Construction of a robust prognostic model for adult adrenocortical carcinoma: Results from bioinformatics and real- world data

Xi Tian1,2 iD Wen-Hao Xu1,2 Aihetaimujiang Anwaier1, 1,2 Hong-Kai Wang1,2

Fang-Ning Wan1,2 Da-Long Cao1,2 I Wen-Jie Luo1,2 iD Guo-Hai Shi1,2

Yuan-Yuan Qu1,2 Hai-Liang Zhang1,2 | Ding-Wei Ye1,2

1Department of Urology, Fudan University Shanghai Cancer Center, Shanghai, China

2Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, China

Correspondence

Yuan-Yuan Qu, Hai-Liang Zhang and Ding- Wei Ye, Department of Urology, Fudan University Shanghai Cancer Center, No. 270 Dong’an Road, Shanghai, 200032, China. Emails: quyy1987@163.com (Y .- Y. Q.); zhanghl918@163.com (H .- L. Z.); dwyelie@163.com (D .- W. Y.)

Funding information

This work is supported by Grants from National Key Research and Development Project (No.2019YFC1316000) and the National Natural Science Foundation of China (No.81802525).

Abstract

This study aims to construct a robust prognostic model for adult adrenocortical car- cinoma (ACC) by large-scale multiomics analysis and real-world data. The RPPA data, gene expression profiles and clinical information of adult ACC patients were obtained from The Cancer Proteome Atlas (TCPA), Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). Integrated prognosis-related proteins (IPRPs) model was constructed. Immunohistochemistry was used to validate the prognostic value of the IPRPs model in Fudan University Shanghai Cancer Center (FUSCC) cohort. 76 ACC cases from TCGA and 22 ACC cases from GSE10927 in NCBI’s GEO database with full data for clinical information and gene expression were utilized to validate the effectiveness of the IPRPs model. Higher FASN (P = . 039), FIBRONECTIN (P <. 001), TFRC (P <. 001), TSC1 (P <. 001) expression indicated significantly worse overall sur- vival for adult ACC patients. Risk assessment suggested significantly a strong predic- tive capacity of IPRPs model for poor overall survival (P < . 05). IPRPs model showed a little stronger ability for predicting prognosis than Ki-67 protein in FUSCC cohort (P = . 003, HR = 3.947; P = . 005, HR = 3.787). In external validation of IPRPs model using gene expression data, IPRPs model showed strong ability for predicting prog- nosis in TCGA cohort (P = . 005, HR = 3.061) and it exhibited best ability for predict- ing prognosis in GSE10927 cohort (P = . 0898, HR = 2.318). This research constructed IPRPs model for predicting adult ACC patients’ prognosis using proteomic data, gene expression data and real-world data and this prognostic model showed stronger pre- dictive value than other biomarkers (Ki-67, Beta-catenin, etc) in multi-cohorts.

KEYWORDS

adult adrenocortical carcinoma, biomarker, predictive model, proteomics, real-world data

Tian, Xu and Anwaier contribute equally to this work.

@ 2021 The Authors. Journal of Cellular and Molecular Medicine published by Foundation for Cellular and Molecular Medicine and John Wiley & Sons Ltd.

1 INTRODUCTION |

Adrenocortical carcinoma (ACC) is a rare and aggressive endo- crine malignancy with high risk of relapse, poor survival and lim- ited treatment options. The Surveillance, Epidemiology and End Results (SEER) database estimates that the annual incidence rate of ACC is approximately 0.72 per million cancer cases, resulting in 0.2% of all cancer deaths in the United States.1 However, ACC shows highly aggressive biological behaviour with less than 35% of patients surviving 5 years after initial diagnosis.2 Therefore, ap- propriate treatment is extremely important. The current preferred treatment of ACC is based on surgical resection of the primary tumour that is usually the first and most effective therapeutic strategy.3-5 Currently, there are very few drugs to treat this dis- ease and mitotane remains the only medication approved by the US Food and Drug Administration for ACC treatment.6 Thus, new treatment options and drug targets are urgently needed, espe- cially for clinical management of patients with ACC who are resis- tant to mitotane.

Proteomics is a powerful tool for detecting unknown protein species, exploring absolute quantified protein abundance, and iden- tifying biomarkers for pathogenic process.7,8 Proteomics has been used widely to explore biomarkers for various diseases9 and the latest developments in proteomics have made it possible to con- duct more comprehensive examinations of protein biomarkers in various cancers.10 For instance, Bouchal et al11 used transcriptome and proteomic analysis to identify potential biomarkers associated with metastatic breast cancer, and several proteomic studies have focused on identifying new diagnostic biomarkers in patients with prostate cancer.12,13 Thus far, very few studies have used a large- scale sequencing proteomic approach to identify potential protein biomarkers for ACC.14

Bioinformatics studies have generated large amounts of complex biological data through combinations of computer science, informa- tion technology and biology. For example, The Cancer Proteome Atlas (TCPA) database provides researchers with reverse-phase protein array (RPPA) data.15 The RPPA technique is a powerful pro- teomic approach for economical, sensitive and high-throughput evaluation of sizable numbers of selected protein markers, which made it possible to explore protein biomarkers using bioinformat- ics.16,17 Because there is a big difference between adult patients and child patients with ACC, in this study, we focused only on adult pa- tients. This study constitutes the first large-scale proteomic analysis combined with transcriptome data to describe the protein landscape of ACC in adult patients.

To explore novel protein biomarkers of potential prognostic value and develop a protein-derived predictive model in adult patients with ACC, we analysed the survival of proteins and constructed an integrated prognosis-related proteins model on risk assessment. Gene expression profiles also were analysed to reveal the under- lying biological interaction networks. The goal of this study was to provide potential novel therapeutic targets and a high performing prognostic predictive model for clinical management of adult ACC.

2 MATERIALS AND METHODS |

2.1 Data downloading and processing

The RPPA data (level 4) of adult ACC were obtained from The Cancer Proteome Atlas (TCPA). The gene expression profiles and clinical in- formation of patients with ACC were downloaded from The Cancer Genome Atlas (TCGA). Preprocessing and normalization of the raw biological data were performed using R software to remove noise and ensure the integrity of the data. By matching the sample IDs, we obtained 46 ACC cases with full data for clinical information, protein abundance and gene expression. We also obtained 76 ACC cases from TCGA (Table 1) and 22 ACC cases from GSE1092718 (Table 2) in NCBI’s GEO database with full data for clinical information and gene expression. All the cases were patients over 18 years old.

2.2 Survival analysis of candidate proteins

Kaplan-Meier analysis was performed based on the median protein abundance value and univariate Cox regression was used to evalu- ate the prognostic value of candidate proteins. For both statistics, P-values < . 05 were considered significant. The volcano plot was obtained using the ggplot2 package in R software.19 Red indicates negative association between protein abundance and survival, green indicates positive association between protein abundance and sur- vival, and black indicates no statistical significance. Survival curves were drawn using the survival package in R software. Red indicates high-risk group, and blue indicates low-risk group.20

2.3 Screening of candidate proteins and construction of a predictive multivariate Cox model

Lasso Cox regression was used to further narrow the proteins with prognostic significance using the glmnet package in R software.21 Multivariate analysis was performed using the Cox proportional haz- ards regression model to identify candidate proteins and evaluate the risk score based on candidate protein abundance and survival rates. An integrated prognosis-related proteins (IPRPs) model was then constructed (Risk score = 2.743 x fibronectin abundance (ref. Low) + 0.781 x FASN abundance (ref. Low) + 1.091 x TFRC abun- dance (ref. Low) + 3.043 x TSC1 abundance (ref. Low)). Median risk score of the predictive IPRPs model was used as the cut-off value and patients were classified into high-risk or low-risk groups.

2.4 | Assessing the prognostic significance of the IPRPs model in TCPA cohort

Besides the risk score of the IPRPs model for the patients with ACC, the covariables for the univariate and multivariate Cox regression models included age, gender, pTstage, pNstage, pMstage and pathologic stage.

TABLE 1 Clinicopathological characteristics 76 adult ACC patients (TCGA cohort)
CharacteristicsEntire cohort (N = 76)
N (%)
Age
< 70 years73(96.1)
≥ 70 years3 (3.9)
Gender
Male30 (39.5)
Female46 (60.5)
Laterality
Left42 (55.3)
Right34 (44.7)
Stage
I-II45 (59.2)
III-IV29 (38.2)
Censored2 (2.6)
T stageª
T1 - T248 (63.2)
T3 - T426 (34.2)
Censored2 (2.6)
N stageª
N066 (86.8)
N18 (10.5)
Censored2 (2.6)
M stageª
M059 (77.6)
M115 (19.7)
Censored2 (2.6)
Mitotic rate
> 5/50 HPF39 (51.3)
≤ 5/50 HPF28 (36.8)
Censored9 (11.8)
Weiss score
≤ 422 (28.9)
> 436 (47.4)
Censored18 (23.7)
Invasion of tumour capsule
Present41 (53.9)
Absent29 (38.2)
Censored6 (7.9)
Necrosis
Present40 (52.6)
Absent32 (42.1)
Censored4 (5.3)

ªTNM scoring system: Tumour size, Lymph Nodes affected, Metastases. AJCC, American Joint Committee on Cancer.

TABLE 2 Clinicopathological characteristics of 22 adult ACC patients (GSE10927 cohort)
CharacteristicsEntire cohort (N = 22)
N (%)
Age
<70 years20 (90.9)
≥70 years2 (9.1)
Gender
Male6 (27.3)
Female16 (72.7)
Laterality
Left10 (45.5)
Right9 (40.9)
Unknown3 (13.6)
Stage
I-II12 (54.5)
III-IV10 (45.5)
TABLE 3 Clinicopathological characteristics of 39 adult ACC patients (Fudan University Shanghai Cancer Center cohort)
CharacteristicsEntire cohort (N = 39)
N (%)
Age
<70 years34 (87.2)
≥70 years5 (12.8)
Gender
Male19 (48.7)
Female20 (51.3)
AJCC stage
I-II14 (35.9)
III-IV25 (64.1)
T Stageª
T1 - T220 (51.3)
T3 - T419 (48.7)
N stageª
N019 (48.7)
N120 (51.3)
M Stageª
M021 (53.8)
M118 (46.2)
Necrosis
Present25 (64.1)
Absent14 (35.9)

ªTNM scoring system: Tumour size, Lymph Nodes affected, Metastases. AJCC, American Joint Committee on Cancer.

FIGURE 1 Survival analysis and screening of proteins. In the volcano plot (A), red and green separately represent high- and low-risk candidate protein biomarkers. 42 proteins with both P-value < . 05 (Kaplan-Meier analysis and univariate Cox regression analysis) were selected and listed in Table 1. The model of Lasso cox regression (B-C)

23

22

22

23

23

23

23

19

18

15

11

9

9

8

B

3

A

TFRC

B

4

8

P27_pT157

FIBRONECTIN

Partial Likelihood Deviance

40

ERK2

CYCLINB1

30

CMET

JAB1

PKCALPHA_pS657

3

PRDX1

TSC1

2

ERALPHA_pS118

NRAS

SNAIL

SMAC

PARP1

TAZ

-log10(pvalue)

PKCDELTA_pS664

1

ARAF_pS299

BCL2A1

Sig

P70S6K_pT389

CD31

XBP1

ECADHERIN

RBM15

CASPASE3

High risk

-8

-7

-6

-5

-4

-3

-2

-1

Log(2)

2

CIAP

EPPK1

CABL

Low risk

23

22

23

21

18

10

7

0

RAB25

FASN

DVL3

BID

· Not

C

X1433BETA

CI

INPP4B

YAP_pS127

RAD50

19

5

P53

PARPCLEAVED

PEA15

JAK2

M

ERALPHA

IGF1R_pY1135Y1136

0

1

Coefficients

A

1

3

2

-100

12

-20

-10

0

10

log2(HR)

-8

-7

-6

-5

-4

-3

-2

-1

Log Lambda

A receiver operating characteristic (ROC) curve was constructed to an- alyse the diagnostic accuracy of the logistic model and the area under curve (AUC) was calculated. Co-abundance analysis was performed using Pearson’s test to identify proteins associated with the logistic model with 0.4 set as the correlation coefficient cut-off value. Survival curves and a scatter diagram were used to explore the correlation be- tween risk score and patient’s prognosis, and a heat map of candidate protein abundance in the high-risk and low-risk groups was drawn.

2.5 | Validation of the IPRPs model in a cohort from the Fudan University Shanghai Cancer Center (FUSCC) in China

Real-world data were collected to validate the prognostic value of the IPRPs model. The cohort included 39 adult patients with ACC (Table 3) from the FUSCC between 2013 and 2019, and tumour specimens were obtained with informed consent. Anti-Ki67 (ab16667, Abcam, USA) anti- fatty acid synthase (ab128870, Abcam, USA), anti-fibronectin (ab2413, Abcam, USA), anti-TSC1 (ab217328, Abcam, USA), and anti-transferrin receptor (ab84036, Abcam, USA) antibodies were used to detect the abundance of the corresponding proteins by immunohistochemistry (IHC). Positive or negative staining of a certain protein in one FFPE slide was independently assessed by two experienced pathologists and

determined as follows. The staining intensity level was graded from 0 to 3. Samples with no staining, weak, median and strong staining denote to the level of 0, 1, 2 and 3. Based on the coverage percentage of immuno- reactive tumour cells (0%, 1-25%, 26-50%, 51-75%, 76-100%), the stain- ing extent was ranging from 0 to 4. The overall IHC score grading from 0 to 12 was evaluated according to the multiply of the staining intensity and extent score. Negative staining represented grade 0 to 3 and posi- tive staining from 4 to 12 for each sample. Risk score of each patient was calculated using the formula generated by the IPRPs model. The Kaplan-Meier method was applied to validate the prognostic value of the model, and the median of the risk score was set as the cut-off value.

2.6 Comparing the IPRPs model with other biomarkers using gene expression data

The number of patients with proteomic data was low, and therefore we used the gene expression data for the prognostic validation. The IPRPs model was compared with other biomarkers in the TCGA co- hort (76 cases) and GSE10927 (22 cases). Survival analyses were carried out using the Kaplan-Meier method and median of gene ex- pression was set as the cut-off value. AUC, C-index and net reclassi- fication improvement (NRI) were calculated to compare IPRPs model with other biomarkers.

TABLE 4 Kaplan-Meier analysis and univariate Cox regression analysis of proteins (both P-value < . 05)
ProteinP value (KM)P value (unicox)HR
P27_pT157.004.0000 (0-0.001)
ERALPHA_pS118.007.0010 (0-0.007)
NRAS.002.0020 (0-0.041)
X1433BETA.018.0220.001 (0-0.34)
CMET.001.0010.003 (0-0.077)
SNAIL.020.0020.009 (0-0.183)
CD31.039.0040.01 (0-0.23)
ARAF_pS299.001.0030.02 (0.002-0.254)
PRDX1.004.0020.021 (0.002-0.238)
XBP1.006.0030.026 (0.002-0.291)
P70S6K_pT389.003.0040.027 (0.002-0.311)
PKCDELTA pS664.023.0030.036 (0.004-0.324)
PARPCLEAVED.044.0340.039 (0.002-0.781)
CIAP.007.0100.055 (0.006-0.498)
P53.012.0250.071 (0.007-0.713)
RAB25.006.0100.075 (0.01-0.542)
JAB1.007.0010.08 (0.018-0.347)
ECADHERIN.006.0040.133 (0.033-0.527)
INPP4B.002.0190.137 (0.026-0.722)
ERALPHA.004.0500.164 (0.027-0.999)
SMAC.003.0020.188 (0.065-0.543)
PKCALPHA pS657.024.0010.281 (0.133-0.594)
EPPK1.027.0110.343 (0.151-0.779)
FASN.039.0151.943 (1.139-3.316)
CYCLINB1.000.0012.198 (1.407-3.434)
PEA15.028.0332.209 (1.066-4.577)
TFRC.001.0002.771 (1.681-4.567)
YAP_pS127.014.0302.984 (1.109-8.031)
PARP1.002.0023.23 (1.553-6.72)
ERK2.000.0005.629 (2.168-14.618)
RBM15.008.0086.531 (1.64-26.011)
FIBRONECTIN.001.0007.007 (2.517-19.511)
DVL3.002.01010.978 (1.754-68.72)
IGF1R_ pY1135Y1136.012.04714.431 (1.033-201.583)
TSC1.001.00116.639 (2.965-93.39)
JAK2.042.04318.739 (1.098-319.658)
BCL2A1.003.00520.533 (2.432-173.352)
CASPASE3.042.00651.697 (3.028-882.701)
CABL.007.00868.124 (2.962-1566.613)
BID.038.01077.659 (2.787-2164.284)

(Continues)

TABLE 4 (Continued)
ProteinP value (KM)P value (unicox)HR
RAD50.005.019158.062 (2.325-10744.279)
TAZ.013.003768.971 (10.143-58295.167)

2.7 Gene set enrichment analysis (GSEA)

To explore potential associated signal pathways, the TCGA data- sets of the high-risk and low-risk groups (according to risk score of the IPRPs model) were analysed using the GSEA software (version 3.0) with the number of permutations set to 1000. False discovery- adjusted P-values were obtained using the Benjamini and Hochberg method.22 Significant differential expression was defined as an ad- justed P-value of < . 01 and a false discovery rate of < 0.25.

The DEGs (adjusted P-value < 0.01; fold change at least 2x) be- tween the high-risk and low-risk groups were identified using the Limma package.23 A heat map was drawn according to the expres- sion matrix of the samples to show the differences in gene expres- sion between the two groups. The Search Tool for the Retrieval of Interacting Genes (STRING; http://string-db.org) (version 10.0)24 online database was used to predict protein-protein interaction (PPI) networks of the DEGs. Cytoscape (version 3.5)25 is an open-source bioinformatics software platform for visualizing molecular interac- tion networks. We used MCODE (version 1.4.2),26 a Cytoscape plug- in, to find the most significant hub genes with MCODE Score ≥ 20. Functional enrichment analysis of the hub genes was completed using the ClusterProfiler package.27

|

3 RESULTS

In this work, we aimed to explore new prognostic biomarkers for adult patients with ACC using proteomics and transcriptomics data. A flow chart of the methods used in this study is given in Figure S1.

3.1 | Selection for candidate proteins with significant prognostic value

From the volcano plot (Figure 1A), 42 candidate protein biomarkers with P-values < 0.05 in both the Kaplan-Meier analysis and univari- ate Cox regression analysis were selected and are listed in Table 4. The Lasso Cox regression results for the selected proteins are shown in Figure 1B, C.

age0.2361.025(0.984-1.067)
gender0.1110.351(0.097-1.271)
stage<0.0013.700(1.833-7.466)
T stage<0.0013.823(1.889-7.736)
M stage0.0016.856(2.152-21.841)
N stage0.1492.639(0.706-9.867)
riskScore0.0061.015(1.004-1.025)
FIGURE 2 Construction of IPRPs model. In univariate Cox regression analysis (A), pathological stage (P < . 001), pTstage (P < . 001), pMstage (P = . 001), risk score of IPRPs model (P < . 01) were associated with shorter OS. However, only risk score (P < . 05) was still significantly correlated with worse outcome in multivariate Cox regression and C-index (0.939, 95% CI:0.916-0.962), NRI (0.235, 95% CI:0-0.597) indicated that our model is stable. (B). This means that our IPRPs model has independent prognostic significance. The red line represents risk score with AUC of 0.933, indicating the diagnostic accuracy and consistent predictive ability (C). Various types of proteins (D) may be associated with the candidate proteins

A

C

pvalue

Hazard ratio

1.0

C-index=0.939

0.8

95%CI:0.916-0.962

True positive rate

NRI =0.235

0.6

95%CI:0-0.597

0.4

0.2

AUC=0.933

P<0.05

0

5

10

15

20

0.0

Hazard ratio

D

0.0

0.2

0.4

0.6

0.8

1.0

B

False positive rate

pvalue

Hazard ratio

FRAN

age

0.295

1.027(0.977-1.080)

0.609(0.135-2.748)

3

gender

0.519

stage

0.216

4.026(0.443-36.579)

PIRRONECTIN

[RALPHA_dit:18

T stage

0.385

1.667(0.526-5.283)

M stage

0.329

0.264(0.018-3.829)

TFRC

MALAS

127_p157

N stage

0.919

1.099(0.177-6.818)

riskScore

0.018

1.009(1.002-1.017)

TSC1

0

5

10

15

20

25

30

35

Hazard ratio

3.2 Construction of the IPRPs model

In the univariate Cox regression analysis (Figure 2A), the pathologi- cal stage (P < . 001), pTstage (P < . 001), pMstage (P = . 001) and risk score of the IPRPs model (P < . 01) were associated with shorter overall survival. However, in the multivariate Cox regression analy- sis, only risk score (P < . 05) was significantly correlated with worse outcome (Figure 2B). C-index (0.939, 95% CI:0.916-0.962) and NRI (0.235, 95% CI:0-0.597) indicated that our model is stable. These results indicate that our IPRPs model has independent prognostic significance. The risk score with AUC of 0.933 indicates the diag- nostic accuracy and consistent predictive ability of our IPRPs model (Figure 2C).

3.3 | Survival analysis of the IPRPs model in the TCPA cohort

Kaplan-Meier survival curves (Figure 3A) revealed that high abun- dances of fatty acid synthase (FASN) (P = . 039), fibronectin (FN) (P < . 001), transferrin receptor (TFRC) (P < . 001) and tuberous scle- rosis 1 (TSC1) (P < . 001) indicated a worse outcome. The formula used to predict overall survival was generated by multivariate Cox

regression models as integrated risk score = 2.743 x FN abundance (ref. Low) + 0.781 x FASN abundance (ref. Low) + 1.091 x TFRC abundance (ref. Low) + 3.043 x TSC1 abundance (ref. Low). The heat map shows that the abundances of FASN, FN, TFRC and TSC1 in the high-risk group were higher than they were in the low-risk group (Figure 3B). The survival time of the high-risk group was significantly shorter than that of the low-risk group (P < . 001), and the increased risk score corresponded to shorter survival (Figure 3C-E).

3.4 | Validation of the prognostic value of the IPRPs model in the FUSCC cohort

Representative IHC plots for the ACC samples are displayed in Figure 4A-E (Abundances of A: Ki-67, B: Fatty acid synthase (FASN), C: Fibronectin (FN), D: Tuberous sclerosis 1 (TSC1) and E: Transferrin receptor (TFRC)). The Ki-67 protein abundance and high-risk (HR) score (Figure 4F, G) were both significantly correlated with worse outcome for patients in the FUSCC cohort (P = . 005, HR = 3.787; P = . 003, HR = 3.947). The IPRPs model predicted the prognosis better than the Ki-67 protein in the FUSCC cohort. A high-risk score was significantly correlated with higher Stage, T stage and N stage (Figure 4H, I).

|

WILEY

FIGURE 3 IPRPs model showed strong ability for predicting prognosis in TCPA cohort. Kaplan-Meier survival curves (A) revealed that higher FASN (P = . 039), FIBRONECTIN (P <. 001), TFRC (P <. 001), TSC1 (P <. 001) expression indicated worse outcome. The formula for predict OS was generated by multivariate Cox regression models: Integrated risk score = 2.743 x FIBRONECTIN expression (ref. Low) + 0.781 x FASN expression (ref. Low) + 1.091 x TFRC expression (ref. Low) + 3.043 x TSC1 expression (ref. Low). It can be seen from the heat map that the expression of FASN, FIBRONECTIN, TFRC, TSC1 in high-risk group is higher than that of low-risk groups (B). The survival time of high-risk group is significantly shorter than the low-risk group (P < . 001) and the increased risk score corresponds to shorter survival (C-E)

A

a

FASIN level

high

low

b

FIBRONECTIN level

high

low

B

type

1.00

1.00

Survival probability

0.75

Survival probability

0.75

FIBRONECTIN

0.50

0.50

type

high

0.25

p=3.8946-02

2.25

P=9.882e-04

low

0.00

0.00

0

1

2

8

9

10

12

FASN

y

+

5

11

A

~

3

m

¥

00

0

1

Time(years)

FASN level

IBRONECTIN level

0

S

S

12

Q

Time(years)

2

53

23

22

15

14

10

23

1

1

1

20

15

1

A

0

0

& 3

23 23

22

16

13

0

4

3

NO

S

1

23

10

5

B

23

2

2

N

0

0

A

3

2

3

4

6

6

7

8

9

10

11

12

0

1

2

3

A

3

6

9

10

11

12

0

Time(years)

Time(years)

C

TFRC level

high

low

d

TSC1 level

hig

low

TFRC

-1

1.00-

1.00-

-2

-3

Survival probability

0.75

Survival probability

0,75

0.50

0.50

TSC1

0.25

p=7.659-04

.25

p=9.532e-04

0.00

0.00

1

TCGA-PA-ASYG

TCGA-OR-A5LH

TCGA-OR-A5JR

TCGA-OR-A5J6

TCGA-OR-A5LK

TCGA-OR-A5JT

TCGA-OR-A5LT

TCGA-PK-A5HA

TCGA-OR-A5JW

TCGA-OR-A5K1

TCGA-OR-A5LP

TCGA-OR-A5LN

TCGA-OR-A5LM

TCGA-OR-A5JZ

TCGA-OR-A5K8

TCGA-OR-A5KU

TCGA-OR-A5J3

TCGA-OR-A5LS

TCGA-OR-A5K3

TCGA-OR-A5JV

TCGA-OR-A5J9

TCGA-PK-A5H8

TCGA-OR-A5K4

TCGA-OR-A5LG

TCGA-OR-A5KO

TCGA-OR-A5KX

TCGA-OR-A5LO

TCGA-OR-A5LL

TCGA-OR-A5KW

0

TCGA-PK-A5H9

TCGA-OR-A5J2

TCGA-OU-A5PI

TCGA-OR-A5JS

TCGA-OR-A5J7

TCGA-OR-A5K6

TCGA-OR-A5JP

TCGA-OR-A5KO

TCGA-OR-A5LD

TCGA-OR-A5JA

TCGA-OR-A5LJ

TCGA-OR-A5KY

TCGA-OR-A5J8

TCGA-OR-A5KZ

TCGA-OR-ASJY

TCGA-OR-A5K5

TCGA-PB-A5OG

4

5

6

1

8

9

10

11

Time(years)

12

0

1

4

S

6

Y

S

11

Time(years)

10

12

TFRC level

TSC1 level

5

23

23

O

6

A

N

e

0

2

15

13

0

0

10

1

9

1

0

1

1

0

1

/

2

?

1

0

0

23

20

K

.

N

14

10

1

1

2

1

1

0

f

2

3

4

E

6

1

8

9

10

11

12

0

1

2

3

4

5

S

B

9

10

Time(years)

11

12

Time(years)

C

Risk

High risk

Low risk

D

1.00

10

J

High risk

8

low Risk

Survival probability

Risk score

0.75

4

*

0.50

~

0

·

0.25

0

10

20

30

40

p=3.085e-06

E

Patients (increasing risk socre)

0.00

Survival time (years)

~

Dead

Alive

0

1

2

3

4

5

6

7

8

9

10

11

12

0

Time(years)

8

0

Risk

High risk

23 23

22

13 22

11

6

Low risk

23

2

1

0

05

18

05

03

o-

01

~

14

14

10

7

0

0

1

2

3

4

5

6

7

8

9

10

11

12

0

10

20

30

40

Time(years)

Patients (increasing risk socre)

3.5 External validation of the IPRPs model and comparison with other biomarkers using gene expression data

In the TCGA cohort (Figure 5A-F), the IPRPs model showed stronger ability for predicting prognosis than the expression levels of CTNNB1 (beta-catenin gene), IGF2 and TP53 (P = . 005, HR = 3.061; P = . 012, HR = 2.768; P = . 162, HR = 0.574; P = . 033, HR = 2.336), whereas the MKI67 (Ki-67 protein gene) and NR5A1 (SF-1 protein gene) ex- pression levels had stronger predictive ability than the IPRPs model (P < . 0001, HR = 9.238; P = . 003, HR = 4.084). In the GSE10927 cohort (Figure 5G-K), which lacked IGF2 expression data, the IPRPs model showed better ability for predicting prognosis than the ex- pression levels of TP53, CTNNB1, NR5A1 and MKI67 (P = . 0898, HR = 2.318; P = . 73, HR = 1.187; P = . 16, HR = 1.983; P = . 36, HR = 1.57; P = . 22, HR = 1.824). AUC, C-index and NRI of various biomarkers were listed in Table 5, and it indicated that IPRPs model may act better than other biomarkers in RPPA data and IHC.

3.6 Significantly involved pathways of the IPRPs

The top 100 genes that were most significant positively and neg- atively correlated with the risk score are depicted in a heat map (Figure 6A). Besides an ACC progressive phenotype, the GSEA in- dicated that significant alteration of the IPRPs model involved chro- mosome separation, metaphase-anaphase transition of the cell cycle and protein modification by small protein removal. Hub genes with prognostic implications associated with the IPRPs were involved mainly in regulation of cell-cycle pathways (Figure 6B-D).

3.7 Identification of DEGs associated with the IPRPs

A significant difference was detected between the gene expression in high-risk and low-risk groups as shown in the heat map (Figure 7A). A PPI network of the DEGs was constructed and the identified hub

|

FIGURE 4 IPRPs model showed a little stronger ability for predicting prognosis than Ki-67 protein in FUSCC cohort. Representative plots of IHC in ACC samples were displayed in (A-E) (A: Ki-67 protein expression, B: FASN protein expression, C: FIBRONECTIN protein expression, D: TSC1 protein expression, E: TFRC protein expression). Both Ki-67 protein expression and high-risk score (F-G) were significantly correlated with worse patients' outcome in FUSCC cohort (P = . 005, HR = 3.787; P = . 003, HR = 3.947). IPRPs model showed a little stronger ability for predicting prognosis than Ki-67 protein in FUSCC cohort. And high-risk score was significantly correlated with higher Stage, T stage and N stage (H-I)

A

Anti-Ki67

B

Anti-FASN

C

Anti-FIBRONECTIN

D

Anti-TSC1

E

Anti-TFRC

F

FUSCC cohort

FUSCC cohort

Low Ki-67 expression

G

Overall survival (%)

Low risk score

100

Overall survival (%)

High Ki-67 expression

100

High risk score

n=19

n=19

50-

50-

n=20

n=20

0

p=0.005, HR(high)=3.7.87.

p=0.003, HR(high)=3.947

Time (years)

0

Time (years)

H

*, ANOVA P=0.02

I

80

80

**

**

ns

Risk score

60.

Risk score

60-

40-

40-

20

20

Stage I

Stage II

Stage III

Stage IV

T1-T2 T3-T4

NO

N1

MO

M1

genes were CENPM, NDC80, DLGAP5, SPC25, CENPF, ZWILCH, AURKB, CENPA, CDC20, CCNA2, KIF4A, BUB1B, CCNB2, UBE2C, AURKA, PLK1, CASC5, RANGAP1, BIRC5, CEP55, NEK2, SGOL2, KIF18A, CCNB1, SKA1, RRM2, ASPM, SGOL1, KIF2C, CDCA8, CENPI, KIF11, BUB1, CDCA5, CDK1, SPC24, SPAG5 and NUF2 (Figure 7B,C). The functional enrich- ment analyses indicated the hub genes were enriched mainly in cell cycle, mitotic nuclear division, chromosome, centromeric region and microtubule binding (Table 6 and Figure 7D, E).

3.8 | Correlation analysis between the IPRPs and other potential signatures

Various types of proteins may be associated with the candidate proteins as shown in Figure 2D. The analysis detected 27 kinds of proteins (correlation coefficients from - 0.64 to 0.66, P < . 001) that were correlated with FN abundance; among them, BID abundance was highly positively correlated (correlation coefficient = 0.66) with FN abundance (Figure S2A). Twenty-one kinds of proteins (correla- tion coefficients from - 0.597 to 0.742) were correlated with TSC1 abundance; among them, PARP1 abundance was highly positively

correlated with TSC1 abundance (correlation coefficient = 0.742) (Figure S2B). Seven kinds of proteins (correlation coefficients from - 0.534 to 0.581) were correlated with FASN abundance; among them, EEF2 abundance was highly positively correlated with FASN abundance (correlation coefficient = 0.0.581) (Figure S2C). Four kinds of proteins (correlation coefficients from - 0.52 to 0.607) were correlated with the TFRC abundance; among them, CYCLINB1 abundance was highly positively correlated with TFRC abundance (correlation coefficient = 0.607) (Figure S2D).

4 DISCUSSION |

The prognosis of ACC is poor because most patients with ACC have locally advanced or metastatic diseases and cannot be treated by surgery. Approximately 66% of patients with localized diseases ex- perience recurrence and usually require systematic treatment.28,29 Although there are diagnostic and prognostic molecular detection methods for ACC, including IGF2, p53, and the Wnt/B-catenin and PI3K signalling pathways, they have not been well applied in mor- phological evaluation, auxiliary diagnosis, or prognostic modelling

|

FIGURE 5 IPRPs model showed stable predictive ability in external validation and comparing with other biomarkers using gene expression data. In TCGA cohort (A-F), IPRPs model showed stronger ability for predicting prognosis than gene expression of CTNNB1 (Beta-Catenin), IGF2 and TP53 (P = . 005, HR = 3.061; P = . 012, HR = 2.768; P = . 162, HR = 0.574; P = . 033, HR = 2.336). While MKI67 (Ki-67 protein) and NR5A1 (SF-1 protein) showed stronger predictive ability (P < . 0001, HR = 9.238; P = . 003, HR = 4.084). In GSE10927 cohort (G- K), apart from lacking IGF2 gene expression, IPRPs model showed better ability for predicting prognosis than TP53, CTNNB1, NR5A1, MKI67 (P =. 0898, HR = 2.318; P = . 73, HR = 1.187; P = . 16, HR = 1.983; P = . 36, HR = 1.57; P = . 22, HR = 1.824)

A

TCGA cohort

B

TCGA cohort

C

TCGA cohort

Overall survival (%)

100-

Low-risk score

Overall survival (%)

- Low IGF2 expression

High-risk score

100

Low CTNNB1 expression

Overall survival (%)

- High CTNNB1 expresion

100

- High IGF2 expresion

n=38

n=38

n=38

50-

50.

50

n=38

n=38

0

p=0.005, HR(high)=3.061

p=0.012, HR(high)=2.768-

p= 0.162. HR(high)=0.574

Time (years)

0

Time (years)

0

Time (years)

D

TCGA cohort

E

TCGA cohort

F

TCGA cohort

Overall survival (%)

100

Low TP53 expression

Overall survival (%)

100

n=38

Low MKI67 expression

n=38

- High TP53 expression

High MKI67 expresion

Overall survival (%)

100

- Low NR5A1(SF-1) expression

n=38

- High NR5A1(SF-1) expresion

50-

n=38

50

50

n=38

n=38.

0

p=0.033, HR(high)=2.336

p< 0.0001, HR(high)=9:238.

p=0.003, HR(high)=4.084

Time (years)

0

Time (years)

0

Time (years)

G

GSE10927

H

GSE10927

I

GSE10927

Overall survival (%)

100-

Low-risk score

Overall survival (%)

100

Low MKI67 expression

Overall survival (%)

- Low NR5A1(SF-1) expression

- High-risk score

High MKI67 expresion

100-

High NR5A1(SF-1) expresion

n=11

n=11

50

n=11

50

50

n=11

n+11

n=11:

0

p=0.0898, HRYhigh

=2:318

Time (years)

0

p=0.22 — HR(high)=1,824

p=0.36; HR(high)=j.57

Time (years)

0

Time (years)

J

GSE10927

K

GSE10927

Overall survival (%)

100

Low TP53 expression

Overall survival (%)

High TP53 expression

100-

Low CTNNB1 expression

High CTNNB1 expresion

n=11:

n=11

50

50

n=11.

n=11

0

p=0.73. HAihigh)=1.187

Time (years)

0

p=0:16 :- HR/high):1,983

Time (years)

TABLE 5 Evaluating biomarkers of ACC in multiple cohorts
Biomarker evaluationAUCC-indexNRI
CohortsTCGA (RPPA)TCGA (RNAseq)GSE10927 (RNA seq)FUSCC IHC)TCGA (RPPA)TCGA (RNAseq)GSE10927 (RNAseq)FUSCC (IHC)TCGA (RPPA)TCGA (RNAseq)GSE10927FUSCC
(RNAseq)(IHC)
CTNNB10.40.7040.732-0.5640.6590.657-ns-0.020.06-
IGF2-0.313---0.575---ns--
TP530.4890.870.571-0.6760.6590.551--0.02nsns-
MKI67-0.8620.7860.621-0.8530.6710.694-0.37nsns
SF-1-0.71---0.665----0.07ns-
Risk Score0.9330.8850.7050.6490.9390.7890.6570.72----

of ACC.30 Early diagnosis and appropriate treatment play key roles in the management of ACC; thus, effective biomarkers are urgently needed.31 Proteomics has unique advantages and our study is the first large-scale proteomic analysis of ACC with RPPA data. We found that FASN, FN, TFRC and TSC1 abundance levels were of high prognostic value. To explore the underlying biological mecha- nism, we performed a GSEA of high-risk and low-risk groups and the results indicated that the most significant pathways associated with candidate proteins included chromosome separation, metaphase- anaphase transition of cell cycle and protein modification by small protein removal. These pathways are worth further study.

FASN is a key enzyme in mammals that is needed for ab initio palmitic acid synthesis. In most normal non-adipose tissues, the abundance and activity of FASN are largely inhibited by adequate di- etary fat, but in many human cancers, FASN abundance and activity are abnormally increased and are associated with poor prognosis.32 The increased abundance of FASN potentially confers tumour cells an advantage in survival and growth.33 For instance, Ueda et al34 found that FASN expression promoted cell survival and growth of tumour cells in gestational trophoblastic neoplasms, and Nguyen et al35 found that increased intratumoral FASN expression led to more aggressive prostate cancers. In this study, we found that a high

A

B

FIGURE 6 IPRPs model was strongly correlated with chromosome separation, metaphase-anaphase transition of cell cycle. Top 100 most significant genes positively and negatively correlate with risk score were depicted in heat map (A). Besides ACC progressive phenotype, GSEA indicated that significant alteration of IPRPs model involved in chromosome separation, metaphase-anaphase transition of cell cycle and protein modification by small protein removal. Hub genes with prognostic implications associated with IPRPs model mostly involved in regulation of cell-cycle pathways. (B-D)

TCGA-P6-A50G-014-11-A39K-20

TCGA-OR-A5K5-014-21-A39K-20

TCGA-OR-A5JY-01A-21-A39K-20

TCGA-OR-A5K7-01A-21-A39K-20

TCGA-OR-A5J8-014-21-A39K-20

TCGA-OR-A5KY-01A-21-A39K-20

TCGA-OR-A5LJ-01A-21-A39K-20

TCGA-OR-A5JA-01A-21-A39K-20

TCGA-OR-ASLD-01A-21-A39K-20

ICGA-OR-A5KO-01A-21-A39K-20

TCGA-OR-A5JP-01A-21-A39K-20

TCGA-OR-A5K6-01A-21-A39K-20

TCGA-OR-A5J7-01A-21-A39K-20

TCGA-OR-A5JS-01A-21-A39K-20

TCGA-OT-A5PT-01A-21-A39K-20

TCGA-OR-A5J2-01A-21-A39K-20

TCGA-PK-A5H9-01A-21-A39K-20

TCGA-OR-A5KT-01A-21-A39K-20

TCGA-OR-A5LL=01A-21-A39K-20

TCGA-OR-A5L0-01A-21-A39K-20

TCGA-OR-A5KY-01A-21-A39K-20

TCGA-OR-A5K0-014-21-A39K-20

TCGA-OR-ASLG-01A-21-A39K-20

TCGA-OR-A5K4-01A-21-A39K-20

TCGA-PK-A5M8-01A-21-A39K-20

TCGA-OR-A5J9-01A-21-A39K-20

TCGA-OR-A5TH-01A-21-A39K-20

TCGA-OR-A5K3-01A-21-A39K-20

ICGA-OR-ASLS-01A-21-A39K-20

ICGA-OR-A5J3-01A-21-A39K-20

ICGA-OR-A5KUI-01A-21-A39K-20

TCGA-OR-A5K8-01A-21-A39K-20

TCGA-OR-A5J7-01A-21-A39K-20

TCGA-OR-ASIM-01A-21-A39K-20

TCGA-OR-ASLY-01A-21-A39K-20

TCGA-OR-A5LP-01A-21-A39K-20

TCGA-OR-A5K1-01A-21-A39K-20

TCGA-OR-A5JT-01A-21-A39K-20

TCGA-PK-A5HA-01A-21-A39K-20

TCGA-OR-A5LT-01A-21-A39K-20

TCGA-OR-A5JT-014-21-A39K-20

TCGA-OR-A5LK-01A-21-A39K-20

TCGA-OR-A516-01A-41-A39K-20

TCGA-OR-A5 TR-01A-21-A39K-20

TCGA-OR-ASLH-01A-21-A39K-20

TCGA-PA-ASYG-014-21-439K-20

Enrichment plot: GO_CHROMOSOME_SEPARATION

0.6

Enrichment score (ES)

0.5

0.4

0.3

0.2

0.1

SampleName

0.0

HAIISS

KATMBL1

SE3B4

STX5

Ranked list metric (Signal2Noise)

AGAP1

JARID2

1.0

‘h’ (positively correlated)

R3HDM1

TYTS

0.5

TTS1

SAP130

0.0

LIM9

Zero cross at 13137

CHOT9

0.5

ZMF143

ADR74

-1.0

‘T (negatively correlated)

MARK3

RPLA

0

2,500

5,000

7,500

10,000

12,500

15,000

17,500

20,000

7-Mar

Rank in Ordered Dataset

TLKAP

SRPRA

Enrichment profile - Hits

Ranking metric scores

IS¥1

COPS7B

TAPS

C

MAFE

STK25

Enrichment plot: GO_METAPHASE_ANAPHASE_TRANSITION_OF_CELL_C YCLE

SPTAM1

CKAP2L

RIT1

ZIF 496

AMMECR1L

SIIV39H2

0.7

CCHIB2

SPC25

Enrichment score (ES)

0.6

AKTRTU2

0.5

PHESA

DDx50

0.4

HIST1H4T

0.3

HTRA2

KCTD20

0.2

MITFK

0.1

REM1 7

SIIRPD3

0.0

TET20

SART1

PRPF40A

Ranked list metric (Signal2Noise)

KUSTRIL

CCDC9OR

MICAPH

1.0

‘h’ (positively correlated)

CASC10

CIDSPL2

0.5

PPP2BSC

0.0

TIMES

Zero cross at 13137

GRTA3

0.5

TOUR

-1.0

TMEM232

‘I’ (negatively correlated)

CISA

0

2,500

5,000

7,500

10,000

12,500

15,000

17,500

20,000

PALM

SSBP2

Rank in Ordered Dataset

CYSTM1

EMPP4

Enrichment profile - Hits

Ranking metric scores

ARAT

AMIKEF1

SPATA4

D

CASC1

GLIM2

Enrichment plot: GO_PROTEIN_MODIFICATION_BY_SMALL_PROTEIN_RE MOVAL

PAKS

AMIE3858

CYP27A1

COXZR

TTC25

PRKG1

0.5

STAC

SERPTIT1

Enrichment score (ES)

0.4

SMTM1012B

PIFO

0.3

GIGZ

PRLR

EMPPS

0.2

ISCIL

TIGD4

0.1

CPME4

HS6ST3

0.0

PTPRA

DMAT1

TSC2204

Ranked list metric (Signal2Noise)

ATFM1

TAK3

PI3K

1.0

‘h’ (positively correlated)

AMY2R

MSRA

0.5

SYCP3

0.0

SPEF1

Zero cross at 13137

MAGED2

0.5

LACTE

-1.0

ECHTH2

‘T (negatively correlated)

ICEP2L

0

2,500

5,000

7,500

10,000

12,500

15,000

17,500

20,000

APSS1

IMT12

Rank in Ordered Dataset

SPAG17

ZDHHC17

Enrichment profile - Hits

Ranking metric scores

PYROXD1

A

FIGURE 7 Identification of differentially expressed genes (DEGs) associated with IPRPs model. As shown in the heat map, there is a significant difference between the gene expression of high-risk and low-risk groups (A). Protein-protein interaction network of DEGs was constructed and the selected hub genes are CENPM, NDC80, DLGAP5, SPC25, CENPF, ZWILCH, AURKB, CENPA, CDC20, CCNA2, KIF4A, BUB1B, CCNB2, UBE2C, AURKA, PLK1, CASC5, RANGAP1, BIRC5, CEP55, NEK2, SGOL2, KIF18A, CCNB1, SKA1, RRM2, ASPM, SGOL1, KIF2C, CDCA8, CENPI, KIF11, BUB1, CDCA5, CDK1, SPC24, SPAG5 and NUF2 (B-C). Hub genes are mostly enriched in cell cycle, mitotic nuclear division, chromosome, centromeric region and microtubule binding (D-E)

Type

15

Type

high

B

C

10

low

5

0

CDCA5

-5

CENPI

SPC25

SGOL2

RANGAP1

CENPM

CENPA

SKA1

ZWILCH

BUB1

NUF2

BUB1B

CASC5

CDCA8

NDC80

SPC24

BIRC5

PLK1

SGOL1

CCNA2

CCNB1

CENPF

CDC20

CDK1

KIF18A

CEP55

KIF2C

KIF4A

CCNB2

AURKB

RRM2

AURKA

DLGAP5

KIF11

ASPM

SPAG5

UBE2C

NEK2

mitotic nuclear division

TCGA-OR-ASK4-01A-21-A39K-20

TCGA-PK-ASH8-01A-21-A39K-20

TCGA-OR-A519-01A-21-A39K-20

TCGA-OR-A5JV-01A-21-A39K-20

BAJOR-A5K3-01A-21-A39K-20

TCGA-OR-ASLS-01A-21-A39K-20

TCGA-OR-A5J3-01A-21-A39K-20

TCGA-OR-A5KB-01A-21-A39K-20

TOGA-OH NABOK-20

TCGA-OR-A5IZ-01A-21-A39K-20

PRIM-018-21-A19K-20 TCGA-OR-ASLN-01A-21-A39K-20

TCGA-OR-ASLP-01A-21-A39K-20

TCGA-OR-A5JW-01A-21-A39K-20

MORAM04-014-21-A39K-20

TCGA-PK-A5HA-01A-21-A39K-20

ePAOP-ALT-01A-21-A39K-20

TOGA-OR-ASJT-01A-21-A38K-20

TCGA-OR-A5LK-01A-21-A39K-20

TCGA-OR-A5J5-01A-41-A39K-20

TCGA-OR-A5JR-01A-21-A39K-20

TCGA-OR-A5LH-01A-21-A39K-20

TCGA-PA-ASYG-01A-21-A39K-20

TCGA-P6-ASOG-01A-11-A39K-20

NON 005-01A-21-A39K-20

TCGA-OR-A5JY-01A-21-A39K-20

TCGA-OR-A5KZ-01A-21-A39K-20

TOGA-OR-ASKY-01A-21-A39K-20

TCGA-OR-ASLJ-01A-21-A39K-20

TOGA-OR-A5LD-01A-21-A39K-20

TCGA-OR-45JA-014-21-A39K-20

TCGA-OR-A5KO-01A-21-A39K-20

TCGA-OR-ASJP-01A-21-A39K-20

TCGA-OR-ASK6-01A-21-A39K-20

TCGA-OR-A5J7-01A-21-A39K-20

TCGA-OR-A5J8-01A-21-A39K-20

TCGA-OU-A5PI-01A-21-A39K-20

TCGA-OR-A5.12-01A-21-A39K-20

TCGA-PK-ASH9-01A-21-A39K-20

TOGA-OR-A5KW-01A-21-A39K-20

TOUR OR POLY-04-21-A39K-20 TCGA-OR-A5LO-01A-21-A39K-20

TCGA-OR-A5KX-01A-21-A39K-20

TCGA-OR-ASLG-01A-21-A39K-20

TOGA-OR-A5K0-01A-21-A3BK-20

E

nuclear division

organelle fission-

mitotic sister chromatid segregation

chromosome segregation-

sister chromatid segregation

4

nuclear chromosome segregation-

regulation of chromosome segregation-

regulation of mitotic nuclear division-

D

spindle organization

Cell cycle

chromosome, centromeric region

Progesterone-mediated oocyte maturation

kinetochore

Oocyte meiosis

chromosomal region

p.adjust

p53 signaling pathway

Cellular senescence

condensed chromosome, centromeric region-

Human T-cell leukemia virus 1 infection

condensed chromosome

0.0025

FoxO signaling pathway

condensed chromosome kinetochore

8

0.0050

Viral carcinogenesis

Human immunodeficiency virus 1 infection-

spindle

p.adjust

condensed nuclear chromosome, centromeric region

0.0075

Ubiquitin mediated proteolysis-

Hepatitis B

Apoptosis - multiple species

midbody

Glutathione metabolism-

0.1

condensed chromosome outer kinetochore

Pyrimidine metabolism-

Acute myeloid leukemia

Platinum drug resistance

0.2

microtubule binding

histone kinase activity

Drug metabolism - other enzymes

Colorectal cancer

0.3

tubulin binding-

Gap junction

protein serine/threonine kinase activity

AMPK signaling pathway

microtubule motor activity

Purine metabolism

Apoptosis

cyclin-dependent protein serine/threonine kinase regulator activity-

Hippo signaling pathway

motor activity

RNA transport

Transcriptional misregulation in cancer-

microtubule plus-end binding

Epstein-Barr virus infection

ATP-dependent microtubule motor activity, plus-end-directed-

0

2

4

6

8

cyclin-dependent protein kinase activity

0

5

10

15

20

25

abundance of FASN also was significantly correlated with worse prognosis of ACC. Previous studies have established the anti-tumour effects of the first-generation FASN inhibitors.36,37 Thus, FASN may be a potential therapeutic target in ACC.

FN is a large extracellular matrix protein in bones, which can combine with itself and collagen to form a network.38 Studies have shown that the abundance of FN in breast cancer is higher than in normal tissues and FN abundance is significantly related to the in- vasiveness of the disease.39 Knowles et al40 found that FN matrix formation was associated with kidney tumour cell spreading. Besides the prognostic value of FN in ACC, we also found that the abun- dances of JAB1, SCD1 and PRDX1 were negatively correlated with FN abundance, whereas the abundances of HEREGULIN, TIGAR and BID were positively correlated with FN abundance. Thus, FN is also a candidate target for new therapeutic drugs.

Iron is a basic trace element involved in cell metabolism, division and proliferation, and iron also has been considered as an important factor in the development of cancer.41 TFRC is a cell surface receptor that is responsible for transferrin-mediated iron uptake; thus, TFRC may play a key role in the energy supply of cancer cells.42 Shpyleva

et al43 found a high abundance of TFRC in breast cancer, and TFRC antibodies have been used to inhibit tumour growth. 44 We found mu- tual inhibition between TFRC and SMAC, and that the abundances of X1433ZETA, ERK2 and CYCLINB1 were positively correlated with TFRC abundance. Modulation of PPIs is a promising new idea in drug development45,46; thus, the design of TFRC inhibitors based on the interaction modes may create new therapeutic drugs.

TSC1, in a complex with tuberous sclerosis 2, inhibits the nutrient- mediated or growth factor-stimulated phosphorylation of S6K1 and EIF4EBP1 by negative regulation of mTORC1 signal transduction.47,48 We also found interactions between various types of proteins and TSC1. Among them, PARP1 abundance showed the highest correla- tion with TSC1 (correlation coefficient = 0.742) and it attracted our attention because of its key role in DNA repair.49 Maintaining the integrity of the genome is the basis of cell survival, and PARP inhibi- tors kill tumours mainly by inhibiting DNA repair and destroying the genomes of tumour cells.50 Inhibiting PARP1 also may inhibit TSC1, suggesting a potential strategy for the treatment of ACC.

Moon et al also established a model for predicting prognosis using RPPA data. They focused on the patients with distant metastasis. But

TABLE 6 GO and KEGG pathways enrichment analysis of hub genes
TermDescriptionCount in gene setP value
GO:0 140 014Mitotic nuclear division211.26E-30
GO:0 000 280Nuclear division221.81E-28
GO:0 048 285Organelle fission221.61E-27
GO:0 000 775Chromosome, centromeric region245.20E-41
GO:0 000 776Kinetochore211.48E-37
GO:0 098 687Chromosomal region251.04E-36
GO:0 008 017Microtubule binding81.65E-08
GO:0 035 173Histone kinase activity42.65E-08
GO:0 015 631Tubulin binding81.83E-07
hsa04110Cell cycle87.26E-12
hsa04914Progesterone- mediated oocyte maturation71.13E-10
hsa04114Oocyte meiosis77.02E-10

Note: Abbreviations: DEGs, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

the C-index of their model (maximum:0.86) is much lower than ours (0.939). Guo J et al identified 9 hub genes (CCNB1, CDK1, TOP2A, CCNA2, CDKN3, MAD2L1, RACGAP1, BUB1 and CCNB2) with prog- nostic value. But the data they used were different from our research. They only focused on gene expression data, which is usually consid- ered unstable than protein data. And they just identified 9 hub genes with prognostic value without any further validation. In our study, we used protein data to establish a model for predicting prognosis and validate its value successfully in multiple cohorts. The main strength of this study lies in the first attempt to explore the prognostic role of protein biomarkers based on quantitative proteomic analysis of ACC in adult patients. An IPRPs model was constructed with AUC values equal to 0.933 and our results show that it distinguished itself from previous prognostic predictive models of ACC.

This study had several limitations. The validation of IRPPs model in the transcriptome dataset may lead some bias as the model derived from the proteomic data. The nature of retrospective research limits the clinical value of this work. Further validation cohorts in multicentre or prospective studies are needed to verify the findings. And the more advanced ACC patients in FUSCC cohort may lead to unbalanced baseline. However, it is difficult to conduct randomized controlled tri- als for ACC because of the rarity of these tumours. There is also an urgent need for in vitro and in vivo experiments to explore potential effective functions of IPRPs and reveal the underlying mechanisms.

5 CONCLUSION

We constructed an IPRPs model for predicting the prognosis of adult patients with ACC using proteomic data, gene expression data and

real-world data. The prognostic model showed a stronger predic- tive value for prognosis than other biomarkers (eg Ki-67 and beta- catenin) in multi-cohorts. Our results distinguished FASN, FN, TFRC and TSC1 from previously identified tumour promoters and revealed novel prediction model IPRPs that outperformed the currently es- tablished prognostic parameters for anticipating disease course and better clinical management of adult ACC.

ACKNOWLEDGEMENTS

We thank the TCPA, TCGA and GEO database for providing RPPA data and gene expression profiles of ACC.

CONFLICT OF INTERESTS

The authors declare no competing interests.

AUTHORS’ CONTRIBUTIONS

All authors: Work present carry out in collaboration. YDW, ZHL and QYY: Define the theme of the study and discussed analysis, inter- pretation and presentation. TX and XWH: Manuscript draft; data analysis; development of the algorithm; and explanation of the results. Aihetaimujiang, WHK and WFN: Participation in the collection of rele- vant data and manuscript draft. CDL and LWJ: Help to perform the sta- tistical analysis. SGH: Help to revise the manuscript and provide guiding suggestions. All the authors: Read and approval of the final manuscript.

ETHICAL APPROVAL

The Ethics approval and consent to participate of the current study was approved and consented by the ethics committee of Fudan University Shanghai Cancer center.

DATA AVAILABILITY STATEMENT

The datasets analysed during the current study available from the corresponding author on reasonable request.

ORCID Xi Tian İD https://orcid.org/0000-0003-1965-0647

Wen-Jie Luo İD

https://orcid.org/0000-0003-1133-9870

REFERENCES

1. Kebebew E, Reiff E, Duh Q-Y, et al. Extent of disease at presen- tation and outcome for adrenocortical carcinoma: have we made progress? World J Surg. 2006;30(5):872-878.

2. Else T, Williams AR, Sabolch A, et al. Adjuvant therapies and pa- tient and tumor characteristics associated with survival of adult patients with adrenocortical carcinoma. J Clin Endocrinol Metab. 2014;99(2):455-461.

3. Icard P, Goudet P, Charpenay C, et al. Adrenocortical carcinomas: surgical trends and results of a 253-patient series from the French Association of Endocrine Surgeons study group. World J Surg. 2001;25(7):891-897.

4. Bilimoria KY, Shen WT, Elaraj D, et al. Adrenocortical carcinoma in the United States: treatment utilization and prognostic factors. Cancer. 2008;113(11):3130-3136.

5. Wängberg B, Khorram-Manesh A, Jansson S, et al. The long-term sur- vival in adrenocortical carcinoma with active surgical management and use of monitored mitotane. Endocr Relat Cancer. 2010;17(1):265-272.

|

6. Schteingart DE, Doherty GM, Gauger PG, et al. Management of patients with adrenal cancer: recommendations of an international consensus conference. Endocr Relat Cancer. 2005;12(3):667-680.

7. Rahim MAA, Rahim ZHA, Ahmad WAW, et al. Can saliva proteins be used to predict the onset of acute myocardial infarction among high-risk patients? Int J Med Sci. 2015;12(4):329-335.

8. Li X, Li S, Lu M, et al. Proteomic profiling of iron overload-induced human hepatic cells reveals activation of TLR2-mediated inflamma- tory response. Molecules (Basel, Switzerland). 2016;21(3):322.

9. WuL, Sun Y, Wan J, et al. A proteomic analysis identifies candidate early biomarkers to predict ovarian hyperstimulation syndrome in polycys- tic ovarian syndrome patients. Mol Med Rep. 2017;16(1):272-280.

10. Borrebaeck CAK. Precision diagnostics: moving towards protein biomarker signatures of clinical utility in cancer. Nat Rev Cancer. 2017;17(3):199-204.

11. Bouchal P, Dvořáková M, Roumeliotis T, et al. Combined pro- teomics and transcriptomics identifies carboxypeptidase B1 and nuclear factor KB (NF-KB) associated proteins as putative biomark- ers of metastasis in low grade breast cancer. Mol Cell Proteomics. 2015;14(7):1814-1830.

12. Geisler C, Gaisa NT, Pfister D, et al. Identification and validation of potential new biomarkers for prostate cancer diagnosis and prog- nosis using 2D-DIGE and MS. Biomed Res Int. 2015;2015:454256.

13. Khan AP, Poisson LM, Bhat VB, et al. Quantitative proteomic profil- ing of prostate cancer reveals a role for miR-128 in prostate cancer. Mol Cell Proteomics. 2010;9(2):298-312.

14. Kim HM, Lee YK, Koo JS. Proteome analysis of adrenal cortical tu- mors. Expert Review of Proteomics. 2016;13(8):747-755.

15. Li J, Akbani R, Zhao W, et al. Explore, visualize, and analyze func- tional cancer proteomic data using the cancer proteome atlas. Can Res. 2017;77(21):e51-e54.

16. Nishizuka S, Charboneau L, Young L, et al. Proteomic profiling of the NCI-60 cancer cell lines using new high-density reverse-phase lysate microarrays. Proc Natl Acad Sci USA. 2003;100(24):14229-14234.

17. Tibes R, Qiu Y, Lu Y, et al. Reverse phase protein array: validation of a novel proteomic technology and utility for analysis of primary leukemia specimens and hematopoietic stem cells. Mol Cancer Ther. 2006;5(10):2512-2521.

18. Giordano TJ, Kuick R, Else T, et al. Molecular classification and prog- nostication of adrenocortical tumors by transcriptome profiling. Clin Cancer Res. 2009;15(2):668-676.

19. Wickham H. Ggplot2: Elegant Graphics for Data Analysis. Incorporated: Springer Publishing Company; 2009.

20. Lin H, Zelterman DJT. Modeling survival data: extending the cox model. Technometrics. 2000;44(1):85-86.

21. Friedman J, Hastie T, Tibshirani R. Regularization paths for general- ized linear models via coordinate descent. J Stat Softw. 2010;33(1).

22. Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrich- ment analysis: a knowledge-based approach for interpret- ing genome-wide expression profiles. Proc Natl Acad Sci USA. 2005;102(43):15545-15550.

23. Ritchie ME, Phipson B, Wu D, et al. limma powers differential ex- pression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47.

24. Franceschini A, Szklarczyk D, Frankild S, et al. STRING v9.1: protein- protein interaction networks, with increased coverage and integra- tion. Nucleic Acids Res. 2013;41(Database issue):D808-D815.

25. Smoot ME, Ono K, Ruscheinski J, et al. Cytoscape 2.8: new fea- tures for data integration and network visualization. Bioinformatics (Oxford, England). 2011;27(3):431-432.

26. Bandettini WP, Kellman P, Mancini C, et al. MultiContrast Delayed Enhancement (MCODE) improves detection of subendocardial myocardial infarction by late gadolinium enhancement cardiovas- cular magnetic resonance: a clinical validation study. J Cardiovasc Magn Reson. 2012;14:83.

27. Yu G, Wang L-G, Han Y, et al. clusterProfiler: an R package for compar- ing biological themes among gene clusters. Omics. 2012;16(5):284-287.

28. Miller BS, Ammori JB, Gauger PG, et al. Laparoscopic resection is inappropriate in patients with known or suspected adrenocortical carcinoma. World J Surg. 2010;34(6):1380-1385.

29. Leboulleux S, Deandreis D, Al Ghuzlan A, et al. Adrenocortical car- cinoma: is the surgical approach a risk factor of peritoneal carcino- matosis? Eur J Endocrinol. 2010;162(6):1147-1153.

30. Mete O, Gucer H, Kefeli M, et al. Diagnostic and prognos- tic biomarkers of adrenal cortical carcinoma. Am J Surg Pathol. 2018;42(2):201-213.

31. Lacombe AMF, Soares IC, Mariani BMP, et al. Sterol O-Acyl transfer- ase 1 as a prognostic marker of adrenocortical carcinoma. Cancers. 2020;12(1):247.

32. Liu H, Liu J-Y, Wu X, et al. Biochemistry, molecular biology, and phar- macology of fatty acid synthase, an emerging therapeutic target and diagnosis/prognosis marker. Int J Biochem Mol Biol. 2010;1(1):69-89.

33. Albiges L, Hakimi AA, Xie W, et al. Body mass index and metastatic renal cell carcinoma: clinical and biological correlations. J Clin Oncol. 2016;34(30):3655-3663.

34. Ueda SM, Mao T-L, Kuhajda FP, et al. Trophoblastic neoplasms ex- press fatty acid synthase, which may be a therapeutic target via its inhibitor C93. Am J Pathol. 2009;175(6):2618-2624.

35. Nguyen PL, Ma J, Chavarro JE, et al. Fatty acid synthase polymor- phisms, tumor expression, body mass index, prostate cancer risk, and survival. J Clin Oncol. 2010;28(25):3958-3964.

36. Kridel SJ, Axelrod F, Rozenkrantz N, et al. Orlistat is a novel in- hibitor of fatty acid synthase with antitumor activity. Can Res. 2004;64(6):2070-2075.

37. Knowles LM, Axelrod F, Browne CD, et al. A fatty acid synthase blockade induces tumor cell-cycle arrest by down-regulating Skp2. J Biol Chem. 2004;279(29):30540-30545.

38. Sottile J, Hocking DC. Fibronectin polymerization regulates the composition and stability of extracellular matrix fibrils and cell- matrix adhesions. Mol Biol Cell. 2002;13(10):3546-3559.

39. Park J, Schwarzbauer JE. Mammary epithelial cell interactions with fibronectin stimulate epithelial-mesenchymal transition. Oncogene. 2014;33(13):1649-1657.

40. Knowles LM, Gurski LA, Maranchie JK, et al. Fibronectin matrix formation is a prerequisite for colonization of kidney tumor cells in fibrin. J Cancer. 2015;6(2):98-104.

41. Jones NR, Ashmore JH, Lee SY, et al. Association studies of HFE C282Y and H63D variants with oral cancer risk and iron homeosta- sis among whites and blacks. Cancers. 2015;7(4):2386-2396.

42. Richardson DR, Baker E. The effect of desferrioxamine and ferric ammonium citrate on the uptake of iron by the membrane iron- binding component of human melanoma cells. Biochem Biophys Acta. 1992;1103(2):275-280.

43. Shpyleva SI, Tryndyak VP, Kovalchuk O, et al. Role of ferritin al- terations in human breast cancer cells. Breast Cancer Res Treat. 2011;126(1):63-71.

44. Brooks D, Taylor C, Dos Santos B, et al. Phase Ia trial of murine im- munoglobulin A antitransferrin receptor antibody 42/6. Clin Cancer Res. 1995;1(11):1259-1265.

45. Lu S, Jang H, Zhang J, et al. Inhibitors of Ras-SOS interactions. ChemMedChem. 2016;11(8):814-821.

46. Jiang H, Deng R, Yang X, et al. Peptidomimetic inhibitors of APC- Asef interaction block colorectal cancer migration. Nat Chem Biol. 2017;13(9):994-1001.

47. Tee AR, Fingar DC, Manning BD, et al. Tuberous sclerosis complex-1 and -2 gene products function together to inhibit mammalian tar- get of rapamycin (mTOR)-mediated downstream signaling. Proc Natl Acad Sci USA. 2002;99(21):13571-13576.

48. LimJS, Gopalappa R, Kim SH, et al. Somatic mutations in TSC1 and TSC2 cause focal cortical dysplasia. Am J Hum Genet. 2017;100(3):454-472.

49. Ahel I, Ahel D, Matsusaka T, et al. Poly(ADP-ribose)-binding zinc finger motifs in DNA repair/checkpoint proteins. Nature. 2008;451(7174):81-85.

50. Fong PC, Yap TA, Boss DS, et al. Poly(ADP)-ribose polymerase inhibition: frequent durable responses in BRCA carrier ovar- ian cancer correlating with platinum-free interval. J Clin Oncol. 2010;28(15):2512-2519.

How to cite this article: Tian X, Xu W-H, Anwaier A, et al. Construction of a robust prognostic model for adult adrenocortical carcinoma: Results from bioinformatics and real-world data. J Cell Mol Med. 2021;25:3898-3911. https:// doi.org/10.1111/jcmm.16323

SUPPORTING INFORMATION

Additional supporting information may be found online in the Supporting Information section.