The following protocol describes the development and optimization of a high-throughput workflow for worm culturing, fluorescence imaging, and automated image processing to quantify polyglutamine aggregates as an assessment of changes in proteostasis.
A rise in the prevalence of neurodegenerative protein conformational diseases (PCDs) has fostered a great interest in this subject over the years. This increased attention has called for the diversification and improvement of animal models capable of reproducing disease phenotypes observed in humans with PCDs. Though murine models have proven invaluable, they are expensive and are associated with laborious, low-throughput methods. Use of the Caenorhabditis elegans nematode model to study PCDs has been justified by its relative ease of maintenance, low cost, and rapid generation time, which allow for high-throughput applications. Additionally, high conservation between the C. elegans and human genomes makes this model an invaluable discovery tool. Nematodes that express fluorescently tagged tissue-specific polyglutamine (polyQ) tracts exhibit age- and polyQ length-dependent aggregation characterized by fluorescent foci. Such reporters are often employed as proxies to monitor changes in proteostasis across tissues. Manual aggregate quantification is time-consuming, limiting experimental throughput. Furthermore, manual foci quantification can introduce bias, as aggregate identification can be highly subjective. Herein, a protocol consisting of worm culturing, image acquisition, and data processing was standardized to support high-throughput aggregate quantification using C. elegans that express intestine-specific polyQ. By implementing a C. elegans-based image processing pipeline using CellProfiler, an image analysis software, this method has been optimized to separate and identify individual worms and enumerate their respective aggregates. Though the concept of automation is not entirely unique, the need to standardize such procedures for reproducibility, elimination of bias from manual counting, and increase throughput is high. It is anticipated that these methods can drastically simplify the screening process of large bacterial, genomic, or drug libraries using the C. elegans model.
Age-dependent neurodegenerative protein conformational diseases (PCDs) such as Alzheimer's, Parkinson's, and Huntington's diseases, or amyotrophic lateral sclerosis, are characterized by protein misfolding that leads to aggregation, cell death, and tissue degeneration1. While protein misfolding is recognized as the culprit, the etiology of these diseases is not clear. As such, the development of effective therapies has been hindered by the lack of knowledge regarding the factors and conditions that contribute to disease onset and progression. Recent studies suggest that changes in the microbiome influence the onset, progression, and severity of PCDs2,3,4. However, the complexity of the human, or even murine, microbiome makes it difficult to conduct studies that would reveal the exact influence of microbes on their host. Therefore, simpler organisms, such as Caenorhabditis elegans, are often used as a discovery tool5,6,7,8. Recent studies have employed C. elegans to investigate the effect of bacteria on host proteostasis and disease pathogenesis9,10. Bacterial colonization, hormesis, and genomic changes are among exemplar conditions that affect the aggregation of polyglutamine (polyQ) tracts9,11,12. Additionally, these misfolded protein clusters exhibit polyQ length- and age-dependent accumulation within the host and are associated with impaired motility9,13. The relatively simple approach of quantifying fluorescently labeled puncta can generate important data on conditions, factors, or drugs that affect protein folding and aggregation.
Though quantification of fluorescent puncta has proven to be a reliable and relatively simple procedure, the challenge remains to develop a protocol that would facilitate large-scale screening of compounds, bacteria, or conditions that affect protein aggregation. The concept of automated C. elegans image processing and puncta quantification is not entirely novel, as a number of practical support tools have been developed14,15. However, the integration of culturing, image acquisition, and a processing pipeline are essential in eliminating variability in results and allowing for higher-throughput screens.
As such, the intent of this manuscript is to standardize the procedure used to quantify polyQ aggregation in C. elegans as a proxy to detect changes in proteostasis. This task was accomplished by employing CellProfiler, an open-source image analysis software16 capable of automated worm and aggregate identification, and is integrated into a larger protocol for culturing worms, acquiring images, and processing data.
The described protocol outlines procedures for C. elegans culturing, imaging, and image processing that incorporates CellProfiler, an open-source image analysis software. The representative results demonstrate reproducibility, reduction of bias, and scalability. This standardized procedure will improve screening strategies employed with large bacterial, genomic, or drug libraries. While other automated C. elegans methods of object detection exists, the described technique offers a standardized, higher-throughput pipeline that integrates culturing, image acquisition, and analysis.
Several variations of worm cultivation had to be tested to optimize the protocol described herein. Initially, worms were transferred to sample bacteria immediately post age-synchronization (L1 stage). However, such an approach resulted in a population of worms with variable sizes, even among worms within the same well. C. elegans are known for pathogen avoidance19, which could have contributed to the observed variability in size and ultimately affect downstream imaging-worm detection, in particular. To eliminate such variability, the entire NGM area in each well was covered with test bacteria. Furthermore, worms were fed E. coli OP50 and allowed to fully develop into young adults for 48 h at 25 °C. Allowing worms to reach adulthood on E. coli OP50 prior to transferring them onto test bacteria resulted in more consistent body size. Additionally, overcrowding and rapid food depletion by progeny were eliminated by supplementing NGM agar with FUDR. The implementation of FUDR removed progeny and enhanced automated worm identification, which was obscured by progeny mixing in with the parental population. However, it is important to be cautious and use appropriate controls when utilizing FUDR, as the compound is known to affect C. elegans proteostasis and lifespan20,21. Under the conditions described in this protocol, FUDR did not affect intestinal polyQ aggregation (Supplemental Figure 5); therefore, its utilization was suitable and beneficial to the described method.
Freezing samples prior to imaging turned out to be a critical step in the successful employment of the pipeline. The aggregate counts prior to freezing were significantly higher than manual counts (Figure 3B). Keeping worms at -20 °C for 18-48 h prior to imaging reduced background fluorescence and ultimately improved aggregate detection (Figure 3A). The effects of freezing on aggregate detection have only been investigated for polyQ and should not be generalized to other models without further investigation of such effects.
Despite all the conditions being kept the same, it was observed that the average number of aggregates per worm could vary between different runs, while the ratio between the number of aggregates in animals colonized with OP50 versus MPAO1 remained consistent (Figure 6, Supplemental Figure 2, Supplemental Figure 5). Therefore, it is essential to always include E. coli OP50 control, or any additional suitable reference controls, in every run. Such variability in aggregate counts between experiments could be influenced by environmental conditions (temperature, humidity)22,23 or genetic background8. In fact, it was observed that after prolonged culture, intestinal fluorescence drastically decreased or was completely lost, which required thawing a new strain from frozen stock. The observed decrease in fluorescence could be a result of genetic changes that suppress toxic transgenes, such as those expressing polyQ. Nonetheless, the exceptional reproducibility of the results observed between different experimenters (Supplemental Figure 2), between biological replicates (Figure 5), and within the same sample (Supplemental Figure 3) emphasize on the strength of this approach.
Numerous reports have employed intestinal polyQ to study proteostasis9,11,12,13,24,25. However, a direct comparison between results cannot be made due to variability between experimental approaches and readout methods. Nonetheless, a few results from previously published data are recapitulated by the automated quantification described herein, including bacterial induction of aggregation9,13 and a comparable number of aggregates11. Collectively, the described pipeline offers a valuable tool to study proteostasis.
The method described herein has some inherent challenges. For example, it requires sufficient time to master all components of this protocol, which is especially true for section 8 of the protocol, which requires familiarity with the assay to determine if the images acquired are appropriate for pipeline analysis. Deviations from the image acquisition settings used in this protocol are possible; however, modification of the settings and worm training set will likely be required. This pipeline can distinguish aggregates of various sizes and those that are touching, which limits the "blending" of aggregates and ultimately increases the detection sensitivity. However, issues may arise when attempting to identify large aggregates that exceed the accepted size range, as expanding the upper size threshold may lead to errors caused by poor identification, such as the inability to differentiate aggregates that are touching. A balance between accuracy, size, and intensity must be found prior to image analysis. The efficiency of aggregate identification could be further improved by incorporating machine learning to create a neural network capable of enhancing aggregate detection. Such improvements are currently being explored and will greatly assist in addressing current issues such as the detection of aggregates that lie on different focal planes or that have abnormal shapes.
One notable weakness of the described method is the variability in automated aggregate counts, as they are not always recapitulated by manual counts in worms fed different bacterial strains. For example, based on automated counts, worms fed P. aeruginosa mutant 53 (M53) had significantly fewer aggregates compared to the wild-type strain (MPAO1) (Figure 7); however, confirmation of the hit showed no significant difference (Supplemental Figure 4). In general, high-throughput drug screens have a high rate of false-positive hit detection, and the described method is no exception26. Thus, it is a critical part of the protocol to confirm all potential hits.
While this protocol was optimized to fit a screening strategy to identify bacteria that affect host proteostasis, each step can be further modified to test the effect of genomic RNAi libraries, small molecules, or other conditions. Additional modifications can be made at each step to suit the requirements of a specific screening strategy. Furthermore, this technique provides a level of flexibility that allows for the optimization of each step to suit a specific model. For example, this approach can be extended to polyQ aggregation in other tissues or extracting other features detected in images such as monitoring gene expression using inducible fluorescent reporters (e.g., heat shock genes), assessing subcellular localization of proteins (e.g., nuclear localization of DAF-16), studying aggregation in other disease models (Aβ1-42, α-synuclein, TDP-43, etc.) or assessing physiological phenotypes, such as worm size.
The authors have nothing to disclose.
This work was supported by the National Institutes of Health (1RO3AG069056-01) and the Infectious Diseases Society of America funding to DMC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We thank the members of the Czyz Lab for proofreading the manuscript. Cartoon figures were generated using BioRender paid license.
1.7 mL Microtubes | Olympus Plastics | Cat#24-282 | Microcentrifuge tubes |
10 mL Serological Pipettes | GenClone | Cat#12-104 | Plastic pipettes |
15 mL Centrifuge Tubes, Racked | Olympus Plastics | Cat#28-101 | Conical tubes |
5-Fluoro-2′-deoxyuridine | Fisher Scientific | D2235100MG | FUDR |
Accu-jet Pro Pipette Controller | Genesee Scientific | 91-600RB | Pipette gun |
Agar | Fisher Scientific | Cat#BP1423-2 | Granulated agar |
BioRender | BioRender | Graphical figure generator | |
Bleach (Regular) | Clorox | Bar# 044600324111, Splash-less Bleach | 4.5% sodium hypochlorite |
C. elegans: AM140: rmIs132 [unc-54p::q35::yfp] | Morimoto Lab (Northwestern University) | AM140, Q35::YFP | Muscle polyQ |
C. elegans: AM738: rmIs297[vha-6p::q44::yfp; rol-6(su1006)] | Morimoto Lab (Northwestern University) | AM738, Q44::YFP | Intestinal polyQ |
CaCl2·2H2O | Fisher Scientific | Cat#C79-500 | Calcium chloride dihydrate |
CellProfiler | Broad Institute | Image analysis software | |
Cholesterol | Fisher Scientific | Cat#ICN10138201 | Cholesterol |
Circulating Water Bath Head | Lauda | 26LE | Lauda E 100 |
CoolLED pE300lite 365 dir mount STEREO | CoolLED | 8114931 | Fluorescent light control and emitter |
16 mL Culture Tubes | Olympus Plastics | 21-129 | Culture tubes, 17 mm x 100 mm |
Escherichia coli OP50 | Caenorhabditis Genetics Center | OP50 | E. coli control strain |
Ethanol, 200 Proof | Decon Labs, Inc. | 2701 | |
Eyepiece 10x/23B, adjustable, 3d gen | Leica Microsystems | 10450910 | Eyepiece set |
Filter Set ET GFP – MZ10F | Leica Microsystems | 10450588 | Filter cube |
GraphPad Prism v9.2.0 | GraphPad Software, Inc. | Statistical analysis tool | |
Heratherm Incubator IMP180 | Thermo | 51031562 | Refrigerated incubator |
Innova 4000 | New Brunswick Scientific | M1192-0000 | Shaking incubator |
K5 Camera | Leica Microsystems | 11547112 | Stereomicroscope camera |
KH2P04 | Fisher Scientific | Cat#P285-3 | Potassium phosphate monobasic |
LAS X Imaging Software | Leica Microsystems | Microscope imaging software | |
Leica MZ10 F Optics Carrier | Leica Microsystems | 10450103 | Stereomicroscope |
Levamisole | Fisher Scientific | Cat#0215522805 | Levamisole hydrochloride |
Luria Broth (Lennox) | Apex Bioresearch Products | Cat#11-125 | LB |
Magnetic Stir Plate | Fisher Scientific | 11-100-49S | Stir plate |
MgSO4·7H2O | Alfa Aesar | A14491 | Magnesium sulfate heptahydrate |
Microscope Slides | Premiere | 8205 | Single frosted microscope slides |
Na2HPO4·7H2O | Fisher Scientific | Cat#S373-500 | Sodium phosphate dibasic heptahydrate |
NaCl | Fisher Scientific | Cat#S671-500 | Sodium chloride |
NaOH | Fisher Scientific | Cat#S318-3 | Sodium hydroxide pellets |
Objective Achromat, f = 100 mm | Leica Microsystems | 10411597 | Objective microscope lens |
Petri Dishes | Genesee Scientific | Cat#32-107G | 100 mm x 15 mm |
Pseudomonas aeruginosa Mutant Library | Manoil Lab (University of Washington) | P. aeruginosa mutant library | |
Suspension Culture Plate 24-Well, Flat Bottom | Olympus Plastics | 25-102 | Used for worm growth and imaging |
Trinocular Tube 100% M-series | Leica Microsystems | 10450043 | |
Trypticase Peptone | ThermoFisher, Difco | Cat#211921 | |
TX-400 Rotor | Thermo Scientific | Cat#75003181 | Swing bucket rotor |
Vacuum Driven Filter System | GenClone | Cat#25-227 | 500 mL, PES Membrane, .22 µm |
Video Objective with C-Mount | Leica Microsystems | 10447367 | 0.63x camera adapter tube |