Dissecting the functional contributions of the intrinsically disordered C-terminal tail of B. subtilis FtsZ

Abstract

FtsZ is a bacterial GTPase that is central to the spatial and temporal control of cell division. It is a filament-forming enzyme that encompasses a well-folded core domain and a disordered C-terminal tail (CTT). The CTT is essential for ensuring proper assembly of the cytokinetic ring and its deletion leads to mis-localization of FtsZ, aberrant assembly, and cell death. In this work, we dissect the contributions of modules within the disordered CTT to assembly and enzymatic activity of B. subtilis FtsZ (Bs-FtsZ). The CTT features a hypervariable C-terminal linker (CTL) and a conserved C-terminal peptide (CTP). Our in vitro studies show that the CTL weakens the driving forces for forming single-stranded active polymers and suppresses lateral associations of these polymers whereas the CTP promotes the formation of alternative assemblies. Accordingly, in full-length Bs-FtsZ the CTL acts as a spacer that spatially separates the CTP sticker from the core thus ensuring filament formation through core-driven polymerization and lateral associations through CTP-mediated interactions. We also find that the CTL weakens GTP binding while enhancing the catalytic rate whereas the CTP has opposite effects. The joint contributions of the CTL and CTP make Bs-FtsZ an enzyme that is only half as efficient as a truncated version that lacks the CTT. Overall, our data suggest that the CTT acts as an auto-regulator of Bs-FtsZ assembly and as an auto-inhibitor enzymatic activity. Based on our results, we propose hypotheses regarding the hypervariability of CTLs and compare FtsZs to other bacterial proteins with tethered IDRs.

Graphical Abstract

Introduction

Cell division in rod-shaped bacteria is a complex process that begins with assembly of the tubulin-like cell division protein FtsZ at the site of cell division [1]. In vivo studies focused on Bacillus subtilis and Escherichia coli have shown that FtsZ, which is a GTPase, is responsible for recruiting ~20 additional proteins that together orchestrate assembly of the division septum [2–9]. Because FtsZ serves as the foundation on which the rest of the division machinery is assembled, the machinery itself is frequently referred to as the “Z-ring” [1]. As the first protein at the nascent division site, FtsZ is a target for factors controlling the timing, formation, and location of the Z-ring via direct or indirect interactions with FtsZ [3, 10]. Recent efforts have uncovered modulatory proteins that: (1) prevent Z-ring assembly at aberrant locations (via the functions of nucleoid occlusion proteins such as EzrA and Noc in B. subtilis, SlmA in E. coli, and the Min proteins in both) [11–14]; (2) tether the Z-ring to the membrane (E. coli ZipA and FtsA in both E. coli and B. subtilis) [15, 16]; and (3) facilitate the assembly of FtsZ into higher-order structures (the Zap proteins in E. coli and B. subtilis) [10, 17]. Although the key roles of many of the essential proteins are yet to be fully elucidated, it is clear that FtsZ is a central player in the divisome and that FtsZ-mediated assembly of the cytokinetic ring is a tightly regulated process [18].

The formation of Z-rings is governed, in part, by the intrinsic ability of FtsZ to hydrolyze GTP and self-assemble into linear, single-stranded polymers that laterally associate to form higher-order structures [19–23]. Key aspects of the assembly process have been uncovered through systematic structural and mechanistic studies performed in vitro and in vivo focusing predominantly on FtsZs from E. coli and B. subtilis, and more recently Caulobcater crescentus [3, 5, 6, 24–28]. Monomers of FtsZ bind to GTP and this facilitates the formation of single-stranded linear polymers that are also referred to as protofilaments [29, 30]. The active site of FtsZ is formed at the interface between a pair of FtsZ monomers, and this implies that the dimer is the minimal active form of the enzyme [29, 31]. GTP hydrolysis drives the exchange of monomeric subunits thereby controlling the length distribution and dynamics of FtsZ polymers [9, 32–35]. GTP-dependent assembly of FtsZ in vitro depends on a variety of factors including solution conditions and FtsZ concentration [36–39].

FtsZ is a prominent archetype of filament-forming enzymes [40], which are also referred to as living polymers in the physical literature [41, 42]. The assembly of filament-forming enzymes often requires the crossing of a threshold concentration to facilitate the formation of active filaments. Previous studies identified the presence of at least two threshold concentrations for GTP-dependent assembly of FtsZ, referred to hereafter as c_A and c_B, that are in the micromolar range [20, 21, 43–46]. Increasing the concentration above the threshold concentration c_A is required for the formation of active protofilaments, which are single-stranded polymers. Crossing of the threshold concentration c_B > c_A leads to the formation of higher-order assemblies, that appear to be characterized by the lateral association of linear polymers [47]. From a thermodynamic standpoint, the existence of threshold concentrations implies that both the formation of single-stranded polymers and the lateral association of these polymers are cooperative processes [41].

Considerable effort has been invested into uncovering the determinants of cooperativity in FtsZ assembly [20, 21, 43–46, 48, 49]. These efforts have been motivated in part by the observation that dimers of FtsZ are thought to be the cooperative units that control the assembly of single-stranded linear polymers [45] and yet the nucleus for polymerization appears to be independent of the binding of GTP and of GTP hydrolysis [49]. Additionally, there is the notion that the formation of single-stranded polymers cannot be governed by a threshold concentration [50]. However, this view ignores several nuances whereby conformational transitions and / or anisotropic interactions among subunits, especially in ligand-dependent systems such as FtsZ can give rise to a threshold concentration for the formation of single-stranded protofilaments [40–42, 51–55].

FtsZ has a tripartite architecture consisting of a globular catalytic domain and an intrinsically disordered C-terminal tail (CTT) that encompasses two modules viz., a 40–300 residue hypervariable linker (CTL) and a highly conserved 10–20 amino acid peptide at the extreme C-terminus called the C-terminal peptide (CTP) [50]. Figure 1A shows a schematic of this architecture using FtsZ from B. subtilis (Bs-FtsZ) as an archetype. Recent studies have established the in vivo importance of CTTs in B. subtilis, E. coli, and C. crescentus [3, 47, 50, 56–62]. In all FtsZs studied to date, the CTT is not required for forming linear polymers or hydrolyzing GTP [56, 58, 62]. However, there are clear morphological differences in the assemblies that are formed and the bacterial phenotypes that are observed for cells expressing wild type FtsZs vs. variants where the CTT has been deleted [56, 58, 62]. Deletion of the CTL reduces GTP hydrolysis and subunit exchange, inhibits cell division, and eventually causes cell death via a filamenting phenotype [50, 57, 58, 61]. Systematic in vitro studies have also begun to uncover the mechanisms by which CTTs influence FtsZ assembly [3, 47, 50, 56–62]. Specifically, in vitro studies show that the CTP coordinates the lateral associations among protofilaments and serves as a binding motif for modulatory proteins that regulate FtsZ assembly in vivo [47, 63–65].

Figure 1: Architecture of FtsZ includes three distinct domains.

(A) The Bacillus subtilis FtsZ (Bs-FtsZ) core (residues 1–315) is represented using a pink ribbon structure rendering of the model deposited in the protein data bank (PDB id: 2vxy) and rendered here using PyMOL (The PyMOL Molecular Graphics System, Version 1.2r3pre, Schrödinger, LLC). The 68-residue CTT of Bs-FtsZ spans residues 315 – 383 and is depicted using 50 green beads for the CTL and 18 cyan beads for the CTP. (B) Schematics showing the block architectures of the four constructs viz., WT, ΔCTT, ΔCTL, and ΔCTP.

In order to understand why the disordered CTT is essential for FtsZs in vivo, we focus here on in vitro biophysical and biochemical investigations to uncover the functions of each of the FtsZ modules / domains using FtsZ from B. subtilis (Bs-FtsZ) as an archetypal system. Our studies, which build on previous foundational contributions [3, 47, 50, 56–62], are directed at four different constructs viz., full-length Bs-FtsZ (designated as WT), ΔCTT which lacks the C-terminal tail, ΔCTL which lacks the C-terminal linker, and ΔCTP which lacks the C-terminal peptide (Figure 1B). For each of these constructs, we quantify the concentration dependence and morphologies of different assemblies, the apparent affinity for GTP, and the efficiency of GTP hydrolysis. Taken together, the results from our systematic experiments provide clear insights regarding the contributions made by each of the modules of Bs-FtsZ in controlling assembly, cooperativity, and regulating GTPase activity.

RESULTS

The CTL and CTP modules impact Bs-FtsZ assembly differently

In the presence of GTP, FtsZ can form single-stranded linear polymers that associate to form higher-order assemblies. The sizes and morphologies of assemblies are likely to be heterogeneous and dependent on FtsZ concentration. We used dynamic light scattering (DLS), a correlation spectroscopy method [66, 67], to uncover the size distributions of assemblies formed by FtsZ as a function of protein concentration in an MES reaction buffer (50 mM MES, 50 mM KCl, 2.5 mM MgCl₂, 1 mM EGTA, 1 mM GTP, and pH 6.5). The time correlation functions from raw scattering data yielded distributions of scattering intensities that were converted to number density distributions. These distributions quantify the number densities of hydrodynamic diameters (D_H) that are observed in solution. The conversion between intensity and number distributions is obtained using an instrument-specific algorithm based on Mie’s theory for scattering [67].

First, we analyzed the size distributions of assemblies that are formed by the WT construct, as a function of protein concentration. Data for the number density distributions are shown in Figure 2A. At a protein concentration of 2 μM, we observe a bimodal distribution. We interpret this bimodality to imply the presence of a concentration threshold of ~ 2μM where we observe the coexistence of two different types of assemblies viz., 10 nm sized assemblies that are most likely protofilaments that coexist with higher-order assemblies that are likely to be bundles of filaments and / or other types of structures. The DLS data are further summarized by integrating the number distribution to quantify the populations associated with assemblies that correspond to specific size intervals. Although number density distributions are complete representations of the DLS data, they are probability density functions (PDFs). To enable comparative assessments across different solution conditions and constructs, we summarize the DLS number density distributions as interval-specific cumulative distribution functions (CDFs), which are integrals of PDFs i.e., areas under the number density distributions calculated across specific intervals. CDFs are optimal choices because they quantify the total probability rather than the probability density. The CDFs, calculated as integrals over the PDFs across specific intervals, are displayed in the form of heatmaps, where the color within each box quantifies the total percent probability of observing scatterers within the size interval corresponding to the box in question (Figure 2B). The advantages of CDFs over PDFs have been well established [68], especially for comparative assessments across different distributions. We use these as devices that enable comparative assessments of the different observed assembly sizes at various protein concentrations.

Figure 2: The CTL and CTP differently impact FtsZ assembly.

(A) Dynamic light scattering (DLS) measurements are used to estimate the fraction of the total number of scatterers that are of a specific apparent diameter (D_H). The percent number density is plotted as a function of D_H for Bs-FtsZ concentrations of 0.5, 1, 2, 4, 6, 8, and 10 μM. This analysis shows that the presence of a threshold concentration of 2 μM is evident in the bimodality of the population distribution (bolded cyan curve). (B) The number density function inferred from DLS data are integrated and represented using a checkerboard heat map, where each box shows the percent occupancy for species that lie within specific intervals of D_H values (ordinate) for a specific protein concentration (abscissa). Boxes shown in white correspond to intervals with zero occupancy. (C) Transmission electron microscopy (TEM) indicates the formation of higher-order assemblies such as multi-filament bundles and wreath-like structures at 6 μM FtsZ. (D) For ΔCTT, the number density probabilities as a function of D_H shows that most structures fall between 10¹ and 10² nm. There is a bimodality in the distribution at 8 μM (bolded orange curve) (E) Checkerboard heatmap of the ΔCTT D_H distribution as measured by DLS. The particle size is enhanced at low concentrations compared to WT and largely maintains size with an increase in concentration (F) TEM reveals that these sizes correspond to long, linear, and curved protofilaments. (G) Number density (%) vs. D_H for various ΔCTP concentrations. No significant bimodalities are observed. (H) Heatmap of the ΔCTP DLS data shows a low instance of large assemblies and mostly small structures at least an order of magnitude smaller than the structures formed by the ΔCTT variant. (I) This assembly is predominantly short single-stranded protofilaments, as observed with TEM. (J) Plot of the number density as a function of D_H for various ΔCTL concentrations. The populations are more heterogenous than ΔCTT, and a bimodality is observed at both 8 μM and 10 μM (K) DLS measurements of ΔCTL shows that the distributions at each protein concentration more closely match ΔCTT than that of WT. However, enhanced assemblies are shown at lower concentrations. At high concentrations, these assemblies are still smaller than those observed by WT (L) TEM shows that the higher-order assemblies formed by the ΔCTL variant are rings that are visibly smaller than WT rings. These are referred to as mini rings.

Both Figure 2A and 2B show evidence for a threshold protein concentration of 2 μM beyond which the assemblies that form in solution are larger than 100 nm. Transmission electron microscopy (TEM) images obtained at a concentration of 6 μM, a representative of which is shown in Figure 2C, show the bundling of filaments into wreath-like structures that are larger than 100 nm in diameter. In contrast, the assemblies that form at concentrations below 2 μM are smaller than 100 nm. The concentration dependent DLS data for WT Bs-FtsZ point to the existence of a threshold concentration, c_B ≈ 2 μM for the formation of higher-order assemblies of filaments in the solution conditions deployed in our experiments.

Next, we asked how the deletion of the CTT, the CTP, and the CTL impact the assemblies formed under conditions that were identical to those used for the study of the WT. Figures 2D and 2E summarize the DLS data for ΔCTT. At low protein concentrations, 0.5 μM – 2 μM, ΔCTT forms assemblies that are larger than those observed for WT. The concentration at which we observe a bimodal distribution for the number density of scatterers shifts upward to 8 μM vis-à-vis the threshold of 2 μM for the WT. This is likely similar to previous observations of the condensation of ΔCTT filaments in the presence of crowders and stabilizers [23, 56, 58]. Our data suggest that while ΔCTT forms larger filaments when compared to WT, the absence of the CTT weakens its ability to form higher-order assemblies. Support for this inference comes from direct visualization using TEM, which shows the formation of long, single-stranded filaments and the absence of bundles or wreaths at a protein concentration of 6 μM (Figure 2F). These morphologies are distinct from those observed for WT under identical solution conditions (Figure 2C).

In order to understand why ΔCTT is able to form long, single-stranded polymers more robustly than WT while showing a weakened ability to form higher-order assemblies, we analyzed the DLS and TEM data for ΔCTP and ΔCTL. These constructs help us separate contributions made by the CTL and the CTP to the overall behavior of WT. Surprisingly, we observe a significant diminution in the sizes of assemblies formed by ΔCTP. It appears that the disordered CTL suppresses assembly on all length scales, leading to the formation of assemblies that are at least an order of magnitude smaller than ΔCTT across protein concentrations ranging from 0.5 μM – 8 μM (Figures 2G and 2H). Unlike the WT and ΔCTT constructs, we do not observe the onset of bimodality in any of the number density distributions (Figure 2G). The DLS data suggest that, in the concentration range we study here, ΔCTP likely forms short linear polymers that do not form bundles or other higher-order assemblies. This conjecture is confirmed by TEM analysis, performed at a ΔCTP concentration of 6 μM, which is identical to the concentrations used for WT and ΔCTT. The TEM image shows the formation of significantly shortened linear polymers and an absence of higher-order assemblies (Figure 2I). We conclude, in accord with previous findings of Huecas et al. [56], that the CTL acts as an excluded volume spacer that substantially weakens the formation of single-stranded polymers and inhibits lateral associations.

Finally, we studied the assemblies and morphologies formed by ΔCTL, the construct lacking the C-terminal linker (Figures 2J–L). In the concentration range between 0.5 μM and 6 μM we observe a clear upward shift in the sizes of assemblies formed by ΔCTL when compared to those of ΔCTT and ΔCTP. TEM analysis, performed at a ΔCTL concentration of 6 μM, shows the formation of alternate morphologies. We refer to these morphologies as mini rings because they appear to be miniature versions of the wreath / ring-like structures that we observe for the WT under similar conditions. Direct comparisons of the DLS data for WT and ΔCTL show that the mini rings are at least an order of magnitude smaller in size than the structures formed by WT. The number density distributions become bimodal and even multi-modal above protein concentrations of 8 μM and this points to association or condensation of the mini ring structures. The DLS data and TEM analysis suggest that the CTP module is akin to a sticker which engages in cohesive intramolecular interactions to drive the formation of alternative assemblies. The implication is that in the WT, the CTL provides spatial separation between the core and the CTP that enables the formation of larger assemblies by facilitating polymer formation through core-mediated interactions and lateral associations via CTP-mediated interactions.

Taken together, the totality of data presented in Figures 2A–L lead to a coherent picture for how the different modules / domains of FtsZ contribute to assembly of Bs-FtsZ. Of the four constructs we studied, the core (see data for ΔCTT) forms the longest single-stranded polymers. Absence of the CTT weakens the ability of filaments to form higher-order structures. Our data suggest that the CTP and CTL modules may be thought of as stickers and spacers, respectively [56, 69–71]. The CTP sticker provides cohesive interactions and appears to engage in interactions with the core and with itself thereby giving rise to mini ring structures that are fundamentally different from the single-stranded polymers formed by ΔCTT (compare Figures 2F and 2L). The observation of bi- and multimodality in the DLS data suggests that the CTP can also promote lateral associations and lead to the formation of higher order assemblies. However, without the CTL providing the necessary spacing between the core and the CTP, the higher-order assemblies formed by ΔCTL are smaller than those formed by WT. In direct contrast, the CTL acts as a spacer that carves out an excluded volume thus ensuring spatial separation of the CTP sticker from the core and suppressing assembly on all length scales [56].

Increased salt concentration weakens the formation of higher order assemblies for WT Bs-FtsZ

Previous work showed that a positively charged CTP is necessary to drive the bunding of FtsZ protofilaments [47]. Studies have also shown that increases to the concentration of monovalent salts can reduce the extent of FtsZ assembly [36, 47]. Therefore we hypothesize that CTP-mediated interactions that are spatially separated from the core are likely to be electrostatic in nature and that bundling must be influenced by alterations to these interactions. To test this hypothesis, we analyzed the salt dependence of higher order assembly formation by WT Bs-FtsZ and comparing these results to the variant that lacks the CTP.

We measured the size distribution of FtsZ assemblies as a function of increased KCl concentration for WT Bs-FtsZ. These measurements were performed in otherwise consistent MES reaction buffer at 6 μM WT Bs-FtsZ. Under these conditions, WT Bs-FtsZ forms large wreath-like structures (Figure 2C). As the concentration of KCl in the buffer increases, we observe a concomitant reduction in the sizes of assemblies formed by WT Bs-FtsZ. This is summarized in the number density distributions extracted from DLS measurements (Figure 3A) and the integrals of these distributions presented as checkerboard heatmaps (Figure 3B). Sizes of the dominant assemblies in solution decrease by two orders of magnitude as the salt concentration increases from 50 mM KCl to 200 mM KCl (Figure 3B). The size distributions at relatively high salt are similar to those formed by ΔCTP at similar protein concentrations. Comparison of TEM images, collected using material extracted from solutions with 50 mM KCl (Figure 2C) versus 200 mM KCl (Figure 3C) shows that higher order structures are disassembled at high salt and the structures that persist are short single-stranded polymers. In fact, the structures observed in the presence of 200 mM KCl are most similar in morphology to those observed for ΔCTP (Figure 2I). The simplest interpretation of data shown in Figure 3 is that lateral associations among protofilaments can give rise to bundled higher-order assemblies and these are promoted primarily by electrostatic interactions, mediated mainly by the CTP sticker. These interactions are screened at high salt thereby unmasking the assembly inhibiting effects of the CTL spacer.

Figure 3: Increasing the KCl concentration screens interactions mediated by the CTP.

(A) DLS measurements are used to estimate the percentage of the total number of particles that are of a specific apparent diameter (D_H) for KCl concentrations of 50, 75, 100, 150, and 200 mM. (B) Heatmap of the DLS data for different salt concentrations shows that the apparent size of the assemblies decreases with increasing salt concentration. FtsZ assembly size decreases by over an order of magnitude as the salt concentration is increased to 200 mM. The size of the assemblies formed in 150 and 200 mM KCl is similar to the size formed by ΔCTP at the same concentration of 6 μM. (C) In high salt (200 mM KCl), TEM images show the formation of single-stranded protofilaments.

Our data thus far suggest that within the WT Bs-FtsZ, core-driven interactions give rise to single-stranded linear polymers. These interactions are weakened by the CTL which also spatially separates the core from CTP-mediated electrostatic interactions. This spatial separation appears to help drive interactions that lead to the formation of wreath and ring-like structures. We next asked how the distinct domains in the tail contribute to the formation of active polymers.

The CTL weakens the driving forces for forming active polymers

FtsZ is a filament-forming enzyme and the formation of active polymers requires the crossing of an apparent concentration threshold (c_A). We refrain from referring to this as a critical concentration because this parlance has precise implications for concentration fluctuations at critical points in systems that undergo first-order and continuous phase transitions [41, 72]. The presence of an apparent threshold concentration is the hallmark of a cooperative transition [72]. Lower apparent threshold concentrations for forming active polymers imply stronger driving forces for forming these polymers. By estimating the values of c_A for each of the four constructs, we investigated the effects of the CTT and modules within the CTT on the strength of driving forces for cooperative assembly into active polymers. It is difficult to estimate the value of c_A using DLS. This is because the scattering intensity decreases as the concentration is lowered leading to significant reduction of signal vs. noise. Additionally, the use of purely spectrophotometric methods based on spinning samples down and separating them into supernatant and pellet are confounded by the lack of tyrosine and tryptophan residues in Bs-FtsZ requires either amino acid substitutions or the incorporation of fluorescent dyes at suitable positions that do not disrupt WT Bs-FtsZ behavior. These are non-trivial modifications that can confound interpretations, especially given the subtle interplay between CTT modules and the core domain. Therefore, we leveraged extant methods in the FtsZ literature and used the onset of activity as a function of FtsZ concentration as a proxy for quantifying the apparent concentration threshold for the onset of active polymers [38]. Specifically, we measured the GTPase activity (millimoles of GTP hydrolyzed per minute) as a function of FtsZ concentration. The intercept along the abscissa of a linear fit to the activity data is used as an estimate of c_A. This is valid because GTP binding and dimer formation are obligatory for polymer formation and GTPase activity. In our analysis, we only fit points that have a GTPase activity value greater than zero, taking into account the error in our measurements. Our analysis yielded an estimate for c_A to be 0.92 ± 0.12 μM for Bs-FtsZ in 50 mM KCl, which is consistent with estimates from previous studies showing that c_A lies between 0.5 and 1 μM [20, 21, 43–46].

Figure 4B shows our data for the GTPase activity of the ΔCTT construct. Our analysis of these data yields an estimate of 0.21± 0.18 μM for ΔCTT. The lower value of c_A for ΔCTT vis-à-vis WT and its ability to assemble into long single-stranded polymers suggests that active polymers form via a cooperative process that is driven mainly by interactions between Bs-FtsZ cores. Our analysis of the GTPase data for the ΔCTP construct yields an estimate of ~ 1.76 ± 0.11 μM for c_A (Figure 4C). This is almost twice the value of c_A for WT Bs-FtsZ and an order of magnitude higher than that of ΔCTT. These data indicate that the CTL spacer weakens the driving forces for the assembly of single-stranded active polymers whilst maintaining the cooperative nature of this assembly process. Finally, analysis of the activity data for ΔCTL (Figure 4D) yields an estimate of 0.67 ± 0.07 μM for c_A, and this value is slightly less than, that of the WT construct. Overall, our estimates for c_A follow the trend whereby c_A(ΔCTT) < c_A(ΔCTL) < c_A(WT) < c_A(ΔCTP). These data show that as a whole, the CTT weakens the driving forces for forming active polymers. This weakening of GTPase activity and the driving forces for forming active polymers derives exclusively from the CTL spacer and is overcome by the CTP sticker, although not enough to lower the value of c_A to be equivalent to that of ΔCTT.

Figure 4: The modules of the tail impact the concentration of Bs-FtsZ required to form active polymers (c_A).

The onset of activity pinpoints the location of the threshold concentration for the formation of active polymers. The FtsZ concentration vs. GTPase activity data were fit using a linear regression model, and the x-intercept or inflection point shows the location of c_A (A) Measurement of GTPase activity (millimoles of GTP hydrolyzed per minute) as a function of WT concentration. GTPase activity is measured using a regenerative-coupled GTPase assay. The c_A is 0.77 ± 0.23 μM. (B) GTPase activity as a function of ΔCTT concentration. The intercept along the abscissa for the formation of active polymers has an error of 0.33 μM, and the estimated c_A appears to lie between 0 and 0.25 μM. (C) GTPase activity as a function of ΔCTL concentration, which has an apparent c_A of 0.52 ± 0.11 μM. (D) GTPase activity as a function of ΔCTP concentration shows the clearest evidence for an apparent threshold concentration (c_A) of 1.23 ± 0.08 μM.

The CTL and CTP have opposing effects on the enzymatic activity of Bs-FtsZ

The Michaelis-Menten formalism [73] has been used as a minimal model for analyzing enzyme kinetics of FtsZs [38, 74]. However, FtsZ is a filament-forming enzyme and its overall activity will convolve contributions from the full range of polymers and higher order assemblies. Extant methods do not allow us to parse the species-specific contributions to enzyme kinetics. Accordingly, we analyzed measurements of enzyme kinetics, for each of the four constructs, using two minimal models viz., the Michaelis-Menten model [73] and the Hill model [75], respectively. In general, the data for enzyme kinetics measured for all constructs at two different concentrations (4 μM and 6 μM) are optimally fit using the Michaelis-Menten model (Figures S1A–D) when compared to the Hill equation (Figures S2A–D). It is worth noting that while the data from measurements of GTPase activity vs. GTP concentration are shown to obey Michaelis-Menten kinetics for all four constructs (Figures S1A–D), the inferred value of the apparent Michaelis constant K_M viz., the concentration of GTP at which 50% of the sites from of all active species in the ensemble are occupied increases with FtsZ concentration for WT (Figure 5A). This reflects the complexities associated with analyzing enzyme kinetics using a simple model that does not account for species-specific contributions from different types of assemblies formed by the WT enzyme. Our analysis in Figure 2 shows that the WT enzyme forms a mixture of single-stranded polymers and higher order structures at concentrations of 4 μM and 6 μM. At these concentrations, ΔCTT, ΔCTP, and ΔCTL higher-order assembly is either minimal or non-existent. Accordingly, for each ΔCTT, ΔCTP, and ΔCTL the apparent K_M values we obtain within error do not show a dependence on enzyme concentration.

Figure 5: The CTT functions as an auto-inhibitor.

(A) Bar plots of the apparent Michaelis-Menten constants (K_M) for each variant at 4 and 6 μM enzyme show that the ΔCTP enzyme has the highest apparent K_M. (B) Bar plots of the catalytic rates (k_cat) for each variant at 4 and 6 μM. These parameters show that the CTP slows the catalytic rate (ΔCTL data) in comparison to the ΔCTT enzyme. (C) Bar plots of the catalytic efficiency (k_eff) for each variant at 4 and 6 μM.

With the caveat that the enzyme kinetics for WT likely reflect a convolution of multiple species, we compared the construct-specific values for K_M and v_max and used these to quantify enzyme efficiencies. Overall, our measurements show that FtsZs that lack the CTL, as in ΔCTL and ΔCTT, have the lowest apparent K_M values (Figure 5A). Addition of the CTL, which suppresses polymerization and inhibits higher order assembly, leads to an increase in apparent K_M (see data for ΔCTP). For the WT, increased bundling increases the apparent K_M. A parsimonious interpretation of the inferred values for the apparent K_M is that the CTL weakens GTP binding whereas the CTP enhances GTP binding.

The catalytic rate constant (k_cat) or turnover number calculated as the ratio of v_max to the total concentration of the enzyme, is higher for ΔCTT when compared to WT (Figure 5B). Since the rate of GTP hydrolysis is coupled to the rate of subunit turnover, this likely corresponds to an enhanced rate of subunit exchange [32, 33]. We find that the CTP reduces the catalytic rate of the core, as shown by data obtained for ΔCTL, indicating that the subunit exchange is slowed by the CTP. Data for the ΔCTP construct suggest that having only the CTL and not the CTP restores, and even slightly enhances, the catalytic rate of the core alone. The WT construct falls in the middle of these two extremes, indicating that the reduced subunit exchange due to the presence of the CTP sticker can be offset by the presence of the CTL spacer.

Our data suggest that, the modules of the CTT contribute to an auto-inhibitory function as far as overall enzyme activity is concerned. Indeed, the ΔCTT is the most efficient enzyme, where efficiency (k_eff) is quantified by the ratio k_cat/K_M (Figure 5C) at both concentrations (4 μM and 6 μM). While the k_cat of ΔCTL is less than that of ΔCTP, the opposite is true for the apparent K_M. This leads to an approximately equivalent k_eff for ΔCTP and ΔCTL. WT Bs-FtsZ is the least efficient enzyme, emphasizing auto-inhibition by the tail through a combination of the contributions of the ΔCTL and ΔCTP modules and the facile ability of WT to form bundles that lead to a reduction of overall activity.

Discussion

The intrinsically disordered CTT of Bs-FtsZ has two distinct modules viz., a hypervariable CTL and a well conserved CTP [76]. In this work, we focused on dissecting the functional contributions of each of these modules within the CTT and the overall contribution of the CTT. Taken together, our data suggest that the CTT acts as an auto-regulator of assembly and an auto-inhibitor of enzymatic activity. We find that the CTL weakens the driving forces for forming single-stranded polymers thereby leading to shorter polymers when compared to ΔCTT. The CTL also inhibits the formation of higher order assemblies. From a functional standpoint, the CTL acts as a spacer with a finite excluded volume and helps in spatially separating the CTP from the core of Bs-FtsZ – a feature that has been established by Huecas et al. for different FtsZs [56]. The ability of the CTL to function as a spacer also helps alleviate interactions involving the CTP, which if left unregulated will give rise to alternative mini ring structures (Figure 2L). The assembly suppressing activities of the CTL appear to ensure that the CTP engages primarily in interactions that drive higher-order assemblies. Additionally, in vivo, the spatial delocalization of the CTP from the core should help ensure that the CTP is free to coordinate the network of interactions involving FtsZ and modulatory proteins that help regulate the Z-ring assembly, localization, anchoring, and dynamics [3]. It stands to reason that there will be a competition between homotypic CTP-mediated lateral associations that promote higher-order assemblies of FtsZ and heterotypic interactions that involve the CTP and modulatory proteins. There could also be a synergy between homotypic associations giving rise to lateral associations and vacant CTP sites engaging in heterotypic interactions. Experiments designed to test the interplay between homotypic and heterotypic CTP-mediated interactions for FtsZs extracted from different bacteria would be of immense utility.

Our data suggest that while the CTL may be thought of as a spacer, the CTP plays the role of a sticker. In ΔCTL, the CTP enables the formation of alternative mini ring structures (Figure 2L) whereas in the WT, the CTP enables the formation of higher order assemblies. The latter likely arises through CTP-mediated electrostatic interactions (Figure 3). Our data are in accord with previous findings showing that the CTP promotes lateral bundling of protofilaments in B. subtilis [47, 57] and in E. coli [62]. We interpret the weakening of higher order assemblies at high salt concentrations to imply that CTP-dependent electrostatic interactions are among the main drivers of lateral associations that promote higher-order assemblies. Taken together, our data suggest that the spacer activities of the CTL and the sticker nature of the CTP make the CTT an auto-regulator of FtsZ assembly.

Previous studies have quantified the effects of KCl concentrations on inferred size distributions of E. coli FtsZ (Ec-FtsZ) polymers formed in the presence of GTP [36]. Comparisons between our data and those of Ahijado-Guzmán et al., would appear to point to discrepancies but closer scrutiny reveals mutual consistency. Our DLS data for the size distributions show that Bs-FtsZ assemblies become narrow and shift to lower values in 200 mM KCl vs. 50 mM KCl (Figure 3A). This implies a weakening of higher-order assemblies at higher concentrations of KCl. The data of Ahijado-Guzmán et al., [36] obtained for Ec-FtsZ in KCl concentrations between 100 mM and 500 mM suggests the opposite behavior where the narrow distributions shift toward smaller sizes in 100 mM KCl vs. 500 mM KCl. This is a reflection of the shorter linear polymers seen at 100 mM KCl in comparison to 500 mM KCl, as Ec FtsZ forms minimal bundles under these conditions. Therefore, the experiments of Ahijado-Guzmán et al., are probing a different phase boundary when compared to our assay. However, in conditions where bundling is enabled in the presence of a non-hydrolysable analog, the behavior we observe wherein higher salt decreases assembly size is recapitulated for Ec-FtsZ as well. It is worth emphasizing that the impact of salt concentration, ion valence, and ion hydrophobicity will likely contribute differently to the driving forces for and the assemblies of FtsZs derived from different orthologs. The totality of our data and those of Ahijado-Guzmán et al., [36] point to subtle interplays between the conformations of the FtsZ cores (which are very similar between Ec-FtsZ and Bs-FtsZ) and the CTTs, which are very different, especially across the CTLs, between Ec-FtsZ and Bs-FtsZ [76]. Clearly, a comparative analysis of the salt and osmolyte dependence of assemblies and driving forces for assembly across FtsZ orthologs will be imperative for uncovering differences in synergies among GTPase cores and disordered CTTs.

As far as enzyme activity is concerned, the presence of the CTL in ΔCTP weakens GTP binding vis-à-vis ΔCTT whereas the presence of the CTP in ΔCTL enhances GTP binding. Inferences regarding GTP binding are drawn from values for the apparent K_M estimated for different constructs (Figure 6A). In contrast to the apparent K_M values, the inferred k_cat values suggest near ΔCTT-like values for ΔCTP, and a significantly diminished value for ΔCTL (Figure 6B). It is possible that the effect on k_cat derives from the formation of shorter polymers when compared to ΔCTT. These data suggest that the CTP serves as an auxiliary binding site for GTP, especially when it is not spatially separated from the core via the CTL. Support for this inference comes from the higher k_cat values we infer for WT when compared to ΔCTL. The differing effects of the CTL and CTP lead to a WT enzyme that is roughly half as efficient as the core domain alone. This is seen in comparisons of inferred values for k_eff of ΔCTT vs WT (Figure 6C). Indeed, the CTT appears to act as an auto-inhibitor of enzymatic activity, a feature that is realized via convolution of the different contributions of the CTL and CTP modules. Further, it is known that the hydrolysis state of the bound nucleotide modulates the curvature of the protofilament [77]. Therefore, the presence of the CTL and CTP modules not only regulates activity but could also impact the ability of the filament to curve and likewise impact the ability to form a ring.

In many proteins, IDRs tethered to folded domains are known to contribute as auto-inhibitors of the functions controlled by folded domains [78–83]. Along these lines, Trudeau et al., showed that the activities of auto-inhibitory IDRs are more frequently modified by other modulatory proteins than the folded domains to which they are tethered [84]. Indeed, it is noteworthy that the auto-regulatory and auto-inhibitory functions of the CTT, due to differing contributions from the CTL and CTP, are congruent (from a phenomenological standpoint) with data reported by Li et al., for a series of variants of the glucocorticoid receptor [85]. The impacts of different types of spacers were quantified on the effects of hormone binding on allosteric coupling and transcriptional activation. These results were analyzed using a quantitative framework for ensemble allostery developed by Mothlagh and Hilser [86]. Li et al., substituted linkers that they deemed to be inert and used this to uncover positive coupling between one of the tethered IDRs and the structured domain while also uncovering evidence for autoregulation through negative coupling between the two tethered IDRs [85]. The studies of Li et al., provide a template for probing the effects of different CTLs on the nature of the coupling between the CTP and the core domain. In this context, it is worth emphasizing that the CTL is a hypervariable module characterized by an assortment of lengths and compositional biases [76]. Accordingly, designing an inert linker becomes non-trivial because there are likely to be cryptic interactions involving the CTL that can only be dissected using a combination of simulations [87] and phenotypic characterization [88]. Specifically, we will need to titrate sequence features within the CTL of different FtsZ orthologs to probe the impact of changes to sequence patterns within CTLs. This approach, inspired by recent efforts of Das et al., [89] and Sherry et al., [90] are already underway for the CTL of Bs-FtsZ [88]. Specifically, since the CTL of Bs-FtsZ is a polyampholyte [91], we are investigating the effects of altering the linear patterning of oppositely charged residues away from the patterning observed for the WT CTL. These designs, which maintain the amino acid composition, show clear cellular phenotypes [88], and our findings in the current study provide a clear basis for interpreting the emerging data. A detailed analysis of these designs and interpretations using the framework of Motlagh and Hilser [86] will be forthcoming.

The active role of the CTT as a modulator of Bs-FtsZ assembly and enzymatic activity also mirrors recent discoveries regarding the distinct roles of disordered C-terminal linkers and conserved C-terminal tips in single-stranded DNA binding proteins referred to as SSBs [92]. The SSBs share coarse-grained architectural similarities with FtsZs. Their oligomerizing N-terminal folded oligonucleotide binding (OB) domains are connected via hypervariable C-terminal linkers (CTLs) to conserved C-terminal tips that encompass acidic residues and a short linear motif that helps coordinate the interactions of SSB interaction proteins (SIPs) [93]. Recent molecular dissection studies on SSBs from E. coli (Ec-SSB) [92, 94, 95] show that the CTL in Ec-SSB is essential for cooperative binding of single-stranded DNA. Importantly, the CTL of Ec-SSB helps with spatial delocalization of the tip from the OB fold – a feature that is shared with the CTL of Bs-FtsZ suggesting a converging theme of disordered linkers serving as active spacers that separate stickers such as short linear motifs from folded domains.

As noted above, the CTLs of FtsZs are hypervariable, and this is akin to that of the CTLs of SSBs as well as other oligomerizing / self-assembling systems such as the bacterial RNA chaperone Hfq [80]. In the context of FtsZs, the lengths, compositions, and sequence features of CTLs vary considerably across FtsZ orthologs [50, 76]. In direct contrast, the sequences of CTPs and core domains remain largely conserved [76]. The hypervariability of CTLs in FtsZs might be an example of convergent evolution. In this scenario, a large number of disparate sequences, providing they are intrinsically disordered, might be able to serve as modulators of cooperative assembly of FtsZ while spatially separating the conserved core and CTP, weakening GTP binding, and enhancing the catalytic rate. On the other hand, the hypervariability of CTLs might also be an example of divergent evolution. Previous studies have shown that CTLs from different bacteria are not always interoperable with one another [59, 60, 62]. Furthermore, Sundararajan and Goley showed that the ΔCTT of C. crescentus FtsZ is less efficient than the WT. This implies that the mechanism that is operative in C. crescentus FtsZ is the opposite of what we have uncovered here for Bs-FtsZ. In C. crescentus FtsZ, the catalytic inefficiency of the core, possibly due to interactions among core domains that drive alternative inactive assemblies, is alleviated by alternative CTL and CTP modules. Indeed, analysis using CIDER [96] shows that the CTLs of Bs-FtsZ vs. C. crescentus FtsZ are very different from one another in terms of their sequence lengths (N = 50 vs. 176), fraction of charged residues (0.41 vs. 0.24), net charge per residue (0.06 vs. −0.05), and Kyte-Doolittle hydropathy (3.12 vs. 3.72). These results suggest that the sequences of CTLs and cores might covary in order to achieve a requisite level of enzymatic efficiency and assembly giving rise to distinct rules / features that underlie the sequence features of CTLs in distinct classes of bacteria.

CTL length is another source of variability among FtsZs and it is thought to be the key determinant of differences among FtsZs [56, 60, 62]. However, informatics analysis has established an inverse correlation between CTL length and the fraction of charged residues within CTLs [50, 76]. This feature, which is evident even in the comparisons of the CTLs from Bs-FtsZ and C. crescentus FtsZ, is important because recent studies have established a distinction between sequence length (also referred to as apparent length [87]) and functional length (also referred to as effective length [76] or thermodynamic length [87]). While the apparent length is simply the number of amino acids in a sequence, the functional length reflects the fact that intrinsically disordered regions (IDRs) of different sequence lengths can have similar end-to-end distance and shape distributions because amino acid compositions as well as sequence patterning effects, and not just apparent length, will control the conformational properties of IDRs [97].

In ongoing analysis, we have discovered that twelve out of the ~20 essential proteins that are involved in bacterial cell division encompass IDRs that are longer than 30 residues. This points to the potential importance of these regions as regulators of cell division. Furthermore, despite having considerably smaller proteomes when compared to eukaryotes, bacteria exert tight control over all processes that are vital to their life cycles. In contrast to eukaryotic systems and viruses, where the functional importance of IDRs is well established [98–101], the prevailing view is that bacterial proteins conform to the classical sequence-structure-function paradigm [102, 103]. This view has emerged mainly from bioinformatics analysis, which shows that only a small percentage of bacterial proteins encompass long IDRs [104–106]. However, several recent studies have demonstrated that while IDRs make up a small fraction of proteins / protein regions in bacterial proteomes, the synergies between IDRs and folded domains contribute directly to an assortment of functions. Prominent bacterial IDRs include regions within proteins that are involved in regulating cell division [50, 57], single-stranded DNA binding [92, 93], protein and RNA quality control [107–113], bacterial warfare [114–116], biofilm formation [117–119], and chemotaxis [120–122]. In each of these systems, it becomes imperative to undertake molecular dissections of IDR functions and uncover the coevolution between folded domains and IDRs.

Materials and Methods

Protein expression and purification:

Plasmids for Bs-FtsZ and the tail variants were acquired from the Levin Lab. These were cloned into the pET-21b(+) expression vector through E. coli strain AG1111. The resulting plasmids were mini-prepped and freshly transformed into C41(DE3) cells and made into glycerol stocks. Growth and purification were performed using the protocol deployed in Buske et al., [57]with minimal adjustments. 500 mL of LB medium was inoculated 1:100 with an overnight culture. Cells were grown at 37°C until A600 ~0.6–0.8, and then cells were induced with 1 mM isopropyl β-d-1-thiogalactopyranoside. Cells were grown for an additional 4h at 37°C. The cells were then harvested by centrifugation and the cell pellets were stored at −80°C. Purification was performed on an AKTA FLPC system using a Superdex 75 size exclusion column. Peak fractions were analyzed by SDS-PAGE, pooled together, and dialyzed overnight in 1 L of FtsZ dialysis buffer (50 mM MES 50 mM KCl, 2.5 mM MgCl2, 1 mM EGTA, pH 6.5) Protein concentration was quantified using UV-Vis via phenylalanine absorbance, and confirmed using a Pierce 660nm assay with BSA as a standard (Thermo Fisher Scientific). Protein preparations were concentrated as needed, separated into aliquots, and flash frozen on liquid N₂, and stored at −80°C. Prior to use, FtsZ aliquots were thawed on ice, well mixed, and the concentration was reconfirmed using UV-Vis.

Dynamic Light Scattering Assay:

FtsZ was diluted into an MES reaction buffer (50 mM MES, 2.5 mM MgCl₂, 1 mM EGTA, 1 mM GTP, pH 6.5) with KCl concentrations varying from 50–200 mM KCl as specified) to a total volume of 50 μL. All samples were incubated at 30°C with 1 mM GTP for 5 minutes prior to measurement on a Malvern ZEN3600 Zetasizer instrument (Nano Research Facility, Washington University in St. Louis). A standard operator procedure for aggregating proteins was used that was developed by the instrument manufacturer. The inherent output of dynamic light scattering measurements is a correlation function of intensity, which deconvolves the contributions of each size to the overall intensity. This distribution is then converted to a percent number distribution.

Transmission Electron Microscopy (TEM):

Samples were prepared in MES reaction buffer with 6 μM of FtsZ and the specified KCl concentration. Prior to preparing the glow-discharged copper grids, each sample was incubated for 5 minutes in the presence of 1 mM GTP to allow for adequate assembly. Each sample was quickly rinsed with DI water then stained 3x with 1% uranyl acetate for 5-seconds each, with the solution wicked away and followed by a 10 second wait time in between stains. Samples were visualized using a FEI Transmission Electron Microscope (Nano Research Facility, Washington University in St. Louis).

GTPase Assay:

GTP hydrolysis activity was monitored using a coupled GTPase assay [123]. Using a 96-well TECAN plate reader, the assay was conducted in MES reaction buffer and included 1 mM phosphoenolpyruvate, 250 μM NADH, and 40 units/ml of both lactose dehydrogenase and pyruvate kinase. From the following equation, the linear decline of NADH absorbance at 340 nm was monitored over 30 minutes. The steepest decline rate for a 5-minute consecutive stretch was related to the GTPase activity by the following manipulation of Beer’s law, which yields: $\left(\frac{\text{moles of GTP hydrolyzed}}{\text{min}}\right)=\left(\frac{\Delta {\text{A}}_{340}}{{\epsilon }_{\text{NADH}}L{V}_a}\right)$. Here, ΔA₃₄₀ is the slope of the decline, ε_NADH is the extinction coefficient for NADH at 340 nm (6220 M⁻¹cm⁻¹), L is the path length of the cuvette (0.401 cm), and V_a is the observation volume (150 μL). Each trial was performed in at least three times.

Determination of c_A:

The threshold concentration (c_A) and its variance were determined by randomly selecting a trial at each considered concentration and prescribing a linear fit. This process was repeated 10³ times. We only considered points that correspond to non-zero values along the ordinate. The mean and variance were then calculated from the resulting vector of intercepts along the abscissae. All analyses were performed using MATLAB.

Michaelis-Menten Fits:

Data for enzyme kinetics were fit to: $a=\left(\frac{v_{\text{max}}S}{{\text{K}}_{\text{M}}+S}\right)$. Here, a is the measured GTPase activity, S is the concentration of the substrate, which is GTP, and K_M is the apparent Michaelis constant. A random trial was selected from each tested GTP concentration, and the data were fit to equation for enzymatic activity. Values of K_M and v_max were determined using a nonlinear regression model, the fitnlm function in MATALB. This process was repeated 10³ times using bootstrapping with replacement to extract a mean and variance. The data were also fit using a Hill model, where $a=\left(\frac{v_{\text{max}}{S}^{\text{n}}}{{\text{K}}_{\text{M}}^{\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$\text{n}$}\right.}+S^{\text{n}}}\right)$. Here, the residuals from the two models were compared to determine the best fit. In all cases, the Michaelis-Menten model provided the best fit to the activity data.