Leptolyngbya BL0902 was from Dr. Golden JW (University of California-San Diego). Leptolyngbya BL0902 and derivatives were grown in BG11 in flasks, with manual agitation 3–4 times a day, at 30°C under the illumination of 30 μE·m⁻²·s⁻¹. For selection with antibiotics, spectinomycin (10 μg/mL) or neomycin (25 μg/mL) was added to liquid or solid media as appropriate.

The growth of Leptolyngbya BL0902 and a 50.8 kb-deletion mutant in Zarrouk medium (3 biological replicates) was compared in vertical column-type photobioreactors (48.5 cm × 2.0 cm) bubbled with air supplemented with 1% CO₂ in the light of 100 μE·m⁻²·s⁻¹. The OD₇₃₀ value was adjusted to 0.05 at the beginning and measured every 12 h.

Conjugation was performed as described by Taton et al. (2012), using Escherichia coli HB101 containing pRL443 (conjugative plasmid), pRL623 (helper plasmid) and the plasmid to be transferred into Leptolyngbya BL0902 as the donor strain. The helper plasmid is not required for the conjugal transfer of RSF1010-based editing plasmids, but it is not necessary to remove this plasmid from the donor strain.

The 100 L-scale semi-continuous cultivation of Leptolyngbya BL0902 P_psbA -desD was carried out in Zarrouk medium bubbled with 4% CO₂ (v/v) at 26°C ~ 29°C, in a 1-m² open raceway photobioreactor, with the constant illumination of 100 μE·m⁻²·s⁻¹ from both upper and lower sides (Figure 1). The depth of the culture was kept at 10 cm by replenishing water every day to compensate for the evaporative loss. An electric motor-driven paddle wheel, 0.3 m in diameter, rotating at 20 rpm, was used to propel the circulating movement of the culture. The initial OD_730nm of the culture was ~0.1. Starting from the 5th day after the inoculation, cells were harvested once every 2 days; at each harvest time point, 50 L of the culture were collected, and the same volume of fresh medium was supplemented to the culture. The growth of cells was recorded based on the dry weight of biomass per liter in two biological replicates. At each time point, 50 mL of cells taken from the culture were vacuum filtered, washed with 0.5 N HCl and dried at 105°C for 4 h, and the dry weight was measured.

Fatty acid composition was analyzed as previously described (Laurens et al., 2012) with 3 biological replicates. Lipids were extracted from 10 mg freeze-dried cells and simultaneously transesterified, with sequentially added chloroform/methanol (2:1) and 5% HCl in methanol, and analyzed by GC (Trace Ultra, Thermo Electric) equipped with a DB-23 capillary column (60 m in length, 0.25 mm in diameter, 0.25 μm in film thickness) coupled to a FID detector. For quantification of the GLA content in Leptolyngbya strains, tridecanoic acid (C13:0) was added to the dried cells as an internal standard before extraction and transesterification of lipids.

Molecular cloning manipulations were performed according to standard protocols, but some ligation reactions were performed using ClonExpress Ultra One step Cloning Kit V2 (Vazyme, Nanjing, China) based on in vitro homologous recombination. DNA fragments generated by PCR were confirmed by sequencing after cloning in plasmids. Details of plasmid construction and strain generation are provided in Supplementary Table S1 but briefly described as below.

For interruption of desB with the omega cassette (Prentki and Krisch, 1984) or P_psbA-desD-omega in Leptolyngbya BL0902, pHB6115 and pHB6127 were constructed, with these fragments inserted at the Bal I site of Leptolyngbya desB, cloned into pRL271, a sacB-bearing vector (Cai and Wolk, 1990). The desD gene was generated by PCR using Synechocystis PCC 6803 DNA as the template.

For generation of deletions in the chromosome of Leptolyngbya BL0902, a DNA fragment with the two homologous arms for double crossover recombination was generated by overlap PCR (Horton et al., 1989) and cloned into the Bgl II-BamH I site of a pCpf1 plasmid, and a short dsDNA with the gRNA sequence was cloned into the plasmid replacing the DNA fragment between the two Aar I sites. Such constructed plasmids pHB7912, pHB7913, pHB7914 and pJS2529 were used to generate the ΔhetR, Δ[patX-hetR], ΔpatX and Δ07990-08580 mutants, respectively.

These plasmids were introduced into Leptolyngbya BL0902 by conjugation (Taton et al., 2012), and exconjugants were randomly picked and streaked on plates with appropriate antibiotics, followed by selection in liquid medium. Genomic DNA was extracted from these exconjugants with a quick mini-preparation method (Cai and Wolk, 1990) for PCR examinations.

High-molecular-weight genomic DNA was extracted from Leptolyngbya BL0902 using the cetyltrimethylammonium bromide (CTAB) method (Murray and Thompson, 1980), and RNA was removed by using RNase A. The genome sequence was determined by third generation long read sequencing technology on PacBio platforms (Pacific Biosciences, CA, United States). A total of 283,232 reads were obtained, encompassing 1.7 Gb, with an average read length of 5,990 bp. In addition, Illumina sequencing was performed on Illumina HiSeq 2,500 (Illumina Inc., San Diego, CA, USA) to obtain high quality reads, producing 3 Gb high quality data. The genome was assembled using HGAP3 (Chin et al., 2013) based on PacBio sequencing reads and polished using Illumina reads. Finally, the Leptolyngbya BL0902 genome was assembled into a complete circular chromosome with no gap and five circular plasmids.

Protein-coding genes were predicted by using Glimmer3 (Delcher et al., 2007) with default parameters, and the predicted proteins were annotated by searching against NCBI non-redundant protein (NR),¹ Swiss-Prot (Wu et al., 2006), COG (Tatusov et al., 1997), KEGG (Kanehisa and Goto, 2000) and GO (Ashburner et al., 2000) databases. The rRNA sequences were identified using RNAmmer (Lagesen et al., 2007), tRNA genes identified using tRNAscan-SE (Lowe and Eddy, 1997), tandem repeats (> 6 bp) identified using Trf407b.linux in RepeatMasker,² and regularly interspaced short palindromic repeat (CRISPR) arrays identified using MinCED.³

Leptolyngbya strains grown to OD_730nm ~ 0.8 were collected by centrifugation, quickly frozen and stored in liquid nitrogen. Total RNA was extracted from cells using the RNA pre-purification Cell/Bacteria Kit (Tiangen Biotech Co., Ltd., Beijing, China), then DNA was removed with DNase I, and rRNA was removed using the Ribo-off rRNA Depletion Kit V2 (Bacteria) (Vazyme, Nanjing, China). One microgram of total RNA was used to construct RNA libraries after retro-transcription, and 150 nt paired-end sequencing was performed on Illumina Novaseq 6,000.

The raw paired-end reads were trimmed and quality controlled by fastp version 0.21.0 (Chen et al., 2018) with default parameters, then clean reads (over 3 G for each strain) were separately aligned to reference genome using hisat2 version: 2.0.1-beta (Kim et al., 2015). The featureCounts function from the Subread package (Liao et al., 2013) was used to count reads that mapped to each one of the protein-coding genes. Raw count data was then used as input into DESeq2 v.1.42.0 (Love et al., 2014) for differential expression analyses (3 biological replicates). Differential expression was considered significant if the absolute FoldChange value was >2 or < 0.5 and the false discovery rate (FDR) adjusted p-value was <0.05. After applying a regularized-logarithm transformation to the raw count gene matrix, we calculated the average expression value of each gene across three biological replicates, then standardized the expression of each gene along the samples to generate a standardized matrix and visualized the result as a heatmap using the package ggplot2 v.3.4.4 (Wickham, 2016), in R.

The genome of Leptolyngbya BL0902 (GenBank accession numbers: CP046155-CP046160) is composed of six circular DNA molecules, the chromosome and five plasmids (Table 1), with a total size of 4.71 Mb. The 4.309 Mb-long chromosome is predicted to have 3,981 protein-coding genes, 6 rRNA (two copies of rRNA operons) and 43 tRNA genes; the five plasmids, with sizes of 158.483 kb, 93.510 kb, 78.547 kb, 44.466 kb and 25.969 kb, contain 177, 84, 66, 68 and 27 protein-coding genes, respectively. To our best knowledge, this is the smallest among the Leptolyngbya genomes (Table 2) and those of the closely related genus Nodosilinea (Supplementary Table S2). As seen in some cyanobacteria with rRNA operon variants, such as Anabaena sp. PCC7120 (Iteman et al., 2000), the two 16S-23S internal transcribed spacer regions of Leptolyngbya BL0902 contain either tRNA-Ile or tRNA-Ala gene. In addition, 571 tandem repeats (> 6 bp) and 16 clustered regularly interspaced short palindromic repeat (CRISPR) arrays were identified in the genome (Table 1).

Some cyanobacterial species produce cyanotoxins that pose risks to human and animals, and the syntheses of cyanotoxins are typically dependent upon conserved gene clusters. We searched the genome of Leptolyngbya BL0902 for all the eleven representative cyanotoxin biosynthesis gene clusters (Pearson et al., 2016) but found no similar one. Type II restriction enzymes may cut foreign DNA and greatly reduce gene transfer efficiency in cyanobacteria (Elhai et al., 1997). However, no genes (homologs) for type II restriction enzymes from REBASE (Roberts et al., 2010) were found in the genome of Leptolyngbya BL0902.

Cyanobacteria can be classified into 4 groups according to the types of fatty acid desaturases (Murata and Wada, 1995): (1) those only with DesC that generates a single double bond at position 9 (Δ9, from the C-terminus) on the fatty acid chain in acyl-lipids; (2) those with DesA (Δ12), DesB (Δ15) and DesC; (3) those with DesA, DesC and DesD (Δ6); (4) those with DesA, DesB, DesC and DesD. In the genome of Leptolyngbya BL0902, we found genes encoding DesA (GFS31_05880), DesB (GFS31_39000) and DesC (GFS31_27570), therefore it should be a strain of group 2.

In filamentous cyanobacteria, there is a monophyletic group that form specialized cells termed heterocysts for nitrogen fixation (Wolk et al., 1994). Some non-heterocystous cyanobacteria can also perform nitrogen fixation (Rippka and Waterbury, 1977), and these diazotrophic cyanobacteria all possess the nif (nitrogen fixation) gene cluster (Tsujimoto et al., 2014). Leptolyngbya BL0902 is a non-heterocystous cyanobacterium without the nif gene cluster; however, it possesses genes (GFS31_32400, GFS31_16620-GFS31_16630) similar to hetR (Buikema and Haselkorn, 2001) and hetZ-patU (Zhang et al., 2007), which play central roles in regulation of heterocyst differentiation. In heterocyst-forming cyanobacteria, HetR directly regulates the expression of hetZ via a HetR-binding site, while HetZ regulates the expression of hetR and two genes encoding RG(S/T)GR-containing peptides, namely patS and patX, directly or indirectly via the DIF1 promoter (Du et al., 2020). HetR also activates or inhibits the expression of some other genes, such as hetP and the hetP-like gene alr3234 in the heterocyst-forming cyanobacterium Anabaena sp. PCC 7120 (Hou et al., 2015). The pentapeptide RG(S/T)GR, derived from PatS (Yoon and Golden, 1998), PatX (Elhai and Khudyakov, 2018) and a protein called HetN (Higa et al., 2012), is an inhibitor of HetR. In Leptolyngbya BL0902, patS and hetN are not found, but a protein-encoding gene, GFS31_32390, meets the definition of patX (Elhai and Khudyakov, 2018); of the three ORFs similar to hetR, the one (GFS31_32400) with the highest similarity is located immediately downstream of patX and appears to be co-transcribed with patX from the predicted DIF1 promoter (Supplementary Figure S1).

A gene transfer system based on wide-host-range plasmids has been established in Leptolyngbya BL0902 (Taton et al., 2012; Poole et al., 2020). However, manipulations of genetic loci on the chromosome are required for gene function analyses and generation of more stable cell factories. To this end, we tested integration of genes into the chromosome by homologous double-crossover. In Leptolyngbya BL0902, the three fatty acid desaturases generate double bonds at Δ9, Δ12 and Δ15 positions on C₁₈ fatty acid chains of acyl-lipids, leading to the formation of α linolenic acid (ALA). First, we constructed the plasmid pHB6125 (Supplementary Table S1) for interrupting desB with the omega cassette (Sp^r) in Leptolyngbya BL0902 (Figures 2A,B). A DNA fragment containing desB was cloned in the plasmid, and the omega cassette was inserted at the Bal I site of desB. sacB on the vector can provide the positive selection for double-crossover recombinants. We planned to introduce the plasmid into Leptolyngbya BL0902 by conjugation to obtain single-crossover recombinants, then to select double-crossover mutants on sucrose-containing plates. However, when we checked 9 randomly picked spectinomycin-resistant exconjugants by PCR examinations, one of them was already double-crossover mutant (ΔdesD), whereas the rest 8 were single-crossovers, therefore the selection on sucrose-containing plates was omitted. Then, we tried to interrupt desB with P_psbA -desD and the omega cassette, so as to replace α-linolenic acid with γ-linolenic acid (GLA, C₁₈ fatty acid with double bonds at Δ6, Δ9 and Δ12) in Leptolyngbya BL0902. P_psbA is a strong promoter from the chloroplast of Amaranthus hybridus (Elhai, 1993), while desD is from Synechocystis PCC 6803. The plasmid pHB6127 was constructed in a structure similar to pHB6125, but a fragment with P_psbA -desD and the omega cassette was inserted into the BalI site of desB. Of 11 randomly picked exconjugants, 10 formed single-crossover recombination between pHB6127 and the chromosome, but one directly formed double-crossover, with P_psbA-desD and the omega cassette inserted within desB, and this desD-overexpressing strain was called P_psbA-desD (Figures 2A,B).

Gas chromatography analyses of fatty acid composition showed that unlike in the wild type, ALA was no longer formed in the ΔdesB mutant (Figures 2C,D; Supplementary Table S3). In the P_psbA-desD strain, ALA (18:3α) was replaced with GLA (18:3γ), which accounted for 18.02% ± 0.92% of long chain fatty acids. Compared to the fatty acid composition in the wild type, C18:3 increased in this strain at the expense of mono- and di-unsaturated C18 fatty acids (Figure 2D; Supplementary Table S3). The desD-overexpressing strain and GLA content were stable over past 6 years under laboratory conditions. To demonstrate the potential for commercial production, we evaluated the biomass and GLA productivities of this strain under semi-continuous cultivation conditions in a 1-m² open raceway photobioreactor (Figures 1, 2E). On the 5th, 7th and 9th day, half of the culture was collected for cell harvest, and fresh medium of the same volume was supplemented to the culture. Productivities were calculated based on the increase from the 5th day to the 7th day and that from the 7th day to the 9th day (3 samples × 2). The cultivation was extended to the 16th day. The GLA contents gradually increased at the four time points, reaching ~2% of the biomass (dry weight). Based on two independent cultivation experiments (3 samples × 2 × 2), we calculated that the productivity of biomass was 19.1 ± 4.9 g·m⁻²·d⁻¹, and that of GLA was 0.36 ± 0.14 g·m⁻²·d⁻¹ (p < 0.05).

For analyses of gene functions or biotechnological genetic manipulations, markerless deletions or insertions are sometimes required. The CRISPR/Cpf1 editing system is suitable for such purposes (Ungerer and Pakrasi, 2016; Niu et al., 2018). hetR and patX are two genes found in almost all filamentous cyanobacteria, heterocyst-forming or not (Elhai and Khudyakov, 2018). The functions of these two genes in those species that do not form heterocysts remain a mystery. We constructed Cpf1-based editing plasmids pHB7912, pHB7913 and pHB7914 (Supplementary Table S1), for deleting hetR, patX-hetR and patX, respectively. The editing plasmids were introduced into Leptolyngbya BL0902 by conjugation, and exconjugants were checked with PCR using specific primers (Figures 3A–D; Supplementary Table S1). In the resulted mutants ΔhetR, Δ[patX-hetR] and ΔpatX, the predicted DIF1 promoter upstream of patX-hetR (Supplementary Figure S1) remained unchanged. The editing plasmids were then removed from the mutants by positive selection on sucrose-containing plates, and the removal of plasmids was confirmed by PCR examination and antibiotic-resistance assay (data not shown).

In heterocyst-forming cyanobacteria, PatX is thought to be one of the precursors for the RG(S/T)GT-containing inhibitor of HetR (Elhai and Khudyakov, 2018; Khudyakov et al., 2020). We wondered whether PatX counteracts the regulatory effects of HetR, either activating or inhibitory, on gene expression in Leptolyngbya BL0902. Transcriptomic analyses were performed to compare the transcriptional profiles of the WT and mutants. The results showed that there were 993 genes differentially expressed in ΔhetR vs the wild type, 428 in ΔpatX, 477 in Δ[patX-hetR], including genes involved in photosynthesis, respiration, nitrogen metabolism, etc. Of these genes, 126 were up- or down-regulated in all these mutants compared to the wild type. According to the differential expression patterns in mutants compared to the wild type, these 126 genes can be classified into 5 groups (Supplementary Figure S2, with 2 groups shown in Figure 3E as examples). It appeared that HetR downregulates genes in the first and second groups but upregulates those in the third, fourth and fifth groups. Only in groups 2 and 3, PatX showed opposite effects on gene expression compared to HetR. If PatX acts as a (pre)inhibitor of HetR, the effect of patX on gene expression should be dependent upon that of hetR. However, in groups 2 and 3, Δ[patX-hetR] was similar to ΔpatX but contrary to ΔhetR in differential expression. Apparently, the effects of PatX on gene expression are basically independent of HetR in Leptolyngbya BL0902.

Deletions of large non-essential regions in the chromosome are required for genome streamlining in cyanobacteria, which would reduce the genomic complexity and may improve some productive traits (Hou et al., 2023; Sengupta et al., 2024). By BLAST searching against the essential genes experimentally identified in Synechococcus elongatus PCC 7942 (Rubin et al., 2015), we obtained a list of genes that are probably essential in Leptolyngbya BL0902. Then we tried to delete some chromosomal regions without these genes, for example, a 50.8-kb region (chromosomal bp 825,418–876,218), extending from GFS31_07990 to GFS31_08580. The Cpf1-based editing plasmid pJS2529 (Supplementary Table S1) was constructed and transferred into Leptolyngbya BL0902 by conjugation. Of 20 randomly picked exconjugants, 6 exhibited complete deletion of the 50.8-kb region. The complete segregation of the resulted mutant, Δ07990-08580, was confirmed by PCR using 3 pairs of primers (Figure 4). When grown in Zarrouk medium in column photobioreactors with aeration, Δ07990-08580 showed a slightly reduced growth rate compared to that of the wild type (Figure 4), but this does not compromise the feasibility of genome streamlining in Leptolyngbya BL0902.