Interaction between time-of-day and oxytocin efficacy in mice and humans with and without gestational diabetes

Abstract

Due to significant risks of peripartum complications, pregnancies complicated by diabetes often require labor induction or augmentation with synthetic oxytocin. However, the efficacy of oxytocin is often compromised in diabetic pregnancies. Given that diabetes deregulates the body's circadian timekeeping system, our objective was to determine how time of day and the circadian clock gene, Bmal1, gate oxytocin efficacy. We compared oxytocin uterotonic efficacy in a smooth muscle-Bmal1 conditional knockout mouse (cKO), and a mouse model of food-induced gestational diabetes. We found that in wild-type mice, the oxytocin receptor is expressed in a time-of-day-dependent manner and is under the control of BMAL1. Both Bmal1 cKO and food-induced gestational diabetes mice, which presented with a downregulation of Bmal1 in the uterus, had decreased uterine contractility in response to oxytocin. To establish the translational value of these findings, we utilized an immortalized term human myometrial cell line. We determined that the time-of-day impacted oxytocin-induced myometrial contractility in vitro. Furthermore, we conducted a retrospective medical record analysis of 2,367 pregnant patients ≥39 weeks gestation undergoing induction of labor. We assessed the timing of labor induction and the impact of gestational diabetes mellitus on labor duration. Induction of labor in the morning compared to midnight was associated with a ∼1.5-hour and ∼7-hour shorter labor duration in controls and patients with gestational diabetes mellitus, respectively. In conclusion, circadian timing plays a key role in induction of labor and oxytocin responsiveness and should be considered when managing labor induction.

Graphical abstract

Highlights

• •Oxytocin-induced uterine contractions vary by time-of-day in mice and humans.
• •BMAL1 directly regulates oxytocin receptor expression in the gravid mouse uterus.
• •Bmal1 knockout or gestational diabetes reduces murine uterine sensitivity to oxytocin.
• •Human myometrial cells show circadian rhythms in oxytocin responsiveness.
• •Labor in women is shorter when initiated in the morning, especially with gestational diabetes.

1. Introduction

Diabetes during pregnancy represents an obstetric concern, contributing significantly to fetal, neonatal, and maternal morbidity and mortality [1]. Pre-gestational diabetes and gestational diabetes mellitus (GDM) are associated with an elevated risk of adverse maternal and neonatal outcomes compared to non-diabetic pregnancies. These complications include an increased incidence of preterm birth, prolonged labor, dysfunctional uterine contractions, and postpartum hemorrhage [[2], [3], [4], [5]].

Diabetes-induced alterations in uterine smooth muscle (myometrial) contractility have been implicated in adverse pregnancy outcomes [6,7]. Specifically, the efficacy of synthetic oxytocin, when administered to induce and augment labor [8], is reduced in diabetic pregnancies [3]. This reduced efficacy translates to prolonged labor which is a risk factor for postpartum hemorrhage and cesarean delivery. The variability in synthetic oxytocin's effectiveness [9] prompts further investigation into potential underlying biological factors not currently considered in clinical management protocols.

One potential contributing factor to the variable efficacy of oxytocin for labor enhancement and labor induction is the time-of-day when labor induction and oxytocin administration is initiated. Using retrospective medical health record data, we identified that labor duration was impacted by induction of labor with oxytocin, and dependent on time-of-day, gestational age, BMI, and parity in non-diabetic pregnant women [10]. Such time-of-day effects on drug efficacy are well-documented in the literature [11,12]. Failure to account for these circadian variations in drug response may result in inconsistent clinical outcomes and, requiring higher doses to achieve therapeutic effects [[13], [14], [15]].

A mechanism by which the time-of-day effects of drug efficacy are generated is through the circadian expression of the drug's target. The cell's endogenous molecular clock, a cell-autonomous transcriptional-translational feedback loop, drives circadian gene expression [16,17], influencing cellular functions throughout the day [[18], [19], [20]]. At the core of the molecular clock are a small set of transcriptional regulators, including Bmal1, Clock, Cry1/2, Per1/2/3 and Nr1d1/2. Of these transcriptional regulators, loss of Bmal1 disrupts molecular clock function, leading to a loss in circadian rhythm [17,21]. Dysregulation of the molecular clock has been associated with numerous pregnancy complications in women, such as GDM [22] and preeclampsia, with and without preterm birth [[23], [24], [25]]. Notwithstanding, the role of the molecular clock within reproductive tissues during pregnancy remains largely unknown. Several mouse studies indicate the essential role of a functional molecular clock in the uterus during pregnancy [26,27], and its contributions to fetal resorption and labor dystocia [28]. Conditional knockout of Bmal1 in the smooth muscle (including the myometrium, referred to as Bmal1 cKO) increased the rate of abnormal birth timing (∼16% preterm, and ∼16% postterm, with term pregnancy in the mouse ∼ gestation day (GD)19.5). This increase in abnormal birth timing was not associated with changes in the contractile-associated protein connexin-3, oxytocin receptor (Oxtr), or progesterone, the latter of which is known to reduce uterine contractions [29]. The mechanism causing this mistiming in birth in mice was not identified. As BMAL1 regulates the circadian expression of 10–30% of transcripts in a cell, the absence of BMAL1 in the myometrium could alter the daily pattern of expression of a wide variety of genes coding signaling proteins, hormones, and immune modulatory signals, which, through an unknown mechanism, could cause mistiming of labor onset and birth.

Together, these prior studies indicate that the loss or dysregulation of the molecular clock during pregnancy is associated with adverse pregnancy outcomes and potentially labor dystocia due to uterine malfunction. The goal of this study was to determine how Bmal1 and time-of-day modulate uterine contractions and the uterotonic efficacy of oxytocin. To this end, we conducted mechanistic studies using Bmal1 conditional knockout mice, and a mouse model of food-induced gestational diabetes. In addition, we evaluated the impact of the time of induction in the presence of GDM on labor duration in a retrospective medical record analysis of over 2,000 pregnancies.

2. Methods

2.1. Mouse breeding and timed matings

Mice were maintained on a light/dark cycle of 12 h light, 12 h dark, with lights ON at Zeitgeber Time 0 (ZT0) and lights OFF at ZT12 with lights spectral composition as previously described [30]. Mice had food and water ad libitum. PER2::Luciferase [Tg(Per2-luc)1Jt, JAX #006852], Bmal1^flox [B6.129S4(Cg)-Arntl^tm1Weit/J, JAX #007668] and Telokin^{Cre+/− 29} were used. All experimental mice were studied on gestation day (GD) 18 or GD19, where the day of the presence of a vaginal plug was defined at GD1. We confirmed the specific deletion of Bmal1 in smooth muscle via PCR (Supplemental Fig. 1). Mice were euthanized through cervical dislocation.

2.2. Food-induced gestational diabetes mouse model

Controls were maintained on standard chow for the entire study. To generate mice with food-induced gestational diabetes (FID), 5-6-week-old females had blood glucose levels monitored using One Touch Verio Blood Glucose Monitoring System Meter (OneTouch Verio) and body weight measured pre- and post–conception. After basal glucose levels were determined, the mice were given a high (45%) fat diet (TEKLAD:TD.06415), with 30% corn syrup in autoclaved water. Mice were maintained on this diet for 5 weeks. After 5 weeks, both control and high fat, high sugar diet females were mated with a male on standard chow. The day of the presence of a vaginal plug was defined at GD1. A glucose tolerance test (GTT) was done on GD16. Briefly, pregnant females were food deprived from ZT0-ZT6. A bolus of sucrose (2 g/kg weight) was given through intraperitoneal injection and blood glucose was monitored at 0', 15', 30', 60', 90', and 120’. Mice were euthanized by cervical dislocation on GD19 at ZT5±2 h, and uterine samples were prepared for uterine contraction studies. Pups were sexed, weighed and had insulin and blood glucose measured at euthanasia. The high fat, high sucrose diet induced gestational diabetes in 60% of the mice (FID), whereas 40% of the females did not develop diabetes (FN).

2.3. Organotypic circadian PER2::Luciferase luminescence recording

PER2::luciferase (PER2::LUC) recordings were done as we have previously described [21]. In brief, the uterus was removed from a GD18, ZT19±3 h mouse, and uterine samples were collected. For samples with endometrium removed, the endometrium was gently scraped from the myometrium with a scalpel. Uterine explants were placed on a MilliCell membrane in 35-mm culture dishes and placed into a LumiCycle (Actimetrics) at 35.5 °C, 5%/95% CO₂, in a non-humidified environmental chamber. The bioluminescence signal was counted every 10 min for 6 days and analyzed on days 1–6 of recording time. Data were normalized by subtraction of the 24-hour running average from the raw data and then smoothed with a 1 h running average (Luminometer Analysis, Actimetrics) and analyzed blind to the experimental group.

2.4. Mouse uterine contractions

Pregnant females were euthanized at ZT5±3 h or ZT17±3 h. The uterus was dissected along the longitudinal axis and subdissected into full-thickness uterine strips of ∼ 2 × 5 mm² in ice-cold oxygenated physiological saline [154 mM NaCl, 5.6 mM KCl, 1.2 mM MgSO₄, 10.2 mM HEPES, 2 mM CaCl₂, and 8 mM glucose]. Uterine strips were mounted in organ baths (DMT Muscle Strip Myograph System- 820MS) containing oxygenated physiological saline at 36.0 ± 0.5 °C. Each strip was stretched to a tension of 6 mN and allowed to equilibrate until 15 min of regular spontaneous contractions occurred. The tissues were treated with vehicle (Milli Q water, at 1/2000 dilution), 500 nM atosiban (A3480-10 MG, Sigma Life Science) or 1 nM oxytocin (O3251-1000IU, Sigma Life Science). LabChart software (ADInstruments) analyzed contractions by evaluating the area under the curve (AUC), amplitude, and frequency of contractions over 10 min during baseline or drug application periods.

2.5. In vitro PHM1-41 cell contraction study

This study used immortalized pregnant human (Homo Sapiens) elective cesarean section myometrium/fibroblast-like cells derived from 39 weeks of gestation, termed PHM1-41 [31] (American Type Culture Collection, VA, #CRL-3046, RRID:CVCL_8566). Cells were cultured in DMEM (Mediatech) containing 10% fetal bovine serum (Gemini Bio), and 1x penicillin-streptomycin (Life Technologies/Invitrogen) in a humidified 5% CO₂ incubator at 37 °C. PHM1-41 cells were seeded into 35 cm plates (Nunc) at 18 million cells per well. To synchronize the molecular clock, cells were first treated with low serum media [DMEM containing 0.1% bovine serum albumin (BP9706-100, Fisher)] for 24 h (serum starvation) [32.33]. After 24 h serum starvation, cells were then treated with regular cell culture media. Beginning at 12 h after synchronization (set as circadian time point 00:00 h), cells were treated at the indicated time-points with either vehicle (water) or 4.5 nM oxytocin (O3251-1000IU, Sigma Life Science). Images of cells were captured at 0 and 15 min after treatment using a Leica DMi1 microscope (Leica) under bright-field illumination at 20× magnification. Images were analyzed for cellular surface area as a measure of contraction using Fiji [34]. Percent cell size was calculated as cellular surface area at 15 min divided by the corresponding cellular surface area at baseline (0 min). For each image 5–10 cells were analyzed.

2.6. Cell culture conditions, transient transfections and luciferase assays

Mouse embryonic fibroblast NIH-3T3 (American Type Culture Collection, VA, #CRL-1658. RRID:CVCL_0594) cells were cultured in DMEM (Mediatech), containing 10% fetal bovine serum (Gemini Bio), and 1x penicillin-streptomycin (Life Technologies/Invitrogen) in a humidified 5% CO₂ incubator at 37 °C. For luciferase assays, NIH-3T3 cells were seeded into 24-well plates (Nunc) at 50,000 cells per well. Transient transfections for luciferase assays were performed using PolyJet™ (SignaGen Laboratories), following the manufacturer's recommendations. NIH-3T3 cells were co-transfected as indicated in the figure legends with 200 ng/well luciferase reporter plasmids, as well as 100 ng/well thymidine kinase-β-galactosidase reporter plasmid, which served as an internal control [35.36].

The plasmids used were mouse −1000 to +200bp Oxtr-luciferase (200 ng/uL, VectorBuilder), mouse Bmal1 overexpression plasmid (200 ng/uL, Addgene, #31367), pcDNA 3.1 (between 0ng/uL and 200 ng/uL), and pGL4-luciferase (200 ng/uL). Site-directed mutagenesis of the Oxtr-luciferase reporter was performed using the NEB Q5 Site-Directed Mutagenesis Protocol (New England Biolabs Inc.), following manufacturer's instructions, to mutate each E'-box to TGACGA. Primers for NEB Q5 site-directed mutagenesis were designed using NEBaseChanger [37] (Supplemental Fig. 2). We systematically equalized plasmid concentrations by adding the corresponding inactive plasmid backbone to equalize the amount of DNA transfected into cells. Cells were given their respective ligands 24 h after transfection and were then harvested 48 h after transfection in lysis buffer [100 mM potassium phosphate (pH 7.8) and 0.2% Triton X-100]. Luciferase values were normalized to β-galactosidase values to control transfection efficiency. Values were further normalized by expression as fold change compared to pGL4 control plasmid, as indicated in the figure legends.

2.7. RNAscope multiplex in situ hybridization assay

We completed RNAscope® multiplex in situ hybridization detection of mouse (Mus musculus) mRNAs with RNAscope® LS Multiplex Fluorescent Reagent Kit (Advanced Cell Diagnostics, cat no. 322800) following the vendor's standard protocol for FFPE tissue sections with minor modifications. RNAscope® assays were performed on a Leica Bond autostainer as described [38.39] with the following probes:RNAscope® 2.5 LS Probe – Mm-Arntl (also known as Bmal1) [aryl hydrocarbon receptor nuclear translocator-like (Arntl) transcript variant 1 mRNA, cat no. 438748-C1] and RNAscope® 2.5 LS Probe – Mm- Oxtr-C2 [oxytocin receptor (Oxtr), cat no. 412178-C2]. Tissue slides were counterstained with DAPI and scanned with Aperio Versa imaging system with 20× objective with customized narrow-width band excitation and emission filter cubes as described [38]. The Aperio Cellular IF Algorithm (Leica Biosystems, No:23CIFWL) was used for automated cell enumeration and segmentation based on nuclear DAPI staining. Cells were classified based on the expression levels of one or more mRNAs (cell classes were threshold based solely on increasing intensity value per pixel within each segmented object/cell).

2.8. Quantitative real-time PCR

Full-thickness uterine strips were first dissected from pregnant females at designated time points on GD19. Total RNA was extracted using E.Z.N.A. Total RNA Kit I (Omega Bio-tek). cDNA was obtained by reverse transcription of total RNA using a Lunascript RT Supermix cDNA synthesis kit (New England Biolabs). cDNA products were detected using an iQ SYBR Green Supermix (Bio-Rad Laboratories) on a qRT-PCR CFX real-time detection system (Bio-Rad Laboratories). qRT-PCR primers were on Bmal1 (F:TGACCCTCATGGAAGGTTAGAA; R:GGACATTGCATTGCATGTTGG), Clock (F:ATGGTGTTTACCGTAAGCTGTAG; R:CTCGCGTTACCAGGAAGCAT), and Cry2 (F:CACTGGTTCCGCAAAGGACTA; R:CCACGGGTCGAGGATGTAGA). Data were expressed as fold change using the ΔΔCT method by normalizing Bmal1, Clock, and Cry2 to the housekeeping genes H2afz (F:TCACCGCAGAGGTACTTGAG; R:GATGTGTGGGATGACACCA) and Gapdh (F:GGCAAATTCCATGGCACCGT; R:GCAAATGAGCCCCAGCCTTC). Data represent mean fold change ± SEM.

2.9. Human participants

2.9.1. Human study population

The Sparrow cohort (n = 7,804 consecutive deliveries at the Sparrow Health System, East Lansing, MI from February 2019 to March 2022) was used. Deidentified data of pregnant patients were collected from the Electronic health records. Exclusion criteria:Maternal age<18 years old, multiple gestation, spontaneous labor or labor type unknown, gestational age <39 weeks, pre-gestational diabetes utilizing ICD code = O24.x not including O24.4x), fetal demise, or absent neonatal anthropomorphic data. These exclusions led to the final dataset of 2367 patients for downstream analyses. A sub-analysis was completed where only patients receiving oxytocin for labor induction were included. Both vaginal and cesarean sections were combined. This final dataset included 985 patients for downstream analyses.

2.9.2. Measurements of major variables

Gestational diabetes mellitus (GDM) patients were defined as those with ICD10 code = O24.4x, and the remaining patients without ICD 024.x as Non-GDM. Labor induction mode was classified into 3 groups: oxytocin only, Cytotec only, and both oxytocin and Cytotec. Induction of labor duration was calculated by subtracting the time of induction initiation, which was defined by the time of the initial dose of Cytotec or oxytocin. Induction of labor method including cervical balloon, Cytotec, oxytocin or combinations thereof were at the discretion of the provider and based on Bishop score at the time of admission. Labor induction start time was grouped into six 24-hour clock intervals:00–04, 04–08, 08–12, 12–16, 16–20, and 20–24 h. Body Mass Index (BMI) refers to BMI at delivery and was grouped into 4 groups – underweight (BMI <18.5), Normal (BMI = 18.5 to 25), Overweight (BMI = 25 to 30), and Obese (30+). Maternal age was categorized as <30 versus ≥30 years old. Parity was dichotomized into “0" (nulliparous) versus "≥1" (parous). Race was coded as:"White”, “Hispanic”, “Black”, “Asian”, and “Others”. Delivery method was coded as “Cesarean-section” or “Vaginal” and newborn sex as “Female” vs. “Male”. Fetal weight was measured in grams. Initial cervical dilation and bishop scores were not available for this study.

2.9.3. Statistical analysis of retrospective data

All categorical variables were summarized as numbers and percentages, and the continuous variables were presented as median and interquartile range (IQR) due to the non-normal distribution of the labor duration (Supplemental Figures 3, 4 and 5). The generalized additive model (GAM) [40] backward-elimination algorithm was applied to select covariates for examining the association between labor duration as the outcome and multiple covariates, including maternal age as a continuous covariate (a smooth term), with an interaction term of induction start time with or without GDM. Comparisons of estimated marginal means among interactions of factors after adjusting for other factors in the GAM model were conducted with the emmeans R package [41]. The adjusted p-values were calculated by using the Benjamini-Hochberg method to correct multiple comparisons. All data management and analyses were conducted in R-4.4.1 [42].

2.10. Study approvals

The Animal Use Committee and the Institutional Animal Care of Michigan State University approved animal procedures. The Institutional Review Board of Michigan State University approved human research under STUDY0007199.

2.11. Statistical analysis of experimental data

Statistical analyses were done as described in the figure legends. Statistical analyses were done using Prism 9 (GraphPad Software, San Diego, CA, USA), and R (R Development Core Team).

3. Results

3.1. Oxytocin uterotonic efficacy depends on the time of day in the gravid mouse

To determine how time-of-day impacts the uterotonic efficacy of oxytocin, we placed uterine explants from late gravid mice (gestation day 18, GD18) in a myograph and recorded their contractile response to oxytocin during the rest phase (Zeitgeber time 5, ZT5, 5 h after lights on ± 3 h), and active phase (ZT17, 5 h after lights off ± 3 h). Oxytocin increased contraction frequency at both studied time points (Figure 1A). A Two-way ANOVA comparing time-of-day to drug (vehicle and oxytocin) showed a significant effect of time-of-day [F(1,35) = 15.28, p = 0.004] and drug [F(1,62) = 21.48, p < 0.001], but no significant interaction between the two (p = 0.68). The effect of oxytocin was specific, as evidenced by the capacity of the oxytocin receptor (OXTR) antagonist, atosiban, to antagonize its effect at both ZT5 and ZT17 (Figure 1A–C). Interestingly, the sole application of atosiban significantly reduced uterine contraction frequency at ZT17 (p = 0.0006) but not at ZT5 (p = 0.92, Figure 1A). Two-way ANOVA analysis comparing time-of-day to drug (vehicle and atosiban) showed a significant effect of time-of-day [F(1, 42) = 4.41, p = 0.04), drug (F(1,65) = 14.87, p < 0.0003), and a significant interaction between the two [F(1,42) = 6.77, p = 0.01).

Figure 1.

Time-of-day specific effect of OXTR-ligands on uterine contractions in the pregnant mouse. A) Uterine samples were mounted in a myograph during the dark phase ZT17 (ZT17 $\pm 3$ hours) or light phase ZT5 (ZT5 $\pm$ 3 h), and contractions were evaluated in response to vehicle (water), oxytocin 1 nM or atosiban 500 nM in wild-type mice at GD18. Two-way ANOVA, n = 7–17, with samples in duplicate. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001. B, C) Example tracing of uterine contractions in response to the indicated treatments at GD18, ZT17. D) Representative RNAscope® ISH staining for Oxtr in the GD18 uterus at ZT5 and ZT17, and E) Quantified data. n = 7–9/group. T-test, ∗, p < 0.05.

To determine if the time of day impacted Oxtr expression at GD18, we detected Oxtr mRNA by ISH. Oxtr expression was significantly lower during the rest phase of the day (ZT5) than the active phase (ZT17, Figure 1D,E). In contrast, Bmal1 transcript levels trended lower at ZT17 than ZT5 (t(14) = 2.06, p = 0.059, Supplemental Fig. 6A–C), whereas the number of Oxtr⁺/Bmal1⁺ cells were lower at ZT17 than ZT5 (t(3) = 3.59, p = 0.04, Supplemental Fig. 6D–F).

3.2. Bmal1 is required for circadian rhythms and appropriate oxytocin receptor expression in the mouse myometrium

To further support the finding that the molecular clock gates the time-of-day effect of OXTR-regulated uterine contractions, we conditionally deleted the Bmal1-flox/flox allele from smooth muscle cells using the Telokin-cre+/− allele. These mice are referred to as Bmal1 cKO and have been previously validated [29]. To confirm the specific abolishment of circadian rhythms in the myometrium, we generated triple transgenic circadian reporter mice, where we crossed the Bmal1 cKO with the validated PER2::Luciferase [43] mice. As expected, Bmal1 cKO only impaired circadian rhythms in the myometrium (Figure 2A, and Supplemental Fig. 7). We identified highly conserved regions of the Oxtr promoter region (Supplemental Fig. 8A), including four semi-conserved BMAL1 binding sites (E′-boxes, Supplemental Fig. 8B). To test BMAL1 use of these sites, we performed an in vitro reporter gene assay in mouse NIH-3T3 cells. We found that Bmal1 overexpression increased Oxtr-luciferase expression in a dose-specific manner (Supplemental Fig. 8C). Transient transfections of the Oxtr-luciferase reporter with and without site-directed mutagenesis (labeled μE) to the four E′-boxes identified that mutations at sites μE2 and μE4 abolished Bmal1-induced Oxtr-luciferase expression (Figure 2B). These findings agree with the time-of-day specific binding of BMAL1 to the Oxtr regulatory region in male mouse liver (Supplemental Fig. 8D). This shows that BMAL1 can directly bind to the Oxtr regulatory region to drive Oxtr expression. To determine the regulation by BMAL1 of the Oxtr-luciferase reporter in vivo, we interrogated Oxtr mRNA expression in uterine tissue from Bmal1 cKO animals (Figure 2C,D). We found that Oxtr and Bmal1 had significantly reduced expression in the Bmal1 cKO compared to the control at ZT17 (Figure 2C,D). Note that Oxtr is expressed in the Bmal1 cKO myometrium at ZT17, but the number of cells with Oxtr expression is significantly reduced (Figure 2C, p = 0.03). Due to the low Oxtr expression at ZT5 in the control, no difference in Oxtr transcript between control and Bmal1 cKO was identified at this time (not shown).

Figure 2.

BMAL1 regulates Oxtr expression. A) Representative PER2::LUC recording of GD18 myometrium from control (Telokin^Cre:PER2::LUC) and Bmal1 cKO:PER2::LUC. Data represent n = 5/group. B) To determine if BMAL1 regulated Oxtr expression, in vitro NIH-3T3 cells were transiently transfected with Bmal1 overexpressing plasmid and Oxtr-luciferase plasmids with or without mutated E′-boxes (μE1, μE2, μE3, μE4). Two-way ANOVA, n = 4–5 per group in duplicate, ns:non-significant, ∗p < 0.05, ∗∗p < 0.05, ∗∗∗p < 0.005. C) Histogram of Oxtr expression levels and D) representative RNAscope® ISH staining of GD18 uterus at ZT17 in control and Bmal1 cKO uterus. Two-way ANOVA, n = 3–4/group. ∗, p < 0.05.

3.3. Bmal1 cKO uterus is hyposensitive to oxytocin-induced contractions

To test the capacity of oxytocin to promote uterine contractions in Bmal1 cKO, we first evaluated ex vivo spontaneous uterine contraction frequency. We found the Bmal1 cKO uterus was hypercontractile (Figure 3A,F(1,101) = 7.26, p = 0.008). In controls, spontaneous uterine contraction frequency was significantly higher during the active phase (ZT17) than in the inactive phase (ZT5, p = 0.05), a time-of-day difference lost in the Bmal1 cKO (Figure 3A, p = 0.51). Due to the time-of-day effect on spontaneous uterine contraction frequency, we normalized the oxytocin-induced contraction data to vehicle control within each sample for the time points studied (Oxytocin/Vehicle). The capacity of oxytocin to increase uterine contraction frequency was reduced at ZT5 in Bmal1 cKO (Figure 3B,C, p = 0.002), with an overall reduced efficacy of oxytocin in the Bmal1 cKO (p < 0.01).

Figure 3.

Time-of-day and Bmal1 cKO impacts oxytocin's uterotonic efficacy in pregnant mice. A) GD18 uterine samples were mounted in a myograph, and spontaneous uterine contraction frequency was analyzed at ZT5 and ZT17 in control and Bmal1 cKO. n = 11–15, in duplicate. Two-way ANOVA, ∗, p < 0.05; ∗∗, p < 0.01. B) Histogram shows the effect of oxytocin 1 nM-promoted uterine contraction frequency normalized to baseline frequency in GD18 uterine samples harvested at ZT5 and ZT17 from control and Bmal1 cKO. Two-way ANOVA, n = 11–14, in duplicate. ∗∗, p < 0.01. C) Example uterine contraction recordings in response to oxytocin at GD18, ZT17 in control and Bmal1 cKO.

3.4. Food-induced gestational diabetes (FID) reduced oxytocin-induced contractions and molecular clock transcript expression in the mouse uterus

GDM is known to deregulate the molecular clock and reduce oxytocin efficacy. Based on this, we asked how GDM in the mouse impacted molecular clock expression and oxytocin efficacy. To generate females with GDM, we placed virgin females on standard chow or high fat, high sucrose diets (Figure 4A). Females on the high fat, high sucrose diet gained more weight over the study period than females on standard chow (Supplemental Fig. 9A). All females had their basal glucose tested pre-pregnancy and during gestation (Supplemental Fig. 9B). All high-fat, high-sucrose diet mice had their fasting glucose levels tested and responses to a glucose tolerance test (GTT) completed on GD16. Based on the GTT, females were labeled as Food-normal GTT (FN) if they had a GTT <180 mg/ml and as Food-induced gestational diabetes (FID) if their GTT was >180 mg/ml (Supplemental Fig. 9C). The GD19 fetuses from mothers in the FID group had significantly higher blood glucose and insulin levels than in the FN group (Supplemental Fig. 9D and E). To determine how the high-fat, high-sucrose diet with (FID) and without (FN) failed GTT impacted molecular clock gene expression, we collected GD19 ZT5 uterine samples for qPCR. Based on previous work we focused on Bmal1, Clock, and Cry2 [[22], [23], [24], [25]]. We found that the uterus of females on standard chow (Ctr) had higher levels of Bmal1 and Clock than FN, FID and FN + FID (FN and FID were merged as they were not significantly different, Figure 4B,C). BMAL1-CLOCK drives the expression of the clock gene Cry2, a transcriptional regulator which was downregulated in the FN + FID group as compared to Ctr (Figure 4D). To determine how the reduction in molecular clock transcripts in the FN and FID mice was associated with the capacity of oxytocin to promote uterine contractions, we placed GD19 ZT5 uterine explants on the myograph. Spontaneous uterine contraction frequency was comparable across vehicle-treated groups (Figure 4E). The Ctr and FN uterine samples responded with increased uterine contraction frequency to oxytocin, an effect lost in the FID group (Figure 4E,F).

Figure 4.

Food-induced gestational diabetes (FID) reduces molecular clock expression and the uterotonic efficacy of oxytocin. A) The experimental timeline for generating control (Ctr), Food-normal GTT (FN), and Food-induced gestational diabetes (FID) mice. B-D) qPCR of the uterine samples at GD19, ZT5. One-way ANOVA, n = 3–7/group, ∗, p < 0.05; ∗∗, p < 0.01. E) Histogram and, F) example recording of uterine contractions in response to vehicle or oxytocin 1 nM at GD19, ZT5. Two-way ANOVA, n = 3–5, with samples in duplicate. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001.

3.5. Human myometrial cells have time-of-day-specific responses to oxytocin

To determine if the findings from mice translate to human myocytes, we tested in vitro how time-of-day impacted the uterotonic efficacy of oxytocin in a cell line derived from a term pregnant human uterus. We used a validated protocol to synchronize the molecular clock [32.33] (Figure 5A) and assessed the capacity of oxytocin to promote human myocyte (PHM1-41 cell [31]) contractions. Vehicle (water) did not impact PHM1-41 cell size (used as a measure of contraction); in contrast, oxytocin induced PHM1-41 cell contractions in a time-of-day dependent manner, with a significant effect on cell contraction at circadian time points 00, 04, 10, and 24 h (Figure 5B), with a reduction in cell size as compared to control of 31% at 00 h, 28% at 04 h, and 20% at 10 h, as compared to 9% at 13 h and 12% at 16 h.

Figure 5.

Time-of-day dependent changes in uterine response to oxytocin.A) Experimental timeline of synchronization and treatment of PHM1-41 cells. After treatment with 0.1% BSA media for 24 h (serum starvation), cells were treated with regular cell culture media to synchronize their circadian rhythms. Beginning at 12 h after synchronization (time point 0:00 h), each culture dish was treated with either vehicle (water) or 4.5 nM oxytocin for 15 min at the indicated time points. Images were captured at 0 and 15 min after treatment and analyzed for cellular surface area to measure cellular contraction. B) Changes in PHM1-41 cell size (% from 0 min treatment) to vehicle or oxytocin at the indicated time points. Two-way ANOVA with mixed effects, n = 3 replicates per time point with 5–10 cells/replicate, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. The average labor duration for each 4-hour time of induction bin across the 24-hour day is summarized for women with C) vaginal and cesarean delivery combined, with (Yes, n = 253) and without (No, n = 2114) GDM, D) vaginal delivery only, with (Yes, n = 204) and without (No, n = 1580) GDM, E) oxytocin only (vaginal delivery and Cesarean combined), with (Yes, n = 204) and without (No, n = 1580) GDM with GAM regression comparing labor duration between GDM yes (grey boxes) vs. no (black boxes), ∗∗∗p < 0.001. For complete statistics, see Table 2, Table 3, S2-S6. F) Time-of-day distribution of % of patients induced with Cytotec alone, oxytocin alone, or both. Two-way ANOVA comparing the number of women induced in each time window across induction methods, p < 0.05.

3.6. Time-of-day of labor induction impacts labor duration in a GDM-specific manner

To determine whether GDM influences the effect of labor induction timing on labor duration, we analyzed health record data from a tertiary referral hospital in the Midwest USA (Sparrow). As summarized in Table 1 (Vaginal delivery and Cesarean section combined) the cohort was composed of 4.8% Hispanic, 45.2% maternal age≥30 years old, 52.5% parity≥1, 64.7% obese (BMI>30), 10.7% GDM = Yes, 41.7% labor induction mode = oxytocin only, 24.6% Maternal Method = Cesarean Section, as well as 49.9% fetal sex = female and overall median of the fetal weight at delivery = 3480.0 g (IQR = 573 g). No underweight individuals at the time of delivery were present in the data. A similar distribution was found when the cohort was defined by vaginal deliveries only (Table S1).

Table 1.

Frequency distribution of the patients across all covariates (Sparrow, N = 2367). The endpoint is for vaginal delivery and Cesarean section combined.

	All n (%)	Duration of Labor	GDM (N = 253)
		Median (IQR)	n (%)
Race:
White	1641 (69.3)	14.8 (13.1)	158 (62.0)
Hispanic	114 (4.8)	14.6 (11.6)	18 (7.1)
Black	311 (13.1)	15.4 (12.5)	25 (9.9)
Asian	97 (4.1)	14.6 (15.4)	26 (10.0)
Others	204 (8.6)	14.8 (12.3)	26 (10.0)
Maternal age:
<30	1297 (54.8)	15.5 (13.6)	106 (42.0)
≥30	1070 (45.2)	14.3 (12.6)	147 (58.0)
Parity:
0	1125 (47.5)	20.3 (14.2)	101 (40.0)
≥1	1242 (52.5)	11.2 (8.5)	152 (60.0)
BMI:
Normal (BMI<25)	142 (6.0)	12.4 (10.1)	11 (4.3)
Overweight (BMI 25–30)	694 (29.3)	13.8 (12.1)	71 (28.0)
Obese (BMI>30)	1531 (64.7)	15.7 (13.8)	171 (68.0)
Induction mode:
Pitocin only	985 (41.7)	9.8 (7.2)	94 (37.0)
Cytotec only	275 (11.6)	13.6 (7.3)	28 (11.0)
Both	1104 (46.7)	22.0 (13.4)	131 (52.0)
Time of induction:
0-4	331 (14.0)	15.5 (13.5)	40 (16.0)
4-8	192 (8.1)	13.8 (11.0)	19 (7.5)
8-12	297 (12.5)	14.6 (11.6)	26 (10.0)
12-16	553 (23.4)	14.6 (13.2)	66 (26.0)
16-20	490 (20.7)	15.6 (13.5)	56 (22.0)
20-24	504 (21.3)	15.3 (14.0)	46 (18.0)
GDM:
No	2114 (89.3)	14.8 (12.7)	NA
Yes	253 (10.7)	15.8 (14.9)	NA
Delivery method:
Vaginal	1784 (75.4)	13.5 (11.1)	204 (81.0)
C-section	583 (24.6)	20.9 (15.4)	49 (19.0)
Fetal sex$:
Female	1181 (49.9)	14.3 (12.7)	129 (51.0)
Male	1186 (50.1)	15.5 (13.4)	124 (49.0)
Fetal weight at delivery:
Median (Q1,Q3) (grams)	3480.0 (3200.0, 3773.0)

To determine how the timing of labor induction impacted labor duration, we used a Generalized Additive Model (GAM). We found that the timing of labor induction has a significant impact on labor duration, whether we examined the entire cohort or excluded cesarean sections (Figure 5C, D, Table 2 and S2). Labor induction method significantly impacted labor duration, where women induced with oxytocin alone, had a 1.6 h shorter labor duration than women induced with Cytotec, whereas women induced with both Cytotec and oxytocin, had a 9.2 h longer labor duration than women induced with Cytotec alone (Table 2). When combining cesarean and vaginal delivery, and controlling for other covariates (GDM, Parity, BMI, Delivery Method, infant weight at delivery, induction mode and interaction between induction time and GDM) the GAM modeling analysis identified a 1.48 h decrease in labor duration, when labor induction was initiated at the 04–08 h interval, as compared to the 00–04 h interval (SE = 0.76 and p = 0.07, Table 2). Meanwhile, compared to non-GDM, the presence of GDM was associated with an increased total labor induction duration of 6.82 h (SE = 1.33 and p < 0.0001, Table 2). The interaction between induction time and GDM also significantly contributes to labor duration (all p-values ≤0.0112, Table 2). Maternal age was not a contributing factor to labor induction duration (p = 0.3710, Table 2). ‘Fetal sex’ did not significantly contribute to the outcome in the GAM model (data not shown) and thus was removed from the final model. A comparable analysis was completed in women with vaginal delivery only (Table S2), where, despite the smaller sample size (N = 1784), the trends remained the same as for the analysis with vaginal delivery and cesarean section combined (N = 2367, Table 2). Gestational age was statistically similar in each cohort (Table 2), and for each of the 6 timeframes in both the vaginal delivery and cesarean section combined (not shown). Utilizing the GAM 46.7% of the differences in labor induction duration at each timeframe were explained by the covariates.

Table 2.

Summary of the association between the outcome variable Labor Duration and the covariates by a GAM model (Sparrow, N = 2367). The endpoint is for vaginal delivery and Cesarean section combined.

	Coefficient (β)	SE of β	t Statistic	p
(Intercept)	1.98	1.47	1.35	0.1789
Induction time:
4–8 h vs. 0–4 h	−1.38	0.76	−1.82	0.0683
8–12 h vs. 0–4 h	−0.59	0.66	−0.89	0.3734
12–16 h vs. 0–4 h	0.49	0.58	0.85	0.3964
16–20 h vs. 0–4 h	−0.16	0.60	−0.27	0.7853
20–24 h vs. 0–4 h	0.34	0.59	0.58	0.5633
GDM:Yes vs. No	6.82	1.33	5.14	<0.0001
Parity:≥1 vs. 0	5.84	0.38	15.56	<0.0001
Induction mode:
Oxytocin vs. Cytotec	−1.60	0.55	−2.93	0.0034
Both vs. Cytotec	9.21	0.53	17.25	<0.0001
Delivery_Method:C-section vs. Vaginal	3.97	0.41	9.77	<0.0001
Infant weight at delivery (grams)	0.00	0.00	5.83	0.0000
Induction_4HRinc4-8:GDMYes	−5.89	2.32	−2.54	0.0112
Induction_4HRinc8-12:GDMYes	−5.86	2.09	−2.80	0.0051
Induction_4HRinc12-16:GDMYes	−7.00	1.68	−4.16	0.0000
Induction_4HRinc16-20:GDMYes	−6.19	1.74	−3.56	0.0004
Induction_4HRinc20-24:GDMYes	−7.13	1.80	−3.96	0.0001
Approximate significance of smooth term (maternal age)
	EDFa	Ref.dfb	F statistic	p
S(Maternal_Age)	1.23	1.43	0.68	0.3710
Deviance explained = 46.4%, adjusted R2 = 46.1%

^aEDF (Effective Degrees of Freedom) from the GAM model represents the flexibility or complexity of a smooth term in the model. EDF ≈1:The smooth term is approximately linear. EDF >1:The smooth term is non-linear, with higher values indicating greater flexibility.

^bRef.df (reference degrees of freedom) represents the degree of freedom that would be used if the smooth term were treated as a parametric term.

To understand how the factors in our GAM model work together, we compared the average labor duration across different groups of the factors that have a significant interaction. We found that women with GDM experience a longer labor duration than those without GDM, when labor is induced between 00 and 04 h in the groups with both vaginal and cesarean section combined (Figure 5C) and vaginal delivery only (Figure 5D). When labor induction is initiated between 00 and 04 h the average labor lasted about 6.82 h longer in the GDM group (Vaginal delivery and Cesarean section combined, Table 3), and 7.33 h (Vaginal delivery only, Table S3), a difference that was highly statistically significant (Vaginal delivery and Cesarean section combined, p < 0.0001; Vaginal delivery only, p < 0.0001).

Table 3.

Multiple comparisons of estimated marginal means of labor duration between GDM and non-GDM across different induction time intervals in Sparrow cohort (N = 2367). The endpoint is for vaginal delivery and Cesarean section combined.

Contrast (No vs. Yes)	Induction Time Interval	Estimate	SE	t.ratio	adj.p
No - yes	0–4	−6.82	1.33	−5.13	<0.0001
No - yes	4–8	−0.94	1.90	−0.49	0.6224
No - yes	8–12	−0.97	1.62	−0.60	0.5495
No - yes	12–16	0.18	1.04	0.17	0.8654
No - yes	16–20	−0.63	1.12	−0.56	0.5742
No - yes	20–24	0.31	1.22	0.25	0.8003

When we compared average labor induction duration for each induction frame within each group (GDM vs. non-GDM, vaginal delivery, and Cesarean section combined), we found clear differences in labor duration. For women with GDM, starting labor induction between 00 and 04 h resulted in the longest labor duration (the differences of labor duration between the 00–04 h interval and all other intervals = 6.36–7.27 h, all p-values ≤0.005). The 04–08 h interval presents the shortest labor duration (Table S4). In contrast, for women without GDM, no significant difference of labor duration between the different groups were identified (Table S4, Figure 5C). Similar trends were identified in the smaller cohort of women with vaginal deliveries only (Figure 5D, Table S5). Finally, to have a more detailed analysis of the time of day-difference in labor duration between women with and without GDM in the subgroup of women who were induced with only oxytocin, we completed a sub-analysis of these women (Table S6). As shown in Table S7, with a GAM modeling analysis and adjusting for other covariates (GDM, Parity, BMI, Delivery Method, infant weight at delivery, and interaction between induction time and GDM), there was no significant difference in labor duration across the groups. Meanwhile, compared to non-GDM, GDM increased labor duration by 8.00 h (SE = 1.73 and p < 0.0001, Table S7). The interaction between induction time and GDM also significantly contributed to the labor duration (all p-values ≤0.0016, Table S7). The covariate ‘Fetal Sex’ did not significantly contribute to the outcome in the GAM model (data not shown) and thus was removed from the final model. Importantly, we identified again significantly increased labor duration in the GDM patients compared to the non-GDM peers in the 00–04 h induction time interval (estimated marginal mean of labor duration = 8.00 with SE = 1.73 and adjusted p < 0.0001, Table S8, Figure 5E). It should be noted that there was no significant difference in the number of women induced with oxytocin, Cytotec, or both across the different time windows of the day. In contrast, the number of women induced during each time window of the day was different, with most women being initiated for labor induction between noon and midnight (Two-way p < 0.0001, Figure 5F).

4. Discussion

Induction of labor in diabetic pregnancies remains clinically challenging, often complicated by reduced responsiveness to uterotonic agents. Our data identified BMAL1 as a key regulator of temporal uterine OXTR function in mice, offering a new molecular target to optimize labor induction. Indeed, our retrospective human data show that time-of-the day of labor induction plays a role in labor duration in women with and without GDM.

4.1. BMAL1 regulation of Oxtr expression defines the daily uterotonic efficacy of oxytocin in mice

Understanding how the molecular clock shapes tissue-specific drug responses is fundamental to advancing precision medicine. Here we revealed that Oxtr is under BMAL1 transcriptional control and confirm that in the Bmal1 cKO mouse, Oxtr is significantly reduced at ZT17, compared to control. This finding contrasts with a previous study, which did not identify a difference in Oxtr expression [29]. This discrepancy could be due to the sample harvest time, where we found very low levels of Oxtr at ZT5, which were comparable between control and Bmal1 cKO. Independent of Oxtr changes in the Bmal1 cKO, changes in OXTR signaling could also be impacted by time-of-day and gestational diabetes. Indeed, BMAL1 targets a broad range of transcripts which can impact OXTR signaling and localization [44], including signaling molecules (such as RhoA), inflammatory signaling molecules (such as CCL2 and Il6) and proteins regulating receptor localization (such as β Arrestin) [[45], [46], [47]].

Of note, the Bmal1 cKO presented with a higher rate of spontaneous uterine contractions than controls. This increase in spontaneous contractions was not associated with Oxtr expression, which was lower at all the studied timepoints. Although it is of interest to identify why the Bmal1 cKO uterus is hypercontractile, this topic is outside the scope of the presented work. Significantly, the time-of-day difference in Oxtr expression in the control was associated with the capacity of the OXTR antagonist, atosiban, and the OXTR agonist, oxytocin, to modulate uterine contractions. Using the Bmal1 cKO mouse, we demonstrated that loss of BMAL1 causes a loss of the time-of-day changes in the uterotonic efficacy of oxytocin, demonstrating that BMAL1 is required to generate the increased sensitivity to oxytocin at ZT17. These findings support that BMAL1 defines time-of-day specific efficacy of drugs targeting OXTR and may explain the variable efficacies in clinical studies using OXTR ligands [48,49].

4.2. A mouse model of food-induced gestational diabetes (FID) shows an association between uterine molecular clock downregulation with reduced oxytocin-promoted contractions

Diabetes during pregnancy is known to deregulate the molecular clock [22,50,51], and is associated with uterine dysfunctions, including diminished contractile response to oxytocin [3]. Identifying the molecular contributors to the abnormal uterine contractions in GDM is required to develop strategies to prevent or overcome these challenges. Here we identify for the first time that a mouse model of food-induced gestational diabetes (FIN), and moderate overweight without gestational diabetes (FN) drives a down-regulation of Bmal1 and additional molecular clock transcripts in the GD19 uterus (term mouse pregnancy is ∼GD19.5). Based on our finding that the Bmal1 cKO has reduced Oxtr expression and reduced sensitivity to oxytocin, as expected, the FIN mouse also had reduced uterine contractions in response to oxytocin. However, the FN uterus retained its sensitivity to oxytocin, despite a downregulation of Bmal1. This indicates a complex interaction between molecular clock expression and uterine sensitivity to oxytocin and highlights some significant differences between uterine function in mice on a high fat, high sucrose diet with and without failed GTT. A recent study reported similar data demonstrating mice on a long-term high-caloric diet, causing obesity, significantly reduced uterine contractions before and during labor in vivo [52]. This study linked the mice reduced uterine contractile function to increased long-chain fatty acid uptake by the myometrium. This link is relevant because altered lipid handling, including excess fatty acid uptake, can interfere with cellular energy metabolism and has been shown to be present in conditions with metabolic disorders, which can disrupt circadian rhythms [53]. Thus, in metabolic disorders, impaired circadian rhythms and abnormal fatty acid metabolism may act together to impair myometrial contractility. In the previously published high-calorie diet study, there were no significant reduction in Bmal1, Clock, Cry2, or Oxtr expression at the studied timepoint (not shown, GSE268397) [52], a difference potentially linked to different diets and durations of diets in the two studies. Together, these results show a differential impact of diet, body weight and gestational diabetes on uterine contractile function. Further studies are needed to fully understand the molecular mechanisms involved in diet and GDM-induced changes in uterine contractile function. Still, they indicate that disrupted molecular clock expression combined with a failed GTT reduces uterine sensitivity to oxytocin.

4.3. The time-of-day of induction of labor initiation defines induction of labor duration in a GDM-dependent manner

While there is evidence in women that spontaneous uterine contractions and sensitivity to oxytocin change depending on the time of day [9,54], little is known about what regulates this daily change in uterine function. In agreement with this prior work, we show that human myometrial cells in vitro have defined daily time windows of increased contractile response to oxytocin. Our work in a larger human cohort identified a circadian rhythm to labor duration based on the time of labor induction [55]. Here, we extend this line of inquiry by focusing on women with GDM, where labor duration is significantly longer than that of non-diabetics. Although the retrospective clinical study was not designed to measure individual patients’ responsiveness to pharmacologically administered oxytocin, total induction of labor duration represents responsiveness to both endogenous and exogenous oxytocin. In our cohort, we found that when labor induction was initiated from midnight to 4 AM (00–04 h interval), labor duration was significantly longer (∼7 h) in women with GDM than in non-diabetic women. Importantly, we found that the shortest labor duration was obtained independently of the presence of GDM when labor induction was initiated between 8AM and early afternoon. Together, these data identify a potential benefit in considering the time of day of labor induction for women with and without GDM, and specifically show that inducing women with GDM closer to midnight should be avoided due to the significantly longer labor duration resulting from induction at this time of day. Interestingly, the decision to initiate labor induction with oxytocin alone, Cytotec alone, or both was not influenced by the time of day. Understandably, women who require both Cytotec and oxytocin had the longest labors. Future work will determine if labor duration can be influenced by the timing of oxytocin in patients who receive mechanical cervical ripening with Cytotec. Despite these exciting findings, it is essential to note some limits of the retrospective clinical cohort study. We were unable to stratify labor duration by admission bishop score, mechanical ripening (cervical balloon) or compare separately the duration of cervical ripening, latent phase, or active phase, all of which are important in the assessment of labor management [56,57]. To this end, the cohort was designed to be as homogeneous as possible. We included only gestations greater than 39 weeks, which generally excludes significant medical and fetal co-morbidities requiring earlier induction of labor. Our data shows that across all examined timeframes, patients had similar mean gestational ages, and there was no departmental policy for induction of labor scheduling based on Bishop score, parity, or patient request, which serves as our proxy for similar bishop scores and the basis for comparing induction of labor durations. We acknowledge that individual practitioners may have had scheduling preferences that could have contributed to potential bias. Future prospective studies will be required to determine if our identification of an in vitro time-of-day difference in oxytocin efficacy in a human cell line translates to the clinic.

5. Conclusion

Our findings identify that the time of day of oxytocin administration in mice, or initiation of labor induction in women, defines uterine contraction frequency in mice and labor duration in women. This work highlights the importance of time of day for initiation of labor induction in pregnancies with and without GDM.

CRediT authorship contribution statement

Thu Van-Quynh Duong: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Alexandra M. Yaw: Writing – review & editing, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Guoli Zhou: Writing – review & editing, Visualization, Software, Methodology, Investigation, Data curation, Conceptualization. Niharika Sinha: Writing – review & editing, Methodology, Investigation, Data curation. Aneesh Sai Cherukuri: Writing – review & editing, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Duong Nguyen: Writing – review & editing, Supervision, Methodology, Data curation. Kylie Cataldo: Writing – review & editing, Formal analysis. Nicollette Ly: Formal analysis, Data curation. Aritro Sen: Writing – review & editing, Conceptualization. Lorenzo F. Sempere: Writing – review & editing, Formal analysis, Data curation. Cara Detrie: Writing – review & editing, Conceptualization. Robert Seiler: Writing – review & editing. I. Nicholas Olomu: Writing – review & editing, Data curation, Conceptualization. Rene Cortese: Writing – review & editing, Formal analysis, Conceptualization. Robert Long: Writing – review & editing, Formal analysis, Conceptualization. Hanne M. Hoffmann: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Funding sources

This work was funded by the USDA National Institute of Food and Agriculture Hatch project MICL1018024 (H.M.H.), March of Dimes Grant no 5-FY19-111 (H.M.H.), the NIH/National Institute of Environmental Health Sciences project R01ES035691 (H.M.H.), and a sub-award from the Michigan Diabetes Research Center through the National Institute of Diabetes and Digestive and Kidney Diseases P30DK020572 (H.M.H.). A.M.Y. was supported by the Eunice Kennedy Shriver National Institute of Child Health & Human Development (NICHD) of the National Institutes of Health under Award Numbers F32HD107852 and K99HD113843 with additional support of training by T32HD087166.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.