Perinatal Oxidative Stress May Affect Fetal Ghrelin Levels in Humans

In vitro cell model studies have shown that oxidative stress may affect beta-cell function. It is unknown whether oxidative stress may affect metabolic health in human fetuses/newborns. In a singleton pregnancy cohort (n = 248), we studied maternal (24–28 weeks gestation) and cord plasma biomarkers of oxidative stress [malondialdehyde (MDA), F2-isoprostanes] in relation to fetal metabolic health biomarkers including cord plasma glucose-to-insulin ratio (an indicator of insulin sensitivity), proinsulin-to-insulin ratio (an indicator of beta-cell function), insulin, IGF-I, IGF-II, leptin, adiponectin and ghrelin concentrations. Strong positive correlations were observed between maternal and cord plasma biomarkers of oxidative stress (r = 0.33 for MDA, r = 0.74 for total F2-isoprostanes, all p < 0.0001). Adjusting for gestational age at blood sampling, cord plasma ghrelin concentrations were consistently negatively correlated to oxidative stress biomarkers in maternal (r = −0.32, p < 0.0001 for MDA; r = −0.31, p < 0.0001 for F2-isoprostanes) or cord plasma (r = −0.13, p = 0.04 for MDA; r = −0.32, p < 0.0001 for F2-isoprostanes). Other fetal metabolic health biomarkers were not correlated to oxidative stress. Adjusting for maternal and pregnancy characteristics, similar associations were observed. Our study provides the first preliminary evidence suggesting that oxidative stress may affect fetal ghrelin levels in humans. The implications in developmental “programming” the vulnerability to metabolic syndrome related disorders remain to be elucidated.

Plasma unacylated ghrelin was measured by a human unacylated ghrelin immunoassay kit (SPI-BIO, Bertin). The intra-and inter-assay coefficients of variation (CVs) were in the range of 2.8% to 9.2%. We chose to measure unacylated ghrelin since it accounts for about 99% of total ghrelin, while acylated ghrelin accounts for only about 1% of total ghrelin in cord blood 15 . Also, for accurate measurement of acylated ghrelin, specimens need special treatment to prevent the degradation of acylated ghrelin, but this was not done during specimen collection and processing in the study. Without timely anti-deacylation treatment, acylated ghrelin would be rapidly converted to unacylated ghrelin in blood samples 16 . Therefore, the measurement of plasma unacylated ghrelin is effectively the measurement of total ghrelin in the present study. For simplicity, we used the term ghrelin in lieu of unacylated ghrelin in results presentation.
Data on maternal and cord plasma concentrations of leptin and adiponectin, insulin-like growth factor I (IGF-I), IGF-II and surrogate indicators of fetal insulin sensitivity (cord plasma glucose-to-insulin ratio) and beta cell function (proinsulin-to-insulin ratio) are available in the study cohort 12,17,18 . The intra-and inter-assay CVs in the assays of plasma glucose, insulin, proinsulin, IGF-I, IGF-II, leptin and adiponectin were in the range of 2.0% to 10.4%.
All assays were blinded to the labs. The lab staff had no information on clinical characteristics of study subjects.
Oxidative stress biomarkers. Maternal and fetal oxidative stress levels were represented by plasma total F2-isoprostanes and MDA concentrations. F2-isoprotanes, the products of lipid peroxidation, are reliable biomarkers of in vivo oxidative stress 19 . MDA, another product of lipid peroxidation, was used for evaluating the consistency of associations between oxidative stress and fetal metabolic health biomarkers. Many previous studies used enzyme-linked immuno-assays to measure only one specific isomer of F2-isoprostaness -the 8-iso-PGF2α which accounts for only a small fraction of total F2-isoprostanes 14 , as an oxidative stress biomarker. The use of our recently developed technique for measuring F2-isoprostanes by HPLC-MS/MS allows the simultaneous measurements of seven isomers. In addition to total F2-isoprostanes, we explored the associations of specific F2-isoprostane isomers with fetal metabolic health biomarkers.
Fetal metabolic health biomarkers. Fetal metabolic health biomarkers included cord plasma glucose-to-insulin ratio (a surrogate indicator of fetal insulin sensitivity) 20 proinsulin-to-insulin ratio (a surrogate indicator of beta-cell function 21 ), insulin, IGF-I and IGF-II (key fetal growth hormones) 17 , leptin, adiponectin and ghrelin concentrations. Leptin, adiponectin and ghrelin are important hormones in regulating insulin sensitivity and energy balance 22,23 . We hypothesized that perinatal oxidative stress may "program" long-term metabolic risk through affecting one or more of these fetal growth and metabolic health biomarkers. Birth weight (z score, according to the Canadian sex-and gestational age-specific birth weight standards 24 ) was also examined because oxidative stress has been associated with impaired fetal growth in some studies 25,26 . Statistical analysis. Median (inter-quartile range) and Mean± SD (standard deviation) are presented for plasma ghrelin, MDA and F2-isoprostanes concentrations. Biomarker data (positively skewed crude data distribution) were log-transformed in partial correlation analyses adjusting for gestational age at blood sampling (maternal or cord blood) and glucose concentration (an important indicator of metabolic status at the time of blood sampling), and in t test or analysis of variance for differences between groups where appropriate. Generalized linear models were used to assess the associations of oxidative stress with cord plasma metabolic health biomarkers in log-transformed data adjusting for maternal and pregnancy characteristics. The effects were expressed in percentage change [for a dependent variable (y) in log scale, the regression coefficient (β) effectively represents the proportion of change in the original scale [log y 1 -log y 0 = β, then log (y 1 /y 0 ) = β, thus y 1 /y 0 = e β , and the percentage change is (e β -1)*100%]. The co-variables considered for inclusion in the adjusted models were maternal race (White, others), age (< 35, ≥ 35 years), parity (primiparous: yes/no), pre-pregnancy BMI (kg/m 2 ), smoking (yes/no), alcohol use (yes/no), gestational diabetes (yes/no), gestational hypertension (yes/no), infant sex (boy, girl), gestational age (week), ponderal index (birth weight/length 3 , kg/m 3 ), mode of delivery (cesarean, vaginal) and cord plasma glucose concentration (SD score). Co-variables with p values < 0.20 were retained in the parsimonious final adjusted models. Birth weight was not included in the adjusted models since it was highly correlated with ponderal index, a surrogate indicator of fetal adiposity associated with insulin sensitivity 12,27 . Data management and analyses were conducted using Statistical Analysis System (SAS), Version 9.2 (SAS Institute, Cary, North Carolina). P values < 0.003 were considered statistically significant, accounting for 16 primary comparisons of interest (8 cord plasma biomarkers as the primary outcomes, 2 cord plasma oxidative stress biomarkers as the primary exposures, total number of primary comparisons = 8*2 = 16, adjusted p value cutoff = 0.05/16 = 0.003). The study had a power of 99% to detect an absolute correlation efficient of 0.3 or greater, taking into account of multiple tests.
Scatter plots revealed negative relationships between maternal or cord plasma F2-isoprostanes and cord plasma ghrelin concentrations (tests for linear trends in log transformed data, all p < 0.0001) (Fig. 1), and between maternal plasma MDA and cord plasma ghrelin concentrations (p < 0.0001). There was also a negative relationship between cord plasma MDA and ghrelin concentrations, but the association was not statistically significant (p = 0.066).
Adjusting for gestational age at maternal and cord blood sampling and glucose concentration, significant positive correlations were observed between maternal and cord plasma concentrations in ghrelin (partial r = 0.33, p < 0.0001) and biomarkers of oxidative stress (partial r = 0.33 for MDA, r = 0.74 for F2-isoprostanes, all p < 0.0001) ( Table 2). Consistent negative correlations were observed between cord plasma ghrelin levels and indices of oxidative stress in both maternal (partial r = − 0.32, p < 0.0001 for MDA; r = − 0.31, p < 0.0001 for total F2-isoprostanes) and cord plasma (partial r = − 0.13, p = 0.04 for MDA; r = − 0.32, p < 0.0001 for total F2-isoprostanes). There were no significant correlations in maternal or cord plasma concentrations of MDA or total F2-isoprostanes with cord plasma concentrations of insulin, leptin, adiponectin, IGF-I or IGF-II. Similarly, there were no significant correlations in oxidative stress biomarkers with cord plasma glucose-to-insulin and proinsulin-to-insulin ratios (Table 2), or with cord plasma proinsulin concentrations (data not shown).
Exploratory analyses showed that cord plasma ghrelin concentrations were negatively correlated with four specific F2-isoprostane isomers (partial r = − 0.34      Table 2. Partial correlation coefficients (n = 248 mother-newborn pairs) between oxidative stress and fetal (cord plasma) metabolic health biomarkers. ξ ξ Partial correlations adjusting for gestational age at blood sampling (maternal and cord blood) and glucose concentration (an important indicator of metabolic status at the time of blood sampling); data were log-transformed for biomarkers with skewed crude data distribution before the correlation analyses. Correlation coefficients in bold remained significant after accounting for multiple tests at p < 0.001. *p < 0.05, †p < 0.01, ‡p < 0.001, τ p < 0.0001. MDA = malondialdehyde; F 2 -Isop = total F 2 -isoprostanes.
correlations with the other three F2-isoprostane isomers (r = − 0.09 for iPF2α -VI, p = 0.15; r = − 0.03 for 5-iPF2α -VI, p = 0.63; r = 0.05 for 5-8,12-iso-iPF2α -VI, p = 0.49). There were no significant correlations between all specific F2-isoprostane isomers and cord plasma leptin or adiponectin concentrations (all p > 0.1), except for that cord plasma concentrations of 5-iPF2α -VI and adiponectin were weakly correlated (r = 0.16, p = 0.02). Adjusting for maternal plasma ghrelin concentration, pre-pregnancy BMI, infant sex, gestational age, ponderal index and mode of delivery (other co-variables were excluded at p > 0.2 in multivariate generalized regression models), significant negative associations remained in maternal or cord plasma F2-isoprostanes and maternal plasma MDA concentrations with cord plasma ghrelin concentrations (Table 3). For each log unit increase in plasma concentration, cord plasma ghrelin decreased by 14.2% for maternal plasma MDA (adjusted p < 0.001), by 10.7% for maternal plasma total F2-isoprostanes (adjusted p = 0.039), and by 14.0% for cord plasma total F2-isoprostanes (adjusted p = 0.002), respectively. The association between cord plasma MDA and ghrelin became not statistically significant (p = 0.206) after the multivariate adjustment, but the point estimate also showed a negative association. There was a strong influence of maternal ghrelin on fetal circulating ghrelin concentration: each log (pg/ml) unit increase in maternal ghrelin was associated with a 68% increase in cord plasma ghrelin (p < 0.001).
There were no significant correlations between birth weight (z score) and oxidative stress biomarkers (MDA or F2-isoprostanes) (all p > 0.05, data not shown). Cord plasma concentrations of MDA, total F2-isoprostanes or specific F2-isoprostane isomers were similar comparing SGA vs. AGA newborns, except for that SGA infants had lower 5-iPF2α -VI concentrations (median: 228.3 vs. 262.6 pg/ml, p = 0.01). There were no significant differences in cord plasma biomarkers of oxidative stress in LGA vs. AGA infants.
There were no significant differences in maternal and cord plasma concentrations of ghrelin or biomarkers of oxidative stress comparing gestational diabetic (n = 25) versus non-diabetic pregnancies, or gestational hypertensive (n = 14) versus non-hypertensive pregnancies, or preterm (n = 11) vs. term births (all p > 0.05, data not shown). Excluding these patients from the analyses, the results were similar. There were no significant differences in cord plasma biomarkers of oxidative stress in caesarean-section (n = 70) vs. vaginal deliveries, or between boys and girls (all p > 0.05, data not shown).

Discussion
Main findings. To our knowledge, this is the first study demonstrating a consistent negative correlation in perinatal (maternal or fetal) oxidative stress with circulating ghrelin levels in human fetuses/newborns. Also, this is the largest pregnancy cohort demonstrating that maternal and fetal oxidative stress levels are strongly positively correlated.

Data interpretation in comparisons with previous studies.
The developmental programming hypothesis is well recognized 2 , but data on perinatal biomarkers that may be related to the mechanisms of metabolic programming remain scanty in humans 28 . The present study provides important evidence suggesting that perinatal oxidative stress does not affect most fetal metabolic health indicators as measured at birth in a relatively low-risk pregnancy cohort (without Per log unit increase in concentration of: Adjusted % change (95% CI) in cord plasma plasma ghrelin concentration ξ P  Table 3. Adjusted percentage change (95% CI)ξ in cord plasma ghrelin concentration in relation to maternal plasma ghrelin, maternal and cord plasma MDA and F2-isoprostanes concentrations. ξ Covariables considered for inclusion in the adjusted models were maternal race, age, parity, pre-pregnancy body mass index, smoking, alcohol use, gestational diabetes, gestational hypertension, infant sex, gestational age, ponderal index, mode of delivery and cord plasma glucose concentration; only co-variables with p values < 0.20 were retained in the final adjusted models (pre-pregnancy BMI, infant sex, gestational age, ponderal index and mode of delivery). The models for the effects of MDA and F 2 -isoprostanes on cord ghrelin concentration were further adjusted for maternal ghrelin concentration. Separate models were fitted for maternal and cord blood oxidative stress biomarkers (MDA, total F2-isoprostanes) as they were strongly correlated. Model's R squares are in the range of 0.34 to 0.39. The adjusted % change was calculated from the regression coefficient of the dependent variable (y) in log scale per log unit increase in the independent variable (x), because the regression coefficient (β) represents the proportion of change in y in the original scale (ghrelin concentration): log y 1 -log y 0 = β, then log (y 1 /y 0 ) = β, thus y 1 /y 0 = e β , and the percentage change is (e β -1)*100%. MDA = malondialdehyde; CI = confidence interval.
pre-gestational serious chronic illnesses such as diabetes or hypertension). However, oxidative stress was associated with decreased fetal ghrelin levels. In addition, there is a strong influence of maternal ghrelin levels on fetal ghrelin levels.
Ghrelin is an important hormone in regulating food intake and energy balance, and is primarily secreted by the fundus of the stomach in adult humans and rodents, but is primarily secreted by the pancreas during the perinatal period 11 . Ghrelin circulates in two forms: unacylated and acylated. Acylated ghrelin is considered the biologically active form with respect to its growth hormone secretion stimulation and orexigenic effects through binding to growth hormone secretagogue receptor 29 . However, unacylated ghrelin is not biologically inactive, and may counter the metabolic effects of acylated ghrelin 30 . Unacylated ghrelin in the predominant form in fetal circulation 11 , but little is known about its biological relevance. A recent intervention study demonstrated a negative association between oxidative stress and circulating ghrelin concentrations in adults 31 . There is an absence of data on perinatal oxidative stress in relation to ghrelin levels in fetuses/newborns. We observed a consistent negative correlation between maternal or fetal oxidative stress and fetal ghrelin levels. On the other hand, it has been proposed that both acylated and unacylated ghrelin may have a protective effect against oxidative stress in cell and animal model studies 32,33 . There is a possibility of reverse causality that lower ghrelin levels might be a cause of higher oxidative stress levels in the newborns. However, the consistent negative correlations of both maternal and cord blood oxidative stress biomarkers with cord ghrelin levels are reassuring, since it is unlikely that fetal ghrelin would cause oxidative stress in the mothers. The implications of fetal ghrelin concentration for postnatal metabolic health are unknown. Considering the plurality of biological effects of ghrelin, it is possible that the oxidative stress associated changes in fetal ghrelin levels may be related to a developmental metabolic programming effect on long-term susceptibility to obesity and metabolic syndrome in postnatal life.
Interestingly, we observed that the associations between F2-isoprostanes and fetal circulating ghrelin levels appear to be restricted to certain isomers (8-iso-PGF2α , 15(R)-PGF2α , 8-iso-15(R)-PGF2α , and iPF2α -IV). This finding requires confirmation from other independent cohort studies. F2-isoprostane isomers are biological active molecules. The 8-iso-PGF2α is the most studied F2-isoprostane isomer which has been associated with potent vascular constrictive effects, while research on the biological effects of other isomers has been limited, and isomer-specific biological effects may exist 34 . Much remains to be known about the biological effects of various F2-isoprostane isomers. There is an absence of data on the physiological relevance of various F2-isoprostane isomers in metabolic health. Our preliminary data indicate that certain F2-isoprostanes may be related to ghrelin synthesis or secretion during fetal development.
Apart from ghrelin, we did not observe any significant association between oxidative stress and other fetal metabolic health biomarkers including leptin, adiponectin, glucose-to-insulin ratio, proinsulin-to-insulin ratio, insulin, IGF-I and IGF-II. Also, we could not confirm an association between oxidative stress and poor fetal growth. Caution is warranted in data interpretation since the number of SGA infants is small (n = 14) in the study cohort. It should be noted that there is a tendency of under-reporting null association in the literature 35 . It has been recognized that fetal metabolic programming may occur within normal birth weight ranges 2 . It appears that in this relatively low risk pregnancy cohort (free of major pre-gestational illnesses), there was no significant impact of perinatal oxidative stress on fetal growth and most fetal metabolic health biomarkers. Preterm delivery has been associated with elevated insulin levels at birth and early childhood in a recent large prospective cohort study 36 . We could not detect any significant differences in cord blood oxidative stress and metabolic health biomarkers in preterm vs term births, but caution is warranted in data interpretation since the number of preterm infants is small.

Strengths and limitations.
Main study strengths are the largest pregnancy cohort on maternal and fetal oxidative stress, and objective biomarker measurements (assay labs' blindness to patients' clinical characteristics). Main study limitation is that we only have data on total ghrelin, but could not detect different forms of ghrelin. However, it is known that concentrations of acylated and unacylated ghrelin are highly correlated, and unacylated ghrelin accounts for about 99% of total ghrelin in cord blood 15 . The study is observational in nature, and we could speculate, but not affirm that the observed associations are causal.