A cell line derived xenograft, usually shortened to CDX, is a living tumor grown inside a laboratory animal by injecting established human cancer cell lines into an immunodeficient host, most often a mouse. The “xeno” in xenograft means “foreign,” reflecting the fact that human cells are being placed into a different species. Because the cancer cells come from cell lines that have been cultured and maintained in the lab, sometimes for decades, a CDX is distinct from a patient-derived xenograft (PDX), which uses tissue taken directly from a patient’s tumor. CDX models remain one of the most widely used tools in preclinical cancer research, yet their strengths and weaknesses depend heavily on how they are built and what questions researchers are trying to answer.
How a CDX Model Is Built
The starting material for a CDX is a human cancer cell line, a population of tumor cells that has been grown and passaged in plastic dishes or flasks under controlled laboratory conditions. Some of these cell lines have been around for many years. The NCI60 panel, for instance, is a collection of 60 human tumor cell lines developed in the late 1980s as a drug-screening tool, and many of those lines are still in active use today.1Nature Reviews Cancer. The NCI60 human tumour cell line anticancer drug screen Once researchers select a cell line that matches their cancer type of interest, they prepare a suspension of those cells and inject them into a host animal.
The host animal needs to lack a functioning immune system, because a normal mouse would recognize human cancer cells as foreign and destroy them. Researchers use specially bred immunodeficient mouse strains for this purpose. These mice have been genetically engineered or selectively bred to be missing key immune components, which allows the human tumor cells to engraft and grow without rejection.2PubMed Central. Considerations for selecting immunodeficient mouse strains for cancer xenograft models The degree of immune suppression varies by strain: some lack only T cells, while others are missing T cells, B cells, and natural killer cells. The choice of strain can affect how well the tumor takes and how it behaves once established.
Where the Cells Get Injected Matters More Than You Might Think
One of the most consequential decisions in building a CDX is where the cancer cells are placed. The two main approaches are subcutaneous injection, where cells go just under the skin (often on the mouse’s flank), and orthotopic injection, where cells are placed into the organ that matches the cancer’s origin. A breast cancer cell line, for example, can be injected into the mammary fat pad for an orthotopic model or under the skin of the back for a subcutaneous one.3PubMed. Large-scale Characterization of Orthotopic Cell Line-Derived Xenografts Identifies TGFβ Signaling as a Key Regulator of Breast Cancer Morphology and Aggressiveness
Subcutaneous tumors are easier to inject, easier to measure with calipers, and cheaper to maintain. They grow as a discrete lump under the skin, which makes them convenient for tracking tumor size over time. But convenience comes at a cost. Studies comparing the two approaches head-to-head find that subcutaneous tumors often look and behave quite differently from tumors in their native organ. In a gastric cancer model, for instance, subcutaneous tumors grew as well-defined masses with little interaction with surrounding tissue, while orthotopic tumors in the stomach wall were more invasive, had more stroma, and showed stronger cell-to-cell signaling with the surrounding tissue.4Journal of Toxicologic Pathology. Difference in morphology and interactome profiles between orthotopic and subcutaneous gastric cancer xenograft models
Prostate cancer xenografts show a similar split. In one comparison, subcutaneous tumors had roughly three and a half times the level of oxygen deprivation as orthotopic tumors, and only the orthotopic group developed micrometastases.5Scientific Reports. Comparative Study of Subcutaneous and Orthotopic Mouse Models of Prostate Cancer: Vascular Perfusion, Vasculature Density, Hypoxic Burden and BB2r-Targeting Efficacy That finding gets at a fundamental limitation of the subcutaneous approach: if the tumor never spreads, you cannot study the biology of metastasis in that model.
A transcriptomic analysis comparing the two sites found that the tumor cells themselves stayed fairly consistent regardless of where they were implanted; the scores for processes like epithelial-to-mesenchymal transition, blood vessel formation, and stemness were strongly correlated across sites. But the mouse stroma surrounding the tumor responded very differently depending on location, reflecting how the local tissue environment shapes the tumor’s neighborhood.6PubMed Central. Impact of Subcutaneous Versus Orthotopic Implantations on Patient-Derived Xenograft Transcriptomic Profiles In practical terms, this means the tumor cells carry their core identity with them, but the stromal conversation around them changes based on the tissue they land in.
CDX Versus PDX
The most common point of comparison for a CDX is the patient-derived xenograft, or PDX, where a piece of a patient’s actual tumor is implanted directly into a mouse without ever passing through a culture dish. PDX models are generally considered more faithful to the original human tumor because they retain the tumor’s original cellular diversity, its architecture, and much of its genetic complexity. CDX models, by contrast, have been shaped by the pressures of growing on plastic, which tends to select for cells that thrive in that artificial environment and can cause the cell population to drift away from the biology of the original tumor over time.
This matters for drug development. Research has shown that pharmacology studies conducted with PDX tumors tend to predict clinical outcomes more accurately than studies using CDX models, especially when drugs are tested at doses that match what patients actually receive.7PubMed. Advances in patient-derived tumor xenografts: from target identification to predicting clinical response rates in oncology The high failure rate of cancer drugs in clinical trials has been blamed, in part, on overreliance on CDX screening, where a drug might kill cultured cells in a mouse but fail in a real patient whose tumor has a different cellular makeup.
But PDX models have their own headaches. They are expensive, slow to establish, and not every patient’s tumor successfully engrafts. Certain cancer subtypes take months to grow, if they take at all. PDX models can also lose their original heterogeneity over serial passages in mice, experience replacement of human stroma by mouse stroma, and are subject to clonal evolution over time.8PubMed Central. Challenges and Prospects of Patient-Derived Xenografts for Cancer Research So the choice between CDX and PDX is not simply “worse vs. better.” It depends on the question being asked. If you need a fast, reproducible system to screen hundreds of compounds for initial activity against a particular cancer pathway, CDX models are hard to beat. If you need to mimic a specific patient’s tumor biology to test a personalized therapy, a PDX is the better fit.
The Stroma Problem
Regardless of whether you build a CDX or PDX, one biological reality is inescapable: the human tumor grows inside a mouse, and the supportive tissue around it, known as stroma, is mouse tissue. Studies using immunohistochemistry have confirmed that while the tumor cells in a xenograft are human, the blood vessels, connective tissue, and immune-like stromal cells that develop around them are murine in origin.9Clinical Cancer Research. Establishment in Severe Combined Immunodeficiency Mice of Subrenal Capsule Xenografts and Transplantable Tumor Lines from a Variety of Primary Human Lung Cancers This creates a kind of species mismatch: the tumor is talking to mouse stroma using human signaling molecules, and some of those molecular conversations don’t translate perfectly across species.
This mismatch can affect how drugs perform. A drug designed to target human signaling molecules in the tumor microenvironment might not bind well to the mouse versions of those molecules, which means the model could underestimate or overestimate how effective the drug would be in a patient. For drugs that directly target the tumor cells themselves, this is less of a concern. For drugs that work by modifying the environment around the tumor, it can be a real problem.
Modeling Metastasis
Cancer kills most patients not because of the primary tumor but because it spreads. CDX models can be adapted to study metastasis in several ways. The most straightforward is orthotopic implantation and waiting: place breast cancer cells in the mammary gland and see whether they eventually show up in the lungs, bones, or brain. This approach mirrors the natural metastatic cascade, including the steps of local invasion, entry into the bloodstream, survival in circulation, and colonization of a distant organ. The disadvantage is that it takes time, and not every cell line reliably metastasizes.
A faster, more controlled route is tail-vein injection, where cancer cells are injected directly into the mouse’s bloodstream. This skips the early stages of metastasis and delivers cells straight to the lungs, brain, or other organs where they can colonize. Tail-vein injection is one of the most commonly used methods for studying breast cancer lung metastasis, alongside orthotopic implantation.10PubMed Central. Is tail vein injection a relevant breast cancer lung metastasis model? It has also been adapted for brain metastasis research, where inflammatory breast cancer cell lines injected via the tail vein can robustly and consistently form brain metastases in mice.11PubMed Central. Modeling Brain Metastasis Via Tail-Vein Injection of Inflammatory Breast Cancer Cells
The trade-off is biological realism. Because tail-vein injection bypasses invasion and intravasation, it only models the later steps of the metastatic process. If you are studying a drug that blocks cancer cells from breaking away from the primary tumor in the first place, a tail-vein model is not going to tell you much. Researchers pick the approach that matches the biology they want to interrogate.
Tracking Tumors in Real Time
One practical advantage of CDX models is that established cell lines can be genetically modified before implantation. A common modification is inserting the firefly luciferase gene into the cancer cells. Once inside the mouse, these cells emit light when a substrate is injected, allowing researchers to image the tumor non-invasively using a sensitive camera system. This technique, called bioluminescence imaging, lets scientists track tumor growth and the appearance of metastases over time without having to sacrifice animals at each time point.
In breast cancer CDX models, bioluminescence imaging has been used to quantify primary tumor development in the mammary fat pad and to detect early spontaneous lymph node metastasis while the mouse is still alive.12PubMed Central. Bioluminescent human breast cancer cell lines that permit rapid and sensitive in vivo detection of mammary tumors and multiple metastases in immune deficient mice In lung cancer models, the bioluminescence signal has been shown to correlate strongly with actual tumor volume measured by histology, making it a reliable readout of how the tumor is progressing.13PubMed. Bioluminescence imaging correlates with tumor progression in an orthotopic mouse model of lung cancer This capability is one of the reasons CDX models remain popular despite the acknowledged limitations of cultured cell lines: they are easy to engineer, easy to image, and produce data quickly.
The Immunotherapy Blind Spot and Humanized Mice
CDX models have an inherent limitation when it comes to studying cancer immunotherapies: the immunodeficient mice that allow the human tumor to grow are, by definition, missing the very immune system the therapy is supposed to activate. If you are testing a checkpoint inhibitor that works by unleashing T cells against the tumor, a standard CDX mouse with no functional T cells cannot tell you whether the drug works.
Researchers have addressed this gap by developing humanized mouse models, where immunodeficient mice are reconstituted with a human immune system before the tumor is implanted. One approach uses fresh cord blood stem cells to rebuild human immune populations in the mouse. In one such model, mature human white blood cells were detectable within four weeks, and over time the mice developed human T cells, B cells, natural killer cells, dendritic cells, and other immune subsets across multiple tissues.14PubMed Central. An Improved Patient-Derived Xenograft Humanized Mouse Model for Evaluation of Lung Cancer Immune Responses
These humanized mice can then receive either CDX or PDX tumors. When the checkpoint inhibitor pembrolizumab was tested in humanized mice carrying a triple-negative breast cancer CDX, it produced significant tumor growth inhibition that did not occur in non-humanized mice. Further experiments showed the anti-tumor effect depended specifically on human CD8-positive T cells.15PubMed Central. Humanized mice in studying efficacy and mechanisms of PD-1-targeted cancer immunotherapy This confirmed that the humanized mouse system can model immune-mediated tumor killing in a way standard CDX models cannot. The technology is still expensive and technically demanding, but it has opened a door for CDX models to re-enter the immunotherapy conversation.
CDX Models in Combination Therapy Research
Beyond single-agent drug testing, CDX models are routinely used to evaluate combination therapies, including pairings of drugs with radiation. In a recent study of esophageal cancer, a repurposed drug was tested alongside radiotherapy in both a CDX model (using immunodeficient mice) and a syngeneic model (using immunocompetent mice with mouse tumors). In the CDX model, the combination suppressed tumor growth, while the syngeneic model revealed an additional benefit: the combination enhanced the infiltration of CD8-positive T cells into the tumor.16PubMed Central. Repurposing cepharanthine as a radiosensitizer in esophageal squamous cell carcinoma through dual metabolic intervention and direct targeting of p70s6K The point is that a CDX model can establish whether a combination physically shrinks a tumor, and a parallel syngeneic or humanized model can test the immune dimension. Researchers often use both side by side.
Genomic Drift in Cell Lines
A persistent concern with CDX models is that the cell lines used to create them have evolved in culture. Cancer is inherently unstable, its cells accumulate mutations and rearrangements over time, and those changes are amplified by the selective pressures of growing on plastic. A cell line passaged hundreds of times may have a very different genetic makeup from the original patient’s tumor. The extent of this genomic drift in cancer models has historically received limited attention, even though it can meaningfully affect how cells respond to drugs.17PubMed Central. Genomic evolution of cancer models: perils and opportunities
This does not make CDX models useless, but it does mean that findings from a CDX should be interpreted with an awareness of what the model does and does not represent. A CDX made from an HCT-116 colon cancer cell line tells you about HCT-116 biology, which may or may not reflect the diversity of colon cancers seen in the clinic. The reproducibility that makes CDX models attractive for controlled experiments is, paradoxically, also their weakness: they represent a narrow slice of cancer biology.
Beyond the Mouse
Mice are the dominant host for xenograft studies, but they are not the only option. The chorioallantoic membrane of a developing chick embryo, or CAM, is a naturally immunodeficient tissue that can accept human tumor cells without immune rejection. Researchers can grow CDX tumors on the CAM to study angiogenesis, drug efficacy, and even metastasis at much lower cost and higher throughput than mouse studies.18PubMed. The chick embryo chorioallantoic membrane patient-derived xenograft (PDX) model In many countries, experiments terminated before a certain embryonic age do not require formal animal ethics approval, which further lowers the barrier to entry.
Zebrafish offer another alternative, especially for visualizing tumor cell behavior in real time because their larvae are transparent. However, their small size limits certain applications. Chick CAM models, for example, can be used with PET imaging to screen for radiotracer uptake in tumors, something a zebrafish is simply too small to accommodate.19npj Imaging. The chicken chorioallantoic membrane as a low-cost, high-throughput model for cancer imaging Both alternative hosts serve as early-stage screening platforms, with the understanding that findings still typically need to be validated in mouse models before moving toward clinical application.
Ethical Frameworks and the 3Rs
Any research involving live animals raises ethical questions, and CDX studies are no exception. The prevailing framework in laboratory animal research is the 3Rs: Replacement, Reduction, and Refinement. Replacement asks whether an animal experiment can be substituted with a non-animal method, such as cell culture or computational modeling. Reduction pushes researchers to use the smallest number of animals that can still produce statistically meaningful results. Refinement requires minimizing pain and distress. Published CDX studies increasingly cite adherence to these principles explicitly, with researchers justifying their chosen sample sizes based on prior data and opting for the smallest groups that still yield reliable results.20Nature Communications. Tumor-agnostic cancer therapy using antibodies targeting oncofetal chondroitin sulfate
The rise of non-murine hosts like the chick CAM and zebrafish fits neatly into the Replacement pillar. So does the increasing use of organoids and other three-dimensional cell culture systems that can model some aspects of tumor biology without a living animal. CDX mouse experiments have not disappeared, but they are increasingly positioned as one layer in a tiered screening strategy where simpler, cheaper, and less ethically fraught models handle early-stage questions and animals are reserved for the questions that only an intact organism can answer.