L. casei strain T21 was obtained from the stock culture of the Department of Microbiology, Faculty of Medicine, Chulalongkorn University. Bacterial stock culture was maintained in deMan Rogosa Sharpe (MRS) broth (Oxoid, Hampshire, United Kingdom) containing 20% (vol/vol) glycerol at –80°C. L. casei T21 was cultured on MRS agar under anaerobic conditions using gas generation sachets (Anaero Pack-Anaero, Mitsubishi Gas Chemical, Japan) at 37°C for 48 h. C. difficile ATCC BAA1870 (ATCC, Manassas, VA, United States) was cultured anaerobically on Brucella agar (Becton Dickinson, France) supplemented with 5% (vol/vol) sheep blood at 37°C for 48 h.

The experimental protocol in accordance with the US National Institutes of Health standards (NIH publication no. 85–23, revised 1985) was approved by the Institutional Animal Care and Use Committee of the Faculty of Medicine, Chulalongkorn University (SST006/2560). Male 8-week-old C57BL/6 mice were purchased from the Nomura Siam International Co., Ltd. (Lumphini, Pathumwan, Bangkok, Thailand). CDI mouse model as previously developed (Chen et al., 2008) and recently published (Panpetch et al., 2019) was performed with modifications. Briefly, 500 μL of the antibiotic cocktail (Sigma-Aldrich, St. Louis, MO, United States) containing gentamicin (3.5 mg/kg), colistin (4.2 mg/kg), metronidazole (21.5 mg/kg), and vancomycin (4.5 mg/kg) was administered by oral gavage twice a day from day –6 to day –4 before C. difficile infection (D-6–D-4) (Figure 1A). Mice were free from antibiotic administration for 2 days and received an intraperitoneal injection of a single dose of clindamycin (10 mg/kg) at 1 day before infection (D-1). After the treatment with antibiotics (ATB), mice were gavaged with either 0.5 ml of normal saline solution (NSS) in the ATB-administered uninfected group (ATB uninfected group; n = 12) or 1 × 10¹⁰ colony forming units (CFU) of C. difficile vegetative cells in 0.5 ml of NSS once daily for 2 days (D0 and D1) in the C. difficile group (n = 24). Mice were observed and monitored daily for weight, stool consistency, and survival until D7. Blood was collected through tail vein nicking for enumeration of bacteria at D2, D4, and D7. According to our pilot study, C. difficile-infected mice developed severe symptoms for 3 days, some succumbed to infection, and the survivors gradually recovered from CDI. At D3, mice from each group (n = 8) were then tested for gut leakage and sacrificed with cardiac puncture under isoflurane anesthesia for determining CDI severity by using serum pro-inflammatory cytokines as markers. The stool consistency was semi-quantitatively evaluated using the following scoring; 0, normal; 1, soft or loose; and 2, diarrhea, as previously published (Kim et al., 2012).

For Lacticaseibacillus treatment in the CDI mouse model, mice were randomly divided into three groups: ATB uninfected group (n = 18), C. difficile-infected mice treated with NSS (NSS group; n = 24), and C. difficile-infected mice treated with L. casei T21 (T21 group; n = 20), which received 1 × 10⁸ CFU of L. casei T21 in 0.5 ml of NSS once daily for 4 days from D-1 (started at 6 h after clindamycin injection), D0 and D1 (together with C. difficile), and D2 (Figure 2A). Mice were observed and monitored daily until D7 as described above. Feces were collected for microbiome analysis at D-6 for baseline and D-1 (before clindamycin injection and gavage with L. casei T21) as time of post-ATB administration. At D3, some mice from each group (n = 8) were tested for gut leakage and sacrificed for determining the parameters of CDI severity. Blood samples and cecal and ascending colonic tissues were collected for the assessment of cytokine levels, which represented systemic and local inflammation, respectively. Cecal and ascending colonic tissues were also used for histopathologic evaluation, while luminal content including feces in the cecum and colon were used for quantitation of C. difficile and microbiome analysis. All mice were sacrificed on D7 at the end of experiment.

The pro-inflammatory cytokines as previously described (Leffler and Lamont, 2015), including IL-1β, tumor necrosis factor (TNF)-α, macrophage inflammatory protein 2 (MIP-2), and keratinocyte chemoattractant (KC), in serum and homogenized tissue of the cecum and colon were measured by an enzyme-linked immunosorbent assay (ELISA) (PeproTech, NJ, United States). For determining tissue pro-inflammatory cytokines, tissue samples (approximately 100 mg) were weighed and homogenized using an Ultra-Turrax homogenizer (IKA, Staufen, Germany) in 500 μL of phosphate-buffered saline (PBS, pH 7.4) containing protease inhibitor and centrifuged at 12,000 × g for 15 min at 4°C to separate the supernatant for analysis.

For quantitation of C. difficile in luminal content including feces, quantitative real-time polymerase chain reaction (qPCR) was performed as previously published (Panpetch et al., 2019). Briefly, genomic DNA of C. difficile was extracted from cecum and colon contents using the High Pure PCR Template Preparation Kit (Roche, NJ, United States), quantified by NanoDrop™ 1000 Spectrophotometer (Thermo Fisher Scientific, Inc., United States), and amplified with tcdB (C. difficile toxin B) primers (forward, 5′-GGAAAAGAGAATGGTTTTATTAA-3′ and reverse, 5′-ATC TTTAGTTATAACTTTGACATCTTT-3′) as previously described (Lemee et al., 2004) in the QuantStudio 5 Real-Time PCR System (Thermo Fisher Scientific, Inc., United States) using the QuantiNova^® SYBR^® Green PCR Kit (QIAGEN, Hilden, Germany). The standard curve was created by using 10-fold serially diluted plasmids containing 1–10⁹ copies of tcdB. The number of C. difficile was calculated using the standard curve and shown as bacterial copy number.

Histopathological analysis was performed by fixing the sample in 10% buffered formalin, embedding in paraffin, sectioning, and staining with hematoxylin and eosin (H&E) before grading (score 0–4) as previously published (Reeves et al., 2011; Erikstrup et al., 2015) based upon epithelial damage, edema, and cellular infiltration by two pathologists in a blinded manner.

Gut microbiota composition was determined as previously reported (Panpetch et al., 2020). Feces collected at D-6 as baseline, D-1 as time of post-ATB administration, and luminal content including feces at D3 as time of sacrifice were used for microbiome analysis. Briefly, fecal samples or luminal content (0.25 g/mouse; three mice/group) were extracted for total DNA with the DNeasy PowerSoil Kit (Qiagen GmbH, Hilden, Germany). The quality and concentration of the extracted DNA were monitored by agarose gel electrophoresis and NanoDrop spectrophotometry. Libraries of the V4 hypervariable region of 16S rRNA gene were amplified by PCR using primers 515F (forward; 5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (reverse; 5′-GGACTACHVGGGTWTCTAAT-3′), modified with the Illumina adapter and Golay barcode sequences as previously described (Caporaso et al., 2012). PCR was run in triplicate, and the products from the triplicate reactions were pooled and visualized on agarose gel. Amplicons of approximately 381 base pairs were purified by PureDireX PCR Clean-Up & Gel Extraction Kit (BIO-HELIX Co., Ltd., Keelung City, Taiwan) and quantified using PicoGreen fluorescence with the Qubit dsDNA HS assay kit (Invitrogen, Eugene, OR, United States). The amplicon pool was sequenced with the Illumina MiSeq300 platform (Illumina, San Diego, CA, United States) (Caporaso et al., 2012). Sequences were analyzed with Mothur version 1.3 (Schloss et al., 2009). Briefly, quality filtering and trimming were performed to remove low-quality bases and short reads from the raw sequences. Quality-filtered sequences were then aligned to each other and binned into operational taxonomic units (OTUs) with a minimum of 97% similarity. Each representative OTU sequence was compared to the SILVA rDNA sequence database (version 1.32) and assigned a taxonomical annotation. Alpha diversity (total OTUs, Chao1 index, and Shannon diversity) and beta diversity (non-metric multidimensional scaling) were calculated using Mothur (Schloss et al., 2009). The 16S rDNA sequences in this study were deposited in an NCBI open access Sequence Read Archive database with accession number SRP336496.

Intestinal epithelial permeability defect (gut leakage) was determined using a single oral administration of 12.5 mg fluorescein isothiocyanate–dextran (FITC-dextran; molecular weight 4.4 kDa) (Sigma-Aldrich, St. Louis, MO, United States), a non-intestinal-absorbable marker, before the determination in serum at 3 h later as previously described (Leelahavanichkul et al., 2016; Panpetch et al., 2018). Serum FITC-dextran was measured by the fluorospectrometry (Thermo Fisher Scientific, Wilmington, DE, United States) with the excitation and emission wavelengths at 485 and 523 nm, respectively, against a standard curve of serially diluted FITC-dextran. For the enumeration of live bacteria, blood (25 μL) was collected through tail vein nicking and spread directly onto blood agar (Oxoid, Hampshire, United Kingdom) and incubated at 37°C for 24 h before counting bacterial colonies.

The conditioned medium of L. casei T21 was tested for immunomodulation of IL-8 production in colonic epithelial cell lines as previously described (Panpetch et al., 2016, 2018). In brief, Lacticaseibacillus-conditioned medium (LCM) was prepared by growing L. casei T21 with an OD₆₀₀ of 0.1 in MRS broth anaerobically for 48 h. The supernatant was collected and subjected to filtration with a 0.22-μm membrane (Minisart, Sartorius Stedim Biotech GmbH, Goettingen, Germany), concentrated by speed vacuum drying, resuspended in cell culture medium with equal volume, and stored at –20°C until use. In parallel, human colonic epithelial cell lines Caco-2 (ATCC HTB-37) and HT-29 (ATCC HTB-38) were maintained (5 × 10⁴ cells/well) in supplemented Dulbecco’s Modified Eagle Medium (DMEM) and McCoy’s 5A modified medium, respectively. Colonic epithelial cells were then incubated with viable cells of C. difficile ATCC BAA1870 at multiplicity of infection (MOI) 1:300 either alone or with 5% (vol/vol) LCM for 24 h in 5% CO₂ at 37°C. Subsequently, the supernatant was collected by centrifugation (125 × g, 4°C for 7 min), and the levels of IL-8 were measured by using a Human CXCL8/IL-8 ELISA kit (R&D Systems, Minneapolis, MN) according to the manufacturer’s instructions.

In addition, colonic epithelial cells at 2 and 4 h from the incubation time were collected for performing quantitative reverse transcription-polymerase chain reaction (qRT-PCR) as previously described (Panpetch et al., 2018). In short, the total RNA of treated colonic epithelial cells was extracted by TRIzol reagent (Invitrogen, United States), prepared for complementary DNA (cDNA) from total RNA (50 ng) by SuperScript^® VILO™ cDNA Synthesis Kit (Invitrogen), and subjected to qPCR measurement in a QuantStudio™ Design & Analysis Software v1.4.3 (Thermo Fisher Scientific) with the following primers: IL-8 (forward 5′-ACACTGCGCCAACACAGAAATTA-3′, reverse 5′-ACACTGCGCCAACACAGAAATTA-3′) and human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (forward 5′-GCACCGTCAAGGCTGAGAAC-3′, reverse 5′-ATGGTGGTGAAGACGCCAGT-3′) (Imaoka et al., 2008; Panpetch et al., 2016). The expression of IL-8 relative to GAPDH was calculated according to the 2^−ΔΔCp method (Pfaffl, 2001).

The conditioned medium of L. casei T21 was tested for its effect on the expression of C. difficile-activated host genes SLC11A1, HuR, and MUC2 by using qRT-PCR as described above with the following primers: SLC11A1 (forward 5′-CTGGACGAATCCCACTCTGG-3′, reverse 5′-CGCGCCACCACATACTCAT-3′), HuR (forward 5′-GCTTGGGCTATGGCTTTGTGAACT-3′, reverse 5′-CGCTG ATGTACAAGTTGGCGTCTT-3′) (Feng and Cohen, 2013), and mucin2 (MUC2) (forward 5′-CCTGCCGACACCTGCTGCAA-3′, reverse 5′-ACACCAGTAGAAGGGACAGCACCT-3′) (Xue et al., 2014). In parallel, the pH of cell culture medium was measured at multiple time points using a pH meter (Orion 4-star, Benchtop pH/Conductivity, Thermo Fisher Scientific).

TEER was performed according to a previous report (Gao et al., 2017). In short, Caco-2 cells (ATCC HTB-37) at 5 × 10⁴ cells per well were seeded onto the upper compartment of a 24-well Boyden chamber transwell using high-glucose DMEM supplemented with 20% fetal bovine serum (FBS), 1% HEPES, 1% sodium pyruvate, and 1.3% penicillin/streptomycin under 5% CO₂ at 37°C for 15 days with daily medium replacement to establish the confluent monolayer. The cells were then treated with 5% (vol/vol) LCM of L. casei T21 or medium alone together with viable C. difficile cells (5 × 10⁶ CFU/well) with MOI at 1:100 for 24 h. Next, TEER was measured by an EMOM² Epithelial Volt/Ohm Meter with a chopstick electrode (World Precision Instruments, Inc., Sarasota, United States) that was placed at a 90° angle with one tip in supernatant at the basolateral chamber and another tip at the apical chamber. The TEER value in control media without cells was used as a baseline subtracted from all measurements. The value of TEER was reported as ohm (Ω) × cm².

Mean ± standard error of mean (SEM) was used for data presentation. The difference between groups was examined for statistical significance by one-way analysis of variance (ANOVA) followed by Tukey’s analysis or unpaired t tests for comparisons of multiple groups or two groups, respectively. Survival analysis was performed by log-rank test. All statistical analyses were performed with GraphPad Prism version 9.0 software (La Jolla, CA, United States). A p-value of < 0.05 was considered statistically significant.

In the CDI murine model (Figure 1A), C. difficile-infected mice began to develop CDI symptoms such as weight loss and soft stool on day 1 after the first oral gavage with C. difficile on day 0. After the second oral gavage on day 1, CDI symptoms became worse on day 2, and mice were moribund on day 3 with maximum weight loss (Figure 1B) and significant diarrhea (loose stools) as compared with the ATB uninfected group (Figure 1C). By day 3, 41.67% (10/24) of mice succumbed to infection (58.33% survival rate) (Figure 1D). Eight mice were sacrificed on day 3, and the remaining six mice gradually gained weight while still having diarrhea and surviving until the end of the experiment (Figures 1B,C). Additionally, C. difficile damaged intestinal integrity as demonstrated by the increased serum FITC-dextran levels (Figure 1E), causing gut leakage-induced bacteremia (Figure 1F) that enhanced the production of systemic inflammatory cytokines (serum IL-1β, TNF-α, KC, and MIP-2 levels as markers) (Figures 1G–J). In contrast, mice in the ATB uninfected group demonstrated loose stools for a few days (days 1–3) without weight loss, death, gut leakage, bacteremia, or systemic inflammation (Figures 1B–J).

For the treatment of L. casei T21 in the murine model of CDI, C. difficile-infected mice in the T21 group received 1 × 10⁸ CFU of L. casei T21 once daily for 4 days (D-1–D-2), whereas infected mice in the NSS group received NSS (Figure 2A). All mice in the T21 and the ATB uninfected group survived, while only 54.17% (13/24) of the NSS group survived by day 3 (Figure 2B). Compared to the NSS group, the T21 group had significantly lesser weight loss (except at days 4 and 5) (Figure 2C) and diarrhea (Figure 2D), which were monitored for 7 days. The average weight at each day (except at days 4 and 5) between the T21 group and the ATB uninfected group showed no significant difference (Figure 2C). However, mice in the T21 group still had soft stool with a stool consistency index significantly different from the ATB uninfected group (Figure 2D). Treatment with T21 also reduced gut leakage and systemic inflammation. The T21 group had significantly decreased levels of FITC-dextran (Figure 2E) and pro-inflammatory cytokines IL-1β, TNF-α, MIP-2, and KC in sera (Figures 2F–I). Likewise, L. casei T21 also attenuated intestinal injury in the cecum and colon as evaluated by histopathology (Figures 3A,B) and reduced the levels of pro-inflammatory cytokines in the intestinal tissue (Figures 3C–F). Notably, C. difficile-infected mice without T21 treatment (NSS group) demonstrated several characteristics of severe intestinal injury, including loss of villi, villous edema, numerous neutrophil infiltration (Figure 3A), and neutrophils in feces (data not shown). In addition, T21 treatment reduced the abundance of C. difficile in the intestinal content. The analysis of luminal content including feces by q-PCR of the tcdB gene demonstrated that the T21 group had significantly decreased copies compared to the NSS group (Figure 3G). Moreover, the absence of the tcdB gene in the ATB uninfected mice (Figure 3G) revealed the reliability of the experiment.

Gut dysbiosis of the model was evaluated by a fecal microbiome analysis of mice in the ATB uninfected, NSS, and T21 groups at multiple time points. Alpha diversity measures used in this study included total OTUs (the simplest measure of richness), Chao1 (a measure of richness that gives more weight to rare taxon), and Shannon (a measure of richness and evenness). Antibiotic cocktail treatment significantly reduced the diversity of fecal bacteria as the values of the total OTUs (Figure 4A) and Chao1 index (Figure 4B) were significantly lower in all groups of mice at post-ATB compared to baseline. After clindamycin injection, antibiotic-induced dysbiosis worsened at day 3 of the experiment as determined by total OTUs (Figure 4A), but not by Chao1 index (Figure 4B). C. difficile infection did not lead to a significant decrease of bacterial diversity as demonstrated by total OTUs (Figure 4A) and Chao1 index (Figure 4B). Surprisingly, T21 treatment did not increase the diversity of fecal bacteria as the values of total OTUs (Figure 4A) and Chao1 index (Figure 4B) of the NSS and T21 groups were not significantly different. The T21 group did show an increased Chao1 index, although the value was not significantly different from that of the NSS group. The values of Chao1 index were also not significantly different in the T21 and ATB uninfected groups (Figure 4B). Notably, the bacterial diversity in all groups of mice at day 3, post-ATB, and baseline was not significantly different as determined by the Shannon index (Figure 4C). To examine the beta diversity, non-metric multidimensional scaling (NMDS) based on Thetayc dissimilarity was performed. The NMDS (Figure 4D) demonstrated similar results in all groups at baseline (D-6) (blue-colored symbols at the upper left quadrant) and at post-ATB (D-1) (red-colored symbols at the lower right quadrant). In contrast, there were some differences at D3 from the experiments between groups of mice with or without L. casei T21 (green-colored symbols of the NSS and T21 groups), suggesting a possible impact of L. casei T21 on gut microbiota. The ATB uninfected and T21 groups showed similar results.

To characterize the microbiome composition in each group of mice, the relative taxa abundances at each time point were compared. The gut microbiome of mice at baseline was predominated by Bacteriodetes, followed by Firmicutes and Proteobacteria, as shown by the average relative abundances of microbiota at the phylum level in each group of mice (Figure 5A) and the relative abundances of Bacteroidetes (Figure 5D), Firmicutes (Figure 5E), and Proteobacteria (Figure 5F) at each time point. In contrast, antibiotic cocktail treatment induced a significant decrease in the relative abundance of Bacteriodetes and Firmicutes and a significant increase in the abundance of Proteobacteria and Verrucomicrobia at D-1 (Figures 5A,D–G). Antibiotic pre-conditioning of the model thus caused fecal dysbiosis as indicated by a decrease in Firmicutes and Bacteroidetes together with an increase in Proteobacteria and Verrucomicrobia. By day 3, microbiota composition at the phylum level almost turned to the baseline without T21 treatment. The relative abundance of Bacteroidetes (Figure 5D), Firmicutes (Figure 5E), and Verrucomicrobia (Figure 5G) in the ATB uninfected group and the NSS group was not significantly different from the baseline, whereas the abundance of Proteobacteria was still significantly different from the baseline (Figure 5F). L. casei T21 treatment led to an increase in the relative abundance of Firmicutes in the T21 group with a significant difference compared to post-ATB, but not from the NSS group (Figure 5E). In addition, T21 treatment resulted in a significant decrease in the relative abundance of Verrucomicrobia in the T21 group compared to the NSS group and post-ATB, and the abundance of Verrucomicrobia in the T21 group was not significantly different from the baseline (Figure 5G). Although the relative abundance of Proteobacteria in the T21 group significantly decreased compared to post-ATB, this change also occurred in the NSS group (Figure 5F).

The analysis of microbiome at the lower taxon levels (class, order, family, genus, and species) was performed. The average relative abundances of microbiota at baseline, post-ATB, and among groups of mice at sacrifice (day 3) are shown at the class level (Figure 5B) and the genus level (Figure 5C). There were significant changes in class Clostridia (Figure 5H) and family Enterobacteriaceae (Figure 5I) in the T21 group at sacrifice (D3) as compared with mice at post-ATB. Although the NSS group had changes in the relative abundance of these taxa as compared with mice at post-ATB, the differences were not statistically significant. Microorganisms of interest in class Clostridia (phylum Firmicutes) were the families Lachnospiraceae and Ruminococcaceae (Figure 5C) that have been reported to protect C. difficile colonization (Reeves et al., 2012; Lee et al., 2017). The relative abundances of the family Lachnospiraceae (Figure 5J) were not significantly different among the ATB uninfected, the NSS, and the T21 groups, while there was a slight increase in the abundance of the family Ruminococcaceae in the T21 group compared to the NSS group, although the difference was not statistically significant (Figure 5K). The relative abundances of other taxa (at the class and genus levels) were also not significantly different among these groups of mice at sacrifice (data not shown). Data on the relative abundances of microbiota from individual mouse are shown at the levels of phylum (Supplementary Figure 1), class (Supplementary Figure 2), and genus (Supplementary Figure 3) in Supplementary Material.

Probiotic bacteria, such as Lactobacillus spp., have been reported to produce biologically active compounds that can suppress inflammation (Thomas et al., 2012, 2016; Panpetch et al., 2016, 2018). To strengthen the beneficial effects of L. casei T21 in the murine model of CDI, the LCM of T21 was tested for its ability to attenuate inflammation and modulate the expression of important host genes involved in the pathogenesis of CDI in colonic epithelial cell lines. The LCM of T21 suppressed the production of IL-8 (Figures 6A,B) and down-regulated the expression of IL-8 (Figures 6C,D) in C. difficile-stimulated Caco-2 and HT-29 colonic epithelial cells, respectively. In C. difficile-stimulated Caco-2 cells, the LCM of T21 down-regulated the expression of associated genes of toxin lethality SLC11A1 (Figure 6E) and HuR (Figure 6F) while up-regulated a mucosal protective gene MUC2 (Figure 6G). Similar benefits exerted by the LCM of T21 were also demonstrated in HT-29 cells (Figures 6H–J). Additionally, the LCM of T21 strengthened mucosal integrity as shown by TEER values in C. difficile-stimulated differentiated Caco-2 cells (Figure 6K). However, TEER could not be determined in the HT-29 cell line due to limitation in the generation of polarized monolayers (Le Bivic et al., 1988). Since the enhanced mucin production by enterocytes is also influenced by the acidity in gut content (Velcich and Augenlicht, 1993; Shekels et al., 1996), lactic acid produced from L. casei T21 might stimulate the up-regulation of MUC2 expression. To determine whether the up-regulated MUC2 could be attributable to the acidity, the pH of the cell culture medium (McCoy’s 5A modified medium) was measured at multiple time points. The LCM of T21 did not significantly reduce the pH of the cell culture medium at 2 and 4 h after incubation with HT-29 cells as compared to controls (Figure 6L). A simple acidification of the cell culture medium with lactic acid also did not upregulate MUC2 expression (data not shown). This suggested that mucin production was induced by other substances produced by L. casei T21.