pDawn-Ag43 Expression Drives Light-Regulated Biofilm Formation in E. coli.

We postulated that E. coli cells expressing Ag43 from the blue light-responsive promoter pDawn should transition from a planktonic to biofilm phenotype when illuminated by blue light. To test our hypothesis, we designed a construct (termed pDawn-Ag43) where a ribosomal binding site and the Ag43 coding sequence have been inserted downstream of the pDawn transcriptional control element (Fig. 1A). We then transformed pDawn-Ag43 into the MG1655 strain of E. coli, a weak native biofilm former (29) that is known to form biofilm with Ag43 overexpression (28).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

pDawn control of the adhesin Ag43 enables light-controlled deposition of biofilm onto surfaces. (A) Ag43 was inserted downstream of pDawn control. Escherichia coli with pDawn-Ag43 can be optically stimulated with a projector to express Ag43 and to form biofilm at surfaces. (B) Cells engineered with the pDawn-Ag43 construct form biofilm contingent upon illumination (***P < 0.001). Wild-type and pDawn controls fail to form significant biofilm, and pBAD-Ag43 expression forms biofilm regardless of illumination. Representative 96-well crystal violet stains are shown below. (C) Increasing illumination intensity increases biofilm formation, saturating past ∼41 μW/cm². Dashed curve, fit with Monod model. (D) Increasing illumination time increases biofilm formation up to approximately 8 h. Dashed curve, linear fit. (B–D, error bars represent SD, n = 4 wells.)

We seeded cultures of MG1655/pDawn-Ag43 inside a polystyrene well plate with M63 growth media and used a portable LED projector (Ivation Pro4) to provide blue light illumination during an overnight growth incubation (Fig. 1A). Using crystal violet staining for quantification (30), we found that MG1655 transformed with pDawn-Ag43 formed robust biofilm when grown under illumination, compared with weak biofilm when grown in the dark (Fig. 1B). We further verified using qPCR that light stimulation of MG1655/pDawn-Ag43 increased Ag43 mRNA transcript levels approximately 15-fold (Ag43 mRNA Transcript Levels Measured by qRT-PCR) and that this elevated gene expression did not cause a growth rate defect (MG1655+pDawn-Ag43 Growth Curves: Dark vs. Illuminated).

To verify that the pDawn-regulated Ag43 expression is in fact responsible for this light-dependent phenotype, we ran a series of controls: (i) untransformed MG1655; (ii) MG1655 transformed with pDawn-sfGFP—a pDawn plasmid expressing superfolder green fluorescent protein (sfGFP) instead of Ag43; and (iii) MG1655 transformed with pBAD-Ag43—a plasmid expressing Ag43 under the control of the pBAD promoter, induced with 10 mM arabinose (Fig. 1B). As expected, native MG1655 and MG1655 transformed with pDawn-sfGFP both failed to form strong biofilm regardless of dark or illuminated growth conditions. This indicates that at the intensity used (40 μW/cm²), blue light illumination on its own is unable to stimulate biofilm growth in MG1655, even with stimulation of the YF1/fixJ two-component sensor involved in pDawn. We verified that pDawn works as expected in MG1655/pDawn-sfGFP by confirming sfGFP expression (pDawn-sfGFP Control). On the other hand, MG1655 transformed with pBAD-Ag43 induced with arabinose formed biofilm regardless of illumination intensity, indicating that Ag43 expression is sufficient to induce biofilm formation. We conclude that pDawn-regulated expression of Ag43 is responsible for the light-dependent biofilm phenotype in MG1655/pDawn-Ag43.

Biofilm Formation Can Be Tuned by Illumination Intensity and Time.

We characterized the influence of illumination intensity and time on the extent of biofilm formation. By stimulating cells overnight for 16 h across a range of intensities from 0 to 115 μW/cm², we found that biofilm formation increases with brighter illumination in a saturating manner with a characteristic illumination intensity constant of 41 μW/cm² when fit with the Monod equation (Fig. 1C; see fit/model choice discussion in Fitting Biofilm Formation vs. Illumination Intensity/Time). When illuminating with 40 μW/cm², we found a linear increasing relationship between biofilm formation and illumination time up to ∼8 h, beyond which biofilm formation appears to slow down (Fig. 1D). For the remainder of this paper, we use standard illumination conditions of 40 μW/cm² for 16 h unless stated otherwise.

Biofilm Can Be Patterned Using Light in Various Environments.

Next, we developed a protocol to form and visualize patterned biofilms based on pDawn-Ag43. We cotransformed MG1655/pDawn-Ag43 with a red fluorescent protein (mRFP) expression plasmid. Applying the same culture conditions as before, we set up the projector using Microsoft Powerpoint to illuminate various patterns (e.g., stripes, polka dots, pictures) on bacterial cultures in polystyrene wells and subsequently verified that the patterns were recapitulated as bacterial biofilms using fluorescence microscopy (Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Biofilm Lithography enables the structured patterning of viable biofilms over multiple days and inside closed-culture chambers. (A) Various illumination images (stripes, polka dots, pictures) are recapitulated as biofilm patterns; patterns (shown on Left for reference) were illuminated from below using a projector (Fig. 1A). (B) Patterned biofilms remain stable for at least 3 d under ambient conditions in PBS buffer. (C) In culture media, growth and eventual detachment of viable biofilm can be observed over 3 d. Growth regions with increased mRFP signal and instances of biofilm detachment are marked by white rectangles and triangles, respectively. (D) Biofilms can be patterned with light inside sealed, transparent culture chambers, without requiring direct physical access to the substrate. Image contrast appears reduced compared with B and C due to a secondary biofilm image that forms at the PDMS ceiling, and planktonic bacteria were harder to wash out from the chamber without perturbing the patterned biofilm on the polystyrene.

We next investigated the long-term stability of the patterned biofilms. After overnight incubation, samples were removed from optical illumination and rinsed with PBS, and subsequent daily imaging demonstrated that the biofilm pattern remained stable over a period of 3 d in PBS (Fig. 2B). When biofilm was maintained in M63 growth medium over the same time span, we observed growth and shedding of viable biofilm (Fig. 2C), analogous to expansion and dispersal processes in natural biofilms (29). Using qPCR, we confirmed that Ag43 mRNA levels remain stable over the course of these long-term experiments (Ag43 mRNA Transcript Levels Measured by qRT-PCR).

Next, we tested this tool’s ability to pattern other materials besides polystyrene. Using glass coverslips and PDMS coupons placed into well-plates, we confirmed that our engineered cells are able to form patterned biofilms on both glass and PDMS (Patterned Biofilm Formation on Other Surfaces). For all tested substrates, no surface pretreatment or patterning was required.

Furthermore, we tested whether our method works on samples inside completely enclosed environments. This is in contrast to the requirement of direct surface access for other patterning methods such as inkjet-based printers or microcontact printing (12⇓⇓–15). We used molded PDMS cavities (31) bonded to polystyrene to create enclosed culture chambers with dimensions 19 mm × 13 mm × 1.5 mm (Fig. 2D). Bacteria were cultured in this chamber using the same illumination conditions as before. We found that patterned biofilms form as expected on the polystyrene substrate, with a faint secondary biofilm image present on the PDMS ceiling (Fig. 2D). Hence, our method can pattern biofilm inside enclosed environments.

Biofilms Can Be Optically Patterned with 25 𝝁m Resolution.

To more quantitatively characterize pDawn-Ag43–mediated biofilm patterning, we collected volumetric biofilm data via confocal laser scanning microscopy (Fig. 3A) and measured the biofilm’s average thickness to be 14.4 μm (Quantifying Biofilm Thickness and Density from Confocal Biofilm Images). Assuming an average E. coli cell volume of 1 μm³ (32) and that cells constitute 10% of total biofilm volume (33), this approximately corresponds to an average thickness of seven cells and a surface density of 1.4 × 10⁶ cells per mm². From this, we can estimate an average biofilm deposition rate on the order of 25 cells per mm²⋅s over the course of a 16 h incubation (Quantifying Biofilm Thickness and Density from Confocal Biofilm Images).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Biofilm Lithography enables patterning with a lateral resolution down to 25 μm, which is explained by transient, light-independent cell–surface attachment followed by light-activated adhesin expression. (A) Confocal microscopy reveals average biofilm thickness of 15 μm, surface roughness coefficient of 0.33, and characteristic roughness length scale of 5.4 μm. (B) Step-response analysis across light–dark boundary with high-resolution microscopy of striped illumination sample indicates ∼45μm spatial resolution; white triangles point to artifacts due to projector pixels. (C) Using electrical tape as field stop (as opposed to a projector) enables a spatial resolution of ∼25μm. (D) Biofilm patterned over large area and at high resolution are possible with photomask (shown example was originally designed for patterning microfluidic channels). (E) Higher magnification imaging of D confirms feature sizes of ∼25μm; white triangle points to an example of a spurious individual cell. (F) Schematic of Monte-Carlo simulation with cell swimming, adsorption/desorption, and light-regulated levels of adhesin A. In model 1, adhesin level is directly proportional to the adsorption rate of planktonic cells, whereas in model 2, adhesin level is inversely proportional to the desorption rate of adsorbed cells. (G) No clear biofilm boundary can be observed at the light–dark border in model 1 where optically regulated adhesin production increases adsorption rate ka. (H) Clear biofilm boundary can be observed in model 2 where optically regulated adhesin production decreases desorption rate kd. (For G and H, adsorbed cells are labeled red, illuminated region is marked in blue, and histogram of cell positions plotted below.)

The confocal images also reveal that the biofilm surface is not smooth, and we determine a surface roughness coefficient (34) of ∼0.33 (Quantifying Biofilm Roughness from Confocal Biofilm Images). This is in general agreement with the value of 0.31 estimated by assuming biofilm deposition to be a purely Poisson process (Quantifying Biofilm Roughness from Confocal Biofilm Images). Using autocorrelation analysis, we derive an approximate length scale for the surface roughness on the order of 5.7 μm (Quantifying Biofilm Roughness from Confocal Biofilm Images). We speculate that clustering on this length scale may be a result of cell division and intercellular Ag43 homodimerization leading to a different effective affinity for cell–cell vs. cell–surface binding.

Next, we determined the spatial resolution (smallest feature size) that could be patterned with this method by measuring the step response of the red fluorescence signal across a light–dark illumination boundary. Using the striped illumination sample (Fig. 2A, Top), we estimate resolution of ∼45 μm (Fig. 3B and Determining Width of Transition Region Between Dark and Illuminated Regions). This is on the order of the optical resolution limit of the projector setup, which has a pixel–pixel distance of ∼80 μm; the effect is visible in corresponding periodic structural artifacts at this length scale (Fig. 3B, white triangles). To improve this resolution, we applied electrical tape to the bottom of the well plate to act as a field stop and repeated the step-response analysis across the tape boundary. Based on this analysis, we estimate the spatial resolution to be 25 μm, measured as the width of the transition region across the light–dark boundary (Fig. 3C and Determining Width of Transition Region Between Dark and Illuminated Regions).

To validate this point further, we taped a film photomask [originally designed for microfluidic circuit fabrication (35)] to the bottom of the culture chamber. This structure was faithfully recapitulated over the area of multiple mm² (Fig. 3D) with feature sizes at the scale of 25 μm (Fig. 3E and Profile Plot of High-Resolution Biofilm Patterning). By comparing the peak RFP intensity in this image to basal RFP expression in nonilluminated regions, we estimate a contrast of 3.3 (Profile Plot of High-Resolution Biofilm Patterning). We noted that the feature sizes of the final patterned biofilm are reduced compared with the underlying photomask—that is, 25 μm-wide stripes were observed in the biofilm corresponding to 75 μm transparent stripes in the photomask (Images of Photomask). This reduction is roughly consistent with the 25 μm light–dark transition region measured earlier and suggests that the transition region lies on the illuminated side of the light–dark boundary. We also noted spurious cells attached to the surface (Fig. 3E, white triangle), which along with the transition region and surface roughness length scales constitute a spatial resolution limit for Biofilm Lithography as currently implemented.

Biophysical Modeling of Biofilm Formation Explains High Spatial Resolution.

Finally, we sought to understand mechanistically how MG1655/pDawn-Ag43 biofilm patterning can achieve high spatial resolution across illumination boundaries, given the inherent motility of the MG1655 strain.

Our initial working hypothesis (model 1, Fig. 3F) was that pDawn-Ag43 works by optically regulating Ag43 expression in freely swimming planktonic cells. The expression of surface appendages on cell membranes can help overcome cell-substrate electrostatic repulsion (36⇓–38), which could lead to increased adsorption rate and higher levels of biofilm formation in illuminated regions. However, a back-of-the-envelope calculation suggests that this model is inconsistent with the observed 25 μm spatial resolution. Motile E. coli have an effective diffusivity due to motility of Deff∼200μm2/s (39), while a very conservative lower bound on pDawn-Ag43 stimulation delay based only on gene expression time can be estimated at Tdelay∼100s (40⇓–42). Bacterial swimming during the time delay between stimulation and attachment would then blur features below a length scale of at least lblur=Tdelay×Deff∼140μm. A realistic estimate for spatial resolution would be even larger due to additional delays related to pDawn-signaling and protein accumulation/export, suggesting that model 1 is inconsistent with the observed spatial resolution given cell motility (Estimating Spatial Resolution Limits Due to Diffusion).

For another hypothesis (model 2, Fig. 3F), we note that pDawn-Ag43–mediated patterning occurs within the context of other biofilm-related processes—in particular, bacteria at a liquid–solid interface naturally exhibit reversible adsorption and desorption (43), switching between planktonic and transiently adsorbed subpopulations. Instead of only stimulating the planktonic subpopulation (by increasing their adsorption, as discussed above), pDawn-Ag43 could also have a significant effect on the desorption rate of transiently adsorbed cells. Left unstimulated in the dark, these cells readily desorb from the surface (43), but in illuminated regions, membrane-expressed Ag43 reduces the desorption rate of adsorbed cells, effectively anchoring them to the surface (38). Model 2 is consistent with high spatial resolution patterning that is not limited by bacterial motility since the adsorbed cells are immotile.

To quantitatively confirm these models, we developed a Monte-Carlo simulation of cell swimming and adsorption/desorption (Fig. 3F; see also Monte-Carlo Modeling). Individual cells are simulated to swim in a virtual culture chamber, one side of which is “illuminated,” causing increased production of adhesin A. Bacterial cells adsorb and desorb from the surface with respective rates ka and kd, which are functions of their adhesin expression level A. We simulated model 1 by setting kd constant and ka an increasing function of A and observed no clear transition at the light–dark boundary (Fig. 3G). We simulated model 2 by setting ka constant and kd a decreasing function of A and observed a clear increase in adsorbed cells in the illuminated region, with a sharp transition across the light–dark boundary (Fig. 3H). Notably, this model did not need to incorporate additional features such as increased adsorption, 3D cell–cell interactions, or a permanently attached cell state. Therefore, we establish that a biophysical model with just two key features, (i) optically controlled adhesin expression and (ii) adhesin decreasing desorption rate of adsorbed cells, is sufficient to explain light-regulated biofilm formation with high spatial resolution despite bacterial motility. Additionally, the light–dark transition region in the simulations is less than 1 μm (Higher Resolution Analysis of Light–Dark Boundary in Monte-Carlo Simulations)—smaller than the experimentally observed 25 μm transition region. This discrepancy points to unaccounted-for experimental sources of noise such as colony growth/clustering, optical scattering, and 3D effects (e.g., cell stacking), which may inform future experimental strategies to improve spatial resolution even further.