Library Preparation Is the Major Factor Affecting Differences in Results of Immunoglobulin Gene Rearrangements Detection on Two Major Next-Generation Sequencing Platforms

Minimal residual disease (MRD) assessment via next generation sequencing (NGS) of immunoglobulin (Ig) and T-cell receptor (TR) gene rearrangements for lymphoid malignancies is currently under extensive development. NGS MRD has a potential to overcome the limitations of current techniques; laboriousness and difficult interpretation of qPCR for Ig/TR and low sensitivity of flow cytometry. However, amplicon-based NGS MRD has potential pitfalls that have to be addressed before it can be safely introduced for clinical decision making. Multi-center concordance in the experimental setting, quality control and interpretation of the results need to be achieved in order to surpass the advantages of qPCR, which is currently rigorously standardized within the EuroMRD consortium.

Our aim was to test the stability and reproducibility of an optimized Ig heavy chain (IGH) based NGS approach for MRD assessment in a multi-center setting within the EuroClonality NGS Consortium on two different sequencing platforms.

A one-step PCR library preparation approach was tested in seven institutions (Kiel, Salamanca, Milano, Bristol, London, Prague, Torino). Serial dilutions (10^-1 to 10^-5) of diagnostic DNA into polyclonal DNA as well as follow-up samples of 30 B-cell precursor ALLs with known complete IGH rearrangements were sequenced on the MiSeq. Serial dilutions of five different diagnostic ALL samples and libraries from polyclonal control were sequenced in parallel on both the MiSeq and Ion Torrent platforms. All samples were spiked with pre-defined copy numbers of five reference IGH sequences as a calibrator. FR2 primers, harboring platform-specific sequencing adapters, were used during the one-step PCR with 500ng of DNA per sample (75,000 copies). Negative and positive controls (27 pooled B-cell lines) were used for testing assay stability and reproducibility among the labs. Purpose-built bioinformatics methods were applied to analyze data. MRD results were compared to results of EuroMRD-based qPCR results.

A total of 333 libraries were sequenced in 29 deep sequencing runs producing 194 million reads. The IGH gene rearrangements of all 27 pooled positive B-cell line controls were identified in all centers. NGS MRD analysis in 116 ALL follow-up samples revealed MRD positivity in 69/116 samples vs. 66/116 samples in qPCR, with discrepancies concerning samples with low MRD (R²=0.81). The dilution experiments gave similar results for both platforms, with a minimum sensitivity of 10^-4 (as currently required by most treatment protocols using qPCR) for all tested assays. The correlation between MRD levels obtained by the two NGS platforms was good (R²=0.84). Ratios of reads containing reference IGH sequences were highly consistent in intra- and inter-laboratory analyses, independent of the total number of reads in the sample. When comparing platforms, in 10^-1 dilution samples sequenced on MiSeq the ratio of reads harboring reference sequences was 2.1 to 2.7 times lower than in remaining dilutions, while on the Ion Torrent it was only 0.9 to 1.3 times, reflecting the competition with the leukemic clone. The correlation of the amounts of spiked-in sequences with the representation of reads harboring these sequences was slightly better for the Ion Torrent (R²=0.88) than for the MiSeq (R²=0.79). Amplification efficiency of each primer was checked by analyzing libraries from healthy polyclonal control. All primer sequences were present in all samples on both platforms, however, the differences between four libraries prepared from the same sample sequenced on the MiSeq were 2.6 times higher than in one library from this sample sequenced in five replicates on the Ion Torrent.

The newly developed IGH assay shows robust intra and inter-laboratory reproducibility, which is the first step towards the safe use of this new MRD technique in a multi-center setting. The distribution of reference sequences and sequences of primers confirmed that the main source of differences between platform strategies is the library preparation and not the platform itself. Using the same amount of DNA, the sensitivity of the method is similar to qPCR. The performance and costs of the assay are similar for both the MiSeq and Ion Torrent. MRD analysis via NGS has therefore a great potential to replace qPCR as the gold standard for MRD-guided therapy in ALL, provided that thorough standardization can be achieved.

Support: NV15-30626A, GBP302/12/G101.