Genetics of aldosterone-producing adenomas with pathogenic KCNJ5 variants

Abstract

Somatic variants in genes that regulate intracellular ion homeostasis have been identified in aldosterone-producing adenomas (APA). Although the mechanisms leading to increased aldosterone production in APA cells has been well studied, the molecular events that cause cell proliferation and tumor formation are poorly understood. In the present study, we have performed whole exome sequencing (WES) to characterize the landscape of somatic alterations in a homogeneous series of APA with pathogenic KCNJ5 variants. In the WES analysis on eleven APA, 84 exonic somatic events were called by 3 different somatic callers. Besides the KCNJ5 gene, only two genes (MED13 and ZNF669) harbored somatic variants in more than one APA. Unlike adrenocortical carcinomas, no chromosomal instability was observed by the somatic copy-number alteration and loss of heterozygosity analyses. The estimated tumor purity ranged from 0.35 to 0.67, suggesting a significant proportion of normal cell infiltration. Based on the results of PureCN analysis, the KCNJ5 variants appear to be clonal. In conclusion, in addition to KCNJ5 somatic pathogenic variants, no significant somatic event that would obviously explain proliferation or tumor growth was observed in our homogeneous cohort of KCNJ5-mutated APA. The molecular mechanisms causing APA growth and tumorigenesis remain to be elucidated.

INTRODUCTION

Primary aldosteronism (PA) has been recognized as a major cause of endocrine and drug-resistant hypertension that is associated with significant morbidity and mortality (Funder, et al. 2016). Aldosterone-producing adenoma (APA) is one of the major subtypes of PA (Young 2007). Over the past decade, there has been a considerable focus on defining the genetic bases of excess aldosterone production in PA. After the initial identification of somatic and germline KCNJ5 variants in PA patients (Choi, et al. 2011), different germline and somatic variants have been identified in familial and sporadic forms of PA (Monticone, et al. 2018; Zennaro, et al. 2017). The immediate functional consequences of these alterations are disruption of intracellular ion homeostasis, which ultimately lead to increased aldosterone synthase (CYP11B2) transcription, and inappropriate aldosterone production (Monticone et al. 2018; Zennaro et al. 2017). KCNJ5 mutations are the most common somatic variants in APA with greater frequency in East Asian populations (Zennaro et al. 2017). However, the mechanism by which these variants provide growth advantage to adrenocortical cells and promote APA formation remains poorly understood.

Recent studies have demonstrated that in patients with PA and multiple adrenal nodules, somatic variants in different aldosterone-driver genes can be present in distinct CYP11B2-expressing, but not in CYP11B2-negative nodules within the same adrenal (Dekkers, et al. 2014; Fernandes-Rosa, et al. 2015; Nanba, et al. 2016). Additionally, in-vitro studies indicate that mutated KCNJ5 channels do not confer growth advantage when expressed in the human adrenocortical HAC15 cell line (Oki, et al. 2012). These observations suggest that the processes of nodule formation and autonomous aldosterone production may be independent events, the so-called “two-hit” model (Zennaro et al. 2017). According to this model, an initial somatic event that confers growth advantage would induce tumor formation, and a second mutational event (in any of the already characterized genes associated with APA) would be responsible for increased aldosterone production. However, no recurrent secondary mutations in growth-promoting genes have been reported in APA with mutations in aldosterone-driver genes so far. In this study, we used whole-exome sequencing (WES) to characterize the landscape of somatic alterations in a homogeneous series of APA with KCNJ5 variants. We were particularly interested in identifying novel recurrent mutations and somatic alterations that may contribute to APA tumorigenesis, and to further characterize KCNJ5-mutated tumors in terms of tumor purity, ploidy, and clonal composition.

PATIENTS AND METHODS

Patients and samples

Fourteen APA from fourteen patients with PA who underwent adrenalectomy at the University of Michigan or Yokohama Rosai Hospital were included in the present study. To study genetic alterations related to APA tumorigenesis, relatively large APA (median: 30 mm, interquartile range: 23.8–30.0 mm) were studied. Clinical characteristics of studied patients are shown in Supplementary Table 1. For somatic variant identification, control germline DNA either from peripheral blood (n=5) or matched adjacent normal adrenal (n=6) was used based on the availability. Peripheral blood or matched adjacent normal adrenal and APA tissue DNA was isolated using commercial kits (Nucleospin Blood [Macherey-Nagel], Nucleospin Tissue [Macherey-Nagel], or AllPrep DNA/RNA/Protein kit [QIAGEN]). This study was approved by Institutional Review Board of the University of Michigan and Yokohama Rosai Hospital. An informed consent was obtained from each individual patient after full explanation of the purposes of this research.

Immunohistochemistry

Formalin-fixed paraffin-embedded tissue blocks were available from all but one patient (Case 10, APA-Y182). Immunohistochemistry (IHC) was performed using specific antibodies against CYP11B2 (mouse monoclonal, Millipore, #MABS1251) (Gomez-Sanchez, et al. 2014), 11β-hydroxylase (CYP11B1, rat monoclonal) (Gomez-Sanchez, et al. 2017), 17α-hydroxylase (CYP17A1, mouse monoclonal) (Dharia, et al. 2004), 3β-hydroxysteroiddehydrogenase/Δ4–5 isomerase type II (HSD3B2, mouse monoclonal) (Gomez-Sanchez et al. 2017), and β-catenin (mouse monoclonal, BD Biosciences, #610153) essentially as described previously (Monticone, et al. 2012; Nanba et al. 2016).

Whole exome sequencing

DNA was fragmented to 300 bp using standard Covaris sonication (Covaris). Fragmented DNA was then prepared as a standard Illumina library using IntegenX (now Wafergen) reagents on the Apollo instrument, where the fragments were end repaired, A tailed, and adapter ligated. Then the samples were PCR amplified. Libraries were checked for quality and quantity by TapeStation (Agilent) and pooled for capture; 6 samples per pool. Each pool underwent exome capture using the Roche Nimblegen SeqCap EZ v3.0 using manufacturer’s protocols (Roche NimbleGen). In brief, a total of 1μg of prepared libraries was hybridized to the SeqCap EZ library for 72 hours, then washed and the captured DNA was recovered using Streptavidin Dynabeads (Life Technologies). The captured DNA was then amplified by PCR. The purified captured DNA was then clustered using the cBot (Illumina) and sequenced on a HiSeq 2500 (Illumina) high output paired end 125 bp run using version 4 reagents and manufacturer’s protocols.

Data analysis

Paired-end reads were aligned to the hg19 assembly of the human genome using bwa-mem (Li 2013). Aligned reads were sorted and duplicates were marked with bamsort and bammarkduplicate tools, respectively, from biobambam2 (Tischler and Leonard 2014). Indel realignment and quality score recalibration were performed with GATK (McKenna, et al. 2010). Coverage metrics and quality-control analysis of the alignments were performed with Qualimap2 (Okonechnikov, et al. 2016). As an additional quality-control measure, we used NGSCheckMate (Lee, et al. 2017) to rule out cross-contamination, sample mislabeling, and unexpected similarities among samples from different patients. Somatic variants were called with three different callers: MuTect2, Lancet, and LoFreq (Cibulskis, et al. 2013; Narzisi, et al. 2017; Wilm, et al. 2012). Somatic calls were annotated with Annovar (Wang, et al. 2010). We considered somatic variants all those called by at least two of the callers. We used the Integrative Genome Viewer (IGV) to visually confirm the somatic alterations and the reported allele frequencies. We compared the final list of somatic variants to an internal database generated from other projects using our in-house pipeline to further reduce the possibility of calling recurrent sequencing artifacts. For samples which paired germline DNA were not available, we calculated the allele frequencies of pathogenic KCNJ5 variants from the VCF files generated with samtools mpileup/bcftools (Danecek, et al. 2011; Li, et al. 2009). Finally, we used PureCN (Riester, et al. 2016) perform somatic copy-number alteration (SCNA), loss of heterozygosity (LOH) analyses, and to estimate tumor purity, ploidy, and clonal composition. PureCN infers tumor purity and ploidy by simultaneously analyzing depth-of-coverage and variant allele frequencies from germline and tumor DNA.

RESULTS

Histologic characteristics of KCNJ5-mutated APA

In accordance with previous studies (Azizan, et al. 2012; Monticone, et al. 2015), the KCNJ5-mutated tumor cells were mainly composed of lipid-rich clear cells. IHC was performed to assess tumor expression of CYP11B2, which is essential for aldosterone biosynthesis, and to confirm the tumor cell capacity to produce aldosterone (Figure 1A and 1C). IHC for other steroidogenic enzymes, including CYP17A1, CYP11B1, and HSD3B2 was also performed (n=13). The CYP11B2 expression intensity was variable and most commonly exhibited a heterogeneous expression pattern (Supplementary Table 2). APA tumor cells also showed strong expression of HSD3B2 (Figure 1D). The expression of CYP17A1 and CYP11B1, which are required for cortisol synthesis, was also observed in KCNJ5-mutated APA with some degree of variability (Figure 1E and 1F, Supplementary Table 2). The results of 1mg dexamethasone suppression test were available in 10 cases (Supplementary Table 1). Of those, three cases showed non-suppressed cortisol after a dexamethasone suppression test (serum cortisol levels higher than 1.8 μg/dL). The APAs from those patients exhibited positive but variable expression of CYP17A1 and CYP11B1. We further performed IHC for β-catenin to determine a potential contribution to tumorigenesis or growth. The β-catenin IHC was not successful in one case (Case 1). All other cases (n=12) showed similar patterns of predominantly membranous staining, indicating the absence of Wnt/β-catenin activation in the majority of the tumor cells (Supplementary Figure 1). Only a few cases showed focal nuclear and/or cytoplasmic reactivity, which did not reach the staining intensity, usually observed in adrenocortical carcinomas with genetic changes known to activate the Wnt/β-catenin pathway.

Figure 1:

Representative histologic characteristics of KCNJ5-mutated aldosterone-producing adenoma. A. Low magnification view of the adrenal gland with CYP11B2 IHC (Case 5, APA-Y80). The APA was identified by CYP11B2 IHC. Scale bar, 5mm. B-F. High magnification view of the APA, B. H&E, C. CYP11B2 IHC, D. HSD3B2 IHC, E. CYP17A1 IHC, F. CYP11B1 IHC. Scale bars, 100 μm. APA, aldosterone-producing adenoma; IHC, immunohistochemistry; H&E, hematoxylin and eosin staining.

Somatic mutation analysis

We performed WES on 11 APA with paired germline DNA and 3 unpaired APA samples. The mean coverage and the percentage of targeted regions that were covered at least 20x were 90.45x (range 67.47 – 118.95x) and 91.12% (range 88.59 – 92.81%) for tumors, and 89.45x (range 63.46 – 120.11x) and 91.45% (range 89.07 – 92.65%) for germline. Our NGSCheckMate analysis ruled out cross contamination, sample mislabeling, and unexpected similarities among samples from different patients (Supplementary Figure 2). We identified 120 exonic somatic events that were called at least by 2 callers, including 108 single nucleotide variants (SNVs) (22 synonymous, 81 nonsynonymous, and 6 stopgain), 6 indels (4 frameshift deletions, 1 frameshift insertion, and 1 nonframeshift insertion), and 5 splicing site variants, averaging 10.9 events per sample (range 5–16). Among these, 84 events were called by the three callers, 32 were called by Mutect2 and Lofreq, 3 by Mutect2 and Lancet, and 1 by Lancet and LoFreq (Supplementary Figure 3). Apart from KCNJ5 variants, which were detected by the three callers in all samples, only two other genes had variants in more than one APA: MED13 (NM_005121:exon26:c.T5866G:p.Ser1956Ala, [APA-78 and APA-86]) and ZNF669 (NM_001142572:exon4:c.A1052C:p.Glu351Ala, [APA-78 and APA-86]) (Supplementary Figure 4–7). Of note, the MED13 and ZNF669 variants were not confirmed by direct Sanger sequencing likely due to the low variant frequencies.

Among the identified mutated genes, 6 of them have been reported as cancer-associated (Futreal, et al. 2004) (COSMIC v85): JAK1, NAB2, FOXP1, ATM, MEN1, and USP6. Mutations in these genes have been shown to affect cellular processes that are frequently dysregulated in cancer and are thought to be tumor promoting. Additionally, we identified variants in PDE11A and CYP17A1, genes that are essential components of the adrenal steroidogenesis machinery and have been implicated in other adrenal diseases, such as forms of adrenal micronodular hyperplasia, cortisol-producing adenomas (CPA), and congenital adrenal hyperplasia. Finally, we identified variants in other ionic channels such as CACNA1E and KCNA1 (Supplementary Table 3–6). Noteworthy, exome sequencing confirmed the presence of four previously identified KCNJ5 variants in out cohort: c.T503G (p.L168R), c.G451A (p.G151R), c.G451C (p.G151R), and c.477_478insATT (p.T149delinsTI) with allele frequencies ranging from 14–48% (Table 1).

Table 1.

Genetic characteristics of KCNJ5-mutated aldosterone-producing adenoma

Case	Sample name	Tumor purity	Tumor ploidy	KCNJ5 mutation	KCNJ5 mutation allele frequency	# of mutated alleles per cell	Clonality
1	APA-59*	0.57	2.00	p.G151R	0.44	1	clonal
2	APA-62*	0.64	2.00	p.L168R	0.29	1	clonal
3	APA-78*	0.67	2.00	p.L168R	0.37	1	clonal
4	APA-86*	0.64	2.00	p.T149delinsTI	0.34	1	clonal
5	APA-Y80*	0.36	2.00	p.L168R	0.14	1	clonal
6	APA-Y108#	0.59	1.98	p.G151R	0.26	1	clonal
7	APA-Y110	NA	NA	p.G151R	0.24	NA	NA
8	APA-Y179#	0.55	2.00	p.G151R	0.31	1	clonal
9	APA-Y180	NA	NA	p.L168R	0.48	NA	NA
10	APA-Y182*	0.35	1.99	p.L168R	0.26	1	clonal
11	APA-Y183#	0.55	2.02	p.L168R	0.18	1	clonal
12	APA-Y184	NA	NA	p.G151R	0.22	NA	NA
13	APA-Y189#	0.59	2.00	p.L168R	0.35	1	clonal
14	APA-Y198#	0.58	2.00	p.G151R	0.40	1	clonal

^*DNA from matched adjacent normal was used as a control.

^#peripheral blood DNA was used as a control.

NA, not available

SCNA and LOH analyses

In agreement with previous study (Choi et al. 2011), our SCNA and LOH analyses revealed that unlike a subset of ACA with CTNNB1 mutations and adrenocortical carcinomas (ACC) (Goh, et al. 2014; Zheng, et al. 2016), KCNJ5-mutated APA did not exhibit chromosomal instability (characterized by arm-level segmental loss and gains, and/or alterations in ploidy). The diploid heterozygous state was maintained in all samples (Table 1, Supplementary Figure 8).

Tumor purity and clonal composition

KCNJ5-mutated APA appear to bear a significant proportion of normal cell infiltration. Tumor purity in our cohort ranged from 0.35 – 0.67, which contrasts with ACC that exhibit high purity according to a recent molecular profiling study (Zheng et al. 2016). Furthermore, PureCN analysis also demonstrated that KCNJ5 variants are heterozygous and likely clonal in all tumors, suggesting that these variants are early events in adenoma formation (Table 1, Supplementary Figure 8).

DISCUSSION

In recent years, high-throughput molecular profiling studies have identified recurrent variants targeting different regulators of intracellular calcium homeostasis, including KCNJ5, ATP1A1, ATP2B3, and CACNA1D in over half of APA. Despite the well-characterized effects of these variants in abnormal CYP11B2 expression and aldosterone production, the contribution to cell growth remain poorly understood (Zennaro et al. 2017). So far, a growth-promoting effect of KCNJ5 variants has not been demonstrated in-vitro studies using HAC15 cells (Oki et al. 2012). Although it is difficult to assess a cell proliferation effect of these variants using highly proliferative H295R/HAC15 cells which have a β-catenin mutation (Tissier, et al. 2005), these in vitro studies suggest that alternative mechanisms might play a role in tumor formation. Interestingly, areas of cortical cell hyperplasia and other nodular lesions are frequently observed near APA. These areas are usually CYP11B2-negative and do not bear aldosterone-driving gene variants (Dekkers et al. 2014). Furthermore, APA often exhibit a heterogeneous immunostaining for CYP11B2 (Nakamura, et al. 2014), suggesting that mutated cells could be subclonal. Based on these observations, a “two-hit” model in which tumor growth and increased aldosterone production are caused by independent molecular hits has been proposed (Zennaro et al. 2017). In fact, we observed at least two variants of uncertain significance in genes that have been previously associated with adrenocortical tumorigenesis: the MEN1 NM_130804:exon3:c.T364C (p.Ser122Pro) and the PDE11A NM_001077197:exon5:c.C514T (p.Arg172Cys) variants, both predicted to be pathogenic by SIFT, polyphen2, and CADD. However, both of these variants were somatic and, according to our PureCN analysis, none of these variants are in focal areas of LOH, suggesting that both are “passenger” events. Although somatic variants in MED13 and ZNF669 were identified in more than one APA, the pathological role of these variants is unclear. MED13 is a component of the Mediator Complex, a multiprotein complex critical for transcriptional activation. Mutations and altered expression of different Mediator components have been described in rare neurodevelopmental disorders and in a broad spectrum of benign and malignant tumors (Clark, et al. 2015). The function and physiological relevance of ZNF669 are currently unknown. Noteworthily, we did not identify variants in other genes previously described in APA, such as CTNNB1 (encoding β-catenin), suggesting a dominant role for KCNJ5 variants. These results are consistent with previous studies on APA genetics, in which KCNJ5 and CTNNB1 variants were mutually exclusive (Akerstrom, et al. 2016; Wu, et al. 2017). Previous reports have demonstrated that the activation of Wnt/β-catenin pathway is frequent in APA regardless of the mutation status of CTNNB1 (Berthon, et al. 2014; Boulkroun, et al. 2011). However, the nonactived state of β-catenin was predominantly observed in our cohort, suggesting that, unlike ACC (Mete, et al. 2018), activation of Wnt/β-catenin pathway may not be a common feature of KCNJ5-mutated APA. For better understanding of the role of Wnt/β-catenin signaling pathway activation in APA development, further analysis including APA with variants in other aldosterone-driver genes, such as CACNA1D, ATP1A1, and ATP2B3 will be needed.

Finally, our clonality analysis suggests that KCNJ5-mutated cells are a clonal population within the tumor rather than a subclone that acquired the ability to secrete aldosterone. Certain KCNJ5 variants, including p.G151R and p.T158A appear to have potential impact on cell proliferation considering the fact that these variants have been identified both in sporadic APA (as somatic variants) and in patients with massive hyperplasia (as germline variants), whereas a KCNJ5 variant p.G151E has only been documented in individuals without massive adrenal hyperplasia as a germline variant (Choi et al. 2011; Mulatero, et al. 2012; Scholl, et al. 2012). Furthermore, in an early-onset PA patient, the identical somatic KCNJ5 p.G151R variant has been identified in multiple CYP11B2-expressing hyperplastic lesions and APA in bilateral adrenals and the variant has not been observed in non-hyperplastic lesions (Tamura, et al. 2017), supporting the potential role of the KCNJ5 variant on cell proliferation.

Taken together, our results do not support the proposed “two-hit” model. However, since our study included only a homogeneous population of APA with KCNJ5 variants, the findings cannot be generalized. It is also important to note that the use of DNA from adrenal tissue adjacent to the tumors could have limited detection of somatic mutations considering a previous observation that small nodules are sometimes present in adrenal tissue adjacent to APA (Boulkroun, et al. 2010). The development of APA might be complex and heterogeneous. A recent case report describes the co-occurrence of a germline APC variant and a somatic KCNJ5 variant in a young PA patient with multinodular adrenal hyperplasia (Vouillarmet, et al. 2016). In the resected adrenal from the patient, the KCNJ5 variant was identified only in a CYP11B2-expressing nodule, whereas biallelic APC inactivation due to LOH was observed in both CYP11B2-positive and negative nodules (Vouillarmet et al. 2016), indicating “two-hit” model as a possible mechanism of APA development at least for some cases. Epigenetic event may also be involved in adrenal nodule formation in some APA. A recent study of integrated transcriptome and methylome analysis has demonstrated global hypomethylation in APA and up-regulated genes with CpG hypomethylation in APA, including five potentially tumorigenesis-related genes (Murakami, et al. 2015).

Interestingly, while our analysis is consistent with a relatively high proportion of normal cells in the tumors, we do not observe a significant infiltration of fibroblasts or inflammatory cells. In fact, although CYP11B2 expression is heterogeneous in some tumors, the majority of the cells, including CYP11B2-negative cells, exhibit morphological features consistent with steroidogenic cells and express steroidogenic enzymes such as CYP17A1, HSD3B2, and CYP11B1 (Figure 1). This observation suggests that the normal cell type that infiltrates KCNJ5-mutated APA are normal steroidogenic cells, consistent with a role for paracrine factors in mediating tumor growth.

In conclusion, our study did not identify recurrent somatic events that could explain proliferation or tumor formation in APA with KCNJ5 variants. The mechanism by which KCNJ5-mutated tumors form and grow remains elusive.