Mutational profile of primary clear cell renal cell carcinoma predicts recurrence and potential candidacy for adjuvant immune checkpoint inhibition

Introduction

Renal cell carcinoma (RCC) is a heterogenous group of kidney cancers originating from the nephron.¹ Renal cell carcinoma ranks among the ten most frequently diagnosed malignancies worldwide, with an estimated 400.000 new diagnoses and more than 170.000 deaths annually.^2,3 Clear cell RCC (ccRCC) is the most common RCC subtype, accounting for about 70-75% of cases and has a distinct molecular profile.^1,4

Localized ccRCC is treated with surgical resection, either partial or radical nephrectomy. Nevertheless, one-third of these patients experience recurrence.⁵ So far, ccRCC recurrence, disease progression and mortality are being predicted using clinicopathological criteria.⁶ While various recurrence models have been proposed, they only marginally outperformed standard staging.⁷ Further, they demonstrate statistically significant variability in their predictive ability over time, rendering implementation into clinical practice and clinical trial design challenging.⁷

Immune checkpoint inhibition with the use of pembrolizumab, a monoclonal antibody against programmed death-1 (PD-1), is approved as adjuvant therapy for patients with resected ccRCC who have a high risk of recurrence.⁸ This was based on results of a phase III randomized double-blind study comparing pembrolizumab with placebo, which demonstrated a significant improvement in disease-free survival (DFS).⁹ Patient selection criteria for a high risk of recurrence included tumor stage II with nuclear grade 4 or sarcomatoid differentiation, tumor stage III or higher, regional lymph-node metastasis, or stage M1 without evidence of disease after combined nephrectomy and metastasectomy either concurrently or within a year from primary tumor resection.⁹

What currently remains an unmet need is the ability to predict which patients with ccRCC will relapse using a single or composite molecular biomarker that would be more directly related with tumor biology.

In this ongoing prospective study, we examined the mutational profile of patients with non-metastatic ccRCC who underwent nephrectomy, followed by observation or adjuvant immunotherapy with pembrolizumab depending on established clinical and histopathological criteria. We studied associations of mutated genes with high-risk features and assessed the prognostic relevance of somatic mutations with regard to DFS after nephrectomy with or without adjuvant immunotherapy.

Methods

Results

Mutations in non-VHL genes are associated with more aggressive disease

We first assessed the frequency of mutations in primary ccRCC tumors. In the discovery cohort (n=37), patients’ clinical and histopathological characteristics are described in Table 1. VHL was the most frequently mutated gene (n=19; 51%), followed by PBRM1 (n=10; 27%), BAP1 (n=5; 13%), SETD2 (n=5; 13%), KDM5C (n=2, 5%), ATM (n=2, 5%), MTOR (n=2, 5%), and PTEN (n=1, 3%) (Table 2). Variant types per gene are listed in the data file.³⁵ 11 patients (30%) did not have any somatic mutations within the 15-gene targeted panel.

Table 1. Clinical and histopathological characteristics of ccRCC patients (discovery cohort, n=37).

Characteristic	Number (%)
Age
median (range)	63 (42-87)
Sex
males	28 (75)
Tumor diameter (cm)
median (range)	5.6 (1.6 – 15)
Pathological T stage
T1	21 (57)
T2	6 (16)
T3	10 (27)
ISUP grade
2	13 (35)
3	15 (41)
4	8 (24)
Vascular invasion	5 (14)
Necrosis	13 (35)
Sarcomatoid differentiation	0
AJCC TNM stage
I	21 (57)
II	6 (16)
III	10 (27)

Table 2. Frequency of mutated genes in primary ccRCC tumors (discovery cohort, n=37).

Gene	Frequency, n (%)
VHL	19 (51)
PBRM1	10 (27)
BAP1	5 (13)
SETD2	5 (13)
KDM5C	2 (5)
ATM	2 (5)
MTOR	2 (5)
PTEN	1 (3)

Tumors harboring no mutations at all or only VHL mutations (n=19, 51%) were associated with smaller size (pT1-2 n=17, 89%) and earlier stage (I/II n=17; 89%), whereas presence of any other gene mutations in various combinations with or without VHL was enriched in larger (pT3, n=8; 44%; p=0.02) and more advanced tumors (III, n=8; 44%; p=0.02) (Table 3). There was also a trend towards higher frequency of ISUP grade, vascular invasion, and necrosis in these tumors (Table 3).

Table 3. Associations between histopathological characteristics and mutational profile (discovery cohort, n=37).

Variable	Mutated genes		P value
	None or VHL-only	Non-VHL
Pathological T stage			0.02
T1/T2	17 (89)	8 (44)
T3	2 (11)	10 (56)
AJCC TNM stage			0.02
I/II	17 (89)	8 (44)
III	2 (11)	10 (56)
ISUP grade			0.93
2/3	15 (79)	14 (78)
4	4 (21)	4 (22)
Necrosis			0.64
no	13 (68)	11 (61)
yes	6 (32)	7 (39)
Vascular invasion			0.58
no	17 (89)	15 (83)
yes	2 (11)	3 (17)

In the validation TCGA cohort (n=451), similar mutations frequencies were noted, including VHL in 50% of patients/samples, PBRRM1 in 29%, SETD2 in 11%, BAP1 in 9%, MTOR in 7%, KDM5C in 6%, PTEN in 4%, and ATM in 2.7%, respectively (Figure 1). Larger and higher stage tumors, particularly T3a, T3b, and stage III tended to have a higher frequency of non-VHL mutations (p=0.282; q=0.437) (Figure 2).

Figure 1. Frequency of mutated genes in primary ccRCC tumors (validation cohort, n=451).

Figure 2. Associations between histopathological characteristics and mutational profile (validation cohort, n=451).

Mutations in non-VHL genes predict recurrence

No recurrences were noted in patients with unmutated tumors or VHL-only mutations whereas three patients (17%) with other somatic mutations relapsed (p=0.06) (Table 4). Another three patients received and completed adjuvant PD-1 inhibition with pembrolizumab until present, none of whom recurred despite the presence of PBRM1 and SETD2 mutations in 2/3 and 1/3, respectively. Presence of somatic mutations in PBRM1, BAP1, SETD2, KDM5C, ATM, MTOR, or PTEN genes in 451 TCGA ccRCC patients was associated with a significantly shorter DFS compared to those with unaltered tumors (p<0.001; q=0.01) (Figure 3).

Table 4. Associations of mutational profile (none or VHL-only versus non-VHL mutated genes) and disease recurrence.

Outcome	Mutated genes		P value
	None or VHL-only	Non-VHL
Recurrence			0.06
no	19 (100)	15 (83)
yes	0 (0)	3 (17)

Figure 3. Kaplan-Meier analysis of DFS in patients with unmutated or VHL-only versus non-VHL mutated tumors (validation cohort, n=451).

Discussion

Cancer is a disease caused by the accumulation of genomic alterations.¹¹ This is a multistep process during which certain somatic mutations confer a survival advantage by activating signaling pathways that are associated with several hallmarks of cancer, including proliferative signaling, evasion of growth suppression, resistance to cell death, replicative immortality, angiogenesis, invasion and metastasis, reprogramming of energy metabolism and evasion of immune recognition and destruction.¹² Since these biological principles apply to the majority of solid tumors, including RCCs, we hypothesized that the higher the number of pathogenic mutations within a primary renal tumor, the more aggressive clinical behavior could be observed.

The tumor suppressor VHL is the most frequently mutated gene in ccRCC and is a major player in renal cell carcinogenesis. However, VHL mutations alone are insufficient to drive disease progression, and have no prognostic or predictive value.^13,14 Mounting evidence has revealed an emerging role of other genes, heavily involved in chromatin rearrangement and epigenetic DNA modifications, including PBRM1, SETD2, BAP1, and KDM5C in ccRCC progression.^15,16 Inactivating mutations or/and low expression of these genes in primary renal tumors have been associated with poor outcomes.^17,23

Interestingly, many of these frequently non-VHL genes, including PBRM1, BAP1, SETD2, have been investigated in the metastatic setting as potential predictors of response or resistance to immune checkpoint inhibitors (ICIs). For example, mutated PBRM1 was included in a composite biomarker together with tumor infiltrating lymphocytes (TILs) and absence of necrosis predicting a favorable response to ICIs.^24,25 Likewise, SETD2 loss results in greater vulnerability to immune checkpoint blockade compared to SETD2-proficient tumors, in vitro and in vivo.²⁶ BAP1-mutated tumors, while they portend a worse prognosis, they are more likely to be PD-L1 positive and demonstrate a more inflamed immune microenvironment suggesting that immune-targeting approaches could benefit these patients.²⁷ Another DNA damage response and repair gene, ATM, was the second most frequently altered gene in ICI-responders with advanced RCC.²⁸ Pancancer analyses of tumors from ICI-treated patients suggested that those with KDM5C alterations have a substantially higher tumor mutational burden (TMB) level and a significantly higher level of CD8+ T cell infiltration and T effector signature which were associated with prolonged OS compared to the wild-type group.²⁹ Expression of PTEN/PI3K/mTOR pathway genes was significantly associated with numerous immune cells and immune-evasive mediators such as CD274/PD-L1, TGFBR1, CSF1R and PDCD1 in patients with ccRCC,³⁰ suggesting that alterations in this pathway could also play a role in shaping these patients’ outcomes after treatment with ICIs.

Our ongoing prospective study examined the mutational profile of patients with ccRCC on primary tumors after nephrectomy followed by observation or adjuvant immunotherapy with pembrolizumab and assessed for associations of mutated genes with high-risk features and DFS. We hypothesized that primary RCC tumors either umutated or harboring VHL mutations might be associated with a more benign clinical course after nephrectomy compared to those with a burden of mutations in other, non-VHL genes.

In this preliminary analysis of the first 37 patients, non-VHL mutations, including mutations in PBRM1, SETD2, BAP1, KDM5C, MTOR, PTEN, or ATM genes in a targeted 15-gene NGS panel were significantly associated with more aggressive histopathological characteristics including larger size, and higher stage. Presence of mutations in any of those genes as opposed to completely unmutated or VHL-only mutated tumors was also associated with higher ISUP grade, necrosis, sarcomatoid differentiation, vascular invasion and predicted recurrence. In ccRCC tumors from TCGA, non-VHL mutations in the same genes were also associated with high-risk features and predicted a significantly shorter DFS compared to VHL-only mutations or complete absence thereof.

Due to the retrospective nature of previous studies and lack of reproducibility, particularly across immunohistochemical assessments^31,32 in an era when observation was the only available modality post-operatively even in high-risk patients, there has been a paucity of data to support testing of these genes as a molecular tool to assist in selection of patients who might benefit from adjuvant therapy.

Our ongoing prospective study addresses this gap by demonstrating that patients who had either no mutation or mutations in the most frequently altered gene, VHL, were more likely to have smaller tumors and experienced a more benign course without relapse, compared to those patients with tumors that harbored mutations in other genes, including PBRM1, BAP1, SETD2, KDM5C, ATM, MTOR, or PTEN. Thus, this study suggests that patients with mutations in these high-risk genes might be more suitable candidates and should be prioritized for post-operative immunotherapy with pembrolizumab, which has demonstrated clinical benefit in PFS and OS and is currently approved as adjuvant treatment.³³

Our findings are in line with a previous work with slightly different design, whereby primary RCC tumors were segregated into VHL+0, VHL+1, VHL+2, and VHL+≥3 mutations.³³ In both the discovery and validation cohorts of the study, those patients with a VHL+0 tumor had longer 5-year DFS and were proposed as candidates for surveillance. Conversely, patients with VHL+2 and VHL+≥3 tumors experienced shorter DFS rates of less than 50% and were deemed candidates for adjuvant therapy.³⁴

Our study was limited by small size and relatively short follow-up of patients. Additionally, the lack of transcriptional or/and epigenetic analyses is another limitation of our study. Nevertheless, presence of an early “signal” of high-risk genes in this preliminary report will be further studied in additional patients being accrued as part of this ongoing prospective study.

Collectively, our study combined with emerging evidence on the genomic landscape of RCC might open new avenues for both prognostication and better selection of a subgroup of patients with RCC that could benefit from adjuvant anti-PD1 immunotherapy.