Activation by Allostery in Cell-Wall Remodeling by a Modular Membrane-Bound Lytic Transglycosylase from Pseudomonas aeruginosa

SUMMARY

Bacteria grow and divide without loss of cellular integrity. This accomplishment is notable, as a key component of their cell envelope is a surrounding glycopeptide polymer. In Gram-negative bacteria this polymer—the peptidoglycan—grows by the difference between concurrent synthesis and degradation. The regulation of the enzymatic ensemble for these activities is poorly understood. We report herein the structural basis for the control of one such enzyme, the lytic transglycosylase MltF of Pseudomonas aeruginosa. Its structure comprises two modules: an ABC-transporter-like regulatory module and a catalytic module. Occupancy of the regulatory module by peptidoglycan-derived muropeptides effects a dramatic and long distance (40 Å) conformational change, occurring over the entire protein structure, to open its active site for catalysis. This discovery of the molecular basis for the allosteric control of MltF catalysis is foundational to further study of MltF within the complex enzymatic orchestration of the dynamic peptidoglycan.

INTRODUCTION

The assembly of the bacterial cell wall is among the least understood biological events. Although we have a credible understanding of the molecular and macromolecular components of the cell wall—the lipids, the structural proteins, the enzyme catalysts, and the peptidoglycan polymer—we do not understand how these components individually integrate into a dynamic structural edifice that allows a bacterium to replicate itself within minutes, while preserving throughout structural integrity. In this particular respect the peptidoglycan component of the cell wall of the bacterium presents a structural dilemma. The peptidoglycan is a polymer that surrounds the bacterium, and yet it expands and divides as a polymer in coordination (and in the face of considerable internal pressure) with both the cytoplasmic and membrane events of bacterial reproduction (Rojas et al., 2014, Tuson et al., 2012). While the chemical bonding of the oligomeric peptidoglycan is known—glycan strands consisting of a repeating β-1,4-linked N-acetylglucosamine-N-acetylmuramic acid (NAG-NAM) disaccharide, cross-linked to adjacent glycan strand(s) through the peptide stem emanating from the N-acetylmuramic acid saccharide—the three-dimensional arrangement of these oligomers and the enzymatic events of its growth and division are not known (Lovering et al., 2012, Turner et al., 2014). Notwithstanding its identity as a polymer, the peptidoglycan is a dynamic structure. Indeed, this structure is in constant turnover by processes that involve a large number of enzymes (Johnson et al., 2013). The enzymatic activities that fabricate the peptidoglycan structure in Gram-negative bacteria correspond to activities that build the polymer (notably transglycosylases and transpeptidases) and activities that cleave (or remodel) the polymer (notably amidases and lytic transglycosylases). Each of these activities corresponds to a family of enzymes, having seemingly extensive functional redundancy within each family. Accordingly, experimental design to clarify the catalytic activity of any one member of these families is challenging.

Of the four primary enzymatic families that act on the peptidoglycan, arguably that of lytic transglycosylases (LTs) is the most enigmatic (Scheurwater et al., 2008). The reaction catalyzed by these enzymes is the non-hydrolytic cleavage of the glycan strand of the peptidoglycan, with formation of the non-reducing N-acetyl 1,6-anhydromuramic and the N-acetylglucosamine as the two termini (Figure 1). The inability to create a viable phenotype upon attempted genetic deletion of the LT family of Escherichia coli implicates aspects of their collective activities as essential (Scheurwater and Clarke, 2008). The importance of the LT family is further demonstrated by the substantially decreased viability and antibiotic resistance observed as result of the genetic deletion of selective members of this family in both E. coli (Heidrich et al., 2002) and Pseudomonas aeruginosa (Cavallari et al., 2013, Lamers et al., 2015). Indeed, LT activity in Gram-negative bacteria contributes directly to peptidoglycan growth, to peptidoglycan cleavage in septation (Cloud and Dillard, 2004, Jorgenson et al., 2014, Yahashiri et al., 2015), to the structural integrity of the cell wall (Lamers et al., 2015), to the expression of virulence as a result of the release of inflammatory peptidoglycan entities (Cloud-Hansen et al., 2008, Cloud-Hansen et al., 2006), to the creation of space within the peptidoglycan polymer to allow the penetration of the macromolecular complexes required for bacterial viability (Cloud-Hansen et al., 2008, Scheurwater and Burrows, 2011, Scheurwater and Clarke, 2008), to the regulation of β-lactamase expression (Cavallari et al., 2013, Fisher and Mobashery, 2014, Johnson et al., 2013, Korsak et al., 2005, Kraft et al., 1999), and lastly (as a disregulated activity) to the lytic events consequent to antibiotic exposure (Kohlrausch and Holtje, 1991, Lamers et al., 2015). These observations have led to the presumption that (notwithstanding proven functional redundancy) individual members of the LT family preferentially associate with particular activities. Apart from the LTs that complex with the penicillin-binding proteins and thus are participants in peptidoglycan growth and remodeling (Holtje, 1996, Legaree and Clarke, 2008, Nikolaidis et al., 2012, Romeis and Höltje, 1994, Vollmer et al., 1999, von Rechenberg et al., 1996), there is scant direct evidence toward the identification of these activities or for the mechanisms regulating the activity of any of the LTs. It is with this objective in mind that our attention was drawn to the MltF LT of P. aeruginosa.

Figure 1. Scheme showing different activities for peptidoglycan-degrading enzymes.

The sites of reactions are indicated by the same colors designating each activity.

MltF of E. coli was characterized first by Scheurwater and Clarke (Scheurwater and Clarke, 2008) as an outer membrane-adhering protein having a unique (as compared to other LTs) primary structure. MltF was interpreted by Scheurwater and Clarke as two conjoined protein modules, with one module having an unknown function (but annotated by its structure as a Periplasmic_Binding_Protein_Type_2 superfamily module by the Conserved Domain Database of the NCBI, and as a SBP_bac_3 bacterial extracellular solute-binding protein family 3 (PF00497) by the Pfam Protein Families Database, with sequence homology to the ligand recognition domain of ABC transporters), and the second module as the lytic transglycosylase. This N-terminal domain with homology with ABC transporters has been shown to modulate the lytic activity of the LT domain to permit the continued lysis of insoluble peptidoglycan at a constant rate (Scheurwater et al., 2008).

Genetic deletion of MltF affected neither the shape nor the doubling time of the planktonic growth of E. coli (Scheurwater and Clarke, 2008). Pseudomonas aeruginosa also has an MltF enzyme with substantial (although not complete, as will be discussed) similarity to the MltF of E. coli, notably including the linked bimodule structure. In addition to MltF, the P. aeruginosa family of LTs comprises nine additional LT enzymes. This LT ensemble divides between six lipoproteins (Mlt enzymes, which are covalently attached either to the inner membrane leaflet or to the outer membrane leaflet, both facing the periplasm) and four soluble enzymes (the Slt enzymes of the periplasm). MltF is a lipoprotein LT. Its lipobox sequence (…LAGCSEA….) is consistent with its retention in the inner membrane of P. aeruginosa (Lewenza et al., 2008, Narita et al., 2007, Remans et al., 2010). Genetic deletion of MltF in PAO1 P. aeruginosa likewise gave no obvious phenotype with respect to planktonic growth, nor an effect on the β-lactam sensitivity of the bacterium (Lamers et al., 2015). Collectively, these observations seemingly exclude a role for MltF in peptidoglycan growth, remodeling, and recycling in both E. coli and P. aeruginosa. Nonetheless, the MltF gene is retained not just in eight pathogenic P. aeruginosa strains—LESB58 (excluding the signal sequence, 100% sequence identity compared to PAO1), DK2 (100%), PA7 (97.3%), NCGM2 (100%), PA14 (UCBPP) (100%), B136 (100%), M18 (100%), PACS2 (100%)—but in all of the sequenced P. aeruginosa genomes we examined (Figure S1). This retention suggested to us that the MltF function contributes directly to P. aeruginosa viability and/or pathogenicity, and that the relationship of its two linked modules to this attribute was likely controlled regulation of its LT catalytic activity. Here we address this hypothesis through the comprehensive analyses of an ensemble of P. aeruginosa MltF X-ray structural states and by characterization of the enzymatic reactions that are modulated by muropeptide effectors.

Results and Discussion

Structure of the Catalytically Inactive Conformation of MltF

P. aeruginosa MltF was expressed as a soluble protein (without its N-terminal lipobox) and was crystallized. The structure was solved by using the SAD method with a SeMet derivative diffracting up to 2.6-Å resolution (Table 1). The resulting model was then used as initial model in the molecular replacement method for phasing high-resolution (1.8 Å) wt MltF diffraction data (Table 1). The MltF structure showed high-quality electron density maps (Figure S2) and confirmed that the enzyme was organized into two linked modules, one non-catalytic and the other catalytic (CM). The non-catalytic module—which we will refer to as the regulatory module (RM)—is indeed homologous to the substrate-binding modules from ABC-transporters (Figure S3). In this crystal structure disposition of the RM module with respect to the CM precludes the possibility of catalysis (Figure 2). Although CM presents a large active-site cleft (30 Å long by 10 Å wide) for substrate binding (Figure. S4), centered on the exposed catalytic Glu316 residue, access to this cleft is blocked. Superimposition of the catalytic module alone with the closest structural homologues in complex with substrate analogues—the MltC:tetrasaccharide complex (Artola-Recolons et al., 2014) and the Slt70:anhNAM-NAG complex (van Asselt et al., 1999)—indicates that six saccharide rings (positions −4 to +2) could be accommodated within this cleft (Figure S4). However, access to this cleft is blocked by three separate segments of the MltF protein: by a loop of the CM (L_C, residues 317–337), by the twelve-amino-acid linker peptide (residues 266–277) spanning the two modules, and by a loop of the RM (L_R, residues 49–66) (Figure 2C). The L_R loop blocks the distal side of the active site and engages, with its Arg60, the catalytic Glu316 in a salt-bridge interaction (Figure 2C and Figure S2B). In this conformational state access of the peptidoglycan substrate to the active site of the LT module is not possible.

Table 1.

Crystallographic Data

	SeMet-MltF	MltF	MltF:1 complex	MltF:2 complex	MltF:3 complex	MltF:3:NAG-NAM-NAG-NAM complex
Data collection
Wavelength (Å)	0.98088	0.98088	0.97950	1.00001	0.97936	0.979340
Space group	P21212	P21212	P21	P21	P21	P21
Unit Cell a, b, c (Å)	166.993 134.977 48.749	166.993 134.977 48.749	65.85, 135.75, 138.03	65.59, 135.65, 137.46	66.12, 136.77, 138.10	150.60, 136.38, 195.39
Unit Cell α,β,γ (°)	90, 90, 90	90, 90, 90	90, 92.01, 90	90, 92.06, 90	90, 92.34, 90	90, 111.36, 90
T (K)	100	100	100	100	100	100
X-ray source	Synchrotron	Synchrotron	Synchrotron	Synchrotron	Synchrotron	Synchrotron
Resolution (Å)	48.79(2.71–2.6)	54.17(1.83–1.80)	47.25(2.98–2.89)	48.26–(2.90–2.80)	48.57–(2.44–2.40)	48.48–(3.25–3.20)
Total reflections		559415	192583	131278	631146	790332
Unique reflections		103026(4898)	49859(5000)	59073(5869)	94338(4752)	120807(5982)
Average I/σ(I)	30.6(14.3)	15.5 (1.9)	12.5 (1.7)	5.1 (0.5)	13.8(1.7)	4.7(1.3)
Completeness (%)	99.8(99.2)	99.7(98. 0)	91.3(100)	99.7(99.5)	98.5(99.9)	99.6(99.8)
Redundancy	13.1(13.5)	5.5(5.3)	3.9(3.9)	3.3(3.1)	6.7(6.7)	6.5(6.8)
Rpim	0.02(0.05)	0.03(0.41)	0.06(0.46)	0.07(0.65)	0.04(0.50)	0.10(0.56)
CC1/2	0.99(0.99)	0.99(0.99)	1.00(0.99)	0.99(0.99)	1.00(1.00)	0.99(0.90)
Refinement
Resolution (Å)		54.17–1.80	47.25–2.89	48.26–2.80	48.57–2.40	14.99–3.20
Rwork/Rfree		0.18/0.21	0.16/0.24	0.16/0.23	0.18/0.23	0.24/0.30
No. of Atoms
Protein		7340	13625	13589	13609	40237
Ligand		–	42	79	67	–
Water		658	129	71	166	–
B-factor (Å2)
Protein		39.60	64.50	60.10	70.20	72.80
Ligand		–	83.30	80.80	80.00	–
Water		46.00	42.40	44.20	55.50	–
r.m.s. deviations
Bond length (Å)		0.007	0.009	0.008	0.016	0.014
Bond angles (°)		1.05	1.35	1.15	1.60	1.68
PDB CODE		5A5X	5AA1	5AA2	5AA4	5AA3

Figure 2. The three-dimensional structure of apo MltF.

(A) Cartoon depicting the modular arrangement of MltF: Signal Peptide (SP) with a Lipobox, the ABC-transporter module (ABC), the linker, and the catalytic module (CM). The beginning and the ending of each segment are specified by the amino-acid positions. (B) Three-dimensional structure of apo MltF (colored as in panel A). The catalytic Glu316 is depicted as yellow-capped sticks in the center of the catalytic domain. (C) The active site of MltF is entirely blocked by the specific interactions, as depicted by Connolly surfaces for both models. Right, an expansion of the catalytic site showing the salt-bridge interaction between Arg50 and the catalytic Glu316.

According to the classification of substrate-binding proteins (SBPs) (Berntsson et al., 2010), the RM of MltF belongs to the Cluster F-IV module that binds typically amino acids as ligands. The RM comprises two domains (Domains 1 and 2), interconnected by a conformationally flexible hinge. SBPs exist customarily in an “open” conformation in the absence of ligand, exposing a large space between domains (Bermejo et al., 2010, Oldham et al., 2008) (Figure. S5). In typical SBPs ligand binding closes the domains by a “Venus Fly-Trap” mechanism to entrap the ligand (Mao et al., 1982). However, the RM module in MltF is different. Our MltF structure shows a closed RM conformation in the absence of ligand (Figure 3A and Figure S5) as a result of a strong network of salt-bridge interactions on the opposite side (with respect to the hinge) of the module. In this closed arrangement a large T-shaped cavity (13 × 21 Å and a volume of 766.8 ų as calculated by CASTp server (Dundas et al., 2006)) traverses the length of the RM and connects its outer surface near the Lββ loop to Tyr266 of the linker region (Figure 3B). This cavity presents a nearly homogeneous width of around 8 Å and is built by polar residues (Tyr122, Tyr94, Ser160, Thr111, and Thr55), many hydrophobic residues (Leu90, Leu110, Val155, Val185, Leu188, Leu229, Leu127, and Phe66) and some charged residues (His161, Arg50, Glu206, Glu67 and Asp122) that results in a specific bipolar pattern for the electrostatic potential inside the cavity (as calculated by the Adaptive Poisson-Boltzmann Solver (APBS)) (Figure 3B). Our ensuing crystallographic studies revealed that ligand muropeptide occupancy of the RM cavity controls access to the catalytic module of MltF.

Figure 3. The regulatory module of MltF showing the structure of a closed, but unoccupied, conformation of an SBP/ABC-transporter module.

(A) The two domains of the regulatory module are colored in two shades of green. Right, an expansion of the interface between the domains highlighting the residues involved in inter-domain salt-bridge interactions (depicted as capped sticks). This strong network of interactions (Arg50-Glu184, 2.8 Å and 3.1 Å; Arg50-Glu206, 3.6 Å; Asp91-Lys157, 3.4 Å) keep the RM in a closed conformation in the absence of ligand. (B) A cross-section of the ligand-binding groove of the RM of MltF showing the Poisson-Boltzmann electrostatic-potential surface (color bar range ± 5 kT/e). The dimensions of the cavity are labeled. The Tyr266 residue of the linker is located at the bottom of the cavity, and is depicted as capped sticks. (C) The regulatory module of MltF (same view as that of panel B) with the three muropeptides (1, 2 and 3 as observed in MltF:1, MltF:2 and MltF:3 complexes respectively) superimposed. Muropeptides are represented as capped sticks and are labeled. (D) Electron density maps for the peptide stem observed in the different complexes: MltF:1 complex, MltF:2 complex and MltF:3 complex. Peptide stems are presented as capped sticks and colored in yellow, orange and purple respectively. Electron density corresponds to the (2F_o–F_c) FEM map contoured at 1.0 α. In all these cases electron density was observed only for the peptide portion of the ligand.

The Structure of MltF in Complex with Muropeptides at the Regulatory Module

The presumption that the RM module could recognize muropeptides was tested using synthetic muropeptides. Exposure of the MltF crystal to each of three muropeptides—NAG-anhNAM-L-Ala-γ-D-Glu-mDAP-D-Ala-D-Ala (compound 1), NAM-L-Ala-γ-D-Glu-mDAP-D-Ala-D-Ala (compound 2) or L-Ala-γ-D-Glu-L-Lys-D-Ala (compound 3)— changed the space group of the resulting crystals and showing four monomers in the asymmetric unit (instead of the two chains observed in the apo structure) (Table 1). The three complexes present different structures compared to the apo structure (rmsd range of 1.13 Å-1.39 Å for MltF:1 complex, 1.02 Å-1.32 Å for MltF:2 complex and 0.96 Å-1.26 Å for MltF:3 complex, values obtained by superimposition of all Cα atoms of the complexes with chain A of the apo structure). These structures show the muropeptides bound to the RM (Figure 3C, 3D and Figures S6–S8) and at a distance of 40 Å from the active site (from Tyr94 residue at the entrance of the RM cavity to catalytic Glu316). Only the peptide segment inserted into the RM cavity, and not the glycosidic segment, of the first two muropeptide structures is crystallographically visible; the best electron density observed for the compound 3. The structures document a progressive conformational transformation of the entire protein en route to activation. As exemplified by the MltF:3 complex (Figure 4A and Movie S1), the RM shows an altered conformation of the Lββ loop and small rotations of each domain around the hinge (~2° for Domain 1 and ~3° for Domain 2). These changes propagate to the linker region of the RM. The presence of the muropeptide ligand at the allosteric site affects the conformation of Tyr266. This tyrosine is a residue of the linker, and is located at the bottom of the RM cavity (Figures 3B, 4A and Figure S9). The linker region also changes upon muropeptide binding and acts as a pendulum to amplify movement of the CM (a rotation of ~11° of the entire CM centered at Cα of Tyr266) that results in up to 4 Å residue displacements. Similar effects were observed for the MltF:1 and MltF:2 complexes. It is worth mentioning that in the MltF complexes Tyr266 presents different rotamers (Figure S9C) and that part of the linker region was disordered for some of the chains. Despite these structural changes, the catalytic site remained occluded in each complex. Simultaneous exposure of the crystals to ligand 3 (targeting the RM) and to the substrate analog NAG-NAM-NAG-NAM (targeting the CM) cracked the crystal, consistent with profound conformational change. Rapid freezing of one such crystal provided crystallographic data (Table 1) indicating that a larger unit cell (having 12 independent molecules per asymmetric unit cell, compared to two for the apo crystal and four in each of the previous complexes) was formed. All 12 molecules in the unit cell showed conformational change compared to the apo form (rmsd values ranging from 1.09 Å to 1.32 Å for Cα atoms) (Figure 4B), notwithstanding the absence of clear electron density either for the ligand or for the substrate analog in the CM. In the most profound case (Figure S9B) the changes in the RM resulted in a larger rotation (~15° around Tyr266) for the CM and up to a 5-Å displacement in its residues. The 24 independent structures for the MltF complexes reported here are interpreted using a sequence of eight representative snapshots (Figure S9D), which show the progression of the RM-CM conformations toward enzyme activation. None of the eight, however, fully exposes the active site. This inability (to expose the active site) is likely a consequence of the crystal packing not allowing the conformational change required for full exposure of the active site.

Figure 4. Allosteric signal propagation from the RM to the CM upon muropeptide binding.

(A) Superimposition of the crystal structure of apo MltF (light-brown ribbon) with the crystal structure of MltF:3 complex (green ribbon). The muropeptide is depicted as black capped-sticks. Arrows indicate molecular motion in both RM and CM (see also Movie S1). (B) Superimposition of the crystal structure of apo MltF (black ribbon) with the 12 independent monomers observed in the MltF:3:NAG-NAM-NAG-NAM complex. (C) Structural comparison between the inactive conformation of MltF (left, this work) and of the fully open conformation for the same enzyme (right, PDB 4P11). Ribbons are colored as in panel A. Both structures have the same orientation for the RM. Activation of the protein involves rotation of 129° and displacements of >55 Å by the CM. The catalytic Glu316 is depicted as capped sticks with its carbon atoms colored in yellow. In the inactive conformation the linker is present in an extended-coiled conformation spanning 30 Å from Cα of Tyr266 to Cα of Tyr278. In the active conformation the linker organizes as a helical structure having half (14.5 Å) of the original length (see also Movie S2).

From Inactive to the Active Conformation of MltF

Independent of our work, complementary (but as yet unpublished) crystal structures for P. aeruginosa MltF were deposited to the PDB (codes 4P11, 4OYV, 4OZ9, 4OXV, 4OWD and 4P0G). These structures crystallize in the apo form and in complex with single amino acids (Leu, Ile, Cys and Val) in the RM, or in complex with a ligand occupying the active site. Surprisingly, all of them are very similar (rmsd ranging from 0.2 Å to 0.8 Å) and show the active site of the CM as unoccluded and thus substrate accessible. This circumstance points to two conclusions. First, the MltF used in these crystallizations was somehow activated during purification and/or the crystallization process. Second, once activated the structure remains in this conformation even in absence of a ligand in the RM or in the CM. Our efforts to reproduce crystal formation using the different crystallization conditions reported for these PDB files were not successful. Structural comparison of our apo structure to those deposited in PDB indicates that the conformational changes in the transition from the inactive to the active states occur to both modules and to the linker (Figure 4C and Figure S10). The CM rotates as a rigid body in this transition by 129° around the linker (Figure S10) in order to accomplish a >55 Å movement. The linker region transitions from an extended coiled conformation of ~30 Å (our structure) to a distorted α-helix of half the length (14.5 Å) (Figure 4C). Simulation of this dramatic structural transition by targeted molecular dynamics (Movie S2) reveals the extensive flexibility of the linker that is required to accommodate the two modules in the two (inactive and active) completely different conformations.

In order to demonstrate the potential role of peptide stems from muropeptides in MltF activation in solution, we examined the reactivity of crystalline MltF toward peptidoglycan as potential substrates using liquid chromatography and mass spectrometry (LC/MS). To guarantee a homogeneous inactive population of MltF monomers, large MltF crystals were collected, washed with crystallization mother liquor (25% ethylene glycol), and then used for the reaction with substrates. A synthetic NAG-NAM(pentapeptide)-NAG-NAM(pentapeptide) tetrasaccharide structure was unreactive either in the presence or absence of an added muropeptide effector. This result indicated that MltF requires either a substrate greater than four saccharides in its length, or prefers substrates with a shorter peptide stem. We then examined 4—a mixture of (NAG-NAM)_n-NAG-anhNAM natural glycans having a length of 6–20 saccharides and lacking peptide stems—as a substrate. This choice allowed us to clearly differentiate between recognition and catalytic transformation as a muropeptide substrate, from any effect on the muropeptide effector. The closed conformation crystals of MltF were unreactive to 4 (Figures 5A and B). In agreement with this lack of reactivity, crystals remained intact after incubation with 4. However, simultaneous incubation of the crystals with muropeptide 3 and substrate 4 resulted in the fracturing of the crystals and the catalytic transformation of 4 to the disaccharide product NAG-1,6-anhydroNAM 5 and the tetrasaccharide product NAG-NAM-NAG-1,6-anhydroNAM 6 (Figure 5C). As these reactions use a crystalline enzyme catalyst, their turnover rate is modest. In the presence of 3 peak intensity for both the hexamer (7) and octamer (8) also increased (Figure 5C). However, this increase is not evident in the chromatogram of Figure 5C, as both compounds were present already in the substrate mixture (corresponding to the n = 2 and n = 3 compounds of mixture 4, Figure 5A) used in this experiment. For this reason, this experiment was repeated with substrate 9, a mixture of glycans with a length of 10–18 saccharides and thus lacking both 7 or 8 (Figure 5D). As expected, substrate 9 alone was unreactive to the MltF crystals. (Figure 5E). In the presence of both muropeptide 3 and 9 with the MltF crystals, the expected formation of glycan products, 5–8 was observed (Figure 5F). This set of experiments clearly demonstrates MltF activation from inactive MltF crystal triggered by muropeptide in solution.

Figure 5. Reaction of the crystals of apo MltF with (NAG-NAM)_n-NAG-anhNAM (4, n = 2–9 and 9, n = 4–8) in the presence or absence of compound 3.

Extracted-ion-chromatograms of the substrate/product mixture obtained with compounds 4 (A), 4 with MltF_crystal (B), 3 + 4 with MltF_crystal (C), compounds 9 (D), 9 with MltF_crystal (E), 3 + 9 with MltF_crystal (F) demonstrate the positive effect of the presence of the muropeptides as effectors of the catalytic transformation of 4 (or 9) as a MltF substrate. Upon activation MltF acts as a primarily endolytic enzyme.

Discussion

The lytic transglycosylases catalyze, through a single chemical reaction, myriad transformations of the peptidoglycan of Gram-negative bacteria. Understanding how these transformations are temporally (with respect to the growth cycle), spatially (location to the sidewall or septum), and functionally controlled is a frontier for the understanding of the dynamic peptidoglycan structure of the Gram-negative bacterium. For the MltF LT of P. aeruginosa we demonstrate here an allosteric mechanism for the control of its catalytic activity, as a result of the dramatic conformational transition that occurs following occupancy of its regulatory module by peptidoglycan-derived effectors. The outstanding question concerning MltF is therefore the purpose of this regulation.

The primary structure of the P. aeruginosa MltF is highly conserved, including an invariant sequence for the linker between the two modules. This constancy befits the role the linker has in the propagation of the signal from the RM to the CM to enable the LT activity. Interestingly, another substrate-binding protein (an Arg transporter) has recently been reported for which the helical region after the SBP, equivalent to our linker region, is involved in protein oligomerization and changes its conformation upon ligand binding (Ruggiero et al., 2014).

A comparison of the sequences of the MltF enzymes of P. aeruginosa and E. coli (Figure S11) confirms the latter as an RM-linked LT. Notwithstanding this similarity, two contrasts exist between the P. aeruginosa and the E. coli MltF enzymes. The P. aeruginosa MltF is a lipoprotein whose presumptive “avoidance lipobox” retains it in the inner membrane, while the E. coli mltF gene does not encode a lipobox (and thus the protein cannot be lipidated). In contrast, the E. coli MltF has an N-terminal amphipathic helix that localizes this enzyme to the outer membrane of E. coli (Scheurwater and Clarke, 2008). Although the difference in membrane location of these two enzymes might appear substantive, as the peptidoglycan is found between the two membranes and activation of these MltF enzymes involves substantial structural change, the peptidoglycan may be regarded as equally accessible from either membrane. An alternative possibility is that the functional homolog in P. aeruginosa for the MltF of E. coli is the P. aeruginosa MltF2 enzyme, which is also an RM-linked-CM LT and whose N-terminus is homologous to that of the E. coli MltF (Fig. S11).

MltF is the second LT of P. aeruginosa brought to a deeper understanding of its function as a regulated enzyme: RlpA is a lipoprotein LT that is recruited (as a consequence of prior amidase removal of the peptide stems of the peptidoglycan) by its SPOR domain to the septal peptidoglycan (Jorgenson et al., 2014, Yahashiri et al., 2015). The presence of numerous peptidoglycan amidases in the periplasmic space is well recognized (Büttner et al., 2015, Yahashiri et al., 2015). These amidases include AmpDh2 and AmpDh3 (Juan et al., 2006, Martinez-Caballero et al., 2013, Moya et al., 2008, Schmidtke and Hanson, 2008), as well as other peptidoglycan amidases (such as AmiB) that participate in cell division (Juan et al., 2006, Martinez-Caballero et al., 2013, Moya et al., 2008, Schmidtke and Hanson, 2008). Given that prior studies with MltF exclude it from cell division, a mechanistic connection between AmpDh2/AmpDh3 and MltF is possible. AmpDh2 is a dimeric protein that anchors to the inner leaflet of the outer-membrane (Büttner et al., 2015). AmpDh3 is a tetrameric and soluble enzyme that latches on to the peptidoglycan itself (Lee et al., 2013). Both enzymes hydrolyze the peptide stem of the peptidoglycan at the lactyl moiety of the NAM unit without preference for the length of the stem (Zhang et al., 2013). Their hydrolytic reactions were assumed to contribute to cell-wall recycling, but the activation of MltF by the binding of the peptide stems in the RM module resulting in MltF-dependent cleavage of substrates 4 indicates their possible relevance to the control of the catalytic function of MltF. In such a circumstance, AmpDh2- and/or AmpDh3-amidase-dependent release of regulatory stem peptides from the cell wall would effect the dramatic transformation of MltF from a catalytically inactive (our structures) to the catalytically active structure of PDB 4P11 (Figure 6A). Regulation of cell-wall turnover is complex and entirely un-elucidated at present. The peptide stem that is liberated by AmpDh2 and AmpDh3, especially the tetrapeptide, which is incidentally the most abundant in P. aeruginosa, is the likely effector. Under what biological condition this peptide is liberated and why it might regulate MltF is presently not known.

Figure 6. Muropeptide-induced activation of the catalytic activity in MltF.

(A) Proposed model for the muropeptide-induced activation of the catalytic activity in MltF is depicted as a cartoon (top) and associated crystallographic structures (below). Muropeptides bound within the RM are depicted as spheres. (B) MltF interacts with long PG chains in the middle. Three-dimensional structure of active MltF (PDB code 4P0G) is superimposed with the NMR-based structure (Meroueh et al., 2006) of the crosslinked peptidoglycan (salmon sticks). A structural model for a bilayer lipidic membrane is depicted for comparison purposes.

In the catalytically active state of MltF the linker segment has folded into an α-helix and the CM has oriented towards the polymeric peptidoglycan substrate. Our cell-wall degradation experiments indicate that, once activated, MltF acts as a primarily endolytic enzyme. This is the ability to fragment the peptidoglycan strand in the middle (Figure 6B). Both the crystallographic clarity of the peptide effectors found within the RM module of our structures and our in crystallo activation experiments provide compelling evidence that these peptide moieties indeed correspond to the endogenous effector, and thus define the experimental directions necessary to clarify the complete character of the MltF enzyme.

How does this transformation relate to the physiological function of MltF? Again, the studies of Scheurwater and Clarke (Scheurwater and Clarke, 2008) and Lamers et al. (Lamers et al., 2015) exclude a role for MltF in either the sidewall growth or the septal cleavage of planktonic E. coli and P. aeruginosa. The third task given to LTs, the punctuation of the peptidoglycan (Scheurwater et al., 2008, Scheurwater and Burrows, 2011), remains as a possibility for MltF. Thus MltF activation may coordinate insertion through the peptidoglycan of the flagella, pili, fimbriae, secretion systems, or transporters (such as for the recovery of muropeptides from the medium). Given the observation that the E. coli mltF (yfhD) gene is transcriptionally activated by the Rob resistance transcription factor (Bennik et al., 2000), the importance of the pili and fimbriae for bacterial surface adhesion (Persat et al., 2015a, Persat et al., 2015b), and the notable adhesive capability of P. aeruginosa (Giltner et al., 2012), an involvement in these roles of MltF (as a genetically retained and allosterically regulated enzyme) is possible.

This is first study demonstrating how an SBP RM modulates the cell-wall remodeling machinery of P. aeruginosa. The richness of the structural information on MltF reveals this unique regulatory mechanism and provides a foundation for further inquiry in understanding of its contribution toward the complex orchestration of the structure of the peptidoglycan of the P. aeruginosa cell wall.

Experimental Procedures

Protein design and purification

The mltF(PA3764) gene corresponding to amino acids 33–490 of the native enzyme was cloned. This construct omits the cysteine (Cys27) that is S-lipidated in vivo. Cloning from residue 33 gave the soluble recombinant protein then purified to homogeneity.

Protein crystallization

Crystals were obtained at 18 °C in a precipitant solution consisting of 25% (v/v) ethylene glycol (EG) in 25 mM Hepes pH 7.4 buffer. Complexes were produced by soaking of the ligands, as detailed in the Supporting Information.

Data collection, phasing and model refinement

The X-ray diffraction data set for Se-Met crystal was collected on beamline ID23-1 at the European Synchrotron Radiation Facility (ESRF, France) using a Pilatus 6M-F detector and a wavelength of 0.98088 Å at 2.6 Å. The structure was solved by the SAD technique (statistics shown in Table 1). The asymmetric unit of MltF consists of two monomers. Each monomer is comprised of three distinct regions: the ABC-transporter module (residues 31 to 265), the linker (residues 266 to 277), and the catalytic module (residues 278 to 490)

All complexes were collected at the beamline XALOC at the ALBA synchrotron. These structures were solved by the Molecular Replacement method using the coordinates of our apo MltF structure as a search model, followed by refinement (statistics shown in Table 1). These methodologies are detailed in the Supporting Information.

Molecular-Dynamics Simulations

The closed (this work) and open [PDB ID: 4P11] crystal structures were prepared for molecular-dynamics simulation using the Schrödinger MAESTRO program, as detailed in the Supplementary Information.

Synthesis of Muropeptide Ligands

Ligands 1 and 2 were prepared using methods reported previously (Hesek et al., 2009, Lee et al., 2010) Ligand 3 has not been reported previously. It was prepared by these same general methods (Hesek et al., 2010): ¹H NMR (500 MHz, D₂O) δ 1.24–1.56 (m, 2H), 1.46 (d, J = 7.2 Hz, 3H), 1.62 (d, J = 7.0 Hz, 3H), 1.70–1.92 (m, 4H), 2.03–2.15 (m, 1H), 2.23–2.34 (m, 1H), 2.48 (tq, J = 15.3, 7.5 Hz, 2H), 3.05 (t, J = 7.5 Hz, 2H), 4.19 (q, J = 7.0 Hz, 1H), 4.33–4.47 (m, 3H); ¹³C NMR (126 MHz, D₂O) δ 16.4 (q), 16.8 (q), 22.2 (t), 26.5 (t), 26.9 (t), 30.8 (t), 31.6 (t), 39.3 (t), 49.0 (d), 49.3 (d), 52.8 (d), 53.8 (d), 170.9, 173.9, 175.1, 175.3, 176.7; MS (ESI-QTOF) m/z 418.2334 [M+H]⁺ (calculated for C₁₇H₃₁N₅O₇ 418.2296).

Reaction of MltF crystals with compounds 4 and 9 analyzed by liquid chromatography and mass spectrometry

MltF crystals were grown in the condition described above. Large single crystals (~0.3 × 0.3 × 0.2 mm) were picked and washed with 25% ethylene glycol. Muropeptide 3 was added to a suspension of MltF crystals (~20 μg) in buffer containing 25% ethylene glycol. After 20 min compound 4 (or 9) was added to the suspension. The suspension was incubated for 2 h at 23 °C. The reaction was quenched by the addition of 2% trifluoroacetic acid. The supernatant of the quenched reaction mixture was analyzed by LC/MS. The control reaction was carried out the same way except without the addition of 3. Our LC/MS analytical methods were described previously (Lee et al., 2013).