|Home | About | Journals | Submit | Contact Us | Français|
Here, we investigated novel interactions of three global regulators of the network that controls biofilm formation in the model bacterium Escherichia coli using computational network analysis, an in vivo reporter assay and physiological validation experiments. We were able to map critical nodes that govern planktonic to biofilm transition and identify 8 new regulatory interactions for CRP, IHF or Fis responsible for the control of the promoters of rpoS, rpoE, flhD, fliA, csgD and yeaJ. Additionally, an in vivo promoter reporter assay and motility analysis revealed a key role for IHF as a repressor of cell motility through the control of FliA sigma factor expression. This investigation of first stage and mature biofilm formation indicates that biofilm structure is strongly affected by IHF and Fis, while CRP seems to provide a fine-tuning mechanism. Taken together, the analysis presented here shows the utility of combining computational and experimental approaches to generate a deeper understanding of the biofilm formation process in bacteria.
Bacteria shift from their free-swimming lifestyle to adopt a community structure to benefit from the micro-environment created in a biofilm1, gaining protection against hazardous substances, and leading in some cases, to antibiotic resistance1–3. It is well-understood that when differentiating from the planktonic to a biofilm structure bacteria transit through specific stages, with each functional stage accompanied by changes to expression of specific genes of the flagella-biofilm regulatory network1–3. In Escherichia coli, the complex transcriptional regulatory network of flagella function and curli fimbriae production, the principal biofilm structure indicator, has been investigated in various reports4–7. In this organism, the process is controlled by the RpoS sigma factor and FlhDC regulator. These master regulators receive major regulatory inputs from c-di-GMP, cAMP and ppGpp, which are modulated by a series of environmental and physiological signals8–10. For example, the flhDC genes are expressed at post-exponential phase and their products control more than 60 genes involved in flagella synthesis and related functions, such as chemotaxis11. Also at stationary phase, the general stress response master regulator RpoS is produced and controls over 500 genes in a highly complex regulatory network7. Therefore, the interplay between those master regulators and downstream-activated genes modulates the complex transition between planktonic and biofilm stages.
In the case of planktonic bacteria, mainly found in the exponential phase, flagella maintenance is due to the activation of effector molecules such as FliA sigma factor. Whilst, at stationary phase, biofilm formation is initiated by increasing levels in the secondary messenger c-di-GMP, as well as expression and activation of rpoS, which in turn leads to the activation of csgD 5,12. Additionally, other mechanisms also leading to csgD activation and scale-up of curli fiber production; such as CpxR and ClpX which play a complex dual role during bacterium development to inhibit or activate both programs8,10,13. Undoubtedly, the interplay between these molecules acts to modulate the proper execution of the flagella-biofilm program14,15. Experimentally, the transition between flagella and biofilm formation is evaluated using a “batch cells” assay, in which cells grow and attach to the chemically-inert surface of a microliter dish under static conditions, generating a biofilm in an “aquatic environment”. Additionally, the “macro-colonies” assay allows a bacterial population to grow over extended periods of time on agar plates, leading to striking morphological patterns, resembling biofilms growing in biological materials.
While hundreds of reports have dissected the molecular mechanisms leading to biofilm formation in several bacterial groups, a full understanding the dynamics of the complex regulatory network controlling planktonic to biofilm transition still lies out of reach. This is particularly true since most studies analyzed the effect of particular regulators on the expression of specific sets of genes, without addressing the complexity of the network using a more systematic approach. In particular, since a few global regulators (GRs) are able to control most of the genes in genome of E. coli 16, it is anticipated that the role of these transcriptional factors in the biofilm formation network is rather underestimated. GRs are expressed or activated differentially during growth or under specific input signals17. Specifically, CRP is related to the control of metabolic processes in bacteria and it is differentially activated by substrates18, while IHF (growth rate dependent) and Fis (expressed at early exponential phase) are dual nucleoid associated proteins (NAPs) that can work both as activators and repressors19,20. During the transition from motile cells to biofilm structure, a mixture of different developmental stages occur. In this sense, small molecules and GRs direct the activation of signaling proteins that are sensed by neighboring cells, coordinating the construction of this complex community structure14,15. Therefore, mapping hidden interactions in the flagella-biofilm regulatory network would provide a better understanding of this developmental-like program and provide new targets for intervention and/or engineering21,22.
In this work, we used computational tools and molecular biology, together with microbiology approaches to investigate the effect of the GRs CRP, IHF and Fis on the E. coli flagella-biofilm network. First, to analyze regulatory interactions already reported, we reconstructed the flagella-biofilm transcriptional regulatory network in silico using previous data and we analyzed its structure. Next, by using promoter expression assays, we mapped novel regulatory interactions between the three GRs and key genes controlling biofilm formation. Finally, novel regulatory interactions mapped using this approach were added to the original network to generate an enhanced appreciation of biofilm formation. The general strategy used here is depicted in Fig. S1. These results suggest that IHF and Fis are important modulators of the process while CRP seems to exert a fine-tuning effect. Additionally, using motility, adherence and biofilm formation assays, we validated key regulatory effects observed using transcriptional fusions. All together, the systematic investigation presented here adds new understanding of the complex regulatory program that controls biofilm formation in E. coli, revealing a novel player in this network.
To reconstruct the flagella-biofilm regulatory network, information regarding the transcriptional regulation of E. coli was gathered from scientific articles and compiled with that from RegulonDB23 and KEGG DB24, and from the extensive analysis of the literature. We used only experimentally validated interaction and not large-scale data or computational predictions. The resulting network consisted of 37 nodes representing the connections between different transcription factors (TFs), GRs and small molecules (Table S1). The degree analysis (which shows the most connected nodes) of the flagella-biofilm transcriptional regulatory network identified three major hubs, rpoS, flhD and csgD (Fig. 1A). The betweenness and edge betweenness analysis which provides information about the most important nodes and paths in the network, shows that from 37 nodes, 8 seem to be the principal effectors that modulate the proper gene expression of the network (rpoS, csgD, c-di-GMP, rpoE, cpxR, clpX, flhD, fliA). Interestingly, rpoS is the main effector in the network and it is modulated by different inputs. Among the inputs of rpoS, c-di-GMP presents high betweenness and edge connection values (Fig. 2). From the 37 nodes, rpoD and CRP were the main nodes, which exert more out-regulatory signals to other nodes, while all others possessed fewer out-connections (Fig. 1B). Additionally, this analysis shows three paths that seem to be important to maintain balance in the network. The first critical connection that maintains flagella-biofilm program equilibrium lies between rpoS and rpoE, activating both fliA for motility and its repressor matA, to eventually develop adherence. Secondly, rpoS with the TF cpxR, which modulates rpoE, but also has an important connection with the TF nsrR, which represses fliA expression. Thirdly, the activation of clpX ATPase by rpoS that exerts flhD inhibition at post-transcriptional and/or post-translational levels25. From the analysis presented here, we selected ten genes in order to perform a systematic investigation of the effect of CRP, IHF and Fis, which are among the main GRs of the regulatory network of E. coli 26. We expected that this approach would reveal some hidden interactions controlling the critical nodes of the flagella-biofilm network. The selected genes of interest were rpoS, rpoE, csgD, cpxR, flhD, fliA, matA, ompR, adrA and yeaJ, which were found to have critical roles in the network studied or as they could serve as indicators of the output of the network. In order to analyze the promoter activity of the selected genes, the promoter regions of the genes were cloned upstream of a GFPlva reporter system27 and used to transform wild-type and Δihf and Δfis mutant strains of E. coli. As control, an empty reporter plasmid and a strong synthetic promoter named BBa_J23100 (http://parts.igem.org/Part:BBa_J23100, here referred as Pj100) were also used to transform the same strains (Fig. S2). The 36 reporter strains were then analyzed as presented in the next section.
We started our investigation by considering the three putative master regulators of the transition from planktonic to biofilm phases. The sigma factor RpoS has been proposed as a main regulator of the flagella-biofilm network7,8, in agreement with our network analysis. Thus, we were interested to understand the effect of these three GRs on the promoter region of the rpoS gene. For this, overnight cultures of E. coli wild-type strain carrying the transcriptional PrpoS::GFPlva fusion were diluted 1:100 in fresh M9 minimal medium supplemented with glycerol as sole carbon source, and GFP fluorescence was measured at 20minute time intervals over 8hours at 37°C. As can be seen in Fig. 3A, the analyzed promoter presented a growth-phase-dependent activity with increased activity toward the end of the growth curve. When the same reporter system was analyzed in E. coli Δihf and Δfis mutant strains, we observed a significant increase in promoter activity in the absence of IHF and a strong activity in the absence of Fis (Fig. 3A left panel, red and green lines, respectively). These data strongly suggest that IHF and Fis GRs are acting to repress the rpoS promoter. To understand the role of CRP over the rpoS promoter region, we first employed the glucose inducible CCR system to down-modulate the activity of CRP by adding glucose in the same experimental conditions as described above28. When the three reporter strains were analyzed in the presence of 0.4% glucose, we observed the same general expression profile as in Fig. 3A left panel but a remarkably reduced promoter activity (Fig. 3A right panel), in agreement to the previously reported positive role of CRP on the rpoS promoter29. However, when this promoter was analyzed in a mutant strain of crp, we observed an increase in rpoS promoter, indicating that this regulator could in fact have a negative role in the expression of this target (Fig. S3). It is worth to notice that previous reports have also reported contradictory effect of carbon source and crp on rpoS expression, and two CRP-binding sites with potential antagonistic roles are reported for the promoter of this gene (REF). Therefore, the mutant assayed reported here indicated that the final role of CRP on rpoS promoter is repressive.
When we analyzed the activity of the rpoE promoter, which controls the expression of a sigma factor linking RpoS activity with flagella genes (Fig. 2), we observed a strong increase in promoter activity in the Δihf and Δfis strains of E. coli when compared to the wild-type strain (Fig. 3B left panel), indicating that IHF and Fis have a negative effect on the activity of this promoter. However, contrary to the observation for rpoS, the addition of glucose (Fig. 3b right panel) to the media (which inhibits CRP activity) did not result in any significant change in rpoE promoter activity. Additionally, when the same promoter was assayed in E. coli Δcrp mutant strain, only a minor increment in promoter activity was noticed, in agreement with the result found in glucose (Fig. S3).
After investigating the activity of rpoS and rpoE promoters, we analyzed the promoter of flhD gene, which encodes a master regulator of flagellar genes and of the flagella-related fliA sigma factor30. Using the same experimental conditions as described above, we first observed that flhD promoter activity has a strong growth-dependence, with activity increasing during the growth curve when assayed in the wild-type strain of E. coli (Fig. 3C left panel). When, we assayed the promoter activity in Δihf and Δfis mutant strains, we observed a similar expression profile as in the wild-type, with the exception that in both strains the initial promoter activity was significantly higher than in the wild-type. When we assayed promoter response in the three strains in the presence of glucose (to trigger CRP inactivation), activity was strongly impaired in all strains (Fig. 3C right panel), confirming the reported role of CRP as activator of flhD expression31. The same profile was observed in a mutant strain for crp, where flhD promoter activity was strongly reduced (Fig. S3).
Once we had investigated the effect of the GRs of flagella to biofilm transition, we proceeded to analyze the promoters of genes related to biofilm formation and an additional lower flagellum regulator (fliA). CsgD is a central protein related to, and an indicator of, biofilm formation12,32. When we analyzed the activity of csgD promoter in the wild-type strain, we observed a peak of expression at about 200min of growth with a reduced activity later in the growth phase (Fig. 4A left panel). Investigating this promoter in Δihf and Δfis mutants, we observed a strong decay in promoter activity, indicating that both proteins positively modulate the activity of the promoter. Moreover, when we added glucose to the growth media (Fig. 4A right panel), we observed a generalized decay in promoter activity in all three strains, indicating a positive role of CRP for this promoter. This result was also observed in the Δcrp mutant strain of E. coli (Fig. S3). Reflecting on these results, it is interesting to note that, while CRP and IHF have been reported as positive regulators of the csgD promoter33,34, the positive effect of Fis has never been demonstrated before.
In the case of the fliA gene that codes a sigma factor specifically related to flagella genes, analysis of promoter activity in both wild-type and Δfis mutant strains revealed a very low level of activity throughout the growth curve, both in the presence and absence of glucose (Fig. 4B). However, in the Δihf mutant strain, this promoter displays a strong increase (30-fold) in activity in the absence of glucose (left panel), while this level was lower when glucose was added (right panel). In agreement to this later observation, expression of fliA promoter was strongly reduced in the Δcrp mutant strain of E. coli (Fig. S3). Since fliA expression is dependent on FlhD30, the observed decay of fliA promoter activity during growth in the presence of glucose could be the result of a cascade process, with the apparent IHF repression of this promoter suggesting a previously unreported regulatory interaction. To test the direct effect of IHF on FliA, we identified two putative binding sites for this global regulator at fliA promoter Fig. S4A. We then constructed a mutant version of this promoter with 11 point mutations that fully abolish both sites and tested this new variant as before. As shown in Fig. S4B, expression of the mutated version of fliA promoter was low in the wild-type and Δfis mutant strains at the same level as the original promoter, and displayed a higher activity in the Δihf mutant, both in the absence and presence of glucose. These data suggest that the effect of IHF on fliA promoter is indirect.
We next analyzed the effect of GRs on the expression of yeaJ, a diguanylate cyclase coding gene, which modulates the levels of c-di-GMP in the cell. In E. coli, deletion of yeaJ results in strains with reduced motility at 37°C8, however no regulatory proteins have been demonstrated to play a role in the expression of this gene. The expression profile of the yeaJ promoter in wild-type and Δihf mutant strains showed a strong peak of activity at 200min of experiment (Fig. 4C). However, when this promoter was assayed in the Δfis mutant, we observed a significant decrease in promoter activity, indicating that the Fis regulator could play a positive role in its expression. Furthermore, addition of glucose to the growth media (Fig. 4C right panel) resulted in an increase in the promoter activity in the first minutes of growth. However, analysis of this promoter in the Δcrp mutant strain of E. coli demonstrated a completely different result, were yeaJ promoter expression was drastically reduced in the mutant (Fig. S3). Finally, the additional three promoters selected for investigation (from adrA, cpxR and ompR genes, Fig. S5) displayed maximal activity very similar to that of the negative control (the empty reporter vector, pMR1 in Fig. S2), which precludes unequivocal conclusions concerning their regulation.
In order to generate fresh insight into the regulatory network controlling biofilm formation in E. coli, we added the new regulatory interactions identified in the previous sections to the network from Fig. 2. For this, the effect of CRP, IHF and Fis over the promoter regions obtained by our promoter activity assay was transformed into activation or inhibition links. From the 27 interactions tested, we suggested 8 new interactions, and were able to confirm 5 described interactions (Table 1). We integrated these new interactions into the network and re-performed the centrality measurements. Using this approach, we observed that the integration of the experimental data changed the topology of the network, while maintaining the three major hubs identified (Fig. S6A). Yet, the out-degree analysis showed that rpoD, CRP, IHF, and Fis in minor degree are the main nodes exerting the most out-connections to all other nodes (Fig. S6B), while the edge betweeness analysis indicates that, upon the addition of the new interactions, the critical nodes of the network remain the same (Fig. 5). Taken together, these data suggest the existence of previously unidentified regulatory interactions in the biofilm regulatory network that play a critical role in the planktonic to biofilm transition.
Flagella play an important function during the different developmental phases of biofilm formation such as attachment at a surface, as well as bacterial cell cohesion within the biofilm2,30. Therefore, we were interested in understanding the effect of the GRs CRP, IHF and Fis on the flagella function. As it can be seen in Fig. 7S, after 18h of incubation in a low agar plate, E. coli BW25113 wild-type strain showed reduced motility in the presence of glucose. However, analysis of the ihf strain in similar conditions revealed a strong motile phenotype, with this effect more apparent after 24h of incubation with full spreading over the plate observed (Fig. 6). Finally, under similar conditions, the fis strain presented reduced motility, which is more evident at 24h. It is important to notice that addition of glucose to the media generated a general decrease in motility of all strains (Figs 6 and S7, right panels). Interestingly, the observed strong phenotypic effect of ihf on the motility of E. coli can be traced directly to the strong increase of fliA promoter activity in this mutant strain, since FliA sigma factor controls several genes related to flagella formation in this organism35.
We next analyzed the capability of the different strains to attach to a solid surface, which is indicative of early biofilm formation, using the protocol described by O’Toole et al., 2011. For this, we performed the experiments at 30°C and 37°C, since different reports have used different temperatures36,37. In general, E. coli cells adhered to the wall of the 96-well plates in all conditions. At 37°C, two- or three-fold increases in biofilm yields were observed when compared with E. coli cells cultured at 30°C, as can be seen by comparing the biofilm indexes in Fig. 7A. In general terms, the Δihf mutant strain of E. coli displayed decreased adhesion at both temperatures, which is in agreement with the increased motility observed for this strain in Fig. 6. By the same token, adhesion of the Δfis mutant strain at 37°C was significantly increased when compared to the wild-type, which could be due to the reduced motility observed before. In most of the cases, the general adhesion observed for the strains was decreased in the presence of glucose, with the exception of the adhesion level of Δihf mutant at 30°C. These data suggest that, at 37°C, CRP and IHF are important effectors to activate the adherence program. They also suggest that Fis is acting as a repressor of the early biofilm program, since in its absence, motility is higher at 37°C. Altogether, adherence yields are better when induced at 37°C than at 30°C, and the three GRs CRP, IHF and Fis participate differentially to regulate the adherence capabilities of E. coli.
As depicted in Fig. S8, all the strains were capable of developing biofilm structures at 30°C and 37°C in the presence or absence of glucose, as previously reported. In general, E. coli cells grown at 30°C produced a biofilm structure resembling that reported for 28°C, while those grown at 37°C conditions showed a more visually complex biofilm structure with three morphologically distinct zones, as reported by Serra et al.7 (Fig. 7B). More explicitly, in wild-type cells a concentric ring delineates zone 3, which presents an intense red color suggesting curli formation by the different layers of stationary phase cells. Zone 2 represents the intrinsic capability of wild-type community to produce wrinkles and curli. Finally, zone 1 presents weak red color, an indicative of bacteria at exponential phase related with colony expansion but not with curli production7. In the presence of glucose, the biofilm structure formed by the wild-type strain presented slight differences. When we analyzed the Δihf mutant strain, we observed a significant reduction in the colony size and degree of pigmentation (that could be related to lower curli production) when compared to the two other strains (Fig. 7B). This size reduction can be evidenced by the systematic decrease in the three zones of the colony in this strain. Yet, analysis of biofilm formation in this strain in the presence of glucose showed a recovery in colony size generated by a high expansion of zone 1, while pigmentation was apparently not affected. Finally, the analysis of Δfis mutant strain revealed a colony where the zone 1 (expansion zone) could not be detected. Additionally, this strain presented a darker zone 3, which could be indicative of higher production of curli fimbriae than wild-type strain, with addition of glucose to the growth media only generated small changes in the colony such as an apparent reduction in size.
Altogether, morphology assays have shown that IHF and Fis GRs are important contributors to proper biofilm development, as could be indicated by drastic changes in the three well-characterized zones in these mutants. In contrast, the role of CRP protein (which was indirectly assayed by the addition of glucose to the media) during mature biofilm formation could not be dependably inferred, since the change in the nutritional state of the colonies could have a strong impact in the process. Taken together, these data indicate that these GRs could have an important role in the final biofilm structure in E. coli that could be the result of the detected change in gene expression of key regulatory elements in the planktonic to biofilm regulatory network.
The work presented here describes the systematic investigation of the regulatory interaction mediated by GRs during the transition from planktonic to biofilm in the bacterium E. coli. While several works have addressed this issue before, data generated has been fragmented and few targets have been investigated at each time. The systematic investigation used here allowed the identification of novel interactions mediated by CRP, Fis and IHF, as presented in Table 1. More importantly, the modulation of some critical nodes, such as fliA by IHF, could explain the strong phenotypic effects observed for motility and attachment assays, as this regulator has also been found to be a modulator of genes involved in the biofilm program as rpoS, matA and csgD 33,38–40. The molecular mechanism involved in the different phases of biofilm formation by IHF remains to be determined. However, in this work, we present evidence that IHF is a key element to biofilm formation by modulating gene expression level of the flagella-biofilm program, suggesting that IHF could be a good candidate to disrupt/engineer bacterial biofilm structure. Here, we also demonstrated that Fis plays a more major role in the biofilm formation network than anticipated, both using promoter assays as well as physiological tests. The evidence in this study indicates that, CRP, IHF and Fis are important effectors that modulate the flagella-biofilm network. Interestingly, no condition in this study shows absence of motility or biofilm formation, suggesting that, while these GRs are important modulators of these processes, they are not essential genes to completely suppressed the flagella-biofilm program. Additionally, it has been demonstrated that bacteria without the main effectors rpoS or csgD (also rpoE, csgB, flhDC or fliA) are still capable of producing a biofilm7,41. Altogether, the high plasticity observed in the flagella-biofilm program clearly shows that the network has evolved the robustness to compensate for the loss of critical nodes, indicating the importance of this program for bacterial survival. Finally, the systematic identification of novel connections for a specific network will allow prediction of the behavior of the biological system in the absence of different TFs. It opens the possibility of using genetic engineering to achieve a balanced connectivity between synthetic circuits and the entire bacterial network when performing a specific task. In other words, that rational interactive synthetic circuits can operate in accordance with the natural network of a specific organism, thereby allowing enhanced performance of a desired biological task.
The list of bacterial strains, plasmids and primers is presented in Table 2. Escherichia coli wild-type (BW25113) or strains genetically depleted of IHF or Fis GRs42 were used as the host of all plasmids. E. coli DH5α or DH10b were used as cloning strains. DNA of E. coli BW25113 and MG1655 were used as templates. E. coli cells were grown in Luria Broth (LB) or M9 minimal media (6.4g/L Na2HPO4•7H2O, 1.5g/L KH2PO4, 0.25g/L NaCl, 0.5g/L NH4Cl) supplemented with 2mM MgSO4, 0.1mM CaCl2, 0.1mM casamino acids, and 1% glycerol or 0.4% glucose as the carbon source. Chloramphenicol was added (34μg/mL) to ensure plasmid maintenance. Cells were grown at 37°C with constant shaking at 220rpm overnight.
Plasmids and primers used in this work are listed on Table 2. As a control of all experiments we transformed E. coli BW25113 wild-type and mutant strains with an empty pMR1 vector27 and pMR1-Pj100. To analyze the signal integration of GRs at the studied genes, we amplified the promoter regions by PCR and cloned into pMR1 vector using standard protocols43. DNA from E. coli BW25113 wild-type was used as a template for all constructions except for flhD promoter, for which E. coli MG1655 was used. PCR was performed using 50 ng of DNA template with 50pmol of each primer (purchased from Sigma and extended) and 1U Phusion® High-Fidelity DNA Polymerase (New England Biolabs M0530L) in a 50μl reaction volume. Initial denaturation was at 98°C for five minutes; subsequently, amplification was done over 30 cycles at 98°C, 60°C and 72°C, each with duration of 30seconds. The final extension step was performed at 72°C for ten minute. All PCR products were gel purified and cloned as EcoRI/BamHI fragments in pMR1, previously digested with the same enzymes. The resulting plasmids were DNA sequence verified and correct plasmids were used to transform E. coli BW25113 (wild-type), E. coli JW1702 (Δihf) and E. coli JW3229 (Δfis).
For promoter activity measurements, single colonies of wild-type and mutant E. coli strains harboring the different plasmids were picked from fresh plates. Each strain was inoculated and grown overnight at 37°C in 2mL of M9 medium with glycerol or glucose containing chloramphenicol with shaking at 225rpm. Stationary phase cultures were diluted to a final optical density (OD) of 0.05 in 200μL of M9 medium containing either glycerol or glucose as required, and supplemented with chloramphenicol. Cell growth and GFP fluorescence was quantified every 20minutes over an 8 hr incubation at 37°C using Victor×3 plate reader (PerkinElmer, Waltham, Massachusetts, USA). Under these conditions, all strains presented similar growth rates (Fig. S9). Data were analyzed from at least three biological samples, arbitrary units were calculated by dividing GFP-corrected by OD600-corrected, subsequently, the mean and standard error for each promoter was calculated and graphed using Microsoft Excel and R software.
The effect of GRs on motility was evaluated as following. E. coli wild-type or mutant strains harboring pMR1 plasmid were grown overnight at 37°C on LB plates. Inoculation of a single colony onto motility plates (tryptone 1%, NaCl 0.25%, agar 0.3% and when indicated glucose was added) was done by using a toothpick44. The motility halos were measured at 18h and 24h. Each strain was evaluated in triplicate from independent plates.
Biofilm formation was assayed using the Microtiter Dish Biofilm Formation Assay45. Single colonies of E. coli wild-type or mutant strains harboring all the constructions were grown overnight at 37°C in M9 supplemented with glycerol or glucose. The cultures were diluted 1:100 in 200μl of fresh M9 media with either glycerol or glucose. A 96-well plate was incubated at 37°C for 72h, after which cells were gently washed four times with distilled water and left at room temperature for 15minutes with 200μL of crystal violet 0.1% w/v solution. Subsequently, vigorous washing was performed with distilled water (four times) and the plates left to dry at room temperature for 24h. Finally, 200μL acetic acid 30% w/v was added and after 15minutes, the solution was transferred to a new plate and Optical Density at 550nm was measured in FLUOstar Omega (BMG Labtech) plate reader. All the culture dilutions, staining and quantification steps were performed by using procedures developed for a robotic platform called Edwin46. Data from biological triplicates were analyzed by Microsoft Excel software. Average values and standard deviation was determined.
Biofilm morphology was evaluated by Congo red assay47. Chloramphenicol was added during all steps to ensure plasmid maintenance. Wild-type and mutant E. coli strains with pMR1 plasmid were grown overnight at 37°C on LB agar plates. Single colonies were picked and grown for 14h at 37°C on 1ml LB at 220rpm. The cultures were washed twice with MgSO4 and resuspended in M9 media with either glycerol or glucose as required, to an OD at 600nm (~0.5). 5μL drop of each culture was added into YESCA-CR plates48. YESCA media (1g/L yeast extract, 20g/L agar) was complemented with Congo red 50μg/ml diluted in KPi buffer (50mM potassium phosphate buffer, pH 7.2, 28.9mM KH2PO4, 21.1mM K2HPO4 in water), where indicated glucose was added at 0.4% final concentration. YESCA-CR seeded plates were allowed to dry in a sterile environment and, left to grow at 37°C. Biofilm development was followed (documented) for 6 days using a fluorescence stereoscope (Leica MZ16). Biological triplicate images were analyzed by ImageJ and prisma software was used to determine the statistical significance of the experiments by using one-way ANOVA.
To generate the transcriptional regulatory network that will define our genes of interest, we gathered information from research articles that contained information on genes that modulate the transition between planktonic to biofilm formation. Eventually, the data were compared with information available at Regulon DB, EcoCyc DB and Kegg pathway DB. Cytoscape 3.4.1 was used to generate the transcriptional network with slight modifications49. Briefly, we transformed our mined data into activation or inhibition statements for each of the partner molecules. Secondly, to construct the network graph, we used Cytoscape 3.4.1 in which different connections were added to each node on the collected data. Next, networks were analyzed using different algorithms, such as degree, out-degree, betweenness centrality by using Cytoscape 3.4.1 software. Finally, we transformed our GFP dynamics gene expression experimental results into functional interactions (inhibits or activates). Results with same function as reported were maintained without changes in the rewired network, while those with different activity were changed in the network. During the data collection from the literature, some TFs were reported with dual effect (activation and repression) for the same target gene. For instance, rpoS expression was reported to be activated but also inactivated by the CRP global regulator50–52. Additionally, most of the genes that were reported as being regulated by different TFs, but in the web-databases the same gene did not present a cis-element to any TFs. To overcome those constrains we considered for the network that any TF, master regulator, global regulator, small molecule or RNAs could be a node. This last, despite being gathered, were not included because RNA-node analysis of regulatory gene expression significantly increases the complexity of the network12,53 and because of the complex requirements of molecular tools to evaluate this topic. For edges direction analysis, we considered the activation or inhibition most reported, or both when divergent reports were found. The data were plotted in a power-law graph using an organic algorithm to give cluster visualization generating the transcriptional regulatory network of both planktonic and biofilm structure. Furthermore, we performed degree analysis in order to present a general view of the most connected nodes, followed by edge and node betweenness analysis54 in order to disclose the main effectors and the logic pathway that could describe the network. Finally, we also performed the out degree analysis at the integrated network to identify the potential talkative nodes.
The authors express their thanks to members of the Silva-Rocha & Elfick labs for insightful discussion and suggestions in developing this work. This work was supported by a FAPESP Young Research Award (grant number 2012/22921-8) and by a Royal Society Newton International Exchange award (NI140137). G.R.A. was supported by a CAPES PhD fellowship while A.S.M. was supported by FAPESP Scientific Initiation Award (2015/22386-3). A.d.H. was supported by an EPSRC Programme Grant (EP/J02175X) to the Flowers Consortium.
G.R.A. and R.S.R. conceived the work. G.R.A., A.d.H. and A.S.M. performed the experiments. G.R.A., A.d.H. and A.S.M analyzed the data. G.R.A. and R.S.R. drafted the manuscript. A. d.H. and A.E. corrected the manuscript. A.E. and R.S.R. provided funding for the execution of the work. All authors read and approved the manuscript.
The authors declare that they have no competing interests.
Electronic supplementary material
Supplementary information accompanies this paper at 10.1038/s41598-017-17114-6.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.