Prognostic significance of E-cadherin and ZEB1 expression in intraductal papillary mucinous neoplasm

Abstract

There is an urgent need to investigate the genetic changes that occur in intraductal papillary mucinous neoplasm (IPMN), which is a well-known precursor of pancreatic cancer. In this study, gene expression profiling was performed by removing unwanted variation to determine the differentially expressed genes (DEGs) associated with malignant progression of IPMN. Among the identified DEGs, zinc finger E-box binding homeobox 1 (ZEB1) and E-cadherin, a crucial regulator of epithelial-to-mesenchymal transition (EMT), was validated among identified DEGs. A total of 76 fresh-frozen tissues were used for gene expression profiling and formalin-fixed, paraffin-embedded blocks from 87 patients were obtained for immunohistochemical analysis. Loss of E-cadherin expression (p = 0.023, odd ratio [OR] = 4.923) and expression of ZEB1 in stromal cells (stromal ZEB1, p < 0.001, OR = 26.800) were significantly correlated with degree of dysplasia. The hazard of death was significantly increased in patients with loss of E-cadherin expression (hazard ratio [HR] = 13.718, p = 0.004), expression of epithelial ZEB1 (HR = 19.117, p = 0.001), and stromal ZEB1 (HR = 6.373, p = 0.043). Based on the results of this study, loss of E-cadherin and expression of stromal ZEB1 are associated with increased risk of malignant progression. Epithelial and stromal ZEB1, as well as E-cadherin may be strong predictors of survival in patients with IPMN. Our finding suggests that these EMT markers may be utilized as potential prognosticators and may be used to improve and personalize treatment of IPMN.

Keywords: ZEB1; biomarker; differentially expressed gene; epithelial-to-mesenchymal transition; intraductal papillary mucinous neoplasm.

Figure 1. Relative log expression (RLE) plots for all control genes

The RLE plots of control genes with 1-8 factors were compared to determine the appropriate number of factors that were consistent with the gene expression patterns of control genes. (A) RLE plot by the RMA normalization. (B) RLE plot normalized by RUV, number of factors (κ) = 1. (C) RLE plot normalized by RUV, κ = 2. (D) RLE plot normalized by RUV, κ = 3. (E) RLE plot normalized by RUV, κ = 4. (F) RLE plot normalized by RUV, κ = 5. (G) RLE plot normalized by RUV, κ = 6. (H) RLE plot normalized by RUV, κ = 7. (I) RLE plot normalized by RUV, κ = 8. (Red boxplots indicate low-grade dysplasia and green boxplots indicate invasive IPMNs. Lines inside the boxplots indicate median values).

Figure 2. P-value plots of control genes

(A) Unadjusted P-value plots. Some p-values of the control genes are less than 0.05. (B) P-value plots normalized by RUV, κ = 1. (C) P-value plots normalized by RUV, κ = 2. (D) P-value plots normalized by RUV, κ = 3. (E) P-value plots normalized by RUV, κ = 4. (F) P-value plots normalized by RUV, κ = 5. (G) P-value plots normalized by RUV, κ = 6. (H) P-value plots normalized by RUV, κ = 7. (Green dots indicate control genes).

Figure 3. Relative log expression (RLE) plots for all genes

(A) RLE plot by RMA normalization. (B) RLE plot normalized by RUV-4 method with number of factors (κ) = 1. (C) RLE plot normalized by RUV-4, κ = 2. (D) RLE plot normalized by RUV-4, κ = 3. (E) RLE plot normalized by RUV-4, κ = 4. (F) RLE plot normalized by RUV-4, κ = 5. (G) RLE plot normalized by RUV-4, κ = 6. (H) RLE plot normalized by RUV-4, κ = 7. (I) RLE plot normalized by RUV-4, κ = 8. (Red boxplots indicate low-grade dysplasia, green boxplots indicate invasive IPMN. Lines inside the boxplot indicate median values).

Figure 4. P-value histograms of each removing unwanted variation methodology

(A) The histogram obtained from t-tests show exponential distribution. The variance of the statistic was calculated by (B) the standard method, (C) empirical bayes, (D) rescaled variance, and (E) rescaled variance with empirical Bayes. (F) Distribution and peak of p-values when variance of the statistic was calculated by the empirical variance

Figure 5. Kaplan-Meier survival curves for disease-specific survival according to immunohistochemical markers

(A) Survival curves according to E-cadherin expression (p < 0.001). (B) Survival curves according to epithelial ZEB1 expression (p = 0.025). (C) Survival curves according to stromal ZEB1 expression (p = 0.009).

Figure 6. Receiver operating characteristic (ROC) curves of E-cadherin and stromal ZEB1

AUC for E-cadherin was 0.603 (p = 0.182), and stromal ZEB1 was 0.680 (p = 0.019).

Figure 7. Representative immunohistochemical images showing expressions of E-cadherin and ZEB1

(A) 5–50% loss of E-cadherin in IPMN with high-grade dysplasia, 224× magnification (grade 1). (B) ≥50% loss of E-cadherin in IPMN with moderate-grade dysplasia, 232× magnification (grade 0). (C) Nuclear expression of ZEB1 in the epithelial cells of invasive IPMN, 264× magnification (intensity grade 2, area grade 3, score 6). (D) Positive stromal ZEB1 expression in invasive IPMN, 264× magnification. (E) 5–50% expression of vimentin in IPMN with moderate-grade dysplasia, 264× magnification (grade 2). (F) <5% expression of vimentin in non-neoplastic lesion of IPMN with moderate-grade dysplasia, 264× magnification (grade 1).