Whole Genome Sequencing Identifies Single Nucleotide Variants, Structural and Copy Number Abnormalities with a High Sensitivity in Hematological Neoplasm: An All in One Diagnostic Tool Tested on 3241 Cases

Background: Novel targeted treatment approaches for hematological malignancies require a comprehensive genetic characterization of patient samples. So far combinations of various techniques are used in different entities. As gross structural variants (SV) and copy number aberrations (CNA) as well as molecular mutations have to be assessed - in best case genome wide - to date no single technique is able to provide all information in a routine diagnostic setting. Whole genome sequencing (WGS) is a technology able to provide all this information in a single approach.

Aim: To evaluate whether WGS qualifies as a diagnostic tool in a routine setting.

Patients and Methods: 3241 bone marrow or blood samples from patients (pts) diagnosed with hematological neoplasm (including AML, ALL, MDS, CML, CLL) were evaluated by WGS. Samples had been sent for routine diagnostic work-up to our laboratory between 2005 and 2017. For WGS, 150bp paired-end sequences where generated on Illumina HiseqX and NovaSeq 6000 (Illumina, San Diego, CA). A mixture genomic DNA from multiple anonymous donors was used as normal controls. The median coverage was 104x (range: 47-196). Only cases with an estimated tumor fraction of at least 20% were included. WGS was validated against chromosome banding analysis (CBA), which was available in 2752 pts with an aberrant karyotype detected in 1513. For 334 pts genomic array data (GA) was available. CNA were called using GATK4 and SV using MANTA software accounting for missing matched-normal samples. For the validation of single nucleotide variants (SNV) and Indels we compared WGS data produced by BaseSpace WGS and Tumor/Normal app to variants classified as pathogenic during routine diagnostics using targeted amplicon sequencing (median coverage 1800x) in 70 genes known to be recurrently mutated including ASXL1, DNMT3A, RUNX1, SRSF2, TET2 , and TP53.

Results: In total 475 recurrent reciprocal structural rearrangements (38 different rearrangements including BCR-ABL1, PML-RARA, CBFB-MYH11, RUNX1-RUNX1T1, IGH-BCL2, IGH-MYC, IGH-CCND1) were identified by CBA. Of these 455 (96%) rearrangements were identified by WGS. Due to the significantly lower resolution of CBA compared to WGS and the fact that in complex karyotype the precise determination of CNA in CBA is not possible the comparison with respect to CNA between CBA and WGS was restricted to 843 cases with non-complex karyotype (<4 abnormalities). 289 trisomies, 48 monosomies and 464 recurrent deletions (del) (including del(5q), del(7q), del(11q), del(17p)) were identified by CBA. Of these WGS detected 210 (73%) trisomies and 42 (88%) monosomies. For 74 of the 79 trisomies undetected by WGS the percentage of cells harboring the respective trisomy was determined by interphase FISH and was in median 8%. FISH data was available for all 6 missed monosomies, median clone size was 14%. WGS identified 420/446 (81.5%) del detected by CBA. FISH data was available for 31/44 del missed by WGS. The median proportion of cells harboring the respective del was 11%. In order to test the CNA detection of WGS on a higher resolution level GA data from 334 cases was compared to WGS data. These included 135 cases with normal and 194 with aberrant karyotype in CBA (no CBA: 5), respectively. Comparing 18,337,602 positions 18,031,728 (98%) yielded the same result with both technologies with respect to gain, loss or normal, respectively.For SNV/Indel calls we investigated 2074 mutations in 1022 pts (harboring at least 1 pathogenic mutation (range 1 - 12)). 1892/2074 (91%) were concordant between amplicon sequencing and WGS. 132 from the missed 182 mutations had a variant allele frequency of <10%, which is on the verge of the limit of detection for 100x WGS data. Only 50 cases were missed due to low coverage or very complex alterations.

Conclusions: WGS can provide in an "all in one test" all relevant information required for classification and treatment decisions in hematological neoplasms with a high potential to substitute current genetic evaluation based on CBA, FISH and targeted mutation analysis. The next steps on the road towards a diagnostic tool are the validation of CNA, SV and SNV/Idel identified in addition to standard diagnostics and the determination of the coverage necessary to detect small clones relevant for patient care. Thus, a first step is taken towards a completely automated genotyping enabling a broad access to state of the art diagnostics.