Abstract
The second messenger signaling molecule cyclic diguanylate monophosphate (c-di-GMP) drives the transition from planktonic to biofilm growth in many bacterial species. Pseudomonas aeruginosa has two surface sensing systems that produce c-di-GMP in response to surface adherence. The current thinking in the field is that once cells attach to a surface, they uniformly respond with elevated c-di-GMP. Here, we describe how the Wsp system generates heterogeneity in surface sensing, resulting in two physiologically distinct subpopulations of cells. One subpopulation has elevated c-di-GMP and produces biofilm matrix, serving as the founders of initial microcolonies. The other subpopulation has low c-di-GMP and engages in surface motility, allowing for exploration of the surface. We also show that this heterogeneity strongly correlates to surface behavior for descendent cells. Together, our results suggest that after surface attachment, P. aeruginosa engages in a division of labor that persists across generations, accelerating early biofilm formation and surface exploration.
Introduction
Pseudomonas aeruginosa is an opportunistic pathogen that engages in a range of surface-associated behaviors and is a model bacterium for studies of surface-associated communities called biofilms. Biofilms are dense aggregates of cells producing extracellular matrix components that hold the community together. The biofilm mode of growth confers cells protection from a variety of environmental stresses including nutrient limitation, desiccation, and shear forces, as well as engulfment by protozoa in the environment or phagocytes in a host (1).
The secondary messenger signaling molecule cylic-di-GMP (c-di-GMP) drives the transition from the planktonic to the biofilm mode of growth. In many bacterial species, including P. aeruginosa, elevated c-di-GMP results in repression of flagellar motility genes, while promoting expression of genes involved in producing a biofilm matrix(2). The P. aeruginosa biofilm matrix is composed of a combination of polysaccharides (including Pel and Psl), proteins (including the adhesin CdrA), and extracellular DNA (3–8). Biofilm matrix production is an energetically costly process that is regulated at multiple levels (9). The cdrA, pel and psl genes are all transcriptionally induced under conditions of high c-di-GMP(10).
For many species, the initial step in biofilm formation involves adherence of free swimming planktonic cells to a surface and the initiation of surface sensing. P. aeruginosa has at least two distinct surface sensing systems, the Wsp and the Pil-Chp systems, that when activated, lead to biofilm formation. The Wsp system senses an unknown surface-related signal (recently proposed to be membrane perturbation (11)) through WspA, a membrane-bound protein homologous to methyl-accepting chemotaxis proteins (MCPs). Activation of this system stimulates phosphorylation of the diguanylate cyclase WspR, which leads to the formation of aggregates of phosphorylated WspR (WspR-P) in the form of visible subcellular clusters. This aggregation of WspR-P potentiates its activity, increasing c-di-GMP synthesis (12). In comparison, the Pil-Chp chemosensory-like system initiates a hierarchical cascade of second messenger signaling in response to a surface (13). First, an increase in cellular cAMP levels occurs through activation of the adenylate cyclase CyaB by the chemotaxis-like Pil-Chp complex. This increases expression of genes involved in type IV pilus biogenesis, including PilY1. PilY1 is associated with the type IV pilus and harbors a Von Willebrand motif, which is involved in mechanosensing in eukaryotic systems(14). Thus, it has been proposed that this protein may be involved in the mechanosensing of surfaces (15). The output of this second signal is through the diguanylate cyclase, SadC, resulting in an increase in cellular c-di-GMP levels. Unlike the Wsp system, which localizes laterally along the cell (16), PilY1 is required to be associated with polarly-localized type IV pili in order to stimulate c-di-GMP production (13, 14), suggesting that P. aeruginosa deploys both polar and laterally localized systems to promote c-di-GMP synthesis in response to a surface.
Here, we examined the dynamics of c-di-GMP production and bacterial surface motility at the single-cell level during early stages of biofilm formation. We used a plasmid-based, transcriptional reporter of intracellular c-di-GMP to follow the downstream fate of cells producing varying levels of c-di-GMP in response to surface attachment. Within a clonal population of P. aeruginosa, we found that levels of c-di-GMP vary among individual cells as they sense a surface, leading to a division of labor between two energetically costly behaviors associated with early biofilm formation: surface exploration and polysaccharide production.
Results
Cellular c-di-GMP levels rapidly increase upon surface attachment
We initially compared levels of c-di-GMP between P. aeruginosa PAO1 cells growing attached to a silicone surface and subjected to constant flow for 4 hours to those grown planktonically for 4 h. As expected, we observed that PAO1 cellular c-di-GMP levels are 4.4-fold higher (± 0.78 SD, N = 3, p ≤ 0.05) after 4h of growth attached to a surface compared to planktonic growth (Figure 1A). Because direct measurement of c-di-GMP by LC-MS/MS is limited by our ability to generate enough biomass at earlier time points, we used qRT-PCR to monitor pel transcript levels as a readout of c-di-GMP. We found that after just 30 min of surface attachment, pelA transcript levels had increased almost 10-fold compared to planktonically grown cells (Figure 1 – Supplement 1). This is consistent with previously published literature showing that transcription of the pel operon is directly and positively controlled by high cellular levels of c-di-GMP (17, 18).
The PcdrA::gfp reporter detects heterogeneity in c-di-GMP during surface sensing
Next, we sought to visualize early c-di-GMP signaling events at the single cell level. To this end we used a plasmid-based, c-di-GMP responsive transcriptional reporter, pPcdrA::gfpASV (19) in two commonly-studied P. aeruginosa strains, PAO1 and PA14. Planktonic cells (a condition where the reporter is inactive due to low c-d-GMP levels) were used to inoculate flow cell chambers. We imaged individual cells of each reporter strain hourly for up to 6 hours after surface attachment (Figure 1B and Figure 1 – Supplement 2). As expected, we saw minimal GFP fluorescence at the 0 h time point (right after surface attachment). However, by 1 h, the reporter was activated in a subset of surface attached cells, as defined by GFP fluorescence greater than twice that of background levels (referred to as reporter “on” subpopulations). Interestingly, between 4 and 6 h post inoculation, we consistently observed that the c-di-GMP reporter was only active in a subset of cells in both strains (Figure 1C). In PA14, the reporter was activated in 10% of the population over 6 h, whereas PAO1 displayed greater reporter activity, with 40-60% of the cells displaying reporter activity through 12 h (Figure 1 – Supplement 2). We confirmed these results using flow cytometry to assess the proportion of attached cells that were fluorescent (Figure 1 – Supplement 3D,E). To be sure that the promoter of cdrA is representative of c-di-GMP-regulated gene expression, we replaced PcdrA with the promoter of siaA, a gene that is also highly expressed under conditions of elevated c-di-GMP (10, 20). We found that pPsiaA::gfpASV reporter activity resembled that of pPcdrA::gfpASV in response to a surface (Figure 1 – Supplement 4). Thus, reporter activity is indeed linked to cellular levels of c-di-GMP.
Cyclic di-GMP heterogeneity leads to phenotypic diversification at early stages of biofilm formation
We then wanted to confirm that subpopulations of surface-attached P. aeruginosa cells with high and low c-di-GMP reporter activity are truly physiologically distinct from one another. We used TRITC-labeled lectins to stain for two c-di-GMP-induced exopolysaccharides, Psl and Pel (7, 21), the presence of which is indicative of biofilm formation by PAO1 and PA14, respectively. After 4h of attachment to glass, we observed an enrichment of TRITC-conjugated lectin staining in the population of cells with high c-di-GMP reporter activity (Figure 1D and Figure 1 – Supplement 5), demonstrating that the subpopulation of cells with high c-di-GMP is producing more exopolysaccharide than their low c-di-GMP counterparts. As a complementary approach, we separated 4 h surface-grown cells of the reporter strain into reporter “on” and “off” subpopulations using flow-assisted cell sorting (FACS; Figure 1 – Supplement 6). We then applied qRT-PCR to compare Pel and Psl transcript levels in these two populations. Both the pel and psl operon transcripts were elevated in the reporter “on” subpopulation, relative to the reporter “off” subpopulation (Figure 1E). These data support that, with respect to c-di-GMP signaling, there are at least two distinct subpopulations that arise shortly after surface attachment.
The Wsp system is required for surface sensing
We next evaluated the relative contributions of the Wsp and Pil-Chp surface sensing systems to surface-induced c-di-GMP production. Strains with mutations in the Pil-Chp chemosensory system were not significantly defective in surface sensing activity. Deletion of the diguanylate cyclase activated through the Pil-Chp system (PAO1 ΔsadC) and the gene encoding the putative sensor PilY1 (PAO1 ΔpilY1) did not significantly influence reporter activity in response to a surface (Figure 2 – Supplement 1A,B). Whereas both the SadC and PilY1 mutants displayed wild type levels of reporter activity, a mutant lacking the main Type IV pilus filament protein (PAO1 ΔpilA) did show a statistically significant defect in reporter activity by 6 h (Figure 2 – Supplement 1B; p <0.05 by T-test). We then mutated the c-di-GMP cyclase gene, wspR, to inactivate the Wsp system. In addition, we deleted the gene encoding the methylesterase wspF, which locks the system into the active state, regardless of whether cells are surface-associated. We found that PAO1 ΔwspR strain exhibited extremely low levels of reporter activity during the first 6 h after surface attachment (Figure 2A and Figure 2 – Supplement 2). Complementation of PAO1 ΔwspR restored wild type levels of activity at all time points (Figure 2 – Supplement 3). As expected, PAO1 ΔwspF had a high proportion of reporter active cells (Figure 2A). We repeated these experiments in the lab strain PA14 and saw a similar trend for Wsp and Pil-Chp mutants (Figure 2 – Supplement 4).
Since the Pil-Chp surface sensing apparatus is polarly localized and the Wsp system is localized laterally along the length of the cell body, we examined whether reporter activity correlated with polar versus lateral attachment to the surface. We found that reporter activity was very low in polarly attached cells, while cells attached along the entire length of the cell body displayed a higher proportion of activated cells (Figure 2B). This finding is also consistent with the localization of the Wsp system and its role for early c-di-GMP signaling during surface sensing.
Heterogeneity in c-di-GMP levels among cells correlates with Wsp system activity
The specific activity of purified WspR increases as a function of WspR concentration when the protein is treated with beryllium fluoride to mimic phosphorylation, supporting the idea that formation of subcellular clusters of WspR-P potentiates its diguanylate cyclase activity and leads to elevated c-di-GMP(12). Fewer than 1% of wild-type cells grown in broth have a visible WspR-YFP cluster. However, after a short period of growth on an agar surface, WspR-YFP clusters were visible in 30-40% of wild type PAO1 cells, and this is dependent on sensing by the membrane-bound protein WspA, which is laterally distributed in cells (9). To directly link WspR cluster formation with diguanylate cyclase activity at the cellular level and with surface sensing, we constructed a version of the c-di-GMP reporter that expresses mTFP1 instead of GFP (pPcdrA::mTFP1) to avoid the issue of spectral overlap with WspR-YFP. We monitored reporter activity in two point mutants of WspR (L170D and E253A) that are driven by an inducible promoter, translationally fused to eYFP and have been previously shown to form large subcellular WspR clusters in a higher percentage of cells than wild-type WspR. The WspR[L170D] protein is highly active for c-di-GMP production, and it forms subcellular clusters in about 75% of agar surface-grown cells. A WspR[E253A] point mutation abolishes diguanylate cyclase activity, but this protein still forms clusters in about 70% of surface-grown cells (12). As expected, in the presence of inducer, we observed a large increase in c-di-GMP reporter activity in WspR[L170D], but not WspR[E253A] (Figure 3A, B). We then asked whether the heterogeneity in reporter activity in response to surface attachment correlates with WspR clustering in the WspR[L170D] strain. We found that pPcdrA::mTFP1 activity was significantly higher in cells with at least one subcellular WspR-eYFP focus in the WspR[L170D] strain compared to cells without a WspR-eYFP focus (Figure 3C; median mTFP1 fluorescence of 345 vs. 320 RFU respectively, Mann-Whitney test, p < 0.001). These data indicate that the heterogeneity observed in c-di-GMP signaling after surface attachment is due to the heterogeneity in the activity of the Wsp system, as reflected by subcellular clustering of active WspR-P.
We next asked whether the observed heterogeneity in c-di-GMP signaling in response to a surface has a meaningful influence on biofilm formation. This was particularly important since previous published results indicated that a wspR mutation had only a small impact on biofilm production (22). However, these studies assessed biofilm formation at later stages of biofilm growth that were well beyond initial surface attachment. Therefore, we chose to compare a wspR mutant to wild type at earlier biofilm stages. We performed in vitro biofilm assays and observed that a PAO1 ΔwspR mutant was defective for biofilm formation relative to wild type PAO1 at 2, 4, and 6 hours post-attachment (Figure 4A). However, at later stages of development (∼24 h), the wspR mutant caught up and produced similar amounts of biofilm biomass relative to wild type levels. Complementation of the ΔwspR strain in trans restored wild type levels of biofilm formation at all time points. These data suggest that the Wsp system rapidly responds to surface contact to generate elevated levels of c-di-GMP, which accelerates biofilm production. Given the importance of c-di-GMP signaling in biofilm production, the fact that the ΔwspR strain can ultimately attain wild-stype levels of biofilm biomass suggests that one of the many other known c-di-GMP cyclases present in P. aeruginosa may ultimately compensate for c-di-GMP production in the absence of WspR.
Cyclic di-GMP heterogeneity leads to diversification in surface exploration at the lineage level
We hypothesized that heterogeneity in c-di-GMP signaling dictated by the Wsp complex could impact the surface behavior of the two observed subpopulations. We predicted that the subpopulation of cells with high c-di-GMP after surface attachment would produce biofilm matrix exopolysaccharides and contribute to initial microcolony formation, while the cells with low c-di-GMP would exhibit increased surface motility and detachment, which is known to be inhibited by exopolysaccharide production. To test this hypothesis, we tracked both reporter activity and surface behavior for cells within a single field of view for 40 h. From our single-cell tracking data, we generated family trees across at least four generations of cells, using a previously described technique (23). We tracked the time-averaged PcdrA::gfpASV reporter activity (Ic-di-GMP), surface motility behavior (Fmotile, defined as the fraction of time that cells are motile), and detachment behavior (tree asymmetry λ where λ = 0 represents both daughter cells remaining attached to the surface and λ = 1 represents when one daughter cell detaches or travels outside the field of view).
In P. aeruginosa, surface exploration is mainly accomplished by twitching motility, mediated by type IV pili, and does not appear to be influenced by levels of intracellular c-di-GMP when analyzing single cells (24). Interestingly, we found that correlations between c-di-GMP and motility during the lifetime of individual cells are weak. However, when analyzing entire lineages in family trees rather than individual cells, we found clear inverse correlations between Ic-di-GMP and Fmotile (Figure 4B, ρ = - 0.53, p = 0.0012) and between Ic-di-GMP and λ (Figure 4C, ρ = −0.45, p = 0.0068), suggesting that c-di-GMP levels is strongly inversely correlated with surface motility behavior and detachment behavior over multiple generation of cells. To illustrate these correlations, we chose three representative families, with either high, intermediate, or low Ic-di-GMP and plotted their family trees (Figure 4D) and spatial trajectories (Figure 4E). Families with the highest Ic-di-GMP had the lowest Fmotile and λ (Family 1, Figure 4B-E). In these families, daughter cells remained attached following cell division, exhibited continuously elevated c-di-GMP, did not move appreciable distances on the surface, and ultimately produced small microcolonies. In contrast, families of cells with low Ic-di-GMP had the highest Fmotile and λ. For these families, daughter cells frequently detached or traveled outside the field of view, had lower c-di-GMP levels, traveled larger distances on the surface, and ultimately did not form microcolonies (Family 3, Figure 4B-E).
One important question is what happens to early biofilm development if we were to effectively remove heterogeneity in c-diGMP output rooted in the WspR surface sensing system. To address this question, we used a strain in which c-di-GMP production could be easily controlled using an optogenetic system. The precise control of c-di-GMP expression in individual cells was made possible by the use of a chimeric protein that fused a diguanylate cyclase domain to a bacteriophytochrome domain. Flow chambers were seeded with the optogenetic strain encoding a heme oxygenase (bphO) and light-responsive diguanylate cyclase (bphS)(25). Initially, cells attached on the glass surface were tracked and continuously stimulated with red-light over ∼8 h using adaptive tracking illumination microscopy (ATIM), which allows for precise stimulation of the initial attached cells and their offspring and ensures sustained intracellular c-di-GMP production for a fixed number of surface cell generations (Figure 5 – Supplement 1). Cellular lineages (a cell and all of its offspring) and c-di-GMP expressions were continually monitored for at least 12 h. Families that were not stimulated with light demonstrated a heterogeneous surface response (Figure 5B,D) similar to that of Families 1-3 in Figure 4B-E. Some lineages were dominated by surface explorers, whereas others were seen to commit to microcolony formation. In contrast, in families stimulated with light for more than 1 generation, the resulting c-di-GMP production artificially forced lineages to have low surface motility and commit to microcolony production (Figure 5A,C) similar to that of Family 1 in Figure 4B-E. Families stimulated with light in this manner had higher Ic-di-GMP and lower λ values than those that were not stimulated (Figure 5 – Supplement 2). We also found that optogenetic control of c-di-GMP results in phenotypes that are consistent with the wild-type behavior presented in Figure 4, with illuminated cells (high c-di-GMP) displaying the least motility and control (non-illuminated) displaying comparatively greater surface motility (Figure 5 – Supplement 2). Interestingly, families stimulated with light for 1 generation or less are not significantly different from un-illuminated controls (data not shown). Our data show that the generation of c-di-GMP can deterministically lead to the creation of an entire lineage of sessile cells with post-division surface persistence, low motility, and initiation of microcolony formation. Altogether, these results show that c-di-GMP levels, surface motility, and detachment are inversely correlated at the lineage level, and that the time scale for this occurs over multiple generations.
Discussion
Collectively, our data show that heterogeneity in cellular levels of c-di-GMP, generated by the Wsp system in response to surface sensing, leads to two distinct physiological subpopulations. Phenotypic heterogeneity of single cells is a common phenomenon in bacteria that is thought to be beneficial at the population level by allowing a single genotype to survive sudden environmental changes and by promoting a division of labor between costly behaviors that support the growth and survival of the population (26). Sources of phenotypic heterogeneity include bistability (27) and stochasticity (28) of gene expression, unequal partitioning of proteins during cell division due to low abundance (28), epigenetic modifications resulting in phase variation (29), or through asymmetrical cell division (30, 31). In this study, we show that the Wsp system generates heterogeneity in c-di-GMP signaling, and it is never fully activated in 100% of wild-type, surface-attached cells. Moreover, we show that such heterogeneity results in phenotypic changes for entire family lineages of descendent cells. It is interesting that correlations between c-di-GMP, surface motility, and surface detachment probability are strong when considered for an entire lineage in a bacterial family tree, but weak when considered at the individual cell level. This form of correlation suggests that the enforcement of surface sensing outcomes (ex: the activation of DGCs, attenuation of motility) is slow compared to the cells’ division times, and that c-di-GMP signaling is propagated across multiple generations. Additionally, proteins such as DGCs activated by surface sensing may not be passed down to daughter cells equally after division, especially if their number is not large or if they are assymetrically partitioned, which may be one mechanism that leads to the heterogeneity in c-di-GMP levels.
If we overwhelm WspR-generated c-di-GMP heterogeneity by using optogentically-induced sustained c-di-GMP production, we find that phenotypic heterogeneity is lost, and that illuminated cells deterministically become sessile and form microcolonies. Interestingly, our optogenetic experiments show that sustained c-di-GMP production for more than one generation is required before commitment to the sessile lifestyle. This observation is consistent with the fact that we see strong correlations between c-di-GMP levels and motility behavior at the lineage level and not at the individual cell level. Moreover, since the WspR surface sensing system generates heterogeneous c-di-GMP levels, this requirement of sustained c-di-GMP production for more than one generation is inherently difficult for wild-type cells to meet, and virtually guarantees the simultaneous existence of motile and sessile subpopulations. This phenotypic heterogeneity, which has been ‘hardwired’ into the structure of c-di-GMP surface sensing networks, allows for a division of the labor during early biofilm formation, with one subpopulation committing to initiating the protective biofilm lifestyle, while the other subpopulation is free to explore the surface and potentially colonize distant, perhaps more favorable, locations.
Materials and Methods
Bacterial strains and growth conditions
The strains, plasmids, and primers used in this study are listed in Table 1. Escherichia coli and P. aeruginosa strains were routinely grown in Luria–Bertani (LB) medium and on LB agar at 37°C. For the flow cell experiments, P. aeruginosa was grown in either LB or FAB minimal medium supplemented with 10mM or 0.6mM glutamate at room temperature (7). For flow cytometry experiments, P. aeruginosa was grown in either LB medium or in Jensen’s defined medium with glucose as the carbon source (21). For the tube biofilm and c-di-GMP measurements, P. aeruginosa strains were grown in Vogel-Bonner Minimal Medium (VBMM; (32)). Antibiotics were supplied where necessary at the following concentrations: for E. coli, 100 µg/mL ampicillin, 10 µg/mL gentamicin, and 10 or 60 µg/mL tetracycline; for P. aeruginosa, 300 µg/mL carbenicillin, 100 µg/mL gentamicin, and 100 µg/mL tetracycline. PcdrA::gfpASV reporter and vector control plasmids were selected with 100 µg/mL gentamicin for P. aeruginosa strains and 10 µg/mL gentamicin for E. coli.
PAO1 ΔpilY1 was constructed using two-step allelic exchange following conjugation of wild type PAO1 with E. coli S17.1 harboring pENTRPEX18Gm::ΔpilY1 (a gift from Joe Harrison) as previously described (33). PAO1 ΔpilY1 was identified by colony PCR using primers PAO1pilY1-SEQ-F and PAO1pilY1-SEQ-R. PAO1 ΔdipA was constructed similarly by conjugation of wild type PAO1 with E. coli S17.1 harboring pENTRPEX18Gm::ΔdipA (a gift from Joe Harrison). PAO1 ΔdipA was identified by colony PCR using primers PAO1dipA-SEQ-F and PAO1dipA-SEQ-R. PA14 ΔwspR and ΔwspF deletion mutants were confirmed by PCR using primers PA14wspR-SEQ-F and PA14wspR-SEQ-R or PA14wspF-SEQ-F and PA14wspF-SEQ-R, respectively.
To create MPAO1 attTn7::P(A1/04/03)::GFPmut, the miniTn7 from pBT270 was integrated into the chromosome of P. aeruginosa PAO1 with the helper plasmid pTNS2, as previously described (34). pBT270 was created by introducing the constitutive A1/04/03 promoter (35) and removing the trc promoter from pBT223 using the QuikChange Lightning Kit (Agilent Technologies) and the oligonucleotides OBT314 and OBT315. pBT223 was constructed via recombineering of pBT200, pUC18-miniTn7T2-Gm-GW, and pBT212 using Multisite Gateway technology (Invitrogen). pBT212 was constructed by cloning the gfpmut3 from AKN66 using OBT268 and OBT269, and recombining the PCR product with pDONR221 P1-P5r.
Construction of optogenetic, c-di-GMP reporter strain in P. aeruginosa
Chromosomal insertion of bphS was achieved using the mini-CTX system and these strains were marked with different fluorescent proteins by mini-Tn7 site-specific transposition essentially as previously described (34, 36). First, a bphS fragment obtained from the plasmid pIND4 was cloned into the vector mini-CTX2 with the PA1/O4/O3 promoter upstream of the MCS via a two-piece ligation. The constructed plasmid was electroporated into PAO1 and the corresponding recombinant strain was identified by screening on LB agar plates containing 1mM IPTG and 100 µg/mL tetracycline. Then, the strains were electroporated with a pFLP2 plasmid and distinguished on LB agar plates containing 5% (w/v) sucrose for the excision of the resistance marker. The c-di-GMP reporter plasmid and mCherry/EGFP marked bphS mutants were constructed as described above. The c-di-GMP reporter plasmid (PcdrA::gfpASV) was electroporated into the mCherry marked bphS mutant to monitor the intracellular c-di-GMP level.
Cyclic di-GMP measurement and qRT-PCR of tube biofilms
Measurement of c-di-GMP in tube biofilm cells was performed as previously described (4). Transcriptional analysis of PelA expression in tube biofilms was performed as described in the “FACS and qRT-PCR of c-di-GMP reporter cells” section.
Crystal violet attachment assays
Crystal violet assays were performed essentially as previously described to measure biofilm biomass, except using gentle washing after 2-6 hours of static incubation (8). To measure biofilm biomass at 24 hours, the crystal violet assay was performed as previously described without gentle washing (37).
Flow cell time course experiments and confocal microscopy
P. aeruginosa cells harboring the pPcdrA::gfpASV reporter plasmid or a promotorless vector control (pMH489) were grown to mid-log in LB with 100 µg/mL gentamicin (Gm100) from LB Gm100 plates or from FAB + 10mM glutamate overnight broth cultures in FAB + 10mM glutamate. Mid-log cells were back diluted into 1% LB or FAB + 0.6mM glutamate and flow chambers were inoculated at a final OD600 0.1 and inverted for 10 minutes to allow cells to attach before induction of flow. Clean media was used to wash non-attached cells by flow at 40mL per hour for 20 minutes. Flow was then reduced to a final constant flow rate of 3mL per hour and bacteria were imaged immediately on a Zeiss LSM 510 scanning confocal laser microscope (t=0h). Flow cells were incubated at a constant flow rate at room temperature and imaged hourly for up to 24 hours. For every strain and time point, 5 fields of view and a minimum of 300 cells were captured using identical microscope settings to image GFP fluorescence across all experiments. Images were analyzed using using Volocity software (Improvision, Coventry, UK). Cells were counted as pPcdrA::gfpASV reporter “on” if their mean GFP fluorescence intensity per pixel was greater than two-fold above the background GFP fluorescence intensity (approximately 340). Data are presented in terms of the percentage of cells with an average GFP fluorescence per pixel twofold more intense compared to the background (pPcdrA::gfpASV reporter “on”). Microscopy images were artificially colored to display GFP fluorescence as green.
Construction of pPsiaA::gfp
A region 259 bp upstream through 21 bp into the coding sequence of siaA was amplified from PAO1 genomic DNA using primers BamH1-Psia-F and SiaA-BamH1-R, then gel purified using a QIAquick gel extraction kit (Qiagen, Hilden, Germany) digested with BamH1, then column purified with a QIAquick PCR purification kit (Qiagen, Hilden, Germany) to remove BamH1. The GFP expression vector pMH487, which contains the gfpmut3 gene with an RNase III splice site and lacking a promoter (38), was digested with BamH1, treated with Antarctic phosphatase (New England Biolabs, Ipswich, MA), then column purified with a QIAquick PCR purification kit (Qiagen, Hilden, Germany) to remove BamH1. The PsiaA allele was ligated into digested pMH487, then transformed into E. coli DH5α, purified, and sequenced using primer M13F(−21) (Genewiz). The reporter pPsiaA::gfp was electroporated into P. aeruginosa as previously described and maintained under gentamycin selection at 100 µg/mL.
Multi-generation single cell tracking of type IV motility and c-di-GMP reporter activity
Wild type PAO1 harboring the pPcdrA::gfpASV reporter was grown shaking for 20 hours in FAB media with 6mM glutamate. The flow cell inoculum was prepared by diluting the culture to a final OD600 of 0.01 in FAB with 0.6mM glutamate. The flow cell inoculum was injected into the flow cell (Department of Systems Biology, Technical University of Denmark) and allowed to incubate for 10 minutes at 30°C prior to flushing with media at 30mL/h for 10 minutes. Experiments were performed under a flow rate of 3mL/hour for a total of 40 hours.
Images were acquired with an Olympus IX81 microscope equipped with a Zero Drift Correction autofocus system, a 100× oil objective with a 2× multipler lens, and an Andor iXon EMCCD camera using Andor IQ software. Bright-field images were recorded every 3 seconds and GFP fluorescence every 15 minutes. Acquisition continued for a total recording time of 40 hours, which resulted in approximately 48000 bright-field images, and 160 fluorescence images.
Images were analyzed in MATLAB to track bacterial family trees, GFP fluorescence, and surface motility essentially as previously described(23) with the following modifications. Image analysis, family tracking and manual validation, family tree plotting, and tree asymmetry λ calculations were performed as previously described(23) without modification. GFP fluorescence intensities were normalized by calculating the distribution of intensities per cell per frame (extracted by using the binary image as a mask) and then setting the minimum and maximum intensities to the 1st and 99th percentiles of this distribution for each dataset. Ic-di-GMP (relative normalized c-di-GMP reporter intensity) was calculated by averaging the normalized fluorescence intensities across all members of a family. Fmotile (fraction of time that cells in a family are motile) was calculated as follows. For each family, every cell trajectory in the family was divided into time intervals. For each time interval, presence or absence of motility was determined using a combination of metrics, including Mean Squared Displacement (MSD) slope, radius of gyration, and visit map. MSD slope quantifies the directionality of movement relative to diffusion. Radius of gyration and visit map are different metrics for quantifying the average distance traveled on the surface. Fmotile was then calculated by the fraction of these time intervals that have motility. This calculation was modified from the “TFP activity metric” previously described(23).
Setup of Adaptive Tracking Illumination Microscopy
Figure 5 – Supplement 1 shows a schematic of the Adaptive Tracking Illumination Microscopy (ATIM) setup. An inverted fluorescent microscope (Olympus, IX71) was modified to build the ATIM. The modification includes: 1) a commercial DMD-based LED projector (Gimi Z3) was used to replace the original bright-field light source, in which the original lenses in the projector were removed and three-colored (RGB) LEDs were rewired to connect to an external LED driver (ThorLabs) controlled by a single chip microcomputer (Arduino UNO r3); 2) the original bright-field condenser was replaced with an air objective (40× NA = 0.6, Leica); and 3) an additional 850 nm LED light (ThorLabs) was coupled to the illumination optical path using a dichroic mirror (Semrock) for the bright-field illumination. Note that 850 nm LED light is safe light to ensure that the bright-filed illumination does not affect optogenetic manipulation. The inverted fluorescent microscope (Olympus, IX71) equipped with a 100×oil objective and a sCMOS camera (Zyla 4.2 Andor) was used to collect bright-field images with 0.2 frame rate. The bright-field images were further analyzed to track multiple single cells in real time using a high-throughput bacterial tracking algorithm coded by Matlab. The projected contours of selected single cells were sent to the DMD (1280 × 760 pixels) that directly controlled by a commercial desktop through a VGA port. The manipulation lights were generated by the red-color LED (640 nm), and were projected on the single selected cells in real time through the DMD, a multi-band pass filter (446/532/646, Semrock) and the air objective. Our results indicated that feedback illuminations could generate projected patterns to exactly follow the cell movement (Figure 5 – Supplemental 1B) or single cells divisions (Figure 5 – Supplemental 1C) in real time.
Manipulation of c-di-GMP expression in single initial-attached cells
The bacterial strain PAO1-bphS-PcdrA-GFP-mCherry was inoculated into a flow cell (Denmark Technical University) and continuously cultured at 30.0 ± 0.1°C by flowing FAB medium (3.0 mL/h). The flow cell was modified by punching a hole with a 5 mm diameter into the channel, and the hole was sealed by a coverslip that allows the manipulation light to pass through. An inverted fluorescent microscope (Olympus, IX71) equipped with a 100× oil objective and a sCMOS camera (Zyla 4.2 Andor) was used to collect bright field or fluorescent images with 0.2 or 1/1800 frame rate respectively. The power density of the manipulation lights was determined by measuring the power at the outlet of the air objective using a power meter (Newport 842-PE). GFP or mCherry was excited using a 480 nm or 565 nm LED lights (ThorLabs) and imaged using single-band emission filters (Semrock): GFP (520/28 nm) or mCherry (631/36 nm). Initial-attached cells were selected to be manipulated using ATIM with the illumination at 0.05 mW/cm2, which allowed us to compare the results arising from illuminated or un-illuminated mobile cells in one experiment. The c-di-GMP levels in single cells were gauged using the ratio of GFP and mCherry intensities.
Lectin staining and flow cytometry
Glass culture tubes were inoculated with 1mL of P. aeruginosa in LB or Jensen’s minimal media at an OD600 0.8 and incubated statically at 37°C for 4 hours. Non-adhered cells were removed by washing three times with 2mL sterile phosphate buffered saline (PBS). Biofilm cells were harvested by vortexing in 1mL PBS with fluorescein-labeled lectins (WFL lectin (100 µg/mL; Vector Laboratories) for Pel, TRITC-labeled HHA (100 µg/mL; EY Laboratories) for Psl) and incubated on ice for 5 minutes. Cells were washed 3 times to remove non-adhered lectin, resuspended in PBS, and immediately analyzed for GFP and TRITC fluorescence on a BD LSRII flow cytometer (BD Biosciences). Events were gated based on forward and side scatter to remove particles smaller than a single P. aeruginosa cell and large aggregates.
We used PAO1 cells that did not express GFP (wild type PAO1; Figure 1 – Supplement 3A) or constitutively expressed GFP (PAO1 Tn7::P(A1/04/03)::GFPmut; Figure 1 – Supplement 3B) to define a gate for high GFP fluorescence. We validated this gate using a strain in which we expect very high levels of reporter activity (surface grown PAO1 ΔwspFΔpelAΔpslBCD harboring pPcdrA::gfpASV) and saw that 91.6% of cells had high GFP levels (Figure 1 – Supplement 3C), in agreement with our flow cell characterization of this strain (Figure 2A). We determined gating for TRITC using cells that had not been stained with TRITC-conjugated lectin (Figure 1 – Supplement 5A), as well as two strains that overproduced either Psl (Figure 1 – Supplement 5B) or Pel (Figure 1 – Supplement 5C) that were stained with the appropriate TRITC-conjugated lectin. Our flow cytometry gating procedure accurately gated 99.7% of wild type PAO1 cells (without the PcdrA reporter or lectin-staining) as low GFP and low TRITC (Figure 1 – Supplement 5D).
FACS and qRT-PCR of c-di-GMP reporter cells
Static biofilm reporter cells were grown as described above and harvested without lectin staining. Cells were fixed with 6% paraformaldehyde for 20 minutes on ice, then rinsed once with sterile PBS prior to analysis with a FACSAriaII (BD Biosciences, San Jose, CA). Events were gated first to remove debris and large cellular aggregates, and then gated into cells with low and high GFP fluorescence intensity. The low GFP gate was drawn using wild type PAO1 cells without the gfp gene (Figure 1 – Supplement 6A) and the high GFP gate was drawn using both PAO1 Tn7::P(A1/04/03)::GFPmut (Figure 1 – Supplement 6B) and PAO1 ΔwspF ΔpelA ΔpslBCD PcdrA::gfpASV reporter (Figure 1 – Supplement 6C). As expected, wild type PAO1 pPcdrA::gfpASV reporter cells that had been harvested after 4 hours of surface attachment to glass in static LB liquid culture displayed subpopulations of high GFP, reporter “on” cells (30.8% of the population) and “off” (57.2%) cells (Figure 1 – Supplement 6D), whereas this same strain grown to mid-log planktonically in LB displayed mostly reporter “off” cells (Figure 1 – Supplement 6E). Cells were sorted at 4°C by flow assisted cell sorting (FACS) to collect 100,000 events into TRIzol LS (Thermo Fisher Scientific, Waltham, MA). RNA was extracted from sorted cells by boiling immediately for 10 minutes and following the manufacturer’s instructions for RNA isolation. DNA was digested by treating with RQ1 Dnase I (Promega, Madison, WI) and samples were checked for genomic DNA contamination by PCR to detect rplU. Expression of pelA, pslA, and ampR was measured by quantitative Reverse Transcriptase PCR (qRT-PCR) using the iTaq Universal SYBR Green One-Step kit (Biorad, Hercules, CA) and a CFX96 Touch Real-Time PCR detection system (Bio- Rad, Hercules, CA). The ΔΔCq was calculated for 3 independent samples of sorted wild type PAO1 PcdrA::gfpASV reporter biofilm cell populations by normalizing PelA and PslA to relative levels of AmpR expression. Data were presented as the average fold change in PelA or PslA expression in the PcdrA::gfpASV sorted “on” population (high GFP) relative to the “off” population (low GFP) for the three biological replicates.
WspR-YFP foci and pPcdrA::mTFP1 reporter
A version of the pPcdrA reporter was constructed in the pBBR1MCS5 plasmid to express mTFP1 instead of GFP, for use with YFP-tagged WspR proteins. The PcdrA promoter and an enhanced ribosomal binding site from the gene 10 leader sequence of the T7 phage (g10L) was amplified from pUC18-miniTn7T2-PcdrA-RBSg10L-gfpAGA using primers SacI-PcdrA-F and SOE-PcdrA-RBSg10L-R. The primers mTFP1-F and KpnI-mTFP1-R were used to amplify the mTFP1 gene from plasmid pNCS-mTFP1 (Allele Biotech, San Diego, CA). The PcdrA::RBSg10L::mTFP1 allele was constructed by SOE- PCR using primers SacI-PcdrA-F and Kpn1-mTFP1-R, then pBBR1MCS5 and the SOE PCR product were doubly digested with SacI/KpnI. Digested pBBR1MCS5 was treated with Antarctic phosphatase, then both digests were gel purified and ligated. The ligation was transformed into E. coli DH5α, and plasmid from clones growing on LB with 10 µg/mL gentamycin were sequenced with primers M13F and M13F(−21) (GeneWiz). Fluorescence of the pPcdrA::mTFP1 reporter was measured in Wsp mutants in a fluorimeter (BioTek Synergy H1 Hybrid Reader, BioTek Instruments, Inc., Winooski, VT, USA) and in flow cells to confirm its activity resembled that of pPcdrA::gfpASV. The pPcdrA::mTFP1 reporter was electroporated into P. aeruginosa strains with the native WspR deleted and harboring an arabinose-inducible copy of WspR-YFP on its chromosome (12). Cells were grown on LB agar plates with 100 µg/mL gentamycin and 1% arabinose for 10 hours, then transferred to an agar pad for imaging. WspR-YFP foci and mTFP1 fluorescence was imaged using a Nikon Ti-E inverted wide-field fluorescence microscope with a large-format scientific complementary metal-oxide semiconductor camera (sCMOS; NEO, Andor Technology, Belfast, United Kingdom) and controlled by NIS-Elements. WspR-YFP foci were detected as previously described (12).
Movie 1. Single cells are precisely illuminated by ATIM via in situ analysis and tracking of bacteria. The left panel shows the merged images of gfpASV and mCherry fluorescence microscopy images over time. The right panel shows the merged images of red LED projected patterns and bright field images corresponding to the left panel. The fluorescence intensity of gfpASV in the illuminated cells and their offspring (colored red in right panel) is significantly increased after using ATI for 460 mins. In contrast, the gfpASV fluorescence intensity of the un-illuminated cells remains low and these cells remain motile.
Acknowledgements
We thank Drs. Julie Cass and Paul Wiggins providing the wide-field microscope and cMOS camera to image WspR-eYFP clusters, Drs. Joe J. Harrison and Yasuhiko Irie for the gift of bacterial strains, and Dr. Keiji Murakami for performing c-di-GMP measurements.