Optimized Targeted NGS With Multiplex Competitive PCR Amplicon Libraries for Reliable Diagnostic Testing of Cancer FFPE Samples

Abstract
Background:  A recently described targeted NGS method using multiplex competitive PCR amplicon libraries (Blomquist et al, PLOS one, 2013) was used to quantify genes comprised by the Lung Cancer Diagnostic Test (LCDT), a three gene biomarker (MYC x E2F1/p21) intended to augment accuracy of lung cancer diagnosis in transthoracic fine needle aspirate (FNA) cell block FFPE samples. The results from targeted NGS were compared to those from qPCR and revealed excellent inter-platform reproducibility (R2= 0.99; slope = 0.95). In this method a) due to consumption of primers for abundantly expressed genes in earlier PCR cycles, there is normalization (convergence toward equivalent representation) of PCR products for lowly and highly expressed genes, and b) each target is measured relative to a known number of synthetic internal standard molecules which controls for variation in loading and/or amplification efficiency. In an effort to increase throughput, reduce reagent consumption, and further promote normalization through increased PCR efficiency of primers that perform poorly in multiplex, we implemented this method on the Fluidigm Access Array SystemTM. 
Methods:  A normal bronchial epithelial cell cDNA sample was mixed with an internal standard (IS) mixture comprising 6,000 molecules of IS for each of 70 respective transcript targets. After amplification with a mixture of primers targeting each of the targets we split the product into two aliquots. One aliquot was barcoded for sequencing (Library 1) and the second was amplified in a second round on the Access Array System with only one or two targets per reaction, products re-combined into a single library and barcoded (Library 2). Platform adaptors were added, and Libraries 1 and 2 were mixed then analyzed on Illumina HiSeq™. 
Results:  Of 70 gene targets, 32 demonstrated at least 4-fold and up to 6,000-fold increase in target sequence depth. Differences in loading and product amplification led to widely variant sequencing depth between the two libraries for each assay. However, measured relative to the known number of respective internal standard molecules (6,000 copies), there was high inter-library correspondence for each assay (R2=0.95 ; Slope = 0.95). Further, supplemental amplification on the Access Array System better normalized amplicon yield with a 2-log reduction in coverage breadth. 
Conclusions:  Generation of competitive PCR amplification libraries on the Access Array System a) increased sequence yield for most targets, thereby reducing potential stochastic sampling effects, and b) promoted normalization , leading to reduced consumption of sequencing space/sample. In addition, measurement of each target relative to a known number of internal standard molecules controlled variation resulting from low sample loading and/or inefficient amplification. We conclude that this is an optimized strategy for clinical molecular diagnostic testing, including in FFPE samples.