Dear Editor,

Natural competence requires a contractile pilus system. Here, we provide evidence that the pilus biogenesis and natural competence in cyanobacteria are regulated by the second messenger c-di-AMP. Furthermore, we show that the ComFB signaling protein is a novel c-di-AMP-receptor protein, widespread in bacterial phyla, and required for pilus biogenesis and DNA uptake.

Cyclic di-AMP (c-di-AMP) is one of the recently discovered di-nucleotide-type second messengers1. In cyanobacteria, c-di-AMP controls diurnal metabolism via its binding to the carbon control protein SbtB to regulate glycogen metabolism2. Although important roles for c-di-AMP have been acknowledged since its discovery (e.g., in osmoregulation)1,2,3,4, recent studies suggested broader regulatory impacts of c-di-AMP signaling with further functions yet to be elucidated3,4 (see Supplementary Text). For instance, a role for c-di-AMP in controlling natural competence has been speculated5, although the molecular mechanism remained elusive5. In this study, we aimed to investigate the involvement of c-di-AMP in natural competence.

Natural competence involves a contractile pilus system and an assemblage of competence-accessory proteins6,7 (see Supplementary Text). To test the involvement of c-di-AMP in cyanobacterial natural competence, we compared the ability of the wild-type (WT) Synechocystis and the c-di-AMP-free ΔdacA mutant2 to take up DNA. The ΔdacA mutant showed significantly lower transformation efficiency than the WT, implying an essential role for c-di-AMP in natural competence (Fig. 1a; Supplementary Fig. S1). Complementing ΔdacA restored the transformation efficiency to WT levels, while c-di-AMP overexpression (WT::petE-dacA strain) did not affect the transformation efficiency (Fig. 1a; Supplementary Fig. S1). These results indicate that the absence of c-di-AMP affects cyanobacterial natural competence negatively, whereas high c-di-AMP does not. A similar result was obtained using the c-di-AMP-null ΔcdaA mutant8 of Synechococcus elongatus (Supplementary Fig. S1), indicating that the c-di-AMP-dependent control of natural competence is a common trait among cyanobacteria.

Fig. 1: Involvement of c-di-AMP in cyanobacterial natural competence via ComFB signaling protein.
figure 1

a Transformation efficiency of WT, ∆dacA, ∆dacA::petE-dacA, WT::petE-dacA and ∆sbtB strains (see also Supplementary Fig. S1). b Immunodetection of PilA1 in the exoproteome of WT and ∆dacA. c Pulldown experiment using immobilized c-di-AMP and extracts of Synechocystis cells grown under day-night cycles, showing the enriched proteins in the day phase. Potential new c-di-AMP receptors are highlighted in orange. d Phylogenetic tree showing that ComFB proteins are widespread among different bacterial phyla (detailed tree Supplementary Fig. S4). e Dissociation constant (KD) of c-di-AMP binding to ComFB and enthalpy (ΔH) are obtained from sigmoidal fitting curve of all ITC experiments with different monomeric ComFB concentrations. f DRaCALA assay showing the binding of [32P]c-di-AMP to purified ComFB in a concentration dependent manner as indicated. The upper panel shows a representative of one replicate from four technical replicates. The lower panel shows the calculated mean ± SD of the quantification of the bound fraction of [32P]c-di-AMP to ComFB from the four replicates and the best fitting curve with the obtained KD value. g DRaCALA competition binding assay showing the competition of [32P]c-di-AMP with different nucleotides to bind ComFB. NC refers to no competitor. SbtB and cell extract of E. coli harboring an empty plasmid were used as positive and negative control, respectively. h Transformation efficiency of WT, ∆comFB, and ΔcomFB::petE-comFB strains (see also Supplementary Fig. S1).

Full size image

Next, we checked how the lack of c-di-AMP affects pilus biogenesis. A proteome analysis of ΔdacA compared to WT under day-night cycle, a condition trigger pili biogenesis and natural competence7, revealed a strong downregulation of several proteins involved in pilus biogenesis and DNA uptake in ΔdacA4. These changes were marked by a reduced abundance of PilT1 (Slr0161), PilM (Slr1274), PilN (Slr1275), PilO (Slr1276), and Sll0180 proteins (Supplementary Table S1). The cellular levels of the other pilus machinery proteins were not significantly altered in the ΔdacA mutant (Supplementary Table S1). The assembly of a functional pilus requires two motor ATPases, PilB1 and PilT1. PilT1 is located at the pilus base and is required for pili retraction and depolymerization. Therefore, pilT1 mutant is nonmotile, hyperpiliated and loses natural competence6. Similarly, the pilM, pilN and pilO mutants are nonmotile and non-transformable. The PilMNO proteins form the alignment complex6, connecting the components of the pilus machinery in the inner and outer membranes by forming a ring-structure in the periplasm. Sll0180 is an accessory protein, needed for PilA1 (Sll1694) glycosylation and S-layer secretion, and thereby the correct assembly of the pilus machinery6. Non-functional PilA1 causes a non-transformable phenotype6. Notably, our transcriptome analysis3 showed a partial downregulation of pilT1 and sll0180, while pilT2 (sll1533) was strongly downregulated in ΔdacA (Supplementary Table S2). These findings explain why ΔdacA mutant lost the natural competence.

The striking decrease of PilT1 levels in ΔdacA suggested a strong defect in pilus assembly and retraction. To test this assumption, we examined negatively stained ΔdacA and WT cells by transmission electron microscopy (TEM). While we could detect both thick and thin pili in the WT, only thick pili were obvious in ΔdacA (Supplementary Fig. S2). Additionally, ΔdacA mutant showed a hyperpiliation phenotype in analogy to pilT1 mutant. The quantification of the major pilin PilA1 in ΔdacA exoproteome revealed an accumulation of PilA1 compared to WT cells (Fig. 1b), further supporting the notion of a c-di-AMP-dependent control of pilus biogenesis and natural competence.

The lack of non-retractable pili explains the inability of the ΔdacA to take up DNA. Interestingly, the downregulation of the above-mentioned proteins was not detected in ΔsbtB mutant4, which lacks the only known cyanobacterial c-di-AMP receptor2 (Supplementary Table S3). Furthermore, ΔsbtB9,10 behaved like the WT in our competence assays (Fig. 1a; Supplementary Fig. S1), suggesting the involvement of an additional, yet unknown, c-di-AMP receptor, required for natural competence. To identify this new c-di-AMP receptor, we performed c-di-AMP-dependent pulldowns using Synechocystis cells growing under day-night cycles (Fig. 1c; Supplementary Fig. S3a–c). The identification of several known c-di-AMP targets: SbtB2,9,10,11, as well as the transporters TrkA, KrtA, MthK, MgtE and NhaS5, validated our pulldowns2. The Slr1970 protein was also enriched and correlated with the intracellular c-di-AMP levels, where it was more abundant in the day than in the night (Supplementary Fig. S3c). This protein was annotated as ComFB12, and we found that ComFB proteins are widespread among different bacterial phyla (Fig. 1d; Supplementary Fig. S4), implying a fundamental role in cell physiology. In Bacillus, comFB forms an operon with comFA and comFC, which are known to be involved in DNA uptake12. In cyanobacteria, comFB forms an operon with hfq6, which is also involved in DNA uptake and motility (Supplementary Fig. S5), strongly suggesting a potential function of ComFB in natural competence.

To validate ComFB as a novel c-di-AMP-binding protein, we used several biophysical methods. Size exclusion chromatography coupled to multiangle light scattering showed a species of 40.5 kDa (Supplementary Fig. S6a), indicating that ComFB protein is dimeric (theoretical monomer mass 20.4 kDa). Using isothermal titration calorimetry (ITC)2,10, we found that c-di-AMP binds endothermically to ComFB with high affinity of a KD 2.6 ± 0.11 μM (Fig. 1e; Supplementary Fig. S6b), while no binding was observed for ATP, ADP, AMP, cAMP, and cGMP (Supplementary Fig. S7), indicating that ComFB binds c-di-AMP specifically. This result was confirmed using nanoDSF and thermal shift assays (Supplementary Figs. S8, S9), where c-di-AMP thermally stabilized ComFB in a concentration-dependent manner. Moreover, DRaCALA titration assays revealed strong binding of [32P]c-di-AMP to ComFB with a KD of 3.6 ± 5.4 µM (Fig. 1f). In the competition assays, the unlabeled c-di-AMP competed with [32P]c-di-AMP for binding to ComFB, which was not the case for ATP, ADP, and cAMP, confirming that c-di-AMP binding to ComFB is specific. However, this assay revealed that ComFB could additionally bind c-di-GMP, as c-di-GMP efficiently competed with [32P]c-di-AMP (Fig. 1g). Remarkably, a recent study showed that the ComFB homolog (named CdgR) controls cell size by binding c-di-GMP13 in the multicellular cyanobacterium Nostoc, which is regarded as being not naturally competent, implying that ComFB or CdgR might play different roles in multicellularity lifestyle.

To ascertain whether c-di-AMP binding to CdgR is also of physiological relevance, we performed a c-di-AMP-dependent pulldown but using Nostoc cell extract (Supplementary Fig. S3d). Indeed, we identified the ComFB homolog (CdgR; Alr3277) as one of the highly enriched proteins along with other known c-di-AMP receptor proteins2. This result further confirms that both ComFB and CdgR specifically bind both cyclic di-nucleotides in both organisms. Additionally, ComFB was found to bind c-di-GMP with comparable affinity (1.7 ± 0.5 µM) to that of c-di-AMP13. The existence of a crosstalk between c-di-AMP and c-di-GMP on ComFB awaits, however, further investigation. Crosstalk between second messenger nucleotides is perhaps a more common phenomenon than so far realized2,9 (see Supplementary Discussion).

To rule out that ΔdacA transformation deficiency (Fig. 1a) is due to a downstream effect on the intracellular c-di-GMP content, which is known to regulate motility-related functions1, we measured the c-di-GMP levels. The c-di-GMP levels were comparable between ΔdacA and WT cells within the light/dark phases (Supplementary Fig. S10), thus confirming that DNA uptake is influenced by c-di-AMP specifically.

To clarify whether ComFB plays a role in natural competence, we created a Δslr1970 deletion mutant (ΔcomFB; Supplementary Fig. S11). Like ΔdacA, ΔcomFB showed reduced transformation efficiency as compared to the WT and ΔsbtB10,11 cells (compare Fig. 1h with a; Supplementary Fig. S1c). Complementation of ΔcomFB restored the competence phenotype (Fig. 1h). Interestingly, no impairment in DNA uptake was observed for ΔsbtB10,11, which lacks another c-di-AMP receptor protein and showed a similar transformation efficiency to the WT cells (Fig. 1a; Supplementary Fig. S1c). These results further support that the natural competence depends on c-di-AMP-signaling and is controlled by a pathway that involves ComFB as a c-di-AMP-receptor. In contrast to ΔsbtB2,9,11, ΔcomFB did not show any impairment under diurnal growth (Supplementary Fig. S12), supporting the notion that c-di-AMP plays different signaling functions through binding to different receptors.

To gain insights into the molecular basis of how ComFB controls the natural competence, we carried out a comparative proteome analysis of ΔcomFB mutant (Supplementary Fig. S13). Surprisingly, PilA1 (Sll1694) was the most upregulated protein in ΔcomFB, implying a hyperpiliation phenotype like ΔdacA (Fig. 1b; Supplementary Fig. S2). Also, Sll1693 (methyl transferase potentially involved in PilA1 and PilA2 methylation) and Sll1696, which are part of pilA1 operon (sll1693–sll1696), showed upregulation. The minor pilin PilX2 (Slr0442) was also upregulated, further supporting the hyperpiliation hypothesis. The S-layer protein (Slr1704) and the outer membrane porin (Sll1550), which are required for the cell envelop and the correct assembly of pilus machinery, were also upregulated in ΔcomFB mutant. Moreover, the Deg protease (Sll1679), Sll0141 and Sll1581 were also upregulated in ΔcomFB mutant. The Deg protease (Sll1679) is known to be involved in motility and piliation6, while the Sll0141 is an accessory protein needed for pilin glycosylation and secretory machinery6 and the Sll1581 is important for the production of cell-surface exopolysaccharides.

Additionally, PixJ1 (Sll0041)6 and PixL (Sll0043)6 of TaxD1, and PilJ (Sll1294) and CheA (Sll1296)6 of TaxD2, which are motility-related proteins and involved in phototaxis6, were down-regulated in ΔcomFB. Consistent with PilA1 upregulation, it was reported previously that ΔpixL mutant shows pilA1 upregulation. The ΔpilJ mutant was shown previously to be non-transformable and non-motile. Moreover, several proteins (e.g., Sll0445, Sll0446, Slr0362, and Slr0442) which are under the control of LexA or SyCRP1/2, transcription factors known to regulate motility and pilus biogenesis6, were deregulated in ΔcomFB mutant. In analogy to ΔdacA, we detect a downregulation in PilN, which is part of pilMNO operon, explaining the reduced transformability of both ΔdacA and ΔcomFB mutants. To further confirm this result, we analyzed the transcript levels of pilM using RT-PCR. In fact, pilM was downregulated in both ΔcomFB and ΔdacA mutants (Supplementary Fig. S14a). Moreover, we could not detect comFB mRNA in ΔcomFB, while hfq showed a normal expression in all strains. Since the levels of hfq mRNA were similar to that of the WT, we can safely assume that comFB mutation does not cause a polar effect on the upstream hfq gene, which is required for pilus assembly6, highlighting the specific effect of comFB and dacA mutations on pilMNO operon. As a negative control, we checked for pilB1 mRNA, which did not change.

To test whether ΔcomFB causes hyperpiliation analogs to ΔdacA, we examined negatively stained ΔcomFB cells by TEM. Indeed, ΔcomFB cells were hyperpiliated (Supplementary Fig. S2), and PilA1 quantification revealed a strong accumulation of PilA1 in ΔcomFB similar to ΔdacA (Supplementary Fig. S14b, c). Altogether, these results confirm that c-di-AMP and ComFB are involved in a similar pathway with both being required for pilus biogenesis and natural competence.

Finally, to determine whether the cellular role of ComFB is conserved among cyanobacteria, we created a ΔcomFB mutant in S. elongatus (Synpcc7942_1924; Supplementary Fig. S11). The DNA uptake was impaired in this strain (Supplementary Fig. S1e, f), confirming a conserved role for ComFB in natural competence. To gain insight into the pathways that ComFB could coordinate in S. elongatus, we checked for co-fitness scores of an Random Barcode Transposon Insertion Site Sequencing (RB-TnSeq) library, which indicates likelihood that two genes participate in similar pathways and respond alike under different growth conditions14 (see Supplementary Discussion). Genes, which possess co-fitness values > 0.75 in the RB-TnSeq library, are considered to possess robust co-fitness and likely to participate in similar pathways. S. elongatus ΔcomFB mutant showed a very high co-fitness (0.85–0.98) for mainly genes involved in natural competence and pilus machinery, including pilA, pilA2, pilB1, pilC, and rntAB (Synpcc7942_2484–2486) operon (Supplementary Fig. S15). Similar to Synechocystis ΔcomFB, S. elongatus ΔcomFB showed also a strong association with pilMNOQ operon (Synpcc7942_2450–2453). Also, several regulatory genes of pilus biogenesis (e.g. hfq, sigF1 and esbA) showed strong co-fitness association with S. elongatus ΔcomFB mutant15. This result further confirms that ComFB is a new player in controlling cyanobacterial natural competence and pilus biogenesis.

In conclusion, our results show that the regulation of pili biogenesis and natural competence is a new unexplored role of c-di-AMP, which requires the receptor protein ComFB. In a broader context, natural competence is a primary mode of horizontal gene transfer, which plays an important role in spreading multidrug resistance. It would therefore be highly interesting to determine whether the influence of c-di-AMP and ComFB signaling on DNA uptake and/or motility-related functions extends to other bacteria, especially those of clinical relevance. Collectively, we identified ComFB as a novel widespread c-di-NMP-receptor, which turned out to be a pivotal competence-accessory protein, at least in cyanobacteria, regulating the pili biogenesis.