The gene devA of the filamentous heterocyst-forming cyanobacterium Anabaena sp. strain PCC 7120 encodes a protein with high similarity to ATP-binding cassettes of ABC transporters. Mutant M7 defective in the devA gene is arrested in the development of heterocysts at an early stage and is not able to fix N2 under aerobic conditions. The devA gene is differentially expressed in heterocysts. To gain a better understanding of the structural components of this putative ABC transporter, we determined the complete nucleotide sequence of the entire gene cluster. The two additional genes, named devB and devC, encode proteins with similarities to membrane fusion proteins (DevB) of several ABC exporters and to membrane-spanning proteins (DevC) of ABC transporters in general. Site-directed mutations in each of the three genes resulted in identical phenotypes. Heterocyst-specific glycolipids forming the laminated layer of the envelope were identified in lipid extracts of M7 and in the site-directed mutants. However, transmission electron microscopy revealed unequivocally that the glycolipid layer is missing in mutant M7. Ultrastructural analysis also confirmed a developmental block at an early stage of differentiation. The results of this study suggest that the devBCA operon encodes an exporter of glycolipids or of an enzyme that is necessary for the formation of the laminated layer. The hypothesis is proposed that an intact envelope could be required for further heterocyst differentiation.
The filamentous cyanobacterium Anabaena spp. protects the extremely oxygen-sensitive nitrogenase by spatially separating N2 fixation and oxygenic photosynthesis in two different cell types, the oxygen-evolving vegetative cell and the N2-fixing heterocyst. Upon deprivation of inorganically bound nitrogen, about 5–10% of the vegetative cells differentiate into heterocysts, resulting in a semi-regular spacing of these specialized cells along the trichome. Upon differentiation, morphological as well as biochemical changes in the developing heterocysts lead to the establishment of a microaerobic environment tolerated by nitrogenase (reviewed in Haselkorn, 1978; Wolk, 1982; Fay, 1992; Gallon, 1992). A thick envelope is formed outside of the Gram-negative cell wall to reduce the diffusion of gases into the heterocyst (Walsby, 1985; Murry and Wolk, 1989). The inner laminated layer is composed of heterocyst-specific glycolipids, which are derivatives of hexoses containing long-chain polyhydroxylalcohols (Bryce et al., 1972). The outer homogeneous and probably the outermost fibrous layer, too, are built of specific polysaccharides (Cardemil and Wolk, 1979). To minimize the diffusion of O2 into heterocysts from adjacent vegetative cells, the junctions between neighbouring cells are reduced to a narrow septum, which may be traversed by microplasmodesmata (Lang and Fay, 1971). Massive deposition of envelope material in this region leads to the formation of a pore channel. Close to the poles, the nitrogen storage molecule, cyanophycin, forms the so-called polar granule (Lang et al., 1972; Jäger et al., 1997). In addition to the cytoplasmic membranes and thylakoids, a third type of membrane is found close to the polar granule, forming the so-called honeycomb region (Lang and Fay, 1971; Braun-Howland et al., 1988). In this area, light-independent oxidation of diaminobenzidine (DAB) can be observed, indicating high concentrations of respiratory enzymes (Murry et al., 1981). Photosynthetic O2 production and CO2 fixation are restricted to vegetative cells. Therefore, heterocysts depend on these cells for a supply of reductant, probably delivered in the form of disaccharides (Schilling and Ehrnsperger, 1985). In return, heterocysts provide the vegetative cells with fixed nitrogen, presumably as glutamine (Thomas et al., 1977).
Only a few genes have been identified that are required for the formation of the heterocyst envelope. The hepA (former hetA) gene functions in the stabilization or synthesis of the polysaccharide layer (Holland and Wolk, 1990; Leganés, 1994). Black et al. (1995) described mutant strain 543 of Anabaena 7120, in which the hglK gene had been inactivated. HglK might be involved in the formation of the glycolipid layer because, in the hglK mutant, heterocyst-specific glycolipids can be found in cell extracts, but the laminated layer is not formed. A gene cluster consisting of hetM (or hglB ), hglC and hglD was identified that is required for heterocyst glycolipid synthesis as shown by mutational analysis (Black and Wolk, 1994; Bauer et al., 1997; I. Maldener unpublished).
Transposon mutagenesis of Anabaena 7120 provides a useful tool for exploring the genes that are involved in heterocyst differentiation (Borthakur and Haselkorn, 1989; Wolk et al., 1991). Using this technique, many mutants which are unable to fix N2 under aerobic conditions, owing to incompletely developed heterocysts, were isolated (Ernst et al., 1992). One of these mutants, M7, is arrested relatively early in the differentiation of heterocysts. The mutant is able to fix N2 (Fix+) under anaerobic but not under aerobic (Fox−) growth conditions. This defect is caused by an aberrant heterocyst envelope (Hen−) and an arrest in protoplast maturation, which results in a lack of heterocyst-specific oxidation of DAB (Dab−) (Ernst et al., 1992). Maldener et al. (1994) showed that the pleiotropic phenotype of mutant M7 was caused by the transposition of Tn5-1063 into an open reading frame (ORF) named devA. Expression studies, using luxAB as a reporter, showed that devA expression increases approximately eightfold in whole filaments about 14 h after nitrogen stepdown; the increase in differentiating cells was even greater. The deduced amino acid sequence of DevA shows striking similarity to the ATP-binding subunit of ABC (ATP-binding cassette) transporters (Higgins et al., 1990). Transport systems of this family facilitate ATP-dependent translocation of a great variety of substrates and are common in bacteria and in eukaryotes. Prokaryotic ABC transporters often comprise several subunits encoded by genes that are organized in an operon (Ames, 1986). The subunits include a periplasmic binding protein in the case of importers or, in the case of several exporters, a membrane fusion protein (MFP) working as a homodimer, connecting the outer and cytoplasmic membranes of Gram-negative bacteria. In addition to an ATP-binding subunit, in most cases working as a homodimer, both types of transporters possess one or two membrane proteins that traverse the cytoplasmic membrane as a hetero- or homodimer (Ames, 1986; Dinh et al., 1994).
Speculating that the devA gene is one component of an operon encoding the complete transporter, the DNA flanking devA was analysed. A gene cluster of three genes, devB, devC and devA, was found possibly encoding an ABC exporter. Anabaena 7120 strains were constructed bearing an insertionally inactivated devA, devB or devC gene respectively. The genotypical and phenotypical characterization of these mutants and the results of ultrastructural analysis of mutant M7 by transmission electron microscopy are presented.
The devBCA cluster
Figure 1 shows the map of the devBCA gene cluster. Two genes which are located upstream of devA have the same orientation and form a gene cluster of 3540 bp including devA. The nucleotide sequences are available from the GenBank-EMBL database under accession number X99672. The first ORF upstream of devA is separated by 180 bp of a non-coding stretch. This ORF was named devC and comprises 1155 bp (it corresponds to orfA in Maldener et al., 1994). The devC gene could encode a protein of 384 amino acids with a molecular mass of 43.4 kDa. Upstream of devC, spaced by a non-coding region of 41 bp, the first ORF of the gene cluster with 1425 bp, devB, was identified. The devB gene may encode a protein of 474 amino acids with a molecular mass of 51.6 kDa. The DNA sequence 1000 bp further upstream from devB does not reveal any ORF of significant size. A stem–loop structure, which may be part of a rho-independent signal for the termination of transcription, was identified 33 bp downstream from the devA stop codon.
Besides a high similarity to the ATP-binding protein of importers, the sequence of DevA further shows extensive similarities to the ATP-binding subunits of several exporters (36% identity with the C-terminal half of HlyB, a haemolysin exporter of Escherichia coli ).
The amino acid sequence comparison program BLASTP found membrane fusion proteins (MFPs) with similarity to the N-terminal part of the sequence of DevB (Fig. 2A). Dinh et al. (1994) made a multiple alignment of various MFP sequences, which revealed a consensus sequence in the C-terminal part of these proteins (Fig. 2B). Table 1 summarizes the overall sequence similarity of DevB to prokaryotic MFPs.
Table 1. . Comparison of the amino acid sequences of DevB and various membrane fusion proteins.
The hydropathy profile of DevB fits the profile of typical MFPs (data not shown). DevB contains a hydrophilic region at the N-terminus (amino acids 1–20), followed by a highly hydrophobic part of about 20 residues that might be responsible for anchoring the MFP in the cytoplasmic membrane. Adjacent to the transmembrane stretch is a region of moderate hydrophobicity, followed by a strikingly hydrophilic part that possibly traverses the periplasmic space. The C-terminus consists of a slightly hydrophobic β-strand, typical of outer membrane-associated domains (Dinh et al., 1994). These data suggest that DevB is the membrane fusion protein of an ABC export system.
The hydropathy profile of the deduced amino acid sequence of DevC shows five significantly hydrophobic stretches of about 20 amino acids, which may form transmembrane helices (data not shown); a similar profile with five helices was obtained with a different transmembrane prediction program (TMPP; Hofmann and Stoffel, 1992). DevC might be the integral membrane component of a putative ABC protein-mediated exporter.
Mutation of the dev gene cluster
Mutant M7 was reconstructed by directed mutagenesis of the devA gene in the wild type of Anabaena 7120 using two different approaches. The first approach used the recovered transposon and flanking Anabaena DNA (plasmid pRL1340) to create mutant DR238 (Maldener et al., 1994). In the second approach, the devA gene cloned from a library of wild-type DNA as an SspI–DraI fragment in pIM13 (Maldener et al., 1994) was mutagenized directly by the insertion of the kanamycin resistance cassette C.K3 (Elhai and Wolk, 1988a) into the unique NheI site (Fig. 1). After cloning into a suicide plasmid (pRL271; Black et al., 1993), the interrupted gene (plasmid pIM22) was transferred to the wild type of Anabaena 7120 via conjugation; double recombinants (DR22) were obtained using sacB as positive selection marker (Cai and Wolk, 1990). Five randomly chosen recombinants showed the same phenotype as mutant strains M7 and DR238 with the characteristics Fox−, Het+, Dab− and Hen−, the last determined by light microscopy.
To check whether the three dev genes are functionally related, devB and devC were mutagenized directly by insertion of the C.K3 cassette in the sites shown in Fig. 1. The resulting constructs were transferred to wild-type Anabaena 7120 on sacB containing suicide plasmids (details about the construction of plasmids is in Experimental procedures). For each mutant strain, sucrose-resistant recombinants (DR42 mutated in devC or DR74 in devB respectively) were analysed and showed the same phenotype as the devA mutants, i.e. Fox−, Het+, Dab− and Hen−. To confirm that the phenotype of devC mutant DR42 was caused by disruption of devC and not by repression of the devA gene by insertion of the cassette into the putative regulatory 5′ region of devA, an intact devA gene was transferred to the devC mutant on a shuttle vector that could complement mutant M7 (pIM27 in Maldener et al., 1994). This plasmid could not complement the devC mutant, showing that its phenotype was not caused by insufficient transcription of the devA gene located downstream.
An ORF, 400 bp downstream of devA, oriented in the opposite direction was identified (Maldener et al., 1994) and named orf2 (accession no. X99672). The deduced amino acid sequence does not show similarity to other known sequences. An insertion mutant (DR29) was created using plasmid pIM29 (see Experimental procedures). This mutant showed a phenotype that was indistinguishable from that of the wild type, being able to form mature heterocysts and to grow on N2 as the sole source of nitrogen. The genotype of each of the mutant strains was confirmed by Southern blot analysis (data not shown).
Analysis of heterocyst-specific glycolipids in the devBCA mutants
According to Ernst et al. (1992), mutant M7 lacks the heterocyst-specific glycolipids (Hgl−), as determined by thin-layer chromatography (TLC) of extracts of filaments that had been nitrogen starved for 48 h. Cells of wild-type Anabaena 7120, devC and devB mutants, as well as cells of mutant M7, were analysed in the same way. As shown in Fig. 3, in wild-type Anabaena 7120 and in each mutant that had been deprived of nitrogen, a spot was detected in a region that has been attributed to heterocyst-specific glycolipids (Winkenbach et al., 1972), whereas this spot was not detected in extracts of cells grown on NO3− (data for the devB mutant are not shown). In extracts of the Hgl− mutant, P2 (Ernst et al., 1992), the specific glycolipid spot did not appear after the induction of heterocyst formation (Fig. 3). In conclusion, the phenotypes of M7, the devB and devC mutants have to be defined as Hgl+.
Ultrastructure of the heterocysts of mutant M7
Despite the presence of heterocyst-specific glycolipids in extracts of whole filaments, the heterocyst envelope of the dev mutants looks thin and less refractile under the light microscope (Ernst et al., 1992; Maldener et al., 1994). This prompted us to examine the ultrastructure of induced filaments of mutant M7 by electron microscopy. For the induction of heterocyst differentiation, NO3−-grown Anabaena cultures were washed three times with NO3−-free medium and incubated for 48 h (M7) or 36 h (wild type) in the same medium. Figure 4 shows the ultrastructure of heterocysts of wild-type Anabaena 7120 and mutant M7 after fixation with glutaraldehyde and permanganate. With this fixation procedure, the glycolipids are retained during dehydration (Lang and Fay, 1971; Winkenbach et al., 1972). Heterocysts of wild-type and mutant M7 possess the homogeneous layer consisting of polysaccharides. The innermost laminated layer composed of heterocyst-specific glycolipids is completely absent in the mutant; in the wild type, however, this layer can be clearly seen near the junction between heterocyst and vegetative cell. The empty space between laminated layer and cell wall is an artefact that occurs frequently during preparation. The distribution of intracytoplasmic membranes in heterocysts of mutant M7 is more confluent than in vegetative cells. However, the formation of the honeycomb region near the poles does not take place. As expected for a Fix− mutant, no polar granule is built up in the mutant, whereas the characteristic small septum at the junctions with vegetative cells is clearly visible. Densely packed glycogen granules can be seen in both cell types of the mutant, which may indicate a depletion of fixed nitrogen (Ernst et al., 1984); in the wild type, glycogen granules are present in vegetative cells only. Vegetative cells of M7 cannot be distinguished from vegetative cells of the wild type. In both strains, the large carboxysomes, containing the paracrystalline RUBISCO structure, are present in vegetative cells only.
Three closely linked ORFs, devB, devC and devA, could form an operon encoding the subunits of an ABC protein-mediated transporter. The relatively long non-coding sequences between devB and devC and between devC and devA are not unusual for cyanobacterial operon structures. Omata et al. (1993), for example, describe a gene cluster encoding an ABC transporter for NO3− with a stretch of 197 bp between two of the genes that are transcribed polycistronically (see also Bartsevich and Pakrasi, 1995). To determine the total size of the transcript, Northern blot analysis and reverse transcriptase–polymerase chain reaction (RT–PCR) were attempted intensively; but all efforts were without success, probably because of a very short half-time or low transcription rates (not shown). However, from the lack of any termination signals between devB, devC and devA, it could be predicted that the genes are transcribed polycistronically. The functional linkage between the three dev genes was shown by site-directed mutagenesis of each gene, resulting in mutants with identical phenotypes. These data are consistent with the idea that the three genes form an operon.
The devC-encoded protein shares the general size and hydropathy profile of typical integral membrane proteins of ABC transporters. The low similarity of the primary sequence of DevC to other known membrane domains is not surprising, as sequence conservation of these components is very low. The ‘Dassa–Hofnung’ consensus sequence of membrane proteins of ABC importers is not present in ABC exporters and does not appear in the DevC sequence (Dassa and Hofnung, 1985). The sequence and hydropathy profile of DevB shows all the typical features of peptide or protein exporters described by Dinh et al. (1994). A signal peptide guiding the protein to the periplasmic space is present neither at the N-terminus of DevB nor in other MFPs. The presence of an MFP-like protein suggests that DevBCA belongs to the class of ABC protein-mediated exporters rather than to the class of ABC importers. A gene that could encode a periplasmic binding protein, an essential component of bacterial ABC importers, was not present in the neighbourhood of the devBCA gene cluster. No gene for an additional component of the transporter or a putative gene that could encode the substrate of the ABC exporter is linked to the devBCA exporter genes, as was found for some other MFP-coupled systems (Létofféet al., 1996), e.g. for the haemoprotein exporter (hasDE ) of Serratia marcescens (Létofféet al., 1994), for the haemolysin exporter (hlyBD) of E. coli (Mackman et al., 1986) and for the polysialic acid exporter (kpsMTE ) of E. coli (Bliss and Silver, 1996).
The MFPs of ABC exporters of carbohydrates, such as the polysialic acid exporter, do not show similarity to the MFPs of peptide/protein exporters (Bliss and Silver, 1996). Subunits of carbohydrate exporters are encoded by single genes, like the subunits of the DevBCA exporter. On the contrary, in the known peptide/protein exporters (with only one exception), membrane-spanning and ATP-binding domains reside on the same polypeptide (Fath and Kolter, 1993). Sequence similarity of the DevB protein to MFPs of peptide/protein exporters suggests a proteinaceous substance as substrate of the DevBCA exporter. However, the structural organization of this exporter resembles that of the carbohydrate exporter. No predictions about the nature of the transported substrate of the DevBCA exporter can be made from the sequence comparison data.
Winkenbach et al. (1972) showed the identity of the heterocyst-specific spot on TLCs (the heterocyst-specific glycolipids) with the laminated layer of the envelope. The same lipids were found in extracts of the devBCA mutants after nitrogen stepdown. However, envelope glycolipids cannot be assembled as a laminated layer outside the cell wall, as seen in electron microscopic sections. Our conclusion is that the DevBCA exporter translocates heterocyst-specific glycolipids or an enzyme, which might be required for the assembly of the laminated layer. The presence of the homogeneous layer in the heterocyst envelope of mutant M7 showed that the formation of the two layers is regulated independently.
The phenotype of the hglK mutant (Black et al., 1995) is similar to that of M7. However, in contrast to mutant M7, not only heterocysts but also vegetative cells of the hglK mutant were affected; structures similar to the ‘thylakoid lacunae’ described for the hglK mutant were not found in the heterocysts of mutant M7. Sequence analysis of HglK did not reveal similarities to known transport proteins. Therefore, we conclude that hglK and the devBCA cluster influence different aspects of the formation of the glycolipid layer.
The absence of the laminated layer serving as a primary barrier to the diffusion of oxygen (Murry and Wolk, 1989) explains well the Fox− phenotype of the devBCA mutants. The pleiotropic phenotype of mutant M7 could be explained by speculating that the maturation of the protoplast depends on the formation of the laminated layer. The profound changes in the distribution of intracytoplasmic membrane structures is blocked in mutant M7, resulting in a lack of the honeycomb region, which agrees well with the lack of DAB oxidation. The hypothesis is proposed that the establishment of a barrier to oxygen, resulting in a decreased PO2, could trigger the process of maturation. This hypothesis awaits proof in future experiments.
Strains and growth conditions
Strains of Anabaena 7120 and derivatives (Table 3) were grown under photoautotrophic conditions at 30°C in the light as described previously (Ernst et al., 1992) with the following changes: dilution of liquid medium (A&A) was 1:4, and undiluted medium was solidified with 1.5% agar. The mutant strains were grown in the presence of 5 mM NO3− and 50 μg ml−1 neomycin sulphate in liquid or 200 μg ml−1 in solidified medium. Medium for strain DR42 (pIM27) additionally contained streptomycin dihydrochloride and spectinomycin sulphate (2.5 μg ml−1 each). Heterocyst differentiation was induced by washing the cultures three times with A&A/4 and resuspending the filaments in the same medium. Chlorophyll content was estimated from methanolic extracts according to Mackinney (1941). Strains of E. coli were grown on LB medium under standard conditions (Maniatis et al., 1982). Transfer of plasmids by conjugation between Anabaena 7120 and E. coli was achieved as described earlier using RP4 bearing strain J-53 and cargo strain HB101 bearing helper plasmid pRL528 in triparental matings (Wolk et al., 1984; Elhai and Wolk, 1988b). Selection for recombinants was performed as described previously (Cai and Wolk, 1990).
Table 3. . Strains of Anabaena used in this study. Nm, neomycin; DR, double recombinant; SR, single recombinant.Others as in Table 2.
Table 2. . Relevant plasmids used in this study. Ap, ampicillin; Bm, bleomycin; Cm, chloramphenicol; Em, erythromycin; Km, kanamycin; Sm, streptomycin; Sp, spectinomycin.
DNA isolation and analysis
Total DNA from Anabaena strains was isolated as described previously (Cai and Wolk, 1990). Plasmids were purified from E. coli with the Qiagen-plasmid kits. Transposon Tn5-1063, together with flanking Anabaena DNA, was recovered from mutant M7 on a ClaI fragment according to Wolk et al. (1991), creating plasmid pIM11. Sequencing was done with the T7 sequencing kit from Pharmacia using the M13 reverse and universal primers and oligonucleotides complementary to the Anabaena DNA. Templates were derived from subclones of pIM11 as described here: EcoRV fragments of pIM11 were cloned into the SmaI site of pUC19 resulting in pIM25 (1.7 kb insert) and pIM23 (3.6 kb insert). By digestion of pIM25 with XbaI or HincII and recircularization, pIM31 (1.5 kb fragment) and pIM33 (1.0 kb fragment) were created. pIM23 was subcloned by restriction with SapI, followed by mung bean nuclease treatment, digestion with HincII and cloning of the fragment into the SmaI site of pUC19. A 1.5 kb fragment of pIM23 was subcloned into the SmaI site of pUC19, and a 2.3 kb fragment was subcloned into the XbaI site of the same vector, giving pIM48 and pIM53. Sequence analysis was performed with the UWGCG package of the University of Wisconsin Genetics Computer Group, version 7.3 (Devereux et al., 1984). Sequence comparisons were made using the BLASTP program of the Heidelberg Unix Sequence Analysis Resources, version 4.
Mutation of devA, devB, devC and orf2
The orf2 gene on pRL1451 (Maldener et al., 1994) was disrupted by the insertion of C.K3, derived as a SmaI fragment from pRL448 (Elhai and Wolk, 1988a), into the XmnI site replacing a 45 bp fragment. After NdeI digestion, the insert was ligated into the AseI site of pRL271 (Black et al., 1993), resulting in pIM29. The devA gene was disrupted by the insertion of C.K3 into the NheI site, resulting in pIM16. An SphI–Ecl 136II fragment of pIM16 was ligated into the NruI and SphI sites of pRL271, creating pIM22. The devC gene derived from plasmid pIM11 was subcloned first as a 1.4- kb SpeI–HindIII fragment into pUC19 cut with XbaI and HindIII (pIM40). After the insertion of C.K3 into the EcoRV site of pIM40, the Ecl 136II–AseI fragment was ligated into pRL271, resulting in pIM42. To construct a devB mutant, plasmid pIM48 was digested with SpeI, and C.K3, derived as an XbaI fragment from pRL448, was inserted. Insertion into pRL271 was achieved in the same way as with pIM42 creating pIM74.
Analysis of glycolipids
From 50 ml cultures that had been deprived of nitrogen for 48 h, total lipids were extracted twice with methanol–chloroform (1:2). The extracts were evaporated in a stream of air, dissolved in 200 μl of chloroform and chromatographed on thin-layer plates of silica gel (Merck) as described previously (Winkenbach et al., 1972).
Fixation with 2.5% glutaraldehyde and 2% KMnO4 and dehydration was performed as described previously (Black et al., 1995). After dehydration, the samples were incubated in a 1:1 mixture of Durcupan (Fluka) and propylene oxide overnight at 37°C, followed by embedding in Durcupan for 24 h at 37°C and 48 h at 60°C in BEEM capsules with an open lid. Thin sections of 70–90 nm were collected on copper grids and stained with uranyl acetate for 20 min and with lead citrate for 5 min. The samples were examined with a Zeiss EM109 electron microscope at 80 kV.
Present address: Universität Heidelberg, Institut für Biochemie I, Im Neunheimerfeld 328, D-69120 Heidelberg, Germany.
The authors would like to thank Professor Dr C. Peter Wolk, Michigan State University, for plasmids pRL271 and pRL448, Professor Dr Juergen Boeckh and Janna Streck from the Lehrstuhl für Zoologie, University of Regensburg for access to and assistance with electron microscopy and Dr Eckhard Loos and Dr Margret Sauter for critical reading of the manuscript. The work was supported by the Deutsche Forschungsgemeinschaft (DFG).