The Evolving Role of Cytoreductive Nephrectomy - Incorporating Genomics of Metastatic Renal Cell Carcinoma into Treatment Decisions

Abstract

Purpose of review

Recent publications evaluating cytoreductive nephrectomy (CN) in the era of targeted therapy emphasize the importance of patient selection. We reviewed the predictive role of genetic alterations in patients with metastatic renal cell carcinoma (mRCC) undergoing CN.

Recent Findings

Studies evaluating the association between genetic alterations and outcomes following systemic treatment for mRCC include mainly patients after CN. Expression of pro-angiogenic genes, single nucleotide polymorphisms involving genes of the vascular-endothelial growth factor (VEGF) pathway and somatic mutations of chromatin remodeling genes were associated with response to VEGF-targeted therapy. Outcomes following treatment with mTOR inhibitors were initially associated with mTOR/TSC1/TSC2 mutations; however, subsequent studies did not validate these findings but rather found an association between loss of PTEN expression and PBRM1 mutations and improved outcomes. Loss of PBRM1 was initially linked to response to immunotherapy however larger studies question this association and showed high expression of T-effector gene signature predicted improved outcome. Primary tumors with low intratumor heterogeneity but elevated somatic copy-number alterations were associated with rapid progression at multiple sites.

Summary

Genetic alterations may help select patients for CN and optimize timing of treatment. Intratumor heterogeneity and genetic discordance between primary and metastatic tumors may limit clinical applicability. Future studies should evaluate approaches to overcome these limitations.

Introduction

Initial studies by Flanigan et al. and Mickisch et al. established the benefit of cytoreductive nephrectomy (CN) for patients with metastatic renal cell carcinoma (mRCC) treated with interferon α−2b [1–3]. However, the role of CN has recently been questioned with the publication of the CARMENA trial, a randomized phase 3 trial, which did not show a survival benefit when CN was performed prior to treatment with sunitinib [4]. The findings of this study, which included patients with intermediate-poor risk disease, emphasize the importance of patient selection for CN [5].

Clinical and pathological information may help identify patients who would benefit from CN [6, 7]. Culp et al. reported an association between increased risk of death after CN and lactate dehydrogenase level greater than the upper limit of normal, albumin level less than the lower limit of normal, symptoms at presentation, liver metastases, retroperitoneal adenopathy, supradiaphragmatic adenopathy, and clinical stage ≥T3 [6]. Margulis et al. developed two models for accurate prediction of cancer-specific survival after CN; a pre-operative model including serum albumin and serum lactate dehydrogenase and a post-operative model including the previously mentioned parameters and final pathologic tumor stage, nodal stage, and receipt of blood transfusion [7]. However, when validated in a separate cohort, the preoperative model showed a low discrimination with minimal clinical utility, suggesting there is need for improved models to risk-stratify patients [8].

Incorporating genetic data into risk models showed improved stratification of mRCC patients who received targeted therapy [9]. In the current review we delineate the genomic landscape of mRCC and explore the added value of genomic information in identifying patients who will benefit from treatment and specifically CN. Finally, we describe limitations with the use of genetic information and future directions to overcome these limitations.

The Genomic Landscape of Metastatic Renal Cell Carcinoma

Several landmark studies, using next generation sequencing (NGS) of primary renal tumors, elucidated the common genomic alterations in clear cell renal cell carcinoma (ccRCC). Loss of chromosome 3p and VHL mutations are the foundational events in this disease occurring in upwards of 90% of patients [10–13]. Additional mutations involve PBRM1 (33–41%), SETD2 (11–12%) and BAP1 (10%) all of which are chromatin modifier genes [10, 11]. These findings led to a better understanding of the biology underlying ccRCC and possible targets for treatment.

Recent publications describe the mutational landscape of mRCC [12, 14, 15]. De Velasco et. al. evaluated two separate cohorts of patients with unmatched metastases and primary ccRCC who underwent NGS. Common mutations in metastatic lesions were VHL (74–78%), PBRM1 (33–47%), SETD2 (25–31%), BAP1 (13–14%) and TP53 (15–16%). The frequency of mutations in primary tumors and metastases were comparable with no significant difference in mutation type and burden. While TP53 mutations were more prevalent in metastatic disease (15% vs. 9%), this was not significant when controlling for multiple testing [14]. Becerra et. al. evaluated the results of NGS of matched pairs of primary and metastatic lesions. In this study 47/60 patients carried nonidentical cancer gene mutations within their matched primary-metastatic pair. The study concludes that mutation profiles of the primary tumor alone could compromise precision in selecting effective targeted therapies [15]. When combining mutational data from these two publications, a higher rate of KDM5C and PBRM1 mutation were found among metastases (p=0.028 and 0.033, respectively, Figure 1). Further support for the diverse mutational landscape of primary and metastatic RCCs comes for the recently published study by Turajlic et. al. which identified 3 groups of tumor clones within the primary tumor: clones that were not selected and absent in the metastases (51%), clones that were maintained and appeared both in the primary and metastatic tumor (15%), and clones that were selected which were sub-clonal or absent in the primary tumor and clonal in the metastases (34%) [12].

Figure 1 -.

A comparison of the mutational landscape of metastatic and primary ccRCC as reported by Becerra et al. and de Valesco et al. * denotes p= 0.033 and ** denotes p= 0.028

Using Genomic Classification to Identify Response in Patients with Metastatic RCC

Initial NGS studies have shown an association between PBRM1, BAP1 and SETD2 mutations in the primary tumor and pathological and clinical outcomes [10, 11]. Furthermore, when accounting for the presence and clonality of specific driver mutations and copy number events, Turajlik et al. was able to identify 7 distinct evolutionary subtypes which differed in their clinical phenotypes and outcomes [13]. Multiple studies have evaluated the association between genetic classifiers and the outcome of mRCC patients treated with VEGF inhibitors, mTOR inhibitors and immunotherapy.

Genetic Markers Associated with VEGF Inhibitor Response

The most common mutation in ccRCC involves VHL whose loss of function results in constitutive activation of HIF-α, which stimulates production of VEGF and increases activity of VEGFR, resulting in increased angiogenesis and metastatic potential [16]. VEGF inhibitors prevent the pro-angiogenic expression of tumors, therefore, it can be posited that patients with genetic mutations resulting in upregulation of these receptors would be more likely respond to VEGF inhibition [17].

Meta-analyses have shown that VHL and VEGFR1 genetic alterations are not associated with pathologic and oncologic outcomes in patients with ccRCC [18, 19]. However, high expression of pro-angiogenic genes in the primary tumor of ccRCC patients treated with first-line sunitinib were associated with oncologic outcome and overall survival [20]. Another study found that two SNPs in VHL were associated with decreased overall survival when controlling for clinicopathological characteristics in patients undergoing first-line treatment with TKIs. Both SNPs were associated with higher level of sarcomatoid dedifferentiation and one was found within a 3’ untranslated region of VHL thought to prevent binding of mi-RNAs regulating VHL expression, resulting in decreased VHL production [16]. Additional SNPs involving genes from the angiogenesis pathway including VEGFA, VEGFR2, VEGFR3, PDGFRA, PDGFRB, PRKAR1B and IL8 were associated with worse response rate and survival, while SNPs associated with VEGFR2 and VEGFB showed a protective effect [21].

Mutations in genes other than those directly involved in the VEGF pathway have also been implicated in differing outcomes with VEGF inhibitors. Genotyping of primary renal tumors in patients treated with first-line sunitinib in the RECORD-3 trial revealed that BAP1 mutations were associated with reduced PFS, whereas patients with KDM5C mutations had longer PFS [22].

An additional gene associated with outcome to VEGF targeted therapy is PTSG2, which encodes for COX-2, a cyclooxygenase that has been associated with increased microvessel density, stage, and grade of tumor. The activation of angiogenic pathways outside that of the VHL-pathway may account for resistance to VEGF inhibitors. Indeed, the minor allele of SNP rs5275 located in the untranslated region of PTSG2 was found to affect mi-RNA binding resulting in increased expression of the gene and decreased PFS and cancer-specific survival [17]. Additional agents which may provide a genetic basis for resistance to sunitinib include histone methyl transferase EZH2 [23], miRNA-21 [24], and MUC13 [25], a cell surface mucin glycoprotein. Silencing expression of these genes resulted in reversal of acquired resistance to VEGF inhibitors.

Genetic Markers Associated with mTOR Inhibitors

mTOR is a serine/threonine protein kinase implicated in protein, lipid, and nucleotide synthesis that results in increased cell growth and division. Certain regulators of mTOR exist, such as the tumor suppressor genes TSC1 and TSC2. As a result, patients with activating mutations in mTOR or loss of function mutations in TSC1/TSC2 would be expected to best respond to mTOR inhibitors. Kwiatkowski et. al. reported that mutations in mTOR, TSC1, or TSC2 were more common in responders (12/43, 28%) than non-responders, (4/36, 11%), (p=0.06) [26]. Contrary to this finding, other studies reported no association between PFS and mTOR/TSC1/TSC2 mutations in patients treated with everolimus [27, 28].

Unlike mTOR pathway mutations, loss of PTEN expression, seen in 50% of patients in the RECORD-3 trial, was associated with increased PFS in patients treated with everolimus [28]. PBRM1 mutations were also associated with increased PFS while BAP1 mutations were associated with a decreased PFS in patients treated with first-line everolimus [22]. Thus, patients with loss of function mutations in PTEN and PBRM1 may represent a subset who better respond to everolimus.

Genetic Markers to Predict Response to Immunotherapy

The discovery of checkpoint inhibitors has resulted in multiple new therapeutic options for mRCC. Studies aggregating RCC with other solid malignancies have suggested high overall mutation burden to be positively associated with response to immunotherapy due to stimulation of TH1-cells with increased neoantigens [29]. Similarly, tumors which are microsatellite instability high and those with deficiencies in mismatch repair genes are more sensitive to checkpoint inhibitors due to a buildup of somatic mutations in tumor cells, a high tumor mutation burden and increased expression of neoantigens [30].

When evaluated specifically in ccRCC, an additional marker of positive response to immunotherapy is high endogenous retroviral load, which was associated with increased immune infiltration, checkpoint pathway upregulation, higher CD8+ T cell fraction in infiltrating leukocytes and BAP1 mutation [31]. Conversely, transcription elongation mutations were associated with poor response to immunotherapy in mRCC and metastatic melanoma patients [32].

The IMmotion150 study, a randomized phase 2 study, compared atezolizumab (anti-PD-L1) alone or combined with bevacizumab (anti-VEGF) versus sunitinib in 305 patients with treatment-naive mRCC, most of whom (87%) underwent CN prior to enrollment. An exploratory analysis of molecular correlates of clinical outcome showed distinct subgroups based on angiogenesis, immune, and myeloid inflammation-associated genes. The subgroup with high expression of T-effector gene signature was positively associated with protein expression of PD-L1 and CD8 T-cell infiltration, indicative of pre-existing adaptive antitumor immunity. High expression of the angiogenesis gene signature was associated with improved overall response rate and PFS within the sunitinib treatment arm. High expression of T-effector gene signature was associated with improved overall response rate and PFS within the atezolizumab+bevacizumab arm while high expression of myeloid inflammatory gene signature was associated with an opposite effect. When evaluating alteration at the gene level, no association was found between VHL mutations status and PFS in any group while PBRM1 mutations were associated with improved PFS in the sunitinib group and was not associates with outcome in the atezolizumab monotherapy group [33]. This is in contrary to a previous report by Miao et al. who found that loss of function mutations in PBRM1 was associated with clinical benefit in 35 patients who received anti-PD1 monotherapy (p=0.012). This finding was confirmed in an independent validation cohort of 63 ccRCC patients treated with PD-1 or PD-L1 blockade therapy alone or in combination with anti-CTLA-4 therapies (p=0.0071) [34]. Current studies are underway to better define the role of PBRM1 mutation status in predicting the outcome to immunotherapy.

Despite the heterogeneity of evaluated studies, ultimately, several genetic mutations have been identified as being associated with differential clinical outcomes with different targeted therapies and immunotherapies (Table 1). Given the fact most of these markers were evaluated in cohorts in whom most patients underwent CN, an argument can be made that cytoreductive surgery should be avoided in patients with markers pertaining to an adverse outcome, especially if they are poor surgical candidates, however further studies are required to evaluate the true role of CN within this setting. Moreover, data from the recent SURTIME trial suggest that a deferred approach to CN in which patients are initially treated with sunitinib and offered nephrectomy only if their disease does not progress may be superior to performing upfront CN [35]. Within this context the use of genomic markers may aid in identifying patient who will respond to treatment and possibly guide the timing of CN.

Table 1 -.

Genetic predictors of outcome in recently published studies of patients with metastatic renal cell carcinoma most of whom underwent cytoreductive nephrectomy prior to systemic treatment

Study	No. Cytoreductive/ No. Total	Sample Source	No. ccRCC/ No. Total	Treatment	Predictors of outcome
VEGF targeted therapy
Verbiest et al. 2018 [16]	187/199	Germline DNA from normal kidney (n=106) or blood (n=93)	199/199	Sunitinib (74%); pazopanib (15%); sorafenib (11%)	The minor alleles of the VHL SNPs rs1642742 (HR=2.07, 95% CI 1.13–3.77, p=0.018) and rs1642743 (HR=2.12, 95% CI 1.21–3.73, p=0.009) were associated with poor OS on multivariable analysis
Cebrian et al. 2017 [17]	59/75	Germline DNA from blood	56/75	Sunitinib	The minor allele of the PTGS2 SNP rs5275 was associated with decreased PFS (HR=3.03; 95% CI 1.1–8.33, p=0.053) and CSS (HR=5.22, 95% CI 1.7–15.98, p=0.01)
Beuselinck et al. 2018 [20]	104/104	Primary kidney tumor	104/104	Sunitinib (28% received prior cytokines)	mRNA expression of genes associated with angiogenesis (HIF1A, HIF2A, PDGFRB, VEGFA, VEGFR1, VEGFR2, VEGFR3) were associated with OS and oncologic outcomes (PFS, response rate)
Garrigos et al. 2017 [21]	43/50*	Primary kidney tumor	41/50*	Tyrosine kinase inhibitor (86%), mTOR inhibitor (14%)	In the advanced setting, 5 SNPs determined inferior overall survival (IL8: rs2227543, PRKAR1B: rs9800958, PDGFRB: rs2302273; p=0.05) or worse response rate (VEGFA: rs699947, rs3025010 p ≤ 0.01)). One SNP in VEGFB (rs594942) predicted better response rate (p=0.03)
Hsieh et al. 2017 [22]	199/220	Primary kidney tumor	220/220	Sunitinib (50.5%); everolimus (49.5%)	First-line sunitinib – BAP1 (HR=1.69; 95% CI 0.9, 3.2) and KDM5C mutations (HR=0.57; 95% CI 0.3, 1.1) were associated with PFS First-line everolimus - PBRM1 (HR=0.53; 95% CI 0.3, 0.8) and BAP1 (HR=1.84; 95% CI 1.1, 3.2) were associated with PFS
mTOR inhibitors
Kwiatkowski et al. 2016 [26]	NA	Primary kidney tumor	69/79	Everolimus (48%); temsirolimus (52%), (56% received prior VEGF-targeted therapy)	Mutations in MTOR, TSC1, or TSC2 were more common in treatment responders (12/43, 28%) than non-responders (4/36, 11%), (p=0.06)
Gao et al. 2018 [27]	22/24	Metastatic lesions	18/24	Everolimus (100% and 25% pre-treated with VEGF-targeted therapy and immunotherapy, respectively)	There was no difference in PFS between patients with and without mutation in the PI3K-AKT-mTOR pathway
Voss et al. 2019 [28]	184/184	Primary kidney tumor	154/184	Everolimus	PFS did not differ based on TSC1, TSC2, or mTOR mutations status (HR, 1.1; p=0.806) Retained PTEN expression on IHC was associated with decreased PFS (HR, 2.5; p<0.001)
Immunotherapy
McDermott et al. 2018 [33]	265/305	Primary kidney tumor	288/305	Atezolizumab (34%); Atezolizumab + Bevacizumab (33%); Sunitinib (33%)	High expression of the angiogenesis gene signature was associated with improved overall response rate and PFS within the sunitinib treatment arm High expression of T-effector gene signature was associated with improved overall response rate and PFS within the atezolizumab + bevacizumab arm and high expression of myeloid inflammatory gene signature was associated with an opposite effect PBRM1 mutations were associated with improved progression free survival in the sunitinib group only
Miao et al. 2018 [34]	NA	Primary kidney tumor	35/35 (discovery) and 63/63 (validation)	anti-PD1 monotherapy (discovery); anti-PD-1 or anti-PD-L1 +/− anti-CTLA-4 (validation)	In the discovery cohort, biallelic PBRM1 loss was associated with clinical benefit (p=0.012), prolonged OS (p=0.0074) and PFS (p=0.029) The clinical benefit was confirmed in the validation cohort (p=0.0071)

^*Includes part of the cohort which had metastatic disease.

ccRCC = clear cell renal cell carcinoma; VEGF = vascular endothelial growth factor; HR = hazards ratio; CI = confidence interval; PFS = progression free survival; SNP = single nucleotide polymorphism; CSS = cancer specific survival; OS = overall survival; NA = not available

Within the setting of a CN, metastatecomty has been used as an adjunctive procedure in carefully selected patients with low tumor burden. However, patients with rapid progression may not be suitable for CN and metastatectomy. When evaluating ccRCC evolutionary subtypes, the group of rapid progressors was enriched for the “multiple clonal driver”, “VHL wild-type” and “BAP1 driven” subtypes and associated with lower intra-tumoral heterogeneity and elevated weighted genome instability index relative to the cases with attenuated progression [12].

Genomic Classification in the Context of Cytoreductive Nephrectomy

Tennenbaum et al. directly evaluated the role of genomic classifiers within the context of CN [36]. The association between VHL, PBRM1, BAP1, SETD2 AND KDM5C mutation status and overall survival was evaluated in 167 patients with metastatic ccRCC who underwent genomic sequencing of their primary tumor. Mutations in SETD2 (HR=0.58, 95% CI 0.35–0.94, p=0.027) and KDM5C (HR=0.43, 95% CI 0.22–0.85, p=0.019) were associated with a reduced risk of death while BAP1 mutations were associated with an increased risk of death (HR=1.81, 95% CI 1.16–2.83, p=0.008). It is important to note, that within this cohort, aside from the genetic factors evaluated, grade of the primary tumor was the only clinic-pathologic characteristics evaluated that was significantly associated with outcome [36].

Limitations to the use of Genomic Classifiers

Intra-tumoral heterogeneity (ITH) poses an obstacle for the use of genetic data. Sequencing of tumor samples from multiple geographically distinct regions highlighted the diversity of genetic findings within a tumor [12, 13, 37]. Moreover, multiregional sampling is required to define mutation clonality within a tumor [13]. While 2 distinct renal biopsies may be sufficient to characterize the genetic landscape of small renal masses, large renal tumors, as is the case in CN, may require 4–8 geographically distinct biopsies to capture ≥75% of molecular events [13]. Further complicating this issue is the discordance between genetic information obtained from the primary tumor and metastatic sites [12, 15].

A possible approach to overcome limitations associated with genetic heterogeneity while avoiding the need for repeated invasive biopsies is the use of circulating tumor DNA (ctDNA) [38]. Highly sensitivity detection methods and prior knowledge of mutations are required to successfully identify genetic alterations in ctDNA in RCC [39]. Pal et al. evaluated the ctDNA profile of a large cohort of 220 consecutive patients with mRCC using a 73-gene sequencing platform with average raw coverage depth of 15,000x. Genomic alterations were detected in 79% of patients the most frequent of which were TP53 (35%), VHL (23%), EGFR (17%), NF1 (16%), and ARID1A (12%). The frequency of most genomic alterations differed between patients receiving different lines of treatment [40]. While this study did not cover all prognostic genes in ccRCC, it affirms that NGS of ctDNA may provide a mutation profile of mRCC [38]. When comparing mutations identified on ctDNA to those identified on tumor tissue sequencing a similar number of mutations were found however there was discordance in mutation type suggesting ctDNA may better reflect genomic ITH and that the two tests complement each other [41]. Furthermore, patients with detectable ctDNA had a higher tumor burden compared to those without [42], and positive ctDNA and fragmentation of ctDNA were significantly associated with poor cancer-specific survival [43].

Additional limitations precluding widespread application of genomic markers is the heterogeneity of studies conducted, which may contribute to inconsistent results [18]. Furthermore, without a control group treated with a placebo, it is unclear whether these genetic alterations truly represent predictive markers as opposed to prognostic markers. For instance, BAP1 mutations are associated with poorer clinical outcomes in patients taking everolimus and sunitinib [22], but in fact may just be a marker of more aggressive disease [44].

Conclusion

The current role of CN is being re-evaluated, emphasizing the importance of patient selection. Genomic markers have shown associations with treatment outcome for patient with mRCC who underwent systemic treatment. While these markers may assist in choosing patients who may benefit the most from CN, future studies are needed to evaluate how best to incorporate these markers in clinical practice, while overcoming limitations associated with intra-tumoral heterogeneity.

Key Points.

• Recent publications emphasize the importance of patient selection prior to performing cytoreductive nephrectomy for metastatic renal cell carcinoma.

• Genetic alterations were associated with outcome in patients undergoing cytoreductive nephrectomy followed by systemic treatment.

• These markers may serve to choose patients who are candidates for CN and select the timing of surgery.

• Intra-tumoral heterogeneity poses an obstacle for the use of genetic findings and may be overcome by multiregional sampling or sequencing of circulating tumor DNA.