Long-term non-invasive drug treatments in adult zebrafish that lead to melanoma drug resistance

Over the past two decades zebrafish have emerged as an important model for drug discovery, directly leading to drugs that have entered clinical trials or for compassionate use (Patton et al., 2021b). These include therapies that promote haematopoietic stem cell renewal (North et al., 2007) or prevent antibiotic-induced ototoxicity (Kitcher et al., 2019), and those that treat childhood epilepsy (Baraban, 2021), cancer (Mandelbaum et al., 2018; White et al., 2011; Yan et al., 2019), lymphatic anomaly (Li et al., 2019), arteriovenous malformation (Al-Olabi et al., 2018) and fibrodysplasia ossificans progressive (Yu et al., 2008), among other diseases and disorders. This success is, in part, because zebrafish are vertebrates and share over 80% of disease genes with humans (Howe et al., 2013), as well as have shared drug metabolism, physiology and pharmacology (MacRae and Peterson, 2015). Thus, zebrafish pre-clinical disease models are important platforms for drug discovery and repurposing, even at times leading to new treatment strategies for patients directly from zebrafish models.

With some exceptions, most zebrafish drug discovery, gene–drug screens and compound–phenotype evaluation studies are performed using zebrafish embryos. However, embryos and larval stages do not fully recapitulate adult disease states and lack a complete immune system. Drug screening and discovery in adult zebrafish – and modelling the impact of long-term drug treatment – has been limited by methods of drug administration, which can be complex, harmful and imprecise. Current methods for experimental drug discovery in adult zebrafish involve adding the drug to the fish water, which can irritate exposed mucosal surfaces (e.g. eyes, gills), are not appropriate for water-insoluble compound administration and involve administrating large quantities of drug to fish water with unknown final drug concentrations absorbed by the fish. Alternatively, drugs can be administered by oral gavage or through injection (retro-orbital or intraperitoneal), which are more precise for dosing, but invasive and sometimes fatal, and require repeated anaesthesia (Dang et al., 2016; Kinkel et al., 2010; Pugach et al., 2009). Although these methods are generally acceptable for short-term drug treatments, they are problematic for longer-term drug protocols because they can lead to accumulative distress and injury to the animal. Adding antibiotics to jelly-like food has been used to manage zebrafish colony health (Chang et al., 2017), however, these methods were not designed to administer precise drug treatments to individual fish in the experimental and pre-clinical disease model context, and are therefore not appropriate to investigate dose-based drug responses. Thus, non-invasive and precise drug delivery protocols that permit longitudinal experimental drug treatments for adult zebrafish are not well developed.

Our laboratory uses zebrafish to model the progression of melanoma, the deadliest form of skin cancer (Travnickova and Patton, 2021). Zebrafish melanoma models have provided key insights into the origins, progression and new drug targets for melanoma (Baggiolini et al., 2021; Ceol et al., 2011; Kaufman et al., 2016; Patton et al., 2005; Travnickova et al., 2019; Venkatesan et al., 2018; White et al., 2011). However, a significant gap in the field has been to generate zebrafish melanoma models that recapitulate the development of acquired drug resistance as seen in patients (Patton et al., 2021a). Indeed, acquired resistance is one of the major challenges limiting the progression-free survival time of melanoma patients on therapies that specifically target BRAF^V600E and MEK (MAP2K) signalling (Chapman et al., 2011; Larkin et al., 2014; Long et al., 2015; Luke et al., 2017; Ribas et al., 2014; Robert et al., 2015; Sosman et al., 2012). This gap is unfilled in zebrafish melanoma models due to the lack of sustainable long-term drug administration methods for adult fish, despite evidence for the strong potential of zebrafish models to recapitulate many human melanoma plasticity states, including residual disease (Travnickova and Patton, 2021).

Here, we present a novel drug formulation and administration method for adult zebrafish that enables the delivery of precise drug concentrations directly via food pellets. As a proof of principle, we fed pellets containing vemurafenib to zebrafish with BRAF^V600E melanoma and showed that they caused immediate and on-target efficacy in reducing melanoma growth. Long-term treatments (>2 months/daily treatment) led to drug resistance and melanoma progression, enabling zebrafish models of melanoma drug resistance for the first time. Our drug-pellet method enables modelling of the effects of drug dose, administration and long-term treatments in adult zebrafish, is non-invasive and limits animal handling, and we anticipate that it will be widely applicable across a wide range of zebrafish disease models.

We wanted to generate drug pellets that would deliver consistent drug doses to treat melanoma while also being quickly and freely consumed by the zebrafish. To begin, we prepared vemurafenib pellets using a drug-pellet mould that we designed and 3D printed, so that each drug pellet would be 2 mm in diameter (comparable to the size of a zebrafish egg and easily consumed by adults) (Fig. 1A). We suspended dry fish food in water and mixed this with agar–agar and gelatine powder to create a food paste to generate the base of the food pellet. To prepare the pellets, 10 cm³ dry fish food was added to water up to 50 ml in a conical centrifuge tube, completed with 1 g food-grade agar–agar plus 2 g gelatine powder, and shaken well to generate a red-coloured mix (Fig. 1B). This recipe was optimised to achieve a balance between congelation for laboratory handling and a palatable texture for the fish. The mixture was then transferred into a 100 ml borosilicate container and microwaved for ∼1 min to reach boiling, to generate a smooth texture (Fig. 1B). When the mixture was cooled yet before it congealed, we added the desired drug (i.e. vermurafenib) or dimethyl sulfoxide (DMSO) as a control (Fig. 1A). We prepared our vemurafenib pellets to each deliver 100 mg/kg vemurafenib, as has been shown to induce tumour regression when administered by oral gavage (Dang et al., 2016). Briefly, 10 mg vemurafenib powder was resuspended in 300 μl DMSO and then mixed well with 700 μl agar–fish food mixture while warm in a 1.5 ml microcentrifuge tube, generating a pink-coloured paste (Fig. 1C). Each well of the 3D-printed pressing mould is 5 μl, and therefore each pellet contained 0.05 mg vemurafenib. Given that the average weight of each fish is 0.5 g, one vemurafenib pellet per day will deliver a dose of 100 mg/kg.

A sheet of parafilm large enough to cover the 3D-printed drug pressing mould and an appropriately sized plastic roller were prepared as tools to press drug pellets (Fig. 1D). Once the food–drug paste began to congeal, the paste was transferred onto the drug-pressing mould, covered with the parafilm sheet (to form a ‘sandwich’ with the mould between the parafilm sheet and the backing paper) and then pressed into the mould with the plastic roller (Fig. 1E). The parafilm sheet was lifted, and the mould carefully removed to reveal the drug pellets now formed and adhering to the backing paper (surface tension will trap most of the drug pellets neatly on the backing paper) (Fig. 1E-G). This approach enables batches of >100 cylindrical tablet-like drug pellets to be conveniently made during a single preparation. DMSO control pellets were prepared in a similar way. In this experiment, drug pellets were prepared once every week and stored at −80°C as individual pellets in PCR tubes for daily dose aliquots (Fig. 1H). The entire process of pressing drug pellets is shown in Movie 1.

We confirmed that the 3D-printing technique produced consistent drug pellets by weighing pellets produced within the same batch or across different batches (Fig. S1). Further, we addressed the stability of vemurafenib in pellets stored at −80°C using high-performance liquid chromatography (HPLC) and found that vemurafenib was stable in the pellets for up to 2 weeks (Fig. S1). The pipeline concept of 3D design and printing the pressing mould can be found in Fig. S1, and the handling and processing of 3D-printing the pressing mould is shown in Movie 2. Overall, this method can produce stable and dose-controlled drug pellets, which are suitable for snap-frozen storage for up to 2 weeks.

Next, we fed the drug pellets to individual adult zebrafish with melanoma that were individually housed in 1 l tanks. As shown in Movie 3, zebrafish actively sought for and consumed the drug pellets voluntarily. In general, zebrafish consumed their drug pellets immediately; however, any pellets that were ignored for more than 15 min were removed and replaced by fresh pellets to ensure consistent dosing in drug administration. Using a pipette to gently deliver the drug pellets encouraged zebrafish to spot and consume the pellets. We observed no distress or toxicity in the fish from either the drug-pellet formulation or the administration (Fig. S2), and no adverse effects on normal feeding behaviour were reported by independent facility staff even after long-term treatment (daily treatment up to 10 weeks).

To test the drug efficacy, we provided DMSO or vemurafenib drug pellets to adult zebrafish with spontaneous BRAF^V600E p53 mutant melanomas (Patton et al., 2005) and compared the tumour responses between the two groups. To track the tumour size over time, we imaged and analysed the brightfield images of each tumour from these individuals every week (Fig. 2A). DMSO-treated zebrafish showed continuous lesion expansion and tumour growth (Fig. 2B,C), whereas fish treated with 100 mg/kg vemurafenib daily showed tumour regression over 3 weeks (Fig. 2D,E), with the average tumour size reduced by 60% in 2 weeks, and 70% in 3 weeks, compared to the pre-treatment state (Fig. 2E). This result is highly comparable to the observation from the previous 100 mg/kg oral-gavage regime (Dang et al., 2016), from which a 2-week daily treatment reduced the melanoma size by 70% on average. Treatment response was heterogeneous, as observed in patients (Chapman et al., 2011; Flaherty et al., 2010; Joseph et al., 2010; Larkin et al., 2019; McArthur et al., 2014) and other genetic models of melanoma (Dang et al., 2016; Perna et al., 2015). We observed no apparent difference between tumour appearance (pigmented versus unpigmented, bulk versus superficial) and response to vemurafenib.

Next, we increased the dose of vemurafenib by feeding the fish two drug pellets per day and observed that a higher proportion of lesions had improved regression (Cohort I: 4/9 lesions reduced by greater than 80% in 3 weeks, compared to 1/6 lesions in fish treated with one pellet/day), with some lesions developing signs of resistance following the initial regression (Fig. 2D-G). To validate the drug response kinetics and potential for acquired drug tolerance, we repeated the two drug-pellet treatments on a larger cohort of fish (Cohort II: 22 lesions in ten fish) and observed similar response patterns (Fig. 2H,I). Thus, drug pellets are highly effective at treating zebrafish melanoma, are well tolerated and can be used in dose escalation studies.

Patients with BRAF mutant melanoma and receiving targeted therapy will often show a dramatic reduction in melanoma burden, followed by recurrent melanoma growth from residual disease (or persister cells) (Marine et al., 2020; Shen et al., 2020b). We and others have shown that persister cell states are heterogeneous and consist of cell states that pre-exist in the primary tumour and emerge de novo (Rambow et al., 2018; Travnickova et al., 2019). We have studied these states in zebrafish by conditional expression of the master melanocyte transcription factor MITF, while others have used mouse xenograft studies to follow human melanoma cells following administration of BRAF inhibitors (Rambow et al., 2018; Shen et al., 2020a; Travnickova et al., 2019, 2022 preprint). However, there are no zebrafish models of BRAF inhibitor resistance over time because of the limitations of long-term drug delivery methods to adults.

We noticed that one tumour developed resistance and regrowth at the end of week 3 of vemurafenib treatment at 200 mg/kg in the short-term treatment protocol (Fig. 2G). We reasoned that drug pellets could be used to investigate the effect of long-term vemurafenib treatment on BRAF^V600E zebrafish melanoma and generate models of drug resistance. Similar patterns of tumour responses were observed in mouse xenograft models, which showed that human melanoma tumours regressed following daily MAPK-inhibitor treatment, and then entered a stable or residual disease stage at ∼18 days, followed by recurrent growth as drug-resistant tumours for ∼50 days (Rambow et al., 2018; Wang et al., 2018).

We first treated zebrafish with 100 mg/kg vemurafenib pellets daily and followed the tumour response over 5 weeks (Fig. 3A,B). In our experiment, we noticed that, by 4 weeks of vemurafenib treatment at 100 mg/kg, the melanomas began to regrow, so we increased the dose to 200 mg/kg at week 5 (Fig. 3B). We found that melanomas were initially responsive to the increased vemurafenib concentration, but that they again began to regrow by week 8 (Fig. 3B). Next, we treated a cohort of zebrafish with melanomas with 200 mg/kg vemurafenib for 4 weeks and found that the melanomas responded rapidly to the treatment, with drug resistance and progressive disease observed as early as 4 weeks (Fig. 3C,D), and almost all the lesions developed drug tolerance or resistant recurrent growth on the longer treatment course (Fig. 3D). These studies indicate that our drug-pellet delivery methods can model long-term treatment response adaptation, from the initial melanoma regression through to stable disease, and finally recurrent and drug-resistant disease.

To assess the on-target efficacy of vemurafenib to inhibit BRAF^V600E activity in zebrafish melanoma, we performed immunofluorescence staining of melanoma sections to assess the MAPK pathway activity using phospho-Erk1/2, a downstream target of activated BRAF signalling. Melanoma samples collected from the early-responding stage of vemurafenib pellet treatment (weeks 2 and 3) had significantly reduced levels of phospho-Erk1/2 compared to those of DMSO-treated control samples (Fig. 4A,B), indicating that vemurafenib pellets have sufficient bio-availability to target BRAF^V600E in melanoma and to lead to melanoma regression.

Next, we analysed tumours that showed melanoma recurrence on vemurafenib pellets, and found that although the responses were varied, on average, tumours exhibited increased levels of phospho-Erk1/2, consistent with those seen in patients (Manzano et al., 2016; Proietti et al., 2020) (Fig. 4A,B). Total Erk1/2 levels measured by immunofluorescence staining showed no significant changes across all samples (Fig. 4C). These results validate the on-target efficacy of the drug compound delivered by our pellet feeding method.

Zebrafish are a powerful model system for drug discovery, yet although drug treatments for embryos and larvae can be easily administered through the water, drug discovery in adult zebrafish is limited by a lack of efficient, non-invasive and long-term permissive drug administration methods. Here, we provide a new method to generate drug pellets that can be easily fed to adult zebrafish to administer a controlled and precise drug dose in a non-invasive process. We apply this method to our zebrafish BRAF mutant melanoma models (Patton et al., 2005) and demonstrate that the melanomas respond to vemurafenib drug-pellet therapy. We validate the on-target efficacy of the drug by showing a reduction in total phospho-Erk1/2 in the melanoma following treatment. Long-term studies (>2 months) demonstrate that, upon drug treatment, zebrafish melanomas undergo regression followed by recurrent disease, as seen in patients (Marine et al., 2020; Shen et al., 2020b; Travnickova and Patton, 2021). For the first time, this method enables us to model long-term melanoma drug treatment and resistance stages in zebrafish genetic melanoma models, an immunocompetent model system.

We found no toxicity or side effects from the drug-pellet method, indicating that this method supports the tenets of the 3Rs (Replacement, Reduction and Refinement) in Animal Research (https://nc3rs.org.uk/). Specifically, by reducing the animal handling and total exposure required for drug administration, our method is a Refinement for drug delivery because it minimises zebrafish stress and improves welfare.

In conclusion, we provide a drug-pellet method to administer precise doses of drugs to adult zebrafish in a non-invasive, free feeding-based procedure. The drug pellets can be individually frozen so that experiments can be controlled for batch effects, are suitable for drugs with low solubility in water (such as vemurafenib, which is hydrophobic), and provide a platform for drug combinations and screens. For a broader spectrum of application, pellet size and number can be adapted easily by modifying the 3D-printing mould and modifying the quantities of the food paste recipe. Although our experiments here focus on cancer studies in zebrafish, we expect that this method will be applicable to a wide range of zebrafish disease models and will open new doors for drug discovery within the context of complex adult zebrafish in vivo biology.

Reagents or resources used in this study are listed in Table S1.

Zebrafish were maintained in accordance with UK Home Office regulations, UK Animals (Scientific Procedures) Act 1986, under project license P8F7F7E52. All experiments were approved by the Home Office and AWERB (University of Edinburgh Ethics Committee). Zebrafish were fed daily with artemia, and fed the drug pellets in the evening (18:00-20:00).

Zebrafish were genotyped using DNA extracted from fin-clipped tissue by PCR to establish the mutant allele status tp53^M214K or mitfa-BRAF^V600E as described in our previous publications (Travnickova et al., 2019). The emergence of melanoma is usually observed in individuals aged 3-6 months. Individuals used for DMSO versus vemurafenib drug-pellet treatment in this experiment were siblings and aged 5-6 months when entering the treatment scheme. Both female and male individuals were admitted into the treatment course.

The recipe for our routine dry fish food mix consists of ZM flakes, ZM Medium Premium Granular, ZM Small Granular (all ZM Fish Food and Equipment) and Hikari MicroPellets (Kyorin Food Industries, Ltd.), mixed at a weight ratio of 2:3:2:5. Food-grade agar–agar and gelatine powder were purchased from local grocery stores. Vemurafenib (SelleckChem, CAS#918504-65-1) powder was resuspended in DMSO before mixing with fish-food paste as described.

The 3D modelling and design of the drug-pressing mould was carried out on Tinkercad (STL file in Dataset 1) followed by slicing set-up using Ultimaker Cura and 3D-printed via Ultimaker 3 with AA 0.25 generic PLA. More information about 3D modelling and printing can be found on the open resource link provided by UCreate Team website at the University of Edinburgh (https://www.ucreatestudio.is.ed.ac.uk/workshop_recordings).

Vemurafenib dissolved in DMSO or extracted from drug pellets by DMSO was measured by HPLC, using the Agilent 1260 Infinity II Prime LC platform (Agilent InfinityLab Poroshell 120 EC-C18, 1000 bar, 3×100 mm, 2.7 μm). The eluent and settings were as follows: flow rate, 1 ml/min; injection volume, 10 μl; eluent A, water with TFA (0.1%); eluent B, acetonitrile with TFA (0.1%); A/B=95:5 to 5:95 in 10 min, 5:95 isocratic for 2 min. Between each sample, a sample of methanol was measured to serve as a technical blank control. The peaks and measurement numerics were generated by the platform integrated software, Openlab CDS.

Fish were briefly anaesthetised (Tricaine in PBS 1:10,000 concentration) once every week for imaging purposes to follow the tumour burden changes during the experiment. Each fish was anaesthetised for no longer than 10 min per session and fully recovered in fresh system water. Brightfield images were taken for each fish positioned on both sides. Images of fish lesions were captured at the same magnification scale on the same microscope every week. The size of each lesion was quantified by using the manual field selection in Fiji on each tumour image, then compared to the matching pre-treatment lesion to calculate the relative percentage change. Lesions that could be observed from both sides of the fish were measured by combining the area number averaged from both sides.

Zebrafish melanoma samples were collected, fixed and processed (including H&E staining) as described in our earlier publications (Lister et al., 2014). MAPK activity was assessed using: phospho-p44/42 MAPK (ERK1/2) (Thr202/Tyr204) primary antibody (1:200; rabbit, Cell Signaling Technologies, #9101), total p44/42 MAPK (ERK1/2) primary antibody (1:200; rabbit, Cell Signaling Technologies, #9102), and Alexa Fluor 488 secondary antibody (1:1000; goat-anti-rabbit IgG, Life Technologies, #A-11034). Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; 1:1000; Life Technologies, #62248).

We are grateful to Cameron Wyatt and the Medical Research Council (MRC) Zebrafish Facility for zebrafish management and husbandry, the MRC Imaging Facility for supporting the imaging experiments, and Helen Caldwell and Elaine McLay for histology. We appreciate the constructive inputs from Cameron Wyatt and Jana Travnickova during the development of this project, and are grateful to Melissa Van de L'isle and Asier Unciti-Broceta for assistance with HPLC. We are also grateful to the uCreate Studio team of the University of Edinburgh for providing the equipment and materials for the 3D-printing self-service.