The logic of circadian organization in Drosophila

Abstract

Background: In the fruit fly Drosophila melanogaster, interlocked negative transcription/translation feedback loops provide the core of the circadian clock that generates rhythmic phenotypes. Although the current molecular model portrays the oscillator as cell autonomous, cross-talk among clock neurons is essential for robust cycling behavior. Nevertheless, the functional organization of the neuronal network remains obscure.

Results: Here we show that shortening or lengthening of the circadian period of locomotor activity can be obtained either by targeting different groups of clock cells with the same genetic manipulation or by challenging the same group of cells with activators and repressors of neuronal excitability.

Conclusions: Based on these observations we interpret circadian rhythmicity as an emerging property of the circadian network and we propose an initial model for its architectural design.

Graphical abstract

Figure 1

Different Groups of Neurons Affect Rhythmic Behavior (A) Operational classification of clock neurons according to the expression of the Pdf and cry promoters (see also text and Figure S1). (B) Period differences between CRYΔ-expressing and control flies. Horizontal black bars refer to the period change (in hours) compared to controls when CRYΔ (UAS-cryΔ14.6) is expressed in particular groups of neurons (shown on the left) as a result of different GAL4/GAL80 combinations (shown on the right). (C) Average locomotor activity profiles of CRYΔ-expressing flies showing 4 days in LD 12:12 and 12 days in DD. Genotypes and statistics are as reported in Table S1; control CRYΔ was w, UAS-cryΔ14.6.

Figure 2

Circadian Logic We operationally divided the circadian neurons into PDF⁺CRY⁺, PDF⁻CRY‡, PDF⁻CRY^∗, and PDF⁻CRY⁻ groups (see also Figures 1 and S1). (A) The PDF⁺CRY⁺ cells have a large influence on the network as they can communicate with a significant number of neurons through PDF (green arrows). Probably, the PDF⁻CRY‡ neurons are particularly responsive to PDF signaling (thick green arrow) as they express the PDF receptor (PDFR) at the highest level [37]. Activation of the PDF⁺CRY⁺ cells results first in a longer period of locomotor activity and then, as activation increases, in complex rhythms (Figure S3). Activation of the PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ cells results in a shorter period of locomotor activity and then in arrhythmicity. The PDF⁻CRY‡ and the PDF⁻CRY^∗ cells have an inhibitory role toward the output of the PDF⁺CRY⁺ and the PDF⁻CRY⁻ neurons, respectively, thus balancing slower and faster components in the network. This could provide a simple mechanism to achieve phase changes under different environmental conditions. (B) The PDF⁺CRY⁺ also use PDF-independent connections (black arrow) to promote the activity of the PDF⁻CRY⁻ cells (see also Figure S2). The latter exert a negative feedback on the former through activation of the PDF⁻CRY^∗ neurons. Thus, activation of PDF⁺CRY⁺ results in direct (PDF-independent) activation and in indirect (mediated and amplified by PDF and the PDF⁻CRY‡ neurons) repression of the PDF⁻CRY⁻ cells. These integrate both pathways and feedback to regulate the activity of the PDF⁺CRY⁺ neurons. This could explain how the PDF⁺CRY⁺ and PDF⁻CRY⁻ groups synchronize together. (C) Within the PDF⁻CRY⁻ group, the N neurons (including the majority of DN1s) are involved in stabilizing behavioral rhythms through synchronization with the PDF⁺CRY⁺, which suggests they might have an independent connection with those cells (broken blue line). The F neurons (which include the DN2s and perhaps unrecognized PDF⁻CRY⁻ neurons) have an intrinsically faster molecular rhythm. We assume that the signal to inhibit the PDF⁺CRY⁺ group can be passed on to the PDF⁻CRY^∗ before the PDF⁻CRY‡ cells have time to exert their repression. This would explain how the fast PDF⁻CRY⁻ counteract the slower cycling of the PDF⁺CRY⁺ neurons, resulting in a 24 hr period. However, the signal for a faster rhythm would not usually reach the N neurons before the repression from the PDF⁻CRY‡ cells takes effect, suggesting a delay in the connection between the F and N groups of PDF⁻CRY⁻ neurons. The activation of the PDF⁻CRY^∗ and PDF⁻CRY⁻ groups (or a reduction in the activation of the PDF⁻CRY‡ cells) would overcome that delay, causing the N cells (such as the DN1s) to cycle in synchrony with the F cells (such as the DN2s); see Figure 4. The N neurons would then pass on the shorter-period signal to the PDF⁺CRY⁺ group (such as the s-LNvs, Figure 4), causing a shorter behavioral period.

Figure 3

Locomotor Activity Profiles after Manipulation of Clock Neurons Average locomotor activity profiles of flies showing 3–4 days in LD 12:12 and 12 days in DD. Genotypes and statistics are reported in Table S2. (A) NaChBac overexpression, line NaChBac4 is shown. Control: w; +/ UAS-NaChBac4;+/+. (B) Kir2.1 overexpression, line Kir2.1(III) is shown. Control: w; +/+; UAS-Kir2.1/+. (C) HID, RPR overexpression. Control: w, UAS-hid,UAS-rpr; +/+;+/+. (D) CLKΔ overexpression, line CLKΔ1 is shown. Control: w; UAS-ClkΔ1/+; +/+.

Figure 4

PDP1ε Immunoreactivity after Expression of CRYΔ in Different Groups of Neurons Staining index (SI, error bars correspond to the standard error of the mean) for PDP1ε at CT0, CT6, CT12, and CT18 during day 2 (green) and 5 (magenta) in DD. ANOVA showed a significant effect (p < 0.05) of time of day on SI values, with the following exceptions. Control CRYΔ, DN3s at day 5. PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, LNds at day 5 and DN3s at day 5. PDF⁻CRY‡ ∩ PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, LNds at day 5 and DN3s at day 2 and 5. For the two CRYΔ-expressing genotypes, we also tested the interaction among factors by comparing the SI of cell types s-LNvs, DN1s, and DN2s at different time points, at day 2 and 5. (PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, ANOVA, Day, F_1,215 = 31.09, p < < 0.01; Time, F_3,215 = 24.83, p < < 0.01; Cell type, F_2,215 = 6.29, p < 0.01; Day^∗Time, F_3,215 = 18.26, p < < 0.01; Day^∗Cell type, F_2,215 = 0.88, p = 0.42; Time^∗Cell type, F_6,215 = 4.50, p < < 0.01; Day^∗Time^∗Cell type, F_6,215 = 3.50, p < 0.01. PDF⁻CRY‡ ∩ PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, ANOVA, Day, F_1,193 = 2.33, p = 0.13; Time, F_3,193 = 50.73, p < < 0.01; Cell type, F_2,193 = 36.67, p < < 0.01; Day^∗Time, F_3,193 = 2.71, p = 0.046; Day^∗Cell type, F_2,193 = 0.83, p = 0.44; Time^∗Cell type, F_6,193 = 3.06, p < 0.01; Day^∗Time^∗Cell type, F_6,193 = 0.82, p = 0.56.) PDP1ε did not cycle in l-LNvs, so they are not shown. We did not measure the PDF-negative LNv and the lateral posterior neurons (LPNs) because we were unable to identify them unequivocally in all preparations. For each cluster of neurons, we calculated the average SI per hemisphere (each considered as an independent observation) and we used those values for statistical comparisons. The number of hemispheres analyzed are given below as GENOTYPE, DAY [cell type1 (time points), cell type2 (time points), etc.]. Cell types are s-LNvs, LNds, DN1s, DN2s, DN3s, respectively. Time point are CTO, CT6, CT12, CT18, respectively. Control CRYΔ, DD2 [(9, 10, 11, 8), (9, 8, 11, 7), (10, 9, 11, 9), (10, 6, 10, 8), (10, 10, 10, 6]; DD5 [(11, 9, 12, 14), (11, 9, 11, 14), (9, 10, 9, 11), (10, 5, 9, 10), (10, 9, 10, 12)]. PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, DD2 [(11, 9, 10, 9), (11, 9, 10, 10), (12, 7, 10, 10), (11, 8, 9, 10), (11, 8, 8, 10)]; DD5 [(11, 10, 12, 9), (11, 10, 13, 11), (10, 10, 10, 11), (10, 9, 10, 11), (9, 10, 10, 8)]. PDF⁻CRY‡ ∩ PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, DD2 [(9, 13, 10, 8), (9, 10, 10, 8), (8, 7, 10, 8), (8, 5, 10, 7), (8, 7, 9, 8)]; DD5 [(11, 10, 10, 11), (12, 11, 10, 11), (9, 9, 8, 11), (9, 6, 8, 12), (10, 9, 9, 9)]. Genotypes: Control CRYΔ, w, UAS-cryΔ14.6; +/+; +/+. PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, w, UAS-cryΔ14.6; tim-GAL4/+; cry-GAL80_2e3m/+. PDF⁻CRY‡ ∩ PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRYΔ, w, UAS-cryΔ14.6; tim-GAL4/Pdf-GAL80_96a; +/+.

Figure 5

Network Perturbations Independently of Development (A and B) Expression of CRYΔ in PDF⁺CRY⁺ neurons after activation of the Geneswitch system by the drug RU486. (A) Average locomotor activity profiles (3 days LD, 8 days DD) of Geneswitch PDF⁺CRY⁺ > CRYΔ flies that were never subjected to drug treatment (DO), that were subjected to treatment as adults only (DA), or that were exposed to the drug since early development (DC). (B) The period of locomotor activity of Geneswitch PDF⁺CRY⁺ > CRYΔ flies and their parental controls were compared across the three different treatments: DO (black), DA (red), and DC (blue). ANOVA showed a nonsignificant effect of genotype (F_2,477 = 0.554, p = 0.58) but a significant effect of treatment (F_2,477 = 7.22, p < 0.01) and (asterisk) of the interaction term (genotype × treatment, F_4,477 = 2.53, p = 0.04). Post-hoc analyses revealed significant differences comparing DA (Bonferroni, p < 0.01) and DC (Bonferroni, p = 0.04) to DO but not between DA and DC (Bonferroni, p = 1). Genotypes: Geneswitch PDF⁺CRY⁺ > CRYΔ, w, UAS-cryΔ14.6; UAS-CD8GFP/+; Pdf-GS/+; Control 1, w; UAS-CD8GFP/+; Pdf-GS/+; Control 2, w, UAS-cryΔ14.6; +/+; +/+. See also Table S4. (C and D) Expression of the temperature sensitive cation channel TRPA1 in PDF⁺CRY⁺ and in PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ cells under restrictive (18°C) and permissive (28°C) temperature. (C) Average locomotor activity profiles (3 days LD, 7 days DD) of both genotypes under both conditions. (D) The period of locomotor activity of PDF⁺CRY⁺ > TRPA1 (top) and PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > TRPA1 (bottom) flies were compared, at the two temperatures (18°C, blue and 28°C, red), to their parental controls. The increase in period length for PDF⁺CRY⁺ > TRPA1 flies at higher temperature did not reach significance compared to controls as ANOVA showed a significant effect of genotype (F_2,104 = 57.69, p < < 0.01) and temperature (F_1,104 = 9.35, p < 0.01) but not of the interaction term (genotype × temperature, F_2,104 = 1.45, p = 0.24). Conversely, we observed a significant decrease (asterisk) in period length for PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > TRPA1 flies at higher temperature (ANOVA, genotype, F_2,75 = 6.22, p < 0.01; temperature, F_1,75 = 6.20, p = 0.02; genotype × temperature, F_2,75 = 6.32, p < 0.01). Genotypes: PDF⁺CRY⁺ > TRPA1, yw; Pdf-GAL4/+; +/UAS-TrpA1; Control 1, yw; Pdf-GAL4/+; +/+; Control 2 w; +/+; +/ UAS-TrpA1; PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > TRPA1 yw; tim-GAL4/+; cry-GAL80_2e3m/ UAS-TrpA1; Control 3, w; tim-GAL4/+; cry-GAL80_2e3m/+. See also Table S4.

Figure 6

Ectopic Expression of CRY Reveals a Functional Interaction with SGG (A) Average locomotor activity profiles (4 days LD, 11 days DD) of PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > SGG flies and controls. (B) Same data as in (A) but limited to the last 2 days of LD and first 2 days of DD. The profile for PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > SGG flies (blue) shows earlier anticipation of the dark-to-light transitions (blue arrows) compared to the other genotypes (PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ Control, red and SGG Control, green). (C) Average locomotor activity profiles (4 days LD, 11 days DD) of PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRY, CRY Control and PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > SGG, CRY flies. Only the latter genotype showed a shorter period of locomotor activity (see Table S5). Genotypes: PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > SGG, w, UAS-sgg; tim-GAL4/+; cry-GAL80_2e3m/ +; SGG Control, w, UAS-sgg; +/+; +/+; PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ Control, w; tim-GAL4/+; cry-GAL80_2e3m/+; PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > CRY, w; tim-GAL4/+; cry-GAL80_2e3m/UAS-HAcry16.1; CRY Control, w; +/+; +/UAS-HAcry16.1; PDF⁻CRY^∗ ∩ PDF⁻CRY⁻ > SGG, CRY, w, UAS-sgg; tim-GAL4/+; cry-GAL80_2e3m/UAS-HAcry16.1. See also Table S5.