Sustained maternal inflammation during the early third-trimester yields intrauterine growth restriction, impaired skeletal muscle glucose metabolism, and diminished β-cell function in fetal sheep

Abstract

Maternal inflammation causes fetal intrauterine growth restriction (IUGR), but its impact on fetal metabolism is not known. Thus, our objective was to determine the impact of sustained maternal inflammation in late gestation on fetal inflammation, skeletal muscle glucose metabolism, and insulin secretion. Pregnant ewes were injected every third day from the 100th to 112th day of gestation (term = 150 d) with saline (controls) or lipopolysaccharide (LPS) to induce maternal inflammation and IUGR (MI-IUGR). Fetal femoral blood vessels were catheterized on day 118 to assess β-cell function on day 123, hindlimb glucose metabolic rates on day 124, and daily blood parameters from days 120 to 125. Fetal muscle was isolated on day 125 to assess ex vivo glucose metabolism. Injection of LPS increased (P < 0.05) rectal temperatures, circulating white blood cells, and plasma tumor necrosis factor α (TNFα) concentrations in MI-IUGR ewes. Maternal leukocytes remained elevated (P < 0.05) and TNFα tended to remain elevated (P < 0.10) compared with controls almost 2 wk after the final LPS injection. Total white blood cells, monocytes, granulocytes, and TNFα were also greater (P < 0.05) in MI-IUGR fetuses than controls over this period. MI-IUGR fetuses had reduced (P < 0.05) blood O₂ partial pressures and greater (P < 0.05) maternofetal O₂ gradients, but blood glucose and maternofetal glucose gradients did not differ from controls. Basal and glucose-stimulated insulin secretion were reduced (P < 0.05) by 32% and 42%, respectively, in MI-IUGR fetuses. In vivo hindlimb glucose oxidation did not differ between groups under resting conditions but was 47% less (P < 0.05) in MI-IUGR fetuses than controls during hyperinsulinemia. Hindlimb glucose utilization did not differ between fetal groups. At day 125, MI-IUGR fetuses were 22% lighter (P < 0.05) than controls and tended to have greater (P < 0.10) brain/BW ratios. Ex vivo skeletal muscle glucose oxidation did not differ between groups in basal media but was less (P < 0.05) for MI-IUGR fetuses in insulin-spiked media. Glucose uptake rates and phosphorylated-to-total Akt ratios were less (P < 0.05) in muscle from MI-IUGR fetuses than controls regardless of media. We conclude that maternal inflammation leads to fetal inflammation, reduced β-cell function, and impaired skeletal muscle glucose metabolism that persists after maternal inflammation ceases. Moreover, fetal inflammation may represent a target for improving metabolic dysfunction in IUGR fetuses.

Introduction

Low birthweight due to intrauterine growth restriction (IUGR) increases perinatal morbidity and mortality in livestock (Wu et al., 2006) and is associated with postnatal metabolic dysfunction that reduces G:F, increases fat deposition, and diminishes carcass merit (Greenwood et al., 2005). In humans, IUGR has been linked to obesity, hypertension, type 2 diabetes, and other metabolic and cardiovascular disorders that shorten lifespan and reduce quality of life (McMillen and Robinson, 2005; Barker et al., 2007). Fetal sheep studies in which IUGR was induced by maternal hyperthermia have identified several components of stress-associated fetal programming that result in metabolic dysfunction late in pregnancy and after birth (Yates et al., 2018). These include impairment of whole-body glucose oxidation (Limesand et al., 2007), β-cell function (Leos et al., 2010; Macko et al., 2016; Limesand and Rozance, 2017), islet development (Brown et al., 2016; Kelly et al., 2017), and insulin responsiveness (Limesand et al., 2007; Thorn et al., 2013; Camacho et al., 2017). Fetal adaptations that result in IUGR are often in response to hypoxemia and hypoglycemia caused by placental insufficiency (Greenwood and Cafe, 2007; Cox and Marton, 2009), and the resulting fetal hypercatecholaminemia is a well-known mediator of these developmental changes (Macko et al., 2013, 2016; Chen et al., 2014, 2017; Limesand and Rozance, 2017). However, elevated circulating concentrations of inflammatory cytokines have also been observed in IUGR fetal sheep, mice, and humans (Xu et al., 2006; Bertucci et al., 2011; Lausten-Thomsen et al., 2014). Thus, we postulate that chronic inflammation plays a role in metabolic programming of the IUGR fetus as well.

Studies by us and others show that skeletal muscle is disproportionately affected by IUGR fetal mechanisms aimed at metabolic thrift (Yates et al., 2014, 2016; Rozance et al., 2018). Skeletal muscle is a large consumer of glucose, and reducing its mass and metabolic rates repartitions nutrients to essential tissues (DeFronzo et al., 1981; Brown, 2014). We recently found that incubation of primary rat muscle with inflammatory cytokines disrupted insulin-signaling and insulin-stimulated glucose oxidation (Cadaret et al., 2017). Other in vivo and in vitro studies further indicate that inflammation reduces the glucose oxidative capacity of skeletal muscle (Frisard et al., 2015; Remels et al., 2015). Glucose oxidation is similarly reduced in pancreatic β cells from maternal hyperthermia-induced IUGR fetal sheep (Limesand et al., 2006). Because it is the impetus for glucose stimulus-secretion coupling, this in turn impairs glucose-stimulated insulin secretion (GSIS; Limesand and Rozance, 2017). Cytokines and other inflammatory factors inhibit insulin secretion by disrupting β-cell glucose oxidation in a similar fashion (Zhang and Kim, 1995; Mukhuty et al., 2017; Barlow et al., 2018; Lee et al., 2019), and thus we postulate that they contribute to hypoinsulinemia in the IUGR fetus. We recently found that chronic maternal inflammation in pregnant rats induced by serial administration of bacterial endotoxin resulted in IUGR fetuses that exhibited asymmetric growth and metabolic adaptations (Cadaret et al., 2019). Therefore, the objective of this study was to determine whether maternal inflammation-induced IUGR (MI-IUGR) fetal sheep exhibited chronic inflammation, metabolic dysfunction, and poor growth late in pregnancy. Our hypothesis was that MI-IUGR fetuses would develop skeletal muscle-centric reductions in glucose oxidation and insulin responsiveness as well as impaired insulin secretion. We further hypothesized that these observations would coincide with indicators of persistent fetal inflammation.

Materials and Methods

Animals and Experimental Design

This study was approved by the Institutional Animal Care and Use Committee at the University of Nebraska–Lincoln (UNL), which is accredited by AAALAC International. Polypay ewes were purchased from a single commercial source and at 18 to 24 mo of age were timed-mated to a single sire. Nutritional management, housing, and husbandry were practiced as previously described (Macko et al., 2016). Twenty ewes carrying twin pregnancies were identified via ultrasonography and assigned via simple randomization to the control group or MI-IUGR group. Four pregnancies were lost for undiagnosed reasons, leaving 9 ewes as controls and 7 ewes carrying MI-IUGR fetuses. The MI-IUGR fetuses were produced by injecting ewes with bacterial lipopolysaccharide (LPS) endotoxin (100 ng/kg bodyweight in 0.6-mL saline; Escherichia coli O55:B5; Sigma–Aldrich, St. Louis, MO) i.v. every 72 h from the 100th to the 112th day of gestational age (dGA). Ewes carrying control fetuses were injected every 72 h with 0.6-mL saline carrier only. Maternal body (rectal) temperatures and blood samples (jugular venipuncture) were collected 0, 3, 6, 12, 24, and 48 h after each injection to assess maternal responses to LPS administration. Fetal surgical preparations were performed on one fetus for each ewe on dGA 118 ± 1, and once-daily maternal (venous) and fetal (arterial) blood samples were collected from dGA 120 to 125 at 8:00. As described below, square-wave hyperglycemic clamps were performed on dGA 122 ± 1 and hyperinsulinemic-euglycemic clamps were performed on dGA 124 ± 1. Ewes were euthanized by double barbiturate overdose and fetuses were necropsied on dGA 125 ± 1. Organs and tissues were collected from the same fetus in which the in vivo studies were performed.

Fetal Surgical Hindlimb Preparation

Fetal femoral artery and vein catheters were surgically placed as previously described (Macko et al., 2016). Ewes were fasted overnight, induced by i.v. injection of ketamine (10 mg/kg) and diazepam (0.11 mg/kg), intubated, and maintained under anesthesia by inhalation of 1% to 5% isoflurane gas in oxygen. The hindlimbs of 1 fetus per ewe were exteriorized by partial cesarean and indwelling catheters (Tygon ND-100–80 Flexible Plastic Tubing; outer diameter 1.4 mm and inner diameter 0.9 mm, US Plastics, Lima, OH) were surgically placed in the descending aorta and inferior vena cava via the femoral artery and vein of the left hindlimb for blood sampling and intravenous infusions. A catheter was placed in the distal femoral vein of the right hindlimb with the tip advanced to the external iliac vein for blood sampling. The deep circumflex iliac artery and vein of the right hindlimb were ligated and severed to isolate blood flow to the external iliac artery and vein. Thus, in vivo metabolic studies were performed on the right hindlimb. A Precision S-series Flow Probe (2 to 4 mm; Transonic Systems, Inc., Ithaca, NY) was positioned around the external iliac artery of the right hindlimb. Catheters were filled with heparinized saline (30 U/ml, 0.9% NaCl, Nova-Tech, Inc., Grand Island, NE) and along with the flow probe cable were tunneled subcutaneously to the flank, exteriorized through a skin incision, and kept in a plastic mesh pouch sutured to the skin. Ewes were administered 6,600 U/kg penicillin G procaine, 2.2 mg/kg Ketofen, 10 mg/kg phenylbutazone, and 3 mg/kg Atropine at the time of surgery. Postoperative phenylbutazone was continued for 3 d following surgery, and ewes were allowed to recover for 4 d before performing studies. Catheters were flushed daily with heparinized saline.

Maternal and Fetal Blood Sample Analyses

Maternal venous blood was collected via jugular venipuncture and fetal arterial blood was collected via catheter into EDTA-treated syringes (~1.25 mL) and heparinized syringes (~0.25 mL). Concentrations of total white blood cells, lymphocytes, monocytes, and granulocytes, as well as red blood cells, hematocrit, hemoglobin, and platelets were determined from 125-µL aliquots of whole blood collected with the EDTA syringe with a HemaTrue Veterinary Hematology Analyzer (Heska, Loveland, CO) using the ovine software specifications as recommended by the manufacturer and as previously described (Cadaret et al., 2018). Blood plasma was then isolated from the remaining EDTA-treated whole blood by centrifugation at 14,000 × g for 5 min. Plasma was stored at −80 °C and commercial ELISA kits were used to determine concentrations of tumor necrosis factor α (TNFα; Sheep TNFα Fine Test, Wuhan Fine Biotech Co., Ltd., Wuhan, China) and insulin (Ovine Insulin, Alpco Diagnostics, Windham, NH) from duplicate aliquots as previously described (Yates et al., 2012a). Intra-assay and interassay coefficients of variance for both assays were less than 10%. Maternal and fetal insulin was measured for each daily sample. Maternal TNFα was measured 3 h after each injection and on dGA 120 and 125 only. Fetal TNFα was measured on dGA 120 and 125 only. Whole-blood concentrations of glucose, lactate, pH, HCO₃, and partial pressures of O₂ (pO₂) and CO₂ (pCO₂) were determined with an ABL90 FLEX blood gas analyzer (Radiometer, Brea, CA) from 90-µL aliquots of filtered whole blood collected in heparinized syringes as previously described (Cadaret et al., 2018).

In Vivo Metabolic Studies

Glucose-stimulated insulin secretion

Square-wave hyperglycemic clamps were performed as previously described (Yates et al., 2012a) to compare fetal insulin secretion (i.e., circulating insulin concentrations) under basal and steady-state hyperglycemic conditions. Due to the failure of some catheters and/or blood flow probes, GSIS was measured in 7 controls and 6 MI-IUGR fetuses. Three basal arterial blood samples were collected from fetuses in 10-min intervals to determine basal glucose and insulin concentrations. Hyperglycemia was then initiated by infusing the fetus with an intravenous dextrose bolus of 250 mg/kg, followed by a constant infusion of 33% dextrose solution. Blood glucose was determined every 10 min and the infusion rates were adjusted to achieve steady-state arterial glucose concentrations that were ~200% basal glucose concentration of each fetus. After steady-state hyperglycemia had been achieved for a minimum of 20 min, 3 additional blood samples were collected at 10-min intervals to determine second-phase insulin concentrations. Hyperglycemia was considered to be at steady state when glucose values from these 3 samples varied by less than 10% from their overall mean.

Hindlimb glucose metabolic flux

Hyperinsulinemic-euglycemic clamps were performed as previously described (Limesand et al., 2007; Brown et al., 2015) with some modifications (Cadaret et al., 2018) to determine fetal hindlimb-specific glucose metabolic rates. Due to the failure of additional catheters and/or blood flow probes, hindlimb glucose metabolism was measured in 5 controls and 5 MI-IUGR fetuses. To begin, [¹⁴C(U)]-d-glucose (37.2 µCi/mL; PerkinElmer Life Sciences, Boston, MA) suspended in saline was infused into the fetus at a constant rate of 1 mL/h following a 1-mL bolus. After 40 min of infusion, 4 sets of simultaneous arterial and venous blood samples were collected from the fetus at 10-min intervals to determine basal rates of glucose uptake and oxidation by the fetal hindlimb tissues. Fetal hyperinsulinemia was initiated with a 2-mL bolus of insulin (250 mU/kg; HumulinR; Lilly; Indianapolis, IN) followed by a constant infusion at 4 mU/min/kg. Fetal euglycemia was concurrently maintained with a 33% dextrose infusion that was adjusted in response to arterial plasma glucose concentrations measured every 5 to 10 min until fetal glycemic conditions were at steady state. Beginning 1 h after hyperinsulinemia/euglycemia were initiated, 4 additional sets of simultaneous arterial and venous blood samples were collected at 10-min intervals. Throughout the study, blood flow into the hindlimb through the exterior iliac artery was measured with the Transonic flow probe and recorded using LabChart software (ADInstruments, Colorado Springs, CO).

Hindlimb metabolic fluxes were estimated by the Fick Principle as the product of arterial blood flow and arteriovenous difference (Rozance et al., 2018). Specifically, hindlimb fetal glucose utilization rates were estimated by calculating the difference in glucose concentrations between simultaneously collected arterial and venous samples. These differences were then normalized to the blood flow rate at the time of sampling and the hindlimb weight at necropsy. To determine glucose oxidation rates, 3 technical replicates of whole blood from each arterial and venous sample were added to micro-centrifuge tubes, each containing 2 M HCl and suspended inside sealed 20-mL scintillation vials over a pool of 1 M NaOH. The HCl caused CO₂ to be released from the blood, where it was then captured by the NaOH at the bottom of the scintillation vial. After a 24-h incubation at room temperature, the centrifuge tube was removed and UltimaGold scintillation fluid (PerkinElmer Inc, Waltham, MA) was added to the scintillation vial. Concentrations of ¹⁴CO₂ captured from each blood sample were determined using a Beckman-Coulter 1900 TA LC counter (Beckman Coulter, Fullerton, CA). Glucose oxidation rates were determined from the difference between venous and arterial ¹⁴C specific activities. The amount of glucose oxidized in nmol was calculated by normalizing these blood ¹⁴C values to the ¹⁴C-specific activity of the infusate, determined from triplicate aliquots. The amount of glucose oxidized was then normalized to blood flow rate and hindlimb weight. The fraction of glucose taken up by the fetal hindlimb that was utilized for oxidation (i.e., fractional glucose oxidation rates) were calculated as the rate of glucose oxidation divided by the rate of glucose uptake. Blood pH, pO₂, pCO₂, and HCO₃ were also determined by ABL90 FLEX and plasma insulin was determined by ELISA.

Ex Vivo Skeletal Muscle Glucose Metabolism

At necropsy, the flexor digitorum superficialis (FDS) muscle was collected tendon-to-tendon from the left fetal hindlimb and separated into intact longitudinal strips (454 ± 19 mg) to measure ex vivo skeletal muscle glucose uptake and oxidation as previously described (Cadaret et al., 2017). Muscle strips were washed in ice-cold phosphate-buffered saline (PBS) and then preincubated for 1 h at 37 °C in Krebs–Henseleit bicarbonate buffer (KHB) containing 0.1% bovine serum albumin, 5 mM d-glucose (Millipore Sigma), and either 0 (basal) or 5 mU/mL insulin. Muscle strips were then washed in basal or insulin-spiked KHB with no glucose for 20 min at 37 °C. All KHB media were O₂ saturated (i.e., gassed with 95% O₂, 5% CO₂).

Glucose uptake

Skeletal muscle glucose uptake rates were determined from intracellular accumulation of [³H]2-deoxyglucose in incubated muscle strips as previously described (Cadaret et al., 2017, 2018). After the preincubation and wash steps, muscle strips were incubated at 37 °C for 20 min in basal or insulin-spiked KHB containing 1 mM [³H]2-deoxyglucose (300 µCi/mmol; PerkinElmer) and 39 mM [1-¹⁴C] mannitol (1.25 µCi/mmol; PerkinElmer). Muscle strips were then removed from the media, washed 3 times in ice-cold PBS, blotted dry and weighed, and then lysed in 2 M NaOH (Sigma–Aldrich) at 37 °C for 1 h. Lysates were vortexed and mixed with UltimaGold scintillation fluid, and specific activities of ³H and ¹⁴C were determined by liquid scintillation. Specific activities of the media were also determined from triplicate 10-µL aliquots. Mannitol concentrations in the lysates were used to estimate the amount of extracellular fluid in each muscle strip, and intracellular accumulation of 2-deoxyglucose was calculated as total radiolabeled 2-deoxyglucose in the lysate minus the amount in the extracellular fluid (estimated volume of extracellular fluid multiplied by its radiolabeled 2-deoxyglucose concentration, which was expected to be equivalent to the media). Four technical replicates/condition were performed for each fetus.

Glucose oxidation

Skeletal muscle glucose oxidation rates were determined by ex vivo oxidation of radiolabeled glucose as previously described (Cadaret et al., 2017) with some modifications. Each muscle strip was placed in one side of a sealed dual-well chamber and incubated for 2 h at 37 °C in basal or insulin-spiked KHB containing 5 mM [¹⁴C-U]-D-glucose (0.25 µCi/mmol). The adjacent side of the chamber contained 2 M NaOH to capture the CO₂ produced by the muscle strip. Following the incubation, chambers were incubated at −20 °C for 2 min, 2 M HCl was injected into the media through a rubber seal to release media-bound CO₂, and the chambers were incubated for an additional 1 h at 4 °C. Muscle strips were then removed, blotted dry, and weighed. The NaOH was collected from the chamber and mixed with UltimaGold scintillation fluid. Specific activity of ¹⁴CO₂ was determined from the NaOH via liquid scintillation. Glucose oxidation in pmol was calculated by normalizing dpm counts for ¹⁴CO₂ to the specific activity of radiolabeled glucose in the media, which was determined from three 10-µL aliquots. Rates were calculated by normalizing the amount of glucose oxidized to the mass of the muscle strip and time in incubation. Four technical replicates/condition were performed for each fetus.

Akt Western Immunoblots

Insulin-signaling responsiveness in FDS muscle was estimated from the proportion of phosphorylated Akt to total Akt as previously described (Cadaret et al., 2017). Muscle strips were incubated in basal or insulin-spiked KHB media for 20 min at 37 °C, snap frozen, and homogenized in RIPA buffer containing manufacturer-recommended concentrations of protease and phosphatase inhibitors (Thermo Fisher). Homogenates were sonicated for 15 s and centrifuged (14,000 × g, 5 min, 4 °C), and total protein concentrations were determined from the supernatant using a Pierce BCA Protein Assay Kit (Thermo Fisher). Protein (30 µg) was combined with BioRad 4x Laemmli Sample Buffer to make a 1× solution, heated for 5 min at 95 °C, and separated by SDS–PAGE. Gels were transferred to polyvinylidene fluoride low-fluorescent membranes (BioRad Laboratories, Hercules, CA), incubated in Odyssey block buffer (LI-COR Biosciences, Lincoln, NE) for 1 h at room temperature, and washed with 1× TBS-T (20 mM Tris-HCl + 150 mM NaCl + 0.01% Tween 20). Membranes were incubated with rabbit antiserum raised against Akt (1:1,000; Cell Signaling) or phosphorylated Akt (Ser⁴⁷³; 1:2,000; Cell Signaling) diluted in Odyssey block buffer + 0.05% Tween-20 at room temperature for 1 h (Akt) or at 4 °C overnight (phosphorylated Akt). Membranes were then incubated with goat anti-rabbit IR800 IgG secondary antiserum (LI-COR) diluted in Odyssey block buffer containing 0.05% Tween-20 and 0.01% SDS at room temperature for 1 h. Blots were scanned with the Odyssey Infrared Imaging System and analyzed with Image Studio Lite Software Ver 5.2 (LI-COR).

Myosin Heavy-Chain Electrophoresis

Proportions of myosin heavy chain (MyHC)-I and MyHC-II were determined in protein isolates from FDS muscles by electrophoresis as previously described (Yates et al., 2016). For each sample, 15 µg of protein was added to BioRad 4x Laemmli Sample Buffer (BioRad Laboratories) to make a 1× solution, which was incubated at room temperature for 10 min, heated to 70 °C for 10 min, and then loaded into a gel. The MyHC isoforms were separated by SDS–PAGE. We previously reported the respective compositions of stacking and separating gels (Yates et al., 2016). Electrophoresis was performed on a Mini-PROTEAN Tetra Cell (BioRad Laboratories) at 4 °C for 30 h at a constant 100 V. Gels were stained overnight with Gel-Code Blue (Thermo Fisher), destained in distilled water, and imaged on an Odyssey infrared imaging system (LI-COR). MyHC-I and collective MyHC-II bands were measured by densitometry (Image Studio Lite Ver 5.2; LI-COR).

Statistical Analysis

All data were analyzed using SAS 9.4 (SAS Institute, Cary, NC). Body temperatures and components from daily maternal and fetal blood samples were analyzed for the effects of experimental group, day (and time of sample), and their interactions by ANOVA using the Mixed procedure with repeated measures. Appropriate covariance structures were selected based on best-fit statistics. The main effect of fetal sex was not tested due to insufficient power, but fetal sex ratios were 6 males, 3 females for the control group and 4 males, 3 females for the MI-IUGR group. Data from in vivo metabolic studies and ex vivo metabolic studies were likewise analyzed using the Mixed procedure, with study period and incubation condition as the respective repeated measures. For each period of the in vivo metabolic studies, the 4 samples were averaged and the mean is reported. Similarly, the 4 technical reps/incubation condition in the ex vivo metabolic studies were averaged, and the mean is reported. Fetal biometric data and fiber type ratios were analyzed by 1-way ANOVA. Ewe was considered the experimental unit for all maternal data, and fetus was the experimental unit for all other analyses. The threshold for significance was α = 0.05, and all data are presented as the mean ± SE.

Results

Maternal Inflammatory Responses

Experimental group × day × sample time interactions were observed (P < 0.05) for maternal body temperatures and for circulating concentrations of white blood cells, lymphocytes, monocytes, and granulocytes over the period in which maternal endotoxin was serially administered (dGA 100 to 115). Body temperatures were increased (P < 0.05) in MI-IUGR ewes (i.e., those administered LPS) compared with control ewes (i.e., those administered saline) within 3 h of each injection and were not different by 12 h for all but the first injection (Supplementary Fig. 1A). Maternal blood concentrations of total white blood cells (Supplementary Fig. 1B), lymphocytes (Supplementary Fig. 1C), monocytes (Supplementary Fig. 1D), and granulocytes (Supplementary Fig. 1E) were reduced (P < 0.05) in MI-IUGR ewes compared with control ewes immediately following each injection but were increased (P < 0.05) otherwise. Maternal platelet concentrations were less (P < 0.05) in MI-IUGR ewes compared with control ewes, but were not affected by day or sample time (Supplementary Fig. 1F). Hematocrit (32.9% ± 2.2%) and red blood cell concentrations (9.8 ± 0.7 cells/µL) did not differ between experimental groups or among days or sample times. Maternal plasma TNFα concentrations were elevated (P < 0.05) in MI-IUGR ewes compared with control ewes and did not differ among dGA (Supplementary Fig. 2).

Daily Maternal and Fetal Physiological Parameters

No experimental group × day interactions were observed for any maternal or fetal blood parameters measured daily from dGA 120 to 125. Maternal plasma TNFα tended to be elevated (P = 0.07) in MI-IUGR ewes (Fig. 1A) and fetal plasma TNFα was elevated (P < 0.05) in MI-IUGR fetuses compared with controls (Fig. 1B). During this period, total white blood cell concentrations were increased (P < 0.05) in MI-IUGR ewes (Fig. 2A) and fetuses (Fig. 2B) compared with controls. MI-IUGR increased (P < 0.05) maternal lymphocytes (Fig. 2C), but not fetal lymphocytes (Fig. 2D), and increased (P < 0.05) maternal and fetal monocytes (Fig. 2E and F, respectively) and granulocytes (Fig. 2G and H, respectively) compared with controls. No differences were observed between experimental groups or among days for maternal or fetal concentrations of platelets, hematocrit, or red blood cells. Maternal blood glucose did not differ between experimental groups but was greater (P < 0.05) at dGA 125 than 120 in all ewes (4.1 ± 0.5 vs. 3.0 ± 0.4 mM). Maternal blood lactate (0.7 ± 0.2 mM), pH (7.431 ± 0.021), pCO₂ (40.7 ± 2.1 mmHg), HCO₃ (27.7 ± 2.4 mM), and pO₂ (81.3 ± 8.8 mmHg) did not differ between experimental groups or among days. Fetal blood glucose (1.4 ± 0.3 mM), lactate (1.8 ± 0.4 mM), pH (7.351 ± 0.020), pCO₂ (52.7 ± 3.1 mmHg), and HCO₃ (31.0 ± 5.6 mM) did not differ between experimental groups or among days. However, fetal blood pO₂ tended to be reduced (P = 0.10) in MI-IUGR compared with control fetuses (26.9 ± 2.2 vs. 32.8 ± 2.6 mmHg) and maternofetal gradient (i.e., fetal pO₂/maternal pO₂) was reduced (P < 0.05) in MI-IUGR compared with control fetuses (0.33 ± 0.02 vs. 0.41 ± 0.03 mmHg/mmHg). The difference between maternal oxygen and fetal oxygen (i.e., maternal pO₂ − fetal pO₂) was also greater (P < 0.05) in MI-IUGR fetuses compared with controls (28.3 ± 1.0 vs. 23.5 ± 1.6 mmHg).

Figure 1.

Plasma TNFα concentrations in ewes and fetuses during and after serial administration of lipopolysaccharide (LPS) to dams during the early third trimester of pregnancy. Ewes were injected i.v. with E. coli O55:B5 LPS (MI-IUGR) or saline (controls) every 72 h from dGA 100 to 112. Data are shown for dams during the period of LPS administration (dGA 100 to 112; 3 h after each injection; controls, n = 9, MI-IUGR ewes, n = 7) (A), for dams after LPS administration (dGA 120 and 125; controls, n = 9, MI-IUGR ewes, n = 7) (B), and for fetuses after LPS administration (dGA 120 and 125; controls, n = 7, MI-IUGR fetuses, n = 6) (C). Effects of maternal injection (GRP), day, and the interaction are noted when significant (P < 0.05).

Figure 2.

Maternal and fetal circulating leukocyte concentrations after serial administration of lipopolysaccharide (LPS) to dams during the early third trimester of pregnancy. Ewes were injected i.v. with E. coli O55:B5 LPS (MI-IUGR) or saline (controls) every 72 h from dGA 100 to 112. Data are shown for maternal (controls, n = 9; MI-IUGR, n = 7) and fetal (controls, n = 7; MI-IUGR, n = 6) whole-blood concentrations of total white blood cells (A and B, respectively), lymphocytes (C and D, respectively), monocytes (E and F, respectively), and granulocytes (G and H, respectively). Effects of maternal injection (GRP), day of gestation (DAY), and the interaction (GRP × DAY) are noted when significant (P < 0.05).

Fetal Biometry at Necropsy

At necropsy, fetal weights were reduced (P < 0.05) in MI-IUGR (2.53 ± 0.28 kg) compared with control (3.23 ± 0.18 kg) fetuses (Supplementary Table 1). Absolute weights for fetal hearts, lungs, livers, kidneys, and brains did not differ between experimental groups. Moreover, heart, lung, liver, and kidney weights relative to fetal weights did not differ between groups. However, brain weight/fetal weight tended (P = 0.10) to be greater in MI-IUGR fetuses compared with controls.

Fetal GSIS

No experimental group × glycemic period interactions were observed for fetal blood glucose or plasma insulin concentrations during the square-wave hyperglycemic clamp. Fetal blood glucose concentrations did not differ between experimental groups but by design were greater (P < 0.05) in the hyperglycemic period than in the basal period (Fig. 3A). Fetal plasma insulin concentrations were reduced (P < 0.05) in MI-IUGR fetuses compared with controls across both periods and were greater (P < 0.05) during the hyperglycemic period compared with the basal period in all fetuses (Fig. 3B).

Figure 3.

Fetal glucose-stimulated insulin secretion after serial administration of lipopolysaccharide (LPS) to dams during the early third trimester of pregnancy. Ewes were injected i.v. with E. coli O55:B5 LPS (MI-IUGR) or saline (controls) every 72 h from dGA 100 to 112 and square-wave hyperglycemic clamps were performed on dGA 123. Data are shown for fetal (controls, n = 7, MI-IUGR, n = 6) whole-blood glucose concentrations (A) and plasma insulin concentrations (B) during resting (basal) glycemic conditions and steady-state fetal hyperglycemia. Effects of maternal injection (GRP), glycemic period (PERIOD), and the interaction are noted when significant (P < 0.05).

Fetal Hindlimb Glucose Metabolism

An experimental group × insulinemic period interaction was observed (P < 0.05) for fetal hindlimb glucose oxidation and fractional glucose oxidation during the hyperinsulinemic-euglycemic clamp, but interactions were not observed for hindlimb glucose uptake or arterial blood lactate, pH, pO₂, pCO₂, HCO₃, or plasma insulin concentrations. Fetal hindlimb glucose uptake rates did not differ between experimental groups but were greater (P < 0.05) during the hyperinsulinemic period than the basal period (Fig. 4A). Hindlimb glucose oxidation rates were not different between experimental groups during the basal period, but were less (P < 0.05) in MI-IUGR fetuses than in controls during the hyperinsulinemic period (Fig. 4B); hyperinsulinemia increased (P < 0.05) hindlimb glucose oxidation rates by 4.1-fold in controls and by 2.5-fold in MI-IUGR fetuses compared with their respective basal periods. The estimated fraction of glucose utilized by the hindlimb for oxidation did not differ between experimental groups during the basal period (0.25 ± 0.09 mmol/mmol) but was less (P < 0.05) in MI-IUGR fetuses than in controls during the hyperinsulinemic period (0.43 ± 0.04 vs. 0.81 ± 0.13 mmol/mmol). Fetal arterial lactate concentrations did not differ between experimental groups or insulinemic periods (Fig. 4C). Fetal arterial blood pH, pO₂, pCO₂, and HCO₃ likewise did not differ. By design, fetal plasma insulin concentrations were greater (P < 0.05) during the hyperinsulinemic period than the basal period (15.0 ± 1.5 vs. 0.51 ± .08 ng/ml) but did not differ between experimental groups.

Figure 4.

Fetal hindlimb glucose metabolism after serial administration of lipopolysaccharide (LPS) to dams during the early third trimester of pregnancy. Ewes were injected i.v. with E. coli O55:B5 LPS (MI-IUGR) or saline (controls) every 72 h from dGA 100 to 112 and hyperinsulinemic-euglycemic clamps were performed on dGA 124. Data are shown for fetal (controls, n = 5; MI-IUGR, n = 5) hindlimb glucose uptake rates (A), hindlimb glucose oxidation rates (B), and arterial blood lactate concentrations (C) during resting (basal) insulinemic conditions and steady-state fetal hyperinsulinemia. Effects of maternal injection (GRP), glycemic period (PERIOD), and the interaction (GRP × PERIOD) are noted when significant (P < 0.05). ^a,b,cWhen interactions were observed, means with different superscripts differ (P < 0.05).

Ex Vivo Skeletal Muscle Glucose Metabolism and Fiber-Type Ratios

An experimental group × incubation condition interaction was observed (P < 0.05) for ex vivo skeletal muscle glucose oxidation rates, but not for glucose uptake rates. Skeletal muscle glucose uptake rates were less (P < 0.05) in muscle from MI-IUGR fetuses compared with muscle from controls across incubation conditions and were greater (P < 0.05) in insulin-spiked media compared with basal media regardless of experimental group (Fig. 5A). Skeletal muscle glucose oxidation rates did not differ between experimental groups in basal media but were less (P < 0.05) in muscle from MI-IUGR fetuses compared with muscle from controls when incubated in insulin-spiked media (Fig. 5B). Insulin-spiked media increased (P < 0.05) glucose oxidation rates in muscle from controls by 1.7-fold compared with basal media, but did not affect glucose oxidation rates in muscle from MI-IUGR fetuses. Phosphorylated Akt/total Akt was reduced (P < 0.05) in muscle from MI-IUGR fetuses compared with muscle from control fetuses across incubation conditions and was greater (P < 0.05) in insulin-spiked media than in basal media for all fetuses (Fig. 5C). Fiber-type ratios in the fetal FDS muscle were estimated at necropsy by MyHC content (Supplementary Fig. 3A), but did not differ between control (78.5% Type II, 21.5% Type I) and MI-IUGR (74.9% Type II, 25.1% Type I) groups (Supplementary Fig. 3B).

Figure 5.

Ex vivo glucose metabolism by primary fetal skeletal muscle after serial administration of lipopolysaccharide (LPS) to dams during the early third trimester of pregnancy. Ewes were injected i.v. with E. coli O55:B5 LPS (MI-IUGR) or saline (controls) every 72 h from dGA 100 to 112 and fetal flexor digitorum superficialis muscles were isolated on dGA 125. Data are shown for fetal (controls, n = 9, MI-IUGR, n = 7) skeletal muscle glucose uptake rates (A), glucose oxidation rates (B), and Akt phosphorylation (C) after incubation in media without (basal) or with 5 mU/mL insulin. Effects of maternal injection (GRP), media (MEDIA), and the interaction (GRP × MEDIA) are noted when significant (P < 0.05). ^a,b,cWhen interactions were observed, means with different superscripts differ (P < 0.05).

Discussion

In this study, we found that sustained maternal inflammation during the early third trimester of pregnancy resulted in impaired fetal growth and glucose metabolism. Specifically, MI-IUGR fetuses exhibited a reduction in glucose oxidation that was inherent to skeletal muscle and occurred independently of glucose uptake rates, which were also reduced. The deficit in glucose oxidation was characterized by poor responsiveness to insulin, as basal glucose oxidation rates in MI-IUGR hindlimb tissues and primary skeletal muscle were normal but insulin-stimulated glucose oxidation rates were diminished by about 50%. Conversely, glucose uptake rates in MI-IUGR fetal skeletal muscle were reduced equally under basal and insulin-stimulated conditions. MI-IUGR fetuses also exhibited impaired β-cell function, as circulating insulin concentrations were less in MI-IUGR fetuses than controls under resting glycemic conditions. This disparity was even greater under hyperglycemic conditions, which would indicate impaired responsiveness of MI-IUGR fetal β cells to glucose stimulation. Circulating leukocyte and TNFα concentrations were greater in MI-IUGR fetuses when measured 8 to 13 d after the induction of maternal inflammation, which indicated persistent systemic fetal inflammation. The sustained nature of this inflammation was consistent with the findings of previous studies in which fetal sheep were directly administered LPS (Newnham et al., 2002; Kramer et al., 2010). However, E. coli-derived LPS endotoxins and maternal cytokines both have low placental permeability (Aaltonen et al., 2005; Gomez-Lopez et al., 2018), and thus we speculate that the fetal inflammation we observed was mediated by placental inflammation. Lower fetal blood O₂ and greater maternofetal O₂ gradients in MI-IUGR fetuses would be consistent with placental insufficiency. However, MI-IUGR fetuses did not exhibit the hypoglycemia that is often associated with placental insufficiency (Rozance et al., 2018; Yates et al., 2018), and thus more comprehensive assessments of placental function in this model are warranted. Nevertheless, we conclude from these findings that maternal inflammation induced by serial LPS injections in the early third-trimester restricted fetal growth and yielded a fetal metabolic phenotype similar to that observed in maternal hyperthermia-induced IUGR fetal sheep (Limesand et al., 2007; Brown et al., 2015). Furthermore, systemic fetal inflammation subsequent to maternal inflammation represents a potential therapeutic target for improving metabolic outcomes of IUGR.

Pregnant ewes consistently exhibited sustained inflammatory responses to serial administration of LPS endotoxin. This included a transient febrile response lasting several hours after each injection, elevated plasma TNFα, and an acute drop in circulating leukocytes followed by an increase that persisted until the next injection, after which the dip/rise repeated. This pattern is typical of white blood cell responses to LPS (Yates et al., 2011; Dickson et al., 2019) and reflects acute migration of cells out of circulation followed by increased mobilization of additional cells into the bloodstream. It is worth noting that the responses were of a similar general magnitude for each of the 5 LPS injections. In an earlier study, we found that pregnant rats exhibited a robust inflammatory response to the first in a series of daily LPS injections at midgestation, but that their responses to the subsequent injections were diminished substantially (Cadaret et al., 2019). We attribute the sustained responsiveness to LPS observed in the present study to the greater amount of time between injections (72 h) compared with the previous study (24 h), although differences between species and the gestational age at which LPS was administered cannot be ignored. Nevertheless, the heightened inflammatory state did not resolve quickly, and maternal leukocyte and TNFα concentrations were still elevated (albeit to a lesser extent) almost 2 wk after the final LPS injection. One limitation of this study is that it did not assess fetal responses during the period in which maternal LPS was administered. However, when fetal responses were assessed over a 6-d period beginning 8 d after the final LPS injection (i.e., dGA 120 to 125), fetal plasma TNFα, and circulating concentrations of total white blood cells were both almost 50% higher in MI-IUGR fetuses than in controls. This persistent fetal inflammation, which included a ~35% increase in circulating monocytes and a ~200% increase in circulating granulocytes, was reasonably consistent with the blood leukocyte profiles of the MI-IUGR dams over the same time period. It was also comparable to leukocyte profiles observed in fetal sheep 1 wk after direct infusion with LPS, which included elevated monocytes, neutrophils, lymphocytes, and total white blood cells (Newnham et al., 2002; Kramer et al., 2010; Maneenil et al., 2015). In addition, monocytes isolated from these fetuses 7 d after LPS administration produced substantially more TNFα in response to ex vivo stimulation (Kramer et al., 2010). Although our study did not identify how maternal LPS resulted in chronic fetal inflammation, it is unlikely that the observed changes were a fetal response to LPS proper. Studies in rodents showed that placental permeability to LPS from E. coli is low (Kohmura et al., 2000; Gomez-Lopez et al., 2018), which recapitulated earlier findings in human placental cells (Romero et al., 1987). Although differences in placental structure exist among rodents and ruminants, we are not aware of any evidence that would indicate LPS permeability is greater in the sheep placenta. It is also unlikely that the fetuses in our study encountered maternal cytokines, which also have low placental permeability (Zaretsky et al., 2004; Aaltonen et al., 2005). Rather, we speculate that fetal inflammation was primarily driven by the placenta itself, which produces and secretes large amounts of inflammatory factors from both the maternal-facing and fetal-facing surfaces in response to maternal LPS (Boles et al., 2012; Bloise et al., 2013; Lei et al., 2015). Moreover, the ~20% reduction in O₂ partial pressures in our MI-IUGR fetuses along with the ~25% greater maternofetal O₂ gradients would be consistent with placental insufficiency resulting from a placental insult (Rozance et al., 2018). Indeed, structural and functional placental insufficiency was observed in pregnant rats administered LPS in the early third trimester of pregnancy (Cotechini et al., 2014a,b), and it is worth noting that hypoxia itself can be a stimulator of inflammation (Taylor and Colgan, 2017). Interestingly, our MI-IUGR fetuses did not present hypoglycemia, which is perhaps an indication of milder placental stunting compared with what would be expected in other sheep models of IUGR (Beede et al., 2019). Nevertheless, MI-IUGR fetal sheep were ~22% smaller than controls by dGA 125, which was comparable to several other ovine models of IUGR (Beede et al., 2019) but was more profound than the fetal growth restriction observed by us and others in MI-IUGR rodents (Liu et al., 2014; Cadaret et al., 2019).

Our in vivo assessment of MI-IUGR fetal hindlimb glucose metabolism revealed reduced insulin-stimulated glucose oxidation rates but normal glucose uptake, which was consistent with previously observed changes in whole-body glucose metabolism for maternal hyperthermia-induced IUGR fetuses (Limesand et al., 2007; Brown et al., 2015). However, our ex vivo assessments showed that skeletal muscle-specific glucose uptake and oxidation were both impaired in MI-IUGR fetuses. Similar to the hindlimb, the deficits in skeletal muscle-specific glucose oxidation were characterized by reduced responsiveness to insulin, as rates were comparable to controls in basal media but ~22% lower in insulin-spiked media. However, glucose uptake rates and Akt phosphorylation were lower for MI-IUGR fetal muscle in both basal and insulin-spiked media, indicating additional deficits beyond impaired responsiveness to insulin and perhaps related to mTORC signaling (Posont et al., 2017). Skeletal muscle is a principle utilizer of glucose (DeFronzo et al., 1981; Brown, 2014), and disproportionate targeting of skeletal muscle glucose metabolism is an adaptive mechanism for increasing fetal metabolic thrift (Yates et al., 2012b; Posont and Yates, 2019; Reynolds et al., 2019). Less utilization by muscle spares glucose for vital brain and heart tissues, and reduced oxidation shifts utilization toward glycolytic production of lactate that can be utilized for gluconeogenesis (Jones et al., 2019). Blood lactate concentrations were not increased in our MI-IUGR fetuses, but previous observations in maternal hyperthermia-induced IUGR fetal sheep would indicate that this was due to greater clearance by the fetal liver (Limesand et al., 2007; Thorn et al., 2010, 2013). The changes in skeletal muscle glucose metabolism that we observed in our MI-IUGR fetuses were consistent with what we would expect under inflammatory conditions. We previously found that incubation of skeletal muscle with inflammatory cytokines diminished insulin-stimulated glucose oxidation and Akt phosphorylation (Cadaret et al., 2017) and that LPS-induced systemic inflammation in adult sheep caused them to be insulin resistant (Yates et al., 2011). Others have shown that 3- to 5-d incubation of muscle cell lines with inflammatory cytokines reduced O₂ consumption rates, decreased expression of oxidative enzymes, and increased glycolytic lactate production (Zentella et al., 1993; Remels et al., 2010, 2015). Reduced muscle glucose oxidation in our MI-IUGR fetuses was not associated with changes in muscle fiber types, as they exhibited normal proportional concentrations of MyHC-I, which is indicative of slow oxidative muscle fibers. We previously found that the proportions of MyHC-I protein and of muscle fibers expressing MyHC-I were reduced in the hindlimb muscle of maternal hyperthermia-induced IUGR fetuses (Yates et al., 2016). However, we did not observe such reductions in the hindlimb muscles of MI-IUGR fetal rats (Cadaret et al., 2019).

Our MI-IUGR fetal sheep exhibited a 33% reduction in resting plasma insulin concentrations but a 58% reduction in second-phase GSIS. This was comparable to findings in maternal hyperthermia-induced IUGR fetal sheep (Leos et al., 2010; Macko et al., 2013) and indicates a disruption in stimulus-secretion coupling (i.e., the ability of β cells to respond to elevated blood glucose). Studies have demonstrated a role for hypercatecholaminemia in impaired IUGR fetal β-cell function by rescuing it with adrenergic blockers (Leos et al., 2010; Macko et al., 2013) or adrenal demedullation (Yates et al., 2012a; Macko et al., 2016) and by replicating it with norepinephrine infusion (Chen et al., 2014; Chen et al., 2017). In fact, we found that acute hypoxemia had no impact on GSIS in fetuses that had undergone adrenal demedullation to prevent the hypercatecholaminemic response (Yates et al., 2012a). However, whole-transcriptome analyses of pancreatic islets from maternal hyperthermia-induced IUGR fetuses identified TNFα signaling as one of the top enhanced pathways (Kelly et al., 2017), which was not the case in islets from norepinephrine-infused fetuses (Kelly et al., 2018). Thus, we speculate that inflammation plays a parallel role in suppressing β-cell function in IUGR fetuses and that impaired insulin secretion can independently result from abhorrent adrenergic or inflammatory activity. Inflammatory regulation of insulin secretion is complex and is dependent on severity and duration of exposure. In adult mice, chronic administration of LPS at low doses increased insulin concentrations under resting glycemic conditions but not during hyperglycemia (Cani et al., 2007). When the LPS dose was doubled, resting insulin concentrations were normal but GSIS was increased (Nohr et al., 2016). Conversely, 24-h incubation of mouse islets with LPS reduced GSIS in a dose-dependent fashion (Cucak et al., 2014), and palmitate-induced inflammation of Min6 β cells and primary human islets reduced GSIS via NFκB-mediated pathways (Mukhuty et al., 2017; Lee et al., 2019). Moreover, acute exposure to TNFα had no effect on insulin secretion in INS1 β cells (Zhang and Kim, 1995), but chronic exposure to TNFα or other inflammatory cytokines resulted in dose-dependent reductions of basal secretion and GSIS by inhibiting glucose oxidation (Zhang and Kim, 1995; Barlow et al., 2018).

The findings of this study allow us to conclude that chronic maternal inflammation results in IUGR fetuses with reduced capacity for skeletal muscle glucose metabolism and impaired β-cell function. This phenotype is similar to that of maternal hyperthermia-induced IUGR fetuses and includes evidence for mild to moderate placental insufficiency. The metabolic dysfunction observed in MI-IUGR fetuses was reflective of their inflammatory state, which was heightened even 2 wk after ceasing maternal LPS administration. Although more work is needed to identify the factors that mediate fetal inflammation, this study shows that developmental programming of fetal skeletal muscle by chronic inflammation results in metabolic dysfunction that is consistent with what is observed in IUGR-born humans and animals, and that fetal inflammation may be a target for therapeutic interventions to improve outcomes of IUGR.