Abstract
Background/Aim: The aim of this study was to examine clonal heterogeneity, to test the utility of liquid biopsy in monitoring disease progression and to evaluate the usefulness of ex vivo drug screening in a BRAF L597Q-mutated colorectal cancer (CRC) patient developing metastases during adjuvant therapy. Materials and Methods: Next generation sequencing (NGS) and droplet digital PCR (ddPCR) were performed in samples from tumor tissues and liquid biopsies. Live cancer cells from a metastatic lesion were used in ex vivo drug sensitivity assays. Results: We found evidence of continued dependence of MEK/MAPK pathway activation, but different activating mutations in primary tumor and metastases. Liquid biopsy based BRAF L597Q ddPCR testing was a sensitive personalized biomarker predicting the rise of clinically aggressive metastatic disease. Ex vivo drug sensitivity assays with BRAF L597Q mutated cells showed response to MEK/MAPK targeted therapies. Conclusion: The rare BRAF L597Q mutation may be associated with aggressive tumor behavior in CRC. Liquid biopsy can be used to capture clinically relevant tumor features.
Understanding oncogenic drivers and molecular pathways promoting cancer development, progression and drug resistance is crucially important for precision oncology. As tissue-based molecular pathology analyses are subject to sampling bias, provide only a snapshot of tumor heterogeneity, and cannot be obtained repeatedly, liquid biopsies are suggested as an alternative to tissue-based analytics. Circulating cell-free tumor DNA (ctDNA) assays have been shown to capture many of the clinically relevant molecular features of cancer. However, it is still unclear how the presence of rare variants should be interpreted and whether sensitivity of ctDNA assays is sufficient for detection of actionable alterations when tumor burden is low (1-3).
In recent years, it has become clear that MEK/MAPK-pathway is frequently activated in the pathogenesis of colorectal cancer (CRC) (4). The hotspot BRAF mutation V600E leads to very high and constitutive BRAF kinase activity and sustained MAPK pathway signaling.
The prevalence of the BRAF V600E mutation is 5-10% in metastatic colorectal cancer (mCRC) and it is a biomarker for poor prognosis (5). In BRAF V600E–mutant mCRC, patients median overall survival was only 10.4 months compared to 34.7 months in patients without this mutation (6). Increasing use of comprehensive tumor genomic profiling with next generation sequencing (NGS) technology has led to the detection of many BRAF nonV600E mutations. Their prevalence in mCRC is reported to range from 1.6% to 5.1% (7-11). The BRAF V600E mutation still represents 60% to 80% of all BRAF-mutations (11, 12).
Clinical and prognostic implications of rare BRAF mutations are largely unknown. Based on in vitro kinase activity data, BRAF-mutations can be classified into activating, intermediate activating and impaired activity subgroups, as compared to wild type-BRAF (13, 14). These data indicate that there are different types of BRAF-mutations, not all leading to a similar phenotype. Until now (14), little attention has been devoted to the functional characterization of rare mutations and their potential implications in the treatment of CRC.
Today, multiple BRAF-directed targeted therapies have been adopted into standard clinical practice. Unlike melanoma (15), BRAF and MEK-inhibitors used as a monotherapy show modest responses in BRAF V600E mCRC. In metastatic melanoma the response rate is up to 50%, whereas only dismal 5% response rates are reported in mCRC. However, combining EGFR-antibody with BRAF-MEK-inhibitor doublet in BRAF V600E mutant mCRC patients finally resulted in an overall response rate of 48% for the binimetinib-encorafenib-cetuximab triplet. This led to the first FDA-approval for BRAF-inhibitors alone or in combinations in BRAF-mutated mCRC (16).
Here, we describe the molecular analysis and ex vivo functional characterization of a BRAF L597Q-mutated colorectal cancer patient that developed metastatic disease during standard adjuvant therapy.
Materials and Methods
Case history. A 74-year-old female with no significant past medical history presented with severe anaemia (Hemoglobin 58 g/l). Colonoscopy biopsy samples revealed an adenocarcinoma located in the ascending colon. No unequivocal metastases were observed in computer tomography (CT) scan and preoperative carcinoembryonic antigen (CEA) was low (3.1 μg/l). Elective hemicolectomy was performed in February 2017. Histological examination revealed a grade II adenocarcinoma with lymphatic invasion, and lymph node status 3/16, staging pT3N1b. An adjuvant therapy with planned eight cycles of capecitabine-oxaliplatin (CAPOX) was initiated. A peritoneal lesion was found in a CT-scan 2 months after primary surgery and after the first chemotherapy cycle. Retrospective review showed that this lesion was detectable also in the preoperative CT scan. Nevertheless, this lesion was not metabolically active in the fludeoxyglucose(18F) positron emission tomography-computed tomography (PET-CT) scan 6 months after the surgery and showed no growth tendency. Instead, the PET-CT scan demonstrated a new metabolically active lymph node in the para-aortal area. After the seventh cycle of CAPOX (7 months after primary surgery) the active lymph node was removed and histologically verified as metastatic adenocarcinoma. One month later a new lesion in the right ovary was detected in CT with the same histology. During severe abdominal pain period, a metastasis infiltrating vena cava and multiple pulmonary metastases were detected. No further chemotherapy was initiated due to the rapid progression and patient deterioration. She died about a year after the initial diagnosis of colon cancer.
Study design. The patient was recruited in a prospective study that is designed for colorectal cancer patients resected with curative intent in Tampere University Hospital in Finland. The study was approved by the ethics committee of Tampere University Hospital (R15085) and the patient provided written informed consent. The trial identifier is NCT03189576. The drug sensitivity and resistance testing were implemented within another clinical trial recruiting in Central Finland Health Care District (DNRO 3U/2015).
Blood samples for ctDNA and CEA were planned to be collected preoperatively (0-3 weeks before), at 4 weeks and every 3 months up to 2 years and every 6 months up to 3 years or until trial withdrawal or death. Altogether, five plasma samples were obtained from this patient before her death.
Algorithm for classifying ctDNA status. Plasma collection: Peripheral whole blood (2×10 ml) was collected into Vacuette Blood Collection Tubes (Bio-Greiner GmbH, Kremsmünster, Austria). Plasma was separated by centrifugation for 10 min at 1,100 × g. Separated plasma was then stored at −70°C until cfDNA extraction.
DNA extraction from plasma: Plasma cfDNA was extracted from 4 ml of blood plasma using QIAamp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany). Extraction was performed according to the protocol provided by the kit manufacturer.
DNA extraction from tissue samples: DNA was extracted from representative FFPE tissue sections with QIAamp DNA FFPE Tissue Kit (Qiagen) according to the protocol provided by the kit manufacturer.
NGS: Qiaseq Human Comprehensive Cancer Panel (Qiagen) including 275 cancer related genes was used to prepare NGS amplicon gene library according to the protocol provided by kit manufacturer. Unique molecular identifiers (UMI) were used to tag individual DNA strands. Sequencing was performed with Illumina NextSeq500 instrument (Illumina, San Diego, CA, USA) according to standard protocol. Data was demultiplexed and fastq files created with bcl2fastq software (Illumina). The data was processed in CLC Biomedical Genomics Workbech (Qiagen) with workflow provided by Qiagen and using Hg19 human reference genome to call the gene variants. Gene annotations were performed according to the vcf files in OmnomicsNGS software (Euformatics, Espoo, Finland).
Droplet digital PCR: To detect the mutation BRAF c.1790T>A (p.Leu597Gln), droplet digital PCR (ddPCR) testing was performed in cfDNA samples. PCR mix contained 5 μl cfDNA, 1× ddPCR Supermix for probes (no UTP; Bio-Rad, Hercules, CA, USA), 1× target (FAM) and wild-type (HEX) primers/probes (Bio-Rad), 2 U Hae III (New England Biolabs, Ipswich MA, USA) and water in a 20 μl volume. Droplets were then prepared with QX200 Droplet Generator (Bio-Rad) and then the reactions were subjected to PCR cycling. The PCR cycling included a 95°C enzyme activation step for 10 min followed by 40 cycles of a two-step cycling protocol (94°C for 30 sec and 55°C for 1 min) and finally a 10 min enzyme inactivation step at 98°C. The ramp rate between these steps was slowed to 2°C/sec. After PCR, the droplets were counted using QX200 Droplet Reader (Bio-Rad) and data analysis was performed with QuantaSoft software (Bio-Rad). Each sample was prepared in duplicates. Mutation positive and negative samples were prepared as controls for test performance.
Drug sensitivity and resistance testing. A sample of metachronous para-aortal lymph node metastasis was dissected into three pieces by a pathologist: 1) formalin fixed sample for histological evaluation, 2) liquid nitrogen frozen sample for sequencing and 3) fresh tissue for drug testing.
For drug screening, tumour sample was rinsed with cold Hank's balanced salt solution (HyClone HBSS GEHealthCare Life Sciences, Marlborough, MA, USA) and dissected into 1 mm3 pieces with scalpel. Pieces were further digested enzymatically into a uniform cell suspension by incubating for 2.5 h in HBSS with 1 U/ml Dispase (Corning Bedford, MA, USA), after which the enzyme was inactivated by repeated washing and centrifugation of cells in RPMI-1640 cell culture medium (Lonza Ltd, Basel, Switzerland) supplemented with 5% FCS (Gibco by Life Technologies, Waltham, MA, USA). Cells were seeded on ready-made 384-well drug plates that contained altogether 146 FDA-approved and investigational drugs in four different concentrations. The ex vivo drug screening library drugs were purchased from a commercial vendor (Cat no L1300, Selleck Chemicals, Houston, TX, USA). Cells were incubated for 5 days in the presence of drugs in four 2-fold dilutions with 5 μM as the highest concentration. DMSO (Amresco Inc., Solon, OH, USA) only samples were used as negative control, 5 μM Staurosporin (Fisher Scientific, Waltham, MA, USA) as cell death control and 2 μM Aphidicolin (MP Biomedicals, Santa Ana, CA, USA) was used for normalization of growth rate. Cell viability was measured by using a CellTiter-Glo reagent (CTG, Promega Corporation, Madison, WI, USA) and a luminescence reader (Labrox, Turku, Finland). The drug response data were analysed with Microsoft Excel using the normalized growth rate inhibition (GR) approach, which yields per-division metrics for drug potency (17, 18). Absolute IC50 values were calculated with GraphPad Prism 7 software using a non-linear curve fit equation modified using previously described parameters (19).
Drug sensitivity testing (cell line derived from metachronous para-aortal lymph node metastasis). Adherent cell line derived from primary sample was cultured in RPMI-1640 medium (Lonza) containing 5% FCS (Gibco)+1% L-glutamine (Gibco)+1% Penicillin-Streptomycin (Gibco). Cells were seeded on a 384-multiwell plate (2000 cells/ well) and were allowed to adhere and grow for 8 days before the addition of drugs. Wells were aspirated and drug dilutions were added on cells, which were then incubated at 37°C in a 5% CO2 incubator for 5 days. Viability of cells was detected with a luminescence based CTG-assay. Mean results of normalized data from three parallel wells are shown. All experiments were performed two independent times in triplicate. Drugs used in drug combination tests were obtained from the National Cancer Institute (NCI)/Division of Cancer Treatment and Diagnosis (DCTD)/Developmental Therapeutics Program (DTP) repository (Rockville, MD, USA). Combination index (CI) which defines drug synergism (CI<1) and antagonism (CI>1) was calculated with Microsoft Excel from replicate, fixed-ratio, dose escalation experiments using the Chou and Talalay method (20). CI values were reported at 50% inhibitory values (CI50).
Results
Mutational analysis. Targeted sequencing of the primary tumor revealed multiple somatic mutations (Table I, Figure 1). KRAS G13D has established clinical significance and guided therapy choices in this case. Co-occurrence of a KRAS mutation with BRAF F595L was detected in the primary tumor sample. Sequencing of the whole blood sample revealed no germline mutations.
Sequencing of the synchronous locoregional lymph node metastasis revealed a different assembly of somatic mutations (Table I, Figure 1). A BRAF L597Q had replaced the BRAF F595L and the KRAS G13D mutation had disappeared. In addition, nine other somatic mutations were found. The mutational landscape of metachronous para-aortal lymph node resembled the synchronous locoregional lymph node at the time of primary operation with only exception in SMAD4. A high concordance between mutations in these two lymph node metastases was detected in a plasma sample taken 9 months after primary surgery. The same driver mutations were observed in this liquid biopsy sample.
Ovarian metastasis revealed mutations in concordance with mutations found in the primary tumor: a BRAF F595L mutation with KRAS G13D mutation (Table I, Figure 1).
Other known somatic mutations were found in CRC tissue samples in genes such as MAP2K4, MAP2K2, APC, AMER1, BCOR, BLM, AKT1, HIST1H3B, SMAD4, ASXL1 and CCNE1 (Table I, Figure 1).
Droplet digital PCR. Serial plasma samples (N=5) were tested for driver mutation, BRAF L597Q, with ddPCR. The ctDNA for BRAF L597Q was found in 4 out of the 5 time points (Figure 2).
Targeted drug testing. Viable tumor cells were successfully extracted and cultured from the metachronous para-aortal lymph node metastasis sample taken 7 months after primary surgery. These cells containing the BRAF L597Q mutation were assayed for response to various drugs and drug combinations ex vivo (Figures 3 and 4). The most effective single-agent compound was the MEK-inhibitor, trametinib, with an IC50 of 0.69 μM. Also, the MEK1/2-inhibitor, AZD6244, and the pan-EGFR-inhibitor, AZD8931, showed activity in killing tumor cells with an IC50 of 1.40 μM and 1.86 μM, respectively. These results were consistent with the clinical experience that metastatic disease was resistant to adjuvant therapy of CAPOX showing tumor response in vitro with an IC50 of 25.81 μM for capecitabine and 11.11 μM for oxaliplatin (Figure 3). The control tumor cells grew well without the presence of an active antitumor agent. Drugs active against MEK/MAPK pathway targets were tested alone or in combinations in the tumor cell line derived from the metachronous para-aortal lymph node metastasis sample. Trametinib showed activity alone with an IC50 of 1.664 μM. Cetuximab as a single agent showed only modest activity with an IC50 of 157.900 mg/ml but the combination showed synergism (CI50=0.67) with an IC50 of 1.009 μM for trametinib, 10.090 mg/ml for cetuximab (Figure 4A). Adding dabrafenib to cetuximab showed no benefit with an IC50 of 9.237 μM for dabrafenib, 92.370 mg/ml for cetuximab (CI50=1.07) (Figure 4B). Similarly, the combination dabrafenib-trametinib had no synergism in these experiments with an IC50 of 1.279 μM for the combination and an IC50 of 1.244 μM for trametinib only (CI50=1.14) (Figure 4C).
Discussion
Results presented here emphasize multiple aspects of tumor heterogeneity. It has been established that the molecular makeup of tumors evolves dynamically in response to treatment and correlates with disease progression (21). Here, we showed that the primary tumor and metastatic sites may show dependence on the same signaling pathway, but mutations in target genes may show heterogeneity. These results illustrate that currently dominating and therapeutically relevant clonality may not be accurately captured by relying on primary tumor samples as a tissue source for treatment-guiding molecular pathology analyses. CtDNA has been suggested to be useful as a surrogate marker for metabolically active, dominant cancer clones. Previous studies have shown good correlation of ctDNA alterations with tissue biopsies, particularly in EGFR-driven cancers (22-24). There are also promising results in the non-metastatic, postoperative setting (25, 26). These findings have led to the suggestion that identification of these mutations in patient's blood, enables characterization of molecular heterogeneity, molecular residual disease and tumor evolution in real time. Results presented here illustrate that this approach is feasible.
It can be speculated that activating the MEK/MAPK pathway is crucial for colorectal cancer cell survival, yet different activating mutation combinations can result in a similar tumor phenotype. Interestingly, BRAF L597Q appeared to be the most suitable ctDNA biomarker for our patient. This ctDNA was detected in preoperative samples, the amount rapidly decreased after surgery and resurgence predicted clinical relapse.
For this patient BRAF L597Q ctDNA was a more sensitive biomarker than CEA or radiological imaging with CT. This is in agreement with the suggestion that ctDNA could work as an early indicator of tumor dynamics and serve to guide treatment strategies by revealing dominant tumor clonality. In vitro evidence also points to the role of BRAF L597Q as an oncogenic driver mutation since it has been described to be sufficient for MEK/MAPK pathway activation but to a lesser extent compared to BRAF V600E (13, 27, 28). There are no previous functional data available regarding this rare mutation in CRC. BRAF L597Q has been described in studies cataloguing BRAF mutations (27, 29, 30). The prevalence of BRAF L597Q mutation has been highest in melanoma, occurring at 4.1% frequency in two melanoma series (27, 31).
It is interesting to note that KRAS G13D mutation was lost in the metastatic sites with the BRAF L597Q mutation but retained in ovarian metastasis that contained the BRAF F595L mutation. Co-operation of the BRAF F595L and KRAS is consistent with the reported literature (32, 33). BRAF V600E and KRAS-mutations tend not to be expressed together and the mutual exclusivity is speculated to be due to senescence (34). In contrast, weak and intermediate kinase activity BRAF nonV600E mutant tumors are more likely to occur with concomitant KRAS-mutations (6, 10, 12, 32, 33, 35, 36).
Other mutations found in primary tumor and metastatic sites (APC, AMER1, SMAD4, MAP2K2, MAP2K4, AKT1, BCOR and BLM) have all been previously described somatic mutations in CRC (4, 12, 37-41). APC and MAP2K2 mutations appeared to be present in all samples suggesting that these were truncal mutations in this cancer, whereas variability was observed for other mutations suggesting their subclonal nature.
Mutations appear to be enriched in key oncogenic signaling pathway target genes. It has been shown that new mutations in the EGFR gene conferring drug resistance may be detected when evolutionary pressure in the form of EGFR TKI therapy or EGFR monoclonal antibody suppresses drug-sensitive clones (42, 43). These findings stress the importance of a system biology-based approach for cancer therapy, i.e. establishing the major molecular switch that promotes tumor survival and focusing on therapeutically shutting down this cascade. In the context of CRC, our results point to the key role of MEK/MAPK pathway as an oncogenic switch.
The best strategy for MEK/MAPK pathway inhibition in mCRC remains to be determined. BRAF inhibitor monotherapy as well as dual blockage of the MEK/MAPK pathway have had limited efficacy in mCRC (44-49). One reason for treatment failure seems to be the negative feedback network that leads to reactivation of MAPK signaling following treatment with a BRAF inhibitor. In CRC, this mechanism of resistance seems to be at least partially mediated via EGFR, while the whole underlying cascade has not yet been discovered (46, 48). Combining a third target, EGFR-inhibition, to the BRAF-MEK-inhibitor- doublet in BRAF V600E mutant patients resulted in a response rate of 21% with dabrafenib-trametinib-panitumumab triplet (50). Recently, binimetinib-encorafenib-cetuximab triplet resulted in an overall response rate of 48% and led to FDA-approval (16).
It has been suggested that one approach to improve personalized oncology is to utilize patient-derived live cancer cells, cancer cell clusters or organoid models to directly screen for optimally active drugs and drug combinations. Here, we employed this strategy to test effectiveness of a panel of cancer drugs on patient-derived tumor cells carrying BRAF L597Q driver mutation. Importantly, these cells were resistant to fluorouracil + oxaliplatin regimen recapitulating clinical tumor behavior. A MEK-inhibitor, trametinib was the most effective compound as single agent with an IC50 of 0.69 μM. An EGFR-inhibitor, cetuximab, alone showed very modest activity, but the combination of trametinib-cetuximab resulted in quantitative synergy (CI50=0.67) (22) with decreased IC50 of both drugs. In contrast, a BRAF-inhibitor, dabrafenib, appeared ineffective. Our results are in concordance with previous findings showing that intermediate activating BRAF-mutations signal as dimers and do not respond to BRAF-inhibitors as single agents. EGFR + MEK inhibition is probably required to target such BRAF mutations (14).
Combination of NGS and ex vivo testing appears to be a promising strategy for hypothesis-generating experiments in precision oncology. We suggest to perform NGS sequencing in the primary tumor to derive a list of driver genetic alterations and create affordable PCR-assayable personalized biomarkers to monitor treatment response with liquid biopsies. At the time of progression, extended molecular profiling could be used to analyze clonal evolution and detect novel genetic alterations. These data could then be used to engineer personalized drug panels to be tested with ex vivo assays on cancer cells derived from tissue biopsies.
Acknowledgements
National Cancer Institute (NCI)/Division of Cancer Treatment and Diagnosis (DCTD)/Developmental Therapeutics Program (DTP) (http://dtp.cancer.gov) is acknowledged as source of the Oncology Drugs Set. This work has also been supported in part by AZ-SLL-KI open innovation grant #18122013 and Seppo Nieminen's fund (grant # 160627). TheAuthors would like to thank Tuula Vierikko for providing assistance for reviewing radiological images.
Footnotes
Author's Contributions
KL, PL-KL, JK: study design, oversight, data analysis; LL: NGS experiments and data analysis; NR, JKR, RM: ex vivo drug testing; TK, IK: pathology workup; All Authors contributed to manuscript writing and accepted the final manuscript.
This article is freely accessible online.
Conflicts of Interest
The Authors declare that no competing interests exist regarding this study.
- Received September 21, 2019.
- Revision received October 22, 2019.
- Accepted October 24, 2019.
- Copyright© 2019, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved