NGS Techniques to Detect TRK Fusions
About the Speaker
Dr. Long Phi Lee is a practicing molecular pathologist in the MGH Center for Integrated Diagnostics, and also the associate chief of pathology informatics at the Massachusetts General Hospital. His clinical and research interests include development of novel target enrichment, bioinformatics, and medical informatics solutions for next generation sequencing, and their application for clinical molecular diagnostics. He has a strong interest in applying big data, descriptive and predictive analytics and healthcare with the goal of efficiently delivering laboratory results, and clinical decision support.
Dr Lee was co-founder of Archer Diagnostic and remained as consultant for the company. Archer Dx had commercialized the sequencing technique Dr Lee developed at MGH, which was named anchored multiplex PCR.
The sequencing and molecular testing have become tried-and-true standard of care, including EGFR in lung adenocarcinoma, KRAS mutations in colon adenocarcinoma, BRAF V600E mutations in melanoma. There were growing data and association between mutations and either diagnosis or treatment response. Prior to last decades, the gene fusions were studied in hematological diseases mainly. Lymphomas and sarcomas were regarded as being more prone to having rearrangements. . It wasn't until the TCGA projects and other molecular profiling efforts did we realize that gene fusions extended well beyond those. They were found in breast cancer, brain cancer, lung cancer, colon cancer, as well.
Structural chromosome rearrangements may result in the exchange of coding or regulatory DNA sequences between genes. Many such gene fusions are strong driver mutations in neoplasia and have provided fundamental insights into the disease mechanisms that are involved in tumorigenesis. In addition, increasing numbers of chimeric proteins, encoded by the gene fusions serve as specific targets for treatment, resulting in dramatically improved patient outcomes. For example, at MGH back in early 2000, ALK and ROS1 patients were first identified by FISH and were placed in clinical trials, respectively, treating lung cancers.
Back in 2007-2008, we were fortunate enough to have FISH assay that screened patients by detecting the intrachromosomal rearrangement where 20 mega bases were involved. On the other hand, some complex mutation events were difficult to detect by FISH- those in the form of TACC3-FGFR3 (in GBM), SEC16A-NOTCH1 (in breast cancer), EGFRvIII (in GBM), as well as NOTCH1 Del (in breast cancer). People were looking for alternative generic assays that were high-throughput, scalable, and sensitive.
NTRK Fusion Genes
NTRK gene fusions result from intra-chromosomal or inter-chromosomal rearrangements that juxtapose the 3’ region of the NTRK gene with the 5’ sequence of a fusion partner gene expressed by the tumor cell progenitor.2 The NTRK gene fusion transcript encodes a protein composed of the N-terminus of the fusion partner with the TRK partner tyrosine kinase domain.2 In most characterized fusions, the 5’ partner gene sequence encodes one or more dimerization domains,7 resulting in a constitutively active fusion protein. This constitutive activation results in uninterrupted downstream signaling messages, thereby acting as a true oncogenic driver. Although fusions may occur in any of the three NTRK genes, most of those identified to date involve either NTRK3 or NTRK1.
Testing Methods for TRK Fusion Cancers
Approaches that may be used to directly or indirectly detect the presence of a gene fusion in clinical tissue samples include immunohistochemistry (IHC), fluorescence in situ hybridization (FISH), reverse transcriptase polymerase chain reaction (RT-PCR) and next-generation sequencing (NGS) using DNA or RNA. The features of each technique were reviewed in Table 1. 
Anchored Multiplex PCR
The target enrichment method for next-generation sequencing, termed anchored multiplex PCR (AMP), that is compatible with low nucleic acid input from formalin-fixed paraffin-embedded (FFPE) specimens. AMP is effective in detecting gene rearrangements (without prior knowledge of the fusion partners), single nucleotide variants, insertions, deletions and copy number changes. A unique advantage of AMP compared to other PCR methods is its ability to assess for unique reads. As a result, AMP may be used to assess sequence read complexity based on random start sites, in contrast to other PCR-based enrichment techniques that demand prior knowledge of reported fusion partners in the literature. By targeting sequences with a one-sided nested primer approach, AMP offers the distinct ability to agnostically detect gene rearrangements by simply targeting one of the consistently involved fusion partners. Based on a core set of standard molecular biology reagents, AMP utilizes primers that may be quickly designed and synthesized as part of a facile, custom targeted sequencing solution wherein library construction can be completed in 1–2 days. 
Figure 1: AMP for targeted RNA and DNA sequencing. A double-stranded cDNA (dscDNA) synthesis starts with total nucleic acid or RNA from fresh or FFPE material without ribosomal RNA or genomic DNA (gDNA) depletion. Solid-phase reversible immobilization (SPRI)-cleaned double-stranded cDNA or fragmented genomic DNA is processed with end repair and dA tailing, directly followed by ligation with a half-functional adapter. SPRI-cleaned, ligated fragments are amplified with 10–14 cycles of multiplex PCR1 using gene-specific primers (GSP1 pool) containing a PCR multiplexing tag (brown) and a primer complementary to a portion of the universal ligated adapter (short green). An unknown fusion partner or target sequence is indicated by a question mark. SPRI-cleaned PCR1 amplicons are amplified with a second round of 10-cycle multiplex PCR2 using a combination of GSP2 pool nested gene-specific primers (3′ downstream of GSP1), which are tagged with the second adapter sequence specific for Ion Torrent (black-blue) or Illumina (black-red, subsequently tagged with an indexing primer (red-orange-purple)), and a second nested primer against the ligated universal adapter (long green). After a final SPRI cleanup, the target amplicon library is ready for quantitation, downstream clonal amplification and sequencing.
The limitations and challenges of NTRK fusion detection by RNA based AMP include (1) lack of bioinformatic tools for chimeric fusion detection; (2) analytical sensitivity compared with serial dilutions in DNA assay; (3) limited specimens for validation due to low prevalence rate of NTRKs; (4) difficult target validation; (5) quality control across genes, tissues, expression profiles across various tumor types.
NTRK experience sharing from MGH
MGH has been levering 55-gene solid tumor panel since 2016 to detect fusion genes. It takes 2-3 weeks to have results returned. Out of approximately 8000 specimens, 43 (0.5%) NTRK fusions (NTRK 1/2/3 variants) positive with 19 gene partner combinations were identified across tumor types encompassing thyroid papillary carcinoma, lung adenocarcinoma, lung IMT, GBM, melanoma, pancreatic carcinoma and so on. The NTRK fusions were proven to be actionable drivers that associated with clinical response in 50% of patients upon treatment of entrectinib as an example. It is noteworthy that AMP helped confirm tricky atypical unusual NTRKs, ALK, ROS1 specimens showing misleading FISH results.
Questions and Discussion
How do you best integrate fusion testing into mainstream testing? There still seems to be a gap where solutions are not detected until very late in the testing process.
I would say the lab running in parallel DNA and RNA makes it a bit more cumbersome, obviously is not ideal. You would want one assay that perform DNA and RNA at the same time. Illumina in their Two-Seq I think 500 kit can do both RNA and DNA at the same time, I believe. I know Archer does have some panels that perform RNA and DNA, same time, I think it comes down to user education. And, you know, I know it's cumbersome to run two assays in parallel, but we've done this over the years, and we found the it just worked quite well. It's one extraction, two separate techniques. And we are able to focus on just the DNA and just the RNA. It makes the analysis on the bioinformatics easier as well. But I would say you know, user education, get everybody on board in terms of adopting NGS. But yes, I agree. I would say most labs don't do this routinely.
From my experience, we're getting 50% actionable calls and fusion detections, whether it drives diagnosis or treatment. That's something that will be missing if you only focus on DNA alone. And trying to detect fusions from DNA I hopefully I've brought the point home is that it's very challenging if you're dealing with just introns very challenged to map very challenged to know what's functional or not. And I didn't get into all the details, but a lot of our colleagues have sent over cases that have been detected on DNA for us to confirm an RNA because RNA side, were able to detect the breakpoint at the RNA level, and pairing up two axons next to each other, we have the information of gene expression, but to also whether the two axons are in-frame. So that's a functional readout that we actually adhere to, when we try to find a fusion that is relevant for clinical reporting is that if it's in frame most likely as biologically real and meaningful for the patient.
What do you see as a possibility for finding more consistency among testing and protocols? Is there's a network for finding rare validation samples? Is there a way that you could see feasibly educating more pathologists and practices and making this fusion detection common practice?
A network would be awesome. I think in terms of fusion detection from solid cancer types, I think, proficiency testing and validation samples and whatnot, is still being worked on. And I'm talking about this 5-10 years later than we'd launched the assay at MGH. A network-raised setting involving education, sharing of knowledge sharing of samples and validation experiences would be vital. Helping out a small lab alone, without the supporting cast of colleagues and folks who've done this for 10 years like us, I take every opportunity I can do to talk and share our experience. It does help the next lab in terms of the point that we can have a more formal setting, whether it's through webinars like this, meetings and whatnot. Also sharing of samples, knowledge, and content is vital. And in a network setting, I think that's the best-case scenario to help democratize this type of technology. We have quite an experienced base, at least on this one technology that can potentially resonate with other technologies out there as well.
What do you do, clinically, to advocate testing? Do you test everything at once or would you stage it sequentially?
Right now, we don't we don't stage as much anymore as I showed you for ROS1 target where we'll do breadth, if it's negative up front. Nowadays, we just run the 55-gene panel up front, and most solid tumors coming in definitely for lung for brain or colon. They just get both DNA and RNA based assays at the same time. The DNA assay is 91 genes focusing more on the point mutations, the hotspots and whatnot, copy number change, and then the RNA as they mainly it's for the intergenic fusions as well as the intragenic. So namely, majority of samples came from lung cancer, EGFRvIII and GBM, so we just do both up front from-the-get-go. About the next iteration of fusion panel, I didn't highlight but we also have a fusion panel for AML and lymphoma, as well as a sarcoma fusion panel that the same technology underneath just different mix in terms of primers and targets. For the next iteration of our solid tumor panel, we're thinking about just wanting sarcoma targets, along with the solid tumor target. Oftentimes we struggle over whether we should run it on the solid panel, or on the soft tissue sarcoma panel. Sometimes some of the-get-go is just a query. They don't know exactly what tumor does. It would be nice if they would do one run and be able to cover all the relevant soft tissue targets as well as solid tumor targets. The next generation will combine the two into roughly 78-gene panel. And that just makes things more efficient as well for the lab technicians to run it. In our practice at MGH, we run up front everything in parallel -RNA and DNA.
Do you currently change your testing patterns based on tumor types? Do you aggressively test for lung but less for other tumor types?
When I was at MGH we drove a lot of clinical trials and phase I and they're very much so invested in the technique and in the testing and doing this is the right thing to do for our patients. Across the board, I think, the plot to show you in terms of the distribution across different tissue types, mainly is the type of disease testing that we do here based on clinical volume. If splitting this between lung, thyroid, brain or so, it makes it very hard from a management standpoint for technicians going into the freezer grabbing the lung primers grabbing the thyroid ones. I think I'd rather not do that. This confuses them more and leads to potentially false negative results and mistakes that could be had. But when you're talking about homology- based diseases versus solid tumor versus sarcoma, I think it is potentially biologically meaningful and also practically meaningful to separate families at that level. We do have a Hema-based panel, a solid tumor-based panel. And like I mentioned, on the fusion side, a lot of it crossover in terms of trying to detect something that is meaningful for diagnostic purposes on the soft tissue side, it's probably better just to do the 70 genes up front with one panel. And on the DNA side, we don't have a sarcoma DNA panel. We just run the solid tumor panel for those cases as well. I think that's where you can draw the line-solid tumor versus blood-based cancers and lymphomas. Going more granular than that. I think at some point, it makes it challenging at some point, it's just not worth it. And the savings to be had from cleaning out a 10 to 20 genes as a minimum. Also validation wise, validating one assay for all is a lot easier from a technical standpoint, and also from a professional standpoint.
Which technique would you suggest to detect gene fusions and liquid biopsies?
We had tried for a year or two, to extract RNA from plasma. And I know maybe there are some studies out there that say otherwise, but we were not successful. Either labile, very transient and even if you get some, I think the RNA is so short, it is not amenable to a technique like AMP here to be able to detect. We do need at least 20 bases right to anchor the primers on the known end. And we need at least 20 to 30 bases on the other side. Beyond that you want much longer sequences as well to drive the random start site so that we have unique reads to power your bioinformatic detection. From plasma, first off, we're already hampered by the size of the fragments. DNA is much more stable than RNA, and we get on average like 120 bases or so, if we're lucky in terms of circulating plasma DNA. We have not been successful in deploying this technique for RNA from plasma. We have been developing a DNA base to capture same thing where I am but targeting DNA from plasma. it's challenging because we have to code across the intron. Sometimes the breakpoint involves a very large intron. The mapping is is difficult. Because there's no RNA in circulation. We can't use an RNA technique. Maybe DNA is the way to go. Exosomes is another sort of medium that's been substrate that we could look into where maybe more pristine RNA reserved and encapsulated. I would say it's not mainstream. There's only one company I believe, I know, that's really pushing for this. Finally, circulating tumor cells is another one but that comes with its own challenges as well as that you really need a very solid technology to isolate this CTC. And if you can get to CTC and enough of them, perhaps, with pristine RNA, it will be very minimal to a technique like ours in terms of AMP or RNA Seq.
What about IHC to screen for NTRK fusion? Do you feel that it would be appropriate to do IHC and then request NGS?
This is my personal opinion, I think no, depending on the setting. If you're a small community hospital, small pathology practice., you have three professionals who are signing out and they can find out IHC if the volume is low, I think it makes sense to deploy IHC. There are antibodies for Trk A/B/C, suggesting three different IHCs to perform. If you have the large sample volume, I just think it does not make sense.
It is a challenge to have multiple different fusion partners. Which NTRK fusion, left right partner combination, do you see most frequently in your practice?
NTRK2 is quite rare, based on our experience while NTRK1 and NTRK3 are more prevalent, Implementing RT-PCR, the very simple PCR based assay, I think sensitivity clinically at best is like 10% within the cohort of patients who are positive for NTRK fusion. Consequently, you'll be missing out on quite large portion. The agnostic way of knowing that you have NTRK involved and unknown, not knowing the other side, is to resort to our technique. One of the problems is NTRKs are so rare, heterogeneous that we need a very nice NTRK technique that can be one probe for all to be able to achieve this clinical sensitivity that you want, so that you can pair up these patients with the right treatment. But no, I would say I'm hard pressed to find one that I would recommend t design one simple assay around. Unfortunately, I think it requires a highly multiplies technique kind of like NGS to be able to do.
You talked about turnaround time for assays, what turnaround times are oncologists Looking for currently? What are the possibilities of significantly decreasing those in the future?
I would say bread and butter wise for the most part, 90 to 95% of situations that we deal with, one to two-week turnaround time would be ideal. Patients don't come back frequently. For a new diagnosis, there's the anxiety of knowing what' s behind the scenes in terms of treatment selection and whatnot. Two weeks, I think is sort of a reasonable expectation for most labs. Sometimes it's not the lab itself, but the lab technique is the bottleneck when it comes down to acquiring the samples, from the outside institution, digging through the different blocks, which one has the most high tumor contents. Those are intangibles before the specimens hit the lab. Oftentimes these cause the delay in terms of the turnaround time.
In some settings, there are some targets that we do need the same-day turnaround time, such as AML. There are some clinical trials with, for example, decision of chemotherapy or molecular targeted therapy. The oncologists do appreciate if we can do it within the day. In this case, Next Gen Seq is not going to cut it due to library construction, sequencing. The three to five-day turnaround time is a bit too long for them. In that scenario, we've gone with quantitative real time PCR techniques that can be extraction done in 30 minutes to an hour or put it on the instrument an hour or two later, we get the result. We just can't do next gen sequencing and focus on that one target.
The improvement can be made upon optimized techniques on the lab enzymes that work faster, less steps for the technicians to handhold during the process. The other piece I think we can improve upon is the sequencing. At the end of the day, I think three-day turnaround is at best what we can look at if you're going to use a more conventional NGS techniques.
What are your thoughts on the US and international adoption of the Archer approach?
Even though I'm involved with a company, as a consult with them as my disclosure, I actually don't have details on that. In general Europe is usually slower to adopt. I think the healthcare systems are quite different between Europe and the US and even in the US is not easy. Europe probably has other intentions in terms of reimbursement and political action really compared to United States. I don't know the answer in terms of how popular this assay is in Europe, versus the United States. But we are a believer in our assay, we deployed it, and we are excited to know that we're able to help a much larger patient population by being able to deploy this across the entire world.
I would venture to say we're probably one of the first clinical labs in the United States to launch a targeted RNA Seq panel for the purpose of fusion detection. Some of the drawbacks of implementation is reimbursement. We were doing this as a laboratory developed tests. I think the NTRK story with entrectinib, IGNYTA would not have been possible without our technique that drove the screening in their clinical testing to get this drug approved, and then it opened up the avenue for other inhibitors across other companies to also get on the way.
 J Clin Pathol. 2019 Jul; 72(7): 460–467.
 Nat Med. 2014 Dec;20(12):1479-84
Detects TRKA, B and C
Turnaround time 1–2 days
May not be specific for NTRK gene fusion as it detects both wild-type and fusion proteins
Possible false positives
Possible false negatives for fusions involving TRKC
There is no standardization of scoring algorithms
The location of the target within the cell is visible
Several targets can be detected in one sample using several fluorophores
Requires knowledge of only one of the two fusion partners when using break-apart probes
NTRK gene fusions with unknown partners can be detected using break-apart FISH
FISH is readily available in most laboratories and institutions
The target sequence must be known for conventional FISH otherwise three separate tests are required for NTRK1, NTRK2 and NTRK3
Complex chromosomal translocations can result in false positive signals
False negative results may be above 30%
High sensitivity and specificity
Low cost per assay
Target sequences must be known (i.e., cannot readily detect novel fusion partners)
A comprehensive multiplex RT-PCR assay might be challenging because of the potentially large number of possible 5’ fusion partners
May detect novel fusion partners (depending on the assay used)
Can be used to evaluate multiple actionable targets simultaneously while preserving limited tissue
Currently used for NTRK testing
RNA-NGS is preferred over DNA-NGS as sequencing for RNA-based testing is focused on coding sequences not introns
Commercially available DNA-based NGS platforms may not be capable of identifying all NTRK gene fusions, especially those involving NTRK2 and NTRK3, which have large intronic regions
DNA-NGS is limited by intron size
RNA-NGS is limited by RNA quality