Fig 1.
Identification of metabolite rhythms at baseline and during the night shift condition.
(A) Heatmap of metabolites identified as rhythmic by mixed-effects cosinor analysis (FDR < 0.05) at baseline (left) and during the night shift condition (right). Metabolites that are rhythmic in both conditions are shown in bold-italic. (B) Number of rhythmic metabolites across the continuum of FDR cutoffs. Dotted vertical line represents an FDR cutoff of 0.05. (C) Venn diagram depicting the number of metabolites identified as significantly rhythmic at baseline, during the night shift condition, and the overlap. (D, E) Histogram displaying the distribution of q-values obtained by mixed-effects cosinor analysis on (D) the 19 metabolites that were rhythmic at baseline but not during the night shift condition and (E) the 21 metabolites that were rhythmic during the night shift condition but not at baseline. (F) Phase distribution of metabolites identified as rhythmic at baseline (left) and during circadian misalignment (right). Shaded areas represent the scheduled sleep periods in the two conditions. Numerical data underlying the results presented in this figure are available in S2 Data. FDR, false discovery rate; TMAO, trimethylamine N-oxide.
Fig 2.
Identification of circadian- and behaviour-influenced metabolites.
(A, B) Examples of metabolites that are influenced by the circadian cycle (A: tryptophan and methionine sulfoxide) and that are influenced by the behavioural cycle (B: hexanoylcarnitine [C6] and alanine). Data are presented as mean ± SEM. Relative clock time reflects the time since the start of the individual habitual rest period during baseline. (C) Phase shift during the night shift condition relative to baseline of the 32 metabolites that are significantly rhythmic at both baseline and during the night shift condition. Symbols represent whether the metabolites were classified as circadian-influenced, behaviour-influenced, or neither. The direction of the arrow represents the average phase shift of the behaviour-influenced metabolites. Length of the arrow represents the mean resultant vector length, a measure of the spread of the circular data. No arrow is shown for the circadian-influenced metabolites, because the phase shift of the metabolites was not nonuniformly distributed (p = 0.16, Rayleigh test). (D) Amplitude of the 32 metabolites that were identified as rhythmic during baseline and the night shift condition. Numerical data underlying the results presented in this figure are available in S2 Data. C6, hexanoylcarnitine.
Fig 3.
Individual variability in metabolite rhythms.
(A) Number of significantly rhythmic metabolites (corrected p < 0.05, individual cosinor analysis) per metabolite class and per subject at baseline (left panel) and in the night shift condition (right panel). (B) Change in number of rhythmic metabolites at baseline and during the night shift condition. Different lines and colours represent different subjects. (C) Percentage of rhythmic metabolites that overlap between each pair of subjects. (D, F) Rhythmic alanine (D) and LysoC28:1 (F) profiles per subject at baseline and the night shift condition. Symbols on top of each panel indicate the phase of the rhythm. Different colours represent different subjects. Only rhythmic profiles (uncorrected p < 0.05) are shown. (E, G) Initial and final phases of alanine (E) and LysoC28:1 (G) profiles per subject. Colours, representing different subjects, matched with the colours in B, D, and F. (H) Circular variance of individual phase estimates per metabolite at baseline and during the night shift condition. Circular variance ranges from 0 to 1, with lower values indicating tighter clustering of the phase estimates around the mean and higher values indicating a higher degree of dispersion around the 24-h cycle. Only metabolites that were rhythmic (uncorrected p-value <0.05) in at least 3 subjects were taken into account. Numerical data underlying the results presented in this figure are available in S2 Data. LysoC, lysophospholipid.
Fig 4.
Phase shifts of rhythmic metabolites per subject.
Panels show the data for each of the 9 subjects. Dots represent the phase shifts of metabolites that are rhythmic (uncorrected p < 0.05) both at baseline and during the night shift schedule in that subject. By convention, a negative phase shift denotes a phase delay; a positive phase shift denotes a phase advance. The p-value for the significance of the mean direction of the phase shift (Rayleigh test) is shown in each panel, as well as the average phase shift (Δφ) if the Rayleigh test p-value is <0.05. Arrow length represents the resultant vector length, an indication of the strength of the mean direction. The colour of the dots represents the metabolite class, as indicated in the legend. Numerical data underlying the results presented in this figure are available in S2 Data.