Creation and Validation of Patient-Derived Cancer Model Using Peritoneal and Pleural Effusion in Patients with Advanced Ovarian Cancer: An Early Experience

Abstract

Background: The application of personalized cancer treatment based on genetic information and surgical samples has begun in the field of cancer medicine. However, a biopsy may be painful for patients with advanced diseases that do not qualify for surgical resection. Patient-derived xenografts (PDXs) are cancer models in which patient samples are transplanted into immunodeficient mice. PDXs are expected to be useful for personalized medicine. The aim of this study was to establish a PDX from body fluid (PDX-BF), such as peritoneal and pleural effusion samples, to provide personalized medicine without surgery. Methods: PDXs-BF were created from patients with ovarian cancer who had positive cytology findings based on peritoneal and pleural effusion samples. PDXs were also prepared from each primary tumor. The pathological findings based on immunohistochemistry were compared between the primary tumor, PDX, and PDX-BF. Further, genomic profiles and gene expression were evaluated using DNA and RNA sequencing to compare primary tumors, PDXs, and PDX-BF. Results: Among the 15 patients, PDX-BF was established for 8 patients (5 high-grade serous carcinoma, 1 carcinosarcoma, 1 low-grade serous carcinoma, and 1 clear cell carcinoma); the success rate was 53%. Histologically, PDXs-BF have features similar to those of primary tumors and PDXs. In particular, PDXs-BF had similar gene mutations and expression patterns to primary tumors and PDXs. Conclusions: PDX-BF reproduced primary tumors in terms of pathological features and genomic profiles, including gene mutation and expression. Thus, PDX-BF may be a potential alternative to surgical resection for patients with advanced disease.

Keywords: ascites; ovarian cancer; patient-derived xenograft; pleural effusion; sequence analysis.

Figure 1

Protocols used to prepare Patient-Derived Xenografts (PDXs) using the primary tumor and peritoneal and pleural effusion samples. The tissue of primary tumors was immediately washed and minced. The tumor fragments mixed with DMEM nutrient mix F-12 (DMEM/F12) and Matrigel were subcutaneously injected into immunodeficient mice (F0). The collected peritoneal and pleural effusion samples were centrifuged several times at 3000 min⁻¹ for 3 min. Isolated cancer cells with DMEM/F12 and Matrigel were subcutaneously injected into immunodeficient mice (PE0 or A0).

Figure 2

Pathological and immunohistochemical findings in primary tumors (P), patient-derived xenografts (PDXs, F0), and PDX from pleural effusion (PE0) and peritoneal effusion (A0) samples. (a,b) Similar pathological findings among P, F0, and A0 for cases 1 and 3 (i.e., serous carcinoma). The tumor cells mainly had a solid architecture. Papillary growths and a fibrous branch were often observed. The nuclei had strong atypia. The tumor cells were positive for AE1/AE3 and negative for p53. The percent Ki-67 labeling of each tumor ranged from 30% to 33%. (c) In case 5 (i.e., carcinosarcoma), the primary tumor had two malignant components: epithelial and sarcomatous components. The high-grade epithelial carcinoma component was positive for AE1/AE3, while the sarcomatous component was positive for CD10. Interestingly, similar pathological and immunohistochemical features were found between P and A0. The percent Ki-67 labeling was 26% in P and 27% in A0. (d) In case 6 (i.e., low-grade serous carcinoma), similar pathological and immunohistochemical features were found between P and A0; the tumor cells were small papillae-containing cells with uniform nuclei and inconspicuous mitotic activity. The tumor cells were positive for AE1/AE3 and negative for p53 and had 6% Ki labeling. (e) In case 7 (i.e., clear cell carcinoma), similar pathological and immunohistochemical features were found among P, F0, and PE0. The tumors exhibited a solid pattern. The solid architecture consisted of sheets of clear cells separated by delicate septa. The tumor cells were positive for AE1/AE3 and negative for p53. Ki-67 labeling was 7% in P, 7% in F0, and 6% in PE0. The scale bar indicates 100 µm.

Figure 3

Summary of the relationships between somatic mutations in primary tumors and matched xenograft tumors based on the DNA-seq results. (a) Heat map analysis of DNA profiling in primary tumors (P), patient-derived xenografts (PDXs, F0), and PDXs from body fluid, including peritoneal effusion (A0) and pleural effusion (PE0) samples, using Ion AmpliSeq Cancer Hotspot Panel v2. In all cases, xenograft tumors had almost the same mutation as the primary tumors. (b–d) Variant allele frequencies (VAFs) of somatic mutations identified in primary and xenograft tumors. Scatter plots and linear regression of the VAFs levels in the primary and xenograft tumors. This diagonal graph shows the kernel density estimation (KDE). In all cases, primary and xenograft tumors show a positive correlation.

Figure 4

Gene expression in primary tumors (P), patient-derived xenograft (PDX, F0), and PDX from body fluid (A0 or PE0) based on RNA sequencing. (a) Clustering and heat map analysis of mRNA profiling in tissues of the primary and xenograft tumors using Ion AmpliSeq RNA Cancer Panel. In most cases, the gene expression in primary and xenograft tumors exhibited a similar pattern, whereas in cases 4 and 7, PDX-BF exhibited partially different results. (b–d) Pair plot showing gene expression in primary and xenograft tumors. Scatter plots and linear regression of the gene expression in the primary and xenograft tumors. The diagonal graph shows the kernel density estimation (KDE). In all cases, primary and xenograft tumors show a positive correlation. The normalized data have been changed to base 10 logarithms and z-score.