High-Throughput Phenotyping of C. elegans

Animals were grown in 96-well microtiter plates to allow for automated liquid handling to speed assay set up and phenotyping. In each well, the final volume was 50 µL of S medium solution and HB101 bacterial food. HB101 was prepared in large batches (greater than 20 L) to reduce assay-to-assay variability. The bacteria were prepared from cultures grown for 24 hr in Superbroth and then centrifuged to concentrate. HB101 bacteria were resuspended at 20% volume per volume in S medium and frozen at −80° (without glycerol) in 1-mL aliquots. Bacteria were thawed and fed to animals at a final concentration of 2% volume per volume. This amount of food was sufficient to sustain the number of offspring derived from a single fourth larval stage hermaphrodite with food left over at the end of the assay. For any mapping or causality experiment, we used the same HB101 culture to reduce the substantial variation that arises from bacteria grown in independent preparations. Solutions of S medium and bacterial food were prepared and then split for control and paraquat conditions.
After assay plates were prepared, one mid to late fourth larval stage animal was singled using a platinum wire to each 50-µL well. In the plate set up, we separated different genotypes with a wash well containing only S medium to decrease carry-over and mixing of independent genotypes from one well to the next. After all wells were populated with animals, the microtiter plate was sealed with a Breath-Easy film (USA Scientific) and placed in a humidity chamber lined with damp towels, closed, and then sealed with parafilm. We observed less than 5% of the well volume evaporated after four days under these conditions. Humidity chambers were placed into incubators set to 20° and shaking at 180 rpm. These conditions ensured that cultures and bacterial food was constantly mixing so animals would never enter hypoxia nor encounter regions of depleted food in the well. Animals were grown for 96 hr in these conditions and then prepared for measurement on the COPAS BIOSORT (Union Biometrica).
Two minutes before the animals were loaded on the COPAS BIOSORT, 200 µL of M9 with 50 mM sodium azide was added to the wells. The sodium azide kills and straightens the animals to ensure proper measurement of body length as the paralyzed animal passes through the flow cell. The COPAS BIOSORT sheath flow rate was kept constant at 9.8 mL per minute to decrease variability in length measurements as much as possible. Then, 96-well microtiter plates were aspirated using the ReFLx module with the BISORT set to “no-bubble-trap” mode. When the system is run in “no-bubble-trap” mode, all wells from a single microtiter plate can be measured in approximately 25 min and well-to-well contamination is less than 1%, which is further mitigated by the use of wash wells between unique genotypes. Extinction and time of flight minimums were 50 and 20, respectively. Green, yellow, and red photomultiplier tubes were set to 700, 700, and 900, respectively. Signal multipliers were set to 1.0 and signal gains set to 3.0. The COPAS BIOSORT software had Profiler II enabled. Flat comma-separated value files were analyzed using custom scripts in the R statistical computing environment. The COPAS BIOSORT software can not differentiate bubbles from animals as objects pass through the flow cell. For this reason, we trained a support vector machine to differentiate these two types of objects with 99.97% accuracy (Supporting Information, Figure S1). The support vector machine primarily uses optical density raw values to generate a binary linear classifier separating bubbles from animals. The raw data were read in, processed, and plotted using the COPASutils R package (Shimko and Andersen 2014 (link)). All of the raw and processed data (File S2) can be accessed on the Andersen lab GitHub page. High-throughput assay plates with paraquat were prepared at 1.5 mM paraquat (Chem Service, Inc.) or no paraquat from the same dilution of HB101 and S medium.