Helpful Criteria When Implementing NGS Panels in Childhood Lymphoblastic Leukemia

The development of Next-Generation Sequencing (NGS) has provided useful diagnostic, prognostic, and therapeutic strategies for individualized management of B-cell precursor acute lymphoblastic leukemia (BCP-ALL) patients. Consequently, NGS is rapidly being established in clinical practice. However, the technology’s complexity, bioinformatics analysis, and the different available options difficult a broad consensus between different laboratories in its daily routine introduction. This collaborative study among Spanish centers was aimed to assess the feasibility, pros, and cons of our customized panel and other commercial alternatives of NGS-targeted approaches. The custom panel was tested in three different sequencing centers. We used the same samples to assess other commercial panels (OncomineTM Childhood Cancer Research Assay; Archer®FusionPlex® ALL, and Human Comprehensive Cancer Panel GeneRead Panel v2®). Overall, the panels showed a good performance in different centers and platforms, but each NGS approach presented some issues, as well as pros and cons. Moreover, a previous consensus on the analysis and reporting following international guidelines would be preferable to improve the concordance in results among centers. Our study shows the challenges posed by NGS methodology and the need to consider several aspects of the chosen NGS-targeted approach and reach a consensus before implementing it in daily practice.


Mapping genome reference
The use of different reference genomes may lead to differences in the clinical impact of the variants Choose carefully which transcript set and software will be used for annotation

Variant prioritization and annotation
Not established criteria. Differences in bioinformatics pipelines and prioritization of the variants may result in discrepancies It is important to standardize bioinformatics algorithms and the annotation

Absence of validation of the variants may produce misleading results
Visualization of the aligned sequencing data and confirmation by Sanger sequencing when possible (VAF >15%)

Supplementary Figure 1.
Overview of the study: workflow of samples, panels and sequencing platforms.

Supplementary Figure 2.
Applicability of different targeted-gene panels for BCP-ALL.

BCP-ALL custom panel development
An exhaustive literature search was performed by clinical and molecular experts in ALL from the different centers, selecting those genes related to diagnostic, prognostic, or therapeutic interest recurrently (more than 1% of patients) identified in BCP-ALL.

Library preparation and data analysis
Library preparation and subsequent sequencing runs were processed independently at each center, following the standardized workflows for each panel and instrument. Custom BCP-ALL libraries were generated and sequenced in 3 centers (centers 1, 2, and 3) on the Illumina NextSeq500/550 Instrument, using a NextSeq500/550 High Output reagent kit v2 (300 cycles). The

FusionPlex ALL® kit was sequenced on a MiSeq instrument (Illumina) (center 2) and on an Ion
Torrent Personal Genome Machine (PGM) (center 4). The OCCRA was sequenced on an S5 sequencer at center 4. Finally, the Human Comprehensive Cancer GeneRead Panel v2® was sequenced on a NextSeq500/550 in center 3. Data analysis of the BCP-ALL custom panel was performed using in-house pipelines. For commercial panels, we used software provided by each company.

1) BCP-ALL custom panel
The BCP-ALL custom panel was assessed following the Nextera® Rapid Capture Enrichment protocol (Illumina). Based on amplicon concentration, equal amounts of amplicons were pooled to ensure that the optimal cluster density was achieved. After library pooling, two rounds of hybridization were performed, by binding targeted regions of the DNA with capture probes.
Finally, size control was carried out with a 2100 Bioanalyzer (Agilent Technologies, California, USA) to ensure the library's good quality.

Bioinformatics pipeline for samples tested in Centre 1-HSJD
The variant detection protocol was developed at the Bioinformatics Unit from the Molecular Genetics Department at HSJD. The variant calling pipeline consisted of the following steps: first, read quality was assessed (FastQC v.0.11.5), low-quality adapters and reads were removed (Cutadapt v1.13) and remaining reads were aligned to the human reference genome

Bioinformatic pipeline for samples tested in centers 2 & 3 (HUS and HNJ-ISCIII)
The first step was the bcl file generation, the base-calling which occurred within the sequencing instruments. After base-calling, the raw reads were generated in a FASTQ file format. Moreover, base-calling software supplied a quality metric for each base (Q) that indicates base-calling error probability. Usually, a quality threshold is set in a Q30, which means that the probability of the base being incorrectly called is 1 in 1000. The FASTQ files were analyzed by an in-house pipeline:

2) Archer® FusionPlex® ALL kit
For library generation, two different protocols were applied following the manufacturer´s specification depending on the sequencing platform (Illumina or Ion Torrent) used. In both cases, 250 ng of total input RNA was used. Libraries were quantified with the KAPA Universal Library Quantification Kit (Roche). Equimolecular libraries were pooled, amplified, and sequenced following standard procedures for Illumina MiSeq or Ion Torrent PGM sequencing. Results were analyzed on the Archer Analysis Software 6.0.

3) Oncomine TM Childhood Cancer Research Assay
Amplicon-based libraries for mutation and CNV detection were generated from 10ng of DNA  10 .

4) Human Comprehensive Cancer GeneRead Panel v2® (Qiagen)
Briefly, the amplicon libraries were prepared from 10ng of DNA according to the manufacturer's specifications and sequenced in a NextSeq500 sequencer (Illumina) in 2x150bp run format, to obtain an average depth of 1,000x reads. Raw data analysis for the identification of mutations was carried out with the GeneRead software, exon enrichment panel data analysis (Qiagen). Besides the automated analysis, the aligned files were visually reviewed using Integrative Genomics Viewer 2.5 (Broad Institute and the Regents of the University of California) 10 .

Sanger sequencing
Sanger sequencing was performed in patient-matched samples to confirm all the identified variants with ≥15-20% VAF. Different primer sets were designed using PRIMER3plus software Retrotranscription kit (Applied Biosystems). Real-time PCR was performed as previously described 11 using GUS as a control gene. Primer sequences for LMO-RIC1 fusion were LMO1-E2-F GCTCCACCCTCTACACCAAG and RIC1-E2-R CAGGAGTCTGGCCATCTGAG.

Confirmation of somatic nature of variants
To confirm the somatic nature of the variants identified by NGS, we used patient-matched bone marrow samples in complete morphological complete remission (CR) with negative or very low (<0.05%) measurable minimal residual disease (MRD), as assessed by 8-color flow cytometry or by quantitative PCR.

Confirmation of CNVs
CNVs were screened by MLPA using the SALSA MLPA P335 ALL-IKZF1, P181 and P182 kits (MRC Holland, Amsterdam, The Netherlands) according to the manufacturer's instructions.