Abstract

Background

   Despite progression in its treatment, the clinical outcome of patients
   with clear cell renal cell carcinoma (ccRCC) remains not ideal. Anoikis
   is a unique form of programmed apoptosis, owing to insufficient
   cell-matrix interactions. Anoikis plays a crucial role in tumor
   migration and invasion, and tumor cells could protect themselves
   through the capacity of anoikis resistance.

Methods

   Anoikis-related genes (ARGs) were obtained from Genecards and
   Harmonizome portals. The ARGs related to ccRCC prognosis were
   identified through univariate Cox regression analysis, then we utilized
   these ARGs to construct a novel prognostic model for ccRCC patients.
   Moreover, we explored the expression profile of ARGs in ccRCC using the
   Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression (GTEx)
   database. We also conducted Real-Time Polymerase Chain Reaction
   (RT-PCR) to probe ARGs expression of the risk score. Finally, we
   performed correlation analysis between ARGs and tumor immune
   microenvironment.

Results

   We identified 17 ARGs associated with ccRCC survival, from which 7
   genes were chosen to construct a prognostic model. The prognostic model
   was verified as an independent prognostic indicator. The expression of
   most ARGs was higher in ccRCC samples. These ARGs were closely
   correlated with immune cell infiltration and immune checkpoint members,
   and had independent prognostic value respectively. Functional
   enrichment analysis demonstrated that these ARGs were significantly
   associated with multiple types of malignances.

Conclusion

   The prognostic signature was identified to be highly efficient in
   predicting ccRCC prognosis, and these ARGs were closely related to
   tumor microenvironment.

   Keywords: anoikis, clear cell renal cell carcinoma, prognostic model,
   tumor microenvironment, bioinformatics

Introduction

   Renal cell carcinoma (RCC) is one of the predominant tumors in
   urological system, and there is an increasing number of new patients
   with RCC year by year ([41]1, [42]2). Clear cell RCC (ccRCC) is the
   most frequent pathological subtype of RCC, with a ratio of 75-80%
   ([43]3). The clinical prognosis of RCC patients is closely associated
   with disease pathological stage, and relapse-free survival rate is
   decreasing from stage I to stage IV ([44]4). Early surgical resection
   is considered as the predominant treatment for ccRCC patients ([45]5).
   However, about 40% of patients with surgical treatment occur distant
   metastases, contributing to poor endpoint ([46]4). For patients with
   metastatic or unresectable RCC, targeted therapy drugs were
   demonstrated to prolong survival time effectively, including sorafenib
   and sunitinib ([47]6, [48]7). However, various of adverse effects
   decrease the life quality of patients during target treatment, such as
   hypertension ([49]8). Thus, there is an urgent need to perform early
   diagnosis of ccRCC, finding novel diagnosis biomarkers and prognostic
   signatures have great clinical benefit.

   Epithelial cells in human body keep normal tissue structure by
   cell-to-cell adherence and cell-to-extracellular matrix interactions
   ([50]9). Four cell adhesion protein molecules are involved in these
   interactions, including: cadherins, immunoglobulins, selectins, and
   integrins. Anoikis is a kind of specific programmed apoptosis which
   occurs when cell-to-cell adherence and cell-to-extracellular matrix
   interactions are destroyed, and it plays an important role in
   maintaining tissue homeostasis by killing misplaced cells ([51]10).
   Mitochondria- and cell surface death receptor-mediated pathways are
   necessary for triggering anoikis ([52]11). In diverse tumors, anoikis
   serves as a protective physiological barrier inhibiting dissemination
   of metastatic cancer cells. However, cancer cells have the power of
   anoikis resistance to guarantee their survival via different signaling
   pathways, resulting in the distant metastasis ([53]12). Previous
   studies have demonstrated that anoikis-related genes (ARGs) were
   closely correlated with metastatic cascade and progression of cancer.
   As a member of ARGs, the expression of Fas apoptotic inhibitory
   molecule 2 (FAIM2) was positively linked to tumor stage and poor
   endpoint of non-small cell lung cancer, and FAIM2 knockdown may inhibit
   anoikis resistance ([54]13). Kruppel-like factor 5 (KLF5) was verified
   to promote anoikis resistance, and elevated KLF5 expression was
   correlated with poor clinical outcome in colorectal cancer patients
   ([55]14).

   Past studies have validated that several ARGs were involved in
   reversing anoikis resistance and invasion and migration in RCC ([56]15,
   [57]16). Nevertheless, few studies have systematically explored the
   role of ARGs in ccRCC. In our study, multiple databases are integrated
   to screen seven ARGs. Then, the anoikis prognostic model (APM) based on
   these seven ARGs was constructed and validated. Finally, we performed
   several analyses to explore the expression profile of these ARGs and
   the correlation between immune cell infiltration and ARGs expression in
   ccRCC.

Materials and methods

Data collection and processing

   We downloaded the [58]GSE53757 (72 normal renal samples and 72 ccRCC
   samples), [59]GSE40435 (101 normal renal samples and 101 ccRCC
   samples), and [60]GSE66272 (27 normal renal samples and 27 ccRCC
   samples) datasets from the GEO database
   ([61]https://www.ncbi.nlm.nih.gov/geo/, accessed on 12 October 2022).
   Then we obtained 316 anoikis-related genes from the genecards website
   ([62]17) ([63]https://www.genecards.org/, accessed on 12 October 2022)
   and Harmonizome portals ([64]18)
   ([65]https://maayanlab.cloud/Harmonizome/, accessed on 12 October
   2022). The above data analysis was performed by R software. The ARGs
   expression data was normalized via the “edgeR” package ([66]19). We
   obtained the differentially expressed genes (DEGs) between ccRCC
   tissues and normal kidney samples in the [67]GSE53757, [68]GSE40435,
   [69]GSE66272, and KIRC datasets by “limma” package. Furthermore, we
   collected the tumor prognostic genes (TPGs) via “survival” package, and
   overall survival (OS) of these genes was verified using univariate Cox
   analysis ([70]20).

Construction and validation of the APM

   We utilized the least absolute shrinkage and selection operator (LASSO)
   Cox regression analysis to screen out ARGs which were closely
   associated with prognosis of ccRCC patients, and optimal lambda (λ) was
   seen as the optimal value via cross validation ([71]21, [72]22). Then
   we computed the risk score by integrating the expression profile of
   ARGs and paired multivariate Cox regression values (β). We divided the
   ccRCC patients into two groups (low-risk group and high-risk group)
   based on the media risk score. The “umap” package and “Rtsne” package
   were used to demonstrate the distribution patterns of ccRCC patients.
   We utilized the “survminer” package to perform Kaplan-Meier (K-M)
   survival analysis. The “time” ROC package was used to perform
   time-dependent receiver operating characteristic (ROC) analysis.

Predictive role of the APM

   A heatmap was employed to demonstrate the APM and clinical
   characteristics including age, grade, gender, and stage of patients. We
   used the univariate and multivariate Cox regression analyses to
   evaluate the predictive value of the APM and several clinical factors.

Construction and validation of the nomogram

   The “survival” package and “rms” package were employed to build a
   nomogram predicting prognosis based on the Cox regression analysis
   results. We used the calibration curve to estimate the predictive
   effectiveness of the nomogram. The “time” ROC package was utilized to
   perform the time-dependent ROC analysis which could estimate the
   precision of the APM.

Gene expression analysis of ARGs

   In our study, we utilized R (version3.6.3) to analyze the mRNA
   expression levels of the ARGs in tumor tissues and paracancerous tissue
   samples in The Cancer Genome Atlas (TCGA) and normal kidney tissues in
   Genotype-Tissue Expression (GTEx) database which were downloaded from
   UCSC Xena website ([73]https://xena.ucsc.edu/) ([74]23). The expression
   profile was visualized with ggplot2 (3.3.3). Then we evaluated the
   statistical significance of the differential expression by using the
   Wilcoxon test. The total protein expression levels of ARGs were
   collected from Clinical Proteomic Tumor Analysis Consortium (CPTAC)
   dataset in the UALCAN website
   ([75]http://ualcan.path.uab.edu/analysis-prot.html), a convenient tool
   for exploring cancer omics data ([76]24).

Immune infiltration analysis

   R package “estimate” was employed to investigate the abundance of
   stromal components (StromalScore), immune cell infiltration levels
   (ImmuneScore) and ESTIMATEScore of ccRCC. The associations between ARGs
   expression and immune/stromal/estimate scores were evaluated by
   Spearman’s correlation analysis. The CIBERSORT algorithm was utilized
   to conduct the immune cell infiltration estimations ([77]25). The
   partial correlation (cor) values and p values were calculated via the
   purity-adjusted Spearman’s rank correlation test.

Functional and pathway enrichment analysis

   With the help of GeneMANIA website ([78]http://www.genemania.org), we
   explored and displayed the interaction network among these seven ARGs.
   We used WebGestalt to perform Gene Ontology (GO) and Kyoto Encyclopedia
   of Genes and Genomes (KEGG) analysis about ARGs in ccRCC.

Cell culture

   We obtained three human ccRCC cell lines 786-O, Caki-1 and ACHN and
   human renal proximal convoluted tubule cell HK2 from American Type
   Cultural Collection (ATCC). 786-O and Caki-1 were cultured in RPMI 1640
   media (11875119, Gibco, Foster, CA, USA). HK2 and ACHN were kept in
   DMEM high glucose media (11965092, Gibco, Foster, CA, USA). All media
   contained 1% penicillin-streptomycin (15140163, Gibco, Foster, CA, USA)
   solution and 10% fetal bovine serum (FBS) (10100147, Gibco, Foster, CA,
   USA). All cells were maintained in an incubator at 37°C with a 5%
   CO[2].

Reverse transcription and PCR analysis

   Total RNA from cells was extracted by using the Spark easy cellular RNA
   extraction kit (AC0205) (Shandong Sparkjade Biotechnology Co., Ltd).
   The Evo M-MLV RT Kit (AG11711) was obtained from Accurate Biotechnology
   (Hunan) Co., Ltd. The Hieff^®qPCR SYBR Green Master Mix (11201ES08)
   were obtained from Yeasen Biotechnology (Shanghai) Co., Ltd. (Shanghai,
   China). The primer sequences of RT-qPCR were displayed in [79]Table S1.

Statistical analysis

   All data were demonstrated by mean ± standard deviation. Statistical
   analysis was conducted by using SPSS 13. The value of p < 0.05 was
   considered significant (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p
   < 0.0001). The “ns” represented no statistically significant.

Results

Identification of the ARGs

   In our study, a total of 935 DEGs were identified from three GEO
   databases, including [80]GSE40435, [81]GSE53757, and [82]GSE66272. Then
   we obtained 37 significant DEGs from GEO dataset, TCGA database, and
   ARGs. In addition, we identified 168 TPGs based on ARGs and TPGs
   database. Finally, we integrated the 37 significant DEGs and 168 TPGs
   and screened out 17 genes ([83]Figure 1A). In these genes, 12 genes
   expression were upregulated in ccRCC tissues compared with normal
   kidney samples in TCGA and GTEx, including UBE2C, PYCARD, TIMP1, CEBPB,
   MMP9, LGALS1, PECAM1, CSPG4, COL4A2, BNIP3, CCND1, TLR3. While 5 genes
   were downregulated in tumor tissues, including PLAT, ERBB2, CEACAM1,
   NOX4, and UCHL1 ([84]Figure 1B). Univariate Cox regression analysis
   demonstrated that 17 ARGs were closely linked to clinical outcome of
   ccRCC patients ([85]Figure 1C).

Figure 1.

   [86]Figure 1
   [87]Open in a new tab

   Identification of ARGs. (A) Identifying differentially expressed ARGs
   associated with OS via Venn diagram. (B) Heatmap of the expression of
   17 ARGs in KIRC dataset of TCGA and GTEx. (C) Forest plot displaying
   univariate Cox regression analysis of 17 ARGs in ccRCC patients.

Construction of the APM

   LASSO regression analysis was employed to identify seven
   prognosis-related genes including BNIP3, CCND1, CEBPB, ERBB2, PECAM1,
   TIMP1, and UBE2C from 17 ARGs ([88]Figure 2A). By weighting the
   expression of these seven genes with LASSO regression coefficient, we
   constructed the APM based on the following algorithm. Risk score =
   (-0.08912 * BNIP3 expression) + (-0.00457 * CCND1 expression) +
   (0.04290 * CEBPB expression) + (-0.05204 * ERBB2 expression) +
   (-0.25383 * PECAM1 expression) + (0.16386 * TIMP1 expression) +
   (0.16675 * UBE2C expression). We firstly checked the clinical
   information and gene expression profiles of the ccRCC patients in TCGA,
   13 cases without gene expression and unabridged clinical data were
   deleted. Then 526 patients were grouped into a high-risk cluster (n =
   263) and a low-risk cluster (n = 263) according to the media risk
   score. The survival time of patients in high-risk group was shorter
   than patients in low-risk group ([89]Figure 2B). The UMAP and t-SNE
   analysis demonstrated that the distribution of ccRCC patients in two
   groups was in line with the expectation ([90]Figure 2C). The K-M curve
   indicated that the OS of high-risk group was lower than that of
   low-risk group ([91]Figure 2D). ROC analysis was employed to evaluate
   the predictive performance of the APM, and the area under curve (AUC)
   values in 1-year, 3-year, 5-year were 0.735, 0.702, and 0.705,
   respectively, suggesting that the APM could be an ideal prognosis
   indicator ([92]Figure 2E).

Figure 2.

   [93]Figure 2
   [94]Open in a new tab

   Construction of the risk score model in TCGA. (A) LASSO regression
   analysis and LASSO coefficient profiles of ARGs. (B) Risk score
   distribution and ccRCC patients’ survival status. (C) UMAP plot and
   t-SNE plot of the ccRCC patients exhibiting the distribution in high
   and low-risk groups. (D) K-M curves of patients in high and low-risk
   groups. (E) AUC of the time-dependent ROC curves.

Independent prognostic value of the APM and construction of the nomogram

   The heatmap demonstrated the correlation between risk group, stage,
   grade, gender, age, status, and expression levels of seven ARGs
   ([95]Figure 3A). A higher risk score was found to be associated with
   higher expression levels of UBE2C, CEBPB, TIMP1, and lower expression
   levels of ERBB2, BNIP3, CCND1, PECAM1. A univariate Cox regression
   analysis indicated that age, cancer grade, risk score and tumor stage
   were closely correlated with OS ([96]Figure 3B). Meanwhile, a
   multivariate Cox regression analysis displayed that risk score, age,
   and tumor stage were significantly related to OS ([97]Figure 3C). These
   results indicated that the APM had independent prognostic value in
   ccRCC patients. Next, we constructed a nomogram about age, stage, and
   risk score to compute the survival probability rates ([98]Figure 3D).
   The 1-, 3-, and 5-year AUCs for the nomogram were 0.867, 0.810, and
   0.777, respectively ([99]Figure 3E). Furthermore, the calibration plot
   also demonstrated the ideal prediction effect at 1-, 3-, and 5-year OS
   ([100]Figure 3F).

Figure 3.

   [101]Figure 3
   [102]Open in a new tab

   The risk score model could be an independent prognostic indicator for
   ccRCC. (A) Heatmap demonstrating the association between ARGs
   expression and clinical factors and risk groups. (B) Univariate Cox
   regression analysis of ccRCC patients in TCGA. (C) Multivariate Cox
   regression analysis of ccRCC patients in TCGA. (D) Nomogram-based gene
   signature for predicting 1-, 3- and 5-year OS. ROC curves (E) and
   calibration curves (F) for predicting the 1-, 3-, and 5-year
   performance of the nomogram.

Validation of the APM in E-MTAB-1980 cohort

   The clinical feature and gene expression profile of the patients in
   E-MATB-1980 cohort were checked, and patients with abridged information
   were deleted from the cohort. Then we chose 92 cases to verify the
   prognostic value of the APM, which were divided into a low-risk group
   (n = 33) and a high-risk group (n = 59) based on the media risk score.
   [103]Figure 4A demonstrated that most dead patients were distributed
   into the high-risk group, suggesting the ideal predictive value of the
   risk score. UMAP and t-SNE analysis displayed that the ccRCC patients
   of two groups could be placed into two sets suitably ([104]Figure 4B).
   The K-M curve indicated that the high-risk group patients had a worse
   clinical outcome, and the AUC values in 1-year, 3-year, 5-year were
   0.787, 0.761, and 0.762, respectively ([105]Figures 4C, D). The heatmap
   demonstrated the correlation between expression levels of these seven
   ARGs and risk group, clinical information, and pathological
   characteristic ([106]Figure 4E). The univariate Cox regression analysis
   showed that stage and risk score were closely related to OS
   ([107]Figure 4F). Meanwhile, the age, grade, stage, and risk score had
   an intimate relationship with OS in multivariate Cox regression
   ([108]Figure 4G). These results further verified the independent
   prognostic value of the APM.

Figure 4.

   [109]Figure 4
   [110]Open in a new tab

   Validation of the risk score model. (A) Risk score distribution and
   ccRCC patients’ survival status. (B) UMAP plot and t-SNE plot of the
   ccRCC patients exhibiting the distribution in high and low-risk groups.
   (C) K-M curves of patients in high and low-risk groups. (D) AUC of the
   time-dependent ROC curves. (E) Heatmap demonstrating the association
   between ARGs expression and clinical factors and risk groups.
   Univariate Cox regression analysis (F) and multivariate Cox regression
   analysis (G) of ccRCC patients.

Expression profiles of ARGs in ccRCC

   Next, we explored the expression patterns of these seven ARGs in ccRCC
   tissues and paired normal kidney samples based on TCGA dataset. As
   graphed in [111]Figure 5A, the mRNA expression levels of BNIP3, CCND1,
   CEBPB, PECAM1, TIMP1, and UBE2C were higher than that in normal kidney
   tissues, while ERBB2 was downregulated in ccRCC. The protein expression
   levels of BNIP3, CCND1, PECAM1, TIMP1, and UBE2C were elevated in
   ccRCC, while CEBPB and ERBB2 protein expression were downregulated in
   ccRCC ([112]Figure 5B).

Figure 5.

   [113]Figure 5
   [114]Open in a new tab

   ARGs expression profiles in ccRCC and normal tissues. (A) Expression
   levels of ARGs in ccRCC and paired normal tissue samples. The red color
   represents tumor tissue, while the blue color represents normal tissue.
   (B) Total protein levels of ARGs in ccRCC and normal tissue samples
   from CPTAC. ** p < 0.01, *** p < 0.001, **** p < 0.0001. ns represents
   no significance.

   Furthermore, we chose human ccRCC cell lines 786-O, Caki-1 and ACHN and
   human renal proximal convoluted tubule cell HK2 to detect the
   expression of ARGs via RT-qPCR experiments. As displayed in
   [115]Figure 6, the mRNA expression levels of CCND1 and CEBPB were
   higher in three ccRCC cell lines than that in HK2, and BNIP3 and PECAM1
   were higher expressed in 786-O and Caki-1 than HK2. We also conducted
   immunohistochemistry (IHC) staining to investigate the UBE2C protein
   level in ccRCC and par-carcinoma tissue. The result demonstrated that
   UBE2C protein level was significantly upregulated in ccRCC compared to
   par-carcinoma tissue, which was consistent with previous result from
   CPTAC ([116]Figure S1).

Figure 6.

   [117]Figure 6
   [118]Open in a new tab

   The mRNA expression of ARGs in HK2 and three ccRCC cell lines
   quantified by real-time PCR. (A) BNIP3; (B) CCND1; (C) CEBPB; (D)
   ERBB2; (E) PECAM1; (F) TIMP1; (G) UBE2C. * p < 0.05, ** p < 0.01, *** p
   < 0.001, **** p < 0.0001. ns represents no significance.

Clinicopathological stages and independent prognostic value of ARGs in ccRCC

   As displayed in [119]Figure S2, the expression levels of these seven
   ARGs were strongly correlated with pathological stages of ccRCC
   patients. Notably, the expression levels BNIP3, CCND1, ERBB2, and
   PECAM1 were downregulated from stage I to stage IV in ccRCC, while
   expression of CEBPB, TIMP1, and UBE2C were elevated in higher stage. To
   further explore the independent prognostic value of single ARGs in
   ccRCC, we performed survival analysis using K-M plotter. The result
   demonstrated that high expression levels of BNIP3, CCND1, ERBB2, and
   PECAM1 were correlated with better clinical endpoint of ccRCC, while
   high CEBPB, TIMP1, and UBE2C expression were linked to worse clinical
   outcome ([120]Figure S3). The K-M result was consistent with previous
   Cox regression analysis.

TME and immune infiltration analysis of ARGs

   An increasing number of evidences have illustrated that tumor
   microenvironment (TME) plays a significant role in tumor occurrence and
   progression ([121]26, [122]27). Therefore, it is essential to dig out
   the relationship between expression of ARGs and TME. In our study, the
   ESTIMATE algorithm was utilized to evaluate the immune and stromal
   scores of ccRCC. [123]Figure 7 demonstrated that the expression of
   CEBPB, PECAM1, TIMP1, and UBE2C were significantly positively
   associated with stromal score, as well as with immune score, while
   ERBB2 expression was negatively correlated with these scores. The
   relationship between BNIP3 and CCND1 had no statistical significance
   ([124]Figure S4). Moreover, we examined the association between
   expression of ARGs and immune cell infiltration level in ccRCC using
   CIBERSORT algorithm. As displayed in [125]Figure 7, the expression of
   BNIP3, CCND1, ERBB2, and PECAM1 were closely negatively related to
   infiltration levels of regulatory T cell, follicular helper T cell, M0
   macrophage, and memory B cell, and were positively correlated with M2
   macrophage and monocyte. CEBPB, TIMP1, and UBE2C expression levels were
   strongly positively associated with infiltration levels of regulatory T
   cell, CD4+ memory activated T cell, and M0 macrophage ([126]Figure 8).

Figure 7.

   [127]Figure 7
   [128]Open in a new tab

   The correlation analysis between StromalScore, ImmuneScore and
   ESTIMATEScore and expression levels of (A) CEBPB; (B) ERBB2; (C)
   PECAM1; (D) TIMP1; (E) UBE2C.

Figure 8.

   [129]Figure 8
   [130]Open in a new tab

   Relationship between immune cells infiltration and ARGs expression
   based on CIBERSORT. The red color indicates a positive correlation,
   while the blue color represents a negative correlation.

   In addition, we evaluated the relationship between expression of ARGs
   and immune subtype in ccRCC, and the result displayed high
   heterogeneity. Immune subtype consisted of six types, including C1
   (wound healing), C2 (IFN-gamma dominant), C3 (inflammatory), C4
   (lymphocyte depleted), C5 (immunologically quiet), C6 (TGF-β dominant)
   ([131]25). BNIP3 and UBE2C were highest expressed in C1, CCND1 and
   PECAM1 were highest expressed in C3, CEBPB and TIMP1 were highest in
   C6, and ERBB2 was highest expressed in C5 ([132]Figure 9A). We also
   evaluated the relationship between ARGs and significant immune
   checkpoint genes including CD274 (PD-L1), CTLA4, PDCD1LG2 (PD-L2), LAG3
   and PDCD1 (PD-1) in ccRCC. As displayed in [133]Figure 9B, the
   expression levels of CEBPB, TIMP1 and UBE2C were positively correlated
   with most immune checkpoint genes except for CD274, indicating that
   they could be potential targets for immunotherapy.

Figure 9.

   [134]Figure 9
   [135]Open in a new tab

   Correlation analysis between ARGs expression levels and immune subtypes
   and immune checkpoints in ccRCC. (A) ARGs expression in different
   immune subtypes in ccRCC. C1, wound healing; C2, IFN-g dominant; C3,
   inflammatory; C4, lymphocyte depleted; C5, immunologically quiet; C6,
   TGF-β dominant. (B) Relationship between ARGs expression and main
   immune checkpoint genes. The red color indicates a positive
   association, while the blue color represents a negative association.

Correlation and enrichment analysis of ARGs

   The correlation analysis of ARGs was conducted in ccRCC, and result was
   displayed in [136]Figure 10A. GeneMANIA is an online site which have
   abundant genetic information and can analyze the correlation among
   several genes ([137]28). Then, the GeneMANIA website was utilized to
   explore the several potential interaction partners of ARGs. As showed
   in [138]Figure 10B, the expression of ARGs were closely correlated with
   EARS2, IL3RA, ENC1 and so on. We next employed GO and KEGG analysis to
   investigate the biological process, cellular component, molecular
   function and related signaling pathways of seven ARGs. GO analysis
   indicated that biological regulation and metabolic process were
   included in biological process, and protein binding was involved in
   molecular function ([139]Figure 10C). The result of KEGG analysis
   demonstrated that ARGs were involved in occurrence and development of
   multiple types of tumors, endocrine resistance, and HIF-1 signaling
   pathway ([140]Figure 10D).

Figure 10.

   [141]Figure 10
   [142]Open in a new tab

   Enrichment analysis of ARGs and related partners. (A) The correlation
   analysis between the ARGs. Blue color represents a negative correlation
   and red color represents a positive correlation. (B) Protein–protein
   interaction network of the ARGs and relevant genes by GeneMANIA tool.
   (C) GO analysis of related biological processes, cellular components,
   and molecular functions of ARGs. (D) KEGG enrichment analysis of ARGs.

   Then, the single cell dataset KIRC_GSE121636 in the TISCH database was
   employed to explore the expression of ARGs in TME. In the
   [143]GSE121636 dataset, CEBPB, PECAM1 and TIMP1 were mainly expressed
   in monocyte, while ERBB2 was almost not detected in immune cell
   ([144]Figure S5).

Discussion

   ccRCC was considered as the most common pathological subtype of
   urologic malignances ([145]29). Several approaches of treating RCC have
   been employed in clinical work including surgery, radiofrequency
   ablation, immunotherapy, chemotherapy, and targeted therapy. After
   radical or partial nephrectomy, plenty of ccRCC patients still develop
   distant metastases, contributing to higher mortality ([146]30). No
   enough prognostic indicator or signature has been verified for the
   clinical outcome of ccRCC patients. Although doctors have made a
   remarkable progress in treating ccRCC, the endpoint of advanced and
   metastatic patients is still not ideal ([147]31). The 3-year survival
   rate of patients with lymph node metastasis is 20-30% after surgery.
   Targeted therapy including sunitinib has been utilized to treat
   advanced RCC patients, while large number of patients could acquire
   drug resistance within 6-15 months, and the OS remains unsatisfactory
   ([148]32–[149]34). Therefore, it is necessary to explore novel
   biomarker as the prognostic indicator or therapy target of ccRCC.

   Anoikis is a significant defense mechanism, restraining the readhesion
   of shedding cells to an inappropriate matrix and inhibiting their
   aberrant growth ([150]35). Previous studies have illustrated that the
   occurrence of anoikis relied on extrinsic pathway and intrinsic pathway
   ([151]36). Anoikis plays a significant role in tumor metastasis and is
   closely correlated with tumor development and therapy failure
   ([152]37). Malignance cells deprive from primary tissue and adapt to
   abnormal growth elsewhere, contributing to tumor migration. Two main
   strategies have been demonstrated to achieve anoikis resistance in
   tumor, including the regulation of the cell adhesion molecules
   expression and employment of oxidative stress ([153]38). Nevertheless,
   the accurate mechanisms of anoikis resistance in ccRCC are still need
   for further study. Past evidence has indicated that anoikis-resistant
   ccRCC cells displayed insensitiveness to detachment-induced apoptosis
   and uncontrolled proliferation. Meanwhile, the expression of Tyrosine
   receptor kinase B (TrkB) in them was higher than parental ccRCC cells
   ([154]39). Quinazoline-based drugs were described to induce anoikis in
   RCC through activating focal adhesion survival signaling pathway, which
   could be a novel therapy method of RCC ([155]40). Moreover, an
   extracellular matrix deprivation system (EDS) was developed to reverse
   anoikis resistance of RCC cells via downregulating FAK phosphorylation
   ([156]15).

   In this study, we constructed and verified the APM containing seven
   ARGs, namely BNIP3, CCND1, CEBPB, ERBB2, PECAM1, TIMP1, and UBE2C. An
   increasing number of studies have demonstrated that these genes were
   strongly correlated with occurrence and pathogenesis of cancer. As a
   mitophagy receptor, upregulation of BNIP3 was associated with poorer
   melanoma patient’s endpoint, and BNIP3 knockout in melanoma cells
   inhibited tumor growth in vivo ([157]41). CCND1, a positive cell cycle
   regulator, was frequently deregulated in tumor and was an indicator of
   cancer phenotype and disease development ([158]42). CEBPB was found to
   be involved in aerobic glycolysis and promote growth of breast cancer
   ([159]43). On the other hand, overexpression of MIR503 host gene
   regulated invasion and proliferation of cervical cancer cells and
   accelerated cell apoptosis by miR‐191/CEBPB axis ([160]44). ERBB2 was
   replicated through elevated gene transcription in various of cancer
   cells, and tumor growth depended on ERBB2 amplification heavily
   ([161]45). A series of studies identified that PECAM1 was linked to
   development of several malignances and had the potential to be a
   diagnostic and prognostic biomarker in gastric cancer patients
   ([162]46, [163]47). The knockdown of TIMP1 expression inhibited
   metastasis, and proliferation but prompted apoptosis of colon cancer
   through activating FAK-PI3K/AKT and MAPK pathway ([164]48). And TIMP1
   could be an independent prognostic biomarker of OS and disease-free
   survival for colon cancer patients. As for UBE2C, suppression of UBE2C
   expression restrained E2-promoted invasion, migration and epithelial
   mesenchymal transition in vitro and in vivo, and UBE2C elevation was
   significantly correlated with higher histologic grade, recurrence, and
   worse OS in endometrial cancer ([165]49).

   The Infiltration of Treg cells was often correlated with worse
   prognosis in diverse tumors, and decreased Treg cells was found to
   evoke and promote anti-tumor immune response ([166]50). In our study,
   the expression levels of BNIP3, CCND1, ERBB2, and PECAM1 were
   significantly negatively correlated with infiltration level of Treg
   cells, and CEBPB, TIMP1, and UBE2C were positively associated with
   infiltration level of Treg cells, suggesting that these genes might
   affect ccRCC patients’ prognosis through regulating infiltration of
   Treg cells. To explore the potential mechanisms or functions of the
   ARGs, functional enrichment analysis was conducted. KEGG analysis
   demonstrated that the seven ARGs were mainly enriched in several types
   of malignances, including bladder cancer, endometrial cancer, non-small
   cell lung cancer, pancreatic cancer, and prostate cancer, indicating
   the intimate relationship between ARGs and different types of tumors.

   In summary, we systematically constructed and assessed the APM of ccRCC
   based on 7 ARGs, and this model could well predict the OS of ccRCC
   patients. Moreover, we found that these 7 ARGs were closely correlated
   with immune cell infiltration and immune checkpoint genes. Our study
   could provide insights for further exploring the role of ARGs in ccRCC
   and contribute to the progression of personalized and precise therapy
   strategies.

Data availability statement

   The datasets presented in this study can be found in online
   repositories. The names of the repository/repositories and accession
   number(s) can be found in the article/[167]Supplementary Material.

Author contributions

   CS and HS designed the overall study. JZ, ZZ, and ZX revised the paper.
   YL and ZS drafted the manuscript and performed the data analysis. YB,
   MF and YD participated in the data collection. All authors contributed
   to the article and approved the submitted version.

Glossary

   Abbreviation                    Full name
   APM          Anoikis prognostic model
   ARGs         Anoikis-related genes
   ATCC         American Type Cultural Collection
   AUC          Area under curve
   BNIP3        Bcl-2/adenovirus E1B 19kDa interacting protein 3
   CCND1        Cyclin D1
   ccRCC        Clear cell renal cell carcinoma
   CEBPB        CCAAT/enhancer-binding protein beta
   CPTAC        Clinical Proteomic Tumor Analysis Consortium
   EARS2        Glutamyl-tRNA synthetase 2
   EDS          Extracellular matrix deprivation system
   ERBB2        Epidermal growth factor receptor 2
   FAIM2        Fas apoptotic inhibitory molecule 2
   FBS          Fetal bovine serum
   GO           Gene Ontology
   GTEx         Genotype-Tissue Expression
   IHC          Immunohistochemistry
   KEGG         Kyoto Encyclopedia of Genes and Genomes
   KLF5         Kruppel-like factor 5
   K-M          Kaplan-Meier
   LASSO        Least absolute shrinkage and selection operator
   OS           Overall survival
   PECAM1       Platelet endothelial cell adhesion molecule-1
   RCC          Renal cell carcinoma
   ROC          Receiver operating characteristic
   TCGA         The Cancer Genome Atlas
   TIMP1        Tissue inhibitor of metalloprotease 1
   TME          Tumor microenvironment
   TPGs         Tumor prognostic genes
   TrKB         Tyrosine receptor kinase B
   UBE2C        Ubiquitin-conjugating enzyme E2C
   [168]Open in a new tab

Funding Statement

   This work was supported by the National Natural Science Foundation of
   China (No.81803044, 81972373), the Natural Science Foundation of Fujian
   Province of China (No.2019J01016), and the funding of Science and
   Technology Planned Project of Medical and Health of Xiamen City
   (No.3502Z20194038).

Conflict of interest

   The authors declare that the research was conducted in the absence of
   any commercial or financial relationships that could be construed as a
   potential conflict of interest.

   The reviewer YZ declared a shared parent affiliation with the authors
   to the handling editor at the time of review.

Publisher’s note

   All claims expressed in this article are solely those of the authors
   and do not necessarily represent those of their affiliated
   organizations, or those of the publisher, the editors and the
   reviewers. Any product that may be evaluated in this article, or claim
   that may be made by its manufacturer, is not guaranteed or endorsed by
   the publisher.

Supplementary material

   The Supplementary Material for this article can be found online at:
   [169]https://www.frontiersin.org/articles/10.3389/fimmu.2023.1171883/fu
   ll#supplementary-material
   [170]Click here for additional data file.^ (3.4MB, docx)
   [171]Click here for additional data file.^ (16.1KB, docx)

References