Cpf1 Is A Versatile Tool for CRISPR Genome Editing Across Diverse Species of Cyanobacteria

Cyanobacteria are the ideal organisms for the production of a wide range of bioproducts as they can convert CO2 directly into the desired end product using solar energy. Unfortunately, the engineering of cyanobacteria to create efficient cell factories has been impaired by the cumbersome genetic tools that are currently available for these organisms; especially when trying to accumulate multiple modifications. We sought to construct an efficient and precise tool for generating numerous markerless modifications in cyanobacteria using CRISPR technology and the alternative nuclease, Cpf1. In this study we demonstrate rapid engineering of markerless knock-ins, knock-outs and point mutations in each of three model cyanobacteria; Synechococcus, Synechocystis and Anabaena. The markerless nature of cpf1 genome editing will allow for complex genome modification that was not possible with previously existing technology while facilitating the development of cyanobacteria as highly modified biofactories.

Prokaryotes are being widely employed as microbial cell factories for the production of value added compounds ranging from biofuels to polymers to therapeutics. Most prokaryotic production systems rely on heterotrophs which require expensive carbohydrate feedstocks. Cyanobacteria are of particular interest among the prokaryotes for their potential role as carbon neutral platforms for the production of such chemicals while eliminating the need for expensive feedstocks. These diverse organisms can fix atmospheric carbon using sunlight and water. Carbon fixation can then be coupled to the direct conversion of CO 2 into a wide range of products. Cyanobacteria have been engineered to produce commodities such ethylene 1 , isoprene 2 , and sugars 3,4 ; biofuels such as alkanes 5 , hydrogen 6 and terpenoids 7 ; bioplastics such as polyhydroxybutyrate 8 ; and bioactive compounds such as pharmaceuticals 9 and vitamins 10 . One major hurdle to engineering these production systems is the lack of precise, modern genetic tools that exist for other extensively studied prokaryotes such as Escherichia coli.
In recent years, CRISPR genome editing technology has revolutionized the field of biotechnology by enabling precise, efficient modification of DNA sequences in a single step, in a wide variety of organisms from mammals 11 to plants 12 to bacteria [13][14][15] . This technology is ideally suited to engineer markerless knock-ins, knock-outs or specific point mutations in numerous species. Unfortunately, CRISPR technology has not been widely used in cyanobacteria due to the apparent toxicity of the Cas9 nuclease in these organisms 15 . We sought to overcome this obstacle by employing Cpf1 from Francisella novicida, a novel RNA directed dsDNA nuclease that we determined to be nontoxic to cyanobacteria.
Cpf1 is a type V-A nuclease of the class II family of CRISPR systems 16 . Cfp1 is not homologous to the commonly used CRISPR nuclease, Cas9 and employs a mechanism that is different from that of Cas9 17 . As such, there are several major differences between cas9 and cpf1 systems. Cpf1 is a dual nuclease that is specific to both the repeats in the pre-crRNA of the CRISPR array transcript as well as the DNA target specified by the mature crRNA and PAM sequence 18 . Cpf1 possesses specific ribonuclease activity that cleaves the 36 bp repeat of the pre-crRNA 4 nucleotides upstream of a hairpin in a sequence, structure, and in a Mg 2+ dependent manner 18 . The mature crRNA then guides Cpf1 to its DNA target where its nuclease activity induces a 5 bp staggered double stranded break 17 nucleotides downstream from the YTN PAM sequence. In cas9 systems the PAM is 3′ to the crRNA while in cpf1 systems the PAM is 5′ to the crRNA 16 . Cas9 typically uses a G rich PAM sequence such as NGG; while Cpf1 from Francisella novicida utilizes a more relaxed YTN PAM sequence 18,19 . The cut site also differs between the two nucleases. Cas9 makes a blunt cut directly adjacent to the PAM while Cpf1 generates a 5 bp staggered cut 17 nucleotides downstream of the PAM 16 .
There are several advantages of using Cpf1 instead of Cas9 for genome editing in bacteria. Cas9 cleaves directly adjacent to the PAM sequence so that an indel resulting from nonhomologous end joining (NHEJ) in species that undergo this repair method results in disruption of the PAM and prevents proper editing via homology directed repair (HDR). In the cpf1 system, the cut is 17 nucleotides distal to the PAM, so that an indel resulting from NHEJ would not disrupt the PAM and the resulting sequence can be recut for a second chance at HDR. Overall, this increases the efficiency of the system. Additionally, in cas9 systems a tracrRNA is required for processing of the pre-crRNA. Therefore, cas9 systems require both a crRNA and tracrRNA to mediate interference. In Cpf1 based systems, the CRISPR array is processed independently of other factors; requiring only Cpf1 and a pre-crRNA to mediate interference, which significantly simplifies the system 16,18 . In synthetic biology studies, the cpf1 system is more cost effective as it uses only a 42 nt RNA component which is significantly cheaper to synthesize than the > 100 nt gRNA required by cas9 systems 16 . Furthermore, in cas9 based systems, a separate tracrRNA-crRNA fusion must be introduced for every target. In contrast, in cpf1 systems a single pre-crRNA array with tandem spacer-repeat sequences can be introduced for multiple targets. The pre-crRNA is subsequently processed by Cpf1 into individual mature crRNAs to target multiple genes, thus facilitating multiplex gene editing. In addition, cpf1 is 20% smaller than cas9, which allows for more efficient editing 16 . Finally, Cas9 requires a NGG PAM sequence which reduces the number of possible targets, especially in AT rich genomes. The YTN (CTN or TTN) PAM sequence that is recognized by Cpf1 is significantly more abundant and allows for a more precise selection of the cleavage target. Importantly, it has recently been demonstrated that the relaxed PAM does not lead to increased off target cutting; at least in an eukaryotic system 20 .

Results
Toxicity of cpf1. We have previously demonstrated that Cas9 poses toxicity in cyanobacteria 15 . We sought to circumvent the toxicity issue by employing an alternative RNA guided DNA nuclease, Fncpf1 from Francisella novicida 16 . We chose to utilize the FnCpf1 gene as this variant has been previously demonstrated to cleave DNA in bacterial systems 16 . We first compared the toxicity from cpf1 to that of cas9. Either promoterless cas9 or cpf1 with a lac promoter were cloned into pVZ321 21 , a replicating vector based on RSF1010. No other editing machinery or homologous repair template was included so that we could assess the toxicity of the two proteins alone. Presumably, Cas9 expression was greatly reduced relative to Cpf1 because it lacked a promoter while cpf1 had a functional promoter. Since both genes are cloned in the same vector in the same orientation, any additional background expression should be the same in both cases and relative expression can be compared. We conjugated both constructs as well as an empty vector into Synechococcus 2973. Only 3 colonies were obtained from the vector containing cas9, while the vector containing cpf1 yielded about half as many colonies as the empty vector (Supplementary Figure 1). Even with lower expression levels, a high degree of toxicity was observed from cas9, while cpf1 showed far less toxicity. This suggested that cpf1 is a suitable nuclease for genome editing in cyanobacteria and therefore, we sought to develop a cpf1 based editing system for these organisms.
Markerless Editing with Cpf1 in Synechococcus UTEX 2973. To facilitate rapid cloning of editing plasmids we constructed a vector, pSL2680 (Supplementary Figure 2) based on the broad host range plasmid RSF1010. RSF1010 replicates well in most gram negative bacteria and would allow us to perform genome editing without integrating cfp1 into the chromosome. This strategy served to simplify the system and allowed us to make truly markerless mutations after the mutants are cured of the editing plasmid. The pSL2680 vector expresses cpf1 from a lac promoter and an endogenous Francisella novicida CRISPR array from a J23119 promoter (Biobrick #BBa_ J23119). The native array had 3 repeat sequences with 27-30 nt spacer sequences separating them. The first spacer in the array is replaced with lacZ flanked by AarI sites that allow lacZ to be swapped for annealed oligos 24 nt in length; while later spacers would remain endogenous to Francisella novicida with no target (Supplementary Fig. 3). Using AarI, the annealed oligos are scarlessly placed between two repeats where they form the new targeting segment of a crRNA in a CRISPR array. Cpf1 will then process the CRISPR array transcript into mature crRNA which is used to target Cpf1 to a specific spot in the genome. Following the CRISPR array are unique KpnI and SalI sites to linearize the plasmids so that homologous repair templates can be inserted using Gibson assembly. We have made this base vector for genome editing available through Addgene (plasmid #85581).
We performed our initial test of genome editing in Synechococcus 2973, a cyanobacterial strain with an ability for rapid growth 22 . We attempted all three modes of gene editing by creating a markerless point mutation, a knock-out mutation or a knock-in mutation. We elected to engineer an S264A point mutation in the psbA1 gene that encodes the D1 protein of photosystem II. This specific mutation is interesting as it has been identified as the mutation that gives rise to resistance to the herbicide DCMU 23 . As such, we would have an easy phenotype to screen for. For the knock-out we created a markerless deletion of nblA. NblA functions to mediate degradation of the phycobilisome antenna complexes which can comprise up to 50% of the total cell protein in cyanobacteria. During conditions of nitrogen starvation, phycobilisomes are degraded to provide supplemental nitrogen for the cell. As such, this protein plays and important role in adaptation to changing environmental conditions. A knockout of the nblA gene is easily identified due to its obvious phenotype. WT Synechococcus 2973 bleaches under nitrogen deprivation due to the degradation of antenna complexes. When the nblA gene is knocked out, the resulting strains remain green upon the removal of nitrate from BG11, the growth medium, because the antenna complexes are not degraded 24 . For the knock-in, we inserted eYFP under the control of the trc promoter into NSI without the aid of antibiotic selection.
Plasmids to generate the three edits were constructed by inserting annealed oligos that target psbA, nblA or NSI into pSL2680, using golden gate assembly. A homologous repair template was synthesized as left and right fragments (or left, right and middle fragments for the eYFP knock-in) with 1 kb of homology to the upstream and downstream sequences. When making a point mutation, the editing plasmid would also be a target for Cpf1 cleavage because the homologous repair template would also contain the target of the crRNA. To prevent cleavage of the editing plasmid, we made a second silent mutation in the homologous repair template to eliminate the PAM site. This mutated PAM site will be incorporated into the chromosome upon editing and prevent further cleavage of the chromosome. This strategy will serve to drive editing to completion and force segregation. The eYFP insertion was designed such that the it would split the sequence targeted by the crRNA to prevent targeting of the editing plasmid as well as cleavage of the chromosome after eYFP has been inserted. The demonstration of markerless knock-ins and knock-outs is of special importance as markerless mutations will allow the introduction or removal of an unlimited number genes while eliminating concerns of growing antibiotic resistant bacteria.
The editing plasmids were conjugated into Synechococcus 2973 and a few colonies resulting from each experiment were collected for further analysis. Sequencing of the point mutation revealed that only 2 of the 8 colonies contained the point mutation; however, when streaked on BG11 + 5 uM DCMU, 6 of the patches showed DCMU resistance (Fig. 1A). Apparently, selective pressure from DCMU allowed us to rescue a small subpopulation of cells that had acquired the point mutation. This suggests that some cells in each patch had received the point mutation, but editing had not gone to completion in all cells. To allow the editing process to go to completion, we repatched the colonies onto BG11 Km 10 three additional times. After the three additional patchings 6 of the 8 colonies had the point mutation as determined by Sanger sequencing. The knock-in of eYFP yielded similar results as initially 2 of the 10 colonies contained the insertion of eYFP as indicated by colony PCR (Fig. 1C). After two additional patches on BG11Km 10 the number of patches containing the insertion of eYFP had risen to 6 of 10 ( Fig. 1D). These data indicate that editing is driven towards completion after a few passages on selective media. While correct colonies could be easily identified from the initial conjugation, repeated passage on selective media significantly increased the proportion of properly edited colonies. The degree of segregation of the mutants was determined by examining the deletion of nblA as only a fully segregated mutant would show the nonbleaching phenotype. After three generations on BG11Km 10 plates, 9 of the 10 nblA deletion mutants showed the nonbleaching phenotype indicating that they were fully segregated (Fig. 1B). Together, these findings suggest that while ~20% of the colonies are edited upon first appearance, maintaining them on selective media forces editing to go to completion and drive complete segregation.
Patches with the insertion of eYFP could not be assayed for eYFP expression because they contained the editing plasmid which also expressed eYFP from the repair region. To clarify the issue, we cured the edited strains of the plasmid. We also cured the editing plasmid from an edited colony from the nblA deletion and the psbA point mutation at this time as well. Toward this goal, an edited patch of each mutant strain was grown in BG11Km 10 to an OD 720 of 1.0. The culture was then diluted 1:2500 into BG11 without antibiotics and grown to an OD 720 of 1.0 to allow spontaneous loss of the editing plasmid. To obtain single cells, the culture was serially diluted and plated on BG11. 50 single colonies of each were picked and patched on BG11Km 10 and also BG11 to identify colonies that had become sensitive to kanamycin and thus lost the editing plasmid. We found that the editing plasmid was lost in ~8% of the cells for the nblA deletion, 14% of the cells for the knock-in of eYFP and 36% of the cells for the point mutation in psbA (Supplementary Table 1). Once we rid the eYFP knock-in strain of the plasmid borne copy of eYFP, we verified that eYFP was properly expressed from its chromosomal location using fluorescence microscopy (Fig. 2).

Markerless Deletions in
Synechocystis 6803 and Anabaena 7120. We examined the versatility of the cpf1 system by applying the technology to editing of two additional genus of cyanobacteria. We chose Synechocystis 6803 and Anabaena 7120 because they are model organisms for the study of photosynthesis and nitrogen fixation, respectively, and genome editing has not yet been described for these strains. We constructed derivatives of the pSL2680 editing plasmid to generate a deletion mutation, a point mutation and a knock-in mutation in each strain.
In Synechocystis 6803 we chose to delete the nblA gene as this would generate the same non-bleaching phenotype that is observed in Synechococcus 2973 25 . Synechocystis 6803 has two adjacent copies of the nblA gene as opposed to the single gene that exists in Synechococcus 2973, requiring a much larger deletion to eliminate both copies. In Anabaena 7120 we chose to delete the first 400 bp of the nifH gene which encodes nitrogenase reductase. The nifH gene is the first gene in the operon that encodes the structural genes for nitrogenase and a knock-out of this gene with conventional methods would disrupt expression of the entire operon. NifH is essential for nitrogen fixation and a nifH mutant would be incapable of diazotrophic growth. We ligated oligos targeting each gene into the pSL2680 vector and cloned a region containing the deletion of nblA or nifH with 1 kb of upstream and downstream sequences into the KpnI site on the vectors containing the modified CRISPR array. The resulting plasmids were conjugated into either Anabaena 7120 or Synechocystis 6803 and 16 colonies were collected from each conjugation. After three rounds of repatching onto BG11Nm 20 or BG11Km 10 , respectively, we performed PCR from upstream of nblA or nifH to a region on the chromosome that is not present on the editing plasmid to determine if the deletion had occurred in the chromosome.
In Synechocystis 6803 we found that 7 of the 16 colonies had a segregated deletion of nblA1/2 (Fig. 3B); while in Anabaena 7120 we found that 10 of the 16 colonies had obtained the deletion of nifH and appeared fully segregated via PCR (Fig. 3A). The lack of more fully segregated colonies in Synechocystis 6803 can be attributed to the extreme level of ploidy in this organism. Ploidy is highly variable in Synechocystis 6803 with one study demonstrating ~200 copies 26 while another shows that copy number is as high as 50 in liquid culture but  varies depending on growth conditions 27 . In contrast, Synechococcus 2973 and Anabaena 7120 maintain less than 10 copies of their chromosome 26 . The high degree of ploidy likely impaired full segregation in 3 patchings in Synechocystis 6803. To further verify segregation, we grew two colonies that contained the deletion of nblA or nifH and subjected them to nitrogen deprivation. Both colonies of Synechocystis 6803 exhibited the expected non-bleaching phenotype while the WT control bleached upon removal of nitrate, indicating they were fully segregated for the nblA deletion (Fig. 3D). Two Anabaena 7120 mutants were spotted onto BG11 -N alongside wildtype. Both edited colonies yellowed while the wildtype Anabaena 7120 spot grew darker and greened indicating the mutants were nif-and thus contained a segregated deletion of nifH (Fig. 3C).

Markerless Knock-ins in Anabaena 7120 and Synechocystis 6803. To engineer a point mutation in
Anabaena 7120, we made a H197G active site mutation in nifD that has been previously shown to abolish nitrogenase activity 28 . The nifD point mutation would also result in a nif-phenotype when segregated. Such a mutation would serve as a good example of the utility of CRISPR technology for studying protein active sites. In parallel, in Synechocystis 6803, we changed the GTG start codon of the isiA gene to GCG to abolish translation initiation of that gene. IsiA is a light harvesting protein that associates with photosystem I and is expressed under conditions of iron limitation. IsiA also forms empty rings that are not associated with PSI where they play a protective role in energy dissipation 29 . When expressed, IsiA generates a 77 K fluorescence emission peak at 685 nm when excited at 430 nm; thus an isiA knockout is easily identified by the absence of this characteristic peak.
Editing plasmids were constructed for each point mutation in the same fashion as was done for the 2973 editing plasmids. Homologous repair templates containing the desired point mutation were included on the editing plasmid. Conjugation of the editing plasmids into the respective strains yielded hundreds of colonies, of which 16 were collected. After 3 rounds of repatching, the nifD or isiA genes of 8 colonies were sequenced to verify the presence of the point mutation. Editing was more efficient in Anabaena 7120 and Synechocystis 6803 compared to Synechococcus 2973 as 7 of the 8 Synechocystis colonies had the mutation while 6 of the 8 Anabaena colonies had the point mutation as determined by Sanger sequencing. Two colonies edited for the nifD H197A mutation were examined for a nif-phenotype to verify segregation. Both colonies yellowed when spotted onto BG11 -N while the WT strain greened indicating that both were fully segregated for the point mutation (Fig. 4A). 77 K fluorescence emission scans of colonies edited for the isiA start codon mutation were examined for the absence of the 685 nm peak to verify segregation of the isiA start codon mutation. All colonies did not exhibit the characteristic peak at 685 nm under conditions of iron starvation indicating that they were fully segregated for the point mutation (Fig. 4B). Anabaena 7120, respectively, with a promoterless eYFP. The gene replacement occurs precisely from start codon to stop codon. These knock-ins would result in a markerless direct gene replacement and place eYFP under the control of the respective promoter. This presents a new capability that cannot be done in a single step with current technology. In fact, such a replacement takes months to perform with existing techniques but can be completed in 2-3 weeks with CRISPR. To engineer the knock-ins, we employed the same crRNA that was used to make the previously discussed deletions of nifH or nblA. The homologous repair template for the deletion was substituted for one in which the coding region of either nifH or nblA is replaced with that of eYFP. Conjugation of the editing plasmids into their respective strains resulted in strains in which eYFP was properly inserted into nifH or nblA operons as verified by PCR between eYFP and a chromosomal region lying outside the homologous repair template (Fig. 5A,B). Sanger sequencing was also used to verify that the eYFP insertions were properly aligned with respect to the nifH or nblA start and stop codons. Six of the 8 Anabaena 7120 colonies had the insertion while 7 of the 8 of the Synechocystis 6803 colonies had the eYFP insertion, after 3 rounds of patching (Fig. 5A,B).
We cured the eYFP knock-ins of their editing plasmids in the same way that is described for Synechococcus 2973 except that in Anabaena 7120 an additional sonication step was used before plating to break up the filaments so that colonies resulting from single cells could be obtained. A cured colony of each strain was examined for eYFP expression under nitrate replete and nitrate deplete conditions as both genes replaced (nblA or nifH) are only expressed upon nitrogen deprivation. In Synechococcus 6803, eYFP expression was observed under nitrate deplete but not nitrate replete conditions (Fig. 5C) indicating that eYFP had taken on the expression pattern of nblA. In Anabaena 7120, eYFP was expressed specifically in heterocysts and only upon removal of nitrate (Fig. 5D). Swapping coding regions in such a way allows one to replace a gene with one with altered activity while maintaining proper regulation. Such gene replacements will be a valuable tool for metabolic engineering as one can directly replace any gene with another to modify metabolic pathways.

Discussion
We previously demonstrated CRISPR genome editing in Synechococcus 2973 by adapting pCRISPOmyces, a cas9 based system from Streptomyces, to function in Synechococcus 13,15 . However, this system was suboptimal because of toxicity issues. Due to cas9 toxicity, the system in our previous work is challenging to generate colonies with. We have never obtained more than 10 colonies from a conjugation and we often obtain no colonies even when using a drawn out and convoluted conjugation protocol. The previously published protocol takes twice as long as conventional conjugations and we have struggled to generate exconjugates in other cyanobacteria or make certain types of edits beyond a deletion. Due to reduced toxicity, with the new Cpf1 based system we obtain hundreds of colonies per conjugation. Only 20% are initially edited; however, it is possible to find a correct colony without further repatching. An additional drawback of the previous system is that up to 9 subculturing steps are required to cure the editing plasmid while the editing plasmid described herein can be cured without perpetual repatching on non-selective media. In the current study, we sought to develop an optimal genome editing system that can function universally across diverse species of cyanobacteria. We used an alternative nuclease that is non-toxic to cyanobacteria and demonstrated different modes of editing in three model cyanobacteria across three separate genera. CRISPR genome editing is a major advance in the field of cyanobacterial genetics. The new tool facilitates rapid and specific modification of target genomes. Furthermore, CRISPR significantly simplifies and accelerates markerless modifications that were previously cumbersome and time consuming. Metabolic engineering projects that typically require months to complete can be completed in just a few weeks using CRISPR. Moreover, edited mutants are fully segregated which further reduces the time and effort needed to generate a clean mutant strain. The system presented here works well across diverse genera of cyanobacteria and may serve as a universal tool for genome editing of these organisms.
Additionally, the RSF1010 based vector that the CRISPR system resides on encodes all three proteins essential for replication of the vector in its host. This allows the vector to replicate in most gram-negative bacteria and some gram-positive bacteria, independent of the host replication apparatus. The RSF1010 vector backbone is known to replicate well in diverse prokaryotes including Salmonella 30 , Pseudomonas 31 , E. coli, Streptomyces 32 , Bacillus 33 , Mycobacterium 32 , Rhizobium 34 and Agrobacterium 35 species. It is reasonable to assume that the versatility of this genome editing system can be extended to other more diverse prokaryotes.
Typically, genes are knocked out using insertional inactivation. A major drawback of this strategy in a bacterial system is that the antibiotic selection cassette generates polar effects on downstream gene on the operon. This prevents one from separating the function of different genes in an operon. Using Cpf1/CRISPR, no operon is disrupted in this process and there are no polar effects from insertion of an antibiotic resistance cassette; thus markerless gene deletions will enable characterization of individual genes in an operon without creating the polar effects of including an antibiotic resistance cassette. Additionally, there is virtually no limit to the number of knock outs that one can make. The ability to remove multiple genes will be an invaluable tool when engineering metabolism to redirect carbon into desired products. The ability to rapidly make markerless single nucleotide changes will be a valuable tool for future efforts in protein engineering, analysis of active sites, structural studies of interesting proteins, and modifications to transcription factor binding sites. Another function of this CRISPR system, direct gene replacement, provides a useful tool for refactoring genomes to generate novel metabolic pathways for the production of biofuels and other value added chemicals. Additionally, the use of markerless knock-ins will allow numerous genes to be inserted into the genome. Overall, the use of markerless modifications will expand the size and complexity of synthetic metabolic pathways that can be assembled leading to the engineering of highly modified cyanobacterial autotrophic cell factories.

Materials and Methods
Strains and Culture Conditions. Synechococcus 2973 was grown on BG11 36 agar plates and in BG11 liquid at 38 °C with 150 μ E• m −2 • s −1 light. Synechocystis 6803 and Anabaena 7120 were grown on BG11 agar plates and in BG11 liquid at 30 °C under 50 μ E• m −2 • s −1 light. Exconjugates of the respective strains were grown under the same conditions on BG11 agar plates supplemented with 10 μ g/mL kanamycin for Synechocystis 6803 and Synechococcus 2973 or 20 μ g/mL neomycin for Anabaena 7120. Cloning was performed in E. coli XL1-blue strain on L-agar with 50 μ g/mL kanamycin. Conjugation was performed with HB101 containing either pRL443 or pRL623 37 with the editing plasmid on 82mm HAF Milipore filters overlayed on BG11 agar supplemented with 5% Luria Broth. Conjugation of editing plasmids into all strains was performed by mixing 100 μ L of overnight cultures of HB101 pRL443 and HB101 pRL623 + editing plasmid with 200 μ L of the cyanobacteria strain adjusted to an OD 720 0.8. After 24 h incubation at 38 °C under 150 μ E• m −2 • s −1 light, Synechococcus 2973 conjugation filters were transferred onto BG11 supplemented with 50 μ g/mL kanamycin. After 48 h incubation at 30 °C under 50 μ E• m −2 • s −1 light, Synechocystis 6803 and Anabaena 7120 conjugation filters were transferred onto BG11 supplemented with either 50 μ g/mL kanamycin or 40 μ g/mL neomycin. Colonies appeared within 3 days for Synechococcus 2973 or 8 days for Synechocystis 6803 or Anabaena 7120.

Construction of Strains.
The plasmid containing cpf1 and the native Francisella novicida CRISPR array, pY002 (pFnCpf1_min) 16 , was obtained as a kind gift from Feng Zhang (Addgene plasmid # 69975). The cpf1 gene was amplified from pY002 with the cpf1 lac-L/cpf1-R primers (Supplementary Table 2) which also fuse a lac promoter onto cpf1. The resulting fragment was cloned into the ApaLI/EcoRI sites on pVZ321 38 to replace the Cm R cassette to generate pSL2668. Next, overlap extension PCR was used to introduce a pair of AarI sites into the first spacer in the CRISPR array of pY002 by amplifying the left and right halves using the J23119ecoL/directrepeat aarI-2 or directrepeat aarI-1/directrepeat-R primers followed by amplification using the J23119ecoL/directrepeat-R primers. The resulting PCR fragment was cloned into the EcoRI/SalI sites on pSL2668 to generate pSL2683. LacZ was amplified from the pCrispomyces-2 39 plasmid using the lacZaarI-L/ lacZaarI-R primers. The resulting fragment was then cloned into the AarI sites on pSL2683 to generate pSL2680, which served as the base plasmid for construction of editing plasmids expressing a full length pre-crRNA. Editing plasmids were constructed by cloning annealed oligos into the AarI sites on pSL2680. The following annealed oligos were ligated into the AarI sites on pSL2680: 7942nblAKOgRNAL/7942nblAKOgRNAR to yield pSL2682; 7942s264agRNAL/7942s264agRNAR to yield pSL2723; NS1gRNAL/NS1gRNAR to yield pSL2724; 6803nblAKOgRNAL/6803nblAKOgRNAR to yield pSL2726; 7120nifHgRNAL/7120nifHgRNAR to yield pSL2728; 7120nifDgRNAL/7120nifDgRNAR to yield pSL2833; and 6803isiAgRNAL/6803gRNAR to yield pSL2834. Next, PCR was used to synthesize the homology regions which were then cloned into the KpnI site on the plasmids containing the matching crRNA. The Synechococcus 7942 nblA homology region containing the deletion of nblA was synthesized from pSL2470 15 using nblAdelRkpnI/nblAdelLkpnI and cloned into the KpnI sites on pSL2682 and pSL2684 to yield pSL2691 and pSL2689 respectively. The homology region containing the Synechococcus 7942 psbA S264A mutation was synthesized using fusion PCR with the 7942psbAL1/7942psbAR2 Scientific RepoRts | 6:39681 | DOI: 10.1038/srep39681 and 7942psbAL2/7942psbAR1 primers followed by PCR with the 7942psbAL1/7942psbAR1 primers. The resulting PCR fragment was cloned into pSL2723 to yield pSL2796. The homology region targeting eYFP to NS1 was synthesized using fusion PCR with the pAM1303NS1L1/pAM1303NS1R2, pAM1303NS1L2/ pAM1303NS1R3 and pAM1303NS1L3/pAM1303NS1R1 primers followed by PCR with the pAM1303NS1L1/ pAM1303NS1R1 primers. The resulting PCR fragment was cloned into pSL2724 to yield pSL2801. The Synechocystis 6803 nblA homology region containing the deletion of nblA1A2 was synthesized using fusion PCR with the 6803nblAdelL1/6803nblAdelR2 and 6803nblAdelL2/6803nblAdel R1 primers followed by PCR with the 6803nblAdelL1/6803nblAdelR1primers. The resulting PCR fragment was cloned into pSL2726 to yield pSL2773. The Anabaena 7120 nifH homology region containing the deletion of nifH was synthesized using fusion PCR with the 7120nifHL1a/7120nifH R2 and 7120nifHL2/7120nifHR1 primers followed by PCR with the 7120nifHL1a/7120nifHR1 primers. The resulting PCR fragment was cloned into pSL2728 to yield pSL2739. The Anabaena 7120 nifD point mutation homology region was constructed in two pieced using nifDL/nifDMR or nifDML/nifDR primers. The homology template was then assembled into pSL2833 linearized with kpnI using Gibson assembly to generate pSL2839. The Synechocystis 6803 isiA point mutation homology region was constructed in two pieces using 6803isiAL/6803isiAMR or 6803isiAML/6803isiAR primers. The homology template was then assembled into pSL2834 linearized with KpnI using Gibson assembly to generate pSL2834. The homologous repair template to insert eYFP into nifH of Anabaena 7120 was synthesized as three fragments using the primers 7120eYfplgibs/7120eYFPR1 or 7120eYFPL1/7120eYFPR2 or 7120eYFPL2/7120eYFPRgibs. The three fragments were assembled into pSL2728 linearized with KpnI using Gibson assembly to generate pSL2840. The homologous repair template to insert eYFP into nblA of 6803 was synthesized as three fragments using the primers 6803eYfplgibs/6803eYFPR1 or 6803eYFPL1/6803eYFPR2 or 6803eYFPL2/6803eYFPRgibs. The three fragments were assembled into pSL2726 linearized with KpnI using Gibson assembly to generate pSL2841. Nitrogenase activity. Cultures were grown to an OD 720 of 0.5 at which time they were washed 1X with BG11 -N and 50 μ l was spotted onto BG11 -N agar plates and grown for 72 hours at 30 °C with 50 μ E• m −2 • s −1 light.

Bleaching Experiments. Synechococcus 2973 and
Fluorescence Microscopy. WT and eYFP containing mutant strains were concentrated 10-fold from mid log phase cultures. Samples were deposited onto glass slides that were coated with 2% polyethyleneimine. Cells were imaged using a Nikon Eclipse 80i microscope equipped with a Photometrics Cool Snap ES CCD camera (Roper Scientific). Filter sets (Chroma) were as follows: YFP was detected using a 480/30 nm excitation filter, a 505 nm dichroic beam splitter, and a 535/40 nm emission filter. Chlorophyll fluorescence was detected using a 560/40 nm excitation filter, a 595 nm dichroic beam splitter, and a 630/60 nm emission filter. A 100 ms exposure time was used for imaging chlorophyll fluorescence and a 1 s exposure time was used to image eYFP expression. 77 K Fluorescence Spectroscopy. Log phase Synechocystis 6803 cells were washed 3X with BG11 -iron and used to start 50 mL cultures in BG11 with or without iron. After 2 days of growth at 30 °C, 50 μ E• m −2 • s −1 light, whole cell fluorescence was observed. Fluorescence emission spectra at 77 K were recorded on a Fluoromax-2 fluorometer (JobinYvon, Longjumeau, France) with excitation at 435 nm. Fluorescence emission curves were normalized as F/F 720 .