Current Approach to Undifferentiated Neoplasms, With Focus on New Developments and Novel Immunohistochemical Stains

Given that a differentiated tumor usually shows characteristic features resembling a primary or normal tissue type, in most cases it is relatively easy to identify the tumor when it presents in a metastatic site and may be most challenging in the primary site given its resemblance to the normal tissue. A great illustrative example is that of differentiated thyroid tumors without nuclear atypia, such as follicular thyroid carcinomas, that may be challenging to distinguish from a hyperplastic colloid nodule, but relatively easy to diagnose when metastatic to the bone given that it is clearly not native tissue you would expect in the bone regardless of the atypia.⁵ 

Undifferentiated neoplasms in this review will include tumors that are morphologically ambiguous or poorly differentiated without a clear primary site or subtype based on initial cytologic or histologic review. After a review of imaging and appropriate immunostains, some of these tumors will be classifiable with respect to the site of origin and/or subclassification (eg, neuroendocrine, squamous) and therefore not truly be an undifferentiated tumor on final diagnosis. Furthermore, some of these ambiguous tumors may represent a dedifferentiated tumor or high-grade transformation from a prior differentiated tumor, not truly undifferentiated; but without a good clinical history or in the setting of a limited tissue sample without the transition to the differentiated tumor present, these may be morphologically undifferentiated and not definitively classifiable (Figure 1, A through D). This review will focus on the thought process and tools available from initial review of a morphologically undifferentiated tumor in order to try to reach a classification or site of origin.

Undifferentiated neoplasms can show a variety of histomorphologic and cytomorphologic features that often necessitate immunostains to evaluate for cell of origin and potential primary site of origin, in addition to looking for biomarkers that are targetable for therapy. These tumors can sometimes first manifest in a small biopsy that cannot definitively subclassify the tumor or determine the origin due to ambiguous immunohistochemical findings, lack of obvious transition to a well-differentiated tumor that it arose from, or lack of sufficient tissue overall for additional testing, leading to indeterminate morphology-based or descriptive diagnoses. In addition, these undifferentiated tumors may require additional tissue for further characterization and molecular or other biomarker studies, which may necessitate repeat procedures or larger tissue samples. These can be challenging both on first diagnosis and on repeat sampling, where there may already be a significant delay in treatment, leading to anxious patients and families and frustrated clinicians.^3,4 

When approaching these small biopsies, rapid on-site evaluation can be helpful to ensure there is lesional tissue present and to triage the sample appropriately, especially in the repeat setting given that prolonging the time to diagnosis can lead to treatment delays and patient dissatisfaction. For example, in the setting of a new diagnosis of a undifferentiated tumor or in a repeat biopsy scenario where a tumor was previously challenging to characterize, obtaining material for at least 2 cell blocks or using different needle approaches (eg, fine-needle aspiration and concurrent core needle biopsy) can be helpful to optimize the ability to have at least 2 formalin-fixed, paraffin-embedded (FFPE) blocks for ancillary studies and potential molecular studies.⁶  Thus, if multiple immunostain panels are required for characterization, there will be another FFPE block preserved for any additional testing. Furthermore, having 2 FFPE blocks allows the pathologist to select the optimal block for different studies, such as preserving the block with the greatest tumor cellularity and least necrosis for molecular studies, and allocating the block with less tumor cellularity, more background stroma, and/or more necrosis for immunostains. Some providers will also communicate the need to procure material for certain studies on the requisition when ordering a repeat biopsy, which is important to convey to the pathology team to optimize appropriate processing. In addition, some institutions have protocols in place to limit the histologic levels and immunostains performed, and to minimize tissue loss when trimming the block, by cutting blank slides for ancillary studies up front in a stepwise fashion so that the best tissue sections are saved for the most important studies or those having more stringent tumor cellularity requirements.³ 

The importance of a good clinical history and thorough radiologic imaging cannot be underestimated in cases of undifferentiated neoplasms (Figure 2). Knowledge of any prior malignancies or surgeries, medical conditions, smoking status, age, biological sex, serologic markers, and other clinical findings are key factors in order to narrow the differential diagnosis. In particular, if there is a history of prior malignancies, the treatment history can be important since this may alter the morphologic, immunophenotypic, or molecular features of a tumor. These key elements can be critical to minimize a large exhaustive panel of immunostains, and to focus on the stains that are statistically most likely based on the clinicoradiologic features and morphology. With tissue samples getting smaller and the need for ancillary studies getting larger, this need for clinicoradiologic correlation is even more important.⁷  This can be problematic in patients who see physicians in different health care systems without integrated electronic health care records and may require reaching out to other healthcare institutions to acquire important radiology or pathology reports for correlation.

Finally, there is an increasing need to close the loop on morphologically ambiguous tumors that are sent for molecular or other biomarker studies in case these studies help in further characterization of the tumor. An example would be an undifferentiated tumor arising in a metastatic setting without any clear site of origin. If molecular testing has been performed, the results should be compared to any prior tumors the patient had with molecular studies to see if there are similar aberrations that allow us to determine if the tumors are clonally related.⁵  An illustrative example is a poorly differentiated adenocarcinoma with cytokeratin 7 (CK7) positivity but lacking site-specific markers that undergoes molecular testing and shows a KRAS mutation. Given that KRAS mutations are more common in pancreatic tumors, particularly KRAS G12R mutations, this could be used to favor a pancreatic origin in the correct clinical setting, as opposed to KRAS G12C mutations that are more commonly seen in lung adenocarcinomas.⁸  This ability to use molecular testing to favor or definitively determine a site of origin is becoming increasingly common as we improve our understanding of the molecular signature of different tumors. Thus, as more patients are beginning to have molecular characterization of tumors, and may undergo treatments that increase survival, these patients can develop new tumors that are morphologically ambiguous but have a similar molecular signature. Given the complexity of these correlations, some institutions now have diagnostic management teams including molecular and anatomic pathologists, in addition to oncologists and other healthcare professionals, who discuss follow-up results on cases after initial sign-out at tumor boards or other multidisciplinary conferences in order to reach a consensus opinion and reduce medical errors, given the endorsement by the Institute of Medicine in 2015 as a high-quality practice that reduces medical errors.⁹  These discussions can be very important in order to decide on further treatment or management options, and are increasingly valuable in modern personalized medicine.¹⁰ 

Undifferentiated neoplasms can show a variety of morphologic patterns but can broadly be divided dichotomously into epithelioid and spindle cell proliferations. This can be helpful in determining the initial immunostain panels to order for classification with regard to cell of origin and primary site. In addition, if there is a nonspecific immunophenotype with no clear tissue of origin, the broad morphologic descriptive diagnosis of “malignant spindle or epithelioid neoplasm” may be utilized with a comment regarding the diagnostic possibilities (Figure 2).

In tumors with a predominantly round cell or epithelioid pattern, the main differential diagnosis includes carcinomas, epithelioid melanoma, epithelioid mesothelioma, lymphoproliferative disorders, and epithelioid sarcomas. Epithelioid tumors usually appear either more cohesive with carcinomas and mesotheliomas, or more discohesive with melanomas, lymphomas, germ cell tumors, and sarcomas, which can help with the initial immunopanel selected. In many cases it may be helpful to remember that a malignancy is far more likely to be a carcinoma than a lymphoma, sarcoma, mesothelioma, or melanoma, and may simply require an extended panel of cytokeratins to determine positivity, along with proving that other key markers are negative.¹¹  The initial approach that the pathologist takes in this balancing act of a diagnostic workup is a critical step. At times, even with the best clinical history and radiologic correlation, he or she is bewildered by the seemingly unclassifiable cytologic and histologic features of the tumor; it may be prudent in these instances to take the route of caution, starting with a small panel of broad screening markers and using several unstained levels for a second, “differentiation-specific” panel, especially in small biopsy scenarios.^12,13 

Traditional site-specific markers remain helpful for characterizing carcinomas with poorly differentiated or undifferentiated patterns, but some of these tumors can lack staining of these commonly used markers. In the case of anaplastic thyroid carcinoma, loss of thyroid markers (such as thyroglobulin and thyroid transcription factor 1 [TTF1]) is common, although traditionally it had been quoted that paired box gene 8 (PAX8) expression was retained in more than 80% of these tumors. Recent studies have shown that PAX8 expression is retained in only half of all anaplastic thyroid carcinomas.¹⁴  In scenarios that are PAX8-negative, the recently discovered marker insulin-like growth factor 2 (IGF2) mRNA binding protein 1 shows significant promise, as it is expressed in more than 75% of anaplastic thyroid carcinomas.¹⁵  For breast carcinomas that are negative for estrogen and progesterone receptors and are human epidermal growth factor receptor 2 (HER2) negative (ie, “triple negative”), or for metaplastic breast carcinoma, confirming the site of origin can be challenging, particularly in the metastatic setting. GATA3 is a generally useful marker of breast origin, but in the “triple-negative” or metaplastic setting, the sensitivity of GATA3 can range from 54% to less than 20%.^16,17  A new marker showing great preliminary success is trichorhinophalangeal syndrome type 1 (TRPS1), which has been shown to be positive in more than 85% of breast carcinomas, including metaplastic types, and shows a lack of expression in the majority of other tumors, including other GATA3-positive tumors such as urothelial carcinomas.^17,18 

Renal tumors can show significant morphologic and immunophenotypic overlap, leading to potential diagnostic confusion. This can be particularly worrisome in cases where the differential diagnosis includes benign and malignant entities, such as renal oncocytoma and chromophobe renal cell or other oncocytic renal cell carcinomas. Immunohistochemistry for c-kit and CK7 can often narrow the differential diagnosis, but the varying degree of CK7 expression in chromophobe renal cell carcinoma and the varying degree of CK7 negativity in oncocytoma still leave diagnostic ambiguity between benign and malignant. Thus, the new immunostain for homeobox protein NKX6-1 (NKX6-1) shows moderate sensitivity with high specificity for chromophobe renal cell carcinoma and can greatly help solidify this diagnosis in these situations.¹⁹ 

One area of traditional difficulty concerns the diagnosis of primary intrahepatic cholangiocarcinoma versus metastatic carcinoma to the liver. Often the clinical and radiologic presentation could be compatible with either, as both metastases and intrahepatic cholangiocarcinoma can present with multiple discrete lesions. Pathologists traditionally resorted to cytokeratins (CK7, CK19, and CK20), but unfortunately both intrahepatic cholangiocarcinoma and many other metastatic carcinomas share a CK7⁺/CK20⁻ immunophenotype; the result was a diagnosis of exclusion, involving extensive additional imaging and clinical procedures by the clinical team to rule out another source. Furthermore, poorly differentiated hepatocellular carcinomas can also lose expression of HepPar1 and α-fetoprotein (AFP), which limits the ability to make a definitive diagnosis. Albumin RNA in-situ hybridization (ISH) is a new diagnostic modality to confirm hepatic origin with high sensitivity and specificity.²⁰  Subsequent studies have confirmed the high sensitivity of albumin RNA ISH for hepatocellular carcinomas and intrahepatic cholangiocarcinomas, but have also noted that a subset of lung, gallbladder, pancreatic hepatoid and acinar cell carcinomas, and breast carcinomas can also show positivity.²¹  Nevertheless, as part of a wider panel, albumin RNA ISH is a valuable tool to save patients additional procedures and delays in treatment for intrahepatic cholangiocarcinomas (Figure 3, A through D). Furthermore, it can be helpful when patients have other CK7⁺ adenocarcinomas, such as lung carcinoma, to help exclude a metastasis that would be negative for albumin ISH and lead to more accurate staging. An upper gastrointestinal primary may also be considered in the CK7⁺ adenocarcinomas present in the liver, and in these scenarios, homeobox protein CDX-2 (CDX2) and cadherin-17 (CDH17) have been shown to be helpful.²² 

Sometimes even the most reliable markers of origin can show loss of staining in some poorly or undifferentiated variants. In the case of prostatic carcinoma, the modern nuclear marker homeobox protein NKX3.1 has a greater than 90% sensitivity reported and is generally accepted to be the most reliable indicator of prostatic origin. But cases with divergent differentiation (such as small cell morphology) can bring this sensitivity down to 0%, and the diagnosis of metastatic small cell or neuroendocrine carcinoma of the prostate is often made via exhaustive exclusion and clinical correlation. Puzzlingly, many cases will show conventional differentiated morphology with NKX3.1 expression within the primary tumor, even with extensive sampling, without an associated small cell component, and then demonstrate the high-grade neuroendocrine component in metastatic disease with a different morphology and immunohistochemical profile.²³  Immunostains to confirm neuroendocrine differentiation in these scenarios are critical for diagnosis and may impact the treatment regimen to include platinum-based chemotherapy.

Tumors with a small round blue cell or basaloid morphology could be considered a subset of epithelioid tumors, and they encompass both epithelial and mesenchymal entities. Ewing sarcoma is perhaps the quintessential round blue cell tumor, characterized by rearrangement of the EWRS1 gene, usually with the FLI1 partner gene; traditional immunohistochemical markers such as cluster of differentiation 99 (CD99) are sensitive but nonspecific, and thus, primary diagnosis has been traditionally made via fluorescence in situ hybridization (FISH). Additional research into downstream products of this gene fusion has led to the discovery of modern markers with improved specificity such as homeobox protein NKX2.2 (84%) and protein kinase c beta (PRKCB) (96%).^24–27  Additional considerations among the rare Ewing-like tumors include the CIC sarcomas and the BCOR sarcomas; these unfortunately show significant morphologic and immunophenotypic overlap, and definitive diagnosis requires gene rearrangement studies.²⁸  NUT (nuclear protein in testis) midline carcinoma, characterized by rearrangement of the NUT gene, has a varied morphology (squamoid to basaloid) and is a rare entity, but highly sensitive and specific immunohistochemistry for NUT can rule it in or out.²⁹  SWI/SNF-related matrix-associated actin-dependent regulator of chromatin subfamily B member 1 [SMARCB1]-deficient sinonasal carcinomas are generally basaloid tumors with the appearance of a poorly differentiated carcinoma; the diagnosis is secured by immunohistochemistry showing the specific loss of expression of SMARCB1.³⁰  Similar morphologies are again observed in SMARCA4-deficient tumors, with the subset of SMARCA4-deficient non–small cell lung carcinomas expressing cytokeratins and with SMARCA4-deficient undifferentiated tumors lacking significant cytokeratin expression; all should show loss of SMARCA4 with variable loss of SMARCA2 by immunohistochemistry.^31,32  Furthermore, thymic carcinomas can be uncommon undifferentiated tumors with ambiguous morphology that are frequently positive for CD5, CD117, and squamous markers, and may show epithelioid or spindled tumor cells with associated lymphoid cells (eg, lymphoepithelial type appearance). Finally, poorly differentiated subsets of tumors, such as poorly differentiated colonic carcinomas with medullary features and frequent microsatellite instability, can show loss of expression of CK20 and CDX2, while being positive for newer markers such as special AT-rich sequence binding protein 2 (SATB2) and CDH17.²² 

For spindle cell proliferations, the main differential diagnosis includes mesenchymal tumors (including sarcomas), spindle cell melanoma, sarcomatoid mesothelioma, and sarcomatoid carcinoma. Typically, cytokeratins are critical for confirming a carcinoma or mesothelioma; however, cytokeratin expression has been reported in melanomas (1%–40%) and malignant vascular tumors (such as angiosarcoma [31%] and epithelioid hemangioendothelioma [67%]), which is why a panel of stains is always encouraged.^33–35  For spindle cell proliferations that are cyokeratin positive, especially those within the abdominal or chest wall, the possibility of malignant sarcomatoid mesothelioma should always be considered before rendering a diagnosis of carcinoma. In this scenario, traditional mesothelial markers (calretinin, Wilms tumor protein, CK5/6) in conjunction with more recent markers of malignancy (p16 FISH; BRCA1-associated protein 1 [BAP1], methylthioadenosine phosphorylase [MTAP], nuclear 5-hydroxymethylcytosine [5-hmC], and ubiquitin-like with plant homeodomain and ring finger domains-1 [UHRF1] immunostaining) are helpful in confirming the diagnosis. Malignancy markers are crucial to exclude reactive spindle cell proliferations that are common and notoriously overlap with mesothelioma. BAP1 has been helpful as an immunostain, particularly in laboratories that may not have p16 FISH studies available, and loss of expression is seen in the majority of mesothelioma cases in fluid cytology and small biopsy samples.^36,37  The loss of BAP1 is also more commonly seen in metastatic mesothelioma than in metastatic carcinoma and may help favor mesothelial origin in a metastatic malignancy of unknown primary.³⁸  Given the medicolegal implications of a mesothelioma diagnosis, a confirmatory panel of stains is prudent even when the disease is not morphologically undifferentiated. Additional stains that are helpful as malignancy markers for mesothelioma include MTAP, 5-hmC, and UHRF1.³⁹  Although some of these stains may have inferior performance in sarcomatoid mesotheliomas, UHRF1 was recently shown to have better performance as a positive marker of sarcomatoid mesothelioma than of epithelioid variants or benign proliferations.⁴⁰  Some of these markers may even include potential therapeutic options in mesothelioma patients, such as with the use of EZH2 inhibitors in BAP-mutated mesotheliomas or inhibitors of UHRF1.⁴¹ 

Additionally, some sarcomas with characteristic chromosomal rearrangements traditionally diagnosed via FISH now have immunostains available as surrogate markers for characterization, given that the fusion oncoproteins tend to be overexpressed by translocation-positive tumor cells and can potentially be targeted with antibodies. Ewing sarcoma, as previously discussed, can be targeted with downstream products of EWRS1 gene rearrangements, such as NKX2.2 and protein kinase C beta type (PRKCB).^25–27  Myxoid liposarcoma is characterized by a fusion of DDIT3, usually with a FUS fusion partner, and immunohistochemistry for the DDIT3 oncoprotein is highly sensitive and reasonably specific for myxoid liposarcoma.^42,43  Other similar examples include immunohistochemistry for CAMTA1 and TFE3 fusion products in epithelioid hemangioendotheliomas,^43,44  synovial sarcoma X chromomsome (SSX) C-terminus and synovial sarcoma translocation (SYT)-SSX fusion-specific staining for synovial sarcoma,⁴⁵  and immunohistochemistry for STAT6 fusion products in solitary fibrous tumors.⁴⁶ 

Neuroendocrine tumors (NETs) can be a diagnostic difficulty on both the well-differentiated and poorly differentiated sides of the spectrum. In some cases, it is challenging to even recognize a tumor as neuroendocrine, given that vague gland formation in adenocarcinomas may resemble a rosette formation in NETs and some basaloid tumors. In addition, some poorly differentiated tumors may show only focal neuroendocrine staining, and be best classified as carcinomas with neuroendocrine features, as opposed to true neuroendocrine carcinomas. This is seen in lung non-small cell carcinomas, where a subset of lung adenocarcinomas or large cell carcinomas can show neuroendocrine features or differentiation.⁴⁷ 

Challenging subsets of these tumors occur not infrequently in the lung. In poorly differentiated NETs with a markedly elevated proliferation index, staining with TTF1 and with traditional neuroendocrine markers (such as synaptophysin) can be variable. Recent RNA expression studies have shown at least 4 subtypes of small cell lung carcinomas defined by expression of achaete-scute homologue 1 (ASCL1), neurogenic differentiation factor 1 (NEUROD1), POU class 2 homeobox 3 (POU2F3), and yes-associated protein 1 (YAP1).⁴⁸  The majority of small cell lung carcinomas have been shown to express ASCL1 and/or NEUROD1, and in these tumors, there is high expression of neuroendocrine markers and TTF1. However, approximately 10% of cases show a lack of expression for both ASCL1 and NEUROD1, which is frequently associated with low expression or loss of conventional neuroendocrine markers and TTF1, and positivity for the new immunohistochemical stain for POU2F3 (Figure 4, A through D). These are the frustrating tumors that every pathologist has likely encountered, where there is small cell morphology without immunohistochemical confirmation. Despite the diagnostic challenge, new nuclear markers of neuroendocrine differentiation (such as insulinoma-associated protein 1 [INSM1]), or markers for other small cell lung carcinoma subtypes (POU2F3) can be advantageous for confirming the diagnosis.⁴⁸ 

In the pancreas, well-differentiated neuroendocrine neoplasms can overlap in morphology with normal islet cells, normal acinar cells, and acinar cell carcinomas given the moderate amounts of cytoplasm with granularity, round nuclei, and relative cellular monotony. A new marker to help in this distinction from normal pancreatic islet cells is inhibin-alpha, which appears to be predominantly negative in NETs of the pancreas, but positive in normal islet cell hyperplasia, serous cystadenomas, and some acinar cell carcinomas.⁴⁹  Another new marker to aid in the diagnosis of acinar cell carcinoma of the pancreas is carboxypeptidase 1, which has been shown to be a highly sensitive and specific marker, especially in the metastatic setting given that normal acinar cells are also positive.⁵⁰  Lymphoid enhancer-binding factor 1 (LEF1) is another relatively new marker that can help in the diagnosis of pancreatic solid-pseudopapillary neoplasms, which can show neuroendocrine morphologic features and stain with neuroendocrine markers like CD56, to help distinguish them from NETs or acinar cell carcinomas.⁵¹ 

In well-differentiated NETs and neuroendocrine carcinomas of the pancreas, a metastatic tumor from the lung or other source should be excluded. Mutations in DAXX, ATRX, and ALT have been associated with a poor prognosis, but more recently it has been shown that loss of death-associated protein 6 (DAXX) and ATP-dependent helicase ATRX (ATRX), with expression of ALT is moderately sensitive but highly specific for neuroendocrine carcinomas of pancreatic origin.⁵²  Conversely, orthopedia homeobox (OTP) is a highly sensitive and specific marker for pulmonary origin in neuroendocrine tumors (ie, carcinoids), often helpful in cases where these tumors have lost TTF1 expression.⁵³ 

Basaloid neoplasms or small round cell tumors that may have a matrix component and lack obvious high-grade features can also be challenging to diagnose. The differential diagnosis in these tumors includes a variety of tumors, such as NETs and mesenchymal tumors, in addition to epithelial-myoepithelial carcinomas (EMCs). The cases that are most challenging are those without a matrix component that appear to have exclusively basaloid or round blue cell morphology, and can overlap with neuroendocrine carcinomas, Ewing sarcoma, EMCs, and other round cell tumors. Once there is recognition of a matrix component to the tumor, then salivary gland–type tumors and EMCs become more probable and require exclusion. Recently, the RAS Q61R mutant-specific immunostain was found to show diffuse cytoplasmic/membranous expression in the myoepithelial cells in 65% of EMC cases, specifically those with the HRAS Q61R mutation.⁵⁴  This illustrates how specific molecular characterization of tumors can lead to new immunostains that are easier and faster to use in laboratories, which is an increasingly frequent phenomenon allowing us to reach more definitive diagnoses in undifferentiated tumors.

Historically, the classification of salivary gland neoplasms was largely based on morphologic patterns, type of stroma or background material, and other features such as lymphoid infiltrate, and even with all this information, some tumors were given a descriptive diagnosis with a list of diagnostic possibilities, especially in cytology specimens. However, with new discoveries of characteristic rearrangements in a variety of salivary gland tumors, definitive classification on cytology fine-needle aspirations and small biopsies is possible and more accurate. Furthermore, the development of new immunostains and ISH in salivary gland tumors that correspond to the rearrangements makes characterization easier, more widely available, and faster than with FISH or molecular studies. In particular, the classification of basaloid tumors of the salivary gland is enhanced with new markers, in addition to traditional markers. For instance, adenoid cystic carcinomas are characterized by MYB rearrangements, and there are now MYB immunostains and chromogenic ISH available for use.⁵⁵  More recently, novel immunostains, such as Notch homolog 1, translocation-associated (NOTCH1), the alleged MYB target, and brain-derived neurotropic factor (BDNF) have also been shown to be helpful. Interestingly, BDNF is involved with neurogenesis and shows cytoplasmic staining in almost all adenoid cystic carcinomas and may be involved with this tumor’s notorious proclivity for perineural invasion.⁵⁶  NOTCH1 staining pattern is cytoplasmic, membranous or nuclear. It is thought that the characteristic MYB-NFIB translocation may lead to Notch pathway activation given that all MYB-induced adenoid cystic carcinomas appear to show Notch activation (as in Notch pathway). In addition to helping with classification and correlating with the solid histology of adenoid cystic carcinoma, NOTCH1 mutations have been shown to have prognostic significance, given the correlation with advanced disease stage, higher incidence of liver and bone metastasis, and shorter relapse-free and overall survival.^57–59  LEF1 is a nuclear stain that is positive in basaloid salivary gland neoplasms like basal cell adenomas, in addition to the cribriform morular variant of papillary thyroid carcinoma, solid pseudopapillary tumor of the pancreas, pancreatoblastoma, and chronic lymphocytic leukemia/small lymphocytic lymphoma, but its positivity essentially excludes the diagnosis of adenoid cystic carcinoma.^51,60 

Diagnosis of other salivary gland neoplasms has also seen promising results in immunohistochemical surrogates for molecular diagnostics. In acinic cell carcinoma, pathologists have traditionally relied upon DOG1 and S100 for diagnosis, but since normal salivary gland acini also express these markers, this panel is unable to separate generous sampling of salivary gland acini versus acinic cell carcinoma. FISH studies for NR4A3 fusions have resolved this dilemma in the past, but new immunohistochemistry for NR4A3 (NR4A3/NOR-1) fusion products is faster and outperforms FISH.⁶¹  Additional stains such as carbonic anhydrase VI (CA6), Mist1, and p-STAT5 are emerging markers that can be used in a panel for acinic cell carcinomas if available.^62–64 

Pleomorphic adenoma/mixed tumor has been characterized by rearrangements of PLAG1, resulting in overexpression that can be targeted with excellent success using the PLAG1 immunostain; however, caution is still recommended as PLAG1 overexpression is also seen in pleomorphic low-grade adenocarcinoma.⁶⁵ 

Several specific tumors are characterized by NTRK fusions, including (mammary analogue) secretory carcinoma of the salivary gland, secretory carcinoma of the breast, and infantile fibrosarcoma. Initial trials of immunohistochemistry for Pan-TRK found the marker to be highly sensitive and specific for NTRK fusions and helpful diagnostically for these entities.⁶⁶  However, as TRK-inhibitors have become more prevalent, use of the Pan-TRK immunostain has widened to many different organs to predict response to TRK-inhibitor therapies. Notably this has called the sensitivity of the Pan-TRK stain into question, as staining can be observed in adenocarcinoma of the lung and colon, papillary thyroid carcinoma, olfactory neuroblastoma, uterine leiomyosarcoma, and adenoid cystic carcinoma.⁶⁷ 

While a pathologist’s goal is to make the most accurate diagnosis possible, the penultimate goal of accurate diagnosis is to give the clinical team the information and tools they need to care for the patient. Despite the frustrations encountered by both the pathologist and the oncologist with an undifferentiated cancer or cancer of unknown primary, the oncologist does not necessarily need to know the site of origin or type of differentiation to be able to provide an effective treatment plan. In these cases, the value of molecular profiling by next-generation sequencing or immunohistochemistry to uncover exploitable targets (EGFR, NTRK, ERBB2/HER2, programmed death ligand-1, or mismatch repair proteins) cannot be understated, as identification of any of these has been shown to offer significant survival benefit to the patient.⁶⁸ 

The pathologist is the steward of the patient’s tumor tissue, and in these difficult cases he or she must know when “enough is enough.” There is no hard and fast rule for when it is appropriate to exhaust tissue in pursuit of an elusive diagnosis or when it is better to give a descriptive diagnosis after a cursory panel of immunostains and surrender what remains to the cause of finding targetable mutations. However, as every clinical situation is unique, clear communication between the pathologist and the clinical team, regarding what can be done with the tissue that remains and/or how easily or safely additional tissue can be procured, is the approach most congruent with the goals of maximizing patient benefit and minimizing patient harm.

The characterization of undifferentiated and poorly differentiated neoplasms is challenging regardless of the amount of tissue received, but particularly problematic in the primary diagnostic setting and in the situation of limited tissue. These cases typically require a robust attempt at characterization via a panel of immunohistochemical stains, which may exhaust tissue from limited small biopsies and could potentially lead to repeat biopsies and delays in patient care. Thus, using the available clinicoradiologic findings and history to guide the approach to these tumors is important to ensure that the pathologist has enough tissue for the most likely diagnostic possibilities.

This review highlights some of the progress being made in new site-specific markers in tumors that have notoriously been challenging to characterize (eg, triple-negative breast carcinomas and neuroendocrine carcinomas), and new malignancy markers (eg, BAP1, MTAP, and UHRF1 in mesothelioma) that aid in separating reactive mimics of malignant tumors. These markers have helped us to be more definitive in challenging entities. Furthermore, we are now seeing a new class of immunostains emerging from RNA expression, FISH, and molecular studies, whereby immunomarkers of expression correlate with the molecular or genetic findings. This is demonstrated by the DDIT3 immunostain for myxoid liposarcoma, POU2F3 in a subset of small cell lung carcinomas, and some mutation-defined small round blue cell tumors (such as SMARCA1-deficient sinonasal carcinomas, SMARCA4-deficient tumors, and NUT midline carcinomas).

In the future, as more characteristic molecular and genetic abnormalities are described in specific entities, trying to determine if there is a correlating immunostain to adopt will be helpful to potentially classify some tumors that are currently perceived as undifferentiated morphologically or lacking expression of available markers for characterization. This will allow for less expensive and more widespread adoption of these markers in practice, and lead to faster definitive diagnoses in these dismal settings where patients and providers want more rapid diagnoses, than the current practice of utilizing in-house or referral laboratory testing for FISH or molecular studies.⁶⁹ 

Furthermore, the issue of undifferentiated neoplasms is problematic in today’s health care environment, where there seems to be an emphasis on targetable biomarkers for treatment, without necessarily the need to determine the exact subtype or source of the tumor. As pathologists, we are trained to use the tools needed to determine where the tumor is coming from and to provide precise subtyping, in order to correlate with the clinical findings and to provide the correct staging and tumor synoptic reports, but this may conflict with the need to have enough material for biomarker studies that may be helpful in conveying prognostic or therapeutic information. Thus, when characterizing these undifferentiated neoplasms on small biopsies (Figure 2), there appears to be a shift toward excluding the most likely diagnoses and resorting to a descriptive diagnosis if clinicians want to perform theranostic biomarker studies to provide systemic treatment options in patients with widely metastatic disease who may not be surgical candidates. Given the complexity of the current next-generation sequencing panels and the subsequent results, there is an increasingly important need to have diagnostic management teams to discuss the final findings and render the best ultimate management decisions for optimal patient care, and for pathologists to play a key role in these discussions.