Mutations in NOTCH pathway genes in OC were studied using the cBio Cancer Genomics Portal (http://cbioportal.org) [21, 22, 23]. We performed visualization and analysis queries based on references [23], with the studied proteins being DLL1 [ENSG00000198719], DLL3 [ENSG00000090932], DLL4 [ENSG00000128917], HES1 [ENSG00000114315], HEY1 [ENSG00000164683], JAG1 [ENSG00000101384], JAG2 [ENSG00000184916], NOTCH1 [ENSG00000148400], NOTCH2 [ENSG00000134250], NOTCH3 [ENSG00000074181], and NOTCH4 [ENSG00000204301].

R (version 3.6.3, statistical analysis and visualization, R Core Team, Vienna, Austria) and ggplot2 (version 3.3.3, for visualization, R Core Team, Vienna, Austria) were used [24]. Using the Toil process, UCSC XENA (https://xenabrowser.net/datapages/) RNAseq data was uniformly processed into TPM format for TCGA and GTEx [22, 24, 25, 26, 27, 28]. Data for OC was extracted from TCGA, while GTEx was used for the corresponding normal tissues [24]. RNAseq data were converted into TPM (transcripts per million reads) format by log2 transformation for the comparison of gene expression between samples [24].

Correlations between all pairs of genes in the NOTCH pathway were assessed using Pearson’s correlation coefficient [24]. R (version 3.6.3) and ggplot2 (version 3.3.3) were used for the analysis [24].

R (version 3.6.3) and Basic R package were used for this analysis [24]. The grouping condition was the median.

R (version 3.6.3), survminer package (version 0.4.9, for visualization, R Core Team, Vienna, Austria), and survival package (version 3.2-10, R Core Team, Vienna, Austria) were used for the statistical analysis of survival data [24]. The two subgroups studied were 0–50 and 50–100 [24]. Overall survival (OS) and progression-free interval (PFI) [24] were used for prognostic evaluation. Prognostic information for the controls was provided as supplemental data [20].

R (version 3.6.3) and survival package (version 3.2-10) were used for regression analysis [24]. Cox regression was used as the statistical method [29, 30]. PFI was used for the type of prognosis. Included variables were FIGO stage, primary therapy outcome, race, age, histologic grade, anatomic neoplasm subdivision, venous invasion, lymphatic invasion, tumor residual, tumor status, and 11 NOTCH pathway genes [23]. Prognostic information from the reference was provided as supplemental data [31].

R (version 3.6.3) and ggplot2 package (version 3.3.3) were used for Forest plots [24, 30].

R (version 3.6.3), pROC package (version 1.17.0.1, R Core Team, Vienna, Austria) and ggplot2 package [version 3.3.3] were used for analysis [24]. The clinical variables were tumor and normal [24].

R (version 3.6.3) and stat base package (version 3.6.3) were used for this analysis [24].

R (version 3.6.3), ggplot2 package (version 3.3.3) and clusterProfiler package (version 3.14.3, R Core Team, Vienna, Austria) were used [24].

R (version 3.6.3) and GSVA package (version 1.34.0, R Core Team, Vienna, Austria) were used for this analysis [24, 31]. The immune-infiltration algorithm used was ssGSEA (built-in algorithm in the GSVA package) [24]. The immune cell types studied here include aDC (activated DC), B cells, CD8 T cells, cytotoxic cells, DC, eosinophils, iDC (immature DC), macrophages, mast cells, neutrophils, NK CD56_bright cells, NK CD56_dim cells, NK cells, pDC (Plasmacytoid DC), T cells, T helper cells, Tcm (T central memory), Tem (T effector memory), Tfh (T follicular helper), Tgd (T gamma delta), Th1 cells, Th17 cells, Th2 cells, and Tregs [23, 32]. Markers for the 24 immune cell types were obtained from a previous reference [33].

The online tool cBioPortal was used to examine NOTCH pathway gene alterations in OC. As shown in Fig. 1, the frequency of genetic alterations in OC was: DLL1, 3%; DLL3, 10%; DLL4, 2.8%; HES1, 19%; HEY1, 5%; JAG1, 4%; JAG2, 5%; NOTCH1, 4%; NOTCH2, 8%, NOTCH3, 14%; and NOTCH4, 5%. As shown in Fig. 2, the frequency of genetic alterations in NOTCH pathway genes in OC patients from TCGA, Firehose Legacy were: mutation, 1.2%; amplification, 52.05%; deep deletion, 6.85%; and multiple alterations, 1.71%. For TCGA, Pan-Cancer Atlas: mutation, 3.25%; structural variant, 0.34%; amplification, 37.84%; deep deletion, 5.48%; and multiple alterations 2.91%. For TCGA, Nature 2011: mutation, 1.23%; amplification, 33.74%; deep deletion, 3.68%; and multiple alterations, 1.43%. In general, amplifications accounted for most of the genetic alterations detected in OC.

Fig. 1.

Genetic alterations of NOTCH pathway genes in ovarian cancer (OC) as observed using cBioPortal (RNA Seq V2 RSEM).

Fig. 2.

The distribution of genetic alterations in NOTCH pathway genes in OC cases based on the cancer type summary in cBioPortal.

A total of 427 OC tumors and 88 samples of normal ovarian tissue were used to assess NOTCH pathway gene expression (Fig. 3). The expression levels of DLL1, DLL4, HEY1, JAG1 and NOTCH4 were all significantly lower in OC samples than in normal ovarian tissues (1.064 $\pm$ 0.832 vs. 1.991 $\pm$ 0.827; 1.688 $\pm$ 0.631 vs. 2.206 $\pm$ 0.878; 1.33 $\pm$ 0.876 vs. 1.51 $\pm$ 0.632; 3.339 $\pm$ 1.027 vs. 4.916 $\pm$ 0.835; and 2.113 $\pm$ 0.568 vs. 3.514 $\pm$ 0.689, respectively; all p 0.001). In contrast, the expression levels of DLL3, HES1, JAG2, NOTCH1 and NOTCH3 were all significantly higher in OC tumor samples (1.075 $\pm$ 0.884 vs. 0.25 $\pm$ 0.246; 4.9 $\pm$ 1.119 vs. 4.383 $\pm$ 1.168; 3.121 $\pm$ 0.929 vs. 2.284 $\pm$ 0.731; 2.82 $\pm$ 1.063 vs. 2.487 $\pm$ 0.676; and 5.086 $\pm$ 1.253 vs. 3.933 $\pm$ 1.353, respectively; all p 0.001 except NOTCH1, p = 0.002). No significant difference was seen in the level of NOTCH2 expression between OC tumor tissues and healthy ovarian tissues.

Fig. 3.

TCGA mRNA levels for the NOTCH pathway genes in OC tissue and in unpaired normal ovarian tissue. ns, not significant (p $\ge$ 0.05); **, p 0.01; ***, p 0.001.

Pearson correlation analysis was used to test for relationships between genes involved in the NOTCH pathway. A significant negative correlation was found between the expression of DLL3 and NOTCH2, while the expression levels of most other NOTCH pathway genes showed a significant positive correlation with each other (Fig. 4).

Fig. 4.

Correlations between the expression level of genes in the NOTCH pathway.

The clinical characteristics and gene expression data for 379 OC tumor samples were downloaded from the TCGA database (Supplementary Table 1). DLL4 and NOTCH4 expression were associated with FIGO stage (p 0.001 and p = 0.036, respectively), while NOTCH3 expression was associated with race (p = 0.039) and age (p = 0.044). The expression levels of other NOTCH pathway genes (DLL1, DLL3, HES1, HEY1, JAG1, JAG2, NOTCH1, and NOTCH2) did not correlate significantly with the clinical data of OC patients.

The expression level of 11 NOTCH pathway genes (DLL1, DLL3, DLL4, HES1, HEY1, JAG1, JAG2, NOTCH1, NOTCH2, NOTCH3, and NOTCH4) showed no significant correlation with the overall survival of OC patients (Fig. 5). The expression of JAG2 (p = 0.011) and NOTCH1 (p = 0.039) were found to be significantly associated with the PFI of OC patients, whereas the remaining 9 genes (DLL1, DLL3, DLL4, HES1, HEY1, JAG1, NOTCH2, NOTCH3 and NOTCH4) showed no significant correlation with the PFI (Fig. 6). Higher expression of JAG2 and NOTCH1 were correlated with significantly better PFI.

Fig. 5.

Overall survival (OS) analysis for NOTCH pathway gene expression in OC patients using Kaplan Meir plot web tool. (A) DLL1, (B) DLL3, (C) DLL4, (D) HES1, (E) HEY1, (F) JAG1, (G) JAG2, (H) NOTCH1, (I) NOTCH2, (J) NOTCH3, and (K) NOTCH4.

Fig. 6.

Progression-free interval (PFI) analysis for NOTCH pathway gene expression in OC patients using Kaplan Meir plot web tool. (A) DLL1, (B) DLL3, (C) DLL4, (D) HES1, (E) HEY1, (F) JAG1, (G) JAG2, (H) NOTCH1, (I) NOTCH2, (J) NOTCH3, and (K) NOTCH4.

Univariate Cox regression analysis (Fig. 7A) revealed that primary therapy outcome (HR: 0.401; 95% CI: 0.304–0.528, p 0.001), JAG2 expression (HR: 0.735; 95% CI: 0.580–0.932, p = 0.011), and NOTCH1 expression (HR: 0.779; 95% CI: 0.614–0.987, p = 0.039) were predictors of good PFI in OC patients. In contrast, tumor residual (HR: 1.695; 95% CI: 1.219–2.358, p = 0.002) and tumor status (HR: 10.045; 95% CI: 5.758–17.526, p 0.001) were predictors of worse PFI in OC patients. Multivariate analysis revealed that primary therapy outcome (HR: 0.613; 95% CI: 0.451–0.832, p = 0.002), tumor status (HR: 14.164; 95% CI: 6.909–29.039, p 0.001), DLL1 expression (HR: 2.096; 95% CI: 1.522–2.886, p 0.001), and NOTCH1 expression (HR: 0.711; 95% CI: 0.514–0.983, p = 0.039) were independent predictors of PFI.

Fig. 7.

Forest plots showing the results of Cox regression analysis for NOTCH pathway genes and other clinical features as predictors of progression free interval (PFI) in OC patients. (A) Univariate regression analysis. (B) Multivariate regression analysis.

As shown in Fig. 8, the expression of several NOTCH pathway genes showed some tendency for differentiating normal and OC tumor tissues (DLL1: AUC = 0.823, CI = 0.785–0.861; DLL3: AUC = 0.839, CI = 0.801–0.876; JAG1: AUC = 0.883, CI = 0.849–0.916; JAG2: AUC = 0.769, CI = 0.718–0.820; NOTCH3: AUC = 0.739, CI = 0.679–0.799). Other genes showed low tendency (DLL4: AUC = 0.668, CI = 0.599–0.737; HES1: AUC = 0.642, CI = 0.576–0.709; HEY1: AUC = 0.621, CI = 0.560–0.683; NOTCH1: AUC = 0.603, CI = 0.549–0.657; NOTCH2: AUC = 0.532, CI = 0.482–0.583), while high accuracy was shown by NOTCH4 (AUC = 0.948, CI = 0.928–0.968).

Fig. 8.

ROC curves for NOTCH pathway genes in OC and normal ovarian tissues. The area under the curve (AUC) ranged from 0.5 to 1. When the AUC is closer to 1, diagnosis is more accurate for the differentiation of OC. An AUC between 0.5 and 0.7 has low accuracy, an AUC between 0.7 and 0.9 has some accuracy, while an AUC above 0.9 has high accuracy.

The top five negatively associated genes with each NOTCH pathway gene are shown in Fig. 9, together with the top 5 positively associated genes. The genes most positively associated with DLL1 were ADAMTS7, DENND2A, ZNF697, RHOBTB3 and DCHS1, while those most negatively associated with this gene were RTP4, EVA1C, EIF4EBP3, PDZK1IP1, and PLAAT4. The genes most positively associated with DLL3 were SELENOV, TIMM50, ECH1, RYR1 and MRPS12, while those most negatively associated were VNN1, MVP, KCTD12, SERPINB1, and TMEM87A. The genes most positively associated with DLL4 were NOTCH4, CDH5, VWF, PCDH12 and ROBO4, while those most negatively associated were RPL39, HIGD2A, ATP5MD, RPL36A, and RPS27. The genes most positively associated with HES1 were SMOX, DACT2, JAG1, ITM2C, and PDE9A, while those most negatively associated were HPSE, LTC4S, NPR1, CYP4B1, and EVA1C. The genes most positively associated with HEY1 were PGF, PEAR1, TSPAN11, HSPA12B, and SIX2, while those most negatively associated were KLK7, PDZK1IP1, C19orf33, FOLR1, and MSLN. The genes most positively associated with JAG1 were NOTCH1, BTBD3, NOTCH3, ATRN, and CDS2, while those most negatively associated were COX5B, ATP5MC2, LAMTOR4, RPL32, NDUFA1. The genes most positively associated with JAG2 were CEP170B, BRF1, AKT1, PACS2, and PLCB3, while those most negatively associated were OST4, COPS9, BLOC1S1, COX7C, and PFDN5. The genes most positively associated with NOTCH1 were SEC16A, CAMSAP1, PLXNA1, RXRA and PRRC2B, while those most negatively associated were ATP5MF, UQCRQ, NDUFA1, FAU and COX5B. The genes most positively associated with NOTCH2 were CTDSP2, HIPK1, GIT2, ATXN1L and BAZ2A, while those most negatively associated were RNF181, ATP5MD, COX5B, ATP5F1E and RPS21. The genes most positively associated with NOTCH3 were BRD4, WIZ, MYO9B, TET3 and CEP250, while those most negatively associated were VAMP8, UQCR11, PLAAT4, COX7B, and NDUFA1. The genes most positively associated with NOTCH4 were CDH5, ROBO4, PCDH12, DIPK2B, and DLL4, while those most negatively associated were RPL12, RPL36A, RPL31, RPS21 and RPS27.

Fig. 9.

Heatmap showing the top 10 genes that were correlated with NOTCH pathway genes. (A) DLL1, (B) DLL3, (C) DLL4, (D) HES1, (E) HEY1, (F) JAG1, (G) JAG2, (H) NOTCH1, (I) NOTCH2, (J) NOTCH3, and (K) NOTCH4.

GO and KEGG enrichment analyses were conducted on the NOTCH pathway and related genes (102 genes in total) (Supplementary Table 2). As shown in Fig. 10 and Supplementary Table 3, the top five biological processes identified were the regulation of NOTCH signaling pathway, NOTCH signaling pathway, negative regulation of NOTCH signaling pathway, NOTCH signaling involved in heart development, and positive regulation of NOTCH signaling pathway. The top five cytological components were the mitochondrial inner membrane, polysomal ribosome, cytosolic ribosome, mitochondrial proton-transporting ATP synthase complex, and cytosolic part. The top five molecular functions were NOTCH binding, structural constituent of ribosome, electron transfer activity, cytochrome-c oxidase activity, and heme-copper terminal oxidase activity. As shown in Fig. 11 and Supplementary Table 3, the top five pathways identified were the NOTCH signaling pathway, breast cancer, endocrine resistance, Th1 and Th2 cell differentiation, and oxidative phosphorylation.

Fig. 10.

GO analysis of genes involved in the NOTCH pathway.

Fig. 11.

KEGG analysis of genes associated with NOTCH pathway genes.

As shown in Fig. 12, NOTCH pathway gene expression was correlated with tumor-infiltrating immune cells (TIIC) in OC. DLL1 gene expression correlated positively with eosinophils, iDC, macrophages, mast cells, NK cells, Tcm, Tem, TFH and Tgd, and negatively with aDC, cytotoxic cells, T cells, Th1 cells, and Th17 cells. DLL3 expression correlated positively with Th2 cells, and negatively with CD8 T cells, cytotoxic cells, DC, iDC, macrophages, neutrophils, NK CD56bright cells, NK CD56dim cells, pDC, T cells, Tcm, Tem, Th1 cells and Th17 cells. DLL4 expression correlated positively with eosinophils, iDC, macrophages, mast cells, neutrophils, NK cells, pDC, T helper cells, Tcm, Tem, TFH and Tgd, and negatively with cytotoxic cells. HES1 expression correlated positively with eosinophils, NK cells and Th2 cells, and negatively with aDC, B cells, cytotoxic cells, DC, neutrophils, T cells, Tem, Th1 cells, Th17 cells and Tregs. HEY1 expression correlated positively with eosinophils, iDC, macrophages, mast cells, neutrophils, NK cells, T helper cells, Tem, TFH, Tgd and Th2 cells, and negatively with cytotoxic cells, NK CD56bright cells and Th17 cells. JAG1 expression correlated positively with eosinophils, iDC, NK cells, T helper cells, Tcm, Tem, TFH and Th2 cells, and negatively with cytotoxic cells. JAG2 expression correlated positively with eosinophils, NK cells, T helper cells, Tcm and Tem. NOTCH1 expression correlated positively with DC, eosinophils, iDC, macrophages, mast cells, neutrophils, NK CD56dim cells, NK cells, T helper cells, Tcm, Tem, TFH and Th2 cells, and negatively with cytotoxic cells. NOTCH2 expression correlated positively with DC, eosinophils, iDC, macrophages, mast cells, neutrophils, NK cells, T helper cells, Tcm, Tem, TFH, Tgd and Th17 cells. NOTCH3 expression correlated positively with eosinophils, NK cells, Tcm and Tem, and negatively with aDC, CD8 T cells, cytotoxic cells, pDC, T cells, Th1 cells, Th17 cells, and T regs. NOTCH4 gene expression correlated positively with eosinophils, iDC, macrophages, mast cells, neutrophils, NK cells, pDC, T helper cells, Tcm, Tem, TFH and Tgd.

Fig. 12.

Analysis of the correlation between expression of each NOTCH family gene and 24 types of TIICs in OC. (A) DLL1, (B) DLL3, (C) DLL4, (D) HES1, (E) HEY1, (F) JAG1, (G) JAG2, (H) NOTCH1, (I) NOTCH2, (J) NOTCH3, and (K) NOTCH4. Results are shown in the form of lollipop plots. The darker the color, the smaller the p-value and hence greater statistical significance. Correlations are represented by bubble sizes, with larger bubbles indicating stronger correlation.