The Bacillus subtilis cell division proteins FtsL and DivIC are intrinsically unstable and do not interact with one another in the absence of other septasomal components

Authors


Summary

The bacterial septum appears to comprise a macromolecular assembly of essential cell division proteins (the ‘septasome’) that are responsible for physically dividing the cell during cytokinesis. FtsL and DivIC are essential components of this division machinery in Bacillus subtilis. We used yeast two-hybrid analysis as well as a variety of biochemical and biophysical methods to examine the proposed interaction between Bacillus subtilis FtsL and DivIC. We show that FtsL and DivIC are thermodynamically unstable proteins that are likely to be unfolded and therefore targeted for degradation unless stabilized by interactions with other components of the septasome. However, we show that this stabilization does not result from a direct interaction between FtsL and DivIC. We propose that the observed interdepend-ence of DivIC and FtsL stability is a result of indirect interactions that are mediated by other septasomal proteins.

Introduction

Cell division in rod-shaped bacteria such as Escherichia coli and Bacillus subtilis is characterized by the formation of a division septum at the centre of the elongating cell. The septum appears to comprise a macromolecular assembly of essential cell division proteins (collectively referred to as the ‘septasome’ or ‘divisome’) that act in concert to facilitate the co-ordinated ingrowth of the cytoplasmic membrane, the rigid peptidoglycan layer and, in the case of Gram-negative bacteria, the outer membrane of the cell envelope. The septasome can thus be envisaged as a macromolecular machine that divides the cell into two discrete daughter cells.

The initiating event in bacterial cytokinesis is the formation at mid-cell of a circumferential ring of polymerized FtsZ (the ‘Z ring’) (Lutkenhaus and Addinall, 1997). FtsZ, the bacterial homologue of tubulin (Erickson, 1997; Löwe and Amos, 1998), is a GTPase that may provide the contractile force for septal invagination. Additionally, the Z ring acts as a scaffold that is essential for the recruitment of the remaining septasomal proteins (Rothfield et al., 1999). In E. coli, the cytoplasmic septasomal components FtsA and ZipA are first recruited to the Z ring in a mutually in-dependent fashion (Hale and de Boer, 1999) by direct interaction with the C-terminal region of FtsZ (Wang et al., 1997; Ma and Margolin, 1999; Hale et al., 2000). The remaining septasomal proteins then localize to the division site in the following strict temporal order: FtsK, FtsQ, FtsL, FtsW, FtsI (penicillin binding protein 3; PBP 3) and FtsN (see Chen and Beckwith, (2001) and references therein).

Many of the B. subtilis septasomal proteins are homologues of those in E. coli, including FtsZ and FtsA (Beall et al., 1988), DivIB (a homologue of FtsQ; Harry et al., 1994), FtsL (Daniel et al., 1998), PBP 2B (a homologue of FtsI; Yanouri et al., 1993) and SpoVE (a homologue of FtsW; Ikeda et al., 1989). However, B. subtilis has no identifiable ZipA or FtsN homologue and contains at least one septasomal protein, DivIC (Levin and Losick, 1994), that has no sequence homologue in E. coli. The order of loading of septasomal proteins onto the Z ring in B. subtilis is unclear. As with E. coli, FtsZ initially localizes to the nascent division site followed by FtsA. However, in contrast to the linear order of assembly of the membrane-bound division proteins that occurs in E. coli, it appears that these proteins in B. subtilis (DivIB, DivIC, FtsL and PBP 2B) may be recruited to the division site in a cooperative manner (Daniel et al., 1998; 2000; Daniel and Errington, 2000; Katis et al., 2000; Sievers and Errington, 2000; Feucht et al., 2001). This implies that there are distinct differences in the septation process, or at least in the assembly of the septasome, between the two organisms.

All the B. subtilis septasomal proteins are essential for growth with the exception of DivIB, which is only required for normal cell division at high temperatures (Beall and Lutkenhaus, 1989; Katis et al., 2000). Remarkably, we do not know the function of any of the membrane-bound septasomal proteins with the exception of PBP 2B, which plays a role in septal peptidoglycan synthesis (Daniel et al., 2000). However, the absence of the membrane-bound division proteins in wall-less bacteria suggests that FtsL, DivIC and DivIB also play a role in septum-specific peptidoglycan synthesis or in the regulation of this event (Katis and Wake, 1999; Margolin, 2000).

FtsL, DivIB, DivIC and PBP 2B are simple bitopic membrane proteins containing a small cytoplasmic domain, a single transmembrane domain and a larger extracellular domain. The extracellular domains of both FtsL and DivIC contain putative leucine zipper coiled-coil regions, suggesting a role for these extracellular domains in either self-association or heteroassociations with other septasomal proteins. Consistent with the latter possibility, it was found that depletion of cellular FtsL resulted in a rapid loss of DivIC (but not DivIB) and markedly impaired localization of DivIC (and DivIB) to nascent division sites (Daniel et al., 1998).

Recently, a yeast two-hybrid (Y2H) approach was used to examine the possibility of a direct protein– protein interaction between DivIC and FtsL (Sievers and Errington, 2000). A positive result was observed with the full-length proteins but not when only the extracellular domains were used (Sievers and Errington, 2000). This was surprising, as it implies a key role for the cytoplasmic and/or transmembrane domains in the putative DivIC–FtsL interaction. This is inconsistent with the previous observation that cells still divide normally when both the cytoplasmic and the transmembrane domains of DivIC are replaced with the corresponding domains of the unrelated TolR protein (Katis and Wake, 1999).

As part of our long-term goal to structurally characterize the bacterial septasome, we decided to investigate further the proposed interaction between DivIC and FtsL. We show that FtsL and DivIC are both thermodynamically unstable proteins that are likely to be degraded in vivo unless stabilized by protein– protein interactions. However, we show here using Y2H analysis as well as a variety of biochemical and biophysical assays that there is no direct interaction between B. subtilis DivIC and FtsL, nor is there any direct interaction between the orthologous proteins from the thermophile Bacillus stearothermophilus. Rather, we propose that the stability of DivIC and FtsL is controlled by indirect interactions that are mediated by other septasomal proteins.

Results

FtsL and DivIC are intrinsically unstable

Most experiments reported here used protein constructs corresponding to the entire predicted extracellular do-mains of B. subtilis FtsL (residues 54–117; Sub-FtsL54–117), B. subtilis DivIC (Sub-DivIC58–125), B. stearothermophilus FtsL (Ste-FtsL63–126), and B. stearothermophilus DivIC (Ste-DivIC58–123). An alignment of these domains is given in Fig. 1A; if conservative substitutions are included, the sequence homology between the extracellular domains of the two FtsL and two DivIC orthologues is 58% and 66% respectively. Even though we overproduced these domains in a protease-deficient strain of E. coli (BL21, lon, ompT) and used protease inhibitors during the purification procedure, SDS–PAGE analysis revealed noticeable degradation of all proteins (except B. stearo-thermophilus DivIC) after storage at 4°C for 1 week. Thus, all subsequent experiments were conducted within 1–2 days of protein purification, and SDS–PAGE gels recorded after the experiments revealed no significant protein degradation.

Figure 1.

FtsL and DivIC are thermodynamically unstable.

A. Primary structures of the extracellular domains of B. subtilis (Bsub) and B. stearothermophilus (Bste) FtsL and DivIC. These domains were used for all the studies reported here with the exception of sedimentation equilibrium analysis of B. subtilis DivIC, which used a slightly shorter construct (residues 66–125). The predicted coiled-coil region of these domains is boxed, and the individual residue positions within each heptad repeat are labelled a–g.

B. Far-UV circular dichroic spectra of the extracellular domains of B. subtilis and B. stearothermophilus FtsL and DivIC. Note that only B. stearothermophilus DivIC lacks a dominant random coil signature (minimum at 200 nm).

C. Thermal denaturation profiles of the extracellular domains of B. subtilis and B. stearothermophilus FtsL and DivIC. Each domain is predominantly unfolded at 0°C with the exception of B. stearothermophilus DivIC.

Far-UV circular dichroic (CD) spectra of each of these domains are shown in Fig. 1B. It is immediately apparent that the B. subtilis protein domains as well as Ste-FtsL63–126 are predominantly unstructured, containing large amounts of random coil (which yields a deep minimum at 200 nm) and very small amounts of α-helical secondary structure (minima at 208 and 222 nm) (Woody, 1996). The percentages of α-helical secondary structure calculated from the ellipticities at 222 nm (Pelton and McLean, 2000) were ≈ 6% (Sub-FtsL54–117), ≈ 8% (Sub-DivIC58–125) and ≈ 12% (Ste-FtsL63–126). In marked contrast, the CD spectrum of Ste-DivIC58–123 is characteristic of a protein with significant α-helical secondary structure; the percentage α-helix calculated from the prominent minimum at 222 nm is ≈ 40%.

The high percentage of random coil structure is reflected in the thermal denaturation profiles of these protein domains (see Fig. 1C). The thermal denaturation curve for Sub-FtsL54–117 has a negative slope, which is characteristic of an almost completely random coil peptide (Woody, 1992); we therefore conclude that this protein domain is essentially completely unfolded even at 0°C. The thermal denaturation profiles for both Sub-DivIC58–125 and Ste-FtsL63–126 reveal that they are completely unfolded by the time the temperature reaches 30°C, which indicates that both proteins will be completely unfolded at the optimal growth temperatures for these two organisms. As anticipated from its higher α-helical content, the thermal denaturation profile of Ste-DivIC58–123 reveals that it is more thermodynamically stable than the other three protein domains. However, the melting temperature (Tm, defined as the temperature at which the domains are half unfolded) is still <40°C, and the protein is almost completely unfolded at 60°C, the optimal growth temperature for B. stearothermophilus.

It seems unlikely that the instability of these proteins results from the absence of the cytoplasmic and transmembrane domains as, at least in the case of B. subtilis DivIC, it has been shown that these regions can be replaced without loss of function (Katis and Wake, 1999). Furthermore, in stark contrast to DivIC and FtsL, the extracellular domain of B. stearothermophilus DivIB (which has a very similar bitopic membrane topology to DivIC and FtsL) is very stable (Tm≈ 70°C; S. A. Robson and G. F. King, unpublished data). Thus, we conclude that the extracellular domains of DivIC and FtsL are both thermodynamically highly unstable and are likely to be rapidly degraded in vivo unless stabilized by interaction with other septasomal proteins.

Bacillus subtilis FtsL and DivIC do not self-associate

The extracellular domains of both FtsL and DivIC have predicted coiled-coil regions (see Fig. 1A) that might conceivably be involved in either self-association of these proteins or their interaction with other septasomal proteins to form a heteromeric coiled-coil. During the final stage of purification of these protein domains using size exclusion chromatography (SEC), it was apparent they each eluted with a retention time corresponding to an apparent dimer or trimer (data not shown), consistent with the formation of homodimeric or homotrimeric coiled-coils. However, even short dimeric coiled-coils such as the leucine zipper from the yeast transcription factor GCN4, which contains only four heptad repeats, give CD spectra with characteristically intense minima at both 208 and 222 nm (e.g. see Fig. 2A in O’Shea et al., 1989). Such strong minima are not evident in the CD spectra of the extracellular domains of B. subtilis FtsL and DivIC and, furthermore, the helical contents calculated from the CD spectra of these domains (6–8%; see Eqn 1 in Experimental pro-cedures) are much lower than the values expected (51–55%) if their leucine zipper regions were involved in coiled-coil formation.

Figure 2.

The extracellular domain of B. subtilis DivIC does not self-associate. Bottom: absorbance versus radial distance data obtained from ultracentrifugation of DivIC66–125 in deionized water, pH 6.0. The solid line shows the fit to the data of a model incorporating a single ideal species, which yielded an apparent weight-average molecular weight of 7300 Da. The residuals for this fit (top) are small and randomly distributed.

Estimating molecular masses using SEC is hazardous because the retention time of the protein depends on both its shape and its propensity to interact with the column matrix. If a protein is largely random coil, as is the case for B. subtilis DivIC and FtsL, its average hydrodynamic radius will be much larger and its column elution time earlier than that of a globular protein of equivalent mass. As a result, the apparent mass of the protein estimated from its retention time by comparison with a set of globular SEC standards will always be too high. We therefore examined the self-association properties of the extracellular domains of B. subtilis DivIC and FtsL using two solution-based methods (thus avoiding the potential problem of interactions with a solid support) that are able to estimate molecular mass in a shape-independent fashion.

The oligomeric state of DivIC66–125 was determined using sedimentation equilibrium experiments. The absorbance versus radial distance data obtained for the protein dissolved in deionized water (pH 6.0) were fitted with a model incorporating a single ideal species (Fig. 2, bottom) to yield an apparent weight-average molecular weight of 7300 Da (with 95% confidence limits of 7100 and 7500 Da), which is very close to the predicted monomer weight of 7118 Da. The residuals for this fit (Fig. 2, top) are small and randomly distributed, indicating that the monomer model fits the data well. When a monomer–dimer model was fitted to the data, the calculated proportion of DivIC66–125 in the dimer form was extremely low, and the protein was predicted to be essentially completely monomeric up to the highest concentration tested (630 μM). Similar results were obtained for DivIC66–125 in phosphate and Tris buffer systems, although the protein was not stable in these buffer systems for more than about 36 h (data not shown).

The oligomerization state of Sub-FtsL54–117 was determined using multiangle laser light scattering (MALLS). In combination with on-line measurement of protein concentration using UV absorbance and/or refractive index, MALLS enables determination of the absolute molecular mass of proteins in a shape-independent manner and without reference to standards (Folta-Stogniew and Williams, 1999). MALLS analysis of Sub-FtsL54–117 (injected at a concentration of ≈ 275 μM in 20 mM HEPES, 150 mM NaCl, 1 mM EDTA, pH 8.0) indicated that it was monodisperse with a mass of 7400 Da, which is very close to the predicted monomer molecular weight of 7318 Da.

Thus, we conclude that the extracellular domains of B. subtilis FtsL and DivIC are monomeric. These results are consistent with a previous Y2H analysis (Sievers and Errington, 2000), as well as with our own Y2H results (see below), which show that the full-length proteins do not self-associate.

FtsL and DivIC do not interact in yeast

Previous experiments have shown that only the C-terminal portion of DivIC (DivIC66–125) is required for its function in B. subtilis (Katis and Wake, 1999). For this reason, and to avoid any misfolding/aggregation of these proteins in yeast resulting from the presence of the highly hydrophobic transmembrane domains, we initially tested the C-terminal portions of both proteins, Sub-DivIC66–125 and Sub-FtsL54–117, in a Y2H system in which a protein– protein interaction is essential to confer growth on His-deficient plates (see Experimental procedures). In these experiments, both proteins were tested as fusions to the GAL4 activation and DNA-binding domains. Consistent with the results reported above, no self-association of these protein domains was detected in the Y2H assay. Furthermore, no interaction was detected between the extracellular domains of DivIC and FtsL. During the course of this work, another group reported an interaction between full-length DivIC and FtsL from B. subtilis using a slightly different Y2H system, suggesting an interaction between these proteins in vivo (Sievers and Errington, 2000). We therefore tested fusions containing full-length DivIC or FtsL fused to both GAL4 domains using the Y2H system described here. No His+ colonies could be isolated after the yeast transformation, indicating that B. subtilis FtsL and DivIC do not interact with one another under these conditions.

CD analysis of the putative FtsL–DivIC interaction

The discrepancy between our Y2H results and those reported previously (Sievers and Errington, 2000), as well as the susceptibility of the Y2H system to false positives, prompted us to investigate the putative FtsL–DivIC interaction using more direct and rigorous biophysical and biochemical methods. As the CD spectra of Sub-FtsL54–117 and Sub-DivIC58–125 indicated that both proteins are predominantly random coil, we reasoned that any interaction between these proteins should be accompanied by an increase in structure, which would be reflected in the CD spectrum of the complex. An example of such a structural transition upon complex formation is the dramatic random coil to α-helix transition that occurs when the DNA-binding domains of the transcriptional activators c-Jun, c-Fos and GCN4 bind their cognate enhancer elements (Patel et al., 1990; Weiss et al., 1990).

Figure 3A shows CD spectra of 10 μM solutions of Sub-FtsL54–117 (cyan) and Sub-DivIC58–125 (red) and an equimolar mixture of the two protein domains (both proteins at 10 μM, orange). It can be seen that the CD spectrum of the mixture is not significantly different from the mathematical sum of the spectra of the individual proteins (green). In other words, the structure of neither protein is affected by the presence of the other; both proteins remain predominantly unstructured. Similar results were obtained when the proteins were mixed at concentrations of 18, 50 and 100 μM (data not shown). Thus, the CD experiments argue strongly against an interaction between these two protein domains, as it is difficult to envisage a complex in which both proteins remain predominantly unstructured. Furthermore, it seems very unlikely that the proteins would interact without experiencing any structural perturbations relative to their un-complexed state.

Figure 3.

FtsL and DivIC do not interact directly.

A. Far-UV CD spectra of the extracellular domains of B. subtilis (Bsub) FtsL (cyan) and DivIC (red), as well as an equimolar mixture of the two proteins (orange). The mathematical sum of the spectra of the individual domains (green) is almost identical to the spectrum of the equimolar mixture of the two proteins (orange).

B. Far-UV CD spectra of the extracellular domains of B. stearothermophilus (Bste) FtsL (blue) and DivIC (red), as well as an equimolar mixture of the two proteins (orange). The mathematical sum of the spectra of the individual domains (green) is almost identical to the spectrum of the equimolar mixture of the two proteins (orange).

We performed a similar series of experiments using the B. stearothermophilus protein domains (Fig. 3B). Once again, the CD spectrum of a 10 μM mixture of the two proteins (orange) was not significantly different from the mathematical sum of the spectra of the individual proteins (green), indicating that the structure of neither protein domain is perturbed by the presence of the other. Thus, the CD data argue strongly against a direct interaction between DivIC and FtsL.

Chromatographic analysis of the putative FtsL–DivIC interaction

The predicted isoelectric points of Sub-FtsL54–117 (pI ≈ 8.9) and Sub-DivIC58–125 (pI ≈ 5.5) are markedly different, which enables them to be readily separated using ion exchange chromatography. Under the conditions used here (pH 6.7, see Experimental procedures), Sub-FtsL54–117 was significantly retained on a MonoS cation exchange column (Fig. 4A, centre), whereas Sub-DivIC58–125 eluted in the void (Fig. 4A, bottom). We reasoned that, if there was an interaction between these two protein domains, then the resultant complex would have an intermediate pI (e.g. the predicted pI of a 1:1 complex is 8.0) and therefore should elute with a retention time between that of the individual protein domains. However, application of an equimolar mixture of the two protein domains to the MonoS column yielded a chromatogram (Fig. 4A, top) that only contained peaks at the positions expected for uncomplexed Sub-FtsL54–117 and Sub-DivIC58–125, which argues against an interaction between the two proteins.

Figure 4.

FtsL and DivIC do not interact directly.

A. Chromatograms obtained from applying the extracellular domains of B. subtilis DivIC (bottom), FtsL (middle) and an equimolar mixture of the two domains (top) to a Pharmacia MonoS cation exchange column. FtsL is significantly retained under these experimental conditions, whereas DivIC elutes close to the column void. Note that no additional peaks are apparent in the mixture of the two domains.

B. SDS–PAGE gel analysis of the retained peak (RP) and void peak (VP) from each of the chromatograms shown in (A). Note that the void peak in the equimolar mixture contains only DivIC and the retained peak contains only FtsL.

It is conceivable, however, that a binary complex with a stoichiometry highly biased towards one of the two domains would elute close to the retention time of that domain. Fortunately, despite their very similar molecular weights, Sub-FtsL54–117 and Sub-DivIC58–125 migrate differently during SDS–PAGE (Fig. 4B, lanes 6 and 7) and, hence, we were able to examine this possibility. It is clear from the gel shown in Fig. 4B that the void peak (VP) in the chromatogram of the mixture (see lane 4) contains only DivIC, whereas the retained peak (RP; see lane 5) only contains FtsL. Thus, we can safely conclude that, under these conditions, there is no interaction between the extracellular domains of B. subtilis FtsL and DivIC.

NMR analysis of the putative FtsL–DivIC interaction

The 15N-edited heteronuclear single-quantum coherence (HSQC) spectrum of a uniformly 15N-labelled protein contains a single cross-peak for each backbone amide group as well as pairs of cross-peaks for each Gln and Asn side-chain amide moiety. The 1H and 15N resonance frequencies of these cross-peaks are exquisitely sensitive to the chemical environment of the amide group and can be substantially altered by the proximal binding of even small ligands (Maciejewski et al., 2001). Thus, we reasoned that the HSQC spectrum of 15N-labelled Sub-FtsL54–117 should be markedly altered if it forms a complex with Sub-DivIC58–125, and vice versa.

Figure 5A and B shows the 15N-edited HSQC spectra of uniformly 15N-labelled Sub-FtsL54–117 and Sub-DivIC58–125, respectively, each at a concentration of ≈ 230 μM. As anticipated from the CD experiments reported above, most of the amide–proton chemical shifts are concentrated in the expected range (8.0–8.5 p.p.m.) for largely random coil proteins (Wüthrich, 1986). Figure 5C shows the HSQC spectrum acquired from an equimolar mixture of the labelled protein domains (each domain at ≈ 230 μM). Figure 5D shows an overlay of the HSQC spectra of the individual proteins on the spectrum of the equimolar mixture. It can be seen that the spectrum of the protein mixture is essentially identical to the sum of the spectra of the individual protein domains; there is no difference in the resonance frequencies of any of the cross-peaks in either of the proteins when they are mixed compared with when they are alone. Given the extreme sensitivity of these resonance frequencies to even minor changes in the chemical environment, these data provide very strong evidence that there is not even a transient interaction between the extracellular domains of B. subtilis FtsL and DivIC.

Figure 5.

FtsL and DivIC do not interact directly. Two-dimensional HSQC NMR spectra of 15N-labelled extracellular domains of B. subtilis FtsL (A), DivIC (B) and an equimolar mixture of the two proteins (C).

D. The spectra of the two individual domains have been overlaid on the spectrum of the equimolar mixture, with preservation of colours. The near-perfect superposition of the individual spectra on the spectrum of the mixture indicates that the NMR spectrum of neither protein is perturbed by the presence of the other.

We performed an identical set of experiments with the extracellular domains of the B. stearothermophilus proteins (i.e. Ste-FtsL63–126 and Ste-DivIC58–123) and, again, the spectrum of an equimolar mixture was essentially identical to the sum of the spectra of the individual protein domains (data not shown). Hence, we conclude that, as for the B. subtilis proteins, there is no interaction between the extracellular domains of B. stearothermophilus FtsL and DivIC.

Discussion

DivIC and FtsL do not interact with one another directly

The bacterial division apparatus appears to comprise a macromolecular protein assembly that is responsible for physically dividing a cell into two smaller daughter cells (Chen and Beckwith, 2001). FtsL and DivIC are essential components of this division machinery in B. subtilis. The extracellular domains of both proteins contain leucine zipper motifs that commonly mediate protein–protein interactions. Thus, the fact that depletion of cellular FtsL caused a rapid decline in DivIC levels (Daniel et al., 1998) led to the suggestion that these proteins interact with one another via their leucine zipper regions (Sievers and Errington, 2000).

In this paper, we used several methods to directly examine the possibility that FtsL and DivIC interact with one another in the absence of other septasomal components. In contrast to a previous Y2H result obtained using the B. subtilis full-length proteins (Sievers and Errington, 2000), we were unable to detect any protein–protein interaction between the full-length proteins or between the extracellular domains of DivIC and FtsL of B. subtilis using Y2H analysis. Although we cannot explain this difference, it raises the possibility that the interaction observed in yeast may not be genuine. We therefore used more direct and rigorous methods to further investigate the proposed FtsL–DivIC interaction. Our Y2H results were confirmed using cation exchange chromatography as well as CD and nuclear magnetic resonance (NMR) spectroscopic analyses using the extracellular domains of the B. subtilis proteins. We also used CD and NMR spectroscopy to show that there is no interaction between the extracellular domains of the orthologous proteins from B. stearothermophilus.

A native gel assay was used previously to demonstrate an interaction between the extracellular domains of B. subtilis FtsL and DivIC (Sievers and Errington, 2000); however, we were unable to detect any interaction between these domains using essentially identical experi-mental conditions (data not shown). The reason for the difference between the two studies is unclear, but our native gel results are consistent with the CD, NMR, Y2H and ion exchange measurements reported here that all failed to demonstrate any interaction between FtsL and DivIC.

Thus, we conclude that Bacillus FtsL and DivIC do not interact with one another in the absence of other septasomal proteins. Although we cannot discount the possibility that the full-length proteins behave differently in vivo, our results help to explain a number of previous in vivo observations, including the marked instability of FtsL and DivIC under certain experimental conditions. Most importantly, as outlined below, our conclusion that FtsL and DivIC do not interact directly with one another provides an explanation for the apparently paradoxical non-reciprocal interdependence of the stabilities of the two proteins.

How are FtsL and DivIC stabilized in vivo?

Our results indicate that: (i) both FtsL and DivIC are intrinsically unstable proteins that are likely to be unfolded and therefore targeted for degradation unless stabilized by interactions with other components of the septasome; (ii) this stabilization does not result from a direct interaction between FtsL and DivIC. These conclusions are consistent with several previously reported observations: (i) although FtsL is rapidly degraded in a divIB null at the non-permissive temperature, DivIC is actually more stable under these conditions (Katis et al., 2000). If DivIC were stabilized solely by a direct interaction with FtsL, then DivIC levels would be expected to decrease in concert with the decline in FtsL at the non-permissive temperature; (ii) although FtsL depletion results in a rapid decline in cellular DivIC levels (Daniel et al., 1998), depletion of DivIC (which occurs in the divIC355 strain at 48°C) does not cause such a dramatic loss of FtsL (Daniel and Errington, 2000).

How then are FtsL and DivIC stabilized in vivo if not by direct interaction with one another? It is clear that DivIB plays critical, but distinctly different, roles in determining the stability of FtsL and DivIC. FtsL is markedly destabilized both in the absence of functional DivIB and in a divIB null strain (Daniel and Errington, 2000), suggesting a stabilizing interaction between these two proteins. Consistent with this possibility, the filamentous phenotype of a divIB null at 48°C can be largely suppressed by overexpression of ftsL (Daniel and Errington, 2000). In marked contrast to FtsL, DivIC levels are enhanced in a divIB null strain (Katis et al., 2000), indicating that DivIC is most likely stabilized by interactions with proteins other than FtsL and DivIB. PBP 2B is unlikely to play a role in stabilizing DivIC, as FtsL, DivIC and DivIB levels are unaffected by PBP 2B depletion (Daniel et al., 2000).

It appears impossible to rationalize all the current data concerning the interdependent stabilities and localizations of the known B. subtilis septasomal proteins without invoking the existence of additional cell division proteins that remain to be discovered, one of which is likely to play a role in stabilizing DivIC. One intriguing possibility that might explain some of the seemingly contradictory stability data is that DivIB localizes a protease to the division site that specifically targets DivIC (and possibly other septasomal proteins), but FtsL protects DivIC, indirectly, from the action of the protease. This could occur via a direct interaction between FtsL and the putative protease. Depletion of FtsL would allow the protease to target DivIC, thereby explaining why DivIC is rapidly degraded when FtsL is depleted. In contrast, the protease would not localize to the division site in the absence of DivIB, consistent with the observation that DivIC stability increases when DivIB is depleted. In support of such a model, it was shown recently that the FtsH protease localizes to the mid-cell division site in B. subtilis during vegetative growth and to the polar septum during sporulation, where it is reported to be involved in the degradation of SpoIVFA (Wehrl et al., 2000). Such targeted proteolysis of DivIC might represent an important regulatory step in septasome assembly/disassembly.

How are FtsL and DivIC assembled at the division site?

In E. coli, the septasomal proteins localize sequentially to the nascent division site; there appears to be no co-dependent localization of any of these proteins (Rothfield et al., 1999; Chen and Beckwith, 2001). In striking contrast, there is considerable evidence for interdependent localization of FtsL, PBP 2B, DivIC and DivIB in B. subtilis. First, depletion of cellular FtsL levels leads to severely impaired septal localization of both DivIC and DivIB, even though DivIB levels are unaffected by FtsL depletion (Daniel et al., 1998). Secondly, proper localization of FtsL appears to require DivIB, DivIC and PBP 2B (Daniel and Errington, 2000; Sievers and Errington, 2000). Thirdly, localization of PBP 2B is impaired when FtsL, DivIB or DivIC levels are depleted, and PBP 2B depletion leads to impaired localization of DivIB and DivIC. Finally, it was shown recently that DivIB localization is dependent on DivIC (Katis et al., 2000).

No linear order of septasomal protein localization can be proposed that satisfactorily explains these data. Rather, we propose two possibilities for how the division complex might form in B. subtilis. The first possibility is that all four membrane-bound proteins independently recognize a binding site at mid-cell and assemble simultaneously at this site. The second possibility, and one that we favour, is that these proteins (or at least some of them) are preassembled within the membrane before localization at the division site. The thermodynamic instability reported here for DivIC and FtsL suggests that both proteins are bound directly to other division proteins throughout the cell cycle and therefore are already present in a complex before localization. In support of this suggestion, it is known that DivIC, which we have shown here is intrinsically unstable in the absence of other septasomal proteins, is stable in vivo even under conditions (such as FtsZ depletion) in which it does not localize to the division site (Daniel et al., 1998). That is, DivIC is stable even when not localized to the Z ring, which indicates that it must be stabilized by interactions with other proteins before localization at the division site.

It is clear that a complete appreciation of the mechanism of septasome assembly in B. subtilis will require a much better understanding of the protein–protein interactions that take place between septasomal proteins both before localization at the division site and during assembly of the septasome. In view of this, we are currently attempting to develop some of the biophysical methods outlined here for more rapid screening of interactions between septasomal proteins.

Experimental procedures

Construction of FtsL and DivIC overexpression strains

DNA encoding the extracellular domains of B. subtilis FtsL (residues 54–117; Sub-FtsL54–117) and DivIC (residues 58–125; Sub-DivIC58–125) were polymerase chain reaction (PCR) amplified from a template of B. subtilis strain 168 chromosomal DNA using primers containing 3′BamHI and 5′EcoRI restriction sites for directional cloning of the PCR products into pGEX-2T. The resulting plasmids (pSAR11 and pSAR12) encode Sub-FtsL54–117 and Sub-DivIC58–125, respectively, as in frame fusions to the C-terminus of Schistosoma japonicum glutathione S-transferase (GST), with an intervening thrombin cleavage site between the two protein-coding sequences. The plasmids were transformed into E. coli BL21 cells for IPTG-inducible expression of the GST fusion proteins.

The amino acid sequences of FtsL and DivIC from B. subtilis were used to identify coding sequences for these proteins in the unfinished B. stearothermophilus genome (see http://www.genome.ou.edu/bstearo.html). Sufficient matching contigs were found to allow the design of primers for PCR amplification of the extracellular domains of these proteins. DNA encoding the extracellular domain of B. stearothermophilus FtsL (residues 63–126; Ste-FtsL63–126) was PCR amplified from a template of B. stearothermophilus strain NGB101 chromosomal DNA using primers Ste-L-F (5′-CGTGGATCCAACCAAGTGCGCATTTACG-3′) and Ste-L-R (5′-GGAATTCTCATTCTCATTCCTGCACAACTTTCA CATGG-3′). DNA encoding the extracellular domain of B. stearothermophilus DivIC (residues 58–123; Ste-DivIC58–123) was PCR amplified from the same template using the primers Ste-C-F (5′-CGTGGATTCCAAGCGAA-GGCGATCGATGC-3′) and Ste-C-R (5′-GGAATTCTTATTTTTCCGGCAAGACG-3′). The primers contained 3′BamHI and 5′EcoRI restriction sites for directional cloning of the PCR products into pGEX-2T. The resulting plasmids (pSAR13 and pSAR14) encode Ste-FtsL63–126 and Ste-DivIC58–123, respectively, as in frame fusions to the C-terminus of S. japonicum GST, with an intervening thrombin cleavage site between the two protein-coding sequences. The plasmids were transformed into E. coli BL21 cells for IPTG-inducible expression of the GST fusion proteins.

Purification of FtsL and DivIC for biochemical and biophysical analyses

Escherichia coli BL21 cells harbouring the appropriate FtsL/DivIC overexpression plasmid (pSAR11-14) were grown in 1 l of 2× TY medium containing 100 μg ml−1 ampicillin at 37°C. When required, the proteins were uniformly labelled with 15N by growing cells in a defined minimal media with 15NH4Cl (Isotec) as the sole nitrogen source (Weber et al., 1992). Expression of the fusion protein was induced in mid-log phase by the addition of 100 μM IPTG. Cells were harvested 3 h after induction and stored overnight at –80°C. Harvested cells were thawed and resuspended in a lysis buffer (50 mM Tris, 100 mM NaCl, 2.5 mM EDTA, pH 8.0) to which a protease inhibitor cocktail (Sigma P8465) had been added at the appropriate dilution. Cells were then lysed by three passes through a French press, and the GST fusion protein was purified from the soluble cell fraction using affinity chromatography on GSH–sepharose. The column was equilibrated in thrombin cleavage buffer (TCB; 50 mM Tris, 100 mM NaCl, 2.5 mM CaCl2, pH 8.0), and then the GST fusion protein was cleaved on the column by the addition of 50 U of bovine thrombin (Sigma) followed by incubation at 37°C for 1.5 h with mixing by inversion every 15 min. The cleavage product was eluted from the column with 15 ml of TCB and purified to >98% homogeneity (as judged by SDS–PAGE analysis) using SEC on a Pharmacia Sup-erdex 75 HR 10/30 column (optimal separation in the range 3–70 kDa). SEC was performed at a flow rate of 0.5 ml min−1 using a buffer comprising 20 mM sodium phosphate, 150 mM NaCl, pH 6.7. All biochemical/biophysical experiments were conducted within 1–2 days of protein purification.

Overproduction of DivIC66–125 for analytical ultracentrifugation

The C-terminal portion of DivIC (Sub-DivIC66–125) from B. subtilis was overproduced as a His6-tagged fusion protein using E. coli strain EC512 (see Katis et al., 1997). This fusion protein contains a Factor Xa cleavage site between the His tag and the C-terminal portion of DivIC. Overproduction of this fusion protein was induced as described previously (Katis et al., 1997) and, after centrifugation, cell pellets were stored at –80°C. Cells were resuspended in lysis buffer (50 mM Tris, 100 mM NaCl, 5 mM CaCl2, 5 mM imidazole, pH 8.0), and lysis was completed by two further freeze–thaw cycles. After ultracentrifugation (32 000 r.p.m. in a Beckman 60 Ti rotor at 4°C for 1 h), the supernatant was collected and incubated with 2 ml of Ni-NTA resin (Qiagen) with shaking for 1 h at 4°C. The protein–resin mix was loaded onto a column, washed thoroughly with Factor Xa cleavage buffer (50 mM Tris, 100 mM NaCl, 5 mM CaCl2, pH 8.0) and then resuspended in cleavage buffer (5 ml total volume). On-column cleavage of DivIC66–125 from the His6 tag was performed by incubation with 20 U of Factor Xa (Novagen) for 16 h at room temperature with gentle rocking. The cleaved protein was eluted from the column in cleavage buffer (total volume of 10 ml), and the preparation was cleared (16 000 g for 5 min). Sub-DivIC66–125 was purified from a secondary cleavage product (resulting from cleavage at a non-canonical Factor Xa recognition site present in Sub-DivIC66–125) using reversed phase high-performance liquid chromatography (rpHPLC) on a Vydac C18 analytical column. The protein was eluted using a linear gradient of 10–60% acetonitrile in 0.1% trifluoroacetic acid over 15 min, then the purified protein was lyophilized. CD spectra acquired from the protein before and after rpHPLC were identical, indicating that the protein structure was not affected by the rpHPLC conditions. The mass of the final product determined using mass spectrometry was 7118 Da, which is identical to the theoretical value calculated for Sub-DivIC66–125. The purity of the protein was judged by SDS–PAGE to be >98%.

Yeast two-hybrid analyses

DNA encoding full-length DivIC and FtsL from B. subtilis or the C-terminal domains of these proteins (DivIC 66–125 and FtsL54–117) was cloned into both the gal4AD fusion vector, pGAD10, and the gal4DBD fusion vector, pGBT9 (Clontech). The HF7c strain of Saccharomyces cerevisiae (Clontech) was transfected with prey and bait clones, and interactions between full-length proteins or C-terminal domains were assayed according to the manufacturer’s instructions (Clontech). Transformants were selected on Trp/Leu-deficient media and tested for interactions by patching onto Trp/Leu/His-deficient plates. Growth was monitored over 7 days at 30°C. Controls containing both vectors without DNA inserts were also conducted and were negative for growth during the 7 day period. A positive control using two known interacting transcription factors (BKLF and mCtBP2; Turner and Crossley, 1998) showed growth on Trp/Leu/His-deficient plates after 1 day.

Analytical ultracentrifugation

Sedimentation equilibrium experiments were carried out using a Beckman Optima XL-A analytical ultracentrifuge. Sub-DivIC66–125 was dissolved in deionized water, and the pH was adjusted to 6.0 to give samples with loading concentrations of 630 μM and 315 μM. Sub-DivIC66–125 samples with loading concentrations of 440 μM, 220 μM and 110 μM were also constituted in 50 mM sodium phosphate buffer con-taining 100 mM NaCl, pH 6.0, and in 50 mM Tris containing 100 mM NaCl, pH 8.0. Data were recorded at 20°C using six-channel centrepieces in an An-60Ti rotor at 25 000 r.p.m., 30 000 r.p.m. and 42 000 r.p.m. Data were acquired as scans of absorbance (280 nm and 360 nm) versus radius (0.001 cm increments). Scans were collected at 3 h intervals, and the samples were judged to have attained both sedimentation and chemical equilibrium when successive scans were indistinguishable by eye. After subtracting the 360 nm scans from the 280 nm scans, the data for the scans that had reached sedimentation equilibrium for each buffer type were fitted to single-species and homoassociation models using the program NONLIN (Johnson et al., 1981). The best model and final parameters were determined by examination of the residuals for each fit. The partial specific volume of Sub-DivIC66–125 was determined from its amino acid sequence (Perkins, 1986), and the density of the solvents was taken to be 1.00379 g ml−1 for the Tris buffer and 1.00715 g ml−1 for the phosphate buffer (20°C). These parameters were calculated using the program SEDNTERP (Hayes, 1995–2001).

Multiangle laser light scattering

Analysis of Sub-FtsL54–117 using multiangle laser light scattering (Wen et al., 1996) was performed at the HHMI and W. M. Keck Foundation Biotechnology Resource Laboratory at Yale University, New Haven, CT, USA, as described previously (Folta-Stogniew and Williams, 1999) except that SEC was performed on a Superdex 75 HR 10/30 column using 20 mM HEPES, 150 mM NaCl, 1 mM EDTA, pH 8.0, as the elution buffer. Weight-average molecular weights were determined using the Debye fitting method (ASTRA software package; Wyatt Technology) as described previously (Folta-Stogniew and Williams, 1999).

Circular dichroism

Far-UV CD spectra were acquired using a Jasco J-710 spectropolarimeter. Proteins were first dialysed into a 1 mM sodium phosphate buffer containing 50 mM NaF. Final spectra were the average of eight transients collected at 10°C in a 0.1 cm rectangular quartz cell using a scan rate of 20 nm min−1. A spectrum of the dialysis buffer collected using identical acquisition parameters was subtracted from the spectrum of each protein. The percentage of α-helical secondary structure was calculated from the mean residue ellipticity at 222 nm ([θ]222), using the following equation (Pelton and McLean, 2000):

image

CD thermal denaturation curves were acquired as described previously (Bains et al., 1999) by monitoring the ellipticity at 222 nm as a function of temperature. Data were recorded every 1°C as the temperature was increased from 4°C to 90°C at a rate of 1°C min−1.

Cation exchange chromatography

Cation exchange chromatography was performed using a Pharmacia MonoS HR 5/5 column. Buffer A was 20 mM sodium phosphate, pH 6.7, whereas buffer B additionally contained 2 M NaCl. Chromatography was performed at a flow rate of 1 ml min−1 using isocratic conditions (100% buffer A) for the first 6 min, followed by a linear gradient of 0–20% buffer B over the next 12 min. Under these conditions, Sub-FtsL54–117 elutes at ≈ 15% buffer B (300 mM NaCl), whereas Sub-DivIC58–125 elutes in the void. Both void and retained peaks were collected, lyophilized and analysed using SDS–PAGE. Proteins were injected individually at a concentration of 50 μM or as an equimolar mixture (each 50 μM).

NMR spectroscopy

NMR experiments were performed at 25°C on a Varian INOVA 600 MHz spectrometer using uniformly 15N-labelled Sub-FtsL54–117, Sub-DivIC58–125, Ste-FtsL63–126 and Ste-DivIC58–123. The labelled proteins were concentrated to 500 μM in NMR buffer (20 mM sodium phosphate, 100 mM NaCl, pH 6.8) using 3 kDa cut-off centrifugal concentrators (Amicon). The pH was checked and adjusted to 6.8 where necessary; then the samples were clarified using 0.45 μm microfuge filters (Gelman Sciences). Samples of individual proteins were obtained by adding 125 μl of protein solution, 125 μl of NMR buffer and 20 μl of 99.96% D2O to a susceptibility-matched microcell (Shigemi) to give a final pro-tein concentration of ≈ 230 μM. Equimolar mixtures (≈ 230 μM of each protein) were prepared by adding 125 μl of Sub-FtsL54–117 and 125 μl of Sub-DivIC58–125 (or the corresponding pair of B. stearothermophilus FtsL and DivIC extracellular domains) to 20 μl of 99.96% D2O in a susceptibility-matched microcell (Shigemi). 15N-edited, sensitivity-enhanced two-dimensional HSQC spectra were acquired as described previously (King et al., 1999). NMR data were processed with nmrPIPE (Delaglio et al., 1995) using a script generator available on the UConn Health Center Structural Biology website (http://sbtools.uchc.edu/nmr/).

Acknowledgements

We thank Dr B. Setlow for the gift of B. stearothermophilus NGB101 chromosomal DNA, Dr E. Folta-Stogniew for MALLS analysis of Sub-FtsL54–117, and Dr Bruce Roe for providing sequences of B. stearothermophilus contigs before official release.

Ancillary