating ATP and CoA. Typical intracellular concentration of ATP lies in the millimolar range 45 , exceeding the K M value for ATP as well. The concentration of the possibly stress-induced DltC 44 may also reach saturating level in bacteria when such stress is present. As implied by the CoA-triggered dramatic change in K M values for ATP and D-alanine, the two adenylation substrates are likely incorporated by the enzyme with prebound CoA rather than with merely the other adenylation substrate. In addition, three-dimensional modelling suggests the CoA-bound state of the enzyme can only exist in its thiolation conformation. We therefore hypothesize a two-conformation model for the enzymatic cycle catalyzed by DltA ( Figure 6) in the presence of saturating concentration of the three substrates. Our model differs from the three-conformation model proposed for DltA and adenylation domains in non-ribosomal peptide synthetase 16 which includes a third substrate-free conformation. This model also differs from the two-conformation model proposed for and for 4-chlorobenzoate-CoA ligase 26,46 which includes substrate-free state of the enzyme. The two previously proposed models are consistent with a typical Ping-Pong mechanism. Neither a substrate-free conformation nor a substrate-free state of the enzyme is required in our model for the enzymatic cycle of DltA once the protein enters the reaction cycle. The adenylation reaction starts from a ternary complex with ATP and D-alanine and proceeds with the release of pyrophosphate. A domain rotation around the aspartate hinge residue follows transforming into the thiolation conformation which binds CoA, or DltC in bacteria, and catalyzes the thiolation reaction with D-alanine by displacing AMP. In our model, the resulting complex with the two thiolation products proceeds with AMP release, D-alanine-CoA/CoA exchange, D-alanine binding, and ATP binding. In the quadruple complex, the ATP-bound magnesium ion would replace Arg-408 of BcDltA in forming a bridge with Glu-298, as observed in the crystal structure BcDltA/ATP complex 24 . The disruption of the Arg-408 to Glu-298 salt bridge would destabilize the thiolation conformation and facilitate a reverse domain rotation and release of CoA which is not compatible with the adenylation conformation. The hypothesized D-alanine-CoA/CoA exchange step is the central piece of this enzymatic cycle, which reflects the finding that CoA affects DltA's apparent affinities with D-alanine and ATP. If DltA were to become substrate-free for a long enough period of time, it would form a ternary complex with ATP and D-alanine, which contradicts the observed effect by the thiolation substrate CoA. It is worth noting that the sequence of D-alanine and ATP binding is uncertain. Similarly, AMP release has to occur before ATP binding but not necessarily before exchange with CoA or D-alanine binding.
The hypothesized enzymatic cycle involves a second CoA binding step, and a CoA release step in addition to the Ping-Pong mechanism previously proposed for DltA and its homologs. Both additional steps seem unnecessary for the enzymatic reaction itself, but are required to explain our enzymatic data and are consistent with the three-dimensional models of BcDltA and BsDltA. A Ping-Pong mechanism would require that the binding of CoA and the binding of either adenylation substrate be uncompetitive, and therefore the apparent K M and k cat values for either adenylation substrate both would become larger at higher CoA concentration. While the k cat values for both ATP and D-alanine did become larger at saturating CoA concentration, the K M values actually became smaller, therefore contradicting the typical Ping-Pong mechanism. Another property of aminoacyl-CoA synthetases including DltA is the inhibitory effect of CoA at high concentration 42 , which is difficult to explain by a typical Ping-Pong mechanism unless the enzymatic cycle includes an additional CoA-dissociation step as in our model. The closest homologs of DltA are amino acid-activation domains found in non-ribosomal peptide synthetases 42 . Similar to DltA, these homologs also pass the adenylated intermediate to the 4'-phosphopantetheine group attached to a serine residue on a peptide carrier domain. It is possible that such amino acid activation domains may also act as DltA.
The extra binding step for 4'-phosphopantetheine D-alanyl carrier could serve as the sensor for the availability of the carrier. The maximum rate catalyzed by BcDltA is approximately seven times faster in the presence of CoA than in its absence, with respective k cat values 10.9 min -1 and 1.5 min -1 . Moreover, the intracellular concentration of D-alanine is typically found in the 100 micromolar range 43 , which lies above the K M value for D-alanine in the presence of CoA (~30 μM) but below the K M value in the absence of CoA (1100 μM) ( Table 2). The reaction rate should become slower by approximately 100 fold when the thiol carrier is absent. It is worth noting that the adenylated D-alanine intermediate generated by the adenylation reaction is not covalently attached to the enzyme and could be released and wasted when the thiolation substrate is absent. The significant slowing down of the adenylation reaction in the absence of the 4'-phosphopantetheine carrier therefore provides a biological advantage.  (1) an adenylation reaction in which ATP is joined to D-Ala to form D-Ala-AMP (PPi is released), and (2) a thiolation reaction in which D-Ala is transferred from D-Ala-AMP to a phosphopantetheine group of D-alanyl carrier protein (DltC). The reaction can be mimicked in vitro by substituting CoA for DltC (does this also happen in vivo?). Previous crystal structures have indicated that the two reactions are promoted by two distinct conformational states of the enzyme, which differ by a large rearrangement of a C-terminal domain that closes over the active site. In this work, Du & Luo provide kinetic data to suggest that binding of the two substrates for the adenylation reaction (ATP and D-Ala) is promoted by the presence of the substrate (CoA) for the second reaction. The selectivity for binding of D-Ala as opposed to L-Ala is also enhanced significantly by the presence of CoA. A new scheme for the reaction cycle is proposed in which instead of forming an empty enzyme when the final product D-Ala-CoA is released, free CoA is rapidly "exchanged" in for D-Ala-CoA to keep the enzyme in its thiolation conformation. Then new D-Ala and ATP substrates bind (with the enzyme still in its CoA-bound thiolation conformation), and CoA is released to allow the enzyme to adopt the adenylation conformation, which is required for ATP and D-Ala to react. Binding of CoA to the newly formed D-Ala-AMP complex then puts the enzyme back in the thiolation conformation. The new reaction scheme is considerably more complicated than one would expect, but seems to account for the observed kinetic data. The authors propose that the biological purpose for this type of mechanism is to limit binding (and reaction) of ATP and D-Ala to conditions in which the 2 substrate (presumably DltC in vivo) is available, so as to avoid non-productive formation and release D-Ala-AMP.

Grant information
A crystal structure of the apo form of DltA is also determined, which shows that in the absence of substrates the enzyme adopts the adenylation conformation, and that several loops on the enzyme are considerably more disordered than in previous structures of various substrate or product-bound states. The new structure does not add a lot, but is meant to support the concept that ATP and D-Ala do not bind tightly to the enzyme unless CoA is around to reduce the overall flexibility.
The manuscript is reasonably well written, scientifically sound, and should be of fairly broad interest, particularly for those studying other types of two-reaction enzymes, such as in non-ribosomal peptide synthesis, and possibly even enzymes like amino-acyl tRNA synthetases. I found the manuscript to be rather uninteresting, until I got to the end of the discussion, where the newly proposed reaction scheme of Figure 6 is presented, and a biological purpose of the observed phenomena is proposed. If there is some way to bring these aspects into the Abstract or the Introduction, a potential reader might be drawn in a little bit earlier than I was. nd 1.
The proposed scheme is quite unusual in that CoA supposedly promotes binding of ATP and D-Ala, but subsequent release of CoA is required for the two to react. Presumably this means that the rate of the catalytic step for the adenylation reaction must be considerably faster than the rates of ATP and D-Ala dissociation, or else the enzyme could return to the apo state, and presumably the adenylation conformation, when CoA is released. Are these parameters known? Also, is the physiological concentration of DltC acceptor protein, which presumably takes over the role of CoA in vivo, high enough to be consistent with the proposed reaction cycle? It would need to rebind to the enzyme in the exchange step before the enzyme reverts to the adenylation conformation. Presumably the rate of DltC binding is faster than the rate of reversion back to the adenylation conformation? The authors should discuss/clarify these points.
Minor points: The data in this paper demonstrate the effect on the enzyme by CoA, a mimic of the in vivo substrate DltC, but not for DltC itself. Is there some reason why this is difficult to demonstrate with DltC experimentally?
Was it previously known from other structures that the apo form of the enzyme is in the adenylation conformation, or is this a new insight? In general the authors could do a little bit of a better job describing the previous structures in the Introduction. Why were the previous structures of the B. enzyme found to be in the thiolation conformation despite the absence of CoA (or DltC)? Is subtilis there no structure available with CoA?
References 16 and 18 are the same. As far as we know, the rate constants of substrate binding, reaction and product dissociation steps have not been measured for DltA or its closely related homologs that transfer the activated amino acyl or peptidyl group to a carrier protein. We do not have expertise in measuring such rate constants.
It is generally accepted that DltC is the D-alanine carrier . CoA, however, is a more feasible in vivo analog of DltC for enzymatic assays. If DltC were used, the assay would require either millimolar concentration of DltC or addition of a thioesterase to regenerate DltC, for which we haven't resolved the technical difficulties.

The
DltA in thiolation conformation has been co-crystallized with DltC, although DltC is B. subtilis