Painting with light-powered bacteria

Abstract

Self-assembly is a promising route for micro- and nano-fabrication with potential to revolutionise many areas of technology, including personalised medicine. Here we demonstrate that external control of the swimming speed of microswimmers can be used to self assemble reconfigurable designer structures in situ. We implement such 'smart templated active self assembly' in a fluid environment by using spatially patterned light fields to control photon-powered strains of motile Escherichia coli bacteria. The physics and biology governing the sharpness and formation speed of patterns is investigated using a bespoke strain designed to respond quickly to changes in light intensity. Our protocol provides a distinct paradigm for self-assembly of structures on the 10 μm to mm scale.

Fig. 1

Response to light. a Population-averaged speed vs. time for initially stationary E. coli AD10, light on at t = 0 and off at t ≈ 1,400 s (black circles); green shading = illuminated, and the data point just after switch-off are 1 s apart. This sample was measured soon after the cells had run out of O₂. One hour later, cells accelerated less rapidly (magenta crosses). Full lines: fits to $\overline{v}\left(t\right)={\overline{v}}_{\mathrm{sat}}t\mathrm{∕}\left(t+{\tau }_{\mathrm{on}}\right)$. The dot-dashed line shows the acceleration (0.05 μm s⁻²) obtained from ‘box emptying’ experiments starting from stationary cells. Inset: dependence of ${\overline{v}}_{\mathrm{sat}}$ on incident light intensity (points) fitted to ${\overline{v}}_{\mathrm{sat}}\left(\mathcal{I}\right)={\overline{v}}_{\max }\mathcal{I}\mathrm{∕}\left(\mathcal{I}+{\mathcal{I}}_{1∕2}\right)$ (see Supplementary Note 1). b Strain AD10 with unc-cluster deletion shows a rapid decrease in $\overline{v}$ when illumination ceases (black circles). In a strain without unc-cluster deletion (AD57), a significant fraction of the population (see inset for the motile fraction α = 1 − β, where lines are guides to the eye) continues to swim at reduced speed, resulting in a much slower decay of $\overline{v}$ (red crosses; solid red line: offset exponential fit, giving τ ≈ 60 s). c Average swimming speed of AD10 as green light is switched off (at t = 0): τ_off ≈ 1 s independent of the initial swimming speed (at the resolution limit of our technique). d AD10 bacteria adapt their swimming speed within 1 s as the green light is alternated between low and high intensity. In this case τ_on ≈ τ_off. Error bars in all these plots depict SD from the mean speed over a range of q = 0.5–1.5 μm⁻¹

Fig. 2

Smart templated active self-assembly. Positive and negative masks projected onto initially stationary cells: a sample illuminated for 1 min with a positive ‘UoE’ (University of Edinburgh) pattern shown in inset, b same after 9 min of illumination, c 1 min after switching to the negative pattern shown in inset (with green = bright), as well as d 6 min and e 12 min after switching. Yellow square = boundary between light and dark regions of the ‘o’. f Samples illuminated with pattern of squares shown in inset. g Negative smiley pattern. h Negative strip together with i time evolution of the strip’s density profile. Scale bar in h is 100 μm and applies throughout

Fig. 3

Characterising the dynamics of pattern formation. Time evolution of normalised bacterial density (as measured by DDM, see Methods) inside squares of different dimensions L (inset, Fig. 2f), starting from a diffusive state and b uniformly swimming state. Inset: data plotted against time; main plots: data collapse when scaling time by $\sqrt{L_0\mathrm{∕}L}$ in a and L₀/L in b (both with L₀ = 540 μm). c Scaling of ‘box emptying times’ t_D, t_S, as well as ${\overline{v}}_{\mathrm{D,max}}$ (the maximum swimming speed reached when starting with diffusive cells) with square size L. d Shift of the peak position over time in the dark stripe (Fig. 2h); dashed line = power law fit of slope 0.5. Error bars in a, b are SE of the q-averages and in c, d the SD estimates returned by the fitting routines

Fig. 4

Effect of stopping distance. Response to a step illumination pattern (left bright, right dark) projected at time t = 0 by two strains starting from initially uniformly swimming cells. Faint lines show the normalised variance (a measure of cell density) profile, whereas bold lines show numerically smoothed profiles. a AD10 with τ_off ≲ 1 s and b AD57 with τ_off ≳ 1 min (corresponding images in Supplementary Fig. 3)

Fig. 5

Spatial resolution for AD10 strain. A pattern of dark lines of different width (see inset; nominal line width on the sample from left to right: 45, 22, 6, 3, 11, 45 and 90 μm) is projected onto a dense (OD ≈ 8) sample of AD10. At low swimming speeds (v = 2.7 μm s⁻¹) the 11 μm-wide line still shows up clearly while the density contrast for the thinner lines is very low. Scale bar: 100 μm