Gene expression profiling of placentae from women with early- and late-onset pre-eclampsia: down-regulation of the angiogenesis-related genes ACVRL1 and EGFL7 in early-onset disease

Abstract

The underlying mechanisms behind the obstetric condition pre-eclampsia (PE) are still unclear. Manifestation of PE is heterogeneous and it has therefore been proposed to be a syndrome with different causes rather than one disease with a specific aetiology. Recently, we showed differences in circulating angiogenic factors between two subgroups—early- and late-onset PE. To further elucidate the differences between the two, we investigated placental gene expression profiles. Whole genome microarray technology and bioinformatic analysis were used to evaluate gene expression profiles in placentae from early- (24–32 gestational weeks, n = 8) and late-onset (36–41 gestational weeks, n = 7) PE. The results were verified by using quantitative real-time (qRT)–PCR. We found significant differences in the expression of 196 genes in early- compared with late-onset PE, 45 of these genes showing a fold change above 2. Bioinformatic analysis revealed alterations in angiogenesis and regulation of cell motility. Two angiogenesis-associated transcripts (Egfl7 and Acvrl1) showed lower expression in early-onset PE versus late-onset PE (P = 0.037 and P = 0.003) and versus gestational age-matched controls (P = 0.007 and P = 0.011). We conclude that angiogenesis-associated genes are regulated in a different manner in the two subgroups, and that the gene expression profiles of early- and late-onset PE diverge, supporting the hypothesis of early- and late-onset PE being at least partly two separate entities.

Introduction

The obstetric syndrome pre-eclampsia (PE) is characterized by new onset of elevated blood pressure together with proteinuria, which manifests itself in the second half of gestation. It affects 2–8% of all pregnancies (Duley, 2009) and is one of the leading causes of maternal and perinatal mortality and morbidity worldwide. Despite its high incidence, the aetiology of PE remains unclear. However, the disorder is believed to originate in the placenta, since PE ends with delivery and the removal of placental tissue.

One major abnormality found in PE is poor placentation, resulting in uneven placental perfusion, consequently causing oxidative stress and inflammation (Redman and Sargent, 2009). This may lead to the release of biologically active placental factors into the maternal circulation. These factors then bring about the clinical manifestation of PE (Roberts and Hubel, 2009): general maternal endothelial dysfunction and systemic inflammation.

Imbalance between pro- and anti-angiogenic factors has been shown to be important in the pathogenesis of PE. Members of the vascular endothelial growth factor (VEGF) family are important factors involved in placental angiogenesis, and since they deviate in women with PE, they have been suggested as the link between stage I and stage II, where the clinical manifestations appear. Levels of placental growth factor (PlGF) are decreased in the maternal circulation several weeks before the onset of PE (Tidwell et al., 2001). In contrast, concentrations of the (soluble) VEGF receptor soluble fms-like tyrosine kinase 1 (sFlt1) are increased in the maternal circulation in women destined to develop PE (Levine et al., 2004). This receptor binds to and antagonizes both VEGF and PlGF by preventing them interacting with their endogenous membrane receptors.

PE is a heterogeneous disorder. Early-onset PE, defined as onset of clinical symptoms before 35 weeks of gestation, is associated with more pronounced symptoms and a poorer outcome than late-onset PE. A growing body of evidence suggests that early-onset PE diverges from late-onset PE to such an extent that in fact they may be two different diseases or subgroups sharing a similar clinical outcome (von Dadelszen et al., 2003; Vatten and Skjaerven, 2004). Signs of abnormal placentation are more frequently observed in early- than in late-onset PE. In addition, early-onset PE is associated with intrauterine growth restriction and abnormal changes in the blood flow of the umbilical arteries, whereas late-onset PE is associated with a normally growing fetus and usually no changes in umbilical artery blood flow (Huppertz, 2008). Despite these differences, early- and late-onset PE are not separated in most studies of the condition.

Our previous results support the subgrouping of the syndrome into early- and late-onset PE. Both the increase of anti-angiogenic sFlt1 and the decrease of pro-angiogenic PlGF in PE are more pronounced in early- than in late-onset PE (Wikstrom et al., 2007). Moreover, we recently reported that circulating levels of the potent antagonist of angiogenesis, endostatin, are increased in early- but not in late-onset PE (Wikstrom et al., 2009a), and that early- but not late-onset PE is associated with increased placental oxidative stress and an increased plasminogen activator inhibitor (PAI)-1 to PAI-2 ratio (Wikstrom et al., 2009b).

Microarray technology is a powerful tool that can be used to study complex disorders such as PE. The gene expression profile of the placenta in cases of PE has been investigated in a number of microarray studies (Hansson et al., 2006; Centlow et al., 2008; Founds et al., 2008). In this study, we used whole genome microarray technology to investigate differences in placental gene expression between early- and late-onset PE. We show differences in the RNA expression of 196 genes, of which several are associated with angiogenesis.

Materials and Methods

The study was approved by the Ethics Committees at Uppsala University, Sweden and Lund University, Sweden. Informed consent was obtained from each study subject before tissue sampling.

Sample collection

Placental tissue samples from four subject groups were included in the study. For the microarray experiment, samples from eight women with early-onset PE and seven women with late-onset PE were collected. PE was diagnosed as new onset of hypertension (>140/90 mmHg) and proteinuria (≥ 2 on a dipstick, or a 24-h urine sample showing ≥ 300 mg/24 h; Milne et al., 2005). Between gestational Weeks 24 and 32 (Days 161–224), PE was defined as early-onset and between gestational Weeks 36 and 41 (Days 245–287) as late-onset. Women with PE between gestational Weeks 33 and 35 were excluded. In addition, women with chronic disease such as chronic hypertension, diabetes and renal disease were excluded. For qRT–PCR verification, two control groups of pregnant women without PE were included: four early controls and six late controls. Samples from these two groups are from gestational days 160–182 and 267–298, respectively. Early controls were delivered prematurely either due to uterine malformations or cervical insufficiency. Women with clinical or laboratory signs of infection were not included. Cases and controls who delivered at term were recruited from the Women's Clinic at Lund University Hospital, Sweden. Cases and controls who delivered prematurely were recruited from the Women's Clinic at Uppsala University Hospital, Sweden. Using the Statistical Package for the Social Sciences (SPSS), statistical differences between the groups were assessed by Fisher's exact test and the Kruskal–Wallis test with post hoc Mann–Whitney testing. Placental tissue was collected immediately after either vaginal or Caesarean section delivery. In Lund, tissue was dissected from a central part of the placenta and was immediately frozen on dry ice. In Uppsala, tissue samples were taken from five different cotyledons from the central part of each placenta in order to avoid gene expression variation within the same organ. Samples were rinsed three times in saline, and were thereafter frozen in liquid nitrogen. All tissue samples were stored at −70 to −80 °C.

RNA extraction

Placental total RNA from early-onset PE, late-onset PE and late controls was extracted and washed using TRIZOL^® (Invitrogen, Carlsbad, USA) and E.Z.N.A. total RNA Kits (Omega Biotek, Doraville, USA) as previously described (Centlow et al., 2009). Total RNA from early control placentae was extracted by using RNeasy Mini Kits (Qiagen Inc., Valencia, USA) according to the manufacturer's instructions. RNA concentrations were measured spectrophotometrically by using a NanoDrop instrument (NanoDrop Technologies, Wilmingon, USA). RNA integrity was assessed by electrophoresis, using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, USA).

Microarrays

Complementary DNA (cDNA) probes were prepared and labelled with the fluorescent dyes Cy3-dCTP and Cy5-dCTP (Amersham Biosciences, Buckinghamshire, UK) using the Corning ChipShot™ labelling system according to the manufacturer's instructions (Corning, Acton, USA). A pool of reference Cy5-labelled cDNA was prepared using total RNA from a healthy placenta. Samples were labelled with Cy3 using 5 μg of total RNA from each sample. Before hybridization, 50 pmol Cy3-labelled cDNA from each experimental sample was pooled with 50 pmol of reference Cy5-labelled cDNA.

Arrays were printed with a whole genome (27 000 oligos) oligo set (OPERON v 2.1 human 70-mer oligo set, Operon, Huntsville, USA) at the Swegene DNA Microarray Resource Center, Department of Oncology, Lund University, Sweden, as described before (Centlow et al., 2008). The arrays were pretreated, hybridized with the pooled labelled probes and washed using a Pronto! Microarray Hybridization kit according to the manufacturer's instructions (Corning, Acton, USA).

An Agilent G2565AA microarray scanner (Agilent Technologies, Palo Alto, USA) was used to read the fluorescence intensities of the hybridized arrays at 635 (Cy5) and 532 (Cy3) nm. Spot intensities were visualized by using GenePix Pro 4.1 software (Axon Instruments Inc., Foster City, USA) and spots affected by veils, grains of dust or other contaminants were removed from further analysis. The data generated via GenePix Pro were analysed by using the Bio Array Software Environment BASE (Saal et al., 2002). Spots with intensities below 100 in both of the two channels were excluded. Individual spot intensities were also required to be two times as high as the background to be included in further analysis. Data were normalized for labelling and reference bias by using lowess normalization and global median centring. Fluorescence ratios were calculated as Cy3/Cy5 and log₂-transformed. To compare expression between early- and late-onset PE, significance analysis of microarrays (SAM; Tusher et al., 2001) with the false discovery rate set to 5% (q-value of <0.05) was used. Only spots that were reported in all 15 arrays were included in the group comparison analysis.

Primary microarray data were submitted to the Gene Expression Omnibus (GEO), accession number GSE22526.

Bioinformatic analysis

To extract biological meaning from the list of up- and down-regulated genes, the database for annotation, visualization and integrated discovery (DAVID) was used. Significant changes were uploaded to the database and analysed with the whole human genome as background (Dennis et al., 2003; Huang da et al., 2009). Genes were classified [gene ontology (GO)] on the basis of biological process, molecular function and cellular component. To identify the pathways in which significantly changed gene expressions were involved, the Kyoto Encyclopedia of Genes and Genomes (KEGG) was used. Annotations for protein domains were extracted from InterPro. A P-value of ≤0.05 and a fold change (FC) of ≥2 were used as cut-offs.

Quantitative real-time PCR

cDNA was synthesized from RNA with SuperScript™ III First-Strand Synthesis SuperMix (Invitrogen, Carlsbad, USA) according to the manufacturer's protocol. Briefly, 0.1 µg of total RNA was reverse-transcribed using oligo (dT)₂₀, random hexamers and dNTPs. The cDNA samples were stored at −20°C until further use.

Relative mRNA levels of the genes of interest were measured by qPCR as previously described (Helmestam et al., 2010). Primers (Table I; CyberGene AB, Stockholm, Sweden), complementary to an exon boundary to avoid amplification of genomic DNA, were designed according to guidelines from Applied Biosystems. Melt curve analysis was carried out after each experiment to ensure specific amplification.

Table I.

Real-time PCR primers used to verify the microarray results.

Gene	NCBI accession number	Forward primer 5′ → 3′	Reverse primer 5′ → 3′
ACVRL1	NM_000020.2	CCGGACCATCGTGAATGG	GTCATTGGGCACCACATCATAG
EGFL7	NM_016215.3	GTGGACCTGCTGGAGGAGAA	TCCGGGAGCCCATGCT
IDO1	NM_002164.4	GCTGGGCATCCAGCAGACT	TCTTCTCATGTCCTGGAGGAACT
ROBO4	NM_019055.5	GATGCCATCCTAAAACACAGGAT	CCGAGAGCCAGAGGTGGAA

The genes for TATA box-binding protein (TBP), tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein, zeta polypeptide (YWHAZ) and succinate dehydrogenase complex, subunit A (SDHA) were used as reference genes and have been reported as being stable in pre-eclamptic placental tissue (Meller et al., 2005). LinRegPCR software (Version 11.3) was used for Ct extractions and efficiency calculations for all the PCRs (Ruijter et al., 2009). Mean normalized expression (MNE) for each gene was calculated by using the following equation:

where E is the mean experimental efficiency of the primer and Ct is the mean Ct of triplicate reactions. Using SPSS software, statistical differences in expression between all groups were determined (Kruskal–Wallis test). If the test was significant, the post hoc Mann–Whitney test was used to evaluate differences between individual groups. A P-value of <0.05 was considered significant.

Results

In order to find differences in placental gene expression between early- and late-onset PE, we compared the gene expression profiles in eight cases with early-onset PE with those in seven cases with late-onset PE, using whole genome microarrays. The clinical data of the cases and their controls are presented in Table II. There were no differences in maternal age or parity. Systolic and diastolic blood pressures did not differ between the two groups of PE cases. However, all but one of the women with early-onset PE were treated with anti-hypertensive drugs, compared with the late-onset PE group in which nobody was treated. The mode of delivery differed between the two groups. All eight in the early-onset group delivered by Caesarean section, compared with one in the late-onset PE group. The median length of gestation in the early-onset PE group was 205 days, and in the late-onset PE group, 279 days.

Table II.

Clinical characteristics at delivery for all subjects included in the study.

	Late control (n = 6)	Early control (n = 4)	Late PE (n = 7)	Early PE (n = 8)
Maternal age (years)	24 (21–30)	34 (30–39)	30 (22–45)	31.5 (29–39)
Nulliparous n (%)	5 (83)	3 (75)	6 (86)	6 (75)
Systolic pressure (mmHg)a,c	120 (100–140)	123 (102–126)	160 (140–168)	150 (140–186)
Diastolic pressure (mmHg)a,b	70 (60–80)	79 (68–84)	105 (95–110)	99.5 (81–111)
Hypertensive treatment n (%)c,d	0 (0)	0 (0)	0(0)	7 (87.5)
Proteinuria (0–3)a,c	0 (0–0)	0 (0–0)	2 (1–3)	3 (1–3)
Gestational age at delivery (days)b,d	281.5 (267–298)	170.5 (160–182)	279 (259–285)	205 (172–214)
Caesarean section n (%)b,e	2 (33)	0 (0)	1 (16)*	8 (100)
Birthweight (grams)d	3612 (3300–4000)	649.5 (490–963)	3290 (2695–3760)	990 (428–1320)
Placental weight (grams)d	605 (450–800)	235 (130–280)	650 (500–850)	237.5 (82–325)*

Values are expressed as the median (range) for continuous variables and n (%) for categorical variables. For continuous variables, the Kruskal–Wallis test followed by the (post hoc) Mann–Whitney test was used to evaluate differences between the four groups. For categorical variables, Fisher's exact test was used.

^aSignificant difference between Late Control versus Late PE (P <0.01).

^bSignificant difference between Early Control versus Early PE (P <0.05).

^cSignificant difference between Early Control versus Early PE (P <0.01).

^dSignificant difference between Late PE versus Early PE (P <0.01).

^eSignificant difference between Late PE versus Early PE (P <0.05).

*Data missing from one subject.

SAM, with a false discovery rate set to 5% revealed 196 genes with significantly altered expression in early- compared with late-onset PE placentae. Of these, 88 genes were up-regulated and 108 genes down-regulated in early-onset PE. Among these, 45 showed an FC of ≥2 (Table III).

Table III.

Genes showing significantly altered expression in the microarray experiment (q-value of <5%, FC ≥2).

Gene symbol	GenBank accession	FC	q-value (%)
Up-regulated in early PE
TCL6	NM_020553	2.00	0.00
C15orf29	NM_024713	2.60	0.00
RAB11FIP5	NM_015470	2.56	0.00
N/A	AF090099	2.27	0.00
LIFR	NM_002310	3.46	0.00
CREBZF	NM_001039618	2.09	0.00
DEPDC1B	NM_018369	2.83	0.00
DAB2	NM_001343	3.19	0.00
N/A	AL365412	2.21	0.00
HOPX	NM_139212	4.29	0.00
MAP3K9	NM_033141	2.33	1.67
CYP19A1	NM_000103	3.31	1.65
EXPH5	NM_015065	3.18	1.71
PSPC1	NR_003272	2.23	2.10
DAB2	NM_001343	2.19	2.10
CCDC21	NM_022778	2.04	2.47
TREML2	NM_024807	3.11	2.89
N/A	AL121777	3.78	3.84
HES2	NM_019089	2.07	3.86
N/A	AK023465	3.44	4.10
TRIM25	NM_005082	2.09	4.70
CGA	NM_000735	2.28	4.83
Down-regulated in early PE
KRT3	NM_057088	−2.27	0.00
FADS3	NM_021727	−2.08	0.00
ACTA2	NM_001613	−2.78	0.00
ID1	NM_002165	−2.17	0.00
CORO6	NM_032854	−2.56	0.00
APOLD1	NM_030817	−4.17	0.00
TMEM175	NM_032326	−2.22	0.00
HBB	NM_000518	−7.69	0.00
KLF2	NM_016270	−2.78	0.00
AFF3	NM_002285	−2.33	0.00
SDPR	NM_004657	−2.38	1.53
EGFL7	NM_201446	−2.04	1.93
TMEM100	NM_001099640	−2.78	1.93
CLEC3B	NM_003278	−2.04	1.93
ACVRL1	NM_000020	−2.04	1.93
C4orf31	NM_024574	−2.04	1.93
DDIT4	NM_019058	−2.56	1.93
RCAN1	NM_004414	−2.08	1.93
IDO1	NM_002164	−2.27	2.20
SNCA	NM_000345	−2.38	3.79
N/A	AK024295	−2.13	4.70
MAGEA11	NM_005366	−2.08	4.87
HBA2	NM_000517	−3.70	4.87

Genes up-regulated or down-regulated in early- compared with late-onset PE are listed, by increasing q-value.

To give biological meaning to the differentially expressed genes, we used DAVID bioinformatic resources. GO analysis was carried out to investigate the biological processes, molecular functions and cellular components with which the significantly changed gene expression patterns were associated. Two different gene lists were used: the first contained all 196 significantly changed genes and the second the 45 genes with an FC of ≥2. The results from the analysis of all significantly changed genes are presented in Table IV and the results from the analysis of the 45 significantly changed genes with an FC of ≥2 in Table V. Among the GO groups found by way of this analysis were GO0001525 angiogenesis (first list, P = 0.0015, FC = 5.7, second list, P = 0.022, FC = 12.39), GO0051270 regulation of cell motility (first list, P = 0.022, FC = 6.6), GO0051271 negative regulation of cell motility (second list, P = 0.042, FC = 45.44), GO:0015671 oxygen transport (second list, P = 0.029, FC = 65.64) and GO:0020037 haem binding (first list P = 0.028, FC = 4.3; second list P = 0.00201, FC = 15.29). Pathway analysis was carried out to investigate whether the significantly altered genes were involved in any specific signalling pathways. Clustering of transcriptionally altered genes was found in the Alzheimer's disease pathway (hsa05010; P = 0.0311, FC = 10.5). No pathway association was found when solely analysing the genes with FC ≥ 2. Protein domain analysis of the first list resulted in significant clustering of genes in the following groups: IPR001092 Basic helix-loop-helix dimerization region bHLH (P = 0.0002, FC = 8.06), PR003961 Fibronectin, type III (P = 0.015, FC = 4.12) and IPR006662 Thioredoxin-related (P = 0.038, FC = 9.60). In the second list, genes belonging to the protein domain groups IPR012292 Globin (P = 0.019, FC = 99.14) and IPR000971 Globin, subset (P = 0.021, FC = 91.51) were enriched, i.e. there was significant clustering of genes in these two groups compared to a random set of genes.

Table IV.

Gene ontological analysis of the differentially expressed genes (q<5%) in the microarray experiment with the whole human genome as background.

GO term	GO ID	P-value	Fold enrichment
Biological process
Positive regulation of metabolic process	GO:0009893	0.0005	3.4
Positive regulation of transcription	GO:0045941	0.0005	3.9
Blood vessel morphogenesis	GO:0048514	0.0006	5.6
Positive regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolic process	GO:0045935	0.0006	3.8
Positive regulation of cellular metabolic process	GO:0031325	0.0009	3.3
Blood vessel development	GO:0001568	0.0012	4.9
Vasculature development	GO:0001944	0.0013	4.8
Angiogenesis	GO:0001525	0.0015	5.7
Anatomical structure morphogenesis	GO:0009653	0.0016	2.1
Regulation of locomotion	GO:0040012	0.0038	7.8
Locomotion	GO:0040011	0.0040	7.7
Anatomical structure formation	GO:0048646	0.0045	4.5
Positive regulation of cellular process	GO:0048522	0.0051	2.1
Positive regulation of biological process	GO:0048518	0.0065	2.0
Organ morphogenesis	GO:0009887	0.0096	2.8
Generation of precursor metabolites and energy	GO:0006091	0.010	2.3
Negative regulation of protein metabolic process	GO:0051248	0.010	5.9
Cellular catabolic process	GO:0044248	0.014	2.3
Regulation of cell migration	GO:0030334	0.016	7.6
Regulation of cell motility	GO:0051270	0.022	6.6
Electron transport	GO:0006118	0.023	2.4
Positive regulation of transcription. DNA-dependent	GO:0045893	0.026	3.1
Catabolic process	GO:0009056	0.027	2.0
Regulation of focal adhesion formation	GO:0051893	0.034	57.7
Negative regulation of metabolic process	GO:0009892	0.035	2.4
Morphogenesis of an epithelium	GO:0002009	0.036	5.5
Macromolecule catabolic process	GO:0009057	0.037	2.3
Tube development	GO:0035295	0.038	3.9
Negative regulation of translation	GO:0017148	0.044	8.9
Regulation of cell differentiation	GO:0045595	0.048	3.6
Molecular function
Oxidoreductase activity	GO:0016491	0.0003	2.6
Growth factor binding	GO:0019838	0.0024	8.8
Electron carrier activity	GO:0009055	0.0025	4.3
Protein N-terminus binding	GO:0047485	0.0036	12.9
Copper ion binding	GO:0005507	0.0048	7.3
Heme binding	GO:0020037	0.028	4.3
Tetrapyrrole binding	GO:0046906	0.028	4.3
mRNA binding	GO:0003729	0.035	10.1
Iron ion binding	GO:0005506	0.039	2.8
Oxygen binding	GO:0019825	0.039	9.4
Transcription activator activity	GO:0016563	0.040	2.8
Cellular component
Endosome membrane	GO:0010008	0.040	5.3
Endosomal part	GO:0044440	0.040	5.3

Groups with P <0.05 and fold enrichment ≥2 are listed, by increasing P-value.

Table V.

Gene ontological analysis of the 45 differentially expressed genes (q<5%) with FC ≥2 in the microarray experiment with the whole human genome as background.

GO term	GO ID	P-value	Fold enrichment
Biological process
Regulation of locomotion	GO:0040012	0.0064	24.0
Locomotion	GO:0040011	0.0066	23.6
Developmental process	GO:0032502	0.0088	2.17
Multicellular organismal development	GO:0007275	0.0090	2.51
Angiogenesis	GO:0001525	0.022	12.39
Oxygen transport	GO:0015671	0.029	65.64
Blood vessel morphogenesis	GO:0048514	0.030	10.68
Blood circulation	GO:0008015	0.030	10.61
Circulatory system process	GO:0003013	0.030	10.61
Gas transport	GO:0015669	0.032	59.08
Negative regulation of cellular process	GO:0048523	0.033	3.13
Anatomical structure formation	GO:0048646	0.034	9.90
Negative regulation of cell migration	GO:0030336	0.035	53.71
Blood vessel development	GO:0001568	0.037	9.43
Vasculature development	GO:0001944	0.038	9.28
Negative regulation of biological process	GO:0048519	0.039	3.00
Negative regulation of cell motility	GO:0051271	0.042	45.44
Negative regulation of locomotion	GO:0040013	0.043	43.76
Molecular function
Tetrapyrrole binding	GO:0046906	0.00201	15.29
Haem binding	GO:0020037	0.00201	15.29
Oxygen binding	GO:0019825	0.00222	41.39
Iron ion binding	GO:0005506	0.01732	7.00
Oxygen transporter activity	GO:0005344	0.03199	59.54
Transcription regulator activity	GO:0030528	0.04214	2.58
Cellular component
Haemoglobin complex	GO:0005833	0.02894	66.07

GO groups with P <0.05 and fold enrichment ≥2 are listed, by increasing P-value.

Of the genes with an FC ≥2, we selected two angiogenesis-associated genes and one gene associated with placental development for verification of the microarray results by qRT–PCR (Fig. 1). We also included one angiogenesis-associated gene, Roundabout homolog 4, magic roundabout (Robo4, q-value = 3.34%, FC = −1.80), from the first list in the verification step. The angiogenesis-associated genes Activin A receptor type II-like 1 (Acvrl1, P = 0.003), EGF-like-domain, multiple 7 (Egfl7, P = 0.037) and Robo4 (P = 0.015) as well as a gene associated with placental development, Indoleamine 2, 3-dioxygenase 1 (Ido1, P = 0.011), were down-regulated in early-onset PE, confirming the microarray data.

Figure 1.

qRT–PCR verification of the microarray data and mRNA expression in control placentae from preterm and term pregnancies. Results are presented as box-plots showing the median, the 25th and the 75th percentiles. Whiskers depict range. MNE was calculated by using the geometric mean of the reference genes TBP, YWHAZ and SDHA. The Kruskal Wallis test followed by post hoc Mann–Whitney U-test was used to determine statistical significance. *P <0.05. **P <0.01.

Gene expression varies with normal placental development. To investigate if the discovered alterations in gene expression were related to PE and not gestational age, we compared the mRNA levels of the genes verified by qRT–PCR in the two PE groups with those in gestational age-matched controls. Length of gestation at delivery was slightly longer in the early-onset PE group than in their controls (P = 0.040; Table II). Real-time PCR analysis showed differences in the expression of two of the four genes validated earlier, the angiogenesis-associated transcripts Acvrl1 (P = 0.011) and Egfl7 (P = 0.007)—both decreased in early PE compared with early controls (Fig. 1).

There were no differences in mRNA levels of the four selected genes when comparing women with late-onset PE with late controls (Fig. 1).

Discussion

Recently it has become clear that the heterogeneous disease PE may be divided into different subtypes, and the importance of studying PE pathology from this perspective has been highlighted (Roberts and Hubel, 2009). There are previous studies in which placental gene expression in early- and late-onset PE have been compared (Nishizawa et al., 2007; Sitras et al., 2009). The present study is, to the best of our knowledge, the first involving exclusion of women who delivered in Weeks 33–35, a period when it is difficult to decide whether the disorder is to be considered early- or late-onset. By doing this, we obtained a more distinct sample set. Several genes were found to be differentially expressed in early- and late-onset PE (Table III). Verification of the microarray data by qRT–PCR showed similar results for the four selected genes, implying that the microarray data are reliable (Fig. 1). Pathway analysis of the altered transcripts showed involvement of the Alzheimer's disease pathway. Interestingly, our data are compatible with that of van Dijk et al. (2010), in which they have proposed that PE and Alzheimer's disease share a conserved pathway involving the PE-susceptibility gene STOX1.

Biological function analysis of the altered transcripts revealed a difference in expression of angiogenesis-related genes. This was an expected finding, since previous studies have shown that angiogenesis is involved in the pathology of PE (Wikstrom et al., 2007, 2009a). However, we did not anticipate finding such a major difference in functional GO analysis, suggesting that angiogenesis and angiogenic factors, such as Acvrl1 and Egfl7, not yet associated with PE, are important in the pathogenesis of early-onset PE. These findings further support the hypothesis of angiogenesis being involved in the pathogenesis of early-onset PE.

Of the altered genes four genes with possible links to PE pathology were particularly interesting: Acvrl1, Egfl7, Robo4 and Ido1. Therefore, they were selected for verification of the microarray results and gestational age-matched comparisons with control placentae by means of qRT–PCR. Comparison with controls to determine if the observed differences in gene expression are due to differences between early- and late-onset PE, and not gestational age per se, is an approach that not has been used in previous studies. The importance of this approach became obvious when some of the genes that were differentially expressed when comparing early- and late-onset PE were found to be gestational age-related rather than PE-related when including early and late controls. We found that the two angiogenesis-associated factors EGF-like-domain, multiple 7 (Egfl7) and Activin A receptor type II-like 1 (Acvrl1) were down-regulated in early-onset PE compared with both late-onset PE and gestational age-matched controls.

The Egfl7 transcript is up-regulated in physiological angiogenesis and during vascular injury (Parker et al., 2004; Campagnolo et al., 2005). The secreted protein EGFL7 represses smooth muscle but not endothelial cell migration in vitro (Soncin et al., 2003), and EGFL7 mutant mice show defects in vascular morphogenesis. Furthermore, EGFL7 is part of the extracellular matrix deposited on the basal side of vascular sprouts, and it has been proposed that it ensures an appropriate microenvironment for endothelial cell migration (Schmidt et al., 2007). The role of Egfl7 in pregnancy remains to be studied, but it is tempting to speculate that down-regulation of its expression in early-onset PE contributes to dysfunctional angiogenesis described in early-onset PE placentae.

Acvrl1 encodes ALK1, a co-receptor in the transforming growth factor (TGF)-β signalling pathway. It partners TGF-β receptor type II (TβR-II) and binds TGF-β1 and TGF-β3. ALK1 is primarily expressed in endothelial cells and stimulates endothelial cell migration, proliferation and tube formation (Goumans et al., 2003). Mutations in this gene cause hereditary haemorrhagic telangiectasia, characterized by telangiectases and arteriovenous malformations (Johnson et al., 1996). Interestingly, mutations in the Endoglin (Eng) gene, a gene associated with PE in a number of reports (Venkatesha et al., 2006), represent the only other cause of this disease (McAllister et al., 1994). Endoglin is expressed in vascular endothelial cells at sites of active angiogenesis and in syncytiotrophoblasts. Like ALK1, it is also involved in the TGF-β signalling pathway. Both placental Eng and a soluble form of the protein, sEng, found in plasma, are increased in PE (Venkatesha et al., 2006), and sEng is associated with early and/or severe disease (Lambert-Messerlian et al., 2009). It is hypothesized that sEng, together with sFlt, contributes to the maternal syndrome of PE. This soluble form of endoglin antagonizes the cell surface variant by binding TGF-β family ligands. It has been reported to interfere with TGF-β signalling and endothelial nitric oxide synthase, thereby causing endothelial dysfunction. The decreased levels of placental Acvrl1 transcripts we observed in the present study may interfere with the TGF-β signalling pathway in a similar way as sEng, thereby contributing to the pathophysiology of PE. Additional studies need to be conducted to clarify the involvement of Acvrl1 in TGF-β signalling and its connection to PE pathogenesis. Activin A is another protein in the TGF-β superfamily. The gene encoding one of its two receptors, Activin receptor type 2 (Acvr2), has recently been suggested as a candidate susceptibility gene associated with PE in a large Norwegian population-based study (Roten et al., 2009). Our data strengthen the theory that TGF-β superfamily genes and/or pathways are involved in PE.

Indoleamine 2,3-dioxygenase 1 (Ido1) was down-regulated in early-onset PE compared with late-onset PE. The tryptophan-degrading enzyme IDO1 has a regulatory effect on T-cells. By inhibiting T-cell proliferation, IDO1 mediates maternal tolerance of the allogenic fetus (Jalili et al., 2007). In contrast to Kudo et al. (2003) and Santoso et al. (2002), we did not see a decrease of Ido1 mRNA or protein in term pre-eclamptic placentae. Instead, the decreased mRNA expression in early compared with late-onset PE seems to be a physiological effect of gestational age, because there was no difference in Ido1 in the PE groups compared with their respective gestational age-matched controls (Fig. 1). Supporting this interpretation is the trend towards a difference in Ido1 expression between early and late control placentae (P = 0.055). The difference in mRNA expression found in Roundabout homolog 4, magic roundabout (Robo4), between early- and late-onset PE (Fig. 1) also seems to be gestational age-related, because there was similar mRNA expression in PE and control groups (Fig. 1).

The transcript AL365412 is a non-coding RNA close to or part of the non-coding PLAC2 gene on chromosome 19 and was up-regulated in early-onset PE (Table III). Epigenetic regulation has been suggested to be important in PE pathogenesis (Chelbi and Vaiman, 2008) and our finding indicates that a non-coding RNA seems to be functionally involved in the pathogenesis of early-onset PE.

In previous work, we have shown that PE is associated with over-expression of the haemoglobin genes alpha2 and gamma and accumulation of the protein in the vascular lumen of the placenta (Centlow et al., 2008). In addition, the mean plasma concentrations of both fetal and adult haemoglobin are significantly increased in pre-eclamptic women and the levels of total plasma haemoglobin correlate strongly with systolic blood pressure (Olsson et al., 2010). Based on these results, we have hypothesized that cell-free haemoglobin leaks from the placenta into the maternal circulation and contributes to the endothelial damage and symptoms of PE by inducing oxidative stress. Therefore, an interesting finding in the present study is that genes associated with oxygen transport and haem binding were transcriptionally altered in early- compared with late-onset PE.

There are limitations in the design of this study that deserve attention. First, the mode of delivery differed between the two PE groups. There are conflicting data regarding the effect of labour on placental gene expression. Cindrova-Davies et al. reported that labour is a powerful inducer of placental oxidative stress, inflammatory cytokines and angiogenic regulators, and they related these to intermittent perfusion of the placenta during delivery (Cindrova-Davies et al., 2007). In contrast, Sitras et al. concluded that gene expression in near-term placenta is not altered by labour (Sitras et al., 2008). In our study, we assumed that the effect of labour on placental gene expression is not of major importance. Whether this is correct or not is a matter of discussion.

Secondly, the physiological gene expression pattern changes during placental development. For example, genes involved in regulation of the cell cycle are up-regulated in the first trimester versus term placentae, whereas genes involved in cell communication are down-regulated (Mikheev et al., 2008). Therefore, the differences in gene expression we have found in this study may partly reflect normal gestational age-related discrepancies rather than differences related to the PE subgroup. However, it has been reported that gestational age-dependent gene regulation is most pronounced in the first trimester, and GO groups revealed by bioinformatics in our experiment do not match GO groups previously reported to differ between second trimester and term placentae (Mikheev et al., 2008). Taken together, these facts suggest that the GO groups found in our study are related to the PE subgroup and not to changes related to gestational age. In a more extensive microarray study, the inclusion of both early and late control placentae would possibly give deeper insight into the divergence of pathology between early- and late-onset PE.

In the present study, we included early and late control placentae in the verification step of four selected genes (Fig. 1). A limitation of this design is related to the known association between preterm delivery and abnormal placentation (Arias et al., 1993), and data obtained from early control placentae may therefore not reflect normal gene expression at this gestational age. Nevertheless, in order to discover PE-related changes, gestational age-matched placental samples were the most appropriate controls that we could obtain and we believe that they can be used with some caution. Furthermore, gestational age at delivery was slightly longer among women with early-onset PE than their controls (Table II). This must be taken into account when interpreting the results, since the differences in gene expression seen between early controls and early-onset PE may in part be due to variations in gestational age-related placental gene expression. Placental gene expression in the four studied genes was higher in term than in pre-term pregnancies. One could therefore speculate that expression of these genes increases during pregnancy. Since the early control group delivered at an earlier gestational age than the early-onset PE group, the differences in expression of Egfl7 and Acvrl1 between the two groups may in fact be underestimated.

Gene expression can vary in different parts of the placenta. For instance, VEGF expression is enhanced in the subchorionic lateral border of the placenta compared with the medial basal area (Wyatt et al., 2005). The samples in the present study were collected at two different clinics, with slightly different protocols for sample collection. However, placental biopsy samples were always taken from a central part of the placenta in order to minimize the risk of expression variation due to sampling site.

In summary, differences in gene expression profiles between early- and late-onset PE support the concept that they are at least partly two different entities. Differences in the expression of angiogenesis-associated genes strengthen the theory of impaired angiogenesis being part of the aetiology of early-onset PE. We have shown that the two angiogenesis-associated genes Egfl7 and Acvrl1 are down-regulated in placentae from women with early-onset PE versus late-onset PE and versus gestational age-matched control placentae. The specific roles of these two genes in the aetiology of PE remain to be studied.

Authors' roles

K.J. wrote the article, and contributed substantially to design, acquisition of data, analysis and interpretation of data. M.C. contributed substantially to conception and design, acquisition of data, and analysis and interpretation of data. A.-K.W. and I.L. contributed substantially to acquisition of data. S.R.H. and M.O. contributed substantially to conception and design, and analysis and interpretation of data. M.C., A.-K.W., I.L., S.R.H. and M.O. contributed to revising the article for important intellectual content. All authors approved the final manuscript.

Funding

This work was supported by the General Maternity Hospital Foundation, the Uppsala-Örebro Regional Research Council, and the Swedish Research Council (projects 8683).