Cyclic-di-GMP signalling mutants drive ecological succession and self-generated diversity in experimentally evolved biofilms of Pseudomonas aeruginosa

Abstract

Biofilms represent a discrete form of microbial life which are physiologically distinct from free-living planktonic cells. The altered phenotypic manifestations of the biofilm may also elicit lifestyle-dependent adaptive responses to selective pressures. In this work, an experimental evolution model was used to study the adaptation to a biofilm lifestyle in Pseudomonas aeruginosa PA14. The serial passage of biofilms selected for biofilm hyperproduction in a stepwise fashion characterized by increased biomass production and phenotypic diversification was not associated with reduced susceptibility to antibiotics. Adaptation to a biofilm lifestyle selected for mutations causes constitutive increases of intracellular c-di-GMP concentrations via mutations in the phosphodiesterase dipA, the yfiBNR signalling complex and the bifunctional diguanylate cyclase/phosphodiesterase morA. Furthermore, selection for biofilm hyperproduction also gave rise to self-generated diversity by eliciting morphotypic diversification into complex community structures. Individual morphotypes were not associated with specific mutations and lineages dynamically switched between morphotypes despite possessing conserved mechanisms of biofilm hyperproduction. This work provides insights into the evolutionary importance of self-generated diversity to the biofilm and reveals the genetic control and phenotypic dynamics which contribute to the characteristically rugged fitness landscape associated with a sessile lifestyle.

Keywords: diguanylate cyclase; emergent property; experimental evolution; hyperbiofilm; niche adaptation; phosphodiesterase.

Fig. 1.. (a) Biofilm formation as measured via crystal violet staining of experimentally evolved lineages of P. aeruginosa adapted to growth planktonically or as a biofilm on glass, PVC or stainless steel substrates in LB for 30 transfers. Statistical differences between groups were determined via a two-way ANOVA at a 0.05 significance level. Significant main effects of substrate [F(4, 379)=45.89, P<0.0001] and number of transfers [F(2, 379)=25.32, P<0.0001] on biofilm formation were identified. Post hoc testing with Šidák correction identified significant increases in biofilm formation relative to the ancestor by transfer 20 in all biofilm-adapted lineages (glass: P=0.0005, PVC: P<0.0001, stainless steel: P<0.0001). Further increases were observed from transfer 20 to transfer 30 in all selective substrates except glass (glass: P=0.0578, PVC: P<0.0001, stainless steel: P<0.0001). No significant change was observed at any timepoint in the planktonically adapted lineages (transfer 10: P=0.9973, transfer 20: P>0.9999, transfer 30: P=0.9978). (b) Cellular productivity of experimentally evolved lineages grown on their selective substrate. A one-way ANOVA found no significant effect of selective substrate on productivity [F(2, 591)=1.821, P=0.1628]. Data are shown as (a) mean OD₅₉₅ and (b) as mean c.f.u. per mm², box limits show first and third quartiles, whiskers show ±1.5× interquartile range, large points show lineage mean and small points show outliers, n=16.

Fig. 2.. Growth kinetics of experimentally evolved lineages during planktonic growth. Statistical differences between groups were determined via a two-way ANOVA. A significant main effect of selective substrate on AUC [F(3, 180)=5.508, P=0.0010] and a significant interaction effect associated with substrate and number of transfers [F(6, 180)=5.895, P<0.001] were identified; however, no main effect of transfers [F(2, 180)=0.162, P=0.8510] was detected. Post hoc testing with Šidák correction identified that all biofilm-selective conditions were significantly less fit than the planktonically adapted lineages by transfer 30 (P<0.0001) driven by an increase in fitness in the planktonically adapted lineages between transfers 20 and 30 (P=0.0040). Adaptation to a biofilm lifestyle was not associated with a reduction in fitness from baseline (glass: P>0.9999, PVC: P>0.9999, stainless steel: P=0.0590). Curves show mean OD₆₀₀±sd. Boxplot shows AUC, box limits show first and third quartiles, whiskers show ±1.5× interquartile range, large points show lineage mean and small points show outliers, n=16.

Fig. 3.. Colony morphotypes identified from experimentally evolved lineages after a 10-day incubation at 20 °C on bacteriological agar supplemented with 1% w/v tryptone, 20 µg ml⁻¹ Congo red and 40 µg ml⁻¹ Coomassie brilliant blue. Data are shown as representative colony dimensions in pixels, n=4.

Fig. 4.. (a) Succession of colony morphotypes in experimentally evolved lineages. The morphotypic trajectory of each lineage was plotted using parallel sets v2.1. (b) Distribution of morphotypes across selective substrates and timepoints. Associations between colony morphotypes and selective substrate and timepoint were identified with a chi-square test for independence from 500,000 Monte Carlo simulations. Significant associations between both selective substrate and number of transfers on morphotype presentation were detected (substrate: χ²=241.56, P<0.0001, transfer: χ²=51.574, P<0.0001). Specific combinations driving these effects were determined using a Z-test for proportions on standardized residuals from the chi-square analysis controlling the false discovery rate using the Benjamini–Hochberg procedure. Hyperrugose colonies were significantly more likely in steel-adapted (residual=5.63, P<0.0001) and glass-adapted lineages (residual=2.63, P=0.0189) but were significantly underrepresented in PVC-adapted lineages (residual=−3.38, P=0.0020). Filiform and radial morphotypes were significantly overrepresented in PVC-adapted lineages (filiform: residual=5.75, P<0.0001, radial: residual=2.41, P=0.0293), and the diffuse morphotype was significantly associated with glass-adapted lineages (residual=2.3611, P=0.0312). The circumscribed morphotype, however, was not associated with any individual substrate (residual=1.6681, P=0.1271). At transfer 10, hyperrugose morphotypes were significantly overrepresented (residual=3.67, P=0.0011), while filiform morphotypes were significantly underrepresented (residual=−3.52, P=0.0016). At transfer 20, hyperrugose colonies were significantly underrepresented (residual=−4.59, P<0.0001) while both circumscribed and filiform morphotypes were strongly overrepresented (circumscribed: residual=4.09, P<0.0001, filiform: residual=4.02, P<0.0001). Data shown are as a proportion of observed morphotypes, n=4.

Fig. 5.. Colony rugosity morphometry of experimentally evolved lineages. The colony rugosity of each lineage was determined by measuring the coverage of rugose folds using a linear threshold model in Fiji and plotting the threshold area relative to the total colony area. Statistical differences in rugosity between selective substrates were determined by a two-way ANOVA. A significant main effect of selective substrate on rugosity was detected [F(4, 183)=47.622, P<0.0001], and a significant interaction effect was identified between substrate and timepoint [F(6, 183)=8.761, P<0.0001]; however, a main effect of timepoint was not detected [F(2, 183)=1.330, P=0.267]. Post hoc testing with Dunnett’s test identified that all biofilm substrates gave rise to higher rugosity than the ancestor (glass: P=0.0003, PVC: P=0.0196, stainless steel: P<0.0001), but not the planktonically adapted lineages (P=0.9925). According to Šidák’s post hoc test, the glass- and PVC-adapted lineages demonstrated a significant increase in rugosity between transfers 10 and 20 (PVC: P=0.0018, stainless steel: P=0.0016), with no further significant changes thereafter (glass: P=0.1357, PVC: P>0.9999). There was no increase in rugosity over time in the stainless steel-adapted lineages (P>0.9999). Colony rugosity also varied significantly by morphotype according to a one-way ANOVA [F(5, 179)=79.701, P<0.0001]. Post hoc testing with Tukey’s HSD test identified that all adapted morphotypes possessed significantly greater rugosity than the ancestor (circumscribed: P<0.0001, diffuse: P=0.0040, filiform: P<0.0001, hyperrugose: P<0.0001, radial: P<0.0001). The hyperrugose lineages had significantly greater rugosity than all other morphotypes (circumscribed: P=0.0266, filiform: P<0.0001, radial: P<0.0001, diffuse: P<0.0001), and the diffuse lineages were the least rugose among all morphotypes (circumscribed: P=0.0059, filiform: P<0.0001, hyperrugose: P<0.0001 radial: P=0.0420). Data are shown as mean coverage of rugosity relative to colony area, box limits show first and third quartiles, whiskers show ±1.5× interquartile range, large points show lineage mean and small points show outliers, n=16.

Fig. 6.. Gene targets under parallel selection in experimentally evolved lineages identified from Illumina short-read whole-genome sequencing data using snippy v4.6.0. Data are shown as the number of times a mutation co-occurred in the four parallel lineages sequenced, n=4.

Fig. 7.. Pathways with mutations involved in c-di-GMP signalling identified in biofilm-adapted lineages. The YfiBNR system is a tripartite regulatory complex that modulates intracellular c-di-GMP levels in response to envelope stress. During surface adhesion, YfiB sequesters the periplasmic repressor YfiR, releasing the diguanylate cyclase YfiN to produce c-di-GMP. In this study, mutations in yfiR likely disrupted its ability to bind YfiN, while a Gly24–Leu27 in-frame deletion in the first transmembrane domain of yfiN may enhance its activity or stability, promoting biofilm hyperproduction. Mutations across multiple domains were also selected in MorA, a bifunctional diguanylate cyclase–phosphodiesterase, likely shifting its activity towards net c-di-GMP synthesis, and in DipA, a phosphodiesterase critical for biofilm dispersal, which likely results in the loss of function, preventing a return to a planktonic lifestyle. The Gac/Rsm pathway regulates the motile–sessile switch via the sensor kinase GacS and response regulator GacA, which activates transcription of the small noncoding RNAs (sRNAs) rsmX, rsmY and rsmZ. These sRNAs sequester the post-transcriptional repressor RsmA, which in turn lifts translational repression of the diguanylate cyclase SadC, leading to increased c-di-GMP production and transition to a sessile lifestyle. Multiple independent mutations were detected in gacA and gacS, which likely act to constitutively activate GacA thereby de-repressing SadC. Finally, the Wsp pathway detects mechanical stimuli at the inner membrane with the chemoreceptor WspA which, in turn, activates the sensor kinase WspE. WspE phosphorylates the GGDEF domain-containing response regulator WspR, promoting production of c-di-GMP. Normally, this is controlled through a feedback system with the methylesterase WspF, which is phosphorylated to an active state to negatively regulate the activity of WspA. In this study, a frameshift mutation was detected in wspA, which likely disrupts signal detection or triggers constitutive activation; a 14-residue in-frame deletion in wspE which may alter kinase function or specificity; and an early stop codon in wspF which may abolish its negative regulatory function, resulting in constitutive Wsp pathway activation.