A network of filament-forming proteins maintains multicellular shape in the cyanobacterium Anabaena sp. PCC 7120

The determinants of bacterial cell shape are extensively studied in unicellular forms. Nonetheless, the mechanisms that shape bacterial multicellular forms remain understudied. Here we study coiled-coil rich proteins (CCRPs) in the multicellular cyanobacterium Anabaena sp. PCC 7120 (hereafter Anabaena). Our results reveal two CCPRs, Alr4504 and Alr4505 (termed LfiA and LfiB for linear filament), which assemble into a heteropolymer in vivo and in vitro. Two additional CCRPs, Alr0931 (termed CypS for cyanobacterial polar scaffold) and All2460 (termed CeaR for cyanobacterial elongasome associated regulator), form a polar proteinaceous scaffold and are associated with MreB activity, respectively. Deletion mutants of these CCRPs are characterized by impaired trichome (i.e. cyanobacterial filament) and cell shape and decreased viability. All four CCRPs interacted with each other, with the septal junction protein SepJ and all but CypS interacted with MreB. Our results indicate that filament-forming CCRPs are present in cyanobacteria and that they, likely in cooperation with SepJ and MreB, could form a proteinaceous network that stabilizes the Anabaena trichome. We propose that this network is essential for the manifestation of the linear trichome phenotype in Anabaena. Importance The phylum Cyanobacteria is characterized by a large morphological diversity, ranging from coccoid or rod-shaped unicellular species to species forming multicellular morphology, which comprise several cells connected into a linear form. Despite this diversity, very few molecular mechanisms underlying the cyanobacterial morphological diversity are known. Among these, the cytoskeletal proteins FtsZ and MreB are important regulators of cyanobacterial cell shape and viability. The multicellular phenotype of cyanobacteria has been linked also to the septal junctions, which comprise a pretentious complex dividing between neighboring cells in the linear form. In our research we identified and characterized four proteins that are involved in cell and trichome shape regulation in multicellular cyanobacteria. We show that two of those proteins are interdependent for polymerization, revealing a novel feature for prokaryotic filament-forming proteins. Our study leads to a broader understanding of the underlying principles of cyanobacterial morphological diversity.


In vitro and in vivo filamentation of CCRP candidates 154
Out of the 13 candidates, four CCRPs showed self-association and protein filamentation 155 properties, including LfiA together with LfiB, CypS and All4981 (that will be investigated in a 156 separate report). The remaining nine candidates failed to form filamentous structures in vivo 157 and in vitro and were excluded from further analysis. An exception is CeaR whose in silico 158 prediction suggested similarities to the well-characterized prokaryotic IF-like protein 159 crescentin8 (Supplementary Table 1). Notably, none of those four proteins has homologs in the 160 SynProCya clade of unicellular cyanobacteria (Fig. 1a), hinting for a function in more complex 161 cyanobacterial morphotypes. To evaluate the ability of the four candidates to self-associate, 162 we ectopically expressed C and N-terminally tagged (His6, YFP, eCFP or GFP) recombinant 163 proteins and investigated in vitro polymerization properties and in vivo localization pattern. The 164 assembly of CCRPs into filaments in vitro was tested by fluorescence microscopy using the 165 NHS-Fluorescein dye, which was previously successfully used to visualize in vitro FtsZ protein 166 filaments36. Similar to previous investigations of filament-froming CCRPs8,9,20, we purified His6-167 tagged CCRPs by Ni-NTA affinity chromatography under denaturing conditions and renatured 168 them by dialysis followed by NHS-fluorescein staining. We note that the His6-tag has previously 169 been shown to have no impact on the in vitro polymerization properties of the CCRP FilP20. As 170 a positive control for our approach we used crescentin. The NHS-fluorescein staining of 171 crescentin revealed an extensive filamentous network in our in vitro assay ( Supplementary Fig.  172 2), showing that our approach is applicable for filament-forming CCRPs. As negative controls 173 we included empty vector-carrying BL21 (DE3) cells, GroEL1.2 from Chlorogloeopsis fritschii 174 PCC 6912 (known to self-interact37) and the maltose binding protein (MBP), all of which were 175 tested negatively for filament formation in vitro using our approach. While neither, the cell-free 176 extract nor the highly soluble MBP protein formed any discernible structures in vitro, GroEL1.2 177 aggregates could be indicative for an uncontrolled oligomerization ( Supplementary Fig. 2). 178 Additionally, we observed similar clumps of protein aggregates from other negatively tested 179 Anabaena CCRPs, suggesting that this is a common observation for putative oligomerizing 180 proteins. 181

LfiA and LfiB are interdependent for polymerization in vitro and in vivo 182
Since the candidate proteins were annotated as hypothetical proteins, we initially investigated 183 and confirmed the transcription of all four genes under standard (BG11) and diazotrophic 184 (BG110) growth conditions ( Supplementary Fig. 3a,b). An additional inspection of the genomic 185 loci suggested that lfiA and lfiB are encoded in an operon structure, however, RT-PCR data 186 indicated that they are not co-transcribed ( Supplementary Fig. 3a,c). Applying our in vitro 187 polymerization assay to renatured and purified LfiA revealed amorphous non-filamentous 188 protein aggregates while LfiB assembled into aggregated and sheet-like structures (Fig. 1b). 189 We note, however, that the vast majority of LfiB protein precipitated into clumps of aggregates 190 upon renaturation, suggesting that LfiB has only a partial capacity to form filaments or is 191 unstable in vitro. Inspired by the close genomic localization of lfiA and lfiB, we next tested for 192 co-polymerization of both proteins. Upon co-renaturation, LfiA and LfiB co-assembled into a 193 meshwork of protein heteropolymers (Fig. 1c). While both, LfiA and LfiB renatured alone 194 formed aggregates in the dialysis tubes that were detectable with the naked eye (similar to 195 GroEL1.2), the co-renatured LfiA/B sample remained in solution, a known property of 196 eukaryotic IFs38. We also observed that this co-filamentation is dosage-dependent as only GFP from their respective native promoters (as predicted using BPROM41) revealed no 204 discernible expression of LfiB-GFP ( Supplementary Fig. 5a). Consequently, we investigated 205 the in vivo localization of both proteins from the copper-regulated petE promoter (PpetE), which 206 has previously been used to study the localization of FtsZ and MreB in Anabaena26,39,40. We 207 generally observed that the PpetE-driven gene expression does not always lead to expression 208 of the fusion protein in every cell under standard growth conditions. Notably, this was not 209 observed under diazotrophic growth conditions (BG110) or upon supplementation with 210 additional CuSO4. The expression of LfiA-GFP and LfiB-GFP from PpetE in Anabaena 211 independently did not reveal filamentous structures (Fig. 1d). However, upon co-expression of 212 LfiA-eCFP and LfiB-GFP from PpetE, a distinct filamentous structure along the longitudinal cell 213 axis could be observed (Fig. 1e). To confirm that the localization of LfiA-GFP and LfiB-GFP is 214 not affected by the wildtype (WT) lfiA or lfiB alleles, we localized both proteins individually in a 215 ΔlfiAΔlfiB double mutant; this revealed the same localization pattern as in the WT 216 ( Supplementary Fig. 5b), suggesting that co-polymerization is a dosage-dependent process. . 217 We further validated the in vivo co-polymerization of LfiA and LfiB by heterologous expression 218 in E. coli, which also revealed an interdependent polymerization pattern (Supplementary Fig.  219 5c 7). The intracellular localization of the LfiA/B heteropolymer in Anabaena suggests that the 220 polymer is either anchored at the cell poles or specifically broken up during cell division, as 221 LfiA/B filaments were never observed to cross cell-cell borders and only traversed through not 222 yet fully divided cells (Figs. 1e inlay and 1f). 223

CypS localizes to the cell poles in Anabaena 224
Applying our in vitro polymerization assay to CypS revealed that CypS assembled into star-225 like filamentous strings (Fig. 1b). The expression of CypS-GFP in Anabaena WT cells from the 226 predicted native promoter (PcypS; using BPROM) did not reveal coherent fluorescence signals 227 ( Supplementary Fig. 6a). When expressed from PpetE, CypS-GFP was localized to the cytosol 228 and the cell envelope ( Supplementary Fig. 6a). The same localization to the cell envelope and 229 the cytoplasm was also observed upon expression of CypS-GFP from PcypS in a ΔcypS mutant 230 background ( Supplementary Fig. 6a), suggesting that CypS concentration is strictly controlled 231 in the cells. Notably, CypS-GFP only partially complemented the ΔcypS mutant swollen cell 232 phenotype ( Supplementary Fig. 6a). Consequently, we examined whether the addition of a C-233 terminal His6-tag can reconstitute the WT phenotype and found that CypS-His6 expressed from 234 PpetE can complement the ΔcypS mutant ( Supplementary Fig 9a,c). To localize CypS-His6 in 235 the cell, we performed anti-His immunofluorescence staining and found that CypS-His6 forms 236 plugs at the cell poles (Fig. 1g) that appeared to displace the thylakoid membranes 237 ( Supplementary Fig. 6b). Notably, similar polar plugs could also be observed for CypS-GFP 238 but only after additional induction of protein expression ( Supplementary Fig. 6a), 239 demonstrating that the GFP-tag only partially interferes with CypS localization. Further 240 induction of CypS-His6 expression led to the formation of swollen cells ( Supplementary Fig.  241 6b), indicating that CypS has morphogenic properties. Since we found that CypS forms polar 242 plugs and that the LfiA/B filament generally ended at the cell poles, we tested for a possible 243 nonetheless no filamentous CeaR in vitro structures were observed. We note that genomes of 251 unicellular cyanobacteria do not have a homologous gene to ceaR (Fig. 1a), and furthermore, 252 unlike CypS, LfiA and LfiB, recombinant expression of CeaR-GFP in Synechocystis was 253 unsuccessful. This indicates that CeaR function is specific to the multicellular cyanobacterial 254 phenotype. Expression of a functional CeaR-GFP fusion protein (functionality of the fusion 255 protein is shown in Supplementary Fig. 11b,d,e) from PceaR and from PpetE in Anabaena WT 256 showed that the protein localized to the cell periphery in a patchy pattern (Fig. 1h, 257 Supplementary Fig. 8a), yet it also accumulated at the septa or at the Z-ring. Z-ring or septal 258 localization was found in 25% of cells (589 out of 2301 counted cells) carrying PpetE::ceaR-gfp 259 and in 17% of cells (206 out of 1237 counted cells) carrying PceaR::ceaR-gfp. In addition, we 260 observed that the expression of CeaR-GFP from PpetE led to a swollen cell phenotype in a large

308
The round and swollen cell phenotypes of the ∆cypS and ∆lfiA∆lfiB mutant strains are 309 indicative of an impairment in cell wall integrity and/or defects in PG biogenesis as well as an 310 elevated sensitivity to turgor pressure28,42. Consequently, we tested the sensitivity of the CCRP 311 mutants to cell wall degrading enzymes and osmotic stressors. This showed that the ∆cypS 312 mutant had an elevated sensitivity to lysozyme, suggesting a defect in cell wall integrity in both 313 mutants. No increased sensitivity to Proteinase K was identified in any of the mutants (Fig 2d). 314 . An increased sensitivity to lysozyme has previously been associated with a defect in 315 elongasome function26, suggesting that CypS could be associated with the Anabaena 316 elongasome. Furthermore, ∆cypS and ∆lfiA∆lfiB mutants were unable to grow in liquid culture 317 ( Supplementary Fig. 10a), with ∆cypS mutant cells readily bursting upon transfer to liquid 318 culture ( Supplementary Fig. 10b), hinting for an elevated sensitivity to fluid shear stress or 319 turgor pressure. In contrast, the ∆ceaR mutant was unaffected by the presence of cell wall 320 stressors (Fig. 2d) and grew well in BG11 growth medium (Fig. 2b). However, upon nitrogen 321 stepdown (i.e., transfer into BG110), the ΔceaR mutant readily fragmented into shorter 322 trichomes that aggregated into large-scale cell clumps (Fig. 2b, Supplementary Figs. 10a and  323 11d,e). Cells in those clumps also gradually lost their chlorophyll auto-fluorescence signal (an 324 indicator for cell viability) and ultimately died within a few days (Fig 2b, Supplementary Fig.  325 11a,b,d), revealing a viability defect of the ΔceaR mutant under diazotrophic conditions. The 326 defect in trichome viability could be complemented with pRL25C carrying PceaR::ceaR or 327 PceaR::ceaR-gfp ( Supplementary Fig. 11b, Fig. 14a-c). We suggest that the decreased 347 nanopores per septal disk may be responsible for the decrease in solute diffusion in both 348 mutants. In addition, we observed that some nanopores in the ∆ceaR mutant strain were large 349 and irregular ( Supplementary Fig. 14B), which as well could contribute to the altered efficiency 350 in solute diffusion. 351

Anabaena CCRPs affect MreB localization 352
The swollen cell phenotype of an Anabaena ΔmreB mutant has been previously reported to 353 have no effect on intracellular structures26. To assess whether the altered cell and trichome 354 shape of Anabaena CCRP mutant strains had any effect on intracellular arrangements, we 355 compared ultrathin sections of Anabaena WT and CCRP mutants. Except for ΔceaR mutant 356 cells that contained linear seemingly void structures, which did not represent thylakoid 357 membranes, intercellular ultrastructures of the CCRP mutants were unaffected regardless of 358 their impact on trichome viability and shape ( Supplementary Fig. 15). Using epifluorescence 359 microscopy, the void linear structures in the ΔceaR mutant were observed as prominent red 360 autofluorescence signals (Fig. 3d, Supplementary Fig. 11e and 9e), as such, we named them 361 red fluorescent filaments, whose nature is yet to be elucidated. Nonetheless, the observed cell 362 wall and cell and trichome defects indicated that CypS, LfiA/LfiB and CeaR function is related 363 to PG biogenesis, possibly through association with FtsZ or MreB. To test for a link with the 364 FtsZ-driven divisome, we visualized Z-ring placement in the WT and the mutants by 365 immunofluorescence. However, no alterations in Z-ring placement were observed, indicating 366 that Z-ring formation is unaffected in the mutants ( Supplementary Fig. 16). To test for an 367 association with the elongasome, we localized a functional PpetE::gfp-mreB fusion26 in 368 Anabaena WT and the CCRP mutants. Unlike in the previously reported PpetE::gfp-mreBCD 369 overexpression strain26, we never saw polar aggregates in our GFP-MreB-expressing strain 370 CCRPs were all seemingly linked to MreB localization or PG biogenesis, we next investigated 423 whether the four proteins interact with each other and with other known morphological 424 determinants in Anabaena. Using bacterial two hybrid assays (BACTH), we found that all our 425 four CCRPs were able to self-interact (Fig 4a). Additionally, all four CCRPs could cross-interact 426 with each other and we found that LfiA, LfiB and CeaR but not CypS, interacted with MreB. interaction of CeaR with SepJ in Anabaena (Fig. 4b). Corroborating a role of CeaR in PG 450 biogenesis and MreB function, CeaR was also found to be associated with three penicillin 451 binding proteins (Fig. 4b), which are known regulators of PG synthesis and are part of the 452 elongasome50. Furthermore, both, CeaR and LfiA, co-precipitated ParA, and CeaR was 453 additionally found to be associated with MinD (Fig. 4b). Both ParA and MinD belong to a protein 454 family of Walker-A-type ATPases and mediate plasmid and chromosome segregation51. To 455 test for a similar function in our CCRPs, we compared the DNA distribution among the CCRP 456 mutant cells as measured by distribution of 4′,6-Diamidin-2-phenylindol (DAPI) staining 457 intensity ( Supplementary Fig. 19). For that, we calculated the width of the DAPI focal area as 458 the range of DAPI staining around the maximum intensity focus (±10 grey intensity in arbitrary 459 units). This revealed that the staining focal area size was significantly different among the four 460 tested strains (P=3.14x10-41, using Kruskal-wallis). Post-hoc comparison showed that the focal 461 area size in the ΔceaR mutant was larger than the others, and the area size in Anabaena WT 462 was not significantly different than ΔcypS. The DAPI signal observed in the ΔlfiAΔlfiB mutant 463 appears as the most condensed, and indeed, the ΔlfiAΔlfiB mutant focal DAPI area was 464 smallest in comparison to the other strains (alpha=0.05, using Tukey test; Supplementary Fig.  465 19a,b). Unlike the ΔceaR mutant and the WT, DAPI signals in the ΔlfiAΔlfiB and ΔcypS mutant 466 strains were also observed between two neighboring cells ( Supplementary Fig 19a), indicating 467 that DNA distribution is not properly executed in those strains. 468   (Fig. 3d, Supplementary Fig. 9e).  Note: Structural similarities and conserved domains are inferred from in silico-based prediction tools and are not based on actual experimentally identified protein structures but simply serve as a mean to identify similarities between known filament-forming proteins and candidate CCRPs. The identified similarities might be subject to prediction bias due to the repeated nature of coiled-coil motifs. The first column indicates the respective locus tags of protein candidates and Crescentin. The second and third column indicate the respective subsection of the corresponding genus according to Rippka et al. (1979)