Introduction

Kidney cancer is the ninth most commonly occurring cancer in men and the 14th most commonly occurring cancer in women. There were over 400,000 new cases in 2018 worldwide [1]. In view of the multiple cell types present in the kidney, all with specialized function, it is not surprising that renal tumors represent a heterogeneous group. In the Vancouver consensus conference of the International Society of Urologic Pathology the foundation for the 2016 World Health Organization (WHO) renal tumor classification was laid [2]. This was a revision of the 2004 renal tumor classification that was deemed necessary because knowledge on pathology, genetics and epidemiology had greatly improved. Whereas renal tumor subtypes were formerly named and categorized based on cytoplasmic, morphologic and anatomic locations, a number of new entities have been added on the basis of distinctive molecular alterations. This shows that molecular characterization needs to be integrated in the diagnosis of renal cancer.

The WHO classification of tumors of the kidney distinguishes Renal Cell tumors (N = 16), metanephric tumors (N = 3), Nephroblastic and cystic tumors (N = 3), Mesenchymal tumors (N = 4 occurring mainly in children, N = 16 occurring mainly in adults), Mixed tumors (N = 2), Neuroendocrine tumors (N = 4) and Miscellaneous tumors (N = 2). Renal Cell Carcinoma (RCC) is the most prevalent adult Renal Cell tumor [2].

Worldwide, RCC represents the sixth most frequently diagnosed cancer in men and the tenth in women, accounting for 5% and 3% of all oncological diagnoses, respectively [3]. It affects nearly 300,000 individuals worldwide annually and is responsible for more than 100,000 deaths each year. The incidence of RCC varies worldwide, with higher incidence in developed countries. Several risk factors have been identified, most notably smoking, obesity, and hypertension [4]. Initially RCC was considered one entity, but it is now clear that RCC is also a very heterogeneous group of tumors, each with its own clinical and molecular manifestations. These differences are a reflection of the different cells of origin. As such, implementation of accurate biomarkers appears to be mandatory to guide clinical management.

Molecular Markers in the Diagnosis of Renal Cancer

In the 1970s clear cell RCC and granular RCC were distinguished based solely on morphological features. In 1997 the Heidelberg classification was published: the first attempt to include molecular features of RCC to subdivide RCC [5]. Clear cell RCC (ccRCC) remained as a single entity (common or conventional RCC), but the granular RCC cases were subclassified based on the then available genetic knowledge into papillary RCC, chromophobe RCC and RCC, unclassified. Papillary RCC was further divided into type 1 and type 2 pRCC based on morphological characteristics [6].

Systematic analysis of The Cancer Genome Atlas (TCGA) cohort of 894 RCC cases of various histological types revealed nine major genomic subtypes [9]. Interestingly, this included three different subtypes of ccRCC, four different subtypes of pRCC, chRCC and mixed or unclassified RCC. Thus, even within histological homogeneous groups subclassification was possible based on molecular characteristics. Such further subclassification has major ramifications for the clinical management of patients who were formally treated homogeneously. Interestingly, analysis of this TCGA cohort has revealed correlation of specific somatic alterations and metabolic pathways with subtype-specific decreased survival, as well as common signatures correlating with decreased survival within all subtypes [10]. Indeed, personalize medicine, which is foreseen as the next step forward in cancer management, is likely to become possible for RCC by including such molecular information.

Clear Cell RCC

Clear cell RCC (ccRCC) represents the most common type of RCC and comprises approximately 80% of RCCs that metastasize. Large molecular studies ultimately lead to the recognition that mutations in the Von Hippel Lindau (VHL) gene were drivers of ccRCC [11]. The investigators first studied families suffering from the autosomal dominant VHL syndrome. Affected family member frequently develop multiple, bilateral ccRCC, among others. Once it was evident that mutations in the VHL gene played a critical role in the development of ccRCC in these families, VHL mutations and VHL promoter silencing were shown to play a critical role in sporadic ccRCC [11, 12]. VHL mutations in combination with VHL promoter silencing has been reported in 60–90% of cases. This variation in detection reported between different studies is likely to be caused by sampling in combination with low coverage rate [13]. Thus, the most common alterations found in ccRCC are loss of 3p, the region where the VHL gene resides, combined with inactivation of the VHL gene in the other allele. Expression of an aberrant VHL protein or complete loss of VHL gene expression results in aberrant stabilization of hypoxia-inducible factor (HIF), a transcription factor which is responsible for transcription of numerous genes also involved in tumor formation under hypoxic conditions [14].

Other somatic mutations are now being recognized as important driver genes in ccRCC, owing to large scale sequencing efforts [15–18]. In a multi-center prospective study seven molecular subtypes were described that correlated with diverse clinical phenotypes in ccRCC [13], suggesting that a molecular subtype can serve as a potential biomarker to guide clinical management. Interestingly, one of the subtypes concerned a subtype of RCC with wild-type VHL gene, a previously unrecognized variant. Intratumor heterogeneity was one of the defining characteristics of different subtypes [13, 19] and this feature may be difficult to capture in regular clinical practice as it requires analysis of multiple tumor areas.

The most frequently involved genes in ccRCC are PBMR1, SETD2, BAP1 and KDM5C [13, 16, 17]. Remarkably, the genes encode for chromatin modifiers and loss can lead to dramatic changes. For instance, SETD2 loss leads to depletion of nucleosomes, loss of DNA methylation, aberrant splicing, and expression of abnormal intragenetic RNAs [18]. PBMR1, SETD2 and BAP1 are all located in close vicinity of VHL on 3p emphasizing the importance of 3p loss in ccRCC. The importance of loss of VHL and PBMR1 in driving renal transformation was also shown in the mouse kidney: kidney-specific deletion of Vhl and Pbrm1, but not either gene alone, resulted in bilateral, multifocal, transplantable clear cell kidney cancers [20, 21].

Interestingly, PBMR1, SETD2 and BAP1 levels have been associated with worse outcome in ccRCC [10, 22–24]: survival of patients with BAP1^mut tumors is lower compared to patients with PBMR1^mut tumors and survival of patients with BAP1^mut /PBMR1^mut tumors is very poor, but this entity is also very rare [25]. It also appears that complete SETD2 loss may be a later event: SETD2 protein expression levels were lower in ccRCC metastases compared to primary ccRCC, suggesting that complete loss of SETD2 protein expression may not be required for the development of ccRCC, but that the decrease is related to tumor progression or adaptation [26]. Finally, the frequency of SETD2^mut/PBMR1^mut is high, suggesting that these genes corporate in ccRCC tumorigenesis [27].

The large sequencing efforts have also revealed that a metabolic shift can occur in ccRCC cells. Gene expression in high-grade, high-stage ccRCC tumors reflects increased lactic acid fermentation and decreased oxidative phosphorylation. In ccRCC with a worse prognosis, the cellular metabolic activity involved increased dependence on the pentose phosphate shunt, decreased AMPK, decreased Krebs cycle activity, increased glutamine transport and fatty acid production [10]. Whether this information can lead to alternative treatment strategies remains to be established.

In several biomarker studies the potential prognostic value of epigenetic alterations has been examined (reviewed in [28]). These comprise different aberrations, such as changes in histone modifications and DNA methylation, and in view of the importance of histone modifying genes in ccRCC epigenetic alterations are regarded as potential biomarkers for the early detection of disease and for prediction of prognosis and treatment response. TCGA studies showed that promoter DNA hypermethylation frequency increases with ccRCC stage and grade [16]. In other studies, integration of global transcription levels with massive parallel sequencing data revealed a gene signature of 4 genes that correlated with poor survival [29]. This signature was validated in 2 independent cohorts. The biological relevance of the 4 genes included in this signature in ccRCC is unclear and still needs be elucidated. Moreover, independent validation by other investigators is still warranted, and the effect of tumor heterogeneity on the prognostic value of this epigenetic biomarker needs to be established.

Papillary RCC

Papillary RCC (pRCC) accounts for about 10–20% of RCC tumors and is the second most common renal neoplasm. Disease progression and patient outcome can be highly variable, and two histological subtypes have been distinguished. pRCC type 1 is often multifocal, with a quite homogeneous histological appearance, whereas the histological defined type 2 pRCC exhibits a rather diverse morphologic spectrum, and several features are shared with other non-ccRCC tumors, often leading to inconsistent diagnosis. Hereditary pRCC is a rare disorder associated with an increased risk of type 1 pRCC, characterized by activating germline mutations in c-MET [30]. Comprehensive molecular analysis confirmed that type 1 and type 2 pRCC are indeed two different entities, and that type 2 could be further stratified into three different subgroups on the basis of patient survival [31]. Similar to the hereditary type 1 pRCC, sporadic pRCC was characterized by alterations in the c-MET pathway [31]. Detailed analysis revealed 10 significantly mutated genes (MET, SLC5A3, NF2, PNKD, CPQ, LRP2, CHD3, SLC9A3R1, SETD2 and CRTC1) with somatic c-MET mutations occurring most frequently (13–15%) [32]. Activation of the NRF2-ARE pathway was associated with pRCC type 2 as earlier described [33]. Remarkably, in pRCC type 2 tumors chromatin-modifying genes SETD2, PBMR1 and BAP1 are frequently mutated, similar to ccRCC [31]. Within pRCC type 2 loss of CDKN2A, a gene playing an important role in cell cycle regulation, and CIMP, a phenotype characterized by simultaneous hypermethylation of numerous promoters, correlated with poor prognosis. CDKN2A alterations may serve as an independent prognostic marker associated with type 2 tumors, but this requires validation.

Based on genome-wide profiling of transcription-binding events providing insight in gene expression programs, pRCC was also subdivided in 3 clusters. Overexpression of MECOM, a transcriptional regulator was significantly associated with poorer OS [34]. Importantly, MECOM overexpression could not readily be explained by previously identified subgroups of pRCC. This shows the value of various molecular approaches and the results suggest that pRCC with MECOM activation identifies a subgroup of pRCC patients with adverse outcomes. Collectively pRCC type 2 is a very heterogeneous from a molecular standpoint (Table 33.1): many different events can ultimately culminate in the occurrence of pRCC type 2.

Based on the observation that c-MET plays a prominent role in type 1 pRCC foretinib, an oral multikinase inhibitor targeting MET, VEGF, RON, AXL, and TIE-2 receptors was tested in a phase 2 clinical trial [35]. Disappointingly, responses were very limited: only one of five patients with somatic MET mutation had a PR, responses were absent in 2 patients with MET amplification and only one of 18 patients with a gain of chromosome 7 experienced a PR. Possibly, other molecular aberrations lead to a foretinib-resistant phenotype. In contrast, the presence of a germline c-MET mutation was highly predictive of a response: 5/10 patients experienced a PR, 4/10 had a SD as best response. Thus, in this cohort where c-MET was possibly the most relevant driver c-MET inhibition resulted in a substantial anti-tumor effect. However, as mentioned by the investigators, this is a rare entity: germline mutations in a pRCC population from the Mayo clinic approached 0% [36], and development of foretinib for pRCC treatment was discontinued.

Chromophobe RCC

Chromophobe RCC (chRCC) is a rare type of kidney cancer accounting for approximately 5% of kidney cancer cases. ChRCC can occur in patients suffering from the Birt-Hogg-Dubé (BHD) syndrome, a rare genetic disorder. In approximately one-third of BHD patients chRCC develops and this is associated with germline mutations of the FLCN gene, a gene that encodes for a protein involved in MAPK and mTOR pathways [37]. Molecular studies of sporadic chRCC have been hampered by its rare nature. The most extensive study was reported by Davis et al. who examined 66 cases of chRCC [38]. The study confirmed that chRCC could be distinguished from ccRCC at the molecular level. Importantly, TP53 and PTEN, both tumor suppressor genes that normally regulate cell growth and apoptosis, were frequently mutated. Moreover, structural rearrangements were discovered in the TERT gene promoter. Because telomerase plays a pivotal role in senescence, deregulation of telomerase provides cancer cells the opportunity to divide indefinitely.

Because nearly all genes encoding enzymes in the Krebs cycle showed increased expression over normal in chRCC mitochondrial DNA analysis has been performed. Interestingly, mutations were observed in many chRCC with up to 18% of cases with alterations in electron transport chain Complex I genes leading to a complex metabolic phenotype [38].

Molecular Markers in the Treatment of Metastatic ccRCC

Our increased understanding of molecular events underlying ccRCC has resulted in the discovery and implementation of new treatment for patients with metastatic ccRCC (mccRCC). The central role of the HIF/VHL dysregulation lead to the development of multiple vascular endothelial growth factor (VEGF)–targeted therapies (bevacizumab, sorafenib, sunitinib, pazopanib, axitinib, and cabozantinib) and mTOR inhibitors (temsirolimus and everolimus). However, only a minority of mccRCC patients respond to a given treatment and molecular biomarkers able to predict the likelihood to respond to a particular treatment is greatly needed to help clinicians select optimal treatment for individual patients.

Several studies have shown an association between PBRM1 mutation status and patient outcome. In the RECORD-3 study, a randomized trial comparing first-line sunitinib with everolimus in patients with mccRCC, PBRM1 mutations were shown to be associated with a longer progression-free survival (PFS), whereas in the sunitinib treated patients longer PFS correlated with KDM5C mutations [39]. Similarly, mccRCC patients with PBRM1 mutant tumors treated with either sunitinib or pazopanib had significantly improved OS and PFS compared with patients with PBRM1 wildtype ccRCC [40]. Finally, in the phase II IMmotion 150 trial PBRM1 mutations were associated with improved PFS in the sunitinib treated patients but not in the other treatment groups [41]. The possibility that PBRM1 status may serve as a predictive biomarker for VEGF-targeted therapy is intriguing. It is possible that ccRCC mutated in VHL as well as PBRM1 are extraordinary dependent on HIF signaling [21], resulting to a more anti-angiogenic sensitive phenotype.

Unsupervised transcriptome analysis has identified gene signatures related to different responses to sunitinib treatment [42]. Based on this analysis a 35-gene classifier was developed which correctly classified the samples. Sunitinib response differed significantly between the identified groups. Remarkably, subtype classification was the only significant covariate in multivariate analyses for PFS and OS [42]. Similarly, these molecular subtypes were associated with outcome on pazopanib as first-line therapy [43]. Using another approach, this group studied whether mRNA-expression of genes associated with angiogenesis was correlated with sunitinib treatment outcome. On multivariate analysis, HIF2A-, PDGFRB-, VEGFC-, VEGFR1- and VEGFR2-expression were correlated with PFS and HIF1A-, HIF2A-, VEGFR1- and VEGFR2-expression with OS. VEGFR2-expression showed the strongest association with outcome, but prognostic impact was lacking [44]. Independent validation of these observations is still pending, and therefore it is unclear whether the described 35 gene signature can be clinically implemented.

Immunotherapy with immune checkpoint inhibitors has recently been developed in ccRCC. Treatment of a population of previously treated mccRCC patients with nivolumab, a PD-1 checkpoint inhibitor demonstrated OS and ORR benefits compared with everolimus in patients who had prior anti-angiogenic therapy [45]. To improve its efficacy nivolumab has been combined with other immunomodulatory agents [46, 47]. Combination of nivolumab with ipilimumab, a cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) checkpoint inhibitor for treatment-naïve patients with advanced ccRCC showed a significant improvement in OS and ORR compared with sunitinib [46]. Biomarker analysis for immunotherapy with checkpoint inhibitors has focused on gene signatures related to angiogenesis, T-effector/Interferon-ɣ response, inflammatory gene signatures, neoantigen burden and associated with a high cytolytic CD8+ T cell expression signature [41, 48, 49].

Interestingly clinical benefit has been associated with loss-of-function PBRM1 gene mutations in mccRCC patients treated with PD-1 or PD-L1 blockade therapy alone or in combination with anti-CTLA-4 [50].

The Impact of Tumor Heterogeneity on Molecular Markers

Intratumor heterogeneity has been suggested as a major hurdle for molecular subtyping: study of a single tumor-biopsy can lead to underestimation of the tumor genomics landscape [51]. Indeed, using ultra-deep sequencing on multiple regions of ccRCC intratumor heterogeneity was identified in all cases. Importantly, in 75% of cases ccRCC driver aberrations were subclonal [52], emphasizing the importance of multiregional analysis to capture all relevant driver events. In a recent 3D tumor sampling study, it was calculated that on average, two biopsies were required to detect ≥50% of all variants and seven were required to detect ≥75% of all variants [13]. For large tumors, four to eight biopsies may be required to capture the majority of events (≥75% detection). Clearly, important driver events may still go unnoticed even when ccRCC samples quite extensively. Obviously, this presents a major challenge to personalized-medicine and biomarker development for ccRCC.

Another aspect that deserves attention is the fact that almost invariably studies have focused on analysis of primary RCC. However, understanding of the molecular landscape of RCC metastasis is needed to ultimately be able to discover and validate predictive biomarkers for response. Studies on paired primary ccRCC and metastases have revealed substantial differences at the protein level and transcriptome level [53–55]. In the largest study conducted thus far, 575 primary and 335 metastatic biopsies across 100 patients with metastatic ccRCC were included [55]. Importantly, the overall number of driver events in metastases was lower compared to primary tumors and metastases were significantly more homogeneous. Across all primary-metastasis pairs, the majority of driver events were shared between primary tumors and metastases (62.5%), followed by driver events private to primary tumors (31.7%), and driver events private to metastases were limited (5.4%). This shows that sufficient sampling of the primary tumor will most likely reveal the vast majority of relevant driver events [13].

The Impact of Molecular Profiling on Therapy Choice

Implementation of molecular markers to guide therapy choice has not occurred in RCC, despite the exquisite molecular knowledge that has been accrued. It has well been established that genome-driven treatment can guide clinical management and lead to impressive responses. For instance, due to the implementation of imatinib for the treatment of chronic myelogenous leukemias (CML) harboring the BCR–ABL translocation patients with CML have life expectancies approaching that of the general population today [56]. Similarly, specific targeting agents have dramatically improved outcomes in solid tumors such as a monoclonal antibody against HER2 in epidermal growth factor receptor 2 (HER2)-expressing breast cancer [57], vemurafenib in BRAF V600-mutant melanoma [58], and gefitinib in EGFR-, ALK-, and ROS1-mutant lung cancer [59].

Our understanding of RCC’s detailed molecular profile has already lead to the implementation of numerous small molecule drugs such as tyrosine kinase inhibitors (TKI) and mTOR inhibitors, but personalized medicine approaches are lacking. Various VEGFR inhibitors can be used as first line treatment modality for metastatic ccRCC [60] and it is important to realize that these inhibitors in essence do not target the tumor cells but are anti-angiogenic. In general almost all patients show progression of disease after a longer progression free survival period. This progression can either be caused by vascular co-option, a process where tumors use normal blood vessels to sustain their metabolic needs [61, 62], or to activation of alternative pathways. In most cases patients are therefore treated with another small molecule drug as second line therapy such as cabozantinib, a VEGFR2/MET/AXL/RET inhibitor that has recently been approved. However, ultimately almost invariably patients progress and die of metastatic RCC.

Because molecular profiling is not common, it is not unlikely that responses can be improved by molecular profiling before treatment initiation. However, oncogenic mutations in a gene are often not limited to a single codon, and this may lead to different drug sensitivity. The results in pRCC with a c-MET inhibitor are exemplary: whereas mutations in the target gene c-MET were molecularly defined, responses in sporadic pRCC were absent, whereas responses were observed in germline pRCC patients. This study underscores that the benefits of genomically targeted therapy are conditioned by a multitude of factors. Facilitating the matching of highly relevant molecular characteristics with treatment modalities is going to be of utmost importance in the near future.

In conclusion, the extensive sequencing efforts have led to the recognition of multiple molecular RCC subtypes within the three main RCC subtypes. The improved understanding of the molecular landscape of RCC has led to the development of more effective therapies for metastatic RCC. However, because only subsets of patients with metastatic RCC respond to a given treatment, predictive biomarkers are needed to guide treatment selection. It is envisioned that molecular markers can play a key role in personalizing treatment [19].