Alpha-Fetoprotein Stimulates Cleavage of Membranal MICA/B on Liver Cancer Cell Lead to Escape Immune Surveillance of Natural Killer Cells.

1
Introduction
Hepatocellular carcinoma (HCC), the most common primary liver cancer, was characterised by high rates of recurrence and metastasis. Global cancer incidence, mortality, and prevalence data for 2020 indicate that HCC was the sixth most common cancer worldwide and the third leading cause of cancer‐related deaths [1]. Risk factors for HCC include chronic viral hepatitis infection, alcoholic fatty liver, metabolic dysfunction‐associated fatty liver, and aflatoxin exposure [1, 2]. Standard treatments for HCC include surgical resection, transarterial tumour embolization, radiotherapy, chemotherapy, and targeted therapy. However, these treatments often face challenges due to late disease diagnosis or rapid metastasis, leading to an advanced stage at the time of diagnosis [3]. Currently, studies have reported that sorafenib and lenvatinib are first‐line targeted therapies for advanced HCC, but they offer limited benefits to patients, extending survival by only approximately 3 months [4, 5]. In recent years, with the development of precision medicine, immunotherapy for HCC has garnered widespread attention [6]. When an organism develops a tumour, the immune system can recognise and eliminate the tumour cells in the microenvironment. However, under pathological conditions, tumour cells attempt to evade immune surveillance and clearance by releasing molecules that are hindered by immune cell recognition, such as major histocompatibility complex class I chain‐related protein A/B (MICA/B). The release of soluble MICA/B (sMICA/B) into the cell supernatant can prevent immune cells from recognising and killing tumour cells, thereby promoting their immune escape [7]. MICA/B is considered a target for cancer vaccines [8] and thus plays a crucial role in immunotherapy.
The immune system plays a significant role in HCC [9]. Despite the presence of numerous immune cells in HCC tissues, immune dysfunction can still exist in immunosuppressive regions of the tumour microenvironment, inhibiting the activation of immune effectors [10]. Natural killer (NK) cells are an essential component of the innate immune system. NK cells can kill tumour cells by recognising their surface receptors [11]. Therefore, NK cells are regarded as the first line of defence against cancer and have received considerable attention in secondary cancer immunotherapy [12]. NK‐92, the first NK cell approved by the US and FDA for clinical trials, was a homogeneous immortalised NK lymphoma cell line that can be cultured in vitro. The cytotoxicity of NK cells primarily depends on their activated receptor, natural killer group 2 member D (NKG2D), which binds to NKG2D ligands on target cells and mediates cytotoxicity [13]. MICA/B was a highly glycosylated membrane protein belonging to the NKG2D family of ligands [13]. MICA/B acts as a ligand for NKG2D, specifically activating NK cells and inducing an immune response.
Alpha‐fetoprotein (AFP) was a crucial biomarker for primary HCC and was closely associated with its development [14]. Clinically, high serum AFP levels are typically associated with a high risk of HCC [15]. In recent years, it has been reported that AFP can promote the immune escape of HCC cells [16], thereby facilitating HCC growth. However, it remains unclear whether AFP promotes HCC cell immune escape by regulating MICA/B. This study aimed to explore the mechanism by which AFP regulates the expression of MICA/B and its distribution in HCC cell membranes, and it sought to elucidate the mechanism by which AFP shields HCC cells from the attack of NK‐92.

2
Materials and Methods
2.1
Databases and Analysis Tools
HCC RNA‐Seq data were downloaded from the TCGA database (424 cases in August 2022) (https://portal.gdc.cancer.gov) and GEO database (GSE121248). Data were converted to ID using the vlookup function, differentially expressed genes were screened, and volcano plots were plotted using R (4.2.1) software with screening conditions of |LogFC| > 1.5 and adjusted p < 0.05. Co‐expressed genes were screened using the STRING database, KEGG enrichment analysis was performed using R software, visualised by bubble plots, immune infiltration analysis was conducted using the single‐sample gene set enrichment analysis (ssGSEA) algorithm, and Spearman correlation was used for the correlation analysis between two genes.

2.2
Immunohistochemistry
Tumour tissues from 30 HCC patients hospitalised at the First Affiliated Hospital of Hainan Medical University from January 2021 to December 2022, as well as adjacent normal liver tissues, were collected. All tissue specimens were immersed in 4% paraformaldehyde, fabricated into tissue microarrays, and then soaked in xylene for 10 min for dehydration. Antigen repair was performed using citric acid antigen repair buffer (pH 6.0), endogenous peroxidase was blocked using 3% hydrogen peroxide solution, and tissue microarrays were blocked for 30 min using 3% bovine serum albumin (BSA) sealing solution. Tissue microarrays were then incubated with antibodies against AFP (Cat #ab169552, Abcam, USA) and MICA/B (Cat #ab224702, Abcam, USA) overnight at 4°C. The following day, microarray tissues were retrieved using PBS and incubated with secondary antibodies (Cat #GB23301, Servicebio, China) at room temperature for 1 h. After washing with PBS, a DAB kit (Cat #G1212, Servicebio, China) was used for colour development. The nuclei were stained with haematoxylin staining solution, and finally, the tissue microarrays were dehydrated and sealed for image capture.

2.3
Cells and Reagents
The human HCC cell lines HLE (Cat #CBP60200) and HuH‐7 (Cat #CL‐0120) were purchased from NanJing BIOSCIENCES Co. Ltd. and Wuhan Pricella Biotechnology Co. Ltd., respectively, and natural killer‐92 (NK‐92) cells (Cat #GDC0052) were purchased from the China Center for Type Culture Collection. Human HCC cell line, HuH‐7 (preserved in our laboratory) and HLE cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) (Cat #C3113‐0500, Viva cell, China) supplemented with 10% fetal bovine serum (Cat #164210–50, Procell, China) and 1% penicillin–streptomycin (Cat #15140‐122, Gibco, United States); NK‐92 cells were cultured in a special medium (Cat #GDM1001). All the cells were placed in an incubator at 37°C and 5% CO2.

2.4
Lentiviral Transfection
Human HCC cell lines in the logarithmic growth phase were uniformly plated in 6‐well plates at approximately 5 × 104 cells per well. Lentiviral vectors overexpressing AFP were transfected into HLE (HLE‐AFP) cells, transfected with negative control vectors HLE (HLE‐NC), and lentiviral vectors interfering with AFP were transfected into HuH‐7 (HuH‐7‐shAFP) cells; negative control vectors were transfected into HuH‐7 (HuH‐7‐NC). The required lentiviral volume was calculated as follows: viral volume (μL) = (MOI × number of cells)/viral titre (TU/mL) × 1000, where the MOI represents the multiplicity of infection. The HiTransG A Transfection Enhancer was used in HLE cells, and the HiTransG P transfection enhancer was used in HuH‐7 cells. Lentivirus was added to the human HCC cell line for 16 h, replaced with fresh medium, and incubated for 48 h. A green fluorescence greater than 80% under a fluorescence microscope indicated successful transfection.

2.5
Western Blotting
Protein precipitates from human HCC cells were collected for protein extraction. Protein concentration was determined using a BCA kit (Cat #P0009, Beyotime, China). Equal volumes and masses of proteins were subjected to 12% sodium dodecyl sulfate‐polyacrylamide gel electrophoresis (SDS‐PAGE) and the proteins on the gel were transferred onto a PVDF membrane. After washing for 10 min, the PVDF membrane was blocked with a protein‐free rapid blocking buffer (Cat #PS108, Epizyme, China) and incubated with the primary antibodies AFP (Cat #ab169552, Abcam, USA) at a dilution of 1:1000, MICA/B (Cat #ab224702, Abcam, USA) at a dilution of 1:1000, MMP9 (Cat #13667, Cell Signalling Technology, USA) at a dilution of 1:1500, PI3K (Cat #ab302958, Abcam, USA) at a dilution of 1:1000, AKT (Cat #ab8805, Abcam, USA) at a dilution of 1:500, p‐AKT (Cat #4060, Cell Signalling Technology, USA) at a dilution of 1:500, GAPDH (Cat #10494‐1‐AP, Proteintech, China) at a dilution of 1:10,000, and ATP1A1 (Cat #14418‐1‐AP, Proteintech, China) at a dilution of 1:10,000 at 4°C overnight. After washing twice with PBST for 10 min each, the membrane was incubated with secondary antibody (Cat #SA00001‐2, Proteintech, China) at 37°C for 1 h. It was then washed twice again, soaked in the ECL luminescent solution, and placed on the luminescent equipment for detection.

2.6
Flow Cytometry
Human HCC cells were collected and centrifuged at 1500 rpm at 4°C for 10 min; this process was repeated twice. Next, the cells were resuspended in 50 μL of phosphate‐buffered saline (PBS), and 5 μL of MICA/B (Cat #320907, Biolegend, USA), CD3 (Cat #981002, Biolegend, USA), and CD56 (Cat #985906, Biolegend, USA) antibodies were added to each sample. The samples were incubated for 20 min at room temperature in the dark, washed twice with 1 mL of PBS, and centrifuged again at 1500 rpm and 4°C for 10 min. After washing, the cells were resuspended in 1 mL of PBS, and the cell suspension was filtered through a strainer and placed on a flow cytometer to detect MICA/B expression.

2.7
Immunofluorescence
HCC cells (5 × 104) were evenly placed in a special dish for immunofluorescence analysis. After the cells adhered to the wall, the medium was discarded and the cells were washed twice with PBS for 5 min each time. The cells were then fixed in a 4% paraformaldehyde solution for 15 min. After fixation, the cells were washed twice with PBS for 5 min each, 200 μL goat serum was added, and the cells were blocked for 30 min. Subsequently, the cells were washed twice with PBS for 5 min each, 200 μL MICA/B antibody (Cat #PA5‐109323, Thermo Fisher Scientific, USA) was added, and the cells were incubated at 4°C for 16 h. The cells were then washed twice with PBS for 5 min each time, 200 μL of secondary antibody (Cat #A0468, Beyotime, China) was added, and the cells were incubated at room temperature for 1 h. The cells were then washed twice with PBS for 5 min each, 80 μL of DAPI staining solution (Cat #C1006, Beyotime, China) was added, and the cells were incubated at room temperature for 10 min. Images were acquired using a confocal laser microscope.

2.8
Enzyme‐Linked Immunosorbent Assay (ELISA)
The soluble MICA (sMICA) and MICB (sMICB) in the cell supernatants were determined using human MICA‐precoated ELISA (Cat #F111358‐B, FANKEW, China) and human MICB‐precoated ELISA (Cat #F111362‐B, FANKEW, China) kits, according to the manufacturer's instructions. Perforin, Granzyme B, and IFN‐γ in the co‐culture cell supernatants were detected using a human perforin precoated ELISA kit (Cat #F1149‐A, FANKEW, China), human Granzyme B precoated ELISA kit (Cat #F0430‐A, FANKEW, China), and human IFN‐γ precoated ELISA kit (Cat #F0033‐A, FANKEW, China) according to the manufacturer's instructions.

2.9
Quantitative Real‐Time PCR (qRT‐PCR)
Total RNA was extracted from HCC cells using a TaKaRa MiniBEST Universal RNA Extraction Kit (Cat #9767, TaKaRa, Japan). RNA was reverse‐transcribed into DNA using the PrimeScript RT Reagent Kit (Cat #RR047A, TaKaRa, Japan). PCR (Cat #RR820A, TaKaRa, Japan) was performed using a Roche 480 instrument. The PCR was divided into three steps: the first step was an initial denaturation step (95°C for 30 s), the second step was a PCR reaction (5 s at 95°C and 20 s at 60°C) for 40 cycles, and the third step was a melting/dissociation curve (95°C for 0 s, 65°C for 15 s, and 95°C for 0 s). GAPDH was used as an internal reference gene for MMP9, and the data were analysed using the 2−ΔΔCt method. The primer sequences are shown in Table 1.

2.10
Cell Counting Kit‐8(CCK‐8) Assay
First, cells in the logarithmic growth phase were digested with trypsin and evenly spread in 96‐well plates, with approximately 5000 cells per well and six compound pores. The cells were incubated overnight in a cell incubator. On the second day, the medium was discarded, fresh medium was added, and various concentrations of GM6001 (Cat #HY‐15768; MCE, USA) and TAPI‐1 (Cat #HY‐16657; MCE) were added to each well. Eight concentration gradients were set, which were 0 μg/mL, 5 μg/mL, 10 μg/mL, 15 μg/mL, 20 μg/mL, 25 μg/mL, 30 μg/mL and 35 μg/mL, respectively. The cells were cultured in an incubator for 24 h, after which the medium was discarded, fresh serum‐free medium (100 μL per well) was added, and CCK8 detection reagent (Cat #CK04, DOJINDO, Japan) was added (10 μL per well). The 96‐well plate was placed in a microplate reader to determine the absorption wavelength at 450 nm.

2.11
Animal Tumour Model
Three‐week‐old female NOD/SCID mice were purchased from Guangzhou Medical Center and maintained in a specific‐pathogen‐free (SPF) animal facility. Animal experiments were approved by the Institutional Animal Care and Use Committee of Hainan Medical University. HCC cells in the logarithmic growth phase were collected and injected into NOD/SCID mice; 5 × 106 cells were injected into each mouse to construct an in vivo HCC tumour model. After tumour formation, the mice were euthanized, and tumour tissues were removed for Western blotting and IHC experiments.

2.12
Cytotoxicity Assay
The cytotoxicity of NK‐92 cells to HCC cells was detected using a lactate dehydrogenase (LDH) release cytotoxicity assay kit at E:T ratios of 2:1 and 4:1 (Cat #C0017, Beyotime, China), according to the manufacturer's instructions. Specifically, NK‐92 cells were co‐cultured with HCC cells, and the cytotoxicity of NK‐92 cells to HCC cells was assessed by measuring the amount of LDH released in the supernatant of the co‐cultured cells using an LDH release cytotoxicity assay kit after 24 h. NK‐92 cells were then treated with an NKG2D neutralising antibody (Cat #MAB139, Bio‐Techne, USA) for 30 min and co‐cultured with HCC cells to measure cytotoxicity. NK‐92 cells were co‐cultured with HCC cells after treatment with MMP9 inhibitors for 48 h, and cytotoxicity was measured. Cytotoxicity was calculated using the following formula: Cytotoxicity (%) = [(ODExperimental value − ODSpontaneous value) / (ODMaximum value − ODSpontaneous value)] × 100.

2.13
Statistical Analysis
Statistical software R (version 4.2.1) was used. The data of the two groups that satisfied normal distribution and homogeneity were analysed using the t‐test, those that satisfied normal distribution but did not satisfy homogeneity were analysed using the Welch t‐test, and those that did not satisfy the normal distribution were analysed using the Wilcoxon rank sum test. Data from three or more groups were analysed using one‐way ANOVA. Indicators with missing data of ≥ 60% were excluded from the statistical analysis. p < 0.05 was considered a statistically significant difference.

3
Results
3.1
AFP And MICA/B Were Highly Expressed in HCC Tissues and Their Regulatory Relationship May Be Related to MMP9
The volcano plot results of TCGA data showed that 829 genes were upregulated and 203 genes were downregulated in the HCC RNA‐Seq data (Figure 1A). The expression levels of AFP and MICA/B in HCC tissues and adjacent normal liver tissues showed that the expression levels of AFP, MICA, and MICB were higher in HCC tissues than in adjacent normal liver tissues in TCGA data (p < 0.001) (Figure 1B). The volcano plot results of the GEO data showed that 123 genes were upregulated and 339 genes were downregulated in HCC RNA‐Seq data (Figure 1C). The expression levels of AFP and MICA/B in HCC tissues and adjacent normal liver tissues showed that the expression of AFP (p < 0.05), MICA (p < 0.01), and MICB (p < 0.001) was significantly higher in HCC tissues than in adjacent normal liver tissues (Figure 1D). Furthermore, immunohistochemical results showed that both AFP and MICA/B had higher cell densities in HCC tissues than in the adjacent normal liver tissues (p < 0.05) (Figure 1E). The results indicated that the expression of AFP and MICA/B was higher in HCC tissues than in the adjacent normal liver tissues. Finally, protein–protein interaction (PPI) results of AFP and MICA/B showed that AFP, MMP9, A disintegrin and metalloproteinase 17 (ADAM17), matrix metalloproteinase 14 (matrix metalloproteinase 14, MMP14), disintegrin metalloproteinase 10 (A disintegrin and metalloproteinase 10, ADAM10), MICB, and MICA were present (Figure 1F).
KEGG enrichment analysis showed that the enriched pathways mainly included the IL‐17, PI3K‐AKT, TNF, Toll‐like receptor, and NF‐κB signalling pathways. The related diseases were mainly novel coronavirus and hepatitis B, and their biological functions were mainly associated with NK cell‐mediated toxicity (Figure 1G). Immunofluorescence analysis showed that AFP was mainly associated with helper T cell 2 (r = 0.271, p < 0.001), follicular helper T cells (r = 0.241, p < 0.001), T cells (r = 0.207, p < 0.001), CD56bright NK cells (r = 0.199, p < 0.001), helper T cells 17 (r = −0.187, p < 0.001), NK cells (r = −0.209, p < 0.001), central memory T cells (r = −0.210, p < 0.001), and neutrophils (r = −0.288, p < 0.001) (Figure 1H). A scatter plot of the correlation between AFP and NK cell infiltration is shown in Figure 1I. The results of the correlation scatter plot of genes showed that AFP was positively correlated with MMP9 (r = 0.283, p < 0.001) (Figure 1J) but not with ADAM17 (r = 0.089, p = 0.086), MMP14 (r = 0.003, p = 0.946), and ADAM10 (r = 0.087, p = 0.094) (Figure 1J).

3.2
AFP Inhibits MICA/B Expression on the Membrane of HCC Cells
Western blotting results showed that overexpression of AFP had no effect on the expression level of total MICA/B proteins (p > 0.05); AFP could significantly inhibit the content of MICA/B in the cellular membrane (p < 0.001), and interference with AFP had no effect on the expression level of total MICA/B proteins (p > 0.05) but significantly promoted the content of MICA/B in the cellular membrane (p < 0.001) (Figure 2A). Flow cytometry results also showed that overexpression of AFP significantly inhibited the content of MICA/B in the cellular membrane (p < 0.001), and interference with AFP significantly increased the content of MICA/B in the cellular membrane (p < 0.01) (Figure 2B). The results of cellular immunofluorescence showed that MICA/B was mainly located in the cell membrane and cytoplasm, and that the fluorescence intensity of MICA/B was decreased by overexpression of AFP, but the fluorescence intensity of MICA/B was increased by interference with AFP (Figure 2C).

3.3
AFP Upregulates MMP9 Expression to Promotes MICA/B Shedding on the Membrane of HCC Cells Through Activating PI3K/AKT Signalling Pathway
Western blotting showed that MMP9 expression levels in the HLE‐AFP group were significantly higher than those in the HLE and HLE‐NC groups (p < 0.001); however, MMP9 expression levels in the HuH‐7‐shAFP group were significantly lower than those in the HuH‐7 and HuH‐7‐NC groups (p < 0.001) (Figure 3A,B). qRT‐PCR experiments showed that MMP9 mRNA expression levels in the HLE‐AFP group were significantly higher than those in the HLE and HLE‐NC groups (p < 0.001). However, MMP9 mRNA expression levels in the HuH‐7‐shAFP group were significantly lower than those in the HuH‐7 and HuH‐7‐NC groups (p < 0.001) (Figure 3C).
Western blotting experiments further revealed that sMICA/B was observed in the cell supernatants of HLE, HLE‐NC, and HLE‐AFP with two different molecular weights, 50 kDa and 17 kDa (Figure 3D). Soluble MICA/B (sMICA/B) was similarly observed in the cell supernatants of HuH‐7, HuH‐7‐NC, and HuH‐7‐shAFP with two different molecular weights of 50 kDa and 17 kDa (Figure 3E). The ELISA results showed that sMICA/B in the cell supernatant was significantly increased when AFP was overexpressed in HLE cells. However, sMICA/B in the cell supernatant was significantly decreased by interference with AFP in Bel7402 cells (Figure 3F). To analyse the influence of MMP9 on the shedding of MICA/B in HCC cells, in the present study, MMP9 inhibitors (GM6001 or TAPI‐1) were applied to treat HLE cells transfected with AFP‐expressing vectors (HLE‐AFP). CCK‐8 assay showed that the IC50 of GM6001 was 14.32 μg/mL, 12.95 μg/mL, and 20.65 μg/mL in HLE, HLE‐NC, and HLE‐AFP cells, respectively; and that of TAPI‐1 was 21.62 μg/mL, 24.59 μg/mL, and 33.14 μg/mL, respectively (Figure 3G). The flow cytometry results showed that the expression level of MICA/B was significantly lower in the HLE‐AFP group compared to the HLE‐NC group (p < 0.001); after GM6001 and TAPI‐1 were used to treat the HLE‐AFP group, the expression level of MICA/B was significantly rebounded (p < 0.01) (Figure 3H,I). Further transient knockdown of MMP9 expression in HLE‐AFP cells was performed using siRNA, followed by detection of MICA/B expression was performed using flow cytometry; results show the expression level of MMP9 was significantly increased in the HLE‐AFP group (p < 0.01), while it was significantly decreased in the HLE‐AFP‐siMMP9 group (p < 0.001) (Figure S1A). Further flow cytometry results showed that the expression level of MICA/B was significantly decreased in the HLE‐AFP group (p < 0.001). After interfering with MMP9, the expression level of MICA/B significantly increased again (p < 0.001) (Figure S1B). This result was consistent with the findings observed when HLE‐AFP cells were treated with GM6001 and TAPI‐1. ELISA results showed that the sMICA/B levels in the cell supernatant of the HLE‐AFP group were significantly increased (p < 0.01). After treatment with GM6001 and TAPI‐1 in the HLE‐AFP group, the promoting effect of AFP on sMICA/B was reversed, and the sMICA/B in the cell supernatant returned to its original level, which was consistent with that in the HLE and HLE‐NC groups (Figure 3J). Western blotting experiments showed that there was no significant change in the expression of PI3K and AKT after the overexpression of AFP and PI3K/AKT pathway inhibitor (Ly294002) treatment in HLE cells (p > 0.05). However, after AFP overexpression in HLE cells, the expression of MMP9 was significantly elevated (p < 0.01); while treatment of the cells with Ly294002, the expression of p‐AKT (p < 0.01) and MMP9 (p < 0.001) were significantly decreased. The PI3K/AKT pathway inhibitor reversed the effect of AFP on upregulation of the expression of MMP9 in HLE cells (Figure 3K,L). These results implied that AFP may stimulate the expression of MMP9 to shed MICA/B through activating PI3K/AKT signalling pathway.

3.4
Silencing AFP Was Able to Downregulate MMP9 Expression and Promote the Membrane Protein Level of MICA/B in the NOD/SCID Mouse Tumour Model
Western blotting results showed that the expression level of AFP in the HuH‐7‐shAFP group was significantly lower than that in the HuH‐7 group (p < 0.05), the expression level of MMP9 in the HuH‐7‐shAFP group was also significantly lower than that in the HuH‐7 group (p < 0.01), and the MICA/B membrane protein levels in the HuH‐7‐shAFP group were significantly higher than those in the HuH‐7 group (p < 0.05) (Figure 4A). IHC experiments showed that the expression level of AFP in the HuH‐7‐shAFP group was significantly lower than that in the HuH‐7 group (p < 0.01), and the expression level of MMP9 was significantly lower than that in the HuH‐7 group (p < 0.05). However, the expression level of MICA/B membrane protein in the HuH‐7‐shAFP group was significantly higher than that in the HuH‐7 group (p < 0.01) (Figure 4B). These results indicated that AFP could stimulate the expression of MMP9 to shed MICA/B in the HCC cell membrane.

3.5
AFP Inhibits the Cytotoxicity of NK‐92 Cells to HCC Cells
Flow cytometry results showed positive CD56 and negative CD3 expression in the NK‐92 cells (Figure 5A). The lactate dehydrogenase (LDH) release assay showed that the cytotoxicity of NK‐92 cells in the HLE‐AFP group was significantly lower than that in the HLE and HLE‐NC groups (p < 0.01), and the cytotoxicity in the HuH‐7‐shAFP group was significantly higher than that in the HuH‐7 and HuH‐7‐NC groups (p < 0.01) (Figure 5B). The promotion effect of interfering with AFP was reversed after treatment of NK‐92 cells with NKG2D neutralising antibody (p < 0.001) (Figure 5C), and the inhibitory effect of AFP overexpression was reversed after treatment of HCC cells with MMP9 inhibitor (GM6001+TAPI‐1) (p < 0.05) (Figure 5D). These results indicated that AFP inhibited NK‐92 cell to attack HCC cells.

3.6
AFP Inhibits Cytokine Release From NK Cells and NK Cells Attack HCC Cells
ELISA was used to detect cytokines, such as Perforin, Gzms‐B, and IFN‐γ. The results showed that Perforin, Gzms‐B, and IFN‐γ were significantly reduced in co‐cultured NK‐92 with HLE‐AFP cells compared to co‐cultured NK‐92 with HLE cells and co‐cultured NK‐92 with HLE‐NC cells (p < 0.05); Perforin, Gzms‐B and IFN‐γ were significantly increased in co‐cultured NK‐92 with HuH‐7‐shAFP cells compared to co‐cultured NK‐92 with HuH‐7 cells and co‐cultured NK‐92 with HuH‐7‐NC cells (p < 0.05) (Figure 6A). In the co‐cultured NK‐92 with HLE‐NC group, only a large number of cell fragments without in HLE‐NC cells could be observed under an inverted microscope; however, in the co‐cultured NK‐92 with HLE‐AFP group, the morphology of HLE‐AFP cells was intact, and some cell fragments adhered to the surface of the cells. Statistical analysis showed that the number of surviving HCC cells in the co‐cultured NK‐92 with HLE‐AFP group was significantly higher than that in the co‐cultured NK‐92 with HLE‐NC group (p < 0.001) (Figure 6B).
The co‐cultured cells were dynamically monitored and imaged using an intelligent live‐cell high‐throughput imaging analyser, and dynamic monitoring diagrams were obtained when NK‐92 cells were co‐cultured with HuH‐7‐NC cells (Figure 6C). Dynamic monitoring diagrams were obtained when NK‐92 cells were co‐cultured with HuH‐7‐shAFP cells (Figure 6D). The number of killed HCC cells in five randomly counted fields of view in each group was statistically analysed, and the results showed that the number of HuH‐7‐shAFP cells killed by NK‐92 cells was significantly more than that of HuH‐7‐NC cells (p < 0.001) (Figure 6E). The dynamic monitoring NK‐92 cells to attack HuH‐7‐NC or HuH‐7‐shAFP cells were shown in Supporting Information (Video S1 and S2). These results indicated that AFP could inhibit NK‐92 cells from attacking HCC cells.

4
Discussion
HCC has the biological characteristics of high recurrence and metastasis, and its five‐year survival rate of only 18.1% [17]. Traditional therapies and drugs have failed to significantly extend the overall survival of patients with HCC. Therefore, there is an urgent need for effective therapies to reduce the recurrence and metastasis rates of HCC and improve patient survival. In 2022, Badrinath et al. [18]. Designed a universal vaccine that offered new hope for liver cancer immunotherapy. Vaccines have been reported to possess anti‐tumour immunity throughout the body [19, 20, 21, 22]. It can prevent the cutting of MICA/B protein on tumour cell membranes to increase the level of MICA/B protein on the surface of tumour cells [18, 23, 24]. Thus, NK cells in vivo can recognise tumour cells and enhance their cytotoxicity to kill HCC cells.
It was well known that AFP was the most important diagnostic marker for liver cancer. To explore the relationship between AFP and MICA/B, we simultaneously analysed the expression of AFP and MICA/B in HCC and adjacent normal liver tissues using TCGA and GEO data. The results showed that the expression levels of AFP and MICA/B in HCC tissues were higher than those in adjacent normal liver tissues. To verify these results, tumour tissue and adjacent normal liver tissue samples were collected from 30 patients with HCC. The expression levels of AFP and MICA/B in HCC and adjacent normal liver tissues were detected using IHC. The results showed that the expression of AFP and MICA/B in HCC tissues was significantly higher than that in adjacent normal liver tissues. The experimental results are consistent with the data analysis results. We then constructed a protein interaction network using the STRING database and found that AFP and MICA/B may be co‐expressed at the protein level and linked by four cutting enzymes (MMP9, MMP14, ADAM10, and ADAM17). The correlation between AFP, MMP9/14, and ADAM10/17 was analysed using the TCGA database. The results showed that AFP was only positively correlated with MMP9 but not with the other three cutting enzymes. KEGG analysis indicated that the correlation of AFP with MMP9 and MICA/B mainly involved the PI3K signalling pathway. Therefore, we speculated that AFP might regulate the expression of MMP9 through the PI3K/AKT signalling pathway, thereby further regulating the expression and shedding of MICA/B.
To verify the above inference, we constructed HCC cell lines that stably overexpressed AFP and interfered with the expression of AFP, and then detected the expression level of MICA/B. The results showed no significant change in total MICA/B protein whether AFP was overexpressed or absent of AFP. However, MICA/B levels in the cellular membrane decreased significantly after AFP overexpression and increased significantly after AFP knockdown. sMICA/B in the cell supernatant was detected using ELISA. The results showed that sMICA/B levels significantly increased after AFP overexpression and significantly decreased after AFP interference. These results imply that AFP inhibits the content of MICA/B on the HCC cell membrane and increases sMICA/B levels in the cell supernatant. The expression level of MMP9 was detected using Western blotting and qRT‐PCR. The results showed that the expression of MMP9 was significantly increased after AFP overexpression and significantly decreased after interference with AFP. These results prove that AFP may influence the MICA/B expression pattern by regulating MMP9 expression.
To further verify the relationship between AFP and MMP9, MICA/B HCC cells were treated with MMP9 inhibitors, and changes in sMICA/B were detected by ELISA. The results showed that the upregulation of sMICA/B expression by AFP was reversed after MMP9 inhibitor treatment. MICA/B expression in the HCC cell membrane was detected by flow cytometry, and the results showed that after MMP9 inhibitor treatment, the inhibitory effect of AFP on MICA/B on the HCC cell membrane protein was weakened; however, the expression level of MICA/B membrane protein was not restored to that of the untreated group. Furthermore, after knocking out MMP9 using small interfering RNA, the MICA/B on the surface of HCC cells also significantly increased, and the increase was greater than that achieved by the MMP9 inhibitor. Further, we observed immunoreactive bands of approximately 17 kDa and 50 kDa corresponding to sMICA/B in the cell supernatant. This was consistent with the well‐documented phenomenon of protease‐mediated shedding of the extracellular domain of MICA/B [25, 26]. The membrane‐bound MICA/B has a molecular weight of approximately 50–60 kDa. The primary soluble fragment released after cleavage, depending on its glycosylation status and cleavage sites, has been reported in the literature to have an apparent molecular weight mostly ranging between approximately 30–40 kDa [25, 26, 27]. The 50 kDa band we detected may correspond to a more highly glycosylated variant; whereas the 17 kDa band likely represents a stable product of further proteolytic degradation of this soluble fragment. Although the exact band sizes vary across experimental systems, the presence of these double bands collectively confirms that AFP upregulation of MMP9 promotes active cleavage and shedding of MICA/B, thereby increasing the generation of sMICA/B. We explored the effect of the PI3K/AKT signalling pathway on the expression of MMP9 in HCC cells. The PI3K inhibitor Ly294002 was used to treat HCC cells. The results showed that the upregulation effect of AFP overexpression on MMP9 expression was inhibited. These results indicate that AFP stimulates the expression of MMP9 by activating the PI3K/AKT signalling pathway, thus regulating the expression and shedding of MICA/B in HCC cell membranes. The results of animal experiments further confirmed that interference with AFP inhibited the expression of MMP9 and promoted the membrane protein content of MICA/B in HCC tissues.
To explore the relationship between AFP and NK cells, TCGA data was used to analyse the correlation between AFP and NK cell immune infiltration. The results showed that AFP was negatively correlated with the immune infiltration of NK cells. To verify this hypothesis, HCC cells were co‐cultured with NK‐92 cells, and the cytotoxicity of NK‐92 cells toward HCC cells was detected using an LDH release assay. The results showed that AFP inhibited the cytotoxicity of NK‐92 cells toward HCC cells, which was consistent with the results of TCGA data analysis. In addition, NKG2D neutralising antibodies could reverse the effect of increased AFP interference on NK‐92 cytotoxicity, and MMP9 inhibitors could reverse the effect of AFP inhibition on NK‐92 cytotoxicity. Further studies showed that AFP inhibited the cytotoxicity of NK‐92 cells to HCC cells by inhibiting the release of Perforin, Granzyme B, and interferon‐γ from NK‐92 cells. The survival of HCC cells overexpressing AFP was significantly increased according to live‐cell imaging. The survival of HCC cells that interfered with AFP expression was significantly reduced.
Overexpression of AFP significantly inhibited the content of MICA/B in the HCC cell membrane, whereas the MMP9 inhibitor reversed the effect of AFP inhibition and increased the expression of MICA/B but did not recover to the level observed in the untreated group. Overexpression of AFP also significantly increased sMICA/B levels, which were reversed by the MMP9 inhibitor. These results suggest that AFP inhibits the content of MICA/B in HCC cell membranes by stimulating the expression of MMP9. Fu et al. [28] found that inhibition of the aerobic glycolysis pathway increased the expression of MICA/B membrane proteins, indicating that tumour cells could inhibit the expression of MICA/B through the glycolysis pathway. Aerobic glycolysis in HCC cells was also known as the Warburg effect, leading to HCC drug resistance [29], and studies have shown that the Warburg effect was increased in HCC [30, 31, 32]. AFP was a diagnostic marker for HCC and was widely present in the peripheral blood [33]. Although no study has revealed a relationship between AFP and the Warburg effect, existing research suggests that AFP may be positively correlated with the Warburg effect. Therefore, whether AFP regulates MICA/B expression via aerobic glycolysis requires further investigation. Recently, we analysed the molecular structure of AFP, accurately locating the amino acid sites of AFP that carry metal ions, fatty acids, and glycosylation [34], and found that tumour‐derived AFP could inhibit the phagocytosis of macrophages in liver cancer cells [35]. In this study, AFP promoted the expression of MMP9 by activating the PI3K/AKT signalling pathway, shedding MICA/B (sMICA/B) on the cellular membrane of liver cancer cells. sMICA/B can bind to NKG2D in NK cells, blocking the recognition of MICA/B on liver cancer cell membranes by NKG2D, resulting in the inability of NK cells to recognise liver cancer cells, leading liver cancer cells to escape the attack of NK cells. These results suggested that AFP was a key molecule that promotes immune escape from liver cancer. Figure 7 shows the mechanism by which AFP inhibits MICA/B levels in liver cancer cell membranes and promotes cancer cells to escape from NK cell attacks.
This study also has certain limitations. First, we only elucidated the mechanism by which the AFP‐MMP9‐MICA/B axis resists the cytotoxicity of NK‐92 cells in vitro co‐culture experiments, while the in vivo experiments merely validated the regulatory relationship within the AFP‐MMP9‐MICA/B axis. Therefore, future studies should employ humanised immune system mouse models to verify the relationship between the AFP‐MMP9‐MICA/B axis and NK cells, thereby further clarifying its regulatory mechanism. Second, this research was conducted based on the established mechanism that “sMICA/B hinders the binding of NKG2D to MICA/B on the tumour cell membrane, thereby reducing NK cell recognition and killing capacity [36, 37],” and “high levels of plasma sMICA serve as an independent prognostic factor for poor outcomes in HCC [38].” We only verified that an NKG2D neutralising antibody could reverse the effect of interfering with AFP in promoting NK‐92 cell cytotoxicity, without elucidating the role of sMICA/B in this process. Consequently, future research should involve treating NK cells with human recombinant sMICA/B to examine whether the NKG2D‐MICA/B axis was blocked and whether this reverses the immunosuppressive effect of AFP.

5
Conclusion
In summary, the present study elucidated that AFP upregulated MMP9 expression through the PI3K/AKT signalling pathway. MMP9 acts as an enzyme to shed MICA/B on the cellular membrane of liver cancer. This in turn partially inhibited the level of MICA/B on the membranes of HCC cells, increased the content of sMICA/B in the cellular supernatant, and blocked the interaction between MICA/B and NKG2D, thereby inhibiting the activation of NK‐92 cells and increasing the resistance of HCC cells to NK‐92 cells. Targeted inhibition of AFP expression is a promising strategy to promote liver cancer immunotherapy.

Author Contributions

Xiaowei Li: investigation (equal), methodology (equal). Siren Feng: investigation (equal). Xueqin Wu: investigation (equal), writing – original draft (equal). Yinglian Pan: investigation (equal), methodology (equal). Yuli Zhou: data curation (equal), project administration (equal). Kun Liu: formal analysis (equal), investigation (equal). Bo Lin: project administration (equal). Wei Li: methodology (equal), project administration (equal). Mengsen Li: conceptualization (equal), investigation (equal), project administration (equal), writing – review and editing (equal). Mingyue Zhu: investigation (equal), project administration (equal), writing – review and editing (equal).

Funding
This work was supported by the National Natural Science Foundation of China (Nos. 82560459, 81960519, 82060514, and 82460602), the Natural Science Foundation of Hainan Province (Nos. 824RC517, 822RC700 and 2019CXTD406), the Hainan Province Science and Technology Special Foundation (No. ZDYF2021SHFZ222), and the Research Project of Take off the Proclamation and Leadership of the Hainan Medical University Natural Science Foundation (No. JBGS202106).

Consent
All authors have read and agreed to publish this manuscript.

Conflicts of Interest
The authors declare no conflicts of interest.

Supporting information

Figure S1. Overexpression of AFP upregulates the expression of MMP9 and inhibits the expression of MICA/B on the cell membrane of HLE cells. (A) The Western blotting experiment was used to detect the expression of AFP and MMP9 in the HCC cell line. (B) Flow cytometry was applied to detect the expression of MICA/B in the HCCr cell lines. **p < 0.01; ***p < 0.001; MICA/B, major histocompatibility complex class I chain‐related proteins A and B; MMP9, matrix metallopeptidase 9; siMMP9, small interfering matrix metallopeptidase 9.

Video S1. HuH7‐NC cells and NK‐92 cells were co‐cultured, and the attack effect of NK‐92 cells on HuH7‐NC cells was captured dynamically every 2 h using intelligent high throughput live cell imager.

Video S2. HuH7‐shAFP cells and NK‐92 cells were co‐cultured, and the attack effect of NK‐92 cells on HuH7‐shAFP cells was captured dynamically every 2 h using intelligent high throughput live cell imager.