Core-shell commensal biocapsules for in situ gut microbiome engineering

Abstract

Despite the profound influence that gut microbiome composition has on human health, the development of live microbe treatments is constrained by a lack of knowledge on single-species functions within the gastrointestinal (GI) tract. A key barrier is the absence of broadly accessible tools capable of overcoming colonization resistance and enabling spatiotemporal control of microbiome ecology. To address this gap, we have developed a core-shell capsular material, termed a biocapsule, designed to promote the engraftment of a defined bacterial payload to a modified GI niche, thereby enabling precision engineering of commensal populations in situ. To achieve this, biocapsules employ a sequential kill-and-replace strategy in which local native flora are transiently cleared by the capsule before a delivered commensal consortium is introduced to occupy the vacated niche. This targeted antagonism approach is a unique departure from traditional methods that utilize broad-spectrum antibiotics or heterogeneous stool microflora to alter microbial populations, neither of which provide the fine control needed to carefully shape the composition of an established community. Consequently, biocapsules offer a self-assembling, biocompatible platform capable of reshaping microbiome composition in situ to advance translational opportunities in materials-enabled commensal engineering.

Graphical abstract

Highlights

• •Core–shell biocapsules enable targeted, antibiotic-free microbiome engineering.
• •Capsules protect probiotics through GI transit and release payloads in the colon.
• •Capsule coating transiently clears local flora to promote commensal engraftment.
• •Biocapsules outperform free probiotics in pathogen clearance and niche replacement.
• •Commensal-loaded capsules reshape gut ecology in vivo, enriching beneficial taxa.

1. Introduction

The human gastrointestinal (GI) tract hosts an ecologically diverse bacterial community which forms a commensal consortium alongside resident archaea, fungi, and viruses. The ∼10¹³ bacterial residents of the colonic microbiome occupy a variety of niches and collectively influence host digestion, immune training, and defense against opportunistic pathogens [1]. Gut microbial ecology also impacts the functions of systems outside the GI tract, including the brain where neurologic processes are altered via communication through the ‘gut-brain axis’ [2,3]. Yet, despite its link to human health, progress in understanding how microbiome ecology influences human physiology and disease has been slowed by a lack of facile and accessible tools that can be used to manipulate microbial composition in situ [4].

The field of synthetic biology has offered a possible solution. ‘Smart’ microbes, genetically engineered to colonize specific bowel locations and dynamically respond to their environment, have the potential to transform the way we detect, diagnose, and treat GI disease [5]. However, current formulation approaches for probiotics or live biotherapeutics are heavily reliant on lyophilization, which may sacrifice cell viability [6,7], and technologies that successfully deliver microbial products to specific GI regions are lacking. Without a delivery vehicle, unprotected commensals are either destroyed by the astringent gastric environment, cleared by host immune cells and resident flora, or excreted in the continuous GI flow, ultimately leading to transient health benefits. Consequently, recent efforts have focused on packaging microbiota into orally administered enteric-coated capsules, dry polymeric matrices or micro-emulsions [8]. These technologies significantly improve the viability and mucoadhesion of delivered microbes during GI transit, while addressing palatability concerns to improve patient compliance. Indeed, formulating microbiota into gelatine capsules coated with pH-sensitive polymers has become the clinical standard for oral transplants. However, despite progress made by these advanced formulation strategies, niche competition and host rejection remain key barriers to translation of oral microbiota therapies. Consequently, new materials-enabled approaches are needed to realize the full therapeutic potential of probiotic interventions for microbiome-related GI disorders.

Inspired by this challenge, here we report dual-compartment, core-shell commensal biocapsules that replace native gut flora and/or pathogens with defined probiotic payloads, thus allowing precision engineering of gut ecology without the need for antibiotics or stool donors (Fig. 1a). This materials-enabled approach leverages well-established principles of invasion biology to engineer microbiome ecology in situ via antagonistic interactions. Our findings demonstrate that biocapsules improve commensal colonization efficiency in vitro through a sequential kill and replace strategy that diverges from current methods focused on supplanting established flora. In vivo results further demonstrate that biocapsules exert a ‘soft kill’ that transiently reduces commensal levels to promote engraftment of the delivered probiotic cargo. By utilizing inexpensive, biocompatible, and biodegradable ingredients, these materials can be broadly adopted into a variety of research and translational settings.

Fig. 1.

Biocapsule design and characterization. a, Biocapsules are prepared with a commensal-loaded liquid core (brown) spatially segregated from an antimicrobial surface coating (blue). Contact of gut flora and pathobionts (red) with the biocapsule surface leads to rapid bacteriolytic death and triggers the release of the commensal payload to the lower GI tract to modify microbiome ecology. b, Methodologic schematic of biocapsule fabrication. HA leakage from the alginate capsule (middle) leads to its electrostatic surface assembly with environmental polycationic polymers to create the antimicrobial coating. c, Minimum inhibitory concentration (MIC) of screened cationic polypeptides and polyethyleneimine (PEI) against the indicated gut pathogen. Polypeptides were tested as a function of molecular weight: Lysine_LMW = 6.3 kDa; Lysine_HMW = 150 – 300 kDa; Ornithine_LMW = 5.6 kDa; Ornithine_HMW = 78 kDa; Arginine_LMW = 6.2 kDa; Arginine_HMW = 83 kDa. PEI at MW = 25 kDa was tested as a branched or linear arrangement. d,e, Representative dark field micrographs of uncoated (d) or PLL coated (e) core-shell capsules. Yellow arrows highlight PLL surface coating. Insets: Brightfield particle images. Scale bars = 500 μm. f – k, Representative false colored scanning electron micrographs of alginate capsule surface (f), interface between capsule core and alginate shell (g), topographic composition of alginate shell (h) and core (i), interface between PLL coating (blue) and alginate shell (brown) (j), and magnified view of PLL coating from region of interest identified by the dashed box in panel j (k). Insets: Schematic showing the region of the particle imaged. Scale bars = 5, 25, 20, 10, 200 and 30 μm for panels f – k, respectively.

2. Results

2.1. Fabrication of core-shell, probiotic-loaded biocapsules

The design of commensal biocapsules began by selecting biocompatible and generally recognized as safe (GRAS) building blocks. The core-shell capsule is prepared by dispersing commensals into a solution of calcium chloride and hyaluronic acid (HA) (Fig. 1b). This solution is then drop-cast into an alginate bath to rapidly induce alginate-calcium crosslinking [[9], [10], [11], [12], [13]], and form a hydrogel capsule at the surface of the droplet. HA is included as both a non-immunogenic and non-toxic polysaccharide nutrient source for the loaded commensals [14,15], and is a prerequisite to form the antimicrobial surface coating. As the capsule shell continues to crosslink, diffusion of alginate into the interior of the droplet is inhibited, ultimately leading to a liquid core that serves as a habitat for the microbial cargo and a barrier between the payload and antimicrobial surface coating.

To coat the particles, capsules are transferred to a calcium-containing solution of a polycationic antibacterial polymer. HA, which does not participate in Ca²⁺ cross-linking, diffuses out of the liquid core and through the alginate shell to electrostatically assemble with the polymer and form a supramolecular coating at the capsule surface (see Fig. 1b). Selection of an effective bactericidal coating was determined by screening a series of polycationic peptides and polymers against four organisms of interest: Escherichia coli, Shigella flexneri, Salmonella enterica, and Staphylococcus aureus (Fig. 1c). Results identified high molecular weight polylysine (150 – 300 kDa; hereafter referred to as PLL) as a lead compound with activity against all four of the tested bacteria. Conversely, the HA coating component did not appreciably alter the growth kinetics of these gut pathobionts (Supplementary Fig. 1) and therefore serves as a biologically benign co-assembling formulant.

Next, using our capsule fabrication method, we prepared PLL coated capsules and performed optical and electron microscopy to visualize the material architecture (Fig. 1d–k). Cryo-scanning electron microscopy revealed that the alginate capsule has a porous outer shell and a liquid interior core suitable for the loading and retention of probiotic microflora (Fig. 1f–i) with an antimicrobial surface coating that is spatially segregated to avoid harming the bacterial payload (Fig. 1j and k). Importantly, the porous nature of the alginate capsule shell suggests that nutrients and metabolic waste products can exchange across this interface to support the growth of the internal commensal payload, which was further corroborated through dye diffusion experiments (Supplementary Fig. 2).

Using multiparametric experimental design, we next determined the fabrication conditions that yield capsules optimized for both structural stability and loading viability of three model probiotic strains, namely Lacticaseibacillus rhamnosus, Lactiplantibacillus acidophilus and Bifidobacterium longum. To do this, we tested the effects of component concentration on both stability of the core-shell capsule (Fig. 2a) and surface coating integrity (Fig. 2b) 1 h after fabrication. Here, we use the classifiers stable, transitional, or unstable, with the term transitional used to define combinations that yield quasi-stable materials that were ultimately too weak to be manually manipulated. These results were then collated with 24-h growth kinetics of each model commensal strain in the corresponding reagent solutions, with color-coded classifiers of growth, stasis, and death determined from proliferation curves shown in Supplementary Figs. 3–8. As demonstrated in Fig. 2a, the region of 300-800 mM calcium chloride and 2-8 mg/mL alginate was identified as suitable to produce both stable capsules and maintain viability of all three loaded microbial strains. Calcium chloride concentrations >800 mM caused alginate precipitation, which prevented capsule formation, and was toxic to both Lactobacilli cultures. Conversely, B. longum growth was uninhibited at all tested reagent concentrations. Next, coating studies showed that HA and PLL concentrations ≤1.0 mg/mL and ≥0.13 mg/mL, respectively, were optimal for producing competent capsule coatings without inhibiting commensal growth (Fig. 2b). Unexpectedly, we found that, although B. longum was sensitive to PLL, both L. rhamnosus and L. acidophilus generally tolerated exposure up to 0.5 mg/mL. This may reflect an intrinsic resistance of Lactobacillus spp. to PLL given that some strains produce these polymers as a secondary metabolite [16]. Conversely, we found that the tested Lactobacilli were inhibited by exposure to ≥2.0 mg/mL of HA, which does not appear to be a sensitivity reported in the literature. These multiparametric studies ultimately identified optimal reagent concentrations that can generate stable biocapsule materials while maintaining viability of the encapsulated commensal cargo (see dashed boxes in Fig. 2a and b). Finally, in addition to examining bacterial growth in the presence of biocapsule components, we screened pectin and glucose as two nutrient sources for possible addition into the core to further support commensal viability (Supplementary Figs. 4, 6, and 8). These screens showed that pectin supported robust growth of all three strains tested at concentrations up to 2.0 mg/mL and was therefore chosen for incorporation within the final formulation.

Fig. 2.

Multiparametric optimization of capsule design. a,b, Concentration of the indicated inner capsule (a) and surface coating (b) components that yield stable (black outer circle), quasi-stable/transitional (grey outer circle), or unstable (white outer circle) material compartments. Inner circle color represents conditional growth behavior of the indicated commensal from optical density assays (see Supplementary Figs. 3–8). Red = growth, pink = stasis, white = death. c – e, CFU/mL of L. rhamnosus (c), L. acidophilus (d) or B. longum (e) cultures encapsulated within uncoated (blue) or PLL-coated capsules (red) over a 48-h growth period. Data presented as mean ± s.d. from n = 3 biologic replicates. f, Representative image of B. longum loaded biocapsules following a 24-h incubation in culture media. Dark interior demonstrates colonization of the core volume by loaded bacteria. Scale bar = 3 mm. g, Representative false colored scanning electron micrograph of the interior of L. rhamnosus loaded biocapsules, showing EPS of bacteria (pink) engaged with the alginate/HA capsule constituents (brown). Scale bar = 5 μm.

Next, we evaluated the growth of each model test organism encapsulated within the complete PLL-coated and pectin-loaded biocapsules over 48 h (Fig. 2c–e). In general, we observed that encapsulated bacterial growth was markedly different than the response to the individual material components. For example, although L. rhamnosus was the most sensitive of the tested strains to HA (Fig. 2a and b and Supplementary Fig. 3), it grew rapidly within the complete capsule and achieved a core density of 10¹¹ CFU/mL 24 h after loading (Fig. 2c). Conversely, L. acidophilus cultures were static in the capsule over 48 h (Fig. 2d), while B. longum grew within the capsule interior over the first 24 h before declining in viability between 24 and 48 h. These growth profiles were unaffected by coating status, recapitulating the general insensitivity of all three strains towards PLL observed in Fig. 2b. Based on these findings, L. rhamnosus was prioritized in preparing commensal-loaded biocapsules for further study. Additional imaging studies demonstrated that, 24 h after encapsulation, loaded commensals grew to fill the entire interior volume of the capsule (Fig. 2f) and formed fibrous assemblies that may be exopolysaccharide (EPS) biofilm structures (Fig. 2g and Supplementary Fig. 9). Further, the biocapsule fabrication process allows us to compress a dense suspension of bacteria into a small volume in the core, and the porous nature of the hydrogel shell enables the exchange of both nutrients and toxic metabolic waste, creating a dynamic growth environment suitable for maintaining a high-density bacterial community.

2.2. Biocapsule antimicrobial activity and specificity

In parallel with the formulation work described above, we assessed the activity of sterile capsules (e.g., no loaded bacteria) against four model pathogens, Staphylococcus aureus, Escherichia coli, Salmonella enterica, and Shigella flexneri. Here, we measured viability of the pathogens 24 h after treatment with either uncoated or PLL-coated capsules; initial inoculation densities of 10² – 10⁷ CFU/mL were employed to model the typical concentration of bacteria detected in clinical gastroenteritis stool samples [17]. Results in Fig. 3a confirm that, as expected, uncoated capsules are inactive towards all four of the pathogens at the tested inoculation densities. Conversely, capsules coated with PLL demonstrate significant inhibition of S. aureus and S. flexneri cultures at 10² – 10⁷ CFU/mL. E. coli was inhibited until 10⁵ CFU/mL, at which point the capsule activity was overwhelmed by increasing bacterial inoculum. Finally, PLL-coated capsules inhibited S. enterica growth at a 10² CFU/mL inoculum and experienced decreasing effects as inoculation concentration increased.

Fig. 3.

Biocapsule antibacterial activity and specificity. a, Antimicrobial activity of uncoated (blue) and PLL-coated (green) capsules against four common Gram-positive and Gram-negative GI pathogens, including Staphylococcus aureus (S. aureus USA300), Escherichia coli (E. coli 101-1), Salmonella enterica (S. enterica ATCC 14028), and Shigella flexneri (S. flexneri 2a 457T). Microbial viability was determined via optical density (OD₆₀₀) measurements following a 24-h co-incubation of each microbe with the indicated capsule formulation at varying bacterial seeding densities (CFU/mL). Statistical significance between uncoated and PLL coated capsules at each condition was determined using Student's t-test with equal variance. For visual clarity, only non-significant (n.s.) comparisons are shown, all others are defined by p ≤ 0.05. b, Difference between zone of inhibition diameter and biocapsule diameter for uncoated (Unc., blue) and PLL-coated (green) sterile biocapsules on agar lawns of the indicated pathogen. c, Time-dependent growth curves of commensal E. coli (MG1655) and Clostridium scindens (DSM 6597) strains in the absence (untreated, red) and presence of uncoated (blue) and PLL-coated (green) capsules. Data in panels a - c presented as mean ± s.e.m. from n ≥ 3 biologic replicates. d, Relative abundance of the 10 most represented commensals in a human fecal microbiome community before (stock, t = 0) and after a 1 – 4 day exposure to BHI CHVR media in the absence (untreated) or presence of uncoated and PLL-coated capsules. Data presented as the mean of n ≥ 2 biological replicates. e, Relative viability of Caco-2 human colorectal epithelial cells following a 24-h incubation with uncoated (Unc., grey) or PLL-coated (black) capsules. Data presented as mean ± s.e.m. from n = 6 biologic replicates. Statistical significance determined using Student's t-test with equal variance; n.s. = non-significant, ∗p < 0.05, ∗∗∗p < 0.001.

Next, we performed zone of inhibition (ZOI) studies to evaluate the contact-dependence of PLL-biocapsule antimicrobial activity (Fig. 3b–c, Supplementary Fig. 10). This study was modeled after disk diffusion assays, with sterile capsules pinned to a bacterial lawn on an agar substrate, followed by visual evaluation of clearance distance from the capsule surface after 24 h of incubation. Similar to the results from our liquid assays, S. flexneri was inhibited by PLL capsules more significantly than uncoated capsules, with PLL formulations reaching a ZOI of 0.3 ± 0.1 cm, while S. aureus, E. coli, and S. enterica displayed no significant difference in sensitivity between uncoated and PLL-coated capsules. Interestingly, while uncoated capsules were inactive towards S. flexneri, they showed weak inhibition of S. aureus and E. coli (ZOI of 0.2 – 0.4 cm), and moderate inhibition of S enterica (ZOI = 1.4 cm). This suggests possible co-inhibitory effects of the biocapsule components such as alginate, HA, and/or calcium towards these specific pathogens. Taken together with our liquid broth assays, these results indicate that the material components may slowly diffuse away from the capsule surface, exerting distal antibacterial activity that does not explicitly require direct contact of certain pathogens with the material surface. Yet, ZOIs for most of the pathogens are relatively small (≤1.5 cm), suggesting that antimicrobial effects of the material remain generally localized around the capsule. This, in turn, may enable the material to clear a local niche in the GI tract that allows for targeted engraftment of the delivered commensal.

To evaluate the influence of capsules on GI microbiome ecology, we next performed a series of growth inhibition and sequencing assays using both mono- and polymicrobial probiotic cultures, beginning with the two model commensal organisms E. coli (MG1655) and Clostridium scindens (DSM6597) (Fig. 3c). Interestingly, although PLL-coated capsules inhibited the growth of pathogenic E. coli (Fig. 3a), they were inactive towards the commensal strain (MG1655). Similarly, PLL capsules did not negatively impact the growth kinetics of C. scindens, relative to untreated controls. Together, this suggests that the biocapsules may exert taxa-specific, and potentially even strain-specific, antimicrobial activity. Subsequently, we assessed capsule-mediated changes in a polymicrobial fecal microbiome community obtained from a healthy human donor. Here, we monitored shifts in the relative abundance of >2000 bacterial species over a four-day incubation with either uncoated or PLL-coated capsules (Fig. 3d). Compiling data for the 10 most abundant species showed few differences in the composition of capsule-treated communities relative to culture controls. Taken in context with results from our antibacterial assays (Fig. 3a and b), these results suggest capsules exert localized and taxa-specific clearance of gut microbes in a spatially confined, contact-dependent fashion, thereby avoiding the systemic and harmful perturbation of GI commensal ecology associated with use of broad-spectrum antibiotics.

Lastly, we examined the in vitro toxicity of the capsules toward undifferentiated Caco-2 human colorectal epithelial cells after a 24-h incubation (Fig. 3e). Neither formulation yielded a statistically significant change in Caco-2 viability, suggesting that orally administered biocapsules are unlikely to harm the host GI epithelium during transit.

2.3. Capsule GI stability and commensal delivery

We next investigated the region-specific viability and release of the L. rhamnosus biocapsule payload in varying GI environments (Fig. 4). These in vitro experiments utilize simulated gastric (Fig. 4a), small intestinal (Fig. 4b), and colonic (Fig. 4c) fluids, and incubation periods relevant to the physiological transit time of solid materials through these individual tracts in humans (i.e., 3 h for gastric, 4 h for small intestine, and 36 h in the colon). Using optical density and plating assays, we differentiated the density of bacteria retained in the core or released to the supernatant to assess the time-dependent delivery of the payload. Results show that capsules are generally stable in the gastric (Fig. 4a) and small intestinal (Fig. 4b) fluids, demonstrating maintenance of encapsulated L. rhamnosus viability relative to the initial time point control (t = 0), with minimal extracapsular delivery. In contrast, capsules erode in the simulated colonic environment to release approximately 9.9 × 10⁸ CFU of loaded bacteria over the 36-h transit period. This is complemented by a gradual reduction of bacterial density within the capsule core as the payload is delivered. Taken together, this suggests that biocapsules maintain structural integrity and preserve encapsulated L. rhamnosus viability during upper GI transit, thereby enabling preferential delivery of the commensal payload to the colon as intended. Moreover, given that these static assays do not mimic the influence of mechanical agitation and interaction with host and commensal gut flora in the true colonic environment, we expect even greater release of biocapsule contents to the large intestine in vivo.

Fig. 4.

Simulated GI payload stability and release. a-c, Viability of encapsulated L. rhamnosus in the biocapsule core (black) or released to the supernatant (grey) when incubated in simulated gastric (a), small intestinal (b), and colonic (c) fluids. Data shown as stacked bar plots. Top: Volume normalized L. rhamnosus optical density (OD₆₀₀ x mL) at physiologically relevant GI transit time points. Bottom: Total colony forming units (CFU) of initial and endpoint plated L. rhamnosus. Statistical significance of n ≥ 3 technical replicates determined relative to initial time point (t = 0) using Student's t-test with equal variance; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. For visual clarity, p value indicators are shown only for statistically significant comparisons (p ≤ 0.050) and are differentiated by color for the core (white) or released (black) compartments.

2.4. Biocapsule microbial clearance and replacement

Engineering gut microecology in a targeted in situ niche requires clearing local flora and replacing it with the delivered commensal. Here, we perform a series of studies to assess the ability of biocapsules to clear the model gut pathobiont S. flexneri and replace it with an L. rhamnosus payload (Fig. 5a). During these experiments, we observed that contact of external S. flexneri with the capsule surface led to gradual erosion of the material and subsequent release of the commensal cargo (Fig. 5b). Differential plating of the supernatant solution was then used to track the extracapsular density of each microbe, with MHB2 containing 100 μg/mL ampicillin used to selectively grow S. flexneri and MRS with 5 μg/mL vancomycin used for L. rhamnosus.

Fig. 5.

Biocapsule in vitro clearance and replacement. a, Schematic demonstrating triggered release of biocapsule contents following erosion of the material via surface contact of external flora (red) with the antimicrobial coating. b, Optical images of L. rhamnosus loaded biocapsules incubated in a 10⁵ CFU/mL solution of S. flexneri at 0 (top) and 3 (bottom) days. Magnified region of interest indicated by the black dashed line demonstrating release of the L. rhamnosus biocapsule payload. The inner capsule core is demarcated by the white dashed line. Scale bars = 2.5 mm. c-e, Time-dependent density of L. rhamnosus (green) or S. flexneri (red) in the extracapsular supernatant as individual untreated planktonic cultures (c, control), S. flexneri solutions treated in the presence of sterile (unloaded) capsules (d), or L. rhamnosus loaded biocapsules (e). Planktonic carrying capacity of each microbe in the broth solution was determined from panel c and is represented in panels d and e by the appropriately colored dashed line. Limit of detection (L.O.D.) is indicated by the dashed black line. Data presented as mean ± s.d. from n = 3 biologic replicates.

We began with control cultures grown separately in the broth solution to establish a carrying capacity (Fig. 5c). Next, we assessed the S. flexneri inhibition efficacy of empty (sterile) PLL biocapsules, which demonstrated a ∼1 log reduction in the pathobiont's stationary phase over the 3-day incubation period (Fig. 5d). We then compared the growth of S. flexneri when in co-culture with L. rhamnosus (Supplementary Fig. 10) to its growth when exposed to L. rhamnosus biocapsules (Fig. 5e). We observed that the L. rhamnosus loaded biocapsules fully inhibited S. flexneri within 2 days after biocapsule addition and supported an extracapsular L. rhamnosus concentration of 1.3 × 10⁹ CFU/mL – a density greater than the ∼10⁸ CFU/mL carrying capacity achieved in planktonic controls. Additional experiments showed that the efficiency of S. flexneri clearance by the empty PLL capsules is dependent on the inoculum density, with concentrations of S. flexneri >10⁷ CFU/mL leading to a reduction in the antimicrobial potency of the biocapsule's PLL surface (Supplementary Fig. 12). It is important to emphasize that these clearance assays were conducted in aerobic conditions and broth compositions optimized for S. flexneri growth, which is not representative of the highly anaerobic in vivo GI. Nevertheless, treatment of S. flexneri solutions with a single commensal-loaded biocapsule led to clearance of the pathobiont density below the limit of detection within 48 h of exposure, and its subsequent replacement with a high density of L. rhamnosus (∼10⁹ CFU/mL). These proof-of-concept studies suggest that biocapsules can clear a specific niche within the GI flora and substitute it with defined commensal payloads. This motivated us to further investigate the in vivo engraftment performance of this technology in C57BL/6 mice with an established gut microbiome.

2.5. In vivo engraftment

Mice were randomly assigned to receive an oral gavage of either a liquid bacterial stock (free bacterial control) or a single commensal-loaded biocapsule, prepared with equivalent amounts of bacteria (Fig. 6a). A separate group received sterile capsules that did not contain a commensal payload as a material control. For these studies, capsules were miniaturized from an original diameter of 5 mm (Fig. 2f) to ∼0.7 mm to accommodate oral gavage for mice. Growth experiments confirmed a similar volume-adjusted carrying capacity of L. rhamnosus within the miniature biocapsules (7.8 × 10⁶ ± 5.3 × 10⁶ CFU) relative to the larger formulations used in prior studies. Time-dependent engraftment of delivered L. rhamnosus was subsequently monitored through both fecal sampling and endpoint analysis of GI contents using selective plating on vancomycin-containing MRS agar. Fecal sampling demonstrated that biocapsules delayed the GI excretion of delivered L. rhamnosus by approximately 3 h relative to the free bacterial gavage (Fig. 6b), with clearance of the commensal dropping below the limit of detection at 24 h. These excretion kinetics match prior studies in BALB/C mice following oral gavage with L. rhamnosus GG [18]. Increasing the dose to 15 capsules per animal resulted in similar excretion kinetics and GI engraftment (Supplementary Fig. 13). Conversely, co-delivery of free L. rhamnosus and a solution of PLL, thus spatially decoupling the probiotic cargo from the antimicrobial agent, led to GI transit profiles that matched the free bacterial bolus (Supplementary Fig. 14). This suggests that the ability of biocapsules to co-localize commensal payload release with PLL-mediated microbial clearance plays a causative role in the material's ability to delay L. rhamnosus GI excretion. Parallel toxicology assays found that dosing mice with PLL, either as the free agent given in combination with bacteria, or as part of the biocapsule formulation, did not produce statistically significant changes in most hepatic functional markers (Supplementary Fig. 15). The notable exception was a decrease in cholesterol levels for animals receiving free PLL. However, a similar effect was observed for mice dosed with the bacteria alone, and so is likely a consequence of altered food intake as a result of the gavage procedure. Finally, to assess the functional consequences of biocapsule-mediated delays in L. rhamnosus excretion, we measured the time-dependent change in specific short chain fatty acids (SCFAs) in the feces (Supplementary Fig. 16). No statistically significant differences were observed across all tested time points between biocapsule treated animals and controls, suggesting L. rhamnosus residence time in the gut was insufficient to exert functional changes in metabolite production.

Fig. 6.

Biocapsule-mediated modulation of gut commensal ecology. a, In vivo experimental design schematic outlining treatment groups and fecal sampling frequency. b, Time-dependent L. rhamnosus fecal load (in CFU/g) from C57BL/6J mice treated with free L. rhamnosus (red), L. rhamnosus-loaded biocapsule (green), or sterile capsule control (blue). c-e, Treatment-dependent changes in Lipocalin 2 (Lcn2) concentration in the feces (c) and the proximal (d) and distal (e) colonic tissues. Inset in panel c shows average Lcn2 concentration at 6 h post-administration for each color-differentiated treatment group. Statistical significance determined using Student's t-test, with n.s. = non-significant, ∗p < 0.05, ∗∗p < 0.01. Comparisons in panels d and e are made relative to untreated control (grey), with lines reflecting a similar statistical outcome for all treatment groups. f, Venn diagram of ASVs (amplicon sequence variants) detected in the input inoculum vs. hFMT recipient fecal contents. A small number of ASVs (7) overlapped between input and hFMT, while the majority are unique to each set. g, Time-dependent fecal alpha diversity of the indicated treatment group, represented as Faith's phylogenetic diversity (Faiths_PD), observed ASVs, and Shannon diversity. h, Taxonomic abundance of specific human fecal microbiota transplant (hFMT)-derived strains over time in each treatment group. i,j, Volcano plots showing differential abundance of microbial taxa determined by ALDEx2 in the feces at 48 (i) and 168 (j) hours post-administration of hFMT vs. hFMTCap. Data presented in panels b – e, g and h as mean ± s.d. from n = 4 biological replicates. Statistical significance in c-e performed by Welch's t-test (FDR corrected).

Although we did not achieve long-term L. rhamnosus engraftment from the biocapsule delivery vehicle, it is important to note that the density of bacteria administered to animals in our study is ∼2 log lower than the standard in the field. This was necessitated by limitations in the core loading volume of the capsules, which had to be miniaturized for in vivo experiments to enable oral gavage to mice. Larger capsules, which would be suitable for human dosing, can encapsulate up to 10¹¹ CFU/mL of a commensal payload (Fig. 2c). Nevertheless, the ability of biocapsules to delay excretion of the delivered commensal is notable and may be partly due to the vehicle's ability to spatially sequester loaded probiotics from host immune cells to limit payload rejection. This is further supported by follow-up ELISA assays measuring the concentration of the innate inflammatory marker lipocalin-2 (Lcn2) in feces (Fig. 6c), as well as proximal (Fig. 6d) and distal colon (Fig. 6e) tissues. Fecal contents show that mice treated with free L. rhamnosus experience a large spike in Lcn2 expression 3 – 6 h post administration (Fig. 6c), matching the bacterial GI transit kinetics (Fig. 6b). This inflammatory response was blunted when delivering the commensal via the biocapsule, with fecal Lcn2 levels below the assay detection limit 6 h after capsule administration. While comparisons between the groups were not statistically significant due to high data variance, the results suggest a possible protective mechanism of the biocapsules that may enhance the viability of loaded commensals during GI transit. This is corroborated by Lcn2 levels measured in the proximal colon tissue for capsule-treated animals (Fig. 6d) which, despite a small and transient increase at 1 h post administration, were not significantly elevated relative to untreated controls at 7 days. Conversely, treating mice with free L. rhamnosus yielded a significant increase in proximal colon Lcn2 levels at the same detection time point. Distal colon tissue did not show a statistically significant change in Lcn2 levels across all tested treatments and time points relative to controls (Fig. 6e).

Next, we sought to test the efficacy of our biocapsules when delivering a diverse consortium of commensals to conventional mice rather than a single species. We therefore repeated our in vivo studies using a human fecal microbiome transplant (hFMT) matched in overall microbial density between the free hFMT and hFMT-loaded biocapsule groups, the latter hereafter referred to as hFMTCap. Sterile capsules were used as a biologic control to investigate the effects of capsule material on gut flora composition. Initial high-throughput fecal sequencing demonstrated only a small fraction of amplicon sequence variants (ASVs) from the input fecal inoculum could be detected in hFMT-treated mice, with an overlap of 0.9% ASVs between input and recipient (Fig. 6f). This highlights the colonization resistance of the gut microbiota without use of a conventional broad-spectrum antibiotic, or alternatively the non-antibiotic approach described here. To assess the engraftment efficiency of hFMT delivered by biocapsules we analyzed the gut microbial diversity across treatment groups, as well as microbial diversity (alpha diversity) within each group (Fig. 6g). While mice receiving hFMT alone exhibited moderate reductions in alpha diversity, as measured by multiple indices within the first 24 h of treatment, the hFMTCap group showed the most pronounced shift in microbial diversity. In the sterile capsule control group, a minor and transient decrease in diversity was measured over the initial 24-h period, likely as the result of the antimicrobial effects of PLL. Further analysis of strain-specific bacterial colonization showed that hFMTCap treatment transiently generated a favorable environment for Romboutsia ilealis (Fig. 6h) which is implicated in gut carbohydrate metabolism to produce SCFAs and stimulate insulin secretion [19,20]. In contrast, the colonization of Bacteroides uniformis in hFMTCap-treated mice was significantly reduced compared to hFMT treated controls. Taken together, this suggests that the antimicrobial biocapsule coating exerts a ‘soft kill’ on particular colonic microflora, temporarily modulating the consortium density and diversity to lower colonization resistance and enhance microbial restructuring following delivery of the hFMT cargo.

To further assess whether this microbial perturbation would facilitate niche availability for donor strains, we performed differential abundance analyses at 48- and 168-h post-transplantation. Results for hFMT and hFMTCap groups revealed that members in Clostridia UCG-014 and the Ruminococcaceae family were selectively enriched in the hFMTCap group, whereas Bacteroides uniformis was more abundant in free hFMT treated animals. While biocapsule-mediated depletion of B. uniformis is concerning given its importance to gut health [[21], [22], [23]], enrichment of SCFA-producing Ruminococcaceae and Clostridia UCG-014 species are noteworthy due to their emerging roles in positively regulating gut barrier function and mucosal inflammation [[24], [25], [26]]. This suggests that hFMT-loaded biocapsules may bias the GI consortium towards Clostridia species that may be able to therapeutically manage inflammatory bowel disorders.

3. Discussion

Overcoming colonization resistance remains a primary barrier to the GI persistence of engineered therapeutic microbes, where delivered commensals must compete with established flora for both habitat and nutrient sources. Current methods to reduce colonization resistance prior to fecal transplant mainly rely on antibiotics and pre-enematic bowel lavage approaches, which are invasive and can lead to systemic changes in microbial ecology that negatively impact patient health outcomes [27,28]. This urges materials-enabled strategies that can transiently manipulate gut ecology to promote the installation of beneficial species in a spatiotemporally precise manner. The biocapsule technology reported here leverages a clear and replace strategy to augment the colonization efficiency of the commensal payload, while protecting the microbial cargo during GI transit. While the biocapsule's antimicrobial coating is designed to exert a localized “soft kill” effect, its potency and selectivity toward the manifold of GI resident microbes remains incompletely understood. Our current data suggests that the coating transiently suppresses specific taxa within the colonic niche, creating a window for engraftment without causing widespread microbiota depletion. This targeted approach stands in contrast to conventional antibiotics, which broadly disrupt the gut ecosystem and often impair colonization by beneficial strains. By modulating microbial density at a local level, the biocapsule platform may offer a more precise and ecologically balanced strategy for microbiome engineering. Additional incorporation of microbiome-activated release strategies, as has been recently reviewed [29], could further enhance the anatomical precision of microbial payload release. Moreover, the versatile composition of the material itself could allow loading of diverse antimicrobial or immunomodulatory compounds, as well as provide a privileged habitat for an ecologically varied cocktail of probiotic microbes pre-adapted to the gut environment. In fact, when delivering a complex mixture of bacteria, such as a human fecal microbiota transplant, biocapsules demonstrate a capacity to preferentially support the outgrowth of non-pathogenic Clostridia, which may have therapeutic relevance in treating certain GI disorders and bacterial infections. Screening other commensal bacteria and complex consortia may identify additional species amenable to this targeted engraftment strategy and enable scientists to better deconvolute the roles each member in the GI microbiome play in regulating human health.

It is important to recognize that successful engraftment of delivered probiotics depends on multiple, interrelated factors. Although the density of native flora represents a substantial barrier, and one that our delivery strategy is specifically designed to overcome, exogenous commensals must also acclimatize to the GI mucosal environment and metabolize available carbon sources. This ecological adaptation is not directly addressed by our current capsule design and necessitates careful selection of bacterial strains, or the development of engineered organisms, that are tolerant of, and competitive within, the target niche. Such considerations may help explain why encapsulated L. rhamnosus, despite exhibiting delayed excretion kinetics relative to unencapsulated controls, did not achieve sustained colonization in our animal experiments. This may be overcome by delivering higher bacterial loads and/or repeated dosing.

That said, although our goal was to promote long-term commensal engraftment, there are several applications where transient functional modulation is preferred, and so may benefit from our delivery approach. Thus, our findings demonstrate that biocapsules have the potential to functionally modulate probiotic residence time within the host environment to improve functional outcomes, and may enable more persistent integration of commensal species when paired with appropriate strains and given in a multidose format. Collectively, this approach may help reinvigorate both research and commercial efforts in probiotic development and enhance the durability of next-generation microbiome-based therapies.

4. Materials and methods

4.1. Bacterial and mammalian cell culture

Lacticaseibacillus rhamnosus 4R2127 was grown in MRS with 0.55 mM L-cysteine in a conical tube at 37°C and was used at 24 h. For differentially plating this strain, MRS agar (VWR 90004-084) with 50 μg/mL vancomycin was employed. Lactiplantibacillus acidophilus L917 liquid cultures were grown under the same conditions and were harvested at 24 h. Bifidobacterium longum (ATCC 15707) was grown in TSB (Fisher Scientific ICN1010717) with 0.55 mM L-cysteine, and microaerobic culture conditions were maximized through the use of a glass vial with an airtight cap. Escherichia coli 101-1, Shigella flexneri 2a 457T, Salmonella enterica subsp. enterica (ATCC 14028), and Staphylococcus aureus ST8:USA300 were cultured in MHB2 (Sigma Aldrich 90922) in a flask under 200 rpm shaking at 37°C and were used at 24 h. GFP Shigella flexneri (ATCC 12022GFP) was cultured in MHB2 with 100 μg/mL ampicillin under the same conditions as described previously. For differential plating of GFP S. flexneri, MHB2 agar with 100 μg/mL ampicillin was used. When L. rhamnosus and S. flexneri were grown together for kill and replace studies, a ratio of 1:1 MRS:MHB2 was used as the growth medium, and differential plating according to each strain's above specifications allowed for strain-specific quantification. Anaerobic strains E. coli (MG1655) and C. scindens (DSM 6597) were cultured anaerobically in BHI CHVR under static growth conditions and used at 18 h after inoculation.

Human FMTs were prepared by the Bisanz Lab using donors JBZ16 and JBZ77 and were handled under anaerobic conditions. Fecal samples of 200 mg were suspended in 2 mL of BHI CHV with 15% glycerol, allowed to sit for 5 min after which settled solids were discarded, and frozen for storage. Human fecal matter was collected under an approved study protocol (PSU IRB STUDY00018171).

Caco2 (ATCC HTB-37) cells were cultured using DMEM (Corning 10013CV) with 4 mM L-glutamine, 1x Pen/Strep, 1x MEM NEAA (Gibco), and 10% FBS. A culture at 500 cells per μL was used to seed cells into a 48-well plate at 500,000 cells per well.

4.2. Biocapsule fabrication and characterization

A solution of 30 mM CaCl₂, 0.5 mg/mL hyaluronic acid (100 kDa), 1.0 mg/mL pectin, L. rhamnosus at 20 OD₆₀₀, and 2.0 mM L-cysteine in MRS with (biocapsule) or without (sterile capsule) L. rhamnosus was extruded into low viscosity sodium alginate (8 mg/mL) using an 18-gage needle until the droplet shape reaches a stable sphere with a diameter of 3 mm, at which point the droplet was held suspended in solution for 5 s to increase droplet stability. The needle was then pulled out of solution, causing the droplet to rebound from the liquid-air interface as an independent unit. The formed capsule was then moved using a trimmed pipette tip to a well plate where it was suspended in 2 mL of sodium alginate solution and set to shake at 180 rpm for 30 min. Biocapsules were flipped 10 min after addition to alginate to ensure uniform gelation and spherical architecture. Next, biocapsules were moved to 0.0125% w/v solution of PLL with 30 mM CaCl₂ in water for 30 min with continued shaking to form a stable coating. Uncoated biocapsules used as controls were formed in 30 mM CaCl₂ only. All steps were conducted at room temperature. Next, capsules were transferred to MRS broth containing 30 mM CaCl₂ and incubated for 24 h at 37°C under shaking at 40 rpm. During particle optimization experiments, concentrations of CaCl₂ at 100 - 1600 mM, alginate at 0.5 - 8.0 mg/mL, PLL at 0.03 - 0.50 mg/mL and HA at 0.5 - 8.0 mg/mL were tested. To evaluate bacterial viability within the capsule, each capsule component was loaded into a 96-well plate and a two-fold serial dilution was used to achieve the final concentrations previously mentioned in PBS, before addition of the representative commensal at an OD₆₀₀ of 0.1. Wells of PBS alone served as a control. The plate was then covered with a BreatheEasy plate seal, before OD₆₀₀ measurements taken every 10 min to track growth.

Response of bacterial strains to biocapsule components was tested in a 96-well format, growing four representative GI species, E. coli 101-1, S. aureus, S. enterica, and S. flexneri, over 24 h at 37°C under medium-speed double orbital shaking. Monocultures were seeded at an OD₆₀₀ = 0.01 into HA diluted 10-fold in MHB2, with HA concentrations ranging from 0.004 to 0.5 mg/mL. In a similar experiment, model commensals were seeded at OD₆₀₀ = 0.2 into biocapsule reagents diluted in their corresponding media. The commensals L. acidophilus and L. rhamnosus were grown in MRS, and B. longum in TSB. HA was tested over a range of 0.5 - 15 mg/mL, alginate over 0.5 - 15 mg/mL, and PLL over 0.02 - 0.5 mg/mL.

To assess diffusion of nutrients across the material matrix, blue (FD&C blue 1) and red (FD&C red 3) food dyes were used as model molecules. For each dye, 30 μL of a 1:10 dilution in distilled water was injected into capsule cores 5 min into alginate gelation. Fabrication then continued as standard. PLL-coated capsules were immediately placed into distilled water and the supernatant sampled during the course of 7 h. Blue dye in the supernatant was detected at an absorbance of 630 nm and red at 495 nm.

Finally, growth of L. rhamnosus, L. acidophilus, and B. longum in the capsule core was monitored by forming biocapsules containing the strain of interest and transferring the capsule into wells of the corresponding growth medium with 30 mM CaCl₂ before incubation at 37°C. At 0, 24, and 48 h of incubation the core solutions were extracted and measured for volume then plated in a tenfold dilution series. Plates were incubated at 37°C for 24 h (L. acidophilus and B. longum) or 48 h (L. rhamnosus), at which time colonies were counted to obtain CFU.

4.3. Electron microscopy

Cryo Scanning electron microscopy (SEM) was performed using a Zeiss SIGMA VP-Field emission SEM equipped with Gatan Alto-2500. The particles were placed into rivets mounted onto the cryo-SEM sample holder and then plunged into slushed liquid nitrogen for vitrification. To view intracapsular structures, capsules were subjected to a freeze-thaw protocol to fracture the material. Samples were sublimated at −95°C for 10 min prior to sputter coating with gold for 120 s. Imaging was performed at −195°C using an accelerating voltage either at 10 kV or 20 kV with working distance of 10 mm.

4.4. Antimicrobial testing

Minimum inhibitory concentration (MIC) assays were performed by dissolving polymers at 200 μg/mL in sterile nuclease-free water and two-fold serially diluting the treatment stock in broth (50 μL volume after dilution) in a sterile 96-well round bottom plate in triplicate. Bacterial growth was determined by visual inspection.

To characterize the antimicrobial capabilities of the PLL biocapsule coating, S. aureus, E. coli, S. enterica, and S. flexneri were each grown in planktonic cultures in MHB2 in a 24-well plate format, with starting concentrations ranging from 10² to 10⁷ CFU/mL. An uncoated or coated capsule was manually added to appropriate wells and plate covered with a BreatheEasy plate seal. Untreated bacteria, or blank broth, were used as negative and positive controls, respectively. After a 24-h incubation at 50 rpm shaking, OD₆₀₀ area scans were collected and capsule contribution to the signal excluded during post-processing. Assays employing gut commensals utilized cultures of E. coli (MG 1655) and C. scindens (DSM 6597) grown in an anaerobic chamber. Monocultures were seeded at OD₆₀₀ = 0.1 in BHI CHVR and tracked for viability over 36 h, with plating at 0, 6, 24, and 36 h. Biocapsule impact on a complex community was assessed in vitro by adding both coated and uncoated capsules to a suspension of hFMT. The fecal sample (JBZ77) was suspended in modified GMM (mGMM) media containing 2 mg/mL tryptone peptone, 1 mg/mL yeast extract, 0.4 mg/mL D-glucose, 1 mg/mL fructose, 5 mg/mL meat extract, and 1% w/v L-cysteine. 100 μL hFMT was transferred in fresh 5 mL mGMM and a capsule was added to the suspension and cultured at 37°C for 4 days. At every 24 h, the culture was homogenized via vortexing, and 1 mL of each culture was collected and frozen. 100 μL of overnight culture was transferred to a new tube of fresh 5 mL mGMM and a new capsule was added. 24 hr and 96 h samples were analyzed through 16S sequencing.

Zone of inhibition was measured by plating test bacteria as lawns on MHB2 agar. Once the lawns dried, uncoated and coated capsules were pinned to the agar to maintain contact with the plate during incubation. Plates were incubated for 24 h at 37°C after which the clearance diameter was measured.

4.5. Mammalian cell viability

Capsules were formed both with (coated) and without (uncoated) PLL and added to wells pre-seeded with 500,000 cells/well of human Caco2 cells. Untreated cells, or those incubated with 20% DMSO, served as negative and positive controls, respectively. Cultures were incubated for 48 h at 37°C and 5% CO₂, after which MTT conversion was used to evaluate cell viability.

4.6. Simulated intestinal environment

Biocapsules were submerged in either simulated gastric fluid (SGF, Ricca R7108000, pH of 4.5, 0.1 mg/mL L-cysteine), fasted state simulated small intestinal fluid (FaSSIF), or simulated colonic fluid (SCoF) and incubated at 37°C under 50 rpm shaking over time periods commensurate with gastrointestinal transit in these compartments (see composition of simulated intestinal fluids in Table 1). Viability of encapsulated bacteria in the supernatant or capsule core solution were subjected to plating to determine CFU, and CFU/mL determined from the relevant compartment volume, as previously described.

Table 1.

Compositions of simulated intestinal fluids used for in vitro biocapsule testing.

FaSSIF Components	Concentration
sodium taurocholate	3 mM
lecithin	0.75 mM
sodium phosphate monobasic	3.95 g/L
sodium hydroxide	0.35 g/L
sodium chloride	6.19 g/L
Properties
pH	6.5
SCoF Components	Concentration
potassium chloride	0.2 g/L
sodium chloride	8 g/L
potassium phosphate monobasic	0.24 g/L
sodium phosphate dibasic	1.44 g/L
Properties
pH	7

4.7. Polymicrobial studies

L. rhamnosus loaded biocapsules, or sterile controls, were added to cultures of S. flexneri (ATCC 12022GFP) seeded in 1:1 MRS:MHB2 at 0.0001 OD₆₀₀. Experimental controls included sterile biocapsules and a L. rhamnosus-loaded capsule. Biocapsules were subjected to 40 rpm shaking at 37°C and imaged over time to observe degradation. Additionally, kill and replace mechanisms were measured using differential plating. Here, aliquots were removed from the supernatant every 24 h, and replaced with an equal volume of broth. This analysis was conducted in two stages – first over a three-day period to observe higher resolution growth characteristics, then over a seven-day period to visualize extended growth patterns. Both experimental sets utilized controls including sterile MHB2, GFP S. flexneri alone, biocapsules alone, and sterile capsules in GFP S. flexneri, along with the experimental condition of biocapsules in GFP S. flexneri. All conditions were incubated at 37°C under 40 rpm shaking. Each time point collection required pipet mixing of the supernatant which agitated the broth. Growth was assessed via differential plating as previously described.

4.8. Capsule miniaturization and in vivo studies

Biocapsule miniaturization was conducted to achieve a target loading capacity of 10⁷ CFU L. rhamnosus, as is typical in delivering commensals to mice, while maintaining a diameter that allowed capsules to be delivered through a 17G oral gavage needle. The standard biocapsule fabrication procedure was carried out with the exceptions of using a 17G needle for extrusion and shortening the time in the alginate bath to 1 min. Additionally, biocapsules were transferred with a trimmed 100 μL pipet tip and washed in distilled H₂O between baths. Next, 16-week-old male C57BL/6J mice weighing at least 30 g were purchased from JAX. The mice were acclimatized to the Penn State University animal facility for 5 days, after which they were housed individually for the study with ad libitum water and food (Lab Diet 5021) under a 12-h light/dark cycle. All animal studies were conducted under an approved study protocol (PSU IACUC PROTO202101826). Mice were administered L. rhamnosus loaded biocapsules in four separate studies. First, a 24-h study to analyze the release profile of a single biocapsule dose and track distribution of the bacteria payload in the gastrointestinal contents, followed by three 7-day studies to observe long-term trends in engraftment which examined L. rhamnosus engraftment for a single biocapsule dose, impacts on treatment longevity when dosing with multiple L. rhamnosus biocapsules, and single hFMT (JBZ16) biocapsules for screening commensal candidates for future applications.

In the 24-h study, mice were organized into 3 groups and gavaged with either a sterile biocapsule, free L. rhamnosus, or an L. rhamnosus biocapsule. Feces were collected at 0 - 24 h after treatment. Mice were euthanized at 24 h and their intestinal tracts sectioned into duodenum, ileum, cecum, and colon. The contents of these segments were removed and stored separately from the tissue, both flash frozen for storage. Fecal pellets and GI contents were weighed then dissolved in 1 mL PBS, after which they were serially diluted in PBS and plated on MRS + vancomycin agar with four technical replicates to determine the fecal/intestinal load of L. rhamnosus on a CFU basis. For mice in the 7-day single biocapsule study, groups remained the same as above, while the multi-dose study animals received either L. rhamnosus, L. rhamnosus + free PLL, or 15 L. rhamnosus biocapsules. Animals in the hFMT study were administered either a sterile biocapsule, free hFMT, or an hFMT biocapsule. In all cases, culture controls were normalized using 600 nm absorbance to the microbial load delivered via biocapsule(s). Free PLL was normalized to the amount of polypeptide present on biocapsules, which determined using a BCA assay with a PLL standard curve. The same procedures were used for the 7-day in vivo experiments as in the 24-h study, with fecal collections now at 0 h, 3 h, 6 h, 9 h, 1d, 2d, 3d, 5d, and 7d, and endpoint dissection at 7 d. Fecal samples and intestinal contents were again plated for CFU quantification. Results were normalized against sample mass. For mice dosed with single L. rhamnosus biocapsules, inflammation in the colon was then quantified by measuring lipocalin-2 (Lcn2) in both the feces and the proximal and middle + distal colon segments. First, GI tissues were homogenized, then both suspended feces and dissolved tissue were centrifuged at 4°C for 15 min at 12,000g. The supernatants were extracted and their Lcn2 concentrations measured using a Lcn2/NGAL ELISA kit. Fecal Lcn2 was normalized according to feces mass, while colon tissues were normalized according to each sample's protein concentration, which was found using BCA assays. For mice dosed with multiple L. rhamnosus biocapsules, liver toxicity markers and short chain fatty acids (SCFAs) were quantified using blood serum and cecal contents, respectively. Blood serum was extracted by centrifuging whole blood at 5000 RCF for 5 min, 2 h after collection and kept at 4°C, then analyzed using the Vetscan VS2 Chemistry Analyzer (Abaxis, Inc., Union City, CA, USA) with mammalian liver profile rotors according to the manufacturer instructions. SCFAs were quantified using NMR, conducted in cooperation with core facilities in the Huck Institutes of the Life Sciences at Penn State University.

For our final in vivo experiment, capsules were prepared with human fecal microbiota transplant (hFMT) samples of 100 mg. Here, the fecal mass was suspended in 1 mL MRS, dissolved, fibers allowed to settle, and 1 μL of the supernatant moved into 999 μL MRS to create the loading culture. 1 mL of the supernatant was then extracted and centrifuged on a desktop spinner until a significant pellet formed, about 3 min. This supernatant was used to create the core solution that, once dissolved, had L-cysteine added and was used to resuspend the hFMT pellet. Fabrication and preparation of hFMT biocapsules was performed in an anaerobic chamber, and the biocapsules were transported to the animal facility in a well plate covered in a plate seal to minimize gas exchange. Feces were collected at 0 h, 3 h, 6 h, 9 h, 1d, 2d, 3d, 5d, and 7d, and endpoint dissection was conducted at 7 d. Feces were analyzed using qPCR to track fecal load of hFMT strains.

4.9. Gut microbiome characterization

Fecal microbiome samples were sequenced to determine the microbial composition following the standard protocol in the One Health Microbiome Center Co-Laboratory at Penn State. The complete protocol can be found at https://github.com/BisanzLab/OHMC_Colaboratory. In short, DNA was extracted using Qiagen DNeasy PowerSoil Pro Kits (47014) by bead lysis performed on Qiagen Tissuelyzer III. DNA purification process was automated on QIAcube HT. Primary PCR was performed with 515F and 806R primers with overhangs for i7 and i5 adapters using KAPA HiFi HotStart PCR kit (KR0370). Amplification was monitored real-time on a Sybr-based method on QIAquant 384 to ensure successful amplification of samples and a lack of successful amplification in negative controls. Amplicons were further diluted 100X and indexed using 10 nt unique dual indexes based on Illumina IDT TagmentationSet ABCD before quantification and equal-molar pooling. The final library went through two rounds of size selection using Ampure XP beads (A63881) before sequencing on an Illumina Nextseq 2000 using a XLEAP P1 600 cycle flow cell.

Raw sequencing data were processed using Quantitative Insights Into Microbial Ecology 2 (Qiime2 [30]) version 2023.5. Primer sequences were removed using Cutadapt. Forward reads were trimmed at 220 bases and reverse reads were trimmed at 150 bases based on the read quality. DADA2 [31] as implemented in QIIME2 was applied to trimmed reads to denoise the data, merge paired reads, remove chimeric sequences, and resolve amplicon sequence variants (ASVs). Taxonomic classifications of representative sequences were performed using DADA2's classifier with the SILVA 138 database for 515F/806R sequences [32]. ASVs belonging to host mitochondrial or chloroplast sequenced were removed. Four extraction controls were included, and all had total reads below 25 indicating a minimal level of probable cross-contamination. Zymo extraction control (D6302) was included and found no reads belonging to strains that do not belong to the extraction control were found. Experimental samples contained 112,851 ± 66,126 reads after processing. Data processing and visualization was performed using qiime2r v 0.99.6. Diversity analysis was performed using Vegan v 2.7-1 and differential abundance calculations was done by ALDEx2 [33] v 1.40.0.

Ethics approval and consent to participate

Human fecal matter was collected under an approved study protocol (PSU IRB STUDY00018171). Written/oral informed consent was obtained from all sample donors. All animal studies were conducted under an approved study protocol (PSU IACUC PROTO202101826).

CRediT authorship contribution statement

Lily Foley: Formal analysis, Investigation, Methodology, Writing – original draft. Sonika Kohli: Formal analysis, Investigation. Shuchang Tian: Investigation, Methodology, Writing – review & editing. Logan C. Eisaman: Investigation, Writing – review & editing. Harshita Ganga: Investigation, Writing – review & editing. Gabrielle Hamner: Investigation. Matthew R. Aronson: Investigation. Jordan E. Bisanz: Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing – review & editing. Scott H. Medina: Conceptualization, Data curation, Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing – original draft, Writing – review & editing.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Scott H. Medina reports financial support was provided by National Science Foundation. Scott H. Medina reports financial support was provided by National Institutes of Health. Jordan E. Bisanz reports financial support was provided by National Institutes of Health. Scott H. Medina has patent #2019-5012 pending to The Penn State Research Foundation. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data collected for this study is found in the text or in the Supplementary Information. Raw data is available upon reasonable request to the corresponding author.