Metabolic gene profile in early human fetal heart development

Abstract

The primitive cardiac tube starts beating 6–8 weeks post fertilization in the developing embryo. In order to describe normal cardiac development during late first and early second trimester in human fetuses this study used microarray and pathways analysis and created a corresponding ‘normal’ database. Fourteen fetal hearts from human fetuses between 10 and 18 weeks of gestational age (GA) were prospectively collected at the time of elective termination of pregnancy. RNA from recovered tissues was used for transcriptome analysis with Affymetrix 1.0 ST microarray chip. From the amassed data we investigated differences in cardiac development within the 10–18 GA period dividing the sample by GA in three groups: 10–12 (H1), 13–15 (H2) and 16–18 (H3) weeks. A fold change of 2 or above adjusted for a false discovery rate of 5% was used as initial cutoff to determine differential gene expression for individual genes. Test for enrichment to identify functional groups was carried out using the Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG). Array analysis correctly identified the cardiac specific genes, and transcripts reported to be differentially expressed were confirmed by qRT–PCR. Single transcript and Ontology analysis showed first trimester heart expression of myosin-related genes to be up-regulated >5-fold compared with second trimester heart. In contrast the second trimester hearts showed further gestation-related increases in many genes involved in energy production and cardiac remodeling. In conclusion, fetal heart development during the first trimester was dominated by heart-specific genes coding for myocardial development and differentiation. During the second trimester, transcripts related to energy generation and cardiomyocyte communication for contractile coordination/proliferation were more dominant. Transcripts related to fatty acid metabolism can be seen as early as 10 weeks and clearly increase as the heart matures. Retinol receptor and gamma-aminobutyric acid (GABA) receptor transcripts were detected, and have not been described previously in human fetal heart during this period. For the first time global gene expression of heart has been described in human samples to create a database of normal development to understand and compare with known abnormal fetal heart development.

Introduction

Six to eight weeks after fertilization and toward the end of the embryonic period, the primitive cardiac tube, with four chambers, starts beating. At this time, the remaining cardiac stem cells became differentiated cardiomyocytes. Through the remaining pregnancy period, the cardiomyocytes then undergo differentiation and specialization into specific cardiac tissues (Wheater et al., 1987; Moore, 1989; Gratacos et al., 2007). Due to the obvious difficulty in obtaining human cardiac tissue, most of the literature in early cardiac profiling is based on animal models, such as mice and sheep. Studies looking for common pathways that may explain cardiac failure have been performed in human fetal hearts compared with failing adult heart profiles (Razeghi et al., 2001), but such studies were limited to the 18- to 28-week period of gestation.

Previous cardiac development studies have elucidated, in part, the metabolic profile of the early human heart. Fetal heart energy production has been described as dependent on glucose metabolism during the whole fetal life, which changes to fatty acid oxidation during the first weeks of the neonatal period (Lopaschuk and Jaswal, 2010). The adult heart in turn depends on fatty acid oxidation for energy generation. Previous experiments using mouse models have also assessed the transition in energy profile from embryonic stem cells to fetal cardiomyocyte during the first stages of pregnancy. Cardiac stem cells appear to obtain energy from anaerobic metabolism that transforms into a mitochondria-mediated oxidative metabolism to meet increases in energy demand during cardiomyocyte differentiation and development during fetal life (Chung et al., 2007; Lopaschuk and Jaswal, 2010). No data are available to verify if this is also true in humans.

The regulation of fetal cardiomyocyte growth is essential since the number of cardiac cells at birth predetermines the number of cardiomyocytes for life, which is a predictor for cardiac disease (Thornburg et al., 2010). Studies in sheep have shown that proliferation of cardiomyocytes is stimulated by angiotensin II, cortisol and insulin-like growth factor-1 in the womb (Thornburg et al., 2010). Thus, in order to understand cardiac disease in humans, it is imperative to study if these same endocrine systems are present in early fetal life.

In the current study, the gene expression profile of human fetal heart between 10 and 18 weeks' gestation is presented. The working hypothesis was that there is a developmental change in gene profile between fetal hearts in the late first trimester (10–12 weeks' gestation) compared with the early second trimester (16–18 weeks' gestation), which would be detectable by transcript analysis and/or pathway analysis. To this end we have analyzed fetal heart mRNA using Affymetrix arrays in an attempt to both investigate changes in gene expression in human fetal heart and provide a much needed reference database to guide future studies in normal and abnormal human fetal development.

Methods

Tissue collection and RNA isolation

Human fetal cardiac tissue was collected after informed consent following institutional guidelines, from patients older than 18 years of age undergoing an elective dilation and curettage for termination of pregnancy due to unwanted pregnancy between 10 and 18 weeks of gestational age. Patients had already undergone the surgical consent process with their health team before they were approached to donate fetal tissue samples under an IRB approved protocol. Those who expressed willingness for such research tissue donation then underwent a research informed consent process. All pregnancies had a dating ultrasound done within a week of the procedure. The suction was generated by a vacuum with a standard suction unit providing variable controlled suction up to 550 mmHg, and a suction flow rate of 30 lpm.

Before the procedure, collection bottles were placed at 4°C for 1 h. At the time of the procedure, 100 ml (at 4°C), RNAse-free water solution (BP561-1 Water, Sterile for RNA work, Fisher Scientific, USA) mixed with 10× PBS buffer (Ambion, USA) at a final concentration of 10%, was added to the collecting bottle.

The procedure was performed within 5 min of adding the cold water. After the procedure, the heart was grasped with sterilized surgical instruments and carefully isolated from the rest of the thoracic content to be placed in RNAse-free plastic tubes. Fetal brain was also collected in the same fashion. No karyotype was done to confirm normal chromosomes. The brain tissue was used for comparison looking for major differences among the study group and the brain group as an internal validation.

These tubes were immediately placed in liquid nitrogen for transport. The tissue was then stored in a −80°C freezer.

Tissue RNA was extracted using RNA STAT-60 reagents, as recommended by the manufacturer (Tel-Test, Inc., Friendswood, TX, USA). All procedures were run on ice. A starting ratio of 1 ml reagent/100 mg tissue was used to ensure efficient, intact RNA recovery. RNA concentration and purity were initially determined using spectrophotometric A₂₆₀ and A_260/280 ratios (NanoDrop Products, Thermo Fisher Scientific, Inc., Wilmington, DE, USA). RNA quantification and quality control were performed using the Agilent Bioanalyzer (Agilent Technology, Santa Clara, CA, USA) 260/280 ratios and standard RIN-values (RNA integrity number). A value of 1.7–2.1 was considered workable for further testing.

Array preparation

Fourteen hearts RNA samples were divided into three Affymetrix ST 1.0 arrays according to gestational age. Array #1 analyzed six hearts RNA samples from 10 to 12 weeks' gestation. Array #2 analyzed six hearts RNA samples from 13 to 15 weeks' gestation, and Array #3 analyzed two hearts RNA samples from 16 to 18 weeks' gestation. Ten brain samples from the same fetuses were used as internal controls. cDNA was synthesized from extracted RNA (300 ng) using the Ambion WT Expression Kit (Expression Kit Manual P/N 4425209 Rev. C). cDNA yield quantity and quality were assessed using NanoDrop Spectrophotometer (NanoDrop Products, Thermo Fisher Scientific, Inc., Wilmington, DE, USA) and Agilent Bioanalyzer (Agilent Technology, Santa Clara, CA, USA). Extraction RNA samples with an A_260/280 ratio of 1.7–2.1 were further analyzed by gel electrophoresis at the Biotech Center. Samples were considered acceptable for testing when 28S and 18S rRNA bands resolved into two discrete bands that had no significant smearing below each band. The 28S rRNA band had intensity approximately twice that of the 18S rRNA band. These samples were then further tested by cDNA generation for hybridization. cDNA was checked for yield and size distribution using NanoDrop Spectrophotometer (NanoDrop Products, Thermo Fisher Scientific, Inc., Wilmington, DE, USA) and Agilent profile (Agilent Bioanalyzer, Agilent Technology, Santa Clara, CA, USA). Fragmentation check revealed a cDNA <500 bp reliable for hybridization to Human Gene 1.0 ST arrays. All subsequent experiments were then undertaken using RNA samples matching these requirements.

Fragmentation and end-terminus labeling of cDNAs were done at the Biotechnology Center of the University of Wisconsin, Madison, using the Affymetrix WT Terminal Labeling Kit [see GeneChip^® WT Terminal Labeling and Hybridization User Manual (P/N 702808 Rev. 4)]. The samples were then hybridized to Human Gene 1.0 ST arrays at 45°C for 16 h O/N, following all procedures outlined in the GeneChip^® WT Terminal Labeling and Hybridization User Manual (P/N 702808 Rev. 4) for Human Gene 1.0 ST Arrays. GeneChips post-processed were done on the AFX Fluidics 450 Station, according to all AFX protocols and procedures defined for the Human Gene 1.0 ST Array (FS450_0007), as outlined in the GeneChip Expression Wash, Stain and Scan User Manual (P/N 702731 Rev. 3). GeneChips were scanned on the GC3000 G7 Scanner. Data were extracted and processed using Affymetrix Command Console version 3.1.1.1229.

Heart tissue was compared with brain tissue looking for organ-specific genes. The heart's top signals detected were Myosin light chain 7 (MYL7), Myosin heavy chain 7 (MYH7), T-box 20 and Popeye domain which are clearly appropriate to the heart. The brain's top signals detected were Internexin, Septin, Doublecortin, Stathmin and NEL2 which are appropriate to the heart.

Brain tissue was compared with heart tissue looking for organ-specific genes. In the brain, the top signals detected are summarized in Fig. 1B. Note the abundant expression of Internexin, Septin, Doublecortin, Stathmin and NEL2, which are all clearly appropriate.

Figure 1.

(A) Genes with the highest differential expression in the heart. This figure is based on pooled tissue samples from 14 fetal hearts from 10 to 18 weeks of gestational age for genes with 20-fold change or greater expression level. Human Gene 1.0 ST array (Affymetrix) was used. Genes were chosen based on fold change compared with mRNA, after confirming a differential expression >95% with an false discover rate (FDR) of 5%. * = genes verified by qRT–PCR. (B) Genes with the highest differential expression in the brain. This figure is based on pooled tissue samples from 10 fetal brains from 10 to 18 weeks of gestational age for genes with 20-fold change or greater expression level. Human Gene 1.0 ST array (Affymetrix) was used. Genes were chosen based on fold change compared with mRNA, after confirming a differential expression >95% with an FDR of 5%. * = genes verified by qRT–PCR. (C) Differentially expressed genes measured by fold change in mRNA levels between the hearts at different gestational ages. The late first trimester hearts (H1) showed an increased number of differentially expressed genes (green bars) when compared with the early second trimester hearts (H3). Tissue samples correspond to hearts from 10–12 weeks of gestational age (H1) and 16–18 weeks of gestational age (H3). Differences are measured by fold change (>2), after adjusting for an FDR of 5% with a differential expression >95% using EBarrays (Newton et al., 2001; Kendziorski et al., 2003), to identify differentially expressed genes in any two conditions. (D) Genes with the highest differential expression in the late first trimester hearts (H1). Tissue samples correspond to hearts from 10–12 weeks of gestational age (H1) and 16–18 weeks of gestational age (H3) pooled on two separate Human Gene 1.0 ST array, according to gestational age. Differences are measured by fold change (>2), after adjusting for an FDR of 5% with a differential expression >95% using EBarrays (Newton et al., 2001; Kendziorski et al., 2003), to identify differentially expressed genes in any two conditions. * = gene verified by qRT–PCR.

The array data can be found at GEO; the accession number is GSE56048.

Statistical methods

All analyses of GeneChip data were carried out in R (R Development Core Team, 2007), a publicly available statistical analysis environment (available at www.r-project.org). Specific software packages (affy, EBarrays, allez) are available at Bioconductor (http://bioconductor.org/), an online suite of tools for the analysis of genomic data (Huber and Gentleman, 2004). The Affymetrix probe level data were processed using Robust Multi-Array Analysis (RMA) as implemented in the affy package, in order to obtain normalized summary expression scores for each probe set on each array (Irizarry et al., 2003). Cutoffs for sensitivity, ensuring a reliable positive detection result, were 0.95. Cutoffs for a significant change in expression were initially set at 2-fold.

Initial pools of the brain and heart derived from RNA were prepared for validation using methods allowing for verification of the characteristics of both the brain and heart. Results were those expected in the brain and heart RNA.

In order to draw any conclusion, parameters had to first be established as to how the output chips would be analyzed. In this presentation, the probe sets are referred to as genes. RMA fits a linear model to the log probe intensities for each probe set. The linear model includes a sample effect (the parameter of interest), a probe effect and an error term. The standard errors are normalized so that each probe set has the same median standard error across arrays.

EBarrays was used to identify differentially expressed (DE) genes (Newton et al., 2001; Kendziorski et al., 2003). EBarrays is an empirical Bayes approach which models the probability distribution of a set of expression measurements. It accounts generally for differences in specific genes with regard to their true underlying expression levels, measurement fluctuations and distinct expression patterns for a given gene among conditions (Kendziorski et al., 2003). An expression pattern is an arrangement of the true underlying intensities (μ) in each condition. The number of patterns possible depends on the number of conditions from which the expression measurements were obtained. For example, when measurements are taken from two conditions, two patterns of expression are possible; equivalent expression (!μ1 = μ2) and differential expression (!μ1 ≠ μ2). Since there is no priori stating which genes are in which patterns, the marginal distribution of the data is a mixture of all possible patterns with model parameters being determined by the full set of array data. This approach utilizes information across a set of arrays optimizing model fit and is, therefore, more efficient than a number of methods that make gene inferences one gene at a time (Kendziorski et al., 2003). Fitted model parameters provide information on the number of genes expected in each expression pattern. Furthermore, the fitted model is used to assign probability distributions to every gene. Each gene-specific distribution gives the posterior probability of that gene's individual expression pattern.

Thresholds were chosen to target an FDR of 5% when more than one sample was available within the condition (e.g. heart versus brain), the lognormal normal with moderated variance (LNNMV) model was used in EBarrays; otherwise, a lognormal normal model was used.

EBarrays then provides the posterior probability of differential expression (DE) as well as equivalent expression (EE) for every gene (Table I). A gene was defined as DE if its posterior probability of DE exceeded 0.95. This is analogous to its posterior probability of EE being ≤0.05. This threshold targets an overall false discovery rate (FDR) of 5%. In some cases (specified in the text), a second fold-change based filter was applied to the list of DE genes obtained from EBarrays.

Table I.

Transcripts detected by array analysis with a fold change >5.

Entrez ID	Gene	Gene.Name	PP.LNNMV. EE	PP.LNNMV. DE	FC
253017	TECRL	trans-2,3-enoyl-CoA reductase-like	0	1	63
58498	MYL7	Myosin, light chain 7, regulatory	0	1	56
64091	POPDC2	Popeye domain containing 2	0	1	46
57057	TBX20	T-box 20	0	1	46
4878	NPPA	Natriuretic peptide precursor A	0	1	43
5350	PLN	phospholamban	0	1	38
4625	MYH7	Myosin, heavy chain 7, cardiac muscle, beta	0	1	36
7137	TNNI3	Troponin I type 3 (cardiac)	0	1	30
115701	ALPK2	alpha-kinase 2	0	1	28
4888	NPY6R	Neuropeptide Y receptor Y6 (pseudogene)	0	1	25
4633	MYL2	Myosin, light chain 2, regulatory, cardiac, slow	0	1	24
28965	SLC27A6	Solute carrier family 27 (fatty acid transporter), member 6	0	1	23
7134	TNNC1	Troponin C type 1 (slow)	0	1	21
4634	MYL3	Myosin, light chain 3, alkali; ventricular, skeletal, slow	0	1	21
137835	TMEM71	Transmembrane protein 71	0	1	20
51086	TNNI3K	TNNI3 interacting kinase	0	1	19
10699	CORIN	Corin, serine peptidase	0	1	19
4624	MYH6	Myosin, heavy chain 6, cardiac muscle, alpha	0	1	18
93649	MYOCD	Myocardin	0	1	17
4607	MYBPC3	Myosin binding protein C, cardiac	0	1	17
27063	ANKRD1	Ankyrin repeat domain 1 (cardiac muscle)	0	1	16
27129	HSPB7	Heat shock 27 kDa protein family, member 7 (cardiovascular)	0	1	16
7139	TNNT2	Troponin T type 2 (cardiac)	0	1	16
150572	SMYD1	SET and MYND domain containing 1	0	1	15
4151	MB	Myoglobin	0	1	14
51725	FBXO40	F-box protein 40	0	1	14
2170	FABP3	Fatty acid binding protein 3, muscle and heart (mammary-derived growth inhibitor)	0	1	13
376132	LRRC10	Leucine rich repeat containing 10	0	1	13
145781	GCOM1	GRINL1A complex locus	0	1	13
6523	SLC5A1	Solute carrier family 5 (sodium/glucose cotransporter), member 1	0	1	12
23676	SMPX	Small muscle protein, X-linked	0	1	12
29119	CTNNA3	Catenin (cadherin-associated protein), alpha 3	0	1	12
8048	CSRP3	Cysteine and glycine-rich protein 3 (cardiac LIM protein)	0	1	12
8736	MYOM1	Myomesin 1, 185 kDa	0	1	11
5318	PKP2	Plakophilin 2	0	1	11
1160	CKMT2	Creatine kinase, mitochondrial 2 (sarcomeric)	0	1	11
11149	BVES	Blood vessel epicardial substance	0	1	11
10529	NEBL	nebulette	0	1	11
2274	FHL2	Four and a half LIM domains 2	0	1	10
23493	HEY2	Hairy/enhancer-of-split related with YRPW motif 2	0	1	10
285025	CCDC141	Coiled-coil domain containing 141	0	1	10
27302	BMP10	Bone morphogenetic protein 10	0	1	10
64208	POPDC3	Popeye domain containing 3	0	1	9
221662	RBM24	RNA binding motif protein 24	0	1	8
10611	PDLIM5	PDZ and LIM domain 5	0	1	8
146862	UNC45B	Unc-45 homolog B (C. elegans)	0	1	8
1410	CRYAB	Crystallin, alpha B	0	1	8
6262	RYR2	Ryanodine receptor 2 (cardiac)	0	1	8
123722	FSD2	Fibronectin type III and SPRY domain containing 2	0	1	8
8988	HSPB3	Heat shock 27 kDa protein 3	0	1	8
7135	TNNI1	Troponin I type 1 (skeletal, slow)	0	1	7
23566	LPAR3	Lysophosphatidic acid receptor 3	0	1	7
6331	SCN5A	Sodium channel, voltage-gated, type V, alpha subunit	0	1	7
27145	FILIP1	Filamin A interacting protein 1	0	1	7
167681	PRSS35	Protease, serine, 35	0	1	7
91624	NEXN	Nexilin (F actin binding protein)	0	1	7
84803	AGPAT9	1-acylglycerol-3-phosphate O-acyltransferase 9	0	1	7
11155	LDB3	LIM domain binding 3	0	1	7
116496	FAM129A	Family with sequence similarity 129, member A	0	1	7
1E+08	OCR1	Ovarian cancer-related protein 1	0	1	7
6450	SH3BGR	SH3 domain binding glutamic acid-rich protein	0	1	7
59272	ACE2	Angiotensin I converting enzyme (peptidyl-dipeptidase A) 2	0	1	7
91807	MYLK3	Myosin light chain kinase 3	0	1	7
2201	FBN2	Fibrillin 2	0	1	6
140456	ASB11	Ankyrin repeat and SOCS box-containing 11	0	1	6
6546	SLC8A1	Solute carrier family 8 (sodium/calcium exchanger), member 1	0	1	6
NA	NA	NA	0	1	6
NA	NA	NA	0	1	6
8470	SORBS2	Sorbin and SH3 domain containing 2	0	1	6
6910	TBX5	T-box 5	0	1	6
84675	TRIM55	Tripartite motif-containing 55	0	1	6
116362	RBP7	Retinol binding protein 7, cellular	0	1	6
84940	CORO6	Coronin 6	0	1	6
1573	CYP2J2	Cytochrome P450, family 2, subfamily J, polypeptide 2	0	1	6
144100	PLEKHA7	Pleckstrin homology domain containing, family A member 7	0	1	6
84676	TRIM63	Tripartite motif-containing 63	0	1	6
4286	MITF	Microphthalmia-associated transcription factor	0	1	6
345557	PLCXD3	Phosphatidylinositol-specific phospholipase C, X domain containing 3	0	1	6
9200	PTPLA	Protein tyrosine phosphatase-like (proline instead of catalytic arginine), member A	0	1	6
26548	ITGB1BP2	Integrin beta 1 binding protein (melusin) 2	0	1	6
88	ACTN2	Actinin, alpha 2	0	1	6
1832	DSP	Desmoplakin	0	1	6
8490	RGS5	Regulator of G-protein signaling 5	0	1	6
NA	NA	NA	0	1	5
10398	MYL9	Myosin, light chain 9, regulatory	0	1	5
283358	B4GALNT3	beta-1,4-N-acetyl-galactosaminyl transferase 3	0	1	5
161247	FITM1	Fat storage-inducing transmembrane protein 1	0	1	5
387700	SLC16A12	Solute carrier family 16, member 12 (monocarboxylic acid transporter 12)	0	1	5
11030	RBPMS	RNA binding protein with multiple splicing	0	1	5
90523	C6orf142	Chromosome 6 open reading frame 142	0	1	5
5322	PLA2G5	Phospholipase A2, group V	0	1	5
2702	GJA5	Gap junction protein, alpha 5, 40 kDa	0	1	5
51705	EMCN	endomucin	0	1	5
8395	PIP5K1B	Phosphatidylinositol-4-phosphate 5-kinase, type I, beta	0	1	5
606553	C8orf49	Chromosome 8 open reading frame 49	0	1	5
442721	LMOD2	Leiomodin 2 (cardiac)	0	1	5
358	AQP1	Aquaporin 1 (Colton blood group)	0	1	5

Tests for enrichment of common function among sets of differentially expressed genes were carried out using data from the Gene Ontology (GO) annotations and the Kyoto Encyclopedia of Genes and Genomes (KEGG). In GO, transcripts genes are categorized at varying levels of biological detail. The three broadest levels are molecular function, cellular component and biological process. There are many subcategories within each detail. The R package allez was used to perform tests of enrichment for each GO category and KEGG pathway (Newton et al., 2001). In general, the interpretation of P-values resulting from enrichment tests is not straightforward due to the many dependent hypotheses tested. Furthermore, the enrichment test tends to result in small P-values when groups with few transcripts are considered. The statistical methods underlying allez adjust for these factors, allowing increased power and sensitivity for identifying sets that are biologically meaningful (Tables II and III).

Table II.

Genontology classification.

	GO	GO.term	n.genes	z.score
GO:0042627	CC	Chylomicron	12	12.01
GO:0010886	BP	Positive regulation of cholesterol storage	6	11.26
GO:0008307	MF	Structural constituent of muscle	43	10.29
GO:0032982	CC	Myosin filament	18	10.08
GO:0042159	BP	Lipoprotein catabolic process	6	9.96
GO:0033033	BP	Negative regulation of myeloid cell apoptosis	6	9.92
GO:0008035	MF	High-density lipoprotein binding	7	9.81
GO:0034361	CC	Very-low-density lipoprotein particle	20	9.09
GO:0034385	CC	Triglyceride-rich lipoprotein particle	20	9.09
GO:0031983	CC	Vesicle lumen	38	8.64

Table III.

Geontology classification by heart group.

H1			H3
GO:0008307	Structural constituent of muscle	43	GO:0042627	Chylomicron	12
GO:0032982	Myosin filament	18	GO:0010886	Positive regulation of cholesterol storage	6
GO:0033033	Negative regulation of myeloid cell apoptosis	6	GO:0042159	Lipoprotein catabolic process	6
			GO:0008035	High-density lipoprotein binding	7
			GO:0034361	Very-low-density lipoprotein particle	20
			GO:0034385	Triglyceride-rich lipoprotein particle	20
			GO:0031983	Vesicle lumen	38
	Total genes	67		Total genes	109

H1: 10-12wks; H3: 16-18wks.

qRT–PCR

Seven selected genes (FABP4, MYH7B, APOA2, TNNI3, TMOD2, NRXN1, RPLO) were further validated by quantitative real-time PCR (qRT–PCR) in order to validate the array data. Human reference cDNA, oligo-dT primed (Clontech) was used to create a standard dilution series for analysis in parallel with samples on a Roche LightCycler480 using sequence-specific primers. Each assay per sample was performed in triplicate on 96-well plates with 480 Sybr Green I Master (Roche), and 500 nM forward and reverse primers. Cycling conditions were as follows: 10 min at 95°C; 40 cycles at 95°C for 10 s, 57°C for 10 s and 72°C for 5 s; followed by melt curve analysis.

Ethical approval

Sample collection was started after receiving Institutional Review Board approval (H-2009-0071) over a 2-year period from 2009 to 2011.

Results

Confirmation of cardiac tissue

Pooled RNA from all cardiac tissue (10–18 weeks) was compared with a control tissue (fetal brain) to confirm specificity of the tissue. Genes like natriuretic peptide (NPPA), troponin (TNNI3) and myosin heavy chain 7 (MYH7) cardiac muscle were readily detected (differential expression >95%, after adjusting for an FDR of 5%) and correctly reported as ‘up-regulated’ relative to brain tissue (Fig. 1A). Subsequent qRT–PCR confirmed expression of TNNI3 and MYH7. Brain tissue genes are shown in Fig. 1B (Iruretagoyena et al., 2014).

In order to simplify the analysis of the data, samples were then subgrouped by gestational age within 2-week periods. H1 (10–12 weeks), H2 (13–15 weeks) and H3 (16–18 weeks). Given the magnitude of the data set, statistical analysis was done comparing groups H1 versus H3 (H2 group is shown in the figures to further illustrate the normal progression from first trimester to second trimester).

Changes according to gestational age on global expression of genes

When fold changes of 2, 5 and above 20 were used to compared H1 to H3 gene expression, there were 43, 13 and 5 individual gene transcripts specifically up-regulated, respectively. Conversely, when H3 was compared with H1 24, 4, 0 genes were up-regulated, respectively (Fig. 1C).

Heart contractile proteins

MYL1, MYH3, MYBPC1, SLN and ACTG2 genes coding for sarcoplasmic reticulum, actin and essential myosin contractile protein constituents of cardiac muscle were more highly expressed (differential expression >95%, after adjusting for an FDR of 5%) in hearts of a younger gestational age (Figs 1D and 2A). When gene ontology analysis was applied, terms for structural constituents of muscle and myosin filament were revealed, further supporting single transcript analysis (Fig. 2B). qRT–PCR analysis was also confirmatory when MYH7 was used as a representative target.

Figure 2.

(A) mRNA gene expression of regulators of cardiomyocyte differentiation. All transcripts are increased in the late first trimester (H1) versus the early second trimester (H3). Differences are measured by mRNA level on genes with a differential expression >95%, after adjusting for an FDR of 5% using EBarrays (Newton et al., 2001; Kendziorski et al., 2003), to identify differentially expressed genes in any two conditions. SLN, Sarcolipin; MYL1, Myosin light chain 1; MYH3, Myosin heavy chain 3; ACTG2, Actin gamma. * = gene verified by qRT–PCR. (B) Bar chart showing terms from Gene Ontology in the heart, classified according to biological process, cellular component and molecular function using test for enrichment of common functions among sets of differentially expressed genes (Gene Ontology Annotations and the KEGG). Gene families involved in common functions for muscular development like myosin filament and fatty acid metabolism including chylomicron, cholesterol, lipoprotein binding proteins and triglyceride lipoprotein are the highest ranked in the cardiomyocytes from 10 to 18 weeks of gestational age (pooled data). (C) Genes with the highest differential expression in the early second trimester hearts (H3). Tissue samples correspond to hearts from 10–12 weeks of gestational age (H1) and 16–18 weeks of gestational age (H3) pooled on two separate Human Gene 1.0 ST array, according to gestational age. Differences are measured by fold change (>2), after adjusting for an FDR of 5% with a differential expression >95% using EBarrays (Newton et al., 2001; Kendziorski et al., 2003), to identify differentially expressed genes in any two conditions. *= gene verified by qRT–PCR.

Energy generation metabolism in the heart

Individual transcript analysis showed that genes coding for FABP4, FABP2, NRAP and ACE2, all part of either cardiac fatty acid metabolism or structural remodeling, were among the highest expressed in H3 hearts. (Figs 2C and 3A). Gene ontology analysis showed ontology terms in agreement with single transcript finding involving fatty acid metabolism including chylomicron, positive regulation of cholesterol, lipoprotein catabolic process, high and very low lipoprotein and triglyceride-rich lipoprotein (Fig. 2B). qRT–PCR confirmed expression of FABP4 and APOA2.

Figure 3.

(A) mRNA gene expression of regulators of the cardiac fatty acid metabolism and remodeling. All transcripts are increased in the early second trimester as the heart matures, compared with the late first trimester. Differences are measured by mRNA level on genes with a differential expression >95%, after adjusting for an FDR of 5% using EBarrays (Newton et al., 2001; Kendziorski et al., 2003), to identify differentially expressed genes in any two conditions. *= genes verified by qRT–PCR. (B) mRNA expression of regulators of glucose energy generation by the cardiomyocytes. All transcripts are high in both groups showing no significant differential expression (<0.005), after adjusting for an FDR of 5%. (C) mRNA gene expression of insulin growth factor and angiotensin. Specific transcripts were examined due to the their suggested importance in previous studies (see introduction). mRNA levels detected in each group are shown. Differential expression >95%, after adjusting for an FDR of 5% using EBarrays (Newton et al., 2001; Kendziorski et al., 2003), was not achieved suggesting that these genes are not differentially controlled across early gestation. (D) Reference genes in the heart. The graph shows mRNA expression of two candidate genes without significant change in expression, regardless of the experimental group (H1, H2 or H3). Reference genes were validated by the method of Stern-Straeter et al. (2009). All H1, H2 and H3 groups of hearts were analyzed under exact conditions. Differential expression was <0.005 for both genes under the two conditions. RPLP0: Ribosomal protein, large; TBS: TATA box binding protein.

Complementing our findings on transcripts for proteins involved in fatty acid metabolism as an energy source, single gene transcript analysis for genes involved in glucose energy production including PPARα, PPARβ, PD, LD, LDH-A, HK-1 and HIF-1 also was present but not differentially expressed between H1-H3 groups, after adjusting for an FDR of 5% (Fig. 3B). Thus expression of the mRNA encoding these pathways appears more constant with gestation.

Miscellaneous transcripts

GABRA4, GABRB1 and RBP7 genes were highly expressed in the H3 group, with a differential expression >95%, after adjusting for an FDR of 5% (Fig. 2C). GABRA4 or GABRA1 were not found in the gene ontology terms. Although RBP7 is not a direct participant in the ontology terms described, it is indirectly involved in the Chylomicron and lipoproteins terms as part of fatty acid metabolism.

Retinol receptors (RAR, RXR), angiotensin receptors 1-2, IGF1-2, IGFR1-2, IGFBP1 and glucocorticoid receptor NR3C1 were found to be present but not differentially expressed (differential expression <95%, after adjusting for an FDR of 5%) (Fig. 3C).

Housekeeping genes

For validation purposes, mRNA expression of reference genes RPLO and TBT (Stern-Straeter et al., 2009) did not show changes among the different groups as expected (Fig. 3D).

Discussion

Our findings have shown that there are characteristic dynamic changes in gene expression during early development in the human heart. While H1 heart gene expression profiles showed dominance by genes involved in myocardium development and differentiation, H3 hearts showed relatively increased expression of genes not only actively involved in energy generation, but also advanced cellular functions like cardiomyocyte communication and proliferation.

Heart contractile and remodeling proteins

A heart's ultimate destiny is myocardial contraction and so it is clearly of interest that various subtypes of myosin, the principal component of muscle, are highly and differentially expressed in H1 hearts compared with H3 hearts. Light and heavy chains of myosin and actin genes involved in the cytoskeleton formation dominate this period (Tsuchimochi et al., 1986; Yazaki et al., 1987). Genes like the Secreted Frizzled-Related Protein 2 (SFRP2), which function to inhibit fibroblast growth (He et al., 2010) and sarcolipin, which function to reduce the accumulation of calcium in the sarcoplasmic reticulum (Babu et al., 2007) are differentially expressed in the young hearts, potentially contributing to the right proportion of muscle and collagen, as well as calcium dynamics based in the findings of our study.

The renin-angiotensin system, specifically through ACE2 (angiotensin converting enzyme), is one of the pathways hypothesized to play an increased role in H3 hearts when compared with the H1 heart, based on differential gene expression revealed in our studies. The H3 heart expresses genes involved in advanced functions such as contractile coordination, angiogenesis and utilization of energy. Studies in ovine animal models have already shown angiotensin II acts in conjunction with insulin-like growth factor (IGF) to control proliferation of cardiomyocytes during the prenatal period through late gestation, when cardiomyocytes undergo terminal differentiation and proliferation diminishes (Reini et al., 2009). Data from our study further suggest the initiation of cardiomyocyte proliferation and remodeling in humans occurs around weeks 16–18 of gestation. Prior to this period, in the second trimester, more specifically at the end of the first trimester (10–12 weeks' gestation), the major events occurring in the cardiomyocyte relate more to the early differentiation of stem cells into active contractile cardiomyocytes.

Energy generation metabolism in the heart

Fetal cardiac metabolism has been characterized in animal models as dependent on glucose metabolism for energy from the very early stages of cardiac differentiation. This profile changes shortly after birth when the heart energy metabolism depends on fatty acid oxidation in the mitochondria, and persists into adulthood (Kong et al., 2008). Our findings using transcriptome analysis approaches in humans are entirely consistent with Kong et al. findings.

Single gene expression analysis showed H3 heart genes were differentially expressed, favoring fatty acid oxidation metabolism (FABP 4, FABP 2 and RBP) (Figs 2C and 3A) during the H3 period otherwise not apparent in the H1 hearts. Nonetheless this pathway was revealed when grouped analysis was performed. This may be interpreted as further evidence that an early sign of emerging fatty acid oxidation metabolism in the H1 hearts that is not significant when transcripts are evaluated individually. At a later gestational age, when single gene expression is robust enough to create a difference, such changes can be observed individually.

When gene ontology was further applied to a pooled sample of hearts from 10 to 18 weeks of gestational age (Tables II and III), genes involved in fatty acid metabolism either in cell components, molecular function or biological processes were again identified as the major active pathways. One remarkable finding was that genes driving metabolism of fatty acid were highly differentially expressed in the H3 hearts, compared with the H1 hearts. KEGG pathway analysis also showed pathways involved in fatty acid metabolism and oxidative stress controls (ascorbate and aldarate metabolism, retinol metabolism and PPAR signaling) are more dominant in the H3 period.

In contrast to our findings on fatty acid metabolism, key genes coding for HIF, HK1, LDH-A, and lactate dehydrogenase, as well as PPAR-alpha and beta were more uniformly present and highly expressed in both H1 and H3 hearts. The apparent lack of differential gene expression for genes involved in glucose metabolism is again consistent with previous findings in animal studies of the heart as being dependent on glucose utilization for basal energy generation during the fetal period (Lopaschuk and Jaswal, 2010). Combined, these observations may reflect the fact that the fetal period is dominated by glucose energy production until after birth, and no differential changes in the genes regulating this process should be expected from 10 to 18 weeks of gestation. Interestingly, however, three genes involved in fatty acid oxidation ranked among those with more than a 5-fold change in the H3 period compared with H1. These genes are the FABP 4, FABP 2 and RBP.

Other specific transcripts

Beyond energy metabolism, transcript analysis also revealed many other developmental changes of relevance. Retinol receptors are grouped in two families; the RXRs and the RARs (Mangelsdorf et al., 1992). RARs may function as receptors for thyroid hormones, vitamin D and peroxisome proliferators (PPAR) and co-operate with RXRs. Retinol and its receptors may play an important role in cardiac development through hormonal signaling and as cofactors in fatty acid oxidation for energy generation (Sucov et al., 1994). Although RXR presence has been documented in adult hearts, it has not been shown in the human fetal heart. In this study, experiments found the presence mRNA coding for both RXR and RAR in the human embryonic heart. There does not appear to be a difference in mRNA expression when compared by gestational age. However, when KEGG pathways are implemented, the retinol metabolism and PPAR signaling pathway appear to be among the main pathways activated in fetal hearts from 10 to 18 weeks. This finding may be linked, since retinol is a cofactor in many peroxidation reactions where PPAR is involved (Lopaschuk and Jaswal, 2010). Fatty acid metabolism and gluconeogenesis are major processes regulated by these pathways.

A further intriguing finding in the H3 heart was the higher expression of GABA receptors. These receptors have been extensively described in the brain and pre/post synaptic environment as neurotransmitters, but have not been described in the heart. Kang et al. (2011) recently showed an association between neurotransmitters as GABA and sympathoexcitation of the heart in rats with heart failure. It would be reasonable to associate the high expression of GABA receptors in the H3 heart with sympathetic innervation control of the heart and advanced functions such as heartbeat variability and blood pressure control.

Insulin growth factor-1 (IGF-1), cortisol receptor (NR3C1) and angiotensin receptors 1 and 2 (AGT1R, AGT2R), previously described by Thornburg et al. (2010) in sheep as important factors stimulating cardiomyocyte proliferation, were found to be present in human fetal heart as early as 10 weeks of gestation. These genes were not differentially expressed when comparison was made among groups based on gestational age. The mRNA expression though, showed that IGF-1 and IGFBP1 were present in early gestational age (10–12 weeks) as opposed to IGF-1 receptor that appeared to be mostly present at a later stage (16–18 weeks).

The angiotensin receptor 1 (AGT1R) mRNA expression appears to decrease with gestational age, the opposite of AGT2R which increases with gestational age. These changes may be related to the function of angiotensin II in cardiac remodeling during fetal life. AGT1R has been associated with vasoconstriction and hypertrophy while AGT2R has been associated with vasodilation counteracting AGT1R (Savoia and Volpe, 2011). IGF and AGTR may be related. IGF has been previously described (Thornburg, 2011) to inhibit the expression of AGT2R during hypoxic conditions in fetal rats. To our knowledge, none of these genes have been studied during early human fetal life.

Reference genes

In all array studies it is important to run negative controls. Interrogated housekeeping genes, or reference genes, were indeed found to be unchanged in both groups. While it was not clear in advance that this would be the case for all housekeeping genes, it is reassuring that this is the outcome, since it provides an internal quality control for this and indeed future analysis. Previous studies validating reference genes in myoblast development, utilizing different software, indicated that TBP (TATA box binding protein) and RPLP0 (ribosomal protein large P0) were the only two consistent genes that could be used to validate changes in other genes (Stern-Straeter et al., 2009).

In conclusion, these experiments using fetal human hearts spanning between 10 and 18 weeks of gestation confirmed previous findings in metabolic profile showing, during this period, that energy derived from glucose was the main source of fuel for the heart. Although the major source of energy is glucose, it is not the only source. Fatty acid metabolism appears to also be a source of energy and transcripts related to that metabolic pathway can be found to be present as early as 10 weeks of gestation. These genes involved in fatty acid metabolism thereafter show a stronger expression as the fetus develops, and this finding is consistent with our current knowledge showing a transition to a full utilization of fatty acid oxidation as the energy source in the neonatal period and adulthood. Many other individual transcript changes have also been identified that relate to the altered endocrine control of the heart and contractile proteins underlying its function. These findings not only serve as a database for comparison with older gestational ages and adults, but give a picture of the heart at a developmental stage not previously well characterized and yet critical to developmental success. It is tempting to hypothesize that, between the gestational ages of 10–18 weeks, the heart fully differentiates and starts specializing, remodeling and adjusting to finally become the hemodynamic controller. Further studies will be necessary to both validate these findings at a functional level and establish physiologic correlates suggested by this data set.

Authors' roles

J.I.I.: substantial contributions to conception and design, acquisition of data, analysis and interpretation of data, drafting the article and final approval of the version to be published. W.D.: substantial contributions to analysis and interpretation of data, drafting the article and final approval of the version to be published. C.B.: substantial contributions to acquisition of data, analysis and interpretation of data, drafting the article and final approval of the version to be published. J.O.: substantial contributions to analysis and interpretation of data, drafting the article and final approval of the version to be published. R.R.: substantial contributions to acquisition of data, drafting the article and final approval of the version to be published. A.T.B.: substantial contributions to analysis and interpretation of data, drafting the article and final approval of the version to be published. C.K.: substantial contributions to design, analysis and interpretation of data, drafting the article and final approval of the version to be published. S.S.B.D.: substantial contributions to analysis and interpretation of data, drafting the article and final approval of the version to be published. T.G.: substantial contributions to conception and design and interpretation of data, drafting the article and final approval of the version to be published. I.B.: substantial contributions to conception and design, analysis and interpretation of data, drafting the article and final approval of the version to be published. D.S.: substantial contributions to conception and design, acquisition of data, analysis and interpretation of data, drafting the article and final approval of the version to be published.

Funding

Maternal Fetal Medicine Research and Development Grant. Dept. OBGYN, University of Wisconsin.

Conflict of interest

None declared.