Introduction

Fetal growth restriction (FGR), defined as the failure of the fetus to reach its genetically determined growth potential, is one of the most common causes of perinatal mortality and morbidity [1]. It results from multiple causes, such as genetic and epigenetic alterations, the environment, hormonal dysregulation, or placental vascular dysfunction. More than 150 genetic disorders have been associated with FGR [2].

Silver–Russell syndrome (SRS, OMIM #180860) is a rare but well known imprinting disorder [3]. Clinical diagnosis of SRS is considered if a patient shows at least four of the six criteria of the Netchine–Harbison clinical scoring system (NH-CSS) [3, 4], which includes pre- and postnatal growth retardation, relative macrocephaly at birth, body asymmetry, protruding forehead, and early feeding difficulties. An underlying molecular cause is identified in ~60% of patients with SRS [3, 5]. Among them, loss of methylation (LOM) at H19/IGF2:IG-DMR (also called ICR1) on chromosome 11p15 (11p15 LOM) is observed in 40 to 60% and maternal uniparental disomy (mUPD) for chromosome 7 (upd(7)mat) in ~5 to 10% [3, 5,6,7]. The recent international consensus of SRS recommends additional molecular testing in cases of normal methylation on chromosomes 11 and 7, including screening of cyclin D kinase inhibitor 1c (CDKN1C) and insulin-like growth factor 2 (IGF2) genes [8, 9]. Since the first consensus on SRS, new molecular defects have been identified in high mobility group AT-hook 2 (HMGA2) and pleiomorphic adenoma gene 1 (PLAG1) in patients with a clinical presentation of SRS [10, 11]. The recent use of next-generation sequencing (NGS) for SRS patients has improved the molecular diagnosis [12,13,14,15,16], but more than 30% of patients with SRS remain without an identified molecular cause.

Temple Syndrome (TS) is another rare cause of prenatal and postnatal growth restriction caused by disruption of the 14q32 imprinted region. In this region, MEG3/DLK1:IG-DMR is normally methylated on the paternal allele [17], resulting in Delta like non-canonical Notch ligand 1 (DLK1), retrotransposon Gag like 1 (RTL1), and Iodothyronine Deiodinase 3 (DIO3) expression from the paternal allele [18]. In contrast, long noncoding RNAs (maternally expressed 3 (MEG3) and maternally expressed 8 (MEG8)), microRNAs, and small nucleolar RNAs are expressed by the unmethylated maternal allele (Fig. 1). mUPD of chromosome 14 (upd(14)mat), hypomethylation of MEG3/DLK1:IG-DMR, and paternal deletion of this region all lead to the phenotype of TS. Clinical overlap between SRS and TS has been described and Geoffron et al. reported 73% of patients with 14q32 disruption scoring positively for SRS, with a NHCSS ≥4/6 [19,20,21,22]. According to Geoffron et al., 14q32 disruption may be considered to be an alternative molecular cause of SRS and MEG3/DLK1:IG-DMR methylation should be tested in cases of negative results for other molecular testing of SRS patients [3].

Fig. 1: Schematic representation of the imprinted domain of the 14q32 region.
figure 1

The black line “IG-DMR” indicates the differentially methylated region (the imprinting control center of 14q32, named IG-DMR methylated on the paternal allele). The star represents methylated DMR. Black boxes indicate genes expressed from the paternal (pat) allele (DLK1, RTL1, and DIO3). White boxes indicate genes expressed from the maternal (mat) allele (the non-coding genes MEG3 and MEG8, and a cluster of snoRNAs and miRNAs).

Full size image

DLK1 is widely expressed during fetal development. It encodes a transmembrane glycoprotein with six epidermal growth factor (EGF)-like motifs in its extracellular domain, a juxtamembrane region with a TACE-mediated cleavage site, a single transmembrane domain, and a short cytoplasmic tail [18]. The exact function of DLK1 is uncertain but it is involved in adipogenesis and appears to play an important role in preserving the pool of various progenitor cells until they differentiate [18]. With the generation of Dlk1-knockout mice, Moon et al. demonstrated overlapping phenotypes between Dlk1-null mice and human upd(14)mat, including growth retardation. They hypothesized that loss of Dlk1 expression may be responsible for most of the symptoms observed in human upd(14)mat [23]. Paternally inherited DLK1 variants have been recently identified in patients with central precocious puberty (CPP) but no growth retardation [24, 25]. Genetic defects of DLK1 have never been described in patients with FGR.

We screened DLK1 variants in a cohort of patients born SGA with a SRS phenotype using a NGS approach to search for new molecular defects responsible for SRS and assess the role of DLK1 in fetal growth.

Methods

Population studied

Patients included were referred to our molecular laboratory because they were born SGA (birth length and/or weight with a standard deviation score (SDS) < −2 [26]) with a clinical suspicion of SRS. Patients born SGA with a NH-CSS ≥4/6 or a NH-CSS = 3/6 with a strong clinical suspicion of SRS (relative macrocephaly and/or protruding forehead), negative molecular testing for H19/IGF2:IG-DMR (11p15.5) and DLK1/MEG3:IG-DMR (14q32.2) LOM and upd(7)mat, and negative molecular testing for CDKN1C, IGF2, HMGA2, and PLAG1 variants were included in this study. Written informed consent for participation was received from all patients or parents, in accordance with national ethics rules (Assistance Publique–Hôpitaux de Paris authorization no. 681). Patients were either followed at Armand Trousseau Children’s Hospital or referred by other clinical centers for molecular analysis. Postnatal growth parameters are expressed as SDS according to charts by Sempé and Pedron [27]. Blood samples were collected during routine biological follow-up at clinical visits. DNA was extracted in our laboratory from peripheral blood samples using an in-house protocol after cell lysis by a salting-out procedure, as previously described [28, 29].

Next-generation sequencing

DLK1 was sequenced as a SRS candidate gene using targeted sequencing. Library preparation, gene enrichment, sequencing, and data analysis were performed by IntegraGen SA (Evry, France) or by our laboratory with a pipeline designed by SOPHiA GENETICS (Lausanne, Switzerland).

Sanger sequencing

Variations of DLK1 identified through NGS were verified for the probands and their parents by Sanger sequencing using the ABI PRISM Big Dye Terminator v3.0 Cycle Sequencing Kit and an ABI 3100 Genetic Analyzer (Life Technologies, Courtaboeuf, France).

In silico analysis

The allele frequency was checked in the GnomAD database (online https://gnomad.broadinstitute.org/) to predict the functional consequences of any identified DLK1 variants. Interspecies alignment of DLK1 was performed using Clustal Omega (online tool from The European Bioinformatics Institute (EMBL-EBI), http://www.ebi.ac.uk/Tools/msa/clustalo/) and damage prediction scores were obtained using the Polyphen-2 bioinformatic tool [30]. Variants were also classified as benign or likely benign, pathogenic or likely pathogenic, or of uncertain significance following the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) classification of variants [31]. Six main categories are evaluated according to these guidelines: population data (prevalence of the variant in control populations), computational in silico predictive data, functional characterization, segregation, de novo data, and allelic data.

Statistical analysis

Characteristics of the population are described as percentages for qualitative variables or as the SDS and mean (range) for quantitative variables.

Results

Clinical characteristics of patients screened for DLK1 variants

Samples from 132 patients referred for molecular genetic testing for SRS and without disturbances of 11p15 and 14q32 methylation or upd(7)mat were analyzed using a targeted NGS-based approach (93 by Integragen and 39 using the pipeline of SOPHiA GENETICS). Clinical data for at least four NH-CSS criteria were available for all patients and was complete (all six criteria available) for 61 (46%). All patients were born SGA. Patient characteristics are presented in Tables 1 and 2.

Table 1 Anthropometric data at birth for patients included in the analysis.
Full size table
Table 2 Clinical features and NH-CSS for patients included in the analysis.
Full size table

Screening for DLK1 variants

Among the 132 patients born SGA with a clinical suspicion of SRS, only two rare heterozygous variants of DLK1 were identified in two independent patients (NM_003836.6:c.103 G > C, p.(Gly35Arg) and NM_003836.6: c.194 A > G, p.(His65Arg)).

Phenotype of patients with identified DLK1 variants

Patient 1 was the third female child of two non-consanguineous parents. The proband’s parents were healthy. The final height of the father was 176 cm (0.2 SDS) and that of the mother 152 cm (−2.0 SDS). Proband’s mother was not born SGA and had no clinical features of SRS. The proband’s sisters were healthy, without prenatal or postnatal growth restriction. Patient 1 was born at 36 + 2 weeks of amenorrhea (WA). Her birth weight was 1930 g (−1.9 SDS), birth length 41 cm (−3.6 SDS), and head circumference 31 cm (−1.6 SDS). She did not experience catch up growth, with a height of 98.5 cm (−3.3 SDS), weight of 13.5 kg (−3.4 SDS), and head circumference of 49 cm (−1.5 SDS) at 6 years of age. She had no other remarkable features. She was not treated by growth-hormone (GH) therapy. She had feeding difficulties and a protruding forehead and fulfilled five of the six criteria of the NH-CSS. NGS sequencing of DLK1 revealed the heterozygous NM_003836.6: c.103 G > C variant located in exon 2, predicting an amino acid substitution at codon 35 (p.(Gly35Arg)). Gly35 is located within the first EGF-like motif in the extracellular domain of DLK1. This variant was inherited from her healthy mother, who carried the same heterozygous variant (Fig. 2).

Fig. 2: Intrafamilial segregation of variants.
figure 2

Intrafamilial segregation of variants NM_003836.6:c.103 G > C p.(Gly35Arg) of DLK1 in family 1 (a) and NM_003836.6: c.194 A > G p.(His65Arg) of DLK1 in family 2 (b). The two variants were inherited from the mother of the patients. In black, patients born SGA.

Full size image

Patient 2 was the first female child of two non-consanguineous parents. The mother’s final height was 169 cm (1.0 SDS). The proband’s father was born SGA at 35 WA (birth weight and height were 1290 g (−3.4 SDS) and 38 cm (−4.7 SDS)), but his head circumference at birth was uknown and we do not know if he had protruding forehead between 1 and 3 years. The proband’s father had no feeding difficulties during childhood and initially experienced catch-up growth with a height at −1 SDS between 10 and 14 years of age, but his final height was only 163 cm (−2.0 SDS). He was not treated with GH therapy. NH-CSS of proband’s father was 1/4. Patient 2 was a 29 WA-preterm girl with a birth weight of 920 g (−1.9 SDS), birth length of 34 cm (−3.0 SDS), and head circumference of 25 cm (−1.2 SDS). SGA was diagnosed in the second trimester of gestation. She did not develop feeding difficulties but had a protruding forehead. By 16 months of age, she had not experienced catch-up growth, with a height of 68.5 cm (−3.0 SDS) and a head circumference of 44.5 cm (−1.0 SDS). She was suspected of having SRS, with a NH-CSS = 4/6 and fifth finger clinodactyly. At 5 years of age, she experienced premature adrenarche without precocious puberty, responsible for catch-up growth, with a height of 105.4 cm (−0.8 SDS). NGS sequencing of DLK1 revealed that she carried a heterozygous NM_003836.6:c.194 A > G variation in exon 3 of DLK1, predicting an amino acid substitution at codon 65 (p.(His65Arg)). His65 is located within the second EGF-like motif of the extracellular domain of DLK1. This variant was inherited from her healthy mother who carried the same heterozygous variant (Fig. 2).

In silico analysis of the two DLK1 variations

The two variants NM_003836.6:c.103 G > C p.(Gly35Arg) and NM_003836.6: c.194 A > G p.(His65Arg) are described in GnomAD and dbSNP (rs762558665 and rs147224004) with an allele frequency in the general population of 7.0 × 10−6 and 4.7 × 10−4, respectively. Interspecies alignment of the amino acid sequences of DLK1 showed that residue Gly35 is invariant in vertebrates. The variation NM_003836.6:c.103 G > C p.(Gly35Arg) is predicted to be probably damaging, with a score of 1.000 by the Polyphen-2 bioinformatic tools of variation damage prediction. This variant is classified as a variant of uncertain significance (class 3) according to the ACMP/AMP classification (PM2-PP3).

Residue His65 is conserved only within Pan Troglodytes. Polyphen-2 predicted the variation p.His65Arg to be benign, with a score of 0.215. Variant NM_003836.6: c.194 A > G p.(His65Arg) is classified as likely benign (class 1) according to the ACMP/AMP classification (PM2-BP4-BS4).

The two variants NM_003836.6:c.103 G > C p.(Gly35Arg) and NM_003836.6: c.194 A > G p.(His65Arg) were not described in ClinVar. We submitted it (submission SUB9482112 and SUB9433818).

Discussion

We found two rare heterozygous variants of DLK1 in a cohort of 132 SGA patients with clinical suspicion of SRS and no identified molecular defects. The variants have already been reported in databases but with a low frequency. However, caution should be paid about variants frequencies regarding imprinted genes, as such a variant might have a different clinical impact depending on the maternal or paternal inheritance. Segregation analysis did not favor a pathogenic effect of these two variants, as they were both on the maternal allele, which is silent due to the maternal imprint of this gene. Thus, we did not identify any pathogenic DLK1 variants in our cohort.

No variants of DLK1 have been reported in SGA patients. Dauber et al. identified a complex defect of DLK1 (14-kb deletion and 269-bp duplication) in four patients with familial CPP. The four patients did not show prenatal or postnatal growth failure, and other classical clinical features of Temple or SRS, such as feeding difficulties, facial dysmorphia, precocious obesity, and relative macrocephaly, were excluded [24]. Gomes et al. identified three frameshift variants of DLK1 (NM_003836:c.594_594delC p.(Gly199Alafs*11), NM_003836:c.810_810delT p.(Val271Cysfs*14), and NM_003836:c.479_479delC p.(Pro160Leufs*50)) in five women from three families with CPP. Among them, three experienced postnatal growth failure, but no data were available about birth weight or length [25]. Montenegro et al. described a deletion (c.401_404 + 8del) in the splice-site junction of DLK1 in a girl with sporadic CPP without postnatal growth failure. No data about prenatal growth were available for this patient [32].

The role of DLK1 in fetal growth is not well established. Murine models have suggested a role for Dlk1 in fetal and postnatal growth. Indeed, Dlk1-null mice and heterozygous mice with paternal inheritance of the Dlk1-knockout allele showed prenatal and postnatal growth restriction [23, 33]. By contrast, heterozygous mice with maternal inheritance of the Dlk1-knockout allele did not experience growth restriction. Moreover, it has been demonstrated that Dlk1 promotes Gh expression. Dlk1-null mice showed reduced pituitary GH content and mice overexpressing Dlk1 had excessive pituitary and circulating levels of GH [33, 34]. The modulation of GH levels could explain, at least in part, the postnatal growth failure of mice lacking Dlk1, but not the prenatal growth failure. During fetal life, Dlk1 is expressed in the placenta in the endothelial cells of the placental labyrinth but is not required for its development [18]. Mice with a conditional deletion of Dlk1 in placental endothelial cells did not show FGR [35]. Reduced DLK1 levels in maternal blood samples have been shown in the second and third trimester of gestation with FGR [36, 37]. However, a causal relationship between low DLK1 levels and FGR has not been demonstrated or whether DLK1 levels simply reflect fetal weight.

To date, less is known about the contribution of individual genes of the 14q32 domain in the TS phenotype and the overlapping features with SRS. FGR could, for example, be explained by the action of several genes of the 14q32 domain in concert with genes in other imprinted domains [38]. Indeed, Abi Habib et al. demonstrated that overexpression of MEG3 and MEG8 in TS patients with 14q32 hypomethylation is associated with downregulation of IGF2 transcription from the 11p15 imprinting region [28].

In conclusion, we did not identify any variants in DLK1 in a cohort of 132 patients with suspected SRS. Although we screened a large cohort of patients for DLK1 variants, we cannot rule out the possibility of a role of DLK1 in fetal growth and the SRS phenotype. However, a frequent contribution of DLK1 variants among the molecular causes of SRS is unlikely. We did not identify a new molecular cause of SRS by the targeted NGS approach. Whole exome and genome sequencing and characterization of the entire methylome offer promising perspectives for the identification of new molecular causes of SRS.