Medicine ®
OPEN
Identification of prognostic metabolic genes in adrenocortical carcinoma and establishment of a prognostic nomogram
A bioinformatic study
Qing Chen, MMª(D, Ziyu Ren, MMª, Dongfang Liu, MDª, Zongrui Jin, MMb, Xuan Wang, MMª, Rui Zhang, MMª, Qicong Liu, MMª, Wei Cheng, MMª,*
Abstract
Adrenocortical carcinoma is an invasive malignancy with poor prognosis, high recurrence rate and limited therapeutic options. Therefore, it is necessary to establish an effective method to diagnose and evaluate the prognosis of patients, so as to realize individualized treatment and improve their survival rate.
This study investigated metabolic genes that may be potential therapeutic targets for Adrenocortical carcinoma (ACC). Level 3 gene expression data from the ACC cohort and the relevant clinical information were obtained from The Cancer Genome Atlas (TCGA) database. To verify, other ACC datasets (GSE76021, GSE19750) were downloaded from the Gene Expression Omnibus (GEO) database. The ACC datasets from TCGA and GEO were used to screen metabolic genes through the Molecular Signatures Database using gene set enrichment analysis. Then, the overlapping metabolic genes of the 2 datasets were identified.
A signature of five metabolic genes (CYP11B1, GSTM2, IRF9, RPL31, and UBE2C) was identified in patients with ACC. The signature could be used to divide the patients with ACC into high- and low-risk groups based on their median risk score. Multivariate Cox regression analysis was performed to determine the independent prognostic factors of ACC. Time-dependent receiver operating characteristic (ROC) curve analysis was conducted to assess the prediction accuracy of the prognostic signature. Last, a nomogram was established to assess the individualized prognosis prediction model.
The results indicated that the signature of 5 metabolic genes had excellent predictive value for ACC. These findings might help improve personalized treatment and medical decisions.
Abbreviations: ACC = adrenocortical carcinoma, GO = gene oncology, HR = hazard ratio, KEGG = Kyoto Encyclopedia of Genes and Genomes, OS = overall survival, ROC = receiver operating characteristic, TCGA = the cancer genome atlas.
Keywords: adrenocortical carcinoma, bioinformatics, metabolize, mRNA, prognosis, the cancer genome atlas
Editor: Maya Saranathan.
Chongqing social undertakings and people’s livelihood security science and technology innovation special project. NO. Cstc2017shmsA130069.
The authors have no conflicts of interests to disclose.
The datasets generated during and/or analyzed during the current study are publicly available.
ª Department of Endocrinology, the Second Affiliated Hospital of Chongqing Medical University, Chongqing, China, b Guangxi Medical University, Nanning, China.
Correspondence: Wei Cheng, Department of Endocrinology, the Second Affiliated Hospital of Chongqing Medical University, Chongqing 400000, China (e- mail: chengwei.cq@outlook.com).
Copyright @ 2021 the Author(s). Published by Wolters Kluwer Health, Inc. This is an open access article distributed under the Creative Commons Attribution License 4.0 (CCBY), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
How to cite this article: Chen Q, Ren Z, Liu D, Jin Z, Wang X, Zhang R, Liu Q, Cheng W. Identification of prognostic metabolic genes in adrenocortical carcinoma and establishment of a prognostic nomogram: a bioinformatic study. Medicine 2021;100:50(e27864).
Received: 21 March 2021 / Received in final form: 9 October 2021 / Accepted: 3 November 2021
http://dx.doi.org/10.1097/MD.0000000000027864
1. Introduction
Adrenocortical carcinoma (ACC) is a rare but aggressive endocrine malignancy with an incidence worldwide of 0.7 to 2.0 cases/million/year.[1] It can occur at any age and has a bimodal incidence, predominantly affecting adults in their 40 to 50 second and children aged 1 to 5 years.[2] ACC is a heterogeneous disease with multiple manifestations. For 40% to 60% of patients, the major presenting complaints are symptoms and signs of hormone excess.[3,4] Roughly 10% to 15% of ACCs are incidentally diagnosed by imaging procedures for unrelated medical issues.[5,6] Although our understanding of this disease has progressed tremendously over the last decade, the rate of early diagnosis of ACC is still low due to its rarity and lack of specific diagnostic criteria. Currently, surgical resection offers the only hope of a curative intervention in ACC. Initial staging is the most important factor in determining prognosis. However, even in patients with negative resection margin resection, recurrence and disease progression is still commonly seen, so that adjuvant therapy is frequently recommended. Currently, the only other adjuvant treatment shown to have effect is radiotherapy, however overall survival was not improved.[7] Therefore, it is very important to find effective methods to improve the diagnosis and prognosis strategies of ACC.
Evidence suggests that the occurrence of tumors involves a variety of changes in cell metabolism. Fast-growing cancer cells produce adenosine triphosphate at a high rate of glycolysis, independent of the supply of oxygen, a process known as the Warburg effect.[8] This change in metabolic pathways is a hallmark of cancer because it helps to absorb large amounts of nutrients into cellular building blocks, leading to an overproduction of the antioxidant glutathione, which leads to the creation of new cells. [9] These changes lead to overexpression of metabolic genes. In most cases, metabolic changes are driven by oncogene-oriented metabolic reprogramming,[10] which seems to be a common feature of highly invasive tumors, independent of carcinogenic origin.[11] This change can be found in many tumors, such as pancreatic and colon cancers.[12,13] However, there isn’t research shows the relationship between the metabolic genes and ACC. In this study, metabolic genes associated with the prognosis of ACC were identified on the basis of RNA-seq data from the cancer genome atlas (TCGA) through the Molecular Signatures Database (MSigDB),[14] and a risk score model for ACC prognosis was constructed. A prognostic nomogram that combined prognostic gene trait risk models and clinical prognostic factors was established to predict overall survival (OS). The validity of the model was further verified through the geo database.
2. Materials and methods
2.1. Data source
The common shared data sets of 79 ACC patients and 128 normal tissues were downloaded from TCGA database (https:// portal.gdc.cancer.gov/, accessed July 20, 2020) and GETx (Genotype-Tissue Expression database, https://www.gtexpor tal.org/home/index.html, accessed July 20, 2020). GTEX data- base is a high-throughput sequencing database covering many normal human tissues. Common data sources from 2 different data sets are standardized and combined. The clinical and pathologic data contain age, gender, tumor stage, TNM stage, survival time. All patient data involved in current study were acquired from the TCGA database and GETx database, and data retrieval and application complied with the TCGA and GETx publication guidelines and data access policies. Therefore, additional study approval by the local Ethical Review Committee was not required. other ACC datasets (GSE76021, GSE19750) were downloaded from the Gene Expression Omnibus (GEO) database. Additional ACC datasets (GSE76021, GSE19750) for validation were downloaded from the gene expression compre- hensive database (GEO), which contained 29 ACC samples (GSE76021), 44 ACC samples (GSE19750) and 4 normal samples (GSE19750).
2.2. Metabolic gene screening
A list of metabolic genes was extracted from the datasets of Metabolic process (c2.cp.kegg.v7.0.symbols) from MSigDB (http://software.broadinstitute.org/gsea/msigdb/index.jsp).[14]
The expression data of these genes were screened from the ACC cases of TCGA and GEO. Metabolic prognostic genes were further evaluated with univariate Cox proportional hazard regression by using a “survival” package (version 3.2-3) on the R platform (version 3.6.3). Genes with P <. 05 and |hazard ratio (HR)| > 1.00 were considered to be prognostic risk genes, and their expression levels were significantly associated with OS in ACC.
2.3. Signature development
The prognostic metabolic genes were analyzed using multivariate Cox regression analysis with OS as the dependent variable to evaluate their roles in predicting ACC survival. A prognostic risk score model was prepared via the linear combination of the expression levels of metabolic genes with the multivariate Cox regression coefficient (B) as the weight.[15] The risk scores were calculated using the prognostic gene signatures. The risk score formula was as follows: risk score = expression of gene1 x ß1 + expression of gene2 × 32 + ·· · expression of gene nx Bn.[16-18] A total of 79 cases were divided into high- and low-risk groups based on the median risk score. |HR|>1.0 and P <. 05 were selected among the TCGA and GEO datasets as a cut-off. Then, 5 genes were selected for signature development. Receiver operating characteristic (ROC) curve refers to the receiver operating characteristic curve. It is a comprehensive index reflecting the continuous variables of sensitivity and specificity. It reveals the relationship between sensitivity and specificity by composition. A ROC curve was established over time on the R platform to assess the accuracy of the risk score model for predicting the prognosis of ACC.
2.4. Functional and pathway enrichment analyses
Gene ontology (GO) simply annotates the gene products from the aspects of function, involved biological pathways and localiza- tion in cells. Go enrichment analysis can be used to understand the biological functions, pathways or cell localization of gene enrichment. Kyoto Encyclopedia of Genes and Genomes (KEGG) is a comprehensive database integrating genome, chemistry and system function information. KEGG pathway stores the information of many gene pathways in different species. The functional enrichment analyses of the metabolic genes mainly involving GO terms and KEGG pathway analysis were carried out using the “cluster Profiler” R package.[19] GO analysis revealed the functions of the metabolic genes in biological processes (BP), cellular components, and molecular functions, and the KEGG analysis showed the pathway enrichment of the metabolic genes. P <. 05 was considered statistically significant.
2.5. Predictive nomogram construction and validation
Prognostic characteristics based on metabolic gene expression were analyzed to further evaluate the prognostic model. After collinearity was tested, a Norman diagram was predicted using a stepwise Cox regression model to predict the 1-, 3-, and 5-year OS rates of the patients with ACC the datasets from TCGA and GEO. Kaplan-Meier analysis and area under the curve comparison of the ROC curve were applied to predict and observe the OS rate for assessing the performance of the prognostic nomogram. We not only compared the clinical outcomes of the low risk and high risk groups, but also assessed the prognostic value of ACC using nomogram’s risk score. The potential application of risk scores in the prediction of clinical status was also discussed.
2.6. Statistical analysis
Lasso (least absolute shrinkage and selection operator) method is a compression estimation method based on the idea of reducing variable set (order reduction). By constructing a penalty function, it can compress the coefficients of variables and make some regression coefficients become 0, so as to achieve the purpose of variable selection. Compared with ordinary logistic regression, a bias term is
added to the regression optimization function to reduce the influence of collinearity and reduce the model variance. Therefore, it is used to screen important genes related to ACC. Kaplan-Meier survival analysis statistical method is a survival analysis method that can investigate single factors and hierarchical control confounding factors, which is suitable for large samples or small samples. Kaplan-Meier survival analysis by log-rank test was conducted to identify the metabolic genes associated with the prognosis of ACC. Univariate, multivariate, and Cox regression analyses were executed in R. Univariate and multivariate Cox regression analyses as much performed to assess survival, HRs and 95% confidence intervals were calculated to identify OS-associated genes. Statistical significance was set at P <. 05.
3. Results
3.1. Identification of metabolic genes with prognostic value
Clinical information and gene expression profiles of 79 ACC cases and 128 normal tissues downloaded from the database of TCGA were used for further analysis. A total of 36521 metabolic
genes were selected from MSigDB v4.0[14] Metabolic process (c2. cp.kegg.v7.0.symbols). For validation, gene expression micro- array datasets GSE76021 and GSE19750 were downloaded from the GEO dataset. We extracted 3527 metabolism related genes expression quantities from TCGA and GTEX, and then compared the expression quantities by volcano map of differen- tial in tumor tissues and normal tissues (Fig. 1B), resulting in 37 differential genes (Fig. 1C), and taked heatmap of differentially expressed metabolic genes in the Gene Expression Omnibus cohort (Fig. 1A). Then, 5 best modeled prognostic metabolic related genes (11ß-hydroxylase P450c11 [CYP11B1], glutathi- one-S-transferases mu 2 [GSTM2], interferon regulatory factor 9 [IRF9], ribosomal protein L31 [RPL31] and ubiquitin-conjugat- ing enzyme E2C [UBE2C]) were identified by univariate analysis, multivariate analysis and lasso regression analysis, and were used to establish the metabolic gene signature (Fig. 1 DEFGH) and the results are shown in Figure 2, Tables 1 and 2.
3.2. Prognostic model construction and receiver operating characteristic curve analysis
The signatures of 5 metabolic genes were developed using a risk- scoring approach. The 79 ACC patients in the database from
Volcano
8
8
-log10(fdr)
2
15
0
5
0
B
-4
-2
0
2
4
logFC
12
-
5
CYP11B1 expression
·
GSTM2 expression
2
0-
..
0
..
D
Normal
Tumor
Turbor
F
Normal
-
s
75
A
IRFB expression
RPL31 expression
0
25
type @NET
2
0.0
..
1
D
-
E
Normal
Tumor
G
Norma
00
I
D
.
Gene expression
O
N
O
.
-
G
1
..
-
UBERC espression
2%
0
:
.
2%
ABCC3
ADH1B
CCL2
CHKB
CYP11B1
DCN
FTH1 GSTM2
GTF21
H3F3A
HSD3B2
IGFBP3
ILK
IRF9
MAN2C1
NDST2
NDUFV2
PIK3R2
PLA2G1B
POLR2F
PSMC1
PTGDS
PTHIR
RAP1B
RPL27A
RPL31
RPL37A
RPL39
RPS29
SDHC
SEMA3B
SLC11A1
TH
TNXB
TOP38
U2AF1
UBE2C
C
H
Normal
Tumor
| pvalue | Hazard ratio | |
|---|---|---|
| RPS29 | 0.027 | 1.984(1.081-3.642) |
| RPL31 | <0.001 | 4.498(2.041-9.912) |
| IRF9 | 0.012 | 2.226(1.193-4.151) |
| RPL27A | 0.005 | 2.025(1.241-3.306) |
| UBE2C | <0.001 | 2.616(1.809-3.783) |
| H3F3A | 0.046 | 2.859(1.021-8.005) |
| GSTM2 | 0.012 | 0.577(0.375-0.888) |
| RPL37A | 0.001 | 2.666(1.464-4.852) |
| CYP11B1 | 0.018 | 0.870(0.775-0.977) |
8
8
8
8
8
8
8
6
5
5
5
5
5
2
1
1
9.5
H
Partial Likelihood Deviance
9.0
8.5
0
2
4
6
8
A
Hazard ratio
8.0
8
8
5
4
0
B
2
-5
-4
-3
-2
-1
2.0
Log (x)
1.5
Hazard ratio
RPL31
(N=79)
2.47
(1.05 - 5.83)
F
0.039 *
Coefficients
1.0
5
IRF9
(N=79)
1.96
4
0.5
(0.85 - 4.52)
+
0.116
3
8
UBE2C
(N=79)
2.05
0.0
(1.34 - 3.13)
<0.001
9
-0.5
GSTM2
(N=79)
0.57
(0.35 - 0.93)
0.024 *
4
7
-1.0
1
CYP11B1
(N=79)
0.87
(0,76 - 0,99)
-
0.034 *
T
-5
-4
-3
-2
-1
C
0.1
0.2
0.5
2
5
Log Lambda
D
1
# Events: 28; Global p-value (Log-Rank): 4.2579e-09
AIC: 178.05; Concordance Index: 0.87
TCGA and GETx were divided into high- and low-risk groups based on their median risk score. The Kaplan-Meier analysis in the GEO cohort are shown in Figure 3AB, and in the TCGA cohort are shown in Figure 3FH. Heatmap of differentially expressed metabolic genes in TCGA cohort (Fig. 3C) and Gene Expression Omnibus (GEO) cohort (Fig. 3I). Survival analysis showed that the OS of the patients in the high-risk group was shorter (median survival time of high risk and low risk: 748 vs 1547 day) than it was in the low-risk group (P <. 001, Fig. 3D). Validation by the GEO database further confirmed that the survival rate of the high-risk score group was lower than that of the low-risk group (P <. 001, Fig. 3G). Our results showed that the prognostic signature from the database from TCGA in the current study performed well in predicting 1-, 3-, and 5-year survival rates. The area under the curve values for 1-, 3-, and 5- year survival were 0.865, 0.929, and 0.867, respectively (Fig. 3E). The prognostic signatures of the GEO database also performed well in predicting 1-, 3-, and 5-year survival rates (Fig. 3J).
3.3. Predictive nomogram construction and validation
The nomogram calculates the predictive value for an outcome event for that individual by constructing a multivariate regression model, allowing visual representation of the model’s influence on the outcome event. Nomogram survival analysis was used to investigate the relationship between OS risk score and clinical characteristics of ACC patients. A nomogram was drawn with RMS and its auxiliary packages based on the clinical information of ACC and the risk score. The results showed that the prognostic indicators of the risk score significantly influenced the risk points, whereas other clinical features had a lower effect on the risk points (Fig. 4). In our nomogram, the disadvantage was that the prognostic signature of tumor stage could not perform well in ACC.
3.4. Functional enrichment analysis of genes
Metabolic genes were subjected to functional enrichment analysis by applying the cluster Profiler R package. The metabolic genes
| Id | HR | HR.95L | HR.95H | P value |
|---|---|---|---|---|
| RPS29 | 1.983898167 | 1.080620945 | 3.642213262 | .0270979 |
| RPL31 | 4.49757919 | 2.040765418 | 9.912074358 | .000192074 |
| IRF9 | 2.225511025 | 1.193282913 | 4.150649664 | .011881593 |
| RPL27A | 2.025251643 | 1.240650243 | 3.306043938 | .004766747 |
| UBE2C | 2.616123087 | 1.809031407 | 3.783295293 | 3.231040182 |
| H3F3A | 2.858683906 | 1.020821803 | 8.005387082 | .045588031 |
| GSTM2 | 0.577364214 | 0.37549014 | 0.887771475 | .012340497 |
| RPL27A | 2.665524947 | 1.464434062 | 4.851719466 | .001335247 |
| GSTM2 | 0.870139614 | 0.775185453 | 0.976724918 | .018303571 |
| Id | Coef | HR | HR.95L | HR.95H | P value |
|---|---|---|---|---|---|
| RPL31 | 0.904238627 | 2.470050577 | 1.047129791 | 5.826545957 | .038909596 |
| IRF9 | 0.671920087 | 1.957993232 | 1.047129791 | 4.521446836 | .115586995 |
| UBE2C | 0.717969594 | 2.050266109 | 1.343038837 | 3.129910321 | .000879668 |
| GSTM2 | -0.561156263 | 0.570548978 | 0.351050053 | 0.927292655 | .023537575 |
| CYP11B1 | -0.141209604 | 0.868307292 | 0.761979656 | 0.989472026 | .034110122 |
from the database from TCGA in the BP analysis were mainly enriched in nuclear-transcribed mRNA catabolic process and protein targeting to ER. The genes in the cellular component analysis were significantly enriched in cytosolic part and mitochondrial inner membrane. The genes in the molecular function analysis were mainly enriched in structural constituent of ribosome and intramolecular oxidoreductase activity (P <. 05, Fig. 5A, B, C). KEGG analysis suggested that most of the metabolic gene pathways were significantly linked to ribosome and focal adhesion (P <. 05, Figure 5D). Gene set enrichment analysis was used to clarify the main biological functions of 5 node genes. The results showed that they were mainly associated with condensed chromosome, regulation of chromosome segregation, and cell cycle transition (Fig. 6).
4. Discussion
Adrenal tumors are very common, and the majority are small benign nonfunctional adrenocortical adenomas (ACA).[20] ACC, in contrast, is a rare cancer of the adrenal cortex. Despite its rarity, outcomes for patients diagnosed with ACC remain dismal, with a 5-year overall survival of ~35%.[21] Clinically apparent adrenocortical hormone production is evident in up to 45% to 70%.[22] However, syndromes of hormone excess are often not readily recognized by physicians, leading to delay in diagnosis and subsequent surgical and/or medical therapy. Typically, the ideal treatment for ACC is surgical intervention, but ~75% of all patients will ultimately develop metastases.[23] Patients with metastatic disease mainly receive radiation therapy, chemothera- py, or immunotherapy. Mitotane is a unique antineoplastic agent used solely in the treatment of ACC, sometimes paired with palliative surgery. Several mechanisms have been postulated for its action including alteration of mitochondrial respiratory chain activity by inducing cytochrome c oxidase defect in adrenocorti-
cal cells. However, these treatments are often ineffective. These challenges call for an urgent need to find more effective methods to improve strategies for ACC diagnosis and prognosis.
Metabolic reprogramming is considered a hallmark of cancer because it is the most common physiological change in cancer cells.[24] In recent years, several metabolic gene models have been established for various types of cancer, such as colon cancer,[25] thyroid cancer,[26] and ovarian cancer.[27] Therefore, under- standing changes in metabolite levels associated with ACC will help in the search for accurate and clinically useful biomarkers and therapeutic targets. Comprehensive analysis of metabolo- mics and transcriptomics data is an advanced method for finding reliable metabolic biomarkers. However, as far as we know, no metabolic gene signature has been set up for ACC. So, this study attempted to identify differentially expressed metabolic genes and establish a useful metabolic gene signature to predict clinical outcomes for patients with ACC.
In this study, we identified a total of 37 dysregulated metabolic genes in ACC using GEO and TCGA data. Then, 5 identified genes, CYP11B1, GSTM2, IRF9, RPL31, and UBE2C were found to be associated with a high risk in the databases from TCGA and GEO. The signature of 5 metabolic genes was developed using a risk score method, and the patients with ACC were divided into low- and high-risk groups based on their median risk score. Analysis of the gene set enrichment analysis indicated that these genes were mainly enriched in cell growth and division in the high-risk groups. And these pathways of active cell growth may have an important relationship with highly active tumor growth in the high-risk group. Multivariate Cox regression analysis was conducted to determine the independent prognostic factors of ACC. Then, the prognostic signature was comprehensively analyzed on the basis of metabolic gene expression. Kaplan-Meier analysis and the area under the curve comparison of the ROC curve confirmed that the 5 metabolic
Risk
High risk
Low risk
Risk score
8
…
-
High risk
Survival probability
1.00-
. low Risk
0
5
0.75
៛
(
0.50
0
20
40
60
80
A
Patients (increasing risk socre)
0.25
p=1.907e-09
0.00
Survival time (years)
0
1
2
3
4
5
6
7
8
9
10
11
12
Time(years)
~
Dead
.
Alive
8
Risk
39
35
20
14
9
5
2
1
1
1
0
0
0
0
Low risk
40
40
38
30
21
19
14
10
7
6
4
2
2
D
0
1
2
3
4
5
6
7
8
9
10
Time(years)
11
12
A
NV
0
0
20
40
60
80
B
Patients (increasing risk socre)
1.0
type
type high
3
0.8
IRF9
2.5
low
2
GSTM2
1.5
Sensitivity
0.6
1
CYP11B1
0.5
0
0.4
RPL31
UBE2C
0.2
C
AUC of 1 year survival:
0.865
0.0
AUC of 3 year survival:
0.929
AUC of 5 year survival:
0.867
E
0.0
0.2
0.4
0.6
0.8
1.0
False positive rate
Risk score
-
High risk
0 2 3 4 56
low Risk
Risk
High risk
Low risk
Survival probability
1.00
0.75
0
10
20
30
40
50
F
Patients (increasing risk socre)
0.50
0.25
p=1.219e-02
Survival time (years)
0.00
0
1
N
00
4
un
6
7
8
9
10 11
12
13
14
15
16
17
18
19
20
Time(years)
5
Dead
Alive
Risk
21 15 10
5
5
5
5
5
3
3
3
3
3
2
1
1
0
0
0
0
C
Low risk
30
25 21
18
17
16
14
12
11
10
7
6
5
4
4
3
3
1
1
0
0
5
G
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Time(years)
0
0
10
20
30
40
50
H
Patients (increasing risk socre)
1.0
0.8
type
3.5 type
3
high
Sensitivity
RPL31
low
0.6
2.5
2
CYP11B1
1.5
1
0.4
UBE2C
0.5
0
IRF9
0.2
GSTM2
AUC of 1 year survival:
0.694
0.0
AUC of 3 year survival:
0.795
AUC of 5 year survival: 0.823
I
J
0.0
0.2
0.4
0.6
0.8
1.0
False positive rate
| 0 10 | 20 | 30 | 40 | 50 | 60 | 70 | 80 | 90 | 100 | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Points Gender | female | ||||||||||||||
| male | |||||||||||||||
| Age | >65 | ||||||||||||||
| ≤65 | |||||||||||||||
| Stage | II | IV | |||||||||||||
| I | III | ||||||||||||||
| T2 | T4 | ||||||||||||||
| T | T1 T3 | ||||||||||||||
| N | N1 | ||||||||||||||
| NO | |||||||||||||||
| M | M1 | ||||||||||||||
| MO | |||||||||||||||
| laterality | left | ||||||||||||||
| right | |||||||||||||||
| RiskScore | |||||||||||||||
| 2.5 3 | 3.5 | 4 | 4.5 | 5 | 5.5 | 6 | 6.5 | 7 | 7.5 | 8 | |||||
| Total Points | |||||||||||||||
| 0 20 | 40 | 60 | 80 | 100 | 120 | 140 | 160 | 180 | |||||||
| 1-year survival | |||||||||||||||
| 0.99 | 0.9 | 0.7 | 0.5 0.3 | 0.1 | |||||||||||
| 3-year survival | |||||||||||||||
| 0.99 | 0.9 | 0.7 | 0.5 0.3 | 0.1 | |||||||||||
| 5-year survival | |||||||||||||||
| 0.99 | 0.9 | 0.7 | 0.5 | 0.3 | 0.1 | ||||||||||
mRNA catabolic process
RNA catabolic process
viral transcription
cytosolic ribosome-
viral gene expression
ribosomal subunit-
SRP-dependent cotranslational protein targeting to membrane
cytosolic part- ribosome-
cotranslational protein targeting to membrane
protein targeting to ER-
nuclear-transcribed mRNA catabolic process, nonsense-mediated decay
pvalue
cytosolic large ribosomal subunit-
large ribosomal subunit-
establishment of protein localization to endoplasmic reticulum
focal adhesion-
protein localization to endoplasmic reticulum
0.0004
cell-substrate adherens junction-
Count
translational initiation-
0.0008
cell-substrate junction-
1
protein targeting to membrane
0.0012
mitochondrial inner membrane
2
nuclear-transcribed mRNA catabolic process-
0.0016
polysomal ribosome
polysome-
3
establishment of protein localization to membrane
negative regulation of cellular component movement-
smooth endoplasmic reticulum-
4
negative regulation of locomotion
Count
terminal bouton-
5
protein targeting
2
rough endoplasmic reticulum-
response to peptide hormone
3
respiratory chain complex-
cytoplasmic translation
antimicrobial humoral response
4
mitochondrial respiratory chain-
pvalue
5
respiratory chain-
positive regulation of calcium ion transport
oxidoreductase complex-
0.005
mineralocorticoid biosynthetic process-
6
axon terminus-
0.010
mineralocorticoid metabolic process-
1
inner mitochondrial membrane protein complex-
0.015
glucocorticoid biosynthetic process
neuron projection terminus
0.020
response to corticosterone-
proteasome regulatory particle, base subcomplex-
0.025
positive regulation of myoblast differentiation-
RNA polymerase I complex-
innate immune response in mucosa
secondary lysosome
benzene-containing compound metabolic process-
melanosome membrane
glucocorticoid metabolic process-
chitosome-
circadian sleep/wake cycle
pigment granule membrane
0.05
0.10
0.15
0.20
RNA polymerase Il, core complex-
GeneRatio
cytoplasmic side of endoplasmic reticulum membrane-
.
A
B
0.025
0.050
0.075
GeneRatio
0.100
0,125
structural constituent of ribosome
iron ion binding
isomerase activity
carboxylic acid binding
Ribosome
organic acid binding
ferric iron binding
ferrous iron binding
intramolecular oxidoreductase activity
Count
Count
monocarboxylic acid binding
monooxygenase activity
1
2
electron transfer activity
2
3
3
Focal adhesion
oxidoreductase activity, acting on the CH-OH group of donors, NAD or NADP as acceptor oxidoreductase activity, acting on CH-OH group of donors
4
4
5
integrin binding
heme binding
5
ferroxidase activity
pvalue
oxidoreductase activity, oxidizing metal ions, oxygen as acceptor-
pvalue
RNA polymerase Il activity
0.01
glutathione binding
Tyrosine metabolism
0 02
RNA polymerase I activity
0.01
0.03
insulin-like growth factor I binding
0.02
0.04
oligopeptide binding-
dopamine binding
0,03
protein phosphatase activator activity
intramolecular oxidoreductase activity, transposing C=C bonds
1-phosphatidylinositol-3-kinase regulator activity-
Steroid hormone biosynthesis
RNA polymerase III activity:
alpha-mannosidase activity
heparan sulfate sulfotransferase activity
C
mannosidase activity-
0.100
D
0.075
0.100
GeneRatio
0.125
0.025
0.050
0.075
0.125
GeneRatio
Enrichment plot: GO_MITOTIC_NUCLEAR_DIVISION
Enrichment plot: GO_CONDENSED_CHROMOSOME
0.7
Enrichment score (ES)
0.7
0.6
Enrichment score (ES)
0.6
0.5
0.5
0.4
ES=0.61
0.4
0.3
0.3
ES=0.68
0.2
FDR<0.01
0.2
FDR<0.01
0.1
P<0.01
0.1
P<0.01
0.0
0.0
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
1.0
1.0
-
0.5
Zero cross at 17096
0.5
0.0
Zero cross at 17098
0.0
0.5
T (negatively comelated)
-0.5
T (negatively comeisted)
0
5.000
10.000
15.000
20.000
25.000
30.000
35.000
Rank in Ordered Dataset
0
5.000
10.000
15,000
20.000
25.000
30.000
35,000
Rank in Ordered Dataset
A
Enrichment profile - Hits
Ranking metric scores
B
Enrichment profile - Hits
Ranking metric scores
Enrichment plot: GO_CHROMOSOME_SEGREGATION
Enrichment score (ES)
Enrichment plot: GO_CELL_CYCLE_G2_M_PHASE_TRANSITION
0.6
Enrichment score (ES)
0.6
0.5
0.5
0.4
ES=0.66
0.4
0.3
FDR<0.01
ES=0.58
0.3
0.2
P<0.01
0.2
FDR<0.01
0.1
0.1
P<0.01
0.0
0.0
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
1.0
-
1.0
0.5
0.5
0.0
Zero cross at 17000
0.0
Zero cross at 17008
0.5
0.5
” (negatively comeisted)
Y (negatively comelated)
0
5.000
10.000
15.000
20.000
25.000
30.000
35.000
0
5.000
10,000
15.000
20,000
25.000
30,000
35.000
Rank in Ordered Dataset
Rank in Ordered Dataset
D
Enrichment profile - Hits
Ranking metric scores
E
Enrichment profile
Hits
Ranking metric scores
Enrichment plot: GO_REGULATION_OF_CHROMOSOME_SEGREGATION
Enrichment plot: GO_ORGANELLE_FISSION
0.6
Enrichment score (ES)
0.8
Enrichment score (ES)
0.7
0.5
0.6
0.4
0.5
0.4
ES=0.76
ES=0.59
0.3
FDR<0.01
0.3
FDR<0.01
0.2
P<0.01
0.2
P<0.01
0.1
0.1
0.0
0.0
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
1.0
1.0
0.5
0.5
0.0
Zero cross at 17008
Zero cross at 17008
0.0
0.5
0.5
7. (negatively correlated)
T (negatively comrelated)
0
5.000
10.000
15,000
20,000
25,000
30.000
35,000
0
5,000
10,000
15.000
20.000
25,000
30,000
35,000
Rank in Ordered Dataset
Rank in Ordered Dataset
G
Enrichment profile
Hits
Ranking metric scores
H
Enrichment profile
Hits
Ranking metric scores
Enrichment plot: GO_CELL_CYCLE_CHECKPOINT
Enrichment score (ES)
0.6
0.5
0.4
0.3
ES=0.64
0.2
FDR<0.01
0.1
P<0.01
0.0
Ranked list metric (Signal2Noise)
-
1.0
0.5
Zero cross a: 17098
0.0
-0.5
” (negatively correlated)
0
5,000
10.000
15.000
20,000
25.000
30.000
35.000
Rank in Ordered Dataset
C
Enrichment plot: GO_NUCLEAR_CHROMOSOME_SEGREGATION
Enrichment score (ES)
0.6
0.5
0.4
ES=0.66
0.3
FDR<0.01
0.2
P<0.01
0.1
0.0
Ranked list metric (Signal2Noise)
1.0
0.5
Zero cross at 17008
0.0
-0.5
T (negatively comelated)
0
5.000
10,000
15,000
20.000
25,000
30.000
35.000
Rank in Ordered Dataset
Enrichment profile - Hits
Ranking metric scores
Enrichment plot: GO_REGULATION_OF_CELL_CYCLE_PHASE_TRANSITI
ON
Enrichment score (ES)
0.5
0.4
0.3
ES=0.53
0.2
FDR<0.01
0.1
P<0.01
0.0
Ranked list metric (Signal2Noise)
1.0
0.5
Zero cross at 17008
0.0
-0.5
” (negatively correlated)
0
5.000
10.000
15.000
20.000
25,000
30,000
35,000
Rank in Ordered Dataset
I
Enrichment profile - Hits
Ranking metric scores
gene signatures were reliable for OS prediction. Next, a nomogram was established and integrated with a signature of 5 metabolic genes and clinical data, and OS was accurately predicted. Furthermore, a prognostic nomogram was established based on the risk score and 7 clinicopathological factors. The prognostic accuracy of the model was confirmed by the ROC curve and the corresponding calibration curve. These results indicate that the signature not only serves as a biomarker for ACC, independent of clinicopathological features, but also predicts clinical outcomes in ACC patients.
In our current study, 5 genes associated with prognosis of ACC were identified: CYP11B1, GSTM2, IRF9, RPL31, and UBE2C. Most of the 5 genes of the signature have been reported in various
types of cancer. CYP11B1 is a Protein Coding gene which provides instructions for making an enzyme called 11-beta- hydroxylase. This enzyme is found in the adrenal glands, which are located on top of the kidneys. The 11-beta-hydroxylase enzyme is a member of the cytochrome P450 family of enzymes. These enzymes are involved in the formation and breakdown of various molecules within cells. The 11-beta-hydroxylase enzyme helps produce hormones called cortisol and corticosterone. Mutations in this gene cause congenital adrenal hyperplasia due to 11-beta-hydroxylase deficiency. Mitotane as an adjuvant therapy for ACC patients, it is metabolites can inhibit several enzymes in the adrenocortical steroidogenesis pathway, mainly at the level of the cholesterol side-chain cleavage enzymes CYP11A1
Enrichment profile - Hits
Ranking metric scores
F
and CYP11B1.[28,29] CYP11B1 is reliable biomarkers for the histopathological subtyping of functional benign adrenocortical tumors (ACT) and may offer some value in the histopathological diagnosis of malignant ACT.[30] Compared with normal tissue, this gene is low expressed in ACC.
The GO annotations related to GSTM2 include protein homodimerization activity and enzyme binding. GSTM2 is a Protein Coding gene. This gene encodes a glutathione S- transferase that belongs to the mu class. The mu class of enzymes functions in the detoxification of electrophilic compounds, including carcinogens, therapeutic drugs, environmental toxins and products of oxidative stress, by conjugation with glutathi- one. Genetic variants of glutathione S-transferase as possible risk factors for carcinoma. The genetic variations can change an individual’s susceptibility to carcinogens and toxins as well as affect the toxicity and efficacy of certain drugs. [31,32] Diseases associated with GSTM2 include Barrett’s Adenocarcinoma[33] and Hepatocellular Carcinoma.[34] Therefore, the low GSTM2 expression in ACC might be the cause of ACC pathogenesis and progression.
Diseases associated with IRF9 include Immunodeficiency 65 Viral Infections[35] and Lymphocytic Choriomeningitis.[36] Among its related pathways are Interferon Pathway and JAK- STAT signaling pathway. GO annotations related to this gene include DNA-binding transcription factor activity. Studies have shown that the triple negative breast cancer inherent interferon regulatory factor IRF9 as a marker of active intratumoral type I and II interferon signaling and reduced risk of distant relapse.[37]
RPL31 encodes a ribosomal protein that is a component of the 60S subunit. The protein belongs to the L31E family of ribosomal proteins. It is located in the cytoplasm. Higher levels of expression of this gene in familial adenomatous polyps compared to matched normal tissues have been observed. Diseases associated with RPL31 include Diamond-Blackfan Anemia[38] and Combined Oxidative Phosphorylation Deficiency 1. Among its related pathways are rRNA processing in the nucleus and cytosol and Viral mRNA Translation. GO annotations related to this gene include structural constituent of ribosome.
UBE2C encodes a member of the E2 ubiquitin-conjugating enzyme family. The encoded protein is required for the destruction of mitotic cyclins and for cell cycle progression, and may be involved in cancer progression. Diseases associated with UBE2C include Breast cancer[39] and Gastric Cancer. 40] Among its related pathways are Gastric Cancer Network 2 and Regulation of activated PAK-2p34 by proteasome mediated degradation.
This study has some limitations. First, our clinical information was mainly obtained from databases from TCGA and GEO. Some patients’ clinical information was incomplete, and detailed data on patient prognosis were unavailable. Second, we were unable to evaluate the predictive accuracy of the nomogram by external validation. The expression level of prognosis-related genes and their molecular mechanisms in the pathogenesis and progression of ACC should be further explored experimentally. The screened genes could be verified through real-time PCR and Western blot.
5. Conclusion
We successfully established a novel metabolic gene signature in ACC. This signature was not only able to predict OS in ACC patients but was also used as a guide for the diagnosis and therapy of ACC.
Author contributions
Formal analysis: Qing Chen, Zongrui Jin, Xuan Wang.
Funding acquisition: Dongfang Liu.
Investigation: Ziyu Ren.
Methodology: Qing Chen, Rui Zhang.
Resources: Qing Chen, Qicong Liu.
Software: Qing Chen.
Supervision: Dongfang Liu.
Writing - original draft: Qing Chen.
Writing - review & editing: Dongfang Liu, Wei Cheng.
References
[1] Kebebew E, Reiff E, Duh QY, et al. Extent of disease at presentation and outcome for adrenocortical carcinoma: have we made progress? World J Surg 2006;30:872-8.
[2] Wajchenberg BL, Albergaria Pereira MA, Medonca BB, et al. Adreno- cortical carcinoma: clinical and laboratory observations. Cancer 2000;88:711-36.
[3] Allolio B, Fassnacht M. Clinical review: adrenocortical carcinoma: clinical update. J Clin Endocrinol Metab 2006;91:2027-37.
[4] Luton J, Cerdas S, Billaud L, et al. Clinical features of adrenocortical carcinoma, prognostic factors, and the effect of mitotane therapy. N Engl J Med 1990;322:1195-201.
[5] Dietel M, Schirmacher P. Deficits in the management of patients with adrenocortical carcinoma in Germany. Statements cannot be substanti- ated. Deutsches Arzteblatt international 2011;108:339-40.
[6] Dackiw A, Lee JE, Gagel RF, et al. Adrenal cortical carcinoma. World J Surg 2001;25:914-26.
[7] Fassnacht M, Kim AC, Sabolch A, et al. Efficacy of adjuvant radiotherapy of the tumor bed on local recurrence of adrenocortical carcinoma. J Clin Endocrinol Metab 2006;91:4501-4.
[8] Warburg O. On the origin of cancer cells. Science (New York, NY ) 1956;123:309-14.
[9] Hirschhaeuser F, Sattler U, Mueller-Klieser W. Lactate: a metabolic key player in cancer. Cancer Res 2011;71:6921-5.
[10] Shi Y, Wei J, Chen Z, et al. Integrative analysis reveals comprehensive altered metabolic genes linking with tumor epigenetics modification in pan-cancer. Biomed Res Int 2019;2019:6706354.
[11] Bustamante E, Morris H, Pedersen P. Energy metabolism of tumor cells. Requirement for a form of hexokinase with a propensity for mitochondrial binding. J Biol Chem 1981;256:8699-704.
[12] Chaika NV, Yu F, Purohit V, et al. Differential expression of metabolic genes in tumor and stromal components of primary and metastatic loci in pancreatic adenocarcinoma. PLoS One 2012;7:e32996.
[13] Ren J, Feng J, Song W, et al. Development and validation of a metabolic gene signature for predicting overall survival in patients with colon cancer. Clin Exp Med 2020.
[14] Zuo S, Wei M, Wang S, et al. Pan-cancer analysis of immune cell infiltration identifies a prognostic immune-cell characteristic score (ICCS) in lung adenocarcinoma. Front Immunol 2020;11:1218.
[15] Huet S, Tesson B, Jais JP, et al. A gene-expression profiling score for prediction of outcome in patients with follicular lymphoma: a retrospective training and validation analysis in three international cohorts. Lancet Oncol 2018;19:549-61.
[16] Zhou M, Zhao H, Wang Z, et al. Identification and validation of potential prognostic lncRNA biomarkers for predicting survival in patients with multiple myeloma. J Exp Clin Cancer Res 2015;34:102.
[17] Liao X, Huang K, Huang R, et al. Genome-scale analysis to identify prognostic markers in patients with early-stage pancreatic ductal adenocarcinoma after pancreaticoduodenectomy. Onco Targets Ther 2017;10:4493-506.
[18] Wang X, Gao J, Zhou B, et al. Identification of prognostic markers for hepatocellular carcinoma based on miRNA expression profiles. Life Sci 2019;232:116596.
[19] Yu G, Wang LG, Han Y, et al. Cluster Profiler: an R package for comparing biological themes among gene clusters. OMICS 2012;16:284-7.
[20] Mansmann G, Lau J, Balk E, et al. The clinically inapparent adrenal mass: update in diagnosis and management. Endocr Rev 2004;25:309-40.
[21] Else T, Kim AC, Sabolch A, et al. Adrenocortical carcinoma. Endocr Rev 2014;35:282-326.
[22] Abiven G, Coste J, Groussin L, et al. Clinical and biological features in the prognosis of adrenocortical cancer: poor outcome of cortisol- secreting tumors in a series of 202 consecutive patients. J Clin Endocrinol Metab 2006;91:2650-5.
[23] Glenn J, Else T, Hughes DT, et al. Longitudinal patterns of recurrence in patients with adrenocortical carcinoma. Surgery 2019;165:186-95.
[24] Fouad Y, Aanei C. Revisiting the hallmarks of cancer. Am J Cancer Res 2017;7:1016-36.
[25] Ren J, Feng J, Song W, et al. Development and validation of a metabolic gene signature for predicting overall survival in patients with colon cancer. Clin Exp Med 2020.
[26] Ma B, Jiang H, Wen D, et al. Transcriptome analyses identify a metabolic gene signature indicative of dedifferentiation of papillary thyroid cancer. J Clin Endocrinol Metab 2019;104:3713-25.
[27] Kanakkanthara A, Kurmi K, Ekstrom TL, et al. BRCA1 deficiency upregulates NNMT, which reprograms metabolism and sensitizes ovarian cancer cells to mitochondrial metabolic targeting agents. Cancer Res 2019;79:5920-9.
[28] Cai W, Counsell RE, Schteingart DE, et al. Adrenal proteins bound by a reactive intermediate of mitotane. Cancer Chemother Pharmacol 1997;39:537-40.
[29] Lindhe O, Skogseid B, Brandt I. Cytochrome P450-catalyzed binding of 3-methylsulfonyl-DDE and o,p’-DDD in human adrenal zona fascicu- lata/reticularis. J Clin Endocrinol Metab 2002;87:1319-26.
[30] Yang Y, Xiao M, Song Y, et al. H-score of 11beta-hydroxylase and aldosterone synthase in the histopathological diagnosis of adrenocortical tumors. Endocrine 2019;65:683-91.
[31] Zhu K, Zhang CL, Ma XL, et al. Analyzing the interactions of mRNAs and ncRNAs to predict competing endogenous RNA networks in osteosarcoma chemo-resistance. Mol Ther 2019;27:518-30.
[32] Federici L, Lo Sterzo C, Pezzola S, et al. Structural basis for the binding of the anticancer compound 6-(7-nitro-2,1,3-benzoxadiazol-4-ylthio) hexanol to human glutathione s-transferases. Cancer Res 2009;69: 8025-34.
[33] Peng D, Razvi M, Chen H, et al. DNA hypermethylation regulates the expression of members of the Mu-class glutathione S-transferases and glutathione peroxidases in Barrett’s adenocarcinoma. Gut 2009;58:5-15.
[34] White D, Li D, Nurgalieva Z, et al. Genetic variants of glutathione S- transferase as possible risk factors for hepatocellular carcinoma: a HuGE systematic review and meta-analysis. Am J Epidemiol 2008;167:377-89.
[35] Hernandez N, Melki I, Jing H, et al. Life-threatening influenza pneumonitis in a child with inherited IRF9 deficiency. J Exp Med 2018;215:2567-85.
[36] Huber M, Suprunenko T, Ashhurst T, et al. IRF9 prevents CD8 T cell exhaustion in an extrinsic manner during acute lymphocytic choriome- ningitis virus infection. J Virol 2017;9122:
[37] Brockwell N, Rautela J, Owen KL, et al. Tumor inherent interferon regulators as biomarkers of long-term chemotherapeutic response in TNBC. NPJ Precis Oncol 2019;3:21.
[38] Quarello P, Garelli E, Carando A, et al. A 20-year long term experience of the Italian Diamond-Blackfan Anaemia Registry: RPS and RPL genes, different faces of the same disease? Br J Haematol 2020;190:93-104.
[39] Psyrri A, Kalogeras KT, Kronenwett R, et al. Prognostic significance of UBE2C mRNA expression in high-risk early breast cancer. A hellenic cooperative oncology group (HeCOG) study. Ann Oncol 2012;23:1422-7.
[40] Zhang J, Liu X, Yu G, et al. UBE2C is a potential biomarker of intestinal- type gastric cancer with chromosomal instability. Front Pharmacol 2018;9:847.