Orthotopic PDX models 09:00 - 10:30

Introduction

Breast cancer is the most frequently diagnosed malignancy worldwide. The risk of developing breast cancer is related to the reproductive history, namely the exposure to steroid hormones that women experience during their life. However, little is known about the roles that hormones play in the human breast epithelium, partially because of the lack of preclinical models to study them. Patient-Derived Xenografts (PDXs) are the models that best recapitulate the complexity of human tissues, however their usage is challenged by the presence of both human and murine cells, which can lead to data misinterpretation.

Material and methods

Breast tissues were obtained from women undergoing reduction mammoplasties with no previous history of breast cancer and examined by a pathologist to be free of malignancy. Breast specimens were processed to single cells, infected using GFP-luc2-expressing lentiviruses, injected to the mammary ducts of immunosuppressed mice and exposed to different hormones. The in vivo growth of human breast epithelial cells (HBECs) was followed by bioluminescence measurements. Xenograft-bearing mouse mammary gland were fixed in 4% paraformaldehyde and paraffin-embedded for histological analysis. Paraffin blocks were cut and 4 µm-thick sections were stained with hematoxylin and eosin (H&E). Slides were scanned with a slide scanner using a 20x/0.75 objective connected to a color camera.

Results and discussion

We show that primary human breast epithelial cells establish themselves in the mouse milk ducts and grow there without hormone supplementation while retaining their hormone receptor expression and hormone responsiveness. Using the Mouse INtraDuctal (MIND) approach, we derived xenografts of normal breast epithelial cells from 40 women and show that different contraceptives have distinct biological activities in the breast epithelium depending on their androgenic properties. To automatically quantify the amount of xenografted cells in H&E-stained histological sections, we developed Single Cell Classifier (SCC), a machine learning-based approach that reaches up to 96% of accuracy in discriminating normal or malignant human cells from cells of the murine host.

Conclusion

Coupling the MIND model and SCC provides a robust framework for the study of the role that steroid hormones play in the normal breast epithelium, with potential in breast cancer prevention.

Stefan Hutten¹, Roebi de Bruijn¹, Catrin Lutz¹, Madelon Badoux¹, Timo Eijkman¹, Xue Chao¹, Esther Lips¹, Jelle Wesseling¹, Colinda Scheele² and Jos Jonkers¹

¹The Netherlands Cancer Institute, Amsterdam, The Netherlands

²Center for Cancer Biology (VIB), Leuven, Belgium

Introduction

Ductal Carcinoma in Situ (DCIS) is a non-invasive non-obligate precursor of invasive breast cancer (IBC). DCIS is usually treated by surgery combined with radiotherapy, which can have a large impact on the life of patients. However, many of these DCIS lesions would never progress into IBC. To reduce the overtreatment of DCIS, but assure proper treatment for high risk DCIS, it is crucial to understand the biology underlying DCIS. Unfortunately to this day there is a lack of models to study DCIS. The same holds true for estrogen receptor positive (ER) IBC even though the vast majority of breast cancers ER+ .

Material and Methods

Therefore we established Mouse INtraDuctal (MIND) patient-derived xenograft (PDX)models by intraductally injecting patient DCIS or IBC material into the mammary ducts of female immunocompromised mice. We engrafted 130 DCIS samples as well as 314 IBC samples, which have been incubated in vivo for a period of 12 months.

Results and Discussion

For our DCIS models we obtain a take rate of 89% with 45% of our models showing progression. Histology and molecular subtyping by PAM50 classification are well preserved in the MIND models compared to the primary counterpart, ensuring that our MIND models represent the patient disease well. For 102 primary samples we obtained RNAseq profiles as well as for 64 matched MIND-PDX models. In addition Whole exome- /panel sequencing data is generated from the same primary DCIS samples together with 12 matched MIND-PDX WES profiles as well as 60 matched Copy Number Variation (CNV) MIND-PDX profiles.

Together these data revealed multiple biomarkers related to invasive progression, including factors such as high grade, solid growth, a high copy number aberrations burden, HER2, PTK6 & MYC amplifications and a high Ki67. On top of this we used whole mount imaging of the injected mammary glands extracted from our MIND-PDX models, showing two distinct growth patterns correlated with invasion.

We have also successfully passaged 42 MIND-PDX models which showed minimal changes in pheno- and genotype over time indicating invasive behavior is an intrinsic phenotype of DCIS with minimal evolution, supporting a multiclonal evolution model. Moreover, this provided a collection of 19 stable sequentially transplantable DCIS MIND models including Luminal A, Luminal B, ER+/HER2+ and ER-/HER2+ models.

For our ER+ IBC models we obtain a take rate of 58% with around 20% transplantable models resulting in 35 ER+ models. This greatly exceeds the take rates previously found for ER+ IBC models, which could be as low as 2%.

Aniello Federico^1,2, Apurva Gopisetty^1,2, Justyna Wierzbinska³, Norman Mack^1,2, Benjamin Schwalm^1,2, Natalie Jäger^1,2, Jan Koster⁴, Gudrun Schleiermacher⁵, Stefan M Pfister^1,2, Marcel Kool^1,2,6

¹ Hopp Children’s Cancer Center (KiTZ), Heidelberg, Germany

² German Cancer Research Center (DKFZ) and German Cancer Consortium (DKTK), Heidelberg, Germany

³ Bayer AG, Pharmaceuticals, Research and Development, Berlin, Germany

⁴ Department of Oncogenomics, Amsterdam University Medical Centre, Amsterdam, the Netherlands

⁵ INSERM U830, Équipe Labellisée LNCC, Genetics and Biology of Pediatric Cancers, PSL Research University, SIREDO Oncology Centre, Institut Curie Research Centre, Paris, France

⁶ Princess Máxima Center for Pediatric Oncology, Utrecht, the Netherlands

Introduction

Pediatric malignancies exhibit some peculiarities, such as high intra- and intertumoral heterogeneity, low tumor mutational burden and low neoantigen load, that limit the effectiveness of the current therapeutic options. For this reason, increasing research efforts aiming at the definition of suitable preclinical models and an improvement of therapeutic strategies are needed. The main objectives of the “Pediatric Preclinical Proof of Concept Platform” (ITCC-P4) project are: a) to develop a large-scale platform comprising >400 preclinical models, including patient-derived xenograft models (PDXs) covering the most common pediatric high-risk entities; b) to investigate the biology of pediatric cancer in a robust and systematic manner; c) to identify suitable biomarkers for future clinical stratification of patients across entities; d) to support a proof-of-concept preclinical in vivo drug testing.

Material and Methods

We performed a comprehensive molecular characterization of 251 PDXs, representing both brain (n=69) and non-brain (n=182) models. Molecular characterization included DNA methylation profiling, lcWGS/WES sequencing, as well as RNAseq and Affymetrix gene expression profiling of all the PDX models. Matching human material (primary tumor, n=203/251; germline sample, n=145/251) has been characterized as well.

Results and Discussion

The Integration of multi-omic analysis allowed a more in-depth characterization of the molecular features exhibited in the whole PDX cohort. PDX models showed high reliability when reproducing the heterogeneity of the pediatric tumors and overall strong molecular fidelity compared to the original human tumor. The main source of molecular discrepancy between paired human tumor and PDX samples was represented by a different degree of tumor purity and stromal composition, observing a frequent impairment of the stromal components and a resulting higher tumor purity in the PDX samples. Bulk and single-cell transcriptomic analysis showed the downregulation, in the PDX samples, of genes linked with the human stroma; on the other hand, the gene expression analysis of the murine tissue found in the PDXs indicated an activation of the stromal signatures, with an increase of several functional stromal markers, such as C1qa, C1qb, Cd83, Fgd2, Itgam, Pecam1 and Cd34. Our data suggested that in these models the human stromal component is lost and functionally replaced by the surrounding murine host.

Conclusion

The PDX models established within the ITCC-P4 platform represent a valid tool for preclinical research, to investigate the oncologic therapy response, drug delivery and tumor-host interactions.

Anaïs Oudin^1*, Pilar M. Moreno-Sanchez^1,2*, Yahaya A. Yabo^1,2, Eliane Klein¹, Virginie Baus¹, Aurélie Poli³, Simone P. Niclou¹, Anna Golebiewska¹

¹ NORLUX Neuro-Oncology Laboratory, Department of Cancer Research, Luxembourg Institute of Health, Luxembourg, Luxembourg;

² Faculty of Science, Technology and Medicine, University of Luxembourg, L-4367 Belvaux, Luxembourg;

³ Neuro-Immunology Group, Department of Cancer Research, Luxembourg Institute of Health, L-1526 Luxembourg, Luxembourg

Introduction

Glioblastoma (GBM) is the most deadly primary brain tumor despite a very aggressive treatment, which consists of surgical resection followed by radiotherapy and chemotherapy. Most GBM clinical trials fail due to inappropriate selection of the compounds at the preclinical stage. Therefore, appropriate preclinical models are crucial for achieving better treatment outcomes. GBM patient-derived othotopic xenografts (PDOXs) faithfully recapitulate the molecular and histopathological profiles of patient tumors. Orthotopic implantation allows for analogous tumor microenvironment and drug distribution though the blood brain barrier. However, current xenograft models suffer from the reduction of immune system components, which represents a bottleneck for adequate immunotherapy testing. Recapitulation of the standard-of-care regimen is challenging and it remains to be seen whether procedures combining surgical resections with clinically-relevant treatment protocols are practicable.

Material and Methods

We regularly derive GBM PDOXs by intracortical implantation of primary patient-derived organoids in immunodeficient mice (NSG, Nude). To introduce the human adaptive immune system, GBM organoids were implanted in the brains of the human CD34+ hematopoietic stem cell-engrafted NSG (HU-CD34+) mice. Tumor growth was followed by Magnetic Resonance Imaging. Detailed characterization of the tumor microenvironment was performed by single cell transcriptomics, immunohistochemistry and multicolor flow cytometry. To mimic tumor recurrence, we performed surgery and tumor resection at the early tumor development stage. Robust in vivo treatment protocols were established for standard-of-care, novel drugs and combinatory treatments.

Results and Discussion

We show that PDOX models recapitulate major components of the tumor microenvironment found in human GBM, including microglia, astrocytes, oligodendrocytes and endothelial cells. Mouse cells reciprocally crosstalk with human GBM cells and switch towards GBM-instructed phenotypes. Missing adaptive immune system can be introduced by tumor growth in HU-CD34+ mice. HU-CD34+ mice present more than 50% of human immune cells in the bone marrow and blood. We observed an influx of human immune cells to the tumors developed in the brain. We further show that surgical resection can be applied on mouse brain and GBM tumors regrow after surgery similarly to patient GBM, leading to delayed survival time. Magnetic Resonance Imaging allows for robust assessment of the tumor growth and drug responses in vivo.

Conclusion

Our data provide insights into the recapitulation of the heterogeneous tumor microenvironment instructed by GBM cells in PDOXs. This work further confirms the wide applicability of GBM PDOXs for recapitulation of standard-of-care protocols and testing of novel treatments including immunotherapeutics.