The membrane localization domains of two distinct bacterial toxins form a 4‐helix‐bundle in solution

Abstract

Membrane localization domain (MLD) was first proposed for a 4‐helix‐bundle motif in the crystal structure of the C1 domain of Pasteurella multocida toxin (PMT). This structure motif is also found in the crystal structures of several clostridial glycosylating toxins (TcdA, TcdB, TcsL, and TcnA). The Ras/Rap1‐specific endopeptidase (RRSP) module of the multifunctional autoprocessing repeats‐in‐toxins (MARTX) toxin produced by Vibrio vulnificus has sequence homology to the C1‐C2 domains of PMT, including a putative MLD. We have determined the solution structure for the MLDs in PMT and in RRSP using solution state NMR. We conclude that the MLDs in these two toxins assume a 4‐helix‐bundle structure in solution.

Short abstract

PDB Code(s): 2N9V; 2N9W

Abbreviations

MARTX

multifunctional autoprocessing repeats‐in‐toxins

MLD

membrane localization domain

PMT

Pasteurella multocida toxin

RRSP

Ras/Rap1‐specific endopeptidase

Introduction

Pathogenic bacteria produce a number of toxins that influence host–pathogen interactions to aid survival within a host.1 These toxins range in size from small molecules to large proteins and confer toxicity to hosts by targeting different cellular components and processes. Many bacterial toxins consist of an effector domain that requires a receptor‐binding domain to reach specific cells and an additional translocation domain to facilitate delivery of the effector domain into the cytosol, where it gains access to its intracellular targets.

The heterotrimeric G‐protein‐deamidating toxin from Pasteurella multocida (PMT) is a 150 kDa protein comprised of N‐terminal receptor‐binding and translocation domains (residues 1–568)2, 3 and C‐terminal activity domains (residues 569–1285). The crystal structure of the C‐terminal region is resolved into C1 (residues 569–719), C2 (residues 720–1104), and C3 (residues 1105–1284) domains.4 The Gα‐protein‐deamidase activity of PMT was localized to the C3 domain.5, 6 The C1 domain contains a 4‐helix‐bundle motif (residues 589–668) that was shown to have membrane‐targeting properties capable of bringing GFP to the plasma membrane.4, 7 This structural motif was also observed in the clostridial glycosylating toxins (Fig. 1), including TcdA, TcdB, TcsL, and TcnA.8, 9, 10, 11

Figure 1.

Comparison of sequences and crystal structures of putative MLDs in bacterial toxins. (A) Amino acid sequence alignment with conserved residues color coded (purple: hydrophobic, brown: basic, green: acidic, orange: polar) and (B) the corresponding crystal structures of the individual and aligned 4‐helix‐bundle motifs found in PMT (PDB ID: 2EBF), TcdA (PDB ID: 3SS1), TcdB (PDB ID: 2BVL), TcsL (PDB ID: 2VL8), and TcnA (PDB ID: 2VK9).

Similar sequences have been found in other bacterial toxins,12 a number of which have been shown to confer membrane localization of adjacent effector domains, and are referred to as membrane localization domains (MLDs).13 The MLD and its adjacent effector domain from a MARTX toxin of Vibrio vulnificus (Ras and Rap1‐specific endopeptidase RRSP; formerly DUF5) have the strongest homology to the C1‐C2 domains of PMT.14, 15 A crystal structure of the MLD_RRSP alone (PDB 4ERR) is available, but this structure differs from the structure observed in the crystal structures with other domains present (PDB ID: 2EBF,4 3SS1,10 2BVL,9 2VL8,8 2VK911).

Considering the importance of this MLD motif in three classes of diverse toxins,13 a question arises regarding the solution structure of the MLDs. Thus, we determined the solution structures of MLD_PMT and MLD_RRSP using solution state NMR spectroscopy and compare the solution structures to the MLD moiety to that found in the crystal structures.

Results

Structure of MLD_PMT

MLD_PMT contains 79 residues [Fig. 2(A)]. In a previous study, predicted dihedral angles, determined by TALOS+, indicated that MLD_PMT contains four helices.16 For this study, the improved TALOS‐N program was used to predict dihedral angles and order parameters [Fig. 2(B)].17 Similar to the TALOS+ results, the dihedral angle prediction indicates that MLD_PMT forms four helices. Furthermore, the order parameter predictions indicate that the loop and termini residues experience greater mobility as compared to the helical residues.

Figure 2.

Secondary structures and dynamics determined by TALOS‐N. (A) Amino acid sequence alignment of MLD_PMT and MLD_RRSP. Non‐native termini residues are in grey, loop residues are in blue, and helix residues are in red. Dihedral angles (φ, ψ) and order parameters of MLD_PMT (B) and MLD_RRSP (C), predicted from TALOS‐N are plotted.

The overall fold of the solution structure is a 4‐helix‐bundle [Fig. 3(A)]. Of the top 20 structures (out of 200 total), the backbone RMSD of the ordered residues (corresponding to the helix residues) was determined to be 0.99 Å. Table 1 summarizes the statistics regarding the ensemble of structures. The ensemble of structures has an average of 2.6 ± 1.3 energy violations. Ramachandran analysis determined that 86.1% ± 2.8% of the residues were in favored regions, where violations were found in the mobile loop regions.

Figure 3.

Alignment of top 20 solution structures of MLD_PMT and MLD_RRSP. Three‐dimensional superposition of the top 20 structures of MLD_PMT (A) and MLD_RRSP (B), as determined using solution state NMR. The backbone RMSD of the helix residues of MLD_PMT is 0.99 Å and of MLD_RRSP is 0.96 Å.

Table 1.

Restraints and Statistics of Top 20 Structures

NMR distance restraints	MLDPMT	MLDRRSP
Total NOE	1485	1256
Intra‐residue NOE	606	492
Inter‐residue NOE	503	445
Sequential (|i − j| = 1)	238	180
Medium‐range (|i − j| < 4)	156	151
Long‐range (|i − j| ≥ 5)	109	114
Ambiguous NOE	376	319
Dihedral angle restraints
ϕ (TALOS‐N)	75	73
ψ (TALOS‐N)	75	73
Violations
Total violations	2.6 ± 1.3	0.1 ± 0.3
NOE violations	0	0
CDIH violations	2.3 ± 1.2	0
van der Waals violations	0.3 ± 0.6	0.1 ± 0.3
Deviations from idealized geometry
Bond lengths (Å)	0.003 ± 0.000	0.002 ± 0.000
Bond angles (°)	0.449 ± 0.010	0.380 ± 0.006
Impropers (°)	0.337 ± 0.013	0.246 ± 0.015
Ramachandran plot statistics (Molprobity)
Favored region (%)	89.6	98.7
Allowed region (%)	8.1	1.2
Disallowed region (%)	2.2	0.1
RMSD (ordered residues)
Backbone RMSD (Å)	0.97	0.97
Heavy Atom RMSD (Å)	1.76	1.73

As reported in Table 1, 1485 NOE restraints were used to calculate the structure. Of those NOE peaks, 109 were long‐range peaks that helped to define the interhelical contacts. For example, contacts between E7 to L28 define the helix 1 and 2 interface, I34 to L46 for helix 2 and 3, L47 to L77 for helix 3 and 4, and L9 to F70 for helix 1 and 4. These contacts are depicted on a representative structure [Fig. 4(A)], with the lowest RMSD with respect to the bundle of 20 structures.

Figure 4.

Interhelical NOEs of MLD_PMT and MLD_RRSP. Interhelical NOE contacts are mapped onto the structures of MLD_PMT (A) and MLD_RRSP (B). Contacts in MLD_PMT include E7 to L28 (H1‐H2), I27 to Y57 (H2‐H3), I34 to L46 (H2‐H3), M47 to L77 (H3‐H4), I54 to R74 (H3‐H4), and L9 to F70 (H1‐H4). Contacts in MLD_RRSP include V7 to L29 (H1‐H2), I28 to Y57 (H2‐H3), V35 to A46 (H2‐H3), I58 to N67 (H3‐H4), S47 to V77 (H3‐H4), and L10 to L73 (H1‐H4).

Structure of MLD_RRSP

Similar to MLD_PMT, MLD_RRSP also forms four helices and contains mobile loop and termini residues as predicted by TALOS‐N dihedral angles and order parameters [Fig. 2(B)]. We have found that MLD_RRSP also forms a 4‐helix‐bundle [Fig. 3(B)]. The backbone RMSD (ordered residues) of the top 20 structures is 0.96 Å. The structures have an average of 0.1 ± 0.3 violations, where 18 out of 20 structures have no violations. Ramachandran analysis determined that 96.8% ± 2.0% of the residues are within favored regions. Details regarding the statistics of the ensemble of structures can also be found in Table 1.

For MLD_RRSP, 1256 NOE peaks were used for the structure calculation where 114 long‐range peaks defined the interhelical contacts. The interhelical contacts were defined by the following NOE restraints: V7 to L29 for helix 1 and 2; V35 to A46 for helix 2 and 3; S47 to V77 for helix 3 and 4; L10 to L73 for helix 1 to 4. The interhelical contacts are depicted on a representative structure [Fig. 4(B)].

Comparison of the solution structures of MLD_PMT and MLD_RRSP to the PMT crystal structure

Two representative solution NMR structures, with the lowest RMSD with respect to the bundle of 20 structures [Fig. 5(A,B)], were superimposed [Fig. 5(C)]. Comparison of these two MLDs revealed that the orientation of the helices is mostly preserved between the two structures with a backbone RMSD for the helix residues of 1.25 Å.

Figure 5.

Comparison of the two solution structures of the MLD_PMT and MLD_RRSP with each other. Shown are representative solution structures of (A) MLD_PMT alone in red, (B) MLD_RRSP alone in cyan, (C) superimposed with each other with all four helices aligned, and (D) superimposed with each other with helix 3 and 4 aligned.

Helices 3 and 4 can be more closely aligned [Fig. 5(D)], where the backbone RMSD of helix 3 and 4 is 0.84 Å. The differences were primarily observed within the loops and end residues for helices 1 and 2. PMT contains a shorter helix 1 and a longer loop 1 as well as a shorter loop 2, which may account for the differences in the overall position of helix 1 and 2 in the bundle.

The C1 domain of PMT contains the 4‐helix‐bundle motif [Fig. 1(B)], as was previously determined by X‐ray crystallography with a 1.9‐Å resolution (PDB ID: 2EBF).4 Superposition of the representative solution structure of MLD_PMT [Fig. 5(A)] with the crystal structure [Fig. 1(B)] showed that helices 3 and 4 are structurally similar [Fig. 6(A)], with a backbone RMSD of 0.75 Å. Helix 1 is in a very different orientation while helix 2 has a slightly different orientation, which may be the result of decreased mobility in the crystal as compared to solution.

Figure 6.

Comparison of the representative solution structures of MLD_PMT and MLD_RRSP with the corresponding region in the crystal structure C1 domain of PMT. Shown are (A) the superposition of solution MLD_PMT (red) and the crystal structure (blue) and (B) the superposition of solution MLD_RRSP (cyan) and the crystal structure. Helices 3 and 4 were aligned in each superposition.

Superposition of MLD_RRSP [Fig. 5(B)] with the PMT crystal structure likewise showed that the structures are similar, with a backbone RMSD for helices 3 and 4 of 0.43 Å [Fig. 6(B)]. The greatest difference in structure is found in helix 1, where the helix of RRSP is slightly longer than the helix of PMT.

Discussion

Despite the great differences in the origins and functions of PMT and RRSP, both possess a structurally homologous MLD. Both MLDs are 4‐helix‐bundles in solution, confirming the previously predicted secondary structures based on TALOS+ chemical shift calculations.16, 18 These results agree with the corresponding MLDs found in the crystal structures of several clostridial toxins.8, 9, 10, 11 Although there is precedent of the MLD_RRSP alone forming a non‐4‐helix‐bundle dimeric structure (PDB ID: 4ERR), we confirm here that MLD_RRSP is a 4‐helix‐bundle in solution at ambient temperature.

The structures of MLD_PMT and MLD_RRSP are in overall agreement with each other. Differences are confined to the orientation of the mobile loops and ends, as well as slight differences in the lengths of the helices. In MLD_RRSP, helix 3 is slightly longer and loop 2 is slightly shorter than that of MLD_PMT. Furthermore, as indicated by the predicted order parameters (Fig. 3), loop 2 of MLD_RRSP is more rigid than loop 2 of MLD_PMT. MLD_PMT in solution forms a 4‐helix‐bundle similar to that observed in the crystal structure of the C1 domain of full‐length PMT (PDB 2EBF). The lengths and orientations of the helices 2, 3, and 4 are comparable between the structures. However, the orientation of helix 1 and loop 1 exhibit some differences, presumably due to the higher mobility of loop 1 in solution, which could impact the position of helix 1. As mentioned earlier, the loops are mobile and loop 1 in particular is more flexible because it is the longest unrestrained part of the sequence. Likewise, in comparing MLD_RRSP [Fig. 4(B)] to the MLD motif in the PMT crystal structure, the structures are fairly similar, except with regard to the length of helix 1, where helix 1 of MLD_RRSP in solution is longer than helix 1 in the crystal structure of MLD_PMT. The similarity in structure of MLD_RRSP to MLD_PMT reaffirms the homologous relationship between RRSP and PMT, and by analogy with MLDs of other clostridial glycosylating toxins.

Formation of the 4‐helix‐bundle in solution in the absence of any other domains indicates that the 4‐helix‐bundle conformation is preferred and may be important for localization of the catalytic activity domain to the membrane target site. One possibility is that the 4‐helix‐bundle is required for localizing to the correct membrane associated target,15 presumably through recognition by an as‐yet‐unidentified receptor in the membrane. Alternatively, membrane localization may involve a conformational change of the MLD. In solution, the structures of MLD_RRSP and MLD_PMT correlate with all other MLD crystal structures previously determined for the full‐length toxins. However, the crystal structure of MLD_RRSP (PDB ID: 4ERR) is significantly different. In this crystal structure, the protein appears to form an antiparallel dimer, where the globular 4‐helix‐bundle opens up into a planar structure and the nonpolar faces of helices 1 and 4 of one molecule are interacting with the nonpolar faces of helices 2 and 3 of the other molecule.

This crystal structure may indicate a dimerization process that occurs when binding to the membrane, so that the 4‐helix‐bundle opens into a planar structure exposing the nonpolar faces of the protein and facilitating the nonpolar faces of the protein to embed into the hydrophobic section of the membrane. However, this may also be an artifactual crystal structure of a domain swapped19 structure of MLD_RRSP. Without further structural data of the membrane‐bound form of MLD, we are unable to confirm the extended conformation of this putative intermediate, unfolded structure of the MLD. Having now solved the structure of the MLD in solution, we are in a position to study the structure of the membrane‐bound form and determine whether MLDs undergo large conformational changes upon localizing to and interacting with target membranes.

Materials and Methods

Expression and purification of MLD_PMT and MLD_RRSP

Uniformly 13C,15N‐labeled recombinant MLD_PMT and MLD_RRSP proteins were expressed and purified from E. coli BL21(DE3) cells (Promega, Fitchburg, WI) grown in medium containing 13C,15N‐BioExpress (Cambridge Isotopes Laboratories, Andover, MA), U‐¹⁵N NH₄Cl, and U‐¹³C glucose, as previously described.16

NMR spectroscopy

Solution NMR spectra were collected on a Varian INOVA (600 MHz ¹H) spectrometer with a 5‐mm triple resonance (¹H–13C–15N) triaxial gradient probe using VnmrJ 2.3 with the BioPack suite (School of Chemical Sciences NMR Facility at the University of Illinois at Urbana‐Champaign).

Spectra of solution state MLD_PMT were collected at 30ºC with a 1 mM protein sample in 20 mM bis‐Tris buffer, pH 6.0, containing 100 mM NaCl, 10% D₂O (v/v) and 0.01% 4,4‐dimethyl‐4‐silapentane‐1‐sulfonic acid (DSS), as previously described,16 or on a 99% D₂O back‐exchanged sample (1 mM in 20 mM bis‐Tris buffer, pH 6.0, containing 100 mM NaCl, 99% D₂O, and 0.01% DSS). For MLD_RRSP, data was also collected at 30ºC with a 1 mM sample in 20 mM Tris buffer, pH 7.4, containing 500 mM NaCl, 10% D₂O, 0.01% DSS, and 2 mM EDTA, as described previously,18 or on a 99% D₂O back‐exchanged sample (1 mM sample in 20 mM Tris, pH 7.4, containing 500 mM NaCl, 99% D₂O, 0.01% DSS, and 2 mM EDTA). Distance measurements for both proteins were made using 13C‐HSQC‐NOESY spectra with a mixing time of 150 ms and 15N‐HSQC‐NOESY spectra with a mixing time of 150 ms. Spectra were processed with NMRPipe20 and analyzed in Sparky.21

Distance restraints and structure calculation

A total of 200 structures were calculated in XPLOR‐NIH using the NOE distance restraints and dihedral angles (Table 1),22, 23 and the top 20 structures are presented here. The structures of MLD_PMT were calculated with 1,485 NOE restraints and 75 dihedral angles [Fig. 3(A)], while those of MLD_RRSP were calculated with 1,256 NOE restraints and 73 dihedral angles [Fig. 3(B)]. NOE distance restraints were determined using the PASD algorithm.24, 25 NOESY peaks (data height and peak position) from carbon and nitrogen edited 3D NOESY experiments were imported into the PASD algorithm, where the peaks were categorized into standard distance bins: strong (80%+ intensity): 1.8 Å to 2.7 Å; medium (50% to 80% intensity): 1.8 Å to 3.3 Å; weak (20% to 50% intensity): 1.8 Å to 5.0 Å; very weak (0% to 20% intensity): 1.8 Å to 6.0 Å. Furthermore, the algorithm increases the upper bound of the methyl peak distance ranges by 0.5 Å to compensate for the larger intensities of methyl peaks. Dihedral angles and order parameters were determined using TALOS‐N.17 The chemical shifts used for these calculations were previously published.16, 18

RMSD calculations were made using VMD software for both backbone and heavy atoms for the top 20 structures.26 For comparison of any pair of structures, the RMSD calculations were based on global alignments for the region of interest. Ramachandran analysis was calculated using Molprobity software through the PSVS server to verify angular geometry.27, 28

Data deposition

The ensemble of structures of both proteins was deposited into the Protein Data Bank. The PDB ID for MLD_PMT is 2N9V and the PDB ID for MLD_RRSP is 2N9W.