Pan-TRK Immunohistochemistry: An Example-Based Practical Approach to Efficiently Identify Patients With NTRK Fusion Cancer Open Access

Because IHC is considered a screening method, we must use the highest-sensitivity assays. This is particularly relevant in this setting because of the low prevalence of NTRK fusions: patients with a negative IHC result are unlikely to be tested again for this biomarker. Therefore, the universal preanalytic recommendations for IHC are also valid here.²⁰  A properly fixed tissue block should be selected to avoid heterogeneous staining. In therapy-naïve patients, intertumor heterogeneity should not be an issue, because NTRK fusions are founder alterations.¹³  The expression of the 3 TRK proteins (TRKA, TRKB, and TRKC) must be identified simultaneously using a pan-TRK antibody. The clone EPR17341 is the most frequently used and well characterized.^21–23  A positive control must be included in all the slides to ensure the proper functioning of the analytic phase, minimizing the risk of a false-negative result. The most accessible positive control is the appendix.²²  The neural structures (ganglion cells) in its wall should be positive (positive control), in contrast to the rest of the completely blue tissue (negative control) (Figure 1). To facilitate the presence of a relevant number of these nerves, there are 2 measures that help: (1) do not select an appendix with appendicitis and (2) include 2 sections of the positive control in each slide to increase the likelihood of having these neural cells present in the section. The staining of the external positive control does not guarantee the absence of preanalytic problems. The only reliable pan-TRK IHC in situ positive controls are peripheral nerves (Figure 2). Because most slides will lack a pan-TRK IHC in situ positive control, pathologists should continue their search for NTRK fusions whenever there is a reason to question the optimal preanalytic conditions of the specimen (eg, decalcified samples or large surgical resections) or in specific circumstances that increase the likelihood of finding an NTRK fusion (see Table; eg, pathognomonic/characteristic histology, nuclear IHC staining, and solid tumors in therapy-naïve patients without driver gene mutations/fusions/amplifications after limited biomarker testing).^10,11,16,24 

The interpretation of pan-TRK IHC may be more challenging than that of other IHC-based biomarkers. Findings from a literature review are difficult to harmonize because of the different preanalytic and analytic conditions of the published series. Although there is no interpretation guide for any of the available IHC clones, the literature does suggest the level of IHC positivity that requires genomic confirmation. A positive cutoff has been defined in large pan-tumor series as staining above background in at least 1% of tumor cells.^14,15 

Cytoplasmic staining is seen in most NTRK fusion–positive tumors,^21,25  but it is also the staining pattern most frequently observed in false-positive cases.¹⁴  In addition, 3 different subcellular staining patterns have been described with a certain degree of fusion partner specificity: nuclear (eg, ETV6), perinuclear/nuclear membrane (eg, LMNA, MUC2), and membranous (eg, TPM, TPR, TRAF2).^21,25–27  The reduced IHC sensitivity that has been reported for the detection of NTRK3 fusions^14,15  could be due to the overrepresentation of ETV6-NTRK3 fusions, as this nuclear pattern of pan-TRK IHC staining can be quite heterogeneous (see below: secretory breast carcinoma, secretory carcinoma of the salivary gland, and papillary thyroid carcinoma). Conversely, to the best of our knowledge, relevant false-positive pan-TRK IHC nuclear staining has been described in only a minority (8%–10%) of nonsecretory breast and salivary gland carcinomas,^28,29  always in a subdiagnostic manner (ie, focal and/or weak). Finally, it must be emphasized that there is pan-TRK physiologic cytoplasmic expression in neural and smooth muscle tissue and their malignant counterparts.²⁵  Therefore, tumors with these differentiations (for example, gastrointestinal stromal tumors, neuroblastomas, leiomyosarcomas, or glioblastomas) should not be screened with pan-TRK IHC.^11,25 

Pan-TRK IHC is not entirely specific for NTRK fusions, so positive pan-TRK IHC should always be followed by a second assay. It must be emphasized that even the most expensive and sophisticated methodologies have their own sensitivity and specificity issues.¹²  A clinically concerning situation would be a true-positive pan-TRK IHC result not confirmed by the orthogonal method. Knowing which positives are true for NTRK fusions is not intuitive, but certain histologic (ie, rare tumors with a very high incidence of NTRK fusions), immunohistochemical (ie, nuclear pan-TRK IHC) and molecular features (for example, solid tumors without driver gene mutations/fusions/amplifications or triple-negative colorectal carcinomas; see Table) can identify good candidates. Although it is beyond the scope of this paper to describe the advantages and disadvantages of the different methodologies (reviewed elsewhere^{8–12,25,30–32} ), some comments are helpful. Break-apart FISH can give false-negative results in NTRK1 fusions because of insufficient separation of the signals.¹²  The performance of reverse transcriptase–polymerase chain reaction can be limited because of the large number of partners that have been described.¹²  In routine clinical work, the use of FISH or reverse transcriptase–polymerase chain reaction is restricted to 3 situations: (1) confirming the alteration in those rare tumors where NTRK fusions are pathognomonic/very frequent, (2) confirming an NTRK3 fusion after a nuclear IHC result, and (3) addressing DNA/RNA failure of the NGS assay if no additional tissue is available.^12,25,30  Finally, even the most comprehensive NGS panel might not detect all NTRK fusions, so a precise knowledge of the width of the assay and its real-world performance can help rule out a false-negative NGS result.^9,12,30–32  Along those lines, it has been reported that there is a significant risk of false negatives when using a DNA-based NGS approach.^5,16,30,31  Conversely, the RNA quality of formalin-fixed tissue can influence the analysis of any PCR-based assay.¹²  An unbiased review¹²  of the different NGS approaches was recently released.

In summary, there are 3 questions that pathologists ask most frequently when scoring pan-TRK IHC: (1) Do I need to trigger orthogonal testing based on this level of positivity? (2) The confirmatory method result came back negative after a clearly positive pan-TRK IHC; is there something else I need to do? and (3) Are there any histologic or molecular features that can help me suspect a pan-TRK IHC false-negative result? In other words, when do we need to persevere after a negative or inconclusive pan-TRK IHC result? To further help pathologists answer these questions, illustrative examples are shown below, classified using a 3-tier approach: (1) rare tumors with a very high incidence of NTRK fusions (infantile fibrosarcomas and secretory carcinomas of the breast and salivary glands), (2) common carcinomas with an extremely low probability of harboring NTRK fusions (lung adenocarcinomas, colorectal carcinomas, and papillary thyroid carcinomas), and (3) tumor types with frequent and relevant TRK protein expression not associated with NTRK fusions (leiomyosarcomas, olfactory neuroblastomas, and adenoid cystic carcinomas).

Infantile fibrosarcoma (Figure 3, A) is a very rare pediatric tumor that is characterized by an ETV6-NTRK3 fusion in 80% to 90% of cases.^33–36  Accordingly, most cases show a characteristic pan-TRK IHC nuclear staining in addition to diffuse cytoplasmic positivity, given that ETV6 encodes a transcription factor (Figure 3, B).³⁴  However, diffuse pan-TRK staining can also be found in many other pediatric spindle cell tumors, either NTRK rearranged (even outside lipofibromatosis-like neural tumors; see below)^18,37  or not harboring any of the NTRK fusions (eg, primitive myxoid mesenchymal tumors of infancy, fibrous hamartomas of infancy, fibrosarcomatous dermatofibrosarcoma protuberans, synovial sarcomas).^33,34  Taking into account the overlapping clinical and pathologic attributes of NTRK-rearranged and nonrearranged tumors in this population, it is sensible to perform pan-TRK IHC in all pediatric patients with mesenchymal tumors.^18,34,38  Likewise, the expanding number of NTRK partners described in such tumors (and for this matter also in congenital mesoblastic nephroma, an NTRK-rearranged renal tumor of infancy) argues against the traditional use of ETV6 FISH as an orthogonal method, particularly if pan-TRK IHC nuclear staining is not clearly identified.^18,34,39  Therefore, the use of NTRK3 FISH or NGS has been recommended.³⁹ 

Secretory carcinoma (Figure 4, A) is a rare type of breast carcinoma (<0.05% of all invasive breast carcinomas) that is characterized by an ETV6-NTRK3 fusion.⁴⁰  In clinical practice, these carcinomas are frequently underdiagnosed because they are almost always triple negative or at most weakly estrogen receptor (ER)/progesterone receptor (PR) positive.^28,41,42  Because the identification of all breast secretory carcinomas in each surgical pathology practice seems unlikely, the use of pan-TRK IHC in both triple-negative breast carcinomas and invasive carcinomas with eosinophilic secretions could be a good strategy to identify these patients. Secretory breast carcinomas typically show strong cytoplasmic and nuclear pan-TRK staining (Figure 4, B).²⁸  However, because of suboptimal fixation, heterogeneous staining for pan-TRK IHC can be seen in resection specimens (Figure 4, C). Nonsecretory breast carcinomas can exhibit either faint cytoplasmic staining (7%–18%)^14,43  or barely visible nuclear staining in less than 5% of the cells (10% of cases in another series).²⁸  To rule out the influence of preanalytics on the intensity or the percentage of the nuclear staining, it is useful to select the best-fixed tissue block or compare the result of the pan-TRK IHC in the core needle biopsy versus the resection specimen if both are available.

Secretory carcinoma of the salivary gland (Figure 5, A), initially described as mammary analogue secretory carcinoma, usually contains an ETV6-NTRK3 fusion.^44,45  Pan-TRK IHC usually shows nuclear staining with a TTF-1–like or silver in situ hybridization–like staining pattern, associated with a weak cytoplasmic positivity in 100% of the cells (Figure 5, B). Although nuclear pan-TRK IHC is very specific,^29,46  pathologists should be aware of several facts: (1) cytoplasmic staining alone has also been described at a lower frequency in larger series,^46,47  and unfortunately this cytoplasmic pattern can also be found in many other benign and malignant salivary gland neoplasms (see below)^29,46–48 ; and (2) the nuclear pan-TRK staining can be weak and/or focal,^29,46,47  and a suboptimal sensitivity of pan-TRK IHC for NTRK3-rearranged tumors has been suggested.¹⁴ 

The estimated frequency of NTRK fusions in non–small cell lung carcinomas is less than 1%.^14,16,17  Most cases are adenocarcinomas (Figure 6, A), but squamous cell carcinomas and carcinomas with neuroendocrine differentiation have also been described.¹⁷  Signet ring cells can also be found, as in other rearranged lung adenocarcinomas.^4,17,49  Pan-TRK IHC–positive cases usually show diffuse strong cytoplasmic staining in tumor cells (Figure 6, B). NTRK fusions are usually mutually exclusive with other known oncogenic drivers, so frequencies are higher in therapy-naïve patients without driver gene mutations/fusions/amplifications.^16,50  If all advanced non–small cell lung carcinomas are not routinely screened for NTRK fusions using a large NGS panel, the use of pan-TRK IHC is particularly useful in such patients after the initial limited testing (ie, only results from main drivers or from a small NGS panel are available). A positive pan-TRK IHC result in lung adenocarcinomas is an indication for further NGS testing.^24,51  Nonetheless, given the large number of therapeutic targets in this tumor type, NGS should always be prioritized over IHC if tissue is limited. Finally, NTRK1 fusions have been proposed as a resistance mechanism to EGFR tyrosine kinase inhibitors.⁵² 

Screening all colorectal carcinomas (Figure 7, A) with pan-TRK IHC may not be feasible (0.02%–0.3% positivity in series with several thousand patients),^14,16,27  but a higher frequency of NTRK fusions (∼5%) in BRAF/RAS wild-type, microsatellite instability–high colorectal carcinoma provides a rationale for a more sensible approach.^27,53,54  Therefore, a molecular testing workflow for colorectal carcinomas has been proposed.⁵³  After the initial microsatellite instability/mismatch repair assessment with IHC, deficient patients are sequentially tested for MLH1 promoter hypermethylation and BRAF status before fusion testing is considered.⁵³  Pan-TRK IHC–positive cases usually show diffuse strong cytoplasmic staining in all tumor cells (Figure 7, B).²⁷  In addition, there could be positivity in other cell compartments in a partner-specific manner: membrane (eg, TPM3, TPR), nuclear membrane (eg, LMNA, MUC2), or nuclear (eg, ETV6).^21,26,27,54 

The overall frequency of NTRK fusions in thyroid carcinomas is around 2.3%,^13,14,16  but this figure can be much higher in children (26%).³⁵  In fact, kinase fusions have been linked to radiation-induced thyroid carcinomas in pediatric patients.³⁵  Most NTRK-positive thyroid cases are papillary carcinomas (Figure 8, A), but poorly differentiated and anaplastic carcinomas have also been described.^15,55  In fact, NTRK fusions may be associated with higher stage,^56,57  so a recommendation could be made to start by screening those papillary thyroid carcinomas with lymph node metastases. In addition to the characteristic cytoplasmic and nuclear positivity of the ETV6-NTRK3 fusion, the patterns of IHC expression include multiple nonnuclear NTRK1 and NTRK3 partners.^13,55,57  It is worth remembering that this nuclear staining can be quite focal, as in the case reported herein (Figure 8, B through D), but it is extremely specific for this rearrangement.

Olfactory neuroblastoma (Figure 9, A) is a malignant neuroectodermal tumor, and as such it can show very intense and diffuse pan-TRK IHC staining (Figure 9, B) not associated with an NTRK fusion. In fact, several lines of evidence involve the sonic hedgehog (SHH) signaling pathway in its pathogenesis.⁴⁴  Because NTRK fusions are particularly common in pediatric tumors, the frequent and intense pan-TRK IHC positivity of NTRK fusion–negative neuroectodermal tumors (eg, Ewing sarcomas or neuroblastomas) should caution pathologists against overexpectations when waiting for the NGS results.^14,58 

Strong and diffuse cytoplasmic pan-TRK IHC expression has been described in 2% of bona fide uterine spindle cell leiomyosarcomas (Figure 10, A and B) without NTRK rearrangements.⁵⁹  Because NTRK fusions seem to be singularly prevalent in gynecologic sarcomas,^59–63  and leiomyosarcomas are the most frequent uterine malignant mesenchymal tumors, it seems plausible that practicing pathologists will encounter this scenario. Identifying NTRK-rearranged soft-tissue and visceral sarcomas is extremely difficult for a number of reasons: (1) less than 2% of sarcomas contain an NTRK fusion⁶⁴ ; (2) in most centers diagnosis is not systematically interrogated in sarcomas through NGS, given the traditional lack of targeted therapies in these tumor types⁶⁵ ; (3) NTRK fusion–positive sarcomas display a very wide clinical, morphologic, and IHC spectrum^58,65 ; and (4) the sensitivity and specificity of pan-TRK IHC to detect fusions in sarcomas has been considered the lowest for all tumor types, because of the significant expression of the wild-type TRK protein in many of the mesenchymal lines of differentiation, including smooth muscle, as shown above.^14,58,59  Therefore, awareness that several morphologic patterns have been linked with kinase fusions in sarcomas can help pathologists narrow their investigation of these patients⁶⁵ : (1) lipofibromatosis-like neural tumors³⁷ ; (2) spindle cell tumors with S-100 and CD34 coexpression, sometimes resembling malignant peripheral nerve sheath tumors^19,63 ; (3) adult-type fibrosarcomas located in the uterus without desmin, ER, and PR expression^59,60 ; and (4) spindle cell sarcomas with an hemangiopericytic or myopericytoma-like pattern.⁶⁶  Along these lines, a provisional category encompassing lipofibromatosis-like neural tumors and tumor that closely resemble peripheral nerve sheath tumors has been included in the recently released World Health Organization³⁶  classification of soft tissue and bone tumors as NTRK-rearranged spindle cell neoplasm (emerging). Accordingly, it may be wise to screen for NTRK fusions in all adult sarcomas without obvious differentiation or driver gene mutations/fusions/amplifications, particularly those with S-100 and CD34 coreactivity and/or located in the gynecologic tract.^{24,60–62,67,68} 

Because of the perception that NTRK fusion–positive tumors are frequent in the salivary glands and the relative abundance of relentless adenoid cystic carcinomas (Figure 11, A) in most institutions (<10% of all salivary gland neoplasms),⁴⁴  it is plausible that pathologists will be asked to perform and score pan-TRK IHC in these tumors.¹⁴  At least 40% of adenoid cystic carcinomas can show moderate to strong cytoplasmic staining in the outer layer of tumor cells (Figure 11, B), which never corresponds to an NTRK fusion.^14,48  Interestingly, adenoid cystic carcinomas can also show overexpression of KIT and EGFR, which are rarely mutated or amplified, further preventing druggable opportunities.⁴⁴ 

Through this paper we have emphasized the key role of pan-TRK IHC in the identification of NTRK fusions. The application of these histology-based and genomic-based triaging strategies is illustrated with 9 examples from our daily practice and is summarized in the Table. Although NTRK inhibitors are being approved in a tumor-agnostic manner, pathologists should take the clinical, histologic, and molecular context into consideration when scoring this marker.

This work is dedicated to all the COVID-19 victims. Dr Lopez-Rios thanks T. Crean for his constant support.