A total of 19 women (between 23 and 42 years of age) participated in the present study, who experienced MA between 5–12 weeks of gestation and who were treated in the Department of Obstetrics and Gynecology of the Chinese People's Liberation Army (PLA) General Hospital (Beijing, China) between March 2017 and June 2017. Patient details are given in Table I. The inclusion criteria were as follows: The participants with spontaneous abortions in early pregnancy with unexplained etiology prior to the 12th week of gestational age and lack of any successful pregnancy in the previous history were included in the study. The exclusion criteria were as follows: The patients with a history of risk factors, including chronic infections, thrombosis, autoimmune diseases, endocrinological disorders or genital malformation were not included in the present study.

All procedures performed in studies involving human participants were in accordance with the ethical standards of the Chinese PLA General Hospital's research committee and with the 1964 Helsinki declaration and its later amendments, or comparable ethical standards. The present study was approved by the ethical committee of Chinese People's Liberation Army (PLA) General Hospital. Written informed consent was obtained from each participant.

The present study was conducted on 19 preserved chorionic villus samples with the normal chromosome number as determined by next-generation sequencing (NGS). WES was performed on chorionic villus genomic DNA from miscarriage samples of unknown cause.

Each exome was captured using Roche Nimblegen SeqCap EZ Exome v3.0 kit (Roche Applied Science, Madison, WI, USA) according to the manufacturer's protocol. Subsequently, the enriched exomes were sequenced using the Illumina HiSeq ×10 platform (Ilumina, Inc., San Diego, CA, USA). Reads were mapped against the human reference genome hg38 (https://genome.ucsc.edu/index.html) using Burrows-Wheeler Aligner (http://bio-bwa.sourceforge.net/). The single nucleotide variants (SNV) were called by SAMTools (version 0.1.19, http://samtools.sourceforge.net/) and the Genome Analysis Toolkit (GATK, version 4.0,4.0, Broad Institute, http://software.broadinstitute.org/gatk/), and ANNOVAR (http://annovar.openbioinformatics.org/en/latest/) was used for SNV annotation and filtering. Variants fulfilling the following criteria were retained: i) Missense, nonsense, frame-shift, or splice site variants; ii) absent in the dbSNP (http://www.ncbi.nlm.nih.gov/snp/), 1000 Genomes (http://browser.1000genomes.org/index.html), ESP6500 (http://evs.gs.washington.edu/EVS/), Exome Aggregation Consortium (ExAC; http://exac.broadinstitute.org/) and the Genome Aggregation Database (gnomAD; http://gnomad.broadinstitute.org/) databases. Four online functional prediction tools including Polyphen2 (http://genetics.bwh.harvard.edu/pph2/), SIFT (http://sift.jcvi.org/), MutationTaster (http://mutationtaster.org/) and FATHMM-MKL (http://fathmm.biocompute.org.uk/fathmmMKL.htm), were used to predict the variant effect on protein function. Constraint Metrics Z score for missense variation (9), Loss Intolerance (pLI) (10) and Haploinsufficiency Score (11) were used for evaluating the haploinsufficiency effect of each gene. DECIPHER database (https://decipher.sanger.ac.uk/) was used for identifying the previous published copy number variations and single nucleotide variants and the associated disease phenotypes. DAVID Bioinformatics Resources 6.7 (https://david-d.ncifcrf.gov/) was used for conducting the GO analysis.

Sanger sequencing was used to validate the WES results and verify whether the potentially disease-causing variants identified were true variants or sequencing artifacts. Sanger sequencing for the LDB1 variant in the POC QW013 was performed using gene-specific primers as follows; the forward primer was 5′-AGGAGTGTCACAATGCTCAGATGAT-3′ and the reverse primer was 5′-GTAAACGGAGACTCAGATGGGAGAG-3′. Cycling parameters were an initial denaturation at 94°C for 5 min followed by 35 cycles of denaturation at 94°C for 20 sec, annealing at 60°C for 30 sec and extension at 72°C for 1 min, followed by a final extension at 72°C for 5 min. TransStart FastPfu DNA polymerase (TransGen Biotech Co., Ltd., Beijing, China) was used in the PCR reaction.

A total of 19 patients (between 23 and 42 years of age) with MA participated in the present study (Table I). A total of 19 MA POCs (chorionic villus from 5–12 gestational weeks) were examined. The karyotypes of all POC were normal (Table I). WES was performed for each POC. The sequencing depth of each WES is listed in Table I. Therefore, WES-detected sequence variants identified in each embryo were focused upon. The present study aimed to identify direct sequence variants causing MA, which should not exist in live human beings. Therefore, polymorphisms with a minor allele frequency absent in the dbSNP, 1000 Genomes, ESP6500, ExAC and gnomAD databases were retained. Subsequently, all variants were further filtered, according to the list of 286 selected candidate genes that were associated with early embryonic lethality and MA. A total of 36 sequence variants in 32 genes potentially associated with MA were identified in 15 out of 19 patients (data not shown). All variants were in the heterozygous state.

GO analysis suggested that these 32 genes were enriched in biological processes in early embryonic development, including ‘chordate embryonic development’, ‘cell proliferation’ and ‘forebrain development’ (Fig. 1). In silico analysis predicted that 12 of 36 variants were considered to be pathogenic alleles by four online prediction tools, including Polyphen-2, SIFT, Mutation Taster and FATHMM-MKL (Table II). As all the variants were heterozygous, the present study aimed to determine whether the heterozygous state of the variants influenced disease tolerance. The variation intolerance scores were analyzed using three scoring systems, including the Constraint Metrics Z score for missense variation (9), Loss Intolerance (pLI) (10) and Haploinsufficiency Score (11). Out of the 12 genes, LIM domain binding 1 gene (LDB1) was the only gene that was predicted as intolerant to variation by the three scoring systems (Table II). Sequence variant c.3064C>T; p.P1022S (Table II) in another gene, death induced obliterator-1 (DIDO1), was additionally a potential candidate gene causing MA. This variant was predicted to be a pathogenic allele by Polyphen-2, MutationTaster and FATHMM-MKL. The variant in DIDO1 was considered loss-of-function-intolerant, as predicted by the Constraint Metrics Z score for missense variation and pLI (Table II). Therefore, it was hypothesized that the variant in LDB1 (c.662C>T; p.S221L) and DIDO1 (c.3064C>T; p.P1022S) was likely to be associated with embryo lethality and MA.

LDB1 serves important roles in the regulation of a variety of processes in early embryonic development, including heart formation (12), head and brain development (12–17), limb patterning (18), and eye development (16). In previous studies, the development of LDB1 null mutant mice was arrested at embryonic day 8.5 (E8.5) and mice succumbed at E9-E10 (12,19). Therefore, LDB1 may be a good candidate gene for human early embryonic lethality and MA. In the present study, the heterozygous c.662C>T variant in LDB1 was also validated by Sanger sequencing (Fig. 2A). DNA samples from the patient and her husband were unavailable, therefore it was not possible to analyze whether this variant was inherited or generated de novo. This variant and the flanking region were additionally highly conserved in humans and other animals including zebrafish (Fig. 2B), which reflects the highly conserved role of LDB1 in early embryonic development across different species.