A Localized Complex of Two Protein Oligomers Controls the Orientation of Cell Polarity

ABSTRACT Signaling hubs at bacterial cell poles establish cell polarity in the absence of membrane-bound compartments. In the asymmetrically dividing bacterium Caulobacter crescentus, cell polarity stems from the cell cycle-regulated localization and turnover of signaling protein complexes in these hubs, and yet the mechanisms that establish the identity of the two cell poles have not been established. Here, we recapitulate the tripartite assembly of a cell fate signaling complex that forms during the G1-S transition. Using in vivo and in vitro analyses of dynamic polar protein complex formation, we show that a polymeric cell polarity protein, SpmX, serves as a direct bridge between the PopZ polymeric network and the cell fate-directing DivJ histidine kinase. We demonstrate the direct binding between these three proteins and show that a polar microdomain spontaneously assembles when the three proteins are coexpressed heterologously in an Escherichia coli test system. The relative copy numbers of these proteins are essential for complex formation, as overexpression of SpmX in Caulobacter reorganizes the polarity of the cell, generating ectopic cell poles containing PopZ and DivJ. Hierarchical formation of higher-order SpmX oligomers nucleates new PopZ microdomain assemblies at the incipient lateral cell poles, driving localized outgrowth. By comparison to self-assembling protein networks and polar cell growth mechanisms in other bacterial species, we suggest that the cooligomeric PopZ-SpmX protein complex in Caulobacter illustrates a paradigm for coupling cell cycle progression to the controlled geometry of cell pole establishment.

morphology, replication competency, and size (1). Prior to cytokinesis in Caulobacter, distinct sets of signaling proteins localize to opposite cell poles where they dictate the cell fate of the nascent daughter cells.
One of the progeny, the sessile stalked cell, immediately enters S phase and initiates chromosome replication. The other progeny, the motile swarmer cell, enters G 1 phase and is incapable of DNA replication. The swarmer cell undergoes a period of differentiation to become a stalked cell, culminating in the generation of further progeny (Fig. 1). During this G 1 -S transition, the developing swarmer cell releases the PleC phosphatase from the flagellated cell pole, sheds its polar flagellum, begins biogenesis of the polar stalk appendage, and initiates DNA replication. Synthesis of the DivJ histidine kinase marks the end of the G 1 -S transition, enabling the initiation of chromosome replication and driving the stalked cell genetic program. Newly synthesized DivJ is positioned at the nascent stalked pole, where it remains localized to propagate the stalked cell fate throughout future divisions (2)(3)(4).
Localization of DivJ to the nascent stalked pole depends on several factors. A microdomain composed of the PopZ polymeric network, which marks the flagellated pole in swarmer cells (5,6), is necessary for the polar localization of DivJ in addition to many other cell fate factors that localize to the cell poles (6,7). One PopZ-dependent factor, SpmX, colocalizes with PopZ immediately upon synthesis at the beginning of the G 1 -S transition (8)(9)(10)(11). SpmX is necessary for the stalked pole localization and activation of DivJ (8). However, the biochemical basis of SpmX and DivJ localization to the stalked pole has not been elucidated.
Here, we investigate the mechanism of the ordered recruitment of SpmX and DivJ to the incipient stalked pole. We found that in vitro, PopZ directly binds SpmX, that SpmX in turn directly binds DivJ, and that these three proteins are sufficient for polar colocalization when coexpressed heterologously in Escherichia coli. SpmX forms higherorder oligomers, suggesting a structural integration between the PopZ and SpmX polymers. Cell cycle-controlled copy number is critical for the geometry of cell pole establishment, as overexpression of SpmX leads to the generation of ectopic growth zones, with the accumulation of SpmX at these sites facilitating the establishment of a   (13,14). SpmX and DivJ were purified without their TM regions. To assay binding of SpmX to PopZ and DivJ, we first sparsely labeled lysine residues on SpmX with Atto-488 dye (indicated as SpmX*). We then measured the PopZ-or DivJ-dependent change in the thermophoresis of SpmX* over a 2-fold serial dilution of either PopZ or DivJ ( Fig. 3A and B). Direct binding was observed between SpmX* and PopZ (K D [equilibrium dissociation constant] ϭ 600 Ϯ 100 nM) and between SpmX* and DivJ (K D ϭ 510 Ϯ 90 nM).
Dye-labeled SpmX-L*, bearing only the lysozyme domain, also directly bound to PopZ and DivJ but with affinities altered from those observed with wild-type (WT) SpmX* ( Fig. 3C and D). The 10-fold-weakened affinity of SpmX-L* for DivJ (K D ϭ 6.3 Ϯ 1.1 M) is consistent with the reduced capacity of SpmX-L-eYFP to recruit DivJ when The purified proteins were allowed to incubate together at room temperature for 10 min prior to the binding assay. The data report the fraction of SpmX* that is bound at each concentration of PopZ. The binding curve is fitted according to the law of mass action and is described in Text S1 in the supplemental material. (B to D) Direct binding was also assayed between WT SpmX* and DivJ (B), SpmX-L* and PopZ (C), and SpmX-L* and DivJ (D). The maximum concentrations of PopZ and DivJ are 10 M and 40 M, respectively. Dissociation constants are listed for each interaction, and each point represents the mean fraction bound of at least 3 independent experiments, with error bars representing the standard deviation.
Cell Polarity Control by a Complex of Oligomers ® coexpressed in E. coli (Fig. 2H). The affinity of SpmX-L for PopZ was 5-fold stronger (K D ϭ 120 Ϯ 40 nM) than that of WT. The proline-rich domain of SpmX is highly negatively charged, as is PopZ, suggesting that the interaction between SpmX-L and PopZ may be strengthened when this electrostatic repulsion is alleviated.
Surface plasmon resonance (SPR) experiments corroborated our finding that PopZ directly binds to both WT SpmX and SpmX-L (Fig. S2). PopZ was immobilized onto a biosensor chip, and 2-fold serial dilutions of WT SpmX or SpmX-L were injected over PopZ to assay for binding. The overall response of SpmX-L binding to PopZ was approximately 2-fold higher than that of WT SpmX at comparable concentrations, consistent with the tighter K D of SpmX-L binding to PopZ as measured via MST.
Cumulatively, our results demonstrate a hierarchical localization pathway whereby PopZ at the flagellum-bearing cell pole serves to recruit SpmX via a direct interaction during the swarmer-to-stalked-cell transition. SpmX bound to PopZ then recruits DivJ to the cell pole via a second interaction. Thus, SpmX serves as a bridge between the polar PopZ matrix and the DivJ kinase in order to localize DivJ to a specific cell pole.

Function of SpmX domains in Caulobacter.
To interrogate the functions of each domain of SpmX in Caulobacter, we examined the phenotypes of Caulobacter cells expressing SpmX domain deletion alleles fused to eYFP, from the native spmX promoter, and as the only copy of SpmX present in the cell. We assayed the subcellular localization of WT and two mutant SpmX alleles fused to eYFP: SpmX-L and SpmX missing its TM domains (SpmX-ΔTM) (Fig. S3A). Wild-type SpmX-eYFP displayed the previously reported nascent stalked pole localization pattern (8). In addition to the stalked pole signal, we also observed 18% of cells displaying a second, less intense focus at the new cell pole (Fig. S3A).
Replacing native SpmX with the truncated variant SpmX-L-eYFP or SpmX-ΔTM-eYFP yielded morphological defects, including filamentation and the formation of minicells (Fig. S3A). Additionally, SpmX-L-eYFP and SpmX-ΔTM-eYFP formed bipolar foci in 54% and 59% of cells, respectively, indicating that the SpmX TM domains contribute to unipolar SpmX localization. Western blotting assays showed that the relative levels of the SpmX-L-eYFP-and SpmX-ΔTM-eYFP-tagged mutant variants were slightly higher than WT ( Fig. S4A and B). Despite robust recruitment to PopZ, these SpmX variants failed to recruit DivJ-mCherry to the Caulobacter cell pole ( Fig. S3B and C), consistent with previous findings (8). Even though the in vitro binding assays showed that purified SpmX does not require its TM domains to interact with DivJ in vitro ( Fig. 3B and D), the requirement for the TM domains in vivo is likely due to the space-filling nature of the PopZ polar microdomain (7, 15) (Text S1). These results demonstrate that the SpmX lysozyme domain alone cannot complement WT SpmX function, despite its ability to directly bind to both PopZ and DivJ.

Point mutations in the lysozyme-like domain of SpmX affect SpmX polar localization and function in Caulobacter.
Lysozyme catalytic activity requires a conserved glutamic acid residue in the active site (16)(17)(18). The SpmX lysozyme-like domain conserves this residue at position E19 (Fig. S5A). Despite extensive efforts, we have not been able to detect SpmX lysozyme activity in vitro, though we were able to observe lysozyme activity for hen egg white lysozyme in identical assays (Fig. S5C). Nevertheless, we asked if mutation of E19 would affect SpmX localization and function in vivo. We mutated E19 to either alanine or arginine and expressed the mutant alleles as eYFP fusions as the sole chromosomal copy of spmX at the native locus (Fig. S5B). SpmX E19A-eYFP displayed bipolar localization in 59% of cells (compared to 18% for WT SpmX-eYFP) and mild cell filamentation defects (Fig. S5B). Cells harboring replacements of spmX with the point mutant SpmX E19R-eYFP did not maintain the protein at detectable levels ( Fig. S4B), suggesting that it is degraded. Purified SpmX E19R failed to interact with either PopZ or DivJ in vitro via MST binding assays ( Fig. S6A to C), and a homology model for the SpmX lysozyme domain, based on T4 endolysin, predicts that E19 participates in an electrostatic interaction that may contribute to the protein's stability and interactions (Fig. S6E). These data suggest that the catalytic region of this lysozyme domain is critical for normal interaction with PopZ and DivJ.
SpmX constitutive overexpression generates ectopic growth zones. The majority of SpmX remains at the stalked pole for the duration of the cell cycle ( Fig. 1A) (8,11). We reasoned that constitutive overexpression of SpmX could disrupt its asymmetric polar localization and normal function. Indeed, inducing spmX expression from the chromosomal xylX promoter was sufficient to produce minicells and, more strikingly, to initiate ectopic growth zones in~10% of cells (Fig. 4A). Transmission electron microscopy of these cells indicated that these ectopic growth zones had diameters and shapes resembling those of wild-type cells (Fig. 4B).
To determine if overexpressed SpmX localized to the ectopic poles, we performed time-lapse microscopy of cells overexpressing spmX-ecfp (Fig. 4C). SpmX-eCFP accumulated at the cell poles in cells displaying ectopic growth. Many of these cells had one cell pole with strikingly diminished SpmX-eCFP intensity, including a subset of poles lacking SpmX that budded off as minicells. Critically, we observed the ectopic accumulation of SpmX-eCFP signal in nonpolar regions of the cell that preceded the formation of ectopic pole growth (Fig. 4C, open arrowheads), and cells extending from ectopic growth zones were capable of division for multiple replication cycles (Fig. 4C, black arrowheads). Subsequent progeny continued to produce ectopic growth zones, indicating that they are capable of sustaining growth while exhibiting markedly altered axes of symmetry.
Polar identity of ectopic poles is maintained during SpmX overexpression. We visualized cells coexpressing SpmX-eCFP with other fluorescently labeled proteins normally positioned at the cell poles. We observed mCherry-PopZ colocalization with overexpressed SpmX-eCFP at many cell poles (Fig. 4D). Specifically, cells containing SpmX-eCFP accumulations at lateral regions sometimes colocalized with PopZ but PopZ never arrived at lateral regions without SpmX (Fig. S7). This finding suggests that during SpmX overexpression, SpmX can recruit PopZ to lateral sites prior to growth of an ectopic pole.
To determine the identities of the ectopic poles, we examined the subcellular localization of the mCherry-tagged stalked and swarmer pole markers DivJ and TipN, respectively. DivJ-mCherry foci colocalized with SpmX-eCFP at the ectopic cell poles (Fig. 4E), while TipN-mCherry did not (Fig. 4F, white arrowheads). Strikingly, cell poles that lacked SpmX-eCFP signal did not contain a corresponding DivJ-mCherry focus but did contain a polar focus of TipN-mCherry ( Fig. 4E and F), and we did not observe multiple polar TipN-mCherry accumulations. These data indicate that, during SpmX overexpression, most poles contain a stalked pole protein composition but suggest that one "swarmer pole" composition is maintained.
We additionally found that the division plane marker FtsZ-eYFP localized as a discrete band at medial positions of the cell body (Fig. 4G), with minor accumulations at the new pole, as has been previously observed (19). Cumulatively, our data demonstrate that cells undergoing abnormal growth during SpmX overexpression retain polar identity and the ability to divide and are able to sustain growth.
PopZ is critical for ectopic pole growth during SpmX overexpression. Given that SpmX interacts with both PopZ and DivJ, we asked whether formation of ectopic poles driven by SpmX overexpression requires PopZ and/or DivJ. Accordingly, we overexpressed SpmX-eCFP in either a ⌬popZ or a ⌬divJ background. In the absence of DivJ, SpmX-eCFP overexpression still induced ectopic pole formation (9% of cells, compared to~10% in an otherwise isogenic background) (Fig. 5A), indicating that DivJ is not required for ectopic growth under these conditions. However, ectopic poles were very rarely observed when SpmX-eCFP was overexpressed in the absence of PopZ (0.7% of cells). The SpmX-eCFP signal under this condition exhibited a patchy distribution throughout the cell, with large clusters of signal present at variable subcellular locations (Fig. 5A). Thus, PopZ is critical for ectopic pole formation during SpmX overexpression and is needed to cluster SpmX into tight foci as visualized by fluorescence It has been reported that overexpression of TipN also results in the growth of ectopic poles (20). We asked if SpmX overexpression induces ectopic pole growth via a mechanism that is distinct from that caused by TipN overexpression. Accordingly, we overexpressed SpmX-eCFP in a ⌬tipN background and observed that ectopic pole growth still occurred in the absence of TipN (5% of cells [ Fig. 5A, right]). While there are indeed fewer cells displaying ectopic poles in a ⌬tipN background (5% compared tõ 10% for WT), the observation that TipN-mCherry specifically does not colocalize at ectopic poles with SpmX-eCFP (Fig. 4F) suggests that TipN does not play a direct or critical role in the formation of SpmX-seeded ectopic growth zones.
Because the production of the SpmX protein is tightly cell cycle regulated during normal Caulobacter growth (8,11,21), we asked if the constitutive timing of expression contributed to the formation of ectopic cell poles. Overexpressing spmX-eYFP from its native promoter on a high-copy-number plasmid in a ⌬spmX background generated ectopic poles in 8% of cells, compared to approximately 10% when spmX-eYFP was expressed from the chromosomal xylX locus (Fig. 5B). Given this relatively small difference, these data indicate that the expression levels of overall levels of SpmX and not its timing of expression are critical for maintaining an axis of symmetry.

The transmembrane tethers of SpmX are necessary to generate ectopic poles.
To interrogate the molecular features of SpmX needed for the formation of ectopic growth zones, we overexpressed three SpmX alleles (SpmX E19R, SpmX-L, and SpmX-ΔTM) fused to eYFP via the chromosomal xylose promoter in a ⌬spmX background. Overexpression of both SpmX-ΔTM-eYFP and SpmX-L-eYFP failed to produce cells undergoing ectopic growth (0% of cells for both alleles) (Fig. 5C), indicating that the SpmX transmembrane domains are necessary for ectopic pole growth. We also assayed the effect of the mutation E19R on the ability of overexpressed SpmX to form ectopic growth zones. While the SpmX E19R variant was not stable at native expression levels, overexpression stabilized the variant as weak polar accumulations ( Fig. S4B and C;  Fig. 5C), possibly due to protection from defects in oligomerization, as described below. Overexpression of SpmX E19R-eYFP yielded cells with ectopic growth zones but at lower levels than with WT (3% to 11%, respectively) (Fig. 5C), perhaps reflecting the impaired ability of this SpmX mutant to interact with PopZ in vitro. The drop in ectopic pole formation and reduced capacity to localize to the cell poles suggests that SpmX E19R has a diminished ability to generate ectopic poles upon overexpression, perhaps due to its impaired ability to interact with PopZ (Fig. S6).

SpmX oligomerizes via interactions in the lysozyme and proline-rich domains.
The observation that the SpmX-eCFP overexpression resulted in accumulations of lateral SpmX foci in the absence of PopZ (Fig. 5A) suggested that SpmX might self-associate, providing a mechanism to seed ectopic growth zones. To test this, we measured the apparent molecular mass of purified SpmX variants by analytical size exclusion chromatography. While the predicted monomeric molecular mass of WT SpmX is 40.1 kDa, the majority of WT SpmX eluted in the void volume, suggestive of a molecular mass greater than 600 kDa, with smaller peaks corresponding to estimated molecular masses of 136 kDa and 44 kDa (Fig. 6A). Strikingly, the SpmX E19R variant eluted primarily at an apparent molecular mass of 158 kDa, close to the predicted molecular mass of a tetrameric SpmX species. SpmX-L eluted as a monomer, indicating that the proline-rich domain is necessary for oligomeric assembly (Fig. 6A). While secondary structure predictions indicate that much of this region forms a random coil, a 19-residue stretch in this region is free of prolines and predicted to contain an extended alpha-helix that could assist in oligomerization (Fig. S8) (22,23).
Native gel analysis of the same proteins showed that WT SpmX migrated in a ladder-like pattern composed of discrete pairs of bands (Fig. 6B). While the elution of SpmX complexes in the void volume of the gel filtration column (Fig. 6A) could suggest that this protein exists as nonfunctional aggregates, the well-resolved, repeated addition pattern of the native gel indicates that the protein exists in discrete species formed by regularly sized additions. And, though the displayed molecular mass standards of the native gel do not evenly account for charge or shape of distinct substrates under nondenaturing conditions, these species were centered near the 720-kDa marker. SpmX E19R ran as a single smeared band at a lower apparent molecular mass than the WT, consistent with a defect in higher-order assembly. Purified PopZ protein is capable of refolding into its native oligomeric structure after denaturation (5,6). We asked whether SpmX also has this property. Following denaturation and refolding in nondenaturing buffer, native gel electrophoresis showed that WT SpmX partially reformed the fastest-migrating native band, while a majority of the protein migrated as a band similar to nondenatured SpmX E19R. Refolded SpmX E19R migrated in a pattern similar to its nondenatured form. These results indicate that the mutation E19R in the lysozyme-like domain of SpmX specifically disrupts the assembly of higher-order oligomeric structures and suggest that the intermediate assembly product trapped by E19R is capable of refolding after denaturation.
To further validate the SpmX self-association, we assayed the binding of purified, dye-labeled WT SpmX* to unlabeled SpmX by MST. We observed a single oligomeric transition of labeled SpmX* binding to itself and a weakened interaction between SpmX-L* and SpmX (Fig. S9A and B) (K D ϭ 360 Ϯ 50 nM and 630 Ϯ 70 nM, respectively). SpmX E19R did not show any binding to WT SpmX (Fig. S6). These measured interac-tions between the lysozyme-like domain and the WT SpmX species indicates that the lysozyme-like domain of SpmX makes contact with a WT SpmX species, suggesting that it can incorporate into an oligomeric structure when the full-length protein is present and that compromising this domain via mutagenesis partially disrupts SpmX selfassociation. The finding that SpmX undergoes one observable binding transition in MST experiments suggests that the assembly into an intermediate state occurs at a concentration below the 25 nM level at which we probed oligomerization with SpmX*, consistent with our inability to observe an E19R oligomeric transition. Notably, the Cell Polarity Control by a Complex of Oligomers ® SpmX E19R mutant displays a reduced ability to form ectopic foci and cell poles when overexpressed in vivo (Fig. 5C), suggesting that its ability to form multimers is compromised and that this oligomerization contributes to SpmX function. Cumulatively, the in vitro self-association experiments suggest that SpmX forms oligomers that assemble through a multistep process (Fig. 6C). The proline-rich domain is strictly required for oligomerization, and the mutation E19R in the lysozyme-like domain disrupts further higher-order assembly, suggesting that this domain mediates additional contacts. Further, comparison of the SpmX variants' ability to form oligomers in vitro with their relative ability to form fluorescent foci in vivo suggests a possible mechanism for both the polar localization of SpmX and seeding of new SpmX-PopZ accumulations and the initiation of ectopic cell pole growth when SpmX is overexpressed.

DISCUSSION
Here, we have shown that the lysozyme-like domain of SpmX mediates the ordered assembly of a signaling complex through direct interactions with the PopZ polymeric matrix and the DivJ histidine kinase. In addition to mediating contact with PopZ and DivJ, the lysozyme-like domain contributes to oligomerization of SpmX. When overexpressed, SpmX seeds ectopic cell poles, recruiting the DivJ histidine kinase, with outgrowth of these poles dependent on the establishment of a localized PopZ microdomain. The PopZ-SpmX complex thus coordinates cellular polarity through its coordinated regulation of cell fate and localized cell growth.
The repurposed lysozyme-like factor SpmX bridges the assembly of a microdomain signaling complex with the PopZ polymeric network. SpmX and DivJ are produced sequentially at the beginning of the swarmer-to-stalked-cell transition (8). We recapitulated the polar localization dependencies of these proteins in an in vivo heterologous E. coli expression system (Fig. 2D), indicating that SpmX forms a bridge between PopZ and DivJ in the polar complex. Further, we demonstrated that purified SpmX directly interacts with both PopZ and DivJ in vitro (Fig. 3). We conclude that SpmX directly bridges PopZ to DivJ in Caulobacter, localizing the kinase at the incipient stalked pole.
The N-terminal lysozyme-like domain of SpmX has been shown to be a polar localization determinant for SpmX in Caulobacter (8). Consistent with this result, we show that the SpmX lysozyme-like domain is sufficient for colocalization with PopZ at the heterologous E. coli cell pole in vivo as well as for a direct interaction with PopZ in vitro ( Fig. 2G and 3C). We have not detected in vitro catalytic activity for purified SpmX constructs, but we did find that mutation of the conserved, catalytic residue E19 disrupted the abilities of SpmX to localize to one pole or to form ectopic poles upon overexpression in vivo ( Fig. 5C; also see Fig. S5 in the supplemental material). While we found that WT SpmX forms oligomers in vitro, the mutant SpmX E19R cannot oligomerize into the highest-order structures observed in size exclusion chromatography and native gel electrophoresis ( Fig. 6A and B). This mutant is also incapable of interacting with both PopZ and DivJ, raising the possibility that SpmX oligomerization is important for its interaction with other proteins. Together, these data suggest that the lysozymelike fold of SpmX has been repurposed as a mediator of oligomeric assembly and protein recruitment. We additionally found that the proline-rich domain of SpmX was necessary for initial oligomerization into a tetrameric state, as indicated by the comparison of the apparent molecular weights (MWs) of the full-length E19R point mutant and the isolated lysozyme-like domain ( Fig. 6A and C). This initial oligomerization event may occur through an isolated, extended ␣-helix, which is predicted to be an island of local structure amid the otherwise disordered region (Fig. S8). Holmes et al. recently suggested that the disordered proline-rich region of PopZ critically contributes to its ability to localize up to 11 different protein substrates in vivo (7), and conservation of this motif in SpmX further supports a critical role for it in organizing polarity factors.
Models for the generation of ectopic growth zones during SpmX overexpression. Two primary models can explain the mechanism by which overproduction of SpmX generates ectopic poles: SpmX is the central actor in the generation of an ectopic pole or lateral aggregates of SpmX recruit additional factors that build the new poles. The lysozyme-like domain suggests that SpmX could degrade or perhaps modify peptidoglycan (PG) through a process such as transglycosylation (17), directly generating outgrowths by acting on the cell wall. While it has been suggested that SpmX exists in the periplasm (8), where lysozyme typically acts, SpmX lacks periplasm secretion motifs via either the Sec or twin-arginine transport pathways (24,25). Further, the direct interactions between SpmX-L and the cytoplasmic protein PopZ and the cytoplasmic portion of DivJ ( Fig. 3C and D) suggest that SpmX likely exists largely in the cytoplasm, though it remains possible that there is an additional method for transiently transporting SpmX to the periplasm. Alternatively, SpmX may participate in an upstream step in PG assembly that occurs in the cytoplasm prior to lipid II flipping (26). The fact that the E19R mutant still retains some ability to generate ectopic poles strongly argues against this hypothesis. However, because we have no evidence that the SpmX lysozyme domain has catalytic activity, we favor the hypothesis that aggregates of SpmX attract the formation of a PopZ microdomain and recruit factors involved in cell wall growth to create new regions of localized outgrowth.
Cell growth normally occurs via the addition of new PG at two cellular locations in Caulobacter: at the midcell and at the base of the stalk, where SpmX typically resides (27)(28)(29). Specifically, SpmX may coopt cell growth machinery normally dedicated to stalk biogenesis and repurpose it to build new cell poles. Further, overexpression of SpmX may inactivate the transcription factor TacA, which promotes stalk biogenesis (30), instead forcing this normally localized cell wall growth to ectopic sites. Depletion of the cytoskeletal protein MreB or the PG synthase RodA (31) has been shown to yield ectopic poles in Caulobacter at the expense of normal stalk biogenesis (32).
While SpmX is important for stalk formation and placement in the related genus Asticcacaulis (33), it is not in Caulobacter (8). However, PopZ is a critical regulator of both stalk biogenesis and the formation of ectopic cell poles (Fig. 5A) (5,6). PopZ defines a polar microdomain, recruiting at least 11 other proteins which may additionally include cell wall assembly proteins normally dedicated to stalk biogenesis (7). These growth factors would thus be repositioned to ectopic SpmX/PopZ foci during SpmX overexpression, leading to the formation of new cell poles.
The dependence on PopZ for the outgrowth of ectopic poles reflects similar findings in several other organisms. Notably, in the alphaproteobacterium Agrobacterium tumefaciens, which adds new PG at the tip of one cell pole, PopZ colocalizes with a transpeptidase at the growing cell pole, and mislocalization of PopZ causes ectopic growth zones (34,35). More broadly, homologs of the Bacillus subtilis polymeric protein DivIVA, which has many similarities to PopZ, have been linked to polar growth in Streptomyces coelicolor and Mycobacterium smegmatis through its regulation of PG addition at regions of high membrane curvature (36)(37)(38)(39).
Self-organizing polarity factors. When overexpressed, SpmX forms ectopic oligomeric aggregates that recruit PopZ to nonpolar sites, opposite from the typical order of assembly. Indeed, PopZ never formed ectopic accumulations without SpmX also being present ( Fig. 6C; Fig. S7). The overexpression of SpmX in Caulobacter mirrors the coexpression of PopZ and SpmX in E. coli in the ability of SpmX to seed additional PopZ foci (Fig. 2B, 4D, and 6C). This phenotype differs from the overexpression of PopZ in Caulobacter or in an E. coli test system (Fig. 6C), which simply increases the size of existing PopZ regions and is not sufficient to generate new foci (6,9,40). Recent work additionally demonstrated that the PopZ binding protein ZitP can also seed a second PopZ focus when coexpressed in E. coli, in a manner that requires both PopZ binding and the ZitP transmembrane tether (41). Critically, this similarity between SpmX, ZitP, and other proteins (7) indicates that membrane-bound factors that interact with PopZ may be generally adept at seeding new PopZ microdomains. This potentially widespread mechanism for generation of ectopic microdomains underscores the need for tight spatial and temporal control of polarity factors to prevent cell polarity aberrations.
Our work demonstrates the ordered assembly of the PopZ-SpmX-DivJ signaling complex and that disruption of the relative copy number of the protein components of this complex can drastically alter cell polarity. Ectopic aggregates of SpmX nucleate new poles upon acquisition of colocalized PopZ microdomains, and they specify the identity of those poles based on the recruitment or exclusion of other polar identity proteins. While it is intuitive that self-organizing proteins such as SpmX and PopZ are the most upstream regulators of cell polarity, many questions remain about how these master organizers interact to robustly establish, remodel, and propagate cell polarity through many generations.

MATERIALS AND METHODS
Bacterial strains and growth. For detailed information on strains, plasmids, and growth conditions, consult the supplemental material (see the supplemental methods and tables in Text S1). All C. crescentus strains used in this study are derived from the synchronizable wild-type strain CB15N (42) and were grown at 28°C in M2G medium (43). Plasmids used were constructed via Gibson assembly or described in previous studies (45)(46)(47)(48)(49).
Microscopy. C. crescentus and E. coli strains were imaged on M2G-1.5% agarose pads. Phase-contrast and fluorescence microscopy images were obtained using a Leica DM6000 B microscope with an HCX PL APO 100ϫ/1.40 oil PH3 CS objective, Hamamatsu electron-multiplying charge-coupled device (EMCCD) C9100 camera, and MetaMorph microscopy automation and image analysis software. For all image panels, the brightness and contrast of the images were balanced with ImageJ (NIH) or Adobe Photoshop CS6 to represent foci and diffuse fluorescent signal.
For computational image analyses, MicrobeTracker (44) was used to determine cell outlines from phase images. Fluorescent signal was integrated along the length of the cell outlines. For fluorescence profiles, a Matlab (MathWorks) script was used to interpolate the integrated fluorescent signal into 50 segments along the cell length. These profiles were averaged for 46 to 210 cells for each experiment. The averaged profiles were normalized such that the highest signal intensity was equal to 1. For E. coli cells displaying only diffuse localization, position 0 was randomly set to one cell pole. For E. coli cells displaying a fluorescent polar focus, position 0 was set to the cell pole containing the focus. For Caulobacter cells, position 0 was set to the stalked pole as visualized via phase images. If no stalk was visible, the pole containing the most intense SpmX-eYFP focus was set at position 0. For all experiments, only cells containing fluorescent signal were analyzed.
Negative-stain electron microscopy. Mid-log-phase cultures were applied to glow-discharged carbon-coated grids, stained with 1.5% uranyl acetate, blotted, and air dried. Images were taken at 80 kV on a JEOL TEM1230 transmission electron microscope equipped with a Gatan 967 slow-scan, cooled CCD camera.
MST binding assays. SpmX variants were fluorescently labeled on lysine residues with N-hydroxysuccinimide-functionalized Atto-488 (Sigma-Aldrich) at approximately one dye molecule per protein. Direct binding between fluorescently labeled SpmX variants and the SpmX targets PopZ, SpmX, and DivJ was probed in 2-fold serial dilutions via microscale thermophoresis (MST) (13, 14) (NanoTemper Technologies). Data analysis was performed as described previously (13). Detailed methods can be found in the supplemental material. Protein purification was performed as described in detail previously (50) and in the supplemental material.
Size exclusion chromatography. SpmX variants were assayed for the apparent molecular weight of their assemblies by size exclusion chromatography, using a GE Healthcare Superdex 200 Increase 10/300 GL, connected to a Bio-Rad NGC chromatography system. Molecular weights of the complexes were assigned based on a standard curve derived from the elution volumes of a Bio-Rad premixed gel filtration standard (catalog no. 151-1901). For detailed assay conditions, consult the supplemental material.
Native gel protein assembly assays. SpmX variants were assayed for their ability to form higherorder assemblies using nondenaturing gel electrophoresis. Three micrograms of SpmX variant was loaded into each well of a TGX gel (4 to 15%; Bio-Rad), and SpmX complexes were separated by gel electrophoresis at 80 V, for at least 2.5 h, at 4°C. Gels were stained for protein using SafeStain (Invitrogen). Approximations of molecular weight were made using a NativeMark (Thermo Fisher) ladder. Detailed procedures can be found in the supplemental material.