Oral nano-delivery of a gut microbial metabolite enhances T cell stemness for cancer immunotherapy

Abstract

Gut microbial metabolites play crucial roles in regulating systemic immunity, but their mechanisms and limited drug-like properties remain unresolved. Here we report an oral nano-formulation that leverages gut microbial metabolites to modulate T cell metabolism and amplify antitumour immunity. Through an in vitro screening of gut microbial metabolites, we identified 3,4-dihydroxybenzoic acid that improved adoptive T cell therapy and enhanced CD8⁺ T cell stemness by suppressing glycolysis and regulating the Akt-mTORC1-Myc pathway. To harness the potency of 3,4-dihydroxybenzoic acid for systemic cancer immunotherapy, we engineered a 3,4-dihydroxybenzoic acid prodrug nano-emulsion, significantly increasing its oral absorption and half-life. In multiple murine tumour models, the oral nano-emulsion enhanced the expansion of antigen-specific, stem-like CD8⁺ T cells, sensitizing tumours to anti-PD-1 blockade and exerting robust antitumour efficacy. By integrating nanotechnology with microbial-metabolite-based immunotherapy, this study establishes a mechanistic link between the gut microbiota and T cell immunity, offering a promising approach for cancer immunotherapy.

A nano-emulsion that delivers the gut microbial metabolite 3,4-dihydroxybenzoic acid in a prodrug form enhances T cell stemness and boosts the efficacy of immunotherapy approaches in different murine cancer models.

Main

Immune checkpoint blockade (ICB) antibodies have revolutionized cancer immunotherapy, but their overall low response rates in patients remain a major challenge¹. Emerging evidence suggests that microbial metabolites are critical for the crosstalk between the gut microbiota and systemic immunity through the regulation of T cell phenotype and effector functions^2–5, suggesting their therapeutic potential to improve antitumour immunity. However, the role of microbial metabolites in T cell immunity and fitness remains largely unexplored, particularly in the context of ICB cancer therapy.

We previously reported that oral inulin gel enhances antitumour immunity and synergizes with ICB via the gut microbiota–short-chain fatty acids (SCFAs)–CD8⁺ T cell axis⁶. Interestingly, we observed that inulin gel substantially altered a library of non-SCFA microbial metabolites, leading to our hypothesis that these metabolites may control T cell immunity. In particular, gut microbial metabolites may serve as a rich source of new compounds for drug development, as they offer favourable safety profiles, well-defined chemical structures, scalability for manufacturing and quality control⁷. However, many microbial metabolites are highly hydrophilic with a low molecular weight, showing poor oral absorption and pharmacokinetic (PK)⁸ properties with a short plasma half-life and high clearance rate. Their clearance from the systemic circulation can occur within minutes, thereby limiting their bioavailability and effective regulation of T cell immunity in vivo⁹.

In this paper, we report a microbial-metabolite-based oral nano-emulsion that regulates T cell metabolism and triggers robust T cell immune responses in multiple mouse tumour models and preserves the stemness of CD8⁺ T cells (Fig. 1a). By utilizing an in vitro screening of these altered non-SCFA metabolites on inulin gel administration, we identified 3,4-dihydroxybenzoic acid (DHB; also known as protocatechuic acid) as a key microbial metabolite that promotes the generation of memory T cells via the suppression of glycolysis. To improve the oral absorption and PK properties of the metabolite, we developed DHB into a synthetic prodrug in an oral nano-emulsion formulation, which substantially increased its half-life and bioavailability. The oral DHB prodrug formulation sensitized mice to anti-PD-1 blockade by boosting antigen-specific stem-like T cells in vivo, and demonstrated how to improve the drug-like properties of microbial metabolites. As small-molecule prodrugs have been widely used in modern pharmaceutics, our DHB prodrug formulation shows promising translational potential for systemic cancer immunotherapy.

Fig. 1. DHB promotes memory CD8⁺ T cell differentiation.

a, Schematic of DHB in modulating intracellular GSH content, reprogramming T cell metabolism and promoting memory T cell generation. b, Schematic of the in vitro co-culture of OT-I CD8⁺ T cells with various metabolites. Splenocytes (5 × 10⁴ cells per well) from OT-I mice were cultured with SIINFEKL peptide (4 nM) and metabolites for 2 or 3 days, and then, the cells were washed twice and further cultured with metabolites in the presence of IL-2 (10 ng ml⁻¹). c, Frequency of CD44⁺CD62L⁺ Tcm CD8⁺ T cells after 5 days. d, Frequency of Tcf1⁺CD8⁺ T cells after co-culture for 3 days. The dashed line in c and d represents the mean value of the PBS control group. e, After co-culturing for 2 days in the presence of IL-2, cells were washed to remove IL-2. Viability of CD8⁺ T cells (Annexin V⁻PI⁻) after 36 h of incubation in the absence of IL-2 is shown. The viability fold change was calculated using the level of Annexin V⁻PI⁻CD8⁺ T cells in the PBS group as the baseline. f, After co-culture for 5 days, cells were reactivated with the SIINFEKL peptide. The frequencies of IFN-γ⁺CD8⁺ T cells before and after antigen stimulation are shown. The fold change of IFN-γ⁺CD8⁺ T cells was calculated as the ratio of their frequency after antigen restimulation normalized to that before restimulation. g, Splenocytes were treated as that in d, and the frequency of Sca-1⁺CD8⁺ T cells on day 6 is shown. h, DHB was incubated with OT-I CD8⁺ T cells, as shown in b, and nuclear (unphosphorylated) Foxo1⁺CD8⁺ cells and phospho(Ser256)-Foxo1⁺ were measured by flow cytometry on day 8 and at the 6-h time point, respectively. i,j, OT-I splenocytes were incubated with SIINFEKL and DHB or its isomers. The frequency of CD44⁺CD62L⁺CD8⁺ Tcm cells on day 5 (i) and the amount of IL-2 secreted in the supernatant on day 3 (j) are shown. Data represent the mean ± s.e.m. (n = 10 (c), n = 5 (d), n = 3 (g), n = 8 (h); n = 3 for PBS, n = 4 for pantothenate and sebacate, n = 5 for other groups (e); n = 3 for ferulate, n = 5 for other groups (f); n = 6 for PBS, n = 4 for other groups (i and j) biologically independent samples). Data were analysed using a one-way ANOVA with Bonferroni’s multiple comparisons test (c–f, i and j) or unpaired, two-sided Student’s t-test (g and h). Illustrations created in BioRender: a, Moon, J. https://BioRender.com/qj9nivn (2026); b, Na, Y. https://BioRender.com/s0yv8bw (2026).

Source data

DHB induces stem-like CD8⁺ T cell differentiation

We have previously reported that oral inulin gel plus α-PD-1 IgG therapy expanded CD8⁺ T cells with stemness in part by modulating the gut microbiome and SCFA levels⁶. As the metabolomics data⁶ revealed that inulin gel plus α-PD-1 IgG therapy also altered various non-SCFA microbial metabolites, we sought to examine whether these non-SCFA microbial metabolites could regulate the phenotype of CD8⁺ T cells towards stem-like CD8⁺ T cells, known to promote long immunological memory; possess features of self-renewal, persistence and multipotency; and improve ICB efficacy¹⁰. In particular, we focused on glycerol, N-acetylneuraminate, γ-aminobutyric acid (GABA), pantothenate, disodium sebacate, hippurate, 5-aminoimidazole-4-carboxamide ribonucleotide, ferulate and DHB, based on our prior metabolomics data showing their markedly altered levels in the faeces after treatment with inulin gel plus α-PD-1⁶. Some other substantially altered metabolites were excluded because of their potential toxicity (for example, phenylacetaldehyde), the reported immune-activating effects (for example, ursodeoxycholate¹¹, linoleate¹² and arginine¹³). Splenocytes from naïve OT-I mice were activated with SIINFEKL antigen peptide, followed by incubation with these microbial metabolites and interleukin-2 (IL-2) (Fig. 1b). Among these microbial metabolites, DHB significantly increased the frequencies of CD44⁺CD62L⁺CD8⁺ central memory T (Tcm) cells (Fig. 1c and Supplementary Fig. 1a) and T cell factor 1⁺ (Tcf1) CD8⁺ T cells, in which CD62L and Tcf1 were encoded by the stem-like markers Sell and Tcf7, respectively, and reported to be positively associated with ICB outcome in cancer patients¹⁴ (Fig. 1d and Supplementary Fig. 1b). Moreover, DHB maintained T cell viability even when IL-2 was withdrawn for 36 h, compared with all the other groups (Fig. 1e and Supplementary Fig. 1c). In line with this stem-like phenotype, DHB treatment decreased the frequency of CD44⁺CD62L⁻CD8⁺ T effector memory cells, GzmB⁺CD8⁺ T cells and IFN-γ⁺CD8⁺ T cells, indicating less terminal effector differentiation (Supplementary Fig. 2 and Extended Data Fig. 1a,b); however, DHB-treated T cells exhibited a robust IFN-γ recall response on antigen restimulation (Fig. 1f and Supplementary Fig. 2). In addition, DHB-treated OT-I CD8⁺ T cells showed an increased expression of Sca-1, a hematopoietic stem cell antigen marker within memory-precursor-like T cells^15,16 (Fig. 1g and Extended Data Fig. 1c). DHB treatment also upregulated the expression of Forkhead O transcription factor 1 (Foxo1) and decreased Foxo1 phosphorylation, indicating that DHB engages Foxo1 to sustain stem-like T cell function and maintenance^8,17,18 (Fig. 1h and Extended Data Fig. 1d). Besides, DHB did not affect the proliferation of CD8⁺ T cells (Supplementary Fig. 3). Interestingly, several DHB isomers, including 2,4-DHB and 2,3-DHB, also increased the generation of CD8⁺ Tcm cells (Fig. 1i). However, among these DHB isomers, only DHB significantly enhanced IL-2 secretion when OT-I splenocytes were incubated with the SIINFEKL peptide and DHB isomers (Fig. 1j), suggesting the unique molecular activity of DHB. Additionally, OT-I CD8⁺ T cells incubated with DHB and anti-CD3/anti-CD28 antibodies expanded CD44⁺CD62L⁺ Tcm cells (Extended Data Fig. 1e), indicating that DHB-mediated Tcm expansion was antigen independent. Moreover, DHB induced the differentiation of CD44⁺CD62L⁺CD8⁺ Tcm cells in a dose-dependent manner (Extended Data Fig. 1f).

Extended Data Fig. 1. DHB treatment modulates the phenotype of CD8 + T cells.

(a) Splenocytes (5×10⁴ cells/well) from OT-I mice were cultured with SIINFEKL peptide (4 nM) and DHB for 3 days, and then cells were washed twice and further cultured with DHB (100 µM) in the presence of IL⁻2 (10 ng/mL). Shown are the frequencies of CD44⁻CD62L⁺, CD44⁺CD62L⁺, and CD44⁺CD62L⁻ CD8⁺ T cells after 5 days. (b) OT-I CD8⁺ T cells were activated with anti-CD3, anti-CD28 antibodies and IL-2 for 3 days. Cells were washed for three times and further incubated with IL-2 for 4 days. 100 µM DHB was added during this in vitro assay. Shown is the level of Gzmb in CD8⁺ T cells. The gating strategy of CD8⁺ T cells and representative flow cytometry plots of Sca-1⁺CD8⁺ T cells (c), Foxo1⁺CD8⁺ T cells and pFoxo1⁺CD8⁺ T cells (d). (e) OT-I CD8⁺ T cells were activated with anti-CD3, anti-CD28 antibodies and IL-2 for 3 days. Cells were washed for three times and further incubated with IL-2 for 5 days. 100 µM DHB was added during this in vitro assay. Shown is the frequency of CD44⁺CD62L⁺CD8⁺ Tcm cells on day 8. (f) OT-I splenocytes were treated as described in a with various concentrations of DHB, and shown are the frequencies of CD44⁻CD62L⁺, CD44⁺CD62L⁺, and CD44⁺CD62L⁻ CD8⁺ T cells. Data represent the mean ± s.e.m. (n = 3 (a,f), 4 (b), 5 (e) biologically independent samples). Data were analysed using two-way ANOVA with Bonferroni’s multiple comparisons test (a), one-way ANOVA with Bonferroni’s multiple comparisons test (f), unpaired, two-sided Student’s t-test (b,e).

Source data

DHB reprograms T cell metabolism and promotes stemness

T cell activation and differentiation are largely controlled by their immunometabolic adaptation¹⁹. Hence, we performed targeted metabolomics analysis to delineate the potential mechanism of DHB-induced stem-like T cell differentiation. The DHB treatment of OT-I CD8⁺ T cells decreased many intermediate metabolites within the canonical metabolic pathways, including glycolysis, glutaminolysis and citric acid cycle, compared with phosphate-buffered saline (PBS) control (Fig. 2a and Extended Data Fig. 2a). This metabolically quiescent status corresponds to the distinct metabolic profile in stem-like memory CD8⁺ T cells²⁰. Specifically, DHB substantially reduced the levels of glucose-6-P, fructose 6-P, P-enolpyruvate and pyruvate in the glycolytic pathway (Extended Data Fig. 2b). Moreover, Seahorse assay for bioenergetics performed on CD8⁺ T cells isolated from lymph nodes of CT26-tumour-bearing mice treated with DHB showed that DHB treatment decreased the extracellular acidification rate (ECAR) of CD8⁺ T cells (including basal ECAR, glycolytic capacity and glycolytic reserve) in vivo (Fig. 2b and Extended Data Fig. 2c), indicating the suppression of glycolysis. In addition, in vitro DHB treatment of OT-I CD8⁺ T cells decreased ECAR as well as the oxygen consumption rate (reflecting oxidative phosphorylation²¹; Extended Data Fig. 2d). Taken together, these results suggested that DHB suppresses glycolysis in CD8⁺ T cells and skews CD8⁺ T cells into a metabolically quiescent status.

Fig. 2. DHB programs T cell metabolism for stem-like T cell differentiation.

a, OT-I CD8⁺ T cells were treated with DHB as shown in Fig. 1b, followed by targeted metabolomics analysis on day 8. The z-score heat map of significantly altered metabolites in T cells is shown. b, CT26-tumour-bearing BALB/c mice were orally administered with PBS or DHB (2.4 mg per dose) for five times (every other day), and CD8⁺ T cells isolated from inguinal lymph nodes on day 11 were analysed for ECAR. c, CD8⁺ T cells isolated from C57BL/6 mice were activated with anti-CD3/CD28 antibodies in the presence of IL-2 (10 ng ml⁻¹) for 2 days. Cells were washed twice and incubated with IL-2 for another 5 days. DHB and various inhibitors (10 µM of BPTES, 5 µM of UK5099, 40 μM of (+)-etomoxir, 0.5 mM of 2-DG or 1 µM of oligomycin) were added. d, DHB and GSH were reacted in 10 mM of PBS with or without NADPH for 60 min, and DHB-GSH adduct and MS/MS fragmentation pattern were identified by QqQ-MS/MS analysis. e, OT-I CD8⁺ T cells were treated with DHB as shown in Fig. 1b, and intracellular GSH levels were measured on day 8. f, OT-I CD8⁺ T cells were treated with DHB as shown in Fig. 1b, and after 6 h, western blot analysis for phosphorylated p-mTOR (Ser2448), total mTOR, p-Akt (Ser473) and total Myc was performed. g, Fold change of genes related to progenitor-like signature between DHB- and PBS-treated T cells in bulk RNA-seq analysis. h–j, OT-I CD8⁺ T cells were treated as shown in Fig. 1b, and cells were collected on day 8 for scRNA-seq analysis. h, Gene expression mapped in two dimensions via uniform manifold approximation and projection (UMAP) of all samples (left) and heat map showing the marker representation in each cluster (right). i, Relative size change of various cell clusters for DHB versus PBS. j, scATAC-seq plots for Tcf7, Sell and Ccr7 from all clusters. k,l, Leading edge plots of GSEA for glycolysis (k) and hypoxia (l) (DHB versus PBS). FDR, false discovery rate. GSEA was performed using GSEA preranked analysis with MSigDB mouse Hallmark gene sets (v. 2023.1). Genes were ranked using a signed–log-transformed nominal P value, where genes upregulated in that group were assigned positive scores and genes downregulated in that group were assigned negative scores. Nominal P values used for ranking were obtained from differential expression analysis. The ranked gene lists were analysed using a weighted scoring scheme with 10,000 permutations. Enrichment significance was evaluated using permutation-based P values and FDR q values. m, ATAC-seq analysis of cluster 3 plus cluster 6 shows the fold changes of chromatin accessibility between PBS- and DHB-treated T cells. n, ATAC peak tracks at the Nr4a1, Pdcd1 and Icos loci. The orange-shaded regions in j and n indicate the corresponding peaks to each gene locus in the ATAC-seq track. Data represent the mean ± s.e.m. (n = 3 (a, b and g), n = 5 (e); n = 4 for PBS without inhibitors, and n = 3 for groups (c); biologically independent samples). Data were analysed using one-way ANOVA with Bonferroni’s multiple comparisons test (c), or unpaired, two-sided Student’s t-test (e).

Source data

Extended Data Fig. 2. T cell metabolism analyses.

DHB were incubated with OT-I CD8⁺ T cells as shown in Fig. 1b, cells were collected on day 8 for targeted metabolomics analysis. Shown are (a) the analysis of various pathways and (b) further analysis of glycolysis pathway. (c) The basal ECAR, glycolytic capacity, and glycolytic reserve analyses for Fig. 2b. (d) DHB was incubated with OT-I CD8⁺ T cells as shown in Fig. 1b, and on day 8, OT-I CD8⁺ T cells were seeded on the Cell-Tak-coated plate and the ECAR and OCR were measured by the Seahorse assay. FCCP: Carbonyl cyanide-p-trifluoromethoxyphenylhydrazone; R/A: rotenone + antimycin A. (e) The roles of various inhibitors in metabolic pathways. Data represent the mean ± s.e.m. (n = 3 (c), 6 (d) biologically independent samples). Data were analysed using unpaired, two-sided Student’s t-test (c). Illustration in e created in BioRender; Moon, J. https://BioRender.com/d5de5y9 (2026).

Source data

Next, we examined the impact of metabolic changes on DHB-induced stem-like T cell differentiation using a series of metabolic pathway inhibitors (Fig. 2c). We tested the effects of bis-2-(5-phenylacetamido-1,3,4-thiadiazol-2-yl)ethylsulfide (BPTES), which depletes glutamate for the TCA cycle²²; UK5099, a mitochondrial pyruvate carrier (MPC) inhibitor²³; etomoxir44, a fatty acid oxidation inhibitor²⁴; 2-deoxy-ᴅ-glucose (2-DG)²⁴, a glycolysis inhibitor; and oligomycin, known to promote glycolysis by inhibiting ATP synthase and oxidative phosphorylation²⁵ (Extended Data Fig. 2e). Treatment with BPTES, UK5099 or etomoxir did not alter DHB-induced Tcm differentiation (Fig. 2c). By contrast, 2-DG increased the frequency of Tcm in the PBS group to a comparable level as the DHB treatment (Fig. 2c), highlighting the critical role of glycolysis inhibition in CD8⁺ Tcm differentiation. In particular, oligomycin markedly decreased the level of CD8⁺ Tcm in the PBS control, but not in the DHB group (Fig. 2c). Consistently, GSK2837808A, a lactate dehydrogenase inhibitor²⁶ that blocks the conversion of pyruvate to lactate, promoted Tcm generation to a comparable level to that observed in DHB-treated T cells (Extended Data Fig. 3a).

Extended Data Fig. 3. CD8⁺ Tcm induction. OT-I CD8⁺ T cells were activated with anti-CD3, anti-CD28 antibodies and IL-2 for 3 days.

Cells were washed twice and further incubated with IL-2 for 4 days. 100 µM DHB, 25 µM BSO, 15 µM GSK2837808A, or 10 nM rapamycin was added during the in vitro assay. Shown are the frequencies of CD8⁺ Tcm cells in the presence of GSK2837808A or rapamycin (a), or BSO (b). Data represent the mean ± s.e.m. (n = 4 biologically independent samples). Data were analysed using one-way ANOVA with Bonferroni’s multiple comparisons test.

Source data

To delineate the underlying mechanism of the DHB-mediated suppression of glycolysis and Tcm differentiation, we focused on reduced glutathione (GSH). This was based on the dramatic decrease (~97%) in the intracellular GSH level in DHB-treated T cells, compared with PBS-treated T cells (Fig. 2a). To examine how DHB treatment decreased the GSH levels, we performed biomolecular mass spectrometry with triple-quadrupole (QqQ) tandem mass spectrometry (MS/MS) on GSH incubated with free DHB or DHB-quinone oxidized by nicotinamide adenine dinucleotide phosphate (NADPH). The analysis revealed rapid DHB-GSH adduct formation via Michael addition reaction between the thiol group of GSH and DHB (Fig. 2d and Extended Data Fig. 4a,b). Interestingly, among its various isomers, only DHB formed a stable adduct showing the highest signal intensity, as the ortho-catechol position favours the formation of stable quinone intermediates (Extended Data Fig. 4c,d). In vitro GSH trapping assay further confirmed the reduced level of intracellular GSH in DHB-treated OT-I CD8⁺ T cells (Fig. 2e). GSH is a major determinant of intracellular redox potential that regulates T cell immunity via the Akt/mammalian target of rapamycin (mTOR) pathway²⁷. Partial loss of intracellular GSH by buthionine sulfoximine (BSO; known as a glutamate cysteine ligase inhibitor²⁸) increased the Tcm to a similar level observed in DHB-treated T cells (Extended Data Fig. 3b). Immunoblot analysis revealed that DHB treatment decreased the phosphorylation of Akt (Ser473) and mTOR (Ser2448) in OT-I CD8⁺ T cells (Fig. 2f and Supplementary Fig. 4), indicating the DHB-mediated downregulation of the Akt-mTORC1 pathway. In line with these findings, treatment with the mTOR inhibitor rapamycin induced CD8⁺ Tcm cells to a level comparable to that observed in DHB-treated cells (Extended Data Fig. 3a)²⁷. Moreover, DHB substantially decreased the expression of Myc, a crucial transcription factor that activates glycolysis and glutaminolysis²⁸ (Fig. 2f and Supplementary Fig. 4). These results showed that DHB decreases intracellular GSH by forming an adduct with GSH and reprograms T cell metabolism to switch off glycolysis via the Akt-mTORC1-Myc pathway.

Extended Data Fig. 4. DHB-GSH adduct.

(a) Schematic for the formation of DHB-GSH adduct from DHB-quinone via Michael addition reaction. (b) QqQ-MS/MS analysis of DHB-GSH adduct formation from incubating DHB and GSH, with or without NADPH. The circled peaks correspond to the adduct. (c) LC-MS/MS chromatograms of GSH incubated with different DHB isomers. Positive ion MS/MS scanning for precursors of m/z 460. (d) High resolution product ion MS/MS spectra of protonated molecules of GSH conjugates.

Source data

We performed RNA-sequencing (RNA-seq) analyses to examine DHB-mediated transcriptional control of activated OT-I CD8⁺ T cells. Bulk RNA-seq analysis showed that DHB markedly increased the expression levels of genes associated with progenitor-like signature, including Ly6a, Ccr7, Sell and Tcf7 (Fig. 2g). In addition, single-cell RNA-sequencing (scRNA-seq) analysis of DHB-treated OT-I CD8⁺ T cells revealed 13 distinct clusters (Fig. 2h). Among those clusters, DHB-induced stem-like T cells were identified in cluster 3 with a high expression of Tbx21, Tcf7, Tox, Mki67, Ikzf2 and bcl6, and cluster 6 with a high expression of Tcf7, Sell, Tox, Ly6c2, bcl6 and Pecam1 (Fig. 2h,i and Extended Data Fig. 5a). In cluster 6, we also noted a small but pronounced increase in the expression levels of Tox, which suggests that DHB-induced stem-like CD8⁺ T cells shared molecular features with progenitor exhausted T cells.

Extended Data Fig. 5. scRNA-seq and ATAC-seq analyses.

(a) Violin plots representing the expression of Tcf7 and Sell in PBS control and DHB induced stem-like CD8⁺ T cells in clusters 3 and 6 from the scRNA/ATAC-seq analysis shown in Fig. 2. For the corresponding box plots, the center line represents the median, the bounds of the box represent the upper and lower quartiles (interquartile range, IQR), and the whiskers extend to 1.5 times the IQR (or to the minimum and maximum data points if within that range). The P-values were obtained from two-sided Wilcoxon rank-sum test. (b) UMAP of scATAC-seq data integrated with scRNA-seq UMAP. (c) Distribution of PBS- and DHB-treated cells in scATAC-seq UMAP of all T cells. (d) Downregulated pathways of DHB versus PBS in cluster 3 plus cluster 6 as well as the normalized enrichment score (NES). (e) Differential gene expression analysis, between DHB- and PBS-treated cells within cluster 3 plus cluster 6 as shown in Fig. 2h, was performed using the Seurat::FindMarkers(). Statistical significance was assessed using a two-sided Wicoxon rank-sum test, and P-values were adjusted for multiple comparisons using Bonferroni correction. Representative upregulated and downregulated genes are labeled in the panel. Genes with P < 0.05 are highlighted in blue color, otherwise are colored in gray. The dashed line indicates the threshold of log₁₀(0.05). (f) scATAC-seq plots for Tcf7, Sell, and Ccr7 from clusters 3 and 6. (g) scATAC-seq tracks showing chromatin accessibility at the Tcf7, Ccr7 enhancer 1, and Sell enhancer 1, 2, 5 regions in PBS control compared with DHB-treated cells. For each group, we pooled 3 wells of cells for n = 1 sequencing sample.

Source data

Using the same cells analysed by scRNA-seq described above, we additionally performed paired single-cell assay for transposase-accessible chromatin with high-throughput sequencing (scATAC-seq) to assess the impact of DHB on chromatin accessibility at single-cell resolution (Extended Data Fig. 5b,c). DHB-treated CD8⁺ T cells showed increased accessibility at the target gene loci of Tcf7, Sell and Ccr7 compared with the PBS control, consistent with the transcriptomic data (Fig. 2j). In particular, Gene Set Enrichment Analysis (GSEA) performed on cluster 3 plus cluster 6 showed a substantial downregulation of glycolysis (normalized enrichment score (NES), −1.49) as well as hypoxia (NES, −1.60) pathways (Fig. 2k–m and Extended Data Fig. 5d,e). As hypoxia contributes to increased glycolysis and T cell exhaustion and hinders the effectiveness of immunotherapy²⁹, these results suggest the potential role of DHB in promoting T cell stemness and preventing T cell exhaustion. Meanwhile, DHB-induced stem-like T cells in cluster 3 plus cluster 6 exhibited decreased chromatin accessibility for Pdcd1 and Icos, known to drive T cell exhaustion^30,31 as well as increased chromatin accessibility for Sell and Nr4a1, a recently reported transcription factor that promotes the stemness of T cells and progenitor exhausted T cells-related genes³² (Fig. 2n and Extended Data Fig. 5f). Furthermore, DHB-induced stem-like T cells consistently showed an increase in chromatin openness in enhancer regions of Tcf7, Ccr7 and Sell (Extended Data Fig. 5g). These data show that DHB induces a unique epigenetic landscape that supports progenitor-like gene expression programs.

Oral DHB Prodrug 201 augments the efficacy of α-PD-1 therapy

On the basis of the strong potential of DHB to induce stem-like CD8⁺ T cells in vitro (Figs. 1 and 2) and accumulating evidence demonstrating stem-like CD8⁺ T cells as the crucial determinant for successful ICB immunotherapy³³, we set out to examine whether DHB can improve the antitumour efficacy of systemic α-PD-1 IgG and how to further improve its therapeutic potential. In an ICB-resistant spontaneous CDX2-Cre NLS-APC^fl/fl colorectal cancer model³⁴, the oral administration of DHB combined with α-PD-1 IgG therapy markedly reduced the burden of colon adenocarcinoma (Fig. 3a). On the other hand, in CT26 subcutaneous (s.c.) tumour model, oral DHB administration did not improve the antitumour efficacy of α-PD-1 IgG (Fig. 3b), potentially due to the low oral bioavailability of DHB. By contrast, increasing the local concentration of DHB by direct intratumoural administration into CT26 s.c. tumour resulted in similar antitumour efficacy as systemic α-PD-1 monotherapy (Fig. 3b). The combination of intratumoural DHB administration and systemic α-PD-1 IgG significantly delayed the growth of CT26 s.c. tumour (Fig. 3b). Taken together, these results suggested that DHB, when administered appropriately, could achieve strong synergy with α-PD-1 ICB.

Fig. 3. Oral DHB prodrug synergizes with α-PD-1 therapy.

a, CDX2-Cre NLS-APC^fl/fl mice were treated as shown, and the average tumour number, size and the total tumour size in the colon were measured. b, CT26-tumour-bearing BALB/c mice were treated as shown with DHB via either intratumoural or oral administration, followed by the monitoring of tumour growth. c,d, Various oral DHB prodrug formulations (molar equivalent to 0.11 mg DHB per dose) were tested with α-PD-1 therapy in CT26-tumour-bearing BALB/c mice. Tumour volume (c) and frequencies of AH1-specific CD8⁺ T cells (d) among PBMCs on days 17 and 22. e, Hydrodynamic size and morphology of Prodrug 201 formulation. PDI, polydispersity index. f, Plasma concentration–time plot of Prodrug 201 and free DHB after a single oral dose in rats (9.7 mg kg⁻¹, DHB-equivalent dose) and the summary of PK parameters. g,h, Therapeutic regimen of oral Prodrug 201 or oral DHB (0.11 mg of DHB per dose, or its molar-equivalent dose for Prodrug 201) in combination with α-PD-1 therapy. Tumour growth curves in NOOC1 squamous cell carcinoma (g) and B16F10 melanoma models (h) are shown. i, MMTV-PyMT mice were treated with either PBS, α-PD-1 antibody or Prodrug 201 plus α-PD-1 antibody, and the average tumour growth curves and the representative picture of mammary tumours on days 88 are shown. Data represents the mean ± s.e.m. (n = 3 (e and f), n = 5 (b, c, g and i); n = 4 for Prodrug 201 + α-PD-1, n = 5 for other groups in the left panel, or n = 5 in the right panel (d); n = 7 for DHB + α-PD-1, 8 for α-PD-1 (a); n = 5 for PBS, DHB + α-PD-1, and n = 6 for other groups (h) biologically independent samples). Data were analysed using a two-way ANOVA (b, c and g–i) or one-way ANOVA with Bonferroni’s multiple comparisons test (d), or two-sided Student’s t-test (a). Mouse icons created in BioRender: a, g and h, Na, Y. https://BioRender.com/j0elmnw (2026); b, c and i, Na, Y. https://BioRender.com/r7xpzgq.

Source data

Encouraged by these results, we sought to optimize the oral delivery of DHB for systemic cancer immunotherapy. DHB is a hydrophilic small molecule that can be cleared rapidly in vivo. Prodrug is a well-established strategy to modify the parent drug for increased hydrophobicity, protection of functional groups and improved PK profiles. We synthesized seven variants of DHB prodrugs and found that Prodrugs 863, 142 and 201 were well encapsulated with high stability in an oral nano-emulsion formulated with oleic acid/Tween 80 (ref. ³⁵) (Extended Data Fig. 6). Oral administration of these prodrug nano-emulsion formulations improved the therapeutic efficacy of α-PD-1 IgG in CT26 s.c. tumour-bearing mice (Fig. 3c). In particular, oral prodrug 201 formulation (Prodrug 201) combined with α-PD-1 therapy exerted the strongest antitumour efficacy and significantly increased the frequency of CD8⁺ T cells specific to AH1, an immunodominant epitope from CT26 gp70 (H-2L^d-restricted SPSYVYHQF), among the peripheral blood mononuclear cells (PBMCs; Fig. 3c,d). On the basis of these results, we focused on Prodrug 201 in the subsequent studies.

Extended Data Fig. 6. Prodrug structures and formulation preparation.

(a) A library of synthesized DHB prodrugs. (b) The scheme of the preparation of orally administered DHB prodrug nano-emulsion.

Prodrug 201 nano-emulsion exhibited a homogeneous spherical morphology with a hydrodynamic size of ~160 nm with a polydispersity index of 0.19 (Fig. 3e). We examined the PK profiles of Prodrug 201 administered orally. First, liquid chromatography MS/MS analysis of Prodrug 201 exposed to rat plasma showed that ethyl-DHB (M4) and DHB were the major metabolized products of Prodrug 201 (Extended Data Fig. 7a,b), supporting our prodrug-based strategy. PK studies in rats demonstrated that the half-life of DHB in the plasma after oral gavage is only 0.4 h with a plasma concentration (C_max) of 34.7 ng ml⁻¹. In sharp contrast, the oral administration of Prodrug 201 markedly extended the plasma half-life to 4.1 h, increased the C_max to 182 ng ml⁻¹, improved the area under the curve by 14.3-fold, and improved the oral bioavailability (%F) by 9.72-fold, compared with free DHB (Fig. 3f and Extended Data Fig. 7c). This improved oral absorption and PK profile of Prodrug 201 led to greater antitumour efficacy in various murine tumour models. Compared with free DHB plus α-PD-1 therapy, oral Prodrug 201 plus α-PD-1 combo-therapy exerted significantly enhanced antitumour efficacy in mice with established α-PD-1-resistant NOOC1 squamous cell carcinoma (Fig. 3g and Extended Data Fig. 7d) as well as in mice bearing B16F10 melanoma (Fig. 3h), demonstrating the superiority of Prodrug 201 over the parent drug, DHB. Besides, in both spontaneous CDX2-Cre NLS-APC^fl/fl colorectal cancer model and mammary-specific polyomavirus middle T antigen overexpression mouse model (MMTV-PyMT), oral Prodrug 201 boosted the therapeutic efficacy of α-PD-1 antibody, resulting in markedly reduced tumour burden in the colon or abdomen/axilla (Fig. 3i and Extended Data Fig. 8). In particular, the MMTV-PyMT model closely recapitulates human breast cancer progression³⁶, including metastasis, thereby underscoring the translational relevance of the Prodrug 201 combination therapy.

Extended Data Fig. 7. The structure and stability of metabolized products of Prodrug 201.

(a,b) Prodrug 201 incubated in rat plasma was analyzed for their metabolites by LC-MS/MS. (c) Oral bioavailability analysis of Prodrug 201 and free DHB after single dose of Prodrug 201 or DHB administered via IV or oral routes in rats (19.4 mg/kg, DHB equivalent dose). (d) The overall Kaplan–Meier survival curves in NOOC1 model as shown in Fig. 3g. Data represent the mean ± s.e.m. (n = 3 (b), 5 (d) biologically independent samples). Data were analysed using a log-rank Mantel–Cox test (d).

Source data

Extended Data Fig. 8. Prodrug 201 formulation boosted the antitumor immunity.

16-week-old CDX2-Cre NLS-APC^fl/fl mice were treated with either PBS, α-PD-1 antibody, or Prodrug 201 plus α-PD-1 antibody, and shown are the total tumor number, total tumor size in colon. Data represent the mean ± s.e.m. (n = 6 for PBS, 8 for α-PD-1, and 7 for Prodrug 201 + α-PD-1 group biologically independent samples). Data were analysed using a one-way ANOVA with Bonferroni’s multiple comparisons test. Mouse icon created in BioRender; Na, Y. https://BioRender.com/j0elmnw (2026).

Source data

Prodrug 201 promotes antigen-specific stem-like CD8⁺ T cells

On the basis of the strong antitumour efficacy of oral Prodrug 201 combined with α-PD-1, we further investigated its impact on antitumour immune responses. In CT26-tumour-bearing mice, oral Prodrug 201 plus α-PD-1 therapy effectively suppressed tumour growth and prolonged the animal survival (Fig. 4a–c). Analyses of splenocytes from CT26-tumour-bearing mice on day 25 and tumour-free survivors on day 65 showed that the stemness features persisted in antigen-specific, AH1-tetramer⁺CD8⁺ T cells with higher frequencies of CD44⁺CD62L⁺AH1-tetramer⁺CD8⁺ T cells and Sca-1⁺PD-1⁻AH1-tetramer⁺CD8⁺ T cells in tumour-free survivors on day 65 (Extended Data Fig. 9). Moreover, when tumour-free survivors in the oral Prodrug 201 plus α-PD-1 group were rechallenged with CT26 tumour cells on the contralateral flank on day 90, they were completely resistant to tumour growth (Fig. 4d). This indicated that oral Prodrug 201 plus α-PD-1 therapy induced a long-term protective antitumour response. In particular, this combination therapy induced negligible changes in the body weight or complete blood count in CT26-tumour-bearing mice (Supplementary Figs. 5 and 6), indicating good biosafety profile of the oral Prodrug 201 plus α-PD-1 therapy.

Fig. 4. Oral Prodrug 201 expands antigen-specific memory CD8⁺ T cells.

a, Therapeutic regimen in CT26-tumour-bearing BALB/c mice (Prodrug 201 dose: molar equivalent to 0.11 mg DHB per dose). b,c, Average and individual tumour growth curves (b) and overall Kaplan–Meier survival curves (c). d, Survivors in the Prodrug 201 plus α-PD-1 group were rechallenged with 1.5 × 10⁵ CT26 cells inoculated s.c. on the contralateral flank on day 90, followed by tumour monitoring for another 60 days. e, CT26-tumour-bearing BALB/c mice were treated as shown and administered i.p. with various depleting antibodies against CD8⁺ T (α-CD8), Treg (α-CD25) cells or NK (α-asialo GM1) cells, or isotype IgG2b control on days 9, 12 and 15. The average tumour growth curves are shown. f, Frequencies of AH1-specific CD8⁺ T cells among PBMCs on days 14 and 20, and the representative flow cytometry plots. g,h, Frequencies of CD44⁺CD62L⁺CD8⁺ Tcm cells (g) and CD44⁺CD62L⁻CD8⁺ Tem cells (h) on day 20 in PBMCs. i, IFN-γ enzyme-linked immune absorbent spot (ELISPOT) analysis on PBMCs isolated on day 21 and restimulated with AH1 epitope. j,k, AH1-tetramer⁺CD44⁺CD62L⁺CD8⁺ Tcm cells (j) and AH1-tetramer⁺CD44⁺CD62L⁻CD8⁺ Tem cells (k) on days 20 in PBMCs. Data represent the mean ± s.e.m. (n = 5 (b, c, e–h, j and k), n = 6 (i), n = 3–9 (d) biologically independent samples). Data were analysed using a two-way ANOVA (b and e) or one-way ANOVA with Bonferroni’s multiple comparisons test (f–k), or a log-rank Mantel–Cox test (c and d). Mouse icons in a and e created in BioRender; Na, Y. https://BioRender.com/r7xpzgq (2026).

Source data

Extended Data Fig. 9. Analysis of antigen-specific CD8⁺ T cells in tumor-free survivors.

CT26 tumor-bearing mice were treated either PBS or Prodrug 201 plus α-PD-1 antibody as shown in a. Spleens were collected on day 25 (PBS control and Prodrug 201 plus α-PD-1 antibody treated tumor-bearing mice) and on day 65 (tumor-free surviving mice from the Prodrug 201 plus α-PD-1 antibody treatment group), and AH1-specific CD8⁺ T cells were analyzed by flow cytometry. Shown are (b) the frequencies of CD3⁺CD8⁺ T cells among live CD45⁺ cells, and CD44⁺CD62L⁺, Sca-1⁺PD-1⁻ (c) among AH1-tetramer⁺CD8⁺ T cells. Data represent the mean ± s.e.m. (n = 8 for tumor-free survivors (day 65), and 6 for other groups (b,c) biologically independent samples). Data were analysed using a one-way ANOVA with Bonferroni’s multiple comparisons test. Mouse icon in a created in BioRender; Na, Y. https://BioRender.com/r7xpzgq (2026).

Source data

Next, we examined which immune cell type was responsible for the antitumour efficacy of the combination therapy. Depletion of natural killer (NK) cells or CD4⁺ regulatory T cells (Tregs) with α-Asialo GM1 antibody or α-CD25 antibody, respectively, had only minor effect on the antitumour effect of the Prodrug 201 plus α-PD-1 therapy (Fig. 4e). By contrast, the administration of α-CD8 antibody abrogated the antitumour efficacy of the combination therapy (Fig. 4e), suggesting CD8⁺ T cells as the key antitumour effector cells. Indeed, oral Prodrug 201 plus α-PD-1 combo-therapy significantly increased the frequencies of AH1 antigen-specific CD8⁺ T cells, CD44⁺CD62L⁺CD8⁺ Tcm cells and CD44⁺CD62L⁻CD8⁺ Tem cells in PBMCs (Fig. 4f–h), compared with α-PD-1 monotherapy. The combination therapy also promoted IFN-γ response from CD8⁺ T cells on in vitro stimulation with AH1 epitope (Fig. 4i) and increased AH1-specific CD44⁺CD62L⁺CD8⁺ Tcm cells, and CD44⁺CD62L⁻CD8⁺ Tem cells (Fig. 4j,k).

Moreover, in CT26-tumour-bearing mice, oral Prodrug 201 plus α-PD-1 combo-therapy markedly increased the intratumoural levels of IFN-γ, interferon-inducible chemokine (C-X-C motif) ligand 10 (CXCL10) and granzyme B, compared with PBS treatment (Fig. 5a,b and Extended Data Fig. 10a). The combination therapy also drove the substantial infiltration of CD45⁺ leucocytes, CD3⁺ T cells and CD8⁺ T cells into the tumour microenvironment (TME), whereas increasing the ratio of M1-like macrophages to M2-like macrophages and decreasing immune suppressive Tregs (Fig. 5c–f and Extended Data Fig. 10b–e). Prodrug 201 plus α-PD-1 also elevated the frequencies of CD3⁺ and CD8⁺ T cells in tumour-draining lymph nodes (tdLNs), compared with α-PD-1 monotherapy (Extended Data Fig. 10f,g).

Fig. 5. Oral Prodrug 201 promotes antigen-specific stem-like CD8⁺ T cell generation in tumour and tdLNs.

a–p, CT26-tumour-bearing BALB/c mice were treated with Prodrug 201 (molar equivalent to 0.11 mg DHB per dose) starting on day 9, and some mice were administered i.p. with α-PD-1 antibody on days 11, 15 and 19. The tumour tissues were analysed on day 23 for IFN-γ (a) and CXCL10 (b). The tumour and tdLN tissues were isolated on day 21, and shown are the frequencies of CD3⁺CD45⁺ T cells (c), CD8⁺CD3⁺ T cells (d), the ratio between M1- and M2-like macrophages (e) and Foxp3⁺CD3⁺CD8⁻ Tregs (f) in the tumour tissues. g–j, Analyses of CD8⁺ T cells in tumour tissues on day 23. Numbers or frequencies of CD8⁺ Tcm cells (g), AH1-specific CD8⁺ T cells (h), CD44⁺CD62L⁺AH1-tetramer⁺CD8⁺ T cells (i), Tcf1⁺AH1-tetramer⁺CD8⁺ T cells (j), Foxo1⁺ AH1-tetramer⁺CD8⁺ T cells (k), Sca-1⁺AH1-tetramer⁺CD8⁺ T cells (l) and Sca-1⁺PD-1⁻AH1-tetramer⁺CD8⁺ T cells (m). n–p, Analyses of CD8⁺ T cells in tdLNs on day 23. Frequencies of CD44⁻CD62L⁺, CD44⁺CD62L⁻ and CD44⁺CD62L⁺ among AH1-specific CD8⁺ T cells (n), Sca-1⁻PD-1⁺, Sca-1⁺PD-1⁺, Sca-1⁺PD-1⁻ and Sca-1⁻PD-1⁻ among AH1-specific CD8⁺ T cells (o) and CD44⁻CD62L⁺, CD44⁺CD62L⁻ and CD44⁺CD62L⁺ among Sca-1⁺PD-1⁻AH1-tetramer⁺CD8⁺ T cells (p). Data represent the mean ± s.e.m. (n = 6 (a, b, l and n–p); n = 4 for Prodrug 201 + α-PD-1, and n = 5 for other groups (c–f); n = 5 for α-PD-1, and n = 6 for other groups (g and h); n = 5 for Prodrug 201 + α-PD-1, and n = 6 for other groups (i–k and m) biologically independent samples). Data were analysed using a two-way ANOVA (n–p) or one-way ANOVA with Bonferroni’s multiple comparisons test (a–m).

Source data

Extended Data Fig. 10. TME analyses.

(a) The tumor tissues were analysed for the level of granzyme B. (b-j) CT26 tumor-bearing BALB/c mice were treated with PBS or Prodrug 201 starting on day 9, and some mice were administered i.p. with α-PD-1 antibody on days 11, 15 and 19. The tumor and tdLN tissues were isolated on day 21. Shown are the frequencies of (b) CD45⁺ cells, (c) macrophages, (d) M1-like macrophages, (e) M2-like macrophages in tumors, and (f) CD3⁺CD45⁺ T cells, (g) CD8⁺CD3⁺ T cells in the tdLNs, (h) the number of AH1-specific CD8⁺ T cells in tumor. (i) NOOC1 tumor-bearing mice were treated as shown in Fig. 3g. Shown are the frequencies of CD3⁺ T cells, CD62L⁺CD45⁺CD3⁺CD8⁺ T cells, and Tcf1⁺CD45⁺CD3⁺CD8⁺ T cells in NOOC1 tumor after various treatments. (j) B16F10 tumor-bearing mice were treated as shown in Fig. 3h. Shown are the frequencies of CD45⁺CD3⁺ T cells, and (k) CD45⁺CD3⁺CD8⁺ T cells, and CD44⁺CD62L⁺CD45⁺CD3⁺CD8⁺ T cells in B16F10 tumor after various treatments. (l) The representative flow cytometry plots with the percentage of Sca-1⁻PD-1⁺, Sca-1⁺PD-1⁺, Sca-1⁺PD-1⁻, and Sca-1⁻PD-1⁻ among AH1-specific CD8⁺ T cells, and CD44⁻CD62L⁺, CD44⁺CD62L⁻, and CD44⁺CD62L⁺ among Sca-1⁺PD-1⁻AH1-tetramer⁺CD8⁺ T cells. Data represent the mean ± s.e.m. (n = 6 (a), 5 (f,g,i-k); n = 4 for Prodrug 201 + α-PD-1, and 5 for other groups (b-e); n = 5 for α-PD-1, and 6 for other groups (h) biologically independent samples). Data were analysed using one-way ANOVA with Bonferroni’s multiple comparisons test.

Source data

Immunological profiling of the TME showed that Prodrug 201 plus α-PD-1 substantially increased the frequencies of CD44⁺CD62L⁺CD8⁺ T cells and AH1-specific CD8⁺ T cells (Fig. 5g,h and Extended Data Fig. 10h). Meanwhile, the combination therapy significantly expanded the intratumoural frequencies of AH1-specific CD8⁺CD44⁺CD62L⁺ T cells, AH1-specific Tcf1⁺CD8⁺ T cells, AH1-specific Foxo1⁺CD8⁺ T cells, AH1-specific Sca-1⁺CD8⁺ T cells and AH1-specific Sca-1⁺PD-1⁻CD8⁺ T cells, compared with PBS treatment (Fig. 5i–m). Similar results were also observed in NOOC1 and B16F10 tumour models, where tumour-infiltrating CD8⁺ T cells exhibited high levels of CD62L and Tcf1 expressions following the Prodrug 201 plus α-PD-1 combo-therapy (Extended Data Fig. 10i–k).

We also examined the impact of the combo-therapy on tdLNs (Fig. 5n–p). Prodrug 201 plus α-PD-1 increased the CD44⁻CD62L⁺ and CD44⁺CD62L⁺ subsets among AH1-specific CD8⁺ T cells in tdLNs, compared with PBS or α-PD-1 groups (Fig. 5n). In addition, the combination therapy increased the stem-like biomarkers in AH1-specific CD8⁺ T cells in tdLNs, as evidenced by the increased frequency of Sca-1⁺PD-1⁻CD8⁺ T cells and decreased frequency of Sca-1⁻PD-1⁺ CD8⁺ T cells, compared with the PBS group (Fig. 5o and Extended Data Fig. 10l). Specifically, among these AH1-tetramer⁺Sca-1⁺PD-1⁻CD8⁺ T cells, the frequencies of CD44⁺CD62L⁺CD8⁺ T cells and CD44⁻CD62L⁺CD8⁺ T cells were significantly higher in the combo-therapy group, compared with α-PD-1 group (Fig. 5p and Extended Data Fig. 10l). In particular, CD44⁻CD62L⁺Sca-1⁺PD-1⁻CD8⁺ T cells are stem-like CD8⁺ T cells characterized by their longevity and stemness^15,16. Collectively, our data showed that Prodrug 201 is a potent oral immunomodulator that expands antigen-specific stem-like CD8⁺ T cells in the systemic compartment as well as in the TME, thereby contributing to the antitumour efficacy of α-PD-1 ICB therapy.

DHB improves adoptive T cell immunotherapy

Lastly, as another application of DHB-based expansion of stem-like CD8⁺ T cells, we evaluated whether DHB treatment can improve the therapeutic efficacy of adoptive T cell therapy (ACT). Recent studies suggest that Foxo1 plays a critical role in promoting T cell longevity through stem-like programming, underscoring its importance in T cell therapy^8,37. DHB treatment of OT-I CD8⁺ T cells ex vivo significantly increased the frequency of CD127⁺PD-1⁻CD8⁺ T cells (Fig. 6a), a subset of T cells reported as long-lived memory T cells³⁸. Within this subset of T cells, DHB significantly increased the expression of Foxo1 and Tcf1 (Fig. 6b,c) and concomitantly upregulated Ki67, a proliferation marker (Fig. 6d). We tested if OT-I CD8⁺ T cells expanded ex vivo in the presence of DHB could exert strong antitumour efficacy in B16F10-OVA-bearing mice. Adoptive transfer of DHB-trained OT-I CD8⁺ T cells significantly delayed B16F10-OVA tumour growth and induced complete tumour regression in two out of ten mice, whereas the adoptive transfer of control OT-I CD8⁺ T cells failed to slow the tumour growth (Fig. 6e). The analysis of T cells in the TME on day 21 revealed that ACT with DHB-trained OT-I CD8⁺ T cells significantly increased intratumoural infiltration of transferred (Thy1.1⁺) CD8⁺ T cells and their Ki67 expression (Fig. 6f,g). Besides, the adoptive transfer of DHB-trained OT-I CD8⁺ T cells also inhibited NOOC1-OVA tumour growth, with increased frequencies of Tcf1⁺CD8⁺ T cells, and CD44⁺CD62L⁺CD8⁺ T cells among antigen-specific CD8⁺ T cells in PBMCs (Supplementary Fig. 7).

Fig. 6. DHB improves ACT.

a–g, OT-I CD8⁺ T cells were isolated from splenocytes from OT-I mice and cultured with SIINFEKL peptide (4 nM) and DHB for 3 days. Cells were then washed twice and further cultured with DHB in the presence of IL-2 (10 ng ml⁻¹) for 5 days. Frequencies of CD127⁺PD-1⁻CD8⁺ T cells (a), Foxo1⁺ (b), Tcf1⁺ (c) and Ki67⁺ (d) cells among CD127⁺PD-1⁻CD8⁺ T cells. e, Therapeutic regimen of the adoptive cell therapy with DHB- or PBS-treated OT-I CD8⁺ T cells in B16F10-OVA-tumour-bearing mice, and the average tumour growth curves. f,g, Analyses of CD8⁺ T cells in the TME on day 21. Frequencies of Thy1.1⁺CD8⁺ T cells (f) and Ki67 mean fluorescence intensity (MFI) values (g) among Thy1.1⁺CD8⁺ T cells. Data represent the mean ± s.e.m. (n = 16 (a–d); n = 10 for DHB-trained OT-I CD8 T cells and n = 9 for other groups (e); n = 8 for DHB-trained OT-I CD8 T cells, and n = 9 for other groups (f and g) biologically independent samples). Data were analysed using a two-sided, unpaired Student’s t-test (a–d), two-way ANOVA with Bonferroni’s multiple comparisons test (e) or a one-way ANOVA with Bonferroni’s multiple comparisons test (f and g). Mouse icon in e created in BioRender; Na, Y. https://BioRender.com/j0elmnw (2026).

Source data

On the basis of these results, we evaluated whether DHB-based Prodrug 201 could improve ACT in situ. Indeed, in the B16F10-OVA tumour model, the oral administration of Prodrug 201 increased the antitumour efficacy of adoptive OT-I CD8⁺ T therapy (Supplementary Fig. 8a,b). Smart-sequencing-based bulk RNA-seq analysis of tumour-infiltrating OVA-specific OT-I CD8⁺ T cells further revealed that genes associated with glycolysis, inflammatory signalling, oxidative stress and dysfunctional T cell states were downregulated, including Acacb³⁹, Slc16a5 (ref. ⁴⁰), Ifnab⁴¹, Nos1 (ref. ⁴²), Pla2g5 (ref. ⁴³) and multiple Klra family members⁴⁴ (Supplementary Fig. 8c,d). Taken together, DHB or Prodrug 201 treatment endows CD8⁺ T cells with a persistent stem-like phenotype and enhances their tumour infiltration, thereby improving their antitumour efficacy.

Conclusions

In summary, we have synthesized a DHB-derived prodrug nano-emulsion and shown that oral prodrug therapy induced antigen-specific stem-like T cell responses in vivo and amplified the therapeutic efficacy of α-PD-1 immunotherapy in multiple ICB-refractory murine tumour models. To the best of our knowledge, this is the first demonstration of a microbial-metabolite-based oral drug designed to improve stem-like T cell immunity. Our oral prodrug formulation with a well-defined composition and chemical structure is well positioned for large-scale manufacturing, quality control and patient compliance, highlighting its strong translational potential. More broadly, since microbial metabolites play critical roles in host-immune interactions and are associated with various immune-related diseases, this study provides a general pathway for harnessing microbial metabolites in innovative drug discovery and development.

Methods

In vitro differentiation and survival of OT-I T cells with various metabolites

Splenocytes from OT-I mice (6–8 weeks old) were obtained aseptically. The cells were cultured in 96-well plates (1 × 10⁵ cells per well). For the CD8⁺ Tcm differentiation study, cells were activated with SIINFEKL peptide (4 × 10⁻⁹ M) in the presence of various metabolites (100 µM of N-acetylneuraminate, disodium sebacate, DHB, pantothenate, GABA, glycerol, 5-aminoimidazole-4-carboxamide ribonucleotide or ferulate) for 3 days. Then, the cells were washed several times with a T cell medium and further cultured with IL-2 (10 ng ml⁻¹, PeproTech) in the presence of metabolites for 5 days. The suspension cells were collected, blocked with CD16/32 antibody, and stained with the fixable viability dye eFluor 450, FITC-CD44 rat anti-human/mouse (clone IM7, eBioscience), PE-CD62L monoclonal antibody (clone MEL-14, BioLegend) and APC-CD8 rat anti-mouse (clone 53-6.7, BD Biosciences). For CD8⁺ T cell survival study, cells were activated with an antigen peptide as above, followed by thorough washing, and incubation in the presence of IL-2 for 2 days. Metabolites were added during this incubation (100 µM of N-acetylneuraminate, disodium sebacate, DHB, pantothenate, GABA, glycerol or 5-aminoimidazole-4-carboxamide ribonucleotide). Then, the cells were cultured in the absence of both IL-2 and metabolites for 36 h. CD44⁺CD62L⁺ and live T cells (stained with Annexin V-APC and PI) were detected via flow cytometry. For the Tcf1⁺CD8⁺ differentiation study, cells were activated with antigen peptide as above, followed by thorough washing, and cultured with metabolites in the presence of IL-2 for 3 days. Cells were stained with the PE-Tcf1/Tcf7 rabbit monoclonal antibody (clone C63D9, Cell Signaling Technology). For the isolated CD8⁺ T cell differentiation study, CD8⁺ T cells isolated from the spleen of OT-I mice were activated with antigen peptide as above, washed with the T cell medium, and further cultured with DHB in the presence of IL-2 for 5 days. For isolated CD8⁺ T cell co-cultured with BSO, rapamycin or GSK2837808A, CD8⁺ T cells were isolated from splenocytes in C57BL/6 mice with CD8⁺ T cell isolation kit, cultured in 96-well plates (1 × 10⁵ cells per well), and activated with anti-CD3 and anti-CD28 (1 µg ml⁻¹) antibodies in the presence DHB (100 µM), BSO (25 µM), rapamycin (10 nM) or GSK2837808A (15 µM) for 3 days. Then, the cells were washed twice and further cultured with IL-2 (10 ng ml⁻¹) in the presence of the same drug mentioned above for 4 days. Cells were stained with BV570-anti-mouse CD8a antibody (BioLegend, clone 53-6.7), PerCP-Cy5.5-anti-mouse CD127 antibody (Fisher Scientific, clone A7R34), FITC-anti-mouse CD279 (PD-1) antibody (eBioscience, clone RMP1 30), PE-anti-mouse/human/rat Foxo1 antibody (Cell Signaling, clone C29H4), APC-anti-mouse/human Tcf1/Tcf7 antibody (Cell Signaling, clone C63D9), AF700-anti-mouse Ki67 antibody (BioLegend, clone 16A8) and live/dead fixable near-IR (Invitrogen, number L10119). For the anti-CD3/CD28 antibodies activation setting, CD8⁺ T cells were isolated and seeded in a hamster IgG precoated plate at a density of 10⁵ cells per well. DHB, anti-CD3/CD28 antibodies (1 µg ml⁻¹) and IL-2 (100 U ml⁻¹) were added to activate the T cells. After 48 h, cells were washed for two times, and the cells were further incubated with DHB and IL-2 (10 ng ml⁻¹) for 5 days. Tcm cells were quantified by flow cytometry.

In vivo cancer immunotherapy of prodrug formulation plus α-PD-1

For the formulation screening study in vivo, s.c. CT26 tumour model was established as mentioned above. The formulation containing either Prodrug 201, Prodrug 863 or Prodrug 142 (molar equivalent to 0.11 mg of DHB per dose) was orally administered every other day starting on day 7. α-PD-1 antibody (100 µg per dose) was injected intraperitoneally (i.p.) on days 9, 12, 15 and 19. The PBMCs were obtained on days 17 and 22 and stained with PE-tagged peptide-MHC-I tetramer (H-2L^d-restricted SPSYVYHQF). In the B16F10 tumour model, 3 × 10⁵ B16F10 cells were inoculated in C57BL/6 mice (5–6 weeks old), Prodrug 201 or free DHB was orally administered every other day starting on day 6. α-PD-1 antibody (100 µg per dose) was injected i.p. on days 7, 9, 11 and 14. In the NOOC1 tumour model, 1.5 × 10⁵ NOOC1 cells in the matrix gel were inoculated in C57BL/6 mice, Prodrug 201 or free DHB was orally administered every other day starting on day 8. α-PD-1 antibody (100 µg per dose) was injected i.p. on days 14, 20 and 26. For antitumour efficacy comparison between Prodrug 201 and Prodrug 201 plus α-PD-1, Prodrug 201 was orally administered every other day, starting on day 7. α-PD-1 antibody (100 µg per dose) was injected i.p. on days 10, 14, 18 and 22. The PBMCs were collected on days 14 and 20 for further analysis. The body weight, tumour length and width, and animal survival were recorded. For the CT26 tumour rechallenge study, the surviving animals after Prodrug 201 plus α-PD-1 treatment were inoculated with 1.5 × 10⁵ CT26 tumour cells on the contralateral flank by s.c. injection on day 90, and the tumour growth was monitored for additional 60 days.

Statistical analysis

The results are expressed as the mean ± s.e.m. A one-way or two-way analysis of variance (ANOVA), followed by Bonferroni’s multiple comparisons test was used for testing differences among groups. Data were approximately normally distributed, and variance was similar between the groups. The experiments were repeated multiple times as independent experiments, as indicated in the figure captions. A complete dataset from one representative, independent experiment is shown in the figures. No samples were excluded from the analyses. Statistical significance is indicated as *P < 0.05. GraphPad Prism 10 was used for the statistical analyses.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41565-026-02235-9.

Author contributions

K. Han, Y.S.C. and J.J.M. designed the study. K. Han and Y.S.C. performed the experiments. H.E.D., Q.W., X.Z., F.X., J.W., C.L., H.J., J.X. and Y. Xu assisted with the flow cytometry experiments. Y.K. assisted with the ACT in vivo study. J.C., G.Y.C. and Z.W. assisted with the in vivo efficacy studies. M.T. and E.C. assisted in the WB assay. K. Hutchings and M.C. assisted with the prodrug synthesis. C.X. assisted with the TEM imaging. Y.N. and A.S. assisted with the MS experiments. M.N. and F.W. assisted in the Seahorse assay. A.A., H.W., L.Z., O.A., C.A.L. and D.N. assisted in the metabolism experiments. Y.W., Y. Xie, K.D. and Y.L.L. assisted with the scRNA-seq and ATAC-seq experiments and analyses. K. Han, Y.S.C. and J.J.M. interpreted the data and wrote the paper.

Peer review

Peer review information

Nature Nanotechnology thanks Li Tang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Funding

J.J.M. discloses support for the research of this work from NIH (R01DE030691, R01CA271799, R01NS122536, R01DE031951 and P30CA046592). Y.L.L. discloses support for the research of this work from NIH (R01DE026728). D.N. discloses support for the research of this work from NIH (R01CA271369). K. Han discloses support for the research of this work from the National Natural Science Fund for Excellent Young Scientists Fund Program (Overseas) of China, the National Natural Science Foundation of China (number 52473160), and the Project Program of Jiangsu Key Laboratory of Drug Design and Optimization, China Pharmaceutical University (number DDOZZ202301). D.N. is supported by a Rogel Cancer Center grant and a Forbes Scholar Award. J.C. is supported in part by the NIH T32 training grant (T32GM145304).

Data availability

The single-cell RNA-seq and ATAC-seq data generated in this study have been deposited in the Sequence Read Archive (SRA) of the National Center for Biotechnology Information (NCBI) under BioProject accession number PRJNA1468711. The bulk RNA-seq analysis data of SIINFEKL-activated OT-I splenocytes treated with DHB have been deposited in the SRA of the NCBI under BioProject accession number PRJNA1475719. The Smart-sequencing-based bulk RNA-seq data have been deposited in the SRA of the NCBI under BioProject accession number PRJNA1475007. All other data are available in the Article or the Supplementary Information. Source data are provided with this paper.

Code availability

The codes used in paired single-cell RNA-seq/ATAC-seq (BioProject accession number, PRJNA1468711), bulk RNA-seq analysis of SIINFEKL-activated OT-I splenocytes treated with DHB (BioProject accession number, PRJNA1475719), Smart-sequencing-based bulk RNA-seq (BioProject accession number, PRJNA1475007) are available via GitHub at https://github.com/yuyingxie/DHB_Prodrug.git.

Competing interests

Patent applications for microbial metabolite prodrug-based oral formulation to improve the efficacy of ICB have been filed, with J.J.M., K. Han, D.N., K. Hutchings and M.C. as inventors. J.J.M. declares financial interests for board membership, as a paid consultant, for research funding, and/or as an equity holder in EVOQ Therapeutics, and Saros Therapeutics. C.A.L. has consulted for Astellas Pharmaceuticals, Odyssey Therapeutics, Third Rock Ventures and T-Knife Therapeutics, and is an inventor on patents pertaining to Kras-regulated metabolic pathways, redox control pathways in pancreatic cancer and targeting the GOT1-ME1 pathway as a therapeutic approach (US patent number 2015126580-A1, 05/07/2015; US patent number 20190136238, 05/09/2019; international patent number WO2013177426-A2, 04/23/2015). The other authors declare no competing interests.

Extended data

is available for this paper at https://doi.org/10.1038/s41565-026-02235-9.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41565-026-02235-9.