Wild-type female Balb/c mice (6–10 weeks old) were obtained from Lingchang (Shanghai, China). All mice were bred and maintained under specific pathogen-free conditions at Shanghai Jiao Tong University (Shanghai, China). Animals were housed under a 12-h light/dark cycle with stable temperature (20 °C–24 °C) and humidity (45%–65%). All animal experimental procedures were conducted in accordance with the “Guidelines for the Management of Laboratory Animals” (2024 Revision) issued by the National Science Commission.

The genomic analysis dataset GSE271243 was downloaded from the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo). This dataset contained 18 completely resected HCC tissue samples from three groups of HCC mouse models: HC vs. LB vs. HB (HB group: BBR 30 mg/kg; LB group: BBR 10 mg/kg; HC group: blank solvent only, n = 6 per group). From the HC and HB groups, one tissue sample each (GSM8372134 and GSM8372142) was randomly selected for data processing and analysis.

The Seurat (v5.3.0, RRID: SCR_016341) R package was employed for comprehensive analysis of the scRNA-seq data. Highly variable genes within the dataset were selected for Principal Component Analysis (PCA) dimensionality reduction. Significant principal components (PCs) were selected based on the elbow plot, and the Louvain algorithm was applied for cell clustering. Clusters were visualized using Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction. The Wilcoxon rank-sum test was used to compare gene expression profiles between clusters to identify marker genes for each cluster. Cell types were annotated by referencing the expression of well-established marker genes and by comparison with publicly available single-cell datasets. Differentially expressed genes between clusters of interest were identified using the FindMarkers function.

Bone marrow-derived macrophages (BMDMs) were isolated from mouse femurs and tibias according to previous studies (Zhang et al., 2024). Briefly, cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Gibco, USA) supplemented with 10 ng/mL recombinant mouse M-CSF for 6 days. On day 7, cells were treated with various methods for further assays. The mouse macrophage cell line RAW 264.7 and mouse-derived hepatocellular carcinoma cell line H22 were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). RAW 264.7 and H22 cells were cultured in Roswell Park Memorial Institute (RPMI) medium (RPMI-1640, Gibco) or DMEM supplemented with 10% FCS (Gibco/ThermoFisher) and 1% penicillin/streptomycin (Yeasen Biotech), respectively.

H22 cells (5 × 10⁵) in PBS were subcutaneously implanted into the left flank of Balb/c mice. Starting from day 6 post-implantation, tumor length (A) and width (B) were measured every 2 days. The H22 tumor volume was calculated using the formula: Tumor volume=(A × B²)/2. The mice were humanely euthanized via cervical dislocation, and the tumors were excised and photographed. The maximal tumor size was in line with the animal experimental protocol approved by the Ethics Committee of Shanghai Jiao Tong University. The tumor weight was measured. For the therapeutic studies, mice were randomized into treatment groups, and tumor-bearing mice received intragastric administration (i.g.) of BBR (molecular formula C₂₀H₁₈NO₄ ⁺, molecular weight 336.4) 5 mg/kg or 10 mg/kg every 3 days starting from day 6 for three doses. For immune checkpoint inhibitor treatment, tumor-bearing mice were administered anti-PD-L1 antibody (10 mg/kg) via intraperitoneal injection (i.p.) every 3 days for three doses. The synergy between BBR and anti-PD-L1 was quantitatively assessed using the Bliss independence model, where the expected additive effect (E_Exp) was calculated as EExp = E_A + E_B − (E_A × E_B), with E_A and E_B representing the tumor growth inhibition rates of BBR and anti-PD-L1 monotherapies, respectively. Bliss score = E_AB − E_Exp.

Tumor tissues were shredded and digested with collagenase D (1.0 mg/mL) and DNase I (50 U/mL) (Sigma-Aldrich, St. Louis, USA). Single-cell suspensions were obtained using a gentleMACS Octo dissociator (Miltenyi Biotec, Bergisch Gladbach, Germany) with the program for tumors. The cells were then incubated with FACS buffer and stained with antibodies. For tumor-associated macrophages (TAMs), the following antibodies were used: Fixable Viability Stain 510 (564,406, BD Pharmingen), APC-Cy7-conjugated CD45 (557,659, BD Pharmingen), PE-Cy7-conjugated Ly6C (560,593, BD Pharmingen), BV650-conjugated Ly6G (740,554, BD Pharmingen), FITC-conjugated CD11b (561,688, BD Pharmingen), BV605-conjugated F4/80 (743,281, BD Pharmingen), AF647-conjugated I-A/I-E (562,367, BD Pharmingen), PE-conjugated CD206 (568,273, BD Pharmingen). For T cells, the following antibodies were used: FITC-conjugated CD3 (553,061, BD Pharmingen), BUV737-conjugated CD4 (612,761, BD Pharmingen), PE-CF594-conjugated CD8 (562,315, BD Pharmingen). All Sample data were collected using the LSRFortessa™ X-20 flow cytometer (BD Biosciences, San Jose, USA), followed by analyses using FlowJo software (Version 10.8.1; BD Life Sciences, RRID: SCR_008520).

cDNA was synthesized using the ReverTra Ace qPCR RT kit (Toyobo). RT-qPCR was performed by mixing primers with SYBR green RT-PCR master mix (Toyobo) on the StepOne Plus system (Thermo Fisher Scientific). Data were normalized using GAPDH as a reference. The primer sequences are shown in the table below.

BMDM cells were seeded in 6-well plates and pretreated with BBR (0, 3, 10, 30 μM) for 1 h. Cells were then stimulated with H22 tumor cell-conditioned medium or IL-4 (20 ng/mL) for 1 h. BMDM cells were lysed in RIPA buffer (Sangon) containing protease inhibitors (Sigma-Aldrich). Lysates were centrifuged at 4 °C, and supernatants were collected as protein extracts. Protein concentrations were measured using a BCA kit (Beyotime). Samples were separated on 8% or 10% acrylamide gels by electrophoresis at 120 V. Proteins were transferred to nitrocellulose membranes (GE Healthcare, Little Chalfont, UK) and blocked with 5% skim milk for 1.5 h to prevent nonspecific binding. Primary antibodies were incubated overnight at 4 °C, including: p-STAT6 (56554S, CST), t-STAT6 (9362S, CST), p-JAK1 (3331S, CST), t-JAK1 (3344S, CST), β-actin (8457S, CST), Arg-1 (89872SF, CST), Retnla (Abcam, AB39626). Secondary antibodies were incubated for 2 h. After washing, proteins were visualized using ECL detection reagent (ThermoFisher Scientific, Waltham, USA) and a ChemiDoc XRS + system (Bio-Rad, Hercules, USA).

BMDM cells were seeded in 10-cm dishes and pretreated with BBR (30 μM) for 1 h, followed by stimulation with IL-4 (20 ng/mL) for 30 min. Control cells were treated with vehicle alone or IL-4 only. After treatment, cells were washed twice with PBS and lysed in in mild lysis buffer (NP-40 lysis buffer) supplemented with protease and phosphatase inhibitors. Lysates were cleared by centrifugation at 20,000 × g for 20 min at 4 °C. For each sample, 50 µL of supernatant was mixed with 50 µL of 2×loading buffer as input control. The remaining supernatant was equally divided for incubation with either 2 µg of mouse IgG or 2 µg of anti-JAK1 antibody (3344S, CST) overnight at 4 °C. Protein A/G magnetic beads were then added and incubated for 2 h. Beads were washed four times with NP-40 lysis buffer, and bound proteins were eluted in loading buffer for subsequent Western blot analysis.

The mouse JAK1 structure (PDB code: 7T6F) was modeled by Glassman (Stanford University School of Medicine, USA). The mouse STAT6 structure (PDB code: 4Y5U) was modeled by Jing Li (University of Chinese Academy of Sciences, China). BBR was drawn using ChemDraw Professional (v22.0.0.22; PerkinElmer, RRID: SCR_016768). JAK1 and STAT6 were preprocessed with PyMOL (v2.6; Schrödinger, RRID: SCR_000305), then docked with BBR using AutoDock (v1.5.7; RRID: SCR_012746). Docking results were analyzed and visualized with PyMOL. Interaction diagrams were generated using LigPlot+ (v2.2.5; EMBL-EBI, RRID: SCR_018290) to illustrate hydrogen bonds, hydrophobic interactions, and other critical contacts for binding mode analysis.

For the intact cell assay, RAW 264.7 cells were treated with BBR (30 μM) for 1 h, harvested, and washed twice with PBS to remove unbound molecules. The supernatant was removed, and cells were resuspended in PBS containing protease inhibitors. The cell suspension was evenly distributed into 0.2 mL PCR tubes and heated at a temperature gradient for 3 min. After rapid freeze-thaw cycles in liquid nitrogen (three repetitions), cells were centrifuged at 20,000 × g for 20 min, and the supernatant was collected. A loading buffer was added for subsequent Western blot analysis.

For the cell lysate assay, after washing with PBS, cells were resuspended in PBS containing protease inhibitors and subjected to three rapid freeze-thaw cycles in liquid nitrogen. The lysate was centrifuged at 4 °C and 20,000 × g for 20 min to remove debris. The supernatant was collected, treated with BBR (30 μM) at room temperature for 30 min, and then denatured at a temperature gradient for 3 min. A loading buffer was added for subsequent Western blot analysis.

Cells were washed twice with PBS and lysed in mild lysis buffer (NP-40 lysis buffer) containing 100× protease inhibitor cocktail (PIC) on ice for 10 min. Cells were scraped and centrifuged at 4 °C and 18,000 × g for 10 min to obtain the protein supernatant. The supernatant was divided into separate tubes and incubated with DMSO or various concentrations of BBR (0, 3, 10, 30 μM) at room temperature for 1 h. Samples were then digested with Pronase (Roche) (enzyme-to-protein ratio = 1:800) at room temperature for 30 min. The digestion reaction was stopped by adding 100× protease inhibitor PMSF, followed by incubation on ice for 10 min. A loading buffer was added for subsequent Western blot analysis.

Data are presented as the mean ± SEM, with a sample size of n ≥ 3 for all experiments. Detailed statistical parameters of the analyses are provided in Figure Legends. For two groups, data were compared using a two-tailed Student’s test. For multiple groups, data were compared using one-way analysis of variance (ANOVA) Dunn’s post hoc analysis or two-way ANOVA (Two-way ANOVA) Tukey’s post hoc analysis. Significance is presented as *P < 0.05, **P < 0.01, and ***P < 0.001.

To examine whether BBR has an inhibitory effect on HCC, we subcutaneously implanted mouse-derived hepatocellular carcinoma cells H22 into the left flank of Balb/c mice. The BBR treatment protocol is shown in Figure 1A. We found that both 5 mg/kg and 10 mg/kg BBR effectively inhibited tumor growth, and the inhibitory effect was positively correlated with BBR dosage (Figures 1B–D). At 16 days after H22 cell inoculation in Balb/c mice, tumor volume in the PBS group reached 600 mm³. The final tumor volume in the BBR (10 mg/kg) group was less than 200 mm³. BBR treatment significantly reduced tumor volume by more than threefold (Figure 1C). Tumor weight in the BBR (10 mg/kg) group decreased by at least 50% compared to the PBS group (Figure 1D). Notably, BBR administration showed no significant impact on mouse body weight (Figure 1E), confirming its safety profile. These data unequivocally underscore the potent anti-tumor effects of BBR in HCC.

To investigate the impact of BBR treatment on the tumor immune microenvironment in HCC, we analyzed single-cell data from BBR-treated and untreated mouse HCC tissues in the GSE271243 dataset (one randomly selected sample from each group). Unsupervised clustering of the single-cell data identified six distinct cell populations, including TAMs, neutrophils, DCs, T cells, cancer-associated fibroblasts (CAFs), and B cells (Figure 1F). Notably, the proportion of TAMs was significantly reduced in BBR-treated mice. In contrast, the proportions of T cells and neutrophils were higher in BBR-treated mice compared to untreated mice (Figure 1G). Further analysis of macrophage subsets via scRNA-seq revealed four types: M1-TAMs (characterized by high expression of Ptgs2, Tnf, and cytokines Il1b, Ccl2, Osm), M2-TAMs (characterized by high expression of Arg1, Mrc1, Chil3, and Tgfb1), M1/M2 Transitional TAMs (M1/M2-TAMs, characterized by co-expression of these genes), and Angio-TAMs (characterized by high expression of Col1a1) (Figure 1H). Among these subsets, the proportion of M2-TAMs was notably reduced in BBR-treated mice. Whereas the proportion of M1-TAMs was slightly elevated, the difference in the number of cells was not significant, suggesting that the reduced macrophage population was predominantly M2-TAMs (Figure 1I). Differential gene expression analysis using FindMarkers on tumor-associated macrophages revealed significant downregulation of M2 polarization markers (Arg1, Tgfb1) post-BBR treatment (Figures 1J,K). These data suggest that BBR treatment alters the composition of the tumor microenvironment in HCC mice, particularly affecting immune cell infiltration.

To further investigate how BBR modulates immune cells in the tumor microenvironment, we analyzed immune cell populations of mouse-derived hepatocellular carcinoma cells H22 tumor-bearing xenograft mice given BBR treatment by flow cytometry (Figure 2A). First, we found that BBR treatment increased the proportion of CD45⁺ immune cells among live cells, with the low-dose group reaching approximately 20% of total cells (Figure 2B). Moreover, BBR promoted increased proportions of CD3⁺ T cells, CD4⁺ T cells and CD8⁺ T cells, with the high-dose group showing more than 2-fold increases (Figures 2C–E). Simultaneously, BBR treatment reduced the proportion of monocytic myeloid-derived suppressor cells (M-MDSCs) (Figure 2F). Compared to the PBS group, BBR treatment decreased the proportion of TAMs in immune cells, with the low-dose group showing a significant 50% reduction (Figure 2G). While no significant effect was observed on MHC-II⁺ M1-macrophages (Figure 2H), BBR treatment markedly reduced the proportion of CD206⁺ M2-macrophages (Figure 2I). These findings suggest that BBR treatment can enhance CD8⁺ T cell infiltration and reduce M2 polarization of tumor-infiltrating macrophages, thereby promoting anti-tumor immunity in vivo.

To investigate whether BBR can inhibit M2 polarization of macrophages in vitro, we co-cultured BMDM cells with H22 tumor cell-conditioned medium. We found that H22 tumor cell-conditioned medium promoted the expression of M2 macrophage marker genes, while BBR significantly suppressed these genes at the mRNA level. The mRNA levels of Arg1, Retnla, Il10, Mrc1, Tgfb1, and Chil3 were all inhibited by BBR (Figures 3A–F), with Arg1 expression being suppressed by up to 3-fold. Consistently, at the protein level, BBR markedly suppressed tumor-conditioned medium-induced upregulation of Arg-1 and Retnla (Figures 3G–I). In addition, at the protein level, BBR inhibited tumor-conditioned medium-induced phosphorylation of STAT6 (Figures 3J,K), and administration of 30 μm BBR resulted in a 30% decrease in the phosphorylation status of STAT6 compared to the PBS group. Also, a certain concentration of BBR inhibited the phosphorylation activation of JAK1 (Figures 3L,M). These results suggest BBR inhibits H22 tumor cell-conditioned medium-induced M2 polarization of macrophages through the JAK1/STAT6 pathway.

Since IL-4 can induce macrophage differentiation into the M2 phenotype by activating the JAK1/STAT6 pathway, we examined whether BBR could inhibit IL-4-induced macrophage polarization. We found that IL-4 treatment significantly upregulated the mRNA expression levels of M2-type macrophage marker genes. BBR treatment dose-dependently inhibited the mRNA lev els of Arg1, Retnla, Il10, Mrc1, Tgfb1, and Chil3 (Figures 4A–F). Furthermore, 10 μM BBR significantly inhibited IL-4-induced phosphorylation of STAT6 (Figures 4G,H) and JAK1 (Figures 4I,J). Furthermore, co-immunoprecipitation assays revealed that BBR substantially disrupted the physical interaction between JAK1 and STAT6 induced by IL-4 stimulation (Figure 4K), providing mechanistic insight into how BBR attenuates JAK1-STAT6 signaling. These findings indicate that BBR reduces M2 polarization of macrophages by inhibiting the IL-4-JAK1-STAT6 axis.

Given that BBR can disrupt the JAK1-STAT6 interaction and inhibits phosphorylation events along the IL-4-JAK1-STAT6 signaling axis, we hypothesized that BBR might directly bind to JAK1 and inhibit its activity. Through structure-based simulated docking, we used PyMOL software to perform molecular docking verification of active compounds against core targets and constructed a specific binding model (Figure 5A). Additionally, Ligplot was used to identify hydrogen bond and binding sites between them, with detailed binding energy (−6.7 kcal/mol) listed (Figure 5B). Specifically, BBR binds to JAK1 protein by forming a hydrogen bond with residue Leu150 in the FERM domain (Figure 5C), and the key amino acid residues involved in docking are highly conserved among different species of JAK1/2 (Figure 5D). The FERM domain is primarily responsible for the binding between JAK kinase and the intracellular Box1 portion of cytokine receptors, and the interaction between JAK1 and cytokine receptors is crucial for signal activation (Haan et al., 2006). Therefore, BBR’s direct binding to the FERM domain of JAK1 may impair JAK1-cytokine receptor interaction (Figure 5E).

Cellular thermal shift assay (CETSA) results showed that in intact RAW 264.7, at 50.3 °C, JAK1 protein in the DMSO-treated group had degraded by over 50% compared to low temperature. In comparison, the BBR-treated group required 59.4 °C to reach the same degradation. Moreover, in both intact cells and cell lysates, JAK1 protein in the DMSO-treated group was completely degraded at 59.4 °C, whereas the BBR-treated group required 62.2 °C for complete degradation. These results indicate that BBR enhances the thermal stability of JAK1 protein in both intact cells and cell lysates (Figures 5F,G). Correspondingly, the drug affinity responsive target stability (DARTS) assay showed that in the presence of BBR, the proteolytic capacity of protease against JAK1 protein was also weakened (Figures 5H,I). As BBR concentration increased, the anti-proteolytic ability of JAK1 protein gradually improved, suggesting that JAK1 protein is a potential direct molecular target for BBR’s anti-tumor activity.

To investigate the potential synergistic effects of combining BBR with anti-PD-L1 antibody in treating HCC in mice, we administered anti-PD-L1 antibody alongside BBR to H22-bearing mice (Figure 6A). Results showed that both anti-PD-L1 antibody and BBR monotherapies exhibited anti-H22 solid tumor effects in vivo, with BBR monotherapy demonstrating superior efficacy compared to anti-PD-L1 antibody monotherapy. Moreover, the combination of these two therapies was more effective in inhibiting tumor growth than either BBR or anti-PD-L1 antibody monotherapy alone (Figure 6B), as quantitatively demonstrated by a calculated Bliss synergy score of +0.0045, indicating synergistic interaction. Compared to BBR monotherapy and anti-PD-L1 antibody monotherapy groups, H22 tumor-bearing mice receiving combination therapy showed significantly reduced solid tumor volume and weight (Figures 6C,D). We further analyzed immune cell infiltration in tumor tissues. As shown in Figures 6E–L, compared to PBS group, the combination therapy group showed significant increase in CD45⁺ immune cells, accounting for approximately 30% of total cells (Figure 6E). The tumor-suppressive T cell populations also markedly expanded, with CD3⁺ T cells increasing by over 10% (Figure 6F). Additionally, the proportion of CD4⁺ T cells among CD45⁺ T cells increased (Figure 6G). Notably, CD8⁺ T cell numbers increased approximately 3-fold compared to PBS group (Figure 6H). The combination of anti-PD-L1 antibody and BBR further reduced M-MDSC cell numbers (Figure 6I) and decreased the proportion of TAMs (Figure 6J). Specifically, MHC-II⁺ anti-tumor M1-macrophages showed minimal changes (Figure 6K), while tumor-promoting CD206⁺ M2-macrophages were significantly reduced after combination treatment (Figure 6L). These results indicated that the combination therapy enhanced the efficacy of anti-PD-L1 antibody therapy in HCC treatment.