Digital PCR can provide improved BCR-ABL1 detection in chronic myeloid leukemia patients in deep molecular response and sensitivity of standard quantitative methods using EAC assays

BCR-ABL1 molecular detection using quantitative PCR (qPCR) methods is the golden standard of chronic myeloid leukemia (CML) monitoring. However, due to variable sensitivity of qPCR assays across laboratories, alternative methods are tested. Digital PCR (dPCR) has been suggested as a robust and reproducible option. Here we present a comparison of droplet dPCR with routinely used reverse-transcription qPCR (RT-qPCR) and automated GeneXpert systems. Detection limit of dPCR was above 3 BCR-ABL1 copies, although due to background amplification the resulting sensitivity was 0.01% BCR-ABL1 (MR4.0). Nevertheless, in comparison with GeneXpert, dPCR categorized more than 50% of the patients into different MR groups, showing a potential for improved BCR-ABL1 detection.

evaluating the suitability of dPCR as an alternative technique for BCR-ABL1 monitoring suggest that dPCR has increased sensitivity up to MR5.5, although higher false positivity rate (FPR) could lower the detection limit of this method [6][7][8][9][10]. Therefore, whether dPCR is indeed suitable for monitoring of CML patients remains unclear.
The objectives of our study were to evaluate the performance of droplet dPCR using the Europe Against Cancer (EAC) standardized BCR-ABL1 and ABL1 assays and to evaluate the suitability of dPCR for molecular monitoring of CML patients with low or undetectable levels of BCR-ABL1 in comparison with routinely used methods.

Samples
A retrospective analysis was performed on 70 clinical samples from chronic phase CML patients and 15 samples from healthy volunteers used as BCR-ABL1 negative controls. All samples were obtained from the Department of Internal Medicine, Haematology and Oncology, University Hospital Brno. Characteristics of CML patients are listed in Supplementary Table 1. Informed consent was obtained from all enrolled subjects prior to participation in the study. The study protocol and subject sampling were reviewed and approved by an institutional ethical committee in agreement with the Declaration of Helsinki. Sample handling and storage is described in detail in Supplementary Data.

Calibrators
ERM AD623 BCR-ABL pDNA (Sigma-Aldrich) plasmid set [11] was used to assess the linearity of EAC assays by dPCR, according to the manufacturer's recommendations. A Qiagen Ipsogen BCR-ABL1 Mbcr Fusion Gene Standard kit and an Ipsogen ABL Control Gene 3 Standard kit were used for quantification of BCR-ABL1 transcript by RT-qPCR in accordance with the manufacturer's recommendations.

Quantitative PCR
All qPCR measurements were performed at the Centre of Molecular Biology and Gene Therapy, University Hospital Brno. Two methods, routinely performed to assess treatment response, were used: conventional RT-qPCR and the GeneXpert-based assay.
RT-qPCR quantification of BCR-ABL1 in K562 cells was performed on an Applied Biosystems 7300 Real-Time PCR System. Cepheid Xpert BCR-ABL Monitor test was used for quantification of BCR-ABL1 in clinical CML samples, according to the manufacturer's recommendations. Reverse transcription and qPCR reaction conditions are described in detail in Supplementary Data and Table 1.

Digital PCR
All dPCR measurements were performed at the Central European Institute of Technology of Masaryk University Brno, Department of Molecular Medicine on Bio-Rad QX200 Droplet Digital PCR System (dPCR) according to the manufacturer's recommendations. Reaction conditions are described in Supplementary Data and Table 1.

Statistical analysis
The formulae for calculation of FPR, limit of blank (LOB) and limit of detection (LOD) are concluded in Supplementary Data. Statistical analyses were generated using GraphPad Prism v.9 software. P values < 0.05 were considered significant.

Detection limit of dPCR using EAC assays
Evaluation of EAC assays' linearity on dPCR showed near perfect correlation of expected and measured BCR-ABL1 and ABL1 copies across the dilution range of the ERM AD623 standard down to 10 copies, with R 2 BCR-ABL1 ¼ 0.998 and R 2 ABL1 ¼ 0.999, respectively (Supplementary Fig. 1).
To further assess the sensitivity of dPCR, we performed quantification of BCR-ABL1 in 3 independent dilution series. We used a cDNA sample from a CML patient with 10% BCR-ABL1 IS , which was serially diluted into BCR ABL1 negative control. Digital PCR detected BCR-ABL1 transcript all sample dilutions down to 0.0001% BCR-ABL1 IS , however after correction to LOD (3.3 copies/sample), the final sensitivity was corresponding to MR4.0 (0.01% BCR-ABL1 IS ) (Fig. 1A, Supplementary Table 3).
The discrepancy of copy numbers was also observed in samples from patients in major and deep molecular response (MR3) (n ¼ 12). RT-qPCR still quantified significantly more ABL1 copies than dPCR in these samples (median copies of 133,971 by RT-qPCR vs. 73,333 by dPCR) (Supplementary Table 6, Supplementary Fig. 4). The BCR-ABL1 copy numbers showed similar trend.

Discussion
The sensitivity of dPCR for BCR-ABL1 quantification has already been demonstrated, showing its superiority compared to RT-qPCR (LOD up to MR5.5) [12]. In our study, we showed that dPCR can attain sensitivity above MR5.0. However, we also observed a false positive amplification in BCR-ABL1 negative controls, not distinguishable from the lowest copy numbers in positive samples. Therefore, final sensitivity was decreased to MR4.0.
The assessment of the false positivity and the detection limit is especially important for the assays, monitoring minimal residual disease. The EAC assays for BCR-ABL1 detection have been validated on dPCR in the studies of Franke and Maier [9,10], revealing up to 5% false positivity. In our study, we observed a fairly high FPR of 6%. Alikian [7] described similarly increased false positivity in a study comparing three dPCR platforms, using the same combination of an EAC qPCR assays in multiplex, suggesting that this combination and type of assays might not be ideal for dPCR [7,8,13]. Indeed, Andersen and Pallisgaard reported that using a combination of GUS and BCR control gene assays, or GUS alone, resulted in sensitivity of MR5.5 [14]. Most recently, Maier's group validated new multiplex BCR-ABL1 and ABL1 dPCR assays with low-background, 98% specificity, and sensitivity of MR5.0 [10].
Herein, the BCR-ABL1 and ABL1 copy numbers quantified in K562 cell line and in CML patients by dPCR significantly differed from those detected by RT-qPCR. Franke et al. obtained similar results for samples with 0.1 BCR-ABL1% measured by dPCR, with significant redistribution of the patients into MR categories relative to RT-qPCR [8]. In addition, similar to our observation, Folta reported the discrepancy in ABL1 copies measured by two methods (median copies of 180,511 by RT-qPCR vs. 39,960 by dPCR) [15]. Since the cDNA synthesis was performed using the same RT enzyme, we hypothesize that differences among the two methods are likely caused by quantification principles (plasmid standard versus absolute quantification). In addition, dPCR outperformed RT-qPCR in quantifying the lowest transcript levels, which are most likely at the lower detection limit of RT-qPCR. We observed analogous observations when comparing dPCR with GeneXpert. In clinical samples with 0.1 BCR-ABL1% and BCR-ABL1 negative samples dPCR yielded more sensitive measurements, GeneXpert assay used for this comparison was validated to MR4.0 [16], which might influence the discrepancy at the lowest transcript levels (<0.01% BCR-ABL1 IS ). Still, more than half of the patients in MR3.0 -MR4.0 were categorized into different MR classes by dPCR. Similarly, Folta reported consistent MR levels only in 29/50 samples in a comparison of dPCR with more sensitive GeneXpert Ultra kit with sensitivity of MR4.5. [15] Our study demonstrated that droplet dPCR, tested with standard EAC assays, provided a detection limit of above 3 BCR-ABL1 copies/ sample, which corresponded to sensitivity of conventional quantitative methods. Nevertheless, dPCR categorized more than 50% of the CML patients into different MR categories compared to quantitative GeneXpert. CML patients in deep molecular response considered for TKI cessation could benefit from more sensitive monitoring using well optimized dPCR assays, however an international standardization independent from RT-qPCR would be required to promote dPCR as an alternative method.
CRediT authorship contribution statement

Declaration of competing interest
Authors DS, DD, ZR, and MR have no competing interests.