Bacterial strains, plasmids and oligonucleotides used are listed in Table 1. L. reuteri (ATCC 23272) and Lactobacillus rhamnosus GG (ATCC 53103) were grown in MRS (de Man, Rogosa, Sharpe) medium (De Man et al., 1960) (BD, Franklin Lakes, NJ) for 16 h at 37°C, 5% CO₂. Salmonella typhi (strain JSG698) and Citrobacter rodentium (ATCC 51459) were grown in Lysogeny broth (LB) at 37°C, 5% CO₂. Clostridium difficile (strain R20291) was grown in degassed brain-heart infusion (BHI) medium (BD, Franklin Lakes, NJ) at 37°C in an anaerobic chamber (Thermo Forma Scientific, 1025 Anaerobic System, Hampton, NJ) established with an atmosphere of 5% H₂, 85% N₂, and 10% CO₂. DLD-1 (ATCC CCL-221) human colonic cells were grown in RPMI medium supplemented with 10% fetal bovine serum at 37°C, 5% CO₂. FHs 74 Int (ATCC CCL-241) human fetal small intestinal cells were grown in Hybri-Care medium (ATCC 46-X) supplemented with 30 ng/ml epidermal growth factor (EGF) and 10% fetal bovine serum at 37°C, 5% CO₂. The gtfW deletion strain (LMW500) was constructed by insertion of a chloramphenicol resistance cassette (cat) into the gtfW open reading frame by allelic exchange as described previously (Mashburn-Warren et al., 2012). Briefly, 1kb fragments upstream and downstream of gtfW were amplified by PCR using oligos oSG1082-1083 and oSG1084-1085, followed by cloning into pFED760 (Mashburn-Warren et al., 2012) using NotI/SalI and SalI/XhoI restriction sites, respectively. The cat cassette was amplified from pEVP3 (Mashburn-Warren et al., 2012) using oligos LMW34-35, followed by cloning into pFED760 that contained the upstream and downstream fragments of gtfW using the SalI restriction site. The resulting gtfW knock-out construct plasmid (pWAR500) was then introduced into L. reuteri ATCC 23272 by electroporation. L. reuteri electrocompetent cells were prepared by growing 5 ml of culture in MRS at 37°C with 5% CO₂ until OD_600nm of ~1.0. Cells were then pelleted and resuspended in 10 ml of sterile cold 0.5 M sucrose and 10% glycerol twice, followed by a final resuspension in100 μl sterile cold 0.5 M sucrose and 10% glycerol. To this resuspension 1 μg of pWAR500 was added and the cell/DNA mixture was placed into an ice cold 2 mm electroporation cuvette (BioRad, Hercules, CA). Cells were electroporated at 2500V, 25 μF and 400 Ω using a BioRad Gene Pulser Xcell (BioRad, Hercules, CA). Immediately after electroporation, cells were resuspended in 1 mL of MRS and incubated at 30°C for 2 h, followed by serial dilution and plating onto MRS agar containing 5 μg/ml chloramphenicol and incubated at 30°C. The mutant was selected and confirmed as previously described (Chang et al., 2011).

To estimate transcription from the gtfW promoter (P_gtfW), the P_gtfW-CBluc reporter plasmid was constructed by amplifying the promoter region 350 bp upstream of the gtfW start codon (including the native ribosome binding site) by PCR using oligos oSG1102-1103. The resulting DNA fragment was inserted into pJC156 using the XhoI/SalI restriction sites. The click beetle luciferase (CBluc) gene was amplified from the Streptococcus mutans strain ldhCBGSm (Merritt et al., 2016) using oligos oSG1067-1068 and inserted downstream of the gtfW promoter region in pJC156 using SalI/NotI restriction sites. The resulting reporter plasmid pWAR501 was transformed into L. reuteri 23272 as described above to create the reporter strain LMW501.

The E. coli gtfW overexpression strain (LMW 502) was created by amplifying the L. reuteri gtfW open reading frame (including the stop codon) using primers oSG1120-1126. The resulting DNA fragment was inserted into pTXB1 (New England BioLabs, Ipswich, MA) using NheI/SapI restriction sites. The resulting plasmid, pWAR502 was then transformed into the E. coli expression strain ER2566 (New England BioLabs, Ipswich, MA) and selected on LB agar containing 100 μg/ml ampicillin and confirmed by DNA sequencing. This strain allows the overexpression of tagless GTFW protein.

To produce a L. reuteri strain constitutively expressing click beetle luciferase, a reporter plasmid was constructed by amplifying the promoter region 250 bp upstream of the elongation factor Tu (EF-Tu) start codon (including the native ribosome binding site) by PCR using oligos oSG1069-1070. The resulting DNA fragment was inserted into pJC156 using the XhoI/SalI restriction sites. The click beetle luciferase (CBluc) gene was amplified from the S. mutans strain ldhCBGSm (Merritt et al., 2016) using oligos oSG1067-1068 and inserted downstream of the EF-Tu promoter region in pJC156 using SalI/NotI restriction sites. The resulting reporter plasmid pWAR503 was transformed into L. reuteri 23272 as described above to create LMW503.

Anhydrous dextranomer microspheres (DMs; Sephadex® G-25 Superfine) were purchased from GE Healthcare Life Sciences (Pittsburgh, PA). Anhydrous cellulose microspheres (CMs; Cellulobeads D50) were obtained from Kobo Products, Inc. (South Plainfield, NJ). Anhydrous microspheres were hydrated in growth medium or water at 50 mg/ml then autoclaved for 20 min. For conditions with microspheres that contained maltose, sucrose, fructose, or glucose only, microspheres previously autoclaved in water were removed from solution on a vacuum filter apparatus and approximately 50 mg were collected via sterile loop into 1ml of filter-sterilized 1M solution of the sugar (see Figure S1). The microsphere mixture was then vortexed vigorously and incubated for 24 h at room temperature to reach equilibrium.

For application with L. reuteri, microspheres loaded with water, 1M maltose, 1M sucrose, 1M glucose, or 1M fructose were removed from solution on a vacuum filter apparatus and collected via a 10 μl sterile loop. Approximately 5 mg of hydrated microspheres were then added to 1 ml of 2 × 10⁹ CFU L. reuteri from an overnight culture that had previously been pelleted by centrifugation at 3220 × g for 10 min, washed twice with sterile 0.9% saline, and resuspended in 1 ml sterile saline. For experiments involving eukaryotic cell lines, 2 × 10⁹ CFU of bacteria were resuspended in 1 ml RPMI instead of saline. For experiments with no microspheres but equivalent volume of cargo, 10 μl of cargo was added to 1 ml of bacteria either in sterile saline or RPMI. For all experiments, the bacteria and microsphere mixture were incubated together at room temperature for 30 min (unless otherwise stated) to facilitate bacterial adherence and biofilm formation on the microsphere surface prior to use in assays.

L. reuteri culture was grown and prepared as described above and incubated with microspheres filled with either: water, 1M maltose, 1M sucrose, 1M fructose, or 1M glucose. To examine bacterial adherence to the microspheres, 300 μl of bacteria (from an overnight culture containing ~2 × 10⁹ CFU) in sterile saline and 5 mg of microspheres were combined and incubated for 5 min in a Micro Bio-Spin column (BioRad, Hercules, CA) (see Figure S2). The columns were then centrifuged (100 × g) for 1 min. The flow-through was serially diluted and plated to calculate the total number of non-adhered bacteria, and this value was subtracted from the total number of starting bacteria to derive the total number of adhered bacteria. For all experiments, a control preparation that consisted of bacteria with no microspheres was used.

The reporter strain LMW501 was grown at 37°C with 5% CO₂in MRS or MRS containing 3% glucose, sucrose, fructose, or maltose and optical densities (OD_600nm) of the cultures were measured throughout growth using an Epoch Microplate Spectrophotometer (BioTek Instruments Inc., Winooski, VT). At the indicated times, 80 μl aliquots of the bacterial cultures were mixed with 20 μl 2 mM D-luciferin in 0.1M citrate buffer, pH 6.0 and placed in a Falcon white flat-bottom 96-well plate (Becton, Dickinson Labware, Franklin Lakes, NJ), followed by luminescence detection using a Veritas Microplate Luminometer (Turner BioSystems Inc., Sunnyvale, CA).

S. mutans was grown in Todd Hewitt Broth at 37°C with 5% CO₂ until early log phase (OD_600nm ~0.3), L. reuteri WT and the ΔgtfW mutant were grown in MRS at 37°C with 5% CO₂ until late log phase (OD_{600 nm} ~1.0) for optimal gtf expression, and the E. coli gtfW overexpression strain was grown in LB broth at 37°C shaking (200 rpm) until mid-log phase (OD_{600 nm} ~0.4) followed by the addition of 1 mM IPTG to induce gtfW expression and was then grown at 37°C shaking for an additional 2 h. Whole cells of S. mutans, L. reuteri WT, L. reuteri ΔgtfW, and the E. coli gtfW overexpression strain were assayed for GTF activity as previously described (Bai et al., 2015) using Periodic acid-Schiff staining of SDS-PAGE gels.

The rate of cargo diffusion out of the microspheres was determined by tracking crystal violet, a small molecular weight dye (407.979 g/mol) (Fisher Scientific, Hampton, NJ). The microspheres were loaded with a 0.1% solution of crystal violet by incubating 20 mg of microspheres in 1 ml of 0.1% crystal violet solution either with or without added glycerol (40 or 80% v/v) overnight to reduce the diffusion rate by increasing viscosity. After 16 h, excess crystal violet solution was removed from the microspheres as described above using a vacuum filter apparatus. The crystal violet-loaded microspheres were then placed into 1 ml of water, and aliquots of water were removed and analyzed for diffusion of crystal violet into solution using an Epoch Microplate Spectrophotometer (BioTek, Winooski, VT) at OD_590nm every hour for 16 h. Percent diffusion was calculated using the equivalent amount of crystal violet within the microspheres (10 μl) in water as a control equivalent to 100% cargo diffusion.

Production of reuterin by L. reuteri was measured via a quantitative colorimetric assay (Cadieux et al., 2008). As this assay did not differentiate between similar aldehyde products, measurements included 3-HPA and any potential derivatives, such as acrolein and 3-HPA hydrate. L. reuteri was grown overnight in MRS as described above, 1 ml aliquots of 2 × 10⁹ CFU were pelleted at 3,220 × g for 10 min, washed twice with sterile saline, and resuspended in either 1 ml of sterile saline or 1 ml sterile saline containing 2% v/v glycerol. DM containing 0, 2, 10, 20, 30, 40, 50, 60, 70, or 80% glycerol were prepared as described above for other cargo, and added to the resuspended L. reuteri in saline (so that the only source of glycerol available for reuterin production was via the microsphere cargo) for 1 h at 37°C. Cells were then pelleted again and the reuterin-containing supernatant was removed, filtered through a 0.45 μm filter, and assayed for reuterin as described in Cadieux et al. (2008) without modification. A standard curve using reuterin at known concentrations was used to extrapolate bacterial-produced reuterin concentrations from DM-glycerol and the 2% v/v glycerol control experimental conditions.

Overnight cultures of WT L. reuteri were aliquoted into microcentrifuge tubes, centrifuged, washed twice with sterile saline, and resuspended in either 1 ml saline or 1 ml MRS medium. Five mg of either DM-water or DM-80% glycerol were then added to the tubes and incubated at 37°C. At hourly intervals the tubes were mixed thoroughly and aliquots were taken for subsequent serial dilution and plating for viable CFU of bacteria.

Production of histamine from L-histidine by L. reuteri was measured via ELISA (Enzo Life Sciences, Inc., Farmingdale, NY). L. reuteri was grown overnight in MRS as described above, 1 ml aliquots of 2 × 10⁹ CFU were pelleted at 3220 × g for 10 min, washed twice with sterile saline, and resuspended in one of the following conditions: sterile saline, saline with 3% maltose, saline with 2% v/v glycerol, 4 mg/ml L-histidine (Sigma-Aldrich, St. Louis, MO), 4 mg/ml L-histidine with 3% maltose, or 4 mg/ml L-histidine with 2% v/v glycerol. 5 mg of DM containing either 4 mg/ml or 30 mg/ml L-histidine were added to media lacking L-histidine, so that the only source of L-histidine for L. reuteri was as cargo diffusing out of the DMs. Each condition was then incubated at 37°C for 2 h, after which time the contents were pelleted and the supernatant was removed for histamine quantification via a histamine ELISA kit (Enzo Life Sciences, Inc., Farmingdale, NY) following the manufacturer's instructions without modifications. All conditions were done in at least triplicate.

Bacteria were exposed to a synthetic gastric acid equivalent to determine survival at pH 2. Gastric acid equivalent is a modified version of synthetic gastric fluid (Cotter et al., 2001), composed of 0.1M HCl, 0.1M NaCl, and 0.01M KCl, with pH adjusted to 2 using 0.1M NaOH. For the assay, 1 ml of 2 × 10⁹ CFU of L. reuteri from a fresh overnight culture were pelleted at 3,220 × g for 10 min, washed twice with sterile saline, and resuspended in 1 ml 0.9% sterile saline. The cells were incubated for 30 min with approximately 5 mg of loaded or unloaded microspheres as described above, and the bacteria-microsphere mixture was diluted 1:100 directly into gastric acid equivalent. Aliquots of the inoculated acid solution were mixed, serially diluted, and plated at hourly time points for 4 h to determine the number of viable bacteria. Bacteria without microspheres were used as a control.

DLD-1 colonic cells and FHs 74 small intestinal cells were cultured as described above. When the adherent epithelial cells reached confluence, the growth medium was removed, cells were washed twice with sterile phosphate buffered saline (PBS), and trypsin-EDTA (0.25%) was added for 10 min at 37°C to dislodge the cells from the culture flask surface. Total epithelial cells were counted using a hemacytometer (Hausser Scientific, Horsham, PA). Cells were then diluted to a concentration of 5 × 10⁵ cells/ml and 1 ml per well was seeded into a 24-well plate and incubated at 37°C, 5% CO₂. After either 48 h (for DLD-1 cells) or 120 h (for FHs 74 cells) of growth, the spent medium was removed and replaced with 1 ml of RPMI or Hybri-Care medium containing 2 × 10⁹ CFU of L. reuteri alone, L. reuteri with 5 mg water-filled DMs, L. reuteri with 5 mg sucrose-filled DMs, or L. reuteri with 5 mg maltose-filled DMs. After a 1 h incubation, the spent medium was removed and the well was washed with 1 ml of sterile PBS 3 times to remove non-adhered bacteria. The remaining epithelial cells, with adhered bacteria, were then trypsinized as described above, serially diluted, and plated onto solid MRS medium for enumeration of total adhered bacteria. For confocal microscopy experiments with DLD-1, Nunc Lab-Tek 8-well borosilicate chamber slides (Fisher Scientific, Hampton, NJ) were used in place of 24-well plates. The chamber slides were treated with collagen prior to DLD-1 seeding to improve cellular adherence using the following protocol: a mixture of 100 μl of 7.5% BSA (Sigma-Aldrich, St. Louis, MO), 50 μl of 3.79 mg/ml collagen (Millipore, Temecula, CA), 100 μl of 1 mg/ml rat fibronectin (Biomedical Technologies, Stoughton, MA), and 9.75 ml of PBS was prepared, and 200 μl of this solution was added per chamber slide well. After incubation for 1 h at 37°C, the solution was removed from the well, and epithelial cells were seeded and grown as described above.

Mucin agar plates were created using porcine stomach mucin (Sigma-Aldrich, St. Louis, MO). Mucin agar plates contained 2% mucin and 0.8% agar to simulate the consistency of the mucus layer found in vivo (Macfarlane et al., 2005; Van den Abbeele et al., 2009). To assess L. reuteri's ability to bind mucin, 2 × 10⁹ CFU of L. reuteri that contained a plasmid that encoded expression of the click beetle luciferase enzyme either planktonically or bound to 5 mg DM-water, DM-sucrose, or DM-maltose were incubated on both mucin agar and agar without mucin stationary at room temperature. After 60 min, the non-adhered L. reuteri were removed by washing the plates twice with sterile saline. The luciferase substrate D-luciferin (Sigma-Aldrich, St. Louis, MO) was then added to the plates at a concentration of 0.4 mM to visualize the remaining adhered bacteria. Relative luminosity generated from the bacteria on the plates was measured using a FluorChem E system (ProteinSimple, San Jose, CA) with a 20 min exposure setting. To assess the number of bacteria bound to the mucin within the plate (and not any background binding that may occur to the agar within the plate), the amount of luminescent signal from the agar-only plates was subtracted from the mucin agar plates.

All confocal laser scanning microscopy (CLSM) was performed using a Zeiss LSM 510 confocal microscope (Ziess AG, Oberkochen, Germany). For fluorescent staining, dextranomer and cellulose microspheres were pre-stained with Congo Red (Fisher Scientific, Hampton, NJ) prior to incubation with the cargo (e.g., sucrose) and experiments with bacteria. L. reuteri was stained with SYTO 9 (Life Technologies, Carlsbad, CA). Differential fluorescent visualization was performed using the following settings: Congo Red excitation 554 nm/emission 568 nm, and SYTO 9 excitation 490 nm/emission 525 nm. Samples were fixed using a custom biofilm fixative containing 1.5% paraformaldehyde, 0.025% glutaraldehyde, 4.0% acetic acid, and 0.1M phosphate buffer at pH 7.4 (Devaraj et al., 2015). All microscopy was performed on samples in Nunc Lab-Tek 8-well borosilicate chamber slides (Fisher Scientific, Hampton, NJ). For CLSM experiments with DLD-1 epithelial cells, DLD-1 was stained with 4′,6-Diamidino-2-Phenylindole (DAPI, Life Technologies, Carlsbad, CA), L. reuteri was stained with carboxyfluorescein succinimidyl ester (CFSE, Life Technologies, Carlsbad, CA). AxioVision software (Ziess AG, Oberkochen, Germany) and ICY (de Chaumont et al., 2012) were used to analyze images and create figures from CLSM images. COMSTAT (Heydorn et al., 2000) software was used to quantify bacterial biomass in CLSM images.

For in vitro biofilm assays, overnight cultures of WT and ΔgtfW L. reuteri were diluted into fresh MRS growth medium to 0.01 OD_600nm, incubated at 37°C 5% CO₂ for 2.5 h until reaching 0.65 OD_{600 nm}, diluted 1:2,500 into either MRS, MRS + 3% sucrose, or MRS + 3% maltose, seeded into 8-well borosilicate chamber slides and incubated for 1, 3, or 6 h at 37°C 5% CO₂. At the designated time intervals, the bacteria were stained for viability with LIVE/DEAD stain, fixed, visualized via confocal microscopy, and quantified via COMSTAT analysis of the fluorescent signal.

All scanning electron microscopy (SEM) was performed using a Hitachi S-4800 field emission SEM (Hitachi, Tokyo, Japan). Samples were prepared as described in “Adherence to colonic cells,” with the exception that DLD-1 human colonic epithelial cells were grown on 15 mm diameter thermanox coverslips (Electron Microscopy Sciences, Hatfield, PA) placed within the well of a 12-well plate. Samples of DLD-1 cells and adhered bacteria were fixed overnight at 4°C in a solution of 2.5% glutaraldehyde in 0.1M phosphate buffer (pH 7.2). Samples were then washed with double distilled water and stained with a 1% solution of osmium tetroxide (Sigma-Aldrich, St. Louis, MO) in 0.1M phosphate buffer (pH 7.2) for 1 h, washed for 5 min, stained with a 1% solution of thiocarbohydrazide (Sigma-Aldrich, St. Louis, MO), washed for 5 min, and further stained with 1% osmium tetroxide for 30 min. Samples were then dehydrated using a graded series of ethanol: 25% ethanol for 15 min, 50% ethanol for 15 min, 70% ethanol for 30 min, 95% ethanol for 15 min (twice), 100% ethanol (twice), a 1:1 mixture of 100% ethanol to 100% hexamethyldisilazane (HMDS, Sigma-Aldrich, St. Louis, MO) for 100 min, 100% HMDS for 15 min, and a final immersion in 100% HMDS that was allowed to air dry overnight. Dehydrated sample coverslips were then mounted onto 15 mm diameter metal SEM specimen stubs (Electron Microscopy Sciences, Hatfield, PA) using colloidal silver (Electron Microscopy Sciences, Hatfield, PA). The outer edge, where the stub and coverslip meet, was then coated with a light layer of colloidal silver, and allowed to dry overnight. Samples were sputter coated with gold and palladium for 2 min at 25 mA using an Emitech K550X sputter coater (Quorum Technologies Ltd., Laughton, United Kingdom).

All experiments were conducted a minimum of three times and statistical analysis was performed via a Student's t-test using GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA), wherein a P-value less than 0.05 was accepted as significant.

Our strategy was to have probiotic bacteria adhere to a biocompatible surface to induce the formation of a biofilm (Figure 1). To investigate this, we differentially stained DMs with Congo Red and L. reuteri with SYTO 9, and examined binding via confocal laser scanning microscopy (CLSM). As shown in Figure 2, aggregates of bacteria were associated with the surface of numerous DMs which indicated that L. reuteri was able to adhere to the DM surface within the time allotted. Since DMs are cross-linked glucan similar to the native reuteran produced by L. reuteri, we hypothesized that either an increase in GTFW (for enhanced binding to DMs) or production of glucan to stimulate aggregation and biofilm formation would facilitate the adhered state of L. reuteri. To this end, we compared adherence of L. reuteri to DMs that contained luminal cargo of either sucrose (an inducer of gtfW expression but not a substrate for GTFW; see Figure S3) or maltose (the sole substrate of GTFW). As shown in Figures 2B,C, compared to DMs that contained only water within the lumen (Figure 2A) there were greater numbers of L. reuteri adhered to DMs with either sugar as cargo.

To further investigate L. reuteri's ability to bind DMs, we tested other DM lumen compounds that we hypothesized should not affect GTFW protein mediated binding and thus unlikely to support increased adherence to DMs. For this assay we chose the monosaccharide subunits of maltose and sucrose (e.g., glucose for maltose, glucose and fructose for sucrose), which the GTF enzyme cannot utilize to catalyze glucan polymers. Interestingly, fructose (and not glucose) was shown to induce gtfW expression at a rate similar to sucrose, but did not result in enhanced binding to DMs as was found with sucrose (Figure S3A, Figure 3A).

To determine if this GTFW-dependent binding is specific to the glycosyl linkages of DMs, we compared L. reuteri binding to cellulose microspheres (CMs), as DMs are composed of polymers of glucose with α-linkages while CMs possess β-linkages between the glucose units (Updegraff, 1969; Kralj et al., 2002). As shown in Figure 3A, only ~10% of L. reuteri adhered to CMs regardless of luminal contents. Collectively the data in Figure 3 indicated that L. reuteri does not bind to CMs, binding to DMs was GTFW-dependent and further, that inclusion of maltose or sucrose significantly enhanced the binding of L reuteri to DMs. We hypothesized that the predicted glucan binding domain of GTFW is a necessary component of L. reuteri's ability to adhere to DMs. To further test if the adherence to DM is GTF-dependent, we created a mutant strain of L. reuteri (LMW500) with a chloramphenicol resistance gene inserted in place of the gtfW gene. As shown in Figure 3B, the ΔgtfW strain was not able to bind to DMs as effectively as the wild type (WT) in our spin column assay, regardless of the cargo within the DM lumen. To further demonstrate the difference between the WT and ΔgtfW, we examined biofilm formation on glass chamber slides in media supplemented with sucrose or maltose (Figure S4). After a 1 h incubation, the WT had more bacteria present and noticeably more bacterial aggregation when sucrose or maltose was added to the growth medium (Figures S4A,B). After 3 and 6 h with sucrose or maltose supplemented media, the WT displayed a significantly more robust biofilm with greater biomass compared to the gtfW mutant under every condition, with significantly more cells present when sucrose or maltose was in the growth medium (Figures S4A,C,D).

We next tested whether bacteria that do not express a similar GTF would lack the adherent phenotype shown in Figures 3A,B. To examine this, we performed our DM adherence assay with another probiotic bacterium and three enteric pathogens that L. reuteri would likely encounter within the gastrointestinal tract: Lactobacillus rhamnosus GG, a Gram-positive bacterium commonly found in the genitourinary system and sold commercially as a probiotic; Salmonella typhi, a Gram-negative bacterium responsible for typhoid fever in humans; Citrobacter rodentium, a Gram-negative bacterium that causes colitis in rodents; and Clostridium difficile, a Gram-positive spore-forming bacterium that can cause severe colitis and recurring infections in humans. As shown in Figure 3C, all of the non-GTF expressing bacteria showed minimal adherence to DMs, regardless of cargo present within the DM lumen.

Initial binding of bacteria to DMs is a critical component of our formulation, however equally important is the ability to co-deliver beneficial luminal cargo needed by the adherent bacteria during transit of DMs through the gastrointestinal tract. Targeted delivery of maltose (or any other beneficial compound) via diffusion out of the DMs directly to the probiotic bacterium over time was a desired feature of our system (Figure 1). However, since the method of cargo delivery would be diffusion through the porous surface of the microsphere and not its degradation, such as occurs in poly(lactic-co-glycolic) acid (PLGA) microspheres (Danhier et al., 2012), the rate of diffusion is dependent upon the size of the microsphere, the mass of the solute, and the viscosity of the diluent. As proof of concept, we filled the DMs with crystal violet, a small molecular weight stain (407.979 g/mol), and tested the diffusion rate of the dye out of the DMs with and without changing the viscosity of the solution in the DM lumen. As shown in Figure 4, the crystal violet diffused out of the DM lumen with a half-life of ~6 h. When the viscosity was increased by adding 40% glycerol, the half-life of release was increased to ~8 h. At 80% glycerol, the half-life of crystal violet release was further enhanced to 12 h. By 16 h >95% of all of the crystal violet had been released under all tested conditions.

An important feature of L. reuteri's function as a probiotic bacterium is its ability to compete with pathogenic bacteria within the host potentially via production of antimicrobials e.g., extracellular reuterin (Cleusix et al., 2007; Spinler et al., 2008). Due to limited glycerol availability, suboptimal endogenous concentrations of glycerol in the GI tract would likely limit adequate reuterin production. In order to obviate the need to provide high levels of glycerol to satisfy L. reuteri's optimal needs, we provided targeted delivery of glycerol directly to the bacteria attached to the surface of DMs. To test this in vitro, we utilized a colorimetric assay for reuterin production (Cadieux et al., 2008). As shown in Figure S5, DMs filled with glycerol concentrations ranging from 10 to 80% were able to induce reuterin production. Compared to the 2% glycerol solution control, DMs filled with 80% glycerol produced on average 53% more reuterin in 1 h (average concentration of reuterin produced: 2% glycerol = 40 mM, DM-80% glycerol = 61 mM). To determine if the 80% glycerol or the resulting reuterin/downstream metabolites of glycerol fermentation produced by L. reuteri is toxic to L. reuteri, we compared hourly colony forming units (CFU) of L. reuteri incubated with either DM-water or DM-80% glycerol, in either sterile saline or MRS growth medium. As shown in Figure S6, there was no loss of CFU regardless of DM cargo when L. reuteri was incubated in MRS. Incubating L. reuteri in saline did result in a steady loss of viable CFU over time, though there was no difference in viability between the DM-water and DM-80% glycerol over this time, suggesting the loss of CFU was not due to any potentially toxic compounds, such as reuterin or acrolein, from glycerol fermentation (Figure S6). As acrolein in particular is known to be toxic to humans and is a byproduct of reuterin production, we next calculated the maximum possible amount of acrolein that could be produced from the dosage of L. reuteri and volume of glycerol provided via DMs in our formulation, assuming all available glycerol was converted 1:1 into acrolein. As shown in Figure S7, the amount of acrolein that could possibly be produced via our formulation is a nominal ~6 μg (for reference, the World Health Organization recommends less than 7.5 μg/kg body weight per day) (Gomes et al., 2002). From these results and the data presented in Figure 4, we hypothesized that DMs loaded with glycerol would have two beneficial effects in vivo, namely slowing the release of beneficial cargo and providing a substrate for reuterin production.

Histamine produced by L. reuteri has previously been shown to inhibit pro-inflammatory cytokines such as TNF via H₂ receptors and reduce colitis in an animal model (Thomas et al., 2012; Gao et al., 2015). Our microsphere-based approach provides a unique method for delivery of the histamine precursor substrate L-histidine to L. reuteri. To test this in vitro, we filled DMs with 30 mg/ml and 4 mg/ml L-histidine and measured the amount of histamine produced by the bacteria when the only source of L-histidine was via diffusion out of the DMs. As shown in Figure 5, DM- L-histidine (4 mg/ml) resulted in histamine levels only slightly lower than those produced when bacteria were incubated in 4 mg/ml L-histidine solution without DMs. When the DMs were loaded with a higher concentration of L-histidine, the amount of histamine produced was 6–7 times greater than the lower 4 mg/ml concentration, consistent with the DM-L-histidine (30 mg/ml) providing ~7 times more L-histidine than the DM-L-histidine (4 mg/ml) (Figure 5). In addition, we tested whether other cargo relevant DM cargo substrates, such as maltose and glycerol, would negatively affect histamine production. Addition of glycerol did not result in reduced histamine production, regardless of whether the L-histidine was in solution or provided via DMs (Figure 5). With addition of maltose, histamine production actually increased when L-histidine was provided in solution, but statistically unchanged when L-histidine was provided via DMs (Figure 5).

Orally consumed probiotics face a significant pH challenge upon reaching the stomach, where pH values are as low as 1.5 when the stomach is empty (Dressman et al., 1990). Enhancing the ability to deliver a maximal number of viable L. reuteri to the colon is crucial to its sustainability and effectiveness as a probiotic. We thereby hypothesized that L. reuteri bound to the surface of DMs in the form of a biofilm would increase survival upon exposure to acid, and that DMs filled with sucrose or maltose would result in even greater survival in a GTFW-dependent manner. As shown in Figure 6, less than 0.1% of WT L. reuteri without DMs survived in synthetic gastric acid after 4 h at pH 2, which resulted in a nearly 3 log loss of viable probiotic. Addition of water-filled DMs did not significantly alter the survival rate of WT L. reuteri in gastric acid; however, when either DM-sucrose or DM-maltose was delivered with WT, nearly 1 log more survived the acid stress (Figure 6). To show that the protective effect is dependent on the microspheres and not the cargo within the DM lumen, we also incubated L. reuteri with the equivalent amount of diffusible cargo without the DMs. Acid survival in the presence of cargo only was no different than L. reuteri alone (Figure 6), which strongly indicated the importance of the bacterial biofilm-on-DM delivery system for the observed protective effect.

To investigate whether this phenotype is GTFW-dependent, we also tested synthetic gastric acid survival using the ΔgtfW strain of L. reuteri and found that the beneficial effect of DM-sucrose and DM-maltose was lost (Figure 6). Interestingly, the mutant also showed deficiency in acid survival without DMs compared to WT, which indicated that GTFW's role in cellular aggregation and biofilm formation (Figure S4) may contribute significantly to survival in synthetic gastric acid.

Next, we examined what effect the DMs, the DM luminal cargo and the product of the gtfW gene have on the relative adherence of L. reuteri when delivered as planktonic cells or as biofilms on DMs to the human intestinal cell lines DLD-1 (adult human colonic epithelial cells) and FHs 74 Int (3–4 months gestation, small intestine epithelial cells) in vitro. As shown in Figure 7A, after a 1 h incubation on DLD-1 cells, significantly more WT L. reuteri (without DMs) adhered to the colonic cells compared to ΔgtfW either with or without DMs, which indicated that GTFW contributed to host cell adherence. When L. reuteri adhered to DMs that contained sucrose or maltose were added to colonic cells, relative adherence of WT L. reuteri to the colonic cells was increased by 4.7-fold for DMs that contained sucrose and by 5.2-fold for DMs that contained maltose (Figure 7A). Although overall fewer WT L. reuteri adhered to the FHs 74 cells than to DLD-1 cells, delivering the bacteria with either DM-sucrose or DM-maltose resulted in 1.8-fold (DM-sucrose) or 2.7-fold (DM-maltose) more adhered bacteria compared to WT bacteria without DM (Figure 7B).

To further show that DM luminal cargo of maltose and sucrose improved relative adherence of L. reuteri to epithelial cells in vitro, we analyzed WT and ΔgtfW L. reuteri adherence after 1 h incubation on DLD-1 cells visually, using CSLM (Figure 8). As with the CFU data presented in Figure 7, delivery of WT L. reuteri as a biofilm on maltose or sucrose-loaded DMs supported greater adherence to the DLD-1 cells than those delivered on water-loaded DMs or with no DMs, both by visual inspection (Figure 8A) and when analyzed by quantification of bacterial biomass using COMSTAT analysis of CSLM images (Figure 8B). The observed adherence was significantly diminished in the ΔgtfW mutant compared to the wild type, consistent with measured CFUs (Figure 7A).

Finally, we tested the effect of DM adhered WT L. reuteri's ability to bind to mucin. While cellular binding of probiotics likely plays a role in colonization, a healthy GI tract has a mucus layer on the apical surface of epithelial cells, of which the primary constituent is mucin (Turner, 2009). Indeed it is believed that healthy commensals are found primarily within this layer so it is imperative that our formulation maintains it enhanced probiotic effects in the presence of mucin. As mucin adherence is not GTF-dependent, but rather controlled by specific mucin-binding proteins (Miyoshi et al., 2006; Lukic et al., 2012), we hypothesized that being bound to DMs would not have an effect on the ability of L. reuteri to adhere to mucin. As shown in Figure S8, there is no significant difference in relative adherence of WT L. reuteri to mucin when delivered as either a planktonic bacterial suspension or as a biofilm adhered to DMs after a 60 min incubation on mucin agar plates.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.