Fine-needle aspiration (FNA) offers a suitable alternative to biopsy in a variety of clinical settings, and there are many studies demonstrating the possibility of using it not only for diagnosis but also to study response to therapy. Immunocytochemistry (ICC) and, more recently, molecular techniques are ancillary tools that contribute to diagnosis, prognosis, and prediction of tumor behavior. Therapies are now being directed toward individual molecular targets; therefore, the use of ancillary techniques in cytology is a challenge, demanding an increase in standardization of preanalytic and analytical methods. In cytology, the amount of material is a limiting factor for the use of ancillary techniques. It is fundamental to have precise clinical information and a good morphologic workup for interpretation and strategic use of the sampling to select the correct test. Recently, for the first time, a recommendation for good practice on lung cancer states that tissue specimens should be managed not only for diagnosis but also to maximize the amount of tissue available for molecular studies.1 ICC is a relatively easy technique to establish; hence, most laboratories have implemented it in their routine. Unfortunately 1 of the most common mistakes when using ICC is to jump from a histologic adapted technique directly to cytologic preparation. Several variables can affect ICC results that are performed in cytologic preparations, and they should carefully be taken in account to obtain accurate and reliable result. Special attention should be given to the type of material collected, the type of fixative used, the antibodies selected, and the appropriate controls. Internal positive controls usually are difficult to obtain in cytology samples. ICC and molecular techniques can be performed with good results directly in cytologic smears, in cytospins, or in liquid-based thin-layer preparations. In their review Fowler and Lachar2 report that each of these types of material has advantages and disadvantages. FNA material embedded in formalin-fixed cell blocks has demonstrated excellent results when applied in ICC studies. A correlation study performed by Fetsch et al3 revealed a great advantage of the cell block technique for immunostaining compared with cytospins and smears. If cell block is not feasible, then material should saved for cytospins or monolayer preparations. The same approach can be used for molecular techniques that, although they demand even more sophisticated procedures, can be applied to different types of cytologic material as indicated in the literature.3-5 In our experience, better results concerning DNA and RNA extraction are obtained with liquid-based preparations.7, 8
Chromosomal translocations are 1 of the most common genetic alterations studied in pathology. Several methods exist to evaluate them, including conventional cytogenetics, reverse transcription-polymerase chain reaction (RT-PCR), and fluorescence in situ hybridization (FISH). FISH offers several advantages over conventional cytogenetics. Its main advantage is that nondividing (interphase) nuclei can be evaluated, making it unnecessary to evaluate the neoplastic cells in culture. This allows a retrospective analysis of alcohol-fixed smears or formalin-fixed, paraffin-embedded cell blocks. FISH is especially advantageous in small samples, in which it may not be possible to submit tissue for cytogenetics. FISH also offers several benefits over RT-PCR. For a given translocation with multiple breakpoints, multiple primers are necessary to evaluate all possible fusion transcripts; whereas, with FISH this can be achieved by using break-apart probes (with split signals reflecting the presence of gene fusion) or dual-fusion translocation probes (with fusion of signals indicating the presence of gene fusion). Therefore, 1 set of FISH probes can identify essentially all known breakpoints of a given translocation, resulting in increased sensitivity. In addition to the advantages of evaluating multiple breakpoints, FISH also is helpful in situations in which 1 translocation partner is constant but the second translocation partner changes. This event occurs in Ewing sarcoma/primitive neuroectodermal tumor (ES/PNET), in which the ES receptor 1 gene (EWSR1) translocates with different partners, and in myxoid/round cell liposarcoma, in which the DNA-damage-inducible transcript 3 (DDIT3) gene in chromosome 12 translocates with chromosome 16, and the chromosome 22 translocation t(12,16), and the t(12;22).9 When analyzing FISH results, the examiner must be aware that the same gene often is involved in translocations present in different soft tissue neoplasms. Therefore, the same set of probes also can be used for a gene rearranged in multiple diagnostic entities: EWSR1 is rearranged in ES/PNET, in desmoplastic round cell tumor, in extraskeletal myxoid chondrosarcoma in some myxoid/round cell sarcomas, and in clear cell sarcoma.
Fixative and reagent concentrations should be adapted for cytology. Certain markers, such as gross cystic disease fluid protein 15 (GCDFP-15), S100 protein, or Hep Par 1, are leached from alcohol fixatives and can render false-negative results.2 Reagent concentrations should be customized for cytologic specimens; otherwise, you can have false-positive results because of the excess of antibodies. Another concern is the use of controls. Positive and negative controls must be performed with each test sample. A recent meta-analysis demonstrated that >50% of the articles published in the last 15 years about ICC on cytology do not even mention the use of controls.10 The authors of that report did not identify similar studies on molecular techniques. The ideal control should be a comparably fixed cytologic sample. The precise diagnosis is the first goal of a cytologic examination. In the current review, we address the use of ICC and molecular techniques in the diagnosis and correct classification of selected epithelial and mesenchymal tumors. Special attention is given to lung, thyroid, kidney, pancreatic, and soft tissue tumors, because other organs will be discussed in another article in the same issue of this journal.
Lung cancer is the solid tumor for which diagnosis and therapeutic decisions rely more frequently on morphologic evaluation performed in small biopsies or cytologic material. The identification of epidermal growth factor receptor (EGFR)-positive adenocarcinomas (ADCs) permits the use of tyrosine kinase inhibitors (TKIs); also, the recognition of squamous cell carcinoma (SqCC) avoids the use of bevacizumab, which has been linked to serious bleeding in this subset of lung cancer patients.1, 11
Nonsmall cell lung cancer (adenocarcinoma vs squamous cell carcinoma)
Often, FNA is the only material available for diagnosis in lung cancer; therefore, a panel of well defined markers must to be used to refine a diagnosis of nonsmall cell lung cancer (NSCLC) to ADC or SqCC (Fig. 1). Nicholson et al12 demonstrated that a panel composed of thyroid transforming factor 1 (TTF1), cytokeratin 5/6 (CK5/6), and tumor protein 63 (P63) solve the majority of dubious cases. P63 and CK5/6 detect most SqCCs, and TTF1 detects most ADCs. P63 appears to be more sensitive, whereas CK5/6 appears to be more specific.13 More recently, another antibody, Napsin-A, was proposed as an alternative marker for ADC. Napsin-A has a specificity similar to that of TTF1; however, it displays a higher sensitivity and is almost exclusively positive in ADC.14 This marker also is useful in cytology to identify primary and metastatic lung ADCs among poorly differentiated carcinomas.15
Nonsmall cell lung cancer versus small cell lung cancer
ICC also has an important role in differentiating NSCLC from small cell lung carcinomas (SCLC) and large cell neuroendocrine tumors (LCNEC). In these tumors, a panel of immunostains including either TTF1, cluster of differentiation 56 (CD56 [neural cell adhesion molecule]), and synaptophysin or chromogranin, CK5/6, and P63 will help to differentiate SCLC from a small cell variant of SqCC, respectively.16 In cytology LCNECs are characterized by smears with intermediate-to-large, pleomorphic single or grouped cells in a “dirty” background. These groups have peripheral palisading and frequently form rosette-like structures. However, poorly differentiated ADCs also can share some of these characteristics; therefore, neuroendocrine markers are essential for the differential diagnosis between these 2 entities. In this situation, it is important to be aware that SCLC and LCNEC also can stain with TTF1.16
Primary lung cancer versus metastasis
Lung metastases are frequent; and, in most instances, a previous diagnosis of malignancy already is available. However, pulmonary metastases can be the first sign of disease and may require careful investigation to avoid an inappropriate diagnosis of primary lung carcinoma. There are many published algorithms that can distinguish between primary and metastatic carcinoma in the lung.2, 17 Metastatic lesions from the colon, breast, prostate, thyroid, ovary, and others should be excluded. TTF1 and Napsin-A together are an excellent choice of markers that allow for the differential diagnosis between primary and metastatic lung carcinoma in most situations. Metastases from colon are positive for cytokeratin 20 (CK20) and the homeobox protein CDX2. Other potential useful markers are: estrogen receptor (breast), prostate-specific antigen (prostate), thyroglobulin (thyroid), Wilms tumor 1 (WT-1), and paired box 8 (PAX8) (ovary), among others. Although gross cystic disease fluid protein 15 (GCDFP-15) and mammoglobin are indicated as markers of breast carcinoma, there are limitations to their use in cytologic material. GCDFP-15 does not work in alcohol-fixed material, and RT-PCR is much better than ICC to assess mammoglobin. New opportunities have emerged to use molecular techniques to characterize metastasis of carcinoma of unknown primary (CUP). This group of tumors is a clear diagnostic challenge and usually is left untouched by traditional workups; and, without appropriate therapy, patients with these tumors have a very poor survival rate. The next diagnostic steps often are launched by multivariate analysis, using a cumulative approach to examine at hundreds or thousands of genes to determine which expression pattern best fits the tumor type to a database. The microarray-based tests available on the market use messenger RNA or microRNA. MicroRNAs control the gene expression after transcription. They are remarkably tissue-specific, can identify the cancer tissue origin, are well preserved in fixed cells, and the protocols for extraction are relatively simple.8 In a recent article, investigators demonstrated that microRNAs can be used as biomarkers for tracing the tissue of origin of CUPs with high levels of accuracy.18
The most important molecular discoveries associated with lung cancer therapeutics are related to the use of selective TKIs targeting EGFR mutation and the echinoderm microtubule-associated protein-like 4/anaplastic lymphoma kinase (EML4-ALK) fusion gene. Today, all ADCs must be tested routinely for the presence of exclusive EGFR or v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation; and, if both of these markers are negative, then tests for EML4-ALK translocation should be performed.1 EGFR mutation in lung ADCs is associated mainly with nonsmoking women and is identified more often in better differentiated ADCs with or without a bronchioloalveolar component. These patients benefit from EGFR-TKI therapy, and first-line treatment with EGFR-TKI is currently the standard of care in these patients.
Currently, the molecular detection of EGFR mutation is the most accurate method for selecting these patients.19 A recent review of the literature indicated that different molecular techniques—such as FISH; quantitative PCR; and direct sequencing with fresh cells, scraped cells from archival slides, and cell blocks—have similar or greater accuracy and sensitivity than surgical specimens.20 This is fundamental, because most of lung cancers are diagnosed based on cytologic material. In this review, we provide evidence that such material is suitable for detecting EGFR status using different methodologies and preparations.20
The most common oncogenic mutations in the tyrosine kinase (TK) domain of EGFR are small in-frame deletions in exon 19, and a point mutation (leucine→arginine at codon 858 [L858R]) in exon 21.1, 19, 21 Together, these 2 mutations account for 90% of all EGFR mutations in NSCLC.21 These mutations are likely to cause constitutive activation of the kinase and confer dramatic sensitivity to the TKIs gefitinib and erlotinib. Unfortunately, the effect of these drugs is limited in time because of the emergence of drug resistance. A second mutation, a threonine→methionine substitution at codon 790 (T790M) in exon 20, appears in approximately 50% of all patients with acquired resistance to TKIs. Screening for EGFR mutations is used not only to select patients for treatment but also to detect the resistance mutation. The most common method of mutation detection—involving DNA purification from the whole tumor sample, PCR-based amplification, and sequencing—has several limitations. The most important of these limitations is the need of large-sized samples.22 Cytologic samples, as mentioned above, are used frequently to diagnose NSCLC and often are the only available samples (Fig. 2). Recently, Molina-Vila et al developed a method for detecting EGFR mutations in samples with limited numbers of tumor cells and in cytologic specimens. This method involves 3 steps: 1) direct microdissection of tumor cells into PCR buffer; 2) a first round of PCR for each EGFR exon and determination of EGFR status by length analysis; and 3) a TaqMan Assay using the first PCR product as a template. This method is complemented by further analysis using nested PCR and sequencing. The results have has demonstrated that 4 cells are sufficient to determine EGFR mutation status.23 Archival cytology slides also can be used after scraping the cells from the smear to determine EGFR mutations in lung cancer.24 Whole genomic amplification is a technique that allows high-fidelity in vitro reproduction of quality template DNA, opening the door to further array-based analysis of multiple genes. However, this technology requires a great amount of starting DNA material and is not used routinely.22 Recently, the feasibility of RNA amplification was demonstrated in material obtained from low-volume lung specimens for the detection of mutations in multiple genes that have prognostic value in NSCLC.
In 30% of lung ADCs and 5% of SqCCs, KRAS mutations are identified. These KRAS ADCs are mainly of the mucinous type and have been associated with smokers. These 2 types of molecular alteration, KRAS and EGFR, are mutually exclusive, and patients with the KRAS mutation do not benefit from the above-mentioned therapy.19, 25
Fusion of the ALK gene with EML4 recently was detected in 3% to 5% of NSCLCs.1 Both ALK and EML4 are located in the short arm of chromosome 2; they are separated by 12 Mb and are oriented in opposite 5′-to-3′ directions. Two different variants of the EML4-ALK fusion gene have been characterized, both of which involve exons 20 to 29 of ALK fused to exons 1 to 13 (variant 1) or exons 1 to 20 (variant 2) of EML4. Patients who have tumors with ALK gene fusion generally are men and are resistant to EGFR inhibitors.26 Currently, a drug is in trial that inhibits this oncogene (PF-02341066-crizotinib; a dual met proto-oncogene/anaplastic lymphoma kinase [MET/ALK] inhibitor). These EML4-ALK ADCs have lower levels of ALK protein than other neoplasms and also are associated with ALK mutations (anaplastic T-cell lymphoma); and they are not detected so easily by conventional ICC. The determination of ALK protein expression using either ICC or FISH probes in EGFR–TKI-negative ADCs is fundamental for determining treatment and can be done in cytologic material.19 At this moment, it is important to encourage the proper collection and handling of cytologic samples in new prospective clinical studies so that these novel techniques can be validated in large patient cohorts. It is important to emphasize that discrepancies in KRAS and EGFR mutation status between primary tumors and corresponding metastases can appear. This may have significant implications in determining TKI therapy for patients with NSCLC. Therefore, in validating these findings (eg, the presence of KRAS in metastatic colon cancer), such analyses also are necessary in lung carcinoma metastases.27 However, studies on the clinical impact of discrepancies between primary and metastasis still are needed.
FNA is the most accurate and cost-effective method for evaluating thyroid nodules. A new system of reporting FNA diagnosis in thyroid was created to clarify the meaning and interpretation of cytologic reports and to standardize clinical procedures.28
Follicular-patterned lesions, including papillary carcinoma
The Bethesda Classification System of thyroid lesions correlates each category with a malignancy risk, suggesting appropriate clinical management. In this classification system, subcategories like such as atypia of undetermined significance/follicular lesion of undetermined significance and follicular neoplasm/suspicious of follicular neoplasm encompass several heterogeneous lesions. These follicular patterned lesions include hyperplasic and adenomatoid nodules, follicular adenomas and carcinomas, and the follicular variant of papillary carcinoma (PTC).
Ancillary techniques have been proposed to improve diagnostic accuracy in this group of lesions. ICC markers, such as CK19, galectin-3, and human mesothelial cell 1 (HBME-1), are used most often. The presence of these markers can be detected easily in smears, methanol-fixed thin layer preparations, and cell blocks.29 CK19 is expressed strongly and diffusely in PTC, and several authors have emphasized its utility in discriminating between dubious or suspicious follicular variants of PTC. However, CK19 staining is not specific and stains follicular cells of lymphocytic thyroiditis and of follicular adenomas or carcinomas.30
Galectin-3, a protein involved in cell-to-cell and cell-to-matrix interaction, is considered a marker of malignancy in thyroid follicular lesions. Galectin-3 is positive in most PTCs and anaplastic thyroid carcinomas. Both CK19 and galectin-3 are poor specific markers, and false-positive results can be obtained in lymphocytic thyroiditis, follicular adenomas or carcinomas, and medullary carcinomas.29, 31, 32
Most PTCs have diffuse, positive immunostaining for HBME-1. This positivity is not exclusive to PTCs and, thus, does not specifically indicate PTC.33 In summary, none of these markers is sufficiently specific to be used as a single marker. The use of a panel with CK19, galectin-3 and HBME-1 amplifies sensitivity and specificity.
The finding that some thyroid lesions are associated with specific genetic alterations raises the possibility of improving FNA cytology diagnoses with access to molecular techniques. Papillary carcinomas frequently are diploid lesions and can display nonoverlapping mutations of the genes v-raf murine sarcoma viral oncogene homolog B1 (BRAF), ret proto-oncogene (RET), and RAS in 46% to 75%, 3% to 85%, and 0% to 21% of cases, respectively.34 Although the prevalence of these mutations varies among the published series, their presence is very specific and indicates the existence malignancy. The follicular variant of PTC has a mutational framework that differs slightly from other PTCs. A distinct BRAF mutation is observed (lysine→glutamic acid at codon 601 [K601E]) in 7% of tumors, RAS mutations are more frequent (approximately 25% of tumors), and they can have a PAX8-peroxisome proliferator-activated receptor gamma (PPARγ) mutation (38% of tumors).35 Follicular neoplasia is effectively an aneuploid lesion and has a high prevalence of RAS and PAX8-PPARγ gene mutations (33% vs 45%, respectively, in follicular adenomas vs follicular carcinomas). Characteristic follicular neoplasia mutations are less specific than those described in PTCs and do not necessarily point to malignancy.
Somatic rearrangements of RET are considered specific of PTCs. There are 2 types of mutations in RET/PTC most commonly related to PTCs: RET/PTC1 and RET/PTC3. RET/PTC1 prevails in classic and diffuse sclerosing variants, whereas RET/PTC3 predominates in solid and follicular variants of PTC.
RET rearrangement can be evaluated with immunostaining or PCR techniques in FNA material. Although there is a close correlation between ICC and PCR detection of RET-PCT, molecular testing seems to be more reliable in identifying those rearrangements.36-38 Mutations of BRAF also are highly prevalent and specific to PTC. Their detection appears to be helpful in preoperative dubious FNA diagnosis of PTC. Recently, Nikiforov et al demonstrated that using a panel of markers to detect BRAF, RAS, PAX8/PPAR, and RET/PTC mutations increased the overall accuracy of FNA.39 Their study also indicated that molecular testing applied to undefined cases increased the probability of cancer detection from 40% to 100%. The BRAF valine→glutamic acid mutation at codon 600 (V600E), in particular, is a highly specific marker of PTC; however, the lack of this mutation does not rule out the diagnosis.40, 41 According to the revised American Thyroid Association guidelines for the management of patients with thyroid nodules, the use of molecular markers in thyroid FNA is now considered decisive in clinical decisions for patients who have indeterminate FNA cytology results.41, 42
In thyroid aspirates other differential diagnoses need to be considered. Medullary thyroid carcinomas (MTCs) express calcitonin, carcinoembryonic antigen (CEA), synaptophysin, and chromogranin. Calcitonin is more specific than CEA; however, some MTCs can be only focally positive or even negative for this marker. Calcitonin is lost with dedifferentiation of MTCs, whereas CEA expression is retained. False-positive results have been demonstrated in cytologic material.43 MTC is sporadic in 75% of patients and, in 25%, can be associated with inherited endocrine syndromes. Regardless of the clinical context involved, RET mutations frequently are associated with MTC. MTCs associated with inherited endocrine syndromes have specific RET mutations that help to identify these patients. This knowledge will guide clinicians in the appropriate management of patients.
Kidney tumors are assessed by FNA for initial diagnosis and therapeutic planning. Renal cell carcinoma (RCC) subtypes can exhibit overlapping features; and, in that event, immunostains are crucial in daily routine differential diagnosis.
Eosinophilic subtypes of kidney tumors
Renal cell tumors with granular and eosinophilic features represent a group that encompasses lesions like clear cell RCC (eosinophilic subtype), chromophobe RCC (eosinophilic subtype), PRCC type II, oncocytoma, and angiomyolipoma (eosinophilic variant). These lesions share some morphologic characteristics, which creates problems for their differential diagnosis. These tumors have a specific immunocytochemical profile and also have molecular alterations that can be detected by FISH techniques in FNA cytology samples.44 (Table 1).
|Ancillary Techniques||Cromophobe RCC (Eosinophilic Subtype)||Clear Cell RCC (Eosinophilic Subtype)||Oncocytoma||Papillary (Type II) RCC||Angiomyolipoma (Eosinophilic Subtype)||Medullary RCC|
|Coloidal iron||+ Diffuse||−||−||−||−||−|
|Cytogenetics||Losses in chromosomes 1, 3, 6, 7, 9, 13, 17, 18, 21||Del 3p||Losses in chromosome 1;14q;Y; rearrangements in chromosomes 11p12-13 and 12q12-13||Trisomy 7;17; loss of chromosome Y|
Medullary renal cell carcinoma
Medullary RCC is a highly aggressive tumor that affects young African patients with sickle cell trait. This tumor is sometimes included in the differential diagnosis of eosinophilic-like tumors of the kidney because of the presence of large epithelioid cells with abundant, granular, eosinophilic cytoplasm. These tumors also can bear prominent nuclei and rhabdoid features and can mystify rhabdoid tumors of the kidney, lymphomas, or even germ cell tumors. Together with rhabdoid tumors, they are the only 2 known tumors that share the loss of expression for the tumor suppressor gene INI1 (ring finger-like protein) antibody. However, rhabdoid tumor of the kidney affects a much younger population, and medullary RCC generally has a more heterogeneous population with foci of rhabdoid features. Yolk sac tumors and embryonal carcinoma are easily excluded by the absence of cluster of differentiation 30 (CD30 [tumor necrosis factor receptor]) or alpha-fetoprotein immunoexpression in medullary carcinoma.
Renal cell carcinoma associated with Xp11.2 translocations
Recently, a subset of RCCs in young patients was associated with Xp11.2 genetic alterations and with distinctive ICC features. These carcinomas are characterized by nuclear labeling for transcription factor E3 (TFE3) antibody and have a slightly different immunoprofile from that of clear cell RCC. They do not express CK7, epithelial membrane antigen (EMA), or vimentin; but they do express E-cadherin, and cluster of differentiation 10 (CD10) (neutral endopeptidase). In cytology, these tumors are characterized by smears with numerous sheets and papillary groups of large, polyedric cells; nuclei are round and central with prominent nucleoli and abundant, ill defined, clear or granular cytoplasm (Fig. 3). The presence of psammoma bodies also is frequent and should provide a clue about the diagnosis.45 These carcinomas are characterized by various translocations involving the transcription factor E3 (TFE3) gene in chromosome Xp11.2. At least 6 different Xp11.2 translocation RCCs have been identified until now.46 The most frequent translocations described in these tumors are t(X; 17)(p11.2; q21), and t(X; 1)(p11.2; p34). RCC with t(X; 17)(p11.2; q21) shares a similar TFE3 gene translocation point with alveolar soft-part sarcoma. The gene fusion protein product binds with the MET promoter, inducing activation of the tyrosine kinase receptor MET, making it a therapeutic target in this group of RCCs.
The differential diagnosis between renal tumors with sarcomatoid features is important with regard not to their diagnosis and prognosis but also to therapeutic selection. Differential diagnosis with angiomyolipoma (human melanoma black 45 [HMB45] positive), leiomyosarcoma (muscle markers positive), and solitary fibrous tumors (cluster of differentiation 34 [CD34] [protein tyrosine phosphatase, receptor type C] positive) is achieved well in immunostains.44
Wilms tumor (WT) is by far the most common renal tumor in childhood. This makes the diagnosis of WT the most probable whenever an abdominal mass is detected in the kidney of a child. Under this assumption, many institutions start chemotherapy even without pathologic confirmation. The cytologic diagnosis of WT is performed easily whenever epithelial, stromal, and blastemal components are all present or when at least some are present in smears. When 1 of these components predominates or is the only 1 present, a differential diagnosis with other less frequent tumors of the kidney must be posed. ICC is not very helpful in this differential diagnosis, because WT has its origin in a pluripotential cell (blastema) and can differentiate into various epithelial and mesenchymal tissues. However, some of the tumors that should be considered in the differential diagnosis of WT are characterized by specific genetic alterations, namely, mesoblastic nephroma and rhabdoid tumor.
Mesoblastic nephroma is a benign renal tumor that is diagnosed essentially in the first 3 months of life. Cytologic smears reveal single spindle cells or spindle cells in cohesive sheets with fibroblastic/myofibroblastic, bland nuclei. This pattern can simulate the mesenchymal component of a nephroblastoma. PCR-based molecular techniques can be applied to FNA samples to detect a t(12;15), which is characteristic of mesoblastic nephroma.47
Renal malignant rhabdoid tumor
The cytologic differential diagnosis of renal malignant rhabdoid tumor can be very demanding, because the rhabdoid phenotype is not exclusive to this entity. Other “renal” tumors, such as PNET, clear cell sarcoma of the kidney, mesoblastic nephroma, and WT with rhabdomyosarcoma-like features, can share this phenotype. Malignant rhabdoid tumor is characterized by the absence of expression of INI1, which is generally expressed in normal cells or in cancer cells. This is because of a characteristic monosomy/deletion of chromosome 22 in this tumor.48 All other tumors, with the exception of renal medullary carcinomas, are positive for INI1.
Endoscopic ultrasound-guided FNA biopsy (EUS-FNA) has been used increasingly for the assessment of diverse intra-abdominal tumors. EUS-FNA not only allows for a meticulous representation of both extramural and intramural structures of the gastrointestinal tract but also permits tissue sampling from tumors in these locations. In this section, we discuss the applications of molecular cytopathology to refine the diagnosis of pancreatic tumors and to set a precise treatment in gastrointestinal stromal tumors (GISTs) and colon cancer.
FNA is a well established technique to sample tissues for the diagnosis of patients with unresectable pancreatic cancer and in those patients who may be eligible for therapeutic protocols. For practical purposes, we can divide pancreatic tumors into solid lesions and cystic lesions; and the former can be divided into exocrine tumors and endocrine tumors. In general, the diagnosis of ductal ADC is straightforward; however, in some situations, additional markers can be helpful. Recently, it was demonstrated that the expression of K homology domain-containing protein (KOC), an oncofetal RNA-binding protein that regulates insulin-like growth factor and cell proliferation during embryogenesis, is a new marker of malignancy for pancreatic ADC. KOC expression can be studied easily in cytologic specimens obtained by FNA.49 KOC expression is limited to invasive pancreatic ductal ADC and may serve as an adjunct to cytologic cases in which atypical features preclude a definitive diagnosis of malignancy. Loss of SMAD4 also can supplement the traditional cytologic diagnosis, distinguishing between primary and metastatic ADC. This tumor suppressor gene of the transforming growth factor beta (TGF-β) signaling pathway (SMAD4) is lost in >80% of pancreatic ductal ADCs in contrast to ADCs of the colon, ovary, endometrium, and lung.50
When diagnosing endocrine pancreatic tumors, markers like chromogranin or synaptophysin should be preferred to CD56 and neuron-specific enolase (NSE), which are less specific.2 It is important to emphasize that the main differential diagnosis of such tumors is the solid pseudopapillary neoplasm. This tumor has a peculiar immunocytochemistry profile and expresses multiple antibodies, namely, CD56, vimentin, alpha-1-antitrypsin, CD10, beta-catenin, and progesterone receptor.2 In addition, focal positivity has been observed for NSE and other neuroendocrine markers.
The diagnosis of pancreatic cysts requires a multidisciplinary approach.51 Imaging plays an essential role in identifying cystic morphologic details and conducting the screening between probably benign and probably malignant cysts. Recent advanced techniques are extremely accurate in predicting malignancy and separating mucinous from nonmucinous cysts. The introduction of EUS-FNA with rapid on-site cytologic evaluation gave cytology a great role in the management of pancreatic cysts. Supported by imaging techniques, the role of the cytopathologist is first to discriminate pseudocysts from neoplastic cysts and, then, to separate mucinous cysts from serous cysts, avoiding a possible misdiagnosis with contaminant stomach or duodenum epithelia.51 Cytologic interpretation can be very misleading. A differential diagnosis with dragged contaminant epithelium of the stomach and duodenum can be a dilemma, especially in lesions with low-grade dysplasia. The cystic evaluation of amylase CEA and in selected cases of K-RAS mutations or DNA content can be of a great aid, separating benign from malignant lesions.52 K-RAS mutations or loss of heterozygosity in >2 loci is associated with mucinous cysts and helps predict malignancy.52 CEA is a good marker for distinguishing between mucinous and nonmucinous cysts, but it does not distinguish benign from malignant lesions.53 In negative K-RAS samples, the CEA level can capture almost 70% of cases. In scanty samples (<1 mL of fluid), it may be tempting to rely too heavily on the detection of K-RAS mutations; however, recent studies have demonstrated that molecular analysis has lower performance in predicting mucinous cysts than CEA analysis. The ideal procedure stems from the combined use of both techniques.53
Gastrointestinal stromal tumors
GISTs are characterized in >80% of cases by a mutation of the CKIT proto-oncogene with subsequent autonomous activation of the tyrosine kinase receptor that induces tumor development and proliferation.54 In 5% of GISTs, there is also an active mutation of the platelet-derived growth factor receptor (PDGFRA) gene. These mutations are mutually exclusive and represent 2 different alternative oncogenic events that lead to similar biologic consequences. Imatinib is first-line therapy for patients with advanced GISTs. It acts as an inhibitor of the tyrosine kinase receptor KIT and PDGFRA. The most common mutation, and also the most responsive to TKIs, is on exon 11 (juxtamembrane domain) of CKIT. The other involved mutations, which are less responsive in GISTs, are related to exon 9 (extracellular domain), exon 13, and exon 17 (tyrosine kinase domain). Also, PDGFRA mutations vary according to the exon involved. Exon 18 (tyrosine kinase domain) is much more responsive to imatinib than exon 12 (juxtamembrane domain). This is where FNA has its main molecular application in selecting patients for therapy. Our group previously demonstrated the feasibility of performing molecular analysis of CKIT and PDGFRA genes in cytologic material obtained by EUS-FNA from 85 patients with intramural gastrointestinal mesenchymal tumors.55 These patients also were studied for the expression of CKIT (CD117) by ICC in cell blocks. Strong membrane and cytoplasm immunostaining was observed in the majority of tumors that were preclassified as GISTs; also, PCR amplification followed by DNA sequencing revealed exon 11 mutations in 60% of cases.55 Recently, it was also demonstrated that the use of discovered on GIST 1 (DOG1) antibody in cytology cell blocks was more sensitive and specific than CKIT in the diagnosis of GIST.56 Indeed, those studies demonstrated that a precise preoperative diagnosis of GIST can be achieved with EUS-FNA, and the detection of mutations in cytologic samples also allows the prediction of therapeutic response, enabling greater efficiency in the use of neoadjuvant therapy.
Therapeutic target in colon cancer
Cetuximab and panitumumab are EGFR monoclonal antibodies that, either alone or in combination with chemotherapy agents, have been proven effective in the treatment of patients with metastatic colorectal cancer.57 KRAS gene mutations have been pointed out as possible molecular diagnostic markers for predicting the sensitivity of tumors to anti-EGFR therapeutics. Patients with tumors that harbor KRAS mutations have a lower drug response rate; however, a significant percentage of patients with KRAS-WT tumors also do not respond to anti-EGFR drugs.57 One of the many possible reasons for this is the discrepancy between KRAS status in the primary tumor and in metastasis. In a series of 250 patients with sporadic colon cancer who were analyzed in our institution, we detected 17% discordance between mutation status in KRAS and BRAF genes when comparing primary and metastatic samples.58 These results are of fundamental importance, because they may represent 1 of the resistance mechanisms interfering with the response of patients with KRAS-WT metastatic colorectal cancer to anti-EGFR monoclonal antibody therapy. Moreover, it is estimated that 20% of the target patient population will present with metastatic disease and there will not be access to archive material from the primary tumor. In both situations, FNA material obtained from metastatic sites is a good, safe, and cheap alternative for determining KRAS status in metastatic colon cancer. Recently, it was demonstrated by different groups that it is possible to analyze KRAS mutation status in cytology specimens obtained from colon cancer and lung cancer by using a simple and reproducible methodology.25, 59 In summary, we recommend the use of FNA samples from metastatic sites of colon cancer to assess KRAS mutation status and to improve the selection process by detecting more patients who will benefit from treatment.
Soft Tissue Tumors
Small, blue, round cell tumor
Many soft tissue tumors harbor characteristic translocations or gene region amplification that can be assessed by molecular methods and that have a great impact on the cytologic diagnosis.60-62 In this field, we make special reference to the small, blue, round cell tumor group. Tumors like ES/PNET, alveolar rhabdomyosarcoma, synovial sarcoma (undifferentiated type), mesenchymal chondrosarcoma, and desmoplastic small round cell tumor are included in this group and are characterized by specific molecular changes. Morphologically these tumors are similar, presenting an undifferentiated small round cell pattern, and only an experienced cytopathologist will detect minor details that may lead to the correct diagnosis. Although ICC is the ancillary method of choice in a first approach to the diagnosis of these tumors, it does not always present conclusive findings. In these situations, molecular diagnosis increases diagnostic accuracy for differentiating lesions like Ewing sarcoma, rhabdomyosarcoma, synovial sarcoma, and desmoplastic small round cell tumor.
In >85% of Ewing family tumors (EFTs), there is a specific tumor-associated translocation, t(11; 22)(q24; q12), juxtaposing the EWS gene on chromosome 22 with the friend leukemia integration 1 transcription factor (FLI1) gene on chromosome 1 (Fig. 4). This translocation appears to emerge very early in tumorigenesis and is deeply involved in tumor initiation. The identification of EWS/ETS fusion in problematic cases is determinant in cytologic differential diagnosis, as demonstrated in the literature.63
Alveolar rhabdomyosarcoma also is characterized by 2 main specific translocations: t(2;13)(q35;q14) in 70% of cases and t(1;13)(p36;q14) in 10% to 20% of cases. These translocations, apart from determining the cytologic diagnosis, reflect a hint for prognosis. Patients who had alveolar rhabdomyosarcoma with t(1;13) reportedly had a better prognosis than those who had t(2;13)(q35;q14).60 In a study using FISH probes, it was demonstrated that only 66% of patients with alveolar rhabdomyosarcoma had rearrangement of forkhead box 01 a (FOXO1A) (13q14). Therefore, a negative FOXO1A FISH study alone does not exclude the diagnosis of alveolar rhabdomyosarcoma; in these patients, the diagnosis rests on the recognition of characteristic histologic features and a supportive ICC profile. Therefore, it is important to state that, if sufficient morphologic and ICC evidence is present to suggest a diagnosis, then a negative FISH test does not necessarily exclude it, and additional ancillary studies may be necessary.9
Undifferentiated synovial sarcoma with a primitive, small, round cell pattern occurs in 15% of the cases and represents a challenge in the differential cytologic diagnosis with all the other small round cell tumors. These tumors express CD99 (a single-chain type 1 glycoprotein) and fail to immunoexpress most epithelial markers other than EMA. This profile makes a differential diagnosis with EFTs imperative. The detection through RT-PCR of a reciprocal translocation t(X;18)(p11.2;q11.2) or the demonstration of a synovial sarcoma translocation-chromosome 18 (SYT) gene rearrangement by FISH is observed in >90 % of cases, is highly specific for synovial sarcoma, and thus is essential for achieving a correct diagnosis.9 Recently, the use of dual-color break-apart chromogenic in situ hybridization was demonstrated in liquid-based cytology to demonstrate the translocations associated with EFTs and synovial sarcomas.63
Desmoplastic small round cell tumor (DSRCT) is a highly malignant mesenchymal neoplasm in which cytologic smears also present a malignant small round cell pattern. This tumor has a distinctive immune profile, coexpressing vimentin and epithelial, muscle, and neural markers. Characteristically, the expression of desmin typically has a dot-like, paranuclear pattern. Muscle-specific markers, like MyoD1 or myogenin, are negative. CD99 also can be expressed in 35% of the cases, diffusely staining the cytoplasm. Cytogenetically, DSRCT is characterized by a reciprocal translocation, t(11;22)(p13;q12).9, 60, 61, 64
Tumors with rhabdoid phenotype
In soft tissue tumors, as indicated above for renal tumors, a rhabdoid phenotype is uncharacteristic and raises difficulties in the diagnosis of different entities. Extrarenal rhabdoid tumors, except for the absence of immunoexpression for the INI1 marker, do not have any specific marker that can helpful in the differential diagnosis with other rhabdoid phenotype-sharing tumors, such as PNETs and medulloblastomas. The karyotype can be obtained directly from FNA cytology samples, or DNA can be isolated and analyzed by PCR-based microsatellite analysis for loss of heterozygosity using probes that map to 22q. From 30% to 40% of medulloblastomas have an isochromosome 17q; and, to date, this molecular change has not been observed in rhabdoid tumors. Recently, the accuracy of FNA was demonstrated in the diagnosis of rhabdoid tumors using ancillary techniques like karyotyping and FISH analysis to detect the loss of probe signals for the switch/sucrose nonfermentable-related, matrix-associated, actin-dependent regulator of chromatin, subfamily b, member 1 (SMARCB1) gene locus.65
Low-grade myxoid tumors
Low-grade myxoid neoplasms represent another group of soft tissue tumors in which the cytologic aspects are not enough to achieve a correct diagnosis. The differential diagnosis in this group includes low-grade fibromyxoid sarcoma, myxoma, myxofibrosarcoma, and myxoid liposarcoma. FISH probe set for the fused in sarcoma (FUS) gene region (16p11) will detect the split apart in either t(7,16) or t(11,16) in low-grade fibromyxoid sarcoma or t(12,16) in myxoid liposarcoma. The FISH test for the DDIT3 gene region (12q13) that is split apart in t(12,16) and t(12,22) is extremely useful and specific in the diagnosis of myxoid and round cell liposarcoma.9
Atypical lipomatous tumors
The differential diagnosis between atypical lipomatous tumors (well differentiated liposarcoma) and lipoma using FNA is puzzling. Murine double-minute 2 homolog (MDM2) amplification detected by FISH is very helpful in this situation. Well differentiated liposarcomas are characterized by MDM2 gene amplification, whereas all variants of lipomas (including spindle cell and pleomorphic lipomas) are negative for amplification of MDM2.9 Although immunohistochemistry for MDM2 is available, it is not as sensitive or specific as MDM2 FISH for the diagnosis of atypical lipomatous tumor. MDM2 FISH also is used to provide support for the diagnosis of dedifferentiated liposarcoma when the well differentiated component is not sampled. Recently, the feasibility of using MDM2 FISH in FNA smears was demonstrated for the diagnosis of dedifferentiated liposarcoma.66
Clear cell tumors
Although they are rare, sometimes, the problem of a differential diagnosis between primary clear cell sarcoma and metastatic, malignant melanoma can be a reality in FNA. Morphology and ICC can be identical; and, in these cases, apart from clinical data, the detection of rearrangement in the EWSR1 gene region by FISH permits the diagnosis of a clear cell sarcoma that harbors the t(12,22) translocation.9
FNA is a valid alternative to surgical biopsy in a variety of clinical settings. If sufficient clinical and morphologic evidence is present to suggest a cytologic diagnosis, then a negative molecular test does not necessarily exclude the diagnosis, and other ancillary tests sometimes may be necessary. Perhaps in the near future, with new high-throughput technology, all recurrent genetic events will be described and will change pathologic diagnoses.67 However, until then, the morphologic appearance and ICC profile of these entities, in the hands of experienced pathologists, should be considered the gold standard for prognosis and therapeutic guidance. Currently, cytopathologists are expected to include specific prognostic and predictive information in their reports, to order ancillary tests, and to contribute their expertise to clinical trials. Table 2 summarizes the main molecular markers and techniques actually used that have practical application in the solid tumors that were sampled by FNA in this review.
|Solid Tumor Type||Molecular Markers and Techniques|
|Lung||SCLC: Chromogranin, CD56, synaptophysin by ICC|
|NSCLC: P63/CK5/6 and TTF1/Napsin-A by ICCa; EGFR mutation status by sequencing mutation or FISHb; EML4-ALK translocation by FISHb|
|Thyroid||Medullary carcinoma: TTF1; calcitonin; chromogranin; CEA by ICCa|
|Follicular/papillary carcinoma: CK19, galectin-3, and HBME-1 by ICCa; BRAF, RAS, PAX8/PPAR and RET/PTCa|
|Kidney||See Table 1|
|Gastrointestinal tumors||Pancreatic ductal tumors: KOC, loss of SMAD4 by ICCa|
|Pancreatic endocrine tumors: Chromogranin, CD56, synaptophysin by ICCa|
|Pancreatic solid pseudopapillary tumors: Vimentin, alpha-1-antitrypsin, CD10, beta-catenin and progesterone receptor, chymotrypsin, and trypsin by ICCa|
|Cystic lesions: Amylase, CEA, and K-RAS mutationsa|
|GISTs: DOG and CKIT by ICCb; CKIT and PDGFR mutationa|
|Colon cancer: KRASb|
|Soft tissue||Small round cell tumors: CD56, NSE, chromogranin, synaptophysin, CD99, FLI1, EMA, caveolin, BCL2, miogenin, MYO D1, WT1 by ICCa; N-Myc, del1p, ploidy, search for characterizing specific translocationsb|
|Other sarcomas: Specific immunomarkers and characterizing specific translocations (see text)ab|
At the time of tumor progression, the assessment of newly acquired genetic alterations also may considered in selecting the next line of therapy. FNA sampling of surgical specimens is an effective tissue-sparing method for tissue collection and banking.68 High-quality RNA can be obtained from these cells and allows molecular studies even in small tumors. The use of FNA to obtain cells for genome sequencing is a great and possible path for additional exploration.67 However, because therapies are now being directed toward individual molecular targets, the big challenge for the use of all this technology in cytology is the need to increased standardization of preanalytic and analytical methods.