Unprecedented biomass and fatty acid production by the newly discovered cyanobacterium Synechococcus sp. PCC 11901

Cyanobacteria, which use solar energy to convert carbon dioxide into biomass, are potential solar biorefineries for the sustainable production of chemicals and biofuels. However, yields obtained with current strains are still uncompetitive compared to existing heterotrophic production systems. Here we report the discovery and characterization of a new cyanobacterial strain, Synechococcus sp. PCC 11901, with potential for green biotechnology. It is transformable, has a short doubling time of ≈2 hours, grows at high light intensities and a wide range of salinities and accumulates up to 18.3 g dry cell weight/L of biomass – 2-3 times more than previously described for cyanobacteria - when grown in a modified medium containing elevated nitrate, phosphate and iron. As a proof of principle, PCC 11901 engineered to produce free fatty acids did so at unprecedented levels for cyanobacteria, with final yields reaching over 6 mM (1.5 g/L), comparable to those achieved by heterotrophic organisms.


Introduction. 49
The current production of commodity chemicals from fossil fuels results in the release of greenhouse 50 gases, such as CO2, into the atmosphere. Given concerns over the possible link between greenhouse 51 gases and climate change and the limited abundance of cheap fossil fuels 1 , alternative sustainable 52 approaches need to be developed to produce carbon-based chemicals on an industrial scale. Yeast and 53 bacteria are widely used biotechnology production platforms. However, their growth relies on the 54 addition of carbohydrates to the growth medium leading to a 'food vs. fuel' dilemma, which will likely 55 drive the price of most carbon feedstocks up, fueled by the negative impact of global warming on food 56 crops 2 . Cyanobacteria have the potential to provide a completely sustainable solution 3,4 . These 57 evolutionary ancestors of algal and plant chloroplasts are gram-negative prokaryotic 58 oxyphotoautotrophs, able to convert carbon dioxide and inorganic sources of nitrogen, phosphorous 59 and microelements into biomass 5 . 60 Among cyanobacterial strains, the marine strain Synechococcus sp. PCC 7002 6 and several freshwater 61 strains, namely Synechocystis sp. PCC 6803 7 , Synechococcus elongatus PCC 7942 8 and more recently 62 Synechococcus elongatus UTEX 2973 9 , have become model organisms for both basic photosynthesis 63 research as well as the photoautotrophic production of different chemicals such as bioplastics 10 , biofuels 64 (ethanol 11 and free fatty acids 12,13 ) and specialized compounds like terpenoids 14,15 . However, the average 65 yields are often low compared to heterotrophic microbes, partly due to relatively slow growth and low 66 biomass accumulation. 67 In this work, we report the isolation and detailed characterization of the novel cyanobacterial strain 68 analyses revealed that the axenic cyanobacterial strain was a member of the Synechococcales group 96 (now deposited in the Pasteur Culture Collection as Synechococcus sp. PCC 11901) and that the closest 97 phylogenetic relatives to the companion heterotrophic bacterial strain belong to the marine 98 Thalassococcus genus 23 . Both the axenic and xenic strains of PCC 11901 grew equally well in the 99 presence of cobalamin, reaching an OD730 ≈23 after 72 hours (Fig. 1a). In contrast, growth of the axenic 100 strain was severely inhibited in the absence of added cobalamin with the observed residual growth 101 probably reflecting the retention of intracellular cobalamin or methionine reserves 24 in the inoculum 102 (Fig. 1a). These results imply that PCC 11901 is auxotrophic, and that its growth depends on the 103 availability of cobalamin in the growth medium. 104 Although we did not sequence the genome of the isolated Thalassococcus strain, another 105 Thalassococcus sp. strain (SH-1) possesses cobalamin biosynthesis genes (GenBank accession number 106 CP027665.1). Given the previous evidence for the mutualistic interaction between cobalamin-107 dependent microalgae and bacteria 22 , it is very likely that the Thalassococcus contaminant in the xenic 108 culture could serve as a natural symbiotic partner for the PCC 11901 strain providing it with cobalamin, 109 while consuming nutrients excreted by the cyanobacterium. 110 As in the case of PCC 7002 25 , the PCC 11901 strain can grow mixotrophically in the presence of glycerol 111 and photoheterotrophically in the presence of glycerol and the herbicide DCMU, which inhibits 112 photosynthetic electron flow, although growth is impaired (Fig. S3c). Interestingly, this strain could not 113 utilize glucose for growth, as very little or no growth was observed after one week of incubation with 114 glucose and DCMU (Fig. S3d). 115 Analysis of cell morphology revealed that PCC 11901, though unicellular, could form short filaments of 116 2-6 cells, depending on growth phase (Fig. S11). Individual cells are elongated with sizes ranging from 117 1.5 to 3.5 µm in length and 1-1.5 µm in width. On average (n=45), cells contain 4 to 6 concentric layers 118 of thylakoids around the cytoplasm with visible convergence zones on the cell periphery (Fig. 2a-c). In negatively stained cells, long (1-1.5 µm) fibres of pili-like structures, similar to those seen previously in 120 other cyanobacteria 26,27 , were observed extending from the outer cell wall membrane (Fig. 2d-e).  were compared to the PCC 11901 strain genome using BLAST analysis in the CGView Server 28 tool. 150 Rather unusually, there are several major insertions in the genome of the PCC 11901, depicted by white 151 gaps in the BLAST search scores shown in Fig. 1d, the biggest of approximately 24 kbp, which are not found in other cyanobacterial genomes. Analysis of these fragments showed similarity to genes from 153 other cyanobacteria outside the Synechococcales taxonomy order group, with a strong prevalence of 154 genes encoding predicted glycosyltransferase genes (14 genes), ABC transporter components, 155 transposases, toxin-antitoxin system components and an alcohol dehydrogenase (Table S5). 156 In one of the endogenous plasmids (GenBank accession number: CP040361.1) we identified a gene 157 encoding a 5-methyltetrahydropteroyltriglutamate-homocysteine S-methyltransferase homologue 158 (MetE, GenPept: QCS51047.1) involved in cobalamin-independent methionine synthesis. Although most 159 cyanobacteria are capable of synthesizing cobalamin or pseudocobalamin required by the cobalamin-160 dependent methionine synthase (MetH), some strains, such as PCC 7002, lack MetE, and are therefore 161 strict cobalamin auxotrophs 24,29 . Surprisingly, despite having both variants of methionine synthase in the 162 genome, PCC 11901 is dependent on added cobalamin for growth (Fig. 1a), suggesting potential 163 mutations in either the metE gene itself or in a corresponding cobalamin riboswitch leading to either a 164 loss of activity or expression of this gene. Interestingly, two genes encoding enzymes in the cobalamin 165 biosynthetic pathway cobA (encoding uroporphyrinogen-III C-methyltransferase, GenPept: QCS48576.1) 166 and cobQ (encoding cobyric acid synthase, GenPept: QCS50783.1) can be found in the chromosome, 167 suggesting that they could have been either acquired from other strains by horizontal gene transfer or 168 the cobalamin biosynthesis pathway could have been lost during evolution. 169 To test the tolerance of PCC 11901 to high salt concentrations, growth was assessed in the presence of a 176 wide range of NaCl concentrations, from 0 to 10% (w/v), where sea water is equivalent to ≈3.5 % NaCl 177 (Fig. 1b). PCC 11901 growth rates did not differ significantly in the 0-2.5% NaCl concentration range, 178 with a maximum of 0.28 h -1 at 1.8% NaCl. Higher NaCl concentrations had a noticeably negative impact 179 on growth rates, although cultures grown in the presence of 7.5% NaCl were still able to reach an OD730 180 of 17.4 after 3 days of cultivation and even in the presence of 10% NaCl were able to grow to a final 181 OD730 = 6.3 with an average doubling time of 12 hours, a similar growth rate to that of the halotolerant 182 microalga Dunaliella 30 . Different combinations of temperature, light intensity and CO2 conditions were tested in order to 184 investigate their effect on growth (Fig. 1c). As a reference, PCC 11901 was grown alongside two other 185 fast-growing and high-temperature tolerant strains -UTEX 2973 and PCC 7002 (Fig. S2). All strains 186 grown in atmospheric CO2 conditions exhibited similar doubling times with the shortest of 3.35 ± 0.12 187 hours at low light (100 μmol photons·m −2 ·s −1 ) observed for PCC 11901 (Table S1) tested strains and was unable to exceed OD730 ≈ 10 after 4 days (Fig. S4), an effect that was also 198 previously observed 31 . Regardless of light conditions, PCC 11901 accumulated more biomass after 4 days 199 of growth (4.9 g/L) than PCC 7002 and UTEX 2973 (3.7 and 2.5 g/L respectively). However, all strains 200 accumulated less biomass under higher light, especially PCC 7002 (Fig. S4). 201 202

Transformability and availability of molecular tools for protein expression 203
For a new strain to be useful for synthetic biology and metabolic engineering applications, it must be 204 amenable to transformation and have molecular tools for controlled protein expression. Considering 205 that PCC 11901 is a close relative of PCC 7002, we tested several molecular tools already established for 206 promoter (a truncated cpcB promoter from the PCC 6803 strain functional in PCC 7002) 32 and inserted 208 between flanking regions of the acsA (acetyl-CoA synthetase) gene. As previously reported, targeting of 209 genes to the acsA locus in PCC 7002 allows for the generation of markerless mutants using acrylic acid 210 for counterselection 33 . In the second construct, the yfp gene and a spectinomycin-resistance cassette 211 were cloned between flanking regions of the psbA2 (encoding the D1 subunit of photosystem II) gene 212 and YFP expression was controlled by the synthetic IPTG-inducible Pclac143 promoter 32 . PCC 11901 was 213 then transformed by natural transformation (with transformation efficiency similar to PCC 7002) and 214 positive clones were confirmed by colony PCR (Fig. S10a-b). Fluorescence microscopy confirmed 215 successful YFP expression in both constructs (Fig. 3b), with expression in the ΔpsbA2::Pclac143-YFP strain 216 detectable only after induction with IPTG. Though the targeted genome loci are different for these two 217 constructs, the Pcpt promoter appeared to be stronger than the induced Pclac143 promoter, as the relative 218 fluorescence unit (RFU) per OD730 ratio was higher (Fig. 3c). YFP under control of the Pclac143 promoter 219 was expressed upon induction at 45-fold greater levels than the non-induced control (Fig. 3c). 220

Identification of nutrient constraints 236
Modifying the composition of the growth medium is a common approach to improve biomass and 237 secondary metabolite production using heterotrophic microorganisms 34 . In contrast, the media routinely 238 used to cultivate cyanobacteria were formulated over 40 years ago and are not optimised for high 239 biomass production 29 . Indeed modification of the basic AD7 medium by Clark et al. 31 was shown to 240 increase biomass production of PCC 7002 by approximately 2-fold with further improvements 241 theoretically possible. Other work has also described an enriched BG-11 medium for high-biomass and 242 cyanophycin production by the freshwater strain PCC 6803 35 . 243 Inspired by these reports we adopted a systematic approach to improve growth of PCC 11901 by 244 independently changing the levels of nitrate, phosphate and iron in AD7 medium (Fig. 4) (Fig. 5a-b). Among the freshwater cyanobacteria, PCC 6803 accumulated the highest biomass 275 (6.9 gDW/L), followed by PCC 7942 and UTEX 2973 -both around 6.5 gDW/L. In regard to culture 276 fitness, loss of the light-harvesting phycobilisome complex, which is symptomatic of nitrogen stress, was 277 already apparent in the case of UTEX 2973 after 3-4 days of cultivation but less apparent in the PCC 7942 278 and 11901 strains, even after 10 days of cultivation (Fig. 5c-d). 279 Despite numerous efforts, we were unable to grow any of the freshwater cyanobacteria used in this 280 study to biomass levels exceeding 7 gDW/L. Medium optimization appeared to be more challenging, 281 possibly due to a lower tolerance to higher ionic strength in the media by these strains. We tested five 282 different media formulations for UTEX 2973, PCC 7942 and PCC 6803 (Fig. S7, Table S2). Supplementing 283 BG-11 medium with more nitrate, phosphate, ammonium iron (III) citrate and magnesium sulphate was 284 met with limited success, as, in spite of a fast intial growth, cultures declined dramatically after a few 285 days (Fig. S7a-c). Cultures grown in modified MAD medium lacking sodium chloride did not perform well 286 either (Fig. S7d). In our hands, 5xBG was the most successful formulation (see Fig. 5 and S7e-f) and to 287 our surprise supplementing this medium with phosphate and nitrate concentrations to the levels in 288 MAD medium (modified 5xBG, 5xBGM) led to complete bleaching of the PCC 6803 strain after 10 days 289 ( Fig. S8a-b). Since the 5xBGM medium contains more nitrate and phophate than 5xBG, it is very unlikely 290 that the cultures bleached as a result of the nutrient starvation. Lippi et al. 35 have previously shown that supplementation of BG-11 medium with 65 mM nitrate and 10 mM phosphate allows PCC 6803 to grow 292 to high cell densities (OD=40) though, again, not surpassing the density here observed with 5xBG. 293 Therefore, it is possible that these freshwater strains may have some limitations related to low 294 tolerance to high concentration of inorganic salts or a regulatory mechanism (e.g. quorum sensing) 295 preventing them from growing further. 296 were grown in the MAD medium at initially 150 µmol photons·m -2 ·s -1 , which was then increased to 750 301 µmol photons·m -2 ·s -1 after 1 day. Synechococcus elongatus UTEX 2973 was grown in 5xBG medium under 302 the same conditions. Synechococcus elongatus PCC 7942 and Synechocystis sp. PCC 6803 were also 303 grown in 5xBG medium, with light intensity gradually increased from 75 to 150 µmol photons·m -2 ·s -1 304 (day 1) and eventually set to 750 µmol photons·m -2 ·s -1 (day 2). Having demonstrated that the PCC 11901 strain is amenable to genetic manipulation, a proof of concept 312 demonstration of the strain's biotechnological potential was designed. For that purpose, we chose to 313 modify PCC 11901 for photoautotrophic production of FFA, which can potentially be used as renewable 314 industrial feedstocks 16 . We therefore inserted a truncated version of the E. coli thioesterase ('tesA) 315 gene 12 , codon-optimized for PCC 7002, under control of the inducible Pclac143 promoter 32 , in the genomes 316 of both PCC 11901 and PCC 7002, by simultaneously knocking-out of the long-chain-fatty-acid-CoA ligase 317 (fadD), to generate 11901 ΔfadD::tesA and 7002 ΔfadD::tesA strains (Fig. 6a). Knockout strains of the 318 fadD gene alone (ΔfadD) and WT were used as production controls. In order to check if the MAD 319 medium would also allow an improvement of the production yields, all strains were grown side-by-side 320 using either regular AD7 or MAD medium. 321 Both 11901 and 7002 ΔfadD::tesA strains grown in AD7 medium produced similar amounts of FFA (0.41 322 and 0.34 mM respectively) 3 days after induction (Fig. 6b), while the non-induced strains produced less 323 than 0.07 mM of FFA. The difference in productivity became much more apparent when the strains were grown in the MAD medium. The engineered 11901 ΔfadD::tesA strain was able to grow faster than 325 the 7002 counterpart and 4 days after IPTG induction had already excreted 3.97 mM of FFA, a yield 326 almost twice that achieved by the 7002 ΔfadD::tesA strain in the same time frame (2.01 mM) (Fig. 6c). 327 While the 7002 ΔfadD::tesA strain was able to continue growing for two more days, achieving final FFAs 328 yields of 3.54 mM, the corresponding 11901 strain declined and yields did not increase significantly after 329 the 4 th day, reaching only 4.32 mM. When taking into account cell growth, the overall FFA productivity 330 was higher for the 11901 strain (0.12 mmol/L/OD on day 5) compared to 7002 (0.05 mmol/L/OD and 331 0.08 mmol/L/OD on day 5 and 7 respectively). The control strain 7002 ΔfadD grown in optimized 332 medium produced 0.1 mM of FFA, almost twice as much as the PCC 11901 control strain and production 333 levels in the WT were negligible, within statistical error (Fig. 6d). The large error bars in all FFA 334 measurements can be explained by the low solubility of FFA in water, which makes them float on top of 335 the culture surface and adhere to the surface of culture vessels (Fig. 6f). 336 To test whether production yields could be further improved by using different extraction methods, the 337 11901 ΔfadD::tesA strain was grown in MAD medium and FFA were extracted with hexane either 338 directly from cell cultures or from the cell-free medium supernatant (Fig. 6e). For the first few days, the 339 difference between concentration of FFA in the cell culture and medium extracts was statistically 340 insignificant. However, after 7 days, samples extracted directly from the cell culture contained 341 significantly more FFA (6.16 mM) than the ones from the supernatant. 342 The composition of produced FFA was evaluated by Gas Chromatography analysis in both 11901 and 343 7002 ΔfadD::tesA engineered strains (Fig. S9). Despite having different final productivities, there was no 344 major difference in the lipidomic profile of the two strains. For the 11901 strain, palmitic acid (C16:0) 345 constitutes on average approximately 65% of the total FFA produced, followed by myristic acid (C14:0) -346 23% and stearic acid (C18:0) -9% (Fig. 6g). 347 In this work we report the discovery of a novel cyanobacterial strain, Synechococcus sp. PCC 11901, from 370 the Johor Strait (Singapore) and describe a new medium for its cultivation. PCC 11901 displays fast 371 growth, high biomass accumulation and resistance to various abiotic stresses (such as high light and 372 salinity), which are all important criteria for potential industrial uses of any cyanobacterial strain, 373 especially when cultivated outdoors in a non-sterile environment and exposed to contamination that 374 could potentially overgrow it. It was recently shown that engineering cyanobacteria to utilize 375 unconventional phosphorous and nitrogen sources can dramatically reduce the risk of contamination 36 . 376 However, increasing salinity of the growth medium is a more conventional method to decrease viability 377 of less adaptable competing strains. We show that this euryhaline strain tolerates up to 10% (w/v) of 378 NaCl and can grow at high light intensities greater than 750 µmol photons·m -2 ·s -1 (the limit of our 379 equipment). Although PCC 11901 tolerates high temperatures, up to a maximum of 43 °C, the optimal 380 temperature for its growth may be around 30 °C, the average temperature in its natural environment in Singapore with low fluctuations (28-32 °C) throughout the year. Though its cobalamin auxotrophy may 382 be inconvenient for industrial cultivation, this can easily be overcome by heterologous expression of a 383 cobalamin-independent methionine synthase (MetE) 24 . Interestingly, PCC 11901 strain already possesses 384 a metE gene in one of the endogenous plasmids, though it seems to be inactive. This is possibly an 385 evolutionarily recent mutation, most likely due to the ready availability of cobalamin in seawater 21 . 386 In our hands, PCC 11901 shows a similar doubling time to PCC 7002 of 2.14 ± 0.06 hours, slightly longer 387 than that for UTEX 2973 (1.93 hours) (Table S1). However, compared to both UTEX 2973 and PCC 7002, 388 PCC 11901 can grow to much higher cell densities and accumulate more biomass overall (Fig. 5a-b, S4  389

and S6). 390
Experimental testing of nutrient limitations allowed us to tap into an unprecedented potential for high 391 biomass accumulation. We show that PCC 11901 can grow to OD730 of 101 and produce 18.3 g/L of dry 392 weight biomass, which are the highest reported values for any cyanobacterium. When grown alongside 393 PCC 7002, PCC 7942, PCC 6803 and UTEX 2973 strains, PCC 11901 accumulated 2-3 more biomass than 394 these strains despite using the same nutrient concentrations. It is also possible that PCC 11901 could 395 produce even more biomass if exposed to greater light irradiances than we could provide, especially 396 given that residual phosphate and iron precipitates were still present in the medium and no bleaching or 397 pigment loss was observed even at the highest cell density. Growing marine strains in both MAD and the 398 previously described MMA medium 31 at higher temperatures (37-38 °C), led to a decline in biomass after 399 10 days of cultivation (Fig. S6). In contrast, when grown at 30 °C, strains could sustain growth to higher 400 densities, possibly due to the higher solubility of CO2 at lower temperatures or a hitherto unknown 401 regulatory mechanism. A comparison of the compositions of MAD and MMA media suggests that 402 cyanobacteria do not require such a high excess of nutrients as found in MMA (especially phosphate, 403 which is approximately 26x greater in MMA) to accumulate high biomass.
As a biotechnological chassis strain, PCC 11901 has several attractive traits: it is naturally transformable, 405 which facilitates genetic manipulation; synthetic biology tools already existing for the PCC 7002 406 strain 32,33 are also compatible with this new strain, facilitating an easy switch of metabolic engineering 407 constructs to the PCC 11901 strain and it shows high growth rates and high rates of biomass production. 408 As a proof of concept, we expressed a truncated 'tesA gene in order to produce FFA, a valuable biodiesel 409 feedstock. FFA yields per litre of culture using MAD medium were found to be approximately 10-fold 410 higher in comparison to using the regular AD7 medium under the same growth and induction 411 conditions, even though the final OD730 was only 1.5 times higher. Our results demonstrate that growth 412 medium optimization can dramatically increase the production yields of FFA though further 413 investigations on the mechanism behind the observed changes to productivity will be necessary to 414 further understand this phenomenon. 415 The maximum FFA titre achieved in this study of 6.16 mM (≈1.54 g/L, 5 days after induction) is the 416 highest reported to date for cyanobacteria, with values of 0.64 g/L (when using isopropyl myristate 417 overlay for FFA removal) 13 , 0.19 g/L 37 and 0.13 g/L 12 previously reported for PCC 7942, PCC 6803 and PCC 418 7002 respectively after 16-20 days of cultivation. It is also on a par with the production levels achieved 419 by heterotrophic organisms. Introduction of different heterologous 'tesA genes in a ΔfadD background 420 strain of E. coli resulted in the production of ≈2 g/L FFA 2 days post-induction 38 and impairment of 421 glycogen metabolism in Yarrowia lipolytica led to production of 1.44 g/L of FFAs (up to 2.04 g/L when 422 using a dodecane overlay) after 5 days 39 . Though these values are much less than the highest ever 423 reported FFA production titre of 33.4 g/L, using heavily engineered Saccharomyces cerevisiae strains, 424 this particular system required supplementation with >300 g/L of glucose over 10 days of cultivation 40 425 and it is quite likely that further engineering of our strain will increase titre and cell viability.
Overall this study shows that the newly discovered cyanobacterium Synechococcus sp. PCC 11901, in 427 conjunction with the use of modified growth medium, has the potential to become a relevant industrial 428 biotechnology platform for the sustainable production of carbon-based molecules. 429   Table S1, as required. Cells growth was monitored by measuring the optical density at 730 nm (OD730) 472 in a 1-cm light path with a Cary 300Bio (Varian) spectrophotometer. Doubling times were calculated as 473 mean within the logarithmic range only.
For the cobalamin auxotrophy experiment, a starter culture of the axenic strain was grown in medium 475 AD7 supplemented with cobalamin, but the inoculum was washed 3 times in order to remove any 476 remaining cobalamin from the medium. 477 To facilitate acclimation to high salt concentrations cultures were incubated in low light for 1 week. 478 Once the cultures were adapted to the respective conditions, biological triplicates were inoculated to 479 starting OD730 of 0.1 and grown at 38 °C and 300 µmol photons·m -2 ·s -1 light intensity. For the comparison of all cyanobacterial strains 3x 33 mL of either MAD or 5xBG (for details see Table  486 S2) media were inoculated with cells to starting OD730 of 0.1 and grown with shaking at 180 rpm, 30 °C, 487 with RGB LED ratio 1:1:1. For growth of PCC 11901, PCC 7002 and UTEX 2973, the initial light intensity 488 was set to 150 µmol photons·m -2 ·s -1 increased after 1 day to 750 µmol photons·m -2 ·s -1 . In the case of PCC 489 7942 and PCC 6803 the initial light intensity was set to 75 µmol photons·m -2 ·s -1 , changed to 150 µmol 490 photons·m -2 ·s -1 after 1 day and further increased to 750 µmol photons·m -2 ·s -1 on the next day. To 491 compensate for water evaporation 700 µL of sterile MilliQ water was added to the cultures on daily 492 basis. For dry biomass evaluation, 0.5 to 1 mL of cultures were transferred into pre-weight Eppendorf 493 tubes and centrifuged at 20,000 × g for 2 minutes. Cell pellets were washed three times with 1 mL of 494 MilliQ water in order to dissolve and remove remaining salt precipitates and dried at 65 °C for 24 hours. AD7 medium supplemented with 10 mM glycerol or 0.15% (w/v) glucose was inoculated with a single 514 colony of PCC 11901 and grown for one week at low light intensity 50 µmol photons·m -2 ·s -1 , 38°C, 1% 515 CO2. Growth was observed after a few days and strains were restreaked on AD7 agar plates with either 516 10 mM glycerol or 0.15% (w/v) glucose. For the photoheterotrophy assay glycerol and glucose adapted 517 strains were grown to OD730 ≈5 and 15 µL dilution series were transferred onto AD7 agar plates 518 containing different combinations of 10 µM DCMU, 10 mM glycerol or 0.15% (w/v) glucose. Plates were 519 dried and incubated for 7 days at 50 µmol photons·m -2 ·s -1 , 38°C, 1% CO2. 520 521 4.5. Cloning of constructs.
All fragments needed for cloning of pSW036, pSW039, pSW040, pSW068 and pSZT025 vectors were 523 amplified using Q5® High-Fidelity DNA Polymerase (New England Biolabs) according to manufacturer's 524 protocol. All primers and templates used for PCR amplification are listed in After embedding in 2% (w/v) low-gelling-temperature agarose, samples were cut in 1-2 mm cubic 554 blocks, and post-fixed with 1% (w/v) osmium tetroxide in distilled water for 1 h. Samples were washed 555 twice with distilled water, and dehydrated through a graded ethanol series (1 × 15 min 50%, 1 × 15 min 556 70%, 1 × 15 min 90% and 3 × 20 min 100%). Two 5 min washes with acetone were performed prior to 557 infiltration with araldite for 1 h and with fresh Araldite overnight. Polymerisation was achieved by 558 incubation at 60-65°C for 48 h. Ultrathin sections were cut with a diamond knife at a Reichert Ultracut E 559 microtome and collected on uncoated 300-hexagonal mesh copper grids (Agar Scientific). High contrast 560 was obtained by post-staining with saturated aqueous uranyl acetate and Reynolds lead citrate for 4 min 561

each. 562
Negative staining was performed on 300-mesh copper carbon supports grids (Agar Scientific) that were 563 previously rendered hydrophilic by glow discharge (Easy-Glow, Ted Pella). Glutaraldehyde fixed bacteria 564 were adsorbed to TEM grids by direct application of 5 µl of the suspension for 1 minute and stained by 565 floating the loaded grid onto a drop of 1% uranyl acetate for 20 seconds. The grids were examined in a 566 JEOL JEM-1230 transmission electron microscope at an accelerating potential of 80 kV. 567 568 4.9. Free fatty acid production assay. 569 (production) and ΔfadD (control) strains respectively. PCC 7002 was transformed with pSZT025 and 571 pSW072 to generate 7002 ΔfadD::tesA (production) and ΔfadD (control) strains. Successful 572 transformants were screened for complete segregation by colony PCR (Fig. S10). A large number of the 573 screened colonies for the PCC 11901 strain did not carry the designed insert, suggesting that kanamycin 574 may not be the antibiotic of choice for selection in this strain, as it exhibits partial resistance to 575 kanamycin at low concentrations. 576 All engineered and control strains were grown in 33 mL cultures, in biological triplicates, (except for PCC 577 11901 ΔfadD::tesA grown in biological duplicates) using either basic AD7 or MAD medium. Cultures were 578 inoculated with cells to starting OD730 of 0.1 and grown with shaking at 180 rpm, 30 °C, with RGB LED 579 ratio 1:1:1 at 150 µmol photons·m -2 ·s -1 light intensity. After 1 day, cultures were induced with 1 mM 580 IPTG and the light was increased to 750 µmol photons·m -2 ·s -1 . 1 mL medium aliquots were collected for 581 both OD730 and FFA quantification. For FFA quantification, cell cultures were centrifuged for 2 minutes at 582 20,000 × g and medium supernatants were carefully collected to avoid any disruption of the cell pellets. 583 FFA were quantified in technical duplicates using the EZScreen™ Free Fatty Acid Colorimetric Assay Kit 584 (384-well) (BioVision, USA) according to manufacturers' protocol. 585 For the GC analysis both cell extract and medium supernatant were acidified with 1 M HCl to pH ≈ 2 in 586 order to protonate FFA and facilitate extraction. Samples were extracted with n-hexane, evaporated and 587 dried using a centrifuge vacuum concentrator. Dried samples were resuspended in 100 µL of hexane and 588 aliquots were transferred on the TLC Silica gel 60 plates (Merck, Germany). Plates were resolved for 589 approximately 30 minutes in hexane, diethyl ether, formate solvent mixture at the 70:30:2 ratio. Sample 590 preparation and the GC analysis was performed as previously described 46 .