Solution Structure of IseA, an Inhibitor Protein of dl-Endopeptidases from Bacillus subtilis, Reveals a Novel Fold with a Characteristic Inhibitory Loop

Background: IseA from Bacillus subtilis is an inhibitor protein against dl-endopeptidases including synthetic lethal autolysins.

Results: We solved the solution structure of IseA using NMR. Titration with LytF dl-endopeptidase indicated interaction sites around the loop region.

Conclusion: The IseA structure revealed a novel “hacksaw”-like fold with a characteristic inhibitory loop.

Significance: The results suggest a new inhibition mechanism of IseA with a unique loop.

Abstract

In Bacillus subtilis, LytE, LytF, CwlS, and CwlO are vegetative autolysins, dl-endopeptidases in the NlpC/P60 family, and play essential roles in cell growth and separation. IseA (YoeB) is a proteinaceous inhibitor against the dl-endopeptidases, peptidoglycan hydrolases. Overexpression of IseA caused significantly long chained cell morphology, because IseA inhibits the cell separation dl-endopeptidases post-translationally. Here, we report the first three-dimensional structure of IseA, determined by NMR spectroscopy. The structure includes a single domain consisting of three α-helices, one 3₁₀-helix, and eight β-strands, which is a novel fold like a “hacksaw.” Noteworthy is a dynamic loop between β4 and the 3₁₀-helix, which resembles a “blade.” The electrostatic potential distribution shows that most of the surface is positively charged, but the region around the loop is negatively charged. In contrast, the LytF active-site cleft is expected to be positively charged. NMR chemical shift perturbation of IseA interacting with LytF indicated that potential interaction sites are located around the loop. Furthermore, the IseA mutants D100K/D102K and G99P/G101P at the loop showed dramatic loss of inhibition activity against LytF, compared with wild-type IseA, indicating that the β4–3₁₀ loop plays an important role in inhibition. Moreover, we built a complex structure model of IseA-LytF by docking simulation, suggesting that the β4–3₁₀ loop of IseA gets stuck deep in the cleft of LytF, and the active site is occluded. These results suggest a novel inhibition mechanism of the hacksaw-like structure, which is different from known inhibitor proteins, through interactions around the characteristic loop regions with the active-site cleft of enzymes.

Introduction

The cell envelope of the Gram-positive bacterium, Bacillus subtilis, is composed of a single cytoplasmic membrane covered by a thick cell wall. The thick cell wall behaves as an exoskeleton, protecting the bacterium from many stresses. However, degradation of the cell wall is necessary for cell growth (1, 2). In B. subtilis, proteins in the cell wall include not only synthetic enzymes of peptidoglycan but also autolytic enzymes of the peptidoglycan, which are involved in cell wall turnover, cell growth, and cell separation (2, 3). From the autolytic enzymes of B. subtilis, we have studied especially dl-endopeptidases, which hydrolyze the linkage of d-γ-glutamyl-meso-diaminopimelic acid in peptidoglycan. B. subtilis has four major vegetative dl-endopeptidases, LytE, LytF, CwlS, and CwlO (4–7), that are members of the NlpC/P60 protein domain family (Pfam entry, PF00877), a family of cell wall cysteine peptidases (8). LytE, LytF, and CwlS are cell separation autolysins (4–6), and the lytE cwlO genotype showed synthetic lethality (1, 2). These enzymes have similar domain structures consisting of putative cell wall-binding domains, which are known as LysM repeats in LytE, LytF, and CwlS or an uncharacterized domain in CwlO, in the N termini, and the catalytic domains of dl-endopeptidase in the C termini (30–70% amino acid identities). In addition, another dl-endopeptidase, PgdS from B. subtilis, is a poly-γ-glutamic acid degradase, which also belongs to the NlpC/P60 family (9).

Salzberg and Helmann (10) reported that the cell surface protein, IseA (YoeB) from B. subtilis, was induced by cell envelope-targeting antibiotics. The iseA (yoeB) mutant exhibited an increased rate of autolysis in response to cell wall-targeting antibiotics or nutrient depletion, whereas a phenotypic difference between the deletion mutant and the wild type did not appear in the peptidoglycan structure or the levels of autolysins. Bisicchia et al. (1, 11) reported that iseA is one of the direct targets for negative regulation by YycF (WalR), a response regulator of the essential two-component system YycFG (WalRK).

In our previous study (12), we demonstrated that IseA is a proteinaceous inhibitor of the vegetative dl-endopeptidases, LytE, LytF, CwlS, and CwlO, through direct interaction with their catalytic domains. Immunofluorescence microscopy with 3×FLAG tag indicated that IseA localizes specifically at cell separation sites and cell poles on the vegetative cell surface in a similar manner as the dl-endopeptidases. Overexpression of IseA caused significantly long chained cell morphology in the exponential growth phase, because IseA regulates the cell-separation dl-endopeptidases post-translationally.

Among proteinaceous inhibitors against cell wall lytic enzymes, Ivy family proteins, which inhibit vertebrate lysozymes and/or endogenous bacterial lytic transglycosylases, have been reported in Gram-negative bacteria (13–15). However, the amino acid sequences of Ivy and IseA show no detectable homology, suggesting that the structure and inhibition mechanism of IseA should be different from Ivy. Many bacterial cell wall autolytic enzymes, including dl-endopeptidases and lytic transglycosylases, play essential roles in cell wall turnover, cell growth, and cell division (2, 3, 12, 16). For instance, in Listeria monocytogenes, a food-borne pathogen, deletion of iap encoding p60, a cell wall lytic enzyme from the NlpC/P60 family, led to abnormal cell division and loss of actin-based motility, and the iap mutant was strongly attenuated in a mouse model after intravenous injection, demonstrating the importance of p60 during infection (17). Furthermore, the lytE cwlO genotype, whose protein products LytE and CwlO can be inhibited by IseA (12), led to synthetic lethality of B. subtilis (2). The bacterial cell wall autolytic enzymes should be tightly controlled within the producing cells because of their autolytic potential (16). These findings imply that understanding the inhibition and regulation mechanism of the cell wall lytic enzymes and their inhibitor proteins may contribute to development of antibacterial agents against new targets.

IseA homologs are present within several Bacillus and Paenibacillus species. For example, the B. subtilis IseA protein shares 77, 76, 65, 50, 49, 47, and 32% amino acid identity with the IseA homologs from Bacillus amyloliquefaciens, Bacillus atrophaeus, Bacillus licheniformis, Bacillus pumilus, Bacillus sp. NRRL, Bacillus megaterium, and Paenibacillus polymyxa, respectively (Fig. 1). However, IseA and its homologs do not show any significant similarity with protein domains of known three-dimensional structures. To understand the IseA inhibition mechanism, it is vital to solve its three-dimensional structure. Here, we report the first three-dimensional structure of IseA from B. subtilis, determined by solution nuclear magnetic resonance (NMR) spectroscopy. The structure shows a novel fold with a characteristic loop, and we suggest an inhibition mechanism of IseA against dl-endopeptidases.

FIGURE 1.

Amino acid sequence alignment of homologs of B. subtilis IseA (YoeB). B. subtilis IseA (YoeB) homologs are derived from B. amyloliquefaciens FZB42, B. atrophaeus 1942, B. licheniformis ATCC 14580, B. pumilus ATCC 7061, Bacillus sp. NRRL B-14911, B. megaterium DSM 319, and P. polymyxa SC2. Their deduced signal peptide sequences, predicted by SignalP 4.0 (55), are omitted. The alignment was generated by ESPript (56) with ClustalW2 (57). The secondary structures of the IseA structure determined by NMR are shown above (α, α-helix; β, β-strand; η, 3₁₀-helix; TT, β-turn) as analyzed by DSSP (Define Secondary Structure of Proteins) (58). Green symbols at the bottom indicate a disulfide bond. The amino acid sequence of B. subtilis (NMR) is the B. subtilis IseA protein expression construct used for the NMR experiments in this study. The first residue, Gly-1, is a cloning artifact residue. The full-length sequence of the original IseA (YoeB) from B. subtilis (strain 168) consists of 181 amino acid residues, including 23 residues of the signal peptide (Met-1–Asn-23) predicted by SignalP. In this study, the original residue numbers Lys-24–Ile-181 of B. subtilis IseA were renumbered Lys-2–Ile-159 of the IseA expression construct for the NMR experiments.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification

The DNA fragment encoding IseA (the original amino acid sequence from Lys-24 to Ile-181) was amplified from pQEYoeB (12) by PCR. The amplified DNA fragment was subcloned into the expression vector pCR2.1 (Invitrogen) as a fusion with an N-terminal histidine affinity tag and a tobacco etch virus protease cleavage site. The uniformly ¹³C/¹⁵N-labeled fusion protein was synthesized by a cell-free protein expression system (18–21). The cell-free reaction solution was first adsorbed to a HisTrap column (GE Healthcare), which was washed with buffer A (20 mm Tris-HCl buffer, pH 8.0, containing 300 mm NaCl and 20 mm imidazole) and eluted with buffer B (20 mm Tris-HCl buffer, pH 8.0, containing 300 mm NaCl and 500 mm imidazole). To remove the histidine tag, the eluted protein was incubated at 297 K for 16 h with the tobacco etch virus protease (22), producing a protein consisting of IseA residues Lys-2–Ile-159 (corresponding to the original residue numbers Lys-24–Ile-181 of B. subtilis IseA) with the first Gly-1 of a cloning artifact residue. After desalting with buffer A, the sample was applied to a HisTrap column. The flow-through fraction was fractionated on a HisTrap SP column by a concentration gradient of 20 mm Tris-HCl buffer, pH 8.0, containing 300 mm NaCl and 1 m NaCl, and the fractions containing IseA were collected. For NMR measurements, the purified protein was concentrated to 1.12 mm in 20 mm deuterium-labeled Tris buffer, pH 7.0, ¹H₂O/²H₂O (9:1), containing 300 mm NaCl and 0.02% NaN₃.

NMR Spectroscopy and Resonance Assignments

All NMR data were collected on Bruker AVANCE 600, 700, and 800 MHz spectrometers equipped with a triple-resonance CryoProbe at 298 K, at RIKEN NMR Facility (Tsurumi, Yokohama, Japan). Backbone and side-chain assignments were obtained by standard triple resonance experiments (23, 24). Two-dimensional ¹H-¹⁵N HSQC,³ and three-dimensional HNCO, HN(CA)CO, HNCA, HN(CO)CA, HNCACB, and CBCA(CO)NH spectra were used for the ¹H, ¹⁵N, and ¹³C assignments of the protein backbone. Side-chain ¹H and ¹³C assignments were obtained using two-dimensional ¹H-¹³C HSQC, and three-dimensional HBHA(CO)NH, C(CCO)NH, HC(C)H-total correlation spectroscopy, (H)CCH-total correlation spectroscopy, and HC(C)H-correlation spectroscopy spectra. All the assignments were checked for consistency with three-dimensional ¹⁵N- and ¹³C-edited HSQC-NOESY spectra. Three-dimensional NOESY spectra were recorded with mixing times of 80 ms.

The NMR data were processed with the NMRPipe program (25). The programs KUJIRA (26) and NMRView (27, 28) were used for visualization of the NMR spectra and for chemical shift assignments.

Structure Calculations

Peak lists for the NOESY spectra were generated by interactive peak picking, and peak intensities were determined by the automatic integration function of NMRView (27, 28). The three-dimensional structure was determined by the combined automated NOESY cross-peak assignment and structure calculation with torsion angle dynamics implemented in the CYANA program (29). Restraints for the backbone torsion angles φ and ψ were determined by a chemical shift database analysis with the TALOS program (30). The structure calculations were started from 100 extended conformers, and the 20 conformers with the lowest final CYANA target function values were selected. The 20 structures from the CYANA calculation were subjected to restrained energy refinement with the AMBER program (31). PROCHECK_NMR (32) was used to validate the final structures. Structure figures were prepared with the MOLMOL program (33) and PyMOL (DeLano Scientific).

Measurements of Dynamic Parameters and Residual Dipolar Coupling

The measurements of the nitrogen relaxation times, T₁ and T₂, and the proton-nitrogen heteronuclear NOEs were performed on a spectrometer using the ¹⁵N,¹³C-labeled IseA at a concentration of 1.12 mm. Relaxation delays of 10, 70, 150, 250, 370, 530, 760, and 1150 ms were recorded for the measurement of ¹⁵N T₁, and relaxation delays of 14.4, 28.8, 43.2, 57.6, 72.0, 100.8, 144.0, and 187.2 ms were recorded for the measurement of ¹⁵N T₂, using standard two-dimensional methods (34). The proton-nitrogen heteronuclear NOE values were calculated as the ratio between the cross-peak intensities with and without ¹H saturation during relaxation delays. Residual dipolar coupling was obtained using ¹H-¹⁵N IPAP HSQC spectra under the isotropic and anisotropic (polyacrylamide gel) conditions (35).

NMR Titration Experiments

The His₆-tagged LytF catalytic domain (LytF) was expressed and purified with a HisTrap HP column (GE Healthcare), as described previously (5). LytF was further purified by cation exchange chromatography (20 mm MES buffer, pH 6.5, with a linear gradient of NaCl from 0.1 to 2 m) with a Poros HS/M column (PerSeptive Biosystems) and gel filtration chromatography (20 mm MES buffer, pH 6.5, containing 150 mm NaCl) with a Superdex 75 10/300 GL column (GE Healthcare). For the amide chemical shift titration experiments, the catalytic domain of purified LytF was prepared in 20 mm deuterium-labeled Tris buffer, pH 7.0, ¹H₂O/²H₂O (9:1), containing 300 mm NaCl and 0.02% NaN₃. Two-dimensional ¹H-¹⁵N HSQC spectra were recorded on Bruker AVANCE 600 and 700 MHz spectrometers with LytF domain concentrations relative to the IseA concentration (0.06 mm) to the ratio of IseA/LytF = 1:0, 1:0.5, 1:1, and 1:2.

Cell Wall Hydrolase Assay

Expression vectors of the IseA mutants were made from pQEYoeB (12), using megaprimer PCR. The IseA mutants with His₆ tags were expressed in Escherichia coli and purified using TALON metal affinity resin (Clontech), and cell wall hydrolase assays were performed as described previously (12). Briefly, the reaction mixture included B. subtilis cell wall (final A₅₄₀ = 0.3) in 20 mm MES-NaOH buffer, pH 6.5, containing 100 mm KCl, and was preincubated at 37 °C. The protein mixtures containing the catalytic domain of LytF (1 nmol) and each IseA variant (2 nmol) (mutants D100N/D102N, D100K/D102K, and G99P/G101P or wild type) were preincubated in the buffer at 37 °C for 5 min. Then each protein mixture was added to the reaction mixture (final volume of 2 ml) in a cuvette with a stirrer. The A₅₄₀ of the reaction mixture was measured at 37 °C with a spectrophotometer (V-560; Jasco, Tokyo, Japan).

Homology Modeling and Docking Simulation

The structure models of the catalytic domains were predicted by homology modeling using the SWISS-MODEL program (36–38). The template structure was the structure of lipoprotein Spr from E. coli (residues 36–164; PDB code 2K1G) (39). Moreover, a structure model of IseA and the LytF catalytic domain complex was predicted by docking simulation using the docking web server GRAMM-X (40, 41). The protein docking software GRAMM-X and its web interface extended the original GRAMM Fast Fourier Transformation methodology (42–44) by employing smoothed potentials, refinement stage, and knowledge-based scoring (40, 41).

RESULTS

Structure Determination

First, we tried to crystallize the IseA protein to solve the three-dimensional structure by x-ray crystallography. However, we obtained only thin microcrystals with almost no x-ray diffraction. Thus, we performed NMR spectroscopy experiments to determine the structure of IseA.

We measured a complete set of standard two- and three-dimensional NMR spectra (23, 24). The two-dimensional ¹H-¹⁵N HSQC spectrum of the IseA showed dispersed, sharp NH signals with similar intensities, which are characteristic of a folded protein (Fig. 2A). One disulfide bond between Cys-40 and Cys-56 was confirmed by their proton nuclear Overhauser effect (NOE) peaks and the chemical shifts of their Cβ atoms, which were 44.59 and 47.33 ppm, respectively (45).

FIGURE 2.

Structural analysis of B. subtilis IseA using NMR spectroscopy. A, two-dimensional ¹H-¹⁵N HSQC spectrum of IseA. The spectrum was recorded at a ¹H frequency of 800 MHz, pH 7.0, at 298 K. Cross-peaks are labeled with the residue numbers. Side-chain resonances of asparagine and glutamine residues are indicated by horizontal lines. B, backbone traces of the 20 energy-minimized conformers of the solution structure of IseA (stereoview). C, ribbon representation of the IseA structure (stereoview). Color gradient (blue, green, yellow, and red) is from the N to the C terminus. A disulfide bond between Cys-40 and Cys-56 is shown as stick models.

More than 20 NOE distance restraints per residue, including 1320 long range distance restraints, were used in the final structure calculations with the CYANA program (29). The final structures were energy-refined with the AMBER program (31). The final 20 energy-minimized conformers that represent the solution structure of IseA were well defined (Fig. 2B) and showed excellent agreement with the statistical data (Table 1). The precision of the structure was characterized by root-mean-square deviation values to the mean coordinates of 0.40 Å for the backbone and 0.66 Å for all heavy atoms, excluding the unstructured regions of residues Gly-1–Asp-7 at the N termini (Fig. 2B). In the ordered regions, which include Gln-10–Met-28 (α1), Thr-44–Tyr-47 (β1), Leu-50–Tyr-54 (β2), Lys-62–Leu-69 (α2), Lys-75–Lys-83 (α3), Thr-88–Ser-90 (β3), Lys-93–Pro-97 (β4), Trp-107–Lys-109 (η1(3₁₀-helix)), Thr-111–Lys-118 (β5), Thr-121–Val-128 (β6), Ala-137–Lys-145 (β7), and Trp-150–Val-152 (β8), the root-mean-square deviation values to the mean coordinates were 0.25 Å for the backbone and 0.53 Å for all heavy atoms, respectively. The quality of the structure is also reflected by the fact that 89.8% of the (φ,ψ) backbone torsion angle pairs were found in the favored regions and 9.9% were within the additionally allowed regions of the Ramachandran plot, according to the PROCHECK_NMR program (32).

TABLE 1.

Summary of conformational restraints and statistics of the final 20 structures

NOE upper distance limits
Total	3659
Intraresidue (|i − j| = 0)	852
Sequential (|i − j| = 1)	836
Medium range (1 < |i − j| < 5)	651
Long range (|i − j| ≥ 5)	1320
Torsion angle restraints	154
CYANA target function value	0.24
Distance restraint violations
Number >0.1 Å	0
Maximum (Å)	–
Torsion angle restraint violations
Number >5°	0
Maximum (°)	–
Energies of AMBER calculation (kcal/mol)
Mean AMBER energy	−4076.34
Ramachandran plot statisticsa
Residues in favored regions	89.8%
Residues in additionally allowed regions	9.9%
Residues in generously allowed regions	0.2%
Residues in disallowed regions	0.1%
Root mean square deviation to the averaged coordinates
All regionsb
Backbone atoms	0.40 Å
Heavy atoms	0.66 Å
Ordered regionsc
Backbone atoms	0.25 Å
Heavy atoms	0.53 Å

^a The region for the PROCHECK calculation includes Leu-8–Ile-159.

^b All regions include Leu-8–Ile-159.

^c Ordered regions include Gln-10–Met-28 (α1), Thr-44–Tyr-47 (β1), Leu-50–Tyr-54 (β2), Lys-62–Leu-69 (α2), Lys-75–Lys-83 (α3), Thr-88–Ser-90 (β3), Lys-93–Pro-97 (β4), Trp-107–Lys-109 (3₁₀), Thr-111–Lys-118 (β5), Thr-121–Val-128 (β6), Ala-137–Lys-145 (β7), Trp-150–Val-152 (β8).

Moreover, to confirm the reliability of the IseA structure, we performed residual dipolar coupling analysis with polyacrylamide gel (supplemental Fig. S1). This analysis also supported the high accuracy of the IseA structure determined by NMR.

Overall Structure of IseA

The solution structure of the IseA protein includes a single domain consisting of three α-helices (α1, α2, and α3), two anti-parallel β-sheets (β1, β2, β4, and β3; β5, β6, β7, and β8), and one 3₁₀-helix (η1) (Figs. 1 and 2C). One disulfide bond, Cys-40–Cys-56, forms a bridge between the two loops, α1-β1 and β2-α2. The three α-helices constitute a backbone structure in the central region. Each of the two anti-parallel β-sheets is located on the two polar ends of the α-helices. A particularly noteworthy region is the loop (Gly-99–Asn-103) between β4 and 3₁₀-helix (η1).

The structural homology search results against the IseA structure by the Dali server (46) revealed several structures whose Z-scores were above 5 (Fig. 3). However, unlike the separated two β-sheets of IseA, these structures have one continuous β-sheet, suggesting that the IseA structure has a novel fold.

FIGURE 3.

Structural homology search results of the IseA structure. Representative structures of three-dimensional structural homologs (Z-score >5) searched by the Dali server (46). These structures are shown as a monomer. A, solution structure of IseA from B. subtilis. B, crystal structure of hypothetical protein YybH from B. subtilis (Northeast Structural Genomics Consortium, unpublished data.). C, crystal structure of the association domain of Ca²⁺/calmodulin-dependent kinase II from Mus musculus (59). D, crystal structure of nuclear transport carrier NTF2 from Rattus norvegicus (60). E, crystal structure of scytalone dehydratase from Magnaporthe grisea (61).

The electrostatic potential distribution (Fig. 4A) and the highly conserved residues (Fig. 4B) on the solvent-accessible surface of the IseA structure clearly show the negatively charged and conserved region around the β4–3₁₀ loop (Gly-99–Asn-103), suggesting that the region is probably important for the function. Interestingly, the loop and the rest of the structure make a through-hole within the IseA structure (Fig. 4). The uncommon framework structure of IseA looks like a “hacksaw,” in which the three α-helices constitute a backbone frame, the two β-sheets make a side frame, and the acidic loop resembles a “blade.”

FIGURE 4.

Surface presentation of the IseA structure. A, electrostatic potential distribution on the solvent-accessible surface of the IseA structure. Blue and red surfaces represent positive and negative potentials, respectively. The potentials were calculated at ±3 kT/e using the PyMOL APBS Tools (M. Lerner, unpublished data). B, residue conservation mapping on the surface of the IseA structure. Magenta and orange surfaces represent completely and highly conserved residues on the solvent-accessible surface of the IseA structure, respectively. The views at both sides were made by ±90° vertical axis rotation. The arrowheads indicate the strictly conserved Asp-100–Gly-101 region at the β4–3₁₀ loop.

Dynamic Properties of the IseA

To determine the dynamic properties of the IseA structure, the ¹⁵N T₁ and T₂ relaxation times and the steady-state ¹H-¹⁵N NOEs were measured at 298 K (Fig. 5). The average NOE values for the α- and 3₁₀-helices, β-sheets, and C-terminal regions were above 0.8. These values are in agreement with the result that these regions are tightly packed to form a rigid body in solution. The NOE values for the N-terminal extension before α1 decreased gradually when approaching the chain terminus, indicating that the N-terminal region is intrinsically disordered in solution. The lower NOE values, shorter T₁ and longer T₂, for the two loop regions, α1-β1 (Lys-36–Gly-43) and β4–3₁₀ (Gly-99–Asn-103), compared with the above-mentioned rigid core, indicate that the two loop regions are flexible and dynamic. Incidentally, these flexible and dynamic N-terminal and loop regions may be the cause for the difficulty to crystallize IseA.

FIGURE 5.

NMR analysis of dynamic properties of IseA. A, ¹⁵N T₁ relaxation time of IseA. B, ¹⁵N T₂ relaxation time of IseA. C, steady-state ¹H-¹⁵N NOEs of IseA. These data suggest that the N-terminal region and the two loop regions, α1-β1 (Lys-36–Gly-43) and β4–3₁₀ (Gly-99–Asn-103), are flexible and dynamic.

Titration of IseA with the Catalytic Domain of LytF

To investigate the interaction site of IseA with the catalytic domain of dl-endopeptidases, we performed two-dimensional ¹H-¹⁵N HSQC titration experiments with the addition of the LytF catalytic domain (LytF). Fig. 6A shows the two-dimensional ¹H¹⁵N HSQC spectra of samples whose molar ratios of IseA/LytF were 1:0 (black), 1:0.5 (red), 1:1 (green) and 1:2 (blue). The addition of LytF to IseA induced chemical shift changes in the HSQC spectra of IseA. The spectra, in the cases of the 1:1 and 1:2 ratios, were very similar, suggesting that IseA mostly forms the 1:1 stoichiometric complex with LytF under the 1:1 and 1:2 ratio conditions in the NMR experiments. The close-up view of the dotted line rectangle in Fig. 6A is shown in Fig. 6B. The left panel of Fig. 6B shows that the addition of the LytF to IseA (IseA/LytF = 1:0.5) induced a reduction in peak intensity of the original IseA (e.g. A, B, C, and A157) and an emergence of new peaks (e.g. A′, B′, C′, and D′), probably derived from the IseA-LytF complex in the two-dimensional ¹H-¹⁵N HSQC spectrum. In the case of IseA/LytF = 1:2 (Fig. 6B, right panel), the original peaks of IseA (A, B, C, and A157) disappeared, and the intensity of the new peaks (A′, B′, C′, and D′) increased. These new peaks of IseA/LytF = 1:0.5 and 1:2 completely overlapped, suggesting that the interaction between IseA and LytF is in slow exchange on the NMR time scale. Furthermore, after the addition of LytF, there were several peaks such as A157, which we could not trace in the two-dimensional ¹H-¹⁵N HSQC spectrum because their chemical shift changes were too big. These untraceable peaks, including His-23, Phe-24, Trp-25, Met-28, Ser-29, Gly-30, His-31, Ile-98, Gly-99, Gly-101, Asp-102, Leu-105, Trp-107, Thr-130, Asp-132, Gly-133, Asn-153, Phe-155, Ala-157 and Ile-159 (Fig. 6A), suggest that these residues are involved in the interaction between IseA and LytF, directly or indirectly. These residues are located at the β4–3₁₀ loop, the β6-β7 loop, the latter half of α1, and the C terminus of the IseA structure (Fig. 7), suggesting that the surface regions interacting with LytF are distributed around the β4–3₁₀ loop.

FIGURE 6.

Titration of IseA with the catalytic domain of LytF. A, two-dimensional ¹H-¹⁵N HSQC spectra of the samples whose molar ratios of IseA/LytF (catalytic domain) were 1:0 (black), 1:0.5 (red), 1:1 (green), and 1:2 (blue). The residue numbers of the untraceable peaks are shown as magenta text. B, close-up view of two-dimensional ¹H-¹⁵N HSQC spectra in the dotted line rectangle in A. The left panel shows the spectra after the addition of LytF to IseA at a ratio of 1:0.5 (IseA/LytF), and the right panel shows the spectra after the addition of LytF to IseA at a ratio of 1:2.

FIGURE 7.

Mapping of the untraceable peak residues in the titration experiment on the IseA structure (stereoviews). The residues of the untraceable peaks in the two-dimensional ¹H-¹⁵N HSQC spectra are shown as magenta stick models with labels. B, view was made by 90° vertical axis rotation of the A view.

Cell Wall Hydrolase Assay

Based on these results, we speculated that the acidic and dynamic β4–3₁₀ loop of IseA is important for the inhibition of dl-endopeptidases. To test this speculation, we performed site-directed mutagenesis at the β4–3₁₀ loop and inhibition analysis of cell wall hydrolase assay. We made three IseA mutants as follows: D100N/D102N and D100K/D102K changed the negative charge on the β4–3₁₀ loop to a neutral and positive charge, respectively, and G99P/G101P restricted the flexibility and structure of the β4–3₁₀ loop. Fig. 8 shows the inhibition activities of the IseA mutants against LytF cell wall lysis. The IseA mutants D100K/D102K and G99P/G101P showed dramatic loss of inhibition activity against LytF, compared with wild-type IseA, suggesting that the β4–3₁₀ loop plays an important role in the inhibition of the enzyme, although the D100N/D102N mutant showed almost the same inhibition activity as wild-type IseA.

FIGURE 8.

Cell wall hydrolytic curves of dl-endopeptidases with the IseA mutants. The cell wall hydrolytic curves. LytF catalytic domain (LytF) only (gray line), LytF and wild-type IseA (black line), LytF and IseA mutants D100N/D102N (green line), D100K/D102K (blue line), or G99P/G101P (red line) are shown.

DISCUSSION

Structural and Surface Properties of Catalytic Domains of dl-Endopeptidases, Predicted by Homology Modeling

To speculate about the interaction mechanism between IseA and the catalytic domains of the target dl-endopeptidases, model structures of the catalytic domains were predicted by homology modeling using the SWISS-MODEL program (36–38). The template structure was the structure of lipoprotein Spr from E. coli (residues 36–164; PDB code 2K1G) (39), which is a member of the NlpC/P60 protein domain family (Pfam entry, PF00877) (8) and shares 38, 34, 33, 30, and 36% identity with the catalytic domains of LytF, LytE, CwlS, CwlO, and PgdS, respectively (Fig. 9A). Global model quality estimation of these homology modeling structures by QMEAN (47) showed reasonable QMEAN scores (LytF, 0.62; LytE, 0.59; CwlS, 0.59; and PgdS, 0.61) and QMEAN Z-scores (LytF, −1.8; LytE, −2.1; CwlS, −2.1; and PgdS, −2.0) except for CwlO (QMEAN score, 0.47; QMEAN Z-score, −3.5), suggesting that we can discuss global structure and surface properties of some dl-endopeptidases using the homology modeling structures (Fig. 9).

FIGURE 9.

Homology modeling of catalytic domains of dl-endopeptidases. A, sequence alignment of catalytic domains of lipoprotein Spr from E. coli; dl-endopeptidases LytF, LytE, CwlS, CwlO, and PgdS from B. subtilis; AvPCP from Anabaena variabilis, and BcYkfC from B. cereus. The deduced catalytic triad residues are shown as red arrowheads below the alignment. The secondary structures of the Spr structure (PDB code 2K1G) (39) are shown above the alignment (α, α-helix; β, β-strand; η, 3₁₀-helix; TT, β-turn) as analyzed by the DSSP program. B, superimposition of the main-chain structure models of the catalytic domains of dl-endopeptidases, predicted by homology modeling using SWISS-MODEL. LytF (residues 370–488; green), LytE (residues 230–334; cyan), CwlS (residues 297–414; magenta), CwlO (residues 340–472; orange), PgdS (residues 64–286; yellow), and Spr (residues 36–164; gray; PDB code 2K1G). The deduced catalytic triad residues are shown as stick models and indicated by red arrowheads. C, electrostatic potential distribution on the solvent-accessible surface of the IseA structure. Blue and red surfaces represent positive and negative potentials, respectively. The potentials were calculated at ±3 kT/e using the PyMOL APBS Tools. The dashed line ellipse indicates the β4–3₁₀ loop region of IseA. D, electrostatic potential distribution on the solvent-accessible surface of the homology structure models of the catalytic domains of dl-endopeptidases. Blue and red surfaces represent positive and negative potentials, respectively. The potentials were calculated at ±3 kT/e using the PyMOL APBS Tools. The dashed line ellipses indicate their active-site cleft regions.

Structural analyses of Spr (39), AvPCP and NpPCP endopeptidases from cyanobacteria (48), and YkfC from B. cereus (BcYkfC) (49), from the NlpC/P60 family, showed a characteristic Cys-His-His catalytic triad in the active sites of these cysteine peptidases. In the deduced active sites of LytF, LytE, CwlS, CwlO, and PgdS (shown as red arrowheads in Fig. 9, A and B), the first and second residues, Cys and His, are strictly conserved (Fig. 9A), but the third residue is His in LytE and CwlS, Asn in LytF and CwlO, and Gln in PgdS. Indeed, we demonstrated that the conserved cysteine residues of LytE and CwlO are essential for their enzymatic activity (2). Generally, in cysteine proteases/peptidases, the two residues of the active site, Cys and His, catalyze the hydrolysis reaction (50). Thus, the deduced catalytic triads of LytF, LytE, CwlS, CwlO, and PgdS probably work as the catalytic residues.

Fig. 9D shows the electrostatic potential distribution on the solvent-accessible surface of the homology modeling structures of the dl-endopeptidases. The dashed line ellipses in Fig. 9D indicate their active-site cleft regions. These figures suggest that the surface regions around the active-site cleft of LytF, LytE, CwlS, and CwlO are significantly positively charged. However, the surface region around the active-site cleft of PgdS is neutral or negatively charged, relative to the others.

Inhibition Mechanism of IseA against the Catalytic Domains of dl-Endopeptidases

As shown in Fig. 9D, the surface regions around the enzymatic active sites of the dl-endopeptidases, except PgdS, are positively charged. The negatively charged region of IseA around the β4–3₁₀ loop (Fig. 9C), which widely overlaps with the potential LytF-interacting regions according to the NMR titration experiment (Fig. 7B), is complementary to the positively charged regions around the active sites of the dl-endopeptidases, except PgdS. This is consistent with the preliminary experimental result that IseA did not inhibit the enzymatic activity of PgdS,⁴ whereas IseA inhibited the enzymatic activity of LytF, LytE, CwlS, and CwlO (12). Furthermore, in the surface configurations, the narrow and stretch structure of the IseA β4–3₁₀ loop, looking like the blade of a hacksaw, is a complementary shape to the narrow and long cleft of the enzymatic active site of the dl-endopeptidases (Fig. 9D). Therefore, we propose the hypothesis that the acidic and dynamic IseA β4–3₁₀ loop occludes specifically the active-site cleft of dl-endopeptidases, and also the acidic surface regions around the loop widely interact with the basic surface around the active site of the enzyme, as suggested by the NMR titration experiment (Fig. 7), i.e. IseA inhibits the enzymatic activity by not only the β4–3₁₀ loop occupying the active site but also IseA widely blocking the regions around the active site and the substrate-binding site.

To partially test this hypothesis, we performed site-directed mutagenesis (D100N/D102N, D100K/D102K, and G99P/G101P) of IseA and inhibition analysis against LytF. Fig. 8 demonstrates that the IseA mutants D100K/D102K and G99P/G101P showed dramatic loss of inhibition activity compared with wild-type IseA, indicating that the β4–3₁₀ loop plays an important role in the inhibition mechanism. However, the D100N/D102N mutant showed almost the same inhibition activity as the wild-type IseA. The inhibition activity of the D100N/D102N mutant seemed to be contradictory to the above-mentioned hypothesis about charge complementation. However, we speculate that Asp-100 and/or Asp-102 of IseA may be recognized by LytF not through basic residues but polar residues, which are compatible with the D100N/D102N mutation. In the crystal structure of BcYkfC complexed with the reaction product, l-Ala-γ-d-Glu, the α-carboxyl of d-Glu was recognized by hydrogen bond interactions with the side chains of Ser-239 and Ser-257 (49). Because the Ser-239 residue next to the catalytic cysteine is strictly conserved among the homologs (shown by a blue arrowhead in Fig. 9A), Ser-401 of LytF, corresponding to Ser-239 of BcYkfC, may also recognize the α-carboxyl of d-Glu of the peptidoglycan substrate. In addition, we speculate that the β-carboxyl of l-Asp (Asp-100 or Asp-102) of IseA may mimic the α-carboxyl of d-Glu of the substrate, recognized by a hydrogen bond with Ser-401 of LytF, and a side-chain amide of Asn (D100N or D102N) of the IseA mutant (D100N/D102N) may be compatibly recognized by a hydrogen bond with Ser-401 of LytF.

In contrast, the structures and inhibition mechanisms of several proteinaceous inhibitors such as cystatin and chagasin against cysteine proteases/peptidases have been reported (51, 52). The structure of cystatin from chicken egg white consists of one α-helix and one anti-parallel β-sheet, and the structure of chagasin from the parasite Trypanosoma cruzi consists of two anti-parallel β-sheets known as an immunoglobulin-like β-sandwich fold. Although the overall structures of cystatin and chagasin are different, they bind to the same class of cysteine proteases via remarkably similar tripartite interactions using short loops and the N terminus (51). However, the structure of IseA is clearly distinct from the structures of cystatin and chagasin, and our interaction and mutagenesis analyses suggest that the inhibition mechanism of IseA is different from that of cystatin and chagasin.

Moreover, the structures of cathepsin L cysteine proteases with their propeptides from human and Toxoplasma gondii have been reported (53, 54). The amino acid sequences of the propeptides bound to the active-site cleft are MNGFQ (residues 75p–79p) from human and YLGFK (residues 177p–181p) from T. gondii. The structures show that the binding regions of the propeptides to the active-site cleft of cathepsin L form strained conformations, and the binding mode contains hydrogen bonds and hydrophobic interactions. The β4–3₁₀ loop sequence of IseA is GDGDN (residues 99–103), which plays an important role in the inhibition mechanism. The IseA β4–3₁₀ loop also has glycine residues, and it is a flexible and dynamic loop without secondary structure. However, the IseA β4–3₁₀ loop has no hydrophobic side chains in contrast to the binding regions of the cathepsin L propeptides, suggesting that the binding mechanism of IseA is different from that of the propeptides of cathepsin L.

In addition, the structure of E. coli Ivyc, a lysozyme inhibitor protein in the Ivy family, complexed with hen egg white lysozyme has been reported (13). Through the complex formation, the lysozyme active site becomes occluded by a loop protruding from the Ivyc molecule. The loop is relatively rigid because it consists of the conserved amino acid sequence, CKPHDC (residues 57–62), in which there is a disulfide bridge between the two Cys residues and a salt bridge between the Lys and Asp residues. Conversely, the IseA β4–3₁₀ loop, which looks like a “hacksaw blade,” is flexible and dynamic (Figs. 2B and 5). These contrasts suggest that the inhibition mechanism of IseA is also distinct from that of Ivyc.

Finally, to visualize the proposed novel inhibition mechanism, we present a model of the complex structure of IseA and the LytF catalytic domain, predicted by the GRAMM-X docking simulation program, which employs a smoothed Lennard-Jones potential on a fine grid during the global search FFT stage, followed by the refinement optimization in continuous coordinates and rescoring with several knowledge-based potential terms (40, 41). The complex structure model of IseA-LytF shows that the β4–3₁₀ loop of IseA gets stuck deep in the cleft of LytF, and the active site is physically occupied (Fig. 10). Fig. 10B shows that most of the potential interacting residues (shown as magenta stick models), indicated by the NMR titration experiment, are proximately nestled around the wide surface of LytF, suggesting that the complex structure model is plausible.

FIGURE 10.

Structure model of the complex of IseA and the LytF catalytic domain. A, ribbon representation of a complex structure model of IseA (red) and the LytF catalytic domain (cyan), predicted by GRAMM-X docking simulation (40, 41) (stereoview). The deduced catalytic triad (Cys-400, His-449 and Asn-461) of LytF is shown as stick models and indicated by arrowheads. B, complex structure model is represented as the electrostatic potential surface of LytF, and ribbon representation of IseA with the potential interacting residues (shown as magenta stick models) indicated by the NMR titration experiment. The view of IseA is the same as in Fig. 7A. Blue and red surfaces of LytF represent positive and negative potentials, respectively. The potentials were calculated at ±3 kT/e using the PyMOL APBS Tools.

In conclusion, the first three-dimensional structure of IseA determined by NMR reveals a novel fold with a characteristic inhibitory loop, suggesting a novel inhibition mechanism against dl-endopeptidases. This study provides the mechanistic insights into the B. subtilis cell growth and separation events controlled post-translationally by the molecular interactions between IseA and dl-endopeptidases. To understand the recognition mechanism in atomic level, further studies are necessary.