Harnessing the hygroscopic and biofluorescent behaviors of genetically tractable microbial cells to design biohybrid wearables.

Cells' biomechanical responses to external stimuli have been intensively studied but rarely implemented into devices that interact with the human body. We demonstrate that the hygroscopic and biofluorescent behaviors of living cells can be engineered to design biohybrid wearables, which give multifunctional responsiveness to human sweat. By depositing genetically tractable microbes on a humidity-inert material to form a heterogeneous multilayered structure, we obtained biohybrid films that can reversibly change shape and biofluorescence intensity within a few seconds in response to environmental humidity gradients. Experimental characterization and mechanical modeling of the film were performed to guide the design of a wearable running suit and a fluorescent shoe prototype with bio-flaps that dynamically modulates ventilation in synergy with the body's need for cooling.

INTRODUCTION

In nature, the basic units of life are cells, which can dynamically adjust their behaviors biochemically () or biomechanically () in response to signals indicating environmental change, such as nutrient levels (), shear stress (), and electrical pulses (). Among those stimuli, the moisture gradient is an intriguing factor, which can trigger shape transformation in plant due to mechanical amplification of moisture-induced strains in hygroscopic tissues [for example, pine cone scales () and wheat awns ()]. Inspired by this hygroscopic information-processing concept (), researchers previously developed active materials with unique geometric constraints or structural heterogeneities, including polymers for fabricating hygrovoltaic generators (), hydrogels for forming complex micropatterns (), spores for harvesting energy (), and wood for building innovative pavilions (). These previous studies primarily focused on innovative applications by using nonliving materials, the chemical performance and mechanical performance of which are predefined before fabrication. For instance, spores or harvested wood materials are not biologically active and are insusceptible to direct genetic engineering that can introduce new properties or responsiveness to the system. Therefore, the application domains involving customization are limited because of the nature of the raw materials. By contrast, microbes are composed of a portfolio of single-celled living organisms that can be easily genetically modified to acquire new functions and are amenable to production in large-scale bioreactors, but have not yet been fully exploited to create environment-responsive materials.

Here, we propose to use genetically tractable microbial cells to create multifunctional, moisture-responsive interfaces. Generally, bio-derived materials, including nucleic acids (), proteins (), and polysaccharides (), can change their structures or volumes when surrounding water activity is altered. Because these three materials are the basic biological components that form cellular structure, we hypothesized that microbial cells can be used as functional building blocks for constructing moisture-responsive materials. We speculated that a bilayer-structured biohybrid film could change shape and functions (fluorescence as an example) simultaneously with humidity change (Fig. 1). The concept of dual responsiveness is illustrated at the macroscopic level (Fig. 1, A and B), microscopic level (Fig. 1, C and D), cellular level (Fig. 1, E and F), and molecular level (Fig. 1, G and H). If this hypothesis were true, we could easily embed other functionalities into the existing moisture-responsive cells through modern genetic engineering technologies, such as clustered regularly interspaced short palindromic repeats (CRISPR)–Cas9 (), for instance, to facilely introduce fluorophores (), colors (), odors (), or other metabolic responsiveness induced by environmental change (). This will open up a new application space for using hygroscopic transforming materials to tackle real-world problems related to moisture (for example, indoor humidity control and warts caused by hyperhidrosis). One intriguing application domain is designing wearable devices () to accommodate body sweat () for high-intensity working professionals including athletes and soldiers, to prevent frostbite or hypothermia. This research domain for regulating body condition has not been intensively pursued previously because of the lack of multifunctional hygroscopic materials that are facilely accessible, easy to process, and low-cost.

Fig. 1

An illustration showing the reversible transformation induced by the moisture gradient at different scales for a bilayer biohybrid film.

(A and B) The film bends tangibly at low humidity levels (A) and becomes flat and glows at high humidity levels (B). (C and D) A cross section of the biohybrid film at the microscopic level, where dehydration of the cell layer (dark green and light green) coated on top of an inert thin film (black) enables the bending of the film at low humidity levels (C), whereas the film becomes flat at high humidity levels (D) through rehydration. The box on the top left corner indicates the contraction (C) and expansion (D) of cells under two conditions. (E and F) The change of cell size and cellular fluorescence with humidity levels due to moisture desorption (E) and adsorption (F) at the cellular level. (G and H) An example of the conformational change of intracellular eGFP at the molecular level due to water removal (G) and water binding (H) at different humidity levels.

RESULTS

To prove that microbial cells can be used to create a hygroscopic structure that exhibits reversible shape transformation, we used the most common type of Escherichia coli cells to fabricate the biohybrid film. First, the cells were cultured and harvested to prepare a concentrated suspension for printing, before which all the nutrients (for example, sugar) were removed. Uncoated natural latex sheets (cis-1,4-polyisoprene, 150 to 500 μm thick) with rough surfaces (fig. S1) were selected as the moisture-inert materials, which demonstrated good bacterial biofilm adhesion through hydrophobic interactions () in the absence of additional chemical modification. By microprinting these cells (thickness, 1 to 5 μm) in parallel lines on the latex surface (fig. S2), a multifunctional biohybrid film was produced. When the film was introduced into dry conditions [relative humidity (RH), 15%], it immediately started to bend because of the contractile forces generated by cell dehydration (Fig. 2A), whereas the film flattened (Fig. 2B) with increasing environmental humidity (RH, 95%) by regaining water. This bending behavior (movie S1) was reversible over several cycles showing no hysteresis or distortion (Fig. 2C), as characterized in a customized humidity-controlled chamber (fig. S3). Using atomic force microscopy (AFM) imaging with humidity control (fig. S4), we observed similar phenomena at the cellular level, where single cells swelled or shrank along all three axes in response to changes in humidity (Fig. 2, D to F). This indicated that the bending effect of the tangible film was induced by the net force generated through the volume change of each individual cell (movie S2).

Fig. 2

Characterization of bilayer biohybrid films at different RH levels.

(A and B) Shape transformation of a biohybrid film (1.2 cm × 0.9 cm) with a bilayer structure. The top layer is composed of E. coli cells (1 μm thick), and the bottom layer is a latex sheet (200 μm thick). It bends at 15% RH (A) and becomes flat at 95% RH (B). (C) The bending curvature of this biohybrid film at different RH levels. (D and E) Topological images of a cell at 15% RH (C) and 95% RH (D) obtained from AFM. (F) The cell volume change at different RH levels scanned by AFM. (G and H) Fluorescence images of a biohybrid film coated with E. coli with eGFP expression. It shows little fluorescence at 15% RH (G) compared with that at 95% RH (H). (I) Fluorescence intensity varies along with the bending curvature of the film when exposed to humid air in a dry environment. RFU, relative fluorescence units. (J) Bending angle for different types of cells. (K) Bending angle for major cellular biological components. (L and M) Simulation (symbol S) and theoretical model (symbol T) are consistent with experimentally measured bending (symbol E) for varying numbers of cell layer (symbol L) thickness (L) or latex substrate thickness (M).

Beyond mechanical response to moisture, we sought to introduce additional functionalities through genetic engineering of the E. coli cells. Here, we used enhanced green fluorescent protein (eGFP) to demonstrate this concept. It is known that GFP will lose all of its fluoresce ability in ethanol because of dehydration (Fig. 1, G and H) (). In our experiment, we found that the fluorescence intensity of biohybrid films using GFP-expressing E. coli (fig. S5 and table S1) was correlated with humidity (Fig. 2, G to I). The fluorescence intensity of the cell significantly decreased under dry conditions, consistent with dehydration-induced conformational changes of eGFP at low levels of water activity (). When environmental humidity increased, eGFP renatured and the film re-exhibited fluorescence (movie S3). We further validated that eGFP itself changed fluorescence intensity under different conditions by imaging the cell lysate through confocal microscopy (fig. S6).

These sets of experiments verified our hypothesis that E. coli cells are naturally hydroscopic, which can process environmental information through shape transformation, whereas the multifunctionality can be easily introduced by genetic engineering. To generalize this concept to other types of microbial cells to suit the needs of different application domains, we measured the bending angles of a biohybrid film consisting of two Gram-positive bacteria (Bacillus subtilis and Rhodococcus erythropolis), two Gram-negative bacteria (E. coli and Pseudomonas nitroreducens), and baker’s yeast (Saccharomyces cerevisiae), respectively (Fig. 2J). The results showed that different types of cells have similar moisture responsiveness, which indicates that this property may arise from macromolecular building blocks (nucleic acids, proteins, carbohydrates, and lipids) that are common to the cells. To examine this hypothesis, we investigated the bending behavior of the biohybrid films prepared by printing each of the four major intracellular macromolecules on latex (Fig. 2K). The results showed that with the same dry weight, proteins induced a higher contractile force than nucleic acids and polysaccharides. As expected, lipids did not change the bending angles because the hydrophobicity of lipids prevented water absorption. This provides us with a pragmatic approach to further improve the bending performance by overexpressing proteins intracellularly using genetic engineering. The enlarged pool of hygroscopic biological materials including generally regarded as safe (GRAS) materials will definitely facilitate material selection in application and function customization. For instance, S. cerevisiae can be found on the surface of many fruits, and B. subtilis exists in soils—both sources are something humans are routinely in contact with—whereas nonpathogenic E. coli is a major member of natural human gut microbiota.

To quantitatively understand the mechanical performance of these hygroscopic materials, we measured the bending angles of the bilayer-structured biohybrid films (geometry in note S1) with varied cell layer and substrate thicknesses under different humidity conditions. The experimental data of the bending angle were then compared with a theoretical formula (Eq. 1) based on the classical Stoney formula () and the results from a three-dimensional finite-element model (fig. S7). The theoretical formula relates the bending angle, θ, to the contraction strain of the cell layer due to drying (decrease of the RH to a certain value at zero stress), , and can be expressed as (note S2)where L is the length of the strip; R is the bending radius; ts is the thickness of the latex substrate; tc is the thickness of the microbial cell layer under zero stress; α is the fraction of surface area printed with the cells; Es, νs and Ec, νc respectively denote Young’s modulus and Poisson’s ratio of the substrate and the cell layer; is the contraction strain due to the drying (decrease of the RH) of the cell layer; and . AFM revealed that the volume depends linearly on humidity at the single-cell level (Fig. 2F), which can be described by the empirical relation in terms of the normalized volume . The volume of the fully wet state is based on the measurement at 95% RH, the maximum value possible while avoiding the formation of water droplets.

The total volume change for the cell layer should also follow this relation, because the cells are densely packed to form a continuous film. Although the single cell exhibits anisotropic contraction, the average effect over a large amount of cells results in the cell layer deforming isotopically in each dimension. The contraction strain of the cell layer can thus be expressed as . Assuming that the thickness of each cell layer is 1 μm, νc = 0.3, and using a constant value of the cell layer modulus as a fitting parameter, the proposed formula is in agreement (Fig. 2, L and M) with the experimentally measured bending angle for biohybrid films with different cell layer numbers and substrate thicknesses, except for the thinnest latex layers where large strains in the cell layer may lead to rearrangements or loss of contact between cells. The fitted elastic modulus of the cell layer is 70 MPa, consistent with previous reports (); the stress and strain of the cell layer in the experiments were also estimated (fig. S8). Furthermore, the bending of the biohybrid films is well described by numerical simulation (movie S4 and note S3).

After characterizing the system, we turn to application of the biohybrid films for making wearables that respond to body sweat during motion. Because the bilayer structure curls naturally under ambient conditions (30 to 70% RH), which makes it hard to handle (for example, cutting, shaping, and assembling), we developed a sandwich structure where cell layers are coated on both sides of a moisture-inert material (Fig. 3A). This structure allows the film to respond only to a localized moisture gradient across the film (Fig. 3B) while ensuring the flatness of the film with balanced contractile forces on both sides (movie S5) in homogeneous environments (two sides exposed to the same condition). This biohybrid film with the sandwich structure is a robust fabric that responds to body sweat, where the bending degree can be adjusted and simulated by changing the thickness and elasticity of the middle supporting layer. By applying this responsive fabric that can reversibly change shape with the response to sweat production during exercise (Fig. 3C), we designed sports apparels with ventilating bio-flaps that can automatically adjust the moisture transfer and the heat resistance of the fabric through change of skin exposure percentage by modulating the area of exposed skin via shape transformation. Mechanical simulation was used to facilitate the design of the ideal shape deformation. In high heat–producing regions on the body, ventilating flaps with large unit size were used to enhance the air convection for fast moisture removal, whereas in high-sweat regions, high opening percentage was applied to ensure the permeability of the fabric (Fig. 3D) (). This design, combined with the dynamic response of the fabric to moisture gradients, could smartly reflect and respond to the body’s condition to enhance body core temperature control. With this technology, two wearables—a running suit and a shoe—were designed. First, the suit focusing on back ventilation (fig. S9) was designed on the basis of both a sweat map () and a heat map () (fig. S10) following the abovementioned principles (Fig. 3E). The flaps opened up after 5 min of exercising (movie S6) when the test participants started to feel humid (Fig. 3, F and G). In the meantime, the temperature and humidity profile on the body was tracked using iButton sensors when wearing the garment in a temperature- and humidity-stable environment (Fig. 3, H and I). It showed that the suit with functional flaps could effectively remove sweat from the body and lower the temperature of the still air between the body and the fabric, compared with nonfunctional flaps with the same geometry (control). Following the first fabrication example, we designed a running shoe with multifunctional fluorescent flaps on the sole (fig. S11). It exhibited both shape and fluorescence intensity change under different humidity conditions (Fig. 3, J to O), where uncomfortable feeling and hygiene risk due to accumulation of feet sweat (for example, warts) can be innovatively avoided.

Fig. 3

Performance of sandwich-structured biohybrid film for making sweat-responsive wearables.

(A) Shape transformation of a flat sandwich-structured biohybrid film when exposed to moisture. (B and C) Stress simulation (B) and experimental bending behavior (C) of a ventilating flap at the open stage when exposed to skin with high humidity. (D) Garment design principle considering both the amount of sweat and body temperature gradient during exercise (note S4). (E) Design of a female garment prototype based on heat maps (left, unit size) and sweat maps (right, percentage of opened area) of the back (note S5). (F and G) Images of garment prototype before exercise with flat ventilation flaps (F) and after exercise with curved ventilation flaps (G). (H and I) Temperature (H) or RH (I) profiles of stagnant air layer near volunteer skin when she wears the female garment with either functional flaps (blue) or nonfunctional flaps (orange). (J to L) The image of the shoe under transmitted light (J) and the flap on the sole at low (K) or high humidity (L). (M to O) The image of the shoe under fluorescence light (M) and the flap on the sole at low (N) or high humidity (O).

DISCUSSION

We harnessed the hygroscopic behavior of microbial cells, which provides us with a new perspective in using living materials for making moisture-responsive wearables that are multifunctional, interactive, and programmable. By characterizing and simulating the shape change effects in both homogeneous and heterogeneous environments, we designed two functional biohybrid prototypes that contain self-reproductive and genetically tractable cellular materials. These unique traits support the feasibility of scalable production of these functional materials for diverse applications.

Safety was our priority when we designed the prototypes. The garment used B. subtilis, which has the status of GRAS given by the U.S. Food and Drug Administration. Active B. subtilis can even be found in food products, such as Japanese natto soybean food, which has been shown to provide various health benefits to consumers (). Hence, we conclude that wearing the garment or being exposed to it should not incur any safety concern. In the second application, the shoes are covered by genetically modified E. coli, which does not have GRAS status, though it is a nonpathogenic strain and is commonly used in the laboratory. We used E. coli here for proof of the concept, because genetic modification of this species was easy. The use of E. coli is strictly limited to the prototype stage and will be replaced by genetically modified GRAS species or natural fluorescent microbes in the commercialization stage. An additional safeguard strategy to be explored in the future is to replace active cells with cellular components, such as proteins and nucleic acids, which have also been shown in this study to have moisture-responsive properties.

Future works should focus on the stability of the wearables. Although we have shown that the use of fabrics with microbes through 100 dry-wet cycles did not affect the bending performance of the fabrics (fig. S12), we believe that washability and durability of the wearables in different environments can be further improved by using novel design and chemistry, for example, using fibers with embedded bacteria and/or immobilizing microbes onto surfaces through covalent bonds. Additionally, wearable function maintenance (for example, cleaning and storage) should also be innovatively designed.

We foresee that this scalable and adaptable technology can be used in industries beyond fabricating wearable devices, where moisture gradient is a key factor, such as hydration enhancement in skin care products, humidity control in a smart home, and moisture removal in biomedical textiles.

MATERIALS AND METHODS

Chemicals, materials, and cells

Luria-Bertani (LB) medium, yeast extract–peptone-dextrose (YPD) medium, and agar were purchased from Becton Dickinson. The latex sheets with different thicknesses were purchased from both McMaster-Carr and MJTrends. Muscovite mica and sample mounting disks were purchased from Electron Microscopy Sciences (71856-01) and Bruker (SD-101). Glass slides for confocal imaging were purchased from VFM (SAH-2260-03A). Fluorescence source (Sola) and excitation filters for the camera were purchased from NIGHTSEA. The sensors for measuring body humidity and temperature (iButton DS1923) were purchased from Maxim Integrated. “Milliskin Matte” spandex fabric and black patterned spacer fabric were purchased from Spandex World Inc. Tuftane thermoplastic polyurethane was purchased from Perfectex Plus LLC. All other chemicals and reagents were purchased from Sigma-Aldrich. All other tools and materials were purchased from McMaster-Carr. We used the E. coli strain “MG1655_ΔrecA_ΔendA_DE3,” harboring the plasmid “pET11a-eGFP-lacZ,” which was obtained by cloning eGFP (Sequence 1) and E. coli lacZ into the commercial plasmid “pET11a” as a fusion protein. The B. subtilis strain “(pLS19) (pLS20) wild-type isolate” was purchased from Bacillus Genetic Stock Center. Baker’s yeast S. cerevisiae was isolated from Red Star. We used R. erythropolis [ATCC (American Type Culture Collection) 53968] and P. nitroreducens HBP1 [DSMZ (German Collection of Microorganisms and Cell Cultures) 8897].

Preparing printing sample

All Gram-positive bacteria (B. subtilis and R. erythropolis), Gram-negative bacteria (E. coli and P. nitroreducens), and baker’s yeast (S. cerevisiae) were cultured in 250-ml shake flasks. The cells were inoculated from a glycerol stock stored at −80°C into 50 ml of culture medium (LB for bacteria and YPD for yeast) in an Erlenmeyer shake flask and kept at 30°C with a shaking speed of 200 rpm overnight. For E. coli, additional ampicillin (100 mg/liter) was added to the medium in prior, and 0.1 mM isopropyl-β-d-thiogalactopyranoside was used to induce the expression of eGFP for another 24 hours when the culture’s optical density at 600 nm (OD600) reached 1.5 to 2.0. The cells were then harvested through centrifugation at 3000g for 5 min. Phosphate-buffered saline (PBS) was used for cell washing to remove all the medium components. Cell suspension was kept at 4°C for short-term storage. Before printing, the cells were resuspended to deionized water to prepare a ready-to-print cell suspension at a certain OD600 (ranging from 21.2 to 106). For other biological samples, it was prepared by dissolving lyophilized powder with a certain weight into deionized water or organic solvents depending on the solubility of the substance.

Deposition of biological materials onto humidity-inert substrate through bioprinting

Samples (cell suspension or biological solutions) were deposited onto the latex sheet through a microextrusion system (). The latex substrate was framed on a flat printing platform, where the gap between the printing tip and substrate was about 0.5 mm. Typically, the extrusion speed was set between 5 and 8 μl/min, whereas the printer head moving speed was set at 20 cm/min. The biological sample’s concentration or suspension density was estimated on the basis of the printing thread width, printing time, and required thickness. For the sandwich-structured biofilm, after printing one side of the substrate after it dried, the frame was flipped and the other side was then printed. Samples for bending curvature testing were cut by a rectangular cutting die.

Theoretical calculation regarding cell coating layers

On the basis of the scanning electron microscopy (SEM) images, we assumed that each individual cell coated on top of the latex film has a rectangular projection, with dimensions of 1 μm (width) × 3 μm (length), such that the area covered by each cell (Acell) is 3 μm2. During 1-min printing time, the extrusion tip will move 20 cm [tip moving speed (vtip) = 20 cm/min], whereas the volume of extruded solution is 5 μl [dispensing volumetric rate (Vdispensing) = 5 μl/min]. From literature, we found that at OD600 = 1.0, the colony-forming unit (CFU/ml, number of cells) is 1.625 × 108. In general, we printed three times in the same position to minimize the coffee stain effect. With the above setting, the width of the cell film (Wfilm) after three prints with a cell suspension OD600 at ~20 is 0.775 mm (measured data). Thus, to cover a single layer of cells on top of a certain printing length (Lfilm) when printing three times, the OD600 required can be calculated on the basis of the following equation

By substituting the parameters in the current setting, the OD600 of cell suspension required to print a single layer is 21.2. Similarly, for coating multiple cell layers (N) on the inert substrate, the concentration (OD600) of the cell suspension can be calculated (N × 21.2).

Measurement from AFM

Before AFM, 1 ml of cell suspension (OD600 = 5.0) in PBS was centrifuged at 3000g for 5 min. The supernatant was carefully removed, followed by adding 1 ml of water, and the precipitated cell was dispersed. This washing process was repeated three times. The obtained sample (2 μl) was dropped to a freshly cleaved mica substrate and incubated at room temperature. After 10 min, the sample on the substrate was washed by 50 μl of water, and this washing process was repeated two times. The sample was dried, followed by scanning on AFM (Veeco Multimode with NanoScope V) in tapping mode with an NSC15/AIBS cantilever (MikroMasch). The AFM measurements were performed in 15 to 95% RH, with 20% as a step with a customized humidifier system (similar to fig. S3 but with an outlet tubing pointed to the sample). The measurements were conducted three times with different cells. The sizes (length, width, and height) of the cell in the different RHs were analyzed on NanoScope Analysis software.

Sample bending curvature measurement

Sample bending curvature was measured in a humidity-controlled chamber (fig. S4). Before sample placement, the film was exposed to a 100% RH environment and then placed under room conditions (RH, 50%). This procedure was repeated for at least three cycles. The film was further transferred to the humidity chamber, and the bending curvature was monitored from 15 to 95% RH, with 10% as a step. The humidity was controlled through an RH sensor, which is located on the sample measurement platform. It can automatically adjust the chamber humidity to a target value by dry air and 100% humid air. The image of the bended sample was captured by a Canon EOS 5D Mark III camera, and the curvature was analyzed by customized curvature-measuring software built on top of a Rhino and Grasshopper platform.

Scanning electron microscopy

The film was coated with a 5- to 10-nm gold nanoparticle using a Gatan High Resolution Ion Beam Coater, and it was then imaged by a Zeiss Merlin high-resolution scanning electron microscope to observe high-resolution images. Noncoated samples were imaged using an FEI/Philips XL30 FEG emission scanning electron microscope.

Dark-field microscopy

Images were collected by a Nikon TE2000 inverted microscope with a 40×/0.60 numerical aperture objective using transmitted light from a 100-W halogen light source. Imaging was performed with MetaMorph software (version 7.8.4) and a Hamamatsu Orca-ER camera. Stream acquisition was performed with an exposure time of 20 ms. The images were collected with the camera chip center quadrant.

Confocal measurement

E. coli cells containing eGFP were lysed using the freeze-thaw method. The cell suspension was frozen in a −20°C freezer, and then the material was thawed at room temperature for several cycles. Cell pellets were removed by centrifugation at 3000g, while the green supernatant (containing eGFP) was transferred to a glass slide for confocal imaging. The humidity of the sample was adjusted through dry air or humid air during confocal imaging by Olympus FV300 (100×). The images were processed using Olympus FluoView version 4.2.a viewer.

Fluorescence measurement

The sample coated with cells containing eGFP was measured with a fluorescence source using a Canon EOS 5D Mark III camera with a Tiffen 49-mm yellow 12 filter with an excitation filter. The fluorescence image was analyzed by ImageJ to obtain the relative fluorescence unit before and after the humid air was applied to the samples.

Sensor on body testing

The iButton sensors (which monitor humidity and temperature) were mounted on the supporting frame of the garment with the sensing side toward skin. The resolution of the temperature measurement was set at 0.5°C, whereas the resolution of humidity was set at 0.6% RH. The data recording speed was set at 30 data points per minute. Two garments were used for testing: one with functional ventilation bio-flaps (functional garment) and the other one with nonfunctional ventilation flaps (control garment). The weight and the geometrical configuration of the two garments were the same. The temperature and humidity profiles of the participants were tracked by iButtons while each garment was worn during exercise.