Maternal stress retards fetal development in mice with transcriptome-wide impact on gene expression profiles of the limb

Abstract

The environment of a pregnant mother has a life-long impact on later life of offspring. Maternal stress is known to cause low birth weight and programs several physiological dysfunctions in offspring. However, the direct effects of maternal stress on the developing fetus remain largely unknown. The present study focused on the effect of chronic maternal stress on the developmental program and its molecular mechanisms. Pregnant mice were given 6-hour immobilization stress every day from 8.5 days post coitum. Fetal body weight was significantly decreased by maternal stress throughout development. Importantly, developmental events were retarded in the stressed fetuses. Around embryonic day 13.5 (E13.5), the developmental increment of somite numbers was delayed, although this difference recovered by E15.5. Limb bud formation and regression of interdigital webbing were also retarded by approximately 0.5 days. Subsequently, transcriptomes of developing limbs were analyzed by cDNA microarrays. Approximately, one-tenth of detected transcripts were significantly influenced by maternal stress. Q-PCR AQ analyses further demonstrated that the expression of a subset of limb development-associated genes, including Igf1, Aldh1a2, and Acta1, was changed in the stressed fetus. In conclusion, our findings suggest that maternal stress can retard limb and somite development in mice, with profound impacts on the developmental genetic program of limb.

Keywords::

• Maternal stress
• environmental programing
• embryonic development
• developmental retardation
• gene expression
• growth retardation

Introduction

It is a common notion that the environment confronted by a pregnant mother can program the quality of her offspring's life in a prominent and persistent manner (Waterland and Michels 2007; Seckl 2008). One of the most compelling pieces of evidence for this notion has been provided by studies on the programing effect caused by stress imposed on pregnant females, the so-called maternal stress (Cottrell and Seckl 2009; Lupien et al. 2009). Adult offspring of stressed mothers have been reported to exhibit various physiological dysfunctions, including abnormal hypothalamus–pituitary–adrenal (HPA) axis activity and altered brain function and behavior (Chung et al. 2005; Son et al. 2006, 2007; Mueller and Bale 2008; Weinstock 2008).

Maternal stress may influence the developmental plan of the offspring in both prenatal and postnatal periods. Prenatally, stress applied to the pregnant female can influence fetal development via stress hormones (Weinstock 2005; Seckl 2008) or reduced uteroplacental blood flow (O'Donnell et al. 2009). Dams often continue to influence the postnatal development of her pups through maternal care behavior (Meaney 2001) and milk (Wan et al. 2007). Although postnatal manipulations, including adoption and boosting maternal care by handling, have been reported to reverse the effect of maternal stress (Maccari et al. 1995; Barbazanges et al. 1996), it remains unclear whether the reversal of the adverse effect of maternal stress by postnatal manipulation is caused by preventing the transmission of maternal stress in the postnatal period or by recovery from the adverse effects encoded in the prenatal period. However, the observation that low birth weight for gestational age is caused by maternal stress (Son et al. 2006) implies that the developing fetus can be directly influenced by maternal stress.

For better evaluation of the prenatal effects of maternal stress, it is necessary to analyze the developmental processes in the fetus in the womb. Since the consequences of maternal stress barely lead to anatomical malformation of organs in adulthood, it is important to examine the development of specific organs as model systems, which can provide critical milestones and in which molecular mechanisms have been intensively scrutinized. The developing limb is a prime candidate for this purpose because the formation of limbs has been the subject of extensive developmental studies for decades (Mackem 2006). Two marked events characterize limb development: growth of the limb bud and formation of the digits. In mouse, the limb buds start to bulge out of the trunk through an interaction between the apical ectodermal ridge (AER) and mesenchymal cells from embryonic day 9.5 (E9.5). Around E14.5, formation of fully separate digits is achieved by the removal of interdigital tissue evoked by apoptotic cell death. Several molecular and cellular events regulating the limb development, such as the cascade of signaling molecules, cellular proliferation, programed cell death, and differentiation into various cell types, have been well documented, providing an ideal experimental model for studies on the encoding processes of programing by maternal stress.

Although adult offspring have been a focus of extensive studies regarding the consequences of maternal stress, the effect of maternal stress on the developing fetus in the uterus remains largely unknown. In the present study, we explored the development of the fetus, primarily focusing on limb development, and gene expression profiles therein to gain insight into the direct effects of maternal stress on the developing fetus.

Materials and methods

Animals

ICR mice, obtained from the Laboratory Animal Center at Seoul National University, were kept in temperature-controlled (22–23°C) quarters under a 12/12-h light/dark photoperiod (lights on at 07:00 h). Standard mouse chow and water were available ad libitum. All animal procedures were approved by the Animal Care and Use Committee of Seoul National University.

Preparation of pregnant mice

Females were injected with pregnant mare's serum gonadotropin (PMSG; 14:00–15:00 h of the day) at the age of 6 weeks, and mated with males 2 days after the PMSG injection. Female mice whose embryos were to be collected at E10.0 were allowed to be in the same cage with stud males only for 2 h (04:00–06:00 h) to ensure temporal resolution among groups. Otherwise, females were allowed to mate for an entire night (18:00–09:00 h of the next day) and then randomly assigned to control or stressed groups. Plug formations were checked by visual inspection and females with plugs were considered to be 0.5 days post coitum (dpc).

Maternal stress procedure

The maternal stress procedure was performed as previously described (Chung et al. 2005; Son et al. 2006, 2007). Briefly, pregnant female mice were randomly divided into control and stress groups. Mice in the stress group were individually placed in a restrainer (a transparent plastic cylinder, 3 cm in diameter and 9 cm long) daily for 6 h (10:00–16:00 h) from 8.5 dpc to the day on which they were sacrificed. When female mice were placed in the restrainer, stress groups were isolated in another room to avoid stressing the control group.

Measurement of food intake

Food intake was determined by weighing the food and spillage collected on a filter paper placed at the bottom of each cage. Minimal amount of bedding to form a nest was also added to reduce unexpected stress. Food intake was assessed as the difference between the sum of the remaining and spilled food weight before and after each feeding period.

Measurement of serum corticosterone and norepinephrine

Mice were sacrificed between 13:00 and 14:00 h, and their trunk blood was collected and centrifuged at (10,000 g, 20 min) to obtain sera. Corticosterone (CS) levels were assayed using a commercial radioimmunoassay kit according to the manufacturer's instructions (Diagnostic Products, Los Angeles, CA, USA). Serum concentrations of norepinephrine (NE) were determined by HPLC coupled to an electrochemical detector (HPLC-ECD) system (Gilson, Middleton, WI, USA) as previously described (Kang et al. 1998). Briefly, serum samples were diluted in 0.1 N perchloric acid (Sigma, St Louis, MO, USA), containing 0.04 M sodium metabisulfite (Sigma) and centrifuged at 10,000 g; supernatants were recovered, microfiltered, and subjected to HPLC-ECD.

Skeletal analysis

Embryos were harvested by Caesarean section. The fetuses were skinned, eviscerated, and fixed in 95% ethanol. Subsequently, acetone was used to remove fat. The skeletons were then stained with Alcian blue/Alizarin red (Sigma) and sequentially cleared in 1% potassium hydroxide. Stained embryos were photographed and the number of somites was counted.

Whole-mount immunohistochemistry

Whole-mount immunohistochemistry was performed on embryos fixed with 4% paraformaldehyde and 0.15% picric acid. Fixed embryos were incubated for 2 days with a polyclonal antibody that recognizes FGF8 (N-19, Santa Cruz Biotech., Santa Cruz, CA, USA) diluted with blocking solution (5% skim milk, 0.5% Triton X-100, 5% DMSO in 0.1% PBS-T). Primary antibody binding was subsequently detected by incubation with secondary antibodies conjugated with horseradish peroxidase (Santa Cruz Biotech.) and visualized by incubation with a 3,3′-diaminobenzidine (DAB) tablet (Sigma) solubilized in PBS-T.

Analysis of cell death

Cell death was detected by vital dye staining with Nile Blue A (Sigma). Embryos were collected in ice-cold PBS, incubated in pre-warmed DMEM containing 0.001% Nile blue A at 37°C for 1 h, washed in PBS for 5 h at 4°C, and photographed. Activated caspase-3, a marker of apoptotic cells, was detected by Cleaved Caspase-3 (Asp175) antibody (Cell Signaling, Beverly, MA, USA).

Microarray

cDNA microarray analysis was performed as described previously, with minor modifications (Sun et al. 2007). Autopods from 2 to 3 l were excised from both forelimb and hindlimb at E13.5 and 14.5 for controls and at E14.5 for the stressed animals. Total RNA was isolated by the single-step acid guanidinium thiocyanate-phenol-chloroform method. Each reaction involving a single GeneChip hybridization was started with 5 μg RNA, which was reverse transcribed, labeled, hybridized, and stained according to standard Affymetrix protocols (Affymetrix Gene Chip Expression Analysis Technical Manual). Affymetrix GeneChip Mouse 430A 2.0 arrays (Affymetrix, Santa Clara, CA, USA) were used. An Affymetrix GeneChip scanner operated with GeneChip Operating Software version 1.3 (GCOS; Affymetrix) was used to generate original array images. The average difference of each probe set, which is a measure of the relative abundance of a transcript, and signals and detection calls such as present or absent were computed by GCOS. All hybridization intensities were corrected by a set value for a total intensity of 500, and the scaling factors were between 1.29 and 2.83. Nomenclature, sequence, and gene ontology (GO) information was obtained from the NetAffx^TM Analysis Center http://www.affymetrix.com/analysis/netaffx, Entrez Gene http://www.ncbi.nlm.nih.gov/gene, and GO http://www.geneontology.org. The changes in transcripts between groups with change P-values of less than 0.01 or more than 0.99 (calculated by signed rank analysis using Affymetrix GCOS) were considered to be significant changes. The detailed list of changed transcripts with the log fold of normalized mean value is provided as supplemental data. Functional classification of changed transcripts was based on MGI GO slim database (http://www.spatial.maine.edu/∼mdolan/MGI_GO_Slim.html), which harbors minor modifications with original GO classification.

Real-time Rt-pcr

Analyses of mRNA levels by real-time RT-PCR were carried out as previously described (Son et al. 2008). Briefly, total RNA was isolated by the single-step acid guanidinium thiocyanate-phenol-chloroform method. Each RNA (500 ng) was reverse-transcribed with moloney murine leudemia virus (MMLV) reverse transcriptase (Promega, Madison, WI, USA). Aliquots of cDNA were then subjected to quantitative real-time PCR using the 7300 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) in the presence of SYBR Green I (Sigma). Gene expression levels were normalized with the level of Tbp. Primer sequences used for real-time PCR are listed in Table I.

Table I.  Primer sequences for real-time PCR.

Symbol	Forward primer	Reverse primer	PCR product size (bp)
Igf1	TGG ATG CTC TTC AGT TCG TG	GCA ACA CTC ATC CAC AAT GC	113
Aldh1a2	CCA TTG GAG TGT GTG GAC AG	ACG GTG TTA CCA CAG CAC AA	94
Ctgf	GCA GAC TGG AGA AGC AGA GC	GCA GCC AGA AAG CTC AAA CT	137
Nov	GCA GAA CAT GAG CCT CTT CC	GCG CAG GTC GGT GAT ATA CT	146
Acta1	TCA CTG AGC GTG GCT ATT CC	TGG CCA TCT CAT TCT CGA AG	108
Actn2	CAC CCA GGA GCA GAT GAA TG	TTC ACC CAA GTC ATA GCC CA	124
Cryab	CTG GAT TGA CAC CGG ACT CT	CCC CAG AAC CTT GAC TTT GA	109
S100a6	CTA GCC CAG TGA TCA GTC AT	AAC TCC TTC AGC TCC TTC TT	134
Apoe	CTG ATG GAG AAG ATA CAG GC	GGA CAG GAG AAG GAT ACT CA	89
Snai2	ACA TTG CCT TGT GTC TGC AA	CAG TGA GGG CAA GAG AAA GG	110
Nipbl	TCC TCC AGT GCT CTG TGC TC	TTA GCA GGG GAC GTT TTC CT	299
Smc2	ACC AAG CGC AAA GAG CTA CA	TCG TTA CTG GAC AAG CCA GC	188
Tbp	GGG AGA ATC ATG GAC CAG AA	CCG TAA GGC ATC ATT GGA CT	113

Statistical analyses

All data, including serum levels of CS and NE, food intake, abortion rate, body weight of fetus, number of somites, and mRNA expression levels at given time points, were statistically evaluated using Student's t-test. Statistical significance was set at p < 0.05.

Results

Maternal responses to chronic immobilization stress

Serum levels of stress hormones, including CS and NE, were determined in pregnant females at 18.5 dpc to confirm the validity of the stress scheme (Figure 1A, B, respectively). Serum CS levels increased over twofold as a result of chronic immobilization stress (controls (CTL), 0.848 ± 0.084 vs. stressed (STR) females, 2.097 ± 0.087μg/ml; p < 0.01; n = 5 for each group). Approximately, twofold increase was observed for serum NE (controls, 100.00 ± 8.00 vs. stressed females, 176.99 ± 11.00% of control; p < 0.01; n = 4 for each group).

Figure 1.  Maternal responses to repeated restraint stress. Level of serum CS (A) and NE (B) were measured at 18.5 dpc from pregnant mothers (n = 5 for each group for CS, 4 for NE; **: P < 0.01 vs. CTL). (C). Body weight of pregnant female was measured at 14.5 dpc (n = 8∼9 for each group). (D). Food intake of pregnant female was measured from 7.5 dpc to 20.5 dpc (n = 8∼9 for each group; *: P < 0.05 vs. CTL). Data are presented as means ± SE.

Because chronic stress can influence the nutritional status of pregnant females, food intake of pregnant mothers, an important determinant of fetal growth and development, was assessed from 7.5 to 20.5 dpc. During the first 4 days of stress sessions, food intake of stressed females decreased to 70% of that of controls (Figure 1D), but food intake of stressed females soon recovered to a normal level. There was no significant difference in body weight between stressed mothers and controls at 14.5 dpc, when the differences in cumulative food intake were maximized (controls, 48.27 ± 1.37 vs. stressed females, 45.74 ± 1.15 g; n = 8 for controls and 9 for stressed; Figure 1C).

Fetal growth and abortion rate in stressed dams

The abortion rate at E14.5 was more than threefold higher in the stressed group (control, 11.37 ± 2.21%, n = 156; stressed, 36.24 ± 8.47%, n = 164; p < 0.01; Figure 2A), consistent with previous reports that maternal stress often results in smaller litter sizes (Mulder et al. 2002). Along with the abortion rate, the size and body weight of the fetuses and neonates were measured to examine fetal growth. At E10.0, E14.5, and postnatal day 1 (P1), stressed fetuses were strikingly smaller than control fetuses (Figure 2B). Accordingly, body weight of stressed fetuses decreased significantly to approximately half of that of control fetuses at all time points examined (Figure 2B).

Figure 2.  Increased abortion rate and growth retardation in maternally stressed fetus. (A). Abortion rates were determined at E14.5. Numerals on the bars indicate the number of fetus in each group (**: P < 0.01 vs. CTL). (B). Body weights were measured in E10.0, 14.5 embryos, and P1 neonates. Numerals on the bars indicate the number of examined mice (**: P < 0.01 vs. CTL). Data are presented as means ± SE. All experiments were performed on at least three different litters.

Effect of maternal stress on the development of somites

To explore changes in developmental processes caused by maternal stress, we counted the number of somites in Alcian blue- and Alizarin red-stained embryos around E14.5 (Figure 3A). The number of somites is known to increase until E15.5, reaching approximately 65 (Kaufman 1998). The number of somites in control fetuses increased from E13.5 to E15.5, as expected (Figure 3B). An increase was also observed in stressed fetuses, but the number of somites in stressed fetuses was significantly smaller than that in controls at E13.5 (control, 57 ± 0.5 vs. stressed 53 ± 0.5; p < 0.05; n = 8 for each group). Because the difference in the number of somites became less and statistically insignificant after E14.5, it appears that the normal schedule of somite development was regained in the stressed fetuses.

Figure 3.  Somite development in control and maternally stressed embryo. (A). Skeleton of E14.5 fetus was stained. Bone was stained with Alcian blue and cartilage was stained with Alizarin red. (B). Number of somites was counted at E13.5, 14.5, and 15.5. Data are presented as means ± SE (n = 8 for each group; *: P < 0.05 vs. CTL). Consistent results were obtained in two different litters.

Limb development in maternally stressed fetuses

To further scrutinize the effect of maternal stress on the developmental process, we explored the development of limbs and underlying molecular events. First, we examined outgrowth of the limb bud, an embryonic structure forming the future limbs. Lateral plate mesoderm initiates the formation of the limb bud by gradual swelling and sequentially inducing a specialized ectodermal structure known as the AER. Later, the AER serves as a signaling center for the determination of the proximodistal axis and the proliferation of the mesenchymal cell mass. During this process, Fgf8 expressed in the AER plays a key role in the maintenance of AER function, such as supporting proximodistal growth of limb buds and inducing downstream signaling pathways (Mackem 2006). In control fetuses, an extended limb bud and immunoreactivity against Fgf8 in the forelimb and rather a smaller limb bud and distinct Fgf8-immunoreactivity in the hindlimb around the AER were observed at E9.75 (Figure 4B). In contrast, it was only at E10.0 that distinct Fgf8-immunoreactivity could be found in the AER of the forelimb in the stressed fetuses. In the hindlimbs of stressed fetuses, limb buds and Fgf8-immunoreactivity could be found only after E10.25. This indicates that maternal stress caused retardation in the onset of Fgf8 expression and limb bud outgrowth by approximately half a day.

Figure 4.  The growth of limb bud in control and maternally stressed embryo. (A) and (B). E9.75, 10.0, and 10.25 fetuses were fixed and immunostained with anti-Fgf8 antibody as described in the Materials and Methods. Fgf8-immunoreactivity was visualized as brown color by the DAB method. The arrowhead indicates developing forelimb and the red arrow indicates the developing hindlimb. These results were reproduced in three independent litters.

Around E14.5, separate digits begin to be sculpted by the regression of interdigital tissue mass via apoptotic cell death. In the control fetuses, separate digits were formed at E14.5 in both fore and hindlimbs (Figure 5A). However, interdigital tissue mass still existed at E14.5 in the stressed fetuses. Separate digits were found only after E15.5 in both fore and hindlimbs. To examine whether lack of cell death between E13.5 and E14.5 was involved in the retardation of the regression of the interdigital tissue mass, fetal limbs at E14.0 were stained with Nile Blue A, a vital dye. While dead cells, which stained blue, were found in the interdigital tissue in control limbs, staining was limited to the distal border region in the limbs of stressed fetuses, indicating that regression of the AER, but not the interdigital tissue, had started (Figure 5B). Whole-mount immunohistochemistry against activated caspase-3, an apoptosis marker, also showed decreased immunoreactivity in the interdigital region of the stressed fetuses compared with that of the controls at E14.0 (Figure 5C). Consistent with the delayed development of the somites, maternal stress delayed two key events in limb development, namely, the onset of limb bud development and the regression of the interdigital tissue mass.

Figure 5.  Retardation of digital morphogenesis in maternally stressed embryos. (A). Morphology of forelimb and hindlimb from E13.5, 14.5, and 15.5 was compared. (B). Cell death was detected in the forelimb and hindlimb at E14.0 by Nile Blue A staining, as described in the Materials and Methods. (C). A 15-μm-thick section of hindlimb from an E14.0 fetus was immunostained with anti-activated caspase-3 antibody. These results were independently reproduced in more than three litters.

Effects of maternal stress on transcriptome expression

In the next set of experiments, we sought to assess the influence of maternal stress on global gene expression profiles using the cDNA microarray analysis. For this experiment, groups were set up as follows after considering developmental stage based on limb morphology and embryonic day: control E13.5 (morphology-matched control), control E14.5 (age-matched control), and stressed E14.5 (n = 2 with pooled samples each from three different litters for each group). Among 22,690 transcripts present on the chip, 14,685 transcripts were detected to be significantly expressed and were subjected to further analysis. Unsupervised hierarchical clustering (MAS 5.0, Affymetrix) revealed a distinct molecular pattern of the transcriptome in stressed E14.5 fetuses, which could be distinguished from a continuum between control E13.5 and E14.5 (Figure 6A), suggesting that maternal stress caused a marked and unique disruption in the transcriptome profile of developing limbs.

Figure 6.  Analysis of global gene expression in the fetal limb using a cDNA microarray. (A). Expression levels were clustered according to their expression pattern. Among 22,690 transcripts included in the microarray, 14,685 transcripts detected significantly were subjected to further analyses. (B) and (C). Functional classification of differentially regulated gene transcripts. Up-regulated or down-regulated genes in the limb of maternally stressed fetus compared with that of the morphology-matched control (control E13.5, B) or gestational age-matched control (control E14.5, C), respectively, were classified into functional classes according to the MGI slim database. Numerals in the pie charts represent the percentage of each class of genes of the total number of up- or down-regulated genes in the maternally stressed group.

To quantitatively assess the effect of chronic maternal stress on the transcriptome profiles in the developing limb, we have counted the number of significantly changed transcripts based on the P-value calculated by signed rank analysis (Table II). Approximately, 10% of the total expressed transcripts in the limb of the stress group showed significant changes compared with either morphology-matched control (control E13.5, 6.7% increased and 2.9% decreased) or gestational age-matched control (control E14.5, 8.7% increased and 1.9% decreased), strongly suggesting that the altered transcriptome profile by maternal stress is not merely a corollary to the developmental stage or the gestational age.

Table II.  Pair-wise comparison of global gene expression profiles.

Experimental group	Control group	Number of increased transcripts^*	Number of decreased transcripts^*^*
Control E14.5	Control E13.5	1428 (9.2%)	1061 (6.8%)
Stressed E14.5	Control E13.5	1119 (7.2%)	242 (1.6%)
Stressed E14.5	Control E14.5	930 (6.0%)	347 (2.2%)

*: P < 0.01 by signed rank analysis calculated by GCOS (Affymetrix); **: P>0.99 by signed rank analysis

Functional categorization of significantly changed transcripts between control E13.5 and stressed E14.5 (Figure 6B), and between control E14.5 and stressed E14.5 (Figure 6C) was conducted to delineate the effects of maternal stress on global gene expression as described in the Materials and Methods. Comparison between the global expression profiles of the limb of stressed E14.5 fetuses and its morphology-matched control revealed that up-regulated transcripts were particularly enriched in transport, signal transduction, and protein metabolism, and down-regulated transcripts in gene expression, protein metabolism, and organelle organization (Figure 6B). Comparison between the transcriptome of stressed E14.5 limb and its gestational age-matched control showed that there were preferential changes in up-regulated transcripts involved in gene expression and large changes in protein metabolism, organelle organization, development, and transport. For down-regulated transcripts, organelle organization was ranked first, and transport, protein metabolism, and development were also highly ranked (Figure 6C).

Twelve differentially expressed genes related to embryonic development (Igf1 and Aldh1a2), cell proliferation (Cryab, Nov, Ctgf, and S100a6), cell death (Apoe and Snai2), muscle differentiation (Acta1 and Actn2), and cell cycle (Smc2 and Nipbl) were selected on the basis of p-values between controls and stressed at E14.5 by signed rank analysis and further evaluated by real-time RT-PCR (Figure 7A, B). Expression profiles of eight genes (Igf1, Ctgf, Nov, Acta1, Actn2, Cryab, S100a6, and Apoe) showed a lack of induction at E14.5 in the limbs of maternally stressed fetuses compared with those in control fetuses. Thus, mRNA expression levels in the stressed limb at E14.5 of these genes appear comparable with those in the limb of morphology-matched control. In contrast, the profiles of two genes (Aldh1a2 and Nipbl) showed different expression profiles of mRNA levels in comparison with both morphology-matched control and gestational age-matched control.

Figure 7.  Expression profiles of development-associated genes. A. Expression patterns of representative genes in selected GO categories (embryonic development and regulation cell proliferation, regulation of apoptosis, muscle differentiation, and cell cycle) are shown as a heat map diagram. B. Developmental mRNA expression profiles of selected genes were determined by real-time RT-PCR at E12.5, 13.5, and 14.5. Data are presented in arbitrary units by normalization with Tbp (n = 4 for each point; **: P < 0.01 and *: P < 0.05 between CTL and STR at the indicated time point). Open circles represent controls and closed circles represent the maternally stressed group.

The expression profiles of Igf1, a growth factor implicated in limb development (Allan et al. 2001), and Aldh1a2, the rate-limiting enzyme in converting retinoic acid from its precursor (Yashiro et al. 2004), were altered by maternal stress, raising possible involvement of these signaling molecules in the transduction of maternal stress. Among the differentially expressed genes, Nov, Ctgf, and Cryab have been reported to regulate the differentiation of mesenchymal tissues such as chondrocytes and myotubes (Kamradt et al. 2002; Luo et al. 2004; Arnott et al. 2007; Minamizato et al. 2007; Song et al. 2007). Acta1 and Actn2, components of mature skeletal muscle, were not induced at E14.5 in the stressed fetus during the limb development. These results indicate that developmental retardation was not limited to the separation of each digital unit, but could be extended to the differentiation of premature limbs into mature forms. In addition to global expression profiles, differential expression profiles of several development-related genes also suggest that the information of maternal stress is encoded in the offspring, presumably by modulating the manifestation of the genetic program in the developmental stage.

Discussion

In the present study, we sought to examine the effects of maternal stress on fetal development. Pregnant mice were subjected to 6 h of restraint stress every day, and the development of fetuses in the womb was analyzed. Sizes and body weights of the stressed fetuses were markedly decreased compared to those of the controls at all time points examined. Interestingly, the development of the somites and limbs were markedly retarded by maternal stress. In particular, delayed expression of Fgf8 in the limbs provides insight concerning the molecular effects of maternal stress on fetal development.

As shown in the cases of the proximodistal outgrowth of the limb bud (Figure 4) and digital morphogenesis (Figure 5), maternal stress appears to retard key molecular events underlying the milestones of fetal limb development. Although it is usually assumed that ordered and punctual molecular signaling cascades underlie the development of an organism, it is surprising that the onset of Fgf8 expression, a key signaling molecule in the limb development, was delayed by approximately half a day. The importance of Fgf8 in limb development has been established in a number of studies (reviewed in Mackem 2006). Fgf8 starts to be expressed in the intermediate mesoderm just before the occurrence of limb bud outgrowth (Crossley et al. 1996). Exogenous Fgf8 application supported the growth of limb buds and regulated the expression of other signaling molecules, including Shh (Vogel et al. 1996). More importantly, limb-specific conditional knock-out of Fgf8 showed deficits in limb bud development, although an abnormal limb was formed because of compensation by other members of the FGF family, including Fgf4 (Lewandoski et al. 2000; Moon and Capecchi 2000), indicating essential roles of Fgf8 in the limb development. Indeed, reduced distal outgrowth of the limb bud was accompanied by delayed expression of Fgf8 in both forelimbs and hindlimbs (Figure 4B). Interestingly, Fgf8 is also expressed in other regions such as the posterior ectoderm in the last primitive streak stage, the brachial arches, the ventral diencephalon, and the isthmic organizer and plays important roles in the development of the somites, the middle ear, the pituitary, the hindbrain, and the midbrain, including dopaminergic neurons (Rosenfeld et al. 2000; Mallo 2001; Roussa and Krieglstein 2004; Duester 2007); these imply that the molecular retardation may play wide roles in alterations in developmental programing by maternal stress.

Although changes in global gene expression patterns cannot distinguish the causal and resultant genes in maternal stress-evoked developmental attenuation, several altered genes suggest plausible explanations for the molecular basis underlying the retardation of limb development in the maternally stressed fetus. Among the differentially expressed genes, it should be noted that the increment in Igf1 mRNA expression in the fetal limb during normal development was apparently suppressed by maternal stress (Figure 7). Igf1 encodes a secreted polypeptide, which is implicated in prenatal and postnatal growth and development. The biological actions of IGF-1 are primarily mediated via its receptors (Igf1rs) and modulated by binding proteins (Igfbps), constituting the insulin-like growth factor (IGF) system. The role of the IGF system in limb development has been suggested based on expression patterns and mutation studies (Allan et al. 2001). Our cDNA microarray data revealed that in contrast to significant decrease in Igf1 expression, there was no significant change in the expression levels of other components of the IGF system, such as Igf1r or Igfbps (data not shown). Considering the growth and differentiation-promoting roles of Igf1, attenuated weight gain, and limb development may be attributable to the decreased Igf1 mRNA levels resulting from maternal stress.

Another differentially expressed gene, Aldh1a2, encodes retinaldehyde dehydrogenase, the rate-limiting enzyme in the synthesis of retinoic acid from its precursor. Retinoic acids are additional important signaling molecules in the limb development, playing a variety of roles, including limb bud initiation, proximodistal outgrowth, apoptosis of interdigital tissue, and chondrogenesis. In control fetuses, the abundance of Aldh1a2 mRNA increased from E12.5 to E13.5 and then decreased by E14.5. Interestingly, mRNA levels of Aldh1a2 were lower in stressed limbs at E12.5 and 13.5 than those in controls, whereas they were significantly higher in the stressed limb at E14.5 due to a slighter decline (Figure 7B). Retinoic acid is known to induce apoptosis and to inhibit chondrogenesis in mesenchymal cells, as exemplified in the Cyp26b1 knockout mouse; the Cyp26b1 gene encodes an enzyme that degrades retinoic acid (Yashiro et al. 2004). Indeed, apoptotic cell death appeared to be delayed in the limbs of stressed fetuses compared with that in controls, (Figure 5). Furthermore, reduced chondrogenesis might be also implicated by the mRNA profiles of Nov and Ctgf (Figure 7B), which are involved in the regulation of differentiation in mesenchymal cells (Luo et al. 2004; Arnott et al. 2007; Minamizato et al. 2007; Song et al. 2007). Considering the persistent manner of maternal stress, it seems less plausible that Aldh1a2 may serve as a direct molecular target for maternal stress due to the delayed mRNA level profile of Aldh1a2. Rather, the expression pattern of Aldh1a2 seems to be well consistent with both delayed and catch-up features of limb development observed in the stressed fetuses.

The present study showed the profound impact of maternal stress on fetal development and related molecular markers in this process, but it remains unresolved how maternal stress is transmitted to the fetus. Because glucocorticoids (GCs) such as CS are pivotal stress hormones mediating the diverse effects of stress, intensive studies have focused on elevated GC levels in the circulation as a mediator of the programing effects of maternal stress (Weinstock 2005). Experimental evidence shows that exogenous administration of the synthetic GC dexamethasone to pregnant females reduces birth weights (Seckl 2004). However, abundant expression of 11β-HSD2 in the placenta complicates the contribution of endogenous GC on fetal development in stress. On the other hand, transport by or the metabolic capacity of the placenta can be influenced by maternal stress, because stress can modulate the vascular tone by humoral and neuronal factors. Although the transmission mechanism of maternal stress into the womb remains unclear, the molecular markers identified in the present study provide valuable information on potential molecular links mediating the programing effects of maternal stress.

Additionally, it remains unaddressed how the environmental information from early life is stored, maintained, and decoded in the adulthood of offspring. For persistent and broad effects on the expression of genetic information, epigenetic mechanisms have been suggested as a language of maternal programing (Waterland and Michels 2007). For example, methylation of the GC receptor (GR) promoter has been shown to be modified by early life experiences, thus altering HPA axis reactivity (Weaver et al. 2004; McGowan et al. 2009). To assess the possible roles of epigenetic modifications in the programing by maternal stress, methylation states of differentially expressed gene promoters were examined by real-time PCR after digestion with a methylation-specific restriction enzyme. However, there were insignificant changes in the levels of global methylation and the promoter methylation of the several tested genes (Figure S1), although these results do not exclude epigenetic mechanisms in maternal programing. It is still possible that the modification of chromosomal structure may occur at other developmental stages. To further evaluate the roles of epigenetic modification in the programing of maternal stress, a more systematic approach appears to be required.

In summary, the present study demonstrates that maternal stress causes the retardation of fetal development and reduces body weight. Focused analyses of limb development show that delays and/or alterations in molecular events were accompanied by morphological attenuation throughout limb formation. In this context, we provide evidence of a broad distortion of global gene expression profiles in the limb of stressed fetuses and a list of differentially expressed genes, supporting the notion that maternal stress can disturb the manifestation of the genetic blueprint, eliciting developmental plasticity in the fetus. Although the quality of life in adulthood can be modulated by postnatal experiences, our results suggest that molecular encoding processes of maternal programing onto the offspring of stressed dams can impact fetal development in the uterus.

Acknowledgements

This work was supported by the Korea Ministry of Education, Science, and Technology (MEST) through the Brain Research Center of the 21st Century Frontier Research Program, and by the National Research Foundation of Korea (NRF; 2010-0020566). H. K. Choe and S. Chung were supported by Brain Korea 21 Research Fellowships from the Korea MEST.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.