Proximity to Delivery Alters Insulin Sensitivity and Glucose Metabolism in Pregnant Mice

During pregnancy, maternal metabolism adapts to support offspring growth. In particular, there are changes in insulin sensitivity, which affects the availability and fate of nutrients in both mother and conceptus (1,2). The specific adaptations depend on the stage of pregnancy, as metabolic demands increase with expansion of the gravid uterus (3,4). In humans and rats, early pregnancy is a period of lipid accumulation and unchanged or increased insulin sensitivity, whereas later pregnancy is characterized by lipid mobilization and insulin resistance, common features of overt type 2 diabetes (1,2,5,6). Indeed, whole-body resistance to the hypoglycemic action of insulin has been reported during late pregnancy in a wide range of species, including rabbits, dogs, sheep, and horses, as well as rats and humans (7–13). It is often accompanied by reduced maternal glucose utilization, particularly in skeletal muscle, although there is less consensus about the actions of insulin on hepatic glucogenesis during pregnancy (10,12,14,15). Most studies of insulin sensitivity during late pregnancy have been performed between 60 and 85% of gestation, with few measurements closer to term when fetal nutrient demands are maximal yet maternal nutrient requirements may also be changing in preparation for the imminent onset of labor and lactation.

Total conceptus mass varies not only with increasing gestational age but also between species both in total and as a percentage of maternal mass (16). Fetal growth rate is high during late mouse pregnancy and results in the gravid uterus accounting for 30% of maternal mass at term (16). This is a higher percentage than found in monotocous species like humans and sheep (5–9%) or the values of 12–25% seen in other polytocous animals like dogs, pigs, and rats (16). Despite this, little is known about insulin sensitivity in pregnant mice, even though they are used widely in genetic and developmental studies in which variations in maternal insulin sensitivity may affect the ensuing offspring phenotype (17,18). The molecular basis of insulin resistance during late pregnancy is also still poorly understood in many species. Tissue insulin receptor (IR) abundance appears to be unaffected by pregnancy, although there is evidence for defects in the early stages of insulin signal transduction downstream of the IR in skeletal muscle of both rats in late pregnancy and pregnant women insulin resistant due to obesity or gestational diabetes (19–23). Hence, the aims of this study were to measure glucose-insulin dynamics, whole-body insulin sensitivity, and tissue insulin signaling proteins in nonpregnant (NP) mice and in pregnant dams in late gestation and close to term. C57B1/6 mice were used in this study because this strain has been used extensively to investigate the genetic and environmental regulation of feto-placental development (17,24).

C57Bl/6 females (n = 123) were group housed at 21°C under 12-h dark/12-h light conditions with free access to water and food (RM3; Special Diet Services). Aged 8–12 weeks, females were time mated with C57Bl/6 males with the presence of a copulatory plug defined as day 1 (D1) of a D20.5 pregnancy. Pregnant (n = 77) and remaining NP (n = 46) mice were weighed every 5 days while food intake was measured every 3 days. Mice were allocated to one of the following procedures: 1) tissue and blood collection (n = 30), 2) DEXA (n = 29) scanning, 3) glucose tolerance tests (GTTs) (n = 28) or insulin tolerance tests (ITTs) (n = 20), or 4) a hyperinsulinemic-euglycemic clamp (HEC) together with D-[³H]glucose and 2-deoxyglucose (2DG) administration (n = 18). All experiments were performed under the U.K. Animals (Scientific Procedures) Act 1986 after ethical review by the University of Cambridge.

Between 0800 and 1000 h in fed conditions, pregnant dams at D16 and D19 and age-matched NP females (n = 10 mice per group) were weighed and anesthetized (10 µL/g of fentanyl-fluanisone:midazolam in sterile water, 1:1:2, i.p.; Jansen Animal Health). A cardiac blood sample was taken before euthanasia by cervical dislocation. In NP mice, liver, heart, kidneys, skeletal muscle (biceps femoris), and retroperitoneal fat were dissected, weighed, and snap frozen in liquid nitrogen. In pregnant dams, the gravid uterus, hysterectomized carcass, and individual fetuses and placentas were weighed before maternal tissue collection. Blood glucose concentrations were measured on a handheld glucometer (One Touch Ultra; LifeScan, Livingstone, U.K.). After centrifugation, the plasma was stored at −20°C to measure metabolite and hormone concentrations.

Whole-body fat and lean mass content were determined by DEXA scanning (Lunar PIXImus densitometer; GE Lunar Corp., Madison, WI) in intact NP females (n = 10) and hysterectomized pregnant mice killed by cervical dislocation between 0800 and 1000 h (D16, n = 7; D19, n = 10). Values were expressed as a proportion of total body weight in NP mice and of hysterectomized weight in pregnant dams.

Conscious NP and pregnant mice received either a GTT (NP, n = 11; D16, n = 9; D19, n = 8) or ITT (n = 6–8 mice per group) after fasting from 0800 h for 6 h or 3.5 h, respectively. Blood samples (≤5 μL) were taken from the tail vein immediately before intraperitoneal administration of either glucose (10% weight for volume, 1 g/kg body weight) or insulin (0.25 units/kg, human insulin, Actrapid; Novo Nordisk) and thereafter at 15–30-min intervals for 120 min to measure blood glucose concentrations as above. The mice were then killed by cervical dislocation.

The HEC was performed as described previously (25). In brief, NP and pregnant mice fasted for 2.5 h (NP, n = 7; D16, n = 5; D19, n = 6) were anesthetized with a mixture of ventranquil:dormicum:fentanyl (1:2:10 in 3 units of water, 10 μL/g body weight, i.p.; Janssen-Cilag, Tilburg, the Netherlands) and maintained at 37°C using a servo-controlled thermopad (Harvard Apparatus, U.K.). After catheterizing a tail vein, d-[³H]glucose was infused continuously (0.006 MBq/min in PBS, 50 μL/h, i.v., 370–740 GBq/mmol; PerkinElmer, Seer Green, U.K.). After steady state was achieved at 60 min (basal state, ∼3.5 h fasted), two blood samples (≤50 μL each) were taken 10 min apart from the tail. Insulin was then injected as a bolus (3.3 mU, i.v., human insulin, Actrapid) followed by infusion (0.09 mU/min) together with the d-[³H]glucose. Blood glucose levels were monitored every 5 min for the first 20 min after insulin administration and then at 10-min intervals until the end of the protocol (≤5 µL per sample). When a decrease in blood glucose concentration was detected 5–10 min after beginning the insulin infusion, a variable glucose infusion rate (GIR) (12.5% weight for volume PBS, i.v.; Sigma-Aldrich) was begun and adjusted every 5–10 min thereafter to maintain blood glucose concentrations at mean basal levels. At 50 min after insulin infusion, 2-deoxyglucose ([¹⁴C]2DG, specific activity 9.25–13.0 GBq/mmol; PerkinElmer) was injected intravenously. By 70 min of insulin administration, blood glucose levels were clamped at basal concentrations and a further three blood samples (≤50 μL each) were collected from the tail at 10-min intervals. The mice were then killed by cervical dislocation and samples of biceps femoris and retroperitoneal fat were collected from all animals together with the fetus and placenta adjacent to the cervix in each horn (n = 2 fetuses and placentas per litter) for analysis of tissue [¹⁴C]2DG content. The rates of glucose utilization and production in basal and hyperglycemic states together with whole-body and hepatic sensitivity to insulin were calculated as described previously (25,26).

Plasma d-[³H]glucose concentrations were measured by scintillation counting (Hidex 300SL; LabLogic Ltd., Sheffield, U.K.), after samples were deproteinized with 20% trichloric acid and dried to eliminate tritiated water. Plasma leptin and insulin concentrations in the fed state were measured simultaneously using a 2-plex specific immunoassay (Meso Scale Discovery). The interassay coefficients of variation (CVs) were 10.8 and 9.7%, respectively. Plasma insulin concentrations during the HEC were measured by ELISA (90090; Crystal Chem, Inc.), which detected both murine and human insulin. The interassay CV was ≤10%. Plasma insulin-like growth factor 1 (IGF1) levels were also measured by ELISA (ImmunoDiagnostic Systems) with an interassay CV of 4.6%. Enzymatic assay kits were used to determine the plasma concentrations of triglycerides (TGs), total cholesterol (CV 5.5 and 4.9%, respectively; Siemens Healthcare), and free fatty acids (FFAs) (CV 4.5%; Roche).

Hepatic glycogen content was measured enzymatically, using amyloglucosidase as reported previously (27). The total fat content of the liver and pooled samples of skeletal muscle were measured using the modified Folch method (28). To determine tissue phosphorylated 2DG (p2DG) content, tissues were homogenized in 0.5% perochloric acid and the homogenates neutralized to separate p2DG from 2DG by precipitation, as described previously (29).

Proteins were extracted (∼100 mg; NP, n = 10; D16, n = 5; D19, n = 5–6) from the liver and skeletal muscle and quantified using Western blotting as described previously (30). Successful transfer and equal protein loading was confirmed by Ponceau S staining of membranes before incubation with antibody (Table 1). Protein abundance was determined by measuring pixel intensity of the protein bands using ImageJ analysis software (National Institutes of Health, Bethesda, MD).

All statistical analyses and calculations were performed on the GraphPad Prism 4.0. For most of the data, differences between NP, D16, and D19 animals were analyzed by one-way ANOVA with Bonferroni post hoc test. When the ANOVA indicated an effect of pregnancy, differences between D16 and D19 of pregnancy were assessed separately by unpaired Student t test. During the HEC, changes within the same group were assessed by paired Student t test or by t test of the mean change differing from zero. For GTT and ITT protocols, the changes in glucose concentration were analyzed by two-way ANOVA with time as a repeated measure. The area above the curve (AAC) in the GTT and area under the curve (AUC) in the ITT for the changes in glucose concentrations were calculated using the trapezoid rule. Fetal and placental data were averaged for each litter before calculation of mean values at each gestational age.

Pregnant dams were heavier than NP females due to the gravid uterus and increased weights of several maternal tissues; however, when weights were expressed as a percentage of total or hysterectomized body weight, only the liver and retroperitoneal fat pads were proportionately heavier during pregnancy (Table 2). However, DEXA scanning showed no significant changes in body fat content during pregnancy (Table 2). As expected (31), the gravid uterus and individual fetuses weighed less while the placentas weighed more at D16 than D19 with no difference in litter size between the two groups (Table 2). Hepatic fat and glycogen content were higher in total during pregnancy in line with the increased tissue weight but not when expressed per gram tissue (Table 2). Skeletal muscle fat content of pooled samples appeared to be greater in pregnant than NP animals when expressed per gram tissue (Table 2). Pregnant dams increased their food intake relative to NP females from D9 of pregnancy (data not shown).

Blood glucose concentrations were unaffected by pregnancy in the fed state but were significantly lower than NP values in both pregnant groups after 3.5 h of fasting (Table 3). At 6 h of fasting, blood glucose levels were similar to those seen at 3.5 h of fasting in NP and D19 groups but were significantly higher than at 3.5 h of fasting in D16 dams (Table 3). In fed animals, plasma FFA concentrations were significantly lower in both pregnant groups than in NP females, with the lowest values in D16 dams (Table 3). Cholesterol concentrations were also significantly lower during pregnancy and declined significantly between D16 and D19 (Table 3). Insulin concentrations were significantly higher in D16 dams than NP females, with intermediate values in D19 dams (Table 3). Plasma leptin concentrations were higher whereas plasma IGF1 levels were lower in pregnant than NP groups, with no significant differences with gestational age (Table 3).

The increment in blood glucose concentrations, the time course of the concentration changes, and the AUC did not differ with pregnancy or gestational age (Fig. 1A). Blood glucose concentrations remained elevated for the entire 120 min after glucose injection in all three groups (Fig. 1A).

Blood glucose concentrations were significantly lower than baseline 15 min after acute insulin injection in all three groups and remained depressed for up to 120 min (Fig. 1B). However, the decrement in blood glucose concentrations was greater in NP than both pregnant groups from 15 to 90 min after insulin injection (Fig. 1B). In D16 dams, blood glucose levels returned to baseline by 60 min and were significantly greater than baseline 120 min after insulin injection (Fig. 1B). The AAC also differed significantly with pregnancy and gestational age and indicated that pregnant dams were less sensitive to insulin than NP females, particularly at D16 (Fig. 1B).

Whole-body and hepatic insulin sensitivity were investigated in more detail by HEC coupled with [³H]glucose infusion. In all three groups, blood glucose levels were clamped at basal, euglycemic levels by 70–90 min after beginning insulin infusion (Fig. 2A). At this time, plasma insulin concentrations were within the postprandial range and five to sixfold higher than the basal values in all three groups (Fig. 2B). However, steady-state insulin concentrations during the clamp were significantly lower in the D19 than NP or D16 groups (Fig. 2B). Whole-body insulin sensitivity, measured as GIR, was lower in D16 dams than in the other two groups and similar in NP and D19 pregnant groups (Fig. 2C). The findings were identical when GIR was adjusted for the differences in the increment in insulin concentration during the clamp period (data not shown). Whole-body insulin sensitivity was also calculated as the difference between glucose utilization in hyperinsulinemic (R_d) and basal states (R_a). This difference varied widely between individuals, particularly in D16 dams, and was only a significant increment in NP females, indicative of insulin resistance in both D16 and D19 dams (Fig. 2D). During hyperinsulinemia, endogenous glucose production (EGP) continued at a significant rate in all three groups (P < 0.02, greater than zero, all groups, t test) and occurred at the highest rate in D16 dams (Fig. 2D). Hepatic insulin sensitivity, calculated as a significant difference in glucose production between basal and hyperinsulinemic states, was only detected in D19 dams and was greater in absolute value at D19 than D16 (Fig. 2D).

Whole-body glucose utilization was significantly greater in D19 dams than NP females in basal conditions (R_a), whereas during hyperinsulinemia (R_d), it was higher in D16 dams than NP females with intermediate values in D19 dams (Fig. 2C). Tissue glucose utilization during hyperinsulinemia, measured as p2DG content, in skeletal muscle and adipose tissue varied widely, particularly in NP females, but was not significantly different between pregnant and NP animals or between D16 and D19 dams (Table 4). Fetal p2DG content was similar to that of maternal tissues at both ages (Table 4). Placental p2DG content was significantly higher than in the fetus or maternal tissues at D19, but not at D16 (Table 4).

The hepatic insulin-IGF signaling pathway was largely downregulated in D16 dams compared with NP females (IR, IGF1 receptor [IGF1R], pAkt T308, and phosphorylated mitogen-activated protein kinase [pMAPK]) (Fig. 3A). In contrast, at D19, the insulin signaling pathway was largely upregulated, compared with NP and D16 pregnant groups (IGF1R, total Akt, pAkt S473, total Gsk3, pGsk3, total S6 kinase [S6K], total MAPK, and pMAPK) (Fig. 3A). Pregnancy had less effect on insulin signaling pathways in skeletal muscle, with increased abundance of p85α at D16 and of pAkt T308 and total MAPK at D19 compared with NP values and decreased total Akt and pS6K abundance at D19 compared with either NP or D16 values (Fig. 3B).

Hepatic abundance of lipogenic sterol regulatory element binding protein (SREBP) was reduced at D16 but normalized by D19, whereas peroxisome proliferator–activated receptor α (PPARα), peroxisome proliferator–activated receptor γ (PPARγ), and lipoprotein lipase (LPL) were reduced in both pregnant groups relative to NP values (Fig. 3C). Hepatic abundance of PPARγ and LPL was significantly lower in D19 than D16 dams (Fig. 3D). Hepatic fatty acid transport protein 1 (FATP1) was lower in D19 dams than NP females (Fig. 3C). In skeletal muscle, SREBP was reduced in D16 but not D19 dams whereas PPARγ and LPL abundance was lower in both pregnant groups relative to NP values (Fig. 3D). Skeletal muscle expression of LPL was greater and FATP1 was less at D16 than D19 (Fig. 3D).

This is the first study to measure insulin sensitivity of glucose metabolism in pregnant mice and shows significant changes in insulin-glucose dynamics with both pregnancy and gestational age over the last 20% of gestation. In particular, there were changes in insulin concentration and glucose utilization and production and in whole-body and tissue insulin sensitivity between D16 and D19 of pregnancy. Protein abundance in the hepatic and skeletal muscle insulin signaling pathways also differed during pregnancy in line with the gestational changes in insulin sensitivity and glucose metabolism. Overall, insulin resistance and EGP capacity were more pronounced at D16 than D19, which indicates that mice adapt their metabolic strategy for supporting pregnancy as delivery approaches. There were also changes in body composition, tissue lipogenic signaling proteins, and circulating concentrations of leptin and IGF1 during pregnancy, which indicate that like other species (2,5), mice accumulate fat in specific deposits during early but not late pregnancy and are resistant to the actions of leptin, as well as insulin, in late pregnancy, when their food intake increases despite hyperleptinemia.

Whereas glucose tolerance was unaffected by pregnancy, insulin resistance was evident in D16 pregnant mice. These dams were hyperinsulinemic in both fed and fasted states, had a significantly smaller hypoglycemic response to acute insulin administration, and required 80–85% less exogenous glucose to maintain euglycemia during the HEC than the other two groups of animals. This decrement in GIR from NP values was greater than seen in rats, dogs, and women at an equivalent stage of pregnancy (7,9,12). In addition, insulin failed to inhibit EGP in D16 mouse dams. Similar findings have been made in pregnant rats, but in women, rabbits, and dogs with a proportionately smaller gravid uterus, insulin continues to reduce EGP during late pregnancy, although not always as effectively as in the NP state (9,10,12,14,15). Furthermore, whole-body glucose utilization varied widely and did not increase significantly during hyperinsulinemia in D16 pregnant mice, unlike the significant increment seen in NP females. Although there were no significant changes in tissue p2DG content, skeletal muscle expression of p85α, a known regulator of human muscle insulin resistance (32), was increased whereas hepatic expression of IR and several downstream insulin signaling proteins were decreased in D16 mouse dams, consistent with their reduced whole-body and liver insulin sensitivity. Collectively, these findings indicate that insulin resistance occurs at whole-body, tissue, and molecular levels at D16 of mouse pregnancy.

By D19, insulin concentrations in fed and fasted states had returned to NP levels and the hypoglycemic response to acute insulin administration was greater than at D16, although still less than in NP females. Insulin concentrations were also lower during the clamp in D19 dams, which suggests increased insulin clearance consistent with findings in dogs during late pregnancy (12). A degree of hepatic insulin sensitivity was restored in D19 pregnant mice as insulin infusion now reduced EGP. Hepatic expression of several proteins in the insulin signaling pathway were also upregulated at D19 relative to the other two groups. Improved hepatic insulin signaling at D19 is also suggested by normalized expression of the insulin-regulated transcription factor SREBP (33). Furthermore, the GIR required to maintain euglycemia during hyperinsulinemia was significantly greater at D19 than D16, indicative of improved whole-body insulin sensitivity near term. However, tissue p2DG content showed no change between D16 and D19, and whole-body glucose utilization was unresponsive to insulin at D19, which suggests that a degree of insulin resistance still persists in pregnant mice close to delivery. These apparent contradictions probably reflect the noninsulin-dependent, transplacental flux of glucose, which would increase GIR and lead to overestimation of the actual maternal insulin sensitivity, particularly when fetal glucose demands are high near term (2,34). Insulin sensitivity, therefore, appears to change differentially in individual maternal tissues with proximity to delivery in pregnant mice.

In basal conditions, the rate of EGP (R_a) in the NP females fasted for 3.5 h in the current study was higher than that seen previously in older male mice fasted overnight (25). Insulin also had little effect on the rate of EGP in NP females in the current study compared with its inhibitory actions in males published previously (25). Whether these differences are sex linked or due to the differing ages and length of fasting remain unclear, but insulin sensitivity is known to be greater in juvenile than adult animals and in adult women than men (35,36). In the current study, basal glucose production (R_a) was higher in pregnant than NP females as seen previously in dogs, sheep, and women during late pregnancy (11,12,14). In mice, this is consistent with the greater total availability of hepatic glycogen associated with the increased relative size of the liver during pregnancy. Particularly at D16, EGP was activated readily and sustained for several hours. Blood glucose concentrations were higher after fasting for 6 h than 3.5 h and could be increased above pretreatment levels 2 h after initiating an acute hypoglycemic challenge at D16 but not D19. EGP was also significantly higher during hyperinsulinemia in D16 dams than in the other two groups. Since hepatic glycogen content and activity of glucose-6-phosphatase are similar at D16 and D19 (37), the results indicate that, in addition to differences in hepatic insulin sensitivity, there may also be a greater gluconeogenic capacity and/or a more robust counterregulatory response to hypoglycemia at D16 than D19. Mice may, therefore, appear to rely more heavily on glucose production to meet the glucose demands of the gravid uterus at D16 than D19. Although the causes of the changes in insulin sensitivity and glucose production with proximity to delivery remain unknown, they closely parallel gestational changes in maternal concentrations of corticosterone, a known insulin antagonist (4,37), with maximal values at D16 and progressively declining concentrations thereafter toward term (38).

Fetal glucose utilization tended to be less at D19 than D16, consistent with the changes in glucose metabolism seen in fetal sheep toward term (39). At D19, weight-specific rates of glucose utilization by the fetus (∼39 μmol/min/kg) and placenta (∼120 μmol/min/kg), estimated from their p2DG contents, were within the range of values reported previously for other species at ≥90% of gestation (40–42). Estimation of total feto-placental glucose utilization by the whole litter indicates that this increases by 15–20% between D16 and D19 while total conceptus mass increases by 120%. Mouse pups must, therefore, be using nutrients other than glucose to sustain their growth rate during late gestation. Indeed, previous studies have shown that their growth is positively correlated to placental glucose transport at D16, but not at D19, and becomes progressively more dependent on placental amino acid transport toward term (43). A greater ATP requirement for active amino acid transport by the D19 placenta is consistent with its high rate of glucose utilization relative to other fetal and maternal tissues at this age. In addition, the lower maternal FFA and cholesterol concentrations during pregnancy indicate that lipids may also provide alternative substrates for feto-placental tissues during late gestation (44).

In summary, the marked insulin resistance of glucose utilization and production at D16 will favor fetal glucose acquisition when fetuses are entering their maximal growth phase (31). When the fetuses have nearly reached term weight and can use other substrates at D19, insulin sensitivity of maternal tissues like the liver improves, which will conserve proportionally more glucose for maternal use in anticipation of the imminent demands of labor and lactation. This late change in insulin-glucose dynamics is likely to be particularly important in mice in which the gravid uterus accounts for such a large proportion of maternal weight at term but may also have a significant role in the maternal metabolic preparations for delivery in other species when the primary site of maternal-offspring nutrient allocation switches from the uterus to the mammary gland.

Acknowledgments. The authors thank the staff of the animal facilities for their care of the mice.

Funding. The authors are grateful to the Medical Research Council for funding the research through a studentship to B.M. and an in vivo skills award (MR/J500458/1 and MRC CORD G0600717).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. B.M. designed the study, collected the tissues, and performed the clamp studies, GTTs, ITTs, and Western blotting. D.S.F.-T. and P.V. performed the clamp studies. O.R.V. collected the tissues. S.E.O. designed the study. A.N.S.-P. designed the study, collected the tissues, and performed the GTTs and ITTs. A.L.F. designed the study, collected the tissues, and performed the clamp studies, GTTs, and ITTs. All authors contributed to writing the paper. A.L.F. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.