Food choice behavior. (A) Food choice of wild-type L1 larvae. There are three arrangements of bacterial food, detailed on the left. In the ‘easy’ arrangement, colonies of different food touch each other; in the ‘harder’ and ‘hardest’ arrangements they are separate and 2 mm and 6 mm from the center of the plate, respectively. At time point 0, worms were placed in the center of the plate (marked with ×) at equal distance from the bacterial colonies. Worms were killed and counted at the indicated time points. (Right) Each area diagram represents a time course of the food choice between two bacterial species. Light gray and dark gray areas depict the fraction of worms in each food, and the white area depicts the fraction of worms outside the bacteria. The two bacterial species in each test are listed beneath and above the panels. In almost all cases where choice develops, it is in favor of the bacteria that better support growth. (B) Food choice of the eat mutants eat-2 and eat-5 between two Escherichia coli strains, HB101 and DA837. In all arrangements of bacteria the preference of eat mutants for the better food, HB101, is stronger than that of the wild type, consistent with DA837 being far worse food for both eat mutants (Fig. S1 in supplementary material). (C) Food choice of eat mutants between DA837 and Bacillus megaterium. While the wild type shows clear preference for DA837, the choice of eat mutants is not as strong and does not show a distinct trend in the easy and the harder arrangements, consistent with both DA837 and B. megaterium being bad foods for eat mutants. Mean ± s.e.m., N=5, except four cases with N=3 and 4, as indicated. Each data point is derived from a distribution of 70–150 worms on one assay plate. *Different from wild-type on the same food combination, P<0.05, Student’s t-test. (D) Food choice of wild-type L1 larvae in the biased food preference assay. At time point 0, worms were placed outside the circle, as shown by the ×. With time, animals crossed the circle and located the food in the center (y-axis of the plot). At indicated times, worms were killed and their distribution was counted. Results are expressed as a fraction of animals in the center. Worms migrate to the central colony only if the central colony is good food, Comamonas or E. coli HB101, whereas the circle is mediocre food, B. megaterium or E. coli DA837. Values are means ± s.e.m. Numbers of assays are 6–18 for different pairs of food.

Leaving behavior. (A) Leaving frequency of the wild type and two eat mutants, eat-2 and eat-5, in a population leaving assay (eat-5 was tested on E. coli foods only). Worms were placed near the small chunk of bacteria and allowed to enter it; the time when the first worm entered the colony was time 0. The plot shows mean leaving frequency between 1 and 2h after the start. On poor quality food, leaving is more active, and leaving behavior of eat-2 and eat-5 is more active than that of the wild type. Values are means ± s.e.m. (N=8). *Different from the wild type on the same bacteria (P<0.05). ^†Different from P(leaving) of the same worm strain on good food, HB101 and Comamonas (P<0.05). All comparisons are by Student’s t-test. (B) Leaving behavior of five individual eat-2 worms on E. coli DA837. A late-stage egg was placed near the colony. Time 0 is when the hatched larva first entered the colony. On the y-axis, ‘in’ is the time spent in the bacterial colony, while ‘out’ is the time spent outside the colony. (C) Leaving behavior of five individual wild-type worms on Bacillus megaterium; procedure as in B. Three out of five worms at some point, marked with arrows, left the colony and never returned. (D) Sample leaving trajectories of an individual wild-type worm on B. megaterium. Five typical segments of the animal’s trajectory are shown with different colors, and the direction of movement is shown with arrows.

Effect of dietary experience on food seeking behavior. (A,B) Effect of experience on food choice. (A) Wild-type L1 larvae were kept for 2·h in one of the four conditions, and then each group was tested in the ‘harder’ choice assay with three food combinations. The only two cases where the consistent time course of the food choice was observed were groups conditioned on Comamonas and tested on pairs of E. coli DA837 vs E. coli HB101, and E. coli DA837 vs Comamonas; in both cases worms chose better food. The effect of experience seems mild, but note that without conditioning, no preference develops with these pairs in a harder arrangement (Fig. 1A). Values are means ± s.e.m. (N=5). *Different from worms conditioned on E. coli DA837 and empty plate (P<0.02; Student’s t-test). (B) Wild-type larvae were exposed to one of the listed conditions for 2·h and then tested in a biased choice assay with a circle of DA837 surrounding a central colony of HB101. Worms that have experienced the high quality food, Comamonas and E. coli HB101, show the strongest food choice. Values are means ± s.d. (N=20; 10 plates with two circle assays on each). At all time points groups conditioned on food are different from those conditioned on the control empty plate (P<0.01; Student’s t-test). (C) Effect of food experience on leaving behavior. 50–80 naive eat-2 L1 larvae were conditioned in one of the five indicated conditions for 3·h, washed and transferred to another plate for a leaving assay. The time when the first worm entered the colony is time point 0. The x-axis shows the time intervals within which leaving probability was determined: 0–30·min, 30·min–1·h, 1–2·h, and in 1-h increments thereafter. After exposure to high quality food, Comamonas or HB101, leaving behavior was increased as compared to conditioning on the same food, worse food, or without food (empty plate). Values are means ± s.e.m. (N=6). *Different from worms conditioned in any of the other conditions (P<0.05; Student’s t-test).

Trajectories of wild-type worms on different bacterial food sources. (A) on Comamonas, (B) on E. coli HB101, (C) on E. coli DA837 and (D) on Bacillus megaterium. A single L1 larva was placed on a roughly rectangular bacterial lawn, and its movement trajectory during the subsequent 10–15·h was recorded. (Here, trajectories during the interval from 2–10·h of the experiment are shown.) The width of the field of view is approximately 10.3 mm; the bacterial lawn fits into the video field. The trajectory on B. megaterium is fragmented because of poor contrast; this problem was most severe at the edges, where bacteria tended to be the thickest. Roaming periods are shown in blue, dwelling in orange; there was much more roaming on mediocre food. Note that, since by definition the worms move slowly or not at all during dwelling, the extent of the orange traces doesn’t reflect the proportion of time spent dwelling. (E) Sample speed of locomotion and turning angle (direction of movement change) traces of the wild type on E. coli foods HB101 and DA837. Roaming periods (green bars), when the turning angle stays low and the speed is high, are common on mediocre food, E. coli DA837, but very rare on good food, E. coli HB101. (F) Speed of locomotion and (G) movement duration distributions. On mediocre foods, the speed of locomotion and the movement duration is increased. In F and G, data from 10 worms on E. coli DA837, 11 on E. coli HB101, eight on Comamonas sp. and five on B. megaterium were averaged; trajectories in the interval from 2 to 10·h from the start of the recording were analyzed for each animal. Data are expressed as mean ± s.e.m. between individual animals. In G, error bars for E. coli foods only are shown. Because the trajectory on B. megaterium is fragmented (D), the population of long events is artificially decreased. The increased roaming duration on B. megaterium is still obvious, but the real effect is even bigger, as suggested by the speed analysis (F).

AIY neurons function to extend food-seeking periods. Trajectory on mediocre food, E. coli DA837, of (A) a ttx-3 mutant and (B) an animal whose AIY neurons have been killed. Compare to wild-type in Fig. 4C. ttx-3 mutant trajectories did not span the whole lawn; and there were far fewer long straight roaming events. Trajectories of AIY worms also had fewer straight long movements than wild-type controls. (C,D) Movement duration distribution of (C) wild type, ttx-3, osm-6, osm-6;ttx-3 and (D) AIY-ablated animals, all tested on E. coli DA837 food. N=10 for WT, 10 for ttx-3, 6 for osm-6, 6 for osm-6;ttx-3, 10 for AIY ablations and 8 for ttx-3p::GFP controls. (E) ttx-3 was defective in the food preference behavior if bacterial foods were located at a small distance from each other. By 3·h, all ttx-3 worms found food, but there was no preference in the harder arrangement. In contrast to ttx-3, osm-6 animals took longer to discriminate between good and bad food, but they finally managed to make the right choice even if foods were located at a distance. Values are means ± s.e.m. *Different from the wild type (P<0.01); ^†different from ttx-3 (P<0.01; Student’s t-test). (F) Biased food preference for E. coli HB101 over B. megaterium of mutants and animals with laser-ablated neurons. The fraction of animals that reached the central colony of good food, E. coli HB101, was determined. ttx-3 mutants and AIY-ablated animals performed worse than controls. In laser ablation experiments, worms were counted after 20·h. For tests on mutants, the number of assays is 18 for WT, 15–17 for ttx-3 alleles and 6–15 for various mutants tested. For laser ablations, number of worms found in the center and the total number of worms tested is indicated next to the bars. Values are means ± s.e.m. *Different from the wild type (P<0.01; Student’s t-test); ^†Different from the ablation control (P<0.01; χ² test of independence).