Histological chorioamnionitis shapes the neonatal transcriptomic immune response

Abstract

Background

Histologic chorioamnionitis (HCA) is commonly associated with preterm birth and deleterious post-natal outcomes including sepsis and necrotizing enterocolitis. Transcriptomic analysis has been used to uncover gene signatures that permit diagnosis and prognostication, show new therapeutic targets, and reveal mechanisms that underlie differential outcomes with other complex disease states in neonates such as sepsis.

Aims

To define the transcriptomic and inflammatory protein response in peripheral blood among infants with exposure to histologic chorioamnionitis.

Study Design

Prospective, observational study.

Subjects

Uninfected preterm neonates retrospectively categorized based on placental pathology with no HCA exposure (n = 18) or HCA exposure (n = 15).

Outcomes measures

We measured the transcriptomic and inflammatory mediator response in prospectively collected whole blood.

Results

We found 488 significant (p<0.001), differentially expressed genes in whole blood samples among uninfected neonates with HCA exposure that collectively represented activated innate and adaptive immune cellular pathways and revealed a potential regulatory role for the pleotropic microRNA molecule miR-155. Differentially secreted plasma cytokines in patients with HCA exposure compared to patients without HCA included MCP-1, MPO, and MMP-9 (p < 0.05).

Conclusions

Exposure to HCA distinctively activates the neonatal immune system in utero with potentially long-term health consequences.

Introduction

Preterm birth, defined as birth prior to 37 weeks of gestation, is the leading cause of neonatal death worldwide and a major contributor to the Global Burden of Disease (1, 2). The most common cause for spontaneous preterm birth is chorioamnionitis, affecting up to 80% of deliveries below 28 weeks of gestation (3, 4). Chorioamnionitis is an ascending infection of the intrauterine cavity during pregnancy. Histological chorioamnionitis (HCA) results in a fetal inflammatory response syndrome (FIRS) that can reprogram the developing immune system causing lifelong effects (5-8). A recent study in rhesus monkeys found that acute chorioamnionitis significantly reduced the ratio of T regulatory cells (Tregs) to interleukin (IL)-17-producing cells in lymphoid organs and was associated with inflammation and structural changes in the neonatal lung (9). Umbilical cord blood from human neonates with clinical evidence of perinatal infection had a higher proportion of T_H1 cells than umbilical cord blood from uninfected neonates (10). Furthermore, among patients with preterm premature rupture of membranes (PPROM), the percentage of IFN-γ-secreting T_H1 cells in the infected neonates correlated with the duration of membrane rupture before the onset of labor (11). While this evidence supports the capability of a robust intrauterine fetal innate and adaptive host response to HCA, specific changes in whole blood gene expression have not been characterized in human neonates. Thus, the use of transcriptomic endotyping in neonates exposed to HCA has significant potential to identify distinct pathways and host responses that will likely reveal previously undescribed pathophysiology.

We have recently used this methodology and approach to determine a unique transcriptome among infants with sepsis (12). In this study, we determined the transcriptomic response in prospectively collected whole blood and compared uninfected preterm neonates without HCA exposure to neonates with HCA exposure early after birth. Here we report a characteristic gene expression profile in uninfected neonates with exposure to HCA that indicates immune priming in utero.

Materials and Methods

Study population

As previously reported, this study was approved by the Institutional Review Boards at Duke University and Vanderbilt University before their initiation (12). Neonates admitted to the neonatal intensive care unit (NICU) and evaluated for sepsis on the basis of the judgment of the supervising clinician were eligible and enrolled once the parents gave informed consent (12). Whole blood was collected prospectively as part of the evaluation for early neonatal sepsis within 24 of birth for all the patients in this analysis (12). Complete blood cell counts were analyzed to identify the relative proportions of leukocyte subsets (12). Neonates with a known or suspected syndrome, abnormal karyotype or congenital malformation, or without available placental pathology were excluded from this analysis (12). Only neonates retrospectively classified as uninfected based on meeting all of the following signs and laboratory criteria were included in this analysis [the absence of clinical signs consistent with sepsis, negative blood culture results, serial CRP values (at least two values 24 h apart) < 10 mg/L), and discontinuation of antimicrobial therapy within 48 hours after birth]. Neonates were retrospectively classified as “HCA exposed” if placental membranes were scored as ≥ stage 1 histologic HCA (13). Alternatively, the neonate was classified as “no HCA” if the placenta lacked any histologic inflammatory changes.

Microarray

Whole blood samples (400–500 μL) were collected in ethylenediaminetetraacetic acid microtainer tubes followed by immediate RNA (ribonucleic acid) stabilization in PaxGene™ reagent as reported previously (12). Total RNA was isolated from stored human neonatal whole blood as previously described (12). Briefly, fifteen nanograms of total RNA were amplified using the Nugen Ovation Pico WTA System V2 (Nugen Technologies) and hybridized labeled cDNA targets to Affymetrix GeneChip^® GGh3 Transcriptome Array per standard protocols (12). Arrays were normalized by using Robust Multi-array Average (RMA) and only annotated probe sets representing 20,322 unique genes were used in the subsequent analysis (12).

Plasma inflammatory protein analysis

Plasma was isolated from whole blood immediately and stored for subsequent batch processing at − 80°C (12). As previously described, multiplex bead-based arrays were used to determine concentrations of multiple plasma proteins (RND; eBioscience) (12).

Statistical Analysis

As previously described, based on whether the data passed or failed tests of normality, we used the Student's t test or Wilcoxon signed rank test to compare results from the two groups (no HCA versus HCA exposed) (12). Binary variables were compared using Fisher's exact test with two-tailed p-value. Values were considered significant if p < 0.05. Microarray data were analyzed as previously described (12). Gene expression are available in GEO (#GSE69686).

Results

Neonates with and without exposure to HCA were similar in gestational age, sex, birth weight and mode of delivery (Table 1). All neonates had C-reactive protein values of < 10 mg/L. Median C-reactive protein values were 3.3 and 2.8 in the HCA and non-HCA group, respectively (p=0.84 by Wilcoxon signed rank test). Similarly, there was no statistically significant difference in either exposure to antenatal steroids (at least one dose before delivery) or absolute number of neutrophils, lymphocytes or monocytes between groups. Pathology-confirmed stages of HCA were 1, 2, or 3 in seven, five, and two cases, respectively. Five samples showed funisitis.

Table 1. Patient Demographics.

	No HCA, n = 18	HCA exposure, n = 15	p-value
Antenatal steroid exposure	56	73	0.16*
Vaginal delivery (%)	33	47	0.30*
Gestational age at birth, mean (range)	31.5 (27-36)	29.5 (25-34)	0.06ˆ
Birth weight, g, mean (range)	1586 (590-2480)	1342 (640-2200)	0.36ˆ
Male %	28	40	0.49*
C-reactive protein peak (mg/L)	2.8	3.3	0.84#
WBC/uL, mean (Range)	9441 (2400-18100)	14042 (3100-43600)	0.1#
Neutrophil no. (range)	3885 (698-9313)	8066 (403-35752)	0.22#
Lymphocyte no. (range)	4313 (1251-9072)	4276 (1953-7107)	0.90#
Monocyte no. (range)	869 (96-2025)	1182 (341-2275)	0.11#

^*-Fisher's exact test,

^#-Wilcoxon rank,

^ˆ-Paired t-test

We performed unsupervised and supervised analyses of the whole blood transcriptome as shown previously for neonates with and without sepsis (12). Importantly, although the approach we used was very similar to our previous study (12), this study focused exclusively on the comparison of the whole blood transcriptome measured within 24 hours of birth in uninfected neonates with and without HCA exposure. The unsupervised analysis showed 2964 probe sets with a coefficient of variation (CoV) > 0.5. Hierarchical clustering of these probe sets revealed a clear distinction between gene expression patterns in subjects exposed to HCA and those that did not. A supervised analysis was subsequently performed which yielded 488 significant (P<0.001) probe sets (Figures 1A/B). Leave one out cross validation and a Monte Carlo simulation revealed that differentially expressed genes could be used to identify HCA-exposed neonates at a probability greater than expected by chance p<0.004. Transcriptomic profiling of HCA-exposed neonates revealed greater evidence of activation (331 up-regulated genes) versus suppression (157 down-regulated). A principle component analysis (PCoA) (Figure 2) of the 488 probe sets revealed exposure to HCA as the major node of separation.

Figure 1. Unsupervised and supervised analysis of normalized expression data.

A Unsupervised analysis showed 2964 probe sets with a coefficient of variation (CoV) > 0.5. Hierarchical clustering of these probe sets revealed a clear distinction between gene expression patterns in subjects exposed to HCA and those that did not. B. A supervised analysis was subsequently performed which yielded 488 significant (P<0.001) probe sets. Leave one out cross validation and a Monte Carlo simulation revealed that differentially expressed genes could be used to identify HCA-exposed neonates at a probability greater than expected by chance p<0.004.

Figure 2. Principle component analysis (PCoA) by group revealed exposure to histologic chorioamnionitis as the major node of separation.

Histologic chorioamnionitis (HCA).

We uploaded the expression for the 488 probe sets identified by supervised analysis for each group into Ingenuity Pathway Analysis™ software to identify the top canonical pathways affected (Table 2). Specifically, Ras-related C3 botulinum toxin substrate (Rac) signaling, N-formyl-methionyl-leucyl-phenylalanine (fMLP) signaling in neutrophils, F_C-Receptor-mediated phagocytosis in macrophages and monocytes, F_C epsilon receptor I (F_CεRI) signaling, G beta-gamma (G_βγ) signaling. The top regulator effect network differentially expressed in HCA exposed neonates is shown in Figure 3. Top regulatory pathways included transcription factors SPI1 and GATA1, as well as mirRNA miR-155, all of which control of parts of hematopoiesis including myelopoiesis, erythropoiesis, and lymphopoiesis (14, 15).

Table 2. Top canonical pathways for differentially expressed genes in uninfected infants with HCA.

Name	p-value
Ras-related C3 botulinum toxin substrate (Rac) signaling	8.35 × 10-6
N-formyl-methionyl-leucyl-phenylalanine (fMLP) signaling in neutrophils	1.20 × 10-5
FC-Receptor-mediated phagocytosis in macrophages and monocytes	1.86 × 10-5
FC epsilon receptor I (FcεRI) signaling	6.80 × 10-5
G beta-gamma (Gβγ) signaling	7.22 × 10-5

Figure 3. The top regulatory effect network differentially expressed in histologic chorioamnionitis exposed neonates.

Ingenuity Pathway Analysis™ was used to examine expression of 488 significant genes from neonates with and without histologic chorioamnionitis exposure. GATA1, miR-155, SPL1 emerged as the most critical regulatory effect network with a consistency score of 8.1.

Differentially expressed genes are listed in Supplemental Table 1. The top 10 up-regulated genes (≥1.7-fold) in non-infected neonates with HCA exposure included: (1) olfactomedin 4 (OLFM4), a fundamental component of neutrophil granules (16) and potentially critical in neutrophil extracellular trap [NET] formation (17), (2) CD302, a C-type lectin receptor involved in cell adhesion and migration, as well as endocytosis and phagocytosis (18), (3) Lin-7 homolog A (LIN7A), a regulator of cell migration (19), (4) NME8, also known as thioredoxin domain-containing protein 3 (TXNDC3) implicated in ciliary function (20), (5) liver glycogen phosphorylase gene PYGL, (6) S100A12, (other names include calgranulin C, MRP6, or EN-RAGE (extracellular newly identified receptor for advanced glycation endproducts binding protein), is a calcium-binding pro-inflammatory protein predominantly secreted by granulocytes (21), (7) probable phospholipid-transporting ATPase IM (ATP8B4), (8) complement component 5a receptor 1 (C5AR1), expressed on neutrophils and a potential contributor to fetal cortical brain injury (22, 23) (9) CD93, another C-type lectin transmembrane receptor with a role in cell–cell adhesion processes and neonatal host defense (24, 25), and (10) cystatin A / stefin A (CSTA), which plays a role in epidermal cell development and host defense of the neonatal skin (26-28) (Supplementary Table S1). Of note, with the exception of OLFM4, the top ten differentially expressed and upregulated genes in whole blood from uninfected HCA-exposed neonates were entirely distinct from the top ten in whole blood from neonates with sepsis using identical methods (12). Top down-regulated genes included SLC6A19, the gene encoding the sodium-dependent neutral amino acid transporter B(0)AT1, which mediates transport of neutral amino acids across renal and intestinal epithelia (29) and mitochondrial acyl-CoA thioesterase THEM5, which is involved in diseases such as metabolic syndrome, diet-induced obesity, and diabetes (30).

Plasma cytokine analysis revealed statistically significant alterations in patients with HCA exposure for MCP-1, MPO, and MMP-9 (Figure 4; all p < 0.05 by Wilcoxon signed rank test). Overall the specific plasma protein profile of uninfected HCA-exposed neonates differed strikingly from that of septic neonates as we found in our previous work (12), suggesting a different fetal host response program after HCA compared to neonatal sepsis.

Figure 4. Concentration (pg/mL) of plasma inflammatory mediators by group.

Median with interquartile range (25-75^th percentile) are shown. *-P<0.05 by t-test. Histologic chorioamnionitis (HCA), (Monocyte chemoattractant protein-1 (MCP-1), myeloperoxidase (MPO), and matrix metalloproteinase- 9 (MMP-9).

Discussion

To the best of our knowledge we report the first distinguishing gene expression profile in uninfected preterm neonates exposed to HCA to define immune priming following HCA. Collectively, we detected activated innate immunity and T cell pathways likely contributing to the frequently lasting inflammatory sequelae in preterm neonates exposed to HCA (5, 7).

The significant influence of the intrauterine environment on the developing immune system is evident from the anti-inflammatory cytokine profile and protection from atopic sensitization in offspring after maternal exposure to farming activities and farm dairy products during pregnancy (31, 32). Understanding how in utero inflammation programs the immune response in preterm neonates has the potential to uncover new ways of modulating immune responses with the goal to reduce inflammatory injury and/or improve vaccine effectiveness. Importantly uninfected preterm neonates with exposure to HCA demonstrated a significantly different transcriptomic profile compared to infected preterm neonates at similar gestation and age (12). For example we identified microRNA miR-155 as a top upstream regulator and top regulator effect network in uninfected HCA exposed neonates. MicroRNAs are epigenetic elements binding to mRNAs and down-regulating their translation to protein. MiR-155 is an inflammatory response microRNA induced by endotoxin and during the macrophage inflammatory response (33, 34). MiR-155 plays a key role in the homeostasis and function of the immune system and increased expression in humans has been associated with chronic inflammatory states such as atopic dermatitis, multiple sclerosis, and rheumatoid arthritis (35, 36). MiR-155 is expressed in activated CD4⁺ T cells and promotes development of T_H1 and T_H17 biased responses (35). This data correlates well with increased proportions of T_H1 cells in umbilical cord blood of neonates with evidence of perinatal infection and offspring of mothers with preterm PROM (10, 11). A recent study in rhesus monkeys found that acute chorioamnionitis significantly reduced the ratio of T regulatory cells to IL-17-producing cells in lymphoid organs and was associated with inflammation and structural changes in the neonatal lung (9). Overall prenatal inflammation alters innate and adaptive immune pathways in the preterm host known to contribute to immune dysregulation and organ injury. While transcriptomics from small blood samples may be useful in identifying preterm neonates at risk for perinatal inflammatory injury from HCA soon after birth, the relative paucity of differences in plasma profiling between uninfected neonates with or without HCA make plasma protein profiling likely less practical.

Our study has several limitations. First, we performed transcriptomics on whole blood, which does not allow us to discern the relative contributions of specific circulating immune cells. Second, our sample size was limited. However we purposely utilized technologies that allowed measurement of gene expression and peptide profiling from small volumes of neonatal peripheral blood (instead of cord blood) to more accurately reflect ongoing neonatal rather than prenatal fetal physiology. Collection of peripheral blood samples from preterm infants is more challenging than cord blood studies. Third, while we attempted to match preterm neonates by factors that can influence immune cell function, we are aware that other variables in the maternal and intrauterine environment (e.g. maternal smoking, intrauterine growth restriction, exposure to antimicrobials) as well as neonatal factors may have influenced the transcriptome. Despite these limitations, this is the first study that reports the transcriptomic response to HCA among preterm infants. Our data may lead to the design of specific tools for diagnosis, treatment, or prevention of HCA-associated sequelae in preterm neonates.

Supplementary Material

1

Supp Table 1: Differentially expressed genes between groups (n = 488).

Highlights.

• 488 significant, differentially expressed, genes in whole blood among uninfected neonates with histologic chorioamnionitis exposure

• Activation of both innate and adaptive immune cellular pathways was shown and there was a potential regulatory role for the pleotropic microRNA molecule miR-155.

• Plasma cytokines in patients with HCA exposure compared to patients without HCA included MCP-1, MPO, and MMP-9.

Abbreviations

NICU

Neonatal intensive care unit

HCA

histologic chorioamnionitis

CRP

C-reactive protein

CoV

coefficient of variation

GEDI

Gene Expression Dynamics Inspector

PCoA

principle component analysis