|Home | About | Journals | Submit | Contact Us | Français|
Coiled-coil domains in eukaryotic and prokaryotic proteins contribute to diverse structural and regulatory functions. Here we have used in silico analysis to predict which proteins in the proteome of the enteric pathogen, Salmonella enterica serovar Typhimurium, harbor coiled-coil domains. We found that coiled-coil domains are especially prevalent in virulence-associated proteins, including type III effectors. Using SopB as a model coiled-coil domain type III effector, we have investigated the role of this motif in various aspects of effector function including chaperone binding, secretion and translocation, protein stability, localization and biological activity. Compared to wild type SopB, SopB coiled-coil mutants were unstable, both inside bacteria and after translocation into host cells. In addition, the putative coiled-coil domain was required for the efficient membrane association of SopB in host cells. Since many other Salmonella effectors were predicted to contain coiled-coil domains, we also investigated the role of this motif in their intracellular targeting in mammalian cells. Mutation of the predicted coiled-coil domains in PipB2, SseJ and SopD2 also eliminated their membrane localization in mammalian cells. These findings suggest that coiled-coil domains may represent a common membrane-targeting determinant for Salmonella type III effectors.
Some pathogenic Gram-negative bacteria inject proteins, known as type III effectors, into host cells to manipulate host cell processes and establish their infective niche. Specialized injection devices, type III secretion systems (T3SS), span both bacterial membranes and puncture the host cell membrane to deliver these effectors. The enteric bacterium Salmonella enterica servoar Typhimurium has three major T3SS virulence determinants: the flagellar apparatus provides motility and is required for the optimal infection of non-phagocytic cells; T3SS1 is essential for the invasion of non-phagocytic cells; and T3SS2 is necessary for systemic spread of bacteria. To date, 39 Salmonella effectors have been identified as translocated substrates of these T3SS (McGhie et al., 2009, Niemann et al., 2011). While the activities of these effectors are extremely diverse, a universal theme is that these bacterial proteins must be delivered to a specific intracellular location in order to execute their function within host cells. Host cell membranes, both the plasma membrane (Cain et al., 2004) and internal membranes (Kuhle et al., 2002, Knodler et al., 2003, Freeman et al., 2003, Brumell et al., 2003), are an especially frequent target for Salmonella type III effector activity. Yet very little is known about how these bacterial proteins localize to host cell membranes.
Coiled-coil (CC) domains are structural elements that comprise two or more right-handed amphipathic α-helices that wind around each other into a supercoiled structure (Cohen et al., 1990). Despite their apparent simplicity, these structures show great diversity in their helix orientation (left- or right-handedness), alignment (parallel or anti-parallel), and mode of complex formation (homo- or hetero-complex) (Burkhard et al., 2001, Grigoryan et al., 2008, Parry et al., 2008). CC motifs, which provide protein-protein interaction sites, are found in a diverse array of proteins including those involved in transcription, membrane fusion and cell adherence (Burkhard et al., 2001), and are widely distributed in proteins from Eukarya, Archaea and Bacteria. Based upon two prediction programs (COILS and MultiCoil), up to 5% and 10% of all proteins encoded by prokaryotic and eukaryotic genomes, respectively, contain CC domains (Liu et al., 2001, Rose et al., 2005). Interestingly, three studies have revealed a higher prevalence of CC-containing proteins associated with T3SS, including needle proteins, translocators and effectors, compared to the average prokaryotic proteome (Pallen et al., 1997, Delahay et al., 2002, Gazi et al., 2009). This increased incidence has also recently been extended to T2SS and T4SS (Gazi et al., 2009). It has been suggested that the presence of CC domains in type III effectors might promote their unfolding prior to translocation and re-folding in the host cell (Gazi et al., 2009). Alternatively, since CC domains are often involved in protein-protein interactions, their presence may increase the likelihood of type III effector interaction(s) with host cell targets. Here we used two widely used CC prediction programs, Marcoil and Pepcoil, to evaluate the prevalence of CC domain containing proteins in the 4741 open reading frames (ORFs) of the pathogenic bacterium, Salmonella enterica serovar Typhimurium (S. Typhimurium). We identified 660 proteins to potentially containing one or more CC motifs, corresponding to 13.9 % of the entire proteome. Notably, and in agreement with previous reports (Pallen et al., 1997, Delahay et al., 2002, Gazi et al., 2009), the majority of known Salmonella type III effectors were predicted to contain at least one CC domain, suggesting that this motif could be a major determinant in their function or targeting.
We tested this hypothesis using SopB as a model CC domain-containing protein. sopB is transcribed in an operon from Salmonella pathogenicity island 5 (SPI5) together with its cognate chaperone, sigE (Hong et al., 1998), and is translocated via T3SS1, and is associated with host cell membranes upon translocation (Marcus et al., 2002). SopB has inositol phosphatase activity, is essential for activation of the mammalian pro-survival kinase, Akt/PKB, in epithelial cells (Steele-Mortimer et al., 2000) and promotes actin rearrangements at the plasma membrane (Zhou et al., 2001). A CC domain has been predicted in this type III effector between amino acid residues 118 and 142 (Marcus et al., 2002). We altered this putative CC domain by either deleting it entirely (SopBΔCC) or mutating three key hydrophobic residues at the interface of the α-helix (SopB-L3D). We assessed these mutants for chaperone binding, stability, secretion, translocation, localization, and biological activity in host cells. We discovered that the CC domain influences SopB stability inside bacteria and after translocation, and facilitates the membrane association of SopB in mammalian cells. Many Salmonella effectors that are membrane-associated upon translocation are predicted to contain CC domains (Table 1). We extended our analysis to other Salmonella type III effectors and found that these α-helical domains are necessary, but not sufficient, for subcellular targeting. Therefore CC domains function cooperatively with other domains to mediate the membrane localization of type III effectors in host cells.
Previous analyses have suggested an increased prevalence of CC domain-containing proteins in the proteomes of Gram-negative bacterial species harboring T3SS and T4SS (Pallen et al., 1997, Delahay et al., 2002, Gazi et al., 2009). In order to identify the ensemble of S. enterica proteins containing CC domains, the sequences of predicted open reading frames (ORFs) of S. Typhimurium SL1344 were obtained from the Sanger Institute (www.sanger.ac.uk/Projects/Salmonella/SL1344_web.tab) and putative functions were assigned by BLAST, based upon similarities with the S. Typhimurium LT2 proteome and manual searches in genomic databases (Ibarra et al., 2010). 4741 annotated ORFs were analyzed for potential CC motifs by the Pepcoil and Marcoil prediction programs. The Pepcoil program calculates the probability of a CC structure for windows of n residues (14, 21 or 28) in a protein sequence using the method of Lupas et al. (Lupas et al., 1991). Marcoil predicts CC domains based on a hidden Markov model-based algorithm (Delorenzi et al., 2002). Together, these two programs predicted 660 CC-domain containing proteins, corresponding to 13.9% of the complete ORFs annotated in the SL1344 proteome (see Supplementary Table 1). These CC domain-containing proteins can be categorized into into five functional classes according to their COG assignation and BLAST searches to the S. Typhimurium LT2 genome (Supplementary Table 2). The first group includes proteins that currently have no assigned functions (32%). One-quarter (26%) of the predicted proteins are involved in multiple metabolic pathways, transport, catabolism and biosynthesis. The third category (26%) comprises either transcriptional regulators, including various two-component system sensor proteins, such as PhoQ, SsrA, NarX, and BarA, DNA and RNA polymerase components, and/or components of the protein synthesis pathways (ribosomal proteins, tRNA synthetases, etc.). The fourth category (8.4%) includes proteins devoted to cell processes, such as motility and cell wall/membrane biogenesis. Lastly, 8.2% of the predicted proteins are involved in motility, signal transduction, trafficking and secretion and/or virulence processes. This last functional class includes structural components of the three T3SS in S. enterica as well as the flagellins, FliC and FljB, and numerous type III effectors.
In order to investigate further the prevalence of CC domains associated with T3SS substrates, we applied Marcoil and Pepcoil analyses on all known S. Typhimurium type III effectors (McGhie et al., 2009, Niemann et al., 2011)(Table 1), instead of restricting the target proteins to S. Typhimurium SL1344 as we had done in the initial predictions. Overall, 49% of the Salmonella effectors were predicted by Pepcoil and/or Marcoil to possess at least one CC domain, including all but one of the T3SS1-associated effectors, SopE2, and over one-third (36%) of the T3SS2-associated effectors. Seven effectors can be translocated by T3SS1 and T3SS2 and both programs predicted SspH1 and SlrP to harbor multiple CC domains. Pathogens use either T3SS or type IV secretion systems (T4SS) to deliver proteins into host cells. By comparison, the prevalance of CC domains in Pseudomonas aeruginosa type III effectors and Legionella pneumophila type IV effectors has been reported to be 20% and 24%, respectively (Gazi et al., 2009). Therefore, of the proteomes that have been analyzed to date, there is a much higher CC content in Salmonella effectors than those from other pathogenic bacteria.
Of the Salmonella type III effectors predicted to contain CC domains, there are currently four crystal structures available; SipA(48-264) (Lilic et al., 2006), SopA (163–782) (Diao et al., 2008), SptP(161-543) (Stebbins et al., 2000) and SspH2(166–783) (Quezada et al., 2009). These structures revealed that all of the regions predicted by both Pepcoil and Marcoil to harbor CC domains (Table 1) adopt α-helical structures (Supplementary Figure S1). Furthermore, these α-helices are localized to a solvent accessible area on the surface of their respective proteins, indicating that there is a strong potential for their association with other α-helices to form a CC structure. Overall, these structural observations reinforce our sequence predictions that Salmonella type III effectors contain domains capable of forming CC structures.
We chose SopB as a model protein to study the functional role of CC domains in Salmonella type III effectors. SopB has a cognate chaperone, SigE, and is translocated into host cells by T3SS1. A short region in the N-terminus of SopB is predicted by Pepcoil (amino acid residues 117–138) and Marcoil (residues 100–169) to form a CC domain (Table 1). In Figure 1A, a ribbon model of the predicted three-dimensional structure of SopB is shown. The corresponding CC motif identified in α-helix 10 is indicated and corresponds to amino acid residues 118–138 of SopB (residues 100–120 in the modeled sequence).
Typically, CC domains consist of one or more heptad repeats, denoted as (a-b-c-d-e-f-g)n. The “a” and “d” residues are usually small, hydrophobic amino acids such as leucine, valine and isoleucine. CC dimerization is driven by hydrogen bonding between these residues, shielding them in a hydrophobic core at the helix interface. Positions “b”, “c”, “e”, “f” and “g” are normally occupied by hydrophilic residues and form the solvent-exposed part of the CC. The COILS program (Lupas et al., 1991) (http://www.ch.embnet.org/software/COILS_form.html) predicted that five hydrophobic residues of SopB; Ala124, Leu127, Leu131, Leu134 and Leu145, occupy the “a” and “d” positions of their consecutive heptad repeats (Figure 1B). Using this prediction framework, we constructed two SopB CC mutants; an in-frame deletion of amino acids 118–142 (SopBΔCC) and a triple leucine-to-asparate mutant, L127D L131D L134D (SopB-L3D). These mutants were compared against wild-type SopB in a number of different assays for type III effector function.
To determine whether the SopB CC domain is required for effector-chaperone interaction, we used a pulldown assay based on AviTag™ technology. This 15 amino acid peptide tag is biotinylated in vivo and in vitro by the biotin ligase, BirA (Schatz, 1993), which is widely distributed in eubacteria and archaea, including S. enterica (Rodionov et al., 2002). Biotinylated proteins can then be detected or purified using streptavidin derivative products. ΔsopB-sigE bacteria were complemented with plasmids encoding SopB and SigE (pWSKDE), SopB tagged at the C-terminus with AviTag™ and SigE (pWSKDE-AviTag) or SopBΔCC tagged at the C-terminus with AviTag™ and SigE (pWSKDEΔCC-AviTag). The SopB-AviTag™ fusions were biotinylated in vivo and competent for secretion and translocation (data not shown). SopB binding to SigE was analyzed by streptavidin-Avitag™ affinity purification from bacterial lysates. Proteins bound to streptavidin agarose beads were eluted and subject to immunoblotting with antibodes against AviTag™, SopB and SigE. Equivalent amounts of SopB-AviTag and SopBΔCC-AviTag were affinity purified from bacterial lysates, whereas no untagged SopB was purified (Figure 2). Likewise, similar quantities of SigE co-purified with SopB-AviTag and SopBΔCC-AviTag, demonstrating that residues 118-142 are not required for the interaction between SopB and SigE.
While T3SS1 effectors are primarily translocated into host cells, they can also be secreted into culture medium under certain growth conditions. Because secretion and translocation are distinct transport mechanisms, we next characterized whether either of these processes was affected by mutation of the SopB CC domain. Salmonella were grown under microaerophilic conditions in LB-Miller broth for 8 h. Five strains were directly compared; wild type, ΔsopB-sigE, ΔsopB-sigE pWSKDE, ΔsopB-sigE pWSKDEΔCC and ΔsopB-sigE pWSKDE-L3D. Intrabacterial (bacterial lysates) and secreted proteins (culture medium) were collected, separated by SDS-PAGE and subject to immunoblotting with antibodies against SopB (Figure 3A). Confirming the deletion of sopB, no SopB signal was detected for the ΔsopB-sigE strain. Intrabacterial levels of SopB were increased by 4-fold for ΔsopB-sigE pWSKDE bacteria compared to wild type bacteria, in agreement with the multicopy nature of the complementing plasmid. Compared to ΔsopB-sigE pWSKDE, less intracellular SopB was detected for ΔsopB-sigE pWSDKEΔCC and ΔsopB-sigE pWSKDE-L3D bacteria, 1.7-fold and 1.3-fold respectively. Intrabacterial levels of the cytosolic chaperone, DnaK, were unchanged for all strains. We next evaluated the five Salmonella strains for SopB secretion by measuring the fraction of the total SopB pool that is secreted. For wild type and ΔsopB-sigE pWSKDE bacteria, approximately one-quarter of the SopB pool was secreted; 26±9.7% for wild type and 28±10% for ΔsopB-sigE pWSKDE. However, SopB secretion efficiency was significantly reduced for both CC mutants, to only 5.5±1.9% of the total pool for SopBΔCC and 9.9±2.7% for SopB-L3D. Altogether, these results establish that the CC region of SopB affects both its intrabacterial accumulation and secretion efficiency.
Translocation of SopB can be detected using a 13 amino acid tag derived from human glycogen synthase kinase-3β (GSK), which is phosphorylated following translocation (Garcia et al., 2006). We used this methodology to compare the translocation of wild type SopB with its CC mutant derivatives. ΔsopB-sigE bacteria complemented with pWSKDE-GSK, pWSKDEΔCC-GSK or pWSDKE-L3D-GSK were grown under conditions that induce invasiveness. The amount of SopB in these bacterial subcultures was initially compared by immunoblotting (Figure 3B). Similar to what was observed in the overnight standing cultures (Figure 3A), the amounts of SopBΔCC-GSK and SopB-L3D-GSK in the intrabacterial pool were found to be approximately 3-fold less than SopB-GSK. To compensate for this, the MOI was increased by 3-fold for pWSKDEΔCC-GSK and pWSKDE-L3D-GSK bacteria, so that equivalent amounts of SopB-GSK were present during the initial interaction with host cells. HeLa cell lysates were prepared from a time course of infection (0.5 h – 4 h p.i.) and subject to immunoblotting with phospho-Ser9-GSK3β and GSK tag antibodies to detect translocated and total (intrabacterial and translocated) SopB-GSK, respectively (Figure 3B). The amounts of total and translocated SopB-GSK accumulated over time, similar to what we have recently reported (Knodler et al., 2009). By comparison, the levels of total and translocated SopBΔCC-GSK and SopB-L3D-GSK were comparable to those of SopB-GSK at the earliest time points only (0.5 h and 1 h p.i.), but rapidly declined thereafter. The migration pattern of SopBΔCC and SopB-L3D also differed compared to wild type SopB. Translocated SopB-GSK was primarily detected as three bands, representing unmodified, mono- and di-ubiquitinated species (Knodler et al., 2009, Patel et al., 2009). However, the CC mutants migrated as single bands, at ~60 kDa, suggesting a lack of ubiquitination after delivery to mammalian cells. Collectively, these data show that, although the CC domain is not essential for SopB translocation per se, it is required for the post-translational modification and persistence of translocated SopB.
In bacterial cultures, the intrabacterial SopB pool for the CC mutants was reduced compared to their wild type counterpart (Figure 3), suggesting that the stability of SopB was affected for these proteins. To assess this, we measured the intrabacterial stability of SopB, SopBΔCC and SopB-L3D following chloramphenicol treatment to prevent de novo protein synthesis. ΔsopB-sigE bacteria complemented with pWSKDE, pWSKDEΔCC or pWSKDE-L3D were grown to mid-log phase in LB-Miller broth, then samples were collected at various times post-antibiotic addition. Proteins were separated by SDS-PAGE and subject to immunoblotting with polyclonal SopB antibodies (Figure 4A). For SopB, there was a slow, gradual decline in the intrabacterial pool with time, whereas SopBΔCC and SopB-L3D decayed faster. The calculated half-life of SopB was 80 min, while that of SopBΔCC and SopB-L3D was 20 min and 40 min, respectively. Therefore, the SopB CC mutants have a stability defect inside bacteria.
We also evaluated the stability of SopB, SopBΔCC and SopB-L3D following translocation into mammalian cells. Again, we used a modified pulse-chase experiment in combination with SopB-GSK fusions to estimate their half-lives. HeLa cells were infected with ΔsopB-sigE bacteria harboring either pWSKDE-GSK (10 µl subculture), pWSKDEΔCC-GSK (50 µl subculture) or pWSKDE-L3D-GSK (50 µl subculture). Increased MOIs were used to ensure adequate signal intensity in immunoblots for accurate densitometry analysis. At 1.5 h p.i., tetracycline was added to stop bacterial protein synthesis (t0) and samples were collected at hourly intervals thereafter. Lysates were subject to immunoblot analysis with phospho-Ser9-GSK3β antibodies to detect translocated SopB (Figure 4B). As previously estimated (Knodler et al., 2009), the half-life of translocated SopB was 2 h. By comparison, the half-lives of the CC mutants were much shorter, approximately 55 min. Little to no SopBΔCC-GSK or SopB-L3D-GSK was detected within 2 h of tetracycline addition. Taken together these experiments establish that the CC domain is required for optimal SopB stability, both in Salmonella and after translocation into mammalian cells.
We next assessed whether the CC domain affected the activity of SopB in mammalian cells. SopB is an inositol phosphatase that targets host cell phospholipids. This enzymatic activity is essential for activation of the mammalian pro-survival kinase, Akt/PKB, in epithelial cells (Steele-Mortimer et al., 2000). We therefore used the phosphorylation status of Akt as a reporter for the biological activity of SopB in mammalian cells. Whole cell lysates from HeLa cells expressing Myc-tagged wild type SopB, catalytically inactive SopB (SopB C460S), or the two CC mutants, SopBΔCC and SopB-L3D, were collected and subject to immunoblotting for SopB, phospho-Akt and total Akt. While SopB and SopB C460S appeared as two bands in immunoblots, representing unmodified and mono-ubiquitinated species (Figure 5A; (Knodler et al., 2009)), SopBΔCC and SopB-L3D migrated as a single band only, indicating lack of ubiquitination. Therefore, neither ectopically expressed nor bacterially delivered SopB CC mutants are post-translationally modified in mammalian cells (Figures 3B, ,5A).5A). Akt phosphorylation was induced upon ectopic expression of wild type SopB, SopBΔCC or SopB-L3D, but not the catalytically inactive SopB mutant, SopB C460S (Figure 5A). We also examined CC domain-dependence of the biological activity of SopB after translocation by Salmonella. HeLa cells were infected with increasing amounts of ΔsopB-sigE bacteria expressing SopB-GSK (pWSKDE-GSK) or SopBΔCC-GSK (pWSKDEΔCC-GSK). At 1 h p.i., monolayers were solubilized and subject to immunoblotting for translocated SopB-GSK, total SopB-GSK, phospho-Akt and total Akt. Confirming an absolute requirement for SopB in HeLa cells (Steele-Mortimer et al., 2000), ΔsopB-sigE bacteria failed to induce the phosphorylation of Akt, even at a high MOI (Figure 5B). In contrast, increasing the MOI for ΔsopB-sigE bacteria expressing SopB-GSK resulted in a proportional increase in the amount of translocated SopB-GSK and phospho-Ser473-Akt, demonstrating the dose-dependence of Salmonella-induced Akt phosphorylation. A similar trend was observed for ΔsopB-sigE bacteria expressing SopBΔCC-GSK, although equivalent amounts of translocated SopB-GSK induced higher levels of Akt phosphorylation (Figure 5B). Therefore, the CC deletion mutant is able to induce Akt phosphorylation, but less efficiently than wild type SopB.
We hypothesized that the instability of SopBΔCC might contribute to its reduced ability to effectively induce Akt phosphorylation at the plasma membrane. We therefore monitored a time course of Akt phosphorylation in the presence of the proteasome inhibitor, MG-132. HeLa cells were infected with ΔsopB-sigE bacteria expressing either SopB-GSK (pWSKDE-GSK) or SopBΔCC-GSK (pWSKDEΔCC-GSK), in the presence or absence of MG-132. Samples were collected over a 4 h time course and subject to immunoblotting (Figure 5C). For wild type SopB, increasing amounts of translocated SopB were evident with time, in agreement with our earlier study (Knodler et al., 2009). Sustained phosphorylation of Akt was detected up to 4 h p.i., with a peak at 1 h p.i. (Knodler et al., 2005a). Surprisingly, inhibiting proteasomal degradation led to an increased accumulation of translocated SopB and prolonged Akt phosphorylation kinetics, shifting the peak of phosphorylation to ≥ 4 h p.i. SopBΔCC induced low levels of Akt phosphorylation in untreated cells. However, in the presence of MG-132 Akt phosphorylation was restored, with an intensity and longevity comparable to wild type SopB in the absence of MG-132 (Figure 5C). Collectively, these experiments show that translocated SopB and SopBΔCC are subject to proteasomal degradation and the CC domain of SopB is not essential for its biological activity in mammalian cells.
It has previously been shown that SopB targets host cell membranes after translocation (Marcus et al., 2002, Knodler et al., 2009, Patel et al., 2009). Furthermore, residues 117–167 are required for membrane association of an enhanced green fluorescent protein (EGFP)-SopB fusion as determined by detergent fractionation of transiently transfected cells (Marcus et al., 2002). To further examine the role of the CC domain in membrane association, we constructed an EGFP-full length SopB fusion and various mutants, including the catalytically inactive mutant, EGFP-SopB C460S, and the CC mutants, EGFP-SopBΔCC (deleted for residues 118–142) and EGFP-SopB L3D (L127D L131D L134D mutations). These plasmid constructs were transfected into HeLa cells and protein localization was then monitored by fluorescence microscopy (Figure 6A). By direct fluorescence microscopy, EGFP-SopB and EGFP-SopB C460S associated with vesicles scattered throughout the cytoplasm, in agreement with previous reports (Dukes et al., 2006). In contrast EGFP-SopBΔCC and EGFP-SopB L3D were evenly distributed throughout the cytoplasm and also concentrated in membrane ruffles at the periphery of some cells. We next used a sequential detergent fractionation procedure to determine whether SopB and its mutants were membrane associated or cytosolic. HeLa cells were transfected with Myc-SopB, Myc-SopB C460S, Myc-SopBΔCC or Myc-SopB-L3D for 18 h and then treated with 0.1% (w/v) saponin, which permeabilizes the plasma and internal membranes and releases cytosolic content (Wassler et al., 1987), followed by incubation with 1% (v/v) Triton X-100 (TX-100) to solubilize peripheral and integral membrane proteins. Samples were subject to immunoblotting with antibodies against SopB, the lysosomal membrane protein, LAMP-1, and the cytosolic protein, Hsp27 (Figure 6B). Myc-SopB and Myc-SopB C460S were mostly present in the TX-100 soluble fraction, indicating their membrane association. Notably, the small portion of SopB and SopB C460S present in the saponin-soluble fraction was unmodified, whereas both unmodified and ubiquitinated forms partitioned to the TX-100 soluble fraction. In contrast, Myc-SopBΔCC and Myc-SopB L3D were primarily detected in the saponin-soluble fraction, hence are cytosolic.
We also tested the intracellular localization of the SopB CC mutants after bacterial translocation. For infected cells, we used a different fractionation technique since TX-100 has been shown to compromise bacterial integrity (Gauthier et al., 2000). HeLa cells were infected with ΔsopB-sigE bacteria complemented with pWSKDE-GSK, pWSKDEΔCC-GSK or pWSKDE-L3D-GSK. At 1 h p.i., cells were subject to mechanical lysis followed by a low speed centrifugation step to remove intact bacteria and then a high speed centrifugation step to separate cytosolic from membrane proteins (Figure 6C). Equal volumes of the cytosolic and membrane fractions were subject to immunoblotting to detect translocated SopB-GSK, the membrane protein, LAMP-1, and the cytosolic protein, Hsp27. In agreement with previous reports (Marcus et al., 2002, Knodler et al., 2009, Patel et al., 2009), SopB localized exclusively to host cell membranes after translocation, while deletion or mutation of the CC domain caused a significant amount of translocated SopB to partition to the cytosolic fraction (Figure 6C). Together, these data indicate an intact CC domain is required for the membrane localization of SopB in mammalian cells.
We next tested whether CC domains are important for the membrane localization of three other Salmonella type III effectors; PipB2, SseJ and SopD2, which are predicted by the COILS program to have a CC domain either at their N-terminus (PipB2) or C-termini (SseJ and SopD2) (Table 1). Specifically, residues 51–66 of PipB2, 308–325 of SseJ and 275–291 of SopD2 are predicted to adopt a CC conformation (data not shown). Using a similar approach to SopB, we constructed EGFP fusions to each of these effectors and also made triple amino acid substitutions to asparate in positions “a” or “d” of consecutive heptad repeats to disrupt this potential structural motif. The fusion proteins were expressed in HeLa cells and their localization determined by fluorescence microscopy and sequential detergent fractionation (Figure 7).
PipB2 is targeted to host cell membranes and localizes to punctate vesicles that are concentrated at the cell periphery (Figure 7A) (Knodler et al., 2005b). However, we found that the corresponding CC mutant, PipB2 Y54D L57D M61D (PipB2 YLM3D), was spread diffusely throughout the cytoplasm and concentrated in a juxtanuclear position (Figure 7A). SopD2 localizes to vesicular membranes (Brumell et al., 2003) (Figure 7A) and like for PipB2, its association with vesicular structures was abrogated by mutation of the predicted CC domain (SopD2 Y281D Y284D K288D, SopD2 YYK3D) (Figure 7A). SseJ also localizes to cellular membranes (Freeman et al., 2003, Ruiz-Albert et al., 2002) and was detected on the plasma membrane when ectopically expressed (Figure 7A) (Nawabi et al., 2008), but mutation of Asn309 Leu312 Val316 residues (SseJ NLV3D) shifted its localization pattern to the cytoplasm (Figure 7A).
We independently confirmed these observations by differential detergent solubilization. Transfected HeLa cells were subject to sequential incubations with saponin and TX-100 as described above. Samples from both fractions were subject to immunoblotting with antibodies against GFP, the membrane marker, LAMP-1, and the cytosolic protein, Hsp27 (Figure 7B). EGFP-PipB2 was detected only in the TX-100 soluble fraction representing cellular membranes, but the EGFP-PipB2 YLM3D mutant was solely present in the saponin-soluble, or cytosolic, fraction. EGFP-SseJ and EGFP-SopD2 were enriched in the TX-100 soluble fraction whereas their respective CC mutants were present exclusively in the saponin-soluble fraction. Therefore disruption of the potential CC domains of PipB2, SseJ and SopD2 render these proteins cytosolic. Collectively, the fluorescence microscopy and fractionation data support the concept that CC domains in Salmonella type III effectors facilitate their targeting to host cell membranes.
To determine whether the predicted CC domains were sufficient for membrane targeting, we constructed EGFP fusions to amino acid residues 113–147 of SopB, 41–73 of PipB2, 303–335 of SseJ or 268–300 of SopD2. HeLa cells were transfected with the EGFP-CC domain fusions and their localization determined by direct immunofluorescence and sequential detergent fractionation. All of the fusion proteins showed localization patterns indistinguishable from EGFP, with fluorescence detected in the nucleus and cytosol (Supplemental Figure S2A). Detergent fractionation confirmed that the EGFP-CC domain fusions, like EGFP, were cytosolic (Supplemental Figure 2B). Therefore the predicted CC domains of Salmonella type III effectors are necessary, but not sufficient, to confer membrane targeting.
To test the interchangeability of type III effector CC domains, we constructed a number of chimeric fusion proteins where we replaced the predicted CC domain of one effector with another. One Salmonella effector, SseK1, has a region that is predicted by Marcoil to adopt a CC conformation (Table 1), but is not targeted to membranes (Kujat Choy et al., 2004). To see whether the predicted CC domain of SopB could confer membrane targeting of SseK1, we replaced amino acid residues 157–177 of SseK1 with residues 117–137 of SopB and compared the localization of EGFP-SseK1 and EGFP-SseK1-SopBCC after transfection of HeLa cells. By direct immunofluorescence microscopy (Supplemental Figure S2B), EGFP-SseK1 localized to the nucleus and cytosol (Kujat Choy et al., 2004), whereas the SseK1-SopBCC chimera was largely excluded from the nucleus and accumulated primarily in the cytosol. However, both proteins were only detected in the saponin-soluble fraction after detergent fractionation (Supplemental Figure S2B), confirming their cytosolic status.
Having demonstrated that disruption of the predicted CC domains of PipB2, SseJ and SopD2 affected their membrane targeting, we hypothesized that these regions could functionally replace the CC domain of SopB and target it to membranes. To test this, we replaced residues 117–137 of SopB with the regions in these proteins predicted to adopt a CC structure i.e. residues 47–67 of PipB, 309–329 of SseJ or 274–294 of SopD2. Plasmids encoding these chimeric fusions, EGFP-SopB-PipB2CC, EGFP-SopB-SseJCC and EGFP-SopB-SoPD2CC, were transfected into HeLa cells and protein localization was monitored by fluorescence microscopy and sequential detergent fractionation. All three chimeras showed the same localization pattern, accumulating in the cytosol and membrane ruffles at leading edges of the plasma membrane (Figure 8A), which is comparable with the SopB CC mutants, SopBΔCC and SopB L3D, and not wild type SopB. In immunoblots, EGFP-SopB-PipB2CC, EGFP-SopB-SseJCC and EGFP-SopD2CC migrated as single species on immunoblots, indicating a lack of ubiquitination, and were primarily solubilised by saponin (Figure 8B). Collectively, this data shows that none of the chimeric proteins were efficiently targeted to host cell membranes.
Like SopB, the Shigella homologue, IpgD, hydrolyzes host cell phospholipids at the plasma membrane, leading to the sustained activation of Akt in infected epithelial cells (Pendaries et al., 2006). This type III effector is predicted by the COILS program to have three CC domains (data not shown), one at the N-terminus in a region analogous to the predicted CC domain of SopB. To investigate whether these residues could function as a membrane targeting determinant in SopB we transfected HeLa cells with a construct that expresses a protein with residues 117–137 of SopB replaced by residues 118–144 of IpgD. We found that EGFP-SopB-IpgDCC exhibited a staining pattern comparable to EGFP-SopBΔCC and EGFP-SopB L3D, and not wild type SopB, (Figure 8A) and localized primarily in the saponin-soluble fraction (Figure 8B). Therefore the predicted CC domain of IpgD cannot functionally substitute for that of SopB.
In this study, we used in silico analysis to predict CC domains in the proteome of the pathogenic bacterium, Salmonella enterica. We found a high propensity for CC domains in virulence-associated proteins, especially type III effectors and used one of these, SopB, to experimentally determine the role of CC domains in Salmonella type III effectors. We chose SopB since it is a relatively well-characterized effector and its biological activity in mammalian cells can be readily detected. We found that the CC domain is required for the efficient secretion of SopB and enhances SopB stability, both intrabacterial and inside host cells after translocation. In addition, the CC domain promotes SopB membrane association in mammalian cells. There was no requirement for the CC domain in chaperone binding or type III translocation. These results indicate that the CC domain of SopB serves multiple functions, both inside bacteria and host cells.
Previously, Letzelter and others (Letzelter et al., 2006) have shown that the membrane localization domain (MLD) overlaps with the chaperone-binding domain (CBD) for the Yersinia effectors YopO, YopE and YopT. Inside bacteria, binding of a class I chaperone to the CBD prevents type III effector aggregation. Upon translocation, the chaperone is no longer associated with the effector, and this freed region acts as a membrane-targeting domain in host cells (Letzelter et al., 2006). As proof of this hypothesis, deletion of the CBD in YopO prevented the interaction with its chaperone, SycO, and its association with the plasma membrane (Letzelter et al., 2006). However, our data suggest that these findings cannot necessarily be extended to Salmonella effectors. The SopBΔCC mutant retains the ability to bind its class I chaperone, SigE (Figure 2), but no longer localizes to host cell membranes (Figure 6), implying that the CBD and MLD do not coincide for SopB. Of note, the CBD/MLD regions of YopO, YopE and YopT are not predicted by COILS (Lupas et al., 1991) to adopt CC conformations. Therefore it appears that the specific role of chaperone binding in masking an aggregation-prone membrane localization signal is divergent for Salmonella and Yersinia type III effectors.
Translocated SopB targets both plasma and vacuolar membranes (Knodler et al., 2009, Patel et al., 2009) and has inositol phosphatase activity that acts on membrane phospholipids. This enzymatic activity is essential for phosphorylation of the mammalian pro-survival kinase, Akt/PKB, at the plasma membrane (Steele-Mortimer et al., 2000). Molero et al. have previously demonstrated that a SopBΔ118–142 mutant does not activate Akt at 30 min p.i. (Molero et al., 2009). By contrast, here we show that a SopB CC mutant can induce Akt phosphorylation when equivalent amounts of SopB and SopBΔCC are translocated (Figure 5). Therefore, the CC domain is not essential for the biological activity of SopB, but does contribute to its efficacy. This is most likely related to the defective intracellular localization of the SopB CC mutants. SopB is a peripheral membrane protein (Marcus et al., 2002, Knodler et al., 2009, Patel et al., 2009) and our cellular fractionation data shows that the SopB CC mutants are no longer enriched in membrane fractions. Instead, at steady state, a considerable portion of these mutant proteins partition to the cytosolic fraction (Figure 6), demonstrating that the CC domain enhances the association of SopB with host cell membranes. Overall our data supports the hypothesis that the biological activity of SopB requires a stable association with host cell membranes, and the CC domain increases its membrane affinity, likely via protein-protein or protein-lipid interaction(s). Interestingly, the mammalian inositol phosphatase, myotubularin (Mtmr2), also requires a CC domain to enhance its membrane association (Berger et al., 2003). Homodimerization of the CC domain combines two lipid-binding pleckstrin homology-GRAM (PH-G) domains, increasing the binding affinity of Mtmr2 for membrane phosphoinositides. In an analogous manner, homodimerization of SopB might function to stabilize its interaction with host cell membranes.
Translocated SopBΔCC and SopB-L3D are not completely deficient in membrane association (Figure 6) implying that additional membrane-targeting domains exist within SopB. In agreement with this, a short hydrophobic domain in the C-terminus of SopB (residues 288–309) was recently identified as being required for its membrane targeting in host cells (Patel et al., 2009). This is not surprising since CC domains alone are rarely able to impart membrane targeting (Stam et al., 1997, Raiborg et al., 2001) (Supplemental Figure S2) and usually function with other domains to effect the subcellular targeting of mammalian proteins (Buchsbaum et al., 1996, Hume et al., 2006, Stam et al., 1997, Berger et al., 2003). However it was unexpected that predicted CC domains from other Salmonella effectors could not substitute for the SopB CC domain in directing membrane association (Figure 8). From these domain-swapping experiments we conclude that it is not simply the secondary structure adopted by a CC domain, but also specific amino acid residues within this region of SopB that provide a protein-protein interface for interaction with other proteins, thus directing its specific subcellular localization. This is analogous to the specificity of pleckstrin homology (PH) domains for distinct membrane phosphoinositides within mammalian cells (Lemmon, 2008). Therefore the targeting of SopB to host cell membranes requires the cooperative function of multiple domains, one of which is the putative CC domain.
Previously, two targeting mechanisms had been identified for the membrane association of bacterialproteins in host cells: (i) the covalent attachment of fatty acids or lipids promoting membrane binding affinity, (ii) motifs that mediate membrane binding via protein-protein interactions. We propose that CC domains represent a third membrane-targeting determinant for type III effectors. Examples of the first category include bacterial proteins with sequence motifs that are recognized by host cell enzymes responsible for post-translational modifications; prenylation of CAAX-like motifs at the C-terminus of the Salmonella type III effector SifA (Reinicke et al., 2005) and numerous L. pneumophila type IV effectors (Ivanov et al., 2010) is required for their membrane anchoring. Similarly, modification of eukaryotic-like acylation sites in three type III effectors from the plant pathogen Pseudomonas syringae drives their efficient membrane association (Nimchuk et al., 2000). In the second category, small protein modules facilitate the membrane recruitment of bacterial proteins. For example, short hydrophobic stretches in the N-termini of YopE from Y. pseudotuberculosis and ExoS from P. aeruginosa serve as membrane localization signals in host cells (Krall et al., 2004). A region in the N-terminus of SopD that is conserved in the Salmonella translocated effector (STE) family, the WEKI(I/M)xxFF motif, is required for membrane targeting of this effector (Bakowski et al., 2007), but whether it is involved in the membrane association of other STE family members has not been investigated. Finally, mutations in a predicted α-helix in the C-terminus of the P. aeruginosa type III cytotoxin ExoS disrupt its plasma membrane localization (Veesenmeyer et al., 2010). Interestingly, of these three categories, only targeting determinants that lead to a permanent post-translational modification such as acylation or lipidation are sufficient to mediate membrane attachment. For example, the C-terminal hexapeptide of SifA containing the CAAX motif (Boucrot et al., 2003) and a CAAX motif in conjunction with a nearby stretch of positively charged amino acids for Legionella effectors (Ivanov et al., 2010) is sufficient for their membrane targeting. However, while the N-terminal hydrophobic regions of YopE and ExoS are sufficient for directing membrane association (Krall et al., 2004), SopD requires a CC-domain in addition to its WEKI(I/M)xxFF motif for membrane targeting (Bakowski et al., 2007). Furthermore, a CC-rich region of LidA encompassing residues 191–541 is sufficient for membrane association (Derre et al., 2005), but not the predicted CC domains of Salmonella effectors (this manuscript, (Bakowski et al., 2007)).
Targeting to specific subcellular compartments is essential for type III effector function. Here we have shown that CC domains of four Salmonella type III effectors, SopB, PipB2, SopD2 and SseJ, are required for their membrane association. SopD is another Salmonella effector that binds to host cell membranes (Bakowski et al., 2007) and is predicted to contain a CC domain at its C-terminus (Table 1). When heterologously expressed in mammalian cells, a SopD truncation mutant lacking the predicted CC domain localized to the cytosol (Bakowski et al., 2007). Additionally, the T3SS1 translocon component and effector, SipC/SspC, is predicted to have multiple CC domains (Table 1). Individual point mutations in one CC domain at the C-terminus of SipC/SspC disrupted its plasma membrane association after bacterial translocation (Scherer et al., 2000). Therefore, CC domains facilitate the membrane localization of at least six Salmonella type III effectors. Recently, a high prevalence of CC domains was also predicted for type IV effectors (Gazi et al., 2009). Although, little data exists validating a connection between CC domains and the membrane association of type IV effectors, the Legionella pneumophila effector LidA provides one example (Derre et al., 2005). Collectively, the in silico analyses and experimental data support a correlation between the presence of α-helical CC domains in type III and type IV effectors and their ability to associate with host cell membranes. Notably, a few Salmonella effectors are not predicted to have CC domains, but do target host cell membranes (SopE2, PipB, SseF, SseG; Table 1), so must utilize CC-independent targeting mechanisms for their intracellular localization.
Several studies have demonstrated that monoubiquitination regulates the function of SopB in mammalian cells (Marcus et al., 2002, Patel et al., 2009) and up to nine lysine residues in SopB serve as ubiquitin conjugation sites (Patel et al., 2009, Knodler et al., 2009). Here we demonstrate that the SopB CC mutants, SopBΔCC and SopB L3D, are defective for monoubiquitination. Another SopB mutant, SopBΔ288–309 (Patel et al., 2009), and the chimeric proteins SopB-PipB2CC, SopB-SseJCC, SopB-SopD2CC and SopD-IpgDCC (Figure 8) are also not ubiquitinated and fail to associate with host cell membranes after translocation. Within the predicted CC domain are four lysine residues (Lys118, Lys125, Lys128, Lys129) and one lysine residue is within the hydrophobic domain (Lys297) but none of these residues are targets of ubiquitin (Patel et al., 2009, Knodler et al., 2009). This argues that lack of ubiquitination of the SopB mutants and chimeras is an indirect effect of their inability to associate with membranes. Interestingly, this bears a striking similarity to the mammalian endosomal scaffold protein, CIN85. CIN85 directly interacts with the membrane phospholipid, phosphatidic acid (Zhang et al., 2009), and is monoubiquitinated by the ubiquitin ligase, Cbl, on endosomal membranes (Haglund et al., 2002). Deletion of a predicted CC domain in the C-terminus of CIN85 abolishes phosphatidic acid binding and reduces interaction with Cbl. As a result, the CC deletion mutant of CIN85 is cytosolic and not monoubiquitinated (Haglund et al., 2002, Zhang et al., 2009). Further studies are required to clarify which ubiquitin ligase is responsible for the modification of SopB when it is bound to membranes.
Based upon previous reports that translocated SopB is not subject to polyubiquitination and degradation by the proteasome (Marcus et al., 2002, Patel et al., 2009), we did not expect MG-132 treatment to affect SopB stability and activity (Figure 5). However, proteasome inhibition blocked the degradation of translocated SopB and SopBΔCC and significantly enhanced the longevity of SopB- and SopBΔCC-induced signaling in epithelial cells, suggesting that translocated SopB is subject to polyubiquitination. Indeed, we have previously reported that immunoprecipitation of translocated SopB indicates mono-, di- and poly-ubiquitinated species (Knodler et al., 2009). Notably, a similar function for proteasomal regulation of bacterial protein activity in host cells has been reported for two other T3SS1 effectors, SopE and SptP (Kubori et al., 2003). Therefore, in addition to the documented monoubiquitination-dependent effects on SopB localization (Patel et al., 2009), our data supports a role for polyubiquitination in regulating the kinetics of SopB activity inside mammalian cells.
In conclusion, we report that a high percentage of Salmonella T3SS effectors are predicted to have one or more CC domains. Using SopB as a model type III effector, we identified that this domain has a multi-faceted role, and is central in defining the cellular location of translocated SopB. We extended this functional characterization to other Salmonella effectors and demonstrated that CC motifs facilitate membrane association of these proteins too. Given the high prevalence of putative CC domains in type III and type IV effectors, we believe that this is a commonly used mechanism for targeting bacterial proteins to host cell membranes.
The complementing plasmid, pWSKDE, (Knodler et al., 2009) is a low copy number plasmid that encodes SopB and its cognate chaperone, SigE, under the control of the sopB promoter. To create the CC deletion mutant, pWSKDEΔCC, residues 118–142 of SopB were removed by overlap extension PCR (Horton et al., 1989), using pWSKDE as a template. The initial round of PCR used SigD-Hind-F (see Supplementary Table 3 for oligonucleotide sequences) with SigD d118-142-R and SigD d118-142-F with SigD-Nhe-R. These amplicons were purified and used as a template for a second round of PCR with SigD-Hind-F and SigD-Nhe-R. The resulting fragment was digested with HindIII/NheI and cloned into HindIII/XbaI digested pWSK29 (Wang et al., 1991). The pWSKDE-L3D mutant (pWSKDE L127D L131D L134D) was generated by site-directed mutagenesis with the Quikchange kit (Stratagene).
Overlap extension PCR was also used to construct pSopB-AviTag, a plasmid that expresses a SopB-AviTag™ fusion (Avidity Inc.) under the control of the sopB promoter. Two PCR fragments were generated using pWSKDE as a template with the oligonucleotides SigD-Hind-F and SigD-Rv-AP to amplify sopB and its promoter region, and AP-SigD-Fw and SigD-Nhe-R to amplify sigE. The resulting amplicons were gel purified and used in a second round of PCR with the oligonucleotides SigD-Hind-F and SigD-Nhe-R. This amplicon was gel purified, digested with HindIII/NheI and ligated into HindIII/XbaI-digested pWSK29 to generate pSopB-AviTag. pSopBΔCC-AviTag was created in a similar manner, but using pWSKDEΔCC as a template.
To detect translocated SopB by immunoblotting, a 13-residue peptide tag derived from human GSK-3β was fused to the C-terminus of SopB. This plasmid, pWSKDE-GSK, has been described previously (Knodler et al., 2009). The corresponding CC deletion mutant, pWSKDEΔCC-GSK, was constructed as described for pWSKDE-GSK (Knodler et al., 2009), except that pWSKDEΔCC was used as the template for amplification. The pWSKDE-L3D-GSK mutant was generated by Quikchange site directed mutagenesis (Stratagene).
pCMV-Myc-SopB encodes for N-terminal Myc-tagged SopB (Knodler et al., 2009) and was used for ectopic expression studies. Overlap extension PCR was used to create Myc-SopB deleted for the CC domain, pCMV-Myc-SopBΔCC. The initial PCR products were amplified from S. Typhimurium genomic DNA with the oligonucleotides Myc-SigD-F and SigDΔ118-142R, SigDΔ118-142-F and pGAD-SigD-R. These two amplicons were mixed for the second round of PCR with the oligonucleotides Myc-SigD-F and pGAD-SigD-R. The amplicon was digested with EcoRI and BglII and ligated into the corresponding sites of pCMV-Myc (Clontech). pCMV-Myc-SopB L3D and pCMV-Myc-SopB-C460S were generated by Quikchange site-directed mutagenesis (Stratagene).
Alternatively, the codon usage of sopB was optimized for expression in human cells and this plasmid, EX-SopB-M29 (GeneCopoeia), was used as a template for amplification. To create EGFP-HuSopB, the humanized sopB coding sequence was amplified from EX-SopB-M29 with the oligonucleotides Kpn-HuSigD-F and BglII-HuSigD-R. The amplicon was purified, digested with KpnI/BglII and ligated into KpnI/BamHI digested pEGFP-C1 (Clontech). To generate EGFP-HuSopBΔCC, EX-SopB-M29 was used as a template for amplification with the oligonucleotides Kpn-HuSigD-F and HuSigD d118–142-R, and HuSigD d118-142-F and BglII-HuSigD-R. These two amplicons were mixed together for a second round of PCR with Kpn-HuSigD-F and BglII-HuSigD-R. The PCR product was purified, digested with KpnI/BglII and ligated into KpnI/BamHI digested pEGFP-C1. EGFP-HuSopB C460S and EGFP-HuSopB L3D were generated by Quikchange site-directed mutagenesis (Stratagene).
For ectopic expression of other Salmonella effectors, EGFP fusions were constructed. EGFP-PipB2 has been described previously (Knodler et al., 2005b). sseJ was amplified from S. Typhimurium genomic DNA with the oligonucleotides GFP-SseJ-F and GFP-SseJ-R. The resulting amplicon was digested with BglII/KpnI and ligated into the corresponding sites of pEGFP-C1 (Clontech) to create EGFP-SseJ. sopD2 was amplified with the oligonucleotides GFP-SopD2-F and GFP-SopD2-R, the amplicon digested with BglII/KpnI and ligated into BglII/KpnI-digested pEGFP-C1 to create EGFP-SopD2. Likewise, sseK1 was amplified with EGFP-SseK1-Bgl and EGFP-SseK1-Kpn, digested and ligated into BglII/KpnI-digested pEGFP-C1 to create EGFP-SseK1. EGFP-PipB2 Y54D L57D M61D (EGFP-PipB2 YLM3D), EGFP-SseJ N309D L312D V316D (EGFP-SseJ NLV3D) and EGFP-SopD2 Y281D Y284D K288D (EGFP-SopD2 YYK3D) were generated by site-directed mutagenesis.
To assess the sufficiency of the CC domains for membrane localization, the predicted CC domains of SopB, PipB2, SseJ and SopD2 were fused to the C-terminus of EGFP. The following oligonucleotide pairs were used in PCR amplification with S. Typhimurium SL1344 genomic DNA; HuSopB-113F and HuSopB-147R, EGFP-PipB2-41F and EGFP-PipB2-73R, EGFP-SopD2-268F and EGFP-SopD2-300R, EGFP-SseJ-303F and EGFP-SseJ-335R. The amplicons were digested with BglII/KpnI and ligated into BglII/KpnI-digested pEGFP-C1 to create EGFP-SopBCC, EGFP-PipB2CC, EGFP-SopD2CC and EGFP-SseJCC, respectively.
Overlap extension PCR was used to replace amino acids 117–137 of SopB with the predicted CC domains of IpgD, PipB2, SseJ and SopD2. EGFP-HuSopB was used as a template for amplification with the oligonucleotides Kpn-HuSigD-F and HuSopB-IpgDCC-R, HuSopB-PipB2CC-R, HuSopB-SseJCC-R or HuSopB-SopD2CC-R, and BglII-HuSigD-R with HuSopB-IpgDCC-F, HuSopB-PipB2CC-F, HuSopB-SseJCC-F or HuSopB-SopD2CC-F. The resulting amplicons were gel purified and mixed for a second round of PCR with Kpn-HuSigD-F and BglII-HuSigD-R. The PCR products were gel purified, digested with KpnI/BglII and ligated into KpnI/BamHI digested pEGFP-C1 to create EGFP-HuSopB-IpgDCC, EGFP-HuSopB-PipB2CC, EGFP-HuSopB-SseJCC and EGFP-HuSopB-SopD2CC, respectively. To replace amino acid residues 157–177 of SseK1 with the SopB CC domain, EGFP-SseK1 was used as a template with EGFP-SseK1-Bgl and SseK1-SopBCC-R, and EGFP-SseK1-Kpn and SseK1-SopBCC-F. The resulting amplicons were purified and mixed for a second round of PCR with EGFP-SseK1-Bgl and EGFP-SseK1-Kpn, followed by BglII/KpnI digestion and ligation into pEGFP-C1 to create EGFP-SseK1-SopBCC. All plasmid sequences were verified by DNA sequencing.
The predicted ORFs of S. Typhimurium SL1344 were obtained from the Sanger Institute (http://www.sanger.ac.uk/Projects/Salmonella/SL1344_web.tab) and putative functions assigned by BLAST based upon the S. Typhimurium LT2 proteome. 4741 annotated ORFs (Ibarra et al., 2010) were analyzed for CC propensities using two prediction programs, Pepcoil and Marcoil. Pepcoil calculates the probability of a CC domain for windows of 14, 21 or 28 residues in a protein sequence using the method of Lupas et al (Lupas et al., 1991). Here we have used Pepcoil with a window size of 14 and a probability of 1.0. Marcoil is a hidden Markov model-based program that predicts the presence and location of CC domains in protein sequences (Delorenzi et al., 2002). Marcoil was used with parameters by default and a threshold of 2.0. In addition, all known S. Typhimurium type III effectors (McGhie et al., 2009, Niemann et al., 2011) were analyzed for CC domains using Marcoil and Pepcoil algorithms as described above.
To evaluate the performance of the CC predictions, two additional data sets comprising 420 proteins with well-characterized CC domains (true-positive data set) were compared to 420 randomly generated protein sequences with no known CC domains (true-negative data set). This comparison was useful to calculate the following values: (1) true-positives (TP), proteins with at least one CC structure identified by any method; (2) false-positives (FP), proteins whose well-known CC was not identified; (3) false-negatives (FN), proteins with a CC structure not identified by any method; (4) sensitivity of the fraction of proteins recovered in the inferred CC search and was calculated as Sn = TP/(TP + FN); (5) positive predictive value (PPV) is the fraction of the proteins and CC motifs that belong to the well-known data set, calculated as PPV = TP/(TP + FP); and (6) accuracy (Ac) is the PPV and Sn average, calculated as Ac= (Sn + PPV)/2. Results for performance evaluation are as follows:
Overall both methods detected true-positive proteins with a high accuracy. In order to have a better prediction performance, we defined proteins with at least one CC motif as “positive” if they were identified by both prediction algorithms.
A computational model of the three-dimensional structure of SopB was built by the CPHmodels 2.0 Server (http://www.cbs.dtu.dk/services/CPHmodels) using a probable sugar ABC transporter from Streptococcus pneumoniae (PDB entry 2w7y:A) as a template. SopB and 2w7y:A are 31% identical at the amino acid level. In order to refine the model, it was minimized using Gromacs server (http://lorentz.immstr.pasteur.fr/gromacs). Finally, the RAMPAGE program (http://mordred.bioc.cam.ac.uk/~rapper/rampage.php) was used to validate the stereochemical quality of the resulting three-dimensional model. After analyzing the Ramachandran plot, 92.4% of the residues are in favored and allowed regions and only 7.6% are in disallowed regions. All the bond distances, angles and dihedrals fulfill the normal limits for polypeptide chains. The model includes 463 of the 561 residues of SopB. The lack of a signal peptide in the template protein was likely responsible for the inability to model the first 18 amino acid residues of SopB.
SopB and SopBΔCC binding to SigE were analyzed by a pulldown assay. Wild-type Salmonella, ΔsopB-sigE, ΔsopB-sigE complemented with plasmid borne SopB-AviTag and SigE (pSopB-AviTag), or SopBΔCC-AviTag and SigE (pSopBΔCC-AviTag) were grown shaking (225 rpm) overnight at 37°C in LB-Miller broth, then subcultured 1:33 into 10 ml of fresh LB-Miller broth. After 3.5 h shaking at 37°C (OD600~3.5), bacteria were pelleted by centrifugation and resuspended in 1 ml phosphate buffered saline (PBS) supplemented with protease inhibitor cocktail set II (Calbiochem), followed by sonication in an ice bath for 2 min (cycles of 20 sec pulse/10 sec rest) using a S4000 sonicator (Misonix). Bacterial extracts were cleared by centrifugation at 16,000 × g for 20 min at 4°C. Equivalent amounts of SopB-AviTag and SopBΔCC-AviTag were used as the reaction mixture input (3.3-fold more SopBΔCC-AviTag culture), determined from immunoblotting of bacterial extracts, and incubated with 60 µl of streptavidin-agarose beads (Invitrogen) for 2 h at 4°C. Beads were washed three times with PBS containing 0.5% (v/v) Triton X-100. Bound proteins were eluted by boiling in 1.5× SDS-PAGE sample buffer, and separated on 12% or 15% SDS-PAGE gels and transferred to 0.2 µm or 0.45 µm nitrocellulose (Bio-Rad). Blots were blocked in Tris-buffered saline supplemented with 0.1% (v/v) Tween-20 (TBST) and 5% (w/v) skim milk powder (TBST-milk) for 2 h at room temperature. Immunoblotting was with mouse monoclonal anti-AviTag™ antibodies (Avidity Inc.; 1:10,000) or rabbit polyclonal anti-SigE antibodies ((Darwin et al., 2001); 1:2,000). Blots were washed three times with TBST and incubated with HRP-conjugated goat anti-rabbit IgG or horse anti-mouse IgG (Cell Signaling Technology; 1:10,000) in TBST-milk for 1 h at room temperature. Immunoblotting with affinity purified rabbit polyclonal anti-SopB antibodies ((Marcus et al., 2002); 1:40,000) was as described previously (Knodler et al., 2006). Blots were developed using enhanced chemiluminescence according to manufacturer’s protocol (SuperSignal West Femto Maximum Sensitivity detection system, Thermo Scientific) and imaged using either a Kodak Image Station 440 system or Kodak BioMax Light film.
Wild-type or ΔsopB-sigE Salmonella harboring pWSKDE, pWSKDEΔCC or pWSKDE-L3D were grown under microaerophilic conditions for 8 h in 10 ml LB-Miller broth for the analysis of intrabacterial and secreted protein levels. Preparation of samples has been described previously (Knodler et al., 2006). Aliquots of intracellular (15 µl, equivalent to 0.6% original sample) and secreted (6 µl, equivalent to 2.3% original sample) proteins were separated by SDS-PAGE, transferred to 0.45 µm nitrocellulose (Bio-Rad) and subject to immunoblotting with rabbit polyclonal anti-SopB antibodies ((Marcus et al., 2002); 1:40,000) and mouse monoclonal anti-Escherichia coli DnaK antibodies (Stressgen; 1:20,000). Blots were incubated with affinity purified HRP-conjugated goat anti-rabbit or horse anti-mouse IgG (Cell Signaling Technology; 1:20,000) and developed using enhanced chemiluminescence as above. To calculate SopB secretion efficiency, identical exposure times were used for chemiluminescence detection of intracellular and secreted protein samples and band intensity was quantified using a Kodak Image Station 440 system with 1D Image Analysis software. Secretion efficiency was calculated as follows: secreted/[(3.8xintracellular)+secreted] x100%.
HeLa adenocarcinoma epithelial cells (ATCC CCL-2) were grown in Eagle’s modified medium (Mediatech, Herndon, VA) containing 10% (v/v) heat-inactivated fetal calf serum (Invitrogen, Carlsbad, CA). Cells were seeded in 24-well tissue-culture treated plates (Costar) 12–18 h prior to infection. Cells were serum-starved for 3 h prior to infection and incubated in serum-free media post-infection. Preparation of invasive bacteria and infection conditions has been described in detail previously (Knodler et al., 2006). For all infections, three-fold more SopBΔCC and SopB-L3D bacteria (MOI~150) than SopB bacteria (MOI~50) were added to monolayers to compensate for the deficit in mutant SopB protein levels in bacterial subcultures. For some experiments, cells were pretreated for 45 min with 10 µM MG-132 (Calbiochem), and the inhibitor was maintained for the duration of the experiment.
HeLa cells were seeded in 6-well tissue-culture treated plates (Costar) at 2.4×105 cells/well the day prior to the experiment. Cells were serum-starved for 3 h prior to infection, and maintained in serum-free media for the duration of the experiment. Cells were infected with invasive bacteria as described above (MOI~50 for SopB, MOI~150 for SopBΔCC and SopB-L3D). Infected monolayers were solubilized in boiling 1.5× SDS-PAGE sample buffer and proteins separated by SDS-PAGE. Immunoblotting was with rabbit monoclonal antibodies against phospho-Akt (Ser473) (D9E, Cell Signaling Technology; 1:20,000) and Akt (pan) (11E7, Cell Signaling Technology: 1:2,000) as described previously (Knodler et al., 2009).
The intra-bacterial stability of SopB, SopBΔCC and SopB-L3D was measured by a modified-pulse chase experiment (Knodler et al., 2006). ΔsopB-sigE bacteria harboring pWSKDE, pWSKDEΔCC or pWSKDE-L3D were inoculated from freshly streaked LB agar plates into 10 ml LB-Miller broth containing 50 µg/ml carbenicillin. Bacteria were grown, standing, at 37°C for 4–5 h to mid-log phase (OD600~0.4). Chloramphenicol (60 µg/ml) was then added to stop de novo protein synthesis and at various times post-antibiotic addition 1 ml aliquots were collected, centrifuged at 16,000 × g for 2 min and the pellets resuspended in 100 µl (pWSKDEΔCC or pWSKDE-L3D) or 200 µl (pWSKDE) boiling 1.5× SDS-PAGE sample buffer. Proteins (15–20 µl from each sample) were separated by SDS-PAGE and subject to immunoblotting with rabbit polyclonal antibodies against SopB (1:20,000) as described previously (Knodler et al., 2006).
The stability of SopB was also measured after its translocation into host cells (Knodler et al., 2009). Serum-starved HeLa cells were infected with ΔsopB-sigE bacteria harboring pWSKDE-GSK (10 µL inoculum, MOI~50), pWSKDEΔCC-GSK (30 µL inoculum, MOI~150) or pWSKDE-L3D-GSK (30 µL inoculum, MOI~150). At 1.5 h p.i., 10 µg/ml tetracycline was added to stop bacterial protein synthesis. At hourly intervals thereafter, infected cells were lysed in 150 µl boiling 1.5× SDS-PAGE sample buffer. Samples were subject to immunoblotting with rabbit polyclonal anti-phospho-GSK-3β (Ser9) (Cell Signaling Technology; 1:2,000) or rabbit polyclonal anti-GSK-3β tag (Cell Signaling Technology; 1:5,000) antibodies as described previously (Knodler et al., 2009).
The partitioning of SopB and the SopB CC mutants to host cell membranes was assayed by the mechanical fractionation of infected HeLa cells as described previously (Knodler et al., 2003), with minor modifications. HeLa cells were seeded in 10 cm tissue culture treated dishes (Corning), serum-starved for 3 h prior to infection and maintained in serum-free media throughout. HeLa cells were infected with 50 µl invasive ΔsopB-sigE bacteria complemented with pWSKDE-GSK or 400 µl ΔsopB-sigE bacteria complemented with pWSKDEΔCC-GSK or pWSKDE-L3D-GSK per 10 cm dish (four dishes per condition). At 1 h p.i., cells were mechanically disrupted in homogenization buffer containing protease inhibitor cocktail set III and phosphatase inhibitor cocktail set II (EMD Biosciences). Intact nuclei and bacteria and unbroken cells were removed by a low speed centrifugation at 6,000 × g for 10 min at 4°C, and the supernatant was subject to high-speed ultracentrifugation at 100,000 × g for 30 min at 4°C to separate host cell membranes from cytosol. Equal volumes of these samples were subject to immunoblotting with rabbit polyclonal anti-phospho-GSK-3β antibodies (Ser9) (Cell Signaling Technology; 1:2,000), mouse monoclonal anti-LAMP1 antibodies (H4A3 developed by J.T. August and J.E.K. Hildreth was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa; 1:5,000) and mouse monoclonal anti-Hsp27 antibodies (G31, Cell Signaling Technology, 1:20,000) as described previously (Knodler et al., 2009).
Plasmid DNA was prepared using the Perfectprep® Plasmid Midi kit according to the manufacturer’s instructions (Eppendorf). HeLa cells were seeded in 6-well plates or on acid-washed glass coverslips in 24-well tissue culture plates 8 h prior to transfection with FuGENE® 6 reagent (Roche). Cells were lysed in boiling 1.5× SDS-PAGE sample buffer or fixed for microscopy 18–24 h post-transfection.
For analysis of the membrane association of Salmonella effector proteins and their CC mutants when ectopically expressed, transfected HeLa cells were subject to sequential detergent extraction with 0.1% (w/v) saponin and 1% (v/v) TX-100 as described previously (Marcus et al., 2002). Saponin- and TX-100-soluble fractions were analyzed by immunoblotting with rabbit polyclonal anti-GFP antibodies (Invitrogen; 1:50,000) and mouse monoclonal anti-LAMP1 and anti-Hsp27 antibodies (see above).
Transfected cells were fixed in 2.5% (w/v) paraformaldehyde for 10 min at 37°C. DNA was labeled with 5 µM DRAQ5™ (Biostatus Limited) for 5 min at room temperature and coverslips mounted in ProLong Gold Antifade reagent (Invitrogen). Images (1024×1024 pixels, 0.35 µm optical sections) were acquired on a Zeiss LSM510 or LSM710 confocal microscope equipped with a Plan APOCHROMAT 63×/1.4 NA objective. Flat maximum intensity projections of 3–6 sections were assembled using LSM510 AIM v4.2 or Zen 2008 software (Carl Zeiss MicroImaging) and final figures were prepared using Adobe Photoshop CS and Canvas X.
Supplementary Figure S1: Location of predicted CC domains (shown in red) in crystal structures of Salmonella type III effectors. Structures for the following Salmonella effectors are shown: SipA(48–264) (Protein Data Bank (PDB) identification (ID): 2FM9), SptP(161–543) (PDB ID: 1G4W), SopA(163-782) (PDB ID: 2QYU), SspH2(166–783) (PDB ID: 3G06). Residues 144–168 of SipA, 275-294 of SptP, 368-384 of SopA, 673–693 and 720–736 of SspH2 are predicted by Pepcoil to form CC domains and are highlighted in red.
Supplementary Figure S2: CC domains are not sufficient for membrane association. (A) Upper panel. HeLa cells were transfected with plasmids expressing EGFP alone or the following EGFP-CC domain fusions; EGFP-SopBCC (amino acid residues 113–147 of SopB), EGFP-PipB2CC (residues 41–73 of PipB2), EGFP-SseJCC (residues 303–335 of SseJ) or EGFP-SopD2CC (residues 268–300 of SopD2). Monolayers were fixed after 20 h. Confocal microscope images show EGFP fusions in grey scale. Scale bar, 10 µm. Lower panel. Transfected HeLa cells were harvested and subject to sequential detergent fractionation. Equal volumes of the saponin-soluble (S, cytosolic proteins) and TX-100 soluble (T, membrane proteins) fractions were separated by SDS-PAGE and subject to immunoblotting with antibodies against GFP, LAMP-1 (membranes) and Hsp27 (cytosol). (B) Left panel. HeLa cells were transfected with plasmids expressing EGFP-SseK1 or EGFP-SseK1-SopBCC. Monolayers were fixed after 20 h. Confocal microscope images show EGFP fusions in grey scale. Scale bar, 10 µm. Right panel. Transfected cells were subject to sequential detergent fractionation and immunoblotting as described in (A).
Supplementary Table S1: Salmonella enterica serovar Typhimurium SL1344 proteins predicted by Pepcoil and Marcoil algorithms to contain CC domains. For each protein, the assigned SL1344 gene is correlated with the homologous gene in S. Typhimurium LT2 and its corresponding annotation.
Supplementary Table S2: Functional classification of S. Typhimurium proteins with a predicted CC motif. A search in the cluster of orthologous groups database (COGs, http://www.ncbi.nlm.nih.gov/COG/) was performed for each protein in Supplementary Table S1 (COG_number column). COG categories are as follows: Metabolism C, E, F, G, H, I, P and Q; DNA/RNA processes J, K, L and O; Cell processes D, M and N; Defense and intracellular trafficking T, U and V. Those with no identity to any COG were classified as No_Cogs. Bar graph shows the proportion of proteins calculated as a percentage of the proteins per COG divided by the total number of proteins in the predicted CC group (CC_COGs blue bars) or by the total number or proteins in the genome (COGs_Genome orange bars). A Welch two sample t-test analysis shows that there is no significant different between CC_COGs and COGs_Genome. The pie chart shows the percentage of the total number of predicted CC domain-containing proteins for each of the functional groups.
Supplementary Table S3: Oligonucleotide sequences used in this study.
We kindly thank Victor Cid for helpful discussions, Jean Celli for critical reading of this manuscript, Seth Winfree for IT assistance, members of the Steele-Mortimer laboratory for helpful discussions, Gary Hettrick and Austin Athman for graphics assistance, and the Genomics Core Facility at Rocky Mountain Laboratories for DNA sequencing analysis. This research was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases. E.P-R was financed by a grant (IN-217508) from DGAPA-UNAM.