Stratification of G2 pancreatic neuroendocrine tumors using a Ten Percent Ki-67 index cut-off based on clinicopathological and molecular analyses

Abstract

Pancreatic neuroendocrine tumors (PanNETs) often have low malignancy, but some develop distant metastasis or recurrence. This study aimed to investigate the prognostic utility of a Ki-67 index 10% cut-off using clinicopathological and multiomics analyses. Eighty-seven resected PanNETs were classified as G1, G2-low, G2-high, and G3 according to the World Health Organization classification, incorporating a 10% Ki-67 threshold. These groups comprised 29 (33%), 33 (38%), 16 (18%), and 9 (10%) patients, respectively. Comprehensive analyses evaluated tumor characteristics and mutation profiles across the groups. Significant differences in tumor size, recurrence, and overall survival were observed between the G2-low and G2-high groups. Genomic profiling of 19 samples revealed that the G3 (4 patients) and G2-high (6 patients) groups showed greater tumor mutation burdens and more driver gene mutations than the G2-low (6 patients) and G1 (3 patients) groups. Gene expression profiling revealed distinct patterns between the G3/G2-high and G2-low/G1 groups, with the oncogene SERPINA1 significantly upregulated in the G3/G2-high group. The multiomics approach identified key genes, emphasizing the significance of the Ki-67 index 10% cut-off in predicting tumor behavior. The findings support using the Ki-67 index of 10% as a critical threshold for stratifying PanNETs and guiding prognosis and treatment.

Keywords: Gene mutations; Ki-67 index; Molecular genetics; Pancreatic neuroendocrine tumors; Whole-exome sequencing.

Fig. 1

Associations between four-group classification and tumor size (A), lymph node metastasis (B), distant metastasis (C), vascular invasion (D), perineural invasion (E). Tumor size, distant metastasis, vascular invasion, perineural invasion correlated with the four-group classification (p < 0.001).

Fig. 2

Kaplan-Meier survival curves of pancreatic neuroendocrine tumors. Overall survival (A) and recurrence-free survival (B) of patients subdivided into four groups (G1, G2-low, G2-high, and G3) with Ki-67 indices of 3%, 10%, and 20% as the cutoff values. The differences between the G2-low and high groups were significant for both overall survival and recurrence-free survival (p = 0.004).

Fig. 3

Oncoprint showing the cancer-related mutations obtained by WES and CCP for 19 cases of pancreatic neuroendocrine tumors (PanNETs). The status of recurrence, mortality, grade, Ki-67, tumor mutation burden (TMB), copy number variants (CNV) size, microsatellite instability (MSI), sequencing method, and purity are shown above the mutation data. Tier 1, driver; Tier 2, likely driver; Tier 3, predicted driver; Tier 3-g, predicted driver (germline); Tier 4, variant of uncertain significance. Tier 4 includes only genes previously reported to be associated with PanNETs.

Fig. 4

Immunostaining in 2 cases of truncating mutations of DAXX and 1 case of ATRX mutation with a variant of uncertain significance. (a, b) Immunostaining showed loss of DAXX. (c) Immunostaining showed loss of ATRX.

Fig. 5

Heatmap showing hierarchical clustering of differentially expressed genes in 17 PanNETs. Cluster 1: upregulation in the group with Ki-67 indices of ≥ 10% and includes 4 oncogenes; Cluster 2: upregulation in the groups with Ki-67 indices of < 10%, including 2 oncogenes and 2 tumor suppressor genes; and Cluster 3: downregulation mainly in the groups with Ki-67 indices of ≥ 10% and upregulation in some of the groups with Ki-67 indices of < 10%, including a tumor suppressor gene and 7 genes expressed in normal pancreatic islets.

Fig. 6

RT-PCR validation of mRNA expression levels obtained by microarray. Delta Ct values (ΔCt), with lower values indicating higher expression, were shown for 4 oncogenes upregulated in the groups with Ki-67 indices of ≥ 10% (G3 and G2-high, red circles) and 3 tumor suppressor genes downregulated in the groups with Ki-67 indices of < 10% (G2-low and G1, dark blue circles). For samples producing no amplicons after 40 cycles of PCR, the Ct values were considered as 40 to calculate the ΔCt values.