Marfan syndrome (MFS, MIM # 154700) is a multisystem autosomal dominant connective tissue disorder caused by heterozygous variants of the fibrillin-1 gene (FBN1, 15q21.1). Both dominant-negative effects and haploinsufficiency are reported in its pathogenesis (Aubart et al., 2018) and are expected by the constraint metrics of FBN1 (Z = 5.06; pLI = 1; gnomAD v2.1.1).

Confirmation of FBN1 alterations, mainly missense and loss of function (Lof) variants or rare chromosomal rearrangements (Colovati et al., 2012; Dordoni et al., 2017; Schnause et al., 2021), is achieved in approximately 90% of cases meeting the Ghent II nosology criteria applied for the clinical diagnosis of MFS (Loeys et al., 2010; Zeigler et al., 2021). Alterations of FBN1 linked to MFS have also been identified outside the FBN1 coding region and were proven to be causal through functional analysis (Guo et al., 2023). However, in approximately 10% of patients with distinctive clinical signs of MFS, including positivity to GHENT II criteria, no alteration of FBN1 is detectable by routine DNA investigation. In this regard, the familial case reported by Pagnamenta (Pagnamenta et al., 2022) is exemplary: a 1.97 Mb inversion with a distal breakpoint in intron 4 of FBN1 was highlighted after years of genetic testing in the mother and proband, who were recruited with a diagnosis of “familial thoracic aortic aneurysm disease.” We present a similar case of a male boy long suspected of suffering from MFS complicated by ID. A de novo complex rearrangement involving chromosomes 3, 4, and 13 was discovered at amniocentesis; however, the search for FBN1 alteration explaining his postnatal MFS phenotype gave negative results after MLPA, array-CGH, and gene panel sequencing. The involvement of chromosome 15 with a copy-neutral insertion of a portion of FBN1 at the breakpoint of chromosome 3 was eventually detected by OGM. This case not only confirms the superiority of this technological approach in revealing structural variants (SVs) but also stresses the burden of the copy-neutral structural variants underlying genetic disorders.

There is growing evidence that OGM can integrate all forms of SV, even throughout complex loci, thanks to the uninterrupted assembly of long-range molecules, allowing anchoring and resolving of most SVs regardless of sequence composition (Porubsky et al., 2022).

Indeed, in the case we studied, the insertional translocation of 2.7 Mb leading to the breakage of FBN1 has been detected thanks to OGM and, after 7 years of vain investigations, explained the reasons for the main patient’s features, i.e., those coinciding with the Marfan phenotype. On the other hand, no clear basis for the patient’s ID could be highlighted. PE-WGS revealed no like-pathogenic or pathogenic SNVs in known ID-associated coding genes, thus pointing to the other genes altered by the rearrangement, namely, ROBO2, PARP4, and RNF17. ROBO2, which was partially lost as a consequence of the 3p deletion, and is of interest, being mainly expressed in the brain (GTEx, V6 release) and intolerant to Lof variants (pLI = 1; o/e = 0.08; gnomAD v2.1.1). Furthermore, a significant association of rs7642482, near ROBO2-3’, with expressive vocabulary in infancy was demonstrated (St Pourcain et al., 2014), whereas a decreased expression was observed in the brains of individuals with autism spectrum disorders (Suda et al., 2011). These findings are in agreement with the ROBO2 function as a receptor for SLIT2, and probably SLIT1, which are thought to act as a molecular guide in cell migration, including axonal navigation at the ventral midline of the neural tube and axon projection in different regions during neuronal development (RefSeq NM_002942). To date, heterozygous ROBO2 pathogenic variants have been associated with autosomal dominant VUR2. The ultrasound scan of the patient’s abdomen, specifically requested to highlight any possible alteration of the kidney and urinary tract, associated with the partial loss of ROBO2, did not reveal any abnormality, and no clinical signs such as recurrent urinary tract infections were evident.

SNVs associated with VUR2 are located along the entire gene (Supplementary Figure S8) and are mainly of the missense type, suggesting that VUR2 syndrome is the result of a dominant-negative effect. In our patient, the deletion involving ROBO2 removes exons 1–14 (Figure 3E; Supplementary Figure S1), which encode most of the extracellular domains of the protein (Supplementary Figure S8). Analysis of the spared sequence revealed the presence of a start codon at position +18 of exon 15, possibly indicating that the C-terminal portion of the protein is translated. This portion retains the transmembrane domain and cytoplasmic region, which is characterized by three intrinsically disordered regions (Supplementary Figure S8). It is, however, difficult to speculate on the effective production and functionality of this truncated protein. Even if translated, the protein is predicted to lack the signal peptide, a short N-terminal sequence that drives trafficking to the cell membrane via the endoplasmic reticulum (Guna and Hegde, 2018). Mutations or deletions in the signal peptides of other human proteins were shown to result in reduced protein targeting to the cell membrane, retention in the endoplasmic reticulum, or very low-level intracellular expression (Albers et al., 2014; Uetz-von Allmen et al., 2018; Potorac et al., 2019). In any case, even if translated and spared from degradation, the protein would miss the extracellular domains required for interaction with SLIT2, an important mediator of neuronal migration (Bagri et al., 2002). Thus, in our patient, the deletion is likely to have resulted in a reduced amount of functional ROBO2 protein, which may have affected his neurological development by impairing the axon-guiding function, as demonstrated in the anterior cingulate cortex (Suda et al., 2011) and in lymphocytes of individuals with autism (Anitha et al., 2008).

PARP4 and RNF17, which are also broken by the transposition from 13q to 3p, are not so far disease-associated, and both of them are almost not expressed in the brain. The good tolerance of PARP4 to Lof variants (pLI = 0; gnomAD v2.1.1) makes its involvement in the intellectual disability of the patient unlikely. In contrast, RNF17 appears highly intolerant to Lof variants (pLI = 1; o/e = 0.03; gnomAD v2.1.1); however, according to its expression limited to testis, may be involved in spermiogenesis only. In conclusion, an explanation for the ID of the patient is missing, although ROBO2 appears to be the best candidate. Moreover, we must consider that the reshuffling of topologically associating domains (TADs) may have caused misexpression of intact genes around the breakpoints, especially those within 100 kb (Schöpflin et al., 2022). In silico analysis of the three-dimensional organization of chromatin (Supplementary Figure S9) showed that none of the identified rearrangement breakpoints altered TADs organization or occurred within highly conserved non-coding sequences associated with developmental regulators (Lowther et al., 2022). However, we cannot rule out that other genes in the 100 kb surrounding breakpoints have unbalanced allelic expression (Schöpflin et al., 2022).

Our patient has received a clinical diagnosis of MFS since he was 2^6/12 years old. However, the lack of confirmation at the molecular level and the concomitant presence of ID created uncertainties about the real cause of the clinical condition. OGM, which was performed 5 years after the first molecular investigations ended the diagnostic odyssey by showing that the patient was actually suffering from MFS due to the transposition of part of the FBN1 gene from 15q to 3p, even if the basis of the patient’s ID remains vague and considering possible alterations of the TADs as a consequence of the rearrangement’s breakpoints. While more evidence is needed, ROBO2 appears to be the best candidate for the ID observed in our patient.

Our study further emphasizes the role of neutral SVs in missing diagnoses of Mendelian disorders, including those that are clinically and molecularly confident. This is the case for subchromosomal-size inversions for which OGM has been shown to be much more frequent than previously estimated (Porubsky et al., 2022). OGM highlights that even copy-neutral insertions increase the burden of genetic disorders, demonstrating that some of them are found in regions refractory to sequencing (Sabatella et al., 2021; Yang and Hao, 2022; Zhang et al., 2023). A separate condition concerns the so-called copy-neutral loss-of-heterozygosity (CN-LoH). This condition has been reported in aging, cancer, and increasingly in congenital disorders. In all cases, variants with a lower cell fitness, disease variants, or chromosomal imbalances can be removed by somatic recombination, resulting in segmental uniparental disomy and CN-LoH (Loh et al., 2020; Papenhausen et al., 2021). CN-LoHs are detected through trio SNP analysis and are unquestionable of postzygotic origin. However, the expected mosaicism with the cell line presumably present in the zygote or at early embryogenesis is not always detected, sometimes making it difficult to correlate the CN-LoH region with the patient’s phenotypic abnormalities. At least in the case of CN-LoHs in mosaic with SVs for the same chromosome regions, OGM technologies could indeed be superior to other sequencing techniques (Noyes et al., 2022).

This study provides further overwhelming evidence that chromosomal rearrangements, although not necessarily complex, require OGM testing, even before any other molecular investigation. This approach might be suitable both in the presence of congenital disorders and in apparently healthy infants, considering the long-term morbidity that can be unpredictable at birth (Li et al., 2014; Halgren et al., 2018; Fjorder et al., 2019).

Even if the breakpoint sequence is not provided, the OGM software we used can capture and highlight complex genomic rearrangements in a very intuitive and effective manner. The combination of the CNV and SV pipelines allows one to both highlight unbalanced variants >500 bp and to reconstruct the final order and orientation of the displaced regions. In contrast, short-reads genome sequencing, without knowing a priori which regions have to be investigated, requires endless manual visual inspection with the risk of losing regions masked by segmental duplications/high copy repeats that are not captured by the methodology. Based on the present study, the use of OGM before any other molecular analysis is recommended for the complex chromosomal rearrangement, and perhaps for the apparently simple ones, both associated with congenital clinical disorders or present in apparently healthy newborns, taking into consideration long-term morbidity that can be unpredictable at birth (Li et al., 2014; Halgren et al., 2018; Fjorder et al., 2019). This study also emphasizes the need to approach the diagnosis of genetic diseases in the context of the entire genome rather than key genes, as is common practice for most medical classes (Mahmoud et al., 2023).