Next-generation sequencing and high DNA input identify previously missed measurable residual disease in peripheral blood of B-cell precursor acute lymphoblastic leukaemia

Abstract

In acute lymphoblastic leukaemia (ALL), measurable residual disease (MRD) is the most important prognostic factor for relapse.^{1, 2} Traditionally, MRD is quantified in bone marrow (BM) aspirates using real-time quantitative polymerase chain reaction (RQ-PCR) according to EuroMRD standards.³ The emergence of modern molecular methods has opened the possibility of MRD assessment from peripheral blood (PB), which causes less discomfort to patients and can therefore be performed more frequently, allowing a close monitoring of MRD levels. Furthermore, MRD assessment from PB could bypass technical challenges of BM aspiration, such as skewed or non-representative MRD values due to haemodilution and heterogenous distribution of leukaemic cells in different body compartments.^{4, 5} However, while comparable MRD values of BM and PB are observed for T-ALL,^{6, 7} in B-cell precursor ALL (BCP-ALL), MRD determination from PB is still a challenge. In both paediatric and adult cases, mean MRD levels in PB are consistently significantly lower compared to BM and frequently escape detection.^{6, 7} Additionally, paired samples exhibit a marked variability in the MRD level ratio.⁶

Nevertheless, BCP-ALL patients with MRD positivity in the BM likely also harbour circulating leukaemic cells, even when missed by traditional methods. The implementation of more sensitive assays may enhance MRD detection from PB in these cases. To target this question, we retrospectively selected paired BM-PB samples of patients with BCP-ALL and BM MRD positivity where conventional RQ-PCR of clonal immunoglobulin heavy chain (IGH) VJ complete/DJ incomplete gene rearrangements did not detect PB MRD and reanalysed them with heightened sensitivity by using increased DNA input and amplicon-based next-generation sequencing (NGS).

In total, 69 patients with pre/c-B-ALL (n = 57) or pro-B-ALL (n = 12) treated according to protocols of the German Multicentre Study Group on Adult ALL (GMALL) were included. All patients provided written informed consent for storage and use of the leftover material for medical research purposes. The Ethics Committee of the Medical Faculty at Christian-Albrechts-University of Kiel approved that there are no ethical or legal concerns about the conduct of the study (reference number D-402/21). Patient age ranged from 19 to 72 years with a median of 44 years. Of the analysed sample pairs, 52 were taken in first-line therapy and 17 during salvage treatment. IGH MRD markers were identified in the diagnostic samples using the EuroClonality amplicon-based NGS assay⁸ and the respective interpretation guidelines leading to a total of 80 IGH RQ-PCR assays (mean of 1.2 IGH assays/patient). RQ-PCR-based MRD quantification as part of the GMALL reference diagnostics was performed by standard RQ-PCR with 1.5 μg DNA in BM and PB, respectively, according to EuroMRD guidelines.³ In cases where more than one MRD marker was used per patient, the highest measured value was considered the cumulative MRD value. In all included cases, BM was MRD positive (41 with quantifiable MRD positivity and 28 with positivity below quantitative range) and PB was MRD negative using standard RQ-PCR with a sensitivity of at least 10⁻⁴. BM samples with a non-quantifiable MRD positivity were only included if subsequent BM samples were positive within the quantifiable range.

For MRD re-evaluation, PB samples were reanalysed using the following approaches: (1) For high DNA input RQ-PCR, a total 6.6 μg DNA (corresponding to the DNA content of one million cells) was used and analysed in 10 separate reactions. (2) Amplicon-based NGS was performed with standard DNA input using 1.5 μg DNA, as well as with (3) high DNA input using 6.6 μg DNA in four different reactions. NGS MRD analyses were performed with the EuroClonality amplicon-based NGS assays.⁸ A 100-fold coverage of the spike-in rearrangements in the respective library or a minimum read amount of 300 000 was reached for all samples. Identification and quantification of leukaemia-derived clonotypes were done using ARResT/Interrogate.^{9, 10} In the NGS analyses, MRD values were termed not quantifiable when the percentage of positive marker reads corresponded to less than one cell equivalent (4 × 10⁻⁶ for standard input NGS, 10⁻⁶ for high input NGS).

To assess the effect of high DNA input alone (approach 1), RQ-PCR from PB was performed for 57 of the 69 patients, for which sufficient leftover DNA was available. With this approach, at least one positive PCR signal was detected in 18/57 patients (32%; Figure 1A,B). None of the cases qualified as quantifiable MRD positive according to EuroMRD guidelines, which would require all 10 replicates to produce a positive signal within the specificity range.

To assess the effect of NGS, 69 PB samples were NGS tested with standard (approach 2) and high DNA input (approach 3). With standard DNA input, MRD was detected in 16/69 patients (23%) (Figure 1A). In the high DNA input approach, MRD was detected in 33/69 patients (48%) (Figure 1A,B). All quantifiable MRD values were below 10⁻⁴, and in each case lower than in the corresponding BM sample.

In these direct comparisons, only those IGH VJ/DJ markers that were also used for RQ-PCR were tracked by NGS. However, an important advantage of NGS is the ability to monitor all potential markers of a respective rearrangement type. For 37 of the 69 patients (54%), at least one additional valid IGH VJ/DJ marker was detected at initial diagnosis, which had not been used for routine RQ-PCR. In total, a mean of 2.3 valid IGH markers per patient was identified by NGS at initial diagnosis, but only a mean of 1.2 IGH markers per patient was used for MRD assessment by RQ-PCR. Considering all possible IGH markers improved NGS MRD detection in PB (Figure 1A) and identified MRD in three additional PB samples using high DNA input (Figures 1A and 2). Overall, MRD from PB was detected in 36/69 patients (52%) using this approach.

Monitoring all potential IGH markers with NGS, therefore, increases the chance of detecting MRD. Although the majority of IGH clonotypes detected in the follow-up, PB samples derived from highly abundant diagnostic IGH clonotypes (52% of MRD-positive IGH markers were present in more than 50% of the target reads in the diagnostic sample) and also the lower abundant clonotypes contributed to MRD detection in PB (11% of MRD-positive IGH markers were present in only 5–10% of the target reads at initial diagnosis; Figure S1).

However, the strategy of tracing not only the most abundant IGH clonotypes but also IGH rearrangements with lower abundance must be applied with caution. Low abundant clonotypes may not represent the leukaemia but minor accompanying B-cell clones leading to incorrect MRD results. As a precaution in accordance with the EuroClonality NGS guidelines, only markers with an abundance of >5% reads should be used, which could be clearly assigned to the leukaemia in a validation study.⁸ Nevertheless, the presence of monoclonal B-cell lymphocytosis (MBL) clones in individual patients cannot be ruled out. The risk of monoclonal B-cell expansions increases with patient age, with MBL being reported in up to 10% of healthy adults aged 50–59 years.¹¹ Therefore, before implementation of this strategy into clinical routine, strict criteria have to be established for which markers can be used for MRD assessment. This is currently done within the EuroMRD Consortium.

It should be noted that the distinction between quantified and non-quantified MRD values also stems from the varying methods of quantification. While the guidelines for RQ-PCR evaluation have been standardized for years, there are still no established definitions for quantitative range or non-quantifiable positivity in NGS.

Overall, our results indicate that PB can be valuable for early MRD detection in BCP-ALL when adjusting traditional MRD assessment strategies. Still, MRD values in PB were significantly lower compared to BM and NGS with high DNA input only allowed MRD detection in half of the investigated cases, although it is very likely that tumour cells circulate in virtually all BM MRD-positive patients. Potentially, the frequency of leukaemia cells in the blood in these cases is so low that the sample volume limits MRD detection, as the denominator of the detection limit is the number of cell equivalents tested. Even with adjusted methods, PB therefore cannot replace BM for MRD assessment in BCP-ALL yet and BM monitoring should be sustained at critical stages of treatment protocols and during long-term follow-up, as relying on PB monitoring alone could delay the detection of an early relapse. However, in cases where a BM sample is not available, high DNA input and NGS should be considered to enhance the assay sensitivity and, consequently, increase the probability of detecting low-level MRD in PB.

MB, NG, AS and MK designed the study, MB, MK, BK and SBu collected and selected patient data. SBu, SBe and BF performed the experiments. MB, CDB, MK, SBu, SBe and GC interpreted the data. CD, FD, BK and ND performed the bioinformatic data analysis. NG provided patient information. SBe wrote the first draft of the manuscript. All authors approved the final version of the manuscript.

CB is consulting Astellas, BMS, AstraZeneca, Amgen, Jazz Pharmaceuticals, Gilead and Jannsen. NG received research funding from Novartis and research funding/Honoria from Pfizer, Jazz Pharmaceuticals, Incyte, Clinigen, Servier, Gilead and Amgen and honoraria from Autolus. MB is a member of Incytes and Amgens Board of Directors or advisory committees and the Speakers Bureaus of BD, Janssen, Pfizer and Amgen and received research funding from Amgen, Affimed and Regeneron. For the remaining authors, no relevant conflicts of interest were declared.