Multi-omics landscape and clinical significance of a SMAD4-driven immune signature: Implications for risk stratification and frontline therapies in pancreatic cancer

Abstract

SMAD4 mutation was recently implicated in promoting invasion and poor prognosis of pancreatic cancer (PACA) by regulating the tumor immune microenvironment. However, SMAD4-driven immune landscape and clinical significance remain elusive. In this study, we applied the consensus clustering and weighted correlation network analysis (WGCNA) to identify two heterogeneous immune subtypes and immune genes. Combined with SMAD4-driven genes determined by SMAD4 mutation status, a SMAD4-driven immune signature (SDIS) was developed in ICGC-AU2 (microarray data) via machine learning algorithm, and then was validated by RNA-seq data (TCGA, ICGC-AU and ICGC-CA) and microarray data (GSE62452 and GSE85916). The high-risk group displayed a worse prognosis, and multivariate Cox regression indicated that SDIS was an independent prognostic factor. In six cohorts, SDIS also displayed excellent accuracy in predicting prognosis. Moreover, the high-risk group was characterized by higher frequencies of TP53/CDKN2A mutations and SMAD4 deletion, superior immune checkpoint molecules expression and more sensitive to chemotherapy and immunotherapy. Meanwhile, the low-risk group was significantly enriched in metabolism-related pathways and suggested the potential to target tumor metabolism to develop specific drugs. Overall, SDIS could robustly predict prognosis in PACA, which might serve as an attractive platform to further tailor decision-making in chemotherapy and immunotherapy in clinical settings.

Keywords: Pancreatic cancer; Prognosis; SMAD4 mutation; Signature; Therapeutic response.

Graphical abstract

Fig. 1

The workflow of our research. Our detailed process for obtaining SMAD4-driven immune genes and establishing SDIS is as follows: (1). According to the SMAD4 mutation status of PACA samples, we first obtained 2279 SMAD4 mutation-driven genes by differential analysis. (2). The abundance of 28 immune cell types in PACA samples was assessed using the single sample gene set enrichment analysis(ssGSEA) algorithm, after which the samples were stratified into two immune subtypes (Immune-H and Immune-L) using consensus clustering, and the black module (containing 1446 immune-related genes) most associated with immune subtypes was further identified using the weighted correlation network analysis (WGCNA) algorithm. (3). Using Venn diagram, 180 SMAD4-driven immune genes were obtained by intersection of 2279 SMAD4-driven genes and 1446 immune-related genes. (4). Afterwards, we screened 3 risk and 14 protective eligible genes (P < 0.2 and all of the hazard ratio (HR) > 1 or < 1) in at least 3/4 of the cohorts by univariate Cox regression analysis. (5). Finally, we generated a 7-gene prognostic signature in the ICGC-AU2 training cohort by backward stepwise regression machine learning algorithm using the 17 eligible genes obtained in step (4) and named: SMAD4-driven immune signature (SDIS).

Fig. 2

Mutational landscape and genes driven by SMAD4 mutation of pancreatic cancer (PACA) patients in TCGA cohort. (A) Frequency and type of mutations in the top 30 genes in PACA. Genes are sorted according to frequency of mutations. (B) Identification the differentially expressed genes between the SMAD4^MUT and SMAD4^WT patients. Orange dots represented up-regulated genes, green dots represented down-regulated genes and grey dots represented genes with no significance.

Fig. 3

Screening immune-related genes. (A) The consensus score matrix of all samples when k = 2. A higher consensus score between two samples indicates they are more likely to be grouped into the same cluster in different iterations. (B) The cumulative distribution functions of consensus matrix for each k (indicated by colors). (C) The proportion of ambiguous clustering (PAC) score, a low value of PAC implies a flat middle segment, allowing conjecture of the optimal k (k = 2) by the lowest PAC. (D) Two-dimensional principal component plot by the abundance of 28 immune cell types in the two clusters. The orange dots represented C2, and blue dots represented C1. (E) Analysis of network topology for various soft-thresholding powers. (F) The infiltration heatmap of 28 immune cell types in the two subtypes. (G) The correlation heatmap of the modules obtained by WGCNA and clinical traits. (H) The correlation between genes within black modules with immune subtypes.

Fig. 4

The development and validation of SMAD4-driven immune signature (SDIS) for pancreatic cancer (PACA). (A, B) Kaplan-Meier analysis for OS (A) and RFS (B) between the high- and low-risk groups in ICGC-AU2 cohort. (C, D) Multivariate Cox regression analysis of OS (C) and RFS (D) in ICGC-AU2 cohort. (E-J) The ROC curves of SDIS for predicting 1-, 2-, and 3-years OS in ICGC-AU2 cohort (E), TCGA cohort (F), ICGC-AU cohort (G), ICGC-CA cohort (H), GSE62452 cohort (I), GSE85916 cohort (J). OS, overall survival; RFS, relapse-free survival; ROC, receiver operating curve.

Fig. 5

The genomic alterations landscape of the high- and low-risk groups in TCGA cohort. (A) The waterfall plot of top 20 mutation genes in the two groups. Each column represented individual patients. The right barplot and number indicated the mutation frequency in each gene. The color of each column indicated the mutation type of this gene in the sample. (B) Heatmap of mutation frequency of top 20 mutation genes in the high- and low-risk groups. (C) The top 10 amplification and homozygous deletion (HOMDEL) genes in the high- and low-risk groups.

Fig. 6

The KEGG and GO enriched pathways and immune landscapes in the high- and low-risk groups. (A-B) The top five KEGG enriched pathways in the high (A) and low (B) risk groups. (C-D) The top five GO enriched pathways in the high (C) and low (D) risk groups. (E) The heatmaps of 8 immune cell types and 27 immune checkpoints profiles in the high- and low-risk groups. KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, Gene Ontology.

Fig. 7

Efficacy evaluation of chemotherapy and immunotherapy in the high- and low-risk groups patients. (A-E) The estimated IC50 of several common drugs such as gemcitabine (A), paclitaxel (B), cisplatin (C), JNJ.26854165 (D) and CGP.60474 (E) in the high- and low-risk groups. (F) The TIDE tool was used to predict the sensitivity of the two subtypes to immunotherapy in TCGA cohorts. (G) Submap analysis of the two subtypes and 47 pretreated patients with comprehensive immunotherapy annotations in TCGA cohort. For Submap analysis, a smaller p-value implied a more similarity of paired expression profiles. (H-J) Kaplan-Meier survival analysis of high SDIS and low SDIS groups (H), the ROC curve of SDIS for predicting 1- and 2-year OS (I), and ROC curve of SDIS for predicting immunotherapy response (J) in GSE91061 cohort. IC50, half-maximal inhibitory concentration; TIDE, Tumour Immune Dysfunction and Exclusion; SDIS, SMAD4-driven immune signature.