Savolitinib

The Role of Circulating Tumor DNA in Renal Cell Carcinoma

Introduction

Over the past decade, a multitude of therapies have been approved for metastatic renal cell carcinoma (mRCC). These include agents that abrogate signaling through the vascular endothelial growth factor receptor (VEGF) and the mammalian target of rapamycin (mTOR) [1–10]. Most recently, multikinase inhibitors and novel immu- notherapeutic agents have gained FDA approval [10, 11]. This has created a unique challenge in treating the disease, as no optimal protocol for these therapies exists [12•, 13•]. For example, the National Comprehensive Cancer Network (NCCN) guidelines offer a category 1 recommendation for sunitinib, pazopanib, or bevacizumab with interferon as first-line therapy [14], but there is no specification as to which is preferred. In the second-line setting, the challenge is even more im- mense, with agents such as nivolumab and cabozantinib garnering category 1 recommendations as well [14]. These agents work via a different mechanism and are likely active in different subsets of patients, but no bio- markers exist to discern which patients are ideal candi- dates for these therapies [12•, 13•, 15].

The Cancer Genome Atlas (TCGA) is one exam- ple of many efforts that have been made to charac- terize the genomic profiles of patients with renal cell carcinoma (RCC) [16]. These tissue-based stud- ies have provided detailed interrogation at the DNA and RNA levels. TCGA data is available for multiple histologic subtypes of RCC, including clear cell, papillary, and chromophobe [16]. One inherent challenge is that this data set is constituted primar- ily by patients with localized disease. Multiple stud- ies suggest that there may be a temporal evolution of genomic profile as patients progress from local- ized to metastatic disease and [17] subsequently through different lines of systemic therapy [18– 20]. From a clinical perspective, it is challenging to obtain serial tissue samples as patients progress through various stages of disease [21]. In addition to issues related to financial cost, there are risks associated with biopsy such as bleeding and infec- tion [22]. Furthermore, biopsies may be difficult to obtain due to complex anatomic locations (i.e., bone and/or brain metastases) [22]. Circulating tu- mor DNA (ctDNA) may offer a unique tool with which to obtain serial samples mitigating some of the noted challenges with tissue biopsy [22] (Fig. 1). However, the use of ctDNA in RCC is in a relative infancy. The intent of the review is to define current techniques for assessing ctDNA, discuss the poten- tial uses of ctDNA for localized RCC, and then review the role of this modality in the metastatic setting.

Assessment of ctDNA

Multiple commercially available platforms exist to assess ctDNA, including Guardant360, FoundationAct, and GeneStrat [17, 23]. Guardant360 interro- gates ctDNA for single nucleotide variants (SNVs) in 73 cancer-related genes, indels in 23 genes, copy number amplifications (CNAs) in 18 genes, and fusions in six genes. Similarly, FoundationAct detects four distinct classes of genetic alterations in 62 cancer-related genes. In addition to commercially available platforms, many reported clinical research studies utilize hotspot panels that sequence specific genes of interest and are not commercially avail- able [24, 25]. To date, all studies of ctDNA in RCC have utilized either Guar- dant360 or hotspot panels.

As an example of how ctDNA NGS is performed, we describe the Guar- dant360 assay in more detail [26••, 27••]. The process starts with a routine blood draw (two, 10-mL Streck tubes which should have 5 to 30 ng of cfDNA), with no need of refrigeration or local centrifugation. After receipt, DNA un- dergoes library preparation for complete digital sequencing of targeted exons in 73 cancer-related genes. This methodology achieves analytical specificity of 100% for SNVs, fusions, and CNAs and 96% for indels and analytical sensitivity of 9 99.9% for SNVs, fusions, and indels, and 95% for CNAs [27••]. Clinically significant data, such as quantitative mutant allele fractions and gene copy numbers, are then reported to guide clinical decisions (Fig. 2) [26••].

Fig. 1. Outline of the potential advantages of doing ctDNA assessment versus tissue base genomic testing.

Fig. 2. The process for genomic testing.

Potential roles for ctDNA in localized RCC

While 25% of patients with RCC have metastatic disease at presentation, another 20–40% with localized disease will eventually develop mRCC [28, 29]. The timing and location of metastases in RCC are difficult to predict, which makes surveillance difficult. Earlier detection of metastatic disease may improve clinical outcomes. ctDNA has potential as a surveillance biomarker for patients with localized RCC after nephrectomy. In a study of 30 patients with RCC preparing for nephrectomy, ctDNA NGS was used to interrogate 14 commonly mutated genes [30]. Twenty of the 30 patients had detectable somatic mutations in at least one of the 14 genes assessed. This suggests that even the low tumor burdens seen in localized RCC shed detectable quantities of ctDNA.

Another study used quantitative real-time PCR to measure the level of ctDNA in 92 patients with clear-cell RCC across different stages of disease [21]. The authors found that ctDNA was higher in patients with mRCC than localized RCC (6.04 vs. 5.29, p = 0.017). They also showed that recurrence of RCC had higher levels of ctDNA (p = 0.024). These studies suggest that ctDNA can be routinely monitored at set intervals to monitor for disease recurrence. Use of ctDNA could decrease the potential harms associated with screening CT scans, including contrast nephropathy and radiation exposure.

In the setting of localized RCC, a small number of studies have looked at novel uses of ctDNA for disease monitoring. Hauser et al. found that the CPG island hypermethylation of cfDNA is common in RCC compared to healthy patients and could serve as a potential biomarker for initial diagnosis of RCC [31•]. In a separate study, investigators analyzed the concentrations of genomic and mitochondrial ctDNA concentrations across the disease stages of RCC [32]. Using a combination of genomic and mitochondrial ctDNA, they developed promising diagnostic and prognostic models for RCC [32]. Both of these novel techniques using ctDNA need further validation in larger cohorts to clearly delineate their clinical utility.

ctDNA in mRCC

To date, there are many studies on ctDNA in metastatic non-small cell lung cancer (NSCLC) and colorectal cancer (CRC) [33, 34]. However, few studies have explored ctDNA in mRCC. In metastatic NSCLC or CRC, specific targeted therapies are approved for patients harboring driver mu- tations, such as alectinib for ALK rearrangement positive NSCLC patients [35]. mRCC exists in a unique paradigm where targeted therapies, such as VEGF tyrosine kinase inhibitors or mTOR inhibitors, are routinely used in clinical practice, but no predictive biomarkers exist to guide selection of these targeted therapies. Here, we will discuss the mutational landscape of mRCC in ctDNA, how the landscape evolves with treatment, concordance of tissue and ctDNA in mRCC, and the role for ctDNA as a biomarker of response to immunotherapy.

Mutational landscape of ctDNA in mRCC

A large study of 220 patients with mRCC was the first to assess the mutational landscape of ctDNA in mRCC [36••]. In that study, genomic alterations (GAs) in ctDNA were detected using Guardant360. Of the 220 patients assessed, at least one GA was detected in 78.6% of patients, indicating that mRCC sheds detectable quantities of ctDNA. The most frequent GAs detected in ctDNA were TP53 (35%), VHL (23%), EGFR (17%), NF1 (16%), and ARID1A (12%) (Table 1). In comparison to a historical tissue NGS control, TCGA RCC study, the landscape of GAs detected in RCC was similar; however, the frequency of GAs differed between the two studies. For example, in TCGA study, less than 3% of patients had a TP53 alteration, whereas, 35% of patients had TP53 alterations in the ctDNA cohort [16]. The difference in frequency of GAs between tumor tissue and ctDNA NGS studies is likely due to differences in patient population, i.e., localized vs. metastatic. Additionally, select prognostic genes in RCC, such as PBRM1, SETD2, and BAP1, are not currently included in the G360 panel. However, this study demonstrated that ctDNA NGS can provide a contempo- rary profile of the mutations present in mRCC.

Since ctDNA NGS is capable of detecting a contemporary profile of alterations after multiple lines of treatment, ctDNA could be used to identify patients eligible for biomarker-guided clinical trials. Biomarker- guided clinical trials are already being used in mRCC and are expected to become more prevalent. For example, MET alterations are common in type 1 papillary RCC (pRCC) [41•]. In a phase II trial of savolitinib for pRCC, MET-driven pRCC showed a strong response to therapy, whereas MET- independent disease showed no response (p = 0.002) [42•]. This finding led to an ongoing phase III trial clinical trial of savolitinib vs. sunitinib in MET-driven pRCC (NCT03091192). Guardant360 detects MET alterations, so ctDNA could be used to identify patients for this biomarker-guided clinical trial.

In the study of 220 patients with mRCC by Pal et al., the evolution of ctDNA between first and subsequent lines of treatment was also assessed [36••]. The authors showed that ctDNA is a valid tool for obtaining real-time genomic data after disease progression, allowing for more informed therapeutic decisions based on the contemporary mutational profile. In mRCC, detection of an accurate mutational profile is challenging with tissue-based NGS due to the high degree of intrapatient tumor heterogeneity present. This study also con- firmed that treatment results in significant clonal evolution of ctDNA between first-line therapy and subsequent lines. For example, only 3% of patients had GAs in NF1 at time of first-line treatment, but post-first-line 21% of patients had GAs in NF1. Recognition of these GAs can lead to personalized selection of salvage-line treatments. NF1 encodes for a key regulator of the mTOR pathway and is increased in exceptional responders to everolimus [43]. These alterations may eventually occur and been identified through ctDNA with exacerbated mutation rates, and it should be considered in future researches. While the use of ctDNA to personalize treatment selection after systemic treatment is exciting, prospective validation is necessary before this is adopted for routine use in the clinic.

The only comparison of GAs detected by tumor tissue and ctDNA NGS was a study of 19 patients with mRCC [37••]. When controlling for GAs detected by both tests, the median number of GAs detected was similar between tissue and ctDNA (median 3.0 vs. 1.0, p = 0.14). However, the concordance rate between the two tests was only 8.6%. The clinical utility of the study is limited by prolonged time between tissue and ctDNA NGS, which could allow for signif- icant clonal evolution. In studies from other malignancies that controlled for time between tests, the reported concordance rate is much higher [44]. Still, the significant discordance between ctDNA and tissue NGS likely reflects intrapatient tumor heterogeneity. In a 2012 study of four patients with mRCC, biopsies from multiple sites in the same patient demonstrated significant spatial heterogeneity in somatic alterations [45]. ctDNA is likely a more accurate reflection of the intrapatient tumor heterogeneity than tissue NGS and may more accurately identify driver mutations. This would make ctDNA an ideal test to identify contemporary GAs in patients with new metastatic disease and only archival primary tumor tissue available.

Nivolumab is a monoclonal antibody against PD-1 that is currently approved for salvage-line treatment of mRCC [14]. More immune checkpoint inhibitors are anticipated to receive approval for treatment of mRCC in the near future. Recently, CheckMate 214 showed that the combination of nivolumab plus ipilimumab improved overall survival (OS) compared to sunitinib alone for first-line treatment of mRCC [46]. Furthermore, many combinations of im- mune checkpoint inhibitors and targeted therapy are being studied in the first- line setting [47]. Biomarkers predictive of superior response to immune check- point inhibitors are not routinely used in the clinic, yet they could help
personalize treatment sequencing in malignancies such as mRCC. ctDNA has shown potential as a predictive biomarker for immune checkpoint inhibitors.

Nivolumab was recently approved for the treatment of any cancer with mismatch repair deficiency [48]. Tumors with mismatch repair deficiency have exceptionally high TMB. Furthermore, a phase 2 clinical trial of atezolizumab in metastatic urothelial carcinoma showed that patients with high TMB had superior response to atezolizumab [49]. A recent study demonstrated that ctDNA can act as a surrogate for TMB [50]. In a study of 69 patients with diverse solid tumors, including three patients with mRCC, who were treated with immune checkpoint inhibitors, hypermutated ctDNA, defined as ≥ 6 GAs, was predictive of improved ORR (40.9% vs. 15.9%, p = 0.025) and PFS (2.85 vs.
2.19 m, HR = 0.59, p = 0.025). This clearly shows that more GAs in ctDNA act as a surrogate for high TMB and is predictive of response to immune checkpoint inhibitors.

A unique case report demonstrates the value of ctDNA as a predictive biomarker of response to immunotherapy in mRCC [51•]. A 44-year-old man was diagnosed with mRCC as he had a solitary metastasis in his brain. He was sequentially treated with high-dose interleukin-2, pazopanib, bevacizumab, nivolumab plus ipilimumab (discontinued due to hepatotoxicity), and cabozantinib. After progression on cabozantinib, ctDNA NGS was assessed and showed a high number of GAs, 6, so he was re-challenged with nivolumab and experienced a deep radiographic response. ctDNA NGS performed during his deep response showed that he now had no detectable GAs. This case illustrates two potential roles for ctDNA in treatment of mRCC. Consistent with previously discussed studies, a high number of GAs in ctDNA can act as a surrogate for high TMB and predict response to immunotherapy. Second, the number of GAs detected in ctDNA after initiating treatment may reflect treat- ment response. This observation requires further validation in larger cohorts.

ctDNA and radiographic disease progression

In most clinical settings, patients with mRCC are monitored for treatment response with a CT chest/abdomen/pelvis every 3 months. Recurrent CT scans are time consuming, costly, and expose cancer patients to high levels of radia- tion. ctDNA has the potential to act as a surrogate for radiographic disease progression in mRCC. In a recent study of 34 patients with mRCC, patients with detectable ctDNA had higher radiographic tumor burden than patients with no detectable ctDNA (p = 0.01) [52•]. While these findings suggest that ctDNA has the capability to complement or replace frequent CT scans in patients with mRCC, these results are from a small, retrospective study and warrant further evaluation in larger, prospective cohorts.

Conclusion

NGS of ctDNA is a novel technology that has the potential to alter clinical practice by providing a contemporary profile of a tumor’s mutational land- scape. While there are extensive studies of ctDNA in NSCLC and CRC, experience with ctDNA in RCC is still in its infancy. In localized RCC, ctDNA has potential as a screening tool for recurrence of disease after nephrectomy. In mRCC, ctDNA has promise as a predictive biomarker for response to immunotherapy by acting as a surrogate for TMB. Furthermore, ctDNA provides insight into clonal evaluation after systemic treatment. Across all stages of RCC, liquid biopsies are attractive to patients and providers as they are less invasive than tumor tissue NGS and can be serially repeated without significant harm.