PRENATAL PSYCHOSOCIAL STRESS EXPOSURE IS ASSOCIATED WITH INSULIN RESISTANCE IN YOUNG ADULTS

Abstract

Objective

To examine the association in humans between maternal psychosocial stress exposure during pregnancy and measures of glucose-insulin metabolism in the adult offspring.

Study Design

Healthy young adults whose mothers experienced major stressful life events during their pregnancy (n=36, Prenatal Stress, PS group, mean age 25 ± 5.14 (SD) years) and a comparison group (n=22, CG, mean age 24 ± 3.7 (SD) years) underwent an oral glucose tolerance test.

Results

Glucose levels were not significantly different across the groups, however, prenatally-stressed subjects showed significantly elevated 2h insulin (p=.01) and C-peptide levels (p=.03). These differences were independent of other major risk factors for insulin resistance, including birth phenotype (birth weight, length of gestation), family history of diabetes, gestational diabetes, body mass index, pro-inflammatory state, and smoking.

Conclusions

Higher insulin responses reflect relative insulin resistance in these prenatally-stressed young adults. This study is the first to provide evidence for a link in humans between prenatal psychosocial stress exposure and alterations in glucose-insulin metabolic function.

Introduction

A large number of epidemiological studies across the world have reported associations between markers of an individual’s birth phenotype such as low birth weight or small body size and subsequent risk of disease in adult life, including glucose intolerance, type 2 diabetes, dyslipidemia, hypertension, and insulin resistance.^1–4 These associations are independent of adult size and other established risk factors. Moreover, these effects extend continuously across the normal range of birth phenotype and are not merely a function of adverse birth outcomes such as low birth weight (LBW) or small-for-gestational age at birth (SGA).^{5, 6} It is, however, unlikely that birth phenotype per se plays a causal role in increasing risk of adult disease. Instead, birth phenotype is more likely a crude reflection of developmental processes in intrauterine life that may also influence the structure and function of physiological systems that underlie health and disease risk in later life.^{7, 8} Several studies of the effects of adverse early environment have focused on the role of pre- and perinatal nutritional factors. We and others have proposed that prenatal stress exposure represents yet another adverse environment that may contribute to both birth phenotype and the physiology of the developing organism.⁹ Maternal psychosocial stress during pregnancy has been shown in humans to predict low birth weight ^10–12 and preterm delivery,^{10, 12, 13} and other studies have linked these birth phenotypes with metabolic changes such as altered glucose tolerance (for recent review see ⁴). However, to the best of our knowledge, no human studies to date have examined the direct relationship between prenatal psychosocial stress exposure and changes in metabolic systems later in life.

Thus, the major objective of the present study was to examine the association between maternal psychosocial stress exposure during pregnancy and measures of physiology of their offspring in adult life. In this report we focus on metabolic markers related to glucose-insulin function and regulation. Because major predictors of insulin resistance and type 2 diabetes mellitus are family history of diabetes,^{14, 15} exposure to gestational diabetes,¹⁶ birth weight/size at birth,⁴ and prematurity,¹⁷ we adjusted for their potential effects in our study design. Because alterations in glucose-insulin metabolism may be secondary to high levels of adiposity,¹⁸ pro-inflammatory state,¹⁹ or behavioral factors such as smoking,²⁰ we also assessed the potential mediating role of these factors in the link between prenatal stress exposure and insulin resistance. In addition, postnatal measures of adversity including poor maternal care, adverse childhood experience and depression were assessed.

Material and Methods

Subjects

The study sample included a total of 58 subjects. Thirty-six young adults (mean age 25 ± 5.14 (SD) years, 31 women and 5 men) whose mothers experienced a high level of psychosocial stress (negative life events during pregnancy) constituted the prenatal stress group (PS). A sample of 22 subjects (mean age 24 ± 3.7 (SD) years, 17 women, 5 men) constituted the comparison group (CG). Subjects were recruited through announcements in local newspapers and to students at the university. Before entering the study, the absence of acute or chronic health problems was ascertained by self-report and confirmed by a medical examination. All subjects were non-smokers and reported to be medication free at the time of the metabolic assessments. As for maternal factors, obstetric complications were excluded. A copy of the prenatal medical record of the subjects’ mothers was obtained from each participant. By screening those records we identified one mother of a PS subject who developed gestational diabetes. This subject was excluded from all analyses. Participants received a modest monetary incentive on completion of the study. Written informed consent was obtained from all subjects. To avoid a potential self-selection bias, subjects were not informed about the hypothesized directions of the findings. This investigation was conducted in accordance with the guidelines described in the declaration of Helsinki, and the study protocol was approved by the ethics committee of the German Psychological Society, DGPs.

Conceptualization and assessment of prenatal psychosocial stress exposure

We adopted a conservative strategy for the conceptualization of prenatal stress in the present study. We defined a high level of prenatal psychosocial stress exposure in our “prenatal stress group” as the presence of major negative life events that occurred to the mother while she was pregnant (see Table 1 for list and frequency of events in the PS group). Psychosocial stress is a multi-component construct that includes the occurrence of negative life events, appraisal of the stress (e.g., degree of predictability and control), and psychological symptoms such as anxiety and negative affect. Because retrospective assessment of stress appraisals and symptoms is known to be unreliable, we focused on only the presence of negative life events during the index pregnancy. Moreover, we selected those events that are considered as highly stressful across individuals (see Table 1).

Table 1.

List of negative life events in the “prenatal stress group (PS)” that occurred to their mothers during pregnancy. One life event is listed for each mother. One mother reported marital infidelity of her husband followed by a divorce during her pregnancy is only listed once under the category “relationship conflicts”. Three mothers lost a family member after suffering a disease and are listed once in the “death of someone close” category and not under “severe illness of someone close” category.

Event		N (36)	%
Relationship conflicts	-divorce	16	44
	-break up
	-paternity denial
	-marital infidelity
Death of someone close	-partner	7	19
	-parent
	-other child
Severe illness of someone close	-cancer	5	14
	-heart attack
	-stroke
Severe financial problems	-loss of house by flooding	3	8
	-sudden unemployment of husband
	-foreclosure
Car accident		2	5
Unmarried, father not accepted by family		1	3
Becoming political refugee		2	5

In all subjects we conducted semi-structured interviews based on a questionnaire about exposure to major negative life events during the prenatal period that subjects were instructed to review with their mothers prior to the interview. In most of the cases (70%) we were able to verify this information by communicating directly with the mothers by phone, e-mail or letters. The subjects that were recruited to constitute the comparison group were asked to review the same questionnaire with their mothers to ascertain that their mothers had not experienced any negative life events during pregnancy.

Additional Questionnaires

In order to obtain an important aspect of the family environment during the postnatal period we administered the maternal care scale of the Parental Bonding Inventory (PBI; German version by Lutz et al.,²¹ originally developed by Parker et al.²² The PBI measures the self reported perception of being parented till 16 years of age. Studies assessing re-test reliability of the PBI suggest that this evaluation of parenting is a stable measure, which is not affected by confounding variables like dysthymia, neuroticism, depressive episodes or gender.^23–25 Furthermore, good validity of the PBI can be inferred from high agreement between sibling ratings.²³ Subjects’ and subjects’ mothers socio-economic status (SES) was assessed by educational level. A translated version of the Childhood Traumatic Events Survey ²⁶ was administered. The instrument screens adverse experience during childhood in six questions. In addition, a German version of the Centre for Epidemiological Studies Depression Scale (CES-D),²⁷ and the NEO Five Factor Inventory ²⁸ were administered.

Experimental procedures

Participants reported to the laboratory at 8 am after fasting overnight for 12 h. An intravenous catheter was inserted into an antecubital vein. After the fasting blood sample was drawn, the subjects were instructed to drink a glucose solution (75 g glucose, Caelo, Hilden, Germany) within 2 min. Blood samples were drawn 30, 60, 90 and 120 min following the glucose load. During a medical exam on a day prior to the oral glucose tolerance test (OGTT) subjects’ height, body weight, and waist-to-hip ratio were measured and body mass index (BMI) was computed. Percentage of body fat was assessed by a commercial digital electronic scale.

Serum blood samples were kept at room temperature for 15 min before they were centrifuged at 2000 × g and 4° C for 10 min. Serum was divided into aliquots and then stored at −80° C until analysis. Indicators of insulin resistance (levels of glucose, insulin, C-peptide), and the fat-derived hormones leptin and adiponectin were measured at all time points. In addition, from the fasting sample a lipid profile (triglycerides, total cholesterol, LP (a) cholesterol, HDL, VLDL, LDL) and two reliable markers of pro-inflammatory state (IL-6, TNF-α), that are associated with insulin resistance and type 2 diabetes, were assessed.

Biochemical analysis

Plasma glucose concentrations were measured by a standardized enzymatic photometric assay on an Olympus 2700 analyzer (Olympus, Hamburg, Germany). Insulin, C-peptide, leptin, adiponectin and cytokines were assessed using Luminex technique. Three multiplex kits were obtained from Linco Research Inc. (St. Charles, MO). Human Endocrine LINCOplex kit was used to assess insulin, C-peptide and leptin simultaneously. Adiponectin was analyzed using the Human Serum Adipokine LINCOplex Kit. Cytokines were assessed by multiple cytokine analysis kits (Human Cytokine LINCOplex Kit). Millipore multiscreen 96 well filter plates (Bedford, MA) were used for all LINCOplex kits. Assays were run in triplicates according to the manufacturers’ protocols. Data was collected using the Luminex-100 system Version 1.7 (Luminex, Austin, TX). Data analysis was performed using the MasterPlex QT 1.0 system (MiraiBio, Alameda, CA). A five-parameter regression formula was used to calculate the sample concentrations from the standard curves. Determinations of triglycerides, total cholesterol, LP (a) Cholesterol, HDL, VLDL, and LDL were obtained using the SPIFE^® VIS Cholesterol kit Pro 185 (Helena Laboratories, Beaumont, USA) according to the manufacturer’s instructions. For all analytes intra- and inter-assay coefficients of variance varied between 2.6 and 10.4 %.

A homeostatic model assessment for insulin resistance index (HOMA-IR) was calculated.²⁹

Statistical analyses

We first examined the associations between prenatal psychosocial stress and glucose, insulin, and C-peptide levels. Next, we examined the associations between prenatal stress and other risk factors for insulin resistance, including BMI, hormones in adipose tissue (leptin, adiponectin), inflammatory state (measured by levels of pro-inflammatory cytokines), socioeconomic status, birth phenotype (birth weight, length of gestation, birth weight adjusted for length of gestation as expressed by growth percentiles at birth according to a US reference group (US Natality dataset) and family history of diabetes. Last, the risk factors that were significantly associated with prenatal stress exposure were included in multivariate analyses to examine whether or not the effects of prenatal stress on insulin resistance were independent of these factors.¹

To approximate normal distribution, glucose, insulin, C-peptide, leptin and adiponectin levels were log-transformed before entering in the analyses. The interrelationship between birth weight and length of gestation and parameters of insulin resistance was assessed applying Pearson correlations. General Linear Model (GLM) analyses were performed to assess differences between the two groups in birth phenotype, anthropometric measures, blood lipid levels, SES, depression score and maternal care. Chi-square analyses were conducted to assess differences between the two groups in the frequency of the events assessed by the Childhood Traumatic Events Survey. Further GLMs were computed to assess the repeated measure effect time, the between-subject effect group as well as the interaction time × group for endocrine responses to the glucose load. Since in clinical practice fasting and 2h levels of glucose, insulin and C-peptide are used as markers of insulin resistance and diabetes, group differences in those single time points were also analyzed using GLMs. In addition, GLM procedures that controlled for the impact of BMI were computed. Greenhouse-Geisser corrections were applied were appropriate and only adjusted results are reported. All results are presented in original units as the mean ± standard error of the mean (SEM).

Results

Glucose, insulin and C-peptide concentrations

Figure 1 depicts mean glucose, insulin and C-peptide responses to the oral glucose tolerance test. As expected, glucose, insulin and C-peptide rose significantly in response to the glucose load (main effect time: glucose: F_{(2.68;150.08)} = 60.57, p<.001; insulin: F_{(2.82;138.07)} = 123.87, p<.001; C-peptide: F_{(1.8; 84.64)} = 40.35, p<.001). While glucose levels did not differ significantly between the two groups, insulin levels were higher in PS subjects (main effect group: F_(1;49) = 4.21, p=.046, trend for interaction group × time: F_{(2.82; 138.07)} = 2.33, p=.08). These findings were paralleled by a trend for higher C-peptide levels in PS subjects in response to the glucose load (trend for interaction group × time F_{(1.8; 84.64)} = 2.39, p = .10; main effect group n.s.). Fasting insulin levels were on average 58% and 2h levels 59% higher in PS subjects (fasting levels 93.23±15.41 vs. 58.94±10.63 pmol/L, F_{(1; 51)}=3.34, p=.07; 2h levels: 208.56±22.96 vs. 130.89±19.63 pmol/L, F_{(1; 52)} = 7.12, p=.01). 2h C-peptide levels were 40% higher (530.5±62.72 vs. 379.68±75.45 pmol/L, F_{(1; 51)} = 5.28, p=.03), while basal C-peptide levels were not different. There was a statistical trend for a 64% higher HOMA-IR (2.97±.5 vs. 1.81±.35, F_{(1; 51)} = 3.51, p=.07.

Figure 1.

Mean glucose, insulin, C-peptide and leptin responses (± SEM) to an oral glucose tolerance test (OGTT) in prenatally stressed (PS, “black circles”) and comparison group (CG, “white triangles”) subjects. Glucose levels were not significantly different across the groups, however, PS subjects showed significantly elevated 2h insulin (p=.01) and C-peptide levels (p=.03), as well as higher leptin levels at all time points during the OGTT (p=.05).

Anthropometric measures

BMI was significantly higher in the PS group than in the CG (24.6±.74 vs. 22.5±.49, F_(1;51)=3.51; p=.04). Percentage of body fat was slightly but not significantly higher in PS subjects (30.77±1.32 vs. 28.32±1.59, n.s.), and there was no difference in waist-to-hip-ratio between the two groups.

Leptin, adiponectin

As expected, leptin and adiponectin levels did not change significantly in response to glucose stimulation. Leptin levels were higher in PS subjects at all time points (main effect group: F_(1;52) =3.89, p=.05, see figure 1), PS subjects showed on average 98% higher levels at baseline (768.24±122.76 vs. 391±62.52 pg/ml) compared to CG subjects while adiponectin levels did not differ at any time during the OGTT.

Proinflammatory cytokines

The two groups did not differ significantly in basal levels of proinflammatory cytokines (IL-6: PS: 64.19±12.88, CG: 98.87±21.77 pg/ml, F_(1;48)=2.14, p=.15; TNF-α: PS: 3.62±.16, CG: 4.09±.22 pg/ml, F_(1;52)=2.57, p=.16).

Cholesterol concentrations

There were no differences in total cholesterol, triglycerides and LPa cholesterol between the two groups. However PS subjects showed 16% lower HDL levels (50.36±2.43 vs. 59.82±4.11 mg/dl, F_(1;51) =4.41, p=.04) and 138% higher VLDL levels (63±4.48 vs. 28±4.72 mg/dl, F_(1;51)=28.47, p<.001), while LDL levels were on average 33% lower in PS subjects (65.71±4.92 vs. 99.27±6.66 mg/dl, F_{(1; 51)}=15.98; p<.001).

Family history of diabetes

One subject in the PS group and two subjects in the comparison group reported family history of diabetes. Neither inclusion nor exclusion of these subjects changed the size or direction of the above reported effects.

Birth outcomes

PS subjects showed slightly lower birth weight (3305.86±99.01 vs. 3536.00±88.14 g) but this difference was not statistically significant (F_(1;49)=2.11; p=.15). There were no differences in length of gestation (PS: 39.26±.33, CG: 39.94±.29 wk, F_(1;49)=1.67; p=.20), nor in birth weight adjusted for length of gestation as expressed by growth percentiles (PS: 41±4.15, CG: 37±4.36, F_(1;49)=.30, p=.59). There was no association between birth weight or length of gestation and parameters of insulin resistance (glucose, insulin and C-peptide levels, HOMA-IR).

Questionnaires

Subjects and subjects’ mothers did not differ in SES, assessed by educational level, and there were no differences in maternal care scores between the two groups (PS: 23±1.86, CG: 26±1.88, F_(1;49).=1.74, p=.19). The two groups neither differed significantly in the frequency of the events assessed by the Childhood Traumatic Events Survey (p > .25 for all events), nor in their depression score (PS: 10±7.6, CG: 8.5±5.9, F_(1;56) = 1.75, p = .20) or their degree of neuroticism (F_(1;56) = 1.29, p =.24)

Adjustment for BMI

Because prenatal stress was associated with insulin resistance as well as with BMI, GLM procedures that controlled for the impact of BMI were computed. After controlling for BMI all associations between insulin and PS were stronger (main effect group: F_(1;48)=5.78, p=.03, interaction group × time: F_{(4; 133.7)}=2.86, p=.04), as were associations between PS and 2h levels of C-peptide (p=.02) and HOMA-IR (p=.05). The trend for higher leptin levels in PS subjects disappeared after controlling for BMI. Adjusting for BMI only marginally reduced the size of the effects for VLDL (F_(1;51)=27.02, p<.001) and LDL (F_(1;51)=13.8, p=.001) whereas the difference in HDL disappeared after including BMI in the model (F_{(1; 51)}=1.44, p=.24).

Comment

Our results suggest an association between exposure to prenatal psychosocial stress and glucose-insulin metabolism in young adulthood. PS subjects exhibited higher insulin and C-peptide levels in response to an OGTT. In the pathophysiology of type 2 diabetes, relative insulin resistance and high insulin secretion are followed by pancreatic failure and insulin deficiency. Thus, the earlier stage of type 2 diabetes is characterized by high insulin secretion in response to a glucose load.⁴ Since our subjects were young adults, it is possible that we may have been observing a relatively early stage of the pathophysiological process leading to glucose intolerance and type 2 diabetes.

BMI was significantly higher in the PS group, while waist-to-hip-ratio and percentage of body fat did not differ between the two groups. This suggests that the differences in BMI may be due to differences in subcutaneous rather than abdominal fat. Since the typical centralisation of fat distribution occurs with advancing age, differences in abdominal fat and waist to hip ratio may become more pronounced in later adulthood.

We found higher VLDL and lower HDL levels in subjects exposed to prenatal stress, while total cholesterol and triglycerides were not different, and LDL was lower. These findings suggest differences in fat storage and mobilisation in PS subjects.

PS subjects had higher leptin levels at all time points during the OGTT. This finding, together with the higher BMI, may reflect leptin resistance in those subjects. According to Breier et al.,³⁰ leptin resistance, in addition to insulin resistance, is a critical endocrine defect in the pathogenesis of programming-induced obesity and metabolic disorders.

The precise mechanisms underlying the development of insulin resistance are still poorly understood. Among the major risk factors for insulin resistance and diabetes are maternal/paternal history of diabetes,^{14, 15} exposure to gestational diabetes,¹⁶ socio- economic status,³¹ and size/weight at birth.⁴ In our sample, very few subjects had a family history of diabetes or exposure to gestational diabetes, and neither inclusion nor exclusion of these subjects changed the size of the observed effects. Socio-economic status of the subjects and their mothers was not different between the two groups. All markers of birth phenotype, such as length and weight at birth, length of gestation, and head circumferences, were obtained from the mother’s obstetric record, were in the normal range, and there was no correlation of birth weight or length of gestation with markers of insulin resistance. It is well-recognized that in some instances insulin resistance may be secondary to other disease states such as obesity and disordered fat storage and mobilisation,¹⁸ inflammation,¹⁹ and behavioral and lifestyle factors such as smoking, diet and physical inactivity.^{20, 32} In our sample the effects of prenatal stress on insulin resistance were independent of BMI, inflammatory state and smoking. Although BMI was higher in PS subjects, the association between prenatal stress and markers of insulin resistance did not change or was even stronger after adjusting for BMI. There were mo differences between the two groups with respect to the pro-inflammatory cytokines TNF-α and IL-6 that have been linked to insulin resistance and diabetes mellitus.¹⁹ From among the behavioral factors known to influence insulin resistance we excluded smoking, however, we did not assess diet or physical activity.

The concepts of the developmental origins of health and disease and of predictive adaptive responses (PAR) propose several explanations for how insulin resistance may be programmed in early life (see ^{8, 33} for recent overviews). Fetal programming may occur via exposure of the fetus to excess glucocorticoid hormones.³⁴ Maternal stress during pregnancy results in higher glucocorticoid exposure of the fetus and has been proposed as one example of early adversity. We and others (e.g.,^{35, 36}) have discussed possible mechanisms of how prenatal stress may be transduced from the pregnant mother to her fetus, such as transplacental transport of maternal stress hormones (cortisol) to the fetus; maternal stress-induced release of placental hormones (CRH) that enter the fetal circulation; and maternal stress-induced alterations in utero-placental blood flow. Changes in the utero-placental blood flow produced by maternal stress hormones may result in nutrient restriction in the fetus, which, in turn, may program insulin resistance in the developing organism. Furthermore, cortisol, acting as a catabolic hormone, induces several processes leading to increased concentrations of blood glucose. It is therefore possible that the hormonal changes in the fetal circulation induced by maternal stress elevate levels of glucose in maternal and fetal circulation and contribute to the development of insulin resistance. Experiments in several animal models and one human study show that insulin resistance and other manifestations of the metabolic syndrome can be induced by manipulating maternal nutrition or exposing the mother to synthetic glucocorticoids or prenatal stress.^37–42

There are certain limitations to our study. First and foremost, prenatal stress exposure was assessed retrospectively. Although retrospective assessments of psychosocial factors such as stress are prone to biases such as “after-the-fact” reporting, biases produced by personality/mood, and memory biases, we believe it is unlikely that any of these biases operated in our study sample for the following reasons: Subjects with adverse health outcomes are more prone to retrospectively reporting higher levels of prior adverse exposures (i.e., after-the-fact retrospective reporting bias), however, it is unlikely that this bias was present in the current study because all subjects received the same information before entering the study; the subjects were not informed about the hypothesized direction of the effects; subjects (as well as the experimenters) were blind to and had no a priori knowledge about the results of the study outcome (i.e., measures of glucose-insulin metabolism); and none of the subjects had any underlying disease. In terms of reporting bias produced by personality/mood state, the subjects in the two groups did not differ in either neuroticism or depression scores (i.e., the major personality/mood constructs that underlie self-report bias). Last, retrospective assessments also are prone to memory biases, but it also is unlikely that this potential bias affected one group of subjects more than the other. If at all, memory bias and under-reporting stress in the comparison group could only bias the results in the direction of possibly dilute the observed effects.

Prenatal psychosocial stress was operationalized as the occurrence of major negative life events experienced by the mother at any time during her pregnancy. Retrospective life event assessments certainly are less biased than retrospective assessments of other components of stress such as perceived severity of stress appraisals and symptoms. While it is known that stressful life events are more likely to occur in women of lower social class, there were no differences in SES in our two study groups. Finally, it is possible that prenatal stress exposure is associated with adverse postnatal experiences such as poor maternal care and presence of other stressors during childhood. We, however, assessed these constructs in our sample and found no differences across the two study groups in perceived maternal care after birth, in the frequency of early traumatic events assessed by the Childhood Traumatic Events Survey, in depression and neuroticism levels.

Women are overrepresented in our study and we did not have the power to adequately examine possible sex differences as modulators of the effect of prenatal stress exposure on glusocu-insulin metabolism. However, Newsome et al.⁴ concluded in their review that programming effects of glucose and insulin and the prevalence of type 2 diabetes may be similar for males and females.

Extensive research is under way to advance our understanding of prenatal stress and its possible consequences for development and subsequent disease risk. Similarly, the pathogenic changes leading to type 2 diabetes mellitus comprise a broad array of factors, many of which are not yet fully understood. This study contributes new data suggesting a link between prenatal stress exposure and metabolic changes in early adulthood, thus adding further evidence to the growing awareness of pre-disease pathways not only being multi-faceted but also having their foundations very early in life.