LAB strains L33 and L125 (Pavli et al., 2016) and LGG (DMSZ, Braunschweig, Germany) were cultivated in de Man, Rogosa and Sharp (MRS, Applichem, Darmstadt, Germany) broth at 37°C, under anaerobic conditions. S. aureus and E. coli were obtained from the microbial collection of the Laboratory of Clinical Microbiology, University Hospital of Alexandroupolis and were routinely cultured in Tryptic Soy Broth (TSB, Condalab, Madrid, Spain) at 37°C under aerobic conditions.

HT-29 human colorectal adenocarcinoma cell line (ATCC, Manassas, VA, United States) was maintained in Roswell Park Memorial Institute GlutaMAX^™ (RPMI)-1640 medium supplemented with 10% fetal bovine serum (FBS), 100 μg/mL streptomycin and 100 U/mL penicillin (all from Thermo Fisher Scientific, Waltham, MA, United States). For HT-29-bacteria co-incubations, a modified cell culture medium consisted of 10% FBS and 20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES; Thermo Fisher Scientific) was used. Cells were incubated in a humidified atmosphere at 37°C, 5% CO₂ under sterile conditions.

The cytotoxic capacity of pathogens was initially evaluated using the Sulforhodamine B (SRB) (Invitrogen, Waltham, MA, United States) assay. HT-29 cells were seeded at a density of 7,500 cells per well in a 96-well plate (SPL Life Sciences, Pochon, South Korea) and incubated O/N in standard cell culture conditions. The following day, cells were washed with phosphate-buffered saline (PBS, Thermo Fischer Scientific) and incubated with viable pathogens or conditioned media (CM). For viable cell treatments, pathogens were added at a concentration of 10⁷ CFU/mL for 1, 2 or 4 h. CM was prepared as follows: pathogens were incubated in RPMI-1640 medium supplemented with 10% FBS and 20 mM HEPES for 24 h. Cell pellets were discarded after centrifugation at 8,000 g for 5 min, and the supernatants were sterile filtered using a 0.2 μm filter (Merck, Rahway, NJ, United States). Cells were then treated with pathogen CM for 2 or 4 h. HT-29 cell viability was determined using a previously published protocol (Chondrou et al., 2018). Cell survival (%) is expressed as: [(Sample OD₅₇₀ – Blank OD₅₇₀)/(Control mean OD₅₇₀ – Blank OD₅₇₀)] × 100. Cells treated with standard culture medium served as the untreated control.

The cytotoxic capacity of pathogen CM was also evaluated using the HoloMonitor Live Imaging System (PHI, Boston, MA, United States). In this case, 2 × 10⁵ cells per well were seeded into 24-well plates. The following day, cells were treated with undiluted, sterile-filtered pathogen CM. Cells treated with cell culture medium were included as an untreated control. Cell proliferation was assessed over a 24 h period with the Kinetic Cell Proliferation Assay software (PHI).

The potential protective effects of viable lactobacilli were evaluated against pathogen-induced toxicity, under two different conditions: (i) co-treatment of cells with lactobacilli (10⁷ CFU/mL) and pathogens (10⁸ CFU/mL), (ii) pre-treatment (2 or 4 h) of cells with lactobacilli (10⁷ CFU/mL) prior to the addition of pathogens (10⁸ CFU/mL). Pathogen challenge lasted for 4 or 2 h for S. aureus and E. coli, respectively. Cell survival was calculated as described above. Untreated cells served as a control, while cells exposed only to pathogens were used as a positive control.

Propidium iodide (PI) staining, and flow cytometry were used to examine membrane permeability and cell death. HT-29 cells were seeded at a density of 2 × 10⁵ cells per well in 6-well plates. The following day, cells were pre-treated with lactobacilli (10⁷ CFU/mL) for 4 h after which S. aureus (10⁸ CFU/mL for 4 h) or E. coli (10⁸ CFU/mL for 2 h) were added. After incubation, cells were collected by trypsinization and were centrifuged at 600 g for 5 min. Then, cell pellets were washed twice with PBS, and stained with 50 μg/mL PI (Biotium, Fremont, CA, United States) for 3 min. Untreated cells served as a negative control, and pathogen-only treated cells as a positive control. Flow cytometry was performed using the Attune NxT Flow Cytometer (Thermo Fischer Scientific) and results were analyzed with FlowJo V10 software (BD Biosciences, San Jose, CA, United States).

Cell culture inserts were used to determine whether direct contact between L125 and HT-29 cells is required for protection against pathogen-induced cytotoxicity. Specifically, HT-29 cells were seeded at a density of 5 × 10⁴ cells per well into 24-well plates. The following day, L125 cells (10⁷ CFU/mL) were added to either side of a 0.4 μm trans-well insert (SPL Life Sciences) for 4 h. Then, pathogens were inoculated in the lower compartment, in direct contact with HT-29 cells. Cell survival was determined using the SRB assay, after 4 h challenge with S. aureus or 2 h challenge with E. coli.

HT-29 were seeded in 24-well plates at a density of 4 × 10⁵ cells per well and incubated until reaching 100% confluency, following a previously published protocol (Plessas et al., 2020). Lactobacilli at a concentration of 10⁷ CFU/mL were then added for 2 or 4 h. After incubation, the monolayers were washed twice with sterile PBS, and cells were detached using 1× Trypsin (Thermo Fischer Scientific). The resulting cell suspensions were serially diluted in Ringer’s solution and plated onto MRS agar for enumeration. Plates were incubated at 37°C until visible colonies formed. Attached bacteria are expressed as Log CFU/mL. Lactobacilli-HT-29 interactions were further visualized using an inverted microscope (ZEISS, Jena, Germany).

For the competitive exclusion assay, HT-29 cells were seeded into 24-well plates (4 × 10⁵ cells/well). The following day, cells were pretreated with lactobacilli for 4 h prior to pathogen challenge. Then, pathogens (10⁸ CFU/mL) were added and incubated for 2 h. At the end of the incubation period, the monolayers were washed twice with sterile PBS and cells were disassociated with 1× Trypsin. The suspension was serially diluted in Ringer’s solution and plated onto TSA for colony enumeration of pathogens. The plates were incubated at 37°C, under aerobic conditions, until visible colonies formed. Attached bacteria were expressed as Log CFU/mL. Controls included pathogen-only treated cells and untreated cells to account for potential contamination.

HT-29 cells were seeded in 24-well plates at a density of 5 × 10⁴ cells per well. The following day, cells were pre-treated with lactobacilli at a concentration of 10⁷ CFU/mL for 4 h. Pathogens were, then, added at a concentration of 10⁸ CFU/mL for 1 h. Monolayers were washed twice with PBS and incubated in RPMI medium supplemented with 10% FBS and 100 μg/mL gentamicin (all from Thermo Fischer Scientific) for 1 h. Cells were then lysed using 1% (v/v) Triton-X (Thermo Fischer Scientific), the resulting suspension was serially diluted in Ringer’s solution and plated onto TSA plates for colony enumeration of internalized pathogens. Plates were incubated at 37°C, under aerobic conditions, until visible colonies formed. Internalized bacteria were expressed as Log CFU/mL.

To examine global changes in the expression of genes in L125 and HT-29 cells during co-incubation, dual RNA-seq experiments were performed. Specifically, HT-29 cells were seeded at a density of 8 × 10⁵ cells in 100 mm plates and incubated O/N. The following day, L125 cells were added at a concentration of 10⁷ CFU/mL for 4 h. HT-29 cells treated with cell culture medium were utilized as non-treated control samples. After treatments, plates were washed with PBS, and cells and attached bacteria were collected by trypsinization. Samples were centrifuged at 8,000 g for 5 min and incubated in a lysis buffer containing 1 M Τris-Cl pH 8, 0.5 M EDTA pH 8, 0.1% Triton-X and 100 mg/mL lysozyme, at 37°C for 30 min to maximize the efficacy of RNA isolation from the Gram-positive bacterium L125. Trizol reagent (Sigma-Aldrich, Saint Louis, MO, United States) was finally added, and RNA isolation was performed following manufacturer’s instructions. RNA quality was determined via agarose gel electrophoresis and spectrophotometrically (Thermo Fisher Scientific, NanoDrop 1000 Spectrophotometer). RNA concentration was measured with Qubit according to the Qubit^™ RNA Broad Range protocol (Thermo Fisher Scientific). Dual RNA-seq (metatranscriptomics) was performed with Illumina NovaSeq X Plus (Strategy PE150). Bioinformatic analysis was performed on the Galaxy server (Abueg et al., 2024). The quality of the obtained reads was determined with FASTQC (v0.74; Andrews, 2010), and low-quality reads were discarded with Trimmomatic (v0.39; Bolger et al., 2014). HISAT2 (v2.2.1; Kim et al., 2019) was used for read assembly against the L125 or human (GRCh38/hg38) genomes. Gene counts and lengths were determined with featureCounts (v2.0.3; Liao et al., 2014) and differential expression with Limma (v3.25.1; Ritchie et al., 2015) and DESeq2 (v1.1.0; Love et al., 2014). Goseq (v1.50.0; Young et al., 2010) and EGSEA (v1.20.0; Alhamdoosh et al., 2017) were used to classify the genes into KEGG pathways and functional categories and for gene ontology analysis, respectively. Transcripts per million (TPM) were calculated for L125 transcripts with StringTie (v2.1.5; Pertea et al., 2015). Based on TPM values expression was characterized as high (>1,000 TPM), medium (10 < TPM < 1,000), low (0.5 < TPM < 10) or below cut-off (TPM <0.5).¹

The WGS of L125 (Accession number: JAKJPP000000000.1) was searched for S-layer proteins (SLAPS) and proteins containing cell-surface associated domains (LysM, SH3, WxL) and motifs (LPxTG and LPxTG-like) using EggNOG (Huerta-Cepas et al., 2019) and InterPro (Blum et al., 2021). Proteins containing signals for translocation to the bacterial cell surface were predicted using SignalP 6.0 (Teufel et al., 2022). Their localization and topology were also validated with PSORTb (v3.0; Yu et al., 2010) and DeepTMHMM (Hallgren et al., 2022), respectively. Putative surface proteins were selected for further analysis based on the presence of motifs and domains implicated in interactions with host cells and ECM, including mucin-binding domains (MuBPs), fibronectin-binding domains (FnBPs), collagen-binding domains (CnBPs) or proteins containing Ig-like folds and leucine-rich repeats (LRRs). Furthermore, known moonlighting proteins (Jeffery, 2019) were also included. Sequence homology between the annotated cell surface proteins and proteins encoded by S. aureus and E. coli was examined using Blastp (Camacho et al., 2009).

The structure of the putative cell-surface associated proteins containing adhesin-related domains and motifs was predicted with ColabFold (Mirdita et al., 2022). Protein structures of human receptors and proteins participating in the interactions were downloaded from PDB or AlphaFold and used for docking experiments. These include: human fibrinogen chain A (FnA, 1BBR), fibrinogen fragment D (FnB, 1FZA), fibrinogen chain G (FnG, 1DUG), mucin 1 (MUC1, 1SM3), mucin 3b (MUC3b, AF-Q9H195), mucin 4 (MUC4, AF-Q99102), mucin 7 (MUC7, AF-Q8TAX7), mucin 13 (MUC13, AF-Q9H3R2), mucin 16 (MUC16, 7SA9), Toll-like receptor 2 (TLR2, AF-O606063), Toll-like receptor 4 (TLR4, AF-O00206), Toll-like receptor 5 (TLR5, AF-O60602), Toll-like receptor 6 (TLR6, AF-Q6Y2C9), carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1, AF-P13688), carcinoembryonic antigen-related cell adhesion molecule 7 (CEACAM7, Q14002), cadherin 1 (CDH1, AF-P12830) and collagen alpha-1(I) chain (COL1A1, AF-P02452). Docking experiments were performed on the HDOCK server (Yan et al., 2020).

High throughput determination of cytokines produced by HT-29 cells after co-incubation with L125 was performed using the Proteome Profiler Human Cytokine Array Kit (R&D Systems, Minneapolis, MN, United States). HT-29 cells were treated with the LAB strain at a concentration of 10⁷ CFU/mL for 4 h. Then, culture supernatants were collected, centrifuged at 11,000 g for 5 min and stored at −80°C until further analysis. Supernatants from three independent experiments were pooled and analyzed using the protein microarray, following manufacturer’s instructions. Briefly, the membranes spotted with antibodies against 36 different cytokines, chemokines and acute phase proteins were blocked for 1 h while cell culture supernatants were mixed and incubated with the detection antibody cocktail for 1 h at RT. Then, the membranes were incubated with the sample/antibody mixtures overnight at 4°C on a rocking platform. The following day, membranes were washed three times, incubated with Streptavidin-HRP for 30 min at RT and washed again three times prior to chemiluminescence detection. Membranes were imaged using the ChemiDoc Imaging System (Bio-Rad, Hercules, CA, United States). Mean pixel density was calculated using a protein array analyzer plugin for ImageJ (National Institute of Health, United States).

For the statistical analysis of the experimental data, Student’s t-test was performed using GraphPad PRISM 9 (GraphPad Software Inc., CA, United States). All experiments were performed in triplicate unless otherwise stated. Results are represented as mean ± standard deviation. A p-value <0.05 was considered statistically significant.

First, we sought to determine the virulence effects of clinically isolated S. aureus and E. coli strains by their ability to induce cytotoxic effects on the colon adenocarcinoma cell line HT-29 at 1, 2 and 4 h (Figure 1). As shown in Figures 1A,C, a time-dependent reduction in cell viability was recorded for both pathogens. After 4 h-treatment of HT-29 cells with S. aureus or 2 h-treatment with E. coli, cell viability was reduced by >80% (p < 0.005). These timepoints were selected for subsequent experiments. These findings were further validated by PI staining and flow cytometry (Figures 1B,D). The effect of pathogen-derived CM on cell viability after 4 h (S. aureus, SACM) or 2 h (E. coli, EC CM), was initially investigated with the SRB cytotoxicity assay. A smaller reduction (p < 0.005) in cell survival was recorded (Figure 1E), compared to treatments with viable pathogens. These effects were further monitored for 24 h with the HoloMonitor live cell imaging system, using the Kinetic Cell Proliferation Assay software. It was shown that CM derived from either pathogen reduced cell confluency in a time dependent manner (Figure 1F).

The ability of the potential probiotic LAB strains to limit pathogen-induced cell death was then assessed using various experimental designs (Figures 2A,3A). Initially, cells were co-incubated with lactobacilli and pathogens for 4 h (S. aureus-treated cells) or 2 h (E. coli-treated cells), and viability was determined with the SRB assay. No difference in cell survival (%) was observed between pathogen-only treated cells and those treated with any of the three lactobacilli (Figures 2B,C). Then, we investigated whether lactobacilli pretreatment could prevent pathogen-induced cell death (Figure 3A). As shown in Figure 3B, L125 and LGG significantly limited cell death, while L33 had no significant effect. These results were also confirmed through flow cytometry analysis (Figure 3C). It should be noted that lactobacilli treatment did not affect cell survival (Supplementary Figure S1).

A trans-well cell culture system was employed to determine whether direct contact of L125 with the epithelial cells is necessary for protection against pathogen-induced cytotoxicity (Figure 4A). Indeed, direct contact for 4 h prior to the addition of the pathogens resulted in a 2-fold reduction in cell death (p < 0.05). However, pretreatment with L125 in the upper compartment of the trans-well system, where no direct contact occurred, did not confer any protective effect (Figure 4B).

To further explore the protective phenotype, we measured the capacity of L125 to adhere to HT-29 cells and limit pathogen attachment and invasion, using established microbiological assays. As shown in Figure 5A, L125 and LGG exhibited strong binding to HT-29 cells, while L33 demonstrated markedly lower adherence capacity (p < 0.05). Notably, L125 formed aggregates and adopted a net-like morphology over the HT-29 cells (Figure 5B). Pretreatment with L125 or LGG for 4 h reduced the attachment of S. aureus (p < 0.05), having no effect against E. coli (Figure 5C). L33 did not limit the adhesion ability of either pathogen, but did limit internalized S. aureus counts. L125 significantly inhibited the invasion of both pathogens in HT-29 cells, reducing counts by more than 1.5 Log CFU/mL (Figure 5D). A slight reduction in pathogen viability was also recorded at the same timepoints (Supplementary Figure S2).

Dual RNA seq was utilized to analyze the gene expression changes in L125 and HT-29 cells during a 4 h co-incubation period. Specifically, for L125, a total of 108 genes exhibited high expression levels (>1,000 TPM), 2.568 genes displayed medium expression levels (10 < TPM < 1,000), and 497 genes showed low expression levels (0.5 < TPM < 10). Additionally, 68 genes were not expressed in neither of the two independent experiments (TPM: 0). The 108 highly expressed genes clustered into 14 KEGG functional categories, 23 KEGG pathways (Figure 6) and 16 clusters of orthologous groups (COGs) (Table 1). The most represented KEGG functional category was “genetic information processing” with “translation” being most prominent KEGG pathway and COG category.

To identify transcriptional changes in cell-surface exposed proteins, the WGS of L125 was, firstly, re-annotated in silico. Genome mining was performed to identify proteins potentially involved in the competitive exclusion phenotype, and the strain’s ability to limit pathogen internalization. By filtering the EggNOG output, 77 proteins containing motifs and domains indicative of cell-wall exposure were identified (Supplementary Table S1). Specifically, these proteins included: 17 proteins carrying the WxL domain, 19 proteins with a Gram-positive anchor (i.e., LPxTG or LPxTG-like motifs), 2 proteins with an S-layer homology (SLH) domain and 7 proteins with a SlaP (S-layer assembly protein) domain, 13 proteins containing the LysM domain and 9 proteins with the SH3 domain. Among these proteins: 7 possess Ig-like motifs (Big_2 and Big_3 groups), 5 carry LRR motifs, 3 carry collagen-binding domains (Cna_B), 13 contain mucin-binding domains (MucBP), 2 possess a fibronectin-binding domain (FbpA), 2 carry a FIVAR domain, and 1 has a domain belonging to the concanavalin A-like lectin/glucanases superfamily. Additionally, a protein belonging to the invasin/intimin bacterial adhesion mediator protein family was annotated. Out of these, 29 proteins were predicted to have extracellular localization or be exposed on the cell wall (Table 2), 15 were predicted to be localized to the cytoplasmic membrane, 6 to the cytoplasm, and 31 proteins to have unknown or multiple localization sites. Extracellular and cell wall proteins were selected for further analysis. Most predicted surface proteins were found to utilize a Sec/SPI signal recognized by signal peptidase I (Table 2), and one protein (LP125_RS11350) carries a YSIRK processing signal for transport outside the cell. Finally, 16 moonlighting proteins with adhesin function were identified in the L125 genome (Table 3).

To investigate possible exclusion events, the homology of L125 surface exposed proteins to those encoded by S. aureus or E. coli was analyzed in silico (Supplementary Tables S2, S3). Several proteins encoded by L125 contain domains with high alignment scores to those found in the two pathogens. Specifically, LP125_RS10075, LP125_RS11095 contain LPXTG anchor domains that show homology to those present in S. aureus (HDP6314765.1, HDM8613333.1), LP125_RS11095, LP125_RS11335, LP125_RS01150 and LP125_RS06780 carry MucBP domains that present high similarity to domains encoded by S. aureus (HDM8613333.1, EZX22480.1, MDU3924263.1). Furthermore, LRR repeats are contained in proteins encoded by L. plantarum L125 (LP125_RS02875, LP125_RS02875) and a BspA family leucine-rich repeat surface protein encoded by S. aureus (MDF4035866.1). Accordingly, LRR repeats found in L125 protein LP125_RS11350 share similarity to BspA family leucine-rich repeat surface protein (MBL0960876.1) and a bacterial Ig-like domain-containing protein (WP_142456104.1), respectively. Accordingly, LP125_RS11385, LP125_RS01150, LP125_RS07170 contain MucBP domains with similarity to MucBP domain-containing proteins encoded by E. coli (MBC8929979.1, MBC8921713.1). Furthermore, two proteins (LP125_RS01490 and LP125_RS09530) exhibited similarity to collagen-binding proteins (PPI97406.1, HCC5232179.1). The majority of moonlighting proteins annotated in the genome of L125, except for maltose phosphorylase (MapA, LP125_RS10155 and LP125_RS06445), share high sequence similarity with proteins encoded by S. aureus. On the contrary, L125 proteins generally exhibit lower query cover and/or similarity with E. coli proteins (Table 4).

Concerning the expression at the transcriptional level, most of the putative cell surface proteins exhibited medium expression levels (100 < TPM < 1,000), with the exception of two extracellular proteins LP125_RS14880 and LP125_RS09840, which had high expression levels (>1,000 TPMs). No transcripts were detected for LP125_RS11935 and very low expression levels were recorded for LP125_RS08445 (TPM < 0.5) (Table 3). As expected, several of the moonlighting adhesins exhibited particularly high expression, reflecting their crucial role in bacterial homeostasis (Table 3). The highest expression levels were recorded for EF-Tu (LP125_RS08655), GAPDH (LP125_RS02930), and Hsp60 (LP125_RS03215). Notably, no transcripts were measured for ornithine carbamoyltransferase (LP125_RS10650).

The ability of the annotated, putative cell-exposed proteins to interact with host proteins was determined with in silico docking experiments. To this end, docking partners were determined based on the presence of domains and motifs involved in host–microbe interactions (Table 2). As shown in Figures 7, 8 proteins from L125 demonstrated the lowest docking scores (<−300) when interacting with host receptors and ECM components indicating strong interaction potential. More specifically, the LRR or Ig-like domains contained in LP125_RS10075, LP125_RS11335, LP125_RS12380 and LP125_RS04395 interact with CEACAM1 with a high confidence score (>95%). Accordingly, LP125_RS10075 and LP125_RS11350 interact with FnB (confidence score >95%) and LP125_RS12380 with FnA (confidence score 95.3). Furthermore, the MucBP and LRR domains present in LP125_RS11350 interact with MUC1, 4 and MUC16, as well as with LRR domains present in TLR2, 4, 5, 6 with a confidence score >97%. Additionally, the Ig-like folds of LP125_RS07170 demonstrated interaction with MUC4, TLR4 and 5 with a confidence score >94%. Interestingly, the highly expressed LP125_RS14880 and LP125_RS09840 proteins, which do not contain specific adhesin-related domains or motifs, may still contact extracellular domains of TLR4 and −5. Moreover, LP125_RS09840 may also form interactions with MUC1 and LP125_RS14880 with MUC3b and 7, as indicated by the low docking scores and high confidence scores (Figures 7A,B). Protein–protein interactions of partners with the lowest free energy are presented in Figure 7C.

RNA seq was utilized to explore the capacity of L125 to prime antimicrobial responses in the host cell. As shown in Figure 8, treatment of cells with L125 for 4 h led to the upregulation of 600 genes and the downregulation of 1,137 genes in HT-29 cells, while 26,658 genes remained unaffected. The top over-represented categories in “biological function” and in “molecular function” are visualized in Figures 8C,D. Interestingly, L125 did not significantly modulate immune-related pathways, including TLR- or NOD- signaling cascades, nor did it affect the production of cytokines and chemokines (Figure 9A). At the protein level, L125 was shown to limit the secretion of immunological markers in pooled cell culture supernatants, collected from three independent experiments (Figures 9B–D). Pathway analysis revealed that L125 primarily exerted inhibitory effects (Supplementary Table S4), negatively impacting bacterial invasion (Figure 10), adherens junction (Supplementary Figure S3), and endocytosis (Supplementary Figure S4). These phenotypes are mediated by central processes related to the regulation of actin cytoskeleton, internalization and vacuole formation. In this context, genes involved on the mitotic spindle assembly pathway and G2/M progression were also downregulated, further emphasizing L125’s regulatory role in cellular processes.