Spores of Bacillus subtilis are dormant cell types that are formed when the bacterium encounters starvation conditions. Spores are encased in a shell, termed the coat, which is composed of approximately seventy different proteins and protects the spore's genetic material from environmental insults. The structural component of the basement layer of the coat is an exceptional cytoskeletal protein, termed SpoIVA, which binds and hydrolyzes ATP. ATP hydrolysis is utilized to drive a conformational change in SpoIVA that leads to its irreversible self-assembly into a static polymer in vitro. Here, we characterize the middle domain of SpoIVA, the predicted secondary structure of which resembles the chemotaxis protein CheX but, unlike CheX, does not harbor residues required for phosphatase activity. Disruptions in this domain did not abolish ATP hydrolysis, but resulted in mislocalization of the protein and reduction in sporulation efficiency in vivo. In vitro, disruptions in this domain prevented the ATP hydrolysis-driven conformational change in SpoIVA required for polymerization and led to the aggregation of SpoIVA into particles that did not form filaments. We propose a model in which SpoIVA initially assumes a conformation in which it inhibits its own aggregation into particles, and that ATP hydrolysis remodels the protein so that it assumes a polymerization-competent conformation.
Endospores produced by the bacterium Bacillus subtilis are dormant, hardy cell types that are formed to protect the organism's genetic material when it faces adverse environmental conditions (Stragier & Losick, 1996). Upon sensing the imminent onset of starvation conditions, the rod-shaped B. subtilis initiates the sporulation program which ultimately results in the production of a spore. During sporulation, the cell divides asymmetrically to produce two dissimilar-sized daughter cells that initially lie side-by-side. Next, the asymmetrically placed septum curves as the larger ‘mother cell’ compartment migrates around the smaller ‘forespore’ compartment. As a result, a roughly spherical forespore eventually resides in the rod-shaped mother cell cytosol as a double membrane-bound ‘organelle’. The mother cell then nourishes the forespore as it matures, whereupon the mother cell lyses, thereby releasing the mature (now largely dormant) spore into the environment, where it may remain dormant for decades (Setlow, 2007). During forespore maturation, the mother cell deposits a thick protein shell, termed the ‘coat’, onto the surface of the forespore. The coat is a complex structure composed of approximately seventy different proteins and participates in protecting the spore from environmental insults (Driks, 2002; Henriques & Moran, 2007; McKenney et al., 2013).
Coat morphogenesis initiates with the assembly of a basement layer on top of which the rest of the coat assembles (McKenney et al., 2010; McKenney & Eichenberger, 2012). The structural component of the basement layer is a soluble protein named SpoIVA (hereafter, simply ‘IVA’) (Roels et al., 1992; Driks et al., 1994; Price & Losick, 1999; Catalano et al., 2001) that is anchored to the surface of the forespore by a small amphipathic α-helical protein named SpoVM that senses positive membrane curvature (Ramamurthi et al., 2006, 2009; Ramamurthi, 2010). Although IVA is a structural protein, it curiously exhibits an ATPase activity (Ramamurthi & Losick, 2008). Indeed, the N-terminal half of the protein harbors motifs that are required for ATP binding and hydrolysis and assumes a tertiary fold that is predicted to be similar to the TRAFAC class of P-loop GTPases (Castaing et al., 2013). SpoIVA's preferential binding of ATP over GTP is likely due to an evolutionary alteration of a motif in its TRAFAC domain that typically confers specificity for GTP (Leipe et al., 2002; Castaing et al., 2013), an alteration which is also found in members of the myosin/kinesin family of ATPases which are also classified as members of the TRAFAC class of GTPases (Leipe et al., 2002).
Cytoskeletal proteins may be classified into two broad categories: (1) dynamic structures, like those formed by globular proteins such as actin or tubulin and their prokaryotic homologs, which require nucleotide binding to assemble and wherein nucleotide hydrolysis is linked to disassembly (Kueh & Mitchison, 2009); and (2) static structures found in the intermediate filament family of fibrous proteins that spontaneously assemble without a nucleotide requirement via coiled coil interactions, but do not exhibit the rapid disassembly and assembly behavior that is characteristic of dynamic structures (Goldman et al., 2012). Previously, we have proposed that SpoIVA defines a third category of cytoskeletal proteins, in that ATP hydrolysis, not simply ATP binding, results in the static, not dynamic, assembly of SpoIVA in vitro into filament-like structures (Ramamurthi & Losick, 2008; Castaing et al., 2013). Specifically, we have demonstrated that the energy released by ATP hydrolysis drives a massive conformational change in SpoIVA that places it in a polymerization-competent state in which it subsequently assembles once it surpasses a critical concentration for assembly.
Here, we investigate a second domain of IVA that lies downstream of its N-terminal TRAFAC domain. We termed it the ‘middle domain’. We found that the predicted secondary structure of this domain resembles the CheC/CheX/FliY (CXY) family of phosphatases, a component of the bacterial chemotaxis signal transduction pathway (Sircar et al., 2013). However, the middle domain does not harbor critical residues that are required for phosphatase activity, suggesting that it is likely a structural, not an enzymatic, domain. Disruption of several conserved structural regions of the middle domain resulted in the ATP-independent aggregation of IVA into particles that did not assemble into filaments in vitro and the misassembly of the spore coat in vivo. Furthermore, such IVA variants were unable to undergo the characteristic ATP hydrolysis-dependent conformational change that places IVA in a polymerization-competent state, even though they were folded properly enough to hydrolyze ATP. We propose a model in which SpoIVA initially assumes a conformation in which it inhibits itself from prematurely polymerizing, and that this conformation requires the middle domain. We suggest that a function of ATP hydrolysis is to remodel IVA so as to remove this autoinhibition, thereby allowing it to assemble in an orderly manner.
Materials and methods
Sequence profile searches were performed using the psi-blast program that run against the nonredundant (NR) protein database of National Center for Biotechnology Information (NCBI), to identify further homologs (Altschul et al., 1997). The multiple sequence alignment of IVA was built using the kalign program (Lassmann & Sonnhammer, 2005, 2006). Secondary structure was predicted using the jpred programs. The hhpred program was used for profile–profile comparisons (Soding, 2005; Soding et al., 2005). Structural manipulations were performed using the pymol program (DeLano, 2002).
Strain construction, cell growth, and general methods
Bacillus subtilis strains used are otherwise isogenic derivatives of strain PY79 (Youngman et al., 1984). Sporulation efficiencies were determined by growing cells in Difco Sporulation Medium (KD Medical) for at least 24 h, followed by exposure to 80 °C for 20 min to kill nonsporulated cells. For production of His6-tagged SpoIVA variants in Escherichia coli for purification, mutations in IVA were introduced using the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent) using plasmid pKR145 (Ramamurthi & Losick, 2008) as the template. For insertion of IVA alleles at ectopic loci in B. subtilis, site-directed mutagenesis was performed using either pKR13 (IVA under control of its native promoter for insertion at thrC) or pKR130 (IVA under control of its native promoter for insertion at amyE) (Ramamurthi & Losick, 2008) as templates (Guerout-Fleury et al., 1996). Mutagenesis was confirmed by DNA sequencing. Cellular protein levels were determined by immunoblotting extracts from sporulating cells as described previously (Ramamurthi & Losick, 2008) using rabbit antiserum raised against purified His6-IVA or σA-His6 (Covance, Inc.).
His6-IVA purification, dynamic light scattering, electron microscopy, ATP hydrolysis, limited proteolysis
His6-IVA (wild type and variants) was overproduced in E. coli BL21(DE3) pKR145 and derivatives and purified using Ni2+ affinity chromatography (GE Healthcare) followed by ion-exchange chromatography (Mono-Q; Pharmacia) exactly as described previously (Castaing et al., 2013). ATPase activity was measured by incubating purified protein with [α-32P]-ATP and monitoring ADP formation by thin-layer chromatography as described previously (Castaing et al., 2013); nonlinear regression analysis to estimate Vmax values used to calculate kcat was performed using prism 5 software (Graphpad). Trypsin-resistance assays were performed as described previously (Castaing et al., 2013). Briefly, 2 μM of purified protein was incubated in the presence or absence of 4 mM ATP for 4 h at 37 °C and then exposed to 1 ng mL−1 trypsin (Sigma) for the indicated times. Digestion was stopped by addition of sample buffer; samples were separated by PAGE and analyzed by Coomassie staining. Polymerization of IVA and derivatives at 4 μM final concentration was measured in the presence and absence of 4 mM ATP in 50 mM Tris (pH 7.6), 400 mM NaCl, 10 mM MgCl2, by exposure to laser light in a DynaPro NanoStar System photometer (Wyatt) as described previously (Castaing et al., 2013). Scattered light, measured as photons per second, was analyzed using dynamics v6 software (Novell). Polymerized IVA (4 μM) and derivatives were prepared for ultrastructure analysis by electron microscopy using a Hitachi H7650 microscope equipped with an Advanced Microscopy Techniques CCD camera as described previously (Castaing et al., 2013).
Overnight cultures of B. subtilis harboring GFP fusions to IVA and variants were induced to sporulate by the resuspension method (Sterlini & Mandelstam, 1969) in medium containing 1 μg mL−1 of the fluorescent membrane dye FM4-64. Cells were harvested and prepared for microscopy using an agarose pad as described previously (Eswaramoorthy et al., 2011). Cells were viewed with a DeltaVision Core microscope system (Applied Precision). Images were captured with a Photometrics CoolSnap HQ2 camera and deconvolved using softworx software (GE Healthcare) as described previously (Ebmeier et al., 2012).
The predicted structure of the IVA middle domain resembles the CheX family of phosphoesterases
To better understand the role of the middle domain of IVA, we prepared a multiple sequence alignment of all available orthologs of this protein and used it in profile–profile comparisons against a panel of HMMs derived from either the Pfam database or from structures in the PDB database with the hhpred program. Both these searches recovered the CheC/CheX/FliY (CXY) domain as the best hits (P = 10−5–4 × 10−4; probability of match 60–85%). Comparison of the secondary structure of the SpoIVA middle domain predicted using the jpred program with the known structure of the CXY superfamily (e.g. 3H2D) revealed a congruent pattern of structural elements (Fig. 1a). These observations suggested that the middle domain of the SpoIVA is structurally related to the CXY domain, which includes the CheC, CheX and FliY proteins that act as aspartyl phosphatases (Sircar et al., 2013). However, the CXY domain in another protein, FliM, appears to have lost it phosphatase activity and functions instead as a structural protein in the C-ring or switch complex of the rotor component of flagellar basal body (Paul et al., 2011; Sircar et al., 2013). As a part of the C-ring, FliM assembles into a 34-mer ring suggesting that catalytically inactive versions of the CXY domain might possess the ability to form multimeric structures. The crystal structures of the CXY superfamily reveal that they share a globular core formed by the duplication of a 3-stranded unit, each with a long helix (Fig. 1b and c) (Paul et al., 2011; Sircar et al., 2013). We accordingly labeled the cognate predicted helices and strands in IVA middle domain as H1 and S1-3 in the first unit and H1′ and S1′-3′ in the second unit. Examination of the known CXY domain structures suggested that dimer interfaces are potentially formed by helix H1, H1′ and strand S1′-2′.
To test whether such a multimerization role might be retained by the cognate structural elements in the IVA middle domain, we asked whether residues in the predicted Helix 1, Helix 1′, and the downstream strand S1-2′ are required for sporulation. Additionally, we tested the role of a highly conserved Trp at position 248 (W248) found at the boundary between the TRAFAC domain and H1 of the middle domain. Accordingly, we deleted the native copy of IVA and complemented the mutation in trans by expressing either wild-type or mutant alleles of IVA at an ectopic locus (thr). H1 was disrupted by deleting residues 256–258 (IVAH1), H1′ was disrupted by deleting residues 349–351 (IVAH1′), and the interface strand was disrupted by deleting residues 378–394 (IVAINT). Strains harboring a deletion in IVA are unable to sporulate (Roels et al., 1992); this defect was rescued when wild-type IVA was expressed ectopically (Table 1, strains A-C). In contrast, expression of IVAW248A, IVAH1, IVAH1′, or IVAINT alone was largely unable to restore sporulation (Table 1, strains D-G), even though the IVA variants were expressed from the ectopic locus at a level near to that of wild-type IVA (Fig. 2a). These results are consistent with a previous study (Catalano et al., 2001), in which a random mutagenesis of IVA revealed that two residues in what we now define as the interface strand (I383 and L393) and one residue in H1 (H256) are critical for IVA function. Taken together, the predicted secondary structure of the middle domain of IVA and our mutagenic analysis of key elements of that predicted structure suggest that H1, H1′, and the interface strand are required for IVA function in vivo.
Table 1. Sporulation efficiencies of strains harboring various IVA alleles
Spores per milliliter recovered, relative to PY79 (WT). PY79 routinely yielded approximately 1.5 × 108 spores per milliliter.
4.85 × 10−5
1.7 × 10−7
Disruption of H1′, the interface strand, and W248 do not abolish ATPase activity of IVA
To test whether disruptions introduced to the middle domain abolish ATP hydrolysis by the N-terminal ATPase domain of IVA, we overproduced in E. coli and purified from cell extracts, middle domain variants of N-terminal six-histidinyl tagged IVA. Although IVAH1 was stable and present at normal levels in B. subtilis extracts (Fig. 2a), we were unable to solubilize it at concentrations that were sufficient for downstream analyses when we attempted to purify it from E. coli. Therefore, the IVAH1 variant was not examined in subsequent in vitro experiments. The W248A and H1′ variants were severely reduced in their solubility but we were able to purify enough of each to perform selected experiments in vitro. We incubated the purified IVA variants with α-32P-ATP and measured the production of ADP by separating the products of hydrolysis by thin-layer chromatography. At Vmax, wild-type IVA hydrolyzed ATP with a turnover rate (Kcat) of 0.95 ± 0.03 pmol min−1 pmol−1 of IVA [(Castaing et al., 2013); Fig. 2]. As a negative control, IVAA*, which harbors a disruption in the nucleotide binding site of IVA, was almost completely defective in ATP hydrolysis [(Ramamurthi & Losick, 2008); Fig. 2]. By comparison, IVAW248A and IVAH1′ displayed Kcat values of 0.49 ± 0.08 and 0.56 ± 0.06 pmol min−1 pmol−1 of IVA variant, respectively, approximately half as efficiently as the wild-type IVA. IVAINT hydrolyzed ATP with an efficiency of 0.92 ± 0.07 pmol min−1 pmol−1 of IVAINT, a turnover rate that was nearly identical to the wild type. Given the low turnover rate of wild-type IVA to begin with, the low solubility of IVAW248A and IVAH1′, and the almost undetectable turnover rate of the IVAA* variant, we conclude that disruption of the hinge region (W248A) between the ATPase and the middle domain, H1′, and the interface strand abolish IVA function in vivo, but do not abolish ATP hydrolysis in vitro. These results also suggested that these disruptions to the middle domain did not disrupt the folding of the nucleotide binding pocket at the N-terminus of IVA.
Disruption of H1′ and the interface strand prevent ATP-dependent structural rearrangement of IVA
The hydrolysis of ATP leads to a conformational change in IVA, as evidenced by a change in the exposure of trypsin cleavage sites upon incubation of ATP that renders IVA largely resistant to cleavage by trypsin, even under nonpolymerization conditions (Castaing et al., 2013). To test whether features of the middle domain are required for the structural reorganization of IVA upon ATP hydrolysis, we examined the tertiary structures of middle domain variants by limited proteolysis by trypsin.
To this end, purified IVA, IVAH1′, or IVAINT were incubated, below the critical concentration for polymerization, in the presence or absence of ATP for 4 h, digested with trypsin for the 0–10 min, after which the products of digestion were separated by gel electrophoresis and visualized with Coomassie staining of the gel. In the absence of ATP, full-length IVA was largely degraded within just 2 min of exposure to trypsin; in the presence of ATP, however, IVA became largely resistant to trypsin, even after 10 min of exposure to the protease [(Castaing et al., 2013); Fig. 3]. Like wild-type IVA, IVAH1′ and IVAINT were also largely degraded, in the absence of ATP, after just 2 min of incubation with trypsin. However, in the presence of ATP, unlike wild-type IVA, both proteins exhibited sensitivity to digestion by trypsin: IVAH1′ was degraded in less than 2 min (similar to its behavior in the absence of ATP), and IVAINT was completely degraded in less than 5 min. Taken together, we conclude that H1′ and the interface strand are required for the structural reorganization of IVA upon ATP hydrolysis that places IVA in a polymerization-competent state.
Disruption of the interface strand prevents proper IVA polymerization
To monitor the effect of disruptions in the middle domain of IVA on polymerization, we measured polymerization of purified IVAH1′ over time, in the presence and absence of ATP, using dynamic light scattering. We were unable to perform this assay using the other middle domain variants of IVA because we were unable to purify them in sufficient enough quantities required for this assay. In the presence of ATP, but not in its absence, the light-scattering signal produced upon incubation of purified IVA increased over time, suggesting that it polymerized in an ATP-dependent manner (Fig. 4a). However, incubation of IVAH1′ with ATP led to a rapid increase in the light-scattering signal which, at its maximum, was nearly tenfold higher than that produced by wild-type IVA. Moreover, this rapid and extreme rise in light-scattering signal occurred even when IVAH1′ was incubated in the absence of ATP. The ATP-independent increase in light scattering suggested that IVAH1′ may be prone to aggregation, rather than proper polymerization.
To test whether this was simply nondescript aggregation, we examined the ultrastructure of different IVA polymers using transmission electron microscopy. Incubation of wild-type IVA with ATP yielded narrow filaments that extended for tens of microns as described previously (Ramamurthi & Losick, 2008; Castaing et al., 2013), which were not observed in the absence of ATP (Fig. 4b). However, incubation of IVAH1′ or IVAINT, both in the presence and absence of ATP, yielded particles that were approximately 50 nM in length that did not polymerize into a filament-like structure. Interestingly, such particles were qualitatively similar to those seen in the background of EM images of wild-type IVA that was incubated with ATP, suggesting that they did not represent a completely misfolded and unspecific aggregate of the protein, but rather a particular misfolding of IVA. This notion that the IVA middle domain variants were not simply globally misfolding and unspecifically aggregating was consistent with our observation that these IVA variants still retained ATPase activity (Fig. 2b), suggesting that at least the complex nucleotide binding pocket at the N-terminus was folded properly. Combined with the data from the light-scattering experiments, we conclude that disruptions of the middle domain of IVA result in the unregulated aggregation of the protein into larger particles, instead of the ATP hydrolysis-dependent assembly of the protein into filaments that are normally seen with the wild-type protein.
W248, H1′, and interface strand are required for proper morphogenesis of the coat
To determine whether variants in the middle domain of IVA that appear to mediate its polymerization are required for the morphogenesis of the spore coat in vivo, we examined the subcellular localization of GFP fused to IVA and its variants in sporulating cells of B. subtilis. Wild-type IVA when fused to GFP tracks the engulfing membrane and ultimately forms a uniform shell that surrounds the developing forespore (Price & Losick, 1999) (Fig. 5a). In contrast, although GFP-IVAW248A largely localized properly to the forespore, its localization pattern was perturbed relative to wild type. Rather than forming a uniform shell, most forespores displayed two discontinuous caps of GFP-IVA at the mother cell distal and proximal sides that were often connected by a band of GFP-IVA that traversed the forespore across the short axis of the cell (Fig. 5b). Disruptions to H1′ and the interface strand resulted in more severe phenotypes as GFP-IVAH1′ and GFP-IVAINT formed a single focus near the forespore and failed to track the engulfing membrane (Fig. 5c and d), a mis-localization pattern that is reminiscent of IVA localization when it fails to interact with SpoVM, which anchors IVA to the membrane (Ramamurthi et al., 2006).
To test whether later coat assembly steps were also affected upon disruption of the middle domain of IVA, we examined the localization of CotE (a later arriving coat protein) fused to GFP. In the presence of wild-type IVA, the CotE-GFP fusion localizes to the forespore surface with a biased accumulation on the mother cell-proximal surface of the forespore. In contrast, in the presence of IVAW248A, the majority of the CotE-GFP signal was localized to a focus near the forespore on its mother cell-proximal face with a residual amount of CotE-GFP localized to the remaining surface of the forespore. In the presence of IVAH1′ or IVAINT, however, even this residual amount of CotE-GFP did not remain and the protein completely mis-localized, usually as a single focus, in the mother cell cytosol. Taken together, we concluded that disruption of the middle domain of IVA resulted in the mis-localization of IVA in vivo and, by extension, abrogated assembly of the coat.
The structural component of the B. subtilis spore coat basement layer is composed of an exceptional cytoskeletal protein called IVA, which utilizes the energy released by ATP hydrolysis to drive a massive structural reorganization that ultimately results in its irreversible polymerization. IVA is composed of three distinct domains: an N-terminal ATPase domain that resembles members of the TRAFAC class of GTPases (Leipe et al., 2002; Castaing et al., 2013); a middle domain whose predicted tertiary structure resembles that of members of the CXY superfamily of phosphoesterases; and a small C-terminal domain with a predicted hydrophobic terminal strand, which has been implicated in anchoring IVA to the surface of the forespore via a small amphipathic helical protein (Ramamurthi et al., 2006). In this report, we examined the role that the middle domain of IVA plays during IVA polymerization in vitro and during formation of the basement layer of the spore coat in vivo. Guided by its predicted secondary structure, we disrupted three major regions of the middle domain of IVA: a conserved Trp residue (W248) that is located near the boundary of the N-terminal ATPase domain and the middle domain; three residues that reside in the H1′ helix of the middle domain; and seventeen residues in the interface strand (Fig. 1). Disruption of the middle domain resulted in IVA variants that were stably produced in vivo in B. subtilis and harbored ATPase activity in vitro, but were largely nonfunctional in that they were unable to undergo a characteristic conformational change upon ATP hydrolysis that places the protein in a polymerization-competent state. Instead, middle domain variants aggregated into particles of defined size in an ATP-independent manner instead of ATP hydrolysis-dependent filaments. Ultimately, cells harboring these mutant alleles of IVA were largely unable to construct the spore coat and, as a result, did not sporulate efficiently.
Taken together, we propose a working model in which IVA, when initially synthesized, assumes a conformation in which it prevents its own premature aggregation. In this model, the energy released by ATP hydrolysis is required to remodel the protein in such a way that it is competent for polymerization, but resistant to aggregation into particles. We propose that disruptions of the middle domain specifically disrupt the ability of IVA to assume this initial conformation, and as a result, molecules of these variants incorrectly interact with each other and aggregate into particles. Currently, in the absence of formal structural data, we are unable to conclude whether the middle domain represents the actual polymerization domain of IVA or an autoinhibitory domain that prevents the premature aggregation of a polymerization domain elsewhere in the protein. Nonetheless, the data are consistent with a model in which IVA has a propensity to aggregate, that this behavior may be curtailed by an initial conformation of IVA that is resistant to aggregation, that this conformation might be disrupted by amino acid changes in the middle domain, and that a role of ATP hydrolysis may be to remodel SpoIVA in an orderly fashion to place it in a polymerization-competent state.
We thank members of our laboratories for discussion and comments on the manuscript and Ulrich Baxa of the Electron Microscopy Laboratory (Frederick National Laboratory for Cancer Research) for assistance with EM. This work was funded by the Intramural Research Program of the National Institutes of Health, National Cancer Institute, Center for Cancer Research (J-P.C, S.L., and K.S.R.), National Institute of Allergy and Infectious Diseases (G.E.R.), and National Library of Medicine (V.A. and L.A.).