A cosmid containing a wild-type Anabaena PCC 7120 DNA fragment was found to suppress heterocyst differentiation, creating a Het phenotype in an otherwise wild-type strain. Curing of the cosmid restored the full wild-type Het+ Nif+ phenotype. The cosmid contains at least four genes encoding proteins with significant sequence similarity to enzymes involved in the synthesis of fatty acids. Selection for Nif+ revertants of the suppressed strain yielded modified cosmids, one of which contained a 10.2-kb transposon, Tas1, inserted into the promoter region of a gene encoding a protein with acyl carrier and β-keto reductase domains. This gene, called hetN, was shown previously by Black and Wolk (J. Bacteriol. (1994) 176, 2282–2292) to inhibit heterocyst differentiation when present alone on a plasmid. Oddly, hetN gene transcription is detected later than 6 h into heterocyst differentiation.
Some filamentous cyanobacteria can differentiate a fraction of their vegetative cells into nitrogen-fixing cells called heterocysts. In Anabaena PCC 7120 filaments, approximately every tenth cell differentiates when fixed nitrogen is removed from the medium. Heterocysts provide a low oxygen environment for nitrogen fixation. Vegetative cells provide a carbon source for heterocyst metabolism and heterocysts provide a source of fixed nitrogen to the vegetative cells. Reviews of metabolite exchange, models of pattern formation and heterocyst development are available [1–3].
To date, several genes whose products significantly affect heterocyst spacing have been identified [4–7]. The patA gene has a domain related to the cheY gene of Escherichia coli and is thought to be the response regulator of a two-component system. Mutations in patA lead to filaments with mostly terminal heterocysts instead of the normal spacing pattern . The gene hetR is required for the initiation of heterocyst development. Null mutations and a mutation which converts a serine residue to an asparagine in HetR prevent the initiation of heterocyst development [8, 4]. The patB gene encodes a protein with ferredoxin and helix-turn-helix domains. Mutation of patB results in a strain that has greatly delayed heterocyst formation, but eventually the pattern shows extra heterocysts .
Transposon insertion identified the N10 locus (hetN), which is also involved in the regulation of heterocyst differentiation . The original transposon insertion in the hetN gene was characterized as producing a Het (absence of heterocysts) phenotype. However, further experiments with cassette insertions in this gene indicated that hetN mutants are not Het, but produce supernumerary heterocysts instead. Black and Wolk refer to this phenotype as Mch, for Multiple contiguous heterocysts . The Mch strains produced by interruption of the hetN gene were unstable. They usually acquired secondary mutations blocking heterocyst development, such as occurred in all likelihood in the original N10 strain . The hetN gene suppressed heterocyst development when alone on a plasmid, but this suppression was relieved when a second gene, hetI, downstream and antisense to hetN was included on the plasmid. HetI is similar in sequence to certain regulatory proteins. It was not determined whether the protein HetI or the transcription of hetN antisense from the hetI promoter is responsible for relief of suppression of heterocyst development by HetN . HetN contains a domain similar to β-ketoacyl reductases .
The hglB (formerly hetM) gene is located upstream of hetN. hglB encodes a protein with two apparent functional domains: the amino terminus is similar to an acyl carrier protein and the central portion is similar to keto reductases. An insertion mutation in the hglB gene results in inability to produce heterocyst glycolipids .
2Materials and methods
Anabaena PCC 7120 was grown in modified Kratz and Myers medium C (K&M)  or BG-11 medium . In place of Na2HPO4, 1.125 mM each of Na2HPO4 and K2HPO4 were added to the K&M media. The nitrogen sources added for N+ growth were either 2.5 mM (NH4)2SO4 (K&M+NH4) or 17.6 mM NaNO3 (K&M+NO3). Plates contained K&M or BG-11 media with 1.3% agar (BBL purified) and 17.6 mM NaNO3 if a nitrogen source was included. Cultures were grown photoautotrophically under 30–40 μE/m2/s cool white fluorescent lighting at 25–30°C in an incubator with 2% CO2 in air. Large scale cultures were bubbled with 2% CO2 in air. Mid-log phase cells refer to cultures containing 2–6 μg/ml of chlorophyll corresponding to 0.7–2.0×107 cells/ml. Chlorophyll was determined by extraction with methanol and measurement of absorption as described .
For selective growth of E. coli DH5α, antibiotics were used at concentrations of 100 μg/ml ampicillin and 50 μg/ml kanamycin. For selective growth of recombinants, 100 μg/ml neomycin was used for maintenance of single recombinant gene interruptions, and 30 μg/ml neomycin and 20 μg/ml each of spectinomycin (Spc) and streptomycin (Str) for plasmid-based replicating vectors.
Filaments were collected for RNA isolation at 6 h time intervals during a large scale heterocyst induction of the wild-type strain. Vegetative cells from 80 ml of a 7-day-grown BG-11+NO3 culture were transferred to 8 liters of fresh medium containing NaHCO3 (10 mM), NH4NO3 (2.5 mM), and MOPS (5 mM, pH 8.0). Growth conditions were as described above, monitored by optical density at 750 nm and light microscopy. The culture was induced to form heterocysts after 4 days growth (OD750 of 0.35). The vegetative filaments were collected by centrifugation at 3000×g for 5 min at room temperature, washed twice with sterile water, and transferred to 8 liters of fresh medium containing 10 mM NaHCO3 but lacking NH4NO3 and MOPS. At 6 h intervals, filaments from 1.1-liter samples were collected by centrifugation and frozen at −80°C for subsequent RNA extraction. Filaments were examined microscopically during the induction; proheterocysts were present at 12 h and mature heterocysts containing polar granules were present at 15–16 h after nitrogen stepdown.
2.3RNA isolation and Northern analysis
N+ RNA was prepared from wild-type cells grown in 100 ml of BG-11+NH4NO3 for 5 days (OD750 approximately 0.5). RNA was prepared from frozen samples of the large-scale induction of Anabaena PCC 7120 by centrifugation through 5.7 M CsCl. For Northern blots, approximately 20 μg samples of total RNA were denatured with formaldehyde, separated by electrophoresis on a 1.2% formaldehyde-agarose gel  and transferred to MagnaCharge membranes (MSI) with 10×SSPE . The blots were hybridized with random primer-labeled probes at 55°C in 50% formamide-5×SSPE-1% sodium dodecyl sulfate (SDS) and washed at 65°C in 0.5×SSPE-0.1% SDS.
2.4Molecular biology techniques
Preparation, restriction enzyme digestion, and ligation of hybrid plasmid DNAs used previously described techniques . Protein and DNA sequence comparisons used the GenBank databases via the National Center for Biotechnology Information's network services and the BLAST program . Multiple sequence alignments used the Clustal V program .
2.5Isolation of HDIC36
A twenty-thousand member cosmid bank containing Anabaena PCC 7120 chromosomal DNA fragments was conjugated by tri-parental mating to wild-type Anabaena PCC 7120 and exconjugants were selected with 30 μg/ml neomycin on nylon filters on N+ plates . After two weeks the nylon filters were transferred to K&M plates containing 30 μg/ml neomycin. Colonies that grew poorly on these plates were picked and rescued by streaking on K&M+NO3. The Fox phenotypes (unable to Fix nitrogen in the presence of OXygen), or poor growth phenotypes, were then confirmed by re-streaking on K&M. Colonies were grown in BG-11+NO3 and plasmids were extracted and used to transform E. coli. Re-introduction of these cosmids by conjugation into wild-type cells was used to determine whether the cosmids were responsible for the growth phenotypes (Table 1).
Table 1. Representative cosmids selected on the basis of interference with heterocyst development
Color on N− plate
HindIII bands in common
Phenotype when cosmid reintroducedc
Phenotype when cosmid curedd
aPhenotype, in N− medium, of wild-type cells carrying the cosmid named in the first column. All of these strains are unable to fix nitrogen. Cosmids 11 and 16 overlap with respect to nearly all of their DNA fragments, as do cosmids 17, 36 and 55.bRatio of wild-type to mutant colonies on N− plates.cPhenotype resulting when the cosmid was extracted from the first Het strain, used to transform E. coli, then conjugated back into wild-type Anabaena PCC 7120. Cosmids 1, 10, 11, and 16 failed to inhibit heterocyst differentiation when tested this way, indicating that the original exconjugant was Nif, Het due to chromosomal mutation.dPhenotype of the original mutant exconjugants following curing of the HDIC cosmid. Only the strains harboring cosmids 19, 36 and 55 were restored to the full wild type by curing. The chromosomal mutations that remained in the other cured strains could not be complemented by a cosmid library of wild-type DNA fragments .
Yellow wild type
Hets fall off
Yellow wild type
Yellow wild type
Yellow, Het, long cells
Green, wild type
Green, wild type
Green, single cells
Strains with poor growth phenotypes on nitrogen-deficient media were cured by bi-parental mating with E. coli cells containing a non-compatible cosmid vector, pDB21. pDB21 was derived from the original pDUCA7 cosmid vector by the addition of a Spc/Str resistance cassette inserted into the PstI sites of pDUCA7 . After selecting for Spc/Str resistance, the colonies were grown in 10 ml liquid BG-11+NO3 cultures with Spc at 20 μg/ml and tested for neomycin resistance on K&M+NO3 plates at 30 μg/ml. Those strains that were Spc/Str-resistant and Neo-sensitive, considered to be cured of the original cosmid, were then re-tested for growth on nitrogen-deficient medium. Mutants that became wild type after plasmid segregation were thus confirmed as having contained heterocyst-suppressing cosmids. All mutants were also examined for structural defects or the absence of heterocysts by light microscopy.
2.6Reversion of heterocyst suppression
A lawn of Anabaena PCC 7120 cells carrying pHDIC36 was incubated on K&M+neomycin 30 μg/ml medium for three to four weeks. Nif+ colonies were picked, grown in BG-11+NO3 and cosmids were isolated. Restriction digests of these cosmid isolates were compared to the original pHDIC36. In one case, additional bands were found, indicating the translocation of a DNA element into pHDIC36. The transposon in the larger cosmid, pHDIC36a, was located using restriction analysis with HindIII and ClaI. The ends of the putative transposon were then defined by sequencing ClaI subclones using the universal and reverse primers of pUC19.
2.7Sequencing of the transposon insertion region
The DNA insert in cosmid pHDIC36, called Tas1 for transposable Anabaena sequence, was located in a 1.9-kb ClaI fragment of pHDIC36. This fragment and an adjoining 2.5-kb ClaI fragment were sequenced on both strands with a SequiTherm Cycle sequencing kit (Epicentre Technologies) and [α-32P]dATP (DuPont, NEN Research Products). Sequencing reactions were primed with custom-made oligonucleotides. These sequences (hetN, hglB and ORF 552) are available from GenBank under the accession number U04436.
A 6.5-kb internal ClaI fragment of Tas1 was sequenced in part by using restriction fragment deletions and the appropriate forward or reverse primers. Custom-made oligonucleotide primers were used to fill in gaps. This sequence is available from GenBank under the accession number U13767. The remaining sequences of Tas1, including two copies of a 1.6-kb Insertion Sequence in direct repeat orientation at the ends of the transposon, will be published elsewhere (C. Bauer, L. Scappino, W.J. Buikema and R. Haselkorn, manuscript in preparation).
Insertional mutagenesis used a suicide vector called pCCB111aa carrying internal gene fragments, as described previously [15, 16]. From the two genes, hglB and ORF552, located in the sequenced ClaI fragments, internal HindIII fragments of 500 and 450 bp, respectively, were isolated and used in attempts to interrupt each of these ORFs in the chromosomes of wild-type Anabaena PCC 7120. Single recombinants that interrupted hglB were selected using 100 μg/ml neomycin. Southern analysis confirmed complete chromosomal replacement, utilizing the internal HindIII fragment as probe. Interruptions of ORF552 were not isolated.
3Results and discussion
3.1Cosmids that suppress heterocyst differentiation
To search for cosmids that perturb the ability of wild-type Anabaena PCC 7120 to form functional heterocysts and to fix nitrogen, a twenty-thousand member cosmid bank of wild-type DNA fragments was conjugated into Anabaena PCC 7120 by tri-parental mating and exconjugants were selected on neomycin plates . The filters from these plates were then transferred to K&M medium  and 60 colonies that could not grow without fixed nitrogen were detected and rescued on K&M-NO3. These strains were named HDIC1 to HDIC60 (Heterocyst Differentiation Inhibition Cosmid). Each cosmid was re-isolated, introduced into wild-type Anabaena PCC 7120, and the strains re-tested for their ability to fix nitrogen. In addition, an incompatible plasmid, pDB21, containing the same pDU1 replicon as the cosmid vector but different antibiotic resistance genes, was introduced into 24 of the 60 HDIC cosmid-containing strains. In each case, pDB21 cured the cosmids from these strains. Strains containing the five cosmids discussed below were restored to wild-type after selection for pDB21 with spectinomycin and streptomycin, while the 19 others retained their mutant phenotypes, indicating that they were defective due to chromosomal mutations rather than to the cosmids they contained. Of the 60 cosmids tested, only five, HDIC17, HDIC36, HDIC54, HDIC55, and HDIC56, produced the same phenotype when reintroduced into wild-type Anabaena PCC 7120. Three of these cosmids, HDIC17, HDIC36, and HDIC55, had overlapping restriction maps with only a 10-kb difference among them (Table 1). Cosmids HDIC11 and HDIC16 failed to suppress differentiation when re-introduced into wild-type cells, so although they had restriction fragment patterns similar to each other, they were not characterized further. In addition, the exconjugants originally isolated with these cosmids displayed different defects and were still crippled after curing the HDIC cosmid, so these were concluded to bear chromosomal mutations. Based on a great deal of anecdotal observation, we believe that conjugation of pDU1-based plasmids into Anabaena PCC 7120 is mutagenic. Indeed, the mutations induced in these experiments are either extensive or multiple, because they could not be complemented with a library of wild-type DNA fragments, whereas that same library could be used efficiently for complementation of chemically induced point mutations .
Two additional cosmids that suppress heterocyst differentiation, 3E10 and 8E11, were identified in an independent cosmid library in experiments designed to complement mutants that form heterocysts constitutively . The insert in cosmid 3E10 shows extensive overlap with HDIC36 and cosmid 8E11 overlaps with HDIC56 (Table 1).
3.2Relief of heterocyst suppression by insertion of a transposon into hetN
Each of the 24 strains used in the plasmid curing experiments was also analyzed for frequency of reversion of the phenotype to Het+, Nif+. All strains exhibited some degree of reversion, but the range varied widely as shown in Table 1. HDIC36 was chosen for further study. Twelve cosmids were isolated from HDIC36 revertants, i.e. colonies that formed on N− plates. One of these cosmids, pHDIC36a, showed a change in its HindIII restriction pattern due to a 10.2-kb insert. The insert was sequenced in part and found to contain a novel transposable element, Tas1. This element contains two copies of a 1.6-kb IS element in direct repeat orientation at its ends and five open reading frames in its interior. The Tas1 internal sequence is available from GenBank under the accession number U13767.
The transposon insertion was initially localized to a 1.9-kb ClaI fragment in pHDIC36. Sequence analysis located it more accurately downstream of the hglB gene, within the 5′-flanking region of the hetN gene (Fig. 1). Numerous insertions of a selectable cassette were made by partial digestion of cosmid HDIC36 with Sau3a and ligation of the omega fragment (Fig. 1). Among these insertions, only those in or flanking the hetN gene relieved the suppression of heterocyst differentiation. Moreover, various subclones of the related cosmid 3E10 were tested for heterocyst suppression and all subclones that lacked the 1.9-kb fragment containing hetN failed to suppress.
3.3Other genes on the hetN-containing cosmid
The sequence of the 2.5- and 1.9-kb ClaI fragments revealed a 552-amino acid ORF upstream from the 506-amino acid ORF of hglB, separated by 105 bp, both in the same orientation as hetN (Fig. 1). The predicted translation product of ORF552 was compared with the combined protein databases at NCBI utilizing the blastp and blastn programs , but no significant matches were found.
Downstream from ORF552 is the hglB gene. A region of dyad symmetry indicative of a strong rho-independent transcriptional terminator is located 50 bp after the hglB ORF. The hglB gene sequence predicts a protein with similarity to both polyketide synthases and to fatty acid synthases, as has been noted previously . The domains located in HglB show similarity to acyl carrier proteins, as evidenced by a 4′-phosphopantetheine-binding motif, and to keto reductases with an NAD(P)H binding site.
Two regions of heptamer repeats were found upstream of ORF552. A heptamer of GTTAAC(A/C/T) was repeated six times and a heptamer of CC(C/T)ATTA was repeated four times. These are found within the intergenic region of 260 bp between the end of an upstream ORF, hglC, and the putative start of translation of ORF552 (Fig. 1).
Attempts were made to construct single recombinant null mutations utilizing internal fragments of ORF552 and hglB in the non-replicating vector pCCB111aa [15, 16]. Neomycin-resistant colonies were tested on nitrogen-deficient medium and analysed by light microscopy. Complete gene inactivation was confirmed by Southern analysis. After screening 20 plates, which should have been sufficient to obtain at least one hundred mutants based on observed recombination frequency in other inactivations, no insertions in ORF552 were obtained, suggesting that the product of this ORF is essential for growth. Strains with insertions in hglB are unable to grow aerobically on nitrogen-deficient medium, although they do differentiate heterocysts. These heterocysts are defective, lacking the normal glycolipid envelope and missing heterocyst-specific glycolipids as revealed by thin-layer chromatography (data not shown). Thus, HglB is required for heterocyst glycolipid synthesis.
3.4Transcription of ORF552, hglB and hetN
Total RNA was collected from vegetative cells, purified heterocysts and cells taken 6, 12, 18, 24 h after transfer to N− medium and probed with HindIII fragments of 0.45 kb and 0.5 kb internal to ORF552 and hglB. With either probe, a smear running from 3 kb to smaller sizes was seen in the 12-, 18-, 24-h and purified heterocyst lanes, but no message was seen in the 6-h or the vegetative cell lanes (data not shown). This result indicates that a single transcript could contain both ORF552 and hglB. The timing of appearance was the same for the hetN gene transcript, whose size of 1.4 kb is sufficient for that gene alone (Fig. 2).
We are thus faced with a dilemma: how can a gene, hetN, that is normally expressed late in heterocyst differentiation prevent that differentiation when carried on a plasmid? Note also that the suppression of heterocyst differentiation can be seen either when the hetN gene is alone on a plasmid  or when it is in the full context of its surrounding genes on the cosmid HDIC36. Black and Wolk suggested that the hetN gene product might synthesize a secondary metabolite involved in intercellular signalling, regulating heterocyst differentiation negatively . Overproduction of this substance, due to the extra copy of hetN on a plasmid, would totally suppress heterocyst differentiation. On the other hand, inactivation of the hetN gene would lead to ectopic or precocious development of too many cells, a potentially lethal situation, so strains in which hetN was interrupted would accumulate second site mutations rendering them Het, allowing them to survive on N+ medium. These suggestions account for all of the data except the rather late expression of the hetN gene. It is possible that hetN has two modes of transcription, and the low level of expression in vegetative cells (essential if HetN plays a role in regulating differentiation) is not detected in our Northern blot.
The HetN protein contains an unmistakable β-keto reductase domain, suggesting that it is part of the biosynthetic machinery for fatty acids. Three of the upstream ORFs, hglB, hglC and hglD, have been shown to be required for heterocyst glycolipid synthesis (manuscript in preparation). Null mutations in any of the latter three genes result in strains that make heterocysts lacking the glycolipid layer and lacking the heterocyst-specific glycolipids; they cannot fix nitrogen under aerobic conditions .
We are grateful to W.J. Buikema for advice on all aspects of this work. The work was supported by research grants GM21823 and GM40685 from the NIH and a predoctoral traineeship (GM07183) to C.C.B. The two ClaI fragments were sequenced on both strands in the Advanced DNA Technologies Laboratory in the Department of Biology, Texas A & M University.