Detection of BRAFV600E Mutation in Melanoma Patients by Digital PCR of Circulating DNA

Abstract

Aims:

About 50% of melanomas have the BRAFV600E mutation. This mutation is an attractive therapeutic target. The aims of our study were to detect BRAFV600E mutations within circulating cell-free DNA in plasma (“liquid biopsy”) by a droplet digital PCR (ddPCR) method, and to investigate how well the Breslow-Clark score can be predicted by ddPCR.

Materials and Methods:

We analyzed 113 patients with malignant melanoma. ddPCR was performed using the QX200 system (BIO-RAD^®, Hercules). All samples were tested in duplicate. Besides the results of the liquid biopsy, we have collected data on gender and age of the patients, as well as the mitotic activity of the tumor; the tumor subtype and localization, and the Breslow-Clark score. The limit of detection (LoD) was determined by the method of Tzonev. The LoD was found to be five events per well.

Results:

The BRAFV600E mutation was detected in 37 of 113 samples. A moderate predictive accuracy of the Breslow-Clark score can be attained with the mitotic activity and the type of melanoma as the most important predictors.

Conclusion:

Our results show that ddPCR is a highly sensitive method and could be used for a routine laboratory detection of the BRAFV600E mutation as well as for follow-up monitoring to determine the treatment response in patients with malignant melanomas.

Introduction

Cell-free circulating DNA (cfDNA) is DNA that occurs freely in the bloodstream of healthy individuals. Most of the cfDNA in plasma is from hematopoetic stem cells that undergo apoptosis (Stroun et al., 2000; Lui et al., 2002). Circulating tumor DNA (ctDNA) is cfDNA derived from cancer cells undergoing apoptosis or necrosis, whereas cfDNA is derived mainly from apoptotic processes in healthy individuals (Suzuki et al., 2008; Salvi et al., 2016). Apart from the blood, cfDNA was also detected in other body fluids such as urine, saliva, cerebrospinal fluid, and pleural fluid (Momtaz et al., 2016; Pu et al., 2016; Salvi et al., 2016; Husain et al., 2017). The length of noncancerous cfDNA is around 166 bp, corresponding to the DNA wrapped around nucleosome associated with histone H1. The average length of ctDNA is 134-144 bp, ctDNA is more fragmented than cfDNA (Mouliere and Rosenfeld, 2015). The detection of mutations in circulating cell-free DNA (cfDNA) serves as a surrogate for detecting and analyzing tumor DNA and is thus termed as “liquid biopsy.” Measurement of the fraction of ctDNA relative to normal circulating DNA has been shown to be prognostic and predictive of response to treatment, recurrence, or progression in some cancers (Polivka et al., 2015).

BRAF is a member of the Raf kinase family and is a downstream target of RAS, playing an important role in the MAPK/ERK signaling pathway. Activating the mutated BRAF protein is a result of conformation changes within the protein structure (Davies et al., 2002). The majority of malignant melanomas have BRAF mutation (∼40-60% of cases) (Dhomen and Marais, 2009). More than 90% of BRAF mutations are at codon 600 of the BRAF gene (Long et al., 2011). For malignant melanomas, BRAFV600E mutation is an attractive therapeutic target for the treatment with tyrosine kinase inhibitors such as vemurafenib and dabrafenib. They have been shown to improve the survival of melanoma patients bearing this mutation (Sosman et al., 2012; Banzi et al., 2016). Absence of BRAF mutation excludes using tyrosine kinase inhibitors (Goldlinger et al., 2013). Thus, precise detection of BRAFV600E is essential to guide a management of patients with malignant melanomas. Several methods are available to detect BRAFV600E mutation including Sanger sequencing, high-resolution melting (Goldlinger et al., 2013), allele-specific PCR (Sapio et al., 2006; So et al., 2014), Cobas^® 4800 (Mourah et al., 2015), next-generation sequencing (Miraflor et al., 2010), real-time PCR (Pinzani et al., 2010), and droplet digital PCR (ddPCR) (Hudecova, 2015).

Digital PCR is a cost-effective and sensitive method for detection of mutated DNA in patient's samples (Dube et al., 2008). ddPCR can quantify small amounts of ctDNA and there is no need for a standard curve. The ddPCR provides absolute quantification of target DNA and background DNA, which are randomly distributed among the droplets (Weaver et al., 2010). This system combines water-oil emulsion droplet technology with microfluids, allowing partitioning of DNA molecules into ∼20,000 droplets where every droplet represents an individual PCR (Day et al., 2013). Each droplet is read individually with the QX200 droplet reader. Positive and negative droplets are counted to provide absolute quantification (Hindson et al., 2011). In this study, we investigated whether detection and quantification of BRAFV600E mutations in plasma from patients with malignant melanomas using ddPCR are achievable. We also assessed predictive accuracy of BRAFV600E mutations and clinical parameters for predicting the Breslow-Clark score.

Materials and Methods

Patients

This study was conducted at the Clinic of Dermatovenerology, Department of Plastic Surgery and Department of Pathology of Jessenius Faculty of Medicine and the University Hospital of Martin, Slovakia. One hundred thirteen plasma consecutive samples, histologically identified as melanomas, were selected for this study. All subjects before their biopsies agreed with study participation and signed written informed consent. This prospective study was approved by the University Ethics Committee. The BRAFV600E mutation was analyzed by the ddPCR; details are given hereunder.

DNA extraction

Blood samples, collected in EDTA tubes, were centrifuged at 4000 g for 10 min at 4°C. Next, plasma samples were stored at −80°C until cfDNA extraction was performed from 1.5 mL of plasma using the DSP virus kit. After washing several times, the nucleic acids are eluted from the membrane by kit-provided Elution Buffer. DNA was quantified using the Qubit 2.0 fluorometer and the Qubit dsDNA HS assay kit (Life-Technologies, Gent-Brussels, Belgium). Tumor tissue sample analysis for BRAFV600E mutations was conducted in our previous study. In this study, DNA was extracted by using the blackPREP FFPE DNA kit (Malicherova et al., 2018b).

Droplet digital PCR

For preparation of the ddPCR mixture, we first created optimal amplification conditions for thermal cyclization for each test. For separation of mutation-negative and mutation-positive droplets, we used the BRAFV600E mutated human colorectal cancer (RKO) cell line (Dr. N.A.P. Franken from Academic Medical Center, University of Amsterdam in the Netherlands) diluted to 40 ng/μL; the wild-type (WT) DNA sample was also diluted to 40 ng/μL. The Taq polymerase reaction mixture consisted of 2 × ddPCR Supermix for Probes (Bio-Rad Laboratories), 20 × assay PrimePCR™ ddPCR™ Mutation Assay Kit BRAF WT for p. V600E, and BRAF p.V600E (Bio-Rad Laboratories). Subsequently, various concentrations of DNA prepared from the respective RKO line were prepared in seven dilutions. One microliter sample from each dilution and 1 μL of WT DNA were mixed with mastermix from the first step in 18 μL. After this optimization, we analyzed plasma-derived DNA samples. ddPCR mixtures with a final volume of 22 μL were prepared, consisting of 10 μL Supermix for Probes (no deoxyuridine triphosphate), 1 μL fluorescein (FAM) probe and 1 μL hexachlorofluorescein probe, and 5 μL of DNA sample isolated from plasma. The negative control contained 5 μL of purified water. The WT-only samples contained 2 μL of DNA from FFPE samples. ddPCR was performed using the QX200™ ddPCR system according to the manufacturer's instructions (Bio-Rad Laboratories). The DG8 (Bio-Rad) cartridge was filled with 20 μL of MasterMix from the previous step and 70 μL of the oil droplet generator for each sample. After the emulsification of samples in Droplet Generation, we transferred 40 μL of created emulsion to the 96-well plate according to the instructions of the manufacturer. The 96-well plate was secured on the support block and covered with one sheet of special pierceable foil seal. The 96-plate from droplet foil sealer was carried to the thermal cycler and reaction in each microbeads ran under the following conditions: activation of polymerase (95°C 10 min); 40 cycles with steps: denaturation (94°C 30 s), annealing/extension (58°C 60 s), and final extension (98°C 10 min) with ramp 2.5°C/s. PCR products were held at 4°C until the next step of analysis. After completion of the reaction, the ddPCR plate was inserted into the Droplet Reader, which analyzed the droplets in each well. Analysis of the ddPCR data was inserted in the analytical software QuantaSoft Analysis Pro (Bio-Rad Laboratories) based on two probes: one that emits an FAM fluorescence signal, with the specific mutated target and the second that emits an HEX fluorescence signal with the specific WT target. A series of dilutions of genomic mutant DNA labeled with FAM probe and also with WT DNA with a HEX-labeled probe were performed in duplicate. Every drop of the sample appears on the chart as fluorescence intensity versus number of drops.

Predicting the Breslow-Clark score

The data on age, gender, mitotic activity, type of melanoma, localization of the tumor, and the Breslow-Clark score were collected for all the patients, together with the BRAFV600E as detected by ddPCR. The multinomial logistic regression model was fit to the data with the objective of identifying the statistically significant predictors of the Breslow-Clark score. To obtain a realistic estimate of the receiver operating characteristic (ROC) curve, we have used the Random Forest machine learning algorithm with the feature selection performed by the minimum depth method within the nested cross-validation procedure.

Statistical analysis

Statistical analysis was performed in R (R Core Team, 2015) version 3.2.3, using the libraries nnet (Venables and Ripley, 2002) to perform the multinomial logistic regression, randomForestSRC (Ishwaran and Kogalur, 2017), to run the Random Forest classification algorithm and a feature selection. Reduction of the full multinomial logit model (i.e., the model with all the considered predictors) was done by the Akaike Information Criterion. Statistical significance of the predictors in the reduced multinomial logit model was assessed by the type III ANOVA test from the library car (Fox and Sanford Weisberg, 2011). A result was considered to be statistically significant if the p-value was <0.05.

Results

Limit of detection of ddPCR

We analyzed plasma samples by using QuantaSoft Analysis Pro software (Bio-Rad Laboratories). We used the method of Tzonev to determine the limit of detection (LoD). The method of Tzonev determines LoD in terms of the number of events (Tzonev, 2016). To investigate the analytic sensitivity of ddPCR, we used RKO line containing BRAFV600E that was serially diluted with WT DNA (range 50% to 0.000001%) as described previously (Malicherova et al., 2018a). The ddPCR was able to detect the mutation with 0.001% sensitivity. No BRAFV600E-positive droplets were detected in negative control samples. The number of positive versus negative droplets in a sample was used to calculate the concentration of target DNA in terms of copies/μL.

cfDNA quantification by ddPCR

Plasma samples from 113 patients were analyzed by using ddPCR. The amount of ctDNA was quantified, ranging from 5 to 12.9 ng/μL. The BRAFV600E mutation was detected in 37 of 113 plasma samples (Table 1). In total, 87 of 113 patients also provided tissue samples. Eighty-seven tumor tissue samples were previously tested by Cobas 4800, Sanger sequencing, allele-specific PCR, and ddPCR. (Malicherova et al., 2018a). All samples were tested in duplicate.

Table 1.

Breslow Stages and Number of Cases in Each Stage and Percentage of Positive Cases in Each Stage

Stage	Number of cases	Positive cases
I	5	0% (0)
II	11	18% (2)
III	19	36.8% (7)
IV	70	37.1% (26)
V	8	25% (2)

Tumor tissue versus plasma samples

The BRAFV600E mutation was detected in 31 of 87 FFPE samples (Malicherova et al., 2018a) and in 37 of 113 plasma samples. Thirty-one patients had detected BRAFV600E mutation in both tissue-derived genomic DNA and plasma cfDNA. Six patients with malignant melanomas had detected BRAFV600E mutation in only plasma cfDNA.

Predicting the Breslow-Clark score

The full multinomial logit model that involved age, gender, mitotic activity, type of melanoma, localization of the tumor, and ddPCR quantification of BRAF V600E was reduced by the Akaike Information Criterion, which identified the mitotic activity and the type of melanoma as the only relevant predictors. The type III ANOVA test has found both the predictors to be statistically significant (the p-value <0.0001, 0.0072, respectively).

The feature selection by the minimum depth method in Random Forest has also identified the mitotic activity and type of melanoma as the most important predictors, and has also selected age, localization, and ddPCR as relevant predictors. Thus, the machine learning algorithm has effectively excluded solely the gender, as the predictor irrelevant for predicting the Breslow-Clark score. The predictive accuracy, as quantified by the area under ROC curve, was 89.3%, 72.5%, 71.5%, and 68.8% for the Breslov-Clark I versus II, III, and IV together; II versus I, III, and IV together; III versus I. II, and IV together; and IV versus I, II, and III together, respectively.

Discussion

In the recent past, liquid biopsy has revolutionized cancer detection. Liquid biopsy is useful not only for diagnosis but also to monitor tumors more effectively and for assessing tumor heterogeneity. If tumor cells coexist with different molecular features, it is known as the intratumor heterogeneity. This kind of heterogeneity is present in most solid tumors, including malignant melanomas (Yancovitz et al., 2012). Analysis of ctDNA can capture all mutations present in all tumors in one patient and provide a more complex picture of the mutational landscape (Diaz and Bardelli, 2014).

Malignant melanomas belong to cancer with increasing incidence and mortality in Slovakia. This disease is treated with a combination of surgery, chemotherapy, targeted therapy, and immunotherapy. Melanoma is a highly aggressive cancer that tends to spread to other parts of the body—metastasize. Precise detection of BRAFV600E mutation is needed in BRAF-positive patients with melanomas for treatment with vemurafenib and dabrafenib (BRAFV600E target therapy) (Chapman et al., 2011; Atkinson, 2015). The other common mutations in codon 600 of the BRAF gene are BRAFV600K and V600R. They are also associated with malignant melanomas (Klein et al., 2013). In addition, mutations in other codons were described (Wu et al., 2017). It was demonstrated that patients with the rarer BRAF mutations can be treated with BRAF inhibitors (McArthur et al., 2014; Casadevall et al., 2016), so screening tests for these mutations may be needed.

A correlation between amount of circulating BRAFV600E and overall survival (OS) was observed. Patients who had higher plasma BRAFV600E had shorter OS (Sanmamed et al., 2015). BRAFV600E mutation levels in plasma were lower in patients, who respond to the target therapy (Gray et al., 2015).

In our study, we investigated whether BRAFV600E can be detected by ddPCR in plasma of melanoma patients and how well can the Breslow-Clark score be predicted from gender, age, mitotic activity, type of melanoma, localization of the tumor, and the tumor quantification by ddPCR. Our study demonstrated that we can detect BRAFV600E mutation in plasma samples with high sensitivity (0.001%). ddPCR can quantify low concentrations of ctDNA in a high background DNA, without using standard curves. A higher level of ctDNA was reported to be associated with advanced stages of melanomas (Shinozaki et al., 2007).

However, there are also limitations associated with applying ddPCR to ctDNA analysis. Low amount of mutated DNA or poor ctDNA quality can lead to false negative results. It is difficult to detect low concentration of ctDNA in a large background of normal DNA. The plasma samples from all 113 patients in our cohort were analyzed for BRAFV600E mutation. ddPCR identified 24 primary tumors, 10 metastases, and 3 micrometastases. Mean age of the diagnosed patients was 63.9 years in males and 64 years in females. Most melanomas were superficially spreading melanomas and they were located on the back (n = 54) of the body. Seven patients, included in our study, were having acral lentiginous melanoma, 4 patients were having lentigo malignant melanoma, 16 patients were having an unspecified type of malignant melanomas and 7 patients were having in situ melanomas, 24 patients were having nodular malignant melanomas, and 1 patient was having desmoplastic malignant melanoma. Furthermore, five of the eight melanoma patients whose tumors were only identified as BRAFV600E positive using ddPCR in FFPE tissues (Malicherova et al., 2018a) and plasma later developed subsequent metastatic relapse (Table 1).

We hypothesized that ddPCR can predict Breslow-Clark score. However, we did not confirm that ddPCR from plasma samples can predict Breslow thickness. In the multinomial logistic regression predictive model, there were only two statistically significant predictors: mitotic rate and type of malignant melanoma. However, a machine learning Random Forest algorithm had included ddPCR quantifications among the important predictors, though as the least important predictor. The most important predictors were mitotic activity and type of melanoma. The predictive accuracy of the model was moderate, implying that the Breslow thickness cannot be predicted very reliably by means of the studied factors.

Footnotes

Acknowledgments

We would like to sincerely thank all patients. This work was supported by the “Biomedical Center Martin (BioMed Martin)” ITMS code 26220220187 project, which is cofinanced from EU sources and by the Slovak Research and Development Agency under contract nos. APVV-16-0066 and APVV-14-0273.

Author Disclosure Statement

No competing financial interests exist.

References

Atkinson

(2015) Medical management of malignant melanoma. Aust Prescr, 38:74-78.

Banzi

, De Blasio

, Lallas

, et al. (2016) Dabrafenib: a new opportunity for the treatment of BRAF V600-positive melanoma. Onco Targets Ther, 9:2725-2733.

Casadevall

, Vidal

, Gallardo

, et al. (2016) Dabrafenib in an elderly patient with metastatic melanoma and BRAF V600R mutation: a case report. J Med Case Rep, 10:158.

Chapman

, Hauschild

, Robert

, et al. (2011) Improved survival with vemurafenib in melanoma with BRAFV600E mutation. N Engl J Med, 364:2507-2516.

Davies

, Bignell

, Cox

, et al. (2002) Mutations of the BRAF gene in human cancer. Nature, 417:949-954.

Day

, Dear

, McCaughan

(2013) Digital PCR strategies in the development and analysis of molecular biomarkers for personalized medicine. Methods, 59:101-107.

Dhomen

, Marais

(2009) BRAF signaling and targeted therapies in melanoma. Hematol Oncol Clin North Am, 23:529-545.

Diaz

Jr ., Bardelli

(2014) Liquid biopsies: genotyping circulating tumor DNA. J Clin Oncol, 32:579-586.

Dube

, Qin

, Ramakrishnan

(2008) Mathematical analysis of copy number variation in DNA sample using a digital nanofluidic PCR on a device. PLoS One, 3:e2876.

10.

Fox

, Sanford Weisberg

(2011) An {R} Companion to Applied Regression, Second Edition. Sage, Thousand Oaks, CA. Available at: http://socserv.socsci.mcmaster.ca/jfox/Books/Companion (accessed October 24, 2018).

11.

Goldinger

, Murer

, Stieger

, et al. (2013) Targeted therapy in melanoma—the role of BRAF, RAS and KIT mutations. EJC Suppl, 11:92-96.

12.

Gray

, Rizos

, Reid

, et al. (2015) Circulating tumor DNA to monitor treatment. response and detect acquired resistance in patients with metastatic melanoma. Oncotarget, 6:42008-42018.

13.

Hindson

, Ness

, Masquelier

, et al. (2011) High-throughput droplet digital PCR system for absolute quantitation of DNA copy number. Anal Chem, 83:8604-8610.

14.

Hudecova

(2015) Digital PCR analysis of circulating nucleic acids. Clin Biochem, 48:948-956.

15.

Husain

, Nykin

, Bui

, et al. (2017) Cell-free DNA from ascites and pleural effusions: molecular insights into genomic aberrations and disease biology. Mol Cancer Ther, 16:948-955.

16.

Ishwaran

, Kogalur

(2017) Random Forests for Survival, Regression and Classification (RF-SRC), R package version 2.4.2. Available at: https://kogalur.github.io/randomForestSRC/theory.html (accessed October 24, 2018).

17.

Klein

, Clements

, Menzies

, et al. (2013) BRAF inhibitor activity in V600R metastatic melanoma. Eur J Cancer, 49:1073-1079.

18.

Long

, Menzies

, Nagrial

, et al. (2011) Prognostic and clinicopathologic associations of oncogenic BRAF in metastatic melanoma. J Clin Oncol, 29:1239-1246.

19.

Lui

, Chik

, Chiu

, et al. (2002) Predominant hematopoietic origin of cell- free DNA in plasma and serum after sex-mismatched bone marrow transplantation. Clin Chem, 48:421-427.

20.

Malicherova

, Burjanivova

, Grendar

, et al. (2018a) Droplet digital PCR for detection of BRAF V600E mutation in formalin-fixed, paraffin-embedded melanoma tissues: a comparison with Cobas^® 4800, Sanger sequencing, and allele-specific PCR. Am J Transl Res, 10:3773-3781.

21.

Malicherova

, Burjanivova

, Minarikova

, et al. (2018b) Detection of driver mutations in FFPE samples patients with verified malignant melanoma. Neoplasma. [Epub ahead of print]; DOI: 10.4149/neo_2018_180115N31.

22.

McArthur

, Chapman

, Robert

, et al. (2014) Safety and efficacy of vemurafenib in BRAF(V600E) and BRAF(V600K) mutation-positive melanoma (BRIM-3): extended follow-up of a phase 3, randomised, open-label study. Lancet Oncol, 15:323-332.

23.

Miraflor

, de Abreu

, Peterson

, et al. (2010) Somatic mutation analysis in melanoma using targeted next generation sequencing. Clin Chim Acta, 411:1319-1324.

24.

Momtaz

, Pentsova

, Abdel-Wahab

, et al. (2016) Quantification of tumor-derived cell free DNA(cfDNA) by digital PCR (DigPCR) in cerebrospinal fluid of patients with BRAFV600 mutated malignancies. Oncotarget, 7:85430-85436.

25.

Mouliere

, Rosenfeld

(2015) Circulating tumor-derived DNA is shorter than somatic DNA in plasma. Proc Natl Acad Sci U S A, 112:3178-3179.

26.

Mourah

, Denis

, Narducci

, et al. (2015) Detection of BRAF V600 mutations in melanoma: evaluation of concordance between the Cobas^® 4800 BRAF V600 mutation test and the methods used in French National Cancer Institute (INCa) platforms in a real-life setting. PLoS One, 10:e0120232.

27.

Pinzani

, Salvianti

, Cascella

, et al. (2010) Allele specific Taqman-based real-time PCR assay to quantify circulating BRAFV600E mutated DNA in plasma of melanoma patients. Clin Chim Acta, 411:1319-1324.

28.

Polivka

Jr. , Pesta

, Janku

(2015) Testing for oncogenic molecular aberrations in cell-free DNA-based liquid biopsies in the clinic: are we there yet?. Expert Rev Mol Diagn, 15:1631-1644.

29.

, Liang

, Wei

, et al. (2016) Evaluation of a novel saliva-based epidermal growth factor receptor mutation detection for lung cancer: a pilot study. Thorac Cancer, 7:428-436.

30.

R Core Team (2015) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available at www.R-project.org (last accessed September 14, 2017).

31.

Salvi

, Martignano

, Molinari

, et al. (2016) The potential use of urine cell free DNA as a marker for cancer. Expert Rev Mol Diagn, 16:1283-1290.

32.

Sanmamed

, Fernández-Landázuri

, Rodríguez

, et al. (2015) Quantitative cell-free circulating BRAFV600E mutation analysis by use of droplet digital PCR in the follow-up of patients with melanoma being treated with BRAF inhibitors. Clin Chem, 61:297-304.

33.

Sapio

, Posca

, Troncone

, et al. (2006) Detection of BRAF mutation in thyroid papillary carcinomas by mutant allele-specific PCR amplification (MASA). Eur J Endocrinol, 154:341-348.

34.

Shinozaki

, O'Day

, Kitago

, et al. (2007) Utility of circulating B-RAF DNA mutation in serum for monitoring melanoma patients receiving biochemotherapy. Clin Cancer Res, 13:2068-2074.

35.

, Chan

, Chung

, et al. (2014) A comparison of high resolution melting, allele-specific priming and Sanger sequencing for the detection of BRAFV600E mutation in hairy cell leukaemia from different haematological specimens. Pathology, 46:533-537.

36.

Sosman

, Kim

, Schuchter

, et al. (2012) Survival in BRAF V600-mutant advanced melanoma treated with vemurafenib. N Engl J Med, 366:707-714.

37.

Stroun

, Maurice

, Vasioukhin

, et al. (2000) The origin and mechanism of circulating DNA. Ann N Y Acad Sci, 906:161-168.

38.

Suzuki

, Kamataki

, Yamaki

, et al. (2008) Characterization of circulating DNA in healthy human plasma. Clin Chim Acta, 387:55-58.

39.

Tzonev

(2016) Sensitivity, specificity and limit of detection in dPCRBIO-RAD; (Powerpoint slides). Retrieved from Eddie van Collenburg, DBC Summit.

40.

Venables

, Ripley

(2002) Modern Applied Statistics with S. Fourth Edition. Springer, New York; ISBN 0-387-95457-0.

41.

Weaver

, Dube

, Mir

, et al. (2010) Taking qPCR to a higher level: Analysis of CNV reveals the power of high throughput qPCR to enhance quantitative resolution. Methods, 50:271-276.

42.

, Yan

, Dai

, et al. (2017) Mutations in BRAF codons 594 and 596 predict good prognosis in melanoma. Oncol Lett, 14:3601-3605.

43.

Yancovitz

, Litterman

, Yoon

, et al. (2012) Intra- and inter-tumor heterogeneity of BRAF(V600E) mutations in primary and metastatic melanoma. PLoS One, 7:e29336.