Research Article
Identification of a Ferroptosis-Related Signature Associated with Prognosis and Immune Infiltration in Adrenocortical Carcinoma
Xi Chen (D,1,2 Lijun Yan D,3 Feng Jiang D,4 Yu Lu [D,2 Ni Zeng D,5 Shufang Yang (D,2 and Xianghua Ma @D1,6
1 Department of Endocrinology, The First Affiliated Hospital of Nanjing Medical University, Nanjing 210029, China
2Department of Endocrinology, Taizhou Clinical Medical School of Nanjing Medical University (Taizhou People’s Hospital), Taizhou 225300, China
3Department of Hepatology, Nantong Third People’s Hospital Affiliated to Nantong University, Nantong 226000, China
4Department of Neonatology, Obstetrics and Gynecology Hospital of Fudan University, Shanghai 200011, China
5 Department of Dermatology, Affiliated Hospital of Zunyi Medical University, Zunyi 563000, China
6Department of Nutriology, The First Affiliated Hospital of Nanjing Medical University, Nanjing 210029, China
Correspondence should be addressed to Shufang Yang; 47607369@qq.com and Xianghua Ma; xianghuama@njmu.edu.cn
Received 9 April 2021; Revised 22 June 2021; Accepted 30 June 2021; Published 21 July 2021
Academic Editor: Giuseppe Reimondo
Copyright @ 2021 Xi Chen et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Adrenocortical carcinoma (ACC) is a rare malignant tumor with poor prognosis. Ferroptosis, a new form of cell death, differs from other forms of cell death and plays a vital role in tumor progress. Our study aimed to establish a ferroptosis-related signature with prognostic value in ACC. RNA-seq data and corresponding clinical characteristics for ACC were downloaded from TCGA and GEO databases. Genes included in ferroptosis risk signature were assessed by univariable and multivariable Cox regression analysis as well as lasso regression analysis. The prognostic value of the ferroptosis risk signature was assessed using K-M and ROC curves. Furthermore, we performed GSEA to discover the enriched gene sets in high-risk group. Additionally, TIMER website was applied to detect a possible connection between the signature and immune cells infiltration. ssGSEA was performed to evaluate scores of immune cells and immune-related pathways in two groups. A ferroptosis signature comprised of six genes (SLC7A11, TP53, HELLS, ACSL4, PCBP2, and HMGB1) was constructed to predict prognosis and reflect the immune infiltration in ACC. Patients in high-risk group were inclined to have worse prognosis. The ferroptosis model performed well in predicting prognosis and could be served as an independent indicator in ACC. GSEA revealed that gene sets correlated with biological processes including cell cycle, DNA replication, base excision repair, and P53 signaling pathway were highly enriched in high-risk group. In addition, we discovered that the expressional levels of hub genes were linked to six immune cells’ infiltration in ACC tumor. ssGSEA revealed that contents of most immune cells significantly decreased in the high-risk group. In conclusion, the novel ferroptosis risk signature could be useful in predicting prognosis and reflecting immune infiltration in ACC. It also brings us new insights into the possible value of targeting ferroptosis during the therapy of ACC.
1. Introduction
Adrenocortical carcinoma (ACC) is a rare and aggressive malignant tumor derived from adrenal cortex with dismal prognosis [1]. Although it is advantageous for ACC patients to receive complete surgical resection or treatment with mitotane, the 5-year survival rate is less than 40% [2, 3]. Meanwhile, prognosis varies based on age, scope of surgery, mitotic intensity, and hormone secretion. The present
tumor, lymph node, and metastasis (TNM) classification method is unreliable in predicting prognosis owing to the heterogeneous features of ACC patients. It is challenging to make accurate prediction for ACC patients because of the diverse pathogenic factors, high heterogeneity, and poor prognosis. Hence, it is imperative to identify more effective biomarkers for predicting prognosis of ACC patients.
Ferroptosis is a novel iron-dependent form of regulated cell death along with iron accumulation and lipid
peroxidation [4]. In the aspect of morphology, biochemistry, and genetics, ferroptosis differs from other forms of cell death such as apoptosis, necroptosis, autophagy, and pyroptosis [5]. Emerging researches demonstrated that ferroptosis is implicated in neurous system disorders and plenty of cancers [6, 7]. Several research studies have re- cently mined the online databases for identifying prognostic signatures based on ferroptosis-related genes in diverse cancers. Luo et al. produced a new ferroptosis-related sig- nature that may predict prognosis in uveal melanoma pa- tients [8]. For predicting the prognosis of low-grade gliomas, Zheng et al. constructed a risk signature that included 12 ferroptosis-associated genes [9]. However, no research has yet determined whether ferroptosis-related genes are linked to ACC patients’ prognosis.
Firstly, we downloaded the mRNA expression data and relevant clinical information of ACC patients from public datasets. Then, based on the ferroptosis-related genes as- sociated with overall survival (OS) from The Cancer Ge- nome Atlas (TCGA) cohort, we developed a prognostic risk signature and validated it in Gene Expression Omnibus (GEO). We further performed Gene Set Enrichment Analysis (GSEA) to explore the underlying mechanisms. Finally, we evaluated the potential associations between the prognostic genes and immune cells based on the Tumor Immune Estimation Resource (TIMER).
2. Methods
2.1. Data Collection. As a training set, TCGA database was applied to collect the mRNA expression and related clini- copathological data of 79 ACC patients. In addition, 21 ACC patients with survival information from GEO database (GSE19750) were retrieved as a validation set. Supple- mentary Material Table S1 listed the detailed clinical in- formation. A list of 130 ferroptosis-associated genes was downloaded from GeneCards detailed in Table S2. No ethical approval was required because the data we utilized were obtained from public databases.
2.2. Establishment and Validation of a Prognostic Ferroptosis Risk Signature. To construct a ferroptosis risk signature in ACC patients, univariate Cox regression analysis was applied to assess the association between ferroptosis-associated genes and OS. PPI network diagram of the candidate prognostic ferroptosis-related genes was drawn using STRING online database to explore the relationships between these genes. Ferroptosis-related genes with a p value < 0.05 in univariate analysis were considered as candidate genes and recruited into lasso-penalized Cox regression analysis to narrow the gene extent with independent prognostic value. Further multivariate Cox regression analysis was then used to elim- inate the possible interaction among the candidate genes and obtain the coefficients. The risk score value of each patient was calculated by the following formula:
risk score = > (coef mRNAi * expression of mRNAi). (1) i=1 n
Coef was the coefficient calculated by multivariable Cox regression. Risk score was calculated for each individual, and total patients in the TCGA and GEO databases were allo- cated into high- and low-risk groups according to the median risk score value. PCA and t-SNE were adopted to explore whether the risk model had reliable clustering ability. Kaplan-Meier (K-M) curves were generated to compare the survival difference in two groups. Moreover, receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) were applied to evaluate the prognostic value of risk signature for OS in ACC patients. Furthermore, we performed univariate and multivariate Cox analysis to determine whether the risk score could serve as an independent factor for OS. We further applied GEO data to verify the above results through the same methods.
2.3. GSEA. To assess the potential molecular mechanisms underlying our risk signature, GSEA was applied to identify enriched terms correlated with KEGG pathway in high-risk group. Significant gene sets were classified as those with normalized enrichment score (NES) >1 and minimal p value < 0.05.
2.4. TIMER and ssGSEA. TIMER is an integrated website, which could measure immune infiltrate levels in various cancers, including ACC. In our study, we assessed the correlation between the hub ferroptosis-related genes with the contents of six immune cells, including CD4+ T cells, CD8+ T cells, B cells, neutrophils, dendritic cells, and macrophages in AGG via the TIMER. ssGSEA was per- formed to evaluate scores of 13 immune-related pathways and 16 immune cells in two risk groups.
3. Results
3.1. Construction of a Ferroptosis Risk Signature. Flow chart of our research was displayed in Supplementary Figure S1. To construct a ferroptosis risk signature and explore its prognostic value in ACC patients, a total of 94 overlapping ferroptosis-associated genes derived from the TCGA and GEO database were preserved for further analysis. Then, univariate Cox regression analysis was performed to assess the association between the expression levels of these 94 genes with ACC clinical survival information in the TCGA dataset. We found 31 ferroptosis-associated genes correlated with OS of ACC patients (Figure 1(a), p<0.05). Figure 1(b) displayed the PPI network suggesting that TP53, HMGB1, CDKN2A, and MAPK1 were the hub genes. The association among these genes was exhibited in Figure 1(c). 11 fer- roptosis-associated genes were finally reserved after lasso regression analysis (Figure 1(d)). In addition, through further multivariate Cox regression analysis to eliminate the possible interaction among the candidate genes, six genes (SLC7A11, TP53, HELLS, ACSL4, PCBP2, and HMGB1) were screened out (p<0.05), suggesting their strong cor- relations to the OS of ACC patients (Figure 1(f)). The multivariate Cox regression coefficients and expression
levels of the six ferroptosis-associated genes were used to construct a ferroptosis risk model. The formula for risk score calculation is as follows: risk score = (0.88 * SLC7A11) + (0.50 * TP53) + (1.37 * HELLS) - (1.37 * ACSL4) + (1.06 * PCBP2) + (1.55 * HMGB1).
3.2. Prognostic Performance of the Ferroptosis Risk Signature in TCGA. As seen in Figure 2(a), most of ferroptosis-associ- ated genes were linked with a higher risk score in TCGA cohort. Based on the median risk score, individuals in TCGA were assigned to high-/low-risk group (Figure 2(b)). Patients in high-risk class had higher mortality than those in low-risk class (Figures 2(c) and 2(d)). PCA and t-SNE showed that individuals in disparate categories were distributed in dif- ferent directions (Figures 2(e) and 2(f)). Furthermore, pa- tients in high-risk group were inclined to have shorter OS than those in low-risk group (Figure 2(g)). In addition, the AUC in TCGA reached 0.909 at 1 year, 0.947 at 3 years, and 0.968 at 5 years, respectively (Figure 2(h)). These results suggested that the novel ferroptosis risk signature had a definite effect in predicting the prognosis of ACC patients. K-M curves for each of the six hub genes included in the risk model in ACC are demonstrated in Supplementary Figure S2, showing that lower expression of ACSL4 and higher expression of SLC7A11, TP53, HELLS, PCBP2, and HMGB1 were associated with poor survival. Considering heterogeneity of ACC patients, we further assess whether the ferroptosis risk signature had good performance in pre- dicting the prognosis of ACC patients in different stages. As shown in Supplementary Figure S3, patients in high-risk group had shorter OS than those in low-risk group in both stage 1-2 group (p=0.00062) and stage 3-4 group (p<0.0001), suggesting that our ferroptosis risk signature also performed well in predicting prognosis of ACC patients in different stage.
3.3. Validation of the Ferroptosis Risk Signature in GEO. In GEO cohort, the heatmap revealed that most of the ferroptosis-associated genes were also related with a higher risk score (Figure 3(a)). In the same way, ACC patients in GEO were categorized into two groups (Figure 3(b)). Similarly, PCA and t-SNE analysis showed that patients in GEO with different risk scores were distributed in disparate directions (Figures 3(e) and 3(f)). As shown in Figure 3(g), the OS of patients in high-risk class was obviously shorter than that of patients in low-risk group in GEO cohort (p = 0.006). In addition, the ROC curve was drawn to assess the predictive performance of the risk model. The AUC at 1-, 3-, and 5-year in GEO cohort were 0.618, 0.899, and 0.945, respectively, indicating that the risk signature had a favourable capacity in predicting prognosis of ACC patients (Figure 3(h)).
3.4. Independent Prognostic Value of the Ferroptosis Risk Signature. We performed univariate and multivariate Cox regression analyses to observe whether clinical charac- teristics (such as age, gender, T, N, M, and stage) and the
risk score are independent prognostic factors for OS. We discovered that the risk score and T staging were inde- pendent prognostic predictors for OS in TCGA cohort. Owning to incomprehensive clinical parameters, a further Cox regression was not conducted to assess the prognostic value in GEO cohort. In addition, we investigated the correlation between the six ferroptosis-associated genes with the pathological T staging in ACC patients. Heatmap showed the expressional profiles of the six ferroptosis- associated genes at different T staging in TCGA cohort (Figure 4(c)). As drawn in Figure 4(d), the expressional levels of the major ferroptosis-associated genes, except ACSL4, which was considered as a protective gene, were generally higher in ACC patients at advanced T staging. We further compared the prognostic efficiency of our ferroptosis risk signature with other common prognostic factors, including age, gender, T, N, M, and stage. As shown in Figure 4(e), our ferroptosis risk signature (AUC=0.909) demonstrated significantly better predic- tion of ACC patients’ OS at 1 year than age (AUC=0.707), gender (AUC=0.438), T staging (AUC=0.649), N staging (AUC=0.438), M staging (AUC=0.528), and stage (AUC=0.587). These results suggested that our risk sig- nature performed better than other common prognostic characteristics.
3.5. GSEA for Identifying the Ferroptosis-Associated Signaling Pathways. We conducted GSEA to compare the biological signaling pathways between two groups. It was noteworthy that enriched gene sets linked to cell cycle, base excision repair, DNA replication, and P53 signaling pathway were highly enriched in high-risk group in both TCGA and GEO datasets (Figures 5(a) and 5(b).
3.6. Relationships between the Ferroptosis-Associated Genes and Immune Infiltration. Numerous studies demonstrated that the infiltration of cancer-related immune cells is asso- ciated with tumor development and prognosis. To identify whether there was a link between immune infiltration and the expressional levels of hub genes, we used TIMER to assess the correlation between the 6 hub genes and tumor purity along with six types of immune cells. As exhibited in Figure 6(a), we found the association between SLC7A11 expression and in- filtration levels of B cells (r=0.273, p=1.94e-02), CD8+ T cells (r =- 0.007, p=9.55e-01), CD4+ T cells (r =- 0.037, p=7.55e-01), macrophages (r=0.106, p=3.72e-01), neu- trophils (r=0.051, p=6.69e-01), and DCs (r=0.206, p = 8.06e-02) in ACC. Besides, the other 5 hub ferroptosis- associated genes (TP53, HELLS, ACSL4, PCBP2, and HMGB1) included in the signature also showed significant correlation with the infiltrating levels of B cells (r = 0.159 to 0.387, p<0.001), CD8+ T cells (r =- 0.078 to 0.211, p<0.001), CD4+ T cells (r =- 0.019 to 0.293, p<0.001), macrophages (r =- 0.155 to 0.304, p<0.001), neutrophils (r=0.044 to 0.257, p < 0.001), and DCs (r=0.206, p<0.001) in ACC (Figures 6(b)-6(f)). In sum, these results revealed that these 6 hub genes were in varying degrees related to tumor- associated immune cells in the ACC microenvironment.
p value
Hazard ratio
PRONS
STEAP3
HSPB1
0.009 0.684 (0.514-0.911)
2
RIPK1
PRDX6
HSPBI
BAP1
SLC7A11
<0.001 2.216 (1.568-3.130)
HILPDA
RIPK1
0.033 2.408 (1.071-5.413)
HISPB1
E
RIPKI
ELAVLI
<0.0017.029 (2.471-19.994)
BAP1
EGLNI
PRC1
2.246 (1.652-3.053)
F
NF2
AURKA
<0.001 2.496 (1.819-3.426
HMG81
PCBP2
MAPKI
YAPI
0.025 1.857 (1.080-3.195)
MAPK1
VDAC2 HMG B1
SLC7A11
TP53
ATF4
MIF
2.346 (1.488-3.698)
MYC
HELLS
<0.001 5.289 (3.116-8.980)
CE
ACSLA
<0.001 0.542 (0.390-0.754
YAPI
ATF4
HSPAS
ELAVLI
ARNTI L
FANOD2 HSPA5
TFRC
PRDX6
0.014
YAP1
2.073
1.158-3.712
0.002 2.295 (1.358-3.879
F
NF2
<0.001 3.533 (1.681-7.425)
SLCTA11
HILPDA
HELLS
VDAC2
AURKA
MUCI
PCBP2
SLC40A1
0.001
0.002
2.532 (1.451-4.420)
0.632
0.473-0.843
11
CDKNZA
W
BAP1
0.016 1.991 (1.138-3.484
H
TP53
PRCI
ATF4
AIFM2
CDKNZA
HMGB1
CDKN2A
<0.0014.756
1.958-11.556
<0.001
773
5.465
1.289-2.438
MUC1
AIFM2
FANCD2
HILPDA
<0.001
3.047-9.802
MYC
ELAVIL1
MIT
0.017
1.407 (1.064-1.862
TP53
AIFM2
<0.001 2.120 (1.369-3.281
STEAP3
0.011 1.454 (1.090-1.939)
IH
H
EGLNI
HELLS
MYC
<0.001
1.780 (1.325-2.391)
PCBP2
SLC40A1
TFRC
<0.001 1.682 (1.271-2.225
FANCD2
€
MAPK1
0.013 2.176 (1.181-4.011)
H
C
AURKA
ACSLA
MIF
0.013
1.427
1.078-1.889
STEAPS
HSPA5
0.005 2.314(1.292-4.1
.144
IH
ACSLA
1
0.5
0
-0.5
-1
MUC1
<0.001 1.375 (1.155-1.637)
50
ARNTL
PRC1
TFRC
ARNTL
0.003
2.180
1.312
.622
TH
U
SLCADA1
EGLNI
0.020
1.767
1,095
.853
VDAC2
0.025 2.193 (1.104-4.358)
0
5
10
15
Hazard ratio
(a)
(b)
(c)
16
4 12 12
14
11
10
8
8
7
5
4
3
3
1
16
12
10
7
3
0
10.0
p value
Hazard ratio
1:
9.5
8
SLC7A11
<0.001
2.386 (1.533-3.714)
1
Partial liklihood deviance
0.5
9.0
24
TP53
0.042
1.647 (1.018-2.666)
4
Coefficients
8.5
9
0.0
HELLS
<0.001
3.916(2.150-7.133)
A
8.0
ACSL4
<0.001
0.402 (0.250-0.648)
7.5
-0.5
PCBP2
0.010
2.882 (1.292-6.430)
7.0
HMGB1
0.011 4.697 (1.425-15.481)
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
-3.5
-3.0
-2.5
-2.0
-1.5
-1.0
0
2
4
6
8
10
14
Log(A)
Log lambda
Hazard ratio
(d)
(e)
(f)
Type
3
10
ACSL4
2
8
1
Risk score
6
TP53
0
4
-1
PCBP2
-2
2
-3
0
SLC7A11
0
20
40
60
80
Patients (increasing risk socre)
HELLS
High risk
Low risk
HMGB1
Type
High
Low
(a)
(b)
Survival time (years)
10
Low risk
98%
2%
8
6
4
2
High risk
31%
69%
0
0
20
40
60
80
Patients (increasing risk socre)
Dead
25
50
75
Alive
0
100
Percent
Status
Dead
Alive
(c)
(d)
30
2
20
S
10
PC2
tSNE2
0
0
-2
-10
-20
-4
-2
0
2
4
-20
-10
0
10
20
PC1
30
tSNE1
Risk
Risk
High
High
Low
Low
(e)
(f)
1.00
1.0
Survival probability
0.75
0.8
0.50
Sensitivity
0.6
0.25
0.4
P < 0.001
0.00
0
1
0.2
2
3
4
5
6
7
8
9
10
11
12
Time (years)
Risk
High risk Low risk
39
35
21
16
8
4
3
2
1
1
1
0
0
0.0
40
40
37
28
22
20
13
9
7
6
3
2
2
0
1
2
3
4
5
6
7
8
9
10
11
12
0.0
0.2
0.4
0.6
0.8
1.0
Time (years)
1-specificity
Risk
AUC at 1 year: 0.909
High risk
AUC at 3 years: 0.947
+ Low risk
AUC at 5 years: 0.968
(g)
(h)
To identify the association between our signature and immune microenvironment condition in ACC, we further performed ssGSEA to evaluate the scores of 16 immune cells and 13 immune-related pathways in TCGA cohort. We found that the contents of most immune cells in high-risk group, including aDCs, B cells, CD8+T cells, iDCs, mast cells, neutrophils, NK cells, pDCs, T helper cells, Tfh, Th2 cells, and Treg, were significantly lower than those in low-risk group (Figure 7(a)). Moreover, the scores of the most immune- related pathways were lower in high-risk group (Figure 7(b)). The above results suggested an immune suppressive micro- environment in ACC patients with high-risk scores.
4. Discussion
ACC is a rare endocrine malignancy with poor prognosis. For early diagnosis, more effective therapy, accurate prog- nosis of ACC, novel biomarkers, and prognostic signatures are required. Although there are some single gene and risk models linked to ACC patients’ prognosis, no ferroptosis- related risk signature was reported for predicting prognosis in ACC. In our study, the associations between 103 fer- roptosis-related genes and OS as well as immune cells in- filtration were investigated in ACC patients. A new prognostic risk signature including six ferroptosis-
Type
10
2
TP53
8
1
Risk score
6
ACSI4
0
4
-1
HELLS
-2
2
0
.
.
.
5
6
HMGB1
5
10
15
20
Patients (increasing risk socre)
SLC7A11
High risk
PCBP2
Low risk
Type
High
Low
(a)
(b)
Survival time (years)
15
Low risk
22%
78%
10
5
High risk
17%
63%
0
5
10
15
20
Patients (increasing risk socre)
Dead
0
25
50
75
100
Alive
Percent
Status
Dead
Alive
(c)
(d)
100
1
50
0
PC2
tSNE2
-1
0
-2
-50
-3
-100
-3
-2
-1
0
1
2
3
-100
-50
0
50
PC1
tSNE1
Risk
Risk
High
High
Low
Low
(e)
(f)
1.00
1.0
Survival probability
0.75
0.8
0.50
Sensitivity
0.6
0.25
P = 0.006
0.4
0.00
0
0.2
1
2
3
4
5
6
7
8
9
10
11
12
13
1
15 5
6
1
18
19
20
Time (years)
0.0
Risk
High risk
12
8
6
4
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
Low risk
9
9
9
8
8
8
8
6
5
5
3
3
3
3
3
2
2
1
1
0
0
0.0
0.2
0.4
0.6
0.8
1.0
0
1
2
3
4
5
6
7
8
9
10
1
12
13
1
1
1
7
18 8 1
19
20
1-specificity
Time (years)
AUC at 1 year: 0.618
Risk
AUC at 3 years: 0.899
High risk
AUC at 5 years: 0.945
Low risk
(g)
(h)
p value
Hazard ratio
p value
Hazard ratio
Age
0.328 1.012 (0.988-1.038)
Gender
0.972 0.986 (0.451-2.154)
T
0.016
3.458 (1.255-9.528)
T
<0.001 3.378 (2.110-5.407)
M
0.554
1.744 (0.277-10.979)
N
0.152 2.038 (0.769-5.400)
M
<0.001 6.150 (2.710-13.959)
Stage
0.841
0.852 (0.176-4.113)
H
Stage
<0.001 2.914 (1.860-4.565)
Risk score <0.001 1.017 (1.011-1.023)
Risk score
<0.001
1.016(1.009-1.023)
0
2
4
6
8
10
12
0
2
4
6
8
10
Hazard ratio
Hazard ratio
(a)
(b)
T
3
SLC7A11
TP53
HELLS
ACSI4
PCBP2
HMGB1
ACSL4
2
·0.6-
-0.43
0.11.
0.68
0.94
0.24
1
0:2
0:37
-le - 04.
0:4
0.16
0.014
TP53
0
0.22
-0.93
0.24
-0.2-
0.41.
0.49
-1
0.38
0.017
0.35
0.095
PCBP2
2.9e -05
0.067
10
-2
0.42
0.093
:0.0079
0.2
0.54
0.48
SLC7A11
Gene expression
0.99
0.031
0.069
0.65
0.43
0.72
-3
HELLS ***
HMGB1
5
T
T1
T3
T2
T4
T1
T2 T3 T4 T1
T2
T3
T4
T1
T2
T3
T4
T1 T2 2 T3
T4
T1
T2
T3
T4
T1
[2 T3
T4
T
T
T1
T3 T4
E T2
(c)
(d)
1.0
0.8
Sensitivity
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
1-specificity
Risk, AUC = 0.909
N, AUC = 0.438
Age, AUC = 0.707
M, AUC = 0.528
Gender, AUC = 0.438
Stage, AUC = 0.587
T, AUC = 0.649
(e)
associated genes correlated with immune cells infiltration was firstly constructed in ACC patients.
Ferroptosis, a novel form of cell death, emphasizes the importance of iron synthesis and metabolism, which was firstly proposed in 2012. On account of discovering the unique cell death form, numerous researches are focusing on the exploration of the potential mechanisms and therapy related to ferroptosis in multiple cancers. Previous
studies show that ferroptosis also opens up a new potential avenue for cancer development and treatment [10, 11]. Furthermore, the expressional level of ferroptosis-related gene GPX4 and the sensitivity to ferroptosis were signifi- cantly increased in ACC, indicating that ACC patients may be susceptible to induction of ferroptosis [12]. In this study, we constructed a novel ferroptosis risk signature with powerful value in predicting ACC patients’ prognosis for
TCGA
Enrichment score (ES)
Enrichment plot: KEGG_CELL_CYCLE
Enrichment score (ES)
Enrichment plot: KEGG_BASE_ EXCISION_REPAIR
Enrichment score (ES)
Enrichment plot: KEGG_DNA_REPLICATION
Enrichment score (ES)
Enrichment plot: KEGG_P53 SIGNALING_PATHWAY
0.7
0.7
0.6
0.6
0.7
0.6
0.5
NES1.82
0.5
NES1.74
0.6
0.5
0.5
NES1.65
0.4
NES1.71
0.4
0.4
0.3
p-val=0.006
0.3
p-val=0.008
0.4
0.2
0.3
p-val=0.036
0.3
0.2
0.2
p-val=0.006
0.1
0.1
0.0
0.1
0.1
0.0
0.0
0.0
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
2
h’ (positively correlated)
2
‘h’ (positively correlated)
2
h’ (positively correlated)
2
h’ (positively correlated)
1
Zero cross at 32237
1
Zero cross at 32237
1
Zero cross at 32237
1
Zero cross at 32237
0
0
0
0
-1
T’ (negatively correlated)
-1
T’ (negatively correlated)
-1
T’ (negatively correlated)
-1
‘T’ (negatively correlated)
0
10,000 20,000 30,000 40,000 50,000
0
10,000 20,000 30,000 40,000 50,000
0
10,000 20,000 30,000 40,000 50,000
0
10,000 20,000 30,000 40,000 50,000
Rank in ordered dataset
Rank in ordered dataset
Rank in ordered dataset
Rank in ordered dataset
Enrichment profile
Enrichment profile
Enrichment profile
Enrichment profile
Hits
Hits
Hits
Hits
Ranking metric scores
Ranking metric scores
Ranking metric scores
Ranking metric scores
(a)
Enrichment score (ES)
Enrichment plot: KEGG_CELL_CYCLE
Enrichment score (ES)
Enrichment plot: KEGG_BASE_ EXCISION_REPAIR
Enrichment score (ES)
Enrichment plot: KEGG_DNA_REPLICATION
Enrichment score (ES)
Enrichment plot: KEGG_P53 SIGNALING_PATHWAY
0.7
0.6
0.6
0.7
0.5
0.5
NES1.74
0.5
NES1.70
0.6
0.4
0.4
0.4
0.5
NES1.84
0.3
0.3
0.3
NES1.66
0.4
0.2
p-val=0.008
0.2
p-val=0.012
0.3
p-val=0.002
0.2
p-val=0.002
0.1
0.1
0.2
0.0
0.0
0.1
0.1
0.0
0.0
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
Ranked list metric (Signal2Noise)
2
h’ (positively correlated)
2
‘h’ (positively correlated)
2
h’ (positively correlated)
2
h’ (positively correlated)
1
Zero cóósá at 32237
1
Zero cross at 10637
1
Zero cross at 10637
1
0
0
0
0
Zero cross at 10637
-1
T’ (negatively correlated)
-1
10,000 20,000 30,000 40,000 50,000
-1
T (negatively correlated)
-1
T (negatively correlated)
T (negatively correlated)
0
0
5,000
10,000
15,000
20,000
0
5000
10,000
15,000
20,000
0
5,000
10,000
15,000
20,000
Rank in ordered dataset
Rank in ordered dataset
Rank in ordered dataset
Rank in ordered dataset
Enrichment profile
Enrichment profile
Enrichment profile
Enrichment profile
Hits
Hits
Hits
Hits
Ranking metric scores
Ranking metric scores
Ranking metric scores
Ranking metric scores
GEO
(b)
SLC7A11 expression level (log2 TPM)
Purity
B cell
CD8 + T cell
CD4 + T cell
Macrophage
Neutrophil
Dendritic cell
cor = 0.156
p = 1.84e-01
partial.cor = 0.273
p = 1.94e-02
partial.cor = - 0.007
p = 9.55e-01
partial.cor = - 0.037
p = 7.55e-01
partial.cor = 0.106
3
p = 3.72e-01
partial.cor = 0.051
p = 6.69e-01
partial.cor = 0.206
p = 8.06e-02
..
2
ACC
1
0
-1
0.2
0.4
0.6
0.8
1.0
0.11
0.12
0.13
0.20
0.25
0.30
0.35 0.07
0.09
0.11
0.13
0.15
0.08
0.12
0.16
0.12
0.14
0.16
0.18
0.49
0.50
0.51
0.52
0.53
Infiltration level
(a)
Purity
B cell
CD8 + T cell
CD4 + T cell
Macrophage
Neutrophil
Dendritic cell
TP53 expression level (log2 TPM)
cor = 0.309
partial.cor = - 0.019
partial.cor = - 0.06
7
p = 7.45e-03
partial.cor = 0.159
p = 1.78e-01
partial.cor = 0.209
P = 7.56e-02
p = 8.75e-01
p = 6.13e-01
partial.cor = 0.176
p = 1.35e-01
partial.cor = 0.124
p = 2.97e-01
6
ACC
5
4
:
3
0.2
0.4
0.6
0.8
1.0
0.11
0.12
0.13
0.20
0.25
0.30
0.35 0.07
0.09
0.11
0.13
0.15
0.08
0.12
0.16
0.12
0.14
0.16
0.18
0.49
0.50
0.51
0.52
0.53
Infiltration level
(b)
Purity
B cell
CD8 + T cell
CD4 + T cell
Macrophage
Neutrophil
Dendritic cell
HELLS expression level (log2 TPM)
6
cor = 0.453
p = 5.13e-05
partial.cor = 0.236
p = 4.47e-02
partial.cor = - 0.078
p = 5.12e-01
partial.cor - 0.018
p = 8.83e-01
partial.cor — 0.155
p = 1.91e-01
partial.cor - 0.044
p = 7.14e-01
partial.cor - 0.138
p = 2.46e-01
4
2
ACC
0
-2
0.2
0.4
0.6
0.8
1.0
0.11
0.12
0.13
0.20
0.25
0.30
0.35 0.07
0.09
0.11
0.13
0.15
0.08
0.12
0.16
0.12
0.14
0.16
0.18
0.49
0.50
0.51
0.52
0.53
Infiltration level
(c)
Purity
B cell
CD8 + T cell
CD4 + T cell
Macrophage
Neutrophil
Dendritic Cell
ACSL4 expression level (log2 TPM)
cor = - 0.188
p = 1.09e-01
partial.cor = 0.277
p = 1.76e-02
partial.cor = 0.157
p = 1.85e-01
partial.cor = 0.066
p = 5.80e-01
partial.cor = 0.304
p = 9.04e-03
partial.cor = 0.187
p = 1.14e-01
partial.cor = 0.216
p = 6.66e-02
7.5
ACC
5.0
2.5
0.2
0.4
0.6
0.8
1.0
0.11
0.12
0.13
0.20
0.25
0.30
0.35 0.07
0.09
0.11
0.13
0.15
0.08
0.12
0.16
0.12
0.14
0.16
0.18
0.49
0.50
0.51
0.52
0.53
Infiltration level
(d)
Purity
B cell
PCBP2 expression level (log2 TPM)
CD8 + T cell
CD4 + T cell
Macrophage
Neutrophil
Dendritic Cell
10
cor = 0:317
p = 5.850-03
partial.cor = 0.387
p = 7.28e-04
partial.cor = 0.22
P = 6.21e-02
partial.cor = 0.293
P .= 1.19e-02
partial.cor = 0.207
p = 7.96e-02
partial.cor = 0.207
p = 7.87e-02
partial.cor = 0.334
p = 3.84e-03
9
8
ACC
7
6
0.2
0.4
0.6
0.8
1.0
0.11
0.12
0.13
0.20
0.25
0.30
0.35 0.07
0.09
0.11
0.13
0.15
0.08
0.12
0.16
0.12
0.14
0.16
0.18
0.49
0.50
0.51
0.52
0.53
Infiltration level
(e)
HMGB1 expression level (log2 TPM)
Purity
B cell
CD8 + T cell
CD4 + T cell
Macrophage
Neutrophil
Dendritic cell
cor = 0.322
p = 5.21e-03
partial.cor - 0.343
p = 2.98e-03
partial.cor - 0.211
P = 7.29e-02
partial.cor - 0.232
p = 4.85e-02
partial.cor - 0.275
p = 1.86e-02
partial.cor - 0.257
partial.cor - 0.421
8
p = 2.80€-02
p =2.10e-04
7
6
ACC
5
4
0.2
0.4
0.6
0.8
1.0
0.11
0.12
0.13
0.20
0.25
0.30
0.35 0.07
0.09
0.11
0.13
0.15
0.08
0.12
0.16
0.12
0.14
0.16
0.18
0.49
0.50
0.51
0.52
0.53
Infiltration level
(f)
the first time. Furthermore, the risk signature had better prognostic efficiency than other common prognostic fac- tors, including age, gender, T, N, M, and stage of ACC patients. The six ferroptosis-related genes adopted in the risk signature contained the risk-related genes (SLC7A11, TP53, HELLS, PCBP2, and HMGB1) and the protective gene (ACSL4).
These hub genes can be crudely categorized into four classes, including iron metabolism (PCBP2), lipid meta- bolism (ACSL4, HELLS), (anti)oxidant metabolism (SLC7A11, HMGB1), and cancer metabolism (TP53) [11]. Among these, tumor protein p53 (TP53) acts as an im- portant tumor suppressor in cancer development and progression. Apart from the impact on apoptosis and cell cycle, TP53 could regulate cancer ferroptosis in a dual manner at transcriptional or posttranslational levels [13]. By targeting DPP4 and inducing P21 expression, TP53 could inhibit the ferroptosis. Conversely, ferroptosis could be enhanced by the inhibitory effect of TP53 on solute carrier family 7 member 11 (SLC7A11) in cancers [14]. Moreover, overexpression of SLC7A11, which had a high expression in several cancers, including ACC, inhibited the ferroptosis induced by ROS [14, 15]. Helicase lymphoid specific (HELLS, known as LSH), a chromatin remodeler, was shown to be linked with advanced stage and worse prognosis in pancreatic carcinoma, hepatocellular carcinoma, and na- sopharyngeal carcinoma [16-18]. In lung carcinoma, Jiang et al. demonstrated that HELLS could inhibit ferroptosis by stimulating ferroptosis-associated genes SCD1 and FADS2
and lipid metabolism-related gene GLUT1 [19]. Poly(rC) binding protein 2 (PCBP2), an RNA-binding adaptor pro- tein, could bind and deliver iron to ferritin for storage. Higher levels of PCBP2 are connected with worse prognosis in glioblastoma and gastric cancer [20, 21]. High mobility group box 1 (HMGB1), a nuclear protein, releases under the exposure to ferroptosis activators [22]. Ye et al. found that HMGB1 is a novel regulator of ferroptosis through RAS- JNK/p38 pathway in leukemia [23]. In addition, acyl-CoA synthetase long-chain family member 4 (ACSL4), a vital protein in ferroptosis, was overexpressed and served as an independent prognostic indicator in various cancers [24]. It is reported that ACSL4 is both a sensitive regulator and an effective inducer of ferroptosis [25]. In short, previous re- search revealed that these six genes are closely connected with ferroptosis and tumorigenesis, providing a powerful theoretical foundation for our risk model based on fer- roptosis-related genes. Moreover, it was reported that ACC was associated with abnormal p53 signaling and frequent genetic alterations in TP53 [26]. In the study based on the comprehensive genomic characterization of 91 ACC pa- tients, TP53 somatic alterations were reported to be the most frequent gene with genetic alterations [27]. TP53 p.R337H mutation is highly prevalent among children with ACC, accounting for 90% of ACC cases in Southern Brazil [28]. Although limited research focusing on the effects of these six hub genes on ACC have been published, we found that lower expression of ACSL4 and higher expression of SLC7A11, TP53, HELLS, PCBP2, and HMGB1 were related to poor OS
1.00
**
∗
∗
ns
**
ns
∗
**
∗
**
∗
ns
**
0.75
Score
0.50
0.25
0.00
aDCs
B_cells
CD8+_T_cells
DCs
iDCs
Macrophages
Mast_cells
Neutrophils
NK_cells
pDCs
T_helper_cells
Tfh
Th1_cells
Th2_cells
TIL
Treg
Risk
Low
High
(a)
1.0
**
**
**
**
ns
**
**
**
ns
0.8
Score
0.6
0.4
APC_co_inhibition
APC_co_stimulation
CCR
Check-point
Cytolytic_activity
HLA
Inflammation-promoting
MHC_class_I
Parainflammation
T_cell_co-inhibition
T_cell_co-stimulation
Type_I_IFN_Response
Type_II_IFN_Response
Risk
Low
₿ High
(b)
of patients with ACC in our study, and the underlining mechanism needed further study. In addition, it is reported that adrenal cortex cells are extremely sensitive to ferroptosis due to their steroidogenic properties. Mitotane, as the only available drug applied in the treatment of ACC, is unable to induce ferroptosis [15]. Hence, it might be very promising in developing new drugs by inducing ferroptosis for ACC in years to come.
According to recent studies, ferroptosis could play a vital role in tumor immunotherapy [29-31]. IFNy, released by CD8+ T cells, regulates lipid peroxidation and ferroptosis- related pathways in tumors. Several studies have shown that cells under the condition of ferroptosis could modulate anticancer immunity by releasing chemotaxis interacted with immune cells, such as NK and CD8+ T cells [30, 32]. Furthermore, a previous study reported that iron
metabolism-related genes FPN1 and CP might participate in tumor immune microenvironment of ACC [33]. Of note, we mentioned that the ferroptosis-related genes in our signa- ture have significant associations with immune cells, im- plying that ferroptosis and immunity in ACC microenvironment are complicated. Meanwhile, the lower scores of immune cells and immune-related functions in high-risk group suggested an immune suppressive micro- environment in ACC patients with high-risk scores, im- plicating that the poor prognosis of patients in the high-risk group might be caused by the immunosuppressive status. However, the underlying mechanisms between ferroptosis- related genes and tumor immunity in ACC remain poorly understood and warrant further investigation.
As far as we know, this is the first study aiming at constructing and validating a ferroptosis risk signature in patients with ACC. These results revealed that our ferrop- tosis risk signature could be considered as a powerful tool for predicting prognosis and reflecting the immune infiltration in ACC. However, there are several limitations in our study. Firstly, our study was retrospective based on public datasets, so further large-scale prospective researches and clinical trials are needed for validation of the prognostic ability. Besides, the mechanism of how ferroptosis modulates the development of ACC was not verified by functional ex- periments. Thus, further researches are required to confirm our findings and explore the underlying mechanisms before application in clinical practice.
Data Availability
RNA-seq data and clinical information used to support the findings of this study were collected from the Cancer Ge- nome Atlas (TCGA) (https://cancergenome.nih.gov/) and Gene Expression Omnibus (GEO) repository (GSE19750).
Conflicts of Interest
All authors declare no conflicts of interest.
Acknowledgments
This work was supported by the Key Social Development Program of Jiangsu Province (No. BE2017736).
Supplementary Materials
Figure S1: flow chart of the study. Figure S2: K-M curves for each of the 6 hub genes in ACC patients. Figure S3: Kaplan-Meier curves for patients assigned to high- and low- risk groups based on the risk score in TCGA cohort. Table S1: clinical characteristics of ACC patients in the TCGA and GEO. Table S2: 103 ferroptosis-related genes downloaded from the GeneCards. (Supplementary Materials)
References
[1] S. Jasim and M. A. Habra, “Management of adrenocortical carcinoma,” Current Oncology Reports, vol. 21, no. 3, p. 20, 2019.
[2] S. H. Tella, A. Kommalapati, S. Yaturu, and E. Kebebew, “Predictors of survival in adrenocortical carcinoma: an analysis from the national cancer database,” The Journal of Clinical Endocrinology & Metabolism, vol. 103, no. 9, pp. 3566-3573, 2018.
[3] E. Mirallié, C. Blanchard, C. Caillard et al., “Adrenocortical carcinoma: impact of surgical treatment,” Annales d’Endoc- rinologie, vol. 80, no. 5-6, pp. 308-313, 2019.
[4] T. Hirschhorn and B. R. Stockwell, “The development of the concept of ferroptosis,” Free Radical Biology and Medicine, vol. 133, pp. 130-143, 2019.
[5] B. R. Stockwell, J. P. Friedmann Angeli, H. Bayir et al., “Ferroptosis: a regulated cell death nexus linking metabolism, redox biology, and disease,” Cell, vol. 171, no. 2, pp. 273-285, 2017.
[6] Y. Qiu, Y. Cao, W. Cao, Y. Jia, and N. Lu, “The application of ferroptosis in diseases,” Pharmacological Research, vol. 159, Article ID 104919, 2020.
[7] A. Weiland, Y. Wang, W. Wu et al., “Ferroptosis and its role in diverse brain diseases,” Molecular Neurobiology, vol. 56, no. 7, pp. 4880-4893, 2019.
[8] H. Luo and C. Ma, “A novel ferroptosis-associated gene signature to predict prognosis in patients with uveal mela- noma,” Diagnostics (Basel), vol. 11, 2021.
[9] Y. Zheng, Q. Ji, L. Xie et al., “Ferroptosis-related gene sig- nature as a prognostic marker for lower-grade gliomas,” Journal of Cellular and Molecular Medicine, vol. 25, no. 6, pp. 3080-3090, 2021.
[10] C. Liang, X. Zhang, M. Yang, and X. Dong, “Recent progress in ferroptosis inducers for cancer therapy,” Advanced Ma- terials (Deerfield Beach, Fla.), vol. 31, Article ID e1904197, 2019.
[11] B. Hassannia, P. Vandenabeele, and T. Vanden Berghe, “Targeting ferroptosis to iron out cancer,” Cancer Cell, vol. 35, no. 6, pp. 830-849, 2019.
[12] A. Belavgeni, S. R. Bornstein, A. von Mässenhausen et al., “Exquisite sensitivity of adrenocortical carcinomas to in- duction of ferroptosis,” Proceedings of the National Academy of Sciences, vol. 116, no. 44, pp. 22269-22274, 2019.
[13] R. Kang, G. Kroemer, and D. Tang, “The tumor suppressor protein p53 and the ferroptosis network,” Free Radical Biology and Medicine, vol. 133, pp. 162-168, 2019.
[14] L. Jiang, N. Kon, T. Li et al., “Ferroptosis as a p53-mediated activity during tumour suppression,” Nature, vol. 520, no. 7545, pp. 57-62, 2015.
[15] I. Weigand, J. Schreiner, F. Röhrig et al., “Active steroid hormone synthesis renders adrenocortical cells highly sus- ceptible to type II ferroptosis induction,” Cell Death & Dis- ease, vol. 11, no. 3, p. 192, 2020.
[16] X. Hou, L. Yang, K. Wang et al., “HELLS, a chromatin remodeler is highly expressed in pancreatic cancer and downregulation of it impairs tumor growth and sensitizes to cisplatin by reexpressing the tumor suppressor TGFBR3,” Cancer Medicine, vol. 10, no. 1, pp. 350-364, 2021.
[17] X. He, B. Yan, S. Liu et al., “Chromatin remodeling factor LSH drives cancer progression by suppressing the activity of fu- marate hydratase,” Cancer Research, vol. 76, no. 19, pp. 5743-5755, 2016.
[18] C. T. Law, L. Wei, F. H. C. Tsang et al., “HELLS regulates chromatin remodeling and epigenetic silencing of multiple tumor suppressor genes in human hepatocellular carcinoma,” Hepatology, vol. 69, no. 5, pp. 2013-2030, 2019.
[19] Y. Jiang, C. Mao, R. Yang et al., “EGLN1/c-Myc induced lymphoid-specific helicase inhibits ferroptosis through lipid
metabolic gene expression changes,” Theranostics, vol. 7, no. 13, pp. 3293-3305, 2017.
[20] K. Luo and K. Zhuang, “High expression of PCBP2 is asso- ciated with progression and poor prognosis in patients with glioblastoma,” Biomedicine & Pharmacotherapy, vol. 94, pp. 659-665, 2017.
[21] C. Chen, J. Lei, Q. Zheng, S. Tan, K. Ding, and C. Yu, “Poly(rC) binding protein 2 (PCBP 2) promotes the viability of human gastric cancer cells by regulating CDK2,” FEBS Open Bio, vol. 8, no. 5, pp. 764-773, 2018.
[22] Q. Wen, J. Liu, R. Kang, B. Zhou, and D. Tang, “The release and activity of HMGB1 in ferroptosis,” Biochemical and Biophysical Research Communications, vol. 510, no. 2, pp. 278-283, 2019.
[23] F. Ye, W. Chai, M. Xie et al., “HMGB1 regulates erastin- induced ferroptosis via RAS-JNK/p38 signaling in HL-60/ NRASQ61L cells,” American Journal of Cancer Research, vol. 9, pp. 730-739, 2019.
[24] X .- J. Sun and G .- L. Xu, “Overexpression of Acyl-CoA Ligase 4 (ACSL4) in patients with hepatocellular carcinoma and its prognosis,” Medical Science Monitor, vol. 23, pp. 4343-4350, 2017.
[25] H. Yuan, X. Li, X. Zhang, R. Kang, and D. Tang, “Identifi- cation of ACSL4 as a biomarker and contributor of ferrop- tosis,” Biochemical and Biophysical Research Communications, vol. 478, no. 3, pp. 1338-1343, 2016.
[26] C. D. C. Kamilaris, F. Hannah-Shmouni, and C. A. Stratakis, “Adrenocortical tumorigenesis: lessons from genetics,” Best Practice & Research. Clinical Endocrinology & Metabolism, vol. 34, Article ID 101428, 2020.
[27] S. Zheng, A. D. Cherniack, N. Dewal et al., “Comprehensive pan-genomic characterization of adrenocortical carcinoma,” Cancer Cell, vol. 29, no. 5, pp. 723-736, 2016.
[28] G. Custódio, G. A. Parise, N. Kiesel Filho et al., “Impact of neonatal screening and surveillance for the TP53 R337H mutation on early detection of childhood adrenocortical tumors,” Journal of Clinical Oncology, vol. 31, no. 20, pp. 2619-2626, 2013.
[29] X. Lang, M. D. Green, W. Wang et al., “Radiotherapy and immunotherapy promote tumoral lipid oxidation and fer- roptosis via synergistic repression of SLC7A11,” Cancer Discovery, vol. 9, no. 12, pp. 1673-1685, 2019.
[30] W. Wang, M. Green, J. E. Choi et al., “CD8+ T cells regulate tumour ferroptosis during cancer immunotherapy,” Nature, vol. 569, no. 7755, pp. 270-274, 2019.
[31] B. R. Stockwell and X. Jiang, “A physiological function for ferroptosis in tumor suppression by the immune system,” Cell Metabolism, vol. 30, no. 1, pp. 14-15, 2019.
[32] L. Zitvogel and G. Kroemer, “Interferon-y induces cancer cell ferroptosis,” Cell Research, vol. 29, no. 9, pp. 692-693, 2019.
[33] B. Zhu, Q. Zhi, Q. Xie et al., “Reduced expression of ferro- portin1 and ceruloplasmin predicts poor prognosis in adre- nocortical carcinoma,” Journal of Trace Elements in Medicine and Biology, vol. 56, pp. 52-59, 2019.