Eleven nonconsanguineous families with MFS were recruited from International Peace Maternity and Child Health Hospital in this study. The ethics committee of International Peace Maternity and Child Health Hospital approved the project and investigators followed the principles of the Declaration of Helsinki. Informed consent was obtained from each patient and their related families before genetic testing. MFS was diagnosed according to Ghent criteria by cardiologists, ophthalmologists, internists, and geneticists.

Clinical data were retrospectively collected based on patients’ medical records kept at our hospital. Case inclusion criteria and clinical data inclusion scope are (Radke and Baumgartner, 2014): diagnosed patients or family history of MFS (Ramirez and Dietz, 2007); Cardiovascular system phenotype: aortic dilatation, aortic dissection, or mitral valve prolapse (Ammash et al., 2008); Ocular system phenotype: high myopia >6.0D or ectopia lentis (Yetman et al., 2003); Skeletal system phenotype: arachnodactyly, scoliosis, pectus excavatum, or flatfeet.

Genomic DNA was isolated from peripheral blood or using the MagNA Pure LC DNA Isolation Kit (Roche Diagnostics, GmbH, Mannheim, Germany). Whole-exome sequencing library construction and sequencing were performed using the Illumina platform by Beijing Genomics Institute (BGI) according to the manufacturer’s protocols. The detection covers exons (over 180,000) and 10bp flanking sequences of 22,000 genes. Exome sequencing was performed on the HiSeq2000 sequencing platform (Illumina). In-solution whole-exome capture and massively parallel sequencing was performed using the Agilent SureSelectXT All Exon Kit 51 Mb. Sequenced reads were collected, filtered for quality, and aligned to the human reference sequence (UCSC Genome Browser hg19) with the Burrows-Wheeler Aligner. On average, over 95% of exons were covered at >20×. Sequence variants including single-nucleotide variants (SNVs) and small insertions or deletions (indels) were annotated by ANNOVAR software. Common variants (defined as 10% frequency in 1,000 Genomes) were excluded if they were present in the dbSNP (v.142) database, the 1,000 Genomes Project, or the Exome Aggregation Consortium (ExAC) Browser. The detected variants were annotated and filtered with Annovar based on public databases [such as Mendelian Inheritance in Man (OMIM), Exome Aggregation Consortium (ExAC) Browser and MutationTaster2] in accordance with the criteria set by the American College of Medical Genetics and Genomics (ACMG)/Association for Molecular Pathology (AMP) guidelines. We focus on screen mutations of the FBN1 (NM_000138.5) gene with bioinformatics analysis of FASTQ files. Each mutation we found will be confirmed by bidirectional Sanger sequencing.

The sequences of screened variations sites in FBN1 (NM_000138.5) were obtained from UCSC Human Genome Browser. Pedigree analysis was performed to identify the disease-causing mutation. The variants were detected in probands and his/her family members by polymerase chain reaction (PCR). All variants were classified according to the FBN1-Specific Guideline for the Interpretation of Sequenced Variants in the FBN1 gene. The involved databases and criteria are as follows:

1) Prediction tools which are used to assess mutation pathogenicity contained SIFT (http://sift.jcvi.org), PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2), Rare Exome Variant Ensemble Learner (REVEL) (https://sites.google.com/site/revelgenomics/), ClinPred (https://sites.google.com/site/clinpred/), Human Splicing Finder (http://www.umd.be/HSF3/index.html/), and NNSPLICE (http://www.fruitfly.org/seq_tools/splice.html).

Proband in family five is a female (III-2). She was a 26-year-old woman who was diagnosed with Marfan syndrome (MFS). Her mother and grandfather are all MFS patients and are treated for to aortic dilatation and mitral valve prolapse. She and her mother are very tall. All three MFS patients in this family suffered cardiac anomalies and skeletal dysplasia. Her father, husband, and aunt were apparently healthy. The variant is classified as uncertain significance with a pathogenic possibility of 67.5%–81.2. The patient had a strong desire to block the disease and to have an unaffected child via PGT-M. After expert consultation, informed consent from the patient and her husband, and approval from the ethics committee, the couples were determined to be pregnant through IVF and PGT-M, and signed an informed consent form for the PGT-M cycle.

Proband in family eight is a Male (I-2). He was a 36-year-old man who was diagnosed with Marfan syndrome (MFS). He had a c.5498G > T [p(Cys1833Phe)] mutation in FBN1 gene. He was a typical MFS patient with disease phenotype in the Cardiovascular system, Ocular system, and Skeletal system. The variant is classified as pathogenic which is a pathogenic possibility over 99.7%. The MFS patient and his wife decided to undergo an IVF cycle associated with PGT-M and signed their informed consent.

Family five and family eight obtained 18 and 10 embryos respectively; embryo biopsy was performed on Day 3 (cleavage stage) and blastocysts biopsy was performed in embryos of grade 3 or higher according to Gardner’s grading scale on day 6. The multiple displacement amplification (MDA) products and gDNA libraries were prepared and captured using a 1.5 Mb customized probe covering 350 kb upstream to 350 kb downstream of the FBN1 gene. The SNPs identified in the couple and their parents were used for haplotype construction. Embryos diagnosed as unaffected were selected for transfer. Prenatal molecular diagnosis was performed through amniocentesis at the 13th-20th gestational week. The fetal genotype was confirmed by Sanger sequencing.

Clinical information of MFS probands and patients was summarized in Table 1. Affected patients from these families exhibited similar clinical symptoms of MFS. All the healthy family members had no features of MFS. We recruited 11 probands with MFS and recorded the phenotype of all MFS patients. The phenotype and family history of affected patients with FBN1 mutation coincided with MFS except in Family 4 and Family 6, the clinical manifestations of the two families are mainly ocular defects including ectopia lentis and high myopia >6.0D. The clinical feature of the proband in Family 11 is missing since the proband died before birth.

In this study, we analyzed the genomic DNA of 11 probands with MFS. The quality and reliability of targeted NGS data were evaluated based on the percentage of readable bases and the coverage depth in the targeted region, to ensure complete sequencing coverage of all coding regions in candidate genes. Altogether, 11 potentially disease-causal FBN1 gene variants were screened out in 11 probands, 6/11 (54.5%) variants had not been reported before and 5/11 (45.5%) variants had been reported in MFS patients. We showed all mutations and their location in the protein domain of FBN1 (Figure 1). The hot spot mutations were considered to happen frequently in exon 1–16 and secondarily exon 32–40. Among 11 variants, six variants marked in red were first reported through this study. The location and basic information of all mutations are summarized in Table 2. Except for one splicing variant, nine variants are in Ca-binding EGF-like domains and one is not in a Ca-binding EGF-like domain. There are eight missense variants, and four of eight (50%) missense variants affected conserved Cysteine residues of fibrillin 1 protein.

Pedigree analyses were performed to obtain familial segregation data and determine whether the mutation is a de novo mutation. Each variant, considered as a causative candidate and pathogenic mutation, was further validated using the Sanger sequencing method in other family members. Figure 2 is the pedigree chart of 11 MFS families. FBN1 mutations in Family 3 and Family 11 were do novo mutations. Figure 3 showed the validation result of FBN1 mutations in MFS families by sanger sequencing.

To identify disease-causing mutations for MFS families, stringent criteria according to ACMG guidelines were performed. We carefully examined all available literature and mutation-related database for sequence variant interpretation. In Table 3, allele frequencies of variants detected in populations, pieces of computational evidence of pathogenicity prediction, the clinical significance of variants in Clinvar, and whether the variant is a de novo variant are listed. We explored the classification evidence for each variant and did sequence variant interpretation according to the ACMG criterion. In Figure 4, conservation analysis of the related homologous proteins in eight FBN1 missense mutation sites was conducted by referring to the UniProt database. Orthologous protein sequence alignment for eight missense mutations showed the eight mutation sites happened in a highly conserved region of FBN1 among different species. Finally, we confirmed 11 causative candidate and pathogenic heterozygous mutations including five known mutations of the FBN1 gene in the patients, including c.4955G > A [p(Cys1652Tyr)], c.4210+1G > A [p(?)], c.7559C > T [p(Thr2520Met)], c.364C > T [p(Arg122Cys)], and c.3602G > A [p(Cys1201Tyr)], as well as six novel mutations never reported before, including c.4469A > C [p(Glu1490Ala)], c.6615A > G [p(Glu2205 = )], c.3244G > T [p(Gly1082Cys)], c.5885_5895del [p(Tyr1962Serfs*11)], c.5498G > T [p(Cys1833Phe)], and c.6695G > T [p(Cys2232Phe)].

For the FBN1 gene, missense variants are a common mechanism of MFS disease (PP2). c.4955G > A [p(Cys1652Tyr)] is a missense variant (PP2), it had been reported by several groups in patients with MFS (PS4_Moderate). The variant is not present in GnomAD and ExAC (PM2). The variant results in a Cysteine substitution in the critical function domain of the calcium-binding EGF-like domain of FBN1 protein (PM1). In Family 1, the mutation was segregated with the phenotype in five affected patients and was absent in unaffected individuals (eLOD = 1.20, PP1_Moderate). Phenotype and family history of this family and reported patients were consistent with FBN1-induced MFS (PP4). Cys1652 is conserved between species and predicted to be disease-causing according to in silico prediction (PP3). In summary, this variation is classified as “likely pathogenic variation”. In August 2016, the patient and his wife became pregnant naturally, and the fetus (II-2) was diagnosed with FBN1 c.4955G > A mutation after prenatal diagnosis.

c.4210+1G > A [p(?)] is in intron 34 and affects the canonical +1 donor splice site most likely leading to abnormal splicing (PVS1). It has been reported previously in association with MFS (PS4_Supporting). The mutation was not detected in the samples of proband’s parent (PM6) and was not present in GnomAD (PM2). In summary, this variation is classified as “pathogenic variation”. In December 2019, the patient and his wife became pregnant naturally, and the fetus (III-1) did not carry this mutation after prenatal diagnosis.

c.7559C > T [p(Thr2520Met)] is a missense variant (PP2), and is found at a very low frequencies in GnomAD (0.00004246) (PM2). Although the variant had been reported in an MFS patient (PS4_Supporting), in our study, the mutation was not segregated with the phenotype in five affected patients and was detected in unaffected individuals (BS4). In summary, this variation is classified as “Uncertain significance variation”.

c.364C > T [p(Arg122Cys)] is a missense variant (PP2) and located in a critical function domain of the not calcium-binding EGF-like domain of FBN1 protein (PM1). The variant is found at a very low frequency in GnomAD (0.000003986) (PM2) and many pieces of prediction software predict this variant to be damaging (PolyPhen2, SIFT, REVEL) (PP3). The variant had been reported in MFS patients before (PS4_Moderate) and was segregated with the phenotype in eight affected patients (eLOD = 2.00, PP1_Strong). In Family 10 and the previous reports, phenotypes and family histories of patients were consistent with MFS (PP4). In summary, this variation is classified as “pathogenic variation”.

c.3602G > A [p(Cys1201Tyr)] is a missense variant (PP2) and results in a Cysteine substitution in the critical function domain of the calcium-binding EGF-like domain of FBN1 protein (PM1). The mutation was not present in GnomAD (PM2) and was predicted to be disease-causing by in silico analysis (PP3). In Family 11, the mutation is a de novo mutation (PM6), and the mutation had been reported in the literature (PS4_Moderate). In summary, this variation is classified as “likely pathogenic variation”. In May 2019, the parents of the probands conceived naturally, and the fetus (II-2) did not carry FBN1 c.3602G > A mutation after prenatal diagnosis.

c.4469A > C [p(Glu1490Ala)] is a novel missense variant (PP2) and located in a critical function domain of the calcium-binding EGF-like domain of FBN1 protein (PM1). The mutation was not present in GnomAD (PM2) and was predicted to be disease-causing by in silico analysis (PP3). In Family 2, the mutation was segregated with the phenotype in four affected patients and was absent in unaffected individuals (eLOD = 0.90, PP1). Moreover, mutation at Glu1490 locus [p(Glu1490Lys)] had been classified as likely pathogenic (PM5). In summary, this variation is classified as “likely pathogenic variation".

c.6615A > G [p(Glu2205 = )] is a novel synonymous variant that had not been reported before, all three MFS patients had c.6615A > G mutation in FBN1 but the other healthy members in her family did not carry this mutation (eLOD = 0.90, PP1). The variant is absent from all population databases (PM2), and multiple lines of computational evidence support a deleterious effect on the gene product through disturb splicing (PP3). The phenotype and family history of the proband in this family are highly consistent with MFS characteristics (PP4). All three MFS patients in this family suffered cardiac anomalies and skeletal dysplasia. In Figure 5, we can see three affected patients have arachnodactyly, flatfeet, pectus excavatum, or pectus carinatum. In summary, this variation is classified as “Uncertain significance variation”. The variant is classified as uncertain significance with a pathogenic possibility of 67.5%–81.2. The patient had a strong desire to block the disease and to have an unaffected child via PGT-M. After expert consultation, informed consent from the patient and her husband, and approval from the ethics committee, the couples were determined to be pregnant through IVF and PGT-M, and signed an informed consent form for the PGT-M cycle.

c.3244G > T [p(Gly1082Cys)] is a novel missense variant (PP2) and located in a critical function domain of the calcium-binding EGF-like domain of FBN1 protein (PM1). The mutation was not present in GnomAD (PM2) and was predicted to be disease-causing by in silico analysis (PP3). In Family 6, the mutation was segregated with the phenotype in six affected patients and was absent in unaffected individuals (eLOD = 1.50, PP1_Strong). In summary, this variation is classified as “likely pathogenic variation”.

c.5885_5895del [p(Tyr1962Serfs*11)] is a novel frameshift variant that likely leads to a truncated protein (PVS1). The variant is absent from all population databases (PM2), and the phenotype of affected patients is consistent with MFS (PP4). In summary, this variation is classified as “pathogenic variation”.

c.5498G > T [p(Cys1833Phe)] and c.6695G > T [p(Cys2232Phe)] are all novel missense variants (PP2) and result in a Cysteine substitution in the critical function domain of the calcium-binding EGF-like domain of FBN1 protein (PM1). The mutations were not all present in GnomAD (PM2) and were predicted to be disease-causing by in silico analysis (PP3). In Family eight and Family 9, the phenotype of affected patients is consistent with MFS (PP4). Moreover, mutation at Cys1833 locus [p(Cys1833Arg), p(Cys1833Ser)], and mutation at Cys2232 locus [p(Cys2232Tyr), p(Cys2232Arg)] had been reported to be pathogenic (PM5_Strong). In summary, the two variations are classified as “pathogenic variation".

In Family 5, we defined the haplotype linked to c.6615A > G and associated with MFS as F0, and the haplotype linked to wildtype allele as F1. In Family 8, we defined the haplotype linked to c.5498G > T and associated with MFS as M0, while the haplotype linked to wildtype allele was defined as M1. The genotypes of 16 embryos were all successfully determined using the Hidden Markov model approach (Figure 6; Table 4). In Family 5, five embryos were free of maternal FBN1 c.6615A > G variant; embryos 2, 7, and 8 were carriers of c.6615A > G variant. In Family 8, five embryos were free of paternal FBN1 c.5498G > T variant; embryos 1, 6, and 8 were carriers of c.5498G > T variant.

Embryo 6 in Family 5 and embryo 2 in Family 8 were selected for transfer and a successful pregnancy was confirmed by human chorionic gonadotropin (hCG) and ultrasound examination. Sanger sequencing results showed that the fetus did not carry mutations. This confirmed the accuracy of PGT-M. The chromosome imbalance anomaly results showed that no copy number variant (CNV) larger than 100 kb was identified in the fetus.