4. VALIDATION STUDY OF A NGS-BASED PGT-A WORKFLOW FOR MOSAICISM AND SEGMENTAL COPY NUMBER VARIATION

Abstract

Introduction Recently, with the introduction of Next Generation Sequencing (NGS), preimplantation genetics test for aneuploidy (PGT-A) has been greatly advanced. While copy number variation (CNV) calling of full chromosome aberrations has been well developed by many NGS-based PGT-A workflows, the calling algorithms for mosaicism and segmental CNVs are still not well established. In this study, we assess the specificity and accuracy of a newly developing PGT-A workflow for calling mosaicism and segmental CNV. Material & methods Thirteen Coriell cell lines with known karyotypes were analyzed in this study. To test the accuracy of mosaic calling, 5 cells were precisely picked under dissection microscope and processed using the protocol for the Ion ReproSeq™ PGS Kits for the Ion GeneStudio S5™ System (Thermo Fisher Scientific). Mosaic samples with mosaicism of 0%, 20%, 40%, 60%, 80% and 100% were generated by combining defined number of cells from two cell lines. For evaluating full and segmental chromosome CNV calling, 4-5 cells of each cell lines were sampled. The sequence data was analyzed using decimal ploidy value copy number calling (mosaic) algorithm in Ion Reporter v5.10. This algorithm was combined in a workflow with either default (2Mb), 1Mb or 0.5Mb tile size reference baseline. In total, 144 samples were analyzed with 4 replications for each mosaic sample and 3 replications for all the other samples. Results To evaluate specificity and sensitivity of mosaic calling, two sets of mosaic samples were generated by mixing (i) 46,XY with 47,XX,+21, and (ii) 46,XY,del(4p) with 47,XY,+13 cell lines. Mosaic profiles of Chr13 closely matched expected values at all tested combinations, and in contrast, those of Chr21 showed expected mosaic values at 60% and 80% but not 20% and 40% groups. Meanwhile, mosaic calling workflow with 0.5Mb baseline could not detect 20% segmental mosaic loss of del(4p)27Mb, and displayed 9%, 15% and 17 % overestimation of mosaicism at 40%, 60% and 80% groups, respectively. By choosing different tile size reference baselines while keeping other parameters at default values, we tested the ability of the workflow to detect full and segmental chromosome CNV. With default baseline, trisomy of Chr13, Chr18, Chr21 and ChrX, monosomy of Chr21, dup(3q)99.1Mb, del(5p)32Mb, del(4p)27Mb and dup(6p)20.97Mb could be correctly called. With 0.5Mb baseline, dup(22q)14.75Mb and dup(9p)11Mb but not dup(16p) 3.6Mb, del(5q) 2.3Mb, dup(6p) 2.19Mb, del(9p) 1.81Mb and del(14q) 1.18 Mb, could be consistently reported with expected size and breakpoints. Of note, the workflow with 0.5Mb baseline identified some unexpected segmental CNV events of various sizes and locations, such as dup(1q)4.6Mb, del(15q)2.6Mb, del(4p)1.4Mb, del(17q)1.4Mb, dup(2q)0.9Mb, del(22q)0.8Mb and del(15q)0.7Mb. Conclusion Accuracy and sensitivity of both mosaic calling and segmental CNV callings may vary among different chromosomes, different sizes and regions of the chromosomes. Segmental mosaic calling is less sensitive and accurate than full chromosome mosaic calling. Choosing higher resolution baseline could enhance sensitivity of small segmental CNV calling but reduce specificity. Further modification and optimization of the calling algorithm for mosaicism and small segmental CNV events are needed. This study provides useful information for establishing NGS-based PGT-A guidelines.