High-throughput technologies, in particular WGS, have shown huge potential to identify the genomic aberrations that define genetic subgroups in ALL  and are increasingly being introduced in the diagnostic setting for hematological malignancies [10–15]. The present study was designed to explore the possibility to extend the diagnostic utility of WGS beyond genetic characterization. Since MRD assessment is the most important parameter in risk adapted treatment protocols, we investigated the feasibility to use WGS to identify genomic structural rearrangements that involve breakage and reunion of the genome and use the novel sequences generated at the SV junctions as leukemia-specific unique markers. We hypothesized that genes recurrently rearranged in ALL are likely to be important for leukemic evolution and/or leukemia maintenance and would thus be preserved during the course of the disease and therefore constitute suitable targets to monitor treatment response.
The patients in the study were selected to challenge WGS; none of the BCP-ALL harbored the recurrent rearrangements that result in fusion genes mandatory to investigate in the NOPHO 2008 treatment protocol. Nonetheless, WGS identified suitable targets in all BCP-ALL patients and in the two T-ALL patients included. WGS enabled the detailed characterization of the breakpoints and directly provided the leukemia-specific junction sequences. Some of the chromosomal rearrangements first selected had occurred at genomic sites with highly repetitive sequences and were unsuitable as targets. This was particularly evident in patient 3, a BCP-ALL with high hyperdiploidy, where all structural events but one, had occurred in such regions. This phenomenon is a known feature of the human genome: SV are enriched in regions that contain repetitive regions. Patient 3 was the only patient where the one suitable clonal marker was not in a genomic region reported to be involved in the pathogenesis of ALL. In order to boost the sensitivity, a single assay with double probes was designed and resulted in a LoQ and LoD comparable with the other patients in the study.
The current “gold standard” for molecular monitoring of residual disease in ALL is to measure clonal IG/TR rearrangements with allele-specific quantitative PCR, a demanding method that has been extensively standardized by the EuroMRD consortium. In addition, this method requires a substantial amount of BM DNA and was the reason why BM DNA samples were not available from all time points in the patients with T-ALL. We used the criteria recommended by EuroMRD to test the performance of the SV ddPCR assays on BM. Four out of six patients LoQ reached 10− 5, and in the remaining two patients the LoQ was 10− 4. All patients had at least one assay able to detect one leukemic cell in 100,000 normal cells and the limiting factor for the sensitivity was the input DNA. Thus, the assays were at least as good as and potentially superior to IG/TR-PCR in regards of specificity, sensitivity and range of quantifiability and had no background amplification from the germline. They also showed good concordance with routine MRD methods, although for a few samples ddPCR indicated higher MRD values. This was particularly evident for patient 4, where ddPCR showed MRD above the routine methods at both EoI and CB1 timepoints, despite the fact that this patient's assays had the lowest LoD/LoQ.
All targets selected were present in a high proportion of leukemic blasts, and when possible recurrent genetic aberrations, and thus likely represented truncal events. The use of several targets per patient minimizes the risk that a genetic event will be lost during clonal evolution. In order to maximize the accuracy and sensitivity when few copies of the targets are present the assays were successfully multiplexed for all patients except in patient 3 where the high hyperdiploid clone only rendered one target. In all, ddPCR assays based on SVs were highly specific which enabled precise MRD quantification with very little optimization and no need for standard curves. Hence, this WGS-based approach enabled simultaneous genetic characterization of the leukemic blasts with regards to recurrent genetic markers and identification of specific molecular targets to monitor therapy response.
The successful clinical implementation of WGS for the diagnostics of germline conditions has paved the way for introduction of WGS in the diagnostic setting of malignancies[39, 40]. In pioneering studies, WGS, often combined with WTS, has been evaluated as a diagnostic tool in acute leukemia with promising results. Compared with standard of care multi-testing, WGS/WTS performed equally well or better in identifying clinically relevant genetic aberrations in acute leukemia patients and changed the risk classification in a proportion of cases [15, 41]. Within Genomic Medicine Sweden, a national study evaluating WGS and WTS in acute leukemia diagnostics is ongoing, with the ultimate aim to replace standard of care methods  (Berglund et al., unpublished results). With decreasing sequencing costs, these type of studies will hopefully provide the required impetus to gradually implement these powerful techniques in the diagnostic setting.
A test strategy for personalized treatment protocols based on WGS is often considered premature for clinical routine due to the still relatively time-consuming methods. However, the strategy outlined in Fig. 1 is suitable for MRD testing since there is enough time to perform and analyze the WGS data and to design and evaluate different ddPCR targets before MRD monitoring is of clinical relevance.
The results also show that with our approach, MRD can be successfully measured in cfDNA from a blood sample with a sensitivity comparable to that in BM. To our knowledge, this is the first time that MRD has been monitored in cfDNA using patient-specific targets determined from WGS on BM in ALL. Using targeted ddPCR, we observed extremely high levels of ctDNA in plasma samples at diagnosis and the kinetics showed a progressive decline during therapy. However, total cfDNA is unspecific as a MRD marker since it can be affected by other medical conditions as seen in three of our patients. On the other hand, sensitive detection and quantification of the patient-specific SV targets were possible. The selected targets decreased in plasma during the induction and initial consolidation, reflecting the leukemic burden.
The plasma markers rose from being undetectable in plasma to very high levels 2 months before overt recurrence in the only patient in this series that suffered a relapse. This result is in line with the results from other studies that suggest that ctDNA may be a sensitive tool to monitor therapy response in hematological malignancies[43, 44] and that ctDNA not only reflects the circulating malignant cells but rather represents the entire disease burden . Although residual leukemic cells can be detected in peripheral blood, they do not consistently represent the residual disease in the bone marrow and reliable MRD monitoring today calls for bone marrow sampling. However, monitoring leukemia burden through plasma ctDNA in ALL could potentially be more representative than BM aspiration and has the additional advantage of being less invasive than BM sampling which, in children, is performed under general anesthesia.
Of note, the targeted SVs could be detected in CSF from half of the patients and 1 ml CSF was sufficient. Our results suggest that the presence of ctDNA in CSF is not solely the consequence of leakage from the plasma, but rather correlates to the presence of malignant cells in the CSF. There were no traumatic punctures in the study, and as the CSF was collected in STRECK tubes containing cell stabilizing agents any leukemic cells from minimal puncture bleedings are not likely to have caused false positives. The current methods for detection of CNS involvement based on morphology miss a large number of patients with leukemic cells in the CNS and therefore all children are treated with intrathecal prophylaxis[7, 8]. Studies on FCM to improve CNS diagnostics are ongoing and the potential contribution of ctDNA analysis on CSF also needs to be evaluated in a larger cohort.
A potential limitation of our approach is that SVs can occur in genomic regions harboring repetitive sequences as was the case with patient 3 with a had high hyperdiploid karyotype. Despite detection of multiple SVs in the WGS data only one suitable ddPCR target was identified. Thus, in patients where the class defining aberration is an aneuploidy it may be difficult to find several suitable targets. A limitation for all molecular methods is the turn-around time; assays need to be available by end of induction to be useful for clinical decision-making. As WGS provides data for both genetic classification and SV target identification in one seamless workflow, and SV ddPCR requires minimal optimization, this approach is time efficient.