- Email: [email protected]
Optimized Digital Droplet PCR for BCR-ABL
Accepted Manuscript Optimized digital droplet PCR for BCR-ABL Jacqueline Maier, Thoralf Lange, Michael Cross, Kathrin Wildenberger, Dietger Niederwieser, Georg-Nikolaus Franke PII:
S1525-1578(18)30071-0
DOI:
https://doi.org/10.1016/j.jmoldx.2018.08.012
Reference:
JMDI 742
To appear in:
The Journal of Molecular Diagnostics
Received Date: 26 February 2018 Revised Date:
5 July 2018
Accepted Date: 22 August 2018
Please cite this article as: Maier J, Lange T, Cross M, Wildenberger K, Niederwieser D, Franke G-N, Optimized digital droplet PCR for BCR-ABL, The Journal of Molecular Diagnostics (2018), doi: https:// doi.org/10.1016/j.jmoldx.2018.08.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Optimized digital droplet PCR for BCR-ABL Jacqueline Maier,* Thoralf Lange,*† Michael Cross,* Kathrin Wildenberger,* Dietger Niederwieser,* and Georg-Nikolaus Franke*
RI PT
From the Department of Haematology and Oncology,* Leipzig University, Leipzig; and the Department of Oncology and Haematology,† Asklepios Klinik Weissenfels, Weissenfels, Germany
Footnote: Portions of this work were presented in oral form at the Droplet Digital PCR Expert Meeting held May 30, 2017, in Frankfurt/Main, Germany.
SC
Running head: ddPCR signal intensity and background
M AN U
Disclosures: T.L. received fees and presentations of BMS and Novartis, research funding from Novartis. D.N. received research funding from Novartis. G-N.F. received fees and presentations of BMS and Novartis, research funding from Novartis. This work was supported by Novartis Pharma GmbH (Project HTAS-181, AN 72255619 to G-N.F. and D.N.).
AC C
EP
TE D
Corresponding author: Jacqueline Maier Leipzig University Department of Haematology and Oncology Johannisallee 32A 04103 Leipzig Email: [email protected]
1
ACCEPTED MANUSCRIPT Abstract Quantitative real-time-PCR methods are commonly used to monitor BCR-ABL transcript levels in patients with chronic myelogenous leukemia. However, standard techniques involve separate measurements of target and reference DNAs, require standard curves, and are
RI PT
susceptible to PCR inhibition. An optimized duplex droplet digital (dd)PCR should provide absolute quantification without the need for standard curves. The combination of high sensitivity and low background is particularly important for reliable monitoring of minimal residual disease. We report here primer probe set testing ad step by step optimization of a
SC
duplex ddPCR for BCR-ABL/ABL. The optimization of ddPCR parameters increased ABL and BCR-ABL fluorescence signals by 2 and 5 fold, respectively, and enhanced the
M AN U
resolution between positive and negative drops. The optimized procedure generates a background false positive rate of 5% of samples and reliably detects BCR-ABL/ABL down to 1/100000 (CV<10%), with a single BCR-ABL copy being detected in 54% of reactions performed in duplicate from an MR5 sample (limit of detection = 1). Assay of duplicates
TE D
resulted in a detection rate of 100% and 92% for MR4 and MR4.5, respectively. Detection of MR4.5 was increased to 100% by analyzing quadruplicates. Selection of an optimal primer/probe combination and stepwise optimization of the ddPCR conditions has yielded a
AC C
EP
robust low background duplex ddPCR procedure for BCR-ABL/ABL.
2
ACCEPTED MANUSCRIPT Introduction Digital droplet PCR (ddPCR) offers a means of quantifying low-level mutations within a high background of wild-type sequences that, in contrast to real time PCR, is relatively robust to PCR inhibition [1–3]. The qualitative positive/negative readout of single droplets largely
RI PT
excludes variation arising from differences in amplification efficiency and obviates the need for standard curves. Digital PCR is characterized by a high degree of sensitivity and precision [1] that make it particularly suitable for detection of BCR-ABL transcripts as
SC
evidence for the presence of residual malignant cells in minimal residual disease (MRD) diagnostics during the treatment phase of chronic myeloid leukemia (CML). In the recent past
M AN U
we compared the performance of duplex ddPCR for BCR-ABL/ABL with that of real time PCR using the Europe Against Cancer (EAC) primer/probe set on patient samples according to a published method [4] standardized to the international scale using a laboratory specific conversion factor [5, 6]. Although the sensitivity of this primer set is high, there is typically a background in ddPCR of one to two BCR-ABL positive droplets in up to 30% of negative
TE D
controls. This means that clinical samples need to generate at least three positive droplets (corresponding ideally to three copies of the specific target) to be reliably scored positive [5, 6]. In contrast, the false positive rate (FPR = 1 - Specificity = false positives/(false positives +
EP
true negatives) of ABL PCR was consistently around 2% of water controls, allowing contamination to be effectively ruled out as a cause of the background signal. Our aim was to
AC C
design a duplex BCR-ABL/ABL primer/probe system and PCR conditions that combine high sensitivity with a false positive rate not exceeding 5% in negative control reactions containing no cDNA or wild-type (BCR-ABL negative) cDNA (Supplementary Figure S1).
Materials and Methods
Samples
3
ACCEPTED MANUSCRIPT K-562 cells (Leibnitz-Institut Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany, DSMZ no.: ACC 10), originally derived from a CML patient in blast crisis, were used as a source of BCR-ABL b3a2 (also known as e14/a2) transcript for positive control experiments [7]. The BV-173 cell line (DSMZ no.: ACC20), originally derived
RI PT
from a 45-year–old male with CML in blast crisis was used as a source of BCR-ABL b2/a2 (e13/a2) transcript for positive control experiments [8]. HEK 293T (human embryonic kidney) cells (DSMZ no.: ACC 305) and mononuclear cells from umbilical cord blood (UCB) or peripheral blood were used for BCR-ABL negative control experiments and dilution of CML
SC
patient cDNA for sensitivity experiments. Total RNA was extracted in the QIAcube using QIAamp RNeasy Mini QIAcube Kit (240), QIAGEN Germany, Hilden, 74116) according to the
M AN U
manufacturer’s instructions but including an additional DNase digest step (RNase-Free DNase Set (50), QIAGEN, 79254) for removal of genomic DNA. The RNA concentration and purity was measured spectrophotometrically. Reverse transcription was performed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems Darmstadt, Germany,
EP
Primer/Probe Design
TE D
4368813) according to the manufacturer’s instructions.
Human mRNA sequences for BCR (reference sequence: NM_004327.3) and ABL
AC C
(M14752.1) were downloaded from the nucleotide database NCBI (National Center for Biotechnology Information, date of last access: 03. April 2018) https://www.ncbi.nlm.nih.gov/nuccore and cross referenced with BCR-ABL real time-PCR standard pME2 including BCR-ABL translocation sequence (e13/a2 and e14/a2 transcript forms). After testing the background signals generated by three different BCR-ABL1 primer/probe combinations the choice fell on the primer system published 2014 by the group of Jennings [9] with a FPR of 2% (2/92 replicates). To establish a duplex BCR-ABL/ABL PCR, a total of 10 different ABL primers and probes for exons 4, 8, 10, and 11 were tested for optimal annealing temperature and background signals. An ABL exon 10 primer/probe 4
ACCEPTED MANUSCRIPT combination generating an amplicon length of 106bp was chosen on the basis of a low FPR and clear separation of the ABL positive droplets.
Primers and Probes for BCR-ABL PCR (IDT - Integrated DNA Technologies London, UK)
B-A_for (BCR exon 13 (b2)) 5’-CATTCCGCTGACCATCAATA-3’;
SC
B-A_rev (ABL exon 3) 5’-ACACCATTCCCCATTGTGAT-3’;
RI PT
Amplicon length b2a2 (e13/a2) transcript: 247bp; b3a2 (e14/a2) transcript: 330bp
B-A_probe (ABL exon 2) 5’-FAM-CCCTTCAGCGGCCAGTAGCATCTGA-MGBEclipse or
M AN U
MGBNFQ-3’.
Note that B-A for does not take account of a polymorphism at the penultimate base. Although the Jennings primer is suitable for K562 cDNA, reliable analysis of patient material requires a
TE D
Y at this position (5’-CATTCCGCTGACCATCAAYA-3’).
Primers and Probes for ABL PCR (IDT)
EP
Amplicon length ABL: 106bp
ABL_for (ABL exon 10) 5’-GGAAAAGGAGCTGGGGAAAC-3’;
AC C
ABL_rev (ABL exon 10) 5’-GTGCTCTGCAGCTCTCCTG-3’; ABL_probe (ABL exon 10) 5’-HEX-AGCTGCCCACCAAGACGAGGACCTC-ZENIBFQ-3’ (Probes were purified with HPLC and primers with standard desalting.)
ddPCR
Optimized reactions were performed in 20µL duplex ddPCR reaction mix consisting of 1x ddPCR Supermix for Probes (Bio-Rad München, Germany), primers B-A_for, B-A_rev, 5
ACCEPTED MANUSCRIPT ABL_for and ABL_rev (fc 1360nM), probes B-A_probe, ABL_probe (fc 450nM) and 5µL template cDNA (500ng reverse transcribed RNA). After sealing and centrifugation (1min/86xg), the 96well plates were transferred to the QX200 Droplet Generator (Bio-Rad) to generate around 20000 droplets per well. The PCR was performed in a Thermocycler T100
RI PT
(Bio-Rad) using the following protocol: 95 °C/10min , 70 cycles of denaturation 94 °C/30sec and annealing/extension 58 °C/96sec (ramp rate sett ings 2.5 °C; heated lid set to 105 °C; sample volume 40µL) finishing with one cycle 98 °C/ 10min and holding at 12 °C. Positive and negative droplets were counted using the QX200 Droplet Reader (Bio-Rad) and
SC
analyzed by Poisson statistics using the Software QuantaSoft 1.7.4.0917. Validation of the BCR-ABL/ABL dPCR assay was performed according the dMIQE (Minimum Information for
M AN U
Publication of Quantitative digital PCR Experiments) [3]. The dMIQE checklist is shown in Supplemental Table S1.
BCR-ABL levels were expressed as BCR-ABL/ABL ratios. Since the ABL exon 10 primers bind additionally to BCR-ABL cDNA, effectively increasing the ABL signal, the calculated
TE D
BCR-ABL/ABL ratio tends to be underestimated at very high BCR-ABL copy numbers [4].
EP
Results
AC C
Annealing temperature and background prior to PCR optimization
The optimum annealing temperature for the BCR-ABL/ABL duplex ddPCR was determined by gradient PCR over an annealing/extension temperature range from 55.1 °C to 63.5 °C (Bio-Rad manufacturer´s note). Optimum results were obtained at 58.3 °C (Figure 1A and 1B). The BCR-ABL PCR generated false positive rates of 2% using water controls (singleplex BCR-ABL PCR: 2/92; duplex BCR-ABL PCR: 1/64). The corresponding false positive rate for BCR-ABL using wild-type cDNA controls was around 5% (5/92 using HEK 293T cells and 2/37 using healthy donor). An annealing temperature of 58 °C was chosen for further experiments. 6
ACCEPTED MANUSCRIPT
Choice of quenchers
To improve the resolution in fluorescence between positive and negative droplets, the probes
RI PT
specific for each of the duplex PCR products (B-A1_FAM-MGBNFQ probe (BCR-ABL PCR) and ABL_HEX-BHQ probe (ABL-PCR) were tested against newly available and stronger quencher modifications (B-A1_probe FAM-ZEN-IOWA Black FQ; ABL_probe HEX-ZENIOWA Black FQ; Figure 2). The ZEN-IOWA Black FQ quencher appeared to be stronger than
SC
MGBNFQ, but the fluorescence signals of both the positive and the negative drops were suppressed similarly. There was therefore no improvement in the BCR-ABL PCR droplet
M AN U
separation using the new quencher (Figure 2A). Furthermore an increased frequency of background signals was detected in water controls when using the B-A1_probe FAM-ZENIOWA Black FQ (21%, 20/94) compared to B-A1_FAM-MGB probe (4%, 4/94). In contrast, the ZEN-IOWA Black FQ probe for ABL resulted in an improved separation
TE D
compared to that achieved with the ABL BHQ probe used to date (Figure 2B). All further experiments made use of the BCR-ABL MGB probe and the ABL ZEN-IOWA
EP
Black FQ probe.
AC C
Optimizing the PCR conditions
To further improve the resolution of positive and negative droplets in the BCR-ABL ddPCR, the following PCR conditions were tested first singly and then in combination using K562 cDNA as target. (To present more rain a high concentration of K562 cDNA was used as target to maximize potential PCR inhibition effects.): Ramp rate [°C/sec]: >3; 3; 2; 1.5; 1 (2.5); Number of cycles [cycle x-fold]: 40; 50; 60; 70; Annealing/Extension times [min]: 0.5; 1; 1.5; 2; 2.5; Denaturing time [sec]: 30; 45; 60 (Conditions used in initial testing are underlined).
7
ACCEPTED MANUSCRIPT The results are shown in Figure 3 and identify the following conditions to be optimal: Ramp rate: 1 °C/sec, resulting in a 30% increase in fluo rescence in BCR-ABL ddPCR (1.5 °C/sec led to the best separation in ABL ddPCR); Cycle number: 70, resulting in a 100% increase in fluorescence in BCR-ABL ddPCR and in improved separation. Annealing/extension time: 2.5
time: 30 sec, longer times resulted in no further improvement.
RI PT
mins, 50% increase in fluorescence in BCR-ABL ddPCR, improved separation. Denaturation
It thus proved possible to further optimize three of the four parameters separately. These
SC
were then tested in combination (Figure 4). Interestingly, conditions in which the shorter annealing times were used failed to generate double positive droplets, even when the
M AN U
resolution between positive and negative droplets in 1D plots was distinct. This suggests competitive inhibition, most probably of the BCR-ABL PCR by the ABL PCR. This would have the effect of shifting the BCR-ABL/ABL ratio to indicate lower levels of residual disease. However, the use of longer annealing/extension times overcomes this problem (Figure 4B, 4-
TE D
8), the best separation being seen under the conditions shown in Figure 4B, 8. Under these conditions, the maximum BCR-ABL fluorescence signal was nearly three-fold that seen before optimization and the BCR-ABL/ABL copy number was increased significantly from
EP
3360/21080 copies to 10780/21000 copies (Figure 4B, 1 and 4B, 8 respectively). Although the maximum level of ABL fluorescence was essentially unchanged, the increased cycle
AC C
number of conditions 2-8 effectively increased the fluorescence from the weaker droplets. This reduces the “rain” effect caused by droplets that do not clearly belong to the positive or negative cluster [10, 11]. The rain is generated by droplets of intermediate fluorescence intensity resulting from low-yield PCR amplifications that in turn result from a suboptimal PCR efficiency or a premature termination. The following optimum conditions were used in further experiments: ramp rate 1 °C/s, cycle number 70, and annealing/extension time 2.5 min.
Optimizing primer/probe concentration 8
ACCEPTED MANUSCRIPT To further reduce the rain effect in the BCR-ABL reaction, a range of primer/probe concentrations were tested under the previously optimized PCR conditions (Figure 5). Increasing the primer to a final concentration (fc) of 1.36µM led to minor improvement in
RI PT
BCR-ABL fluorescence (Figure 5). However, increasing the probe concentration from 250nM to 450nM resulted in a doubling of the fluorescence signal from both BCR-ABL and ABL, with an associated reduction in the rain effect and improvement of the separation (Figure 5). This suggests that the concentration of the probe is limiting.
SC
For further experiments a primer concentration of 1.36µM and a probe concentration of
M AN U
450nM was chosen.
Background signals generated by the optimized assay in no-template control (NTC) and wildtype cDNA controls
Having performed numerous modifications to the PCR conditions during the optimization, it
TE D
was necessary to re-test the background signals using the new conditions. The FPR was calculated as false positive samples/(false positive + true negative samples). Although the false positive rate was marginal in NTCs at 3% (5/188) it was very high for wild-type controls,
AC C
(Table 1).
EP
with 64% to 100% positivity from healthy donor and a BCR-ABL negative cell line cDNA
To identify the parameter(s) responsible for the decreased specificity (specificity = true negatives/(true negatives + false positives)), a series of reactions were performed using a high level of wild-type target (HEK 293T cDNA, 70.000 ABL equivalents, Figure 6). As expected, no BCR-ABL background was seen under the original conditions (Figure 6A). Increasing the primer/probe concentration alone or in combination with a decreased ramp rate also resulted in no detectable background. However, several positive BCR-ABL signals were generated at the longer annealing/extension times. The increase from 40 to 70 PCR 9
ACCEPTED MANUSCRIPT cycles resulted in only one BCR-ABL positive droplet from six reaction wells but in combination with a lower ramp rate (which effectively increases extension time) resulted in background noise around the BCR-ABL negative fraction. This suggests that the BCR-ABL background signals are favored by longer amplification times (Figure 6B), so rapid ramping
RI PT
and shorter annealing and extension times “focus” the reaction on the specific product. This was confirmed in a separate set of reactions employing a rapid ramp rate (2.5 °C/s) and shorter annealing/extension time (1min), which eliminated the unspecific signals in all wild-
SC
type controls (Figure 6B).
A ramp rate of 2.5 °C/s and an annealing/extension time of 1min were chosen for further
M AN U
work.
BCR-ABL FPR and Copy Number (CN) controlled by time of annealing/extension
TE D
A final concern was that the reduction in annealing/extension time necessary to eliminate the non-specific background may have the undesired effect of increasing the competition between the ABL and BCR-ABL amplification reactions as noted above (Figure 4), once
EP
again eliminating the double positive droplets. It was therefore important to determine whether an optimal annealing/extension time maintains both sensitivity and specificity.
AC C
A comparison of sensitivity (BCR-ABL and ABL copy numbers) and specificity at annealing times between 1 and 2.5 min is shown in Table 2. An annealing/extension time of 1.6 mins maintained a specificity of 98% (FPR 2%), while detecting BCR-ABL copy numbers close to those seen using real time-PCR after conversion to the international scale. The shorter annealing/extension time resulted in absolute specificity but a very low rate of detection, whereas longer times reduced specificity to unacceptable levels. After balancing of specificity and sensitivity an annealing/extension time of 1.6min was chosen for the final, fully optimized assay. 10
ACCEPTED MANUSCRIPT
Final assay, Linearity, and Precision
A 2D plot of fluorescence signals generated using the final PCR conditions compared to
RI PT
those used at the outset (Figure 7A) is shown in Figure 7B. The sequential modifications result in a 5-fold increase in BCR-ABL fluorescence, a doubling of ABL fluorescence and the resolution of a population of double-positive droplets indicating little or no competition
SC
between the targets.
Six log10 dilutions of BCR-ABL positive K562 cDNA were prepared to generate ABL copy
M AN U
numbers from 100000 – 1 (Figure 7C). High linearity for ABL was seen over the complete range. BCR-ABL detection was also highly linear at copy numbers below 4000 (dilutions 2 6). However the most concentrated BCR-ABL dilution (40000 copies) generated a detected copy number of only 18,000 copies, suggesting PCR inhibition at high target concentration. With quantitative detection of between 1 and 4000 BCR-ABL copies in patient sample
TE D
replicates containing 40000 ABL copies, linearity would be given between 10% and 0.0025% BCR-ABL/ABL corresponding to MR4.5. Although standard deviation of the BCR-ABL/ABL ratios was increased in samples 5 and 6 that contained limiting amounts of target (31 and
EP
30% BCR-ABL/ABL), those in dilutions 1 to 4 were between 1% to 4%. The median intraassay coefficient of variation of all six dilutions was below 10% (ABL CV = 8%; BCR-ABL CV
AC C
= 6%; BCR-ABL/ABL CV = 9%), indicating that the assay provides reliable and accurate data with high precision.
As expected from the Poisson distribution, the variation in BCR-ABL copy numbers increases markedly as the copy number drops below 100, equivalent to the range MR3 to MR5.5. Accordingly, the analysis of patient samples (n=48) revealed a BCR-ABL inter-assay CV increasing from 22% for MR3 samples to 288% for MR5.5 samples (Table 3).
Limit of Blank (LOB), False-Positive-Rate, and Specificity of final assay 11
ACCEPTED MANUSCRIPT The FPR required for LOB determination was analyzed in 92 replicates of no template controls. The observed BCR-ABL FPR of 1% (1/92) is equivalent to a LOB of 0 for >95% of the replicates.
RI PT
To check the background noise in samples from healthy donors, cDNA from two independent white blood cell (buffy coat) preparations were assayed as follows: 188 aliquots of cell lysates (each 1x 10E7 cells) were extracted (final 40µL RNA eluate, mean 224ng/µL RNA)
SC
and reverse transcribed (each 4µg RNA to final 40µL cDNA). The cDNAs were then pooled and diluted (factor 1.6). 174 technical replicates of 5µL cDNA were analyzed (mean 63000
M AN U
ABL copies per replicate). The FPR (presence of a single positive drop in a ddPCR reaction) for BCR-ABL in wild-type cDNA was 4% (3/84) and 7% (6/90) resulting in an overall FPR of 5% (9/174) and a specificity of 95% (“exact” Clopper-Pearson 95% confidence interval: 90%
TE D
to 98%).
Limit of Detection (LOD) and Sensitivity of final assay
EP
LOD and sensitivity were measured in serial dilutions of CML patient cDNA (9.5% BCRABL/ABL (IS) as determined by real time-PCR standardized to the international scale) in
AC C
healthy donor cDNA to simulate MR3 down to MR5.5 (Table 4). Three batches of 16 replicates for each MRD dilution were analyzed at three different time points. The ddPCR method yielded MR levels identical to those expected, confirming the comparability of the two assay platforms. The sensitivity is defined as detection rate, calculated as true positive samples/(true positive + false negative samples). Analyses in duplicate resulted in BCR-ABL detection rates of 100% and 92% from MR4 (0.009% BCR-ABL/ABL) and MR4.5 (0.003%) samples respectively. Analysis of four wells resulted in 100% and 75% detection of MR4.5 (0.003%) and MR5 (0.001%), respectively.
12
ACCEPTED MANUSCRIPT The LOD was one copy of BCR-ABL per set of duplicates or per four well analyses (Table 4). The duplex ddPCR assay can therefore detect one BCR-ABL transcript per 120,000 wild-
Comparison of different internal control sequences
RI PT
type copies per replicate with high specificity.
Given the evidence of ABL product interference with the BCR-ABL reaction under suboptimal
SC
conditions seen in the 2D plots in Fig. 4B, the incidence of double positive drops was
assessed using alternative internal controls. Five different master mixes (MM) were used
M AN U
under pre- and post-optimization ddPCR conditions: (MM1) employs the EAC BCR-ABL/ABL primers and probes used in the international standard real-time PCR reaction, (MM2 to MM5) employ the Jennings BCR-ABL primers/probes together with EAC GUS primers (MM2), the exon 10 ABL probe to the opposite strand of the published assay (MM3), exon 8 ABL primer/probes (MM4), and the assay published in this paper (MM5) for comparison. All
dilution (Table 5).
TE D
master mixes were tested in the same run and with the same BCR-ABL positive K562 cDNA
The EAC primer probe set (MM1) was highly sensitive even under the pre-optimization
EP
conditions that, however, generate a high background as described in the introduction. Each of the primer/probe sets MM2 to MM5 generated low frequencies of double positives under
AC C
pre-optimization conditions that were markedly increased by optimization. Furthermore, the frequency of BCR-ABL copies detected under optimized conditions did not vary with the internal control sequence, demonstrating that the optimized procedure is robust in this respect.
Discussion
13
ACCEPTED MANUSCRIPT The key advantage of ddPCR over quantitative real-time-PCR lies in the detection of endproducts generated by single target molecules. A drop containing amplified product is scored positive, regardless of the efficiency of the reaction, making the procedure relatively resistant to the effects of PCR inhibition [1, 2]. Given a suitable primer set, accurate and reliable
RI PT
BCR-ABL monitoring should be achievable without the need for standard curves. Furthermore, analysis of both BCR-ABL and ABL in the same (duplex) PCR reaction avoids inaccuracies introduced by pipetting variations.
The relative insensitivity of the digital PCR end-point readout to PCR inhibition may be
SC
particularly relevant to treatment decisions that depend on accurate determination of MRD. The discontinuation of TKI therapy in CML patients who have achieved stable deep
M AN U
molecular remission is one example of this. Here, partial inhibition of a real time-PCR reaction would, by delaying the accumulation of reaction products, deliver a read out below the level of MRD actually present and may lead to premature discontinuation. Since partial inhibition and delay of amplification in a ddPCR reaction has little or no effect on the endpoint
decision.
TE D
read out, ddPCR should in this case provide a more reliable basis for informing the therapy
A comparison of three primer/probe sets identified one previously published by the group of
EP
Jennings to be the most specific and suitable. Starting with this, a duplex PCR was designed incorporating the wild type control gene ABL. The resolution between positive and negative
AC C
droplets was initially poor, mostly due to the relatively low fluorescence of positive droplets. However, three strategies were identified to improve this separation. The clearest advantage came from increasing the concentration of fluorescent probe, which is typically kept low due to cost considerations. Here, a 2- to 3-fold increase in probe concentration was found to generate a comparable increase in fluorescence intensity. Notably, increasing the cycle number up to 70 proved to be an effective way of boosting weak signals, reducing the rain effect, and thereby improving the separation between positive and negative samples. Finally, the choice of Quencher may also have a minor effect on separation due to probe-specific differences in quenching efficiencies. In this way, it proved possible to reduce the rain effect 14
ACCEPTED MANUSCRIPT and generate the double-positive fraction that is predicted on the basis of stochastic segregation of target between droplets. Changing the PCR reaction conditions for a given set of primers often has reciprocal effects on specificity and sensitivity, so that optimization comes down to finding the best balance
RI PT
between the two. Accordingly, low ramp rates and extended annealing/extension times tended to increase the frequency of false positive BCR-ABL signals in the form of weakly positive drops generated from wild-type (non–BCR-ABL) target cDNA. On the other hand, increasing the ramp rate (to 2.5 °C/s) and reducing the annealing/extension time (to 1 min)
SC
eliminated the non-specific signals but reduced the sensitivity, leading to underestimation of BCR-ABL in positive controls. Fortunately, it proved possible to combine high specificity
M AN U
(98%) with a degree of sensitivity that generated BCR-ABL/ABL ratios comparable to those of the current (IS) real time-PCR standard by adjusting the annealing/extension time to 1.6 min. This dependence of the BCR-ABL/ABL ratio readout on the annealing time suggests that it may be beneficial to employ PCR- and laboratory-specific conversion factors for
TE D
ddPCR in the same way that they are currently used for real time-PCR. Using clinical CML samples, the optimized ddPCR assay detected one BCR-ABL copy in 54% of reactions (performed in duplicate) from an MR5 sample (≤0.0010% BCR-ABL/ABL),
EP
whereas under the same conditions an MR4.5 (≤0.0032% to ≥0.0010% BCR-ABL/ABL) sample detected three copies in 92% of the assays. This is close to the frequency expected
AC C
from the Poisson distribution, which predicts 95% of samples taken from a solution with three targets per unit volume to contain at least one copy of the target [12]. In summary, we report the development of a low-background, duplex ddPCR BCR-ABL/ABL assay with specificity in blank controls of 98%. The cross-reaction with wild-type cDNA does not exceed 5%, making the use of a cut-off unnecessary. Furthermore, we optimized the conditions of the assay so that positive and negative droplets are well-separated and the rain effect is either markedly reduced (BCR-ABL) or eliminated (ABL). The assay conditions have been adjusted to allow BCR-ABL and ABL PCR reactions to proceed with similar efficiency in a single drop, evidenced by a well-separated fraction of double-positive droplets. The assay 15
ACCEPTED MANUSCRIPT provides accurate data with high precision and high linearity between 1 and 4000 copies of BCR-ABL and between 1 and 100000 copies of ABL. The linearity and ratio were reduced at BCR-ABL copy numbers above 4,000, but this is not relevant for samples with molecular response between MR2 and MR5. The BCR-ABL amplification described here is more
RI PT
sensitive to changes in PCR conditions than is the ABL reaction, but stepwise optimization resulted in a detection rate of 92% for MR4.5, close to the theoretical ideal of 95%. Clearly, sensitivity could be increased further by extending the analysis to more replicates.
The observations reported here should be of general use in optimizing ddPCR protocols,
SC
especially in the diagnostic setting in which primer choice is necessarily limited and even the
AC C
EP
TE D
M AN U
best available primer sets can require careful optimization of reaction conditions.
16
ACCEPTED MANUSCRIPT
References Yang R, Paparini A, Monis P, Ryan U: Comparison of next-generation droplet digital PCR (ddPCR) with quantitative PCR (qPCR) for enumeration of Cryptosporidium oocysts in faecal samples. Int J Parasitol 2014:1105–1113. 2 Blaya J, Lloret E, Santisima-Trinidad AB, Ros M, Pascual JA: Molecular methods (digital PCR and real-time PCR) for the quantification of low copy DNA of Phytophthora nicotianae in environmental samples. Pest Manag Sci 2016:747–753. 3 Huggett JF, Foy CA, Benes V, Emslie K, Garson JA, Haynes R, Hellemans J, Kubista M, Mueller RD, Nolan T, Pfaffl MW, Shipley GL, Vandesompele J, Wittwer CT, Bustin SA: The digital MIQE guidelines: Minimum Information for Publication of Quantitative Digital PCR Experiments. Clin Chem 2013:892–902. 4 Gabert J, Beillard E, van der Velden, V H J, Bi W, Grimwade D, Pallisgaard N, Barbany G, Cazzaniga G, Cayuela JM, Cave H, Pane F, Aerts JL, Micheli D de, Thirion X, Pradel V, Gonzalez M, Viehmann S, Malec M, Saglio G, van Dongen, J J M: Standardization and quality control studies of 'real-time' quantitative reverse transcriptase polymerase chain reaction of fusion gene transcripts for residual disease detection in leukemia - a Europe Against Cancer program. Leukemia 2003:2318–2357. 5 Franke G, Maier J, Wildenberger K, Cross M, Frank O, Giles F, Hochhaus A, Dietz CT, Müller MC, Niederwieser D, Lange T: Quantification of BCR-ABL with Digital PCR Results in a Significantly Lower Rate of Deep Molecular Responses When Compared to RT-qPCR in CML Patients Treated in the ENEST1st Trial. Blood 2015:135. 6 Maier J, Franke G, Schubert K, Wildenberger K, Cross M, Niederwieser D, Lange T: A Comparison of Droplet Digital PCR and Quantitative RT-PCR for Low Level BCR-ABL in CML Patients with Molecular Responses. Blood 2014:1792. 7 Lozzio CB, Lozzio BB: Human chronic myelogenous leukemia cell-line with positive Philadelphia chromosome. Blood 1975:321–334. 8 Pegoraro L, Matera L, Ritz J, Levis A, Palumbo A, Biagini G: Establishment of a Ph1positive human cell line (BV173). J Natl Cancer Inst 1983:447–453. 9 Jennings LJ, George D, Czech J, Yu M, Joseph L: Detection and quantification of BCRABL1 fusion transcripts by droplet digital PCR. J Mol Diagn 2014:174–179. 10 Jacobs BK, Goetghebeur E, Vandesompele J, Ganck A de, Nijs N, Beckers A, Papazova N, Roosens NH, Clement L: Model-Based Classification for Digital PCR: Your Umbrella for Rain. Anal Chem 2017:4461–4467. 11 Jones M, Williams J, Gartner K, Phillips R, Hurst J, Frater J: Low copy target detection by Droplet Digital PCR through application of a novel open access bioinformatic pipeline, 'definetherain'. J Virol Methods 2014:46–53. 12 Cross NC, White HE, Colomer D, Ehrencrona H, Foroni L, Gottardi E, Lange T, Lion T, Machova Polakova K, Dulucq S, Martinelli G, Oppliger Leibundgut E, Pallisgaard N, Barbany G, Sacha T, Talmaci R, Izzo B, Saglio G, Pane F, Muller MC, Hochhaus A: Laboratory recommendations for scoring deep molecular responses following treatment for chronic myeloid leukemia. Leukemia 2015:999–1003.
AC C
EP
TE D
M AN U
SC
RI PT
1
17
ACCEPTED MANUSCRIPT Figure Legends Figure 1. A: Optimal annealing temperature for BCR-ABL and ABL ddPCR in 1D Plot (A) and 2D Plot (B) at temperature optimum. Representative ddPCR results are shown for BCR-ABL positive K562 cDNA. Channel 1 amplitude showing BCR-ABL positive (blue) and
RI PT
negative (black) droplets, separated by the threshold (pink line). Channel 2 amplitude showing ABL positive (green) and negative (black) droplets, separated by the threshold (pink line). Yellow dashed lines separate different reaction wells. A temperature gradient from 63.5
SC
°C to 55.1 °C was used and annealing temperature of 58.3 °C (asterisk) was selected as
M AN U
optimal.
Figure 2. A: Comparison of BCR-ABL signal and background fluorescence levels obtained using MGBNFQ and Zen-IBFQ quenched probes (A) and comparison of ABL signal and background fluorescence levels obtained using BHQ and Zen-IBFQ
TE D
quenched probes (B). A: Results from MGBNFQ and Zen-IOWA black probe. B3a2 and b2a2 indicate reactions using the two respective BCR-ABL transcripts. Channel 1 amplitude (y-axis) showing fluorescence units of BCR-ABL positive (blue) and negative (black) droplets, separated by the threshold (pink line). B: Signals obtained from the BHQ quenched probe
EP
are shown on the left and those from the Zen-IOWA black quenched probe on the right.
AC C
Channel 2 amplitude (y-axis) showing ABL positive (green) and negative (black) droplets, separated by the thresholds (pink lines). Yellow dashed lines separate different reaction wells. MGBNFQ (Minor Groove Binder Nonfluorescent Quencher); BHQ (Black Hole Quencher); ZEN-IBFQ (ZEN-IOWA Black FQ, Dark Quencher Modification (IDT).
Figure 3. Optimal ddPCR conditions according to ramp rate (A), cycle number (B), annealing/extension (C), and denaturation (D). Ramp rate (A), cycle number (B), annealing/extension time (C), and denaturation time (D) were tested for both the BCR-ABL (blue) and ABL (green) reactions. Pink line separates amplification positive from negative 18
ACCEPTED MANUSCRIPT droplets and yellow dashed lines separate different runs. Initial conditions are underlined. The condition identified as optimal in each case is denoted with an asterisk.
Figure 4. Optimization of ddPCR conditions ramp rate, cycle number, and
RI PT
annealing/extension time. A: Eight duplex BCR-ABL and ABL ddPCR reactions (denoted 1 to 8, separated by yellow dashed lines) were performed using variable ramp rates, cycle numbers, and annealing/extension times as shown in the table (gray shaded sections). The
SC
fluorescent read-outs are shown as one-dimensional plots directly below the respective condition. B: Two dimensional plots of the results of reactions (1-8) shown in A. Initial
M AN U
conditions are underlined (1). Condition 8 was determined to provide the optimal signal to noise and is marked with an asterisk. Pink line separates amplification positive from negative droplets.
TE D
Figure 5. Optimization of primer/probe concentrations. A set of eight duplex BCR-ABL and ABL ddPCR reactions were performed using variable concentrations of primers and probes (gray shaded sections at the top). Yellow dashed lines separate different reaction wells. The results are shown in a one-dimensional plot directly below the respective
EP
condition. The condition chosen as optimal is marked with an asterisk. Pink line separates
AC C
amplification positive from negative droplets.
Figure 6. Background cause analysis using different combinations of ddPCR conditions and primers and probe concentrations (A) and reduction of background (B). A: A set of six reaction conditions was chosen to determine the effects of primer/probe concentration, cycle number, ramp rate, and annealing time on the background frequency of positive droplets. Initial conditions are shown in the table in red and changed conditions in blue numbers above the plots. Each reaction condition was used to amplify NTC (n=2), positive control K562 (n=2), and wt HEK 293T (n=6) samples. Increased background was 19
ACCEPTED MANUSCRIPT associated with a lower ramp rate (2.5 °C/s vs 1 °C /s) and a longer annealing/extension time (2.5 minute vs 1 minute). B: The effects of ramp rate and annealing/extension time were confirmed in an independent set of reactions amplifying NTC (n=2), positive control K562 (n=2), and wt HEK 293T (n=92). Pink line separates amplification positive from negative
RI PT
droplets and yellow dashed lines separate different reaction wells.
Figure 7. Digital PCR assay M-BCR-ABL/ABL before (A) and after optimization (B) and
SC
linearity of optimized ddPCR duplex Assay (C). 2D Plot. A: ramp rate 2.5 °C/s, cycle number 40x, annealing/extension 1min 58 °C, Primers fc. 0.91µM, Probes fc. 250nM; 2D
M AN U
Plot. B: Ramp rate 2.5 °C/s, cycle number 70x, annealing/ex tension 1.6 min 58 °C, Primers fc. 1.36µM, Probes fc. 450nM; BCR-ABL positive droplets (blue); ABL positive droplets green; negative for BCR-ABL and ABL black; doubled positive (red) were separated by pink lines. C: Copy number of BCR-ABL and ABL are shown on the primary axis (left y-axis) and
AC C
EP
TE D
the ratio is shown on the secondary axis (right y-axis)
20
ACCEPTED MANUSCRIPT Table 1. False-positive rate (FPR) of NTC and wild-type with changed conditions.
BCR-ABL FPR [%] (n)
BCR-ABL/10000 ABL [copies]
NTC
3 (5/188)
-
HEK 293T
100 (92/92)
2.0
UCB
78 (70/90)
0.8
Healthy donor
64 (59/92)
0.3
Median ABL [copies]
RI PT
Template
0
M AN U
SC
82000
21000
48000
Conditions: ramp rate 1 °C/sec; cycles 70; annealin g/extension 2.5min; Primers fc. 1.36µM; Probes fc. 450nM; NTC (no template control); HEK (human embryonic kidneys); UCB
AC C
EP
TE D
(umbilical cord blood).
21
ACCEPTED MANUSCRIPT Table 2. BCR-ABL FPR and copy number controlled by time of annealing/extension. Ann./ext. [min]
Specificity/FPR (wt) [%]
BCR-ABL CN (1*)
ABL CN (*1)
% Ratio (*1)
ddPCR ddPCR ddPCR ddPCR ddPCR RT-PCR(IS)
1 1.6 1.7 1.8 2.5 1
100/0 98/2 93/7 86/14 0/100 100/0
44 707 918 1246 1649 827
76700 60383 11340 47360 54898 58808
0.1 1.2 2.1 2.6 3.0 1.4
RI PT
Assay
*1 - BCR-ABL positive sample; CN (copy number); FPR (False-Positive-Rate); wt (wild type);
SC
ddPCR (Droplet Digital PCR); RT-PCR (Real-Time PCR); IS (international scale) - for
standardization to IS BCR-ABL CN was multiplied with a laboratory-specific conversion
AC C
EP
TE D
M AN U
factor.
22
ACCEPTED MANUSCRIPT Table 3. Interassay Variation of MR3 to MR5.5 CML patient samples (n = 48).
%CV ABL (mean CN)
%CV BCRABL/ABL (mean)
MR3
22 (58.3)
9 (64600)
20 (0.09036)
MR4
53 (5.7)
8 (62175)
52 (0.00922)
MR4.5
82 (1.6)
8 (60493)
82 (0.00265)
MR5
143 (0.6)
8 (62641)
142 (0.00100)
MR5.5
288 (0.3)
9 (63053)
287 (0.00039)
SC
RI PT
Patient Sample
%CV BCR-ABL (mean CN)
n = 3x 16 replicates for each MRD level determined at three different time points; ddPCR
M AN U
(Droplet Digital PCR); MRD (molecular residual disease); CN (copy number); MR (molecular
AC C
EP
TE D
remission).
23
ACCEPTED MANUSCRIPT Table 4. Sensitivity of CML patient samples (n = 48)
2-well Analysis
4-well Analysis
% Sensitivity
BCR-ABL CN ABL CN
MR3
117
129201
100 (86 - 100) 233
MR4
11
124350
100 (86 - 100) 23
MR4.5
3
120987
92 (73 - 99)
7
MR5
1
125283
54 (33 - 74)
2
MR5.5
0,5
126106
21 (7 – 42)
1
M AN U
(*95% CI)
BCR-ABL CN ABL CN
% Sensitivity (*95% CI)
258402
100 (74 - 100)
248700
100 (74 - 100)
241973
100 (74 - 100)
SC
MRD level
RI PT
ddPCR
250565
75 (43 - 95)
252212
42 (15 - 72)
n = 3x 16 replicates for each MRD level determined at three different time points; *Confidence intervals for sensitivity are “exact” Clopper-Pearson confidence intervals; ddPCR (Droplet Digital PCR); MRD (molecular residual disease); CN (copy number); MR
AC C
EP
TE D
(molecular remission).
24
ACCEPTED MANUSCRIPT Table 5. Test of changed and unchanged PCR conditions within different assays.
Assay
Before/after optimization BCR-ABL
CG
BCR-ABL/CG
droplets
copies
copies
[%]
MM1
3559 / 3782
8930 / 9340
10410 / 11450
86 / 82
MM2
199 / 508
4990 / 5590
1584 / 3800
315 / 147
MM3
974 / 1707
4360 / 5770
12620 / 12280
35 / 47
MM4
195 / 1516
2990 / 5370
12670 / 11650
24 / 46
MM5
81 / 1343
2920 / 5410
12310 / 11840
24 / 46
AC C
EP
TE D
M AN U
SC
RI PT
Ch1+Ch2+ 20µl reaction
25
ACCEPTED MANUSCRIPT A
°C 63.5 62.9 61.8 60.2 *58.3 56.7 55.7 55.1
M AN U
SC
RI PT
BCR-ABL
EP AC C
B
TE D
ABL
BCR-ABL
ABL
ACCEPTED MANUSCRIPT
MGBNFQ
A
b2a2
b3a2
RI PT
b3a2
ZEN-IBFQ
B
AC C
ABL
ZEN-IBFQ
EP
BHQ
TE D
M AN U
SC
BCR-ABL
b2a2
ACCEPTED MANUSCRIPT
A
B
Ramp rate °C/s
>3
3
2
1*
1.5
Cycle number X-fold
40
50
60
SC
RI PT
BCR-ABL
Annealing/extension 1
1.5
2
AC C
BCR-ABL
0.5
ABL
2.5*
EP
min
TE D
C
M AN U
ABL
D
Denaturation
sec
30*
45
60
70*
ACCEPTED MANUSCRIPT A 1
2
3
4
5
6
7
8*
ramp rate [°C/s]
2.5
2.5
1.5
1
1.5
1.5
1.5
1*
cycle [x-fold]
40
70
70
70
70
70
70
70*
ann./ext. [min]
1
1
1
1
1.5
2
2.5
2.5*
RI PT
condition
M AN U
SC
BCR-ABL
ABL
AC C
2
EP
1
TE D
B
5
6
3
7
4
8*
0.45
0.91
1.36*
0.91
0.91
0.91
0.91 1.36*
Probes fc [nM]
250
250
250
150
250
350
450*
M AN U
SC
Primers fc [µM]
AC C
EP
TE D
BCR-ABL
ABL
RI PT
ACCEPTED MANUSCRIPT
450*
ACCEPTED MANUSCRIPT
A 2.5 40 1 1.36 450
2.5 40 1 0.91 250
1 40 1 1.36 450
B
TE D
1 70 2.5 1.36 450
AC C
EP
rampe rate [°C/s] cycle [x-fold] ann./ext. [min] Primers fc [µM] Probes fc [nM]
ABL
M AN U
ABL
BCR-ABL
2.5 70 1 1.36 450
SC
BCR-ABL
2.5 40 2.5 1.36 450
RI PT
rampe rate [°C/s] cycle [x-fold] ann./ext. [min] Primers fc [µM] Probes fc [nM]
2.5 70 1 1.36 450
1 70 1 1.36 450
ACCEPTED MANUSCRIPT
B
A
cycle number 40x annealing/extension time 1min Primers fc. 0.91µM Probes fc. 250nM
TE D
M AN U
SC
RI PT
cycle number 70x annealing/extension time 1.6min Primers fc. 1.36µM Probes fc. 450nM
C
Linearity
10000
R² = 0.9999
AC C
1000
1000 100
100 10
1 0
10
R² = 0.9962
1
2
3
4
5
6
ABL
85333
9643
982
99
10
1
BCR-ABL
18051
4076
471
44
5
1
21
42
48
44
53
81
BCR-ABL/ABL
Log10 Dilutions of K562 cDNA
1
% BCR-ABL/ABL
EP
Copies per Reaction
100000
S1525-1578(18)30071-0
DOI:
https://doi.org/10.1016/j.jmoldx.2018.08.012
Reference:
JMDI 742
To appear in:
The Journal of Molecular Diagnostics
Received Date: 26 February 2018 Revised Date:
5 July 2018
Accepted Date: 22 August 2018
Please cite this article as: Maier J, Lange T, Cross M, Wildenberger K, Niederwieser D, Franke G-N, Optimized digital droplet PCR for BCR-ABL, The Journal of Molecular Diagnostics (2018), doi: https:// doi.org/10.1016/j.jmoldx.2018.08.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Optimized digital droplet PCR for BCR-ABL Jacqueline Maier,* Thoralf Lange,*† Michael Cross,* Kathrin Wildenberger,* Dietger Niederwieser,* and Georg-Nikolaus Franke*
RI PT
From the Department of Haematology and Oncology,* Leipzig University, Leipzig; and the Department of Oncology and Haematology,† Asklepios Klinik Weissenfels, Weissenfels, Germany
Footnote: Portions of this work were presented in oral form at the Droplet Digital PCR Expert Meeting held May 30, 2017, in Frankfurt/Main, Germany.
SC
Running head: ddPCR signal intensity and background
M AN U
Disclosures: T.L. received fees and presentations of BMS and Novartis, research funding from Novartis. D.N. received research funding from Novartis. G-N.F. received fees and presentations of BMS and Novartis, research funding from Novartis. This work was supported by Novartis Pharma GmbH (Project HTAS-181, AN 72255619 to G-N.F. and D.N.).
AC C
EP
TE D
Corresponding author: Jacqueline Maier Leipzig University Department of Haematology and Oncology Johannisallee 32A 04103 Leipzig Email: [email protected]
1
ACCEPTED MANUSCRIPT Abstract Quantitative real-time-PCR methods are commonly used to monitor BCR-ABL transcript levels in patients with chronic myelogenous leukemia. However, standard techniques involve separate measurements of target and reference DNAs, require standard curves, and are
RI PT
susceptible to PCR inhibition. An optimized duplex droplet digital (dd)PCR should provide absolute quantification without the need for standard curves. The combination of high sensitivity and low background is particularly important for reliable monitoring of minimal residual disease. We report here primer probe set testing ad step by step optimization of a
SC
duplex ddPCR for BCR-ABL/ABL. The optimization of ddPCR parameters increased ABL and BCR-ABL fluorescence signals by 2 and 5 fold, respectively, and enhanced the
M AN U
resolution between positive and negative drops. The optimized procedure generates a background false positive rate of 5% of samples and reliably detects BCR-ABL/ABL down to 1/100000 (CV<10%), with a single BCR-ABL copy being detected in 54% of reactions performed in duplicate from an MR5 sample (limit of detection = 1). Assay of duplicates
TE D
resulted in a detection rate of 100% and 92% for MR4 and MR4.5, respectively. Detection of MR4.5 was increased to 100% by analyzing quadruplicates. Selection of an optimal primer/probe combination and stepwise optimization of the ddPCR conditions has yielded a
AC C
EP
robust low background duplex ddPCR procedure for BCR-ABL/ABL.
2
ACCEPTED MANUSCRIPT Introduction Digital droplet PCR (ddPCR) offers a means of quantifying low-level mutations within a high background of wild-type sequences that, in contrast to real time PCR, is relatively robust to PCR inhibition [1–3]. The qualitative positive/negative readout of single droplets largely
RI PT
excludes variation arising from differences in amplification efficiency and obviates the need for standard curves. Digital PCR is characterized by a high degree of sensitivity and precision [1] that make it particularly suitable for detection of BCR-ABL transcripts as
SC
evidence for the presence of residual malignant cells in minimal residual disease (MRD) diagnostics during the treatment phase of chronic myeloid leukemia (CML). In the recent past
M AN U
we compared the performance of duplex ddPCR for BCR-ABL/ABL with that of real time PCR using the Europe Against Cancer (EAC) primer/probe set on patient samples according to a published method [4] standardized to the international scale using a laboratory specific conversion factor [5, 6]. Although the sensitivity of this primer set is high, there is typically a background in ddPCR of one to two BCR-ABL positive droplets in up to 30% of negative
TE D
controls. This means that clinical samples need to generate at least three positive droplets (corresponding ideally to three copies of the specific target) to be reliably scored positive [5, 6]. In contrast, the false positive rate (FPR = 1 - Specificity = false positives/(false positives +
EP
true negatives) of ABL PCR was consistently around 2% of water controls, allowing contamination to be effectively ruled out as a cause of the background signal. Our aim was to
AC C
design a duplex BCR-ABL/ABL primer/probe system and PCR conditions that combine high sensitivity with a false positive rate not exceeding 5% in negative control reactions containing no cDNA or wild-type (BCR-ABL negative) cDNA (Supplementary Figure S1).
Materials and Methods
Samples
3
ACCEPTED MANUSCRIPT K-562 cells (Leibnitz-Institut Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany, DSMZ no.: ACC 10), originally derived from a CML patient in blast crisis, were used as a source of BCR-ABL b3a2 (also known as e14/a2) transcript for positive control experiments [7]. The BV-173 cell line (DSMZ no.: ACC20), originally derived
RI PT
from a 45-year–old male with CML in blast crisis was used as a source of BCR-ABL b2/a2 (e13/a2) transcript for positive control experiments [8]. HEK 293T (human embryonic kidney) cells (DSMZ no.: ACC 305) and mononuclear cells from umbilical cord blood (UCB) or peripheral blood were used for BCR-ABL negative control experiments and dilution of CML
SC
patient cDNA for sensitivity experiments. Total RNA was extracted in the QIAcube using QIAamp RNeasy Mini QIAcube Kit (240), QIAGEN Germany, Hilden, 74116) according to the
M AN U
manufacturer’s instructions but including an additional DNase digest step (RNase-Free DNase Set (50), QIAGEN, 79254) for removal of genomic DNA. The RNA concentration and purity was measured spectrophotometrically. Reverse transcription was performed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems Darmstadt, Germany,
EP
Primer/Probe Design
TE D
4368813) according to the manufacturer’s instructions.
Human mRNA sequences for BCR (reference sequence: NM_004327.3) and ABL
AC C
(M14752.1) were downloaded from the nucleotide database NCBI (National Center for Biotechnology Information, date of last access: 03. April 2018) https://www.ncbi.nlm.nih.gov/nuccore and cross referenced with BCR-ABL real time-PCR standard pME2 including BCR-ABL translocation sequence (e13/a2 and e14/a2 transcript forms). After testing the background signals generated by three different BCR-ABL1 primer/probe combinations the choice fell on the primer system published 2014 by the group of Jennings [9] with a FPR of 2% (2/92 replicates). To establish a duplex BCR-ABL/ABL PCR, a total of 10 different ABL primers and probes for exons 4, 8, 10, and 11 were tested for optimal annealing temperature and background signals. An ABL exon 10 primer/probe 4
ACCEPTED MANUSCRIPT combination generating an amplicon length of 106bp was chosen on the basis of a low FPR and clear separation of the ABL positive droplets.
Primers and Probes for BCR-ABL PCR (IDT - Integrated DNA Technologies London, UK)
B-A_for (BCR exon 13 (b2)) 5’-CATTCCGCTGACCATCAATA-3’;
SC
B-A_rev (ABL exon 3) 5’-ACACCATTCCCCATTGTGAT-3’;
RI PT
Amplicon length b2a2 (e13/a2) transcript: 247bp; b3a2 (e14/a2) transcript: 330bp
B-A_probe (ABL exon 2) 5’-FAM-CCCTTCAGCGGCCAGTAGCATCTGA-MGBEclipse or
M AN U
MGBNFQ-3’.
Note that B-A for does not take account of a polymorphism at the penultimate base. Although the Jennings primer is suitable for K562 cDNA, reliable analysis of patient material requires a
TE D
Y at this position (5’-CATTCCGCTGACCATCAAYA-3’).
Primers and Probes for ABL PCR (IDT)
EP
Amplicon length ABL: 106bp
ABL_for (ABL exon 10) 5’-GGAAAAGGAGCTGGGGAAAC-3’;
AC C
ABL_rev (ABL exon 10) 5’-GTGCTCTGCAGCTCTCCTG-3’; ABL_probe (ABL exon 10) 5’-HEX-AGCTGCCCACCAAGACGAGGACCTC-ZENIBFQ-3’ (Probes were purified with HPLC and primers with standard desalting.)
ddPCR
Optimized reactions were performed in 20µL duplex ddPCR reaction mix consisting of 1x ddPCR Supermix for Probes (Bio-Rad München, Germany), primers B-A_for, B-A_rev, 5
ACCEPTED MANUSCRIPT ABL_for and ABL_rev (fc 1360nM), probes B-A_probe, ABL_probe (fc 450nM) and 5µL template cDNA (500ng reverse transcribed RNA). After sealing and centrifugation (1min/86xg), the 96well plates were transferred to the QX200 Droplet Generator (Bio-Rad) to generate around 20000 droplets per well. The PCR was performed in a Thermocycler T100
RI PT
(Bio-Rad) using the following protocol: 95 °C/10min , 70 cycles of denaturation 94 °C/30sec and annealing/extension 58 °C/96sec (ramp rate sett ings 2.5 °C; heated lid set to 105 °C; sample volume 40µL) finishing with one cycle 98 °C/ 10min and holding at 12 °C. Positive and negative droplets were counted using the QX200 Droplet Reader (Bio-Rad) and
SC
analyzed by Poisson statistics using the Software QuantaSoft 1.7.4.0917. Validation of the BCR-ABL/ABL dPCR assay was performed according the dMIQE (Minimum Information for
M AN U
Publication of Quantitative digital PCR Experiments) [3]. The dMIQE checklist is shown in Supplemental Table S1.
BCR-ABL levels were expressed as BCR-ABL/ABL ratios. Since the ABL exon 10 primers bind additionally to BCR-ABL cDNA, effectively increasing the ABL signal, the calculated
TE D
BCR-ABL/ABL ratio tends to be underestimated at very high BCR-ABL copy numbers [4].
EP
Results
AC C
Annealing temperature and background prior to PCR optimization
The optimum annealing temperature for the BCR-ABL/ABL duplex ddPCR was determined by gradient PCR over an annealing/extension temperature range from 55.1 °C to 63.5 °C (Bio-Rad manufacturer´s note). Optimum results were obtained at 58.3 °C (Figure 1A and 1B). The BCR-ABL PCR generated false positive rates of 2% using water controls (singleplex BCR-ABL PCR: 2/92; duplex BCR-ABL PCR: 1/64). The corresponding false positive rate for BCR-ABL using wild-type cDNA controls was around 5% (5/92 using HEK 293T cells and 2/37 using healthy donor). An annealing temperature of 58 °C was chosen for further experiments. 6
ACCEPTED MANUSCRIPT
Choice of quenchers
To improve the resolution in fluorescence between positive and negative droplets, the probes
RI PT
specific for each of the duplex PCR products (B-A1_FAM-MGBNFQ probe (BCR-ABL PCR) and ABL_HEX-BHQ probe (ABL-PCR) were tested against newly available and stronger quencher modifications (B-A1_probe FAM-ZEN-IOWA Black FQ; ABL_probe HEX-ZENIOWA Black FQ; Figure 2). The ZEN-IOWA Black FQ quencher appeared to be stronger than
SC
MGBNFQ, but the fluorescence signals of both the positive and the negative drops were suppressed similarly. There was therefore no improvement in the BCR-ABL PCR droplet
M AN U
separation using the new quencher (Figure 2A). Furthermore an increased frequency of background signals was detected in water controls when using the B-A1_probe FAM-ZENIOWA Black FQ (21%, 20/94) compared to B-A1_FAM-MGB probe (4%, 4/94). In contrast, the ZEN-IOWA Black FQ probe for ABL resulted in an improved separation
TE D
compared to that achieved with the ABL BHQ probe used to date (Figure 2B). All further experiments made use of the BCR-ABL MGB probe and the ABL ZEN-IOWA
EP
Black FQ probe.
AC C
Optimizing the PCR conditions
To further improve the resolution of positive and negative droplets in the BCR-ABL ddPCR, the following PCR conditions were tested first singly and then in combination using K562 cDNA as target. (To present more rain a high concentration of K562 cDNA was used as target to maximize potential PCR inhibition effects.): Ramp rate [°C/sec]: >3; 3; 2; 1.5; 1 (2.5); Number of cycles [cycle x-fold]: 40; 50; 60; 70; Annealing/Extension times [min]: 0.5; 1; 1.5; 2; 2.5; Denaturing time [sec]: 30; 45; 60 (Conditions used in initial testing are underlined).
7
ACCEPTED MANUSCRIPT The results are shown in Figure 3 and identify the following conditions to be optimal: Ramp rate: 1 °C/sec, resulting in a 30% increase in fluo rescence in BCR-ABL ddPCR (1.5 °C/sec led to the best separation in ABL ddPCR); Cycle number: 70, resulting in a 100% increase in fluorescence in BCR-ABL ddPCR and in improved separation. Annealing/extension time: 2.5
time: 30 sec, longer times resulted in no further improvement.
RI PT
mins, 50% increase in fluorescence in BCR-ABL ddPCR, improved separation. Denaturation
It thus proved possible to further optimize three of the four parameters separately. These
SC
were then tested in combination (Figure 4). Interestingly, conditions in which the shorter annealing times were used failed to generate double positive droplets, even when the
M AN U
resolution between positive and negative droplets in 1D plots was distinct. This suggests competitive inhibition, most probably of the BCR-ABL PCR by the ABL PCR. This would have the effect of shifting the BCR-ABL/ABL ratio to indicate lower levels of residual disease. However, the use of longer annealing/extension times overcomes this problem (Figure 4B, 4-
TE D
8), the best separation being seen under the conditions shown in Figure 4B, 8. Under these conditions, the maximum BCR-ABL fluorescence signal was nearly three-fold that seen before optimization and the BCR-ABL/ABL copy number was increased significantly from
EP
3360/21080 copies to 10780/21000 copies (Figure 4B, 1 and 4B, 8 respectively). Although the maximum level of ABL fluorescence was essentially unchanged, the increased cycle
AC C
number of conditions 2-8 effectively increased the fluorescence from the weaker droplets. This reduces the “rain” effect caused by droplets that do not clearly belong to the positive or negative cluster [10, 11]. The rain is generated by droplets of intermediate fluorescence intensity resulting from low-yield PCR amplifications that in turn result from a suboptimal PCR efficiency or a premature termination. The following optimum conditions were used in further experiments: ramp rate 1 °C/s, cycle number 70, and annealing/extension time 2.5 min.
Optimizing primer/probe concentration 8
ACCEPTED MANUSCRIPT To further reduce the rain effect in the BCR-ABL reaction, a range of primer/probe concentrations were tested under the previously optimized PCR conditions (Figure 5). Increasing the primer to a final concentration (fc) of 1.36µM led to minor improvement in
RI PT
BCR-ABL fluorescence (Figure 5). However, increasing the probe concentration from 250nM to 450nM resulted in a doubling of the fluorescence signal from both BCR-ABL and ABL, with an associated reduction in the rain effect and improvement of the separation (Figure 5). This suggests that the concentration of the probe is limiting.
SC
For further experiments a primer concentration of 1.36µM and a probe concentration of
M AN U
450nM was chosen.
Background signals generated by the optimized assay in no-template control (NTC) and wildtype cDNA controls
Having performed numerous modifications to the PCR conditions during the optimization, it
TE D
was necessary to re-test the background signals using the new conditions. The FPR was calculated as false positive samples/(false positive + true negative samples). Although the false positive rate was marginal in NTCs at 3% (5/188) it was very high for wild-type controls,
AC C
(Table 1).
EP
with 64% to 100% positivity from healthy donor and a BCR-ABL negative cell line cDNA
To identify the parameter(s) responsible for the decreased specificity (specificity = true negatives/(true negatives + false positives)), a series of reactions were performed using a high level of wild-type target (HEK 293T cDNA, 70.000 ABL equivalents, Figure 6). As expected, no BCR-ABL background was seen under the original conditions (Figure 6A). Increasing the primer/probe concentration alone or in combination with a decreased ramp rate also resulted in no detectable background. However, several positive BCR-ABL signals were generated at the longer annealing/extension times. The increase from 40 to 70 PCR 9
ACCEPTED MANUSCRIPT cycles resulted in only one BCR-ABL positive droplet from six reaction wells but in combination with a lower ramp rate (which effectively increases extension time) resulted in background noise around the BCR-ABL negative fraction. This suggests that the BCR-ABL background signals are favored by longer amplification times (Figure 6B), so rapid ramping
RI PT
and shorter annealing and extension times “focus” the reaction on the specific product. This was confirmed in a separate set of reactions employing a rapid ramp rate (2.5 °C/s) and shorter annealing/extension time (1min), which eliminated the unspecific signals in all wild-
SC
type controls (Figure 6B).
A ramp rate of 2.5 °C/s and an annealing/extension time of 1min were chosen for further
M AN U
work.
BCR-ABL FPR and Copy Number (CN) controlled by time of annealing/extension
TE D
A final concern was that the reduction in annealing/extension time necessary to eliminate the non-specific background may have the undesired effect of increasing the competition between the ABL and BCR-ABL amplification reactions as noted above (Figure 4), once
EP
again eliminating the double positive droplets. It was therefore important to determine whether an optimal annealing/extension time maintains both sensitivity and specificity.
AC C
A comparison of sensitivity (BCR-ABL and ABL copy numbers) and specificity at annealing times between 1 and 2.5 min is shown in Table 2. An annealing/extension time of 1.6 mins maintained a specificity of 98% (FPR 2%), while detecting BCR-ABL copy numbers close to those seen using real time-PCR after conversion to the international scale. The shorter annealing/extension time resulted in absolute specificity but a very low rate of detection, whereas longer times reduced specificity to unacceptable levels. After balancing of specificity and sensitivity an annealing/extension time of 1.6min was chosen for the final, fully optimized assay. 10
ACCEPTED MANUSCRIPT
Final assay, Linearity, and Precision
A 2D plot of fluorescence signals generated using the final PCR conditions compared to
RI PT
those used at the outset (Figure 7A) is shown in Figure 7B. The sequential modifications result in a 5-fold increase in BCR-ABL fluorescence, a doubling of ABL fluorescence and the resolution of a population of double-positive droplets indicating little or no competition
SC
between the targets.
Six log10 dilutions of BCR-ABL positive K562 cDNA were prepared to generate ABL copy
M AN U
numbers from 100000 – 1 (Figure 7C). High linearity for ABL was seen over the complete range. BCR-ABL detection was also highly linear at copy numbers below 4000 (dilutions 2 6). However the most concentrated BCR-ABL dilution (40000 copies) generated a detected copy number of only 18,000 copies, suggesting PCR inhibition at high target concentration. With quantitative detection of between 1 and 4000 BCR-ABL copies in patient sample
TE D
replicates containing 40000 ABL copies, linearity would be given between 10% and 0.0025% BCR-ABL/ABL corresponding to MR4.5. Although standard deviation of the BCR-ABL/ABL ratios was increased in samples 5 and 6 that contained limiting amounts of target (31 and
EP
30% BCR-ABL/ABL), those in dilutions 1 to 4 were between 1% to 4%. The median intraassay coefficient of variation of all six dilutions was below 10% (ABL CV = 8%; BCR-ABL CV
AC C
= 6%; BCR-ABL/ABL CV = 9%), indicating that the assay provides reliable and accurate data with high precision.
As expected from the Poisson distribution, the variation in BCR-ABL copy numbers increases markedly as the copy number drops below 100, equivalent to the range MR3 to MR5.5. Accordingly, the analysis of patient samples (n=48) revealed a BCR-ABL inter-assay CV increasing from 22% for MR3 samples to 288% for MR5.5 samples (Table 3).
Limit of Blank (LOB), False-Positive-Rate, and Specificity of final assay 11
ACCEPTED MANUSCRIPT The FPR required for LOB determination was analyzed in 92 replicates of no template controls. The observed BCR-ABL FPR of 1% (1/92) is equivalent to a LOB of 0 for >95% of the replicates.
RI PT
To check the background noise in samples from healthy donors, cDNA from two independent white blood cell (buffy coat) preparations were assayed as follows: 188 aliquots of cell lysates (each 1x 10E7 cells) were extracted (final 40µL RNA eluate, mean 224ng/µL RNA)
SC
and reverse transcribed (each 4µg RNA to final 40µL cDNA). The cDNAs were then pooled and diluted (factor 1.6). 174 technical replicates of 5µL cDNA were analyzed (mean 63000
M AN U
ABL copies per replicate). The FPR (presence of a single positive drop in a ddPCR reaction) for BCR-ABL in wild-type cDNA was 4% (3/84) and 7% (6/90) resulting in an overall FPR of 5% (9/174) and a specificity of 95% (“exact” Clopper-Pearson 95% confidence interval: 90%
TE D
to 98%).
Limit of Detection (LOD) and Sensitivity of final assay
EP
LOD and sensitivity were measured in serial dilutions of CML patient cDNA (9.5% BCRABL/ABL (IS) as determined by real time-PCR standardized to the international scale) in
AC C
healthy donor cDNA to simulate MR3 down to MR5.5 (Table 4). Three batches of 16 replicates for each MRD dilution were analyzed at three different time points. The ddPCR method yielded MR levels identical to those expected, confirming the comparability of the two assay platforms. The sensitivity is defined as detection rate, calculated as true positive samples/(true positive + false negative samples). Analyses in duplicate resulted in BCR-ABL detection rates of 100% and 92% from MR4 (0.009% BCR-ABL/ABL) and MR4.5 (0.003%) samples respectively. Analysis of four wells resulted in 100% and 75% detection of MR4.5 (0.003%) and MR5 (0.001%), respectively.
12
ACCEPTED MANUSCRIPT The LOD was one copy of BCR-ABL per set of duplicates or per four well analyses (Table 4). The duplex ddPCR assay can therefore detect one BCR-ABL transcript per 120,000 wild-
Comparison of different internal control sequences
RI PT
type copies per replicate with high specificity.
Given the evidence of ABL product interference with the BCR-ABL reaction under suboptimal
SC
conditions seen in the 2D plots in Fig. 4B, the incidence of double positive drops was
assessed using alternative internal controls. Five different master mixes (MM) were used
M AN U
under pre- and post-optimization ddPCR conditions: (MM1) employs the EAC BCR-ABL/ABL primers and probes used in the international standard real-time PCR reaction, (MM2 to MM5) employ the Jennings BCR-ABL primers/probes together with EAC GUS primers (MM2), the exon 10 ABL probe to the opposite strand of the published assay (MM3), exon 8 ABL primer/probes (MM4), and the assay published in this paper (MM5) for comparison. All
dilution (Table 5).
TE D
master mixes were tested in the same run and with the same BCR-ABL positive K562 cDNA
The EAC primer probe set (MM1) was highly sensitive even under the pre-optimization
EP
conditions that, however, generate a high background as described in the introduction. Each of the primer/probe sets MM2 to MM5 generated low frequencies of double positives under
AC C
pre-optimization conditions that were markedly increased by optimization. Furthermore, the frequency of BCR-ABL copies detected under optimized conditions did not vary with the internal control sequence, demonstrating that the optimized procedure is robust in this respect.
Discussion
13
ACCEPTED MANUSCRIPT The key advantage of ddPCR over quantitative real-time-PCR lies in the detection of endproducts generated by single target molecules. A drop containing amplified product is scored positive, regardless of the efficiency of the reaction, making the procedure relatively resistant to the effects of PCR inhibition [1, 2]. Given a suitable primer set, accurate and reliable
RI PT
BCR-ABL monitoring should be achievable without the need for standard curves. Furthermore, analysis of both BCR-ABL and ABL in the same (duplex) PCR reaction avoids inaccuracies introduced by pipetting variations.
The relative insensitivity of the digital PCR end-point readout to PCR inhibition may be
SC
particularly relevant to treatment decisions that depend on accurate determination of MRD. The discontinuation of TKI therapy in CML patients who have achieved stable deep
M AN U
molecular remission is one example of this. Here, partial inhibition of a real time-PCR reaction would, by delaying the accumulation of reaction products, deliver a read out below the level of MRD actually present and may lead to premature discontinuation. Since partial inhibition and delay of amplification in a ddPCR reaction has little or no effect on the endpoint
decision.
TE D
read out, ddPCR should in this case provide a more reliable basis for informing the therapy
A comparison of three primer/probe sets identified one previously published by the group of
EP
Jennings to be the most specific and suitable. Starting with this, a duplex PCR was designed incorporating the wild type control gene ABL. The resolution between positive and negative
AC C
droplets was initially poor, mostly due to the relatively low fluorescence of positive droplets. However, three strategies were identified to improve this separation. The clearest advantage came from increasing the concentration of fluorescent probe, which is typically kept low due to cost considerations. Here, a 2- to 3-fold increase in probe concentration was found to generate a comparable increase in fluorescence intensity. Notably, increasing the cycle number up to 70 proved to be an effective way of boosting weak signals, reducing the rain effect, and thereby improving the separation between positive and negative samples. Finally, the choice of Quencher may also have a minor effect on separation due to probe-specific differences in quenching efficiencies. In this way, it proved possible to reduce the rain effect 14
ACCEPTED MANUSCRIPT and generate the double-positive fraction that is predicted on the basis of stochastic segregation of target between droplets. Changing the PCR reaction conditions for a given set of primers often has reciprocal effects on specificity and sensitivity, so that optimization comes down to finding the best balance
RI PT
between the two. Accordingly, low ramp rates and extended annealing/extension times tended to increase the frequency of false positive BCR-ABL signals in the form of weakly positive drops generated from wild-type (non–BCR-ABL) target cDNA. On the other hand, increasing the ramp rate (to 2.5 °C/s) and reducing the annealing/extension time (to 1 min)
SC
eliminated the non-specific signals but reduced the sensitivity, leading to underestimation of BCR-ABL in positive controls. Fortunately, it proved possible to combine high specificity
M AN U
(98%) with a degree of sensitivity that generated BCR-ABL/ABL ratios comparable to those of the current (IS) real time-PCR standard by adjusting the annealing/extension time to 1.6 min. This dependence of the BCR-ABL/ABL ratio readout on the annealing time suggests that it may be beneficial to employ PCR- and laboratory-specific conversion factors for
TE D
ddPCR in the same way that they are currently used for real time-PCR. Using clinical CML samples, the optimized ddPCR assay detected one BCR-ABL copy in 54% of reactions (performed in duplicate) from an MR5 sample (≤0.0010% BCR-ABL/ABL),
EP
whereas under the same conditions an MR4.5 (≤0.0032% to ≥0.0010% BCR-ABL/ABL) sample detected three copies in 92% of the assays. This is close to the frequency expected
AC C
from the Poisson distribution, which predicts 95% of samples taken from a solution with three targets per unit volume to contain at least one copy of the target [12]. In summary, we report the development of a low-background, duplex ddPCR BCR-ABL/ABL assay with specificity in blank controls of 98%. The cross-reaction with wild-type cDNA does not exceed 5%, making the use of a cut-off unnecessary. Furthermore, we optimized the conditions of the assay so that positive and negative droplets are well-separated and the rain effect is either markedly reduced (BCR-ABL) or eliminated (ABL). The assay conditions have been adjusted to allow BCR-ABL and ABL PCR reactions to proceed with similar efficiency in a single drop, evidenced by a well-separated fraction of double-positive droplets. The assay 15
ACCEPTED MANUSCRIPT provides accurate data with high precision and high linearity between 1 and 4000 copies of BCR-ABL and between 1 and 100000 copies of ABL. The linearity and ratio were reduced at BCR-ABL copy numbers above 4,000, but this is not relevant for samples with molecular response between MR2 and MR5. The BCR-ABL amplification described here is more
RI PT
sensitive to changes in PCR conditions than is the ABL reaction, but stepwise optimization resulted in a detection rate of 92% for MR4.5, close to the theoretical ideal of 95%. Clearly, sensitivity could be increased further by extending the analysis to more replicates.
The observations reported here should be of general use in optimizing ddPCR protocols,
SC
especially in the diagnostic setting in which primer choice is necessarily limited and even the
AC C
EP
TE D
M AN U
best available primer sets can require careful optimization of reaction conditions.
16
ACCEPTED MANUSCRIPT
References Yang R, Paparini A, Monis P, Ryan U: Comparison of next-generation droplet digital PCR (ddPCR) with quantitative PCR (qPCR) for enumeration of Cryptosporidium oocysts in faecal samples. Int J Parasitol 2014:1105–1113. 2 Blaya J, Lloret E, Santisima-Trinidad AB, Ros M, Pascual JA: Molecular methods (digital PCR and real-time PCR) for the quantification of low copy DNA of Phytophthora nicotianae in environmental samples. Pest Manag Sci 2016:747–753. 3 Huggett JF, Foy CA, Benes V, Emslie K, Garson JA, Haynes R, Hellemans J, Kubista M, Mueller RD, Nolan T, Pfaffl MW, Shipley GL, Vandesompele J, Wittwer CT, Bustin SA: The digital MIQE guidelines: Minimum Information for Publication of Quantitative Digital PCR Experiments. Clin Chem 2013:892–902. 4 Gabert J, Beillard E, van der Velden, V H J, Bi W, Grimwade D, Pallisgaard N, Barbany G, Cazzaniga G, Cayuela JM, Cave H, Pane F, Aerts JL, Micheli D de, Thirion X, Pradel V, Gonzalez M, Viehmann S, Malec M, Saglio G, van Dongen, J J M: Standardization and quality control studies of 'real-time' quantitative reverse transcriptase polymerase chain reaction of fusion gene transcripts for residual disease detection in leukemia - a Europe Against Cancer program. Leukemia 2003:2318–2357. 5 Franke G, Maier J, Wildenberger K, Cross M, Frank O, Giles F, Hochhaus A, Dietz CT, Müller MC, Niederwieser D, Lange T: Quantification of BCR-ABL with Digital PCR Results in a Significantly Lower Rate of Deep Molecular Responses When Compared to RT-qPCR in CML Patients Treated in the ENEST1st Trial. Blood 2015:135. 6 Maier J, Franke G, Schubert K, Wildenberger K, Cross M, Niederwieser D, Lange T: A Comparison of Droplet Digital PCR and Quantitative RT-PCR for Low Level BCR-ABL in CML Patients with Molecular Responses. Blood 2014:1792. 7 Lozzio CB, Lozzio BB: Human chronic myelogenous leukemia cell-line with positive Philadelphia chromosome. Blood 1975:321–334. 8 Pegoraro L, Matera L, Ritz J, Levis A, Palumbo A, Biagini G: Establishment of a Ph1positive human cell line (BV173). J Natl Cancer Inst 1983:447–453. 9 Jennings LJ, George D, Czech J, Yu M, Joseph L: Detection and quantification of BCRABL1 fusion transcripts by droplet digital PCR. J Mol Diagn 2014:174–179. 10 Jacobs BK, Goetghebeur E, Vandesompele J, Ganck A de, Nijs N, Beckers A, Papazova N, Roosens NH, Clement L: Model-Based Classification for Digital PCR: Your Umbrella for Rain. Anal Chem 2017:4461–4467. 11 Jones M, Williams J, Gartner K, Phillips R, Hurst J, Frater J: Low copy target detection by Droplet Digital PCR through application of a novel open access bioinformatic pipeline, 'definetherain'. J Virol Methods 2014:46–53. 12 Cross NC, White HE, Colomer D, Ehrencrona H, Foroni L, Gottardi E, Lange T, Lion T, Machova Polakova K, Dulucq S, Martinelli G, Oppliger Leibundgut E, Pallisgaard N, Barbany G, Sacha T, Talmaci R, Izzo B, Saglio G, Pane F, Muller MC, Hochhaus A: Laboratory recommendations for scoring deep molecular responses following treatment for chronic myeloid leukemia. Leukemia 2015:999–1003.
AC C
EP
TE D
M AN U
SC
RI PT
1
17
ACCEPTED MANUSCRIPT Figure Legends Figure 1. A: Optimal annealing temperature for BCR-ABL and ABL ddPCR in 1D Plot (A) and 2D Plot (B) at temperature optimum. Representative ddPCR results are shown for BCR-ABL positive K562 cDNA. Channel 1 amplitude showing BCR-ABL positive (blue) and
RI PT
negative (black) droplets, separated by the threshold (pink line). Channel 2 amplitude showing ABL positive (green) and negative (black) droplets, separated by the threshold (pink line). Yellow dashed lines separate different reaction wells. A temperature gradient from 63.5
SC
°C to 55.1 °C was used and annealing temperature of 58.3 °C (asterisk) was selected as
M AN U
optimal.
Figure 2. A: Comparison of BCR-ABL signal and background fluorescence levels obtained using MGBNFQ and Zen-IBFQ quenched probes (A) and comparison of ABL signal and background fluorescence levels obtained using BHQ and Zen-IBFQ
TE D
quenched probes (B). A: Results from MGBNFQ and Zen-IOWA black probe. B3a2 and b2a2 indicate reactions using the two respective BCR-ABL transcripts. Channel 1 amplitude (y-axis) showing fluorescence units of BCR-ABL positive (blue) and negative (black) droplets, separated by the threshold (pink line). B: Signals obtained from the BHQ quenched probe
EP
are shown on the left and those from the Zen-IOWA black quenched probe on the right.
AC C
Channel 2 amplitude (y-axis) showing ABL positive (green) and negative (black) droplets, separated by the thresholds (pink lines). Yellow dashed lines separate different reaction wells. MGBNFQ (Minor Groove Binder Nonfluorescent Quencher); BHQ (Black Hole Quencher); ZEN-IBFQ (ZEN-IOWA Black FQ, Dark Quencher Modification (IDT).
Figure 3. Optimal ddPCR conditions according to ramp rate (A), cycle number (B), annealing/extension (C), and denaturation (D). Ramp rate (A), cycle number (B), annealing/extension time (C), and denaturation time (D) were tested for both the BCR-ABL (blue) and ABL (green) reactions. Pink line separates amplification positive from negative 18
ACCEPTED MANUSCRIPT droplets and yellow dashed lines separate different runs. Initial conditions are underlined. The condition identified as optimal in each case is denoted with an asterisk.
Figure 4. Optimization of ddPCR conditions ramp rate, cycle number, and
RI PT
annealing/extension time. A: Eight duplex BCR-ABL and ABL ddPCR reactions (denoted 1 to 8, separated by yellow dashed lines) were performed using variable ramp rates, cycle numbers, and annealing/extension times as shown in the table (gray shaded sections). The
SC
fluorescent read-outs are shown as one-dimensional plots directly below the respective condition. B: Two dimensional plots of the results of reactions (1-8) shown in A. Initial
M AN U
conditions are underlined (1). Condition 8 was determined to provide the optimal signal to noise and is marked with an asterisk. Pink line separates amplification positive from negative droplets.
TE D
Figure 5. Optimization of primer/probe concentrations. A set of eight duplex BCR-ABL and ABL ddPCR reactions were performed using variable concentrations of primers and probes (gray shaded sections at the top). Yellow dashed lines separate different reaction wells. The results are shown in a one-dimensional plot directly below the respective
EP
condition. The condition chosen as optimal is marked with an asterisk. Pink line separates
AC C
amplification positive from negative droplets.
Figure 6. Background cause analysis using different combinations of ddPCR conditions and primers and probe concentrations (A) and reduction of background (B). A: A set of six reaction conditions was chosen to determine the effects of primer/probe concentration, cycle number, ramp rate, and annealing time on the background frequency of positive droplets. Initial conditions are shown in the table in red and changed conditions in blue numbers above the plots. Each reaction condition was used to amplify NTC (n=2), positive control K562 (n=2), and wt HEK 293T (n=6) samples. Increased background was 19
ACCEPTED MANUSCRIPT associated with a lower ramp rate (2.5 °C/s vs 1 °C /s) and a longer annealing/extension time (2.5 minute vs 1 minute). B: The effects of ramp rate and annealing/extension time were confirmed in an independent set of reactions amplifying NTC (n=2), positive control K562 (n=2), and wt HEK 293T (n=92). Pink line separates amplification positive from negative
RI PT
droplets and yellow dashed lines separate different reaction wells.
Figure 7. Digital PCR assay M-BCR-ABL/ABL before (A) and after optimization (B) and
SC
linearity of optimized ddPCR duplex Assay (C). 2D Plot. A: ramp rate 2.5 °C/s, cycle number 40x, annealing/extension 1min 58 °C, Primers fc. 0.91µM, Probes fc. 250nM; 2D
M AN U
Plot. B: Ramp rate 2.5 °C/s, cycle number 70x, annealing/ex tension 1.6 min 58 °C, Primers fc. 1.36µM, Probes fc. 450nM; BCR-ABL positive droplets (blue); ABL positive droplets green; negative for BCR-ABL and ABL black; doubled positive (red) were separated by pink lines. C: Copy number of BCR-ABL and ABL are shown on the primary axis (left y-axis) and
AC C
EP
TE D
the ratio is shown on the secondary axis (right y-axis)
20
ACCEPTED MANUSCRIPT Table 1. False-positive rate (FPR) of NTC and wild-type with changed conditions.
BCR-ABL FPR [%] (n)
BCR-ABL/10000 ABL [copies]
NTC
3 (5/188)
-
HEK 293T
100 (92/92)
2.0
UCB
78 (70/90)
0.8
Healthy donor
64 (59/92)
0.3
Median ABL [copies]
RI PT
Template
0
M AN U
SC
82000
21000
48000
Conditions: ramp rate 1 °C/sec; cycles 70; annealin g/extension 2.5min; Primers fc. 1.36µM; Probes fc. 450nM; NTC (no template control); HEK (human embryonic kidneys); UCB
AC C
EP
TE D
(umbilical cord blood).
21
ACCEPTED MANUSCRIPT Table 2. BCR-ABL FPR and copy number controlled by time of annealing/extension. Ann./ext. [min]
Specificity/FPR (wt) [%]
BCR-ABL CN (1*)
ABL CN (*1)
% Ratio (*1)
ddPCR ddPCR ddPCR ddPCR ddPCR RT-PCR(IS)
1 1.6 1.7 1.8 2.5 1
100/0 98/2 93/7 86/14 0/100 100/0
44 707 918 1246 1649 827
76700 60383 11340 47360 54898 58808
0.1 1.2 2.1 2.6 3.0 1.4
RI PT
Assay
*1 - BCR-ABL positive sample; CN (copy number); FPR (False-Positive-Rate); wt (wild type);
SC
ddPCR (Droplet Digital PCR); RT-PCR (Real-Time PCR); IS (international scale) - for
standardization to IS BCR-ABL CN was multiplied with a laboratory-specific conversion
AC C
EP
TE D
M AN U
factor.
22
ACCEPTED MANUSCRIPT Table 3. Interassay Variation of MR3 to MR5.5 CML patient samples (n = 48).
%CV ABL (mean CN)
%CV BCRABL/ABL (mean)
MR3
22 (58.3)
9 (64600)
20 (0.09036)
MR4
53 (5.7)
8 (62175)
52 (0.00922)
MR4.5
82 (1.6)
8 (60493)
82 (0.00265)
MR5
143 (0.6)
8 (62641)
142 (0.00100)
MR5.5
288 (0.3)
9 (63053)
287 (0.00039)
SC
RI PT
Patient Sample
%CV BCR-ABL (mean CN)
n = 3x 16 replicates for each MRD level determined at three different time points; ddPCR
M AN U
(Droplet Digital PCR); MRD (molecular residual disease); CN (copy number); MR (molecular
AC C
EP
TE D
remission).
23
ACCEPTED MANUSCRIPT Table 4. Sensitivity of CML patient samples (n = 48)
2-well Analysis
4-well Analysis
% Sensitivity
BCR-ABL CN ABL CN
MR3
117
129201
100 (86 - 100) 233
MR4
11
124350
100 (86 - 100) 23
MR4.5
3
120987
92 (73 - 99)
7
MR5
1
125283
54 (33 - 74)
2
MR5.5
0,5
126106
21 (7 – 42)
1
M AN U
(*95% CI)
BCR-ABL CN ABL CN
% Sensitivity (*95% CI)
258402
100 (74 - 100)
248700
100 (74 - 100)
241973
100 (74 - 100)
SC
MRD level
RI PT
ddPCR
250565
75 (43 - 95)
252212
42 (15 - 72)
n = 3x 16 replicates for each MRD level determined at three different time points; *Confidence intervals for sensitivity are “exact” Clopper-Pearson confidence intervals; ddPCR (Droplet Digital PCR); MRD (molecular residual disease); CN (copy number); MR
AC C
EP
TE D
(molecular remission).
24
ACCEPTED MANUSCRIPT Table 5. Test of changed and unchanged PCR conditions within different assays.
Assay
Before/after optimization BCR-ABL
CG
BCR-ABL/CG
droplets
copies
copies
[%]
MM1
3559 / 3782
8930 / 9340
10410 / 11450
86 / 82
MM2
199 / 508
4990 / 5590
1584 / 3800
315 / 147
MM3
974 / 1707
4360 / 5770
12620 / 12280
35 / 47
MM4
195 / 1516
2990 / 5370
12670 / 11650
24 / 46
MM5
81 / 1343
2920 / 5410
12310 / 11840
24 / 46
AC C
EP
TE D
M AN U
SC
RI PT
Ch1+Ch2+ 20µl reaction
25
ACCEPTED MANUSCRIPT A
°C 63.5 62.9 61.8 60.2 *58.3 56.7 55.7 55.1
M AN U
SC
RI PT
BCR-ABL
EP AC C
B
TE D
ABL
BCR-ABL
ABL
ACCEPTED MANUSCRIPT
MGBNFQ
A
b2a2
b3a2
RI PT
b3a2
ZEN-IBFQ
B
AC C
ABL
ZEN-IBFQ
EP
BHQ
TE D
M AN U
SC
BCR-ABL
b2a2
ACCEPTED MANUSCRIPT
A
B
Ramp rate °C/s
>3
3
2
1*
1.5
Cycle number X-fold
40
50
60
SC
RI PT
BCR-ABL
Annealing/extension 1
1.5
2
AC C
BCR-ABL
0.5
ABL
2.5*
EP
min
TE D
C
M AN U
ABL
D
Denaturation
sec
30*
45
60
70*
ACCEPTED MANUSCRIPT A 1
2
3
4
5
6
7
8*
ramp rate [°C/s]
2.5
2.5
1.5
1
1.5
1.5
1.5
1*
cycle [x-fold]
40
70
70
70
70
70
70
70*
ann./ext. [min]
1
1
1
1
1.5
2
2.5
2.5*
RI PT
condition
M AN U
SC
BCR-ABL
ABL
AC C
2
EP
1
TE D
B
5
6
3
7
4
8*
0.45
0.91
1.36*
0.91
0.91
0.91
0.91 1.36*
Probes fc [nM]
250
250
250
150
250
350
450*
M AN U
SC
Primers fc [µM]
AC C
EP
TE D
BCR-ABL
ABL
RI PT
ACCEPTED MANUSCRIPT
450*
ACCEPTED MANUSCRIPT
A 2.5 40 1 1.36 450
2.5 40 1 0.91 250
1 40 1 1.36 450
B
TE D
1 70 2.5 1.36 450
AC C
EP
rampe rate [°C/s] cycle [x-fold] ann./ext. [min] Primers fc [µM] Probes fc [nM]
ABL
M AN U
ABL
BCR-ABL
2.5 70 1 1.36 450
SC
BCR-ABL
2.5 40 2.5 1.36 450
RI PT
rampe rate [°C/s] cycle [x-fold] ann./ext. [min] Primers fc [µM] Probes fc [nM]
2.5 70 1 1.36 450
1 70 1 1.36 450
ACCEPTED MANUSCRIPT
B
A
cycle number 40x annealing/extension time 1min Primers fc. 0.91µM Probes fc. 250nM
TE D
M AN U
SC
RI PT
cycle number 70x annealing/extension time 1.6min Primers fc. 1.36µM Probes fc. 450nM
C
Linearity
10000
R² = 0.9999
AC C
1000
1000 100
100 10
1 0
10
R² = 0.9962
1
2
3
4
5
6
ABL
85333
9643
982
99
10
1
BCR-ABL
18051
4076
471
44
5
1
21
42
48
44
53
81
BCR-ABL/ABL
Log10 Dilutions of K562 cDNA
1
% BCR-ABL/ABL
EP
Copies per Reaction
100000