Functional dissection of the three-domain SepJ protein joining the cells in cyanobacterial trichomes


  • Vicente Mariscal,

    1. Instituto de Bioquímica Vegetal y Fotosíntesis, Consejo Superior de Investigaciones Científicas and Universidad de Sevilla, Américo Vespucio 49, E-41092 Seville, Spain
    Search for more papers by this author
  • Antonia Herrero,

    1. Instituto de Bioquímica Vegetal y Fotosíntesis, Consejo Superior de Investigaciones Científicas and Universidad de Sevilla, Américo Vespucio 49, E-41092 Seville, Spain
    Search for more papers by this author
  • Anja Nenninger,

    1. School of Biological and Chemical Sciences, Queen Mary University of London, Mile End Road, London E1 4NS, UK
    Search for more papers by this author
  • Conrad W. Mullineaux,

    1. School of Biological and Chemical Sciences, Queen Mary University of London, Mile End Road, London E1 4NS, UK
    Search for more papers by this author
  • Enrique Flores

    Corresponding author
    1. Instituto de Bioquímica Vegetal y Fotosíntesis, Consejo Superior de Investigaciones Científicas and Universidad de Sevilla, Américo Vespucio 49, E-41092 Seville, Spain
    Search for more papers by this author

E-mail; Tel. (+34) 95 448 9523; Fax (+34) 95 446 0065.


Heterocyst-forming cyanobacteria grow as filaments of cells (trichomes) in which, under nitrogen limitation, two interdependent cell types, the vegetative cells performing oxygenic photosynthesis and the nitrogen-fixing heterocysts, exchange metabolites and regulatory compounds. SepJ is a protein conspicuously located at the cell poles in the intercellular septa of the filaments that has three well-defined domains: an N-terminal coiled-coil domain, a central linker and a C-terminal permease domain. Mutants of Anabaena sp. strain PCC 7120 carrying SepJ proteins with specific deletions showed that, whereas the linker domain is dispensable, the coiled-coil domain is required for polar localization of SepJ, filament integrity, normal intercellular transfer of small fluorescent tracers and diazotrophy. An Anabaena strain carrying the SepJ protein from the filamentous, non-heterocyst-forming cyanobacterium Trichodesmium erythraeum, which lacks the linker domain, made long filaments in the presence of combined nitrogen but fragmented extensively under nitrogen deprivation and did not grow diazotrophically. In contrast, a chimera made of the Trichodesmium coiled-coil domain and the Anabaena permease allowed heterocyst differentiation and diazotrophic growth. Thus, SepJ provides filamentous cyanobacteria with a cell–cell anchoring function, but the permease domain has evolved in heterocyst formers to provide intercellular molecular exchange functions required for diazotrophy.


Multicellularity has evolved independently several times in the course of evolution (Carroll, 2001) and each independent case of multicellularity is therefore a unique biological phenomenon. Although most bacteria are unicellular, multicellular forms are also found in bacterial phyla such as Actinobacteria (Flärdh and Buttner, 2009), Cyanobacteria (Flores and Herrero, 2010) and Myxococcales (Zusman et al., 2007). An important aspect of multicellularity is that it permits the development of organisms with specialized cells. Whereas differentiated structures in Actinobacteria and Myxococcales produce spores, some multicellular cyanobacteria are capable of different developmental processes. These include the production of akinetes (a type of spores resistant to cold and desiccation), hormogonia (small motile filaments) and heterocysts (metabolically specialized cells).

Some cyanobacteria make linear chains of cells termed filaments or trichomes, which constitute their organismic unit of growth (Rippka et al., 1979; Flores and Herrero, 2010), and some filamentous cyanobacteria, such as those of the genera Anabaena or Nostoc, can produce heterocysts. In the absence of combined nitrogen, a one-dimensional pattern of single heterocysts separated by approximately 10–15 vegetative cells is established to form a multicellular organism composed of two interdependent cell types (Xu et al., 2008). Whereas vegetative cells remain undifferentiated and perform oxygenic photosynthesis, differentiation of vegetative cells into heterocysts involves morphological and metabolic changes to produce a terminal cell specialized in N2 fixation (Flores and Herrero, 2010; Kumar et al., 2010). In the mature diazotrophic trichome, an exchange of metabolites takes place that results in a net transfer of carbon and reducing equivalents from vegetative cells to heterocysts and of combined nitrogen from heterocysts to vegetative cells (Wolk et al., 1994). Additionally, the development of heterocyst-containing filaments involves the intercellular transfer of regulators such as a PatS-related compound (Yoon and Golden, 1998). Key steps in the evolution of this form of prokaryotic multicellularity must have included the development of efficient mechanisms for keeping the cells together in a filament and for intercellular communication.

The cyanobacteria carry a Gram-negative type of cell envelope, and in filamentous cyanobacteria, whereas each cell is surrounded by its cytoplasmic membrane and peptidoglycan layer, the outer membrane is continuous along the filament (Flores et al., 2006). The periplasm, which lies between the cytoplasmic and outer membranes, is therefore also continuous in these organisms and could represent a conduit for intercellular communication (Mariscal et al., 2007). The outer membrane of an Anabaena strain has been shown to be a permeability barrier for metabolites (such as glutamate and sucrose) that can be exchanged between cells, thus ensuring that metabolites important in the diazotrophic physiology, if released from the cytoplasm to the periplasm, are not lost from the trichome (Nicolaisen et al., 2009). Although the outer membrane might contribute to keep the cells together in the filament, specific cell–cell joining structures that have been termed microplasmodesmata appear to be present in the intercellular septa (Lang and Fay, 1971; Giddings and Staehelin, 1978; Flores et al., 2006).

Several genes whose inactivation results in a filament fragmentation phenotype have been identified in the model cyanobacterium Anabaena sp. strain PCC 7120 (Bauer et al., 1995; Flores et al., 2007; Nayar et al., 2007; Merino-Puerto et al., 2010). The fraC-fraD-fraE operon encodes proteins that are needed for filament integrity mainly under nitrogen deprivation, whereas sepJ (also known as fraG) encodes a protein essential for filament integrity both in the presence and in the absence of combined nitrogen, although sepJ mutants fragment most extensively under the latter conditions. As shown with GFP fusions, SepJ localizes to the cell poles in the intercellular septa (Flores et al., 2007; Mariscal and Flores, 2010), and FraC and FraD are also located at the intercellular septa, although not as focused in the cell poles as SepJ, and are needed for proper localization of SepJ (Merino-Puerto et al., 2010). Molecular intercellular exchange has been recently probed in Anabaena spp. with calcein, a fluorescent tracer that can be loaded into the cytoplasm (Mullineaux et al., 2008). Intercellular exchange of calcein is impaired in sepJ, fraC and fraD mutants, suggesting that these proteins contribute to make a channel that connects adjacent cells and allows the intercellular transfer of small molecules. Because it is essential for filament integrity and conspicuously present at the intercellular septa in the trichomes of Anabaena sp. strain PCC 7120, SepJ could be a central component of such cell–cell joining channels. In this work, we have addressed the role of SepJ through the analysis of the function of its different protein domains and the study of SepJ chimeras.


SepJ domain architecture

Analysis of the sequence of the SepJ protein from Anabaena sp. strain PCC 7120 shows that it has three well-differentiated domains: (i) a 200-residue N-terminal domain with a strongly predicted coiled-coil structure, (ii) an internal 211-residue region rich in Pro and Ser, which shows sequence similarity to plant extensins and mammalian TITIN/connectins, and (iii) a 340-residue C-terminal domain that is homologous to cytoplasmic membrane proteins of the drug/metabolite exporter (DME) family (Flores et al., 2007). Topology predictions and a GFP fusion to the C-terminus of the protein indicate that SepJ is a cytoplasmic membrane protein in which the coiled-coil and linker domains reside external to the membrane in the periplasm and the C-terminus is cytoplasmic (Flores et al., 2007; Mariscal and Flores, 2010). blast searches using as a query the Anabaena SepJ protein showed that a genuine SepJ carrying the predicted three-domain structure is restricted to the filamentous, heterocyst-forming cyanobacteria (Fig. 1). However, a similar protein carrying the coiled-coil and permease domains is found in the filamentous, non-heterocyst-forming cyanobacteria whose genomic sequence is available, whereas a protein homologous to the permease domain, and therefore to proteins in the DME family, is also encoded in the genome of many unicellular cyanobacteria. The distribution of homologous proteins among the cyanobacteria suggests that evolution of SepJ may have comprised the acquisition of new domains starting from a DME family protein: a coiled-coil domain may have first been added to the permease to produce the protein that is found in filamentous non-heterocyst-forming cyanobacteria, and the internal domain, which we will here denote as linker domain, may have been then added to this protein to produce the SepJ protein found in the heterocyst formers. Interestingly, the linker domain is the least conserved of the three domains in the SepJ protein (Fig. 1).

Figure 1.

Domain arrangement of cyanobacterial SepJ proteins. Cyanobacterial SepJ homologous proteins were obtained by homology searching in the NCBI database using SepJ from Anabaena sp. strain PCC 7120 as a query. CC denotes the coiled-coil domain; L, linker domain; and P, integral membrane domain. Length is indicated as number of amino acid residues. The SepJ protein or its homologues are from the following cyanobacteria: Anabaena sp. strain PCC 7120; Anabaena variabilis strain ATCC 29413; Nodularia spumigena strain CCY9414; Nostoc punctiforme strain ATCC 29133/PCC 73102; Nostoc azollae strain 0708; Trichodesmium erythraem strain IMS101; Lyngbya sp. strain PCC 8106; Arthrospira platensis strain NIES-39; Microcoleus chthonoplastes strain PCC 7420; Acaryochloris marina strain MBIC 11017; Synechocystis sp. strain PCC 6803; Cyanothece sp. strain ATCC 29141/PCC 7425; Synechococcus sp. strain PCC 7335.

Generation of sepJ deletion mutants

In order to investigate the requirement of the different SepJ domains for filament integrity, heterocyst differentiation and diazotrophic growth, Anabaena mutants that produce SepJ proteins lacking specific domains or polypeptide fragments were constructed by gene replacement with versions of the sepJ gene bearing in-frame deletions, which we generally denote (Δ)sepJ genes (see Experimental procedures for details). Strains producing SepJ proteins with most of the coiled-coil domain deleted (CSVM25), with two deletions of different lengths in the linker domain (CSVM57, CSVM85), with the coiled-coil and linker domains deleted (CSVM26) and with two transmembrane segments deleted (CSVM36) were prepared (see Fig. 2). Additionally, a strain lacking most of the SepJ protein was also constructed (CSVM34) and used as a control. PCR analysis confirmed that all these strains were homozygous for the mutant chromosomes (see Fig. S1 and Experimental procedures for details). Because the in-frame deletions are unmarked mutations, all these strains were regularly grown without antibiotics, and repeated PCR analysis showed that they were stable mutants. All these strains grew with combined nitrogen (nitrate or ammonium) at rates similar to those of the wild type (not shown).

Figure 2.

Phenotype of Anabaena strains carrying different (Δ)SepJ proteins, Trichodesmium SepJ or Anabaena/Trichodesmium SepJ chimeras. (Left) Diagram of the different (Δ)SepJ versions constructed in this work, Trichodesmium SepJ (in red colour) and Anabaena/Trichodesmium SepJ chimeras (Trichodesmium SepJ fragments shown in red). A dotted line denotes the fragment deleted, with the number of deleted amino acids indicated above. The SepJ fragments deleted in the different strains were: CSVM34, from amino acid residue 40 to amino acid residue 747; CSVM25, from 40 to 201; CSVM26, from 40 to 410; CSVM36, from 623 to 687; CSVM57, from 323 to 410; CSVM85, from 223 to 410. (Right) Summary of phenotypic characteristics. To quantify filament length, BG11C-grown filaments or filaments incubated for 18 h in BG110C media were observed under the light microscope. Filament length is expressed as the mean number of cells per filament (see Fig. S2 for more details). For staining heterocyst polysaccharides (Hep) with Alcian blue, filaments incubated for 36 h in the absence of combined nitrogen were used (see also Fig. S3A). Glycolipids (Hgl) were extracted from filaments after 48 h of incubation in the absence of combined nitrogen and analysed by TLC (see also Fig. S4). The presence of heterocysts (Hets) was checked by light microscopy in cultures incubated for several periods of time (22–48 h) in the absence of combined nitrogen. As a quantitative indication of heterocyst functionality, nitrogenase (acetylene reduction) activity under oxic conditions was determined in filaments after 24 h of combined nitrogen deprivation. Nitrogenase (mean and standard deviation of the mean, determined in three independent cultures) is expressed as µmol ethylene produced (mg chlorophyll a)−1 h−1. All these parameters were determined in filaments from bubbled cultures. Diazotrophic growth rate constants were determined in shaken cultures in BG110 medium; figures (µ, in day−1) are the mean and standard deviation of the mean of three or four independent determinations. A minus signal denotes absence of the indicated characteristic or no growth, a plus signal indicates its presence and (+) denotes a weak positive.

To check whether the mutant proteins were produced at appreciable levels in the different (Δ)sepJ mutants, Western analysis was performed. First, we obtained a polyclonal antibody against a polypeptide including amino acids 1–226 of SepJ, which comprises the whole coiled-coil domain (see Experimental procedures). In cell-free extracts from the wild type, this antibody detected a protein (indicated by a black arrowhead in Fig. 3A) and several possible degradation products that were mostly present in the membrane fraction. This protein was present in extracts from mutant strain CSVM57 (which lacks part of the linker domain but retains the coiled-coil domain) but not from strains CSVM25 or CSVM26 (which lack most of the coiled-coil domain) or from CSVM34 (which lacks most of the SepJ protein) identifying it as a SepJ-related protein. Only a faint band was detected in the membrane fraction from strain CSVM36, which lacks two transmembrane segments but retains the coiled-coil domain, suggesting the presence of only a low amount of (Δ)SepJ in this strain. Finally, in extracts from strain CSVM85 (which lacks the linker domain but retains the coiled-coil domain), a larger band but similar degradation products were detected. Although the predicted size of SepJ in the wild-type, CSVM57 and CSVM85 strains is 81.4, 72.4 and 61.9 kDa, respectively, the protein identified as SepJ migrated in SDS-PAGE between the 50 and 36 kDa markers (wild type and CSVM57) or between the 64 and 50 kDa markers (CSVM85). We cannot explain the difference in mobility between the (Δ)SepJ proteins from CSVM57 and CSVM85 observed in this analysis. However, the anomalous migration of SepJ or the (Δ)SepJ proteins can be generally explained by the known altered mobility of membrane proteins in SDS-PAGE (Rath et al., 2009).

Figure 3.

Detection of proteins produced by the Anabaena (Δ)sepJ and (Δ)sepJ–gfp strains.
A. Western analysis of soluble (S) and membrane (M) subcellular fractions from the indicated Anabaena strains using an anti-SepJ-CC antibody. WT, strain PCC 7120; 25, CSVM25; 26, CSVM26; 34, CSVM34; 36, CSVM36; 57, CSVM57; 85, CSVM85.
B. Western analysis of cell-free extracts from the indicated Anabaena strains using an anti-His-tag antibody. WT-H, strain PCC 7120-H producing SepJ-6His; 25-H, CSVM25-H; 26-H, CSVM26-H; 57-H, CSVM57-H; 85-H, CSVM85-H.
C. Western analysis of soluble (S) and membrane (M) subcellular fractions from the indicated Anabaena strains using an anti-SepJ-CC antibody. WT, strain PCC 7120 used as a positive control; CSAM137, derivative of strain PCC 7120 producing a full SepJ–GFP; 34, strain CSVM34 used as a negative control; 25-GFP, CSVM25–GFP; 26-GFP, CSVM26–GFP; 36-GFP, CSVM36–GFP; 57-GFP, CSVM57–GFP.
D. Direct detection of the GFP fluorescence in SDS-PAGE gels of the soluble (S) or membrane (M) fractions from the indicated Anabaena strains (nomenclature as above). The wild-type strain PCC 7120 was used as a negative control.
Black arrowheads to the left in (A) and (B) or internal to (C) and (D) point to the different putative SepJ or (Δ)SepJ proteins (6His- or GFP-tagged when indicated); white arrowheads in (D) point to soluble, free GFP; diamonds in (D) indicate possible GFP-carrying membrane-associated degradation products of SepJ–GFP or (Δ)SepJ–GFP proteins.

To test whether the deletions introduced in (Δ)sepJ mutants CSVM25 and CSVM26 affected production of the corresponding protein, strains carrying genes encoding His-tagged versions of (Δ)SepJ were prepared and named, respectively, CSVM25-H and CSVM26-H. Additionally, strains PCC 7120-H, CSVM57-H and CSVM85-H carrying the corresponding His-tagged SepJ or (Δ)SepJ proteins were prepared (Experimental procedures and Table S2). Similar levels of the different His-tagged SepJ proteins could be detected with a His-tag antibody in cell-free extracts from the different strains (Fig. 3B). In spite of the different sizes of the deleted fragments, a similar migration of these proteins in SDS-PAGE (between the 50 and 36 kDa markers) was observed, which would be consistent with migration being mainly dependent on the binding of detergent to the hydrophobic, permease domain of the protein.

Taken together the Western analyses performed indicate that the (Δ)SepJ proteins are produced at substantial levels in strains CSVM25, CSVM26, CSVM57 and CSVM85. In strain CSVM36, however, deletion of two transmembrane segments appears to result in an unstable protein that is largely lost from the cells. Finally, as expected, the control strain CSVM34 that carries a deletion of most of the sepJ gene did not produce a SepJ protein detectable with the anti-SepJ-CC antibody.

Phenotype of the sepJ deletion mutants

Insertional mutant SR2787a of sepJ shows a filament fragmentation phenotype (Flores et al., 2007). Extensive filament fragmentation was also observed, as expected, in liquid cultures of the sepJ deletion mutant CSVM34. The fragmentation was observed in nitrate-grown cultures (Fig. 2; Fig. S2A), and it increased after removal of combined N resulting in filaments with an average length of about two cells per filament (Fig. 2; Figs S2B and S3A). Among the strains carrying (Δ)SepJ proteins, a similar fragmentation phenotype was observed in strains CSVM25, CSVM26 and CSVM36, although CSVM25 fragmented less extensively in nitrate-containing media (Fig. 2; Figs S2 and S3A). In contrast, strains CSVM57 and CSVM85 formed long filaments using either nitrate or N2 as nitrogen source (Fig. 2; Figs S2 and S3A).

Heterocyst differentiation was impaired in strains CSVM34, CSVM25, CSVM26 and CSVM36, which consistently failed to grow diazotrophically (Fig. 2, Fig. S3B). In the sepJ insertional mutant SR2787a, heterocyst differentiation aborts at a stage between deposition of the heterocyst polysaccharide layer and biosynthesis of heterocyst-specific glycolipids (Flores et al., 2007). To ascertain the stage at which heterocyst differentiation aborted in the (Δ)sepJ mutants, filament suspensions of these mutants incubated in the absence of combined nitrogen were stained with Alcian blue to visualize the polysaccharide layer, and their glycolipids were extracted and analysed. Whereas the short filaments of strains CSVM34, CSVM25, CSVM26 and CSVM36 contained cells that were stained with Alcian blue, strains CSVM34, CSVM25 and CSVM26 lacked heterocyst-specific glycolipids and strain CSVM36 synthesized a small amount of them (Fig. 2; see Figs S3A and S4). In contrast, strains CSVM57 and CSVM85 could grow diazotrophically at rates similar to that of the wild type (Fig. 2; Fig. S3B). N2-grown filaments of both strains contained heterocysts that could be stained with Alcian blue, and these filaments contained heterocyst-specific glycolipids and exhibited significant nitrogenase activity under oxic conditions, although this activity in strain CSVM57 was only about 20% that in wild type (Fig. 2; see Figs S3 and S4). Nonetheless, this activity was not limiting for growth. Thus, whereas removal of only the linker domain had little or no effect on the investigated parameters, removal of the coiled-coil domain notably affected filament integrity and diazotrophy. Removal of two transmembrane segments (as in strain CSVM36) impaired filamentation and diazotrophy too but, as discussed above, accumulation of the (Δ)SepJ protein was impaired in this strain.

Intercellular molecular exchange

Intercellular molecular exchange in the Anabaena filament can be investigated using as a tracer calcein AM (Mullineaux et al., 2008) or 5-carboxyfluorescein diacetate AM (C.W. Mullineaux and collaborators, in preparation), the latter producing a compound, 5-carboxyfluorescein (5-CFDA), smaller than calcein (374 versus 622 Da) that can also be transferred between cells. Intercellular transfer of these tracers can be quantified by fluorescence recovery after photobleaching (FRAP) experiments (Mullineaux et al., 2008). Intercellular transfer of calcein and 5-CFDA was investigated in nitrate-grown filaments of the (Δ)sepJ mutants using the wild type as a control (Table 1A). Intercellular calcein transfer was hampered by 70–90% in the different mutants. However, whereas transfer of 5-CFDA was impaired by 54–68% in strains CSVM25, CSVM26 and CSVM34 and by 26% in CSVM36, it took place at high levels in CSVM57 and CSVM85. As previously shown for calcein (Mullineaux et al., 2008), intercellular transfer of the probes was higher after incubation of the filaments in the absence of combined nitrogen (compare strain PCC 7120 in Table 1A and B). Under these conditions, intercellular calcein transfer was severely hampered in filaments of strains CSVM57 and CSVM85, which however showed high levels of intercellular 5-CFDA transfer (Table 1B). Although other channel(s) or mechanism(s) might contribute to intercellular transfer of the tracers, especially of 5-CFDA, these results indicate a dependence on a functional SepJ protein for a substantial transfer of calcein.

Table 1.  Exchange coefficients for transfer of calcein and 5-CFDA between vegetative cells in filaments of Anabaena sp. strain PCC 7120 and the (Δ)sepJ mutants.
StrainCalcein E (s−1)5-CFDA E (s−1)
  1. Filaments grown in BG11C (A) or grown in the same medium and incubated in the absence of combined nitrogen (medium BG110C) for 72 h (B) were loaded with calcein or 5-CFDA, washed and subjected to FRAP analysis as described in Experimental procedures to determine the exchange coefficient E (Mullineaux et al., 2008). Figures are the mean and standard deviation of the mean of the number of experiments indicated in parenthesis.

(A) Nitrate-grown filaments
PCC 71200.0301 ± 0.0011 (n = 22)0.0204 ± 0.0005 (n = 12)
CSVM340.0046 ± 0.0007 (n = 11)0.0093 ± 0.0006 (n = 14)
CSVM250.0093 ± 0.0007 (n = 20)0.0065 ± 0.0004 (n = 14)
CSVM260.0051 ± 0.0005 (n = 15)0.0083 ± 0.0006 (n = 15)
CSVM360.0090 ± 0.0006 (n = 16)0.0151 ± 0.0011 (n = 17)
CSVM570.0033 ± 0.0004 (n = 19)0.0251 ± 0.0014 (n = 12)
CSVM850.0026 ± 0.0005 (n = 13)0.0296 ± 0.0012 (n = 19)
(B) N2-induced filaments
PCC 71200.0408 ± 0.0030 (n = 8)0.0590 ± 0.0017 (n = 16)
CSVM570.0090 ± 0.0023 (n = 8)0.0650 ± 0.0049 (n = 11)
CSVM850.0000 ± 0.0000 (n = 9)0.0608 ± 0.0024 (n = 12)

Subcellular localization of (Δ)SepJ–GFP fusions

In order to study the requirement of each of the SepJ domains for SepJ localization, translational fusions to the GFP of each of the SepJ truncated versions were prepared. In all cases, the GFP was fused at a position nine amino acids from the C-terminus, which is predicted to position the GFP adjacent to the cytoplasmic face of the cytoplasmic membrane (Flores et al., 2007). The generated strains were homozygous for the chromosomes carrying the (Δ)sepJ–gfp gene (see Experimental procedures for details). Whereas strains CSVM57–GFP and CSVM85–GFP could grow diazotrophically, strains CSVM25–GFP, CSVM26–GFP and CSVM36–GFP could not (not shown), indicating that, as has been previously reported for wild-type SepJ (Flores et al., 2007), fusion to the GFP does not affect the diazotrophic function of SepJ.

Figure 4 shows the subcellular localization of the different (Δ)SepJ–GFP proteins using a fusion of the GFP to a wild-type SepJ (carried in strain CSAM137) as a control. The GFP fluorescence was hardly visible in strains CSVM25–GFP and CSVM26–GFP, although in the latter some random spots of fluorescence were observed in the surface of the cells, suggesting that (Δ)SepJ–GFP was not localized at the cell poles in the intercellular septa in these strains. In strains CSVM57–GFP and CSVM85–GFP, the GFP fluorescence was observed in the septa, more focused in the case of CSVM85–GFP. These results indicate that the coiled-coil domain, but not the linker domain, is necessary for localization of SepJ at the cell poles. On the other hand, in strain CSVM36–GFP, the GFP fluorescence was found inside the cells, suggesting that the deletion of two transmembrane segments impeded the insertion of SepJ into the cytoplasmic membrane.

Figure 4.

Subcellular location of SepJ–GFP, the (Δ)SepJ–GFP proteins and the Trichodesmium coiled-coil/Anabaena permease SepJ–GFP chimera. Filaments from the indicated strains were taken from cultures grown for 72 h in bubbled BG11C medium and analysed by confocal microscopy. Figures are an overlay of red autofluorescence and green GFP fluorescence. 85-GFP, strain CSVM85-GFP; 117-GFP, CSVM117-GFP; all other strains as in the legend to Fig. 3.

To corroborate the results obtained by microscopy, cell-free extracts were prepared from the different strains, and the soluble and membrane-containing fractions were obtained and analysed by SDS-PAGE. The anti-SepJ-CC antibody detected appreciable levels of a reacting protein (migrating between the 50 and 36 kDa markers) in the membrane fractions from strains CSAM137 and CSVM57–GFP (strain CSVM85–GFP was not analysed) and a faint band in the membrane fraction from strain CSVM36–GFP (Fig. 3C). The levels in strains CSAM137 and CSVM57–GFP were similar to those found in the wild type, strain PCC 7120 (Fig. 3C), indicating that fusion to the GFP does not alter SepJ expression or accumulation.

Because the GFP fluorescence can be observed directly from gels (Drew et al., 2006), the subcellular localization was also analysed by detection of GFP fluorescence from the SDS-PAGE gels (Fig. 3D). In these gels, a background of fluorescent bands that should correspond to photosynthetic pigments is observed in the wild-type strain PCC 7120. Additionally, as observed in extracts of the control strain CSAM137, fluorescent bands that should correspond to free GFP present in the soluble fraction (about 27 kDa in size; white arrowheads) and two membrane-associated proteins, one migrating between the 50 and 36 kDa markers (black arrowheads) and another migrating between the 36 and 22 kDa markers (diamonds), can be detected. The membrane-associated proteins could correspond to SepJ–GFP and a GFP-carrying membrane-associated degradation product respectively. Whereas only the soluble GFP was observed in strain CSVM36–GFP, in strains CSVM25–GFP, CSVM26–GFP and CSVM57–GFP both the soluble GFP and the two putative membrane-associated (Δ)SepJ–GFP proteins were observed (strain CSVM85 was not analysed). Whereas these results indicate a low stability of the SepJ–GFP fusion always producing some free GFP, they corroborate the presence of the (Δ)SepJ–GFP protein in the membrane fractions of strains CSVM25–GFP, CSVM26–GFP and CSVM57–GFP [the reason for largest (Δ)SepJ–GFP appearing as a doublet in strain CSVM57–GFP is unknown]. The (Δ)SepJ–GFP protein in strain CSVM36–GFP, however, is missing from the membrane fraction, suggesting again that it is processed resulting in accumulation of free GFP in the cytoplasm, as observed by confocal microscopy (Fig. 4).

Anabaena strains carrying Trichodesmium SepJ and Anabaena/Trichodesmium SepJ chimeras

To further investigate the features of SepJ, we asked whether a two-domain SepJ from a filamentous non-heterocyst-forming cyanobacterium could provide SepJ function to Anabaena. The sepJ gene from Trichodesmium erythraeum was cloned downstream of the sepJ promoter in Anabaena, producing strain CSVM114 that was homozygous for chromosomes carrying Trichodesmium sepJ (see Experimental procedures and Fig. S5 for details). Strain CSVM114 showed long filaments in nitrate-containing medium (Fig. 5). However, upon incubation in the absence of combined N, this strain fragmented extensively and was unable to grow diazotrophically (Fig. 5). After 24 h of incubation without combined nitrogen, these fragmented filaments showed no detectable nitrogenase activity (see summary of phenotypic characteristics in Fig. 2). Although these cultures contained cells that were stained with Alcian blue (Fig. S6A), heterocyst-specific glycolipids were not produced (Fig. S6B) indicating that heterocyst differentiation was aborted between the steps of polysaccharide and Hgl biosynthesis, as is the case for sepJ mutants. These results show that Trichodesmium SepJ provides Anabaena with the capability to make long filaments when growing with combined nitrogen but not with the SepJ function(s) necessary for completing heterocyst differentiation and diazotrophic growth.

Figure 5.

Filaments of Anabaena strains carrying Trichodesmium erythraeum SepJ or the Anabaena/Trichodesmium SepJ chimeras. Light micrographs of strains CSVM114 (carrying Trichodesmium SepJ), CSVM116 (carrying the Anabaena coiled-coil and linker/Trichodesmium permease SepJ chimera) and CSVM117 (carrying the Trichodesmium coiled-coil/Anabaena permease SepJ chimera) grown in bubbled BG11C medium (NO3-) or incubated for 24 h in BG110C medium (N2). Original magnification, ×400.

To investigate whether one particular domain was required for diazotrophic SepJ function, Anabaena/Trichodesmium chimeras were constructed. Anabaena sp. strain CSVM116 carries a chimeric sepJ gene producing a SepJ protein composed of amino acid residues 1–412 (the coiled-coil and linker domains) from Anabaena SepJ and 256–583 (the permease domain) from Trichodesmium SepJ. Conversely, Anabaena sp. strain CSVM117 produces a SepJ protein composed of amino acid residues 1–255 (the coiled-coil domain) from Trichodesmium SepJ and 413–751 (the permease domain) from Anabaena SepJ. Both strains were homozygous for the chimeric sepJ genes and lacked wild-type sepJ (Fig. S5). Both strains made long filaments in nitrate-containing medium (Fig. 5). However, upon incubation in medium without combined nitrogen, whereas strain CSVM116 fragmented extensively and failed to develop nitrogenase activity and diazotrophic function, strain CSVM117 made short filaments that contained heterocysts, showed a low but consistent nitrogenase activity and grew diazotrophically at a rate about 50% that of the wild type (Figs. 2 and 5). Consistently with these observations and in line with what was observed with other sepJ mutants, both strains produced heterocyst polysaccharides (positive Alcian blue staining in some cells) but only CSVM117 produced heterocyst-specific glycolipids (Fig. S6). These results show that the Anabaena SepJ permease domain has a specific role in diazotrophy.

An Anabaena strain carrying the GFP fused to the Trichodesmium coiled-coil/Anabaena permease SepJ chimera (strain CSVM117–GFP; the GFP was fused at a position nine amino acids from the C-terminus of the SepJ chimera) was constructed. In this strain, the GFP fluorescence was observed at the cell poles in the intercellular septa (Fig. 4), which is consistent with SepJ function being provided by the SepJ chimera in strain CSVM117. We failed, however, in preparing similar GFP fusions with the SepJ proteins that carried the Trichodesmium permease, which might indicate some differences in the topology of the Trichodesmium protein as compared with the Anabaena protein.


The SepJ protein from heterocyst-forming cyanobacteria bears three well-defined domains, namely, an N-terminal coiled-coil domain, a linker domain and a C-terminal permease domain. Anabaena sp. strain CSVM25, which produces a (Δ)SepJ protein that lacks the coiled-coil domain, fragments extensively upon nitrogen deprivation and does not develop diazotrophic functions. As shown by fractionation of cell-free extracts of strain CSVM25–GFP, the (Δ)SepJ–GFP protein lacking the coiled-coil domain could be incorporated into the membrane fraction of the cells (Fig. 3D), but no GFP fluorescence could be observed focused at the cell poles in this strain (Fig. 4) indicating delocalization of the protein. Thus, the coiled-coil domain is essential for polar localization of SepJ. Strains that lack the coiled-coil and linker domains (CSVM26) or most of the SepJ protein (CSVM34) show similar phenotypic traits as CSVM25, indicating that deletion of the coiled-coil domain is sufficient to produce a non-functional SepJ. Coiled-coil domains are implicated in protein–protein interactions (Lupas and Gruber, 2005), and therefore the SepJ coiled-coiled domain could have a role joining adjacent cells in the Anabaena filament. Interaction of the coiled-coil domains of SepJ proteins from adjacent cells could in turn contribute to keep SepJ focused at the cell poles. Because heterocyst differentiation is hampered in strain CSVM25, proper localization of SepJ appears also needed for the intercellular interactions that support the development of a diazotrophic filament.

With regard to the SepJ linker domain, deletion of a minor (as in strain CSVM57) or a major (as in strain CSVM85) part of this domain results in a slight or no fragmentation of the filaments and does not significantly impair SepJ–GFP localization, heterocyst differentiation or diazotrophy. Thus, consistent with its low conservation as shown in Fig. 1, the linker domain seems to be dispensable. Finally, in strain CSVM36, which produces a (Δ)SepJ protein that lacks two transmembrane segments of the permease domain, only a small amount of membrane-bound SepJ protein is detected (Fig. 3A), as is also the case in strain CSVM36-GFP (Fig. 3C). On the other hand, in the latter strain a substantial amount of soluble GFP is found (Fig. 3D) that apparently fills the cytoplasm as observed by confocal microscopy (Fig. 4). Therefore, in strain CSVM36–GFP the (Δ)SepJ–GFP protein that is not inserted into the cytoplasmic membrane appears to be degraded releasing a soluble GFP. Nonetheless, strain CSVM36 produces a low level of heterocyst glycolipids, which could be related to a small amount of (Δ)SepJ inserted into the membrane as discussed above.

SepJ appears to contribute to an intercellular channel that has been probed with the fluorescent tracer calcein (Mullineaux et al., 2008; Merino-Puerto et al., 2010). In this work we have used, in addition to calcein, a smaller but otherwise similar tracer, 5-CFDA. Mutants lacking the coiled-coil domain (CSVM25, CSVM26, CSVM34) are affected in the transfer of both compounds. This impairment could result from lack of proper SepJ localization or from a need of the coiled-coil domain as a component of the channel. In contrast, linker domain mutants (CSVM57, CSVM85) exhibit a low transfer of calcein (622 Da) but a high transfer of 5-CFDA (374 Da). It is possible that, although not required to make a SepJ channel, the linker domain influences the effective size of the channel, which allows a high transfer of calcein in the presence of the linker but not in its absence. On the other hand, a channel independent of SepJ might contribute essentially to transfer of 5-CFDA. Because CSVM57 and CSVM85 grow well with N2, the transfer of molecules smaller than calcein appears to be sufficient for heterocyst differentiation and diazotrophic function. Prior to this work, the upper size limit for these possible channels was set by work with the GFP (27 kDa), which is not transferred between the cytoplasm of adjacent cells (Yoon and Golden, 1998; Mariscal et al., 2007).

Because some unicellular cyanobacteria carry genes encoding permeases homologous to the SepJ permease domain, it is possible that such a permease, as a cytoplasmic membrane protein, was evolutionarily recruited by addition of a coiled-coil domain to anchor the cells in the filament. The resulting protein would correspond to the SepJ found in the filamentous, non-heterocyst-forming cyanobacteria. SepJ from T. erythraeum provides Anabaena with the capability to make filaments in the presence of combined nitrogen, which is consistent with an anchoring role of Trichodesmium SepJ. Filament integrity under nitrogen deprivation and heterocyst differentiation appear however to require additional specializations met by SepJ from the heterocyst formers. No Anabaena strain carrying the Trichodesmium SepJ protein or fragments thereof makes long filaments in the absence of combined nitrogen. It is possible that interactions of SepJ with other septum-localized proteins such as FraC and FraD, which have a role in filamentation mainly under combined nitrogen deprivation and influence SepJ localization (Merino-Puerto et al., 2010), are needed to make long filaments. The Trichodesmium or chimeric SepJ proteins might fail in these interactions with Anabaena proteins. Remarkably, however, whereas a chimeric SepJ composed of the Anabaena coiled-coil and linker domains and the Trichodesmium permease does not allow heterocyst differentiation and diazotrophic growth, a chimeric SepJ composed of the Trichodesmium coiled-coil domain and the Anabaena permease shows heterocyst differentiation and diazotrophic growth. These results indicate that the permease domain of SepJ from heterocyst formers has an important role in diazotrophic function.

To summarize, the dissection of Anabaena SepJ shows a strong correlation between proper localization of SepJ at the cell poles in the intercellular septa, filament integrity, intercellular molecular exchange probed with fluorescent tracers and diazotrophic functions (including heterocyst differentiation, nitrogenase activity and the capability to grow with N2). SepJ appears to contribute not only to maintain the cells together in the filament but also to the making of a channel for intercellular communication. Periplasmic coiled-coil and cytoplasmic-membrane permease domains play distinct roles making SepJ a unique protein essential for multicellularity in heterocyst-forming cyanobacteria.

Experimental procedures

Bacterial strains and growth conditions

Anabaena sp. strains were grown in BG11 (containing NaNO3) or BG110 (lacking nitrate) medium at 30°C in the light (25 µE m−2 s−1 from fluorescent lamps), in shaken (100 r.p.m.) liquid cultures or in medium solidified with 1% Difco agar. Alternatively, the medium was supplemented with 10 mM of NaHCO3 and termed BG11C or BG110C, and some cultures (denoted bubbled cultures) were also supplemented with a mixture of CO2 and air (1% v/v) and illuminated with light from fluorescent lamps (75 mE m−2 s−1). Antibiotics were used at the following concentration: Sm 2 µg ml−1, Sp 2 µg ml−1 for liquid cultures; and Sm 5 µg ml−1, Sp 5 µg ml−1 and Nm 150 µg ml−1 for solid cultures.

Escherichia coli DH5α was used for plasmid constructions. It and strains HB101 and ED8654, used for conjugations with Anabaena sp., were grown in LB medium, supplemented when appropriate with antibiotics at standard concentrations (Ausubel et al., 2010).

Plasmid construction and genetic procedures

To introduce in-frame deletions into sepJ (ORF alr2338 of the genome of Anabaena sp. strain PCC 7120; Kaneko et al., 2001), fragments upstream and downstream of the desired deletion were amplified by PCR using as template DNA from Anabaena sp. strain PCC 7120 and oligodeoxynucleotide primers listed in Table S1. DNA was isolated from Anabaena sp. by the method of Cai and Wolk (1990). The upstream and downstream DNA fragments were cloned together in pMBL-T (Dominion MBL, Spain) producing the plasmids listed in Table S2, and their sequences were confirmed by sequencing. To clone the sepJ gene from T. erythraeum, the open reading frame was amplified by PCR using as template genomic DNA from this cyanobacterium. The product was fused between the upstream and downstream sequences flanking Anabaena sepJ and cloned in pMBL-T producing the plasmids listed in Table S2, whose sequences were confirmed by sequencing. To prepare Anabaena/Trichodesmium SepJ chimeras, fragments from the sepJ genes from each cyanobacterium were cloned together as described in Table S2.

Cloned DNA fragments were transferred to pRL278 or pCSRO, which can be mobilized by conjugation and carry the sacB gene for positive selection and antibiotic resistance determinants (see Table S2; Black et al., 1993). The produced plasmids, pCSVM25, pCSVM26, pCSVM34, pCSVM36, pCSVM57, pCSVM85, pCSVM114, pCSVM116 and pCSVM117, were transferred by conjugation into Anabaena sp. strain PCC 7120 (Elhai et al., 1997). Exconjugants that had incorporated the transferred plasmid by single recombination were selected as clones resistant to the corresponding antibiotic. Cultures of these exconjugants were used to select for clones resistant to 5% sucrose (Cai and Wolk, 1990). Individual SucR colonies were checked by PCR looking for clones that had substituted the wild-type sepJ gene by a modified sepJ.

To fuse the gfp-mut2 gene to the 3′-terminal region of the deleted sepJ versions, plasmid pCSAM137 (Flores et al., 2007) was transferred by conjugation into strains CSVM25, CSVM26, CSVM57, CSVM85 and CSVM117. To fuse the gfp-mut2 gene to the 3′-terminal region of the sepJ gene encoding a SepJ protein lacking two transmembrane segments, a DNA fragment of 690 bp of the sepJ 3′-terminal region was PCR-amplified and cloned into pMBL-T, producing pCSVM37 (Table S2). The cloned DNA fragment was then transferred to pCSEL24, which contains the gfp-mut2 gene expressed from the ntcA gene promoter (Olmedo-Verd et al., 2006), yielding pCSVM38. Finally, the translational fusion was transferred to pCSV3 producing pCSVM39, which was transferred to Anabaena strain CSVM36 by conjugation. Exconjugants that had incorporated the plasmid by single recombination were selected as clones resistant to Sm and Sp.

To fuse the 6-His tag to the C-end of SepJ, a DNA fragment of the 3′-terminal region of sepJ was PCR-amplified from Anabaena sp. strain PCC 7120 and cloned into pMBL-T producing plasmid pCSVM101 (Table S2). The amplified DNA fragment was cloned into pCSV3 yielding pCSVM102, which was transferred by conjugation to Anabaena sp. strains PCC 7120, CSVM25, CSVM26, CSVM57 and CSVM85 respectively. Exconjugants that had incorporated the plasmid by single recombination were selected as clones resistant to Sm and Sp.

Growth tests and phenotypic analysis

To calculate growth rates, protein concentration was determined by a modified Lowry procedure (Markwell et al., 1978) in 0.2 ml of samples from shaken cultures. The growth rate constant, µ, corresponds to ln2/td, where td is the doubling time. Under our culture conditions, cultures grow exponentially between cell densities corresponding to about 5–100 µg of protein (0.2–4 µg of chlorophyll a) per ml. Chlorophyll a content of cultures was determined by the method of Mackinney (1941).

Filament length, Alcian blue staining, heterocyst glycolipids and nitrogenase activities were determined in bubbled BG11C or BG110C cultures, as indicated. To determine filament length, samples taken with great care to prevent disruption were visualized by standard light microscopy. For staining with Alcian blue, a filament suspension was mixed 1:1 with a 1% Alcian blue (Sigma) preparation in water. For detection of heterocyst glycolipids, lipids were extracted with chloroform–methanol (2:1 v/v), concentrated under N2 and analysed by thin-layer chromatography (Merino-Puerto et al., 2010). Nitrogenase activity was determined under oxic conditions by the acetylene reduction technique as described (Montesinos et al., 1995).

Confocal microscopy and calcein and 5-CFDA labelling

For confocal microscopy, samples from cultures of Anabaena sp. set atop solidified medium (BG11 or BG110) were visualized using a Leica HCX PLAN-APO 63× 1.4 NA oil immersion objective attached to a Leica TCS SP2 confocal laser-scanning microscope. GFP was excited using 488 nm irradiation from an argon ion laser. Fluorescent emission was monitored by collection across windows of 500–540 nm (GFP imaging) and 630–700 nm (cyanobacterial autofluorescence).

Calcein staining and FRAP analysis were performed as previously reported (Mullineaux et al., 2008). For 5-CFDA staining, 1 ml of cell cultures were harvested, washed and resuspended in 1 ml of fresh growth medium, and then mixed with 4 µl of 5-carboxyfluorescein diacetate AM (5 mg ml−1 in dimethylsulphoxide). The suspension was incubated in the dark at 30°C for 60 min in a shaker (100 r.p.m.), and filaments were then harvested and washed three times in fresh, dye-free growth medium. The suspension was then incubated in the same conditions for a further 60 min before imaging. Filament suspensions were spotted onto agar and placed in a custom-built temperature-controlled sample holder with a glass coverslip on top. All measurements were carried out at 30°C.

For both calcein and 5-CFDA, cells were imaged with a laser-scanning confocal microscope as previously described for calcein (Mullineaux et al., 2008) with a 488 nm line argon laser as the excitation source. Fluorescent emission was monitored by collection across windows of 500–520 nm or 500–527 nm in different experiments and a 50 µm pinhole. After an initial image was recorded, the bleach was carried out by switching the microscope to X-scanning mode, increasing the laser intensity by a factor of 10 and scanning a line across one cell for 1–2 s. The laser intensity was then reduced, the microscope was switched back to XY-imaging mode and a sequence of images recorded typically at 3 s intervals.

FRAP data analysis was made using Image Pro Plus 6.2 software (Media Cybernetic) and modelling by fitting the predicted fluorescence recovery of the bleached cell to the experimental recovery curve by adjusting the time axis to obtain an estimate of the exchange coefficient, E (Mullineaux et al., 2008).

Expression and purification of SepJ coiled-coil domain and production of antibodies

A soluble form of the coiled-coil domain of SepJ (SepJ-CC) was produced and purified as follows. DNA encoding a polypeptide covering amino acids 1–226 from SepJ was PCR-amplified and cloned into expression vector pET28-b, producing a C-terminal fusion to a 6-His tag in plasmid pCSVM98 (Table S2). One litre of LB medium containing 50 µg of kanamycin ml−1 was inoculated with 30 ml of an overnight culture of E. coli BL21 (pCSVM98) and incubated at 37°C to reach an OD600 of 0.6, and 1 mM IPTG was added. After 3 h at 37°C, the cells were collected and resuspended at 5 ml per gram of cells in a buffer containing 50 mM Tris-HCl (pH 8.0), 200 mM NaCl and 10% glycerol. About 2 mg of DNase and 1 mM PMSF were added just before breakage of the cells by passage twice through the French pressure cell at 20 000 p.s.i. After centrifugation at 15 000 g, (10 min, 4°C) the SepJ-CC-6His protein was purified from the supernatant by two chromatographic steps through HisTrapTM HP columns from GE Healthcare, using an imidazole ingredient to elute the retained proteins. The purified protein was concentrated using an AMICON Ultra filter device (Millipore) and subjected to SDS-PAGE. The SepJ-CC protein band was excised from the gel, electro-eluted and concentrated. An amount of 1.4 mg of the purified protein was used to raise antibodies in the ‘Centro de Producción y Experimentación Animal’, Universidad de Sevilla (Seville, Spain), and antiserum was recovered 90 days after the first subcutaneous injection of a rabbit. The antibodies were purified using Aminolink Plus Immobilization kit (Thermo Scientific) following the instructions from the supplier.

Anabaena cell-free extracts

For obtaining cell-free extracts, filaments from 500 ml of a culture grown in BG11C medium were harvested, washed with buffer A (20 mM HEPES-NaOH, pH 7.6, 10 mM MgCl2, 5 mM CaCl2, 20% glycerol) and resuspended in 10 ml of the same buffer. Filaments were broken by two passages through a French pressure cell at 9000 p.s.i. Cell debris was discarded after centrifugation at 32 000 g (10 min, 4°C), and the supernatant was centrifuged at 120 000 g (40 min, 4°C) in order to separate soluble material (blue supernatant) from membrane fraction (green precipitate). Membranes were resuspended in 500 µl of buffer A and frozen at −20°C for subsequent analysis. Samples were treated with 1% (v/v) β-dodecyl maltoside at 4°C and insoluble material was eliminated by centrifugation at 16 000 g (10 min, 4°C). Protein was quantified using the Dc Protein Assay (Bio-Rad).

Immunoblotting and GFP visualization

For immunoblotting, samples were mixed with 0.1 volume of 5× sample buffer, incubated at 65°C for 5 min, loaded and run in a 10–14% Laemmli SDS-PAGE system, and transferred to PVDF membrane filters. For detection of SepJ-CC, the filters were incubated overnight in blocking buffer containing 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5% non-fat milk powder and 0.025% Nonidet P-40. Primary antibody (anti-SepJ-CC) was added at a dilution 1:500 and incubated for 3 h at 30°C. The secondary antibody was anti-rabbit IgG conjugated to peroxidase (Sigma), and it was used at a dilution 1:40 000 in 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.05% Tween-20 and 0.8% non-fat milk powder. For detection of the 6-His tag, anti-His HRP-conjugated antibodies (Quiagen) were used and Western analysis was performed following the instructions of the supplier. Detection was performed with a chemiluminiscence kit (ECL plus, GE Healthcare) and exposure to hyperfilm (GE Healthcare).

For direct GFP visualization, samples were mixed with 0.1 volume of 5× sample buffer, incubated at 37°C for 5 min, and loaded and run in a 10–14% Laemmli SDS-PAGE system. GFP fluorescence was directly detected from gels in an Ettan DIGE Imager (GE Healthcare) using a SYPRO Ruby filter (excitation 488/30, emission 595/25).


We thank Gustaf Sandh and Birgitta Bergman for a sample of T. erythraeum genomic DNA. Use of DNA sequences from the National Center for Biotechnology Information (USA), DOE-Joint Genome Institute (USA) and Kazusa DNA Research Institute (Japan) databases is acknowledged. Work in Seville was supported by Grants BFU2008-03811 from Ministerio de Ciencia y Tecnología, co-financed by FEDER, and CVI1896 from Junta de Andalucía (Spain). The project also used equipment purchased with grants to C.W.M. from the Wellcome Trust and Biotechnology and Biological Sciences Research Council (UK).