The organismal size is determined by multiple inputs from various biological pathways that affect the size, structure, and organization of the cells comprising the tissues and organs (Bar et al., 2016; Tuck, 2014; Uppaluri et al., 2016). During development, various genetic pathways that regulate cell growth, proliferation, and apoptosis need to be activated appropriately to define the body’s dimension and proportion, ultimately determining an individual’s body size (Nagashima et al., 2017). Although the regulation of body size by internal and external factors in various organisms has been extensively studied, the genetic pathways contributing to a different aspect of body growth have not been fully identified at the organismal level.

Caenorhabditis elegans, a free-living nematode, is a versatile genetic model system for elucidating genetic pathways that regulate various aspects of organismal growth. C. elegans matures to an adult after multiple molting events and having developed distinctive organs that comprise the entire body structures. A genetic screen, which isolates mutant worms with abnormal body size, contributes to identifying several biological pathways that regulate body growth (Brenner, 1974; Savage-Dunn et al., 2003). Several small (sma) mutant alleles were found to be components of the TGF-β (transforming growth factor-β) pathway, which has been established as one of the primary regulators of body size in worms (Dineen and Gaudet, 2014; Gumienny and Savage-Dunn, 2013). The loss-of-function mutations, such as lon-1 and lon-2, in the negative regulators of TGF-β signaling, results in the long (lon) phenotype. The body size of those mutants of the TGF-β pathway changes proportionally, suggesting that TGF-β signaling plays a crucial role in coordinating organisms’ growth (Harada et al., 2016).

In contrast, dumpy (dpy) mutants exhibit a disproportionate reduction in body size, often with a thicker middle body and a shorter length. Many dumpy mutant alleles are in genes encoding collagens, the cuticle’s significant components, or an exoskeleton, enclosing the entire worm (Page and Johnstone, 2007). The analysis of one long mutant cuticle collagen allele, lon-3, indicates that proper assembly of an organism’s external structure is necessary to build a standard body shape.

Analyzing the characteristics of body shape phenotypes in various mutants, such as Dpy, Sma, and Lon, will be essential for investigating the role of genetic pathways in body growth. However, due to the differential gene expression in different body parts, there is a significant challenge in analyzing body size and shape to determine the phenotype accurately. In the currently available open-source databases, some strains are often reported as both Dpy and Sma, bringing confusion to morphological concepts and phenotype designations (WormBase; https://wormbase.org/). But, at the same time, these records suggest that there may be a growth-regulating gene in complex networks of pathways differentially regulating body shape and size and contributing to organismal growth. Thus, the question of how a gene regulating body size plays a role in constructing an organismal body with optimal size and shape remains unanswered.

In this study, we determined the morphological phenotype, in terms of body size, of 46 strains of C. elegans based on their respective body length, body width, pharynx length, and head size. Among them, 41 strains have been previously reported to have the Dpy, Sma, or Lon, according to the WormBase and published literature. We analyzed the worms’ body size and evaluated each parameter’s ratios, such as the overall head shape, body length, body width, pharynx length, and head size, of each strain to detect previously unknown differences in body morphology. Utilizing this analysis, we found that the ratio of body length to body width and the ratio of head size to pharynx length can be useful parameters to identify a worm’s morphology. We validated the previously reported phenotypes of 37 mutants. We reassigned unc-32(e189) as Sma, but not Dpy that used to be confusingly annotated. We also assigned morphological phenotypes for 2 mutants, previously never characterized as abnormal body phenotype: osm-5(p813) and unc-4(gk668), small and long, respectively. The approach employed in this study provides a simple but useful method to define the morphological phenotypes associated with body size, excavating novel genes that possibly regulate organismal growth.

Strains and maintenance

Worms were grown at 20°C according to the standard protocol (Brenner, 1974). The wild-type N2 animals and the following strains used in this study were obtained from the C. elegans Genetics Center funded by National Institutes of Health (NIH) office of Research Infrastructure Programs (P40 OD010440) or National Bioresource Project Japan, and some strains are generated by the C. elegans Reverse Genetics Facility at the University of British Columbia, which is part of the international C. elegans Gene Knockout Consortium (C. elegans Deletion Mutant Consortium, 2012) : bli-3(e767), calu-1(tm1783), dpy-1(e1), dpy-10 (e128), dpy-11(e224), dpy-13(e184), dpy-17(e164), dpy-4(e1166), dpy-5(e907), dpy-7(e88), hnd-1(q740), unc-104(e1265), unc-120(st365), unc-76(e911), bli-2(e768), egl-30(n686), egl-36(n728), hsf-1(sy441), tax-6(ok2065), unc-13(e1091), unc-11(e47), unc-22(e66), unc-32(e189), unc-68(e587), lon-1(e185), lon-2(e678), lon-3(e2175), sqt-3(e24), sup-9(n180), ser-1(ok345), mod-1(ok103), egl-4(n478), dpy-14(e188), dpy-8(e130), sma-1(ru18), sma-3(e491), sma-6(e1482), unc-32(e189), che-2(e1033), osm-3(p802), osm-5(p813), osm-6(p811), unc-4(e120), unc-4(e2309), unc-4(gk668), unc-4(gk705), and unc-4(wd44) (Supplementary Table S1).

Measurements of body length, body width, pharynx length, and head size

L4 larvae were picked to fresh plates and incubated at 20°C for 24 h to measure the body sizes of adults (Fig. 1). L4 larvae worms with an apparent white crescent surrounding the visible prospective vulva were chosen. One-day adult worms were mounted on a 2% agarose pad, paralyzed with 10 mM levamisole (L9756; Sigma-Aldrich, USA), and imaged under 100× with Zeiss AxioCam using the Normarski optics and an attached AxioCam digital camera. All measurements were performed with AxioVs40 V 4.8.2.0 (Carl Zeiss, Germany).

K-means clustering

We compared the ratio of body length to body width to the ratio of head size to pharynx length to further quantify the parameters of the worm’s morphology. We applied the k-means clustering approach with the Euclidean distance (ED) to measure for similarity. Then, we analyzed 4 formed clusters of wildtype, dumpy, small, and long, to study the existing phenotypic characteristics of the worm’s morphology.

Measuring the distance between data points and cluster centroids was the fundamental first step for all clustering methods. The algorithm used in clustering was the objective function given by:

where ||x _i −v _j ||² is the ED between a data point x _i and its respective cluster center v _j .

Data points are in the space represented by the objects that are being clustered. These points represent initial group centroids in which each object is assigned to the group that has the closest centroid according to the formula given by:

where k represents the number of data points in i ^th cluster.

All objects were reassigned when the number of cluster centers was manually inputted as three. The calculation of the positions of the K centroids was repeated until the centroids became constant. Wolfram Mathematica and R were used to compute the k-means clustering algorithm and ED values.

Body size analysis of morphological mutants

Organismal morphology is determined by various genetic pathways; however, it remains unclear how these genetic pathways play a role in defining certain characteristics of body size and morphology. We investigated the body morphology of mutants with previously reported defects in body size. First, we chose a group of diverse strains, such as dpy (dumpy), sma (small), and lon (long) (Supplementary Table S1, Fig. 2A). The previously categorized dpy mutants include bli-3(e767), calu-1(tm1783), dpy-1(e1), dpy-10(e128), dpy-11(e224), dpy-13(e184), dpy-17(e164), dpy-4(e1166), dpy-5(e907), dpy-7(e88), hnd-1(q740), unc-104(e1265), unc-120(st365), and unc-76(e911). The previously categorized sma mutants include bli-2(e768), egl-30(n686), egl-36(n728), hsf-1(sy441), tax-6(ok2065), unc-13(e1091), unc-11(e47), unc-22(e66), unc-32(e189), and unc-68(e587). The previously categorized lon mutants include lon-1(e185), lon-2(e678), lon-3(e2175), sqt-3(e24), sup-9(n180), ser-1(ok345), mod-1(ok103), and egl-4(n478). Different studies have reported dpy-14(e188), dpy-8(e130), sma-1(ru18), sma-3(e491), sma-6(e1482), and unc-32(e189) to be either dpy or sma.

When the worms were measured 24 h after the L4 larval stage, the average body length of the reportedly dpy mutants was 29.27% shorter than that of the N2 wild type (Figs. 1 and 2B). Similarly, the average body length of the reportedly sma mutants was 13.87% shorter than that of the wild type. In contrast, the average body length of the reportedly lon mutants was 43.56% longer than that of the wild type. These results were consistent with the previous observation that both dpy and sma mutants were shorter, while lon mutants were longer than the wild-type animals (Brenner, 1974).

We also measured the body width of these mutants (Fig. 1). The reportedly dpy and lon mutants’ average body width was 10.78% and 8.97%, respectively, greater than that of the wild type (Fig. 2C). The average body width of the reportedly sma mutants was similar to that of the wild type. These data indicate that dpy mutants are shorter and thicker, lon mutants are thicker and longer, and sma mutants are neither thicker nor thinner but shorter than the wild-type animals. The sma mutants were previously reported to be an accurately scaled-down form of the N2 wild type animals (Brenner, 1974). However, our findings show that the average body width of the reportedly sma mutants was not significantly different from that of the wild type. The result suggests that the average morphology of the previously defined sma mutants was likely disproportionate. We also suspected that the group of sma mutants we investigated may include some mutants that are not actually small and could alter the average measurements.

Some dpy mutants are defective in cuticle development, and their bodies are often squeezed in a small, underdeveloped cuticle cavity (Page and Johnstone, 2007). The malformed cuticle sometimes results in a winding pharynx in a small head (Avery and Shtonda, 2003; Ferrier et al., 2011). Therefore, we examined the pharynx further to characterize the body morphology (Fig. 3A). At 24 h after the L4 larval stage, the pharynx length of the dpy mutants was 19.88% shorter than that of the wild type (Fig. 3B). Surprisingly, in contrary to a previous report in which the pharynx length of the sma mutants was shorter than that of the wild type (Dineen and Gaudet, 2014), the pharynx length of the sma mutants was not significantly different from that of the wild type. On the other hand, the pharynx length of the lon mutants was 25.08% longer than that of the wild type. While measuring, we found that the pharynx of the dpy mutants showed a deformation with a pharynx that was crooked from the procorpus to the terminal bulb (Fig. 3A). Based on this observation, we hypothesized that a small head could not allow the pharynx to develop in full size.

We then measured the head size and found that the average head size of the dpy mutants was 22.46% smaller than that of the wild type. However, the head size of the sma mutants showed no significant difference from that of the wild type. The average head size of lon mutants was 22.12% longer than that of the wild type (Fig. 3C). The ratio of head size to pharynx length of N2, Sma, and Lon phenotypes were all close to 1 (Fig. 3D). However, the ratio of head size to pharynx length of the Dpy phenotype was significantly smaller than that of the wild type, suggesting a disproportionate reduction in the head of the dpy mutants may be responsible for the deformation of the pharynx.

Establishment of standards to define morphological phenotypes

The analysis of body morphology suggests that previous phenotyping may not reflect the actual, i.e., proportionate morphology. We visualized individual mutants’ morphological phenotype by plotting each worm’s body size relative to the wild type by integrating four parameters: body length, body width, pharynx length, and head size (Figs. 4 and 5). Our k-means clustering centroids indicate that the average body lengths of the dpy, sma, and lon mutants are 676.49 μm, 810 μm, and 1,350 μm, respectively. The average body widths of the dpy, sma, and lon mutants are 74.52 μm, 65.42 μm, and 74.67 μm, respectively. The average head sizes of the dpy, sma, and lon mutants are 102.32 μm, 130.50 μm, and 162.06 μm, respectively. Lastly, the average pharynx length of the dpy, sma, and lon mutants are 108.73 μm, 132.55 μm, and 163.37 μm, respectively.

Mutants that belong to their respective morphological phenotype tend to cluster at a specific region of the graph (Fig. 5). The mutants with a defect in a cuticle collagen protein, dpy-1(e1), dpy-4(e1166), dpy-5(e907), dpy-8(e130), dpy-10(e128), dpy-11(e224), dpy-13(e184), dpy-14(e188), and dpy-17(e164), were clustered in a region where the ratio of body length to body width was at least 25% less than that of the wild type, and the ratio of head size to pharynx length was below 2% compared to the wild type (Figs. 4 and 5).

dpy-7(e88) was the only cuticle collagen mutant that fell out of the dpy cluster. While its ratio of body length to body width was reduced by the same degree as other Dpy phenotypes, its ratio of head size to pharynx length was reduced by 38%, indicating a proportional pharynx length relative to its head size. DPY-7 is spatially restricted to the transverse grooves or furrows of the cuticle cortex called the annulations or annular furrows (Thein et al., 2003). DPY-7 is similar in structure to other C. elegans collagen proteins but does not fit perfectly into any of the previously described classes because of its differences in the lengths of the Gly – X – Y repeats and positions of cysteine residues (Johnstone et al., 1992). The DPY-7 encoded by the dpy-7(e88) allele has a glycine-to-arginine substitution at residue 156, suggesting a disruption in the formation of a triple helix with two other collagen helices. However, this alteration in DPY-7 specific to the annular furrows may not result in cuticle shrinkage in the head region or reduced head size.

The plotting revealed that unc-32(e189) would be small rather than dumpy (Fig. 5). unc-32(e189) is previously reported as either small or dumpy (Shephard et al., 2011; WormBase). unc-32 encodes V-type ATPase V_O subunit a and its human orthologs are implicated in many diseases including cutis laxa and osteopetrosis, both of which result in stunted growth and deformity (Lee et al., 2010).

In the plotting, we also included several mutants’ data points to try to determine their morphology, which had never been reported. Among them, osm-5(p813) is clustered in the group of small, whereas unc-4(gk668) is in that of long. osm-5 is an ortholog of mammalian intraflagellar transport 88 (IFT88) that functions in non-motile cilium biogenesis and crucial in craniofacial development. osm-5(p813) exhibits impaired cilium assembly of sensillum, and the defective head structure consequently contributes to moderately retarded growth (Qin et al., 2001). unc-4 encodes a transcription factor that functions in neural wiring and its mammalian ortholog is involved in neurogenesis and somitogenesis (Sammeta et al., 2010; Sewell et al., 2009; Skuntz et al., 2009). unc-4(gk668) exhibited uncoordinated motility that is a phenotype widely reported in other mutant alleles of unc-4 tested together, but was solely clustered as long. unc-4(gk668) contains 1.6 kb deletion spanning a part of the coding region and downstream untranslated region. The deletion mutation may cause further dysregulation of other associated genes that result in the morphological difference. Both osm-5 and unc-4 are critical in nerve system development, and numerous genes functioning in neurons are implicated in body size controls. For example, tax-6 loss-of-function mutants are small, and its activity in neurons is responsible for standard body size (Kuhara et al., 2002; Li et al., 2015). Serotonin signaling in specific neurons controls body size, and serotonin receptor mutants of ser-1(ok345) and mod-1(ok103) are grouped in long cluster (Fig. 5). Therefore, the clustering analysis provides a practical primary analysis tool to assign body size phenotypes.

Morphological analysis of mutant C. elegans has been broadly utilized to investigate the function of genes. The phenotypes, such as Sma, Dpy, and Lon, are commonly used to study body size. The analysis of body size has been useful in defining distinct morphological phenotypes. However, the research is mainly dependent on empirical observation, often accompanied by a simple body length measurement.

Here, we demonstrate a novel method of categorizing morphological phenotypes of Dpy, Sma, and Lon with multiple body parameters. We analyze 46 mutant worms’ body size by measuring their body length, body width, pharynx length, and head size. We evaluated the ratios between each parameter to detect the relatively subtle differences in body morphology and head shape. We also validated some previously reported phenotypes with reassignments and provided the optimal parameters to identify a worm’s phenotype.

WormBase has served as a database for in-depth phenotypic and genetic information of C. elegans. WormBase receives phenotype information from the direct submissions by primary literature, individual researchers, and gene knockout consortiums. The vocabulary used to describe similar or identical phenotypes and the level of precision in the descriptions often vary between these sources. Consequently, different words were used to annotate the same phenotypes. Also, there was no inherent hierarchical organization in the descriptions, making the retrieval of phenotypic information via searches more cumbersome. As a result, there is significant difficulty defining a worm’s phenotype, thus interfering with the useful comparisons between strains.

Extensive phenotyping requires measuring many parameters that quantify body size and organ length and calculate the ratios between different parameters. While mutations that affect cuticle collagen are often responsible for the Dpy phenotype, several major pathways involving TGF-β, spectrin, calcineurin, and feeding efficiency determine the body size phenotypes (Mörck and Pilon, 2006). Though the smaller body size of C. elegans mutants has been recognized as an obvious developmental defective phenotype, clear and distinct standards have not been established to indicate a wild-type worm’s apparent size and shape. Our analyses suggest that the Sma phenotype involves a proportionally smaller size with mild stoutness, and the Dpy phenotype involves a significant increase in stoutness and a disproportionally small head size harboring a winding pharynx. This study utilized K-means clustering, which aims to partition data points into k number of clusters. Each data point belongs to the cluster with the nearest mean represented as a cluster centroid, which serves as a prototype of the cluster (Lloyd, 1982). Although cluster shapes and centroids can change as more data points are included, the algorithm provides practical insight grouping individual data points into classified types. K-means clustering can be further utilized in future research to better understand the specific morphological characteristics of a phenotype. Additional parameters such as egg and larvae size, volume and locomotion can also enhance the phenotypic descriptions (Chung et al., 2020; Levine and Lee, 2020; Yemini et al., 2013). More extensive analyses dealing with many strains may lead to the formation of subgroups within a particular phenotype.

Some strains are provided by the CGC, funded by NIH Office of Research Infrastructure Programs (P40 OD010440) or National Bioresource Project Japan. This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2018R1A2A3074987 to S.K.L.; 2018R1D1A1B07046893 to J.A.) and Research Grant from Hanyang University (201500000001686).

J.Y.C., T.W.C., and S.H.K. performed experiments and collected data. J.Y.C. and S.K.L. analyzed data and generated figures. J.Y.C., S.K.L., and J.A. wrote the manuscript. S.K.L. and J.A. designed the study and secured funding.

The authors have no potential conflicts of interest to disclose.