Metastasis is a clinical term describing the spread of tumour cells from a primary location to distant sites. It is suggested that more than 90% of deaths from cancer are not caused by the primary tumour but by the direct effects of metastatic deposits and from the metabolic burden of a rapidly growing tumour cell mass ( Jemal et al., 2011). Traditionally an orderly cascade of cellular behaviours was presumed to underlie the progression from a well circumscribed and localised tumour growth to distant spread, based on initial local invasion, entry into the vascular or lymphatic system, survival in those fluid channels followed by extravasation and colonisation in a distal site ( Massagué et al., 2017). However, this orderly progression is not borne out by current research and the mechanisms of metastatic spread remain controversial. The role of epithelial to mesenchymal transformation (EMT) is unclear and plasticity of cells that metastasise and their relationship to the primary tumour cells, e.g. stem cells, remains the object of current research ( Pandya et al., 2017). Furthermore, cancers do not appear to disseminate randomly, but exhibit tropism for specific organs, especially lung, liver and bone ( Tarin, 2011). This observation, made over 100 years ago by Steven Paget, led to the "seed and soil" hypothesis which remains unproven. In clinical practice, surgical resection or local treatment of primary tumours is effective, but metastases remain difficult to treat. This is particularly evident for melanoma, where localised and slow-growing metastatic deposits can appear long after apparent cure ( Gershenwald & Scolyer, 2018). Similarly, in prostatic cancer the primary site is rarely a clinical problem in comparison to the pain and pathological fractures from osteolytic vertebral deposits ( Akakura et al., 1996).

Understanding the multi-step processes that regulate cancer metastasis will likely result in new therapeutics to benefit patients with a wide range of cancers at different stages of progression. Although in vitro systems, e.g. the artificial skin model for melanoma ( Hill et al., 2015) can be highly effective for studying primary tumour behaviour, connected organ systems are needed to understand metastasis. The mouse has traditionally been used as a pre-clinical model organism to study cancer under the rationale that they are a mammalian species, with the same organ systems as humans. Although genetically modified animals do spontaneously develop tumours, the introduction of human tumour cells into other species, xenografting, is a vital pre-clinical tool that enables researchers to study tumour metastasis and evaluate drug responses ( van Marion et al., 2016). Xenografts provide greater experimental control and can provide a direct translational link to the patient, particularly when the developmental origin of cancer remains unknown. However within a mouse, metastatic spread from xenografts often occurs late, well after the primary deposit has become distressing to the animal, and further pain can also result from the aggressive invasive nature of the metastases ( Gómez-Cuadrado et al., 2017). Highly metastatic cell lines are often used to accelerate the development of metastatic tumours, but these may not reflect normal metastasis, and therefore several different lines must be used, requiring many more animals ( Cruz-Munoz et al., 2008). It is sometimes possible to surgically remove the primary tumour prior to analysis of metastatic dissemination ( Srivastava et al., 2014); however, this is often associated with excessive tissue damage requiring prolonged post-operative analgesia. Direct injection of cancer cells into the tail vein ( Elkin & Vlodavsky, 2001; Minn et al., 2005), heart ( Kang et al., 2003), illiac artery ( Bos et al., 2009; Wang et al., 2015), spleen ( Morikawa et al., 1988), peritoneum ( Chu et al., 2015) or tibia ( Fisher et al., 2002) have all been used to model local metastatic behaviours, but the mouse model is limited since metastasising single cells cannot be tracked and only relatively large metastatic growths can be detected, precluding study of the earliest metastatic events. Furthermore, mouse models have also had limited success when predicting anti-cancer drug efficacy in human trials ( Day et al., 2015; Kersten et al., 2017).

The zebrafish is a tropical bony fish which for over 30 years has been increasingly used in developmental biology and human disease modelling as it contains almost all human organ systems except lungs ( Penberthy et al., 2002). The zebrafish genome has been sequenced and there is a high degree of conserved genes and genetic signalling pathways compared to humans ( Howe et al., 2013). Importantly for the study of cancer metastasis, embryos are completely transparent, facilitating imaging at single cell level within developing organs whilst also imaging the entire animal. Furthermore, the majority of studies can be carried on early-stage embryos before they are capable of independent feeding (which for the zebrafish is widely considered to be 5 days post fertilisation (dpf)), and protected under the Animals (Scientific Procedures) Act (ASPA) and EU Directive (2010/63/EU). The extra-uterine development of hundreds of eggs also permits a greater number of studies in genetically identical organisms. Since the first reported xenotransplantation of human cells into zebrafish ( Lee et al., 2005), many laboratories have shown that zebrafish embryos are useful for the study of other facets of tumour biology including cancer-induced angiogenesis ( Haldi et al., 2006); cancer cell invasion and metastasis ( de Boeck et al., 2016; Marques et al., 2009); cancer stem cell growth ( Bansal et al., 2014; Chen et al., 2017); interaction of cancer cells with the host ( Feng & Martin, 2015); and drug screening ( Corkery et al., 2011; Gibert et al., 2013). Importantly, the development of human tumours and their response to chemotherapeutic treatment in zebrafish embryos is comparable to that observed in mouse xenograft assays ( Fior et al., 2017). Additionally, while mouse xenograft models require immuno-deficient mice to prevent immune-rejection of the human cancer cells, the lack of a mature adaptive immune system within zebrafish embryos up to 14 days post-fertilisation (dpf) allows analysis of human cancers without rejection ( Lam et al., 2004).

In this article we describe the techniques for performing embryonic zebrafish xenograft experiments and demonstrate the utility of using zebrafish embryos as a model system for studying human cancer metastasis, in particular metastatic melanoma and prostate cancer. We highlight the advantages over mouse xenograft models and provide a practical experimental protocol showing how zebrafish embryos can be used as a replacement for mice to conveniently study metastatic tumour behaviour in the laboratory.

A full step-by-step protocol can be found in Supplementary File 1.

We and others have shown that xenotransplantation of human cancer into zebrafish embryos can be optimally carried out from 48-hour post-fertilisation when gastrulation is complete and the main body plan of the animal is established. The next 72 hours provides sufficient time frame to model key stages of metastatic behaviour, including local invasion, vascular entry, circulation and vascular exit. In this paper we have demonstrated how tumour invasion and/or metastatic dissemination by human cancer cells can be monitored through time-lapse microscopy. Most importantly, metastatic processes of single cells can be visualised at the earliest time points, which is not possible in a mouse model. The zebrafish embryonic xenograft of human cancer therefore directly replaces the need for using mouse xenografts and avoids welfare concerns associated with mouse models, including pain and suffering due to unexpected or excessive primary tumour growth. In the UK alone, it is estimated that over 550,000 mice are used each year for cancer research (UK Home Office statistics). On average 50 mice are used per study of cancer metastasis, and over the past 5 years there have been on average 900 publications per year in this area. We therefore estimate that 45,000 mice are used each year for research of cancer metastasis using mouse xenograft models, many of which could be replaced by embryonic zebrafish at unregulated stages of development, using the model described in this paper. To do this several historical concerns need to be addressed.

Experimentally, it is essential that following xenotransplantation human cells can be distinguished from host cells of the zebrafish. Whilst this can be achieved post-mortem by detecting human-specific antigens using immunocytochemistry ( Bentley et al., 2015), the use of lipophilic fluorescent cell membrane stains in conjunction with zebrafish transgenic lines allows much more data to be collected from animals of greater immaturity. However, quantification of xenografted cancer cell proliferation is equivalent when measured using either membrane stains or fluorescent proteins ( Bentley et al., 2015).

There are concerns that differences in cell size, microenvironmental niches and molecular signalling pathways between human patients and preclinical models (mice as well as zebrafish) could limit the relevance and translational value of data obtained from animal studies. However, our studies show that human cancer cells are able to invade zebrafish blood vessels and form secondary tumours; while previous studies have shown that VEGF and CXCR4 signalling are conserved between human cancers and zebrafish ( He et al., 2012; Tulotta et al., 2016). Nevertheless, further studies to characterise the response of human cells in the zebrafish model organism are required.

The future direction of research using zebrafish embryos for human xenograft studies will likely focus on strategies and methods to increase assay throughput and improve analysis of large data sets. These objectives will benefit from a number of technical innovations, such as devices to orientate the zebrafish for imaging ( Wittbrodt et al., 2014) as well as automated quantification and analysis of tumour cell dissemination ( Ghotra et al., 2012; Heilmann et al., 2015).

Stable cancer cell lines are often dramatically different from patient tumour cells and by definition have been selected for ease of maintenance in the laboratory environment. However, it is likely that heterogeneity and cooperation of cancer cells in patient tumours drives tumour invasion through remodelling of the extracellular matrix ( Chapman et al., 2014). Thus, cancer is represented by cells that vary in their proliferative, invasive and metastatic phenotype, which contributes both to tumour growth and also emergence of drug resistance ( Anderson et al., 2011). However, it is often not feasible to investigate the effect of tumour cell heterogeneity in mouse xenograft models as large numbers of patient primary tumour cells are required for successful engraftment. In contrast, a major advantage of the embryonic zebrafish xenograft assay is the capacity to accurately detect and monitor a small number of cells (100–200 cells per fish), including low-number cancer subpopulations such as cancer stem cells, drug-resistant cells or primary patient tumour tissue where only small numbers of cells can be recovered e.g. circulating tumour cells. Patient-derived tumour xenograft models are therefore a potential solution to the problem of limited intratumoural heterogeneity of cell line derived xenografts, which may improve the accuracy of tumour drug-response studies.

It is becoming increasingly clear that no single pre-clinical model can substitute for actual human trials, and therefore as researchers we must continually reassess and adapt our model assays to improve their relevance, which will likely involve employing an approach that combines multicellular in vitro organoid assays ( Hill et al., 2015) with both zebrafish and mouse in vivo studies. We suggest that this combinatorial approach will reduce the reliance on mouse xenograft models for the study of human cancer metastasis and drug screening. However, the challenge for translational cancer research will be to integrate the multitude of data from different model organisms to identify evolutionary conserved drug-tumour interactions between species so that we may select the most appropriate therapeutics that have the highest chance of providing an effective treatment for patients with cancer.

The data referenced by this article are under copyright with the following copyright statement: Copyright: � 2018 Hill D et al.

Data associated with the article are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

Dataset 1. Raw images used to generate figures shown in this study. Shown are images for Figure 1 and Figure 3; images in Figure 2 were obtained from stills of Supplementary Movie 1– Supplementary Movie 3. DOI: https://doi.org/10.5256/f1000research.16659.d221978 ( Hill et al., 2018).