Retargeting of T cells with CAR against a tumor marker has shown remarkable clinical efficacy in relapsed or refractory B-cell malignancies and multiple myeloma (1). This success has led to the approval of several CAR T cell products targeting hematologic malignancies by the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA). Despite significant clinical efficacy in some cancers, CAR T cell therapy is often associated with several side effects, such as a cytokine storm or neurological toxicity. In addition, CAR T cell products must be produced autologously due to T cell product incompatibility between patients, resulting in high therapy costs. As an alternative, NK cells have recently been recognized as a promising cytotoxic cell type for CAR-based cancer therapy, overcoming many of the obstacles of T cell-based systems. Primary NK cells and the NK cell line NK-92 have been shown to be safe with no adverse effects in clinical trials, and early clinical data in hematologic malignancies are encouraging in terms of efficacy (2–4). NK cells do not express the T cell receptor and therefore an NK cell product can be infused into multiple recipients in an allogeneic scenario without the risk of graft versus host disease (GvHD). In contrast, NK cells express an inherited set of activating or inhibitory receptors that allow them to recognize tumor or infected cells, making them unique cytotoxic cells. Although NK cells can recognize and kill many cancer cells, some cancers can evade NK cell-mediated killing. To overcome this resistance, NK cells can also be genetically engineered with CAR, similar to T cells (5–7). Indeed, several sources of NK cells are in preclinical or clinical development, including primary blood-derived NK cells, NK cells differentiated from hematopoietic or induced pluripotent stem cells (3, 8, 9). In addition, stable NK cell lines such as NK-92 have been established for clinical use (2, 10).

Despite the success of CAR-based therapies in hematological malignancies, solid tumors such as breast cancer remain a challenge. CAR T cells have not shown clinical efficacy and CAR-NK cells are currently in clinical investigation. The difficulties in targeting solid tumors are determined by inter-tumor heterogeneity, heterogeneity in antigen expression, resistance mechanisms to CAR T or NK cells and often generate metastatic diaspores that are difficult to reach by immune cells. Therefore, a personalized approach is required to select the most appropriate CAR system for each patient. Novel CAR-based strategies are rapidly developing and good preclinical models are needed to apply them in a personalized manner. Currently, novel CAR-NK cell therapies are tested in immunodeficient mouse models. However, these models are very expensive, laborious, inflexible and slow to yield results. Furthermore, the lack of transparency of the host tissues hampers in vivo observations of homing processes and the antitumor response. Therefore, in order to bridge the gap from in vitro studies to clinical trials, medium to high-throughput preclinical models with optimal live-cell imaging properties and fast experimental readout are needed to test the efficacy of CAR-NK cells against patient cancer cells.

Zebrafish (Danio rerio) embryos and larvae are prime candidates for the development of such a xenograft screening system. Their small size and transparency allow for medium-throughput injections with the prospect of automated high-throughput upscaling in the near future. Likewise, imaging of cancer and immune cells in transplanted embryos and larvae can be performed at high throughput at low cost. Notably, the lack of a fully functional adaptive immune system during early larval stages prevents rejection of transplanted cells in zebrafish (11). Zebrafish xenografts have been developed for the investigation of human tumor cell proliferation, apoptosis, invasion, extravasation and small molecule anti-cancer drug screening (12, 13). Recent work has shown that the zebrafish model can be used to assess the in vivo efficacy of CAR T cells (14–16). This provides proof of principle that embryonic zebrafish xenografts can be used for preclinical testing of CAR T cell efficacy; however, NK cells are biologically different from T cells in terms of their proliferation, migration, activation, metabolic properties and cytokine requirements (17). Whether zebrafish are suitable to evaluate the antitumor action of CAR-NK cells, especially the relatively large NK-92 cells (18), is currently unclear.

Here, we established zebrafish embryos and larvae as a preclinical xenograft model to assess the efficacy of CAR-NK cells against metastatic cancer cells of solid tumor origin. We validated the CAR-NK cell system with two different specificities and two different breast cancer models. Furthermore, we investigated the live interaction of CAR-NK and cancer cells and the migration of CAR-NK cells through the microvasculature to the site of metastatic breast cancer cells. This demonstrates the functionality and usefulness of CAR-NK cells in cancer therapy and reaffirms the power of zebrafish xenografts in preclinical cancer research.

CAR-NK cell therapies have recently emerged as a promising platform to treat otherwise incurable cancers without severe side effects. Despite promising clinical results in hematological malignancies, solid tumors remain challenging mainly due to their heterogeneity, resistance mechanisms, suppressive microenvironment, immune exhaustion and metastatic properties. Novel CAR or other targeted NK cell systems are rapidly emerging and therefore quick preclinical models to test their in vivo efficacy are required to facilitate the development of NK cell therapy. Currently, immunodeficient mouse xenografts are used, but they are limited in their use by high costs and long read out times.

Here, we describe a preclinical zebrafish xenograft model as an alternative and reliable system to test the efficacy of CAR-NK cells against metastatic breast cancer. Using two different CAR NK-92 systems targeting PD-L1 or ErbB2 molecules, we demonstrate that human NK cells survive in the zebrafish environment and retain specific cytotoxicity against metastatic breast cancer cells. Due to the absence of adaptive immune cells in zebrafish larvae, neither CAR-NK cells nor cancer cells appear to be rejected by the host´s endogenous immune system, although we cannot exclude that innate immune cells such as macrophages diminish CAR-NK and breast cancer cell numbers. In our metastatic model, we observed rapid killing of cancer cells in the periphery within 8 and 24 hours, respectively, depending on the target. We also observed partial elimination of cancer cells by unmodified NK-92 cells in our bulk embryo analysis setup (a time course setup would likely require similar numbers of embryos to demonstrate such activity), suggesting that this model can also provide information on the natural killing of NK cells against certain tumors of interest. While natural cytotoxicity may be difficult to test in short-term in vitro cytotoxicity assays, the zebrafish may certainly provide clinically relevant information on cytotoxicity in vivo. Although unmodified NK cells are unable to eliminate the resistant cancer cells and CAR targeting is required, additional natural cytotoxicity may support CAR-based cancer cell elimination of cells in which the target antigen is less expressed, blocked or shed.

The small size and transparency of the system allowed microscopic tracking of CAR-NK cells as well as of breast cancer cells in individual animals over time to obtain kinetic data on killing efficacy. The small size of the animals and high progeny number allowed us to inject larger cohorts of zebrafish, resulting in reliable data also in a pooled analysis setting. Our system will allow for medium- to high-throughput testing of novel CAR-NK cell products in early developmental stages of zebrafish before initiating time consuming mouse models, which in addition require the injection of a much higher cancer cell number. In this regard, the zebrafish system will be a promising tool to test the reaction of primary patient cancer cells to CAR-NK cell therapy. In fact, most cancer patients do not benefit from immunotherapy, and distinguishing responders from non-responders prior to treatment will allow for a reliable prediction of the treatment outcome. Because of the large number of available zebrafish embryos, low haltering costs and the small number of required patient cancer cells for injection (100-200 cells per embryo) the system is ideally suited for personalized clinical testing. Such a testing model allows fast recapitulation of the patient’s disease and testing of multiple CAR-NK cell products alone or in combination with other cell-based therapeutics such as CAR-T cells, whose action has recently been demonstrated in zebrafish embryos as a basis for T-cell based therapy (14, 15). Of note, zebrafish xenografts also offer the possibility of testing the CAR-NK cells in combination with small drugs for synergistic effects, in analogy to previous studies using zebrafish for screening small drug libraries (23).

The zebrafish embryos’ small size and transparency allowed us to track CAR-NK cells and cancer cells at the single cell level, and to study CAR-NK/cancer cell interaction and cytotoxicity in real-time. In the future, confocal microscopy of cancer cell plus CAR-NK injected embryonic zebrafish could be used to study lytic granule polarization toward the immunological synapse, which is an important parameter associated with cancer cell resistance to NK cell cytotoxicity (20). Furthermore, using transgenic zebrafish with a fluorescently labeled vasculature, we observe homing of metastatic cancer cells to the zebrafish periphery and intravascular migration of CAR-NK cells to these individual tumor cells, suggesting that CAR-NK cells can effectively find and eradicate circulating tumor cells and potentially micrometastases that are difficult to eliminate with standard therapies. In a clinical setting, these capacities can be exploited. For instance, the homing and persistence of novel improved CAR-NK cells including CIS (cytokine-inducible SH2-containing protein) knock out, transgenic IL-15 production and others can be tested in the zebrafish xenograft model (24, 25), as recently done to assess the efficacy of CAR-T cells expressing IL-15 and CCL19 (16). Although active homing of NK cells has previously been shown for solid tumors in a mouse model (26), further studies are required to determine whether CAR-NK cells in the zebrafish xenograft model randomly encounter cancer cells within the bloodstream or whether this is an active migration process involving specific mechanisms. If accurate quantification of NK cells is required, a brighter and photobleach-resistant cell dye or engineered cells expressing a fluorescent protein would be beneficial.

Although the used zebrafish xenograft model has many advantages, it is important to be aware of its limitations, one of which is its brevity itself. The model is very short-term and does not allow for the development of secondary (acquired) resistance mechanisms within the transplanted animal, as ultimately transplants might be rejected after development of innate immunity. However, recently an adult zebrafish model has been introduced that allows a minimum of 14 days of cancer-immune cell interaction albeit it is not clear whether this time is sufficient to study secondary resistance (27). Importantly, patient treatment needs to be started promptly in most cases, which is why short-term assays are generally desirable. Furthermore, the very first response to CAR-NK cell therapy is the first important parameter for healthcare professionals to apply the treatment. In line with this, we have shown that the CAR-NK cell efficacy can be measured as early as 24 hpi, but assessment can be extended to 72 hours if necessary. Similar to other xenograft animal models, zebrafish do not contain human immune and stromal cells, limiting the ability to study complex immune cell interactions. This drawback can be overcome in part by co-injecting human stromal cells along with the cancer cells, as recently shown for CAR-T cells (14) or even by co-injecting immune cells from the patient. In order to enhance the survival of human cells in zebrafish, genetic engineering approaches introducing the production of human cytokines and growth factors can be taken, such as shown for human GM-CSF expression in zebrafish (28). It is also important to note that although zebrafish have a wide temperature range due to temperature variation in their natural habitat, xenotransplanted individuals are usually maintained at a comparatively low temperature of 33 to 34°C, opposite to the standard 37°C of the mouse and human body. This could affect cancer and NK cell growth, proliferation and metabolism. However, we did not observe any effect of the lower temperature on CAR-NK cell killing efficacy, suggesting that this may be a minor issue. In addition, protocols for maintaining adult zebrafish at 37°C were recently established, suggesting that zebrafish xenografts can be further adopted for preclinical cell therapy models in the future (23).

In this work, we successfully assessed the efficacy of the CAR-engineered NK-92 cell line to eliminate breast cancer cells within xenotransplanted zebrafish larvae. NK-92 cells have many similarities with primary NK cells. Several other NK sources are currently being investigated as CAR effectors (3, 9) and further studies are required to test whether zebrafish xenografts can also be used for these cell types. In addition, NK cells have recently been shown to be successfully retargeted by various engagers such as monoclonal antibodies or bispecific antibodies, and the zebrafish model provides an attractive system for high-throughput in vivo testing of these drugs (29–33). Therefore, we propose that the combination of novel engineered CAR-NK cells along with the versatility of zebrafish xenografts may facilitate the future advancement of CAR NK-based therapeutic approaches to the benefit of patients suffering from progressive disease.

Here, we demonstrate that zebrafish embryos and larvae are a suitable in vivo system for testing the killing efficacy of CAR-NK cells against metastatic solid tumor cells. This system is flexible and will allow mid-throughput screening of new drugs at low cost, leading to faster development of highly effective CAR-NK cell products. It can also be refined to personalize testing of CAR-NK cell efficacy against patient-derived cancer cells and as a predictive tool to select the best-possible therapy for the patient.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Ethical approval was not required for the studies on humans in accordance with the local legislation and institutional requirements because only commercially available established cell lines were used. The animal study was approved by Landesdirektion Dresden (DD25-5131/450/4 and 25-5131/564/2 for breeding). The study was conducted in accordance with the local legislation and institutional requirements.

NM: Investigation, Writing – original draft, Writing – review and editing. PM: Writing – review and editing. AK: Writing – review and editing. WR: Writing – review and editing. SK: Writing – review and editing. WW: Writing – review and editing. TT: Conceptualization, Supervision, Writing – review and editing. FK: Conceptualization, Supervision, Writing – original draft, Writing – review and editing. JE: Conceptualization, Supervision, Writing – original draft, Writing – review and editing.