First, we used FBXO22 as a query in the “Tissue Atlas” module of The Human Protein Atlas web service (www.proteinatlas.org) in order to obtain FBXO22 expression levels in different healthy cell and tissue types. Next, we downloaded HTSeq-FPKM data for 33 cancer types from TCGA using the software package TCGAbiolinks (v2.20.0) (17) in R (v4.1.0) to explore FBXO22 expression levels in different cancers. For cancer types containing a number of healthy samples greater than ten, including bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), colon adenocarcinoma (COAD), esophageal carcinoma (ESCA), head and neck squamous cell carcinoma (HNSC), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), prostate adenocarcinoma (PRAD), rectum adenocarcinoma (READ), stomach adenocarcinoma (STAD), thyroid carcinoma (THCA), and uterine corpus endometrial carcinoma (UCEC), we performed expression level analysis of the FBXO22 gene in healthy and tumor tissues using R. For cancer types with no healthy samples, or with a healthy sample number less than ten, including adrenocortical carcinoma (ACC), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), acute myeloid leukemia (LAML), brain lower-grade glioma (LGG), ovarian serous cystadenocarcinoma (OV), testicular germ cell tumors (TGCT), uterine carcinosarcoma (UCS), sarcoma (SARC), pancreatic adenocarcinoma (PAAD), glioblastoma multiforme (GBM), pheochromocytoma and paraganglioma (PCPG), thymoma (THYM), cholangiocarcinoma (CHOL), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), and skin cutaneous melanoma (SKCM), we used the “Expression Analysis-Expression DIY-Box Plots” module of the GEPIA2 web server (18) to obtain box plots of the FBXO22 mRNA levels, as an indicator of gene expression, in tumor and healthy tissues by matching TCGA normal and genotype–tissue expression (GTEx) data. For an expression analysis of FBXO22 protein in BRCA, LUAD, and UCEC, we used data from the Clinical Proteomic Tumor Analysis Consortium (CPTAC) in the UALCAN web service (19, 20).

In the “Survival Analysis” module on the GEPIA2 web server (18), samples were stratified into high- and low-expression groups according to the median expression of FBXO22 gene in each cancer type, and overall and relapse-free survival (RFS) analyses were performed. We performed a “tumor vs. normal” meta-analysis, a “survival” meta-analysis, and an “overall survival” (OS) analysis of FBXO22 gene expression using the LUNG CANCER EXPLORER service (21). We used the Kaplan-Meier plotter to analyze the correlation between the FBXO22 expression and OS/RFS in pan-cancer (22).

To explore the regulatory role of FBXO22 in tumorigenesis, we identified FBXO22-binding proteins using the STRING database (23) and collected FBXO22 substrates through a literature collection. Next, we performed Gene Ontology (GO) biological process enrichment analysis of genes encoding these proteins using the clusterProfiler package (v4.0.2) (24) in R.

Human NSCLC cells (nci-h1975 and A549), human bronchial epithelial cells 16HBE, and HEK293T cells were purified from the Chinese Academy of Sciences Shanghai cell bank (Shanghai, China). A549 and 16HBE cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS, JRH Biosciences, Lenexa, KS, USA). Human NSCLC H1975 cells were cultured in RPMI-1640 medium (Gibco, Life Technologies, Waltham, CA, USA) with 10% FBS. For DNA methyltransferase inhibition assays, SGI-1027 (Shyuanye, Shanghai, China) was added to the culture medium for a final concentration of 2, 4, or 8 μM. All cells were cultured in a humidified cell incubator at 37°C with 5% CO₂.

Cell proliferation was evaluated using a Cell Counting Kit-8 (CCK8) assay (Best Bio, Shanghai, China). In brief, cells were seeded in 96-well plates at a concentration of 3 × 10³ per well and cultured for 1, 2, and 3 days. CCK-8 solution (10 µl) was added into each well at the indicated time points, then the plates were stored for 2 h at 37°C. Next, the number of viable cells was estimated by measuring the absorbance at 450 nm.

Cells were collected and lysed using Qproteome™ Mammalian Protein Prep Kit (Qiagen, Hilden, Germany). After centrifugation at 13,800×g for 10 min, the protein content in the supernatant was determined using the BCA protein assay kit (Bio-Rad, Shanghai, China). Equal amounts of protein were boiled by adding 4× sample loading buffers for 10 min at 100°C and resolved using sodium dodecylsulfate polyacrylamide gel electrophoresis (SDS-PAGE). Antibodies against α-tubulin (11224-1-AP), CDK4 (11026-1-AP), and FBXO22 (13606-1-AP) were purchased from Proteintech (Rosemont, IL, USA), DNTM1 (ab188453) were purchased from Abcam (Cambridge, MA, USA), and PTEN (9188S) and ubiquitin (3936S) were purchased from Cell Signaling Technology (Danvers, MA, USA).

Cells were harvested, lysed, and briefly sonicated. After centrifugation at 12,000×g for 10 min at 4°C, the supernatant (whole-cell lysate) was collected from each sample. PTEN antibodies were incubated with protein A magnetic beads at room temperature for 4 h, and unbound antibodies were washed off with elution buffer. Subsequently, the beads were incubated with cell lysate supernatants at 4°C overnight. The precipitates were washed three times with immunoprecipitation buffer, boiled in sample buffer, and Western blot analysis was then performed.

Plasmids pLV[Exp]-EGFP:T2A:Puro-CMV>FLAG/hFBXO22[NM_147188.3] (VB210508-1064ndh), pRP[shRNA]-EGFP:P2A:Puro-U6>Scramble[shRNA#1] (VB200706-2709njm) and pRP[shRNA]-EGFP:P2A:Puro-U6>hFBXO22[shRNA#1] (VB210408-1046ugc) were purchased from VectorBuilder (Guangzhou, China). For the lentiviral packaging, we used pCMV delta R8.2 (Addgene 12263) and pCMV-VSV-G (Addgene 8454) systems. For plasmid transfection, we used Lipofectamine 2000 (Invitrogen, Waltham, MA, USA).

The Catalogue of Somatic Mutations in Cancer (COSMIC) is the largest source of expert manually curated somatic mutation information relating to human cancers in the world (25). First, we used COSMIC to explore the distribution of different types of mutations in FBXO22. Next, we used the term “FBXO22” in the “Quick Search Beta” module of the cBioPortal web service (26, 27) to analyze the genetic alteration status of FBXO22 in different cancer types from TCGA cohorts. We then examined the association between genetic alterations in FBXO22 and clinical outcomes for UCEC. We used the UCEC (TCGA, PanCancer Atlas) dataset in our query. To explore the relationship between FBXO22 mutations and FBOXO22 substrate gene expression, we used the “Gene_Mutation” module in TIMER2 (28).

To assess the methylation level of FBXO22 in cancer and healthy tissues, we used “TCGA” module in the UALCAN database to obtain boxplots for FBXO22 methylation level in COAD, PRAD, CESC, TGCT, READ, KIRP, LUAD, LUSC, BRCA, UCEC, and SARC (19). DNMIVD is a comprehensive annotation and interactive visualization database for DNA methylation profiles in diverse human cancers (29). The DNMIVD tool was used to investigate the correlation between expression, OS, disease-free interval, and FBXO22 promoter methylation levels. Searches were performed using FBXO22 as the input on the “Home” or “Model” page of the DNMIVD web service. Retrieved data and images were downloaded. The methylation sites cg08290738, cg00942495, and cg05374463 of FBXO22 were also used as inputs in other analyses.

TISIDB, a web portal for tumor and immune system interaction (30), was used to explore relationships between tumor-infiltrating lymphocyte (TILs) abundance and FBXO22 expression, copy number variation (CNV), and methylation. The activity of the tumor immunity cycle is a direct, integrated manifestation of the functions of the chemokine system and other immune regulators (31, 32). Thus, we also used TISIDB to analyze correlations between FBXO22 expression levels and chemokines, receptors, and three kinds of immunomodulators (immunoinhibitors, immunostimulators, and major histocompatibility complexes (MHCs)) across human cancers. For the relationship between the FBXO22 expression and the level of infiltrating natural killer (NK) T cells and myeloid-derived suppressor cells (MDSCs) across the diverse cancer types, we used the “Immune” module of the TIMER2 webserver (28). Data and pictures of the analyzed results were downloaded, and Spearman’s correlation heatmaps were visualized in R using the “pheatmap” package (v1.0.12). To avoid computational errors caused by a single algorithm and different sets of marker genes for TIL, we downloaded immune infiltrate data evaluated using the CIBERSORT (33), CIBERSORT-ABS (34), EPIC (35), MCP-counter (36), quantTIseq (37), xCell (38), TIMER (39), and TIDE (40) algorithms for the 33 cancer types from TCGA database using the TIMER2 web server (28). Similarly, immune infiltrate data evaluated using Tracking Tumor Immunophenotype (TIP) algorithms and immune activity score data for the 33 cancer types were downloaded from the TIP database. In addition, the single-sample gene set enrichment analysis (ssGSEA) based on the “gsva” package (v1.40.1) in R was also used to evaluate differences in the tumor-infiltrating fractions of 28 human immune cell phenotypes in the tumor microenvironment (41–43). Subsequently, Spearman’s correlations between FBXO22 expression and the level of TILs were calculated in R.

Correlations between variables were explored using Spearman’s coefficients. Wilcoxon’s rank-sum tests were used to compare continuous variables between binary groups. Survival curves for categorical variable prognostic analyses were generated using the Kaplan-Meier method, while the log-rank test was used to estimate statistical significance. The significance level was set at p < 0.05, and all statistical tests were two-sided. All statistical data were analyzed using R or online analysis tools described in the relevant Materials and Methods subsections.

First, we aimed to characterize the expression level of FBXO22 in different healthy tissues and cells, based on consensus data from the human protein atlas (HPA) database. The expression of FBXO22 gene was highest in the liver, followed by the placenta and adrenal gland, and lowest in the ductus deferens ( Supplementary Figure S1A ). Among the different cell types, cone and rod photoreceptor cells had the highest expression compared with other cell types, followed by early spermatids ( Supplementary Figure S1B ).

To explore the role of FBXO22 expression in human cancer, we performed a pan-cancer analysis of FBXO22 using Gene Expression Quantification data of 33 cancers from TCGA database. Among the cancer types with a healthy sample number greater than ten, we observed high expression of FBXO22 in tissues from 14 tumor types (UCEC, ESCA, HNSC, KICH, READ, THCA, LIHC, KIRP, BLCA, BRCA, COAD, LUSC, STAD, and LUAD) compared with adjacent healthy tissues, but no significant difference in expression was observed in PRAD or KIRC ( Figure 1A ). For cancer types with no healthy samples or with a healthy sample number less than ten, we used the “Expression DIY” module of GEPIA2 web services to match GTEx datasets as controls. As shown in Figure 1B , the FBXO22 expression level in DLBC, PAAD, GBM, THYM, CESC, and SKCM was relatively high compared with that of corresponding healthy tissues; conversely, it was found to be low in LAML compared with corresponding healthy tissue. Furthermore, we noticed that the expression of FBXO22 also increased in LGG, OV, TGCT, USC, SARC, SARC, PCPG, and CHOL, though the difference was not statistically significant. To assess posttranslational levels of FBXO22, we analyzed FBXO22 protein expression using the CPTAC dataset in the UALCAN database. Consistent with our gene expression results, FBXO22 protein levels were highly expressed in BRCA, LUAD, and UCEC compared with healthy tissue ( Figure 1C ). FBXO22 was also upregulated in the NSCLC cell lines H1975 and A549 compared with the human bronchial epithelial cell line 16HBE ( Figure 1D ). Taken together, these results demonstrate that FBXO22 expression is upregulated in multiple cancers, implying that high FBXO22 expression levels may be associated with tumor progression.

To explore the prognostic value of FBXO22 in people with different cancer types, we used the Survival Analysis module in GEPIA2 to analyze the correlation between FBXO22 expression and percent survival of people with different tumors. As shown in Figure 2A , high FBXO22 expression was associated with poor prognosis and lower OS in patients with KICH (Logrank p = 0.016), ACC (log-rank p = 0.097), UVM (log-rank p = 0.031), PAAD (log-rank p = 0.014), and SARC (log-rank p = 0.21). Conversely, low FBXO22 expression was associated with poor prognosis and OS in patients with ESCA (log-rank p = 0.32) or KIRC (log-rank p = 0.002). We also observed that high expression of the FBXO22 gene was associated with poor RFS in KICH (log-rank p = 0.039), ACC (log-rank p = 0.011), UVM (log-rank p = 0.16), and HNSC (log-rank p = 0.39), but with better RFS in COAD (log-rank p =0.022), ESCA (log-rank p = 0.22), and TGCT (log-rank p = 0.14). Additionally, we used the LUNG CANCER EXPLORER service to perform tumor vs. normal and survival meta-analyses in different lung cancer datasets. We found that FBXO22 expression was positively associated with hazard rate (HR) in most datasets ( Supplementary Figures S2 – S3 ). High expression of FBXO22 was associated with poor OS in the lung cancer datasets GSE37745 (44) and GSE17710 (45) ( Figure 2B ). We then used Kaplan-Meier analysis tool to determine the relationship between pan-cancer and FBXO22 expression; as shown in Figure 2C , high expression of FBXO22 was linked to poor OS and RFS in pan-cancer. Taken together, these data indicate that FBXO22 expression is associated with poor prognosis in cancer.

To explore the regulatory role of FBXO22 in tumorigenesis, we identified 10 FBXO22-binding proteins using the STRING database ( Figure 3A ). Next, we identified 13 proteins as substrates of FBXO22 ubiquitination and degradation ( Table 1 ). FBXO22 regulates the levels of these proteins to alter tumor progression. Finally, we assessed the corresponding biological processes associated with these 23 FBXO22-associated proteins using GO enrichment analysis. We discovered that the genes encoding the FBXO2-associated proteins were significantly enriched in negative regulation of cell cycle and protein ubiquitination-related pathways ( Figure 3B ). To confirm this, a stable A549 cell line overexpressing FBXO22 was constructed. Overexpression of FBXO22 promoted proliferation in these cells ( Figure 3C ) and enhanced the level of cyclin-dependent kinase 4 (CDK4) ( Figure 3D ). This indicates that FBXO22 is related to the cell cycle regulation. Notably, we observed decreased levels of PTEN, a FBXO22 substrate, in FBXO22-overexpressing cells ( Figure 3D ). Subsequently, we explored the potential regulation of PTEN by FBXO22 via ubiquitination. We immunoprecipitated PTEN and found that the level of ubiquitinated PTEN was increased in FBXO22-overexpressing cells compared with wild type ( Figure 3E ). Conversely, the ubiquitination level of PTEN was decreased after interfering FBXO22 ( Figure 3E ). This demonstrates that FBXO22 affects the progression of NSCLC by regulating PTEN ubiquitination levels.

Genetic alterations in FBXO22 may affect its function. Hence, we explored the genetic alteration status of FBXO22 in human cancers. First, we explored the distribution of different mutation types for FBXO22 using COSMIC. As shown in Figure 4A , a missense substitution is the most commonly observed mutation type, observed in the 30.53% of the cancer types. We also used the cBioPortal tool to analyze the genetic alteration status of FBXO22 in different cancer types from the TCGA cohorts. We observed the highest FBXO22 alteration frequency among patients with UCEC, followed by mesothelioma ( Supplementary Figure S4A ). Notably, the FBXO22 alterations in mesothelioma were all identified as “Amplification” ( Supplementary Figure S4A ). Additionally, we found that the frequency of arginine to histidine or cysteine mutations at position 96 in the FBXO22 protein was the highest among all mutations ( Supplementary Figure S4B ). Next, we examined the association between genetic alterations in FBXO22 and clinical outcomes. We discovered that altered FBXO22 was associated with better prognosis in terms of OS (p = 0.021), progression-free (p = 0.0395), and disease-specific (p = 0.091) survival, but not disease-free (p = 0.357) survival, compared with cases without FBXO22 alteration among UCEC cases ( Figure 4B ). This implies that alterations in FBXO22 may be beneficial for UCEC patients.

Finally, we utilized the “gene_Mutation” module in TIMER2 to compare the substrate gene expression according to FBXO22 mutation status. We found that FBXO22 mutation was correlated with the expression of at least one substrate gene in COAD, GBM, LUAD, LUSC, SKCM, STAD, and UCEC ( Figure 4C and Supplementary Figures S4C, D ). We noted an increased expression of MDM2, PHLPP1, and PTEN in LUAD samples with mutated FBXO22 and an increased expression of CDKN1A, KDM4B, STK11, and PTEN in UCEC samples with mutated FBXO22 ( Figure 4C ). These findings suggest that mutant FBXO22 may lose E3 ligase activity, especially in people with UCEC and LUAD. This also explains why FBXO22 mutation was associated with better prognosis in patients with UCEC.

Next, we explored whether FBXO22 methylation was associated with clinical prognosis. We used the UALCAN tool to analyze FBXO22 methylation in various cancers from TCGA database. We found that the level of methylation in the FBXO22 promoter region was significantly decreased in COAD, PRAD, CESC, TGCT, READ, KIRP, LUAD, LUSC, BRCA, UCEC, and SARC compared with normal tissues ( Figure 5A ). This may be related to the high FBXO22 expression in multiple cancer types. Indeed, inhibiting the expression of DNA methyltransferase 1 (DNTM1) in A549 cells resulted in elevated FBXO22 ( Figure 5B ). We subsequently explored whether FBXO22 methylation was associated with prognosis among people with cancer using the DNMIVD tool. We found that hypermethylation of FBXO22 was related to poor OS, disease-free interval, and progression-free interval in PAAD ( Figure 5C ). Similarly, the association of methylated CpG islands and prognosis was analyzed using DNMIVD. Feature importance score calculation using the xgboost algorithm revealed 3 significant methylated CpGs: cg00942495, cg05374463, and cg08290738 ( Supplementary Figure S5A ). We identified a correlation between low cg08290738 methylation and poor OS in CESC and ESCA ( Supplementary Figure S5B ; high levels of cg00942495 or cg05374463 methylation were also found to be associated with better survival in THYM ( Supplementary Figure S5D ).

TILs are an important part of tumor microenvironment (TME) and are involved in the occurrence, development, and metastasis of tumors (46). However, the relationship between FBXO22 expression and TILs/TME is unclear. To understand the relationship between FBXO22 and the immune composition of tumors, we first used the TISIDB to evaluate the relationship between TIL abundance and FBXO22 expression, CNV, and methylation. As shown in Supplementary Figure S6 , the abundance of most TIL types was positively correlated with FBXO22 expression in both GBM and UVM cancers; however, the abundance of almost all TIL types in other cancers was negatively correlated with expression of FBXO22 ( Supplementary Figure S6A ). There was a weak correlation between changes in FBXO22 CNV and TIL levels ( Supplementary Figure S6B ). The methylation of FBXO22 was strongly positively correlated with the abundance of TILs in PRAD and TGCT ( Supplementary Figure S6C ). These lines of evidence suggest that the expression level and genetic alterations of FBXO22 in the tumor may influence antitumor immunity. The activity of the tumor immune cycle is a direct integrated manifestation of the functions of the chemokine system and other immune regulators (31, 32). Thus, we used the TISIDB to analyze correlations between the FBXO22 expression and chemokines and their receptors across human cancers. We found a significant negative correlation between FBXO22 expression and chemokines or receptors in almost all cancers ( Supplementary Figures S7A, B ). Similar analyses have found that three kinds of immunomodulators (immunoinhibitors, immunostimulators, and MHCs) were also significantly inversely correlated with FBXO22 expression ( Supplementary Figures S7C–E ). Taken together, this evidence suggests that tumor FBXO22 expression plays an important role in tumor immune regulation.

To avoid computational errors arising from the use of a single algorithm and different marker gene sets of TIICs, we utilized the CIBERSORT, CIBERSORT-ABS, EPIC, MCP-counter, quanTIseq, TIMER, TIP, and xCell algorithms to evaluate the correlation between FBXO22 expression and the level of immune infiltration of TILs. Although there were differences in the levels of immune cell infiltration calculated using different algorithms, we noted that the calculations produced by at least five of the algorithms showed that B-cell and macrophage infiltration levels were negatively correlated with FBXO22 expression in multiple cancers, including LUAD, LUSC, and SARC ( Supplementary Figures S8A–H ). We discovered significant negative correlations between FBXO22 expression and the NK T-cell infiltration level as estimated by the XCELL algorithm in many types of cancer, including BRCA-Her2 (Rho = −0.471), CHOL (Rho = −0.403), PRAD (Rho = −0.554), and UVM (Rho = −0.67) ( Figure 6 ). Similarly, we found a significant positive correlation between FBXO22 expression and the MDSC infiltration level, as estimated by TIDE algorithm, including in BLCA (Rho = 0.286), ESCA (Rho = 0.365), READ (Rho = 0.435), and SKCM-primary (Rho = 0.511) ( Figure 6 ). NK T cells enhance the function of dendritic cells (DCs), T cells, B cells, and other immune cells by secreting various cytokines and chemokines (47). MDSCs may not only inhibit the antitumor immune response but also directly stimulate tumor growth and metastasis (48). These results suggest that high FBXO22 expression in tumors may be detrimental to the antitumor immune response.

Antitumor immune responses can be conceptualized as a series of stepwise events including the release of cancer cell antigens (step 1), cancer antigen presentation (step 2), priming and activation (step 3), trafficking of immune cells to tumors (step 4), infiltration of immune cells into tumors (step 5), recognition of cancer cells by T cells (step 6), and killing of cancer cells (Step 7) (32). Xu and colleagues evaluated the activities in these steps using ssGSEA based on gene expression in individual samples (32). Therefore, we performed ssGSEA to decipher the involvement of FBXO22 in the immune activation process. We found that FBXO22 expression in LUAD, LUSC, KIRP, KIRC, SARC, and LAML were negatively correlated with the levels of infiltration of multiple immune cells ( Figure 7A ). Notably, these immune response activation steps were significantly activated when FBXO22 expression was low in LUAD, LUSC, and SARC ( Figure 7B ). Together, these data suggest that FBXO22 expression in these tumors may play a critical role in the immune response to tumors.