1. A high proportion of M2-TAMs in bladder cancer patients predicted poor prognosis and correlated with high PD-L1 expression.
To clarify the impact of M2-TAMs on the prognosis of BLCA patients, we analyzed The Cancer Genome Atlas (TCGA) database and found that high expression of CD206 and CD163—biological markers of M2-TAMs—and a high prevalence of M2-TAMs were associated with a poor prognosis in patients with BLCA (Fig. 1A, B, C). Furthermore, CD206 and CD163 were strongly positively correlated with PD-L1 (Fig 1D, E). In addition, M2-TAM in the TCGA database was more prevalent in BLCA T-stages 3 and 4 than stages 1 and 2, high-grade than low-grade, and clinical stages 3 and 4 than stages 1 and 2 (Fig. 1 F, G, H). In subsequent tests, we validated the relationship between M2-TAMs and PD-L1 in 70 samples of BLCA tissues. The results indicated that 46% of the BLCA patients had high expression of CD206 (immunohistochemistry (IHC) score >8), and CD206 was positively correlated with PD-L1 (Fig. 1J). High CD206 expression was associated with poor prognosis (Figure 1K), which was consistent with the findings in the TCGA database. In vitro, an M2-TAM model was also constructed, which confirmed the high expression of CD206, IL-10 and TGF-β, suggesting successful modelling (Supplementary Fig. S1A, B, C, D, E). Two highly malignant G3 BLCA cell lines[10], T24 and J82, were respectively cocultured with M2-TAMs, and the results showed that PD-L1 was significantly overexpressed in the tumor cells (Fig. 1L). In vivo, a murine M2-TAM hyperoccupancy model was constructed successfully, and the high expression of the CD206, CD163 and TGF-β markers was verified (Supplementary Fig. S1M, N, O, Q). Preferential growth of the tumor was also found in M2 hyperoccupant subcutaneous tumor-bearing C57BL/6 N mice (Fig. 1M, N), and IHC analyses also showed that M2 hyperoccupant tissues had high expression of CD206 and PD-L1 (Fig. 1O). Collectively, these findings suggest that M2-TAM hyperoccupancy predicts poor prognosis and is correlated with high PD-L1 expression.
2. M2-TAMs promote glycolysis and PD-L1 expression in BLCA through TGF-β secretion
Transcriptome sequencing was performed for T24 cells after coculture with M2 TAMs, and 1323 differential expression genes (DEGs) were identified, including 742 upregulated genes and 581 downregulated genes (Fig. 2A, B). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of the DEGs revealed that the signatures of “cytokine-cytokine receptor”, “TGF-β signaling”, “central carbon metabolism in cancer” etc. were linked to the regulation of M2-TAMs in the coculture system (Fig. 1C). After coculture with M2-TAMs, tumor cells exhibited a significant increase in the expression of key enzymes (GLUT1, HK1, HK2, PFKFB3, PFK2, ENO1, GPI, GAPDH, PKM2, LDHA) involved in glycolysis and PD-L1 mRNA expression (Fig. 2D and E). This phenomenon, together with the association with TGF-β, was indicated by signatures related to "central carbon metabolism in cancer", "PD-L1 and PD-1 checkpoint in cancer", and "TGF-β signaling". Notably, GSEA revealed significant enrichment of glycolytic metabolic pathways in the high-TGF-β subgroup (TGFB1) (Fig. 2F). Furthermore, in the TCGA database, TGF-β was positively correlated with the glycolytic enzymes PKM2, PFKFB3 and LDHA (Fig. 2G, H, I). In vitro experiments confirmed that coculture with M2-TAMs caused significant upregulation of key glycolytic enzymes and PD-L1 in tumor cells in the coculture system, but these changes were partly reversed by pretreatment of M2-TAMs with siTGF-β, and glycolytic enzymes and PD-L1 were elevated again after the addition of recombinant TGF-β (Fig. 2J). Given the possibility of mutual polarization of macrophage subtypes, it is necessary to clarify the interference of M0 and M1-TAMs to M2-TAMs secreted factors. Then M0, M1 and M2 TAMs were induced successfully from THP-1 cells, and M1 TAM and M2 TAM markers were highly expressed (Supplementary Fig. S1G, H, I, J, K, L), suggesting that the models were successfully constructed. Subsequently, the tumor cells were separately cocultured with M0, M1 or M2 macrophages, and the M2 group exhibited the highest expression levels of TGF-β (Fig. 2K) and PD-L1 (Fig. 2L). This observation was consistent with the bioinformatics analysis of the GEPIA2021 database, in which M2-TAMs displayed superior TGFB1 expression compared to the other subtypes (Fig. 2M). Furthermore, a significant correlation was noted between TGF-β (TGFB1) and CD206 in the TCGA database (Fig. 2N). Finally, an analysis of BLCA patients indicated that those with high TGF-β expression had a poorer prognosis (Fig. 2O), consistent with the hyperoccupancy of M2-TAMs. These collective findings underscore the role of M2-TAMs in promoting glycolysis and PD-L1 expression in BLCA through the influence of TGF-β.
3. TGF-β promotes tumor progression and PD-L1 expression via PKM2
The above study demonstrated that, among the key enzymes involved in glycolysis, PKM2, the last rate-limiting enzyme of the glycolytic process, has the strongest correlation with TGF-β (Fig. 2G). Thus, we further investigated the role of PKM2 in the regulation of M2-TAMs by TGF-β in promoting malignant progression and PD-L1 expression in BLCA cells. PKM2 and PD-L1 expression was significantly greater in BLCA tissues than in paracancerous tissues (Fig. 3A), and the two genes were correlated with each other (Fig. 3B). PKM2 and PD-L1 expression was significantly upregulated by the addition of recombinant TGF-β to BLCA cells (Fig. 3C). Immunofluorescence further confirmed that the expression of PKM2 could be upregulated by TGF-β and partly reversed by SB-431542, an effective TGF-β receptor blocker (Fig. 3D). Further studies revealed that TGF-β increased invasion and migration in BLCA cells, but these effects were partly reversed by knockdown of PKM2 (siPKM2) (Fig. 3E). Moreover, the proliferative effect of TGF-β on BLCA cells was also significantly reversed by shikonin, one of the most potent and selective inhibitors of PKM2[11-13] (Fig. 3F). Through further TGF-β concentration gradient interventions, we found that TGF-β stably promoted the expression of PKM2, PD-L1 and the proliferation-related marker proliferating cell nuclear antigen (PCNA) (Fig. 3G). Further study revealed that high PD-L1 expression activated by TGF-β was significantly downregulated following PKM2 knockdown (siPKM2) in tumor cells (Fig. 3H). Overall, these findings demonstrated that TGF-β can promote tumor progression and positively regulate PD-L1 expression via PKM2.
4. TGF-β regulates PD-L1 by promoting PKM2 dimer nuclear translocation
PKM2 knockdown or inhibition significantly downregulated the promotive effect of TGF-β on PD-L1 expression in BLCA cells (Figs. 4A, B and S2A). However, interestingly, total PKM2 expression was upregulated but PD-L1 expression was downregulated after the addition of TEPP-46, an activator of PKM2, to tumor cells for 24 h (Figs. 4C and S2B), and PD-L1 expression was still downregulated after TGF-β treatment (Fig. 4D). Thus, a protein crosslinking assay of the cytosolic and nuclear components in tumor cells was performed after 1 ng/ml TGF-β treatment increased to 10 ng/ml, and the results showed that the nuclear PKM2 dimer and cytosolic PKM2 tetramer were upregulated gradually, while the nuclear PKM2 dimer and cytosolic PKM2 tetramer showed no significant changes (Figs. 4E and S2C). Immunofluorescence also revealed that PKM2 was highly expressed and translocated to the nucleus after treatment with TGF-β (Fig. 4F). Synthesis the above results implied that TGF-β promoted PKM2 dimer translocation in tumor cells. Cytosolic nuclear isolation was performed, and the results showed that total lysate and cytosolic PKM2 were highly expressed after TGF-β treatment and further upregulated after treatment with TGF-β + TEPP-46. Nuclear PKM2 was also highly expressed after TGF-β treatment but reversely decreased after treatment with TGF-β + TEPP-46, a trend consistent with the total PD-L1 expression that upregulated by TGF-β and reversely downregulated by TEPP-46 (Fig. 4G). A protein crosslinking assay of total lysates was further performed in tumor cells, the results of which revealed that the expression of PKM2 dimer was consistent with the total PD-L1 expression that was upregulated by TGF-β treatment and reversely downregulated by TGF-β + TEPP-46 treatment, suggesting that both the upregulation of PD-L1 expression by TGF-β and the reverse downregulation by TEPP-46 were likely achieved through the PKM2 dimer (Fig. 4H). Further protein crosslinking assays for cytosolic and nuclear components were performed and showed that the cytosolic PKM2 tetramer was upregulated by TGF-β and further upregulated by TEPP-46, but the cytosolic PKM2 dimer did not change significantly. Furthermore, the change in the nuclear PKM2 dimer was consistent with the increase in total PD-L1 protein expression induced by TGF-β and was reversely downregulated by TEPP-46, suggesting that TGF-β upregulates PD-L1 expression but TEPP-46 downregulates PD-L1 expression by regulating nuclear PKM2 dimer translocation. (Figs. 4I and S2D). These results indicated that the expression of PD-L1 was upregulated by nuclear PKM2 dimer translocation in cells treated with TGF-β and reversely downregulated by the suppression of nuclear PKM2 dimer translocation in cells treated with TEPP-46. Further study revealed that PKM2 knockdown (siPKM2) or inhibition by shikonin partly reversed TGF-β-induced PKM2 tetramer overexpression in the cytosol and nuclear translocation of PKM2 dimers (Figs. 4J, K and S2E). In addition, a 3D spatial molecular structure model of PKM2 was constructed and deposited in the online Protein Data Bank archive (Fig. 4L). Molecular docking prediction revealed that shikonin directly bound to PKM2 to exert its function (Fig. 4M), and the two spatial sites predicted for the interaction between shikonin and PKM2 were asparagine and phenylalanine (Fig. 4N), suggesting that shikonin is one of the most potent and selective inhibitors of PKM2. In summary, the results demonstrated that TGF-β not only upregulated PKM2 tetramer expression in the cytosol but also promoted PKM2 dimer nuclear translocation to upregulate PD-L1 expression.
5. M2-TAM-derived TGF-β activates the JAK2-STAT3 signaling pathway in BLCA
GSEA revealed that the IL-6-Janus kinase (JAK)-STAT3 pathway was positively correlated with TGF-β expression and PD-L1 expression (Fig. 5A, B). Further in vitro studies revealed that the levels of phosphorylated JAK2 (p-JAK2) and p-STAT3 were upregulated by TGF-β activation in BLCA cells, and blockade of the TGF-β receptor by SB-431542 partly inhibited the increase in p-JAK2 and p-STAT3 induced by TGF-β (Fig. 5C). Immunofluorescence assays revealed significant increases in p-STAT3 expression and nuclear density after TGF-β treatment in BLCA cells (Fig. 5 D). In addition, TGF-β increased the expression of p-STAT3 and PKM2 dimers in the nucleus, which was partially reversed by SB-431542 or shikonin in BLCA cells (Fig. 5E, F). We further investigated the role of the classical TGF-β-Smad pathway in the activation of the JAK2-STAT3 pathway and found that the increase in the expression of p-smad3, p-JAK2, p-STAT3, PKM2 and PD-L1 activated by TGF-β decreased as the concentration of SIS3 (an inhibitor of smad3 phosphorylation) or siSmad2 interference (Fig. 5G, H). Collectively, these findings indicate that TGF-β activates the JAK2-STAT3 signaling pathway, resulting in increased expression of p-STAT3, PKM2 and PD-L1 via the TGF-β-smad2/3 pathway.
6. TGF-β promotes the formation of the PKM2 and p-STAT3 complex and nuclear translocation to regulate PD-L1 expression
The regulation of PD-L1 by STAT3 was further evaluated, and the results showed that the protein levels of p-STAT3, PKM2 and PD-L1 gradually decreased with increasing concentrations of static, a STAT3 phosphorylation inhibitor (Fig. 6A), and Stattic treatment decreased the expression of PD-L1 at the mRNA level (Fig. 6B). To further clarify the mutual regulatory effects of PKM2, p-STAT3 and PD-L1, we calculated the binding potency of STAT3 and PKM2 using AlphaFold2, and the results revealed the existence of two pairs of binding sites (Fig. 6C). Coimmunoprecipitation (co-IP) was subsequently used to pull down the PKM2 binding protein, which showed that p-STAT3 bound to PKM2 and that TGF-β increased the binding level (Fig. 6D). We also demonstrated that PKM2 bound to p-STAT3 after pulling down the p-STAT3 protein, and TGF-β increased this binding (Fig. 6E). Further cytosol-nucleus separation assays showed that TGF-β promoted the binding of PKM2 to p-STAT3 in the nucleus (Fig. 6F). Moreover, immunofluorescence staining revealed that TGF-β promoted the translocation of PKM2 and p-STAT3 from the cytosol to the nucleus, as indicated by enhanced fluorescence in the nucleus and significant colocalization of these two proteins (Fig. 6J, K). Furthermore, the high level of PKM2 and p-STAT3 binding induced by TGF-β was partially reversed by stattic in the T24 cell line (Fig. 6G, H). In addition, to further validate the regulatory effects of PKM2 and p-STAT3 on PD-L1 in the T24 and M2-TAM coculture system model, SB-431542 and/or shikonin were added, and it was found that blocking either the TGF-β receptor or the inhibition of PKM2 reversed the coculture-induced increase in PKM2, p-STAT3 and PD-L1 expression and was more significant in the combination group (Fig. 6I). Taken together, these findings suggest that TGF-β promotes the binding of PKM2 to p-STAT3 and its nuclear translocation to upregulate PD-L1 expression in BLCA.
7. Combined blockade of the TGF-β receptor and inhibition of PKM2-p-STAT3 complex formation effectively suppresses tumor and antitumour immune responses
In vitro studies confirmed that blocking the TGF-β receptor and inhibiting the PKM2-STAT3 complex can significantly reduce the expression of PD-L1. Further in vivo studies were performed using mouse subcutaneous tumor models (Fig. 7A), and the results revealed that a high percentage of M2-TAMs promoted tumor growth in mice. Routine intraperitoneal injections of SB-431542, shikonin and SB-431542+PKM2 significantly reduced the volume of the loaded tumors, while the combination treatment had a decreased volume and greater inhibitory effect on proliferation (Fig. 7B, C, D). Subcutaneous tumors from mice were dissected and analyzed, and CD206 was found to be highly expressed in the M2-TAM hyperoccupancy group (Fig. 7E). PD-L1 was also highly expressed in the M2-TAM hyperoccupancy group, and this effect was reversed by SB-431542 and/or shikonin (Fig. 7F). Additionally, flow cytometric analysis via a scientific circle-gate strategy revealed that CD3+ and CD8+ T-cell infiltration was significantly reduced in the M2-TAM hyperoccupancy group, whereas CD3+ and CD8+ T-cell infiltration was increased after treatment with SB-431542 or/or shikonin, with the most infiltration observed in the combination group (Fig. 7G, H, I). Furthermore, flow cytometry was employed to distinguish CD3+ T lymphocytes derived from human peripheral blood mononuclear cells (Fig. 7J). These cells were subsequently cocultured with T24 cells. The outcomes revealed a notable decrease in the mortality rate of T24 cells due to the presence of CD3+ T lymphocytes. However, these effects were diminished by TGF-β. Notably, after 24 hours of treatment with SB-431542 in the coculture system, this decrease was partially reversed (Fig. 7K). Moreover, the levels of Granzyme B (GzmB) and interferon-gamma (IFN-γ) in the coculture system were significantly lower in the TGF-β group than in the control group, but these effects were partially reversed by SB-431542 (Fig. 7L). Similar trends were observed in CD8+ T cells isolated using magnetic beads after treatment with both TGF-β and SB-431542 (Fig. 7M). To demonstrate the effects of SB-431542 and shikonin in human BLCA, fresh human BLCA tissues were cultured into organoids and then analyzed by tissue immunofluorescence after 3 weeks of culture. The BLCA organoid tumor markers CD44 and CK20 were highly expressed in organoid tissues, confirming the successful construction of a BLCA organoid (Fig. 7N). Further analysis revealed that TGF-β did not increase the growth rate of the organoids, suggesting that SB-431542 lost its effect. Shikonin inhibited the growth of the organoids, and interestingly, the combination treatment had a greater inhibitory effect on organoid growth than the control treatment, especially at 72 h (Fig. 7O), suggesting that SB-431542 + shikonin can inhibit the growth and deterioration of BLCA cells. In summary, combined blockade of TGF-β and inhibition of PKM2 in vivo suppressed tumor growth and PD-L1-mediated immune escape and enhanced antitumour immunity.