Food-Derived Limosilactobacillus fermentum GR-3 Enhances Anti–PD-1 Immunotherapy Efficacy Authors: Wei Xiang, Qi Yang, Jing Ji, Hui Qiao, Aman Khan, El-Sayed Salama, Chunlan Mao, Ying Wu, Xiangkai Li Published: 2026 ABSTRACT AbstractImmune checkpoint blockade (ICB) has improved the cancer treatment, its application remains constrained by heterogeneous response rates and immune-related adverse events. While gut microbiome modulation represents a promising avenue to optimize ICB, rationally selected probiotics with defined mechanistic actions are critically lacking. Here, we selected Lactobacillus fermentum GR-3 (L. fermentum GR-3), a food-derived strain pre-screened for exceptional gastrointestinal resilience (maintaining > 107 CFU/mL), antioxidant activity (87.4% DPPH-Scavenging efficiency), and immunomodulatory potential, as an oral combination strategy to enhance anti-PD-1 therapy. In the MC38 murine colorectal cancer model, oral L. fermentum GR-3 synergistically enhanced ICB efficacy, yielding significantly greater tumor suppression than anti-PD-1 monotherapy (51.7% volume reduction vs. untreated controls), while attenuating therapy-associated weight loss. Locally within the tumor microenvironment (TME), the combination therapy promoted dendritic cell maturation (MHC-II⁺ DCs reaching 56.0%), favored pro-inflammatory macrophage polarization (M1 increased to 20.8%, while M2 reduced to 2.57%), and suppressed regulatory T cell infiltration (Tregs decreased to 2.32%). These shifts strongly correlated with enhanced cytotoxic responses, with Granzyme B⁺ CD8⁺ T cells reaching 40.4% (doubling the effect of PD-1 monotherapy). Mechanistically, L. fermentum GR-3 colonization partially reversed tumor-associated gut dysbiosis, enriching short-chain fatty acid (SCFA)-producing consortia (e.g., Lachnospiraceae and Bacteroides) and increasing intestinal SCFA concentrations (e.g., propionic and butyric acids recovering to 70.13 and 26.07 μM/g, respectively)). This microbial-metabolic shift was accompanied by upregulated expression of mucosal tight junction genes (ZO-1, Occludin). Furthermore, L. fermentum GR-3 facilitated a balanced immune response. Systemically, it alleviated oxidative stress and buffered peripheral inflammatory signals, contributing to immune homeostasis in non-tumor tissues (spleen and colon). Collectively, these findings position L. fermentum GR-3 as a grounded, translatable strategy to simultaneously enhance ICB efficacy and improve therapeutic tolerability.Supplementary InformationThe online version contains supplementary material available at 10.1007/s12602-026-11048-0.Keywords: L. fermentum GR-3, Immune checkpoint blockade, Tumor microenvironment, Short-chain fatty acids (SCFAs), Immune homeostasis Introduction Cancer remains a leading global cause of death, with projections estimating 28 million new cases and 16.2 million deaths by 2040 [1]. Traditional cancer treatments, including surgical resections, chemotherapy, radiotherapy and immunotherapy, have been widely used [2]. Immunotherapy has transformed the field of oncology by harnessing the body's immune response to fight against cancer [3]. A major milestone in immunotherapy is the development of ICB therapy, which specifically targets the PD-1/PD-L1 pathway, preventing T cell exhaustion and restoring robust anti-tumor immune responses [4]. For example, PD-1 inhibitors like nivolumab and pembrolizumab achieved objective response rates of 40—45% in advanced melanoma and 20—30% in non-small cell lung cancer [5, 6]. Despite these successes, PD-1 blockade monotherapy faces significant limitations. Recent meta-analyses reveal that only 20—40% of patients showed durable responses to PD-1 blockade, with many others developing primary or acquired resistance [7]. Moreover, PD-1 blockade can lead to Treg overactivation, resulting in immune suppression rather than immune reinvigoration [8]. These challenges motivate us to explore innovative strategies to enhance the efficacy of ICB therapy. The gut microbiota is a diverse and dynamic microorganisms ecosystem to regulate host immunity and metabolic pathways [9]. Studies have shown that cancer progression disrupts intestinal microbial homeostasis, leading to a reduction in beneficial bacterial populations [10]. For instance, Faecalibacterium prausnitzii, a primary butyrate producer, is reduced by 60% in colorectal cancer patients [11]. Akkermansia muciniphila, essential for maintaining gut barrier integrity, decreases by approximately 75% in advanced breast cancer cases. Similarly, Bifidobacterium species, crucial for immune regulation, show a 45% decline in melanoma patients [12]. In contrast, levels of certain pathogenic or opportunistic bacteria, such as Clostridiaceae, Faecalibacterium, Ruminococcaceae and Fusobacterium nucleatum, are increased in cancer patients, contributing to inflammation and tumor progression [13, 14]. Furthermore, cancer-induced dysbiosis of gut microbiota results in multiple immunological disruptions, including a 40% reduction in CD8 + T cell infiltration at tumor sites [15], diminished regulatory T cell populations [16], and an approximately 50% decline in protective immunoglobulin A (IgA) level [17]. Research indicates that patients with severe dysbiosis may experience up to a 30% reduction in immunotherapy response rates [18]. Metabolomic analyses reveal a sharp decrease in short-chain fatty acids, particularly butyrate, which are essential for maintaining gut barrier function and supporting anti-tumor immunity [19]. These immune and metabolite alterations are correlated with microbial imbalances induced by cancer [20]. These findings underscore the urgent need for therapeutic strategies that restore gut microbial balance to optimize immunotherapy outcomes. Probiotics offer significant potential for enhancing the efficacy of immunotherapy [21]. Evidence suggests that probiotics can amplify PD-1 therapy by modulating immune responses, improving gut barrier integrity, and restoring microbial balance. For example, Bifidobacterium has been showed to enhance dendritic cell function combined with anti-PD-L1 therapy in melanoma models [22]. Supplementation with Lactobacillus rhamnosus led to a 42% tumor reduction, compared to immunotherapy alone (18%) in colorectal cancer models [23]. Bifidobacterium longum upregulated tight junction proteins such as claudin-3 and occludin in colorectal cancer model, contributing to better intestinal barrier function. Likewise, Lactobacillus rhamnosus GG restores gut microbiota homeostasis by adjusting the Firmicutes/Bacteroidetes ratio [24]. Clinical studies have further supported these findings, with colorectal cancer patients receiving probiotics and PD-1 therapy experiencing a median survival extension of 4.2 months compared to those on PD-1 therapy alone [25]. Despite these promising findings, most probiotic currently used to enhance anti-PD-1 immunotherapy are derived from human gut microbiota, thus relying on the host's native microbiome. Thus, there is a need for food-derived probiotics with established safety profiles that can effectively colonize the gut under varying physiological stresses and enhance immunotherapy efficacy through microbiota modulation. Traditional fermented foods, abundant in probiotics, have attracted considerable interest due to their distinctive and potentially advantageous properties, along with an assumed safety profile [26]. Our research team previously isolated L. fermentum GR-3 from Northwestern traditional fermented food "Jiangshui". This strain has been confirmed in lowering uric acid level, mitigating tumor-induced oxidative stress, enhancing host immune function and modulating the gut microbiota [27]. Clinical trials in hyperuricemia patients have validated the favorable safety and tolerability profile of L. fermentum GR-3, suggesting its potential for broader therapeutic applications [28]. Using this as a starting point, our study investigates how L. fermentum GR-3 can work with PD-1 therapy to improve cancer treatment outcomes by boosting the effectiveness of immune checkpoint inhibitors. 16S rRNA sequencing was used to evaluate shifts in the gut microbiota, and RT-qPCR was employed to quantitatively analyze gene expression related to immune function and barrier protection. Cytokine profiles in tumor tissues and serum, as well as tumor cell apoptosis, were assessed to evaluate the effects of L. fermentum GR-3. The results demonstrate that L. fermentum GR-3 enhances the anti-tumor efficacy of PD-1 immunotherapy by modulating gut microbiota and amplifying immune responses, offering promising strategy to improve cancer treatment outcomes. Materials and Methods Bacterial Strains and Culture Conditions L. fermentum GR-3 was previously isolated and purified from the traditional Northwestern Chinese food “jiangshui” [29]. The strain has been formally deposited at the China Center for Type Culture Collection (CCTCC, No. M2021902). For the in vitro antioxidant capacity comparison, six additional strains (Pediococcus acidilactici S1, S5, and S6; L. fermentum S2; Lactiplantibacillus plantarum S4; and Lactobacillus timonensis S7) were used. These comparator strains were also previously isolated and purified from “jiangshui” by our laboratory. To ensure phenotypic and genetic stability, the bacteria used for all experiments were limited to within three passages from the original − 80 °C glycerol stocks. All strains were cultivated in de Man, Rogosa and Sharpe (MRS) broth at 37 °C without agitation. Bacterial growth was routinely monitored by measuring the optical density at 600 nm (OD₆₀₀). Samples were harvested during the logarithmic growth phase (OD₆₀₀ between 0.8 and 1.0) by centrifugation (8,000 rpm for 5 min at 4 °C) and washed twice with sterile phosphate-buffered saline (PBS). For in vivo administration, the L. fermentum GR-3 bacterial pellet was resuspended in sterile PBS to a final concentration of 1 × 101⁰ CFU/mL (to deliver a target dose of 1 × 10⁹ CFU per 100 μL administration volume). To ensure batch-to-batch consistency, the actual delivered dose was rigorously verified prior to each administration through serial dilution and colony enumeration on MRS agar plates incubated at 37 °C for 48 h. Probiotic Properties of L. fermentum GR-3 The in vitro antioxidant capacities of L. fermentum GR-3 and the comparative strains (S1–S7) were evaluated utilizing both total antioxidant capacity (T-AOC) and DPPH radical-scavenging activity. These parameters were measured using standardized commercial assay kits (Cat. No. A015—2—1 for T-AOC and Cat. No. A153—1—1 for DPPH; Nanjing Jiancheng Bioengineering Institute, Nanjing, China) according to the manufacturer’s instructions. Briefly, T-AOC was determined by the ABTS method measuring absorbance at 405 nm, while DPPH radical-scavenging activity was measured at 517 nm. Cell-free supernatants were prepared by centrifuging overnight bacterial cultures (10,000 × g, 4 °C, 10 min). Following the kit manufacturer’s guidelines, all measurements were performed in triplicate [30, 31]. To assess bile tolerance, L. fermentum GR-3 was cultured in MRS broth supplemented with three specific types of bile salts: bovine bile salt (Product No. S30875), porcine bile salt (CAS No. 8008—63—7), and sodium taurocholate (CAS No. 145—42—6). All bile salts were purchased from Shanghai Yuanye Bio-Technology Co., Ltd. (Shanghai, China). Each was tested at final concentrations of 0.5% and 1.0% (w/v). For pH resistance testing, the strain was cultured in MRS broth adjusted to pH values ranging from 2.0 to 7.0. All media were inoculated with overnight cultures (1% v/v) and incubated statically at 37 °C for 12 h to maintain a microaerobic environment. Bacterial growth kinetics were continuously monitored by measuring the optical density at 600 nm (OD₆₀₀) using a Thermo Scientific Multiskan GO microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). Uninoculated MRS broth (with corresponding pH or bile salts) was used as the blank to correct for background absorbance. Because OD₆₀₀ accounts for both live and dead cellular mass, this metric was used solely to evaluate relative growth trends rather than absolute survival. The relative growth rate was subsequently calculated and expressed as a percentage by comparing the net OD₆₀₀ (sample absorbance minus blank) of the bacteria grown under stress conditions to the net OD₆₀₀ of the control group (bacteria grown under optimal conditions in standard, unadjusted MRS broth) [32, 33]. The gastrointestinal survival capabilities of L. fermentum GR-3 were assessed using simulated digestive conditions. The strain was grown statically in MRS broth (37 °C, 24 h), harvested by centrifugation (10,000 × g, 4 °C, 10 min), and washed three times with PBS (0.1 M, pH 7.2). Subsequently, the bacteria were resuspended in 0.5% NaCl, adjusted to an OD₆₀₀ of 1.0 ± 0.05, and exposed to filter-sterilized solutions simulating gastric fluid (3 mg/mL pepsin, 0.5% NaCl, pH 2.8) or intestinal fluid (1 mg/mL pancreatin, 0.5% NaCl, pH 8.0). Crucially, to accurately quantify true probiotic survivability and avoid the overestimation inherent to OD-based methods, absolute viability was determined strictly via CFU enumeration. Bacterial survival was quantified by serially diluting and plating samples on MRS agar at both the initial time point (0 h) and after 12 h of incubation at 37 °C [34]. Cell Line and Culture Conditions The MC38 mouse colorectal cancer cell line was obtained from Prof. Weilin Jin’s laboratory at the First Hospital of Lanzhou University. The cell cultures were grown in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, and a mix of antibiotics consisting of 100 U/mL penicillin and 100 μg/mL streptomycin. Cultures were maintained in a controlled environment (37 °C, 5% CO₂, humidified atmosphere), and subculturing was performed every 2—3 days. This subculturing frequency was strictly maintained to prevent contact inhibition and ensure that the MC38 cells remained constantly in the exponential growth phase, thus preserving optimal cell viability. Cell confluence was maintained between 30 and 80%. To standardize tumor development, cells from passages 3—7 at 70—80% confluence were utilized for tumor implantation procedures [35]. This specific passage window was selected because cells from very early passages (e.g., passages 1—2) may not have fully recovered from cryopreservation stress, whereas extended in vitro passaging can lead to genetic drift, phenotypic alterations, and variable tumorigenicity. Using passages 3—7 ensures robust, consistent, and reproducible in vivo tumor models. Tumor Model Establishment in Mice All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Lanzhou University (Approval No. EAF2024048). Thirty female C57BL/6 mice (4–6 weeks old) were obtained from the Laboratory Animal Research Center of Lanzhou University and housed in individually ventilated cages (six mice per cage) under controlled conditions (22 ± 2 °C, 50 ± 10% relative humidity, 12 h light/dark cycle), with free access to standard rodent chow and sterile water. Animals were acclimated for one week prior to experimentation. For the tumor model establishment, 24 mice were randomly selected for a subcutaneous injection of 100 μL of cell suspension (1 × 10⁶ cells) into the right axillary region using a 26-gauge needle, while the remaining six mice served as healthy controls (CK). MC38 cells were harvested at 80—90% confluence, washed twice with ice-cold PBS, and resuspended to achieve a concentration of 1 × 10⁷ cells/mL. Cell viability was confirmed using the trypan blue exclusion assay, with viability consistently exceeding 95%. Tumor formation was monitored daily, and tumor dimensions were measured using digital calipers. A stable tumor model was established approximately 7 days post-injection, when palpable tumor volumes reached around 50 mm3. Tumor volume was calculated using the formula: Volume = 0.5 × length × width2. According to the approved IACUC protocol, a maximum in vivo tumor volume of 1500 mm3 was defined as the ethical humane endpoint [36, 37]. Animals were closely monitored throughout the study, and humane endpoints were applied in strict accordance with institutional animal welfare guidelines, including criteria such as excessive tumor burden, impaired mobility, or signs of general distress. Animal Experimental Design Following a one-week acclimation period (from Day −14 to Day −7), mice were subcutaneously injected with 100 μL of cell suspension (1 × 10⁶ MC38 cells) on Day −7 to establish the tumor model. The healthy control (CK) mice were subcutaneously administered an equal volume of sterile PBS (s.c. PBS). After a seven-day monitoring period (Day −7 to Day 0) to confirm successful tumor establishment, mice exhibiting palpable tumors (approximately 50 mm3) were stratified by baseline tumor volume and body weight. As illustrated in the experimental timeline, mice were then randomly assigned to five distinct groups (n = 6 per group, total Group = 5): (1) CK (healthy control); (2) Tumor (tumor-bearing, treat-free control); (3) PD1 (anti-PD-1 monotherapy); (4) PD1/GR-3 (anti-PD-1 combined with L. fermentum GR-3); and (5) GR-3 (L. fermentum GR-3 monotherapy). Therapeutic interventions commenced on Day 0. In the PD1 and PD1/GR-3 groups, mice received intraperitoneal injections of anti-PD-1 monoclonal antibody (100 μg per mouse; BioXcell, #BE0146—5MG, RRID: AB_10949053, Rat IgG2a) every three days (Days 0, 3, 6, 9, and 12). For the microbiota intervention, the PD1/GR-3 and GR-3 monotherapy groups received a daily oral gavage (p.o.) of L. fermentum GR-3 (suspended in 100 μL PBS, corresponding to the dosage indicated in the treatment schematic) from Day 0 until harvest on Day 16. This specific probiotic dosage was selected based on established protocols in murine microbiome studies [38]. To minimize observer bias, tumor growth and treatment responses were monitored through serial measurements of tumor dimensions using digital calipers on days 7, 10, 13, and 16 post-treatment by an investigator blinded to group allocations. Similarly, all downstream analytical procedures (including 16S rRNA sequencing, qPCR, and ELISA) were conducted on coded samples by personnel blinded to the group identifiers. Mice were humanely harvested on Day 16. Animals were euthanized earlier if tumor burden approached the approved ethical limit (strictly defined as a tumor volume exceeding 1,500 mm3 or severe tumor ulceration) or if signs of severe distress were observed, in accordance with the approved IACUC protocol. Sample Collection and Procession All sample collection procedures were conducted under strictly sterile conditions. Fecal samples were collected at three critical timepoints: prior to tumor inoculation (Day −7), 7 days post-tumor establishment but prior to treatment initiation (Day 0), and at the experimental endpoint (Day 16). For fecal collection, individual mice were placed in clean, sterile cages overnight to allow for spontaneous defecation. Fresh fecal pellets were immediately collected, flash-frozen in liquid nitrogen, and stored at −80 °C until analysis. Blood samples were collected via retro-orbital bleeding. After collection, samples were allowed to clot at room temperature for 30 min, then centrifuged at 3000 × g for 15 min at 4 °C. Serum was carefully separated, aliquoted to minimize freeze–thaw cycles, and stored at −80 °C. Following euthanasia by cervical dislocation, subcutaneous tumors, spleens, colons, livers, and kidneys were carefully excised. Crucially, for intratumoral microbiota analysis (a typical low-biomass environment), tumor tissues were harvested using DNA-free sterile instruments within a clean bench to strictly prevent environmental contamination. The external tumor surface was trimmed to remove potential skin-associated microbial carryover. Each tumor was divided aseptically: one portion was immediately placed in a sterile, DNA-free microtube, snap-frozen in liquid nitrogen, and stored at − 80 °C for subsequent 16S rRNA gene sequencing. The remaining tumor tissue was divided for additional analyses. One portion was fixed in 4% paraformaldehyde for histological evaluation, including TUNEL staining for apoptosis assessment. These fixed samples were sent to Wuhan Servicebio Technology Co., Ltd. (Wuhan, China) for sectioning and staining. The remaining unfixed portions of the tumor, spleen, and colon tissues were stored at − 80 °C for subsequent gene expression and cytokine analyses. Real-Time Quantitative PCR (RT-qPCR) Total RNA isolation was conducted using the RNAiso Plus Kit (TaKaRa, China), incorporating supplementary quality control measures. RNA quality was evaluated by examining both concentration and purity using a NanoDrop spectrophotometer (ensuring A₂₆₀/A₂₈₀ ratios of 1.8–2.0) and confirming structural integrity via agarose gel electrophoresis. Subsequently, first-strand cDNA was synthesized using the PrimeScript RT reagent Kit (TaKaRa), processing 1 μg of total RNA per reaction. Gene expression analysis was performed on a QuantStudio 5 platform (Thermo Fisher Scientific, Waltham, MA, USA) using the SYBR Premix Ex Taq (Cat. No. RR820A, TaKaRa). The reactions were set up in a total volume of 20 μL, containing 10 μL of 2 × SYBR Green mix, 0.4 μL each of the forward and reverse primers (10 μM), 2 μL of template cDNA, and 7.2 μL of nuclease-free H₂O. The PCR cycling conditions were as follows: an initial denaturation step at 95 °C for 30 s, followed by 40 cycles of denaturation at 95 °C for 10 s and annealing/extension at 60 °C for 30 s. Melt curve analysis was included to confirm amplification specificity. In this study, we quantified the expression of GzmB and Foxp3 in both colon and spleen tissues, as well as ZO-1 and Claudin3 in colon tissues. The primer sequences for these specific target genes were derived from a previous study [39], and the detailed sequences are listed in Table S1.Relative gene expression was calculated using the 2^ − ΔΔCt method with β-actin as the internal reference for normalization, using mean values from three technical replicates per sample. Enzyme-Linked Immunosorbent Assay (ELISA) Using Elabscience ELISA kits, we analyzed four key cytokines (IFN-γ, IL-6, IL-10, TNF-α and IL-17A) in both tissue and serum samples. Tissue preparation involved homogenizing 0.1 g samples in ice-cold PBS supplemented with protease inhibitors, followed by high-speed centrifugation (12000 g, 15 min, 4 °C) to obtain clear supernatants. Measurements were conducted in duplicate against standard curves, and absorbance values were read at 450/570 nm on a Bio-Rad microplate reader (Bio-Rad Laboratories, Hercules, CA, USA). Cytokine concentrations were quantified using four-parameter logistic regression, with working ranges spanning 16–1,000 pg/mL (IFN-γ), 16–500 pg/mL (IL-6), 16–1,000 pg/mL (IL-10), 16–1,000 pg/mL (IL-17A), and 16–2,000 pg/mL (TNF-α). Samples exceeding these detection ranges were appropriately diluted and re-assayed. 16S rRNA Gene Sequencing and Bioinformatics Analysis Microbial genomic DNA extraction from both fecal and tumor samples was performed using the E.Z.N.A.® soil DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer’s instructions. DNA integrity was verified by agarose gel electrophoresis, and concentration was measured via a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). Extracted samples were maintained at −80 °C prior to downstream processing. Bacterial community profiling targeted the hypervariable V3-V4 region of the 16S rRNA gene, employing the 338 F primer (5’-ACTCCTACGGGAGGCAGCAG-3’) and the 806R primer (5’-GGACTACHVGGGTWTCTAAT-3’) in a T100 Thermal Cycler (Bio-Rad Laboratories, Hercules, CA, USA). The 20 µL amplification mixture contained FastPfu components (buffer, polymerase, dNTPs), primers, and template DNA. The thermal cycling program consisted of an initial denaturation at 95 °C for 3 min, followed by 27–30 cycles of denaturation at 95 °C for 30 s, annealing at 55 °C for 30 s, and extension at 72 °C for 45 s, culminating in a final extension at 72 °C for 10 min. After gel purification using the Omega PCR Cycle Kit (Omega Bio-tek) and quantification via a Qubit 4.0 Fluorometer (Thermo Fisher Scientific), the purified amplicons were pooled in equimolar amounts and subjected to paired-end sequencing (PE300) on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) at Tsingke Biotechnology Co., Ltd. (Beijing, China). The subsequent bioinformatics analysis was conducted utilizing the standard pipelines provided by Tsingke Biotechnology, augmented with strict quality control measures [40]. Following sequencing, raw FASTQ files were demultiplexed and quality-filtered using fastp, and paired-end reads were merged using FLASH. To achieve high-resolution microbiome profiling, the DADA2 plugin within the QIIME 2 framework was employed to denoise sequences, remove chimeras, and generate Amplicon Sequence Variants (ASVs). Taxonomic assignment of ASVs was performed using a Naive Bayes consensus taxonomy classifier trained on the standard SILVA database (release 138). Crucially, given the low-biomass nature of the intratumoral microbiome, stringent contamination control was applied. Putative contaminant ASVs were computationally identified and filtered out based on the sequencing data from negative controls (sterile water and DNA extraction blanks) using the decontam R package. Quantification of L. fermentum Colonization To evaluate the colonization efficacy of L. fermentum GR-3, quantitative real-time PCR (qPCR) was performed on samples from mouse feces and distal colonic mucosa. Upon completion of the in vivo experiments, approximately 1 cm segments of the distal colon were promptly excised and thoroughly rinsed with sterile 0.9% saline. These segments were longitudinally opened, and the mucosal layer was gently scraped to collect mucosal samples. Genomic DNA from fecal and colonic mucosal samples was extracted using the Stool DNA Kit (Tiangen, Beijing, China). All extracted DNA samples were preserved at − 80 °C prior to analysis. The qPCR assays were conducted on a QuantStudio 5 Real-Time PCR System (Applied Biosystems, CA, USA) using specific primers targeting L. fermentum and universal primers for total bacteria (Figure S1) [41]. Standard curves were meticulously generated using tenfold serial dilutions of plasmid DNA containing the specific target sequences at known concentrations. Ultimately, the colonization level of L. fermentum GR-3 in the colonic mucosa and feces was estimated by determining the ratio of L. fermentum to total bacteria, calculated based on the established standard curves [42]. Quantification of Short-Chain Fatty Acids (SCFAs) The concentrations of SCFAs in mouse feces were quantified using gas chromatography with flame ionization detection (GC-FID). The fecal extraction procedure was performed according to previously described methods [43]. Briefly, approximately 600 mg of frozen fecal sample was thawed and completely resuspended in 1.2 mL of PBS (pH 7.4). The suspension was vigorously vortexed and subsequently centrifuged at 15,000 × g for 15 min at 4 °C. Thereafter, 200 μL of the supernatant was collected, filtered through a 0.22 μm syringe filter, and acidified by the addition of 100 μL of 50% (v/v) H₂SO₄. Following a 2-min incubation at room temperature, the acidified mixture was subjected to liquid–liquid extraction using 0.4 mL of diethyl ether. The upper organic phase was carefully collected and analyzed using a Trace 1300 Gas Chromatograph equipped with an FID detector (Thermo Fisher Scientific, Singapore). Chromatographic separation was achieved on a TG-WAXMS capillary column (30 m × 0.25 mm, Thermo Scientific, USA). Nitrogen was utilized as the carrier gas. The injection port and detector temperatures were maintained at 240 °C and 250 °C, respectively. The oven temperature program was initiated at 100 °C (held for 1 min), gradually ramped to 150 °C at a rate of 5 °C/min, and further increased to 240 °C at 20 °C/min, with a final hold for 5 min. The absolute concentrations of specific SCFAs in each sample were determined by comparing their peak areas and retention times against standard reference solutions. In Vivo Intestinal Permeability Assay To evaluate intestinal permeability, mice were fasted for 4 h prior to the oral administration of fluorescein isothiocyanate (FITC)-dextran (4 kDa, Sigma-Aldrich, St. Louis, MO, USA). The FITC-dextran was dissolved in PBS (pH 7.4) and administered via oral gavage at a dose of 100 mg/kg body weight. Following an additional 4 h of fasting post-gavage, blood samples were collected, and serum was isolated. Subsequently, 100 μL of each serum sample was mixed with an equal volume of PBS. The fluorescence intensity was immediately quantified using a Varioskan Flash microplate reader (Thermo Fisher Scientific, Waltham, MA, USA) at an excitation wavelength of 485 nm and an emission wavelength of 535 nm. The absolute concentration of FITC-dextran in the serum was calculated based on a standard curve generated through serial dilutions of FITC-dextran in PBS. Flow Cytometry Analysis of the Tumor Immune Microenvironment The flow cytometry assays, including tissue processing, cell staining, and data acquisition, were performed by Gansu Solivia Biotechnology Co., Ltd. (Lanzhou, China). Briefly, to comprehensively profile the immunological alterations within the tumor microenvironment, single-cell suspensions from subcutaneous tumor tissues were prepared. The excised tumors were minced into small pieces and enzymatically digested with Collagenase IV and DNase I at 37 °C for 45 min, followed by filtration through a 70-μm nylon cell strainer. To ensure rigorous analysis and eliminate non-specific binding, cells were initially incubated with a Fixable Viability Dye to exclude dead cells, alongside a purified anti-mouse CD16/32 antibody (Fc block) for 15 min at 4 °C. For the assessment of surface markers, the pre-blocked cells were incubated in the dark at 4 °C for 30 min with optimized cocktails of fluorochrome-conjugated antibodies. The surface staining panels included markers for T cell subsets (anti-CD45, anti-CD3, anti-CD4, anti-CD8, and anti-CD25), macrophages (anti-F4/80, anti-CD11b, and anti-CD86), and dendritic cells (DCs) (anti-CD45, anti-CD3, anti-CD11c, and anti-MHC-II). For the detection of intracellular and intranuclear targets, cells were subsequently subjected to fixation and permeabilization using the Foxp3/Transcription Factor Staining Buffer Set according to the manufacturer’s instructions. Following permeabilization, cells were stained with antibodies against Foxp3 (to identify regulatory T cells), Granzyme B (to evaluate CD8⁺ T cell cytotoxic activation), and CD206 (to identify M2 macrophage polarization). After extensive washing, the stained samples were acquired on a BD FACSAria III flow cytometer (BD Biosciences, San Jose, CA, USA). All flow cytometry data were subsequently analyzed using FlowJo software (version 10.x, BD Biosciences) [44]. Statistical Analysis Statistical analyses and data visualization were performed using GraphPad Prism software (version 8.0, San Diego, CA, USA). All quantitative data are expressed as the mean ± standard deviation (SD). The normality of the data distribution was initially evaluated using the Shapiro–Wilk test. For normally distributed data, statistical differences between two independent groups were analyzed using the unpaired Student’s t-test, whereas comparisons among three or more groups were assessed using one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test. In cases where the data did not meet the assumption of normality, non-parametric equivalents (the Mann–Whitney U test or Kruskal–Wallis test) were applied. For longitudinal data, such as tumor volume and body weight monitored over time, statistical differences were determined using a two-way ANOVA with Sidak’s multiple comparisons test. Results and Discussion In Vitro Probiotic Characterization of L. fermentum GR-3 as a Candidate Enhancer of PD-1 Immunotherapy The integration of probiotics with immunotherapy has recently gained considerable attention as a strategy to enhance cancer treatment efficacy. In this study, L. fermentum GR-3 was selected as a candidate to enhance the efficacy of PD-1 blockade based on its superior functional properties. First, phylogenetic analysis confirmed the evolutionary classification of L. fermentum GR-3 within the Limosilactobacillus clade (Fig. 1A). Scanning electron microscopy (SEM) showed that L. fermentum GR-3 possesses a typical rod-shaped morphology, with a smooth surface and rounded ends, consistent with the morphological features of Lactobacillus species (Fig. 1B). Physiological profiling demonstrated robust growth over a wide pH range of 4.0–7.0, with optimal proliferation observed at a neutral pH of 7.0 (Fig. 1C). Furthermore, L. fermentum GR-3 exhibited exceptional and specific tolerance to bile salts. While common physiological bile components, such as bovine and porcine bile salts, caused near-complete growth inhibition (Fig. 1D–E), the strain showed remarkable resistance to sodium taurocholate, with only minimal growth inhibition (19.6% and 14.9% at 0.5% and 1.0% concentrations, respectively) (Fig. 1F). Demonstrating its capacity to survive the transit through the host digestive tract, L. fermentum GR-3 maintained high viability under simulated gastrointestinal conditions. Bacterial counts increased from 3.06 × 106 CFU mL−1 to 1.07 × 107 CFU mL−1 in simulated gastric fluid (pH 2.8) and sustained 1.02 × 107 CFU mL−1 in simulated intestinal fluid (pH 8.0) after 12 h of exposure (Fig. 1G). To evaluate its antioxidant potential, we compared L. fermentum GR-3 against several laboratory-preserved isolates. L. fermentum GR-3 exhibited the highest total antioxidant capacity (T-AOC) and DPPH radical–scavenging efficiency (Fig. 1H–I), highlighting its potent capability to mitigate oxidative stress. Collectively, this demonstrated resistance to harsh gastrointestinal conditions (acid and specific bile salts), combined with superior baseline antioxidant activity, firmly supports the rationale for utilizing L. fermentum GR-3 as a robust probiotic enhancer for cancer immunotherapy. Identification and in vitro probiotic characterizations of L.fermentum GR-3. (A) 16S rRNA-based phylogenetic tree. (B) Scanning electron microscopy (SEM) morphology (Scale bar = 3 μm). (C) Growth kinetics across a pH range of 2.0–7.0. (D–F) Tolerance to specific bile salts: bovine bile salt (D), porcine bile salt (E), and sodium taurocholate (F) at indicated concentrations. (G) Viability in simulated gastric (pH 2.8) and intestinal (pH 8.0) fluids over 12 h. (H, I) Comparative antioxidant capacities: T-AOC (H) and DPPH radical scavenging efficiency (I) among screened strains. Data are mean ± SD. *p < 0.05, ***p < 0.001, ****p < 0.0001 (t-test or ANOVA). Different lowercase letters denote significant differences (p < 0.05) by one-way ANOVA with Duncan’s test Recent studies have shown that certain probiotics can enhance the efficacy of PD-1 blockade therapy. For example, Akkermansia muciniphila improves the efficacy of anti-PD-1 antibody therapy in lung cancer by boosting CD8 + T cell infiltration and reducing immunosuppressive Tregs [45]. Bifidobacterium breve and immune checkpoint inhibitors exhibited improved progression-free survival rates in a clinical trial [46]. A key translational challenge is whether probiotics can withstand the harsh and variable conditions of the human gastrointestinal tract. The gastrointestinal tract present significant physiological barriers, with stomach pH as low as 1.5—3.5, the small intestine and colon exhibit pH ranging from 6.0—7.4 and 5.7—6.8, respectively. Meanwhile, bile salt concentrations vary from 0.2% to 2% in the small intestine [47, 48]. In this study, L. fermentum GR-3 survived under simulated gastrointestinal tract conditions, demonstrating robust growth at pH 3.0 and maintaining high viability under 0.5% sodium taurocholate exposure. This ability surpassed that of commonly used probiotic strains. For instance, L. rhamnosus GG showed only about 40% survival, Bifidobacterium longum maintained just 30% viability at pH 3.0, and L. plantarum survival fell to 25% at 1% bile salt concentration [49–51]. Moreover, L. fermentum GR-3 exhibited the highest total antioxidant capacity and DPPH radical-scavenging activity among all screened isolates, further supporting its physiological robustness. Antioxidant-active probiotics have been shown to enhance mucosal defense and reduce oxidative stress–induced immune suppression, which may contribute to improved responses to PD-1 blockade [52, 53]. These findings underscored L. fermentum GR-3’s exceptional ability to withstand environmental stressors, providing a foundation for its potential application in cancer immunotherapy. L. fermentum GR-3 Synergistically Enhances the Pro-apoptotic and Antitumor Efficacy of PD-1 Blockade in Vivo Following the experimental schedule shown in Fig. 2A, subcutaneous MC38 tumors were established in C57BL/6 mice. Therapeutic interventions initiated on Day 0 when nodules reached a mean volume of approximately 50 mm3 (Fig. 2B). Notably, an independent cohort receiving oral L. fermentum GR-3 monotherapy was included to assess its baseline effects; it exhibited a modest but statistically non-significant tumor-inhibitory trend compared to the untreated Tumor group (Tumor control: 1.94 ± 0.49 g vs. GR-3: 1.48 ± 0.48 g), confirming that robust efficacy requires synergy with immunotherapy (Figure S1). Systemic toxicity was evaluated through continuous body weight monitoring. By Day 16, mice in the untreated Tumor group and anti-PD-1 monotherapy (PD1) group experienced substantial weight losses of 12.6% and 10.1%, respectively, compared to the healthy control (CK) mice (p < 0.01). In striking contrast, tumor-bearing mice treated with the combination regimen (PD1/GR-3) maintained the most stable body weight, with only a marginal 6.4% reduction, suggesting the excellent tolerability of the synergistic therapy (Fig. 2C). Tumor growth kinetics (Fig. 2D) revealed that the PD1/GR-3 combination group exhibited the most profound suppression of tumor progression. The final mean tumor volume in this group was restricted to 210.8 ± 96.1 mm3, corresponding to a 51.7% reduction relative to the untreated Tumor group and a remarkable 37.1% additional reduction over PD1 monotherapy. These macroscopic observations were highly corroborated by the final excised tumor weights (Fig. 2E and F). The mean tumor burden plummeted from 2.46 ± 0.32 g in the Tumor group and 1.56 ± 0.23 g in the PD1 group to a mere 1.09 ± 0.28 g in the PD1/GR-3 combination group. To elucidate the cellular mechanisms underlying this enhanced tumor suppression, intratumoral apoptosis was evaluated using TUNEL staining (Fig. 2G). Within the tumor microenvironment, the untreated Tumor group exhibited minimal spontaneous apoptotic activity (8.2 ± 1.2%). While PD-1 blockade alone moderately elevated the apoptotic rate to 26.4 ± 2.9%, the co-administration of L. fermentum GR-3 (PD1/GR-3) dramatically amplified this effect, driving the highest level of apoptosis globally (52.5 ± 3.8%, p < 0.0001) (Fig. 2H). Together, these clinically translatable findings provide compelling in vivo evidence that oral supplementation of L. fermentum GR-3 acts synergistically with anti-PD-1 therapy to potentiate robust antitumor and pro-apoptotic responses. L. fermentum GR-3 synergizes with anti-PD-1 therapy to suppress tumor growth and induce apoptosis in vivo. (A) Schematic of the experimental design. (B) Measurement of subcutaneous tumors. (C) Body weight changes during the 16-day treatment. (D) Tumor growth curves. (E) Macroscopic images of excised tumors on Day 16 (Scale bar = 1 cm). (F) Final tumor weights. (G) Representative TUNEL staining of tumor sections (TUNEL, red; DAPI, blue). (H) Quantification of TUNEL-positive cells. Data are mean ± SD (n = 6 per group). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (one-way or two-way ANOVA) Cancer progression is driven by the uncontrolled proliferation and invasion of malignant cells, which compromise healthy tissues and impair normal physiological functions [54]. The rapid proliferation of cancer cells consumes the host’s metabolic resources, leading to systemic energy depletion, immune dysfunction, and progressive weight loss [55]. Recent advances highlight the role of gut microbiota in modulating cancer therapy efficacy, with probiotic supplementation emerging as a promising strategy to enhance treatment outcomes. For example, patients with higher gut microbiota diversity show approximately 30% better response to immunotherapy compared to those with dysbiotic microbial profiles [56]. Similarly, a balanced gut microbiota is associated with 30% improved tumor control [57]. These observation support our hypothesis that specific probiotics can potentiate PD-1 blockade therapy. In colorectal cancer models, weight loss serves as a key indicator of tumor progression and systemic metabolic disturbance [58]. Patients receiving PD-1 monotherapy typically experience a 5–15% reduction in body weight [59]. Notably, Notably, our PD1/GR3 combination therapy significantly mitigated cancer-induced weight loss to 6.4%, consistent with previous studies where improved weight maintenance by 20–25% in tumor-bearing mice [24]. L. fermentum GR-3 monotherapy alone did not significantly influence body weight or tumor growth, indicating its limited standalone anti-tumor activity. Studies have reported that Bifidobacterium longum combined with anti-PD-1 therapy reduced tumor volume by approximately 40% in colorectal cancer, while Lactobacillus plantarum with immunotherapy led to a 45% reduction in melanoma models [60]. Strikingly, in our model, the L. fermentum GR-3 and anti-PD-1 antibody combination achieved a 51.7% tumor volume reduction, surpassing both anti-PD-1 antibody monotherapy and previously reported probiotic combinations, underscoring a potent synergistic interaction. The absence of anti-tumor activity with L. fermentum GR-3 alone suggests that L. fermentum GR-3 potentiates PD-1 blockade through mechanisms that require concurrent immune checkpoint inhibition. This interpretation is further corroborated by significantly elevated tumor cell apoptosis in the PD1/GR3 group [61]. Together, these results indicate that L. fermentum GR-3 does not exert direct cytotoxic effects but acts as an immunomodulatory enhancer that amplifies PD-1–mediated anti-tumor immunity, positioning it as a novel combinatorial partner for immune checkpoint therapy. L. fermentum GR-3 Potentiates PD-1 Blockade to Extensively Remodel the Tumor Immune Microenvironment To access the enhanced anti-tumor efficacy of the PD1/GR-3 combination, we profiled the innate and adaptive immune landscapes within the tumor microenvironment (TME) using flow cytometry (Fig. 3). Analysis of the adaptive lymphoid compartment revealed that while the PD1 group exhibited a slightly lower CD4⁺/CD8⁺ T cell ratio relative to the Tumor group (0.96 vs. 1.55), the PD1/GR-3 group showed a more pronounced shift toward a cytotoxic phenotype, further reducing this ratio to 0.026 (p < 0.0001) (Fig. 3A and B). Parallel to this, immunosuppressive regulatory T cells (Tregs, CD25⁺Foxp3⁺) were partially diminished in the PD1 group compared to the Tumor group (3.41% vs. 5.82%), but the PD1/GR-3 treatment further suppressed Treg infiltration to 2.32% (p < 0.05) (Fig. 3C and D). This relief of immunosuppression was accompanied by enhanced effector T cell function. The frequency of Granzyme B (GzmB)⁺ CD8⁺ T cells in the PD1/GR-3 group reached 40.4%, which was significantly higher than both the PD1 group (19.6%, p < 0.0001) and the Tumor group (11.9%) (Fig. 3E and F). Within the innate myeloid compartment, both the PD1 and GR-3 groups showed a moderate increase in the maturation of tumor-infiltrating dendritic cells (DCs). However, the PD1/GR-3 treatment yielded a synergistic effect, with MHC-II⁺ DCs reaching 56.0%, outperforming both the Tumor (22.1%) and PD1 (31.6%) groups (p < 0.0001) (Fig. 3G, H). Furthermore, the PD1/GR-3 regimen effectively reprogrammed tumor-associated macrophages (TAMs). Compared to the monotherapy groups, the PD1/GR-3 treatment enriched the proportion of pro-inflammatory M1-like macrophages (CD86⁺; Tumor: 6.42%, PD1: 12.7%, PD1/GR-3: 20.8%, p < 0.0001; Fig. 3J) and concurrently diminished immunosuppressive M2-like macrophages (CD206⁺; Tumor: 8.67%, PD1: 5.59%, PD1/GR-3: 2.57%, p < 0.01; Fig. 3K) (Fig. 3I-K). Together, these data indicate that GR-3 does not merely exert standalone immunomodulatory effects, but effectively cooperates with PD-1 blockade to convert an immunosuppressive TME into an immune-active state. L. fermentumGR-3 potentiates PD-1 blockade to comprehensively remodel the tumor immune microenvironment. Representative flow cytometry plots and corresponding quantifications of intratumoral immune populations: (A, B) The ratio of CD4⁺ to CD8⁺ T cells. (C, D) Regulatory T cells (Tregs, CD25⁺Foxp3⁺). (E, F) Granzyme B⁺ (GzmB⁺) CD8⁺ T cells. (G, H) Mature dendritic cells (DCs, MHC-II⁺). (I–K) Tumor-associated macrophage (TAM) polarization, showing representative plots (I) and the percentages of pro-inflammatory M1-like (CD86⁺) (J) and immunosuppressive M2-like (CD206⁺) (K) TAMs. Data are mean ± SD (n = X per group). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 (one-way ANOVA) The clinical efficacy of immune checkpoint blockade (ICB) is frequently limited by the immunosuppressive nature of the tumor microenvironment, which is actively maintained by tolerogenic myeloid populations and Tregs [62]. Our phenotypic profiling suggests that L. fermentum GR-3 acts as an immunological sensitizer, bridging innate and adaptive anti-tumor immunity to alleviate ICB resistance. Importantly, the intratumoral myeloid compartment underwent functionally significant remodeling. Tumor-associated macrophages (TAMs) were repolarized toward an immunostimulatory M1-like phenotype, while dendritic cell (DC) maturation was concurrently enhanced, collectively fostering a TME more conducive to antitumor immunity [62]. Mechanistically, how does an intestinal probiotic remotely instruct the local TME? Recent studies indicate that the communication between gut commensals and distant tumor beds relies heavily on systemic circulating signals, notably microbial-derived metabolites that can epigenetically or metabolically reprogram infiltrating myeloid cells [15, 63]. Therefore, the profound M1 macrophage and DC priming observed in our model strongly implies that L. fermentum GR-3 colonization may initiate a systemic immunometabolic shift. Activation of the innate compartment is a prerequisite for effective ICB therapy; functional DCs are essential for tumor antigen cross-presentation, while M1 macrophages secrete chemokines that recruit effector T cells [64]. This restructuring of the TME creates a favorable feed-forward loop with PD-1 blockade. By reversing myeloid-driven immunosuppression, L. fermentum GR-3 likely restricts the survival niche for Tregs, consistent with the reduced Treg infiltration observed in the PD1/GR-3 group. The dual effect of relieving Treg-mediated suppression and restoring T cell function via PD-1 blockade ultimately facilitates the expansion and activation of CD8⁺ T cells, as evidenced by increased Granzyme B production. This is consistent with previous reports showing that modulating the gut microbiota with taxa like Bifidobacterium or Akkermansia enhances anti-PD-1 efficacy by increasing DC-mediated CD8⁺ T cell recruitment [22]. In summary, our findings demonstrate that oral supplementation with L. fermentum GR-3 can mitigate the intrinsic limitations of PD-1 monotherapy by systematically dismantling tumor-associated immune tolerance. L. fermentum GR-3 and PD-1 Blockade Jointly Alter the Intratumoral Microbial DNA Signatures Linked to Local Inflammation To determine whether local microbial signatures correlate with the immune awakening within the tumor microenvironment (TME), we conducted 16S rRNA sequencing on tumor tissues (Fig. 4). While α-diversity (Shannon index) remained unchanged across groups (p = 0.29), β-diversity analysis (NMDS) revealed a significant shift in microbial community profiles (PERMANOVA R2 = 0.56, p = 0.026), indicating that the PD-1/GR-3 combined therapy altered the intratumoral microbial DNA composition (Fig. 4A and B). Venn diagram analysis further illustrated a treatment-driven pattern of OTU distribution (Fig. 4C). Taxonomic profiling highlighted distinct compositional shifts. At the phylum level, Pseudomonadota (formerly Proteobacteria) signatures became highly abundant following PD-1–based therapies, rising from 18.0% in the Tumor control to 52.0% (PD1 alone) and peaking at 69.0% in the PD1/GR-3 combination group (Fig. 4D). Conversely, signatures of Bacillota (formerly Firmicutes) and Actinomycetota were nearly absent in the combination group. Genus-level analysis pinpointed that this shift was predominantly driven by a striking enrichment of Salmonella-assigned OTUs. While constituting only 6.0% in untreated tumors, Salmonella DNA signatures escalated to 40.0% with PD-1 monotherapy and 65.0% in the PD1/GR-3 group (Fig. 4E). Crucially, Spearman correlation analyses revealed a tight link between these detected microbial signatures and the local immune tone. Pro-inflammatory cytokines (TNF-α, IFN-γ) exhibited strong positive correlations with Pseudomonadota and specifically the genus Salmonella, while inversely correlating with Bacillota (Fig. 4F, G). Together, these findings indicate that the PD1/GR-3 therapy induces a specific alteration in tumor-associated microbial profiles, which closely parallels the immune-activated state of the TME. Alterations in intratumoral microbiota profiles following L. fermentum GR-3 and anti-PD-1 therapy. (A) Alpha diversity evaluated by the Shannon index. (B) Beta diversity visualized via Non-metric Multidimensional Scaling (NMDS) based on Bray–Curtis distances. (C) Venn diagram illustrating shared and unique operational taxonomic units (OTUs) among groups. (D, E) Relative abundance of major microbial taxa at the phylum (D) and genus (E) levels. (F, G) Spearman correlation heatmaps evaluating the associations between intratumoral cytokine levels and specific microbial taxa at the phylum (F) and genus (G) levels Recent studies have increasingly appreciated the presence of low-biomass microbial signatures within tumors, which can influence local immune tone and treatment responses [65, 66]. In our study, we observed a distinct shift in the intratumoral microbial DNA profile following PD-1-based therapies, characterized by a massive enrichment of Pseudomonadota (predominantly Salmonella) and the depletion of Bacillota, which align with previously reported intratumoral microbiomes [67]. Notably, the accumulation of Salmonella signatures within the PD1/GR-3-treated tumors may not necessarily reflect active local colonization. Instead, this enrichment could be functionally linked to the immune remodeling observed in our flow cytometry data. The combination therapy robustly activated tumor-infiltrating dendritic cells and profoundly polarized macrophages toward the highly phagocytic M1 phenotype. These activated local phagocytes might engulf and accumulate translocated bacterial components (such as DNA or lipopolysaccharides) originating from the leaky gut barrier, a common consequence of advanced tumors and systemic immunotherapy. Regardless of their viability, the presence of these Gram-negative bacterial components serves as potent Pathogen-Associated Molecular Patterns (PAMPs). Accumulating evidence suggests that Proteobacteria-derived components can reprogram antitumor immunity by providing strong immunostimulatory signals that continually activate local myeloid cells and sustain pro-inflammatory cytokine production [68]. This aligns with preclinical models where attenuated Salmonella derivatives-or even their isolated structural components—amplify local TNF-α and IFN-γ production to enhance solid tumor control [69, 70]. In this context, the strong positive correlation we observed between Salmonella signatures and intratumoral IFN-γ/TNF-α levels suggests a bidirectional cross-talk: the therapy-induced inflammatory TME may favor the accumulation of these specific microbial antigens via enhanced phagocytosis, which in turn acts as an immunological immunostimulatory trigger to maintain the “hot” tumor state. L. fermentum GR-3 Colonization Reshapes the Gut Microbiota and Boosts SCFA Metabolism Before investigating the downstream immunological alterations, we first confirmed the successful gastrointestinal engraftment of the orally administered probiotic. Quantitative PCR (qPCR) analysis revealed a significant enrichment of L. fermentum GR-3 relative to total bacteria in the feces of mice receiving L. fermentum GR-3 supplementation (both the GR-3 and PD1/GR-3 groups) compared to the CK group (p < 0.05). Following the confirmation of successful colonization, 16S rRNA sequencing was performed on pooled fecal samples to evaluate the broader effects on gut microbiota composition. Principal component analysis (PCA) revealed distinct microbial clustering among the groups. The microbial community structure of the PD1/GR-3 group positioned closer to the CK group than the untreated Tumor group, indicating a partial restoration of the microbiome (Fig. 5B). At the genus level (Fig. 5C), the Tumor group exhibited a decline in major beneficial genera, particularly Bacteroides (from 12.7% in CK to 5.5%) and Lactobacillus (from 19.3% to 0.4%). Conversely, opportunistic taxa such as Prevotellaceae_UCG-001 (from 0.8% to 4.0%) and Candidatus_Saccharimonas (from 1.3% to 5.9%) were enriched. Treatment with PD1/GR-3 partially reversed these compositional shifts. The relative abundance of Bacteroides increased to 34.7%, and Alloprevotella recovered to 3.5% (compared to 0.6% in the Tumor group). This was accompanied by a reduction in pro-inflammatory genera, with Prevotellaceae_UCG-001 decreasing to 0.35% and Candidatus_Saccharimonas to 0.9%. Furthermore, the PD1/GR-3 group showed an increased abundance of the Lachnospiraceae family (e.g., Lachnospiraceae_NK4A136 group enriching to 11.5%). Because taxa such as Bacteroides and Lachnospiraceae are associated with short-chain fatty acid (SCFA) production, we performed targeted absolute quantification of SCFA concentrations in the intestinal contents. Compared to the suppressed SCFA levels in the Tumor group, the PD1/GR-3 combination significantly elevated the mean concentrations of acetic acid (from 3.02 to 9.17 μM/g; Fig. 5D, p < 0.05), propionic acid (from 22.05 to 70.13 μM/g; Fig. 5E, p < 0.001), and butyric acid (from 15.79 to 26.07 μM/g; Fig. 5F, p < 0.01). Notably, the levels of propionic and butyric acids in the combination group were largely restored to levels comparable to the healthy CK baseline. A similar significant enrichment was observed for isobutyric acid, which increased from 20.60 μM/g in the Tumor group to 35.74 μM/g in the combination group (Fig. 5G, p < 0.05). Isovaleric acid showed a non-significant upward trend (from 11.67 to 18.37 μM/g; Fig. 5H), while valeric acid levels remained severely depleted across all tumor-bearing groups compared to the CK group (33.17 μM/g; Fig. 5I). Collectively, these targeted SCFA measurements indicate that the colonization of L. fermentum GR-3 functionally cooperates with PD-1 blockade to robustly restore and modulate intestinal SCFA metabolism. L. fermentum GR-3 modulates gut microbiota composition and enhances intestinal short-chain fatty acid (SCFA) production. (A) Relative abundance of L. fermentum GR-3 to total bacteria in the gut. (B) Principal component analysis (PCA) of gut microbial communities at the OTU level. (C) Taxonomic composition of the gut microbiota at the genus level. (D–I) Targeted quantification of intestinal SCFAs (μM/g): acetic acid (D), propionic acid (E), butyric acid (F), isobutyric acid (G), isovaleric acid (H), and valeric acid (I). Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant (one-way ANOVA) The clinical efficacy of immune checkpoint blockade (ICB), including anti-PD-1 therapy, is frequently hampered by tumor-induced systemic immunosuppression and gut dysbiosis [71]. A fundamental prerequisite for microbiome-targeted interventions is the successful engraftment of the administered strain. Our qPCR analysis confirmed that L. fermentum GR-3 successfully colonized the murine gastrointestinal tract, providing a stable foundation for subsequent ecological shifts. Correspondingly, our 16S rRNA sequencing revealed that untreated Tumor-bearing mice exhibited severe dysbiosis. This was characterized by a depletion of beneficial commensals, such as Bacteroides and Lactobacillus, alongside an enrichment of opportunistic pathogens. For instance, we observed increased levels of Prevotellaceae, a family often associated with chronic mucosal inflammation, and Candidatus_Saccharimonas, which has been implicated in colorectal cancer progression [72, 73]. Importantly, the combined PD1/GR-3 regimen actively counteracted this tumor-induced dysbiosis, steering the microbial community structure back toward a healthy baseline. While these compositional improvements indicate microbial recovery, the true mechanistic link connecting the gut lumen to the tumor microenvironment lies in functional microbial metabolites. Taxa prominently expanded by the PD1/GR-3 treatment, particularly Bacteroides and members of the Lachnospiraceae family, are renowned as keystone short-chain fatty acid (SCFA) producers in the mammalian gut [74–76]. Specifically, Lachnospiraceae are primary obligate butyrate producers. Our targeted quantification of SCFAs robustly validated this functional translation. We found that the combination therapy mitigated the SCFA depletion seen in the Tumor group, prompting a significant systemic restoration of key immunomodulatory metabolites, including acetic, propionic, and butyric acids. SCFAs are well-documented orchestrators of host immunity. They strengthen the intestinal epithelial barrier to prevent systemic inflammation and directly modulate leukocyte function [77, 78]. SCFAs, especially butyrate and propionate, can inhibit histone deacetylases (HDACs) to epigenetically promote the differentiation of effector T cells, while also directly fueling the metabolic fitness and cytotoxicity of CD8⁺ T cells. Therefore, the restoration of the intestinal SCFA pool provides a compelling mechanistic explanation for the enhanced intratumoral immune activation (e.g., the surge in Granzyme B⁺ CD8⁺ T cells) observed in our model. Recent clinical and preclinical studies highlight that modulating the microbiome with probiotics, such as Bifidobacterium, can enhance the efficacy of anti-PD-L1 therapies by boosting anti-tumor immunity [79, 80]. Consistent with these landmark findings, our data demonstrate that L. fermentum GR-3 acts as a potent immunoregulatory probiotic. By establishing stable colonization, restoring SCFA-producing consortia, and fueling a beneficial SCFA pool, L. fermentum GR-3 effectively shapes the microbiota-metabolite-immune axis. Ultimately, this targeted restoration of ecological and metabolic balance mitigates the intrinsic limitations of PD-1 monotherapy, positioning microbiome normalization as a central pillar for achieving synergistic anti-tumor efficacy. L. fermentum GR-3 Fortifies the Intestinal Barrier and Mitigates Systemic Toxicity While Amplifying Local Anti-Tumor Inflammation Systemic immune-related adverse events (irAEs), particularly immune-mediated colitis and systemic inflammation, remain major clinical hurdles for PD-1 blockade therapies. To evaluate the systemic safety and functional compartmentalization of the combination therapy, we assessed intestinal barrier integrity and multi-tissue immune profiles. We first examined whether the therapy-induced microbial shifts impacted gut barrier function. In vivo assessment using FITC-dextran (FD4) revealed that Tumor-bearing mice exhibited elevated intestinal permeability compared to the CK group, a defect that was effectively reversed by the PD1/GR-3 combination therapy (p < 0.05) (Fig. 6A). Consistent with this macroscopic barrier restoration, the colonic mRNA expression levels of critical tight junction proteins—ZO-1, Occludin, and Claudin-1—were significantly upregulated in the PD1/GR-3 group (p < 0.01 to p < 0.0001) (Fig. 6D-F). Furthermore, the systemic oxidative stress associated with advanced malignancy was alleviated. The PD1/GR-3 regimen enhanced the total antioxidant capacity (T-AOC) (p < 0.05) and strongly suppressed the accumulation of malondialdehyde (MDA) in the systemic circulation compared to the Tumor group (p < 0.001) (Fig. 6B, C). Crucially, this fortified gut barrier translated into systemic immune safety. While high expression of cytotoxic markers (e.g., Granzyme B) within the tumor is essential for efficacy, their aberrant hyperactivation in peripheral tissues drives auto-immune toxicity. Indeed, Tumor-bearing mice exhibited a systemic inflammatory state, characterized by elevated serum levels of TNF-α and IFN-γ (Fig. 6H). While PD1 monotherapy failed to resolve serum TNF-α levels and induced a severe, abnormal spike in GzmB mRNA expression within the spleen (indicative of systemic immune overactivation), the integration of L. fermentum GR-3 successfully attenuated these peripheral abnormalities (Fig. 6H, I). The PD1/GR-3 combination modulated serum TNF-α downwards and normalized splenic GzmB expression back to levels comparable to the healthy baseline. Importantly, the relative mRNA expressions of immune-related genes (Foxp3 and GzmB) within the colon tissues remained near the healthy baseline (1.0) across all treated groups (Fig. 6G, I), indicating that oral L. fermentum GR-3 administration does not trigger colonic hyperactivation or Treg-mediated immunosuppression locally. Strikingly, this peripheral immunoregulation did not compromise local anti-tumor efficacy. In stark contrast to the pacified spleen and serum, the intratumoral expression of the pro-inflammatory cytokine TNF-α was significantly amplified in the PD1/GR-3 group compared to the other treatments (p < 0.01) (Fig. 6J). A concurrent localized increase in IL-10 was also observed, likely reflecting a regulatory feedback mechanism in the highly active tumor microenvironment. Together, these findings demonstrate that L. fermentum GR-3 successfully uncouples systemic autoimmune toxicity from localized anti-tumor efficacy. Combination therapy restores intestinal barrier integrity, mitigates oxidative stress, and regulates systemic and local immune responses. (A) Intestinal permeability assessed by serum FITC-dextran 4 kDa (FD4) levels. (B, C) Systemic oxidative status evaluated by total antioxidant capacity (T-AOC) (B) and malondialdehyde (MDA) (C) concentrations. (D–F) Relative colonic mRNA expression of tight junction genes: Zo-1 (D), Occludin (E), and Claudin-1 (F). (G, I) Relative mRNA expression of Foxp3 (G) and GzmB (I) in the spleen and colon. (H, J) Cytokine profiles in the serum (H) and tumor microenvironment (J) determined by ELISA. Data are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; ns, not significant (one-way ANOVA). The horizontal dashed line (y = 1.0) represents the normalized baseline expression of the healthy control (CK) group A major challenge in the clinical application of immune checkpoint blockade (ICB) is the frequent occurrence of immune-related adverse events (irAEs). Therapies targeting the PD-1/PD-L1 axis often disrupt peripheral tolerance. While unlocking CD8⁺ T cell cytotoxicity (e.g., GzmB expression) is highly desirable within the tumor, ectopic hyperactivation of these cells in healthy peripheral tissues leads to systemic inflammation and immune-mediated colitis. These toxicities frequently necessitate the use of systemic corticosteroids, which can inadvertently blunt anti-tumor efficacy [81, 82]. Therefore, identifying combinatorial strategies capable of uncoupling localized anti-tumor efficacy from systemic toxicity is of paramount clinical importance. In this study, the oral administration of L. fermentum GR-3 achieved precisely this compartmentalized immune response. The systemic protection observed in our study is primarily rooted in the restoration of the intestinal barrier. Advanced malignancies and ICB therapy can induce increased intestinal permeability (“leaky gut”), facilitating the translocation of commensal bacteria and pro-inflammatory endotoxins into the systemic circulation [83, 84]. This drives systemic oxidative stress and inflammatory cascades. Our in vivo functional assays and transcriptional data explicitly demonstrate that L. fermentum GR-3 reduces intestinal permeability and upregulates crucial tight junction proteins. This barrier-fortifying property is a recognized characteristic of specific beneficial Lactobacillus strains [85]. By physically sealing the gut barrier, L. fermentum GR-3 restricts the aberrant systemic dissemination of inflammatory triggers. Consequently, this fundamentally attenuates systemic oxidative stress (MDA) and pacifies the aberrant splenic hyperactivation (downregulating the severe spike in peripheral GzmB induced by PD-1 monotherapy). Furthermore, the stable baseline expressions of GzmB and Foxp3 in colon tissues—remaining comparable to healthy individuals—indicate that L. fermentum GR-3 actively maintains colonic immune homeostasis, preventing ICB-associated auto-immune tissue damage while avoiding unwanted Treg accumulation. Remarkably, this stringent systemic pacification did not impede the local anti-tumor response. On the contrary, intratumoral TNF-α levels were significantly amplified in the PD1/GR-3 combination group. We postulate that this spatial dichotomy—systemic tolerance versus local cytotoxicity—is functionally driven by the restoration of gut-derived short-chain fatty acids (SCFAs). Emerging evidence suggests that circulating SCFAs, such as butyrate, can specifically reprogram the metabolic and epigenetic states of tumor-infiltrating CD8⁺ T cells, enhancing their localized effector functions without triggering non-specific peripheral inflammation [77, 86]. While state-of-the-art clinical strategies often rely on prophylactic cytokine blockade to separate toxicity from efficacy [87], our findings highlight L. fermentum GR-3 as a multifaceted biotherapeutic agent capable of fortifying the gut barrier, mitigating systemic irAEs, and orchestrating a highly localized anti-tumor immune response via the microbiota-metabolite axis. Conclusion This study demonstrates that the food-derived probiotic L. fermentum GR-3 synergistically enhances the efficacy of anti-PD-1 immunotherapy in a murine colorectal cancer model through the coordinated modulation of the microbiota-metabolite-immune axis. Rather than acting as a direct cytotoxic agent, L. fermentum GR-3 functions as an oral immunomodulator that remodels the host macroenvironment to significantly augment PD-1 blockade-mediated tumor regression. Mechanistically, L. fermentum GR-3 colonization reshapes the gut microbiome and enriches intestinal short-chain fatty acid (SCFA) pools. This targeted microbial-metabolic shift fortifies mucosal barrier integrity and effectively attenuates systemic oxidative stress. Consequently, this gut-driven physiological regulation translates into the robust reprogramming of the tumor microenvironment (TME). The combination therapy mitigates intratumoral immune tolerance by suppressing regulatory T cells, while concurrently stimulating local antigen-presenting cells (M1 macrophages and dendritic cells) to promote a potent CD8⁺ cytotoxic T cell response. Crucially, the synergistic PD1/GR-3 regimen orchestrates a spatially compartmentalized immune dynamic. It successfully decouples efficacy from toxicity by maximizing localized intratumoral immune activation while actively suppressing peripheral inflammatory cytokine cascades, thereby achieving a highly favorable therapeutic index. Furthermore, the observed remodeling of the intratumoral microbiota highlights an intricate gut-tumor microbial crosstalk, suggesting that local microbial signatures also actively participate in the therapy-induced immune response. From a translational perspective, L. fermentum GR-3 presents a highly viable and safe strategy for clinical application. Its origin from traditional fermented foods confers an established Generally Recognized as Safe (GRAS) status alongside a scalable production infrastructure, potentially bypassing the complex regulatory bottlenecks often associated with newly isolated, human-derived commensal strains. Collectively, this study provides a comprehensive mechanistic framework for the application of L. fermentum GR-3 as a microbiome-targeted synergistic biotherapeutic. While future clinical investigations and pan-cancer models are warranted to define its broader applicability, our findings firmly position food-grade microbial modulation as a safe, effective, and rapidly translatable strategy to optimize next-generation precision immunotherapy. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 854 KB) Acknowledgements We would like to thank the Central Laboratory of the School of Life Science, Lanzhou University, for providing various instruments and equipment. Abbreviations Malondialdehyde Total antioxidant capacity 2,2-Diphenyl-1-picrylhydrazyl Granzyme B Forkhead box protein P3 FITC-Dextran 4 kDa Major histocompatibility complex class II Dendritic cells Short-chain fatty acids Regulatory T cells Terminal deoxynucleotidyl transferase dUTP Nick End Labeling Tumor-associated macrophages Author Contributions W.X, Q.Y, J. J, H.Q, A.K, and E. S, shared responsibilities in visualization, formal analysis, and manuscript review/editing. Y.W and CL. M jointly led project conceptualization, provided supervision and resources, and managed various aspects including visualization and formal analysis, while contributing to manuscript revision. Xk. L oversaw the project administration, secured funding, provided supervision and resources, and participated in manuscript review. Funding This research was funded by National Natural Science Foundation of China (no.32370110), Gansu Province Science and Technology Major Project (no. 24ZDWA005), and the Postdoctoral Fellowship Program of CPSF (no. GZC20251840). Data Availability No datasets were generated or analysed during the current study. Declarations Ethics Approval The animal study protocol was sanctioned and overseen by the Animal Ethics Committee at the School of Life Sciences, Lanzhou University (No. EAF2024048). All applied methods were completed following the Committee’s regulations/guidelines. Competing Interests The authors declare no competing interests. Footnotes Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Chunlan Mao, Email: maocl@lzu.edu.cn. Ying Wu, Email: ywu18@lzu.edu.cn. References Associated Data Supplementary Materials Supplementary file1 (DOCX 854 KB) Data Availability Statement No datasets were generated or analysed during the current study.