Molecular Signature Of Subtypes Of Renal Cell Carcinomas And Immunotherapy Strategy

Purpose: Classi�ed immunogenomic pro�ling based on therapeutic responses in RCC subtypes. Methods: Tumors from RCC patients from The Cancer Genome Atlas (TCGA) cohort were analyzed, and genomic pro�ling was performed. We classi�ed RCC on the basis of the immunogenomic pro�ling of 29 immune signatures. Results: We investigated the transcriptional changes of three RCC subtypes by RNA-seq. Gene ontology (GO) identify speci�c gene signatures differed signi�cantly between KIRC, KIRP and KICH related to the distinct pathways. Site of origin within the nephron was one major determinant in the molecular and immune classi�cation, re�ecting differences between three subtypes. The Immunity High KIRC and KIRP subtype was enriched not only in immune signatures, but also including PD-L1 expression signaling, NF-kappa B signaling pathway, JAK-STAT signaling pathway and Cell cycle signaling pathway. KICH was a distinct disease that shared little genomic characteristics with KIRC and KIRP. Conclusions: The identi�cation of RCC subtypes based on immune signatures has potential clinical implications for RCC treatment. It is imaginable that patients with higher immunity subtype of KIRC and KIRP would be more likely to respond to anti-PD-1/ PD-L1 therapy than patients with KICH subtype. CCL21 and CCL25 might be a potential target for KICH therapy.


Introduction
Renal cell carcinomas (RCC) comprised 2-3% of all non-cutaneous malignant neoplasms and was the sixth most common cancer in men and the eighth most common cancer in women in 2019 [1,2].Renal epithelial tumors arised from renal tubules and included three main subtypes: clear cell RCC (KIRC) accounted for 75% of RCC, papillary RCC (KIRP) accounted for 15%~20%, and chromophobe RCC (KICH) represented 5% [3].Although surgical resection was common for Patients with RCC, uncontrolled tumor progression and death still occured and RCC was poor response to conventional chemotherapeutics and radiation [4,5].Notably, RCC subtypes were associated with different histology, biological behaviors, prognosis, and responses to therapy.Thus, more effective systemic treatments for RCC subtypes patients are needed to guide treatment decisions.
Recently, cancer immunotherapy with the use of immune checkpoint inhibitors (ICIs) has been reported as an appreciated therapeutic approach for the tumors including RCC [6,7].Indeed, many experimental and clinical trials had explored the possibility of treating RCC patients with immunotherapy and expression of programmed death ligand 1 (PD-L1) in the tumors of patients [8][9][10][11].However, some patients were insensitive to ICIs, this suggested that not all RCC patients could respond to immunotherapy.In fact, tumor mutation burden (TMB), neoantigen expression, PD-L1 expression, PBRM1 loss and de cient DNA mismatch repair, have been associated with cancer immunotherapeutic responsiveness [12][13][14][15][16][17][18].Therefore, a more accurate classi cation system based on a comprehensive immune-related genes is deeply needed to improve prognosis accuracy and direct clinical practice.
In this study, we identi ed the common and speci c molecular characteristics of subtype-speci c treatment and management strategies for these cancer patients.The identi cation of immune signatureassociated RCC subtypes may facilitate the optimal selection of RCC patients responsive to immunotherapy.

Functional enrichment analysis
In order to investigate difference in potential biological role between tumor and normal tissue in KICH, KIRC, and KIRP, we used the DEGs obtained using limma and edgeR package of R to perform exploration with ltering criteria of FDR<0.01 and |logFC|>1.clusterPro ler of R was used to further functional enrichment analysis [19].Only GO categories of 'Biological Process' were considered.Functional annotation with P-value <0.05 was considered be statistically signi cant.

In ltrating Immune Cell Quanti cation
To determine the degree of immune cell in ltration in tumor, we performed the single-sample gene-set enrichment analysis (ssGSEA) score, which applied 28 expression-based gene signatures of immune cell populations to individual tumor samples [20].The immune in ltration score, stromal in ltration score and tumor purity for each RCC sample was calculated using ESTIMATE.MCP-counter computational algorithms were also used to characterize immune cell composition in both tumor and normal tissue [21].

Statistical analysis
Correlations were evaluated using the Spearman correlation test.The signi cance of the differences between continuous variable (immune in ltration level and chemokine expression) and categorical variable (RCC subtypes) was calculated by the Wilcoxon rank-sum test.Statistical analyses were performed using R software and were deemed to be signi cant with a P-value <0.05 (two-sided).

Gene Ontology and Canonical Pathway Analysis in three subgroups
We explore the biological and functional difference in three subgroups based on the differential expressed genes, especially the immune response and tumor microenvironment.Firstly, a total of 5699 DEGs were calculated in KIRC relative to normal tissue, including 3804 up-regulated and 1895 downregulated genes(Fig.1A); In KIRP relative to normal tissue, 4896 genes were differentially expressed including 2899 up-regulated and 1997 down-regulated genes(Fig.1B); 5759 of genes signi cantly differentially expressed in KICH relative to normal tissue, with 2685 up-regulated and 3074 downregulated genes (Fig. 1C).Further exploration, a total of 1208 intersection genes among the three subgroups were screened including 607 up-regulated and 601 down-regulated genes (Fig. 1D).Among all the GO-terms, only 786 (27.7%) intersected among the three subgroups (Supplemental Figure1A).GO-enrichment analysis revealed that among the DEGs were those involved in cell differentiation, meiotic cell cycle, cell maturation, DNA transcription, growth factor activity, leukocyte migration, and DNA repair.These processes were involved in cancer progress and play critical role in the initiation and promotion of tumorigenesis (Supplemental Figure1B).When we looked at the GO-terms in each subgroup of RCC, we noticed that, immune response process was enriched in KIRC, such as T cell differentiation, T cell activation, cell killing, leukocyte proliferation, lymphocyte proliferation and cytokine secretion (Supplemental Figure1C).What's more, we found GO-terms in KIRP were related with metabolism of organic and inorganic matter, such as response to lipopolysaccharide and carbohydrate binding.Interestingly, the immune response also was enriched in KIRP, such as interleukin-1 production, mononuclear cell proliferation and mononuclear cell migration (Supplemental Figure1D).In KICH, more biological functions were seemed to be involve in regulation of synaptic plasticity, kidney morphogenesis, axoneme, and metabolic process (Supplemental Figure1E).
Pathways analysis identi ed 73, 77 and 73 altered pathways in KICH, KIRC, and KIRP, respectively (FDR<0.05)(Supplemental Figure2A).The top 10 pathways of each subgroups were screened compared with normal tissue.Among them, a total of 32 pathways were common to each subgroup.Cytokinecytokine receptor interaction, cell adhesion molecules (CAMs), calcium signaling pathway, ECM-receptor interaction, steroid hormone biosynthesis and PPAR signaling pathway were the most signi cant changed pathways in each subgroup of RCC(Supplemental Figure2B).The tope dysregulated pathways detected in KIRC were more activated mainly associated with immune response (Supplemental Figure2C), such as Th1 and Th2 cell differentiation, Allograft rejection, Graft-versus-host disease, Th17 cell differentiation, Natural killer cell mediated cytotoxicity.Besides, the cancer-related signal pathways also were more activated compared with normal tissue, such as Transcriptional misregulation in cancer, Chemokine signaling pathway, JAK-STAT signaling pathway, and Rap1 signaling pathway.We found the KEGG pathways was only enriched in KIRP (Supplemental Figure2D), such as p53 signaling pathway, cGMP-PKG signaling pathway, Primary immunode ciency, PI3K-Akt signaling pathway, PI3K-Akt signaling pathway, Wnt-catenin formation, and Tyrosine metabolism.This founding indicated that most cancer-related pathways were activated compared with normal tissue.KEGG analysis for all differentially expressed genes in KICH compared with normal tissue revealed TGF-beta signaling pathway, MAPK signaling pathway, cAMP signaling pathway, Chemical carcinogenesis, Cholesterol metabolism, Complement and coagulation cascades, and Cholesterol metabolism at the top cancer-related pathways, not disclosed in KIRC and KIRP(Supplemental Figure2E).Overall, these pathways and GO-terms results indicated that cell migration, cancer-related signal pathways play an important role in KICH.

CD8+ T cell in ltration in RCC is associated with the three subtypes
Based on the GO and KEGG analysis, the immune in ltration was found to be different in three subtypes.CD8+ cytolytic T cells (CTLs) recruitment also re ect immune cell in ltration by measuring CD8A expression.Immune cell cytolytic activity (CYT) by measuring the mRNA expression levels of granzyme A (GZMA) and perforin 1 (PRF1) are a prerequisite for effective antitumor killing.We obtained RNAseq from TCGA to explore if the CD8+ cytolytic T cell was associated with molecular subtypes.Interestingly, we found CD8A was highest in KIRC compared with the rest of two subtypes(Fig.2B), meanwhile, the KIRP has a higher CD8+ cytolytic T cell in ltration relative to KICH, but CYT levels in KIRP was lower than KICH (Fig. 2B), so it's necessary to further explore.In addition, pathological type of RCC are found to have opposite trends in immune cell score and tumor purity based on ESTIMATE package of R(Fig.2A), with tumor purity increasing from KIRC to KICH (KIRC<KIRP<KICH) and immune cell score decreasing from KIRC to KICH (KIRC> KIRP> KICH).In a short, these results may indicate that KIRC raised highest number of CD8+ T cell in ltration, while the KICH contained the lowest levels.

Distinct immune landscape associated with three subtypes in RCC
The different immune-related pathways and CD8+ T cell in ltration among the three subtypes of RCC make us to explore the immune landscape in three subtypes.The ssGSEA con rmed an expected main difference in expression of marker gene sets for immune landscape (Fig. 3A), including gradually increasing expression of marker gene sets for most of T cell subpopulations (CD4 and CD8 T cell, Natural killer T cell, Gamma delta T cell, Type 1 T helper cell, Regulatory T cell (treg), T follicular helper cell) from KICH to KIRC (KICH< KIRP< KIRC).Interestingly, Type 17 T helper cell was enriched highest level in KIRP, while this kind of subtype have lowest dense of Type 1 T helper cell relative to KIRC and KICH.In addition, some of innate immune cells including Macrophage, Mast cell, MDSC, Neutrophil, Activated and Immature dendritic cell were also gradually increasing from KICH to KIRC (KICH< KIRP< KIRC), exclusive of Eosinophil, Monocyte, CD56 natural killer cell, which suggest that the higher activation of effector and professional antigen-presenting immune process in KIRC.On the other hand, T cell priming and activity, including T cell co-stimulation, T cell co-inhibition and IFN expression also was explored based on ssGSEA (Fig. 3B).Besides, exhaustion T cell could indicate the immune process, especially immune evasion, we found KIRC has highest T cell exhaustion level according to gene sets (CTLA4, PD1, TIM3, LAG3, PDL1, and FASLG).Importantly, we found enrichment of T cell in amed gene expression pro les (Parain ammation and In ammation), which could be used as a predictive of response to immunotherapy in tumor(Fig.3C).

Differences in APC related to molecular subtypes
Dendritic cell (DC), as the most important antigen presenting cells (APC), play a critical role in innate immunity and adaptive new immunity, and as the only member of activating naïve T cell, DC has been paid more and more attention.We thus explored the process associated with DC maturation and T cell activation.From our above analysis, we found gradually increasing of immune signaling pathways and DC activation from KICH to KIRC.In addition, APC function could be re ected by the expression of gene sets including MHC Class I, HLA family, APC Co-inhibition, APC Co-stimulation and we observed the KIRC present highest levels compared with others(Fig.4A).In addition, Pre-CDC differentiated into IRF4dependent DC (IRF4-DC) and BATF3-dependent DC (BATF3-DC).Tissue DCS migrate to draining lymph nodes, where they show a mature phenotype, while the DCS stay in lymphoid organs always showed immature phenotype.The BARF3-DC development depends on the transcription factors IRF8 and BATF3, while the expression level RELB and IRF4 were critical to IRF4-DC mature (Fig. 4B).Our results re ected that the expression of BATF3-DC, LC, IRF4-DC signature were increasing from KICH to KIRC (KICH< KIRP< KIRC).Together, these results indicated KIRC was more closely associated with innate immune pathways and APC process which was critical to T cell activation.

The chemokine signatures difference among the three subtypes
Chemokines are increasingly recognized for their chemotactic effects, driving primed effector T cell recruitment into tumor, and the diversity of chemokine expression can mask the contribution of individual chemokine, then the 40 known human chemokines were explored in three subtypes respectively.We explored the relationship of all chemokines with cytolytic T cells in three subtypes(Fig.5A), although active chemokines are similarity between KIRC and KIRP, we still found some chemokines were different, such as CXCL13 and CXCL16 which were more activity in immune response in KIRC than KIRP, while CCL25 and CXCL14 had the opposite trend.Previous report indicated that CXCL13/CXCR5 signaling modulates cancer cell ability to grow, proliferate, invade, and metastasize [22].There are speci c examples of chemokines important in lymphoid development, of which the CCL25-CCR9 axis plays a critical role in T cell development in thymus [23].Our founding suggested that the different active chemokines between KIRC and KIRP may be critical biomarkers of the difference in immune mechanisms between them.The results also show that the active chemokines in KICH have huge variations compared with KIRC and KIRP, which indicated the immune response was different among them.When hematopoietic precursors from bone marrow colonize the thymus, the development of T cells in the thymus begins.Some chemokines participate in the seeding of the thymus rudiment by hematopoietic bone marrow precursors.Previous study found the CCR7 and CCR9, the ligands receptor of CCL19/CCL21 and CCL25, participate in this seeding process [24].Interestingly, we found KIRC has higher the three chemokines in ltration compared with KIRP, while the KICH has the highest in ltration of CCL21 and CCL25, thought above analysis found the CCL25 has lower correlation to cytolytic T cells, which mean that the KICH pay a more activity in this process [Fig.5B].
More and more studies reported that multiple oncogenic mutations can affect the chemokine axes, which correlated with immune escape and tumor progress.The EGFR/Ras mutation in tumor cells induces a distinct chemokine repertoire, and tumors facilitate progression by the EGFR/Ras-induced production of CCL20 [25].From Supplemental Figure2, we found the Ras signal was more active in KICH, which may could explain the reason why the CCL20 was highest active in KICH than the two subtypes.In the other hand, the presence of T cells correlated with the expression of CCL2, CCL3, CCL4, CCL5, CXCL9, and CXCL10[26] and the cognate receptors of these chemokines were found to be upregulated on CD8+ effector T cells, then we explored these chemokines in three subtypes.The results indicated CCL2, CCL3, CCL4, and CCL5 were gradually increasing from KICH to KIRC, but CXCL9, CXCL10 and CXCL11 were signi cantly higher in KICH compared with KIRP (Fig. 5C).From gure 7D, the top 30 mutation genes were found based on the data from TCGA, and we found TP53, MUC4, PTEN, ZAN, TTN were the highest mutation frequency in KICH, while TTN, MUC16, MET, KMT2C, SETD2 in KIRP, and VHL, PBRM1, TTN, SETD2, BAP1 in KIRC.Previous study con rmed that PTEN loss in triple-negative breast cancer increased downstream production of IRF3 targets including CXCL10 expression level [27], which was consistent with our founding.CCR5, receptor of CCL5, CCL3 and CCL4, were elevated in BAP1-mutant clear cell renal cell carcinoma [28], which also explain why KIRC has a higher immune response.In conclusion, we found that oncogenic mutations can affect the chemokine activation in three subtypes, which could change the immune response.

The immune difference in molecular subtypes was replicated in independent studies
Based on expression levels of above gene sets used in TCGA, the three subtypes (KIRC, KIRP, KICH) present the immune in ltration landscape in public gene sets.The similar results were found in our validation datasets (GSE15641 and GSE19949/GSE12090), the two datasets all showed that KIRC has a higher immune score, while KICH has the lowest immune score (Fig. 6B).The heatmap of GSE15641 show signi cant different immune landscape of the three subtypes (Fig. 6A).In conclusion, the immune landscape was distinct in the subtypes, and the immune cell in ltration were increasing from KICH to KIRC (KICH< KIRP< KIRC).

Comparison to normal tissues suggested Cytotoxic lymphocytes activation in KIRC tumors only
In our data, compared with their paired normal tissue, we found B lineage and CD3+ T cells showed either no change or lower population of immune cell in three subtypes relative to their normal tissues (Supplemental Figure3).Interestingly, we found the CD8 T cell and Cytotoxic lymphocytes were signi cantly higher activation in KIRC while KIRP and KICH didn't show obvious activity relative to normal tissue.On the other hand, Monocytic lineage and Myeloid dendritic cells were more activity in KIRC and KIRP, though the latter is relatively lower, while the two kinds of immune cells have lower population of immune cell in KICH.As for NK cells, though our above analysis con rmed that KICH has a lower in ltration relative to KIRP, NK cells was found to be more activity in KICH relative to their normal tissue.

Discussion
RCC subtypes patients with different cell types originating, as well as their histological, genetic and clinical diversity, may have different immue response.Therefore, accurate classi cation of RCC is crucial for targeted therapy selection.In order to nd the unique genes signature in each subtype of RCC, we did RNA-seq analysis between KIRC, KIRP and KICH samples from TCGA.We identi ed 5699, 4896 and 5759 differentially expressed genes in KIRC, KIRP and KICH, respectively(Fig 1).The top upregulated genes including PAEP, HHATL and CRP in both subtypes of KIRC and KIRP play a major part in cancer progression.TERT, LHX9 and NKX6.1 were upregulated genes in KICH.Moreover, it was also noticed that UGT2A3, UGT3A1 and UGT1A9 from UGT family members were also signi cantly downregulation than normal tissues in KICH.UGT family could contribute to cancer risk by modulating the kidney exposure to mutagenic stimuli [29].Expression changes of UGT family genes can affect the development and progression of various types of cancer including the cancer of the kidney cancer [30][31][32][33].Moreover, it was also noticed that Cell migration and immune response process were enriched in KIRC and KIRP by GO and KEGG analysis(Supplemental Figure1,2).Furthermore, Compared to KICH, Th1 and Th2 cell differentiation, Chemokine signaling pathway, JAK-STAT signaling pathway and PI3K-Akt signaling pathway were more in KIRC, and p53 signaling pathway, cGMP-PKG signaling pathway, PI3K-Akt signaling pathway and Wnt-catenin formation in KIRP.But more synaptic plasticity and kidney morphogenesis were involved in KICH, and MAPK signaling pathway, cGMP-PKG signaling pathway in KICH.Our analyses clearly con rmed that KICH was a distinct disease that shared little genomic characteristics with KIRC and KIRP(Figure 1E,supplemental 4,5,6).Our study supported notion that Cells of origin were looked as one major determinant in the classi cation which re ecting molecular pro ling differences of three subtypes.KIRC was thought to arise from cells in the proximal convoluted tubule, while KICH was thought to arise from intercalated cells in the distal convoluted tubule of the nephron, and KIRP was in general most similar to the proximal nephron.Our results suggested that KICH needed speci c therapies, rather than simply adopting therapeutic strategies used for KIRC or KIRP.
Next, our results showed that three subtypes of RCC be different in the immune in ltration(Fig 2).CD8+ cytolytic T cells (CTLs) recruitment also re ect immune cell in ltration by measuring CD8A expression.We found that KIRC subtype was enriched not only in the highest number of immune cells(higher expression levels HLA genes and active CD4 T cell, active CD8 T cell and active B cell) (Fig 2,3,4), but also in many cancer-associated pathways including NF-kappa B signaling pathway, JAK-STAT signaling pathway and Cell cycle signaling pathway.Interestingly, we also fund PD-L1 and PD1 expression signaling showed signi cantly higher expression levels in KIRC (Fig 2, Supplemental Figure1).The Immunity High KIRP subtype contained in not only immune signatures, but also PI3K-Akt signaling pathway, Rap1 signaling pathway, Ras signaling pathway, MAPK signaling pathway, FoxO signaling pathway, JAK-STAT signaling pathway, HIF-1 signaling pathway, NF-kappa B signaling pathway and PD-L1 expression signaling.It were reported Resected RCC tumours were often extensively in ltrated by CD8+ T cells, suggesting immune recognition of the tumour but poor tumour cell killing [34,35].It may partially explain why an increased level of in ltrating CD8+ T cells was associated with worse outcome resulting from PD-L1 increased in tumor cell of KIRC and KIRP.Our studies suggested PD-L1 treatment would be most likely to have a good response in KIRC and KIRP patients.The KICH subtype was also enriched in TGF-beta signaling pathway, p53 signaling pathway, Hedgehog signaling pathway, EGFR tyrosine kinase inhibitor resistance, Notch signaling pathway, Hippo signaling pathway, Wnt signaling pathway, GnRH signaling pathway, VEGF signaling pathway, AMPK signaling pathway, cGMP-PKG signaling pathway and mTOR signaling pathway.Interestingly, we found that the TGF signaling pathway had a signi cantly enrichment in KICH subtype, this observation was in agreement with ndings from previous studies showing that the TGF signaling pathway plays a key role in preventing PD-L1 treatment [36].Although PD-L1 inhibitor showed its e cacy in KIRC and KIRP patients treated with pembrolizumab, the e cacy of immune checkpoint inhibitors in metastatic KICH is low, as no responses were observed [37][38][39].Sanjeev reported TGF-β signaling resulted in exclusion of T cells by the stromal barrier limits, and TGF-β inhibition potentiated the ability of anti-PD-L1 to enhance anti-tumour immunity resulting in optimal T cell positioning [39].Our studies suggested co-inhibition of TGF-β and PD-L1 may be good applicable to KICH.Moreover, we rstly found the higher in ltration of CCL21 and CCL25 in the KICH compared to KIRC and KIRP(Fig 5), CCL21 and CCL25 signaling was proved not only promoting proliferation, invasion, migration and drug-resistance in several types of tumors, but also inducing immune evasion [40][41][42].Our data hinted CCL21 and CCL25 might be a potential target for KICH therapy.Besides, KICH exhibited more active stem cell-associated biological processes(Notch, Hippo and Wnt signaling pathway) than KIRC and KIRP subtypes.However, its mechanism in KICH needs to be further studied.
Previous reports have indicated that the somatic genes mutation regulation that disrupts the function of tumor suppressor genes or oncogenes and was shown to promote carcinogenesis with the alteration of immune signature genes that can signi cantly effect on immunotherapy [43][44][45][46][47]. Thus, We next sought to identify mutation that were speci cally associated with immune in KICH, KIRC, and KIRP.In our studies, the mutation genes were completely different in the top 30 mutated genes in three subtypes of RCC.Firstly, we found the VHL and PBRM1 were the most mutation frequency genes in KIRC.It was known that VHL somatic mutation or deletion was characteristic of most KIRC [48].Loss of VHL usually induce cellular proliferation and promote angiogenesis by actived HIF and STAT3 signal, and STAT3 increased PD-L1 expression in transcriptional level [49,50].Furthermore, We noticed SWI/SNF complex genes including PBRM1, ARID1A and SMARCA4 were high mutation frequency in KIRC and KIRP.Loss ARID1A or PBRM1-de cient elevated PD-L1 expression in tumor cell [51][52].MET, SRRM2 and WDFY3 was higher mutation frequency in KIRP.Loss SRRM2 and WDFY3 reduced immue cell differentiation [53,54].Another interesting founding was that TTN, MUC16, MET, KMT2C were the most mutation frequency genes.In KICH, the vast majority of patients harboured mutations in TP53 and PTEN, which was previously reported as the most frequency mutated gene.MUC4, AGAP4, ZAN, TTN, FAT3, FLT4, and SETD1B mutation was found in KICH, most of which have been reported to be associated with cancer process.
These results suggested that the mutation can be used to distinguish patients with different sensitivities to immunotherapy, making individualized treatment strategies a possibility.

Conclusions
In summary, Understanding the genetic features and immune signature-based classi cation of RCC subtypes will provide effective strati cation of therapy for patients responsive to immunotherapy.Our studies had yielded several important conclusions: KICH was a distinct disease that shared little genomic characteristics with KIRC and KIRP.It is imaginable that patients with KIRC and KIRP would be more likely to respond to anti-PD-1/ PD-L1 therapy than patients with KICH subtype.It was better choose with TGF inhibitor adding PD-L1 inhibitor therapy in KICH.It was unexpected discovery that CCL21 and CCL25 might be a potential target for KICH therapy.Additionally, our study can stratify patients with different mutation and help to predict the sensitivity of patients to immunotherapy.Boxplot of ssGSEA scores differences of T cell exhaustion genes among three subtypes.

Figures
Figures