CM 201 E 6J Pr M og D ra m

The Journal of Molecular Diagnostics, Vol. 18, No. 2, March 2016

jmd.amjpathol.org

Next-Generation Sequencing-Assisted DNA-Based Digital PCR for a Personalized Approach to the Detection and Quantification of Residual Disease in Chronic Myeloid Leukemia Patients Mary Alikian,*y Peter Ellery,* Martin Forbes,y Gareth Gerrard,*y Dalia Kasperaviciute,z Alona Sosinsky,z Michael Mueller,z Alexandra S. Whale,x Dragana Milojkovic,{ Jane Apperley,y Jim F. Huggett,x Letizia Foroni,*y and Alistair G. Reid*y From Imperial Molecular Pathology,* Imperial Healthcare Trust, Hammersmith Hospital, London; the Centre for Haematologyy and the Clinical Genome Informatics Facility,z Faculty of Medicine, Imperial College London, London; Molecular & Cell Biology,x LGC Limited, Queens Road, Teddington; and Clinical Haematology,{ Imperial College Healthcare National Health Institute Trust, London, United Kingdom CME Accreditation Statement: This activity (“JMD 2016 CME Program in Molecular Diagnostics”) has been planned and implemented in accordance with the Essential Areas and policies of the Accreditation Council for Continuing Medical Education (ACCME) through the joint providership of the American Society for Clinical Pathology (ASCP) and the American Society for Investigative Pathology (ASIP). ASCP is accredited by the ACCME to provide continuing medical education for physicians. The ASCP designates this journal-based CME activity (“JMD 2016 CME Program in Molecular Diagnostics”) for a maximum of 36 AMA PRA Category 1 Credit(s)
. Physicians should only claim credit commensurate with the extent of their participation in the activity. CME Disclosures: The authors of this article and the planning committee members and staff have no relevant financial relationships with commercial interests to disclose.

Accepted for publication September 17, 2015. Address correspondence to Mary Alikian, Ph.D., Centre for Haematology, Faculty of Medicine, Imperial College London, Hammersmith Hospital Campus, Du Cane Rd., London W12 0NN, United Kingdom. E-mail: m.alikian@ imperial.ac.uk.

Recent studies indicate that 40% of chronic myeloid leukemia patients who achieve sustained undetectable BCR-ABL1 transcripts on tyrosine kinase inhibitor therapy remain disease-free after drug discontinuation. In contrast, 60% experience return of detectable disease and have to restart treatment, thus highlighting the need for an improved method of identifying patients with the lowest likelihood of relapse. Here we describe the validation of a personalized DNA-based digital PCR (dPCR) approach for quantifying very low levels of residual disease, which involves the rapid identification of t(9;22) fusion junctions using targeted next-generation sequencing coupled with the use of a dPCR platform. t(9;22) genomic breakpoints were successfully mapped in samples from 32 of 32 patients with early stage disease. Disease quantification by DNA-based dPCR was performed using the Fluidigm BioMark platform on 46 follow-up samples from 6 of the 32 patients, including 36 samples that were in deep molecular remission. dPCR detected persistent disease in 81% of molecular-remission samples, outperforming both RT-dPCR (25%) and DNA-based quantitative PCR (19%). We conclude that dPCR for BCR-ABL1 DNA is the most sensitive available method of residual-disease detection in chronic myeloid leukemia and may prove useful in the management of tyrosine kinase inhibitor withdrawal. (J Mol Diagn 2016, 18: 176e189; http://dx.doi.org/10.1016/j.jmoldx.2015.09.005)

The fusion oncogene BCR-ABL1 is the hallmark for chronic myeloid leukemia (CML).1 The BCR-ABL1 tyrosine kinase protein is involved in the cellular phenotype of CML and is therefore a rational target for therapy via tyrosine kinase inhibition, a treatment approach that has revolutionized patient outcome.2 The measurement of BCR-ABL1 transcripts via

Supported by Leading Leukemia Research (LEUKA) charity grant 06/ Q0406/47, the National Institute for Health Research Biomedical Research Center Funding Scheme, and the Imperial College High Performance Computing Service. LGC is the United Kingdom’s designated National Measurement Institute for chemical and bio-measurement. Disclosures: None declared.

Copyright ª 2016 American Society for Investigative Pathology and the Association for Molecular Pathology. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jmoldx.2015.09.005

NGS and dPCR for MRD Monitoring in CML quantitative RT-PCR (RT-qPCR) is the most widely used method of monitoring residual disease in patients with CML, an essential aspect of modern disease management.3 Although tyrosine kinase inhibitors (TKIs) are routinely administered indefinitely, recent studies indicate that 40% of patients who achieve undetectable BCR-ABL1 by RT-qPCR (molecular remission with 5-log reduction on the international scale; MR5) on imatinib that is sustained for at least 2 years, will remain disease-free after drug discontinuation,4e8 raising the possibility of an operational cure in this subgroup of patients. However, the safe introduction of a TKI-withdrawal strategy would require a reliable and cost-effective method for the identification of those patients with the lowest likelihood of relapse. The probability of relapse after withdrawal is likely related to persistence of residual disease, which may include transcriptionally quiescent, TKI-resistant, leukemic stem cells4,9 at a level that is below the threshold of detection by the gold-standard RT-qPCR (105).10,11 A means of detecting these cells that does not depend on oncogene transcription might be clinically valuable. Here we describe a DNA-based method of detecting and quantifying low levels of BCR-ABL1epositive disease that improves on previous methodologies in two key areas.12,13 First, the identification of BCR-ABL1 fusion junctions is undertaken by targeted next-generation sequencing (NGS), allowing for the rapid generation of high-performance DNAbased hydrolysis probe assays that are specific to the unique molecular footprint of each patient’s CML clone. Second, we sought to further enhance the sensitivity of a DNA-based approach by optimizing the technique for use on a digital PCR (dPCR) platform, which provides absolute molecular quantification without the need for a standard curve. When applied to samples with undetectable disease by RT-qPCR, DNA-based dPCR provided a marked improvement in sensitivity, not only over RT-qPCR but also compared to real-time qPCR of DNA and to digital RT-PCR (RT-dPCR).

Materials and Methods Patient Cohort To validate the NGS part of this technique, we studied 32 CML patients treated at the Hammersmith Hospital (London, UK). Twenty-two patients were on treatment with TKI, and 10 had received allogenic stem cell transplants. RNA extracted from diagnostic samples had been previously used for establishing the fusion type using multiplex PCR as previously described.14 Ethical approval was provided by the NHS Research Ethics Committee and informed consent was obtained from all patients in accordance with the Declaration of Helsinki. Forty-six follow-up samples from six of these patients (A1 to A6; median, five samples per patient) collected at least 28 months from diagnosis (follow-up range, 28 to 178 months) were used for comparing the sensitivities of different residual-diseaseemeasuring technologies, as described in Comparison of Quantification Methods. Thirty-six of these

The Journal of Molecular Diagnostics

-

jmd.amjpathol.org

samples were in deep molecular remission (MR), as defined by Cross et al,15 whereas the remaining 10 were at the level of MR3 (previously termed major molecular response) or above. Of the six patients, A1 and A6 received TKI therapy and the remaining four were post-transplantation (A2 to A5).

NGS Identification of Fusion Junctions DNA Extraction Peripheral blood cell lysates, prepared using Buffer RLT (Qiagen, Manchester, UK; catalog no. 79216) and stored at 80 C, were collected from all 32 patients either from the time of diagnosis or when the level of residual disease was 10% by BCR-ABL1 RT-qPCR. DNA was extracted from 200 mL of the peripheral blood cell lysates using the QIAamp DNA Blood Mini Kit (Qiagen; catalog no. 51106) following the manufacturer’s protocol on the QIAcube robotic workstation (Qiagen). DNA was quantified using the Qubit fluorometric method (Qubit dsDNA BR Assay Kit; Invitrogen, Foster City, CA; catalog no. Q32850). DNA was further assessed for impurities using the 260:280 and the 260:230 ratios measured using a spectrophotometer. The integrity of the DNA was assessed on a 2% agarose gel. Targeted NGS Fusion Mapping Targeted NGS Library preparation was performed using the TruSeq DNA sample preparation and PCR kits, according to the manufacturer’s instructions (Illumina, Cambridge, UK; reference nos. 15025064 and 15027084). Briefly, 1 mg of DNA from each sample was fragmented using the Covaris DNA shearing system (Covaris S2, Woburn, MA). The fragments were end-repaired and ligated with Illumina adaptors and barcodes for multiplexing. Individual library quality and size were assessed by running an aliquot of each library on a bioanalyzer using a highsensitivity DNA chip (2100 Bioanalyzer, Agilent Technologies, Santa Clara, CA) and/or capillary gel electrophoresis (Qiaxcel system; Qiagen). The average insert size was 400 bp (range, 100 to 900 bp). Individual libraries were purified using Agencourt Apure XP beads (A63881; Beckman Coulter, Jersey City, NJ), quantified using Qubit, and an equimolar amount of each library was pooled together into a final multiplexed library. The 32 patients were grouped into four batches, comprising multiplexes of 6, 12, 8, and 6 patients, respectively. For the targeted sequencing, each pooled library was hybridized twice overnight with a pool of synthetic oligonucleotide probes custom-designed using Illumina’s DesignStudio online portal to specifically capture the entire coding and noncoding regions of the BCR and ABL1 genes (100 Kbp upstream and downstream from each gene, respectively). Probe details and coordinates are provided in Supplemental Table S1. Sequence capture was followed by extensive washing and purification steps according to the manufacture’s protocol (TruSeq Custom Enrichment Kit, 24, Box 1 and 2; Illumina, reference nos. 15022030 and 15022031). Bridge amplification, cluster generation, and

177

Alikian et al Table 1 Patient

Details of the Primers used for Characterize BCR-ABL1 Fusion Junctions in 32 CML Patients Fusion gene

A1

ABL1-BCR

A2

BCR-ABL1 ABL1-BCR

A3

BCR-ABL1 ABL1-BCR

A4

ABL1-BCR

A5

BCR-ABL1 ABL1-BCR

A6

BCR-ABL1

B1

BCR-ABL1

B2

BCR-ABL1

B3

BCR-ABL ABL1-BCR

B4

BCR-ABL

B5

BCR-ABL

B6

BCR-ABL

B7

BCR-ABL ABL1-BCR

B8

BCR-ABL

B9

BCR-ABL

B10

BCR-ABL ABL1-BCR

B11

BCR-ABL1

B12

BCR-ABL1

C1

BCR-ABL1

C2

ABL1-BCR

C3

BCR-ABL1

C4

ABL1-BCR

Primer F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R

Sequence 0

Strand 0

5 -GTTCCTACCAGCACCCTTGA-3 50 -CTCTCTCCCAACCCCATTC-30 50 -TTCACGCCAGACCACAATTA-30 50 -ACTGAAACCAGCCAATGGAC-30 50 -GCTCTAGGCTTTCCTCAGCA-30 50 -TGTGATGACAGGGATGGGTA-30 50 -CCCATGACACTGGCTTACCT-30 50 -GCTCTTTGCCCACTCCACTA-30 50 -CTTAGGCACCAGCTCGTAGG-30 50 -GGGAAACACCAGCGTTTATG-30 50 -AACCCAAAAAGGAGGACTTGA-30 50 -ACGGCGACACACAATACAAA-30 50 -GATGCTGACCAACTCGTGTG-30 50 -CCCAGGGATGGTAAAAACCT-30 50 -AAAATGTACGGGGGAGAAGC-30 50 -AGATCCAAGGCACAGAGCAT-30 50 -TCTTGCGCAGATGATGAGTC-30 50 -TGAACAACCTCCCTGTTTCC-30 50 -GCTGTTTGCGCTCACATTTAC-30 50 -CAGCCTCCCAAGTAGCTGAG-30 50 -TGTCATCGTCCACTCAGCCAC-30 50 -GGCAGACAGAGTGAGACTCCATC-30 50 -GGAATTGTTTTTCCCGGAGT-30 50 -CCAGGAACAGGCTTTGTTTAA-30 50 -CCATTGGTGAACTGCTCCTT-30 50 -CGGCCTCGAGAAACTTACAC-30 50 -GATGAGTCTCCGGGGCTCTA-30 50 -CTCGTTCTGTCGCTAGGGTG-30 50 -GCTCTTACAGACCATGTGGGT-30 50 -AGGACTGAGGCTGGAAGTCA-30 50 -TTGTGCTGGTTGATGCCTTC-30 50 -TGCAATCTCTCTCTCCAAGGA-30 50 -TGGACAAGGTGGGTTAGGAG-30 50 -AGTAAGAGCTGACGTGTATTGTGC-30 50 -CCAGGGTCTCATAACCAAGG-30 50 -AAGGTTCCAAGGACAGCAGA-30 50 -CATGTCCACTTCTCCCCACA-30 50 -TTCACGTGATCCCTCTGCCT-30 50 -ACTTCTCCAGCACTGAGCTG-30 50 -GTCCCAATTAACGGTGGAAA-30 50 -ATCGTCCACTCAGCCACTG-30 50 -TTATCAACATTCACATCTCACAGG-30 50 -CAGCTGAAGAAGGTCTGGATTAGTA-30 50 -CTGGTAAGCTTTCTGTCTCCACA-30 50 -GTGAAGGCTGGTAACACATGAG-30 50 -CCCAACCAGCTCACTTTACB-30 50 -GGAAACAGGGAGGTTGTTCA-30 50 -GCTACAAGGAACACGCAACA-30 50 -GCCCATGACACTGGCTTACC-30 50 -CACACCAAGCCTCCCAAGTT-30 50 -GAGATGGAGTTTCACCGTGC-30 50 -GCAAACCAGTGACTCGAAGT-30 50 -CCCATGACACTGGCTTACCT-30 50 -TGGCCTGCTTGGTTAACTTT-30 50 -TTTGGTGGTTGGGTTGCAAA-30 50 -TATGAGGCAGCCAGAGACAG-30

þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ  þ 

Amplicon size, bp

Tm,  C

580

61

233

61

299

61

778

61

938

61

336

61

340

61

235

61

932

61

201

61

246

65

286

61

230

61

230

61

284

61

200

61

280

61

310

61

211

60

258

61

664

61

332

61

384

61

671

60

228

61

170

61

224

60

183

61

(table continues)

178

jmd.amjpathol.org

-

The Journal of Molecular Diagnostics

NGS and dPCR for MRD Monitoring in CML Table 1 Patient C5

(continued ) Fusion gene BCR-ABL1 (þ/) ABL1-BCR (þ/)

C6

BCR-ABL1 BCR-ABL1 (þ/) ABL1-BCR (þ/)

C7

BCR-ABL1 ABL1-BCR

C8

BCR-ABL1

D1

BCR-ABL1

D2

BCR-ABL1

D3

BCR-ABL1 ABL1-BCR

D4

ABL1-BCR

D5

BCR-ABL1 ABL1-BCR (þ/)

D6

BCR-ABL1 ABL1-BCR

Primer

Sequence 0

0

5 -CTAGGCAGTGGGCACCTGTA-3 50 -GTAAGAAATCTTTGTGTCTACCCTAAGG-30 50 -GTGGTCTGTGTAAGAAATCTTTGTG-30 50 -GGCAGTGGGCACCTGTAAT-30 50 -CTGGAGTCCGGGTGTCCTC-30 50 -TCACGTTTGAGGCTGTGGAA-30 50 -ATGATGACAGTGAGTGTGGCC-30 50 -TGTCGCATTGAAAGATGACACTTA-30 50 -GCTGCACGATTAGTGTTGTACATT-30 50 -AACTCCCTGGCATGGTGG-30 50 -CCCATGACACTGGCTTACCT-30 50 -CATACTCCGTTCCAGCGG-30 50 -ATGATCTCATCCGCTGGAAC-30 50 -CCAGCTCCCAGGATCTGAG-30 50 -CCCCGTTTCCGTGTACAGG-30 50 -CTAGGACCCTGGAGCACTGT-30 50 -TGCACCTCTTTTCCAACCTC-30 50 -TGCAGCCAGTCCCTTAGTCT-30 50 -TCTGCTGTCCTTGGAACCTT-30 50 -ATGAGGGAAGAGGGAGGAGA-30 50 -TGTGACCTTCTCCATGTCCA-30 50 -TCTGCCACACAAAGAACCTG-30 50 -CAAGGAACTGCCCTATTCCA-30 50 -CAACATTCGTTCACTCAGTCG-30 50 -TTGGGAAGAGAAGGGAACCT-30 50 -AGGCAGTGTCACAGCACAAC-30 50 -TCACGCCAGACCACAATTAG-30 50 -GTCCCAATTAACGGTGGAAA-30 50 -TGTAATCCCAGCAATTTGGGA-30 50 -CACAGCAGGCTGCCTGG-30 50 -GAGTTGGAGACCAGCCTGAC-30 50 -CCTCCCTCACCTCCACAAA-30 50 -CCAACCCCTGCCCTTTTAAA-30 50 -CGATTCTCTTCCCTCAGCCT-30

F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R F R

Strand

Amplicon size, bp

Tm,  C

þ þ þ þ þ      þ  þ  þ  þ  þ  þ  þ  þ  þ  þ þ þ  þ 

2722

61

220

61

206

61

282

61

285

61

300

61

350

61

239

60

248

61

185

61

206

61

300

61

200

61

244

60

189

65

183

61

191

61

Primers were designed based on the results of targeted next-generation sequencing. CML, chronic myeloid leukemia; F, forward; R, reverse.

150-bp paired-end sequencing were performed on an Illumina MiSeq after captured library quantification by qPCR using Kapa Library Quantification Kit (kit code KK4835; Kapa Biosystems Ltd., Bedford Row London, UK). The median read coverage across BCR and ABL1 targets was 100 (range, 50 to 150; minimum base call quality, Q20). Bioinformatics-based prediction of fusion junctions Bioinformatics analysis of sequencing reads was performed according to the following steps. First, TruSeq adaptor sequences were removed using the Cutadapt16 software version 1.9.1. Trimmed reads were then aligned to the human genome reference assembly GRCh37 using BWA-MEM17 Aligner software version 0.7.2. Duplicate reads were marked in the BAM files, and the coverage metrics for targeted regions were collected using Picard software version 1.85 (http://picard.sourceforge.net). CREST18 and BreakDancer19 software were used for extracting split reads (single reads composed of material from two noncontiguous

The Journal of Molecular Diagnostics

-

jmd.amjpathol.org

genomic regions) and discordant pairs of reads (in which individual reads in a pair do not map at the expected distance and/or orientation), respectively, that were likely to mark the site of structural rearrangement. Fusion Junction Confirmation All predicted fusion junction sequences were validated via conventional Sanger sequencing. Primer sequences are listed in Table 1. For the validation of the fusion junctions predicted by split reads, primers were designed at least 200 bp upstream and downstream of the putative fusion junctions, whereas in cases in which the fusion was alluded to via discordant read pairs, primers were designed 200 bp upstream and downstream of the end of the read closest to the predicted junction on either side of the junction. Primer320 online software version 0.40 was used for primer design. The Qiagen Fast Cycling PCR Kit (Qiagen, catalog no. 203743) was used for amplification of junction sequences, which contain a high-fidelity HotStarTaq Plus DNA

179

Alikian et al Table 2 Sequences of Primer and Probe Components of Six Patient-Specific Hydrolysis Probe Assays for the BCR-ABL1 or ABL-BCR DNA Junction Used for dPCR and qPCR Patient A1

Fusion

A5

1. ABL1-BCR 2. 3. 4. BCR-ABL1 5. 6. 7. BCR-ABL1 8. 9. 10. ABL1-BCR 11. 12. 13. BCR-ABL1

A6

14. BCR-ABL1

A2

A3

A4

Assay (digital PCR) A1_ABL1-BCR_F A1_ABL1-BCR_R A1_ABL1-BCR_P A2_BCR-ABL1_F A2_BCR-ABL1_R A2_BCR-ABL1_P A3_BCR-ABL1_F A3_BCR-ABL1_R A3_BCR-ABL1_P A4_ABL1-BCR_F A4_ABL1-BCR_R A4_ABL1-BCR_P A5_BCR-ABL1_F A5_BCR-ABL1_R A5_BCR-ABL1_P A6_BCR-ABL1_F A6_BCR-ABL1_R A6_BCR-ABL1_P

Sequence 0

0

5 -GTTTAGTTGATGACACACCTGACTCTAA-3 50 -CCCAGGCTGGAATGCAGT-30 6FAM-50 -CCTGGCGGAGGTTG-30 50 -GGTGATGTGGAAAAGACCTGTGA-30 50 -CATCCACATATATAGGACTCCCAACAC-30 6FAM-50 -CTTCTCCATGTCCACTTC-30 50 -CAGATCCTGGGAGCTGGTGA-30 50 -GATGGTGTTTCACCACATTAGCC-30 6FAM-50 -CGGATCACAAGGTCA-30 50 -ATCACATAACCTAAAACTTAACATTGACACC-30 50 -CGCTAACAAAGGCAGACAAAAAG-30 6FAM-50 -TGGAAAGAGACTTAAAAAG-30 50 -TGATGGGACTAGTGGACTTTGGTT-30 50 -TCTACACCCATGTGGGAGCAG-30 6FAM-50 -AGAAGGAAGAGCTATGCTT-30 50 -GGATACTACTTTTTTTTTCCTTTCCCTC-30 50 -GTAACATTAACTGTTGGAAAACATGTCTTAG-30 6FAM-50 -CTTAAATAGCTCTAGTTCCCT-30

Tm,  C

Length, bp

58 58.2 70 59 59 70 59.8 59.2 69 60 59 69 59.1 59 70 58.4 58.9 70

116

127

145

109

97

162

F, forward primer; P, probe; qPCR, real-time quantitative PCR; R, reverse primer; Tm, melting temperature.

polymerase capable of functioning over long distances. For all PCRs, two negative controls, one sample of pooled DNA extracted from eight CML patients (pool 1) and DNA extracted from a normal individual were included in addition to a no-template control. PCR was performed using the following temperature-cycling parameters: 94 C for 5 minutes, 30 cycles at 94 C for 30 seconds, 55 C to 65 C for 30 seconds, 72 C for 37 seconds, followed by a final extension step at 72 C for 10 minutes. For Sanger sequencing of PCR products, two independent cycling PCR reactions, one forward and one reverse, were performed per patient sample using the BigDye Terminator version 3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA) under the following conditions: 96 C for 1 minute, 25 cycles at 96 C for 10 seconds, 55 C for 15 seconds, and 60 C for 4 minutes, before storage at 4 C. PCR products were purified with the BigDye XTerminator Purification Kit following the manufacturer’s protocol. Sequencing was performed on the ABI 3500xL Genetic Analyzer (Life Technologies, Foster City, CA). Sequence quality and traces were analyzed using the ABI’s Sequence Analysis software version 5.2. Sequences obtained by Sanger sequencing were aligned to the reference genome for BCR and ABL1 using BLAST21 (Basic Local Alignment Search Tool; NCBI) or the Blast Like Alignment Tool22 (BLAT; University of California at Santa Cruz, Santa Cruz, CA) genome-alignment tools.

of three other quantitative PCR methods, namely, RTqPCR, qPCR, and RT-dPCR. Each was performed as follows.

The sensitivity of dPCR in identifying and quantifying BCR-ABL1epositive residual disease in 46 post-treatment CML samples from six patients was compared to those

DNA-Based dPCR Design and optimization of TaqMan FAM-MGB hydrolysis probe assays For six patients (A1 to A6), qPCR assays were designed specific to each patient’s fusion junction using Primer Express 3 software version 3.0.1 (Life Technologies) and the AutoDimer tool.23 Assays were designed specific to the BCR-ABL1 fusion junction where available (four patients) or to ABL1BCR, if the BCR-ABL1 junction sequence was unavailable (two patients) (Table 2). Primer specificity was assessed using a SYBR Green melt-curve assay (Sigma-Aldrich, St. Louis, MO) following the manufacturer’s protocol, in addition to capillary electrophoresis. DNA extracted from each patient’s diagnostic or highlevel disease sample was used for preparing standard curves to evaluate assay efficiency, quantitative range, and limit of detection. Primer and hydrolysis probe concentrations were optimized by real-time PCR using a matrix of eight different concentrations run on an Applied Biosystems StepOnePlus Real-Time PCR System (Life Technologies). The combination with the highest fluorescence and lowest quantification cycle (Cq) values was subsequently used. The standard curves produced acceptable slopes (3.2 to 3.6) and correlation coefficients (0.98% to 0.99%), acceptable amplification efficiency figures (80% to 110%), and acceptable dynamic ranges, consistent with the expected theoretical sensitivity range. Quantitative ranges (LoQ) varied from 103 to 104 and sensitivity ranges [limit of detection (LoD)] varied from 104 to 5  105 (Supplemental Table S2). SYBR Green assays

180

jmd.amjpathol.org

Residual-Disease Quantification

-

The Journal of Molecular Diagnostics

NGS and dPCR for MRD Monitoring in CML Table 3 dPCR and RT-dPCR Threshold and Cq Parameters Defined for Each of the Six Assays Performed on DNA (A1-A6) and One Assay Performed on cDNA (Wessex Plasmid) Analytical method

Assay

Cq range

Threshold

dPCR dPCR dPCR dPCR dPCR dPCR RT-dPCR

A1 A2 A3 A4 A5 A6 Wessex plasmid

20e32 20e32 20e32 18e40 16e40 20e40 20e32

0.05 0.05 0.05 0.03 0.02 0.025 0.03

Cq, quantification cycle; dPCR, digital PCR; RT-dPCR, digital RT-PCR.

showed a unique melting curve per assay, and capillary gel electrophoresis produced a unique band of the expected size. Preamplification Measurement of residual disease was performed by dPCR in the follow-up samples either directly using unamplified DNA or including a preamplification step according to a previously published protocol24 using TaqMan Preamp Master Mix (Invitrogen, part no. 4391128). In brief, a total reaction mixture of 20 mL was prepared using 10 mL of 2 preamp mix, 5 mL of 0.2 patient-specific assay, 2 mL of distilled water, and 150 ng of DNA in 3 mL. Preamplification was performed using the following thermocycling conditions: 95 C for 10 minutes, then 14 cycles at 96 C for 15 seconds and 60 C for 4 minutes. Amplification products were diluted 1:5 in 1 Tris-EDTA, pH 8, and subsequently used for dPCR quantification. We evaluated the efficiency of preamplification of DNA extracted from 46 follow-up samples from six CML patients using a standard curve of 4 points per assay, prepared from each of the corresponding six patients’ early disease material. The standard curves produced acceptable slopes (3.1 to 3.8), correlation coefficients (0.994% to 0.998%), and amplification efficiency metrics (81% to 114%) (Supplemental Table S3). Of the 46 follow-

up samples, 10 had previously demonstrated RT-qPCR positivity at MR3 or above and thus served as positive controls; all 10 amplified within expected ranges. dPCR quantification Microfluidic dPCR was performed using Biomark HD System with the qdPCR 37k integrated fluidic circuits (Fluidigm, San Francisco, CA). Each integrated fluidic circuit contained 48 panels, each with 770 partitions of approximately 0.84 nL reaction volume (Fluidigm, product code 100-6151). Reaction mixes of 6 mL were prepared containing 3 mL of 2 ABI Fast Advanced Master Mix (Invitrogen; catalog no. 4444965), 0.6 mL of 20 GE Sample Loading Reagent (Fluidigm; catalog no. 85000820), 0.3 mL of 20 patient-specific TaqMan assay, 0.3 mL of distilled water and 1.8 mL of 50 ng/mL template DNA. A partition size of nine panels per sample was used for allowing the quantification of an equal amount of DNA as in qPCR (150 ng). Five microliters of the prepared reaction mixes was loaded into the corresponding inlets on each array chip, and the BioMark IFC controller MX (Fluidigm) was used for uniformly partitioning the reactions across the panels. Thermocycling was performed using the fast mode (5.5 C per second) as follows: 95 C for 60 seconds, 40 cycles at 96 C for 5 seconds, followed by 60 C for 20 seconds. Data were collected using Data Collection software version 4 (San Francisco, CA). Data were analyzed using Digital PCR Analysis software version 4.0.1 (Fluidigm). The positive partitions per nine panels for each assay were grouped (k), and Poisson distribution was used for estimating the mean number of template copies (l) per partition in all of the nine panels (n), where25 l Z ln(1  k/n). Estimated target copy number per microliter and per reaction volume were calculated in addition to the 95% CIs to account for the counting uncertainty. The formulas used for these calculations and are formatted into Supplemental Tables S4 and S5, which could be used as a template, and are as follows. The probability of having more than one

Figure 1 Pictorial representation of breakpoint mapping read types obtained by next-generation sequencing of ABL1 and BCR sequences in a patient with chronic myeloid leukemia. A: An example of split reads. These single reads are composed of material from two noncontiguous regions on the reference genome that map directly across the fusion junction and are thus capable of identifying breakpoints to base pair resolution. Based on the sequence information obtained from these reads, sequencing primers (gray arrows) are designed for confirmation by Sanger sequencing. B: An example of discordant pairs of reads in which individual reads in a pair map to different genes. Clusters of discordant read pairs allow indirect estimation of the position of the fusion site to within up to 1 Kbp. C: The effect of an inversion at the fusion site on read orientation. BCR and ABL1 components of the split read run in opposite genomic directions.

The Journal of Molecular Diagnostics

-

jmd.amjpathol.org

181

Alikian et al Table 4

Target Quantities Assessed Using Four Molecular Quantification Methodologies in 46 Follow-up Samples from Six Patients with CML

Patient demographics

DNA and RNA copy numbers measured on digital PCR platform

DNA and RNA copy numbers measured on real-time quantitative PCR platform

dPCR, RT-dPCR, Month from dPCR, copies/mL, RT-dPCR, copies/mL, Patient Sample therapy start copies/mL preamp copies/mL preamp qPCR A1 A1 A1 A1 A1 A1 A1 A1 A1 A1 A2 A2 A2 A2 A2 A2 A2 A3 A3 A3 A3 A3 A3 A3 A3 A3 A4 A4 A4 A4 A4 A4 A4 A5 A5 A5 A5 A5 A5 A6 A6 A6 A6 A6 A6 A6

A1_46 A1_37 A1_38 A1_39 A1_40 A1_41 A1_42 A1_43 A1_44 A1_45 A2_1 A2_2 A2_3 A2_4 A2_5 A2_6 A2_7 A3_28 A3_29 A3_30 A3_31 A3_32 A3_33 A3_34 A3_35 A3_36 A4_8 A4_9 A4_10 A4_11 A4_12 A4_13 A4_14 A5_15 A5_16 A5_17 A5_18 A5_19 A5_20 A6_21 A6_22 A6_23 A6_24 A6_25 A6_26 A6_27

54 68 70 71 95 134 163 169 175 178 33 63 81 104 116 122 143 48 92 93 96 102 103 106 110 114 33 39 45 55 58 59 62 28 35 66 67 69 77 49 77 102 104 109 110 112

1622 11 1 0 0 1 1 1 0 0 1833 0 0 0 0 0 0 75 3 6 4 4 6 1 2 2 808 127 268 0 1 1 0 235 637 0 0 1 0 64 3 0 0 0 0 0

46 2 0 0 1 1 9 0 11 4 5 2 2 3 52 53 55 38 64 65 72 46 60

416 301 177 0

0 0 4 0

15 212 3 90 6

1475 7 0 0 0 0 0 0 0 0 1167 0 0 0 0 0 0 68 3 0 0 0 0 0 0 0 735 116 243 1 0 0 0 214 579 0 0 0 0 58 2 0 0 0 0 1

67 0 38 452 5 0 0 30 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0

62 1 0 0

0 0 0 1

63 0 0 0 19

BCR-ABL1, ABL1, MR RT-qPCR, IS copies/3 mL copies/3 mL level

0.140000* 3.114583* 8847 0.000814y 0.004485y 3 0.000641y 0.000000z 0 0.000018z 0.000831y 2 0.000284z 0.004530y 4 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.000000z 0.000362y 1 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.680000* 24.91666* 10,000 0.000000z 0.002833y 9 0.000000z 0.003504y 10 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.001281* 0.079929* 409 0.000009z 0.000341y 1 0.000005z 0.000000z 0 0.000019z 0.000000z 0 0.000068z 0.000000z 0 0.000027z 0.000549y 2 0.000022z 0.002072y 7 0.000029z 0.000516y 3 0.000016z 0.000000z 0 0.163000* 3.459297* 4408 0.004000* 0.287023* 695 0.006000* 0.475534* 1460 0.000690* 0.003866y 6 0.000250y 0.000000z 0 0.000101y 0.000000z 0 0.000000z 0.000000z 0 0.045169* 1.194137* 1282 0.022476* 0.586851* 3474 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.000000z 0.000823y 6 0.145103* 0.163604* 348 0.011997* 0.013205* 14 0.000443* 0.000425y 1 0.000128y 0.000906y 2 0.000000z 0.000000z 0 0.000000z 0.000000z 0 0.000000z 0.001349y 3

68,300 20,000 16,000 72,000 26,400 71,300 68,200 82,700 74,900 91,500 12,000 95,000 128000 11,400 31,200 93,200 67,100 153,000 87,800 89,000 91,600 117,000 109,000 101,000 174,000 92,800 38,100 72,400 91,800 46,400 114,000 68,400 26,500 32,100 177,000 89,500 67,100 81,500 218,000 63,600 31,700 70,300 66,000 151,000 24,400 66,500

>MR3 MR4 MR4 MR5 MR4 MR4.5 MR4.5 MR5 MR4.5 MR4.5 >MR3 MR4.5 MR4 MR4 MR4 MR4.5 MR4.5 >MR3 MR5 MR4.5 MR4.5 MR5 MR5 MR4.5 MR5 MR4.5 >MR3 >MR3 >MR3 MR4 MR5 MR4.5 MR4 >MR3 >MR3 MR4.5 MR4.5 MR4.5 MR5 >MR3 >MR3 MR5 MR5 MR5 MR4 MR4.5

Empty cells indicate no preamplification. *Positive value within the limit of quantification of the standard curve. y Positive value within the limit of detection of the standard curve. z Negative value. CML, chronic myeloid leukemia; dPCR, DNA-based digital PCR; MR3, major molecular response; MR4, deep molecular response, specified as 4-log transcript level reduction on the international scale (IS)9; MR4.5, deep molecular response, specified as 4.5-log transcript level reduction on the IS9; MR5, deep molecular response, specified as 5-log transcript level reduction on the IS9; preamp, preamplification; RT-dPCR, RNA-based digital PCR; qPCR, DNA-based quantitative PCR; RT-qPCR, RNA-based quantitative PCR.

182

jmd.amjpathol.org

-

The Journal of Molecular Diagnostics

NGS and dPCR for MRD Monitoring in CML

Figure 2 Comparison of the sensitivity of four methods of disease quantification in 46 samples from six chronic myeloid leukemia (CML) patients. Quantification results are shown for the follow-up samples of six patients with CML according to substrate type and platform used. The four techniques are listed. Samples are represented by circles on a linear timeline (not to scale), with the earliest sample on the far left. The vertical dotted lines represent the point of achievement of molecular remission (MR)4 or lower [based on quantitative RT-PCR (RT-qPCR) analysis of the proceeding sample]. Digital PCR (dPCR) and RTdPCR values are after preamplification. LoD, limit of detection; LoQ, limit of quantification.

target molecule per partition is calculated as P Z k/n, where k is the number of positive partitions, and n is the total number of partitions. SD from the mean of P was estimated as S Z SQRT {[P  (1  P)]/n}. The upper and lower 95% confidence limits of P were estimated as: PL and PH Z [P þ/ (1.96  S)]. The upper and lower 95% confidence limits of l were estimated as lL and lH Z ln(1  PL) and ln(1  PH), respectively. Target copies per reaction volume were estimated as (Est targets) Z l  n. The upper and lower 95% confidence limits of the estimated target copies were calculated as (Est targets-L and Est targets-H) Z lL  n and lH  n. Target copies per microliter were estimated as Est targets/(DNA volume per panel  number of panels). And, the upper and lower 95% confidence limits of the estimated targets per microliter were calculated as Est targets-L/(DNA volume per panel  number of panels) and Est targets-H/(DNA volume per panel  number of panels). Positive and negative controls were run using the same partition size used per sample for each of the six DNA assays to assess platform function, amplification protocol, and level of background noise. Early disease samples from each of the six patients were used as positive controls. DNA from a normal donor and from a pool of eight CML samples were used as negative controls. Three no-template control (NTC) panels were included in each array to rule out contamination. The threshold and the Cq ranges were manually set specific to each assay and its background noise but consistent across all panels of the same assay (Table 3 and Supplemental Figures S1eS7). Linearity and the dynamic range of the Fluidigm platform were previously assessed.24 However, we also performed a limited experiment to assess these two parameters in our hands. We generated a series of dilutions of two synthetic oligonucleotides, each containing the fusion sequence of one CML patient (A3 and A6) (GeneArt; Life Technologies). Twofold dilutions of 4 to 512 molecules per panel were run in duplicate panels on the 48.770 chip array. Linearity was maintained at four molecules per panel (l Z 0.005;

The Journal of Molecular Diagnostics

-

jmd.amjpathol.org

approximately equal to 2 copies per mL); therefore, we considered this a conservative limit for reliable quantification Supplemental Figure S8). All of the normal controls we tested showed complete negativity; additionally, none of the normal controls showed any amplification. We therefore chose a range of 0.5 to 3 molecules per panel to be considered positive outside of the quantitative range. These thresholds have been previously applied to other tests in the measurement of residual disease in acute lymphoblastic leukemia using antigen receptor targets.26 RT-qPCR After each patient’s specific fusion type was identified in the diagnostic sample by multiplex PCR using previously described protocols,11,14 RT-qPCR was used for identifying residual disease in each of the 46 follow samples as previously described using best practice methodology.10,11,27,28 RNA was extracted using the RNeasy Mini Kit (Qiagen; catalog no. 74106) and the QIAcube robot (Qiagen). Reverse transcription and cDNA synthesis were performed using Moloney Murine Leukemia Virus Reverse Transcriptase (Invitrogen; catalog no. 28025-013) and random hexamers (Invitrogen; catalog no. 48190-018) as described in an earlier publication.10 RT-qPCR was performed using a duplex TaqMan MGB assay28 (a modified version of the Europe Against Cancer assay described by Gabert et al29) with 2 TaqMan Fast Advance Master Mix (ABI; catalog no. 4444965) in a final 20-mL reaction. The level of expression of target BCR-ABL1 molecules was expressed as a percentage ratio between BCR-ABL1 and ABL1 to obtain a normalized value for the gene independent of the integrity of the RNA and the efficiency of the RT reaction. RT-dPCR Positive and negative controls were used for assessing the amplification protocol and to establish the Cq range and the quantification threshold. cDNA from each

183

Alikian et al

Figure 3

Comparison of the DNA-based digital PCR (dPCR) and digital RT-PCR (RT-dPCR) readings in 46 samples from six chronic myeloid leukemia patients. DNA dPCR results are shown with blue bars and RT-dPCR results with green bars. The vertical red dashed lines represent the point of achievement of molecular remission (MR)4 or lower [based on quantitative RT-PCR (RT-qPCR) analysis of the proceeding sample]. Precise readings are provided in Table 4 and Supplemental Tables S7 and S8. Plotted dPCR and RT-dPCR values are after preamplification. The bar charts depict five scenarios: i) DNA is detected while RNA is not; ii) RNA is detected while DNA is not; iii) both DNA and RNA are detected, but DNA quantity is higher than that of RNA; iv) both DNA and RNA are detected with higher RNA quantity compared to DNA; and v) neither DNA nor RNA is detected.

patient’s presentation sample in addition to the inhouse standardized Wessex plasmid, which contains the e14a2 BCR-ABL1 cDNA junction, were used as positive controls. cDNA from a healthy individual was used as negative control. A partition size of 15 panels per sample was used for allowing the quantification of an equal volume of cDNA as in RT-qPCR (3 mL). All cDNA samples used for RT-dPCR quantification were the same as those used on the RT-qPCR platform to exclude potential variability introduced by the RT step. Three NTC panels were included in each array to rule out contamination. Using the Digital PCR Analysis software version 4 (Fluidigm), the quantification threshold was determined to be 0.03. The Cq range was manually defined as 20 to 32 (Table 3 and Supplemental Figure S9). Running a series of twofold dilutions prepared using the ERM-AD623 plasmid (produced by the European Commission for Reference Materials), a reference vector containing the BCR-ABL1 exonic fusion, we established linearity and found that accurate quantification was possible down to four molecules per panel (2 copies per mL) as described for DNA-dPCR (Supplemental Figure S10). Total reaction mixes, volume per inlet, thermocycling conditions, and data analysis software were as described for dPCR of DNA (see dPCR Quantification). Threshold and amplification range across the panels of the same assay were set according to the parameters defined during the validation stage.

qPCR of BCR-ABL1 DNA Hydrolysis probe assays were as previously detailed for dPCR (see dPCR Quantification). Assay performance was assessed following Minimum Information for Publication of Quantitative Real-Time PCR Experiments guidelines.30 To ensure accurate disease quantification, we also followed the European Minimal Residual Disease Consortium recommendations for qPCR in Ph-negative acute lymphoblastic leukemia,26 in which DNA extracted from each patient’s diagnostic sample is used for preparing standard curves (Supplemental Table S2). Patient-specific standard curves were prepared by performing 5-log dilutions of each patient’s diagnostic sample, including two half-log dilutions at the end of the scale. The samples were diluted in tRNA buffer (Sigma-Aldrich, St. Louis, MO). We used albumin as a control gene to normalize the total amount of quantifiable DNA included per assay to 150 ng/3 mL per 20-mL reaction. The standards were run in triplicate, and in sextuplicate for values