https://orcid.org/0000-0001-6138-9257
Abstract
While there is great potential for unbiased next-generation sequencing (NGS) approaches—eg, whole transcriptome sequencing (WTS)—for exploration, discovery, and clinical application in the realm of oncology, there are limitations that should be considered when relying on these methodologies for clinical decision making. When using WTS for the detection of clinically relevant gene fusions in tumor specimens, a key consideration is whether a limited coverage depth (approximately 30-50X) is sufficient for detecting these events, especially in samples with low tumor purity. We demonstrate the reduced sensitivity of both a commercial WTS assay for the detection of clinically relevant fusions in analytical validation control samples and of a research use only (RUO) WTS assay for the detection of clinically relevant fusions in real-world clinical samples compared to RNA comprehensive genomic profiling (CGP). Notably, the RUO WTS assay would not have reported 30% (6/20) of fusions detected using RNA CGP assays in fusion-positive tumor samples, highlighting a potential disadvantage of broader sequencing.
Introduction
Gene fusions are strong oncogenic drivers in cancer and have established clinical value for diagnosis, prognostication, and treatment of various malignancies.1 Traditional methods of fusion detection are loci-specific—eg, fluorescence in situ hybridization (FISH) and real-time polymerase chain reaction (RT-PCR)—and often limited to fusions that have been previously characterized. However, with the advent of next-generation sequencing (NGS) technologies, broad profiling for both known and novel fusions is now possible.2
Profiling of RNA for expressed fusion transcripts can be achieved using targeted panels, which assess for rearrangement events in a carefully selected set of clinically relevant fusion genes (10s-100s of genes), or through whole transcriptome sequencing (WTS) which represents a broader approach (> 20 000 genes) and may allow for the detection of atypical, rare, and novel fusions not interrogated by more limited panels.1 However, the breadth of coverage afforded by WTS has a natural corollary in that depth of coverage is sacrificed.3 In this study, we compared the performance of both a commercial WTS assay and a research use only (RUO) WTS assay to RNA comprehensive genomic profiling (CGP) for the detection of clinically relevant fusions.
Methods
RNA CGP sensitivity testing in commercially available fusion-positive controls
RNA CGP was performed using FoundationOne RNA, a laboratory developed test for the detection of fusions which utilizes a hybrid-capture-based sequencing workflow targeting 318 genes in RNA (converted to cDNA) extracted from formalin-fixed paraffin-embedded (FFPE) tumor.4 FoundationOne RNA was used to profile replicates of a fusion-positive FFPE-derived RNA control sample, SeraSeq Fusion RNA Mix v4 (SeraCare, Milford, MA, USA), at titrations of 100%, 10%, 5%, 2.5%, 1%, 0.5%, and 0.25%. A total of 92 replicates were assessed: 2, 5, 5, 20, 20, 20, and 20 replicates at each titration, respectively.
Research use only (RUO) WTS versus RNA CGP sensitivity comparison in fusion-positive clinical samples
WTS was performed using a Foundation Medicine RUO assay which utilizes a hybrid-capture based sequencing workflow targeting the human exome (> 22 000 genes) in RNA (converted to cDNA) derived from FFPE tumor specimens. RNA CGP was performed using either FoundationOne RNA, as described above, or FoundationOneHeme, an assay combining sequencing of DNA (405 genes) and RNA (265 genes).5 FoundationOne Heme was performed on hybrid-captured, adaptor ligation-based libraries using RNA (converted to cDNA) and DNA extracted from FFPE tumor in a Clinical Laboratory Improvement Amendments (CLIA)-certified, College of American Pathologists (CAP)-accredited, New York State-approved laboratory (Foundation Medicine, Inc., Cambridge, MA, USA). RUO WTS, FoundationOne RNA, and FoundationOne Heme sequencing data were analyzed using equivalent computational pipelines and downsampled to 30 million total reads per sample for comparison. Downsampling involved random exclusion of reads for both RUO WTS (approximately 67.5 million median total reads before downsampling) and RNA CGP (approximately 51.4 million median total reads for FoundationOne RNA and approximately 48.8 million median total reads for FoundationOne Heme before downsampling) to ensure consistency in data readouts across specimens and replicates. The same read count thresholds for fusion/rearrangement calls, variable by fusion event, were applied for both the RUO WTS and RNA CGP assays. Approval for this study, including a waiver of informed consent and a HIPAA waiver of authorization, was obtained from the Western Institutional Review Board (Protocol No. 120160955).
Results
Using a commercially available fusion-positive control, FoundationOne RNA detected all (100%) fusions at the 10% titration level (Figure 1), with 14 of 16 fusions having a median of ≥15 supporting reads across replicates. Fifteen of 16 (93.4%) fusions were detected at 5% titration (11/15 with median ≥ 15 reads), 14 of 16 (87.5%) fusions were detected at 2.5% titration (3/14 with median ≥ 15 reads), and 10 of 16 (62.5%) fusions were detected at 1% titration (1/10 with median ≥ 15 reads). Even at very low titrations, select fusions were still detectable (0.5% titration: 6 of 16 fusions [37.5%]; 0.25% titration: 3 of 16 fusions [18.8%]).
Sensitivity of FoundationOne RNA, a 318-gene RNA CGP assay, for detection of fusions in a commercially available fusion positive control. The median number of direct supporting reads across replicates with positive findings for fusions at different titration levels of the SeraSeq Fusion RNA Mix v4 positive control sample is indicated. Borders indicate 95% or greater and <50% positive hit rate across replicates, respectively.
Across 7 fusion-positive clinical samples (sarcoma n = 4, colorectal cancer n = 1, cholangiocarcinoma n = 1, fibroma n = 1) profiled with both FoundationOne RNA and a Foundation Medicine RUO WTS assay, FoundationOne RNA identified all fusions (7/7) with a high number of supporting reads, whereas the RUO WTS assay had lower support for identified (7/7) fusions (median 66.0 reads [IQR 35.8, 93.3] versus median 430.5 [IQR 273.0, 508.3], P = 0.001) with 1/7 (14.3%) fusions detected below reporting confidence thresholds (Figure 2, Supplementary Table S1). Across an additional 8 clinical samples (sarcoma n = 6, hematologic malignancy n = 2) profiled with both FoundationOne Heme and the same RUO WTS assay, FoundationOne Heme identified all fusions (13/13) in RNA with abundant supporting reads, while the RUO WTS assay again showed lower support for identified (11/13) fusions (median 16.5 reads [IQR 8.0, 40.0] vs median 272.0 [IQR 74.0, 470.0], P = 0.05), with 3/13 (23.1%) fusions detected below reporting confidence thresholds and 2/13 (15.4%) fusions lacking any supporting reads (Figure 2, Supplementary Table S2). In total, an RUO WTS assay would not have reported 6/20 (30.0%) fusions detected using RNA CGP assays in fusion-positive clinical samples (Figure 2).
An RUO WTS assay missed 30.0% (6/20) of clinically relevant fusions detected with RNA CGP (FoundationOne RNA or FoundationOne Heme) in clinical samples. Table denotes whether fusions were detected or not detected/(†) detected below reporting confidence thresholds in one or more replicates. Two sample replicates per assay were tested unless indicated. Asterisks (*) indicate samples with only a single replicate due to limited residual sample for testing. CGP, comprehensive genomic profiling; RUO, research use only; WTS, whole transcriptome sequencing.
Discussion
Our study demonstrates that the breadth of coverage provided by WTS comes at the cost of lower depth of coverage (approximately 30-50X5,6 compared to >500X with Foundation Medicine CGP panels) and potentially results in reduced sensitivity for fusion detection. Because WTS targets gene expression across the entire transcriptome, genes that are highly expressed may consume a significant proportion of available sequencing reads. As a result, fusions expressed at low abundance may be missed because they compete for sequencing depth with these highly expressed genes. Therefore, higher sequencing coverage is necessary with WTS to achieve fusion detection with comparable sensitivity to RNA CGP.
We observed that FoundationOne RNA, an RNA CGP assay, detected all fusions in a commercially available fusion-positive control at as low as 10% titration, with 87.5% (14/16) of fusions detected with a median number of supporting reads greater than or equal to 15. Using the same control, a commercial WTS assay (Tempus Labs, Inc., Chicago, IL, USA) reportedly detected all fusions at as low as 50% titration, with 93.8% (15/16) of fusions detected with a median of ≥15 supporting reads.7 At 10% titration, as compared to FoundationOne RNA, 1/16 fusions was missed by the commercial WTS assay and only 31.3% (5/16) of fusions were detected with a median of ≥15 supporting reads. Overall, FoundationOne RNA had higher sensitivity (ie, required a lower minimum titration fraction for detection) for 10 of 16 (62.5%) fusions, while sensitivity was comparable for 5 of 16 fusions (31.3%; CCDC6::RET, EML4::ALK, NCOA4::RET, FGFR3::TACC3, and EGFR::SEPT14). Moreover, the minimum titration fraction for detected fusions using FoundationOne RNA compared to the commercial WTS assay was as much as 10-fold lower for specific fusions: minimum titration fraction was 10-fold lower for 4 of 16 (25.0%; SLC45A3::BRAF, SLC34A2::ROS1, LMNA::NTRK1, and TMPRSS2::ERG), 5-fold lower for 2/16 (12.5%; TFG::NTRK1 and PAX8::PPARG1), and between 1- and 5-fold lower for 4 of 16 (25.0%; KIF5B::RET, ETV6::NTRK3, TPM3::NTRK1, and FGFR2::BAIAP2L1). This difference in sensitivity cautions that fusions in clinical samples with low tumor purity, or fusions with low abundance of expression, may be missed using a WTS methodology.
Exemplifying the potential impact of reduced depth of coverage for clinical fusion testing, an RUO WTS assay would not have reported (defined as either complete lack of evidence or evidence below the threshold for reporting) 14.3% (1/7) of fusions that were detected by FoundationOne RNA in fusion-positive clinical samples. In addition, the same RUO WTS assay would not have reported 38.5% (5/13) of fusions detected in RNA on FoundationOne Heme in a separate set of clinical samples. In total, the RUO WTS assay would not have reported 30.0% (6/20) of fusions detected by RNA CGP. Generally, the level of evidence (ie, read support) for fusions detected using the RUO WTS assay was much lower than for both FoundationOne RNA and FoundationOne Heme even though the total number of analyzed reads was equivalent for all 3 assays at 30 million total reads per sample. In fact, the average coverage of the CGP baited regions was 10.3-fold higher for samples profiled with FoundationOne RNA and 9.8-fold higher for samples profiled with FoundationOne Heme compared to the RUO WTS.
A limitation of this analysis is that the results are not directly translatable to other assays, ie, the proportion of missed fusions using WTS may be higher or lower than reported in our study depending on which specific WTS and CGP assays are compared given differences in bait set design, sequencing methods, reporting rules, etc. Importantly, WTS assays can be designed to address these sensitivity limitations by increasing the depth of coverage in key genomic regions associated with clinically relevant fusions. Therefore, choice of assay matters. Another limitation of this study is the small sample size, such that detection of only a few clinically relevant fusions was tested. Comparative studies using larger cohorts with a broader diversity of fusion-positive samples is needed to refine the estimate of potentially missed fusions. Thus, this estimate is dependent on both assay- and sample-specific factors.
Whereas WTS may be a better approach for research and exploratory analyses to detect atypical or novel rearrangements and fusions, the currently limited set of clinically relevant fusions—those that are targetable, diagnostic, and/or prognostic—can be purposefully baited on RNA CGP panels allowing for highly sensitive and confident detection in the setting of oncology clinical care. The interplay between depth and breadth of coverage applies to other analytes, such as DNA, and is therefore also important to consider for DNA assay selection (eg, DNA CGP vs whole-exome sequencing).
Supplementary material
Supplementary material is available at The Oncologist online.
Acknowledgments
We wish to thank Ena Shinnishi for her contributions to wet lab design and execution of the RUO WTS experiments.
Conflicts of interest
R.B.K.-E., D.M., T.R., S.R., A.S., and R.S.P.H. are employees of Foundation Medicine, a wholly owned subsidiary of Roche, and have equity interest in Roche.
Data availability
The authors declare that all relevant aggregate data supporting the findings of this study are available within the article and its supplementary information files. The data that support the findings of this study originated from Foundation Medicine, Inc. In accordance with the Health Insurance Portability and Accountability Act, we do not have IRB approval or patient consent to share individualized patient genomic data, which contains potentially identifying or sensitive patient information and cannot be reported in a public data repository. Foundation Medicine is committed to collaborative data analysis and has well established and widely used mechanisms by which qualified researchers can query our core genomic database of >900,000 de-identified sequenced cancers. Academic researchers can submit a proposal to the Foundation Medicine Data Collaborations Committee and, if approved, the researcher/institution will be required to complete a Data Usage Agreement. More information and mechanisms for data access can be obtained by contacting the corresponding author or the Foundation Medicine Data Governance Council at [email protected].
