Microbial Nanoculture as an Artificial Microniche

Microbes self-organize in microcolonies while transitioning to a sessile form within a protective biofilm matrix. To enable the detailed study of microbial dynamics within these microcolonies, new sessile culture systems are needed that sequester cells and mimic their complex growth conditions and interactions. We present a new nanoliter-scale sessile culture system that is easily implemented via microfluidics-enabled fabrication. Hundreds of thousands of these nanocultures can be easily generated and imaged using conventional or confocal microscopy. Each nanoculture begins as a several nanoliter droplet of suspended cells, encapsulated by a polydimethylsiloxane (PDMS) membrane. The PDMS shell provides long-lasting mechanical support, enabling long term study, and is selectively permeable to small molecules including antibiotics, signaling molecules and functional fluorescent probes. Thus, as microcolonies mature within the nanocultures, they can be stressed or interrogated using selected probes to characterize cell physiological properties, antibiotic susceptibilities, and antagonistic interactions. We demonstrate this platform by investigating broad ranges of microcolony dynamics, including direct and indirect bacterial-fungal interactions. This versatile new tool has broad potential for addressing biological questions associated with drug resistance, chronic infections, microbiome dynamics, and antibiotic discovery.

decipher the complex nature of the biofilms 16 , and explain how community behavior can be modified by subverting the communication patterns of cells or changing the surrounding milieu, e.g. by forcing them to sense signaling molecules under physical confinement 17 . Furthermore, such systems would provide new avenues to explore, within a controlled space, cross-kingdom microbial interactions and inter-species dynamics, which are dependent on time and growth-phase, and which are critical to find new bioactive metabolites.
Confined cell culture systems that address some of these needs have recently been developed. Diffusion growth chambers are one example; such systems successfully contain microorganisms within a physical barrier, while sustaining nutrient availability and transport of secondary metabolites. One such system, the iChip, has had remarkable success in the screening for new bioactive molecules following the growth of challenging microbial species in situ 18 . However, this system has some limitations. Inoculation can be time consuming and expensive. It requires micromachining of multiple chambers (∼ 200 for iChips) to isolate cells, and allows only a single growth condition to be studied for each microfabricated platform. A second approach, based on microfluidics, generates water-in-oil-in-water (W/O/W) emulsions to encapsulate, isolate and culture bacteria cells 19 ; this is an inexpensive, easy and rapid means of isolating cells in multiple microcompartments (300,000 capsules in 10 min). However, the fragility of the emulsion droplets and limited understanding of mass transfer in these systems pose challenges for robust, long-term studies of cell growth.
Inspired by these studies, we have developed nanoliter-scale growth chambers that combine some of the advantages of both systems into a multifunctional microbial nanoculture system. Using a microfluidics method to control the amount of inoculum, and the size and the generation rate of the nanocultures, we grow microcolonies with variable cell density in a highly reproducible manner. We use polydimethylsiloxane (PDMS), an inexpensive non-toxic material, to form a robust membrane surrounding a cell suspension of 0.5-5 nanoliter; because water can permeate across the PDMS shell, the volume of the nanoculture can be further manipulated via osmotic stress. The robustness of the shell allows these compartments to be exposed to shaking/stirring conditions, which is typically necessary to facilitate gas transfer. In addition, The PDMS-based nanoculture exhibits selective permeability to various low molecular species including antibiotics, fluorescent dyes, and signaling biomolecules. We use the selective permeability as a feature to spur or alter microbial growth via the addition bioactive molecules in the external phase. Thus, we introduce a multifunctional, tunable nanoculture system to encapsulate and grow bacteria and fungi (in single or co-culture models), and to interrogate the cells with specific molecules. We illustrate the versatility of this platform in the study of complex microbial growth dynamics, as well as in inter-kingdom interactions within compartmentalized microenvironments, which could ultimately lead to the discovery of new bioactive molecules.

Results and Discussion
Characterization of PDMS Nanoculture systems. PDMS nanocultures were generated at a frequency as high as 235 Hz using a glass capillary microfluidic device ( Supplementary Fig. S1). The W/O/W double emulsions were collected in a saline solution containing 154 mM NaCl, and incubated at 37 °C for 24 h to induce crosslinking of PDMS, in the presence or the absence of 5% CO 2 according to the cell growth requirements. This procedure results in the formation of capsules with nanoliter chambers; we call each of these capsules a nanoculture (Fig. 1a). PDMS was selected because this biocompatible polymer can be cured to form mechanically robust membranes that withstand manipulation without breaking. More importantly, PDMS is known to be permeable to water as well as small molecules, which will play a crucial role in the functionality of these nanocultures 20,21 . As shown in Fig. 1b, the imposition of a strong osmotic stress (10% NaCl) for 5 h around nanocultures that had been polymerized after 24 h of incubation at 37 °C resulted in wrinkling and collapse of the PDMS nanocultures (Supplementary Video S1) without the loss of cells, indicating that the crosslinked PDMS shell is water-permeable, pliable and mechanically robust. Growth kinetics of model organisms including Pseudomonas aeruginosa and Escherichia coli (bacteria), and Candida albicans (fungus) were investigated in the PDMS nanocultures. For instance, Fig. 1c shows that P. aeruginosa PA14 cells were successfully grown overnight in nanocultures containing ultrafiltered tryptone-yeast extract broth (UFTYE, low molecular weight < 10 kDa). Real-time monitoring of PA14 cells within the PDMS nanoculture demonstrated that the exponential growth phase was reached between 6 h-14 h of incubation (Fig. 1c, Supplementary Video S2), followed by the stationary stage. This trend was very similar to the growth of PA14 in a normal culture. Cell growth within the nanoculture also corroborated with the increased volume of the nanocultures (from 4-5.2 nL), which we attributed to the generation and accumulation of free metabolites, and the associated osmotic pressure gradient across the membrane that drives swelling of the core by water permeation prior to the complete curing of the PDMS shell.
This ability of water to permeate the PDMS shell is essential to tune the microniche dynamically. For example, osmotic pressure differences between exterior and interior can be adjusted to control cell confluence within the nanoculture. Incubation of the microniches in solutions with increasing salinity (154, 600, and 1000 mM NaCl) resulted in osmotic pressure-induced shrinking of the nanocultures, and concomitant increased cell confluence (Fig. 1d1-3). The osmotic annealing also caused the shell thickness to increase (from 7 to 20 μ m), likely enhancing the mechanical durability of the shell, and making the microniche stable for long-term studies.
To test the robustness and durability of the nanoculture system, we evaluated the persistence of P. aeruginosa PA14 Δ pelA in the nanoculture for several months. P. aeruginosa can survive under nutrient-deficient conditions for a period of 12 months 22 . PA14 Δ pelA was chosen for its lack of PelA expression, a major polysaccharide of the PA14 biofilm matrix that also accounts for the initial attachment and aggregation of individual bacterial cells 23 . Preventing the formation of the biofilm matrix facilitates tracking the motion of viable cells within the nanocultures. The cells indeed persisted in the PDMS nanocultures, and remained viable under nutrient limitation over a period of 7 months (Supplementary Video S3). Meanwhile, the nanocultures remained intact, demonstrating that the robustness and durability of our system could enable the cultivation of microorganisms in their natural environment and/or advance the systematic understanding of starvation and bacterial persistence by examining phenotypic changes of the cell.
Growth control through semi-permeability of PDMS nanoculture system. While water permeation is a powerful tool, the ability to introduce small molecules such as fluorescent probes and antibiotics into the cell growth chamber would expand the utility of the nanoculture system. For instance, many bacteria are known to secrete small molecules to spur biofilm growth 24 . If such molecules can be transported across the PDMS membrane, the growth dynamics of cells residing in the nanoculture could be influenced by another community in the external phase; such studies could enable the discovery of new bioactive molecules capable of promoting or impeding cell growth.
To investigate this concept, we tested whether cells growing inside the nanoculture could be affected by the presence of small molecules outside the capsule. E. coli RP437 cells were encapsulated in the nanocultures, incubated in 154 mM NaCl solution, and grown overnight to reach stationary stage at 37 °C (Fig. 2a1). The nanocultures were then transferred to a supernatant from an overnight culture made with the same strain grown in a 250 mL flask. The result, shown in Fig. 2a2, clearly shows that the addition of the supernatant in the external environment of the nanocultures promoted cell growth compared to the control with no supplementation. To confirm that the resulting cell confluence could not be attributed to water outflow via osmotic stresses, the same experiment was repeated, with the sole exception that the nanocultures were re-inoculated in sterile growth medium only, i.e. without signaling molecules, metabolites or waste products. No significant enhancement in cell density was observed (Fig. 2a3), suggesting the increase in the cell density was indeed associated with the flux of bioactive molecules across the PDMS shell.
The utility of the nanoculture system would be further enhanced if one could assess other features like cell activity and cell vitality within the microcolonies using appropriate fluorescent probes. We demonstrate this capability by culturing P. aeruginosa PAO1 cells in the nanoculture, and adding different functional dyes to the external phase. PAO1 cells were successfully stained after re-suspending the 24 h-nanocultures in a solution containing 1.5 μ L/mL Syto 9, a green fluorescent nucleic acid dye (MW = 534.44 g/mol). The staining allowed the visualization of PAO1 accumulation at the inner wall of the nanocultures (Fig. 2b1) as well as a large number of cells that remained active and motile within the nanocultures after incubation for 24 h (Supplementary Video V4). Similarly, acridine orange, another fluorescent probe that emits a green or red fluorescence based on the interaction with double stranded DNA or single stranded RNA expressed by the PAO1 cells, respectively, was successful in staining PAO1 cells within the nanocultures after 24 h incubation (Fig. 2b2). We then characterized cell viability by means of a third functional probe, propidium iodide (PI) that detects dead cells. Surprisingly, the dead cells were detected only when 1.5 μ L/mL PI was introduced in the original cell suspension during encapsulation, instead of to the external phase (PI, MW = 668.39 g/mol). These results suggest that the PI does not cross the shell of the nanoculture, which we attribute to the low solubility of PI (a charged fluorescent dye) in PDMS 25 .
Another potential significant application of the nanoculture would be their exploitation as tools for high throughput screening for new antibiotics. The high frequency and reproducible generation of the nanoculture could facilitate the encapsulation, isolation and growth of multiple unculturable bacterial species, thought to secrete bioactive molecules with antimicrobial properties. The robustness and durability of the nanocultures could enable incubation of the microniches in their original environments (e.g. soil load, sea water). The nanoculture could be introduced in cultures of pathogenic strains to directly characterize the antimicrobial compounds secreted by competing species, especially if we can control the selectivity of the nanocultures. Furthermore, the nanoculture system could facilitate the study of multidrug resistance through screening of a library of mutants exposed to various antibiotics. However, to enable these discoveries, it is critical to understand whether antibiotics can transport through the PDMS shell and to understand what factors determine their permeability in PDMS.
We tested permeability of the nanoculture shell to various antibiotics by monitoring antibiotic susceptibility. Four classes of antibiotics, including β -lactam ampicillin (Amp), aminoglycoside tobramycin (Tob), fluoroquinolone ciprofloxacin (Cip), and tetracycline (Tet), known to have various killing mechanisms 26 , were tested on E. coli RP437 cells. All of these antibiotics exhibit strong inhibitory activities against E. coli RP437 cells when the cells are directly exposed to the drugs (Supplementary Fig. S2a). E. coli RP437 expressing a red fluorescent protein (RFP) was encapsulated in PDMS nanocultures and incubated at 37 °C for 24 h under different antibiotic concentrations. The results in Fig. 2c clearly demonstrate that the growth of RFP-expressing E. coli was affected by 50 μ g/mL Amp and 5 μ g/mL Cip, proving that the β -lactam Amp and the fluoroquinolone Cip crossed the PDMS shell. In contrast, the aminoglycoside Tob exhibited limited transport through the PDMS shell, as no growth inhibition was observed. Also, Tet (tetracylcine) was impermeable to the PDMS nanoculture. The inability of Tob (aminoglycoside) and Tet to permeate through the PDMS shell was also confirmed on nanocultures of P. aeruginosa PA14 cells, as the PA14 cells proliferated in the nanocultures incubated in the presence of 100 μ g/mL Tet or 100 μ g/mL Tob 27,28 ( Supplementary Fig. S2b).
Our first speculation was that the molecular weight of the compounds might be a factor in dictating their flux through the PDMS shell. The permeable antibiotics, Cip and Amp, have a molecular weight of 331.42 and 349.42 g/mol, respectively, smaller than 444.43 and 467.51 g/mol, the molecular weights of the impermeable antibiotics Tet and Tob, respectively. However, we have detected the transport of molecules that have molecular weight as large as 700 g/mol ( Supplementary Fig. S3) and also observed that small molecules such as glucose and sucrose do not cross the shell, suggesting that the permeability of the nanocultures is not trivially linked to the molecular weight of the molecules. Because permeation through PDMS involves the partitioning of a molecule from the external phase to the polymer phase, it is highly plausible that the solubility of the solute in PDMS is an important factor. While the partition coefficient of each molecule between water and PDMS would provide a quantitative description of such solubility 29 , such information is not readily available. As a proxy for solubility, we calculate the Flory-Huggins interaction parameter (χ ), a measure of the magnitude of enthalpic interactions between the repeat unit of PDMS and each antibiotic or solute 30,31 (See Supplementary Information for detailed calculations of χ ).
Our results (Fig. 3) demonstrate that the miscibility of PDMS with Amp (χ = 4.08), Cip(χ = 4.37), and water (χ = 5.83) are significantly greater than that of glucose (χ = 18.72), Tet (χ = 22.22), and sucrose (χ = 30.99). Glucose, Tet and sucrose have very large interaction parameters, suggesting that they do not mix well with PDMS, and therefore are essentially impermeable in PDMS. These results point to the importance of understanding the interactions between the material that forms the physical barrier for the nanoculture and the permeant by, for example, varying the composition of the shell to tailor the selective permeability of the nanoculture system. While we cannot estimate the Flory-Huggins parameter between PDMS and PI (the dye that did not cross the PDMS shell), we suspect that the low solubility of PI may be attributed to its unfavorable enthalpic interactions with PDMS.
Assessing antagonistic inter-kingdom interaction in nanoculture system. The nanoculture provides a platform to investigate cross-kingdom interactions, which are commonly encountered in biofilm-associated infections in humans, including oral, respiratory, and urinary tract diseases 32,33 . Studies of competitive or antagonistic interactions have contributed to the discovery of numerous antibacterial or antifungal compounds [34][35][36] . While high throughput screening of cells and complex biofilm models exist, there remain significant challenges associated with the analysis of inter-kingdom interactions within localized physical niches, or of the chemical exchanges that occur in situ in microcolony-like environments. Spreading the bacteria and fungi on an agar plate to study cross-kingdom interactions also presents several challenges 37 . This method requires significant supplies, limits testing to a few conditions, and may lead to false positives because of the existence of slow-growing cells. Furthermore, this approach fails to address the effects of the physical and chemical fungal-bacterial interactions on cell viability simultaneously and thus it becomes challenging to explore the full potential of the secondary metabolites produced by microorganisms in the presence of pathogens. To overcome these challenges, we chose to spatially sequester one microbial species while investigating how growth dynamics are impacted by chemical exchange (i.e., without physical contacts between cells) and communication with another species across the shell of the nanoculture. Such a situation may be especially important in modeling and studying competition or cooperation among microbes in multispecies communities, factors that are key to identifying new bioactive metabolites and signaling molecules.
Two distinct conditions were tested to illustrate antagonistic (competitive) cross-kingdom dynamics based on physicochemical interactions or chemical exchange between the fungus C. albicans and the bacterium P. aeruginosa PA14. In one batch, the bacterial strains were grown together with C. albicans as nanocultures (both bacteria and C. albicans inside, Fig. 4b1). In the other batch, C. albicans were grown alone in the nanocultures, and the C. albicans nanocultures were incubated in 24-well plates containing the bacterial culture (bacteria outside of C. albicans-containing nanocultures, Fig. 4b2). Control experiments were performed with C. albicans grown within nanocultures but without bacteria (i.e., bacteria neither inside nor in the external phase). In these experiments, confluent colonies of yeast and hyphae formed (Fig. 4a). Yeast and hyphae represent the unicellular and vegetative multicellular (or filamentous) states of C. albicans, respectively, and can be distinguished from each other by their distinct shapes; yeast are single round cells, whereas hyphae have elongated germ tubes 38 .
Our model of cross-kingdom dynamics not only reproduces the current understanding of direct (physicochemical) and indirect (chemical) Candida-Pseudomonas interactions 32 , but also illustrates how spatial isolation influences competition in polymicrobial communities. In both cross-kingdom settings (b1 and b2 in Fig. 4) C. albicans failed to form confluent colonies as occurred in the absence of P. aeruginosa PA14 cells, suggesting some competition. However, strong antagonistic interaction occurred when PA14 physically interacted with C. albicans in the nanoculture (Fig. 4b1), as the bacteria dominated the cocultures. This indicates that P. aeruginosa PA14 inhibited the growth of C. albicans and killed the co-encapsulated yeasts via chemical and/or physical interactions. Since the yeasts commonly secrete farnesol to decrease bacterial virulence 39 , it is more likely that C. albicans killing achieved in the co-encapsulated model was associated with relatively high population density of PA14 cells in the nanocultures, compared to C. albicans. Interestingly, a prior study by Hogan et al. 40 reported that physical interaction between P. aeruginosa and C. albicans in hyphae form provokes killing of the fungus, whereas our work suggest that such interactions could also be detrimental to the yeast, which were eradicated from the nanocultures. However, when the physical interaction between the two cell types is eliminated (Fig. 4b2), a small number of C. albicans in the yeast form are observed with a negligible amount of hyphae in the nanoculture (Fig. 4b2). This finding indicates that some metabolic by-products secreted by the PA14 cells, such as the phenazines 41 , could permeate into the fungal nanoculture and trigger a response by C. albicans. The yeast, sensing the bacterial chemical attack, may have activated their defense mechanism by repressing hyphal growth. Overall, the versatility of the nanocultures allows us to validate that physical interaction is not necessary to inhibit hyphal growth 32 , a finding that is essential to further explore cross-kingdom microbial pathogenesis and screen for new bioactive molecules.

Conclusions
Inspired by bacterial microcolonies that form during the initial stages of biofilm development, we have developed microbial nanocultures that artificially sequester cells within a biocompatible and semipermeable PDMS membrane. Using microfluidic technology, we can create hundreds of nanocultures per second, while controlling their size and cell density to mimic microcolony microenvironments. The nanocultures provide a unique opportunity to confine cells within a physical barrier and to investigate their response to chemical, physical and biological stressors that are selectively permeable to PDMS. We used this novel platform to evaluate cell functionality by infusing or entrapping molecular probes including antibiotics and microbial signaling molecules in the nanocultures, which led to the finding that the selectivity of the nanocultures critically depends on the miscibility of these molecules in the shell material (PDMS). The nanoculture system presented in this work is a versatile vehicle to study the short-and long-term multispecies microbial interactions and to address fundamental questions relevant to human drug-resistant infections, and antibiotic discovery. While the selective permeability of the current nanoculture depends on the interactions between PDMS and permeant, our future work will focus on engineering the encapsulation material in order to achieve selectivity toward molecules of interest. This system can lead to a novel platform to help unveil how microbes orchestrate antibiotic resistance, interspecies social behavior, and competitive cross-kingdom interactions.

Methods
Microorganisms and growth conditions. C. albicans SC 5314 42 , E. coli RP437/pRSH109 and E. coli RP437motB580/pRSH103 (a gift from Dacheng Ren 43 ), P. aeruginosa PAO1, PA14 and PA14 Δ pelA (a gift from Knut Drescher 16 ) were utilized to grow the nanocultures. The cells were cultured in buffered low-molecular weight medium referred to as ultrafiltered tryptone-yeast extract (UFTYE) broth. The UFTYE medium containing 2.5% tryptone, 1.5% yeast extract and 1% glucose had a molecular-weight cut-off of 10 kDa (Millipore). The pH of the medium was adjusted to 5 and 7 to growth C. albicans and bacteria, respectively. Briefly, 25 mL of UFTYE medium were introduced in a flask to culture the cells overnight. Then, 50 μ L aliquots were introduced in 5 mL UFTYE medium to constitute the inner phase of the microfluidics. The middle phase was encapsulated in PDMS to generate nanocultures, which were incubated at 37 °C in the presence or the absence of 5% CO 2 .
Fabrication of microfluidic devices. We make use of a glass-capillary microfluidic device with hydrodynamic flow-focusing and coflowing geometry to generate the W/O/W PDMS capsules, as described previously 44,45 . Briefly, two circular capillary tubes with inner and outer diameters of 0.58 mm and 1.03 mm (World Precision Instrument Inc.) were tapered to desired diameters using a micropipette puller (P-1000, Sutter Instrument Inc.) and a microforge (Narishige MF-830). The inner diameters of tapered tubes for the injection of a bacteria phase and the collection of capsules were 40 μ m and 200 μ m, respectively. The outside of the glass capillary tube for inner fluid was hydrophobically functionalized with octadecyltrichlorosilane (OTS). This chemical treatment enhances the wettability of the PDMS outside the capillary tube, and facilitates the formation of capsules. The two tapered capillary tubes were inserted into a square capillary with an inner dimension of 1.05 mm, and the distance between the two tubes for inner bacteria and collection was adjusted to be 120 μ m. Subsequently, a transparent epoxy was used to seal the tubes where required.
Generation of PDMS capsules with bacteria encapsulated. The microfluidic device was mounted on an inverted optical microscope (Eclipse TE 200, Nikon). Then, the three fluid phases were delivered to the microfluidic device through polyethylene tubing (Scientific Commodities) attached to syringes (SGE) that were driven by positive displacement syringe pumps (Harvard Apparatus, PHD 2000 series). The drop formation was monitored with a high-speed camera (Phantom, Vision Research) attached to the inverted microscope. The inner aqueous phase consists of bacteria suspended in culture medium; the middle phase consists PDMS (Sylgard 184, Dow Corning) with 25 wt% low viscosity silicone oil (50cSt, S159-500, Fisher Scientific) and 10 wt% curing agent. The outside phase comprises 5 wt% poly(vinyl alcohol) aqueous solution (PVA, 87-89% hydrolyzed, average MW = 13 000-23 000, Aldrich). sample. In all cases, staining was performed for 30 min and the samples were gently washed to remove excess dye. The nanocultures were visualized after 24 h of incubation at 37 °C using the Zeiss Axioimager M1 Epifluorescence and Brightfield Microscope (Zeiss, Germany). A LEICA Laser Scanning Confocal Microscope was used to perform fluorescent imaging of nanocultures of Escherichia coli RP437/pRSH 109 and Escherichia coli RP437 (motB) 580/pRSH, as well as nanocultures stained with fluorescent dyes. In all cases, the samples were collected in the Secureseal TM chamber (Grace Bio-labs) for imaging. The chambers were sealed to prevent micromotion due to drying of the samples. Antibiotic susceptibility. To assess the susceptibility of cells within the nanocultures to antibiotics, antibiotics including Amp, Cip, Tet and Tob were introduced in a suspension containing the nanocultures. The antibiotics were added at concentrations comprising between 0-100 μ g/mL before the nanocultures were incubated at 37 °C overnight. The next day, the nanocultures were image to evaluate cell growth within the nanocultures. All experiments were performed at least in triplicate.