A single colony of L. salivarius SNK-6 was inoculated into 1000 mL of MRS medium. The culture was shaken at 220 rpm and incubated under anaerobic conditions at 41 °C for 24 h to achieve a final concentration of approximately 2 × 10⁸ CFU/mL of the L. salivarius SNK-6 mixed solution. LsEVs were obtained from the culture supernatant at 4 °C following previously outlined technique (17). Briefly, to remove cell debris and aggregates, the bacterial culture medium underwent centrifugation at a low speed (12,000 × g for 20 minutes). Following this, the supernatant was collected and filtered with 0.45 μm membrane to remove larger molecules. Then, the filtrate was transferred into a 15 mL 3-kDa ultrafiltration system and centrifuged for 30 minutes at an average speed of 3,000 × g. Finally, total exosome isolation reagent (from cell culture media, Invitrogen, USA) was added to the retained filtrate overnight, and the EVs were separated at 12,000 × g for 1 h, according to the manufacturer’s guidelines. The EVs were then resuspended in 1mL RNA-free water and kept at -80 °C. High concentrations of LsEVs were achieved by merging LsEVs from the culture media collected over four different periods and concentrating them with a 3-kDa ultrafiltration system (Millipore, MA, USA). According to our prior nanoparticle tracking analysis, LsEVs were mostly between 100 and 250 nm in diameter, with the largest reaching 236 nm (17). The BCA protein assay kit (Sangon Biotech, Shanghai, China) is used to determine the protein concentration of LsEVs. Finally, the purified LsEVs were identified by transmission electron microscopy (JEOL, Tokyo, Japan) and details are described below.

Forty 35-day-old male BALB/c mice were selected and classified into four groups: a normal diet group (conventional diet + 100 μL PBS), a high-fat group (HFD + 100 μL of PBS), an SNK-6 treated group (HFD + 100 μL of 1 × 10⁸ CFU/mL L. salivarius SNK-6), and an LsEVs treated group (HFD + 100 μL of 10 μg LsEVs). The mice were provided with a standard diet and kept in an environment with a 12 h light and dark cycle at temperatures ranging from 21 to 25 °C, with humidity levels maintained at 55 ± 10%. Subsequently, mice from the normal diet were treated with conventional diet, while mice in the HFD, SNK-6, and LsEVs treated group were free consumption of a high-fat feed for two weeks, after which, from the 56-day-old onward, each group received gavages of 100 μL PBS, L. salivarius SNK-6, or LsEVs. After 12 weeks (140-day-old), the mice were euthanized, and blood was drawn from their eyes. The serum was obtained by centrifuging the blood at 4,000 rpm for a duration of 10 minutes, after which it was stored at -80 °C for future analysis. Liver tissue was gathered and aseptically frozen at -80 °C. The animal studies adhered to the guidelines and regulations provided by the Ethics and Animal Welfare Committee of the Shanghai Academy of Agricultural Sciences (approval number: SAASPZ0521022). The animal sample size was determined through a post hoc analysis of the achieved power using G*Power 3.1.9.7 (Heinrich Heine Universität, Düsseldorf, Germany). With a significance level (α) set at 0.05, an effect size (d) of 1.5, and a sample size of 10 animals per group, we attained a widely recognized statistical power (1−β) of 0.92.

To assess the effectiveness of different substances, we acquired commercial kits for alanine aminotransferase (ALT), aspartate aminotransferase (AST), total cholesterol (T-CHO), and triglycerides (TG) from Nanjing Jiancheng (Nanjing, China). The serum concentrations of TNF-α, IL-2, IL-6, and IL-4 were determined using enzyme linkedimmunosorbent assay kits following the manufacturer’s instructions (Jingmei Bio, Jiangsu, China).

The livers and jejunum samples were preserved in 4% paraformaldehyde, subjected to dehydration through rising concentrations of ethanol, cleared using xylene, and then encapsulated in paraffin. They were sliced into 5 μm sections, processed with xylene, absolute ethanol, and decreasing ethanol concentrations, dewaxed, rinsed with double-distilled water, and dried. Subsequently, the sections were stained with alkaline hematoxylin and 0.5% eosin, followed by scanning with an Olympus VS200 (OLYMPUS Corporation, Japan) slide scanner.

Livers were initially preserved with 4% paraformaldehyde, then dehydrated in 20% sucrose, and embedded in Tissue Tek OCT (Sakura, Netherlands). They were quickly frozen and sliced into 8 μm sections. Subsequently, Oil Red O staining was performed using Solarbio’s kit (Beijing, China), following the instructions. Image Pro Plus 6.0 software (Media Cybernetics, Bethesda, MD, USA) was used for analyzing positive staining and slice area.

TEM was performed according to established protocols (18). Briefly, purified LsEVs and livers were preserved overnight in a 2.5% glutaraldehyde and 0.05M sodium phosphate buffer solution at pH 7.2. Following the washing of samples in 0.1M diethyl carbonate buffer combined with saline, the next steps involved dehydrating the samples using ethanol, embedding them in epoxy resin, and slicing them into ultra-thin sections measuring 50 to 100 nm. These sections were then placed onto copper grids, subjected to staining with uranyl acetate and lead citrate, and observed under a Hitachi H7700 transmission electron microscope (Tokyo, Japan).

Liver tissues embedded in paraffin were cut into sections of 5 μm thickness and subsequently processed using relevant primary Beclin-1 antibodies (Servicebio, Wuhan, China; GB115741). Concurrently, jejunum sections were incubated with claudin-1 antibody (Servicebio, GB12032). The sections underwent incubation in a blocking solution before being incubated with fluorescent isothiocyanate-labeled secondary antibodies for 30 min at 37 °C. Following washes with PBS, the sections were stained with DAPI for 10 min. Finally, the sections were mounted with neutral resin and examined under a fluorescence microscope.

The NCTC1469 human hepatocyte line (CCL-227, ATCC) was cultured in high-glucose DMEM (Invitrogen, Thermo Fisher Scientific, Inc.), which was enriched with 10% horse serum, under conditions of 37 °C in a 5% CO₂/95% air environment. To verify the protective effect of LsEVs on fatty acids (FFA) challenged liver cells, the NCTC1469 cells were seeded in 12-well plates (approximate 2×10⁵ cells per well) and treated with 1 nM FFA (oleic acid: palmitic acid, 2:1) for 12 h. Following this, 20 nM LsEVs were introduced into the culture media and incubated again at 37 °C for 24 h. Meanwhile, cells treated with PBS in the absence of FFA and LsEVs was used as negative control. Cells were gathered for RT-PCR and western blot analysis.

To verify the protective effect of LsEVs on lipopolysaccharide (LPS) challenged intestinal epithelial cells, human intestinal epithelial Caco-2 cells, sourced from ATCC, were cultivated in DMEM (Invitrogen) along with 10% FBS (Invitrogen). Caco-2 cells (5 × 10⁵ cells/ml) were treated with PBS, or LPS (1 μg/ml), or LPS (1 μg/ml) + LsEVs (10nM) for 24 h. Cell supernatant and cellular samples were then harvested for subsequent analysis. The appropriate amount of LsEVs to be added to cells based on cell viability determined by Cell counting kit-8 (CCK-8, Beyotime, Shanghai, China). Details were shown in Supplementary Figure S1 .

RNA extraction was conducted using Trizol Reagent (TIANGEN Biotech, Beijing, China). Subsequent to this, 1 µg of RNA underwent reverse transcription employing Prime Script™ RT Master Mix (Perfect Real Time) (Applied Biosystems, Waltham, MA, USA). The quantitative analysis of genes was conducted using Power Up SYBR Green MasterMix on a Step One Plus Real-Time PCR Instrument (Applied Biosystems). GAPDH and β-actin served as the housekeeping gene. The primers are listed in Supplementary Table S1 .

Equal volumes of protein samples were carefully loaded onto polyacrylamide gels, subjected to electrophoresis, and transferred onto PVDF membranes (Thermo Fisher Scientific). To detect specific proteins, various primary antibodies were used, such as those targeting LC3 (GB113801), peroxisome proliferator-activated receptor γ (PPARγ, GB11164), sterol regulatory element-binding transcription factor 1 (SREBP1, GB11524), Beclin-1 (GB112053), claudin-1 (GB112543), and β-actin (GB15003), all sourced from Servicebio (Wuhan Servicebio Technology Co., Ltd., Wuhan, China). The presence and quantity of the target proteins were analyzed using a Super ECL detection reagent (Yeasen, Shanghai, China). The resulting bands were then quantified for analysis utilizing Image J software (NIH, Bethesda, MD).

Genomic DNA was obtained from fecal samples of normal diet, HFD, and LsEVs treated mice. The amplification of the bacterial 16S rRNA genes was conducted on the V3–V4 region using the primer pairs 338F (5’-ACTCCTACGGGAGGCAGCAG-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’). This process utilized a T100 Thermal Cycler PCR device (BIO-RAD, USA). Following purification, the amplicons were combined in equal molar concentrations and then subjected to paired-end sequencing on an Illumina Nextseq2000 platform (Illumina, San Diego, USA), in accordance with the standard protocols specified by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China).

Following demultiplexing, quality filtering of the resulting sequences was performed using fastp (version 0.19.6) (https://github.com/OpenGene/fastp) and the merging process was conducted with FLASH (version 1.2.11) (http://www.cbcb.umd.edu/software/flash). Subsequently, high-quality sequences were subjected to de-noising using the DADA2 plugin within the Qiime2 (version 2020.2) framework. This process utilized the recommended settings that ensure single nucleotide resolution by analyzing error profiles within the samples. To reduce the impact of sequencing depth on the measures of alpha and beta diversity, the sequence count from each sample was rarefied to 20,000, which resulted in an average Good’s coverage of 97.90%. The functional capabilities of the microbial communities were predicted employing PICRUSt, leveraging the high-quality sequences. LEfSe was applied to identify bacterial species that contribute to differences among the experimental groups.

Data underwent a one-way ANOVA analysis, subsequently followed by Tukey’s test for multiple comparisons, utilizing SPSS 13.0 software (SPSS, Chicago, IL). Statistical significance between the two mean values was determined by an adjusted P-value of less than 0.05.

Recently, we discovered that L. salivarius SNK-6 exhibits a protective effect against fatty liver in laying hens. In our previous study, we applied a fixed concentration of L. salivarius SNK-6 at 1 × 10⁸ CFU/kg (12). To investigate the effect of LsEVs on MAFLD, mice were subjected to daily oral gavage administration (100 µL) of either L. salivarius SNK-6 or LsEVs for a duration of 84 days. The dose of LsEVs used in this study was determined based on the findings reported by Hao et al. (19) and Hu et al. (20).

Here, we found that L. salivarius SNK-6 and its secreted LsEVs were supplemented to the HFD diet ( Figure 1a ) to evaluate the function of LsEVs on fatty liver prevention. As shown in Figure 1b , we found a significant increase in the body weight on HFD mice, when comparing to the normal diet treated mice (P <0.05). Oral administration of L. salivarius SNK-6 and LsEVs significantly reduced weight gain of mice, comparing with the HFD treated mice (P <0.05). In comparison with the control groups, HFD led to a significant rise in serum T-CHO and TG, potentially contributing to the development of fatty liver ( Figure 1c ). Additionally, the activities of AST and ALT were markedly higher in the HFD groups compared to the other three groups. When L. salivarius SNK-6 and LsEVs were administered orally, a noticeable decrease in TNF-α levels was observed; however, no significant reduction in the pro-inflammatory factors IL-2 and IL-6 occurred in mice treated with L. salivarius SNK-6 and LsEVs when compared to those treated with the HFD. Conversely, a significant increase in the anti-inflammatory factor IL-4 (P<0.05) was noted in mice receiving L. salivarius SNK-6 and LsEVs ( Figure 1d ).

As depicted in Figure 2 , mice in HFD group have livers with a brownish-yellow color, fragile and soft texture, and large vesicular steatosis with many fat vacuoles. In contrast, mice treated with L. salivarius SNK-6 and LsEVs have somewhat yellow livers with a smooth, elastic texture and fewer fat vacuoles. Oil Red O staining reveals significant fat accumulation in the HFD group, with many irregular red lipid droplets, whereas the control group shows minimal droplets. The LsEVs treatment group has some red lipid droplets, but significantly fewer than the HFD group ( Figure 2a ). In comparison to the HFD group, mice that received treatment with L. salivarius SNK-6 and LsEVs exhibited a remarkable decrease in liver lipid droplet counts (P<0.05). Notably, the group administered LsEVs demonstrated a significant reduction in hepatic lipid droplets (P<0.05), suggesting a reduction in lipid buildup and a more substantial improvement in metabolic disorders related to lipid accumulation induced by the high-fat diet ( Figures 2b, c ).

TEM was utilized to observe autophagosomes, mitophagosomes, and autophagic lysosomes. In Figure 3 , no obvious lipid droplets were detected in the liver cells of the control group mice. The integrity of the cellular structure was intact, with cytoplasmic density appearing normal. The mitochondrial structure was also preserved, exhibiting a dense and regular arrangement. Conversely, the liver cells from the HFD group exhibited a higher quantity of dispersed lipid droplets of different sizes, along with structural impairments, sparse cytoplasmic swelling, and notable mitochondrial enlargement. In contrast to the HFD mice, there was a considerable decrease in the number of lipid droplets within the liver cells following treatment with L. salivarius SNK-6 and LsEVs, and autophagosomes developed in the mouse liver. The number of autophagosomes and autophagic lysosomes increased, and mitochondrial damage was reduced ( Figure 3 ). An elevation in autophagy was noted in the mice of L. salivarius SNK-6 and LsEVs treated group, when compared to those on HFD mice.

Immunofluorescence analysis revealed that, Beclin1 proteins were located near the nuclei as depicted in the control group ( Figure 4a ). In contrast, the HFD group displayed a marked reduction in Beclin1expression, while a partial recovery was noted in the groups treated with L. salivarius SNK-6 and LsEVs groups ( Figure 4b ).

Genes crucial for the autophagy-related pathway were analyzed using RT-qPCR ( Figure 5a ). It was noted that the mitophagy-related genes such as FUN14 domain-containing protein 1 (FUNDC1), BNIP3, LC3-Ⅱ, Beclin1, and PTEN induced putative kinase 1 (PINK1) showed significant increase in the livers of HFD mice that received L. salivarius SNK-6 (P<0.05) and LsEVs (P<0.01), in comparison to HFD mice. Furthermore, a marked enhancement in the mRNA levels of Parkin and PINK1 in the LsEVs treated cells. In contrast, the level of LC3-Ⅱ was lower in the HFD mice compared to the control group. Additionally, a reduction was observed in the mRNA levels of BNIP3 and Parkin in the cells treated with FFA. The expression of mitophagy-related proteins LC3 and Beclin1 were increased in the LsEVs treated mice compared to the HFD group ( Figures 5b, c ).

We conducted additional analyses on the mRNA expression levels of genes related to lipid metabolism, which include the fatty acid oxidation genes PPARα and PPARγ, alongside genes linked to fat synthesis, like fatty acid synthase (FASN) and SREBP1 ( Figure 6a ). A notable rise in PPARα and PPARγ levels was detected in mice treated with LsEVs when compared to those subjected to HFD, whereas the mRNA expression of FASN showed a reduction (P < 0.05). In FFA-treated cells, reduced levels of PPARα and PPARγ were noted, alongside increased expression of FASN; however, LsEVs reversed the expression of these genes. Notably, no significant differences were detected in protein levels of PPARγ and SREBP1 in the in vivo and in vitro models, although a decreasing trend in SREBP1 protein levels was noted in LsEVs-treated mice ( Figures 6b, c ).

Research has shown that the functionality of the intestinal barrier is impaired in both mice and individuals suffering from fatty liver disease. In order to explore the effects of LsEVs on the intestinal barrier in mice exposed to HFD, refer to Figure 7 for further details, we observed that in the normal mice, the structure of the intestinal mucosa layers was intact, with villi arranged in an orderly fashion and epithelial cells exhibiting a tall columnar shape. In contrast, the HFD group displayed several pathological changes, including villi atrophy, rupture defects, necrosis and shedding of mucosal epithelial cells, as well as separation of the epithelial layer from the lamina propria in the jejunum. Following treatment with L. salivarius SNK-6 and LsEVs, the morphology of the jejunal villi in these mice showed slight improvement, characterized by mild defects and an intact intestinal mucosal epithelium exhibiting normal morphology. Meanwhile, we used LPS to induce the stress model of gut cells. We found that LsEVs obviously reduced the contents of TNF-α and IL-2, although no significant reduction of IL-6 was found in the LsEVs treated cells, when comparing to the LPS treated cells. Meanwhile, significant increase in the content of IL-4 (P<0.05) was observed in the LsEVs treated cells ( Figure 7c ).

Immunofluorescence staining was employed to observe the distribution of Claduin-1 on the surface of intestinal mucosal epithelial cells, which was consistent with the observed expression of this tight junction protein. In the HFD mice, the arrangement of Claduin-1 protein appeared loose and discontinuous, accompanied by a decrease in fluorescence intensity. Following treatment with L. salivarius SNK-6 and LsEVs, an increase in the fluorescence intensity of Claduin-1 protein was noted, particularly in the LsEVs treatment group, where levels approached those of the control group ( Figures 8a, b ). In the HFD-induced fatty liver mouse model, the protein level of Claudin-1 was significantly reduced, suggesting that the HFD may have disrupted intestinal barrier function ( Figure 8c ).

In light of the recognized relationship between nutrition and the composition of gut microbiota, we examined how LsEVs influence the bacterial populations in the ceca of mice fed a high-fat diet. The results showed that, ace index, chao index, and sobs index of the alpha diversity were significantly changed among the normal, HFD, and LsEVs treated mice ( Figure 9a ). The beta diversity index served as a tool to assess the differences in diversity among the samples ( Figure 9b ). Analyzing the data through principal coordinate analysis (PCoA), along with principal component analysis (PCA), demonstrated a significant alteration in the microbial composition of mice subjected to HFD and LsEVs treatment in contrast to those on a normal diet.

The abundance of five phyla (Fimicutes, Bacteroidota, Patescibacteria, Desulfobacterota, and Campilobacterota) and 14 genera (Muribaculaceae, Blautia, Desulfovibrio, Alistipes, Colidextribacter, Eubacterium_corprostanoligenes_group, norank_f_Oscillospiraceae, Lactobacillus, Lactococcus, and Candidatus_Saccharimonas) was different in the ceca from mice fed normal, HFD, and LsEvs (P < 0.05, Figures 9c , 10 ). For example, at the phylum level, Fimicutes had more than 1.8-fold abundance in HFD mice compared with normal mice (P < 0.01), while Bacteroidota were much less abundant in HFD mice compared with normal mice a (P < 0.01, 11.08% vs 51.46%). LsEVs treatment significantly increased the abundance of Bacteroidota in the ceca of HFD mice (P < 0.01, Supplementary Figure S2 , 27.13% vs 11.08%). At the genus level, the abundance of Desulfovibrio in mice treated with LsEVs was over ten times lower than that in those on HFD (P < 0.05, Supplementary Figure S3 , 1.88% vs 20.23%).

Muribaculaceae, Alistipes, Lactobacillus are much higher in the normal diet treated mice than those in the HFD treated mice, LsEVs treatment tend to increase the abundance of these microorganisms, while no significant difference was found. Overall, Lactobacillus was enriched in the control group, while Desulfovibrio were enriched in HFD mice. LsEVs treatment significantly upregulated the abundance of Eubacterium-coprostagenes ( Figure 10 ).