Identification and Credentialing of Patient Derived Xenograft Models of Invasive Lobular Breast Carcinoma using Multi-omics and Histopathology assessment

Abstract

Breast cancer is a heterogeneous disease with numerous histological subtypes. Invasive lobular cancer (ILC) is the most common special subtype, accounting for 10-15% of all breast cancers. The pathognomonic feature of ILC is the loss of E-cadherin (CDH1), which leads to a unique single-file growth pattern of discohesive cells. Although ILCs show better prognostic factors than the most common No Special Type (NST) of breast cancer, patients with ILC have worse long-term outcomes, which is not well understood. In this study, we aimed to identify and characterize Patient-Derived Xenograft (PDX) models of ILC based upon the presence of truncating CDH1 mutations and/or low CDH1 mRNA expression among 128 human breast cancer PDX models. We selected 8 PDX models for validation using Immunohistochemical (IHC) analysis for E-Cadherin, p120, ER, PR, and HER2. We confirmed that seven of these PDX models are indeed ILC while one was identified as mixed NST-ILC PDX. Molecular analysis of the confirmed ILC PDX models showed enrichment of truncating CDH1 mutations, significantly lower levels of CDH1 mRNA expression and predominantly luminal subtypes compared to NST PDX models, in line with the molecular characteristics of human ILC disease. The commonly altered genes in the ILC PDX models included PIK3CA (57%), CDH1 (57%) and TP53 (57%) among others. Our study confirms and characterizes new ILC PDX models, offering valuable tools to advance our understanding of human ILC biology and support the development of innovative treatment strategies.

Keywords: CDH1; Genomic Profiling; Invasive Lobular Breast Cancer; Patient-Derived Xenograft; Targeted Therapy.

Figure 1:. Identification of potential ILC PDX models using multiomic analysis of Breast Cancer PDX models.

A) Bar plot showing PDX models with highest and lowest CDH1 mRNA level (z-score) categorized discretely as high (z-score > 0.5), medium/mid (−0.5 < z-score <= 0.5) and low (z-score <= −0.5). The low level PDX models, many of which were classified as ‘carcinoma’, were flagged as putative ILC models. B) Lollipop plot showing mutations across CDH1 gene body. C) CDH1 expression in PDX models with CDH1 mutations of various types (truncating, missense and other (intronic)) and or wild type (WT) status. All PDX models with truncating CDH1 mutations were flagged as putative ILC models.

Figure 2:. Pathology assessment of PDX models identifying 7 ILC PDX models in Champions database.

The PDX FFPE slides underwent staining with E-Cadh, p120, ER, PR, or Her2 antibodies on the Ventana Benchmark Ultra staining platform, and signal detection was conducted using Ultraview. A trained pathologist assessed the slides. A) PDXs identified as ILC. B) CTG-3434 was identified as a mixed NST ILC PDX. C) NST PDXs used for comparison.

Figure 2:. Pathology assessment of PDX models identifying 7 ILC PDX models in Champions database.

The PDX FFPE slides underwent staining with E-Cadh, p120, ER, PR, or Her2 antibodies on the Ventana Benchmark Ultra staining platform, and signal detection was conducted using Ultraview. A trained pathologist assessed the slides. A) PDXs identified as ILC. B) CTG-3434 was identified as a mixed NST ILC PDX. C) NST PDXs used for comparison.

Figure 3:. Molecular features of ILC PDX models.

A) PAM50 subtypes in ILC vs NST/IDC PDX models. Luminal subtypes were significantly enriched (Fisher’s Exact Test, p = 0.01) in ILC vs NST/IDC PDX models. B) Gene expression heatmap of top 15% variable genes split by NST and ILC PDX models. C) Frequency of various gene alterations (MUT, MUT;LOH or MUT;GAIN) in all ILC, non-basal NST and basal NST across breast cancer genes frequently altered in breast patient tumors. CDH1 alterations were significantly enriched in ILC vs non-basal NST (fisher’s exact test, p = 0.005) and vs. basal NST (fisher’s exact test, p = 0.0003). Other gene alterations were not significantly different between ILC and non-basal NST PDX models.