All strains used in this study are listed in Supplementary Table S1. Anabaena WT and its derivatives were grown in BG11 (Stanier et al., 1971). For heterocyst induction, cells grown to logarithmic phase in BG11 were transferred to BG11₀ (BG11 without combined nitrogen) medium as described previously (Golden et al., 1991). All strains were cultured in liquid medium at 30°C with shaking at 180rpm and the constant illumination with light intensity at 30μmolm⁻² s⁻¹, the CO₂ concentration is 400ppm. The overexpression strains based on the artificial CT promoter (Xing et al., 2020) were grown first in BG11 medium to logarithmic phase, then transferred to BG11₀ in the presence of copper and theophylline, for the induction of heterocysts development and gene expression. When not specified, the concentration of the copper and theophylline is 0.5 and 2mM, respectively. To modify the c-di-GMP level in vivo, gradient concentrations of copper/theophylline were tested at: 0μM/0mM, 0.125μM/0.5mM, 0.25μM/1mM, 0.5μM/2mM, and 1μM/4mM. Whenever necessary, neomycin (25μg/ml), streptomycin (2.5μg/ml) or spectinomycin (5μg/ml) was supplied in BG11 or BG11₀ medium.

All plasmids used in this study are listed in Supplementary Table S2 and verified by Sanger sequencing. All the oligonucleotides used for plasmid construction and mutant checking in this study are listed in Supplementary Table S3.

All mutant stains with markless deletion of the corresponding genes were generated in this study by the genome editing technique based on CRISPR-Cpf1, and the plasmids were constructed as previously described (Niu et al., 2019). The plasmids for gene overexpression in Anabaena were constructed based on the PCT vector (Xing et al., 2020), allowing control of gene expression by the artificial CT promoter. This promoter was consisted of the petE promoter region and a theophylline riboswitch. Gene expression could thus be induced by adding copper and theophylline in the culture media. The plasmids used for protein purification from E. coli BL21 (DE3) were constructed based on the pHTwinStrep vector, which containing two carboxyl-terminals Strep-tag at the C-terminal.

To construct the mutant and overexpression strains, the corresponding plasmid was transferred into Anabaena by conjugation through triparental mating as previously described (ElhaI et al., 1997). All mutants and overexpression strains were confirmed by PCR.

c-di-GMP in Anabaena was analyzed as previously described (Agostoni et al., 2013). At the indicated time points or the indicated conditions, 2ml of cells were quickly collected by filtering under standard light illumination and resuspended with 500μl ice-cold extraction buffer containing 40% acetonitrile, 40% methanol, and 0.1N formic acid. The cells were broken by crusher FastPrep-24TM5G and the cell lysis efficiency was checked under a microscopy. Once the cells were totally broken, the samples were incubated at −20°C for 30min and then centrifuged at 12,000g for 5min at 4°C. The supernatant was transferred to a new 1.5ml tube and stored at −80°C until further analysis. The protein concentrations of all samples were measured in parallel by Bradford assay. All experiments were repeated three times.

For quantification, the supernatants were evaporated and dried by vacuum manifold. The pellets were resuspended in an equal volume of Milli-Q-purified water. 20μl of each sample was analyzed by UPLC-MS/MS on ACQUITY UPLC H-class-Xevo TQ MS system, equipped with a Waters BEH C18 2.1×50mm column. All samples were filtered with a 0.22μm Ultra free-MC membrane before subjected to the column. The sample was separated at 25°C with a flow rate of 0.3ml/min and the chromatograph process used was as the following: gradient of solvent A (10mM tributylamine plus 15mM acetic acid in 97:3 water: methanol) to solvent B (methanol): t=0min; A-99%: B-1%, t=2.5min; A-80%: B-20%, t=7.0min; A-35%: B-65%, t=7.5min; A-5%:B-95%, t=9.01min; A-99%: B-1%, t=10min (end of gradient). The c-di-GMP was detected with electrospray ionization using multiple reaction monitoring in negative-ion mode at m/z 689.16→344.31. The parameters for mass spectrum were as the following: capillary voltage, 3.5kV; cone voltage, 50V; collision energy, 34V; source temperature, 110°C; desolvation temperature, 350°C; cone gas flow (nitrogen), 50l/h; desolvation gas flow (nitrogen), 800l/h; collision gas flow (nitrogen), 0.15ml/min; and multiplier voltage, 650V. Standard c-di-GMP (Biolog) at the concentrations of 500nM, 250nM, 125nM, 62.5nM, 31.25nM, 15.62nM, and 7.81nM were used to generate the standard curve for calculating the c-di-GMP concentration of each sample.

The microscope images were captured by SDPTOP EX30 microscope. Microscope images taken at the indicated times and conditions were used for analyzing the heterocyst frequency and vegetative cell intervals. All experiments were analyzed with three parallel analyses.

Wide-type Anabaena was cultivated to the log phase in BG11 and then transferred to BG11₀ medium for heterocyst induction as described previously (Golden et al., 1991). To prepare the samples, 30ml of Anabaena were harvested at 0, 3, 16, 48, and 72h after nitrogen starvation, respectively, in triplicate. The total RNA was extracted with Plant Total RNA Isolation Kit (Foregene), and the genomic DNA was removed by treating with RNase-free DNase I (Promega, United States). The obtained total RNA samples were incubated with reverse transcriptase with HiScript Q RT super mix (Vazyme).

Quantitative Real-Time PCR was performed by using ChamQ SYBR qPCR Master Mix (Vazyme) with specific primers listed in Supplementary Table S3. The house keeping gene rnpB was used as an internal control. Primers Pallrs04F334 and Pallrs04R456 (for rnpB), PcdgS4F226 and PcdgSR402 (for cdgS), and PcdgSHF2342 and PcdgSHR2455 (for cdgSH), were used to amplify specific fragments. The RT-PCR products were confirmed by sequencing. Data were analyzed by ABI 7500 SDS software. To be consistent, the baseline was set automatically by the software. The comparative CT method (2^−△△CT method) was used to analyze the expression of cdgS and cdgSH. The data correspond to relative mRNA expression levels, shown as the mean±SE (standard error) of three replicates (N=3) analyzed by t test, and the significance was p<0.05.

The plasmids carrying recombinant genes for Strep-tag fusion proteins were transformed into competent cells of E. coli BL21 (DE3; Novagen) and then grew in LB medium with kanamycin (50μg/ml). When the cell culture reached OD₆₀₀=0.5–0.6, 0.5mM IPTG was added and continued to grow overnight at 18°C. Cells were harvested by centrifugation at 8000g for 5min, resuspended in buffer W (100mM Tris-HCl pH 8.0, 1M NaCl and 1mM EDTA) and lysed by French press. The lysate was cleared by centrifugation at 12000g for 30min at 4°C, and the supernatant was incubated with buffer W equilibrated Strep-Tactin® XT Superflow® resin (IBA) at 4°C for 1h. The resin was then collected by filtration and washed with 10 column volumes (CV) of buffer W. The proteins were eluted with elution buffer (buffer W containing 50mM biotin) and dialyzed against 50mM Tris–HCl (pH 8.0), 10mM MgCl, 0.5mM EDTA, and 300mM NaCl for 12h at 4°C.

The DGC and PDE activity assay was analyzed as previously described (Enomoto et al., 2015). In brief, the total volume of each reaction was 50μl and the reaction mixture contained 50mM Tris·HCl pH 8.0, 10mM MgCl₂, 0.5mM EDTA, 300mM NaCl, 50μM GTP (Thermo scientific) for DGC or 50μM c-di-GMP (Biolog) for PDE. Each reaction was initiated by adding the corresponding protein at a final concentration 2μM, and then incubated at 25°C for 30min. All the reactions were stopped by adding 20mM EDTA and immediately heated at 95°C for 5min. Before subjected to HPLC analysis, all samples were filtered with a 0.22μm Ultrafree-MC membrane by centrifugation. Nucleotides were separated by reverse-phase HPLC through a C18 column (150mm×6mm i.d.; DAISOPAK SP-120-5- ODS-AP; Daiso). 20μl sample from each reaction was injected and eluted with buffer A (50mM potassium phosphate, 4mM tetrabutylammonium hydrogen sulfate, pH 6.0) and buffer B (75% buffer A, 25% methanol, V/V) at 1.2mlmin⁻¹, and the elution protocol as followed: 0min, 80%/20% A/B; 2.5min, 80%/20% A/B; 5min, 70%/30% A/B; 14min, 60%/40% A/B; 25min, 0%/100% A/B; 32min, 0%/100% A/B; 33min, 80%/20% A/B; and 34min, 80%/20% A/B. The nucleotides were detected at 254nm. Standard c-di-GMP and GTP at concentrations of 10, 20, 40, 80, 160, 320, and 640μM in the reaction buffer was used to generate the standard curve for quantifying the c-di-GMP and GTP levels, respectively, in reactions.

To investigate the relationship between c-di-GMP and heterocysts development, we first determined the changes of intracellular concentration of c-di-GMP during this physiological process in Anabaena. c-di-GMP from wild-type (WT) cultures at different time points after nitrogen starvation was extracted and quantified by ultra-high performance liquid chromatography with tandem mass spectrometry detection (UPLC-MS/MS). As shown in Figure 1, the c-di-GMP level in cells in BG11 medium (time 0) was 0.078±0.014nmol/mg protein. After exposure of cells to BG11₀ medium, c-di-GMP levels increased by 3-fold at 3h after the transfer, reaching a peak level of 0.231±0.034nmol/mg protein. After this time point, the levels of c-di-GMP gradually decreased to the basal level in 24h. These results suggested that c-di-GMP may be an important signaling molecule during heterocyst differentiation.

To investigate the role of c-di-GMP in heterocyst differentiation, in-frame markerless deletion of all 16 genes encoding proteins with predicted DGC and/or PDE domains were generated by genome editing technique based on CRISPR-Cpf1 (Supplementary Figure S1; Niu et al., 2019). We determined the ability of heterocyst formation for all mutants grown in BG11₀. When observed under the microscope, 2 of the 16 mutants, ΔcdgSH and ΔcdgS, were affected in heterocyst frequency while all others had a heterocyst pattern and percentage comparable to those of the WT (Table 1). The phenotype of ΔcdgS was similar to that already reported (Neunuebel and Golden, 2008). Indeed, heterocyst frequency of the mutant was similar to that of WT at 24h after the induction (8.0±0.27%) and dropped by 60% (2.5±0.08%) 3days later after the induction as compared to the WT (6.2±0.16%). It can grow in BG11₀, but less well than the WT (Supplementary Figure S6B). While cdgS was required to maintain heterocyst frequency over time, without affecting the initial heterocyst pattern ΔcdgSH showed a very different phenotype. After 24h of growth in BG11₀ medium, ΔcdgSH mutant could differentiate mature heterocysts; however, the percentage of heterocysts showed an increase (10.3%±0.43%) when compared to WT (8.0%±0.33%), and a Mch (multiple-contiguous-heterocysts) phenotype was occasionally observed (Figures 2A,B). To confirm this phenotype, we further quantified the pattern of heterocysts along the filaments. Heterocyst intervals of WT show regularity, with 11 to 13 vegetative cells as the most representative intervals. In ΔcdgSH mutant, the distribution still kept a regularity, but the intervals shortened to 8 to 10 vegetative cells as the most representative ones (Figure 2B). Interestingly, from 48h after induction, the heterocyst frequency of the ΔcdgSH mutant became similar to WT (Figure 2C), indicating that the mutation affected only the initial establishment of the heterocyst frequency, but not its maintenance. The ΔcdgSH can grow well in BG11₀ (Supplementary Figure S6A). The phenotype of ΔcdgSH could be complemented by cdgSH controlled by its own promoter present on a replicative plasmid (Figure 2C, strain CcdgSH). Taken together, among the 16 annotated genes related to c-di-GMP levels on the genome, cdgSH and cdgS appear to be the major ones involved in the regulation of heterocyst pattern.

Although both cdgSH and cdgS regulate heterocyst intervals, their phenotypes were very distinct. To investigate the relationship between cdgSH and cdgS in regulating heterocyst intervals, we constructed a ΔcdgSHΔcdgS double mutant. 24h following combined-nitrogen deprivation, heterocyst frequency of this double mutant was 10.5%±0.21%, similar as ΔcdgSH. However, with prolonged time in combined-nitrogen free medium, the frequency decreased gradually, and reaching 5.2%±0.19, 2.5%±0.17 and 2.1%±0.14% at 48, 72, and 96h, respectively (Figure 2C), a phenotype similar as ΔcdgS. Although the two single mutants displayed opposite effect on the frequency of heterocysts, the results indicated that the phenotype of the double mutant was the sum of the single ones under diazotrophic condition. This indicated that these two genes acted independently at different time points during heterocyst development, with cdgSH for establishment of heterocyst pattern (the first 24h), while cdgS for maintenance of the frequency of heterocysts at the later stages.

Next, to better understand how the two genes could participate at different time points in heterocyst development, we checked the transcription levels of cdgSH and cdgS in WT at different time points after nitrogen starvation by quantitative Real-Time PCR. Our data indicated that these two genes showed opposite transcriptional profiles (Figure 2D). Following nitrogen stepdown, the transcript level of cdgS was kept steady in the first 3h, and then increased after 16h post induction. At 72h after combined-nitrogen deprivation, the relative transcript level of cdgS was increased around 3-fold. In contrast, under the same conditions, the transcriptional activity of cdgSH was dropped to nearly undetectable level at 16h after transfer to BG11₀. Although gradually increased after 16h, it still remained at a much lower at 72h as compared to that observed at time 0 (Figure 2D). The transcriptional profiles of these two genes are in good agreement with the phenotype of their corresponding mutants. Indeed, cdgSH had a higher transcriptional activity at the onset of heterocyst induction, and its inactivation affected the establishment of heterocyst frequency. In contrast, the transcriptional activity of cdgS was low at the beginning of heterocyst induction, but increased steadily afterwards; consistently, heterocyst frequency continuously dropped from the second day overtime after combined-nitrogen deprivation.

CdgSH is predicted to be a multi-domain cytoplasmic protein, with FHA, PAS, GAF, GGDEF and EAL domains arranged in tandem (Supplementary Figure S1). The GGDEF and the EAL domains are both located towards the C-terminus of the polypeptide and contain all the conserved residues required for their corresponding catalytic activity (Supplementary Figure S2). This observation suggests that CdgSH may exhibit both c-di-GMP synthesis and degradation activities. To check this observation, CdgSH and its variants were expressed in E. coli BL21 (DE3) and purified using a strep-tag affinity column (Supplementary Figure S3). The DGC and PDE activities of these proteins were then determined. Our results showed that no matter which of the substrates, GTP or c-di-GMP, was used, only one product, pGpG, could be detected by HPLC (Figure 3). This indicated that CdgSH could synthesize c-di-GMP from GTP, and immediately hydrolyzed c-di-GMP to pGpG. To further confirm this observation, the CdgSH^GGAAF variant, which the catalytic and the metal ion coordination sites were replaced (mutations D530A and E531A), was produced (Chan et al., 2004; Wassmann et al., 2007). This variant kept an intact PDE domain, but a defective DGC domain. As shown in Figure 3, CdgSH^GGAAF could only degrade c-di-GMP to pGpG, and when GTP was added, no activity was detected. The CdgSH^AAA variant, which inactivated the PDE activity of CdgSH by mutating the metal ion coordination sites (mutations E656A and L658A; Barends et al., 2009; Tchigvintsev et al., 2010), was also tested, and it could synthesize c-di-GMP from GTP, but the c-di-GMP hydrolysis activity was abolished (Figure 3).

To further investigate the relationship between the PDE and DGC activities of CdgSH, we quantified the steady-state kinetic parameters (kcat and Km) for CdgSH, CdgSH^GAAEF and CdgSH^AAA under the same experimental conditions. As shown in Table 2, the kcat value for the PDE activity is three times that of the DGC activity in CdgSH, while the DGC Km values is 1.5 times that of the PDE activity. This result indicates that although the binding efficiencies of the two catalytic domains to their respective substrate are similar, the conversion efficiencies are quite different. Indeed, the catalytic activity of PDE in CdgSH, expressed by the kcat/Km ratio is 5.4×10⁻²μM/S⁻¹ (Table 2), which is more than 5-fold higher than that of DGC (1.0×10⁻²μM/S⁻¹). In addition, we found that the catalytic activity (kcat/Km) of PDE in CdgSH and CdgSH^GGAAF are similar to each other, suggesting that the PDE activity of CdgSH was independent of the DGC activity. On the other hand, the catalytic activity (kcat/Km) of DGC in CdgSH^AAA decreased 3-fold when compared to that of CdgSH, suggesting the existence of a feedback inhibition of the DGC activity by c-di-GMP. Such a phenomenon was known for typical GGDEF domains, which can be explained by the presence of the conserved inhibitory site (RXXD motif) present in CdgSH which allows c-di-GMP binding and allosteric feedback inhibition (Supplementary Figure S2; Chan et al., 2004; Wassmann et al., 2007). Taken all together, our results demonstrated that CdgSH was a bi-functional enzyme with both DGC and PDE activities, but its PDE activity dominated over the DGC activity, making the c-di-GMP hydrolysis a prevailing output at the end.

For comparison, we also tested the enzymatic activity of CdgS which possesses an N-terminal REC domain and a C-terminal GGDEF domain under the same experimental conditions (Supplementary Figures S1, S2). Our results indicated that CdgS could convert the substrate GTP to c-di-GMP (Supplementary Figure S4), consistent with the result already published (Neunuebel and Golden, 2008). When the catalytic site and the metal ion coordination site (E229A, and E230A) of DGC was replaced (Chan et al., 2004; Wassmann et al., 2007), the c-di-GMP synthesis activity of CdgS^GGAAF was completely abolished (Supplementary Figure S4). Therefore, CdgS is a typical c-di-GMP synthase.

CdgSH is a protein containing other domains in addition to the two with opposing activities in c-di-GMP synthesis and hydrolysis. To evaluate the contribution of the different activities of CdgSH in regulating heterocyst frequency, we examined the ability of heterologous genes encoding c-di-GMP metabolic enzymes to complement the ΔcdgSH mutant. YdeH (DGC) and YhjH (PDE) are well-studied c-di-GMP metabolic enzymes from E. coli with high catalytic activity in vitro (Pesavento et al., 2008). Each of the two proteins contains only one catalytic domain, without additional regulatory domains (Supplementary Figure S2). Therefore, they are often used to alter the intracellular c-di-GMP level of various bacteria (Tschowri et al., 2014; Li et al., 2018). To examine their activities in Anabaena, plasmids containing ydeH or yhjH under the control of inducible regulatory elements (the CT promoter) was transferred into Anabaena by conjugation, giving the strain OE_CT-ydeH or OE_CT-yhjH, respectively. We then determined the intracellular c-di-GMP levels of the two recombinant stains. Our data revealed that the intracellular c-di-GMP level of the two recombinant strains, growing in BG11 medium and induced by adding 2mM theophylline and 0.5μM Cu²⁺, displayed significant changes as compared with WT after 24h of induction. The c-di-GMP level of the cells in OE_CT-ydeH was at 0.29±0.04nmol/mg total protein, corresponding to an increase by 2-fold relative to that of WT. Under similar conditions, cells of OE_CT-yhjH decreased the level of c-di-GMP by 2-fold, to 0.06±0.01nmol/mg total protein (Figure 4A). Therefore, YdeH and YhjH are active in Anabaena, and their production could modify the intracellular levels of c-di-GMP in this organism.

Next, we analyzed the heterocyst pattern and frequency of the strains expressing yhjH or ydeH in the ΔcdgSH background. To make the data comparable, we used the native promoter of cdgSH to drive the expression of yhjH or ydeH on a replicative plasmid. We found the heterocyst pattern in the recombinant strain ΔcdgSH::yhjH was restored to the WT pattern, with 11 to 13 vegetative cells as the most representative intervals after 24h combined nitrogen step-down and the heterocyst frequency at this time point was also restored to the WT level, to about 8.2±0.35% (Figure 4B). In contrast, the strain ΔcdgSH::yhjH^AAA bearing the same plasmid but containing the mutant allele of yhjH (yhjH^AAA) failed to complement the defect in heterocyst distribution and frequency of ΔcdgSH (Figure 4B). Furthermore, the -strain ΔcdgSH::ydeH was unable to restore the change in heterocyst distribution and frequency of the ΔcdgSH mutant (Figure 4B; Supplementary Figure S3). Altogether, our results indicated that the c-di-GMP hydrolysis activity of CdgSH, and hence the decrease in c-di-GMP level, was the major element that contributes to the regulation of heterocyst frequency of ΔcdgSH mutant at 24h after nitrogen starvation. With prolonged time in combined-nitrogen free medium, ΔcdgSH:: yhjH strain had a slightly lower heterocyst frequency as compared to the WT from 48 to 96h after the transfer to BG11₀ medium (Figure 4C). This phenotype is similar to ΔcdgS mutant which also displayed a drop in heterocyst frequency as compared to the WT at the same time points. This result could be explained by the fact that the gene yhjH in the complemented strain was carried on a replicative plasmid that is known to lead to a higher level expression than that from the chromosome, and also the fact that YhjH possess a high PDE activity.

We also tested whether ydeH could complement ΔcdgS. Indeed, the ydeH gene under the control of the native promoter of cdgS, carried on a replicative plasmid could complement ΔcdgS successfully (strain ΔcdgS::ydeH, Figure 4D). ydeH^GGAAF which encodes an inactive form of YdeH (E208 and E209, Figure 4D) was unable to complement the phenotype of ΔcdgS mutant when controlled by the same regulatory system (Figure 4D). These studies indicated that the c-di-GMP metabolic activities of CdgSH and CdgS were the major reason for the phenotype observed following the inactivation the corresponding genes in Anabaena.

The metabolism of c-di-GMP should be highly regulated during heterocyst development, as demonstrated by the changes of the c-di-GMP levels during the developmental process, as well as the functions of the two genes cdgSH and cdgS, acting at different phases of heterocyst differentiation. To check whether such regulation is important, we examined the effect caused by varying the intracellular levels of c-di-GMP. For this purpose, CdgSH, CdgS, and their corresponding mutant variants were expressed in WT cells of Anabaena, respectively. The expression levels of the genes were titrated by a gradient of inducers as shown in Figure 5 and heterocyst development was observed at 24h after the initiation by transfer to BG11₀. When cdgSH was tested with increasing concentrations of inducers, heterocyst frequency gradually decreased from 8 to 1.3%. A more pronounced results were obtained when CdgSH^GGAAF in which the activity of DGC was abolished, since at the highest concentrations of inducers tested, few heterocysts were formed (0.5%; Figure 5A). These results were consistent with the fact that CdgSH has both c-di-GMP synthesis and hydrolysis activities, but the hydrolysis activity was dominant, while the CdgSH^GGAAF variant had only the PDE activity left (Figure 3; Table 2).

Interestingly, adding increasing gradients of inducers to express either the CdgSH^AAA variant that has only retained the activity of c-di-GMP synthesis or CdgS also led to a reduction of heterocyst frequency. At 24h after combined-nitrogen deprivation, with 4mM theophylline and 1μM Cu²⁺, the heterocyst frequency of both strains was around 1%, while heterocyst frequency is 8% without inducers (Figure 5B). Therefore, ectopic expression of genes encoding either c-di-GMP synthase or hydrolase activities repressed heterocyst differentiation.

To further confirm it is the c-di-GMP metabolic activities of CdgSH that was responsible for such phenotypes, we tested the strain of OE_CT-ydeH or OE_CT-yhjH at the same condition, respectively. With no inducers or with low concentrations of 0.5mM theophylline and 0.125μM Cu²⁺, the heterocysts of YdeH and YhjH overexpression strains were formed with a frequency similar as the WT (Figures 5A,B). However, when the concentrations of inducers increased to 1mM for theophylline and 0.25μM for Cu²⁺, the heterocysts frequency of both strains were reduced to about 1% at 24h after the initiation. Heterocyst developments of both strains were completely suppressed when the concentration of the inducers reached for 2mM theophylline and 0.5μM Cu²⁺ (Figures 5A,B). These results were consistent with those obtained by the ectopic expression of the CdgSH, CdgSH variants and CdgS. To exclude the influence of high dose of inducers or proteins in vivo, effects of production of CdgSH^GGAAF-AAA, CdgS^GGAAF, YdeH^GGAAF_, and YhjH^AAA variants, which c-di-GMP synthesis or degradation activities were abolished, as well as WT strain were also examined under the same experimental conditions (Figure 5C; Supplementary Figure S5A). Even at the highest concentrations of the two inducers added, heterocyst frequency and pattern were hardly affected.

We also compared the growth capacity of different strains under diazotrophic conditions. As the proportion of heterocysts decreased, the diazotrophic growth capacity of the strains decreased as well (Figures 5D–F; Supplementary Figure S5B). The growth capacity of control strains remained similar under different inducer conditions. Taken together, our results showed that depending on the promoter used (native promoters of cdgSH or cdgS, or the synthetic CT promoter), and the concentrations of the inducers employed, expression of the synthase or hydrolase activity of c-di-GMP may have different outcomes. When expressed under the ectopic promoter (CT promoter) to a high level, both the hydrolase activity and the synthase activity could lead to a repressive effect on heterocyst differentiation, suggesting that intracellular c-di-GMP homeostasis in Anabaena is important for heterocyst development.