Abstract Background Tumor cells manipulate the tumor-associated antigens presentation to escape immune surveillance; however, the molecular mechanism is not exactly clear and the measure to intervene is missing. Methods Annexin A2 was knockout by the CRISPR-Cas9 or blocked by the small-molecule matrine, PY60, and hexapeptide. Chemically and genetically induced primary liver cancer models, and the orthotopically implanted liver tumor model were used. Tumor immune environment was analyzed by single-cell sequencing. Annexin A2-interacted proteins and tumor-associated antigens were identified by co-immunoprecipitation coupled with liquid chromatography with tandem mass spectrometry. Tumor cells killing effects were evaluated by co-culture of tumor cells and CD8^+ T cells. Results Targeting Annexin A2 effectively suppressed the progression of liver cancer. The immunosuppressive microenvironment was improved by Annexin A2 inhibition in tumor tissues. The CD8^+ T cells were increased and activated by targeting Annexin A2. Mechanistically, targeting Annexin A2 inhibited its combination with HSP90. The HSP90-mediated tumor-associated antigens presentation was recovered, and the major histocompatibility complex I-presented short peptides were changed, increasing the tumor cells killing by CD8^+ T cells. Interestingly, Annexin A2 was increased in liver cancer tissues and the overall survival was significantly reduced in patients with high expression. However, Annexin A2 was positively correlated with immune cell infiltration in liver cancer, implying that Annexin A2 was used by tumor cells for immune escape and immunotherapy resistance. Hence, we further confirmed that blocking Annexin A2 increased the therapeutic effects of anti-programmed cell death protein-1 both in vitro and in vivo. Conclusions Taken together, our results identified the role of Annexin A2 in the tumor-associated antigens presentation and immune evasion, which could be an actionable target in cancer immunotherapy. Keywords: Hepatocellular Carcinoma, Immunotherapy, Immune modulatory __________________________________________________________________ WHAT IS ALREADY KNOWN ON THIS TOPIC * Tumor-associated antigens are used by tumor cells for immune escape and immunotherapy resistance. WHAT THIS STUDY ADDS * This study shows Annexin A2 is upregulated in tumor cells to impair tumor-associated antigens presentation, thereafter, targeting Annexin A2 increases the immune killing on tumor cells and improves the programmed cell death protein-1 immunotherapy. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY * This study introduces the new role of Annexin A2 in tumor immunity and proposes actionable measures to intervene in immunotherapy resistance. Introduction Liver cancer is one of the most common cancers worldwide. It reaches 4.3% incidence but 7.8% mortality of all cancers, positioning liver cancer as the third leading cause of cancer death.[59]^1 The onset of liver cancer is insidious and lacks clinical symptoms. Most patients bear advanced liver cancer or have distant metastasis at the time of diagnosis, with a high degree of malignancy, difficulties in therapy, and poor prognosis. Systematic therapy is the mainstay of treatment for advanced hepatocellular carcinoma for over a decade. At present, immunotherapy has made a major breakthrough in hepatocellular carcinoma (HCC) treatments, but there are still some drawbacks such as low response rate, drug resistance, and so on.[60]^2 Due to the frequent exposure to gut-derived antigens, the liver forms an immune tolerance microenvironment.[61]^3 Moreover, the occurrence of liver cancer is often accompanied by chronic inflammation, which also makes the immune microenvironment of liver cancer more complex.[62]^4 Compared with other tumors, the immune microenvironment of liver tumor presents more immunosuppressive characteristics. The changes in the immunogenicity of liver cancer cells, like inactive tumor antigens presentation, also promote the formation of the immune suppressive microenvironment of liver cancer, which plays an important role in immune escape and immunotherapy resistance.[63]^5 6 Therefore, it is important to elucidate the immune escape mechanism of liver cancer cells and discover new targets for liver cancer immunotherapy. Annexin A2 (ANXA2) is ubiquitously expressed in various types of normal and tumor cells. It is present in the cytoplasm and on cell surfaces and functions as a binding protein that attaches or bridges proteins, lipids, membranes, and other structures.[64]^7 Importantly, its expression in cancers is elevated and correlates with a poor prognosis.[65]^8 Recent findings revealed that ANXA2 was involved in both the innate and adaptive immune responses.[66]^9 Previously we identified the natural compound matrine as a small-molecule blocker of ANXA2, with a binding affinity of 12.8 µM (K[d]).[67]^10 Besides, the compound PY60[68]^11 and the hexapeptide LCKLSL[69]^12 13 were also reported to be ANXA2 inhibitors. We then used ANXA2 blockers by matrine, PY60, and hexapeptide, and ANXA2 knockout by Crispr-Cas9 to investigate the role of ANXA2 in hepatocellular carcinoma in the present study. Interestingly, targeting ANXA2 reactivated the heat shock protein 90 (HSP90)-mediated tumor antigen presentation process and the killing on liver cancer cells by CD8^+ T cells. These effects of targeting ANXA2 also augmented the therapeutic effects of programmed cell death protein-1 (PD-1) antibody on liver cancer. Hence, our results elucidated a new mechanism of tumor cell immune evasion by tumor antigens presentation, making ANXA2 a potential candidate for cancer therapy in addition to immune checkpoint inhibitors. Results Knockout of Annexin A2 inhibited tumor growth of primary liver cancer To investigate the role of ANXA2 in tumor progression, we observed its effects in primary liver cancer. ANXA2^+/+ and ANXA2^−/− mice were intraperitoneally injected with diethylmitrosamine (DEN) to induce primary liver cancer ([70]figure 1A). The tumor incidence ([71]figure 1B) was not altered by ANXA2 knockout, compare to the control group (ANXA2^+/+). However, the tumor weights were lower ([72]figure 1C) in the mice with ANXA2 knockout. The alpha-fetoprotein (AFP) level in the serum was increased by DEN compared with the control group, but decreased by ANXA2 knockout compared with the ANXA2^+/+ group ([73]figure 1G), while the aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels were increased by DEN ([74]figure 1H,I), with no statistically significant difference between the ANXA2^+/+ and ANXA2^−/− groups. Figure 1. Knockout of Annexin A2 inhibited primary liver cancer. (A) Annexin A2^+/+ (ANXA2^+/+) or Annexin A2^−/− (ANXA2^−/−) mice were intraperitoneally injected with 25 mg/kg DEN at 2 weeks of age to induce primary liver cancer at 10 months. Tumor tissues were stained by H&E, scale bar: 100 µm. (B) The tumor incidence was not changed by Annexin A2 knockout in both male and female mice. (C) Tumor weights were decreased by Annexin A2 knockout in male mice. (D) ANXA2^+/+ or ANXA2^−/− mice were hybridized with miR-122 knockout mice to induce primary liver cancer at 14 months. Tumor tissues were stained by H&E, scale bar: 100 µm. (E) The tumor incidence and (F) tumor weights of miR-122^−/−ANXA2^+/+ and miR-122^−/−ANXA2^−/− mice. (G) The serum contents of alpha-fetoprotein (AFP), (H) aspartate aminotransferase (AST), and (I) alanine aminotransferase (ALT) of Annexin A2^+/+ or Annexin A2^−/− mice with DEN treatment or miR-122 knockout. *p<0.05; ***p<0.001, tested by one-way analysis of variance. n=5 in Annexin A2^+/+ or Annexin A2^−/− group, n=11 in the Annexin A2^+/+ with DEN group, n=13 in the Annexin A2^−/− with DEN group, n=10 in the miR-122^−/−ANXA2^+/+ group, n=19 in the miR-122^−/−ANXA2^−/− group. DEN, diethylmitrosamine. [75]Figure 1 [76]Open in a new tab In addition, we examined the roles of ANXA2 knockout in the genetic hepatocarcinoma mouse model. Mice lacking miR-122 showed spontaneous liver tumor and represented similar characteristics to human liver tumor.[77]^14 We cross-fertilized ANXA2^−/− mice and miR-122^−/− mice to obtain miR-122^−/− ANXA2^+/+ mice and miR-122^−/− ANXA2^−/− mice ([78]online supplemental figure S1). After 14 months, mice were sacrificed and the tumor incidence and weights were recorded ([79]figure 1D). Due to the low incidence of primary liver cancer in female miR-122^−/− mice, we only included the male miR-122^−/− mice. Likewise, the ANXA2 knockout did not affect the tumor incidence of miR-122^−/− mice, as compared to miR-122^−/− ANXA2^−/− group to the miR-122^−/− ANXA2^+/+ group ([80]figure 1E), but the tumor weights were decreased ([81]figure 1F). The serum level of AFP was also increased by miR-122 knockout compared with miR-122 wildtype, and it was decreased by ANXA2 knockout compared with the ANXA2^+/+ group ([82]figure 1G). Furthermore, the AST and ALT levels were increased by miR-122 knockout, with AST level decreased by ANXA2 knockout but no statistical difference in ALT level between the ANXA2^+/+ and ANXA2^−/− groups ([83]figure 1H,I). These results indicated that ANXA2 knockout could restrict tumor growth, although it did not alter tumor oncogenesis. Blocking Annexin A2 inhibited tumor growth and metastasis of orthotopically-implanted and primary liver cancer As we previously found that the natural product matrine could bind to ANXA2 at 12.8 µM (K[d]), we tested the effects of blocking ANXA2 by matrine in liver cancer. First, we orthotopically implanted luciferase-labeled Hepa1-6 cells in the liver to establish a xenograft hepatocarcinoma mouse model.[84]^15 As shown in [85]figure 2A, compared with the control group ((matrine) MA 0, with no addition of matrine), the addition of matrine in water at 0.2 g/L (MA 0.2) and 0.4 g/L (MA 0.4), according to the affinity concentration, effectively inhibited the tumor volume in the liver and lung. Luminescent intensity analysis confirmed that the total number of tumor cells ([86]figure 2B), and the number of tumor cells in the liver ([87]figure 2C) and lung ([88]figure 2D), were all decreased by matrine intervention, with more significant effects in the high-dose group. H&E staining of tumor morphological characteristics was shown in [89]online supplemental figure S2. Infiltrated monocytes in the liver tumor tissues were increased by matrine, which were further validated below, and no obvious metastatic tumors were observed in the lungs with matrine treatment. We also observed the effects of matrine treatment on the primary liver cancer. Mice received DEN injection at 2 weeks of age and matrine treatment at 9 months of age. At 10 months of age, mice were sacrificed and the results showed that matrine decreased tumor weights of DEN-induced primary liver cancer, compared with the group without matrine treatment ([90]figure 2E). H&E staining of tumor tissues showed that the infiltrated monocytes were also increased ([91]figure 2F). Figure 2. Blocking Annexin A2 by matrine suppressed the tumor progression of orthotopically-implanted and primary liver tumors. (A) C57/BL6 mice were orthotopically implanted with luciferase-labeled Hepa1-6 cells and in vivo imaged every 10 days for 50 days. The pictures at the 50-day point were shown. Compared with the control group (MA 0), mice that were treated with matrine (0.2 g/L or 0.4 g/L in drinking water) showed significant reductions in tumor cells of the liver and lung, as confirmed by the luminescent intensity of total tumor cells (B), liver tumor cells (C), and lung tumor cells (D). (E) DEN was used to induce primary liver cancer in C57/BL6 mice. At 9 months, mice were fed matrine 0.4 g/L drinking water for 1 month. Tumor weights were recorded at 10 months. (F) H&E staining of liver tumor tissues. Scale bars: 50 µm. (G) In vivo imaging of C57/BL6 mice implanted with luciferase-labeled Annexin A2^+/+ cells or Annexin A2^−/− liver tumor cells and received matrine drinking water (0.4 g/L). Pictures were taken every 10 days for 50 days. The 50-day point pictures were exhibited. Luminescent intensity of total tumor cells (H), liver tumor cells (I), and lung tumor cells (J) in mice. *p<0.05; **p<0.01; ***p<0.001, tested by one-way analysis of variance with post hoc Tukey’s multiple comparisons, n=5 in (A) and (G), n=7 in (E). ANXA2, Annexin A2; DEN, diethylmitrosamine; MA, matrine. [92]Figure 2 [93]Open in a new tab To confirm the role of ANXA2 in the above effects, we examined the effects of matrine on implanted tumor cells of ANXA2^+/+ and ANXA2^−/−. As shown in [94]online supplemental figure S3A, DEN-induced liver tumors of ANXA2^+/+ and ANXA2^−/− mice were collected for separating ANXA2^+/+ and ANXA2^−/− tumor cells. ANXA2 DNA fragments and messenger RNA (mRNA) expression were undetectable in ANXA2^−/− tumor cells ([95]online supplemental figure S3B,C). Then cells were labeled with luciferase for in vivo imaging. In the mouse model of orthotopically-implanted liver cancer ([96]figure 2G), the knockout of ANXA2 decreased the tumor volume both in the liver and lung compared with the ANXA2^+/+ cells, which proved the antitumor effects of ANXA2 depletion in tumor cells. Importantly, matrine intervention had no more benefits in the condition of ANXA2 knockout, suggesting that matrine acted through ANXA2. Luminescence intensity analysis revealed that the number of tumor cells in total ([97]figure 2H), in the liver ([98]figure 2I), and in the lungs ([99]figure 2J) were significantly decreased by ANXA2 knockout, which were not further decreased by matrine. These results consolidated the role of ANXA2 in inhibiting liver cancer, as well as the blocking on ANXA2 by matrine. Targeting Annexin A2 increased the infiltration and activation of CD8^+ T cells in liver tumors The involvement of ANXA2 in tumor immunity and the obviously increased number of infiltrated monocytes in liver tumors intrigued us to investigate the immune status of liver cancer with ANXA2 blocking or knockout. The DEN-induced primary liver tumor tissues treated by matrine ([100]figure 2E) were collected and proceeded to single-cell sequencing. A total of 9,340 cells in Con-tumors and 8,508 cells in MA-tumors were identified ([101]figure 3A, n=3). The median genes per cell were 2,875 and 2,615, respectively. Cells were clustered and annotated by specific cell markers ([102]figure 3A,B). Then the Ptprc^+ immune cells were further clustered to subtypes ([103]figure 3C). We used the cell markers Cd3d and Cd3e to represent T cells, in which Cd4 represented CD4^+ T cells and Cd8a represented CD8^+ T cells ([104]figure 3D). Besides, we used Ncr1 and Klrk1 to represent NK cells, Cd19 and Cd79a to represent B cells, Adgre1 and Lyz2 to represent macrophages, Siglech, Ccr9, and Ccr2 to represent dendritic cells, Jchain and Eaf2 to represent plasma cells. Statistic results showed that CD8^+ T cells and dendritic cells were increased by matrine treatment, CD4^+ T cells were unchanged, B cells and macrophages were decreased ([105]figure 3E). Due to the enhanced killing effects of CD8^+ T cells on tumor cells by ANXA2 intervention, we focused on the changes in CD8^+ T cells. Gene expression plots exhibited the significant increases or decreases in genes of CD8^+ T cells by matrine treatment ([106]figure 3F). The upregulated genes were enriched in T-cell receptor signaling pathway and chemokine signaling pathway ([107]figure 3G), indicating that matrine treatment also activated CD8^+ T cells functions. Genes that were involved in the T-cell receptor signaling pathway ([108]figure 3H) or chemokine signaling pathway ([109]figure 3I) were exhibited. The most changed genes in the T-cell receptor signaling pathway included Grap2, Fyn, Lcp2, Nfatc1, Nfkb1 ([110]figure 3H), and the most changed genes in the chemokine signaling pathway included Pik3cd, Nfkb1, Adcy7, Foxo3, Rock2 ([111]figure 3I). Figure 3. Blocking Annexin A2 changed the tumor immune microenvironment and activated CD8^+ T cells in vivo. (A) Tumor tissues from diethylmitrosamine-induced primary liver cancer in C57/BL6 mice were prepared for single-cell sequencing (n=3 in each group). A total of 9,340 cells were identified in the control group and 8,508 cells in the MA group. Cells were clustered by cell markers (B). (C) Immune cells were further clustered to subtypes according to immune cell markers (D). (E) The proportions of listed types of immune cells. (F) Volcano plot of gene expressions in CD8^+ T cells. (G) KEGG enrichment of upregulated genes in CD8^+ T cells. (H) Gene-expression differences in T-cell receptor signaling pathway. (I) Gene-expression differences in the chemokine signaling pathway. Con, control; KEGG, Kyoto Encyclopedia of Genes and Genomes; MA, matrine; t-SNE, T-distributed Stochastic Neighbor Embedding; UMAP, Uniform Manifold Approximation and Projection. [112]Figure 3 [113]Open in a new tab To confirm and quantify immune cells in liver tumors, we separated the lymphocytes from tumor tissues by Percoll density gradient centrifugation. Then lymphocytes were gated by FSC/SSC and CD45-positive cells ([114]online supplemental figure S4). In the DEN-induced liver tumors, ANXA2 blocking by matrine only increased the percentage of CD45^+CD3^+CD8^+ T cells ([115]figure 4A), while ANXA2 knockout increased the percentages of CD45^+CD3^+CD8^+ T cells and CD45^+CD19^+ B cells ([116]figure 4B). The ANXA2 knockout did not affect the immune cells in the liver tissues untreated by DEN. In the implanted liver tumors, both CD4^+ T cells and CD8^+ T cells were increased ([117]figure 4C). The number of CD45^+CD19^+ B cells were not significantly affected. CD45^+NK1.1^+ NK cells increased in the low-dose matrine group, but the difference was not statistically significant in the high-dose group. These differences probably resulted from different tumor origins. Further validation by immunohistochemical staining of CD4, CD8, and NK1.1 in liver tumors also confirmed the significant infiltration of CD8^+ T cells by ANXA2 blocking or knockout ([118]figure 4D). The changes in CD4^+ T cells and natural killer (NK) cells were not identical in different liver tumor models under different treatments. CD4^+ T cells were increased by matrine treatment in the DEN-induced or implanted tumors, while NK cells only increased in implanted tumors ([119]figure 4D). Figure 4. Targeting Annexin A2 increased CD8^+ T cells infiltration and activation in liver tumors. (A) The lymphocytes proportions in DEN-induced liver tumors of C57/BL6 mice treated by matrine. (B) The lymphocytes proportions in liver tissues or DEN-induced liver tumors from ANXA2^+/+ and ANXA2^−/− mice. (C) The lymphocytes proportions in orthotopically-implanted liver tumors of mice treated by matrine. (D) Immunohistochemical staining of CD4, CD8, and NK1.1 in DEN-induced or orthotopically-implanted liver tumor tissues. Scale bar: 50 µm. The expressions of genes involved in the T-cell receptor signaling pathway and chemokine signaling pathway in (E) DEN-induced liver tumors of mice treated by matrine, (F) DEN-induced liver tumors of ANXA2^+/+ and ANXA2^−/− mice, and (G) orthotopically-implanted tumors of mice treated by matrine. The sample size was the same as above. (H) NCG mice were orthotopically implanted with Hepa1-6-luc cells and received 0.4 g/L matrine drinking water. Mice were in vivo imaged every 10 days for 50 days. Pictures of the 50-day point are shown. (I) H&E staining of tumor tissues in the liver and lung. White bar: 50 µm. Black bar: 500 µm. Luminescent intensity of total tumor cells (J), liver tumor cells (K), and lung tumor cells (L). *p<0.05; **p<0.01; *** p<0.001, tested by one-way analysis of variance with post hoc Tukey’s multiple comparisons, n=5. ANXA2, Annexin A2; DEN, diethylmitrosamine. MA, matrine. [120]Figure 4 [121]Open in a new tab The most changed genes involved in T-cell receptor signaling pathway and chemokine signaling pathway were also investigated. In DEN-induced liver tumors, matrine treatment increased the expressions of Grap2, Nfatc1, Foxo3, and Rock2 ([122]figure 4E), while ANXA2 knockout exhibited more capacity, increasing the expressions of Grap2, Fyn, Lcp2, Nfatc1, Nfkb1, Adcy7, Foxo3, and Rock2 ([123]figure 4F). In implanted liver tumors, high concentration of matrine showed stronger effects, as it increased the expressions of Grap2, Nfatc1, Nfkb1, Adcy7, and Foxo3 ([124]figure 4G). To further confirm the role of T cells in the antitumor effects of targeting ANXA2, we observed the role of ANXA2 blocking in NCG mice lacking T cells. Interestingly, in vivo imaging ([125]figure 4H) and analysis ([126]figure 4J–L) showed that in T cells-lacking mice, ANXA2 blocking by matrine did not exhibit significant inhibitory effects on tumor growth and metastasis compared with the control group. H&E staining of tumors of the liver and lungs also showed no obvious difference ([127]figure 4I). In conclusion, matrine treatment or ANXA2 knockout increased the infiltration of CD8^+ T cells, and more importantly, activated the cell signaling pathways to restrict tumor growth. Targeting Annexin A2 inhibited its binding with HSP90 and changed the repertoire of MHC I presenting antigens Because we observed that matrine did not affect the expression of ANXA2 in liver tumor tissues ([128]online supplemental figure S5A), to figure out the mechanism of antitumor activities by targeting ANXA2, we used co-IP coupled with liquid chromatography with tandem mass spectrometry (LC-MS/MS) to identify the proteins interacted with ANXA2 ([129]figure 5A). The identification of proteins immunoprecipitated by ANXA2 revealed that 16 proteins were bound with ANXA2 but hindered by ANXA2 blocking, including ribosomal protein RPSA, fibrinogen FGB, HSP90, immunoglobulin IGHG2 and IGKV, and so on (detailed proteins could be seen in [130]online supplemental table S1). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis indicated that the significant pathway related to antitumor effects was antigen processing and presentation, involving the protein HSP90 ([131]figure 5B). The combination of ANXA2 and HSP90, as well as the blocking by matrine, was verified by immunoblotting after immunoprecipitation ([132]figure 5C). The ANXA2-immunoprecipitated HSP90 was decreased by matrine, and vice versa, the HSP90-immunoprecipitated ANXA2 was also decreased. We also tested another ANXA2 inhibitor PY60[133]^11 and the competitive hexapeptide of ANXA2 (LCKLSL)[134]^12 13 to confirm that targeting ANXA2 inhibited its binding with HSP90 ([135]figure 5D). Immunofluorescence images also confirmed the colocalization of ANXA2 and HSP90, which were detached by the treatment of matrine, PY60, or hexapeptide ([136]figure 5E). However, matrine did not affect the expression of ANXA2 and HSP90 ([137]online supplemental figure S5B), indicating that matrine exerted its antitumor functions by inhibiting the interaction of ANXA2 with HSP90. Interestingly, virtual docking of matrine and ANXA2 showed that matrine ideally bound to the active site of ANXA2 ([138]online supplemental figure S6A), which was located between the ANXA2 and HSP90 docking domain and thus hindered the binding of ANXA2 and HSP90 ([139]online supplemental figure S6B). Figure 5. Targeting Annexin A2 changed HSP90-mediated tumor-associated antigens presentation and increased tumor cells killing by CD8^+ T cells. (A) Huh-7 cells were treated with 20 µM matrine (MA) or negative control (NC). Cell lysates were immunoprecipitated by anti-Annexin A2 or anti-IgG, followed by LC-MS/MS identification. The identified proteins were listed in [140]online supplemental tabe S1. (B) The significant KEGG pathways included antigen processing and presentation, in which HSP90 was involved. (C) Validation of Annexin A2 immunoprecipitation and the decreased combination of Annexin A2 with HSP90 by matrine treatment at 20 µM. (D) The decreased Annexin A2-immunoprecipitated HSP90 and the decreased HSP90-immunoprecipitated Annexin A2 under the treatments of 20 µM matrine, 10 µM PY60, and 10 µM hexapeptide LCKLSL (HEXA). (E) Immunofluorescences and co-localization analysis of Annexin A2 and HSP90 in Huh-7 cells treated with matrine, PY60, and hexapeptide. Scale bar: 23.3 µm. (F) Huh-7 cells were treated with matrine or IFN-γ and the lysates were immunoprecipitated by anti-HSP90 or anti-MHC I. The HSP90- and MHC I-collections were then passed through a 10 kD filter to obtain short peptides for antigen identification. (G) The residue protein samples were determined for HSP90, TAP1, and MHC I by immunoblotting. (H) Venn diagrams of 8-10 mer peptides immunoprecipitated by HSP90 or MHC I. (n=3 in each experiment). (I) The immunoprecipitated HSP90, TAP1, and MHC I under the treatments of matrine, PY60, and hexapeptide. (J) The 8-10 mer peptides immunoprecipitated by HSP90 and MHC I under the treatments of matrine, PY60, and hexapeptide. (K) Tumor cells killing by human naïve CD8^+ T cells on Huh-7 cells under the treatments of matrine, PY60, and hexapeptide. (L–N) Tumor cells killing by mouse naïve CD8^+ T cells on Hepa1-6 cells (L), ANXA2^+/+ cells (M), and ANXA2^−/− cells (N). *p<0.05; **p<0.01, tested by t-test, n=6. ANXA2, Annexin A2; IP, immunoprecipitation; LC-MS/MS, liquid chromatography with tandem mass spectrometry; KEGG, Kyoto Encyclopedia of Genes and Genomes; LDH, lactate dehydrogenase; MHC, major histocompatibility complex; TAP1, transporter 1 of antigen peptides. [141]Figure 5 [142]Open in a new tab Because of the important role of HSP90 in antigen processing and presentation, we determined whether the disassociation of ANXA2 and HSP90 influenced the tumor-associated antigens (TAAs) presentation. First, we observed the expression of major histocompatibility complex class I (MHC I), which is responsible for the TAAs presentation. Unfortunately, the membrane location ([143]online supplemental figure S7A,B), the total protein expression ([144]online supplemental figure S7C), and the mRNA expression ([145]online supplemental figure S7D) of MHC I were all unaffected by matrine treatment, although they were significantly increased by interferon (IFN)-γ. These results suggested that the release of HSP90 did not stimulate the expression of MHC I. We then performed TAAs identification to determine whether the release of HSP90 changed the antigens presented by MHC I. Tumor cells that were treated with matrine or IFN-γ were lysed and immunoprecipitated by HSP90 or MHC I. The HSP90-bound or MHC I-bound elution was passed through a 10 kD filter to collect the short-chain peptides. Finally, the collections were identified by LC-MS/MS and the 8-10 mer peptides were sorted by cluster analysis ([146]figure 5F). The immunoprecipitated proteins were verified by immunoblotting, showing that HSP90 and MHC I were successfully immunoprecipitated ([147]figure 5G). Transporter 1 of antigen peptides (TAP1) was co-immunoprecipitated with HSP90 and increased by matrine treatment ([148]figure 5G). LC-MS/MS results showed significant differences in HSP90-bound peptides among matrine-treated cells, IFN-γ-treated cells, and combination-treated cells ([149]figure 5H). MHC I-bound peptides also differed in these cells. The intersection of HSP90 peptides and MHC I peptides, indicating that antigens were transported from HSP90 to MHC I, was changed by matrine or IFN-γ. However, the alterations were not identical to each other ([150]figure 5H, HSP90-enriched and MHC-I-enriched panel), suggesting that the mechanism of ANXA2 blocking was not the same as that of IFN-γ. We also validated the effects of targeting ANXA2 on antigen presentation by matrine, PY60, and hexapeptide. All of them increased the HSP90-immunoprecipitated TAP1 contents ([151]figure 5I). Interestingly, we found that three specific de novo peptides (8-10 mer antigens) that were co-immunoprecipitated by HSP90 and MHC I in all the matrine, PY60, and hexapeptide groups ([152]figure 5J), while there was no same peptide between the matrine and IFN-γ groups ([153]figure 5H), confirmed that targeting ANXA2 would change the TAAs presentation. However, the results also showed that there were 12 unique peptides in the matrine group, 12 unique peptides in the PY60 group, and 13 unique peptides in the hexapeptide group, which presented the differences of matrine, PY60, and hexapeptide, although they all targeted ANXA2. To determine the terminal effects of changes in antigen presentation, we observed tumor cell death by CD8^+ T cells after ANXA2 knockout or blocking. Since the pre-activated CD8^+ T cells have strong cytotoxicity to cover up the effects of TAAs changes, and the CD8^+ T cells isolated from tumor-bearing mice were exhausted to respond to TAAs changes, we thus used naïve CD8^+ T cells, which needed antigens to be activated. The mouse CD8^+ T cells were isolated from normal mouse spleen, and the human CD8^+ T cells were isolated from healthy human blood. These naïve CD8^+ T cells were used immediately at a 10:1 ratio of CD8^+ T cells: tumor cells after being isolated. As shown in [154]online supplemental figure S8, the co-culture of tumor cells and CD8^+ T cells were established. The LDH release experiment showed that the treatments of matrine, PY60, and hexapeptide enhanced the killing of human CD8^+ T cells on human Huh7 cells ([155]figure 5K). In mouse Hepa1-6 cells, targeting ANXA2 by matrine increased the death of Hepa1-6 cells in the presence of CD8^+ T cells, as shown by cell imaging ([156]online supplemental figure S8) and LDH release ([157]figure 5L). In addition, ANXA2^+/+ tumor cells showed the same response to matrine treatment, with increased cell death by CD8^+ T cells ([158]figure 5M and [159]online supplemental figure S8). However, the ANXA2^–/– tumor cells showed increased tumor cell death by CD8^+ T cells alone, with no further increase by matrine treatment ([160]figure 5N, [161]online supplemental figure S8). Together with these results from ANXA2 inhibition and ANXA2 knockout, targeting ANXA2 effectively changed the TAAs presentation and enhanced tumor cells killing by CD8^+ T cells. Liver cancer cells upregulated Annexin A2 to escape tumor immunosurveillance Using The Cancer Genome Atlas (TCGA) database, we found that the mRNA transcript of ANXA2 was higher in the tumor tissues of liver cancer than in adjacent non-tumor tissues ([162]figure 6A). Notably, high ANXA2 expression was positively correlated with poor prognosis ([163]figure 6B). We also determined the expression of ANXA2 in 29 samples of liver tumor tissues and paired adjacent liver tissues that we previously collected.[164]^15 Both the mRNA expression and protein expression of ANXA2 were upregulated in tumor tissues ([165]figure 6C,D). However, a contradictory phenomenon was that in ANXA2 highly expressed tumor tissues, the infiltration of immune cells was also higher, as analyzed by immune cell markers ([166]figure 6E). Spearman correlation analysis showed that ANXA2 expression was positively correlated with the number of T cells, B cells, macrophages, NK cells, and especially DC cells ([167]figure 6F–J), which were important in antigen presentation. These results implied that ANXA2 was increased by tumor cells in response to immune cells. Together with our experimental results, a rational inference could be made that the high expression of ANXA2 was used by tumor cells to restrict TAA presentation, thus escaping the tumor immunosurveillance. Figure 6. Expression of Annexin A2 in liver tumors and its relation to immune resistance. (A) The transcript expression of Annexin A2 (ANXA2) in human liver tumors (T, n=369) and adjacent non-tumor tissues (N, n=50). (B) The overall survival of patients with liver cancer in the low ANXA2 expression group (bottom-quartile, n=91) and high ANXA2 expression group (top-quartile, n=91). (C) The mRNA expression and (D) protein expression of Annexin A2 in liver tumor tissues and adjacent non-tumor tissues. n=29. (E) The expressions of cell markers of T cell, B cell, NK cell, macrophage, and DC cell in human liver tumors of low ANXA2 expression group (bottom-quartile, n=91) and high ANXA2 expression group (top-quartile, n=91). The correlations of ANXA2 expression with T-cell markers expressions (F), B-cell markers expressions (G), macrophage markers expressions (H), NK cell markers expressions (I), and DC cell markers expressions (J) in human liver tumors, n=369. (K) The killing of CD8^+ T cells on Annexin A2^+/+ and ANXA2^–/– cells treated with PD-1 antibodies and matrine. PD-1 antibodies were used at 20 µg/mL, matrine was used at 20 µM. (L) Mice were orthotopically implanted with Hepa 1–6 cells and received 200 µg/mouse PD-1 antibodies for three times and 0.4 g/L matrine water. The liver tumors and lung metastatic nodes were recorded. (M) Tumor weights in the liver of mice. (N) H&E staining of the metastatic nodes in the lung of mice. *p<0.05, tested by one-way analysis of variance with post hoc Tukey’s multiple comparisons. DC, dendritic cells; LDH, lactate dehydrogenase; MA, matrine; NK, natural killer; PD-1, programmed cell death protein-1. [168]Figure 6 [169]Open in a new tab To determine whether targeting ANXA2 improved the immunotherapy by the immune checkpoint inhibitor, we investigated the effects of ANXA2 knockout or blocker in combination with PD-1 antibodies. In ANXA2 wildtype cells, the tumor cells killing by CD8^+ T cells was increased by PD-1 antibodies, which was further enhanced by ANXA2 blocker matrine ([170]figure 6K). The ANXA2 knockout also showed identical efficacy in combination with PD-1 antibodies; however, matrine was not functional in ANXA2 knockout cells ([171]figure 6K). In vivo, PD-1 antibodies significantly reduced the tumor weights of orthotopically implanted Hepa 1–6 cells ([172]figure 6L,M). Meaningfully, the additional treatment of ANXA2 inhibitor matrine further lowered the tumor weights. The lung metastasis nodes of implanted liver tumor were also reduced by PD-1 antibodies and there was no observable metastatic node in mice receiving additional matrine treatment ([173]figure 6L,N). The tumor weights and the lung metastatic nodes were also decreased by matrine treatment alone compared with the control group in the orthotopically implanted tumor model ([174]online supplemental figure S9). These results demonstrated that ANXA2 blocking or knockout altered the TAAs presentation and prevented the immune evasion of liver cancer cells from CD8^+ T cells, which would enhance the therapeutic effects of immune checkpoint inhibitors. Discussion ANXA2 was first recognized as a chaperone protein in the membrane that binds to phospholipid and membrane-binding proteins. In the endothelial cell membrane, its interaction with the protein S100A10 (also named p11) facilitates the conversion of plasminogen to plasmin, leading to angiogenesis.[175]^16 However, the intracellular location of ANXA2 has been identified in various cell types and is associated with more complicated functions. For example, ANXA2-mediated autophagy was newly reported to promote cancer progression in multicancer types.[176]17,[177]19 Also, it impacts multiple intracellular signaling pathways to change malignant phenotypes of cancer cells.[178]20,[179]22 Mice without ANXA2 showed severe bacterial infection and increased mortality, along with an overactivated TLR4-mediated inflammatory response, suggesting a negative role of ANXA2 in innate immune responses.[180]^23 ANXA2 is also involved in NLRC4 activation, NLRP3 inflammasome formation, and neutrophil infiltration.[181]24,[182]26 For adaptive immunity, evidence regarding ANXA2 is rare and controversial. Chao et al reported that the elevation of ANXA2 in tumor cells inhibits DC maturation and proinflammatory interleukin.[183]^27 However, Marlin et al showed that ANXA2 translocation in stressed tumor cells could be recognized by γδ T cells directly through Vγ8Vδ3 TCR.[184]^28 Interestingly, ANXA2 itself is regarded as a tumor antigen that is applied in cancer immunotherapy alone or in combination with anti-PD-1 antibodies.[185]^29 30 However, the role of ANXA2 in liver cancer and its impact on the immune escape of liver cancer cells were not understood. Here, we showed that the knockout or blocking of ANXA2 effectively inhibited the progression of both primary liver cancer and orthotopically-implanted liver cancer. Then we identified the underlying mechanism as targeting ANXA2 could release its combination with HSP90, which mediated the TAAs presentation through TAP1 and MHC I. The changed TAAs presentation thus activated the naïve CD8+T cells and the tumor cells killing by CD8^+ T cells were consequently enhanced. The immune environment was deeply changed. Interestingly, the expression of ANXA2 was positively correlated with the infiltration of T cells, B cells, macrophages, NK cells, and DCs in human liver tumors. However, the higher expression of ANXA2, the lower overall survival time of patients. The increase in immune cells was ineffective in restricting tumor progression because of the inability to recognize tumor cells. Thus, the high expression of ANXA2 is possibly used by tumor cells to escape immunosurveillance, making ANXA2 a potential target for cancer therapy ([186]figure 7, drawn by using the online tool Figdraw). Figure 7. Annexin A2 is significantly upregulated in liver tumor tissues and correlated to poor prognosis and immunotherapy resistance. Targeting Annexin A2 by gene knockout or small-molecule blockers reactivated HSP90-mediated tumor-associated antigens presentation. This process sensitized tumor cells to immune surveillance and increased the tumor cells killing by CD8+T cells, thus enhancing the antitumor therapy by immune checkpoint inhibitors. GzmB, Granzyme B; IFN, interferon; MHC, major histocompatibility complex; PD-1, programmed cell death protein-1; PD-L1, programmed death-ligand 1; PFN, perforin; TAP1, transporter 1 of antigen peptides; TCR, T cell receptor. [187]Figure 7 [188]Open in a new tab Over the past decade, immunotherapy has been the most promising treatment for cancer.[189]^31 The most widely used clinical methods of immunotherapy are immune checkpoint inhibitors and CAR-T cells,[190]^32 representing two directions of targeting tumor cells and reforming T cells in a mutual-interaction relationship of being killed and killing. Tumor cells use many strategies to escape being killed, including the inhibition of T-cell infiltration, suppression of T-cell function, and activation of regulatory T cells.[191]^33 However, the most relevant mechanism of immune evasion is the loss or change in TAAs to avoid recognition by T cells.[192]^31 Like normal cells, malignant cells produce 8–10 residue-long peptides through the immunoproteasome, which are then transported to the endoplasmic reticulum through TAP1 with the assistance of HSP90 and eventually loaded into MHC I.[193]^34 TAA processing and presentation is an important feature of tumor cells to escape being killed by T cells; however, a measure directly targeting TAA presentation by tumor cells is currently lacking. In cutting-edge reports, tumor antigen presentation was found to be altered by RNAi or small-molecule drugs and showed the desired antitumor effects in experimental animals,[194]35,[195]37 guaranteeing the foundation for their clinical application. Recently, the contradictory roles of HSP90 in antigen presentation were identified by different teams. A knockout of HSP90 disabled antigen presentation;[196]^38 however, low-level inhibition of HSP90 enhanced the antigen presentation repertoire and thus led to increased tumor cell death by cytotoxic T lymphocytes.[197]^37 In the present study, we showed that ANXA2 could be a modulator of TAA presentation through binding HSP90. Its knockout or blocking increased tumor cell death and inhibited tumor progression in vivo, which hinted two possible measures for cancer therapy: targeting ANXA2 RNA interference and small molecular blockers such as matrine. Matrine is a natural product of Sophora flavescens that has potential benefits in cancer therapy.[198]^39 Recently, the immune modulation functions of matrine were identified as relievingimmunosuppression through TNFR1,[199]^40 or inhibiting NF-κB signaling pathways.[200]^41 However, the molecular mechanisms of matrine in tumor immunity remain largely unknown. We previously reported matrine was a ligand of ANXA2. The affinity of matrine for ANXA2 was approximately 12.8 µM (K[d]). Our data found that matrine at 20 µM had no inhibitory effect on cell viability and migration. On increasing it to 2,000 µM, matrine treatment inhibited cell viability ([201]online supplemental figure S10), in accordance with the findings of Gao et al and Wang et al that matrine slightly decreased cell viability at 0.5 mg/mL (approximately 2 mM).[202]^42 43 Nevertheless, our results showed that matrine at 20 µM effectively bound to ANXA2 and blocked its interaction with HSP90. Downstream alterations in TAAs presentation, killing of tumor cell, and inhibition of tumor progression were also observed. However, given the multiple targets of matrine, the use of matrine as an ANXA2 inhibitor was further validated by another ANXA2 inhibitor PY60 and ANXA2 competitive hexapeptide LCKLSL. We verified the roles of PY60 and hexapeptide in ANXA2-HSP90 binding, TAAs presentation, and CD8^+ T cells killing. These results not only confirmed the role of ANXA2, but also consolidated the results from matrine single treatment. Hence, these results made the potential for using small-molecule inhibitors of ANXA2 in cancer therapy. However, the μM level of matrine affinity to ANXA2 is still higher than the normally nM level in drug development, which intrigues modifications on matrine structure or findings of new powerful small-molecule blockers of ANXA2. In addition, as we discussed above, matrine may exert off-target effects beyond its interaction with ANXA2, even at the low concentration (20 µM) used in our experiments. While our validation with PY60 and the hexapeptide LCKLSL supports the specificity of ANXA2 targeting, further studies using ANXA2-specific inhibitors or conditional knockout models would help clarify the off-target effects. The need for structural optimization of matrine derivatives to improve ANXA2-binding affinity and minimize off-target interactions is emphasized in future studies. Besides, compensatory immune escape mechanisms, such as upregulation of alternative immune checkpoints (eg, programmed death-ligand 1, cytotoxic T-lymphocyte associated protein 4) or alterations in other immune cells (eg, dendric cells or macrophages), should be comprehensively investigated in response to ANXA2 blockade. The longitudinal studies combining ANXA2 blockade with immune checkpoint inhibitors are warranted. In conclusion, our results introduced the roles of targeting ANXA2 in liver cancer and elucidated the mechanisms by releasing HSP90, which promoted tumor antigens presentation and increased the killing of tumor cells. The high expression of ANXA2 in liver tumor tissues may be an effective target for therapy, but its expression and the role in other cancer types require further verification. In addition, new compounds with stronger affinity to bind ANXA2 are of interest to be explored and validated. Methods Cells and reagents Hepatoma Huh7 and Hepa1-6 cells were obtained from the Stem Cell Bank at the Chinese Academy of Sciences with STR authentication and were detected without mycoplasma using a Myco-Lumi Luminescent Mycoplasma Detection Kit (Cat. C0298 M, Beyotime Biotech, China).[203]^44 Human CD8^+ T cells were purchased from Fuheng Biology (Cat. FNHBCD8005CN). The naïve CD8^+ T cells were negatively selected via immunomagnetic separation from PBMCs of healthy donors and tested for purity and viability by flow cytometry. ANXA2^+/+ and ANXA2^−/− liver tumor cells were isolated from DEN-induced liver tumor tissues of ANXA2^+/+ and ANXA2^−/− mice. Hepa1-6, ANXA2^+/+, and ANXA2^−/− cells were labeled with luciferase by transfecting lentivirus vectors (GenePharma, Shanghai, China) for in vivo imaging. The natural product, matrine, was purchased from Selleck Chemicals (Cat. S2322). ANXA2 inhibitor PY60 (Cat. HY-141644) and ANXA2 competitive hexapeptide (LCKLSL, Cat. HY-P2333) were purchased from MCE. Recombinant human IFN-γ was purchased from PeproTech (Cat. 300–02). In vivo PD-1 antibodies were purchased from Bio X Cell (Cat. BE0146-50MG). Immunoprecipitation and LC-MS/MS Immunoprecipitation (IP) of ANXA2, HSP90, and MHC I in cell lysates was performed using Protein A/G Magnetic Beads (Cat. L-1004, Biolinkedin, China) following the manufacturer’s protocols.[204]^45 Rabbit IgG (cat. 2729, CST, USA) was used as a negative control. ANXA2 antibodies were purchased from Abcam (Cat. ab235939). HSP90 (Cat. sc-13119) and MHC I (Cat. sc-55582) antibodies were purchased from Santa Cruz Biotechnology (USA). The IP collections were quantified by western blotting and normalized to the indicated internal reference. Peptides combined with HSP90 or MHC I were filtered using Amicon Ultra-0.5 Centrifugal Filter Devices (10 kD cut-off; Millipore, Germany). LC-MS/MS was performed as previously described.[205]^46 Animals Mice were reared in the Animal Center of the Second Military Medical University, with normal circadian rhythms and food or water ad libitum in a specific pathogen-free environment. C57BL/6 mice were purchased from Laboratory Animals (Shanghai, China). NCG mice lacking T, B, and NK cells were purchased from the Model Animal Research Center of Nanjing University (Cat. T001475, NOD-Prkdc^em26Il2rg^em26). The generation of ANXA2 knockout mice was commissioned by Cyagen Biosciences, Exon 3 of the gene ANXA2 was deleted using the Crispr-Cas9 system with guide RNAs (gRNA1: GAATGATGGCACCGAGGATTTGG; gRNA2: GGCTCGACTCCAGCGCAGTTAGG). miR-122 knockout mice were bred from our previous project.[206]^45 47 48 Induction of primary liver cancer in mice At 2 weeks of age, C57BL/6 mice or ANXA2^+/+ and ANXA2^−/− mice were intraperitoneally injected once with 25 mg/kg DEN (Sigma-Aldrich). At the age of 9 months, mice received matrine treatment in drinking water at 0.2 g/L and 0.4 g/L. At the age of 10 months, the mice were sacrificed for experimentation.[207]^49 The miR-122 knockout led to spontaneous liver cancer in mice at 14 months of age, which similarly depicts the characteristics of human liver cancer.[208]^14 Tumor incidence and weights were recorded. Serum ALT and AST levels were determined using a Mindray Clinical Chemistry Analyzer (Mindray, Shenzhen, China). Serum AFP content was measured using an enzyme-linked immunoassay kit (Cat. ab210969, Abcam). Single-cell sequencing Tumor tissues of DEN-induced primary liver cancer in C57BL/6 mice were collected (three samples in control group and three samples in matrine treatment group) for single-cell sequencing. Tissues were dissociated into single cells in dissociation solution (0.35% collagenase IV 5.2 mg/ml papain, 120 units/ml DNase I). The resulting cell suspension was filtered by passing through 70–30 µm stacked cell strainer and centrifuged at 300 g for 5 min at 4°C. The cell pellet was resuspended in 100 µL phosphate buffered saline (PBS) (0.04% BSA) and wiped off red blood cells and dead cells by red blood cell lysis buffer (MACS 130-094-183) and Miltenyi Dead Cell Removal Kit (MACS 130-090-101). Trypan blue was used to confirm the viable cells to be above 85%. Single-cell suspensions were counted using a hemocytometer and concentration was adjusted to 1,000 cells/µL. Single-cell suspensions were loaded to 10x Chromium according to the manufacturer’s instructions of 10x Genomics Chromium Single-Cell 3’kit (V.3). The following complementary DNA amplification and library construction steps were performed according to the standard protocol. Libraries were sequenced on an Illumina NovaSeq 6000 sequencing system (paired-end multiplexing run, 150 bp) by LC-Bio Technology, (Hangzhou, China). Orthotopic implantation of tumor cells and in vivo imaging Luciferase-labeled Hepa1-6, ANXA2^+/+, and ANXA2^−/− cells were prepared at 5×10^7 cells/mL in Matrigel (Cat. 356237; Corning) and kept on ice. Orthotopic implantation surgery was performed as previously reported.[209]^15 Mice were anesthetized by isoflurane, and an aseptic micro-incision was made along the midline of the abdomen under the xiphoid process to expose the liver. A 20 µL cell suspension was slowly injected into the largest liver lobe and held for 10 s to solidify the Matrigel, followed by suturing of the abdominal muscles and skin layer by layer. The labeled cells were observed by in vivo imaging every 10 days for 50 days. Mice were anesthetized by isoflurane and intraperitoneally injected with D-Luciferin (150 µL, 20 mg/mL). After 10 min, photographs were taken and analyzed using an in vivo imaging system (IVIS Lumina LT Series III, PerkinElmer, USA). The luciferase substrate D-luciferin was purchased from Sigma-Aldrich (USA). Mouse CD8^+ T cells isolation Untouched and highly purified CD8^+ T cells were isolated from normal mouse spleens using an EasySep Mouse CD8^+ T Cell Isolation Kit (Cat. 19853, STEMCELL Technologies).[210]^15 The naïve CD8^+ T cells needed tumor antigens to be activated, which were suitable for our purpose to verify the changed tumor antigens presentation by targeting ANXA2. Briefly, the mouse spleens were disrupted in PBS containing 2% fetal bovine serum. The cell suspension was passed through a 70 µm strainer and adjusted to 1×10^8/ mL. Undesired cells were excluded using biotinylated antibodies and streptavidin-coated magnetic particles from the kit. The naïve CD8^+ T cells were left in the medium and collected for further analysis. Immunocytes infiltration analysis The total white blood cells from the liver tumors were separated as previously reported.[211]^15 Briefly, tumor tissues were disrupted and passed through a 70 µm strainer to obtain a single-cell suspension. White blood cells were separated by density gradient centrifugation using a Percoll separating medium. The red blood cells were cleared using ammonium chloride and non-specific binding was blocked by Fc receptor antibodies. The white blood cells were then incubated with different surface antigens’ antibodies to determine the proportions of immunocyte subsets by detecting fluorescence signals using flow cytometry (Beckman Coulter, USA). Cells positive for CD45 were regarded as white blood cells, and CD3^+, CD19^+, and NK1.1^+ were used to represent the proportions of T, B, and NK cells, respectively. CD45^+CD3^+CD4^+ and CD45^+CD3^+CD8^+ cells were used to represent CD4^+ and CD8^+ T cells, respectively. All surface antigens and antibodies were purchased from eBioscience (Thermo Fisher Scientific). RNA sequencing analysis and validation The TPM (transcripts per million) of ANXA2 in liver tumors and the overall survival of patients with liver cancer were retrieved from the TCGA database using Gene Expression Profiling Interactive Analysis.[212]^50 A heatmap of immune cell infiltration by immune markers and the correlation analysis with ANXA2 transcripts was generated by using R packages. The mRNA and protein expression of ANXA2 were validated in the liver tumor tissues and non-tumor counterparts from our previous work.[213]^15 Immunohistochemical and immunofluorescence staining The tumor tissues were preperfused by 0.9% NaCl and fixed in 4% paraformaldehyde. After being embedded in paraffin and sliced, the slices were dewaxed, repaired for antigen in antigen repair buffer (Cat. C1032, Solarbio Science and Technology, China), blocked by Normal Donkey Serum (Cat. ab7475, Abcam), and incubated by primary and secondary antibodies.[214]^51 After DAB staining, the nuclei were stained by hematoxylin (Sangon Biotech, China). Antibodies used were as follows: CD4 (1:100, Cat. ab183685, Abcam); CD8 (1:100, Cat. ab217344, Abcam), NK1.1 (1:100, Cat. 70–5941, Tonbo Biosciences). For immunofluorescence staining, cells were washed two times by PBS and fixed by 4% paraformaldehyde for 15 min. Then cells were washed three times by PBS and treated with enhanced permeabilization buffer (Cat. P0097, Beyotime Biotech, China) for 10 min. After being washed by PBS for three times, cells were blocked by BSA and incubated with ANXA2 antibody (1:200, Cat. 11 256–1-AP, Proteintech) and HSP90 antibody (1:50, Cat. sc-13119, Santa Cruz Biotechnology) overnight. Then cells were washed three times by PBS and incubated with Goat Anti-Rabbit IgG H&L (Alexa Fluor 488) (1:500, Cat. ab150077, Abcam) and Goat Anti-Mouse IgG H&L (Alexa Fluor 594) (1:500, Cat. ab150116, Abcam) for 1 hour. After being washed for five times, cells were stained by DAPI, following another washing for three times. Immunofluorescence pictures were taken by STELLARIS 5, Leica. Colocalization was analyzed by Coloc 2 module in ImageJ software. Real-time PCR Detailed real-time PCR protocols can be found in our previous report.[215]^49 In brief, RNA was extracted from cells or tissues by using TRIzol reagent (Cat. 15 596–026, Invitrogen, USA). 1 µg RNA was proceeded to be reverse-transcribed by using PrimeScript RT Master Mix (Cat. RR036A, Takara, Japan). Real-time quantitative PCR was performed by the SYBR Green Kit (Cat. B532955, Sangon Biotech, China) using the QuantStudio 1 system (Applied Biosystems). The mRNA expressions of target genes were normalized by internal reference 18 s. All primers were synthesized and purified by high-performance liquid chromatography at Sangon Biotech (Shanghai, China). The primers sequences were shown in [216]online supplemental table S2. Western blot The total protein content was extracted using a total protein extraction kit (Keygene Biotech, Nanjing, China). Western blotting assays were performed and analyzed as previously described.[217]^52 The antibodies used in this study were: β-actin (Cat. ab8227, Abcam), HSP90 (Cat. sc-13119, Santa Cruz Biotechnology), MHC I (Cat. sc-55582, Santa Cruz Biotechnology), ANXA2 (Cat. 11 256–1-AP, Proteintech), and TAP1 (Cat. 11 114–1-AP, Proteintech). An IRDye secondary antibody (LI-COR, USA) was used, and immunoblots were scanned using an Odyssey dual-color infrared fluorescence imaging system. The grayscale of each band was obtained using the Odyssey software. Statistics Data are presented as the mean±SEM. A t-test was used for two-group comparisons, and one-way analysis of variance was used for multigroup comparisons. Tukey’s multiple comparison test was used for post hoc analysis. Differences were considered statistically significant at p<0.05. Supplementary material online supplemental file 1 [218]jitc-13-6-s001.docx^ (10.6MB, docx) DOI: 10.1136/jitc-2025-011716 Footnotes Funding: This work was supported by the National Natural Science Foundation of China (82003711, 81903306), Shanghai Sailing Program (19YF1459400, 24YF2758300), and Shanghai Natural Science Foundation (19ZR1469700, 23ZR1477600). This work was also supported by the Open Research Fund of Basic Medicine College (JCKFKT-MS-001), the Shanghai Clinical Research Center of Traditional Chinese Medicine Oncology (21MC1930500), the Shanghai Youth Program of Public Health Research (2024GKQ26), and the Natural Science Research Project of Shanghai Jiading District (JDKW-2024- 0013). YT is supported by the Young Elite Scientists Sponsorship Program by CAST (2022QNRC001). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: Animal studies were performed under the approval of the Institutional Animal Care and Use Committee of the Second Military Medical University (NMU-19YF1459400), following the guidelines from the 'Guide for the Care and Use of Laboratory Animals' Data availability free text: The data supporting the findings of this study are available from the corresponding author upon reasonable request. Data availability statement Data are available upon reasonable request. References