|Home | About | Journals | Submit | Contact Us | Français|
Adrenocortical carcinoma (ACC) is a rare, heterogeneous malignancy with a poor prognosis. According to WHO classification 2004, ACC variants include oncocytic ACCs, myxoid ACCs and ACCs with sarcomatous areas. Herein, we provide a comprehensive review of these rare subtypes of adrenocortical malignancy and emphasize their clinicopathological features with the aim of elucidating aspects of diagnostic categorization, differential diagnostics and biological behavior. The issue of current terminology, applied to biphasic tumors with pleomorphic, sarcomatous or sarcomatoid elements arising in adrenal cortex, is also discussed. We additionally present emerging evidence concerning the adrenal cortical tumorigenesis and the putative adenoma–carcinoma sequence as well.
Adrenocortical carcinoma (ACC) is a rare heterogeneous malignancy with an incidence of 0.7–2.0 cases per million population per year . Although the diagnosis of malignancy is easy in most cases, mainly due to advanced stage at presentation, strictly intra-adrenal tumors must be evaluated for malignant potential . In this regard, several multiparametric scoring systems, including the Hough scoring system, the Weiss scoring system, the Van Slooten scoring system and the Weiss revisited index [3, 4], and, also recently, diagnostic algorithms, such as the stepwise discriminate diagnostic system and the simplified diagnostic algorithm [5, 6], have been generated. The latter takes into account the presence of a disrupted reticulin framework, which constitutes the first step of this diagnostic approach, as being the single most sensitive feature of malignancy. Other parameters include mitotic count >5/50 HPF, presence of necrosis and venous invasion .
Currently, the Weiss scoring system is the most popular among multiparametric scoring systems, owing to its reliability and relative simplicity (Table 1) [1, 7]. Nevertheless, the Weiss system suffers particular limitations including (1) lack of reproducibility of several criteria, (2) “borderline” tumors with a Weiss score of 2 or 3, (3–4) oncocytic and myxoid variants of adrenocortical tumors and (5) pediatric adrenocortical neoplasms [2, 7]. The latter are frequently characterized by a limited malignant potential especially in children<5 years of age, despite the presence of impressively atypical histologic features (Fig. 1), which could be attributed either to a low stage upon presentation or to a putative origin from a cell of the fetal adrenal cortex .
The objectives of this review are (1) to present emerging evidence concerning the adrenocortical tumorigenesis in order to address the issue of an adenoma–carcinoma sequential progression as well as to further elucidate the incompletely understood pathophysiology of this aggressive neoplasia and (2) to describe in detail the rare variants of ACC to increase awareness of their clinicopathological features and the potential diagnostic challenges.
A hotly debated issue is whether or not adenomas progress stepwise to carcinoma. Given (1) the differences between benign and malignant tumors in terms of transcriptome and genetic alterations, (2) the epidemiology of adrenal incidentaloma and ACC and (3) the low incidence of progression (5–25%) of clinically inapparent adrenal masses to hyperfunction or an increase in size by at least 1 cm coupled with (4) the lack of clear evolution from benign, non-secreting adenoma to ACC in untreated cases [9, 10], it seems that multistep progression from benign to malignant is unlikely for adrenal cortical neoplasms.
However, not all evidence points in that direction. Fewer genetic aberrations were found in adenomas than in carcinomas. Most (72–86%) distinct alterations seen in the adenomas were also present in carcinomas [11, 12]. Furthermore, the positive correlation between the number of alterations detected by comparative genomic hybridization (CGH) and tumor size [11, 13] supports the notion of an adenoma–carcinoma sequence. Additional evidence in favor of such a sequence derives from X-chromosome inactivation pattern analysis which shows monoclonal ACCs along with ACAs of polyclonal or monoclonal origin. Allelotyping with microsatellites and a few reported cases provide support for an adrenal adenoma–carcinoma sequence [14–20].
With regard to the latter, Bernard et al.  reported on an intriguing case of a localized adrenal neoplasm composed of two different components exhibiting different pathological features: a central zone of a Weiss score (5) and a peripheral zone of a Weiss score (0). The former zone displayed complete LOH at the TP53 (17p13.3) gene locus, uniparental disomy at the 11p15 locus, overexpression of the IGF-II gene and numerous CGH changes, with gains at chromosomes 4, 5p, 10, 12p, 16p, 19 and 20 and losses at chromosomes 2 and 11. The outer zone did not show any of these genetic abnormalities. A similar case was presented by Trezzi et al.  as a dedifferentiated adrenal cortical neoplasm comprising a sharply outlined, hemorrhagic, central nodule (ACC) and an outer ring zone (ACA), while Gaujoux et al.  included one heterogeneous ACT in a small series consisting of an ACC component (Weiss score 6) developed within an ACA component (Weiss score 1). The malignant component exhibited diffuse cytoplasmic and focal nuclear β-catenin accumulation and harbored a β-catenin gene (CTNNB1) mutation, whereas the benign component showed only focal cytoplasmic staining and no such mutation. No TP53 somatic mutation was found in the ACC and ACA components. Similarly, Schmitt et al.  presented an ACA case showing frank heterogeneity both by means of H&E staining (atypia, hyperchromasia, nucleoli and atypical mitoses) and immunohistochemistry, including perinuclear dot-like immunoreactivity for IGF-II, nuclear staining for CDK4 and Ki67 labeling index > 5%. These particular areas within the adenoma were interpreted as small foci of malignant transformation, in accordance with a van Slooten index > 8.
Whether the fact that different scoring systems yield different diagnoses in individual ACTs [20, 21] points to a multistep process remains to be clarified. It is tempting to speculate that in a process of clonal evolution, one dominant clone may overtake neighboring subclones to an extent that the ACA component is no longer visible.
Recent evidence has suggested multistep tumorigenesis within the group of ACCs . In particular, transcriptome analysis has established an ACC subclassification into two distinct subgroups with different outcomes [22–24], consistent with the notion that ACC survival and outcome vary greatly mainly due to different molecular defects modulating the tumor phenotype and, to a lesser extent, to tumor stage at diagnosis . However, ACCs in the poor prognosis group displayed more advanced disease, subsequently confounding the issue of whether these subgroups correspond to distinct stages of the same disease or to distinct ACC types. Statistical independence of transcriptome-based prognostic information from stage seems to argue in favor of the latter .
In an effort to further elucidate the molecular genetics of the poor-outcome ACC group, Ragazzon et al.  identified three subgroups by using unsupervised clustering analysis: (1) p53 group, encompassing all tumors with a TP53 mutation, (2) β-catenin group, containing all tumors with beta-catenin nuclear staining, and (3) x group, with neither p53 nor β-catenin altered pathways, but enriched in cell cycle and metabolism genes. Moreover, IGF-II overexpression has been shown in approximately 90% of ACCs, with paternal unidisomy and loss of imprinting being the underlying mechanisms accounting for this markedly elevated expression in familial ACC cases associated with Beckwith–Wiedemann Syndrome, while additional mechanisms of transcriptional regulation are likely to be involved in sporadic ACCs [9, 23]. In this context, Ragazzon et al.  did refer to widespread IGF-II overexpression among both subgroups, highlighting its importance in adrenocortical tumorigenesis, but concomitantly suggesting a role for additional genomic alterations promoting the development of the most aggressive tumors [25, 27]. Prior to IGF-II overexpression, it appears that ACC loses nephroblastoma overexpressed (nov) expression , in line with the finding of a significantly reduced novH expresssion correlated with the acquisition of a malignant tumoral phenotype .
Several lines of evidence suggest that the activation of the Wnt/β-catenin signaling pathway is important in adrenocortical oncogenesis [19, 25, 30–37]. In fact, it has been demonstrated that this activation, as assessed by abnormal nuclear and/or cytoplasmic β-catenin accumulation with or without somatic activating CTNNB1 mutations, is involved both in benign and malignant adrenocortical tumors [19, 31–33]. It is of particular interest to note that CTNNB1 mutations (1) were detected only in macronodules of primary pigmented nodular adrenocortical disease cases, suggesting that these are secondary genetic events contributing to the nodular development of the disease and potentially implying a more aggressive phenotype [19, 34], (2) were related to a specific phenotype of larger and non-secreting ACAs, indicating a less differentiated state , and (3) were associated with decreased overall and disease-free survival in ACCs, suggesting a specific effect on tumor biology in terms of progression towards a more aggressive phenotype within the group of ACCs [25, 35]. Altogether, these observations, along with data stemming from an experimental mouse model , do not completely clarify whether the activation of Wnt/beta-catenin signaling pathway is an early or a late event in the development of adrenal neoplasms.
Oncocytic adrenocortical neoplasms (OAN) represent a rare group of neoplasms with approximately 115 cases reported in the literature [38–44]. Kakimoto et al.  first described this particular histological variant in 1986, which was followed by numerous reports either in the form of single cases or of small series, with the largest one comprising 13 cases .
Accurate classification of OAN is important. Given that OANs display (1) cells with eosinophilic cytoplasm, clear cell population as a rule consisting of less than 25% of the tumor volume, (2) high-grade nuclear atypia in at least a subgroup of neoplastic cells and (3) almost always diffuse architecture and since these inherent histological parameters do not confer a poor prognosis , applying the Weiss system to this subset of adenocortical neoplasms will lead to a diagnosis of ACC, which is not appropriate given the benign behavior of most of these lesions.
Bisceglia and colleagues  developed criteria for the classification of oncocytic adrenocortical tumors (Table 2). These were based on parameters of the Weiss system and constituted the Lin–Weiss–Bisceglia (LWB) system: (1) major criteria (a mitotic rate of more than 5 mitoses per 50 high-power fields, any atypical mitoses or venous invasion), (2) minor criteria [large size (>10 cm and/or > 200 gr), necrosis, capsular invasion or sinusoidal invasion] and (3) definitional criteria (predominantly cells with eosinophilic-granular cytoplasm, high nuclear grade and diffuse architectural pattern). The latter are not used as they are present in all subgroups of pure oncocytic neoplasms and are not determinants of biological behavior. According to the proposed working rules, the presence of any one of the major criteria indicates malignancy (oncocytic adrenocortical carcinoma), and the presence of one to four minor criteria is indicative of uncertain potential (borderline oncocytic neoplasm of uncertain malignant potential) (Fig. 2), while the absence of all major and minor criteria indicates benign behavior (adrenocortical oncocytoma) . In addition, Bisceglia et al.  proposed categories for oncocytic adrenocortical tumors as follows: (1) pure oncocytic tumor, if a tumor is exclusively or almost entirely composed (greater than 90%) of oncocytic cells, (2) mixed oncocytic tumor, when a clear cell component is also present (ranging from 10% to 50%), and (3) ordinary adrenocortical tumor with focal oncocytic changes, if the oncocytic component is not a predominant one (less than 50% of the tumor mass).
It is critical to extensively sample an oncocytic adrenocortical neoplasm for several reasons. First, one should discriminate a pure oncocytic tumor either from an ordinary adrenocortical tumor with focal oncocytic changes or its conventional counterpart of the compact cell type in order not to incorrectly apply the Weiss system and accordingly inadequately estimate its biological behavior . Second, an oncocytic tumor can only be labeled as pure after quantifying each individual component (oncocytic and clear) since pure and mixed oncocytic tumors do not seem to share similar clinical outcome . In this regard, immunohistochemistry using antimitochondrial antibodies can support the recognition of oncocytic differentiation and help to accurately quantify the oncocytic component. Of note, adrenocortical cells are, by nature, rich in mitochondria and therefore positive, so only strong, diffuse and finely granular staining should be taken as indicative of oncocytic differentiation, as suggested by Wong et al.  with respect to mES-13 staining. Moreover, the differential diagnostic spectrum of OANs also includes the oncocytic variant of pheochromocytoma, the granular cell variant of clear cell renal cell carcinoma, the eosinophilic variant of chromophobe renal cell carcinoma and the hepatocellular carcinoma [45, 46].
In an effort to further delineate and elucidate this subtype of adrenocortical neoplasia, Wong et al.  provided a comprehensive review and retrospectively applied the LWB criteria to all reported cases with sufficient information, along with 13 previously unpublished cases. Of note, they validated the effectiveness of the LWB criteria in predicting malignant potential and provided preliminary evidence supporting the concept of a more favorable prognosis for oncocytic adrenocortical carcinomas in comparison with their conventional counterparts. Moreover, they found approximately 30% of OANs to be functional while emphasizing the frequency of small oncocytes in several cases of their series , accordingly including in their view of the oncocyte not only cells with abundant granular eosinophilic cytoplasm.
In keeping with the current notion that myxoid extracellular matrix is a histological feature that can be found in both physiological and pathological conditions, non-neoplastic (e.g. myxedema) as well as neoplastic (benign/malignant epithelial or mesenchymal tumors) , myxoid change has been recognized rarely also in neoplasms of adrenal cortical origin. The myxoid substance has been shown to be Alcian-blue positive and negative or focally weak positive for PAS stain and mucicarmine. In particular, there have been thus far 46 reports of myxoid adrenocortical neoplasms (MANs) [48–65]: 17 myxoid adrenocortical adenomas [49–58], 2 borderline MANs [48, 50] and 27 myxoid adrenocortical carcinomas [48, 49, 59–65]. These cases were either published as isolated case reports or were presented in small series, with the largest ones each consisting of 14 cases [48, 49].
The reported age range of MANs (46 cases) is 16 to 82 years, with a mean and a median age of 51 and 51.5 years, respectively. In general, cases in females slightly outnumber those in males (26:20). Hormone hypersecretion occurs frequently in these tumors; only 13 out of 42 cases were non-functioning (eight adenomas, one borderline and four carcinomas), whereas five patients had biochemical evidence of hormone production (two adenomas, one borderline and two carcinomas). The remaining 24 cases with an available endocrine evaluation presented with clinical evidence of hormone overproduction: 16 with Cushing-related symptoms, 4 with Conn syndrome, 2 with symptoms due to cortisol and androgens, 1 with gynecomastia and breast pain and 1 with virilization (4 adenomas and 20 carcinomas). Although it had been suggested that the functional status could be associated with the presence of myxoid material, non-functioning tumors displaying a higher percentage of myxoid areas and being lipid depleted , no definite conclusions can be reached on morphological grounds. In fact, eight cases showing a myxoid area > 70 % were hormonally active [48–50, 62, 63], whereas another eight cases with a similar myxoid content were non-functional [48, 51, 53–55, 57, 59, 60]. The proportion of myxoid change varied in all reported cases from 5% to approximately 100% of the total area. Interestingly, neurofilament (NF) expression has been recently found to be related more to the extent of myxoid changes and, therefore, suggested as a potential characteristic of myxoid ACCs with a predominant (> 70 %) myxoid component . This, along with the fact that the expression of neuroendocrine/neuronal markers such as NF and CD56 was mostly restricted to the myxoid areas of conventional ACCs with focal myxoid change, seems to imply a peculiar association between NF expression and myxoid change .
An important point is the distinction of myxoid borderline/malignant ACTs into two subgroups as proposed by Papotti et al. : myxoid ACCs and conventional ACCs with focal myxoid changes. This is based on a different type of myxoid change, growth pattern, cell size, nuclear atypia and cell cytoplasm. Myxoid ACCs are usually characterized by (1) extensive myxoid areas separated from conventional ACC areas, if present, (2) trabecular/microacinar growth pattern, (3) small, uniform cell size, (4) mild nuclear atypia and (5) scant, eosinophilic cell cytoplasm, whereas conventional ACCs with focal myxoid changes are characterized by (1) focal myxoid areas (< 20% of the total area) always merging with conventional ACC areas, (2) solid/diffuse growth pattern, (3) large, heterogenous cell size, (4) moderate to high nuclear atypia and (5) abundant, granular eosinophilic cell cytoplasm. Nevertheless, two cases of the former subset showed overlapping features such as solid growth pattern and pleomorphism as well. In addition, the same investigators  suggested that only the former subgroup actually belongs to the distinctive myxoid variant of ACTs, with the latter potentially representing the result of myxoid degenerative changes (Fig. 3). Whether or not these subgroups have a distinct biologic behavior needs to be further elucidated.
In the histopathological assessment of these lesions, it is critical first to exclude potential morphological mimics, including metastatic tumors affecting the adrenal gland (breast carcinomas, salivary or skin adnexal carcinomas of adenoid cystic type and neuroendocrine tumors)  as well as primary retroperitoneal neoplasms with myxoid areas (chordomas, myxomas, lipomas or liposarcomas, leiomyomas or leiomyosarcomas, benign or malignant nerve sheath tumors, extraskeletal myxoid chondrosarcomas, gastrointestinal stromal tumors and myxoid malignant fibrous histiocytomas) [49, 50, 63]. The presence of typical foci of adrenocortical differentiation, along with an immunohistochemical profile similar to that of the conventional ACT, provides valuable aid in excluding these mimics [48, 50, 51]. In addition, accurate assessment of malignant potential in any given tumor of the myxoid variant is important. In this regard, the issue of whether the Weiss scoring system is adequate to reliably predict malignancy in MANs is currently being questioned because:
According to WHO classification 2004, oncocytic ACCs, myxoid ACCs and ACCs with sarcomatous areas are regarded as rare variants of ACC . Whether the latter variant should be designated carcinosarcoma or sarcomatoid carcinoma, as previously suggested by Sturm and colleagues  in agreement with the WHO terminology applied to biphasic tumors with pleomorphic, sarcomatous or sarcomatoid elements arising in other organs , is still unsettled. Nevertheless, nine cases of this unusual variant have been previously reported [67, 69–76], with only four corresponding to true carcinosarcomas [69, 72–74], consistent with the concept that carcinosarcomas are tumors composed of morphologically malignant epithelial and specialized mesenchymal elements, easily recognizable histologically as forms of mesenchymal malignancy resembling osteosarcoma, chondrosarcoma, rhabdomyosarcoma, angiosarcoma or liposarcoma (Fig. 4) . Adrenal carcinosarcomas, like their counterparts in other organ systems, were mainly characterized by a rhabdomyosarcomatous component (three out of four) [69, 73, 74], while only one case was characterized by a mixture of osteosarcoma and chondrosarcoma . The remaining five biphasic cases without heterologous elements did have a spindle cell component [67, 70, 71, 75, 76], with three out of five also showing variably giant cells in the sarcomatous component [67, 70, 75].
Another unique malignant biphasic neoplasm arising as a primary tumor of the adrenal gland is the adrenocortical blastoma. This functional (virilizing) malignant tumor has been reported in an infant, and it consisted of a mixture of immature epithelial and mesenchymal elements, along with slit-like spaces partially lined by primitive epithelial cells, focally recapitulating the morphology of the normal embryologic development of the adrenal cortex . There have been no additional cases reported, and therefore, the question whether this neoplasm is indeed restricted to the pediatric age group  similar to pleuropulmonary blastoma (virtually restricted) but in contrast to pulmonary blastoma  remains unresolved.
Of interest, only two out of four carcinosarcomas were functional, presenting with clinical signs either of hyperaldosteronism or of virilization [72, 73], whereas all adrenocortical sarcomatoid carcinomas were non-functioning [67, 70, 71, 75, 76].
Although the limited number of reported cases does not allow any definite conclusion, adrenocortical carcinosarcomas and sarcomatoid carcinomas have similar clinical outcome (postoperative survival ranging from 2 days to 12 months; median 6 months)  and therefore have a worse prognosis than their conventional counterparts . This dramatically aggressive behavior may be attributed not only to advanced stage at presentation but also in part to inherent biological properties potentially deriving by progression (dedifferentiation) from a pre-existing better differentiated ACC, in analogy to the undifferentiated (anaplastic) carcinoma of the thyroid gland [66, 80]. This endows cells with migratory and invasive properties, epithelial-mesenchymal transition, stemness, prevention of apoptosis and senescence, and it contributes to immunosuppression and notably confers resistance to chemotherapy and immunotherapy .
Although various theories have been proposed with regard to the histogenesis of carcinosarcomas, namely (1) the composition tumor theory (paradoxically requiring a non-malignant non-epithelial component as a reactive proliferative response induced by the epithelial component via paracrine secretion), (2) the collision or biclonal tumor theory (a collision between two synchronous, histogenetically independent, biclonal tumors), (3) the conversion tumor theory (neoplastic transformation within a monoclonal tumor recapitulating the naturally occurring event of conversion of epithelial to mesenchymal cells during embryogenesis)  and (4) the combination or divergent tumor theory (deriving from a common monoclonal stem cell precursor) [82–85], molecular genetic evidence of monoclonality supports the single pluripotential stem-cell-divergence hypothesis and the epithelial-to-mesenchymal transition as well [83–85]. With regard to ACCs with sarcomatous areas, evidence emerging from the immunophenotypical profile of the sarcomatous component (focal retaining of Melan-A, synaptophysin and calretinin positivity)  and the presence of transitional (intermingled) zones between carcinomatous and sarcomatous components [67, 70, 73] seem to imply such a putative transition. Differences of the mitotic count in the distinct elements [7/10 HPF (epithelioid component) versus 30/10 HPF (pleomorphic/spindle cell component)]  and the presence only of the sarcomatous component either in large veins into the periadrenal fat  or in well-established metastases [71, 74] are consistent with a dedifferentiation process.
Aberrant phenotypical differentiation observed in this particular variant includes both neural and melanocytic differentiation (NSE, S100 and HMB-45 immunoreactivity) [70, 71]. Although melanocytic differentiation, suggestive of a divergent evolutionary pathway, has been reported in carcinosarcomas of other anatomic locations [86–89], pathologists should always keep in mind to first exclude the possibilities of a metastatic melanoma, a collision tumor consisting of an ACC and a primary or metastatic melanoma (not yet reported) and a primary perivascular epithelioid cell tumor . Other differential diagnostic considerations include composite pheochromocytoma with a malignant peripheral nerve sheath tumor component (the latter with or without divergent differentiation into mesenchymal elements such as rhabdomyo-, osteo-, chondro-, angio- and/or fibro-sarcomatous components) [91, 92], metastatic carcinoma with sarcomatous or sarcoma-like areas and metastatic neoplasm with sarcomatoid elements other than carcinomas (melanoma or germ cell tumor), as well as primary retroperitoneal sarcomas . Given that adrenal biphasic tumors, in contrast to those arising in epithelial organs, are unlikely to express cytokeratins while focal smooth muscle differentiation is a common finding, thorough sampling of the resected tumor is of critical importance in order to reveal the co-existent carcinomatous component and subsequently prove the adrenal origin on immunohistochemical grounds [67, 69].
Since ACCs with sarcomatous areas are highly aggressive tumors and a sarcomatous or sarcomatoid component seems to be a predictor of shorter survival in ACC, the question arises whether the Weiss scoring system should be applied to this extremely rare ACC variant. In fact, ACCs with sarcomatous areas regroup the majority of Weiss histological criteria . In this setting, Sturm and colleagues  have proposed that a sarcomatous or sarcomatoid component should be well circumscribed and represent at least 10% of the tumor bulk to establish this unusual diagnosis. Nevertheless, the sarcomatous or sarcomatoid component is not included in the Weiss system, further confounding this issue when dealing with a biphasic tumor demonstrating a predominant non-carcinomatous component [67, 69]; in such cases, the criterion of diffuse architecture (greater than 1/3) is being obviously challenged.
In summary, we discussed the issue of multistep tumorigenesis in adrenocortical neoplasia and provided an overview of the ACN variants in order to make histopathologists aware of the clinicopathological spectrum. It is important to correctly characterize their biological potential, which is occasionally not an easy task for myxoid and oncocytic adrenocortical tumors. In this context, applying the Weiss scoring system on the latter would definitely tilt the diagnostic balance towards malignancy in a similar way as pediatric ACNs (pathologically malignant but benign-behaving tumors) , whereas underdiagnosis could potentially be the case with regard to the myxoid ones. In this setting, even the multistep tumorigenic processes are hard to take apart and analyze, while evidence for a distinct biological behavior is either preliminary  or unestablished  due to the small number of reported cases, limited follow-up period, insufficient data and heterogeneity. Whenever pathologists report on these variants singly or as part of larger series, it is recommended to note the diagnostic subcategory, as has been previously suggested (pure oncocytic tumor/mixed oncocytic tumor/ordinary adrenocortical tumor with focal oncocytic changes, myxoid ACCs/conventional ACCs with focal myxoid changes and carcinosarcoma/sarcomatoid carcinoma) [39, 48].
Conflict of interest statement We declare that we have no conflict of interest.
Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Ronald R. de Krijger, Email: r.dekrijger/at/erasmusmc.nl.
Thomas G. Papathomas, Email: t.papathomas/at/erasmusmc.nl.