Screening microbiota for effects on host tissues

The assembly and function of microbial communities depends on many factors including the local environment and the metabolic properties of the colonizing organisms. Chemical communications or other secreted factors also play a role and are used by different microbial strains both cooperatively and competitively. The spectrum of microbial secretions have various effects on the microbe's respective hosts, both positive and negative. Thus, characterizing the roles of microbial community members and their secretions can yield key mechanistic insights into microbiome function and can lead to new intervention strategies. Focusing on the simple, yet important functional impact of toxicity, we quantify supernatant dosage responses with image data and examine the morphological effects of microbial secretions on skin‐associated host cells. Since the diversity of microbial communities, coupled with the multiplicity of host tissues requires scalable methods, we develop and demonstrate a microfluidic device that enables high‐content screening of microbial secretion effects on adherent cell types.

Bacterial species interact with and remodel their environments to create a niche for their survival in many ways, including through the molecules they secrete. They coordinate behavior through chemical communications by producing quorum sensing molecules (Abisado et al., 2018) and build cooperative or symbiotic relationships by secreting factors that support the growth of other species (McNally et al., 2014;Nogueira et al., 2009). Bacteria produce polysaccharides, nucleic acids, and proteins to form biofilms that function as protective barriers against insults such as a host immune response or antibiotics (Drenkard & Ausubel, 2002;Mann & Wozniak, 2012;den Reijer et al., 2016). Bacteria also secrete virulence factors to kill competitors (Dinges et al., 2000;Korgaonkar et al., 2013). All of these secretions generated by different strains affect the development, stability, and function of microbial communities.
Despite the functional importance of the microbiome, relatively few microbes have been well-characterized. These primarily include prominent pathogens and the most abundant and prevalent commensals. However, contributions from less well-characterized species should not be discounted as mutualistic effects are important and the estimated number of uncultivable species is estimated to be high (Pistone et al., 2021). A broader understanding of the array of secreted factors involved in bacterial community assembly, stability, function, and pathogenesis is a crucial aspect for modifying or engineering microbiome systems (X. Cao et al., 2019;Inda et al., 2019).
Here, we focus on cytotoxic secretions, though a similar approach could be used to examine other types of secretions or host effects. Most assays targeting microbial toxins have been developed with food safety in mind and most are cell-based. The majority of these assays utilize microbial sensors and target-specific toxins in specific media (Lu et al., 2020;Ye et al., 2019). Noncellbased assays include optical approaches (e.g., surface plasmon resonance) and immunoassays (e.g., ELISA) (Reverté et al., 2014).
The requirement to design around a specific target and lack of throughput are common limitations for most of these technologies.
To consider the enormous array of potentially impactful microbial products, we expose cultured cell lines to supernatants from cultures of microbial isolates, rather than focusing on specific, known toxin classes.
Initially, we experimentally characterize effects of supernatants from eight microbial taxa, both pathogenic and commensal (Table 1), on a human fibroblast cell line. We quantify dosage effects on human cell viability, and we use microscopy to examine host specific morphological effects.
Our initial characterizations reveal the potential of these experimental methods for yielding insights into the possible roles of different microbiota and for aiding application development. However, they also point to the key need for improved throughput to characterize large and diverse microbial collections (Carper et al., 2021;Tian et al., 2014;C. M. Timm et al., 2020). Therefore, we developed a microfluidic device that can be used to screen sterile spent media from bacterial cultures for cytotoxic effects, similarly to our manual methods. Indeed, microfluidic assay development is increasingly important for microbiome studies (Hansen et al., 2016;A. C. Timm, Halsted, et al., 2017;A. C. Timm, Warrick, et al., 2017), as microfluidic devices can improve throughput, can employ custom fabrication to enable coculture of organisms or improve environmental control, and can enable novel measurements while reducing cost and improving sensitivity (J. Cao et al., 2022;Tan & Toh, 2020;Valle et al., 2022). The results of our experiments show that the microfluidic device yields results similar to plate assays, but with increased sensitivity and using smaller reagent volumes. Collectively, our efforts offer a path toward broadening the lens of the functional and mechanistic understanding of the microbiome and its diverse members. The cured devices were cut out and inlet/outlets punched out with 1.5 mm biopsy punches (WellTech Rapid-Core), while additional punches to access the cell culture areas were punched with 0.75 mm biopsy punches. The PDMS devices and 2″ × 3″ glass slides were cleaned with isopropyl alcohol, exposed to air plasma (Harrick Plasma), bonded, and baked for 2 h at 65°C. The gradients produced in each device were characterized using the fluorescent tracer molecule, 40 kDa fluorescein isothiocyanate (FITC)-dextran (Millipore Sigma) diluted in phosphate buffered saline (PBS). The flow-rates of the PBS and FITC-dextran/PBS solutions into the devices were controlled using a New Era syringe pump (NE-4000). Image analysis was performed using ImageJ. Briefly, images were flat-field corrected using images taken of fluorescent standard slides and fluorescence was compared to the PBS and FITC-dextra/PBS inlets.

| Supernatant collection from bacterial cultures
The bacterial strains used in this work are summarized in Table 1

| Supernatant exposures and viability determination
For plate assays, Hs27 skin fibroblast cells were plated at 2 × 10 5 cells per well in a six-well tissue-culture treated dish, and allowed to attach overnight. Bacterial supernatants were mixed 1:2 with fresh cell-culture media and incubated with cell-culture monolayers overnight. Fresh TSB media mixed 1:2 with fresh cell-culture media was used as a negative control. Viability was determined by harvesting cells and counting live and dead cells using Trypan Blue (ThermoFisher) after 24 h incubation of bacterial supernatant. Dynamic production of toxic molecules was determined by collecting supernatant from Weisella cibaria every 6 h for 48 h, and then exposing Hs27 cells to each sample for 24 h, as done in quantification of the stationary phase samples.

| Multicell staining postexposure
For plate-based assays, each cell type was plated at 7.5 × 10 4 cells per well in 24-well tissue-culture treated plates. After monolayers were established, growth media was removed, and cells were exposed to cell   (Table 1). Hs27 cell monolayers were exposed to sterile supernatants collected from each microbe to identify strains that secreted substances toxic to these cells ( Figure 1). In these experiments, pH change as indicated by the media color was observed. pH had an effect on cell health, but compared to measured changes in Pseudomonas aeruginosa cultures, pH change was not the dominant cytotoxic effect (Supporting Information: Figure S1). We observed the expected dose-dependent cell-death after addition of supernatents collected from the known pathogens, S. aureus and P.
aeruginosa. Two strains, Amycolatopsis orientalis, which is notable for its production of the antibiotic vancomycin , and Brevibacterium epidermidis, a commensal strain that has been linked to malodor (Leyden, 1993), produced no detectable cytotoxic effects.
Three others, Bacillus subtilis, Corynebacterium xerosis, and S. epidermidis produced slight cytotoxic effects compared to the TSB media control. B. subtillis is a commensal strain. C. xerosis has been linked to infection, but is often misidentified when other Corynebacterium strains are present (Funke et al., 1996). Finally, S.
epidermidis, probably the most ubiquitous skin microbe, is considered a commensal organism, but can have pathogenic tendencies, particularly in different organ systems (Otto, 2009 produces that is most toxic to Hs27 is produced very late in the growth cycle, after stationary phase is reached (Supporting Information: Figure S2). Weisella species are found only at low prevalence on the skin, and the W. cibaria isolate that we used was isolated from the oral epithelia. W. cibaria is not considered pathogenic, making the extreme cytotoxic response oberved unexpected. In fact, some studies indicate its potential as a probiotic organism to treat skin ailments when taken orally (Lim et al., 2017). This obervation may point toward a difference in tissue susceptibility or toward a change in microbial secretions based on environment. Future efforts could include the study of microbial secretomes cultured in skin-tissue environments compared to culture in nutrient media.

| Qualitative measurements show effect of bacterial secretions on different cell types
Bacterial isolates can behave differently when exposed to different tissues (Brown & Horswill, 2020;Otto, 2009). Even within tissues, there can be a differential response by various cell types. Therefore, we selected two strains to demonstrate effects of exposure on the Hs27 skin fibroblasts, as well as a keratinocyte cell line, KerCT, melanocytes, and a model endothelial line, HUVEC. Figure 2 shows the result of an overnight incubation with cell culture media or media mixed with sterile S. epidermidis spent media, sterile S. aureus spent media, or TSB. As expected, the keratinocytes (KerCT) and skin fibroblasts (Hs27) were slightly affected when exposed to spent media from the commensal, S. epidermidis, but survived the exposure.
However, after exposure to S. aureus secretions, all KerCT and Hs27 cells appeared to have lost their nuclear DNA. S. aureus produces a variety of pore-forming enzymes that can disrupt cellular membranes, potentially explaining this observation. The HUVEC cells responded the same way to S. aureus secretions but were also negatively affected by all exposures indicating a particular sensitivity to TSB.
Finally, the melanocytes seemed resistant to S. aureus spent media.
Conformational changes, including a less elongated phenotype indicated by changes to actin filament staining, were observed but these cells retained their nuclear DNA.

| Device designs enable higher throughput microbe characterization
Based on the results above, and due to the time intensive nature of these experiments, we proceeded to develop and demonstrate a microfluidic device that fulfills the following requirements: (1) assessment of multiple host cell types; (2) quantification of dosage responses; (3) maintenance of cells; (4) facilitation of time introduction of reagents for characterization and staining, for example, for image-based analysis.

| CONCLUSION
We present experimental procedures for characterizing the effects of microbial secretions on host cells, and we demonstrate a microfluidic device aimed at higher throughput procedures to accommodate the large microbial diversity associated with many microbiome systems.
While we focus on toxicity, our approach can be applied to a broad range of functions, particularly given that the device facilitates addition of stains and acquisition of images. For demonstration purposes, we focused on bulk supernatants, but the same procedures can be applied to fractionated supernatants to help implicate specific secretions for functional roles once microbes of interest are identified by initial bulk screens. Overall, we envision utility not only in the realm of fundamental characterizations of microbiome systems, but also for a variety of application spaces, including the safety screening of probiotic candidates and the identification of useful natural products.

ACKNOWLEDGMENTS
We thank A. Shaun Francomacaro and Khamphone Inboune for fabricating the microfluidic masters used in this work. This work was supported by the Army Research Office MURI grant W911NF-14-1-0490 and internal APL funds.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ORCID
Andrea C. Timm http://orcid.org/0000-0002-0570-9217 F I G U R E 5 Hs27 cell viability after Phospholipase C exposure. The RealTime-Glo MT cell viability assay was used to measure Hs27 cell viability after exposure to different concentrations of Phospholipase C. For exposure assay in microfluidic chip, 0.25 U/mL was chosen resulting in a dose-dependent increase in cell death as shown qualitatively on the right with sytox-green (dead cells) and cell-mask orange (all cells) stains.