This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Meeks, J. C. Articles by Elhai, J. Search for Related Content PubMed PubMed Citation Articles by Meeks, J. C. Articles by Elhai, J.

Regulation of Cellular Differentiation in Filamentous Cyanobacteria in Free-Living and Plant-Associated Symbiotic Growth States

Section of Microbiology, University of California, Davis, California 95616¹, and ,¹ Department of Biology, Virginia Commonwealth University, Richmond, Virginia 23284²²

SUMMARY

INTRODUCTION

Symbiosis of Cyanobacteria with Plants

What makes cyanobacteria valuable partners in symbioses?

What do plants bring to the symbiosis?

How specific are associations between plants and cyanobacteria?

DEVELOPMENTAL REPERTOIRE OF FREE-LIVING CYANOBACTERIA

Patterned Differentiation to Heterocysts, Well-Spaced Nitrogen-Fixing Cells

What are heterocysts, and how do they work?

What are the regulatory circuits in heterocyst differentiation?

(i) Earliest signals in heterocyst differentiation.

(ii) Signals specific to early differentiating cells.

(iii) Signals specific to mature differentiated cells.

How is the spacing of heterocysts determined?

(i) Characteristics of the one-stage model.

(ii) Multiple contiguous heterocysts.

(iii) Apparent inconsistencies with the one-stage model.

(iv) Two-stage model of pattern formation.

Global Differentiation to Motile Filaments: the Hormogonium Cycle

What are hormogonia, and what is their role in the cyanobacterial life cycle?

What induces the differentiation of hormogonia?

(i) Hormogonium differentiation in response to changes in the chemical environment.

(ii) Hormogonium differentiation in response to changes in light quality.

(iii) Differential gene expression in the hormogonium cycle.

PLANT CONTROL OVER THE INFECTION PROCESS

Plant Induction of Hormogonium Differentiation

Behavioral Characteristics of Hormogonia during Infection

Plant Regulation of the Hormogonium Cycle

PLANT CONTROL OVER THE CYANOBACTERIAL SYMBIOTIC GROWTH STATE

Physiological Characteristics of Cyanobacteria in the Symbiotic Growth State

Do plants control the growth of symbiotically associated Nostoc?

What causes impaired photosynthesis in symbiotically associated Nostoc?

What are the causes and consequences of uncoupled nitrogen fixation and ammonium assimilation in symbiotically associated Nostoc?

(i) Sources of reductant for symbiotic nitrogen fixation.

(ii) Routes of carbon catabolism by symbiotic Nostoc.

(iii) Regulation of ammonium assimilation in symbiotic Nostoc.

Morphological Characteristics of Cyanobacteria in the Symbiotic Growth State

Morphological alteration of vegetative cells.

What are the heterocyst frequencies of symbiotically associated Nostoc?

How Plants Overcome the Normal Controls over Heterocyst Differentiation

Does the high frequency of heterocysts in plant-associated cyanobacteria result from an increased frequency of initiation at discrete sites or a lack of resolution of multiple contiguous heterocysts?

What is the signal(s) initiating heterocyst differentiation in the symbiotic growth state?

CONCLUSIONS AND FUTURE DIRECTIONS

ACKNOWLEDGMENTS

REFERENCES

Top Next References

INTRODUCTION

Top Previous Next References

 
The word "bacteria" generally conjures up images of individual rods or spheres swimming around, every bug for itself. In fact, however, the solitary life is rare in the microbial world; bacteria interact with other microbes and macroorganisms in the environment and with their own siblings, and these interactions often result in complex social interactions. Many bacterial species exchange signals amongst themselves in order to assess cell density and to respond appropriately (62, 71). Myxobacteria take signaling a step further and use it to direct their differentiation into complex multicellular structures, enabling them to survive periods of starvation (47). Rhizobia communicate outside of their species. They respond to signals from leguminous plants (e.g., beans) to differentiate into nitrogen-fixing cells. At the same time, they produce signals that induce the plants to form specialized nodules, microoxic cavities required for the O₂-sensitive process of N₂ fixation (198).

Like myxobacteria, certain N₂-fixing species of filamentous cyanobacteria generate signals to direct their own multicellular development, and, like rhizobia, they respond to signals from plants, initiating or altering the extent of their cellular differentiation. A main objective of this review is to consider the degree to which signals used by cyanobacteria to control their own differentiation are appropriated by plants to achieve stable symbioses. General reviews on the biology of symbioses between cyanobacteria and plants have recently appeared (3, 151).

View larger version (85K): [in this window] [in a new window]   FIG. 1. Photomicrographs of vegetative and heterocyst-containing filaments of the cyanobacterium N. punctiforme. (A) Phase-contrast image of vegetative filaments grown in the presence of ammonium. No heterocysts are visible. (B) Phase-contrast image of filaments grown in the absence of any combined nitrogen source in the culture medium. Heterocysts, identified by arrowheads, are present at well-spaced intervals. (C) Epifluorescence image of the same filaments as in panel B. Excitation was at 510 to 560 nm (green), exciting phycoerythrin, and emission was greater than 600 nm. Heterocysts have negligible fluorescence, while vegetative cells have intense combined fluorescence from phycobiliproteins and chlorophyll a. Bar, 10 µm.

What do plants bring to the symbiosis? The self-sufficiency of heterocyst-forming cyanobacteria with respect to N₂ fixation may account for the relatively simple plant structures that house symbiotic cyanobacteria, in contrast to those that house rhizobia. Infecting rhizobia induce plants to differentiate nodules, which have been likened in their complexity to new plant organs (201). Nodules differ from surrounding root tissue in several regards, many of them related to the control of O₂ tension. One accommodation is the production of plant- and bacterium-encoded leghemoglobin, a protein that binds O₂, keeping it at a concentration low enough to permit N₂ fixation but high enough to permit rhizobial respiration.

In contrast, the plant structures inhabited by cyanobacteria exist even in the absence of the cyanobacterial symbiont, at sites that differ from plant to plant (Fig. 2). Gunnera possesses a unique gland on its stem that attracts cyanobacteria. Cycads have specialized lateral roots (coralloid roots) in which a layer has been prepared for infecting cyanobacteria. Bryophytes come with cavities in the gametophyte tissue that are infected by cyanobacteria. Cyanobacteria associated with Azolla in cavities of the dorsal leaves are unique in that they appear to be obligate symbionts (141), but Azolla cured of cyanobacteria by treatment with antibiotics continues to grow in the presence of combined nitrogen with empty leaf cavities (60). These different plant structures will be referred to collectively as symbiotic cavities, and the cyanobacterial populations within them will be referred to as symbiotic colonies.

View larger version (67K): [in this window] [in a new window]   FIG. 2. Photographs of plant partners showing the location of symbiotic cavities housing associated cyanobacteria. (A and B) Sporophyte (S) and gametophyte (G) generations of the bryophyte Anthoceros punctatus (A) and Nostoc colonies (nc) (B) within the gametophyte tissue. Bar, 1.0 cm. (C and D) Azolla caroliniana floating sporophyte leaves and trailing submerged roots (C). The cavity of a dorsal leaf housing symbiotic Nostoc/Anabaena is shown in a vertical position. A, Nostoc/Anabaena filaments; hc, Azolla hair cells (D). Photographs copyright Gerald Peters. Panel D reprinted from reference 142 with permission of the author. Bar, 0.25 cm and 0.25 mm in panels C and D, respectively. (E and F) Stem and leaves of the cycad Cycas taiwaniana (E) and ventral view and cross section of a coralloid root cluster from C. taiwaniana (F). Nostoc is present in a green annular ring (nc) within the root. Bar, 0.5 m and 0.5 cm in panels E and F, respectively. (G and H) Seedling of stoloniferous Gunnera manicata showing the location of the stem gland (sg) (G) and tangential stem section of a giant Gunnera chilensis showing location of the Nostoc colonies (nc) (H). Photographs copyright Warwick Sylvester. Panel G reprinted from reference 22 with permission of the publisher. Bar, 1.0 cm.

It is a matter of controversy whether all symbiotic cavities, like the nodules of legumes, have O₂ tensions much lower than ambient. Microelectrode measurements have indicated relatively high O₂ tension within coralloid roots of cycads (35), slightly lower than ambient (83%) in symbiotic cavities of Azolla (68), and very low (near to anoxic) in the bryophyte hornwort Anthoceros (34). No plant-derived hemoglobin-like protein has been reported. Cyanoglobin, a heme protein with high affinity for oxygen, is found in many Nostoc strains, but some strains, including a few symbiotic strains, lack the protein (79). Free-living cyanobacteria can maintain high nitrogenase activities while growing under atmospheres ranging from 1 to 50% O₂ (131).

How specific are associations between plants and cyanobacteria? Strains or species of the genus Nostoc are the dominant but not exclusive symbiotic cyanobacteria in plant associations. The relatively low requirements of plant-cyanobacterium symbioses are illustrated by the lack of strain specificity between the partners in reconstitution experiments (22, 49, 85). Of 10 strains of Nostoc isolated from symbioses with Gunnera, 6 achieved an intracellular association with a different species of Gunnera. Of four strains of Nostoc isolated from symbioses with cycads, Anthoceros, or Peltigera (a lichenized fungus), three successfully infected Gunnera. Only one of nine Nostoc strains isolated from various plant or lichen symbioses failed to infect Anthoceros. Free-living isolates have generally not established symbioses, although exceptions have been noted (49, 135). DNA-fingerprinting techniques revealed a wide cluster of symbiotic and free-living isolates from a single field site associated with the hornwort Phaeoceros (202). The isolates included primarily 65 Nostoc strains, but three each of Calothrix and Chlorogloeopsis were also identified illustrating some flexibility in cyanobacterial partners. Except for one Nostoc isolate, all reconstituted a symbiotic association with Phaeoceros or the liverwort Blasia.

DEVELOPMENTAL REPERTOIRE OF FREE-LIVING CYANOBACTERIA

Top Previous Next References

 
The relative lack of discrimination by plants with regard to their Nostoc partners indicates that the properties required for a wide variety of associations are found in most strains. It is difficult to imagine that cyanobacteria, unlike rhizobia, have evolved to exchange signals with multiple partners spanning nearly the entire plant kingdom. The alternative is that different plants have independently learned how to exploit the capabilities common to free-living Nostoc. In all resulting symbioses, plants influence two developmental responses of Nostoc—the formation of hormogonia and the differentiation of heterocysts. Vegetative cells of Nostoc have a third developmental alternative in the differentiation into perennating spore-like cells called akinetes (6). Akinetes are hypothesized to be evolutionary precursors of heterocysts (214). Heterocysts and akinetes may (99) or may not (217) have common developmental regulatory elements. Akinetes are present in large numbers in older Azolla leaves and in the developing sporocarp, where they presumably play a role in generational continuity (141). However, akinetes have not consistently been observed in all symbioses, and there have been no studies on their symbiotically associated differentiation or germination.

In this section, information from studies of free-living cyanobacterial strains on heterocyst and hormogonium developmental responses is considered. In subsequent sections, we will examine possible mechanisms by which plants may draw on the free-living signal transduction systems to assert symbiotic control.

What are heterocysts, and how do they work? The major function of heterocysts is to provide a microoxic environment necessary for the production and proper functioning of nitrogenase and other proteins related to N₂ fixation (214). In essentially all heterocyst-forming cyanobacteria, heterocysts appear to be the only sites for N₂ fixation within the filament (48, 188). The exception is Anabaena variabilis strain ATCC 29413 (and a few closely related strains), which expresses a vegetative cell-specific nitrogenase under anoxic incubation conditions (188). The heterocyst achieves a near anoxic state by at least three means. First, photosystem II, the O₂-producing end of the photosynthetic electron transport chain, is dismantled during heterocyst differentiation (186), so that the heterocyst need contend only against O₂ produced by neighboring vegetative cells and that dissolved in the environment. Second, heterocysts are invested with a specialized envelope (214) that limits the influx of gases (199). Two layers within the envelope have been implicated in O₂ protection (130): an inner layer composed of a hydroxylated glycolipid and an outer layer of polysaccharide. Neither layer is found in vegetative cells. Third, much of the O₂ that overcomes these barriers is consumed by the high oxidase activity associated with heterocysts (144).

Abandoning complete photosynthesis allows the recycling of the amino acids contained in phycobiliproteins, light-harvesting pigments associated primarily with photosystem II that may account for up to 50% of the soluble protein within the cyanobacterial cell (21). Nitrogen deprivation elicits a general increase in proteolysis and degradation of phycobiliproteins (187, 218). At the onset of nitrogenase activity, phycobiliproteins return to their original levels in vegetative cells (23). The new heterocysts themselves, however, are grossly deficient in fluorescence associated with phycobiliproteins (197), as shown in Fig. 1C. After several days, heterocysts in some strains may regain this fluorescence (145, 190), but filaments tend to liberate aged heterocysts as single fluorescent cells. Loss of fluorescence in specific cells can thus be used as a diagnosis of differentiation; even in mutants that cannot complete differentiation, a pattern of well-spaced nonfluorescent cells is indicative of at least initiation of differentiation (48).

In the absence of complete photosynthesis, the heterocyst becomes dependent on adjacent vegetative cells for reduced carbon, just as the vegetative cells are dependent on heterocysts for reduced nitrogen. Nitrogen fixed within the heterocyst as ammonium is first converted to glutamine and then passes as amino acids to adjacent vegetative cells (191). In return, fixed carbon, probably sucrose (166), flows from vegetative cells to heterocysts (208). Sucrose synthase, or perhaps sucrose phosphate synthase (146), is present in high activity specifically in vegetative cells, while alkaline invertase, which converts sucrose to glucose and fructose, is confined largely to heterocysts (166). The resulting hexoses are catabolized in the heterocyst via the oxidative pentose phosphate cycle, providing reductant for nitrogenase and respiratory activities (181).

Heterocyst-forming cyanobacteria have made a number of adaptations in their nitrogen metabolism, which are important in our understanding of the basis of heterocyst differentiation and interactions with vegetative cells (Fig. 3). Like other cyanobacteria and, so far as we know, unlike any other organism on Earth, heterocyst-forming strains store amino nitrogen in a nonprotein polymeric form, as a branched polypeptide called cyanophycin (10, 111). Cyanophycin is a copolymer of aspartate and arginine and is present in both vegetative cells and mature heterocysts. Just as phycobiliproteins may serve as a source of amino nitrogen during differentiation, so may cyanophycin.

View larger version (34K): [in this window] [in a new window]   FIG. 3. Metabolic interactions between heterocysts and vegetative cells. A lighter vegetative cell exchanges metabolites (thin lines) with a darker heterocyst bound by its characteristic thick envelope. The heterocyst has polar plugs at either end. Thick lines indicate metabolic pathways. The dotted line indicates a pathway whose existence is uncertain. Carbon dioxide is fixed in vegetative cells through the dark reactions of photosynthesis (PS), and the resulting triose is metabolized to pyruvate through the partial tricarboxylic acid cycle to isocitrate and then via IDH to -ketoglutarate (KG). The -ketoglutarate combines with glutamine (Gln) via glutamate synthase (GOGAT) to form glutamate (Glu). In heterocysts, carbohydrate from vegetative cells enters the oxidative pentose phosphate (OPP) pathway to produce reductant (H:) used to support the activity of nitrogenase (Nase) to produce ammonium and concurrently yield -ketoglutarate. Ammonium combines with glutamate, derived from the vegetative cell, through a reaction catalyzed by GS to form Gln. These reactions (if confirmed) should serve to reduce the level of -ketoglutarate in the vegetative cell (small type) and increase it in heterocysts (large type). See the text for a discussion of the proposed cellular levels of -ketoglutarate.

Like glutamine, -ketoglutarate may be made primarily in heterocysts; under N₂-fixing conditions, in vitro activity of isocitrate dehydrogenase (IDH) appears to be largely absent from vegetative cell extracts (136), although, based on immunological assays, the IDH protein is present in equal amounts in vegetative cells and heterocysts (113). IDH is highly active in heterocyst extracts, which appears anomalous, since the primary metabolic role of -ketoglutarate is to serve as precursor to glutamate (-ketoglutarate dehydrogenase is absent in cyanobacteria, which thus lack a complete citric acid cycle [170]). However, IDH-generated NADPH may also contribute to the reductant pool for nitrogenase and respiration in heterocysts (109). In brief, glutamine is made in heterocysts, using glutamate transferred from vegetative cells, derived from -ketoglutarate, which may arise in heterocysts. A possible rationale for this state of affairs in which -ketoglutarate also serves as a signaling molecule will be offered in a subsequent section.

What are the regulatory circuits in heterocyst differentiation? Much of what we know about heterocyst differentiation comes from the study of the nonsymbiotic laboratory strain Anabaena sp. strain PCC 7120. Based on DNA-DNA hybridization and sequence analysis of DNA encoding small-subunit rRNA, this strain is closer to those in the genus Nostoc (160), but since the name is entrenched in the literature we will continue to refer to it as Anabaena strain PCC 7120. Within one or two generations (about 18 to 36 h under standard conditions), well-spaced vegetative cells suddenly deprived of a source of combined nitrogen differentiate into N₂-fixing heterocysts. The program of development begins with the sensation of nitrogen deprivation and culminates in the expression of the N₂ fixation apparatus in the mature heterocyst. In between, there is an ordered sequence of events controlled to a great degree at the level of transcription.

Heterocyst differentiation is inhibited by the presence of a usable source of nitrogen, such as NH₄⁺ or NO₃^-. Even 3 to 7 µM NH₄⁺ is sufficient to repress the appearance of heterocysts in Anabaena strain PCC 7120 (122). Since heterocysts persist in Nostoc within symbiotic cavities despite a level of ammonium that is much higher than this (35, 36, 118, 120, 169), it is imperative to understand the basis of the repression. The following discussion will not detail all genes now known to influence heterocyst formation and function; a more thorough listing can be found in references 211 and 212.

(i) Earliest signals in heterocyst differentiation. In Escherichia coli and other proteobacteria, nitrogen starvation is signaled in part by the glnB-encoded protein P_II, which interacts with GlnE to modify GS activity by adenylylation and interacts with NtrB to modify the regulatory activity of NtrC by phosphorylation (124). NtrC-P enhances the transcription of several genes that respond to nitrogen conditions, including glnA, the structural gene of GS. The metabolic status of the cell is sensed directly by P_II through its binding to -ketoglutarate and ATP and through a uridylyltransferase (encoded by glnD) that modifies P_II in response to the levels of glutamine.

P_II is the only component of this regulatory web that has been found in cyanobacteria (65, 70, 108, 192). In unicellular cyanobacteria, the P_II protein is modified by phosphorylation, and not uridylylation as in E. coli (58), in response to the level of -ketoglutarate (83). Unmodified P_II appears to regulate nitrate assimilation in unicellular Synechococcus sp. strain PCC 7942 (96). While a mutant strain of Synechococcus strain PCC 7942 in which glnB has been insertionally inactivated is minimally affected in nitrogen-related functions (59), it has proven impossible to obtain a similar mutant in N. punctiforme, implying that the glnB gene is essential in heterocyst-forming cyanobacteria (70). It should be noted that cyanobacteria, unlike many bacteria, possess only one GlnB-like protein (11) (see http://www.jgi.doe.gov/ and http://www.kuzusa.or.jp/cyano/).

The earliest documented step in the response of cyanobacteria to nitrogen deprivation is activation of the DNA binding protein NtcA (77, 110). Mutants lacking functional ntcA are unable to respond to nitrogen deprivation; they fail to increase levels of GS, to respond to alternate nitrogen sources (such as NO₃^-), to make heterocysts, or to induce nitrogenase expression (61, 200). Many genes whose transcription is altered by nitrogen deprivation (61) are preceded by a consensus sequence (GTAnCaannnnTAC in Anabaena strain PCC 7120 [77]) to which NtcA binds (153). One such gene is the NtcA-regulated gene glnB from the non-heterocyst-forming cyanobacterium Synechococcus strain PCC 7942 (97). Transcription of glnB from N. punctiforme is also induced by nitrogen limitation (70). NtcA binding activity is present in both vegetative cells and heterocysts (153). Although the transcription of ntcA is subject to variation over the course of heterocyst differentiation (200), perhaps owing to a NtcA binding site 5" to the gene (154), the means by which NtcA controls heterocyst differentiation remains unknown.

An unresolved question is the mechanism by which NtcA directly or indirectly senses nitrogen status or, even more generally, how nitrogen status is translated into the decision of a cell to differentiate. It would be convenient if the known ability of P_II to sense levels of -ketoglutarate could account for that decision, but if so, the role of the protein in heterocyst-forming cyanobacteria would be significantly different from that in unicellular strains. The regulation of transcription of nitrogen-regulated genes in unicellular Synechococcus strain PCC 7942 is little affected by the absence or modification of P_II (59), although the state of the protein is critical in mediating the repression of nitrate/nitrite transport by ammonium (96). The nitrogen status in cyanobacteria thus appears to be sensed in two different ways: by P_II, leading to the alteration of enzyme activity, and (perhaps indirectly) by NtcA, leading to the alteration of gene transcription.

The ability of ammonium to suppress heterocyst differentiation is widely believed to be indirect, requiring the activity of GS. When ammonium assimilation in A. cylindrica (164) and other cyanobacteria (210) is blocked by the presence of methionine-DL-sulfoximine, an inhibitor of GS, heterocysts differentiate in the typical spacing pattern in the presence of ammonium. This result implies that the suppressor signal must be a consequence of ammonium assimilation, such as increased glutamine or decreased -ketoglutarate levels. This conclusion has been questioned as the result of experiments exploiting the ability of A. variabilis to express an alternative nitrogenase in vegetative cells under anoxic growth conditions (189). Under these conditions, this strain can grow on N₂, even when the heterocyst-specific nitrogenase is inactivated by mutation. Despite the production of ammonia by the alternative nitrogenase, expressed uniformly in all vegetative cells, heterocysts persist and with the usual spacing. The normal appearance of heterocysts has been taken as evidence that exogenous ammonium is metabolized differently from that produced by endogenous nitrogenase and that only the former inhibits differentiation (189). Alternatively, the minimal level of fixed nitrogen required to suppress differentiation may be markedly higher than the minimal level required to support growth. This is a reasonable supposition, since it is consistent with the behavior of A. cylindrica in more natural circumstances, where the cyanobacterium responds to gradually diminishing supplies of fixed nitrogen well before it is exhausted (111).

Fragmentary evidence points to a crucial role of DNA replication early in heterocyst differentiation. The DNA-damaging agent mitomycin C (5) and the DNA gyrase inhibitor nalidixic acid (M. Gantar and J. Elhai, unpublished results) block heterocyst differentiation when individually applied during the first hour after nitrogen deprivation. Mitomycin C at the effective level has no discernible effect on new RNA synthesis. A mutant of Anabaena strain PCC 7120 defective in a gene, hanA, which encodes a histone-like protein (HU) is blocked in the earliest known stage of heterocyst differentiation (89). HU protein from Anabaena strain PCC 7120 is able to substitute for E. coli HU protein to promote the initiation of DNA synthesis in vitro (132). hanA is preceded by a NtcA binding site. HU protein is by far the most abundant DNA binding protein in vegetative cells, based on autoradiography of acrylamide gels, The vegetative cell band is replaced by a slightly larger DNA binding protein on gels of heterocyst proteins (132), but the protein has not yet been characterized.

Heterocyst differentiation is also blocked by ethionine (93), an analogue of methionine. At the micromolar levels that are effective, ethionine has no effect on bulk protein, DNA, or RNA synthesis or on growth, but it reportedly prevents the rise in hetR expression (P. S. Duggan and D. G. Adams, personal communication) that normally accompanies the earliest stage of differentiation. Ethionine may act by affecting S-adenosylmethionine-dependent methylation of DNA. Such a mechanism has been demonstrated to explain the induction by ethionine of spores of Bacillus subtilis (9).

(ii) Signals specific to early differentiating cells. Several genes have been found that are required for nitrogen deprivation to evoke any visible sign of heterocyst differentiation. One of these, hetR, clearly plays a central role in the decision of a nitrogen-starved cell to differentiate. Mutants with mutations in hetR, unlike those with mutations in ntcA, are able to grow normally in NO₃^- (unaffected in a general response to nitrogen deprivation), but they do not differentiate. On the other hand, strains in which wild-type hetR is present on a multicopy plasmid (94) form heterocysts even in the presence of NO₃^- (27), and the fraction of cells that become heterocysts when hetR is overexpressed in the absence of combined nitrogen can reach 29% (28).

The expression of hetR also is consistent with a role for the protein as an important molecular switch. Transcription of hetR, which is low in vegetative cells, begins to increase as little as 0.5 h after nitrogen deprivation (20, 27, 28, 216). By 3.5 h after deprivation, expression (P_hetR::luxAB) is confined largely to a small fraction of cells, with a spacing in the filaments similar to that of heterocysts, even though morphological differentiation is not evident until many hours later. Presuming that all the increase in hetR expression is confined to cells with high expression, the increase per cell can be estimated to be 20-fold, relative to expression under nitrogen-replete conditions (20). The starvation-dependent increase requires functional ntcA (61, 216) and hanA (89). The increase also requires the function of HetR, positively regulating its own expression (20). A positive autoregulatory switch, which magnifies small differences in expression and locks expression into one of two states, is known to stabilize many developmental decisions: lysogeny by coliphage lambda (147), cell type determination by Bacillus (92, 182), and commitment to different developmental fates by eukaryotic cells (7, 123).

The predicted amino acid sequence of HetR gives no clues to its function. For example, it possesses none of the sequence motifs typical of DNA binding proteins (27). Only one function of the protein, autoproteolytic activity, has been detected, evidently stimulated in vivo by the presence of a nitrogen source (223). The active site for proteolysis is required for heterocyst differentiation although not for the HetR-dependent induction of hetR (44). The protein appears to be modified in vivo, since HetR isolated under nitrogen-depleted conditions is considerably more acidic than the protein isolated from nitrogen-replete cells or than the recombinant protein isolated from E. coli (222).

Two genes, hetF and patA, appear to positively influence heterocyst differentiation by modulating the expression or activity of HetR. The hetF gene, best characterized in N. punctiforme (216), encodes a constitutively expressed protein that is essential for heterocyst differentiation; hetF mutants do not initiate even the earliest stages of heterocyst differentiation. The absence of HetF does not alter NtcA-dependent hetR transcription, but its absence eliminates HetR-dependent hetR transcription and the subsequent accumulation of HetR in differentiating cells. HetF bears no obvious similarity to any known protein. A second protein, PatA, has multiple domains, the carboxyl-terminal of which has similarity to the receiver domain of response regulators (100). Mutants defective in patA form heterocysts only at the ends of filaments and are consequently unable to grow well on N₂ (100). The multiple contiguous heterocyst phenotype of strains overexpressing hetR is suppressed in patA mutants (28). These results imply that PatA is in the same regulatory circuit as HetR and may function downstream of hetR transcription to modulate HetR activity. Since HetR appears to be modified to a more acidic form in nitrogen-limited cultures, it is tempting to speculate that PatA is part of a phosphorelay signal transduction pathway that activates HetR by phosphorylation (28).

Another gene, hetC, encodes a product important in early heterocyst differentiation. Mutants insertionally inactivated in hetC fail to show any morphological signs of differentiation, but, as in wild-type filaments, mutant filaments lose fluorescent pigments in cells spaced as one would expect of differentiating heterocysts (90). Like hetR, hetC is induced early in the course of nitrogen starvation, is dependent on NtcA (128), and requires its own function for full expression.

HetC shows considerable sequence similarity to members of the superfamily of ATP binding cassette (ABC) proteins (90), which couple the cleavage of ATP to the transport of molecules across membranes (51). This similarity, the propensity of hetC mutants to exhibit pairs of nonfluorescent (partially differentiated) cells, and the confinement of expression of hetC to proheterocysts raise the possibility that HetC participates not in the differentiation of heterocysts but, rather, in the export of an inhibitor of differentiation (90).

Other heterocyst-specific genes have been identified that appear to play a role in later heterocyst differentiation (53, 211, 212). Most, like hepA (formerly hetA [82]) and devBCA (54, 112) are needed in the formation of the heterocyst-specific envelope, and so it is not surprising that they are expressed specifically in well-spaced and presumably differentiating cells (112, 213). The epistatic relationships between the genes required for heterocyst differentiation are just beginning to be worked out (28, 29).

The diversity in times after nitrogen step-down at which genes become active implies that there may be a hierarchy of transcriptional regulators, each controlling the expression of a set of genes at a characteristic time. Such ideas are inspired by the interlocking tiers of regulators and their modifiers that govern differentiation in B. subtilis (180) and other bacteria (161).

Much attention has been given to the role of sigma factors, modular subunits of RNA polymerase that determine promoter specificity in bacteria, in defining the course of differentiation. While alternative sigma factors have been identified in filamentous cyanobacteria (25, 32), none have so far been found to markedly influence the course of heterocyst differentiation. In fact, of the eight group 2 sigma factors that can be readily discerned from the sequence of Anabaena strain PCC 7120, all except the essential main sigma factor (24) are dispensable for heterocyst differentiation (88a; I. Khudyakov and J. Golden, personal communication).

Nonetheless, it is very likely that some such differentiation-specific sigma factors or other proteins regulating transcription initiation exist. Several genes are now known that are expressed in both vegetative cells and heterocysts. In such cases, transcriptional start sites have proven to be different under nitrogen-replete and N₂ growth conditions (24, 55, 193). Sites active in nitrogen-replete cultures are preceded by sequences similar to the promoter sequence recognized by the principal E. coli sigma factor, while sites active under N₂ growth conditions are preceded by quite different sequences (41). The principal sigma factor, attached to RNA polymerase from Anabaena strain PCC 7120, binds and transcribes only at sites presumptively active in vegetative cells (152, 167). Taken together, these results are consistent with a few possibilities: (i) combinations of redundant sigma factors are required for differentiation-specific transcription; (ii) Anabaena strain PCC 7120 possesses sigma factors essential for differentiation that cannot be recognized by their sequences; or (iii) proteins distinct from canonical sigma factors significantly alter the specificity of RNA polymerase during differentiation. Regardless of which possibility, if any, is true, an increased understanding of transcriptional regulation during heterocyst differentiation promises to teach us something new about how bacteria regulate their behavior.

(iii) Signals specific to mature differentiated cells. A large suite of genes begins to be expressed at about the same time after the emergence of immature heterocysts (proheterocysts), an event that takes place at about 12 h after the nitrogen step-down. Some of the new gene expression presumably relates to the increase in respiratory capacity that occurs in proheterocysts at about this time. The resulting drop in the intracellular O₂ concentration is required for expression of genes encoding nitrogenase (73), but low O₂ content and nitrogen deprivation are not sufficient for nitrogenase expression (48). Late gene expression is accompanied by at least three DNA rearrangements (72), removing insertions within genes encoding nitrogenase (67); a heterocyst-specific electron donor, ferredoxin (66); and a hydrogenase (37, 114). While ammonium has been reported several times to inhibit nitrogenase activity in heterocysts, this effect appears to be indirect, at least in A. variabilis ATCC 29413, acting by alteration of oxygen levels (50). At present, it is not clear whether nitrogenase synthesis and activity within mature heterocysts in general is directly repressed by nitrogen in the environment.

How is the spacing of heterocysts determined? From the perspective of a developmental biologist, the most striking behavior of Anabaena and Nostoc is the elaboration of spaced heterocysts at nearly regular intervals along their filaments (Fig. 1B). The nonrandomness of heterocyst spacing in wild-type Nostoc is obvious by inspection, but for purposes of comparison, it is important to have a clear picture in mind of what a truly random distribution of heterocysts would look like. Figure 4A shows a random distribution of heterocysts, calculated on the assumption that each cell differentiates, or not, independently of other cells in the filament.

View larger version (31K): [in this window] [in a new window]   FIG. 4. Distribution of randomly spaced heterocysts. (A) Calculated distribution of heterocysts if each cell differentiates with a probability of 0.1 (inspired by references 207 and 209). The graph shows the fraction of intervals with a given number of vegetative cells intervening between two heterocysts, according to the formula Pn = a(1 - a)n, where n represents the number of intervening vegetative cells and a is the probability that a specific cell differentiates. The average heterocyst spacing is (1 - a)/a, which in this case is one heterocyst every 9 vegetative cells. (B) Graphical derivation of the formula that predicts the distribution of heterocysts given random differentiation.

In fact, the peakless function predicted by random differentiation bears little resemblance to the actual distribution of intervals (Fig. 5A), and so heterocyst spacing is demonstrably nonrandom. The bias toward even-numbered intervals is expected if cells destined to vegetative cells status, but not those marked for differentiation, divide precisely once prior to the time of observation. A model that accounts for all spatial influences on heterocyst differentiation should be able to predict the distribution in Fig. 5A. Furthermore, if all the influences comprising the model can be removed (e.g. by mutation), the model predicts that heterocyst spacing should then be random, i.e., as shown in Fig. 4A.

View larger version (18K): [in this window] [in a new window]   FIG. 5. Spacing of heterocysts. Distances between heterocysts, measured as the number of intervening vegetative cells, are given, and their relative frequencies by actual count (solid line) and by calculation presuming no influence of one cell on another (dashed line) (see Fig. 4 for the equation defining P, the probable heterocyst frequency). The value of a, the probability that a cell differentiates, was taken to be the heterocyst density (heterocysts per total cells) from the given data. Graphs inspired by reference 215, using data taken from reference 221. The frequencies of interheterocyst intervals of length zero are not shown in the first two panels. (A) Wild-type Anabaena strain PCC 7120 grown on nitrate and shifted to no fixed nitrogen for 24 h (P = 9.5%). (B) patS Anabaena strain PCC 7120 mutant grown on nitrate and shifted to no fixed nitrogen for 24 h (P = 16.1%). (C) patS Anabaena strain PCC 7120 mutant grown on ammonium and shifted to nitrate for 96 h (P = 5.6%). Owing to the dispersed nature of the data under the conditions in panel C, the frequencies have been displayed as a rolling average, averaging three intervals at a time centered about the given number.

(i) Characteristics of the one-stage model. Wolk and Quine (215) showed by computer simulation that two assumptions were sufficient to reproduce a pattern of spaced heterocysts indistinguishable from that actually observed in filaments of A. cylindrica. First, they assumed that any cell is competent to differentiate at the moment when nitrogen is removed from the environment and that the choice of cells that initiate differentiation is random. Second, they postulated the existence of a diffusible inhibitor made by heterocysts and differentiating cells and consumed by nondifferentiating cells, as predicted by experimental data (127, 204, 205, 207, 210).

The model is attractive for its simplicity and because its elements may be interpreted in ways that are physically appealing. Figure 6 illustrates one interpretation of the one-stage model, in which differentiation is initiated in cells that are randomly distributed through the filament and have attained a critical level of starvation for nitrogen. These cells respond both by synthesizing the components particular to heterocysts and by releasing a compound that inhibits adjacent cells from differentiating. The compound may pass from cell to cell through junctions or through the periplasm, the region between the cytoplasmic membrane and the outer membrane of gram-negative bacteria that includes the peptidoglycan polymer. Because the outer membrane is contiguous in filamentous cyanobacteria and does not envelop individual cells (46), the periplasm is thought to be common to all cells in the filament, at least in the interval between adjacent heterocysts.

View larger version (37K): [in this window] [in a new window]   FIG. 6. Physical interpretation of the one-stage model of heterocyst spacing: Each line represents a filament consisting of many cyanobacterial cells. The color within each cell represents its nitrogen status: the darker the color, the greater the amount of available nitrogen. (A to D) A filament is suddenly starved for nitrogen. Each cell draws on nitrogen reserves, postulated to be available in different cells to different degrees. When a cell has depleted its reserves to the extent that a critical level of starvation is reached (*), it becomes committed to heterocyst differentiation. (E to G) Commitment is postulated to have two effects. The committed cell releases a signal (N) that diffuses to adjacent cells and prevents them from differentiating. In this interpretation, the signal is postulated to be a nitrogenous substance that feeds adjacent cells (symbolized by a darkening of their color). In addition, commitment prevents the committed cell from responding to its own inhibitor. Cells distant from the committed cells continue to starve, until one reaches the critical level. The position of the first cells that initiate differentiation is not critical to spacing. Spacing is determined by lateral inhibition.

Alternatively, the cell may have committed itself genetically to differentiation through a positive-feedback switch. The model of Wilcox et al. (204), similar to the ideas of Turing (194), included not only a diffusible inhibitor, Y, but also a nondiffusible component, X, that acts to increase its own synthesis, to increase the synthesis of inhibitor, and to promote differentiation. Once X has reached a level sufficient to feedback on itself, the cell may be irreversibly locked into a state of differentiation. Reasons to equate HetR with X have already been discussed.

Finally, the inhibitor may be made only after its export from the cell. One can imagine that PatS protein (see below) may be inactive until cleaved outside the differentiating cell (2).

(ii) Multiple contiguous heterocysts. The distinctive attribute of the one-stage model is that the pattern of heterocyst spacing in the filament is an outcome solely of lateral inhibition by differentiating cells. The pattern does not rely on the choice of cells initially selected to differentiate. In its basic form, the one-stage model postulates that initiation is random (215), but this need not be the case and evidently is not so. Wilcox et al. (204) reported seeing strings of adjacent differentiating cells in A. catenula that eventually resolved to single heterocysts, and they interpreted them as normal intermediates of pattern formation. Such intermediates would imply that initiation of differentiation is not random but, rather, takes place in clusters of contiguous cells along the filament.

Apart from the observations of Wilcox et al. (204), strings of differentiating cells have rarely been seen with wild-type strains of cyanobacteria, including Anabaena strain PCC 7120 (212); however, they are common results of chemical and genetic manipulations (Table 2) (for a more extensive discussion, see reference 214). Exceptionally high light intensity (4) and several chemicals often produce strings of two or more heterocysts where single heterocysts would otherwise be expected. Among the chemicals that produce strings of heterocysts are 7-azatryptophan (1, 38, 126), cyclic AMP (172), and rifampin (215), an inhibitor of RNA polymerase. Whether the strings produced chemically occur at random, suggestive of random initiation, or in a pattern, suggestive of a separate pattern-generating mechanism, has to our knowledge never been addressed. Mutant strains that show contiguous heterocysts have been reported (19, 206, 216, 220).

(iii) Apparent inconsistencies with the one-stage model. The one-stage model makes a few predictions that have sometimes not been fully borne out. Chief among these is the existence and behavior of an inhibitory signal emanating from differentiating cells. The one-stage model predicts that abolishing this signal would lead to random spacing of heterocysts while overproducing it might suppress heterocyst formation altogether.

Bauer et al. (13) found two regions of the chromosome that could suppress heterocyst formation when present in Anabaena strain PCC 7120 on a multicopy vector. One of the effective regions was narrowed to patS, a small open reading frame capable of encoding a 13- or perhaps 17-amino-acid polypeptide (220). Controlled overexpression of patS also led to suppression of heterocyst differentiation. Moreover, suppression was observed by exogenous application of a chemically synthesized peptide corresponding to the five C-terminal amino acids of PatS but not by other similar peptides or by the component amino acids. These results indicate that a secreted peptide encoded by patS might serve as the signal mediating lateral inhibition of differentiation.

Deletion of patS resulted in a mutant Anabaena strain PCC 7120 that made heterocysts even in the presence of nitrate and produced strings of heterocysts in the absence of a combined nitrogen source (220, 221). In the latter case, heterocyst clusters are spaced nonrandomly, as judged by the absence of a monotonic decline in frequency as the interheterocyst distance increases (Fig. 5B). Indeed, if one accounts for the fact that contiguous heterocysts diminish the distance between clusters by decreasing the number of dividing, nondifferentiating cells, then the distribution of interheterocyst distances of the patS mutant is much closer to that of the wild-type strain than is at first apparent. The PatS polypeptide, thus, cannot by itself be the signal called for by the one-stage model, since its loss does not lead to random differentiation. Rather, the polypeptide is evidently required to prevent contiguous differentiating cells from producing strings of heterocysts.

The absence of randomly distributed heterocysts in a patS mutant grown on nitrate (221) is equally informative. The presence of the nitrogen source should inhibit the production of the hetN-associated signal (see below), and one might not expect a gradient of nitrogenous compounds emanating from the nitrogenase-deficient heterocysts that appear in the patS mutant. Still, the distribution of interheterocyst distances is not random but, instead, is biased against short intervals; this can be seen by comparing Fig. 5C to the predicted random distribution of heterocysts in Fig. 4A. Thus, there is evidently another pattern-producing signal besides those associated with patS and hetN. The signal may be related to the position of cells in their cell cycles, as discussed below, or the result of heterocysts serving as a sink for glutamate or a source of -ketoglutarate, glutamine, or some other amino acid produced from glutamine (121). The presence of an exogenous nitrogen source does not necessarily imply the absence of a gradient of nitrogenous compounds.

The second heterocyst-suppressing region (13, 19), hetN, encodes a protein with similarity to NAD(P)H-dependent oxidoreductases involved in fatty acid biosynthesis. Mutants with insertions hetN exhibit an unstable phenotype that is initially similar to that of patS mutants but then changes to a total inability to differentiate heterocysts (19), presumably the result of secondary mutations (19,30). HetN is expressed exclusively in heterocysts and only after hetR is induced during nitrogen step-down experiments (13, 30). It therefore plays no role in the initial pattern of heterocyst differentiation. However, conditional hetN mutants exhibit, under nonpermissive conditions, multiple contiguous heterocysts from secondary rounds of differentiation (30).

Mutants that produce nonrandomly placed strings of heterocysts are therefore readily obtained. In contrast, no mutant has been reported in which heterocysts are spaced at random, contrary to what one would expect if pattern were produced by a single mechanism. We take from this the lesson that there is a complicated and dispensable machinery that resolves strings of differentiating cells but that the fundamental mechanism that produces well-spaced heterocysts is embedded in basic cellular functions (and thus immutable) or is the result of multiple redundant processes.

Mitchison et al. (127) observed that differentiation is apparent only in cells where 6 to 8 h have elapsed since cell division in A. catenula. This raises the possibility that the position of a cell within the cell cycle may govern whether it is competent to differentiate (5). Consistent with this idea, synchronization of a mutant of Anabaena strain PCC 7120 that grows as single cells gives a culture that rises and falls twofold in competence to differentiate over the course of a generation time (Gantar and Elhai, unpublished). If competence were tied to position within the cell cycle, then any preexisting pattern in filaments with respect to the cycle would contribute to the ultimate pattern of heterocysts. Such a pattern exists within filaments of Anabaena strain PCC 7120 grown on nitrate (R. Bucheimer, A. Meng, and J. Elhai, unpublished results). On average, neighboring cells tend to divide at similar times while cells separated by six or seven positions along the filament tend to divide about 180° out of phase.

(iv) Two-stage model of pattern formation. The one-stage model places the entire burden of pattern formation on lateral inhibition. The common appearance of contiguous differentiating cells in genetically or chemically perturbed strains raises the possibility that the pattern emerges in two stages: first, the development of a crude pattern consisting of regions of differentiating cells, and second, the resolution of these regions into single heterocysts. Influenced by the requirement of heterocyst differentiation for DNA replication (5, 6) and other results already discussed, we present a two-stage model in which one stage of pattern development is dependent on the position of a cell in the cell cycle (Fig. 7), but it is possible, of course, that the crude pattern may be governed by other influences.

View larger version (19K): [in this window] [in a new window]   FIG. 7. Two-stage model of heterocyst spacing. (A) In response to nitrogen deprivation, four contiguous cells at similar stages in their cell cycles (shown here beginning cell division) initiate differentiation. The string of differentiating cells (dark) is resolved by competitive interaction to a single heterocyst (thick envelope). (B) Either nitrogen deprivation or the presence of a plant signal activates NtcA protein in all cells. The presence of activated NtcA protein and passage of the cell through a critical stage in the cell cycle induce hetR to a middle level of expression, mediated through HetF. Some initiate cells attain a PatA-dependent state (?; perhaps activation of HetR), characterized by a further induction of hetR expression. In these cells, PatS is highly expressed and diffuses to adjacent cells, where it is taken up and inhibits HetR expression. Cells where the positive intracellular HetR feedback loop (thick line) dominates become heterocysts. Cells where the negative intercellular PatS effect dominates revert to vegetative status. As they do, their own positive-feedback loop weakens as PatS inhibits HetR expression and those cells produce less PatS (dotted lines). Cells that are not passing through the critical stage of the cell cycle when nitrogen deprivation takes hold do not initiate differentiation and remain vegetative cells.

(b) Stage 2 (resolution). Expression of hetR prompts the synthesis and release of an inhibitory signal, presumably PatS (and, after long periods of starvation, perhaps a HetN-related product). Each of the cells that have initated differentiation must choose one of two directions away from their unstable state. A relatively high intracellular level of HetR would, through positive feedback, tend to increase the expression of HetR, committing the cell to differentiation. Conversely, a high level of PatS transported from adjacent initiate cells would tend to decrease the level of hetR expression, leading to regression.

The existence of genes whose mutation leads to the failure to initiate heterocyst differentiation when mutated and whose overexpression leads to multiple contiguous heterocysts, or vice versa (Table 2), is explicable by their roles in resolution: an overactive resolution stage eliminates all heterocysts, while an underactive resolution stage leaves the intermediate stage intact. By the same token, the many treatments that lead to multiple contiguous heterocysts also can be seen as interfering with resolution. Mutations in some genes are proposed to act on initiation; these include ntcA, through the requirement for nitrogen deprivation, and perhaps hanA, through the requirement for DNA replication.

Some observations do not fit readily into this simple model. A patA mutant produces heterocysts only at the ends of filaments (100), similar to the behavior of cyanobacteria of the genus Cylindrospermum grown continuously on N₂ (215). A patB mutant of Anabaena strain PCC 7120 (101) and a patN mutant of N. punctiforme (F. C. Wong and J. C Meeks, unpublished results) both exhibit a higher frequency of heterocysts than normal, but not as multiple contiguous heterocysts. However, an increased frequency is also seen in wild-type A. cylindrica exposed to high-intensity light (4). It is conceivable that a metabolic deficiency affecting cell division, such as that described in the patB mutant, might account for the phenotype. Mitchison and Wilcox (125) observed that cell division in A. catenula is asymmetric and that heterocysts arise only from the smaller of the two daughter cells. While asymmetric division has been reported in only one other heterocyst-forming strain (4), asymmetry not apparent by light microscopy is still possible. The rules found by Mitchison and Wilcox (125) make the strong prediction that if adjacent cells differentiate (as in a patS mutant), they will be no closer than three generations distant (i.e., they will share the same greatgrandparent cell). The two-stage model predicts instead that the adjacent cells are most likely to be siblings.

Within the context of the two-stage model, it is possible to see that plants could influence the number of heterocysts by increasing initiation, perhaps by altering the cell's perception of its nitrogen status or by modulating resolution, or both.

View larger version (63K): [in this window] [in a new window]   FIG. 8. Photomicrograph of filaments of N. punctiforme in the N2-dependent vegetative growth state and at different stages of the hormogonium cycle. Phase-contrast images of vegetative filaments (A) and hormogonia (B to D) are shown. Note the process of heterocyst differentiation by the end cells of the hormogonium filaments starting at T48(C), with the appearance of mature heterocysts and an increase in size of vegetative cells by T84(D); this is followed by vegetative cell growth and division and eventually by differentiation of intercalary heterocysts. Bar, 10 µm.

The mechanism by which hormogonia glide is unknown. A protein termed oscillin, which may be essential for gliding, has been identified in vegetative filaments of non-heterocyst-forming Phormidium uncinatum (81). Oscillin forms as fibrils external to the outer membrane of the cell wall and is proposed to passively direct the flow of mucilage between the filament and substratum. A similar mechanism may be involved in hormogonium motility.

The acquisition of motility comes at a price. Hormogonia are unable to fix nitrogen or grow. Phycobiliprotein synthesis ceases (42), leading through attrition to an accompanying decrease in fluorescence. Cells of hormogonia continue to photosynthesize and assimilate exogenous ammonium if it is available, although the rates of CO₂ fixation and NH₄⁺ incorporation are only 70 and 62%, respectively, of those of vegetative filaments (33). The metabolic fates of the photosynthate and newly synthesized or regenerated amino acids in the non-growth hormogonium state are not known. However, much of the metabolic output is probably devoted to the production of the proton motive force (80) and the synthesis and secretion of mucilage (81) thought to be required for gliding motility.

Since cells within hormogonia do not grow, the state by necessity is transient. One can therefore describe a hormogonium cycle, starting with the induction of hormogonia and ending with the return of hormogonia to the vegetative state within the equivalent of two or three generation times (Fig. 8). On induction, cell division begins at a nearly simultaneous rate in all cells of a filament, but not necessarily at the same time in all filaments. The result is that except for a very short period, filaments at a given moment consist of all newly divided cells (reflecting global differentiation) or virtually no newly divided cells (33, 42). There is no significant net synthesis of DNA or protein during hormogonium differentiation (75), although some protein synthesis is still required by the process (33); therefore, the smaller hormogonium cells result from multiple cell divisions in the absence of cell growth. Cell division without DNA synthesis is possible because cyanobacteria, in general, contain multiple copies of their genomes (74). Even in the absence of DNA replication, hormogonium cells are thus likely to receive at least one copy of the genome following multiple septations of the parental cell. Shortly after cell divisions, the filaments begin to fragment in a random manner. In filaments containing heterocysts, fragmentation takes place preferentially between heterocysts and their adjacent vegetative cells (33). The detachment of heterocysts from vegetative cells results in an interruption in the flow of reduced carbon from vegetative cells that supports both nitrogenase activity and respiration, and O₂ protection is lost. Detached heterocysts therefore cannot fix N₂, and nor can hormogonia.

The hormogonia remain in the gliding state for about 36 to 48 h, after which they cease to move. Only at this stage, in the absence of combined nitrogen, do heterocysts differentiate, first at the ends of the filaments and later, as the vegetative cells grow, with the typical spacing pattern within the filaments. Based on the resumption of pigment and total protein synthesis, the conversion of hormogonia back to vegetative filaments is complete within 96 h after their induction. Preliminary experiments indicate that cultures of Nostoc (E. L. Campbell and J. C. Meeks, unpublished data) and Calothrix (184) that have extensively differentiated hormogonia require a vegetative growth period before they are again able to differentiate a substantial number of hormogonia.

What induces the differentiation of hormogonia? Combined nitrogen deprivation is sufficient to induce heterocyst differentiation in all capable free-living cyanobacteria. In contrast, no single environmental factor has been identified that will induce hormogonium differentiation in all capable strains. The one common factor in initiation of hormogonium differentiation appears to be a change in some environmental parameter: an increase or decrease of a nutrient or a change in the quantity or quality of light. The most frequently used inducer of hormogonium differentiation in the laboratory is the transfer of a dense culture, near stationary growth phase, to fresh medium (159). This procedure has the possibility of multiple environmental