The Use of NGS for Molecular Investigations in Leukemia Adds Important Biological, Clinical and Prognostic Information Due to Its Higher Sensitivity in Comparison to the Gold Standard Sanger Sequencing

Introduction: For many decades Sanger sequencing (SS) has been the gold standard for the investigation of molecular mutations in hematology. In recent years, next-generation-sequencing (NGS) has been increasingly used for research and is now also implemented into routine-diagnostic approaches. Usually the sensitivity of SS has been set as 10-20% of malignant cells per sample to allow judgment on mutation status. NGS demonstrates in several studies a higher sensitivity with standard routine assays in the range of 1-3% allele frequency. The possible gain of information for diagnosis, classification, prognostication and even minimal residual disease needs to be established. This also has to take into account the increasing number of variants of unknown significance picked up today.

Aim: Estimate the additional information gained by NGS in comparison to SS due to different sensitivities in a cohort of 16,570 patients and 68 genes investigated at different time points of the respective diseases.

Methods: We here demonstrate our data on 68 genes all rated positive by NGS (454, Branford, CT or Illumina, San Diego CA) for variations at variant allele frequencies (VAF) below 10%. Overall 21,180 samples sent to our laboratory between 2007 and 2016 were accessed in 173,256 gene analyses.

Results: Among the 173,256 analyses 22,834 (13%) variants were called, showing a large number of 21.9% (5,009/22,834) below 10% VAF. The most frequently mutated gene below a threshold of 10% was CXCR4 with 62.1% (36/58) of detected variants (in total 312 analyses were performed), followed by MYD88 (54.4%, 149/274, 882 analyses), BRAF (51.7%, 15/29, 3,088 analyses), FLT3-TKD (47.9%, 35/69, 2,638 analyses) and STAT3 (40%, 4/9, 368 analyses). Most of these mutations associate with lymphoid malignancies, pointing out the importance of sensitive methods to detect genetic variations also in entities where the size of the aberrant clone in peripheral blood or bone marrow aspirate frequently is rather small. Furthermore, FLT3-TKD mutations occur also in disease course of AML, emphasizing the need to catch progressing cell clones at a low mutation burden.

Focusing on prognostically important or therapeutically relevant genes showed VAF <10% in TP53 (33%, 929/2,813, 10,629 analyses), KIT (33.7%, 60/178, 4,606 analyses) and KRAS (34.3%, 173/505, 6,208 analyses).

Validity of the mutations determined were checked against common databases, including COSMIC (v74), ClinVar and dbSNP (v144). In case of TP53, also the IARC TP53 database (r17) was taken into consideration for assessing the variants. Additionally the impact of the variants on the protein structure was determined using prediction software tools (PolyPhen and SIFT). The interpretation of variants remains an issue irrespective of VAF. 213/5,001 (4.3%) detected variants <10% VAF could not be classified based on public data bases.

Sensitive methods are a hallmark need for monitoring of minimal residual disease (MRD). We therefore rescreened in another study samples (with known mutations in former sampling points) reported to be negative using SS. In 187/2,601 (7%) of these samples we found a low level mutation (<10%) using NGS (454, Branford, CT), resulting in the absence of the formerly presumed complete molecular remission. In detail, we were particularly interested in clinically relevant genes and focused on mutations in TP53, RUNX1, ASXL1 and NOTCH1 and found that these genes were mutated in 48/187, 30/187, 23/187 and 8/187 cases, respectively. Furthermore, NGS allows an absolute quantification as well as a value for the respective quantification limit that is not given for SS.

Conclusion: 1) In 68 genes investigated 173,256 times we found in 5,009/22,834 (21.9%) true low level variants <10% VAF, which are not detectable by current gold standard technology. 2) MRD levels between 0.1 and 10% by NGS were found in 187/2,601 investigated genes (7%). 3) The interpretation of variants is even more challenging considering the clinical impact of variants with <10% VAF.