Modularity and determinants of a (bi-)polarization control system from free-living and obligate intracellular bacteria

Although free-living and obligate intracellular bacteria are both polarized it is unclear whether the underlying polarization mechanisms and effector proteins are conserved. Here we dissect at the cytological, functional and structural level a conserved polarization module from the free living α-proteobacterium Caulobacter crescentus and an orthologous system from an obligate intracellular (rickettsial) pathogen. The NMR solution structure of the zinc-finger (ZnR) domain from the bifunctional and bipolar ZitP pilus assembly/motility regulator revealed conserved interaction determinants for PopZ, a bipolar matrix protein that anchors the ParB centromere-binding protein and other regulatory factors at the poles. We show that ZitP regulates cytokinesis and the localization of ParB and PopZ, targeting PopZ independently of the previously known binding sites for its client proteins. Through heterologous localization assays with rickettsial ZitP and PopZ orthologs, we document the shared ancestries, activities and structural determinants of a (bi-)polarization system encoded in free-living and obligate intracellular α-proteobacteria. DOI: http://dx.doi.org/10.7554/eLife.20640.001

The fresh-water bacterium Caulobacter crescentus is a model system for the genetic analysis of a-proteobacterial cell polarity because polar differentiation is tightly coordinated with cell cycle progression and because of the availability of a myriad of genetic tools to study this species compared to the obligate intracellular (rickettsial) pathogens ( Figure 1A) (Curtis and Brun, 2010;Ely, 1991).  temporarily in a non-replicative (G1-like) state. At the SW to ST cell transition, the flagellated and piliated (SW) pole is remodeled into a ST pole and the developing cell acquires DNA replication competence. Replication of the circular chromosome proceeds bi-directionally from the single origin of replication (Cori) located at the nascent ST pole (Curtis and Brun, 2010). Once duplicated, the Cori region is rapidly segregated towards the nascent SW pole by the ParAB chromosome segregation system that targets the parS centromeric sequence located circa 8 kbp from Cori ( Figure 1A) (Mohl and Gober, 1997;Viollier et al., 2004). The cis-encoded ParB protein binds parS and the resulting ParB . parS complex is guided pole-ward by the ParA ATPase, likely reinforced by poorly understood biophysical constraints and properties of the chromosome (Lim et al., 2014;Mohl and Gober, 1997). The PopZ polar organizing protein is thought to assemble a porous homo-polymeric matrix at the cell poles that captures the segregated ParB.parS complex ( Figure 1A) via a direct interaction with ParAB (Bowman et al., 2008(Bowman et al., , 2013Ebersbach et al., 2008;Holmes et al., 2016;Laloux and Jacobs-Wagner, 2013). The polar localization of PopZ is cell cycle-regulated: in newborn cells PopZ is localized to the old cell pole, whereas the newborn pole initially lacks a PopZ cluster (Bowman et al., 2008;Ebersbach et al., 2008). During S-phase, PopZ adopts a bipolar disposition ( Figure 1A) coincident with ParB.parS segregation to facilitate its capture at the opposite pole. Formation of the second polar PopZ cluster may depend on the ParA ATPase and the TipN landmark protein, a coiled-coil protein that interacts with ParA and that marks the new pole as flagellar assembly site (Huitema et al., 2006;Laloux and Jacobs-Wagner, 2013;Lam et al., 2006;Ptacin et al., 2010). Although ParAB are essential for viability in C. crescentus, TipN or PopZ are individually dispensable (Bowman et al., 2008;Ebersbach et al., 2008;Huitema et al., 2006;Lam et al., 2006;Mohl and Gober, 1997), but joint inactivation of both genes arrests growth (Schofield et al., 2010). As most a-proteobacterial genomes encode PopZ and ParAB, but not TipN, TipN-independent control mechanism(s) for PopZ polarization should exist in a-proteobacteria.

T C P E C A S R Y F V D D S K V G P D G R V V R C A S C G N R W T A F K D E A E --M L L T C P K C A L S Y A I D G A Q L G P Q G R T V R C A S C K T T W H A E K P E E P M S I V L S C P S C T T R Y R A N P N A I G T N G R R V R C A S C G H V W T A E V E D P ---M I L T C P E C A S R Y F V D D S K V G P E G R V V R C A A C G H R W T A R N E D A T --M I L T C P A C A T S Y F V P D E A I G P N G R R V R C K T C G H D W R A S L E D
In the Rhizobiales order of the a-proteobacteria, PopZ does not appear to fasten ParB at the cell poles as polar PopZ does not co-localize with ParB: while PopZ is monopolar, ParB localizes in a bipolar fashion (Deghelt et al., 2014;Grangeon et al., 2015). Thus, compared to the bipolar co- Figure 1 continued C-terminal domain-of-unknown function (DUF3426). The green arrowhead points to the codon in the zitP coding sequence harboring the GFP insertion in the zitP::Tn5-GFP strain. All regions are drawn to scale. Numbers indicate residues. (C) Alignment of the ZnR from a-proteobacterial ZitP orthologs (in red) and one d-proteobacterium (in blue) ( [Mx, Myxococcus xanthus]). The four cysteine residues coordinating the zinc ion are highlighted (blue arrowheads). Asterisks indicate the conserved residues promoting ZitP . PopZ complex formation. (D) Overlays of fluorescence and phase contrast images showing the subcellular localization of ZitP Tn-GFP encoded by the zitP::Tn5-GFP allele in WT or DpopZ C. crescentus cells (top). The graphs below show the quantitation of the localization from above. The left graph indicates the distribution of foci along the longitudinal axis. Focus (n = 1048) position is given in relative coordinates from 0 (pole) to 0.5 (midcell). P, pole; M, midcell. The right graph shows the percentage of cells containing at least one focus of ZitP Tn-GFP in WT (n = 1048) or in DpopZ cells (n = 426). (E) Overlay images as in D showing the subcellular localization of the first 90 residues of ZitP from C. crescentus (Cc) and orthologs from A. excentricus (Ae), M. maris (Mm), C. segnis (Cs) in C. crescentus WT (upper panels) or DpopZ (bottom panels) cells. Strains expressing Dendra2-ZitP 1-90 from the chromosomal xylX locus were induced with xylose for 4 hr before imaging (phase contrast and Dendra2-fluorescence). (F) Overlay images as in D showing the subcellular localization of the ZnR of ZitP (Dendra2-ZitP 1-43 ) of C. crescentus (Cc) and orthologs from B. diminuta (Bd) in WT (upper panels) or DpopZ (bottom panels) C. crescentus cells. Strains expressing Dendra2-ZitP 1-43 from the chromosomal xylX locus were induced with xylose for 4 hr before imaging (phase contrast and Dendra2fluorescence). (G) Time-lapse imaging of swarmer cells from DzitP cells expressing Dendra2-ZitP 1-43 from the chromosomal xylX locus after induction for 1 hr with 0.3% xylose. Cells were then synchronized and transferred onto an agarose pad containing 0.3% xylose (t = 0 min), and visualized at 40 min intervals (time in minutes is indicated in the images) by phase contrast and Dendra2-fluorescence microscopy, respectively. Shown above the overlays are the schematics representing ZitP 1-43 localization during the C. crescentus cell cycle. (H) Images of DzitP cells expressing Dendra2-ZitP 1-43 from the chromosomal xylX locus and mCherry-PopZ from the native chromosomal popZ locus. Fluorescence and phase contrast images were acquired after 4 hr of induction with 0.3% xylose. Cells expressing Dendra2-ZitP 1-43 (left panel, green) and mCherry-PopZ (middle panel, red) are shown. Co-localized red and green foci appear yellow in the overlay (right panel). DOI: 10.7554/eLife.20640.002 The following figure supplement is available for figure 1: localization of PopZ and ParB observed in C. crescentus (Bowman et al., 2008;Ebersbach et al., 2008;Ptacin et al., 2014), PopZ seems to have undergone functional specialization in the Rhizobiales, presumably interacting with other (unknown) client proteins. The genomes of the obligate intracellular (rickettsial) lineage also encode PopZ and ParAB orthologs (Andersson et al., 1998), but not several other known client proteins of C. crescentus PopZ that depend on a short N-terminal stretch in PopZ to interact with it (Bowman et al., 2010;Holmes et al., 2016;Laloux and Jacobs-Wagner, 2013;Ptacin et al., 2014).
Here, we unearth a reciprocal, physical and conserved interaction between PopZ and the cytoplasmic N-terminal zinc-finger domain (ZnR) from ZitP (Hughes et al., 2010), a bifunctional and bipolar membrane protein whose C-terminal DUF3426 domain is required for polar pilus biogenesis (Mignolet et al., 2016) (Christen et al., 2016), but dispensable for motility. We locate the structural determinants governing PopZ . ZitP complex formation and we show that this interaction is required to control cytokinesis and centromere positioning from the membrane. The PopZ.ZitP interaction differs for previously described interactions between PopZ and client proteins (Holmes et al., 2016) in that it does not require the aforementioned PopZ N-terminal segment. We show that ZitP induces PopZ bipolarity in the heterologous host Escherichia coli and we reconstitute a bipolar ZitP . PopZ . ParB tripartite complex in this system. Examining rickettsial PopZ and ZitP orthologs, we find that the PopZ.ZitP complex is modular and that the structure-function relationship is maintained in the obligate intracellular lineage.

Results
PopZ-dependent localization of the ZitP zinc-finger domain (ZnR) of aproteobacteria We previously isolated a C. crescentus strain bearing a Tn5-GFP insertion in the 5'-proximal third of the poorly characterized and conserved CC_2215/CCNA_02298 gene (Figure 1-figure supplement 1A, dubbed zitP, zinc-finger targeting PopZ, Figure 1B-D). The resulting strain expresses a bipolarly localized and truncated ZitP-GFP fusion protein and is viable (Hughes et al., 2010). Consistent with this result, genome-wide transposon insertion sequencing (Tn-Seq) revealed that ZitP is not essential for viability, but required for pilus biogenesis and motility (Christen et al., 2011(Christen et al., , 2016Hughes et al., 2010) (Mignolet et al., 2016). The predicted ZitP protein harbors an N-terminal zinc-finger domain (zinc_ribbon_5 or PF13719 superfamily, residues 1-37, henceforth ZnR, Figure 1B-C) and a trans-membrane segment (TM, residues 106-127) preceding the C-terminal DUF3426 (residues 128-245). While the DUF3426 is required for polar pilus biogenesis (Mignolet et al., 2016), it is not required for motility or to direct ZitP to the cell poles since in polar ZitP Tn-GFP the GFP moiety is fused in-frame to codon 49 of ZitP (green triangle in Figure 1B).
Quantitative live-cell fluorescence imaging of ZitP Tn-GFP in zitP::Tn5-GFP cells (n = 1048) revealed monopolar fluorescent foci in non-constricted cells, while bipolar foci are present in constricted cells ( Figure 1D), suggesting that the ZnR and/or adjacent residues are necessary and sufficient for ZitP polarization and that bipolarization is cell cycle-regulated. We engineered strains expressing a Den-dra2-ZitP (1-90) fusion protein (an N-terminal translational fusion of the Dendra2 fluorescent protein to only the cytoplasmic part of ZitP) from the xylose-inducible promoter (P xyl ) at the xylX locus in the wild-type (WT, NA1000) or DzitP background and found an identical localization pattern as for the ZitP Tn-GFP strain, indicating that residues 1-90 carry the necessary information for polar localization of ZitP ( Figure 1E). To test if polarization is also a feature of ZitP from other a-proteobacteria, we expressed orthologous Dendra2-ZitP (1-90) variants in WT or DzitP C. crescentus cells and observed an identical localization pattern of these fusion proteins ( Figure 1E). By contrast, the zinc-finger domain from the unrelated AgmX protein encoded in the d-proteobacterium Myxococcus xanthus (Nan et al., 2010) is not polar in C. crescentus (Figure 1-figure supplement 1B), indicating that the zinc-finger domain of AgmX diverged.

NMR structure of ZitP ZnR and determinants for PopZ binding
To determine if C. crescentus ZitP  and PopZ interact directly, we separately purified soluble ZitP  and PopZ with a C-teminal-hexa-histidine (His 6 )-tag from an E. coli overexpression system. In size exclusion chromatography (SEC), PopZ elutes with an apparent molecular mass of oligomers (~200 kDa), while ZitP (1-43) elutes as a monomer ( Figure 2-figure supplement 1B-D). ZitP  and PopZ can be co-purified by SEC and the resulting ZitP  .PopZ complex has a high apparent molecular mass, suggesting that ZitP binds oligomeric PopZ. Isothermal titration calorimetry (ITC) estimated the dissociation constant (K D ) between ZitP  and PopZ at 700 nM ( Figure 2C), confirming the specific interaction between both proteins in vitro.
To identify determinants within ZitP that are required for the interaction with PopZ, we first resolved the NMR solution structure of ZitP (1-43) by following a classical nuclear Overhauser effect (NOE)-based approach ( Figure 2E, Supplementary file 1, Figure 2-figure supplement 1E, see Materials and methods). The resonance assignment revealed that cysteine residues C5, C8, C28 and C31 are reduced and harbor typical Cb chemical shifts of a zinc ion coordination module comprised between 31.58 ppm and 32.96 ppm (Kornhaber et al., 2006). The 20 NMR structures of ZitP  were overlaid with a root mean square deviation of 0.23 Å over the backbone atoms (Supplementary file 1), revealing two double-stranded antiparallel b-sheets (bbabb), forming a 'crab claw' in which C5, C8, C28, C31 are located at the turns of b1/b2 and b3/b4 chelate a zinc ion ( Figure 2D-E). This ZitP ZnR structure is unusual, though related to the family of the 'ribbon zinc-fingers' (zinc_ribbon_5 or PF13719 superfamily), albeit it lacks the additional b-strand typically located in between the two b-sheets.
Next, we monitored the interaction between 15 N-labeled ZitP  and unlabeled PopZ to locate residues in ZitP  that are influenced by the interaction with PopZ ( Figure 2F). This turned our attention to the aromatic side-chain of an invariant tryptophan at position 35 (W35, Figure 2E) that stacks above the zinc-coordination module of ZitP and that is replaced by a new species upon the addition of PopZ. W35 is interesting because the aromatic side chain at this position is replaced by an isoleucine in the primary structure of the AgmX zinc-finger ( Figure 1C) that does not localize in C. crescentus (Figure 1-figure supplement 1B). Moreover, W35 is surrounded by a surfaceexposed patch of basic residues (K18, R24, R27, R34, Figure 2E) and it is required to bind PopZ in vitro, as determined by ITC with a W35I mutant derivative of ZitP  [ZitP (1-43)W35I , Figure 2C]. Consistent with these ITC experiments, W35I impairs the interaction of Dendra2-ZitP (1-43) with polar PopZ in vivo ( Figure 3A    , we conclude that these residues are required for the formation of a polar ZitP  . PopZ complex. Interestingly, R. sphaeroides  is also largely diffuse in C. crescentus ( Figure 3C) or in PopZ-expressing E. coli, but some faint polar signals are noticeable as well ( Figure 3D). Questioning the basis for this poor polar localization in C. crescentus, we noted that this ZitP ortholog contains a naturally-occurring substitution (corresponding to R27Q in Figure 1C) in the aforementioned basic patch. We therefore asked if polar localization could be improved if this substitution is 'corrected' (i.e. reversed) by a Q27R substitution of R. sphaeroides  . This 'corrected' R. sphaeroides ZitP (1-43)Q27R version is indeed robustly polar in C. crescentus ( Figure 3C) and in PopZ-expressing E. coli cells ( Figure 3D). When the W35I substitution is introduced into this 'corrected' R. sphaeroides variant, polar localization is again lost without compromising abundance (Figure 3-figure supplement 1B). We conclude that W35, the basic patch and the integrity of the Zn 2+ coordinating center are required for the formation of a polar ZitP  . PopZ complex.

ZitP controls PopZ localization
As PopZ interacts with the centromere (parS)-binding protein ParB and is required to anchor parS at both C. crescentus poles (Bowman et al., 2008;Ebersbach et al., 2008;Laloux and Jacobs-Wagner, 2013;Ptacin et al., 2014), we tested if the ZitP . PopZ complex associates, directly or indirectly, with parS in vivo. To this end, we conducted chromatin immunoprecipitation-deep-sequencing (ChIP-Seq) using antibodies recognizing ParB (anti-ParB) or ZitP (anti-ZitP specific for either the N-terminus or the C-terminus, Figure Figure 4A). Quantification revealed that DzitP DpopZ cells expressing mCherry-PopZ KE (impaired in ParB binding, DzitP mCherry-popZ KE ) had fewer bipolar foci compared to zitP + cells and an increase of cells without polar signals. Importantly, these abnormalities were not seen in zitP + cells, indicating that ZitP promotes polar localization of PopZ when the ParAB-dependent (and/or another) localization pathway requiring the PopZ N-terminal region is impaired. Consistent with these results, we found that the efficiency of plating (EOP) of DzitP cells harboring the popZ D26 allele (DzitP mCherry-popZ D26 ) that encodes a PopZ variant lacking the N-terminal 26 residues to bind ParAB and other client proteins (Holmes et al., 2016;Ptacin et al., 2014) is markedly reduced compared to zitP + cells ( Figure 4B  While exploring which region in ZitP is required to control PopZ, we noted that DzitP DpopZ cells expressing the mCherry-PopZ KE have a reduced growth rate in broth and that this defect is rescued by expression of (full-length) Dendra2-ZitP from the xylX locus using the xylose-inducible P xyl promoter ( Figure 4C). The C-terminal DUF3426 that is required for polar pilus biogenesis (Mignolet et al., 2016) is not required to improve growth , as revealed by growth measurements of cells expressing the ZitP (1-133) variant. However, when the interaction with PopZ is abrogated [in the ZitP (1-133W35I) mutant], cells grow poorly. By contrast, no dependency on ZitP is seen in DpopZ cells expressing otherwise unmodified mCherry-PopZ (mCherry-popZ, Figure 4C). Taken together, these results show that ZitP is important for growth and viability for DpopZ cells expressing PopZ that no longer interacts efficiently with ParAB, and that the ability of membrane-anchored ZitP to bind PopZ via the ZnR is important to regulate PopZ. Thus, there exist at least two (redundant) mechanisms of PopZ localization control in C. crescentus: one regulated directly or indirectly by ParAB (Laloux and Jacobs-Wagner, 2013;Ptacin et al., 2014) (or the N-terminal region of PopZ) and another that is modulated by membrane-anchored ZitP [ZitP  ].

ZitP imparts bipolarity upon PopZ in E. coli
Knowing that membrane-anchored ZitP controls PopZ localization in C. crescentus, we attempted to recapitulate these effects in a heterologous system by co-expression of ZitP  or ZitP (1-133) with mCherry-PopZ in E. coli. When ZitP (1-43) is co-expressed with mCherry-PopZ in this system, mCherry-PopZ forms a single cluster (Figure 5-figure supplement 1A) resembling the ones previously described (Bowman et al., 2008;Ebersbach et al., 2008;Laloux and Jacobs-Wagner, 2013) whose location is fairly sporadic and infrequently associated with the cell extremity (  Figure 5B). To confirm that Dendra2-ZitP associates with the polar membrane in E. coli rather than inclusion bodies, we localized it by super-resolution photo-activated localization microscopy (PALM) (Betzig et al., 2006) and observed the fluorescent signals to line exclusively the polar caps ( Figure 5C). By contrast, Dendra2-ZitP is dispersed along the cell envelope when PopZ is absent ( Figure 5C). Thus, these interdependent clusters of PopZ and ZitP at the polar caps are clearly distinct from the internal (cytoplasmic) aggregates typically seen for PopZ when expressed without ZitP in E. coli cells ( Figure 5A) or in division-inhibited (filamentous) E. coli (Laloux and Jacobs-Wagner, 2013) and that resemble fluorescent aggregates of inclusion bodies (Miller et al., 2015). Dendra2-ZitP (1-133) also induces the redistribution of the ParAB-blind version of mCherry-PopZ (mCherry-PopZ D26 ) in E. coli ( Figure 5-figure supplement 1B). By contrast, Dendra2-ZitP (1-43) is unable to do so, yet it still co-localizes (interacts) with mCherry-PopZ D26 . This indicates that Den-dra2-ZitP (1-43) can still interact with mCherry-PopZ D26 , but no longer controls its localization since Dendra2-ZitP (1-43) is not membrane-anchored. Thus, ZitP regulates mCherry-PopZ localization either via the conserved C-terminal DUF2497 domain that is conserved in PopZ orthologs or via sequences distal to the N-terminal (26) residues where the ParAB recognition sites reside. As these findings imply that ZitP and ParB do not compete for the same binding site in PopZ, we reasoned that it should be possible to reconstitute a tripartite bipolar complex by co-expression of PopZ and Den-dra2-ParB together with either mCherry-ZitP (1-133) or mCherry-ZitP (1-133)W35I in E. coli ( Figure 5D).  Figure 5E) revealed bipolar signals of mCherry-PopZ and CFP-ParB in the presence of ZitP  . By contrast only monopolar 'plugs' containing mCherry-PopZ and CFP-ParB were seen with the W35I derived or ZitP (1-90) co-overexpression plasmids at similar expression levels ( Figure 5figure supplement 2). In the latter condition, there is co-localization between PopZ and ParB. However, ParB is no longer at the periphery of bipolar PopZ in cells harboring the PopZ/ZitP or the PopZ/ZitP (1-133) co-overexpression plasmid. Cell length quantification revealed that cells containing monopolar 'plugs' are longer ( Figure 5E) than the PopZ/ZitP (1-133) co-overexpressing cells. As biochemical fractionation experiments confirmed the predicted localization of ZitP (1-43) and ZitP (1-133) ( Figure 5-figure supplement 1C), we conclude that membrane-anchored ZitP robustly and directly controls PopZ localization in E. coli and in C. crescentus.

Discussion
Eukaryotic zinc-finger (ZnR) domains are typically used to bind DNA, but they can also mediate protein-protein interactions (Klug, 2010). Here we described a small (43-residue) ZnR that acts as a conserved polar localization sequence, reminiscent of the nuclear localization signal (NLS) of eukaryotes, by promoting the interaction with the PopZ polar organizer of a-proteobacteria. This ZnR interacts Figure 5 continued ZitP, Dendra2-ZitP 1-133(WT) , Dendra2-ZitP 1-133(W35I) or Dendra2-ZitP 1-133(MalF-TM) . Cells were grown in LB for 2 hr, then Dendra2-ZitP and mCherry-PopZ were induced with 1 mM IPTG and 0.2% L-arabinose, respectively, for 2 hr before imaging. (C) PALM (photo-activated localization microscopy) images of E. coli cells expressing Dendra2-ZitP from pSRK-Km (Khan et al., 2008) and either no PopZ (empty pBAD101 (Guzman et al., 1995) vector, strain EC127, left panel) or untagged PopZ from pBAD101 (strain EC132, right panel). (D) Overlays of Dendra2-and mCherry-fluorescence with phase contrast images showing the co-localization of Dendra2-ParB (from P lac on pSRK-Km) with WT or W35I mCherry-ZitP 1-133 derivatives in E. coli cells co-expressed with untagged PopZ from P ara on pBAD22. (E) Overlays of CFP-and/or mCherry-fluorescence with phase contrast images of C. crescentus popZ:: mCherry-popZ parB::CFP-parB cells co-overexpressing (untagged) full-length ZitP, ZitP 1-90 , ZitP 1-133(W35I) or ZitP 1-133(WT) with (untagged) PopZ under P xyl control from pMT464. Over-expression was induced by growth in xylose 0.3% for 4 hr prior to imaging. mCherry-PopZ (upper panel) and CFP-ParB    Figure 6 continued on next page directly with PopZ and controls its localization from the membrane, even when PopZ no longer carries the interaction sites for ParAB, indicating that PopZ localization is regulated by redundant mechanisms in C. crescentus. One mechanism may require ParAB or at least it depends on the N-terminal region of PopZ that promotes the interaction with ParAB and other PopZ client proteins (Holmes et al., 2016). However, ZitP controls PopZ independently of this region, underscoring the multivalent interactions and multifunctional properties of PopZ. ZitP emerges as new a component of the a-proteobacterial polarity control pathway using an N-terminal ZnR to regulate PopZ localization. Interestingly, ZitP is bifunctional, controlling the localization of polar pilus biogenesis proteins via its C-terminal DUF3426 (Mignolet et al., 2016). While the DUF3426 is less conserved among the a-proteobacteria and dispensable for PopZ localization control, the membrane-anchored ZitP ZnR and PopZ binding is featured by many ZitP orthologs from the branch of obligate intracellular a-proteobacteria (the Rickettsiae) and regulates PopZ subcellular positioning.
The NMR solution structure of the ZitP ZnR revealed an atypical bbabb-architecture, resembling that of the crenarchaeal DNA-binding protein Cren7 (Zhang et al., 2010). However, the ZitP ZnR lacks a fifth b-strand (b5) that is used by Cren7 to establish critical sequence-specific contacts with DNA. An unexpected finding was that PopZ-bound ZitP associates indirectly with sites flanking the ParB-bound centromere (parS), rather than parS itself that ParB binds (Figure 4-figure supplement  1). We found that the interaction between PopZ and ParAB (Bowman et al., 2008;Ebersbach et al., 2008;Laloux and Jacobs-Wagner, 2013;Ptacin et al., 2014Ptacin et al., , 2010, or at least the N-terminal region of PopZ, is required for ZitP to associate with these parS-flanking sites. It is also possible that another (unknown) component(s) of the PopZ complex influences the association of ZitP at these sites or that it underlies an unprecedented mode of occupancy of the macromolecular ZitP.PopZ.ParB complex on the chromosome in proximity to parS. Interestingly, the ParB-centromeric complex was recently proposed to self-assemble via nucleation and 'caging' in vivo, i.e. threedimensional stochastic interactions between ParB and parS, and possibly reinforced or delimited by adjacent nucleoprotein complexes (Sanchez et al., 2015) or accessory factors that may allow supercomplexes to extend into neighboring chromosomal regions.
With the juxtaposition of PopZ . ZitP at the ParB-bound centromere and our reconstitution of the ZitP . PopZ . ParB tripartite complex in E. coli, ZitP is well positioned to exert control of ParB . parS dynamics and to impair segregation by sequestering the adjacent centromere into aberrant PopZ.ZitP clusters. The ParB-control function of PopZ.ZitP depends on the ability of membraneanchored ZitP ZnR to interact with PopZ. Our molecular dissections provide evidence that ZitP Figure 6 continued images of C. crescentus popZ::mCherry-popZ parB::CFP-parB cells over-expressing ZitP 1-104 from R. massiliae with C. crescentus PopZ. Over-expression was induced by the addition of 0.3% xylose for 6 hr prior to imaging. (C) Overlays of Dendra2-fluorescence with phase contrast images of the R. massiliae (Rm) ZitP ZnR version expressed from the xylX locus in WT and DpopZ C. crescentus cells. Synthesis of the Dendra2-ZitP 1-43(WT)Rm or Dendra2-ZitP 1-43(W35I) Rm was induced for 4 hr with 0.3% xylose before imaging. (D) Overlays of Dendra2-and/or mCherry-fluorescence with phase contrast images of E. coli TB28 cells expressing R. massiliae (Rm) Dendra2-ZitP 1-43(WT) (upper panels), Dendra2-ZitP 1-43(W35I) (lower panel) in the presence of mCherry-tagged PopZ from C. crescentus. Cells were grown in LB for 2 hr, then Dendra2-ZitP 1-43 variants and mCherry-PopZ were induced with 1 mM IPTG and 0.2% L-arabinose, respectively, for 2 hr. (E) Overlay of mCherry-fluorescence with phase contrast images of E. coli TB28 cells co-expressing mCherry-ZitP 1-104Rm from R. massiliae (Rm) with C. crescentus PopZ from P ara encoded on the same pBAD22-derived plasmid. Cells were grown in LB for 2 hr, then expression of mCherry-ZitP 1-104Rm and PopZ was induced with 0.2% L-arabinose for 2 hr before imaging. (F) Overlays of Dendra2-and/or mCherry-fluorescence with phase contrast images of E. coli TB28 cells expressing C. crescentus Dendra2-ZitP 1-43 (Cc) in the presence of mCherry-PopZ from R. massiliae (Rm). Cells were grown in LB for 2 hr, then Dendra2-ZitP 1-43 and mCherry-PopZ were induced with 1 mM IPTG and 0.2% L-arabinose, respectively, for 2 hr. (G) Images of E. coli cells co-expressing C. crescentus Dendra2-ZitP 1-43 (Cc, upper panel) or Dendra2-ZitP full-length (Cc, bottom panel) and PopZ from R. massiliae (Rm). Fluorescence (Dendra2) images (left panels) and overlays between phase contrast and Dendra2 fluorescence images (right panels) are shown. Cells were grown in LB to during 2 hr, then Dendra2-ZitP 1-43 variants were induced with 1 mM IPTG and PopZ was induced with 0.2% L-arabinose for 2 hr. (H) The (bi)polar PopZ . ZitP complex of free-living (Caulobacterales) and obligate intracellular (Rickettsiales) aproteobacteria. Pink dots denote PopZ monomers that assemble into a bipolar or monopolar patch, while blue dots denote ZitP molecules. An obligate intracellular rickettsial (rod with bipolar PopZ) cell is depicted within a vacuole (dashed structure) of a host cell (closed structure) and presumed also to polarize PopZ . ZitP (grey arrow). As ZitP is not present in the Rhizobiales, another mechanism of PopZ control is likely operational to drive it into a monopolar disposition. Similarly, we suggest that PopZ localization in C. crescentus can be accomplished by another pathway that operates independently of ZitP and likely involves ParAB and/or another pathway (see Discussion). DOI: 10.7554/eLife.20640.015 targets residues flanking the conserved C-terminal DUF2497 of PopZ. Thus, the DUF2497 region may serve as a key determinant in PopZ to regulate its association with (known and unknown) client proteins and its bipolar localization and/or dynamics.
The bipolar localization of DivIVA of the Gram-positive Firmicutes has been intensively studied and is governed by the biophysical properties of DivIVA that attract it to concavely curved membranes and thus both cell poles when expressed in E. coli (Lenarcic et al., 2009;Ramamurthi and Losick, 2009;Strahl and Hamoen, 2012). The a-proteobacterial PopZ system is of distinct ancestry and structural features compared to DivIVA, and it aggregates into (sporadic) monopolar and nonpolar (cytoplasmic) clusters in (filamentous) E. coli in the absence of ZitP. While these PopZ clusters do not rely on membrane curvature to find the E. coli poles, their localization mechanism is not understood (Laloux and Jacobs-Wagner, 2013) and could reflect aggregation in a partially folded form into cellular inclusion bodies (Miller et al., 2015). A clearly different mechanism underlies the robust assembly of (near native) bipolar PopZ . ZitP complexes at the polar membrane as observed by PALM. It is possible that these PopZ . ZitP complexes scout the poles by a similar mechanism as DivIVA, a possibility that warrants further investigation. As these bipolar PopZ . ZitP complexes in E. coli can also be reconstituted with components from different a-proteobacteria, it is unlikely that it is mediated by specialized phospholipids as the composition differs between E. coli, Caulobacterales (De Siervo, 1985;De Siervo and Homola, 1980) and other a-proteobacteria, for example the Rickettsiales.
There are intriguing evolutionary implications based on the finding that rickettsial ZitP retains the aforementioned PopZ control activities. It suggests that these obligate intracellular pathogens also feature a polarized PopZ . ZitP complex. The rickettsial actin filament organizer Sca2 is polarized and promotes actin-mediated motility of some Rickettsiae inside the host cell cytoplasm (Haglund et al., 2010;Madasu et al., 2013). It is therefore conceivable that the PopZ . ZitP bipolarity system served as a primordial platform ( Figure 6H) for the acquisition of specialized polarity based functions conferred via PopZ-interacting proteins that reinforce bipolarity and/or confer other polarized traits. For example, PopZ is also required for stalk biogenesis in C. crescentus (Bowman et al., 2010), a structure that requires specialized peptidoglycan (PG) biosynthesis enzymes (Kühn et al., 2010). As several Rhizobiales members also sequester PopZ to the site of PG synthesis to promote unipolar growth (Anderson-Furgeson et al., 2016;Brown et al., 2012;Curtis and Brun, 2014;Grangeon et al., 2015), the current challenge is to determine which proteins interact with PopZ in the different a-proteobacterial branches, particularly in the Rhizobiales. PopZ is unipolar in this branch (Deghelt et al., 2014;Grangeon et al., 2015) and no ZitP orthologs are encoded, suggesting that unknown PopZ-control mechanisms exist ( Figure 6H,Figure 1-figure supplement 1A). By contrast, many members of the Rhodobacterales, do not encode a conspicuous PopZ ortholog in their genome. This, along with the structural divergence of the ZitP ZnR in this lineage ( Figure 3C), suggests that ZitP has been appropriated for other functions or interactions with partners differing in primary structure.
Proper control of polarization is critical for efficient cellular proliferation and fitness in many cell types. In yeast, aberrant or misregulated polarity complexes impair the fitness of yeast cells to the extent where it is favorable to cells to eliminate these functions completely than retaining them improperly regulated (Laan et al., 2015). Similar selective forces must have ensured the evolution and retention of related polarity control modules in free-living and obligate intracellular bacteria since misregulation of polarity similarly perturbs the a-proteobacterial cell division cycle.

Strains, plasmids and oligos
Strains, plasmids and oligos are listed in Supplementary file 2. Plasmids expressing Dendra2 variants were used (Holden et al., 2014) and strains are derivatives of Caulobacter crescentus NA1000 whose genome is sequenced (Marks et al., 2010).
Swarmer cell isolation, electroporation, biparental mating (intergeneric conjugations) and bacteriophage jCr30-mediated generalized transduction were performed as described (Ely, 1991). Briefly, swarmer cells were isolated by Ludox or Percoll density-gradient centrifugation at 4˚C, followed by three washes and final re-suspension in pre-warmed (30˚C) M2G. Electroporation was done from 1 ml overnight cells that had been washed three times in sterile water. Biparental matings were done using exponential phase E. coli S17-1 donor cells and C. crescentus recipient cells washed in PYE and mixed at 1:10 ratio on a PYE plate. After 4-5 hr of incubation at 30˚C, the mixture of cells was plated on PYE harbouring nalidixic acid (to counter select E. coli) and the antibiotic that the conjugated plasmid confers resistance to. Generalized transductions were done by mixing 50 mL ultraviolet-inactivated transducing lysate with 500 mL stationary phase recipient cells, incubation for 2 hr, followed by plating on PYE containing antibiotic to select for the transduced DNA.

ZitP purification and production of antibodies
ZitP N-TER or C-TER recombinant protein, comprising only the first 90 amino acids or lacking the last 119 residues respectively, was expressed from pET28a in E. coli BL21(DE3)/ pLysS and purified under native conditions using Ni 2+ chelate chromatography. Cells were grown in LB at 37˚C to an OD 600nm of 0.6 and induced by the addition of IPTG to 1 mM during 3 hr, and harvested at 5000 RPM at 4˚C during 30 min. Cells were pelleted and re-suspended in 25 mL of lysis buffer (10 mM Tris HCl (pH 8), 0.1 M NaCl, 1.0 mM b-mercaptoethanol, 5% glycerol, 0.5 mM imidazole Triton X-100 0.02%). Cells were sonicated in a water-ice bath, 15 cycles 30 s ON; 30 s OFF. After centrifugation at 5000g for 20 min at 4˚C, the supernatant was loaded onto a column containing 5 mL of Ni-NTA agarose resin (Qiagen, Hilden, Germany) pre-equilibrated with lysis buffer. The column was rinsed with lysis buffer, 400 mM NaCl and 10 mM imidazole, both prepared in lysis buffer. Fractions were collected (in 300 mM Imidazole buffer, prepared in lysis buffer) and used to immunize New Zealand white rabbits (Josman LLC).

Fractionation
Fifty mL of an exponential culture of Caulobacter (OD 600nm = 0.4) was harvested by centrifugation for 15 min at 8000g at 4˚C. Cell pellets were re-suspended in 1 mL of lysis buffer (20 mM Tris-HCl pH 7.5, 300 mM NaCL, 0.5 mM EDTA, 5 mM MgCl 2 at 4˚C freshly supplemented with 1 mM DTT, 12500 U ready-lyse (Epicentre technologies), and one tablet of EDTA-free protease inhibitor cocktail ([Complete; Roche] per 50 mL). Ten mL of DNAse 1 mg.mL À1 , 5 mL of RNAseA 20 mg.mL À1 were added before sonication in an ice-water bath, 15 cycles 30 s ON; 30 s OFF. Twenty mL of this was mixed with 20 mL of loading buffer 2X (0.25 M Tris pH 6.8, 6% (wt/vol) SDS, 10 mM EDTA, 20% (vol/ vol) Glycerol) containing 10% (vol/vol) b-mercaptoethanol to obtain the crude extract (CE). Sonicated samples were centrifuged for 30 min at 20,000g at 4˚C, and the supernatant was diluted in 2X loading buffer to obtain the soluble fraction (S). The pellet, containing the insoluble fraction, was resuspended in 1 mL resuspension buffer (20 mM Tris-HCl pH 7.5, 300 mM NaCl, 5 mM EDTA, 1 mM DTT) and 20 mL was diluted in loading buffer 2X to obtain the insoluble fraction (P). The membrane fraction was split in 3 fractions of 300 mL to add 300 mL re-suspension buffer (control) or 300 mL 4M NaCl buffer (20 mM Tris-HCl pH 7.5, 4 M NaCl, 5 mM EDTA, 1 mM DTT) (solubilize proteins associated with membrane) or 300 mL 2% Triton-X100 (20 mM Tris-HCl pH 7.5, 300 mM NaCl, 5 mM EDTA, 1 mM DTT, 2% Triton-X100) (solubilize integral membrane proteins). Samples were incubated 1 hr with shaking at 4˚C and harvested by centrifugation at 20,000g during 30 min at 4˚C. Forty mL of the supernatant was diluted in 40 mL loading buffer 2X to obtain the soluble fraction (S). The pellet, containing the insoluble fraction, was resuspended in 600 ml of resuspension buffer. Forty mL of this was diluted in 40 mL loading buffer 2X to obtain the insoluble fraction (P) All the fraction were analysed by immunoblot using antibodies to ZitP (NTER), DivJ (as control of integral membrane protein) and CtrA (as control of soluble proteins).

Whole-cell extracts preparation
Five hundred mL of an exponential Caulobacter or E. coli cells (OD 600nm = 0.4 and 0.8 respectively) were harvested with 20,000g at room temperature for 5 min. Whole-cell extracts were prepared by resuspension of cell pellets in 75 mL TE buffer (10 mM Tris-HCl pH 8.0 and 1 mM EDTA) followed by addition of 75 mL loading buffer 2X (0.25 M Tris pH 6.8, 6% (wt/vol) SDS, 10 mM EDTA, 20% (vol/ vol) Glycerol) containing 10% (vol/vol) b-mercaptoethanol. Samples were normalized for equivalent loading using OD 600nm and were heated for 10 min at 90˚C prior to loading.

ChIP-SEQ
Mid-log phase cells were cross-linked in 10 mM sodium phosphate (pH 7.6) and 1% formaldehyde at room temperature for 10 min and on ice for 30 min thereafter, washed three times in phosphatebuffered saline (PBS) and lysed in a Ready-Lyse lysozyme solution (Epicentre Technologies) according to the manufacturer's instructions. Lysates were sonicated in a ice-water bath, 15 cycles 30 s ON; 30 s OFF to shear DNA fragments to an average length of 0.3-0.5 kbp and cleared by centrifugation at 14,000 g for 2 min at 4˚C. Lysates were normalized by protein content, diluted to 1 mL using ChIP buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl (pH 8.1), 167 mM NaCl plus protease inhibitors (Roche, Switzerland) and pre-cleared with 80 ml of protein-A agarose (Roche) and 100 mg BSA. To immunoprecipate the chromatin, two mL of polyclonal antibodies were added to the supernatant, incubated overnight at 4˚C with 80 mL of protein-A agarose beads pre-saturated with BSA. Antibodies to ZitP NTER, ZitP CTER and ParB (Mohl and Gober, 1997) were used for ChIP. The immunoprecipitate was washed once with low salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.1) and 150 mM NaCl), high salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl (pH 8.1) and 500 mM NaCl) and LiCl buffer (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA and 10 mM Tris-HCl (pH 8.1)), and twice with TE buffer (10 mM Tris-HCl (pH 8.1) and 1 mM EDTA). The protein DNA complexes were eluted in 500 mL freshly prepared elution buffer (1% SDS and 0.1 M NaHCO3), supplemented with NaCl to a final concentration of 300 mM and incubated overnight at 65˚C to reverse the crosslinks. The samples were treated with 2 mg of Proteinase K for 2 hr at 45˚C in 40 mM EDTA and 40 mM Tris-HCl (pH 6.5). DNA was extracted using phenol:chloroform:isoamyl alcohol (25:24:1), ethanol precipitated using 20 mg of glycogen as carrier and resuspended in 100 mL of water.
Immunoprecipitated chromatin was used to prepare sample libraries used for deep-sequencing at Fasteris SA (Geneva, Switzerland). ChIP-Seq libraries were prepared using the DNA Sample Prep Kit (Illumina) following manufacturer's instructions. Single-end run were performed on an Illumina Genome Analyzer IIx or HiSeq2000, 50 cycles were read and yielded several million reads. The single-end sequence reads stored in FastQ files were mapped against the genome of Caulobacter crescentus NA1000 (NC_011916) and converted to SAM using BWA and SAM tools respectively from the galaxy servor (https://usegalaxy.org/). The resulting SAM was imported into SeqMonk (http:// www.bioinformatics.babraham.ac.uk/projects/seqmonk/, version 0.21.0) to build sequence read profiles. The initial quantification of the sequencing data was done in SeqMonk: the genome was subdivided into 50 bp probes, and for every probe we calculated a value that represents a normalized read number per million. All ChIP-seq data was deposited in the GEO database under accession number GSE79918 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE79918).

Isothermal titration calorimetry (ITC)
ITC experiments were performed on a VP-ITC instrument (Microcal). Both partners were prepared in the same NMR spectroscopy buffer, ZitP at 0.015 mM and PopZ at 0.300 mM. ZitP was titrated using a solution of PopZ by 65 injections of 4 mL every 300 s at 20˚C. The raw data were integrated, normalized for the molar concentration and analyzed using Origin7.0 according to a 1:1 binding model.

Growth measurements in broth
The overnight cultures were started in PYE supplemented with D-xylose 0.3% final. The cultures were diluted to obtain an OD 600nm of 0.1 in PYE supplemented with D-xylose 0.3% final. The OD 600nm was recorded every hour during 9 hr. The graph represents the trend of the growth curve of three independent experiments.

Plasmid constructions pMB123 (pNTPS138-DzitP)
The plasmid construct used to delete zitP (CCNA_02298) was made by PCR amplification of two fragments. The first, a 771 bp fragment was amplified using primers OMB87 and OMB88 (table S?), flanked by an EcoRI and a BamHI site. The second, a 889 bp fragment was amplified using primers OMB89 and OMB90, flanked by a BamHI site and a HindIII site. These two fragments were first digested with appropriate restriction enzymes and then triple ligated into pNTPS138 (M.R.K. Alley, Imperial College London, unpublished) previously restricted with EcoRI/HindIII.
pMB38 (pXdendra2-N2-zitP 1-87(Mm) ) The first 261 nt of the Mmar10_2373 coding sequence were PCR amplified from the synthetic fragment two using the Van and T7pro primers. This fragment was digested with SacI/EcoRI and cloned into SacI/EcoRI-digested pXdendra2-N2.
pMB127 (pMT464-zitP 1-132malF ) The first 396 nt of the ZitP coding sequence was PCR amplified from fragment 13 (containing the MalF transmembrane domain instead the natural transmembrane domain from ZitP) using the Van and T7 primers. This fragment was digested with NdeI/EcoRI and cloned into NdeI/EcoRI-digested pMT464.
pMB156 (pMT464-zitP 1-120WT(Bd) ) The first 360 nt of the HMPREF0185_02600 coding sequence was PCR amplified from fragment 14 (containing the ZnR from B. diminuta and its transmembrane domain) using the Van and T7 primers. This fragment was digested with NdeI/EcoRI and cloned into NdeI/EcoRI-digested pMT464.
pMB114 (pMT464-zitP 1-104WT(Rm) ) The first 312nt of the ZitP coding sequence was PCR amplified from fragment 15 (containing the ZnR from R. massiliae and its transmembrane domain) using the Van and T7 primers. This fragment was digested with NdeI/EcoRI and cloned into NdeI/EcoRI-digested pMT464. pMB177 (pMT464-zitP 1-133W35I -popZ) The PopZ coding sequence was PCR amplified from NA1000 using the OMB78 and OMB79 primers adding an EcoRI plus an RBS and an XbaI site respectively. This fragment was digested with EcoRI/ XbaI and cloned into EcoRI/XbaI -digested pMB113. pSC : pSC is a derivative of pET26b (Novagen) in which expression of the protein of interest is in frame with a coding sequence for C-terminal thrombin cleavage site and a His 6 -tag. The coding sequence can be cloned between NdeI and XhoI. An NheI site was also introduced after the stop codon that enables an easy construction of polycistronic synthetic genes.

pMB64 (pSC-popZ)
The PopZ coding sequence (codon optimized for E. coli) was PCR amplified from fragment 16 using Van and T7 primers. This fragment was digested with NdeI/XhoI and ligated into NdeI/XhoIdigested pSC.
pMB54 (pSC-zitP 1-90 ) The first 270 nt of the ZitP coding sequence (codon optimized for E. coli) was PCR amplified from fragment 17 using Van and T7 primers. This fragment was digested with NdeI/XhoI and ligated into NdeI/XhoI-digested pSC.

pET28a-zitP Cterm
The zitP-coding sequence lacking the first 189 bp from the start codon (nt 1082122-1082481) was PCR amplified from the NA1000 strain using the zitP_Cterm_nde and zitP_Cterm_eco primers. This fragment was digested with NdeI/EcoRI and cloned into NdeI/EcoRI-digested pET28a (Novagen).
pMB140 (pSC-zitP 1-43W35I ) The first 129 nt of the ZitP coding sequence (carrying the W35I mutation and codon optimized for E. coli) was PCR amplified from fragment 19 using Van and T7 primers. This fragment was digested with NdeI/XhoI and ligated into NdeI/XhoI-digested pSC.