Pattern engineering of living bacterial colonies using meniscus-driven fluidic channels

Engineering spatially organized biofilms for creating adaptive and sustainable biomaterials is a forthcoming mission of synthetic biology. Existing technologies of patterning biofilm materials suffer limitations associated with the high technical barrier and the requirements of special equipment. Here we present controlled meniscus-driven fluidics, MeniFluidics; an easily implementable technique for patterning living bacterial populations. We demonstrate multiscale patterning of living-colony and biofilm formation with submillimetre resolution. Relying on fast bacterial spreading in liquid channels, MeniFluidics allows controlled anisotropic bacterial colonies expansion both in space and time. The technique has also been applied for studying collective phenomena in confined bacterial swarming and organizing different fluorescently labelled Bacillus subtilis strains into a converged pattern. We believe that the robustness and low technical barrier of MeniFluidics offer a tool for developing living functional materials, bioengineering and bio-art, and adding to fundamental research of microbial interactions.


INTRODUCTION
Owing to their multifunctional nature, bacterial communities represent sources of inspirations and opportunities for engineers and scientists to create adaptive and programmable living functional materials (1)(2)(3).Self-organized bacterial communities, such as biofilms and swarms, exhibit multicellular behaviours and promote the survival of cells in harsh environments (4)(5)(6)(7).Through metabolic and biophysical interactions, biofilms produce valuable chemicals that can be used for nextgeneration batteries, biodegradable plastics and biomedical applications (8,9).For these collective functions and interactions, the arrangements of microbes in space and time are crucial (10,11).Bacteria also serve as a model system to study active fluids (12).Recent progress in theoretical understanding of topological active matter (13)(14)(15) raised the possibility of creating dynamic and adaptive metamaterials (16), which requires advances in experimental techniques for controlling active constituents and their collectives on multiple scales.Therefore, controlling the spatial distributions of microbes has implications for the fundamental research into engineering, biophysics, microbiology and ecology and holds a promise in material science, environmental restoration and bio-productions.Substantial amount of bioengineering effort in the last decade has been exploited for developing living cells and biomaterials that self-organize into desired patterns (17,18).For example, programable nanoobjects and biomaterials were developed by functionalizing amyloid fibril, the main protein component of biofilm matrix (19,20).Bacterial populations can be organized into patterns using inkjet printers (21), 3D printing (22), biofilm lithography (23) or genetic engineering (24)(25)(26).Yet, controlling the active matter and collective dynamics of living cells largely remains as a challenge.Furthermore, these existing methods are not widely used because of the associated high-technical barriers that demand precise control of tools, customized equipment and/or extensive genetic modifications.A robust and easy method to precisely control microbial patterns in space and time could unleash the full potential of biofilm engineering and facilitate biomaterial innovations.
This study presents a technique for micro-patterning fluid meniscus layers, hereby named as MeniFluidics.We demonstrate that local acceleration of bacterial spreading via MeniFluidics can shape biofilms into desired patterns, arrange multiple stains into controlled proximity and provide a tool to study self-organisation dynamics of active fluids in confinement.

Rationalized gel surface structuring for micropatterned channel formation
Our approach is based on the notion that a structured agar gel with a sharp concave edge is wetted by a fluid meniscus layer (Fig. 1, S1 and S2).The observed meniscus radii of curvature are much smaller than the water capillary length, which is defined by the balance of the fluid surface tension to the gravity force:   = √σ/ρg , where  and  are the surface tension and the density of the fluid, respectively.
The observed small meniscus radii can be attributed to the reduced water pressure in agar (  <   ) due to elastic contraction of the gel.After gelation, the elastic matrix of agar contracts due to evaporation of the water phase.This creates an extensile force acting on the fluid, which gives rise to a reduced pressure in the agar.When we assume a linear response of the elastic agar gel, the pressure   can be estimated as   − Δ, where Δ = / is a ratio of the evaporated volume of water () from the agar to the total volume of agar layer () after gelation, and  is the Young's elastic modulus of the agar gel.The meniscus radii,  1 , can be estimated by the Young-Laplace equation: where  1 and  2 are principal radii of the curvature.In the case of a plane meniscus (Fig. 1A-C, S1A and B), 1/ 2 is negligible and we have  1 =   Δ −1 .Measured variation of  1 as a function of Δ for 1.5% agar gel in water reveals that the decay is slightly slower than Δ −1 (Fig. S1).This indicates that agar does not act as an ideal Hookean spring, but possibly exhibits plastic deformations.The measurements estimated the Young modulus of the agar to be ~ 10 4 − 10 5 , which is in agreement with reported values (27).For Δ in a range of a few percent, we obtain the meniscus radius of curvature  1 to be in the order of a few tens of microns, accounting also for the fact that a complex medium reduces the surface tension compare to water (28).Hence, structuring agar surface should create microns-size meniscus liquid layers that provide a favourable path for bacteria to spread.Such fluidic channels, with a cross section size ℎ ≈ 0.3 1 , can be used for spatial control over biofilm patterning.

Controlled biofilm spreading by MeniFluidics
To implement the above-described approach and create liquid meniscus micro channels, we structured an agar surface by placing a micropatterned PDMS mould on an agar plate during solidification (Fig. 1B, C, see also Fig. S3 and S4).When fluorescently labelled B. subtilis cells were inoculated to the agar, random motions of cells were observed near, i.e. within ~16 m, from the edge of the micropatterned structures (Fig. 1D, Fig. S5 and Movie S1).This observation confirms the formation of meniscus layers along the structured agar surfaces.To examine whether the meniscus channels can accelerate the spreading of B. subtilis colonies, bacteria were inoculated on the flat agar surface approximately 3 mm away from a micropatterned structure (Fig. 1E).The biofilm followed the meniscus pattern as a result of accelerated spreading of the bacterial cells along MeniFluidics (Fig. 1F, G, and Movie S2).The meniscus-accelerated expansion was also confirmed with E. coli (Fig. S6).
The gained spatial control over biofilm expansion posed a question of whether the overall expansion speed can be tuned.To this end, we considered geometrically imposed control over the meniscus radii by introducing a convex structure that creates the second principle radius of curvature,  2 , for the meniscus (Fig. S2).As denoted by the Young-Laplace equation (Eq.1), introduction of the second principal radius of curvature  2 should alter  1 , thus, the MeniFluidics channel size.When Δ and  are constant, a small negative convex  2 leads to smaller  1 , while small positive concave  2 increases  1 .Accordingly, the bacteria translocation along MeniFluidic channels can cease at a small radius of curved convex structure when ℎ decreases to micron size.This suggests that highly curved structure supporting the meniscus should introduce breaks in MeniFluidics.
To test this conjecture, we introduced hinges, which can break the bacterial flow at their high curvature tips (Fig. 2A).More specifically, we designed the structures with periodically spaced hinges with a period   .This design made the biofilm expansion along MeniFluidics to be periodically halted (Fig.

2B, C, and Movie S3
).The distances between halts were in good agreement with the distance between hinges (  ), indicating that the hinge structures slowed down the expansion (Fig. 2B).The overall velocities of biofilm expansion were determined by linear regression of colony edge positions over time (Fig. S7).Examining MeniFluidics with various   , we find that the overall velocity is proportional to   (Fig. 2D).Overall, the above results indicate that MeniFluidics enables controlling the biofilm expansion both spatially and temporally.

Pattern engineering of biofilms and swarming dynamics
To demonstrate the versatility of MeniFluidics, we designed and implemented arbitrary patterns, such as a snowflake-like pattern (Fig. 3A).Bacteria were inoculated at the centre of the design and biofilm growth was monitored by time-lapse imaging.Biofilm was formed following the MeniFluidic channels and emerged into the designed shape (Fig. S8 and Movie S4).Furthermore, structured agar provided a unique platform for biophysics studies on bacterial swarming and active turbulence.We implemented MeniFluidics structure to confine the swarming colony in micron-sized domains consisting of circular, elliptical, and rectangular wells.The structured domains were populated by swarming bacteria, which facilitated bacterial turbulence phenomenon under confinement (Fig. 3B-F and Movie S5).The results revealed an interesting case of spiral vortices, vortex lattices and bacterial turbulence which are reminiscent of what have been reported in microfluidic confinement and isolated droplets (29)(30)(31).An advantage of MeniFluidics here is that it extends the life-time of collective dynamics in the system due to bacteria's unrestricted access to oxygen in contrast to the microfluidics experiments, which enables to implement the method to develop and study biological metamaterials.
The spatio-temporal control by MeniFluidics could also allow organizing different bacterial species and strains into defined spatial patterns.As a proof of concept, we designed an interdigitated pattern in which two bacterial strains can form a stripe pattern in the middle (Fig. 4A).For visualization of different strains, we used fluorescently labelled bacterial strains and inoculated them at the other ends of the pattern.Specifically, we used B. subtilis strains expressing yellow fluorescence protein (YFP) or cyan fluorescence protein (CFP).YFP-labelled and CFP-labelled strains were organized into the designed pattern and formed an alternating pattern in the middle where the distance between two strains is precisely controlled (Fig. 4B, C).A notable feature of MeniFluidics is that it does not require any special equipment or use of genetically modified strains.Even if microfabricated moulds are not readily available, the method can be applied because meniscus layers, and thus MeniFluidics, can be formed by any boundary of an object placed on an agar surface.As an example, we examined and confirmed that a metal piece placed on agar surface accelerated the expansion of a biofilm (Movie S6).

DISCUSSION
We presented a robust method of patterning bacterial populations in space and time.An exciting application of the method, among many others, could be development of living materials for bioremediation via spatial control of microbial communities.For example, MeniFluidics could be designed to facilitate bacteria with biodegradation capability to spread effectivity in a contaminated environment, where biodegradation of toxic compounds requires spatial organization of multiple engineered strains (32).The interdigitated pattern presented in this work could be useful to optimize the distances between two engineered strains for the efficiency.It can also be used to induce autoinhibition of bacterial population via sibling competition mechanism (33,34).Another application could be using the fast boundary propagation of bacteria along the MeniFluidic pattern for a quick assessment of antimicrobial resistance.The antimicrobial resistance agar plate can be modified by parallel ridges structure, while asymmetry in propagation of the colony along the menisci can indicate antimicrobial response.Furthermore, the proof of principle of the method may be extended for tissue engineering with mammalian cells because mammalian cells can only thrive in liquid, such as in MeniFluidic channels.
Our observation is in an interesting conjunction with the phenomenon known as "fungal highways" where fungi hypae and mycelium promote the dispersal of bacteria on cheese, potatoes and various surfaces (35).While the exact physical mechanism behind the fungal highways is unknown, our study suggests that fungal hypae may create natural MeniFluidics in which bacteria spreading is accelerated.
Fungal highways have been observed with various species of fungi and bacteria, which supports the idea that physics underlines this cross-kingdom interaction.MeniFluidics could allow dissecting the biochemical interactions from physical interactions between fungi and bacteria through controlled experimentation of multispecies microbiota.Therefore, this method is not only useful for engineering living materials but also to advance experimental research in microbial biophysics and ecology.
In summary, we demonstrate that MeniFluidics offers a unique method to control dynamics and spatiotemporal organisation of the living matter such as bacteria swarms and biofilms at multiple scales, which can be easily adopted to a plethora of fundamental studies and applications.

PDMS stamps and patterning agar surfaces
Patterns for MeniFluidics were fabricated on four-inch silicon wafers using a conventional soft lithography technique (see Fig. S3).The patterns were transferred to Polydimethylsiloxane (PDMS) (Sylgard 184, Dow Corning Corp) using soft lithography technique.PDMS prepolymer was mixed at a ratio of 10:1 (w/w) with curing agent and poured on top of silicon wafer mould.After degassing the mixture, PDMS was cured at 100°C for 1 hour.The cured PDMS was peeled off from Silicon wafer mould for stamping patterns on agar plates.To transfer the features onto agar plates, MSgg or LB media were poured into plastic dishes (40 mL to 150 mm plates, 20 mL to 90 mm plates) and set to solidify for 30-60 min.Solidified plates were warmed on a 70°C plate while 10 ml (for 150 mm petri dish) and 5 ml (for 90 mm petri dish) media were poured as a second layer.Autoclaved PDMS with features, prewarmed at 50°C, were gently placed on the second layer.Bubbles were removed by tapping with the end of the forceps.After solidification of the second layers, PDMS was removed.Plates were dried at 37°C overnight.

Time-lapse imaging
Biofilms formation was recorded using a digital single-lens reflex camera (DSLR, Nikon D5300) in a 30°C incubator.AF-S DX Micro NIKKOR 40mm f/2.8G macro lens and AF-P DX NIKKOR 10-20mm f/4.5-5.6GVR ultra-wide zoom lens were used.A custom-built imaging platform was assembled to place the camera perpendicular to plates (Fig. S4).The items used for the assembly of the platform are listed in Table 2. Images were taken every hour using the built-in internal time shooting function of D5300.For microscopy, images were obtained by an inverted fluorescence microscope (Leica, DMi8) with HC PL FLUOTAR 40x/0.60Ph2.

Video-microscopy of swarming and menisci radii
The swarming under confinement and the measurements of the menisci radii was recorder by inverted microscope Nikon TE2000U with 20x/0.45Ph1 lens at 50fps using Point Grey GS3-U3-23S6M-C camera (FLIR).The velocity fields from the image sequences were obtained via open source MatPIV toolbox (version 1.61).The menisci radii measurements were undertaken with 160 microns structure on 1.5% agar gel in deionised water.The agar plate has been weighted immediately after gel moulding.
The relative reduced mass of the gel, Δ, has been obtained through the mass measurements with a precision in Δ of 0.6 × 10 −3 .The image was aligned with the inner corner of the structure, then the focus had been moved to the tip of the meniscus.The intensity profile has been averaged along the image symmetry (y-axis) and the tip position has been identified manually with an uncertainty of 0.9 microns (see Fig. S1).

Figure 1 . 1 ,
Figure 1.A) Schematic illustration of meniscus layer (blue) formed on structured agar surface (beige).When 1/ 2 ~0,  1 = /Δ as denoted by the Young-Laplace equation (Eq.1), ℎ =  1 (1 −   4 ) ≈ 0.3 1 ,  is Young's modulus,  is the volume loss due to evaporation,  is the volume of agar,  is surface tension, and  is the pressure.B, C) Schematics of the MeniFluidics formation through PDMS moulding on agar.D) Projected time-lapse microscopy image of meniscus channel populated by bacteria.Colours denote time points of observation.Dashed line indicates the edge of structure (see Figure S3 for corresponding brightfield image).E) Schematic illustration of experiment where bacteria were inoculated ~3 mm away from MeniFluidics.F) Representative time evolution of a biofilm grown on agar with MeniFluidics.G) Distances of biofilm colony edge from the centre on flat agar surface (dashed grey) and MeniFluidics channel (blue).

Figure 2 .
Figure 2. A) Schematic illustration of 3D structure with hinges.  is the distance between hinges.B) Representative graphs of colony edge and C) expansion speed over time along hinged MeniFluidics with   = 4 .D) Mean overall expansion speed as a function of   .Dashed line is a linear regression.Error bars are standard deviations.

Figure 3 A
Figure 3 A) Arbitrary designed MeniFluidics patterns (left) and patterned biofilms formed by the corresponding designs (right).Scale bar, 2 cm.B) Agar well populated by swarming bacteria revealing the chaotic flow of bacterial active fluid.C) The flow field,  ⃗ , averaged over 1000 instantaneous fields reveals a mean spiral flow.D) Averaged fields in the wells of elliptical shape represent a stable elongated vortex.E) Averaged velocity field in a long rectangular groove is organized into a vortex lattice.F) Vorticity field,  = ∇ ×  ⃗ , calculated from panel D. Scale bar, 20  (B-F).Further details of this investigation will be published elsewhere.

Figure 4 .
Figure 4. Engineering of two bacterial strains into a pattern.A) Schematic of interdigitated MeniFluidic design.Yellow and cyan circles are inoculation points of fluorescently labelled bacterial strains.B) Representative fluorescence microscopy image of biofilms formed by YFP (yellow) or CFP (cyan) expressing bacteria.C) Fluorescence intensity across the dashed line in panel B.