Lactobacillus delbrueckii subsp. bulgaricus 2038 and Streptococcus thermophilus 1131 suppress polystyrene nanoplastic transcellular permeability and internalization by intestinal epithelial cells

TL;DR

This study shows that two yogurt bacteria strains can reduce the entry of harmful nanoplastics into intestinal cells, potentially protecting human health.

Contribution

The study identifies specific yogurt strains that suppress nanoplastic internalization and transcellular permeability in intestinal cells.

Findings

L. bulgaricus 2038 and S. thermophilus 1131 significantly reduced PSNP internalization in Caco-2 cells.

Both strains suppressed PSNP transcellular permeability, even when non-viable.

The PSNP suppression effect was strain-dependent, with these strains being the most potent in their species.

Abstract

The plastic is broken into nanoscale particles (nanoplastics) that are harmful to the human body. Nanoplastics orally ingested are internalized into various cells, causing adverse effects such as oxidative stress and apoptosis; however, methods for preventing nanoplastic internalization are lacking. By elucidating this method, it is possible to make a significant contribution to human health. Then, following previous reports that Lactobacillus delbrueckii subsp. bulgaricus 2038 and Streptococcus thermophilus 1131 ameliorated barrier dysfunction in the small intestine, we examined their effects on the internalization of polystyrene nanoplastics (PSNPs) in a differentiated Caco-2 monolayer by flow cytometry and immunofluorescent staining. Both yogurt starter strains significantly suppressed fluorescently labeled PSNP internalization by Caco-2 cells, even when the strains were non-viable. Moreover, both strains significantly suppressed the transcellular permeability of PSNPs, suggesting reduced PSNP accumulation in blood vessels and various tissues. Finally, the suppression of PSNP internalization was strain dependent, and both strains were the most potent strains among multiple strains from the same species. This study highlights the potential of L. bulgaricus 2038 and S. thermophilus 1131 for ameliorating the adverse health effects of PSNP intake.

The online version contains supplementary material available at 10.1038/s41598-026-39631-z.

Figures

Characteristics of polystyrene nanoplastics (PSNPs). (a) PSNP morphology was observed by transmission electron microscopy (scale bar = 100 nm). (b) PSNP size distribution was analyzed by dynamic light scattering (*n* = 3).

Polystyrene nanoplastics (PSNPs) were internalized by Caco-2 cells via macropinocytosis and clathrin-mediated endocytosis. A Caco-2 monolayer was incubated for 18 h in the presence of PSNPs. (a) E-cadherin was visualized by immunofluorescent staining. Internalized PSNPs were observed by confocal laser-scanning microscopy (scale bar = 10 μm, green: PSNP, red: E-cadherin). Arrow heads indicate internalized PSNPs. (b–d) To clarify the internalization pathway, endocytosis inhibitors were added to the medium (*n* = 3–5/group). EIPA, Dynasore, and MβCD concentrations were 0.5 µM, 1 µM, and 100 µM, r

Polystyrene nanoplastics (PSNPs), *Lactobacillus delbrueckii* subsp. *bulgaricus* 2038, and *Streptococcus thermophilus* 1131 affected the expression of various gene sets. Gene set enrichment analysis was performed. Enrichment plots displayed for the gene sets of (a–c) WIKIPATHWAY_OXIDATIVE_STRESS_RESPONSE, (d–f) REACTOME_BASE_EXCISION_REPAIR, and (g–i) REACTOME_GLUCRONIDATION. Comparisons are shown between (a, d, g) the control group and PSNP group, (b, e, h) the PSNP group and PSNP + *L*. *bulgaricus* 2038 group, and (c, f, i) the PSNP group and PSNP + *S. thermophilus* 1131 group.

The suppression of polystyrene nanoplastics (PSNPs) internalization was strain dependent. PSNPs and heat-treated (a) *L. bulgaricus* or (b) *S. thermophilus* were added to the Caco-2 monolayer. After culturing for 18 h, PSNP internalization was quantified by flow cytometry (*n* = 3/group). MFI, median fluorescent intensity.

Full text

Introduction

Amid massive global plastic production, approximately 12 billion metric tons of plastic are predicted to be discarded into landfills and the natural environment by 2050^1^. Plastics are degraded and fragmented into microplastics (MPs; diameter < 5 mm) and nanoplastics (NPs; diameter < 100 nm) through various mechanisms such as mechanical force and ultraviolet radiation^2,3^. Humans then ingest MPs and NPs orally through a variety of routes, including the consumption of marine organisms, tap water, and beverages from plastic bottles^4–6^. Previous reports estimated that human consumes 39,000–52,000 MP particles each year^5^ or 5 g of plastic per week^7^, with clinical studies detecting plastic in human saliva^8,9^ and stool^10^.

Although some ingested MPs and NPs are excreted, some are absorbed into the body^11^. NPs can even be endocytosed into cells, including HeLa (cervical cancer), A549 (lung carcinoma), 1321N1 (brain astrocytoma), and Caco-2 (intestinal epithelial) cells^12–15^. Previously reported adverse effects associated with the cellular uptake of NPs include polystyrene (PS) NPs damaging mitochondria and inducing apoptosis by accumulating in lysosomes in human astrocytoma cells^16^. Moreover, NPs endocytosed in mouse embryonic fibroblasts caused oxidative and inflammatory stress^17^, whereas PSNPs endocytosed in Caco-2 cells induced a disruption of tight junctions^15^. Despite accumulating evidence regarding the health hazards, the full extent of NP endocytosis remains unknown, and no methods have yet been proposed to control NP internalization. Thus, providing a solution to NP internalization has substantial benefits to human health.

Lactic acid bacteria (LABs) are Gram-positive bacteria with thick peptidoglycan layers in their cell walls and mainly produce lactic acid from carbohydrate^18^. LABs have been widely used as safe food ingredients, and some of them play a role as probiotics which provide benefits for intestinal health^19,20^. Lactobacillus delbrueckii subsp. bulgaricus 2038 and Streptococcus thermophilus 1131, which are used as starter strains for commercially available yogurt, also contribute to small intestinal homeostasis. They induced the expression of antimicrobial peptides in the small intestine^21^. In both strains, this mechanism involves stimulation of immune cells in the lamina propria^22^. Moreover, in vitro experiments using Caco-2 cells and human induced pluripotent stem cell-derived small intestinal epithelial cells showed that both strains ameliorated physical barrier destruction caused by tumor necrosis factor alpha and interferon gamma^23,24^. Based on the above, we hypothesize that both strains can protect against the adverse effects of NP exposure. In this study, because the small intestine is frequently exposed to NPs, we evaluated the effects of L. bulgaricus 2038 and S. thermophilus 1131 on PSNP internalization by differentiated Caco-2 cells.

Results

PSNP characteristics

PSNP morphology was observed by transmission electron microscopy. The resulting images showed that PSNPs were spherical and partially aggregated (Fig. 1a). Dynamic light scattering analyses revealed the hydrodynamic size and zeta potential of PSNPs. PSNPs exhibited a unimodal size distribution and mean diameter: 48.8 ± 0.1 (Fig. 1b), suggesting size variability for single PSNPs owing to the aggregation of multiple NPs. PSNPs exhibited a negative charge with a zeta potential of −25.63 ± 0.15.

Fig. 1. Characteristics of polystyrene nanoplastics (PSNPs). (a) PSNP morphology was observed by transmission electron microscopy (scale bar = 100 nm). (b) PSNP size distribution was analyzed by dynamic light scattering (n = 3).

Pathway of PSNP internalization in Caco-2 cells

We examined whether PSNPs were internalized in Caco-2 cells by first assessing the effects of different PSNP amounts (2, 20, and 200 µg/mL) on cell viability and cytotoxicity using WST-1 and LDH assays, respectively. The results showed that PSNPs did not influence the cell conditions (Supplementary Fig. S1). Caco-2 cells were exposed to PSNPs for 18 h and observed using a confocal laser-scanning microscope. The results showed that PSNPs were internalized in Caco-2 cells and mainly located around the bottom of the cells, suggesting that PSNPs were internalized prior to 18 h then accumulated in the cells (Fig. 2a).

Fig. 2. Polystyrene nanoplastics (PSNPs) were internalized by Caco-2 cells via macropinocytosis and clathrin-mediated endocytosis. A Caco-2 monolayer was incubated for 18 h in the presence of PSNPs. (a) E-cadherin was visualized by immunofluorescent staining. Internalized PSNPs were observed by confocal laser-scanning microscopy (scale bar = 10 μm, green: PSNP, red: E-cadherin). Arrow heads indicate internalized PSNPs. (b–d) To clarify the internalization pathway, endocytosis inhibitors were added to the medium (n = 3–5/group). EIPA, Dynasore, and MβCD concentrations were 0.5 µM, 1 µM, and 100 µM, respectively. Statistical comparisons were performed using Welch’s t-test. * P < 0.05. MFI, median fluorescent intensity. (e–i) Inhibition of PSNP internalization by endocytosis inhibitors was observed using confocal laser-scanning microscopy (scale bar = 10 μm, green: PSNP, red: E-cadherin).

We hypothesized that PSNPs were internalized in Caco-2 cells via endocytosis. Therefore, we investigated the endocytosis pathway of PSNPs using inhibitors of the major types of endocytosis: clathrin-mediated (Dynasore), caveolae-mediated endocytosis (methyl-β-cyclodextrin; MβCD), and macropinocytosis (5-N-ethyl-N-isopropyl amiloride; EIPA). We examined the effects of these inhibitors on cell viability and cytotoxicity to determine the appropriate concentration for subsequent experiments (Supplementary Fig. S2). Next, based on the preliminary experiments, we evaluated the effects of endocytosis inhibitors on PSNP internalization. Flow cytometry analysis revealed that EIPA and Dynasore significantly suppressed PSNP internalization, whereas MβCD significantly increased PSNP internalization (P = 0.01, P = 0.02, P = 0.03, Fig. 2b–d). Similar results were observed in confocal laser-scanning microscope images (Fig. 2e–i).

Effects of L. bulgaricus 2038 and S. thermophilus 1131 on PSNP internalization

To investigate whether L. bulgaricus 2038 and S. thermophilus 1131 suppress PSNP internalization, live LABs and PSNPs were added simultaneously to the Caco-2 monolayer, and PSNP intensity in the cells was measured by flow cytometry following incubation for 18 h. The results showed that L. bulgaricus 2038 (P = 0.002) and S. thermophilus 1131 (P < 0.001) significantly suppressed PSNP internalization (Fig. 3a). Similar results were observed in confocal laser-scanning microscope images (Fig. 3b–d). In addition, to clarify whether the effects of live L. bulgaricus 2038 and S. thermophilus 1131 were due to the bacterial components or their metabolites, we evaluated the effects of heat-treated LABs on PSNP internalization. Heat-treated L. bulgaricus 2038 (P = 0.002) and S. thermophilus 1131 (P = 0.009) also significantly suppressed PSNP internalization, showing no significant difference from the live bacteria (Fig. 3a).

Fig. 3Lactobacillus delbrueckii subsp. bulgaricus 2038 and Streptococcus thermophilus 1131 suppressed polystyrene nanoplastics (PSNPs) internalization. (a) PSNPs and live or heat-treated L. bulgaricus 2038 and S. thermophilus 1131 were added to the Caco-2 monolayer. After culturing for 18 h, PSNP internalization was quantified by flow cytometry (n = 6/group). Statistical comparisons were performed using Tukey–Kramer test. * P < 0.05. (b–d) Inhibition of PSNP uptake by live strains was observed by confocal laser-scanning microscopy (scale bar = 10 μm, green: PSNP, red: E-cadherin). (e) Live bacteria were added to the Caco-2 monolayer then incubated for 18 h. Apical sides were washed three times by PBS to remove bacteria. Then, after PSNP exposure for 4 h, PSNP internalization was quantified by flow cytometry (n = 5/group). Statistical comparisons were performed using Dunnett’s test. * P < 0.05. MFI, median fluorescent intensity.

To exclude the possibility that PSNP internalization was suppressed by physical interference from LABs, we examined the effects of pre-treatment with both strains. Caco-2 cells were incubated with both strains for 18 h, then washed three times with phosphate-buffered saline (PBS) to remove the bacterial cells. Then, the PSNPs were added, and internalized PSNPs were measured by flow cytometry following incubation for 4 h. The results confirmed that pre-treatment with L. bulgaricus 2038 (P = 0.01) and S. thermophilus 1131 (P = 0.003) significantly suppressed PSNP internalization (Fig. 3e).

We also hypothesized that the tested strains did not suppress PSNP internalization through aggregation between PSNPs and LABs. Therefore, we evaluated the aggregation potential of PSNPs and LABs by incubating a mixture of PSNPs and LABs for 18 h before passing the mixture through a 0.22-µm filter. Since both strains are trapped by this filter, fluorescence is detected from the trapped LABs if PSNPs and LABs are aggregated. After washing with PBS, PSNPs trapped on the filter were collected and measured using a fluorescence plate reader. No fluorescence was detected from the collected samples. The results showed that PSNPs remained in the flowthrough fraction and did not aggregate with LABs, suggesting that the suppression of PSNP internalization was not due to aggregation between PSNPs and LABs.

Physiological effects of the LAB-induced suppression of PSNP internalization

Transcriptome analysis was performed to explore the physiological effects of PSNP internalization suppression by L. bulgaricus 2038 and S. thermophilus 1131. Gene Set Enrichment Analysis (GSEA) was performed to explore differentially expressed gene sets. PSNP treatment enriched the gene set ‘oxidative stress response’ according to Wikipathway GSEA and significantly enriched ‘base excision repair’ according to Reactome GSEA (Fig. 4a,d; Table 1). Any significant effects of LABs on these gene sets were not observed by microarray analysis (Fig. 4b,c,e,f; Table 1).

Fig. 4. Polystyrene nanoplastics (PSNPs), Lactobacillus delbrueckii subsp. bulgaricus 2038, and Streptococcus thermophilus 1131 affected the expression of various gene sets. Gene set enrichment analysis was performed. Enrichment plots displayed for the gene sets of (a–c) WIKIPATHWAY_OXIDATIVE_STRESS_RESPONSE, (d–f) REACTOME_BASE_EXCISION_REPAIR, and (g–i) REACTOME_GLUCRONIDATION. Comparisons are shown between (a, d, g) the control group and PSNP group, (b, e, h) the PSNP group and PSNP + L. bulgaricus 2038 group, and (c, f, i) the PSNP group and PSNP + S. thermophilus 1131 group.

Table 1. Gene set enrichment analysis.Gene setNESP-valueq-valueOxidative stress response control vs. PSNP1.540.0110.857 PSNP vs. PSNP + L. bulgaricus 2038−1.210.1901.00 PSNP vs. PSNP + S. thermophilus 1131−0.7940.7961.00Base excision repair control vs. PSNP−1.820.0030.191 PSNP vs. PSNP + L. bulgaricus 20381.040.3641.00 PSNP vs. PSNP + S. thermophilus 11310.9400.5701.00Glucuronidation control vs. PSNP−2.24< 0.001< 0.001 PSNP vs. PSNP + L. bulgaricus 20381.600.0391.00 PSNP vs. PSNP + S. thermophilus 11311.080.3541.00NES, nominal enrichment score.

According to Reactome GSEA, PSNP treatment showed significant negative enrichment of the ‘glucuronidation’ gene set, whereas LAB treatment suppressed it (Fig. 4g–i; Table 1).

Effects of LABs on PSNP transcellular permeability

Internalized nanoparticles are released from cells by exocytosis^25^. Therefore, we hypothesized that LABs suppressed the transcellular permeability of PSNPs. Because PSNPs were transported by the paracellular pathway^26^, we first incubated Caco-2 cells exposed to PSNPs in a medium containing ethylene glycol tetraacetic acid (EGTA) to break down intercellular adhesion. Then, the cells were washed with PBS to remove PSNPs from the cell surface, and the absence of fluorescence in the washed PBS was confirmed using a microplate reader. After incubation in the cell culture medium without EGTA for 48 h, we collected the medium at the basolateral side and measured the fluorescence intensity.

Both LAB strains significantly suppressed PSNP internalization, similar to the results of the previous experiment (Supplementary Fig. S3). Moreover, L. bulgaricus 2038 (P < 0.001) and S. thermophilus 1131 (P = 0.02) significantly suppressed the transcellular permeability of PSNPs to the basolateral side (Fig. 5).

Fig. 5Lactobacillus delbrueckii subsp. bulgaricus 2038 and Streptococcus thermophilus 1131 suppressed the transcellular permeability of polystyrene nanoplastics (PSNPs). PSNPs, L. bulgaricus 2038, and S. thermophilus 1131 were added to the Caco-2 monolayer. After culturing for 18 h, the cells were incubated in a medium containing 4 mM ethylene glycol tetraacetic acid (EGTA) for 30 min. Apical and basolateral sides were washed with PBS. The medium was then changed to a medium without EGTA. The medium on the basolateral side was collected after further incubation for 48 h, and the fluorescence was measured using a microplate reader (n = 6/group). Statistical comparisons were performed using Dunnett’s test. * P < 0.05. MFI, median fluorescent intensity.

Strain-dependent effects on the suppression of PSNP internalization

To examine whether the suppression of PSNP internalization by L. bulgaricus 2038 and S. thermophilus 1131 was specific to those strains, we compared the activities of these strains with those of other strains from the same species (L. bulgaricus and S. thermophilus, respectively). Because the strain-dependent growth speed affected the condition of Caco-2 cells when live bacteria were added, heat-treated bacteria were used to compare bacteria activity in this experiment.

For both species, all strains exhibited suppression activity; however, this activity was strain dependent (Fig. 6). Among the tested strains, L. bulgaricus 2038 and S. thermophilus 1131 showed the strongest activity in L. bulgaricus and S. thermophilus, respectively (Fig. 6).

Fig. 6. The suppression of polystyrene nanoplastics (PSNPs) internalization was strain dependent. PSNPs and heat-treated (a) L. bulgaricus or (b) S. thermophilus were added to the Caco-2 monolayer. After culturing for 18 h, PSNP internalization was quantified by flow cytometry (n = 3/group). MFI, median fluorescent intensity.

Discussion

NPs ingested orally pass through the gastrointestinal tract, and those that approach the epithelial cells can be internalized within the cells. Therefore, we focused on the small intestine in this study because of its large surface area, which leads to frequent exposure to NPs. Previous research using a tri-culture model comprising Caco-2, goblet-like, and M cell-like cells reported that smaller plastics were more likely to be internalized in Caco-2 cells^27^. PS is widely used in products such as food packaging and Styrofoam, and PSNPs are often selected as experimental models to examine their biological effects^12–17^ because the density of PS is greater than 1.0 g/mL, allowing the particles to sediment in culture media and come into contact with cells. Therefore, we selected differentiated Caco-2 cells and PSNPs with small diameters for this study.

The present study showed that PSNPs added to a Caco-2 monolayer were internalized within the cells. Regarding the pathway of PSNP internalization, EIPA and Dynasore significantly suppressed PSNP internalization, suggesting that PSNPs are endocytosed by macropinocytosis and the clathrin-mediated pathway in Caco-2 cells. In contrast, MβCD, which inhibits caveolae-mediated endocytosis by depleting cholesterol on the cell membrane, significantly increased PSNP internalization. These results are consistent with previous reports and imply that cholesterol reduction by MβCD may enhance clathrin-mediated endocytosis^15^.

Previous studies^12–15^ have also reported that PSNP was internalized into various cells; however, to the best of our knowledge, there are no reports regarding methods to control the process. This study is the first to report that L. bulgaricus 2038 and S. thermophilus 1131 effectively suppressed PSNP internalization into Caco-2 cells. Although the suppression mechanism remains unknown, no significant difference was observed in the effects of viable and heat-treated strains, suggesting that bacterial cell components were responsible for the observed effects. Additionally, we confirmed that binding between LABs and PSNPs was not involved. Moreover, pre-treatment with both LAB strains significantly suppressed PSNP internalization, precluding physical interference with endocytosis. These observations indicate that both LAB strains activate specific molecular signals in Caco-2 cells to suppress PSNP endocytosis. Previous experiments using mice showed that apical bacteria are endocytosed in intestinal epithelial cells via myosin light chain kinase (MLCK)-dependent fanning of the brush border by interferon gamma^28^. MLCK inhibitors can also reduce the endocytosis of silver nanoparticles and yttrium oxide nanoparticles in HeLa cells^29^. Moreover, L. acidophilus can inhibit MLCK activation in a Toll-like receptor 2 (TLR2)-dependent manner^30^. Components such as peptidoglycan and lipoteichoic acid in the cell wall of Gram-positive bacteria are also recognized by TLR2^31,32^. Based on these previous reports, we hypothesize that L. bulgaricus 2038 and S. thermophilus 1131 suppress PSNP internalization by reducing MLCK activity via TLR2 activation in Caco-2 cells. However, additional future research is required to verify this hypothesis.

We then examined the significance of the effects of both strains on PSNP internalization in Caco-2 cells using microarray transcriptome analysis. First, the results showed that PSNP addition increased the oxidative stress response and decreased DNA repair function, which is consistent with previous reports using different cells^13,17^, suggesting that the PSNP effects detected in this study were reasonable. However, the two strains did not ameliorate PSNP effects at the level of gene expression; thus, the suppression activities of these strains on PSNP internalization may not have been sufficient to fully alleviate PSNP effects. Moreover, optimizing the time at which gene expression was measured may have a mitigating effect on PSNP effects. Second, the results showed that PSNP-induced depletion of glucuronidation function was suppressed by both LAB strains. Glucuronidation is a phase II metabolic pathway for compounds such as drugs and polyphenols^33^. Polyphenols absorbed by small intestinal cells are partially transported into the body as glucuronide is conjugated by intracellular uridine 5’-diphospho-glucuronosyltransferases^34^. Polyphenols play a role in the suppression of antioxidant properties and the prevention of diseases, including type 2 diabetes^35,36^. Caco-2 cells also exhibit a glucuronidation function^37^. Therefore, both strains likely improve PSNP-induced inhibition of polyphenol transport into the body and contribute to health promotion.

Not only NP internalization into the intestinal epithelial cells but also their translocation within the human body have been reported to cause adverse health effects. Carboxylate-modified PSNP treatment to RAW 264.7 cells induced the expression of tumor necrosis factor alpha and IL-6^38^. MPs are reportedly transported in the bloodstream^39^; hence, NPs are likely also transported to organs throughout the body, where they induce adverse effects such as oxidative stress and apoptosis in each organ. Recently, MPs and NPs in carotid artery plaque increased the risk of myocardial infarction, stroke, or death in a clinical study^40^. In this study, L. bulgaricus 2038 and S. thermophilus 1131 suppressed the transcellular permeability of PSNPs in Caco-2 cells, suggesting that both strains may ameliorate excessive immune responses and consequently reduce the risk of various diseases caused by PSNP permeability.

The suppression of PSNP internalization was strain dependent. Assuming that recognition by TLR2 is important for suppression activity, differences in recognition by TLR2 among strains would affect their activities because peptidoglycan and lipoteichoic acid structures can differ among the species or strains^41,42^. In this study, L. bulgaricus 2038 and S. thermophilus 1131 exhibited the most potent suppression among all strains used in the experiments. Thus, consuming both strains could be an effective strategy for protecting against the adverse health effects of PSNPs. In the future, our experimental model could be used to identify bacterial strains with even stronger suppression activity.

This study has some limitations. First, although the internalization of small particles within cells depends on various factors such as particle size^43^, surface charge^43^ and material^44^, we only studied carboxylate-modified PSNPs. In addition to PSNPs with a diameter of 20 nm, we confirmed that both strains also suppressed the internalization of polystyrene MPs with a diameter of 200 nm in Caco-2 cells (Supplementary Fig. S4). Nonetheless, our study only revealed the activities of LAB strains on specific small plastics. Second, only in vitro studies were conducted in this study. In the future, we plan to clarify the activities of both strains through human clinical studies.

To summarize, we demonstrated that L. bulgaricus 2038 and S. thermophilus 1131 significantly suppressed the internalization of PSNPs in Caco-2 cells, even when the strains were non-viable. Transcriptome analysis suggested that the PSNP-induced depletion of glucuronidation function was ameliorated by both strains. Moreover, both strains suppressed the transcellular permeability of PSNPs in Caco-2 cells, suggesting reduced PSNP transport to organs throughout the body. Finally, we clarified that the suppression activities of L. bulgaricus 2038 and S. thermophilus 1131 on PSNP internalization were the strongest among other strains in the same bacterial species. This study highlights L. bulgaricus 2038 and S. thermophilus 1131 consumption as an effective strategy for protecting against the adverse health effects of PSNP intake.

Methods

Bacterial culture

Bacteria were cultured anaerobically with AnaeroPouch-Anaero (Mitsubishi Gas Chemical, Tokyo, Japan) at 37°C for 18 h. L. bulgaricus 2038, JCM 1002^T^, OLL1224, OLL205525, P2306601, P2306602, P2306603, P2306604, P2306605, P2306606, P2306607, and P2306608 were used and grown in de Man Rogosa Sharpe broth medium (Becton Dickinson, Cockeysville, MD, USA). S. thermophilus 1131, JCM 17834^T^, OLS3619, P2306609, P2306610, P2306611, P2306612, and P2306613 were used and grown in M17 broth medium (Becton Dickinson) supplemented with 1% lactose. Bacteria were washed twice with PBS (pH 7.4) and suspended in PBS to achieve an optical density at 600 nm (OD_600_) of 10.0, as determined using a U-2810 spectrophotometer (Hitachi, Tokyo, Japan). Heat-treated bacteria were prepared by incubating at 75°C for 1 h.

L. bulgaricus JCM 1002^T^ and S. thermophilus JCM 17834^T^ were purchased from RIKEN BRC (Ibaraki, Japan) and NCIMB, Ltd. (Aberdeen, Scotland, UK), respectively. The other strains used in this study were obtained from Meiji Co. Ltd (Tokyo, Japan).

Cell culture

Caco-2 cells were purchased from the European Collection of Authenticated Cell Cultures. The cells were cultured in minimum essential medium (MEM; Gibco, Thermo Fisher Scientific, Rochester, NY, USA) supplemented with 10% (v/v) fetal bovine serum (Biowest, Nuaillé, France), 1% (v/v) MEM non-essential amino acid solution (NEAA; Thermo Fisher Scientific), 100 U/mL penicillin, and 100 µg/mL streptomycin (Thermo Fisher Scientific) at 37°C and 5% CO_2_. Medium without penicillin–streptomycin was used during co-culture of LABs and Caco-2 cells.

To assess intestinal barrier function, cells were seeded at a density of 8.9 × 10^4^ cells/cm^2^ on the membranes in Transwell inserts (0.4 μm pore size; Costar, Cambridge, MA, USA) and cultivated for 14 days to form a monolayer. The medium was changed every 2 or 3 days. Trans-epithelial electrical resistance values were measured using a Millicell-ERS voltohmmeter system (Millipore, Burlington, MA, USA); cells with values > 300 Ω·cm^2^ were used in subsequent experiments to ensure sufficient and stable barrier formation.

PSNP characterization

Carboxylate-modified PSNPs labeled with yellow-green fluorescent were commercially obtained (Thermo Fisher Scientific). Particle diameter and density were approximately 20 nm and 1.05 g/mL, respectively, according to manufacturer guidelines.

A droplet of PSNPs was placed on carbon-film grid and dried at 25°C. PSNP morphology was characterized using a transmission electron microscope (JEM-1400Flash; JOEL, Tokyo, Japan) at 100 kV.

PSNP particle size distribution and zeta potential were analyzed by dynamic light scattering with ELSZneo (Otsuka Electronics, Osaka, Japan). MEM (Gibco), with 0.005% Tween 20 supplemented with 1% (v/v) NEAA (Thermo Fisher Scientific) was used as the medium.

Caco-2 cell exposure to PSNPs and LABs

PSNPs were suspended in the same medium used for PSNP characterization, and the mixture was sonicated at an output level of 150 W for 1 min using an Ultrasonic Cleaner VS-150 (Velvo-Clear, Tokyo, Japan) to avoid aggregation. Caco-2 cells were exposed to 2, 20, or 200 µg/mL PSNPs, and LABs with an OD_600_ of 10 were simultaneously added at 1% to the medium to achieve a final OD_600_ of 0.1. OD_600_ of 0.1 corresponds to approximately 1 × 10^7^ CFU/mL for L. bulgaricus 2038 and S. thermophilus 1131. The co-culture was incubated at 37°C under 5% CO_2_. Cell viability and cytotoxicity following PSNP treatment were evaluated using a WST-1 cell proliferation assay system (Takara Bio, Shiga, Japan) and a cytotoxicity LDH assay kit-WST (Dojindo, Kumamoto, Japan), respectively.

To examine the effects of LAB pre-treatment on the cellular internalization of PSNPs, LABs were added to the medium to achieve OD_600_ = 0.1, and Caco-2 cells were incubated for 18 h at 37°C under 5% CO_2_. The apical sides were washed 3 times with PBS to remove LABs. Then, 200 µg/mL of the PSNPs which are treated for anti-aggregation, as described above, were added to the mixture and the cells were incubated for 4 h at 37°C under 5% CO_2_.

Assessment of PSNP and LAB aggregation

PSNPs were suspended in MEM (Gibco) supplemented with 1% (v/v) NEAA (Thermo Fisher Scientific). To avoid aggregation, the mixture was treated as described in Section ‘Caco-2 cell exposure to PSNPs and LABs.’ LABs were added to PSNPs to achieve OD_600_ = 0.1. The mixture was incubated at 37°C under 5% CO_2_ for 18 h then passed through a 0.22-µm polyvinylidene fluoride filter (Merck, Darmstadt, Germany). The filter was washed with PBS 5 times. The trapped LABs were collected by syringe, and the fluorescence emission at 530 nm was measured after excitation at 500 nm using a Synergy H1 microplate reader (BioTek, Santa Clara, CA, USA).

Qualitative assessment of PSNP internalization in Caco-2 cells using immunofluorescent staining

Caco-2 cells exposed to PSNPs on the membrane of a Transwell plate were washed 3 times with PBS. Cells were fixed with 4% paraformaldehyde (Wako) for 10 min at 25°C. After washing with PBS, PBS containing 0.2% Triton X-100 was added, and the cells were incubated for 5 min at 25°C. Image-iT FX Signal Enhancer (Thermo Fisher Scientific) was then added and incubated for 30 min at 25°C. Subsequently, the cells were blocked with 10% normal goat serum for 20 min at 25°C and labeled with anti-E-cadherin (GTX100443; Genetex, Irvine, CA, USA) antibodies for 2 h at 25°C, followed by incubation for 1 h at 25°C with Alexa Fluor 647-conjugated goat anti-IgG (ab150087; Abcam, Cambridge, MA, USA). Finally, the membranes were cut with a scalpel and placed on glass slides. Cells were visualized using a confocal laser-scanning microscope (LSM880; Zeiss, Oberkochen, Germany).

To examine the effects of LABs and endocytosis inhibitors on PSNP internalization, LABs and endocytosis inhibitors were added to the apical side with PSNPs. Dynasore, MβCD and EIPA were used as inhibitors of clathrin-mediated, caveolae-mediated endocytosis, and macropinocytosis, respectively. Cell viability and cytotoxicity were evaluated as described in Section ‘Caco-2 cell exposure to PSNPs and LABs.’

Quantification of PSNP internalization in Caco-2 cells using flow cytometry

Caco-2 cells exposed to PSNPs were washed 3 times with PBS to remove the PSNPs from cell surfaces. Then, Accumax (Nacalai Tesque, Kyoto, Japan) was added to the apical sides, and the cells were incubated at 37°C under 5% CO_2_ for 10 min. The cells were scraped and stained with SYTOX Red Dead Cell Stain (Thermo Fisher Scientific) to identify dead cells. Data were acquired using flow cytometry (FACSVerse; BD Biosciences, Franklin Lakes, NJ, USA).

To examine the effects of LABs and endocytosis inhibitors on PSNP internalization, LABs and endocytosis inhibitors were added as described in Section ‘Qualitative assessment of PSNP internalization in Caco-2 cells using immunofluorescent staining.’

Quantification of internalized PSNP permeability to the basolateral side

Caco-2 cells were cultured with LABs in the presence of PSNPs as described in Section ‘Caco-2 cell exposure to PSNPs and LABs.’ The cells were cultured in MEM supplemented with 1% (v/v) NEAA (Thermo Fisher Scientific), 100 U/mL penicillin, 100 µg/mL streptomycin (Thermo Fisher Scientific), and 4 mM EGTA for 30 min. Then, to remove remaining PSNPs on the cell surface and in wells, the apical and basolateral sides of the membrane were washed with PBS 10 and 3 times, respectively. The absence of fluorescence in the washed PBS was confirmed using a Synergy H1 microplate reader (BioTek). The medium was changed to MEM supplemented with 1% (v/v) NEAA (Thermo Fisher Scientific), 100 U/mL penicillin, and 100 µg/mL streptomycin (Thermo Fisher Scientific). The medium on the basolateral side was collected after 48 h, and the fluorescence was measured as described in Section ‘Assessment of PSNP and LAB aggregation.’

Microarray analysis

Total RNA from Caco-2 cells was prepared using the Maxwell RSC simpleRNA Cells Kit (Promega, Madison, WI, USA), according to the manufacturer’s protocols. RNA concentration was measured using an Agilent Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent Technologies, Palo Alto, CA, USA). RNA samples from the same group were pooled in equivalent amounts. A GeneChip WT PLUS Reagent Kit (Thermo Fisher Scientific) was used to prepare microarray samples according to the manufacturer’s protocol. Samples were hybridized to Clariom S arrays for humans (Thermo Fisher Scientific). All arrays were scanned using the Affymetrix GeneChip Command Console installed on the GeneChip Scanner 3000-7G. Array datasets were normalized using the signal space transformation-robust multichip analysis algorithm implemented in the Affymetrix’s Transcriptome Analysis Console software version 4.0.

For enrichment analysis, GSEA (Broad Institute, Cambridge, MA, USA)^45,46^ was performed. GSEA software version 4.3.2 was used to perform GSEA based on Wikipathway and Reactome databases. To compute the nominal enrichment score (NES), the permutation value was set to 1,000. Gene sets with a false discovery rate q-value < 0.25 were considered to show significant enrichment. P < 0.05 was used to select enriched gene sets when the q-value exceeded 0.25.

Statistical analysis

Data are presented as the mean ± standard error. The Shapiro–Wilk test was performed to determine the normality of distribution. Statistical analysis was performed using two-tailed Welch’s t-test, Dunnett’s test, or Tukey–Kramer test. Statistical significance was set at P < 0.05.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1