A step-by-step in crystallo guide to bond cleavage and 1,6-anhydro-sugar product synthesis by a peptidoglycan-degrading lytic transglycosylase

Lytic transglycosylases (LTs) are a class of enzymes important for the recycling and metabolism of peptidoglycan (PG). LTs cleave the β-1,4-glycosidic bond between N-acetylmuramic acid (MurNAc) and GlcNAc in the PG glycan strand, resulting in the concomitant formation of 1,6-anhydro-N-acetylmuramic acid and GlcNAc. No LTs reported to date have utilized chitins as substrates, despite the fact that chitins are GlcNAc polymers linked via β-1,4-glycosidic bonds, which are the known site of chemical activity for LTs. Here, we demonstrate enzymatically that LtgA, a non-canonical, substrate-permissive LT from Neisseria meningitidis utilizes chitopentaose ((GlcNAc)5) as a substrate to produce three newly identified sugars: 1,6-anhydro-chitobiose, 1,6-anhydro-chitotriose, and 1,6-anhydro-chitotetraose. Although LTs have been widely studied, their complex reactions have not previously been visualized in the crystalline state because macromolecular PG is insoluble. Here, we visualized the cleavage of the glycosidic bond and the liberation of GlcNAc-derived residues by LtgA, followed by the synthesis of atypical 1,6-anhydro-GlcNAc derivatives. In addition to the newly identified anhydro-chitin products, we identified trapped intermediates, unpredicted substrate rearrangements, sugar distortions, and a conserved crystallographic water molecule bound to the catalytic glutamate of a high-resolution native LT. This study enabled us to propose a revised alternative mechanism for LtgA that could also be applicable to other LTs. Our work contributes to the understanding of the mechanisms of LTs in bacterial cell wall biology.


Lytic transglycosylases (LTs) are a class of enzymes important for the recycling and metabolism of peptidoglycan (PG). LTs
cleave the ␤-1,4-glycosidic bond between N-acetylmuramic acid (MurNAc) and GlcNAc in the PG glycan strand, resulting in the concomitant formation of 1,6-anhydro-N-acetylmuramic acid and GlcNAc. No LTs reported to date have utilized chitins as substrates, despite the fact that chitins are GlcNAc polymers linked via ␤-1,4-glycosidic bonds, which are the known site of chemical activity for LTs. Here, we demonstrate enzymatically that LtgA, a non-canonical, substrate-permissive LT from Neisseria meningitidis utilizes chitopentaose ((GlcNAc) 5 ) as a substrate to produce three newly identified sugars: 1,6-anhydro-chitobiose, 1,6-anhydro-chitotriose, and 1,6-anhydro-chitotetraose. Although LTs have been widely studied, their complex reactions have not previously been visualized in the crystalline state because macromolecular PG is insoluble. Here, we visualized the cleavage of the glycosidic bond and the liberation of GlcNAc-derived residues by LtgA, followed by the synthesis of atypical 1,6-anhydro-GlcNAc derivatives. In addition to the newly identified anhydro-chitin products, we identified trapped intermediates, unpredicted substrate rearrangements, sugar distortions, and a conserved crystallographic water molecule bound to the catalytic glutamate of a high-resolution native LT. This study enabled us to propose a revised alternative mechanism for LtgA that could also be applicable to other LTs. Our work contributes to the understanding of the mechanisms of LTs in bacterial cell wall biology.
Lytic transglycosylases (LTs) 3 degrade peptidoglycan (PG) to produce GlcNAc and 1,6-anhydro-N-acetylmuramic acid (1,6anhydro-MurNAc), a key cytotoxic elicitor of harmful innate immune responses (1). Many bacteria have redundant LTs; for example, Escherichia coli has eight identified LTs (MltA, MltB, MltC, MltD, MltE, MltF, MltG, and Slt70). Similarly, Neisseria species can potentially express five LTs (LtgA, LtgB, LtgC, LtgD, and LtgE) (2)(3)(4). Recent studies have provided insights into the specific roles of LtgA and LtgD in Neisseria gonorrhoeae by showing that their primary function is to release PG monomers capable of activating innate immune signaling and ciliated cell death. These studies further demonstrate that LtgA produces cytotoxic PG monomers that are largely taken into the cytoplasm for recycling, whereas LtgD produces PG monomers that are released (5). LtgA and LtgD display unique substrate specificities: unlike other LTs, they can digest synthetic tetrasaccharide dipeptides into disaccharide products (5). These findings imply that LtgA and LtgD are required to release greater amounts of PG monomers during Neisseria sp. infection and therefore lack strict substrate specificity.
LTs are classified into four distinct families based on sequence similarity and consensus sequences. The LTs in subfamily 1 share sequence and structural similarity to the goosetype lysozyme (6). Based on the classification of Blackburn and Clarke (6), family 1 can be further subdivided into five subfamilies (1A to 1E) with distinct secondary structures. Despite the secondary structural differences in LTs, the catalytic residues in the active site are fairly well-conserved.
LT activity is vital to the synthesis and remodeling of PG. Previous structural and biochemical studies revealed that LTs utilize a single catalytic glutamate or aspartate in acid/base catalysis ( Fig. 1) (7). LTs are inhibited by a ␤-hexosaminidase inhibitor (NAG-thiazoline) and bulgecin A (7)(8)(9)(10)(11)(12)(13)(14). The inhibition of LTs by NAG-thiazoline in particular established a direct link between the formation of a 1,6-anhydro-muramoyl residue and the oxazolinium ion intermediate in stage 1 of the proposed mechanism ( Fig. 1) (7). Because LTs have no known second catalytic residue, the reaction is widely accepted to proceed via the anchimeric assistance of the MurNAc 2-acetamido group and the formation of the oxazolinium ion intermediate ( Fig. 1) (7). In stage 2 of the mechanism, the catalytic glutamate acts as a general base to abstract the proton from the C-6-hydroxyl of the MurNAc residue, leading to nucleophilic attack on the anomeric center of the MurNAc residue and the formation of 1,6anhydro-MurNAc ( Fig. 1) (7). The crystal structures of several LTs in complex with non-reactive substrates or product analogs lend support to the proposed mechanism (10,(15)(16)(17)(18)(19)(20)(21). Here, we capture an active LT in the crystalline state cleaving the glycosidic bond to (a) liberate a single GlcNAc sugar and (b) synthesize 1,6-anhydro-GlcNAc-derived products. The signature product of the LT reaction is a 1,6-anhydro-MurNAc derivative. Unexpectedly, LtgA utilized chitopentaose as a substrate and produced 1,6-anhydro-GlcNAc-derived sugars. These high-resolution native structures accompanied by reactions in the crystalline state revealed previously unobserved features and intermediate steps, including, conformational rearrangements of the LT active site during the final stage of 1,6-anhydro-product formation. In the high-resolution native structures of LtgA, a conserved crystallographic water molecule bound to the catalytic glutamate of LtgA was identified. Altogether, this study provides visual insights into how an LT utilizes its substrate during PG metabolism and the conforma-

Visualizing bond cleavage by a lytic transglycosylase
tional path of the enzyme during catalysis and also sheds light on the likely mechanism of LTs. These insights could impact the development of future antibiotics that target LTs.

Discovery of a lytic transglycosylase with chitinase activity in vitro
Most LTs are inactive toward chito-oligosaccharides, although as a polysaccharide, the chitin polymer mimics the chemical composition of the PG glycan strand (Fig. 2) (22). The ␤-1,4-glycosidic bond, which is the chemical site of LT activity, is common to the PG and chitin sugar polymer substructures. However, whereas chitin consists of GlcNAc monomers, PG consists of repeating units of MurNAc and GlcNAc. The Mur-NAc residues linked to the stem peptides are essential to the formation of macromolecular PG and were predicted to be essential to the binding and activity of most LTs.
Serendipitously, we discovered that LtgA can metabolize chitopentaose in vitro. HPLC was used to separate the metabolites formed during the cleavage of chitopentaose by LtgA and was coupled with high-resolution MS to determine the chemical structure of the resulting products (Fig. 3, a and b, and Fig. S1, a and b). The single peak in the HPLC trace and the mass spectrum of the starting substrate revealed that the ligand was chitopentaose and was not contaminated by other chito-oligosaccharides (Fig. S1, a and b). LtgA cleaved chitopentaose to yield chitobiose, chitotriose, chitotetraose, and three newly identified products: 1,6-anhydro-chitobiose, 1,6-anhydro-chitotriose, and 1,6-anhydro-chitotetraose (Fig. 3, a and b, and Fig. S1, a and b). Although 1,6-anhydro-GlcNAc has not been documented as a product of LTs, LtgA can yield this molecule in vitro. To understand the atomic details of how LTs synthesize these anhydro-chitin products and to acquire insights into the LT reaction, we performed crystallographic studies of LtgA complexes.

The structure of native LtgA
Three structures of native LtgA were determined at resolutions of 1.3, 1.4, and 1.85 Å at pH values of 6.5, 7.5 and 9.5, respectively (Tables S1 and S2 and Fig. 4a). The overall structure of LtgA is a heart-shaped, highly superhelical configuration consisting of 37 ␣-helices organized into three domains: U, C, and L (Fig. 4a). LTs have very diverse secondary structures, but their catalytic domains are highly conserved (Fig. 4, a and b, and Fig. S2, a-d) (23). LtgA exhibits an overall weak sequence similarity to Slt70 (25%); however, the active site is conserved (Fig. 4, a and b). The architecture of the active site of LtgA is formed by a total of 10 ␣-helices (␣28, ␣29, ␣30, ␣31, ␣32, ␣33, ␣34, ␣35, ␣36, and ␣37; Fig. S2, a-d). A six-helix bundle (␣29, ␣30, ␣31, ␣32, ␣33, and ␣34) constitutes the active site core, which secures the glycan chain. The proposed catalytic residue Glu 481 is absolutely conserved in the LT family, whereas Gln 480 , Val 492 , Met 501 , Thr 504 , Glu 507 , Asn 528 , Tyr 532 , Glu 580 , Tyr 551 , Arg 557 , and Tyr 584 are highly conserved (Fig. 4, a and b). LtgA has three bona fide saccharide-binding sites (labeled Ϫ1 to Ϫ3) that form the active site core and two weaker binding sites (labeled Ϫ4 and ϩ1) at the entrance and exit of the active site pocket. The majority of the interaction between LTs and the glycan chain would occur within the cleft containing the binding sites Ϫ1 to Ϫ3 (Fig. 4c). Binding site Ϫ1 is buried and protected from the solvent and is closest to the previously identified catalytic residue Glu 481 (Fig. 4c). The ϩ1 and Ϫ1 subsites represent the catalytic center where LtgA catalyzes the reaction.

Catalysis in crystallo
To visualize the reaction of LtgA at the molecular level, the enzymatic reaction of LtgA was initiated in crystallo by soaking a five-moiety glycan sugar (chitopentaose) into native crystals of LtgA grown under crystallization conditions with buffers at pH 6.5, 7.5, and 9.5. At least 30 crystals were monitored over time for the appearance of new density reflecting enzymatic reaction products. The F o F c map was compared with the refined native structures of LtgA. In most cases, the native state was retained (Fig. 5a), but at the 5-min time point at pH 6.5 and 7.5, we visualized a clear LT reaction intermediate or product. At pH 7.5, two trapped intermediates (chitotetraose and a GlcNAc sugar) were observed in the active site (Fig. 5b). The observed chitotetraose intermediate was trapped in the active site core, whereas a single GlcNAc residue was poised in the departure position at the active site exit (Fig. 5b). The chitotetraose intermediate occupied subsites Ϫ4 to Ϫ1 of the active site (Fig. 4c). Cleavage of the glycosidic bond occurred between subsites Ϫ1 and ϩ1 (Fig. 5b). Remarkably, in the absence of the PG stem peptides, LtgA cleaved a single GlcNAc residue. This behavior had not been observed previously and could not be predicted, because the LT reaction normally produces dimers, trimers, or tetramers when PG is used as a substrate (24). In this intermediate, the GlcNAc residue in the Ϫ1 position is significantly distorted and assumes a half-chair position (Fig. 5b). The C1-hydroxyl is rotated into the ␣ position ( Fig. 1). Notably, the N-acetyl of the Ϫ1 GlcNAc appears to be mobile, because the electron density surrounding this N-acetyl group is sparse (Fig. 5b). In the Ϫ1 position, the C6-hydroxyl group of the GlcNAc residue is within hydrogen-bonding distance of the (OE2) oxygen of the catalytic residue Glu 481 . The GlcNAc sugar in the Ϫ1 position is further supported in the active site by Three novel metabolites were identified in peak e: 1,6-anhydro-chitobiose (molecular weight, 406.1588), 1,6-anhydro-chitotriose (molecular weight, 609.2344), and 1,6-anhydro-chitotetraose (molecular weight, 812.3203). Most chitobiose (molecular weight, 427.1925) was found in peak b, chitotriose (molecular weight, 629.2656) was found in peak c, and chitotetraose (molecular weight, 832.3438) was found in peak d. Undigested chitopentaose was identified in peak aЈ and the starting substrate in peak a (please see Fig. S1 for detailed MS analyses). The asterisk signifies the identification of multiple products in the same peak.

Visualizing bond cleavage by a lytic transglycosylase
Asn 552 and Glu 580 (Fig. 5b). Additionally, Glu 580 most likely provides structural support for the 2-acetamido group during catalysis. Ser 490 is hydrogen-bonded to the C1-hydroxyl group of chitotetraose and locks it into the axial position. Tyr 584 is positioned for direct hydrogen bonding to Glu 481 and provides a proton relay with the catalytic residue.
As shown in Fig. 5b, the C4-hydroxyl group of the departing GlcNAc residue in the ϩ1 position is hydrogen-bonded to the oxygen (OE1) of the catalytic residue (Glu 481 ) (Fig. 5b). The nitrogen in the N-acetyl group of the ϩ1 GlcNAc is hydrogen-bonded to the backbone oxygen of the catalytic residue, whereas the NE2 of Gln 480 is hydrogen-bonded to the C6-hydroxyl group. The departing GlcNAc occupies the same space as the GlcNAc sugar that was previously observed in the Slt70 -1,6-anhydro-muropeptide product (21). Normally, the glycosidic oxygen is expected to depart with the ϩ1 GlcNAc after bond breaking in the reaction. However, in X-ray data, an extra oxygen atom was observed. This extra oxygen is potentially derived from a water molecule.
In this trapped chitotetraose/GlcNAc intermediate, the distance between the cleaved glycosidic bond (C1-hydroxyl) of chitotetraose and the catalytic residue is precisely 5.4 Å, whereas the C4-hydroxyl group of the departing GlcNAc residue is precisely at hydrogen-bonding distance (2.5 Å) from the catalytic residue. Previous structural studies of E. coli LTs, such as MltA, MltC, and MltE, with unreactive glycan strand mimics demonstrated that the catalytic residue is perfectly positioned for its role as a base (2.5-3.0 Å) but is too far (3.6 -5.4 Å) from the primary chemistry site (the ␤-1,4-glycosidic bond) (Fig. S3, a and b) (11,(25)(26)(27)(28)(29)(30)(31). MltA undergoes a drastic conformational change to achieve a nearly productive configuration of the active site (28,32). However, the structure of MltA with bound chitohexaose demonstrates that the catalytic residue is still distant from the ␤-1,4-glycosidic bond (32). This distance suggests that further structural rearrangements are necessary for the reaction to proceed. To address this point, we surveyed additional conditions to trap more informative intermediates.

1,6-Anhydro-product formation
At pH 6.5, we observed the formation of 1,6-anhydro-chitotriose with Zn 2ϩ directly bound to the catalytic residue Glu 481 and coordinated by five water molecules. The water-coordinated Zn 2ϩ molecule stabilizes the trapped 1,6-anhydro-chitotriose product after the departure of the GlcNAc sugar(s) (Fig.  5b). Zn 2ϩ was required as an additive to stabilize the native crystals at pH 6.5 but was not required for the activity of LtgA in vitro (Fig. 3, a and b). Additionally, in the native structure of crystals of LtgA grown at pH 6.5, no density corresponding to a Zn 2ϩ atom was observed in the active site of LtgA ( Fig. 5a and Fig. S2).
In this structure, the O6 oxygen of the 1,6-anhydro-GlcNAc product bound in the Ϫ1 position is directly hydrogen-bonded to the OE2 of the catalytic residue Glu 481 . The O7 oxygen of the N-acetyl group of the 1,6-anhydro-GlcNAc is directly hydrogen-bonded to Ser 490 , which was previously involved in locking the C1-hydroxyl group of chitotetraose in the axial position. Overall, the 1,6-anhydro-GlcNAc in the Ϫ1 position forms fewer hydrogen-bonding contacts in the active site than the GlcNAc residue in the Ϫ1 subsite of the chitotetraose (Fig. 5b). For example, the product is no longer hydrogen-bonded to Glu 580 or Gln 480 and does not interact directly with Gln 499 (Fig. 5b). The formation of the 1,6-anhydro product represents the final stage of the catalytic mechanism before the product release (Fig. 1). Interestingly, the 1,6 bond of 1,6-anhydro-GlcNAc is now positioned for direct hydrogen bonding to the catalytic residue, signaling overall structural changes in the active site. Previous studies predict binding sites Ϫ1 and Ϫ3 to be occupied by the 1,6anhydro-MurNAc products.

Visualizing bond cleavage by a lytic transglycosylase Structural rearrangement of the active site of LtgA
The terminal product of the LT reaction is a product containing 1,6-anhydro-muramic acid. To date, all structures of LTs with unreactive glycan chain analogs (MltA-chitohexose, MltE-chitopentaose, and MltC-tetrasaccharide) show distances of 4.66 -5.4 Å between the glycosidic bond and the catalytic residue (Fig. S4) (11,29,32). Because the substrates in these LT complexes are unreactive, they presumably represent nonproductive enzyme-substrate complexes or mimic the resting state of the enzyme. Therefore, for catalysis to occur, (a) elements of the active site must move to bridge the gap, (b) glycan sugars must rearrange into a sterically acceptable conformation, or (c) the conserved water molecule identified in our native structures must form a proton relay between the glycosidic bond and the catalytic residue, which could enable the reaction to proceed without large active site readjustments (Fig.  6, a and b, and Fig. S4). With mounting evidence from known structures suggesting a missing link, it was deduced that the highly decorated PG, the natural substrate of LTs, is likely bound in a more catalytically competent manner than unreactive substrate analogs in complexes with LTs. However, in the crystalline state, LtgA, unlike other known LTs, competently

Visualizing bond cleavage by a lytic transglycosylase
degrades chitopentaose to produce 1,6-anhydro-sugars. Surprisingly, the catalytic domain of the LtgA-1,6-anhydro-GlcNAc complex (Fig. 6a, gray) showed a root mean square deviation of 1.8 Å when superimposed on all other native or trapped complexes (Fig. 6a, pink). Dramatic localized deviations of 3.2-7.0 Å appeared in the catalytic center (Fig. 6, a and  b). Specifically, significant perturbation occurred in ␣30, ␣31, and the loop connecting ␣29 to ␣30 (Figs. 2a and 6a). LtgA is an exolytic enzyme, and in the substrate-bound version, the loop that connects ␣29 to ␣30 restricts the number of residues that the active site can accommodate via steric hindrance (Fig. 6b). Gln 499 is on this flexible loop and supports the departing sugar residues in the ϩ1 position (pH 6.5) during catalysis. However, it does not contact the 1,6-anhydro-sugar product. This flexible loop recedes after product formation (Fig. 6b). Globally, in product formation mode, the active site is in a more open conformation than in all other forms of LtgA (Fig. 6, a-c). The more open conformation of the LtgA-1,6-anhydro-GlcNAc complex may facilitate product release.

Overall view of LtgA PG-binding site
LTs are specific for PG as a substrate. The studies herein were conducted in the absence of the PG stem peptide, and therefore, we wanted to obtain a global view of PG binding to LTs. Because substructures of PG that included the stem peptide were unavailable in the quantity and purity necessary for crystallization, a docking model of LtgA-chitotetraose complexed with a tripeptide that terminates in a meso-diaminopymelyl (meso-DAP) residue was created (Fig. 7).
The majority of the interaction between LTs and chitotetraose occurs at binding sites Ϫ1 through Ϫ4 (Fig. 5b). Binding sites Ϫ2 and Ϫ4 easily accommodate GlcNAc residues, but the attachment of the PG stem peptides would create significant steric hindrance (Fig. 7). Binding sites Ϫ1 and Ϫ3 accommodate the MurNAc residues and attached stem peptides (Figs. 5b  and 7). A groove that can accommodate the stem peptide core extends from binding sites Ϫ1 and Ϫ3. Moreover, binding site Ϫ1 is aligned with conserved residues that were previously identified in Slt70 to be crucial for PG stem peptide binding (22). These residues are conserved in LtgA (Figs. 3b and 5b).
The docked stem peptides bound in binding sites Ϫ1 and Ϫ3 are flanked by the highly conserved residues Glu 580 , Phe 578 , Thr 575 , Trp 561 , and Arg 559 (Fig. 7). This putative binding site of PG stem peptides was previously inferred from the high-resolution crystal structure of Slt70 complexed with 1,6-anhydromurotripeptide (22).

Discussion
This study has captured LtgA in action and provides the first visual snapshots of how LTs liberate 1,6-anhydro-MurNAc and GlcNAc products. LtgA snapshots were captured both with LtgA in its native form and as it liberated a GlcNAc residue with chitotetraose in the process of forming 1,6-anhydro-chitotetraose (Fig. 5b). In our assay, we identified chitotetraose and 1,6-anhydro-chitotetraose as reaction products but did not observe free single GlcNAc (Fig. 3, a and b). We postulate that GlcNAc was eluted close to the void volume of the column in our HPLC analysis and therefore did not appear as a defined, mono-species peak (Fig. 3, a and b). The sugar rings are welladapted to the shape of the active site, where subsites Ϫ1, Ϫ2, and Ϫ3 are locked in the interior of the protein, whereas ϩ1 and Ϫ4 are more flexible on the edges. The flexibility and orientation of GlcNAc in the ϩ1 position would allow unobstructed exit from the active site after cleavage.
In the reported structures of LT complexes with bound substrate analogs to date, the glycosidic bond and catalytic residues are outside hydrogen-bonding range. In the LtgA-1,6-anhydro-chitotriose complex, before product expulsion, we observed localized rearrangement of the active site (Fig. 5c). A possible explanation for this phenomenon is that the LT active site undergoes conformational changes that bring the catalytic residue within hydrogen-bonding distance of the glycosidic bond. However, this rearrangement during the initiation steps (stage 1) of the reaction (Fig. 1) was not observed in this study or in other studies where LTs were crystallized with inactive substrate analogs (22).
Unexpectedly, a conserved crystallographic water molecule was perfectly poised in the native complex to perform a proton relay between residue Glu 481 and the glycosidic bond ( Fig. 5a  and Fig. S3). This relay would ease the need for the substrate or active site to undergo drastic conformation changes to initiate

Visualizing bond cleavage by a lytic transglycosylase
the reaction in stage 1 (Fig. 1). Thus, LtgA and quite possibly other LTs could operate via a water-mediated substrate-assisted catalytic mechanism. In this mechanism, Glu 580 is positioned to provide structural support to the negative charge of the 2-acetamido (Figs. 5, a and b, and 8). During bond cleavage, the water molecule serves as a proton relay to the glycosidic bond and supports the formation of an oxazolium intermediate that deprotonates the catalytic residue Glu 481 (Fig. 8). Glu 481 then acts as a general base to abstract a proton from the C6-OH of the MurNAc residue, thereby allowing the intramolecular nucleophilic attack of the C1, which disrupts the oxazoline intermediate and forms the reaction products (Fig. 8).
LtgA is the first documented case of an LT that not only degrades chitin sugars but is also able to cleave a single sugar residue from the polymeric substrate. Normally, when LTs degrade the PG, the minimal product cleaved from the polymeric PG is a peptide-linked disaccharide (7). Interestingly, lysozyme, a glycosidase that is closely related to the LTs, also cleaves chitin sugar, releasing a single residue at a time (33). When LTs cleave their substrate, they utilize a single catalytic residue, and the reaction proceeds via the anchimeric assistance of the MurNAc 2-acetamido group and the formation of the oxazolinium ion intermediate (7). Lysozyme utilizes a different mechanism, whereby two catalytic residues serve the role of a general acid/base, and the documented intermediate steps include the formation and subsequent break down of a covalent glycosyl-enzyme intermediate (34,35). In the mechanism of lysozyme, there is inversion of the sugar configuration surrounding the anomeric carbon at each step; however, there is a net retention of the stereochemistry when the product is formed. Crystallographic studies of an active lysozyme (hen egg-white lysozyme) in the presence of a chitin sugar have revealed that the equivalent sugar in subsite Ϫ1 is distorted and assumes a half-chair position (35). The mechanism of family 18 chitinases is similar to LtgA and other LTs (36). These chitinases mechanistically utilize substrate participation or anchimeric assistance and cycle through an oxazolinium ion intermediate. However, similar to lysozyme, the stereochemical configuration around the anomeric carbon changes The model highlights the two stem peptide-binding grooves predicted to be involved in PG binding for E. coli Slt70, aligned on the conserved residues (Glu 580 , Phe 578 , Thr 575 , Trp 561 , and Arg 559 ) (Fig. 4b). A citrate molecule (colored in pink) occupies binding site Ϫ2.

Visualizing bond cleavage by a lytic transglycosylase
but retains the original configuration after the product formation step.
In the LtgA-chitotetraose plus GlcNAc complex, each pyranose ring of all the GlcNAc residues in the five subsites, (Ϫ4, Ϫ3, Ϫ2, Ϫ1, and ϩ1) adopts a chair conformation (Fig. 5b). Surprisingly, ring inversion of the GlcNAc residue was only observed in subsite Ϫ3 of the final product complex (LtgA-1,6anhydro-chitotriose). Subsite Ϫ3 is distal from the catalytic center (subsite Ϫ1 and ϩ1), and the GlcNAc residue appears to assume a twisted boat conformation (Fig. 5b). Because there was also structural rearrangement of the LtgA active site after the 1,6-anhydro product formation, subsite Ϫ3 possibly selected a different conformer of the GlcNAc residue because of changes in binding energy between the GlcNAc residue and LtgA (Fig. 6, a-c).
LtgA degrades both chito-oligosaccharides and PG (Fig. 5, a  and b) (5). Additionally, docking studies suggest no direct role of the stem peptides in the LtgA reaction (Fig. 7); however, they probably play a direct role in enzyme-substrate specificity. Structurally, the active site of LtgA is identical to that of E. coli Slt70 (Figs. 3b and 5b). However, there are subtle differences between the overall structures of Slt70 and LtgA. For example, LtgA has a long membrane-bound N termini that attaches itself to the membrane, whereas Slt70 remains localized in the periplasm (Fig. 4a) (5). This structural difference could affect the utilization of substrates in vitro. Overall, LtgA appears to be more permissive in its substrate utilization than other reported LTs.
The identification of 1,6-anhydro-chitotriose in the active site is the first snapshot taken of an LT synthesizing a 1,6-anhydro-sugar in the crystalline state. The F o F c maps contoured at 3 clearly showed the 1,6-anhydro-chitotriose, illustrating the path that the GlcNAc follows to facilitate the formation of the 1,6-anhydro-sugar. Interestingly, the sugar residue in the Ϫ3 position of the 1,6-anhydro-GlcNAc products appears to adopt a twisted boat conformation, suggesting that the glycan strand undergoes multiple conformational changes during the catalytic breakdown of chitopentaose.
These findings reported here reveal snapshots of intermediate steps in the LtgA reaction and provide unprecedented insight into how LTs cleave the glycosidic bond and form cytotoxic PG fragments. The results also suggest an alternative mechanism that could resolve previous discrepancies regarding the initiation of the LT reaction mechanism. Additionally, these trapped intermediates reveal informative details such as loop rearrangements, domain motion, substrate conformation, and transition state stabilization, which are all hallmarks of an active enzyme in motion. This study further emphasizes the pivotal role of LTs in PG biology, and the atomic-level insights into the studied mechanism should facilitate the design of universal inhibitors or antibiotics that target the LT family.

X-ray crystallography
Crystallization screening was conducted by the sitting drop vapor diffusion method with a Mosquito (TTP Labtech) automated crystallization system. All crystals of native LtgA (15-20 mg/ml) were grown at 18°C and appeared within 2-3 days. Native LtgA was crystallized under three different conditions using a 1:1 (v/v) ratio against the following well solutions: (a) 100 mM CHES, pH 9.5, 100 mM trisodium citrate, and 30% (w/v) PEG 3000; (b) 33% (w/v) PEG 6000 and 100 mM HEPES, pH 7.5; or (c) 0.2 M zinc acetate, 0.1 M sodium cacodylate, pH 6.5, and 0.1 M 18% (w/v) PEG 8000. The crystals were rectangular in shape and grew to ϳ200 -300 microns in length. Crystals of selenomethionine-labeled LtgA were grown at 18°C in a 1:1 (v/v) ratio against a well solution containing 0.01 M trisodium citrate, 33% (w/v) PEG 6000 and HEPES, pH 7.5. To initiate the reactions, crystals were soaked with a 10-fold molar excess of chitopentaose over the course of an hour (5,15,30,45, and 60 min) at pH 6.5, 7.5, and 9.5 before flash-cooling. The crystals were cryoprotected in a mixture of 50% Paratone and 50% paraffin oils and flash-cooled in liquid nitrogen.
Single-wavelength anomalous dispersion diffraction data were collected at the selenium absorption peak at the Soleil Synchrotron (Beamline Proxima-1). The highest-resolution selenomethionine crystals diffracted to 1.5 Å with anomalous signal to 1.8 Å (Tables S1 and S2). The initial phases were found using Shelxc/d/e, and the initial model was built using PHENIX autobuild with 13 selenium sites and a figure of merit of 0.70 (37,38). The refined single-wavelength anomalous dispersion structure was used to determine the phases for the native data set. Data processing for all structures was performed using XDS (39). Molecular replacement was performed using PHENIX (37). Building was conducted using Coot (40), and restrained refinement was carried out using a combination of PHENIX and the ccp4 software suite (37,41). MolProbity was used for iterative structure improvements during building and refinement (42).
All structural figures were generated using PyMOL (the PyMOL Molecular Graphics System, version 1.5; Schrödinger). The crystallographic parameters, data statistics, and refinement statistics are shown in Table S1. The coordinates and structure factors of native LtgA, and LtgA intermediates have been deposited in the Protein Data Bank with the accession codes 6FPN, 501J, 5029, 502N, and 5024.

Docking studies
The LtgA chitotetraose complex was modified with PG tripeptide stems terminating in a meso-DAP residue, i.e. Mur-NAc(L-Ala-D-Glu-meso-DAP). The docking and alignments were performed in Coot. The full LtgA-chitotetraose-meso-DAP complex was refined against the structure factors of the LtgA-chitotetraose complex using the ccp4 software refmac (40). The complex was evaluated and modified in Coot to improve the chemical accuracy.

Protein expression and purification
All constructs were created using standard molecular biology techniques as previously described (14). All constructs used in this study were GST fusions from pGEX-4T1 (GE Lifesciences). The native LtgA protein was expressed following transformation of the recombinant plasmids into BL21(DE3) Gold competent cells (Novagen). Protein expression was induced when the cells reached an optical density at 600 nm with 0.6 mM isopropyl Visualizing bond cleavage by a lytic transglycosylase ␤-D-thiogalactopyranoside at an optical density of 0.7-0.8, and harvest was performed after 4 h postinduction with cultures grown at 18°C. Selenomethionine-labeled proteins were expressed in B834 (DE3) and then grown in autoinduction media as described by Studier (43). After glutathione affinity chromatography and thrombin cleavage, the proteins were purified to homogeneity by size exclusion chromatography (Superdex 200; GE Healthcare) in 50 mM HEPES, pH 7.4, 150 mM NaCl, and 1 mM ␤-mercaptoethanol. After gel filtration, the proteins were immediately used for crystallization. Alternatively, the proteins were flash frozen in liquid nitrogen and stored at Ϫ80°C.

LtgA chitopentaose reaction
LtgA activity with chitopentaose was assessed using synthetic chitopentaose (Carbosynth) and purified LtgA. The chitopentaose (10 M) was incubated in the presence of 1.4 M of LtgA in 12.5 mM sodium phosphate buffer, pH 5.6, at 37°C for 5 min. Control reactions lacking chitopentaose or enzyme were also performed. The total reaction volume was 200 l. After 5 min of incubation at 37°C, the reaction was stopped by incubating the samples in a boiling water bath for 3 min. The enzymatic reaction products were separated by reversed-phase HPLC using a Shimadzu LC-20 system with a Hypersil GOLD aQ column (250 ϫ 4.6 mm, 5-m particle size, flow rate of 0.5 ml/min at 52°C; Thermo Fisher Scientific). The mobile phase gradient was H 2 O ϩ 0.05% trifluoroacetic acid with a 0 -25% acetonitrile gradient over 135 min. Chito-oligosaccharide peaks were collected and dehydrated by vacuum centrifugation. A full-spectrum MS scan was obtained on the LTQ Velos Orbitrap at a resolution of 60,000. Collected fractions were further analyzed in positive ion mode by direct nanoelectrospray infusion using a Tri-Versa Nanomate (Advion Biosciences) on an LTQ Velos Orbitrap mass spectrometer equipped with an electron transfer dissociation (ETD) module (Thermo Fisher Scientific). A full set of automated positive ion calibrations was performed immediately before mass measurement. All spectra were acquired in full profile mode. For the MS experiments, the ions were accumulated in the ion trap and then transferred to the Orbitrap for high-resolution mass measurement. For the MS/MS experiments, ions were selected with an appropriate mass window, and higher-energy collisional dissociation was performed at normalized collision energies of 15-25%, with other activation parameters set to default values. The spectra were acquired in the Fourier transform-based mass spectrometer over several minutes with five microscans and a resolution of 60,000 for MS and 30,000 of m/z 400 for MS/MS before being processed with Thermo Xcalibur 2.2. Multiply charged ion spectra were deconvoluted using Xtract.