Lactobacillus casei BL23 (26) was grown in Man, Rogosa, and Sharpe medium at 37°C without agitation overnight. To prepare live bacterial inoculum, strains were washed twice with PBS at 3,000 g and suspended in PBS at a concentration of 5 × 10⁹ colony-forming units/ml plus 15% glycerol.

All experiments were handled in accordance with institutional ethical guidelines, and the study was approved by the COMETHEA ethics committee (“Comité d’Ethique en Expérimentation Animale”) of the Centre INRA of Jouy-en-Josas and AgroParisTech. Female C57BL/6 mice (6–8 weeks old; Janvier SAS, St. Berthevin, France) were maintained in sterile isolators at the INRA animal facility (n = 5 per cage) with 12 h light cycles and fed irradiated normal chow (R 03-40, SAFE) and water ad libitum.

Temperature and moisture were carefully controlled. Mice were separated into three groups. The first two groups were administered PBS, while the last group was orally administered L. casei BL23. As shown in Figure 1, administration of PBS or L. casei BL23 started a week before tumor induction and was performed every day until sacrifice.

Tumors were induced with a single intraperitoneal injection of 8 mg/kg AOM (Sigma-Aldrich) 6 days before the start of a DSS treatment (with the exception of group 1, the negative control). Mice received a 5-day course of 2.5% DSS (TdB) in sterile drinking water, followed by an adjustable recovery period. This schema was followed four times (Figure 1). Intestinal inflammation was assessed daily by measuring the Disease Activity Index (DAI), which included body weight loss (0 = <1%, 1 = 5% > X > 1%, 2 = 10% > X > 5%, 3 = 15% > X > 10%, 4 = > 15%); mouse activity (0 = normal, 2 = hooked back, and 4 = lethargy); stool consistency (0 = absence, 2 = soft and sticky, and 4 = diarrhea); occult/gross rectal bleeding (0 = normal, 1 = occult+, 2 = occult++, 3 = occult+++, and 4 = gross bleeding); and mouse coat state (0 = normal, 2 = ruffed, and 4 = very ruffed). Mice were sacrificed at day 46. Colons were harvested and cleaned with PBS, and tumors were measured with a caliper. Colon sections were stored under conditions appropriate to the subsequent analyses (detailed in the following sections). For each mouse, three sections (~1 cm) were recovered from rectum samples and used for histology, gene expression analysis, and protein analysis. Colic contents were frozen in liquid nitrogen.

Mononuclear cells were isolated from spleens by gentle extrusion of the tissue through a 50-μm-mesh Nylon cell strainer (BD). Cells were resuspended in DMEM medium that was supplemented with 10% fetal calf serum, 2 mM l-glutamine, 50 U/mg penicillin, and 50 U/mg streptomycin (Lonza, Levallois-Perret, France). Erythrocytes were lysed with red blood cell lysing buffer (Sigma-Aldrich).

For stimulation experiments, 2.5 × 10⁶ cells per well were cultured for 48 h (37°C, 10% CO₂) in DMEM medium in P24 plates that were precoated with anti-CD3/CD28 antibodies (4 µg/ml each; eBioscience). Culture supernatant was frozen at −80°C until processing.

Proteins from each colon were extracted with T-PER tissue protein extraction reagent (ThermoFisher Scientific) using a Fastprep instrument at 4,500 g for 30 s (two cycles). Samples were centrifuged at 500 g for 1 min, and supernatants were harvested for cytokine analysis. A cytometric bead array system (LEGENDplex Mouse Th Cytokine Panel, Biolegend) was used, according to manufacturer’s instructions, to determine the levels of the following cytokines: IL-2, IL-4, IL-5, IL-6, IL-9, IL-10, IL-13, IL-17A, IL-17F, IL-21, IL-22, IFN-γ, and TNF-α.

Colon sections were formalin fixed and embedded in paraffin (4%, VWR, France). Epithelial proliferation was assessed by Ki67 staining according to the manufacturer’s instructions, using mouse monoclonal anti-Ki67 antibody (MM1, Leica Biosystems; 1:50). The proliferation index was determined by counting the number of Ki67-positive cells per crypt in three well-aligned crypts.

Two-micron colon sections were used for H&E staining. Histological score was determined using a BX43 Olympus microscope in a blinded manner, via the observation of three parameters: inflammation/cellular infiltration, epithelial lesions, and regeneration.

Colon sections were stored in RNA later (Ambion) at −80°C. RNAs were extracted using the RNeasy mini-kit (Qiagen, Courtaboeuf, France) following the manufacturer’s recommendations. RNA concentration was measured using a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA). cDNA synthesis was carried out from 1 µg of RNA using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems, USA), according to the manufacturer’s instructions. RT-qPCR was carried out in a reaction volume of 25 µl with Taqman probes (β-actin: Mm01963702_S1, caspase-9: Mm00516563_m1, caspase-7: Mm01195085_m1, Bik: Mm00476123_m1) (Life Technologies, France) according to the manufacturer’s instructions, using an ABI Prism 7700 (Applied Biosystems, USA) thermal cycler. To quantify and normalize the expression data, we used the ΔCt method, using the geometric mean Ct value from β-actin as the endogenous reference gene.

Total proteins were extracted from colon sections using T-PER buffer (Thermoscientific) and protease inhibitor mixture (Roche, Germany), through mechanical lysis with the Precellys homogenizer (Ozyme, France; 2 runs of 4,500 g for 30 s). Supernatants were collected, and western blots were performed. Briefly, samples were separated on a mini-PROTEAN TGX precast stain-free gel (4–20%, Biorad). Blots (Trans Blot Turbo Transfer system, Biorad) were incubated with β-actin, caspase-7, and cleaved caspase-7 antibodies (Cell Signaling, Danvers, MA, USA).

DNA was extracted using the Godon technique from stool (T0 = day of AOM treatment) and cecal content (Tf = time final; day of sacrifice) as described by Lamas et al. (27). The DNA pellet was washed with 70% ethanol, dried, and resuspended in 50 µl of Tris–EDTA buffer. DNA suspensions were stored at −20°C until amplification.

16S rDNA was amplified with primers for the V3 and V4 hypervariable regions (PCR1F_460: 5′-CTTTCCCTACACGACGCTCTTCCGATCTACGGRAGGCAGCAG-3′, PCR1R_460:5′-GGAGTTCAGACGTGTGCTCTTCCGATCTTACCAGGGTATCTAATCCT-3′). The reaction mixture contained 10 ng of genomic DNA, 5 U/μl MTP Taq DNA polymerase (Sigma, France), 0.2 mM dNTP, and 0.5 µM (final concentration) of each primer. Reactions were performed using an annealing temperature of 65°C for 30 cycles in a T100 thermocycler (Biorad, France). Sequencing was performed using 460-bp paired-end reads and an Illumina Miseq protocol on the GeT-PLaGe platform (Toulouse, France). Illumina reads were joined using the fastq-join method. The sequences were demultiplexed and quality filtered using the QIIME (version 1.8.0) software package. The sequences were assigned to OTUs using UCLUST algorithm 41 with a 97% threshold of pairwise identity and classified taxonomically using the Greengenes reference database.

Data were analyzed with Prism software (version 5). All normally distributed data were displayed as mean ± SEM. Comparisons between two groups were performed with a Student’s t-test.

To determine the potential beneficial effects of the dairy strain BL23 of L. casei on CRC onset, live bacteria were orally administered to mice treated with AOM and DSS. As shown in Figure 2A, mice fed with L. casei BL23 were protected against tumor development: no mouse in this group developed macroscopic tumors, compared to 67% (6/9) of mice that received only PBS (Figures 2A,B). Mice treated with PBS developed a per-mouse average of 2.3 tumors at least 2 mm in diameter (Figure 2C). Since this CRC model is related to chronic intestinal inflammation, we then assessed DAI and histological scores and intestinal epithelial damage. As shown in Figure 3A, DAI scores increased after each DSS cycle; however, there were no significant differences between any of the treated groups. For the histological scores, mice fed L. casei BL23 showed less damage than control mice did (p = 0.052, Student’s t-test; Figures 3B,C). In addition, Ki67 levels (which are expressed in proliferating cells) were significantly lower (p = 0.044, Student’s t-test) in mice treated with L. casei BL23 than in control mice (Figures 3D,E).

As immunomodulation and the induction of cell apoptosis are among the main probiotic-related protective mechanisms against CRC, we examined changes in both the immune response and apoptosis pathways due to L. casei BL23 treatment. We first determined the levels of both local (i.e., colon and mesenteric lymphoid node (MLN) samples) and systemic (i.e., spleen samples) cytokines that are involved in inflammation and carcinogenesis, including IL-22, IFN-γ, IL-10, IL-21, TNF-α, IL-6, and IL-17A (Figures S1–S3 in Supplementary Material). As shown in Figure 4A, colonic IL-22 (a cytokine that promotes proliferation of cancer cells) (28) levels were lower (p = 0.057, Student’s t-test) in L. casei BL23-treated mice compared to controls. Other cytokines that have been linked to CRC carcinogenesis, such as IFN- γ, TNF-α, IL-6, IL-10, IL-21, and IL-17A, were assessed in MLN (Figure S1 in Supplementary Material), in spleen (Figure S2 in Supplementary Material), and in colon samples (Figure S3 in Supplementary Material), but no significant difference was observed between treated and control mice.

In addition, we performed colic gene expression analysis of genes involved in apoptosis, specifically, caspase-7, caspase-9, and Bik. Our results reveal that L. casei BL23 induced a significant increase compared to controls in the expression of the executioner caspase-7 (Figure 4B, p = 0.017, Student’s t-test) and the initiator caspase-9 (Figure 4C, p = 0.028, Student’s t-test), together with an increase of the apoptotic gene Bik (Figure 4D, p = 0.082, Student’s t-test). Finally, we determined the level of truncated caspase-7 (which corresponds to the active form of caspase-7 in apoptosis). For this, we selected two mice treated with L. casei BL23 and protected against tumor development, which presented high caspase-7 RNA expression levels, and a mouse treated with PBS (not protected against tumors), which presented low levels of caspase-7 RNA expression. As shown in Figure 4E, western blot results confirmed that the mice protected against tumor onset produced higher levels of the active form of caspase-7 compared to the non-protected mouse.

Altogether, these data suggest that L. casei BL23 has an antiproliferative and apoptotic effect in this CRC model; a detailed proposal of the mechanisms of action of L. casei BL23 against CRC is shown in Figure 5.

To determine the impact of L. casei BL23 on the gut microbiota, we analyzed microbiota richness and diversity after BL23 oral treatment. This analysis provided a total of 1,709,762 high-quality and classifiable reads, with an average of 5,000 (n = 51) reads per sample. First, beta diversity (Bray-Curtis distance) was analyzed in microbial samples using principal components analysis, which reduced the dimensionality of the data set. While no grouping was observed before AOM challenge (T0; Figure 6A), the microbiota of mice treated with L. casei BL23 tended to diverge from those of PBS-treated animals at Tf (sacrifice; Figure 6B, p = 0.1, Anosim). Then, we estimated community richness (alpha diversity; Shannon and Simpson indexes) at T0 and Tf. As shown in Figure 6C, AOM injection resulted in a reduction in alpha diversity with respect to that found in the group treated with PBS but not AOM (Shannon index, 6.4 ± 0.22 in PBS/AOM-treated group versus 6.8 ± 0.46 in PBS/no AOM group, p = 0.028, Student’s t-test). Despite the fact that the effects of L. casei strain BL23 were not statistically significant (Simpson index, 483.3 ± 111.8 in BL23/AOM-treated group versus 436.8 ± 54.7 in PBS/AOM-treated group, ns, Student’s t-test; Figure 6D), these results provide intriguing clues about this strain’s interactions with other members of the gut microbiota.

Finally, bacterial communities were characterized at the phylum level (Figure 6E). Firmicutes was the dominant phylum in the L. casei BL23/AOM treated group (51 ± 10%), with Bacteroidetes ranked second (45 ± 10%). In contrast, Bacteroidetes was the most abundant phylum in both PBS/AOM and PBS/no AOM treated groups (52 ± 12 and 53 ± 13%, respectively), followed by Firmicutes (44 ± 12 and 43 ± 13%, respectively). The third most abundant phylum in all groups was Proteobacteria, with approximately 3–4% of reads. By using linear discriminant analysis (LDA) coupled with effect size measurements (linear discriminant analysis effect size), we found that Prevotella, Ruminococcaceae, and Lactobacillus were the key groups that were overrepresented in the L. casei BL23-treated mice (Table 1). At the species level, only Lactobacillus zeae was significantly more abundant (p = 0.004, Student’s t-test) in these mice, with a LDA score of 2.58. This was not surprising, since this species could in fact correspond to strain BL23 of L. casei. Indeed, 16S RNA analysis has revealed a close relationship (99% similarity) between our focal strain and L. zeae (26).