Drug Discovery in Fish, Flies, and Worms

Abstract

Nonmammalian model organisms such as the nematode Caenorhabditis elegans, the fruit fly Drosophila melanogaster, and the zebrafish Danio rerio provide numerous experimental advantages for drug discovery including genetic and molecular tractability, amenability to high-throughput screening methods and reduced experimental costs and increased experimental throughput compared to traditional mammalian models. An interdisciplinary approach that strategically combines the study of nonmammalian and mammalian animal models with diverse experimental tools has and will continue to provide deep molecular and genetic understanding of human disease and will significantly enhance the discovery and application of new therapies to treat those diseases. This review will provide an overview of C. elegans, Drosophila, and zebrafish biology and husbandry and will discuss how these models are being used for phenotype-based drug screening and for identification of drug targets and mechanisms of action. The review will also describe how these and other nonmammalian model organisms are uniquely suited for the discovery of drug-based regenerative medicine therapies.

Discovery of First-in-Class Small Molecule Drugs by Target- Versus Phenotype-Based Screens

Small molecule drug discovery efforts utilize either target- or phenotype-based approaches. Target-based strategies typically employ in vitro high-throughput screening to identify small molecules that alter the activity of an identified candidate protein implicated in a disease process. In contrast, phenotypic approaches screen for the effects of small molecules on an observable characteristic of an animal, tissue, or cell model, typically without a priori knowledge of the target. Prior to the 1990s, most drugs were discovered by phenotype screening. With the emergence of genomics and advances in molecular biology, target-based screens subsequently became the dominant mode of drug discovery (Lee and Berg 2013; Swinney and Anthony 2011; Wagner and Schreiber 2016).

Both target- and phenotype-based screening have their strengths and weaknesses. The strength of a target-based approach is that it allows for the high-throughput identification and optimization of small molecules that have desired properties. However, this approach requires the identification of a target based on existing knowledge of a disease and the mode-of-action by which a drug might treat it. Target selection is thus limited by the depth and breadth of existing knowledge. In addition, since target-based screens are not typically carried out in animal models, problems with efficacy and toxicity are often not revealed until later stages of drug development. The focus on target-based screening in the pharmaceutical industry has coincided with a decline in the number of new treatments approved for patient use. Recent studies have questioned whether target-based screening is the most successful drug discovery strategy (Lee and Berg 2013; Swinney and Anthony 2011).

Phenotype screening has the advantage of not requiring detailed understanding of a disease or of how small molecules might affect that disease. It is also performed under more physiologically relevant contexts in the whole animal or in isolated tissues and organs. Phenotype screens thus provide insight into the overall efficacy of a small molecule, because they reflect critical drug development parameters such as biodistribution, pharmacokinetics, and toxicity. Relative to target-based approaches, phenotype screening is relatively slow and optimizing lead molecules can be difficult if the target and mode-of-action (MoA) are unknown. However, MoA is not necessarily required for regulatory approval, and numerous genetic, genomic, biochemical, and computational approaches exist for identifying targets and MoAs once lead molecules have been identified in phenotype screens (Giacomotto and Segalat 2010; Wagner and Schreiber 2016; Williams and Hong 2011). Figure 1 summarizes strategies for both target- and phenotype-based drug development.

Figure 1.

Strategies for target- and phenotype-based drug discovery. (A) Target-based strategies require target selection based on existing knowledge of the genetic and molecular mechanisms underlying a disease process. Once a putative target is identified, in vitro assays amenable to high-throughput screening strategies are developed. Large numbers of small molecules are then screened for their ability to alter target activity. Once leads or “hits” are identified, they are then subjected to testing in mammalian models and undergo medicinal chemistry modification to optimize their functional properties. (B) Phenotype-based screening in small animal models such as C. elegans, fruit flies, and zebrafish require no prior detailed understanding of the mechanisms underlying a disease process. Instead, phenotype assays are developed by modeling a human disease through genetic manipulation (e.g., mutation of human disease gene orthologs, transgenic expression of human disease genes, or gene knockout) or xenotransplantation of human tumors. In addition, normal cellular and physiological processes such as cell migration, cell viability, and developmental processes can be used for phenotype assays. Once an assay is established, compounds are screened manually or by high-throughput methods. Identification of drug targets can be carried out using forward and reverse genetic approaches, genomic analyses and biochemical and computational methods.

The focus of this article is to briefly review the use of lower vertebrate and invertebrate animals in drug discovery. Zebrafish, fruit flies, and the nematode worm Caenorhabditis elegans will be specifically discussed, and examples of drug candidates discovered in these models will be provided. Most of the drug discovery efforts carried out in zebrafish, flies, and worms begin with phenotype-based screens. The relative advantages and disadvantages of these three models as well as mice for phenotype-based drug discovery are summarized in Table 1. This article will also provide examples of how nonmammalian animal models can be used for target identification, and how they are uniquely suited for target identification and drug discovery in regenerative medicine.

Table 1.

Relative advantages and disadvantages of C. elegans, Drosophila, zebrafish, and mice for drug discovery

C. elegans	Drosophila	Zebrafish	Mice
Advantages
Short life cycle Very low husbandry and animal costs Simple screening assays Ease of disease model development Amenable to high-throughput and high content screening Genetic tractability speeds target and MoA identification Unique opportunities for discovery of antiinfective compounds Mutant and transgenic strains can be archived by cryopreservation	Short life cycle Low husbandry and animal costs Simple screening assays Ease of disease model development Amenable to high-throughput and high content screening Genetic tractability speeds target and MoA identification High conservation of human disease genes Conservation of some physiological processes Unique opportunities for cancer avatar development	Low animal costs High conservation of physiological processes and genes relevant to human disease Amenable to high-throughput and high content screening Unique opportunities for regenerative medicine drug discovery Strains archived by cryopreservation of sperm	High conservation of physiological processes and genes relevant to human disease Strains archived by cryopreservation of embryos and sperm
Disadvantages
Relatively simple anatomy Limited human disease gene orthology Cuticle and intestinal physiology may limit drug permeation	Drug administration can be complicated Strains must be maintained constantly	Life cycle considerably longer than worms and flies Expensive husbandry costs Genetics and molecular biology more complicated than worms and flies	Long life cycle Very expensive husbandry costs High animal costs Relatively low throughput Genetics and molecular biology more complicated and expensive than lower vertebrates and invertebrates Overreliance as preclinical models: many drugs effective in mice do not work in humans

Overview of C. elegans, Drosophila, and Zebrafish Biology and Husbandry

The nematode worm C. elegans was introduced as a model system for biological research by Sydney Brenner in 1963 (Brenner 1988). Its genome was the first multicellular organism genome sequenced. Despite its evolutionary distance from mammals, approximately 40% of C. elegans genes have human orthologs, and the majority of its genes are homologous to human genes (Shaye and Greenwald 2011; Shim and Paik 2010).

C. elegans provides powerful experimental advantages for elucidating gene function and the molecular workings of complex processes, including forward and reverse genetic tractability, a fully sequenced and well annotated genome, and access to extensive molecular and genetic tools and online resources (Chen et al. 2015; Maglioni et al. 2016; Strange 2003) (e.g., http://www.wormbase.org/#012-34-5). A particularly powerful tool in the C. elegans toolkit is in vivo high-throughput genome-wide RNA interference (RNAi) screening. High-throughput genome-wide RNAi screening knocks down gene expression simply by feeding worms Escherichia coli bacteria producing double-stranded RNA targeting specific genes (Simpson et al. 2012).

Adult C. elegans are predominantly hermaphroditic with males making up approximately 0.1% of wild-type populations. Self-fertilized hermaphrodites produce about 300 offspring whereas male-fertilized hermaphrodites can produce over 1000 progeny.

Under optimal laboratory conditions the average lifespan of C. elegans is 2 to 3 weeks. The life cycle is rapid. At 25ºC, embryogenesis, the period from fertilization until hatching, occurs in 14 h. Postembryonic development occurs in four larval stages (L1-L4) that last a total of about 35 h.

Culture of C. elegans in the laboratory is simple and relatively inexpensive. Animals are typically grown in petri dishes on agar seeded with a lawn of E. coli as a food source. C. elegans can also be grown in mass quantities using liquid culture strategies and fermentor-like devices. Worm stocks are stored frozen in liquid nitrogen indefinitely with good viability. The ability to store C. elegans frozen dramatically simplifies culture strategies and reduces costs associated with handling and maintaining wild-type and mutant worm strains.

The anatomy of C. elegans is relatively simple. An adult C. elegans hermaphrodite is about 1 mm long and is comprised of 959 somatic cells including muscle, nerve, and intestinal cells. Despite its simple anatomy, many basic cellular physiological processes are conserved between nematodes and mammals.

The fruit fly Drosophila melanogaster was first proposed as a genetic model organism by Charles Woodsworth in 1900 (Sturtevant 1959). Like C. elegans, Drosophila provides a wealth of experimental approaches for defining gene function and the molecular bases of cellular and whole animal physiological processes. These include forward and reverse genetic tractability, extensive online resources (e.g., http://www.flybase.org), and an extensive molecular and genetic toolkit. Approximately 75% of human disease genes are estimated to have fly orthologs. The fly's complex anatomy allows for the molecular study of physiological processes that are homologous to some of those found in humans, including nervous system functions, neuromuscular junction function, and phototransduction (Millburn et al. 2016; Reiter et al. 2001; Ugur et al. 2016).

An interesting and important approach using Drosophila has been termed the “diagnostic strategy” (Ugur et al. 2016). Numerous efforts exist worldwide to identify genes underlying human disease. A bottleneck in these efforts is determination of whether an identified gene is actually causative. In flies, a homolog or ortholog of a human gene can be readily knocked out and its effect on phenotype characterized. Rescue of the mutant phenotype by the wild type but not the variant human cDNA indicates that the identified human gene variant is likely the cause of the disease (Bellen and Yamamoto 2015; Wangler et al. 2015).

Like C. elegans, the culture of Drosophila is relatively straightforward and economical. Flies are typically grown in glass bottles or vials and maintained in humidified and temperature-controlled incubators. Food consists of a mixture of water, agar, sugar, corn meal, and yeast. Reproduction occurs by mating of males and females. The fly life cycle is about 10 days at 25ºC and consists of embryogenesis, three larval stages, and pupation. Adult flies live about 40 to 50 days. No reliable method exists for archiving fly strains and they must be continuously maintained.

The zebrafish Danio rerio was introduced as a genetic model organism in the late 1960s by George Streisinger (Grunwald and Eisen 2002). Zebrafish have many experimental advantages, including forward and reverse genetic tractability, extensive online resources (e.g., http://www.zfin.org), and transparent embryos that develop outside the mother allowing for the visualization of developmental processes. Many of the organ systems of zebrafish are similar to those of mammals (Fishman 2001; Lin et al. 2016; MacRae and Peterson 2015). For example, human cardiac electrophysiology is more similar to that of zebrafish than rodents (Chi et al. 2008; Genge et al. 2016), and zebrafish have been particularly important for understanding heart disease and heart development (Wilkinson et al. 2014). The zebrafish also provided the first detailed understanding of mechanisms of heart regeneration (Foglia and Poss 2016; Porrello and Olson 2014). Pioneering heart regeneration studies in zebrafish led to the recent groundbreaking observation that the neonatal mouse heart is also capable of regenerating its heart and that it does so via mechanisms remarkably similar to those of zebrafish (Porrello et al. 2011). In addition to physiological similarities, approximately 70% of all human genes and 82% of human disease-associated genes have at least one zebrafish ortholog (Howe et al. 2013).

Zebrafish husbandry is considerably more complex than that of nematodes and fruit flies. Animals must be maintained in centralized aquaculture facilities with controlled water parameters and light/dark cycles (Lawrence and Mason 2012). Animals are typically fed brine shrimp and are supplemented with dry powder diets. Zebrafish husbandry and experimental protocols are subject to vertebrate animal research restrictions and reporting requirements including IACUC oversight in the United States (Sanders 2012).

Reproduction of zebrafish occurs by spawning. Each spawn produces about 400 progeny. Zebrafish develop from an embryo that hatches into a free swimming larval stage 48 to 72 hours after fertilization. Larvae metamorphose into juvenile fish that then mature into adults. Sexual maturity is reached approximately 3 to 4 months after fertilization. Fertility declines as the animals age, and adult zebrafish are typically maintained for approximately 2 years. Mutant and transgenic zebrafish strains are archived by freezing sperm.

Drug Screening and Discovery in C. elegans

C. elegans has a number of advantages for drug discovery, including low husbandry costs, simple screening assays, relative ease of disease model development, and amenability to high-throughput and high content screening methodology. The forward and reverse genetic tractability of C. elegans greatly speeds target identification (Chen et al. 2015; Maglioni et al. 2016; O'Reilly et al. 2014). Disadvantages include its relatively simple anatomy, which makes discovery of drugs focused on specific organ systems challenging if not impossible, and limited human disease gene orthology. In addition, the animal's cuticle and its intestinal physiology can limit drug permeation.

Despite its limitations, C. elegans has been used to identify small molecules with therapeutic potential. For example, C. elegans is an important model for identifying novel antiinfective compounds. The natural environment for C. elegans is rotting plant material where it is exposed to bacterial and fungal pathogens (Frezal and Felix 2015). Many human disease-causing pathogens also infect C. elegans (Cohen and Troemel 2015; Powell and Ausubel 2008).

One of the major advantages of using worms is the ability to identify antiinfective compounds in a living animal using high-throughput methods. Antiinfective compounds are typically identified by assessing the effects of agents on the growth or killing of a pathogen (Wohlleben et al. 2016). Development of new compounds by this approach has been slow. This is due in part to the identification of agents with nonspecific toxicity and agents with MoAs of existing antiinfective compounds.

Screening for antiinfective compounds in C. elegans has the advantages of eliminating compounds early in the discovery phase that may be toxic to the host, identifying compounds that impact pathogen virulence, and identifying compounds that enhance the innate immune response of the host. For example, Moy et al. (2006) carried out a high-throughput screen assaying survival of C. elegans infected with E. faecalis. Of a screen of 7136 synthetic compounds and natural product extracts, they identified 16 compounds and 9 extracts that increased the survival of infected worms. Many of the compounds and extracts identified had little or no effect on bacterial growth in vitro or required much higher doses to inhibit growth compared to the doses that improved worm survival. Numerous screens have subsequently been carried out in C. elegans and have identified compounds effective against other pathogens, including Candida albicans (Breger et al. 2007), Staphylococcus aureus (Kong et al. 2014), and Burkholderia cepacia (Selin et al. 2015). C. elegans has also been an important model for the identification of anthelminthic compounds (Burns et al. 2015; Peddibhotla et al. 2014)

The genetic and molecular tractability of C. elegans provides a powerful tool for defining MoAs of antiinfective agents as well as mechanisms by which pathogens develop drug resistance. For example, Pukkila-Worley et al. (2012) used transcriptome profiling, RNAi screening, and epistasis analysis to define the MoA of the small molecule RPW-24 identified in a previous high-throughout screen of 32,700 compounds (Moy et al. 2009). RPW-24 protects against Pseudomonas aeruginosa infection by stimulating the C. elegans innate immune response via the evolutionarily conserved p38 MAP kinase pathway and the ATF-7 transcription factor. Hu et al. (2010) used quantitative killing assays of wild-type and mutant C. elegans to demonstrate that Bacillus thuringiensis crystal (Cry) proteins and nicotinic acetylcholine receptor agonists act synergistically and exhibit reciprocal hypersusceptibility. Their study defines an important strategy for discovery of anthelmintic combination therapies to treat drug resistant parasite infections. Finally, Burns et al. (2015) combined high-throughput compound screening and forward genetic screens of approximately 19 million C. elegans mutants to identify an antihelmintic lead compound that inhibits complex II of the electron transport chain with nanomolar potency. Importantly, the genetic screens performed in these studies also identified a number of anthelmintic leads for which development of drug resistance appeared unlikely.

C. elegans is an important model for understanding the molecular mechanisms of aging and to identify pro-longevity compounds (Bitto et al. 2015; DiLoreto and Murphy 2015; Labbadia and Morimoto 2014). For example, Petrascheck et al. (2007) screened 88,000 chemicals to identify molecules that extend the lifespan of adults worms. This screen led to the identification of the antidepressant drug mianserin. Analysis of multiple mutant worm strains suggested that the drug was disrupting neurotransmission associated with food sensing and that this in turn led to a perceived state of reduced nutrient intake. It is well known that caloric restriction and compounds postulated to mimic caloric restriction extend lifespan in C. elegans and multiple other species (Gillespie et al. 2016).

A final C. elegans research focus worth noting is the use of transgenic and mutant worm strains to model Alzheimer's, Parkinson's, and Huntington diseases and for the discovery of potential therapeutics to treat these disorders (Chen et al. 2015; Lepesant 2015). Sleigh et al. (2011) isolated and characterized a C. elegans strain with a mutation and phenotype that mimics a mild form of spinal muscular atrophy. Using this strain and an automated phenotyping system that quantified swimming behavior, they screened 1040 chemical compounds. Two FDA-approved drugs and one novel compound were identified that rescued aspects of the mutant phenotype. Braungart et al. (2004) demonstrated that the neurotoxin MPP⁺, which causes severe Parkinson's disease-like symptoms in primates and humans, causes selective damage of C. elegans dopaminergic neurons and reduced mobility. Using automated image analysis, they further demonstrated that several known anti-Parkinson's disease drugs ameliorated MPP⁺-induced motility defects. Finally, C. elegans has been used to model tau protein neurotoxicity. Tau is a microtubule-associated protein that is found in insoluble protein aggregates in the brains of patients with neurodegenerative disorders such as Alzheimer's disease. McCormcik et al. (2013) screened a library of 1120 chemical compounds to determine their effects on mobility defects in a transgenic worm strain expressing a mutant human tau protein. This primary screen identified 16 compounds that reduced mobility defects. One of the compounds, azaperone, reduced neurotoxicity in worms and Tau aggregation in C. elegans and cultured human cells.

Drug Screening and Discovery in Drosophila

Small molecule screening assays have been developed and chemical screens performed in Drosophila for neurodegenerative diseases (Chang et al. 2008; Seguin et al. 2015; Wang et al. 2016)), polycystic kidney disease (Hofherr et al. 2016), seizure disorders (Stilwell et al. 2006; Streit et al. 2016), obesity (Gasque et al. 2013), and sleep disorders (Nall and Sehgal 2013). The advantages of Drosophila for drug discovery include its forward and reverse genetic tractability, amenability to high-throughput and high content screening, relatively low husbandry costs, high conservation of human disease genes, and conservation of multiple physiological processes (Ghaemi and Selvaganapathy 2016; Giacomotto and Segalat 2010; Yadav et al. 2016).

Drosophila has been a uniquely powerful model for understanding cancer and for developing anticancer drugs. Many oncogenes and tumor suppressor genes have been discovered and characterized in Drosophila. These include Notch, Hedgehog, Salvador-Warts-Hippo, and the Janus-activated kinase (JAK)-signal tranducer and activation of transcription (STAT) pathway (Yadav et al. 2016). Approximately 68% of human cancer genes are thought to have fly orthologs (Rubin et al. 2000).

Models of multiple cancer types, including thyroid, colorectal, leukemia, brain, neuroblastoma, glioma, and lung have been developed in Drosophila (Yadav et al. 2016). Lung cancer has been modeled in the Drosophila tracheal system, a network of tubules that supplies oxygen to tissues. Levine and Cagan (2016) induced tumor-like growth in tracheal tubes by expressing the oncogenic isoform of Ras1 alone or in combination with PTEN knockdown by RNAi (PTENi). Ras1,PTENi mutant flies die as larvae. The lethal phenotype was used to screen 1192 FDA-approved drugs. Two hits from the screen, trametinib and fluvastatin, synergistically rescued lethality and suppressed tumor formation. Similar synergistic effects were observed in human A549 adenocarcinoma cells.

Cancer stem cells (CSCs) are thought to underlie some forms of drug-resistant tumor recurrence. Developing drugs that target CSCs is a growing research area that has focused largely on in vitro screens (Park et al. 2009). A drawback of in vitro screens is the absence of the in vivo microenvironment that profoundly affects cell function and behavior. Markstein et al. (2014) circumvented this problem by screening for drugs that target CSCs in adult Drosophila. Expression of the human Raf oncogene in Drosophila intestinal stem cells generated intestinal tumors. Coexpression of Raf with luciferase allowed quantification of tumor size. A screen of 6100 compounds and 88 FDA-approved drugs identified 14 approved chemotherapy drugs and 10 uncharacterized small molecules that inhibited tumor growth. Some of the FDA-approved drugs paradoxically induced hyperproliferation of wild-type intestinal stem cells via the conserved JAK-STAT pathway, which is activated by inflammatory factors. Thus, chemotherapy-induced stimulation of wild-type stem cell proliferation may contribute to tumor recurrence. The findings of this study suggest that tumor recurrence might be reduced by combining certain chemotherapeutics with antiinflammatory drugs that inhibit the JAK-STAT pathway.

Thyroid cancer has been modeled in Drosophila by expressing the Ret oncogene in the fly eye, an epithelial tissue in which signal transduction has been extensively characterized (Das and Cagan 2013). Expression of the Ret oncogene results in cell transformation and a “rough eye” phenotype that is readily scored. The rough eye phenotype allowed for the unbiased screening of nearly 3000 genes to identify genetic modifiers of the Ret oncogene as well as the screening of small molecules. Small molecule screening combined with medicinal chemistry and detailed knowledge of key signaling pathways generated lead compounds with improved efficacy against Ret oncogene-induced transformation as well reduced whole animal toxicity.

Perhaps one of the more remarkable drug screening tools in Drosophila is the development of personalized fly avatars. Avatar mice are mice engrafted with human tumor tissue. A major limitation to the use of avatar mice for development of personalized cancer drug treatments is the time and expense required to create and maintain the animals. In contrast, creating Drosophila avatars costs far less and is much more rapid. Ross Cagan and his colleagues (Rozehal et al. 2016; Scudellari 2015) have developed a method for creating patient-specific thyroid and colorectal tumors in flies. Analysis of gene sequence data from a patient's tumor allows identification of gene variants predicted to give rise to the tumor. Transgenic fly avatars are then generated by expressing 10 to 12 of these gene variants in either the gut or the eye. The avatar flies are screened against single or combinations of 1200 FDA-approved drugs and drugs or drug combinations with the highest efficacy and lowest toxicity are then tested in clinical trials.

Drug Screening and Discovery in Zebrafish

As with worms and flies, the genetic and molecular tractability of zebrafish provides important advantages for drug discovery. In addition, zebrafish are amenable to high-throughput and high content screening methods (Ishaq et al. 2016; van der Ent et al. 2016; Yang et al. 2016) and exhibit high conservation of physiological processes and genes relevant to human disease (Fishman 2001; Howe et al. 2013; Lin et al. 2016; MacRae and Peterson 2015).

Numerous chemical screens have been carried out in zebrafish to identify molecular mechanisms underlying diverse physiological processes and to identify drug candidates for cancer, hearing disorders, muscular dystrophy, cardiomyopathies, inflammation, kidney disease, and neurological disorders (Dang et al. 2016; Fonseka et al. 2016; MacRae and Peterson 2015; Maves 2014; Robertson et al. 2014; Stewart et al. 2014). For example, hair cell death induced by aminoglycoside antibiotics and chemotherapeutics is a major cause of hearing loss (Sedo-Cabezon et al. 2014). The fish lateral line is a mechanosensory organ that contains hair cells with many similarities to mammalian inner ear hair cells (Nicolson 2005). By visualizing zebrafish hair cells with vital dyes, Owens et al. (2008) identified genes underlying antibiotic-induced cell death and screened a library of 10,960 compounds for ones that would protect hair cells from death induced by the antibiotic neomycin. Their screen identified benzothiophene carboxamides that prevent neomycin-induced hair cell death in both zebrafish and the mouse inner ear.

As noted earlier, the zebrafish has been an important model for understanding the development, function, and regeneration of the heart (Chi et al. 2008; Foglia and Poss 2016; Genge et al. 2016; Porrello and Olson 2014; Wilkinson et al. 2014). Liu et al. (2014) recently used the zebrafish to identify drugs that limit cardiotoxicity of the chemotherapeutic agent doxorubicin. In zebrafish larvae, doxorubicin causes pericardial edema, disrupts heart morphology and contractile activity, reduces blood flow, and decreases cardiomyocyte number. Using microscopy measurements of heart contraction and blood circulation as an assay of cardiotoxicity, a library of 3000 compounds was screened to identify molecules that could prevent the toxic effects of doxorubicin. This screen identified visnagin and diphenylurea as cardioprotective compounds. Both compounds rescued the effects of doxorubicin on the zebrafish heart and reduced apoptosis of cultured cardiomyocytes and cardiomyocytes in the zebrafish and mouse hearts. Visnagin also improved heart function in mice treated with doxorubicin. Affinity chromatography and inhibitor studies demonstrated that visnagin inhibits malate dehydrogenase 2, which represents a novel and druggable target for cardiomyopathy induced by doxorubicin.

Like Drosophila, the zebrafish has been an important model for understanding cancer and for developing anticancer therapies (Dang et al. 2016; Tat et al. 2013; White et al. 2013). Tumors similar to those of humans develop in zebrafish by exposure to known mutagens and by creation of transgenic lines expressing oncogenes and/or with altered expression of tumor suppressors. Zebrafish cancer models also include fish lines where processes important to tumorigenesis such as angiogenesis can be studied and used in drug screens.

Xenotransplantation of mammalian cancer cells into zebrafish can be used to study and develop drugs to angiogenesis, cancer cell invasion, and metastasis (Veinotte et al. 2014; Zhang et al. 2015). Zebrafish provide numerous advantages compared to mice for xenotransplantation. These include increased speed, greatly reduced costs, and better in vivo imaging resolution. Cancers modeled in zebrafish include skin, colorectal, pancreatic, liver, blood, testicular, glial, and neural cancers.

Several cancer drug screens have been carried out in zebrafish. For example, White et al. (2011) used zebrafish embryos expressing the human BRAF oncogene and lacking the tumor suppressor p53 to identify compounds that inhibit expansion of neural crest progenitor cells, which give rise to melanoma. In a screen of 2000 small molecules, a predicted inhibitor of dihydroorotate dehydrogenase blocked progenitor expansion. The known but structurally unrelated dihydroorotate dehydrogenase inhibitor leflunomide, which is used to treat arthritis, had the same effect. Leflunomide combined with a BRAF inhibitor suppressed melanoma growth in xenograft mice. Clinical trials of leflunomide combined with the BRAF inhibitor vemurafenib for treatment of melanoma are ongoing.

Gallardo et al. (2015) utilized a transgenic zebrafish line expressing green fluoresecent protein in posterior lateral line primordium cells that undergo a well-defined migration pattern during embryonic development. Migration of these cells can be quantified by automated imaging of embryos arrayed in 96-well dishes. Screening of 2080 compounds identified 165 molecules that disrupted migration without causing obvious toxicity, which was assessed by the presence of developmental defects. Selected molecules were subsequently screened for their ability to disrupt cell migration and invasion in cultured melanoma cells. Three known Src family kinase inhibitors were identified in these screens. CRISPR/Cas9-targeted mutagenesis demonstrated that the Src pathway was the target of these molecules in zebrafish. In vivo assays demonstrated that the Src inhibitor SU6656 significantly inhibited tumor metastasis in mice.

Finally, Williams et al. (2015) screened ~30,000 compounds for their ability to disrupt embryonic patterning in zebrafish embryos. The screen identified eggmanone and structurally related compounds. Analysis of mutant zebrafish demonstrated that eggmanone inhibits Hedgehog signaling, a well-known driver of tumor progression. The single FDA-approved Hedgehog signaling inhibitor as well as inhibitors in clinical trials target the G protein-coupled receptor Smoothened (Smo), which functions downstream of the sonic hedgehog ligand. Smo mutations that induce resistance to these inhibitors (Yauch et al. 2009) as well as oncogenes downstream of Smo have limited the efficacy of these compounds. In vivo and in vitro assays demonstrated that eggmanone inhibits phosphodiesterase 4 (PDE4) and thus represents a new lead for treating Hedgehog signaling-dependent cancers.

Beyond Worms, Flies, and Fish: Comparative Approaches for Target Identification and Drug Discovery in Regenerative Medicine

Regenerative medicine is the field of medical research and practice that focuses on replacing, engineering, or regenerating human cells, tissues, and organs to restore function. The US Department of Health and Human Services has called regenerative medicine the “vanguard of 21st-century healthcare” (2003). Much of the field has to date centered on development of cell, gene, and tissue engineering-based therapies. Despite extensive research, these potential therapeutic strategies remain challenged by their complexity and by efficacy and regulatory hurdles (Abujarour and Valamehr 2015; Chen et al. 2016; Willerson 2015).

Discovery and development of small molecules capable of activating innate tissue repair and regenerative processes is a newly emerging field in regenerative medicine. Small molecules have multiple advantages compared with other regenerative medicine therapeutic strategies, including greatly reduced complexity, likely lower treatment costs, reduced regulatory hurdles, ready reversibility of the therapy, and lack of ethical concerns. However, small molecule discovery and development has to date been constrained by limited understanding of the molecular mechanisms underlying regenerative processes (Längle et al. 2014).

Arguably, the most economical and efficient strategy for development of small molecules with regenerative medicine applications is to use nonmammalian animal models for both target- and phenotype-based drug discovery. Countless invertebrate and lower vertebrate animals including Drosophila and zebrafish exhibit remarkable regenerative capabilities (Birnbaum and Sanchez 2008; Tanaka and Reddien 2011). The zebrafish in particular is able to fully regenerate many lost or damaged body parts (Gemberling et al. 2013; Goessling and North 2014). In contrast, humans and other mammals have limited capacity for regenerating damaged tissues even though they possess the genetic instructions needed for building tissues and organs de novo during embryogenesis.

Identification of potential regenerative medicine drug targets can be obtained by detailed characterization of regenerative processes in nonmammalian animals. For example, studies of spinal cord regeneration in the developing chick identified peptidylarginine deiminase 3 as a potential neuroprotection therapeutic target (Kin Pong et al. 2014; Lange et al. 2011). Lisse et al. (2016) used the zebrafish to model peripheral neuropathy induced by the widely used cancer chemotherapeutic agent paclitaxel. Wild-type zebrafish readily regenerate damaged peripheral neurons, but paclitaxel causes permanent neuronal damage. Paclitaxel-induced neurotoxicity is associated with increased expression of matrix-metalloproteinase 13 (MMP-13). Pharmacological inhibition of MMP-13 in zebrafish rescues the neurotoxic effects of paclitaxel, suggesting a therapeutic strategy for treating peripheral neuropathy in chemotherapy patients. Finally, limb regeneration has been studied extensively in salamanders, and multiple signaling components required for limb regenerative processes have been identified (Tanaka 2016). Detailed knowledge of these mechanisms provides the foundation for developing therapeutic strategies for enhancing tissue repair in limbs and multiple organs.

A systematic comparison of tissue damage and regenerative responses across multiple regenerating and nonregenerating species would likely accelerate regenerative medicine drug target discovery (Figure 2). For example, King and Yin (2016) recently used RNA-seq transcriptome profiling to identify a conserved genetic circuit that controls appendage regeneration in zebrafish, dragon fish (bichir), and salamanders, species that diverged in evolution ~420 million years ago. O'Meara et al. (2015) profiled gene expression changes in in vitro and in vivo models of cardiomyocyte differentiation, an in vitro cardiomyocyte explant model, and in the regenerating neonatal mouse heart. Their studies identified interleukin 13 as a novel regulator of cardiomyocyte cell cycle entry and demonstrated that STAT6, STAT3, and periostin are critical components of interleukin 13 signaling.

Figure 2.

Strategies for discovery and development of drugs that activate tissue repair and regenerative processes. (A) Comparative genomic analyses of injury responses in several strategically chosen regenerating and nonregenerating animal models as well as cell and tissue culture systems can be used to identify genetic circuits and signaling pathways that control repair and regenerative processes. Once these pathways are experimentally validated, they can be used to inform target identification followed by assay development and in vitro compound screening. (B) Phenotype-based screens can be carried out in regenerating animal models. Because of its genetic and molecular tractability, widespread use in drug discovery and robust regenerative capabilities the zebrafish is particularly well suited for regeneration phenotype screens. Screens can be performed to identify compounds that enhance normal regenerative processes or that restore regeneration after it has been inhibited by chemical or genetic means. Once lead molecules are identified through either approach shown here, they would then be tested in mammalian models and modified to improve their properties as shown in Figure 1.

To facilitate regenerative medicine drug target identification using comparative biology strategies, a novel NIH-supported bioinformatics resource, The Comparative Models of Regeneration Database (http://regendb.org/), is being developed by scientists at the MDI Biological Laboratory. The focus of the database is to provide an integrative understanding of regenerative biology and stem cell self-renewal by integrating gene function data across multiple animal, tissue, and cell models. The long-term goals of the database are to validate and inform hypotheses needed to discover and develop new regenerative medicine therapies.

Phenotype-based screening represents a viable strategy for regenerative medicine drug discovery (Figure 2). The genetic and molecular tractability, widespread use in drug discovery, and robust regenerative capabilities of zebrafish make it well suited for such efforts. Several screens relevant to regenerative medicine have recently been carried out. Zon and co-workers screened 2480 small molecules to identify pathways that modulate haematopoietic stem cell (HSC) induction and homeostasis in zebrafish (North et al. 2007). Their studies demonstrated that prostaglandin E2 (PGE2) signaling plays a central role in HSC formation and that dimethyl-PGE2 (dmPGE2) administration enhances HSC formation in zebrafish and mouse models. In follow-up studies, they demonstrated that dmPGE2 enhances human and nonhuman primate HSC formation in vitro and in vivo (Goessling et al. 2011). This work subsequently led to a Phase 1b safety study in patients undergoing umbilical cord blood transplantation for leukemia or lymphoma (Cutler et al. 2013). The encouraging efficacy and safety results of this study have now led to ongoing Phase 2 clinical trials of dmPGE2.

Wang et al. (2015) used a transgenic larval zebrafish line in which insulin-producing pancreatic β-cells are labeled with yellow fluorescent protein to robotically screen 3348 compounds for their ability to increase β-cell mass. Twenty-four drugs stimulated β-cell proliferation. Eighteen of these leads are drugs already approved for human use. Mechanism of action studies identified new roles for NF-κB and serotonergic signaling in β-cell differentiation and proliferation.

Tsuji et al. (2014) also used a fluorescent transgenic larval zebrafish line to identify small molecules capable of increasing zebrafish β-cell mass. A screen of 883 small molecules indentified 20 lead compounds. Two drug classes, retinoic acid and glucocorticoids, increased the regeneration of β-cells after β-cell ablation.

As noted earlier, hair cells in the zebrafish lateral line have properties resembling hair cells in the mammalian inner ear (Nicolson 2005). Using transgenic zebrafish expressing GFP in lateral line hair cells, Namdaran et al. (2012) screened two chemical libraries containing a total of 1680 compounds for their ability to stimulate regeneration of neomycin-ablated hair cells. Two enhancers, the synthetic glucocorticoids dexamethasone and prednisolone, increased hair cell proliferation both in the presence and absence of neomycin-induced damage.

Finally, using zebrafish caudal fin regeneration as a screening assay, Yin and co-workers recently carried out a small molecule screen to identify compounds capable of stimulating complex tissue regeneration (A Smith, K Nguyen, TA Rando, M Zasloff, K Strange, VP Yin, unpublished data). The caudal fin is comprised of bone, nerve, vasculature, connective, and skin tissues, fully regenerates within 10 to 14 days following amputation, and the rate of fin regeneration is readily quantified (Gemberling et al. 2013). Their screen identified the protein tyrosine phosphatase 1B inhibitor MSI-1436. This small molecule stimulates zebrafish caudal fin and heart regeneration 2- to 3-fold, stimulates adult mouse heart regeneration following ischemic injury induced by permanent coronary artery ligation, and stimulates stem cell activation in injured mouse skeletal muscle. MSI-1436 has previously been tested in Phase 1b clinical trials as a potential obesity and type 2 diabetes treatment and was shown to be well tolerated by patients. The doses effective at stimulating tissue regeneration in zebrafish and mice are 50-times lower than the maximum well-tolerated human dose.

Conclusions

Nonmammalian model organisms are powerful tools for understanding the molecular and genetic mechanisms underlying human disease and for drug discovery. They provide particularly unique opportunities for target identification and lead drug candidate discovery for regenerative medicine. It goes without saying that all models including nonmammalian and mammalian animal systems, cell and tissue cultures, and in vitro assays have both strengths and weaknesses for understanding human disease and for drug discovery. An interdisciplinary approach that strategically combines the study of nonmammalian and mammalian animal models with diverse experimental tools is an important investment that can improve the understanding of a disease, its therapeutic targets, drug toxicity, and mechanisms of drug action. This in turn can reduce the probability of drug failure and associated high costs. As eloquently stated by an unnamed reviewer of this article, “in poker, it takes skill, practice, cunning, etc., to successfully play a ‘good hand’ and collect the pot. If dealt a ‘bad hand,’ the only hope is to recognize the fact early and ‘fold’ before wasting additional resources (time, money, animals). And unlike in poker, there is no sense in ‘bluffing’ in pharmaceutical development.”