A Robust, High-Content NAM for Repeatable and Predictive Developmental and Reproductive Toxicity Assessment in C. elegans

A Robust, High-Content NAM for Repeatable and Predictive Developmental and Reproductive Toxicity Assessment in C. elegans | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article A Robust, High-Content NAM for Repeatable and Predictive Developmental and Reproductive Toxicity Assessment in C. elegans Sudip Mondal, Adam Laing, Amber Shen, Evan Hegarty, Abhishri Medewar, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9004834/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 13 You are reading this latest preprint version Abstract Developmental and Reproductive Toxicity (DART) assessment is essential for product safety evaluation and currently relies heavily on vertebrate models that are costly, time-consuming, and resource-intensive. Caenorhabditis elegans has emerged as a promising New Approach Methodology (NAM) for rapid, cost-effective, whole-organism toxicology studies. However, broader adoption has been limited by the lack of high-resolution, rapid imaging approaches, a limited range of endpoints, and insufficient evidence for assay robustness and repeatability. To overcome these limitations, we developed a multiparametric imaging-based assay for assessing DART-related endpoints in C. elegans , building on our previously published microfluidic-based developmental toxicity platform. We expanded the machine learning-based body dimension analysis to include quantification of total embryo number and in utero embryonic development by classifying embryos as early- or late-stage to assess reproductive health. The assay relies on high-resolution brightfield imaging, enabled by the vivoChip microfluidic device, and is compatible with any strain. This study demonstrates highly repeatable results with mean coefficients of variation of 1–5% for developmental endpoints and 6–17% for reproductive-related endpoints, supporting high statistical power. Validation using methylmercury and propiconazole and by phenotyping (scoring and classification) ~ 400,000 embryos across ~ 9,200 worms, demonstrated reproducible, concentration-dependent responses across all endpoints with narrow confidence intervals. Notably, late-stage embryos were the most sensitive endpoint, with effects preceding changes in total embryo count, body size, or viability. Importantly, worms remained > 97% viable and motile across all tested concentrations, indicating that the observed effects reflect DART-specific responses rather than non-specific apical endpoints such as lethality. These results demonstrate a sensitive, repeatable, and scalable DART platform capable of rapid, cost-effective chemical prioritization for safety assessment. Biological sciences/Biological techniques Biological sciences/Biotechnology Biological sciences/Developmental biology Developmental and reproductive toxicology (DART) C. elegans in utero embryo phenotype high-resolution imaging microfluidic technology standard operating procedure (SOP) repeatable assays acceptance criteria new approach methodology (NAM) Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction Developmental and reproductive toxicity (DART) assessment is a vital part of product safety testing for chemicals, consumer products, and pharmaceuticals. Currently, most DART testing is carried out in mammalian models such as rats and rabbits for human health risk assessment [ 1 , 2 ]. However, increased regulatory and societal pressure to reduce animal use has led to new legislative initiatives in the US, EU, and other regions aimed at limiting reliance on vertebrate models. In addition to ethical considerations, mammalian-based DART studies are costly and time-intensive. For example, a standard DART study with rats may require 2–3 months of exposure and monitoring due to the long life cycles, substantial housing infrastructures, and complex experimental design. High costs also restrict sample size, potentially limiting statistical power. Despite their high genetic and physiological similarities to humans, rodent data alone are only 43% predictive of human toxicity outcomes [ 3 ]. The concordance improves to 71% when combined with a non-rodent species, supporting a multi-tier approach. With over 350,000 registered chemicals requiring new or updated safety evaluation [ 4 ], there is an urgent need for New Alternative Methods (NAMs) to complement or replace traditional methods. NAM-based DART approaches include in silico prediction models [ 5 ], in vitro systems [ 6 – 8 ], ex vivo technologies [ 9 ], and alternative invertebrate model organisms not subject to welfare regulations, such as C. elegans [ 10 – 29 ], earthworm [ 30 ], brine shrimp [ 31 , 32 ], and Daphnia [ 33 , 34 ]. Although no single method can fully recapitulate human biology, a multi-tiered approach that integrates these complementary models can provide a more comprehensive and human-relevant toxicity profile for risk assessment. In silico methods remain constrained by limited, high-quality training datasets [ 35 ]. In vitro and ex vivo systems are highly variable and often capture only isolated adverse outcome pathways (AOPs) and lack the integrated tissue- and organ-level interactions required for complex DART responses. Among invertebrates, standardized DART assessment protocols exist for Daphnia and earthworms and are widely used for ecotoxicology risk assessment [ 34 , 36 , 37 ]. Although no regulatory test guidelines currently exist for C. elegans , the small model organism has been extensively used in scientific research, including numerous toxicology studies [ 11 – 29 ]. C. elegans is a powerful, high-throughput model organism that can be grown quickly and inexpensively in large numbers using well-established techniques. Its small size and simple, transparent body make it highly amenable to imaging-based methods for high-content analysis. The C. elegans genome shares extensive conservation with mammals, with up to 60–80% and ~ 83% homology to the human genome and proteome, respectively [ 38 , 39 ]. It has complete neuromuscular, reproductive, and digestive systems with conserved signaling and metabolic pathways, enabling system-level AOPs studies. The reproductive system of a C. elegans hermaphrodite consists of a germline, somatic gonad, uterus, and egg-laying circuitry [ 40 , 41 ]. Many fundamental biological processes are conserved across species and are similarly preserved in C. elegans . Notably, conserved evolutionary mechanisms governing reproduction include germ cell proliferation and differentiation, oogenesis, programmed cell death, chromosomal aneuploidy, and reproductive aging. In addition, C. elegans contains at least 76 cytochrome P450 genes [ 42 ], Phase I-III xenobiotic metabolism pathways [ 14 ], and a functional microbiome, supporting toxicological evaluation of parent compounds and metabolites with unique tissue-level bioavailability profiles [ 14 , 21 , 29 ]. Previous large-scale studies show concordance between C. elegans and mammalian toxicity outcomes. Developmental toxicity endpoints demonstrate ~ 53% balanced accuracy for predicting mammalian toxicity [ 13 ]. Germline integrity studies analyzing chromosome segregation defects in C. elegans embryos achieved 69% balanced accuracy in predicting mammalian reproductive toxicity [ 23 ]. Acute toxicity studies of 17 non-acidic chemicals reported 88.5% concordance with rat lethal dose predictions, outperforming mouse–rat concordance (87.9%) [ 43 ]. However, many of these studies rely on low-throughput, labor-intensive methodologies or low-resolution information to assess gross developmental defects and embryo fates upon chemical exposures [ 44 , 45 ], limiting scalability and reproducibility. Advances in microfluidics [ 46 – 51 ] and machine learning (ML)-assisted image analysis [ 46 ] now enable robust, high-throughput, and multi-parametric analysis of C. elegans . Here, we present a high-content imaging-based assay to quantify DART-related endpoints in C. elegans using an automated high-throughput microfluidic platform that is amenable to automation and large chemical library screening [ 46 – 48 ]. Our assay can be completed within 3 days from initial worm plating and chemical treatment to endpoint imaging. All readouts are derived from brightfield image analysis and are compatible with wild-type strain, transgenic lines, and genetically diverse wild isolates. High-resolution imaging of the entire uterus in 3D enables identification and classification of every in utero embryo, providing a more sensitive measure of toxicity than body size alone. Importantly, in utero analysis of embryo development during early adulthood is unaffected by the confounding effects from age-associated decline in the reproductive health and off-target effects such as altered egg-laying behavior. Using 30 replicates (5 biological × 6 technical replicates), we show that the DART assay is highly robust and produces repeatable results across multiple phenotypes, extended time periods, and independent of operators. As a case study, we evaluated methylmercury and propiconazole, chemicals with well-characterized DART profiles in Daphnia , rodents, amphibians, and fish, to benchmark assay sensitivity and assess translational relevance. Materials and methods Strains and maintenance We used the N2 wild type strain (The Caenorhabditis Natural Diversity Resource, CaeNDR) for all experiments. C. elegans were maintained according to standard methods on nematode growth media (NGM) agar plates seeded with HB101 Escherichia coli bacteria at 20°C [ 52 ]. Frozen stocks of strains were stored at -80°C and thawed annually to generate a new set of Dauer plates. Sealed Dauer plates were stored at 16°C and used to start fresh maintenance cultures every 3 months by chunking, minimizing genetic drift. We cultured worms for > 5 generations before using them in assays in a temperature and humidity-controlled environment. Food preparation We streaked HB101 from − 80°C glycerol stock onto Luria-Bertani (LB) agar plates with streptomycin and incubated them overnight at 37°C to generate single colonies. A single colony was picked and used to inoculate an overnight 5 mL starter culture in LB broth containing streptomycin. To prepare the HB101 food stock, 200 µl of this starter culture was added to 200 mL LB broth and streptomycin in a 1,000 mL baffled flask. The flask was incubated for 16 hours with shaking at 180 rotations per min (rpm) to reach the late log growth phase. We measured the OD 600 using a spectrophotometer (DiluPhotometer™, Implen) and stored the culture at 4°C. Streak plates and starter or food stocks were used within 2 months and 2 weeks, respectively. Immediately before worm culture, we centrifuged a sufficient volume of the bacterial culture to generate the required volume of OD 600 = 3.0 food suspension. The LB supernatant was removed, and the bacterial pellet was resuspended in S media. This density was optimal for growing 80–100 larval 1 (L1) to the day 1 (D1) adult stage in a 500 µl culture volume, with sufficient food still remaining after 72 hours to avoid starvation in the worm population prior to DART analysis. Chemicals Methylmercury (II) hydroxide (Cat# 13395, Alfa Aesar, CAS# 1184-57-2, Batch# M25H004) and propiconazole (Cat# 45642-250MG, Sigma Aldrich, CAS# 60207-90-1, Batch# BCCD0480) were dissolved in dimethyl sulfoxide (DMSO, Cat# D2650, Sigma Aldrich, CAS# 67-68-5, Batch# RNBL9635) to prepare master stocks at 100× the highest treatment concentration. The master stocks were transferred into 0.2 ml tubes as single-use aliquots and kept at -80°C until use. For the concentration-response assays, chemicals were serially diluted in DMSO immediately prior to treatment to generate 100× working dilutions, ensuing a final DMSO concentration of 1.0% (v/v) in each well. Solvent control wells received DMSO at the same final concentration (1.0% v/v). Worm synchronization and large-scale culture To generate large numbers of synchronized L1-stage larvae, two NGM plates were each seeded with four parent larval 4 (P 0 L4) worms. The worms were grown for 48 hours to allow P 0 to become adults and lay a large number of eggs. At this time, all 8 P 0 adults (2 NGM plates × 4 adults/plate) were removed. The plates were incubated for an additional 48 hours, by which point the majority of F 1 generation had reached adulthood. Adult F 1 worms were collected by washing both plates with M9 buffer and treated with an alkaline sodium hypochlorite solution with alternating low- and high-speed shaking on a speed-controllable vortexer to fragment their bodies and release embryos. The bleaching solution was neutralized with 5 washes of M9 buffer. Embryos were allowed to develop and hatch in M9 buffer in a rotating glass conical tube for 24 hours. Hatched synchronized L1s were filtered through a 20 µm cell filter to remove unhatched eggs and debris. The total number of larvae was counted, and the suspension volume was adjusted to make a final density of ~ 100 L1s/20 µl of M9 buffer. Chemical treatment in liquid culture The age-synchronized L1 suspension (~ 100 larvae in 20 µl volumes) was dispensed into individual wells of a standard 24-well plate with 475 µl HB101 suspension (OD 600 = 3.0) prepared in S media [ 53 ]. Chemicals dissolved in DMSO (5 µl) were added to the designated wells to achieve the desired treatment conditions while keeping a constant solvent concentration across all wells. Vehicle control wells received 5 µl of DMSO solvent (1% v/v). The 24-well culture plates were sealed with airtight films (Cat# 232702, Thermo Scientific) to prevent evaporation and cross-contamination and cultured at 20°C for 72 hours. The plates were shaken every 6 hours using an automated shaking platform (Teleshake 95, Inheco) to homogenize the well contents, including the HB101 distribution, shed cuticles from molting, and the intestinal discharge from the defecation cycles over 24 hours. Range-finding assay and live worm quantification For each chemical tested, we first performed a range-finding assay using at least 20 concentrations (0.009–90 µM for methylmercury and 0.01734–1730 µM for propiconazole) based on published work. Age-synchronized L1s were exposed to different chemical concentrations for 72 hours to determine the maximum soluble concentration in culture medium and assess any significant lethality or growth arrest at the L1 stage. Age-synchronized L1s were exposed to these concentrations in 24-well plates for 72 hours, after which culture plates were imaged and manually analyzed prior to chip transfer. The plate images were used to identify the percentage of live worms and to identify the highest concentration at which the worms developed beyond the L4 stage, ensuring suitability for subsequent vivoChip loading and high-resolution imaging for DART phenotyping. Specifically, prior to chip imaging, a single-frame and timelapse images (10 s at 10 fps with a 2×, 0.08 NA objective) were acquired from each 24-well culture plate using the vivoScreen system (described below), generating ~ 10.8 GB of data per plate. Single-frame images were inspected to ensure that control wells developed to D1 adult stage according to our predefined assay acceptance criteria ( Supplementary Table 1 ). Time-lapse images were inspected to count all worms within the 10,744 x 6,290 µm 2 field of view (FOV), and spontaneous movement in liquid culture was used as the criterion for viability. The percentage of moving (alive) worms per well was then calculated from 3 biological replicates on different days. vivoChip-24x design and operation Microfluidic-based C. elegans immobilization chips (vivoChip-24x, vivoVerse), fabricated from a transparent, low-chemical-absorptivity plastic with broad optical transmission spectrum [ 21 ], were used in this study. The device consists of three bonded layers: (i) a top layer containing 24 rectangular wells arranged with 9 mm spacing for compatibility with standard 24-well plates, (ii) an intermediate microfluidic layer containing trapping channels, and (iii) an ultra-thin ~ 80 µm bottom substrate layer enabling high-resolution optical access. Each well has a loading capacity of 250 µL and is fluidically connected at its bottom surface to an array of 40 parallel tapering microfluidic channels, as previously described [ 46 , 47 ]. In total, the chip contains 960 channels that orient and immobilize ~ 40 C. elegans per well, enabling rapid, high-resolution imaging of 24 distinct worm populations. The channels possess a 3D tapering geometry with an aspect ratio (width to height ratio) close to 1.0, promoting lateral worm orientation. Gradual reduction in channel dimensions allows size-dependent trapping along defined channel regions. Two vivoChip-24x designs were used: the vivoChip-24x-D1 design with larger channel geometry (tapering from 98 µm × 102 µm at the channel entrance to 24 µm × 40 µm at the exit) and vivoChip-24x-L4 design based on narrower geometry (tapering further to 10 µm × 17 µm at the narrowest region). The D1 design is optimized to trap day 1 (D1) adult worms containing 2–6 in utero embryos at the narrowest channel region while allowing laid eggs to pass through. The L4 design enables trapping smaller worms, including early L4 stage worms, and is better suited for populations with fewer or no embryos, although it carries a higher risk of blockage from laid eggs and clogging the microfluidic channels. Accordingly, two functional designs were used: the vivoChip-24x-D1 (based on the larger channel geometry) and the vivoChip-24x-L4 (based on the narrower geometry). The D1 chip was preferentially used because its wider channels prevented blockage by laid eggs and therefore had a greater number of channels containing trapped worms. The L4 chip was mainly used for worms treated with high concentrations of chemicals that slowed development, resulting in reduced body size and fewer or no in utero embryos. Following removal of the culture plate seal, worms were transferred from each culture plate to the corresponding vivoChip well using a multichannel pipette. Since vivoChip-24x-L4 can trap L4-stage or larger worms, wells containing populations younger than L4 were filled with molten 2% agarose gel to block the channels in those specific wells. The agar block prevented continuous media flow through the open channels and stabilized pressure across all wells during vivoChip imaging. The chip was placed in a custom vivoChip holder with a sealing gasket to provide controlled fluidic pressurization, as previously described [ 22 ]. The vivoChip holder carrying the vivoChip was placed on the microscope stage for imaging and flow was initiated using our vivoCube+ microfluidic control system. Ten on/off pressure cycles were applied, resulting in complete immobilization of C. elegans across all 960 channels within ~ 3 minutes. Automated image acquisition hardware The vivoScreen system comprises integrated image acquisition hardware, custom software modules (vivoImager), a centralized searchable Postgres SQL database, and hybrid in-house and cloud-based data storage ( Supplementary Fig. 1 ). The image acquisition hardware consists of a customized large FOV inverted microscope (IX73, Evident), a precision xyz motorized stage (MS2000, Applied Scientific Instrumentation), a scientific CMOS camera (IRIS-15, Teledyne), and bright light sources (TLED, Sutter Instrument). The large-FOV setup enables imaging of the entire chip area with a 2×, 0.08 NA objective and 8 microfluidic channels simultaneously using a 10×, 0.4 NA objective, capturing all 40 channels with only 5 FOVs. The 10× objective produces a 2.15 × 1.26 mm 2 FOV, sufficient to capture the full lengths of adult worms across 8 channels with a lateral pixel resolution of 0.425 µm. For DART analysis, volumetric brightfield imaging was performed using 18 z-slices at 6 µm step size to capture the entire worm volume. According to the Nyquist sampling criterion, the theoretical axial resolution for a 0.4 NA objective is 6.5 µm, corresponding to a recommended z-step of 3.25 µm. However, to maintain volumetric imaging speed ~ 1 volume per second with our data readout speed, we used a 6 µm step size. We captured five timelapse 3D hyperstack images at 1-s intervals to analyze twitching in the late-stage embryos if needed. Automated image acquisition software and process To acquire the imaging data, we used custom-developed image acquisition software (vivoImager, Supplementary Fig. 2 ). We captured 10×, 0.4 NA 3D brightfield stacks (18 z-slices) of all worms, centered around the best focal plane of a fiduciary marker ( Supplementary Fig. 3A ). Each vivoChip-24x well contains a cross-shaped fiduciary marker, integrated into the microfluidic channel layer. This feature enables automatic focal plane identification using an edge-detection algorithm ( Supplementary Fig. 3B ). Initially, the software loads a configuration file containing experimental metadata (experiment name, strain name, developmental stage, chemical name, concentration, solvent, etc.) and imaging parameters (illumination intensity, filter sets, camera settings, stage locations, etc.). Then the software initiates all hardware components, calibrates the stage for translation range, and positions the stage at the cross-shaped fiduciary marker of the reference well (B02). Using a pre-defined chip map, the system sequentially navigates to the fiduciary marker locations in each of the 24 wells to determine the best focal plane. Variations in focal position primarily arise from substrate curvature during chip pressurization. While the relative channel locations are known from the chip map, the positions of the trapped worms within the 40 parallel, 3 mm-long channels may vary depending on their size. To optimally center the 1.3 mm × 2.2 mm FOV to capture the entire body of as many worms as possible, a low-magnification overview image of the entire chip is first captured using a 2×, 0.08 NA objective. Then, all immobilized worms are detected using a trained YOLO object detection model [ 23 ], and the relative stage positions are calculated for the 5 FOVs needed to image the worms at 10×, 0.4 NA (8 × 150 µm spaced channels can fit in each 1.3 mm wide FOV) ( Supplementary Fig. 3C ). The system then automatically switches to the 10× objective and captures brightfield z-stack images using high-precision 3-axis translational stages. For each well, 5 FOVs are imaged, each comprising 18 z-planes (6 µm step size) and five timepoints in brightfield for volumetric timelapse acquisition ( Supplementary Fig. 3D ). A single vivoChip experiment generates 12,984 images (~ 390 GB). Raw image data and associated metadata (timestamps, xyz coordinates, image capture and illumination parameters, treatment conditions etc.) are automatically transferred to a high-capacity network attached storage (NAS) server for preprocessing and downstream analysis. Image processing and phenotypic scoring For each FOV with 8 channels, channel boundaries were identified by detecting fiduciary markers within each well and applying pre-defined internal channel spacing from the chip map. Channel coordinates were stored in the centralized database for subsequent multiparametric analysis. Automated worm detection was performed using a deep learning–based method [ 46 ]. In brief, an image slice with optimal focus was identified from each z-stack by maximizing the variance of the Laplace-transformed image [ 54 , 55 ]. An improved 2.5D U-Net architecture, including a classification head at the bottleneck, was trained, tested, and used to first classify each channel as full (entire worm within the FOV), partial (partial worm within the FOV), or none (no worm present) and then predict a segmentation mask for channels classified as full. For the new model, we expanded the training dataset to include small-sized worms close to the channel exit, decoupled the image-level classification and the pixel-wise segmentation models to facilitate independent optimization and model learning with 1 and 5 z-slice images around the best focus image, respectively. The segmentation mask, predicted by the new model, was used to calculate body length (via skeletonization of the mask), area (total segmented pixels), and volume (mask area integrated with channel heights) for each full worm. Embryonic phenotyping was performed manually on all channels predicted as “full” by the ML model. All in utero embryos in these channels were scored and classified according to their developmental stage. Ten trained scorers independently scored and classified embryos using a custom graphical user interface (vivoAnalyzer, Supplementary Fig. 4 ). The software displays individual channels with zoom and z-plane navigation capabilities, allowing users to place a marker on each embryo and classify it as early- or late-stage embryo using the two-fold stage as the classification threshold. All automated image processing outputs (worm classification labels and segmentation masks) and manual annotations (marker positions and classifications of each embryo) were stored in the centralized database and exported for statistical analysis. The majority of embryo scoring was conducted in a blinded manner, with treatment conditions concealed from scorers. Statistical analysis For DART data analysis, datasets were first cleaned to remove any worms for which the body segmentation mask failed, preventing dimensional calculations, or channels marked as unscorable by a user due to intestinal obstruction of the uterus ( Supplementary Fig. 5 ). The remaining worms were then filtered to remove outliers in body volume or total embryo count (sum of early- and late-stage embryos). Outliers were identified using Tukey fences (1.5 × the interquartile range (IQR)) applied to each endpoint, and worms with measurements outside these fences were excluded. Inspection of excluded samples indicated that most were debris composed of shed worm cuticle and laid embryos clogging the channel. Mean phenotype scores were calculated per well and used to fit concentration-response curves from 3 biological replicates. One-way ANOVA was used to identify statistical differences between multiple well average values for different DMSO conditions. To estimate repeatability, the coefficient of variation (CV% = ratio of standard deviation to mean) was calculated for every possible combination of 3 wells drawn from every corresponding well position across 5 independent biological replicates (10 combinations per well position, 24 well positions). The mean CV% and standard error of mean (SEM) across all 10 combinations were reported for each endpoint. To measure the strengths of the difference between 1% DMSO and methylmercury-treated worms, we calculated the Strictly Standardized Mean Difference (SSMD) for all 6 endpoints using 3 biological replicates. From all methylmercury conditions, we selected the lowest concentration to serve as a positive control for our case study that produced the SSMD ≥ 2 for all the DART parameters, indicating strong effects [ 56 – 59 ]. Assay power was estimated using the mean and standard deviations obtained from all 60 possible combinations (10 combinations per well position × 6 well positions with 1% DMSO) for all endpoints from repeatability experiments. Power calculations were performed for a fixed value of n = 3 technical replicates and a series of hypothetical treatment group means representing 2.5% − 50% effect sizes below the control means (one-sample, one-tailed t test, α = 0.05) using GraphPad Prism (version 10.6.1). Concentration-response curves and effective concentration (EC 10 , EC 25 , and EC 50 ) values were fitted using a 4-parameter logistic curve function using the “Find ECanything” nonlinear fit function of GraphPad Prism. The lower asymptote for each endpoint was constrained to the minimum biologically possible values of 0 embryos for Reproductive endpoints and estimated mean L1 body dimensions (length = 150 µm, area = 1,500 µm 2 , volume = 11,775 µm 3 ) for developmental endpoints. For each EC x estimate, the 95% confidence interval (CI) bands were calculated and plotted. Two concentration-response curves were compared using extra sum-of-square model to identify if the best fit values of EC x values and Hillslope differ between two data sets. The p-value was denoted by ns ( p -value ≥ 0.5), * ( p -value ≤ 0.01), ** ( p -value ≤ 0.001), *** ( p -value ≤ 0.0001), and **** ( p -value ≤ 0.0001). Results A robust C. elegans assay to study developmental and reproductive toxicity (DART)-related endpoints We developed a DART assay that quantifies in utero embryos within adult C. elegans at distinct developmental stages (early and late stages) following chemical exposure. C. elegans populations were exposed from the L1 stage for 72 hours and subsequently imaged using vivoChip devices to quantify embryo development relative to the 2-fold stage (Fig. 1 A). Building on our previously established developmental toxicity (DevTox) platform, which uses ML-assisted image analysis [ 46 ] to rapidly quantify developmental endpoints in C. elegans , we extended this high-content imaging and analysis approach to capture reproductive phenotypes with high sensitivity. In addition to measuring worm body dimensions (length, area, volume) with high accuracy, we quantified the total number of embryos in each worm and classified each embryo by developmental stage (early- or late-stage embryo development). This approach of quantifying sub-lethal reproductive phenotypes enabled the determination of effective concentration (EC 50 ) values with high confidence and low coefficient of variance (CV%). The whole-organism 3D imaging enabled by vivoChip microfluidic platform provides sufficient resolution to determine the developmental stage of in utero embryos inside the intact C. elegans uterus [ 60 ]. The vivoChip image-based assay yields 6 DART-related endpoints: body length, body area, body volume, total embryos, early-stage embryos, and late-stage embryos (Fig. 1 B). Additionally, brightfield imaging of culture plates prior to chip loading was used to confirm worm viability and motility. By directly quantifying in utero embryo number and classifying developmental progression during early adulthood, this assay captures sub-lethal reproductive toxicity while minimizing confounding effects from age-associated decline in the reproductive health and off-target effects such as altered egg-laying circuits. This DART platform facilitates rapid, high-content screening of chemicals and sensitive detection of sub-lethal developmental and reproductive perturbations. Automated high-resolution, whole-body imaging to extract DART endpoints The DART assay leverages the unique ability of vivoChip microfluidics platform to rapidly immobilize large numbers of C. elegans for high-resolution imaging of their entire body. Each vivoChip-24x device enables parallel immobilization up to 960 C. elegans (Fig. 2 A), and the closely spaced microfluidic channels maximize information density in each FOV (Fig. 2 B). A pressurized gasket system seals the vivoChip-24x device and applies a pre-defined pressure sequence for loading and immobilization of worms efficiently (Fig. 2 C). Using automated stage navigation and objective switching, all wells are imaged sequentially, and complete volumetric datasets are acquired in ~ 30 minutes per chip. High resolution 10× 0.4 NA brightfield z-stacks (18 planes at 6 µm spacing) captured the full uterus of each immobilized worm, enabling visualization of in utero embryos across developmental stages. This axial sampling accommodates minor variations in worm posture within the channels while preserving imaging speed. Prior to chip imaging, videos of the culture plates were collected to analyze worm motility (Fig. 2 D). Embryos were classified into two developmentally distinct categories: early-stage embryos (≤ 2-fold) and late-stage embryos (> 2-fold). The 2-fold stage provides a clear morphological boundary due to the rapid increase in structural complexity within the eggshell. Late-stage embryos often display fully formed larval structures, particularly in the central region of the uterus near the vulva, where embryos are frequently developed into young larvae. Visualization of individual larvae within the eggshell boundaries facilitates accurate classification. In some cases, movement within the eggshell was observable in time-lapse data (Figs. 2 E–F), further supporting developmental staging. In future implementations, embryo motility could be incorporated as an additional quantitative endpoint. Image analysis pipeline for robust phenotyping Imaging data were first pre-processed by identifying bounding boxes around individual channels, enabling analysis of each worm ( Supplementary Fig. 6A ). Downstream image analysis consisted of two separate processes; (1) automated quantification of worm body dimensions using a deep-learning-based segmentation ( Supplementary Figs. 6B-C ) and (2) semi-automated embryonic phenotyping using a GUI-based analysis software to streamline manual scoring ( Supplementary Figs. 6D-F ). Previously, we developed a 2.5D UNET architecture for image-level classification and pixel-wise segmentation tasks [ 46 ]. Under this architecture, both tasks were jointly learned within a single network sharing a common encoder and bottleneck, trained using three z-slices around the best-focus image. While effective, this design introduced task coupling, limiting independent modification and optimization of the two objectives. To address this limitation, we decoupled classification and segmentation into separate models, each tailored to its respective task. The classification model comprises a convolutional encoder followed by a Vision Transformer (ViT) bottleneck and a classification head. The segmentation model consists of a convolutional encoder, a ViT-based bottleneck, and a convolutional decoder. Since image-level classification relies primarily on global morphological features, present in the best-focus image, we trained the classification model using a single best-focus image. In contrast, pixel-wise segmentation tasks rely on subtle changes in the image contrast to identify the object boundaries. To improve head and tail boundary detection, we fed a total of 5 z-stack images around the best-focus plane into the model. This approach captures features spanning multiple z-slices, improving boundary identification. To further improve robustness under challenging imaging conditions, the training datasets for both models were expanded to include additional channels with smaller worms, located near the exit portion of the immobilization channel from both vivoChip-24x-L4 and D1 devices ( Supplementary Figs. 7 ). For the classification model training, we generated ground truth data by classifying 5,305 channels as full, partial, or no worms class. For the segmentation model training, we used manually segmented masks from 4,360 channels with full worms. The classification model was evaluated on a test set of 538 images (441 full, 50 partial, and 47 channels with no worms) and achieved a weighted F1-score of 98.3%. The segmentation model was evaluated on a separate test set of 439 full worm images, achieving a mean Dice score of 98.5% ( Supplementary Fig. 8 ). High prediction accuracy for both channel classification and worm segmentation was observed for worms immobilized in D1 ( Supplementary Figs. 9A-D ) and 4L ( Supplementary Figs. 9E-F ) devices. From segmented full-worm channels, we calculated body length (defined as the longest skeleton length) and body area. Body volume was then estimated by integrating the segmented area by the known channel height at each pixel. For each worm, these three body parameters were calculated, and population-level metrics were obtained by averaging values across worms within each well. Embryo phenotyping is a challenging task, as each vivoChip-24x can immobilize up to 960 adult C. elegans , corresponding to approximately 40,000 in utero embryos requiring classification. Furthermore, the intertwined morphology of the uterus with the intestinal structures and gut granules requires 3D visualization of the worm uteri for accurate phenotyping. To facilitate robust embryo phenotyping, we developed a custom GUI-based software tool (vivoAnalyzer) that automatically displays each worm-containing channel in sequence, enables interactive zooming and z-stack navigation, and allows users to annotate the xyz centroid of each embryo while assigning early- or late-stage classification ( Supplementary Fig. 4 ). When necessary, the user can use time-lapse frames to confirm late-stage embryos, exhibiting larval movement. Scoring was performed by 10 trained users after following validation of their scoring accuracy on a reference dataset. Worms were randomly assigned to users, and treatment conditions were blinded to minimize scoring bias. Repeatability of the DART assay To evaluate the robustness and statistical reliability of the DART assay in predicting toxicity levels of substances, we studied repeatability across multiple independent experiments using the same test substances. For an assay to replace or augment existing DART methods, it needs to be scalable, reproduceable by different scientists using different batches of worms and capable of providing repeatable results over time. We tested the repeatability of our DART assay by performing a series of 5 independent experiments (biological replicates) over two-month period using DMSO, a commonly used solvent in toxicity assays because of its wide applicability to a range of chemicals. Four DMSO concentrations (0.0, 0.2, 0.5, and 1.0%) were tested to characterize baseline phenotypes under solvent only control conditions. Each concentration was replicated across 6 wells (technical replicates) to assess well-to-well technical variability in addition to inter-experimental variability (Fig. 3 A). We selected a 1% DMSO concentration as the solvent concentration because it allows us to dissolve the highest possible chemical concentrations near solubility limits while modestly increasing membrane permeability, which aids chemical uptake [ 12 , 61 , 62 ]. The worms developed into the D1 adult stage in all wells across all tested DMSO concentrations, including 1% DMSO (Fig. 3 B). To examine variability across both biological and technical replicates, we compared all 30 well averages (5 biological replicates × 6 technical replicates) for all endpoints. For one representative biological replicate, the mean body parameters for 0% DMSO were 1,522 ± 8 µm (length, Fig. 3 C and Supplementary Fig. 10A ), 75.9 ± 0.6 × 10 3 µm 2 (area, Fig. 3 D and Supplementary Fig. 10B ), and 2.96 ± 0.06 × 10 6 µm 3 (volume, Fig. 3 E and Supplementary Fig. 10C ). For 1% DMSO, the corresponding values were 1,596 ± 3 µm (length, p -value < 0.0001), 74.2 ± 0.4 × 10 3 µm 2 (area, p -value = 0.19), and 2.76 ± 0.02 × 10 6 µm 3 (volume, p -value = 0.045). Embryo-related endpoints were similarly consistent. For 0% DMSO, worms exhibited 30.7 ± 0.6 early-stage embryos (Fig. 3 F and Supplementary Fig. 10D ), 14.0 ± 0.5 late-stage embryos (Fig. 3 G and Supplementary Fig. 10E ), and 44.7 ± 0.2 total embryos (Fig. 3 H and Supplementary Fig. 10F ), whereas for 1% DMSO, averages were 29.2 ± 0.5 ( p -value = 0.23) early-stage embryos, 14.5 ± 0.6 ( p -value = 0.86) late-stage embryos, and 43.7 ± 0.4 ( p -value = 0.24) total embryos. Although, there were significant differences between 0% and 1% DMSO conditions for body length and volume, there were no significant differences between these two populations for body area, total embryos, early-stage embryos, or late-stage embryos ( Supplementary Fig. 10 ). Although body length and volume showed statistically significant changes (+ 4.9% and − 7.3%, respectively), the magnitudes of these differences were small. The well mean values obtained from all 30 wells from 5 biological replicates showed a similar trend among all 6 DART parameters ( Supplementary Fig. 11 ). DMSO concentrations ≥ 1% has previously been reported to affect worm internal structure and posture, potentially causing mild changes in the body dimensions [ 33 ]. However, when comparing all 30 wells from the 5 biological replicates, we found no significant differences between 0% and 1% DMSO for early-stage, late-stage, and total embryo counts ( Supplementary Figs. 10D-F ). To test variability in the DART endpoints among technical replicates, we calculated coefficient of variation (CV%) between 6 technical replicates (Columns 1 to 6 in vivoChip-24x) for each condition. For 1% DMSO condition, the mean CV% values between the technical replicates were < 3.8% for developmental endpoints and < 12.5% for embryonic endpoints (Fig. 4 ). This data suggests high reproducibility among worm populations loaded and imaged in the vivoChip, confirming the absence of edge effects or location-based bias. To assess inter-experimental repeatability, we calculated the coefficient of variation (CV%) between 3 corresponding wells from 3 separate experiments (e.g. exp 1 A01, exp 3 A01, exp 5 A01) and repeated this calculation for all 10 possible combinations of the 5 experiments (Figs. 5 A-B). For the 0% DMSO populations, the mean CV% values across different experiments and 60 total combinations were 1.5 ± 0.1% for body length (Fig. 5 C), 3.9 ± 0.2% for body area (Fig. 5 D), 7.8 ± 0.4% for body volume (Fig. 5 E), 11.5 ± 0.8% for early-stage embryos (Fig. 5 F), 17.1 ± 0.9% for late-stage embryos (Fig. 5 G), and 7.4 ± 0.5% for total embryos (Fig. 5 H). For the 1% DMSO populations, the mean CV% were 1.2 ± 0.1% (length), 2.5 ± 0.2% (area), 5.0 ± 0.3% (volume), 5.8 ± 0.3% (early-stage embryos), 16.7 ± 1.1% (late-stage embryos), and 5.24 ± 0.3% (total embryos). All values were substantially below the 30% threshold considered acceptable under OECD test guideline #222 for earthworms [ 37 , 63 ], indicating high inter-experimental repeatability of the assay. Statistical power of DART assay parameters To validate our DART assay design has sufficient statistical power of detecting small proportional changes relative to the control population using 3 biological replicates, we calculated the minimum detectable size effects with ≥ 80% power values. To estimate the power, we used the aggregated means and standard deviations from 60 possible combinations (10 combinations per well locations and 6 wells) and for all parameters measured in the 1% DMSO wells, that will serve as assay controls. With 3 experimental replicates, our assay achieved > 80% power to detect changes of ≥ 5.5% in developmental endpoints (body length, area, and volume) and ≥ 7% in total and early-stage embryos. For late-stage embryos, the minimum detectable effect size at the 80% power threshold was 23% ( Fig. 6 A ) , though increasing biological replicates would enable detection of lower reductions. DART assay with methylmercury We next evaluated the assay using a well-studied toxicant, methylmercury, which has documented toxicity in C. elegans [ 20 , 25 , 64 ] and mammalian models [ 65 ]. The purpose of testing this chemical was to assess repeatability and precision in concentration-response studies, and to find the sensitivity of each endpoint. We ran 3 biological replicates spanning 20 concentrations of methylmercury between 0.0007 and 10 µM using the vivoChip-24x-D1 device. Range-finding assays based on plate imaging showed that concentrations > 10 µM caused developmental arrest between the L1 and L3 stages, below the minimum size trappable in the D1 or L4 devices. Consistently, worms exposed to ≥ 8 µM developed slowly, remained approximately at the L4-stage, and slipped through the D1 chip channels due to their small size ( Supplementary Fig. 12 ). To enable imaging at higher concentrations, we ran a second set of experiments using the L4 devices, which can trap smaller worms down to early L4 stage. For the full concentration-dependent DART study, we used data from the D1 chip for lower concentrations and only used the data from L4 chips for the highest concentrations to complete concentration-response curves. Inclusion of the L4 chip data captured the 50% effect region for developmental endpoints and up to 100% effect region for embryo endpoints. The combined results obtained from the D1 and L4 devices were fit with a 4-parameter logistic curve using the well averages after outlier filtering, and EC 10 , EC 25 , and EC 50 values were estimated for each endpoint ( Fig. 7 and Supplementary Table 2) . For developmental endpoints, EC 50 values (95% CI) were 8.22 µM (7.83–8.69) for body length (Fig. 7 A), 5.33 µM (4.98–5.72) for body area (Fig. 7 B), and 4.61 µM (4.28–4.99) for body volume (Fig. 7 C), indicating that body volume is the most sensitive developmental metric. For the reproductive parameters, EC 50 values (95% CI) were 3.90 µM (3.66–4.16) for early-stage embryos (Fig. 7 D), 1.49 µM (1.33–1.67) for late-stage embryos (Fig. 7 E), and 3.14 µM (2.90–3.38) for total embryos (Fig. 7 F), indicating that embryo classification associated with different stages of development provides additional sensitivity and that late-stage embryos represent the most sensitive reproductive phenotype. Representative worm images are shown in Supplementary Fig. 13 . Inter-experimental repeatability determined from the 1% DMSO control wells was high, with CV% values of 1.2% (length), 8.3% (area), 12.9% (volume), 8.8% (early-stage embryos), 17.1% (late-stage embryos), and 10.2% (total embryos), all well below the 30% threshold. In addition, the EC 50 confidence intervals were narrow ( < ± 0.5 µM), further supporting good assay precision. To study the effect of reducing the number of treatment conditions on EC 50 values, we omitted every other methylmercury condition and reanalyzed the DART data from 10 out of the 20 conditions. We fitted the reduced 10-point dataset with a 4-parameter logistic curve and re-estimated EC 10 , EC 25 , and EC 50 values ( Supplementary Table 3 ). The resulting EC 50 values were not significantly different from those obtained using the full dataset for the most sensitive developmental endpoint (body volume; EC 50 = 4.51 µM; 95% CI: 4.03–5.08, p -value = 0.98, Supplementary Fig. 14A ) and the most sensitive reproductive endpoint (late-stage embryos; EC 50 = 1.51 µM; 95% CI: 1.26–1.80, p -value = 1.00, Supplementary Fig. 14B ). To identify a suitable concentration of methylmercury as a possible positive control for our DART assay, we calculated the SSMD for all 20 concentrations tested and all 6 DART parameters analyzed (Supplementary Table 4) . SSMD provides a quantitative measure of assay quality by evaluating the separation between control and treated populations relative to variability. For 4 µM methylmercury, which produced a clear reduction in both developmental and reproductive parameters ( Supplementary Fig. 13 ), SSMD values were: 5.18 (length), 3.24 (area), 2.77 (volume), 2.96 (late-stage embryos), 5.84 (early-stage embryos), and (total embryos). According to established criteria, SSMD values ≥ 3 indicate excellent assay quality, and values between 2 and 3 indicate good assay quality. Therefore, 4 µM methylmercury demonstrated good-to-excellent performance across all endpoints and was selected as a positive assay control concentration for the subsequent case study. Based on the methylmercury concentration-response data, we further refined our DART assay design to test new chemicals at 10 concentrations per curve. This approach captures concentration-dependent effects across endpoints whose EC 50 values may differ by up to 10 folds. Inclusion of sufficient concentration points is particularly important for endpoints with steep slopes, such as for late-stage embryos, the most sensitive parameter of our assay, to estimate precise EC 50 values with narrow 95% CIs. Case study: DART assessment of an agrichemical Following demonstration of assay precision and repeatability, we performed a concentration-response DART study using the triazole fungicide propiconazole [ 66 – 68 ], a chemical of interest to the agrichemical industry and ecotoxicology. First, we performed a preliminary range-finding assay by treating synchronized L1-stage N2 worms with 22 different concentrations of propiconazole ranging between 0.02-1,730 µM following our standard protocols. Concentrations > 334 µM were lethal ( Supplementary Fig. 15 ). Based on these results, we conducted a full concentration–response study with 10 propiconazole concentrations (0.08–334 µM), along with the solvent control (1% DMSO), and the positive control (4 µM methylmercury). Using 10 test concentrations and 2 controls (12 wells in total), concentration-response curves from 2 different chemicals can be tested on a single vivoChip-24x. At the highest propiconazole concentration (334 µM), worms exhibited delayed growth and were immobilized using the L4 chip. Similar to methylmercury, propiconazole produced distinct concentration-dependent effects (EC 50 responses) across developmental (Figs. 8 A-C) and reproductive (Figs. 8 D-F) endpoints. Among developmental parameters, body volume exhibited a greater sensitivity (EC 50 = 336 µM; 95% CI: 311–369; Fig. 8 C) compared to body length (EC 50 = 504 µM; 95% CI: 392–685; Fig. 8 A). For reproductive endpoints, late-stage embryos were the most sensitive metric (EC 50 = 78 µM; 95% CI: 35–154; Fig. 8 E) with EC 50 less than half that of total embryos (EC 50 = 204 µM; 95% CI: 177–237; Fig. 8 D). These results further support late-stage embryo phenotype as the most sensitive DART endpoint in this assay. The EC 10 , EC 25 , and EC 50 values for 6 DART parameters are listed in the Supplementary Table 4 , and representative worm images are shown in Supplementary Fig. 16 . Comparative endpoint sensitivity analysis for DART assay To further investigate the relative sensitivity and utility of different DART endpoints, we compared concentration-response curves normalized to baseline (no-effect) values for a widely used developmental parameter (worm body length), and total and late-stage embryo counts, for both methylmercury ( Fig. 9 A ) and propiconazole ( Fig. 9 B ) . In addition, we quantified the percentage of worms that were alive, as identified to be moving in culture well plate images captured using 2× objective. All worms remained alive at concentrations up to at least the highest concentration tested in the vivoChip concentration-response experiments. For example, 100% of worms in the plates were alive at 334 µM propiconazole and > 97% alive at 10 µM methylmercury, despite clear developmental impairment. These findings suggest that the observed responses were specifically attributable to DART-specific effects rather than lethality, an apical endpoint studied in C. elegans . We calculated EC 10 values for body length, total embryo, and late-stage embryo parameters for both methylmercury and propiconazole using the normalized concentration-response data (Fig. 9 ). The EC 10 estimates, representing the onset of toxicity for each endpoint, were widely separated and followed the order: late-stage embryos EC 10 < total embryos EC 10 < body length EC 10 < lethality onset. For methylmercury, the EC 10 value for body length was 3.62 µM (95% CI: 3.33–0.92), total embryo was 1.94 µM (95% CI: 1.71–2.19), and late-stage embryo was 1.08 µM (95% CI: 0.92–1.24). For propiconazole, the EC 10 value for body length was 256 µM (95% CI: 230–285), total embryo was 57 µM (95% CI: 43–73), and late-stage embryo was 14 µM (95% CI: 7–24). Notably, the EC 10 for late-stage embryos preceded any measurable response in other endpoints, identifying late-stage embryo count as the most sensitive parameter. For methylmercury and propiconazole, body length EC 10 (3.62 µM and 256 µM), a commonly used developmental endpoint, was significantly different than the most sensitive late-stage embryo EC 10 values (1.08 µM; p -value < 0.001 and 14 µM; p -value < 0.001) and were 3.4× and 18.3× higher, respectively. Viability remained high across the tested concentration ranges, confirming that the observed changes in each DART parameter represent bona fide developmental and reproductive effects rather than secondary consequences of lethality. Together, these findings demonstrate that the multiparametric DART assay provides independent and biologically meaningful assessments of developmental and reproductive endpoints without confounding effects from reduced viability. This significantly lower toxicity threshold and large ratio of developmental and embryo parameters, especially for propiconazole, demonstrates that our multiparametric DART approach can detect reproductive toxicity independent of DevTox and viability. Also, the DevTox parameters are detected independent of viability. Such multiparametric analysis helps us to identify the adverse effects onset with potential benefits for prioritizing safety assessments of substances. Discussion DART toxicity studies are required for hazard assessment of thousands of new and existing substances. However, high costs, long study durations, and a global push to reduce or eliminate mammalian models in routine testing present major bottlenecks. Here, we introduce a novel imaging-based DART assay using the model organism, C. elegans , which is faster, more cost-effective, and eliminates the need for vertebrate animals subject to welfare regulation. Like human females, C. elegans has a complete reproductive system with conserved molecular pathways and several hallmarks of reproductive aging [ 40 ]. In certain toxicological domains, C. elegans have been shown to be as predictive as rodent models [ 13 , 28 , 43 ], supporting the translational relevance of the data obtained with this system for human reproductive toxicity assessment. While C. elegans has been used as a model organism in DART studies, previous approaches were limited by low throughput, labor-intensive methodologies, or reliance on relatively gross phenotypes such as brood size or body length [ 44 , 45 ]. In this work, we developed a high-content, multiparametric imaging platform capable of detecting subtle multi-parametric DART phenotypes. This approach increases assay sensitivity to low-level exposures while improving throughputs, thereby addressing key limitations of existing DART testing paradigms. We previously developed an ML-based imaging platform to assess DevTox using C. elegans as an alternative model organism [ 46 ]. This study expands upon that work by incorporating reproductive toxicity-related endpoints to assess multiple modes of action within a unified DART framework. We leveraged the vivoChip platform, which enables high-resolution, high-throughput imaging of 24 independent worm populations, with 40 worms per population (totaling 960 worms per vivoChip), while generating high-information-density datasets suitable for quantitative assessment of in utero embryo production and development progression. Two different vivoChip-24x designs, each optimized for trapping worms of different sizes, allowed us to analyze DART endpoint changes ranging from delayed L4s to normally developed adults. Integrating data obtained from both device configurations enabled comprehensive coverage across different sections of the concentration-response curve. This capability is particularly essential for multi-parametric assays, in which EC 50 values for distinct endpoints may be widely separated, requiring adequate sampling across multiple exposure ranges to accurately resolve differential sensitivities. For our DART assay, we exposed age-synchronized L1s for 72 hours and performed terminal imaging of C. elegans populations using the vivoChip platform. Worms were treated with multiple conditions, including solvent controls, in which the worms developed to the D1 adult stage during the exposure period. Because toxic chemical exposures can cause slow and asynchronous development, especially at high concentrations, it was necessary to start with precisely age-synchronized worms. Reproductive organ development in C. elegans spans from L1 to early adulthood. This process includes cell division of the two somatic gonad precursors and two primordial germ cells at the L1 stage, reorganization of the somatic gonad primordium during late L2, and formation of the sheath, spermatheca, and uterus during the L3 and L4 stages [ 41 ]. A 72-hour chronic exposure starting at the L1 stage maximizes the likelihood that chemicals reach the germline during critical windows of development. Such exposure may interfere with key biological events, such as mitotic proliferation, meiotic cell cycles, spermatogenesis, oogenesis, oocyte maturation, ovulation, and the entire course of in utero embryogenesis up to the start of neuromuscular movement within the egg [ 69 , 70 ]. Imaging all treatment groups in an experiment within a narrow time interval further ensured accurate comparisons across concentrations and minimized variability introduced by developmental timing differences. To capture adverse effects on reproduction, we identified and quantified all embryos within the uterus of individual C. elegans hermaphrodites and classified them as early- or late-stage embryos based on distinct morphological features, using the 2-fold stage as the classification threshold. We also calculated total embryo counts by adding early- and late-stage embryos for each worm. While the total embryo count serves as an in utero proxy for the traditional brood size metric, the late-stage embryo counts provide an indicator of successful embryonic development, correlating with the hatching efficiency measurements. Compared to traditional reproductive toxicity assays, our assay offers several advantages, including (1) scalability enabled by well-plate-compatible liquid culture, (2) simultaneous testing of 24 independent populations, (3) high-resolution imaging of up to 40 C. elegans per population to quantify in utero embryonic development, and (4) single time point imaging, which reduces assay complexity while preserving multiparametric information content. Traditional brood size assays performed using NGM plates are labor-intensive and low-throughput [ 44 ]. Although microfluidic-based automated egg counting systems have been developed [ 49 – 51 ], these platforms are technically complex, require continuous imaging over 3–6 days, and consume large amounts of test substances due to constant perfusion. In contrast, our DART assay quantifies and classifies all in utero embryos within hermaphrodites on the first day of adulthood, when embryo production enters the peak phase. Assessing DART endpoints at this early adult stage offers several key advantages: (1) it enables scoring of high-quality embryos while avoiding confounding effects from age-related gonadal atrophy; (2) embryos are typically aligned in a single row, reducing phenotyping errors that arise from multilayered stacking in older worms; (3) the readout is independent of the egg-laying circuit, minimizing data drift caused by off-target neuromuscular effects; and (4) the assay is based on direct morphological features of embryogenesis rather than late-stage neuromuscular strength, which influences larval hatching outcomes. In addition, our DART assay relies on image analysis and scoring of brightfield images of wild-type worms and does not require any reporter strain or fluorescent biomarkers. This aspect facilitates application of our DART assay to large numbers of wild isolates with diverse genetic backgrounds, enabling characterization of genetic susceptibility to environmental toxicants using naturally occurring variation in C. elegans populations [ 22 ]. For any newly developed method, intra-laboratory repeatability (and ultimately inter-laboratory reproducibility) is vital for consideration by regulatory bodies [ 10 , 71 ]. To address repeatability, we established robust and detailed protocols for C. elegans husbandry, synchronization, chemical exposure, and culture conditions to minimize inter-experimental variability. Our DART assay is highly data-intensive, generating ~ 11,000 images and nearly ~ 325 GB of data in less than 30 minutes from a single experiment. To enable even higher throughput, we developed a minimal data acquisition protocol to reduce data volume while preserving the ability to produce all the DART data presented in this study. This optimized approach, suitable for large-scale studies, collects 2,184 images (~ 66 GB) from a single vivoChip by limiting imaging data from a single timepoint, 18 z-slice brightfield images. To manage such large volumes of data and support multiparametric phenotypic analysis, we developed the vivoScreen platform, which integrates reliable hardware components with rigorously validated and tested software, a centralized database infrastructure, and both local and cloud-based data storage systems. Every experiment is archived with comprehensive metadata, including experimental details, imaging conditions, analysis outputs, and final reports, ensuring full traceability. Despite the large data volume, image analysis is performed efficiently using an ML-based pipeline for DevTox parameters with an accuracy that is comparable to human scorers while requiring only a fraction of the time. Embryo phenotypes are then scored using semi-automated, user-friendly software designed to streamline manual embryo annotation and improve scoring consistency. The mean coefficient of variation (CV%) for 1% DMSO ranged from 1–5% for developmental endpoints and 6–17% for embryo-related endpoints, values well below thresholds considered acceptable for regulatory-approved assays, including those described for earthworms [ 37 ]. This low variability confers high statistical power, even with relatively few experimental replicates. Power analysis indicates that 3 independent experimental replicates are sufficient to detect > 23% changes across all endpoints, while smaller effect sizes can be reliably identified with additional replicates. While the late-stage embryo phenotype exhibited the highest variability, it was also the most sensitive endpoint in terms of EC 50 . The higher variation likely reflects the oscillatory pattern of egg laying, as embryos are laid in clusters every 1–2 hours [ 72 ]. Depending on timing, worms may be immobilized and imaged immediately before or after an egg-laying event, introducing natural biological variability into embryo counts. In addition, we confirmed that 1% DMSO, used as the solvent control, does not contribute any significant adverse effects on embryo parameters, enabling testing at relatively high treatment concentrations of many chemicals [ 61 , 62 ]. We validated the DART assay using two reference chemicals: methylmercury and propiconazole. Methylmercury was chosen because of its extensive evaluation across variety of species and toxicology domains, including DART, neurotoxicity, metabolism, gene expression, stress response, and cell division [ 20 , 25 , 73 – 76 ]. Propiconazole was chosen as it is a widely used fungicide with well-characterized toxicology profiles in mammals and other model organisms [ 66 , 67 , 77 , 78 ]. Both chemicals showed concentration-dependent effects on all 6 endpoints studied within our DART assay. These endpoints exhibited differential sensitivities based on EC 50 values, with body volume emerging as the most sensitive developmental parameter and late-stage embryos as the most sensitive reproductive parameter. We observed narrow 95% confidence interval bands and low CV% values for the methylmercury control wells, indicating high assay precision. The assay quality, measured using SSMD, was good for ≥ 4 µM methylmercury, supporting its suitability as a positive assay control for all endpoints. The onset of DART toxicity for methylmercury in our assays, based on the late-stage embryo endpoint, occurred at ~ 1 µM, consistent with previously reported C. elegans DART studies [ 20 ]. For propiconazole, adverse effects on fathead minnow fecundity and Daphnia embryonic development have been reported at 1.46 µM (0.5 mg/L), which is within an order of magnitude of our EC 10 value of 14 µM [ 77 , 78 ]. A recent study in zebrafish reported a DART EC 50 of a 1.62 mg/L, (~ 2 µM propiconazole equivalent) for a commercial formulation, and additional studies have observed developmental abnormalities at 0.73-1 µM [ 79 – 81 ]. These comparisons across different species could be due to differences in chemical uptake or accumulation, which can be addressed by measuring tissue-level concentration using analytical methods. In C. elegans , a developmental toxicity assay and a brood-size-based reproductive toxicity assay reported an EC 50 of 223–249 µM and 141 µM, respectively [ 82 , 83 ]. Using the most sensitive late-stage embryos parameter in our assay, these previously reported values are ~ 1.8–3.2× higher than the EC 50 value of 78 µM observed in our assay. These differences could be associated with the differences in the food source, DMSO concentration, exposure duration, and endpoint definition. While absolute effective concentrations of propiconazole in our results are higher than those reported in vertebrate systems, the C. elegans -based DART platform offers substantial advantages in cost, scalability, and throughput. Compared with rodent colonies and fish aquaria, this approach requires significantly lower resources and eliminates labor-intensive and subjective phenotyping of complex reproductive phenotypes on an animal-by-animal basis. Moreover, imaging-based multiparametric design enables objective and quantitative assessment of developmental and reproductive phenotypes. All endpoints studied in our assay are biologically interlinked, and reproductive (embryo) phenotypes may, in part, be affected by developmental perturbations. However, the effective concentrations for the different endpoints were clearly separated, suggesting distinct sensitivities rather than a uniform secondary response. Total embryo numbers declined at lower concentrations than those required to reduce body size, indicating that the germline is specifically affected and that the observed reproductive effects are not simply a consequence of delayed general development. Moreover, reductions in late-stage embryos occurred before a decrease in total embryo counts, consistent with direct impairment of embryonic development rather than a simple reduction in embryo production. Importantly, the worms remained > 97% viable and motile across all tested concentrations. Therefore, the measured effects reflect DART responses rather than the apical endpoints such as lethality. The novel DART platform we have developed is both faster and more cost-effective than traditional vertebrate-based assays. Thanks to the short life cycle of C. elegans , culture and treatment durations are measured in days, while chip loading and imaging times are < 30 min per experiment. In addition to improved efficiency, the assay eliminates the need for vertebrate animals subject to welfare regulations. Data analysis can also be performed rapidly. The ML inference pipeline used for body-dimension quantification can process a single chip (~ 1,000 C. elegans ) in < 5 minutes using standard desktop PCs. Comparable processing times are expected for automated embryo detection once fully optimized ML-based models are implemented. Across these studies, we collected and annotated data from over ~ 389,000 embryos across ~ 9,200 worms under various treatment conditions, including detailed classification of developmental stage and extraction of xyz centroid coordinates within image stacks. This extensive dataset provides a strong foundation for training next-generation ML models to fully automate embryo detection and staging. We are currently developing new ML models that will enable fully automated phenotyping of all DART endpoints within very short processing times, eliminating the current time-intensive manual analysis and substantially increasing scalability. This study demonstrates the sensitivity, repeatability, and precision of our DART assay. The implementation of automated embryo scoring will further enable high-throughput screening, allowing rapid and cost-effective testing of large chemical libraries for hazard prioritization and comparative safety assessment within defined chemical spaces. Declarations Acknowledgments: The authors would like to thank Thao Vu, Laura Murphy, Kyra Kwek, Jennifer Nguyen, Nishi Patel, Ruchi Birur, Cale Heiniger, and Seher Mamun for their help with manual scoring of the brightfield images of worms for in utero embryo scoring. We also thank Nikko Chavez and Dr. Jacob Moore for the initial version of the image acquisition hardware and software development. We thank The Caenorhabditis Natural Diversity Resource (CaeNDR) for providing the N2 strain. Author contributions: A.L., A.S., G.C., and J.B. maintained the worms, setup the 24-well liquid cultures. A.M. developed automated image analysis for body parameters. S.G. developed the image acquisition and graphical user interface (GUI) for embryo scoring. E.H., A.L., and J.B. ran microfluidic device experiments. A.L., J.B., and S.M. analyzed the data, prepared the plots, and performed the statistics. S.M. and A.B. conceptualized the projects and planned the experiments. A.L., S.M., and A.B. prepared the manuscript. A.L., S.M., and A.B. finalized the discussion of the results. Funding: This study was funded by the National Institutes of Health (NIH) and United States Department of Agriculture, National Institute of Food and Agriculture (USDA, NIFA), through Small Business Innovation Research grants from NCCIH (43AT013187), NIEHS (R43ES036387, R43ES033579, and R44ES033579), NIMH (R43MH118841 and R44MH118841), and USDA (2024-33530-41929). Data availability: The datasets generated and/or analyzed during the current study are available from the corresponding authors upon reasonable request. Competing interests statement: A.B., S.M., and E.H. are co-founders of vivoVerse, LLC and are the inventors of the vivoChip microfluidic technology and vivoScreen platform. A.B., S.M., A.L., A.S., and E.H. are inventors on several approved and ongoing patent applications. References Oecd Test No. 422: Combined Repeated Dose Toxicity Study with the Reproduction/Developmental Toxicity Screening Test. OECD Guidelines for the Testing of Chemicals, Section 4 (OECD Publishing, 2015). Oecd Test No. 421: Reproduction/Developmental Toxicity Screening Test. OECD Guidelines for the Testing of Chemicals, Section 4, (2025). Olson, H. et al. Concordance of the toxicity of pharmaceuticals in humans and in animals. Regul. Toxicol. Pharmacol. 32 , 56–67 (2000). Wang, Z. et al. Toward a Global Understanding of Chemical Pollution: A First Comprehensive Analysis of National and Regional Chemical Inventories Vol. 54, p. 2575–2584 (Environmental Science & Technology, 2020). 5. Schenk, B. et al. The ReProTect Feasibility Study, a novel comprehensive in vitro approach to detect reproductive toxicants. Reprod. Toxicol. 30 (1), 200–218 (2010). Rajagopal, R. et al. Beyond AOPs: A Mechanistic Evaluation of NAMs in DART Testing. Front. Toxicol. 4 , 838466–838466 (2022). Piersma, A. H. et al. Evaluation of an alternative in vitro test battery for detecting reproductive toxicants. Reprod. Toxicol. 38 , 53–64 (2013). Oecd, T. No. 455: Performance-Based Test Guideline for Stably Transfected Transactivation In Vitro Assays to Detect Estrogen Receptor Agonists and Antagonists. OECD Guidelines for the Testing of Chemicals, Section 4, (2021). Sychrová, E. et al. In vitro testicular toxicity of environmentally relevant endocrine-disrupting chemicals: 2D vs. 3D models of prepubertal Leydig TM3 cells. Environ. Toxicol. Pharmacol. 93 , 103869–103869 (2022). Höss, S. et al. Interlaboratory comparison of a standardized toxicity test using the nematode Caenorhabditis elegans (ISO 10872). Environ. Toxicol. Chem. 31 (7), 1525–1535 (2012). Behl, M. et al. Comparative Toxicity of Organophosphate Flame Retardants and Polybrominated Diphenyl Ethers to Caenorhabditis elegans. Toxicol. Sci. 154 (2), 241–252 (2016). Boyd, W. A. et al. A high-throughput method for assessing chemical toxicity using a Caenorhabditis elegans reproduction assay. Toxicol. Appl. Pharmacol. 245 (2), 153–159 (2010). Boyd, W. A. et al. Developmental Effects of the ToxCast Phase I and Phase II Chemicals in Caenorhabditis elegans and Corresponding Responses in Zebrafish, Rats, and Rabbits. Environ. Health Perspect. 124 , 586–593 (2016). Harlow, P. H. et al. Comparative metabolism of xenobiotic chemicals by cytochrome P450s in the nematode Caenorhabditis elegans. Sci. Rep. 8 , 13333 (2018). Harlow, P. H. et al. The nematode Caenorhabditis elegans as a tool to predict chemical activity on mammalian development and identify mechanisms influencing toxicological outcome OPEN. (2016). Camacho, J. A. et al. Reproductive-Toxicity-Related Endpoints in C. elegans Are Consistent with Reduced Concern for Dimethylarsinic Acid Exposure Relative to Inorganic Arsenic . J. Dev. Biol. , 11 (2). (2023). Hunt, P. R. et al. Developmental Exposures to Three Mammalian Teratogens Produce Dysmorphic Phenotypes in Adult Caenorhabditis elegans. Toxics 2025, Vol. 13, 13(7). (2025). Hunt, P. R., Olejnik, N. & Sprando, R. L. Toxicity ranking of heavy metals with screening method using adult Caenorhabditis elegans and propidium iodide replicates toxicity ranking in rat. Food Chem. Toxicol. 50 (9), 3280–3290 (2012). Chen, P. et al. Metal-induced neurodegeneration in C. elegans. Front. Aging Neurosci. 5 , 18 (2013). Helmcke, K. J. et al. Characterization of the effects of methylmercury on Caenorhabditis elegans. Toxicol. Appl. Pharmacol. 240 (2), 265–272 (2009). Hartman, J. H. et al. Xenobiotic metabolism and transport in Caenorhabditis elegans. J. Toxicol. Environ. Health B Crit. Rev. 24 (2), 51–94 (2021). Widmayer, S. J. et al. C. elegans toxicant responses vary among genetically diverse individuals. Toxicology 479 , 153292 (2022). Allard, P. et al. A C. elegans screening platform for the rapid assessment of chemical disruption of germline function. Environ. Health Perspect. 121 , 717–724 (2013). Ulaganathan, G. et al. Screening and characterization of 133 physiologically-relevant environmental chemicals for reproductive toxicity. Reprod. Toxicol. 126 , 108602 (2024). Wyatt, L. H. et al. Effects of methyl and inorganic mercury exposure on genome homeostasis and mitochondrial function in Caenorhabditis elegans. DNA repair. 52 , 31–31 (2017). Leung, M. C. et al. Caenorhabditis elegans generates biologically relevant levels of genotoxic metabolites from aflatoxin B1 but not benzo[a]pyrene in vivo. Toxicol. Sci. 118 (2), 444–453 (2010). Tang, B. et al. High-throughput assessment of toxic effects of metal mixtures of cadmium(Cd), lead(Pb), and manganese(Mn) in nematode Caenorhabditis elegans. Chemosphere 234 , 232–241 (2019). Williams, P. L. & Dusenbery, D. B. Using the nematode Caenorhabditis elegans to predict mammalian acute lethality to metallic salts. Toxicol. Ind. Health . 4 (4), 469–478 (1988). Currie, S. D. et al. The impact of early life exposure to individual and combined PFAS on learning, memory, and bioaccumulation in C. elegans. Environ. Pollut . 363 (Pt 2), 125257 (2024). Oecd Test No. 222: Earthworm Reproduction Test (Eisenia fetida/Eisenia andrei). OECD Guidelines for the Testing of Chemicals, Section 2, (2016). Gutner-Hoch, E. et al. Toxicity of engineered micro- and nanomaterials with antifouling properties to the brine shrimp Artemia salina and embryonic stages of the sea urchin Paracentrotus lividus. Environ. Pollut. 251 , 530–537 (2019). Cunningham, P. A. & Grosch, D. S. A comparative study of the effects of mercuric chloride and methyl mercury chloride on reproductive performance in the brine shrimp, Artemia salina. Environmental Pollution 1978. 15(2): pp. 83–99. (1970). Tkaczyk, A. et al. Daphnia magna model in the toxicity assessment of pharmaceuticals: A review. Sci. Total Environ. 763 , 143038 (2021). O., Test No. 211: Daphnia magna Reproduction Test, OECD Guidelines for the Testing of Chemicals . 2012, OECD Publishing, Paris. (2012). Rim, K. T. In silico prediction of toxicity and its applications for chemicals at work. Toxicol. Environ. Health Sci. 12 (3), 191–202 (2020). OECD, Test No. 207: Earthworm, acute toxicity tests . 1984, OECD Guidelines for the Testing of Chemicals. p. Section 2-Section 2. O., Test No. 222: Earthworm Reproduction Test (Eisenia fetida/Eisenia andrei), OECD Guidelines for the Testing of Chemicals . 2016, OECD Publishing, Paris. (2016). Lai, C. H. et al. Identification of Novel Human Genes Evolutionarily Conserved in Caenorhabditis elegans by Comparative Proteomics. Kaletta, T. & Hengartner, M. O. Finding function in novel targets: C. elegans as a model organism. Nat. Rev. Drug Discov . 5 , 387–398 (2006). Athar, F. & Templeman, N. M. C. elegans as a model organism to study female reproductive health. Comp. Biochem. Physiol. Mol. Integr. Physiol. 266 , 111152 (2022). Lints, R. & Hall, D. H. Reproductive system, germ line. , in WormAtlas , L.A. Herndon, Editor. (2009). Larigot, L. et al. Cytochromes P450 of Caenorhabditis elegans: Implication in Biological Functions and Metabolism of Xenobiotics. (2022). Li, Y. et al. Correlation of chemical acute toxicity between the nematode and the rodent The utility of any non-rodent model system for chemical toxicity screening depends on the level of corre. Cite this: Toxicol. Res. 2 , 403–403 (2013). Kwah, J. K. & Jaramillo-Lambert, A. Measuring Embryonic Viability and Brood Size in Caenorhabditis elegans . J. Vis. Exp. , 2023(192). Lee, S. Y. & Kang, K. Measuring the Effect of Chemicals on the Growth and Reproduction of Caenorhabditis elegans . J. Vis. Exp. , 2017(128). DuPlissis, A. et al. Machine learning-based analysis of microfluidic device immobilized C. elegans for automated developmental toxicity testing. Sci. Rep. 15 (1), 15 (2025). Mondal, S. et al. Large-scale microfluidics providing high-resolution and high-throughput screening of Caenorhabditis elegans poly-glutamine aggregation model. Nat. Commun. 7 , 13023 (2016). Mondal, S. et al. High-Content Microfluidic Screening Platform Used To Identify sigma2R/Tmem97 Binding Ligands that Reduce Age-Dependent Neurodegeneration in C. elegans SC_APP Model (ACS Chem Neurosci, 2018). Letizia, M. C. et al. Microfluidics-enabled phenotyping of a whole population of C. elegans worms over their embryonic and post-embryonic development at single-organism resolution. Microsyst. Nanoeng . 4 , 6 (2018). Li, S., Stone, H. A. & Murphy, C. T. A microfluidic device and automatic counting system for the study of C. elegans reproductive aging. Lab. Chip . 15 (2), 524–531 (2015). Banse, S. A. et al. The Egg-Counter: A novel microfluidic platform for characterization of Caenorhabditis elegans egg-laying. bioRxiv, : p. 2023.09.01.555781–2023.09.01.555781. (2023). Brenner, S. The genetics of Caenorhabditis elegans. Genetics 77 , 71–94 (1974). Stiernagle, T. Maintenance of C. elegans. WormBook, : pp. 1–11. (2006). Groen, F. C., Young, I. T. & Ligthart, G. A comparison of different focus functions for use in autofocus algorithms. Cytometry 6 , 81–91 (1985). Sun, Y., Duthaler, S. & Nelson, B. J. Autofocusing in computer microscopy: selecting the optimal focus algorithm. Microsc Res. Tech. 65 , 139–149 (2004). Zhang, X. D. A pair of new statistical parameters for quality control in RNA interference high-throughput screening assays. Genomics 89 (4), 552–561 (2007). Zhang, X. D. Illustration of SSMD, z Score, SSMD*, z* Score, and t Statistic for Hit Selection in RNAi High-Throughput Screens. SLAS Discovery . 16 (7), 775–785 (2011). Zhang, X. D. Optimal high-throughput screening: practical experimental design and data analysis for genome-scale RNAi research 47–53 (Cambridge University Press, 2011). Nlm, C. et al. Advanced Assay Development Guidelines for Image-Based High Content Screening and Analysis. (2017). Chisholm, A. D. & Hardin, J. Epidermal morphogenesis . WormBook. pp. 1–22. (2005). AlOkda, A. & Raamsdonk, J. M. V. Effect of DMSO on lifespan and physiology in C. elegans: Implications for use of DMSO as a solvent for compound delivery. microPublication Biology, 2022: p. (2022). 10.17912/micropub.biology.000634-10.17912/micropub.biology.000634 Calahorro, F., Holden-Dye, L. & O'Connor, V. Impact of drug solvents on C. elegans pharyngeal pumping. Toxicol. Rep. 8 , 1240–1240 (2021). O., OECD (1984), Test No. 207: Earthworm, Acute Toxicity Tests, OECD Guidelines for the Testing of Chemicals . 1984, OECD Publishing, Paris. (1984). Helmcke, K. J. & Aschner, M. Hormetic effect of methylmercury on Caenorhabditis elegans. Toxicol. Appl. Pharmcol. 248 (2), 156–156 (2010). Scheuhammer, A. M. et al. Effects of Environmental Methylmercury on the Health of Wild Birds, Mammals, and Fish. (2007). https://doi.org/10.1579/0044-7447 36[12:EOEMOT]2.0.CO;2, 2007. 36(1): pp. 12–19. Orlandini Costa, N. et al. Evaluation of the reproductive toxicity of fungicide propiconazole in male rats. (2015). Svanholm, S. et al. Developmental reproductive toxicity and endocrine activity of propiconazole in the Xenopus tropicalis model. Sci. Total Environ. 753 , 141940–141940 (2021). Carlier, M. P. et al. Using Caenorhabditis elegans and zebrafish embryos (Danio rerio) in a combined approach to screen chemicals for developmental (neuro)toxicity and reproductive toxicity. NAM J. 2 , 100080 (2026). Killian, D. J. & Hubbard, E. J. A. Caenorhabditis elegans germline patterning requires coordinated development of the somatic gonadal sheath and the germ line. Dev. Biol. 279 (2), 322–335 (2005). McCarter, J. et al. On the Control of Oocyte Meiotic Maturation and Ovulation inCaenorhabditis elegans. Dev. Biol. 205 (1), 111–128 (1999). Fischer, M. et al. Repeatability and Reproducibility of the RTgill-W1 Cell Line Assay for Predicting Fish Acute Toxicity. Toxicol. Sci. 169 (2), 353–364 (2019). Waggoner, L. E. et al. Control of alternative behavioral states by serotonin in Caenorhabditis elegans. Neuron 21 (1), 203–214 (1998). VanDuyn, N. et al. SKN-1/Nrf2 inhibits dopamine neuron degeneration in a Caenorhabditis elegans model of methylmercury toxicity. Toxicol. Sci. 118 (2), 613–624 (2010). Hu, K. et al. Ecotoxicity Risk of Low-Dose Methylmercury Exposure to Caenorhabditis elegans: Multigenerational Toxicity and Population Discrepancy. Chem. Res. Toxicol. 34 (4), 1114–1123 (2021). Xing, X. J. et al. Exposure to lead and mercury in young larvae induces more severe deficits in neuronal survival and synaptic function than in adult nematodes. Arch. Environ. Contam. Toxicol. 56 (4), 732–741 (2009). McElwee, M. K. et al. Comparative toxicogenomic responses of mercuric and methyl-mercury. BMC Genomics 2013. 14(1): pp. 698–698-. (2013) 14:1. Skolness, S. Y. et al. Propiconazole Inhibits Steroidogenesis and Reproduction in the Fathead Minnow (Pimephales promelas). Kast-Hutcheson, K., Rider, C. V. & LeBlanc, G. A. The fungicide propiconazole interferes with embryonic development of the crustacean Daphnia magna. Environ. Toxicol. Chem. 20 (3), 502–509 (2001). Souders, C. L. et al. Sub-lethal effects of the triazole fungicide propiconazole on zebrafish (Danio rerio) development, oxidative respiration, and larval locomotor activity . Neurotoxicol. Teratol. , 74. (2019). Teng, M. et al. Parental exposure to propiconazole at environmentally relevant concentrations induces thyroid and metabolism disruption in zebrafish (Danio rerio) offspring: An in vivo, in silico and in vitro study. Ecotoxicol. Environ. Saf. 242 , 113865–113865 (2022). Sebastião, M. V. R. et al. Ecotoxicological assessment of pesticide mixtures: effects of imidacloprid and propiconazole on Danio rerio and Astyanax lacustris . Aquat. Toxicol. , 287. (2025). Boyd, W. A. et al. Developmental effects of the ToxCast™ phase I and phase II chemicals in caenorhabditis elegans and corresponding responses in Zebrafish, Rats, and Rabbits. Environ. Health Perspect. 124 (5), 586–593 (2016). Wittkowski, P. et al. Caenorhabditis elegans As a Promising Alternative Model for Environmental Chemical Mixture Effect AssessmentA Comparative Study. (2019). Additional Declarations Competing interest reported. A.B., S.M., and E.H. are co-founders of vivoVerse, LLC and are the inventors of the vivoChip microfluidic technology and vivoScreen platform. A.B., S.M., A.L., A.S., and E.H. are inventors on several approved and ongoing patent applications. Supplementary Files vivoDARTPaperSupp.pdf Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 15 May, 2026 Reviews received at journal 11 May, 2026 Reviewers agreed at journal 21 Apr, 2026 Reviewers agreed at journal 20 Apr, 2026 Reviewers agreed at journal 20 Apr, 2026 Reviews received at journal 18 Apr, 2026 Reviewers agreed at journal 17 Apr, 2026 Reviewers agreed at journal 08 Apr, 2026 Reviewers invited by journal 06 Apr, 2026 Editor invited by journal 05 Mar, 2026 Editor assigned by journal 04 Mar, 2026 Submission checks completed at journal 04 Mar, 2026 First submitted to journal 01 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9004834","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":619485241,"identity":"ccf6a10d-e4c2-49e3-b368-33616ea5498b","order_by":0,"name":"Sudip Mondal","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Sudip","middleName":"","lastName":"Mondal","suffix":""},{"id":619485242,"identity":"8521b460-09ea-40c9-a784-20157c773234","order_by":1,"name":"Adam Laing","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Adam","middleName":"","lastName":"Laing","suffix":""},{"id":619485243,"identity":"8c3bf101-7f68-43b6-86a1-68bbf65bf6fa","order_by":2,"name":"Amber Shen","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Amber","middleName":"","lastName":"Shen","suffix":""},{"id":619485244,"identity":"80a48799-d504-4ed6-a5b9-31a74b7baaaf","order_by":3,"name":"Evan Hegarty","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Evan","middleName":"","lastName":"Hegarty","suffix":""},{"id":619485245,"identity":"cd578c2b-5eb2-4117-ae87-60ac09a19f45","order_by":4,"name":"Abhishri Medewar","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Abhishri","middleName":"","lastName":"Medewar","suffix":""},{"id":619485246,"identity":"3aeea3d0-cb8c-4053-86da-31feb534d578","order_by":5,"name":"Sebastian Gomez","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Sebastian","middleName":"","lastName":"Gomez","suffix":""},{"id":619485247,"identity":"684c0cd0-3f9c-42e3-8c32-a020ef72cba8","order_by":6,"name":"Gina Carrion","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Gina","middleName":"","lastName":"Carrion","suffix":""},{"id":619485248,"identity":"f98e9632-d5f2-473d-b021-ff49dc525ea8","order_by":7,"name":"Julia Brown","email":"","orcid":"","institution":"vivoVerse, LLC","correspondingAuthor":false,"prefix":"","firstName":"Julia","middleName":"","lastName":"Brown","suffix":""},{"id":619485249,"identity":"50059dd8-da75-45d2-8c4b-1c7f6cc504b6","order_by":8,"name":"Adela Ben-Yakar","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDUlEQVRIiWNgGAWjYBACAwiVwMDA3gCkChAyjA0EtfAcAFJALg/xWiQSIFyCWszZew8w89SkyZvPfJ344YEBQ+J+9h7Djz8YbGQ3HMCuxbLnXAIzz7Ecwzm3czdLAB2W2MNzLFmahyHNGJcWgxs55r952CoYZ0jnboBokUg+IM3AcDgRp5b7bwyYef5V2M+QPLv5B1iL/MPmnz8Y/uPWcoPHgJm3LSdxhgTvNqgtzMckeBgO4NRi2ZNjwDi3Ly15Bk/uNosEAwnjnjNpadY8BsnGM3FoMWc/Y8Dw5luy7Qz2s5tv/qiwkW1vP2MMZNjJ9uHQAgJMPAi2BMzBuJWDAOMP/PKjYBSMglEw0gEAH+1bDcdNbHQAAAAASUVORK5CYII=","orcid":"","institution":"The University of Texas at Austin","correspondingAuthor":true,"prefix":"","firstName":"Adela","middleName":"","lastName":"Ben-Yakar","suffix":""}],"badges":[],"createdAt":"2026-03-02 02:38:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9004834/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9004834/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106726617,"identity":"1bed47f3-d41b-449a-83c5-cd5cf9c1d280","added_by":"auto","created_at":"2026-04-12 18:36:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":238714,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSchematic of the assay workflow for quantifying DART phenotypes in \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eC. elegans\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003efollowing exposure to test substances. (A)\u003c/strong\u003e Synchronized L1-stage \u003cem\u003eC. elegans\u003c/em\u003e are cultured in liquid media in varying concentrations of a test substance or control chemical. After 72 hours of exposure, when control worms have reached D1 adult stage, worms are loaded into the vivoChip-24x microfluidic trapping and immobilization device for rapid, high-resolution 3D imaging. Software-assisted image analysis is then used to quantify embryonic phenotypes (manual scoring of ~30-40 embryos across ~1,000 worms, requiring ~48 hours) and developmental phenotypes (automated ML-based quantification of body dimensions completed in \u0026lt; 5 minutes) associated with DART toxicity. \u003cstrong\u003e(B)\u003c/strong\u003e Overview of DART and apical (worm motility) endpoints captured using vivoChip imaging data and plate images, respectively. \u003cem\u003eIn utero \u003c/em\u003eembryo number and developmental stage, indicated by different shades of blue objects, are quantified along with body developmental endpoints.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/c37a5cbb83d423f3fc8bdc07.png"},{"id":106726502,"identity":"15804239-fa2f-4546-9ec0-acf33e2a2938","added_by":"auto","created_at":"2026-04-12 18:36:22","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":298989,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003evivoChip-24x design and operation for high-resolution imaging in DART studies. (A)\u003c/strong\u003e A schematic illustration of vivoChip-24x device containing 24 wells, each fluidically connected to 40 microfluidic trapping channels. \u003cstrong\u003e(B)\u003c/strong\u003e Representative image of 40 microfluidic channels under one of the 24 wells, containing N2 adult worms. Worms are trapped at different portions of the tapering channels according to body size. \u003cstrong\u003e(C)\u003c/strong\u003e The vivoChip-24x device mounted in its pressurized holder for controlled loading and immobilization. \u003cstrong\u003e(D)\u003c/strong\u003e Timeline and duration of each vivoChip operation workflow and total data volume produced from a single imaging experiment. \u003cstrong\u003e(E)\u003c/strong\u003eRepresentative 10× image of a trapped adult worm showing 31 early-stage embryos (blue) and 10 late-stage embryos (magenta). \u003cstrong\u003e(F)\u003c/strong\u003e Examples and schematics of embryos at developmental stages used for classification into early- and late-stage categories. Scale bars are 1 mm (B), 20 mm (C), and 100 µm (E and F).\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/9e903d0054c391bd41892ccd.png"},{"id":106727330,"identity":"3f33ab76-6bdf-4860-aacd-7aeae091bdb1","added_by":"auto","created_at":"2026-04-12 18:38:43","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":466593,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRepresentative DART parameters from a single biological replicate for different solvent concentrations. \u003c/strong\u003eA total of 6 technical replicates were performed for each condition (one row per plate), and 5 biological replicates were performed on different days: results from one representative biological replicate is shown here.\u003cstrong\u003e (A) \u003c/strong\u003eExperimental design of a single biological replicate, containing 6 identical technical replicate wells for each 4 concentrations of DMSO. For each parameter, well averages were calculated from up to 40 worms per well.\u003cstrong\u003e (B) \u003c/strong\u003eExample worm images from 0%, 0.2%, 0.5%, and 1% DMSO conditions. Scale bar is 100 µm.\u003cstrong\u003e (C-H)\u003c/strong\u003e Scatter plot of body length (C), body area (D), body volume (E), early-stage embryos (F), late-stage embryos (G), and total embryos (H) with each dot representing data from an individual worm for each well. The worms that passed the outlier filtering process for volume and total embryo for each well are presented in this plot. The mean ± SEM values for each well is represented in red. The vertical dashed lines separate the four DMSO conditions (A01-A06, B01-B06, C01-C06, and D01-D06).\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/a08514b70af26cd6847807ea.png"},{"id":106636208,"identity":"84a529f8-1437-4a75-9b38-136f56628510","added_by":"auto","created_at":"2026-04-10 16:52:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":44584,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCoefficient of variation between technical replicate wells for 1% DMSO conditions.\u003c/strong\u003e Scatter plot for the coefficient of variation (CV%) for 5 biological replicates for all 6 DART endpoints. Each data represents the CV% calculated using 6 technical replicates for 1% DMSO wells. The red line represents mean ± SEM. The dotted line represents the 30% CV line.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/447f91020d436bc6154782cb.png"},{"id":107489038,"identity":"902b9ead-747e-4e96-b842-3da5fe03a257","added_by":"auto","created_at":"2026-04-22 02:46:32","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":464697,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRepeatability of the DART assay across all five biological replicates for all six endpoints. (A) \u003c/strong\u003eExperimental design consisting of 5 independent biological replicates, each containing 6 identical technical replicate wells for 4 concentrations of DMSO. \u003cstrong\u003e(B)\u003c/strong\u003e Schematic, illustrating how we calculated the coefficient of variation (CV%) values for all possible combinations of 3 corresponding wells from 5 experiments. Each well contained up to 40 worms. \u003cstrong\u003e(C-H)\u003c/strong\u003e CV% values body length (C), body area (D), body volume (E), early-stage embryos (F), late-stage embryos (G), and total embryos (H). Each point represents the CV% derived from a unique combination of 3 corresponding wells across different experiments. The red horizontal line represents the mean ±SEM CV% values across all combinations for each DMSO concentration. The horizontal solid black line represents the 30% CV threshold. The vertical dashed lines separate the four DMSO conditions (A01-A06, B01-B06, C01-C06, and D01-D06).\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/62dc95a662406c34349591a7.png"},{"id":106636210,"identity":"4e084c00-ad26-4b08-89bc-49833892196d","added_by":"auto","created_at":"2026-04-10 16:52:53","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":499256,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eStatistical power for six DART parameters using 3 biological replicates.\u003c/strong\u003e Predicted statistical power of the assay to detect different hypothetical changes in each phenotype relative to control (1% DMSO). Power calculations assume 3 biological replicates per condition, the mean and standard deviation measured from all 60 combinations of 3-replicate average values in control populations, and a one-sided significance level of α = 0.05. The dotted horizontal line indicates the 80% power threshold considered acceptable for a good assay.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/cb56f87d0ad0a11affc22902.png"},{"id":106636211,"identity":"d5258733-600b-4573-89b2-0e4398b6616d","added_by":"auto","created_at":"2026-04-10 16:52:53","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":297731,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConcentration-response curves of six DART parameters following methylmercury exposure.\u003c/strong\u003e The effects of methylmercury on \u003cstrong\u003e(A) \u003c/strong\u003ebody length, \u003cstrong\u003e(B) \u003c/strong\u003ebody area, \u003cstrong\u003e(C) \u003c/strong\u003ebody volume, \u003cstrong\u003e(D) \u003c/strong\u003eearly-stage embryos, \u003cstrong\u003e(E) \u003c/strong\u003elate-stage embryos, and \u003cstrong\u003e(F) \u003c/strong\u003etotal embryos. Data for 0-4 µM methylmercury were obtained from 3 biological replicates performed using the D1 vivoChip devices (black symbols). Data for 5-10 µM methylmercury were obtained from 3 separate experiments performed using the L4 vivoChip devices (blue symbols), as worms at these concentrations were too small to be reliably trapped in the D1 devices. The combined well average values from both datasets were fitted with a 4-parameter logistic curve (solid black lines). EC\u003csub\u003e50\u003c/sub\u003e estimates are indicated with dotted lines with corresponding 95% confidence interval bands shaded in light-blue color.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/ce9d4cfac72f1bbe73882256.png"},{"id":106636213,"identity":"6132e302-cb83-4152-888d-40942ea92f1a","added_by":"auto","created_at":"2026-04-10 16:52:53","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":282759,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eConcentration-response curves for DART parameters following propiconazole exposure.\u003c/strong\u003e The effects of propiconazole on \u003cstrong\u003e(A) \u003c/strong\u003ebody length, \u003cstrong\u003e(B) \u003c/strong\u003ebody area, \u003cstrong\u003e(C) \u003c/strong\u003ebody volume, \u003cstrong\u003e(D) \u003c/strong\u003eearly-stage embryos, \u003cstrong\u003e(E) \u003c/strong\u003elate-stage embryos, and \u003cstrong\u003e(F) \u003c/strong\u003etotal embryos. Data for 0-178 µM propiconazole were obtained from 3 biological replicates performed using the D1 vivoChip devices (black symbols). Data for 334 µM propiconazole were obtained from 3 separate experiments performed using the L4 vivoChips (blue symbols), as worms at this concentration were too small to be reliably trapped in the D1 devices. The concentration-response curves were fitted to the combined well-average values from both datasets with a 4-parameter logistic curve (solid black lines). The EC\u003csub\u003e50\u003c/sub\u003e estimates are indicated with dotted lines, with corresponding 95% confidence interval bands shaded in light-blue color.\u003c/p\u003e","description":"","filename":"image8.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/46942873b19763bc79f32f13.png"},{"id":106726674,"identity":"fbc133d2-197f-4de8-8867-c535f54df336","added_by":"auto","created_at":"2026-04-12 18:37:00","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":141496,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eComparison of normalized concentration-response curves for selected DART endpoints. (A-B) \u003c/strong\u003eConcentration-response curves were normalized relative to the control averages for each biological replicate. Four phenotypes were plotted for methylmercury (A) and propiconazole (B). Endpoints included body length, total embryos, late-stage embryos, and the percentage of moving worms in the culture plate wells, used as a measure of viability. Estimated EC\u003csub\u003e10\u003c/sub\u003e concentrations are indicated for body length and total and late-stage embryos.\u0026nbsp; The leftward shift of the late-stage embryo curve relative to other sub-lethal endpoints, and its separation from viability loss, demonstrates its highest sensitivity among all the DART-related endpoints and independence from apical endpoints such as lethality.\u003c/p\u003e","description":"","filename":"image9.png","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/a8170ad76cbe9a4643342097.png"},{"id":107672585,"identity":"6e1a41a7-26a7-4497-829c-e7771d5a52c6","added_by":"auto","created_at":"2026-04-23 22:34:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3319787,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/2446574a-526b-4246-99de-f89d090880a4.pdf"},{"id":106636205,"identity":"737fde06-cdd1-4eb5-a969-1f9a8d516fbd","added_by":"auto","created_at":"2026-04-10 16:52:53","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":1873047,"visible":true,"origin":"","legend":"","description":"","filename":"vivoDARTPaperSupp.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9004834/v1/16d450175dc1113857fcfca6.pdf"}],"financialInterests":"Competing interest reported. A.B., S.M., and E.H. are co-founders of vivoVerse, LLC and are the inventors of the vivoChip microfluidic technology and vivoScreen platform. A.B., S.M., A.L., A.S., and E.H. are inventors on several approved and ongoing patent applications.","formattedTitle":"A Robust, High-Content NAM for Repeatable and Predictive Developmental and Reproductive Toxicity Assessment in C. elegans","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDevelopmental and reproductive toxicity (DART) assessment is a vital part of product safety testing for chemicals, consumer products, and pharmaceuticals. Currently, most DART testing is carried out in mammalian models such as rats and rabbits for human health risk assessment [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. However, increased regulatory and societal pressure to reduce animal use has led to new legislative initiatives in the US, EU, and other regions aimed at limiting reliance on vertebrate models. In addition to ethical considerations, mammalian-based DART studies are costly and time-intensive. For example, a standard DART study with rats may require 2\u0026ndash;3 months of exposure and monitoring due to the long life cycles, substantial housing infrastructures, and complex experimental design. High costs also restrict sample size, potentially limiting statistical power. Despite their high genetic and physiological similarities to humans, rodent data alone are only 43% predictive of human toxicity outcomes [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The concordance improves to 71% when combined with a non-rodent species, supporting a multi-tier approach. With over 350,000 registered chemicals requiring new or updated safety evaluation [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e], there is an urgent need for New Alternative Methods (NAMs) to complement or replace traditional methods.\u003c/p\u003e \u003cp\u003eNAM-based DART approaches include \u003cem\u003ein silico\u003c/em\u003e prediction models [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e], \u003cem\u003ein vitro\u003c/em\u003e systems [\u003cspan additionalcitationids=\"CR7\" citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], \u003cem\u003eex vivo\u003c/em\u003e technologies [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e], and alternative invertebrate model organisms not subject to welfare regulations, such as \u003cem\u003eC. elegans\u003c/em\u003e [\u003cspan additionalcitationids=\"CR11 CR12 CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25 CR26 CR27 CR28\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e], earthworm [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e], brine shrimp [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e], and \u003cem\u003eDaphnia\u003c/em\u003e [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Although no single method can fully recapitulate human biology, a multi-tiered approach that integrates these complementary models can provide a more comprehensive and human-relevant toxicity profile for risk assessment. \u003cem\u003eIn silico\u003c/em\u003e methods remain constrained by limited, high-quality training datasets [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. \u003cem\u003eIn vitro\u003c/em\u003e and \u003cem\u003eex vivo\u003c/em\u003e systems are highly variable and often capture only isolated adverse outcome pathways (AOPs) and lack the integrated tissue- and organ-level interactions required for complex DART responses. Among invertebrates, standardized DART assessment protocols exist for \u003cem\u003eDaphnia\u003c/em\u003e and earthworms and are widely used for ecotoxicology risk assessment [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. Although no regulatory test guidelines currently exist for \u003cem\u003eC. elegans\u003c/em\u003e, the small model organism has been extensively used in scientific research, including numerous toxicology studies [\u003cspan additionalcitationids=\"CR12 CR13 CR14 CR15 CR16 CR17 CR18 CR19 CR20 CR21 CR22 CR23 CR24 CR25 CR26 CR27 CR28\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003cem\u003eC. elegans\u003c/em\u003e is a powerful, high-throughput model organism that can be grown quickly and inexpensively in large numbers using well-established techniques. Its small size and simple, transparent body make it highly amenable to imaging-based methods for high-content analysis. The \u003cem\u003eC. elegans\u003c/em\u003e genome shares extensive conservation with mammals, with up to 60\u0026ndash;80% and ~\u0026thinsp;83% homology to the human genome and proteome, respectively [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. It has complete neuromuscular, reproductive, and digestive systems with conserved signaling and metabolic pathways, enabling system-level AOPs studies. The reproductive system of a \u003cem\u003eC. elegans\u003c/em\u003e hermaphrodite consists of a germline, somatic gonad, uterus, and egg-laying circuitry [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. Many fundamental biological processes are conserved across species and are similarly preserved in \u003cem\u003eC. elegans\u003c/em\u003e. Notably, conserved evolutionary mechanisms governing reproduction include germ cell proliferation and differentiation, oogenesis, programmed cell death, chromosomal aneuploidy, and reproductive aging. In addition, \u003cem\u003eC. elegans\u003c/em\u003e contains at least 76 cytochrome P450 genes [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e], Phase I-III xenobiotic metabolism pathways [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e], and a functional microbiome, supporting toxicological evaluation of parent compounds and metabolites with unique tissue-level bioavailability profiles [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePrevious large-scale studies show concordance between \u003cem\u003eC. elegans\u003c/em\u003e and mammalian toxicity outcomes. Developmental toxicity endpoints demonstrate\u0026thinsp;~\u0026thinsp;53% balanced accuracy for predicting mammalian toxicity [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Germline integrity studies analyzing chromosome segregation defects in \u003cem\u003eC. elegans\u003c/em\u003e embryos achieved 69% balanced accuracy in predicting mammalian reproductive toxicity [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Acute toxicity studies of 17 non-acidic chemicals reported 88.5% concordance with rat lethal dose predictions, outperforming mouse\u0026ndash;rat concordance (87.9%) [\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. However, many of these studies rely on low-throughput, labor-intensive methodologies or low-resolution information to assess gross developmental defects and embryo fates upon chemical exposures [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e], limiting scalability and reproducibility. Advances in microfluidics [\u003cspan additionalcitationids=\"CR47 CR48 CR49 CR50\" citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e] and machine learning (ML)-assisted image analysis [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e] now enable robust, high-throughput, and multi-parametric analysis of \u003cem\u003eC. elegans\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eHere, we present a high-content imaging-based assay to quantify DART-related endpoints in \u003cem\u003eC. elegans\u003c/em\u003e using an automated high-throughput microfluidic platform that is amenable to automation and large chemical library screening [\u003cspan additionalcitationids=\"CR47\" citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Our assay can be completed within 3 days from initial worm plating and chemical treatment to endpoint imaging. All readouts are derived from brightfield image analysis and are compatible with wild-type strain, transgenic lines, and genetically diverse wild isolates. High-resolution imaging of the entire uterus in 3D enables identification and classification of every \u003cem\u003ein utero\u003c/em\u003e embryo, providing a more sensitive measure of toxicity than body size alone. Importantly, \u003cem\u003ein utero\u003c/em\u003e analysis of embryo development during early adulthood is unaffected by the confounding effects from age-associated decline in the reproductive health and off-target effects such as altered egg-laying behavior. Using 30 replicates (5 biological \u0026times; 6 technical replicates), we show that the DART assay is highly robust and produces repeatable results across multiple phenotypes, extended time periods, and independent of operators. As a case study, we evaluated methylmercury and propiconazole, chemicals with well-characterized DART profiles in \u003cem\u003eDaphnia\u003c/em\u003e, rodents, amphibians, and fish, to benchmark assay sensitivity and assess translational relevance.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStrains and maintenance\u003c/h2\u003e \u003cp\u003eWe used the N2 wild type strain (The \u003cem\u003eCaenorhabditis\u003c/em\u003e Natural Diversity Resource, CaeNDR) for all experiments. \u003cem\u003eC. elegans\u003c/em\u003e were maintained according to standard methods on nematode growth media (NGM) agar plates seeded with HB101 \u003cem\u003eEscherichia coli\u003c/em\u003e bacteria at 20\u0026deg;C [\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. Frozen stocks of strains were stored at -80\u0026deg;C and thawed annually to generate a new set of Dauer plates. Sealed Dauer plates were stored at 16\u0026deg;C and used to start fresh maintenance cultures every 3 months by chunking, minimizing genetic drift. We cultured worms for \u0026gt;\u0026thinsp;5 generations before using them in assays in a temperature and humidity-controlled environment.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eFood preparation\u003c/h3\u003e\n\u003cp\u003eWe streaked HB101 from \u0026minus;\u0026thinsp;80\u0026deg;C glycerol stock onto Luria-Bertani (LB) agar plates with streptomycin and incubated them overnight at 37\u0026deg;C to generate single colonies. A single colony was picked and used to inoculate an overnight 5 mL starter culture in LB broth containing streptomycin. To prepare the HB101 food stock, 200 \u0026micro;l of this starter culture was added to 200 mL LB broth and streptomycin in a 1,000 mL baffled flask. The flask was incubated for 16 hours with shaking at 180 rotations per min (rpm) to reach the late log growth phase. We measured the OD\u003csub\u003e600\u003c/sub\u003e using a spectrophotometer (DiluPhotometer\u0026trade;, Implen) and stored the culture at 4\u0026deg;C. Streak plates and starter or food stocks were used within 2 months and 2 weeks, respectively. Immediately before worm culture, we centrifuged a sufficient volume of the bacterial culture to generate the required volume of OD\u003csub\u003e600\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;3.0 food suspension. The LB supernatant was removed, and the bacterial pellet was resuspended in S media. This density was optimal for growing 80\u0026ndash;100 larval 1 (L1) to the day 1 (D1) adult stage in a 500 \u0026micro;l culture volume, with sufficient food still remaining after 72 hours to avoid starvation in the worm population prior to DART analysis.\u003c/p\u003e\n\u003ch3\u003eChemicals\u003c/h3\u003e\n\u003cp\u003eMethylmercury (II) hydroxide (Cat# 13395, Alfa Aesar, CAS# 1184-57-2, Batch# M25H004) and propiconazole (Cat# 45642-250MG, Sigma Aldrich, CAS# 60207-90-1, Batch# BCCD0480) were dissolved in dimethyl sulfoxide (DMSO, Cat# D2650, Sigma Aldrich, CAS# 67-68-5, Batch# RNBL9635) to prepare master stocks at 100\u0026times; the highest treatment concentration. The master stocks were transferred into 0.2 ml tubes as single-use aliquots and kept at -80\u0026deg;C until use. For the concentration-response assays, chemicals were serially diluted in DMSO immediately prior to treatment to generate 100\u0026times; working dilutions, ensuing a final DMSO concentration of 1.0% (v/v) in each well. Solvent control wells received DMSO at the same final concentration (1.0% v/v).\u003c/p\u003e\n\u003ch3\u003eWorm synchronization and large-scale culture\u003c/h3\u003e\n\u003cp\u003eTo generate large numbers of synchronized L1-stage larvae, two NGM plates were each seeded with four parent larval 4 (P\u003csub\u003e0\u003c/sub\u003e L4) worms. The worms were grown for 48 hours to allow P\u003csub\u003e0\u003c/sub\u003e to become adults and lay a large number of eggs. At this time, all 8 P\u003csub\u003e0\u003c/sub\u003e adults (2 NGM plates \u0026times; 4 adults/plate) were removed. The plates were incubated for an additional 48 hours, by which point the majority of F\u003csub\u003e1\u003c/sub\u003e generation had reached adulthood. Adult F\u003csub\u003e1\u003c/sub\u003e worms were collected by washing both plates with M9 buffer and treated with an alkaline sodium hypochlorite solution with alternating low- and high-speed shaking on a speed-controllable vortexer to fragment their bodies and release embryos. The bleaching solution was neutralized with 5 washes of M9 buffer. Embryos were allowed to develop and hatch in M9 buffer in a rotating glass conical tube for 24 hours. Hatched synchronized L1s were filtered through a 20 \u0026micro;m cell filter to remove unhatched eggs and debris. The total number of larvae was counted, and the suspension volume was adjusted to make a final density of ~\u0026thinsp;100 L1s/20 \u0026micro;l of M9 buffer.\u003c/p\u003e\n\u003ch3\u003eChemical treatment in liquid culture\u003c/h3\u003e\n\u003cp\u003eThe age-synchronized L1 suspension (~\u0026thinsp;100 larvae in 20 \u0026micro;l volumes) was dispensed into individual wells of a standard 24-well plate with 475 \u0026micro;l HB101 suspension (OD\u003csub\u003e600\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;3.0) prepared in S media [\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. Chemicals dissolved in DMSO (5 \u0026micro;l) were added to the designated wells to achieve the desired treatment conditions while keeping a constant solvent concentration across all wells. Vehicle control wells received 5 \u0026micro;l of DMSO solvent (1% v/v). The 24-well culture plates were sealed with airtight films (Cat# 232702, Thermo Scientific) to prevent evaporation and cross-contamination and cultured at 20\u0026deg;C for 72 hours. The plates were shaken every 6 hours using an automated shaking platform (Teleshake 95, Inheco) to homogenize the well contents, including the HB101 distribution, shed cuticles from molting, and the intestinal discharge from the defecation cycles over 24 hours.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eRange-finding assay and live worm quantification\u003c/h2\u003e \u003cp\u003eFor each chemical tested, we first performed a range-finding assay using at least 20 concentrations (0.009\u0026ndash;90 \u0026micro;M for methylmercury and 0.01734\u0026ndash;1730 \u0026micro;M for propiconazole) based on published work. Age-synchronized L1s were exposed to different chemical concentrations for 72 hours to determine the maximum soluble concentration in culture medium and assess any significant lethality or growth arrest at the L1 stage. Age-synchronized L1s were exposed to these concentrations in 24-well plates for 72 hours, after which culture plates were imaged and manually analyzed prior to chip transfer. The plate images were used to identify the percentage of live worms and to identify the highest concentration at which the worms developed beyond the L4 stage, ensuring suitability for subsequent vivoChip loading and high-resolution imaging for DART phenotyping.\u003c/p\u003e \u003cp\u003eSpecifically, prior to chip imaging, a single-frame and timelapse images (10 s at 10 fps with a 2\u0026times;, 0.08 NA objective) were acquired from each 24-well culture plate using the vivoScreen system (described below), generating\u0026thinsp;~\u0026thinsp;10.8 GB of data per plate. Single-frame images were inspected to ensure that control wells developed to D1 adult stage according to our predefined assay acceptance criteria (\u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e). Time-lapse images were inspected to count all worms within the 10,744 x 6,290 \u0026micro;m\u003csup\u003e2\u003c/sup\u003e field of view (FOV), and spontaneous movement in liquid culture was used as the criterion for viability. The percentage of moving (alive) worms per well was then calculated from 3 biological replicates on different days.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003evivoChip-24x design and operation\u003c/h3\u003e\n\u003cp\u003eMicrofluidic-based \u003cem\u003eC. elegans\u003c/em\u003e immobilization chips (vivoChip-24x, vivoVerse), fabricated from a transparent, low-chemical-absorptivity plastic with broad optical transmission spectrum [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e], were used in this study. The device consists of three bonded layers: (i) a top layer containing 24 rectangular wells arranged with 9 mm spacing for compatibility with standard 24-well plates, (ii) an intermediate microfluidic layer containing trapping channels, and (iii) an ultra-thin\u0026thinsp;~\u0026thinsp;80 \u0026micro;m bottom substrate layer enabling high-resolution optical access.\u003c/p\u003e \u003cp\u003eEach well has a loading capacity of 250 \u0026micro;L and is fluidically connected at its bottom surface to an array of 40 parallel tapering microfluidic channels, as previously described [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. In total, the chip contains 960 channels that orient and immobilize\u0026thinsp;~\u0026thinsp;40 \u003cem\u003eC. elegans\u003c/em\u003e per well, enabling rapid, high-resolution imaging of 24 distinct worm populations. The channels possess a 3D tapering geometry with an aspect ratio (width to height ratio) close to 1.0, promoting lateral worm orientation. Gradual reduction in channel dimensions allows size-dependent trapping along defined channel regions.\u003c/p\u003e \u003cp\u003eTwo vivoChip-24x designs were used: the vivoChip-24x-D1 design with larger channel geometry (tapering from 98 \u0026micro;m \u0026times; 102 \u0026micro;m at the channel entrance to 24 \u0026micro;m \u0026times; 40 \u0026micro;m at the exit) and vivoChip-24x-L4 design based on narrower geometry (tapering further to 10 \u0026micro;m \u0026times; 17 \u0026micro;m at the narrowest region). The D1 design is optimized to trap day 1 (D1) adult worms containing 2\u0026ndash;6 \u003cem\u003ein utero\u003c/em\u003e embryos at the narrowest channel region while allowing laid eggs to pass through. The L4 design enables trapping smaller worms, including early L4 stage worms, and is better suited for populations with fewer or no embryos, although it carries a higher risk of blockage from laid eggs and clogging the microfluidic channels.\u003c/p\u003e \u003cp\u003eAccordingly, two functional designs were used: the vivoChip-24x-D1 (based on the larger channel geometry) and the vivoChip-24x-L4 (based on the narrower geometry). The D1 chip was preferentially used because its wider channels prevented blockage by laid eggs and therefore had a greater number of channels containing trapped worms. The L4 chip was mainly used for worms treated with high concentrations of chemicals that slowed development, resulting in reduced body size and fewer or no \u003cem\u003ein utero\u003c/em\u003e embryos.\u003c/p\u003e \u003cp\u003eFollowing removal of the culture plate seal, worms were transferred from each culture plate to the corresponding vivoChip well using a multichannel pipette. Since vivoChip-24x-L4 can trap L4-stage or larger worms, wells containing populations younger than L4 were filled with molten 2% agarose gel to block the channels in those specific wells. The agar block prevented continuous media flow through the open channels and stabilized pressure across all wells during vivoChip imaging.\u003c/p\u003e \u003cp\u003eThe chip was placed in a custom vivoChip holder with a sealing gasket to provide controlled fluidic pressurization, as previously described [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. The vivoChip holder carrying the vivoChip was placed on the microscope stage for imaging and flow was initiated using our vivoCube+ microfluidic control system. Ten on/off pressure cycles were applied, resulting in complete immobilization of \u003cem\u003eC. elegans\u003c/em\u003e across all 960 channels within ~\u0026thinsp;3 minutes.\u003c/p\u003e\n\u003ch3\u003eAutomated image acquisition hardware\u003c/h3\u003e\n\u003cp\u003eThe vivoScreen system comprises integrated image acquisition hardware, custom software modules (vivoImager), a centralized searchable Postgres SQL database, and hybrid in-house and cloud-based data storage (\u003cb\u003eSupplementary Fig.\u0026nbsp;1\u003c/b\u003e). The image acquisition hardware consists of a customized large FOV inverted microscope (IX73, Evident), a precision \u003cem\u003exyz\u003c/em\u003e motorized stage (MS2000, Applied Scientific Instrumentation), a scientific CMOS camera (IRIS-15, Teledyne), and bright light sources (TLED, Sutter Instrument). The large-FOV setup enables imaging of the entire chip area with a 2\u0026times;, 0.08 NA objective and 8 microfluidic channels simultaneously using a 10\u0026times;, 0.4 NA objective, capturing all 40 channels with only 5 FOVs. The 10\u0026times; objective produces a 2.15 \u0026times; 1.26 mm\u003csup\u003e2\u003c/sup\u003e FOV, sufficient to capture the full lengths of adult worms across 8 channels with a lateral pixel resolution of 0.425 \u0026micro;m. For DART analysis, volumetric brightfield imaging was performed using 18 z-slices at 6 \u0026micro;m step size to capture the entire worm volume. According to the Nyquist sampling criterion, the theoretical axial resolution for a 0.4 NA objective is 6.5 \u0026micro;m, corresponding to a recommended z-step of 3.25 \u0026micro;m. However, to maintain volumetric imaging speed\u0026thinsp;~\u0026thinsp;1 volume per second with our data readout speed, we used a 6 \u0026micro;m step size. We captured five timelapse 3D hyperstack images at 1-s intervals to analyze twitching in the late-stage embryos if needed.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eAutomated image acquisition software and process\u003c/h2\u003e \u003cp\u003eTo acquire the imaging data, we used custom-developed image acquisition software (vivoImager, \u003cb\u003eSupplementary Fig.\u0026nbsp;2\u003c/b\u003e). We captured 10\u0026times;, 0.4 NA 3D brightfield stacks (18 z-slices) of all worms, centered around the best focal plane of a fiduciary marker (\u003cb\u003eSupplementary Fig.\u0026nbsp;3A\u003c/b\u003e). Each vivoChip-24x well contains a cross-shaped fiduciary marker, integrated into the microfluidic channel layer. This feature enables automatic focal plane identification using an edge-detection algorithm (\u003cb\u003eSupplementary Fig.\u0026nbsp;3B\u003c/b\u003e). Initially, the software loads a configuration file containing experimental metadata (experiment name, strain name, developmental stage, chemical name, concentration, solvent, etc.) and imaging parameters (illumination intensity, filter sets, camera settings, stage locations, etc.). Then the software initiates all hardware components, calibrates the stage for translation range, and positions the stage at the cross-shaped fiduciary marker of the reference well (B02). Using a pre-defined chip map, the system sequentially navigates to the fiduciary marker locations in each of the 24 wells to determine the best focal plane. Variations in focal position primarily arise from substrate curvature during chip pressurization.\u003c/p\u003e \u003cp\u003eWhile the relative channel locations are known from the chip map, the positions of the trapped worms within the 40 parallel, 3 mm-long channels may vary depending on their size. To optimally center the 1.3 mm \u0026times; 2.2 mm FOV to capture the entire body of as many worms as possible, a low-magnification overview image of the entire chip is first captured using a 2\u0026times;, 0.08 NA objective. Then, all immobilized worms are detected using a trained YOLO object detection model [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], and the relative stage positions are calculated for the 5 FOVs needed to image the worms at 10\u0026times;, 0.4 NA (8 \u0026times; 150 \u0026micro;m spaced channels can fit in each 1.3 mm wide FOV) (\u003cb\u003eSupplementary Fig.\u0026nbsp;3C\u003c/b\u003e). The system then automatically switches to the 10\u0026times; objective and captures brightfield z-stack images using high-precision 3-axis translational stages. For each well, 5 FOVs are imaged, each comprising 18 z-planes (6 \u0026micro;m step size) and five timepoints in brightfield for volumetric timelapse acquisition (\u003cb\u003eSupplementary Fig.\u0026nbsp;3D\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eA single vivoChip experiment generates 12,984 images (~\u0026thinsp;390 GB). Raw image data and associated metadata (timestamps, xyz coordinates, image capture and illumination parameters, treatment conditions etc.) are automatically transferred to a high-capacity network attached storage (NAS) server for preprocessing and downstream analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eImage processing and phenotypic scoring\u003c/h2\u003e \u003cp\u003eFor each FOV with 8 channels, channel boundaries were identified by detecting fiduciary markers within each well and applying pre-defined internal channel spacing from the chip map. Channel coordinates were stored in the centralized database for subsequent multiparametric analysis.\u003c/p\u003e \u003cp\u003eAutomated worm detection was performed using a deep learning\u0026ndash;based method [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. In brief, an image slice with optimal focus was identified from each z-stack by maximizing the variance of the Laplace-transformed image [\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. An improved 2.5D U-Net architecture, including a classification head at the bottleneck, was trained, tested, and used to first classify each channel as full (entire worm within the FOV), partial (partial worm within the FOV), or none (no worm present) and then predict a segmentation mask for channels classified as full. For the new model, we expanded the training dataset to include small-sized worms close to the channel exit, decoupled the image-level classification and the pixel-wise segmentation models to facilitate independent optimization and model learning with 1 and 5 z-slice images around the best focus image, respectively. The segmentation mask, predicted by the new model, was used to calculate body length (via skeletonization of the mask), area (total segmented pixels), and volume (mask area integrated with channel heights) for each full worm.\u003c/p\u003e \u003cp\u003eEmbryonic phenotyping was performed manually on all channels predicted as \u0026ldquo;full\u0026rdquo; by the ML model. All \u003cem\u003ein utero\u003c/em\u003e embryos in these channels were scored and classified according to their developmental stage. Ten trained scorers independently scored and classified embryos using a custom graphical user interface (vivoAnalyzer, \u003cb\u003eSupplementary Fig.\u0026nbsp;4\u003c/b\u003e). The software displays individual channels with zoom and z-plane navigation capabilities, allowing users to place a marker on each embryo and classify it as early- or late-stage embryo using the two-fold stage as the classification threshold. All automated image processing outputs (worm classification labels and segmentation masks) and manual annotations (marker positions and classifications of each embryo) were stored in the centralized database and exported for statistical analysis. The majority of embryo scoring was conducted in a blinded manner, with treatment conditions concealed from scorers.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eStatistical analysis\u003c/h2\u003e \u003cp\u003eFor DART data analysis, datasets were first cleaned to remove any worms for which the body segmentation mask failed, preventing dimensional calculations, or channels marked as unscorable by a user due to intestinal obstruction of the uterus (\u003cb\u003eSupplementary Fig.\u0026nbsp;5\u003c/b\u003e). The remaining worms were then filtered to remove outliers in body volume or total embryo count (sum of early- and late-stage embryos). Outliers were identified using Tukey fences (1.5 \u0026times; the interquartile range (IQR)) applied to each endpoint, and worms with measurements outside these fences were excluded. Inspection of excluded samples indicated that most were debris composed of shed worm cuticle and laid embryos clogging the channel. Mean phenotype scores were calculated per well and used to fit concentration-response curves from 3 biological replicates. One-way ANOVA was used to identify statistical differences between multiple well average values for different DMSO conditions.\u003c/p\u003e \u003cp\u003eTo estimate repeatability, the coefficient of variation (CV% = ratio of standard deviation to mean) was calculated for every possible combination of 3 wells drawn from every corresponding well position across 5 independent biological replicates (10 combinations per well position, 24 well positions). The mean CV% and standard error of mean (SEM) across all 10 combinations were reported for each endpoint. To measure the strengths of the difference between 1% DMSO and methylmercury-treated worms, we calculated the Strictly Standardized Mean Difference (SSMD) for all 6 endpoints using 3 biological replicates. From all methylmercury conditions, we selected the lowest concentration to serve as a positive control for our case study that produced the SSMD\u0026thinsp;\u0026ge;\u0026thinsp;2 for all the DART parameters, indicating strong effects [\u003cspan additionalcitationids=\"CR57 CR58\" citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eAssay power was estimated using the mean and standard deviations obtained from all 60 possible combinations (10 combinations per well position \u0026times; 6 well positions with 1% DMSO) for all endpoints from repeatability experiments. Power calculations were performed for a fixed value of \u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3 technical replicates and a series of hypothetical treatment group means representing 2.5% \u0026minus;\u0026thinsp;50% effect sizes below the control means (one-sample, one-tailed t test, α\u0026thinsp;=\u0026thinsp;0.05) using GraphPad Prism (version 10.6.1).\u003c/p\u003e \u003cp\u003eConcentration-response curves and effective concentration (EC\u003csub\u003e10\u003c/sub\u003e, EC\u003csub\u003e25\u003c/sub\u003e, and EC\u003csub\u003e50\u003c/sub\u003e) values were fitted using a 4-parameter logistic curve function using the \u0026ldquo;Find ECanything\u0026rdquo; nonlinear fit function of GraphPad Prism. The lower asymptote for each endpoint was constrained to the minimum biologically possible values of 0 embryos for Reproductive endpoints and estimated mean L1 body dimensions (length\u0026thinsp;=\u0026thinsp;150 \u0026micro;m, area\u0026thinsp;=\u0026thinsp;1,500 \u0026micro;m\u003csup\u003e2\u003c/sup\u003e, volume\u0026thinsp;=\u0026thinsp;11,775 \u0026micro;m\u003csup\u003e3\u003c/sup\u003e) for developmental endpoints. For each EC\u003csub\u003ex\u003c/sub\u003e estimate, the 95% confidence interval (CI) bands were calculated and plotted. Two concentration-response curves were compared using extra sum-of-square model to identify if the best fit values of EC\u003csub\u003ex\u003c/sub\u003e values and Hillslope differ between two data sets. The p-value was denoted by ns (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026ge;\u0026thinsp;0.5), * (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026le;\u0026thinsp;0.01), ** (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026le;\u0026thinsp;0.001), *** (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026le;\u0026thinsp;0.0001), and **** (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026le;\u0026thinsp;0.0001).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003e \u003cb\u003eA robust\u003c/b\u003e \u003cb\u003eC. elegans\u003c/b\u003e \u003cb\u003eassay to study developmental and reproductive toxicity (DART)-related endpoints\u003c/b\u003e\u003c/p\u003e \u003cp\u003eWe developed a DART assay that quantifies \u003cem\u003ein utero\u003c/em\u003e embryos within adult \u003cem\u003eC. elegans\u003c/em\u003e at distinct developmental stages (early and late stages) following chemical exposure. \u003cem\u003eC. elegans\u003c/em\u003e populations were exposed from the L1 stage for 72 hours and subsequently imaged using vivoChip devices to quantify embryo development relative to the 2-fold stage (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). Building on our previously established developmental toxicity (DevTox) platform, which uses ML-assisted image analysis [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e] to rapidly quantify developmental endpoints in \u003cem\u003eC. elegans\u003c/em\u003e, we extended this high-content imaging and analysis approach to capture reproductive phenotypes with high sensitivity. In addition to measuring worm body dimensions (length, area, volume) with high accuracy, we quantified the total number of embryos in each worm and classified each embryo by developmental stage (early- or late-stage embryo development). This approach of quantifying sub-lethal reproductive phenotypes enabled the determination of effective concentration (EC\u003csub\u003e50\u003c/sub\u003e) values with high confidence and low coefficient of variance (CV%).\u003c/p\u003e \u003cp\u003eThe whole-organism 3D imaging enabled by vivoChip microfluidic platform provides sufficient resolution to determine the developmental stage of \u003cem\u003ein utero\u003c/em\u003e embryos inside the intact \u003cem\u003eC. elegans\u003c/em\u003e uterus [\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. The vivoChip image-based assay yields 6 DART-related endpoints: body length, body area, body volume, total embryos, early-stage embryos, and late-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). Additionally, brightfield imaging of culture plates prior to chip loading was used to confirm worm viability and motility. By directly quantifying \u003cem\u003ein utero\u003c/em\u003e embryo number and classifying developmental progression during early adulthood, this assay captures sub-lethal reproductive toxicity while minimizing confounding effects from age-associated decline in the reproductive health and off-target effects such as altered egg-laying circuits. This DART platform facilitates rapid, high-content screening of chemicals and sensitive detection of sub-lethal developmental and reproductive perturbations.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eAutomated high-resolution, whole-body imaging to extract DART endpoints\u003c/h2\u003e \u003cp\u003eThe DART assay leverages the unique ability of vivoChip microfluidics platform to rapidly immobilize large numbers of \u003cem\u003eC. elegans\u003c/em\u003e for high-resolution imaging of their entire body. Each vivoChip-24x device enables parallel immobilization up to 960 \u003cem\u003eC. elegans\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA), and the closely spaced microfluidic channels maximize information density in each FOV (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eB). A pressurized gasket system seals the vivoChip-24x device and applies a pre-defined pressure sequence for loading and immobilization of worms efficiently (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC). Using automated stage navigation and objective switching, all wells are imaged sequentially, and complete volumetric datasets are acquired in ~\u0026thinsp;30 minutes per chip.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eHigh resolution 10\u0026times; 0.4 NA brightfield z-stacks (18 planes at 6 \u0026micro;m spacing) captured the full uterus of each immobilized worm, enabling visualization of \u003cem\u003ein utero\u003c/em\u003e embryos across developmental stages. This axial sampling accommodates minor variations in worm posture within the channels while preserving imaging speed. Prior to chip imaging, videos of the culture plates were collected to analyze worm motility (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eD).\u003c/p\u003e \u003cp\u003eEmbryos were classified into two developmentally distinct categories: early-stage embryos (\u0026le;\u0026thinsp;2-fold) and late-stage embryos (\u0026gt;\u0026thinsp;2-fold). The 2-fold stage provides a clear morphological boundary due to the rapid increase in structural complexity within the eggshell. Late-stage embryos often display fully formed larval structures, particularly in the central region of the uterus near the vulva, where embryos are frequently developed into young larvae. Visualization of individual larvae within the eggshell boundaries facilitates accurate classification. In some cases, movement within the eggshell was observable in time-lapse data (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eE\u0026ndash;F), further supporting developmental staging. In future implementations, embryo motility could be incorporated as an additional quantitative endpoint.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eImage analysis pipeline for robust phenotyping\u003c/h2\u003e \u003cp\u003eImaging data were first pre-processed by identifying bounding boxes around individual channels, enabling analysis of each worm (\u003cb\u003eSupplementary Fig.\u0026nbsp;6A\u003c/b\u003e). Downstream image analysis consisted of two separate processes; (1) automated quantification of worm body dimensions using a deep-learning-based segmentation (\u003cb\u003eSupplementary Figs.\u0026nbsp;6B-C\u003c/b\u003e) and (2) semi-automated embryonic phenotyping using a GUI-based analysis software to streamline manual scoring (\u003cb\u003eSupplementary Figs.\u0026nbsp;6D-F\u003c/b\u003e).\u003c/p\u003e \u003cp\u003ePreviously, we developed a 2.5D UNET architecture for image-level classification and pixel-wise segmentation tasks [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. Under this architecture, both tasks were jointly learned within a single network sharing a common encoder and bottleneck, trained using three z-slices around the best-focus image. While effective, this design introduced task coupling, limiting independent modification and optimization of the two objectives. To address this limitation, we decoupled classification and segmentation into separate models, each tailored to its respective task. The classification model comprises a convolutional encoder followed by a Vision Transformer (ViT) bottleneck and a classification head. The segmentation model consists of a convolutional encoder, a ViT-based bottleneck, and a convolutional decoder. Since image-level classification relies primarily on global morphological features, present in the best-focus image, we trained the classification model using a single best-focus image. In contrast, pixel-wise segmentation tasks rely on subtle changes in the image contrast to identify the object boundaries. To improve head and tail boundary detection, we fed a total of 5 z-stack images around the best-focus plane into the model. This approach captures features spanning multiple z-slices, improving boundary identification.\u003c/p\u003e \u003cp\u003eTo further improve robustness under challenging imaging conditions, the training datasets for both models were expanded to include additional channels with smaller worms, located near the exit portion of the immobilization channel from both vivoChip-24x-L4 and D1 devices (\u003cb\u003eSupplementary Figs.\u0026nbsp;7\u003c/b\u003e). For the classification model training, we generated ground truth data by classifying 5,305 channels as full, partial, or no worms class. For the segmentation model training, we used manually segmented masks from 4,360 channels with full worms. The classification model was evaluated on a test set of 538 images (441 full, 50 partial, and 47 channels with no worms) and achieved a weighted F1-score of 98.3%. The segmentation model was evaluated on a separate test set of 439 full worm images, achieving a mean Dice score of 98.5% (\u003cb\u003eSupplementary Fig.\u0026nbsp;8\u003c/b\u003e). High prediction accuracy for both channel classification and worm segmentation was observed for worms immobilized in D1 (\u003cb\u003eSupplementary Figs.\u0026nbsp;9A-D\u003c/b\u003e) and 4L (\u003cb\u003eSupplementary Figs.\u0026nbsp;9E-F\u003c/b\u003e) devices. From segmented full-worm channels, we calculated body length (defined as the longest skeleton length) and body area. Body volume was then estimated by integrating the segmented area by the known channel height at each pixel. For each worm, these three body parameters were calculated, and population-level metrics were obtained by averaging values across worms within each well.\u003c/p\u003e \u003cp\u003eEmbryo phenotyping is a challenging task, as each vivoChip-24x can immobilize up to 960 adult \u003cem\u003eC. elegans\u003c/em\u003e, corresponding to approximately 40,000 \u003cem\u003ein utero\u003c/em\u003e embryos requiring classification. Furthermore, the intertwined morphology of the uterus with the intestinal structures and gut granules requires 3D visualization of the worm uteri for accurate phenotyping. To facilitate robust embryo phenotyping, we developed a custom GUI-based software tool (vivoAnalyzer) that automatically displays each worm-containing channel in sequence, enables interactive zooming and z-stack navigation, and allows users to annotate the \u003cem\u003exyz\u003c/em\u003e centroid of each embryo while assigning early- or late-stage classification (\u003cb\u003eSupplementary Fig.\u0026nbsp;4\u003c/b\u003e). When necessary, the user can use time-lapse frames to confirm late-stage embryos, exhibiting larval movement. Scoring was performed by 10 trained users after following validation of their scoring accuracy on a reference dataset. Worms were randomly assigned to users, and treatment conditions were blinded to minimize scoring bias.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eRepeatability of the DART assay\u003c/h2\u003e \u003cp\u003eTo evaluate the robustness and statistical reliability of the DART assay in predicting toxicity levels of substances, we studied repeatability across multiple independent experiments using the same test substances. For an assay to replace or augment existing DART methods, it needs to be scalable, reproduceable by different scientists using different batches of worms and capable of providing repeatable results over time.\u003c/p\u003e \u003cp\u003eWe tested the repeatability of our DART assay by performing a series of 5 independent experiments (biological replicates) over two-month period using DMSO, a commonly used solvent in toxicity assays because of its wide applicability to a range of chemicals. Four DMSO concentrations (0.0, 0.2, 0.5, and 1.0%) were tested to characterize baseline phenotypes under solvent only control conditions. Each concentration was replicated across 6 wells (technical replicates) to assess well-to-well technical variability in addition to inter-experimental variability (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). We selected a 1% DMSO concentration as the solvent concentration because it allows us to dissolve the highest possible chemical concentrations near solubility limits while modestly increasing membrane permeability, which aids chemical uptake [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe worms developed into the D1 adult stage in all wells across all tested DMSO concentrations, including 1% DMSO (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB). To examine variability across both biological and technical replicates, we compared all 30 well averages (5 biological replicates \u0026times; 6 technical replicates) for all endpoints. For one representative biological replicate, the mean body parameters for 0% DMSO were 1,522\u0026thinsp;\u0026plusmn;\u0026thinsp;8 \u0026micro;m (length, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC and \u003cb\u003eSupplementary Fig.\u0026nbsp;10A\u003c/b\u003e), 75.9\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6 \u0026times; 10\u003csup\u003e3\u003c/sup\u003e \u0026micro;m\u003csup\u003e2\u003c/sup\u003e (area, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eD and \u003cb\u003eSupplementary Fig.\u0026nbsp;10B\u003c/b\u003e), and 2.96\u0026thinsp;\u0026plusmn;\u0026thinsp;0.06 \u0026times; 10\u003csup\u003e6\u003c/sup\u003e \u0026micro;m\u003csup\u003e3\u003c/sup\u003e (volume, Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eE and \u003cb\u003eSupplementary Fig.\u0026nbsp;10C\u003c/b\u003e). For 1% DMSO, the corresponding values were 1,596\u0026thinsp;\u0026plusmn;\u0026thinsp;3 \u0026micro;m (length, \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026lt;\u0026thinsp;0.0001), 74.2\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4 \u0026times; 10\u003csup\u003e3\u003c/sup\u003e \u0026micro;m\u003csup\u003e2\u003c/sup\u003e (area, \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;0.19), and 2.76\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02 \u0026times; 10\u003csup\u003e6\u003c/sup\u003e \u0026micro;m\u003csup\u003e3\u003c/sup\u003e (volume, \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;0.045). Embryo-related endpoints were similarly consistent. For 0% DMSO, worms exhibited 30.7\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6 early-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eF and \u003cb\u003eSupplementary Fig.\u0026nbsp;10D\u003c/b\u003e), 14.0\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5 late-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eG and \u003cb\u003eSupplementary Fig.\u0026nbsp;10E\u003c/b\u003e), and 44.7\u0026thinsp;\u0026plusmn;\u0026thinsp;0.2 total embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eH and \u003cb\u003eSupplementary Fig.\u0026nbsp;10F\u003c/b\u003e), whereas for 1% DMSO, averages were 29.2\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5 (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;0.23) early-stage embryos, 14.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.6 (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;0.86) late-stage embryos, and 43.7\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4 (\u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;0.24) total embryos. Although, there were significant differences between 0% and 1% DMSO conditions for body length and volume, there were no significant differences between these two populations for body area, total embryos, early-stage embryos, or late-stage embryos (\u003cb\u003eSupplementary Fig.\u0026nbsp;10\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eAlthough body length and volume showed statistically significant changes (+\u0026thinsp;4.9% and \u0026minus;\u0026thinsp;7.3%, respectively), the magnitudes of these differences were small. The well mean values obtained from all 30 wells from 5 biological replicates showed a similar trend among all 6 DART parameters (\u003cb\u003eSupplementary Fig.\u0026nbsp;11\u003c/b\u003e). DMSO concentrations\u0026thinsp;\u0026ge;\u0026thinsp;1% has previously been reported to affect worm internal structure and posture, potentially causing mild changes in the body dimensions [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. However, when comparing all 30 wells from the 5 biological replicates, we found no significant differences between 0% and 1% DMSO for early-stage, late-stage, and total embryo counts (\u003cb\u003eSupplementary Figs.\u0026nbsp;10D-F\u003c/b\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo test variability in the DART endpoints among technical replicates, we calculated coefficient of variation (CV%) between 6 technical replicates (Columns 1 to 6 in vivoChip-24x) for each condition. For 1% DMSO condition, the mean CV% values between the technical replicates were \u0026lt;\u0026thinsp;3.8% for developmental endpoints and \u0026lt;\u0026thinsp;12.5% for embryonic endpoints (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). This data suggests high reproducibility among worm populations loaded and imaged in the vivoChip, confirming the absence of edge effects or location-based bias.\u003c/p\u003e \u003cp\u003eTo assess inter-experimental repeatability, we calculated the coefficient of variation (CV%) between 3 corresponding wells from 3 separate experiments (e.g. exp\u003csub\u003e1\u003c/sub\u003e A01, exp\u003csub\u003e3\u003c/sub\u003e A01, exp\u003csub\u003e5\u003c/sub\u003e A01) and repeated this calculation for all 10 possible combinations of the 5 experiments (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eA-B). For the 0% DMSO populations, the mean CV% values across different experiments and 60 total combinations were 1.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1% for body length (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eC), 3.9\u0026thinsp;\u0026plusmn;\u0026thinsp;0.2% for body area (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eD), 7.8\u0026thinsp;\u0026plusmn;\u0026thinsp;0.4% for body volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eE), 11.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.8% for early-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eF), 17.1\u0026thinsp;\u0026plusmn;\u0026thinsp;0.9% for late-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eG), and 7.4\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5% for total embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eH). For the 1% DMSO populations, the mean CV% were 1.2\u0026thinsp;\u0026plusmn;\u0026thinsp;0.1% (length), 2.5\u0026thinsp;\u0026plusmn;\u0026thinsp;0.2% (area), 5.0\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3% (volume), 5.8\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3% (early-stage embryos), 16.7\u0026thinsp;\u0026plusmn;\u0026thinsp;1.1% (late-stage embryos), and 5.24\u0026thinsp;\u0026plusmn;\u0026thinsp;0.3% (total embryos). All values were substantially below the 30% threshold considered acceptable under OECD test guideline #222 for earthworms [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e], indicating high inter-experimental repeatability of the assay.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eStatistical power of DART assay parameters\u003c/h2\u003e \u003cp\u003eTo validate our DART assay design has sufficient statistical power of detecting small proportional changes relative to the control population using 3 biological replicates, we calculated the minimum detectable size effects with \u0026ge;\u0026thinsp;80% power values. To estimate the power, we used the aggregated means and standard deviations from 60 possible combinations (10 combinations per well locations and 6 wells) and for all parameters measured in the 1% DMSO wells, that will serve as assay controls. With 3 experimental replicates, our assay achieved\u0026thinsp;\u0026gt;\u0026thinsp;80% power to detect changes of \u0026ge;\u0026thinsp;5.5% in developmental endpoints (body length, area, and volume) and \u0026ge;\u0026thinsp;7% in total and early-stage embryos. For late-stage embryos, the minimum detectable effect size at the 80% power threshold was 23% \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e, though increasing biological replicates would enable detection of lower reductions.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eDART assay with methylmercury\u003c/h2\u003e \u003cp\u003eWe next evaluated the assay using a well-studied toxicant, methylmercury, which has documented toxicity in \u003cem\u003eC. elegans\u003c/em\u003e [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e] and mammalian models [\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. The purpose of testing this chemical was to assess repeatability and precision in concentration-response studies, and to find the sensitivity of each endpoint. We ran 3 biological replicates spanning 20 concentrations of methylmercury between 0.0007 and 10 \u0026micro;M using the vivoChip-24x-D1 device.\u003c/p\u003e \u003cp\u003eRange-finding assays based on plate imaging showed that concentrations\u0026thinsp;\u0026gt;\u0026thinsp;10 \u0026micro;M caused developmental arrest between the L1 and L3 stages, below the minimum size trappable in the D1 or L4 devices. Consistently, worms exposed to \u0026ge;\u0026thinsp;8 \u0026micro;M developed slowly, remained approximately at the L4-stage, and slipped through the D1 chip channels due to their small size (\u003cb\u003eSupplementary Fig.\u0026nbsp;12\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eTo enable imaging at higher concentrations, we ran a second set of experiments using the L4 devices, which can trap smaller worms down to early L4 stage. For the full concentration-dependent DART study, we used data from the D1 chip for lower concentrations and only used the data from L4 chips for the highest concentrations to complete concentration-response curves. Inclusion of the L4 chip data captured the 50% effect region for developmental endpoints and up to 100% effect region for embryo endpoints.\u003c/p\u003e \u003cp\u003eThe combined results obtained from the D1 and L4 devices were fit with a 4-parameter logistic curve using the well averages after outlier filtering, and EC\u003csub\u003e10\u003c/sub\u003e, EC\u003csub\u003e25\u003c/sub\u003e, and EC\u003csub\u003e50\u003c/sub\u003e values were estimated for each endpoint \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e \u003cb\u003eand Supplementary Table\u0026nbsp;2)\u003c/b\u003e. For developmental endpoints, EC\u003csub\u003e50\u003c/sub\u003e values (95% CI) were 8.22 \u0026micro;M (7.83\u0026ndash;8.69) for body length (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eA), 5.33 \u0026micro;M (4.98\u0026ndash;5.72) for body area (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eB), and 4.61 \u0026micro;M (4.28\u0026ndash;4.99) for body volume (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eC), indicating that body volume is the most sensitive developmental metric. For the reproductive parameters, EC\u003csub\u003e50\u003c/sub\u003e values (95% CI) were 3.90 \u0026micro;M (3.66\u0026ndash;4.16) for early-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eD), 1.49 \u0026micro;M (1.33\u0026ndash;1.67) for late-stage embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eE), and 3.14 \u0026micro;M (2.90\u0026ndash;3.38) for total embryos (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eF), indicating that embryo classification associated with different stages of development provides additional sensitivity and that late-stage embryos represent the most sensitive reproductive phenotype. Representative worm images are shown in \u003cb\u003eSupplementary Fig.\u0026nbsp;13\u003c/b\u003e. Inter-experimental repeatability determined from the 1% DMSO control wells was high, with CV% values of 1.2% (length), 8.3% (area), 12.9% (volume), 8.8% (early-stage embryos), 17.1% (late-stage embryos), and 10.2% (total embryos), all well below the 30% threshold. In addition, the EC\u003csub\u003e50\u003c/sub\u003e confidence intervals were narrow (\u0026thinsp;\u0026lt;\u0026thinsp;\u0026plusmn;\u0026thinsp;0.5 \u0026micro;M), further supporting good assay precision.\u003c/p\u003e \u003cp\u003eTo study the effect of reducing the number of treatment conditions on EC\u003csub\u003e50\u003c/sub\u003e values, we omitted every other methylmercury condition and reanalyzed the DART data from 10 out of the 20 conditions. We fitted the reduced 10-point dataset with a 4-parameter logistic curve and re-estimated EC\u003csub\u003e10\u003c/sub\u003e, EC\u003csub\u003e25\u003c/sub\u003e, and EC\u003csub\u003e50\u003c/sub\u003e values (\u003cb\u003eSupplementary Table\u0026nbsp;3\u003c/b\u003e). The resulting EC\u003csub\u003e50\u003c/sub\u003e values were not significantly different from those obtained using the full dataset for the most sensitive developmental endpoint (body volume; EC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;4.51 \u0026micro;M; 95% CI: 4.03\u0026ndash;5.08, \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;0.98, \u003cb\u003eSupplementary Fig.\u0026nbsp;14A\u003c/b\u003e) and the most sensitive reproductive endpoint (late-stage embryos; EC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;1.51 \u0026micro;M; 95% CI: 1.26\u0026ndash;1.80, \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;=\u0026thinsp;1.00, \u003cb\u003eSupplementary Fig.\u0026nbsp;14B\u003c/b\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo identify a suitable concentration of methylmercury as a possible positive control for our DART assay, we calculated the SSMD for all 20 concentrations tested and all 6 DART parameters analyzed \u003cb\u003e(Supplementary Table\u0026nbsp;4)\u003c/b\u003e. SSMD provides a quantitative measure of assay quality by evaluating the separation between control and treated populations relative to variability. For 4 \u0026micro;M methylmercury, which produced a clear reduction in both developmental and reproductive parameters (\u003cb\u003eSupplementary Fig.\u0026nbsp;13\u003c/b\u003e), SSMD values were: 5.18 (length), 3.24 (area), 2.77 (volume), 2.96 (late-stage embryos), 5.84 (early-stage embryos), and (total embryos). According to established criteria, SSMD values\u0026thinsp;\u0026ge;\u0026thinsp;3 indicate excellent assay quality, and values between 2 and 3 indicate good assay quality. Therefore, 4 \u0026micro;M methylmercury demonstrated good-to-excellent performance across all endpoints and was selected as a positive assay control concentration for the subsequent case study.\u003c/p\u003e \u003cp\u003eBased on the methylmercury concentration-response data, we further refined our DART assay design to test new chemicals at 10 concentrations per curve. This approach captures concentration-dependent effects across endpoints whose EC\u003csub\u003e50\u003c/sub\u003e values may differ by up to 10 folds. Inclusion of sufficient concentration points is particularly important for endpoints with steep slopes, such as for late-stage embryos, the most sensitive parameter of our assay, to estimate precise EC\u003csub\u003e50\u003c/sub\u003e values with narrow 95% CIs.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eCase study: DART assessment of an agrichemical\u003c/h2\u003e \u003cp\u003eFollowing demonstration of assay precision and repeatability, we performed a concentration-response DART study using the triazole fungicide propiconazole [\u003cspan additionalcitationids=\"CR67\" citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e], a chemical of interest to the agrichemical industry and ecotoxicology. First, we performed a preliminary range-finding assay by treating synchronized L1-stage N2 worms with 22 different concentrations of propiconazole ranging between 0.02-1,730 \u0026micro;M following our standard protocols. Concentrations\u0026thinsp;\u0026gt;\u0026thinsp;334 \u0026micro;M were lethal (\u003cb\u003eSupplementary Fig.\u0026nbsp;15\u003c/b\u003e).\u003c/p\u003e \u003cp\u003eBased on these results, we conducted a full concentration\u0026ndash;response study with 10 propiconazole concentrations (0.08\u0026ndash;334 \u0026micro;M), along with the solvent control (1% DMSO), and the positive control (4 \u0026micro;M methylmercury). Using 10 test concentrations and 2 controls (12 wells in total), concentration-response curves from 2 different chemicals can be tested on a single vivoChip-24x. At the highest propiconazole concentration (334 \u0026micro;M), worms exhibited delayed growth and were immobilized using the L4 chip.\u003c/p\u003e \u003cp\u003eSimilar to methylmercury, propiconazole produced distinct concentration-dependent effects (EC\u003csub\u003e50\u003c/sub\u003e responses) across developmental (Figs.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eA-C) and reproductive (Figs.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eD-F) endpoints. Among developmental parameters, body volume exhibited a greater sensitivity (EC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;336 \u0026micro;M; 95% CI: 311\u0026ndash;369; Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eC) compared to body length (EC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;504 \u0026micro;M; 95% CI: 392\u0026ndash;685; Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eA). For reproductive endpoints, late-stage embryos were the most sensitive metric (EC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;78 \u0026micro;M; 95% CI: 35\u0026ndash;154; Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eE) with EC\u003csub\u003e50\u003c/sub\u003e less than half that of total embryos (EC\u003csub\u003e50\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;204 \u0026micro;M; 95% CI: 177\u0026ndash;237; Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eD). These results further support late-stage embryo phenotype as the most sensitive DART endpoint in this assay. The EC\u003csub\u003e10\u003c/sub\u003e, EC\u003csub\u003e25\u003c/sub\u003e, and EC\u003csub\u003e50\u003c/sub\u003e values for 6 DART parameters are listed in the \u003cb\u003eSupplementary Table\u0026nbsp;4\u003c/b\u003e, and representative worm images are shown in \u003cb\u003eSupplementary Fig.\u0026nbsp;16\u003c/b\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eComparative endpoint sensitivity analysis for DART assay\u003c/h2\u003e \u003cp\u003eTo further investigate the relative sensitivity and utility of different DART endpoints, we compared concentration-response curves normalized to baseline (no-effect) values for a widely used developmental parameter (worm body length), and total and late-stage embryo counts, for both methylmercury \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eA\u003cb\u003e)\u003c/b\u003e and propiconazole \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eB\u003cb\u003e)\u003c/b\u003e. In addition, we quantified the percentage of worms that were alive, as identified to be moving in culture well plate images captured using 2\u0026times; objective.\u003c/p\u003e \u003cp\u003eAll worms remained alive at concentrations up to at least the highest concentration tested in the vivoChip concentration-response experiments. For example, 100% of worms in the plates were alive at 334 \u0026micro;M propiconazole and \u0026gt;\u0026thinsp;97% alive at 10 \u0026micro;M methylmercury, despite clear developmental impairment. These findings suggest that the observed responses were specifically attributable to DART-specific effects rather than lethality, an apical endpoint studied in \u003cem\u003eC. elegans\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eWe calculated EC\u003csub\u003e10\u003c/sub\u003e values for body length, total embryo, and late-stage embryo parameters for both methylmercury and propiconazole using the normalized concentration-response data (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e). The EC\u003csub\u003e10\u003c/sub\u003e estimates, representing the onset of toxicity for each endpoint, were widely separated and followed the order: late-stage embryos EC\u003csub\u003e10\u003c/sub\u003e\u0026thinsp;\u0026lt;\u0026thinsp;total embryos EC\u003csub\u003e10\u003c/sub\u003e\u0026thinsp;\u0026lt;\u0026thinsp;body length EC\u003csub\u003e10\u003c/sub\u003e\u0026thinsp;\u0026lt;\u0026thinsp;lethality onset. For methylmercury, the EC\u003csub\u003e10\u003c/sub\u003e value for body length was 3.62 \u0026micro;M (95% CI: 3.33\u0026ndash;0.92), total embryo was 1.94 \u0026micro;M (95% CI: 1.71\u0026ndash;2.19), and late-stage embryo was 1.08 \u0026micro;M (95% CI: 0.92\u0026ndash;1.24). For propiconazole, the EC\u003csub\u003e10\u003c/sub\u003e value for body length was 256 \u0026micro;M (95% CI: 230\u0026ndash;285), total embryo was 57 \u0026micro;M (95% CI: 43\u0026ndash;73), and late-stage embryo was 14 \u0026micro;M (95% CI: 7\u0026ndash;24). Notably, the EC\u003csub\u003e10\u003c/sub\u003e for late-stage embryos preceded any measurable response in other endpoints, identifying late-stage embryo count as the most sensitive parameter.\u003c/p\u003e \u003cp\u003eFor methylmercury and propiconazole, body length EC\u003csub\u003e10\u003c/sub\u003e (3.62 \u0026micro;M and 256 \u0026micro;M), a commonly used developmental endpoint, was significantly different than the most sensitive late-stage embryo EC\u003csub\u003e10\u003c/sub\u003e values (1.08 \u0026micro;M; \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001 and 14 \u0026micro;M; \u003cem\u003ep\u003c/em\u003e-value\u0026thinsp;\u0026lt;\u0026thinsp;0.001) and were 3.4\u0026times; and 18.3\u0026times; higher, respectively. Viability remained high across the tested concentration ranges, confirming that the observed changes in each DART parameter represent bona fide developmental and reproductive effects rather than secondary consequences of lethality. Together, these findings demonstrate that the multiparametric DART assay provides independent and biologically meaningful assessments of developmental and reproductive endpoints without confounding effects from reduced viability.\u003c/p\u003e \u003cp\u003eThis significantly lower toxicity threshold and large ratio of developmental and embryo parameters, especially for propiconazole, demonstrates that our multiparametric DART approach can detect reproductive toxicity independent of DevTox and viability. Also, the DevTox parameters are detected independent of viability. Such multiparametric analysis helps us to identify the adverse effects onset with potential benefits for prioritizing safety assessments of substances.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eDART toxicity studies are required for hazard assessment of thousands of new and existing substances. However, high costs, long study durations, and a global push to reduce or eliminate mammalian models in routine testing present major bottlenecks. Here, we introduce a novel imaging-based DART assay using the model organism, \u003cem\u003eC. elegans\u003c/em\u003e, which is faster, more cost-effective, and eliminates the need for vertebrate animals subject to welfare regulation.\u003c/p\u003e \u003cp\u003eLike human females, \u003cem\u003eC. elegans\u003c/em\u003e has a complete reproductive system with conserved molecular pathways and several hallmarks of reproductive aging [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. In certain toxicological domains, \u003cem\u003eC. elegans\u003c/em\u003e have been shown to be as predictive as rodent models [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e], supporting the translational relevance of the data obtained with this system for human reproductive toxicity assessment. While \u003cem\u003eC. elegans\u003c/em\u003e has been used as a model organism in DART studies, previous approaches were limited by low throughput, labor-intensive methodologies, or reliance on relatively gross phenotypes such as brood size or body length [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. In this work, we developed a high-content, multiparametric imaging platform capable of detecting subtle multi-parametric DART phenotypes. This approach increases assay sensitivity to low-level exposures while improving throughputs, thereby addressing key limitations of existing DART testing paradigms.\u003c/p\u003e \u003cp\u003eWe previously developed an ML-based imaging platform to assess DevTox using \u003cem\u003eC. elegans\u003c/em\u003e as an alternative model organism [\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. This study expands upon that work by incorporating reproductive toxicity-related endpoints to assess multiple modes of action within a unified DART framework. We leveraged the vivoChip platform, which enables high-resolution, high-throughput imaging of 24 independent worm populations, with 40 worms per population (totaling 960 worms per vivoChip), while generating high-information-density datasets suitable for quantitative assessment of \u003cem\u003ein utero\u003c/em\u003e embryo production and development progression. Two different vivoChip-24x designs, each optimized for trapping worms of different sizes, allowed us to analyze DART endpoint changes ranging from delayed L4s to normally developed adults. Integrating data obtained from both device configurations enabled comprehensive coverage across different sections of the concentration-response curve. This capability is particularly essential for multi-parametric assays, in which EC\u003csub\u003e50\u003c/sub\u003e values for distinct endpoints may be widely separated, requiring adequate sampling across multiple exposure ranges to accurately resolve differential sensitivities.\u003c/p\u003e \u003cp\u003eFor our DART assay, we exposed age-synchronized L1s for 72 hours and performed terminal imaging of \u003cem\u003eC. elegans\u003c/em\u003e populations using the vivoChip platform. Worms were treated with multiple conditions, including solvent controls, in which the worms developed to the D1 adult stage during the exposure period. Because toxic chemical exposures can cause slow and asynchronous development, especially at high concentrations, it was necessary to start with precisely age-synchronized worms. Reproductive organ development in \u003cem\u003eC. elegans\u003c/em\u003e spans from L1 to early adulthood. This process includes cell division of the two somatic gonad precursors and two primordial germ cells at the L1 stage, reorganization of the somatic gonad primordium during late L2, and formation of the sheath, spermatheca, and uterus during the L3 and L4 stages [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e]. A 72-hour chronic exposure starting at the L1 stage maximizes the likelihood that chemicals reach the germline during critical windows of development. Such exposure may interfere with key biological events, such as mitotic proliferation, meiotic cell cycles, spermatogenesis, oogenesis, oocyte maturation, ovulation, and the entire course of \u003cem\u003ein utero\u003c/em\u003e embryogenesis up to the start of neuromuscular movement within the egg [\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e, \u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. Imaging all treatment groups in an experiment within a narrow time interval further ensured accurate comparisons across concentrations and minimized variability introduced by developmental timing differences.\u003c/p\u003e \u003cp\u003eTo capture adverse effects on reproduction, we identified and quantified all embryos within the uterus of individual \u003cem\u003eC. elegans\u003c/em\u003e hermaphrodites and classified them as early- or late-stage embryos based on distinct morphological features, using the 2-fold stage as the classification threshold. We also calculated total embryo counts by adding early- and late-stage embryos for each worm. While the total embryo count serves as an \u003cem\u003ein utero\u003c/em\u003e proxy for the traditional brood size metric, the late-stage embryo counts provide an indicator of successful embryonic development, correlating with the hatching efficiency measurements. Compared to traditional reproductive toxicity assays, our assay offers several advantages, including (1) scalability enabled by well-plate-compatible liquid culture, (2) simultaneous testing of 24 independent populations, (3) high-resolution imaging of up to 40 \u003cem\u003eC. elegans\u003c/em\u003e per population to quantify \u003cem\u003ein utero\u003c/em\u003e embryonic development, and (4) single time point imaging, which reduces assay complexity while preserving multiparametric information content.\u003c/p\u003e \u003cp\u003eTraditional brood size assays performed using NGM plates are labor-intensive and low-throughput [\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Although microfluidic-based automated egg counting systems have been developed [\u003cspan additionalcitationids=\"CR50\" citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e], these platforms are technically complex, require continuous imaging over 3\u0026ndash;6 days, and consume large amounts of test substances due to constant perfusion. In contrast, our DART assay quantifies and classifies all \u003cem\u003ein utero\u003c/em\u003e embryos within hermaphrodites on the first day of adulthood, when embryo production enters the peak phase. Assessing DART endpoints at this early adult stage offers several key advantages: (1) it enables scoring of high-quality embryos while avoiding confounding effects from age-related gonadal atrophy; (2) embryos are typically aligned in a single row, reducing phenotyping errors that arise from multilayered stacking in older worms; (3) the readout is independent of the egg-laying circuit, minimizing data drift caused by off-target neuromuscular effects; and (4) the assay is based on direct morphological features of embryogenesis rather than late-stage neuromuscular strength, which influences larval hatching outcomes. In addition, our DART assay relies on image analysis and scoring of brightfield images of wild-type worms and does not require any reporter strain or fluorescent biomarkers. This aspect facilitates application of our DART assay to large numbers of wild isolates with diverse genetic backgrounds, enabling characterization of genetic susceptibility to environmental toxicants using naturally occurring variation in \u003cem\u003eC. elegans\u003c/em\u003e populations [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFor any newly developed method, intra-laboratory repeatability (and ultimately inter-laboratory reproducibility) is vital for consideration by regulatory bodies [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e]. To address repeatability, we established robust and detailed protocols for \u003cem\u003eC. elegans\u003c/em\u003e husbandry, synchronization, chemical exposure, and culture conditions to minimize inter-experimental variability. Our DART assay is highly data-intensive, generating\u0026thinsp;~\u0026thinsp;11,000 images and nearly\u0026thinsp;~\u0026thinsp;325 GB of data in less than 30 minutes from a single experiment. To enable even higher throughput, we developed a minimal data acquisition protocol to reduce data volume while preserving the ability to produce all the DART data presented in this study. This optimized approach, suitable for large-scale studies, collects 2,184 images (~\u0026thinsp;66 GB) from a single vivoChip by limiting imaging data from a single timepoint, 18 z-slice brightfield images. To manage such large volumes of data and support multiparametric phenotypic analysis, we developed the vivoScreen platform, which integrates reliable hardware components with rigorously validated and tested software, a centralized database infrastructure, and both local and cloud-based data storage systems. Every experiment is archived with comprehensive metadata, including experimental details, imaging conditions, analysis outputs, and final reports, ensuring full traceability. Despite the large data volume, image analysis is performed efficiently using an ML-based pipeline for DevTox parameters with an accuracy that is comparable to human scorers while requiring only a fraction of the time. Embryo phenotypes are then scored using semi-automated, user-friendly software designed to streamline manual embryo annotation and improve scoring consistency.\u003c/p\u003e \u003cp\u003eThe mean coefficient of variation (CV%) for 1% DMSO ranged from 1\u0026ndash;5% for developmental endpoints and 6\u0026ndash;17% for embryo-related endpoints, values well below thresholds considered acceptable for regulatory-approved assays, including those described for earthworms [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. This low variability confers high statistical power, even with relatively few experimental replicates. Power analysis indicates that 3 independent experimental replicates are sufficient to detect\u0026thinsp;\u0026gt;\u0026thinsp;23% changes across all endpoints, while smaller effect sizes can be reliably identified with additional replicates. While the late-stage embryo phenotype exhibited the highest variability, it was also the most sensitive endpoint in terms of EC\u003csub\u003e50\u003c/sub\u003e. The higher variation likely reflects the oscillatory pattern of egg laying, as embryos are laid in clusters every 1\u0026ndash;2 hours [\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]. Depending on timing, worms may be immobilized and imaged immediately before or after an egg-laying event, introducing natural biological variability into embryo counts. In addition, we confirmed that 1% DMSO, used as the solvent control, does not contribute any significant adverse effects on embryo parameters, enabling testing at relatively high treatment concentrations of many chemicals [\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e, \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eWe validated the DART assay using two reference chemicals: methylmercury and propiconazole. Methylmercury was chosen because of its extensive evaluation across variety of species and toxicology domains, including DART, neurotoxicity, metabolism, gene expression, stress response, and cell division [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan additionalcitationids=\"CR74 CR75\" citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e]. Propiconazole was chosen as it is a widely used fungicide with well-characterized toxicology profiles in mammals and other model organisms [\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e, \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e, \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e]. Both chemicals showed concentration-dependent effects on all 6 endpoints studied within our DART assay. These endpoints exhibited differential sensitivities based on EC\u003csub\u003e50\u003c/sub\u003e values, with body volume emerging as the most sensitive developmental parameter and late-stage embryos as the most sensitive reproductive parameter. We observed narrow 95% confidence interval bands and low CV% values for the methylmercury control wells, indicating high assay precision. The assay quality, measured using SSMD, was good for \u0026ge;\u0026thinsp;4 \u0026micro;M methylmercury, supporting its suitability as a positive assay control for all endpoints.\u003c/p\u003e \u003cp\u003eThe onset of DART toxicity for methylmercury in our assays, based on the late-stage embryo endpoint, occurred at ~\u0026thinsp;1 \u0026micro;M, consistent with previously reported \u003cem\u003eC. elegans\u003c/em\u003e DART studies [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. For propiconazole, adverse effects on fathead minnow fecundity and \u003cem\u003eDaphnia\u003c/em\u003e embryonic development have been reported at 1.46 \u0026micro;M (0.5 mg/L), which is within an order of magnitude of our EC\u003csub\u003e10\u003c/sub\u003e value of 14 \u0026micro;M [\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e, \u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e]. A recent study in zebrafish reported a DART EC\u003csub\u003e50\u003c/sub\u003e of a 1.62 mg/L, (~\u0026thinsp;2 \u0026micro;M propiconazole equivalent) for a commercial formulation, and additional studies have observed developmental abnormalities at 0.73-1 \u0026micro;M [\u003cspan additionalcitationids=\"CR80\" citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e]. These comparisons across different species could be due to differences in chemical uptake or accumulation, which can be addressed by measuring tissue-level concentration using analytical methods. In \u003cem\u003eC. elegans\u003c/em\u003e, a developmental toxicity assay and a brood-size-based reproductive toxicity assay reported an EC\u003csub\u003e50\u003c/sub\u003e of 223\u0026ndash;249 \u0026micro;M and 141 \u0026micro;M, respectively [\u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e, \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e]. Using the most sensitive late-stage embryos parameter in our assay, these previously reported values are ~\u0026thinsp;1.8\u0026ndash;3.2\u0026times; higher than the EC\u003csub\u003e50\u003c/sub\u003e value of 78 \u0026micro;M observed in our assay. These differences could be associated with the differences in the food source, DMSO concentration, exposure duration, and endpoint definition. While absolute effective concentrations of propiconazole in our results are higher than those reported in vertebrate systems, the \u003cem\u003eC. elegans\u003c/em\u003e-based DART platform offers substantial advantages in cost, scalability, and throughput. Compared with rodent colonies and fish aquaria, this approach requires significantly lower resources and eliminates labor-intensive and subjective phenotyping of complex reproductive phenotypes on an animal-by-animal basis. Moreover, imaging-based multiparametric design enables objective and quantitative assessment of developmental and reproductive phenotypes.\u003c/p\u003e \u003cp\u003eAll endpoints studied in our assay are biologically interlinked, and reproductive (embryo) phenotypes may, in part, be affected by developmental perturbations. However, the effective concentrations for the different endpoints were clearly separated, suggesting distinct sensitivities rather than a uniform secondary response. Total embryo numbers declined at lower concentrations than those required to reduce body size, indicating that the germline is specifically affected and that the observed reproductive effects are not simply a consequence of delayed general development. Moreover, reductions in late-stage embryos occurred before a decrease in total embryo counts, consistent with direct impairment of embryonic development rather than a simple reduction in embryo production. Importantly, the worms remained\u0026thinsp;\u0026gt;\u0026thinsp;97% viable and motile across all tested concentrations. Therefore, the measured effects reflect DART responses rather than the apical endpoints such as lethality.\u003c/p\u003e \u003cp\u003eThe novel DART platform we have developed is both faster and more cost-effective than traditional vertebrate-based assays. Thanks to the short life cycle of \u003cem\u003eC. elegans\u003c/em\u003e, culture and treatment durations are measured in days, while chip loading and imaging times are \u0026lt;\u0026thinsp;30 min per experiment. In addition to improved efficiency, the assay eliminates the need for vertebrate animals subject to welfare regulations. Data analysis can also be performed rapidly. The ML inference pipeline used for body-dimension quantification can process a single chip (~\u0026thinsp;1,000 \u003cem\u003eC. elegans\u003c/em\u003e) in \u0026lt;\u0026thinsp;5 minutes using standard desktop PCs. Comparable processing times are expected for automated embryo detection once fully optimized ML-based models are implemented. Across these studies, we collected and annotated data from over ~\u0026thinsp;389,000 embryos across ~\u0026thinsp;9,200 worms under various treatment conditions, including detailed classification of developmental stage and extraction of \u003cem\u003exyz\u003c/em\u003e centroid coordinates within image stacks. This extensive dataset provides a strong foundation for training next-generation ML models to fully automate embryo detection and staging. We are currently developing new ML models that will enable fully automated phenotyping of all DART endpoints within very short processing times, eliminating the current time-intensive manual analysis and substantially increasing scalability.\u003c/p\u003e \u003cp\u003eThis study demonstrates the sensitivity, repeatability, and precision of our DART assay. The implementation of automated embryo scoring will further enable high-throughput screening, allowing rapid and cost-effective testing of large chemical libraries for hazard prioritization and comparative safety assessment within defined chemical spaces.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank Thao Vu, Laura Murphy, Kyra Kwek, Jennifer Nguyen, Nishi Patel, Ruchi Birur, Cale Heiniger, and Seher Mamun for their help with manual scoring of the brightfield images of worms for \u003cem\u003ein utero\u003c/em\u003e embryo scoring. We also thank Nikko Chavez and Dr. Jacob Moore for the initial version of the image acquisition hardware and software development. We thank The Caenorhabditis Natural Diversity Resource (CaeNDR) for providing the N2 strain.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u0026nbsp;\u003c/strong\u003eA.L., A.S., G.C., and J.B. maintained the worms, setup the 24-well liquid cultures. A.M. developed automated image analysis for body parameters. S.G. developed the image acquisition and graphical user interface (GUI) for embryo scoring. E.H., A.L., and J.B. ran microfluidic device experiments. A.L., J.B., and S.M. analyzed the data, prepared the plots, and performed the statistics. S.M. and A.B. conceptualized the projects and planned the experiments. A.L., S.M., and A.B. prepared the manuscript. A.L., S.M., and A.B. finalized the discussion of the results.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e This study was funded by the National Institutes of Health (NIH) and United States Department of Agriculture, National Institute of Food and Agriculture (USDA, NIFA), through Small Business Innovation Research grants from NCCIH (43AT013187), NIEHS (R43ES036387, R43ES033579, and R44ES033579), NIMH (R43MH118841 and R44MH118841), and USDA (2024-33530-41929).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability:\u003c/strong\u003e The datasets generated and/or analyzed during the current study are available from the corresponding authors upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests statement:\u003c/strong\u003e A.B., S.M., and E.H. are co-founders of vivoVerse, LLC and are the inventors of the vivoChip microfluidic technology and vivoScreen platform. \u0026nbsp;A.B., S.M., A.L., A.S., and E.H. are inventors on several approved and ongoing patent applications.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eOecd \u003cem\u003eTest No. 422: Combined Repeated Dose Toxicity Study with the Reproduction/Developmental Toxicity Screening Test. OECD Guidelines for the Testing of Chemicals, Section 4\u003c/em\u003e (OECD Publishing, 2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOecd \u003cem\u003eTest No. 421: Reproduction/Developmental Toxicity Screening Test.\u003c/em\u003e OECD Guidelines for the Testing of Chemicals, Section 4, (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOlson, H. et al. Concordance of the toxicity of pharmaceuticals in humans and in animals. \u003cem\u003eRegul. Toxicol. Pharmacol.\u003c/em\u003e \u003cb\u003e32\u003c/b\u003e, 56\u0026ndash;67 (2000).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, Z. et al. \u003cem\u003eToward a Global Understanding of Chemical Pollution: A First Comprehensive Analysis of National and Regional Chemical Inventories\u003c/em\u003e Vol. 54, p. 2575\u0026ndash;2584 (Environmental Science \u0026amp; Technology, 2020). 5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchenk, B. et al. The ReProTect Feasibility Study, a novel comprehensive in vitro approach to detect reproductive toxicants. \u003cem\u003eReprod. Toxicol.\u003c/em\u003e \u003cb\u003e30\u003c/b\u003e (1), 200\u0026ndash;218 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRajagopal, R. et al. Beyond AOPs: A Mechanistic Evaluation of NAMs in DART Testing. \u003cem\u003eFront. Toxicol.\u003c/em\u003e \u003cb\u003e4\u003c/b\u003e, 838466\u0026ndash;838466 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiersma, A. H. et al. Evaluation of an alternative in vitro test battery for detecting reproductive toxicants. \u003cem\u003eReprod. Toxicol.\u003c/em\u003e \u003cb\u003e38\u003c/b\u003e, 53\u0026ndash;64 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOecd, T. \u003cem\u003eNo. 455: Performance-Based Test Guideline for Stably Transfected Transactivation In Vitro Assays to Detect Estrogen Receptor Agonists and Antagonists.\u003c/em\u003e OECD Guidelines for the Testing of Chemicals, Section 4, (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSychrov\u0026aacute;, E. et al. In vitro testicular toxicity of environmentally relevant endocrine-disrupting chemicals: 2D vs. 3D models of prepubertal Leydig TM3 cells. \u003cem\u003eEnviron. Toxicol. Pharmacol.\u003c/em\u003e \u003cb\u003e93\u003c/b\u003e, 103869\u0026ndash;103869 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eH\u0026ouml;ss, S. et al. Interlaboratory comparison of a standardized toxicity test using the nematode Caenorhabditis elegans (ISO 10872). \u003cem\u003eEnviron. Toxicol. Chem.\u003c/em\u003e \u003cb\u003e31\u003c/b\u003e (7), 1525\u0026ndash;1535 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBehl, M. et al. Comparative Toxicity of Organophosphate Flame Retardants and Polybrominated Diphenyl Ethers to Caenorhabditis elegans. \u003cem\u003eToxicol. Sci.\u003c/em\u003e \u003cb\u003e154\u003c/b\u003e (2), 241\u0026ndash;252 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd, W. A. et al. A high-throughput method for assessing chemical toxicity using a Caenorhabditis elegans reproduction assay. \u003cem\u003eToxicol. Appl. Pharmacol.\u003c/em\u003e \u003cb\u003e245\u003c/b\u003e (2), 153\u0026ndash;159 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd, W. A. et al. Developmental Effects of the ToxCast Phase I and Phase II Chemicals in Caenorhabditis elegans and Corresponding Responses in Zebrafish, Rats, and Rabbits. \u003cem\u003eEnviron. Health Perspect.\u003c/em\u003e \u003cb\u003e124\u003c/b\u003e, 586\u0026ndash;593 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarlow, P. H. et al. Comparative metabolism of xenobiotic chemicals by cytochrome P450s in the nematode Caenorhabditis elegans. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 13333 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHarlow, P. H. et al. \u003cem\u003eThe nematode Caenorhabditis elegans as a tool to predict chemical activity on mammalian development and identify mechanisms influencing toxicological outcome OPEN.\u003c/em\u003e (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCamacho, J. A. et al. \u003cem\u003eReproductive-Toxicity-Related Endpoints in C. elegans Are Consistent with Reduced Concern for Dimethylarsinic Acid Exposure Relative to Inorganic Arsenic\u003c/em\u003e. \u003cem\u003eJ. Dev. Biol.\u003c/em\u003e, \u003cb\u003e11\u003c/b\u003e(2). (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHunt, P. R. et al. \u003cem\u003eDevelopmental Exposures to Three Mammalian Teratogens Produce Dysmorphic Phenotypes in Adult Caenorhabditis elegans.\u003c/em\u003e Toxics 2025, Vol. 13, 13(7). (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHunt, P. R., Olejnik, N. \u0026amp; Sprando, R. L. Toxicity ranking of heavy metals with screening method using adult Caenorhabditis elegans and propidium iodide replicates toxicity ranking in rat. \u003cem\u003eFood Chem. Toxicol.\u003c/em\u003e \u003cb\u003e50\u003c/b\u003e (9), 3280\u0026ndash;3290 (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen, P. et al. Metal-induced neurodegeneration in C. elegans. \u003cem\u003eFront. Aging Neurosci.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, 18 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHelmcke, K. J. et al. Characterization of the effects of methylmercury on Caenorhabditis elegans. \u003cem\u003eToxicol. Appl. Pharmacol.\u003c/em\u003e \u003cb\u003e240\u003c/b\u003e (2), 265\u0026ndash;272 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHartman, J. H. et al. Xenobiotic metabolism and transport in Caenorhabditis elegans. \u003cem\u003eJ. Toxicol. Environ. Health B Crit. Rev.\u003c/em\u003e \u003cb\u003e24\u003c/b\u003e (2), 51\u0026ndash;94 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWidmayer, S. J. et al. C. elegans toxicant responses vary among genetically diverse individuals. \u003cem\u003eToxicology\u003c/em\u003e \u003cb\u003e479\u003c/b\u003e, 153292 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAllard, P. et al. A C. elegans screening platform for the rapid assessment of chemical disruption of germline function. \u003cem\u003eEnviron. Health Perspect.\u003c/em\u003e \u003cb\u003e121\u003c/b\u003e, 717\u0026ndash;724 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUlaganathan, G. et al. Screening and characterization of 133 physiologically-relevant environmental chemicals for reproductive toxicity. \u003cem\u003eReprod. Toxicol.\u003c/em\u003e \u003cb\u003e126\u003c/b\u003e, 108602 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWyatt, L. H. et al. Effects of methyl and inorganic mercury exposure on genome homeostasis and mitochondrial function in Caenorhabditis elegans. \u003cem\u003eDNA repair.\u003c/em\u003e \u003cb\u003e52\u003c/b\u003e, 31\u0026ndash;31 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeung, M. C. et al. Caenorhabditis elegans generates biologically relevant levels of genotoxic metabolites from aflatoxin B1 but not benzo[a]pyrene in vivo. \u003cem\u003eToxicol. Sci.\u003c/em\u003e \u003cb\u003e118\u003c/b\u003e (2), 444\u0026ndash;453 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang, B. et al. High-throughput assessment of toxic effects of metal mixtures of cadmium(Cd), lead(Pb), and manganese(Mn) in nematode Caenorhabditis elegans. \u003cem\u003eChemosphere\u003c/em\u003e \u003cb\u003e234\u003c/b\u003e, 232\u0026ndash;241 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliams, P. L. \u0026amp; Dusenbery, D. B. Using the nematode Caenorhabditis elegans to predict mammalian acute lethality to metallic salts. \u003cem\u003eToxicol. Ind. Health\u003c/em\u003e. \u003cb\u003e4\u003c/b\u003e (4), 469\u0026ndash;478 (1988).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurrie, S. D. et al. The impact of early life exposure to individual and combined PFAS on learning, memory, and bioaccumulation in C. elegans. \u003cem\u003eEnviron. Pollut\u003c/em\u003e. \u003cb\u003e363\u003c/b\u003e (Pt 2), 125257 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOecd \u003cem\u003eTest No. 222: Earthworm Reproduction Test (Eisenia fetida/Eisenia andrei).\u003c/em\u003e OECD Guidelines for the Testing of Chemicals, Section 2, (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGutner-Hoch, E. et al. Toxicity of engineered micro- and nanomaterials with antifouling properties to the brine shrimp Artemia salina and embryonic stages of the sea urchin Paracentrotus lividus. \u003cem\u003eEnviron. Pollut.\u003c/em\u003e \u003cb\u003e251\u003c/b\u003e, 530\u0026ndash;537 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCunningham, P. A. \u0026amp; Grosch, D. S. \u003cem\u003eA comparative study of the effects of mercuric chloride and methyl mercury chloride on reproductive performance in the brine shrimp, Artemia salina.\u003c/em\u003e Environmental Pollution 1978. 15(2): pp. 83\u0026ndash;99. (1970).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTkaczyk, A. et al. Daphnia magna model in the toxicity assessment of pharmaceuticals: A review. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cb\u003e763\u003c/b\u003e, 143038 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO., \u003cem\u003eTest No. 211: Daphnia magna Reproduction Test, OECD Guidelines for the Testing of Chemicals\u003c/em\u003e. 2012, OECD Publishing, Paris. (2012).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRim, K. T. In silico prediction of toxicity and its applications for chemicals at work. \u003cem\u003eToxicol. Environ. Health Sci.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e (3), 191\u0026ndash;202 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOECD, \u003cem\u003eTest No. 207: Earthworm, acute toxicity tests\u003c/em\u003e. 1984, OECD Guidelines for the Testing of Chemicals. p. Section 2-Section 2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO., \u003cem\u003eTest No. 222: Earthworm Reproduction Test (Eisenia fetida/Eisenia andrei), OECD Guidelines for the Testing of Chemicals\u003c/em\u003e. 2016, OECD Publishing, Paris. (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLai, C. H. et al. \u003cem\u003eIdentification of Novel Human Genes Evolutionarily Conserved in Caenorhabditis elegans by Comparative Proteomics.\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKaletta, T. \u0026amp; Hengartner, M. O. Finding function in novel targets: C. elegans as a model organism. \u003cem\u003eNat. Rev. Drug Discov\u003c/em\u003e. \u003cb\u003e5\u003c/b\u003e, 387\u0026ndash;398 (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAthar, F. \u0026amp; Templeman, N. M. C. elegans as a model organism to study female reproductive health. \u003cem\u003eComp. Biochem. Physiol. Mol. Integr. Physiol.\u003c/em\u003e \u003cb\u003e266\u003c/b\u003e, 111152 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLints, R. \u0026amp; Hall, D. H. \u003cem\u003eReproductive system, germ line.\u003c/em\u003e, in \u003cem\u003eWormAtlas\u003c/em\u003e, L.A. Herndon, Editor. (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLarigot, L. et al. \u003cem\u003eCytochromes P450 of Caenorhabditis elegans: Implication in Biological Functions and Metabolism of Xenobiotics.\u003c/em\u003e (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Y. et al. Correlation of chemical acute toxicity between the nematode and the rodent The utility of any non-rodent model system for chemical toxicity screening depends on the level of corre. \u003cem\u003eCite this: Toxicol. Res.\u003c/em\u003e \u003cb\u003e2\u003c/b\u003e, 403\u0026ndash;403 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKwah, J. K. \u0026amp; Jaramillo-Lambert, A. \u003cem\u003eMeasuring Embryonic Viability and Brood Size in Caenorhabditis elegans\u003c/em\u003e. \u003cem\u003eJ. Vis. Exp.\u003c/em\u003e, 2023(192).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee, S. Y. \u0026amp; Kang, K. \u003cem\u003eMeasuring the Effect of Chemicals on the Growth and Reproduction of Caenorhabditis elegans\u003c/em\u003e. \u003cem\u003eJ. Vis. Exp.\u003c/em\u003e, 2017(128).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDuPlissis, A. et al. Machine learning-based analysis of microfluidic device immobilized C. elegans for automated developmental toxicity testing. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e (1), 15 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMondal, S. et al. Large-scale microfluidics providing high-resolution and high-throughput screening of Caenorhabditis elegans poly-glutamine aggregation model. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cb\u003e7\u003c/b\u003e, 13023 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMondal, S. et al. \u003cem\u003eHigh-Content Microfluidic Screening Platform Used To Identify sigma2R/Tmem97 Binding Ligands that Reduce Age-Dependent Neurodegeneration in C. elegans SC_APP Model\u003c/em\u003e (ACS Chem Neurosci, 2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLetizia, M. C. et al. Microfluidics-enabled phenotyping of a whole population of C. elegans worms over their embryonic and post-embryonic development at single-organism resolution. \u003cem\u003eMicrosyst. Nanoeng\u003c/em\u003e. \u003cb\u003e4\u003c/b\u003e, 6 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, S., Stone, H. A. \u0026amp; Murphy, C. T. A microfluidic device and automatic counting system for the study of C. elegans reproductive aging. \u003cem\u003eLab. Chip\u003c/em\u003e. \u003cb\u003e15\u003c/b\u003e (2), 524\u0026ndash;531 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBanse, S. A. et al. \u003cem\u003eThe Egg-Counter: A novel microfluidic platform for characterization of Caenorhabditis elegans egg-laying.\u003c/em\u003e bioRxiv, : p. 2023.09.01.555781\u0026ndash;2023.09.01.555781. (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrenner, S. The genetics of Caenorhabditis elegans. \u003cem\u003eGenetics\u003c/em\u003e \u003cb\u003e77\u003c/b\u003e, 71\u0026ndash;94 (1974).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStiernagle, T. \u003cem\u003eMaintenance of C. elegans.\u003c/em\u003e WormBook, : pp. 1\u0026ndash;11. (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGroen, F. C., Young, I. T. \u0026amp; Ligthart, G. A comparison of different focus functions for use in autofocus algorithms. \u003cem\u003eCytometry\u003c/em\u003e \u003cb\u003e6\u003c/b\u003e, 81\u0026ndash;91 (1985).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun, Y., Duthaler, S. \u0026amp; Nelson, B. J. Autofocusing in computer microscopy: selecting the optimal focus algorithm. \u003cem\u003eMicrosc Res. Tech.\u003c/em\u003e \u003cb\u003e65\u003c/b\u003e, 139\u0026ndash;149 (2004).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, X. D. A pair of new statistical parameters for quality control in RNA interference high-throughput screening assays. \u003cem\u003eGenomics\u003c/em\u003e \u003cb\u003e89\u003c/b\u003e (4), 552\u0026ndash;561 (2007).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, X. D. Illustration of SSMD, z Score, SSMD*, z* Score, and t Statistic for Hit Selection in RNAi High-Throughput Screens. \u003cem\u003eSLAS Discovery\u003c/em\u003e. \u003cb\u003e16\u003c/b\u003e (7), 775\u0026ndash;785 (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, X. D. \u003cem\u003eOptimal high-throughput screening: practical experimental design and data analysis for genome-scale RNAi research\u003c/em\u003e 47\u0026ndash;53 (Cambridge University Press, 2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNlm, C. et al. \u003cem\u003eAdvanced Assay Development Guidelines for Image-Based High Content Screening and Analysis.\u003c/em\u003e (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChisholm, A. D. \u0026amp; Hardin, J. \u003cem\u003eEpidermal morphogenesis\u003c/em\u003e. WormBook. pp. 1\u0026ndash;22. (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlOkda, A. \u0026amp; Raamsdonk, J. M. V. \u003cem\u003eEffect of DMSO on lifespan and physiology in C. elegans: Implications for use of DMSO as a solvent for compound delivery.\u003c/em\u003e microPublication Biology, 2022: p. (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.17912/micropub.biology.000634-10.17912/micropub.biology.000634\u003c/span\u003e\u003cspan address=\"10.17912/micropub.biology.000634-10.17912/micropub.biology.000634\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCalahorro, F., Holden-Dye, L. \u0026amp; O'Connor, V. Impact of drug solvents on C. elegans pharyngeal pumping. \u003cem\u003eToxicol. Rep.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 1240\u0026ndash;1240 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eO., \u003cem\u003eOECD (1984), Test No. 207: Earthworm, Acute Toxicity Tests, OECD Guidelines for the Testing of Chemicals\u003c/em\u003e. 1984, OECD Publishing, Paris. (1984).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHelmcke, K. J. \u0026amp; Aschner, M. Hormetic effect of methylmercury on Caenorhabditis elegans. \u003cem\u003eToxicol. Appl. Pharmcol.\u003c/em\u003e \u003cb\u003e248\u003c/b\u003e (2), 156\u0026ndash;156 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScheuhammer, A. M. et al. \u003cem\u003eEffects of Environmental Methylmercury on the Health of Wild Birds, Mammals, and Fish.\u003c/em\u003e (2007). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1579/0044-7447\u003c/span\u003e\u003cspan address=\"10.1579/0044-7447\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e36[12:EOEMOT]2.0.CO;2, 2007. 36(1): pp. 12\u0026ndash;19.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrlandini Costa, N. et al. \u003cem\u003eEvaluation of the reproductive toxicity of fungicide propiconazole in male rats.\u003c/em\u003e (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSvanholm, S. et al. Developmental reproductive toxicity and endocrine activity of propiconazole in the Xenopus tropicalis model. \u003cem\u003eSci. Total Environ.\u003c/em\u003e \u003cb\u003e753\u003c/b\u003e, 141940\u0026ndash;141940 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarlier, M. P. et al. Using Caenorhabditis elegans and zebrafish embryos (Danio rerio) in a combined approach to screen chemicals for developmental (neuro)toxicity and reproductive toxicity. \u003cem\u003eNAM J.\u003c/em\u003e \u003cb\u003e2\u003c/b\u003e, 100080 (2026).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKillian, D. J. \u0026amp; Hubbard, E. J. A. Caenorhabditis elegans germline patterning requires coordinated development of the somatic gonadal sheath and the germ line. \u003cem\u003eDev. Biol.\u003c/em\u003e \u003cb\u003e279\u003c/b\u003e (2), 322\u0026ndash;335 (2005).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCarter, J. et al. On the Control of Oocyte Meiotic Maturation and Ovulation inCaenorhabditis elegans. \u003cem\u003eDev. Biol.\u003c/em\u003e \u003cb\u003e205\u003c/b\u003e (1), 111\u0026ndash;128 (1999).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFischer, M. et al. Repeatability and Reproducibility of the RTgill-W1 Cell Line Assay for Predicting Fish Acute Toxicity. \u003cem\u003eToxicol. Sci.\u003c/em\u003e \u003cb\u003e169\u003c/b\u003e (2), 353\u0026ndash;364 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWaggoner, L. E. et al. Control of alternative behavioral states by serotonin in Caenorhabditis elegans. \u003cem\u003eNeuron\u003c/em\u003e \u003cb\u003e21\u003c/b\u003e (1), 203\u0026ndash;214 (1998).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanDuyn, N. et al. SKN-1/Nrf2 inhibits dopamine neuron degeneration in a Caenorhabditis elegans model of methylmercury toxicity. \u003cem\u003eToxicol. Sci.\u003c/em\u003e \u003cb\u003e118\u003c/b\u003e (2), 613\u0026ndash;624 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu, K. et al. Ecotoxicity Risk of Low-Dose Methylmercury Exposure to Caenorhabditis elegans: Multigenerational Toxicity and Population Discrepancy. \u003cem\u003eChem. Res. Toxicol.\u003c/em\u003e \u003cb\u003e34\u003c/b\u003e (4), 1114\u0026ndash;1123 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXing, X. J. et al. Exposure to lead and mercury in young larvae induces more severe deficits in neuronal survival and synaptic function than in adult nematodes. \u003cem\u003eArch. Environ. Contam. Toxicol.\u003c/em\u003e \u003cb\u003e56\u003c/b\u003e (4), 732\u0026ndash;741 (2009).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcElwee, M. K. et al. \u003cem\u003eComparative toxicogenomic responses of mercuric and methyl-mercury.\u003c/em\u003e BMC Genomics 2013. 14(1): pp. 698\u0026ndash;698-. (2013) 14:1.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSkolness, S. Y. et al. \u003cem\u003ePropiconazole Inhibits Steroidogenesis and Reproduction in the Fathead Minnow (Pimephales promelas).\u003c/em\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKast-Hutcheson, K., Rider, C. V. \u0026amp; LeBlanc, G. A. The fungicide propiconazole interferes with embryonic development of the crustacean Daphnia magna. \u003cem\u003eEnviron. Toxicol. Chem.\u003c/em\u003e \u003cb\u003e20\u003c/b\u003e (3), 502\u0026ndash;509 (2001).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSouders, C. L. et al. \u003cem\u003eSub-lethal effects of the triazole fungicide propiconazole on zebrafish (Danio rerio) development, oxidative respiration, and larval locomotor activity\u003c/em\u003e. \u003cem\u003eNeurotoxicol. Teratol.\u003c/em\u003e, 74. (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTeng, M. et al. Parental exposure to propiconazole at environmentally relevant concentrations induces thyroid and metabolism disruption in zebrafish (Danio rerio) offspring: An in vivo, in silico and in vitro study. \u003cem\u003eEcotoxicol. Environ. Saf.\u003c/em\u003e \u003cb\u003e242\u003c/b\u003e, 113865\u0026ndash;113865 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSebasti\u0026atilde;o, M. V. R. et al. \u003cem\u003eEcotoxicological assessment of pesticide mixtures: effects of imidacloprid and propiconazole on Danio rerio and Astyanax lacustris\u003c/em\u003e. \u003cem\u003eAquat. Toxicol.\u003c/em\u003e, 287. (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd, W. A. et al. Developmental effects of the ToxCast\u0026trade; phase I and phase II chemicals in caenorhabditis elegans and corresponding responses in Zebrafish, Rats, and Rabbits. \u003cem\u003eEnviron. Health Perspect.\u003c/em\u003e \u003cb\u003e124\u003c/b\u003e (5), 586\u0026ndash;593 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWittkowski, P. et al. \u003cem\u003eCaenorhabditis elegans As a Promising Alternative Model for Environmental Chemical Mixture Effect AssessmentA Comparative Study.\u003c/em\u003e (2019).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Developmental and reproductive toxicology (DART), C. elegans, in utero embryo phenotype, high-resolution imaging, microfluidic technology, standard operating procedure (SOP), repeatable assays, acceptance criteria, new approach methodology (NAM)","lastPublishedDoi":"10.21203/rs.3.rs-9004834/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9004834/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eDevelopmental and Reproductive Toxicity (DART) assessment is essential for product safety evaluation and currently relies heavily on vertebrate models that are costly, time-consuming, and resource-intensive. \u003cem\u003eCaenorhabditis elegans\u003c/em\u003e has emerged as a promising New Approach Methodology (NAM) for rapid, cost-effective, whole-organism toxicology studies. However, broader adoption has been limited by the lack of high-resolution, rapid imaging approaches, a limited range of endpoints, and insufficient evidence for assay robustness and repeatability. To overcome these limitations, we developed a multiparametric imaging-based assay for assessing DART-related endpoints in \u003cem\u003eC. elegans\u003c/em\u003e, building on our previously published microfluidic-based developmental toxicity platform. We expanded the machine learning-based body dimension analysis to include quantification of total embryo number and \u003cem\u003ein utero\u003c/em\u003e embryonic development by classifying embryos as early- or late-stage to assess reproductive health. The assay relies on high-resolution brightfield imaging, enabled by the vivoChip microfluidic device, and is compatible with any strain. This study demonstrates highly repeatable results with mean coefficients of variation of 1\u0026ndash;5% for developmental endpoints and 6\u0026ndash;17% for reproductive-related endpoints, supporting high statistical power. Validation using methylmercury and propiconazole and by phenotyping (scoring and classification)\u0026thinsp;~\u0026thinsp;400,000 embryos across ~\u0026thinsp;9,200 worms, demonstrated reproducible, concentration-dependent responses across all endpoints with narrow confidence intervals. Notably, late-stage embryos were the most sensitive endpoint, with effects preceding changes in total embryo count, body size, or viability. Importantly, worms remained\u0026thinsp;\u0026gt;\u0026thinsp;97% viable and motile across all tested concentrations, indicating that the observed effects reflect DART-specific responses rather than non-specific apical endpoints such as lethality. These results demonstrate a sensitive, repeatable, and scalable DART platform capable of rapid, cost-effective chemical prioritization for safety assessment.\u003c/p\u003e","manuscriptTitle":"A Robust, High-Content NAM for Repeatable and Predictive Developmental and Reproductive Toxicity Assessment in C. elegans","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-10 16:52:48","doi":"10.21203/rs.3.rs-9004834/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-05-15T06:39:22+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-11T22:35:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"309211888991359592283367774411553578071","date":"2026-04-21T04:54:26+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"215841215928866465301929040054250094210","date":"2026-04-21T01:33:30+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"108322075561588747301024862663338088119","date":"2026-04-20T05:15:47+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-18T21:37:14+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"162975857109689939136233396400293042384","date":"2026-04-17T11:49:03+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"123797974509286017203935258457438493117","date":"2026-04-08T08:32:36+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-06T10:24:35+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-03-05T15:09:26+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-04T06:44:59+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-04T06:42:16+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-03-02T02:20:58+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f3df0d0c-86a1-4529-9110-c1a7852249c9","owner":[],"postedDate":"April 10th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Revision requested","date":"2026-05-15T06:39:22+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-05-11T22:35:07+00:00","index":96,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[{"id":65917532,"name":"Biological sciences/Biological techniques"},{"id":65917533,"name":"Biological sciences/Biotechnology"},{"id":65917534,"name":"Biological sciences/Developmental biology"}],"tags":[],"updatedAt":"2026-05-15T06:54:46+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-10 16:52:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9004834","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9004834","identity":"rs-9004834","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}