Mass spectrometry-based steroid profiling in primary bilateral macronodular adrenocortical hyperplasia

Abstract

Biochemical characterization of primary bilateral macronodular adrenocortical hyperplasia (PBMAH) by distinct plasma steroid profiles and its putative correlation to disease has not been previously studied. LC-MS/MS–based steroid profiling of 16 plasma steroids was applied to 36 subjects (22 females, 14 males) with PBMAH, 19 subjects (16 females, 3 males) with other forms of adrenal CS (ACS), and an age and sex-matched control group. Germline ARMC5 sequencing was performed in all PBMAH cases. Compared to controls, PBMAH showed increased plasma 11-deoxycortisol, corticosterone, 11-deoxycorticosterone, 18-hydroxycortisol and aldosterone, but lower progesterone, DHEA and DHEA-S with distinct differences in subjects with and without disease-causing variants in ARMC5. Steroids that showed isolated differences included cortisol and 18-oxocortisol with higher (P<0.05) concentrations in ACS than in controls, and aldosterone with higher concentrations in PBMAH when compared to controls. Larger differences in PBMAH than with ACS were most clear for corticosterone, but there were also trends in this direction for 18-hydroxycortisol and aldosterone. Logistic regression analysis indicated four steroids - DHEA, 11-deoxycortisol, 18-oxocortisol, and corticosterone - with the most power for distinguishing the groups. Discriminant analyses with step-wise variable selection indicated correct classification of 95.2% of all subjects of the four groups using a panel of nine steroids; correct classification of subjects with and without germline variants in ARMC5 was achieved in 91.7% of subjects with PBMAH. Subjects with PBMAH show distinctive plasma steroid profiles that may offer a supplementary single-test alternative for screening purposes.

Introduction

Primary bilateral macronodular adrenocortical hyperplasia (PBMAH) is a rare form of adrenocorticotropic hormone (ACTH)-independent Cushing syndrome (CS), responsible for <1% of endogenous hypercortisolemia (Stratakis and Kirschner 1998). PBMAH is typically diagnosed in adults over 40 years of age with clinical and biochemical heterogeneity in presentation, including incidental detection on imaging, with or without asymptomatic, subclinical or overt hypercortisolism (Vassiliadi and Tsagarakis 2019); cortisol, its precursor steroids, and occasionally other hormones, such as aldosterone, are aberrantly regulated by non-mutated eutopic and/or ectopic G-protein coupled receptors (GPCRs) (El Ghorayeb, et al. 2015; Vassiliadi and Tsagarakis 2019). Glucocorticoid production could also be regulated by intra-adrenal ACTH production from a subpopulation of steroidogenic cells in the adrenal cortex (Louiset, et al. 2013). On imaging, PBMAH is characterized by the presence of multiple bilateral macronodules (larger than 1cm associated with internodular zona fasciculata hyperplasia or atrophy) (Stratakis and Kirschner 1998; Vassiliadi and Tsagarakis 2019).

For many decades, PBMAH was considered as a sporadic disease, although familial cases have now been described (Albiger, et al. 2017; Espiard, et al. 2015; Faucz, et al. 2014; Stratakis 2008). Indeed, our group and others have found that biallelic inactivation of the Armadillo repeat containing 5 gene (ARMC5;16p11·2, NM_001288767), a putative tumor suppressor gene and a member of the ARM (armadillo/beta-catenin-like repeat) superfamily involved in protein-protein interactions (Berthon, et al. 2017a), is a frequent cause of PBMAH among all ethnicities (Alencar, et al. 2014; Assie, et al. 2013; Espiard et al. 2015; Faucz et al. 2014; Gagliardi, et al. 2014; Yu, et al. 2018). The association of a germline and somatic tissue-specific disease-causing variants is required for the development of PBMAH, which is consistent with the Knudson two-hit model of tumorigenesis. Individuals with disease-causing variants in ARMC5 manifest with more severe hypercortisolism and larger adrenal glands when compared to predicted benign variants (Faucz et al. 2014).

The biochemical diagnosis of hypercortisolism due to PBMAH is difficult due to lack of a specific biochemical signature and slowly progressive hypercortisolism, which usually develops over several years and accompanied by subtle clinical manifestations. Initial studies have assessed the utility of 24-hr urinary 17-hydroxysteroids as a screening test for PBMAH; however, its lack of specificity and limited availability by most commercial testing laboratories render it impractical (Hsiao, et al. 2009). Nevertheless, reliance on the 24-hr urinary free cortisol (UFC) may also be erroneous, as this measure of adrenocortical activity can be misleadingly normal in PBMAH (Hsiao et al. 2009). Current diagnosis of PBMAH relies on general screening tests for CS, including overnight 1mg dexamethasone suppression test, salivary cortisol and UFCs. Thus, accurate biochemical characterization of PBMAH by distinct plasma steroid profiles holds promise for more accurate diagnosis of the disease, as well as a putative correlation to disease phenotype and disease-causing germline variants in ARMC5.

Indeed, measurements of plasma or urinary multi-steroid profiles by gas chromatography mass spectrometry or liquid chromatography with tandem mass spectrometry (LC-MS/MS) has helped discriminate various forms of endogenous CS (Arlt, et al. 2011; Eisenhofer, et al. 2018; Hines, et al. 2017; Velikanova, et al. 2016). These assays allow simultaneous measurement of multiple low molecular weight analytes with high analytical specificity despite structural similarities. Based on these promising findings of using steroidonomics in CS, we hypothesized that subjects with PBMAH would show differences in plasma steroid profiles when compared to control subjects and to those with other adrenocortical tumors, including cortisol-producing adrenocortical adenoma (CPA) and bilateral adrenocortical nodules manifesting with either primary aldosteronism (PA) or subclinical CS (SCS). We employed a newly developed LC-MS/MS method to measure 15 adrenal steroids in single plasma samples collected from subjects with PBMAH and other causes of adrenal tumors (Eisenhofer, et al. 2017a; Peitzsch, et al. 2015), and compared them to a large age- and gender-matched control group (Eisenhofer, et al. 2017b).

Methods

Subjects

We performed a multicentric cross-sectional study analyzing 36 subjects with biochemical hypercortisolism, with or without clinical CS, as either subclinical or overt PBMAH, and 19 others with other causes of ACTH-independent CS, including 16 patients with cortisol-producing adrenocortical adenomas and 3 patients with micronodular adrenal hyperplasia, all recruited from the German Cushing Registry (Eisenhofer et al. 2017b). For comparison we chose a control group (n=336 subjects) that included a reference population of 227 hypertensive and normotensive volunteers and 109 other patients in whom hypercortisolism was excluded after testing for CS. We did not include individuals with Ethics approval was granted from the Institutional Review Board of each center and informed consent was provided by all participants. The Institutional Review Boards of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) (until 2010) and the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (2010 to present), NIH, approved the research protocol (Clinical Trial Registration no. NCT00005927).

Diagnosis of hypercortisolism

The criterion used to establish the diagnosis of CS, including PBMAH, followed the Endocrine Society guidelines for evaluation of endogenous hypercortisolemia (Nieman, et al. 2008). Subjects were initially examined for clinical features of CS, and were considered positive if they had 2 or more of the following established screening tests for CS: elevated 24-hr urinary free cortisol above the upper cutoffs of reference intervals; elevated midnight salivary cortisol (>1·5 ng/mL), attesting the loss of circadian rhythm of cortisol secretion; and lack of suppression of serum cortisol (fasting serum cortisol >1·8 ug/dL) following the 1 or 2 mg overnight dexamethasone suppression tests (Nieman et al. 2008). Distinction between ACTH-dependent and ACTH-independent CS was made by ACTH measurement, second-tier biochemical testing and radiographic imaging as described in the guidelines (Nieman et al. 2008). The biochemical diagnosis of PBMAH also relied on elevation in 24-hr urinary 17-hydroxysteroids, as detailed elsewhere (Hsiao et al. 2009). All subjects with PBMAH irrespective of ARMC5 status had low to suppressed ACTH values (14 ± 12 pg/mL [0–46 pg/ml]), with normal to mildly elevated UFCs (52 ± 71 ug/24hr [>45 ug/24hr]).

Most subjects with PBMAH (n=31) were evaluated at the NIH Clinical Research Center in the last 14 years (2004–2018). Five subjects were evaluated in Germany and met the biochemical and radiographic criteria for PBMAH. In total, 36 subjects were identified as PBMAH between all centers; all cases had confirmed bilateral macronodules (≥1cm) with hyperplasia and/or internodular atrophy on abdominal imaging. Twenty one (58.3%) subjects with PBMAH underwent bilateral adrenalectomy. Pathological confirmation of PBMAH, as bilateral multi or macronodular hyperplasia, with surgical correction of hypercortisolism, was achieved in all operated subjects. Fifteen (41.6%) subjects remained unoperated, and none were medically treated for hypercortisolism. Cases of bilateral micronodular adrenal hyperplasia were not included in the PBMAH group as they represent other forms of hyperplasias that would have confounded the steroid profiling analysis.

Genetics analysis of ARMC5

Genetic analysis of germline variants in ARMC5 was only performed in the PBMAH cohort. This is primarily because previous reports have shown that most incidentally discovered bilateral adrenocortical adenomas with/without PA have normal ARMC5 sequencing (Emms, et al. 2016; Mulatero, et al. 2016). The details of DNA extraction, sequencing and variant prediction modeling are described in Supplement 1.

Plasma steroid profiling

Blood samples for plasma steroid profiling were collected in the morning (08:00–11:00) into blood tubes containing lithium heparin or ethylenediaminetetraacetic acid. Separated plasma was stored at −80°C until steroid profile analysis by LC-MS/MS according to a previously described method (Eisenhofer et al. 2017a; Eisenhofer et al. 2017b; Peitzsch et al. 2015). The 15 steroids in the panel included cortisol, 11-deoxycortisol, 21-deoxycortisol, corticosterone, 11-deoxycorticosterone, aldosterone, 18-oxocortisol, 18-hydroxycortisol, cortisone, progesterone, 17-hydroxyprogesterone, pregnenolone, androstenedione, dehydroepiandrosterone (DHEA), and DHEA-sulfate (DHEA-S) as previously described (Eisenhofer, et al. 2016; Eisenhofer et al. 2017a; Eisenhofer et al. 2017b).

Immunohistochemistry

Five-micrometer sections were cut and deparaffinized in Histo-Clear (Nationals Diagnostics). After rehydration through serial ethanol dilutions, epitopes were retrieved by 20 minutes of boiling in Vector Antigen Retrieval Solution (H3300; Vector Labs). After 1 hour blocking in 1%BSA/PBS at room temperature, ARMC5 (NBP1–94024; Novus Biologicals), CYP11B1 (PA5–63290, ThermoFischer scientific) or CYP11B2 (NBP2–13891, Novus biologicals) primary antibodies were incubated overnight in a cold room before being recognized by anti-rabbit secondary antibody (MP-7401; Vector Labs). Horseradish peroxidase activity was then detected with 3,3-diaminobenzidine tetrahydrochloride (SK-4105; Vector Labs) (Faucz et al. 2014). Pictures were taken using the microscope BZ-X700 from Keyence.

Statistical analysis

JMP statistics software package (SAS Institute Inc, Cary, NC) was used for statistical analyses. All plasma steroid data were logarithmically transformed before parametric multivariate analyses. Multivariate analyses to test for differences between subject groups included age and sex as covariates consequent to the significant influences of both factors on most of the plasma steroids in the panel (Eisenhofer et al. 2017b). Data in figures for plasma concentrations of steroids are therefore shown as geometric least square means corrected for influences of age and sex. Post-hoc testing for differences between groups utilized the Tukey HSD test with significance determined as P<0.05. Logistic regression and/or discriminant analyses with stepwise variable selection were used to select minimal panels of the most useful steroids for distinguishing subjects of the various groups. For these analyses all data for plasma steroids were normalized for age and sex using previously published data for 525 subjects of a reference population (Eisenhofer et al. 2017b).

Results

Demographics

The 36 subjects with PBMAH in whom plasma samples were available for steroid profiling included 22 females and 14 males (Table 1). The mean age of the PBMAH subjects was 55·2 ± 11·6 years at the time of presentation (Table 1). Although the three groups of subjects with adrenal disease and the control groups did not differ significantly in age, the subjects with other causes of ACS tended to be younger than the other groups, while the subjects with PBMAH who tested negative for germline variants in ARMC5 tended to be older than all other groups, as expected (Table 1). Besides, although there were no overall significant differences in sex ratio between the five groups, the subjects with other causes of ACS showed a tendency to a higher proportion of women (84%) than the other groups (respectively, 42% to 71%). Also among subjects with PBMAH, those without germline variants in ARMC5 tended to have a higher proportion of women than those with germline variants in ARMC5 (respectively, 71% vs 42%).

Table 1.

Demographic data and plasma concentrations of 16 steroids in subjects of the control group (reference population and subjects in whom hypercortisolism was excluded) compared to subjects with adrenal Cushing syndrome (ACS; mainly cortisol-producing adrenocortical adenomas) and subjects with PBMAH with and without damaging germline variants in ARMC5.

	Control Group		ACS Group	PBMAH Group
	Reference	Excluded		ARMC5 -ve	ARMC5 +ve
Demographic data
Sex (M/F)	·	40/69	3/16	7/17	7/5
Age (years)	50	51	45	59	52
	(31–81)	(31–80)	(17–72)	(31–74)	(39–68)
Plasma steroids (ng/mL)
Cortisol	84	97	163 *	100	102
	(66–106)	(71–134)	(85–185)	(75–132)	(80–128)
11-Deoxycortisol	0.13	0.14	0.44 *†	0.44 *†	0.48 *†
	(0.09–0.19)	(0.09–0.24)	(0.33–0.92)	(0.25–0.78)	(0.30–1.14)
Cortisone	16.7	17.7	18.6	13.4	13.4
	(14.4–19.6)	(13.7–22.0)	(14.9–23.5)	(11.7–18.4)	(10.7–19.0)
Corticosterone	1.50	1.41	2.20	2.55	4.71 *†
	(1.04–2.60)	(0.85–2.54)	(1.23–3.86)	(1.24–4.81)	(1.62–7.52)
11-Deoxycorticosterone	0.02	0.03	0.09 *†	0.06 *†	0.12 *†
	(0.01–0.05)	(0.02–0.05)	(0.06–0.20)	(0.03–0.17)	(0.09–0.30)
18-Hydroxycortisol	0.57	0.79 *	0.76	1.11 *†	1.47 *†
	(0.37–0.80)	(0.51–1.27)	(0.45–2.03)	(0.88–1.89)	(0.90–2.95)
21-Deoxycortisol	0.010	0.009	0.011	0.019	0.025
	(0.003–0.022)	(0.011–0.023)	(0.004–0.060)	(0.004–0.045)	(0.002–0.052)
Aldosterone	0.04	0.05	0.06	0.07 *	0.05
	(0.02–0.07)	(0.03–0.10)	(0.03–0.179)	(0.04–0.12)	(0.03–0.12)
18-Oxocortisol	0.007	0.012 *	0.017 *	0.010†	0.010
	(0.001–0.049)	(0.003–0.330)	(0.002–0.500)	(0.010–0.042)	(0.008–0.163)
Pregnenolone	1.29	1.16	1.78	0.67 *	0.96
	(0.84–1.80)	(0.56–1.75)	(0.16–4.09)	(0.46–0.98)	(0.62–1.25)
17-Hydroxyprogesterone	0.32	0.44	0.36	0.24	0.33
	(0.17–0.82)	(0.24–0.78)	(0.17–1.54)	(0.15–0.62)	(0.24–0.50)
Progesterone	0.09	0.09	0.09	0.02 *†	0.03 *†
	(0.07–0.17)	(0.07–0.13)	(0.02–0.21)	(0.00–0.07)	(0.01–0.06)
DHEA	2.33	2.05	0.59 *†	0.22 †	0.07 *†
	(1.48–3.36)	(1.18–3.96)	(0.38–0.95)	(0.02–1.24)	(0.02–1.30)
DHEA-S	1340	1010	204 *†	371 *†	267 *†
	(794–1830)	(621–1895)	(103–419)	(193–939)	(96–664)
Androstenedione	0.82	0.83	0.49	0.50 *†	0.50*
	(0.61–1.13)	(0.55–1.26)	(0.33–1.21)	(0.28–0.91)	(0.27–0.81)
Testosterone	0.31	0.36	0.16 *	0.18 *	1.25
	(0.19–3.55)	(0.17–2.98)	(0.10–0.49)	(0.08–0.52)	(0.17–3.33)

Ages are shown as medians and ranges and all plasma concentrations as medians and interquartiles.

^*P<0·05, different from reference;

^†P<0·05, different from excluded.

Germline variants in ARMC5

Twelve (33.3%) subjects in 34 unrelated and two related individuals carried germline variants in ARMC5 in a heterozygote state that were either disease causing or predicted damaging (Table 2). Two related subjects had frameshift variants p.C579Sfs*50 (c.1736_1739delGCCT). Five missense variants predicted as damaging were identified in 6 subjects (Table 2); three were nonsense, two with p.R173* (c.517C>T), and one with p.R364* (c.1090C>T). The last four were missense and resulted in amino acid substitutions (Table 2). All of these variants have been previously reported in the literature.(Assie et al. 2013; Faucz et al. 2014; Zilbermint, et al. 2015) There were no novel variants in ARMC5 identified in this study.

Table 2.

List of our study subjects with damaging germline variants in ARMC5 causing primary bilateral macronodular adrenocortical hyperplasia and their allele frequency (minor allele) and in silico modeling.

Subjects (N)	DNA change	Protein change	SNP id	Domains	In silico modeling
					Prediction	Scorea	Frequency in the NIH cohort	Allele frequencyb
1	c.466C>T	p.L156F	rs114930262	Armadillo	Possibly damaging	0.527	1.61%	0.16%
2	c.517C>T	p.R173*	no ID	Armadillo	Disease causing		3.23%	N/F
1	c.1084C>T	p.R362W	no ID	Armadillo	Disease causing	1.000	1.61%	0.003%
1	c.1090C>T	p.R364*	no ID	Armadillo	Disease causing		1.61%	N/F
2	c.1223A>G	p.Q408R	rs141923065	Armadillo-like helical	Disease causing	0.990	3.23%	0.31%
2	c.1736_1739delGCCT	p.C579Sfs*50	no ID	Armadillo-like helical	Disease causing		1.61%c	N/F
1	c.1827_1828dup	p.A610Vfs*21	no ID	Armadillo-like helical	Disease causing		1.61%c	N/F
1	c.1928 C>T	p.T643M	rs370836071	Armadillo-like helical	Disease causing	1.000	1.61%	0.006%
1	c.2692C>T	p.R898W	rs587777659		Possibly damaging	0.527	1.61%	0.001%

N/F - no frequency (variant was neither found in gnomAD nor in 1000G);

^aPolyPhen-2 was used as standard. Scores goes from 0.000 to 1.000. Greater score indicates higher probability to impair the protein function. The main factors considered for the calculation of the score are: 1) difference in the thermo-physical properties of the wild type and mutant protein, and; 2) evolutionary preservation of the residue in the corresponding position.

^bBased in the gnomAD database;

^cThis variant was found in two subjects; the frequency was counted as 1.

Plasma steroid profiles

Numerous differences in plasma steroids were observed among the four groups of the study population as initially assessed using the Steel Dwass all-pairs non-parametic test for multiple comparisons (Table 1). However, due to group-variations in age and sex, such differences were further clarified by multivariate analyses taking into account those variables as covariates. The additional comparisons of geometric least square means (Figure 1) mostly confirmed the differences in steroid profiles between groups (Table 1).

Figure 1.

Least square means (with standard errors), as corrected for age and gender and derived from logarithmically transformed data, for plasma concentrations of 12 of the 16 steroids in plasma panel in the control group, subjects with ACS and subjects with PBMAH with and without damaging germline variants in ARMC5. * P<0·05 different from control; † different from ACS; § P<0·05 different from subjects with PBMAH without damaging germline variants in ARMC5.

The steroids showing the most dramatic differences between subject groups included 11-deoxycortisol, corticosterone, 11-deoxycorticosterone, 18-hydroxycortisol, progesterone and the androgens, but particularly DHEA and DHEA-S (Figure 1). Steroids that showed isolated differences included cortisol and 18-oxocortisol with higher (P<0·05) concentrations in subjects with ACS than in controls. Aldosterone showed higher concentrations in subjects with PBMAH and absence of germline variants in ARMC5 than controls. Progesterone was lower (P<0·05) in PBMAH subjects with and without germline variants in ARMC5 when compared to controls, whereas the difference for subjects with ACS did not reach significance. In contrast, cortisone showed no differences between groups. Similarly there were no significant differences between groups for 21-deoxycortisol, pregnenolone and 17-hydroxyprogesterone (data not shown). After correction for gender, plasma testosterone was higher (P<0·05) in controls than other subject groups (data not shown).

The androgens, androstenedione, DHEA and DHEA-S, were all lower in the three subject groups than in controls (Figure 1). There were however, some isolated differences: DHEA showed additional lower (P<0·05) concentrations in subjects with PBMAH and germline variants in ARMC5 than ACS, whereas DHEA-S showed lower (P<0·05) concentrations in subjects with PBMAH and germline variants in ARMC5 as well as ACS and compared to subjects with PBMAH without germline variants in ARMC5. Although there were trends towards larger increases of cortisol, 11-deoxycortisol and 18-oxocortisol above controls for subjects with ACS than PBMAH and the reverse for corticosterone, the differences between subjects with CS and PBMAH did not reach significance.

Aldosterone and CYP11B1/B2 expression

Aldosterone was significantly increased only in subjects with PBMAH who had no damaging germline variants in ARMC5. Therefore, to determine whether this difference was directly associated with the regulation of CYP11B1 and CYP11B2 expression, we analyzed the immunohistochemistry patterns of ARMC5, CYP11B1 and CYP11B2 in two resected PBMAH samples from subjects with damaging germline variants in ARMC5 (p.R173*, p.R364*) in comparison to two subjects without damaging germline variants (one without a germline variant in ARMC5, and one with a synonymous variant (p.L614L) considered as benign). As previously described,(Faucz et al. 2014) ARMC5 expression was decreased in PBMAH harboring a nonsense variant compared to PBMAH samples without damaging germline variants in ARMC5 (Figure 3). Although CYP11B1 expression was not directly correlated with ARMC5 status, CYP11B2 expression was lower in the two PBMAH samples with nonsense germline variants in ARMC5 when compared to those without a germline variant in ARMC5 (Figure 3), which could explain the isolated increase in aldosterone in the latter group.

Figure 3.

Discriminant analysis using a panel of nine steroids (DHEA, 11-deoxycortisol, 18-oxocortisol, corticosterone, aldosterone, progesterone, DHEA-S, 11-deoxycortisosterone, 18-hydroxcortisol) that provided optimal discrimination of the four patient groups: (i) control, (ii) subjects with ACS, (iii) subjects with PBMAH without damaging germline variants in ARMC5, (iv) subjects with PBMAH and damaging germline variants in ARMC5. Three dimensional canonical plots are shown in panel A, ROC curves with areas under curves are shown in panel B, and confusion matrices for classification of predicted versus actual groupings are shown in panel C.

Logistic regression and discriminant analysis

Logistic regression analysis indicated four steroids, DHEA, 11-deoxycortisol, 18-oxocortisol, and corticosterone offering the most power to distinguish controls from subjects with ACS or PBMAH with and without germline variants in ARMC5. Thereafter in order of significance, aldosterone, progesterone, DHEA-S, 17-hydroxyprogesterone, 11-deoxycortisosterone, 18-hydroxcortisol and cortisol provided additional resolving power. Together, this combination of 11 steroids indicated areas under receiver operating characteristic (ROC) curves from 0·9733 to 0·9812. Nevertheless, restricting the panel to just the first four top steroids indicated areas under ROC curves of 0·9140 to 0·9617.

Discriminant analyses with step-wise variable selection indicated correct classification of 95·2% of all subjects of the four groups using a limited panel of nine steroids: DHEA, 11-deoxycortisol, 18-oxocortisol, corticosterone, aldosterone, progesterone, DHEA-S, 11-deoxycorticosterone and 18-hydroxycortisol (Figure 3). With this panel almost all subjects with ACS could be distinguished from those with PBMAH and < 5% of the control group was misclassified as having ACS or PBMAH. Moreover, this panel allowed correct classification of subjects with and without germline variants in ARMC5 for 91·7% of all subjects with PBMAH. Overall diagnostic sensitivities varied from 83·3% to 95·8% at specificities of 97·8% to 99·7%.

Discussion

The present study extends on the simple, fast and versatile use of LC-MS/MS-based steroid profile, allowing simultaneous plasma measurement of 16 adrenal steroids to help establish a distinct steroid metabolome in subjects with PBMAH that clinically and biochemically manifest with subclinical or overt glucocorticoid excess. Our study is aimed at accurately characterizing the plasma steroid profiles in these subjects and their putative correlation to disease phenotype and status of damaging germline variants in ARMC5, to provide a new tool for disease diagnosis and stratification. The perspective would be to use it as a complement to the conventional immunoassay measurements of cortisol and 24-hr urinary 17-hydroxysteroids. Furthermore, we confirm that this subset of PBMAH is a biochemically distinct adrenocortical disease, with a single plasma steroid profile, with high overall diagnostic sensitivity and specificity, making isolated plasma cortisol measurement obsolete in this indication.

The use of LC-MS/MS–based steroid profiling is promising in evaluating common and rare adrenal disorders (Arlt et al. 2011; Eisenhofer et al. 2016; Eisenhofer et al. 2017a; Eisenhofer et al. 2017b). This technique ensures detection of multiple compounds, particularly those in low concentrations, with a high specificity, allowing the separation of compounds with high structural similarities, and therefore subject to cross-reactions in traditional immunoassays, particularly 21-deoxycortisol, corticosterone and 11-deoxycortisol. However, this technique has not yet been used in evaluating ultra rare adrenocortical hyperplasias, such as PBMAH. Thus, we chose to evaluate this technology in subjects with PBMAH, a unique cause of ACTH-independent CS where abnormal regulation of steroidogenesis can be mediated by aberrant adrenal expression of several hormone receptors (G-protein-couple receptors or GPCRs) while fixing their natural ligands (Bourdeau, et al. 2016; Hsiao et al. 2009; St-Jean, et al. 2018). This unique pathophysiological mechanism leads to abnormal steroid production and regulation of cortisol and its precursors, which has not been previously quantified using a robust steroid panel analysis (St-Jean et al. 2018). Historically, abnormal steroidogenesis in PBMAH was quantified using a 24-hr urinary 17-hydroxysteroids measurement, which is a sensitive screening test for PBMAH (Hsiao et al. 2009). However, this test lacks specificity for the diagnosis, and is cumbersome to perform in clinical practice, and limited to a small number of specialized laboratories (Hsiao et al. 2009). Additionally, reliance on the 24-hr urinary free cortisol and other measures of adrenocortical activity including salivary cortisol or plasma cortisol post overnight dexamethasone suppression test is plagued by intra and inter individual variability, collection errors, and can be misleadingly normal in PBMAH as most patients present with no to mild hypercortisolism (normal to mildly elevated 24-hr UFC) (Albiger et al. 2017; Hsiao et al. 2009). Thus, a robust single-test alternative for screening purpose is required in evaluating PBMAH.

Damaging variants in ARMC5 are frequently present in PBMAH and found at the germline and somatic level (Assie et al. 2013). Most variants are frameshift and/or nonsense, leading to loss of function (Espiard et al. 2015). To date, the exact function of ARMC5 is not known. In this study, we identified twelve (33.3%) subjects in 34 unrelated and two related individuals with predicted damaging germline variants in ARMC5, in keeping with data from the literature (Albiger et al. 2017; Assie et al. 2013; Bourdeau et al. 2016; Gagliardi et al. 2014; Yu et al. 2018). In our analysis, progesterone was lower (P<0·05) in PBMAH when compared to controls, and this association was independent of germline ARMC5 status. Moreover, DHEA-S showed lower (P<0·05) concentrations in subjects with ACS and PBMAH harboring a germline variant in ARMC5 when compared to subjects with PBMAH without germline variants in ARMC5, in keeping with internodular atrophy from ACTH suppression.

Surprisingly, higher concentrations of aldosterone were observed in the PBMAH group with negative germline variants in ARMC5 when compared to controls and the largest differences between PBMAH and other causes of ACS were found for corticosterone, with a similar trend for 18-hydroxycortisol and aldosterone. In this group, CYP11B2 expression was increased (Figure 3). These results support our previous findings that CYP11B2 expression is significantly decreased by ARMC5 knock down in H295R cells lines (Zilbermint et al. 2015). In mice, Armc5^+/− develop normally but at the age of 1 year show a transient decrease of corticosterone followed by normalization and hypercorticosteronemia suggesting an important role for this gene in early mouse embryonic development and steroidogenesis (Berthon, et al. 2017b). The decreased CYP11B2 expression in our analysis may, then, be an in situ signature, which contributes to the unique steroidomics profile in PBMAH with damaging variants in ARMC5, although the exact mechanism is yet to be identified. In the contrary, we previously described missense variants in ARMC5 in 10.7% of subjects with primary aldosteronism (Zilbermint et al. 2015). The function of ARMC5 in adrenocortical cells appears to be tightly dependent of ARMC5 protein dosage as well as the origin of cells developing the hyperplasia.

Our study has few limitations. First, we compared our PBMAH cohort with a European reference data, which might not be generalizable to various ethnicities with PBMAH. Moreover, we included the less common subset of PBMAH, that manifests with mild to overt hypercortisolism with low to suppressed ACTH. Most cases of PBMAH are discovered incidentally on imaging without clinical or biochemical CS; therefore, future studies incorporating these subjects would provide an important perspective on steroidonomics, including the sequence of abnormalities in various disease stages and ARMC5 status. Second, although the control populations were selected to match the combined subject populations for age and gender, there was significant variation between the three subject groups. Thus, we analyzed these groups considering both age and gender as covariates and used age- and gender specific reference intervals for data normalization before discriminant analysis. With these approaches, any impact of differences in age and gender should have been minimal. Third, we did not analyze somatic variants in ARMC5 to fully characterize this condition in adrenal tissue; however, approximately half of our subjects did not undergo adrenalectomy and all of our subjects with PBMAH met the biological, radiographic and pathologic criteria, which was sufficient. Fourth, we were unable to compare the performance of this steroid panel with immunoassay measurements of cortisol, aldosterone and 17-hydroxysteroids. Finally, we did not perform steroid profiling on samples, following a systematic in vivo evaluation of expression of various aberrant GPCRs in PBMAH, which would have likely yielded exaggerated steroidogenic responses, something that could be explored in future studies.

In summary, LC-MS/MS serum steroid profiling offers a potentially important advance in the clinical workup of patients with PBMAH that may also assist in prioritizing germline ARMC5 testing. Subjects with PBMAH show distinctive plasma steroid profiles that may offer an additional single-test alternative for screening purposes.

Figure 2.

Comparison of CYP11B1 and CYP11B2 expression between PBMAH with and without damaging germline variants in ARMC5. Histology and immunohistochemistry of ARMC5, CYP11B1 and CYP11B2 in one PBMAH without a damaging germline variant in ARMC5, one with a synonymous variant (p.L614L) considered as benign and two nonsense variants (p.R173*, p.R364*) considered damaging.