INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Cyanobacteria comprise a diverse group of photosynthetic prokaryotes with oxygen-evolving photosynthesis similar to that of higher plants. Many species of cyanobacteria are capable of nitrogen fixation; however, because nitrogenase is very oxygen sensitive, cyanobacteria separate nitrogen fixation from photosynthesis either temporally or spatially (reviewed in references 14 and 16). In Anabaena spp., aerobic nitrogen fixation is confined to differentiated cells called heterocysts that form in a semiregular pattern in a filament in response to nitrogen starvation. Fixed nitrogen in the heterocysts is transported to vegetative cells in the filament, while vegetative cells supply carbon and reductant to heterocysts (reviewed in references 17 and 46). Heterocysts lack oxygen-evolving photosystem II activity (29, 35), have increased respiration, and synthesize a glycolipid layer that is important in protection of nitrogenase from oxygen (28, 42, 46). Hence, heterocysts maintain a relatively anoxic microenvironment in a filament that is predominantly oxic.

Filaments growing with an external source of fixed nitrogen do not contain a significant number of heterocysts. However, removal of fixed nitrogen from the environment, either by washing the cells or by allowing them to deplete low concentrations of fixed N by growth, results in massive degradation of protein followed by de novo differentiation of heterocysts in a spaced pattern (36). Two aspects of heterocyst differentiation are of interest: the mechanisms that give rise to the initial patterned differentiation of heterocysts from apparently identical vegetative cells, and the maintenance of the patterned differentiation of new heterocysts during diazotrophic growth. Since fixed nitrogen, particularly ammonium, represses heterocyst formation, it has been postulated that the differentiation process is controlled by the availability of fixed nitrogen in the vegetative cells (43, 44). In addition, the pattern of heterocyst spacing within a filament may be controlled by a nitrogenous product made by existing heterocysts and metabolized by intervening vegetative cells (43, 44). In such a model, a gradient of fixed nitrogen would emanate from heterocysts, with vegetative cells midway between existing heterocysts becoming starved for nitrogen as the filament grows. Such starved cells would themselves differentiate in response to nitrogen starvation, maintaining the spaced pattern of heterocysts.

The genes involved in early heterocyst differentiation and pattern formation that have been identified (reviewed in references 17 and 45) include ntcA (15, 30, 43), hanA (20), hetR (4, 6), hetC (21, 27), patA (22), patB (23), and patS (47). However, little is known concerning control of the cascade of genes whose expression follows induction of differentiation (7). NtcA, a global nitrogen regulatory protein in the cyclic AMP receptor protein family of transcriptional activators, is required for the utilization of nitrate and for heterocyst differentiation (and hence for nitrogen fixation under oxic growth conditions) (15, 30, 43). NtcA binds to a putative consensus sequence that is found upstream of the promoter of a number of cyanobacterial genes (27) and is presumed to exert its activity by activating expression. It is required for hetR transcription (15) and directly binds to the promoter region of hetC (27). Although it is required for utilization of nitrate and N₂, it is not known whether it has a direct role in activation of nitrogenase.

PatS-5 is a pentapeptide that is processed from the 40-amino-acid precursor polypeptide, the product of the patS gene. PatS-5, which is made within about 6 h after heterocyst induction in spaced cells in the filament, represses heterocyst differentiation. Hence, it is likely that PatS-5 is an inhibitor of heterocyst differentiation that is made in developing heterocysts to prevent the differentiation of nearby vegetative cells (47). It is a good candidate for the repressor that maintains the pattern of spaced heterocysts during diazotrophic growth.

While there is evidence that heterocysts prevent the differentiation of nearby vegetative cells and PatS-5 may be the inhibitor, there is little evidence for a direct role for simple nitrogenous compounds (nitrate, ammonium, urea, or amino acids) in heterocyst differentiation or pattern formation. In fact, there is ample evidence that heterocysts can differentiate in the presence of fixed nitrogen. Anabaena variabilis grows on glutamine as the sole nitrogen source, yet under these conditions patterned heterocysts form in which nitrogenase activity is repressed (presumably by the high intracellular levels of fixed nitrogen) (38). Overexpression of hetR leads to the formation of heterocysts in the presence of nitrate (6), as does a patS mutant (47). Methionine sulfoximine and 7-azatryptophan, as well as other amino acid analogues, allow heterocyst formation in the presence of fixed nitrogen (8, 33). Glutamine synthetase mutants of A. variabilis have heterocysts and high levels of nitrogenase in the presence of ammonium, and they excrete ammonium into the medium (34). In addition to all of this evidence that the availability of fixed nitrogen does not control heterocyst formation is the even more compelling argument that the pattern for heterocyst development is established long before nitrogen fixation begins (46). Thus, while pattern formation may be controlled by a product made in heterocysts, it is unlikely to require a product of nitrogen fixation.

In A. variabilis, two Mo-dependent nitrogenases have been identified (5, 19, 31, 39, 40). One nitrogenase, the product of a nitrogenase gene cluster called nif1 (5, 19), is expressed exclusively in heterocysts and functions under oxic growth conditions (12, 40). The other nitrogenase, encoded by a homologous gene cluster called nif2, is expressed only under anoxic conditions in vegetative cells shortly after nitrogen step-down, long before heterocysts form (31, 39, 40). Both nitrogenases function well, and either enzyme can, under appropriate physiological conditions, support the fixed nitrogen needs of the filament (39).

Nitrogenase synthesis is tightly regulated by the availability of fixed nitrogen; therefore, it seems reasonable that fixed nitrogen produced as a result of expression of one nitrogenase would repress subsequent expression of the other. If the products of nitrogen fixation in a filament control heterocyst differentiation and pattern formation, then the early expression of the Nif2 nitrogenase in vegetative cells under anoxic conditions would be expected to prevent normal heterocyst differentiation. We have examined the role of NtcA in expression of Nif2 and the effects of expression of the Nif1 and Nif2 nitrogenases on heterocyst differentiation in A. variabilis.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains and growth conditions. A. variabilis FD is a derivative of A. variabilis ATCC 29413 that can grow at 40°C and can support the growth of bacteriophages better than the parent strain (9). JE994 (39) is a derivative of a nif1 mutant strain (JE9) that lacks Nif1 nitrogenase activity and grows well under anoxic conditions using the Nif2 nitrogenase. NF-76 (10) is a mutant of FD that fails to differentiate heterocysts and hence lacks Nif1 nitrogenase activity. A. variabilis FD and mutant strains derived from this strain were routinely grown photoautotrophically in liquid cultures in an eightfold dilution of the medium of Allen and Arnon (2) (AA/8) or in AA/8 supplemented with 2.5 mM NaNO₃ and 2.5 mM KNO₃. Cyanobacterial cultures were maintained on AA/8 or on BG-11 medium (3) solidified with 1.5% Difco Bacto agar (41). All strains were grown as 50-ml cultures in 125-ml Erlenmeyer flasks at 30°C on a reciprocal shaker under cool-white fluorescent lights (approximately 50 microeinsteins m¹ s¹).

Growth experiments. Cells were grown aerobically in the light with shaking in AA/8 with 5.0 mM fructose, 5.0 mM NH₄Cl and 10 mM N-tris(hydroxymethyl)methyl-2-aminoethanesulfonic acid (TES) (pH 7.2) to an optical density at 700 nm (OD₇₀₀) of about 0.3. Cells were washed with AA/8, resuspended in 50 ml of AA/8 with 5.0 mM fructose to an OD₇₀₀ of <0.1, and incubated under anoxic conditions in the same medium. Anoxic cultures contained 10 µM dichlorophenyldimethylurea (DCMU) (to inhibit oxygen evolution from photosystem II) in serum-stoppered flasks flushed thoroughly with dinitrogen. Growth experiments and heterocyst frequency determinations were repeated at least three times, and a representative graph is provided.

Acetylene reduction assays. Cells were grown and washed as described for growth experiments, concentrated to an OD₇₀₀ of about 0.8, and incubated under anoxic conditions with DCMU. At various times after removal of ammonium from the medium, 1.0-ml samples were removed for acetylene reduction assays (24). Experiments were repeated at least three times, and a representative graph is provided.

Cloning of ntcA and isolation of an ntcA mutant. A lambda EMBL3 clone containing the ntcA gene of A. variabilis was identified using an internal restriction fragment of the ntcA gene of Anabaena sp. strain PCC 7120 as a probe (kindly provided by J. Golden [43]). The ntcA gene region was subcloned as a 5-kb HindIII/SalI fragment into pUC118 (producing plasmid pMM1), and the ntcA gene was sequenced using standard molecular biology techniques. A mutant was constructed by inserting a neomycin resistance (Nm^r)-kanamycin resistance (Km^r) cassette (C.K3) with AccI ends at the unique ClaI site in ntcA, forming plasmid pMM2. C.K3 contains the npt gene from Tn5 with a promoter from the psbA gene of Amaranthus hybridus that confers high-level Nm^r in Anabaena sp. strain PCC 7120 (11). The mobilizable plasmid pMM3 was constructed by cloning the 6.1-kb HindIII/SalI fragment of pMM2 (containing ntcA interrupted by the Nm^r-Km^r cassette) into pBR322 at the same restriction sites. Methods used for gene transfer from Escherichia coli to A. variabilis as well as for selection and screening of cyanobacterial mutants have been described previously (24, 37). Segregation of the mutant allele was verified by Southern hybridization and by the absence of a wild-type ntcA PCR product in the mutant, MM3, after amplification of DNA from MM3 using primers flanking the gene (data not shown).

Mobility shift assays. NtcA was obtained from E. coli BL21(pREP4, pCSAM70), which overproduces His-tagged NtcA (27), by sonication and centrifugation as previously described (27). Protein concentration was determined by the Coomassie blue protein assay (Pierce product no. 23200). DNA promoter fragments used in mobility shift assays were obtained by PCR. The 448-bp nifH2 promoter fragment of A. variabilis was amplified using oligonucleotides nifH2-448L (5'-GAAGATCTTGATGGGGGAGATATCGAACTGTA-3') and nifH2R (5'-ATACCCGGGACCGATACCACCTTTACCGTAGAA-3'). The 350-bp glnA promoter fragment of Anabaena sp. strain PCC 7120 was amplified with oligonucleotides glnA-5 (5'-CATGATATCTGCTATCTATGGTTTGATAT-3') and glnA-3 (5'-TCGGGATCCGGGTTGTCATTGTTACTCCT-3'). DNA promoter fragments were end labeled using T4 polynucleotide kinase (Promega product no. M4101) with [-³²P]ATP (NEN product no. NEG502A). Each 20-µl binding reaction mixture contained about 80 µg of crude protein extract, 20,000 cpm of end-labeled promoter fragment, and 2 µg of poly(dI-dC) suspended in binding buffer (4 mM Tris-HCl [pH 8.0], 12 mM HEPES [pH 7.9], 12% glycerol, 60 mM KCl, 0.5 mM EDTA, 1 mM dithiothreitol). Binding assay mixtures were incubated for 30 min at 30°C and separated on a 5% polyacrylamide gel. Radioactive bands were visualized using a STORM 860 PhosphorImager (Molecular Dynamics).

Isolation of RNA and determination of nifH2 transcription start site. RNA was isolated from wild-type strain FD 4 h after anoxic induction (see "Growth experiments" above for details) as described previously (37). The transcription start site was determined by primer extension as described previously (37), using a primer (nifH2R; 5'-ATACCCGGGACCGATACCACCTTTACCGTAGAA-3') that binds to the RNA just downstream from the nifH2 initiation codon.

Immunoblots. Cyanobacterial cells were harvested by centrifugation and heated in sodium dodecyl sulfate (SDS) lysis buffer prior to electrophoresis (3). Proteins (45 µg/lane) were separated on SDS-10% polyacrylamide gels and transferred to nitrocellulose membranes using standard procedures (18). Blots were incubated with a 10,000-fold dilution of a 1:1 mixture of two anti-NifH antibodies followed by incubation with a 30,000-fold dilution of alkaline phosphatase-conjugated anti-rabbit antibodies (Sigma product no. A3687). Reactions were detected using the chromogenic reagents nitroblue tetrazolium (Sigma product no. N6639) and 5-bromo-4-chloro-3-indolylphosphate (Sigma product no. B6777) (17). One antibody (kindly provided by Anneliese Ernst) was made against the NifH1 protein of A. variabilis. The other (kindly provided by Paul Ludden) was a universal anti-NifH made against a mixture of purified NifH proteins from Azotobacter vinelandii, Clostridium pasteurianum, Rhodospirillum rubrum, and Klebsiella pneumoniae.

In situ localization of -galactosidase activity. Cells were fixed in 0.01% glutaraldehyde at 25°C for 15 min and washed with water. Cell pellets were resuspended in 15 µl of 100 µM C₁₂-fluorescein--D-galactoside (C₁₂-FDG; Molecular Probes) in 25% dimethyl sulfoxide. Cells were incubated in the dark at 37°C until fluorescence was microscopically visible (15 to 60 min). Filaments were washed, resuspended in one drop of water, and photographed with a fluorescein filter set (excitation, 450 to 490 nm; dichroic, 510 nm; barrier, 520 nm) on a Zeiss epifluorescence microscope, with a 560-nm short-pass filter to block the red fluorescence of the biliproteins (40). Images were acquired using a Photometrics cooled charge-coupled device camera with ScanAnalytics IPLab software. The image acquisition (exposure) time for the fluorescent photograph was about 0.5 s. The image acquisition time for the light micrograph was 0.05 s.

Nucleotide sequence accession number. The sequence reported has been assigned GenBank accession number U89516.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Transcription start site of nifH2 and role of NtcA in nif2 expression. Since the nif2 genes are expressed only under anoxic conditions primarily in vegetative cells, we were interested in analyzing the promoter region and determining whether the nif2 genes required activation by NtcA. The transcription start site of nifH2 was determined by primer extension to be 146 nucleotides upstream of the coding region (Fig. 1A and B). The promoter region had no consensus NtcA binding site (GTAN₈TAC) located 20 to 23 nucleotides upstream from the 10 region (25) (Fig. 1A). In mobility shift assays, NtcA protein bound to the glnA promoter region, which has a consensus NtcA binding site, but did not bind to the nifH2 promoter region (data not shown).

View larger version (27K): [in this window] [in a new window]   FIG. 1.   Promoter region of nifH2 and NtcA-dependent expression of Nif2. (A) Sequence of the region from the end of nifU2 to the start of nifH2. The transcription start site for nifH2 is indicated by a short arrow above nucleotide 4293 (numbering is based on GenBank sequence U49859 for the entire nif2 cluster) (39). (B) The transcription start site was determined by primer extension using a primer that spans nucleotides 4467 to 4490. (C) FD () and MM3 (ntcA mutant) () cells grown with 5.0 mM NH4Cl-10 mM TES (pH 7.2)-5.0 mM fructose were washed with AA/8, resuspended in AA/8 with 5.0 mM fructose, and incubated under anoxic conditions. Nitrogenase activity was determined by acetylene reduction assays at the times indicated.

Growth of strains with the Nif1 or Nif2 nitrogenase. The two Mo-dependent nitrogenases in A. variabilis are expressed under different physiological conditions. Since only the Nif1 nitrogenase is expressed under normal oxic growth conditions, it is sufficient for good diazotrophic growth. However, even under anoxic conditions where the Nif2 nitrogenase genes are expressed, heterocysts differentiate temporally and spatially as they would in filaments starved for fixed nitrogen. One possible explanation for this was that the level of fixed nitrogen produced by the Nif2 nitrogenase was insufficient to support growth and to suppress heterocyst differentiation.

View larger version (11K): [in this window] [in a new window]   FIG. 2.   Growth and heterocyst differentiation in JE994 (), FD (), and NF76 () cultures grown under anoxic, diazotrophic conditions. Cells grown with 5.0 mM NH4Cl-10 mM TES (pH 7.2)-5.0 mM fructose were washed with AA/8, resuspended in AA/8 with 5.0 mM fructose to an OD700 of <0.1, and incubated under anoxic conditions. Optical density (A) and heterocyst frequency (B) were determined for 5 days.

Effect of exogenous ammonium on heterocyst differentiation. Another possibility was that heterocyst differentiation was insensitive to the presence of external fixed nitrogen under anoxic conditions. Heterocyst differentiation in Anabaena cylindrica filaments grown with air was reversed by the addition of 4 mM ammonium chloride up to 8 to 10 h after induction (1). After that time certain cells were irreversibly committed to differentiation, and addition of fixed nitrogen did not repress heterocyst differentiation (1). In the experiments described here, strains FD (wild type) and JE994 (nif1 mutant) were induced under anoxic conditions, and ammonium was added to the culture at various times after induction. At 24 h after induction, the percentage of heterocysts was determined. The addition of exogenous ammonium up to about 8 h after induction prevented heterocyst formation (Fig. 3), indicating that heterocyst differentiation was repressed by external ammonium under anoxic conditions. Lower concentrations (down to about 0.5 mM) of exogenous ammonium similarly repressed heterocyst differentiation in short-term experiments (data not shown); however, because cyanobacteria actively transport ammonium with a K_s of about 3 µM (25), cells can rapidly deplete the available supply when the extracellular concentration is low (25, 32).

View larger version (18K): [in this window] [in a new window]   FIG. 3.   Effect of exogenous ammonium on heterocyst differentiation. Cells grown with 5.0 mM NH4Cl-10 mM TES (pH 7.2)-5.0 mM fructose were washed with AA/8, resuspended in AA/8 with 5.0 mM fructose, and incubated under anoxic conditions; 5.0 mM NH4Cl-10 mM TES (pH 7.2) was added to aliquots of each strain at the times indicated. Incubation was continued under anoxic conditions, and heterocyst frequency was determined after 24 h. The bars labeled "none" indicate the heterocyst frequency for the control culture to which no NH4Cl was added.

Presence of NifH1 and NifH2 proteins under anoxic conditions. Although wild-type strain FD produced heterocysts, the anoxic growth experiments did not reveal whether the Nif1 nitrogenase was produced. However, previous experiments using a strain with a nifH::lacZ fusion have shown that the nifH1 gene is transcribed under anoxic conditions (40). Fortuitously, NifH1 and NifH2 proteins have slightly different electrophoretic mobilities in an SDS-polyacrylamide gel, allowing identification of each protein by immunoblotting. Wild-type cells were induced under anoxic conditions, and cell samples were analyzed at 6 and 18 h after induction. The expression of NifH2 early after induction did not prevent the subsequent expression of NifH1 about the time that heterocysts first formed (Fig. 4A). Conversely, when wild-type cells were grown under oxic diazotrophic conditions and then shifted to anoxic conditions, NifH2 was made within 2 h after the shift (Fig. 4B). Although the amounts of each protein varied with time, these differences may not be significant. Normally, when cells are fixing well using only the Nif1 nitrogenase under oxic conditions, nitrogenase activity declines markedly during growth, even when cells are growing rapidly (data not shown). This is presumably because cells become replete with fixed nitrogen and nitrogenase activity is no longer needed. Thus, it is difficult to attribute the apparent decline in NifH1 protein at 6 h (Fig. 2B) to NifH2 activity. It is clear, however, that either the Nif1 nitrogenase alone or the Nif2 nitrogenase alone produced sufficient fixed nitrogen to support cell growth, but in neither case did the fixed nitrogen repress synthesis of the other nitrogenase.

View larger version (31K): [in this window] [in a new window]   FIG. 4.   Synthesis of NifH1 and NifH2 proteins in cells grown under anoxic conditions. (A) Wild-type strain FD was grown with 5.0 mM NH4Cl-10 mM TES (pH 7.2)-5.0 mM fructose and induced at time zero under anoxic conditions as described in Materials and Methods. Samples were removed at 6 and 18 h after induction, and NifH1 and NifH2 proteins were detected on immunoblots using anti-NifH antibodies. (B) Wild-type strain FD was grown in AA/8 with 5 mM fructose for 48 h to induce heterocysts. Cells were shifted to anoxic conditions at time zero, and NifH1 and NifH2 proteins were detected on immunoblots using anti-NifH antibodies at the times indicated.

In situ localization of nifH2 expression in filaments. Previous studies have demonstrated that nifH2 is expressed in vegetative cells after induction under anoxic conditions (40). One possible explanation for the expression of the Nif2 nitrogenase in filaments already fixing nitrogen using the Nif1 nitrogenase could be that the Nif2 nitrogenase is expressed only in vegetative cells far from existing heterocysts. These vegetative cells would be deficient in fixed nitrogen and thus would express the Nif2 nitrogenase. To test this possibility, strain JE35 (a nifH2::lacZ fusion [40]) was induced under oxic conditions for 48 h to allow heterocysts to form with expression of the Nif1 nitrogenase, and then the cells were incubated under anoxic conditions for 6 h. -Galactosidase activity was visualized in situ by the intracellular cleavage of the fluorogenic substrate C₁₂-FDG. Fluorescent cells expressing -galactosidase activity were identified by epifluorescence microscopy using a fluorescein filter set with a short-pass filter to block biliprotein fluorescence (40). The nifH2 gene was expressed only under anoxic conditions and predominantly in vegetative cells (Fig. 5). While not all vegetative cells were fluorescent, there was no regular pattern of spaced expression of nifH2. Under the exposure conditions of these experiments, the low level of endogenous fluorescence from cell pigments was virtually undetectable. After viewing several thousand filaments, we would describe the expression of nifH2 under these conditions as highly nonrandom, with many instances of blocks of brightly fluorescent cells of highly variable length followed by blocks of dimmer cells. Vegetative cells adjacent to heterocysts were as likely to express nifH2 as were cells far from an existing heterocyst. Thus, the pattern of spacing of existing heterocysts had no effect on expression of nifH2. In addition, most heterocysts showed little or no fluorescence, indicating that nifH2 is not highly expressed in heterocysts.

View larger version (111K): [in this window] [in a new window]   FIG. 5.   In situ expression of nif2. Strain JE35 (nif2::lacZ fusion) was grown in AA/8 with 5 mM fructose for 48 h to induce heterocysts (H) and Nif1 expression. Cells were shifted to anoxic conditions, and samples were removed after 4 h and incubated with C12-FDG as described in Materials and Methods. (A) Fluorescein fluorescence; (B) light micrograph. Bar = 10 µm.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

One of the most intriguing aspects of development in cyanobacteria is the pattern of heterocysts that is maintained throughout diazotrophic growth of the filament. The most obvious explanation for the maintenance of this pattern (but not for its initial formation) is that a nitrogenous product of heterocyst metabolism diffuses along the filament, inhibiting cells near existing heterocysts from differentiating (44, 45). Cells midway between heterocysts would first suffer from a deficiency of this nitrogenous product and would differentiate into a new heterocyst. This model requires that certain cells in the filament be nitrogen starved for differentiation to occur. Glutamine was once thought to be this nitrogenous product; however, we showed that although glutamine can support the nitrogen needs of the filament, nevertheless heterocysts differentiate (38). These heterocysts do not fix nitrogen, and so there is clearly no gradient of fixed nitrogen under these conditions. One problem of these and other earlier experiments that demonstrated heterocyst differentiation in the presence of fixed nitrogen was that the circumstances were unlikely to be replicated in the environment and hence could be considered anomalous.

We demonstrate here that expression of the Nif2 nitrogenase at levels sufficient to support good diazotrophic growth rates did not prevent either heterocyst differentiation or expression of the heterocyst-specific Nif1 nitrogenase. In this case there was no addition of unusual compounds or mutations in genes that may have had pleiotropic effects. It is difficult to reconcile the results presented here as well as results of many previous studies that have shown that spaced heterocysts can form in the absence of nitrogen fixation (6, 8, 33, 38, 47) with a model that requires a gradient of any metabolite of nitrogen fixation. Further, while it has generally been accepted that starvation for fixed nitrogen in a cell is the major trigger for heterocyst differentiation, it is unlikely that cells producing sufficient fixed nitrogen for growth, using the Nif2 system under anoxic conditions, would be nitrogen starved. Yet they differentiated heterocysts as if they were starved. In addition, those heterocysts differentiate in a normal pattern. How could only certain cells in a spaced pattern in the filament be starved for fixed nitrogen when the Nif2 nitrogenase is expressed in all vegetative cells? The most reasonable conclusions to be drawn from this and previous work are that (i) no product of nitrogen fixation controls heterocyst pattern formation, (ii) pattern formation does not depend on a gradient of fixed nitrogen that diffuses along the filament, and (iii) starvation for fixed nitrogen is not a prerequisite for heterocyst differentiation. While we recognize that there could be a complicated physiological explanation invoking different pathways for metabolism of fixed nitrogen in vegetative cells versus heterocysts, our conclusions are the simplest interpretation of our data as well as much other data over the years that has demonstrated a spaced pattern of heterocysts in the absence of normal nitrogen fixation.

One possible explanation for these results is that fixed nitrogen entering the cell from the environment is recognized differently from fixed nitrogen that is produced within cells. If exogenous ammonium is added to filaments that are fixing nitrogen using the Nif2 nitrogenase (under anoxic conditions), heterocyst differentiation is repressed (Fig. 3) just as it is under oxic growth conditions (1); however, the intracellular ammonium produced by the Nif2 nitrogenase in all cells under these conditions had no such effect. The cells in the filament may have a mechanism that distinguishes between external and internal sources of fixed nitrogen, and it may be that only externally derived fixed nitrogen represses differentiation.

How the cells sense the difference in external versus internal fixed nitrogen is not clear; however, it is important that they do so. Nitrogen fixation is metabolically expensive, requiring both ATP and reductant. Any ammonium available from the environment is "free," and it is clearly disadvantageous to the organism to fix nitrogen. However, nitrogen fixation by the organism is, in itself, an indicator of a state of nutritional deprivation. Under such conditions it is advantageous to the organism to have both systems for nitrogen fixation expressed if the proteins can function. There is no danger of wasting energy by fixing too much nitrogen, since nitrogenase activity and expression of the genes are modulated (13, 26). Thus, only exogenous sources of fixed nitrogen are perceived (correctly) by the organism as nitrogen sufficiency. In the absence of sufficiency, it is to the advantage of the organism to fix nitrogen to the best of its ability.

This idea that the filament behaves as an organism requires a different approach to understanding heterocyst differentiation, which is currently based on a model that assumes that filaments comprise connected but fundamentally independent cells. Models for heterocyst differentiation have assumed that an individual cell in the filament senses its own intracellular environment and responds metabolically or developmentally to that environment. Communication, such as it exists, is thought to be via metabolites whose concentrations differ along the filament, leading to differences in the intracellular environment. If the Nif2 nitrogenase provides fixed nitrogen to all vegetative cells in a filament, it is difficult to envision gradients of a fixed nitrogen product that would lead to patterned heterocyst differentiation. Yet such differentiation clearly occurs under anoxic growth conditions both for de novo heterocyst formation and for maintenance of the heterocyst pattern during growth of the filament. Perhaps the cells in a filament behave more as part of an organism than as individuals. That is, the periplasm of the filament senses a common external environment and responds metabolically in a manner that benefits the whole filament. In this case, the filament would sense an external environment devoid of fixed nitrogen and the metabolic response of the filament would be to express all genes that would help the filament to survive, i.e., heterocyst genes and nitrogenase genes. The nutritional state of individual cells would have little impact on the response of the filament. While it is clearly conjecture, this model helps to explain not only the data presented here but also the fact that de novo heterocyst differentiation after cells are first deprived of fixed nitrogen occurs in a pattern. The cells in these filaments are responding to the external environment, without any apparent source of nutrients to help establish an intracellular metabolic gradient that could lead to patterned differentiation. Similarly, vegetative cells in filaments grown under anoxic conditions that are fixing nitrogen using the Nif2 nitrogenase differentiate heterocysts in a pattern that is a response to the external environment, not to the nutritional state of individual cells. If heterocyst differentiation is a response of the filament to the environment, then it is reasonable that certain cells in an undifferentiated filament are predestined to become heterocysts. This model does not preclude cell-signaling molecules (such as PatS [47]); in fact, if the filament behaves as an organism in its response to the environment, then there must be cell-to-cell communication.