Wide spetcrum mutational analysis of metastatic renal cell cancer: a retrospective next generation sequencing approach

Renal cell cancer (RCC) is characterized by histological and molecular heterogeneity that may account for variable response to targeted therapies. We evaluated retrospectively with a next generation sequencing (NGS) approach using a pre-designed cancer panel the mutation burden of 32 lesions from 22 metastatic RCC patients treated with at least one tyrosine kinase or mTOR inhibitor. We identified mutations in the VHL, PTEN, JAK3, MET, ERBB4, APC, CDKN2A, FGFR3, EGFR, RB1, TP53 genes. Somatic alterations were correlated with response to therapy. Most mutations hit VHL1 (31,8%) followed by PTEN (13,6%), JAK3, FGFR and TP53 (9% each). Eight (36%) patients were wild-type at least for the genes included in the panel. A genotype concordance between primary RCC and its secondary lesion was found in 3/6 cases. Patients were treated with Sorafenib, Sunitinib and Temsirolimus with partial responses in 4 (18,2%) and disease stabilization in 7 (31,8%). Among the 4 partial responders, 1 (25%) was wild-type and 3 (75%) harbored different VHL1 variants. Among the 7 patients with disease stabilization 2 (29%) were wild-type, 2 (29%) PTEN mutated, and single patients (14% each) displayed mutations in VHL1, JAK3 and APC/CDKN2A. Among the 11 non-responders 7 (64%) were wild-type, 2 (18%) were p53 mutated and 2 (18%) VHL1 mutated. No significant associations were found among RCC histotype, mutation variants and response to therapies. In the absence of predictive biomarkers for metastatic RCC treatment, a NGS approach may address single patients to basket clinical trials according to actionable molecular specific alterations.


INTRODUCTION
Renal cell carcinoma (RCC) accounts for the 2-3% of all adult malignancies, the seventh most common cancer in men and the ninth in women. [1] Localized RCC can be successfully cured by surgery in most cases. Unfortunately,

Research Paper: Pathology
Oncotarget 7329 www.impactjournals.com/oncotarget about 30% of patients relapse after nephrectomy or present at diagnosis with metastatic disease (mRCC) [2]. Despite the therapeutic improvements made in the past decade, mRCC is still an incurable disease. [3] Several studies outlined that the multiple histological features of RCC corresponded to a molecular intra-and inter-tumoral heterogeneity with prognostic and predictive implications. [3,4,5,6,7] The improved knowledge of the biology and pathogenesis of RCC, lead to major treatment advances through the development of targeted agents. [2] The currently available therapeutic options for mRCC include drugs targeting circulating vascular endothelial growth factor (VEGF) and its receptors (VEGFRs), other tyrosine kinase inhibitors (TKIs) and inhibitors of the mammalian target of rapamycin (mTOR) serine-threonine kinase [2]. Unfortunately, 20 to 30% of mRCC patients do not respond at all to targeted agents while the remaining ultimately progress after an initial benefit. [8] There is current need for molecular biomarkers of RCC for the development of a personalized treatment and predictive approach for patient stratification. [8] According to the Cancer Genome Atlas database the most frequent somatic mutations in ccRCC include mainly alterations of the VHL gene and its partners involved in neo-angiogenesis and response to hypoxia, followed by alterations of the PI(3)K/AKT/mTOR pathway. [9] Conversely, the most frequent somatic genetic changes in non-ccRCC such as the papillary histotype (pRCC) involve mutations or copy number variations of the MET oncogene or loss of the tumor suppressor gene CDKN2A. [3,10] Unfortunately, no correlation was found so far among biomarker expression/alteration and clinical response to single TKIs and mTOR inhibitors. In addition, there is no clear-cut evidence that the ccRCC or the non-ccRCC histotypes respond differently to specific target therapies [9,11] In this study, we have evaluated retrospectively with a next generation sequencing (NGS) approach the mutation burden of 32 primary and/or secondary lesions of 22 patients with mRCC treated with at least one TKI or mTOR inhibitor. We then associated the presence of every identified somatic tumor alteration with response to therapy.

Patients' selection criteria
We have retrospectively analyzed 32 RCC nodules/ metastases from 22 patients with mRCC enrolled and surgically resected at the S.Orsola-Malpighi Hospital (Bologna) and the Policlinico GB Rossi Hospital (Verona) from 2001 to 2013. Informed consent for genetic analyses was obtained from the patients before surgery.
Selection criteria for patients' enrolment were: 1) presence of mRCC at the time of surgery or development of metastases thereafter; 2) Availability of at least one formalin-fixed paraffin-embedded (FFPE) surgical pathology cancer specimen; 3) Systemic treatment with at least one of the anticancer agents registered in Italy for mRCC; 4) availability of at least 6 month follow-up from the beginning of the first line of treatment. When available, tissues from the primary and metastatic tumor sites were analyzed separately. In addition, histological areas of sarcomatoid or rhabdoid change within the same tumor nodule were also analyzed individually. Table  1 summarizes the clinical-pathological features of the selected cases.

Samples' preparation and NGS sequencing
Pathological slides from all patients and all tumor sites have been reviewed blindly by four genito-urinary dedicated pathologists (MB, MF, FG and GM) for independent assessment of RCC histotype, grading and staging, according to the 2016 WHO classification of tumours of the urinary system and male genital organs [12]. Histological diagnosis was confirmed by routine immunohistochemistry in all cases using antibodies anti CA-IX and RCC for ccRCC and anti CK7 and Racemase for pRCC. Histotype call was performed blindly by at least two pathologists with inter-observer variability <5%. (data not shown)All the slides of each case have been pathologically reviewed and the most representative block with highest tumor cell enrichment and highest nucleolar grade was selected. Tumor areas of interest with at least 70% tumor cell enrichment were circled and 10 µm-thick serial sections of the same paraffin block cut in sterility for DNA extraction. Sections were manually microdissected, deparaffinized in xylene and the DNA was isolated using the GeneRead DNA FFPE Kit (Qiagen -Hilden, Germany). DNA was quantified with the Quantifiler ® Human DNA Quantification Kit (Thermo Fisher Scientific -Waltham, MA). The NGS analysis was performed on an Ion PGM™ System (Thermo Fisher Scientific) platform. An amplicon library was produced from 10ng of DNA from each sample using the Ion AmpliSeq™Cancer . Amplification of target sequences was followed by barcode adapter ligation (Ion Xpress Barcode Adapters) to the amplicons that were then purified using Agencourt AMPure XP (Beckman Coulter -Brea, California). Library quantification was performed using the Ion Library TaqMan™ Quantitation Kit. The library was diluted in nuclease-free water to obtain a final concentration of 8pM. Emulsion PCR was performed using Ion PGM™ Template OT2 200 Kit on www.impactjournals.com/oncotarget the Ion OneTouch™ 2 instrument. Library quality was assessed using the Qubit Ion Sphere™ Quality Control Kit. Selective ion spheres (ISPs) with DNA were isolated (Ion PGM™ Enrichment Beads on Ion OneTouch™ ES instrument) and sequenced on a Ion 316™ Chip Kit v2 (5 samples/chip) or a Ion 318™ Chip Kit v2 (10 samples/ chip) using the Ion PGM™ Sequencing 200 Kit v2. Successful sequencing of a sample required at least 500 000 reads with a quality score ≥ Q20. As tumor specimens were admixed with normal tissue, a minimum coverage of 500X with at least 10% frequency was used as cut-off for a variant to be considered true. The following stringent criteria were applied for final variant calling: i) non-synonymous coding; ii) allele frequency of ≥10%; iii) total amplicon coverage of ≥500 reads; iv) a Phred-based quality score of 30 or more and P-value <0.01. The Ion Reporter software incorporated the following databases: ClinVar, dpSNP (National Center for Biotechnology Information, Bethesda, Maryland) and COSMIC (Wellcome Trust Sanger Institute, Cambridge, United Kingdom) databases. The in-silico prediction tools SIFT, PolyPhen, PhyloP and Grantham were also included. The Integrative Genomics Viewer (IGV, Broad Institute) was used to visualize variants.

Clinical-pathological data
Seventeen (85%) of the 22 mRCC patients were males and five (15%) females. The mean age at diagnosis was 57±19. Twenty-one of the 31 RCC samples were from the primary tumor site and 11 from metastatic lesions. The mean time to the development of metastases from the removal of the primary tumor was 13.5 ±14.27 months (range from 0, synchronous, to 53 months). RCC histotypes were distributed as follows: 19 (86%) ccRCC and 4 (18%) papillary RCC (pRCC).. One patient developed two nodules with ccRCC and pRCC histology in the same kidney and the lesions were analyzed separately. Metastatic sites included lymph-nodes, diaphragm, liver, pancreas, omentum, lung and adrenal.   Both the primary and the metastatic sites were analyzed for 6 patients. The rhabdoid and the sarcomatoid tumor components were present and have been tested in 4 and in 1 sample respectively (Table 1).

Mutation variants are different, non-mutually exclusive, and variably associated with pathological features
The mean read length of the NGS was 105 bp and the average reads were approximately 65Mb of sequence per sample. With normalization to 2000 reads per specimen, 97.08% amplicons averaged at least 500 reads. A minimum coverage of 500X with at least 10% frequency was used as cutoff for a variant to be considered true. We identified mutations in the genes VHL, PTEN, JAK3, MET, ERBB4, APC, CDKN2A, FGFR3, EGFR, RB1 and TP53. Table 1 summarizes the clinical-pathological patients' characteristics according to mutation variants and the line of therapy. In particular, the most frequent mutations (7/22 patients 31,8%) affected the VHL1 gene, all were in different loci and except for 3 cases were always associated with other mutations. VHL1 was mutated in 6 primary RCC, in 2 metastatic sites and in 1 sarcomatoid and 2 rhabdoid components of primary RCCs. The VHL1 mutation variant was concordant between primary and the corresponding metastasis in 1 out of 2 cases. PTEN was found mutated in 3/22 (13,6%) patients and only in 1 case was associated with mutations of VHL 1 and FGFR3. Two cases harbored the same PTEN variant (L320S). Two patients (2/22) ( 9%) showed the same JAK3 mutation variant (V722I); it was associated with a VHL 1 mutation and was revealed both in the sarcomatoid and rhadboid component of the same patient, while in the other patient it was present as a single mutation in the primary RCC and the metastatic lesion. Two patients (9%) harboured a mutation of the FGFR3 gene (F386L); in both cases it was associated with a VHL1 mutation (1 cases also with a PTEN mutation) and each patient presented respectively the mutation both in the primary RCC and the corresponding metastasis and the clear cell and in the rhabdoid component. TP53 was found mutated in 2/22 (9%) patients:1 in a metastatic lesion and 1 in the rhabdoid component of a primary RCC. Mutation in MET gene was found in 1/22 patients (4,5%) in the primary RCC and in association with VHL1 mutation. ERBB4 mutation were discovered in 1/22 (4,5%) patients in both primary and metastatic RCC sites, in association with VHL1 mutation. Double mutations in the APC and DCKN2A genes was found in the same patient in the primary RCC. Mutations of EGFR and RB1genes were both detected in 1 patients in the rhabdoid component of the primary RCC.
In 9/22 (20%) patients the tumors were wild type at least for the investigated genes. There was no relevant association between RCC histotype and mutation variants. In particular the four pRCC harbored mutations in the APC, DCKN2A, PTEN and TP53 genes but not in VHL1. Eight (42%) out of the remaining 19 ccRCC tumors habored at least one VHL1 mutation although with different variants.
As depicted in Table 3, the concordance between the primary RCC and its rhadboid/sarcomatoid component was seen in 2/3 cases: FGFR3 F386L and VHL1 V13fs*7 (rhabdoid and primary RCC) and VHL1 P97R and JAK3 (rhabdoid and sarcomatoid component of RCC). In the remaining case the gene mutations between the RCC and its rhabdoid component were completely different. The concordance between RCC and its metastatic lesion was found in 3/6 cases while in the remaining case the genetic alterations were different. Table 1 depicts response to therapy for each patient while Table 2 summarizes the mutation variants found by gene, tumor site, and rhabdoid or sarcomatoid component.
The mean follow-up time was 27,5 months (range 6-78). Four of the 22 (18,2%) patients were alive at the time of last follow-up and only 1 was lost from follow up after 30 months. Of these 5 long survivors 4 (80%) were treated with Sunitinib and 1 (20%) with Sorafenib. Two patients harbouring just the same PTEN L320S mutation survived for at least 27 or 30 months.

DISCUSSION
Renal cell cancer is as an heterogenous disease based on its morphological and molecular features.
The ccRCC is largely the most common histotype and it is tightly associated with alteration of the VHL1-HIF pathway, while non-clear cell tumors are characterized by alterations of the PI3K and MET pathways. However, a large morphological and molecular inter-tumor heterogeneity may occur within the same histotype and intra-tumor in primary and metastatic lesions of the same patient. [4] Here we provided one of the few available reports on wide spectrum mutation analysis on mRCC on an NGS platform. Although we did not find a significant association between histotype and mutation variants we found that approximately 40% of ccRCC harboured mutations of VHL1 gene compared to none in the pRCC group. We found a large number of mutations in different genes. Mutations in some genes such as VHL1, PTEN, JAK3 and TP53 are well recognized. [13,14,15] By contrast, mutations in the genes APC, ERBB4, RB1, EGFR, FGFR3 have been rarely or never found in RCC but were encountered in our series. [16] As far as we can say from our data these mutations are non-mutually exclusive and hit a variable number of loci. In addition, we did not find any significant correlation between sarcomatoid or rhabdoid RCC features and specific gene alterations while concordance between primary RCC and the corresponding metastasis was found in three of six cases. Taken together all these data bring to the conclusion that more gene variants than expected are associated with mRCC, and are non-mutually exclusive. Mutations of VHL1 are the most frequent but involve wide number of loci.
Major limitations affect the present study. The limited number of cases and the even smaller number of patients with available primary tumors and metastases do not allow drawing definitive conclusions on the intratumor heterogeneity in terms of both mixed histotypes or primary/metastases variability. In addition, our mutational analysis is based on a panel generating 207 amplicons covering approximately 2,800 COSMIC mutations from 50 of the most altered oncogenes and tumor suppressor genes in solid tumors. We may have missed mutations in the genes outside the list of our panel as well as genetic rearrangements detectable just at the RNA level. Larger studies covering all the RCC histotypes with paired samples of primary tumors and metastases are required to answer the intra-tumor heterogeneity question.
Since 70 to 80% of patients obtain a benefit from a first line therapy, a large (and presently increasing) number of mRCC patients are treated sequentially with different agents. [17] Unfortunately, there are no current validated predictive biomarkers of response to these target therapies. Our retrospective analysis confirmed that there are no single predictors of response to therapy with TKIs or mTOR inhibitors. In fact, we found that three of the four patients experiencing partial response after first-line therapy were treated with a TKI with anti-angiogenetic