Drug screens using the nematode Caenorhabditis elegans

Abstract

Since its inception as a model system, Caenorhabditis elegans has provided insight about the mechanism of action of drugs through genetic analyses. With the arrival of diverse drug-like small molecule libraries sometime later, the worm also became a platform for drug discovery that was previously inaccessible to academics. Here, the history of larger-scale drug screens using C. elegans is reviewed. The current approaches used to identify the targets and targeted pathways of the novel hits from these screens are also discussed. We focus on the development of small molecule tools for biological investigation, the discovery of novel candidate nematicides and anthelmintics, and touch on screens related to other areas of biology, including neurodegeneration. Finally, we draw attention to the fundamental aspects of C. elegans biology that lends itself to chemical genetic research. When combined with diverse small molecule libraries, the worm's tractability and genetic power make it an unparalleled whole-animal model system for early-stage drug discovery.

This invited review by Roy examines the use of Caenorhabditis elegans as a model organism in drug discovery and chemical genetics. Highlighting its advantages as a cost-effective, whole-animal drug-screening platform, it surveys over 9,600 published studies to identify key research trends. The article revisits Sydney Brenner’s seminal 1974 drug screen, reviews large-scale efforts that uncovered novel candidate drugs, anthelmintics, and nematicides, and shows how genetics can reveal drug targets. By integrating historical context with recent advances, the review underscores the enduring value of C. elegans as a small animal model for early-stage drug research.

Introduction

Why would anyone want to screen drugs against the model nematode Caenorhabditis elegans? Especially nowadays, with such a diversity of biological systems that include human organoids, ingenious CRISPR screens with mice, and zebrafish models of disease aplenty. Despite the many options, biologists typically publish more C. elegans drug studies each year than the year before (Fig. 1). There must be compelling reasons why investigators keep returning to the lowly worm. This review aims to provide insight and perspective into this question.

Fig. 1.

The number of C. elegans drug studies at PubMed over time. In black are the numbers of publications identified on PubMed using keywords “Caenorhabditis elegans” or “C. elegans” and “drug,” “drugs,” “small molecule,” “small molecules,” “compound,” or “compounds” tracked on the left axis as a function of time on the x axis. There are a total of 9,620 papers of this type. In pink are those identified with keywords “Caenorhabditis elegans” or “C. elegans,” the numbers of which are tracked on the right axis. There are a total of 44,720 papers of this type, which include all 9,620 drug-related papers. In each search, the PubMed queries were phrased using appropriate Boolean syntax to yield nonredundant results. Note the dip in the number of C. elegans drug-related papers published after the onset of the COVID-19 pandemic, which is not reflected in the overall numbers of C. elegans papers. Over the past decade, more than 27% of C. elegans papers published included a drug study of some type.

It might be useful to first understand the subject areas in which C. elegans is employed in the context of drug studies. Toward this end, the titles and abstracts of all 9,620 papers involving C. elegans drug studies were collected from PubMed and analyzed. The top 3 subjects focus on longevity, neurobiology, and antimicrobials (Fig. 2). What each of these subjects has in common is that they require an intact whole animal for study. Linda Buck's screen for compounds that extend the life of animals (Petrascheck et al. 2007), our neuromuscular-based screen for candidate nematicides (Harrington et al. 2022), and Fred Ausubel's screen for compounds that defend C. elegans against Enterococcus faecalis (Moy et al. 2006) would be next to impossible without the ability to screen whole animals, and we all did so in a high-throughput manner.

Fig. 2.

A keyword subject analysis from the titles and abstracts of the 9,620 C. elegans drug papers in PubMed. Titles and abstracts of the C. elegans drug papers described in Fig. 1 were downloaded and processed for the occurrence of all words using the Natural Language Toolkit (NLTK) library for Python programming language (Bird et al. 2009) (kindly processed by Matej Usaj from Charlie Boone's lab). A total of 213 common nonrelevant words (and, the, for, etc.), numbers, and bits shorter than 2 characters were automatically filtered. The resulting list of 78,432 terms was manually curated, and similar high-ranking terms were grouped together (eg brain, neuron, nervous system, neural, axon were grouped together into “nervous system”). The top 40 words are indicated with a few related groups color-coded as indicated.

These examples capitalize on a key set of features that make C. elegans such an attractive option for drug studies. The worm's small size (it is ∼1 mm long as an adult), its ability to self-fertilize, and its quick three day life cycle allow us to probe its entire biological program over all life stages for small molecule sensitivities in 96, 384, and 1,532 well plates. This power is unparalleled with any other cell, tissue, or animal model system.

A second key feature of C. elegans is that the entire genetic program of 19,824 protein-coding genes is packaged within an exceptionally simple and transparent body plan. The laboratory wildtype strain is also isogenic. The implications are that phenotype is more readily observable, reproducible, and accessible with fluorescent markers compared to other animal models.

A third feature that makes C. elegans so tractable is that it is exceptionally fecund. Each hermaphrodite generates about 300 replacements in 3 d. With sufficient Escherichia coli food, it would take a single C. elegans parent about 42 d to generate enough progeny equivalent in mass to the Earth's moon. In other words, 100,000 s of animals can easily be generated for screens for drug-resistant mutants, which has repeatedly revealed the mechanism-of-action of drugs.

The final feature introduced here is that drug screens with C. elegans are inexpensive. First, growing worms costs a small fraction compared to culturing human cells, organoids, or zebrafish, for example. Second, worms can be cultured in volumes as little as 50 ml over many days, and unlike other animal model systems, the media does not require refreshment. This saves on both drug and media costs. Third, thousands of samples per week can be processed manually with an inexpensive dissection microscope, although more sophisticated data capture methods exist. Hence, drug studies with C. elegans are more accessible in general compared to many other systems.

Several reviews have been published about drug studies with C. elegans. These include reviews on the use of C. elegans to characterize the mechanism of action of anthelmintics and nematicides (Holden-Dye and Walker 2014; Hahnel et al. 2020; Nixon et al. 2020; Martin et al. 2021), reviews on high-throughput drug discovery and microfluidics (Buckingham et al. 2014; O'Reilly et al. 2014; Zamanian and Chan 2021; Yoon et al. 2024), reviews about antimicrobial drug discovery (Kim et al. 2017; Peterson and Pukkila-Worley 2018), and reviews on natural products (Garcia-Bustos et al. 2019; Ha et al. 2022). There is little point in repeating what has already been written. Instead, this review focuses on the roots of C. elegans chemical genetic analyses, large-scale screens of more than 400 molecules, and how C. elegans genetics has been useful in the characterization of novel hits from these screens.

Brenner's drug screen raises the question of drug sensitivity

The first reported drug screen with C. elegans was performed by the late Nobel laureate Sydney Brenner and published in his seminal 1974 Genetics paper (Brenner 1974). In it, he provided no rationale for the screen. One can only speculate what his underlying motivation was. A decade earlier, Brenner, together with Francis Crick and others, exploited the prolific nature of phage to isolate mutagen-induced suppressors of the rII mutant phenotypes, and suppressors of these suppressors. With brilliant logic, the triplet nature of the genetic code was revealed by combining these suppressors (Crick et al. 1961). Perhaps inspired by this earlier work, Brenner might have sought to show that his new model system was also capable of genetically bending to selective pressure that a drug could provide and was prolific enough to allow the detection of mutants of a desired type. In his fervor to genetically characterize as many loci as possible with his new model organism, Brenner might have also wanted to explore alternative approaches to isolate distinct mutants that might not otherwise be revealed on plates without drugs.

Brenner screened roughly 100 unnamed compounds that were readily available to him and found only 2 molecules that elicited obvious C. elegans phenotypes. He commented that, “most are without effect on the worms, probably because they are not absorbed.” This statement assumes that all 100 molecules could in principle elicit a phenotype, which is likely not true for several reasons (eg target divergence, target redundancy, and target nonessentiality). Despite this caveat, our previous work showed that only a low percentage of molecules (96 of the 1027 surveyed) accumulate robustly within the worm, which is not inconsistent with the spirit of Brenner's statement (Burns et al. 2010).

The typical assumption that is made to explain Brenner's low hit rate or the low accumulation rate is that the cuticle is a major barrier to small molecule absorption. Cuticles do reduce drug sensitivity of the worm, but only modestly so. For example, mutants that disrupt either the exocuticle (eg bus-5 and agmo-1) or the pharynx cuticle (eg pgp-14) increase drug sensitivity only 2-fold (Xiong et al. 2017; Kamal et al. 2023; Rehborg et al. 2023). It turns out that the exocuticle covers a little less surface area than the luminal membrane of the intestine because the latter is covered in microvilli, each of which has a surface area of 0.47 μm² (Zhu et al. 2022). We can estimate the total surface area of an adult hermaphrodite to be roughly 251,000 μm² (1,000 μm length × 2πr, with r = 40 μm). By comparison, the surface area of the intestinal apical (luminal) membrane in an adult hermaphrodite is about 261,000 μm² {(21,250 μm² intestinal lumen surface area, assuming an intestinal length of 850 μm and a luminal perimeter of 25 μm) + [a surface area of 239,700 μm² for the microvilli (24 microvilli/μm² × 21,250 μm² intestinal lumen surface area = 510,000 microvilli, each with a surface area of 0.47 μm²)]}. Given that the pharynx pumps liquid at 4 Hz (Avery and You 2012) and that the residence time in the gut is about 2 min (Ghafouri and McGhee 2007), there is ample opportunity for small molecule absorption via the intestine, regardless of the structural integrity of the cuticle. Another route of small molecule absorption is the 26 sensory neurons that are exposed to the environment through cuticular holes and readily take up hydrophobic molecules (Inglis et al. 2007). Hence, the post-embryonic animal is not encased within a fortress that is largely impenetrable to small molecule absorption.

Given the multiple routes for small molecule absorption, how can we rationalize the relatively low accumulation rate that we observed? C. elegans lives within the microbial world (Schulenburg and Felix 2017). It thrives in the environment of some microbial species and perishes in the presence of others who produce noxious natural products (Samuel et al. 2016). This selective pressure has likely driven the expansion of a robust small molecule detoxification system. The animal has a greatly expanded number of nuclear receptors (267) called nuclear hormone receptors in C. elegans, most of which are orphan receptors that are speculated to respond to environmental compounds (Sural and Hobert 2021) and other debilitating xenobiotics (Lindblom et al. 2001; Jones et al. 2015). By contrast, there are only 48 nuclear receptors in humans (Frigo et al. 2021). C. elegans encodes 76 cytochrome P450s (compared to 57 in humans), many of which have been shown to oxidize xenobiotics as part of phase I drug metabolism (Larigot et al. 2022). The worm also has a wide array of phase II detoxification enzymes (Stasiuk et al. 2019). Finally, C. elegans encodes 14 full P-glycoproteins (PGP) (compared to 4 in humans), many of which have been implicated in xenobiotic and metabolite export from cells (Lespine et al. 2024). Hence, the low hit rate in Brenner's screen and in the accumulation survey may have less to do with absorption than with a robust xenobiotic detoxification system.

At least 3 classes of mutants that compromise xenobiotic defense and increase C. elegans drug sensitivity have been discovered (Table 1). These include mutant genes that disrupt the exocuticle barrier (Gravato-Nobre et al. 2005; Partridge et al. 2008; Xiong et al. 2017), those discovered to disrupt the pharynx cuticle barrier (Kamal et al. 2019, 2023), and others that reduce drug export (Ardelli and Prichard 2013; Jones et al. 2015; Menez et al. 2019; Guerrero et al. 2021). Creating strains that combine all 3 mutant classes might prove additionally useful. While these sensitized mutants might yield additional hits in drug screens, screens focused on identifying novel nematicides or anthelmintics would likely not employ a sensitized background because hits should be capable of circumventing the natural xenobiotic barriers of nematodes.

Table 1.

Mutant genes that confer drug sensitivity.

Drug-sensitizing mutant gene	Mechanism of drug sensitivity	Demonstrated hypersensitivity	Health on nondrug plates	Citations
Disruption of the exocuticle barrier
acs-20; acs-22	Disruption of the exocuticle	Bleach, clofiliuim, E-4031, Hoechst 33342, ibutilide, levamisole, osmotic imbalance	Small, dumpy	Kage-Nakadai et al. (2010); Jiang et al. (2018)
agmo-1	Disrupted production of cuticle lipids	Aldicarb, bleach, fenoxycarb, piperazine, 2-phenoxyethanol, toluene, valproic acid, warfarin	Slow growth, fragile cuticle	Gravato-Nobre et al. (2005); Loer et al. (2015); Xiong et al. (2017)
bus-5	Disruption of the exocuticle	Aldicarb, bleach, boric acid, fenoxycarb, 2-methoxyethanol, piperazine, 2-phenoxyethanol, toluene, valproic acid, warfarin	Skiddy movement and cuticle fragility, poor long-term survival	Gravato-Nobre et al. (2005); Xiong et al. (2017)
bus-8	Disorganized cuticle and hypodermis	Bleach, ivermectin, nicotine, 1-phenoxypropan-2-ol	Slow growth; reduced fecundity	Gravato-Nobre et al. (2005); Partridge et al. (2008); Xiong et al. (2017)
bus-16	Likely disrupted cuticle organization; exact mechanism unknown	Aldicarb, piperazine, valproic acid	Skiddy movement, vulval rupture, poor long-term survival	Gravato-Nobre et al. (2005); Xiong et al. (2017)
bus-17	Likely disrupted cuticle organization; exact mechanism unknown	Bleach	Reduced fecundity, poor long-term survival	Gravato-Nobre et al. (2005); Xiong et al. (2017)
dpy-2, dpy-3, dpy-7, dpy-8, dpy-9, and dpy-10	Exocuticle integrity	Hoechst 33342, ivermectin, levamisole, paraquat	Shorter, some mutants roll, some mutants are uncoordinated	Sandhu et al. (2021)
cat-4, ptps-1	Needed for synthesis of BH4, a cofactor for AGMO-1	Bleach, SDS, levamisole	Cuticle fragility	Loer et al. (2015)
ceh-60	Disrupted exocuticle epicuticle structure	Acridine orange, hydrogen peroxide	Healthy	Van de Walle et al. (2019)
gmap-1	Reduced exocuticle lipid barrier	Levamisole, Na azide, Hoechst 33342, dehydration	Healthy	Njume et al. (2022)
Disruption of the Pharynx Cuticle Barrier
pgp-14	Reduced lipid barrier in pharynx cuticle	188 lethal molecules, Hoechst 33342	Healthy	Kamal et al. (2023)
sms-5	Reduced lipid barrier in pharynx cuticle	169 lethal molecules	Healthy	Kamal et al. (2019)
Disruption of Drug Efflux
nhr-8	Regulation of the expression of drug efflux pumps	Ivermectin, tunicamycin	Healthy	Menez et al. (2019); Guerrero et al. (2021)
pgp-1	Likely drug efflux	Thiabendazole	Healthy	Jones et al. (2015)
pgp-2	Likely drug efflux	Ivermectin	Healthy	Ardelli and Prichard (2013)
pgp-3	Likely drug efflux	colchicine, chloroquine, thiabendazole	Healthy	Broeks et al. (1995); Jones et al. (2015)
pgp-5	Likely drug efflux	Ivermectin, tunicamycin	Healthy	Ardelli and Prichard (2013); Guerrero et al. (2021)
pgp-6	Likely drug efflux	Ivermectin	Healthy	Ardelli and Prichard (2013); Menez et al. (2019)
pgp-7	Likely drug efflux	Ivermectin	Healthy	Ardelli and Prichard (2013)
pgp-11	likely drug efflux	Tunicamycin	Healthy	Guerrero et al. (2021)
pgp-12	Likely drug efflux	Ivermectin, tunicamycin	Healthy	Ardelli and Prichard (2013); Guerrero et al. (2021)
pgp-13	Likely drug efflux	Ivermectin	Healthy	Ardelli and Prichard (2013)

Despite the above discussion, it would be a mistake to leave the impression that C. elegans is exceptionally resistant to small molecule perturbation compared to other model systems. Yeast have 16 drug efflux pumps (Celaj et al. 2020) and often require high drug concentrations for screens (Lee et al. 2014); uniform drug exposure with flies can be difficult to achieve (Vidal et al. 2024), and solubilizing hydrophobic molecules in zebrafish's purely aquatic media can be limiting (Dash and Patnaik 2023). Brenner's 2% hit rate is actually very impressive, especially because we do not know the identity of the other 98 tested molecules, whether those molecules have targets in nematodes, how far those targets have diverged in nematodes relative to their intended target, and what the potency, if any, those 98 substances have in biological systems. Similarly, the finding that fewer than 10% of small molecules accumulate to micromolar concentrations in the worm may seem limiting, but most human drugs have potency in the low nanomolar range [50 nM on average (Gleeson et al. 2011)], so robust accumulation is not a prerequisite for activity. Identifying good hits in any system is a challenge, and likely no more so in C. elegans than in other model systems.

Brenner isolates the first C. elegans drug-resistant mutants

As introduced above, a key advantage of C. elegans as a model system is the ability to rear millions of animals in a short period of time. Given a sufficiently negative selection pressure, this allows for the identification of drug-resistant mutants through powerful selection schemes [see (Jorgensen and Mango 2002) for an excellent overview of genetic screens with C. elegans]. As described previously, mutant genes that confer resistance to a bioactive compound can reveal the target or targeted pathway, irrespective of whether the drug agonizes or antagonizes its target (Burns et al. 2006) (Table 2).

Table 2.

Compound classes and mutant targets that confer drug resistance.

Compound	Compound class	Exemplar gene and allele in target or pathway that confers resistance	Class of mutation that can yield resistance	Citation
Amino-acetonitrile derivatives (monepantel)	Agonist	acr-23(cb79)	RoF or LoF	Kaminsky et al. (2008)
Aminophenylamidines (tribendimine)	Agonist	unc-29(e193)	RoF or LoF	Hu et al. (2009)
Apigenin	Unknown	daf-16(mu86)	RoF or LoF	Kawasaki et al. (2010)
Avermectins (ivermectin)	Agonist	avr-14(ad1305); avr-15(vu227); glc-1(pk54)	RoF or LoF	Dent et al. (2000)
Benzimidazoles (benomyl)	Antagonist	ben-1(e1880)	RoF or LoF	Driscoll et al. (1989)
Carbomates/organophosphates (lannate, aldicarb)	Antagonist	cha-1(p1152)	RoF	Brenner (1974)
Cyclooctadepsipeptides (emodepside)	Agonist	slo-1(js379)	RoF or LoF	Guest et al. (2007)
Dihydropyridines (nemadipine)	Antagonist	egl-19(tr92)	Neomorphic	Kwok et al. (2008)
Imidazothiazole (levamisole)	Agonist	unc-29(x29)	RoF or LoF	Fleming et al. (1997)
Nemacol	Antagonist	unc-17(md414)	Neomorphic	Harrington et al. (2023)
Succinate dehydrogenase inhibitors (fluopyram, wact-11)	Antagonist	sdhc-1(tr393)	Neomorphic	Burns et al. (2015)
Tetrahydropyrimidines (pyrantel, morantel)	Agonist	unc-29(e193)	RoF or LoF	Sleigh (2010)

Neomorphic mutations can confer resistance by altering the ability of the compound to interact with the target. Hypermorphic mutations can confer resistance by increasing the target's activity. RoF, reduction of function; LoF, loss of function; both of which can suppress phenotypes induced by an agonist if in the target or targeted pathway.

Despite the awesome power of C. elegans genetics, screens for resistant mutants can be fruitless in some scenarios. For example, a drug's target may not be mutable to a resistant state without the resulting mutation debilitating the animal (like the carbamates described below). In a second example, a drug may engage multiple targets simultaneously to yield a phenotype. Random mutagenesis is highly unlikely to yield resistant mutants in this case because all drug targets would have to be mutated simultaneously in 1 individual mutant. Given that the mechanism of action is not known in advance, it is impossible to predict whether a bioactive compound will yield informative drug-resistant mutants.

With each of the 2 bioactive compounds that Brenner identified in the first C. elegans drug screen, he was remarkably lucky to isolate resistant mutants for both. One of the 2 molecules, lannate (aka methomyl), is a carbamate. The mode of action of carbamates was discovered in 1921 by Otto Loewi, who showed that a natural product carbamate called physostigmine inhibits acetylcholinesterase (AChE) activity. Loewi won a Nobel Prize related to this discovery in 1936 (McCoy and Tan 2014). Brenner isolated multiple recessive lannate-resistant alleles of the unc-17 gene, which Jim Rand's group showed 19 years later to encode the nematode vesicular acetylcholine transporter (VAChT) (Alfonso et al. 1993). Given that carbamates increase acetylcholine levels at the synapse via AChE inhibition, it is easy to understand how reducing the loading of acetylcholine into presynaptic vesicles confers some level of resistance to this class of drugs. While Brenner's screen did not lead to the discovery of the lannate target, it did ultimately contribute to an understanding of neuronal function, which was the overarching motivation of Brenner's genetic characterization of C. elegans.

Brenner also isolated 2 classes of mutants that resist the paralytic effects of the anthelmintic tetramisole, which is a racemic mixture of 2 enantiomers, including the more potent and commonly used worm anesthetic levamisole. One class of mutants was twitchers that mapped to the unc-22 gene, which are thought to confer resistance through indirect mechanisms. Brenner mapped the other classes of nontwitching mutants to 4 different genes, including unc-29, unc-38, unc-63, and a gene later named lev-1. Of course, methodologies for identifying the sequence of these genes would not be invented until sometime later. Brenner disciple Jim Lewis would spend the better part of 2 decades isolating and characterizing additional levamisole-resistant mutants. In 1997, Lewis's group identified the genomic sequences of lev-1, unc-29, and unc-38, showing them to be nicotinic acetylcholine receptor (nAChR) subunits (Fleming et al. 1997). The group expressed all 3 subunits together in Xenopus oocytes and demonstrated levamisole-induced current that was suppressible with known nAChR antagonists. The full levamisole receptor, including 5 distinct receptor subunits and 3 ancillary proteins, would later be functionally reconstituted by the Bessereau and Richmond groups 11 years later (Boulin et al. 2008). Hence, in the early days of C. elegans chemical genetics, it took 23 years to go from isolating the first drug-resistant mutants to identifying the target components revealed by those mutants. As discussed below, this can now be done in about 60 d with the skills of an undergraduate student.

The first large-scale small molecule screens in C. elegans

The cloning revolution of yeast, worm, and fly genes gained steady momentum throughout the 1980s and 1990s. With it came the revelation that many of the genes in these simple organisms not only had orthologs in humans but controlled orthologous signaling pathways and analogous developmental programs. These organisms truly could be exploited to model our own biology. At around the same time, new techniques in the world of chemistry lowered the cost of synthesizing compounds while increasing structural diversity [ie combinatory chemistry methodology (Blackwell et al. 2001; Galloway et al. 2010)]. In 1997, it was also recognized that most successful drugs fall within a narrow range of physicochemical properties (Lipinski et al. 1997). The scope of chemical synthesis for new drug discovery could now be sharply focused, and libraries of small molecules with drug-like properties became commercially available at this time.

This new chemistry collided with model organism biology in the late 1990s with the founding of biotech startups DevGen, NemaPharm, Exelixis, and Elixir Pharma. These firms focused on drug development using model organisms that included C. elegans (eg Wells 1998; McCarthy 2005). Academic papers describing small molecule screens performed at these companies do not exist, but screens were central to their business model. Beyond these private efforts, there were no published large-scale C. elegans screens of small molecule libraries between the time of Brenner's inaugural screen and papers published 3 decades later.

In the early 2000s, a colleague asked if we were interested in screening their library of 10,000 uncharacterized molecules from ChemBridge Corporation against the worm so that we could compare notes with their screen being carried out against Arabidopsis (Zhao et al. 2007). The idea was exciting. We fantasized about identifying a suite of molecules that would yield obvious “Brenner phenotypes.” These molecules would disrupt the key developmental signaling pathways and might be powerful tools for chemical genetic analyses, not just in C. elegans but in other systems as well.

Given that the vast majority of C. elegans investigations are performed on solid agar substrate, we felt it important to screen the compounds by co-culturing them with animals reared on solid agar substrate in 24-well plates so that we could visualize the emergence of any obvious plate phenotypes. This would also avoid any of the biological peculiarities of worms cultured in liquid (Lev et al. 2019). We ended up screening 14,100 molecules that included other libraries. Through manual inspection of the cultures using a dissection microscope, we identified 308 compounds that induced obvious phenotypes in the worm (Kwok et al. 2006). The majority of hits induced slow growth, embryonic lethality, or larval lethality. At the time, we did not investigate these types of hits further, but some were later included in our worm active (aka wactive) collection that we explored for years.

Among the 308 hits, there was only 1 molecule that elicited an obvious set of post-embryonic phenotypes initially described by Brenner. In addition to slow growth, the hit that we renamed “nemadipine” induced variable morphological abnormalities (aka Vab) in young larvae and an egg-laying defective (aka Egl-d) phenotype in adults (Table 3). Nemadipine is discussed in more detail below in the context of using phenotype to uncover a drug's target.

Table 3.

Large-scale small molecule screens with C. elegans and key resulting compounds.

Main molecule of interest	Phenotype	Mode of action and target	Potential utility	# of molecules in initial screen (and primary vendor)	Citation
Nemadipine	Egg retention (Egl-d), slow growth (Gro), morphological abnormalities (Vab)	Inhibits CaV1 ion channels (EGL-19)	Tool compound	14,100 (ChemBridge Diverset)	Kwok et al. (2006)
Mianserin	Lifespan extension	Suppression of transcriptional drift via modulation of serotonergic signaling	Lifespan extension	88,000 (ChemBridge CNS, ChemBridge Diverset)	Petrascheck et al. (2007); Rangaraju et al. (2015)
Dafadine	Constituative Dauer (Daf-c); distal tip cell migration defects (Mig); lifespan extension	Inhibits cytochrome P450 (DAF-9, aka CYP-22A1)	Tool compound	4,710 (yeast bioactives)	Luciani et al. (2011)
Migrazole	Distal tip cell migration defects (Mig); bivulva (Biv); axon extension defects	Unknown target; chemical genetic interactions with multiple developmental signaling pathways	Tool compound	1,570 (yeast bioactives)	Kwok (2011)
HBAC-B	Uncoordination (Unc); premature starvation-induced death (Sid)	Unknown target; chemical genetic interactions with metabotropic glutamate receptor	Tool compound	4,000 (clickable library assembled by Sean Cutler)	Burns (2013)
ML358	Decreased transcription of SKN-1 targets	Inhibition of SKN-1 pathway	Nematicide	36,4000 (NIH Molecular Libraries Small Molecule Repository)	Leung et al. (2013); Peddibhotla et al. (2015)
NH125	Suppresses microbial-induced killing of C. elegans by MRSA persisters	Disruption of MRSA biofilms	Antimicrobial	85,000 (ICCB Harvard Collection)	Kim et al. (2015)
wact-11	Lethal (Let)	inhibits nematode complex II (succinate dehydrogenase)	Nematicide/anthelmintic	67,012 (ChemBridge NovaCore and Diversets)	Burns et al. (2015)
CID 2747322	Lethal (Let)	Inhibits nematode complex II (succinate dehydrogenase)	Nematicide/anthelmintic	25,986 (ChemBridge and Maybridge)	Mathew et al. (2016)
NP1	Lifespan extension	Promotion of glutamate signaling	Tool compound	30,000 (unknown origin)	Lucanic et al. (2016)
Auranofin	Suppresses microbial-induced killing of C. elegans	Inhibition of the mia40-Erv1 pathway	Antimicrobial	640 (Biomol 4 library, Enzo Life Sciences)	Fuchs et al. (2016); Thangamani et al. (2017)
C22	Embryonic lethality (Emb)	Dependent on LET-607 activity	Tool compound	37,000 (Yale Pilot library, MicroSource Gen-Plus and Natural Products, ChemBridge and Maybridge Diversets)	Weicksel et al. (2016)
wact-86	Lethal (Let); synergistic with aldicarb	Unknown target; synergy is mediated through GES-1 inhibition	Nematicide	67,012 (ChemBridge NovaCore and Diversets)	Burns et al. (2017)
Pimozide	Suppression of motor defects induced by heterologous expression of human TDP-43	T-type calcium channels	Treatment of ALS	3,750 (LOPAC 1280, Spectrum, Biomol, and Prestwick)	Patten et al. (2017)
Prostratin and ingenol-3,20-dibenzoate (repurposed)	Suppression of trafficking-induced suppression of motor defects induced by heterologous expression of a human-worm chimeric hERG potassium channel	Agonism of protein kinase C epsilon	Treatment of long-QT syndrome (LQTS)	10,600 (Prestwick Chemical Library, Spectrum Collection, the Natural Products Library, the ICCB Known Bioactives Library, the IB screening Library, the NIH Clinical Collection, the FDA-Approved Drug Library, the Simga-Aldrich LOPAC, and the AnalytiCon Discovery)	Jiang et al. (2018)
TRVA242	Suppression of motor defects induced by heterologous expression of human TDP-43	Unknown or not public; possibly inhibition of T-type calcium channels	Treatment of ALS	3,765 (Treventis, Zalicus)	Bose et al. (2019)
wact-190	Larval lethal (Lvl)	SMS-5, PGP-14-dependent Crystallization in the pharynx cuticle	Tool compound	67,012 (ChemBridge NovaCore and Diversets)	Kamal et al. (2019)
Rifabutin (repurposed)	Suppression of motor defects induced by heterologous expression of human α-synuclein	Unknown	Treatment of Parkinson's	620 (characterized drugs predicted by IBM Watson for Drug Discovery; Predictive Analytics for inhibiting α-synuclein aggregation)	Chen et al. (2021)
HF-00014 (aka UMW-9729)	Lethal against C. elegans and H. contortus	Unknown	Nematicide/anthelmintic	14,400 (HitFinder, Maybridge)	Taki et al. (2021); Shanley et al. (2024b)
GSK306886A	Larval arrest (Lva); internal hatchlings of animals (Bag)	LET-23 (EGFR)	Nematicide/anthelmintic	2,040 (GSK Published Kinase Inhibitor Sets 1 and 2, APExBIO DiscoveryProbe Kinase Inhibitor Library; Sigma-Aldrich LOPAC; PICR Kinase Inhibitor Library)	Knox et al. (2021)
Trametinib	Sterile (Ste)	MEK-2 (MEK1/2)	Nematicide/anthelmintic	2,040 (custom Kinase Inhibitor Collection—details above)	Knox et al. (2021)
GSK1520489A	sterile (Ste); embryonic lethality (Emb)	PLK-1 (polo-like kinase 1)	Nematicide/anthelmintic	2,040 (custom Kinase Inhibitor Collection—details above)	Knox et al. (2021)
Crotamiton and JM03 derivative	Lifespan extension, stress resistance	OSM-9	Tool compound	1,027 (clinically used drugs)	Bao et al. (2022)
Nementin (aka wact-55)	Lethal (Let); hyperactive; convulsions; potently synergistic with aldicarb	Unknown target; agonizes dense core and synaptic vesicle release	Nematicide	486 (wactive library)	Harrington et al. (2022)
MMV1581032 (aka ABX464)	Lethal (Let)	Tentatively HCON_00074590	Nematicide/anthelmintic	400 [Pandemic Response Box from Medicines for Malaria Venture (MMV)]	Shanley et al. (2022)
Dexrazoxane	Inhibition of proliferation of multiple microsporidia species in C. elegans	Unknown	Antimicrosporidia	2,560 (MicroSource Spectrum Library)	Murareanu et al. (2022)
Nemacol (aka wact-45)	Uncoordinated; coiler; paralysis	Vesicular acetylcholine transporter (VAChT) (UNC-17)	Nematicide	486 (wactive library)	Harrington et al. (2023)
Selectivin (aka wact-1)	Lethal (Let)	Bioactivated into lethal reactive product by CYP-35C1	Nematicide	67,012 (ChemBridge NovaCore and Diversets)	Burns et al. (2023)
Closantel	pharynx crystal suppressor and amyloid disruptor	Inhibits secondary a-beta filament formation in vitro	Amyloid disruption	2,560 (MicroSource Spectrum Library)	Kamal et al. (2024)
NPD8790	Hypoxia-dependent lethality	Inhibits nematode complex I (NADH-ubiquinone oxidoreductase)	Anthelmintic	480 (RIKEN Natural Product Depository)	Davie et al. (2024)
Cyprocide (aka wact-4-5)	Lethal (Let)	Bioactivated into lethal reactive product by CYP-35D1	Nematicide	846 (custom library of disubstituted oxodiazoles)	Knox et al. (2024)
Avocatin A (aka avocado fatty alcohols/acetates)	Lethal (Let)	inhibition of acetyl-CoA carboxylase (ACC, aka POD-2)	Nematicide/anthelmintic	2,320 (MicroSource, Sigma-Aldrich, Cayman)	Fahs et al. (2025)
Benzene-sulfonamides	Inhibition of proliferation of multiple microsporidia species in C. elegans	Unknown	Antimicrosporidia	2,900 (ChemBridge Diverset, Pandemic Response Box)	Huang et al. (2023); Huang et al. (2025)

Included are screens of 400 molecules or more that have revealed a previously uncharacterized hit or a previously characterized hit with novel function, and where at least one of the hits was characterized in detail with respect to its mechanism-of-action or structural constraints.

While nemadipine induced an exciting set of phenotypes, the fact that we obtained only 1 hit that induced a nonlethal post-embryonic phenotype was a disappointment. To achieve our goal of assembling a suite of molecules that induce Brenner phenotypes by screening only 14,100 compounds was obviously naïve in retrospect. Over time and with several different screening approaches, however, the community has established a collection of small molecule tools that not only facilitate biological analyses but also have potential commercial utility (Table 3).

At about the same time as our 2006 publication, Fred Ausubel's group published their screen of 7,136 molecules and extracts for novel antimicrobial candidates using their pioneering C. elegans model of E. faecalis infection (Moy et al. 2006). Their screen was innovative on multiple fronts, including that it was performed in liquid media in 96-well plate format. A year later, Linda Buck's group published their screen of 88,000 compounds for those that extend C. elegans lifespan and identified a series of antidepressants that can do so (Petrascheck et al. 2007). In addition to screening an impressive 6-fold more compounds than anyone before, the group showed that C. elegans screens could be performed in 384-well plate format. Later, Keith Choe's group published the largest C. elegans drug screen to date. They screened 364,000 compounds for those that inhibit the transcription factor SKN-1, which is a key mediator of nematode oxidative stress and a potential anthelmintic target (Leung et al. 2013). Impressively, they used 1,536-well plates and a fluorescent readout to achieve ultrahigh-throughput processivity. The group identified a selective inhibitor of nematode SKN-1 called ML358 that fails to inhibit the human ortholog NRF2 and serves as an anthelmintic candidate (Peddibhotla et al. 2015) (Table 3).

Specific phenotypes can suggest specific drug targets

C. elegans has a relatively simple body plan for an animal (Sulston and Horvitz 1977; Sulston et al. 1983) and has been genetically and phenotypically characterized extensively. Hence, there are many obvious phenotypes that are highly indicative of the underlying biochemical pathway perturbed. In the case of nemadipine, a limited number of genes can be mutated to an Egl-d phenotype, and we knew of only 1 Egl-d mutant that also had a Vab phenotype: reduction of function mutants in the egl-19 gene that encodes the sole C. elegans L-type calcium channel α₁-subunit ortholog (Lee et al. 1997). We therefore had a good hypothesis that nemadipine inhibits EGL-19 function. In parallel to chemical genetic tests with existing egl-19 alleles, we carried out a forward genetic screen for resistant mutants and identified multiple dominant alleles of egl-19 as anticipated (Kwok et al. 2006, 2008; Hui et al. 2009). Nemadipine belongs to the well-characterized dihydropyridine (DHP) class of L-type calcium channel (Ca_v1) inhibitors (Fig. 3). What is useful about nemadipine is that it was the only known DHP that could inhibit EGL-19 within whole animals; other DHPs only work when dissected animals are bathed in the compounds, indicating that nemadipine escapes the ADME (absorption, distribution, metabolism, and excretion) liabilities of other DHPs in C. elegans.

Fig. 3.

Hits that resemble characterized molecules. Hits are on the left (ordered according to their discovery and/or publication date), and previously characterized molecules are on the right. The Jaccard–Tanimoto similarity score between the left and right molecules is in brackets after the names on the left. The score was calculated using RDKFingerprints on the NovoPro online tool. A score of 1.0 is a perfect match, and a score of 0 indicates no structural similarity.

This phenotype-driven hypothesis approach has been repeatedly useful in identifying the target of small molecule hits from our screens. For example, we discovered a new structure called dafadine that induces a constitutive dauer formation (Daf-c) phenotype and a defect in the migration of the distal tip cell (Mig), which results in a clear patch on the ventral side of animals. Mutations in only 2 genes were known to yield that combination of penetrant phenotypes: daf-9 and daf-12 (Gerisch et al. 2001). We went on to provide evidence that supports the hypothesis that dafadine inhibits the cytochrome P450 DAF-9, which makes the hormone that regulates the DAF-12 nuclear receptor (Luciani et al. 2011). Dafadine also extends C. elegans' lifespan. In another example, a novel structure that we discovered called nemacol induces a tight coiling phenotype highly reminiscent of reduction-of-function mutants in the aforementioned UNC-17 VAChT and the CHA-1 choline O-acetyltransferase that produces acetylcholine. Again, we went on to provide multiple lines of evidence that supports the hypothesis that nemacol inhibits UNC-17 (Harrington et al. 2023).

The phenotype-to-hypothesis-to-target approach has been successful, but not always so. Through a screen of 1,570 molecules that were found to be bioactive in the yeast Saccharomyces cerevisiae, we identified 1 molecule that we called migrazole that induced distal tip cell migration defects, a bivulva phenotype, and axon extension defects (Kwok 2011). Despite this specific combination of post-embryonic phenotypes, extensive chemical genetic analyses, and genetic screens for resistant mutants, we failed to identify a candidate target that could satisfactorily explain migrazole's suite of phenotypes. In a similar example, we found 1 molecule from a screen for egg-laying modulators that we call nementin that also induces hyperactivity and convulsions (Harrington et al. 2022). This suite of phenotypes led us to the realization that nementin agonizes dense core and synaptic vesicle release, but despite the robust phenotypes and genetic screens for resistance, we could not identify a specific target. A similar theme emerged for HBAC-B (Table 3) (Burns 2013). We speculate that in all of these cases, the bioactive small molecule may have multiple targets, which confounds the use of genetics to reveal the target. Hence, phenotype can lead to target but, in just as many cases, can also leave the researcher empty-handed.

Screens for novel nematicides

Some have argued that C. elegans is an imperfect model for anthelmintic/nematicide discovery because it lacks several evolutionary innovations adopted by parasitic species. Indeed, if one can screen against the parasite itself, then they should do so. Several groups have provided beautiful examples of this direct approach (Marxer et al. 2012; Preston et al. 2016, 2017a, 2017b; Pasche et al. 2018; Maccesi et al. 2019; Pasche et al. 2019; Wheeler et al. 2023; Elfawal et al. 2025). However, it is undeniable that C. elegans genetics has been instrumental in revealing the molecular mechanisms of some of the most important broad-spectrum nematicidal molecules ever discovered, including the Nobel Prize-winning ivermectin (Cully et al. 1994; Dent et al. 2000), the amino-acetonitrile derivatives (aka AADs, aka monpantel) (Kaminsky et al. 2008), benzimidazole (Driscoll et al. 1989), levamisole (Brenner 1974; Fleming et al. 1997; Boulin et al. 2008), and emodepside (Guest et al. 2007; Crisford et al. 2011). Other nematicidal molecules also have demonstrable activity against C. elegans, including the derquantel precursor (Ondeyka et al. 1990; Ruiz-Lancheros et al. 2011; Koizumi et al. 2023) and the spiroindolines (Sluder et al. 2012). One can easily make the argument that each of these molecules could have been discovered using C. elegans.

In the early 2010s, we embarked on a large screening campaign that would dominate our focus for some time. Colleagues gave us access to 2 of their 50,000-molecule ChemBridge libraries. We prescreened one of these libraries in S. cerevisiae yeast at a high concentration to identify those with bioactivity and then screened the resulting bioactives against C. elegans (Wallace 2011; Luciani 2014). We screened the remaining 50,000-molecule library and molecules from a few other sources directly against worms (Luciani 2014). In total, we screened 67,012 distinct structures against C. elegans in 96-well plate liquid culture manually and identified 627 molecules that induced obvious phenotypes. We re-purchased these hits from vendors to assemble our custom worm active (wactive) library and gave each molecule a trivial name with a “wact” prefix followed by a unique number. We then screened the wactive library against other nematodes, parasitic or not, as well as cultured human cells and zebrafish to determine which molecules might have nematode selectivity. We ended up with 30 structurally distinct scaffolds as candidate nematicides/anthelmintics (Burns et al. 2015).

We reasoned that the most efficient approach to prioritize which of our hits could be understood mechanistically was by determining which might yield resistant mutants. Systematic genome-wide RNAi screens have previously been employed to reveal potential drug targets (Saur et al. 2013), but RNAi's low-throughput relative to forward genetic selection screens, together with RNAi's inability to generate hypermorphic or neomorphic mutations, which are important for revealing the target of inhibitors, made RNAi screens a less desirable option.

We carried out a large-scale campaign of forward genetic selection screens against our lethal hits and by 2015, completed screens against 39 molecules with a minimum of 50,000 mutant genomes screened, typically in both F1 and F2 screens, against each molecule (Burns et al. 2015). An additional 56 molecules were screened against thereafter (Pyche 2019; Harrington 2021). Screens were attempted with an additional 25, but the lethality of these molecules did not fully translate onto the solid substrate format that was used for genetic screens. In the end, robustly resistant mutants were isolated against 8 molecules, which included dafadine (Luciani et al. 2011), wact-86 (Burns et al. 2017), 2 crystal-forming compounds (Kamal et al. 2019; Kamal et al. 2023), and 4 wact-11 family members (Burns et al. 2015) (Table 3).

Our wact-11 study provides a good example of the power of genetic screens to provide mechanistic insight. A total of 33 wact-11 resistant mutants were isolated, and their whole genomes were sequenced. Astonishingly, 32 of the 33 mutants had mutations in 3 different components of the same complex; 10 had missense mutations in succinate dehydrogenase complex subunit B (sdhb-1); 16 had missense mutations in sdhc-1; and 6 had missense mutations in sdhd-1. Together, they defined the ubiquinone binding pocket within complex II of the electron transport chain (ETC). The mutations also revealed the likely binding site of the wact-11 molecule (Burns et al. 2015). Furthermore, several of the nematode residues that can be mutated to confer strong resistance to wact-11 are divergent in mammals, thereby explaining wact-11's nematode selectivity. A few years earlier, Bayer patented a highly related scaffold for nematicidal use and marketed it as fluopyram (Fig. 3) (Greul et al. 2013). Hence, the wact-11 narrative illustrates the power of C. elegans to reveal molecules with real-world utility.

When chemical genetic screens work, whole genome sequencing of a large collection of resistant strains will often reveal multiple alleles within the same genes, which circumvents the laborious mapping of traits. Our experience indicates that these genes typically encode either the target or the components of the targeted pathway. Examples of this include nemadipine, wact-11, and wact-190 (Table 3). These experiments can be executed over the course of about 2 months and are so simple that we have adapted the approach to undergraduate lab classes. It is a testament to the progress of the field of genetics that we went from taking 2 decades of work by experienced geneticists to a couple of months of experimentation by inexperienced undergraduates to uncover the target of a small molecule.

Despite the success of wact-11, our genetic screening campaign yielded resistant mutants for only 8 of 120 bioactive molecules. This is a far cry from Brenner's 100% success rate. While all 8 led to exciting stories, it was clear that approaches beyond genetic screens are needed to understand how most bioactive compounds perturb C. elegans biology.

Second generation screens for candidate nematicides and anthelmintics

After exhausting forward genetic screens, my group developed alternative strategies to characterize the wactives that resisted the genetic screen approach.

The Egl modulator screen

We hypothesized that some of our molecules that kill larvae over 6 d might induce other phenotypes when assayed in different contexts. Identifying those that induce specific motor phenotypes could lead to better hypotheses about their mechanism of action. We therefore established a collaboration with an engineering colleague to build an imaging instrument that measures the ability of C. elegans to lay eggs (aka embryos) (Au et al. 2025). The neuromuscular system that governs egg-laying is controlled by cholinergic, serotoninergic, GABAergic, and peptidergic inputs (Schafer 2006) and is therefore sensitive to a wide variety of perturbations. A screen of our wactive library revealed 58 egg-laying (Egl) modulators, about half of which also induced other motor phenotypes including convulsions, shaking, and coiling (Harrington et al. 2022).

Returning to our “phenotype to hypothesis to target” mantra, the convulsions induced by wact-55 led to an understanding that the molecule agonizes presynaptic vesicle release and has potential utility in selectively enhancing the killing of nematodes by the nonselective acetylcholine esterase inhibitors (Harrington et al. 2022). Similarly, the coiling phenotype induced by the wact-45 family directly led to its vesicular acetylcholine transporter target (Harrington et al. 2023), which is an established anthelmintic target (Sluder et al. 2012). Fruit remains on the Egl-modulator tree, including several molecules that induce convulsions.

The P450 bioactivation screen

Inspired by our earlier work on drug metabolism in the worm (Burns et al. 2010) and the structural similarities of one of our wactive molecules (wact-1) to cytochrome P450-bioactivated molecules, we investigated whether wact-1 is metabolized into a lethal product by one or more of the worm's 76 P450s (Larigot et al. 2022). The oxidation of small molecules by P450 s can pull electrons away from neighboring atoms, making them electrophilic and reactive, which in turn can incapacitate the cell through their interaction with essential nucleophiles (Guengerich 2006). We showed this to be true for wact-1, and together with the support of collaborators, we demonstrated real-world nematicidal potential for the molecule, which we renamed selectivin (Burns et al. 2023).

Our early results inspired us to rescreen the wactive library for other molecules whose lethal effects might be dependent on P450 bioactivation. We did this using established methods to incapacitate the cofactor needed for all microsomal P450 activity (Harlow et al. 2018) and asked whether the wactive lost killing potency. The screen revealed additional P450-bioactivated molecules, including wact-4, which we developed into cyprocide (Knox et al. 2024), and others that we continue to work on.

The crystal screen

From our very first drug screens, we noticed that worms co-incubated with select small molecules develop dark objects near the lumen of the pharynx. Others have reported this phenotype as well [see Fig. 6c in Risi et al. (2019) and Galford and Jose (2020)]. Our curiosity about the dark objects was reignited upon finding that we were able to isolate mutants that resisted both the lethality and the object formation, one of the object-inducing molecules called wact-190 (Kamal et al. 2019). We showed that the dark objects are crystalline in nature (ie they are birefringent in vivo and can form a crystal lattice in vitro), are at least partially composed of the exogenous small molecule, and are seeded within the pharynx cuticle and grow to pierce the underlying epithelium, which is at least in part responsible for their ability to kill young larvae (Kamal et al. 2019, 2023).

At the genetic level, the activity of a sphingomyelin synthesis pathway and the PGP-14 export pump, expressed specifically by the myoepithelium beneath the pharynx cuticle, are necessary for the deposition of polar lipids into the cuticle, which is in turn necessary for the accumulation of the hydrophobic crystalizing molecules (Kamal et al. 2022, 2023).

A detailed screen of 197 wactives revealed that 48 form birefringent crystals. The lethality and crystal formation of 23 of these is suppressed upon mutating the PGP-14 pump (Kamal et al. 2019, 2023). We infer that the remaining 25 crystal-forming wactives have multiple mechanisms by which they can kill the worm.

While we continue to exploit the crystal-forming molecules as tools, the nematicidal utility of crystalizing wactives has yet to be fully explored. In preliminary analyses, we find that wact-190 can form crystals in the pharynx cuticle of some species but not others (Kamal 2023). Nematicidal utility will likely depend on the molecular composition of the species' pharynx cuticle.

The kinase inhibitor screen

In parallel to our work on the wactive library, we assembled a collection of 2040 molecules targeting vertebrate kinases for the purpose of identifying small molecule scaffolds that might be modified to generate selectivity against essential nematode kinases. The pipeline consisted of first identifying vertebrate kinase inhibitors that induced phenotypes in C. elegans that are consistent with the reduction-of-function phenotype of the respective orthologous kinase in C. elegans. We then asked whether the corresponding worm kinase might have a drug-binding pocket that is different in some way from the vertebrate kinase and whether these differences are conserved among parasitic nematodes. With these types of hits, we could in principle modify the initial hit to generate nematode selectivity. We identified nematode EGFR (LET-23), MEK1/2 (MEK-2), and PLK1 (PLK-1) kinases as not only being druggable in the worm but also having distinct differences in their drug-binding pockets that make them good candidate targets for anthelmintic development (Table 3) (Knox et al. 2021).

Novel approaches to identify candidate nematicides and anthelmintics

Several technologies have been developed or employed to advance high-throughput small molecule screening with C. elegans [see Table 4 in Hahnel et al. (2020) with additional highlights shown in Table 4]. Those that have been exploited to screen 400 or more molecules for nematicidal potential and yielded hits that were characterized in detail are described here.

Table 4.

Technologies and protocols that aid throughput of C. elegans small molecule screens.

Technology or protocol	Innovation	Notable publications related to higher-throughput C. elegans drug assays
Animal preparation for screens
Vital dyes	The use of vital dyes to measure the percentage of worms alive in a well	Gill et al. (2003); James and Davey (2007); Ferreira et al. (2015); Phiri et al. (2017)
COPAS Biosort	a worm analyzer and dispenser with multiple utilities, including dispensing worms into multiwell plates for high-throughput screens	Kwok et al. (2006); Mathew et al. (2016); Dilks et al. (2021); Wit et al. (2021); O'Brien et al. (2025)
Live-animal FACS (aka, laFACS)	The use of FACS to sort fluorescently labeled (eg, GFP-expressing) animals based on particular properties (eg genotype) for small molecule screens	Fernandez et al. (2010)
BioTek MicroFlo	Rapid deposition of animals into wells	Leung et al. (2013)
Innovative screening methodologies
Lifespan Extension Assay	High-throughput analysis of lifespan extension in 96-well plate format	Solis and Petrascheck (2011); Lucanic et al. (2018)
Bacterial Ghosts (using dead E. coli to Feed C. elegans)	allows for more reproducible results in high-throughput assays because worm phenotypes are not confounded by overgrowth of E. coli	Dranchak et al. (2023)
Custom Microfluidic Devices	Lower throughput, but innovations allow for unparalleled analyses of single animals over time	Carr et al. (2011); reviewed in Yoon et al. (2024)
Microscope-Based Imaging and Data Capture
Phenalysis Microplate Imaging	A high-throughput imager for the analysis of various traits, including movement	Spensley et al. (2018); Davie et al. (2024)
INVAPP/Paragon Imager	Measures movement of animals in the wells of a 96-well plate at a rate of 100 plates per hour	Partridge et al. (2018)
Microscope in a Box (eg Molecular Devices ImageXpress Nano microscope)	One of several versatile systems for high-throughput and high-content image collection and analyses	Gosai et al. (2010); Shaver et al. (2023); Fahs et al. (2025)
Other Data Capture Methods
PhylumTech's WMicrotracker	A multiwell plate reader that measures movement based on infrared signal disruption	Simonetta and Golombek (2007); Schmeisser et al. (2017); Risi et al. (2019); World Health Organization (2020); Taki et al. (2021); Alberich et al. (2025)
WormScan	a flatbed line scan imager exploited to measure lethality, reduced fecundity, slow growth, and immobility	Mathew et al. (2012); Mathew et al. (2016)
InVivo Biosystems Electropharyngeogram Microfluidic Chip	Rapid determination of a compound’s effect on the neuromuscular activity of the pharynx of free-living and parasitic nematodes	Lockery et al. (2012); Weeks et al. (2018)
2020 Imager	Rapid higher resolution line scanning instrument	Harrington et al. (2022); Au et al. 2025 (in review)
Gustatory Microplate	A microfluidic device that reveals compounds that are attractive and noxious to free-living and parasitic nematodes in a high-throughput manner	Nunn et al. (2023)
Laser-Scanning Cytometry	Rapid high-content screening with fluorescently labeled (GFP reporters, etc.) C. elegans samples	Dranchak et al. (2023)

While most C. elegans drug screens had previously focused on long-term effects of compounds after several days of drug treatment, Andy Fraser's group developed a system to measure acute effects on the scale of minutes to hours. The approach, called Phenalysis Microplate Imaging (PMI), is carried out in 96-well plate format and measures worm movement (Table 4) (Spensley et al. 2018). Two successive images of a well are captured, separated by 500 ms, and pixel displacement from 1 image to the next is measured. Data from 96 wells can be captured in ∼5 min. This allows the user to determine the acute impact of molecules on worm movement in a relatively high-throughput manner. Using this technology, the group discovered that C. elegans employs a rhodoquinone-dependent electron transport chain (ETC) to thrive in little to no oxygen, which is conveniently mimicked by cyanide treatment that blocks the ETC from using oxygen as an electron acceptor (Del Borrello et al. 2019). This alternative to a ubiquinone-dependent ETC allows the use of diverse electron acceptors such as fumarate and is an architectural solution for dealing with hypoxic stress that is shared by only a few types of animals, including soil-transmitted helminth (STH) (nematode) parasites but not their mammalian hosts. Hence, cyanide-treated C. elegans serves as a model for STH viability within the hypoxic milieu of their mammalian hosts. The group went on to use their technology to screen 480 structurally distinct small molecule scaffolds for those able to selectively kill cyanide-tolerant C. elegans (Davie et al. 2024). They identified a novel class of compounds that contain a benzimidazole core and are species-selective inhibitors of complex I of the ETC (Table 3). Excitingly, their lead molecules translated to adult nematode parasites and serve as candidate anthelmintics.

Using an inexpensive flatbed scanner, Don Moerman and colleagues were able to screen roughly 26,000 compounds for their ability to kill, paralyze, or reduce the brood size of C. elegans in a high-throughput manner (Mathew et al. 2016). Their approach, called WormScan (Table 4), relies on 2 successive scans of the wells of a 96-well plate over a total of 2.5 min. Changes in the images for a particular well, or lack thereof, over successive scans reveal phenotype (Mathew et al. 2012). The first WormScan screen yielded 5 compounds with IC50s in the low micromolar range, including one (2747322) that likely inhibits complex II of the electron transport chain as revealed through a forward genetic screen for resistant mutants (Table 3). 2747322 is similar in structure to the aforementioned wact-11 and fluopyram complex II inhibitors. Other hits from the screen have yet to be investigated in detail and remain good nematicidal candidates.

The WMicroTracker instrument, which measures the ability of sample content to interrupt infrared light, has been employed to identify compounds that perturb C. elegans development and movement (eg Liu et al. 2018; Risi et al. 2019). For example, Robin Gasser's group has used the WMicroTracker to screen 14,400 synthetic small molecules in 384-well plate format at a reported throughput of ∼10,000 compounds per week (Taki et al. 2021). A total of 40 hits were identified that reduce adult motility by 70% or more. One of these hits was characterized in detail and was found to translate to the ruminant parasitic nematode Haemonchus contortus (Taki et al. 2021; Shanley et al. 2024b). The Gasser group used the same technology to screen the 400 drug-like compounds from the Pandemic Response Box simultaneously against C. elegans and H. contortus (Shanley et al. 2022). Three hits had low micromolar activity against both species, one of which was characterized in detail including the identification of an aldo-keto reductase (HCON_00074590) as a candidate target in H. contortus (Table 3) using proteomics-based methodology (Shanley et al. 2024a).

There are many excellent options for microscope-in-a-box technologies that come with powerful software for high-content high-throughput analyses of samples in multiwell plate format. The Piano and Gunsalus groups employed one of these, the Cellinsight CX5 instrument from Thermo Fisher Scientific, to screen 2320 compounds from MicroSource Discovery Systems Inc against C. elegans and another free-living nematode species Pristionchus pacificus in parallel (Table 4) (Fahs et al. 2025). The groups discovered a family of molecules derived from avocados, called avocado-derived fatty alcohols/acetates (AFAs), that induce a suite of detrimental phenotypes against both nematode species at low micromolar concentrations while doing little against cultured human cells (U2-OS) or mice (Table 3). A series of chemical genetic experiments indicated that the POD-2 acetyl-CoA carboxylase is likely a key target of the AVAs and its disruption is at least partially responsible for the observed phenotypes. Remarkably, the molecules translate well against multiple parasitic species, H. contortus, Teladorsagia circumcincta, and Heligmosomoides polygyrus. Given that 1 representative structure (Avocatin B) has a favorable safety profile in humans (Ahmed et al. 2019), the AFAs are a promising class of candidate anthelmintics.

While the new screening technologies described above drive discovery and can dramatically increase throughput, the technology's expense can also be a barrier to entry for some. It is important for those without extensive resources to note that with an obvious phenotype, a simple dissection microscope, a multichannel pipette, access to a library of compounds, and a manual drug pinning tool, a throughput of thousands of molecules per week per person can still be achieved.

Beyond nematicides

The investigation of small molecule crystal formation within the pharynx of C. elegans, as strange as crystals may be, led us down a surprising path. In addition to the forward genetic screen that revealed the role of PGP-14 and a sphingomyelin synthesis pathway, we performed a screen of characterized drugs and natural products from the Spectrum library (MicroSource Inc) for molecules that could also suppress crystal-induced death.

Screening molecules with characterized modes of action can sometimes reveal information about a process beyond that which can provided by genetic screens. For example, a small molecule may agonize a target that cannot be mutated to a gain of function or may simultaneously disrupt a family of paralogous proteins that are highly unlikely to be mutated coincidentally in a founding suppressor mutant. Identifying compounds with known mechanisms of action that can suppress a process of interest, like crystal formation within the pharynx cuticle, can immediately lead to hypotheses about how the suppression is occurring and, in turn, lead to new biological insights.

Of the 2,560 spectrum molecules screened, we identified 85 that suppressed crystal-induced death. A search of chemoinformatic databases revealed that 33% of the unique scaffolds from the 85 molecules were known amyloid disruptors, which was 10-fold more than expected by random chance (Kamal et al. 2024). Additional experiments in our lab and that of our collaborators supported the idea that amyloid-like material within the pharynx cuticle was important for seeding crystal formation. This very unexpected result suggested that other molecules identified in the screen might have amyloid-disrupting potential as well, which our collaborators showed to be true (Kamal et al. 2024). Hence, the strange phenomenon of crystal formation fortuitously led to the discovery of a facile in vivo pipeline that yields candidate amyloid suppressors. Given the prevalence of amyloid-based diseases, this work has the potential to lead to real-world utility.

Obviously, we stumbled onto this neurodegeneration-related story. As illustrated in Fig. 2, neurodegeneration-based drug studies are one of C. elegans' most popular uses. A good example of how C. elegans can be exploited in this regard comes from Alex Parker's and Pierre Drapeau's labs. The Parker group generated a worm model of ALS by expressing human mutant TDP-43 in GABAergic motor neurons, which induced age-dependent paralysis (Vaccaro et al. 2012). A set of 3,850 drugs and natural products was then screened for their ability to suppress the paralysis of their model. Hits were tested by the Drapeau group for their ability to suppress fish and mouse models of ALS, and it was found that the neuroleptic pimozide provided significant neuroprotective effects (Patten et al. 2017). Encouragingly, the compound prevented the decline in some measures of motor function in a small trial of 24 human ALS patients (Patten et al. 2017). Efforts are ongoing to optimize pimozide and its analogs (Bose et al. 2019) as a treatment for ALS (Drapeau, personal communication). Similarly, Lorraine Kalia's group found small molecule suppressors of alpha-synuclein-induced motor phenotypes in the worm that translated to a mammalian model (Chen et al. 2021). These examples typify how C. elegans can be used to screen for novel or repurposed drugs against human targets in so-called humanized strains. Other C. elegans models of neurodegeneration are highlighted in recent reviews (Giunti et al. 2021; Caldero-Escudero et al. 2024).

Mining hits for biological and structural insight

Exploring SciFinder

Small molecule vendors like ChemBridge and ChemDiv sell molecules that have physicochemical properties that make them good drug candidates [ie they largely follow Lipinski's rules (Lipinski et al. 1997)]. The majority of their stocks have not been extensively characterized in biological assays, making them ripe for discovery. Yet some fraction of the vendors' offerings are structurally similar to well-characterized molecules. This is presumably because of the limitations of synthetic chemistry combined with the physicochemical limitations imposed by Lipinski's rules. Hence, upon identifying a hit, a key question to ask is whether the hit is likely new in terms of biological activity or whether it is a close structural analog of a previously characterized molecule. If the latter, might the hit still have potential value as a tool compound or as a candidate nematicide? To answer these questions, SciFinder Scholar is an excellent tool with which to explore what is known about hits of interest and their close structural analogs (Somerville 1998) (Table 5).

Table 5.

Computational-related resources used for hit analyses.

Tool	Utility	Academic citation
CFM-ID	Predictor of fragments generated in tandem mass spectrometry (used for drug metabolite analyses in the Roy lab)	Wang et al. (2022)
ChemDraw	visualization of small molecules in specific formats	None
Cytoscape	Visualization of small molecule similarity networks	Shannon et al. (2003)
MolPort	Sourcing hits and structural analogs	None
Open Babel	Accepts small molecule names in different formats and generates extensive physicochemical properties for inputed molecules	O'Boyle et al. (2011)
SciFinder Scholar	Identifying similar molecules in the academic and patent literature, sourcing hits and structural analogs	Somerville (1998)
SwissADME	Facile SMILES visualizer provides downloadable physicochemical properties of inputed list of molecules	Daina et al. (2017)
Zinc	Sourcing hits and structural analogs	Tingle et al. (2023)

SciFinder uses the Jaccard–Tanimoto scoring system (Nikolova and Jaworska 2004) to measure how similar a query is to molecules within the academic and patent literature, as well as to any molecule available from vendors around the globe. Small changes in atom composition, side groups, or ring size can decrease the similarity score. Hence, molecules that look similar to the human eye may have low similarity scores and be far down on the similarity list [eg selectivin and levamisole (Fig. 3)]. Spending time inspecting as many analogs as possible on SciFinder can be revealing. Beyond SciFinder, we and others also take an experimental approach to determine whether a nematicidal hit may have a known mechanism of action by testing whether its associated phenotype can be suppressed by mutants that suppress the lethality induced by established nematicides and anthelmintics (Table 2).

My group has pursued hits that have different degrees of similarity to previously characterized molecules (Fig. 3). Dafadine (Luciani et al. 2011), nementin (Harrington et al. 2022), wact-86 (Burns et al. 2017), wact-190 (Kamal et al. 2019), and cyprocide (Knox et al. 2024) appear to be “first-in-class” molecules. By contrast, nemadipine is similar to the characterized L-type calcium channel inhibitors dihydropyridines but has the unique property in C. elegans of not being rapidly detoxified (Kwok et al. 2006). Nemacol is similar to the VAChT inhibitor vesamicol but has greater selectivity for the nematode ortholog (Harrington et al. 2023). Wact-11 (Burns et al. 2015) and Don Moerman's CID 2747322 (Mathew et al. 2016) are similar to fluopyram, but the 2 molecules led to a better understanding of how these molecules interact with complex II and what residues confer nematode selectivity. Selectivin is similar to levamisole but has a completely different mechanism of action (Burns et al. 2023). Migrazole induces a set of phenotypes in C. elegans that are not induced by the closely related analog fatostatin (Kwok 2011). Hence, there can be value in pursuing hits even if they share some structural similarity with characterized molecules.

Exploring structural analogs

Upon identifying a hit of interest, it can be useful to investigate what parts of the molecule are essential for its activity, what parts can vary without effect, and what changes can be made to improve its potency and, if relevant, its species selectivity. These structure–activity relationships (SARs) can be investigated in a few ways. First, pairwise Jaccard–Tanimoto similarity scores can be calculated for all molecules within a screened library. The resulting structural similarity network can be visualized using Cytoscape network software, and any analogs already screened can be easily visualized (eg Kamal et al. 2024). Second, SciFinder Scholar, Zinc (Tingle et al. 2023), and MolPort can be exploited to identify analogs that are commercially available (Table 5). Third, collaborations with academic chemists can be fruitful in generating useful and unique structural analogs of hits. The analysis of analogs can reveal the key structural features of a hit of interest, and additional rounds of structural optimization can then be carried out. In some projects, activity is only modestly improved [see selectivin (Burns et al. 2023), nementin (Harrington et al. 2022), nemacol (Harrington et al. 2023), UMW-9729 (Shanley et al. 2024b), and ABX464 (Shanley et al. 2024a) for examples]. In others, activity has been dramatically improved [see wact-11 (Burns et al. 2015), cyprocide (Knox et al. 2024), and NPD8790 (Davie et al. 2024) for examples].

Some limitations of C. elegans' utility in drug studies

No one model system is perfect, and C. elegans is no exception. One limitation is that a few commercial nematicides and anthelmintics are ineffective against C. elegans [eg albendazole, fluazaindolizine, and pyrantel (Hu et al. 2013; Lahm et al. 2017)]. Hence, using elegans to identify novel broad-spectrum nematicides will undoubtedly miss some potentially great hits. However, this point is likely true for any single screening platform. For example, the anthelmintics ivermectin and oxibendazole are ineffective in in vitro assays against the Trichuris muris whipworm parasitic model (Keiser et al. 2016; Elfawal et al. 2019). A second issue is that there has yet to be a novel or repurposed hit discovered from a C. elegans screen that has made it to the clinic or market as far as we are aware. The same can be said for academic efforts using nematode parasites in de novo drug screens (J. Keiser, personal communication). Regardless, some molecules from humanized C. elegans drug screens show promise (Patten et al. 2017; Bose et al. 2019; Chen et al. 2021), and some of our nematicidal molecules are being evaluated by agrochemical companies, which is encouraging. Given the growing intensity of drug studies in C. elegans (Fig. 1), there is hope that something will soon be born of the worm with utility beyond academia.

Conclusion

The study of C. elegans has yielded foundational insights that have rippled to most corners of the biomedical world. The earliest of investigations with the worm sought an understanding of the mechanism of action of a single small molecule (levamisole) that laid the groundwork for thousands of drug studies that followed. This review has focused on the largest of C. elegans drug screens and how the resulting hits were characterized by exploiting the worm's awesome genetic power. We hope this review will inspire others to exploit C. elegans for early-stage drug discovery and uncover new molecules with potential utility. It is hard to imagine a more genetically tractable and economically accessible model organism with which to do so.

Funding

The work from my group discussed here has been supported by Canadian Institutes of Health Research grants 68813, 133526, 153024, 173448, and 186156.