Single-cell RNA-seq reveals invasive trajectory and determines cancer stem cell-related prognostic genes in pancreatic cancer

Figures & data

Figure 1. ScRNA-seq analysis reveals a variety of cell types in PDAC and control pancreas

(a) UMAP displayed the main cell types (left). Representative markers across the major cell types are displayed in the bubble diagram (right).(b) UMAP displayed the diverse ductal cell types in PDAC and control pancreas.(c) Organizational contribution rate measured the difference between the ductal cell types.

Figure 2. ScRNA-seq analysis reveals the status and invasive trajectory of ductal cells

(a) Heatmap displaying large-scale copy number variations (CNVs) of ductal cells in PDAC and control pancreases. The normalized CNV levels are shown: red represents a high CNV level and blue represents a low CNV level.(b) Heatmap displaying the expression of specific markers in the different ductal cell types.(c) Functional enrichment analysis of genes specifically expressed in each ductal cell types.(d) Monocle 2 reveals the trajectory of ductal cells in PDAC and control pancreases. Each point corresponds to a single cell.(e) The differentially expressed genes (rows) along the pseudo-time (columns) are clustered hierarchically into three profiles. The color key from blue to red indicates relative expression levels from low to high, respectively.

Table 1. Screening of CRGs

Upregulated CRGs

ADAM9,AHR,AKR1C3,ALDOA,ANG,ANPEP,ANXA1,APOA1,AREG,B2M,BIRC5,CA9,CCL20,CCNB1,CD151,CD55,CD68,CD82,CD99,CDCA7,CDCP1,CDH17,CDKN2A,CEACAM1,CEACAM5,CEACAM6,CEACAM7,CLDN4,COL1A1,CP,CTSB,CTSD,CTTN,CXCL10,CYP3A4,DKK1,DMBT1,EIF4EBP1,ERBB3,EZR,F3,GALNT12,GAPDH,GRN,GSN,HK2,HLA-B,HLA-DQB1,HLA-DRB1,HLA-E,HMGA1,HNF4A,HPGD,ID1,IFI27,IGFBP2,IGFBP3,IL18,IL1RN,IL2RG,ITGA2,ITGA3,ITGA6,ITGB4,JUP,KCNN4,KLF4,KLF5,KRAS,KRT13,KRT17,KRT19,KRT20,KRT7,LAMA3,LAMB3,LAMC2,LCN2,LDHA,LDLR,LGALS1,LGALS3,LY6D, MACC1,MAGEA4,MDK,MECOM,MET,MKI67,MMP1,MSLN,MST1R,MUC1,MUC4,MUC5AC,MVP,MXRA5,NDRG1,NEAT1,NQO1,NTS,PKM,PLAT,PLAUR,PLEC,PMAIP1,PPARG,PSCA,PSMB9,PTGS2,PTTG1,RAC1,RHOC,RRAS,S100A2,S100A4,S100A6,S100A8,S100A9,SAMD9,SDC1,SERPINA1,SERPINB5,SFN,SH3KBP1,SLC16A1,SLC22A18,SLC2A1,SPINK1,ST14,TFF1,TFRC,TGM2,TIMP1,TKT,TMPRSS2,TOP2A,TPM3,TXN,TYMP

Downregulated CRGs.

ALB,APP,BCAM,CADM1,CCL2,CCND1,CD81,CFTR,CLU,COL18A1,CRP,CXCL1,DAB2,DLC1,DLK1,DUSP1,EGR1,EPHX1,FGFR2,FGFR3,GLUL,GMNN,HBA2,HBB,HES1,HNF1B,HSP90AA1,HSPA1A,HSPA8,HSPD1,ID2,IDH2,IGFBP7,IL1R1,INS,JUN,MCAM,MEG3,MEIS1,NFIB,NFKBIA,NOTCH2,NRP1,NTRK2,PBX1,PDCD4,PDGFD,PEBP1,PKHD1,PROX1,PTCH2,RIC3,S100A1,SETBP1,SPHK1,SPP1,TTN,TUBA1A,VCAM1,VTN,WWTR1,ZBTB16

Figure 3. Survival and ROC analysis in training and validation datasets

(a-c) Kaplan–Meier survival curves for patients in high- and low-risk groups of TCGA (a), GSE79668 (b), and GSE62452 (c) datasets.(d-f) Time‐dependent ROC curves at 3, 4, and 5 years for patients in TCGA (d), GSE79668(e), and GSE62452 (f) datasets to evaluate the prediction efficiency of the prognostic signature.

Table 2. Univariate and multivariate survival analysis in the training cohort

TCGA cohort
variables	Univariate analysis		Multivariate analysis
	HR(95%CI)	Pvalue	HR(95%CI)	Pvalue
Age <60(55) ≥60(122)	1.404(0.892–2.209)	0.143
Gender Female(80) Male(97)	0.823(0.548–1.238)	0.35
T_stage T1/T2(31) T3/T4(144)	2.051(1.088–3.868)	0.026*	1.334(0.684–2.605)	0.398
N_stage N-(49) N+(123)	2.112(1.258–3.547)	0.005**	0.965(0.229–4.509)	0.961
History_grade G1/G2(125) G3/G4(50)	1.518(0.984–2.342)	0.059
AJCC_stage I/IIa(49) IIb/III/Ⅳ(125)	2.088(1.241–3.513)	0.006**	1.495(0.335–6.678)	0.599
Race Others(21) White(156)	1.126(0.613–2.068)	0.703
Tumor_site Body or Tail(29) Head(129) Others(19)	2.357(1.2–4.630) 2.306(0.984–5.402)	0.013* 0.054	1.894(0.912–3.930) 1.831(0.743–4.515)	0.087 0.189
Group Low risk(89) High risk(88)	2.367(1.547–3.621)	<0.0001****	1.978(1.257–3.112)	0.003**

Figure 4. Clinical stratification survival analysis

(a,b) Kaplan-Meier curves displaying the difference in PC patient survival rate in T stage.(c,d) Kaplan-Meier curves displaying the difference in PC patient survival rate in N stage.(d,f) Kaplan-Meier curves displaying the difference in PC patient survival rate in histological grade.(g,h) Kaplan-Meier curves displaying the difference in PC patient survival rate in tumor sites.

Figure 5. The landscape of somatic mutation burden between different risk groups

(a) The mutational landscape reveals the frequency of mutation events and the top 10 most frequently mutated genes in the two cohorts.(b) Kaplan-Meier curves displaying the relevance between OS and Kras mutation in each cohort.(c) Heatmap illustrating the co-occurrence and mutually exclusive mutations of the top 10 frequently mutated genes in each cohort.(d) Bar graph revealing chromosome CNV between the two cohorts.WT, wild type; MUT, mutation (·P < 0.05; * P < 0.01; ** P < 0.001).

Figure 6. The translational differences of the key genes between pancreatic cancer tissues and normal pancreatic tissues in the HPA database

Data availability statement

The datasets analyzed was acquired from The Cancer Genome Atlas (TCGA) database(https://portal.gdc.cancer.gov/), Gene Expression Omnibus (GEO) database(https://www.ncbi.nlm.nih.gov/geo/) and Biological Project Library(https://bigd.big.ac.cn/bioproject/browse/PRJCA001063).