Large amplicon droplet digital PCR for DNA-based monitoring of pediatric chronic myeloid leukaemia

2.1 Patients and materials

Cryopreserved mononuclear cells, fresh or frozen blood samples and bone marrow samples collected from patients with CML participating in two consecutive pediatric CML trials (CML-paed I and CML-paed II, EudraCT 2007-001339-69) were included in the analysis. Informed consent of the patients or their legal guardians was obtained, in accordance with the Declaration of Helsinki.

For identification and characterization of genomic BCR-ABL1 fusion sequences, 178 individuals (78 Females, 100 Males; median age: 13 years; range: 0.3-22 years) were analysed. This cohort represents all children and adolescents diagnosed with CML in Germany during the collection period of 15 years.

2.2 Sequencing of genomic BCR-ABL1 fusion sites

DNA was isolated using the QIAamp^® DNA Blood Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions. Genomic fusion sequences were identified using a nested multiplex long-range PCR (MLR-PCR) assay, as described previously.12

2.3 Analysis of breakpoint characteristics

Patient-specific BCR-ABL1 fusion sequences were aligned to the human genome (hg19, UCSC Genome Browser). Breakpoint positions are listed in Table S1. Repeat elements at the fusion sites were identified using the RepeatMasker tool (http://www.repeatmasker.org/). Genomic fusion sites were then analysed for co-localization with repeat elements, recombination-related DNA sequence motifs, topoisomerase II binding sites, translin binding sites, heptamer/nonamer recombination signals, recombination signal sequences (Recombination Signal Sequences Site tool, http://www.itb.cnr.it/rss/), palindromic sequences (EMBOSS explorer [http://emboss.bioinformatics.nl/cgi-bin/emboss/palindrome] and palindromic sequences finder tool [http://www.biophp.org/minitools/find_palindromes/demo.php]), human minisatellite core sequences, human minisatellite conserved sequences, hypervariable minisatellite recombination sequences, DNA polymerase frameshift hotspots, immunoglobulin heavy chain class switch repeats, LTR-IS motifs and human replication origin consensus sequences.

Components of the free software environment R (http://www.r-project.org) were used for kernel density analysis, as described previously.14 WebLogo Version 2.8.2. was used for graphical presentation of nucleic acid multiple sequence alignments of individual patient BCR-ABL1 fusion sequences, as well as patient-specific BCR and ABL1 wild-type sequences at the corresponding positions. 15, 16.

2.4 Quantification of tumour-specific genomic DNA using individual For comparative analysis of RNA- and DNA-based BCR fusion sequences

For comparative analysis of RNA- and DNA-based BCR-ABL1 therapy assessment, we evaluated a total of 687 peripheral blood or bone marrow samples from 55 individuals. Patients were selected based on the availability and quality of follow-up samples.

Chronic myeloid leukaemia cells were analysed by detection of patient-specific BCR-ABL1 fusion genes using breakpoint spanning primer and probe sets. The amplicon length of the personalized assays varied from 80 to 1330 bp. Quantification was performed by ddPCR on a QX200 reader system (BioRad) using probe-based quantification assays with 2× ddPCR Supermix for Probes (no dUTP). A regular QX200 reaction contained 7 µL of DNA (200 ng/µL) for high sensitivity BCR-ABL1 detection and 1 µL of DNA (200 ng/µL) for quantification of the reference gene, albumin (ALB). No template control (NTC) and wild-type negative controls were run on each plate, with adjusted amounts of DNA. All assays were conducted in duplicate. Amplitudes of FAM-TAMRA single quenched probes were compared with those of FAM-ZEN-IBFQ double quenched probes, to identify the best combination for discrimination between amplicon-negative and -positive droplets. A sample was regarded as positive when at least three positive droplets were detected.

For ddPCR assays with amplicon products ≥200 bp, the cycling protocol was extended to a three-step method, with 40 cycles of 94°C for 30 seconds, 60°C for 1 minute, and 72°C for 2 minutes. To calculate the absolute number of BCR-ABL1 copies, fusion-specific probe signals were normalized to that of the single copy human ALB gene. The sensitivities of the patient-specific ddPCR assays for detection of BCR-ABL1 DNA copies were calculated based on the results from samples with the lowest quantifiable BCR-ABL1 copy numbers or patient-individual dilution series, and were 0.0032%-0.00016% (molecular response [MR], 4.5-5.7).

2.5 Statistical analysis

Co-localization of genomic breakpoints with repeat regions and DNA sequence motifs was statistically analysed using the Fisher's exact test. Differences between the ratio of mRNA% (BCR-ABL1 transcript/ABL1 transcript) and DNA% (BCR-ABL1 fusion gene/ALB gene) at the day of diagnosis and 3 months after treatment initiation were assessed using the Mann-Whitney U-test. MRD data from quantification of the BCR-ABL1 fusion at the RNA and DNA levels were compared using Spearman correlation statistics.

3.1 Localization of BCR and ABL1 breakpoints in repeat regions

Kernel density analyses of 178 genomic BCR-ABL1 fusion sequences from pediatric CML samples (breakpoint positions listed in Table S1) confirmed a bimodal distribution of genomic breakpoints within the BCR breakpoint cluster region. One peak of the distribution resulted from an accumulation of breakpoints in and around an Alu repeat element. Interestingly, females and males exhibited different patterns when analysed separately. ABL1 breakpoints were evenly distributed within the breakpoint cluster region (Figure S1).

Statistical analyses revealed no significant over-representation of breakpoints within repeat regions or other recombination-related sequence motifs (data not shown). Nevertheless, due to the high proportion of repeat elements in both genes involved, particularly ABL1, there was a high likelihood of localization of fusion sites in repeat regions. Detailed analyses showed that 23% of BCR breakpoints and 54% of ABL1 breakpoints were located within these elements, mainly Alu repeats or long interspersed nuclear elements (LINEs), in accordance with the expected distribution, calculated from the overall proportion of these elements in the breakpoint cluster regions (Figure 1A). As fusion sequences are composed of both partner genes, 52% of patients with pediatric CML have fusion sites within repeat elements of one involved gene and 12% have fusion sequences located in repeat regions of both genes (Figure 1B). Multiple fusion sequence alignments of the investigated 178 fusion sites illustrated the preferred localization of breakpoints within repetitive sequences, particularly in the ABL1 gene (Figure 1C). Only 36% of fusion sequences in pediatric CML fusion sites were not associated with repeat regions and were therefore perfectly suited for a high sensitivity, DNA-based conventional qPCR assay design.

Extension of the target sequence to 75 nucleotides flanking either side of the DNA breakpoint (the standard sequence range required to place sense and antisense primers, and probes) reduced the proportion of fusion sequences unaffected by complicating repeat elements to 33% in the ABL1 gene and 64% in the BCR gene (Figure 1A).

3.2 BCR-ABL1 fusion sequence quantification by droplet digital PCR

Highly specific and sensitive DNA quantification assays require primer and probe positions outside of repeat regions; therefore, we tested the effects on signal stability and sensitivity of increasing amplicons to lengths exceeding those conventionally used for qPCR (≥150 bp) in our probe-based ddPCR assays.

Various primer/probe combinations binding to the single copy ALB gene achieved highly specific quantification of the region of interest using amplicon lengths up to 925 bp, considerably beyond the typical amplicon size (≤150 bp) used for conventional qPCR assays. The absolute signal amplitude of amplicon-positive droplets reduced with increasing amplicon length; however, separation of amplicon-positive and -negative droplets (the relevant read-out for ddPCR) remained definite and allowed reliable quantification of the target region. The sensitivity reduction on amplification of a very large amplicon (925 bp) relative to the smallest analysed amplicon (81 bp) was minimal, with a factor of 1.9 between them (Figure 2A–C).

In the next step, BCR-ABL1 fusion sequences were quantified in individual genomic DNA samples from 55 patients with pediatric CML. Primers were positioned beyond the usual amplicon length region of 150 bp in 26 (47%) cases, to avoid binding to a repeat element on either side of the breakpoint. As expected, the distance between amplicon-positive and -negative droplets was reduced when very large amplicons were generated (Figure 2D). Nevertheless, the sensitivity of assays remained high, even for quantification of large amplicons (>1000 bp) (Figure 2E).

Reduced signal amplitudes of amplicon-positive droplets were easily compensated for using double quenched probes, which enabled stringent separation between amplicon-positive and -negative droplets (Figure 2F).

3.3 Comparative quantification of genomic BCR-ABL1 fusion genes and BCR-ABL1 fusion transcripts

To evaluate the potential clinical application of DNA-based MRD monitoring for CML patients with genomic breakpoints within, and adjacent to, repeat regions, we compared DNA-based quantification with standard RNA-based monitoring. Amplicon lengths varied from 80 to 1330 bp for DNA-based quantification by ddPCR.

We compared 687 blood or bone marrow samples collected from 55 patients with pediatric CML at initial diagnosis and during the course of treatment. Of those 687 samples, 47 were negative at the DNA and RNA levels, while six samples were quantifiable by RT-qPCR, but not using the DNA-based assay. Conversely, the DNA-based assay was able to quantify 64 samples that tested negative at the RNA level (Figure 3A). Insufficient RNA quality, due to incorrect sample handling, accounted for failed transcript quantification in 9 of 64 samples. The remaining 55 RNA-negative samples had BCR-ABL1 levels around MR4.0, indicating the benefit of supplementary DNA quantification for detection of quiescent, transcript-negative, CML cells. DNA fusion site copy number and standard RNA transcript copy number correlated well for BCR-ABL1 quantification (correlation coefficient, 0.9182 [P < 0.0001]; Figure 3A).

Figure 3B shows an exemplary course of MRD quantification for one pediatric patient, who was initially diagnosed at 6 years old. Four years after initiation of continuous TKI treatment with imatinib, BCR-ABL1 transcript levels decreased to non-detectable (DMR; <MR4.5), and the patient stopped imatinib treatment 32 months later. Five weeks after treatment discontinuation, the patient lost MR4.5, indicating an imminent relapse. Retrospective quantification of the genomic BCR-ABL1 fusion sequence revealed persistence of the CML clone at a detectable level, illustrating the potential of supplementary DNA-based MRD monitoring.

In addition, we investigated whether the RNA transcript/DNA fusion site ratio at diagnosis and 3 months after treatment reflected differences between good and poor responders to TKI treatment. Good responders (MR1.0, <10%, achieved 3 months after treatment initiation and MR3.0, < 0.1%, reached within 12 months of treatment initiation) exhibited lower BCR-ABL1 expression ratios than poor responders (Figure S2). Considering the fusion transcript type, patients with b3a2 transcripts had lower BCR-ABL1 expression ratios those with b2a2 transcripts at the day of initial diagnosis (P = 0.042), which is in contrast to results reported for adult patients.17 These findings require confirmation in larger cohorts.