Neural-circuit basis of song preference learning in fruit flies

Abstract

As observed in human language learning and song learning in birds, the fruit fly Drosophila melanogaster changes its auditory behaviors according to prior sound experiences. This phenomenon, known as song preference learning in flies, requires GABAergic input to pC1 neurons in the brain, with these neurons playing a key role in mating behavior. The neural circuit basis of this GABAergic input, however, is not known. Here, we find that GABAergic neurons expressing the sex-determination gene doublesex are necessary for song preference learning. In the brain, only four doublesex-expressing GABAergic neurons exist per hemibrain, identified as pCd-2 neurons. pCd-2 neurons directly, and in many cases mutually, connect with pC1 neurons, suggesting the existence of reciprocal circuits between them. Moreover, GABAergic and dopaminergic inputs to doublesex-expressing GABAergic neurons are necessary for song preference learning. Together, this study provides a neural circuit model that underlies experience-dependent auditory plasticity at a single-cell resolution.

Keywords: Behavioral neuroscience; Cellular neuroscience; Molecular neuroscience.

Graphical abstract

Figure 1

doublesex-expressing GABAergic neurons are necessary for song preference learning (A) Expression pattern of dsx∩Gad1 driver-1 in the brain (top and middle-left; anterior view) and ventral nerve cord (bottom; ventral view) in females. A magnified view of the cell bodies (arrowheads) is shown in the middle-right. SMP, superior medial protocerebrum; GNG, gnathal ganglia. Scale bars, 50 μm. A, anterior; D, dorsal; L, lateral (the same in the following Figures). See also Figures S2 and S3. (B) Venn diagram of the genetic intersection. The population labeled by the dsx∩Gad1 driver-1 is shown in green. (C) Experimental scheme for song preference learning. In the training session, females in the experienced condition are exposed to the conspecific song for the first 6 days after eclosion, whereas naive females are kept in silence. During the test session on the 7^th day, females of both conditions are paired with mute males and exposed to either conspecific or heterospecific song. (D) Song preference learning in Gad1 knockdown females. The number of trials for each group is shown in parentheses. NC, naive flies tested with the conspecific song; NH, naive flies tested with heterospecific song.; EC, experienced flies tested with the conspecific song; EH, experienced flies tested with heterospecific song. Not significant (N.S.), p > 0.05; ∗, p < 0.05; log-logistic AFT model. See also Figure S1. (E) The learning index (LI) estimated based on the cumulative copulation rate using log-logistic AFT model. The horizontal axis uses a natural logarithm scale (see STAR methods for details). The squares indicate estimated LI, and horizontal lines indicate 95% confidence intervals (CIs) (same in the following Figures).

Figure 2

Synaptic connections between pCd-2 and pC1 (A) GRASP between pCd-2 neurons and pC1 neurons. The brain region shown in the middle to right panels is outlined in the left panel. GRASP signals (green), spGFP_1-10 and spGFP₁₁ expression in pCd-2 neurons and pC1 neurons, respectively (magenta), spGFP_1-10 expression in pCd-2 neurons (light blue), and merged image are shown (see Table S1 for the genotype). Scale bar, 50 μm. (B) Synaptic connections between pC1 neurons and pCd-2 neurons. Red and blue arrows depict the output and input synapses of pCd-2 neurons to/from pC1 neurons, respectively. Numbers on arrows indicate the number of synapses. See also Figure S6. (C–F) Single pCd-2 neurons in the right hemibrain of the FlyEM dataset. See also Figure S8. (G–J) Synaptic connections between pCd-2 neurons and pC1 neurons based on the FlyEM dataset. Numbers on arrows indicate the number of synapses. Weak connections (fewer than six synapses) are shown in dotted arrows. See also Figures S7 and S12.

Figure 3

Involvement of GABA and dopamine receptors in song preference learning (A) Cumulative copulation rates in control and Rdl knockdown groups. NC, naive flies tested with the conspecific song; NH, naive flies tested with heterospecific song.; EC, experienced flies tested with the conspecific song; EH, experienced flies tested with heterospecific song. The number of trials in each group is shown in parentheses. (B) Learning index (LI) in control and Rdl knockdown groups. (C) Cumulative copulation rates in control and Dop1R2 knockdown groups. (D) LI in control and Dop1R2 knockdown groups. (E) Restricted mean time lost (RMTL) of cumulative copulation rate for each group. Plots display the average (circle or triangular dot in each box) and standard errors (horizontal bars). Exp, experienced; Con, conspecific song; Hetero, heterospecific song. Log-logistic AFT model (A–D) and restricted mean survival time with Bonferroni correction (E) were used. N.S., not significant; ∗p < 0.05. See also Figures S9, S10, and S12.

Figure 4

Neural circuit model for song preference learning in flies (A) A model for experience-dependent tuning of song responses in fruit flies. Flies fine-tune their innate IPI preference through auditory experiences. Modified from with permission. Red shaded area is shown in greater detail in (B). (B) A model for the neural circuit mechanism of experience-dependent modulation. Song preference learning in flies involves at least two distinct mechanisms, the experience-dependent suppression of the response to the heterospecific song (blue line in the copulation plot) and the maintenance of the response to the conspecific song (black line in the copulation plot) after experience. These two mechanisms are mediated by GABA and dopamine, respectively, transmitted to pCd-2 neurons. pC1 neurons are excitatory cholinergic neurons transmitting acetylcholine (ACh).