Effective drug combination for Caenorhabditis elegans nematodes discovered by output-driven feedback system control technique

See allHide authors and affiliations

Science Advances  04 Oct 2017:
Vol. 3, no. 10, eaao1254
DOI: 10.1126/sciadv.aao1254


  • Fig. 1 Schematic representation of the experiment.

    The effect of the four anthelmintics (namely, levamisole, pyrantel, methyridine, and tribendimidine) is tested on active, wild-type C. elegans. The dose-response data are fed into an FSC scheme that suggests new combinations of the four anthelmintics to test for worm survivability. Through multiple iterations of trying nonoptimal combinations and using a directed evolutionary search, the winning cocktail is eventually reached that makes the worm inactive, as judged by the values of average centroid velocity and track curvature.

  • Fig. 2 Dose response of the four anthelmintics is shown.

    Here, the percentage response is obtained by calculating the average centroid velocity of a worm population at a certain drug concentration and normalizing it to the velocity from the control experiments with M9 buffer. The EC50 (that is, drug concentration where the percentage response is 50%) values obtained from our microfluidic device are as follows: levamisole, 2.23 μM; pyrantel, 107.5 μM; tribendimidine, 4.68 μM; and methyridine, 1672 μM. Each drug concentration is tested on at least three separate chips with one worm per chamber and three chambers per chip.

  • Fig. 3 Drug concentration keys and FSC iterative process flow.

    (A) The concentration keys (0 to 5) are assigned to six individual concentrations of each anthelmintic. The highest concentration key (that is, key 5) is chosen close to the EC50 values, thus limiting our search to a winning combination having no more than the EC50 value of each compound. (B) Distinct color codes are chosen for each drug for visual representation. (C) In each iteration, the eight combinations are grouped into P group (P1 to P4) and T group (T1 to T4) candidates. The first iteration tests eight combinations. The second iteration retains the four best combinations from the first iteration as the P group and suggested four new combinations as the T group. This procedure is repeated in each iteration until the winning combination is reached (T3 in the fourth round here).

  • Fig. 4 Microfluidic device and worm tracking.

    (A) Movement of a wild-type C. elegans worm is tracked through a 600-s video where it shows distinct periods of being active, semiactive, and inactive. An active worm has relatively high velocity and smaller curvature, whereas an inactive worm displays minimal velocity and larger curvature. (B) Tracks of the same worm [as in (A)] are shown during the active, semiactive, and inactive periods (i to iii). The active worm has relatively straight tracks with a smaller curvature, whereas the inactive worm shows considerable struggle with minimal displacement and a larger curvature. Scale bars, 400 μm.

  • Fig. 5 Average centroid velocity and track curvature.

    (A) The centroid velocity averaged over all the worms for a given drug combination is plotted through all the four iterations. The centroid velocity of a single worm is averaged over the last 120s of the 600-s duration of the worm exposure. (B) Similarly, the track curvature averaged over all the worms for a given drug combination is depicted through the four iterations. Centroid velocity and curvature of worms in control experiments using M9 buffer are also shown. The winning combination was chosen as T3 in the fourth iteration that produced minimal centroid velocity and a larger curvature. Each drug combination is tested on at least three to five different chips. Each chip has three separate chambers with one worm per chamber. Bars are SEM.

  • Fig. 6 Instantaneous centroid velocity and track curvature are plotted for the entire 600 s of drug exposure.

    The data at every second are obtained by averaging the velocity or curvature values for all the worms exposed to a certain drug combination. (A) The winning combination, T3 of the fourth iteration, has minimal velocity (<10 μm/s) and large curvature (>50 mm−1) throughout the 600 s. (B and C) These combinations show medium velocity (50 to 100 μm/s) and curvature (25 to 50 mm−1), suggesting that the compounds affect the worms, but better alternatives are still possible. (D to F) These combinations exhibit high velocity (>100 μm/s) and low curvature (<25 mm−1), indicating that the drugs have a negligible effect on worm movement within the observed time period.

  • Fig. 7 Two-dimensional plot of track curvature versus centroid velocity.

    In the drug environment, the worm is classified as active, semiactive, or inactive as demonstrated in Fig. 4, and the plot is accordingly segregated into the three areas of activity. We are primarily interested in combinations that induce inactivity in the worm. (A) In iteration 1, two combinations (that is, P1 and T3) are able to make the worm inactive. Besides in P3, the active worm does not succumb to any other combination and is classified as being active. (B) In iteration 2, only P1 makes the worm inactive. The rest of the combinations do not seem to affect the active worm. (C) In iteration 3, the worm is in a semiactive or active state for all combinations. (D) In iteration 4, three combinations (T1, T2, and T3) are able to make the worm inactive, and the winning combination is chosen as T3.

Navigate This Article