Comparing Library Preparation Technologies for FFPE-Derived RNA

For Research Use Only. Not for use in diagnostic procedures. © 2024 Watchmaker Genomics. All trademarks are property of their respective owners. A comparative analysis of library preparation technologies for RNA sequencing from FFPE samples Giulia Corbet, Jen Pavlica, Kailee Reed, Thomas Harrison, Travis Sanders, David Gelagay, Abstract #2940 Josh Haimes, Ariele Hanek, Kristina Giorda, Brian Kudlow – Watchmaker Genomics Introduction Formalin-fixed, paraffin-embedded (FFPE) samples are an invaluable resource in the oncology space, providing access to a vast library of archived diseased tissue samples paired with relevant donor information. Despite the broad utility of these samples, RNA extracted from FFPE tissue is typically difficult to process due to the presence of residual crosslinks and its degraded nature. Further, these samples often vary widely in performance as the fixation process, block age, block storage, and extraction method can impart large impacts on resulting template quality. As a result, robust and reproducible RNA sequencing from FFPE-derived RNA remains a challenge with unpredictable and high failure rates. We evaluated four commercially available Whole Transcriptome Analysis (WTA) solutions to determine which performed best with this challenging sample type with respect to library complexity, inter-input and intra-sample concordance, and overall reproducibility. Matched fresh frozen and FFPE liver samples were used such that the fresh frozen data set serves as a comparative truth set for the FFPE data. Figure 1. Whole transcriptome library prep workflow of four commercially available kits. The same fundamental steps are present in all workflows. Some workflows are more streamlined — combining steps, shortening incubations, and bypassing the need for additional SPRI cleanups. Only the Watchmaker chemistry includes an optional FFPE treatment step. Workflow times for the different kits vary, ranging from 4.5 to 8 hours. Materials and Methods Samples: RNA samples from matched fresh frozen and FFPE liver tissues were purchased from Biochain. FFPE RNA quality was assessed by Agilent Tapestation to obtain both a DV200 (64%) and RIN (1.2) score. Library preparation and sequencing: 100 ng and 10 ng of total RNA derived from FFPE or matched fresh frozen tissue was used as input into depletion following each kit’s available protocol. Following depletion, library preparation was performed according to manufacturer instructions. The Watchmaker libraries followed the stubby adapter library preparation recommendations. Libraries were pooled to be equimolar and sequenced on an Illumina NextSeq2000 using the P2, 200 cycle chemistry with 75 nt paired-end reads. Data analysis: For analysis, all libraries were subsampled to 10.8M read pairs per library. Standard RNA-seq metrics such as strandedness, duplication rate, and sequence composition were generated using Picard Tools RNASeqMetrics. Unique genes in each library were detected using featureCounts with an applied deduplicated readcount cutoff of 5. Differential expression analysis utilized DESeq2 which employs a variance stabilizing transformation to stabilize variance for downstream statistical analysis 0 1 2 3 4 5 6 7 8 Time (hrs) FFPE Treatment 35 min Watchmaker RNA Library Prep Kit with Polaris Depletion Hyb & Deplete Probe Digest. FPE 1st SS 2nd SS & AT SPRI Lig SPRI PCR SPRI Illumina Stranded Total RNA Prep, Ligation with Ribo-Zero Plus Hyb Deplete Probe Digestion FPE SPRI 1st SS 2nd SS SPRI AT Lig SPRI PCR SPRI Hyb Deplete Probe Digestion FPE 1st SS 2nd SS AT Lig USER PCR NEBNext® Ultra II Directional RNA Library Prep Kit with NEBNext Globin & rRNA Depletion SPRI SPRI SPRI SPRI Hyb Deplete Probe Digestion FPE 1st SS 2nd SS & AT Lig PCR KAPA RNA HyperPrep Kit with RiboErase (HMR) Globin SPRI SPRI SPRI SPRI SPRI Capture More with Challenging Samples Figure 5. Reduce data distortion at low inputs. Differential expression analysis via DESeq2 between averaged 100 ng FFPE (control) and 10 ng FFPE (experimental) libraries for each of (A) Watchmaker, (B) KAPA, (C) NEBNext and (D) Illumina. Only commonly identified genes were included in the analysis. The number of common genes between the 100 ng and 10 ng data sets is indicated in the lower right of each plot. Dark data points indicate genes identified as differentially expressed. The Watchmaker chemistry outperforms in terms of more common genes detected and fewer genes identified as differentially expressed. 1e-01 1e+00 1e+01 1e+02 1e+03 1e+04 0 2 4 -4 -2 Mean of Normalized Counts Log Fold Change 1e-01 1e+00 1e+01 1e+02 1e+03 1e+04 0 2 4 -4 -2 Mean of Normalized Counts Log Fold Change NEBNext 1e-01 1e+00 1e+01 1e+02 1e+03 1e+04 0 2 4 -4 -2 Mean of Normalized Counts Log Fold Change KAPA 1e-01 1e+00 1e+01 1e+02 1e+03 1e+04 0 2 4 -4 -2 Mean of Normalized Counts Log Fold Change Illumina A B C D 12,013 common genes 8,798 common genes 6,859 common genes 3,817 common genes Conclusions • All chemistries tested generate quality libraries when RNA inputs are of high quality and quantity. Chemistries are differentiated in performance as RNA quality or input decreases. Overall, the Watchmaker chemistry acceses more information with challenging samples. • Libraries prepared from challenging, low-input, and degraded RNA samples demonstrate increased complexity with the Watchmaker chemistry due to core features of the kit, including: ■ Novel FFPE decrosslinking step ■ A reverse transcriptase specifically engineered for improved conversion of RNA to cDNA ■ Fewer bead purification steps, thereby preventing sample loss • The Watchmaker solution demonstrates higher performance and promotes confidence with low-input and degraded samples as evidenced by: ■ Greater gene detection overlap of Fresh Frozen and FFPE samples ■ Less data distortion with differential expression analysis for low-input FFPE samples Library Complexity and Gene Detection Sensitivity Figure 2. Improve library complexity with lower duplication rate and more unique genes detected. (A) Library duplication rate is impacted by the quantity and quality of the RNA input for all chemistries. (B) With high-input (100 ng) and high-quality (fresh frozen) RNA, all chemistries show similar gene detection sensitivity with only slight differences in the number of unique genes detected. As inputs lower and quality decreases, clear differences emerge. The Watchmaker chemistry detects the most unique genes, followed by KAPA and NEBNext with Illumina performing poorly with this sample type and input. Figure 4. Improved gene detection with FFPE samples. (A) Gene overlap analysis of 10 ng fresh frozen and 10 ng FFPE. The fresh frozen data set is used as the truth set to determine how much potential information can be accessed with FFPE samples for a given chemistry. Only genes detected in both technical replicates were included in the analysis. Unique genes were identified using featureCounts with a cutoff of 5 deduplicated raw reads. (B) Pie charts demonstrate the amount of data retained for FFPE analysis for each chemistry. For all chemistries, most genes detected in the FFPE sample are also present in the fresh frozen libraries. The Watchmaker chemistry detects more genes overall, more genes with FFPE samples, and demonstrates the highest overlap (73%) in genes detected in both fresh frozen and FFPE samples. 0 0.3 0.2 0.1 0.5 0.4 0.7 0.8 0.6 0.9 10 ng 100 ng 10 ng 100 ng Fresh Frozen FFPE Watchmaker KAPA NEBNext Illumina Duplication Rate Duplicate Read Fraction 0 5000 10000 15000 20000 Count 10 ng 100 ng 10 ng 100 ng Fresh Frozen FFPE Watchmaker KAPA NEBNext Illumina Unique Genes: Matched FF/FFPE A B 0 0.3 0.2 0.1 0.5 0.4 0.7 0.9 0.8 0.6 1.0 Exonic Intronic Intergenic rRNA Watchmaker KAPA NEBNext Illumina Library Composition 10 ng Fresh Frozen 100 ng Fresh Frozen 10 ng FFPE 100 ng FFPE 10 ng Fresh Frozen 100 ng Fresh Frozen 10 ng FFPE 100 ng FFPE 10 ng Fresh Frozen 100 ng Fresh Frozen 10 ng FFPE 100 ng FFPE 10 ng Fresh Frozen 100 ng Fresh Frozen 10 ng FFPE 100 ng FFPE Fraction of Read s Figure 3. RNA Seq chemistry used changes library composition. The distribution of sequencing reads across exonic, intronic, intergenic, and ribosomal RNA varies by chemistry, sample, and — in some cases — input amount. NEBNext and Illumina libraries generated with FFPE demonstrate high residual ribosomal RNA reads indicating poor depletion and reducing sequencing efficiency. KAPA chemistry demonstrates the lowest residual rRNA reads but shows variance across inputs for both highquality and FFPE samples. Watchmaker chemistry is consistent between 100 ng and 10 ng inputs for both high-quality and FFPE samples, indicating confident representation even for limited sample input. Total Fresh Frozen Genes: 16,092 15,208 14,184 12,660 Watchmaker KAPA NEBNext Illumina Watchmaker 11784 319 4308 Percentage of Fresh Frozen Genes Detected in Matched FFPE KAPA Illumina Genes identified: 30% Genes identified: 57% Genes missed: 43% Genes missed: 70% Watchmaker NEBNext Genes missed: 27% Genes identified: 73% Genes identified: 48% Genes missed: 52% KAPA 26 8798 6710 Percentage of Fresh Frozen Genes Detected in Matched FFPE KAPA Illumina Genes identified: 30% Genes identified: 57% Genes missed: 43% Genes missed: 70% Watchmaker NEBNext Genes missed: 27% Genes identified: 73% Genes identified: 48% Genes missed: 52% NEB 6865 262 7319 Percentage of Fresh Frozen Genes Detected in Matched FFPE KAPA Illumina Genes identified: 30% Genes identified: 57% Genes missed: 43% Genes missed: 70% Watchmaker NEBNext Genes missed: 27% Genes identified: 73% Genes identified: 48% Genes missed: 52% Illumina 3819 1 8841 Percentage of Fresh Frozen Genes Detected in Matched FFPE KAPA Illumina Genes identified: 30% Genes identified: 57% Genes missed: 43% Genes missed: 70% Watchmaker NEBNext Genes missed: 27% Genes identified: 73% Genes identified: 48% Genes missed: 52% A B 10 ng Fresh Frozen 10 ng FFPE M275