Axenic Biofilm Formation and Aggregation by Synechocystis PCC 6803 is Induced by Changes in Nutrient Concentration, and Requires Cell Surface Structures

Phototrophic biofilms are key to nutrient cycling in natural environments and bioremediation technologies, but few studies describe biofilm formation by pure (axenic) cultures of a phototrophic microbe. The cyanobacterium Synechocystis sp. PCC 6803 (hereafter Synechocystis) is a model micro-organism for the study of oxygenic photosynthesis and biofuel production. We report here that wild-type (WT) Synechocystis caused extensive biofilm formation in a 2000 liter outdoor non-axenic photobioreactor under conditions attributed to nutrient limitation. We developed a biofilm assay and found that axenic Synechocystis forms biofilms of cells and extracellular material, but only when induced by an environmental signal, such as by reducing the concentration of growth medium BG11. Mutants lacking cell surface structures, namely type IV pili and the S-layer, do not form biofilms. To further characterize the molecular mechanisms of cell-cell binding by Synechocystis, we also developed a rapid (8 hour) axenic aggregation assay. Mutants lacking Type IV pili were unable to aggregate, but mutants lacking a homolog to Wza, a protein required for Type 1 exopolysaccharide export in Escherichia coli, had a super-binding phenotype. In WT cultures, 1.2x BG11 induced aggregation to the same degree as 0.8x BG11. Overall, our data support that Wza-dependant exopolysaccharide is essential to maintain stable, uniform suspensions of WT Synechocystis cells in unmodified growth medium, and this mechanism is counter-acted in a pili-dependent manner under altered BG11 concentrations. Importance Microbes can exist as suspensions of individual cells in liquids, and also commonly form multicellular communities attached to surfaces. Surface-attached communities, called biofilms, can confer antibiotic resistance to pathogenic bacteria during infections, and establish food webs for global nutrient cycling in the environment. Phototrophic biofilm formation is one of the earliest phenotypes visible in the fossil record, dating back over 3 billion years. Despite the importance and ubiquity of phototrophic biofilms, most of what we know about the molecular mechanisms, genetic regulation, and environmental signals of biofilm formation comes from studies of heterotrophic bacteria. We aim to help bridge this knowledge gap by developing new assays for Synechocystis, a phototrophic cyanobacterium used to study oxygenic phototsynthesis and biofuel production. With the aid of these new assays, we contribute to the development of Synechocystis as a model organism for the study of axenic phototrophic biofilm formation.

protein specific to anchoring Type 1 capsule to the cell surface [30,37]. We did not find 66 a homolog to Wzi in Synechocystis using BLASTP search. This is consistent with 67 previously reported detection of RPS in WT Synechocystis supernatants, and reduced 68 levels of this RPS in supernatants of wzc mutant cultures [29,38]. The role of 69 Synechocystis RPS in cell binding, if any, is not known. 70 Cellulose is a component of EPS associated with the cell surface, promoting adhesion 71 in many heterotrophic bacteria [39]. Cellulose has also been detected in the EPS of 72 several cyanobacterial species [40,41]. Instead of cellulose synthases like BcsA that are 73 common in heterotrophic bacteria, cyanobacteria use CesA, the cellulose synthase 74 protein conserved in higher plants [42,43]. Synechococcus PCC 7002 [41] and 75 Thermosynechococcus vulcanus RKN [40,44] were shown to have cellulose-dependent 76 aggregation. Synechocystis contains one cellulose synthase motif, DDD35QXXRW, in 77 Sll1377, which has homology to the N-terminal region (48% query coverage) of CesA in 78 Thermosynechococcus vulcanus RKN (BLASTP search [45] using default parameters). 79 In a study of 12 diverse cyanobacterial species, cellulose was not detected in 80 Synechocystis or Synechococcus elongatus PCC 7942 (49) in non-aggregated cultures. 81 From these findings, it is inconclusive whether Synechocystis synthesizes cellulose during 82 aggregation. 83 A second line of evidence in the literature ties cellulose-dependent aggregation to 84 environmental signals via highly conserved role of secondary messenger c-di-GMP 85 (cyclic di-guanosine monophosphate) [46,47]. Nutrient limitation such as carbohydrate 86 starvation is frequently correlated with increased c-di-GMP levels in heterotrophs [48]; 87 carbohydrate starvation results in both nutrient and energy limitation in heterotrophs. 88 The cyanobacterium Thermosynechococcus vulcanus RKN undergoes cellulose-dependent 89 aggregation in response to blue light [40,44], an energy-limited condition, in a process 90 requiring formation of c-di-GMP by the protein SesA (Tlr0924). Two studies show that 91 c-di-GMP levels are correlated with aggregation in Synechocystis under different 92 energy-limiting conditions, although cellulose measurements were not reported [49,50]. 93 In addition to the role of EPS in biofilm formation and aggregation, we also 94 investigated the roles of S-layer and pili. The Synechocystis S-layer protein (Sll1951) is 95 glycosylated and forms a surface layer with hexagonal symmetry, which can be imaged 96 October 22, 2018 3/33 via TEM as a honeycomb-like surface texture on WT cells [51,52]. Heterotrophic 97 S-layer mutants have a range of adhesion phenotypes, ranging from super-binding to 98 completely biofilm deficient, depending on species (reviewed in [21]). S-layer mutants 99 have enhanced biofilm formation compared to WT (a super-binding phenotype) in 100 Bacillus cereus [53], Caulobacter crescentus [54], and various Clostridium difficile 630 101 strains [55]. In contrast, S-layer mutants of Streptococcus gordonii are deficient in 102 aggregation [56]. 103 There are six major classes of pili (and / or homologous structures called fimbriae 104 and curli), which pathogens such as E. coli, Pseudomonas aeruginosa, Salmonella 105 enterica, and Neisseria species use for attachment, adhesion, and biofilm formation 106 during infection (reviewed in [57]). Additionally, biofilm initiation depends on motility 107 in many bacterial species (reviewed in [58], including gliding motility conferred by 108 Type IV pili. Synechocystis has thousands of Type IV pili arranged peritrichously and 109 extending several microns beyond the cell surface [29,59]. These Type IV pili are 110 glycosylated along their entire length. Mutations causing altered glycosylation of PilA, 111 the pilin structural subunit, cause defects in gliding motility in Synechocystis [60]. PilC 112 is a predicted cytoplasmic chaperone protein for export of PilA monomer for pilin 113 assembly in diverse bacteria. Consistent with this prediction, Synechocystis pilC 114 deletion mutants (slr0162-0163 ) are apiliate (bald) [ molecule that inhibits secretion of biofilm enhancing proteins in WT S. elongatus [63]. 119 A subsequent study found certain piliated S. elongatus mutants also underwent a degree 120 of sedimentation, adhesion, and biofilm formation compared to WT, indicating that loss 121 of Type IV pili is not a prerequisite for biofilm formation in S. elongatus [64].

122
In this study, we document extensive biofilm formation in a large outdoor 123 photobioreactor used to grow WT Synechocystis. We adapted the crystal violet assay 124 commonly used for biofilm study of heterotrophic bacteria in order to screen conditions 125 leading to biofilm formation by axenic WT Synechocystis cultures. We also developed 126 rapid aggregation and flocculation assays to further characterize environmental signals 127 and cell surface biochemistry of cell binding. We engineered targeted genetic mutations 128 of genes sll1581 (wza), slr0923 (wzc), sll1951 (S-layer), and slr0162-0163 ( h over a period of five days). Biofouling such as representative images in Figure 1 was correlated with using hard tap water to prepare BG11; no biofouling was evident when softened tap water was used. confocal microscopy in Figure 2B revealed that isolated micro-colonies were 168 approximately 200-300 microns wide, 1-2 cells tall, and uniformly distributed on the 169 submerged portion of the coverslips. Figure 2C shows that stained material bound to 170 coverslips included cells and extracellular material.

172
We used allelic exchange of the KmRsacB markers with genes essential for Type IV 173 pili (pilC, (slr0162-0163 )), EPS (sll1581, sll0923 ), and S-layer (sll1951 ) to engineer 174 mutant strains. Three independent isolates of each fully segregated clone were 175 confirmed by PCR and sequencing. Strains and plasmids used in this study are listed in 176 Table 1; primers are listed in Supplementary Material Table S1. We assessed the biofilm 177 phenotypes of mutant strains using the modified crystal violet assay described above.

178
For each condition, we measured four biological replicates (unique cultures). At least 179 four biofilm coupons served as technical replicates for each biological replicate. Figure 3 180 shows that crystal violet staining from the pilC mutants (SD519) (OD 600 of 0.23±0.02 , 181 p-value = 0.02) and S-layer mutants (SD523) (OD 600 of 0.11±0.03 , p-value = 0.01) 182 were significantly lower than WT (SD100) (OD 600 of 0.78±0.23 ). We conclude that 183 Type IV pili and S-layer are essential for biofilm formation by Synechocystis. Our Wza 184 deletion mutants (SD517) appeared to have a growth defect (OD 730 of 0.39±0.05 ) 185 compared to WT (OD 730 of 0.64±0.11 , p-value = 0.01). Growth and/or energy is 186 required for biofilm formation in certain other bacteria [58,67]. We could not determine 187 whether growth or Wza-dependent EPS is required for biofilm formation by 188 Synechocystis from these data.  Our crystal violet data show that pili and S-layer are necessary but not sufficient for 191 biofilm formation: presence of these surface structures did not cause WT cells to form 192 biofilms unless some additional unknown factor was induced, such as by changes in 193 nutrient concentrations. We wanted to improve our understanding of the environmental 194 signals and molecular mechanisms of cell-cell binding in Synechocystis. Aggregation is 195 related to biofilm formation in that it also involves cell-cell binding, and results in 196 multi-cellular structures. Aggregates and biofilms are both relevant to many of the same 197 ecological processes and biotechnology applications [70,71]. Additionally, the amount of 198 cellular material in the Synechocystis biofilms we grew on coverslips was insufficient for 199 convenient biochemical analyses, likely due to the small culture volume (3 mL per 200 biofilm coupon). Therefore, we developed a rapid aggregation assay to further 201 characterize cell-cell binding by Synechocystis. Synechococcus sp. WH8102, producing aggregates when either phosphorus or nitrogen 208 concentrations were lowered [72,73]. Similarly, lowering iron or phosphorus nutrient 209 concentration induced aggregation and synthesis of extracellular material by the marine 210 cyanobacterium Trichodesmium IMS101 [74].  We also investigated the role of cellular energy in aggregation. Compared to positive 212 controls, cultures induced for aggregation by shift to 0.8x BG11 remained suspended 213 when incubated in the dark (5.6±6.0 % aggregation, p-value < 0.01), or when 5 µmol 214 DCMU (3-(3,4-dichlorophenyl)-1,1-dimethylurea)was added to 0.8x BG11 cultures 215 incubated in the light (0.9±1.6 %, p-value < 0.01). Dark conditions and DCMU both 216 prevent photoautotrophic growth [75][76][77]. When cultures in 0.8x BG11 were shifted to 217 illuminated conditions after being incubated eight hours in the dark, they eventually 218 aggregated to the same degree as without dark incubation (46.7±9.8 %). We conclude 219 that the aggregation phenotype requires cellular energy production; i.e. it is not caused 220 solely by change in chemical or physical environment, such as with addition of cationic 221 coagulants for algae dewatering [78,79], or change in ionic strength of media directly 222 affecting the hydrophobicity and adhesiveness of cells, as described in XDLVO theory 223 [80]. 224

225
During our aggregation assay, conditioned supernatant was exchanged for fresh 226 BG11. In this step, the extracellular environment was modified by removal of soluble 227 microbial products (SMP) [81]. SMP of different bacteria include the proposed secreted 228 inhibitors and enhancers of biofilm formation by S. elongatus) [63], and RPS (released 229 polysaccharides) of E. coli, which could influence aggregation. The pH, salinity, and 230 osmolarity of the culture were also altered, and could be signals for aggregation, such as 231 by inducing a stress response [68]. 232 We therefore tested aggregation by altering nutrient concentration without removing 233 SMP, which was accomplished by spiking 100 mL cultures resuspended in supernatant 234 with microliter volumes of concentrated BG11 stock solutions, to approximately 1.2x 235 BG11 final concentration (assuming supernatants of mid-log cultures are approximately 236 1.0x BG11). As a control, we also tested aggregation when nutrient concentration is 237 increased by shift to fresh BG11 at 1.2x concentration. As shown in Figure 4, we found 238 no significant difference in aggregation in 1.2x BG11 in supernatant, compared to that 239 in fresh 1.2x BG11. Furthermore, we show that an increase in nutrient concentration is 240 sufficient to induce aggregation to the same degree as a decrease in nutrient 241 concentration, regardless of presence of SMP (1-factor ANOVA). We conclude that 242 removal of SMP had no effect on cell-cell binding under the conditions tested. 243 We show in Figure 5 and Supplemental Material Video 1 the phenotype of a simple 244 and rapid flocculation assay, where cell-cell binding results in larger, buoyant flocs 245 rather than smaller sinking aggregates. We note that the IUPAC (International Union 246 of Pure and Applied Chemistry) does not distinguish between flocculation and 247 aggregation, both of which refer to the formation of multi-particle clusters due to 248 destabilization of a colloid suspension [82]. We use these different terms here as a 249 convenient way to distinguish between two different assays. It is likely that the data 250 collected from these two assays represent variations in degree of a single phenotype, 251 rather than two different phenotypes.  The flocculation assay affects a higher percent of total biomass and is a more rapid 253 phenotype than the aggregation assay, with a duration of approximately two hours from 254 induction to completed flocculation, compared to the aggregation phenotype which can 255 take up to eight hours (time-course data not shown; see Supplementary Video 1 for 256 flocculation time-lapse video). This assay may be invaluable to characterization of 257 buoyant flocs, which are the cause of cyanobacterial surface blooms, or water blooms 258 (blooms). Blooms and buoyant flocs have been specifically and extensively studied due 259 to their role in harmful algal blooms (HABs), which cause malodorous and/or toxic 260 effects with negative impact on water recreation and ecology (reviewed in (95)(96)(97)).

261
Some phototrophs use gas vesicles to regulate their buoyancy; Synechocystis does not 262 encode genes known for gas vacuole formation [83]. One study described buoyant floc 263 formation by Synechocystis when 5 mmol CaCl 2 (20x levels of 1xBG11) was added, but 264 not BG11 up to 50x. [84] Our Synechocystis flocculation assay uses a nutrient strength 265 of 1.2x BG11, resuspended in autoclaved room-temperature supernatants. biofilm formation compared to WT (Fig. 3). As the S-layer mutant was deficient in 272 biofilm formation but not aggregation, it is possible that S-layer is important for initial 273 attachment, and has greater influence cell-glass binding compared to cell-cell binding.

274
Previous studies have shown that increasing substrate surface roughness decreases the 275 surface interaction energy, facilitating binding [85,86]. Similarly, it may be that 276 Synechocystis binding to smooth glass coverslips during the modified crystal violet assay 277 is less favorable compared to binding to a rough cell surface during the aggregation 278 assay. Additionally, there is precedence for influence of the pH and ionic strength of the 279 growth medium on the binding phenotypes of S-layer mutants, including studies of 280 Lactobacillus, Clostridium, and Geobacillus, three Gram-positive genera [80,87]  facilitates binding by reducing the effective radius at the point of surface contact with 292 pili tips, as predicted by DLVO theory; this is also consistent with pili playing a larger 293 role than S-layer in aggregation ( Figure 5). Since we showed that SMP do not influence 294 aggregation (Figure 4) suspension (sedimented without aggregation) more readily than WT or Wzc mutants.

303
Wza mutants were also shown to have less EPS, and a smaller zeta-potential 304 (approximately -21 mV) than WT (approximately -35mv) [38]. Zeta-potential is a 305 proxy for quantifying the electrostatic charge of cell surfaces [88]. The authors of the 306 study concluded that Wza-dependent EPS production promotes dispersal of planktonic 307 WT cells via electrostatic repulsion, consistent with the super-binding phenotype of the 308 wza deletion mutants reported here. This conclusion is also consistent with a third 309 study reporting constitutive aggregation and binding to glass of Synechocystis mutants 310 lacking EPS [28]. The mutations in wzt/kpsT and wzm/kpsM are predicted to function 311 as ABC-transporters in the same pathway as wza and wzc. concentration of BG11, is required for aggregation to occur. We hypothesized that WT 317 Synechocystis was producing adhesive molecules on its surface in response to changes in 318 growth medium, causing aggregation. To detect these putative adhesive molecules, we 319 isolated outer membrane protein fractions of WT and mutant strains under treatment 320 and control aggregation conditions. Samples were split before loading on gels to include 321 both boiled and unboiled preparations of proteins, in order to detect any heat labile aggregation in this species [40]. We used the published cellulase assay to test for role of 331 cellulose in Synechocystis aggregation. Compared to a negative control, we saw no 332 significant change in degree of aggregation in cellulase-treated cultures (data not 333 shown). As shown in Supplementary Material Figure S2, we also tested for presence of 334 cellulose in the purified extracellular matrices of aggregated and control cultures using 335 the glucose oxidase assay, which quantifies amount of glucose released by digestion with 336 cellulase enzyme, as shown previously with T. vulcanus RKN samples [40]. While our 337 results do indicate that cellulase-liberated glucose was present in our samples compared 338 to negative and positive controls (p-values < 0.03), these glucose equivalent levels 339 showed no significant variation between aggregated and unaggregated samples. A 340 mutational analysis of the biofilm and aggregation phenotypes of Synechocystis mutant 341 lacking the gene sll1377, which contains the putative cellulose synthase domain, would 342 be more definitive.

343
Based on this and previous studies, our current understanding is that Synechocystis 344 aggregation is bioenergy-dependent, is induced directly or indirectly by changes in 345 nutrient concentrations but not SMP removal, in a process that requires Type IV pili 346 but not cellulose, and is inhibited by Wza-dependent cell-bound EPS. Additionally, we 347 show in Supplementary Figure S3 that mutants lacking S-layer and Wza phenocopied 348 WT for gliding motility and phototaxis. This indicates that the influence of these 349 mutations on biofilm formation and aggregation is not due to epistatic effects on motility 350 pathways that may be indirectly required for cell binding. Additionally, our WT Kazusa 351 strain (SD100, see Table 1) with amotile Type IV pili is competent for aggregation and 352 October 22, 2018 14/33 biofilm formation, indicating motility is not required for cell binding in Synechocystis.

353
Although the precise molecular mechanisms of Synechocystis cell binding remain to 354 be determined, our data are consistent with the relative importance of these factors as 355 reported in other bacteria: cellular energy/growth was essential for any aggregation to 356 occur, whereas the smaller effects of S-layer and Wza-dependent EPS on aggregation 357 and biofilm formation suggest their role is due to electrostatic and/or hydrophobic 358 contributions to the cell surface. Likewise, our data indicate that pili have a larger 359 impact than S-layer or Wza-dependent EPS, which could be attributed to their 360 additional roles in increasing effective culture density, and / or reducing the effective 361 radius of the point of surface contact, which is predicted by DVLO theory to increase 362 binding. formation, in addition to aggregation and flocculation assays. These new assays enable 369 much more rapid analysis (<72 hours vs weeks) of WT Synechocystis cell binding 370 phenotypes compared to those published previously. This is due to using changes in 371 nutrient concentration to induce binding of exponentially growing cultures, rather than 372 growing cultures in blue light [50] or light-activated heterotrophic growth (LAHG) [49], 373 which necessitates using slow-growing cultures. 374 We demonstrate the utility of these assays in performing mutational analysis to 375 identify cell surface structures influencing cell-cell binding, namely Type IV pili, 376 Wza-dependent exopolysaccharide, and S-layer. These findings include the report of a 377 non-biofouling strain of Synechocystis, the pilC deletion mutant SD519, which would be 378 an advantageous genotype for feedstock strains cultivated in planktonic PBRs or open 379 ponds. Additionally, we used these assays to determine that change in nutrient limiting self-shading that would otherwise be caused by sedimentation [38] or transient 391 contact in cell suspensions [89]. Under altered nutrient conditions, blue light, or LAHG, 392 this cell-cell repulsion is overcome through an unknown mechanism that, based on 393 studies in other bacteria, likely includes synthesis and export of adhesive molecules to 394 the cell surface.

395
In oligotrophic natural environments such as lakes and pelagic zones of open oceans, 396 cyanobacterial cell density is much lower than that typically used for lab cultures [90], 397 reducing the number and size of aggregates detected in these natural environments [72]. 398 However, migration of cyanobacteria through the water column is a normal part of their 399 seasonal adaptation, forming blooms on lake surfaces in spring, and benthic layers in 400 the winter [90][91][92]. Aggregation under nutrient-limited conditions contribute benefits 401 to phototrophs (reviewed in [93]), including relocation of cells to nearby micro-niches 402 that may not be as nutrient-limited. Cyanobacterial aggregates (particulate organic 403 carbon) also have important ecological implications, as they have recently been 404 identified as important contributors to carbon flux to lower ocean depths, which has a 405 major impact on oceanic food webs [94,95]. Global warming may disrupt these natural 406 cycles in a number of ways, including increased temperatures and ocean acidification; 407 overall, climate change is predicted to increase the growth of cyanobacterial blooms, 408 including those species known to be toxic [96,97]. Additional studies will be helpful in 409 developing strategies to mitigate these negative effects on ocean food webs, and in 410 optimizing cyanobacteria for production of sustainable food, fuel, and other valuable 411 commodities [98,99].

427
A standard crystal violet assay was adapted to Synechocystis from methods described 428 previously ('microtiter dish biofilm formation assay', [66], 'plastic binding assay' [67]. 429 Cultures were diluted in 100 mL of BG11 at a starting OD 730 of approximately 0.15 and 430 grown again to log phase as follows: for negative control cultures (uninduced), growth 431 conditions were as described above; for treatment cultures (induced), the side-arm flask 432 for humidification was removed, resulting in evaporation of culture flask to about 84 mL 433 volume over 24 hours, equivalent to a nutrient strength of approximately 1.20x BG11.

434
Therefore, returning this culture to 1.0x BG11 at the start of a biofilm assay (described 435 below) introduces a shift from higher to lower nutrient condition.  Coverslips were removed and rinsed 10 seconds per side with strong stream of BG11 451 from a squeeze bottle, and excess solution was wicked off by standing the coverslip 452 edgewise on absorbent paper for 5 seconds. Coverslips were then stained by inserting in 453 wells containing 4 mL of 0.01% aqueous crystal violet solution (w\vol) for five minutes 454 in a separate tissue culture plate. Unbound stain was rinsed and wicked away, as above. 455 Coverslips were dried in ambient air overnight in the dark and used for qualitative 456 assessment (imaging of macroscopic staining patterns). The final culture OD 730 of each 457 well was also measured to correlate planktonic culture growth with biofilm growth. For 458 quantifying biofilms, dried coverslips were placed in small weigh boats and crystal violet 459 stain was dissolved in 1 mL of DMSO with platform shaking for 20 minutes in the dark 460 or until coverslip stain was removed (up to 45 minutes). Crystal violet absorbance was 461 measured at 600 nm as a proxy for the amount of cellular material bound to coverslips. 462

463
Biofilms were grown on glass coverslips and rinsed (without staining) as described in 464 the modified crystal violet assay (above) and placed on ice to cool. 100 µL of 1.6%

473
DNA manipulation was carried out using standard procedures [101]. Suicide 474 plasmids for replacing Synechocystis genes with a KmRsacB cassette were constructed 475 by four-part ligation into commercial vector pGEM 3Z (Promega), a pUC18 derivative. 476 The KmRsacB cassette from pPSBA2ks [69] contains markers for kanamycin resistance 477 and sucrose sensitivity. PCR fragments of approximately 500 base pairs (bp) located  Mutants of Synechocystis were generated as previously described [102]. Briefly, cells 504 from log-phase culture were harvested as above and resuspended to 200 µL volume 505 equivalent of OD 730 = 2.5. Four µg of suicide vector was added to Synechocystis cells 506 and incubated for six hours in BG11 without antibiotic, with intermittent shaking. The 507 transformation reaction was plated onto a Nuclepore Track-etch membrane (Whatman) 508 on a BG11 agar plate. These were then incubated for 24-72 hrs at 30 • C with

517
Since natural competence in Synechocystis requires Type IV pili [59], we prepared 518 electro-competent cultures of our apiliate pilC mutants in order to transform them with 519 plasmid pΨ552 for expressing PilC, as described previously [103,104]. We harvested 520 cells from 50 mL of log-phase cultures as described above. Cells were resuspended in  Cells were lysed and fractionated using standard methods [105]. 15 mL volumes of 531 cultures were harvested by centrifugation at 6,000xg for 5 minutes; cells were stored at 532 −80 • C. Cells were resuspended in 1.2 mL of 50 mM ammonium bicarbonate buffer 533 solution with HALT protease inhibitor (ThermoFisher) on ice. 600 µL of sample were 534 added to 2 mL cryovials with 400 µL of 0.1 mm zirconium beads. Cells were lysed by 535 bead-beating (Mini BeadBeater, BioSpec) for 7 cycles at maximum speed (one cycle is 536 30 seconds beating followed by 2 minute incubation on ice). Whole-cell lysates were 537 fractionated using differential centrifugation as follows. Lysates were transferred to new 538 tubes. Unlysed cells were harvested at 1,600xg for 5 minutes. Supernatants were 539 transferred and total membrane fraction was harvested at 16,000xg for 1 hour. Total

Aggregation assay 549
Cells from 100 mL of mid-log culture (OD 730 of 0.6-0.8) were harvested by 550 centrifugation as described above, resuspended in either supernatant (negative control) 551 or 0.8x BG11 (treatment condition), and decanted into 100 mL glass graduated 552 cylinders. The starting OD 730 was measured, and the standing cultures were incubated 553 at 30 • C with illumination as described above. The final OD 730 after eight hours was 554 measured by sampling 1 mL of culture from the 50 mL mark on the cylinder.

555
Aggregation was reported as normalized percentage change in OD as follows: [(Final OD 556 -Starting OD) / Starting OD] x100. Negative % aggregation indicated that the culture 557 density increased over time due to cell growth, while minimal aggregation occurred. Synechocystis cultures were grown to midlog phase (OD 730 between 0.6 to 0.8), and 560 cells were harvested as described above. Supernatants were decanted into large flasks (5 561 liters or more); capped with foil and placed in secondary containment pans.

562
Supernatants were autoclaved for five minutes on gravity cycle with no drying (total 563 time of autoclave cycle should be no more than about 15 minutes.) Supernatants were 564 cooled to 30 • C with ice bath using a digital thermometer to monitor temperature.

565
BG11 stock solutions were added to cooled supernatants to increase medium strength to 566 a final approximate concentration of 1.2x BG11, assuming supernatant contributes 567 approximately 1.0x BG11. Harvested cells were resuspended in the prepared 568 supernatants. Cultures were then decanted into graduated cylinders, and incubated and 569 measured as described above. Cellulase digestion of aggregates was performed as described previously [40]. Three 572 mL of aggregated cultures were transferred to 15 mm glass test tubes, and 100x stock 573 solutions of cellulase were prepared in dH 2 O and added to final concentrations of 574 0.60 U/mL cellulase (Sigma-Aldrich, product number C0615). An equal volume of water 575 was added to a negative control culture. Cultures were mixed and incubated without 576 shaking at 30 • C in the light. After incubation, cultures were gently resuspended, and 1 577 mL samples were transferred to semi-micro cuvettes and allowed to settle for one hour. 578 Dispersal of aggregates was determined by increase in absorbance at OD 730 between 579 treated and control condition. Extracellular matrix (ECM) of cells, including S-layer and exopolysaccharide, were 583 removed and purified by a combination of mechanical and chemical separation from 584 intact cells as described previously [28,29,106]. 100 mL cultures of OD 730 585 approximately 0.6-0.8 were centrifuged for 20 minutes at 6,000xg. Cells are resuspended 586 in 10 mL of supernatant with 60 µL of formaldehyde and incubated at 4 • C for one hour. 587 4 mL of 1 N NaOH were added, mixed gently and incubated again at 4 • C for three 588 hours to disrupt ionic bonding between the ECM and the cell. Cells were centrifuged for 589 20 minutes at 20,000xg to physically separate ECM from cells. Supernatants were  Cellulose from ECM was digested and quantified as described previously [40], with 600 modifications. ECM samples and cellulose positive control (Sigma-Aldrich product 601 number C8002) were digested with cellulase enzyme (Sigma-Aldrich, product number 602 C0615) to liberate glucose, which was detected by a glucose oxidase assay. Reactions  The phototaxis assay was adapted from Bhaya et al. [61]. Log-phase cultures were 622 diluted to OD 730 of about 0.25, and 10 µL volumes were spotted on swarm agar (BG11 623 medium prepared with 0.5% Difco BactoAgar) in grid-lined square petri dishes, such 624 that inocula were oriented directly over grid lines. Plates were sealed in parafilm and 625 incubated at 30 • C and 30% ambient humidity under directional illumination (a dark 626 box with a 30 µmol/(m 2 s) photons PAR light source at one end). Strains were graded 627 as motile and phototactic using qualitative assessment of growth having blurred edges, 628 and having elongated away from grid line, as compared to negative control strain, which 629 grows in a disc with crisp edges centered on top of grid line (See Supplemental Material 630 Figure S3).