Definition, classification and overview of adenoid cystic carcinomas

The term cylindroma, the pathological entity later called ‘adenoid cystic carcinoma’, was coined by Billroth10 to describe a salivary gland tumour composed of entwined cylinders of hyaline stroma and epithelial cells (indeed the illustrations from Billroth's study10 depict the typical architectural pattern of cylindromas/AdCCs, and the cribriform structures with pseudoglands found in AdCCs. The term adenoid cystic carcinoma (carcinoma adenoides cysticum) of the salivary glands was first used by Ewing (1919)11 and applied by Geschickter in 1945 to tumours of the breast.12 Since then, AdCCs have also been described in several other organs such as lacrimal glands, auditory canal, upper respiratory tract and lung, digestive tract, skin, prostate and lower female genital tract.1–7

AdCC belongs to the subgroup of tumours of the myoepithelial lineage and is defined as a tumour where both epithelial (luminal) and myoepithelial (basaloid) cells are neoplastic (figure 1). Multiple architectural patterns have been reported (eg, cribriform, tubular, trabecular and solid) in both mammary and salivary glands, and generally a mixture of different growth patterns are found in AdCCs.1 14 The cribriform growth pattern is the most characteristic and features variably sized and usually smoothly contoured islands of neoplastic cells arranged to compose pseudolumens (which are not true glandular lumens but represent stromal invaginations) and true glandular spaces (formed by the epithelial cells) giving rise to a ‘sieve-like’ appearance (figure 1).1 14 These spaces are filled with eosinophilic hyaline material (Periodic acid-Schiff (PAS) positive, diastase resistant), and/or lightly basophilic myxoid substance (Alcian blue positive): these materials have been demonstrated to represent duplicated basal lamina and glycosaminoglycans by ultrastructural studies (see Cheuk and Chan1 and Bennett et al15 and references therein). Within the cribriform islands, there are occasional true narrow glands lined by cuboidal cells with eosinophilic cystoplasm1 (see figure 1). Histological variants are the glandular (or tubular), reticular (or trabecular) and solid growth patterns (figure 1). The tubular variant is characterised by glandular spaces of elongated tubules lined by epithelial cells and surrounded by single or multiple layers of basaloid cells; the glandular lumens are either empty or contain secretion. In the trabecular variant cells are arranged to form small nests,14 whereas the solid variant is composed of island and sheets of closely packed basaloid cells, with very few or even no pseudocystic spaces; few true glandular spaces can be found.1

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Representative micrographs of the different variants of adenoid cystic carcinoma (AdCC). Breast AdCC showing a mixture of cribriform (A), tubular (B), trabecular (C) and solid (D) growth patterns. Examples of grade 1 (I), grade 2 (II) and grade 3 (III) AdCCs, according to the grading system proposed by Ro et al.13

The two distinct cells types of AdCCs, basaloid and luminal, are best appreciated by immunohistochemical staining (table 1). The basaloid cells express basal cytokeratins (Cks) such as Ck14 and Ck17, vimentin, S-100 protein, actin, calponin and p63. In addition, strong nuclear and cytoplasmic immunoreactivity of maspin, a mammary inhibitory serine protease, has been described in the myoepithelial component of a series of AdCCs of the breast.16 The epithelial cells show strong positivity for luminal cytokeratins such as Ck7, for CEA, EMA and CD117 (c-KIT).1 8 17

The stromal hyaline material is best highlighted by staining for collagen IV and laminin. In addition, PAS and Alcian blue staining may help differentiate the material present in true glandular lumens and pseudolumen, respectively (see above).1 8

Adenoid cystic carcinomas of the salivary glands

Clinical and morphological features

AdCCs are part of the malignant epithelial tumours of the salivary glands and have been recognised as a specific variant of adenocarcinoma of the salivary and mucous glands since 1853.18 19 AdCC affects patients preferentially in the fourth to sixth decade of life, with a slight female predominance (about 3:2).1 17 The parotid gland, submandibular gland and the palate are most commonly involved,1 17 however minor salivary glands in the oesophagus can also be affected.20

AdCCs account for about 10–15% of all parotid malignancies,21 usually present as a slow-growing swelling, and, because of the propensity for perineural invasion, pain may be present (table 2).1 24 In advanced cases, fixation to skin or deeper tissues can occur.1 Owing to its growth pattern, it should be also noted that at time of presentation these tumours have often invaded beyond the clinically apparent borders.25–33 AdCC of the salivary gland is a slow-growing but aggressive cancer which is reflected by the good short-term but very poor long-term outcome of patients with this disease.1 34 The 5-year survival is about 60–75%, while the 10-year survival drops to 30–54%; most patients eventually die of disease after multiple local recurrences and development of distant metastases (distant metastases are more common than regional lymph-node involvement).32 35 36

Histologically, AdCC of the salivary glands presents variable combinations of the three main growth patterns (ie, cribriform, tubular and solid) in each individual case. The cribriform is the most characteristic and is almost invariably found, at least focally.1 Cytologically, the basaloid cells constitute the major cell population, showing mild nuclear pleomorphism and few or no mitoses; in the solid variants these cells usually show a more pleomorphic appearance and mitoses are more commonly found.1

Salivary gland AdCCs are graded using a specific three-tier grading system, originally proposed by Szanto et al18 in 1984, which is solely based on the main type of growth pattern present in the tumour. Grade 1 AdCCs are well differentiated and composed of tubular and cribriform patterns without solid components; grade 2 AdCCs are characterised by a pure cribriform pattern or mixed with less than 30% of solid areas; and grade 3 AdCCs are tumours with marked predominance of the solid pattern.18 This histological grading system has been shown to be associated with prognosis in retrospective studies.1 18 37 38

Recent studies have reported that some low-grade salivary gland carcinomas, including AdCC, can undergo ‘dedifferentiation’.39–41 This histological variant would show two components, a conventional low-grade AdCC and a high-grade ‘dedifferentiated’ carcinoma, which can be either undifferentiated carcinoma or poorly differentiated adenocarcinoma. Dedifferentiation is a well recognised phenomenon in bone and soft tissue tumour pathology42 and is generally associated with a poor prognosis. In AdCCs it is an extremely rare event (11 cases reported to date), whose clinical behaviour and molecular characteristics are still not entirely understood,43 even though recent data seem to support a more aggressive clinical behaviour of dedifferentiated AdCCs.43

Molecular features

The defining molecular feature of AdCCs of the salivary glands and breast appears to be the presence of a recurrent chromosomal translocation t(6;9)(q22–23;p23–24), which generates a fusion transcript involving the genes MYB and NFIB.9 This translocation is described in detail below.

AdCCs of the salivary glands have a stable genome.44 TP53 mutations45 or p53 nuclear expression have been reported,46–48 but in small studies and with discrepant results (see table 3).45 46 The presence of aneuploidy appears to be less common in the cribriform variant (16%) than in the solid lesions (67%, table 3).53 Interestingly, a TP53 point mutation and p53 expression have been demonstrated in the dedifferentiated component of two dedifferentiated AdCCs, together with HER2 overexpression and a high proliferation index.43

Loss of heterozygosis (LOH), comparative genomic hybridisation (CGH) and microarray-based CGH studies show conflicting results. Cytogenetic analysis of AdCCs of salivary glands showed deletions of 6q and 12q as well as rearrangements involving chromosomes 6q and 9p.65 Recently, high-resolution CGH analysis showed that bronchial and salivary AdCCs harbour low levels of genetic instability with few copy number alterations or high-level amplifications. Recurrent gains included 7p15.2, 17q21–25 and 22q11–13, and recurrent losses included 1p35, 6q22–25, 8q12–13, 9p21, 12q12–13 and 17p11–13.61 In addition, Rao et al have recently shown that deletion of 1p32–p36 holds prognostic significance in these lesions.64 Given the rather pervasive presence of recurrent regions of deletions, some have studied the possible involvement of tumour suppressor genes: the minimal region of deletions on chromosome 9p contains CDKN2A and CDKN2B; other possible candidate tumour suppressor genes map to 6q75 and 12q.76

Gene amplifications have been investigated, however studies have yielded conflicting results. The majority of studies where gene amplifications were investigated using fluorescence in situ hybridisation (FISH) or CGH have reported a low prevalence of gene amplifications. The only report that identified frequent amplification of ERBB1, CCND1 and PIK3CA (67%, 46% and 38%, respectively) in AdCCs relied on a PCR-based method,62 which is known to be prone to artefacts when formalin fixed paraffin embedded (FFPE) samples are analysed. Bernheim et al61 and Greer et al63 have recently identified amplifications of MDM2 (12q15), CCND1 (11q13.3) and CTTN (11q13.3) using FISH (table 3); it should be noted, however, that these amplifications affected less than 5% of cases. On the other hand, CCND1 overexpression seems to be more pervasive than CCND1 amplification, as it is found in up to 90% of cases.63

Apart from the recurrent chromosomal translocation described above, additional translocations have been found in single cases of AdCCs (see table 3)55 57 59; however neither the genes involved in these translocations nor their recurrence frequency have been determined.

Two studies have analysed the gene expression profiles of AdCC in salivary glands as compared to normal salivary gland tissues and an AdCC cell line (ACC3): unsupervised hierarchical cluster analysis showed AdCCs separated in three different groups, and interestingly, these groups were not correlated with histological grade.67 Overexpression of several genes encoding transcription factors was demonstrated,67 and among them, SOX4 was found to be up-regulated in both studies.66 67 Patel et al showed that genes associated with morphogenesis, neurogenesis, proliferation, apoptosis, a group of genes encoding extracellular matrix proteins and basement membrane components seem to characterise AdCCs. In addition, Frierson et al67 found epsilon and frizzled-7, both members of the Wnt/β-catenin signalling pathway, to be up-regulated in AdCCs compared to 175 other carcinomas from 10 anatomical sites.

Adenoid cystic carcinomas of the breast

Clinical, morphological and immunohistochemical features

After Geschickter's description of ‘adenocystic basal cell carcinoma’ of the breast,8 three adenocystic basal cell carcinomas were reported by Foote and Stewart in 1946,3 but it was only two decades afterwards that Galloway77 at the Mayo clinic described the first series of mammary adenoid cystic carcinomas.8 Since then several groups have described AdCCs of the breast; however due to the relative rarity of this tumour type, most studies are in the form of case reports or small cohorts.

The existence of AdCC in the breast has been questioned, given its morphological similarity with cribriform carcinoma22 78; however, immunohistochemical and ultrastructural studies have eventually proven the existence of true AdCC in the breast (see below), which is now recognised as one of 17 histological special types of breast cancer by the World Health Organization.15 21–23 79

AdCCs account for 0.1–1% of all breast cancers8 15 23; they are usually diagnosed in adult female patients as unilateral painful masses.8 15 23 This associated pain has been suggested to be due to the contractile myoepithelial component of these tumours, as perineural invasion is not commonly seen in these lesions (table 2).8 15 All quadrants seem to be affected, with a particular trend for the peri-areolar region.8 The most striking feature of AdCCs of the breast, which is in stark contrast with AdCCs of salivary glands, is the excellent long-term prognosis (table 2): a 90–100% 10-year survival rate is reported, and lymph-node metastases are rare, as well as distant metastases, which affect mainly visceral organs.15 54

AdCC can be found in conjunction with other breast lesions, including ipsilateral and controlateral in situ and invasive carcinomas.15 In addition, recent studies have noted an association between AdCC and microglandular adenosis (MGA),80–83 and some authors have suggested that AdCC may develop in a background of and in continuity with MGA. Following this hypothesis there would be a spectrum of lesions with a trend of progression, encompassing MGA, ‘atypical’ MGA (also known as ‘in situ AdCC’) and invasive AdCC.80 An association between AdCC and tubular adenosis (TA) has also been reported in one study, but the molecular analysis performed on these two lesions failed to provide evidence of molecular evolution from TA to AdCC.14

The variety of growth patterns described for AdCCs of the salivary glands is also found when this tumour arises in the breast: the classic cribriform pattern, as well as the glandular, trabecular and solid variants.8 A sebaceous differentiation is found in up to 14% of cases, and foci of adenosquamous differentiation may also be encountered.8 84

In a way akin to salivary gland AdCC, breast AdCCs are graded according to the proportion of solid growth: cases with either cribriform or glandular pattern are considered low grade/G1 tumours, cases with <30% of solid elements are labelled as G2, whereas cases showing >30% of solid growth are classified as high grade/G3 tumours (figure 1).8 13 Ro et al13 reported this grading system to be clinically meaningful, as in their series, tumours with solid components (G2 and G3 cases) were more likely to develop recurrences; in addition, in the same cohort, the only patient who experienced metastases was affected by a high-grade tumour. It should be noted however that histological grade defined by this system was not associated with outcome in two other studies.85 86

Other authors72 have suggested the proliferative indices to be of some relevance as they showed greater values in high-grade when compared with low-grade lesions; however proliferative activity was not found to be significantly related to prognosis.85

Phenotypically, AdCCs are described as hormone receptor negative carcinomas (table 4). In the series of 18 cases analysed by Azoulay et al,73 neither oestrogen receptor (ER) nor progesterone receptor (PR) expression was identified. Similar findings have been reported in other independent series (table 4).49 50 74 One study72 reported that 15% and 10% of cases were positive for ER and PR, respectively (table 4). In contrast, Arpino et al54 demonstrated the presence of ER and PR expression in up to 46% (13/28) and 36% (10/28) of cases, respectively (table 4). Although this cohort represents one of the largest series of AdCCs reported to date (n=28), it should be noted that the cases were collected from different institutions and did not undergo a central review; hence, it cannot be formally ruled out that a substantial number of cases included in that study were cribriform carcinomas, which are usually ER and PR positive. In addition, it should be noted that dextran-coated charcoal assay,87 instead of immunohistochemistry, was used to assess positivity for ER and PR and only 85% agreement between the two techniques is reported.88 The use of the dextran-coated charcoal assay for ER and PR assessment of AdCCs is particularly problematic, given that normal breast lobules and ducts are often entrapped within the bulk of AdCCs, which may lead to false positive results.

Molecular features

In a way akin to salivary gland AdCCs, breast AdCCs consistently display the recurrent chromosomal translocation t(6;9)(q22–23;p23–24), which generates fusion transcripts involving the genes MYB and NFIB.9 This translocation is described in detail below.

Although microarray-based gene expression profiling has been extensively applied to the study of breast cancer, most of these analyses ignored the histological special types of breast cancer (reviewed in Weigelt et al89 and Weigelt and Reis-Filho90), and there is paucity on the transcriptomic features of AdCCs. The molecular subtypes of breast cancer (ie, basal-like, HER2, luminal A, luminal B and normal breast-like), which provide a widely used working model for a breast cancer molecular taxonomy, have been identified by microarray analysis of only the most common types, invasive ductal and invasive lobular carcinomas.91–93 This molecular classification was shown to be of prognostic significance, with tumours of luminal A subtype being associated with the best outcome, and tumours of basal-like or HER2 subtype with the worst outcome.91 94

To date, only one study has formally investigated the transcriptome of breast AdCCs. Using microarray-based gene expression profiling, Weigelt et al49 analysed a series of 113 tumours from 11 special histological types of breast cancer, including four AdCCs. Unsupervised hierarchical cluster analysis showed that AdCCs clustered together with metaplastic and medullary carcinomas, whose similar gene expression patterns were also reflected at the immunohistochemical level (triple negative phenotype; ie, lack of ER, PR and HER2; low levels of Ck19, AR and Ck8/18; high levels of CD117, vimentin, S100, Ck14 and Ck5/6 expression). In addition, molecular subtype analysis using a single sample predictor (ie, centroids) showed that two AdCCs were of basal-like and two AdCCs of normal breast-like phenotype, a molecular subtype which is currently considered to be an artefact of sample representation (ie, high content of normal tissue contamination).95 Indeed, molecular subtype assignment by hierarchical clustering showed that AdCCs consistently displayed a basal-like phenotype, as did medullary and metaplastic carcinomas. These results support earlier immunohistochemical observations by Azoulay et al,73 that breast AdCCs lack expression of ER, PR and HER2 (ie, triple negative phenotype) and express basal cytokeratins (Ck5/6). These observations illustrate the heterogeneity of basal-like breast cancers and emphasise that triple negative and basal-like breast cancer is not a single entity, but rather a spectrum of lesions. Although the majority of triple negative and basal-like breast cancers are high grade cancers (ie, medullary carcinomas, grade 3 invasive ductal carcinomas of no special type (IDC-NSTs), metaplastic carcinomas and apocrine carcinomas49 73 96–99), there is a subgroup of low grade tumours with indolent clinical behaviour that also display a triple negative and basal-like phenotype (ie, AdCCs and secretory carcinomas). Contrary to high grade triple negative and basal-like breast cancers,100 which have been shown to harbour TP53 mutations in >80% of cases,101 two studies have reported that p53 nuclear expression is either absent49 or seen in 25% of AdCCs102; on the other hand, one study reported p53 expression in 6/6 AdCCs analysed.50 Furthermore, AdCCs do not appear to be more prevalent in BRCA1 germ-line mutation carriers.

There is not only a paucity of data on transcriptomic but also on the genomic features of AdCC of the breast. Cytogenetic analyses of a few cases showed 46,XX,t(4;4)(q21;q35),t(5;11)(q13;q21), 46,XX,+1,der(1;16)(q10;p10)103 and 46,XX,inv(9).65 Aneuploidy is reported in less than 10% of cases.54

In a recent case report of a 76-year-old woman affected by a breast AdCC with a kidney metastasis, molecular genetic analysis of the primary and metastatic tumour revealed both a PTEN and a PIK3CA mutation.52 These findings led the authors to speculate that PTEN and PIK3CA mutations may be responsible for the unusually aggressive course of this particular AdCC and suggested identified novel molecular targets for therapeutic intervention. Further studies on larger cohorts are needed to confirm the prevalence of these mutations and their biological role in AdCC.

The limited molecular features of breast AdCC available suggest that these tumours have a relatively stable genome and are characterised by a recurrent chromosomal translocation involving MYB and NFIB. Although AdCCs are of triple negative and basal-like phenotype, a diagnosis of breast AdCC should not necessarily prompt the use of systemic chemotherapy, given the excellent prognosis of patients with this tumour type.

Molecular comparative pathology and therapeutic insights for AdCC: facts and artefacts

c-KIT: therapeutic target or unjustified hope?

The gene KIT maps to 4q11–q12 and encodes c-KIT (CD117), a type III transmembrane tyrosine kinase receptor, which is activated by mutations in gastrointestinal stromal tumours and melanomas. Tumours harbouring KIT mutation but not overexpression of wild-type c-KIT can be targeted by the small molecule inhibitor imatinib mesylate (Gleevec, Novartis) and sorafenib (Nexavar, Bayer).

c-KIT has been shown to be expressed in up to 100% of AdCCs of the breast49 51 72–74 and in 80–100% of AdCCs of the salivary glands arising from the head and neck.68–70 These observations have prompted several groups to investigate whether imatinib would be a valid therapeutic option for those lesions arising in salivary glands. Bold claims were made in 2005 in a study by Faivre et al,104 showing a remarkable response to imatinib treatment in a patient affected by AdCC of salivary glands. These data could, however, never be confirmed by subsequent clinical studies which failed to identify objective clinical responses in patients with AdCC treated with imatinib.69 105 106 In contrast with Faivre et al,104 there are descriptions of progression of metastatic AdCC under treatment with imatinib (table 5).105 These disappointing results should not come as a surprise, given that activating KIT mutations, the determinant for response to imatinib treatment,109 have been repeatedly reported to be vanishingly rare in these cancers.69 70 104 105 Last year, however, Vila et al51 demonstrated the presence of multiple KIT mutations in AdCCs and even multiple mutations in single cases. Owing to the small number of tumours analysed (extraction of DNA was successful in only 8/14 cases) and, most importantly, to the type of sample subjected to mutation analysis (ie, FFPE samples), these findings needs to be confirmed in larger cohorts using optimally processed samples (ie, fresh/frozen samples).

Taken together, the balance of evidence available to date suggest that c-KIT may not constitute a potential novel therapeutic target for AdCC.

A translocation in common

Persson et al9 have recently shown that breast, salivary, lachrymal and ceruminal gland AdCCs harbour a recurrent specific translocation t(6;9)(q22–23;p23–24). This genetic alteration had already been described more than 10 years ago, by the same group, as characteristic of salivary glands. However, the fusion gene partners in this translocation have only now been identified: the oncogene MYB on chromosome 6q22–q23 and the transcription factor NFIB on chromosome 9p23–p24.9 MYB maps to chromosome 6q22–q23 and encodes a transcription factor with an N-terminal DNA binding domain, a centrally located transcription activation domain, and a C-terminal negative regulatory domain, which plays a pivotal role in the control of cell proliferation, apoptosis and differentiation. It is highly expressed in immature, proliferating cells, fetal salivary gland; however it is down-regulated as cells become more differentiated.9 In the t(6;9)(q22–q23;p23–p24), the exon 14 of MYB is fused to the last coding exons of NFIB, most often due to breakpoints in MYB intron 14 and in NFIB intron 8. This translocation was shown to produce distinct types of fusion transcripts due to splice variations of the MYB gene. The common denominator of these genetic aberrations, however, was a deletion of exon 15 of MYB and its 3′-UTR, which contains several highly conserved target sites for miR-15a/16 and miR-150 microRNAs. In normal and cancer cells, these miRNAs have been shown to down-regulate the expression of MYB. Consistent with this hypothesis, transfection of leukaemic cells with premiR-15a/16 and premiR-150 resulted in a significant down-regulation of MYB mRNA expression, whereas transfection of ACC cells did not alter the expression levels of MYB significantly.9

Despite the limited number of samples analysed so far,9 this is the first fusion transcript identified in AdCCs. Although the use of split apart probes for the identification of this translocation are unlikely to be used in diagnostic practice, due to the characteristic histological features of AdCC, one can envisage that these probes may help reclassify some tumours with features overlapping with those of AdCC or mixed salivary gland tumours. This fusion gene may provide new therapeutic avenues for the management of advanced AdCC; however, one must remember that targeting transcription factors is by no means a trivial task. Further functional studies investigating in greater depth the biological consequences of the MYB gain of function due to the MYB-NFIB fusion are eagerly awaited. Importantly, gene silencing experiments to demonstrate that MYB expression is selectively required for the survival of cancer cells with genetically activated MYB have yet to be performed.