Introduction

Primary bilateral macronodular adrenal hyperplasia is a rare cause of Cushing’s syndrome, first described in 1964 [1] with changing terminology over the years. The recent description of clonal molecular alterations has led to consider the neoplastic nature of the disease and the name BMAD for the bilateral macronodular adrenocortical disease was first proposed by Pakbaz and Mete [2] and then by Hodgson et al. and Juhlin et al. [3, 4] and endorsed by the WHO Adrenal Cortex Tumor Classification Consensus Conference 2022 [5].

BMAD is mainly diagnosed in patients in the 5th or 6th decade [6]. It is defined by the development of several bilateral macronodules larger than one centimeter. Some cases can be asymmetrical. This disease is also heterogeneous as demonstrated by the high variability in the clinical signs of Cushing’s syndrome [7], the level of pituitary ACTH-independent hypercortisolism [8, 9], and the number of nodules seen on CT [7]. Steroidogenesis inhibitors (such as metopirone or ketoconazol) have been reported in some cases to control hypercortisolism, but adrenalectomy (bilateral or in selected cases unilateral) is the standard treatment indicated on the basis of the comorbidities and the level of cortisol excess [6, 8, 10].

In contrast with this clinical heterogeneity, to date, pathological descriptions are rather homogeneous in the 55 published case reports of BMAD including a pathological description. Indeed, the enlarged adrenal glands distorted by multiple yellow macronodules larger than 1 cm and up to more than 4 cm, well bounded by fibrous trabeculae appears as the main macroscopic description of BMAD. Stratakis and then Hsiao and collaborators proposed to divide BMAD into two subtypes, type I and II based on the presence of residual adrenal gland [11, 12]. Type I is characterized by nodules separated by residual adrenal gland while type II is characterized by no residual adrenal gland or atrophic cortex around the nodules. The nodules were composed of a majority of large, clear, lipid-rich, microvacuolated cells similar to those of the zona fasciculata. However, the proportion of smaller, eosinophilic compact cells was variable between cases [1, 13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45]. These observations support the hypothesis that a pathological heterogeneity could exist in BMAD.

Most cases are isolated. They can be sporadic or familial [7]. Indeed, rare BMAD cases have initially been described in patients with a genetic syndrome [4, 46] such as multiple endocrine neoplasia type 1 (MEN1 gene) [4, 46,47,48], familial adenomatous polyposis (APC gene) [4, 46, 49], hereditary leiomyomatosis and renal cell cancer (FH gene) [4, 38, 46], and McCune Albright syndrome (GNAS gene) [4].

The recent molecular descriptions demonstrated a genetic heterogeneity in isolated BMAD. To date, these genetic studies divide the isolated BMAD into 3 genetic groups [50]. The first group presents with biallelic inactivation of the ARMC5 (Armadillo containing protein 5) gene [51]. ARMC5 pathological variants are found in about 20 to 25% of isolated BMAD cases [52, 53], 50% of the operated patients (due to more severe forms) [51, 54], and in 80% of cases with a familial presentation [7]. The second group presents with biallelic inactivation of the KDM1A (lysine (K)-specific demethylase 1A/LSD1) gene [50, 55]. The presence of this inactivation leads to aberrant expression of the GIP (gastric inhibitory polypeptide) receptor by the adrenal cortical cells, resulting in food-dependent hypercortisolism [19]. To date, the third group has no identified genetic alteration [50].

So far, no pathological classification of this clinically, molecularly, and genetically heterogeneous disease has been described. We propose the first study based on pathological characteristics and immunoexpression of adrenocortical markers in a large series of 35 BMAD. We then looked for potential correlations between the pathological features identified and the characteristics of the patients, in particular, their genotype.

Material and Methods

Patients

The series concerned 35 index cases with Cushing’s syndrome who underwent surgery for BMAD at the Cochin Hospital Paris, France, between 1998 and 2021. All patients had a constitutional genetic profile study (COMETE network). Leucocyte DNA from BMAD patients has been sequenced for ARMC5, KDM1A, APC, PRKACA, PRKAR1A, and MEN1 genes by next-generation sequencing using the Ion Personal Genome Machine system (Ion Torrent, Thermo Fisher Scientific, USA) or a NextSeq 500 sequencer (Illumina). All clinical, biological, radiological, and pathological information were collected without the knowledge of the genetic data.

All patients gave their written consent for research purposes including genetic analysis in the national COMETE network. This project was approved as a monocentric retrospective study by the data protection office (bureau de la protection des données, registre d’enregistrement AP-HP) (number 20220221155734) and the CLEP, (comité local d’éthique des publications de l’hôpital Cochin) (number AAA-2022–08,019). It complies with the principles of the declaration of Helsinki.

Clinic and Biology

The following clinical information were collected: age, sex, weight, and height. The following biological results were collected: 24-h urinary free cortisol, plasma cortisol at 8:00 am, at midnight, after dexamethasone suppression, and testosterone levels.

Radiology

Computed tomography (CT) scans were available for review in 29/35 patients (83%). An abdominal radiologist (M.B) reported the following morphological criteria: maximal adrenal gland size, number of nodules, and spontaneous adrenal gland attenuation.

Pathology

Macroscopic photographs and stained slides were reviewed by two pathologists (FV, MS) in consensus, without knowledge of clinical, biological, radiological, and genetic data. The following macroscopic criteria were collected: adrenal weight, number and maximum nodule size, color, separate or coalescent nature of nodules, and presence of residual adrenal parenchyma (type I) or not (type II) [12].

A microscopic study was performed on all available nodules on 3 µm sections stained with a Tissue-Tek Prisma Plus automat (Sakura Finetek Europe BV; Zoeterwoude, The Netherlands). We reviewed a median of 10 H&E slides by the case (minimum: 4, maximum: 20). The microscopic criteria collected were architecture (round fibrous septa or sparse fibrous trabeculae inside macronodules, usually evaluated by reticulin [56, 57] or by Sirius red histochemistry [58], presence or absence of trabeculae, pseudoglandular aspects), type and proportion of cells (clear, compact, and oncocytic cells), inflammatory infiltrate, and adipocytic metaplasia.

Exhaustive immunohistochemical study was performed on dewaxed sections of 3 µm thickness on a Leica BOND-III System (Leica, Berlin, Germany). HIER (heat-induced epitope retrieval) treatment was performed for 20 min in a pH6 buffer solution (EDTA buffer, Bond Epitope Retrieval Solution 1, Leica, Berlin, Germany) for anti-CYP17A1 (clone HPA048533, diluted 1:800, Sigma Aldricht, Saint Louis, USA), anti-DAB2 (clone HPA028888, diluted 1:800, Sigma Aldricht, Saint Louis, USA) anti-CYP11B2 (clone 41-17B, diluted 1:500; Millipore, Frankfurt, Germany), and anti-KDM1A antibodies (clone ab-17721, diluted 1:400; Abcam, Boston USA). HIER treatment was performed for 20 min in pH9 buffer (EDTA buffer, Bond Epitope Retrieval Solution 2, Leica, Berlin, Germany) for anti-alpha inhibin (clone R1 IR058 ready to use, Dako, Santa Clara, USA), anti-HSD3B1 (clone WH0003283M1, diluted 1:8000, Sigma Aldricht, Saint Louis, USA), anti-HSD3B2 (clone SAB1402232, diluted 1:8000, Sigma Aldricht, Saint Louis, USA), and anti-CYP11B1 antibodies (clone 80–7, diluted 1:100, Millipore, Frankfurt, Germany). We performed double immunostaining using anti-HSD3B2 and anti-alpha inhibin antibody and with anti-CYP17A1 and anti-DAB2 antibody.

In the normal adult adrenal gland, HSD3B2 is expressed in the zona fasciculata and reticularis and has a lower expression in the zona glomerulosa [59, 60]. HSD3B1 is expressed in the three zones of the adrenal cortex [59, 60]. CYP17A1 is expressed in the zona fasciculata and the zona reticularis [59]. CYP11B1 is expressed in the zona fasciculata [59, 60]. DAB2 is expressed diffusely in the zona glomerulosa and CYP11B2 is expressed by aldosterone-secreting cells in the zona glomerulosa [60,61,62]. Alpha inhibin is expressed in the zona reticularis and can be expressed in the zona fasciculata [60, 63].

Statistical Analyses

The multiple factor analysis was performed with all microscopic criteria using R software (version 4.0.5, R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.) with Factoshiny library. All microscopic criteria, qualitative and quantitative, had the same weight. Clustering was performed without pre-treatment with K-means or consolidation. Metric was Euclidian. Statistical analyses were performed using nonparametric tests appropriate for small numbers. Qualitative data were analyzed using Fisher exact test using R software (version 4.0.5). Quantitative data were analyzed using Kruskal–Wallis and Wilcoxon tests using R software (version 4.0.5).

Results

Patients

Most patients were female (66%), overweight (45%), or obese (20.5%). The median age was in the 6th decade (51.3 years). All patients had pituitary ACTH-independent clinical or subclinical Cushing’s syndrome. CT scans showed enlarged, multinodular adrenal glands with a spontaneous density of less than 20 HU in most cases (88%). Demographic, clinical, biological, and radiological characteristics are presented in Table 1.

Table 1 Cohort characteristics
Full size table

Macroscopic Characteristics

By definition, all 35 patients had multinodular adrenal glands that increased in weight and volume. The nodules were mainly coalescent (86%) and yellow (72%). Some nodules contained red or brown areas. Most patients presented with type II with atrophic or no adrenal parenchyma around the nodules (57%). All macroscopic features are presented in Table 2.

Table 2 Macroscopic characteristics
Full size table

Microscopic Characteristics

The nodules were densely cellular, well-bounded, and separated by fibrous trabeculae that contained capillaries. Most of the cases (63%) showed round fibrous septa around several nests of adrenal cells inside nodules. The architecture within the nodules was nest-like.

In most cases (33/35; 94%), the main cell population was composed of polygonal clear cells with microvacuolized cytoplasm and a medium-sized, round centrally located nucleus. Their proportion varied from 60 to more than 90% except for 2 cases with only 20% of clear cells.

In all cases, there were compact eosinophilic cells of smaller size. They were polygonal, regular with the cytoplasm of variable abundance, and a medium-sized, round centrally located nucleus. Their proportion varied from less than 10% to 40%. They were organized in islands, bands, or sheets intermingled with clear cells. In 11 cases, different cylindrical compact eosinophilic cells forming trabeculae were observed especially at the periphery of the nodules, close to the adipose tissue.

Occasionally, there were large, polygonal, oncocytic cells with a round, nucleolated, eccentric nucleus. Their cytoplasm was abundant, strongly eosinophilic, and granular. These cells frequently demonstrated a marked anisocytosis and anisokaryosis. Their proportion varied from less than 10% to 80%.

Most cases had an inflammatory lymphocytic infiltrate (63%) and adipocytic metaplasia (63%). Some cases had pseudoglandular aspects (cavitations in the center of several nests) of clear cells (29%). In one patient (3%), there was a myelolipomatous territory. All microscopic features are presented in Table 3.

Table 3 Microscopic characteristics
Full size table

Morphological Subtypes

We used all microscopic criteria listed in Table 4 to perform an unsupervised multiple-factor analysis. This analysis subdivided the 35 patients of our cohort into 4 distinct subtypes (Fig. 1).

Table 4 Microscopic criteria used for multiple factorial analysis
Full size table
Fig. 1
figure 1

Clustering divides the 35 BMAD cases into 4 morphological groups. S1, subtype 1; S2, subtype 2; S3, subtype 3; S4, subtype 4

Full size image

Those subtypes showed specific macroscopic (Fig. 2) and microscopic (Figs. 3–7) characteristics that are presented in Table 5.

Fig. 2
figure 2

Macrophotographs of BMAD. 1 white square = 1 cm. Subtype 1 cases (a, e) had no non-nodular adrenal gland in contrast with other subtypes. Subtype 2 cases (b, f) showed a light brown area. Subtype 3 cases (c, g) were composed of homogeneous yellow nodules. Subtype 4 cases (d, h) were made of dark brown nodules separated by the non-nodular adrenal

Full size image
Fig. 3
figure 3

Microphotographs of BMAD subtypes, H&E (a–d), reticulin histochemistry (e–h), and Sirius red histochemistry (i–l) magnification × 25. Subtypes 1 (a, e, i) and 2 (b, f, j) cases had round fibrous septa within the macronodules (arrow). Subtype 3 (c, g, k) and 4 (d, h, l) contained few sparse arciform fibrous trabeculae (arrowhead)

Full size image
Table 5 Pathological characteristics of the subtypes
Full size table

In subtype 1 (n = 17), macroscopically, the adrenal parenchyma was distorted by coalescent, yellow macronodules. Areas of residual non-nodular adrenal gland were rare (type II) (Fig. 2a, e). Microscopically, the macronodules contained round fibrous septa composed of collagen fibers (Fig. 3a, e, i) well demonstrated by reticulin and sirius red histochemistry. Clear cells occupied 70 to 90% of the nodules. Their size and shape were regular. Compact cells represented 10 to 30% of the nodules and formed islands or bands at the periphery and in the center of the nodules (Fig. 4a, e, i). Pseudoglandular aspects inside nodules and trabeculae of cylindrical eosinophilic cells at the periphery and toward the center of the nodules were only seen in this subtype (Fig. 5). Oncocytic cells were rare. Most cases did not show a lymphocytic infiltrate. Areas of adipocytic metaplasia were common (Fig. 4a, e).

Fig. 4
figure 4

Microphotographs of BMAD subtypes, H&E staining, magnification × 12.5 (a–d), magnification × 50 (e–h), and magnification × 400 (i–l). Subtype 1 (a, e, i). Subtype 2 (b, f, j). Subtype 3 (c, g, k). Subtype 4 (d, h, l)

Full size image
Fig. 5
figure 5

Microphotograph of subtype 1 specificities, H&E staining. Eosinophilic cylindrical cells forming trabeculae at the periphery (arrow), magnification × 100 (a). Pseudoglandular aspects with optically empty cavitations in the center of clear cell nests (arrowhead), magnification × 400 (b)

Full size image

In subtype 2 (n = 4), macroscopically, the adrenal gland was predominantly occupied by coalescent yellow nodules with some light brown areas. Most often, areas of residual non-nodular adrenal gland persisted (Fig. 2b, f). Microscopically, the macronodules contained round fibrous septa composed of collagen fibers (Fig. 3b, f, j). Clear cells occupied 60 to 70% of the nodules (Fig. 4b, f, j). Their size and shape were irregular (Fig. 6). Compact eosinophilic cells had abundant cytoplasm and represented 30 to 40% of the nodules (Figs. 4b, f, and 6). They were organized in extensive sheets closely intermingled with clear cells. Oncocytic cells were rare. All cases had a lymphocytic infiltrate. Most cases had areas of adipocytic metaplasia. One of the 4 patients had a myelolipomatous territory.

Fig. 6
figure 6

Microphotograph of subtype 2 specificities, H&E staining. Clear cells irregular in size and shape (arrow), magnification × 400 (a). Compact eosinophilic cells with abundant cytoplasm (arrowhead), magnification × 400 (b)

Full size image

In subtype 3 (n = 9), the adrenal glands were occupied by large, coalescent, yellow nodules. The residual non-nodular adrenal gland was sometimes visible (Fig. 2c, g). Microscopically, the macronodules contained sparse arciform fibrosis (Fig. 3c, g, k). Clear cells occupied more than 90% of the nodules. Their size and shape were regular. Compact cells were uncommon and formed rare small nests scattered in the nodules. Oncocytic cells were rare. Most cases showed a lymphocytic infiltrate and areas of adipocytic metaplasia (Fig. 4c, g, k).

In subtype 4 (n = 5), macroscopically, nodules were separated by areas of the non-nodular adrenal gland. The nodules were yellow and brown (Fig. 2d, h). Microscopically, the macronodules included sparse arciform fibrosis (Fig. 3d, h, l). Oncocytic cells represented 40 to 80% of the nodules (Fig. 7). The proportion of clear cells varied between 20 and 60%. They were regular in size and shape. Compact cells were rare and formed small islands at the periphery or in the center of the nodules. Most cases had a lymphocytic infiltrate and adipocytic metaplasia (Fig. 4d, h, l).

Fig. 7
figure 7

Microphotograph of subtype 4 specificities, H&E staining magnification × 400. Oncocytic cells possess an abundant, granular, eosinophilic cytoplasm with a round, nucleolated, eccentric nucleus. They are larger than compact eosinophilic cells. Anisokaryosis and anisocytosis are often marked

Full size image

Interestingly, this subtype classification concerned both adrenal glands when patients had a bilateral adrenalectomy. The presence of round fibrous septa inside nodules and the proportion of cells (i.e., clear, compact, and oncocytic) were the most discriminating criteria to distinguish these 4 subtypes (Table 6).

Table 6 Main characteristics of BMAD subtypes
Full size table

Immunohistochemistry

Immunohistochemical studies were performed on 34 out of 35 patients (Figs. 8, 9, 10, 11). The remaining patient had non-contributive results due to technical problems.

Fig. 8
figure 8

H&E magnification × 100 (a–e). Immunohistochemistry, double staining with HSD3B2 (red) and alpha-inhibin (brown) magnification × 100 (f–j). Immunohistochemistry, double staining with DAB2 (brown), and CYP17A1 (red) magnification × 100 (k–o). Normal adrenal gland (a, f, k); subtype 1 (b, g, l); subtype 2 (c, h, m); subtype 3 (d, i, n) subtype 4 (e, j, o). In all subtypes, HSD3B2 was preferentially expressed by clear cells and CYP17A1 was preferentially expressed by compact eosinophilic cells. In most subtype 1 cases, HSD3B2 was expressed by all clear cells. In other subtypes, HSD3B2 was mainly heterogeneous on clear cells. In subtype 2 cases, alpha inhibin expression by compact cells was stronger than in other subtypes

Full size image

In all cases, HSD3B2 was preferentially expressed by clear cells (Fig. 8b–e, g–j). Fifteen cases (14 subtypes 1 and 1 subtype 3) showed diffuse staining on all clear cells (Fig. 8g). The remaining 19 cases showed heterogeneous staining with 40–80% of clear cells expressing HSD3B2 (Fig. 8h–j). In all cases, CYP17A1 was preferentially expressed in compact eosinophilic cells (Fig. 8b–e, l–o). HSD3B1 and CYP11B1 were expressed by all nodular cells, regardless of their type (Fig. 9).

Fig. 9
figure 9

H&E magnification × 100 (a–e). Immunohistochemistry CYP11B1 magnification × 100 (f–j), immunohistochemistry HSD3B1 magnification × 100 (k–o). Normal adrenal gland (a, f, k); subtype 1 (b, g, l); subtype 2 (c, h, m); subtype 3 (d, i, n); subtype 4 (e, j, o). In all subtypes, both CYP11B1 and HSD3B1 were expressed by all cell types in all subtypes

Full size image

DAB2 was expressed by peripheral trabecular cylindrical eosinophilic cells, only seen in subtype 1 (Fig. 10e). These cells also express HSD3B2 but did not express CYP17A1 or CYP11B2 (Fig. 10f).

Fig. 10
figure 10

H&E magnification × 100 (a, d). double staining with DAB2 (brown) and CYP17A1 (red) magnification × 100 (b, e). Immunohistochemistry CYP11B2 magnification × 100 (c, f). Normal adrenal gland (a–c) zona fasciculata and reticularis express CYP17A1. Zona glomerulosa express DAB2. Within zona glomerulosa, some aldosterone-producing cells express CYP11B2 (white arrow). In subtype 1 (d–f), islands of polygonal compact cells expressed CYP17A1 (arrowhead). Trabeculae of cylindrical eosinophilic cells expressed DAB2 (black arrow) with no expression of CYP11B2

Full size image

Alpha inhibin was preferentially expressed in islands and bands of polygonal compact cells. It was more abundant in subtype 2 than in the other subtypes (Fig. 8g–j).

KDM1A expression was weaker only in subtype 2 nodules (Fig. 11).

Fig. 11
figure 11

H&E (a) and KDM1A immunohistochemistry (b) magnification × 200. KDM1A expression is weaker in the nodules (arrowhead) compared to adjacent adrenal tissue (arrow)

Full size image

The oncocytic cells took up anti-HSD3B2, CYP17A1, and anti-alpha inhibin antibodies (Fig. 8e, j, o).

In cases where it was visible, the non-nodular adrenal showed an immunohistochemical profile similar to the one observed in a normal adrenal.

In summary, in all subtypes, HSD3B2 was mainly expressed by clear cells and CYP17A1 preferentially stained compact cells. HSD3B1 and CYP11B1 were expressed in all cell types. Specifically, in subtype 1, DAB2 stained trabeculae of cylindrical eosinophilic cells with no co-expression of CYP11B2. Specifically, in subtype 2, alpha inhibin was strongly expressed in compact cells and KDM1A immunoexpression was weaker in the nodules than in the adjacent normal parenchyma.

Genetic Characteristics

Among the 35 patients in the series, 15 (43%) had a constitutional pathogenic ARMC5 variant, 4 (11%) had a constitutional pathogenic KDM1A variant, and 16 (46%) had no specific or recurrent known pathological variant to date. No patient had a PRKAR1A, PRKACA, APC, or MEN1 pathological variant. FH pathogenic variants were not investigated as no patient had clinical suspicion of FH genetic alteration.

Correlations

BMI was significantly higher in subtype 1 patients than in patients from subtype 4. Plasma cortisol at 8:00 am in subtype 2 patients was significantly lower than in the other groups. No significant difference was found in the intensity of 24 h free urinary cortisol. On the CT scan, we found only one significant difference: the spontaneous adrenal density of subtype 4 was higher than in the other groups. The adrenal size of subtype 4 was significantly smaller than in the other subtypes (Table 7). Type II correlated with the presence of a pathogenic ARMC5 variant (p < 0.01). This criterion had a positive predictive value of 62.5% and a negative predictive value of 91.5% to detect a pathogenic ARMC5 variant.

Table 7 Correlation between clinical, biological, radiologic, genetic data, and morphological subtypes
Full size table

We analyzed the correlation between the morphological subtype and the genetic profile of the patients. Subtype 1 contained most patients with a constitutional pathogenic ARMC5 variant. Subtype 2 contained exclusively patients with a constitutional pathogenic KDM1A variant. Subtypes 3 and 4 mainly concerned patients without a known constitutional variant (Table 7).

The clear cells of the 15 patients with an ARMC5 pathogenic variant uniformly express HSD3B2 in contrast to the 4 patients with a KDM1A pathogenic variant and 15 patients with no known genetic cause who express HSD3B2 heterogeneously on clear cells. Weak KDM1A expression in nodule cells was only seen in the 4 patients with a KDM1A pathogenic variant.

Discussion

We report here the first detailed pathological and exhaustive immunohistochemical description of a cohort of 35 patients with BMAD operated by unilateral or bilateral adrenalectomy in our center. In agreement with the literature, the majority of BMAD cases were female and the patients were mainly in their 6th decade when they underwent surgery [7, 11, 64, 65].

The microscopic characteristics found in our series were used to perform an unsupervised multiple-factor analysis. It separated the 35 patients into four distinct histopathological subtypes according to the presence of round fibrous septa and the proportion of cell types (clear, compact, oncocytic).

These morphological subtypes show a strong correlation with the genetic data of our cohort. Patients with a biallelic inactivation of ARMC5 and KDM1A gene belong to two different microscopically homogeneous groups, subtype 1 and subtype 2, respectively. In contrast, patients without a known constitutional variant are morphologically heterogeneous and are distributed between subtypes 1, 3, and 4. All nodules of a patient have the same morphology. Type II was more frequent in patients with a pathological ARMC5 variant. These data strongly support the hypothesis of a causal link between the driver genes and the morphologic appearance of the adrenal glands in BMAD.

In the literature, the microscopic description of the 55 case reports of BMAD is very similar from one publication to another (Table 8). In most of the case reports, the microscopic description is unfortunately not sufficiently detailed to allow classification of the case according to our model. In our series, fibrous septa are frequent within macronodules but were only mentioned in one case report [18]. The presence of cylindrical eosinophilic cell trabeculae at the periphery of the nodules has only been described in one case [17]. To our knowledge, the presence of oncocytic cells has never been mentioned in the microscopic characteristics, and we found them in 5 out of 35 patients. The presence of myelolipomatous territories was reported in two cases [1, 44] and in some patients of a series of BMAD with pathological KDM1A variant [55].

Table 8 The microscopic description of the 55 case reports
Full size table

In BMAD, the two main cell populations (clear and compact) have been studied by several authors, in particular, Sasano et al., using antibodies against steroidogenic enzymes involved in the synthesis of adrenal steroid hormones, in particular: HSD3B2 and CYP17A1 [66]. HSD3B2 is predominantly expressed in clear cells while CYP17A1 is preferentially expressed in compact eosinophilic cells [20, 21, 23, 25, 31, 34, 39]. Our immunohistochemical results, consistent with those previous observations, reinforce that this preferential staining of HSD3B2 in clear cells and CYP17A1 in compact cells as previously described by Sasano et al. [39] may become a diagnostic criterion of BMAD. Indeed, the lack of HSD3B2 and CYP17A1 co-expression may allow to distinguish BMAD from multiple cortisol-producing adenomas (characterized by the co-expression of HSD3B2 and CYP17A1) [59, 67].

BMAD cells produce less cortisol than the normal adrenal gland cells, and hypercortisolism is due to the increase of the number of adrenal cells [68]. Sasano et al. postulated that this low efficiency of cortisol production could be explained by the lack of co-expression of HSD3B2 and CYP17A1. According to this hypothesis, as the different enzymatic steps of steroidogenesis would be performed in different cells, the efficiency of cortisol production would be decreased [39]. Our observations corroborate this hypothesis. We observed that the clear cells of patients with an ARMC5 pathogenic variant uniformly express HSD3B2 in contrast to all other BMAD expressing HSD3B2 heterogeneously. Diffuse HSD3B2 expression could then explain, at least in part, the greater severity of Cushing’s syndromes in ARMC5 mutated patients compared to other BMAD patients [65]. HSD3B1 is expressed by all BMAD nodular cells. It is likely that this isoform of the enzyme, expressed in peripheral tissues, is not sufficient to replace HSD3B2. In conclusion, the partial expression of steroidogenic enzymes observed in BMAD nodule cells may participate to the reduction of cortisol production efficiency.

DAB2, a zona glomerulosa marker distinguishes the two populations of eosinophilic cells observed in subtype 1. The population of cylindrical eosinophilic cells in trabeculae at the periphery expressed DAB2 and did not express CYP11B2 or CYP17A1, whereas the population of compact eosinophilic cells in islands and bands expressed CYP17A1 and not DAB2. None of them appear to be able to produce aldosterone. Consistently, we did not find a correlation between subtype 1 and mineralocorticoid secretion (data not shown). Although hypertension is more frequent in ARMC5 mutated patients than in the other BMAD groups [65], our immunohistochemical and biological data do not support the hypothesis of an aldosterone co-secretion explaining the hypertension in these patients.

Alpha inhibin had never been studied in BMAD to our knowledge. This marker is more expressed in compact cells of subtype 2 compared to other subtypes but we did not find a significant correlation between this subtype and androgen secretion (data not shown). Furthermore, previous studies showed that patients with a biallelic KDM1A inactivation have an overexpression of LH receptors on the surface of adrenal cells [55]. We can hypothesize that the high expression of alpha inhibin comparatively to the other subtypes leads to the inhibition of LH-FSH secretion in this group and then, activates a compensatory mechanism through overexpression of the LH receptor [69]. As observed before in a recent work from our group, weak expression of KDM1A in the nodules is a strong argument in favor of a KDM1A pathogenic variant [50].

The identification of these morphological and immunohistochemical correlations with genetics could have an impact on the genetic sequencing strategy of BMAD specimens from patients who underwent surgery. Patients with a pathogenic ARMC5 variant frequently presented a type II and a microscopic subtype 1. All of them had a diffuse immunoexpression of HSD3B2 on clear cells unlike other patients. Similarly, all patients with a pathological KDM1A variant had a microscopic subtype 2 phenotype and a weak immunoexpression of KDM1A in the nodules compared to the adjacent adrenal gland. Because of their strong correlation with genetic clusters, the pathological characteristics could be incorporated into a genetic search strategy to target either ARMC5 or KDM1A sequencing in adrenal specimens. We previously showed that in the presence of several bilateral nodules on CT scan and a plasma cortisol greater than 50 nmol/L after 1 mg of dexamethasone, the probability of carrying an ARMC5 pathogenic variant is 20% with a sensitivity of 100% [65]. Pathology could guide the identification of this 20% subgroup of ARMC5 mutated patients.

The main observations of this study are summarized in Fig. 12.

Fig. 12
figure 12

Graphical abstract of the main observations

Full size image

In conclusion, our study proposes the first histopathological classification of BMAD into 4 morphological subtypes based on the architecture of the macronodules and the proportion of cell types: clear, compact, and oncocytic. This histopathological classification needs to be validated on a larger number of cases, but the strong correlation with the germinal genetic characteristics of the patients appears as a solid element of validation. This model highlights the heterogeneity of the pathological characteristics of BMAD as well as their link with the genetic characteristics.