Tradeoffs between phage resistance and nitrogen fixation drive the evolution of genes essential for cyanobacterial heterocyst functionality

Abstract Harmful blooms caused by diazotrophic (nitrogen-fixing) Cyanobacteria are becoming increasingly frequent and negatively impact aquatic environments worldwide. Cyanophages (viruses infecting Cyanobacteria) can potentially regulate cyanobacterial blooms, yet Cyanobacteria can rapidly acquire mutations that provide protection against phage infection. Here, we provide novel insights into cyanophage:Cyanobacteria interactions by characterizing the resistance to phages in two species of diazotrophic Cyanobacteria: Nostoc sp. and Cylindrospermopsis raciborskii. Our results demonstrate that phage resistance is associated with a fitness tradeoff by which resistant Cyanobacteria have reduced ability to fix nitrogen and/or to survive nitrogen starvation. Furthermore, we use whole-genome sequence analysis of 58 Nostoc-resistant strains to identify several mutations associated with phage resistance, including in cell surface-related genes and regulatory genes involved in the development and function of heterocysts (cells specialized in nitrogen fixation). Finally, we employ phylogenetic analyses to show that most of these resistance genes are accessory genes whose evolution is impacted by lateral gene transfer events. Together, these results further our understanding of the interplay between diazotrophic Cyanobacteria and their phages and suggest that a tradeoff between phage resistance and nitrogen fixation affects the evolution of cell surface-related genes and of genes involved in heterocyst differentiation and nitrogen fixation.

biosynthesis.This gene is located in the same operon as alr4494, however, to the best of our knowledge, it was not reported to be related to nitrogen fixation or heterocyst formation or functionality.The cells of this strain were 15% narrower than those of the WT, yet no significant change in their length was observed (Fig. 1A and S1A).In addition, a synonymous mutation was observed in this gene in the susceptible wild types (Table S2).
RB4: Similarly to RB2, this strain had a mutation in alr4491.However, while the mutation in RB2 caused a frameshift, the mutation in RB4 caused an amino acid substitution.As in RB2, cells of RB4 were 15% narrower than those of the WT, yet no significant change in length was observed (Fig. 1A and S1A).
RD1: This strain had a single mutation in a gene encoding for a DNA-binding transcriptional regulator (all1719; Table 1,S1).This resistant strain had a reduced heterocyst induction (27%; Table 1) and nearly no nitrogenase gene expression (4%) and activity (3%).This suggests that this gene takes part in the regulation of heterocyst differentiation as well as in nitrogen fixation in the induced heterocyst cells (Fig. 3).Nevertheless, we did not observe an impaired structural appearance of the heterocyst cells (Fig. 3D).Homologs of this gene are found in other members of the Nostocales, as well as in other nitrogen fixing Cyanobacteria and heterotrophic bacteria, however, in a very low similarity, and it is missing from others (Fig. S4).Moreover, this gene clustered, together with a few cyanobacterial genes, with heterotrophic bacteria, while homologs in other Cyanobacteria strains clustered separately (Fig. S4).This suggests that this gene is an accessory gene acquired by lateral gene transfer.To our knowledge, the relation of this gene to nitrogen fixation and heterocyst induction has not been studied yet.
RD2: This strain could not produce heterocysts and had no nitrogenase activity.Moreover, the cells of this strain had an aberrant morphology (Fig. 1B and 3D) both in nitrogen rich and poor environments.It has two SNPs in two loci.The first mutation is intergenic: 518 bp upstream of all1059 and 197 bp upstream of all1060.all1059 encodes for a sucrose synthase, which is involved in disaccharides metabolism.This gene was up-regulated during desiccation and might have a role in the synthesis of osmoprotacnts [5].A previous study showed that a homolog of this gene in Anabaena variabilis is involved in the synthesis of polysaccharides [6], a main component of the bacterial lipopolysaccharide (LPS) that can serve as a receptor for an infection by the phage.Indeed, LPS serves as a receptor for A-4L [3].all1059 had homologs in all Nostocales strains analyzed in this study, except the Cylindrospermopsis and Raphidiopsis strains.all1060 encodes for a 150 bp long hypothetical protein, that looks like a Cyanobacteria specific core gene.The distance of the mutation from both genes (especially from all1059) makes it unclear whether this mutation takes any part in the regulation of these genes, and thus whether it affects any of the phenotypic changes (heterocyst differentiation or phage resistance) of this strain.
The second mutation was found in the gene all2170, which encodes for a Caspase HetF Associated with the Tetratricopeptide repeats (CHAT) domain containing protein.HetF (encoded by alr3546) is an essential regulator of heterocyst differentiation in Nostoc punctiforme and in Nostoc 7120 [7][8][9].Furthermore, HetF also has a role in cell division of Nostoc 7120 cells, though it is not essential for division [10].Both HetF and all2170 belong to the same protein family (pfam12770).The HetF family belongs to the caspase-hemoglobinase fold (CHF) proteases.These CHF proteases can serve as negative regulators in cases where several copies of CHF proteins are expressed, by creating a dimer [11].Homologs of this gene were found in most of the Cyanobacteria strains analyzed, similarly to the hetF gene.However, while some strains The phenotype of RD2 is similar to that of previously described mutants of hetF which might imply a similar function in context of heterocyst differentiation (See main text).
Resistance in this strain was due to a significant reduction (~90%) in the adsorption of the phage to this strain, when compared to the susceptible ancestor.It was previously shown that the O antigen of lipopolysaccharide (LPS) of Nostoc 7120 serves as the receptor of A-4L [3].Multiple genes take part in the construction and modification of the O antigen, and thus a mutation in various loci within these genes can potentially affect the structure of the O antigen and the ability of the phage to identify this receptor.It is thus still unclear which mutation is the resistance conferring mutation in this strain.
RD3: This strain had a significantly impaired adsorption of phage particles to the cell surface (Fig. 1C).Additionally, the induction of heterocyst cells as well as the expression of the nifH gene, under nitrogen starvation, were lower by 68% and 80% respectively, compared to those of the paired control (Fig. S3).Yet, the heterocysts of this strain had a normal appearance and no tradeoff has been observed during nitrogen starvation (Fig. S2, S3).This strain carries a single mutation in all1058.This gene encodes for a protein that has two domains: trehalose and maltose phosphorylase, and β -phosphoglucomutase.This gene was found to be up-regulated in Nostoc 7120 during desiccation, suggesting that it might has a role in the synthesis of osmoprotectans [5], which are often polysaccharide molecules [12].βphosphoglucomutase is a type of isomerase which catalyzes the reaction D-glucose 1-phosphate and D-glucose 6-phosphate.It has been shown that β-phosphoglucomutase formation is induced by the presence of maltose and it is involved in cell wall polysaccharides biosynthesis [13][14][15][16].
Modifications in the cell wall polysaccharides may be the cause for the impaired adsorption of the phage A-4L to this strain [3].It is unclear how this mutation caused a decrease in the ability of strain RD3 to induce heterocyst cells as well as a decrease in nifH expression and at the same time did not significantly affect the growth in nitrogen poor medium.

RD4:
This strain had a significant reduction in growth in comparison to its susceptible paired control, however, while it was able to grow in nitrogen rich medium, its growth was almost arrested under nitrogen starvation (Fig. S2).The phenotypic appearance of the heterocysts showed the lack of cyanophycin polar granules (Fig. S3) which is a nitrogen reserve of fixed nitrogen, that can be transferred to the vegetative cells in the form of arginine and aspartate [17,18].This might suggest that impaired nitrogen fixation activity of RD4 leads to the reduced growth under nitrogen deprivation.We identified in RD4 a single mutation in the gene alr4485, which encodes for an ABC-type polysaccharide/polyol phosphate export permease.This transporter is embedded in the inner membrane in gram negative bacteria, and exports a variety of substrates, among them, cell wall components, such as polysaccharides, and the O-antigenic polysaccharide [19,20].Therefore, a mutation in this gene may affect the cell wall composition and thus cause impaired phage adsorption.A previous study showed that the gene hetC (alr2817), which has a role in the regulation of differentiation to heterocysts, is similar to bacterial ABC membrane transport proteins [21], though aligning the sequences of hetC and alr4485 showed no significant similarity between them (data not shown).
RD5: This strain had reduced growth under nitrogen starvation conditions relative to the susceptible paired control (Fig. S2).Bright field image of RD5 suggests structurally normal heterocyst cells (Fig. S3).We identified SNP in a gene (alr1906) predicted to be a member of the FhaB family (Filamentous hemagglutinin), which is a part of a two-partner secretion system (Tps).This protein (TpsA) is transported by a transporter protein (TpsB) from the periplasmic side and secreted to the extracellular medium, in various gram-negative bacteria [22].A study that examined two TpsB-like proteins in Nostoc 7120, suggested that alr1906 is one of the substrates of the studied TpsB-like proteins [23].Under nitrogen starvation, strains carrying mutations in the TpsB-like proteins, showed structurally normal heterocysts, yet a reduced growth relative to the control strain was noticed.Moreover, this disruption was accompanied by resistance to antibiotics (erythromycin, roxithromycin, and tylosin).

RE1:
This strain had a reduced ability to induce heterocyst cells, however, this reduction was limited in comparison to other resistant strains (Fig. 3B).While this strain had only 27% less heterocysts in comparison to the wild type, it had a very low nifH expression and no nitrogenase activity (Fig. 3), and it could not survive nitrogen starvation (Fig. 2B).The heterocysts of this strain had a deformed morphology and sometimes appeared in groups of 2-3 adjacent cells in different stages of differentiation (Fig. 3D), which may explain their malfunction.Furthermore, this strain had a significantly impaired adsorption of phage A-4L.RE1 has a mutation in the all5019, which encodes for 3-phosphoshikimate 1-carboxyvinyltransferase that is involved in aromatic amino acid biosynthesis (Powell et al. 1992).Phylogeny of this gene suggests that this gene was acquired by lateral gene transfer (Fig. S4) as has been shown previously using phylogeny, structure and activity of this enzyme [24].We identified the same mutation in the substrain RE8 (Table S1), which originated from the same susceptible ancestor (WT-E).
However, the mutation in RE8 was one of five mutations, which suggests that RE8 evolved from RE1.To the best of our knowledge, this gene was not linked previously to nitrogen fixation of heterocyst differentiation or to resistance to phages.
RE3: This strain could not survive nitrogen starvation (Fig. S2) and was able to induce heterocyst cells with a similar phenotype to RE1.We found in this strain a mutation in an intergenic region upstream to all3346 that encodes for a Repeats-in-toxin (RTX) toxin-related Ca 2+-binding protein, and downstream to all3347 that encodes for a CHASE2 domain-containing protein.RTX proteins are a family of proteins, which are secreted via a type I secretion system.RTX proteins have various functions in gram-negative bacteria.In unicellular Cyanobacteria was not predicted to be a promoter site, the morphological structure of its heterocysts suggests a regulatory role of this locus which is common for non-coding regions.
RE4: This strain produced significantly less (60%) heterocysts than the WT (Fig S3).Although this strain produces heterocysts that appear to be structurally normal, the population of this strain had reduced growth under nitrogen starvation (Fig. S2), followed by drastically low nifH expression (Fig. S3).Moreover, our experiments indicate that the resistance of this strain to A-4L was due to significantly impaired adsorption of phage particles (Fig 1C).This strain had a SNP in the gene all7617, which is located on the beta plasmid of Nostoc 7120 (Tables S1, S3) and encodes for a CusA/CzcA family heavy metal efflux RND transporter.A previous study showed that all7617 was upregulated under nitrogen starvation similarly to its two homologs in the Nostoc 7120 genome: all7618 and all763 (Flaherty et al., 2011, suppotrting information).
These two genes are also located in the beta plasmid and encode for CusA/CzcA family heavy metal efflux RND transporters.

RE5:
This strain can induce heterocysts, however, it collapsed during nitrogen starvation (Fig. S2).This strain has a synonymous mutation in alr5237 that encodes for glycosyltransferase that is involved in the synthesis of molecules such as polysaccharides that can affect the structure of the cell surface (Lairson et al. 2008).This SNP is positioned on the 30 th amino acid (leucin, TTA became TTG).Though silent mutations result in the same amino acid, such mutations can affect protein folding [26], which can be crucial for enzyme activity.When analyzing the main chromosome to understand the codon usage in Nostoc 7120, TTA and TTG have almost equal frequency throughout the genome (0.21 and 0.22 respectively).However, alr5237 itself has a different codon usage for leucine, which is compatible with the fact that this gene was acquired by lateral gene transfer (Fig. S4).TTA and TTG have usage frequency of 0.47 and 0.14 respectively, which indicates a codon usage bias.We did not observe any difference between the mRNA secondary structure in silico when replacing the 30 th codon (TTA) with TTG.When scanning the genome for tRNAs, the frequency for TTA is 0.125 while for TTG it is 0.25.Such differences may affect the translation of this gene.Another resistant substrain (RE6) that belongs to the same lineage, had the same mutation (Table S1).Since RE6 had two additional mutations, which suggests that RE6 evolved from RE5.

RE2:
The heterocyst induction of RE2 was lower by 65% (Fig. 3B) than that of the wild type strain.Although this strain could produce structurally normal heterocyst cells (Fig. 3D), no nitrogenase activity was observed (Fig. 3A), the expression of the nifH gene was significantly lower (Fig. 3D), and it could not survive nitrogen starvation (Fig. 2B).
Strain RE2 has three mutations (Table S1).The first mutation is a single base deletion in the gene alr4949, which has two conserved domains: a serine/threonine protein kinase (residues ~20-270) and a porphyrin-binding protein domain GUN4 (residues ~290-410).The mutation was detected in the second domain.GUN4 is involved in chlorophyll biosynthesis regulation and intracellular signaling in photosynthetic organisms such as plants and algae [27,28].Though, previous studies on green algae reported reduced chlorophyll biosynthesis followed by pale green color of the GUN4 mutants [29,30], we did not observe remarkable color change of RE2 compared to the susceptible paired control.
An additional mutation was identified in all1304 (deletion of one nucleotide), which encodes for a bicarbonate transporter (BicA).We identified different mutations in this gene both in the susceptible controls and in an additional resistant strain that was mentioned previously (See RA8).
The third mutation in this strain was in the gene alr4492, which encodes for a glycosyltransferase family 2 (GT2) protein.As mentioned above, glycosyltransferases are involved in the synthesis of polysaccharides, which may affect the cell surface structure and thus the ability of phages particles to adsorb to the cell.We identified a mutation in this gene in the strain RF16 (among another three mutations) which was not part of our phenotypic analysis.This gene is part of an operon enriched with cell surface related genes that were found to be essential for phage infection in other strains isolated in this study.Variovorax boronicumulans J1 1 GCA and GCF represent data taken from GenBank and RefSeq respectively. 2The full names of the organisms used in the heatmap, pangenome analysis, and phylogenetic analysis.S4).
RTX proteins are related to cellular motility (Linhartová et al. 2010, Baumann 2019).Hahn et al. (2015) identified three possible members of the RTX proteins family in Nostoc 7120 (all0275, alr4238 and all7128) but their function remains unclear.Though the mutation position of RE3

Figure S1 :
Figure S1: Phenotypes of the resistant substrains of Nostoc 7120.Cell (A) length of the resistant substrains (light blue) and their susceptible ancestors (grey).Data shown are average +/standard deviation of n replicates.*** p<0.001.B.Bright Field images of resistant substrains and their susceptible ancestors.Scale equals 10µm.

Figure S2 :
Figure S2: Growth cost of resistant substrains.Growth of resistnat substrains (purple for Nostoc 7120, green for C. raciborskii) and wild types (WT; black) and their suceptible paired controls.Growth was measured as chlorophyll a autofluorescece with (left panel) or without (middle panel) combined nitrogen.The data presented are average and standard deviation (+/-) of n biological replicates.The corresponding statistical significance of the difference in the growth of the resistant substrains and their susceptible paired controls with (+N) or without (-N) nitrogen as well as the cost in the presence vs. absence of nitrogen (+N vs. -N) are presented in the tables on the right panels (NS non-significant, * p<0.05, ** p<0.01, *** p<0.001).p-values were calculated using two-way repeated measure ANOVA tests.

Figure S3 :
Figure S3: Cost of resistance in nitrogen fixation.A. Percentage of filaments with heterocysts of the resistant substrains (light blue) relative to their susceptible ancestor (WT, grey), 48 hours after nitrogen stepdown.B. Expression of nifH gene in the resistant substrains (light blue) relative to their susceptible ancestor (WT, grey), 48 hours after nitrogen stepdown.The transcript levels of nifH values are normalized to transcript levels of rnpB.Data shown are average and standard deviation (+/-) of n biological replicates.*p<0.05,**p<0.01,***p<0.001.C. Bright field images and the corresponding fluorescence images of Nostoc 7120 substrains, 48 hours after nitrogen stepdown.White arrows indicate heterocyst cells.Scale equals 10 µm.

Figure S4 :
Figure S4: Phylogeny of mutant genes.Maximum likelihood trees of genes involved in resistance to phages and/or in nitrogen fixation in Nostoc 7120.The locus tag of the analyzed gene is marked above each phylogenetic tree.Blue, analyzed gene; purple, paralogs in Nostoc 7120; pink, homologs in other Nostocales; orange, homologs in non-Nostocales Cyanobacteria; black, homologs in bacteria from different phyla.The database used for this analysis include genomes of 48 bacteria (TableS4).

Figure S6 :
Figure S6: Phylogeny of genes in Cylindrospermopsis phage Cr-LKS4.Maximum likelihood trees of genes with a blastp best hit that was not in phage CrV.Light blue -Cylindrospermopsis, Greenother Cyanobacteria, pinkcyanophage, blackother.The gene of interest in Cr-LKS4 is marked by a phage image.Bootstraps values greater than 50 (out of 100) are shown.Cyl, Cylindrospermopsis; Lepto, Leptolyngbya; Pseud, Pseudomonas; Micro, Microcystis.

Table S6 :
Genetic variation between the phages that infect C. raciborskii.

Table S8 :
Initial assembly of phage genomes.

Table S9 :
Primers used to connect the phage scaffolds.