ELSEVIER

Cancer Genetics xxx (xxxx) xxx

Establishing a human adrenocortical carcinoma (ACC)-specific gene mutation signature

Chinmay Satish Rahanea, Arne Kutznerb, Klaus Heesea,*

a Graduate School of Biomedical Science and Engineering, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 133-791, Republic of Korea; b Department of Information Systems, College of Engineering, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 133-791, Republic of Korea

Abstract

Adrenocortical carcinoma (ACC) is a rare and aggressive tumor whose molecular signaling pathways are not fully understood. Using an in-silico clinical data analysis approach we retrieved human gene mutation data from the highly reputed Cancer Genome Atlas (TCGA). ACC-specific gene mutations were correlated with proliferation marker FAM72 expression and Mutsig along with the algorithmic implementation of the 20/20 rule were used to validate their oncogenic potential. The newly identified oncogenic driver gene set (ZFPM1, LRIG1, CRIPAK, ZNF517, GARS and DGKZ), specifically and most repeatedly mutated in ACC, is involved in tumor suppression and cellular proliferation and thus could be useful for the prognosis and development of therapeutic approaches for the treatment of ACC.

Keywords Adrenal gland, CRIPAK, DGKZ, LRIG1, Oncogene, ZFPM1, Zinc finger. @ 2018 Elsevier Inc. All rights reserved.

Introduction

Adrenocortical carcinoma (ACC) is a rare cancer affecting the adrenal cortex. ACC is a malignancy of the steroid hormone (i.e., mineralocorticoids, glucocorticoids, and androgens)- producing layer of the adrenal gland. ACC occurs mostly in women and has an incidence rate of 0.7-2 cases per million people per year [1]. ACC occurs primarily between 40 and 50 years of age and is often aggressive [2,3]. ACC has a high risk of metastatization even after early diagnosis and surgery [4]. Metastasis is the ability of tumor cells to leave the pri- mary site and migrate to other tissues, leading to the spread of cancer throughout the body [5]. The molecular mecha- nisms of metastasis have been well-studied in carcinomas, in which epithelial to mesenchymal transition (EMT) is con- sidered a critical process [5,6]. Invasion could either occur through the lymphatic system, through the venous system or through the extra-adrenal tissues [7]. Tissue biopsy is unhelp- ful and tumor protein p53 (TP53) expression appeared to be

inadequate as prognostic marker in ACC; instead, quantifica- tion of cell proliferation marker MKI67 is critical for prognosis also in ACC [1,5,8,9]. Several genes have been proposed as cancer drivers in ACC, amongst them: TP53 and catenin B 1 (CTNNB1) [10-12]. TP53 mutations have been observed in more than 50% of child patients and 4% of adult patients of ACC [11,12]. Alterations in the components of the WNT/B- catenin pathway are a prominent marker in ACC. CTNNB1 ac- cumulation has been reported from cases of ACC [13,14] and somatic mutations in CTNNB1 have been observed in large cohort studies [13,15,16]. Activating mutations in CTNNB1 have been observed in approximately 25% of adrenocorti- cal cancers [14]. CTNNB1 and TP53 mutations have been reported to be mutually exclusive in aggressive adrenal can- cers [17]. CTNNB1 knockdown was reported to inhibit EMT, however, that may not apply to adrenal cells due to their mesodermal origins [18]. Other genes associated with ma- lignant ACC include insulin growth factor 2 (IGF2) and splic- ing factor 1 (SF1). IGF2 overexpression is observed in more than 80% cases of ACC and was among the earliest abnor- malities described in ACC [19]. SF1 overexpression is asso- ciated with poor prognosis in childhood and adult sporadic ACC [20].

Despite the importance of the IGF- and WNT/B-catenin signaling pathways, the pivotal driving factors behind ACC

Received August 10, 2018; received in revised form October 20, 2018; accepted October 22, 2018

* Corresponding author.

E-mail addresses: klaus@hanyang.ac.kr, klaus.heese@rub.de

are still unknown and the treatment of ACC remains an unresolved challenge [10,13,19,21,22]. Menin 1 (MEN1), protein kinase cAMP dependent type-I regulatory subunit alpha (PRKAR1A), ribosomal protein L22 (RPL22), telomeric repeat binding factor 2 (TERF2), cyclin E1 (CCNE1) and neurofibromin 1 (NF1) were recently identified as possible cancer drivers in ACC [16]. However, it is highly likely that other genes are involved in ACC tumorigenesis, which have not been reported yet [16].

In our study, we conducted a complete analysis of the mu- tational data of ACC using the comprehensive public cBio- Portal human cancer study database in order to determine new genes involved in ACC oncogenesis. We identified a new ACC-specific gene mutation signature, comprised of the six genes ZFPM1, LRIG1, CRIPAK, GARS, ZNF517 and DGKZ. We validated our findings using the Mutsig (for “Mutation Significance”) [23] data as well as the 20/20 rule proposed by Vogelstein et al. [24]. Our new ACC-specific gene sig- nature ties in with the WNT/B-catenin and MAPK signaling pathways and offers new targets for the effective treatment of ACC.

Materials and methods

Human cancer patient data sources

A publicly available human ACC study (http://www.cbioportal. org/) [25,26] from The Cancer Genome Atlas (TCGA) was an- alyzed for mutations and mRNA expression data (provisional data set). cBioPortal is a human cancer genomics database that contains 169 studies with 40,408 samples (as of Jan- uary 2018)) covering 29 different tissues. cBioPortal com- bines data from TCGA (http://cancergenome.nih.gov/), the In- ternational Cancer Genome Consortium (ICGC; https://icgc. org/), the Wellcome Trust Sanger Institute’s (WTSI) Cancer Genome Project (http://www.sanger.ac.uk/research/projects/ cancergenome/) and the Cancer Genomics Hub (CGHub; https://cghub.ucsc.edu/). TCGA is a collaborative effort be- tween the National Cancer Institute (NCI; http://www.cancer. gov/) and the National Human Genome Research Institute (NHGRI; https://www.genome.gov/). The TCGA ACC dataset on cBioPortal consisted of mutation data from 90 patients. mRNA expression data was available from 79 of those 90 pa- tients. The mutation data consisted of somatic mutations in 6946 genes.

ACC-specific gene signature identification: ACC-specific gene mutation - FAM72 (A-D) paralogs mRNA expression correlation analysis

Complete mutation data for all genes was retrieved from the ACC TCGA study. The data was sorted by the frequency of mutations in each gene across the ACC study. The five genes demonstrating highest number of mutations were selected for display. Mutations in non-oncogenic genes, as described by Lawrence et al. and Greenman et al. [27,28], were not considered for the analysis. The mutations in the five genes demonstrating highest number of mutations in the study were

compared to the mRNA expression of the proliferation marker FAM72 (A-D) paralogs in the ACC study and visualized with the Xena Functional Genomics Explorer [29]. FAM72 (A-D) is a set of four human-specific paralogs associated with neural stem cells [30-32] and is involved with cellular proliferation in cancerous cells where FAM72 expression is triggered by oncogenic mutations and thus they can be useful as cell proliferation markers [33]. Clinical data from the TCGA ACC study was retrieved from cBioPortal for patient-gene-specific analyses.

ACC-specific mRNA expression analysis of the proliferation marker FAM72 (A-D) paralogs

mRNA expression z-scores (RNA sequencing (RNASeq V1/V2) from the 79 patients in the TCGA ACC study were locally computed on the foundation of raw expression data available on cBioPortal. A z-score is a statistical measure- ment indicating how many standard deviations the element is from the mean. The formula is z= (X- m)/o, where zis the z- score, X is the value of the element, m is the numerical mean of the population, and o is the standard deviation [34]. The z- score was normalized for all samples so that they sum to zero. Linear regression was determined between the FAM72A, the proliferation marker MKI67 [35] and genes expressed in the M-phase of the cell cycle, for all available samples across the TCGA ACC study. The regression curve analysis was visualized with the Python-based Bokeh online visualization tool [36].

ACC-specific gene mutation - FAM72 (A-D) paralogs mRNA expression correlation analysis visualized by the bucket method

The mRNA expression z-scores for FAM72 (A-D) paralogs were separately grouped in buckets with a bucket size of 0.7 Z-score units and consequently correlated with genes show- ing high numbers of ACC-tissue-specific gene mutations. The Y-axis denotes the z-score buckets for the selected FAM72 gene. The genes whose mutation numbers are to be visualized, lie on the X-axis. The data was then visualized with the Python-based Bokeh interactive visualization tool [36]. Number of mutations in a gene in the samples within a bucket are denoted by a color code. The color intensity of the buckets is directly proportional to the number of samples, while the colors visualize the relation of samples with a mutation to the total number of samples. Brighter colors indicate more samples in the bucket while paler colors indi- cate fewer samples in the bucket. Colors tending to the red side of the spectrum indicate increasing number of samples with a mutation in relation to the total number of samples in the bucket. Colors tending to the blue side of the spectrum indicate decreasing number of samples with a mutation in the bucket. Black bands denote absence of mutations or lack of expression data in the gene, while bright grey bands indicate absence of samples within the group. Bright pink boxes indicate that only one sample is present in the bucket that contains one mutation in the gene of interest.

ARTICLE IN PRESS

Novel ACC-specific gene signature

Fig. 1
AACCFAM72AFAM72BFAM72DZFPM1LRIG1CRIPAKZNF517GARS
5' 3' 5'3ª 5'3'5' 3' 5'3'
Maximumfor each count + 1) -
No change Minimum No datamean, calculated log2 (normalized sample in columnnullnullnull
B7.3100% mutationC 8MKI67
6.7
6.080% one6ASPM
FAM72A5.3 4.7leastM-phaseBUB1
4.0atand 4CENPE
of3.3showing 60%MKI67CENPF
2.7
expression [z-score]2.0bucket 40%of [z-score] 2CEP55
1.3inKIF14
mRNA0.7 0.0GARSLRIG1ZNF517Percentage of samples 20% >0 =0 S % 30 1111 1 1mRNA expression cell cycle genes 0 -2KIF23 NEK2 NUF2 SGO1
-0.7 -1.3 -2.0CRIPAKZFPM1
Five genes with highest number of mutations in dataset, arranged alphabeticallyPopulation of bucket-2 0 2 4 6 mRNA expression of FAM72A [z-score]8
D Highly mutated genes in all 90 patientsE Highly mutated genes in 57 surviving patients
Gene IDTotal number of mutations in geneTotal number of mutations in samples containinggene in gene Frequencyacross all 90 of mutations samplesGene IDTotal number of mutations in geneTotal number of samples containing in mutations in geneFrequency of mutations gene across all 57 samples
ZFPM11124752.22%ZFPM1642849.12%
LRIG1543134.44%LRIG1412238.60%
CRIPAK372426.67%CRIPAK291729.82%
GARS353437.78%GARS252442.11%
ZNF517343336.67%ZNF517222136.84%
TP53201820%TP538712.28%
F Highly mutated genesin 33 deceased patients with mutationdataG Somatic mutations unique to 33 deceased patients with mutation data
Gene IDTotal number of mutations in geneTotal number of samples containing mutations ingene Frequency of mutations in gene across all 33 samplesGene IDTotal number of mutations in geneTotal number of samples containing in mutations in geneFrequency of mutations gene across all 33 samples
ZFPM1481957.58%DGKZ655.56%
ZNF517121236.36%GOLGA4555.56%
ZNF787121236.36%KCNH7444.44%
PODXL121133.33%NOS3444.44%
TP53121133.33%ADAMTSL4544.44%
SOWAHA111133.33%ZNRF3444.44%
H OVERALL SURVIVAL STATUS
DGKZ6%
GOLGA46%
KCNH74%
NOS34%
ADAMTSL44%
ZNRF34%
TP5320%
MEN110%
GENETIC ALTERATION INFRAME MUTATIONMISSENSE MUTATION MISSENSE MUTATION (UNKNOWN SIGNIFICANCE)
TRUNCATING MUTATION TRUNCATING MUTATION (UNKNOWN SIGNIFICANCE) DECEASED LIVINGNO ALTERATIONS
OVERALL SURVIVAL STATUS

Please cite this article as: C.S. Rahane, A. Kutzner and K. Heese, Establishing a human adrenocortical carcinoma (ACC)-specific gene mutation signature, Cancer Genetics, https://doi.org/10.1016/j.cancergen.2018.10.005

Gene-specific survival analysis

The prognostic significance of selected genes from ACC was analyzed using available Kaplan-Meier curves from the cBio- Portal database and compared by the log-rank test [37].

Application of the ‘20/20’ rule and Mutsig to validate the oncogenic potential of selected genes in ACC

We applied the 20/20 rule as described by Vogelstein et al. [24] to our findings from the mutation data. In brief, the 20/20 rule classifies genes into oncogenes (ONG) or tumor sup- pressors (TSG). If the frequency of gain of function mutations in the gene is greater than 20%, then the gene is an ONG. If the frequency of recurrent loss of function mutations in a gene is greater than 20% and the frequency of gain is func- tion mutations in the gene is less than 20%, then the gene is a TSG [24]. We used this approach on the somatic mu- tation data from the ACC study on cBioPortal and checked whether selected genes were sorted as ONG or TSG. Mutsig data information for each of the highest mutated genes was also retrieved from cBioPortal. Mutsig is a package of tools for analyzing mutation data. It operates on a cohort of patients and identifies mutations, genes, and other genomic elements predicted to be driver candidates [27]. The Mutsig package develops a background mutation model of the tumor and then analyzes mutations in each gene to identify genes that have been mutated at a higher frequency than background. Con- sequently, Mutsig assigns a Q-value to each mutated gene, based on a false discovery rate of mutations, and genes are then sorted based on their Q-value. Lower Q-value indicates a lower false discovery rate of mutational drivers, therefore

the lower the Q-value, the higher the significance of the muta- tion in that gene. Here, our identified genes were validated as potential oncogenes by combining these two (‘20/20’ rule and Mutsig) methods. CTNNB1, PRKR1A, RB1, MEN1 and NF1 were used as positive controls, as they had been described as confirmed oncogenes in ACC from previous studies [16].

Results

Human gene mutation analysis of the TCGA ACC study

We found five genes prominently mutated in ACC: the leucine- rich repeats and immunoglobulin like domains 1 (LRIG1), the cysteine-rich PAK1 inhibitor (CRIPAK), the zinc finger protein 517 (ZNF517), the zinc finger protein FOG family member 1 (ZFPM1) and the glycyl-tRNA synthetase (GARS) (Fig. 1(A), Supplementary File 1). Although mucin 5B (MUC5B) had a higher number of mutations than CRIPAK, gene mutations in mucins have been reported to be passengers and not onco- genic [27]. We thus disregarded the mucins and focused on the remaining top five most mutated genes, namely ZFPM1, LRIG1, CRIPAK, GARS, and ZNF517.

Highest number of mutations were noted for the gene ZFPM1, followed by LRIG1 and GARS. The mRNA expres- sion level of FAM72 paralogs is not linearly correlated with these mutations (Fig. 1(A)). Fig. 1(B) shows that mutations in ZFPM1, LRIG1 and CRIPAK are spread through the samples and no single driver oncogene could probably cause cellular proliferation, but an accumulation of mutations across a set of genes may be responsible for the cause of ACC. The linear MKI67-FAM72 mRNA expression correlation graph clearly demonstrates that FAM72 is highly expressed in proliferat-

Fig. 1 ACC-specific gene-mutation signature. Correlation between mRNA expression of human proliferation marker FAM72 (A- D) paralogs and proto-oncogenes / tumor suppressor genes frequently mutated in the TCGA ACC study comprised of 79 samples that contained both mRNA expression and mutation data. FAM72 paralogs were used as a comparison as they are known to be expressed in proliferative cells [38]. Comparison between the tumor samples sorted by sample in descending order of FAM72A expression (on the left hand), and the ACC-specific gene-mutation signature represented by the five genes demonstrating highest number of mutations (ZFPM1, LRIG1, CRIPAK, ZNF517 and GARS) in the same ACC samples sorted by number of mutations (on the right hand; sorted from left to right) (A). Red bands indicate increased expression, green bands indicate decreased expression and black bands indicate no change in expression. Blue dots (A, right hand) represent missense or in-frame mutations in the indicated gene in a sample, while red dots represent nonsense or frameshift indel mutations in that gene in the sample. ZFPM1 was the most frequently mutated gene in ACC and mutations occurred predominantly at the E444/L446 region. LRIG1 showed the second highest mutation frequency mutated at the same site L24V in its protein sequences. Bucket-wise distribution of mutations in the top five genes demonstrating highest number of mutations in ACC, sorted by FAM72A expression (B). The grey area in the heat maps and bucketed diagrams indicate lack of data. High gene expression correlation between FAM72A and the proliferative marker MKI67 as well as other M-phase-specific cell cycle genes indicates that FAM72A is highly expressed in proliferating ACC cells (C). The sample size was 79, standard error was 0.09, slope was 0.62, and two sided p-value was 0.0 for the linear regression plot. Mutations in selected genes across all 90 patients in the ACC study (D). Mutations in selected genes across 57 survivors (E). Frequency of mutations in ZFPM1 was slightly lower in survivors than overall patients, but mutation frequency in TP53 was 8% lower in samples from survivors. Most frequently mutated genes in 33 deceased patients with mutation data (F). Somatic mutations observed uniquely in deceased patients highlight a novel ACC-specific gene-mutation signature (DGKZ, GOLGA4, KCNH7, NOS3, ADAMTSL4, and ZNRF3) (G). OncoPrint data from the ACC study on cBioPortal for visualization of the relationship between somatic mutations in genes from (G) and survival of patient (H). Mutations in the genes stated in (D), (E), and (F) are well-established oncogenic drivers. However, the genes mentioned in (G) are - thus far - not reported to be oncogenic drivers, but may assist in metastasis. OncoPrint data in (H) clearly shows that mutations occurred in separate patients, all of whom are deceased, and did not overlap, implying that all genes from this novel gene-mutation signature from (G) could have each played a pivotal lethal role as primary driver oncogene in the oncogenic pathway of ACC or in metastasis in conjunction with other driver oncogenes. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

Novel ACC-specific gene signature

JID: CGEN

Alteration frequency

y

4

3

2

2

Alteration frequency

Adrenocortical carcinoma

Neuroblastoma

2

2

Ampullary carcinoma

1

Small cell lung cancer

2

3

ON

Colorectal cancer

Alteration frequency

Mixed cancer types

Cutaneous melanoma

30%

Adrenocortical carcinoma

Head and neck cancer

2

Melanoma

15

CNS cancer

Bladder cancer

16

Small cell lung cancer

Melanoma

2

9

ARTICLE IN PRESS

Mature B-cell neoplasm

Prostate cancer

O

Head and neck cancer

Germ cell tumor

Alteration frequency

Adrenocortical carcinoma

Pheochromocytoma Renal cell carcinoma -

Esophagogastric cancer

Small cell lung cancer

1

Cervical cancer

ZFPM1

Esophagogastric cancer

2

2

CNS cancer

Skin cancer, non-melanoma -

Hepatobiliary cancer

1

Cutaneous melanoma

Non-Hodgkins’s lymphoma

Non-Hodgkins’s lymphoma

2

Melanoma

Bladder cancer

Pancreatic cancer

Endometrial cancer

Alteration frequency

Adrenocortical carcinoma

Endometrial cancer-

Breast cancer

Melanoma

Esophagogastric cancer

Bone cancer

Ovarian cancer

Cutaneous melanoma

Pliocytic astrocytoma- Colorectal cancer

CRIPAK

Cutaneous melanoma

Monoclonal B-cell lymphocytosis

Colorectal cancer

Endometrial cancer

3

EL

Skin canceer, non-melanoma

Multiple myeloma

Thyroid cancer

Fig. 2

?

Endometrial cancer

Skin cancer, non-melanoma

Breast cancer Mesothelioma

Renal cell carcinoma

Cervical cancer

Bladder cancer

Soft tissue sarcoma

Please cite this article as: C.S. Rahane, A. Kutzner and K. Heese, Establishing a human adrenocortical carcinoma (ACC)-specific gene

Alteration frequency

Esophagal stomach cancer, NOS

Non-small cell lung cancer

Leukemia

mutation signature, Cancer Genetics, https://doi.org/10.1016/j.cancergen.2018.10.005

Cutaneous melanoma

Ampullary carcinoma

Cervical cancer

Alteration frequency

Small cell lung cancer

Colorectal cancer

GARS

Pancreatic cancer

2

2

Small cell lung cancer

Non-small cell lung cancer

Non-small cell lung cancer Salivary gland cancer

2

Melanoma

Esophagogastric cancer Multiple myeloma-

Non-Hodgkins’s lymphoma

Glioma

20%

25%

2

7

Skin canceer, non-melanoma Adrenocortical carcinoma

Hepatobiliary cancer

Pancreatic cancer Embryonal tumor

Leukemia

Je

Non-Hodgkins’s lymphoma Head and neck cancer

Cutaneous melanoma

Non-small cell lung cancer

Thyroid cancer

2

Skin canceer, non-melanoma

Ampullary carcinoma

Breast cancer

Soft tissue sarcoma

Cancer of unknown primary

Endometrial cancer

Germ cell tumor

DGKZ

Mature B-cell neoplasm

Hepatobiliary cancer Ovarian cancer

Prostate cancer

Esophagogastric cancer

Wilm’s tumor

Alteration frequency

Adrenocortical carcinoma

Adrenocortical carcinoma

Cutaneous melanoma

Melanoma

Head and neck cancer

Thymic tumor

Peripheral nervous system

30

75

Melanoma

Small cell lung cancer CNS cancer

High grade glioma K27Mut

Non-small cell lung cancer

Renal cell carcinoma

2

Bladder cancer

CNS cancer

Esophagogastric cancer Cervical cancer

Colorectal cancer

2

1

Endometrial cancer

High grade glioma K27Mwt

Prostate cancer

Peripheral nervous system

Endometrial cancer

Bladder cancer

Bone cancer

Prostate cancer

Esophagogastric cancer

Colorectal cancer

Pancreatic cancer

Adrenocortical carcinoma

Colorectal cancer

Prostate cancer, NOS Hepatobiliary cancer

Mature B-cell neoplasm Multiple myeloma

KCNH7

Breast cancer

Hepatobiliary cancer

Embryonal tumor wth rosettes

Non-Hodgkins’s lymphoma

Bladder cancer

Renal cell carcinoma

Embryonal tumor

Prostate cancer, NOS

Bone cancer

LRIG1

Non-small cell lung cancer

Prostate cancer

Mature B-cell neoplasm

Hepatobiliary cancer Cervical cancer

Renal cell carcinoma

CNS cancer

Ampullary carcinoma

Thymic tumor Bone cancer

Ovarian cancer

Bladder cancer

Non-Hodgkins’s lymphoma

Non-Hodgkins’s lymphoma

Leukemia

High grade glioma

Cervical cancer

Multiple myeloma

Head and neck cancer Prostate cancer, NOS Breast cancer

NOS3

Breast cancer

Glioma

Small cell lung cancer

Ovarian cancer

Salivary gland cancer

Alteration frequency

Pancreatic cancer

100%

Esophagogastric cancer

Pancreatic cancer

Mesothelioma-

Glioma

Skin cancer, non-melanoma

Pancreatic cancer Germ cell tumor

80%

ZNF517

61

Hepatobiliary cancer

20%

Mature B-cell neoplasm

Prostate cancer

Glioma

Ovarian cancer Soft tissue sarcoma

2

Cutaneous melanoma

Non-small cell lung cancer

Lymphoid cancer

Colorectal cancer

Skin cancer, non-melanoma

Renal cell carcinoma

Germ cell tumor Leukemia

Alteration frequency

Adrenocortical carcinoma

Endometrial carcinoma

Small cell lung cancer

2

Ependymoma supratentorial Adrenocortical carcinoma Ampullary carcinoma.

Non-small cell lung cancer

Head and neck cancer

3

1

2

Endometrial cancer Cervical cancer

Pancreatic cancer

Renal cell carcinoma

2

Embryonal tumor

Breast cancer

Mixed cancer types

Cutaneous melanoma

Bladder cancer

Glioma

Esophagogastric cancer

Ovarian cancer

Alteration frequency

Small cell lung cancer

Cutaneous melanoma

TOLLE

Head and neck cancer

GOLGA4

Leukemia

Adrenocortical carcinoma

Medulloblastoma Ependymoma infratentorial Mature B-cell neoplasm

Melanoma

3

Skin canceer, non-melanoma

Salivary gland cancer

High grade glioma K27Mwt

Prostate cancer

2

Ampullary carcinoma

7

Osteosarcoma

High grade glioma, K27Mut Bladder cancer

Breast cancer

Esophagogastric cancer

Renal cell carcinoma

Adrenocortical carcinoma

High grade glioma, K27Mwt

Cutaneous melanoma

Melanoma CNS cancer

Melanoma

Leukemia

High grade glioma K27Mwt

Esophagogastric cancer Endometrial cancer

Monoclonal B-cell lymphocytosis

ADAMTSL4

Prostate cancer, NOS-

Glioma

Prostate cancer, NOS

Colorectal cancer Hepatobiliary cancer

Colorectal cancer

Hepatobiliary cancer

Endometrial cancer

Bone cancer

Colorectal cancer

Cervical cancer

Head and neck cancer Ampullary carcinoma

Medulloblastoma

Skin cancer non-melanoma Head and neck cancer

Head and neck cancer

Germ cell tumor

Pheochromocytoma Pancreatic cancer Bladder cancer

Non-small cell lung cancer

Skin canceer, non-melanoma

Small cell lung cancer

Small cell lung cancer

ZNRF3

Pheochromocytoma Renal cell carcinoma

Melanoma

Mature B-cell neoplasm

Multiple myeloma

Thymic tumor

Breast cancer

Mesothelioma

Bone cancer

Non-Hodgkins’s lymphoma Ovarian cancer

Prostate cancer

Hepatobiliary cancer

Non-small cell lung cancer Bladder cancer

Non-small cell lung cancer

Prostate cancer

Renal cell carcinoma

Glioma

Renal cell carcinoma

Prostate cancer Germ cell tumor

Pancreatic cancer

Cervical cancer

Non-Hodgkins’s lymphoma

Breast cancer

Thyroid cancer Breast cancer

Non-Hodgkins’s lymphoma

Ovarian cancer

Glioma

Pancreatic cancer

Cancer of unknown primary-

Glioma

Ovarian cancer

Thyroid cancer Leukemia

Soft tissue sarcoma

Leukemia

Salivary gland cancer

ing ACC cells (Fig. 1(C)). Frequency of mutations in LRIG1, CRIPAK and GARS were slightly higher in survivors as com- pared to total number of patients, while the frequency of mu- tations in ZNF517 remained similar both in total number of patients as well as survivors. However, the frequency of mu- tations in TP53 was lower in survivors than in overall patients (Fig. 1(D)-(E)). We compared the list of frequently mutated genes overall with genes which were frequently mutated only in deceased patients. We found six genes, diacylglycerol ki- nase zeta (DGKZ), golgin A4 (GOLGA4), potassium voltage- gated channel subfamily H member 7 (KCNH7), ADAMTS like 4 (ADAMTSL4), nitric oxide synthase 3 (NOS3), and zinc and ring finger 3 (ZNRF3), which showed a significant num- ber of mutations and adding to the novel ACC-specific gene mutation signature (Fig. 1(G)). We also observed a potential gender effect: mutations in DGKZ, GOLGA4 and NOS3 were observed mainly in women, with female to male occurrence ratios ranging from 3:1 to 4:0 (Supplementary File 1). This hints at a gender-specific role of these genes in ACC. Patient data and list of genes mutated solely in deceased patients is provided in Supplementary File 1.

We analyzed whether mutations in these top-five mutated genes, as well as genes observed only in deceased patients, were specific to ACC by checking alteration frequency of these genes across all human cancer tissue types available on the cBioPortal database. We found that alteration fre- quency of the ZFPM1, LRIG1, CRIPAK, GARS, and ZNF517 was significantly higher in ACC as compared to all other cancer types. This indicates that these five genes were specifically altered in ACC but not in other cancer tissues (Fig. 2). Moreover, the mutations in DGKZ and ZNRF3 were most frequently observed in ACC too, while GOLGA4, KCNH7, ADAMTSL4, and NOS3 did not show increased ACC-specific alteration frequencies (Fig. 2).

Moreover, we found that mutations in the five highest mu- tated genes were localized only to one or two positions in their amino acid (AA) sequences - particularly observed in ZFPM1, LRIG1, GARS and ZNF517; however, this AA loca- tion specificity was not observed in CRIPAK. These locations were specific to ACC and were not observed in any other can- cer tissue (Fig. 3). Recurrent mutations at the same site pro- vided further evidence of the ACC-specific oncogenic poten- tial of these mutations for this specific gene set. The deceased patient-specific gene set did not show such a site-specificity for mutations (Supplementary File 1).

Gene-specific survival analysis in ACC

Kaplan Meier survival curves for ZFPM1, LRIG1, CRIPAK, GARS, and ZNF517 show no change between patients with and without alterations in ACC (Supplementary File

1). However, survival curves for DGKZ, GOLGA4, KCNH7, ADAMTSL4, NOS3 and ZNRF3 show extensive differences between patients with and without mutations in them (Fig. 4). The low number of samples containing mutations in selected genes could be one reason for the large differences between curves.

Application of the ‘20/20’ rule to validate the oncogenic potential of selected genes

Applying the ‘20/20’ test to the mutation data from ACC also confirmed our hypothesis that the genes of interest are po- tential oncogenes. This was additionally verified by the Mut- sig Q-value data from cBioPortal (Supplementary File 1). We used genes reported previously as potential drivers in ACC [14,16] as positive controls and compared their 20/20 re- sults and Mutsig Q-values with our gene set of interest. We found that ZFPM1, LRIG1, GARS, and ZNF517 show lower Q-values than the reported oncogenes. DGKZ qualifies as a potential oncogene by the 20/20 test, however, Q-values for DGKZ were unavailable. CRIPAK had a higher Q-value, even though it fits the parameters for an oncogene by the 20/20 test, while GOLGA4, KCNH7, ADAMTSL4, NOS3 and ZNRF3 were neither oncogenes nor tumor suppressors nor had any Q-value data. This could be due to the low number of muta- tions in the genes. Low Q-values for the genes of interest indi- cate that the genes are significantly mutated in ACC, lending credence to their status as potential ACC-specific oncogenes.

Discussion

Although our understanding of the molecular mechanism driv- ing ACC has advanced, the oncogenic mutations leading to ACC are poorly understood [1]. Thus, we aimed at unravelling novel ACC-specific targets and determined a novel gene set of six genes that were significantly and specifically mutated in ACC.

Kaplan-Meier curves for the gene set from deceased pa- tients show dramatic differences between patients with and without mutations. This indicates that mutations in DGKZ, GOLGA4, KCNH7, ADAMTSL4, NOS3 or ZNRF3 dramati- cally worsen prognosis in ACC and are thus genes of special interest. Although the number of samples containing muta- tions in these genes is very low and could be a possible reason for differences in survival curves, the data correlate very well with the control gene TP53. In addition, GOLGA4, KCNH7, ADAMTSL4, NOS3 or ZNRF3 are not specifically mutated in ACC, nor do they fit 20/20 and Mutsig parameters. Hence we included DGKZ, but not the others, as part of our ACC-specific gene mutation signature. The specific integrated role of this

Fig. 2 Mutation frequency of the selected genes in ACC compared across all available (at cBioPortal) human cancer tissue types (adapted from cBioPortal). Cancer types lacking mutations in the specified genes have not been shown. Bars in red color indicate ACC. Mutation frequency in the five genes with highest number of mutations in ACC (A). Almost 40% of samples in ACC show mutations in ZFPM1, which is significantly higher than in any other cancer type. Similarly, mutations in LRIG1, CRIPAK, ZNF517 and GARS are significantly higher in ACC as compared to all other human cancer tissue types. This indicates that alterations in this particular five-gene set may be important for ACC oncogenesis. Mutation frequency in the top six genes observed only in deceased ACC patients (B). Mutation frequencies of these genes is relatively low in ACC (compared with genes shown in (A)), however, DGKZ and ZNRF3 are altered more in ACC as compared to other cancer types. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

Novel ACC-specific gene signature

Fig. 3 Comparative gene mutation analysis between site-specific AA mutations in ACC and all other available (at cBioPortal) human cancer tissue types. The top five genes with the highest number of gene mutations in ACC are shown. Mutations in ZFPM1 occur primarily at the E444 position in ACC, whereas in other cancer tissues the mutations are not localized to a specific site. Similarly, in ACC, mutations in LRIGI, ZNF517 and GARS occur mainly at the L24, V349 and P42 positions, respectively. Mutations in ZFPM1, LRIG1, ZNF517 and GARS occur at the same site in ACC, while occurring at various sites in all other cancer tissue types. Recurrent mutations at the same site in a gene imply that the location is a mutation hotspot in ACC, and thus could be an oncogenic trigger. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

Mutations in amino acid sequences of selected genes across all cancer tissues

Mutations in amino acid sequences of selected genes in ACC

ZFPM1

ZFPM1

# of mutations

66

E444Afs*227/D/_L446del

# of mutations

66

E444Afs#227/D/_L446del

0

0

0

200

400

600

800

1000

0

200

400

600

800

1000

Amino acid position

Amino acid position

L24V

LRIG1

31

L24V

LRIG1

# of mutations

# of mutations

31

0

1.

0

0

200

400

600

800

1000 1093

0

200

400

600

800

1000 1093

Amino acid position

Amino acid position

C143R

CRIPAK

CRIPAK

# of mutations

10

# of mutations

8

C143R

0

01

0

100

200

300

400 446

0

100

200

300

400 446

Amino acid position

Amino acid position

ZNF517

ZNF517

# of mutations

33

V349A

# of mutations

33

V349A

7

0

01

0

100

200

300

400

492

0

100

200

300

400

492

Amino acid position

Amino acid position

# of mutations

P42A

GARS

GARS

33

# of mutations

33

P42A

0

0

0

200

400

600

739

0

200

400

600

739

Amino acid position

Amino acid position

zinc finger motif

histidyl, glycyl, threonyl and prolyl anticodon binding domain

leucine-rich repeat

Krüppel associated box

leucine rich repeat C-terminal domain

immunoglobulin I-set domain

helix-turn-helix domain

tRNA synthetase class II core domain

ACC-specific gene mutation signature in a convergent ACC cancer cell signaling mechanism is as yet unclear. The sig- naling pathways involving our identified gene set as well as those from Zheng et al. [16] can be interlinked (Fig. 5).

Downstream signaling by tyrosine kinase receptors like EGFR activate multiple pathways including the mitogen- activated protein kinase (MAPK), phosphatidylinositol-4, 5- bisphosphate 3-kinase (PI3K) and WNT signaling path- ways, and there is extensive cross-regulation between these pathways [48]. PAK1 belongs to the p21-activated ser- ine/threonine kinase family and plays a critical role in link- ing multiple signaling pathways, including cell polarity, actin cytoskeleton reorganization and cellular proliferation [49,50]. PAK1 interacts directly with Rac family small GTPase 1 (RAC1) and PI3K, effecting downstream signaling by phos- phorylating other kinases [51]. Increase in PAK1 expres-

sion has been observed in various cancers [49,51]. PAK1 is a promising target for therapy with both ATP-competitive as well as allosteric inhibitors being screened for treatment [52]. The functional role of CRIPAK is inhibition of the PAK1 kinase [43] and disruption in the function of CRIPAK via mu- tations would lead to a dysregulation of PAK1, thereby en- hancing cellular proliferation (Fig. 5). Activation of the PI3K pathway leads to suppression of TP53 via MDM2 activation. AKT also phosphorylates and inhibits glycogen synthase ki- nase 3 (GSK3), thereby stabilizing downstream targets of GSK3 such as the G1 cyclins and TFs such as MYC proto- oncogene (MYC) and Jun proto-oncogene (JUN) [53]. AKT also phosphorylates CDKN1B, inactivating its cell-cycle pre- vention effect [53]. GSK3 phosphorylates CTNNB1, marking it for degradation. Mutation in CTNNB1 makes degradation im- possible; thereby cause constitutive activation of target genes

ARTICLE IN PRESS

Fig. 4

100%

DGKZ

100%

GOLGA4

90%

90%

P-value: 6.316e-5

P-value: 0.00158

80%

80%

Overall survival

Overall survival

70%

70%

60%

60%

50%

50%-

40%

40%

30%

30%

20%

20%

10%

10%

0%

0

10

20

30

40

50

60

70

80

90

100

110

120

130

) 140

150

0%

0

10

20

30

40

50

60

70

80

90

0 100 110 120 130 140 150

Months survival

Months survival

100%

KCNH7

100%

ADAMTSL4

90%

90%

P-value: 2.783e-4

P-value: 4.601e-6

80%

80%

Overall survival

Overall survival

70%

70%

60%

60%

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0

10

20

30

40

50

60

70

80

90

100

110

120

130

140

150

0%

0

10

20

30

40

50

60

70

80

90

100

110

120 1

130

140

150

Months survival

Months survival

100%

NOS3

100%

ZNRF3

90%

90%

P-value: 0.00420

P-value: 4.975e-6

80%

80%

Overall survival

Overall survival

70%

70%

60%

60%

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0

10

20

30

40

50

60

70

80

90

100

110

120

130

140

150

0%

0

10

20

30

40

50

60

70

80

90 100

110

120

130

140

150

Months survival

Months survival

100%

TP53

100%

MEN1

90%

P-value: 0.000682

90%

P-value: 0.0564

80%

80%

Overall survival

Overall survival

70%

70%

60%

60%

50%

50%

40%

40%-

30%

30%

20%

20%-

10%

10%

0%

0

10

20

30

40

50

60

70

80

90

100

110

120

130

0 140

150

0%

0

10

20

30

40

50

60

70

80

90

100

120

130

140

150

Months survival

Months survival

Please cite this article as: C.S. Rahane, A. Kutzner and K. Heese, Establishing a human adrenocortical carcinoma (ACC)-specific gene mutation signature, Cancer Genetics, https://doi.org/10.1016/j.cancergen.2018.10.005

Novel ACC-specific gene signature

such as MYC and cyclin D1 (CCND1). RPL22 is a 60S riboso- mal subunit protein and activates TP53 expression by binding and inactivating MDM2, thereby suppressing cancer cell sur- vival [46]. Inactivation of RPL22 and mutation in TP53 thus acts as a ‘double whammy’ in enhancing proliferation of can- cer cells. The role of TP53 in oncogenesis is well established. TP53 can induce both cell cycle arrest and apoptosis, and mutations in TP53 are the most common ones observed in tumors [42]. PRKAR1A is the regulatory subunit of cAMP de- pendent protein kinase A (PKA) and mutations in PRKAR1A have been known to cause an autosomal dominant neopla- sia syndrome called Carney complex [41]. The downstream signaling of EGFR also regulates CTNNB1, via the mucin-1 (MUC1) protein. Decrease in MUC1 expression suppresses EGFR-dependent tumorigenesis and CCND1 protein expres- sion [54]. The menin (MEN1) protein has both tumor sup- pressive and proliferative functions. Mutations in MEN1 lead to multiple endocrine neoplasia, which causes tumors in en- docrine glands, inactivating mutations in MEN1 are linked to inhibition of CDKN1B, and MEN1 has been reported to sup- press AKT kinase activity in murine cells [55].

LRIG1 is a transmembrane protein which inhibits EGFR kinase receptor signaling by binding to it via its ectodomain [56]. LRIG1 functions as a tumor suppressor in multiple can- cers [57-59] and increased expression of the (non-mutated) proto-oncogene LRIG1 is associated with good prognosis in breast, bladder, lung cancer, melanoma and glioma [60]. Inac- tivating mutations in LRIG1 lead to a ‘switch on’ of the EGFR kinase-signaling pathway, resulting in high cellular prolifer- ation. LRIG1 has already been tested as a therapeutic for glioma. LRIG1 acts as a sensitizing agent for chemotherapeu- tics and injection of soluble LRIG1 has been reported to in- hibit glioblastoma [61-64]. Therefore, LRIG1 could be a highly promising target for immunotherapy for ACC, given that it acts as an inhibitor of EGFR kinase. Ex-vivo sensitized dendritic cells to mutant LRIG1 could be targeted specifically to ACC cells expressing mutant LRIG1 to mediate an endogenous immune response to fight ACC. Antibodies targeting mutant LRIG1 could be generated, thereby killing the ACC cells via a cell-mediated cytotoxic processes. Another approach would be to target other genes on combination with LRIG1. FAM72 is a novel proliferation marker, which is overexpressed in non- neural cancer tissues [65,66]. Singling out FAM72 would me- diate mitotic damage in all growing cells, however, target- ing mutant LRIG1 and FAM72 together will ensure that only cancerous cells expressing mutated LRIG1 and overexpress- ing FAM72 will be targeted while leaving the other non- cancerous cells unharmed.

The zinc finger protein ZFPM1, also known as FOG1, acts as a cofactor with the transcription factor GATA1 and regu- lates differentiation and proliferation of blood cells at various stages [67,68]. ZFPM1 has been reported to be upregulated in myeloid leukemia [69] but no evidence of mutational effects have been reported. Both GATA1 and ZFPM1 interact with

the nucleosome remodeling and deacetylase (NuRD) com- plex, which is comprised of retinoblastoma binding protein 4 (RBBP4), ATPases and histone deacetylases. This protein complex is critical in transcriptional repression of ZFPM1- GATA1 target genes [44,70]. Loss of function mutations in ZFPM1 could prevent differentiation, thereby activating pro- liferative mechanisms in the adrenocortical cells. RBBP4, in turn, forms a complex with RB1 and regulates the function of E2F1 [71]. This directly affects the cell cycle, as E2F1 is a transcriptional activator for G1 and S-phase genes.

GARS mutations are typically associated with the neuro- pathic Charcot-Marie-Tooth disorder [72]. However, there is some evidence to suggest that it also plays a role in the adap- tive immune system by suppressing MAPK signaling in tumor cells [73]. GARS dysregulation has also been reported in a va- riety of cancers [74,75] though oncogenic mutations have not yet been reported. Recent evidence suggests that GARS acts as a chaperone for the proteins directly involved in the ned- dylation pathway [45]. Neddylation is the conjugation of the NEDD8 protein, via GARS, to its targets and is critical in cell cycle regulation: the NEDD8-cullin complex activates cullin- RING ubiquitin ligases, which degrade the p27 cell-cycle in- hibitor and continues cell cycle progression [76]. Inhibition of GARS thus leads to reduction of metastasis [45]. This is of high interest as gain of function mutations in GARS would lead to higher levels of neddylation, thereby increasing cellu- lar proliferation. This makes GARS a very attractive target for therapeutic applications.

There is little evidence about the function of the other zinc finger protein ZNF517. It forms part of the interactome of heat shock 90 kDa proteins (HSP90) [77] so it may be involved in protein folding or intracellular transport, but there is no other report of its cellular function. ZNF517 is a zinc-finger protein, which are the largest family of transcription factors. ZNFs work as transcriptional activators or repressors and play signifi- cant roles in cellular growth, maturation and metabolism [78]. ZNF517 contains Kruppel associated box (KRAB) domains, indicating its potential function as a transcriptional repressor. The KRAB domain binds to other corepressors via direct pro- tein interactions leading to transcriptional repression of genes [79]. KRAB-ZNFs play important roles in cellular differentia- tion, apoptosis and cancer by acting as tumor suppressors and promoters of apoptosis [80]. Mutations in ZNF517 could lead to constitutive expression of downstream signaling path- ways leading to proliferation. DGKZ is a lipid kinase involved in phosphatidic acid synthesis. DGKZ knockdown has been re- ported to inhibit proliferation in glioma cells [81]. DGKZ inhibits RAS guanyl-releasing protein (RASGRP1), thereby modulat- ing RAS activity [47]. Inactivating mutations in DGKZ could lead to increased phosphorylation of RAS, thereby causing continuous activation of the MAPK or PI3K signaling path- ways, ending in cellular proliferation. Mutations in these genes drastically worsen the prognosis, and therefore must act in

Fig. 4 Survival plots showing the prognosis of ACC patients with somatic mutations in DGKZ, GOLGA4, KCNH7, NOS3, ADAMTSL4 and ZNRF3, respectively (adapted from cBioPortal). DGKZ, GOLGA4, KCNH7, NOS3, ADAMTSL4 and ZNRF3 are the genes with highest number of somatic mutations observed only in deceased ACC patients. All patients with mutations in these genes show worse prognosis compared to all other ACC patients. MEN1 and TP53 survival plots are shown as control comparison. Red line: cases with alterations in query gene, blue line: cases without alterations in query gene. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

Please cite this article as: C.S. Rahane, A. Kutzner and K. Heese, Establishing a human adrenocortical carcinoma (ACC)-specific gene mutation signature, Cancer Genetics, https://doi.org/10.1016/j.cancergen.2018.10.005

Fig. 5 ACC-specific gene-mutation signature-activated cell signaling pathways with highlighted proto-oncogenes involved in tumori- genesis. Key to illustration is provided below the figure. Genes mentioned in Zheng et al. [16] have also been illustrated for comparison. LRIG1 binds to EGFR via its ectodomain, leading to suppression of EGFR-mediated signaling [39,40]. Loss-of function mutations in LRIG1 may lead to activation of the EGFR kinase-signaling pathway, thereby leading to constitutive expression of the PI3K and MAPK pathways. PI3K downstream signaling also reduces TP53 activity via MDM2 binding and PAK1 activation leads to cellular proliferation via MAPK signaling. Mutations in PRKR1A lead to an adrenal neoplasia called Carney complex [41]. Mutations in TP53 lead to inhibi- tion of its tumor suppressor and pro-apoptotic role, thereby converting it into an oncogene [42]. Inactivating mutations in CRIPAK may lead to dysregulation of PAK1 [43] resulting in continuous downstream signaling of the MAPK and PI3K pathways. MEN1 mutations will activate AKT and CTNNB1, while also inactivating cyclin-dependent kinase inhibitor 1B (CDKN1B), also known as p27. CTNNB1 mutations lead to transcriptional activation of cell cycle activator like cyclin D. ZFPM1 forms a complex with RBBP4 and GATA1, and regulates cellular differentiation [44]. Mutations in ZFPM1 may dysregulate differentiation and cause abnormal growth of cells. GARS plays a critical role in neddylation and thereby regulates cellular proliferation [45]. The function and location of ZNF517 has not yet been described but it may act as a tumor suppressor. RPL22 inactivates MDM2, leading to TP53 expression [46]. DKGZ modulates cell cycle via DAG metabolization and RASGRP1 inhibition [47]. G0, quiescent or differentiated stage; G1, Gap1 phase; G2, Gap2 phase; M, Mitotic phase; S, Synthesis phase. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

WNT 1/3a/8

cell membrane

LRIG1

EGFR

LRP5/6

Frizzled receptor

DGKZ

polypeptide

PI3K

CAMP

PTEN

MUC1

RAS

GSK3B

GARS

PIP3

RPL22

PRKARIA

mRNA

RACI

BMP7

AKT

CTNNB1

chaperone

ribosome

GARS

NEDD8

RAF

PAK1

CRIPAK

MAP2K6

MAP3K7

NEDD8

MAP2K1/2

MAP3K1

cullin

MDM2

MAPK8

MAPK14

ring ligase

MAPK1/3

cullin

NEDD8

nucleus

MEN1

TP53

CDC25

S

CDKNIB

CDKNIA

Interphase

tumor

+ RB1

CTNNB1

suppression

G2

G1

ZNF517

regulation?

RBBP4

ZFPM1)GATAI

JUN FOS

Mitosis M

CDK4

GO

CCND

phosphatase

protein

multistep inhibition

GTPase

- direct inhibition

translocation

transcription factor

- direct activation

proto-oncogene from Zheng et al

kinase

----- + multistep activation

proto-oncogene from our study

conjunction with our novel gene set as well as other proto oncogenes such as MEN1, CTNNB1 and TP53.

EGFR signaling is the primary node through which the proliferative pathways can be initiated and most of the proto- oncogenes in ACC act downstream of EGFR. Inhibition of EGFR via LRIG1 is thus a key step in regulating, either par- tially or fully, the consequent signaling cascades. LRIG1 mu- tations would cause a continuous expression of the EGFR signaling cascade, thereby causing cellular proliferation. Sim- ilarly, mutations in GARS also serve to increase proliferation via a cascade that is independent of the PI3K/MAPK/WNT

signaling pathways. Mutations in our selected genes thus ap- pear to be more influential in ACC tumorigenesis than those described in earlier studies are.

Our study thus sheds new light on potential driver genes in this rare ACC cancer. Our findings describe a novel ACC- specific gene signature providing a gene set (ZFPM1, LRIG1, CRIPAK, GARS, ZNF517, and DGKZ) and validate its onco- genic potential, which could serve as a powerful therapeutic target.

Novel ACC-specific gene signature

Acknowledgments

This study was supported by Hanyang University by pro- viding a scholarship to C.S.R. and by the Basic Science Research Program through the National Research Founda- tion of Korea (NRF) funded by the Ministry of Education (2015R1D1A1A01057243 and 2016R1D1A1B03932599).

Competing interests

The authors declare no conflicts of interest

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.cancergen.2018. 10.005.

References

[1] Fassnacht M, Kroiss M, Allolio B. Update in adrenocortical car- cinoma. J Clin Endocrinol Metab 2013;98(12):4551-64.

[2] Kebebew E, et al. Extent of disease at presentation and outcome for adrenocortical carcinoma: have we made progress? World J Surg 2006;30(5):872-8.

[3] Kerkhofs TM, et al. Adrenocortical carcinoma: a popula- tion-based study on incidence and survival in the Netherlands since 1993. Eur J Cancer 2013;49(11):2579-86.

[4] Creemers SG, et al. Identification of mutations in cell-free circu- lating tumor DNA in adrenocortical carcinoma: a case series. J Clin Endocrinol Metab 2017;102(10):3611-15.

[5] Lalli E, Luconi M. The next step: mechanisms driving adrenocortical carcinoma metastasis. Endocr Relat Cancer 2018;25(2):R31-48.

[6] Lambert AW, Pattabiraman DR, Weinberg RA. Emerging biolog- ical principles of metastasis. Cell 2017;168(4):670-91.

[7] Else T, et al. Adrenocortical carcinoma. Endocr Rev 2013;35(2):282-326.

[8] Beuschlein F, et al. Major prognostic role of Ki67 in localized adrenocortical carcinoma after complete resection. J Clin En- docrinol Metab 2015;100(3):841-9.

[9] Pereira SS, et al. Mechanisms of endocrinology: cell cy- cle regulation in adrenocortical carcinoma. Eur J Endocrinol 2018;179(2):R95-R110.

[10] Berthon A, et al. Constitutive beta-catenin activation induces adrenal hyperplasia and promotes adrenal cancer development. Hum Mol Genet 2010;19(8):1561-76.

[11] Herrmann LJ, et al. TP53 germline mutations in adult pa- tients with adrenocortical carcinoma. J Clin Endocrinol Metab 2012;97(3):E476-85.

[12] Raymond VM, et al. Prevalence of germline TP53 mutations in a prospective series of unselected patients with adrenocortical carcinoma. J Clin Endocrinol Metab 2013;98(1):E119-25.

[13] Gaujoux S, et al. beta-catenin activation is associated with spe- cific clinical and pathologic characteristics and a poor outcome in adrenocortical carcinoma. Clin Cancer Res 2011;17(2):328- 336.

[14] Tissier F, et al. Mutations of beta-catenin in adrenocortical tu- mors: activation of the Wnt signaling pathway is a frequent event in both benign and malignant adrenocortical tumors. Cancer Res 2005;65(17):7622-7.

[15] Assie G, et al. Integrated genomic characterization of adreno- cortical carcinoma. Nat Genet 2014;46(6):607-12.

[16] Zheng S, et al. Comprehensive pan-genomic characterization of adrenocortical carcinoma. Cancer Cell 2016;29(5):723-36.

[17] Ragazzon B, et al. Transcriptome analysis reveals that p53 and -catenin alterations occur in a group of aggressive adrenocor- tical cancers. Cancer Res 2010;70(21):8276-81. doi:10.1158/ 0008-5472.CAN-10-2014.

[18] Salomon A, et al. Loss of beta-catenin in adrenocortical cancer cells causes growth inhibition and reversal of epithelial-to-mes- enchymal transition. Oncotarget 2015;6(13):11421-33.

[19] Gicquel C, et al. Molecular markers and long-term recurrences in a large cohort of patients with sporadic adrenocortical tumors. Cancer Res 2001;61(18):6762-7.

[20] Sbiera S, et al. High diagnostic and prognostic value of steroido- genic factor-1 expression in adrenal tumors. J Clin Endocrinol Metab 2010;95(10):E161-71.

[21] de Fraipont F, et al. Gene expression profiling of human adreno- cortical tumors using complementary deoxyribonucleic acid mi- croarrays identifies several candidate genes as markers of ma- lignancy. J Clin Endocrinol Metab 2005;90(3):1819-29.

[22] Giordano TJ, et al. Distinct transcriptional profiles of adrenocor- tical tumors uncovered by DNA microarray analysis. Am J Pathol 2003;162(2):521-31.

[23] Lawrence MS, et al. Discovery and saturation analy- sis of cancer genes across 21 tumour types. Nature 2014;505(7484):495-501.

[24] Vogelstein B, et al. Cancer genome landscapes. Science 2013;339(6127):1546-58.

[25] Cerami E, et al. The cBio cancer genomics portal: an open plat- form for exploring multidimensional cancer genomics data. Can- cer Discov 2012;2(5):401-4.

[26] Gao J, et al. Integrative analysis of complex cancer ge- nomics and clinical profiles using the cBioPortal. Sci Signal 2013;6(269):pl1.

[27] Lawrence MS, et al. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature 2013;499(7457):214-18.

[28] Greenman C, et al. Patterns of somatic mutation in human can- cer genomes. Nature 2007;446(7132):153-8.

[29] Goldman M, et al. Abstract 2584: The UCSC Xena system for cancer genomics data visualization and interpretation. Cancer Res 2017;77(13 Supplement):2584. doi:10.1158/1538-7445. AM2017-2584.

[30] Ho NT, et al. Cognitive functions: human vs. animal - 4:1 advan- tage |-FAM72-SRGAP2-|. J Mol Neurosci 2017;61(4):603-6.

[31] Ho NTT, Kutzner A, Heese K. Brain plasticity, cognitive func- tions and neural stem cells: a pivotal role for the brain-spe- cific neural master gene |-SRGAP2-FAM72-|. Biol Chem 2017;399(1):55-61.

[32] Kutzner A, et al. All-or-(N)One - an epistemological character- ization of the human tumorigenic neuronal paralogous FAM72 gene loci. Genomics 2015;106(5):278-85.

[33] Rahane CS, Kutzner A, Heese K. A cancer tissue-specific FAM72 expression profile defines a novel glioblastoma mul- tiform (GBM) gene-mutation signature. J Neurooncol 2018. doi:10.1007/s11060-018-03029-3.

[34] Iverson GL. Z scores. In: Kreutzer JS, DeLuca J, Caplan B, ed- itors. Encyclopedia of clinical neuropsychology. New York, NY: Springer New York; 2011. p. 2739-40.

[35] Scholzen T, Gerdes J. The Ki-67 protein: from the known and the unknown. J Cell Physiol 2000;182(3):311-22.

[36] Team BD. Bokeh: python library for interactive visualization. 2014. Available at: http://www.bokeh.pydata.org.

[37] Bland JM, Altman DG. The logrank test. BMJ 2004;328(7447):1073.

[38] Heese K. The protein p17 signaling pathways in cancer. Tumour Biol 2013;34(6):4081-7.

Please cite this article as: C.S. Rahane, A. Kutzner and K. Heese, Establishing a human adrenocortical carcinoma (ACC)-specific gene mutation signature, Cancer Genetics, https://doi.org/10.1016/j.cancergen.2018.10.005

C.S. Rahane, A. Kutzner and K. Heese

[39] Kou C, et al. LRIG1, a 3p tumor suppressor, represses EGFR signaling and is a novel epigenetic silenced gene in colorectal cancer. Biochem Biophys Res Commun 2015;464(2):519-25.

[40] Stutz MA, et al. LRIG1 negatively regulates the oncogenic EGF receptor mutant EGFRvill. Oncogene 2008;27(43):5741-52.

[41] Kirschner LS, et al. Genetic heterogeneity and spectrum of muta- tions of the PRKAR1A gene in patients with the carney complex. Hum Mol Genet 2000;9(20):3037-46.

[42] Kandoth C, et al. Mutational landscape and significance across 12 major cancer types. Nature 2013;502(7471):333-9.

[43] Talukder AH, Meng Q, Kumar R. CRIPak, a novel endogenous Pak1 inhibitor. Oncogene 2006;25(9):1311-19.

[44] Miccio A, et al. NuRD mediates activating and repressive func- tions of GATA-1 and FOG-1 during blood development. EMBO J 2010;29(2):442-56.

[45] Mo Z, et al. Neddylation requires glycyl-tRNA synthetase to pro- tect activated E2. Nat Struct Mol Biol 2016;23(8):730-7.

[46] Cao B, et al. Cancer-mutated ribosome protein L22 (RPL22/eL22) suppresses cancer cell survival by blocking p53-MDM2 circuit. Oncotarget 2017;8(53):90651-61.

[47] Topham MK, Epand RM. Mammalian diacylglycerol kinases: molecular interactions and biological functions of selected iso- forms. Biochim Biophys Acta 2009;1790(6):416-24.

[48] Mendoza MC, Er EE, Blenis J. The Ras-ERK and PI3K-mTOR pathways: cross-talk and compensation. Trends Biochem Sci 2011;36(6):320-8.

[49] Ong CC, et al. p21-activated kinase 1: PAK’ed with potential. Oncotarget 2011;2(6):491-6.

[50] Guo D, et al. A Rac-cGMP signaling pathway. Cell 2007;128(2):341-55.

[51] Kumar R, Gururaj AE, Barnes CJ. p21-activated kinases in can- cer. Nat Rev Cancer 2006;6(6):459-71.

[52] Semenova G, Chernoff J. Targeting PAK1. Biochem Soc Trans 2017;45(1):79-88.

[53] Manning BD, Toker A. AKT/PKB signaling: navigating the net- work. Cell 2017;169(3):381-405.

[54] Pochampalli MR, Bitler BG, Schroeder JA. Transforming growth factor alpha dependent cancer progression is modulated by Muc1. Cancer Res 2007;67(14):6591-8.

[55] Borsari S, et al. Loss of p27 expression is associated with MEN1 gene mutations in sporadic parathyroid adenomas. Endocrine 2017;55(2):386-97.

[56] Gur G, et al. LRIG1 restricts growth factor signaling by en- hancing receptor ubiquitylation and degradation. EMBO J 2004;23(16):3270-81.

[57] Yokdang N, et al. LRIG1 opposes epithelial-to-mesenchymal transition and inhibits invasion of basal-like breast cancer cells. Oncogene 2016;35(22):2932-47.

[58] Neirinckx V, Hedman H, Niclou SP. Harnessing LRIG1-mediated inhibition of receptor tyrosine kinases for cancer therapy. Biochim Biophys Acta 2017;1868(1):109-16.

[59] Mao F, et al. Lrig1 is a haploinsufficient tumor suppressor gene in malignant glioma. Oncogenesis 2018;7(2):13.

[60] Rouam S, Moreau T, Broet P. Identifying common prognostic factors in genomic cancer studies: a novel index for censored outcomes. BMC Bioinf 2010;11(1):150.

[61] Johansson M, et al. The soluble form of the tumor suppressor Lrig1 potently inhibits in vivo glioma growth irrespective of EGF receptor status. Neuro Oncol 2013;15(9):1200-11.

[62] Qi XC, et al. LRIG1 dictates the chemo-sensitivity of temozolo- mide (TMZ) in U251 glioblastoma cells via down-regulation of EGFR/topoisomerase-2/Bcl-2. Biochem Biophys Res Commun 2013;437(4):565-72.

[63] Wang X, et al. LRIG1 enhances cisplatin sensitivity of glioma cell lines. Oncol Res 2012;20(5-6):205-11.

[64] Mao F, et al. LRIG proteins in glioma: functional roles, molecu- lar mechanisms, and potential clinical implications. J Neurol Sci 2017;383:56-60.

[65] Guo C, et al. Ugene, a newly identified protein that is commonly overexpressed in cancer and binds uracil DNA glycosylase. Can- cer Res 2008;68(15):6118-26.

[66] Nehar S, Mishra M, Heese K. Identification and characterisation of the novel amyloid-beta peptide-induced protein p17. FEBS Lett 2009;583(19):3247-53.

[67] Tanaka M, et al. Differentiation status dependent function of FOG-1. Genes Cells 2004;9(12):1213-26.

[68] Yang HY, et al. The suppression of zfpm-1 accelerates the ery- thropoietic differentiation of human CD34+ cells. Biochem Bio- phys Res Commun 2007;353(4):978-84.

[69] Marcucci G, et al. Prognostic significance of, and gene and mi- croRNA expression signatures associated with, CEBPA muta- tions in cytogenetically normal acute myeloid leukemia with high- -risk molecular features: a cancer and leukemia group b study. J Clin Oncol 2008;26(31):5078-87.

[70] Lejon S, et al. Insights into association of the NuRD complex with FOG-1 from the crystal structure of an RbAp48.FOG-1 complex. J Biol Chem 2011;286(2):1196-203.

[71] Li L, et al. Epigenetic modification of MiR-429 promotes liver tumour-initiating cell properties by targeting Rb binding protein 4. Gut 2015;64(1):156-67.

[72] He W, et al. CMT2D neuropathy is linked to the neo- morphic binding activity of glycyl-tRNA synthetase. Nature 2015;526(7575):710-14.

[73] Park MC, et al. Secreted human glycyl-tRNA synthetase im- plicated in defense against ERK-activated tumorigenesis. Proc Natl Acad Sci USA 2012;109(11):E640-7.

[74] Kim S, You S, Hwang D. Aminoacyl-tRNA synthetases and tumorigenesis: more than housekeeping. Nat Rev Cancer 2011;11(708-18).

[75] Wasenius VM, et al. Hepatocyte growth factor receptor, ma- trix metalloproteinase-11, tissue inhibitor of metalloproteinase-1, and fibronectin are up-regulated in papillary thyroid carci- noma: a cDNA and tissue microarray study. Clin Cancer Res 2003;9(1):68-75.

[76] Pagano M, et al. Role of the ubiquitin-proteasome pathway in regulating abundance of the cyclin-dependent kinase inhibitor p27. Science 1995;269(5224):682-5.

[77] Prince T, Williams H. 494 Heat shock protein 90 interactome is highly altered in adrenocortical carcinoma. European Urology Supplements 2016;15(3):e494.

[78] Jen J, Wang YC. Zinc finger proteins in cancer progression. J Biomed Sci 2016;23(1):53.

[79] Friedman JR, et al. KAP-1, a novel corepressor for the highly conserved KRAB repression domain. Genes Dev 1996;10(16):2067-78.

[80] Lupo A, et al. KRAB-zinc finger proteins: a repressor fam- ily displaying multiple biological functions. Curr Genomics 2013;14(4):268-78.

[81] Diao J, et al. Loss of diacylglycerol kinase-zeta inhibits cell proliferation and survival in human gliomas. Mol Neurobiol 2016;53(8):5425-35.