Figure 1. C. elegans shows both attractive and aversive responses to CO₂.

(A) Animals raised at ambient CO₂ (0.038%) avoid 5% CO₂, while animals raised at high (2.5%) CO₂ are attracted to 5% CO₂. ***p<0.001, Mann-Whitney test. n=12–14 trials per condition.

(B) Adults raised at ambient CO₂ were incubated at high (2.5%) CO₂ for 1, 3, 6, or 9 h and then assayed for their response to 2.5% CO₂ (blue), while adults raised at high CO₂ were put at ambient CO₂ for 3, 6, or 9 h and then assayed for their response to 2.5% CO₂ (orange). A switch in CO₂ environment triggers a rapid change in CO₂ response valence. n=8–24 trials per condition.

(C) BAG sensory neurons are required for CO₂ avoidance and attraction. Wild-type animals raised at ambient CO₂ avoid 2.5% CO₂, while wild-type animals raised at high (2.5%) CO₂ are attracted to 2.5% CO₂. BAG-ablated animals (BAG-) do not respond to CO₂ under either condition. ***p<0.001, two-way ANOVA with Sidak’s post-test. n=8–16 trials per condition.

(D) BAG neurons of animals raised at high (2.5%) CO₂ respond more robustly to CO₂ than BAG neurons of animals raised at ambient CO₂. Graph shows the calcium responses of BAG to 15%CO₂, for animals raised at ambient CO₂ (blue) or high CO₂ (orange), measured using the ratiometric calcium indicator yellow cameleon YC3.60. Solid lines indicate average calcium responses; shading represents SEM. Black line indicates the CO₂ pulse. Animals raised at high CO₂ show an increased BAG response relative to animals raised at ambient CO₂ (***p<0.001, unpaired t test). n=10–15 animals per condition.

(E-F) eat-4 and flp-17 are required for normal CO₂ response. (E) Mutation of eat-4 or flp-17 abolishes CO₂ avoidance in animals raised at ambient CO₂. Responses shown are to 5% CO₂. ***p<0.001, Kruskal-Wallis test with Dunn’s post-test. n=8–14 trials per genotype and condition.

(F) Mutation of either eat-4 or flp-17 reduces CO₂ attraction, and mutation of both genes abolishes CO₂ attraction, in animals raised at high (2.5%) CO₂. Responses shown are to 1% CO₂. ***p<0.001, one-way ANOVA with Dunnett’s post-test. n=8–16 trials per genotype and condition.

For A-C, E and F, graphs depict medians with interquartile ranges. See also Figures S1 and S2.