Background

Preterm birth (delivery before 37 weeks gestation) occurs in ~ 11% of all pregnancies worldwide and is the leading cause of mortality and morbidity for children under 5 years old [1]. When accurately targeted to women at risk of preterm delivery, antenatal corticosteroid therapy (ACS) has been credited with reductions in the risk of perinatal death and respiratory disease [2,3,4]. The largest benefits are seen in preterm infants born at or below 32 weeks’ gestation. Thereafter, treatment effect magnitude tapers with proximity to term, reflected by large increases in the number of patients needed to be treated to prevent a single case of respiratory disease. However, despite over 50 years of clinical use, ACS therapy is still not optimized, and there is no international consensus regarding the ideal drug formulation, dosing regimen, or route of administration.

Three corticosteroids (betamethasone phosphate, dexamethasone phosphate, and betamethasone acetate) comprise most ACS dosing regimens worldwide [5,6,7,8]; these steroids readily cross the placenta and exert extensive genomic and non-genomic effects on the developing fetus [9]. The two most widely studied ACS formulations, dexamethasone phosphate and betamethasone phosphate (used alone or in combination with betamethasone acetate), have significantly different pharmacokinetic profiles [10]. Unfortunately, the risk-benefits for the developing fetus remain poorly understood which is critical as antenatal steroid use has expanded over the past 20 years to include many patients at considerably lower risks than for which the therapy was originally designed [11]. Furthermore, despite receiving ACS therapy, many babies do not demonstrate any respiratory benefit from treatment (non-response) [12, 13]. Increased use persists, despite evidence from population-level cohort studies concluding an association between exposure to ACS and an elevated risk of childhood neurological deficits/behavioral disorders, serious childhood infection [14, 15], and impaired academic performance [16]. Moreover, additional reports have described ACS exposure being associated with an increased risk of neonatal hypoglycaemia [17], HPA axis dysregulation [18,19,20,21,22], reduced birth weight, and head circumference [23, 24]. Each of which is independently linked to diminished health outcomes.

The impact of ACS on the fetal brain is a topic of particular importance. Due to the high density of glucocorticoid receptor expression, the hippocampus could be highly susceptible to off-target, detrimental ACS effects [25,26,27]. These concerns are supported by evidence from clinical studies in preterm neonates showing that a standard full-dose ACS exposure was associated with reduced hippocampal neuron density [28] and bilateral cortex thinning (maximally in areas of the brain related to emotional regulation) in pre-adolescents [29]. Animal studies have also shown that ACS exposure increased neuronal degeneration [25], reduced brain weight [30], delayed central nervous system myelination [31], and suppressed normal neurodevelopmental pathway differentiation within the hippocampus [32, 33].

Based on the concept that current dosing regimens are probably excessively high to achieve the expected lung maturation and could unnecessarily increase the risk of harm, several recent and ongoing efforts to optimize ACS dosing have been proposed [34,35,36]. However, no study has attempted to evaluate the effects of a high dose regimen in current clinical use (2 maternal intramuscular injections of 12 mg dexamethasone phosphate spaced by 12 h, used clinically in Singapore) vs. lower dose pulsatile regimens of ACS on both fetal lung maturation and brain development.

Study aim

We hypothesized that injury-associated changes in the fetal hippocampal transcriptome would be independent of preterm lung function subsequent to antenatal steroid exposure. We test this hypothesis using the sheep model of pregnancy.

Methods

Animal model and antenatal corticosteroid treatment

Merino ewes with singleton fetuses received an intramuscular injection (IM) of 150 mg medroxyprogesterone acetate (Depo-Provera, Pfizer, New York, NY) on 117 ± 2 days gestational age (dGA) to decrease the risk of steroid-induced premature labor. Baseline maternal blood samples were collected for analysis at 0800, 24–48 h prior to allocation to treatment group. Animals were then randomly allocated to one of four groups (Fig. 1), with first treatment given at 122 ± 2 dGA (term = 150 dGA). Lung maturation in the preterm fetal sheep at 124 days’ gestation approximates that of the human at 30–32 weeks’ gestation. First injections occurred at 08:00 to control for circadian effects: (i) Saline Control Group, 4 × 2 ml maternal saline intramuscular (IM) injections at 12-h intervals (n = 11); or (ii) Dex High Group, 2 × 12 mg maternal IM dexamethasone phosphate (DBL Dexamethasone sodium phosphate 4 mg/ml, Hospira NZ, New Zealand) injections at 12-h intervals followed by 2 × 2 ml IM saline injections at 12-h intervals (n = 12; this protocol simulates the standard ACS regimen used clinically in Singapore); or (iii) Dex Low Group, 4 × 1.5 mg maternal IM dexamethasone phosphate injections 12-h intervals (n = 12; this protocol represents a 75% reduction on the Singapore clinical regimen); or (iv) Beta-Acetate Group, 1 × 0.125 mg/kg maternal IM betamethasone acetate injection followed by 3 × 2 ml IM sterile normal saline injections at 12-h intervals (n = 8; this protocol represents a positive control group previously shown to elicit robust preterm lung maturation in the preterm sheep equivalent to standard courses of dexamethasone phosphate or betamethasone phosphate and acetate used worldwide).

Serial maternal plasma samples were collected at 6, 12, 18, 24, and 36 h after first maternal injection. All preterm lambs included in this study for ventilation analysis (122–125 days gestation) were surgically delivered 48 h after first maternal injection at 122–125 dGA.

One ewe from the Saline Control group was removed from the analysis as it was found not to be pregnant at pre-delivery ultrasound. One lamb from the Beta-Acetate Group, which was incorrectly intubated and not ventilated for 15 min, was removed from the ventilation analysis.

Pharmacokinetics

Liquid chromatography-mass spectrometry (LC–MS) was performed on all maternal and fetal plasma samples to assess dexamethasone and betamethasone-acetate concentrations respectively as previously reported [37]. Betamethasone Acetate (Beta-Acetate Group) pharmacokinetic parameters were estimated by non-compartmental analysis using Phoenix WinNonlin software (version 8.1, Certara, Inc.). Dexamethasone (Dex Low and Dex High Groups) simulated pharmacokinetic parameters were generated by Dr Mark Milad based on existing ovine dexamethasone pharmacokinetic data [8].

Delivery and ventilatory assessment

Maternal plasma samples for analysis were taken immediately prior to delivery, i.e., 48 h after first maternal injection. Fetal plasma samples for analysis were obtained during operative delivery and immediately prior to ventilation commencing. Prior to delivery, ewes received an intravenous bolus of midazolam (0.5 mg/kg) and ketamine (10 mg/kg) for the deep induction of anaesthesia. The fetal head was delivered through abdominal and uterine incisions, and a 4.5-mm endotracheal tube was secured by tracheostomy. The fetus was then delivered, and the ewe euthanized under anaesthesia with pentobarbital. Lambs were weighed, dried, and placed in a radiant warmer (Cozy Cot, Fisher & Paykel Healthcare, New Zealand) bed under a plastic insulating wrap (Neowrap, Fisher & Paykel, NZ). Mechanical ventilation (Fabian HFO, Accutronic Medical Systems AG, Switzerland) was immediately commenced with the following settings: peak inspiratory pressure (PIP) of 35 cmH2O, positive end expiratory pressure (PEEP) of 5 cmH2O, respiratory rate (RR) of 50 breaths per minute, inspiratory time (iT) of 0.5 s, using 100% heated and humidified oxygen. We used 100% oxygen which allows comparison of oxygenation through the partial arterial pressure of oxygen among the groups.

Investigators operating the ventilators were blinded to the animal treatment groups. An umbilical artery was catheterized for blood sampling and administration of supplemental anaesthesia with ketamine (5 mg/kg) as necessary. Tidal volume (VT) was continuously measured, and the PIP was adjusted to keep the VT between 7.0 and 8.0 ml/kg, but with PIP limited to 35 cmH2O. At 10, 20, and 30 min of ventilation we measured temperature, blood pressure, ventilator data (PIP, VT, and compliance), and performed blood gas measurements; pH, PCO2 (mmHg), PO2 (mmHg), O2 saturation (SO2, %), and total hemoglobin (Hb, g/dL) levels (Siemens RapidPoint 500, Munich, Germany). Dynamic compliance (Cdyn, ml/cmH2O/kg) was recorded as measured by the ventilator. The ventilation efficiency index (VEI) was calculated using the formula VEI = 3800 / (RR × (PIP − PEEP) × PCO2 (mm Hg)) [38].

Definition of antenatal corticosteroid response

Steroid-treated lambs were defined as having responded to ACS treatment if the PaCO2 level after 30 min of ventilation was two standard deviations below the mean of the Saline group. The mean and standard deviation of the PaCO2 after 30 min of ventilation in the Saline Control group was 118.3 and 31.3 mmHg thus giving rise to 57.0 mmHg (= 118.3 − 2 × 31.3) as cutoff.

Lung functional assessments

After 30 min of ventilation, lambs were euthanized with pentobarbital, disconnected from the ventilator, and the endotracheal tube was clamped for 2 min to achieve atelectasis by oxygen absorption. The lambs were weighed, and the chest was opened for visual evaluation of gross lung injury–pulmonary hemorrhage, pulmonary interstitial emphysema, gas pockets within the lung, or subpleural dissection. These were all performed by a single investigator. A deflation pressure–volume curve was measured after air inflation of the lungs to a pressure of 40 cmH2O. Volume at a pressure of 40 cmH2O was calculated using fetal weight after ventilation (kg) as V40 (ml/kg).

Hippocampal and lung mRNA extraction and RNA bulk sequencing

Twenty-four animals (Six animals from each group -Saline Control, Dex High, Dex Low, Beta-Acetate) were selected for RNA bulk sequencing based on the highest maternal steroid plasma concentration prior to delivery. The same animals were used for both lung and hippocampal RNA extraction and analysis thereby allowing for matched analysis with ventilation outcomes and ACS responsiveness.

The ACS Responder Group for RNA bulk sequencing (total six animals) consisted of three animals from the Dex Low and three animals from Beta-Acetate Groups; ACS non-responder subgroup (total twelve animals) consisted of six animals from the Dex High Group, three animals from the Dex Low Group, and three animals from Beta-Acetate Group. RNA was extracted from snap-frozen hippocampal (entire hippocampus) and lung tissue (inferior aspect of right lower lobe) using RNeasy Plus Mini Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s instructions. The concentrations of extracted RNA and RNA Integrity Number (RIN) were determined using Agilent Technologies RNA Nano Chip as per manufacturer’s instructions. All RNA extracts were diluted in nuclease-free water to yield a final RNA concentration of 30 ng/μl. Hippocampal and lung tissue bulk RNA sequencing was performed by Novogene Singapore.

Bulk RNA-sequencing data processing and quality control

One hundred fifty-base pair paired-end sequenced reads were processed using the nf-core/rnaseq v3.10.1 pipeline [39] with nextflow v22.10.4 [40]. Briefly, raw reads were trimmed using Trim Galore! v0.6.7 [41] to remove low-quality bases and adapters. Trimmed reads were aligned to the Ovis aries reference genome, ARS-UI_Ramb_v3.0 with STAR v2.6.1d aligner [42]. Finally, Salmon v1.9.0 [43] was utilized for assigning reads. Principal component analysis was conducted for quality control assessment and exploratory analysis.

Differential gene expression and gene set enrichment analysis

All downstream analyses were performed using R statistical software version 4.2.1 [44]. Differential gene expression analysis was performed using the DESeq2 package v1.36.0 [45]. Genes with at least 1.5-fold change in expression and FDR < 0.05 were considered significant. The results were visualized using the Enhanced Volcano v1.14.0 [46] and Complex Heatmap v2.12.1 [47]. Gene set enrichment analysis (GSEA) was run on all genes ranked by their signed log p-value using cluster Profiler v4.9.0 [48] and KEGG database [49].

Hematological and biochemical data acquisition

Plasma isolated from maternal and fetal blood samples was used for hematological analyses, including white blood cell counts (WBC,/μl), differential leukocyte counts (%), neutrophil count (%), biochemical parameters; gamma-glutamyl transferase (GGT, U/l), alanine aminotransferase (ALT, U/l), aspartate aminotransferase (AST, U/l), albumin (g/l), cortisol (nmol/L), adrenocorticotropic hormone (ACTH, pg/ml). Those analyses were performed by an independent clinical pathology laboratory (Vetpath, Perth, Australia). Cortisol values under limit of detection (< 5.52 nmol/l) were allocated a value of 5.52 nmol/l for statistical analyses. ACTH values under limit of detection (< 5.0 pg/ml) were allocated a value of 5 pg/ml for statistical analyses.

Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics for Windows, version 25.0 (IBM Corp, Armonk, NY). Chi-square was used to test the differences of nominal values between groups. All numerical data were tested for normality with Shapiro–Wilk tests. In the comparison of the four intervention groups (Dex High, Dex Low, Beta-Acetate, Saline Control Groups) between-group differences in parametric data were tested for significance with one-way ANOVA, while Kruskal Wallis tests were used for non-parametric data. Multiple post hoc comparisons were performed with Tukey’s and Dunnett t (2-sided) tests. p values < 0.05 being significant. Repeated ANOVA test and pairwise comparisons were used to analyze maternal Saline Control Group baseline cortisol and ACTH compared to serial time points.

Results

Reduced dose antenatal steroid regimens accelerate fetal lung maturation

There were no significant differences in delivery data among treatment groups (Additional File 1 and Additional Table 1).

Simulated neonatal steroid plasma concentrations for each steroid-treated group highlight the pulsatile nature of the dexamethasone regimens compared to a constant exposure Beta-Acetate group (Fig. 2A, B with pharmacokinetic profile illustrated in Additional Table 2) [8]. Maternal and cord (delivery) steroid plasma measurements and pharmacokinetic parameters are described in Additional File 1.

Fig. 1
figure 1

Experimental design. Diagram of experimental design and intervention groups

Full size image

Arterial umbilical cord gas PaCO2 values serve as a robust and well-established measure of acute preterm functional lung maturation [50,51,52,53]. The Beta-Acetate Group animals had significantly improved cord PaCO2 values compared to Saline Control Group animals at 10, 20, and 30 min of ventilation (p < 0.05) (Fig. 2C (Additional File 1 )). The Dex Low Group animals had significantly improved cord PaCO2 values compared to Saline Control Group animals at 10 and 20 min (p < 0.05) of ventilation, but not significant at 30 min (p = 0.094/NS) (Fig. 2C). There were no statistically significant differences in cord PaCO2 between Dex High Group animals and Saline Control Group at any ventilation timepoint. Additional lung function data are shown in Additional Figs. 1A-C,2. Our data show, for the first time, that the high-dose, pulsatile, and rapidly cleared dexamethasone ACS regimen (Dex High Group) currently used clinically in Singapore did not convey statistically significant improvements in preterm ovine neonatal lung function. The greatest improvements in preterm neonatal lung function were actually associated with constant, low-concentration and extended-duration fetal steroid exposures represented by the Positive Control (Beta-Acetate) Group.

The effects of different antenatal steroid regimens on fetal lung transcriptome related to lung maturation

Principal component analysis (PCA) plots of the fetal lung transcriptome profile showed that ACS-treated animals were distinct from Saline Control animals (Fig. 3A). Differential analysis identified 1533 significant genes (with at least 1.5-fold change and FDR < 5%) in the steroid-treated groups (Fig. 3B; Supplementary GSEA Table 1). Of these significant differentially expressed genes (DEGs), 567 (37%) were upregulated while 967 (63%) were downregulated (Fig. 3B). There was a marked difference between upregulated and downregulated genes. Upregulated genes common to all steroid-treated groups related to inflammation and immune regulation (IL1B, MARCO, C7, CXCL16, FGL1, TNFRSF13B). In comparison, downregulated genes common to all steroid-treated groups were related to cellular division/ cell cycle (PIMREG, PRR11, FOXM1, PRC1, KIF18B, ASPM, MKI67, CDCA3, CDCA8, CDC20) (Fig. 4B).

Fig. 2
figure 2

Simulated fetal plasma ACS concentrations and arterial PCO2 at 10, 20, and 30 min of ventilation. A Simulated fetal plasma steroid concentrations comparing Dex High, Dex Low, and Beta-Acetate dosing regimens based on maternal plasma concentration and known pharmacokinetics of dexamethasone and betamethasone (Additional Table 2) dosing regimens [8]. B Simulated fetal plasma steroid concentrations comparing Dex Low and Beta-Acetate dosing regimens based on maternal plasma concentration and known pharmacokinetics of dexamethasone and betamethasone dosing regimens (Additional Table 2) [8]. C Umbilical PaCO2 at 10, 20, and 30 min of ventilation (Saline control, Dex High, Dex Low, Beta-Acetate groups). D ACS Responder, ACS Non-responder and Saline Control group PaCO2 at 30 min of ventilation. The Asterix indicates a significant difference among the groups. Error bars represent ± 1 standard deviation

Full size image
Table 1 Summary of the clinically relevant functional lung maturation status, differential gene and KEGG pathway expression for each treatment group compared to Saline Control group
Full size table
Fig. 3
figure 3

Fetal lung RNA bulk sequencing for treatment groups. A Principal component analysis (PCA) plot of differential gene expression for intervention groups Dex High / Dex Low / Beta-Acetate/ Saline Control Groups. B Venn diagram demonstrating differential up and down gene regulation between Dex High/ Dex Low/ Beta Acetate group vs Saline Control group. C Heat map displaying differential gene expression for twenty different genes related to fetal lung maturation [54] for each intervention group (Saline Control, Dex High, Dex Low, Beta-Acetate). D Heatmap demonstrating the significant positively and negatively enriched KEGG Pathways (DEPs) (Fold change > 1.5, False Discovery Rate FDR < 5%) between steroid treated groups (Dex High, Dex Low, Beta-Acetate) compared to Saline Control Group

Full size image

Animals in the Positive Control Beta-Acetate Group (demonstrating the highest degree of lung maturation) had the greatest number of group-specific significant DEGs (708 genes), compared to Dex Low (170 genes) and Dex High (58 genes) (Fig. 3B). Dex Low and Beta-Acetate Group animals, which had the longest predicted duration of fetal ACS-exposure and functional lung maturation, had upregulation of established markers of lung maturation including surfactant production and transport (SFTPAI, SFTPB, SFTPC, CTSH, LPCAT1, LPCAT3, SCNN1A [54]) genes related to immune response/inflammation (LAMP3, C5), and increased vasculature (HIF3A, EPAS1) (Figs. 3C and 4B). Dex Low and Beta-Acetate Group animals also had common downregulation of genes related to extracellular matrix and cell adhesion (MFAP5, CDH12, NFASC, FAP) (Fig. 4B). Consistent with functional studies, these markers of lung maturation were not significantly differentially expressed in Dex High Group animals (Figs. 3C and 4B).

To understand the biological themes associated with the DEGs, we conducted gene set enrichment analysis (GSEA) [48] with KEGG pathways [49] on all genes ranked by the signed negative log p-value. Steroid-treated groups were compared to Saline Control Group (Fig. 3D, Supplementary GSEA Table 2). Overall, 93 (82.8%) KEGG pathways were positively enriched while 16 (17.2%) KEGG pathways were negatively enriched. We identified 22 positively enriched KEGG pathways common to all steroid-treated groups that are related to cellular metabolism, inflammation, and human disease (Fig. 3D). Six negatively enriched pathways were common to all steroid-treated groups related to cellular homeostasis and reproduction (Fig. 3D).

Understanding the transcriptomic fetal lung maturation response to ACS exposures

ACS treated animals were defined and grouped as ACS Responder if the PaCO2 after 30 min of ventilation was more extreme than two standard deviations below the mean PaCO2 value for the Saline Control Group [12]. ACS response rates observed between steroid-treated groups were Dex High Group 8.33% (1/12), Dex Low Group 25% (3/12), and Beta-Acetate Group 42.86% (3/7) (Fig. 2D). When compared to Saline Control Group animals, the ACS response rate was only significantly higher in the Beta-Acetate Group animals (p = 0.048).

PCA plots revealed distinct clusters of Saline Control, ACS Responder, and ACS Non-Responder Groups (Fig. 4A). Comparative analysis comparing DEGs of ACS Responders to Non-Responders showed 101 significant DEGs (Supplementary GSEA Table 3). Responder Group animals had significant upregulation of genes known to be associated with fetal lung maturation and surfactant production (SFTPA1, SFTPC, and SFTPB), inflammation and immune response (CXCL8, CXCL17, C5, PLAUR), and downregulation of genes related to extracellular matrix remodelling (COL12A1, NOG) (Fig. 4B). Analysis of differential KEGG pathway expression between ACS Responders and Non-Responders are presented in Fig. 4C (Supplementary GSEA Table 4). Responder Group animals had positive enrichment of 50 KEGG pathways (related to inflammation, lipid, and sugar metabolism) and negative enrichment of nine KEGG pathways (related to cell and DNA replication).

Antenatal steroid exposure causes neurodegenerative-associated changes to the fetal hippocampal transcriptome

PCA plots for all treatment groups as well as ACS Responder and Non-Responder Groups are shown in Fig. 5A. When compared to Saline Control Group, differential analysis identified 879 significant DEGs (at least 1.5-fold change and FDR < 5%) in the steroid-treated groups (Fig. 5B, Supplementary GSEA Table 5). Common to all steroid-treated groups were 41 upregulated DEGs and 26 downregulated DEGs (Fig. 5B, Additional Fig. 3). The Positive Control Beta-Acetate Group animals had the greatest number of group specific, significant DEGs (371 genes) compared to Dex Low (169 genes) and Dex High (67 genes) (Fig. 5B).

Fig. 4
figure 4

Fetal lung RNA bulk sequencing for ACS responder vs non-responder groups. A Principal component analysis (PCA) plot of differential gene expression for ACS Responder, ACS Non-Responder and Saline Control groups. B Heat map displaying differential gene expression for genes related to fetal lung maturation, angiogenesis, inflammation, extra-cellular matrix, and cell cycle for ACS Responder vs ACS Non-Responder groups and by specific treatment group (Dex High, Dex Low, Beta-Acetate) compared to Saline Control. C Wrap plot demonstrating the significant positively and negatively enriched KEGG Pathways (DEPs) (Fold change > 1.5, False Discovery Rate FDR < 5%) for ACS Responder vs ACS Non-responder

Full size image

Analyses were conducted for genes related to neurodevelopment and neurodegeneration. Upregulated genes common to all steroid-treated groups included NMB, CCN3, APOD, and NPPC, while downregulated genes common to all steroid groups included ISLR2 and NTNG2 (Fig. 5C,D). Dex High and Beta Acetate Group animals had common upregulation of AZIN2 and C4A. Dex Low and Positive Control Group animals had common upregulation of DRD5 and downregulation of SVB2. Finally, Positive Control Group animals had downregulation of GLDN, which was not seen in the other steroid-treated groups.

Steroid-treated groups were compared to Saline Control Group animals in terms of differential KEGG pathway expression. Overall, the majority (75.8%) of KEGG pathways in the steroid-treated groups were negatively enriched (Fig. 6A, Additional File 1 and Supplementary GSEA Table 6). We identified two positively enriched (“Non-Alcoholic Fatty Liver Disease” & “Ribosome”) and 16 negatively enriched DEPs (pathways related to cellular homeostasis & infection) common to all steroid-exposed groups (Fig. 6B).

Fig. 5
figure 5

Hippocampal bulk RNA sequencing data for differentially expressed genes. A Principal component analysis (PCA) plot of differential gene expression for intervention groups Dex High / Dex Low / Beta Acetate/ Saline Control groups and ACS Responder / ACS Non-Responder subgroups. B Venn diagram demonstrating differential up and down gene regulation (DEG) between Dex High/ Dex Low/ Beta Acetate group vs Saline Control group. C Volcano plots of differential gene expression (DEG) for each treatment group Dex High, Dex Low, and Beta Acetate compared to Saline Control group. D Heat map displaying significant differential gene expression of eight different genes related to neurodegenerative/neuropsychiatric disorders for ACS Responder vs ACS Non-Responder and by specific treatment group (Dex High, Dex Low, Beta-Acetate) compared to Saline Control

Full size image

Pulsatile dexamethasone-only exposed groups (Dex High and Dex Low) had three common positively enriched DEPs related to neurodegeneration (“Prion Disease”, “Alzheimer’s Disease”, “Arachidonic Acid metabolism”) and two common negatively enriched DEPs (“Human T-cell leukemia virus 1 infection”, “Cellular Senescence”) (Fig. 6B). Furthermore, in the pulsatile, high amplitude Dex High Group, we identified unique positive enrichment of the KEGG pathway “Spinocerebellar Ataxia” (Fig. 6B) and group specific positive enrichment in the low amplitude, pulsatile Dex Low Group of “Parkinson Disease,” “Huntington disease,” “Pathways of neurodegeneration – multiple diseases,” and “Amyotrophic lateral sclerosis” (Fig. 6C).

Comparisons between Dex Low and Positive Control (Beta-Acetate) Group animals had five common positively enriched DEPs (“Oxidative Phosphorylation” and Chemical carcinogenesis – ROS”) and 11 common negatively enriched DEPs (including “Wnt Signaling”) (Fig. 6B). Dex High and Beta-Acetate Group comparison had no common positively enriched and only four common negatively enriched DEPs (“Motor proteins,” “DNA replication,” “African Trypanosomiasis,” “cell cycle”) (Fig. 6B). The lack of commonality between these groups may reflect the significant differences in pharmacokinetic profiles between these groups (high amplitude pulsatile vs. low amplitude constant exposure) respectively (Fig. 2A). Surprisingly, the constant, low amplitude exposure Beta-Acetate Group had only one additional unique KEGG pathway (“Phagosome”), despite having the greatest number of significant DEGs (Fig. 6B).

Neurodegenerative changes in the fetal hippocampal transcriptome relate to steroid exposure and not ventilation outcome

Analysis of ACS Responder compared to ACS Non-Responder Groups for significant DEGs surprisingly revealed limited differential expression, none of which related specifically to neurodegeneration (one uncharacterized upregulated gene and 37 downregulated genes) (Supplementary GSEA Table 7). By comparison, both steroid-exposed groups (Responder and Non-Responder Groups) when compared to saline control respectively, demonstrated the same pattern of upregulation in the genes related to neurogenerative and neuropsychiatric disorders being NMB, CCN3, APOD, NPPC, AZIN2, DRD5 and downregulation of NTNG2 (Fig. 5D, Supplementary GSEA Table 8). Notably, these genes were not significantly differentially expressed when ACS Responders were compared to Non-Responder Groups.

Finally, KEGG pathway analysis of ACS Responders vs. Non-Responder subgroups did not demonstrate positive or negative enrichment for any pathways related to neurodegenerative disorders (Supplementary GSEA Table 9 and Supplementary GSEA Table 10). These findings therefore suggest that the observed adverse effects of ACS exposure on neurodevelopment are related to ACS drug exposure and not impacted by lung maturation status and subsequent ventilation outcomes.

Cardiac, endocrine, immunological, and hepatic effects

Data describing the effects of different steroid dosing regimens on key homeostatic processes are shown in Additional Figs. 4–6. No outcomes likely to impact the lung maturation or transcriptomic data presented herein were identified.

Discussion

Principal findings

The principal findings of this study are that:

  1. (i)

    An ACS dosing regimen used clinically in Singapore, conveying a high amplitude, pulsatile, and rapidly cleared dexamethasone ACS regimen (represented by the Dex High Group) did not demonstrate statistically significant improvements in fetal lung maturation. Furthermore, consistent with earlier reports [50], a single dose of 0.125 mg/kg Beta-Acetate induced robust fetal lung maturation based on significantly improved PaCO2 (Fig. 2C), dynamic lung compliance, VEI, and lung tidal volumes (Additional Fig. 1). The comparably poor maturation effect observed in the high-dose dexamethasone group is consistent with our in silico modelling and in vivo measurements of materno-fetal dexamethasone exposures following two 12 mg maternal intramuscular injections given at a 12-h interval [8, 11]. In this regimen, dexamethasone is rapidly cleared from the fetal compartment, with fetal plasma concentrations likely at sub-therapeutic levels within 24 h of treatment commencing. These data are consistent with our previous findings, which showed that robust preterm lung maturation in the sheep requires an uninterrupted exposure of 1–2 ng/ml plasma betamethasone for at least ~ 30 h when delivering animals at a 48 h treatment to delivery interval [53, 55, 56]. The most robust maturation was seen in animals that had this level of steroid exposure at the time of delivery and ventilation [56]. These data further support optimizing of ACS therapy by utilizing a reduced dose, extended-duration regimens;

  2. (ii)

    Significant patterns of gene expression corresponding with fetal lung maturation were only observed in animals from the reduced dose, and extended duration ACS regimens (Dex Low and Beta-Acetate Groups). Examples of this pattern of gene expression include the upregulation of genes related to surfactant production/function (SFTPAI, SFTPB, SFTPC, CTSH, LPCAT1, LPCAT3, SCNN1A) [57], and increased vasculature (HIF3A, EPAS1) as well as downregulation of genes related to extracellular remodelling (MFAP5, CDH12, NFASC, FAP). This pattern of transcriptome changes relating to lung maturation was not observed in the Dex High group, which was also demonstrated at a physiological level, with poor ventilation outcomes and therefore poor functional lung maturation. These data are especially relevant since the Dex High regimen is recommended for clinical use in patients at risk of preterm birth in many countries.

  3. (iii)

    ACS-exposed animals had significantly dysregulated hippocampal RNA transcriptomes compared to Saline Control Group animals, suggesting the potential for altered neurodevelopment irrespective of the dose or pharmacokinetic profile of the ACS regimen employed. Despite being exposed to hypercapnia, hypoxia, and more pronounced acidosis, animals in the Saline Control Group from a neurodevelopmental injury perspective, demonstrated a far more benign transcriptome, compared to both ACS-responding and non-responding animals; and

  4. (iv)

    The most concerning findings were seen in the animals exposed to both high and low-dose, pulsatile dexamethasone regimens (Dex High and Dex Low Groups). They had the greatest degree of significant enrichment of KEGG pathways associated with neurodegenerative disorders. For example, both pulsatile dexamethasone exposed groups (Dex High and Dex Low) exhibited positive enrichment of pathways related specifically to neurodegenerative disorders, including “Prion Disease,” “Alzheimer’s Disease,” and “Arachidonic Acid (AA) Metabolism” (Fig. 6B/C). By comparison, none of these pathways were activated in the constant exposure, low-dose Beta-Acetate Group (Table 1).

Fig. 6
figure 6

Hippocampal bulk RNA sequencing data for differentially expressed KEGG pathways. A Venn diagram demonstrating differential up and down KEGG pathway expression (DEP) between Dex High / Dex Low / Beta Acetate groups vs Saline Control group. B Heat map demonstrating the significantly positively enriched and significantly negatively enriched differentially expressed KEGG Pathways (DEP) (Fold change > 1.5, False Discovery Rate FDR < 5%) in common between all steroid-treated groups (Dex High/ Dex Low/ Beta Acetate), Dex High group specific and Beta Acetate group specific, compared to Saline Control. C Heat map demonstrating the Dex Low group specific 14 significantly positively enriched and 65 significantly negatively enriched differentially expressed KEGG Pathways (DEP) (Fold change > 1.5, False Discovery Rate FDR < 5%) compared to Saline Control group

Full size image

Clinical implications

ACS therapy was introduced into clinical practice progressively over the past 50 years although there was initially much resistance to it. In fact, there was so much concern about failure to use it, that the NICHD held a consensus conference to generate publications to warn obstetricians that failure to use steroids for premature labor would be very problematic medicolegally [58, 59]. Recently there has been clinical concern about the overuse of such treatments and deleterious consequences of such. In current practice, ACS is still very commonly administered to women at risk of preterm delivery. Although some 11% of all babies are born preterm, many more pregnant women are judged to be at risk of preterm delivery and are given ACS therapy. As such, a very large (perhaps as much as 20%) of mothers and fetuses are now exposed to high-dose ACS.

Despite the very common usage, optimization of how and when to use ACS has never been achieved, and the effects of drug pharmacokinetics on the brain, lungs, and other vital fetal systems, remains poorly understood.

Understanding the molecular mechanisms and transcriptomic changes behind the fetal lung maturation responses to ACS, as well as the risk of injury to the developing fetal brain, are critical to deciding the proper usage of such powerful pharmaceuticals [12].

We now have molecular tools capable of answering many questions that were not possible when ACS was introduced. In this study, we have highlighted the critical role of the surfactant genes SFTPA1, SFTPB, and SFTPC in the fetal lung maturation response by demonstrating significant upregulation of these genes in the ACS Responder group compared to the non-responder group and Saline control. The late stages of fetal lung development require mesenchymal thinning to promote gas exchange [60]. Correspondingly, we observed significant downregulation in genes related to extracellular matrix (ECM) remodelling and collagen synthesis in the fetal lungs of the Responder group compared to the non-responder group and Saline control which fits with this established process of mesenchymal thinning. Schmidt et al., utilizing a non-human primate model of lung maturation, also observed downregulation in genes related to ECM remodelling and mesenchymal growth factors [61]. These findings help inform on further targeted approaches to inducing preterm fetal lung maturation to improve the response rates to treatment.

Despite concerning associations observed in large population cohort studies, indicating an elevated risk of neurobehavioral/learning disorders in children exposed to ACS in utero, the neurodevelopmental effects of various ACS exposures on the fetal brain are still not well understood [14, 15]. In this study we have demonstrated that all steroid-treated groups had common upregulation of genes related to dysregulation of neurodevelopment including Neuromedin B (NMB), Cellular Communication Network Factor 3 (CCN3), Apolipoprotein D (APOD), Natriuretic Peptide C (NPPC). NMB is a neuropeptide that has been associated with the acute stress response and anxiety in animals [62], while animal models have demonstrated that elevated CCN3 impaired hippocampal differentiation and normal hippocampal neurogenesis [63]. Similarly, elevated levels of APOD have been found in the CSF and hippocampus of patients with Alzheimer’s disease [64], while elevated NPPC has been shown to negatively affect hippocampal neuroplasticity and memory in rat models [65].

All steroid-treated groups also demonstrated downregulation of the genes Immunoglobulin superfamily containing leucine rich repeat 2 (ISLR2) and Netrin G2 (NTNG2). ILSR2 is a key factor for axon guidance and brain development [66, 67]. Deficiency of ILSR2 is associated with the development of hydrocephalus in mice and clinical studies [66, 67]. Reduced mRNA levels of NTNG2 have also been demonstrated in the brains of patients with bipolar disorder and schizophrenia, thereby implicating NTNG2 in the pathophysiology of these diseases [68].

Further, groupwise comparisons of Dex High and Beta-Acetate Group animals revealed common significant upregulation of the genes Antizyme Inhibitor 2-like (AZIN2) and Complement Component 4A (C4A). C4A is elevated in the CSF and brain tissue of patients with schizophrenia and is associated with increased synaptic pruning [69, 70], while AZIN2 has been found in high levels in the brains of Alzheimer’s patients [71]. Dex Low and Beta-Acetate group animals both had significant upregulation of Dopamine Receptor D5 (DRD5), which has also been described in association with Attention-Deficit/Hyperactivity Disorder (ADHD) [72] and common downregulation of Synaptic Vesicle Glycoprotein 2B (SVB2) which is known to be downregulated in the brains of patients with temporal lobe epilepsy (TLE) [73]. In particular, SVB2 has been shown to be downregulated in the brains of patients with pharmaco-resistant TLE which paralleled increased synaptic loss [74]. Finally, although overall of less concern from a developmental perspective, animals in the Beta Acetate group had unique downregulation of Gliomedin (GLDN) which is an important factor in the development of nodes of Ranvier in the peripheral nervous system [75].

Taken together, these data reveal the impact of ACS treatment on the hippocampal transcriptome, with a material number of differentially regulated genes that are associated with neurodegenerative disorders increasingly identified in large, population-level cohort studies [14,15,16]. Further research is needed to establish a causal relationship of these findings with the epidemiological studies that have shown the development of learning difficulties and neurobehavioral disorders in individuals exposed to ACS [14, 15]. Data from our study and others certainly suggest potential connections which warrant further exploration in long-term clinical trial follow up studies.

Research implications

Understanding the molecular pharmaco-molecular relationships between ACS exposure and brain injury is important to the optimisation of ACS intervention. In addition to differential gene expression in the fetal hippocampus, all steroid exposed groups in the present study had downregulation of the KEGG pathways “Gap junction,” “Focal adhesion,” and “ECM -receptor activation” which are known to be involved in normal neuron migration, gene expression, cell proliferation, differentiation, and survival [76,77,78,79,80]. Further analysis of these negatively enriched pathways revealed common downstream signaling processes including PI3K-AKT, Ras-MAPK-ERK, JAK-STAT, and NFҡB signaling. PI3K-AKT signaling is known to be integral for normal brain development, sympathetic neuron growth, and survival [81] with clinical studies demonstrating that AKT gene deletion syndromes exhibit microcephaly and agenesis of the corpus callosum phenotypes [82]. Furthermore, Ras-MAPK-ERK signaling is an established master regulator of neurodevelopment, important for normal neural tissue proliferation and differentiation. Specifically, ERK signaling is an integral component of learning, memory, and behavior [83], with animal studies demonstrating that inhibition and downregulation of ERK signaling can result in repressed long-term potentiation, memory consolidation, depression-like behavior, and impaired fear memory consolidation [83, 84]. Additionally, JAK/STAT signaling is important in CNS synaptic plasticity, memory, and learning [85], with downregulation in the JAK/STAT pathways (in particular JAK2/STAT3) being implicated in age-related memory decline and Alzheimer’s disease [86]. Finally, NFҡB signaling is also important to normal neurodevelopment, with deficits in neural plasticity and memory formation a consequence of negatively enriched NFҡB functioning [87]. These relationships need further clarification.

In summary, our data demonstrate that all ACS exposed groups, including those with substantial dose reductions, had significant downregulation within the preterm hippocampus of crucial signaling pathways (PI3K-AKT, Ras-MAPK-ERK, JAK-STAT, and NFҡB) for normal neurogenesis and development. These findings may also help explain a causal link between ACS exposure and associated increased risks of learning deficits, mental, and behavioral disorders seen in population cohort studies of ACS exposed children [14,15,16].

Limitations of this study

One key limitation for clinical application at this point is our inability to differentiate between potential differences in effect between betamethasone and dexamethasone. Due to agent availability and the clinical regimen of interest, we assessed rapid-release dexamethasone phosphate (pulsatile exposures at high and low doses) against betamethasone acetate which has achieved a slow-release, constant exposure profile. Although both drugs act via the glucocorticoid receptor, and the projected Cmax for the Beta Acetate and Dex Low Groups were approximately equal, the differences in transcriptional effects seen in this study may stem from differences in the activation profile of each drug. This is not a well-studied area of glucocorticoid pharmacogenomics, and aside from differences in receptor affinity, betamethasone and dexamethasone are viewed as being interchangeable from the perspective of signalling action. Whether or not dexamethasone and betamethasone elicit markedly different transcriptional responses at identical drug exposures is an intriguing question. However, as dexamethasone is the front-line ACS agent for much of the world, our data showing that dexamethasone exposure was associated with an increased risk of adverse transcriptional changes in the fetal brain, may have serious implications for how this drug is used in obstetric practice, as well as the supply of the less readily available betamethasone-based preparations. It is also important to note that our studies relate to the acute transcriptomic effects of ACS on the brain and lung. Although we have identified injury-associated changes in the hippocampus independent of respiratory function, it is not clear if these alterations are maintained into childhood or contribute to a neurological phenotype consistent with that reported in the epidemiological and trials literature. It is also important to note that all animals were delivered at a standardized gestational age, following a fixed exposure interval. As lung and brain responses to glucocorticoid exposure are dynamic processes, it is important to note that the different findings may be observed following delivery at different gestations or treatment to delivery intervals. We also acknowledge that we did not do further analysis on fetal cardiac function beyond heart rate and blood pressure (Additional File 1, Additional Figure 4); therefore, we cannot exclude that undetected adverse cardiovascular instability due to ANS exposure may have impacted other organ systems as well as respiratory and neurological development.

Conclusions

In summary, our findings show that, in the preterm fetal sheep, a therapeutically beneficial (for lung maturation) dose of glucocorticoids delivered via the maternal compartment cannot be accomplished without materially disrupting key homeostatic systems in both mother and fetus.

In this study, we show for the first time that a high-dose regimen currently in clinical use in Singapore, fails to achieve its primary goal of improving functional maturation of the preterm fetal sheep lung, but causes marked, neurodegenerative/psychiatric disease-associated transcriptional changes in the preterm hippocampus, and severely disrupts the materno-fetal HPA axis. We additionally show, also for the first time, that adverse changes in the hippocampal transcriptome were not affected by significant dose reduction in the dexamethasone ACS groups, and that these changes occurred independent of functional lung maturation status. These indicate that administering ACS at a dose necessary for functional maturation of the preterm lung may not be achievable without an elevated risk of adverse impacts on the developing brain.

Additionally, although transcriptional changes seen in the hippocampal tissue of Beta-Acetate Group animals were less extreme (to pulsatile ACS-exposures), they still represented material changes consistent with increased risk of harm. Together with additional findings of poor lung maturation in animals exposed to very brief, high amplitude, pulsatile dexamethasone treatments, these data strongly underscore the importance of judicious patient selection, appropriately treatment targeting, and the need to develop constant, extended-duration glucocorticoid regimens to maximize benefit and minimize risk of harm deriving from ACS therapy.

Both the magnitude (dose) of materno-fetal antenatal steroid exposure, and the pharmacokinetic profile of that exposure (i.e., constant vs. pulsatile) contribute quantitatively and qualitatively to transcriptomic perturbation in the preterm fetal hippocampus. Since ACS dosing worldwide is based on pulsatile dosing regimens and high-dose dexamethasone is widely utilized as the antenatal steroid of choice, our data have significant implications for on-going efforts to optimize the safety and efficacy of this important obstetric intervention. Just as the dosing of oxytocin for labor induction and augmentation underwent a complete rethink a few decades ago with dramatic reduction in dosage, these data strongly suggest a need to conceptualize a similar process for antenatal steroids.