Pseudomonas aeruginosa distinguishes surfaces by stiffness using retraction of type IV pili

To investigate the extent to which P. aeruginosa can differentiate surfaces, we assessed responses of the previously characterized PaQa reporter of Vfr activity to a wide range of agarose concentrations. To measure the activity of the PaQa promoter (refered to as P_PaQa response), we recorded individual fluorescence time-lapse movies of P. aeruginosa cells growing on a variety of different gels. First, cells were grown as a liquid culture to the early log phase (optical density OD600 = 0.1), and then a 1 μL droplet of cell suspension was added to a hydrogel. Excess liquid on the surface was allowed to dry on the hydrogel, and then the gel was added to a glass-bottom Petri dish. This confined the cells between the hydrogel and the coverslip to increase substrate interaction. Notably, TFP unspecifically bind to most substrates as evident by TFP-dependent twitching motility, e.g., on agarose gels, or optical tweezers and atomic force microscopy (AFM) experiments using plastic beads or metal cantilevers (14). We then analyzed the increase in P_PaQa fluorescence compared to the constitutive fluorescence of the rpoD house keeping gene P_rpoD over time, where the zero time point represents the P_PaQa expression in liquid. The expression of P_PaQa::yfp in liquid does not change over time (26) but increases and is maximal after approximately three hours on a gel (SI Appendix, Fig. 1 A–C). Consequently, we analyzed the median and interquartile range (IQR) of the distributions of the relative fluorescence values P_PaQa/P_rpoD (YFP/mKate2) at t = 0 (liquid) and the maximum response for each repeat at approximately t = 3 h. The medians of the ratios of these values represent the up-regulation as a response to the hydrogel compared to the expression in liquid (SI Appendix, Fig. 1D).

We found that PaQa induction increased up to a concentration of 1.5% agarose but then decreased at higher concentrations (Fig. 1A). The surprising “peaked” nature of this response could result from sensing increasing stiffness or decreasing pore size for agarose gels with increasing concentration, and it is even possible that the increasing and decreasing response regimes have different dependencies on these two key parameters. To address this question, we utilized two chemically different stiffness-tunable hydrogels: polyacrylamide (PAA) and agarose. By changing the concentration of the polymer acrylamide and its cross-linker bis-acrylamide, the stiffness of PAA gels can be adjusted between 0.1 kPa and 500 kPa (27). Similarly, by adjusting the concentration of agarose, the stiffness of agarose gels can be adjusted between 1 kPa and 1,000 kPa (28). As shown in Fig. 1B and detailed in SI Appendix, Supplementary Methods, the pore size of both gel types differs significantly (29–32). This difference in pore size and stiffness allows us to disentangle whether the PaQa response can be explained by surface stiffness or pore size. We measured the P_PaQa response to a variety of these hydrogels and plotted the results separately as a function of pore size and a function of stiffness (Fig. 1 C and D). If the P_PaQa response is a result of the changes in pore size, then we expect the P_PaQa response data for agarose and PAA to collapse onto each other as a function of pore size, but not as a function of stiffness. In contrast, if the P_PaQa response is a result of a change in stiffness, then the response for agarose and PAA should collapse as a function of stiffness, but not as a function of pore size.

We first compared the P_PaQa response on PAA to that on agarose as a function of substrate stiffness (Fig. 1A). The measured P_PaQa responses for both gel types collapsed almost perfectly onto each other over a 4-order-of-magnitude range in stiffness, suggesting that the P_PaQa response to a hydrogel is a result of a change in stiffness. This response was peaked at an intermediate stiffness of 50 to 100 kPa. To test whether this high stiffness decrease is a possible hidden effect of ever-decreasing pore sizes, we next plotted the same P_PaQa response data as a function of pore size (Fig. 1C). Here, the responses for both gel types did not collapse onto each other. In fact, both responses did not overlap at all and the response to PAA peaked below the smallest pore size of all agarose gels, supporting the idea that the observed high stiffness decrease of P_PaQa is not a hidden effect of a small pore size but is rather a response to high substrate stiffness. In contrast, the doubling time of individual cells during the P_PaQa time-lapse experiments depended on pore size but not stiffness (SI Appendix, Fig. 3). As a further control, we varied the concentration of acrylamide to cross-linker to make gels with the same stiffness but different pore sizes following the method described by Wen et al. (33). Specifically, we made three gels of different pore size for each stiffness of G’ = 1 kPa, 10 kPa, and 160 kPa. As expected, the P_PaQa response changed with gel stiffness but not pore size (Fig. 1D). We also confirmed that the lack of PaQa induction at high stiffness is not due to a general defect in gene induction; both P_PaQa::yfp and P_rpoD::mKate2 fluorescence does not decrease on these gels over time while cells grow and divide, demonstrating that both fluorophores are continuously being made (SI Appendix, Fig. 4 A–F). In addition, we also confirmed that independent promoters such as pBad and pTet are substrate stiffness–independent by including arabinose or tetracycline in the hydrogels (SI Appendix, Fig. 4G). Finally, our findings predicted that PaQa should be poorly induced by highly stiff surfaces, and we confirmed that PaQa was not significantly induced upon association with glass (Fig. 1A). We conclude that the substrate-dependent P_PaQa response is a response to stiffness and that this response is peaked at intermediate substrate stiffness.

We previously demonstrated that the P_PaQa surface response depends on TFP retraction, and we sought to determine whether this is also true for stiffness sensing. As shown in Fig. 1E, the P_PaQa response of a ΔpilTU mutant that still makes TFP but is unable to retract the TFP is independent of surface stiffness. Similarly, disrupting the sensing pathway by deleting the Vfr also abolishes stiffness sensing.

How can individual bacterial cells sense the stiffness of their surrounding substrate? Stiffness is the degree to which a material resists deformation. The stiffness of a spring, for example, is given by the spring constant $k=F/\text{Δ}x$ and is a measure for the spring’s resistance to an applied force F that yields a stretch Δx. Similarly, the 1-dimensional deformation Δx by the force F of the surface layer of a gel with the shear modulus G’ is given by $F=\frac{4}{3}\pi G\prime R\text{Δx}$, where R is the radius of the spot to which the force is applied (SI Appendix, Supplementary Methods and Fig. S5A). In other words, any stiffness-sensing mechanism must be able to apply a force sufficient to deform the surface and measure the resulting deformation. Since the stiffness-dependent P_PaQa response relies on PilTU-dependent TFP retraction (Fig. 1E), we hypothesized that substrate stiffness could be sensed by a molecular sensor that measures the force and/or deformation of the substrate when gel-attached TFP retract.

To directly determine whether P. aeruginosa PAO1 deforms substrates, we next used traction–force microscopy (TFM) to see whether TFP retraction causes local substrate deformation. In TFM, small 40 nm diameter fluorescence beads are embedded densely into the hydrogel substrate and imaged in a time-lapse experiment (34). Substrate deformations become apparent because gel-bound beads get displaced together with the substrate (Fig. 2A). TFP caused substrate deformations in front of the pole of individual cells on soft 4.5 kPa gels, intermediate 16 kPa and 45 kPa gels, and stiff 132 kPa gels (Movies S1–S4). We used single-particle tracking to measure the surface deformation caused by individual cells for the different substrates (SI Appendix, Supplementary Methods). This analysis shows that even for the stiff 132 kPa substrates, TFP can still cause measurable deformation (Fig. 2B and Movie S5). In the example shown in Fig. 2A, pilus retraction stalled slowly over 4 s and deformed the gel by ∼350 nm before slowly relaxing back. Quantifying substrate deformations for a range of gel stiffnesses shows that gel deformation is maximal for soft substrates and declines only for very stiff substrates (Fig. 2B).

Our data also enabled us to determine the size of the TFP region that pulls on the surrounding gel. The fact that substrate deformations are constant for substrate stiffnesses < 100 kPA suggests that deformations are limited by TFP length, because TFP fully retract. Only on stiff gels are surface deformations ae not limited by TFP length but by TFP retraction force. Consequently, the deformation force for stiff 132 kPa gels equals the maximum TFP stall force. Thus, setting the median deformation force F_s equal to the stall force we obtain the TFP attachment spot size $R=\frac{3}{4\pi G^{\prime }\text{Δ}x{F}_s}=0.84\hspace{0.17em}\text{nm}$. Here, the deformation force is the maximum force that TFP generate during retraction. Since this force has been shown to vary between different strains, we estimate the maximum retraction force of P. aeruginosa PAO1 to be 55 pN using optical tweezers (see SI Appendix, Fig. 6) (35–38). Interestingly, the small size of this derived contact region suggests that TFP are unlikely to interact with gels along the length of the pilus but rather attach via a small contact area, as one might expect if binding occurred at the TFP tip. Scaling the substrate deformations in this manner shows that the traction force applied to deform the substrate increases monotonically with substrate stiffness (Fig. 2C). Together, these results show that TFP can deform even stiff substrates, in support of the hypothesis that TFP retraction causes substrate deformation, which can be sensed by a molecular stiffness-sensing mechanism.

The loss of stiffness sensing in TFP retraction mutants indicates that the response to stiffness relies on the retraction of TFP. We therefore imaged TFP retraction on different surface stiffnesses by fluorescently labeling the PilA TFP subunit using a cysteine–maleimide labeling technique (39). PilA can exist in two stable forms: monomers that reside in the inner membrane and polymers that form TFP. TFP extension motors remove PilA from the membrane to polymerize them into TFP at the base of the pilus, while retraction motors depolymerize TFP to release PilA monomers into the inner membrane at the site of the pilus base. Interestingly, we observed that the local concentration of PilA at the base of dynamic TFP changes during TFP extension and retraction. We first measured the local PilA fluorescence intensity at the base of extending TFP that are not in surface contact (Fig. 3 A and B). For example, the pilus shown in Fig. 3A extended for approximately 13 s to a length of approximately 5 μm. During this time, the fluorescence intensity at the pilus base decreased by approximately 20%. Immediately afterward, the pilus retracted completely back into the cell for approximately 7 s and the fluorescence intensity at the base simultaneously increased. Interestingly, upon retraction the local PilA intensity at the pilus base initially increased to higher levels than before pilus extension, only to decrease slowly again to initial levels several seconds after the pilus was fully retracted.

We next examined the local changes in PilA levels at the pilus base in TFP attached to a stiff substrate. We thus imaged local PilA concentrations at the base of TFP bound to a stiff 4.0% (350 kPa) agarose gel (Fig. 3 C and D). Here, the pilus extended to a 3 μm length and retraction stalled immediately as the tip adhered to the gel, indicated by a straining of the pilus and a slight movement of the cell toward the TFP tip (Movie S6). As expected, the fluorescence intensity at the pilus base increased slowly upon stalling but did not overshoot the initial intensity level compared to the unbound pili (Fig. 3 A and B). Once equilibrated by diffusion (t > 15 s), the fluorescence level remained lower than the initial level prior to pilus extension. These findings are consistent with a model in which surface binding slows TFP retraction, thereby reducing the rate and extent to which PilA is recycled back into the membrane.

Our findings that PilA levels change at the pilus base during TFP extension and retraction surprised us because the rapid diffusion of proteins typically equilibrates local fluctuations in protein so quickly that they cannot be observed by traditional microscopy. However, PilA monomers are membrane-associated, and membrane-associated proteins diffuse slower than cytoplasmic proteins. If the rate at which PilA monomers diffuse in the inner membrane is slower than the rate of monomer removal or addition due to TFP extension or retraction, then the local depletion of PilA monomers by TFP assembly could overwhelm diffusion, leading to the observed transient decrease in PilA at the base. Likewise, during retraction, TFP would be locally inserted into the membrane faster than they diffuse away, leading to a transient increase in PilA at the base.

To test this hypothesis, we used fluorescence recovery after photobleaching (FRAP) to measure the diffusion coefficient of PilA in the inner membrane. We photobleached fluorescent PilA at one TFP-free pole of the cell with a focused laser for less than 1 s and measured the dynamics of fluorescence recovery due to the diffusion of unbleached monomers from the rest of the cell (SI Appendix, Supplementary Methods and Fig. 3 E and F) (40). This analysis revealed a PilA monomer diffusion coefficient of D_PilA = 0.22 μm²/s, which is comparable to, but slower than, the diffusion of other inner membrane proteins in Escherichia coli (Fig. 3G and SI Appendix, Supplementary Methods) (41). We also found that the fluorescence intensity at the bleached pole recovered to the initial level (corrected for overall photobleaching), indicating that there is no immobile fraction of PilA monomers that are not diffusing (Fig. 3H).

The rate at which PilA is locally lost during TFP extension can be compared to the effect of diffusion by calculating a corresponding effective diffusion coefficient for PilA extension: A typical pilus with length L = 1 μm and extension velocity v_ext = 350 nm/s can be described by an effective diffusion coefficient D_ext = v_ext · L = 0.35 μm²/s. Even more dramatically, the pilus shown in Fig. 3A, with length 5 μm, yields an effective diffusion coefficient of D_ext = 1.5 μm²/s. This shows that the loss of PilA due to extension of a pilus dominates over the diffusive behavior of PilA in the membrane in governing the concentration of PilA at the pilus base (42).

If changes in surface stiffness result in changes in the dynamics of the local concentrations of PilA at the base of pili, how could those PilA concentration changes be sensed in turn? Chemotaxis systems have evolved to sense rapid changes in ligand concentration, and the Pil-Chp chemotaxis-like system has previously been implicated in surface responses and cAMP induction (26, 43). Furthermore, the PilJ chemosensory receptor-like component of the Pil-Chp system has been shown to directly bind to PilA. Thus, the Pil-Chp system is an ideal candidate for translating surface-dependent changes in PilA concentration dynamics to changes in PaQa responsiveness. Most mutants in the Pil-Chp system eliminate PaQa induction altogether, making it impossible to assess the extent to which they specifically affect stiffness sensing. However, mutants in the chemotaxis adaptation systems, such as those homologous to the ChpB methylesterase from Pil-Chp, are thought to attenuate signaling without eliminating it. Consequently, we analyzed the stiffness-dependent P_PaQa response in a chpB deletion mutant. As shown in Fig. 4, this mutant is still stiffness sensitive but the P_PaQa response is highly attenuated, as it shifted by approximately 1 order of magnitude to a higher stiffness compared to the wild-type (WT) response (in other words, stiffer surfaces are required to induce PaQa in chpB than in WT). This result shows that the Pil-Chp system is important for stiffness sensing, and demonstrates that this system can tune the response to stiffness, which could be used as a form of signal processing to modulate stiffness sensitivity.

Our results show that TFP retraction deforms hydrogel substrates, that the resulting deformation and the applied traction force depends on the stiffness of the substrate, that the concentration of PilA at the base of TFP is dynamic, and that these dynamics change with substrate stiffness. Together, these results support a model where the retraction of gel-bound TFP deforms the substrate and information about the substrate stiffness is read out by a cell via surface-dependent changes in the dynamics of the local PilA concentration at the base of TFP (Fig. 5A). To establish the viability of this hypothesis and help make testable predictions, we developed a quantitative model to simulate changes in PilA concentration during TFP retraction for different substrate stiffnesses.

To explain PilA levels at the pilus base, our model needed to account for the concentration change of PilA monomers driven by polymerization/depolymerization from TFP polymers and by diffusion in the membrane. To model TFP polymerization and depolymerization dynamics, we combined the experimental TFM and optical trapping results to describe the mechanical coupling of the retraction of substrate-attached TFP and the substrate deformation in a numerical simulation (SI Appendix, Fig. 7). In addition to the effect of force on retraction speed we measured using the optical trap (SI Appendix, Fig. 6), TFP also exhibited a force-induced reversible stretching under extensile loads likely due to a conformational change in the PilA monomers. This unfolding-like feature was first demonstrated in gonococcal TFP and has also been shown for TFP in Vibrio cholerae and P. aeruginosa (44–48). As shown in SI Appendix, Fig. 7A, the TFP retraction force that deforms the substrate ramps up more quickly on stiff gels compared to soft gels. Consequently, the tension in PilA monomers increases more rapidly on stiff gels. PilA thus gets stretched more quickly on stiff gels, which increases the ratio of stretched to unstretched monomers in the pilus (SI Appendix, Fig. 7E). More unstretched monomer increases the pilus rest length and thus results in more monomers that can be retracked back into the cell compared to nonstretching TFP. Focusing on the loss of PilA from the inner membrane at a fixed time point after the initiation of TFP retraction reveals a profile that is peaked with stiffness (SI Appendix, Fig. 7H). We empirically determined that a fixed time point of 0.75 s is ideal to fit the model response to the experimental P_PaQa data and determine optimal values for the free fit parameters (the equilibrium folding and unfolding rates of individual monomers and the attachment spot size of the TFP to the substrate).

Having accounted for PilA-level dynamics driven by TFP assembly and disassembly with a numerical mechanics simulation, we next coupled that to a simulation of stochastic PilA diffusion simulation in the inner membrane (Fig. 5A). This 2-tiered combined simulation is necessary to adequately describe the competition on different time scales between PilA diffusion and the TFP extension–retraction dependent change of PilA levels. As shown in Fig. 5 B–D, TFP extension yields a substrate–stiffness independent loss of the concentration of PilA in the inner membrane. The simulated change of PilA levels during TFP retraction on a soft gel yields an increase in the local PilA concentration above the level that was present before TFP extension started, in good agreement with our experimental observation (Fig. 5B, colored lines).

Having established our two-tiered model of PilA dynamics, we next examined the dependence of the system on surface stiffness. Specifically, we analyzed the temporal changes of the local PilA concentration during extension and retraction on different stiffnesses (Fig. 5 E–G). We found that the influx (i.e., the change in the concentration) of PilA was maximal for soft and stiff gels and lower for intermediate stiffnesses. Plotting the maximum influx of PilA during TFP retraction as a function of substrate stiffness revealed that the change in the local PilA concentration is peaked at intermediate stiffness for the stretching model and quantitatively explains the experimentally observed PaQa stiffness response (Fig. 6A). This result confirms that the dynamic changes of PilA concentration that we also observed experimentally (Fig. 3) are biophysically capable of yielding a stiffness-dependent input signal for a PilA sensory system. Interestingly, if the PilA subunits are not allowed to stretch, then the maximum influx of PilA steadily decreases with substrate stiffness and does not show a peak at intermediate stiffnesses (Fig. 4 B–G, black lines). Here, we modeled the unfolding of monomers using a double-well potential function where the rate of unfolding corresponds to the Kramer’s time for first barrier crossing. Although this simple description quantitatively explains the experimental results, it is possible that the unfolding of monomers is better described by a more sophisticated model, which needs to be verified by independent single-molecule stretching experiments. Together, this result suggests that stretching of PilA may be a key aspect of the mechanism that yields a peaked stiffness response.

Our experiments and modeling suggest that PaQa stiffness sensing is determined by the competition between PilA diffusion and the surface stiffness–dependent outflux/influx of PilA during TFP extension–retraction. To further test this hypothesis, we investigated the effect of changes in the extension and retraction velocity on stiffness sensing. Both parameters are experimentally accessible by mutations in the ATPase activity of the PilB and PilT motors. We note that while it would also be interesting to change the diffusion of PilA in the inner membrane, this is not currently experimentally tractable.

We used two genetic mutants that separately affect the extension and retraction velocity of TFP: PilB^D355N reduces the extension velocity on agarose from 350 nm/s to 220 nm/s (SI Appendix, Fig. 8), and PilT^H222A reduces the retraction velocity from 600 nm/s to 250 nm/s (39). Using these parameters in our model predicted that both mutants should attenuate the sensitivity of stiffness sensing, with the pilT mutant having a stronger effect than the pilB mutant (Fig. 5 B and C). To test these quantitative predictions of our model, we measured the stiffness-dependent P_PaQa response in both PilB^D355N and PilT^H222A mutant backgrounds. We found that the model accurately predicted the direction and magnitude of the attenuated stiffness-sensing response in both mutants (Fig. 5 B and C). Further, we note that the difference in the location of the P_PaQa peak as a function of substrate stiffness between agarose and PAA gels (see Fig. 1A) may originate from a difference in the binding of the pilus to the gel, as predicted by our model (SI Appendix, Fig. 9). Together, these results demonstrate that the competition of slow PilA diffusion to fast PilA influx during TFP retraction may determine stiffness sensing and the peaked stiffness response.

Information on cloning, plasmids, and primers used in this study can be found in the SI Appendix, Supplementary Methods and Tables 2–4.

P. aeruginosa PAO1 was grown in liquid lysogeny broth (LB) Miller (Difco) in a floor shaker, on LB Miller agar (1.5% Bacto Agar), on Vogel-Bonner minimal medium agar (200 mg/L MgSO₄ 7H2O, 2 g/L citric acid, 10 g/L K₂HPO₄, 3.5 g/L NaNH₄HPO₄ 4 H₂O, and 1.5% agar), and on no-salt LB agar (10 g/L tryptone, 5 g/L yeast extract, and 1.5% agar) at 30 °C (for cloning) or at 37 °C. E. coli S17 was grown in liquid LB Miller (Difco) in a floor shaker and on LB Miller agar (1.5% Bacto Agar) at 30 °C or at 37 °C. Antibiotics were used at the following concentrations: 200 μg/mL carbenicillin in liquid (300 μg/mL on plates) or 10 μg/mL gentamycin in liquid (30 μg/mL on plates) or 10 μg/mL anhydrotetracycline in liquid for P. aeruginosa, and 100 μg/mL carbenicillin in liquid (100 μg/mL on plates) or 30 μg/mL gentamycin in liquid (30 μg/mL on plates) for E. coli.