Upregulation of ARNTL2 Predicts Poor Survival for Clear Cell Renal Cell Carcinoma and Explores its Associations with PD-L1 and Immune In ltration

Song Wang (  11918351@zju.edu.cn ) Zhejiang University School of Medicine First A liated Hospital https://orcid.org/0000-0003-1129-8044 Xueyou Ma Zhejiang University School of Medicine First A liated Hospital Yufan Ying department of urology the rst a liated hospital of zhejiang university Jiazhu Sun Zhejiang University School of Medicine First A liated Hospital Zitong Yang Zhejiang University School of Medicine First A liated Hospital Jiangfeng Li Zhejiang University School of Medicine First A liated Hospital Ke Jin Zhejiang University School of Medicine First A liated Hospital Xiao Wang Zhejiang University School of Medicine First A liated Hospital Bo Xie Zhejiang University School of Medicine First A liated Hospital Xiangyi Zheng Zhejiang University School of Medicine First A liated Hospital Ben Liu Zhejiang University School of Medicine First A liated Hospital Liping Xie Zhejiang University School of Medicine First A liated Hospital

mainstream immunotherapies, such as monoclonal antibodies targeting PD-L1 and/or CTLA4, are becoming the standard paradigm in advanced ccRCC treatment [6]. However, there are still many patients bene t few from immunotherapies since lacking sensitivity to these immunotherapy agents. Therefore, discovering effective prognostic biomarkers and highly speci c tumor immune related therapeutic targets are of paramount importance for the treatment of ccRCC patients.
Aryl hydrocarbon receptor nuclear translocator like 2 (ARNTL2) pertain to the PAS superfamily and can encode a transcription factor, whose protein structure is remarkably similar to hypoxia-inducible factors, such as HIF1alpha, ARNTL2 was reported to play important roles in circadian and hypoxia process [7]. Emerging evidences have demonstrated the carcinogenesis roles of ARNTL2 in human malignancies such as breast carcinoma and colorectal adenocarcinoma [8,9], there are also two recent researches have showed that ARNTL2 correlated with the poor survival and immune in ltration level of lung adenocarcinoma [10,11]. However, the potential roles of ARNTL2 in ccRCC have not been investigated thus far.
In this study, we demonstrated that ARNTL2 was strikingly elevated in ccRCC through integrated analyses of RNA-seq data from TCGA and GEO database and validated in human ccRCC clinical samples, highly expressed ARNTL2 correlated with advanced clinical tumor stage and poor OS. ARNTL2 can also serve as an independent predictor of ccRCC patient survival. The potential biological functions and mechanisms of ARNTL2 were also explored through GSEA analysis. The ssGSEA and xCell algorithm were subsequently combined to analyze the associations between overexpression of ARNTL2 and TIME in ccRCC. Additionally, we also comprehensively investigated the correlations between ARNTL2 and MSI, TMB, PD-L1 expression, TIME and OS of pan-cancer. We sought to explore the comprehensive roles of ARNTL2 in ccRCC and place major emphasis on its correlations with tumor immunity.

Data acquisition and analysis
The transcriptome and clinical data of the ccRCC (also namely KIRC) patients were downloaded from the Genomic Data Commons (GDC) data portal of TCGA database (https://portal.gdc.cancer.gov/), including 539 ccRCC tissues and 72 normal tissues. The data of GSE15641, GSE46699 and GSE53757 datasets from GEO database (https://www.ncbi.nlm.nih.gov/geo/) were downloaded and used to further validate the expression level of ARNTL2. We also downloaded the genetic mutation data from TCGA and visualized using the package "maftools" in R software [12]. In the section involving the pan-cancer analysis of ARNTL2, the transcriptome and clinical data of thirty-three cancer types were also obtained through GDC portal of TCGA, while data of normal tissues were obtained in GTEx database (https://gtexportal.org/home/datasets) V8 version.
Human clinical samples and qRT-PCR analysis Twenty pairs ccRCC and adjacent nontumorous samples were gathered from patients who had received radical nephrectomy (from 2011 to 2013) at rst a liated hospital of Zhejiang University, informed consents were acquired from every patient and this study was approved by Institutional Ethics Committee in the hospital. All patients' clinical information was shown in Supplementary Table 1. Total RNA was extracted with TRIzol agentia (Takara) and with the help of PrimeScript RT Reagent Kit (Takara), RNA was then reverse transcribed into cDNA. The ARNTL2 mRNA relative expression was detected by RT-qPCR, through utilizing the ABI 7500 fast real-time PCR System (Applied Biosystems) and SYBR Premix Ex Taq (Takara). GAPDH was used as endogenous normalization reference to quantify the relative mRNA expression of ARNTL2. All primers are listed: ARNTL2 F 5'-ACTTGGTGCTGGTAGTATTGGA-3'; ARNTL2 R 5'-TGTTGGACTCGAATCATCAAGG-3'; GAPDH F 5'-CTGGGCTA CACTGAGCACC-3'; GAPDH R 5'-AAGTGGTCGTTGAGGGCAATG-3'.
Univariate and multivariate cox regression analysis and the construction of nomogram model Cox regression analyses including univariate and multivariate analysis were employed to identify independent overall survival-related factors, the forest plot was used to show the P value, HR and 95% CI of each variable through 'forest plot' R package. In order to contribute to predict the OS possibility of ccRCC individuals, a nomogram was subsequently developed based upon the results of multivariate Cox proportional hazards through utilizing the 'rms' R package. The Kaplan-Meier plotter analysis (http://kmplot.com/analysis/) and prognoscan analysis based upon GEO database (http://dna00.bio.kyutech.ac.jp/PrognoScan/index.html/) were utilized to externally validate the prognostic values of ARNTL2 in ccRCC and other cancer types, respectively.

Gene set enrichment analysis (GSEA)
To investigate the underlying mechanisms and biological functions of ARNTL2, we used the "Cluster Pro ler" (version: 3.18.0) package to conduct the GSEA analysis [13] to explore the KEGG signaling pathway analysis of ARNTL2 in ccRCC, in the enrichment results, adjust p value <0.01 and q-values<0.01 were recognized as a signi cantly meaningful pathway.

Tumor immune microenvironment analysis
To perform a reliable assessment of immune in ltration, we preliminarily assess the associations of the expression of ARNTL2 and its top three co-expressed genes with various tumor in ltrated immune cell types by using the ssGSEA algorithm of the GSVA package [14]. The xCell algorithm [15] was subsequently employed to calculate the relative proportions of diverse tumor-in ltrating immune cells in every cancer sample, and the results were implemented by R packages "immunedeconv" and "pheatmap", xCell algorithm was also applied to explore the correlations between ARNTL2 and tumor-immune in ltration cell across thirty-three cancer types. To further estimate the relationships of ARNTL2 expression with the tumor immune microenvironment (TIME) in pan-cancer, we also extract the expression values of seven most common immune checkpoint-related genes (PD1, PD-L1, CTLA4, TIM3, LAG3, PD-L2 and TIGIT) and assessed the correlations of their expression with ARNTL2 across 33 cancer types.
The association of ARNTL2 with tumor mutational burden (TMB), microsatellite instability (MSI) In order to investigate the correlations between ARNTL2 expression and TMB/MSI, we downloaded and analyzed the TMB data from the article [16]  Wilcox test and Kruskal-Wallis test were respectively utilized to implement the intergroup comparisons of two and three groups. Spearman correlation analyses were implemented to assess the correlations of ARNTL2 expression with TMB, MSI, mismatch repair genes and PD-L1 expression. For Kaplan-Meier curves, hazard ratio (HR) and p-values were yielded by univariate Cox proportional hazards regression. The predictive model of ARNTL2 was estimated with the receiver operating characteristic curves (ROC) via using pROC package of R software. p < 0.05 denoted statistical signi cance.

Results
Elevated expression levels of ARNTL2 in ccRCC We preliminarily investigated the expression level of ARNTL2 in pan-cancer by analyzing TCGA and GTEx database. ARNTL2 was observed to be upregulated in various cancer types, including ccRCC ( Fig. 1a, b), a paired line graph of 72 pairs of ccRCC samples and matching nontumorous samples indicated that most ccRCC tissues possessed higher ARNTL2 expression compared to matching normal tissues (Fig. 1c). Three independent datasets from GEO database were utilized to externally illustrated the expression of ARNTL2 in ccRCC, which also demonstrate the highly expressed level of ARNTL2 in ccRCC ( Fig. 1d-e), and increased ARNTL2 expression signi cantly associated with advanced ccRCC clinical stage (Fig. 1f, Table 1) and tumor histologic grade (Table 1). Additionally, we demonstrated that most ccRCC tissues possessed higher ARNTL2 expression compared to adjacent nontumorous in 20 pairs ccRCC tissues via qRT-PCR method (Fig. 1g). These results revealed that ARNTL2 was signi cantly elevated in ccRCC. Multivariate cox regression analysis of ARNTL2 and construction of nomogram model We then evaluated the prognostic values of ARNTL2 in ccRCC, as plotted in Fig. 2a and b, elevated expression of ARNTL2 indicated poor OS in both TCGA (HR = 1.67, P = 0.001) and Kaplan-Meier plotter (HR = 1.91, P = 2.2e-05) database. The cox analyses were also applied to explore associations between ARNTL2 expression and OS in ccRCC, univariate analysis indicated that ARNTL2 expression (HR = 1.45853, p = 0.00029), age (HR = 1.02888, p = 1e-05), pathological TNM stage (HR = 1.86653, p < 0.0001) and tumor grade (HR = 2.29073, p < 0.0001) are signi cantly correlated with OS of ccRCC (Fig. 2c). Multivariate analysis, depicted as a forest boxplot in Fig. 2d, revealed that ARNTL2 expression (p = 0.01016), as well as the age (p = 1e-05), pathological TNM stage (p < 0.0001) and tumor grade (p = 0.00083) were independent predictor of ccRCC patient OS. ARNTL2 expression also showed preferable predective ability as the ROC curve exhibited that the AUC of ARNTL2 expression for predicting OS was 0.773 (Fig. 2e). Furthermore, we developed a nomogram model by integrating ARNTL2 and other independent prognostic variables base on the multivariate cox regression analysis results (Fig. 2f), the nomogram can contribute to quantitatively assess ccRCC patients 1-y, 3-y and 5-y survival probability.

Mutation patterns of SMGs in different ARNTL2 expression level of ccRCC
In order to investigate mutation patterns of signi cantly mutated genes (SMGs) in different ARNTL2 expression level of ccRCC, we downloaded and analyzed the genetic mutation data of ccRCC patients from TCGA database. We identi ed 30 SMGs in ccRCC cohort, among these, VHL, PBRM1, TTN, SETD2, BAP1, MUC16 were the top six frequently mutated genes in ccRCC (Fig. 3a). Subsequently, we divided the ccRCC patients with genetic mutation data into two groups based upon the median expression value of ARNTL2, namely ARNTL2 high and ARNTL2 low group. As shown in Fig. 3b-g, we explore the mutation patterns of top six SMGs in two ARNTL2 groups. Signi cantly higher mutation frequency of PBRM1 was observed in ARNTL2 low expression group(P < 0.001) (Fig. 3b), while other SMGs were found no differences in two ARNTL2 groups (P > 0.05) ( Fig. 3c-g).

Gsea Analysis Of Arntl2 In Ccrcc
To investigate the potential functions and mechanisms of ARNTL2 in ccRCC, we utilized GSEA analysis to explore ARNTL2-related pathways in ccRCC carcinogenesis, GSEA analysis contribute to reveal signi cantly enriched KEGG pathways in highly expressed ARNTL2 samples with a high accuracy. The results revealed that renal cell carcinoma (NES = 1.96614, P-adjust < 0.001), focal adhesion (NES = 2.06698, P-adjust < 0.001), Toll-like receptor signaling pathway (NES = 1.965199, P-adjust < 0.001), JAK-STAT signaling pathway (NES = 1.999494, P-adjust < 0.001), T cell receptor signaling pathway (NES = 2.036177, P-adjust < 0.001) and cell cycle pathways (NES = 1.551655, P-adjust < 0.005) ( Fig. 4a-f, Table 2) were signi cantly enriched in upregulated ARNTL2 samples. These results indicated that the immune response and cell cycle related pathways were strongly correlated with abnormal expression of ARNTL2 in ccRCC. Signi cant correlations between the expression of ARNTL2 and its co-expressed genes and tumor in ltrating immune cells in ccRCC In order to further explore the potential functions of ARNTL2 in the ccRCC carcinogenesis, we employed R software "stat" package to identify ARNTL2 positively co-expressed genes, and the strongly co-expressed genes (spearman analysis, r > 0.70, p < 0.001) were selected to further analysis, as shown in Fig. 5a-5b, the actin related protein 2/3 complex subunit 2 (ARPC2), guanine nucleotide binding protein beta polypeptide 4 (GNB4) and capping-protein muscle Z line alpha 1 (CAPZA1) were the only three positively co-expressed genes with spearman coe cient higher than 0.7 (p < 0.001). Subsequently, we applied ssGSEA algorithm in R package GSVA to preliminarily estimate the relationships between the expression of ARNTL2 and its top 3 co-expressed genes and tumor in ltrating immune cell types, as shown in  (Fig. 5d). In summary, these results indicated that ARNTL2 might be participated in the immune response in ccRCC TIME.
The xCell algorithm was further employed to calculate the fraction of diverse tumor-in ltrating immune cells in the TIME of high and low ARNTL2 expression ccRCC samples. Distinct in ltrating levels of immune cells were observed in two ARNTL2 groups (Fig. 6a). With the rise of the ARNTL2 expression, the immune score (P < 0.001), microenvironment score (P < 0.001) and stroma score (P < 0.05) in ccRCC TIME were enhanced (Fig. 6b), high expression level of ARNTL2 tightly linked to the in ltrating levels of CD8 + T cell, CD4 + memory T cell, Myeloid dendritic cell, macrophage and CD4 + Th2 T cell, while CD4 + Th1 T cell, B naive cell and NK T cell were downregulated in highly ARNTL2 expression group (Fig. 6c). Additionally, the immune checkpoint-related genes such as PD-L1, PD-L2, CTLA4, PD-1, TIM3, LAG3 and TIGIT were also tightly related to the expression level of ARNTL2 (Fig. 6d). These results implied that ARNTL2 might be related with T cell exhaustion and immune evasion in ccRCC.
Associations of ARNTL2 with PD-L1, TMB, MSI and mismatch repair genes We observed that in addition to ccRCC, ARNTL2 was also highly expressed in various tumor types compared to their corresponding normal tissues (Fig. 1a), we subsequently desired to assess whether ARNTL2 has a universal role in immune response in pan-cancer. Considering the GSEA results indicated the potential KEGG pathway of ARNTL2 in regulation of T cell receptor signaling pathway (Fig. 4e), we rstly explore the correlations between PD-L1 and ARNTL2 expression in pan-cancer, as shown in Fig. 7a,  (Fig. 7c). Furthermore, we also explore the relationships between ARNTL2 expression and mismatch repair genes such as MLH1, MSH2, MSH6 and PMS2, the results demonstrated that ARNTL2 tightly related with the expression of mismatch repair genes, especially the expression of MSH2 and MSH6 (Fig. 7d). In summary, ARNTL2 expression showed signi cant correlations with the level of three vital prognostic indicators (PD-L1, TMB and MSI) for cancer immunotherapy in various cancer types.

Associations Of Arntl2 With Time In Pan-cancer
To further investigate the in uence of ARNTL2 on TIME of pan-cancer, xCell algorithm was employed to explore the links of ARNTL2 expression with cancer-in ltrating cells in pan-cancer. The results showed that ARNTL2 was tightly related to the enrichment of immune cells, especially the Th2 CD4 + T cell, CD4 + memory T cell, Macrophages and (Fig. 8a), while remarkably correlated with the decrease of NK T cell and CD4 + central memory cell. Immune checkpoint-related genes had been demonstrated to be involved in immune escape in carcinogenesis, among these, PD-1, PD-L1, PD-L2, CTLA4, LAG3, TIM3, TIGIT were the seven most common immune checkpoint related genes, our results also indicated that ARNTL2 correlated with the expression level of these immune escape markers in various cancers, including ccRCC (Fig. 8b). These results further illustrated the vital role of ARNTL2 in the TIME of pan-cancer.

Discussion
Although substantial advancements have been made in therapeutic strategies such as surgical excision, chemotherapy and targeted therapies of ccRCC in recent years, ccRCC still in uences the health and quality of life of many patients around the world [3]. Immunotherapy has emerged as the last chance for advanced ccRCC patients, especially for patients who cannot tolerate chemotherapy, while most patients unable to obtain durable bene ts from currently immunotherapy represented by anti-PD-1 antibodies [18,19]. It is urgent to identify effective biomarkers and immune-related therapeutic targets to improve the patients' survival. In this study, we rstly discovered that the upregulated ARNTL2 correlated with unfavorable OS and immune in ltration in ccRCC, we also comprehensively investigated the associations between ARNTL2 expression and PD-L1 expression, TMB, MSI, immune in ltration and adverse prognosis in pan-cancer.
ARNTL2 is a member of PAS protein superfamily, located on chromosome 12p11, encodes a transcription factor, whose structure similar to HIF1alpha [7]. The potential roles of ARNTL2 has been explored in several human malignancies. Mazzoccoli et al. discovered that ARNTL2 was upregulated in colorectal cancer specimens and could serve as an independent predictor of colorectal cancer patient survival [9]. Brady et al. demonstrated that highly expressed transcription factor ARNTL2 associated with dismal survival of lung adenocarcinoma patients [20]. However, the clinical values and potential functions of ARNTL2 in ccRCC remains elusive. In this study, we carried out a systematical analysis through integrating available data from GEO and TCGA database and validated the overexpression of ARNTL2 in human ccRCC clinical samples, overexpressed ARNTL2 correlated with advanced clinical tumor stages and grades, we also found that upregulation of ARNTL2 predicted poor overall survival and could serve as an independent prognostic variable in ccRCC. Additionally, in order to contribute to clinical decision making, a nomogram was subsequently developed based upon the results of multivariate Cox analysis including the expression level of ARNTL2 in ccRCC.
Recent studies have recognized PBRM1 as the second-most common mutated gene behind VHL in ccRCC [21,22]. In the present study, we downloaded and analyzed the genetic mutation information of ccRCC patients from TCGA database, PBRM1 also ranked second among 30 signi cantly mutated genes (SMGs) in ccRCC, the enrolled patients with genetic mutation data were subsequently separated into low-ARNTL2 and high-ARNTL2 group, we further analyzed the SMGs between this two ARNTL2 groups. Interestingly, PBRM1 mutation occurred signi cantly higher in ARNTL2 low group compared to ARNTL2 high group, while there were no differences of other SMGs between two groups. Miao et al. reported that PBRM1-mutated ccRCC patients could bene t more from immune checkpoint inhibitor (ICI) therapy [5], Braun et al. demonstrated that PBRM1 mutations could serve as a marker of ICI response in ccRCC in a randomized clinical trial [23]. These results implied that low ARNTL2 expression ccRCC patients with high PBRM1 mutation rates might obtain more clinical bene ts from ICI therapy. Furthermore, two recent studies have revealed that ARNTL2 correlated with prognosis and immune in ltration of lung adenocarcinoma [10,11]. Which also indicated ARNTL2 might in uence the TIME of cancer. Thus, we subsequently explored the possible functions and mechanisms of ARNTL2 through GSEA analysis, the results indicated that ARNTL2 might participated in the signaling pathways including T cell receptor signaling pathway, renal cell carcinoma, focal adhesion, JAK-STAT signaling pathway and cell cycle pathway. JAK-STAT signaling pathway has been implicated in carcinogenesis and immune in ltration of ccRCC [24,25], Fang et al. reported simvastatin could inhibit the growth and metastasis of renal cancer cell growth through JAK2/STAT3, ERK and AKT/mTOR pathway [26]. Miao et al discovered that the alterations of JAK-STAT and immune signaling pathways in PBRM1-de cent renal cancer cells [5]. These ndings revealed that ARNTL2 might in uence the TIME of ccRCC via activating JAK-STAT signaling pathway.
To explore the potential roles of ARNTL2 in the TIME of ccRCC, we initially employed ssGSEA algorithm to assess the associations of the expression of ARNTL2 and its top three co-expressed genes with various tumor in ltrated immune cell types in ccRCC, the results revealed that ARNTL2 and its coexpressed genes remarkably linked to the enrichment of T cells, T helper cells and macrophages in ccRCC. Furthermore, by using the xCell algorithm analysis, we found distinct immune score, microenvironment score, stroma score, in ltrating immune cell types and immune-evasion marker genes such as PD-1, PD-L1 expression between ARNTL2 high and low group. Notably, high ARNTL2 correlated with high in ltration level of CD8 + T cell, macrophage, DC and Th2 CD4 + T cell, as well as of CD8 + T cell predicted poor prognosis of patients with ccRCC [28]. These ndings indicated that ARNTL2 might involve in the formation of CD8 + in amed ccRCC subtype and immune escape.
Previous studies have evidenced that the cancer cells could evade anti-tumor immunity and drive tumor development by overexpressing PD-L1 to interact with immune cells [29,30]. We applied a correlation analysis across 33 cancer types and found that signi cantly positive associations between ARNTL2 and PD-L1 expression, as well as the other immune-checkpoint related genes expression. In addition, we also utilized xCell algorithm to investigate the connections of ARNTL2 expression with the in ltrating level of immune cell across 33 cancer types, the results revealed that ARNTL2 positively associated with Th2 CD4 + T cell and CD4 + memory T cell, while negatively correlated with Th1 CD4 + T cell and NK T cell across diverse cancer types. Accumulating studies have revealed that Th2 leading in ammation was implicated in the development of cancer and facilitated tumor immune escape [31][32][33]. The ratio of Th2/Th1 in the TIME of cancer could act as an independent prognostic biomarker of cancer patients' survival [34]. These ndings indicated that ARNTL2 might implicated in Th1/2-polarizing in cancer.
Furthermore, we also found that elevated ARNTL2 expression correlated with dismal prognosis in a variety of cancer types besides ccRCC through analyzing the clinical data from TCGA and GEO database.
In short, ARNTL2 might be an effective prognostic marker and a potential target for cancer immunotherapy.
Nevertheless, several limitations also merit attentions. Firstly, the associations between the TIME and ARNTL2 were analyzed only based on TCGA database due to paucity of applicable data in our own cohort, although we illustrated the correlations through multiple methods.

Availability of data and materials
The datasets exhibited in present study can be discovered in online repositories, further inquiries can be directed to the corresponding author.     Expression patterns of immune cells and evasion markers in high and low ARNTL2 groups. a The heatmap showed signi cant differences of tumor immune microenvironment between two groups based upon ARNTL2 expression in ccRCC. b High ARNTL2 expression correlated with higher immune score, microenvironment score and stroma score compared to low ARNTL2 group. c The relative expression level of CD8+ naive T cell, CD8+ cell, CD4+ memory cell, myeloid dendritic cell, macrophage, Th2 CD4+ T cell, Th1 CD4+ T cell, naive B cell, NK T cell. d The relative expression level of immune evasion-markers in low and high ARNTL2 expression group.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.