Pregnancy induces an earlier chronotype in both mice and women

Abstract

Daily rhythms generated by endogenous circadian mechanisms and synchronized to the light-dark cycle have been implicated in the timing of birth in a wide variety of species. Although chronodisruption (e.g. shift work or clock gene mutations) is associated with poor reproductive outcomes, little is known about circadian timing during pregnancy. This study tested whether daily rhythms change during full term pregnancies in mice and women. We compared running wheel activity continuously in both non-pregnant (n=14) and pregnant (n=13) 12–24 week-old C57BL/6NJ mice. We also monitored wrist actigraphy in women (N=39) for two weeks before conception and then throughout pregnancy and measured daily times of sleep onset. We found that on the third day of pregnancy, mice shift their activity to an earlier time compared to non-pregnant dams. Their time of daily activity onset was maximally advanced by almost four hours around day 7 of pregnancy and then shifted back to the non-pregnant state approximately one week before delivery. Mice also showed reduced levels of locomotor activity during their last week of pregnancy. Similarly, in women the timing of sleep onset was earlier during the first and second trimesters (gestational weeks 4–13 and 14–27) than before pregnancy and returned to the pre-pregnant state during the third trimester (weeks 28 until delivery). Women also showed reduced levels of locomotor activity throughout pregnancy. These results indicate that pregnancy induces changes in daily rhythms, altering both time of onset and amount of activity. These changes are conserved between mice and women.

Introduction

In many organisms, including humans, reproduction is influenced by the circadian system. Circadian clocks in the brain and tissues generate daily rhythms in sleep/wake, metabolism, hormone secretion, locomotor activity, and many other behaviors. These clocks synchronize (entrain) to daily environmental stimuli such as the light-dark cycle (Takahashi et al. 2008). For example, normal pregnancy requires a functional circadian system, (Rosenwasser et al. 1987; Germain et al. 1993; Honnebier & Nathanielsz 1994; Scribner & Wynne-Edwards 1994; Ducsay 1996; McElhinny et al. 1997; McGregor et al. 1999; Gonzalez-Mariscal et al. 2013; Miller & Takahashi 2014; Perez et al. 2015) and circadian disruption can perturb pregnancy outcomes (Mahoney 2010; Bonzini et al. 2011; Lin et al. 2011; Gamble et al. 2013; Amaral et al. 2014; Vilches et al. 2014; Fernandez et al. 2016). Additionally, in women, spontaneous membrane rupture and the onset of labor most commonly occur between midnight and 4:00 a.m. (Cooperstock et al. 1987). Similarly, rodents are most likely to deliver pups around dawn (Reppert et al. 1987; Viswanathan & Davis 1992). For example, mice with mutations in the circadian genes Clock, Period 1, or Period 2 exhibit altered daily rhythms and can have reduced fecundity, delayed and reduced embryo implantation (Miller et al. 2004), increased embryo resorption (equivalent to miscarriage), decreased litter sizes, and labor dystocia (Pilorz & Steinlechner 2008). In pregnant rats, ablation of the suprachiasmatic nucleus, the master regulator of circadian rhythms, abolishes the preferential clustering of pup deliveries around dawn (Reppert & Schwartz 1983; Reppert et al. 1987). Finally, women who work at night or on rotating shifts are 60% more likely to miscarry or deliver preterm than women who work day shifts (Knutsson 2003; NHS Plus 2009; Bonzini et al. 2011; Lin et al. 2011). To better understand the mechanisms responsible for abnormal pregnancy outcomes following circadian disruption, we first need to understand normal maternal circadian rhythms during pregnancy.

An important aspect of circadian regulation is an individual’s chronotype, the stable phase relationship (in hours) between daily biological and environmental events (Roenneberg et al. 2007). Those who rise early and have an early midpoint of sleep (“larks”) have an early chronotype, whereas those who rise late and have a late midpoint of sleep (“owls”) have a late chronotype. Chronotype depends on age, sex, and genetics (Roenneberg et al. 2004; Patke et al. 2017; Chong 2018), and misalignment between an individual’s chronotype and the external environment (chronodisruption or circadian misalignment) can negatively affect physiology and behavior (Reppert & Weaver 2002; Erren & Reiter 2009; Johnston et al. 2016). For example, chronodisruption can result when an individual with a late chronotype is forced to wake early.

In addition to being affected by the circadian system, pregnancy appears to influence the circadian regulation of behaviors, as pregnant women often report and are treated for sleep disturbances (reviewed in (Lee 1998; Sahota et al. 2003; McLafferty et al. 2018)). Indeed, such daily disturbances have been detected in women of all ages, parities, races/ethnicities, and income levels and across all months of pregnancy (Mindell et al. 2015; Reid et al. 2017). Additionally, pregnancy has variable effects on the time of peak daily expression of some clock genes in the rat brain (Schrader et al. 2010, 2011). However, whether pregnancy affects chronotype has not been addressed. Here, we examined the impact of normal pregnancy on circadian rhythms in both mice and women.

Materials and Methods

Mice, housing, and mating.

All mice (C57BL/6NJ, age 3–6 months, Jackson Laboratories, Bar Harbor, ME) were maintained individually in a cage with a running wheel on a 12h:12h light:dark cycle (2.4 ± 0.5 × 10¹⁸ photons/s*m²; General Electric). Food and water were available at all times in the temperature- and humidity-controlled Danforth Animal Facility at Washington University in St. Louis. All animal procedures were approved by the Washington University Institutional Animal Care and Use Committee and followed NIH guidelines. Females were paired with a male for two hours; those that developed a copulatory plug were considered pregnant. We used an infrared camera (Dlink®, Taipei, Taiwan) to determine the time of delivery of the first pup. Mice that completed their pregnancies were included in the analysis.

Locomotor activity in mice.

We assessed locomotor activity of non-pregnant and pregnant females by measuring wheel running data in 6-minute bins (Clocklab, Actimetrics, Evanston, IL). Daily onset of activity was defined as the first time when activity was counted for at least 1 hour after at least 4 hours of inactivity (Clocklab software, Actimetrics). Chronotype data were statistically analyzed in Prism (GraphPad ®7.02,La Jolla, CA) by using either an independent Student t-test, one-way ANOVA, or repeated measures two-way ANOVA as appropriate. ANOVA results are reported as F_a,b,, where F is the Fisher statistic and a and b are the degrees of freedom between groups and within groups, respectively.

Human subject recruitment.

Institutional Review Board approval was obtained from the Washington University School of Medicine Human Research Protection Office (IRB 201410060, approved October 20, 2014). Women planning pregnancy were recruited from the local community (N=39, Table 1). Women were included if they were at least 18 years of age, preparing to conceive, and willing to wear an actigraphy monitor. Women were excluded if they had multiple gestations, had undergone in vitro fertilization, had a uterine anomaly, used hypnotics/sedatives such as benzodiazepines, were non-English speaking or were shift workers. After signing an informed consent document, enrolled women wore activity monitors for at least two weeks while trying to conceive (pre-pregnancy). Once conception was documented, women wore activity monitors throughout pregnancy until delivery. Women were included in analysis if they had a term delivery with continuous actigraphy data collected for at least one week during each trimester.

Table 1:

Demographic Information of Study Participants

Maternal characteristics	(N=39)
Age	31.3	(4.0)
BMI (kg/m2)	25.4	(5.5)
Ethnicity, n (%)
Non-Hispanic white	32	(82.1%)
Non-Hispanic black	1	(2.6%)
Other	6	(15.4%)

BMI: Body mass index

Actigraphy in women.

We used the MotionWatch 8® actigraphy watch (CamNtech Ltd, Cambridge, UK) to detect wrist movements every minute. We obtained recordings from women before conception and throughout pregnancy. Trimesters were defined as gestational weeks 4 through 13 (Trimester 1), 14 through 27 (Trimester 2), and 28 through delivery (Trimester 3). Data were analyzed with a custom R script (http://www.R-project.org/). Briefly, total daily activity was calculated as the sum of all accelerometer counts each day. Sleep onsets were determined from the actigraphy profile of each 24 h (12 PM to 12 PM next day). We calculated the instantaneous frequency of motion counts as the inverse of the time between each wrist movement. Based on the distribution of instantaneous frequency counts, we identified the time of sleep onset. We excluded weekends and days with more than 13 hours of inactive time and included only patients with at least seven days of usable data per trimester. A custom R script was used to correct for Daylight Saving Time, and Paired Wilcoxon Rank-Sum tests were used for statistical analyses.

Results

Chronotype advances during pregnancy in mice and women.

To determine whether daily rhythms change during pregnancy, we continuously monitored running wheel activity of C57BL/6NJ mice. Non-pregnant dams (n=14) showed a consistent daily onset of activity that aligned with the time of lights off throughout the recording (Figure 1A). In contrast, pregnant dams (n=13) showed a consistent and significant advance in their timing of nightly activity onset (Figure 2A; Pregnant: F_18,228=4.3, P<0.05 vs. Non-pregnant: F_18,247=1.1, P>0.05). The earlier chronotype was apparent by the third day of pregnancy, reached a maximum of ~4h advanced relative to non-pregnant around day 7 of pregnancy and returned to the pre-pregnancy state around the tenth day of pregnancy.

Figure 1. Pregnancy shifts daily locomotor activity patterns in women and mice.

(A) Representative double-plotted actograms from two female mice recorded in a light-dark cycle as indicated by the white and black bars. (B-C) Representative single-plotted actograms of normal (Subject 1), early (Subject 2) and late (Subject 3) chronotype women for two weeks while non-pregnant (NP) and throughout their first, second and third trimesters (sunrise to sunset, yellow shading).

Figure 2. The daily onset of activity in mice and sleep in women becomes earlier during early pregnancy.

(A) The daily activity onset advanced in mice in early pregnancy (open circles, mean ± SEM, n=13 mice) compared to the reliable daily activity onsets of non-pregnant dams (closed circles, n=14 mice) around dusk (light-dark cycle shown as white-black bar; repeated measures two-way ANOVA; P<0.05). (B) Sleep onset was significantly earlier in women (black circles, n=39) during their first trimester (T1) compared to before pregnancy or T2 or T3 (bars; Paired Wilcoxon Rank Sum test, * indicates P<0.05).

To determine the effects of pregnancy on circadian timing in women, we analyzed locomotor wrist actigraphy data from 39 women planning pregnancy. Women who conceived were followed throughout their entire pregnancy. We analyzed the data as the change in sleep onset to account for the individual differences in the baseline chronotype of the participants. Similar to the chronotype change in pregnant mice, the timing of sleep onset in women became earlier during the first (T1) and second (T2) trimesters compared to the sleep onset before pregnancy (P<0.05;) Figures 1B and 2B and Table 2). While the timing of sleep onset significantly advanced, there was not a significant change in the daily onset of activity. Sleep onset returned to the pre-pregnant state during the third trimester. Because all of the 39 recordings included one-hour advance or delay of the time due to daylight savings time, we adjusted for the change to or from standard time. We found that the advance in sleep onset was significant before and after this adjustment (Table 2). We also tested whether the change in chronotype during pregnancy depended on the time of year that subjects delivered. We found the advance in sleep onset in women regardless of season (Table 3). Thus, in both mice and women, chronotype advanced at the beginning of pregnancy and then returned to the pre-pregnancy state during the last one-third of pregnancy.

Table 2.

Measures of sleep and wrist activity (median, quartile 1, quartile 3) before and during pregnancy in women (N=39). Superscripted letters indicate differences between prepregnancy

Work Days without daylight saving adjustment	Prepregnancy	Trimester 1	Trimester 2	Trimester 3
Sleep Onset (Local time)	10:16  (9:52,10:31) PM B,C	9:52 (9:24,10:30) PM A,D	9:58 (9:27,10:31) PM A,D	10:14 (9:53,10:37) PM B,C
Total Activity (Counts x 103)	269.1  (207.5,304.1) B,C,D	230.0 (168.3,263.7) A	206.5 (177.1,269.8) A	227.4 (175.5,260.5) A
Work Days with daylight saving adjustment	Prepregnancy	Trimester 1	Trimester 2	Trimester 3
Sleep Onset (Standard time)	9:28 (9:13,9:57) PM B	9:18 (8:48,9:50) PM A,D	9:22 (8:42,9:51) PM D	9:34 (9:06,10:15) PM B,C
Total Activity (Counts x 103)	269.1 (207.5,304.1) B,C,D	230.0 (168.3,263.7) A	206.5 (177.1,269.8) A,B	227.4 (175.5,260.5) A

^A, first trimester

^B, second trimester

^Cand third trimester

^D(Wilcoxon Rank Sum test, P < 0.05).

Table 3.

Seasonal distribution of sleep onset (mean vector, 95%CI) during pregnancy in women (N=39). Note the chronotype advance during pregnancy in all four seasons.

Sleep Onset by Season of Delivery
	Fall	Winter	Spring	Summer
Total Deliveries (#)	8	10	9	12
Sleep Onset Baseline Mean Vector (95% CI)	10:02 PM (9:45PM,10:20PM)	10:38 PM (10:04PM,11:12 PM)	10:27 PM (9:38PM,11:16PM)	10:09 PM (9:51PM, 10:26PM)
Sleep Onset T1	9:36 PM (8:58PM,10:14PM)	10:16 PM (9:49PM,10:42PM)	10:04 PM (9:16PM,10:53PM)	9:45 PM, (9:21PM, 10:10 PM)
Sleep Onset T2	9:58 PM (9:36PM,10:20PM)	10:15 PM (9:48PM,10:43PM)	10:15 PM (9:25PM,11:05PM)	9:37 PM (9:11PM,10:02PM)
Sleep Onset T3	10:11 PM (9:44PM,10:38PM)	10:25 PM (10:01PM,10:49PM)	10:18 PM (9:20PM,11:15PM)	10:20 PM (9:55PM,10:46PM)

Locomotor activity declines in mice and women during pregnancy.

Shifts in chronotype could reflect changes in circadian phase or in the amount of locomotor activity. Because pregnancy is often accompanied by reduced overall movement (Richards 1966; Albers et al. 1981; Rosenwasser et al. 1987; Scribner & Wynne-Edwards 1994; Kimura et al. 1996; Poudevigne & O’Connor 2006), we measured total daily activity throughout pregnancy. Mice showed significantly reduced levels of locomotor activity beginning on day 12 of pregnancy (Figure 3A; Repeated measures two-way ANOVA, P<0.05). Women showed less movement during all three trimesters of pregnancy than before pregnancy (Figure 3B and Table 2; P<0.00001). Notably, both mice and women showed reduced locomotor activity levels in the last one-third of pregnancy compared to early pregnancy, a time when their chronotype (daily activity onset in mice, daily sleep onset in women) had returned to pre-pregnant times. In addition, chronotype advanced in pregnant mice ~10 days before the mice showed reduced amount of locomotion. This suggests that the shifts in chronotype are independent of changes in daily activity levels.

Figure 3. Total activity diminishes across pregnancy in women and mice.

(A) Pregnant (P) mice (n=13, open circles) reduced their amount of daily wheel revolutions starting around day 12 of pregnancy compared to non-pregnant mice (n=14, NP, closed circles; mean ±SEM; Repeated measures two-way ANOVA; P<0.05). (B) Wrist activity decreased in women during pregnancy (counts x 10⁴; median, quartile 1 and 3, N=39) compared to pre-pregnancy (Paired Wilcoxon Rank Sum test; P<0.00001).

Discussion

Our results indicate that chronotype and activity levels change during pregnancy in both women and mice. Despite differences in baseline chronotype of the study participants, or the time that women wake and sleep, women showed lower activity throughout pregnancy and advanced sleep onset during their first two trimesters. Their advanced chronotype returned to pre-pregnancy its state during the third trimester. In mice, a normal pregnancy also was characterized by three phases: 1) an advance in the onset of daily activity by pregnancy day 3, 2) a return of activity onset characteristic of the pre-pregnant stage around pregnancy day 10, and 3) a decrease in levels of locomotor activity beginning at pregnancy day 12 and lasting until delivery.

Women differed from mice in several ways in our study. Although chronotype advanced during pregnancy in both women and mice, this was manifested in daily sleep onset in women and locomotor activity onset in mice. Sleep onset rather than sleep offset (activity onset) was used as a surrogate for chronotype as social constraints of work (e.g., causing women to use alarm clocks) and familial relationships may mask the effects of pregnancy on women’s timing of sleep offset. Additionally, women, but not mice, showed decreased locomotion starting early in pregnancy. By categorizing data from women by trimester, we may have diminished the ability to detect events in early pregnancy that we were able to resolve with daily measures from inbred mice. Regardless, actigraphy provides an inexpensive and non-invasive method to continuously monitor chronotype during pregnancy. It is intriguing that we find similar phenotypes monitoring wrist actigraphy in women and wheel running in mice. Suppression of general activity during pregnancy has been reported in women using questionnaires, pedometry, and actigraphy (Shima 1990; Sternfeld 1997; Lindseth & Vari 2005; Cioffi et al. 2010; Harrison et al. 2011; Nascimento et al. 2015; Oliveira et al. 2017), and in mice using general activity and wheel running (Richards 1966; Terada 1974; Albers et al. 1981; Rosenwasser et al. 1987; Scribner & Wynne-Edwards 1994; Ducsay 1996). We conclude that non-invasive methods like actigraphy can be used to study how chronotype disruption, including shift work, might affect reproductive and delivery outcomes.

The mechanisms underlying pregnancy-induced changes in chronotype are unknown, but the similar responses in women and mice suggest a conserved mechanism. Two candidates for consideration are ovarian steroid hormones and metabolism. Shifts in chronotype at the time of puberty have been attributed to sex hormones in rats, common degus, and macaques (Golub MS 2002; Hagenauer et al. 2009; Hagenauer et al. 2011a; Hagenauer et al. 2011b; Hagenauer & Lee 2012) and could underlie the pregnancy-induced shifts in daily rhythms. In rats, high levels of estrogen during the proestrus stage advance the timing of daily onset of activity and increase daily wheel running (Albers et al. 1981). In contrast, progesterone can delay activity onset during estrus and lengthen the period of locomotor activity (Albers et al. 1981). Thus, the coordinated increases in estrogen and/or progesterone during early pregnancy in mice and women may underlie the observed changes in chronotype.

Changes in maternal metabolism over pregnancy coincide with alterations in the progesterone/estradiol ratio and or progesterone/estradiol receptor status and may contribute to the findings presented here (Jones & Astwood 1942; Hahn & Hays 1963; Romero et al. 1988; Maclin et al. 1990; Potgieter et al. 1995). For example, during pregnancy, maternal basal and postprandial glucose metabolism adapts to meet maternal and fetal demands (Lain & Catalano 2007). Our observations of the normal advances in chronotype during pregnancy may explain why studies have shown that late chronotype pregnant women are about 2.5 times more likely to develop gestational diabetes in their second trimester than were those who had an earlier chronotype (Facco et al. 2017). Future studies should investigate whether these changes in chronotype reflect changes in circadian timing. Mice will serve as a powerful model to investigate the genes, molecules, and cells underlying circadian changes during pregnancy.

Although some studies report increased sleep duration during pregnancy, others report insomnia in mice and women at term (Driver & Shapiro 1992; Hertz et al. 1992; Brunner et al. 1994; Kimura et al. 1996; Reichner 2015; Hashmi et al. 2016; Reid et al. 2017). Our findings that activity levels decreased during late pregnancy in mice and throughout pregnancy in women are consistent with a large body of literature (Rosenwasser et al. 1987; Scribner & Wynne-Edwards 1994; Kimura et al. 1996; Poudevigne & O’Connor 2005, 2006) and suggest that a common mechanism modulates sleep/wake cycles during pregnancy in the two species. These data also indicate temporally separated changes in the regulation of daily activity and physiology by circadian (time of day gating) and sleep homeostatic (sleep debt accumulated during wakefulness) processes. Future studies assessing changes in other measures of the sleep homeostat including sleep fragmentation throughout pregnancy would be informative. Taken together, these results indicate that pregnancy induces an advance of the circadian system that resolves prior to term delivery and is separable from the processes controlling sleep duration.