Prognostic and Immunological Implications of FAM72A in Pan-Cancer and Functional Validations

The family with sequence similarity 72 Member A (FAM72A) is overexpressed in several types of cancer. However, its contributions to tumorigenesis remain largely unknown. Based on The Cancer Genome Atlas (TCGA) database, FAM72A was upregulated across 33 types of cancer. Accordingly, high levels of FAM72A predicted inferior outcomes in half of the cancer types using survival analysis (the Kaplan-Meier curve and univariate Cox regression model). Receiver operating characteristic (ROC) analysis demonstrated that FAM72A showed high accuracy in distinguishing cancerous tissues from normal ones. FAM72A was correlated with immune and stromal scores and immune cell infiltrations in various tumors. Moreover, FAM72A was also associated with tumor mutation burden (TMB), microsatellite instability (MSI), and immune checkpoint genes. Immunophenoscore (IPS) further validated that the FAM72Alow tumor showed high immunogenicity and tended to respond to anti-PD1/PDL1/PDL2, anti-CTLA4 treatment, and combined immunotherapies. We also investigated the functional role of FAM72A in lung adenocarcinoma (LUAD). In vitro studies demonstrated that the ectopic expression of FAM72A accelerated the proliferation and migration of NSCLC cells, whereas silencing FAM72A showed the opposite effects on them. In short, FAM72A had prognostic potential and correlated with tumor immunogenicity in various tumors. Functional analysis indicated that FAM72A is an oncogene in LUAD.


Introduction
The family with sequence similarity 72 Member A (FAM72A), also known as Ugene, was initially identified in malignant colon cancers and overexpressed in several other common cancers, including breast cancer, lung cancer, uterus, and ovary cancer [1]. Accumulating evidence has indicated the critical role of FAM72A in tumorigenesis, but its functions are largely uncharacterized.
Several studies indicated that FAM72A functions by interacting with uracil DNAglycosylase 2 (UNG2), a crucial enzyme in base excision repair (BER), catalyzing the replacement of deoxyuracil (dU) with deoxycytidine (dC) to avoid mutations [1][2][3][4]. Interestingly, two papers regarding FAM72A's roles in B cell antibody diversification were published back-to-back in Nature [2,3]. With the utilization of a genome-wide CRISPR screen, Feng et al. found that FAM72A induced mutagenic DNA repair during antibody maturation by encountering UNG2. FAM72A promoted U·G mispairs, and B cells from Fam72a knockout mice exhibited somatic hypermutation [3]. Meanwhile, Rogier et al. also reported that FAM72A prompted diversification of the B cell receptor repertoire by facilitating error-prone DNA repair [2]. These findings suggest that FAM72A overexpression may accelerate mutagenesis in cancer. Previous bioinformatic analysis indicated that FAM72A

Landscape of FAM72A mRNA Expression Levels in Pan-Cancer Tissues
The flow chart of this research is shown in Supplementary Figure S1. Compared to normal tissues, significantly elevated mRNA expression levels of FAM72A were observed in 82% of 33 cancer types, including bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), esophageal carcinoma (ESCA), GBM, lung squamous cell carcinoma (LUSC), lung adenocarcinoma (LUAD), and head and neck squamous cell carcinoma (HNSC) ( Figure 1A,C). We next conducted single-sample gene set enrichment analysis (ssGSEA) with 100 FAM72A-correlated genes to calculate FAM72A activity. FAM72A activity was significantly higher in cancers than in normal tissue in tumors with increased FAM72A levels ( Figure 1B). Since the activity score was derived from FAM72A-correlated genes, samples with high FAM72A expression levels are more likely to have high activity scores. We performed the correlation test for the expression and activity of FAM72A in several randomly selected tumor types, and positive correlations were observed (Supplementary Figure S2). Additionally, FAM72A expression levels increased with tumor progression in adrenocortical carcinoma (ACC), BLCA, CESC, kidney chromophobe (KICH), LUAD, kidney renal clear cell carcinoma (KIRC), and kidney renal papillary cell carcinoma (KIRP) ( Figure 1D).  . For tumors lacking corresponding normal tissue samples, the GEPIA web tool, incorporating normal tissues from the Genotype-Tissue Expression (GTEx) database, was used to compare FAM72A expression between the tumor and the normal tissues. (D). Expression of FAM72A in different pathological stages of indicated tumors. * p < 0.05, ** p < 0.01, and *** p < 0.001.

Correlation between FAM72A Expression Level and Cancer Prognosis
The clinical relevance of FAM72A in different tumor types was explored using Kaplan-Meier survival analysis. The results showed that high FAM72A levels were associated with poor OS in ACC, KIRC, KIRP, liver hepatocellular carcinoma (LIHC), LUAD, mesothelioma (MESO), uterine corpus endometrial carcinoma (UCEC), and uveal melanoma (UVM), but had a favorable OS with thymoma (THYM) (Figure 2A, Supplementary Figure S3). Moreover, a univariate Cox regression analyses identified high FAM72A as a risk factor for OS in ACC, KICH, KIRC, KIRP, brain lower grade glioma (LGG), LIHC, LUAD, MESO, pancreatic adenocarcinoma (PAAD), UCEC, and UVM, but a protective factor for overall survival (OS) in THYM ( Figure 2B). Similar analyses were also performed against disease-specific survival (DSS), progression-free interval (PFI), and disease-free interval (DFI), ( Figure 2C-E, Supplementary Figures S4-S6). These results suggest that FAM72A expression has a solid prognostic ability in different tumors.

Correlation between FAM72A Expression Level and Cancer Prognosis
The clinical relevance of FAM72A in different tumor types was explored using Kaplan-Meier survival analysis. The results showed that high FAM72A levels were associated with poor OS in ACC, KIRC, KIRP, liver hepatocellular carcinoma (LIHC), LUAD, mesothelioma (MESO), uterine corpus endometrial carcinoma (UCEC), and uveal melanoma (UVM), but had a favorable OS with thymoma (THYM) (Figure 2A, Supplementary Figure S3). Moreover, a univariate Cox regression analyses identified high FAM72A as a risk factor for OS in ACC, KICH, KIRC, KIRP, brain lower grade glioma (LGG), LIHC, LUAD, MESO, pancreatic adenocarcinoma (PAAD), UCEC, and UVM, but a protective factor for overall survival (OS) in THYM ( Figure 2B). Similar analyses were also performed against disease-specific survival (DSS), progression-free interval (PFI), and disease-free interval (DFI), ( Figure 2C-E, Supplementary Figures S4-6). These results suggest that FAM72A expression has a solid prognostic ability in different tumors.  . Kaplan-Meier survival analyses were performed to determine the association between FAM72A and several clinical endpoints, including overall survival (OS), disease-specific survival (DSS), disease-free interval (DFI), and progression-free interval (PFI) in 33 tumors. Statistical results were summarized and visualized using the "survival" and "ggplot2" R packages. The forest plots of univariate Cox regression analyses for FAM72A regarding OS (B), DSS (C), PFI (D), and DFI (E).

Diagnostic Value of FAM72A for Pan-Cancer
We also used receiver operating characteristic (ROC) to evaluate the discrimination ability of FAM72A expression levels between cancerous and normal tissues. The area under the curve (AUC) values for ROC analysis in each cancer are shown in Figure

Correlation between FAM72A and Tumor Microenvironment in Different Tumors
Intratumoral stromal and immune cells constitute the majority of tumor-associated normal cells. Not only immune cells but also stromal cells regulate tumor growth and progression. The Estimation of Stromal and Immune cells in Malignant Tumor tissues using Expression data (ESTIMATE) algorithm was used to evaluate infiltrating immune and stromal cells by calculating immune scores and stromal scores. FAM72A expression was

Correlation between FAM72A and Tumor Microenvironment in Different Tumors
Intratumoral stromal and immune cells constitute the majority of tumor-associated normal cells. Not only immune cells but also stromal cells regulate tumor growth and progression. The Estimation of Stromal and Immune cells in Malignant Tumor tissues using Expression data (ESTIMATE) algorithm was used to evaluate infiltrating immune and stromal cells by calculating immune scores and stromal scores. FAM72A expression was found to positively correlate with the immune score in KIRC, KIRP, LGG, KIRC, thyroid carcinoma (THCA), UVM, and PCPG, but negatively correlated with the immune score in CESC, STAD, LUSC, UCEC, HNSC, READ, COAD, and ESCA ( Figure 4A). Regarding stromal score, a positive correlation with FAM72A expression was seen in THCA, PCPG, PRA, acute myeloid leukemia (LAML), LGG, KIRP, KIRC, and a negative correlation in BRCA, UCEC, STAD, testicular germ cell tumors (TGCT), LUSC, and HNSC (Supplemental Figure S7A-M).
Previous studies suggest that FAM72A may be implicated in immunogenicity [2,3]. With this in mind, we explored the correlation between FAM72A expression and intratumoral immune cell infiltration. The Estimating Relative Subsets of RNA Transcripts (CIBERSORT) algorithm was used to determine the composition of 22 immune cell subsets based on gene expression profiles.
Results indicated that FAM72A was significantly correlated with immune cell infiltration across different tumor types. For instance, FAM72A was significantly associated with ten or more immune cell subsets in BRCA, KIRC, LIHC, LUAD, LUSC, STAD, THCA, and THYM.  Figure 4B).

The Signaling Pathway Associated with FAM72A
To investigate the potential function of FAM72A Pan-cancer, we carried out a gene set enrichment analysis (GSEA). The top five signaling pathways significantly associated with FAM72A are exhibited in Figure 7. In LGG, FAM72A was associated with inflammatory response, interferon-gamma response, cell cycle, cytokine-cytokine receptor interaction, JAK-STAT signaling, and nature killer cell-mediated cytotoxicity ( Figure 7A,B). In THYM, FAM72A was associated with G2M checkpoint, MTORC1 signaling, MYC targets v1, DNA replication, two DNA repair pathways, oxidative phosphorylation, and primary immunodeficiency ( Figure 7C,D). In UCS, FAM72A was associated with inflammatory response, interferon α response, interferon γ response, and TNFα signaling via NF-κB ( Figure 7E,F). These results suggest that FAM72A plays a fundamental role in tumorigenesis.

In Vitro Validation
We first examine the endogenous expression levels of FAM72A in normal lung and cancer cell lines. A549 cells with relatively low levels of FAM72A were used for gain-offunction analysis, while H1993 cells were adopted for loss-of-function analysis ( Figure  8A-C). The enforced expression of FAM72A promoted lung cancer cell proliferation as indicated by a CCK assay ( Figure 8D). Moreover, Transwell migration and wound healing

In Vitro Validation
We first examine the endogenous expression levels of FAM72A in normal lung and cancer cell lines. A549 cells with relatively low levels of FAM72A were used for gain-of-function analysis, while H1993 cells were adopted for loss-of-function analysis ( Figure 8A-C). The enforced expression of FAM72A promoted lung cancer cell proliferation as indicated by a CCK assay ( Figure 8D). Moreover, Transwell migration and wound healing assays demonstrated enhanced migration in cells overexpressing FAM72A ( Figure 8F,H). On the contrary, FAM72A depletion demonstrated the opposite effects on the proliferation ( Figure 8E) and migration of H1993 cells ( Figure 8G,I). These results verified FAM72A's oncogenic potential in lung cancer. assays demonstrated enhanced migration in cells overexpressing FAM72A ( Figure 8F,H). On the contrary, FAM72A depletion demonstrated the opposite effects on the proliferation ( Figure 8E) and migration of H1993 cells ( Figure 8G,I). These results verified FAM72A's oncogenic potential in lung cancer.

Expression and Prognostic Implication
A pan-cancer analysis revealed that compared to normal tissues, significantly elevated mRNA expression levels of FAM72A were observed in 82% of cancers, including BLCA, . Furthermore, several TCGA-based bioinformatic studies indicated that FAM72A might be involved in cancer. FAM72 consists of four human-specific paralogs (A-D). For instance, the expression of the Ki67 gene (MKI67), the putative proliferation marker, was found to be tightly associated with levels of FAM72A, B, and D across different cancers [5]. Yu and colleagues explored the prognostic value of FAM72A-D and found that all the members were overexpressed in LUAD and showed potential as prognostic markers [6]. Recently, a prognostic classifier of 10 mitochondrial-related genes was identified in hepatocellular carcinoma (HCC), including ACOT7, ADPRHL2, ATAD3A, BSG, FAM72A, PDK3, PDSS1, RAD51C, TOMM34, and TRMU) [13]. Interestingly, Gao et al. recently developed an angiogenesis factors-based prognostic signature in hepatocellular carcinoma, in which FAM72 is also one of six key genes (GRM8, SPC25, FSD1L, SLC386A, FAM72A, and SLC39A10) [14]. Zhou and colleagues validated the upregulation of FAM72A in tumor samples of both HCC patients and a mouse HCC model [15]. These results suggest that FAM72A is a promising prognostic marker in multiple cancer types and deserves extensive investigation.

Tumor Microenvironment
The tumor microenvironment (TME) is composed of tumor cells, stromal cells (e.g., fibroblasts and endothelial cells), a variety of immune cells (including T lymphocytes, NK cells, macrophages, and dendritic cells), and extracellular matrix (e.g., biochemical components released by these cells) [16]. These cells crosstalk with each other and collectively determine tumor fates, regression, or progression. TME is not only crucial for tumor proliferation, invasion, and metastasis but also affects the therapeutic effect [17]. With the development of high-throughput sequencing technology and machining learning, ESTIMATE was generated to infer stromal and immune cells in tumors based on RNAsequencing data [18]. We adopted this method to determine stromal and immune scores for samples across 33 cancer types and found that immune scores derived from the ESTIMATE algorithm were differentially correlated to FAM72A among different cancer types, and this was also the case for stromal cells.
We also used the CIBERSORT algorithm to estimate the proportion of 22 immune cell subsets in each tumor sample. The correlation test indicated that FAM72A was extensively correlated to immune cell infiltrates in BRCA, KIRC, LIHC, LUAD, LUSC, STAD, THCA, and THYM. However, the association of FAM72A with cytotoxic effector cells (e.g., CD8+ T and NK cells) or repressive immune cells (e.g., M2 macrophages and Treg cells) varied among different tumors. These results indicated that the role of FAM72 in tumor immunogenicity is tissue specific.

Correlation to TMB, MSI, and ICPs
Given the tight correlation between FAM72A and TME, we further explored whether FAM72 is related to the response to immunotherapy. A growing body of evidence suggests that MSI, TMB, and programmed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) expression levels are associated with an elevated response rate to immunotherapy and can facilitate the identification of patients suitable for immunotherapy. Our pan-cancer analysis revealed that FAM72A was significantly associated with the TMB in 18 cancer types, including ACC, BLCA, BRCA, COAD, HNSC, KICH, LGG, LUAD, LUSC, PAAD, PRAD, READ, SARC, SKCM, STAD, THCA, THYM, and UCEC. High TMB is consistently used to select proper patients for ICI therapy because somatic mutations in tumor DNA may have a chance to produce neoantigen-containing peptides, which can be processed, displayed onto major histocompatibility complex (MHC) molecules, and recognized by T cells. Although mutation-derived neoantigens are very rare, theoretically, the more somatic mutations a tumor harbors, the more immunogenic neoantigens that can be generated [19]. TMB variations have been characterized across different malignancies. TMB can represent a valuable estimation of tumor neoantigen load. To date, TMB is believed to be a key source of immunogenic neuropeptides presented on the MHC of the tumor cell surface, affecting patient response to ICIs [9]. Response to ICIs in solid tumors with high TMB have been observed in various tumors, including NSCLC, melanoma, and bladder cancer [20]. Mechanistic studies suggested that the reason that TMB predicts ICIs sensitivity might be attributed to TMB-mediated neoantigens and tumor immunogenicity [20].
Our results indicated positive correlations between FAM72A and MSI in COAD, ESCA, LUAD, LUSC, PRAD, READ, STAD, THCA, UCEC, and UCS. Mutations in MLH1, MSH2, MSH6, and PMS2 genes often lead to dMMR. Therefore, MSI is also considered a marker of dMMR. MSI has been observed among 27 tumor types [21]. For instance, MSI has been considered a primary predictive marker for the responses to ICIs in CRC, including nivolumab or pembrolizumab targeting PD-1 [22].
ICIs targeting PD-1/PD-L1 and CTLA-4 have caused a revolution in cancer care by reversing the immunosuppressive tumor microenvironment. Inhibitors of other ICP genes are also underway, such as IDO1 and TIGIT, which also showed correlations with FAM72A in multiple tumors [23][24][25]. Indoleamine 2,3-dioxygenase (IDO) 1, an enzyme converting tryptophan to kynurenine, is an immunosuppressant in TME. Blockade of IDO1 is a promising strategy to reignite antitumor immunity [24]. T Cell Immunoreceptor With Ig And ITIM Domains (TIGIT) belongings to the V-Set and immunoglobulin domain containing (VSIG) family. As a coinhibitory receptor during T-cell activation, TIGIT has received considerable attention as a putative immune checkpoint in TME [25]. The double blocking of the TIGIT and PD-1/PD-L1 pathway showed synergistic antitumor effects in colorectal cancer and ovary cancer by upregulating the effector activity of T cells and NK cells [23].
Furthermore, we found that FAM72A low tumors exhibited a significantly higher IPS than the FAM72A high tumors. Moreover, the FAM72A low group was more likely to benefit from anti-PD1/PDL1/PDL2, anti-CTLA4, and combined immunotherapies in various cancers. The IPS is the most comprehensive estimator of tumor immunogenicity. It is calculated by incorporating four key tumor immunogenicity determining factors, including effector cells, checkpoints/immunomodulators, antigen processing major histocompatibility complex (MHC) molecules (e.g., HLAs), and immunosuppressive cells. The immunophenoscore has shown remarkable performance in terms of predicting response to immunotherapies blocking CTLA and PD1 [8]. Overall, these results indicated that FAM72A might be related to the response to ICP therapies.

Functional Analysis
Finally, we performed in vitro experiments to verify the function of FAM72A in lung cancer. The overexpression of FAM72A promoted proliferation and colony formation and also facilitated the EMT and migration of lung cancer cells. The functional analysis of FAM72A is currently very limited. Guo et al. first identified FAM72A and demonstrated its overexpression in several commonly diagnosed malignancies [1]. They also found that FAM72A interacted with UNG2, a crucial component in the base excision repair pathway, and thereby proposed that FAM72A might participate in tumorigenesis via the BER pathway [1]. Wang et al. reported that enforced expression of FAM72A relieved H 2 O 2 -induced reactive oxygen species production and mitochondria membrane potential (∆ψ) loss in nasopharyngeal carcinoma (NPC) cell lines. FAM72A also stimulated cell cycle progression in the NPC cells [26]. On the contrary, FAM72A depletion in the NPC tumor cell lines significantly reduced the cell population at the G1/S phase but increased the number of cells in the multiploid phase [26]. Recently, FAM72A was shown to induce error-prone DNA repair and assist in mutagenic repair in the process of antibody maturation [2,3]. These results suggested that FAM72A may be a new therapeutic target in cancer. Moreover, Zhou et al. observed that FAM72A depletion suppressed proliferation and inactivated the mTOR signaling in HCC cell lines [15].
Our results are largely dependent on analyzing public databases. Due to the striking advances in high-throughput technology, tons of big data have been generated, coupled with the exciting development of machine learning algorithms. Online databases are mainly based on big data, specific algorithms, and programming languages, which are easy to operate and have a high degree of visualization, rich functions, and fast updating speed. Big data lead to a revolution in various scientific communities, such as life science. For instance, The Cancer Genome Atlas (TCGA) unveiled the genomic landscape of human malignancy. The findings derived from mining data of TCGA projects have dramatically accelerated the understanding of tumor biology, the discovery of predictive biomarkers for early diagnosis and prognosis, as well as the identification of therapeutic targets.
It should be noted that results derived from public data and statistical algorithms may unavoidablely suffer from the significant level of heterogeneity and the possible effect of bias. Findings from a bioinformatics analysis should be interpreted cautiously and cannot be translated directly into the clinical setting. However, they are informative and suggestive for researchers in the field. Our findings can be used as a source for other researchers interested in the role of FAM72A in cancers and to attract more attention to this potentially functionally important molecule. After all, we also performed a functional analysis to validate the oncogenic role of FAM72A on lung adenocarcinoma.
In conclusion, our findings indicated that FAM72A might be a potential biomarker predicting prognosis and tumor immunogenicity in different tumors. High

Data Collection
RNA sequencing data for 33 types of cancers were obtained from The Cancer Genome Atlas (TCGA) Data Portal (https://portal.gdc.cancer.gov/, accessed on 15 October 2022) [27]. Accordingly, we acquired clinical data for patients from the UCSC Xena website (https: //xena.ucsc.edu/, accessed on 17 October 2022), including OS, DSS, DFI, and PFI [28]. First, the expression levels of FAM72A in normal and tumor tissues were compared using the Wilcoxon rank sum test function in the "ggplot2" R package. Moreover, an online tool, the Gene Expression Profiling Interactive Analysis (GEPIA) (http://gepia.cancer-pku.cn/, accessed on 19 October 2022), was adopted to compare FAM72A expression between tumorous and normal tissues for several tumors lacking enough normal tissues [29]. GEPIA comprised gene expression information from TCGA and gene expression profiles for normal tissues from the Genotype-Tissue Expression (GTEx) database (https: //www.genome.gov/Funded-Programs-Projects/Genotype-Tissue-Expression-Project, accessed on 22 October 2022). FAM72A expression was determined for stages I + II and stage III + IV tumors to explore whether its expression increased with tumor progression [29].

Estimation of FAM72A Activity in Patient Samples
We used a metagene approach to assess FAM72A activity in patient samples, as previously published [30][31][32]. A Spearman correlation test was performed to select the top 100 genes mostly correlated with FAM72A through the TCGA datasets, which were used to represent the activity of FAM72A. The ssGSEA was utilized to evaluate FAM72A activity in each sample. The analyses were executed using the "gene set variation analysis (GSVA)" and "GSEABase" R packages based on 100 genes.

Prognostic Analysis
The clinical endpoints selected for this study included OS, DSS, PFI, and DFI. OS was referred to as the time from diagnosis to death, regardless of causes. Unlike OS, DSS does not count patients who died from causes aside from a specific disease. PFI is a period free of disease progression or death from any cause. Patients who died from causes except for a specific disease are excluded from DFI. Patients were divided into FAM72A low and FAM72A high groups using the minimum p-value method. A Kaplan-Meier survival analysis was performed to compare the clinical outcomes of patients with FAM72A low or FAM72A high tumors with the application of "survival" and "survminer" R packages. We also implemented univariate Cox regression analysis to estimate the prognostic value of FAM72A by calculating the hazard ratio (HR) and 95% confidence interval (CI) by executing the "survival" and "foresplot" R packages [33].

FAM72A's Capacity to Distinguish Tumor from Non-Tumor Tissues
A ROC analysis of FAM72A expression levels was performed using the "pROC" R package to examine whether FAM72A expression levels can separate tumors and normal tissue across the 33 types of cancer, and the area under the curve (AUC) was calculated [34]. An AUC value greater than 0.7 is considered acceptable.

Implication of FAM72A Expression in Tumor Immune Microenvironment
The ESTIMATE algorithm [18] was employed to determine the immune score and stromal score with the application of the "ESTIMATE" package in R software. The correlation between FAM72A expression and immune score or stromal score was interrogated in 33 tumors using the Spearman method (p < 0.05 and R > 0.1).
We further dissected the tumor-infiltrating immune cells that constitute a crucial component of tumor tissues. Accumulating evidence demonstrated their clinicopathologic significance in predicting prognosis and therapeutic efficiency [35]. Therefore, we explored the relevance between FAM72A and immune infiltrates in this study. The CIBERSORT algorithm was applied to assess the levels of 22 infiltrating immune cell subtypes. These cells included naive B cells, memory B cells, plasma cells, CD8 T cells, naive CD4 T cells, resting memory CD4 T cells, memory-activated CD4 T cells, follicular helper T cells, T regulatory cells (Tregs), gamma delta T cells, resting NK cells, activated NK cells, monocytes, M0 macrophages, M1 macrophages, M2 macrophages, resting dendritic cells, activated dendritic cells, resting mast cells, activated mast cells, and eosinophils [36,37]. The R packages "limma" and "CIBERSORT" were used. The Spearman method was performed to determine the correlation of FAM72A with 22 immune cell subsets in pan-cancer, and the R package "ggplot2" was used to generate the correlation matrix heatmap.

Correlation Analysis of FAM72A with TMB, MSI, Checkpoint Genes, and Immunophenoscore
TMB measures the mutation number in a specific cancer genome. Numerous studies have explored the significance of using TMB as a biomarker for identifying patients sensitive to checkpoint inhibitors [38]. We downloaded the somatic mutation data for all TCGA patients (https://tcga.xenahubs.net, accessed on 23 October 2022) and calculated TMB scores for each sample using the "maftools" R package. Microsatellite instability (MSI) is featured by the widespread length polymorphisms of microsatellite sequences resulting from DNA polymerase slippage [39]. MSI is used as an indicator of genetic instability for the cancer detection index, and patients with high-MSI cancers have been shown to benefit from immunotherapy [40]. The MSI scores of 33 tumors were obtained from the published literature [41]. Spearman's rank method was used to determine the correlation of FAM72A with TMB and MSI. The correlation results for TMB and MSI were visualized in radar maps.
Charoentong and colleagues calculated the immunophenoscore (IPS) for 20 types of cancer using the TCGA database and deposited them in The Cancer Immunome Atlas (TCIA, https://tcia.at/home, accessed on 24 October 2022) [8]. The IPS varies from 0 to 10, with 0 and 10 indicating the lowest and highest degree of immunogenicity. We made comparisons of IPS for the FAM72A low and FAM72A high groups and subsequently evaluated their responses to anti-PD1/PDL1/PDL2 and anti-CLA4 treatments.

Western Blotting
Cells were harvested, washed twice with phosphate-buffered saline (PBS, Procell, Wuhan, China), and lysed at 4 • C with RIPA (Solarbio, Beijing, China) buffer supplemented with a proteinase inhibitor cocktail (Solarbio, Beijing, China) for 20 min. Equal amounts of proteins were electrophoresed in SDS-PAGE (15%) and transferred to PVDF membranes. After blocking with 5% skimmed milk in PBST at room temperature for 1 h, the membranes were incubated overnight at 4 • C with primary antibodies against FAM72A (1:1000, Proteintech) and GAPDH (1:1000, Absin), as needed. After washing with PBST, the membranes were incubated with a secondary antibody (1:1000, Beijing Zhongshan Golden Bridge Biotechnology Co. Ltd., Beijing, China) for 1 h at room temperature. Finally, an ECL detection system (Tanon, Shanghai) was used to detect targeted protein bands. GAPDH was used as an internal control.

Cell Proliferation
We seeded 3 × 10 3 cells into each well of 96-well plates with the addition of 100 µL of culture medium. Cells grew in an incubator under standard conditions, and a CCK-8 assay (Abbkine Scientific Co, Wuhan, China) was performed at 24 h, 48 h, and 72 h after initial attachment. The OD values were measured at 450 nm with a microplate reader.

Cell Migration Assay
Cell migration was assessed using 8-µm-pore Transwell compartments (Corning Inc, Corning NY, USA). Cell suspensions (3 × 10 4 cells) in serum-free medium were added to the upper compartment. After cells were incubated at 37 • C for 24 h, the translocated cells were fixed by 4% paraformaldehyde for 20 min and then stained with 0.5% crystal violet for 20 min at room temperature. Cells were counted under a light microscope (Nikon, Tokyo, Japan).

Wound-Healing Assay
We also evaluated the effects of FAM72A on cell migration using the wound-healing assay. Briefly, a wound was generated in a 6-well plate by scratching the surface with a 200 µL pipette tip. The wounded areas were photographed under a light microscope (Nikon, Tokyo, Japan) when the wound was created (0 h) and 24 h later. The percentage of wound healing was calculated using the following formula: [1 − (empty area 24 h/empty area 0 h)] × 100%.

Statistical Analysis
All bioinformatic analyses were carried out with the R software version 4.1.3 (www.rproject.org, accessed on 15 October 2022). The ESTIMATE and CIBERSORT algorithms were adopted to assess tumor immune environments. Immunophenoscore (IPS) was utilized to estimate tumor immunogenicity and response to immune checkpoint inhibitors (ICIs). The Spearman method was conducted to determine the correlation of FAM72A with the immune score, infiltrating immune cells, tumor mutation burden (TMB), microsatellite instability (MSI), checkpoint genes, and IPS. R packages (limma, ggplot2, CIBERSORT, ESTIMATE, GSVA, CIBERSORT, org.Hs.eg.db, clusterProfiler, DOSE, and enrichplot) were implemented to visualize the results. Statistics for the CCK8, wound healing, and Transwell migration assays were analyzed using Graphpad Prism 9.0 software (GraphPad, La Jolla, CA, USA). Differences in proliferation and migration between vector and experimental groups were checked using the Student's t-test. The significance threshold was set as 0.05 for all statistical analyses.