New Developments in Salivary Gland Pathology: Clinically Useful Ancillary Testing and New Potentially Targetable Molecular Alterations

Salivary gland tumors are increasingly being found to have characteristic chromosomal rearrangements. Although our understanding of the biology behind these rearrangements and their role in tumorigenesis is not often complete, these rearrangements are being recognized as useful diagnostic markers of salivary gland tumors, especially on small biopsy specimens, including aspiration cytology. Some of these translocations have been known for decades and are only now being considered as potential diagnostic tests, whereas other translocations have been recognized more recently. The ability to detect these rearrangements with FISH, and even with IHC, is making them more clinically significant in the diagnosis of salivary gland tumors (Table 1). This section will focus on the potential diagnostic utility of these chromosomal rearrangements.

Pleomorphic adenoma (PA) is the most common salivary gland neoplasm. Although the diagnosis is fairly straightforward in many cases, at times there is a challenging differential diagnosis between PA and a malignant salivary gland neoplasm, such as adenoid cystic carcinoma (AdCC), polymorphous low-grade adenocarcinoma (PLGA), myoepithelial carcinoma, or epithelial-myoepithelial carcinoma (EMCA). In some cases, IHC may help to resolve the differential diagnosis but immunophenotypic overlap among different diagnostic entities also exists. Another potential ancillary test that has not received much clinical attention to date is an investigation into characteristic translocations in PA.

One of the first chromosomal rearrangements described in salivary gland tumors was described^1,2  in PA nearly 2 decades ago: PA commonly shows rearrangements involving 8q12, resulting in increased pleomorphic adenoma gene 1 (PLAG1 zinc finger) gene expression. The most commonly detected recurrent chromosomal rearrangement detected in PA is the t(3;8)(p21;q12), resulting in increased PLAG1 gene expression because its transcriptional regulation is placed under the control of the constitutively active β-catenin CTNNB1 gene promoter.¹  Another common chromosomal rearrangement in PA is t(5;8)(p11;q12), which, in a similar manner, causes expression of PLAG1 under the regulatory elements of the leukemia inhibitory factor receptor-α (LIFR) gene on chromosome 5.²  Overall, approximately one-quarter of PA cases show detectable rearrangement of PLAG1 by reverse transcription-polymerase chain reaction.³  Other genetic changes resulting in increased PLAG1 expression include insertions and amplifications.⁴ 

Rearrangements involving PLAG1 in PA cause an overexpression of the protein, which can be detected with PLAG1 IHC. In fact, studies have shown that PLAG1 IHC is more sensitive for the diagnosis of PA than FISH is, with nearly all PA results testing positive by PLAG1 IHC.³  When evaluating PLAG1 IHC, one should look for nuclear reactivity to consider a case positive. In addition, PLAG1 IHC is most helpful in small biopsy specimens, when the overall architecture of the lesion cannot be evaluated. In studies examining the use of PLAG1 IHC, most malignant salivary gland tumor results, including those for AdCC and EMCA, which often enter the differential diagnosis, have been shown to be negative for PLAG1.⁵  In contrast, most PA (34 of 36; 94.4%) and all benign myoepitheliomas results (n = 8) are positive for PLAG1.

Unfortunately, PLAG1 IHC results are also positive in approximately one-quarter of cases of PLGA, one of the more difficult malignant lesions to distinguish from PA.⁵  Although negative PLAG1 IHC results argue against a diagnosis of PA, positive staining does not exclude the possibility of malignancy, and it may not be possible to entirely exclude PLGA on biopsy. A biphasic growth pattern, with luminal ducts and abluminal myoepithelial cells, favors PA over PLGA, which is characteristically monophasic. Similarly, although there is usually strong, diffuse S100 reactivity in PLGA, more-specific evidence of myoepithelial differentiation (ie, p40, smooth muscle actin, calponin) is usually lacking or only focal and weak in PLGA. On resections, evidence of invasion, including perineural invasion, is important for the diagnosis of PLGA, as opposed to PA.

Although PLAG1 IHC is less specific for PA than FISH is, even detection of a PLAG1 chromosomal rearrangement by FISH does not always indicate a benign diagnosis because carcinoma ex-PA also shows this genetic change when arising from a preexisting PA with rearrangement.^6–8  In fact, detection of PLAG1 rearrangements have been used to show that many SDCs are actually ex-PA, including those without morphologic evidence of a precursor lesion.⁹ 

PLAG1 gene rearrangements are not limited to salivary gland tumors, and nonsalivary gland tumors with a morphology similar to PA can show similar genetic alterations. Chondroid syringoma of the skin is nearly identical to PA morphologically, and PLAG1 expression can be detected by IHC; while some authors¹⁰  have reported PLAG1 gene rearrangements, this has not been the case in all series.¹¹  The fact that PLAG1 rearrangements were detectable by FISH and not by reverse transcription-polymerase chain reaction suggests the possibility of different partner genes in cutaneous chondroid syringoma.

HMGA2 (previously HMGIC) gene rearrangements or amplifications are also seen in PA and carcinoma ex-PA but are less common^7,8  and are not, to our knowledge, currently used for testing in a clinical setting. In addition, HMGA2 rearrangements are detectable in a variety of soft tissue tumors, including lipoma, leiomyoma, hamartoma, and fibroadenoma of the breast.^12,13 

Mucoepidermoid carcinoma (MEC) is the most-common malignant salivary gland tumor and is associated with one of the more clinically useful translocations, involving the MAML2 gene. A t(11;19)(q21;p13) translocation between MAML2 on chromosome 11 and CRTC1 (previously MECT1; TORC1 [alias]) on chromosome 19 was first described in MEC in 2003.¹⁴  Less commonly, the partner gene with MAML2 is CRTC3 on chromosome 15, t(11;15)(q21;q26).^15,16  The use of MAML2 FISH has been proposed as a useful ancillary test in the routine clinical diagnosis of salivary gland tumors in which MEC enters the differential diagnosis.¹⁷ 

Clinically, detection of a MAML2 rearrangement using break-apart FISH probes is helpful in several situations. In cases of low-grade MEC, differential diagnostic considerations can often include benign entities, such as Warthin tumor (WT), or more rare lesions such as lymphadenoma. In such cases, detection of MAML2 rearrangement confirms the diagnosis of MEC. Although some authors have reported finding MAML2 translocations in WT¹⁸  (some even suggest that MEC can arise from WT [ie, MEC ex-WT]¹⁹), others argue these cases may represent low-grade MEC mimicking WT, and most consider the finding of a MAML2 rearrangement to strongly favor a diagnosis of MEC over WT.^20–25 

A particular scenario in which MAML2 FISH can be extremely useful is in the setting of an oncocytic lesion in which the possibility of an oncocytic MEC is entertained. Because of the prominent oncocytic morphology masking the epidermoid phenotype, the oncocytic variant of MEC, in particular, can pose a challenging differential diagnosis with WT, oncocytic cystadenoma, and acinic cell carcinoma. Although p63 reactivity can be useful in many cases to suggest a diagnosis of MEC, detection of MAML2 rearrangement confirms a diagnosis of MEC.^26,27  Although many of the oncocytic MECs that enter the differential diagnosis with benign mimickers of WT and oncocytic cystadenoma will be low grade and unlikely to metastasize, the diagnosis of malignancy may nevertheless be important in highlighting the potential for more-frequent recurrence.

At the other end of the spectrum, high-grade MEC can mimic a variety of other high-grade cancers, including metastatic squamous cell carcinoma, adenosquamous carcinoma, and SDC. Again, in these instances, identification of a MAML2 rearrangement is diagnostic for MEC, which can be important because, in the parotid gland, this excludes the possibility of a metastasis, and there are treatment and prognostic implications. One caveat is that high-grade MECs are less likely to have MAML2 rearrangements, and so, the absence of a rearrangement does not rule out a diagnosis of MEC.

The presence of MAML2 rearrangement correlates with prognosis and stage because tumors with the rearrangement tend to be less aggressive and of lower histologic grade.^18,21,23,28  MAML2 rearrangements have been detected in up to three-quarters of low- to intermediate-grade MEC, but fewer than one-half of high-grade MECs are positive for this rearrangement in most series. Although MAML2 rearrangements are not required for a diagnosis of MEC, an attempt should be made to exclude other tumors in the differential diagnosis of MAML2⁻, high-grade MEC.^23,29,30  Differential diagnostic considerations for such tumors include SDC, adenosquamous carcinoma, and even squamous cell carcinoma; all of which, in general, have a poorer prognosis than high-grade MEC. SDC frequently stains positive for androgen receptor (AR) and is p63⁻, which contrasts with MEC, which is usually AR negative and has at least focal reactivity with p63. Adenosquamous and squamous cell carcinoma are generally considered to arise from surface epithelium, in contrast to MEC, which does not have a surface component in most cases. The presence of more than just focal keratinization favors a diagnosis of squamous cell carcinoma over MEC.

In addition to salivary gland MEC, MAML2 rearrangement can also be detected in central (intraosseous) MEC³¹  and MEC of the lacrimal gland,³²  thyroid,¹⁸  thymus,³³  lung,^34,35  and cervix.³⁶  MAML2-CRTC1 fusion has also been identified in some cases of cutaneous clear cell hidradenoma.³⁷  MAML2 can fuse with KMT2A (previously MLL) in some cases of acute myelogenous leukemia.³⁸  MAML2 rearrangements are not detected in glandular odontogenic cysts, which can enter the differential diagnosis, especially for a central (intraosseous) MEC.³⁹ 

A recurrent translocation, t(6;9)(q21-24;p13-23), was first recognized in AdCC in the early 1990s^40,41  and was more recently found to result in the fusion of the MYB-NFIB genes.⁴²  Although other tumor-associated translocations are usually identified in considerably more than one-half of cases, MYB rearrangements are reported in only one-quarter to less than one-half of AdCCs in the literature.^43,44  Because MYB rearrangements are not seen in most cases of AdCC and because of some technical challenges with this FISH probe, MYB FISH is not as clinically useful as could be hoped, but larger laboratories may still have the test available.

Although MYB FISH has its challenges, the good news is that a MYB antibody is also available, and studies have shown increased expression of MYB by IHC in a greater proportion of AdCCs, including translocation-negative cases.⁴⁵  Of note, the immunoreactivity of MYB in AdCC is expected to have strong nuclear staining of the abluminal myoepithelial cells, with weaker, patchy reactivity in the luminal ductal cells (despite the chromosomal translocation being detectable in all cell types in translocation-positive tumors) (Figure 1). However, with this increased sensitivity, some of the specificity is lost with MYB IHC because some mimickers of AdCC occasionally express MYB. In a study including 113 non-AdCC tumors, 16 (14.2%) had positive results, including 4 of 5 basaloid squamous carcinomas (80%), 2 of 7 basal cell neoplasms (adenoma/adenocarcinoma; 28.6%), 1 of 3 EMCAs (33.3%), and 2 of 13 PLGAs (15.4%).⁴⁵  By comparison, 56 of 68 (82.4%) of the AdCCs had positive results (>5% of tumor cells with strong nuclear reactivity).

Outside of the head and neck, MYB-NFIB is also seen in AdCC of breast⁴²  and in benign dermal cylindroma.⁴⁶ 

Mammary analogue secretory carcinoma (MASC) is one of the most recently described salivary gland tumors, first reported by Skalova et al⁴⁷  in 2010. The fact that MASC, like its breast counterpart, is associated with rearrangement of the ETV6 gene has been a great benefit in the early identification and characterization of this entity from morphologic mimickers (predominantly acinic cell carcinoma [AciCC] and adenocarcinoma, not otherwise specified), which has allowed a relatively quick and robust recognition of this entity.⁴⁸  The most common translocation partner with ETV6 is NTRK3 because of a t(12;15)(p13;q25) translocation; however, rare tumors have been found to show ETV6 rearrangement involving another, as yet unknown, gene (ie, ETV6-X), and these cases may have more aggressive behavior.^49,50 

Morphologically, MASC is fairly characteristic, with tumor cells often having abundant cytoplasm and multiple, small vacuoles. However, the growth patterns can be quite varied, with solid, cystic, and papillary growth, as well as variable amounts of mucin production. Combined with the morphology, the characteristic finding by IHC of strong coexpression of both S100 and mammaglobin is supportive of the diagnosis. These features are useful in limiting the differential diagnosis and are especially good at separating MASC from the primary morphologic mimickers of AciCC and adenocarcinoma, not otherwise specified; however, other salivary gland tumors, such as PLGA and AdCC, can show a similar immunophenotype.^51,52  The resolution of this differential diagnosis can be challenging morphologically in some cases, especially on small biopsies, but the finding of an ETV6 rearrangement is diagnostic for MASC in a primary salivary gland tumor.⁵³ 

Outside of the salivary glands, ETV6 rearrangements can be detected in tumors arising from all embryologic layers. As noted, secretory carcinomas of the breast also commonly have an ETV6 rearrangement and have a nearly identical morphology to MASC. Other tumors with ETV6-NTRK3 rearrangements include infantile fibrosarcoma and congenital mesoblastic nephroma. Some cases of leukemia also have ETV6 rearrangements but with different partner genes.

Hyalinizing clear cell carcinoma (HCCC) is a rare salivary gland malignancy with squamous differentiation and prominent clear cell morphology. Although the finding of infiltrative nests of low-grade clear tumor cells with hyalinized and cellular fibrous stroma is quite characteristic of HCCC, several other tumor types, such as myoepithelial carcinoma, EMCA, and MEC, can enter into the differential diagnosis.

Recently, it was discovered that most HCCCs have a recurrent chromosomal rearrangement leading to fusion of EWSR1 and ATF1, t(12;22)(q13;q12), but this rearrangement was not detected in myoepithelioma, PLGA, MEC, or EMCA.^54,55 

Initial studies failed to identify EWSR1 rearrangements in other classic salivary gland tumors, but more recently, Skalova et al⁵⁶  reported EWSR1 rearrangements in 25 of 72 (34.7%) of clear cell myoepithelial carcinomas (de novo and ex-PA) as well as in 1 of 11 (9%) of EMCAs. Cases of EWSR1-rearranged myoepithelial carcinoma commonly had at least focal areas of necrosis, squamous pearls, and significant nuclear pleomorphism. In cases in which the differential diagnosis includes an EWSR1-rearranged myoepithelial carcinoma and HCCC, the diagnosis depends more heavily on morphology and immunophenotypic markers because HCCC only infrequently shows focal myoepithelial differentiation with antibodies such as S100.

EWSR1 rearrangements are also seen in a variety of epithelial and soft tissue tumors, including soft tissue myoepithelioma; however, the partner genes that are common in myoepithelioma (FUS, POU5F1, PBX1, and ZNF444) have not been shown to be involved in HCCC.^54,57  EWSR1-ATF1 was reported in one case of soft tissue myoepithelioma in the pelvis.⁵⁸  EWSR1-ATF1 has also been described in angiomatoid fibrous histiocytoma⁵⁹  and clear cell sarcoma,⁶⁰  although the exons involved are different.

Polymorphous low-grade adenocarcinoma occurs most commonly in oral minor salivary glands and has polymorphous architectural growth patterns but is composed of monotonous tumor cells. Recently, a similar entity, termed cribriform adenocarcinoma of minor salivary glands (CAMSG) was described, and there is ongoing debate as to whether these 2 entities are distinct or represent different ends of a morphologic spectrum. The proposed morphologic distinction between these 2 entities is the finding in CAMSG of more prominent solid architecture with glomeruloid proliferation of papillae forming narrow slitlike spaces at the periphery. Molecular alterations of the PRKD genes (PRKD1, PRKD2, and PRKD3) have been described in both entities. Weinreb et al⁶¹  discovered a recurrent translocation involving the PRKD genes in a series of 60 PLGA and CAMSG cases, in which nearly one-half showed a rearrangement of one of these genes, most commonly PRKD1. Partner genes included ARID1A and DDX3X, but the partner gene was not identified in all cases. Most cases with rearrangement were blindly classified as CAMSG or were judged to have indeterminate morphology, whereas only one case categorized as PLGA showed rearrangement of PRKD2.

The use of a clinical assay to detect PRKD rearrangement may eventually be useful in the differential diagnosis of PLGA/CAMSG versus PA and AdCC but has not yet been adopted clinically. PRKD1 point mutations (E710D) were also recently reported to occur in nearly three-quarters of PLGA cases.⁶²  Mutations in PRKD2 and PRKD3 were not found in PLGA.⁶³ 

Frequent cytomorphologic overlap ensures constant challenges on salivary gland tumor fine-needle aspirations (FNAs), and a precise diagnosis is not always possible. In addition, consensus on a standard terminology for indeterminate cases is still lacking. However, a new classification scheme for reporting salivary gland cytopathology—The Milan System—is currently under consideration and may offer a more-unified set of diagnostic terms in the future. Although the goals of this group are not entirely clear yet, it is possible that such a system could provide a risk-stratification–based approach to salivary gland FNAs, similar to that which was provided by the Bethesda System for Reporting Thyroid Cytopathology.

A recent study by Griffith et al⁶⁴  also addressed the possibility of a risk-based stratification scheme, categorized by a pattern-based recognition approach. This study retrospectively reviewed 294 aspirates of primary salivary gland neoplasms with surgical follow-up and separated smears into basaloid or oncocytoid groups according to nuclear features, stromal features, and background characteristics. Basaloid neoplasms were further stratified based on stromal characteristics in monomorphic smears with an additional category of pleomorphic basaloid neoplasms. Similarly, monomorphic oncocytoid neoplasms were subdivided by background features (eg, mucus) and cytoplasmic granularity. As with basaloid neoplasms, a separate group of pleomorphic oncocytoid neoplasms was also defined. The risk of malignancy was calculated for each of the categories. This risk-stratification scheme was applied only to primary epithelial neoplasms of the salivary glands, with lymphoproliferative, mesenchymal, and metastatic lesions excluded from analysis. However, in practice, metastatic lesions to the parotid can be morphologically indistinguishable from a primary, especially among pleomorphic lesions.

The monomorphic basaloid and oncocytoid categories were further divided to provide additional risk stratification.⁶⁴  Monomorphic basaloid neoplasms were separated by stromal characteristics into patterns having fibrillary, hyaline, and mixed/other stroma. Similarly, monomorphic oncocytoid neoplasms were divided based on the background features into those with mucinous backgrounds, cyst content backgrounds, and other backgrounds. In this section of our review, we discuss new ancillary testing on salivary gland FNA smears and/or cell blocks applied to each of the cytomorphologic categories because they provide a pattern-based approach to salivary gland cytology.

The primary differential diagnostic considerations for monomorphic cellular basaloid neoplasms include cellular pleomorphic adenoma, EMCA, basal cell adenoma/adenocarcinoma, and AdCC. Benign and malignant lesions within this group can mimic each other perfectly, although the finding of fibrillary stroma is a lower-risk feature for malignancy than either hyaline stroma or mixed/other stroma. Unfortunately, immunohistochemical or molecular studies to differentiate among these lesions on cytology specimens are not always reliable, but several recent studies suggest clinical utility for some new IHC markers and FISH testing.

One of the challenging issues within the basaloid neoplasm category is the diagnosis of AdCC on cytology aspirate. The perennial differential diagnosis of basaloid neoplasms often included AdCC, but many aspirates of AdCC are not clearly diagnostic for this entity.⁶⁵  Recent studies have suggested that the use of MYB IHC and/or FISH testing on FNA material may be useful in resolving the differential diagnosis.

Pusztaszeri et al⁶⁶  investigated the role of MYB IHC on alcohol-fixed Papanicolaou-stained FNA tests of monomorphic basaloid lesions, including 20 AdCCs and 20 PAs. Positive MYB staining was noted in most AdCCs (80%; 16 of 20), whereas all PA cases were negative. One potential difficulty in the use of MYB IHC, however, is that the results are not purely negative or positive, with only cases showing strong, diffuse reactivity considered positive (scores ≥4 of 6, in this article). Approximately one-quarter of PA smears showed some reactivity for MYB (ie, scores of 2 and 3), indicating that it is important to consider the quantity and quality of staining with this antibody. An additional pitfall when interpreting MYB IHC results on cytology samples is that a subset of lymphocytes can also be strongly MYB⁺, but in most cases, this would not represent most of the cells.

In a similar study, MYB gene rearrangements were detected with break-apart FISH probes on air-dried, modified Wright-Giemsa–stained smears.⁶⁷  Hudson and Collins⁶⁷  detected an MYB abnormality in one-half (5 of 10) of the AdCCs, whereas none of the 13 cases of PA showed any MYB abnormality.

Both of these studies confirm that MYB abnormalities can be detected in cytologic samples to aid in the diagnosis of AdCC. In both series, detection of a MYB abnormality was specific for AdCC, but only cases of AdCC and PA were included. Importantly, other entities that occasionally enter into this differential diagnosis may show some MYB reactivity by IHC, including PLGA.⁴⁵  Additionally, published studies using MYB IHC or FISH have not included cell block material, which is the most commonly used sample in many clinical laboratories. Given that MYB IHC is more sensitive, but less specific, for AdCC, some have suggested a 2-stage testing strategy using positive IHC result as a triage for MYB FISH.⁶⁸ 

Other markers that could be useful in the setting of a basaloid neoplasm in which AdCC is in the differential diagnosis have been investigated on resection specimens, rather than cytologic material. These include CD117, GFAP, PLAG1, and HMGA2. S100 is positive in most benign and malignant lesions within the monomorphic basaloid lesions group; therefore, its usefulness when attempting to distinguish among the monomorphic basaloid neoplasms is limited.⁶⁹  Testing the above IHC markers on FNA cell block material as part of a panel may prove to be useful in some cases, but further studies that include many cases and diagnostic entities will need to be done to confirm diagnostic accuracy.

Pleomorphic basaloid neoplasms are not commonly seen in aspirates of salivary glands and are usually easily diagnosed as malignant by morphology alone. However, the differential diagnosis for these aspirates includes both primary salivary gland tumors and metastatic lesions, such as metastatic high-grade neuroendocrine carcinoma or metastatic nonkeratinizing squamous cell carcinoma. Primary parotid tumors that can show a pleomorphic basaloid neoplasm on aspiration include lymphoepithelial carcinoma, salivary gland carcinomas with high-grade transformations, or occasional low-grade malignancies, such as EMCA. Despite their morphologic similarity, differentiating among these lesions may be possible in selected cases, and this can be important in guided clinical management because primary and metastatic lesions are treated differently in many cases.

Metastatic, nonkeratinizing squamous cell carcinoma (nasopharynx and oropharynx) to the parotid/neck level II lymph node area is uncommon but may be indistinguishable from primary parotid lymphoepithelial carcinoma. Strong and diffusely positive p16 IHC favors metastasis from an oropharyngeal primary but is not as specific as human papillomavirus in situ hybridization in the setting of an unknown primary.^70,71  Negative (or only weak/focal) p16 staining and positive Epstein-Barr virus in situ hybridization is less informative but argues against metastasis from an human papillomavirus–driven, primary oropharyngeal tumor. Although Epstein-Barr virus reactivity often suggests a nasopharyngeal primary, parotid primary lymphoepithelial carcinoma also often stain positively for Epstein-Barr virus in situ hybridization, especially in endemic areas. In the recent cytology case series of parotid lymphoepithelial carcinoma described by Hipp et al⁷² 1 of 6 cases (16.7%) tested for Epstein-Barr virus in situ hybridization had positive results, including one from a Chinese patient. Ultimately, correlation with imaging and clinical exam is recommended to confirm the primary site.

The differential diagnosis for monomorphic oncocytoid neoplasms includes benign and malignant entities, such as WT, oncocytic cystadenoma, oncocytoma, MEC, AciCC, MASC, and metastatic renal cell carcinoma. Although background and cytonuclear features can aid in limiting the differential diagnosis, ancillary studies are also helpful in this setting. Recently, Schmitt et al^73,74  investigated the role of SOX10, DOG1, S100, and p63 IHC in helping to resolve the differential diagnosis of monomorphic oncocytoid neoplasms on FNA cell-block material in some cases. Positive staining for SOX10 and DOG1 reliably distinguished AciCC on FNA cell blocks from its most-common benign mimickers, WT and oncocytoma. In contrast to the strong apical reactivity seen with DOG1 in nonneoplastic serous acini, AciCC is usually characterized by strong, complete, membranous and cytoplasmic staining⁷⁵  (Figure 2). Moreover, MASC was usually positive for S100, SOX10, and mammaglobin but negative for DOG1 and p63.^53,76,77  In contrast, MEC was usually negative for SOX10, S100, and DOG1 but shows at least focal p63 reactivity. Both WT and oncocytoma were consistently negative for DOG1 and SOX10 but frequently are positive for p63; therefore, the differential diagnosis with MEC can often be challenging on FNAs. Pitfalls when interpreting SOX10 and DOG1 on FNA cell blocks include positive staining in nonneoplastic salivary gland acini and rare, focal, and faint staining in MEC.^73,74 

SOX10 and DOG1 can also stain monomorphic basaloid lesions, such as PA.^75,78  Rarely, PA can show a predominance of plasmacytoid cells with oncocytoid cytoplasm, which, when combined with positive DOG1 and SOX10 IHC results, could lead to an erroneous diagnosis of AciCC. In this situation, p63 and S100 can be helpful as sensitive myoepithelial markers that are usually positive in PA and negative in AciCC.^74,79,80 

The phosphorylated form of signal transducer and activator of transcription 5 (p-STAT5) IHC has also been investigated on salivary gland cytology smears by Kawahara et al.⁸¹  p-STAT5 immunoreactivity was detected in all 7 cases of MASC, whereas none of the other 10 salivary gland aspirates were p-STAT5⁺. This study also demonstrated that staining for mammaglobin was less sensitive and less specific for MASC compared with p-STAT5.

Although IHC is often very helpful in the differential diagnosis of salivary gland tumors on FNA, difficulties in interpretation and cross-reactivity among tumors often only suggest one diagnosis over another and may not always provide a definitive finding of malignancy. If a definitive tumor classification is needed preoperatively, ancillary testing with FISH on cytology material may allow this in some cases. The t(12;15)(p13;q25) ETV6-NTRK3 translocation characteristic of MASC can be reliably detected on FNA cell blocks and is diagnostic of MASC in the setting of a primary salivary gland tumor.⁵³  Data on the biologic behavior of MASC is still being collected, but there are indications that MASC may have a slightly more-aggressive behavior than AciCC has, with a greater frequency of lymph node metastasis; if that is confirmed, more-precise preoperative classification may be useful in selecting patients for neck dissection in a neck that otherwise has clinically and radiographically negative results.⁸² 

Pleomorphic oncocytoid neoplasms are usually easily diagnosed as malignant on FNA. The primary differential diagnoses for these lesions include SDC, high-grade MEC, metastatic carcinoma (adenocarcinoma or squamous cell carcinoma), and metastatic melanoma. Although IHC may help resolve this differential (ie, AR positivity in SDC), no IHC is entirely specific, and in most cases, clinical and radiologic correlation are paramount in determining whether a lesion is primary or metastatic. In cases of high-grade MEC that would enter the differential diagnosis of a pleomorphic oncocytoid neoplasm, the cellularity of cytology samples may be sufficient for FISH testing for a MAML2 translocation, although a negative result does not exclude the diagnosis of MEC.

Other primary salivary gland carcinomas that could show overlapping morphology with pleomorphic oncocytoid lesions and pose a diagnostic challenge on FNA include MECs and AciCCs. The MECs show similar IHC profiles to nonkeratinizing squamous cell carcinoma, but the presence of MAML2 rearrangement by FISH could indicate MEC; however, the value of MAML2 FISH testing has not yet been verified on cytology samples. Positive DOG1 and SOX10 and negative p63 staining on FNA cell blocks are supportive of a diagnosis of AciCC, especially if used as a panel, as discussed above.^73,74 

Surgical resection is the primary treatment for patients with salivary gland neoplasia. The classic role of the pathologist is in careful gross examination, macrodissection, and microscopic assessment of the tumor, with the provision of a definitive diagnosis (including grading where appropriate) and identification of key prognostic features for inclusion in the pathologic staging summary, including surgical margin status. These components provide the information upon which virtually all subsequent care is based. For patients with an aggressive histologic variant of salivary gland cancer, advanced pathologic stage, positive surgical margins, and/or regional metastasis, adjuvant radiation therapy is indicated.

Chemotherapy is generally reserved for patients with unresectable, symptomatic, or rapidly progressing, locoregional disease or distant metastasis, with cytotoxic chemotherapy as first-line therapy.⁸³  In these situations, patient selection, therapy selection, and minimization of drug toxicity are critical. Cytotoxic chemotherapy disrupts healthy and neoplastic tissues in a relatively nonspecific manner, and an ideal drug therapy would be tailored to address known salivary gland tumor gene fusions, their downstream targets, or other unique targetable pathways to optimize efficacy and minimize side effects. Targeted drug therapy has been used in the management of other malignant diseases and in combination with preclinical studies of salivary gland tumors; these studies often provide the rationale for use in patients with salivary gland tumors. There are no specifically designed drugs for the treatment of salivary gland malignancies, and the agents used currently are most often those employed in other cancer types; therefore, the assessment of salivary tumor biomarkers before initiation of therapy may optimize success. Pathologists are involved in the triage of tumor tissue subsequently used in testing before targeted drug therapy, as well as in the process of selection and interpretation of testing, such as key IHC stains and/or FISH. The following section explores the results of novel, targeted therapies with a focus on select, common and/or biologically aggressive, salivary gland malignancies.

Multiple biologic markers have been evaluated in AdCC, including epidermal growth factor receptor (EGFR), HER2/neu, KIT, and hormone receptors. EGFR, a transmembrane tyrosine kinase, has an important role in cancer signaling. In AdCC, EGFR may be overexpressed in approximately 33% (6 of 18) to 85% (23/27) of tested cases,^84–88  although other groups have reported EGFR overexpression in fewer than one-quarter of cases.^89–91  Gene amplification and/or activating mutations of EGFR, on the other hand, are quite rare.^85,92,93  Researchers studying multiple anti-EGFR agents in the management of AdCC have found little objective responses, although it is possible such agents improve the response to conventional chemotherapy. In light of studies showing modest rates of stable disease and relative clinical benefit, in some cases, the use of anti-EGFR therapy continues to be studied.^84,94–96 

HER2/ERBB2 overexpression and amplification are virtually absent in AdCC by IHC and FISH.^{84,88,93,97,98}  Moreover, AdCC does not express markers of hormonal sensitivity, such as androgen, estrogen, or progesterone receptor (PR), and as a result, there is insufficient evidence or rationale to support either anti-HER2–based therapy or antihormonal therapies.

Overexpression of c-Kit is identified in more than three-quarters of salivary gland AdCC cases, but activating mutations are exceedingly rare.^{86,90,99–101}  The transmembrane cell surface tyrosine kinase receptor Kit is encoded by the c-Kit proto-oncogene.¹⁰²  Several studies have examined the effects of treating recurrent and/or metastatic AdCC with the small-molecule tyrosine kinase inhibitor imatinib (Table 2). Imatinib was the first agent to be used in a targeted fashion in AdCC¹⁰³  and many trials were completed before description of AdCC molecular biology and/or c-Kit mutational status was known. Although successful in managing metastatic gastrointestinal stromal tumor,¹⁰⁴  targeted c-Kit inhibition has shown only a few cases with objective responses in AdCC^105–109  and a lack of c-Kit mutations in AdCC may be responsible for this lack of efficacy in AdCC.

Success in targeting transcription factors, such as MYB, is thought to be unlikely, therefore attention has turned to targeted inhibition of pathways known or thought to be regulated by MYB, as exemplified by recent studies examining the molecular biology of AdCC through whole exome and genome sequencing.^110–112  Because FGF/IGF/PI3K signaling abnormalities are present in up to one-third of patients with AdCC, the use of multikinase inhibitors, such as dovitinib and sunitinib, is being examined.^111,112  Although the treatment drug dose was poorly tolerated, patients treated with dovitinib exhibited stabilization of progressive AdCC in one recent study.¹¹³  Similarly, although objective responses were not observed, a trial involving sunitinib treatment showed disease stabilization in 62% (8 of 13) of the assessable patients.¹⁰³ 

The PI3K pathway is mutated in several cancer types, including a variety of salivary gland carcinoma types, suggesting that drugs with antitumor activity via inhibition of PI3K/AKT/mTOR pathway may yield a therapeutic response. Kim et al¹¹⁴  reported treatment with everolimus, an mTOR inhibitor, in 34 patients with recurrent/metastatic AdCC. Although no objective responses were seen, clinical benefit in the form of disease stabilization was observed in most patients. In 2013, Ganesan et al¹¹⁵  reported phase I results in the use of lenalidomide and temsirolimus in 4 patients with AdCC. One patient showed stable disease sustained for longer than 6 months in this small study population. Nelfinavir, an antiretroviral drug approved for use in HIV treatment, is thought to exhibit inhibition of the Akt signaling pathway and has shown promise as cancer monotherapy in preclinical studies.¹¹⁶  Hoover et al¹¹⁶  conducted a phase II clinical trial with nelfinavir in patients with advanced AdCC but no objective responses were observed.

Table 2 summarizes studies of targeted therapy in AdCC. Importantly, the slow growth of AdCC may limit traditional use of clinical study endpoints, such as stable disease. Without documented progression of disease before entering a trial, “disease stabilization” may represent antitumor activity of an agent or a period of relative inactivity in the natural history of a disease known to have a protracted course. To date, no targeted agent has shown clear advantage in the form of objective response in the management of patients with recurrent and/or metastatic AdCC.

Similar to AdCC, biomarkers such as EGFR, HER2/neu, c-Kit, and hormone receptors have been studied in SDC. EGFR expression is variably reported in cases of SDC as assessed by IHC^{85,88,91,93,117–119} ; activating mutations are uncommon,^118,120,121  and gene amplification is rare.^85,120  Regardless, several studies have tested anti-EGFR treatment in patients with recurrent and/or metastatic SDC,^84,94–96  and although no objective responses were observed, clinical benefit in the form of radiographic regression of metastases and/or stabilization was seen in some cases.^94,95 

The remarkable morphologic similarity between SDC and mammary carcinoma provides a rationale for testing SDC for amplification of HER2/ERBB2 by IHC and/or in situ hybridization and forms the basis for treatment with the anti-HER2 monoclonal antibody, trastuzumab. Although the reported rate of HER2 overexpression varies widely, reflecting variable methodology and approach to scoring, it appears that approximately 15% (10 of 66) to nearly 40% (64 of 173) of SDCs overexpress HER2.^{118,121–125}  One unresolved controversy revolves around the most efficacious manner in which to assess HER2. The entire gamut of testing strategies has been used in the literature, including testing by IHC only,^126–128  screening by IHC with subsequent FISH testing for borderline IHC cases,¹²⁹  and/or subsequent FISH testing regardless of IHC results (Table 3). Given differences in staining techniques, antibody clones, use of archival tumor blocks, and tumor heterogeneity, only studies using both HER2 IHC and FISH are considered in the following discussion (Table 3). Ten studies* performed FISH regardless of IHC score, and of 64 tumors demonstrating amplification, only 1 (1.6%) was identified among tumors thought to be negative for HER2 (0 or 1⁺) by IHC.¹¹⁸  This suggests good correlation between IHC and amplification, as assessed by in-situ hybridization techniques. Success with trastuzumab-based therapy for SDC has been modest overall, but this assessment must be tempered by the understanding that the literature includes marked heterogeneity in patient population and disease burden, variable treatment regimens, often retrospective small series and case reports subject to bias, and variable methods for assessing HER2 amplification.

Finally, a recent molecular analysis of SDC revealed the coexistence of ERBB2 amplification with either PIK3CA mutations (2 of 9 cases; 22.2%) or PTEN (2 of 9 cases; 22.2%) loss.¹³²  In patients with breast carcinoma, the combination of PIK3CA mutation or PTEN loss in combination with HER2 amplification has been associated with reduced rate of treatment response or trastuzumab resistance, but stage and treatment timing may represent confounding variables.^133,134  It is suggested, therefore, that evaluation of ERBB2 status without assessment of PIK3CA and PTEN status could theoretically undermine the success of trastuzumab-based therapies in SDC.¹³²  Table 4 summarizes studies incorporating anti-HER2 therapy in SDC and, whereas more-vigorous study is still necessary, the finding of multiple cases with an objective response offers hope for this avenue of treatment in a highly aggressive cancer.

Androgen receptor positivity occurs at a high and predictable rate in SDC with between 67% (56 of 84) to nearly 98% (179 of 183) of cases positive for AR nuclear reactivity in the literature.^90,135,136  Other hormonal receptors, such as estrogen and PR, are rarely identified in SDC.^88,131  The AR gene amplification appears to be uncommon, although gain of copy in chromosome X, corresponding to the location of the AR gene, has been identified in 37% (10 of 27) to 58% (7 of 12) of cases studied.^119,137  Although SDC resembles breast ductal carcinoma morphologically (specifically, the apocrine type), SDC shares many immunophenotypic similarities and androgen dependence with prostate carcinoma. In a recent Radiation Therapy Oncology Group study of patients with high-risk prostate carcinoma, neoadjuvant and concurrent treatment with androgen-deprivation therapy (ADT) was concluded to be more beneficial than was radiation alone.¹³⁸  This model of ADT provides a rationale for use of ADT in SDC and, although reported experience with ADT in SDC is limited to single-case reports and case series, results have been encouraging (Table 4).^{119,139–143} 

Although recurrent, fusion oncogenes have not been identified in SDC, identification of molecular alterations in SDC, such as PI3K pathway alterations, may provide novel therapeutic opportunities for patients with SDC. PIK3CA mutations are reported to occur in approximately 10% (1 of 10) to 44% (7 of 16) of SDCs studied,^{119–121,144,145}  and loss of PTEN, a negative regulator of PI3K signaling, has been variably reported.^{119,132,144,146}  Targeting the activity of downstream effectors of PI3K signaling, such as Akt and mTOR, may provide a viable therapeutic option for some patients.^123,144,147  One group¹¹⁹  recently highlighted a possible relationship between PI3K/AKT abnormalities, AR, and potential therapeutic implications for ADT. Although precise mechanisms remain to be clarified, ADT may result in HER3 overexpression, with a resulting increase in AR transcription, followed by PI3K/AKT activation. Recently, Locati et al¹¹⁹  identified HER3 expression in nearly three-quarters of patients with SDC tested. Should this sequence of events lead to ADT resistance, then consideration of ADT in combination with EGFR inhibition becomes a therapeutic possibility.¹¹⁹ 

Similar to AdCC and SDC, biomarkers have also been explored in MEC. EGFR is expressed in one-third to nearly three-quarters of MECs when assessed by IHC, although amplification and/or activating mutations are rare.^{85,86,90,91,97,148}  Despite the latter, it has been shown that some CRTC1-MAML2 fusion-positive MECs have increased EGFR signaling via amphiregulin (AREG), and that may stimulate growth of MEC.¹⁴⁹  This could provide support for testing/determination of MEC fusion status, if elevated levels of AREG and subsequent activation of AREG-EGFR signaling prove to be a pathway for targeting with anti-EGFR therapy.¹⁵⁰ 

In MEC, the reported rate of HER2 overexpression appears to be low. Butler et al¹²⁷  tested more than 70 cases of MEC without a single positive result by IHC. Nakano et al¹⁴⁸  identified ERBB2 gene amplification in approximately 14% (4 of 28) of all MEC, mostly in high-grade MEC, including one case with MAML2 rearrangement. Although HER2 abnormalities, as detected by ancillary testing, have been associated with worse outcome in MEC,¹⁴⁸  the described problem of distinguishing high-grade MEC from SDC could undermine the strength of the association.¹²⁷  Few studies of anti-HER2 therapy have been performed in MEC. Haddad et al¹⁵¹  identified one durable partial response in a case of MEC with 3⁺ HER2 immunoreactivity of 3 patients treated. Expression of AR in MEC is thought to be exceptionally rare,^88,127  and the presence of diffuse AR tumor expression may require consideration for diagnostic reclassification as SDC.¹²⁷  The lack of AR dependence precludes ADT use in MEC. Estrogen receptor and PR expression has little role in MEC tumorigenesis, diagnosis, or therapeutic management.^88,152  Although MEC is the most common type of salivary gland carcinoma, in comparison to AdCC and SDC, few studies have addressed the role of systemic and/or targeted therapy in this tumor type (summarized in Table 5). Most MECs are low to intermediate grade, and patients respond well to surgical treatment or surgery with postoperative radiotherapy.¹⁵³  Continuing focus for targeted therapy in MEC should focus on high-grade cases with an important emphasis on accurate diagnosis to prevent inclusion of other high-grade carcinoma types that can enter the differential diagnosis with high-grade MEC.

Immunohistochemistry and FISH are becoming more useful in the diagnosis of salivary gland tumors. The use of IHC and/or FISH testing can be helpful in the accurate diagnosis of salivary gland tumors in both large resections and, more important, in small biopsy specimens, when combined with clinical information and morphology. These tests are even more useful when applied as part of a panel approach that incorporates tumor morphology to limit the differential diagnosis. Salivary gland tumors also occasionally show targetable molecular alterations, and combined with more-accurate diagnosis, this may allow for more-tailored treatment for the patient and the particular tumor. Specific tumor types show different frequencies of targetable pathways; therefore, ancillary testing should be tailored to the specific diagnosis. Continued clinical work will likely result in the discovery of more characteristic chromosomal rearrangements, useful antibodies, and targetable mutations in these rare tumors.