top of page

Testing for Clinically Actionable Mutations Including TRK in Sarcomas


Presented by Alex Lazar

This event is supported by an educational grant from Bayer

TRK in Sarcomas

Scientific Summary by: Bernice Tsai, Celerity Education Scientific Writer

Sarcoma Background and Classification


The diagnosis and treatment of sarcomas can be challenging as they are a heterogenous group with more than 50 histologic subtypes each with varying clinical phenotypes and behaviors. Some of the tumors are unclassifiable. There are many benign entities, which can be confused for sarcomas.  

Histologically, the diagnosis of sarcoma can be made on the basis of morphologic pattern.  Sarcomas exhibit a wide range of morphological varieties, encompassing round cells, epithelioid cells with abundant clear cytoplasm, and spindle cells. The histologic or morphologic description can be helpful in defining unclassifiable tumors. Yet the shape of cells does not always help determine the malignant potential of tumors.  

Sarcomas can be classified by histogenesis or line of differentiation. Thus different sarcomas will show differentiation toward different tissues in the body. For instance, angiosarcoma shows endothelial differentiation. Malignant peripheral nerve sheath tumor (MPNST) shows peripheral nerve differentiation. Liposarcoma shows  adipocytic differentiation. In contrast, synovial sarcoma doesn't show differentiation toward synovium. And that's just an accident of history and that it was first described in the knee near the synovial lining. For alveolar soft part sarcoma (ASPS), Ewing sarcoma, we don't really know what type of differentiation these show. These unclassified tumors are often associated with unique fusions that drive their biology and differentiation in ways that we don't normally encounter in normal tissues in nature.

To classify sarcomas by line of differentiation or first line of classification is done by looking under the microscope. Additionally, immunohistochemical staining (IHC) often aids in the identification of the presumptive tissue of origin. Some IHC markers are more characteristic than others, and it is the spectrum of markers examined that determines the histological subtype. Here are a few examples-atypical lipomatous tumor with typical nucleus but showing clear adiposity differentiation. Pleomorphic liposarcoma shows high grade sarcoma in the background within these scattered lipoblasts. This is an alveolar rhabdomyosarcoma (A-RMS) We may need additional markers such as myogenin to show that this is rhabdomyoblastic differentiation.


Per biological potential, sarcomas are further classified as benign (e.g., lipoma), intermediate, or malignant (e.g., synovial sarcoma), and intermediate tumors are designated as either locally aggressive (e.g., desmoid) or rarely metastasizing (e.g., angiomatoid fibrous histiocytoma). 

The French Federation of Cancer Centers Sarcoma Group (FNCLCC) system is most commonly used for grading malignant class of sarcomas. It is based on mitotic counts, tumor differentiation, and tumor necrosis. And this correlates well with outcome.


The genomic classification of sarcomas falls into three classes. Many sarcomas with simple karyotype have been found to be associated with recurrent translocations such as synovial sarcoma (SS), rhabdomyosarcoma (RMS), dermatofibrosarcoma protuberans (DFSP), and others. The tumors with simple karyotype have activating gene mutations as well such as GIST and desmoid. The complex karyotype sarcomas are usually associated with TP53 disruption undifferentiated pleomorphic sarcoma (UPS), myxoid fibroblastic sarcomas (MFS), pleomorphic liposarcoma are examples of these. These don't have usually recurrent genomic features associated with them. The well-differentiated (WT) and dedifferentiated (DD) liposarcomas go to intermediate class which have recurrent genomic features.  


Sarcoma and Cancer Genome

Analysis of the 27 cancer types revealed that the median frequency of non-synonymous mutations across cancer types. Tumor types are ordered by their median somatic mutation frequency, with the low frequencies found in Ewing sarcoma, and the high in melanoma induced by carcinogens such as tobacco smoke and UV light.1


The oncogene drivers range from single base substitution (or point mutations), deletions/insertions, and translocations.2For most neoplasms, translocations are relatively uncommon events. In translocation- associated sarcomas, translocations are more common and disease- defining.

These point mutations sometimes will have particular patterns.(1) Different tumor types segregate into different compartments based on their mutation spectra, for instance,  melanoma shows a distinct pattern reflecting the frequent C→T mutations caused by misrepair of UV-induced covalent bonds between adjacent pyrimidines. These types of patterns aren't particularly relevant to sarcoma.


Most of the mutations in cancer either occur outside of the genes or occur within genes. They will often occur in very large genes like replicating genes, or genes that have many different versions such as olfactory receptors1. These are probably not important for driving the biology of the tumor.

The certain types of genes are important for driving the biology of tumor. The PIK3CA and IDH1 are two oncogenes, which tend to be recurrently mutated at the same amino acid positions, constantly turning the protein on as a consequence. The second class is tumor suppressor genes. RB1 and VHL are two examples of them, which are mutated through protein-truncating alterations throughout their length.2 These are mutations that deactivate the function by introducing nonsense mutations that halt the reading frame.

While lots of different mutations can occur in cancer, there's only a few that drive the biology. When we model this across multiple different cancers, common adult tumors, such as pancreatic, colorectal, breast, and brain cancers, appear to have three to six mutations that are important for driving the biology of the cancer at least early on.2

Mutations in all of driver genes do one thing: cause a selective growth advantage, either directly or indirectly. Moreover, there appears to be only a limited number of cellular signaling pathways through which a growth advantage can be incurred.2 The signaling pathways produce the features in cancer that associate with malignant phenotype.   

Sarcoma genomics are associated with translocations specifically. The Cancer Genome Atlas (TCGA) analysis looked at DNA mutations, copy number alterations, gene expression, DNA methylation, miRNAs, RPPA and clinical data, and performed an integrated analysis on all of the above.3 The TCGA has completed f genomic characterization of over 1,000 tumors from 33 cancer types. One of these was sarcoma.

For The Cancer Genome Atlas (TCGA) sarcoma analysis, we focused on 6 major adult soft tissue sarcomas, including 5 with complex karyotypes: (1) dedifferentiated liposarcoma (DDLPS), an undifferentiated sarcoma usually arising in as-sociation with well-differentiated liposarcoma and characterized by 12q13-15 amplification; (2) Leiomyosarcoma (LMS), showing smooth muscle differentiation, arising in both gynecologic(ULMS) and soft tissue (STLMS) sites; (3) undifferentiated pleomorphic sarcoma (UPS), lacking any defined line of differentiation; (4) myxofibrosarcoma (MFS), showing fibroblastic differentiation with myxoid stroma; (5) malignant peripheral nerve sheath tumor (MPNST), which arises in peripheral nerves. The sixth type was a simple-karyotype sarcoma, synovial sarcoma (SS), defined by the translocation t(X;18) (p11; q11).4 Due to small number of MPNST or SS, they are left out of analyses.

On a 2-D tumor “map” visualization of high-dimensional mRNA expression data, leiomyosarcomas (STLMS and ULMS) clustered together, while synovial sarcomas also formed a distinct cluster. The remaining sarcomas largely clustered together with mixing of DDLPS, UPS and MFS, grouped together as a smear.4

Sarcomas with simple karyotypes harboring specific genetic alterations (translocations, activating mutations) and those with complex karyotypes, which were characterized by frequent SCNAs. In contrast, SS displayed very few SCNAs or mutations.4 No sarcomas with both high numbers of copy number alterations and high number of mutations are identified.  The tumors appear to be either driven primarily by copy number alteration, or by gene mutation.

Of the mutations, 90% are attributable to COSMIC5 (53%) and COSMIC1 (37%), which are pre-existing mutations or copies in DNA.4 They are usually not driver mutations but are carried along with the first cell that transforms, likely representing passenger mutations. Their contributions to the mutational profiles are found to be correlated with age at diagnosis.

Adult soft tissue sarcomas have low somatic mutation burden. We applied MuSiC analysist o whole-exome sequencing (WES) data to identify significantly mutated genes (SMGs), i.e., genes with a statistically higher-than-expected mutation prevalence across the entire cohort. This identified only 3 SMGs:TP53, ATRX, and RB1. TP53 mutations were most prevalent in LMS (40 of 80). RB1 mutations were seen in LMS, UPS, and MFS Most mutations are frameshifts in the tumor suppressor genes.4 Sarcomas look like a disease taking your foot off the brakes, i.e. losing tumor suppressors, rather than stepping on the accelerator with proto-oncogenes.

Sarcomas can have varying degrees of copy number alterations that can correlate with outcome. Looking at what the nuclei look like under the microscope, we found that that larger more pleomorphic nuclei are associated with more genomic changes than those that are smaller and more uniform.

Sarcoma groups such as LMS, MFS, UPS, the DDLPS, synovial sarcoma, differ from each other when looking at them at the molecular levels, in terms of DNA copy number, DNA methylation, RNA expression, or mRNA expression, and so on. There are many ways of differentiating these groups.    

Molecular Diagnostics

Karyotyping or classical cytogenetics is the historical gold standard, but it has some problems in detecting complex karyotypes or cryptic translocations. It requires fresh tumor tissues. Having skilled staff interpreting karyotypes is a challenge too. Here is an example that karyotyping helps identify a clear cell sarcoma, with 12-22 translocation. For complex karyotype sarcomas, it is sometimes more difficult to figure out translocation and whether it is a driver. Spectral karyotyping deconvolute things. Here is a case of desmoplastic small round cell tumors, within 11-22 translocation that fuses WT1 and EWS1.

The other way of detecting translocations is by FISH. Two methodologies are developed- the break-apart probes which are more common, and fusion probes. A wide variety of tissues can be used. We don't have to have fresh tissue to grow like we do for karyotypes. It can be performed on paraffin embedded materials; fresh tissues are not mandatory. Failure to Hybridization is perhaps an issue. however. There's an internal usually an internal negative control of the normal tissue. The limitation of FISH is that it only indicates   which locus is rearranged and it does not inform which fusion partner is. Here ‘s an example FISH probe EWSR1 locus in clear cell sarcoma. As mentioned by FISH we are not able to figure out fusion partner. Other morphological immunohistochemical features would help separate these.


PCR-based detection is very sensitive, and it conveys on both participants in the fusion gene. The challenge behind it is the prior knowledge of exact fusion point in product, where the breakpoints are and how they're fused together, in order to design proper PCR primer and for some cases that can be very complex. For instance, for the dermatofibrosarcoma (DFSP), if you want to design RT-PCR for COL1A1 that has more than 50 axons in it. It can require a lot of probes and a multiplexed solution. An example of clear cell sarcoma is presented to illustrate that the appropriate primer against both participating genes is designed in order to capture EWSR1-ATF1 fusion.

There are three common NGS approaches. The first one is hybrid capture that brings down the targeted genes using bait of RNA or cDNA. The bioinformatics is applied in identifying fusions. The amplicon-based approach relies on PCR-amplification of target regions of interest using sequence-specific primers and probes. In other words, the prior knowledge of breakpoints and fusion partners is a pre-requisite. The focused bioinformatics is examining whether the interested genes are valid. The third approach is a hybrid of the above, called semi-anchored PCR where we use one primer for known fusion partner and another universal primer for the other partner.

The molecular diagnostic can assist classify the malignant adipocytic tumors. Within liposarcoma segments, the well differentiate and dedifferentiate liposarcoma both have amplifications of the 12 q 13-15 locus.  By doing a FISH, we would see an amplification of the 12q13-15 locus with containing both MDM2 and usually CDK4 as in CGH array.

Myxoid-round cell liposarcoma is a subtype of liposarcoma characterized by the presence of the reciprocal chromosomal translocation t(12;16) (q13;p11). This translocation creates the FUS-DDIT3 chimeric gene.  We can do break-apart FISH looking for rearrangement of the DDIT3 locus. In addition, there's a lot of different fusion types that can occur here, PCR primers can be designed to be able to detect the vast majority of them. RT-PCR and NGS are also alternatives for detecting these types of fusions. Pleomorphic liposarcoma has a complex karyotype and often has TP53 mutations associated with it.

The nodular fasciitis is one of the most common misdiagnoses in sarcoma. To avoid making a mistake, a small needle biopsy can be performed, if appropriate, to assess whether particular translocations associated with it are present. 

NTRK in Sarcomas

NTRK gene fusions involving either NTRK1, NTRK2 or NTRK3 (encoding the neurotrophin receptors TRKA, TRKB and TRKC, respectively) are oncogenic drivers of various adult and pediatric tumor types. The frequency of these fusions varies from <1% in cancer types including lung, colorectal, pancreatic and breast cancers, melanoma and other solid or hematological cancers, up to 25% in tumors including thyroid, spitzoid and gastrointestinal stromal tumors, to >90% in rare tumors types, specifically secretory breast carcinoma, mammary analogue secretory carcinoma (MASC), congenital infantile fibrosarcoma and cellular which the NTRK fusions are considered practically pathognomonic. Almost all studies describing TRK signaling refer to the MAPK, PI3K and PKC pathways as the main downstream effectors responsible for the activation of neuronal survival and differentiation pathways.5


The history of studying NTRK goes way back to the 50s where neuron growth factor was first discovered in the 50s. And then fusions with these gene were discovered as oncogenes in the early 80s. These intended to accumulate first and colorectal cancer then papillary thyroid cancer then an infantile fibrous sarcoma, and a variety of different genes.5 They are binding to their agonists and then also in their kinase domains. The NTRK fusions which the 3′-region is fused with a 5′ region of a fusion partner,  result in a chimeric receptor protein with dimerization and consecutively uncontrolled activation of the TRK kinase domain. The growing number of genes have been discovered in NTRK fusions.

The fusion partner provides a new extracellular domain and provides a domain that allows the proteins to dimerize so that consecutive signaling can be initiated without binding of agonists or the ligand anymore. That’s what drives the biology of tumor. NTRK inhibitors are inhibiting the kinase domains so that even though they   are turned on consecutively, they're not able to signal.

ALK works alike in sarcoma, and across lots of other tumor types as well. Here is an example of an epithelial IMT, where ALK is fused with RAMBPT2, which targets the nuclear membrane and causes consecutive signaling here. Like NTRK, ALK can be seen fused with lots of different genes and in lots of different neoplasms both in sarcomas and and non-sarcomas.

Larotrectinib and entrectinib are furthest along in clinical development. Entrectinib is an orally available pan- TRK inhibitor with additional activity against ROS1 and ALK6. Larotrectinib is a potent and selective inhibitor of all three TRK proteins.7The waterfall plots of both therapeutics shows efficacy particularly in significant tumor shrinkage.  The swimmerplots of both show the duration of action or the amount of time that disease is controlled. That response of NTRK fused neoplasms to these agent suggests why it is important to detect these fusions.

In general, NTRK mutations can be represented as missense, or point mutations, amplifications, deletions, in-frame fusions, and rearrangements near the NTRK locus without a productive fusion.  Among various types of mutations, the in-frame fusions are activating and predictive of response to TRK inhibitors while NTRK gene mutations do not lead to benefit with TRK inhibiting drugs.

The NTRK fusions so far have been discovered in several types of sarcomas. For example, infantile fibrosarcoma has a disease defining NTRK fusion involving specifically NTRK3-ETV6. They are also seen in infantile, or congenital sarcoma, lipofibromatosis-like neural tumor. uterine undifferentiated sarcomas, or fibroblastic muscle tumors with an unusual phenotype, wild-type gastrointestinal stromal tumor, also unclassified sarcomas in children and young adults.

Christian Curcio shared a 21 year-old male case with a para-vertebal mass. The CD34+ and S-100+ protein are both positive while pan-TRK antibody is positive.   It turned out LMNA-NTRK1 lipofibromatosis-like neural tumor. Typically, this is a locally recurrent disease that does not metastasize. But this particular case was locally aggressive. Sometimes the patient can benefit just from reducing the tumor size by inhibiting NTRK.

Christina Antonescu and Chris Fletcher reported NTRK1 oncogenic activation through gene fusion involving LMNA defines a novel and distinct subset of soft tissue tumors resembling lipofibromatosis (LPF), but displaying cytologic atypia and a neural immunophenotype, provisionally named LPF-like neural tumors.8

One study is performed to distinguish NTRK fusion-positive uterine sarcomas from leiomyosarcoma and undifferentiated uterine sarcoma. NTRK rearrangements were detected by fluorescence in situ hybridization (FISH) and/or targeted RNA or DNA sequencing in 4 undifferentiated uterine sarcomas with spindle cell morphology. FISH and/or targeted RNA or DNA sequencing in the study group showed TPM3-NTRK1, LMNA-NTRK1, RBPMS-NTRK3, and TPR-NTRK1 fusions. All tumors were composed of fascicles of spindle cells. Desmin, estrogen receptor, and progesterone receptor were negative in all tumors, while pan-Trk was expressed in all tumors with concurrent TrkA staining in 3 of them. NTRK fusion-positive uterine spindle cell sarcomas constitute a novel tumor type with features of fibrosarcoma9; patients with these tumors may benefit from Trk inhibition. TrkA and pan-Trk expression in leiomyosarcomas is rare and does not correlate with NTRK rearrangement.9

In the larotrectinib trial, the tumor of a 41-year-old woman with soft-tissue sarcoma metastatic to the lung was found to harbor an LMNA-NTRK1 gene fusion as determined by an in situ proximity ligation assay.10

Multiple studies have indicated that this pan- Trk antibody that will bind to TrkA, TrkB, TrkC. can be useful as a screening tool.11,12 The pan-Trk immunohistochemistry may be a useful initial enrichment tool in certain indications, such as colorectal cancer, but not in others, such as neuroendocrine tumors, where approximately 50% of lesions appeared to express Trk proteins in the absence of NTRK gene fusions. Therefore, particularly in tumors with a low prevalence of NTRK gene fusion events or cell types with baseline expression of TRK proteins, immunohistochemistry positivity should be used with caution as a surrogate biomarker for NTRK gene fusions, and positive cases should be further investigated by reflex NGS or molecular cytogenetic testing.12Diffuse pan‐Trk immunoreactivity is a highly sensitive but not entirely specific diagnostic marker for infantile fibrosarcoma, and may be helpful in selecting patients for TRK‐targeted therapy. As expected, lipofibromatosis‐like neural tumors, which harbor NTRK1 fusions, also show diffuse pan‐TRK immunoreactivity.13 The sensitivity of these is around 80 to 90%, based on studies we look at while  the specificity, general speaking, is relatively poor.

The gastrointestinal stromal tumors (GISTs) show differentiation similar to the interstitial cells of the Cajal. They are identified mainly by expression of the KIT protein and frequently harbor activating mutations in the KIT or platelet-derived growth factor receptor alpha (PDGFRA) genes. The kit mutations and PDGFRA mutation occur in analogous regions of the of the proteins. We can detect these by variety of methods, including next generation sequencing. On top of KIT and PDGFRA mutations, we also see in a subset of GIST SDHB loss, NF1 loss in GISTs.  Moreover, oncogenic drivers such as BRAF mutations, FGFR1 fusions, and NTRK fusions are found as well. The NTRK fusions in GISTs fall on wild-type tumors that lack of KIT mutation, PDGFRA mutation or SDHB loss.

Looking at sarcomas on tissue microarrays, we found that across a wide variety of different translocation and non-translocation associated sarcomas, about 12% of sarcomas had significant expression of pan-TRKs. But only about 10% had a NTRK fusion. Since we didn't sequence everything, the sensitivity and specificity are problematic. Sometimes we encounter tricky cases such as desmoplastic small round cell tumor, which has an EBSR1-WT1 fusion and therefore is not going to have NTRK fusion. However  it stains strongly with TRK positive in virtually every single case. That's because this fusion causes upregulation of in NTRK3 RNA causing overexpression of NTRK3 or TrkC.

Considering the heterogenicity characterized by NTRK fusions, karyotyping, FISH, or RT-PCR are not very technically feasible. For example, within the LMNA-NTRK fusions, the tumors involve a small inversion that's difficult to detect by karyotype. For FISH and RT-PCR, there are challenges in designs of probes and primers. Therefore, NGS testing using cDNAs or RNAs are preferred over other molecular approaches. As addressed earlier in the talk, hybrid capture, amplicon, and semi-anchored approaches coupled with bioinformatics methodologies all can be implemented to detect NTRK fusions.

In summary, sarcoma genomics are complicated. Classifying or diagnosing different subtypes of sarcomas is a huge challenge. The NTRK fusions are seen in a relatively rare subset of sarcomas.  They occur in which tumors show negative for other types of drivers. Pan-TRK immunohistochemistry can be helpful, but the specificity can be a challenge and the sensitivity leaves gaps. For NTRK fusion detection strategies, NTRK partner agnostic is probably preferred over those that demand the knowledge NTRK is used to fuse to.


Questions and Discussion: With Alex Lazar

If you could change one thing about the current state of NGS testing in sarcomas, what would it be?

 Although nowadays we have strong capacity and technical expertise of NGS diagnostic, it is perhaps still discouraging to find NTRK fusions are relatively rare with pre-test probability of 1%. The first suggesting is to receive more reimbursement for NTRK testing to maintain the continuous efforts. Ideally a streamlined platform of genomic profiling can be used to screen sarcomas and other tumor types. Hope the platform is reimbursed as well.

Do you often see any other fusions that occur with TRK in sarcomas or these are usually the only ones when you see them?

I was making points about rearrangements versus productive fusions. In terms of in terms of productive fusions, one great examples in the literature is in Ewing sarcoma driven by EWS-FLI1 fusion, having an NTRK fusion on top of that. The NTRK fusion does not end up making productive fusion oncogene. We would not expect NTRK fusion in a synovial sarcoma with known S18-SSX fusion. Taken together sarcomas with characteristically other types of fusions are not places we would expect to see NTRK fusions, as long as the other fusion that characteristically there is present and documented.

The NTRK mutation or other similar driver mutations tend to be in both the primary and metastases. Any examples you've seen when this isn't true?

For fusion, the end on this is pretty low because there's not  whole lot of experience with  the fusions but in general, I personally have not seen NTRK either being lost between the primary and the metastasis or being gained in the metastases, which are the two main patterns that you might expect to see. In general sarcoma are cancers that are driven by gene fusions. Those tend to happen early in the biology, and they tend to be maintained both in the primary and in them and then the metastasis. So, from biological perspective, if you want to look for a documented a gene fusion, while you always want to use the most recent metastatic disease when looking for these fusions. It doesn't matter whether you look at the primary or look at the metastasis just to document the fusion.

Which TRK antibody do you recommend?

In principle, we would pick one that detects all three of TRKs. You can follow up by email for more details of antibody we use.

How early inpatient diagnosis would you want to be aware of a targetable mutation like NTRK, even if a patient has a surgical option?

For good laboratory management, we will not do a test until we are going to use it to make a clinical decision. For the patient, who have a surgical option, we do not change approach to therapy and molecular testing is not a must.  Only if we have hypothetical sarcoma platform, hopefully reimbursed, it is reasonable to characterize the primary tumors  at that point, enabling us  to be thinking about what treatment options might be down the road if there is tumor progression.

Have you heard of any responses to TRK drug on TRK overexpression only in sarcomas?

it doesn't look like that overexpression is going to be a good marker for response. We have evidence that in some of the early versions of the trials, there were patients who had, amplification and/or overexpression of TRK by variety of different mechanisms. And those with overexpression did not respond. I'd have to have to go back and look at cases to see whether sarcomas were included in that class or not. this class of drugs do not seem to work well against overexpression of the protein and that's not uncommon, we see the same thing and with a variety of different tyrosine kinase inhibitors. For instance, overexpression of KIT does not make a necessarily great target for inhibition with KIT inhibitors.


  1. Nature. 2013 Jul 11;499(7457):214-218

  2. Science. 2013 Mar 29;339(6127):1546-58.

  3. Cell. 2018 Apr 5;173(2):291-304.e6.

  4. Cell. 2017 Nov 2; 171(4): 950–965.e28.

  5. Nat Rev Clin Oncol. 2018 Dec;15(12):731-747.

  6. N Engl J Med. 2018 February 22; 378(8): 731–739.

  7. Cancer Discov. 2017 Apr;7(4):400-409

  8. Am J Surg Pathol. 2016 Oct;40(10):1407-16.

  9. Am J Surg Pathol. 2018 Jun; 42(6): 791–798.

  10. Cancer Discov. 2015 Oct;5(10):1049-57.

  11. J Mol Diagn. 2019 Jul;21(4):553-571.

  12. Am J Surg Pathol. 2017 Nov; 41(11): 1547–1551.

  13. Histopathology. 2018 Oct;73(4):634-644.

Sarcoma Background
Cancer Genome
Molcular Diagnostics
NTRK in Sarcomas
Q&A with Alex Lazar
bottom of page