Autologous humanized PDX modeling for immuno-oncology recapitulates the human tumor microenvironment

Michael Chiorazzi, Jan Martínek, Bradley A. Krasnick, Yunjiang Zheng, Keenan J. Robbins, Rihao Qu, Gabriel Kaufmann, Zachary Skidmore, Laura A Henze, Frederic Brösecke, Adam Adonyi, Jun Zhao, Liang Shan, Esen Sefik, Jacqueline L. Mudd, Ye Bi, S. Peter Goedegebuure, Malachi Griffith, Obi L. Griffith, Abimbola Oyedeji, Sofia Fertuzinhos, Rolando García-Milian, Daniel J. Boffa, Frank C. Detterbeck, Andrew P. Dhanasopon, Justin D. Blasberg, Benjamin L. Judson, Scott Gettinger, Katerina Politi, Yuval Kluger, Karolina Palucka, Ryan C. Fields, Richard A. FlavellbioRxiv 2022 · DOI

TL;DR

This paper introduces a novel autologous humanized mouse model that accurately recapitulates the human tumor microenvironment (TME) and its interactions with the immune system. This system allows for more precise preclinical testing of cancer therapies by overcoming limitations of traditional models, demonstrating its utility in identifying VEGF-A as a key driver of tumor growth in this context.

What Problem Does This Paper Address?

Predicting how well cancer patients will respond to immunotherapies is a major challenge. Current preclinical models often fall short because they don't fully capture the complex interplay between a patient's unique immune system and their tumor. This is partly due to "inter-species differences" between humans and mice, and also because of the vast "inter-person variation" in human immune systems. Existing "humanized" mouse models, which incorporate human immune cells, often use immune cells derived from fetal or neonatal sources, which don't accurately reflect the immune makeup of adult cancer patients. This limits their ability to model the TME and predict treatment responses for the majority of patients. The authors aim to bridge this gap by creating a model that uses an individual patient's own immune cells and tumor, providing a genetically matched and autologous system.

Methods

The core innovation is the development of an "autologous humanized PDX model." This involves:

  1. Humanizing MISTRG6 mice: These mice are genetically engineered to have human genes for key immune-regulating cytokines (M-CSF, CSF2/IL3, SIRPA, THPO) on an immunodeficient background (RAG2-/-, IL2RG-/-). This allows them to develop a more human-like immune system.
  2. Engrafting patient-derived hematopoietic stem and progenitor cells (HSPCs): Bone marrow (BM) aspirates from adult cancer patients are used to isolate CD34+ HSPCs. These are then transplanted into the MISTRG6 mice, allowing the mice to develop a human immune system that is specific to that patient.
  3. Implanting patient-derived xenograft (PDX) tissue: Tumor tissue from the same patient is then implanted into these humanized mice. This creates a "genetically matched autologous model" where the mouse harbors both the patient's immune system and their tumor.
  4. Single-cell transcriptomics: This technique was used to analyze the gene expression profiles of individual cells from both the blood and the tumor microenvironment (TME) of these models, providing a detailed understanding of cell types and their functions.
  5. Ingenuity Pathway Analysis (IPA): Used to identify overrepresented biological pathways and upstream regulators in the TME compared to blood.
  6. Bevacizumab treatment: A targeted therapy was used to block VEGF-A and assess its impact on tumor growth.

Datasets

The study utilized patient samples, including:

  • Bone marrow (BM) aspirates for HSPC isolation.
  • Peripheral blood samples.
  • Tumor tissue for PDX generation.

These samples were collected from patients with melanoma, non-small cell lung cancer (NSCLC), pancreatic adenocarcinoma (PDAC), and head and neck squamous cell carcinoma (HNSCC).

Key Insights and Ideas

  • The MISTRG6 strain is a superior host for adult human HSPCs: This strain, with a humanized IL6 locus, significantly improved hematopoietic engraftment compared to previous models, enabling the use of fewer HSPCs.
  • Autologous models recapitulate the human TME: The developed model successfully engrafted patient-specific immune cells and tumor tissue, creating a TME that closely mirrors the patient's own.
  • Human immune cells in the TME enhance tumor growth: Tumors grown in autologously engrafted mice were significantly larger than those in non-engrafted controls, highlighting the pro-tumorigenic role of the human immune infiltrate.
  • VEGF-A is a key pro-tumorigenic factor in the autologous TME: Single-cell transcriptomics revealed a strong VEGF-A signature in myeloid cells within the TME, and blocking VEGF-A with bevacizumab significantly reduced tumor growth.
  • The model reveals inter-patient variability: The study underscored the importance of individual patient differences, as not all autologous models showed enhanced tumor growth, suggesting the model can elucidate these variations.

Key Results

  • MISTRG6 mice showed significantly increased human hematopoietic cells in peripheral blood and various tissues compared to NSG and MISTRG mice.
  • Engraftment was efficient even with as few as 1,000 human HSPCs.
  • Autologous PDX tumors in MISTRG6 mice grew significantly larger than those in non-engrafted littermate controls.
  • Single-cell transcriptomics identified 16 distinct human cell subtypes in blood and tumor, including various myeloid, T cell, B cell, and NK cell populations.
  • VEGFA was identified as a key upstream regulator of genes in the TME, particularly in myeloid cells.
  • Treatment with bevacizumab (anti-human VEGF-A) significantly abrogated the enhanced tumor growth observed in autologously engrafted mice.

Limitations and Future Work

  • Limited innate immune cell types: Similar to other humanized mouse models, MISTRG6 autologous mice lack large numbers of human neutrophils, basophils, and eosinophils. Further research is needed to identify genes supporting the development of these cell types.
  • PDX generation variability: Not all fresh tumor tissue samples generate efficient PDXs that grow robustly within the experimental timeline.
  • Variable BM aspiration yield: In rare cases, too few HSPCs are obtained from BM aspiration.
  • Chronological lag: The time required for PDX generation can be weeks to months, potentially limiting its use as a predictor of primary therapies. However, the authors note that surgical tissue collection early in a patient's course can allow for early initiation of autologous modeling.
  • Future Work: The authors suggest that further humanization of recipient mice could improve engraftability. The model's ability to elucidate inter-patient differences is highlighted as important for future personalized medicine approaches. The study also opens doors for prospective co-clinical trials and drug comparisons.

Summary updated

Abstract

Abstract Interactions between immune and tumor cells are critical to determining cancer progression and response. In addition, preclinical prediction of immune-related drug efficacy is limited by inter-species differences between human and mouse, as well as inter-person germline and somatic variation. Here we develop an autologous system that models the TME in individual patients. With patient-derived bone marrow, we engrafted a patient’s hematopoietic system in MISTRG6 mice followed by patient-derived xenograft (PDX) tissue, providing a genetically matched autologous model. We used this system to prospectively study tumor-immune interactions in solid tumor patients. Autologous PDX mice generated innate and adaptive immune populations; these cells populated the TME; and tumors from autologously engrafted mice grew larger than tumors from non-engrafted littermate controls. Single-cell transcriptomics revealed a prominent VEGF-A signature in TME myeloid cells, and inhibition of human VEGF-A abrogated enhanced growth, demonstrating the utility of the autologous PDX system for pre-clinical testing.
BibTeX
@misc{chiorazzi2022autologous,
  author = {Michael Chiorazzi and Jan Martínek and Bradley A. Krasnick and Yunjiang Zheng and Keenan J. Robbins and Rihao Qu and Gabriel Kaufmann and Zachary Skidmore and Laura A Henze and Frederic Brösecke and Adam Adonyi and Jun Zhao and Liang Shan and Esen Sefik and Jacqueline L. Mudd and Ye Bi and S. Peter Goedegebuure and Malachi Griffith and Obi L. Griffith and Abimbola Oyedeji and Sofia Fertuzinhos and Rolando García-Milian and Daniel J. Boffa and Frank C. Detterbeck and Andrew P. Dhanasopon and Justin D. Blasberg and Benjamin L. Judson and Scott Gettinger and Katerina Politi and Yuval Kluger and Karolina Palucka and Ryan C. Fields and Richard A. Flavell},
  title = {{Autologous humanized PDX modeling for immuno-oncology recapitulates the human tumor microenvironment}},
  year = {2022},
  doi = {10.1101/2022.08.19.503502},
  howpublished = {bioRxiv},
}
Figure