5th International ACC Symposium: Classification of Adrenocortical Cancers from Pathology to Integrated Genomics: Real Advances or Lost in Translation?

Abstract

For the clinician, despite its rarity, adrenocortical cancer is a heterogeneous tumor both in term of steroid excess and tumor evolution. For patient management, it is crucial to have an accurate vision of this heterogeneity, in order to use a correct tumor classification. Pathology is the best way to classify operated adrenocortical tumors: to recognize their adrenocortical nature and to differentiate benign from malignant tumors. Among malignant tumors pathology also aims at prognosis assessment. Although progress has being made for prognosis assessment, there is still a need for improvement. Recent studies have established the value of Ki67 for adrenocortical cancer (ACC) prognostication, aiming also at standardization to reduce variability. The use of genomics to study adrenocortical tumors gives a very new insight in their pathogenesis and molecular classification. Genomics studies of ACC give now a clear description of the mRNA (transcriptome) and miRNA expression profile, as well as chromosomal and methylation alterations. Exome sequencing also established firmly the list of the main ACC driver genes. Interestingly, genomics study of ACC also revealed subtypes of malignant tumors with different pattern of molecular alterations, associated with different outcome. This leads to a new vision of adrenocortical tumors classification based on molecular analysis. Interestingly, these molecular classifications meet also the results of pathological analysis. This opens new perspectives on the development and use of various molecular tools to classify, along with pathological analysis, ACC, and guides patient management at the area of precision medicine.

Introduction

Adrenocortical cancer (ACC) is a rare aggressive cancer, with an overall poor, but heterogeneous, prognosis [1–3]. Diagnosis of malignancy of adrenocortical tumor needs pathological expertise for localized tumor and can be, in some cases, difficult. Prognostication after complete tumor removal of an ACC is challenging. Local extension or metastasis at initial diagnostic is pejorative prognosis factor. Nevertheless, even in metastatic cases, the outcome can be heterogeneous. In such progressive patients, medical therapy is limited. A better understanding of the molecular alterations of ACC would certainly help to progress in this field. In the last decade, the work of several teams on the pathological diagnosis of adrenocortical tumors and, more recently, at the national and international levels on the genomics of ACC resulted in significant progress to overcome these limitations. This recent progress is discussed in this review.

The Role of Pathology in Adrenocortical Tumor Management

Traditionally, pathologists have tried to provide clinicians with a classifying diagnosis for biopsy material or resected tissue or organ specimens. Adrenocortical tumors largely come in two flavors: benign lesions including hyperplasia and adenomas, and malignant lesions, including carcinomas. While benign lesions bear a large resemblance to the native adrenocortical tissue, showing cells with small dark centrally located nuclei and ample vacuolated and lipidized cytoplasm, carcinomas usually show cytonuclear polymorphism and frequently have eosinophilic cytoplasm. In addition to the “classic” adrenocortical carcinoma (ACC), several variants have now been recognized, including the oncocytic, myxoid, and the sarcomatous variant of ACC. Recognition is not only important with respect to tumor classification but also in view of the use of the distinction between benign and malignant tumors (see below). For classification purposes, several immunohistochemical markers are used to support the adrenocortical origin of lesions, including inhibin, melan-A, calretinin, and steroidogenic factor 1 (SF1).

In addition to classification of lesions, pathologists also try to indicate the expected clinical behavior of tumors. This has been notoriously difficult in endocrine tumors, including adrenocortical tumors. This is in part because many endocrine tumors are rare, and experience is therefore limited. In addition, they do not immediately display similar characteristics as other carcinomas, i.e., invasive growth into the organ itself, such as in breast cancer, or into neighboring structures.

The vast majority of adrenocortical tumors are benign. These lesions are usually small, less than 5 cm in diameter and weigh less than 50 g. Carcinomas, in contrast, are mostly bigger than 5 cm and weigh more than 100 g. In addition, growing lesions and those with specific radiological characteristics, such as Hounsfield units >10, might be suspected to be malignant. While these are good rules of thumb, overlap exists with regard to size and weight between benign and malignant lesions. Therefore, classification systems on the basis of histopathological criteria have been set up. The mostly used of these is the Weiss score, which is based on nine criteria that can be assessed microscopically: diffuse growth pattern, necrosis, capsular penetration, venous invasion, sinusoidal invasion, increased number of mitoses (>5/50 high power fields), atypical mitoses, nuclear pleomorphism, and the presence of <25 % clear cells. Every criterion present scores one point. If it reachs up to two points, tumors are considered benign, and from three points onward, clinically adverse behavior might be expected [4]. A similar classification system was proposed by van Slooten, having slightly different criteria, but essentially leading to a similar distinction in two the groups [5]. Comparison of the two classifications showed that they are equally good in classifying tumors, that in case of doubt, both may be employed, and that they may also be used for prognostic purposes, because patients with a higher tumor score had a worse outcome [6]. More recently, additional scoring systems have become available, largely based on the previous systems and/or with the addition of the Ki67 labeling index [7–9]. These classifications all have similarly high performance, with sensitivity up to 100 % and specificity between 90 and 99 % [9].

It should be noted that the Weiss and van Slooten classification systems have good performance with respect to classic ACC, but are less suitable for the oncocytic and the myxoid variants, in which malignancy may be overestimated and underestimated, respectively. Oncocytic tumors inherently possess three features of the Weiss score: less than 25 % clear cells, diffuse growth, and nuclear polymorphism. Therefore, a modified classification system has been developed for this variant, the Lin-Weiss-Bisceglia system [10]. For the myxoid variant, there is no such separate system, but caution should be exerted when encountering such tumors, firstly, for the correct classification as ACC, using the abovementioned immunohistochemical markers, and secondly, for the potential of underestimating them as benign lesions.

Ki67 labeling is an important addition to the pathologic armamentarium and mainly serves prognostic purposes, for instance in breast cancer, but also through grading of neuroendocrine tumors of the gastrointestinal tract [11, 12]. It has also been used for prognostication and treatment decisions in adrenocortical carcinoma [13, 14]. However, large variation exists in the methods employed, which leads to high inter- and intraobserver variation. Firstly, preanalytical factors, such as tissue fixation, have a large effect. Secondly, the exact antibody and staining method used have a profound effect on the resulting slides. Finally, methods of analysis vary from place to place. In an attempt to overcome this variation, a multiobserver study was performed within the context of the ENSAT cancer network. A total of 61 cases were analyzed by 14 pathologists, showing that digital image analysis was superior to the average of all pathologists, mainly due to different scoring techniques. In addition, there was a clear correlation between Ki67 score and outcome [15].

Another issue is the inherent heterogeneity of ACC, which results in areas with high and areas with low Ki67 labeling. An automated system is ideally equipped to select for areas with the highest proliferation. ACC, like other tumors, are heterogeneous in many respects. This is not only reflected by Ki67 labeling or various histotypes but also by the molecular pattern(s) that can be detected (see below).

Integrated Genomics and the Classification of Adrenocortical Cancer

Advances in sequencing methods and genome-wide genomic studies, particularly with high-throughput sequencing (next-generation sequencing or NGS), have greatly accelerated progress in the study of cancer. The genomics methods allow now to study at the pan-genomic level gene expression profile, and genetic and epigenetic alterations in cancers. It has been a major approach to identify new driver genes in cancer or endocrine tumors, including benign and malignant tumors of the adrenal cortex (see 16 for review).

The power of these pan-genomic methods to classify frequent cancers [17, 18] (breast, lung, melanoma, prostate…), but also endocrine cancer like thyroid cancer [19], has been demonstrated. Interestingly, these molecular classifications could in some cases reveal subgroups of cancers that were not individualized before by conventional pathology. Because of its rarity, the genomic studies of adrenocortical cancer (ACC) have been more recently developed. However, now, more than 25 studies report on the genomics of adrenocortical tumors (see 16 for review). Recently, the integrated genomics of ACC have been developed by two different international efforts: one from Europe (European Network for the Study of Adrenal tumors [ENSAT]) [20] and one from the TCGA ACC Group from America, Europe, and Australia [21]. Both studies were presented in October 2015 at the 5th International Adrenal Cancer Symposium in Michigan. From these studies, subgroups of tumors with different gene expression, miRNome, driver genes, chromosomal, and methylation alterations profiles are individualized. Interestingly, for a translational perspective, these molecular subgroups are associated with different outcome, especially in term of overall survival.

Exome sequencing has established the mutational landscape of ACC in three studies [20–23] that were presented and discussed at the 5th International Adrenal Cancer Symposium in Michigan. This helps to establish now a quite reproducible list of the ACC driver genes. Among these drivers, the E3 ubiquitin ligase ZNRF3 was for the first time reported in cancer. It is the most commonly altered gene in ACC, found inactivated by homologous deletion or mutation in more than 20 % of ACC. Several observations suggest that ZNRF3 inactivation induces an activation of the Wnt/beta-catenin pathway. The Wnt/beta-catenin pathway can also be activated directly in ACC by CTNNB1 activating mutations—encoding the beta-catenin— also commonly found in 15 % of ACC. Interestingly, ZNRF3 and CTNNB1 mutations are mutually exclusive in ACC. Other recurrently mutated genes are related to the cell-cycle regulation. Among these, TP53 mutations are found in 15 % of ACC. Other genes regulating the cell cycle and commonly involved in cancer are also altered in ACC, including the tumor suppressors CDKN2A and RB1, and the oncogenes MDM2 and CDK4. These alterations globally stimulate the cell cycle, one of the key elements of cancer proliferation. Other pathways emerging from the list of recurrently altered genes in ACC are the chromatin remodeling (MEN1, DAXX, and ATRX mutations) and the chromosome maintenance (TERT and TERF2 amplifications). A few other recurrently mutated genes have been identified, including PRKAR1A and RPL22. Finally, other genes are altered, with no clear recurrence among ACC (“private” mutations). The exact role of these genes in ACC pathophysiology remains to be clarified.

It has clearly been demonstrated in transcriptome analysis that gene expression profile of adrenocortical tumors discriminates benign from malignant tumors. Meta-analyses have also confirmed this feature [23, 24]. The genes differentially expressed between benign and malignant adrenocortical tumors are implicated in several processes, like cell cycle, chromosomal maintenance, survival, inflammation, immunity… From this point of view, ACC is comparable to many other cancers not originating in the adrenal gland. Transcriptome also showed that IGF2, that was already reported overexpressed before the genomic area in ACC, is clearly the gene the most up-regulated in malignant by comparison with benign adrenocortical tumors or normal adrenal (see 25 for review). IGF2 ovexpression is also a feature common to various malignancies. Genes controlling cellular process more specific to the adrenal cortex can also be differentially expressed between benign and malignant adrenocortical tumors. In ACC, a decreased expression of steroidogenesis genes is observed, in keeping with the idea that a malignant tumor is less differentiated than a benign one.

Interestingly, in a translational perspective, transcriptome analysis has also being able to identify groups of ACC with different prognosis [26–28]. Indeed, tumor classification on the basis of transcriptome by cluster analysis reveals groups of ACC having specific gene expression profile and different survival. This reflects clearly different tumor biology likely resulting from various genetic and epigenetic alterations. The latter start now to be quite well described by others “omics” methods, as describe latter in this review. Interestingly, the prognostic value of gene expression profile is not directly link to tumor extension (ENSAT staging) suggesting its potential value along the clinical staging for prognostication.

Chromosomal alterations are numerous in ACC genome, especially compared to benign adenomas (see 16 for review). Pan-genomic analysis based on CGH arrays shows many gains (especially in chromosome 5, 7, 12, and 19) and many losses. Chromosomal alterations are also associated with survival [29]. In comparison with CGH arrays, single nucleotide polymorphism (SNP) arrays provide not only the DNA copy number information but also the loss of heterozygosity (LOH). Extensive LOH is observed in ACC and losses can occur at the level of an entire chromosomal arm. Copy number variation analysis also shows the possibility of whole genome doubling at the chromosomal level in ACC. Quite fascinating is the observation that LOH could be associated with this whole genome doubling. This suggests a sequence in which LOH occurs and then the remaining chromosome is gained [21].

In adrenocortical tumors, the involvement of altered DNA methylation in tumorigenesis has been reported before the genomic area with the study of the imprinted 11p15 locus. Alterations of this locus are associated with the Beckwith-Wiedmann syndrome, an overgrowth syndrome predisposing to ACC and are frequent in sporadic ACC. The paternal allele is methylated and expresses IGF2, whereas the maternal allele is normally not methylated and expresses H19, a non-protein coding RNA that is thought to exhibit tumor suppressor activity. In ACC, the H19 promoter methylation is associated with H19 underexpression and IGF2 overexpression, as previously discussed about the transcriptome studies of ACC. Genomic studies of cancers helped to understand that a global hypomethylation inducing genomic instability, loss of parental imprinting, and reactivation of transposable elements that can be observed in some cancers. On the other hand, hypermethylation of CpG islands located in the promoter regions of tumor suppressor genes can also be observed. The two types of alteration are not exclusive. Pan-genomic studies have shown recently that the ACC genome is globally hypomethylated [30]. By contrast, the CpG islands in promoter regions are hypermethylated [31]. This can cause down-regulation of tumor suppressor genes. The methylation levels in CpG islands from promoter regions vary among ACC; some have methylation levels comparable to adenomas, and others being hypermethylated. The hypermethylation profile of some cancers has been reported in other malignancies and referred to as CIMP for “CpG Island Methylator Phenotype.” In colon cancer, CIMP is associated with a worse prognosis. As for colon cancer, hypermethylation was found associated with a worse outcome in ACC [20, 31].

MicroRNA (miRNA) are small non coding RNA (approximately 22 nucleotides). Approximately 2000 miRNA have been identified. Some of these miRNA have been suggested as biomarkers for cancer diagnosis and prognosis. The miRNA plays an important role in the post-translational regulation of gene expressions by targeting mRNA for cleavage or translational repression. A deregulation of the expression of the miRNA is involved in several cancers through activation of oncogenes or silencing of suppressor gene tumors. In ACC, a deregulation of several miRNA has been observed. MiRNA expression profile differs between adenomas and ACC. Some miRNA have been suggested as diagnostic markers of ACC in patients with adrenocortical tumors: high expression of miR-503 and low expression of miR-511 could distinguish adenomas from ACC [32]. A high expression of the miR-483-5p and a low expression of miR-195 and miR-335 are observed in ACC [33–35]. The level of circulating miR-483-5p also distinguishes between ACC and adenomas [36]. MiRNome analysis reveals different miRNA clusters in ACC. In the European Network for the Study of Adrenal Tumors (ENSAT)-integrated genomic study of ACC, three clusters of miRNA have been differentiated: Mi1, Mi2, and Mi3. Mi1 is associated with a group of tumor having a good prognosis [20]. The DLK1-MEG3 miRNA is down-regulated in this group of tumors. Interestingly, integrated genomics showed that this is due to LOH at the locus and that the remaining allele is fully methylated explaining the down-regulation of the miRNA [20].

When all these genomics analysis performed on the same set of tumors were integrated, the major finding that came out was the existence of distinct molecular subgroups. One can expect that the biology of the tumor explains its evolution. Indeed, the molecular subgroups of ACC identified by integrated genomics are strongly associated with survival. To summarize that briefly, (1) transcriptome analysis reveals two main subgroups of ACC, strongly associated with prognosis [26, 27]; (2) miRNome analysis shows three subgroups of ACC, also associated with different outcomes [20]; (3) somatic alterations of driver genes are shown by exome or SNP analysis in more than half of the ACC, and these ACC are associated with a worse outcome [20, 21]; and (4) an hypermethylation of the CpG islands located in regulatory regions of the genes (the so-called CIMP phenotype) is associated with a poor outcome [20, 21, 31]. The molecular classification described by the ENSAT and TCGA studies is rather similar. This underlines the robustness of the classification of ACC based on genomics. Two groups of ACC are therefore identified: one with a good prognosis in term of overall survival and a group of poor prognosis. This last one can be subdivided by the identification of a subgroup of intermediate prognosis. The good prognosis subgroup of ACC with better outcome is characterized by convergent molecular features at the transcriptome, genome, and methylome levels. These tumors have a low mutation rate, and very few of them present alterations of the main driver genes of ACC. The methylation level is similar to normal adrenal or adrenal adenomas; so clearly, these tumors do not present a CIMP phenotype. The poor prognosis subgroup presents a higher expression of cell cycle-related genes. These tumors have a high mutation rate, and driver gene alterations are frequently observed in this subgroup of ACC. This group of aggressive ACC can also be divided according to the methylation levels of the CpG islands: a subgroup of tumors shows hypermethylation at the level of the CpG islands located in the promoter of genes (“CIMP phenotype”). This hypermethylation is associated with a worse prognosis compared with tumors with no hypermethylation of the same group of aggressive ACC. Therefore, methylation analysis allows probably to split the group of aggressive ACC into one subgroup of worst outcome and another subgroup of intermediate outcome.

It is likely that pathology would identify different characteristics between these two groups of ACC. Based on transcriptome analysis, it was already suggested that the poor prognosis group of ACC consists of high-grade tumors [27]. It is also expected that makers of proliferation (e.g., high Ki67 immunohistochemistry and mitotic count) would be higher in this group.

Conclusion

It is now an exiting time for the field of adrenocortical tumors. The efforts of several teams of pathologists to assess in a rigorous and standardized fashion the diagnosis and prognostication of ACC have an immediate impact on patient management. The numerous genetic and epigenetic alterations reported in less than 3 years by ACC genomics studies allow now to have a clear view of the landscape of the molecular alterations occurring in ACC development. Thanks to the power of these pan-genomic methods, we had clearly learned much most in these recent years than we knew previously.

On the short term, this will help to develop molecular markers to strengthen along with pathology prognostication of ACC. We have also realized that most of the ACC driver genes clearly listed by genomics are not the best candidates for the targeted therapies currently used to treat many others cancers. But on the long term, there is no doubt that this will be a great source of idea guiding the development of very innovative therapies targeting ACC molecular alterations.

Compliance with Ethical Standards

Disclosure Summary

The authors have nothing to declare.