Discover

Research

Jincheng Ma1 . Yan Jin1 . Baocheng Gong1 . Long Li1,2 . Qiang Zhao1

Received: 26 January 2022 / Accepted: 3 March 2022

Published online: 21 March 2022

@ The Author(s) 2022 OPEN

Abstract

Background Necroptosis is a novel programmed cell death mode independent on caspase. A number of studies have revealed that the induction of necroptosis could act as an alternative therapeutic strategy for drug-resistant tumors as well as affect tumor immune microenvironment.

Methods Gene expression profiles and clinical data were downloaded from XENA-UCSC (including The Cancer Genome Atlas and Genotype-Tissue Expression), Gene Expression Omnibus, International Cancer Genome Consortium and Chi- nese Glioma Genome Atlas. We used non-negative matrix factorization method to conduct tumor classification. The least absolute shrinkage and selection operator regression was applied to establish risk models, whose prognostic effective- ness was examined in both training and testing sets with Kaplan-Meier analysis, time-dependent receiver operating characteristic curves as well as uni- and multi-variate survival analysis. Principal Component Analysis, t-distributed Sto- chastic Neighbor Embedding and Uniform Manifold Approximation and Projection were conducted to check the risk group distribution. Gene Set Enrichment Analyses, immune infiltration analysis based on CIBERSORT, EPIC, MCPcounter, ssGSEA and ESTIMATE, gene mutation and drug sensitivity between the risk groups were also taken into consideration. Results There were eight types of cancer with at least ten differentially expressed necroptosis-related genes which could influence patients’ prognosis, namely, adrenocortical carcinoma (ACC), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), acute myeloid leukemia (LAML), brain lower grade glioma (LGG), pancreatic adenocarcinoma (PAAD), liver hepatocellular carcinoma (LIHC), skin cutaneous melanoma (SKCM) and thymoma (THYM). Patients could be divided into different clusters with distinct overall survival in all cancers above except for LIHC. The risk models could efficiently predict prognosis of ACC, LAML, LGG, LIHC, SKCM and THYM patients. LGG patients from high-risk group had a higher infiltration level of M2 macrophages and cancer-associated fibroblasts. There were more CD8+ T cells, Th1 cells and M1 macrophages in low-risk SKCM patients’ tumor microenvironment. Gene mutation status and drug sensitivity are also different between low- and high-risk groups in the six cancers.

Conclusions Necroptosis-related genes can predict clinical outcomes of ACC, LAML, LGG, LIHC, SKCM and THYM patients and help to distinguish immune infiltration status for LGG and SKCM.

Qiang Zhao and Long Li contributed equally to this work

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s12672-022- 00477-2.

☒ Qiang Zhao, zhaoqiang@tjmuch.com | 1Tianjin Key Laboratory of Cancer Prevention and Therapy, Department of Pediatric Oncology, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Tianjin’s Clinical Research Center for Cancer, Tianjin, China. 2Key Laboratory of Immune Microenvironment and Diseases of Educational Ministry of China, Department of Immunology, The Province and Ministry Co-sponsored Collaborative Innovation Center for Medical Epigenetics, Tianjin Medical University, Tianjin, China.

Check for updates

Discover Oncology

| https://doi.org/10.1007/s12672-022-00477-2

Keywords Cancer . Necroptosis . Prognosis . Risk score . Tumor immune infiltration

Abbreviations

TCGA	The Cancer Genome Atlas
GTEx	Genotype-Tissue Expression
GEO	Gene Express Omnibus
ICGC	International Cancer Genome Consortium
CGGA	Chinese Glioma Genome Atlas
ACC	Adrenocortical carcinoma
BLCA	Bladder urothelial carcinoma
BRCA	Breast invasive carcinoma
CESC	Cervical squamous cell carcinoma and endocervical adenocarcinoma
CHOL	Cholangiocarcinoma
COAD	Colon adenocarcinoma
DLBC	Lymphoid neoplasm diffuse large B-cell lymphoma
DLBC	Lymphoid neoplasm diffuse large B-cell lymphoma
ESCA	Esophageal carcinoma
GBM	Glioblastoma multiforme
HNSC	Head and neck squamous cell carcinoma
KICH	Kidney chromophobe
KIRC	Kidney renal clear cell carcinoma
KIRP	Kidney renal papillary cell carcinoma
LAML	Acute myeloid leukemia
LGG	Brain lower grade glioma
LIHC	Liver hepatocellular carcinoma
LUAD	Lung adenocarcinoma
LUSC	Lung squamous cell carcinoma
MESO	Mesothelioma
OV	Ovarian serous cystadenocarcinoma
PAAD	Pancreatic adenocarcinoma
PCPG	Pheochromocytoma and paraganglioma
PRAD	Prostate adenocarcinoma
READ	Rectum adenocarcinoma
SARC	Sarcoma
SKCM	Skin cutaneous melanoma
STAD	Stomach adenocarcinoma
TGCT	Testicular germ cell tumors
THCA	Thyroid carcinoma
THYM	Thymoma
UCEC	Uterine corpus endometrial carcinoma
UCS	Uterine carcinosarcoma
UVM	Uveal melanoma
FADD	Fas-associated protein with death domain
TNF-a	Tumor necrosis factor a
NCCD	Nomenclature Committee on Cell Death
TNFR1	Tumor necrosis factor receptor 1
TRADD	TNF receptor 1-associated death domain protein
TRAF2	Tumor necrosis factor and receptor related factor 2
RIPK1	Receptor-interacting protein kinase 1
CIAP1/2	Cellular inhibitors of apoptosis 1 and 2
LUBAC	Linear ubiquitin Chain assembly complex
TGF-ß	Transforming growth factor-beta
TAK1/TAB	TGF-ß activated kinase 1/TGF-ß activated kinase 1 binding protein

Discover Oncology

CYLD Cylindromatosis

RIPK3 Receptor-interacting protein kinase 3

MLKL Mixed lineage kinase domain-like

DAMPs

Damage associated molecular patterns Tumor-associated macrophage

TAM

CXCL1

C-X-C motif chemokine ligand 1 Kyoto Encyclopedia of Genes and Genomes Differentially expressed necroptosis-related genes

KEGG DENGS

NMF Non-negative matrix factorization

Overall survival

DSS

Disease specific survival

PFS Progression free survival

DFS

Disease free survival

LASSO

Least absolute shrinkage and selection operator

ROC

Receiver operating characteristic

PCA

Principal Component Analysis

t-SNE T-distributed Stochastic Neighbor Embedding

UMAP Uniform Manifold Approximation and Projection

GSEA Gene Set Enrichment Analyses

TMB

Tumor mutational burden

MSI

Microsatellite instability

MHC

Major Histocompatibility Complex

FDA Food and Drug Administration

GO CAFs

Gene Ontology

Cancer-associated fibroblasts

Treg

Regulatory T

Tfh

Follicular helper T

TP53

Tumor protein p53

IDH1

Isocitrate dehydrogenase (NADP(+)) 1

CIC

Capicua transcriptional repressor

FUBP1 SMARCA4 SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4 AT-rich interaction domain 1A

ARID1A TTN

Titin

EGFR

Epidermal growth factor receptor

NF1

Neurofibromin 1

PTEN Phosphatase and tensin homolog

RYR2

Ryanodine receptor 2

MUC16

Mucin 16, cell surface associated

ANK3

Ankyrin 3

PKHD1L1

PKHD1 like 1

GTF2I

General transcription factor IIi

HRAS

HRas proto-oncogene, GTPase

CTLA4

Cytotoxic T-lymphocyte-associated protein 4

PD-1

Programmed cell death protein 1

LAG-3

Lymphocyte-activation gene 3

CA125

Carbohydrate antigen 125

Far upstream element binding protein 1

1 Introduction

Although recent authoritative statistics showed that the death rate of cancer declined over the past 30 years, cancer remains one of the primary causes of death worldwide no matter in developed or developing countries, which greatly increases economic burden and seriously affects life quality [1]. The occurrence and development of tumor involves a series of extremely complex biological processes, and the treatment effect of many tumors is still not satisfactory even under the combination of multiple therapies. It is urgent and of great importance to find novel insights and effective agents for cancer.

The resistance to cell death has been identified as one of the most important characters of malignant tumors [2]. Clas- sical theory divided cell death forms into apoptosis and necrosis, according to the whether it’s under the programmed regulation of genetic materials [3]. However, in the 1990s, a new pattern of necrosis-like cell death featured by non-cas- pase dependency gradually emerged. Researchers found that, under the inhibition of key proteins in apoptosis pathway [such as Caspase-8 or Fas-associated protein with death domain (FADD)] and the stimulation of tumor necrosis factor a (TNF-a), the cell morphology was similar to the necrotic cell [4, 5]. Then, at the beginning of the twenty-first century, the concept and process of programmed necrosis or necroptosis was gradually proposed and elaborated [6-8]. In 2018, the Nomenclature Committee on Cell Death (NCCD) officially defined this special form of cell death as necroptosis [9]. Unlike apoptosis which involves kinds of morphological changes, such as cell shrinkage and detachment from the surrounding cells, nucleoplasm concentration, fragmentation of nuclear membrane and nucleolus as well as the appearance of apop- totic bodies, several special biological events occur in cells undergoing necroptosis, including the damage of membranes, disorder of metabolism and the extravasation of inflammatory substances [8]. Necroptosis and apoptosis share the same initiating stage. When tumor necrosis factor receptor 1 (TNFR1) on the cell membrane surface is activated by TNF-a, TNF receptor 1-associated death domain protein (TRADD) and tumor necrosis factor and receptor related factor 2 (TRAF2) will be recruited by its death domain at C-terminal. Subsequently, TRADD and TRAF2 separately recognizes and binds to receptor-interacting protein kinase 1 (RIPK1) and cellular inhibitors of apoptosis 1 and 2 (CIAP1/2), and protein complex scaffold is formed by linear ubiquitin Chain assembly complex (LUBAC). Then, with the combination of these molecules and transforming growth factor-beta (TGF-ß) activated kinase 1/TGF-ß activated kinase 1 binding protein (TAK1/TAB) complex, the supramolecular structure (TNFR1 Complex I) come into being [10]. The deubiquitination of RIPK1 by the cylindromatosis (CYLD) can result in the cleavage of Complex I and the dissociation of RIPK1 as well as TRADD, where different endings of the cell happen. Complex IIa constituted of TRADD, FADD as well as Caspase-8 and Complex IIb composed of PIPK1, receptor-interacting protein kinase 3 (RIPK3), FADD and Caspase-8 would lead cell to apoptosis. The catalytic activity inhibition of caspase-8 would allow RIPK1 to phosphorylate RIPK3, which recruits mixed lineage kinase domain-like (MLKL) to form necroptosome [11, 12]. MLKL migrates to cell membrane to result in necroptosis.

Necroptosis played an indispensable role in the maintenance of internal environment homeostasis and the progression of several inflammation-related diseases, such as neurodegenerative disease, ischemia-reperfusion injury and pathogen infection [10, 13]. A number of studies have also revealed the significance of necroptosis induction at cancer treatment in recent years, which especially worked for the apoptosis-resistant tumors [14]. Meanwhile, with the rise of immuno- therapy, the relationship between different forms of cell death and tumor immunity has gradually attracted extensive attention [15]. There was no effective anti-tumor immune response observed in the tumor area where apoptosis or necrosis occurred. However, increasing number of studies have revealed the influence of necroptosis on tumor immune microenvironment, where the results were opposite in different tumor models. Damage associated molecular patterns (DAMPs) and various cytokines and chemokines which leaked out of necroptotic cells of colon carcinoma and melanoma could strengthen cytotoxic function of CD8+ T cells and the activity of antigen-presenting cells [16-18]. However, the necroptotic cells of pancreatic ductal adenocarcinoma enhanced the immunosuppressive function of tumor-associated macrophage (TAM) by C-X-C motif chemokine ligand 1 (CXCL1) and Mincle signaling [19]. The studies also showed that the synergistic effect of necroptosis-promoting agents and immune checkpoint inhibitors (ICIs) could trigger long-term tumor-suppression effect in mouse models [17, 18], indicating that the necroptosis induction of tumor cell was probably an effective complement to immunotherapy.

In this study, we comprehensively analyzed the necroptosis-related genes in different kinds of cancers based on data from The Cancer Genome Atlas (TCGA), Genotype-Tissue Expression (GTEx), Gene Express Omnibus (GEO), International Cancer Genome Consortium (ICGC) and Chinese Glioma Genome Atlas (CGGA). We developed novel tumor classification and constructed risk models based on necroptosis-related genes to predict patients’ clinical outcomes. Immune infiltra- tion, gene mutation and drug sensitivity were also taken into consideration.

2 Methods

2.1 Gene expression and clinical data collection

We obtained gene profiles, clinical features and survival information of 33 TCGA cancers from XENA-UCSC (https://xena. ucsc.edu/). For thirteen types of cancer with no or very limited number of corresponding normal tissue samples (< 10), we obtained gene expression data of normal samples from GTEx at XENA-UCSC, namely, adrenocortical carcinoma (ACC), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), glioblastoma multiforme (GBM), acute myeloid leukemia (LAML), brain lower grade glioma (LGG), ovarian serous cystadenocarcinoma (OV), pancreatic adenocarcinoma (PAAD), rectum adenocarcinoma (READ), skin cutaneous melanoma (SKCM), testicular germ cell tumors (TGCT), thymoma (THYM) and uterine carcinosarcoma (UCS). Because of no relevant samples for pheochromocytoma and paraganglioma (PCPG) and sarcoma (SARC) found in GTEx, we only used TCGA data for the analysis. Mesothelioma (MESO) and uveal melanoma (UVM) were excluded from this study, for there were no normal samples in neither TCGA nor GTEx. Necroptosis-related gene list (hsa04217) was found in Kyoto Encyclopedia of Genes and Genomes (KEGG). The details of necroptosis-related genes were shown in Supplementary file 1.

The other cohorts with patients’ clinical and survival information were obtained for ACC, CESC, LAML, LGG, liver hepa- tocellular carcinoma (LIHC), PAAD, SKCM from GEO, ICGC and CGGA. The details are as listed:

ACC: GSE19750 [20] https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE19750.

GSE33371 [21] https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE33371.

CESC: GSE44001 [22] https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE44001.

LAML: GSE37642 [23] https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE37642.

LGG: CGGA_693, CGGA_325 [24] http://www.cgga.org.cn/.

LIHC: ICGC (LIRI-JP) https://icgc.org/.

PAAD: ICGC (PACA-AU) https://icgc.org/.

SKCM: GSE65904 [25] https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE65904.

To identify DENGs between tumors and the corresponding normal tissues, the “limma” R package was applied, with |log2 (fold change)| > 1 and false discovery rate (FDR) <0.05 as the thresholds. Then, we conducted survival analysis of DENGs in each particular type of cancer. The cancer types with at least 10 DENGs that significantly influence patients’ overall survival (OS) were selected. Next, we constructed chord diagrams of the prognostic DENGs in the chosen cancers by using “circlize” and “corrplot” R packages, where Pearson correlation analysis was performed. The correlation at protein level was visualized by STRING (Version: 11.5, https://cn.string-db.org/) through “Multiple protein” module with the “Homo sapiens” and “low confidence (0.150)” as the main parameters. Finally, based on prognostic DENGs, we used the non-negative matrix factorization (NMF) to conducted cancer classification. “NMF” R package was used, with “brunet”, “10 iterations” and “clusters k ranks from 2 to 10” as the main parameters. Kaplan-Meier analysis was performed between patients’ survival and the different clusters, where four survival endpoints were taken into consideration, namely, OS, disease specific survival (DSS), progression free survival (PFS) and disease free survival (DFS).

2.3 Construction and validation of DENGs-based risk model

First, batch corrections were performed between TCGA cohorts and the corresponding additional cohorts of the selected cancers by “sva” R package. Then TCGA and additional cohorts were appointed as the training sets and testing sets sepa- rately. For each cancer the training set was used to establish necroptosis-related risk model by the least absolute shrink- age and selection operator (LASSO) regression, employing “glmnet”R package, with fivefold cross-validation applied to optimize the model. Patients were classified into low- and high-risk groups according to the median risk score of training set. Kaplan-Meier analysis of OS and the risk groups was conducted. To assess the predictive efficiency of the risk model,

Fig. 1 Identification of differentially expressed necroptosis-related genes (DENGs) and the investigation of their prognostic effect. Top 8 can- cers with largest number of prognostic DENGs were chosen. The heat maps and forest plots showed the expression state and the prog- nostic effect of DENGs in adrenocortical carcinoma (ACC) (a), cervical squamous cell carcinoma endocervical adenocarcinoma (CESC) (b), acute myeloid leukemia (LAML) (c), brain lower grade glioma (LGG) (d), liver hepatocellular carcinoma (LIHC) (e), pancreatic adenocarcinoma (PAAD) (f), skin cutaneous melanoma (SKCM) (g) and thymoma (THYM) (h). |log2 (fold change)|> 1 and false discovery rate (FDR) <0.05 were used as the screening criteria. Logrank p value and hazard ratio were presented beside each forest plot

ACC

CESC

Type

MAPKID

SOSTMI

pvalue

PLAZG45

PPIA

Hazard ratio

CHUPA

TRAF2

0.003

2.420(1.347-4.349)

STZH2AAJ

FTL

0.017

1.511(1.075-2.123)

pvalue

Hazard ratio

IST2H2AA4 -4

0.422(0.287-0.620)

TRAPS

PLA2G4C

-0.001

STATED

TNF

0.003

1.431(1.128-1.816)

FICAMZ

NUR/-PLA2048

CHMP4B

0.001

0.296(0.142-0.614)

SLC25A5

0.030

0.712(0.524-0.967)

PLA2048

HMGB1

0.018

3.756(1.257-11.225)

JAK3

<0.001

0.604(0.453-0.807)

TRAPE

VDAC1

0.030

1.675(1.050-2.673)

FTH1

CHMP3

STAT3

0.018

0.500(0.281-0.890)

0.026

1.370(1.039-1,807)

STATSA

0.004

0.546(0.363-0.823)

TICAMA

IL1B

<0.001

1.386(1.185-1.621)

IRF9

0.013

1.949(1.150-3.303)

ISTIHZAE

SQSTM1

-0.001

0.446(0.289-0.688)

CHMP4C

0.005

1.848(1.201-2.842)

HSP90AB1 0.003

0.387(0.207-0.723)

IL1A

0.007

1.189(1.048-1.350)

PPIA

<0.001

8.731(3.374-22.596)

H2AFX

<0.001

3.057(2.052-4.554)

BCL2

0.007

0.558(0.365-0.853)

HSPA0AB1

H2AFY

0.016

4.563(1.333-15.617)

HIST1H2AE 0.050

0.809(0.654-1.000)

CHAMPAB

H2AFZ

0.018 1.747(1.100-2.774)

HIST1H2AI 0.028

0.679(0.480-0.960)

IFNAR1

0.0

0.5 1.0

1.5

2.0

2.5

Hazard ratio

LAML

LGG

Type

pvalue

Hazard ratio

Type

TNF

0.024

0.815(0.683-0.973)

-PLAZG45

Hazard ratio

MAPK10

CFLAR

-PLA2048

CYBB

40.001

5.993(3.529-10.178)

pvalue

1.197(1.038-1.380)

FADD

GLUD1

0.013

40.001

0.452(0.375-0.544)

<0.001

1.997(1.323-3.014)

1.418(1.077-1.866)

CH2444

PYGB

ZHZAAS

MAPK10

<0.001

0.607(0.454-0.813)

STATER

RIPK3

0.013

00.001

0.379(0.292-0.491)

APK18

SLC25A4

00.001

6.129(2.294-16.379)

HZAFY

FTL

PLA2G4A

<0.001

1.471(1.269-1.705)

GLUD2

0.002

<0.001 1.736(1.393-2.164)

3.761(1.629-8.681)

CHMPLA

PLA2G4C

0.009

0.687(0.519-0.910)

FTH1

0.012

1.367(1.071-1.744)

CHMP4A

00.001

0.253(0.125-0.514)

1.560(1.277-1.905)

0.587(0.348-0.988)

<0.001

CHMP6

CHMP1B

0.045

PLA2G4A

IL33

0.021

0.460(0.238-0.890)

CASP1

0.005

1.287(1.077-1.536)

0.023

0.823(0.696-0.974)

TICANH

CHMP2A

0.032

1.690(1.047-2.728)

IFNAR1

0.004

2.091(1.267-3.450)

TYK2

0.004

1.826(1.214-2.747)

CHMP4B

0.011

1.982(1.166-3.370)

STAT1

<0.001

1.798(1.530-2.113)

IFNGR1

0.018

1.326(1.050-1.676)

IRF9

0.006

1.537(1.133-2.085)

2.121(1.450-3.105)

JAK3

0.047

0.795(0.634-0.997)

EIF2AK2

<0.001

SQSTM1

0.019

1.682(1.089-2.599)

STAT6

0.017

1.749(1.104-2.771)

CHMPIR

HSP90AB1 PPIA

<0.001

0.430(0.281-0.657)

AIFM1

<0.001

3.270(1.743-6.134)

PLAZGIA

0.008

2.146(1.224-3.763)

HISTSH2A

H2AFX

BCL2

0.034

0.766(0.599-0.979)

HIST3HZA

<0.001

00.001

2.220(1.749-2.817)

0.571(0.492-0.662)

H2AFY2

0.005

1.208(1.060-1.377)

H2AFY2

<0.001

0.513(0.446-0.591)

H2AFY

2.844(1.561-5.181)

H2AFJ

0.009

1.322(1.072-1.631)

<0.001

EF2AK2

H2AFJ

<0.001

1.750(1.340-2.285)

Hazard ratio

LIHC

PAAD

Type

IL33

pvalue

Hazard ratio

pvalue

Hazard ratio

SHARPIN

TRAF2

0.003

1.458(1.136-1.871)

BIRC3

<0.001

1.379(1.148-1.656)

PPIA

RBCK1

0.024

1.319(1.038-1.676)

IOT-PLA2048

CFLAR

0.025

1.686(1.067-2.663)

TRAF2

SHARPIN

0.035

1.270(1.017-1.586)

VDAC1

0.025

1.551(1.058-2.275)

PYGL

<0.001

1.633(1.261-2.116)

H2AFX

PYGB

0.006

1.283(1.074-1.533)

CHOPLA

PYGB

0.006

1.314(1.080-1.600)

H2AFZ

IL33

0.031

0.798(0.649-0.980)

PYCARD

0.047

1.289(1.003-1.656)

RBCK1

USP21

00.001

1.657(1.236-2.221)

CHMP3

0.001

2.914(1.514-5.607)

BAX

SQSTM1

<0.001

1.377(1.172-1.617)

TNFSF10

<0.001

1.737(1.376-2.192)

SQSTM1

HSP90AB1 0.004

1.429(1.124-1.816)

TNFRSF10A0.001

1.632(1.266-2.103)

HSP90AB1

PARP1

0.040

1.294(1.012-1.654)

FAS

0.005

1.597(1.151-2.215)

BAX

0.044

1.258(1.006-1.574)

IFNAR1

0.034

2.038(1.056-3.933)

USP21

TYK2

<0.001

0.358(0.214-0.598)

PARP1

PPIA

<0.001

1.807(1.333-2.449)

STAT1

<0.001

1.516(1.208-1.903)

PYGB

H2AFX

<0.001

1.372(1.156-1.628)

EIF2AK2

0.001

2.194(1.357-3.548)

CAPN2

H2AFZ

<0.001

1.507(1.237-1.837)

PPIA

0.005

2.146(1.256-3.665)

GLUL

0.0

0.5

1.0

1.5

2.0

HIST1H2AC0.002

1.470(1.155-1.871)

HIST1H2AE

Hazard ratio

0 1

Hazard ratio

SKCM

THYM

Type

pvalue

Hazard ratio

CFLAR

<0.001

0.621(0.468-0.824)

Type

pvalue

Hazard ratio

CYBB

<0.001

0.817(0,750-0.891)

TNF

0.002

3.972(1.653-9.547)

SLC25A5

VDAC1

0.028

CFLAR

0.006

4.875(1.580-15.039)

GLUL

<0.001

1.190(1.019-1.389)

0.037

1.507(1,195-1.900)

0.869(0.762-0.992)

CYBB

0.010

1.865(1.164-2.988)

PLA2G4E

<0.001

1.364(1.189-1.564)

SLC25A4

0.048

1.658(1.005-2.733)

PLA2G4D

1.396(1.200-1.624)

PYGM

0.026

0.001(0.000-0,418)

PLA2G4F

<0.001

1.336(1.160-1.539)

FTH1

0.020

3.989(1.241-12.826)

CAPN1

PGAMS

0.029

CHMP4C

<0.001

1.313(1.029-1.677)

0.033

1.578(1.212-2.055)

NLRP3

0.034 6.058(1.141-32.168)

1.193(1.015-1,402)

CASP1

<0.001 4.914(1.992-12.123)

TNFSF10

<0.001

0.822(0.739-0.913)

IL1B

0.049

2.387(1.003-5.681)

FASLG

<0.001

0.720(0.622-0.833)

CHMP7

0.017

0.264(0.089-0.786)

IFNAR1

<0.001

0.635(0,486-0.831)

TNFSF10

<0.001

2.019(1.338-3.046)

IFNGR2

JAK3

0.006

0.002

0.759(0.624-0.924)

IFNG

<0.001

3.848(1.765-8.388)

STAT1

0.828(0.733-0.934)

0.770(0.698-0.849)

IFNAR2

0.008

0.142(0.033-0.606)

IRF9

<0.001

IFNGR2

EIF2AK2

0.005

0.648(0.512-0.819)

<0.001 3.569(1.686-7.558)

0.721(0.573-0.907)

JAK2

<0.001 3.396(1.640-7.032)

HSP90AB

PARP1

0.043

1.217(1.006-1,471)

1.319(1.051-1.655)

STAT4

0.040

2.343(1.041-5.274)

AIFM1

0.017

0.043

1.296(1.009-1.664)

TLR4

0.004

3.629(1.493-8.821)

HIST3H2A

0.011

1.163(1.035-1.308)

ZBP1

0.008

2.149(1.222-3.779)

H2AFJ

0.007

H2AFX

0.013

0.597(0.398-0.895)

H2AFZ

1.149(1.038-1.273)

0.003

1.230(1.072-1.410)

H2AFY2

0.014

2.586(1.212-5.520)

0.0

0.5

1.0

1.5

2.0

Hazard ratio

time-dependent receiver operating characteristic (ROC) curves of 1, 3, 5-years were made using “survivalROC”R package. Uni- and multi-variate survival analyses were employed to examine whether the risk score could independently affect patients’ prognosis. Model genes expression heat maps were constructed with the increase of risk score by “pheatmap” R package, and some clinical factors between patients from low- and high-risk groups were also compared by the use of Fisher’s exact test. Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE) and Uniform

Fig. 2 Correlation among prognostic DENGs. Chord diagrams and protein protein interaction networks (a-h) showed the correlation among the prognostic DENGs at mRNA and protein level in the eight cancers. The width and color of the lines between genes in chord diagrams represents the Pearson correlation coefficients and the sources of the protein-interactions were denoted with lines of distinct colors

ACC

CESC

LAML

LGG

JAKCS

AL 18

PLA2G4A

FIL

HMG81

STAT3

FTMY

FTHI

24A CHMP6 CHAMPID

CASPY

PYGB

MAPK10

CHMP4D

PLA2G4C

STATSA

VDASY

GLUOZ

GLUOT

CHMP4C

SLC25A4

CHMPZA

-1

SLCZSAS

RIPK3

CHMP44B

CFLAR

SOSTMI

FNAR1

FADD

TRAFZ

TYK2

SV06dSH

HAFJ

HRAFJ

INGR$

H2AF2

H2AFY2

H12AFY

PPA

8CL2

SAKS

HAFY2

PPIA HSP90ABTSOSTMI EIFZARE

HEARY

MIST1H2AI

8C12

WST3HZA

H2AFX

HIST 1H2AE

AIFM1

STATS

H2AFX

varDerk

RF9

HMGB1

H2AFJ

SLC2544

AKO

H2AFZ

L30

CFLAR

HZAFY

STATS

STATSA

BCL2

PPIA

FTL

FTH1

HISTIHZA

HISTIHZAE

BCL2

AIFM1

H2AFY2

JAK3

VDACI

TNF

JAK3

CHMP2A

CHMP4B

UNNAT

STATS

H2AFX

HSPOZAB1

IFNGR1

FTH1

TNF

CHMP48

PLAZO4C

IL18

ILSA

PLAZGAA

SPOCAO

TRAP2

SLC25A5

STATO

RIPICI

CHMP4C

VIG

HIST3H34

CASPI

QLUDZ

DYBB

EF240

PYOB

QUOI

SQSTMÍ

FADO

PPIA

OWX

PLAZGACI

FR.

CARA

QAY2

HAPY

LIHC

PAAD

SKCM

THYM

IL33

PYCARD

NLRP3

USP21

PYGB

W TERAMO PLAZCA CAPNIPGAMICHIMPAC

CHMP3

SHARPIN

SLC25A4 PYGM

FTHI

PYGL

CASPI

SOSTMI

GLUL

VOAC!

INFSF 10

TNFSF10

IL18

HISPAMAR1

CHMPY

-1

RBCKT

CFLAR

INFRSFICA

-1

FASLG

-1

IFNAR1

CFLAR

TNFSF10

CFLAR

TRAFZ

PARP1

BIRC3

TNF

IFNG

”

PRASY

H2AFYZ

ZEYNE

BAK

HESTIHZAC

FNART

IFNGRZ

H2AFZ

HZAFJ HIST3HZA AIFMT

STATE

TYAZ

H2AFX

PLA

H2AFX

PerA

EIF2AK2

STATI

JAKO

PAARPTHSP90AB1 EWF2AK2

RF9

2BP1

TLA4

STAT4

IFNAR2

ZBP1

CHMP4C

IFNGR2

EIF2AK2

TNFRSF10A

RBCK1

CFLAR

IFNG

NURPS

PYGB

SHARPIN

PYCARD

PLAZO4E

PLA2G4F

VDACI

TNFSF10

H2AFY2

FAS

CFLAR

GLUL

CASP1

SQSTM1

TRAF2

USP21

HSPSOAB

STAT4

TNFSF10

JAK

IFNART

VDAC1

PLA2040

CAPNI

INFBESO

SLC25AS

ILIB

CHMP7

BAX

BIRC3

PPIA

IL33

STAT1

YBB

POANS

JAK2

CFLAR

HSP90AB1

AK2

LRA

TYK2

HISTIH2AC

IFNAR

AIFM1

PARRI

SLC25A4

CHMP3

PARP1

H2AFZ

PYGL

PPIA

IFNGR2

FASLG

HIST3H2A

CYBB

FTH1

H2AFX

PYGB

IRPP

STATI

H2AFZ

H2AFJ

H2AFX

PYGM

Known Interactions

Predicted Interactions

Others

from curated databases

gene neighborhood

textmining

experimentally determined

gene fusions

co-expression

gene co-occurrence

protein homology

Manifold Approximation and Projection (UMAP) were carried out to verify the risk-group assignments according to the model genes expression data, where “stats”, “Rtsne” and “umap”R packages were used. Distribution of patients’ risk score and survival state was also analyzed. The same procedures were performed in the testing sets.

2.4 Gene set enrichment analyses (GSEA)

In both training and testing sets, GSEA was conducted between low- and high-risk groups by “limma”, “org.Hs.eg.db”, “clusterProfiler”, “DOSE” and “enrichplot” R packages, with “kegg.v7.4.symbols” and “go.v7.4.symbols” downloaded from the MSigDB database. |Normalized enrichment score (NES)|> 1.5 and adjusted p-value < 0.05 were used as the screening criteria.

2.5 Investigation of tumor immune microenvironment

Five algorithms were applied to assess immune infiltration status of each patient in both training and testing sets, namely, CIBERSORT, EPIC, MCPcounter, ssGSEA and ESTIMATE. Then, the immune infiltration level was compared between patients from low- and high-risk groups with Wilcoxon signed-rank test. The Spearman’s correlation analysis of risk score and immune score, stromal score as well as ESTIMATE score was also conducted. Then, we compared tumor mutational burden (TMB) and microsatellite instability (MSI) between the patients from the two risk groups with Wilcoxon signed-rank test, and investigated the relationship of risk score and TMB as well as MSI using Spearman’s correlation analysis. In addition, we explored whether there existed a correlation of risk score and immune related genes expression with Pearson cor- relation analysis, including immunoinhibitor genes, immunostimulator genes, Major Histocompatibility Complex (MHC) genes, chemokine genes and chemokine receptor genes. The corresponding genes were acquired from TISIDB (http:// cis.hku.hk/TISIDB/index.php).

2.6 Analysis of gene mutation

Somatic mutation data based on “VarScan2” software was acquired for TCGA samples. Then, we made oncoplots to show the mutation status of the top 20 most frequently mutated genes in low- and high-risk groups, with “maftools” R package. The mutation rate of the top 20 genes was compared by Fisher’s exact test.

2.7 Drug sensitivity analysis

We downloaded the gene expression and z-score matrix from CellMiner (https://discover.nci.nih.gov/cellminer/home. do) and calculated the risk score of each sample according to the genes and corresponding coefficient of the different cancers’ risk model. Then, we investigated whether there existed any correlation of risk score and the sensitivity of Food and Drug Administration (FDA)-approved drugs with Pearson correlation analysis.

3 Results

3.1 Identification of prognostic DENGs in TCGA-cancers

As shown in Fig. 1, there were eight types of cancer with at least ten prognostic DENGs, namely, ACC, CESC, LAML, LGG, LIHC, PAAD, SKCM and THYM. The situation of other cancers was shown in Fig. S1, and no prognostic DENGs was found in colon adenocarcinoma (COAD) (d), stomach adenocarcinoma (STAD) (t), thyroid carcinoma (THCA) (u) and uterine corpus endometrial carcinoma (UCEC) (v). Notably, there were no DENGs observed in SARC. We also revealed the correlation between the prognostic DENGs in the eight cancers at both transcription and translation level (Fig. 2).

3.2 Tumor classification

We used NMF to classify cancer patients into different subgroups according to the expression profiles of the prog- nostic DENGs. NMF rank survey with multiple parameters and the consensus matrix heat maps were displayed at K value from 2 to 10 for ACC, CESC, LAML, LGG, LIHC, PAAD, SKCM and THYM (Fig. S2). The optimal K value was chosen for each cancer and the corresponding classification was shown (Fig. 3a, c, e, g, i, k, m, o). Notably, there existed sig- nificant difference of OS among the subgroups in all cancers except for LIHC (Fig. 3b, d, f, h, j, l, n, p).

3.3 LASSO regression risk models

The LASSO coefficient spectrum of the selected necroptosis-related genes for ACC, CESC, LAML, LGG, LIHC, PAAD, SKCM and THYM were shown in Figs. 4a, g, m, s and 5a, g, m, s. Figures 4b, h, n, t and 5b, h, n, t showed the fivefold cross-vali- dation. The risk score calculation formulas of the eight cancers were shown in Supplementary file 2. In ACC, LAML, LGG, LIHC and SKCM, low-risk patients had obviously better OS compared with patients from high-risk group (Figs. 4c, o, u, 5c, o), and the time-dependent ROC curves of 1, 3 and 5 years in training and testing sets revealed the good efficiency

Discover Oncology

(2022) 13:17

| https://doi.org/10.1007/s12672-022-00477-2

Research

ACC

DSS

PFS

DFS

1.00

Consensus matrix

1.00-

1.00

0.75

Chefer

Overall survival

0.25

P<0.001

p<0.001

p=0.022

Disease free survival

Progression free survival

Disease specific survival

0.00

a.DE

Time(years)

Number at risk

₾

218

€

Clus/bit

Cluster

Time(years)

CESC

Time(years)

Consensus mavia

1.00

Cluster

1.00

.TS

3 75

0.75

p=0.002

P=0.201

Overall survival

p=0.081

Disease-free survival

P=0.148

Progression free survival

Disease specific sureval

6.2

0.00

01234

10 11 12 13 14 15 98 BF 18 19 20

9 10 11 12 13 14 95 16 17 18 99 20

8 9 10 11 12 13 14 95 06 17 18 99 20

0 1 2 3 4 5 6

8 9 10 11 12 13 14 15 16 17 18 19 20

Time(years)

Number at risk

Cluster

00-0

Cluster

02001

20000

@ 10

Time(yeara)

18 19 20

Time(years)

14 95

Time(years)

10 20

LAML

Time(years)

2 00 11 1

194 16 16 1

Consensus matrix

1.00-

0.75

₸

0.25

Overall survival

p=0.002

Timetyears)

Number at risk

Cluster

LGG

Time(years)

1.00.

1.00

1.00/

Consensus matrix

0.75

₪

Overall survival

0.25

p<0.001

0.26

p<0.001

P=0.012

Disease free survival

Progression free survival

Disease specific survival

p<0.001

0.00

a.DE

0 1

1 1

$ 10 19 12 13 14 15 50 87 10 19 20

3 4 5

17 10 19 20

Time(years)

1 2

៛

Time(years)

Number at risk

10263 30 12 7 3 3 2 1 0 0 0 0 0 0 0 0 0 0 0 0

99 60 29 11 7 3 3 2 1 0 0 0 0 0 0 0 0 0 0 0 0

02 44

40432720814381 61 48 38 23 19 16 12 8 5 4 2 1 1 0 0 0

396022203138 77 58 46 36 21 17 15 12 8 5 4 2 1 1 0 0 0

301 172 102

7 19

3 0

Cluster

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 55 57 18 19 20

123456

9 10 11 12

Time(years)

17 18 99

Time(years)

LIHC

Time(years)

Consensus matrix

1.00-

1.00

0.75

.76

150

0.25

Overall survival

p=0.185

p=0.013

p=0.041

Progression free surviva

P=0.279

Disease free survival

Disease specific survival

0.00

Time(years)

Number at risk

209

143

142

200

178

162

118

158

116

Cluster

Time(years)

PAAD

Time(years)

Consensus matrix

1.00-

1.00

Chuche

0.75

6.75

4.75

Overall survival

0 25

p<0.001

p=0.001

p<0.001

P<0.001

Progression free survival

Disease free survival

Disease specific tureval

0.00

Time(years)

Number at risk

₮

137

Cluster

Ouster

Cluster

Time(yeara)

Time(years)

”

Time(years)

SKCM

Time(years)

Consensus matrix

1.00-

1.004

1.00

0.75

.50

0.25

P<0.001

p<0.001

Overall survival

Progression free sureval

Disease specific survival

p=0.074

0.00

012

Time(years)

10012

Time(years)

Number at risk

Cluster

2009

Cluster

7382900

THYM

Time(years)

Consensus mavia

1.00

0.75

LTS

Cluster

₸

0.25

Overall survival

P<0.001

p<0.001

p=0.027

Progression free survival

Disease specific survival

-031

Time(years)

Number at risk

Cluster

4 .

Time(years)

Timelysans)

Time(years)

Fig. 3 Non-negative matrix factorization (NMF) classification based on prognostic DENGs. The NMF consensus matrix heat maps based on

optimal K value showed the classification status of ACC (a), CESC (c), LAML (e), LGG (g), LIHC (i), PAAD (k), SKCM (m) and THYM (o). Kaplan-

Meier plots (b, d, f, h, j, l, n, p) showed the relationship of different clusters and overall survival (OS), disease specific survival (DSS), progres-

sion free survival (PFS) as well as disease free survival (DFS) in the eight cancers, with logrank p value marked in the graphs

Springer

Fig. 4 Risk model construction and validation based on prognostic DENGs in ACC, CESC, LAML and LGG. LASSO coefficient spectrum of ▸ the selected genes (a, g, m, s) and the fivefold cross-validation (b, h, n, t) for variable selection were shown. Kaplan-Meier plots (c, i, o, u) showed the OS difference between patients from low- and high-risk groups sorted by median risk score of the training set, with logrank p value marked in the graphs. Time-dependent receiver operating characteristic (ROC) curves of 1, 3, 5-years (d, j, p, v) showed the predictive efficiency of the risk model, with area under curve (AUC) values noted in the graphs. The forest plots showed the results of univariate (e, k, q, w) and multivariate (f, l, r, x) survival analyses

of our risk models at predicting patients’ prognosis (Figs. 4d, p, v, 5d, p). The risk score could independently influence patients’ prognosis in both training and testing sets (Figs. 4f, r, x, 5f, r). However, In CESC and PAAD, we failed to observe the statistically significant difference of patients’ OS between low- and high-risk groups in the testing sets (Figs. 4i, 5i). We didn’t find a THYM cohort with sufficient prognostic information, so the analyses were only conducted in TCGA cohort (Fig. 5s-x). For ACC, LAML, LGG, LIHC, SKCM and THYM, the variation trend of model genes expression with the increase of risk score was shown, along with the comparison of some clinical factors between low- and high-risk groups (Fig. 6a, d, g, j, m, p). Dimensionality reduction analysis showed that the risk groups were largely in accordance with the two dimensional pattern of PCA, t-SNE and UMAP distribution, while in the testing set of LGG (CGGA cohort), the results were less satisfactory (Fig. 6b, e, h, k, n, q). With the increase of risk score, patients’ survival period was shortened and the number of deaths increased (Fig. 6c, f, i, l, o, r).

3.4 GSEA result

Gene Ontology (GO) and KEGG pathways related to the cell cycle were enriched in the high-risk group of ACC (Fig. 7a, c) and LIHC (Fig. 7e, g) no matter at training or testing sets, such as cell cycle checkpoint, cell cycle G1-S phase transi- tion, cell cycle G2-M phase transition, chromosome segregation, DNA dependent DNA replication and splicesome, with similar situation observed in low-risk group of THYM (Fig. 7j). In addition, innate and adaptive immune-related pathways were enriched in LGG high-risk group (Fig. 8e, g) and SKCM low-risk group (Fig. 8j, l) no matter at training or testing sets, including activation of immune response, adaptive immune response, antigen presenting and presentation as well as complement and coagulation cascades. Surprisingly, in the analysis of LAML, we found visible enrichment discrepancies in high-risk group at training and testing sets, with immune-related or cell-circle-related pathways separately enriched in the two sets (Fig. 8a, c).

3.5 Immune infiltration analysis of LGG and SKCM

Based on the GSEA results above, we further explored whether there existed any immune infiltration difference between low- and high-risk groups in LGG and SKCM. According to five immune infiltration assessment algorithms, high-risk LGG patients and low-risk SKCM patients had higher level of immune infiltration and function at both training and testing sets, which accorded with the GSEA enrichment results. For LGG patients, the infiltration level of B cells, plasma cells, CD8+ T cells, macrophages, endothelial cells, cancer-associated fibroblasts (CAFs) and dendritic cells was higher in high-risk group (Fig. 9a-d), while the situation of NK cells (Fig. 9a-d) and regulatory T (Treg) cells (Fig. 9a, d) was different between the various algorithms. For SKCM patients, the infiltration level of B cells, plasma cells, CD8+ T cells, CD4+ T cells (Th1 cells, Th2 cells), gammadelta T cells, macrophages, endothelial cells, dendritic cells, follicular helper T (Tfh) cells and Treg cells was higher in low-risk group (Fig. 9f-i). As shown in Fig. 9e, immune score, stromal score and ESTIMATE score were higher in LGG patients from high-risk group at both training and testing sets, which also positively correlated with the patients’ risk score. For SKCM patients, the results were opposite (Fig. 9j).

Then, we took TMB and MSI into consideration and found that high-risk LGG patients possessed higher TMB level (Fig. 10a), and TMB increased with risk score (Fig. 10b). Next, we explored the relationship of risk score and the gene expression of immunoinhibitors, immunostimulators, MHCs, chemokines and chemokine receptors. As shown in Fig. 10i-m, the expression of most immune-related genes positively correlated with risk score of LGG patients in both training and testing sets, while the results were opposite for SKCM patients (Fig. 10n-r).

3.6 Gene mutation status

We explored gene mutation status between low- and high-risk groups in TCGA cohorts of ACC, LAML, LGG, LIHC, SKCM and THYM, and screened out the top 20 genes with the highest mutation frequency. Higher mutation rate of tumor protein p53 (TP53) occurred in ACC and LIHC patients from high-risk group (Fig. 11a, d). For LAML and SKCM

CESC-TCGA ACC-TCGA

ACC

CESC

1.5

*= 0.09

HIAFY

A=0.01

1.0

Parial Likelihood Deviance

A=0.09 X=0.13

0.6

11.8

X-0.01

A-D.10

Coefficients

TRAPT

Coefficients

VDAGE

Parial Likelihood Deviance

0.4

11.6

0.0

HINTINDAN

11.4

0.2

HINT IMZAI

0.5

0.0

11.2

-1

-0.2

11.0

-1.5

10.8

-8

-6

-2

-0.4

-7

-5

-4

log2(lambda)

logž(lambda)

-log2(lambda)

Lasso Cox Risk - High risk

1.00

Low risk

Lasso Cox Risk

High risk

Low risk

1.00

Lasso Cox Risk

High risk

Low risk

Lasso Cox Risk + High risk + Low risk

Survival probability

1.00

Survival probability

0.75

Survival probability

0.75

0.50

0.25

p<0.001

0.25

p<0.001

0.25

P=0.636

0.00

P<0.001

0.00

SD 11 12

0123456

Lasse Cux Risk

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Time(years)

Timetyears)

Time(years)

Laspe Čen Rish

1 -

p 2

10 11 12

19 58 54 11 9 9 8 6 6 5 3 3 3 2 2 2 11 10 0 Time(years)

Timelynarı)

Time(years)

CESC-GSE44001

Sensitivity

AUC at 1 years: 0.858 AUC at 3 years: 0.936

AUC at 1 years: 0.863

AUC at 1 years: 0.804

AUC at 3 years: 0.794

AUC at 3 years: 0.683

AUC at 5 years: 0.873

AUG ant 5 years: 0.828

AUC at 5 years: 0.721

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

Age|>65)

0.393

1.589(0.549-4.601)

Stage[3/4]

0.001 2.421(1,427-4.100)

Gender(Male)

0.972

0.886(0.451-2.154)

M(M1)

<0.001 6.150(2.710-13.959)

N(N1)

0.152

2.038(0.765-5.400)

Grade(High)

0.012

3.315(1.300-8.453)

TỊ3/4)

<0.001 10.286(3.976-26.608)

Smoking history

0.453 0.745(0.322-1.728)

Stage(3/4)

0,015

2.700(1.212-6.016)

Pregnancy

0.634 0.844(0.416-1.713)

Risk score

<0.001

2.965(2.010-4.372)

0.25 0.5 4

Hazard ratio

pvalue

Hazard ratio

pvalue Hazard ratio

Stage(344)

0.013 1.969(1.154-3.362)

Age

0.040

1.537(1.032-3.634)

Stage(3/4)

0.651

0.618|0.076-4.991)

Risk score

<0.001 3.865(2.266-6.594)

0.0625

0.5

Hazard ratio

ACC-GSE19750+GSE33371

Age(>45)

0.818

0.912[0.414-2.009)

Age

0.024 2.062(1.101-3.860)

Gender(Male)

0.551

1.274(0.574-2.827)

Histology

0.354 0.649(0.265-1.619)

Race(White)

0.908 0.965(0.525-1.773)

Stage(3/4)

<0.001 6.476|2.706-15.458)

Risk score

<0.001 4.681(2.889-7.583)

0.25

1 2 4 8 16

0.25 0.5

Hazard ratio

pvalue

Hazard ratio

M(M1)

0.931

1.046|0.377-2.901)

Grade(High)

0.083

2.085(0.878-8.205)

Stage(3/4)

0.001

4.653(1.803-12.005)

Risik score

0.003

2.295(1.335-3.957)

Risk score

<0.001 2.815(1.842-4.209)

Hazard ratio

Risk score

<0.001

2.137[1.414-3.231)

T|3/4)

0.135 5.253(0.597-46.209]

LAML

LGG

A=0.11

W-0.32

A=0.05

A-0.11

12.0

W-0.05

W-0.11

0.4

PLAJCAN

8.8

0.8

Coefficients

Parial Likelihood Deviance

Coefficients

Parial Likelihood Deviance

0.6

11.5

0.2

3.4

0.4

3.2

0.2

11.0

0.0

1.0

0.0

-8

-0.2

10.5

-2

-10

-8

-4

-log2(lambda)

log2(lambda)

-log2(lambda)

log2(lambda)

1.00-

Lasso Cox Risk

High risk

Low risk

1.00

Lasso Cox Risk

High risk

Low risk

1.00

Lasso Cox Risk - High risk - Low risk

Lasso Cox Risk

High risk

Low risk

1.00

Survival probability

0.75

Survival probability

0.75

0.50

0.25

p<0.001

0.25

p<0.001

0.25

p=0.002

P<0.001

0.00.

0.00

0 1 2 3 4

@ 11 12 13 14 18

0123 45 6 7

14 15 16 17 18 19 20

0.00

Timetyears,

0.00

Lansa Con Risk

Lasso Con Risk

Time(years)

Lasas Cox Risk

9 10 11 12 1 ime(years)

Lasse Cox Risk

12 125 187 87 79 73

W 8

158

4 3

+ 15 16 17 1

Timelymars()

1 12 13 14

Time(years(

10 11 12 13

Sensitivity

AUC at 1 years: 0.745

AUC at 1 years: 0.850

AUG at 3 years: 0.751

AUC at 1 years: 0.614

AUC at 3 years: 0.612

AUC at 3 years: 0.885

AUC at 1 years: 0.762

AUC at 3 years: 0.820

AUC at 5 years: 0.841

AUG at 5 years: 0.519

AUC at 5 years: 0.796

AUC at 5 years: 0.749

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity

pvalue Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

Histology(AA)

<0.001

4.347(2.984-6.324)

Age|>41)

<0.001

Age(>65)

3.200(2.098-4.880)

<0.001

3.159|1.980-5.042)

FAB(M4)

<0.001

0.600|0.441-0.780)

Grade(3)

<0.001

2.610(1.789-3.808)

Age[>65]

<0.001

1.880(1.510-2.340)

Grade(3)

<0.001

3.323(2.170-5.089)

Gender[Male)

0.346

1.194(0.826-1.726)

Age(>41)

0.031

1.491(1.036-2.144)

Gender[Male)

0.781

0.939|0.602-1.464)

0.561(0.339-0926)

Gender(Male)

0.820

1.046(0.712-1.537)

Radio status

0.262

1.343(0.802-2.248)

runxinranciti fusion 0.024

Chemo statue

0.182

1.297(0.085-1.901)

<0.001

FAB(M3)

Radiation therapy

0.012

0.225(0.070-0.717)

2.267(1,403-3.664)

IDH mutation

<0.001

0.313(0.215-0.456)

runwi mutation

<0.001

2.077(1.601-2.693)

IDHt mutation

<0.001

0.247(0.167-0.365)

tp19g fetion

<0.001

0.192(0.113-0.326)

MIGMITp methylation

0.020

0.649(0.451-4.833)

Risk score

<0.001 6.625(3.060-14.340)

Risk score

0.002

1.620(1.200-2.185)

Rink score

<0.001

3.686(2.895-4.653)

Risk score

<0.001

3.173(2.439-4.129)

0.0625 0.25

0.25

0.5

0.125

0.5

0.0625

0.25

1 2

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

FAB(M4)

0.627(0.430-0.818)

Histology(AA)

0.352

1.503(0.637-3.545)

<0.001

Age(>41)

<0.001

2.227[1.389-3.547)

Grade(3)

0.221

1.668|0.735-3.787)

Age(>45)

<0.001 2.530(1.581-4.050)

Age[>65]

<0.001

1.712(1.369-2.141)

Grade(3)

0.026

1.744(1.067-2.845)

Agep41)

0.205

1.218|0.835-1.777)

runat-tanattti fusion

0.080

0.631(0.377-1.057)

Radiation therapy

1.133(0.668-1.921)

IDH_mutation

0.941 0.981|0.586-1.641)

FAB(M3]

0.914 0.927(0.236-3.635)

0.643

ip13q_tetion

<0.001

0.323|0.166-0.627)

unx1 mutation

<0.001

1.847(1.416-2.410)

IDH1 mutation

0.913

0.366[0.518-1.500)

0.111

0.734(0.502-1.074)

Risk score

<0.001 5.586(2.179-14.316)

Risk score

<0.001

1.705(1.243-2.339)

Risk score

<0.001

2.661(1.820-3.892)

Risk score

<0.001 2.001|1.373-2.317)

0.125

0.5

0.25

0.5

0.125 0.25 0.5

Hazard ratio

LAML-TCGA

LAML-GSE37642

LGG-TCGA

LGG-CGGA

Fig. 5 Risk model construction and validation based on prognostic DENGs in LIHC, PAAD, SKCM and THYM. LASSO coefficient spectrum of the selected genes (a, g, m, s) and the fivefold cross-validation (b, h, n, t) for variable selection were shown. Kaplan-Meier plots (c, i, o, u), time-dependent ROC curves of 1, 3, 5-years (d, j, p, v) and forest plots (e, f, k, l, q, r, w, x) showed the prognostic effectiveness of the risk models

LIHC

PAAD

0.4

A=0.02

UIPa1

0.4

A=0.04

Parial Likelihood Deviance

12.0

AM9.00

THỨ TƯ 1

0.3

Parial Likelihood Deviance

10.8

Ang.04

0.2

Coefficients

11.9

0.2

Coefficients

Pose

10.6

11.8

MINTIHIAC

0.1

1.31

0.0

11.7

0.2

TYKT

10.4

0.0

11.6

10.2

11.5

-0.4

0.1

11.4

0,0

-0.2

-8

€

-8

-4

-0.6

€

-4

-2

log2(lambda)

-log2(lambda)

1.00

Lasso Cox Risk

High risk

Low risk

Lasso Cox Risk

High risk

Low risk

Lasso Cox Risk

High risk

Lasso Cox Risk

Low risk

Survival probability

1.00

Low risk

1.00

High risk

0.75

Survival probability

0.75

0,75

0.50

0,50

0.50

0,50

0.25

p<0.001

0.25

p<0.001

0.25

p -< 0.001

0.25

p=0.256

0.00

₹

Lasse Cax Risk

Time

Lasse Cox Rik

$ Time(years)

Time(years)

Lasse Cax Risk

Time(years)

182

111

150

124

107

103

gh tính

1 4

finetyears)

Timelymars)

Time(years

Time(year)

PAAD-ICGC (PACA-AU)

Sensitivity

AUC af 1 years: 0,754

at 3 years: 0.654

AUC at 1 years: 0.669

AUC at 3 years: 0.670

AUG at 1 years: 0.730

AUC at 5 years: 0.645

AUC at 5 years: 0.590

AUC at 3 years: 0.729

AUC at 5 years: 0.924

0.0

0.2

0,4

0,6

0.8

1.0

0,0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

Age(>65)

0.233

255(0.864-1.824)

Gender(Male]

0.515(0.270-0.982)

Age(>65)

0.279

1.308(0.805-2.125)

0.044

Gender(Male)

0.590

0.875(0.539-1.421)

Gender(Male)

0.203

0.783|0.537-1.141)

Agep65)

0.647

1.165(0.607-2.235)

Stage(2b-4)

0.004

2.568(1.361-4.847)

Prior malignancy

0.848

1.06610.555-2.045)

Grade(3/4)

0.114

1.495(0.908-2.459)

Stage(3/4)

<0.001

2.454(1.691-3.560)

Stage(3/4]

0.002

2.761(1.464-5.210)

Alcohol history

0.394

1.261(0.740-2.143)

Grade(3/4)

0.427

1.163(0.801-1.689)

Prior malignancy

0.257

1.458(0.692-3.975)

History of diabetes

0.672

0.880(0.486-1.592)

History of chronic pancreatitis

0.855

0.929(0.424-2.039)

Risk score

<0.001

3.558(2.354-6.378)

Risk score

<0.001

3.400(1.799-6.429)

Risk score

<0.001

3.594(2.187-5.906)

0.5

0.25 0.5

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

Gender[Male)

<0.001

0.300(0.149-0.605)

Stage(3/4)

<0.001 2.181(1,498-3.176)

Stage(2b-4)

0.223

1.499(0.781-2.877)

Stage(3/4)

0.008

2.543(1.272-5.085)

Risik score

<0.001 3.314(2.176-5.047)

Risk score

<0.001

3.191(1.601-6.360)

Risk score

<0.001

3.302(1 973-5.528)

0,125

0,5

0.5

Hazard ratio

LIHC-TCGA

LIHC-ICGC (LIRI-JP)

PAAD-TCGA

SKCM

THYM

A=0.02

YOACE

12.3

BOAWE

Parial Likelihood Deviance

Coefficients

Parial Likelihood Deviance

0.2

2.2

150

Coefficients

12.1

100

0,0

2.0

1.9

-5

-0.2

1.8

-10

log2[lambda)

-4

-14

-12

-10

-8

-6

-4

-log2(lambda)

log2(lambda)

-log2(lambda)

1.00

Survival probability

0.75

.50

9.50

0.50

0.25

p<0.001

0.25

p=0.007

0.25

p<0.001

0.00

· 1 2 3 4 5 6 7 8 9 101113 13 6615 16 17 18 19 20 31 22 33 38 25 36 27 30 39 50

Timelynersi)

10 11 12 himelyears!

16 17 18 19 20

Lasse Cex Risk

Lasse Ces Risk

141 56 30 23 13 12 8

Lasto Cen Risk

10879 82 41 23 18 18 14 9 7 4 3 3 2 2 1 1 1 0 0 0

”

: 30.11, 12 13 14 15 16 17 18 18 20

Sensitivity

0.6

Sensitivity

0.4

AUC af 1 years: 0.716

AUC at 3 years: 0.686

AUC at 1 years: 0.634

AUC af 5 years: 0,708

AUC at 3 years: 0.638

AUC at 1 years: 0.854

AUC at 5 years: 0.586

AUC at 3 years: 0.936

AUC at 5 years: 0.966

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

Age[>65]

0.004

1.623|1.172-2.243)

Age

0.185

2.448(0.652-9.185)

Gender[Male)

0.957

1.009|0.734-1.386)

Gender[Male)

0.143 1.366(0.899-2.075)

M(1)

0.112

1.944|0.857-4.409)

Gender

0.513

0.643(0.172-2.409)

N(1/2/3)

<0.001

1.865|1.367-2.545)

Age[>65)

0.441

0.853(0.569-1.279)

Stage

0.033

4.557(1.129-18.396)

T[3/4]

<0.001

2.001(1.463-2.737)

0.884

Stage(3/4)

<0.001

1.771(1.300-2.412)

Stage(General)

<0,001

2.712(1.721-4.276)

Primary site

1.124(0.233-5.419)

0.323

0712(0.363-1.395)

prior

0.558

malignancy

1.874(0.229-15.343)

Risk score

<0.001 3.276(2.373-4.522)

Risk score

0.019

1.656|1.088-2.521)

Risk score

40.001

4.980(2.379-10.459)

0.25 0.5

0.5

0.25

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

pvalue

Hazard ratio

AgepG5)

0,031

1.443(1.035-2.013)

N(1/2/3)

0.148

2.112(0.767-5.811)

Stage(General)

<0.001

2.738(1.735-4.325)

Stage(2b-4)

0.510

1.65%(0.370-7 423)

T(3/4)

0.008

1.562(1.124-2.172)

Stage(2/4)

0.825

0.892(0.325-2.450)

Risk score

<0.001 2.901(2.104-4.001)

Risk score

0.016

1.490(1,101-2.596)

Risk score

<0.001

4.665(2.184-9.965)

0.25 0.5

Hazard ratio

SKCM-TCGA

SKCM-GSE65904

THYM-TCGA

G Springer

Fig. 6 Model genes expression, dimensionality reduction analysis and distribution of risk score and survival state. The heat maps (a, d, g, j, m, p) showed the variation trend of model genes expression with the increase of risk score and the comparison of several clinical factors between low- and high-risk groups in the six selected cancers. Fisher's exact test was used. * p<0.05; ** p<0.01; *** p <0.001. Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP) (b, e, h, k, n, q) confirmed the stratification of patients into low- and high-risk clusters. The scatter diagrams (c, f, i, I, o, r) showed the condition of patients' risk score and distribu- tion of their survival time and state, with dotted line separating patients into low- and high-risk groups

ACC

LAML

FADD

FTH1

TCGA

GSE19750

GSE33371

TCGA

PLA204A

GSE37642

AIFM1

PCA2(17.05%)

PCA2[20.11%)

PÇA2(25.60%)

PCA2(28.67%)

UMAP_2

UMAP 2

UMAP_2

PCA1(38.65%)

PCA1(24.65%)

PCA1(42.63%)

PCA1(36.17%)

ISNE_2

UMAP_1

ESNE_2

UMAP_1

tSNE_2

UMAP_1

ISNE_2

UMAP_1

ISNE_1

High

₣ Low risk

Survival ame years)

High rik Low risk

Survival time (years)

Alive

High risk

Risk score

Survival time (years)

High rá

Survival time (years)

Risk score

bei

Nol

300

500

200

300

Patients (increasing risk score)

LGG

LIHC

Pran

OFLAR

BLUDI

PLAHGAA

PLAZOLA

OFLAKE

PRVA

TCGA

CGGA

TCGA

ICGC (LIRI-JP)

PCA2(19.43%)

PCA2[22.63%)

PCA2[17.17%)

PCA2[18.19%)

UMAP_2

PCA1(39.73%)

PCA1(39.86%)

PCA1(41.06%)

PCA1(43.10%)

ISNE_2

UMAP_1

ISNE_2

UMAP_1

ISNE_2

UMAP_1

ISNE_2

UMAP_1

ISNE-1

ISNE_1

Survival time (years)

High rik Low risk

Survival time (years)

High rink

Survival time (years)

Low ták

High risk

LON BIA

Risk score

BA 10

Risk score

Survival time (years)

100

200

400

200

200 400

200

150

200

Patients (increasing risk score)

SKCM

THYM

PLAY545

PGAMS

CHMINE

PARIS

TCGA

GSE65904

TCGA

PCA2(19.94%)

PCA2(19.50%)

PCA2(19.75%)

UMAP_2

Ring

UMAP_2

PCA1(23.02%)

PCA1(22.41%)

PCA1(34.69%)

ISNE_2

UMAP_1

ISNE_2

UMAP_1

ISNE_2

UMAP_1

ISNE_1

3 High risk Low rik

· Low risk

Low risk

Risk score

Survival time (years)

Risk score

Survival time (years)

Risk score

Survival time (years)

150 200

Patients (increasing risk score)

200

400

100

Patients (increasing risk score)

Fig. 7 Gene Set Enrichment Analyses (GSEA) in ACC, LIHC and THYM. GSEA shows the top 5 gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enriched in low- and high-risk groups of ACC (a-d) and LIHC (e-h) at both training and testing sets. For THYM (i, j), GSEA was only conducted in TCGA cohort. Normalized enrichment score (NES), adjusted p-value and q-value were marked in the plots

ACC-TCGA Enriched in High risk group

Enriched in High risk group

Enriched in Low risk group

ACC-GSE19750+GSE33371

Enriched in High risk group

Enriched in Low risk group

LIHC-TCGA

Enriched in High risk group

Enriched in Low risk group

Rank in Ordered Detmet

LIHC-ICGC (LIRI-JP)

Enriched in High risk group

Enriched in Low risk group

THYM-TCGA

Enriched in High risk group

Enriched in Low risk group

Rankin Crowrue Dronet

Rank in Orteset Dataset

patients from low-risk group and LIHC patients form high-risk group, higher mutation rate of mucin 16, cell surface associated (MUC16) was observed (Fig. 11b, d, e). In addition, isocitrate dehydrogenase (NADP(+)) 1 (IDH1), capicua transcriptional repressor (CIC), far upstream element binding protein 1 (FUBP1), SWI/SNF related, matrix associ- ated, actin dependent regulator of chromatin, subfamily a, member 4 (SMARCA4) and AT-rich interaction domain 1A (ARID1A) were more likely to mutate in LGG patents from low-risk group. However, higher mutation rate of titin (TTN), epidermal growth factor receptor (EGFR), neurofibromin 1 (NF1), phosphatase and tensin homolog (PTEN) and

Fig. 8 GSEA in LAML, LGG and SKCM. GSEA shows the top 5 GO and KEGG pathways enriched in low- and high-risk groups of LAML (a-d), LGG (e-h) and SKCM (i-I) at both training and testing sets. NES, adjusted p-value and q-value were marked in the plots

LAML-TCGA Enriched in High risk group

Enriched in High risk group

Enriched in Low risk group

LAML-GSE37642

Enriched in High risk group

Enriched in Low risk group

Rank in Ordered Distaset

LGG-TCGA

Enriched in High risk group

Enriched in Low risk group

LGG-CGGA

Enriched in High risk group

Enriched in Low risk group

SKCM-TCGA

Enriched in High risk group

Enriched in Low risk group

Rank in Ordered Gotaset

Rank in Dedered Ducaoet

SKCM-GSE65904

Enriched in High risk group

Enriched in Low risk group

Rank in Ordered Dotaset

Rank in Ordored Dotscet:

Park in Dedered Dassset

Rank in Ordered Detsset

Research

CIBERSORTImmune Score

LGG-TCGA

Discover Oncology

B_cells_naive 8.cells_memory amony restin

T_cells_CD4,memur celis

plasma_cells

Y_calls_CD4_memory_ Activated

T cells.CDo-CO8 y calls_C08

” cells_follicular help

Y_cells_regular ya gelt T cells.Samma_della

NK_cells resting

NK cells_activated

Monocytes

EPIC Immune Score

LGG-CGGA

S.2

Macrophages,” Macrophages_StZ Dendritic cells_restin

Dendritic,Con Mast cells_resting Mast_cesis_activated Eosinophils Neutrophils

(2022) 13:17

B_cols_naive

B cells_memory

T_cells_CD4_memory activate T_cells_CDAPcoa

CAFE

MCPcounter Immune Score

r_cells_follicular_helps

…

NK_celly,resting

CD4_Tcells

NK cells_activatss

CD8_Tcells

MacroMonocytes

Endothelial_cells Macrophages

CIBERSORTImmune Score

Macrophages_MY stacrophages_M2

Dendritic cells,resting Dendritic_colis Actu Mast cellye,resting Mast_cells_A Eosinophils

0 A

SKCM-TCGA

| https://doi.org/10.1007/s12672-022-00477-2

R.00

1 10

Neutrophils

T_coils

NK_cella

100

COS_T_cols

Cytotoxic_lymphocytes

B_celu

B_cells_naive a celis_memory

B_Aneage

U SSGSEA Immune Score 1

pissma_cells emory Activated

T_cells_CD4_memory,resting

CAFE

NK_cells

Monocylic_lineage

CD4_Tcela

Y_calls_CO4

cells folicular pelo

Myeloid_dendritic_cells

SKCM-GSE65904

Y_cells_regulatory. T_cells Samma della NK_cells_resting

Neutrophils

NK_cells salva Macrophago Monocyte

Endothelial_cells

Endothelial_cells Macrophages

EPIC Immune Score

Macrophages_Mi Macrophages base del. restin Dendritic dues resting Dendritic, calls.A Mast cells_testing

Fibroblasts

Sosinophis

aDCs

B_cells

r_cells

NK_celha

Neutrophils

CD8+_T_cells

CD8_T_cells

ssGSEA Immune Score

DCS

Cytotoxic_lymphocytes

B_cells_naive 8 cells_memory

IDCs

Macrophages

T_cells_CD4,memory_festins

Mast_cells

8,lineage

T_cells_CD4_memory_activated

Neutrophils

NK_cells

CAFS

T_cells_CD4.CO8

waar helps

Monocylic_lineage

pocs

T_helper_cells

Myeloid_stendritic_cells

MCPcounter Immune Score

NK cells resting

· High

1th

Thi_cells

004_Toelis

Th2_cells

Neutrophils

NK cells_activated - Macrophan inte

Endothelial_cells

CD8_Tcells

TIL

Macrophages,’ Macrophages_Mz

Treg

thing

Fibroblasts

APC_co_inhibition APC_co_stimulation

Erudothelial_celis Macrophages

Dendritic cells_restin Cells_activate Mast cells_resting

Check-point

rosinaphils

Neutrophils

CCR

aDCs

B_cells

Cytolytic_activity

Inflammation-promoting MHC_class_)

CD8+_T_cells

DCS

T_cells

NK_cella

IOCS

CD8_T_cells

Immune Score

HLA

Parainflammation

Macrophages

Mast cells

T_cell_co-inhibition

Neutrophils

Cytotoxic_lymphocytes

B_cela

P + 2.224-18

NK_cells

8_lineage

CAFE

T_cell_co-stimulation

poCs

ssGSEA Immune Score

NK_cells

T_helper_cells

Tth

Monocyte_lineage

Type_I_IFN_Reponse Type_Il_IFN_Reponse

CD4_Tcells

Th2_cells TIL

Thi_cells

Myeloid_dendritic_cells

Low-risk

₱<2.228-18

APC_co_Inhibition

Neutrophils

Treg

Endothelial_cells

Immune Score

Stromal Score

Endothelal_cele Macrophages

High-risk

APC_co_stimulation Check-point

Fibroblasts

ESTIMATE Score

CCR

aDCs

Low-risk

Cytolytic_activity

Inflammation-promoting MHC_class_)

B_cells

T cells

INK_calls

Stromal Score

CD8+_T_cells

HLA

DCS

R = 0.47.9 < 2.26-18

High-risk

COS_T_cells

Parainflammation

ssGSEA Immune Score

IDCS

Cytotoxic_lymphocytes

Risk score

Macrophages

B_lineage

Low-risk

Immune Score

T_cell_co-inhibition

T_cell_co-stimulation

Mast cells

P . 2.220-18

Type ___ IFN_Reponse Type_Il_IFN_Reponse

Neutrophils

ESTIMATE Score

NK_cells

· High

High-risk

Monocytic_Bneage

poCs

T_helper_cells

Myetold_dendritic_cells

Thi_cells

Neutrophils

between risk score and immune score, stromal score as well as ESTIMATE score, with Spearman’s correlation coefficient R value and p value

between low- and high-risk groups of LGG and SKCM patients based on CIBERSORT (a, f), EPIC (b, g), MCPcounter (c, h) and ssGSEA (d, i), with Wilcoxon signed-rank test applied. * p<0.05; ** p<0.01; *** p<0.001; **** p <0.0001. The scatter diagrams (e, j) showed the relationship

Fig. 9 Immune infiltration analysis. The box plots and violin plots showed the difference of immune infiltration level and immune function

Risk score

P <2.2/9-16

Th2_cells

TIL

Endothelial_coils

Immune Score

Stromal Score

Low-risk

High-risk

APC_co_inhibition

Treg

Fibroblasts

APC_co_stimulation

DCS

marked in the plots

Risk score

ESTIMATE Score

Check-point

8_cells

Low-risk

P + 2.220-18

CCR

Cytolytic_activity

·1 -0.48. p < 2.28-18

Stromal Score

High-risk

Inflammation-promoting MHC_class_)

CD8+_T_cells

DCS

IDCS

HLA

Macrophages

Mast_cells

Risk score

Low-risk

Immune Score

Parainflammation

T_cell_co-inhibition

T_cell_co-stimulation

Neutrophils NK_cells

pDCs

T_helper_cells

TIA

ryanodine receptor 2 (RYR2) was found in high-risk LGG patents (Fig. 11c). The mutations of general transcription factor IIi (GTF2I) and HRas proto-oncogene, GTPase (HRAS) were more common in high-risk THYM patients (Fig. 11f).

Risk score

1- 0.01. 9 < 2.28-18

ESTIMATE Score

High-risk

Type ___ IFN_Reponse

Type_Il_IFN_Reponse

Thi_cells

Th2_cells

Low-risk

P + 2.224-5%

APC_co_inhibition

Treg

R=0.4.9<2.2 .- 16

Immune Score

Stromal Score

High-risk

Check-point CCR

APC_co_stimulation

Risk score

ESTIMATE Score

Cytolytic_activity

Low-risk

9 . 2.220-18

MHC_class_) HLA

Inflammation-promoting

3.7 Correlation between risk score and drug sensitivity

Stromal Score

High-risk

Risk score

Parainflammation

Immune Score

T_cell_co-inhibition

Low-risk

P . 2.224-16

R =- 0.47.0 4.2 50-18

ESTIMATE Score

High-risk

Finally, we paid attention to the drug selection. As shown in Fig. 12a, d, with the increase of risk score, ACC and LIHC may be more sensitive to adenine nucleotide analogues, such as nelarabine, clofarabine and cladribine. For high-risk

Stromal Score

T_cell_co-stimulation Type ___ IFN_Reponse Type_I_IFN_Reponse

Risk score

Low-risk

9.79-10

Immune Score

High-risk

Risk score

ESTIMATE Score

Low-risk

0 4.2.2-16

Stromal Score

High-risk

LGG and LAML/SKCM with low-risk score, dasatinib was perhaps a good choice (Fig. 12b, c, e). For THYM, the irofulven sensitivity positively correlated with risk score, but a negative correlation was detected between the sensitivity of vinorelbine, vinblastine as well as eribulin mesilate and risk score (Fig. 12f).

Risk score

Low-risk

ESTIMATE Score

High-risk

Risk score

Springer

H =- 4.7.p <2.26-18

Risk score

4 Discussion

Necroptosis is a novel programmed cell death mode independent on caspase, with increasing evidence of anti-tumor effects discovered in recent years. As we know, traditional chemotherapeutic agents usually induced cell apoptosis to exert anti-tumor effects [26]. However, tumor cells are inherently anti-apoptotic. In spite of the prevalence of heterogeneity in various tumors, there’s a high possibility that the subpopulation of tumor cells with greater anti- apoptotic selection superiority will gradually clone and govern the entire tumor as the treatment proceeds. There- fore, drug resistance has become a common fact during clinical practice, and tumors which relapse or progress after treatment are extremely difficult to deal with [26]. Thus, it became a natural idea to induce other types of cell death for drug-resistant tumors, and alternative choices mainly included ferroptosis, pyroptosis as well as necroptosis [27]. Numerous studies have proven that the transition of apoptosis to necroptosis or the direct induction of necroptosis could make for overcoming drug resistance and inhibiting tumor development for various cancers, such as acute myeloid leukemia [28, 29], breast cancer [30], osteosarcoma [31], nasopharyngeal carcinoma [32], prostate cancer [33, 34] and colon cancer [35, 36].

In this study, based on TCGA and GTEx data, we identified eight types of cancer with the highest number of prog- nostic DENGs and for the first time sorted ACC, CESC, LAML, LGG, LIHC, PAAD, SKCM and THYM patients into different subgroups based on necroptosis-related genes. Kaplan-Meier analysis of four follow-up endpoints showed that the classification was excellent in distinguishing patients’ OS in all cancers above except for LIHC. Then, the risk models were set up. Unfortunately, the risk models didn’t work at testing sets of CESC and PAAD, but we do find a method to efficiently distinguish patients’ OS in ACC, LAML, LGG, LIHC and SKCM. The testing set of LAML (GSE37642) lacked M3-subtype patients and the testing set of LGG (CGGA) only consisted of Asian patients, so there existed some intrin- sic discrepancies between TCGA cohorts (used as training set) and these testing sets. This might cause the inconsist- ency of AUC values between training and testing set. Notably, among these five cancer types, ACC is relatively less studied. As a rare malignancy with great complexity, the 5-year DFS rate of ACC was only about 30%, and there still existed many therapeutic challenges [37, 38]. Due to the heterogeneity of ACC, the prognostic efficiency of the most widely accepted TNM staging was inevitably limited [39]. Thus, it is necessary to seek new risk factors for ACC patients. Our ACC risk model based on necroptosis-related genes has good predictive ability for patient’ survival, which might provide meaningful references for patients’ prognosis in the future clinical practice.

Although kinds of immunotherapies have achieved remarkable success in cancer treatment, only limited number of patients could exhibit long-lasting anti-tumor response, where tumor immune infiltration status played a significant role [40]. Identification of cancer patients with abundant infiltration of immune cells is of great importance to screen out potential candidates for immunotherapy. Our GSEA results of SKCM and LGG cohorts highlighted immune-related GO and KEGG pathways in low- and high- risk groups, which along with results of the estimated immune infiltration level based on five algorithms could contribute to the distinction of “cold” and “hot” tumors.

As we know, immunotherapies have not acquired satisfactory results in glioma patients in recent years, including adoptive lymphocyte transfer, tumor associated vaccine, viral-based therapy and ICIs, where T-cell exhaustion played a dominant role, and tumor heterogeneity, blood brain barrier as well as lack of immune organs in central nerve systems also shared the blame [41]. Although there is a higher CD8+ T cells infiltration level in high-risk LGG patients, we failed to observe the difference of cytotoxic lymphocytes between the two risk groups according to MCPcounter. Noteworthy is the infiltration level of M2 macrophages and CAFs is higher in high-risk LGG patients. Recent studies have revealed the fact that M2 macrophages played a vital part in the development of glioma by promoting tumor invasion and metastasis, facilitating tumor stemness as well as suppressing immunity of the tumor area and the whole body [42, 43]. CAFs were involved in tumor cell replication, angiogenesis, chemotherapy insensitivity and the sup- pression of CD8+ T cell function [44, 45]. M2 macrophages and CAFs have been considered as promising therapeutic targets by number of studies [44-46], and high-risk LGG patients perhaps benefit from the agents which inhibit M2 macrophages or CAFs.

Unlike the situation in LGG, the infiltration level of immune cells widely known for suppressing tumor develop- ment is higher in low-risk SKCM patients, including CD8+ T cells, Th1 cells and M1 macrophages. According to the correlation analysis of risk score and immune-related gene expression, SKCM patients from low-risk group also pos- sessed a higher gene expression level of plenty of immunosuppressive molecules, some of which were identified as immune checkpoints and their therapeutic potential has been proven by numerous studies. ICIs were initially studied and applied for the clinical application in melanoma, and Ipilimumab, targeting cytotoxic T-lymphocyte-associated

▸

Fig. 10 Tumor mutational burden (TMB), microsatellite instability (MSI) and immune-related genes expression analysis. Bar graphs showed the comparison of TMB (a, e) and MSI (c, g) between low- and high-risk groups and scatter diagrams showed the correlation between TMB (b, f) or MSI (d, h) and the risk score of LGG and SKCM patients. Wilcoxon signed-rank test p value and Spearman’s correlation coefficient R value as well p value were marked in the graphs. The correlations between risk score and the expression of immunoinhibitor genes (i, n), immunostimulator genes (j, o), MHC genes (k, p), chemokine genes (I, q) as well as chemokine receptor genes (m, r) were shown, with ”*” representing Pearson correlation p value < 0.05

protein 4 (CTLA4), is the first drug in history to significantly prolong the survival period of patients with this highly malignant tumor [47]. Programmed cell death protein 1 (PD-1) antibody was also approved for the treatment of advanced melanoma by FDA in the year of 2014 and phase 3 clinical trial of Relatlimab, targeting lymphocyte-activa- tion gene 3 (LAG-3), has met its primary endpoint of PFS, which may offer new hope for SKCM patients in the future. It needs to be mentioned that there existed a higher mutation rate of MUC16 in low-risk SKCM patients. MUC16, also known as carbohydrate antigen 125 (CA125), ranks third in the list of gene mutation frequency of cancers, whose mutation occurs most frequently in SKCM [48]. The study also showed that MUC16-mutated melanoma patients treated with ICIs had significantly longer OS. Given that our study could help to recognize SKCM patients with higher level of immune infiltration and immune-checkpoint genes expression as well as higher MUC16 mutation rate, it is reasonable to believe that low-risk SKCM patients are more likely to benefit from ICIs treatment.

Although we failed to find a cohort to check the predictive ability of prognosis in the THYM risk model, there were still some results which could arouse our attention. First, the nine-genes risk model successfully assigned all death cases into high-risk group, and the following time-dependent ROC analysis exhibited an excellent predictive ability of the model with 1, 3, 5-year OS area under the ROC curve up to 0.854, 0.936 and 0.966. Regardless of the application of which dimensionality reduction method, the cases could be obviously divided into low- and high-risk clusters. Thymoma has a low incidence and favorable prognosis, so the associated studies are relatively limited compared with other common or highly malignant tumors. For patients classified as high-risk, their review period perhaps needs to be shortened so that the tumor progression can be detected and treated in time.

For THYM, it is still controversial whether adjuvant radiotherapy or chemotherapy should be applied after surgery. According to our result, some of the patients classified as high-risk might be the potential candidates for postopera- tive adjuvant therapies. We noticed a decline in the sensitivity of tumor cells to vincaleukoblastinum drugs with the increase of risk score based on THYM risk model. However, irofulven exhibited anti-tumor activity in cells with high risk score, which is a kind of cytotoxic drug proven to be an effective agent for tumors with DNA repair deficiency by several studies [49, 50]. This finding may provide some useful information for the clinical chemotherapy of THYM. In addition, we noticed the mutation rate of GTF2I in the high-risk patients was about twice as high as that in low-risk patients. Researchers have found that there existed a high mutation rate of GTF2I in indolent thymomas, which was extremely rare in aggressive thymomas and thymic carcinomas [51]. Mutant GTF2I, identified as a novel tumorigenic driver, can promote growth, proliferation and transformation of epithelial cell as well as alter glucose and lipid metabolism [51, 52], and whether it could work as a therapeutic target requires further research.

5 Conclusions

In summary, this is the first study to comprehensively investigate the genes of necroptosis pathway in all TCGA cancers. We conducted NMF to classify ACC, CESC, LAML, LGG, PAAD, SKCM and THYM patients into subgroups with different prognosis. The risk model based on necroptosis-related genes can effectively predict the prognosis of ACC, LAML, LGG, LIHC, SKCM and THYM patients. The risk score contributes to the identification of immune infiltration level for LGG and SKCM patients, which could help to screen out the potential candidates who might benefit from immunotherapy. Genetic mutation status and drug sensitivity were also different for patients from different risk groups, which may offer meaningful information for the future clinical practice.

LGG-TCGA

SKCM-TCGA

Low-risk High-risk

1.5

1e-08

1.5

R = 0.33, p = 6.5c-14

80-

0.1

R =- 0.11, p = 0.02

Tumor mutational burden

1.0

0.5

0.0

Low-risk

High-risk

D.D

Low-risk

High-risk

Risk score

Low-risk E High-risk

0 Low-risk E High-risk

0.36-

0.23

0.350

0.36

0.3

0.36

…

R=0.068. p=0.15

0.34

Microsatellite instability

0.33

0.325

0.32

0.30

0.300

0.28

0.27

0.275

0.26

R=0.013, p= 0.77

0.24

Low-risk

High-risk

0.24

Low-risk

High-risk

Risk score

LGG-TCGA

LGG-CGGA

SKCM-TCGA

SKCM-GSE65904

Risk score

ADORAZA

HAVGR2

BYLA

HOFFE

IL1DRB

KOR

LGUAL59

VTON

POCDILG2

LAD3

ADORAZA

HAVOR2

ADORAZA

POCDILG

1001

ILSD

ILSORB

Risk soon

KOR

LA43

VTCN1

HTLA

cross

CO274

CSFIR

CTLAN

HAVCH2

LOALS9

TGFER!

VICN1

CSFIR CTL44

HAVVCR2

ILOR8

KDP

LAG3

LGALS9

PDCD1

TGFERI

TIGHT

VIČN1

Fonk score

Fosk score

Risk score

ADORA2

BTLA

ADORAS

BTLA

森

ADORAZ

BTLA

CD160

CD16

0024

CD160

00244

CUZN

CDZN4

5.4

CSF

0.4

CTLA4

CTLA

CTLA4

HAVGRZ

HAVGR

100

HAVCR

IDO

L 10

IDO

LID

IL10

L10RB

₿

KOR

-02

IL TORI

KDR

-02

KOR

KIRZOL1

-02

KOR

KIR2DL#

-42

LAGT

LAG3 LOALS

KIRZDL

-04

KIRZDL

LAALSI

-0,4

-44

POCO

LGALS

POCD1

PDCD1LG2

PDCDILO

-44

PDCD1

-44

Immunoinhibitor

TGFBR

Immunoinhibitor

TGFBR

Immunoinhibitor

PDCDILG2

-08

TGFBRI

PDCD1LG2

DIGIT

-48

Immunoinhibitor

TGFORT

VTCN

-1

VTCN1

-1

VICNI

-1

VICNI

-1

BOY

₥

Immunostimulator

Hak score

B2M

HLA-

HLA-ON HLA-DM

HLA-DO

HLAPLUS

HLA-DRA

HLA-DRCB

HLA-

HLA-1

HLA-D

HLA-DM

HLA-D

HLA-DOE

HLA-DOB

HLA-DRA

HLA-DRS

TAPT

TAPUIF

6 HLA

HLA-

TAPBP

HLA

HLA-

HLACURA

HLA-OR

HLA-

IHLA-F

HLAP

TAPZ

TAPB

Risk score

B2M

000

B2M

HLA-

HLA-A

0.0

0.8

HLA-B

HLA

HLA-

HLA-C

HLA-

HLA

ILA-DMA

0.6

HLA

0.6

HLA-DMA

HLA-DMB

0.4

HLA-DMB HLA-DO

HLA-DOA

0.4

HLA-DIMB

HLA-DO HLA-DOA

HLA-DOA

0.4

HLA-DOB

0.2

HLA-DOB

HLA-DPA

HLA-DOB

HLA-DPM

0.2

HLA-OPA1

HLA-DPA1

HLA-DPB1

HLA-DPB

HLA-DPB1

HLA-DOM1

HLA-DOM

HLA-DQA

HLA-DQA1

HLA-DOA

-0.2

HLA-DOAZ

-0.2

HLA-DOA

HLA-DOB

-0.2

HLA-DOB

-0.2

HLA-DOB1

HLA-ORA

-0.4

HLA-DOB1

HLA-DR

-0.4

HLA-ORD

-0.4

HLA-DRDY

HLA-DRB

-0.6

HLA-ORB

HLA HLA-E

HLA-

-0.6

HUA-

HLA-

-0.8

HLA-F

MHC

HLA-

-0.0

MHC

HLA-

MHC

HLA-G

TAP

MHC

HLA-G

TAP1

TAI

TAPBR

TAPER

TAPOP

8 8

-42

-44

Chemokine

Risk sopre

CCR

CCR2

COR

CCR5

CCR

CCR7

CORO

CCR10

CXCRT

CXCRZ

CXCR

CXCHO

CXCR6 XCR1

CX3CR1

Risk score

CCR

CCRJ

CCR

CCR5

CORT

CCR9

CXCRT

CXCR

CXCI

CNCRA

CAURA

CXCR6

XCR1

CX3CR1

Risk score

CCR

CCR6

CCR7

CCR

CCR9

CORTO

CXCR

CXCRA

CXGRI

XCA1

CX3CR

Risk score

CCR

CCR5

CCR7

CCRA

CXCHT

CXGRA

CXCRO

XCR

CX3CR

Risk score

CCR1

CCRI

OGRI

* O

CCR1

CCR2

CCIO

CORI

CCR4

CORA

CCR4

CORS

CCRS

CORS

0.4

CCR

CCRS

CCR

CCRS

CCR

CCR7

COR

O .

CCR7

CCRS

CORS

CCR8

CORE

CCR10

CCRS

CXCR

CXCRI

CORSO

CCRIO

CXCR2

-02

CXCR1

-0.2

-02

CXCR3

CXCR1

CXCR3

O .

CXCRS

-0.6

CXCRS

-0.6

CXCRS

-0.6

CXCRS

-0.6

Chemokine receptor

CXCR6

Chemokine receptor

CXCR6

XCRI

-0.8

XCRI

Chemokine receptor

CXCR6

-08

XOR1

-0.8

Chemokine receptor

XCRI

CX3OR

CX3CR

CXJCR

CX3CR

Fig. 11 Gene mutation status in low- and high-risk groups. The oncoplots showed the mutation status of the top 20 most frequently mutated genes of ACC (a), LAML (b), LGG (c), LIHC (d), SKCM (e) and THYM (f) at low- and high-risk groups, with different colors referring to gene mutation types. The mutation rate of each gene between the two risk-groups was compared by Fisher's exact test, and the genes with higher mutation rate in low- or high-risk groups were highlighted by blue or red color accordingly

ACC-TCGA

LAML-TCGA

Altered in 21 (52.5%) of 40 samples.

Altered in 36 (92.31%) of 39 samples.

Altered in 28 (66.67%) of 42 samples.

Altered in 38 (79.17%) of 48 samples.

943

592

171

No. of samples

TP53

9 TP53

No. of samples

No. of sampin

28%

DNMT3A

ONMT3A NPM1

19%

CTNNB1

12%

19%

MUC16

GTNNB1

10%

MUC16

21%

10%

18%

NPM1 FLT3

FLT3

12%

MUC4 TIN

MUC4

TIN

21%

TP53

IOH2

TP53

IDH2

5% 5%

PKHO1

18%

10%

6ª%

PKHD1

13%

10%

CNTNAPS

CNTNAP5

1.3%

RUNX1

DST

10%

14%

KIT

20%

OST

WT1

PCDHIS

PCDUHA

PCDH15

10%

OH1

10%

OHT

HMMNT

NF1

NRAS

NBAC

ASXL3

10%

TET2

10%

MUC16

SVEPT

SVEP1

MEN1

10% 13%

KRAS

TETZ

KRAS

PRKARIA

10%

ANK2

2% 0%

PRKARIA

GATA2

ANK2

CMYAS

10%

CEBPA

5% 2% 5%

CEBPA

CMYAS

13% 5%

SPEN

ASKL1

FBN?

ASXL1

SPEN

CCDC168

FAT4

10%

ARMGAP3S

PTPN11

ARHGAPSS

FAT4

PTPN11

Risk

. Missense_Mutation Frame_Shift_Del Frame_Shift_Ins

Nonsense_Mutation

Risk

. Missense_Mutation Nonsense_Mutation Frame_Shift_Ins

In_Frame_Del Frame_Shift_Del

Risk

High Low

· Missense_Mutation Frame_Shift_Del In_Frame_Ins In_Frame_Del

Nonsense_Mutation Frame_Shift_Ins

High Low

Missense_Mutation Frame_Shift_Ins

Frame_Shit_Del · Multi_Hit

Risk

· Multi_Hit

High

Low

· Multi_Hit

Nonsense_Mutation

Low

LGG-TCGA

LIHC-TCGA

Altered in 245 (99.59%) of 246 samples.

Altered in 226 (92.24%) of 245 samples.

Altered in 135 (77.14%) of 175 samples.

Altered in 161 (92%) of 175 samples.

929

1200 -

No. of samples

229

No of samples

151

No. of samples

ICH1

93%

IDH1

62%

TP53

14%

TP53

445%

TP53

CTNNB1

TP53

42%

31%

ATRX

37%

44%

20%

CTNNB1

CIc

ATRX

33%

TTN

23%

TTN

10%

MUC16

23%

33%

MUC16

TTN I

19%

CIC

12%

10%

ALB

PCLO

ALB

PIK3CA

PCLO

11%

FURPI

PIK3CA

FUBP1

APOB

13% 11%

EGFR

11%

APOR RYR2

EGFR

NOTCH1

12%

RYR2

NOTCH1

MUC4

MUCA

NF1

MF1

FLO

Med

MUPAR

ABCA

HORIA

ABCA19

SMARPAL SMARCA4

THE Peut

BIR

10%

WANT

IDH2

XIRP2

9% 5%

CSMD3

RYR2

ICH2

RYR2

XIRP2

FAT3

FLO

296

FLO

FAT3

39%

10%

10% 9%

ZUT820

HMCN1

HUCH1

HMCN1

ZUT820

HMCN1

CACNATE

ARIDIA

CACNATE

ARIDIA

79%

ARIDIA

AXIN1

ARIDIA

AXIN1

15%

Risk

Missense_Mutation Nonsense_Mutation Frame_Shift_Ins

Frame_Shift_Del

Risk

In_Frame_Del

high low

· Missense_Mutation Frame_Shift_Del Nonsense_Mutation In_Frame_Del

· Frame_Shift_Ins In_Frame_Ins Translation_Start_Site

· Multi_Hit

high low

Missense_Mutation Frame_Shift_Del Nonsense_Mutation = Multi_Hit In_Frame_Del

Frame_Shift_Ins In_Frame_Ins

Frame_Shift_Ins In_Frame_Del

Missense_Mutation Frame_Shift_Del “Nonsense_Mutation . Multi_Hit

High

High Low

. Multi_Hit

Low

SKCM-TCGA

THYM-TCGA

Altered in 213 (95.09%) of 224 samples.

Altered in 198 (87.61%) of 226 samples.

Altered in 15 (33.33%) of 45 samples.

Altered in 41 (71.93%) of 57 samples.

3200

13723

646

No. of samples

165

No. of samples

156

No. of samples

TTN

74%

69%

GTF21

20%

G GTF21

49%

MUCTE BRAF

73%

MUC16

HRAS

53%

60%

TTN

HRAS

TTN

14%

ONAHS PCLO

46%

BRAF DNAHS PCLO

48%

MUC16

LRPIB

45%

50%

MUC16

37%

42% 37%

TP53

NRAS

TP53

NRAS

ADORV1

35%

LRP18

ADORV1

37%

BP1

33%

MUCA

MUC4

CSNOI

33%

29%

PCLO GYLD

PCLO

CYLD

CSMDI

DNA,47

31%

DNAHT

31%

NACAD

ANKS

34%

NACAD

ANK3

33%

FAMATA

CHE

BDP1

MINEL

PERL

FAMA

FAT4 APOB FLG

FAT4

THE

33%

32%

APOB FLG

29%

ARE

DNAH17 DMD

09%

DNAHTY DMD

V V

HYDIN

31%

30%

3% 5%

HYDIN

31%

PKHDIL1

30%

PKCHOIL1

0% 4%

35%

27%

RPIL1

NBPF14

CSMD3

MGAM

30%

CSMD3

NBPF14

MGAM

29%

20%

NROB1

MORC2

NROB1

DSCAM

USPOX

MORG2

32%

26%

USPSX

Risk

Missense_Mutation Nonsense_Mutation Frame_Shift_Ins

Nonstop_Mutation In_Frame_Del

Risk

Missense Mutation Frame_Shift_Del

Frame_Shift_Ins In_Frame_Del

Risk

. Missense_Mutation = Multi_Hit

Risk

High Low

Missense_Mutation Nonsense_Mutation In_Frame_Ins

Frame_Shift_Del In_Frame_Del · Multi_Hit

Risk

· Multi_Hit

Nonsense_Mutation . Mulsi_Hit

High

High Low

Low

High Low

Frame_Shift_Del

Fig. 12 Drug sensitivity exploration. We calculated the risk score of each CellMiner sample according to the genes and coefficient of the risk models of the six cancers. The scatter diagrams showed the correlation between risk score and sensitivity (z-score) of Food and Drug Administration (FDA)- approved drugs in ACC (a), LAML (b), LGG (c), LIHC (d), SKCM (e) and THYM (f), with Pearson correlation coefficient (Cor) and p value marked above the graphs

Risk score, Nelarabine Cor=0.470, p<0.001

ACC

Risk score, Clofarabine Cor=0.427, p<0.001

Risk score, Trametinib Cor=0.370, p=0.004

LAML

Risk score, ARRY-162

Cor=0.332, p=0.010

1.0

5.0

0.5

0,0

2.5

-0.5

-1

0.0

-1.0

-2

-1.5

-1

1.5

2.0

2.5

1.5

2.0

2.5

Risk score, Cladribine Cor=0.395, p=0.002

Risk score, Hydroxyurea Cor=0.392, p=0.002

Risk score, Dasatinib Cor =- 0.331, p=0.010

Risk score, Cobimetinib (isomer 1) Cor=0.312, p=0.015

1.5

1.0

0.5

0.0

-1

-0.5

-1

-1.0

-1

1.5

2.0

2.5

1.5

2.0

2.5

Risk score, Irinotecan Cor=0.376, p=0.003

Risk score, Uracil mustard Cor=0.367, p=0.004

Risk score, Selumetinib Cor=0.311, p=0.015

Risk score, JNJ-42756493 Cor =- 0.302, p=0.019

-1

-2

-1

1.5

2.0

2.5

1.5

2.0

2.5

Risk score, Dasatinib Cor=0.479, p<0.001

LGG

Risk score, Ponatinib Cor=0.389, p=0.002

Risk score, Vinorelbine Cor =- 0.347, p=0.007

LIHC

Risk score, VINORELBINE Cor =- 0.330, p=0.010

1.5

1.0

0.5

0.0

-1

-0.5

-2

-1.0

-3

1.0

1.5

2.0

2.5

3.0

-2

1.0

1.5

2.0

2.5

3.0

4.0

4.5

5.0

5.5

4.0

4.5

5.0

5.5

Risk score, TYROTHRICIN Cor =- 0.374, p=0.003

Risk score, Pazopanib Cor=0.320, p=0.013

Risk score, Paclitaxel Cor =- 0.308, p=0.017

Risk score, Cladribine Cor=0.293, p=0.023

-1

-2

-1

1.0

1.5

2.0

2.5

3.0

-3

1.0

1.5

2.0

2.5

3.0

4.0

4.5

5.0

5.5

4.0

4.5

5.0

5.5

Risk score, Midostaurin Cor=0.305, p=0.018

Risk score, JNJ-42756493 Cor=0.281, p=0.030

Risk score, Eribulin mesilate Cor =- 0.282, p=0.029

Risk score, Clofarabine Cor=0.269, p=0.038

-1

-2

-1

-2

1.0

1.5

2.0

2.5

3.0

1.0

1.5

2.0

2.5

3.0

4.0

4.5

5.0

5.5

-2

4.0

4.5

5.0

5.5

Risk score, METHOTREXATE Cor=0.399, p=0.002

SKCM

Risk score, Zoledronate Cor =- 0.382, p=0.003

Risk score, Vinorelbine Cor =- 0.326, p=0.011

THYM

Risk score, Vinblastine Cor =- 0.256, p=0.048

-1

-2

-3

-2

-1

Risk score, 6-MERCAPTOPURINE Cor=0.353, p=0.006

Risk score, JNJ-42756493 Cor =- 0.351, p=0.006

Risk score, Irofulven Cor=0.289, p=0.025

Risk score, Eibulin mesilate Cor =- 0.270, p=0.037

-1

-2

-1

-2

-4

-2

Risk score, Dasatinib Cor =- 0.347, p=0.007

Risk score, Belinostat Cor=0.347, p=0.007

1.5

1.0

0.5.

0.0

-0.5.

-1

-1.0

-2

Acknowledgements We acknowledge TCGA, GTEx, GEO, ICGC and CGGA databases for providing abundant gene profiles and clinical data. We also appreciate Sangerbox tools (http://www.sangerbox.com/tool), a free online platform, for several data analysis.

Authors’ contributions LL and QZ performed the project design. JM and YJ contributed to data collection and analysis. YJ and BG participated in literature search and figure production. JM finished the manuscript. All authors read and approved the final manuscript.

Funding This study was supported by National Key Research and Development Program of China (2018YFC1313000, 2018YFC1313001, 2018YFC1313002, 2018YFC1313004).

Data availability The datasets used in this article can be acquired from the internet. They can be downloaded from the corresponding open databases mentioned in this article.

Declarations

Ethics approval and consent to participate This study is based on the data from TCGA, GTEx, GEO, ICGC and CGGA database. The patients involved in the database have obtained ethical approval. Users can download relevant data for research and publish relevant articles. There is no clinical trial or animal experiment in our research.

Consent for publication All authors agree to publish this paper.

Competing interests The authors declare that there is no competing interest in this work.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adapta- tion, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

References

1. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2021. CA Cancer J Clin. 2021. https://doi.org/10.3322/caac.21654.

2. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011. https://doi.org/10.1016/j.cell.2011.02.013.

3. Buja LM, Eigenbrodt ML, Eigenbrodt EH. Apoptosis and necrosis. Basic types and mechanisms of cell death. Arch Pathol Lab Med. 1993;117:1208-14.

4. Kawahara A, Ohsawa Y, Matsumura H, Uchiyama Y, Nagata S. Caspase-independent cell killing by Fas-associated protein with death domain. J Cell Biol. 1998. https://doi.org/10.1083/jcb.143.5.1353.

5. Wallach D, Varfolomeev EE, Malinin NL, Goltsev YV, Kovalenko AV, Boldin MP. Tumor necrosis factor receptor and Fas signaling mechanisms. Annu Rev Immunol. 1999. https://doi.org/10.1146/annurev.immunol.17.1.331.

6. Chan FK, Shisler J, Bixby JG, Felices M, Zheng L, Appel M, et al. A role for tumor necrosis factor receptor-2 and receptor-interacting protein in programmed necrosis and antiviral responses. J Biol Chem. 2003. https://doi.org/10.1074/jbc.M305633200.

7. Degterev A, Huang Z, Boyce M, Li Y, Jagtap P, Mizushima N, et al. Chemical inhibitor of nonapoptotic cell death with therapeutic potential for ischemic brain injury. Nat Chem Biol. 2005. https://doi.org/10.1038/nchembio711.

8. Galluzzi L, Kroemer G. Necroptosis: a specialized pathway of programmed necrosis. Cell. 2008. https://doi.org/10.1016/j.cell.2008.12.004.

9. Galluzzi L, Vitale I, Aaronson SA, Abrams JM, Adam D, Agostinis P, et al. Molecular mechanisms of cell death: recommendations of the nomenclature committee on cell death 2018. Cell Death Differ. 2018. https://doi.org/10.1038/s41418-017-0012-4.

10. Pasparakis M, Vandenabeele P. Necroptosis and its role in inflammation. Nature. 2015. https://doi.org/10.1038/nature14191.

11. Ofengeim D, Yuan J. Regulation of RIP1 kinase signalling at the crossroads of inflammation and cell death. Nat Rev Mol Cell Biol. 2013. https://doi.org/10.1038/nrm3683.

12. Wang H, Sun L, Su L, Rizo J, Liu L, Wang LF, et al. Mixed lineage kinase domain-like protein MLKL causes necrotic membrane disruption upon phosphorylation by RIP3. Mol Cell. 2014. https://doi.org/10.1016/j.molcel.2014.03.003.

13. Zhou W, Yuan J. Necroptosis in health and diseases. Semin Cell Dev Biol. 2014. https://doi.org/10.1016/j.semcdb.2014.07.013.

14. Gong Y, Fan Z, Luo G, Yang C, Huang Q, Fan K, et al. The role of necroptosis in cancer biology and therapy. Mol Cancer. 2019. https://doi. org/10.1186/s12943-019-1029-8.

15. Tang R, Xu J, Zhang B, Liu J, Liang C, Hua J, et al. Ferroptosis, necroptosis, and pyroptosis in anticancer immunity. Hematol Oncol. 2020. https://doi.org/10.1186/s13045-020-00946-7.

16. Yatim N, Jusforgues-Saklani H, Orozco S, Schulz O, Barreira DSR, Reis ESC, et al. RIPK1 and NF-kappaB signaling in dying cells determines cross-priming of CD8(+) T cells. Science. 2015. https://doi.org/10.1126/science.aad0395.

17. Aaes TL, Kaczmarek A, Delvaeye T, De Craene B, De Koker S, Heyndrickx L, et al. Vaccination with necroptotic cancer cells induces efficient anti-tumor immunity. Cell Rep. 2016. https://doi.org/10.1016/j.celrep.2016.03.037.

18. Snyder AG, Hubbard NW, Messmer MN, Kofman SB, Hagan CE, Orozco SL, et al. Intratumoral activation of the necroptotic pathway com- ponents RIPK1 and RIPK3 potentiates antitumor immunity. Sci Immunol. 2019. https://doi.org/10.1126/sciimmunol.aaw2004.

19. Seifert L, Werba G, Tiwari S, Giao LN, Alothman S, Alqunaibit D, et al. The necrosome promotes pancreatic oncogenesis via CXCL1 and Mincle-induced immune suppression. Nature. 2016. https://doi.org/10.1038/nature17403.

20. Demeure MJ, Coan KE, Grant CS, Komorowski RA, Stephan E, Sinari S, et al. PTTG1 overexpression in adrenocortical cancer is associated with poor survival and represents a potential therapeutic target. Surgery. 2013. https://doi.org/10.1016/j.surg.2013.06.058.

21. Heaton JH, Wood MA, Kim AC, Lima LO, Barlaskar FM, Almeida MQ, et al. Progression to adrenocortical tumorigenesis in mice and humans through insulin-like growth factor 2 and beta-catenin. Am J Pathol. 2012. https://doi.org/10.1016/j.ajpath.2012.05.026.

22. Lee YY, Kim TJ, Kim JY, Choi CH, Do IG, Song SY, et al. Genetic profiling to predict recurrence of early cervical cancer. Gynecol Oncol. 2013. https://doi.org/10.1016/j.ygyno.2013.10.003.

23. Li Z, Herold T, He C, Valk PJ, Chen P, Jurinovic V, et al. Identification of a 24-gene prognostic signature that improves the European Leu- kemiaNet risk classification of acute myeloid leukemia: an international collaborative study. J Clin Oncol. 2013. https://doi.org/10.1200/ JCO.2012.44.3184.

24. Zhao Z, Zhang KN, Wang Q, Li G, Zeng F, Zhang Y, et al. Chinese glioma genome atlas (CGGA): a comprehensive resource with functional genomic data from chinese glioma patients. Genom Proteom Bioinform. 2021. https://doi.org/10.1016/j.gpb.2020.10.005.

25. Cirenajwis H, Ekedahl H, Lauss M, Harbst K, Carneiro A, Enoksson J, et al. Molecular stratification of metastatic melanoma using gene expression profiling: prediction of survival outcome and benefit from molecular targeted therapy. Oncotarget. 2015. https://doi.org/10. 18632/oncotarget.3655.

26. Hu X, Han W, Li L. Targeting the weak point of cancer by induction of necroptosis. Autophagy. 2007. https://doi.org/10.4161/auto.4592.

27. Liu Y, Chen Q, Zhu Y, Wang T, Ye L, Han L, et al. Non-coding RNAs in necroptosis, pyroptosis and ferroptosis in cancer metastasis. Cell Death Discov. 2021. https://doi.org/10.1038/s41420-021-00596-9.

28. Safferthal C, Rohde K, Fulda S. Therapeutic targeting of necroptosis by Smac mimetic bypasses apoptosis resistance in acute myeloid leukemia cells. Oncogene. 2017. https://doi.org/10.1038/onc.2016.310.

29. Chromik J, Safferthal C, Serve H, Fulda S. Smac mimetic primes apoptosis-resistant acute myeloid leukaemia cells for cytarabine-induced cell death by triggering necroptosis. Cancer Lett. 2014. https://doi.org/10.1016/j.canlet.2013.10.018.

30. Han W, Li L, Qiu S, Lu Q, Pan Q, Gu Y, et al. Shikonin circumvents cancer drug resistance by induction of a necroptotic death. Mol Cancer Ther. 2007. https://doi.org/10.1158/1535-7163.MCT-06-0511.

31. Fu Z, Deng B, Liao Y, Shan L, Yin F, Wang Z, et al. The anti-tumor effect of shikonin on osteosarcoma by inducing RIP1 and RIP3 dependent necroptosis. BMC Cancer. 2013. https://doi.org/10.1186/1471-2407-13-580.

32. Liu T, Sun X, Cao Z. Shikonin-induced necroptosis in nasopharyngeal carcinoma cells via ROS overproduction and upregulation of RIPK1/ RIPK3/MLKL expression. Onco Targets Ther. 2019. https://doi.org/10.2147/OTT.S200740.

33. Rizzi F, Naponelli V, Silva A, Modernelli A, Ramazzina I, Bonacini M, et al. Polyphenon E(R), a standardized green tea extract, induces endo- plasmic reticulum stress, leading to death of immortalized PNT1a cells by anoikis and tumorigenic PC3 by necroptosis. Carcinogenesis. 2014. https://doi.org/10.1093/carcin/bgt481.

34. Deeraksa A, Pan J, Sha Y, Liu XD, Eissa NT, Lin SH, et al. Plk1 is upregulated in androgen-insensitive prostate cancer cells and its inhibition leads to necroptosis. Oncogene. 2013. https://doi.org/10.1038/onc.2012.309.

35. Grassilli E, Narloch R, Federzoni E, Ianzano L, Pisano F, Giovannoni R, et al. Inhibition of GSK3B bypass drug resistance of p53-null colon carcinomas by enabling necroptosis in response to chemotherapy. Clin Cancer Res. 2013. https://doi.org/10.1158/1078-0432.CCR-12-3289.

36. Oliver MM, Fuchs D, Tagscherer KE, Grone HJ, Schirmacher P, Roth W. Inhibition of caspases primes colon cancer cells for 5-fluorouracil- induced TNF-alpha-dependent necroptosis driven by RIP1 kinase and NF-kappaB. Oncogene. 2016. https://doi.org/10.1038/onc.2015. 398.

37. Allolio B, Fassnacht M. Clinical review: adrenocortical carcinoma: clinical update. J Clin Endocrinol Metab. 2006. https://doi.org/10.1210/ jc.2005-2639.

38. Else T, Kim AC, Sabolch A, Raymond VM, Kandathil A, Caoili EM. Adrenocortical carcinoma. Endocr Rev. 2014. https://doi.org/10.1210/er. 2013-1029.

39. Fassnacht M, Johanssen S, Quinkler M, Bucsky P, Willenberg HS, Beuschlein F, German Adrenocortical Carcinoma Registry Group, European Network for the Study of Adrenal Tumors. Limited prognostic value of the 2004 International Union Against Cancer staging classification for adrenocortical carcinoma: proposal for a revised TNM classification. Cancer. 2009. https://doi.org/10.1002/cncr.24030.

40. Chen DS, Mellman I. Elements of cancer immunity and the cancer-immune set point. Nature. 2017. https://doi.org/10.1038/nature21349.

41. Fecci PE, Sampson JH. The current state of immunotherapy for gliomas: an eye toward the future. J Neurosurg. 2019. https://doi.org/10. 3171/2019.5.JNS181762.

42. Vidyarthi A, Agnihotri T, Khan N, Singh S, Tewari MK, Radotra BD, et al. Predominance of M2 macrophages in gliomas leads to the sup- pression of local and systemic immunity. Cancer Immunol Immunother. 2019. https://doi.org/10.1007/s00262-019-02423-8.

43. Liu Z, Kuang W, Zhou Q, Zhang Y. TGF-beta1 secreted by M2 phenotype macrophages enhances the stemness and migration of glioma cells via the SMAD2/3 signalling pathway. Int J Mol Med. 2018. https://doi.org/10.3892/ijmm.2018.3923.

44. Biffi G, Tuveson DA. Diversity and biology of cancer-associated fibroblasts. Physiol Rev. 2021. https://doi.org/10.1152/physrev.00048.2019.

45. Chen Y, McAndrews KM, Kalluri R. Clinical and therapeutic relevance of cancer-associated fibroblasts. Nat Rev Clin Oncol. 2021. https:// doi.org/10.1038/s41571-021-00546-5.

46. Komohara Y, Fujiwara Y, Ohnishi K, Takeya M. Tumor-associated macrophages: potential therapeutic targets for anti-cancer therapy. Adv Drug Deliv Rev. 2016. https://doi.org/10.1016/j.addr.2015.11.009.

47. Hodi FS, O’Day SJ, McDermott DF, Weber RW, Sosman JA, Haanen JB, et al. Improved survival with ipilimumab in patients with metastatic melanoma. N Engl J Med. 2010. https://doi.org/10.1056/NEJMoa1003466.

48. Zhang L, Han X, Shi Y. Association of MUC16 mutation with response to immune checkpoint inhibitors in solid tumors. JAMA Netw Open. 2020. https://doi.org/10.1001/jamanetworkopen.2020.13201.

49. Jiang H, Greenberg RA. Morning for irofulven, what could be fiNER? Clin Cancer Res. 2021. https://doi.org/10.1158/1078-0432.CCR-20-4708.

50. Borcsok J, Sztupinszki Z, Bekele R, Gao SP, Diossy M, Samant AS, et al. Identification of a synthetic lethal relationship between nucleo- tide excision repair deficiency and irofulven sensitivity in urothelial cancer. Clin Cancer Res. 2021. https://doi.org/10.1158/1078-0432. CCR-20-3316.

(2022) 13:17

51. Petrini I, Meltzer PS, Kim IK, Lucchi M, Park KS, Fontanini G, et al. A specific missense mutation in GTF2I occurs at high frequency in thymic epithelial tumors. Nat Genet. 2014. https://doi.org/10.1038/ng.3016.

52. Kim IK, Rao G, Zhao X, Fan R, Avantaggiati ML, Wang Y. Mutant GTF2I induces cell transformation and metabolic alterations in thymic epithelial cells. Cell Death Differ. 2020. https://doi.org/10.1038/s41418-020-0502-7.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Quartz 4

Explorer

35312867

Abstract

Keywords Cancer . Necroptosis . Prognosis . Risk score . Tumor immune infiltration

Abbreviations

1 Introduction

2 Methods

2.1 Gene expression and clinical data collection

2.3 Construction and validation of DENGs-based risk model

2.4 Gene set enrichment analyses (GSEA)

2.5 Investigation of tumor immune microenvironment

2.6 Analysis of gene mutation

2.7 Drug sensitivity analysis

3 Results

3.1 Identification of prognostic DENGs in TCGA-cancers

3.2 Tumor classification

3.3 LASSO regression risk models

3.4 GSEA result

3.5 Immune infiltration analysis of LGG and SKCM

3.6 Gene mutation status

4 Discussion

5 Conclusions

Declarations

References

Graph View

Table of Contents

Backlinks

Quartz 4

Explorer

35312867

Pan-cancer analysis of necroptosis-related gene signature for the identification of prognosis and immune significance

Abstract

Keywords Cancer . Necroptosis . Prognosis . Risk score . Tumor immune infiltration

Abbreviations

1 Introduction

2 Methods

2.1 Gene expression and clinical data collection

2.2 Identification of differentially expressed necroptosis-related genes (DENGs), survival analysis and tumor classification

2.3 Construction and validation of DENGs-based risk model

2.4 Gene set enrichment analyses (GSEA)

2.5 Investigation of tumor immune microenvironment

2.6 Analysis of gene mutation

2.7 Drug sensitivity analysis

3 Results

3.1 Identification of prognostic DENGs in TCGA-cancers

3.2 Tumor classification

3.3 LASSO regression risk models

3.4 GSEA result

3.5 Immune infiltration analysis of LGG and SKCM

3.6 Gene mutation status

4 Discussion

5 Conclusions

Declarations

References

Graph View

Table of Contents

Backlinks