Introduction

Neurofibromatosis type 1 (NF1; MIM#162200) is a fully penetrant autosomal disorder with an estimated incidence of 1 in 3500 live births. The main features of NF1 are multiple café-au-lait macules (CALMs), lentiginous macules, and a predisposition to benign and malignant tumors1. Cognitive and psychosocial problems, including learning disabilities, attention-deficit/hyperactivity disorder, and autism spectrum disorders, are more prevalent in the NF1 population compared to the general population2,3. Vasculopathy, such as hypertension and vascular abnormalities of the heart, brain, kidneys, or other major arteries, is an important cause of complications or even early death in patients with NF1. The life expectancy is about 10–15 years shorter than the general population and lifelong medical follow-up is advised, preferably by a specialized NF1 clinic. Diagnosis relies on clinical assessment according to standardized criteria4,5. Sporadic patients with NF1 are diagnosed when presenting two or more of the most characteristic clinical NF1 features or with only one for children, usually CALMs, and an NF1 pathogenic variant. For familial cases, the presence of one clinical NF1 feature, and an affected parent would be sufficient to clinically diagnose the patient, although a molecular test can be recommended to confirm the clinical diagnosis. The disease is marked by poor genotype–phenotype correlation, with the exception of a few specific variations6,7,8,9,10,11, and high inter- and intra-familial variability12,13,14.

In more than 95% of cases, NF1 is caused by autosomal dominant loss-of-function variants in the NF1 gene15. NF1 is located at 17q11.2 and encodes a GTPase-activating protein (neurofibromin) that acts as a negative regulator of the RAS-mitogen-activated protein kinase signaling cascade16. The NF1 gene shows one of the highest mutation rates, with >3700 different pathogenic variants referenced in public databases (Leiden Open Variation Database, LOVD, and Human Gene Mutation Database). The large spectrum of NF1 pathogenic variants is distributed through the entire coding region and splice sites with no hotspot. Point mutations but also large deletions encompassing NF1 and neighboring genes were described11,17,18,19,20.

More than 50% of NF1 germline pathogenic variants arise de novo, and a relatively high frequency of mosaicism is observed in NF1 cohorts. There are different types of mosaicisms that describe which parts of the body harbor the mutated cells and the potential for transmission to offspring. These include isolated germline mosaicism (also called gonadal mosaicism), somatic mosaicism, and a combination of germline and somatic mosaicism21,22,23. Although these classifications are useful in a practical sense, they cannot be conclusively assigned owing to the limitations of tissue sampling. Only somatic cells (mainly from blood and skin) are routinely used in genetic analyses, and a hint about possible germline mosaicism can be obtained only if somatic mosaicism is detected, or if more affected siblings are born with an identical de novo variant.

Patients with mosaic NF1 may more often develop mild NF1 phenotypes or manifestations limited to the affected area of the body with often unilateral manifestations. The presence of mosaic NF1 can complicate molecular diagnosis (with a low variant allele frequency (VAF) of the pathogenic NF1 variant) and specific criteria for mosaic NF1 have been defined, including neurofibroma analysis to help identify the causative NF1 variation4. It has been estimated that ~10% of sporadic NF1 patients have mosaic NF1 caused by postzygotic NF1 mutations that are absent from, or present in, a very low proportion of blood lymphocytes24. Genetic counseling, risk assessment, and fertility options for NF1 mosaic parents present clear challenges25.

Molecular analysis of the NF1 gene is important in clinical practice to confirm the diagnosis, to differentiate from phenocopies (such as Legius syndrome; MIM#611431)26, and to allow genetic counseling27. Each child of an individual with NF1 has a 50% chance of inheriting the disease-causing variant. NF1 is considered to be fully penetrant; thus, a child who inherits an NF1 pathogenic variant is expected to develop features of NF1, but the features may be considerably more (or less) severe in an affected child than in his or her affected parent. In fact, NF1 is considered fully penetrant by the age of eight years with extremely variable clinical manifestations, even within families carrying the same variant [7–8]. A clear genotype–phenotype correlation has only been established for a few NF1 variants and remains an area of active research6,7,8,10,18.

The unpredictable clinical expression in offspring can complicate reproductive decision-making for NF1 patients and their partners28,29,30,31. In France, one of the options for preventing the birth of an affected child is prenatal diagnosis (PND) by chorionic villus sampling (CVS) or amniocentesis, with the option to terminate the pregnancy if the fetus is affected. Another option is preimplantation genetic testing32. PND and preimplantation genetic testing for a pregnancy at increased risk are possible if the disease-causing variant is known, or if there are multiple affected family members, and linkage has been established within the family29; such testing is useful for establishing the presence of the parental mutation in the fetal DNA, but as noted, cannot make any prediction about disease severity.

We report a retrospective 5-year study of PND (2017–2022) in a French diagnosis laboratory for NF1. In each family, the proband’s NF1 pathogenic variant was identified using a targeted next-generation sequencing (NGS) approach. PND consisted of the combination of direct molecular diagnosis using Sanger sequencing or quantitative polymerase chain reaction (qPCR) and indirect diagnostics by linkage analysis based on microsatellite markers segregation in the family, when applicable. In addition, we present two cases of PND in women with mosaic NF1, including the germ line: by PND screening, we showed that the fetuses were not affected but nevertheless carried the risk haplotype associated with the NF1 pathogenic variants in the mosaic-affected parent.

This study illustrates the value of PND in NF1, with crucial consequences for medical and genetic counseling decisions. Our observations also point to the challenges that may be associated with germline mosaics.

Results

French NF1 PND (2017–2022)

For PND, each couple received and understood detailed information on the clinical features of NF1. A reflection period was offered to the couples. The risk of transmission and the high clinical variability of NF1 were discussed. When the couple did not want to take the risk of transmitting the disease to a child, because of the potentially severe complications of NF1, the uncertainty of the disease course, and the medical monitoring, the multidisciplinary PND center authorized a prenatal test, after several multidisciplinary discussions considering the patients’ request and the strong probability and severity of the disease. Finally, PND was discussed and performed when requested.

In 146 women, 205 NF1 PNDs were performed (2017–2022) through CVS biopsies (181; 88%) or amniocentesis (24; 12%): the pathogenic NF1 variant was present in 85 (41%) fetuses and absent in 122 (59%) fetuses. Among these 205 pregnancies (207 fetuses), 135 were carried to term (119, 57.5% unaffected children, and 16, 7.5% NF1 affected children), 69 (33.5%) were pregnancy terminations (affected fetuses), 2 (1%) miscarriages (both NF1 unaffected fetuses), and 1 in utero fetal death (NF1 unaffected fetus) (Table 1). Two of the 205 PNDs corresponded to monozygotic twins (both unaffected) and dizygotic twins (one affected and one unaffected); these two pregnancies were carried to term.

Table 1 Characteristics and previous reproductive history of the couples proceeding with PND (2017–2022): 205 PNDs in 146 patients
Full size table

Among the 146 families, we detail below the molecular analyses for two families in which an NF1 germline mosaic was identified in the women.

Family 1

A woman aged 35 years (Fig. 1: Family 1, individual I.1) with no family history of NF1 was clinically diagnosed with a segmental NF1 (segmental CALMs); the genetic NGS test identified a de novo mosaic pathogenic genetic variant in exon 43 of NF1 (NM_000267.3) c.6641 + 1 G > C p.(?) with a 9% VAF in two independent blood samples, confirming a mosaic NF1. The following criteria classified the NF1 c.6641 + 1 G > C variant as pathogenic (class 5), according to ACMG-AMP criteria33: null variant (canonical splice site) in a gene where loss-of-function is a known mechanism of disease (pathogenic very strong criterion: PVS1), absence of the variant in the population database gnomADv3 (PM2), reputable source (ClinVar ID856149: pathogenic/likely pathogenic 2* and LOVDv.3.0: pathogenic) reporting the variant as pathogenic (PP5), disruption of this splice site observed in individuals with NF134,35,36,37,38 (PS4), and the de novo nature of the variant (PS2).

Fig. 1: Pedigree of families 1 and 2.
figure 1

A Family members clinically diagnosed with NF1 are depicted with blackened symbols, and white denotes healthy individuals. In both families, de novo NF1 mutation has originated in the maternal allele. The arrow indicates the index case in each family. B Schematic representation of the NF1 genomic region. NF1 is depicted as an orange box and adjacent genes as black boxes. Arrows adjacent to gene symbols denote transcriptional orientation. NF1-intragenic (orange) and extragenic (blue) polymorphic microsatellites are indicated under the region.

Full size image

Her son (II.1), clinically diagnosed with NF1 (showing > 10 CALMs and lentigines at 1-year-old) carried the same pathogenic variant. The study of NF1 locus microsatellites in this family allowed the identification of which maternal allele II.1 had been inherited.

The multidisciplinary PND center authorized a PND test at the couple’s request. PND was performed after a CVS biopsy. The affected mother’s NF1 c.6641 + 1 G > C variant was directly tested on the fetal DNA extracted from CVS. The presence of the NF1 pathogenic variant was not detected in the fetal DNA while the 17q11.2 microsatellite genotyping and family segregation analysis ascertained the presence of fetal DNA (Fig. 1); the presence of the maternal NF1 c.6641 + 1 G > C variant in the fetus was therefore excluded and the pregnancy was completed. The maternal NF1 locus haplotype genotyped in the fetus matched the haplotype carrying the pathogenic NF1 variant in the mother, while the variant was not identified in the fetus.

Family 2

A woman aged 27 years (Fig. 1: Family 2, individual I.2) with no family history of NF1 was clinically diagnosed with an NF1 with the following clinical features: >10 CALMs, lentigines, one plexiform neurofibroma, and Lisch nodules. The genetic NGS test identified a de novo mosaic pathogenic genetic variant in exon 38 of NF1 (NM_000267.3) c.5624 C > A p.(Ser1875*) with a 41% VAF in two blood samples. Complementary explorations showed a 36% VAF in a saliva sample, and a 37% VAF in a urine sample, confirming mosaicism for NF1. VAF of SNVs is generally between 45% and 55% in non-mosaic heterozygous cases in our laboratory, using targeted NGS.

The following criteria classified the NF1 c.5624 C > A p.(Ser1875*) variant as pathogenic (class 5), according to ACMG-AMP criteria33: null variant (nonsense) (PVS1), absence of the variant in the population database gnomADv3 (PM2), reputable source (ClinVar ID857730: pathogenic 1* and LOVDv.3.0: pathogenic) reporting the variant as pathogenic (PP5), this nonsense variant observed in individuals with NF139,40 (PS4), and the de novo nature of the variant (PS2).

Her son (II.1), clinically diagnosed with NF1 (showing five CALMs, cutaneous neurofibromas, and lentigines at 2-years-old) carried the same NF1 c.5624 C > A pathogenic variant, with a variant allelic frequency (VAF) of 49% using NGS. The study of NF1 locus microsatellites in this family allowed the identification of which maternal allele II.1 had been inherited.

The woman had three pregnancies with a second partner and the multidisciplinary PND center authorized a PND test at the couple’s request for each pregnancy. PNDs were performed after CVS biopsies. The affected mother’s NF1 c.5624 C > A variant was directly tested on the fetal DNAs extracted from CVS. Examination showed that the three female fetuses did not carry the familial NF1 c.5624 C > A pathogenic variant and the pregnancies were completed. The 17q11.2 microsatellite genotyping and family segregation analysis ascertained the presence of fetal DNA (Fig. 1). Unexpectedly, the maternal NF1 locus haplotype genotyped in the third unaffected fetus (individual II.4) matched the haplotype carrying the pathogenic NF1 variant in the mother, while the variant was not identified in the fetus.

Discussion

As NF1 is inherited following an autosomal dominant pattern, the risk of an affected offspring for an NF1 patient is 50% in every pregnancy. To avoid transmission of the disorder, several reproductive options can be discussed with a couple throughout the process of genetic counseling, such as PND or preimplantation genetic testing for monogenic disorders, among others. In both cases, it is necessary that the NF1-causing variant has been determined in the affected parent27.

Genetic PND of monogenic disorders is a process involving the use of a variety of techniques for the molecular characterization of potential monogenic disease in the fetus during pregnancy. The study of familial segregation of polymorphic markers (microsatellites) initially made it possible to carry out PND by indirect diagnosis using linkage analysis when the causal NF1 variant was unknown and the markers informative, and made it possible to identify and monitor the risk haplotype by familial segregation. Advances in genomic medicine, due to the advent of NGS, allowed direct PND by identification of the pathogenic variant in fetal tissues41. The study of polymorphic markers has been maintained to confirm that (i) the tissue studied is indeed of fetal and not maternal origin and (ii) the haplotype corresponds to that expected according to the allele identified by direct genetic test. Indirect analysis uses four intragenic microsatellites that cover about 50 Kb of the gene, as well as three extragenic markers on both sides of the gene to check for recombination events. The microsatellites are highly polymorphic and are also useful for detecting de novo NF1 locus deletion in sporadic cases.

In the present study, we report on our 5-years (2017–2022) experience as a PND laboratory for NF1. The couples received and understood detailed information on the risk of transmission, the high clinical variability, and the clinical features of NF1. The multidisciplinary PND center authorized a PND test, considering the patients’ request and the strong probability and severity of the disease. Among the 205 NF1 PNDs in 146 women, the majority were performed through CVS biopsies vs amniocentesis (88% vs 12%). A higher proportion of unaffected than affected fetuses have been diagnosed (59% vs 41%), and the main outcomes were either the birth of an unaffected child (58%) or pregnancy terminations of affected fetuses (34%, Table 1). Among the 85 pregnancies in which the PND showed that the fetus was a carrier of the pathogenic variant, 16 (18%) were carried to term. The 16 couples consisted of nine familial and seven sporadic NF1 in one parent, with 12 affected women and 4 affected men.

Among NF1 couples requesting PND in our French cohort, the female was affected in most of the cases (Table 1; 87% vs 13%). With genders being equally affected, a possible explanation could be that reproductive counseling gets more attention from women with a hereditary condition compared to men, making women more aware of their reproductive options. Another factor could be the influence of the gender of the affected partner in the reproductive decision-making itself. A previous qualitative interview study showed that in reproductive decision-making between partners, the woman’s influence is considered greater since she is the individual undergoing the treatment and/or carrying the pregnancy42. Finally, some studies have suggested that male fertility might be reduced in NF1 patients, which could also account for a lower proportion of affected fathers43

We observed a different attitude towards PND in women with an NF1 family history (mother or father and/or other family members affected) than in women who represented sporadic cases, as previously described29: the majority of PND requests came from patients with sporadic NF1 in themselves or in a first affected child, and fewer from couples in which one of the parents had inherited the disease (Table 1; 60% vs 40%). In the NF1 family history group, familiarity with the disease, coupled with increased knowledge and a feeling that complications can be managed, could be reasons for not wishing PND. In the sporadic group, the uncertainty of the clinical outcome and the fear of serious complications could explain for requesting PND.

In this report, genetic PND was performed by invasive (CVS or amniocentesis) methods. Recent advances in the exploration of circulating fetal DNA (cfDNA) have made it possible to carry out a non-invasive prenatal test (NIPT)44,45 where the sample is maternal peripheral blood. NIPT is mostly limited to the exclusion of either paternal or de novo mutations. Determination of fetal status with respect to maternal pathogenic variants remains more challenging, because haploidentical maternal and fetal sequences cannot be easily distinguished. Based on the study of maternal haplotype imbalance in cfDNA, relative haplotype dosage (RHDO) was developed to address this challenge. The analysis of the thousands of single nucleotide variants (SNVs) used in RHDO allows both maternal and paternal inheritance to be determined. RHDO assay tracks the inheritance of haplotypes rather than testing for a familial pathogenic variant directly46: SNV counts obtained from cfDNA NGS are used to determine if the fetus has inherited the reference or alternative haplotype by grouping informative SNVs into statistically significant haplotype blocks.

As the NIPT assay tracks the inheritance of haplotypes, it does not account for the possibility that the risk haplotype does not carry the pathogenic variant in the fetus, in the rare context of maternal germline mosaicism. It is important that practitioners using NIPT are aware of this specific and rare point. In the present cohort, two cases (2/146 tested women) illustrated the pitfall for indirect diagnosis that maternal germline mosaics can represent. In families 1 and 2 presented in Fig. 1, the two women with NF1, who opted for PND, had a mosaic NF1. In both families, direct PND analysis (CVS) showed the absence of the maternal NF1 pathogenic variant in the fetus. However, microsatellite markers analysis showed that the risk haplotype (associated with the NF1 pathogenic variant in the affected parent) had been transmitted. These rare cases of germline mosaicism (which are nevertheless observed in a relatively common disease such as NF1 in which half the cases are de novo) illustrate the pitfall of indirect PND or NIPT-RHDO approaches.

Estimating the prevalence and levels of parental mosaicism and consequently, the recurrence risk is important. For an individual family with a child with a genetic disorder caused by an apparent de novo pathogenic variant, the uncertainty about the presence and the level of parental mosaicism makes counseling about recurrence risk challenging and imprecise. Moreover, the recurrence risk also depends on whether the parental mosaic mutation is present in the paternal or maternal germline and on the proportion of germ cells that harbor the mutation47,48. However, only in an affected man, the existence of a germline mosaic can be demonstrated by screening for the NF1 pathogenic variant in a sperm sample47,49, as limitations prevent detection in maternal germ cells.

The PND techniques currently available are varied and increasingly complex, which poses new analytical, legal, and ethical challenges50. These rapid changes in practices offer new possibilities for non-invasive analyses, such as NIPT-RHDO for monogenic diseases of maternal transmission, which aims to adopt an indirect approach without identifying the pathogenic variant. However, certain specific cases (such as dominant diseases with several potentially mosaic de novo cases) need to be considered to prevent the risk of mistakes that could result from the absence of direct identification of the pathogenic variant. The complexity of the workflow for PND and NIPT-RHDO will necessitate increased communication between reproductive medicine specialists, clinical geneticists, obstetricians involved in PND, and the molecular genetics team to ensure that details regarding embryo/fetal genetic diagnosis are understood by all parties involved in the care of these unique cases.

Methods

Study cohort

The study cohort consisted of couples considering PND because of NF1, who were referred to the Cochin Hospital (Assistance Publique - Hôpitaux de Paris, APHP, Paris, France). Databases from NF1 molecular PND at the Cochin Hospital (APHP, Paris, France) were retrospectively analyzed by reviewing the electronic patient files. Between 2017 and 2022, 146 women with a history of molecularly confirmed NF1 in themselves, their partner, or a previous child of the couple underwent 205 PNDs. The study was approved by the Cochin (APHP) local ethical committee and is in line with the current French legislation on genetic studies (ethical approval code: Comités de Protection des Personnes CPP17/79, A0296746, and 2015-08-11DC). Anamnesis, genetic counseling, and genetic testing were performed in-house according to institutional ethical standards and complying with the 1964 Helsinki Declaration and its later amendments. The main study and the collaborated study protocols were approved by the Cochin APHP institutional review board. A signed informed consent was obtained from all the participating patients.

Study samples and nucleic acid extractions

PND samples were obtained by CVS undertaken from 11 weeks of pregnancy or amniocentesis performed from 15 weeks to 16 weeks of pregnancy. Blood samples were collected in EDTA tubes for leukocyte DNA analysis (Becton Dickinson, Rungis, France). DNA extraction was performed with the Maxwell16 LEV Blood DNA Kit (Promega, Charbonnières-les-Bains, France) on EDTA blood samples according to the manufacturer’s instructions. DNA concentrations were quantified using a NanoDrop 1000 Spectrophotometer v3.8 (Thermo Fisher Scientific, Courtaboeuf, France).

Targeted NGS

For the identification of pathogenic variants in probands, experiments were performed at the NGS facility of Cochin Hospital, Paris (AP-HP, France), as previously described51,52. NF1 and SPRED1 exons and flanking intronic regions were amplified with a custom-made panel (Thermo Fisher Scientific) and sequenced on NextSeq500 (Illumina). Sequence alignment, variant calling, and variant annotation were performed using the MOABI pipeline (AP-HP) or the Polyweb pipeline (Imagine Institute, Necker Hospital, AP-HP). Targeted NGS can detect SNVs, small indels, and single- to multi-exon deletions, and allows a molecular diagnosis of more than 90% of patients17,51,52,53.

Sanger sequencing

PCR was performed on DNA extracted from blood samples, trophoblast samples, or amniotic fluid. Amplified exons of the NF1 gene (NM_000267.3) were sequenced by Sanger sequencing using Big Dye Terminator chemistry and an ABI Prism 3130 Capillary Array Sequencer (Applied Biosystems), as previously described13. Primer sequences are available upon request. Sequences were aligned to the reference sequence with SeqScape analysis software v2.5 (Thermo Fisher Scientific).

Quantitative real-time PCR

Single- or multi-exon deletions were evaluated with qPCR on a LightCycler 480 (Roche Diagnostics). DNAs from deleted and non-deleted patients were diluted to a concentration of 2 ng/µL and mixed with LightCycler® 480 SYBR Green I Master Mix and appropriate primers for targeted regions, following the manufacturer’s instructions (Roche Diagnostics)54. The ALB gene was used as a reference for quantitative analysis. Each sample was normalized based on its ALB content and then compared to a non-deleted reference sample. Exons with ratios below 0.6 were considered deleted.

Variant nomenclature and interpretation

Variants were named according to the Human Genome Variation Society (HGVS) recommendations. An assessment of variants’ pathogenicity was performed according to the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMP) guidelines. Assessment of variants implication was mainly performed based on population databases (gnomADv3), variant databases (ClinVar v20220205 and LOVD-NF1), and predictions software. In silico predictions for the effect of the variant were performed with CADD, SPiP, dbscSNV, Human Splice Finder (HSF), and PROVEAN. NF1 variants are reported according to the reference sequence NM_000267.3 (ENST00000356175.7; NF1-201).

Microsatellite typing and segregation analysis

Familial segregation of four NF1 intragenic polymorphic microsatellites (D17S1307, D17S2163, D17S1166, and GDB:270136) and three NF1 extragenic polymorphic microsatellites (D17S841, D17S1800, and D17S798, respectively 1.88 Mb centromeric, and 0.23 and 1.59 Mb telomeric to NF1) was used54. The primer oligonucleotide sequences are available upon request. DNA samples were diluted at a concentration of 10 ng/mL and amplified using dedicated primers and the Taq GOLD polymerase (Thermo Fisher Scientific). The GS-500LIZ (Thermo Fisher Scientific) marker was used for detection. Maternity and paternity were assessed using a PowerPlex 16 HS System (Promega) according to the manufacturer’s instructions. Scaling was controlled with 2800 M Control DNA (Promega). Microsatellite analysis was performed on an ABI Prism 3130 automatic DNA sequencer (Applied Biosystems). The results were analyzed with the GeneMapper v.4.0 software package (Thermo Fisher Scientific).