JmjC domain proteins modulate circadian behaviors and sleep in Drosophila

Abstract

Jumonji (JmjC) domain proteins are known regulators of gene expression and chromatin organization by way of histone demethylation. Chromatin modification and remodeling provides a means to modulate the activity of large numbers of genes, but the importance of this class of predicted histone-modifying enzymes for different aspects of post-developmental processes remains poorly understood. Here we test the function of all 11 non-lethal members in the regulation of circadian rhythms and sleep. We find loss of every Drosophila JmjC gene affects different aspects of circadian behavior and sleep in a specific manner. Together these findings suggest that the majority of JmjC proteins function as regulators of behavior, rather than controlling essential developmental programs.

Figure 1

Circadian rhythm phenotypes of JmjC genes. (A) KDM2^KO and JMJD5^KO flies showed mildly reduced period lengths (estimated effect sizes compared to wild type at −0.74 ± 0.39 95% confidence interval for KDM2^KO, and −0.40 ± 0.33 for JMJD5^KO respectively). Note, these effect sizes were estimated assuming normal distributions, which these data were not. Thus, all bar graphs shown in this and following Figures are medians with 95% confidence interval error bars. (Kruskal-Wallis test for multiple comparisons with Dunn’s post hoc adjustment to detect significant differences). (B) KDM2^KO and JMJD5^KO flies also showed increased rhythm power compared to w Berlin control flies. (*p < 0.05, **p < 0.01, ***p < 0.001; effect sizes = 0.89 ± 0.40 for KDM2^KO, 0.77 ± 0.34 for JMJD5^KO).

Figure 2

Sleep (and activity) phenotypes of JMJD5^KO and JMJD7^KO flies. (A,B) JMJD5^KO showed moderately reduced daytime sleep (effect size = −0.68 ± 0.40) concomitant with (B) strongly increased daytime activity (effect size = 0.97 ± 40), while nighttime levels of sleep and activity were unaffected (n > 28). In this, and the next two Figures, data from w Berlin controls are in grey. The daytime phenotypes are rescued to wild-type levels for sleep, and beyond wild type for activity. (C) The JMJD5::HA rescue construct (right two brain hemispheres) was expressed in neurons close to the dorsal lateral neurons (arrow). Brains were stained with anti-HA (white) and a negative control is shown at the left. (D) JMJD7^KO strongly reduced daytime sleep (effect size = −1.45 ± 0.33 confidence interval), which was partially rescued towards wild-type by expression of the JMJD7::HA transgene (n > 61). The mutant also mildly affected nighttime sleep (effect size = 0.55 ± 0.30). (E) The JMJD7^KO mutants also displayed a moderate nighttime activity loss phenotype (effect size = −0.78 ± 0.31). (F) The expression pattern (green = anti-HA) of JMJD7::HA genomic rescue construct is shown (red = nuclei). Arrowheads point to the pars intercerebralis and arrows point to the fan shaped body, magnified in inset.

Figure 3

NO66^KO hyperactivity phenotype. (A) NO66^KO flies showed a reduction in sleep, strong for the daytime (effect size = 1.10 ± 0.40) and moderate for the night (effect size = 0.66 ± 0.39). This was completely rescued to wild type by expression of the NO66::HA genomic transgene (n > 31). (B,C) Flies also showed a very strong increase in activity, shown for one example fly (B; individual flies shown in this, and the next Figure have values within half a standard deviation from the mean of that genotypic cohort) and the whole cohort. (C) The effect size was 1.69 ± 0.43 for daytime activity, and 1.15 ± 0.40 for the night. These phenotypes were also completely rescued by NO66::HA expression. (D) The hyperactivity phenotype was in part driven by a strong increase in the activity per waking hour (effect size = 1.10 ± 0.40 for daytime waking activity, and 0.66 ± 0.39 at night). The phenotype in daytime waking activity was rescued by NO66::HA expression, but rescue flies were no different from mutants or wild type for nighttime waking activity. (E–G) The hyperactivity phenotype was also evident in the total starvation activity, which was very strongly enhanced (effect size = 1.81 ± 0.35) and partly rescued towards wild type (n > 47). (E,F) show a single fly example, and (G) the whole cohort. (H) This increased hyperactivity does not cause premature death; NO66^KO flies even showed a mild delay in starvation-induced death (effect size = 0.41 ± 0.32), which was also rescued by the NO66::HA transgene (n > 31). (I) The expression pattern (green = anti-HA) of NO66::HA genomic rescue construct is shown (red = nuclei). Arrowheads point to the pars intercerebralis and arrows point to the fan shaped body, magnified in inset.

Figure 4

KDM4B^KO sleep phenotype. (A–H) Shown are mutant (light blue) and rescue (dark) medians with 95% confidence interval. Averages of double-plotted daily sleep of one representative fly (per genotype, A) and the two cohorts (B) show a strong increase in sleep time in KDM4B^KO (effect size = 1.89 ± 0.46 for daytime sleep, effect size = 0.78 ± 0.42 for night sleep). Average daily activity of one fly each (C) and the two cohorts (D) indicates a concomitant strong reduction in activity (effect size = −1.79 ± 0.45 for daytime activity, effect size = −0.82 ± 0.42 for night sleep). The daytime sleep and activity phenotypes were rescued to wild type by expressing KDM4B::HA, while nighttime sleep and activity were rescued beyond wild-type values. (E) The waking activity was not affected in KDM4B^KO mutants compared to wild type (effect sizes = −0.14 ± 0.41 day, −0.41 ± 0.42 night). Individual fly actograms (F,G) show starvation-induced hyperactivity on non-nutritious agar in rescue flies (G), but not in KDM4B^KO mutants (F). As a group (H), KDM4B^KO flies showed absence of starvation hyperactivity (effect size = −1.22 ± 0.39) but no effect on death (I). Note that in numerous assays shown here, KDM4B::HA rescue flies showed opposite phenotypes from KDM4B^KO mutants (as in B,D,H,I; n >31). (J) The expression pattern of KDM4B::HA genomic rescue construct is shown in (green = anti-HA, red = nuclei) and includes the pars intercerebralis (arrowheads) and fan shaped body (arrows).

Figure 5

Summary of circadian behavior and sleep phenotypes of JmjC^KO mutants. (A) Circadian rhythm, and sleep phenotype strengths of JmjC^KO strains. Phenotype strength (Herge’s g effect size) was determined assuming normal distributions and incorporating data from the primary screen and from follow-ups (Figs 2–4). The phenotype strength is shown as the z-score and percentile in an idealized wild-type distribution (color coded, as in Fig. 5B). Note that the means and standard deviations were calculated for each measure in a large w Berlin population (see Supplemental Table 1), but they are idealized, as some measures were not normally distributed (eg. w Berlin total starvation activity). The point is to illustrate the strength of each JmjC mutant’s strongest phenotype, and that few phenotypes are outside the 10th, or 90th percentiles, corresponding to an effect size of >1.3. This is a reflection of i) the lack of “strong” phenotypes, and ii) the considerable variability of the wild-type behavior. (B) Phenotype table indicating the behaviors of JmjC mutants significantly different from wild type (*p < 0.05, **p < 0.01, ***p < 0.001; Kruskal-Wallis test with Dunn’s post hoc multiple comparison). The strength of those phenotypes is color coded (according to the percentile within the wild-type distribution; see A). Hierarchical clustering dendrograms of these scores are depicted on the left, grouping JmjC^KO mutants, and at the bottom, grouping phenotypic measures. (A/min stands for waking activity per minute; tau for the circadian period length; Pwr. for the power of the rhythm; AR/R for the frequency of arrhythmic flies; TSA for total starvation-induced hyperactivity; and TOD for time of death starting from removal from food).