Umbilical cord blood samples were obtained from 12 pregnant women, six (age range, 35–43 years; mean age, 39.7 years) carrying fetuses with DS (two male and four female) and six (27–36 years, 31.3 years of mean age) carrying fetuses without DS (two male and four female) at a gestational age of 18–22 weeks. The participants were recruited at Shenzhen People's Hospital (Shenzhen, China) between January 2013 and December 2014. Diagnosis of fetal DS was performed by chromosomal examination. In addition, peripheral blood samples were obtained from 12 children, six with DS and six without DS (two males and four females; age range, 5–12 years, mean age, 8 years) at Guilin No. 924 Hospital (Guilin, China) between January 2015 and December 2016. The present study was approved by the Ethics Committees of Shenzhen People's Hospital (Shenzhen, China) and Guilin No. 924 Hospital (Guilin, China). Written informed consent was obtained from all pregnant women and the parents/guardians of the children.

Ultrasound-guided umbilical cord blood extraction was performed and a total of 3 ml extracted blood was placed in EDTA-containing anticoagulant tubes. Peripheral blood (2 ml) of 12 children was extracted with a needle and placed in EDTA-containing anticoagulant tubes. According to the single nuclear cell extraction protocol, 1 ml umbilical cord blood mononuclear cells (PBMCs) was extracted using lymphocyte separation solution (MD Pacific Biotechnology Co., Ltd.), mixed with 1 ml TRIzol^® reagent (Invitrogen; Thermo Fisher Scientific, Inc.), placed in cryopreservation tubes and stored at −80°C until further use.

The frozen PBMCs were thawed at room temperature and total RNA was extracted using TRIzol^® reagent according to the manufacturer's instructions. Total RNA was subsequently treated with DNase I to remove genomic DNA contamination. The purity and concentration of the RNA were determined using a spectrophotometer (NanoDrop ND-1000; Thermo Fisher Scientific, Inc.). RNA integrity was assessed using standard denaturing agarose gel electrophoresis (1%).

Total RNA from each sample was amplified and transcribed into fluorescent cRNA utilizing random primers according to the Arraystar Super RNA Labeling protocol (Arraystar, Inc.). The labelled cRNAs were hybridized onto Arraystar Human circRNA arrays (cat. no. 6×7K; Arraystar, Inc.) and incubated for 17 h at 65°C in an Agilent hybridization oven (Agilent Technologies, Inc.). After washing, slides were scanned with the Axon GenePix 4000B Scanner (Molecular Devices, LLC).

Scanned images were imported into GenePix Pro software (version 6.0; Molecular Devices, LLC) for grid alignment and raw data extraction. Quantile normalization of raw data and subsequent data processing were performed using the R software package heatmap.2 (version 3.2.0; http://mirrors.tuna.tsinghua.edu.cn/CRAN/). The circRNAs that at least 1 out of 2 samples had flagged as ‘expressed’ (greater than 2 times background standard deviation) were retained for further differential analyses. Differentially expressed circRNAs between umbilical cord blood samples obtained from pregnant woman carrying fetuses with DS and without DS were identified through fold-change (FC≥2.0 or ≤-2.0) or volcano plot filtering.

In the present study, sample labelling and array hybridization were performed according to the Agilent One-Color Microarray-Based Gene Expression Analysis protocol (Agilent Technologies, Inc.). Briefly, total RNA from each sample was linearly amplified and labelled with Cy3-UTP (Enzo Life Sciences, Inc.). The labelled cRNAs were purified using an RNeasy Mini kit (Qiagen, Inc.). The concentration and specific activity of the labelled cRNAs (pmol Cy3/µg cRNA) were measured using a spectrophotometer. A total of 1 µg of each labelled cRNA was fragmented by the addition of 11 µl 10X blocking agent (LMAI Bio) and 2.2 µl 25X fragmentation buffer (Agilent Technologies, Inc.), then heated at 60°C for 30 min. Subsequently, 55 µl 2X GEx Hybridization Buffer HI-RPM (Agilent Technologies, Inc.) were added to dilute the labelled cRNA and 100 µl hybridization solution (Agilent Technologies, Inc.) were dispensed into the gasket slide and assembled on the gene expression microarray slide. The slides were incubated for 17 h at 65°C in the aforementioned Agilent hybridization oven. The hybridized arrays were washed, fixed and scanned using the Agilent DNA Microarray Scanner (cat. no. G2505C; Agilent Technologies, Inc.).

The differentially expressed genes (FC≥2.0 or ≤-2.0) were selected using Agilent GeneSpring GX software (version 12.1; Agilent Technologies, Inc.). Then, hierarchical clustering was performed using scripts prepared by Aksomics Inc. Gene Ontology (GO; http://www.geneontology.org) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG; http://www.genome.jp/kegg/) pathway analysis were performed using standard enrichment calculation methods.

Agilent Feature Extraction software (version 11.0.1.1; Agilent Technologies, Inc.) was used to analyse the acquired array images. Quantile normalization and subsequent data processing were performed using the GeneSpring GX software package (version 12.1; Agilent Technologies, Inc.). After quantile normalization of the raw data, genes in which at least 1 out of 2 samples had flags in Detected (‘All Targets Value’) were chosen for further data analysis. Differentially expressed genes between the umbilical cord blood samples obtained from pregnant woman carrying fetuses with and without DS were identified through FC and volcano plot filtering. Hierarchical clustering was performed using the aforementioned R scripts. GO and KEGG analyses were performed using standard enrichment calculation methods.

Six circRNAs (Table I) and three mRNAs (Table II) were randomly selected to verify the accuracy of the results in peripheral blood samples of children with and without DS by RT-qPCR. Total RNA was extracted from the samples by TRI REAGENT BD (Molecular Research Center, Inc.) and reverse transcribed into cDNA using SuperScriptTM III Reverse Transcriptase (Invitrogen; Thermo Fisher Scientific, Inc.), 5X RT buffer, 10 mM dNTP mixture (dATP, dGTP, dCTP and dTTP; 2.5 mM each; HyTest Ltd.) and random primers supplied by Yingjun Biotechnology Co., Ltd., at 65°C for 5 min and on ice for 2 min. qPCR was subsequently performed with 2X SYBRGreen PCR master mix (Arraystar, Inc.) and primers (Table III). The following thermocycling conditions were used: 95°C for 5 min followed by 40 cycles of a denaturing step at 95°C for 10 sec and an annealing/extension step at 60°C for 60 sec. All reactions were run in triplicate. mRNA levels were quantified using the 2^−∆∆Cq method (29) and normalized to the internal reference gene β-actin.

To identify circRNAs acting as miRNA sponges, the circRNA/miRNA interactions were predicted using a miRNA target prediction software (cat. no. AS-S-CR-H-V2.0; Arraystar, Inc.) based on TargetScan (30) and miRanda (31), and the differentially expressed circRNAs were annotated with the circRNA/miRNA interaction information.

All statistical data were analysed using SPSS software (version 19.0). The statistical difference between the two groups was analysed using a Student's t-test. P<0.05 was considered to indicate a statistically significant difference.

circRNAs exhibiting a FC≥2.0 or ≤-2.0 were identified as being differentially expressed between umbilical cord blood samples of pregnant woman carrying fetuses with and without DS. A total of 735 differentially expressed circRNAs were detected between the two groups, of which 414 were upregulated and 321 were downregulated (Table SI). The circRNA expression profile in the two groups is presented in Fig. 1A.

A scatter-plot was used to visualize the variation (or reproducibility) of circRNA expression between two groups of samples. The values of the x- and y-axes in the scatter-plot were the normalized signal values of the samples (log2 scaled) or the averaged normalized signal values of groups of samples (log2 scaled). The green lines represented FCs. The circRNAs above the top green line and below the bottom green line indicated a ≥2.0-FC in circRNA expression between the two groups.

mRNAs with a FC≥2.0 or ≤-2.0 were selected as the differentially expressed ones. In the umbilical cord blood samples obtained from pregnant woman carrying fetuses with and without DS, 6,619 differentially expressed genes were detected, of which 3,411 and 3,208 genes were upregulated and downregulated, respectively (Table SII). The mRNA expression profile of the two groups is shown in Fig. 1B.

A total of six differentially expressed circRNAs in the microarray experiments were randomly selected and the relative expression was calculated using the 2^−∆∆Cq method (29). The circRNAs included three upregulated circRNAs (hsa_circRNA_103135, hsa_circRNA_103127 and hsa_circRNA_103112) and three downregulated circRNAs (hsa_circRNA_103137, hsa_circRNA_104907 and hsa_circRNA_101116). The results were validated using peripheral blood samples obtained from children with and without DS (Table I).

As seen in Table I, there were no significant differences in the expression levels of hsa_circRNA_103135, hsa_circRNA_103137 and hsa_circRNA_101116 (P>0.05), but there were extremely significant differences in the expression levels of hsa_circRNA_103127, hsa_circRNA_103112 and hsa_circRNA_104907 (P<0.01) between six children with DS and six children without DS.

A total of three statistically significant genes [dual specificity tyrosine phosphorylation regulated kinase 1A (DYRK1A), ubiquitin specific peptidase 25 (USP25) and spermatid perinuclear RNA binding protein (STRBP)], corresponding to three circRNAs (hsa_circRNA_103127, hsa_circRNA_103112 and hsa_circRNA_104907) were validated in peripheral blood samples obtained from children with and without DS. The results are shown in Table II.

As seen in Table II, there was no significant difference in the expression levels of DYRK1A and STRBP (P>0.05), but there was extremely significant difference in the expression level of USP25 (P<0.01) between six children with DS and six children without DS.

GO covers three domains: Biological process, cellular component and molecular function. GO analysis of the differentially expressed genes revealed that 805 genes were associated with the biological process domain, of which 315 were upregulated (Fig. 2A) and 490 were down-regulated (Fig. 2B). The five most enriched biological process terms were ‘cellular processes’, ‘primary metabolic processes’, ‘cellular metabolic processes’, ‘biological regulation’ and ‘nitrogen compound metabolic processes’. A total of 189 genes were associated with the cell composition domain, of which 66 were upregulated (Fig. 3A) and 123 genes were downregulated (Fig. 3B). The five most enriched cell composition terms were ‘intracellular’, ‘intracellular part’, ‘organelle’, ‘intracellular organelle’ and ‘membrane-bound organelle’. A total of 155 genes were associated with the molecular function domain, of which 77 were upregulated (Fig. 4A) and 78 were down-regulated (Fig. 4B). The five most enriched molecular function terms were ‘binding’, ‘protein binding’, ‘RNA binding’, ‘ligase activity’ and ‘catalytic activity’.

Pathway analysis of the differentially expressed genes allows the identification of genes related to specific cell pathways. Pathway analysis revealed that the differentially expressed genes were significantly enriched in 73 pathways (Fig. 5A and B). The upregulated genes were involved in 23 pathways and the downregulated genes were involved in 50 pathways. Differentially expressed genes were mainly involved in ‘adhesive spots’, ‘ECM-receptor interactions’, ‘gap junctions’, ‘protein digestion and absorption’, ‘vitamin digestion and absorption’ and ‘TNF signalling pathway’. Additional pathways included ‘primary immunodeficiency’, ‘systemic lupus erythematosus’, ‘rheumatoid arthritis’, ‘type I diabetes mellitus’, ‘arrhythmogenic right ventricular cardiomyopathy (ARVC)’ and ‘homologous recombination.

Endogenous circRNAs function as miRNA sponges, which means that the circRNAs bind miRNAs and repress their function (13,18,25).

A total of 3,263 MREs of differentially expressed circRNAs were described bioinformatically using the Arraystar miRNA target prediction software. Analysis of the MRE sequence showed several binding sites between hsa_circRNA_103112 and various miRNAs. There were five binding sites between hsa_circRNA_103112 and hsa-miR-20b-3p, hsa-miR-93-3p, hsa-miR-370-3p, hsa-miR-520d-3p or hsa-miR-651-3p. The sequence of hsa_circRNA_103112 was complementary to: i) The miR-20b-3p seed region at the 5′ end at position 61–66 with 7mer-m8 binding; ii) the miR-93-3p seed region at the 5′ end of the nucleotide at position 142–147 with 7mer-m8 binding; iii) the miR-370-3p seed region at the 5′ end at position 143–148 with 8mer binding; iv) the miR-520d-3p seed region at the 5′ end at position 172–177 with 7mer-m8 binding; and v) the miR-651-3p seed region at the 5′end at position 112–117 with 8mer binding. This indicated that hsa_circRNA_103112 could function as a sponge for miR-20b-3p, miR-93-3p, miR-370-3p, miR-520d-3p and miR-651-3p (Fig. 6).

The MRE sequence and the target miRNA seed type (7mer-m8, 8mer) and 3′pairing sequence (nucleotides 13–16) are presented in Fig. 6. The precise base positions are shown in the alignments in the upper left and right corners. The seed type 7mer-m8 shows an exact match to positions 2–8 of the target miRNA (the seed + position 8), and 8mer shows an exact match to positions 2–8 of the target miRNA (the seed + position 8). Adenosine/uracil (A/U) content 30 nt upstream and downstream of the seed sequence is shown in the column Local AU. Black bars represent guanine/cytosine and low accessibility whereas red bars represent A/U and high accessibility of the seed. The height of the bars shows the extent of the accessibility. The position column presents the most likely relative MRE position on the linear presentation of the circRNA.

circRNAs use their MRE to bind to miRNAs and consequently repress their function. Therefore, miRNAs can identify the mRNA of a particular coding sequence by base complementary pairing, affecting the translation process and thus playing a role in regulating gene expression (32). The present study investigated the expression profiles of differentially expressed circRNA and mRNA and conducted an association analysis between circRNAs and mRNAs. A total of 254 circRNA transcripts were found in the corresponding differential mRNA expression profile information. Among the 254 circRNA associated with mRNA, there were 153 upregulated circRNAs and 101 downregulated circRNAs, indicating that these identical genes may regulate the occurrence and development of Down's syndrome through circRNA/miRNA/mRNA interactions. The results are presented in Fig. 7.