ALAN advances timing of activity and BMAL1 expression
Daily cycles of activity were strongly affected by the ALAN treatment (GAMM, p = 0.001, Fig. 2A and Fig. S2; Table S4). In the 5 lux group birds were generally active 6–7 h before lights-on, whereas birds in the other two light treatments (0.5 and 1.5 lux) advanced morning activity to a much lesser extent. Because of the advanced onset of activity, 40% of the overall diel activity in the 5 lux group occurred during the night, compared to 11 and 14% in the 0.5 and 1.5 lux groups, and less than 1% in the control dark group. Thus, with increasing ALAN, nocturnal activity also increased (LMM, treatment p < 0.001, Fig. 2A and Table S5).
Activity timing is affected by intensity of ALAN. (A) Shows the proportion of active 2-min intervals in each treatment group per hour (raw mean ± SEM, N = 34). Grey background indicates night-time, white background indicates daytime. (B) and (C) show daily treatment group data (mean ± SEM), for the timing of evening offset and morning onset of activity, respectively (time in min). Activity onset and offset refer to times of lights-on and lights-off, which are shown as horizontal lines crossing zero.
Breaking down this average diel profile (Fig. 2A) by time since first exposure to ALAN (i.e., days from start of the experiment to first sampling, days 0 to 18) shows how the differences in activity timing developed (Fig. 2B,C). Upon exposure to ALAN, the birds’ activity onset (Fig. 2C) advanced in all treatment groups. The groups with intermediate light exposure (0.5 lux, 1.5 lux) responded by a similar instantaneous phase-advance (155 and 142 min, respectively, p > 0.1 for pairwise comparison), and thereafter their timing remained stable. The group exposed to 5 lux showed a much larger instantaneous phase advance of almost five hours (mean ± SEM = 289 ± 21 min), and thereafter continued to gradually phase-advance until reaching a stable phase after 10 days (interaction treatment × day, p < 0.001, Fig. 2C, Table S5). The advance until stabilization could represent gradual entrainment to an early phase. Equally, it could represent temporary free-run of activity under the reduced Zeitgeber amplitude (note that birds were not kept in constant light conditions), as suggested by periodogram analysis. Indeed, we found that in the 5 lux group, prior to stabilization, period length deviated from that of all other groups and from 24 h (mean period length 5 lux group: 23.6 h; LM; Table S6). The individual actograms (Fig. S3) further suggest that the activity rhythm in the 5 lux group may have split into an advancing morning component whereas evening activity remained largely stable.
Changes in the activity offset were much less pronounced (Fig. 2B). The 5 lux group showed an instantaneous delay phase-shift. This initial delay was followed by a gradual advance, similar to but smaller than that of morning onset. At the end of the experiment birds in the 5 lux group ceased their activity before lights-off, and earlier than other groups (treatment × day, p < 0.001, Fig. 2B, Table S5). This advance did not compensate for the earlier onset, as birds in the 5 lux group were more active over the whole 24 h than the remaining birds (treatment × day, p = 0.01, Table S5).
Hypothalamic BMAL1 expression at night parallels advanced activity onset
We next sought to identify whether the profound shifts in activity patterns were paralleled by corresponding shifts in the pacemaker, measured by expression of BMAL1 in the hypothalamus. Day–night differences in transcripts of BMAL1 inverted with increasing ALAN (Fig. S4A), as predicted above (Fig. 1). While BMAL1 expression was higher at midnight than at mid-day for the control birds, increasing ALAN induced a reversal of this pattern, so that birds in the 5 lux group had much higher expression at mid-day than at midnight (treatment × time, p < 0.01, Table S7).
To assess whether changes in day–night BMAL1 gene expression correlated with temporal behavioural shifts, we related BMAL1 levels to onset of activity of an individual once it had stably shifted in response to the ALAN treatment (Fig. 2B,C, after 10 days). Onset was closely predicted by hypothalamic BMAL1 expression at midnight (Gaussian LM, p < 0.001, R2 = 0.71, Fig. 3A). Across ALAN levels, the earliest rising birds had the lowest midnight expression of BMAL1. However, the steep linear regression was largely based on differences between ALAN groups in both activity timing (Figs. 2, 3) and BMAL1 expression (Fig. S4A). Indeed, this relationship was even stronger when we only considered the 0.5, 1.5 and 5 lux group in the analysis (LM p < 0.001, R2 = 0.85), but the association was not present for the dark control birds (LM, p = 0.87). Individual midnight BMAL1 levels also predicted mean offset of activity, albeit less strongly so than onset (LM, p = 0.006, R2 = 0.28, Fig. 3B). Conversely, mid-day BMAL1 levels did not significantly predict variation in either activity traits (LMs, p > 0.1 and R2 < 0.16 for all measures).
BMAL1 expression in the hypothalamus predicts the advance of morning activity. mRNA levels of BMAL1 at midnight correlated with the onset (A) and offset of activity (B), but mid-day levels did not (C and D). Shown are log-transformed mRNA levels, separated by ALAN treatments (blue colour gradient). Points represent individual birds, lines and shaded areas represent model fits ± 95% confidence intervals.
ALAN reverses day–night BMAL1 expression patterns in multiple tissues
ALAN-induced shifts in BMAL1, as detected in the hypothalamus, were remarkably consistent across tissues (Fig. 5A–D). Hippocampal BMAL1 expression profiles resembled those in the hypothalamus (Fig. S5A; interaction of treatment and sampling time p < 0.001, Table S8). Within individuals, mid-day and midnight transcripts in both brain tissues were closely related (LM, p < 0.001, Fig. 4A, Table S9). Liver BMAL1 showed similar effects of ALAN on day–night expression profiles (Fig. S6A; time × treatment, p < 0.001, Table S10), so that within individuals, hepatic and hypothalamic transcripts also correlated closely (LM, p < 0.001, Fig. 4B, Table S9). Similar ALAN effects on BMAL1 expression were found also in the spleen (Fig. S7A; time × treatment, p = 0.003, Table S11), and individual-level transcripts closely correlated with those in hypothalamus (LM, p = 0.011, Fig. 4C) and liver (LM, p = 0.001, Fig. 4D, Table S9).
ALAN effects on BMAL1 expression were comparable in different tissues. Correlation of expression patterns of BMAL1 in different tissues. Shown are log-transformed mRNA levels, separated by sampling time (day vs night) and ALAN treatments (blue colour gradient). Points represent individual birds. Lines and shaded areas depict model estimated means ± 95% confidence intervals. Panels show expression levels of BMAL1 in hypothalamus in relation to (A) hippocampus, (B) liver and (C) spleen, as well as spleen in relation to liver (D).
Inconsistent shifts of expression patterns by ALAN in other genes
We then assessed whether the reversal of day–night expression patterns found for BMAL1 was paralleled in other genes (Table S1). Our analysis revealed that different pathways were differentially affected by ALAN.
Among clock-related genes, hypothalamic expression levels of CK1ε were not affected by the light treatment (p = 0.71, Table S7). Expression was consistently, although not significantly, higher at mid-day (p = 0.09, Fig. 5H, Table S7). Expression of hepatic CK1ε slightly increased with light intensity (p = 0.078, Fig. 5P, Table S10), and was not affected by sampling time (p = 0.13, Table S10). In the liver CRY1 showed no expression trend that aligned with that of BMAL1 and was not affected by treatment or sampling time (p > 0.6 for both variables, Fig. 5O, Table S10).
ALAN effects on gene expression are gene-specific. ALAN does not equally affect all physiological systems. ALAN effects on BMAL1 (A–D) were paralleled by those on three additional genes in the hypothalamus (SIRT1), liver (NRF1) and spleen (TLR4) (E–G), but not by other genes analysed across tissues (H–Q). Shown are log-transformed mRNA levels, separated by sampling time (mid-day: yellow; midnight: dark blue). Large symbols ± SEM connected by lines represent model estimates, whereas small symbols depict raw data points (N = 34 birds).
Among metabolic genes, patterns similar to those in BMAL1 were evident in SIRT1, a gene which is also involved in the modulation of the circadian cycle26. Hypothalamic SIRT1 showed a clear change of day–night expression with increasing ALAN (Fig. 5E; treatment × time, p = 0.029, Table S7), and SIRT1 levels were closely related to those of hypothalamic BMAL1 (LM, p < 0.001, Table S9). In the liver, the metabolic gene NRF1 showed a similar response to ALAN as BMAL1, with reversed day–night expression in the 5 lux group compared to other groups (treatment × time, p < 0.001, Fig. 5F, Table S10), and close correlation with BMAL1 (LM, p < 0.001). In contrast, another hepatic metabolic gene, IGF1, was not significantly affected by light treatment or sampling time (for both, p > 0.11, Fig. 5Q, Table S10). In the hippocampus (Table S8), mid-day and midnight levels of the mineralocorticoid receptor, MR, decreased significantly with increasing ALAN (p = 0.044, Fig. 5M). Conversely, the levels of the glucocorticoid receptor, GR, showed no significant relationship with either light treatment or sampling time (p > 0.33 in both cases, Fig. 5N).
Among immune genes, hypothalamic LY86 levels decreased with increasing ALAN (p = 0.04, Fig. 5K, Table S7), but the same gene in the spleen was not affected by either treatment or sampling time (p > 0.7, Fig. 5L, Table S11). Conversely, another immune gene in the spleen, TLR4, showed the same pattern as BMAL1 (Fig. 5G, time × treatment, p = 0.006, Table S11).
Last, we also analysed genes involved in photoperiodic response in the avian brain. FOXP2, a gene that in birds is involved in learning, song development and photoperiod-dependent seasonal brain growth, showed no significant trends related to ALAN or sampling time (p > 0.32 in both cases, Fig. 5J). DIO2, a thyroid-axis gene involved in photoperiodic reproductive activation, was also not affected by either ALAN or sampling time (p > 0.45 for both variables, Fig. 5I).
Metabolomic profiles support inconsistent reversal of day–night physiology under ALAN
Untargeted LC–MS metabolomic analysis, after filtering, provided abundance values for 755 metabolites, which we tested for effects of ALAN and sampling time by individual linear mixed models (correcting for false discovery rate at 5%). We found that 44.1% (333) differed significantly by sampling time, with higher levels at mid-day in 197, and higher levels at midnight in 136 (all metabolite tables: https://doi.org/10.6084/m9.figshare.12927539.v1). 29 metabolites differed significantly by treatment (Table S12), whereby levels decreased with ALAN in 11 metabolites and increased in 18 metabolites. Finally, 73 (9.7%) of the 755 metabolites showed significant interaction between treatment and sampling time (Fig. 6 and Table S13; 34 of those also differed by sampling time). As this pattern supported reversal of day–night physiology similar to that shown for BMAL1 expression, these metabolites were selected for subsequent focal analyses (hereafter named “interactive dataset”).
Metabolomics analysis supports ALAN-induced shifts in day–night physiology. The 73 metabolites found to be significantly affected by the interaction of treatment and sampling time (interactive dataset) were dissected by pathway analysis and PCA. Heatmaps show the top-25 metabolites at either mid-day (A) or mid-night (B). Heatmaps were created using the Metaboanalyst software27. PCA showed considerable overlap between ALAN groups at mid-day (C), whereas at midnight, ALAN treatment effects were pronounced, particularly for the 5 lux group (D). In all PCA plots, points represent individual samples, and ellipses contain 80% of samples in a group. The first PC of the night cluster significantly predicted the onset of activity (E), but not the offset of activity in the evening (F). In (E) and (F) points represent individual birds, and lines and shaded areas represent model fits ± 95% confidence intervals.
We dissected variation in the interactive dataset by using separate principal component analyses (PCA) on the samples collected at mid-day and midnight (Fig. 6C,D). For mid-day samples, ALAN treatments overlapped considerably (Fig. 6C), but some birds in the 1.5 lux and 5 lux treatments were separated by PC1 (26% of variance explained). PC1 in the mid-day dataset was heavily loaded with metabolites of Arginine biosynthesis pathway, including l-Arginine, Homoarginine and l-Glutamate, and with other important amino acids such as l-Threonine, l-Lysine and l-Tyrosine. Conversely, the midnight samples (Fig. 6D) separated clearly between the 5 lux treatment and the remaining groups. In this midnight PCA, PC1 explained 27% of the variance and was heavily loaded with metabolites of the Glutamate and Arginine pathways, as well as with N-acetyl-l-aspartate. PC2, which explained 21% of variation, was heavily loaded with fatty acids, including Linoleate (factor loading tables: https://doi.org/10.6084/m9.figshare.12927536.v1). The contribution of the Arginine pathway was further confirmed by pathway analysis, conducted with Metaboanalyst27, which indicated “Arginine biosynthesis” as highly significant (p < 0.001). “Aminoacyl-tRNA metabolism” (p < 0.001), “Histidine metabolism” (p = 0.005), and “Alanine, Aspartate and Glutamate metabolism” (p = 0.026) were also indicated as significant pathways.
We finally investigated whether, just like midnight levels of BMAL1 expression (Fig. 4), midnight principal components of metabolites correlated with individual activity timing. PC1 strongly predicted the onset of activity via a linear and quadratic relationship (n = 19, plinear = 0.007, pquadratic = 0.014, R2 = 0.92, Fig. 6E), but did not explain offset of activity (p = 0.63, R2 = 0.04, Fig. 6F). PC2 was related to neither timing trait (p > 0.2).
Source: Ecology - nature.com
