Interfacial residues of SpcS chaperone affects binding of effector toxin ExoT in Pseudomonas aeruginosa: novel insights from structural and computational studies


  • Supratim Dey,

    1. Department of Structural Biology and Bioinformatics, Indian Institute of Chemical Biology, Kolkata, India
    Search for more papers by this author
  • Saumen Datta

    Corresponding author
    1. Department of Structural Biology and Bioinformatics, Indian Institute of Chemical Biology, Kolkata, India
    • Correspondence

      S. Datta, Structural Biology and Bioinformatics Division, Indian Institute of Chemical Biology (IICB), 4 Raja S.C. Mullick Road, Kolkata-700032, India

      Fax: +91(33)24735197, +91(33)24723967

      Tel: +91(33)24995896


    Search for more papers by this author


ExoT belongs to the family of type 3 secretion system (T3SS) effector toxins in Pseudomonas aeruginosa, known to be one of the major virulence determinant toxins that cause chronic and acute infections in immuno-compromised individuals, burn victims and cystic fibrosis patients. Here, we report the X-ray crystal structure of the amino terminal fragment of effector toxin ExoT, in complex with full-length homodimeric chaperone SpcS at 2.1 Å resolution. The full-length dimeric chaperone SpcS has the conserved α-β-β-β-α-β-β-α fold of class I chaperones, the characteristic hydrophobic patches for binding effector proteins and a conserved polar cavity at the dimeric interface. The stable crystallized amino terminal fragment of ExoT consists of a chaperone binding domain and a membrane localization domain that wraps around the dimeric chaperone. Site-directed mutagenesis experiments and a molecular dynamics study complement each other in revealing Asn65, Phe67 and Trp88 as critical dimeric interfacial residues that can strongly influence the effector–chaperone interactions.


The atomic coordinates and structure factors of ExoT–SpcS complex (code 4JMF) have been deposited in the Protein Data Bank Japan (PDBj), Institute for Protein Research, Osaka University.




chaperone binding domain


dynamic cross-correlation matrix




membrane binding domain


molecular dynamics


root mean square fluctuation


response unit


surface plasmon resonance


type III secretion system


Pseudomonas aeruginosa is increasingly recognized as an emerging opportunistic pathogen as it harbors multiple virulence factors that widely manipulate host cell signaling and immune response [1-3], accounting for more than 10.1% of all hospital-acquired infections [4]. Like many other Gram-negative pathogenic bacteria, most clinical isolates of Pseudomonas sp. carry a nanosyringe-like membrane anchored, proteinaceous multicomponent complex called type III secretion system (T3SS), for translocation of the bacterial toxins into eukaryotic host cells. This complex regulon consists of proteins forming injectisome (inner and outer membrane spanning proteins, a proteinaceous hollow needle and translocon), chaperones, regulatory proteins and the most virulent effector proteins [5-7]. Amongst all these proteins, the effector proteins in conjunction with their respective chaperones play a key role in the establishment and maintenance of bacterial pathogenicity and are probably the interest of study of several research groups.

The effector toxin proteins of T3SS present in P. aeruginosa are ExoT, ExoS, ExoU and ExoY [7, 8]. While ExoU and ExoY are involved in host cell membrane disruption [9, 10], both ExoT and ExoS are known as the actual virulence determinants due to the presence of bifunctional GTPase-activating (GAP) and ADP-ribosyltransferase (ADPRT) domains, essential for inhibition of bacterial internalization and epithelial cell migration by altering the actin cytoskeleton [11-14]. ExoT is a 457-residue protein and shares common domain architecture and similar functions with its homolog ExoS. The N-terminal 1–50 residues of ExoT constitute the chaperone binding domain and provide secretion signal to flush the toxin through the T3SS apparatus into the host cell. Due to its high sequence (76%) similarity with ExoS, residues 51–72 of ExoT can be presumed to constitute the membrane localization domain while residues 78–235 and 235–457 respectively form the GAP domain and ADPRT domain [15]. Multiple lines of experimental evidence from different bacterial systems [16-22] suggest that the effector proteins become secretion competent only after they bind their respective class I chaperones. The chaperone SpcS (formerly known as Orf1) binds to both effector proteins – ExoT and its homolog, ExoS [23].

In order to understand the role of chaperones in the protection and release of effector toxin ExoT to cause host cell infections, we have solved the first crystal structure of ExoT–SpcS complex at 2.1 Å resolution. The crystal structure consisting of a stable N-terminal fragment of ExoT in complex with its cognate chaperone SpcS shows the chaperone binding domain and membrane localization domain of effector ExoT tethered around the SpcS homodimer in a 1 : 2 molar ratio. In order to investigate, in greater detail, the effector–chaperone (ExoT–SpcS) interaction mechanism, we constructed in silico and in vitro SpcS mutants and performed molecular dynamics (MD) simulations corroborated by binding affinity studies on the wild-type (WT) and mutants. The results reveal critical chaperone dimerization contacts, which drastically alter the affinity of ExoT binding to its cognate chaperone, SpcS. We finally discuss the structural and functional conservation of ExoT and SpcS amongst all T3SS effectors and their cognate class I chaperones.


Overall structure of ExoT–SpcS complex

Formation of ExoT–SpcS crystals

The effector protein ExoT when expressed recombinantly in Escherichia coli BL-21(DE3) yields a small amount of soluble protein. To increase solubility and yield of recombinant ExoT, the effector protein was coexpressed and co-purified in the presence of its chaperone SpcS, forming the ExoT–SpcS complex. The complex was next purified by gel filtration and then we proceeded with crystallization trials. After 8 months of crystallization set-up, hexagonal bipyramidal shaped crystals of size 0.2–0.3 μm appeared with sharply defined edges (Fig. S1, inset). These crystals diffracted up to 2.1 Å resolution and were crystallized in the P6122 space group. In the absence of a structural orthologue of ExoT, the molecular isomorphous replacement technique was chosen to solve the structure. Crystals were soaked in salts of heavy metals such as mercury, gold and platinum; a few of them were found to be potential derivatives (Table 1) and used for phasing to get an electron density map. The initial electron density map obtained was of good quality and an alanine model could be built automatically in the entire asymmetric unit. The refined ExoT–SpcS complex model revealed an SpcS homodimer (residues 1–116) and an amino terminal fragment of ExoT (residues 28–77, total of 50 residues could be traced), consistent with the mass spectrum data of ExoT–SpcS crystal (Fig. S1) where one peak ~ 13 kDa corresponds to the SpcS monomer while the other dominant peak ~ 6.6 kDa corresponds to the ExoT fragment (6–7 kDa). This implies that the crystallization incubation time of 8 months results in time-dependent degradation of ExoT to finally yield a stable chaperone binding domain (CBD) (residues 28–50) and membrane binding domain (MBD) (residues 51–77) (annotated ∆ExoT28–77), complexed to its cognate chaperone SpcS. The fragment ∆ExoT28–77 is designated as chain A, while the SpcS monomers tethered by MBD and CBD domains of ExoT are designated as chain B and chain C, respectively.

Table 1. X-ray data collection and refinement statistics.
Data collection statistics
 Native ExoT– (SpcS)21 mm KAu(CN)2 ExoT– (SpcS)22 mm K4PtCl6 ExoT– (SpcS)2500 mm HgCl2 ExoT– (SpcS)2600 μm Hg(OAc)2 ExoT– (SpcS)2
Wavelength (Å)1.54181.54181.54181.54181.5418
Space groupP6122P6122P6122P6122P6122
Unit cell parameter (Å)= 78.3= 78.6= 78.8= 78.5= 78.5
= 194.2= 194.1= 194.4= 194.7= 195.4
Resolution (Å)a50–2.1 (2.15–2.10)50–2.07 (2.15–2.07)50–2.3 (2.38–2.3)50–2.2 (2.26–2.2)50–2.5 (2.59–2.5)
Total reflection500 479509 089333 002470 37684 184
Unique reflection21 31120 26316 67018 17911 928
Completeness (%)a98.9 (88.0)81.6 (48.9)99.0 (99.0)84.1 (51.1)91.5 (91.2)
Average I32.6 (1.5)32.8 (1.7)20.8 (1.05)27.3 (1.8)12.5 (1.3)
bRmerge (%)a10.8 (76.6)12.9 (91.7)15.8 (n.a.)15.5 (n.a.)15.3 (n.a.)
  1. a

    The value in parenthesis relates to the highest resolution shell.

  2. b

    a.u., asymmetric unit; n.a., not applicable. Rmerge = Σ(IobsImean)/ΣImean, where Iobs is observed intensity and Imean is the average intensity from multiple observations of symmetry-related reflections after rejections.

Refinement statistics
Resolution (Å)30.5–2.1
Mean B factor or temperature factor (Å2)34.1
Number of atoms
Ligand (glycerol)6
Root mean square deviation
Bond length (Å)0.007
Bond angle (°)1.2
Ramachandran analysis (%)
Most favored94.93

Structure of chaperone SpcS

The dimeric chaperone SpcS belongs to the family of T3SS class I chaperones and shares a common α1-β1-β2-β3-α2-β4-β5-α3 fold topology [6] as shown in Fig. 1A,B. Each SpcS monomer consists of five-stranded antiparallel twisted β sheets, flanked by three α helices. Both monomers possess identical structures (backbone Cα RMSD ~ 0.6 Å), with minor deviations in loop regions (L1 and L4) and dimerization helix (α2) (Fig. S2b). The monomers combine together to form a compact globular dimer with a buried interfacial surface area of 2499.7 Å2. The stretch of alternating hydrophobic and polar residues from His42 to Leu95 of both monomers primarily constitutes the dimeric interface. Structurally, this involves the centrally positioned α2 helix and loop L5 of one monomer stacked with the β3 and β5 strands of the other monomer. A prominent hydrophobic pocket is thus formed at the dimeric interface where loop L5 (Leu66, Phe67) of one monomer intercalates into the strands β3 (Leu48, Phe50) and β5 (Leu86, Trp88) of the other monomer, as shown in Fig. 2. The other hydrophobic contributions to this dimeric association include Ile75 (β4) and Pro92, Leu95 (310 helix). SpcS also possesses a polar cavity of volume 685 Å3 at the dimerization interface, predominantly occupied by polar/charged residues like His46, Gln64, Asn65, Ser68, Asp70, Lys73, Asn89, Arg90 and Gln91 and the entrance to this cavity is guarded by residues Pro92 and Leu95 of 310 helix, as shown in Fig. S2a. In addition, both monomers show extensive inter-hydrogen bonding, as summarized in Table S1. Amongst all these interactions, the most noteworthy is the residue Asn65 that forms a side chain–side chain interaction with Asn65 of the other monomer and is also hydrogen bonded to the main chain carbonyl atoms of Ala59, Thr60 and Trp88. Trp88 is a yet another vital residue for its dual role in mediating hydrophobic interactions as well as maintaining main chain–side chain hydrogen bonding within the monomers.

Figure 1.

Overall structure of P. aeruginosa SpcS chaperone dimer. (A) Ribbon drawing of SpcS dimer, consisting of three α helices and five β antiparallel twisted strands. The chaperone monomers are colored in red and blue. (B) Topology diagram of SpcS where helices, strands and loops are schematically represented by α, β and L, respectively.

Figure 2.

Interactions between chaperone monomers. The β5 residues (Leu86, Trp88) and β3 residues (Leu48, Phe50) from one monomer (yellow) are intercalated by hydrophobic L5 residues (Leu66 and Phe67) of the other monomer (green) to form a hydrophobic pocket. A hydrogen bond is formed between Asn65 (O) and Trp88 (NE1) in both monomers. It is interesting to note that both hydrophobic and hydrogen bonding interactions are mediated by Trp88. The interaction site is boxed (inset) and the important residues are shown as sticks.

Structure of effector ΔExoT28–77

Unlike the rigid globular folds of chaperone SpcS, the crystallized fragment ΔExoT28–77 exists as an extended polypeptide with well-defined secondary structure wrapped around the dimeric chaperone by its CBD and MBD, as shown in Fig. 3A,B. The CBD is constituted by L1, L2 and L3 loops, β1 strand and α1 helix, while MBD is essentially composed of helices α2 and α3 and β2 strand along with loops L4 and L5. An interesting feature of ΔExoT28–77 is the spatial arrangement of the amphipathic MBD (Fig. S2c), where the hydrophobic Leu-rich motif interacts with the chaperone SpcS and the charged residues face the extracellular environment free from any kind of interaction. The importance of this amphipathic nature of the MBD has already been illustrated in ExoS, where the Leu-rich motif associated with the plasma membrane and the free charged residues enhanced RhoGAP activity to interact with host cell proteins Cdc42 and Rac1 [24].

Figure 3.

Structural overview of ExoT–SpcS complex: (A) frontal view and (B) side view of the chaperone binding domain of ExoT (green) wrapped around the SpcS dimer. This CBD domain essentially consists of three α helices and two β sheets. (C) Electrostatic potential mapped onto the molecular surface of SpcS dimer wrapped around by ExoT. A 180° rotation around the vertical axis reproduces (D). Few positively and negatively charged patches are found on the surface of the chaperone. The red electronegative patch is formed by residues Glu54, Asp58, Asp79, Glu63 and Glu103 while residues Lys73 and Arg90 form the central blue patch inside the polar cavity. The chaperone also harbors a polar cavity at the dimeric interface constituted by polar and charged residues. The neutral (white) areas are made up of hydrophobic residues. Two hydrophobic patches labeled 1 and 2 can be identified consisting of Phe13, Leu16, Leu18, Val27, Leu28, Leu30 and Val32 in patch 1 and Leu36, Ala40, Leu47, Leu48, Met49, Phe50, Leu66, Phe67, Leu86 and Trp88 forming patch 2. The hydrophobic patches patch 1 (A, C) and patch 2 (B, D) come in pairs due to the dimeric nature of the chaperone. The ExoT–SpcS interaction is primarily hydrophobic driven as reflected in the hydrophobic path (patches A, B, C and D) that ExoT follows to wrap around the chaperone SpcS.

ExoT–SpcS complex

The complexation of ExoT to SpcS monomers results in burial of a large surface area (3969.2 Å2) at the interface, with a slight difference in individual complexation, the MBD subunit (1399 Å2) being slightly larger than the CBD subunit (862.8 Å2). Figure 3C,D shows the electrostatic potential surface representation of SpcS dimer complexed with ExoT. The path of ExoT around the dimeric chaperone is characterized by a set of four hydrophobic paired-patches (A/B and C/D), curved in, on the chaperone electrostatic potential surface. Each monomer consists of two shallow hydrophobic grooves called patch 1 (A/C) and patch 2 (B/D). Patch 1 is made up of α1, β1 and L1 where the β1 strand residues (Leu16, Val18, Leu28, Leu30, Val32) of SpcS run antiparallel and form hydrophobic interactions with β1 (Ala29, Val32) and β2 (Ala59, Ala60, Leu61) strands of ExoT, as shown in Fig. 4A. On the other hand, hydrophobic residues from twisted antiparallel β strands – β2, β3 and β5 of one monomer, loop L5 of the other monomer – interdigitate with side chains of ExoT α1 and α3 helices forming a ‘helix-binding groove’, also known as patch 2 (B/D). This hydrophobic interaction is mediated by the ExoT residues α1 (Ala38, Ala41, Ala42) and α3 (Ile68, Ile69, Leu72, Leu75, Leu76) along with SpcS residues (Leu36, Ala40, Leu48, Phe50, Leu66), as shown in Fig. 4B. In addition, both effector and chaperone show extensive inter-hydrogen bonding, as summarized in Table S2.

Figure 4.

Insights into the effector–chaperone interactions. (A) The β1 strand of ExoT (green) runs antiparallel to the β1 strand of SpcS (blue) in the patch 1 domain (α1, β1 and L1). Hydrogen bonds are formed between Arg30 (N/O) of ExoT with Gln31 (N/O) of SpcS; Gln31 (NE2) of ExoT with Glu25 (OE2) of SpcS; Thr34 (OG1) of ExoT with Ser29 (N) of SpcS. (B) The alternating hydrophobic residues of the α3 helix of ExoT (red) stacks into the hydrophobic groove of the patch 2 domain of SpcS.

In silico alanine scanning mutagenesis and MD simulations

The structural study reveals a complex network of interactions existing between the chaperone–chaperone dimers and effector–chaperone. To identify the key residues responsible for chaperone dimerization and formation of the effector–chaperone complex, we performed in silico alanine scanning that captures the effect of different energy components (like hydrophobic, hydrogen bonding, salt bridges and electrostatics) by using a simple free energy function and calculated the effect of alanine mutations on the binding free energy of the protein complex. Using this analogy, we identified three alanine replacements (Leu66, Phe67 and Trp88) predicted to destabilize the dimeric chaperone interface and six replacements (Arg44, Leu54, Leu76, Thr34 in ExoT and Ser29, Gln69 in SpcS) predicted to destabilize the effector–chaperone interface by at least 1 kcal·mol−1 (Fig. S3a,b,c; Table S3).

However, in silico alanine scanning is a static representation of the connectivity within chaperone dimers and the effector–chaperone complex. In order to investigate the real-time dynamics of effector and chaperone, we employed 10 ns MD simulation of ∆ExoT28–77–SpcS complex that effectively captures coupled motions of distal domains as exemplified by the dynamic cross-correlation matrix (DCCM) [25]. The DCCM of WT ExoT∆28–77–SpcS complex is shown in Fig. 5, where the intra-effector (residues 28–77) and intra-chaperonic (chain B and chain C) correlations/anti-correlations along the diagonal are respectively marked by large black, red and blue boxes. The strong positive correlated fluctuations (black box) between β3 (residues 45–51) and β5 (residues 87–92) of the monomers corroborate the structural studies with respect to their spatial proximity. The dotted boxes correspond to strong inter-monomeric communication between strands β3, β5 and loop L5. Consistent with the structure studies, dynamical couplings (marked with red boxes) between effector and chaperone were also seen: β1 strand residues (chain B: 30–32 and chain C: 29–32) of SpcS are strongly correlated to strand residues (β2: 59–61 and β1: 29–32) of ExoT, respectively. A more robust direct coupling is noted between the effector helix α1 (38–42) and L5 loop (65–69) of the chaperone (chain B). A similar but moderate positive correlation is found between effector α1 helix (38–42) with β3 (46–51) and β5 (86–88) strand residues of the chaperone (chain C). The DCCM studies thus highlight the importance of chaperonic residues Asn65, Phe67 and Trp88 that were predicted to be destabilizing from alanine scanning mutagenesis in mediating coupled effector–chaperone and chaperone–chaperone interactions.

Figure 5.

DCCMs represent the collective atom fluctuation for all residue pairs within and between the effector and dimeric chaperone. Motion occurring along the same direction corresponds to positive correlation (blue), whereas anti-correlated motion occurring along the opposite direction corresponds to negative correlation (yellow). The correlated residues enclosed in rectangular boxes (red, black and dotted) are the most critical ones, the details of which are mentioned in the text.

To further assess the contribution of these ‘critical’ residues (residues Asn65 and Phe67), we generated two double mutants SpcSF67A/W88A and SpcSN65A/W88A and performed MD simulations (10 ns) along with the WT chaperone complex. The root mean square deviations (RMSDs) of backbone atoms of WT SpcSN65A/W88A and SpcSF67A/W88A complex (Fig. S4a) with respect to their initial energy minimized structure are within 3 Å (2.5 Å, 2.9 Å and 2.8 Å, respectively), indicating not much structural change at the end of the simulation. The superimposed average structures (Fig. S4b,c) of the three complexes at the end of the simulation showed the helices α1, α2, loops L1, L4 and strands β1, β3, β5 to display the sites of maximum variance. Loop L5 and helix α2 known to play a key role in dimeric interactions also showed greater extendibility in mutants than in the WT complex. However, the variations are more evident from per residue backbone fluctuations (RMSF) (Fig. S5) exhibiting a slightly different dynamic behavior of mutants from WT. Subtle variations were also seen at the strand β5 (85–92), loop L5 (66–74) and helix α2 (62–65) implicated in chaperone dimerization, as well as regions close to mutation sites like loop L4 (54–61) and L6 (78–84), but the largest fluctuations were at loop L1 (residues 17–27) of patch 1 domain, quite distal to the site of the point mutations. This is quite remarkable because patch 1 domain is the primary interaction site of effector with the chaperone, implying that the mutation of critical interface chaperonic residues greatly affects the effector binding.

Kinetic and structural characterization of WT SpcS and mutants by surface plasmon resonance and CD

The MD studies were experimentally verified by comparing the binding affinity of the effector ExoT for WT and mutant chaperones (SpcSN65A, SpcSF67A, SpcSW88A, SpcSN65A/W88A and SpcSF67A/W88A) by surface plasmon resonance (SPR). The dissociation constants (KD) of the effector–chaperone complexes as calculated from the ratio of dissociation rates over association rates (kd/ka) were found to be in the micromolar range (Table 2; Fig. S6). The KD value of the WT chaperone with the immobilized effector was measured to be 0.30 μm. Single mutants (SpcSN65A, SpcSF67A and SpcSW88A) mildly affected the effector–chaperone interaction as reflected from a 20- to 30-fold decrease in binding affinity. However, drastic effects were noted for the double mutants SpcSN65A/W88A and SpcSF67A/W88A where the binding affinity of the effector was 100-fold reduced (44.8 μm and 34.8 μm, respectively), supporting the MD simulations. The association rate (ka) for SpcSN65A/W88A was about 30 times lower than that of WT, while those of SpcSF67A/W88A and single mutants were comparatively less affected. In contrast, marginal differences were seen in the dissociation rates (kd) of SpcSN65A/W88A (4-fold increase) and SpcSF67A/W88A (10-fold increase) with respect to the WT. The relatively slow association/dissociation rate of SpcSN65A/W88A compared with SpcSF67A/W88A implies that mutation SpcSN65A/W88A is more critical for docking of chaperone onto the effector.

Table 2. Kinetic parameters of the binding of SpcS and its mutants to ExoT.
Proteinka × 105 (m−1·s−1)k× 10−2 (s−1)KD × 10−6 (m)
SpcSWT2.86 ± 0.088.522 ± 0.060.298 ± 0.06
SpcSN65A0.89 ± 0.0261.94 ± 0.536.96 ± 0.98
SpcSF67A1.09 ± 0.0284.04 ± 0.677.64 ± 0.86
SpcSW88A1.08 ± 0.05103.24 ± 1.239.56 ± 0.75
SpcSN65A/W88A0.08 ± 0.0135.84 ± 0.46044.8 ± 1.25
SpcSF67A/W88A0.32 ± 0.002111.23 ± 1.5434.76 ± 1.68

Further structural evaluation of the mutants was done by far and near UV CD and compared with WT to evaluate any changes in their 3D structural conformations. Analysis of far UV CD spectra and secondary structure compositions at 24 °C predicted from the k2d3 server (WT: 40% α helix, 22% β sheets; SpcSN65A/W88A: 34% α helix, 19% β sheets; SpcSF67A/W88A: 36% α helix, 20% β sheets) suggests similar secondary structural contents and folded conformations of the mutants to those of the WT protein. The WT protein and mutants also exhibited comparable thermal stability as was evident from the minor changes in secondary structure in response to temperature variations, as shown in Fig. S7a–d. The near UV CD spectra of the WT protein and mutants, however, showed a strong tertiary signal of the former as is evident from Fig. S7e. This near UV wavelength range is indicative of aromatic residues that are in an asymmetric environment due to the folded structure of the proteins, which has been substituted by a non-aromatic alanine residue in the case of the mutants.

Thus, based upon in silico Ala scanning mutagenesis studies and MD simulations, our hypothesis of a strong intercommunication network that allows interface residues of dimeric chaperone to regulate the effector–chaperone interaction has been convincingly demonstrated by experimental binding studies of the SpcS mutants SpcSN65A/W88A and SpcSF67A/W88A.


Although the mechanisms by which T3SS of Paeruginosa cause opportunistic infections are a matter of wide biological interest, special interest has been assigned to the effector toxins that impair the innate immune response of the host cell and lead to bacterial dissemination and sepsis. To date, there is only one crystal structure of the effector toxin, ExoU complexed to chaperone SpcU [26, 27]. Here we report the crystal structure of the effector protein ExoT in complex with its cognate chaperone SpcS from Pseudomonas, yet another example of a T3SS class I chaperone–effector complex. The crystallized N-terminal fragment of ExoT (ΔExoT28–77) comprising CBD and MBD is found to wrap around the dimeric SpcS chaperone in a manner similar to the interaction of its homologue YopE–SycE complex [28] (Fig. S8a,b).

Structure-based sequence alignment of T3SS chaperones (Fig. 6) reveals strong structural conservation, reminiscent of the fact that T3SS chaperones evolved from a common ancestral protein. All the T3SS class I chaperones [29-37] possess structural similarities with respect to patch (1/2) domains and an interfacial polar cavity lined by hydrophobic residues, with subtle differences in the volume of the polar cavity, the buried surface area and the presence/absence of α2 helix and loops L4 and L5 at the dimeric interface. The T3SS effector molecules, unlike their chaperone counterparts, have very less sequence and structural identity (data not shown). However, the β strands (β1 for ExoT) of T3SS effector proteins [26-28, 31-34, 38-40] show high structural and sequence conservation with respect to their mode of interaction with the hydrophobic crevice (patch 1) made of chaperone β strands. The antiparallel orientation of the effector β strand (also known as canonical β motif) and the chaperone β1 strand is universally conserved and any point mutation at this site results in hindered complex formation with reduced effector secretion and translocation through the T3SS apparatus [34]. Earlier studies [41] also report a decreased virulence of Pseudomonas upon mutation/deletion of the MBD domain, citing the importance of the amphipathic nature of the MBD in maintaining hydrophobic contacts with SpcS. Structural studies of the MBD of ExoT, YopE, YscM2, YopH, YopN and SptP effectors reveal yet another common mode of binding with the helix-binding groove (patch 2) of chaperones, but no sequence conservation was seen to exist (data not shown). This capping action of the chaperone is apparently important because it allows the effector to remain soluble within the bacterial cytosol and prevents its localization to the bacterial membrane prior to translocation.

Figure 6.

Structure-based sequence alignment of class I chaperones from dali server. The blue boxed residues are conserved hydrophobic residues, while those in red boxes are conserved charged residues. Residues involved in dimeric interaction are enclosed within dotted brackets. The chaperone residues (Leu16, Leu18, Val27, Leu30 and Val32) involved in effector interaction are highly conserved. The chaperonic charged Glu106 residue (marked by red circle), primarily involved in hydrogen bonding with the effector molecule, also exhibits strong conservation.

In the present study we also determined the energetic contributions that drive the effector–chaperone complex and predict the kinetic changes brought about by mutation of key residues. In silico Ala scanning mutagenesis experiments reveal some ‘hot spot’ residues of SpcS and ExoT that are crucial for maintaining stability of the complex. Amongst them, the hydrophobic residues Phe67 and Trp88 and the charged Asn65 residue present at the chaperone dimer interface are shown to be vital from the structural and energetics viewpoint. Both computational and biophysical techniques were employed to study the dynamics and kinetics of WT protein and double mutants SpcSF67A/W88A and SpcSN65A/W88A in the presence and absence of ExoT. Both the mutations significantly reduced the affinity of ExoT for SpcS; the N65A/W88A mutation had a greater effect than the F67A/W88A mutation. While the effect of the F67A/W88A mutation is a mere case of disrupted bonding network with the effector due to their spatial proximity, the perturbation brought about by the N65A/W88A mutation is far more distal yet strongly affects the effector–chaperone coupling, which is manifested on a macromolecular level by a loss of complex stability, as judged by an increase in observed dissociation rates following mutation. The evolutionary conservation scores (Table S3) as calculated by the consurf webserver based on phylogenetic relationships between closely related T3SS class I chaperones also reveals a high conservation score for Asn65 along with other hot-spot residues like His42, Phe50, Arg90 and Glu106. This work thus provides an illustrative example of a strong inter-residue communication network and identifies critical residues at the dimeric interface that can destabilize the effector–chaperone complex.

To summarize, our crystal structure of the ExoT–SpcS complex emphasizes the importance of chaperone SpcS in maintaining ExoT in a secretion competent state within the cytoplasm of P. aeruginosa. The structure provides important insights into the modes of intra-chaperonic interactions between the monomers and strong non-covalent interactions with the effector molecule. Deletion experiments have shown the spcS gene to be important for the secretion of both ExoT and ExoS effectors [23]. This multifaceted and highly efficient function of chaperone SpcS as a ‘multicarrier protein’ makes it a potential drug target and calls for large-scale successful clinical therapeutics. Besides, by designing cell-permeable small molecule inhibitors to neutralize ExoT, one of the greatest virulence determinants of P. aeruginosa, can help reduce cytotoxicity of the pathogen in immuno-compromised individuals. The present study also highlights the importance of using a multidisciplinary approach encompassing computational and biophysical techniques to provide valuable information about a biological system compared with traditional methods.

Experimental procedures

Cloning of exoT and spcS

The gene exoT from P. aeruginosa PA01 2192 was cloned in pET28a+ vector using restriction enzyme NdeI/XhoI. This resulted in the formation of pET28a+-exoT having His6 tag at the N-terminus of protein ExoT (His6-ExoT). To generate a bicistronic construct for coexpression of both exoT and spcS genes, exoT along with thrombin cut site was excised from pET28a+-exoT by XbaI/HindIII enzyme and subcloned into MCS-1 of pETDuet-spcS containing spcS in the MCS-2 region (NdeI/XhoI) of the same vector. The point mutations N65A, F67A, W88A, N65A/W88A and F67A/W88A were introduced in spcS using PCR-based site-directed mutagenesis. The WT spcS gene and its point mutants were cloned into pET28a+ using NdeI/XhoI restriction enzyme resulting in the formation of pET28a+-spcS having His6 tag at the N-terminus of protein SpcS (His6-SpcS). Sequencing was done to check all the recombinant expression vectors.

Protein expression and purification

The expression of full-length N-terminal His tag ExoT–SpcS complex was done in E. coli BL21(DE3). Cells were grown at 37 °C to an optical density at 600 nm (A600) of 0.6. Recombinant gene expression was induced with 1 mm isopropyl-β-d thiogalactoside for 3 h. Induced cells were sonicated by adding protease inhibitors (such as phenylmethylsulfonyl fluoride) and buffer A solution (150 mm NaCl, 50 mm phosphate buffer, pH 7.8, and 10% glycerol). The solution was next centrifuged at 26 000 g for 20 min and the supernatant was charged into an Ni–nitrilotriacetic acid (NTA) agarose prepacked column, pre-equilibrated with buffer A. The column was then washed with a gradient imidazole concentration and buffer A, followed by specific elution of the protein complex with 250 mm imidazole concentration. The last step of purification involved loading into a size exclusion chromatographic column Superdex 200 HR 16/60 and elution with 50 mm phosphate buffer (pH 7.8) and 150 mm NaCl. The resulting purified complex was finally dialyzed using 25 mm Tris/HCl (pH 8.0) buffer and 50 mm NaCl. A similar purification protocol was followed for His tag SpcS and its mutants.

Cultures of E. coli BL21(DE3) transformed with pET28a+-exoT were grown at 37 °C to an optical density of 0.6 at 600 nm. Induction of ExoT was carried out at 30 °C for 2 h in the presence of 0.3 mm isopropyl-β-d-thiogalactopyranoside. Cultures were then centrifuged at 6000 g for 8 min and the pellet was resuspended in buffer A followed by sonication and high speed centrifugation (at 26 000 g for 20 min). Purification of sparingly soluble recombinant ExoT (supernatant) was done by affinity chromatography using an Ni–NTA agarose column. Purification buffer and procedures were similar as for the ExoT–SpcS complex.

Protein crystallization

The purified complex was concentrated using Amicon ultra columns to 25 mg·mL−1. Initial hits were found from wizard 1 sparse matrix crystallization screens. Final crystallization conditions were obtained by vapor diffusion at 22 °C by mixing an equal volume of protein complex and 24%–29% poly(ethylene glycol) monomethyl ether (PEG MME) 5000, 0.15 m (NH4)2SO4 and 0.1 m MES (pH 5.5). Hexagonal bipyramidal shaped single crystals were obtained after 8 months of crystal set-up.

X-ray data collection and processing

The crystals were cryoprotected in 28% PEG MME 5000, 0.15 m (NH4)2SO4, 0.1 m MES (pH 5.5) and 10% glycerol. They were next mounted in fiber loops on a Rigaku X-ray ultrax-18 (using the wavelength of CuKα) generator with a continuous flow of nitrogen at −180 °C as maintained by X-treme 2000. Diffraction data were collected using an R-axis IV+ image detector. Phase determination was done by the multiple isomorphous replacement technique, where the native crystals were soaked in different heavy metals such as 2 mm K4PtCl6, 0.6 mm Hg(OAc)2, 500 mm HgCl2 and 1 mm KAu(CN)2 and later subjected to X-ray diffraction. Diffraction images were indexed and scaled with hkl2000 software [42]. Initial heavy atom binding sites, refinement and phase calculation were done by solve/resolve software [43, 44]. phenix/coot software packages were used to build and refine the final model [45, 46]. molprobity [47] was used for stereochemical verification and secondary structure was assigned by dssp [48]. The complex crystallized in the P6122 space group with one ExoT– (SpcS)2 trimer in the asymmetric unit. The unit cell parameters were = 78.322 Å, = 78.322 Å, = 194.221 Å, α = 90°, β = 90° and γ = 120°. The crystals were found to have a specific volume (VM) of 2.71 Å3·Da−1 and 54.57% solvent content.

In silico Ala scanning mutagenesis

An approximate estimation of individual contributions of amino acid residues involved in the effector–chaperone interaction was obtained by the Rosetta interface computational mutagenesis approach [49], which is similar in principle to the experimental Ala scanning mutagenesis procedure. We utilized a cut-off of ∆∆Gbind ≥ 1.5 kcal·mol−1 to qualitatively identify hot-spot residues that are essential for the interactions.

Molecular dynamics study

MD simulations were performed using the gromacs 4.5.2 software package [50] using gromos96 (ff43a1) force field [51]. The X-ray crystal structure of ExoT–SpcS complex (PDB entry 4JMF), WT chaperone homodimer (chain B and chain C extracted from 4JMF) and the chaperone dimer double mutants SpcSF67A/W88A and SpcSN65A/W88A (mutated using Swiss PDB) were employed as the starting structures for simulations.

The starting structures were kept in a cubic box, pre-equilibrated with single point charge (SPC) water molecules [52], with a minimum distance of 1.0 nm between the protein system and box. In order to neutralize the overall charge of the system, a number of water molecules equal to the protein net charge were replaced by Na+ ions placed at random positions. An initial steepest descent energy minimization (10 000 steps) and 20 ps of solvent equilibration were done at constant temperature (300 K) using a 2 fs time step and 0.1 ps thermal coupling constant. The atomic positions were restrained using a harmonic potential. An initial velocity obtained from a Maxwell distribution at 300 K was given to all the atoms. All the simulations were run in an NVT environment employing V-rescale as temperature coupling algorithm, with reference temperature set at 300 K. The next 100 ps equilibration steps led to the final constant temperature (300 K) and pressure (1 bar).

Productive 10 ns MD simulations were given for the four systems in the NPT ensemble using an external bath with 0.1 ps coupling constant and constant 300 K temperature [53]. All the heavy atom bond lengths were constrained using the lincs algorithm [54]. The particle-mesh Ewald summation scheme [55] was used for calculating electrostatic interactions and Van der Waals and Coulomb cut-offs were kept at 0.9 nm. The non-pair lists were updated every 10 steps and snapshots were collected every 0.5 ps.

Analysis of MD simulations

The RMSDs of the protein main chain atoms were calculated using the g_rms module of gromacs taking the starting structure as the reference. Only backbone atoms were taken into account for computing the RMSD. For quantifying the flexibility of any residue i with respect to the mean square deviation from its position in the average structure, we used the root mean square fluctuation (RMSF) as implemented in the g_rmsf module. The average structures of all four simulations were obtained by comparing with the structure used for initializing the production run.

Matrices with pairwise correlations of atomic fluctuations were generated in the Bio3D module of the r software package [56], taking into account the coordinates of the protein Cα atoms since they contain enough information to describe the largest system motions. The cross-correlation coefficient C(i, j) for each pair of residues i and j is calculated by

display math(1)

where i and j may be any two atoms, residues or domains, Δri and Δrj are displacement vectors of residues i and j, and the angle brackets denote an ensemble average. This function returns a matrix of all atom-wise cross-correlations whose elements C(i, j) may be displayed in a graphical representation frequently termed a dynamical cross-correlation map, or DCCM. If C(i, j) = 1 the fluctuations of atoms i and j are completely correlated, if C(i, j) = −1 the fluctuations of atoms i and j are completely anti-correlated and if C(i, j) = 0 the fluctuations of i and j are not correlated.

Determination of binding affinity by surface plasmon resonance

The binding affinity (KD) and dissociation/association rates (kd/ka) of ExoT binding to WT and SpcS mutants were measured on a BIACORE 3000 system instrument using a sensor Ni–NTA chip. The His-tagged ExoT, SpcS and its mutants were dialyzed in 10 mm Hepes buffer (pH 7.4), 150 mm NaCl and 50 μm EDTA (buffer B). His-tagged WT and SpcS mutants were next digested with thrombin to remove the N-terminal His tag and the molecular weight of the resultant protein was verified by mass spectrometry to determine complete protease activity. His tag-ExoT was next coupled to the Ni–NTA chip in flow cell 2 and the respective response unit (RU) was standardized to predict the kinetic parameters between ExoT and SpcS. Various concentrations of SpcS WT and mutants in buffer B were passed over the ExoT linked Ni–NTA chip. Dissociation of the complex was done in the same buffer B. All SPR experiments were done with a constant flow rate of 10 μL·min−1 and at 25 °C temperature. Experiments with specific SpcS concentrations during each run were repeated thrice. For WT and single mutants of SpcS (SpcSW88A, SpcSF67A and SpcSN65A), ExoT was normalized to 600 RU and WT SpcS concentrations were 5 μm, 1 μm, 350 nm, 200 nm, 125 nm and 75 nm, and those of SpcS single mutants were 25 μm, 5 μm, 750 nm, 400 nm, 250 nm and 150 nm respectively. To perform kinetic analysis of double mutants, ExoT was standardized to 400 RU and the SpcSF67A/W88A concentrations were 750 μm, 200 μm, 100 μm, 50 μm, 25 μm and 15 μm and for SpcSN65A/W88A concentrations were 550 μm, 150 μm, 45 μm, 25 μm, 15 μm and 5 μm respectively. bia evaluation software version 4.1 was used for kinetic analysis and determination of various binding parameters. Various analyte concentrations through running buffer were also passed over flow cell 1 used as the reference cell and were later subtracted from experimental data (flow cell 2) using the above analysis software, in order to reduce signals arising due to non-specific binding to the sensor chip or non-specific changes in refractive index from the sensor chip [57]. Sensograms were fitted to a 1 : 1 binding model to obtain kon and koff rates. Each SPR experiment was repeated thrice to calculate the standard deviations in the resultant kinetic parameters.

Structural characterisation by CD

CD experiments were carried out by dialyzing WT and SpcS mutants (SpcSF67A/W88A and SpcSN65A/W88A) against phosphate buffer (10 mm sodium phosphate, pH 8.0, 30 mm NaCl). Far UV CD spectra for each of the samples were recorded from 200 nm to 250 nm in a Jasco J-815 spectrophotometer at 1 nm intervals, using a 0.1 cm path-length cuvette with a protein concentration of 2.5 μm for WT and 5 μm for mutants. The spectral bandwidth was 1.0 nm and five scans were recorded and averaged for each sample. The spectra were recorded at different temperatures such as 24 °C, 55 °C, 75 °C and 90 °C. Near UV CD spectra for each of the samples were recorded from 250 nm to 350 nm with SpcS protein concentrations of 15 μm for WT and 25 μm for mutants at 24 °C. All the wavelength scans were collected at 50 nm·min−1. The secondary structural contents were estimated using the k2d3 webserver [58].

Structure/sequence alignments and other servers

Structure-based sequence alignment of class I chaperones was generated from dali server [59]. Alignments of effector protein sequences and chaperone protein sequences were constructed using the mutalign program [60] and figures were prepared with espript [61].

The graphic images of molecular structure were made by the pymol program [62]. Domain–domain contacts were calculated with ligplot [63] and contact residues in effector–chaperone and chaperone–chaperone interfaces were identified by the ncont program from the ccp4 package [64]. Conservation of residues was analyzed with consurf server [65]. Webserver castp [66] was used to calculate the volume of the polar cavity.


The research was funded by grants from the Department of Science and Technology and Council of Scientific and Industrial Research, Government of India. Supratim Dey acknowledges Madhurima Das for the help in running gromacs and Samir Roy for assistance in SPR experiments.