Performance of the IHC Screening Test
The level of staining prevalence of the pan-RTK IHC assay across FFPE samples of multiple histologies (n=636) is shown in Table 2. Although some tumor tissues had a higher proportion that demonstrated positive staining (eg, brain 67%, ovary 60%, stomach/GI 75%), the overall average indicated that approximately 60% to 70% of specimens are not demonstrating significant staining by the pan-RTK IHC and would thus be removed from further consideration by NGS-based testing for gene rearrangements. Therefore, an average of 30% of samples would proceed to the custom NGS assay to be investigated for the presence of gene rearrangements within NTRK1, NTRK2, NTRK3, ROS1, or ALK.
NGS Sample Input Quality
To detect gene rearrangements, FFPE samples are sequenced on an RNA-based NGS assay platform. As FFPE RNA is known to be labile, causing it to be susceptible to fragmentation, chemical modifications, and refractory to enzymatic manipulations, there is the possibility that there will be an insufficient quantity of RNA to reliably sequence a gene rearrangement, resulting in a false-negative final result. To address this risk, we performed a qPCR-based RNA QC assay. The RNA QC assay uses the VCP housekeeping gene to determine the integrity of amplifiable RNA in extracted FFPE samples and reports the results as a C t value. We confirmed that the C t values of the RNA QC assay correlate inversely with RNA sequencing coverage in FFPE samples, whereby increased C t values yielded less sequencing coverage and would indicate a poor quality sample (Fig. 3A). To demonstrate that the RNA QC assay provides accurate RNA quality metrics, artificial RNA fragmentation was performed by Covaris shearing on FFPE embedded cell lines harboring gene rearrangements (Fig. 3B) and a time course of zinc (Zn)-mediated fragmentation on varying FFPE tissue types (Supplemental Digital Content 1, http://links.lww.com/AIMM/A116). Covaris shearing generated smaller size RNA fragments in all samples (Fig. 3B). When sheared, the U-118 MG cell line lost detection of the GOPC:ROS1 rearrangement, which was present in the unmodified sample (Fig. 3B and data not shown). These data indicated that detection of gene rearrangements by the NGS assay can be lost in highly fragmented RNA (Fig. 3B and data not shown). The RNA QC assay was performed on each of these samples. The C t values for the unmodified samples for all 3 cell lines ranged from 24.89 to 25.91 (Fig. 3B, left column, data not shown), whereas the C t values for the sheared samples from all 3 cell lines ranged from 27.82 to 28.35 (Fig. 3B, right column, data not shown). Smaller RNA fragments generated through Zn-mediated fragmentation, with longer fragmentation time leading to shorter fragments, correlated with increased RNA QC C t values (Supplemental Digital Content 1, http://links.lww.com/AIMM/A116). Together, these data demonstrate that the RNA QC assay is able to identify the quality, as a function of fragmentation, of the input RNA into the assay.
As FFPE nucleic acid stability is known to decrease with age,15 it is useful to understand the percentage of blocks passing the RNA QC C t threshold as a function of age. The RNA QC C t value increased in FFPE samples of increased age, indicating decreased quality of RNA in older FFPE samples [Fig. 3C; significant linear trend (P<0.0001)]. These results show that older archival samples, such as those that may be used in retrospective cohort studies, generally have C t values of >29, which indicates that they are of lower quality and thus gene rearrangements may be difficult to detect. However, more recent specimens (<1 y from embedding), such as those commonly used in the clinical setting, will generally have C t values of <29.
A primary use of the RNA QC assay is to determine the quantity of RNA input that is required for a reliable NGS result from a FFPE tissue sample. The ability to sequence a sample is based not solely on the overall quality, but also the quantity of RNA fragments capable of amplifying targeted regions. This effect is demonstrated by the fact that if input quantity remains constant at 200 ng across FFPE tissues of varying quality, the percentage of RNA reads is not constant (Fig. 3D). RNA mass alone is inadequate for predicting the quality of an RNA NGS library and the amount of RNA available for library generation varies per nanogram of input across samples. We used a cohort of FFPE tissues and cell lines and categorized them as “high quality” (C t<26), “medium quality” (26≤C t<29), and “low quality” (C t>29). These specimens were run at different input quantities (Fig. 3E). A trend is observed in that, samples of higher initial quality, measured by a lower C t value, required less input quantity and still resided below the RNA QC threshold, whereas to support mid-range to poor quality FFPE clinical samples, a higher amount of input was required to meet the RNA QC threshold (Fig. 3E). These data demonstrate that although it may be possible to use smaller amounts of input material from recent clinical testing specimens, for retrospective studies using older archival tissues, it may be better advised to start with a higher initial test input of 200 ng of RNA. Taken together, performing the RNA QC assay on all samples and starting with an input of 200 ng of RNA enables a confidence in negative results when testing FFPE samples with varying ranges of age, quality, and preanalytical conditions for tissue preparation.
Accuracy and Precision of the NGS Test
To demonstrate accuracy of the test, we qualitatively assessed the presence or absence of gene rearrangements in 66 samples of multiple FFPE tumor types and FFPE-embedded cell line models (Supplemental Digital Content 2, http://links.lww.com/AIMM/A117). The testing cohort included at least 1 known positive sample for each target receptor (NTRK1, NTRK2, NTRK3, ROS1, and ALK). At the exon-exon level, Sanger sequencing confirmed all of the 15 samples found to be positive by NGS sequencing, resulting in 100% positive agreement across all specimens (Supplemental Digital Content 2, http://links.lww.com/AIMM/A117).
To demonstrate the qualitative reproducibility of detecting gene rearrangements, 3 FFPE-embedded cell lines with known gene rearrangements (KM12; TPM3:NTRK1, HCC78; SLC34A2:ROS1, KARPAS-299; NPM1:ALK) and 1 IHC-positive FFPE lung tumor tissue specimen with a known ALK gene rearrangement (EML4-ALK) were run in triplicate through the NGS assay. All samples showed increased, reproducible, expression (as assessed by the median number of RNA reads across primers) at the rearranged receptor kinase domain and no significant expression for all other receptors (Fig. 4). The exceptions in cases of known overexpression of the full length protein (TrkA in KARPAS-299, Fig. 4B) and elevated basal expression (ROS1 in lung, Fig. 4D). These data were additionally reproduced by performing the NGS assay on FFPE tumor tissue samples containing either an ALK rearrangement (EML4:ALK), an ETV6:NTRK3 rearrangement, or negative for gene rearrangements in these targets. Samples were run by 2 different operators on 2 different days and were assessed qualitatively for gene rearrangements in each replicate (Supplemental Digital Content 3, http://links.lww.com/AIMM/A118). Testing resulted in 100% precision of these samples (Supplemental Digital Content 3, http://links.lww.com/AIMM/A118). These data indicate that this test is accurate and robust in detecting gene rearrangements within the regions covered in this diagnostic test.
Demonstration of Overall Test Performance
A subset of specimens (n=192) from Table 2, representing colorectal cancer (n=90), lung cancer (n=74), and thyroid cancer (n=28), was scored by IHC and sequenced to determine the presence of gene rearrangements using the assay methods described here (Fig. 5A). For each of the tumor types, using NGS testing as the reference standard for gene rearrangement detection, the IHC demonstrated 100% positive agreement [CI (colon), 48%-100%; (thyroid): 55%-100%; (lung): 20%-100%], thus there were no specimens that were determined to harbor gene rearrangements that were initially classified as IHC negative (Fig. 5A). Negative agreements ranged from 86% to 95% [CI (colon), 60%-64%; (thyroid): 74%-78%; (lung): 41%-43%), indicating that the majority of IHC-positive samples did not harbor gene rearrangements (Fig. 5A). Nevertheless, these data demonstrated that the pan-RTK IHC test has 100% negative predictive value for gene rearrangement detection in all 3 tumor types tested across 192 specimens [CI (colon), 95%-100%; (thyroid): 95%-100%; (lung): 95%-100%]. Consequently, of 192 samples tested in this cohort, 86 samples were found to be IHC positive (45% positivity) (Fig. 5A). However, the fusion prevalence within the IHC positive population was 9% demonstrating an enrichment of fusion positive samples within a population of clinical samples, which would have been 4% without the addition of IHC screening.
A representative clinical sample of colorectal adenocarcinoma from a 71-year-old female patient was run through the 2-step assay for detection of gene rearrangements in NTRK1, NTRK2, NTRK3, ROS1, or ALK. The clinical sample underwent pan-RTK IHC screening, which resulted in IHC positivity (Fig. 5B upper left panel). Also shown are the staining patterns for the individual antibodies pan-Trk, ROS1, and ALK (Fig. 5B). The IHC for ROS1 and ALK is negative (Fig. 5B, upper right panel and lower left panel, respectively), whereas the IHC staining for pan-Trk expression is positive and strong (Fig. 5B, lower right panel). Sequencing of this sample by the NGS assay resulted in identification of a TPM3:NTRK1 gene rearrangement. The 3′ end of TPM3, at exon 8, is fused to the 5′ end of exon 10 of NTRK1. This particular gene rearrangement is one of the first NTRK1 fusions to have been reported in colorectal cancer and has been shown to be a potential target of therapeutic intervention by the Trk inhibitor, entrectinib.12 Break-apart FISH further confirmed the presence of an NTRK1 fusion (Fig. 5C). Patient-derived cells from this patient have been shown to be sensitive to inhibition by the TrkA inhibitor (entrectinib).12 Together, these data demonstrate the efficacy of using a pan-RTK IHC assay to effectively screen clinical samples for passage to an NGS based assay that provides an accurate assessment of the presence of gene rearrangements.
We have demonstrated that a pan-RTK IHC screen enhances detection of gene rearrangements in a clinical population (9% prevalence vs. ~4% prevalence). This approach has significant therapeutic and clinical implications in that many gene rearrangements are rare, and thus it is costly for health care providers to identify those patients that would respond to targeted therapy. There is now a precedent of IHC screening for individual RTK gene rearrangements and an FDA-approved companion diagnostic IHC test for ALK. Since tissue availability is also limited for many cancer patients, it is not efficient, both for cost and tissue consumption, to run serial IHC tests for each RTK being tested. We have developed a multiplexed IHC method that identifies samples that may harbor gene rearrangements that will be confirmed by subsequent NGS testing. In a clinical setting, the pan-RTK IHC test allows for the rapid reporting of negative specimens, thus allowing patients to find alternative testing. We have proven here that this method is capable of detecting pan-Trk (TrkA, TrkB, TrkC), ROS1, and ALK expression that is associated with gene rearrangements and has a 100% negative predictive value for screening out samples that do not possess gene fusions.
IHC staining has previously been performed using a cocktail of antibodies for many diagnostic tests. The concern has always been that the specificity is insufficient to report a diagnostic result. We see similar results in the specificity for detection of gene fusions with this pan-RTK antibody cocktail. Pan-RTK IHC-positive samples exhibiting strong staining yielded a negative NGS result for gene rearrangement in far more samples than those that yielded a positive NGS gene rearrangement result. This observation is most likely due to other biological contributions that lead to measurable expression by IHC, such as basal expression and the scoring having intentionally been calibrated to favor sensitivity. This effect could explain the higher pan-RTK IHC-positive rates in some tumor types such as brain, skin, and ovary that are reported to express these target genes.16–19 The clinical relevance of variable expression of these receptors remains largely unknown. However, it has been concluded that in certain tissue types, overexpression of Trk proteins is tumor promoting. One example is glioblastomas where TrkB is overexpressed and, when pharmacologically inhibited, tumor burden can be reduced.16 In addition, a more stringent IHC scoring criteria has been subsequently developed to determine positivity based on >50% tumor cells with intermediate to strong staining. The hope is that this increased stringency will enrich the population further, and such an effect will be monitored in future cohorts. Although the apparent discordance between IHC and NGS is primarily based on IHC positivity yielding negative gene rearrangements, it is important to note that data have shown that IHC staining loss due to improper preanalytical conditions could lead to lead to false-negative IHC staining in samples that do actually harbor gene rearrangements.20 However, most preanalytical conditions due to improper fixation parameters, such as time and temperature, can have strong impact on immunohistochemical staining results21 and concomitantly impact RNA quality and the ability to detect gene rearrangements.22
The NGS assay uses RNA-based sequencing by AMP chemistry to detect gene rearrangements. The advantage of RNA sequencing is that less sequencing is required relative to a DNA-based assay due to the splicing of intronic regions during mRNA generation. In addition, fusion transcripts are often amplified, which enhances the sensitivity of an RNA-based assay relative to a DNA-based assay, enabling a more sensitive test for gene rearrangements by removing the issues of reading through large intronic regions. Furthermore, detection of expressed transcripts adds value in that RNA expression often leads to fusion protein expression and disease progression. The primary concern of RNA use is its labile nature, which can be exacerbated during the FFPE archival process.15 Extracted RNA that has been highly damaged consists largely of fragments that are too short to be informative and/or will not contribute to library preparation and subsequent sequencing, making it difficult to confidently discern negative results from a specimen without knowing the quality of the input RNA. We have demonstrated here that by using RNA QC C t values, we are able to characterize RNA quality within a sample; RNA quality represents an inherent limitation to the sensitivity and reproducibility of the test.
We have demonstrated the accuracy of our pan-RTK IHC combined with RNA-based NGS assay through the identification of a TPM3:NTRK1 fusion in a colorectal cancer patient. A PDx mouse model was generated from this patient and patient-derived cells were cultured in the presence and absence of entrectinib, an RTK inhibitor targeting the TrkA receptor.12 These PDCs responded to treatment as monitored by the downregulation of pharmacodynamic markers and decreased tumor growth; demonstrating that detection of this gene rearrangement can enable potential clinical benefit through treatment and that our test can accurately identify patients that have a high probability of responding to RTK inhibitors targeting gene rearrangements.
Gene rearrangements are rare in the overall population; however, targeted RTK therapies provide viable treatment options.1 The diagnostic challenge is to find a method that allows rapid, accurate testing of a large number of patients. The 2-step diagnostic test presented in this paper demonstrates a method to rapidly screen a large population of patients to identify a small percentage who harbor gene rearrangements. This method allows for the screening of negative patient specimens quickly and inexpensively and reflexes putatively positive samples to higher resolution multiplexed NGS-based testing. One can envision that through the use of digital slide imaging and analysis, the IHC portion of the test could be automated to accommodate high volume and reduce any potential intersite or intrasite subjective biases. Although the test described here is based on 5 specific markers that have importance in precision medicine therapeutics, the fundamental approach is more broadly applicable. Both the IHC and the NGS could be modified to incorporate a different set of specific testing biomarkers.
1. Shaw AT, Hsu PP, Awad MM, et al. Tyrosine kinase gene rearrangements
in epithelial malignancies. Nat Rev Cancer. 2013;13:772–787.
2. Shaw AT, Ou Sai-Hong I, Bang Yung-Jue, et al. Crizotinib in ROS1—rearranged non–small-cell lung cancer. N Engl J Med. 2014;371:1963–1971.
3. Kwak EL, Bang Yung-Jue, Camidge D Ross, et al. Anaplastic lymphoma kinase inhibition in non-small-cell lung cancer. N Engl J Med. 2010;363:1693–1703.
4. Bergethon K, Shaw Alice T, Ignatius Ou Sai-Hong, et al. ROS1 rearrangements define a unique molecular class of lung cancers. J Clin Oncol. 2012;30:863–870.
5. Davies KD, Le Anh T, Theodoro Mariana F, et al. Identifying and targeting ROS1 gene fusions in non-small cell lung cancer. Clin Cancer Res. 2012;18:4570–4579.
6. Wu W, Choudry H. Next Generation Sequencing in Cancer Research. Next Gener Seq Cancer Res. 2013;1:75–90.
7. Oliver GR, Hart SN, Klee EW. Bioinformatics for clinical next generation sequencing. Clin Chem. 2015;61:124–135.
8. Metzker ML. Sequencing technologies—the next generation. Nat Rev Genet. 2010;11:31–46.
9. Park G, Kim TH, Lee HO, et al. Standard immunohistochemistry efficiently screens for anaplastic lymphoma kinase rearrangements in differentiated thyroid cancer. Endocr Relat Cancer. 2014;22:55–63.
10. Blackhall FH, Peters Solange, Bubendorf Lukas, et al. Prevalence and clinical outcomes for patients with ALK-positive resected stage I to III adenocarcinoma: results from the European Thoracic Oncology Platform Lungscape Project. J Clin Oncol. 2014;32:2780–2787.
11. Sholl LM, Sun Heather, Butaney Mohit, et al. ROS1 immunohistochemistry for detection of ROS1-rearranged lung adenocarcinomas. Am J Surg Pathol. 2013;37:1441–1449.
12. Lee SJ, Li GG, Kim ST, et al. NTRK1 rearrangement in colorectal cancer patients: evidence for actionable target using patient-derived tumor cell line. Oncotarget. 2015;6:1–8.
13. Zheng Z, Liebers Matthew, Zhelyazkova Boryana, et al. Anchored multiplex PCR for targeted next-generation sequencing. Nat Med. 2014;20:1479–1484.
14. Ardini E, Bosotti Roberta, Borgia Andrea Lombardi, et al. The TPM3-NTRK1 rearrangement is a recurring event in colorectal carcinoma and is associated with tumor sensitivity to TRKA kinase inhibition. Mol Oncol. 2014;8:1–13. doi:10.1016/j.molonc.2014.06.001.
15. Ludyga N, Grünwald Barbara, Azimzadeh Omid, et al. Nucleic acids from long-term preserved FFPE tissues are suitable for downstream analyses. Virchows Arch. 2012;460:131–140.
16. Brodeur GM, Minturn Jane E, Ho Ruth, et al. Trk receptor expression and inhibition in neuroblastomas. Clin Cancer Res. 2009;15:3244–3250.
17. Tapia V, Gabler Fernando, Muñoz Marcela, et al. Tyrosine kinase A receptor (trkA): a potential marker in epithelial ovarian cancer. Gynecol Oncol. 2011;121:13–23.
18. Ren H, Tan ZP, Zhu X, et al. Identification of anaplastic lymphoma kinase as a potential therapeutic target in ovarian cancer. Cancer Res. 2012;72:3312–3323.
19. Polsky D, Cordon-Cardo C. Oncogenes in melanoma. Oncogene. 2003;22:3087–3091.
20. Nkoy FL, Hammond Elizabeth H, Rees William, et al. Variable specimen handling affects hormone receptor test results in women with breast cancer: a large multihospital retrospective study. Arch Pathol Lab Med. 2010;134:606–612.
21. Gundisch S, Annaratone Laura, Beese Christian, et al. Critical roles of specimen type and temperature before and during fixation in the detection of phosphoproteins in breast cancer tissues. Lab Invest. 2015;95:561–571.
22. Bussoloti G, Annaratone Laura, Medico Enzo, et al. Formalin fixation at low temperature better preserves nucleic acid integrity. PLoS One. 2011;6:e21043.
NGS; NTRK; gene rearrangements; IHC screen; entrectinib
Supplemental Digital Content
Copyright 2017 Wolters Kluwer Health, Inc. All rights reserved.