RESEARCH

Adrenal tumors provide insight into the role of cortisol in NK cell activity

Andrew E Greenstein1, Mouhammed Amir Habra2, Subhagya A Wadekar1 and Andreas Grauer1

1Corcept Therapeutics, Menlo Park, California, USA

2Department of Endocrine Neoplasia and Hormonal Disorders, MD Anderson Cancer Center, Houston, Texas, USA

Correspondence should be addressed to A E Greenstein: agreenstein@corcept.com

Abstract

Elevated glucocorticoid (GC) activity may limit tumor immune response and immune checkpoint inhibitor (ICI) efficacy. Adrenocortical carcinoma (ACC) provides a unique test case to assess correlates of GC activity, as approximately half of ACC patients exhibit excess GC production (GC+). ACC multi-omics were analyzed to identify molecular consequences of GC+ and assess the rationale for combining the glucocorticoid receptor (GR) antagonist relacorilant with an ICI. GC status, mRNA expression, and DNA mutation and methylation data from 71 adrenal tumors were accessed via The Cancer Genome Atlas. Expression of 858 genes differed significantly between GC- and GC+ ACC cases. KEGG pathway analysis showed higher gene expression of three pathways involved in steroid synthesis and secretion in GC+ cases. Fifteen pathways, most related to NK cells and other immune activity, showed lower expression. Hypomethylation was primarily observed in the steroid synthesis pathways. Tumor-infiltrating CD4+ memory (P = 0.003), CD8+ memory (P < 0.001), and NKT-cells (P = 0.014) were depleted in GC+ cases; tumor- associated neutrophils were enriched (P < 0.001). Given the pronounced differences between GC+ and GC- ACC, the effects of cortisol on NK cells were assessed in vitro (NK cells from human PBMCs stimulated with IL-2 or IL-12/15). Cortisol suppressed, and relacorilant restored, NK cell activation, proliferation, and direct tumor cell killing. Thus, GR antagonism may increase the abundance and function of NK and other immune cells in the tumor microenvironment, promoting immune response in GC+ ACC and other malignancies with GC+. This hypothesis will be tested in a phase 1 trial of relacorilant + ICI.

Key Words

adrenocortical carcinoma

natural killer cells

glucocorticoids

cortisol

Cushing syndrome

Endocrine-Related Cancer (2021) 28, 583-592

Introduction

Elevated glucocorticoid (GC) activity (hypercortisolism) has been implicated in the pathophysiology of multiple cancer types, including breast, ovarian, squamous cell and cervical cancer, and lymphomas (Mormont & Lévi 1997, Abercrombie et al. 2004, Palesh et al. 2008, Jehn et al. 2010, Sharma et al. 2018). Patients administered synthetic GCs (e.g. prednisone or dexamethasone) prior to immune checkpoint inhibitor (ICI) therapy are reported to experience worse outcomes across multiple oncology

indications (Habra et al. 2019, Head et al. 2019). Cortisol is a potent, abundant, and immunosuppressive endogenous GC and its target, the glucocorticoid receptor (GR), is expressed in nearly every cell in the body (Miller et al. 1998, Nicolaides et al. 2000). Thus, elevated endogenous GC may suppress immune function and ICI efficacy analogous to synthetic GC.

Adrenocortical carcinoma (ACC), a rare endocrine cancer, is often accompanied by excessive secretion of

Endocrine-RelatedA E Greenstein et al.Glucocorticoids affect ACC and28:8584
CancerNK cells

steroid hormones. Transformation of GC-producing cells in the adrenal gland is common and can lead to ACTH- independent constitutive GC production. Approximately half of ACC tumors are hormonally functional, with 43% producing cortisol, either alone or in combination with other hormones (Else et al. 2014). Public ACC multi- omics data include annotations of GC or other hormone production (Assié et al. 2014, Zheng et al. 2016). Thus, ACC provides a unique dataset in which transcriptional profiling, promoter methylation, and tumor mutation burden can be compared between cases with and without excess cortisol.

Therapies to reduce tumor burden for patients with ACC are limited and treatment outcomes are generally poor. GC excess (GC+) is associated with decreased overall survival and disease-free survival in patients with ACC (Else et al. 2014). Furthermore, patients with GC+ ACC show poor responses to ICI therapy (Habra et al. 2019, Raj et al. 2020). Akin to the diminished ICI response caused by synthetic GC, we hypothesize that ICI responses in GC+ ACC are diminished by endogenous GC. The underlying biology responsible for diminished ICI responses may be improved by GR antagonism. Further, understanding the molecular effects of GC in ACC may provide insight into the role of endogenous cortisol in other cancer types with elevated GC activity. Here, we present an ACC multi-omics analysis that identified molecular consequences of GC activity. A key insight from the in silico analysis, related to effects of GC excess on natural killer (NK) cells, was directly assessed in vitro: relacorilant (CORT125134, Corcept Therapeutics), an investigational selective GR modulator (SGRM) that antagonizes the GR, can reverse the effects of cortisol and restore NK cell activation, proliferation, and target cell killing. This work provides a rationale for combining relacorilant and ICI in the treatment of ACC.

Materials and methods

The Cancer Genome Atlas (TCGA) analysis

GC or other hormone status (based on clinical signs and symptoms or biochemical evidence), mRNA expression, DNA mutation, and DNA methylation data from distinct adrenal resections (n= 71) were accessed via TCGA (www. cancer.gov/tcga). The diagnosis of GC excess (GC+/-) or other hormone excess (H+/-) was provided by the participating sites from which each specimen originated. The diagnosis of GC excess (GC+/-) or other hormone excess (H+/-) was provided by the participating sites from

which each specimen originated to classify patients as having either pure hypercortisolism or mixed syndromes in the opinion of the investigator (Zheng et al. 2016). The criteria to diagnose cortisol production was not specified but likely incorporated results of late-night salivary cortisol, serum cortisol, dexamethasone suppression test, and/or 24-h urinary free cortisol (Gilbert & Lim 2008). Two sarcomatoid cases were excluded from the analysis. Using mRNA data in TCGA, genes that differed significantly (>two-fold change in expression and adjusted P ≤ 0.05) by general (i.e. estrogen and androgen) hormone or GC status were identified. Four different comparisons based on presence or absence of GC excess, presence or absence of general hormone excess, and presence of non- glucocorticoid hormone excess (NGC+) were performed. To identify the function of the genes affected by GC status, Kyoto Encyclopedia of Genes and Genomes (KEGG, RRID:SCR_012773) (Kanehisa & Goto 2000) pathway analysis was applied to the genes with >two-fold change and adjusted P ≤ 0.05. Unsupervised clustering (by gene expression alone) and supervised clustering (by gene expression and hormone status) clustering of two select KEGG pathways was performed to better understand the effects of GC excess on NK and T cells. Furthermore, 394,036 methylation probes were analyzed, and data were normalized using beta-mixture quantile normalization (BMIQ). To deconvolute immune cell type abundance, xCell was applied to the mRNA data (Aran et al. 2017). Mutations were called as previously described (Zheng et al. 2016).

Different tumor types available in TCGA were scored using a published GR activity signature (West et al. 2018). Random forests were then used to derive a new gene signature predictive of GC+ tumors. Signature genes were identified by bootstrapping random forests on random subsets comprising 80% of the data and comparing the mean bootstrapped importance of genes with a threshold value. We used bootstrapping to derive a probability distribution of gene importances, which we later used to estimate which features are the most predictive of the GC+/- patients. This method is appropriate with relatively small sample sets like this, and it provides more robust estimates of output parameters, such as feature importance. The threshold value was calculated by applying the same procedure to a random forest predicting randomized labels instead of the true GC+/- labels to simulate lack of signal. The 99.9th quantile of gene importance was selected as the threshold (0.0028). Based on the known distribution of GC+/- cases in ACC, a cutoff score of 0.75 was derived to distinguish GC+/-

Endocrine-RelatedA E Greenstein et al.Glucocorticoids affect ACC and28:8585
CancerNK cells

tumors when the gene signature is applied to other tumor types.

In vitro human NK-cell activation assays

Human NK cells were isolated from peripheral blood mononuclear cells (PBMCs) from three donors using NK cell isolation kits (Miltenyi Biotec, Bergisch Gladbach, Germany) per the manufacturer’s instructions. Ethically approved consent was received from the Queen Square Research Ethics Committee, UK (REC approval number 17/LO/0221) to use the human NK cells in this study. Specifically, the consent includes the use of leukocyte cones, under- or overweight whole blood packs, and buffy coats from anonymized blood donations, given to the National Health Service Blood and Transplant Services (NHSBT) and was obtained after a full explanation of the purpose and nature of the studies for which they were used. NK cells were stimulated for 5 days with 500 U/ml IL-2 ± cortisol and relacorilant. Activation and proliferation were assessed by fluorescence-activated cell sorting (FACS). Supernatants harvested on day 5 were analyzed for levels of IFNy, Granzyme A, and TNFa using a Luminex® kit (MilliporeSigma). RNA was isolated from the cells at day 5 using an RNeasy® RNA kit (Qiagen) according to the manufacturer’s instructions. Gene expression in cells harvested on day 5 was analyzed using an IO360™ panel (NanoString Technologies, Seattle, WA, USA). Significant changes in gene expression by each treatment condition were calculated in nSolver™ 4.0 (NanoString Technologies). After a list of significant genes was compiled, the list was filtered for genes with at least a two-fold change between stimulated without cortisol vs stimulated with cortisol. Data were visualized in GraphPad Prism (RRID:SCR_002798).

In vitro tumor cell killing assays with human NK cells

Human PBMCs were stimulated with 50 ng/ml IL-12+1 ng/ml IL-15 for 24 h ± cortisol and relacorilant. K562 tumor cells were added for a defined ratio, and tumor cell killing was assessed 4 h later by FACS with the eF780 Fixable Viability Dye (ThermoFisher).

Results

Significant differences in gene expression between GC+ and GC- ACC

An analysis of TCGA ACC mRNA data identified the absence vs presence of GC excess as affecting the largest number of

genes (858 genes) among the four comparisons performed (comparison 1, GC- vs GC+, in Fig. 1A and colored regions in Fig. 1B). Absence vs presence of any hormone led to a significant difference in 439 genes (comparison 2, H- vs H+). There was no significant difference between tumors without hormone excess and those with excess in non-GC hormones (comparison 3, H- vs NGC+). A comparison of non-GC vs GC excess tumors revealed 185 significantly different genes (comparison 4, NGC+ vs GC+).

Steroid synthesis pathways are elevated while immune pathways are suppressed in GC+ ACC

The function of the 858 genes that differed by GC status were examined using KEGG pathway analysis. KEGG pathways that were upregulated in GC+ cases included several steroid synthesis pathways (Fig. 1C). This confirmed that GC+ tumors have higher expression of GC-producing genes. The majority of pathways affected by GC excess, surprisingly, represented immune processes suppressed in the GC+ cases. Genes involved in NK cell activity, among other immune activities, were lower in GC+ tumors.

Lymphocyte gene suppression is associated with GC production

To better understand the effects of GC excess on NK and T cells, genes from two KEGG pathways (‘T-cell receptor signaling pathway’ and ‘natural killer cell-mediated cytotoxicity’) were assessed by GC status. Unsupervised clustering of normalized gene expression for these two KEGG pathways showed lower gene expression in GC+ cases (Fig. 2A). When clustering GC+ and GC- tumors separately, GC+ cases trended toward lower expression in these immune-related pathways (Fig. 2B). These visualizations confirm the impact of excess cortisol on the abundance of transcripts in each pathway.

Methylation and cellular infiltration contribute to gene expression differences

Differences in tumor gene expression can be caused by promoter methylation, altered cellular infiltrate, or other mechanisms. To understand the cause of these differences in ACC, first, differences in promoter methylation were analyzed. In GC+ ACC cases, many genes were significantly hypomethylated (Supplementary Fig. 1, light blue, see section on supplementary materials given at the end of this article), while fewer genes were hypermethylated (Supplementary Fig. 1, red). The hypomethylated genes

A Hormone status in TCGA ACC tumors (N=71)

Glucocorticoid absent (GC-, n=40)

Glucocorticoid present (GC+, n=31)

Glucocorticoid status Any hormone status

Any hormone absent (H-, n=25)

Any hormone present (H+, n=46)

Non-GC hormones present (NGC+, n=15)

Comparison 1:
GC- vs GC+858 significantly different genes
Comparison 2:
H- vs H+439 significantly different genes
Comparison 3:
H- vs NGC+O significantly different genes
Comparison 4:
NGC+ vs GC+185 significantly different genes
Figure 1 Gene and pathway expression are altered by GC excess in ACC. (A) Among the four comparisons performed, absence vs presence of GC excess (GC- vs GC+) was associated with the largest number of significantly different genes in ACC. (B) The expression of 858 genes was found to be significantly affected by GC excess (P ≤ 0.05 and >two-fold change in expression compared to GC-), with some genes showing higher expression in GC+ cases (red), and some showing lower expression (blue). (C) Genes affected by GC excess were overrepresented in pathways related to immune and steroid synthesis processes.

B

-log10(Adjusted P Value)

C

Terpenoid backbone biosynthesis

Steroid biosynthesis

Upregulated

2-fold change

Aldosterone synthesis and secretion

in GC+ ACC

Primary immunodeficiency

Downregulated in GC+ ACC

Allograft rejection

4

Graft-versus-host disease

Type I diabetes mellitus

Autoimmune thyroid disease

Viral protein interaction with cytokine and cytokine receptor

Viral myocarditis

T cell receptor signaling pathway

2

Antigen processing and presentation

Th17 cell differentiation

. .

P =: 05

Th1 and Th2 cell differentiation

Natural killer cell mediated cytotoxicity

Hematopoietic cell lineage

Fold Change in Expression Between GC-and GC+

Cytokine-cytokine receptor interaction

Cell adhesion molecules (CAMs)

-4

0

4

T

0

10

20

30

Lower Expression in GC+

Higher Expression in GC+

Number of Genes in Pathway Affected by GC Excess

were primarily associated with aldosterone, GC, and bile synthesis/secretion, similar to the pathways found to be upregulated in GC+ ACC (Fig. 1C). In contrast, the immune pathways with downregulated gene expression identified by mRNA analysis were not enriched in either the hypo- or hypermethylated sets. Thus, differences in methylation may explain the upregulation of steroidogenesis pathways but not the downregulation of immune pathways.

Since promoter hypermethylation could not account for the reduced abundance of immune-related transcripts in ACC tumors, the abundance of various cell types was determined in the ACC tumors. Cell abundance was deconvoluted from the mRNA data using xCell (Aran et al. 2017). GC+ ACC tumors showed lower lymphocyte abundance with higher myeloid and mesenchymal stem cell abundance as compared to GC- tumors. T cells

(P < 0.005) and natural killer T cells (NKT cells, P = 0.014) were less abundant in GC+ cases compared to GC- (Fig. 3). In contrast, mesenchymal stem cells and neutrophils were more abundant in GC+ cases (P < 0.001, Fig. 3). Total missense and nonsense mutation were separately assessed to determine tumor mutation burden (TMB). Higher TMB was observed in the GC+ cases (P = 0.029, Supplementary Fig. 2).

Gene signature can predict tumors resembling GC+ ACC cases

GR activity was assessed in different tumor types using a published GR-driven gene signature (West et al. 2018). These data confirmed that GR activity is high in ACC compared to other tumor types in TCGA (Fig. 4).

A Unsupervised Clustering

Glucocorticoid

Hormone Status

Hormone

RAETIE

MAVZ

HLA-DQA2

GZMB

SINI4

ITGAL

CO247

Genes in Pathways

HLA-OQA1

HLA-CPB1

HLA-ORA

PDCDT

SH2DIA

GD3D CORE

CORA

CORA

PRKCA JAR’S

HLA-A HLA-F

HLA-B

HLALE

CD25

TCGA.OR.A5JO TOGA.OR.A5JK TCGA OR ASLK

TCGA.OR.A5LN

TCGA.OR.A5LA

IL2RA

TOGA.OR.A5L5

TOGA.PK ASHA

TOGA OR ASI 9

TOGA OB AS IM TOGA. OR.ASJV

TCGA. OR.ASJC

TCGA.OR.ASJI

TCGA.OR A5JD

TOGA.OR.A5LP

TCGA.OR.A5JR

TCGA.OR.A5KT

TCGA.OR.ASL.

YOGA OR ASL

TOGA.ORA5JJ

TCGA.OR.ASLG

TCGA.OR.A5KZ

TCGA.OR.A5J2

TCGA.PK.A5H9

TCGA.OR.A5LH

TCGA OR ASK1

TEGA ARKI

TCGA. CH. ADJE

TCGA.OR.ASJL

TOGA.OR.A5L6

TCGA.OR.ASLR

TOGA.OR.A5JT

TCGA.OR.A5K3 TCGA.OR.A5JC

ICGA OR ASIX

TOGA. OH.ASJX

TCGA.OR.ADIM

TCGA.OR ASK8 TCGA.OR.A5J5

TCGA.OR.A5J9 TCGA.OU.ASPI

TCGA.OR.ASJA

TOGA.OR.A5L4

TCGA.OR A5KA

TOGA OR ASKU

TOGA.OD ABY

TOGA.OR.ASLI

TOGA.OR.A5JP

TCGA.OR.A5J1

TCGA.OR.ASKY

TCGAOR A5KB TOGA OR ASJE

TCGA.OR ASK2

TCGA. OH.ADJG

TOGA. OH.ASLL

TCGA.P6.ASOF

TCGA.OR.ASJW

TOGA.OR.A5JF

TCGA OR A5 13

TCGA.OR.ASLC

TOGA OR ASLC

TOGA. OR.ASJS

TOGA OR ASKS TCGA.OR ASKO

TCGA.OR.ASKV

TCGA.OR.A5KW

TCGA OR ASLO

Tech op Ane

ICGA.OCH ROLE

TCGA. Un.ASLO

TOGA. OR. ASLO

TCGA.OR.A5J7

TCGA OR ASKO

TCGA.OR ASK9

TOGA.OR.A5JM

Lower normalized gene expression

2

0

Higher normalized 2 gene expression

ACC Tumors

B Supervised Clustering by GC Status

Figure 2 Expression of 2 KEGG pathways: 'T-cell receptor signaling pathway' and 'natural killer cell- mediated cytotoxicity'. The top 2 rows indicate GC and general hormone status for each tumor (black: GC+/H+, white: GC-/H-), shades of blue/ red show normalized gene expression for each tumor, with darker blue corresponding to lower expression. (A) When unsupervised clustering by gene expression was conducted, GC+ cases appear toward the right of the figure, where many genes show lower expression. (B) When supervised clustering by GC status was conducted, lower expression of genes in these pathways appears in GC+ cases (darker blue cluster on the right).

Glucocorticoid

Hormone Status

Hormone

RAETIE

PRACE

HLA-DQA2

GZMB

STATA

PIPRC

GRAP2

LCK

ITGAL

CD347

ITK

Genes in Pathways

HLA-DQA1

HLA-DPB1

HLA-DRA

PDCDI

GOOG SH2DIA

COD COSE

CORR

PRKCA

MAV3 HLA-A

A

HLA-B

HLA-E

CD28 AL2RA

TCGA.OR.A5JO

TCGA.OR.A5JK TOGA.OR ASLK TCGA.OR.A5LN

TOGA.OR ASLA

TCGA.OR.ASLS

TCGA.PK.ASHA

ICGA OR. ASLS

TCGA.OR.Pod

TOGA OR A5JO

TOGA OR A5LP

TOGA. OR ASJR

TCGA.PK.A5H9

TCGA.OR.A5LH

TOGA OR ASK1

ICGA. OR.ASL6

TOGA ORASUL

TCGA.OR.AS.IZ

TCGA. OR.ASLA

TOGA.OR A5K3

TOGA.OR A5JC

TCGA.OR.A5JX

TCGA.OR.A5JJ

TOGA OR A5J2

TOGA OR A5J1

TOGA OR A5J5

TOGA OR A5J9 TCGA.OU.ASPI

TCGA.OR.A5LT

TCGA OR AS IP

TCGA.OR.ASJA

TOGA. OR ASK4

TOGA. OR.ASKU

TCGA. OR.ASIS

TCGA.OR.A5LO

TOGA.OR A5LB

TCGA.OR.ASLD

TCGA.OR.ASKT

TOGA OD ARO TOGA OR ASL.

TCGA.OR.A5LG

TCGA.OR.ASKZ

ICA OR AS IT

LEGA OR ASLA

TCGA.OR.ASJY

TEGA OR ASJY

TOGRA CIL HOLS

TCGAOR ASKW

TCGA.OR.ASKY

TCGA.OR.A5JG

TOGA OR ASKE

TCGA.OR.MOL

TCGA.PB.ASOF

TOGA.OR.A.W

TCGA.OR.A5JF

TOGA OR A5J3

TCGA.OR.ASL3 TOGA OR A5LC

TOGAOR MER TOGA.OR HET

TOGA. OR.ASKO

TCGA.OR.A5L8

TOGA.OR ASK9 TOGA.OR.A5JM

Lower normalized gene expression

Higher normalized gene expression

-2

0

ACC Tumors

There was no difference between ACC cases with different hormone and GC status (insert in Fig. 4). The published GR-driven gene signature (West et al. 2018) was derived primarily from ER+ breast cancer cells and tumors and thus may not be ideal for the assessment of ACC-specific transcriptional profiles.

As the published GR activity signature scores were consistent between ACC cases, we defined a new gene signature capable of identifying GC+ ACC cases. Random forest methods were used to train a cross-validated model that distinguishes GC+/- ACC cases with a receiver- operator characteristic curve area under the curve (ROC AUC) of 0.87 ± 0.09 (Fig. 5A). The sensor component of the inflammasome (NLRP1) and a mediator of NK activation by IL-15 (ZNF683) were identified as important parts of this signature (insert in Fig. 5A), confirming the relevance of NK

cells in GC biology. The gene signature was then applied to other tumors types in TCGA to identify those with GC+-like transcriptional profiles (Fig. 5B). According to this score, uveal (UVM) and skin cutaneous melanomas (SKCM) may have the highest frequency of cases similar to GC+ ACC of the 13 tumor types assessed (Fig. 5C). This gene signature may be useful in a tissue-based diagnostic to find GC+ ACC cases or similar cases among other tumor types.

Effects of cortisol and relacorilant on NK cell function in vitro

Given the prominent suppression of NK-related genes in GC+ cases, the direct effects of GR modulation of human NK cells were assessed. Human NK cell activation (abundance of CD25+CD69+ cells) was increased by stimulation with

Less abundant in GC+

CD4+ naive T cells

CD8+ T cells

CD4+ memory T cells

0.12

P =. 003

P =. 002

0.3

P =. 004

0.6

XCell score

0.08

XCell score

XCell score

0.2

0.4

0.04

0.2

0.1

0.00

0.0

0.0

GC-

GC+

GC-

GC+

GC-

GC+

CD8+ Tcm cells

NKT cells

0.6

P <. 001

1.00

P =. 014

0.75

0.4

XCell score

XCell score

0.50

0.2

0.25

0.0

0.00

GC-

GC+

GC-

GC+

More abundant in GC+

Figure 3 xCell analysis of cell abundance in ACC tumors. Lymphocyte abundance was lower, while mesenchymal stem cells and neutrophil abundance was higher in GC+ cases. GC+, glucocorticoid excess present; GC-, glucocorticoid excess absent.

Mesenchymal stem cells

Neutrophils

2.0

P <. 001

P <. 001

0.09

1.5

XCell score

XCell score

0.06

1.0

0.03

0.5

0.00

GC-

GC+

GC-

GC+

IL-2, suppressed by cortisol, and restored by relacorilant (Mann-Whitney P = 0.0039, Fig. 6A). Proliferation of NK cells was also increased by stimulation, suppressed by cortisol, and restored by relacorilant (Mann-Whitney P = 0.0099, Fig. 6B). Cytokine production (both transcript and secreted protein) was also increased by stimulation, suppressed by cortisol, and restored by relacorilant (Fig. 6C, D, E and F). Genes that were significantly induced by stimulation, suppressed by cortisol, and restored by relacorilant included key NK-activation genes, including the IL-2 receptor (IL2RA) and the activator LAG3 (Fig. 6F). These data provide experimental confirmation of

the observed effects of GC on NK cell populations in ACC tumors.

Activation, proliferation, and cytokine secretion are all indicative of a functional change in NK cells mediated by cortisol and relacorilant. To determine if this functional change also affected target cell killing, PBMCs stimulated with IL-12+IL-15 were incubated with K562 tumor cells. At various effector:tumor cell ratios, cortisol suppressed tumor cell killing and relacorilant restored it (Fig. 6G). There was a significant improvement in tumor killing when relacorilant was added at the 5:1 effector:tumor ratio (Mann-Whitney P = 0.004) (Fig. 6H). These findings

Figure 4 GR activity score for different tumor types and ACC subsets. Independent of hormone status (see insert), ACC exhibited high GR-driven gene activity relative to other tumors. GR, glucocorticoid receptor; NGC+, non- glucocorticoid hormone excess present; GC+, glucocorticoid excess present; H-, any hormone excess absent; H+, any hormone excess present; ACC, adrenocortical carcinoma; CHOL, cholangiocarcinoma; LUSC, lung squamous cell carcinoma; PAAD, pancreatic adenocarcinoma; LIHC, liver hepatocellular carcinoma; PRAD, prostate adenocarcinoma; OV, ovarian serous cystadenocarcinoma; LUAD, lung adenocarcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; LGG, brain lower grade glioma; UVM, uveal melanoma; BLCA, bladder urothelial carcinoma; SKCM, skin cutaneous melanoma.

5

Kruskal-Wallis, P <. 001

5-

4

4-

GR Summarized Activity

3

!

NGC+ GC+ H- H+

0

0

3

0

0

·

2

. 6

0

0

3

·

¢

9

10

0

1

:

·

ACC CHOL LUSC PAAD LIHC PRAD OV LUAD BRCA CESC LGG UVM BLCA SKCM Cancer Type

confirm that GCs suppress tumor cell killing by human NK cells in vitro.

Discussion

ACC is a grievous disease in which patients face challenges both in tumor and hormone management. ACC patients with GC excess experience Cushing syndrome, which by itself can increase the comorbidities associated with ACC, including thromboembolic, musculoskeletal, cardiovascular, infectious, and metabolic complications (Yaneva et al. 2013). In addition, GCs are potent transcriptional regulators and mediators of immune- cell function. The TCGA ACC data provides a unique dataset in which rich multi-omics data are paired with clinical assessment of GC excess. This exaggerated cortisol physiology was investigated to better understand ACC, select biomarkers for a phase 1 trial with a GR antagonist, and to glean insights into possible subclinical and/or local manifestations of GC activity in other tumor types.

Analysis of the TCGA ACC dataset showed a significant impact of presence vs absence of GC excess. Significant differences in 858 genes were observed between ACC cases with or without GC excess, while fewer genes showed significant differences across our other comparisons, such as cases with or without excess in any steroid hormone. Genes involved in steroid synthesis, including the KEGG pathways ‘steroid biosynthesis’, ‘aldosterone synthesis and secretion’, and ‘terpenoid backbone biosynthesis’, were, higher in cases with GC excess. Increased promoter hypomethylation was observed for steroid synthesis genes in the GC+ cases, consistent with reports of epigenetic regulation of cortisol synthesis (Liu et al. 2004). In contrast, no difference in methylation between GC+ and GC- cases was observed for immune genes in the ‘T-cell receptor signaling pathway’ and ‘natural killer cell- mediated cytotoxicity’ KEGG pathways. Genes in those two pathways, indicative of T-cell and NK cells presence and activation, were higher in the GC- cases. Furthermore, fewer infiltrating immune cells (T cells and NKT cells) were found in GC+ tumors. These findings, consistent with those reported with immunohistochemistry methods (Landwehr et al. 2020), suggest that the observed immune effects in GC+ cases were likely a consequence of poor infiltration of immune cells into GC+ tumors.

Assessment of GR activity via a published gene signature (West et al. 2018) suggested that intratumor GR activity is similar in ACC cases with or without GC excess. This may be driven by high local cortisol levels within the adrenal gland independent of systemic cortisol levels. Thus, the differences in immune infiltration may be due to the systemic effects of GC, including effects on primary and secondary lymphoid organs throughout the body. Effects of GC on lymphoid organs may also be related to the increased TMB observed in GC+ ACC cases, as high GC may increase tolerance toward neo-antigens. These specific effects of GC excess on lymphocyte tumor infiltration may be reversed by a GR antagonist.

Aberrant cortisol production or activity of ACC has been reported (Cirillo & Prime 2011, Cirillo et al. 2017, Sharma et al. 2018), but its assessment is not commonly part of solid tumor care outside of ACC. Since systemic GC excess had such a pronounced effect on transcription, we next looked for other tumors types that resembled GC+ cases. We reasoned that similarities to GC+ ACC could underly local or systemic cortisol activity. To achieve this, we developed a gene signature that can predict GC+- like tumor cases. We note that the accuracy of this gene signature (and other analyses of the TCGA data reported here) hinges upon the correct classification of GC status

Figure 5 Derivation of a gene signature that distinguishes GC+/- ACC cases using random forest. (A) NLRP1 and ZNF683 (highlighted) were identified as important components of the signature. Only signature genes above the threshold of 0.0028 are shown. (B) The ACC gene signature was applied to other tumors types in TCGA. The horizontal line distinguishes GC+ and GC- tumors (derived based on the known distribution of GC+/- cases in ACC). (C) Frequency of tumor cases resembling GC+ ACC by tumor type, predicted by the newly derived gene signature. GC+, glucocorticoid excess present; GC-, glucocorticoid excess absent.

A Random Forest: Importance of Found Genes

B Gene Signature Applied to TCGA Tumor Types

ADGRL3

SPTAN1

1.5

GC-

GC+

Unknown

AGFG1

ZNF683

8

.

TYW5

C3orf80

GC-

MYBPHL

CPNE4

GC+

1.0

RIMKLB

LAMC3

4

ANGPTL2

NDRG4

Expression TMM, log2

Summarized Score

COLCA2

QSOX2

0.5

EML2

NRXN3

0

:

AADAT

GTPBP4

CBWD2

0.0

MAGIX

RTN4R

B9D1

-4

PCOLCE2

ARAP2

EIF2S1

ACC

UVM

SKCM

LIHC

CHOL

OV

BLCA CESC LUSC

PAAD

LUAD B

BRCA

RAD

LGG

PAPSS2

:

Cancer type

IKZF3

·

RPLPO

DENND2C

NLRP1

ZNF683

C Predicted Frequency of GC+ Cases

KIAA0040

ILDR2

CYP3A4

ACC

GC+

GC-

WDYHV1

COMMD6

PIN1

UVM

NR5A1

CPN2

SKCM

SETBP1

LRWD1

LIHC

CHD7

ELOVL7

OV

DLG3

NLRP1

PRAD

FAM166B

Cancer type

LRIG1

PAAD

SERTAD4

PLS3

LUSC

T

T

0.00

0.01

0.02

0.03

LUAD

Gene Importance (a.u.)

LGG

CHOL

CESC

BRCA

BLCA

0.00

0.25

0.50

Relative frequency

0.75

1.00

in the underlying dataset. This diagnosis is not based on a single test with a well-defined threshold; it is polyfactorial and ultimately requires judgment of the diagnosing physician. The derived gene signature could be useful in future efforts to diagnose GC excess from a single test using a tumor biopsy or resected tumor. When non-ACC tumors were scored with our signature, uveal and skin cutaneous melanomas exhibited the highest frequency of cases that resembled the transcriptional signature of GC+ ACC, albeit these cases are still rare. This supports previous reports of local cortisol production in the skin (Vukelic et al. 2011). Based on our findings, it would be reasonable to assess such tumors for immune effects of GR antagonism.

Suppression of NK cells was prominent in the GC+ ACC multi-omics data. NK activation genes were significantly

lower in GC+ cases, and the NK activation gene ZNF683 was among the most important genes distinguishing GC+ from GC- cases. Based on these findings, functional studies assessing the direct effects of GR modulation on human NK cells were conducted. We used the SGRM relacorilant, which antagonizes GR, to counteract the effects of the GR agonist cortisol in human NK cells. These studies confirmed that GR is a key regulator of NK function. Cortisol suppressed NK proliferation, upregulation of cell surface markers of activation, tumor cell killing, IFNy secretion, and IFNy transcription. It also suppressed the secretion of other effector cytokines and expression of the IL-2 receptor (IL2RA). These observations corroborate the decrease in NK activation genes observed in GC+ ACC. Cortisol suppressed, and relacorilant promoted the

Figure 6 Effects of stimulation, cortisol, and/or relacorilant on isolated human NK-cell activity in vitro. In cells stimulated with IL-2 (500 U/mL), cortisol (200 nM) suppressed and the addition of relacorilant (300 nM) significantly improved (A) NK-cell activation and (B) NK-cell proliferation. Addition of relacorilant to IL-2 + cortisol improved (C) IFNy, (D) TFNa, and (E) Granzyme A secretion relative to NK cells treated with cortisol alone. (F) Transcription of IFNG was improved by relacorilant, along with other key regulators of NK activity, including LAG3 and IL2RA. (G) GC also suppressed tumor cell killing by human NK cells in vitro. K562 cell killing is shown at various ratios of effector:tumor cells under the treatment conditions listed above after stimulation with IL-12 (50 ng/ml) + IL-15 (1 ng/ml). (H) At the 5:1 ratio, the significant decrease in tumor cell killing in the presence of cortisol was counteracted by relacorilant. CORT, cortisol (200 nM); RELA, relacorilant (300 nM); NK cell, natural killer cell.

A

NK Cell Activation

B

NK Cell Proliferation

C

INFY

CD25 + CD69 + NK Cells (% Stimulated Control)

125-

125-

15

P =. 0039

NK Proliferation (% Stimulated Control)

P =. 0099

LLOQ = 0.5

ULOQ = 9.82

100-

100-

IFNy (ng/ml)

10

75-

75-

50-

50-

5.

25

25-

*

*

0

0

0

*

*

Unstim

IL-2

IL-2 +

IL-2 +

IL-2 +

CORT

RELA CORT +

Unstim

IL-2

IL-2 + CORT

IL-2 +

IL-2 +

RELA CORT +

Unstim

IL-2

IL-2 +

IL-2 +

IL-2 +

RELA

RELA

CORT RELA CORT +

RELA

D

TNFa

E

Granzyme A

F

NK Cells

2000

LLOQ = 80

250

ULOQ = 4000

LLOQ = 0.036

TNFa (pg/ml)

Granzyme A (ng/ml)

ULOQ = 200.07

12-

CCL3/L1

1500

200

RNA Abundance (Log2 Counts)

TNFRSF9

10-

IL2RA

150

DUSP5

1000

8-

100

IFNG

6

LAG3

500

50

HLA-DQA1

4.

*

0

0

2

Unstim

IL-2

IL-2 +

IL-2 +

IL-2 +

CORT RELA CORT +

Unstim

IL-2

IL-2 +

IL-2 + II

IL-2 +

CORT RELA CORT +

Unstim.

IL-2

IL-2

IL-2

RELA

RELA

+ CORT

+ CORT

+ RELA

G

K562 Cell Killing at Various Effector: Tumor Ratios

H

K562 Cell Killing (5:1 ratio)

60-

200-

Tumor Cell Death (%)

Unstim

IL-2

K562 Killing (% of Positive Control)

175-

40-

IL-2 + CORT

150

IL-2 + RELA

125

P =. 004

IL-2 + CORT

+ RELA

100

20

75

50

25

0

0

20:1

10:1

5:1

2:1

0.5:1

Unstim

IL-12 +

IL-15

IL-12 +

IL-15 +

IL L-12

+

IL-12 +

NK: Tumour Cell Ratio

IL-15 +

IL-15 +

CORT RELA CORT + RELA

expression of LAG3 (CD223, LAG3) and 4-1BB (CD137, TNFRSF9), both targets of experimental agonists intended to improve the anti-tumor immune response. Expression of chemokine ligand 3-like 1 (CCL3L1), a chemokine that attracts lymphocytes, was also suppressed by cortisol in stimulated NK cells, which could also explain the reduced T-cell infiltrate into the GC+ ACC. The observed reduced abundance of immune-related transcripts in GC+ ACC provides insight into the mechanisms by which GC may limit response to ICI therapy.

The data presented here suggest that selective GR antagonism with the investigational SGRM relacorilant may be able to counteract the immune suppression caused by systemic cortisol. This does not exclude a potential benefit of GR antagonism in all ACC cases, particularly as

the local GR activity signature was high in all ACC subsets (Fig. 4, inset) and NK function was suppressed at normal (200 nM) cortisol concentrations (Fig. 6). Selective GR antagonism could both promote anti-tumor efficacy of other immune modulators, such as immune checkpoint inhibitors or more experimental NK-targeting agents and reduce the dangerous sequalae of cortisol excess. This hypothesis is being tested directly in a phase 1 study of relacorilant +pembrolizumab in patients with ACC and GC excess (NCT04373265).

Supplementary materials

This is linked to the online version of the paper at https://doi.org/10.1530/ ERC-21-0048.

Declaration of interest

AEG, SW, and A Gare employees and stockholders of Corcept Therapeutics; M A H is a consultant/advisor for Corcept Therapeutics, HRA Pharma, and Calico, as well as an investigator/researcher for Exelixis.

Funding

The studies presented here were solely funded by Corcept Therapeutics.

Acknowledgements

The results shown here are based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. Editorial support was provided by Tina K Schlafly, PhD, of Corcept Therapeutics. The authors thank Stacie Shepherd for helpful conversations and Charles River Labs (Portishead, UK) and Ardigen (Krakow, Poland) for management and guidance of these experiments. The authors also thank the healthy donors for their willingness to donate the critical components, the NK cells, to this project.

References

Abercrombie HC, Giese-Davis J, Sephton S, Epel ES, Turner-Cobb JM & Spiegel D 2004 Flattened cortisol rhythms in metastatic breast cancer patients. Psychoneuroendocrinology 29 1082-1092. (https://doi. org/10.1016/j.psyneuen.2003.11.003)

Aran D, Hu Z & Butte AJ 2017 xCell: digitally portraying the tissue cellular heterogeneity landscape. Genome Biology 18 220. (https://doi. org/10.1186/s13059-017-1349-1)

Assié G, Letouzé E, Fassnacht M, Jouinot A, Luscap W, Barreau O, Omeiri H, Rodriguez S, Perlemoine K, René-Corail F, et al. 2014 Integrated genomic characterization of adrenocortical carcinoma. Nature Genetics 46 607-612. (https://doi.org/10.1038/ng.2953) Cirillo N & Prime SS 2011 Keratinocytes synthesize and activate cortisol. Journal of Cellular Biochemistry 112 1499-1505. (https://doi. org/10.1002/jcb.23081)

Cirillo N, Morgan DJ, Pedicillo MC, Celentano A, Lo Muzio L, Mccullough MJ & Prime SS 2017 Characterisation of the cancer- associated glucocorticoid system: key role of 116-hydroxysteroid dehydrogenase type 2. British Journal of Cancer 117 984-993. (https:// doi.org/10.1038/bjc.2017.243)

Else T, Williams AR, Sabolch A, Jolly S, Miller BS & Hammer GD 2014 Adjuvant therapies and patient and tumor characteristics associated with survival of adult patients with adrenocortical carcinoma. Journal of Clinical Endocrinology and Metabolism 99 455-461. (https://doi. org/10.1210/jc.2013-2856)

Gilbert R & Lim EM 2008 The diagnosis of Cushing’s syndrome: an Endocrine Society clinical practice guideline. Clinical Biochemistry 29 103-106.

Habra MA, Stephen B, Campbell M, Hess K, Tapia C, Xu M, Rodon Ahnert J, Jimenez C, Lee JE, Perrier ND, et al. 2019 Phase II clinical trial of pembrolizumab efficacy and safety in advanced adrenocortical carcinoma. Journal for ImmunoTherapy of Cancer 7 253. (https://doi. org/10.1186/s40425-019-0722-x)

Head L, Kiseljak-Vassiliades K, Clark TJ, Somerset H, King J, Raeburn C, Albuja-Cruz M, Weyant M, Cleveland J, Wierman ME, et al. 2019 Response to immunotherapy in combination with mitotane in patients with metastatic adrenocortical cancer. Journal of the Endocrine Society 3 2295-2304. (https://doi.org/10.1210/js.2019-00305)

Jehn CF, Kühnhardt D, Bartholomae A, Pfeiffer S, Schmid P, Possinger K, Flath BC & Lüftner D 2010 Association of IL-6, hypothalamus- pituitary-adrenal axis function, and depression in patients with cancer. Integrative Cancer Therapies 9 270-275. (https://doi. org/10.1177/1534735410370036)

Kanehisa M & Goto S 2000 KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Research 28 27-30. (https://doi.org/10.1093/ nar/28.1.27)

Landwehr LS, Altieri B, Schreiner J, Sbiera I, Weigand I, Kroiss M, Fassnacht M & Sbiera S 2020 Interplay between glucocorticoids and tumor-infiltrating lymphocytes on the prognosis of adrenocortical carcinoma. Journal for Immuno Therapy of Cancer 8 e000469. (https:// doi.org/10.1136/jitc-2019-000469)

Liu J, Li XD, Vaheri A & Voutilainen R 2004 DNA methylation affects cell proliferation, cortisol secretion and steroidogenic gene expression in human adrenocortical NCI-H295R cells. Journal of Molecular Endocrinology 33 651-662. (https://doi.org/10.1677/jme.1.01560)

Miller AH, Spencer RL, Pearce BD, Pisell TL, Azrieli Y, Tanapat P, Moday H, Rhee R & McEwen BS 1998 Glucocorticoid receptors are differentially expressed in the cells and tissues of the immune system. Cellular Immunology 186 45-54. (https://doi.org/10.1006/ cimm.1998.1293)

Mormont MC & Lévi F 1997 Circadian-system alterations during cancer processes: a review. International Journal of Cancer 70 241-247. (https:// doi.org/10.1002/(sici)1097-0215(19970117)70:2<241::aid-ijc16>3.0.co;2-l)

Nicolaides NC, Kino T & Chrousos G 2000 AIDS and HPA axis. In Endotext. Eds Feingold KR, Anawalt B, Boyce A, Chrousos G, de Herder WW, Dungan K, Grossman A, Hershman JM, Hofland J, Kaltsas G, et al. South Dartmouth, MA, USA: MDText.com, Inc. (available at: https://www.ncbi.nlm.nih.gov/books/NBK279014/) Palesh O, Zeitzer JM, Conrad A, Giese-Davis J, Mustian KM, Popek V, Nga K & Spiegel D 2008 Vagal regulation, cortisol, and sleep disruption in women with metastatic breast cancer. Journal of Clinical Sleep Medicine 4 441-449. (https://doi.org/10.5664/jcsm.27280)

Raj N, Zheng Y, Kelly V, Katz SS, Chou J, Do RKG, Capanu M, Zamarin D, Saltz LB, Ariyan CE, et al. 2020 PD-1 blockade in advanced adrenocortical carcinoma. Journal of Clinical Oncology 38 71-80. (https://doi.org/10.1200/JCO.19.01586)

Sharma P, Sandhu SV, Bhandari R, Verma I, Bhullar RK & Khangura RK 2018 Estimation of cortisol levels in patients with premalignant disorders and oral squamous cell carcinoma. Journal of Oral and Maxillofacial Pathology 22 27-34. (https://doi.org/10.4103/jomfp. JOMFP_181_16)

Vukelic S, Stojadinovic O, Pastar I, Rabach M, Krzyzanowska A, Lebrun E, Davis SC, Resnik S, Brem H & Tomic-Canic M 2011 Cortisol synthesis in epidermis is induced by IL-1 and tissue injury. Journal of Biological Chemistry 286 10265-10275. (https://doi.org/10.1074/jbc. M110.188268)

West DC, Kocherginsky M, Tonsing-Carter EY, Dolcen DN, Hosfield DJ, Lastra RR, Sinnwell JP, Thompson KJ, Bowie KR, Harkless RV, et al. 2018 Discovery of a glucocorticoid receptor (GR) activity signature using selective GR antagonism in ER-negative breast cancer. Clinical Cancer Research 24 3433-3446. (https://doi.org/10.1158/1078-0432.CCR-17-2793)

Yaneva M, Kalinov K & Zacharieva S 2013 Mortality in Cushing’s syndrome: data from 386 patients from a single tertiary referral center. European Journal of Endocrinology 169 621-627. (https://doi. org/10.1530/EJE-13-0320)

Zheng S, Cherniack AD, Dewal N, Moffitt RA, Danilova L, Murray BA, Lerario AM, Else T, Knijnenburg TA, Ciriello G, et al. 2016 Comprehensive pan-genomic characterization of adrenocortical carcinoma. Cancer Cell 29 723-736. (https://doi.org/10.1016/j.ccell.2016.04.002)

Received in final form 2 June 2021 Accepted 4 June 2021 Accepted Manuscript published online 4 June 2021

@ 2021 Society for Endocrinology Published by Bioscientifica Ltd. Printed in Great Britain