Applicability of droplet digital polymerase chain reaction for minimal residual disease monitoring in Philadelphia‐positive acute lymphoblastic leukaemia

Abstract In Ph+ acute lymphoblastic leukaemia (Ph+ ALL), minimal residual disease (MRD) is the most relevant prognostic factor. Currently, its evaluation is based on quantitative real‐time polymerase chain reaction (Q‐RT‐PCR). Digital droplet PCR (ddPCR) was successfully applied to several haematological malignancies. We analyzed 98 samples from 40 Ph+ ALL cases, the majority enrolled in the GIMEMA LAL2116 trial: 10 diagnostic samples and 88 follow‐up samples, mostly focusing on positive non‐quantifiable (PNQ) or negative samples by Q‐RT‐PCR to investigate the value of ddPCR for MRD monitoring. DdPCR BCR/ABL1 assay showed good sensitivity and accuracy to detect low levels of transcripts, with a high rate of reproducibility. The analysis of PNQ or negative cases by Q‐RT‐PCR revealed that ddPCR increased the proportion of quantifiable samples (p < 0.0001). Indeed, 29/54 PNQ samples (53.7%) proved positive and quantifiable by ddPCR, whereas 13 (24.1%) were confirmed as PNQ by ddPCR and 12 (22.2%) proved negative. Among 24 Q‐RT‐PCR‐negative samples, 13 (54.1%) were confirmed negative, four (16.7%) resulted PNQ and seven (29.2%) proved positive and quantifiable by ddPCR. Four of 5 patients, evaluated at different time points, who were negative by Q‐RT‐PCR and positive by ddPCR experienced a relapse. DdPCR appears useful for MRD monitoring in adult Ph+ ALL.

Indeed, MRD is regarded as the strongest independent prognostic factor both in Ph+ and Ph− ALL. In the latter, MRD can be detected by multiparametric flow cytometry and molecular methods, such as PCR amplification-based methods that use leukaemia-specific (fusion gene transcripts) or patient-specific (immunoglobulin/T-cell receptor (IG/TR) gene rearrangements) molecular markers, [19][20][21] which represent the gold-standard technique for MRD assessment. Recently, in addition to these 'conventional methods', new techniques, namely digital droplet PCR (ddPCR) and next generation sequencing (NGS), [22][23][24][25][26] have been explored, showing an overall higher sensibility in anticipating a relapse.
DdPCR, a third generation PCR, might represent a valid alternative to Q-RT-PCR also for BCR/ABL1 quantification: it is based on a water-oil emulsion which determines the parcellization of the sample into at least 20,000 droplets and then PCR amplification is carried out within each droplet. This technique is highly sensitive and accurate, and it does not require a reference curve; furthermore, it has affordable costs and an easy interpretation of results. ddPCR improves the limit of detection (LOD) and quantification 27-30 of Q-RT-PCR. While this assay has proven valuable compared to Q-RT-PCR in lymphoma and Ph− ALL, its role has been scarcely investigated in Ph+ ALL. In this study, we compared Q-RT-PCR and ddPCR in Ph+ ALL, focusing on the set up of the BCR/ABL1 assay and on the evaluation of the specificity and sensitivity of the method.

| Determination of cDNA input
For cDNA synthesis, 1 μg of RNA was reverse-transcribed by using the SuperScript VILO cDNA Synthesis Kit (Invitrogen™), the same approach used to perform Q-RT-PCR. To define the optimal cDNA input, we tested three different amounts: 1, 2.5 and 5 μL.
In order to simulate a MRD condition, we diluted the cDNA of the diagnostic material in cDNA obtained from mononuclear cells (MNCs) from healthy donors to produce five serial dilutions (10 −1 , 10 −2 , 10 −3 , 10 −4 , 10 −5 ). MNC were obtained after-Ficoll density gradient separation. Each experiment was carried out including negative controls: for every condition tested, a no template control (NTC) and MNC from healthy donors was analyzed at least in triplicate.

| DdPCR CONDITIONS
To perform our experiments, we used the same primers and probes of Q-RT-PCR, according to the BIOMED1 recommendations 31,32 and we followed the Biorad protocols. Probes for target BCR/ABL1 p190 and p210 were labelled with FAM and BHQ1 reporters and probe for control gene was labelled with FAM and HEX reporters. Primers and probes were used at a final concentration of 900 nmol/L and 250 nmol/L, respectively. In each experiment we included negative controls, that is, MNCs, NTCs and a Ph− ALL diagnostic sample; as positive control samples, we used the first and the last dilution point of the plasmid curve, randomly distributed in the plate. The ABL1 gene was used as control gene to evaluate the quality of material and to calculate the ratio between the target and control gene. The reaction mixture, containing the cDNA, ddPCR Supermix (Bio-Rad), primers and probes were loaded into the DG8 cartridge wells, covered with the DG8 gasket, and loaded into the QX200 Droplets Generator together with 70 μL of droplet oil, in order to generate the droplets. During droplet generation, template molecules are distributed randomly into droplets. Due to the random nature of the partitioning, the fluorescence data after amplification are well fit by a Poisson distribution, thus, it can be used to determine the number of template molecules in a droplet given the fluorescence data. After droplets generation, they were transferred from the DG8 cartridge into a 96-well PCR plate which was sealed with a Bio-Rad pierceable foil heat seal and subsequently amplified through a Bio-Rad Thermal Cycler GeneAmp PCR System 9700 (Applied Biosystems). Thermal-ANSUINELLI ET AL.
-681 cycling conditions were the following: one cycle at 95°C for 10 min, 40 cycles at 94°C for 30 s, 40 cycles at 60°C for 1 min, one cycle at 98°C for 10 min and 4°C as holding temperature. Finally, the PCR plate was loaded into the QX200 Droplets Reader and data analyzed by using the QuantaSoft analysis Software version 1.7.4. According to the manufacturer's instructions, only analyses giving a number of droplets ≥9000/replicate were considered acceptable; the correct quantification of each experiment was carried out by setting by manual curation a threshold value with a sufficient distance from the background to ensure suitable sensitivity and specificity, as described by the manufacturer's application.
For ddPCR results, interpretation we followed the guidelines proposed within the EuroMRD Consortium 26 (supporting information).
Results were expressed as [copies/μl of the target gene]/ [copies/μl of the control gene] � 100. The concentrationindicated as the number of copies/μl-is provided by the Quanta-Soft software.

| Population of study
After having established the cDNA input and after having estab-

| Determination of cDNA input
Preliminary experiments showed that using 5 μl of undiluted cDNA were adequate to quantify all serial dilutions, except for 10 −5 ( Figure 1A). The number of copies/μl progressively decreased and they were equal to 43.7, 3.8, 0.24 and 0.03 for the 10 −1 , 10 −2 , 10 −3 and 10 −4 respectively, in agreement with the dilution performed. At, variance, when the starting material was 2.5 μl or 1 μl we were able to quantify up to a dilution of 10 −3 ( Figure 1B and 1C). These experiments indicate that only 5 μL of cDNA guarantees signal detection, even in cases with low levels of BCR/ABL1 transcript prompting us to use this volume in all subsequent experiments, except for the evaluation of the diagnostic samples, for which we used only 1 μl of cDNA, to avoid saturation of the assay.

| Evaluation of LOD, specificity and reproducibility
The experiments performed showed that at 1 � 10 −4 , all replicates scored positive (2/2), at 5 � 10 −5 we observed 6 out of 6 positive replicates, at 1 � 10 −5 we documented six out of 8 positive replicates; at variance, at 5 � 10 −6 only two replicates out of 12 were positive and, finally, at 1 � 10 −6 only 2 out of 14 replicates scored positive. These data showed that ddPCR allowed to reach a sensitivity of 5 � 10 −6 . Nonetheless, we defined the 1 � 10 −5 as the maximum sensitivity since up to this level 75% of replicates scored positive (6/8)-as opposed to what observed at lower concentrations ( Figure 1D and 1E).
Importantly, for both p190 and p210 transcripts, we obtained comparable results.
With regard to specificity, all NTC and healthy donor pools replicates tested always proved negative for BCR/ABL1 transcripts. Furthermore, we also documented a high reproducibility between the replicates of the diagnostic samples and dilutions, defining as reproducible all values which fell within the same logarithm.
Notably, independent experiments analyzing the same samples provided highly similar results between all replicates and runs, thus confirming the robustness of the method ( Figure 1F, 1G and 1H). This aspect has been further evaluated at lowest dilution points ( Figure S1).

| Comparison between Q-RT-PCR and ddPCR values of diagnostic samples
After having established the reaction parameters, specificity and sensitivity of the assay, we focused on the analysis of Ph+ ALL samples previously studied by Q-RT-PCR. First, we analyzed the diagnostic samples to assess the strength of the assay. In every plate, we included-as negative controls-MNC from a pool from healthy donors, NTCs and a Ph− ALL sample; we also included as positive controls three dilution points of the plasmid curve, which in addition to being a positive control are useful to assess the sensitivity achieved in each experiment. Each condition was run in triplicate.
Overall, we analyzed 10 samples at diagnosis and we found a high degree of concordance (Table 1) Table 2. Thus, the overall concordance between the two assays was 41%. Importantly, ddPCR allowed to recover the rate of quantifiable samples in 46% of cases (p < 0.0001).   showed that this approach allowed to recover informative data in some cases. At variance, to date, only one report has compared Q-RT-PCR and ddPCR in Ph+ ALL. 33 In our study, we aimed at developing a robust ddPCR assay for the correct quantification of both p190 and To our knowledge, our study provides for the first time a recommendations for the use of ddPCR analysis for adult BCR/ABL1+

| Comparison between Q-RT-PCR and ddPCR values of follow-up samples
ALL cases, showing that ddPCR allows to recover the quantifiability of MRD in a large proportion of patients, who otherwise would fall into a non-quantifiable range of Q-RT-PCR, and launch the bases for using this approach also in Ph+ ALL. Last, but not least, ddPCR is being set up also for the evaluation of detrimental mutations, that is, T315I mutations, for which a rapid switch in treatment is pivotal for avoiding full-blown relapses. In the forthcoming future, ddPCR might thus improve the overall clinical management of Ph+ ALL patients.