Analytical Validation of a Next-Generation Sequencing Assay to Monitor Immune Responses in Solid Tumors.
Author(s): Conroy JM, Pabla S, Glenn ST, Burgher B, Nesline M, Papanicolau-Sengos A, Andreas J, Giamo V, Lenzo FL, Hyland FCL, Omilian A, Bshara W, Qin M, He J, Puzanov I, Ernstoff MS, Gardner M, Galluzzi L, Morrison C
Publication: J Mol Diagn, 2018, Vol. 20, Page 95-109
PubMed ID: 29061374 PubMed Review Paper? No
Purpose of Paper
This paper investigated the effects of formalin-fixation, section storage, tumor and necrotic content, and assay characteristics on results of next-generation sequencing (NGS) of RNA and DNA and compared NGS results to those obtained with other platforms.
Conclusion of Paper
The normalized reads per million (nRPM) for the 54 targeted genes was strongly correlated between matched frozen and formalin-fixed paraffin-embedded (FFPE) specimens, specimens sectioned and extracted after less than 2 weeks and those processed 20 weeks later, specimens containing 100% tumor RNA and those containing 10-80% tumor RNA, and those containing 0% RNA from necrotic tissue and those containing 10-100% RNA from necrotic tissue. Further, RNA-seq was very robust and was highly reproducible, regardless of the input amount, genomic DNA contamination, batch size, library dilution, operator, days, barcode, or assay. For the 54 target genes, the nRPM on the gene levels were strongly negatively correlated with TaqMan real-time PCR cycle threshold values and strongly correlated with the whole-transcriptome RNA-seq data from matched frozen specimens and with CD8a and PD-L1 immunostaining results.
The mutational burden was affected if <50% tumor was included but was very strongly correlated among specimens with no DNA from necrotic tissue and those with 5-100% DNA from necrotic tissue. While the DNA input amount, batch size, operator, day of assay, and assay had no effect on mutational burden; dilution of the library resulted in lower coverage and fewer identified mutations. DNA-seq variant calls were very strongly correlated with TCGA-called variants. The effect of library dilution on the mutational burden was only observable when there were less than 850,000 reads.
Studies
-
Study Purpose
This study investigated the effects of formalin-fixation and section storage on RNA-seq and the effects of tumor content and percentage necrosis on nRPM and the mutational burden. The study included a total of 167 FFPE non-small cell lung cancer, melanoma, renal cell carcinoma, head and neck squamous cell carcinoma, and bladder cancer specimens obtained during fine-needle aspiration biopsy, punch biopsy, needle core biopsy, incisional biopsy, excisional biopsy, or surgical resection; but details of FFPE processing were not included. Matched frozen specimens stored at -80˚C were obtained from 13 specimens. Some specimens had whole transcriptome and whole exome sequencing data from The Cancer Genome Atlas (TCGA) project. Specimens were confirmed to contain greater than 50% cellularity and <50% necrosis and areas containing nonmalignant, necrotic, or malignant tissue were scraped from 3-5 unstained slides and DNA and RNA were extracted using the truXTRAC FFPE extraction kit. DNA yields were determined using the Quantifiler Human DNA quantification kit and RNA yields were measured using the Quan-iT RNA HS assay. The expression of 54 target genes was determined using the Oncomine Immune Response Research Assay and the Ion AmpliSeq targeted-sequencing technology. Expression was confirmed by TaqMan gene expression assays and immunohistochemistry for PD-L1 (pathologist evaluated) and CD8A (automated). The mutational burden was investigated using the Comprehensive Cancer Panel and the Ion Chef system. The effect of percentage of malignant content was investigated by diluting DNA and RNA from five tumor specimens with DNA and RNA from the normal adjacent specimen. The effect of necrotic content was investigated by adding increasing amounts of DNA and RNA from necrotic areas.
Summary of Findings:
The normalized reads per million (nRPM) for the 54 targeted genes was strongly correlated between matched frozen and FFPE specimens (r=0.943). The RNA-seq results were strongly correlated between specimens sectioned and extracted after less than 2 weeks and those processed after 20 weeks (r>0.96). Diluting nucleic acids from tumor specimens with those from normal adjacent specimens to obtain a mixture RNA or DNA of normal adjacent to tumor specimens of 10%-80% resulted in strong correlations (0.75-0.99) in nRPM value for the 54 targeted genes, but at least 50% tumor RNA was needed to obtain concordant mutational burden information (high for all). When the mixture contained less than 10% malignant cells, specimens showed a higher coefficient of variation (CV) of the 10 housekeeping genes than otherwise observed (0.6068 versus 0.0071-0.1053) and if the mixture contained less than 50% malignant cells, the mutational burden was no longer considered high. Average nRPM values and housekeeping gene levels were very strongly correlated between specimens with 0% and 100% necrotic tissue (r>0.97 and r>0.95, respectively) and the authors report low CV for the specimens, despite the presence of RNA from necrotic tissue. Further, while RNA-seq failed for a single specimen which had 100% necrosis, dilution of RNA with this specimen still produced very strong correlations with specimens with no RNA from necrotic specimens (r=0.911 to 0.955). The mutational burden was comparable in specimens containing no necrotic tissue and those containing 5-100% necrotic tissue.
Biospecimens
Preservative Types
- Formalin
- Frozen
Diagnoses:
- Neoplastic - Melanoma
- Neoplastic - Normal Adjacent
- Neoplastic - Carcinoma
Platform:
Analyte Technology Platform DNA Next generation sequencing RNA Next generation sequencing Pre-analytical Factors:
Classification Pre-analytical Factor Value(s) Biospecimen Preservation Type of fixation/preservation Frozen
Formalin (buffered)
Biospecimen Aliquots and Components Biospecimen components 100% tumor
80% tumor
50% tumor
40% tumor
30% tumor
20% tumor
10% tumor
0% tumor
100% necrotic
80% necrotic
50% necrotic
40% necrotic
30% necrotic
20% necrotic
10% necrotic
0% necrotic
Storage Storage duration < 2 weeks
20 weeks
-
Study Purpose
This study investigated the effects of RNA and DNA input, library dilution, genomic DNA contamination, batch size, operator, day of assay, bar code, and assay on nRPM and mutational burden. The suitability of the controls was also investigated. The study included a total of 167 FFPE non-small cell lung cancer, melanoma, renal cell carcinoma, head and neck squamous cell carcinoma, and bladder cancer specimens obtained during fine-needle aspiration biopsy, punch biopsy, needle core biopsy, incisional biopsy, excisional biopsy, or surgical resection, but details of FFPE processing were not included. Specimens were confirmed to contain greater than 50% cellularity and <50% necrosis and areas containing nonmalignant, necrotic, or malignant tissue were scraped from 3-5 unstained slides and DNA and RNA were extracted using the truXTRAC FFPE extraction kit. DNA yields were determined using the Quantifiler Human DNA quantification kit and RNA yields were measured using the Quan-iT RNA HS assay. The expression of 54 target genes was determined using the Oncomine Immune Response Research Assay and the Ion AmpliSeq targeted-sequencing technology. The mutational burden was investigated using the Comprehensive Cancer Panel and the Ion Chef system. RNA-seq was conducted using 2.5, 5, 10, 15, and 25 ng RNA from a specimen with low expression levels and from one with medium expression levels. DNA-seq was conducted using 1.8, 3.75, 7.5, 15, and 30 ng DNA from two specimens with intermediate mutational burden. To investigate the effects of genomic DNA contamination, genomic DNA was added at 33% and 50% to each of three RNA specimens. The effect of batch size was investigated by including 4, 8, or 16 specimens in a run. The linearity of detection was investigated by diluting a RNA-seq library such that instead of 50 pmol/L only 25, 12.5, 6.25, 3.125, or 1.5625 pmol/L were included and by diluting DNA-seq libraries of six specimens from 100 pmol/L to 50, 24,12.5, 6.25, and 3.25 pmol/L.
Summary of Findings:
nRPM values for the 54 target genes were very strongly correlated and highly reproducible, regardless of the input amount (r=0.997 and CV <0.03). Similarly, changing the DNA input did not alter the mutational burden (CV=0.12). Even after contamination of the RNA with 33% or 50% genomic DNA, nRPM values for the 54 target genes and the 10 housekeeping genes were very strongly correlated with values of uncontaminated RNA (r>0.99, and r>0.97, respectively). Batch size did not significantly affect nRPM of the 54 target genes or the mutational burden. A very strong linear correlation in nRPM for the 54 target genes and 10 housekeeping genes was noted even when the library was diluted to contain only 3.125% of the recommendation. The absolute reads for the 54 genes within the six diluted specimens had a large range (0-137,051 reads), but reads for individual genes displayed a very strong linear correlation (r>0.97). With increasing dilution of the DNA-seq library, the average mean coverage decreased from 1,144,740 reads to 91,210 reads and the max coverage decreased from 98% with 100 pmol/L to 7.8% with 3.125 pmol/L. Further, the mutational burden ranged from 0-12 and decreased with increasing dilution, but the absolute mutation counts were unaffected if there were >850,000 reads. Importantly, the reproducibility demonstrated when using two to four specimens for nRPM was very strong between operators, days, barcodes, and assays and within assays (r>0.99, all) and the mutational burdens were similarly unaffected. The RNA-seq read counts for no template controls ranged from 0-667 but the median across all genes and replicates ranged from 0-7 and all data was corrected for the no template control. Similarly, the DNA-seq no template control generated 6963-138,811 reads, which was much lower than for samples (average 49,529 versus 2489958) and, notably the no template control generated no variants calls.
Biospecimens
Preservative Types
- Formalin
Diagnoses:
- Neoplastic - Melanoma
- Neoplastic - Carcinoma
Platform:
Analyte Technology Platform DNA Next generation sequencing RNA Next generation sequencing Pre-analytical Factors:
Classification Pre-analytical Factor Value(s) Next generation sequencing Specific Template/input amount 2.5 ng RNA
5 ng RNA
10 ng RNA
15 ng RNA
25 ng RNA
1.8 ng DNA
3.75 ng DNA
7.5 ng DNA
15 ng DNA
30 ng DNA
50 pmol/L library
25 pmol/L library
12.5 pmol/L library
6.25 pmol/L library
3.125 pmol/L library
1.5625 pmol/L library
100 pmol/L library
Biospecimen Aliquots and Components Biospecimen components 0% genomic DNA
30% genomic DNA
50% genomic DNA
-
Study Purpose
This study compared RNA-seq data with the results of TaqMan real-time PCR, whole-transcriptome RNA-seq, and CD8A and PD-L1 immunostaining and compared variants called by DNA-seq with those in TCGA. The study included a total of 167 FFPE non-small cell lung cancer, melanoma, renal cell carcinoma, head and neck squamous cell carcinoma, and bladder cancer specimens obtained during fine needle aspiration biopsy, punch biopsy, needle core biopsy, incisional biopsy, excisional biopsy, or surgical resection, but details of FFPE processing were not included. Matched frozen specimens stored at -80˚C were obtained from 13 specimens. Specimens were confirmed to contain greater than 50% cellularity and <50% necrosis and areas containing nonmalignant, necrotic, or malignant tissue were scraped from 3-5 unstained slides and DNA and RNA were extracted using the truXTRAC FFPE extraction kit. DNA yields were determined using the Quantifiler Human DNA quantification kit and RNA yields were measured using the Quan-iT RNA HS assay. The expression of 54 target genes was determined using the Oncomine Immune Response Research Assay and the Ion AmpliSeq targeted-sequencing technology. Expression was confirmed by comparison to whole transcriptome data, TaqMan gene expression assays, and immunohistochemistry for PD-L1 (pathologist evaluated) and CD8A (automated). The mutational burden was investigated using the Comprehensive Cancer Panel and the Ion Chef system and was compared to whole exome sequencing data from The Cancer Genome Atlas (TCGA) project.
Summary of Findings:
For the 54 target genes, the nRPM on the gene levels were strongly negatively correlated with TaqMan real-time PCR cycle threshold values (r=-0.86), but the correlation was lower for three genes (r<-0.7) and, in some specimens, these genes were not detected. The nRPM values for the FFPE specimens and the whole-transcriptome RNA-seq data from matched frozen specimens were strongly correlated (r=0.7739), but while lower than for real-time PCR it is important that these were with frozen specimens. Importantly, specimens with the lowest correlations observed with real-time PCR also had fewer mapped reads, thus the authors suggest a threshold of 200,000 reads for RNA-seq. Strong correlations between nRPM and immunostaining were observed for CD8A (r=0.797) and PD-L1 (r=0.78). The correlations increased with increasing IHC H-score and modified H-score. The authors state that the high nRPM for CD27 in four specimens with 30%, 35%, 0%, or 0% tumor, as determined by IHC for PD-L1, demonstrated that nRPM is more sensitive than IHC. DNA-seq variant calls were very strongly correlated with TCGA-called variants, but the authors state that at least >850,000 bp must be sequenced with at least 20 x depth for accurate mutation calling.
Biospecimens
Preservative Types
- Frozen
- Formalin
Diagnoses:
- Neoplastic - Carcinoma
Platform:
Analyte Technology Platform RNA Real-time qRT-PCR Protein Immunohistochemistry DNA Next generation sequencing RNA Next generation sequencing Pre-analytical Factors:
Classification Pre-analytical Factor Value(s) Next generation sequencing Specific Technology platform Real-time PCR
Whole transcriptome seq
Whole exome Seq
IHC
Biospecimen Preservation Type of fixation/preservation Formalin (buffered)
Frozen
Immunohistochemistry Specific Targeted peptide/protein CD8A
PD-L1