Patient-derived xenografts to study cancer metabolism: when does X mark the spot?

Abstract

A major goal of cancer research is to understand the requirements for cancer growth and progression that can be exploited to treat patients. Model systems reduce the complexity and heterogeneity of human cancers to explore therapeutic hypotheses; however, some relevant aspects of human biology are not well represented by certain models, complicating the translation of preclinical findings to help patients. Here we discuss the advantages and limitations of patient-derived xenografts as a model system to study cancer metabolism, offering a framework to best use these models to address different types of metabolism-specific research questions.

Cancer cells rewire numerous pathways to support proliferation and survival in inappropriate tissue contexts. Of the many tumor growth-enabling adaptations associated with malignancy, metabolic changes are recognized as a hallmark of cancer and, with varying degrees of success, a target for anti-cancer treatments. In the era preceding modern molecular biology, paradigm-shifting discoveries in cancer metabolism include Warburg’s observation of increased glucose uptake and fermentation even when oxygen is available, the use of anti-folates to deprive cells of one-carbon units, and the development of nucleoside analogs to limit DNA replication in dividing cells with high deoxyribonucleotide demands. Advances in the assessment of cancer genomes and improved detection of metabolites has increased understanding of the ways in which metabolism can be adapted to support different tumor types. The renewed interest in better understanding the metabolism of cancer strives toward developing treatments tailored around the unique biology of each patient. Together with rapid advances in technology to assess various cancer phenotypes, it is more important than ever to be thoughtful about the methods used to study cancer in order to increase the likelihood that research findings translate into improved patient outcomes. To this end, what is the role for patient-derived xenografts (PDXs) in tumor metabolism research?

Multiple model systems have been used for the study of cancer including human cancer cell lines and genetically-engineered mouse models. While these model systems have enabled many important discoveries, clear limitations exist. Established human cell lines only represent a subset of human cancers, and have undergone selection to grow in artificial conditions, and serial passaging eliminates the heterogeneity found in human tumors. Mouse models have the advantage of arising in an in vivo tissue environment, but species-specific differences in physiology and the fact that tumors are derived from a limited number of oncogenic events that are introduced into an entire animal tissue can be limitations. PDXs have emerged as a model system that is more closely linked to the biology of human cancer as it is being treated in the clinic. The value of PDXs is best illustrated by examples where these systems have been used most effectively (1). This includes models of rare cancers or unique cancer genotypes where cell lines and autochthonous mouse models are not available, multiple tumor samples derived from the same patient over the course of an individual’s disease progression allowing the study of tumor evolution through treatment, and an opportunity to evaluate treatment responses in model systems as patients receive the same therapies in a clinical trial. A common theme in the successful applications of PDX models is less the models themselves, but rather the nature of the scientific question at hand and the unique ability of PDXs to address that question over other model systems.

For the advantages that they offer, PDXs also come with limitations. Generating a PDX resource is a costly endeavor that requires a clinical-translational research effort with access to fresh patient material and the proper protocols to collect and maintain the models. Additionally, the process of generating PDXs involves serial passaging, in which tumors are dissociated and engrafted into varying anatomic locations of an immunocompromised mouse, which over time imposes selective pressures that may differ from those found in patients. For instance, native tumors contain a diverse collection of cell types, including cancer cells, immune cells, and stromal cells. When passaged as a xenograft, non-cancer cells form the patient are lost, and the selective pressures on passaging the cancer cells may select for adaptions that differ from those in patient tumors. These drawbacks can be partially addressed by pursuing orthotopic implantation of tumors into the original site of tumor origin as opposed to subcutaneous implantation, allowing interactions with tissue specific cells derived from the new host. Nevertheless, differences between species may still lead to different inter-cellular interactions and there is debate over the degree to which PDX models undergo genetic changes that deviate from the original patient tumors (2,3).

Relevant to studying metabolism, the use of immunodeficient murine recipients further introduces two confounding factors: an absent adaptive immune system as well as whole-body differences in metabolism. With regards to the former, a commonly used recipient mouse model lacks the IL2 receptor gamma chain, rendering T cells unable to proliferate and reject xenografted tissue, impairing any examination of the effects of the T cell component on the tumor microenvironment. This artifact—along with the absent stromal compartment intrinsic to the native tumor—diminish the ability of PDXs to evaluate questions about metabolism involving other cell types in the tumor microenvironment and the role that metabolism may play in mediating anti-tumor immune responses, an issue that is of great relevance with respect to ongoing efforts to improve cancer immunotherapy (4). The requirement for xenografting material from human patients to another species (typically mouse) also introduces species-specific differences in total body metabolism that may affect metabolic interactions between the tumor and other tissues, or impact tumor nutrient availability. Differences in whole body metabolism between species include dietary habits, circadian rhythms, and metabolic differences in subsets of cells as well as specific organs.

In considering the utility of PDXs in cancer metabolism research, it is essential to take a step back and first ask what are the major frontiers in cancer metabolism research and subsequently assess how PDX models might be most appropriately used to address key questions. First, there is a growing appreciation that using physiologic nutrient concentrations for cell culture studies can affect metabolism, and that the composition of standard commercially-available cell culture medias with non-physiologic concentrations of nutrients may confound experimental results (5). This recognition has led to the formulation of media with nutrient levels that better match those in blood (6), and metabolite profiling of tumor interstitial fluid to better understand how nutrients available to cells in the tumor may differ from the nutrients found in blood (7). The ultimate goal in formulating medias with more physiological nutrient levels is to better model the metabolic constraints on cancer cell proliferation within the tumor microenvironment. By their nature, cancer cells within PDXs are exposed to a more physiological nutrient environment, including gradients of nutrients that exist within tissues that are not captured by any cell culture models.

Moving beyond the impact of nutrient availability on the intrinsic capacity for cancer cells to proliferate, there is increasing appreciation for the interactions between cancer cells and non-malignant cells in the tumor microenvironment that collectively shape tumor growth and anti-tumor treatment responses (4). Most notable in this regard are the durable long-term remissions that can be achieved with T cell immune checkpoint blockade targeting the PD-1/PD-L1 axis. In efforts to understand mechanisms of treatment response as well as intrinsic resistance and disease control, there is growing interest in the competition between immune cells and cancer cells for a shared nutrient pool in the tumor microenvironment. Myeloid cells and stromal support cells add further complexity in modulating nutrient availability, in some cases limiting nutrients for growth, in other cases potentially supplying metabolites that may confer resistance to various therapies. Improved sequencing technologies enabling cell subset profiling may expand our perspective on as-yet undescribed factors that involve the tumor microenvironment, such as emergent observations about the role of neurohormonal signaling on tumor metabolism (8). PDXs introduce the heterogeneous cancer cell populations from patients into a model where interactions with resident tissue cells can impact growth; however, the necessity of growing PDXs in immunocompromised animals limits the utility of PDXs in answering many of the key questions around how different cell populations compete for nutrients within tumors.

An example illustrates how the use of PDXs can improve understanding of cancer metabolism. A recent study employed PDXs to explore the metabolic factors associated with efficiency of metastasis formation (9). Metastasis is a complicated process that requires escape from a primary tumor, survival in circulation, seeding of distant tissues, and tumor outgrowth in a new tissue site. The process of PDX generation through removal from the patient, dissociation into a cell suspension, and delivery into an immunocompromised mouse is variably efficient, recapitulating some of the biological challenges for tumor cells establishing a new metastasis. The authors stratified a series of melanoma PDX models based on their efficiency of metastasis formation and performed in vivo metabolomics to explore the utilization of isotopically-labeled nutrients. In doing so, they identified enhanced lactate transport as being associated with efficient metastasis formation and nominated the monocarboxylate transporter MCT1 as a therapeutic target. The authors further demonstrated heterogeneity in MCT1 expression in efficiently-metastasizing PDXs, a factor which may explain why some circulating tumor cells are capable of metastasis formation while others are not. The critical insights from this study reflect the ideal alignment of the questions that can be addressed using a PDX model system. That is, by using a model system that is heterogeneous and inherently reliant on niche colonization and tumor outgrowth in a living organism, they could explore metabolism and its contribution to the efficiency of metastasis formation. It is also noteworthy that the study avoids questions extrinsic to the cancer cell for which the PDX model is not well-suited to answer, such as the role of the immune system in restricting melanoma metastasis formation.

Another example involves the use of PDX models to examine heterogeneity in small cell lung cancer and therapy response. Long thought to be a homogenous disease marked by rapid proliferation, TP53 inactivation and RB loss in the neuroendocrine lineage, small cell lung cancer demonstrates unique subtypes based on differential expression of specific transcription factors: ASCL1, NEUROD1, POU2F3 and YAP1(10). The majority of treatment-naïve cases reflect the ASCL1-dominant subtype; however, compelling studies performed in xenografts derived from circulating tumor cells have shown subtype evolution through chemotherapy treatment (11,12). Small cell lung cancers demonstrate sensitivity to upfront treatment with platinum/etoposide doublet chemotherapy, but responses are often short-lived with few effective treatments in the second line and beyond. PDX models are well-suited to capture the evolution of platinum resistance throughout a treatment course, as well as the intratumoral heterogeneity that underlies cancer cell persistence and platinum resistance. Initial work in PDXs derived from circulating tumor cells demonstrates increased ROS and hypoxia gene expression signatures in cases of platinum resistance(13), and further subtype analysis has exposed subtype-specific therapeutic liabilities that include metabolic targets such as arginine limitation(14) and anti-metabolite treatment with nucleoside analogs and anti-folates (12). Simultaneously supporting evaluation of therapeutic efficacy and detailed tumor profiling, PDXs can provide a facile correlate to clinical outcomes (15) or permit validation of complementary, genetically tractable autochthonous murine models to accelerate discovery and pre-clinical development.

As we look forward to addressing the major unanswered questions in cancer metabolism, where do PDXs fit in the roadmap to discovery? Like all model systems, PDXs are most useful when used to ask question that take advantage of their strengths, while bearing in mind their experimental limitations (Table 1). While care should be taken not to over-interpret or overstate conclusions as relevant to the aspect of metabolism being studied, the increasingly translational landscape of cancer metabolism research benefits from a diversity of experimental approaches to answer challenging questions and advance our collective understanding of cancer biology.

Table 1:

Strengths and Weaknesses associated with PDX Model Selection

Strengths	Weaknesses
Derived from human patient tumors	Species differences between mouse and human
Models rare diseases and tumor genotypes	Requires access to patient samples for generation
Captures population-level heterogeneity	Less genetically tractable than many other models
Captures intratumoral heterogeneity	Tumor selection may alter phenotypes with serial passaging
Allows for studies of tumor growth in vivo	Requirement for immunocompromised host limits studies of cancer-immune cell interactions