ABUW_2208 is a negative regulator of A. baumannii biofilm formation

A. baumannii is able to form robust biofilms that help it to colonise surfaces and tolerate antibiotic exposure. However, the mechanisms behind the regulation of biofilm formation in this pathogen are not fully understood. To identify biofilm regulators, we screened the Manoil A. baumannii AB5075 transposon mutant library to identify mutants with an altered biofilm phenotype [24]. After screening more than 10,600 mutants, amongst the hits were three different transposon mutants, bearing independent insertions at different positions within the same gene, ABUW_2208, a predicted adenylate/guanylate cyclase. To further validate the phenotype, all three ABUW_2208::T26 transposon mutants were tested for biofilm formation using a standard microtiter dish assay (Fig 1A). The transposon mutant strains AB05781, AB05784 and AB05783 had significantly higher biofilm biomass than the wild-type (WT) AB5075 (Fig 1A) validating the findings of the screen. Amplification of ABUW_2208 gene by PCR confirmed the T26 transposon insertion in the gene in each of the three strains (S1A Fig). This indicated that a dysfunctional ABUW_2208 enhanced A. baumannii biofilm formation, highlighting it as the most reproducible candidate gene and a potent negative regulator of this phenotype.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CavA (ABUW_2208) is a regulator of biofilm formation and the main functional adenylate cyclase in A. baumannii.

A—Biofilm formation of wild-type (WT) AB5075 compared to three ABUW_2208::T26 transposon mutants labelled with their corresponding number from the Manoil transposon mutant library. B—Proposed cell localisation and function of CavA adenylate cyclase based on domains present in the protein. The transmembrane domains anchor the protein to the inner membrane the cells. The catalytic domain forms a dimer and converts ATP molecules to cyclic AMP (cAMP). Figure was created using PyMOL and BioRender.com. C–Biofilm levels of ΔcavA clean deletion mutant and ΔcavA+cavA complemented strain. D–cAMP concentrations in ΔcavA mutant and complemented ΔcavA+cavA strain, as well as ΔcavB and ΔcavAΔcavB double mutant which was complemented with either cavA (ΔcavAΔcavB+cavA) or cavB (ΔcavAΔcavB+cavB). Cultures grown in LB broth were harvested and lysed via sonication. Bradford assay was used to quantify the total protein concentration and cAMP concentrations in each strain were determined using Cyclic Nucleotide XP Enzymatic Immunoassay kit. Data shown is the mean cAMP concentration per milligram protein from three independent repeats. E—Biofilm formation of ΔcavB deleted mutant compared to WT. Biofilm biomass was assessed after 24 h incubation at 37°C shaking and is presented as the optical density measured at 570 nm (OD₅₇₀). All data represents the averages of three biological replicates ± standard deviations (SD). Bacterial growth was assessed at OD₆₀₀ prior to staining the biofilms (S1 Fig). ns p>0.05, *p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001—One-Way ANOVA with Dunnett (A) or Tukey (C, D) post-hoc tests and Unpaired t-test (D). Strains with empty miniTn7 were used as controls (S1 Fig).

https://doi.org/10.1371/journal.ppat.1012529.g001

ABUW_2208 is the primary functional adenylate cyclase in AB5075

To understand the functional role of ABUW_2208, we investigated its predicted structure and predicted cellular localisation. We found that ABUW_2208 (489 residues) is the only annotated adenylate/guanylate cyclase (AKA31936.1) in A. baumannii AB5075-UW genome (CP008706.1) and is annotated as CyaA in UniProt (A0A059ZQ67). Protein sequence analysis showed that, towards its N-terminus, ABUW_2208 contains several regions predicted to be embedded in the membrane (S1B Fig). Towards its C-terminus, there is an Adenylate and Guanylate cyclase catalytic domain (PF00211) which is part of class III nucleotidyl cyclases and is predicted to dimerise. It contains a metal-binding site, as well as an active site where the conversion of ATP to cAMP occurs (Fig 1B). On the other hand, CyaA is a cytoplasmic class I AC in P. aeruginosa [25]. Therefore, according to its domain architecture, it would be inappropriate to refer to ABUW_2208 as CyaA. Thus, due to its putative cyclase function in A. baumannii and role in virulence, we renamed ABUW_2208 to CavA.

In order to eliminate the possibility of polar effects or off-target transposon insertions in the validated cavA::T26 mutants that may affect their biofilm phenotype, a clean deletion mutant and complemented strain were constructed using a previously developed genome editing strategy for MDR clinical isolates of A. baumannii [26]. As observed with the transposon mutants (Fig 1A), ΔcavA produced significantly more biofilm than the wild-type AB5075 (Fig 1C). The effect on the phenotype was reversed upon chromosomal complementation of the ΔcavA mutant, where ΔcavA+cavA biofilm levels were significantly lower than those of the deleted mutant and the WT (Fig 1C). The miniTn7 empty vector used for the complementation of the mutant was inserted in the WT AB5075 and ΔcavA and had no effect on the bacterial growth or biofilm biomass (S1C and S1D Fig). These results undoubtedly show that CavA plays a key role in the regulation of biofilm formation in A. baumannii.

We next sought to confirm if CavA was indeed a functional adenylate cyclase by measuring the cAMP levels in the deletion mutant compared to the WT and complemented strains. The ΔcavA mutant had the expected significantly lower cAMP levels than the WT (Fig 1D). This was reversed and cAMP concentrations were brought back to WT levels upon complementation in ΔcavA+cavA strain (Fig 1D), confirming the adenylate cyclase function of CavA enzyme. In addition, our data showed that expressing cavA in the WT background (WT+cavA) did not significantly increase cAMP concentrations than the WT (S1E Fig). However, as the deletion of cavA would drop the cAMP concentration compared to the WT, this suggests that under the conditions tested cavA is constitutively functional, leading to high basal cAMP levels in AB5075. Overall, our data demonstrates for the first time that there is a direct link between cAMP levels and surface associated biofilm formation in MDR A. baumannii.

Apart from cavA, which was annotated as an adenylate/guanylate cyclase, there is another adenylate cyclase annotated in the AB5075 genome within the ABUW_1085 locus (AKA30837.1). In agreement with the cavA renaming, we renamed it to cavB. Despite their similar size and functional annotation, no significant similarity was found when CavA and CavB protein sequences were aligned. Unlike CavA, CavB is not predicted to have any transmembrane domains and has a CYTH domain (PF01928) with an active site and metal binding sites near its N-terminus, as well as a CHAD domain (PF05235) near the C-terminus (S1F Fig). Nevertheless, according to their annotation, it is predicted that CavB function is similar to that of CavA in synthesising cAMP. To assess this, we created a cavB clean deletion mutant as well as a double deletion mutant of cavA and cavB (ΔcavAΔcavB) utilising the same strategy as before [26] and tested the cAMP concentrations and biofilm levels. The results demonstrated that deletion of cavB did not significantly alter cAMP concentrations compared to the WT (Fig 1D). In addition, the absence of cavB in the ΔcavA background (ΔcavAΔcavB) did not decrease the cAMP concentrations further than the single cavA mutant (ΔcavA). Complementation of the double ΔcavAΔcavB mutant with cavA, but not with cavB, restored cAMP concentrations to levels comparable to the WT (Fig 1D). This clearly indicates that, under the conditions tested, CavA is the only functional adenylate cyclase in A. baumannii. Thus, in agreement with our results presented above, linking biofilm formation and cAMP levels, the deletion of cavB did not alter bacterial growth (S1G Fig) or biofilm formation in comparison to the WT (Fig 1E). This confirms the major role of CavA in cAMP production and biofilm formation regulation of A. baumannii

In order to assess the prevalence of CavA and CavB, we analysed the frequency of all possible CavA and CavB variants across 9,696 A. baumannii genomes. Our findings revealed that both proteins are highly conserved, but between the two, CavA is less likely to undergo mutations, as fewer variants were identified of this protein compared to CavB (S2A Fig). This was observed regardless of the isolation site of the strains analysed, reinforcing the hypothesis of the major regulatory role of CavA (S2B Fig). In addition, CavB appears to have a different variant in environmental isolates–only 25% have the most frequent variant in humans and hospitals (S2B Fig).

cAMP mediates transcription of various virulence-related genes in A. baumannii

cAMP is a second messenger known to act as a central signalling molecule in other species, regulating multiple processes simultaneously. To unravel what other phenotypes, apart from biofilm formation, could be regulated by cAMP, we performed a differential RNA sequencing (dRNA-seq) analysis of the ΔcavA mutant compared to the complemented mutant strain (i.e. with the gene reintroduced in a neutral chromosomal site). Cells were grown to mid-log phase before RNA extraction and subsequent sequencing and comparison. Our results show that 234 genes were differentially expressed by |Log2(Fold Change)| ≥ 1 in high cAMP conditions (ΔcavA+cavA) compared to low cAMP conditions (ΔcavA empty miniTn7 mutant) (S3 Table). Of these genes 143 were up-regulated and 91 were down-regulated (Fig 2A). To elucidate the main functional gene classes among the differentially expressed genes by cAMP, we performed a Gene Set Enrichment Analysis (GSEA) using FUNAGE-Pro [27]. The GSEA results revealed that cell adhesion and type IV pilus-mediated motility were affected (S4 Table). Our dRNA-seq data showed that genes linked to biofilm formation such as csuA/BABCDE operon, fimbriae coding genes (ABUW_2052–2055) and EPS production related genes (pgaABC) were down-regulated by cAMP. Simultaneously the expression of type IV pili genes (pil and com gene clusters), involved in twitching motility and natural transformation, was increased by this second messenger (Fig 2A). Altogether, our dRNA-seq dataset shows that cAMP acts as a global signalling molecule orchestrating antagonistic cell behaviours and virulence factors in A. baumannii, such as biofilm formation and motility, and multiple sub-processes that are involved in them.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The effect of CavA and cAMP on A. baumannii transcriptome, EPS production and motility.

A—Volcano plot showing dRNA-seq data of global transcription of ΔcavA+cavA complemented strain (having high cAMP levels) compared to the control ΔcavA EV deleted mutant with empty vector (having low cAMP levels). The results indicated 234 differentially expressed genes in total, of which 143 genes were up-regulated (red dots) and 91 were down-regulated (blue dots). Highlighted are up-regulated genes linked to twitching motility (pil and com gene clusters), cAMP signalling (vfr) and c-di-GMP synthesis (ABUW_2135) and degradation (ABUW_1138) and down-regulated genes linked to biofilm formation and attachment (ABUW_2052–55 and csu operons), exopolysaccharides production (pga gene cluster), quorum sensing (abaI), c-di-GMP degradation (ABUW_2631). The crp homologue ABUW_2344 was amongst the not significantly differentially expressed genes (black dots). B & C—The impact of different cAMP levels in ΔcavA mutant and complemented ΔcavA+cavA strain compared to wild-type (WT) AB5075 on exopolysaccharide (EPS) production (B) and twitching motility (C). For the EPS assay colonies of each strain were grown in triplicates on Congo red (40 μg/ml) agar plates at 37°C for 5 days and the images shown are representatives of three independent repeats (B). The twitching zone diameter was measured in millimetres 48 h post-inoculation of soft agar plates. Data represents the mean ± SD of three biological repeats. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001—One-Way ANOVA with Tukey post-hoc test. Strains with empty miniTn7 were used as controls in both assessments (S3A and S4A Figs).

https://doi.org/10.1371/journal.ppat.1012529.g002

cAMP inversely regulates biofilm and motility in A. baumannii

EPS is a major component of bacterial biofilm extracellular matrix [28]. As indicated by the dRNA-seq, expression of the pgaABCD operon, encoding the synthesis machinery for the major exopolysaccharide poly-β-1,6-N-acetylglucosamine (PNAG) in A. baumannii [29], is significantly decreased in high cAMP conditions (Fig 2A). To phenotypically validate this link between cAMP and EPS production, the CavA-related strains were grown on Congo red agar plates for 5 days. As seen in this result (Fig 2B), and in accordance with the transcriptomics results, ΔcavA mutant colonies accumulated more Congo red stain than the WT colonies, which indicates a greater production of EPS. This phenotype was reversed upon complementation in ΔcavA+cavA strain (Fig 2B), as the empty vector used for complementation did not affect EPS production (S3A Fig). These results further support our data presented above indicating the negative impact of cAMP on A. baumannii biofilm formation (Fig 1C). To validate that the effect of CavA on EPS production was attributed to the differential regulation of the pgaABCD operon and PNAG production, a ΔcavA/pgaA::Tn double mutant was created. The results showed a substantial decrease in the accumulated Congo red stain in the double mutant colonies compared to the single ΔcavA mutant (S3B Fig). Hence, cAMP represses expression of EPS related genes, resulting in decreased EPS production by A. baumannii. Further to this, assessment of ΔcavA/pgaA::Tn and ΔcavA/csuC::Tn double mutants showed significant decrease in the elevated biofilm levels caused by the deletion of cavA (S3C and S3D Fig). However, disruption of pgaA, or csuC, in a cavA mutant, did not decrease the high biofilm biomass to WT levels. This suggests the cumulative effect of the increased expression levels of these genes underpins the increased biofilm formation seen in the ΔcavA mutant. Our data highlights that cAMP control of biofilm formation is multifactorial and involves transcriptional regulation of different genes including csu and EPS production related genes.

Biofilm formation and twitching motility are known to be inversely regulated antagonistic lifestyles [20]. In the experiments above, we have established that high cAMP negatively regulates biofilm formation (Figs 1C and 2B). This, together with our findings showing an up-regulation of the type IV pili related genes in high cAMP conditions (Fig 2A), strongly indicated that this second messenger could play a role in the regulation of twitching motility in A. baumannii. In order to validate this cAMP regulation, we assessed A. baumannii motility in the ΔcavA mutant compared to the wild-type strain using a twitching assay (Fig 2C). In accordance with the dRNA-seq results, we observed significant decrease in ΔcavA mutant twitching motility. Chromosomal complementation of the deleted cavA mutant restored motility (Fig 2C), as the empty miniTn7 vector did not affect the phenotype (S4A Fig). In addition to the impairment in twitching motility, we could also observe that the natural transformation capability of the ΔcavA mutant decreased below the detection limit of our natural transformation assay compared to the wild type (S4B Fig). This clearly demonstrates the key role of cAMP in promoting transformability and motility in A. baumannii via activation of type IV pili gene expression to the detriment of biofilm formation.

Vfr regulates biofilm, motility and EPS in a cAMP-dependent manner in A. baumannii

In other bacterial species, such as Pseudomonas aeruginosa and Escherichia coli, several cAMP effector proteins have been identified, including Vfr and CRP [30]. Upon binding to cAMP, Vfr undergoes a conformational change shifting this transcriptional factor from an inactive to active state [31]. In the case of A. baumannii AB5075, a vfr orthologue is annotated in its genome (ABUW_2741 locus, AKA32458.1), but the link between Vfr and cAMP signalling has not been previously established in this pathogen. Moreover, there is also a predicted crp orthologue annotated in AB5075 (ABUW_2344 locus, AKA32071.1) with a predicted cAMP-binding domain. Within our transcriptomic dataset, we identified that vfr expression was significantly increased (Log2(FC) = 1) in high cAMP conditions, whereas the predicted crp gene was not differentially expressed (Fig 2A). To obtain evidence that the AB5075 Vfr orthologue may bind cAMP, we compared its protein sequence to that of the P. aeruginosa PAO1 Vfr protein, for which the residues interacting with the cAMP molecule have been elucidated. As a result, we found that these residues are conserved, along with the DNA binding domain, thus hinting the link between Vfr and cAMP signalling in A. baumannii (S5A Fig). Moreover, alignment of Vfr^AB5075 and Vfr^PAO1 protein structure models resulted in a root mean square deviation (RMSD) of 1.94 indicating the structural similarity of the two proteins (S5B–S5D Fig). Furthermore, we analysed the conservation of Vfr across the A. baumannii pangenome, obtaining as a result that Vfr was present in 81% of the strains (S2A Fig). In contrast, a small proportion (19%) of the strains did not have Vfr, suggesting a possible role of additional alternate cAMP effectors, which requires further investigation. The majority of the strains had only two Vfr sequence variants, indicating a strong negative selection. Additionally, CavA and Vfr present co-conservation (maintaining the same pair of variants in the same strain) in the majority of the analysed strains (S2A Fig), suggesting a link between them in the physiology of A. baumannii.

To establish the role of Vfr as a cAMP effector, we first validated the transposon insertion in vfr::Tn (AB07172) transposon mutant via Sanger sequencing. We then complemented this mutant and verified that neither of the vfr-related derivatives had impaired growth and that the miniTn7 empty vector did not affect the tested phenotypes (S6 Fig). The elevated biofilm phenotype of the vfr::Tn mutant was reversed upon chromosomal complementation with vfr^WT (Fig 3A). To verify the cAMP control of Vfr, we generated a Vfr variant with point mutations in two key cAMP-binding residues (T138A and T139W) in the C-helix of the protein (S5E Fig), which is known to act as the activation mechanism of Vfr in P. aeruginosa [31]. In contrast to vfr^WT, vfr^T138A,T139W variant was impaired for complementation of the vfr::Tn mutant biofilm phenotype. Ectopic expression of vfr in the WT (WT+vfr) did not change biofilm levels compared to the WT and expression of vfr in ΔcavA mutant (ΔcavA+vfr) did not alter the high biofilm levels produced by ΔcavA (Fig 3A). These findings confirm that Vfr acts downstream of CavA on biofilm formation and that its function is cAMP mediated which is in accordance with what is seen in other bacteria [31].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Vfr negatively regulates biofilm formation, EPS production and motility of A. baumannii in a cAMP-dependent manner.

Biofilm formation (A), EPS production (B) and twitching motility (C) of vfr::Tn transposon mutant AB07171, its complemented strain vfr::Tn+vfr^WT, strains overexpressing vfr in the WT (WT+vfr) and ΔcavA backgrounds (ΔcavA+vfr) compared to AB5075 WT. In addition, vfr::Tn mutant was complemented with vfr variant harbouring two modified residues (T138A and T139W) in the cAMP binding site (vfr::Tn+vfr^T138A,T139W). A—Biofilm formation microtiter assay was used as described above and biofilms were grown in LB broth for 24 h at 37°C shaking. Growth of the strains was assessed and strains harbouring empty miniTn7 were used as control. B—Images of colonies on Congo red agar for EPS production assessment were taken 5 days post-inoculation. C—Motility was assessed using soft agar at 48 h as described in the materials and methods section. Data including all controls is presented in S6 Fig. All experiments were repeated three independent times. Bar graphs present averages ± SD. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001—One-Way ANOVA with Dunnet (A) and Tukey (B & D) post-hoc tests.

https://doi.org/10.1371/journal.ppat.1012529.g003

The fact that Vfr channels the CavA-mediated cAMP effect on biofilm formation made us hypothesize whether Vfr may be involved in other CavA-controlled behaviours, including EPS production and twitching motility. Further exploration of this regulation showed that vfr::Tn exhibited an increased EPS production similarly to the ΔcavA mutant (Fig 3B). As observed with biofilm formation, EPS production decreased when the vfr::Tn mutant was complemented with vfr^WT (vfr::Tn+vfr^WT), but not with vfr^T138A,T139W variant (vfr::Tn+vfr^T138A,T139W). Also, the increased EPS production by ΔcavA was unaffected when vfr expression was enhanced in this background (ΔcavA+vfr). Furthermore, we assessed twitching motility in the vfr::Tn mutant compared to the WT and the ΔcavA mutant and determined that vfr::Tn has a similar twitching profile to that of the ΔcavA mutant, with a significant decrease in motility compared to the WT (Fig 3C). Complementation of the vfr::Tn strain with vfr^WT, but not with vfr^T138A,T139W, restored motility to WT levels. Similarly to our observations on the biofilm formation (Fig 3A), increased expression of vfr in the WT (WT+vfr) or ΔcavA mutant background (ΔcavA+vfr) phenocopied WT and ΔcavA motility respectively (Fig 3C). Additionally, disruption of vfr in the double ΔcavA/vfr::Tn mutant did not further alter the ΔcavA biofilm formation, EPS production and motility phenotypes (Fig 3). This was also confirmed by assessing the expression of pilA gene and pgaABCD operon using PpilA::gfpmut3 and Ppga::gfpmut3 transcriptional fusions respectively. Our data showed that the expression of pilA and pgaABCD was significantly decreased in vfr::Tn mutant as well as ΔcavA compared to the WT (S7 Fig).

Altogether, these findings confirm that Vfr function is dependent on the presence of the major adenylate cyclase in A. baumannii, CavA, and highlight the cAMP specificity in Vfr functionality. This clearly demonstrates that, similarly to P. aeruginosa [31], Vfr acts as a cAMP effector in A. baumannii as well.

cAMP negatively regulates QS via abaI expression

Another interesting observation from our dRNA-seq dataset was that the sole QS autoinducer synthase gene encoded in AB5075, abaI, had decreased expression in high cAMP conditions (Fig 2A). Similarly to observations in other bacterial species where a link between second messenger and quorum sensing signalling systems have been reposted [32,33], our dRNA-seq data suggests that there may be an interplay between intracellular second messenger signalling and intercellular communication in A. baumannii too. To explore this further, we sought to verify the AHL production, and thus the QS regulation, by CavA at the phenotypic level using an Agrobacterium tumefaciens AHL biosensor strain (Figs 4 and S8). We could indeed observe that the ΔcavA mutant produced greater amounts of AHL compared to the WT strain, indicating that cAMP is able to reduce QS via transcriptional regulation of abaI (Fig 4). Coherently, the complementation of the cavA mutation abolished the AHL production restoring the WT phenotype. Together, this conclusively indicates that the CavA-mediated cAMP second messenger signalling is able to modulate a primary intercellular communication system as QS in A. baumannii.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Qualitative assessment of AHL secretion by AV-T variants of WT A. baumannii AB5075, deleted ΔcavA mutant and ΔcavA+cavA complemented strain.

Cultures of each strain were spot plated on a soft agar mixed with A. tumefaciens traG-lacZ biosensor strain and supplemented with X-Gal and IPTG. The blue halo indicates AHL production by A. baumannii which triggers the cleavage of X-gal by the A. tumefaciens biosensor strain. Empty miniTn7 controls are presented in S8 Fig.

https://doi.org/10.1371/journal.ppat.1012529.g004

cAMP regulates c-di-GMP Levels in A. baumannii

Second messengers, as signalling molecules, are known to overlay their regulatory targets to modulate the global phenotypic output of the cell. One well-established model in the bacterial cell signalling field is that high cAMP levels correlate with high motility and low biofilm formation (planktonic lifestyle), whereas high c-di-GMP levels are associated with low motility and high biofilm formation (sessile lifestyle) [34].

Our dRNA-seq dataset highlighted that the expression of several genes, that encode both predicted diguanylate cyclases and phosphodiesterases, was increased in high cAMP conditions. This includes the DGC ABUW_2135 and the PDE ABUW_1138 (Fig 2A). We could also observe a decreased expression of the predicted PDE ABUW_2631, however, this is annotated as a pseudogene in the reference AB5075 genome [24], hinting it would probably yield a non-functional protein. Nevertheless, although DGCs and PDEs have opposite effects on c-di-GMP concentrations, our transcriptomics dataset suggested that cAMP may have an effect on the c-di-GMP levels.

To test this, we aimed to quantify the c-di-GMP levels in our mutant and complemented strains, as well as in the WT. To perform these assays, we adapted the previously reported and validated CensYBL c-di-GMP fluorescent biosensor [35] for its use in Acinetobacter. This biosensor is based on the c-di-GMP binding protein BldD, which is able to dimerise via interaction with this second messenger molecule. In the CensYBL biosensor, two copies of the bldD genes are fused to the 5’ and 3’ halves of the fluorescent protein coding gene yfp. After their expression, the BldD fragments of the N-YFP::BldD and C-YFP::BldD fusion proteins will dimerise in the presence of c-di-GMP. This would reconstitute the YFP function, thus producing a level of fluorescence that is proportional to the c-di-GMP levels in the cell. We adapted this biosensor to Acinetobacter by assembling the construct into an Acinetobacter-compatible pWH1266 derivative plasmid (pWH1266-Apr). Furthermore, we cloned the biosensor coding genes downstream of an anhydrotetracycline-inducible promoter. With these modifications, we could not only use the biosensor in any Acinetobacter species, but more specifically, we could select it in any mutant from the Manoil transposon mutant library and induce the biosensor independently of the set of expression systems we routinely use for gene complementation (IPTG-inducible). To validate the functionality of the modified CensYBL biosensor (CensYBL-Ab) in Acinetobacter, we introduced the heterologous DGC coding gene pleD* and the PDE rocR under an IPTG-inducible promoter in AB5075 via miniTn7 insertion. As a result, we could validate our modified c-di-GMP biosensor by measuring a signal increase or decrease after the induction of pleD* or rocR, respectively, compared to an empty vector control (S9A Fig).

In order to test if cAMP levels affect the global c-di-GMP levels in AB5075, we introduced our CensYBL-Ab biosensor in the ΔcavA mutant, as well as in the transposon mutant in the cAMP effector vfr. Strikingly, absence of CavA (Fig 5A) or Vfr (Fig 5B) significantly decreased c-di-GMP levels compared to the WT, with c-di-GMP reduction levels in the ΔcavA and vfr::Tn mutants of 66% and 49% respectively. The c-di-GMP levels were partially restored in the ΔcavA+cavA strain upon cavA complementation, as the empty miniTn7 vector used for the complementation had no effect (S9B Fig). Our data demonstrates that the CavA-Vfr regulatory cascade modulates the c-di-GMP levels in A. baumannii. From a broader perspective, our results challenge a previously established paradigm, presenting cAMP and c-di-GMP as antagonistic and inversely regulated signalling molecules. Instead, our findings demonstrate that the c-di-GMP levels are not only positively regulated by cAMP, but also suggest that they are possibly hierarchically under the control of the Vfr-mediated cAMP regulation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. CavA and Vfr are regulators of global intracellular c-di-GMP levels in A. baumannii.

Cyclic di-GMP levels measured using CensYBL-Ab biosensor in ΔcavA mutant with low cAMP and ΔcavA+cavA complemented strain with high cAMP levels (A) as well as vfr::Tn mutant (B) compared to WT AB5075. Deletion of cavA reduced c-di-GMP levels by 66% (A) while disruption of vfr decreased them by 49% (B). Cultures were diluted in LB broth supplemented with apramycin (100 μg/ml) and biosensor expression was induced with anhydrotetracycline (50 ng/ml). Harvested cells were resuspended in sterile PBS and YFP and mCherry fluorescence signals were measured. The average normalised fluorescence (YFP/mCherry) ± SD from three independent biological repeats is presented. Data was analysed using One-Way ANOVA with Tukey post-hoc test (A) and Unpaired t-test (B) (* p<0.05, ** p<0.01). Strains harbouring empty miniTn7 were used as controls (S9B Fig).

https://doi.org/10.1371/journal.ppat.1012529.g005

cAMP regulates A. baumannii resistance to fosfomycin

Given the role of antibiotic resistance in the ascendancy of A. baumannii to the top of the WHO priority pathogen list [36], we sought to assess if cAMP levels could influence antibiotic resistance in the MDR isolate AB5075. Initially we tested CavA role in A. baumannii resistance to different classes of antibiotics including quinolones (ciprofloxacin), polymyxins (colistin), amphenicols (chloramphenicol) and aminoglycosides (tobramycin) but we saw no significant difference between ΔcavA and WT (S10A Fig). It has been previously suggested that cAMP regulation could be involved in the antibiotic resistance phenotype of MDR A. baumannii clinical isolates. For example, mutations in cavA orthologous genes carried in A. baumannii clinical isolates genomes hinted a link to fosfomycin resistance [37]. However, a phenotypic link between cAMP and fosfomycin resistance has never been established in this pathogen. Hence, we sought to investigate this further using a disc diffusion assay. The results showed that lack of CavA, i.e. low cAMP conditions, or Vfr (in ΔcavA and vfr::Tn mutants respectively) significantly increased the zone of inhibition compared to the WT (Fig 6) and thus, increased AB5075 susceptibility to fosfomycin. This phenotype was reversed upon complementation in the ΔcavA+cavA and vfr::Tn+vfr strains (Figs 6 and S10B). As observed with other phenotypes, increased expression of vfr in the absence of CavA did not change the increased fosfomycin resistance of ΔcavA mutant (Fig 6). Altogether, this demonstrates that the cAMP singalling system may drive resistance to antibiotics, such as fosfomycin, in MDR A. baumannii. Furthermore, this cAMP-mediated regulation is exerted via Vfr.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. A. baumannii resistance to fosfomycin is regulated by CavA and Vfr.

Disc diffusion assay for testing fosfomycin resistance of WT AB5075, ΔcavA deleted mutant, vfr::Tn transposon mutant and their complemented derivative strains (ΔcavA+vfr and vfr::Tn+vfr respectively), as well as ΔcavA mutant overexpressing vfr (ΔcavA+vfr). CAMH agar inoculated with each corresponding strain and having a fosfomycin (50 μg) disc were assessed after 24 h at 37°C. Representative images (A) and the measured zone of inhibition in millimetres (B) are shown. Data represents the averages ± SD of biological triplicates. ns p>0.05, *p<0.05, ***p<0.001, ****p<0.0001 –One-Way ANOVA with Dunnet post-hoc test (B). All controls used are presented in S10B Fig.

https://doi.org/10.1371/journal.ppat.1012529.g006

cAMP mediates A. baumannii virulence in Galleria mellonella

As we have shown above, cAMP regulates virulence associated phenotypes such as biofilm formation and motility, and controls other signalling systems, all of which are related to A. baumannii virulence. In addition, genes related to twitching motility had increased expression in the presence of CavA (Fig 2A). As motility has been known to promote A. baumannii virulence [38], our transcriptomic data clearly suggested that reducing the cAMP levels (i.e. mutating cavA) would attenuate the virulence of A. baumannii AB5075. To confirm the role of cAMP in A. baumannii virulence regulation, we used the G. mellonella in vivo infection model. All larvae were infected with the WT or derivative mutants (a control group was injected with sterile PBS) and larvae survivability was assessed over two days. At 48 h post-infection 40% of larvae infected with WT AB5075 survived, whereas the survival rates were significantly increased in the groups infected with ΔcavA (with lower cAMP levels) and vfr::Tn mutants (63% and 77% respectively, Fig 7). Chromosomal complementation of cavA (ΔcavA+cavA) or vfr (vfr::Tn+vfr) significantly decreased survival rates (40% and 27% respectively) compared to the ΔcavA and vfr::Tn strains harbouring empty miniTn7 vector (EV) (70% and 63% respectively). These results demonstrate that high cAMP levels, and their Vfr-mediated regulation, are key for modulating the in vivo virulence of the priority pathogen A. baumannii.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Cyclic AMP and Vfr increase the virulence of A. baumannii.

Virulence assay was performed three independent times using 30 G. mellonella larvae in total per A. baumannii strain or control (PBS). 48 h post-infection, 100% of the larvae in the control group survived. G. mellonella infected with wild-type (WT) AB5075 had survival rates of 40% while 63% and 77% of the larvae survived infection with ΔcavA mutant and vfr::Tn mutant respectively. Survival rates of larvae infected with the complemented ΔcavA+cavA and vfr::Tn+vfr strains were decreased to 40% and 27% respectively. Strains bearing empty miniTn7 vector (EV) were used as controls. In the interest of visual clarity, the X axis (Time) was set to start at 15 h. Data analysis with Log rank (Mantel-Cox) test indicated p<0.05 for the comparisons of WT vs ΔcavA and ΔcavA EV vs ΔcavA+cavA and p<0.01 for the comparisons of WT vs vfr::Tn and vfr:Tn EV vs vfr::Tn+vfr.

https://doi.org/10.1371/journal.ppat.1012529.g007

Bacterial strains and growth conditions

A. baumannii AB5075-UW and its transposon mutants were obtained from the Manoil laboratory three-allele library [24]. Virulent opaque (VIR-O) variants of the wild-type and its derivative mutants were used due to their clinical relevance, unless otherwise stated [50]. Plasmids were stored in E. coli DH5α or DH5α λpir hosts when necessary. Cultures were grown routinely at 37°C shaking (180 rpm) or statically in LB broth or on LB agar respectively, unless stated otherwise. When required, the medium was supplemented with X-Gal (25 μg/ml), ampicillin (100 μg/ml), apramycin (100 μg/ml for E. coli and 200 μg/ml for A. baumannii), kanamycin (25 μg/ml), tellurite (6 μg/ml and 30 μg/ml for E. coli and A. baumannii respectively) and tetracycline (5 μg/ml). T26 transposon insertion in vfr transposon mutant AB07172 was validated using semi-degenerate PCR as described before [51]. Briefly, T26 fw1 oligo was used in combination with CEKG-2A, CEKG-2B and CEKG-2C for the first stage of the PCR. The PCR product was then amplified with T26 fw2/CEKG-4 primer pair in the second stage of the reaction. DNA fragments with different sizes were purified and sequenced using Sanger sequencing. The sequence of each fragment was aligned to the AB5075 genome to establish T26 transposon insertion position.

Plasmid construction

All plasmids used in this work are listed in S1 Table. For clean deletion mutant construction, we followed the double-recombination strategy described by de Dios et al. involving the construction of pEMGT plasmid derivatives [26]. Approximately 1-kb regions flanking cavA (ABUW_2208), cavB (ABUW_1085) and abaI were amplified from genomic DNA (gDNA) of AB5075 using cavA up fw/rv and down fw/rv, cavB up fw/rv and down fw/rv and abaI up fw/rv and down fw/rv pairs of primers to amplify the upstream and downstream regions of each respective gene. The flanking regions of each gene were joined together via an overlapping PCR and cloned into pEMGT digested with SmaI, resulting in pEMGT-cavA, pEMGT-cavB and pEMGT-abaI.

For gene expression, a miniTn7 transposon derivative bearing the IPTG-inducible lacI^q-Ptac expression system (pUC18T-miniTn7T-Tc-lacI^q-Ptac [52], termed pJM101-Tc) was utilised. For similarly inducing gene expression in Tc-resistant genetic backgrounds (i.e. strains from the Manoil transposon mutant library), we generated a pUC18T-miniTn7T-Gm [53] (Addgene #63121) derivative bearing a tellurite resistance cassette. This cassette was obtained by cloning a fragment from pMo130-Tel [54] (Addgene #50799) cut with SmaI (containing the tellurite marker) into pUC18T-miniTn7T-Gm-lacI^q-Ptac [55] digested with EagI and BsrGI and blunted with Klenow. This construction resulted in pUC18T-miniTn7T-Tel-lacI^q-Ptac, termed PJM101-Tel.

The cavA gene was amplified from AB5075 gDNA using the cavA fw/rv oligos and digested with PstI and KpnI. The PCR product was then cloned in pUC18T-miniTn7T-Tc-lacI^q-Ptac [52] digested with the same restriction enzymes (REs) resulting in pUC18T-miniTn7T-Tc-lacI^q-Ptac::cavA.

To amplify the vfr (ABUW_2741) gene, the vfr fw RBS PstI/rv HindIII primer pair was used and the resulting product was digested with PstI and HindIII. It was then cloned in pUC18T-miniTn7T-Tel-lacI^q-Ptac digested with the same REs. The derivative plasmid was pUC18T-miniTn7T-Tel-lacI^q-Ptac::vfr. To generate a version of vfr carrying mutations in the cAMP binding site (T138A and T139W in the AB5075 vfr allele) as previously reported in P. aeruginosa [31], a joining PCRs were performed. The 5’ and 3’ regions of the vfr gene (approx. 0.4 kb each) were amplified using the vfr fw RBS PstI/vfr T138A T139W intra rv and vfr T138A T139W intra fw/vfr rv HindIII primer pairs, respectively, for introducing the T138A, T139W point mutations. The amplifications generated two overlapping fragments of the vfr gene containing the mutations. The overlapping fragment pair was assembled together by joining PCR, and the resulting mutated vfr gene fragment was digested with PstI and HindIII. This fragment was ligated into pUC18T-miniTn7T-Tel-lacI^q-Ptac digested with the same enzymes, obtaining pUC18T-miniTn7T-Tel-lacI^q-Ptac::vfr^T138A,T139W.

For the construction of Ppga promoter fused to gfpmut3, 1-kb fragment including the promoter region was amplified via PCR using Ppga fw EcoRI/Ppga rv BamHI oligo pairs, digested with EcoRI and BamHI restriction enzymes and cloned in pUC18T-miniTn7T-zeo-gfpmut3 [56] (Addgene #65037) vector digested with the same enzymes. The obtained plasmid was pUC18T-miniTn7T-zeo-Ppga::gfpmut3. To assess pilA expression pUC18T-miniTn7T-zeo-PpilA::gfpmut3 [52] plasmid was used.

To increase the internal c-di-GMP levels in AB5075, pUC18T-miniTn7T-Tc-lacI^q-Ptac and the heterologous DGC coding gene pleD* (constitutively active) [57] was generated. A DNA fragment containing the pleD* coding sequence plus a ribosome binding site was amplified from pMRB165 [58] with primers RBS fw PstI/pleD rv HindIII, digested with PstI and HindIII and cloned into pUC18T-miniTn7T-Tc-lacI^q-Ptac digested with the same enzymes, resulting in pUC18T-miniTn7T-Tc-lacI^q-Ptac::pleD*. To decrease c-di-GMP levels similarly, a pUC18T-miniTn7T-Tc-lacI^q-Ptac derivative bearing the PDE coding gene rocR [59] was generated. The rocR coding sequence was amplified from P. aeruginosa DSM 50071T [60] using primers rocR RBS fw HindIII/rocR rv (the same ribosome binding site as for the pleD* construction was used by including it in the forward primer), digested with HindIII and cloned into pUC18T-miniTn7T-Tc-lacI^q-Ptac cut with HindIII and NruI, resulting in pUC18T-miniTn7T-Tc-lacI^q-Ptac::rocR.

To generate a pWH1266 [61] derivative that could be selected in MDR AB5075 and its derivative transposon mutants from the Manoil library, an apramycin resistance marker was PCR amplified from pFLAG-attP (Addgene, #110095) using primers Apr fw/Apr rv [26] and cloned into pWH1266 digested with ScaI, resulting in pWH1266-Apr.

To construct a variant of the CensYBL c-di-GMP biosensor [35] within an A. baumannii replicative vector, a multi-step cloning strategy was followed. First, a DNA fragment containing the tetR-tetP expression system plus a ribosome binding site was amplified from pYDE009 [62] with primers tetR RBS fw NdeI/tetR rv NruI and ligated into pBluescript-II-SK(+) (pBSK) cut with EcoRV. The orientation of the fragment was assessed to select a clone bearing the tetP promoter neighbouring the HindIII restriction site in the cloning vector, whereas the 3’ end of tetR would be neighbouring the SmaI site. This intermediate vector was digested with HindIII and NdeI, generating a 0.69 Kb fragment containing a truncated segment of the tetR-tetP system plus the rest of the vector backbone. On the other hand, the pCensYBL plasmid was digested with HindIII and NdeI, generating a 2.38 Kb fragment including the promoter-less CensYBL construct. The truncated tetR-tetP fragment, the CensYBL construct and the adapted pBSK vector backbone were ligated together, generating a pBSK-tetR-CensYBL intermediate. This plasmid was digested with SmaI and SalI, and the resulting 3.14 Kb fragment containing the tetR-CensYBL construct was cloned into pWH1266-Apr digested with EcoRV and SalI, resulting in pCensYBL-Ab. An inactive version of this biosensor (pCensYBL*-Ab) was generated as a control, following the same cloning strategy and using pCensYBL* (inactive version of the biosensor) [35] as initial material.

All constructs were verified by Sanger sequencing.

Strain construction

A list of all strains used in this work is given in S1 Table. AB5075 derivatives with in-frame gene deletion were created using our previously described A. baumannii genome-editing toolkit [26]. Briefly, triparental mating was used to transfer pEMGT derivatives (pEMGT-cavA, pEMGT-cavB and pEMGT-abaI) from DH5α λpir into the parental AB5075 strain using pRK2013 as helper plasmid. After selecting the first recombination event, the second event leading to the gene deletion was triggered by conjugating pSW-Apr into the cointegrate strains [26,63]. This created ΔcavA, ΔcavB, ΔcavAΔcavB, ΔabaI, ΔcavAΔabaI, ΔcavA/csuC::Tn and ΔcavA/pgaA::Tn strains respectively. Deletions were verified by PCR.

For complementation, previously described four-parental mating method [64] with pRK2013 and pTNS2 helper plasmids [63,65] was used. Derivatives of miniTn7T-Tc [52] and miniTn7T-Tel bearing the gene of interest or empty were inserted in the chromosome of intended strains. Clones were picked on selective medium depending on the resistance marker in the miniTn7T and insertions were verified using AB5075-glmS fw/Tn7R primers [52,66].

To generate AB5075 derivatives carrying gfp transcriptional fusions to the pilA and pga promoters, pUC18T-miniTn7T-zeo-PpilA::gfpmut3 and pUC18T-miniTn7T-zeo-Ppga::gfpmut3 were introduced in AB5075 by three-parental mating and their integration in the chromosome by single recombination was selected. Selection was performed on LB agar supplemented with gentamycin (20 mg/L) and zeocin (500 mg/L). Insertions were validated by PCR using PpilA fw EcoRI/miniTXC gfp rv Ppga fw EcoRI/miniTXC gfp rv primer pairs.

Protein sequence analysis

Amino acid sequences for CavA (AKA31936.1) and CavB (AKA30837.1) were obtained from the A. baumannii AB5075-UW genome (CP008706.1). Proteins domains and cellular localisation were predicted with InterProScan and TMHMM (v2.0) softwares [67]. IBS 2.0 [68] and BioRender.com were used to create figures. These proteins were searched in a pangenome of 9,696 A. baumannii genomes [69], from which their isolation site was taken and classified into 4 groups: Human, Hospital environment, Environmental, and Other. Protein structures were predicted using AlphaFold [70,71] and visualised with PyMOL (v3.0) software, where “cealign” command was used for protein structures alignment.

Biofilm formation assay

To examine regulators of biofilm formation of AB5075, all biofilm formation assays including the initial screen were performed following a previously described protocol [72] with slight changes. Overnight cultures were diluted to OD₆₀₀ 0.1 and 150 μl were used to inoculate the corresponding wells of a microtiter dish. Following 24 h incubation at 37°C shaking (180 rpm) and growth assessment at OD₆₀₀, planktonic cells were removed by washing the wells three times with distilled water. Biofilms were then stained with 200 μl 0.1% Crystal violet for 15min and the excess stain was removed by five consecutive washings with distilled water. Plates were left to air-dry and biofilm-bound stain was resolubilised in 200 μl 99% Ethanol. Absorbance was measured at 570 nm (OD₅₇₀) using SPECTROstar plate reader (BMG Labtech). The results shown are the average of three biological replicates, with three technical replicates each.

Intracellular cAMP quantification

Intracellular cAMP levels were measured as described before [18] with slight modifications. Bacterial cultures grown to OD₆₀₀ 0.7 were used to harvest cells via centrifugation at 5,000 rpm and 4°C for 20 min. Cells were resuspended in 3ml PBS and lysed by sonication for 5 min (5 sec on, 15 sec off) at 50% amplitude. Samples were centrifuged again as already described to remove cell debris and collect the supernatant. Coomassie (Bradford) Protein Assay kit (23200) was used to quantify the protein concentration of each sample. Briefly, 250 μl Coomassie reagent were added to 5 μl sample or BSA standard, mixed briefly on a plate shaker and incubated at room temperature for 10min. Absorbance was measured at 595 nm and protein concentrations were calculated from a standard curve. Then Cyclic Nucleotide XP Enzymatic Immunoassay (Cell Signalling Technology, MA, USA) was used following manufacturer’s instructions to quantify cAMP concentrations. Fifty microlitres of each sample were transferred in duplicates in the microwell strips provided and 50 μl horseradish peroxidase-linked cAMP were added. Strips were left for 3 h at room temperature on a plate shaker. Liquid was removed from the strips and each well was washed four times with 200 μl wash buffer. After that 100 μl of stop solution (3,3’,5,5’-tetramethylbenzidinesubstrate) were added and absorbance was measured at 490 nm. Cyclic AMP concentrations were calculated from the inverse for the standards from a standard curve and adjusted for the protein concentration of each sample. The data presents three biological repeats with two technical in each.

Differential RNA sequencing and transcriptomic analysis

Cultures of ΔcavA EV and ΔcavA+cavA were grown in LB broth with IPTG (1 mM) to mid-log phase (OD₆₀₀ 0.7) at 37°C, 180 rpm. Cells were pelleted by centrifugation and resuspended in RNAlater. RNAeasy kit with on-column DNase digestion (Quiagen) was used for total RNA isolation. RNA quantity and quality were assessed using Agilent RNA 6000 Nano Kit and Agilent 2100 Bioanalyzer according to the total RNA concentration and the rRNA peak profile and sharpness obtained for each sample. Samples were sequenced using Illumina NovaSeq X Plus with 12 million reads per sample at Microbial Genome Sequencing Centre (Pittsburgh, Pennsylvania, U.S.A). Demultiplexing, quality control, and adapter trimming was done with bcl-convert (v4.1.5). BioJupies [73] was used to create a volcano plot. Genes were defined as differentially expressed if Log2(fold change) ≥ 1 or ≤ -1 and p value < 0.05. FUNAGE-Pro [27] with preset settings was used for Gene Set Enrichment Analysis (GSEA).

The RNA sequencing datasets are available at the Gene Expression Omnibus repository (NCBI) under the accession number GSE250425.

Twitching motility assay

For assessing CavA and Vfr roles in the motility regulation, previously described twitching assay was used [52]. AB5075 and its derivative strains were grown overnight on LB agar. Following autoclaving, twitching medium (10 g Tryptone, 5 g Yeast extract, 10 g Agar per 1 L) was supplemented with IPTG (1 mM) when needed, 10 ml were poured per Petri dish and plates were dried aseptically for 8 min. Each plate was inoculated at the plastic-medium interface by piercing a single colony from the plate culture through the agar. All plates were incubated at 37°C. After 48h the agar was gently removed and the diameter of the movement trace was measured. The experiment was repeated three independent times.

Natural transformation assay

Natural transformation assays were performed following a modification of our previously published protocol [52]. Overnight cultures of A. baumannii AB5075 and the ΔcavA mutant were diluted 1:100 (v/v) in 5 ml of LB broth supplemented with CaCl₂ 2 mM. Cultures were incubated at 37°C, 180 rpm until reaching OD₆₀₀ 0.8. 20 μl of each bacterial culture were mixed with 1 μg of purified genomic DNA from an AB5075 derivative carrying a chromosomal miniTn7T-Tc transposon insertion with a tetracycline resistance cassette (AB5075/miniTn7T-Tc-lacI^q-Ptac). The mixtures were spotted on twitching agar (see twitching motility assay procedure) supplemented with CaCl₂ 2 mM. The mixtures were left to air-dry in a laminar flow hood and were incubated for 4 h at 37°C. After this, the biomass was resuspended in 1 ml of PBS buffer and serial dilutions were plated either on selective LB agar supplemented with tetracycline 5 mg/L or on plain LB agar to assess viability. The transformation frequency was calculated as the number of transformant cells per millilitre divided by the number of viable cells per millilitre. Four biological replicates were performed for each strain.

GFP-based transcriptional fusions assay

The expression from the PpilA and Ppga promoters was measured following a previously described protocol [53] with slight changes. AB5075 WT, ΔcavA and vfr::Tn strains bearing mini-Tn7T-based insertions with the PpilA::gfpmut3 and Ppga::gfpmut3 transcriptional fusions were used. Cultures, grown overnight, were diluted 1:100 in LB broth and grown for 4.5 h at 37°C shaking (180rpm). Cells were harvested via centrifugation, resuspended in PBS and 100 μl were transferred in triplicates in a 96-well plate. Samples growth (OD₆₀₀) and GFP fluorescence (excitation: 485 nm; emission: 535 nm) were measured. The fluorescence was then normalised to the respective growth measurement. The experiment was repeated three independent times.

Congo red exopolysaccharide assay

CavA and Vfr effect on EPS production was tested by adapting a protocol from [74]. Adjusted overnight cultures to OD₆₀₀ 1 were used to inoculate 1% Tryptone with 1% Agar containing Congo red (40 μg/ml) and Coomassie brilliant blue (20 μg/ml) dyes. Plates were incubated at 37°C for 5 days. Example images of the colonies from three independent repeats are presented.

Detection of exogenous acyl-homoserine lactones (AHL)

To assess the effect of CavA on QS, a previously described method for detecting AHLs in A. baumannii [75] was used with some modifications. Briefly, A. tumefaciens traG::lacZ biosensor strain grown to OD₆₀₀ 0.3 was mixed with X-Gal (75 μg/ml) and soft agar (LB 0.7% agar). Plates were dried at 37°C for 35 min. Cultures of avirulent translucent (AV-T) AB5075 and its derivatives were adjusted to OD₆₀₀ 0.3 and 1 μl was spot plated in triplicates on the soft agar. Plates were incubated at 28°C for approximately 36 h. The diameter of the blue halos indicating production of AHL was measured. The experiment was done in three biological repeats.

Quantification of c-di-GMP levels

Cyclic di-GMP synthesis was measured as already described [35] with some changes. Overnight cultures of strains with active (CensYBL-Ab) and inactive (CensYBL*-Ab) c-di-GMP biosensor were diluted 1:100 in fresh LB broth and incubated at 37°C shaking for 1 h. Following biosensor expression induction with anhydrotetracycline (50 ng/ml), strains were incubated for further 2 h. Cells harvested via centrifugation from 300 μl of each culture were resuspended in 1 ml sterile PBS. 100 μl/well in triplicates were loaded in a dark 96-well plate. Fluorescence was measured using CLARIOstar plate reader (BMG Labtech). mCherry and YFP were detected using excitation/emission wavelengths of 570-15/620-20 nm and 497-15/540-20 nm respectively. Cyclic-di-GMP levels were calculated by normalising the YFP to the mCherry fluorescence data. The averages of three biological repeats with three technical replicates are presented.

Virulence in G. mellonella in vivo model

Virulence assays were performed using healthy G. mellonella (UK Waxworms Ltd.) larvae of similar size. A. baumannii AB5075 or mutant derivative cells were diluted to OD₆₀₀ 1.0 and serially diluted in PBS. The number of viable cells (enumerated by plating the serial dilutions on LB agar) of each injected strain per replicate are listed in S2 Table. All larvae were left at 37°C for 48 h. G. mellonella survival was assessed every 2 h between 20–30 h interval post-injection and at 48 h post-infection. Ten larvae were injected in each biological replicate and the experiment was repeated three independent times. Death was assessed by complete lack of response to mechanical stimuli. Probability of survival was assessed using the Long-rank (Mantel-Cox) method on GraphPad Prism.

Antibiotic disc diffusion assay

Cation-adjusted Mueller-Hinton (CAMH) agar (pH 7.4, CaCl₂ 2 mM, MgSO₄ 1mM) (Sigma-Aldrich) was used to assess antibiotic susceptibility. Overnight cultures of A. baumannii AB5075 and its derivatives were adjusted to OD₆₀₀ 0.1 in LB broth. Sterile cotton swabs were used to spread the adjusted bacterial cultures and discs with 50 μg fosfomycin were placed in the centre of each plate. All plates were left at 37°C for 24 h. The experiment was repeated three independent times.

Statistics

Graphs represent average ± standard deviation (SD) of the data. All statistical tests were performed on GraphPad Prism 10.1.0 (316) software (GraphPad Software, San Diego, California USA, www.graphpad.com). One-way ANOVA with Dunnett and Tukey’s post-hoc tests or Student t-test were used when indicated to analyse the data. G. mellonella survival probability was analysed using Long-rank (Mantel-Cox) test.