www.elsevier.com/locate/humpath

Human Pathology (2014) xx, xxx–xxx

Case study

Analysis of BRCA2 loss of heterozygosity intumor tissue using droplet digital polymerasechain reaction☆

Rory L. Cochran BS1, Karen Cravero BS1, David Chu BS, Bracha Erlanger MSc,Patricia Valda Toro BS, Julia A. Beaver MD, Daniel J. Zabransky BS,Hong Yuen Wong BS, Justin Cidado PhD, Sarah Croessmann BS,Heather A. Parsons MD, MPH, Minsoo Kim, Sarah J. Wheelan MD, PhD,Pedram Argani MD, Ben Ho Park MD, PhD⁎

The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University School of Medicine, Baltimore, MD 21287

Received 3 February 2014; revised 17 March 2014; accepted 20 March 2014

VoreFRInA

CB

h0

Keywords:Droplet digital PCR;LOH;Familial breast cancer;BRCA2;VUS

Summary Loss-of-heterozygosity (LOH) analysis of archival tumor tissue can aid in determining theclinical significance of BRCA variants. Here we describe an approach for assessing LOH in formalin-fixed, paraffin-embedded (FFPE) tissues using variant-specific probes and droplet digital polymerasechain reaction (ddPCR). We evaluated LOH in 2 related breast cancer patients harboring a rare missenseBRCA2 variant of unknown clinical significance (c.6966GNT; M2322I). Conventional PCR followed bySanger sequencing suggested a change in allelic abundance in the FFPE specimens. However, we foundno evidence of LOH as determined by the allelic ratio (wild type–variant) for BRCA2 in both patients’archival tumor specimens and adjacent normal control tissues using ddPCR. In summary, theseexperiments demonstrate the utility of ddPCR to quickly and accurately assess LOH in archival FFPEtumor tissue.© 2014 Elsevier Inc. All rights reserved.

☆ We would like to thank the Sandy Garcia Charitable Foundation,irginia Beach,VA,USA.R.L.C. andS.C. are supported byNational Institutesf Health Pre-Doctoral Service Awards (CA168180 and CA167939,spectively). B.H.P. acknowledges support from the Breast Cancer Researchoundation, New York, NY, USA, the Commonwealth Foundation,ichmond, VA, USA, the Santa Fe Foundation, Singapore, the Nationalstitutes of Health/National Cancer Institute, Bethesda, MD, USA, and thevon Foundation, New York, NY, USA.⁎ Corresponding author. The Sidney Kimmel Comprehensive

ancer Center at Johns Hopkins, 1650 Orleans Street, Room 151,altimore, MD 21287.E-mail address: bpark2@jhmi.edu (B. H. Park).1 These authors contributed equally to this manuscript.

ttp://dx.doi.org/10.1016/j.humpath.2014.03.013046-8177/© 2014 Elsevier Inc. All rights reserved.

1. Introduction

The breast and ovarian cancer susceptibility genes,BRCA1 and BRCA2, lead to a significantly increased riskof developing breast and ovarian cancer for carriers [1].Consequently, at-risk individuals often pursue intensivescreening and/or prophylactic treatments, such as oophorec-tomy and/or mastectomy, as these procedures have beenshown to significantly reduce cancer incidence [2,3].Unfortunately, many studies over the past 2 decades havedemonstrated that both BRCA1 and BRCA2 are highly

2 R. L. Cochran et al.

polymorphic genes, with many variants of unclear clinicalsignificance (VUSs) existing within the human population[4]. As a result, BRCA variant classification studies remainongoing. Loss-of-heterozygosity (LOH) analysis of resectedtumor tissue has been tremendously helpful for classifyingparticular BRCA alleles not clearly linked to disease [5–8].Yet, there exist several technical hurdles that hinder archivaltissue−based studies, namely, poor tumor DNA quality,normal DNA contamination within tumor DNA prepara-tions, and inherent shortcomings to the conventionalmethods used for studying LOH.

In particular, formalin fixation leads to a high degree oftissue damage, yielding a limited amount of usable DNAmolecules for downstream studies. Moreover, contaminat-ing normal DNA present within formalin-fixed, paraffin-embedded (FFPE) tumor DNA preparations can hindermany genetic analyses. Additionally, fluorescence in situhybridization has intrinsic shortcomings that make analysisof archival tissue challenging. Specifically, fluorescence insitu hybridization is subjective, is prone to interobservervariability, has limited sensitivity, can be technicallydifficult and time consuming, and cannot identify certaingenomic alterations, such as loss with duplication [9]. Tocircumvent some of these challenges and facilitate futureLOH studies, we used droplet digital polymerase chainreaction (ddPCR) to study genomic DNA derived fromarchival tumor tissues from 2 members of a cancer-pronefamily harboring a BRCA2 VUS.

2. Case study

2.1. Combined clinical history

A 46-year-old woman of non-Ashkenazi ancestry and afamily history of cancer presented to the clinic for geneticcounseling and testing for hereditary breast and ovariancancer syndromes. Three years prior to her presentation, theproband was diagnosed with a grade II invasive ductalcarcinoma of the right breast with ductal carcinoma in situ.The proband’s tumor was found to coexpress estrogen andprogesterone receptors (ER+/PR+) and did not overexpressthe human epidermal growth factor receptor 2 (HER2−). Theproband’s mother was also diagnosed with a grade I invasiveductal carcinoma of the left breast with ductal carcinoma insitu at the age of 76 years. Following a consultation with theproband, she and her mother were encouraged to pursuegermline testing for mutations/alterations of the BRCA1 andBRCA2 genes. Genetic testing for the proband and hermother revealed the presence of a BRCA2 variant ofuncertain clinical significance (c.6966GNT; M2322I). Pa-tients transpose were/consented and enrolled in an institu-tional review board–approved protocol at Johns Hopkinsthat allows for obtaining tissues and bodily fluids in aprospective fashion from breast cancer patients.

2.2. Tissue processing, nucleic acid preparation,and Sanger sequencing

Buccal- and FFPE-derived genomic DNA (gDNA) wasisolated using QIAamp DNA BloodMini and QIAamp FFPEtissue kits, respectively (Qiagen, Valencia, CA). GenomicDNA was isolated from FFPE tissue stored from 8 to24 months’ duration using standard protocols. Briefly,hematoxylin and eosin–stained histology slides wereexamined by the study pathologist (P.A.) to identify areasof at least 70% breast carcinoma cells and adjacent benignlobular epithelial cells with 0% tumor cells greater than10 mm from any invasive component, hereafter termedFFPE tumor and FFPE normal, respectively. Five-microm-eter-thick unstained slides were deparaffinized, and identi-fied regions of interest were macrodissected using the Zymopen and Pinpoint solution (Zymo Research, Irvine, CA), perthe manufacturer’s protocol. Sanger sequencing of gDNAwas carried out following PCR amplification of respectiveloci using Phusion High-Fidelity DNA Polymerase (NewEngland BioLabs, Ipswich, MA). PCR and nested sequenc-ing primers for each locus described herein are listed inSupplementary Table S1.

2.3. ddPCR experiments

All ddPCR experiments were carried out using the QX100Droplet Digital PCR System according to the manufacturer’sprotocols (Bio-Rad, Hercules, CA), as previously described[10]. ddPCR primers were purchased from Integrated DNATechnologies (Coralville, IA), and fluorescently labeledTaqMan probes were purchased through Life Technologies(Carlsbad, CA). Primer and probe sequences are shown inSupplementary Table S2. Prior to analysis of patient DNAsamples, optimized thermocycling conditions were deter-mined for allele-specific binding of the fluorescent probes byinitially cloning 381 base pairs of genomic sequenceencompassing BRCA2 exon 13 (both the wild-type andVUS alleles) with flanking intronic sequence into a high-copybacterial plasmid. The temperature range of 56°C to 58°Cprovided strong fluorescent signals and yielded no probecross-reactivity. Therefore, we carried out the manufacturers’recommended thermocycling protocol with a 58°C annealing/extension step. Supplementary Fig. S1 shows experimentaldata from a representative ddPCR titration experimentdemonstrating 100% probe specificity using linear plasmidDNA as PCR template. All data analysis was performed usingthe accompanying platform software, QuantaSoft (Bio-Rad),with all reported values calculated using Poisson statistics.

2.4. Targeted next-generation DNA sequencing

The proband’s archival tissue−derived genomicDNA (bothtumor and surrounding normal) was subjected to targeted next-generation DNA sequencing using a custom 484 cancer gene

3LOH analysis using droplet digital PCR

panel. Targeted coding exon enrichment was carried out usingAgilent’s SureSelect technology, and DNA sequencing wasperformed on an Illumina HiSeq 2000 platform. Resultant datawere aligned to human genome 19 using BWA 0.6.1 [11], andvariants were called using GATK 1.4 [12].

3. Results

To confirm the presence of the BRCA2M2322I

(c.6966GNT) variant in the 2 affected family members, wesequenced exon 13 of BRCA2 from buccal cell–derivedgDNA. Consistent with previous genetic testing, bothindividuals diagnosed with cancer harbored the VUS, asshown in Fig. 1B. In addition, we identified the presence of aheterozygous single nucleotide polymorphism (SNP) locatedwithin exon 11 of BRCA2, rs543304, for the proband and hermother (Fig. 1B). The rs543304 SNP is located in exon 11approximately 8.7 kilobases upstream on gDNA from theVUS, as illustrated in Fig. 1A. Next, we purified gDNA fromFFPE tissue blocks for the proband and her mother toperform LOH and sequencing analysis on macrodissectedtumor and adjacent normal breast tissue. Using Sangersequencing, we found that both the proband and her motherwere heterozygous for the rs543304 SNP (Fig. 1C).However, as seen in the sequencing traces in Fig. 1C, therewas a high degree of variability in Sanger sequencing resultsfor the VUS locus. Both tumor samples appeared to have a1:1 ratio of wild-type to VUS alleles. In contrast, theproband’s normal adjacent tissue appeared to have a 2:1 wildtype-to-VUS allelic ratio, whereas the mother’s normal tissueseemingly contained a 1:2 wild type−VUS ratio. In our

Fig. 1 A, Diagrammatic representation of BRCA2 exons 11 and 13 showsequencing traces for both patient’ buccal cell–derived gDNA. C, BRCA2

hands, this type of variability in PCR sequencing is commonwith the use of FFPE-derived DNA, particularly when usingminute quantities of DNA. To confirm that our tumor andnormal samples were not inadvertently switched, becauseallelic ratios are more often changed in tumors, we verifiedthat the DNA used for these analyses were indeed fromcancerous tissues by identifying somatic mutations. First, wesequenced exons 9 and 20 of PIK3CA and all coding exonsof TP53, as these genes are frequently mutated in breastcancers [13]. Initial sequencing results revealed a heterozy-gous somatic mutation in exon 20 of PIK3CA (c.3140ANT;H1047L) in the mother’s tumor gDNA, as shown inSupplementary Figure S2. The proband’s DNA preparationsexhibited no PIK3CA mutations, and neither patientdemonstrated a TP53 mutation by Sanger sequencing.Because no somatic mutations were identified for theproband for PIK3CA or TP53, we used the tumor andadjacent normal genomic DNA for targeted next-generationsequencing with a custom 484 cancer gene panel. Somaticalterations in the proband’s tumor were identified and areshown in the Supplementary Information.

Our results with PCR and Sanger sequencing underscoredthe difficulty of determining allelic ratios and LOH usingFFPE-derived DNA. Therefore, to more accurately measurethe BRCA2 allelic ratios, we then tested the relatively newtechnology of ddPCR to perform LOH analyses. Briefly,ddPCR is carried out through the creation of nanoliterdroplets acting as parallel single-molecule PCR reactions(Fig. 2A) [10]. The identity of the DNA molecules withineach droplet can then be resolved using conventional dual-labeled fluorescent oligonucleotide probes. Following end-point PCR on a standard thermocycler, all droplets are thencounted using a fluorescent droplet reader (Fig. 2B). We

ing the rs543304 SNP and c.6966GNT VUS, respectively. B, SangerSanger sequencing traces for both patients’ FFPE-derived gDNA.

Fig. 2 Schematic of ddPCR workflow for assessing LOH. A, The ddPCR reaction involves combining fragmented DNA to be studied withdroplet oil and PCR reagents (PCR master mix, primers, and probes). The droplet generator machine partitions the required reagents and DNAfragments into approximately 2 × 104 individual droplets per well. Following droplet generation, end-point PCR is carried out on a standardthermocycler. B, After PCR amplification and generation of unquenched fluorophores, the droplets are read on a droplet reader. Example 1illustrates the expected distribution of fluorescent droplets for no LOH, where 50% of the gated droplets are positive for the wild-type allele and50% are positive for the VUS. Example 2 illustrates the approximate distribution of fluorescent droplets for LOH (with loss of the wild-typeallele) with contaminating wild type from surrounding normal tissue.

4 R. L. Cochran et al.

reasoned that ddPCR would be well suited for LOH analysisof FFPE-derived DNA for a number of reasons. First, itrequires low amounts of input DNA with very short PCRamplicon sizes (b100 base pairs). Second, ddPCR uses dual-labeled probes that are capable of reliably discriminatingbetween single base pair changes, such as SNPs andmutations.Third, ddPCR has the capacity to easily and accuratelydetermine differences in copy number less than 2-fold, whichis the lower limit of reliable detection using conventionalquantitative real-time PCR. Therefore, we used this approachto effectively count the number of BRCA2 alleles present

Fig. 3 LOH analysis for the proband and her mother. A, Plot showing tDNA patient samples with standard error shown using Poisson statistics.VUS and wild-type alleles for all patient samples.

within our patients’DNA samples using allele-specific probes.By comparing the number ofBRCA2 alleles present within ourpatients’ samples, we reasoned we could simultaneouslyassess LOH and quantify contaminating normal DNA, asillustrated in Fig. 2B. Using conventional allele-specificfluorescent probes, differing by only one nucleotide atBRCA2 c.6966G/T, we quantified the BRCA2 allelic ratioswithin the normal and tumor FFPE gDNA for both individuals.As shown in Fig. 3A, we determined the allelic ratios (wildtype–VUS) to be approximately 1 in all tissue samplesassayed. Absolute percentages over multiple experiments,

he allelic ratio (wild type–VUS) measured for the buccal and FFPEB, Average percentage of fluorescent droplets corresponding to the

5LOH analysis using droplet digital PCR

corresponding to the percentage ofwild-typeDNA–containingand VUS DNA–containing droplets, are shown in Fig. 3B.These data clearly show no evidence of LOH in eitherindividual’s tumor or normal sample and importantlyunderscore the inability of conventional PCR sequencing toaccurately determine allelic ratios using FFPE gDNA.

4. Discussion

LOH information, in combination with other methods, isuseful in classifying uncertain BRCA variants [5,8]. Althoughloss of the wild-type allele in tumor tissues provides strongevidence for a deleterious germline mutation, it should benoted that absence of LOH cannot be used as a definitivecriterion because other mechanisms of inactivation of thewild-type allele exist. This would include epigenetic silencingas well as a de novo inactivating mutation within the wild-typeallele. Nonetheless, given that loss of the second allele is anaccepted common mechanism of carcinogenesis mediated bytumor suppressor genes in familial cancer syndromes, thestudies described here have great potential for definitivelyassessing LOH and helping to ascribe a functional significanceto germline VUS and deleterious mutations.

The arrival of commercially available digital PCRplatforms has the potential to dramatically alter the speedand sensitivity of many molecular biology assays commonlyused in the clinic. Others have shown that ddPCR canaccurately determine both copy number changes andexpression of critical oncogenes using patient samples[14,15]. This approach may prove to be a more objectiveand accurate measure for determining oncogene amplifica-tion and overexpression in cancer patients. ddPCR isespecially useful for analysis of FFPE-derived gDNAbecause it relies on short DNA fragments and FFPE gDNAis generally highly fragmented. Moreover, because ddPCRqueries individual DNA molecules, this approach shouldprove useful in several assays including determining thefractional abundance of somatic mutations within heteroge-neous tumor tissue specimens; validation of somaticmutations identified using whole genome sequencing thatare below the limit of detection by Sanger sequencing; and,as demonstrated here, determining the allelic fraction of agiven gene for LOH analysis. Although, in principle, next-generation sequencing using a targeted gene−captureapproach could also be used for LOH and copy numberanalysis, ddPCR offers significant advantages including amuch faster and cost-efficient platform, need for less inputDNA, superior sensitivity, no artifactual errors inherent withnext-generation sequencing, and no requisite for a bioinfor-matics pipeline for variant calling. To our knowledge, this isthe first report demonstrating the utility of using ddPCR foraccurately assessing allelic ratios via SNPs in FFPE tumorsamples. This approach should greatly facilitate not only

BRCA VUS LOH studies but studies for other variants ofgenes implicated in hereditary cancer susceptibility.

Appendix A. Supplementary data

Supplementary data to this article can be found online athttp://dx.doi.org/10.1016/j.humpath.2014.03.013.

Acknowledgments

Wewould like to thank the patients described in this studyfor their graceful participation.

References

[1] Narod SA, Foulkes WD. BRCA1 and BRCA2: 1994 and beyond. NatRev Cancer 2004;4:665-76.

[2] Rebbeck TR, Friebel T, Lynch HT, et al. Bilateral prophylacticmastectomy reduces breast cancer risk in BRCA1 and BRCA2 mutationcarriers: the PROSE Study Group. J Clin Oncol 2004;22:1055-62.

[3] Rebbeck T, Lynch H, Neuhausen S, et al. Prophylactic oophorectomyin carriers of BRCA1 or BRCA2 mutations. N Engl J Med2002;346:1616-22.

[4] Easton DF, Deffenbaugh AM, Pruss D, et al. A systematic geneticassessment of 1,433 sequence variants of unknown clinical signifi-cance in the BRCA1 and BRCA2 breast cancer-predisposition genes.Am J Hum Genet 2007;81:873-83.

[5] Chenevix-Trench G, Healey S, Lakhani S, et al. Genetic andhistopathologic evaluation of BRCA1 and BRCA2 DNA sequencevariants of unknown clinical significance. Cancer Res 2006;66:2019-27.

[6] Osorio A, de la Hoya M, Rodríguez-López R, et al. Loss ofheterozygosity analysis at the BRCA loci in tumor samples frompatients with familial breast cancer. Int J Cancer 2002;99:305-9.

[7] Osorio A, Milne R, Honrado E, et al. Classification of missensevariants of unknown significance in BRCA1 based on clinical andtumor information. Hum Mutat 2007;28:477-85.

[8] Spearman AD, Sweet K, Zhou X-P, McLennan J, Couch FJ, TolandAE. Clinically applicable models to characterize BRCA1 and BRCA2variants of uncertain significance. J Clin Oncol 2008;26:5393-400.

[9] Karnan S, Mohseni M, Konishi Y, et al. Controversial BRCA1allelotypes in commonly used breast cancer cell lines. Breast CancerRes Treat 2010;119:249-51.

[10] Hindson B, Ness K, Masquelier D, et al. High-throughput dropletdigital PCR system for absolute quantitation of DNA copy number.Anal Chem 2011;83:8604-10.

[11] Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 2009;25:1754-60.

[12] McKenna A, Hanna M, Banks E, et al. The Genome Analysis Toolkit:a MapReduce framework for analyzing next-generation DNAsequencing data. Genome Res 2010;20:1297-303.

[13] Bachman K, Argani P, Samuels Y, et al. The PIK3CA gene is mutatedwith high frequency in human breast cancers. Cancer Biol Ther2004;3:772-5.

[14] Belgrader P, Tanner SC, Regan JF, Koehler R, Hindson BJ, BrownAS. Droplet digital PCR measurement of HER2 copy numberalteration in formalin-fixed paraffin-embedded breast carcinomatissue. Clin Chem 2013;000:1-4.

[15] Heredia NJ, Belgrader P, Wang S, et al. Droplet DigitalTM PCRquantitation of HER2 expression in FFPE breast cancer samples.Methods 2013;59:S20-3.