Biofilm formation by Staphylococcus aureus is triggered by a drop in the levels of a cyclic dinucleotide

Abstract

The bacterial pathogen Staphylococcus aureus forms multicellular communities known as biofilms in which cells are held together by an extracellular matrix principally composed of repurposed cytoplasmic proteins and extracellular DNA. These biofilms assemble during infections or under laboratory conditions by growth on medium containing glucose, but the intracellular signal for biofilm formation and its downstream targets were unknown. Here, we present evidence that biofilm formation is triggered by a drop in the levels of the second messenger cyclic-di-AMP. Previous work identified genes needed for the release of extracellular DNA, including genes for the cyclic-di-AMP phosphodiesterase GdpP, the transcriptional regulator XdrA, and the purine salvage enzyme Apt. Using a cyclic-di-AMP riboswitch biosensor and mass spectrometry, we show that the second messenger drops in abundance during biofilm formation in a glucose-dependent manner. Mutation of these three genes elevates cyclic-di-AMP and prevents biofilm formation in a murine catheter model. Supporting the generality of this mechanism, we found that gdpP was required for biofilm formation by diverse strains of S. aureus. We additionally show that the downstream consequence of the drop in cyclic-di-AMP is inhibition of the "accessory gene regulator" operon agr, which is known to suppress biofilm formation through phosphorylation of the transcriptional regulator AgrA by the histidine kinase AgrC. Consistent with this, an agr mutation bypasses the block in biofilm formation and eDNA release caused by a gdpP mutation. Finally, we report the unexpected observation that GdpP inhibits phosphotransfer from AgrC to AgrA, revealing a direct connection between the phosphodiesterase and agr.

Keywords: Staphylococcus aureus; biofilm formation; cyclic-di-AMP.

Fig. 1.

eDNA release and dependence on gdpP is a shared feature of biofilm formation by multiple strains of S. aureus. (A) The indicated strains were tested for their ability to form glucose- and DNA-dependent biofilms (Top) and to release eDNA (Bottom). gdpP::TnΩ1 (red) was transduced into each strain and the mutant strains were tested for their ability to form glucose-dependent (Top) biofilms and release eDNA (Bottom). DNase I was added to established biofilms for 1 h prior to harvesting to test the dependence on DNA (green). (B) Cells containing a cyclic-di-AMP sensitive riboswitch biosensor were grown under conditions that do not permit biofilm formation (no added glucose, black) or biofilm promoting conditions (plus glucose, gray). Cyclic-di-AMP levels dropped between 6 and 8 h in a glucose-dependent manner.

Fig. 2.

Gene deletions that block reduction of cyclic-di-AMP levels prevent biofilm formation. (A) Cells containing a cyclic-di-AMP sensitive riboswitch biosensor were grown under biofilm promoting conditions (plus glucose). gdpP::TnΩ1 (red), ΔxdrA (yellow), Δapt (blue), ΔxdrA/Δapt (green) mutations caused elevated levels of cyclic-di-AMP compared to that of the wild-type HG003 (gray) under biofilm-forming conditions. (B) eDNA release and biofilm formation of the parental strains containing an empty vector plasmid (pASD132, black) or a plasmid overexpressing gdpP (pASD133, purple) during glucose-dependent biofilm formation. (C) Established biofilms were tested for sensitivity to DNase I treatment (green and orange).

Fig. 3.

c-di-AMP drop is required for biofilm formation in a murine model of infection: (A) Subcutaneously implanted catheters were placed in immunocompetent CFW/Swiss-webster male mice and then strains of S. aureus [HG003 (gray), gdpP::TnΩ1 (red), ΔxdrA (yellow), and Δapt (blue)] were injected into the catheter lumen. After 3 d, the catheter was removed and the colony-forming units (CFU) were measured from the catheter and surrounding tissue. Datapoints represent CFU measurements from individual animals and black lines are the mean of all measurements. Differences in bacterial load in the catheters and surrounding tissue were determined using a two-way ANOVA. (B) To assess the stability of catheter biofilms to DNase treatment, an initial DNase treatment phase was added to the protocol used in A. Subcutaneously implanted catheters in CFW/Swiss-webster male mice that were injected with the HG003 were collected 3 d after infection. The removed catheters were then placed in buffer with DNase (+ DNase, green) or without DNase (– DNase, gray) for 1 h (“DNase phase”) and released CFU were measured. The catheters were then sonicated to enumerate remaining bacteria (“post sonication”). The tissue surrounding each catheter was collected and homogenized and CFU were enumerated (“surrounding tissue”). Datapoints represent CFU measurements from individual animals and black lines are the mean of all measurements. Differences were determined using a two-way ANOVA. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns = not significant.

Fig. 4.

An agr deletion mutation is epistatic to gdpP::TnΩ1. (A) Relative levels of RNA from agrA, RNAIII, and psmα are reported for S. aureus strains ΔgdpP, ΔxdrA, HG003 with pASD133 (gdpP on a plasmid), ΔagrA, and ΔagrA/ΔgdpP, grown in glucose containing medium relative to an HG003 control. Cyclic-di-AMP levels are listed as a reference based on data from Fig. 2A. n.d. indicates that agrA transcripts were not detected in ΔagrA cells. Unpaired two-tailed Student’s t tests were performed, and the levels of significance are indicated. (B) eDNA release and biofilm formation during glucose-dependent biofilm formation. (C) Cells containing a cyclic-di-AMP sensitive riboswitch biosensor were grown under biofilm promoting conditions (plus glucose gray) gdpP::TnΩ1 (red). Mutations in ΔagrA (green) or ΔagrA/gdpP::TnΩ1 (blue) do not affect the level of cyclic-di-AMP compared to the parent strain (HG003 and gdpP::TnΩ1, respectively). The baseline has been normalized to HG003 no glucose.

Fig. 5.

GdpP inhibits AgrC phosphotransfer to AgrA. (A and B) Phosphotransfer from ³²P-AgrC^cyt (lower band, 1.25 µM) to AgrA (upper band, 10 µM) is shown in the absence (Top) and presence (Bottom) of cyclic-di-AMP (1 mM) (A) or GdpP^cyt (20 µM) as indicated by the reaction scheme. Radioactive gels were imaged by phosphorimaging and band intensities were quantified below for ³²P-AgrA signal (Left, which accumulates and then disappears due to spontaneous hydrolysis) and ³²P-AgrC^cyt signal (Right). Data for AgrC^cyt signal were fit to an exponential decay function with rates indicated on the plots. (C) Phosphotransfer from ³²P-KinA from B. subtilis (upper images, 0.1 µM) to Spo0F (lower images, 0.2 µM) is shown in the absence (Top) and presence (Bottom) of GdpP^cyt (20 µM). KinA (69 kDa) and Spo0F (14 kDa) bands are shown cropped from the same gels. Plots show quantification as in (A) and (B). (D) MST dose–response curve (change in absorbance plotted against ligand concentration) for labeled AgrC incubated with serially diluted GdpP^C (Left) in the presence (green) and absence (blue) of 100 µM cyclic-di-AMP. Data were fit to a binding model and dissociation constants are labeled above graphs. (E) MST dose–response curve for labeled AgrC incubated with serially diluted AgrA (Right) in the presence (red) and absence (gray) of 100 µM cyclic-di-AMP, presence (blue) of 200 nM GdpP^C, and presence (green) of 100 µM cyclic-di-AMP and 200 nM GdpP^C. The reaction scheme is diagrammed above the plots. Data are plotted with a logarithmic X-axis. Data were fit to a binding model and dissociation constants are labeled above graphs. Reported errors are the errors of the fits and all plots have a logarithmic X-axis.

Fig. 6.

Model for the role of GdpP in biofilm formation. A schematic is shown summarizing how GdpP controls biofilm formation. gdpP, xdrA, and apt together are required for a biofilm-associated drop in the levels of c-di-AMP. This drop in c-di-AMP is in turn required for biofilm formation through inhibition of Agr. In media lacking glucose or when gdpP, xdrA, or apt are deleted, c-di-AMP levels remain high, preventing inhibition of Agr and biofilm formation.