Tissue microarrays of human tumor xenografts: characterization of proteins involved in migration and angiogenesis for applications in the development of targeted anticancer agents

Abstract

As new target-directed anticancer agents emerge, preclinical efficacy studies need to integrate target-driven model systems. This approach to drug development requires rapid and reliable characterization of the new targets in established tumor models, such as xenografts and cell lines. Here, we have applied tissue microarray technology to patient-derived, re-growable human tumor xenografts. We have profiled the expression of five proteins involved in cell migration and/or angiogenesis: vascular endothelial growth factor (VEGF), matrix metalloproteinase 1 (MMP1), protease activated receptor (PAR1), cathepsin B, and beta1 integrin in a panel of over 150 tumors and compared their expression levels to available patient outcome data. For each protein, several target overexpressing xenografts were identified. They represent a subset of tumor models prone to respond to specific inhibitors and are available for future preclinical efficacy trials. In a "proof of concept" experiment, we have employed tissue microarrays to select in vivo models for therapy and for the analysis of molecular changes occurring after treatment with the anti-VEGF antibody HuMV833 and gemcitabine. Whereas the less angiogenic pancreatic cancer PAXF736 model proved to be resistant, the highly vascularized PAXF546 xenograft responded to therapy. Parallel analysis of arrayed biopsies from the different treatment groups revealed a down-regulation of Ki-67 and VEGF, an altered tissue morphology, and a decreased vessel density. Our results demonstrate the multiple advantages of xenograft tissue microarrays for preclinical drug development.

Figure 1

Imunohistochemical staining of arrayed tumor biopsies. Cathepsin B showed a strong staining in the head and neck cancer HNXF908, especially along fibrous strands (score: 2.33±0.24). Lower levels of the protease were detected in renal cell cancer RXF393 (1.0±0.96). In the latter, granular expression was seen consistent with lysosomal localization. β1 integrin was localized on cell membranes as is shown for the pancreatic cancer PAXF546 (2.25±0.34), and the cervical cancer CEXF610 (3.0±0.00). VEGF showed a cytoplasmic distribution pattern, as evident in pleural mesothelioma PXF1652 (1.67±0.00) and in the breast cancer MAXF1691 (2.75±0.94). Cytoplasmic staining (green, nuclei blue) was observed for MMP1 (LXF 737, 2.5±0.35 and PAXF546, 2.75±0.35) as well as PAR1 (CXF647, 2.0±0.71 and LXFL625, 2.0±0.71).

Figure 2

Mean staining intensities of cathepsin B (A), MMP1 (B), and β1 integrin (C); (D) Cumulative Kaplan Meier survival curve of 91 patients with different solid tumors. Patients with β1 integrin overexpression had a significantly worse prognosis than patients without (p<0.001). VEGF (E) and PAR1 (F) levels sorted by cancer type. The error bars represent the standard deviation.

Figure 3

Chemotherapy of pancreatic cancer xenografts PAXF546 with high VEGF (A) and PAXF736 with low VEGF (B). In both tumors, the combination of HumV833 and gemcitabine was the most potent. Sensitivity of PAXF546 was higher than that of PAXF736, which correlated with VEGF levels and vessel density.

Figure 4

PAXF546 showed high levels of VEGF (A) and a homogenous, nuclear expression of Ki-67 (B). In HuMV833 and HuMV833 + Gemcitabine treated pancreatic cancer PAXF546 (C and D respectively), VEGF expression was substantially lower compared to controls. Photos were taken from the same array slide as in Figure 4A. In HuMV833 + Gemcitabine treated pancreatic cancer PAXF 546 stained for Ki67, occurrence of vital cells was limited to small areas surrounding blood vessels (→) They were lined by cells with hydropic swelling (asterisk) and necrotic tissue (x). (E) magnification ×160, (F) Magnification ×400.