Shift Work Disrupts Circadian Regulation of the Transcriptome in Hospital Nurses

Abstract

Circadian misalignment between sleep and behavioral/feeding rhythms is thought to lead to various health impairments in shift workers. Therefore, we investigated how shift work leads to genome-wide circadian dysregulation in hospital nurses. Female nurses from the University of Alabama at Birmingham (UAB) Hospital working night shift ( n = 9; 29.6 ± 11.4 y) and day shift ( n = 8; 34.9 ± 9.4 y) participated in a 9-day study measuring locomotor activity and core body temperature (CBT) continuously. Additionally, cortisol and melatonin were assayed and peripheral blood mononuclear cells (PBMCs) were harvested for RNA extraction every 3 h on a day off from work. We saw phase desynchrony of core body temperature, peak cortisol, and dim light melatonin onset in individual night-shift subjects compared with day-shift subjects. This variability was evident even though day- and night-shift nurses had similar sleep timing and scheduled meal times on days off. Surprisingly, the phase and rhythmicity of the expression of the clock gene, PER1, in PBMCs were similar for day-shift and night-shift subjects. Genome-wide microarray analysis of PBMCs from a subset of nurses revealed distinct gene expression patterns between night-shift and day-shift subjects. Enrichment analysis showed that day-shift subjects expressed pathways involved in generic transcription and regulation of signal transduction, whereas night-shift subjects expressed pathways such as RNA polymerase I promoter opening, the matrisome, and endocytosis. In addition, there was large variability in the number of rhythmic transcripts among subjects, regardless of shift type. Interestingly, the amplitude of the CBT rhythm appeared to be more consistent with the number of cycling transcripts for each of the 6 subjects than was melatonin rhythm. In summary, we show that shift-work patterns affect circadian alignment and gene expression in PBMCs.

Keywords: chronobiology; circadian disruption; clock genes; endocrinology; microarray; sleep.

Figure 1.

Shift work schedule, and rhythmic indicators of the central circadian clock (A) Shift work schedule of UAB nurses enrolled in the study. Ambulatory activity and sleep from actigraphy and CBT were monitored continuously for 10 days. The shift cycle (indicated in gray) consisted of three, consecutive 12-h shifts. On the last study day 9, participants spent 24-h in the UAB Clinical Research Unit (CRU) for continuous blood draws. (B) Phase clustering of CBT minimum on day 9 for individual day-shift (left) and night-shift (right) nurses plotted against time (in decimal hours). Hourly averages of CBT for individual nurses are shown in Figure S2. (C, E) Average (± SEM) normalized melatonin (C) and cortisol (E) rhythms on study day 9 in day-shift (open circles) and night-shift (closed circles) nurses across time of day (in decimal hours). (D, F) Phase clustering of dim light melatonin onset (DLMO; panel D) and peak cortisol (F) for day-shift (light gray circles) and night-shift (dark gray circles) nurses. Rayleigh test confidence intervals are indicated by lines paralleling circle circumference where clustering of both DLMO and cortisol peak for day-shift nurses indicated statistical significance (P < 0.001) while confidence intervals for night shift nurses were not significant (P > 0.05) for either cortisol or DLMO.

Figure 2.

Circadian transcripts and enriched pathways in day- and night-shift groups (A) Venn diagram shows MetaCycle’s meta3d reported rhythmic transcripts in day-shift (blue/light gray) and night-shift (red/dark gray) groups with the cut-offs of rAMP > 0.1 and P-value < 0.05 (left) and P-value < 0.1 (right). (B) Phase distribution of circadian transcripts in day (blue/light gray; P-value < 0.05) and night (red/dark gray; P-value < 0.05) shift groups. (C) Significantly enriched pathways (q < 0.05) from PSEA analysis of circadian genes in the Day-shift and Night-shift groups, coded by the letters indicated in the circular plot above each list. Note that 6 am and 6 pm are general markers for day and night, respectively, and that the data were collected at various times of the year. (D) The phase distribution of rhythmic genes is shown for two example pathways.

Figure 3.

Canonical clock gene expression in PBMC Quantitative real-time RT-PCR analysis of the 24-h pattern of average transcription of PER1 (panel A), PER2 (panel B), PER3 (panel C), BMAL1 (panel D), and REV-ERBα (panel E) at each time point, analyzing the entire sample pool (n = 9 and 8 for night-shift and day-shift, respectively). Error bars are SEM.

Figure 4.

Phase distributions of circadian transcripts in each individual Vector histograms of gene expression (left) and melatonin rhythms (right) in three, individual day-shift (A) and three, individual night-shift nurses (B). Each ring represents 75 transcripts, and arrows indicate the average peak time in rhythmic expression (as indicated by the direction). For the line graphs, melatonin is plotted with circles using the left Y-axis and core body temperature (CBT) is plotted as triangles using the right Y-axis. Dim light melatonin onset (DLMO) is indicated by the vertical dashed line in each graph. Cosinor amplitude estimates for CBT are as follows (in °C, top to bottom): panel A (day-shift), 0.269, 0.673, and 0.107; panel B (night-shift), 0.272, 0.422, and 0.114.