We were unable to obtain a narrowly defined IC50 value for glutamate, perhaps due to cell-to-cell variation in glutamate receptor content induced by dissociation. However, full inhibition of the response to 10 μM ACh was produced with 10 μM glutamate (n = 6). To test whether GluCl contributes to the inhibitory GSK1210151A clinical trial effects of glutamate on LNvs, we repeated these experiments in a low chloride buffer (Figure 5E). This reduced glutamate inhibition of LNv responses to ACh by 75% ± 13% (n = 12). Therefore, LNvs require

extracellular Cl− for the majority of glutamate-induced inhibition. We also found that applying 500 nM ivermectin, an irreversible GluCl activator (Cully et al., 1994), blocked the response of LNvs to ACh in the absence of glutamate (Figure 5F, n = 4). These in vitro data parallel our in vivo data and support the idea that ACh released from the visual system can only fully activate LNvs in the absence of DN1 glutamatergic signals mediated via GluCl in LNvs. Taking all the larval data in Figure 1, Figure 2, Figure 3, Ku-0059436 in vitro Figure 4 and Figure 5 together, we propose the following model for rhythmic light avoidance (Figure S5).

Around dawn, low CLK/CYC activity increases LNv excitability and reduces DN1 activity. With DN1s releasing minimal glutamate, the LNvs respond strongly to ACh from the visual system and promote the dawn peak in light avoidance. Around dusk, high CLK/CYC activity reduces LNv excitability but increases DN1 activity, causing glutamate release and inhibition of the response of the LNvs to ACh via GluCl, reducing light avoidance. Thus, we propose a mechanism for the morning and evening dual oscillator model (Grima et al., 2004, Pittendrigh

and Daan, 1976 and Stoleru et al., 2004): neuronal excitability peaks in antiphase between excitatory LNvs and inhibitory DN1s to generate robust behavioral rhythms. Although adult clock neurons are more numerous and control more behaviors than their larval counterparts, we sought to test whether the principles we identified in larvae also operate in adult flies, focusing on locomotor activity rhythms in DD. Previous studies suggested that the neurons targeted by cry13-Gal4; Pdf-Gal80 are dispensable for adult DD rhythms because their ablation leaves flies rhythmic, possibly because sufficient PAK6 CRY− non-LNvs remain to support rhythms ( Stoleru et al., 2004). Therefore, we used the tim-Gal4; Pdf-Gal80 combination to target strong transgene expression to all clock neurons except LNvs, i.e., the dorsal lateral neurons (LNds) and the three groups of dorsal neurons. We also used the tim-Gal4; cry-Gal80 combination to target the non-CRY-expressing subset of adult clock neurons (DN2s and subsets of LNds, DN1s, and DN3s). tim-Gal4; Pdf-Gal80 and tim-Gal4; cry-Gal80 drivers both display robust rhythms when crossed to the dORKΔNC control transgene ( Table 1; power > 500; see Experimental Procedures for a description of power).