Expanding the mutational spectrum of ReNU syndrome: insights into 5’ Stem-loop variants

Abstract

A narrow spectrum of heterozygous variants in RNU4-2, encoding the small nuclear RNA (snRNA) U4, underlies ReNU syndrome, a neurodevelopmental disorder (NDD) characterized by moderate to severe developmental delay (DD), intellectual disability (ID), a distinctive facial gestalt, and multisystem involvement. Pathogenic variants have primarily been reported within an 18-nt critical region contributing to stabilizing the U4/U6 snRNA duplex and proper spliceosome assembly. By combining whole genome sequencing reanalysis and targeted direct sequencing in 190 molecularly unexplained NDD cases, we report on five affected individuals carrying pathogenic/putative pathogenic RNU4-2 variants (2.6%). Three individuals harbored the recurrent pathogenic n.64_65insT variant, while two were heterozygous for private/rare maternally inherited variants (n.30 A > T and n.43_44insT) within the 5’ Stem-loop region. Deep clinical phenotyping confirmed a homogeneous constellation of features in all individuals, with global DD, ID, brain malformations, and a recognizable facial gestalt representing core findings. Based on structural homology models and available cryo-EM data, n.30 A > T and n.43_44insT were predicted to disrupt key intra- and inter-molecular interactions critical for spliceosome function. Our findings expand the mutational spectrum of ReNU syndrome, and confirm the 5’ Stem-loop as a second mutational hotspot in RNU4-2. We propose that a more complex genetics likely underlies the inheritance of a subset of disease-causing RNU4-2 variants from an apparently unaffected parent. We anticipate a relatively high proportion of pathogenic RNU4-2 variants among individuals with unclassified NDD despite extensive genomic testing, and propose a set of facial gestalt core features as a clinical screening tool to prioritize patients for RNU4-2 analysis.

Introduction

Neurodevelopmental disorders (NDDs) constitute a broad group of clinically diverse and genetically heterogeneous conditions characterized by deficits in cognitive functions, language impairment, behavioral problems, and psychomotor skills delay. In a significant proportion of cases, developmental delay (DD) and intellectual disability (ID) occur in the context of a complex phenotype characterized by multisystem involvement. Current estimates suggest that approximately 60% of individuals with NDD does not receive a conclusive molecular diagnosis even after extensive genomic testing [1, 2]. Reasons for this so-called “missing heritability” are partially linked to current difficulties in effectively interpreting and prioritizing variants within non-coding portions of the genome (e.g., intergenic, regulatory, and intronic regions) or genes that do not code for proteins.

Recently, pathogenic variants in the non-coding small nuclear RNA (snRNA) 4-2 gene (RNU4-2, NR_003137.3) have been recognized as an emerging cause of syndromic DD/ID These variants underlie ReNU syndrome (MIM #620851), a NDD characterized by moderate to severe global DD, ID, language impairment, abnormal brain morphology, ophthalmological features, and a recognizable facial gestalt [3–6]. Based on the available molecular epidemiology data, this disorder is estimated to be one of the most prevalent NDDs, accounting for up to 0.4% of cases, and possibly higher when considering subjects with severe syndromic conditions [3]. RNU4-2 encodes the U4 snRNA, a core component of the major spliceosome, a complex required for pre-mRNA splicing of U2-type introns [7, 8]. U4 snRNA secondary structure is characterized by seven distinct functional regions: Stem II (n.1-n.15), 5’ Stem-loop (n.16-n.56), Stem I (n.57-n.61), T-loop/quasi-pseudoknot region (n.62-n.67), RNA-binding protein 42 (RBM42) interacting domain (n.68-n.70), Stem III (n.73-n.79), and 3’ Stem-loops (n.80-n.141) (Fig. 1) [7, 8]. U4 and U6 snRNAs bind each other through complementary base pairing to form the U4/U6.U5 tri-snRNP. Studies in yeast showed that the U4/U6 snRNA duplex is complexed with five core proteins, Snu13, Prp3, Prp4, Prp6 and Prp31, via an extensive H-bonding network, with the U4 5’ Stem-loop playing an essential role [7, 8]. Among the major intermolecular interactions involving U4 snRNA, a stretch encompassing the T-loop, RBM42 interacting domain and Stem III is crucial in the U4/U6 duplex to allow the U6 ACAGAGA motif to recognize the 5’ splice site (SS) during the first step of spliceosome assembly [7]. Consistently, heterozygous variants within this ribonucleotide stretch, also defined as the 18-nt “critical region”, have recently been identified as the molecular event underlying ReNU syndrome [3–6]. A second relevant region in U4 snRNA is represented by the 5’ Stem-loop, which interacts with various key residues at the N-terminus of Snu13 (Glu³⁹, Lys⁴² and Arg⁴⁶), the N-terminal α-helix of Prp31, and with α-helices 6 and 39 of Prp6 [7]. Notably, a number of clinically unclassified variants involving this region has been reported [3, 9], whose functional impact and clinical relevance have not been investigated, yet.

Fig. 1. Two-dimensional predicted structure of U4/U6 snRNAs duplex.

The cartoon shows the U4 ribonucleotides defining the Stem II (n.1 to n.15, grey), 5’ Stem-loop (n.16 to n.56, gold), Stem I (n.57 to n.61, blue), T-loop (n.62 to n.70, dark red), RBM42 interacting region (n.71 and n.72, magenta), Stem III (n.73 to n.79, purple) and 3’ Stem-loop (n.80 to n.141, green). The U6 ribonucleotides are also shown (white) with the ribonucleotides defining the ACAGAGA loop (red). The variants identified in this study are indicated by red arrows.

Here, we describe five individuals carrying heterozygous RNU4-2 variants involving the T-loop and 5’ Stem-loop regions. Two subjects were identified by reanalysis of whole-genome sequencing (WGS) data available for 85 molecularly unexplained cases with NDD. Three additional subjects were identified by direct Sanger sequencing assessing 105 individuals from a highly selected cohort of undiagnosed patients with DD/ID who had previously been analyzed by exome sequencing (ES). Three subjects harbored the recurrent n.64_65insT as a de novo event, while two individuals were heterozygous for previously unreported variants located in the 5’ Stem-loop region, which were transmitted from their apparently healthy mothers. Clinical features of the latter well fitted the phenotypic spectrum characterizing ReNU syndrome, including the core group of signs/features defining the facial gestalt of the disorder. Structural analyses predict a disruptive role of these variants. Finally, we confirm the robustness of the published clinical findings and provide a core group of recognizable signs defining the ReNU syndrome facial gestalt.

Materials and methods

Study cohort

A cohort of 190 subjects affected with a molecularly unclassified NDD was included in the study. All probands showed moderate to severe DD/ID and language impairment. The Human Phenotype Ontology (HPO) terms that were used for the selection of the enrolled cases are provided in Table S1. All individuals had previously undergone single nucleotide polymorphism (SNP)/comparative genomic hybridization (CGH) microarray analysis and ES. Among them, 85 subjects had also been analyzed by trio-based WGS. Clinical and biochemical data, neuroimaging, and blood samples were collected, stored, and utilized in accordance with the ethical standards outlined in the Declaration of Helsinki. Informed signed consents were obtained from all participating families in compliance with institutional review board regulations. The study was approved by the Institutional Ethical Committee of Ospedale Pediatrico Bambino Gesù, Rome (1702_OPBG_2018 and 2072_OPBG_2020).

WGS reanalysis

Trio-based WGS data were obtained using a 2×150 bp paired-end read protocol on a NovaSeq 6000 platform (Illumina, San Diego, CA, USA) to guarantee a 30-fold median coverage. Base calling and data analysis were performed using Bcl2FASTQ (Illumina). Paired-end reads mapping to the GRCh38 reference sequence, small variant calling, and joint genotyping were run using Sentieon v.2023-08 (https://www.sentieon.com). Hard filtering was applied using Genome Analysis ToolKit (GATK) v.3.8.0 (Broad Institute, Cambridge, UK) following the best-practice pipeline guidance [10, 11]. WGS data from the selected individuals were queried for any variant at the RNU4-2 locus (hg38 genomic coordinates: chr12:120,291,763-120,291,903).

Targeted sequencing

A PCR-based assay was designed to scan variation encompassing the transcribed RNU4-2 sequence and their upstream flanking region in the subcohort of subjects with uninformative ES due to lack of RNU4-2 coverage. Genomic DNA was isolated from circulating leukocytes, PCR-amplified and sequenced using the following primers: RNU4-2_F: 5’-AAACAACTCACAGTACCCGC-3’; RNU4-2_R: 5’-GTTTTCAAAACCTGAATGTTTGACACG-3’. KAPA2G HotStart ReadyMix (Merck, Darmstadt, DE) was used for amplification, following the manufacturer’s protocol. Sanger sequencing reactions were carried out using the BigDye Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher Scientific, Waltham, MA, USA) on a SeqStudio sequencer (Thermo Fisher Scientific).

Clinical assessment

Clinical assessment included all features reported in Barbour et al. [5]. Deep phenotyping was mainly focused on facial features to identify any core sign that could be considered informative to suspect ReNU syndrome on a gestalt basis. Using this approach, all previously reported individuals having a detailed description or clinical pictures were independently reassessed by three experienced clinical geneticists (FCR, MN, and MP).

Structural analysis

The structural interactions between the U4/U6 snRNA duplex and the core proteins PRPF3, PRPF4, PRPF6, PRPF31 and SNU13, within the U4/U6.U5 tri-snRNP complex (PDB ID: 6QW6) [12], were investigated using UCSF Chimera software v.1.17.3 (https://www.cgl.ucsf.edu/chimera) [13]. Inspection focused on the inter/intra-molecular interactions (salt bridges and H-bonding network) involving the U4 snRNA bases A30, G43, and A44 (reference sequence), U30 (n.30 A > T), U44, and A45 (n.43_44insT), and C37 (n.37 T > C), as well as the polar amino acid residues located at the interface of the complex. U4 structures referring to the n.30 A > T, n.43_44insT, and n.37 T > C mutated sequences were generated by homology modeling using the ModeRNA server (https://genesilico.pl/modernaserver/).

The structure of ribonucleotides 18-54 of the U4 snRNA 5’ Stem-loop was modeled using the AlphaFold Server (https://golgi.sandbox.google.com/) powered by the AlphaFold 3 model [14]. Nineteen modeling replicas were performed (wild-type sequence, N = 10; n.30 A > T, N = 10; n.43_44insT, N = 3; n.37 T > C, N = 3) in the absence of any other nucleic or protein structural elements, selecting a different seed each time to improve prediction accuracy. All replicas were then ranked by score (“ranking_score” as defined by AlphaFold), and the best scoring models for each subset were employed as reference for the root mean square deviation (RMSD) calculation.

Results

Genomic findings

We investigated the occurrence of putative clinically relevant variants within the RNU4-2 gene (NR_003137.3) using two different approaches. First, we reanalyzed WGS data in a subgroup of 85 individuals affected with an apparently molecularly unsolved isolated/syndromic NDD. This in silico scan allowed us to identify two patients sharing the de novo n.64_65insT (Subjects 1 and 2), which had previously been reported as the most common pathogenic RNU4-2 variant [3, 4]. To investigate further the prevalence of RNU4-2 mutations, we performed targeted sequencing in a second subgroup of 105 individuals with isolated/syndromic DD/ID who had remained molecularly unsolved after ES analysis. In these patients, ES data could not be used to analyze the occurrence of variation within the RNU4-2 gene as the target region was not captured by the used enrichment kits. Sanger sequencing allowed to detect one additional case carrying the recurrent n.64_65insT causative variant as a de novo event (Subject 3). We furthermore identified nine variants: n.13 T > A (n = 1), n.30 A > T (n = 1), n.37 T > C (n = 3), n.41 C > T (n = 1), n.42 C > T (n = 1), n.43_44insT (n = 1) and n.92 C > T (n = 1). For each variant, frequency in publicly available population databases, inheritance pattern, and in silico prediction score were assessed (Table S2). Among these variants, only those being private or having rare occurrence in public databases (<10 alleles in gnomAD v.4.1.0 and All of Us) were retained. No significant differences in CADD PHRED and PhyloP100way scores between the previously published bona fide pathogenic variants and those reported in gnomAD v.4.1.0 had been observed. Consequently, these in silico tools were not included as filtering criteria. The three subjects who were heterozygous for the retained variants (n.43_44insT, n.30 A > T, and n.13 T > A) underwent reverse-phenotyping to verify the consistency of their phenotype with the facial gestalt and major clinical features characterizing ReNU syndrome. An extensive clinical overlap was observed for Subjects 4 and 5 (Fig. 2, Table 1, Table S3, Supplemental Clinical Reports). Specifically, both individuals showed severe global DD/ID, language impairment, abnormal brain morphology, including corpus callosum hypoplasia, ventriculomegaly and/or white matter abnormalities, ophthalmological anomalies, and gait dysfunctions. Moreover, their facial features fitted the recognizable facial gestalt of ReNU individuals [3, 5, 15]. Segregation analysis documented that both variants were inherited from the apparently healthy mothers, a finding previously reported in ReNU-causing variants [3, 4, 9]. Differently, the subject carrying the n.13 T > A variant presented with more severe clinical features and a distinct facies. While a possible contributing role of this variant to the observed phenotype in this individual cannot be ruled out in principle, additional molecular investigations are required to exclude the occurrence of other pathogenic variants underlying his condition.

Fig. 2. Clinical and instrumental findings of the five individuals with heterozygous RNU4-2 variants.

A Clinical features. Subject 1 (n.64_65insT) at the age of 6 and 27 years. Subject 2 (n.64_65insT) at 23 years. Subject 3 (n.64_65insT) at 6 years. Subject 4 (n.43_44insT) at 7 years. Subject 5 (n.30 A > T) at the age of 2 and 6 years. Note the hypotonic appearing face, broad sparse eyebrows, open mouth, horizontally oriented philtrum with tented upper lip, flattened cupid’s bow, everted lower lip, and full cheeks. Acral cutis marmorata is appreciable in Subject 1. B Brain MRI findings. Subject 4 (top panel): coronal T1 inversion recovery (a), axial T2 (b) and coronal T2 (c) fat-saturated images. Microcephaly is associated with gyral simplification and thick cortex, various cystic-like lesions and signal hyperintensities in the white matter of centrum semiovale and corona radiata, and asymmetric and enlarged lateral ventricles. Sagittal T1 image (d) showing slightly verticalized tentorium with enlarged cerebellar vermis interfolial spaces. A thin corpus callosum, and enlarged cerebellar hemispheric interfolial spaces with abnormal paravermian folial orientation are also observed. Subject 5 (bottom panel). Sagittal T1 image (e) showing mild hypoplasia of the hystmus and splenium of corpus callosum, and mild cerebellar vermis hypoplasia. Axial T2 images (f, g) showing supratentorial white matter reduction and dilated perivascular spaces; plagiocephaly can also be noted. Coronal T1 inversion recovery image (h) showing disorganized cerebellar hemispheric folia (dysplasia) and supratentorial white matter reduction.

Table 1.

Major clinical features and core facial features characterizing the five individuals with RNU4-2 variants.

Phenotypic features
	Present study					Barbour et al., 2024	Schot et al., 2024	Rosenblum et al., 2024	Total
	Subject 1	Subject 2	Subject 3	Subject 4	Subject 5	(4 subjects)	(1 subject)	(4 subjects)
Sex and age at last evaluation	F, 30 years	F, 25 years	F, 6 years	F, 9 years	M, 6 years
Variant (GRCh38)	12:120291839 T > TA	12:120291839 T > TA	12:120291839 T > TA	12:120291860 T > TA	12:120291874 T > A	12:120291839 T > TA	12:120291839 T > TA	12:120291839 T > TA
HGVS	n.64_65insT	n.64_65insT	n.64_65insT	n.43_44insTa	n.30 A>Tb	n.64_65insT	n.64_65insT	n.64_65insT
Mutation class	insertion	insertion	insertion	insertion	SNV	insertion	insertion	insertion
Inheritance	de novo	de novo	de novo	maternal	maternal	de novo	de novo	de novo
RNU4-2 region	T-loop	T-loop	T-loop	5’ Stem-loop	5’ Stem-loop	T-loop	T-loop	T-loop
Main clinical features
Microcephaly	+	+	+	+	-	4/4	+	3/4	12/14
DD	+	+	+	+	+	4/4	+	4/4	14/14
ID	+	+	+	+	+	4/4	+	4/4	14/14
Speech abnormality	+	+	+	+	+	4/4	+	4/4	14/14
Hypotonia	+	+	+	-	+	4/4	+	4/4	13/14
Abnormal brain MRI	+	+	+	+	+	4/4	+	3/4	13/14
Ophthalmological abnormalities	+	+	+	+	+	4/4	-	4/4	13/14
Dysmorphic phenotypic features
Thin body habitus	Yes	No	Yes	No	Yes	4/4	No	2/3	9/13
Hypotonic appearing face	Yes	Yes	Yes	Yes	Yes	4/4	Yes	3/3	13/13
Sloping forehead	Yes	Yes	Yes	Yes	No	2/4	with age	0/3	7/13
Epicanthal folds	No	No	Yes	Yes	Yes, monolateral	3/4	Yes	4/4	11/14
Broad sparse eyebrows	Yes, medially	Yes, laterally	Yes, medially	Yes	Yes, laterally	3/4	Yes, medially	3/4	12/14
Anteverted nares	No	No	Yes	Yes	Yes	2/4	Yes	3/3	9/13
Flared nares	No	No	Yes	Yes	Yes	2/4	Yes	3/4	9/14
Low-hanging columella	Yes	No	No	Yes, mildly	No	3/4	Yes	1/4	7/14
Large appearing ears	Yes	Yes	Yes	Yes	No	3/4	Yes, with age	2/4	10/14
Full cheeks	Yes	No	Yes	Yes	Yes	3/4	Yes	2/4	10/14
Open mouth	Yes	Yes	Yes	Yes	Yes	4/4	Yes	4/4	14/14
Horinzontally oriented philtrum	Yes	Yes	Yes	Yes	Yes	4/4	Yes	4/4	14/14
Tented upper lip	Yes	Yes	Yes	Yes	Yes	4/4	Yes	4/4	14/14
Flattened cupid’s bow	Yes	Yes	Yes	Yes	Yes	4/4	Yes, mild	3/3	13/13
Everted lower lip	Yes	Yes	Yes	Yes	Yes, at young age	4/4	Yes	4/4	14/14
High arched palate		Yes		Yes	Yes	3/4			6/7
Retro/micrognathia	Yes	Yes	Yes	Yes	Yes	3/4	Yes	1/3	10/13
Hypoplastic distal phalanges	No	No		Yes	No			3/4	4/8
Tapered fingers	No	Yes		Yes	Yes	2/4		3/4	8/12
Clinodactyly	Yes	No		No	Yes	2/4		1	5/7
Narrow hyperconvex nails	No	Yes		Yes	No	2/4		2/3	6/11
Hypoplastic nails	No	No		Yes	No	2/4			3/8
Broad hallux	Yes	Yes		No	No	2/4		4/4	8/12

^aThe variant and its clinical association has been submitted to ClinVar (VCV003390370.1).

^bThe variant and its clinical association has been submitted to ClinVar (VCV003390371.1).

Structural analysis

n.30 A > T and n.43_44insT are located outside the 18-nt “critical region”, as defined by Chen and colleagues [3], and map within the 5’ Stem-loop stretch [7]. To investigate their functional consequences on the U4 snRNA intra- and intermolecular binding network within the U4/U6.U5 tri-snRNP complex, structural analyses were carried out. First, we assessed the perturbing effect of these variants on the intramolecular interactions stabilizing the U4 snRNA 5’ Stem-loop structure. To this goal, folding of the WT, and mutated 5’ Stem-loop core stretches were generated using AlphaFold 3. The most common polymorphism within this stretch, n.37 T > C (143 allele counts in gnomADv.4.1.0), was also considered in the analysis. The replicas documented a well-conserved folding for the WT and n.37 T > C 5’ Stem-loop structures (mean RMSD = 2.082 Å ± 0.879 and 1.098 Å ± 0.372, respectively), which superimposed almost perfectly (RMSD = 1.663 Å ± 0.703) (Fig. 3A, B). In the WT structure, two major assemblies were predicted: H-bonding between A30 and A44 was observed in 60% of replicas, while the two ribonucleotides were placed on different planes and unable to effectively interact in the remaining replicas. Differently, G43 stably interacted with A33 and the ribose of A44. When considering the n.30 A > T and n.43_44insT models, the replicas confirmed a relatively well-conserved folding (mean RMSD = 2.074 Å ± 0.838 and 1.040 Å ± 0.218, respectively), which however showed a higher structural deviation from the WT model (RMSD = 4.064 Å ± 0.776 and 10.227 Å ± 0.283, respectively) (Fig. 3C, D). Of note, the ten replicas for n.30 A > T mutant showed a substantial deviation from the more represented structure of the WT U4 Stem-loop core stretch, with all replicas having a conformation displacing the H-bonding between A30 and A44 and weakening the interaction between G43 and the pentose component of A44. Regarding the three replicas for the n.43_44insT mutant, the insertion of U44 was observed to introduce two stable interactions (A30-U44 and U31-G43) not occurring in the WT model. These findings suggest a relevant local rearrangement of the region in both mutants that is expected to affect the stability and/or function of the entire U4/U6.U5 tri-snRNP complex.

Fig. 3. Superposition of 3D models of the U4 snRNA 5’ Stem-loop (ribonucleotides 18-54) generated by AlphaFold 3.

A Folding replicas (N = 10) of the WT sequence are shown in tan (top, left). Magnifications show the two alternative assemblies predicted for the ribonucleotides A30 and A44. Folding replicas of the n.37 T > C (coral) (N = 3) (B), n.30 A > T (sky blue) (N = 10) (C) and n.43_44insT (orchid) (N = 3) (D) systems are matched to the best scoring model of the WT sequence (tan). Mutated ribonucleotides are magnified in the boxes.

Structural inspection of the cryo-EM maps available for the U4/U6.U5 tri-snRNP complex revealed that ribonucleotide G43 contributes to stabilizing the interaction between the U4 5’ Stem-loop with the core protein SNU13 (Fig. 4). Specifically, G43 directly H-bonds SNU13 Lys⁴⁴ side chain via its O6 and N7, and is predicted to act as donor in the H-bond with either OE1 or OE2 of the SNU13 Glu⁴¹ side chain (Table S4). Locally, A30 and A44 contribute to the U4 intramolecular interactions that maintain G43 in the proper orientation, allowing its binding to SNU13. By introducing an additional uracil in the 5’ Stem-loop, n.43_44insT variant is predicted to hinder proper positioning of G43 and A44. The misalignment of the G43-O2’ bond with SNU13 Glu⁴¹ and G43-N7 interaction with SNU13 Lys⁴⁴, in turn, is expected to lead to a conformational change altering the 5’ Stem-loop scaffold function and impairing spliceosome’s assembly and efficiency (Fig. 4). Likewise, n.30 A > T is predicted to disrupt the H-bond between A30 and A44, weakening the 5’ Stem-loop secondary structure. Interestingly, both variants might locally favor a canonical Watson-Crick interaction involving A30 (by binding to U44, in the n.43_44insT U4 snRNA) and A44 (by binding to U30, in the n.30 A > T U4 snRNA), which are in close contact, facing each other in the 5’ Stem-loop (Fig. 4). A conformational shift may interfere with the appropriate recruitment of the U4/U6 snRNP core proteins, resulting in improper spliceosome assembly or inaccurate splicing site recognition.

Fig. 4. Intramolecular and intermolecular interactions of the A30, G43, A44 ribonucleotides of the 5’ Stem-loop in the U4/U6.U5 tri-snRNP complex generated using UCSF Chimera.

The structure of the complex (PDB ID: 6QW6) is shown with the U4 (gold)/U6 (green) snRNA duplex and the core proteins PRPF6 (light pink), PRPF31 (light purple) and SNU13 (light blue). The relevant interactions involving ribonucleotides A30, G43, A44, and SNU13 E⁴¹ and K⁴⁴ residues are magnified in the boxes, together with the alternative U4 ribonucleotides.

Deep phenotyping analysis

By deep phenotyping, we clinically compared the five probands with nine individuals previously reported in literature for whom a detailed description or clinical picture were available [5, 15, 16] and confirmed a significant overlap in clinical features and facial gestalt, and a similar occurrence of individual signs/features (Tables 1 and  S3, and Fig. 2). We identified a core of facial features occurring in the majority of the affected individuals (75% to 100%), which included hypotonic appearing face (13/13), epicanthal folds (11/14), broad sparse eyebrows (12/14), anteverted nares (9/13), open mouth (14/14), horizontally oriented philtrum with tented upper lip (14/14), flattened cupid’s bow (13/13), and everted lower lip (14/14). Full cheeks were observed in 10/14 individuals (71%), although this sign appeared as one of the most recognizable traits in younger individuals. The four subjects not showing this sign were all adults/young adults (Subject 2, present report; Subject 2, [5]; Subjects 1 and 2, [15]), suggesting a correlation of this trait with age. Consistently, Subject 1 of the present report showed full cheeks at 6 years, a feature that was not evident at 26 years (Fig. 2A). Notably, the two individuals in the present report carrying the inherited variants within the 5’ Stem-loop perfectly scored positive for all the core facial features.

Retrognathia has been reported as a common craniofacial feature in ReNU syndrome. We observed the co-occurrence of retrognathia and micrognathia in almost all the affected individuals of the present cohort (Table 1). By reassessing the clinical pictures of the previously published individuals, we noticed that the latter also represents a common finding, occurring in the majority of individuals, with the exception of individual 3 reported by Barbour et al. [5], and individuals 3 and 4 in Rosenblum et al. [15].

Discussion

By performing a combined strategy based on WGS data reanalysis and direct sequencing of RNU4-2 in a selected cohort of 190 individuals affected with molecularly unexplained NDDs, we report on the identification of pathogenic and putative disease-causing variants in five subjects (2.6%), further documenting the clinical relevance of RNU4-2 in the context of NDD etiology, defining the 5’ Stem-loop as a second RNU4-2 mutational hot-spot underlying ReNU syndrome.

The majority of affected individuals carrying de novo heterozygous variants in RNU4-2 share a recurrent base pair insertion (n.64_65insT) [3, 4]. Our findings confirm the high prevalence of this recurrent change. Additional variants encompassing the 18-nt “critical region” have causally been associated with ReNU syndrome [3, 4, 9]. These nucleotide changes have been predicted to induce steric conformational changes in the T-loop, likely disrupting the U4/U6 interaction and impairing the correct positioning of the U6 recognition motif to the 5’ SS, and were demonstrated to lead to aberrant splicing events [3, 4, 9]. Thus far, there has been comparatively little focus on variants of RNU4-2 that fall outside the 18-nt “critical region”, as only a limited number of de novo variants have been identified [3, 4, 9]. Among these, two nucleotide changes, n.42 C > G and n.45_46insT, were described within the 5’ Stem-loop [3, 9]. Within this region, a large proportion of variants identified in NDD patients has been reported to be inherited [3, 4, 9], suggesting the possibility of incomplete penetrance or a more complex genetics. Incomplete penetrance is a well-established mechanism in the context of syndromic/isolated DD/ID (i.e., ASXL1 [17], ASXL3 [17], ARID1B [17], SETD5 [18, 19], ANKRD11 [17, 20], AUTS2 [17, 21], and SMC3 [22]). Notably, incomplete penetrance is also an emerging feature in disorders causally linked to variants involving RNU genes [4, 23, 24]. While most of the variants in the 18-bp critical region of RNU4-2 have been reported as de novo events, the original work by Chen and coworkers identified 29 additional affected individuals carrying putative pathogenic variants in this region with an unknown/inherited segregation pattern, implying that some variants may indeed be transmitted [3]. Similarly, Greene and colleagues reported other variants affecting the 5’ Stem-loop with a large proportion of affected individuals (>60%) carrying an inherited variant [4]. Only a small proportion of cases (<10%) were reported to harbor a de novo RNU4-2 variant. Overall, variants within the 18-bp critical region are largely de novo and mainly associated with full penetrance, while variants affecting the RNU4-2 5’ Stem-loop region, which are also likely to contribute to ReNU syndrome, commonly exhibit incomplete penetrance. Of note, a bunch of novel pathogenic variants in RNU4-2 and other RNU genes have recently been identified to cause non-syndromic autosomal dominant retinitis pigmentosa [24]. The authors explicitly highlight the presence of incomplete penetrance for these pathogenic RNU4-2 variants reporting a 38% non-penetrant rate [24]. Interestingly, a similar pattern of incomplete penetrance has been observed in a recent preprint by Greene and colleagues involving a newly described syndromic condition with DD/ID caused by pathogenic variants in RNU2-2P. Specifically, 22 out of the 40 affected individuals were found to carry an inherited variant [23]. The mechanisms underlying incomplete penetrance are variegated and complex, with several possible genetic and epigenetic factors influencing gene expression at multiple levels or impacting functional interactions that may modify phenotype presentation [17, 25–28]. Overall, these findings emphasize that incomplete penetrance likely represents a common feature of snRNA-related disorders, particularly for RNU4-2 conditions caused by variants falling outside the 18-bp critical region, suggesting that the variant pathogenicity cannot only be determined a priori by its inheritance pattern.

Cryo-EM structures of the U4/U6.U5 tri-snRNP complex have shown that the U4 5’ Stem-loop plays a pivotal role in the assembly and function of the major spliceosome, serving as a scaffold for protein interactions and RNA-RNA base pairing required for maintaining its inactive conformation [7, 12]. Although yeast and human U4 snRNAs share approximately 50% of sequence homology, a few specific regions with important functional roles are highly conserved. Among these, stretches encompassing ribonucleotides n.28-n.35 and n.40-n.45 within 5’ Stem-loop contribute to the H-bonding network stabilizing U4 snRNA as well as its interaction with SNU13, PRPF6 and PRPF31, core components of the complex. The AlphaFold 3 model prediction considering the isolated U4 5’ Stem-loop and structural inspection of the human U4/U6.U5 tri-snRNP complex highlighted the critical roles of A30, G43, and A44 in stabilizing the 5’ Stem-loop and its interactions with SNU13. These interactions support a perturbing role of the identified variants on the 5’ Stem-loop’s structural integrity and functionality. Consistent with this view, SNU13 Glu⁴¹ and Lys⁴⁴ are invariably conserved from yeast to human, their relative positions are predicted to be intolerant to any amino acid change, and no substitutions affecting these residues have been reported in public databases. We suggest that n.30 A > T and n.43_44insT may disrupt these critical interactions, leading to conformational changes, possibly resulting in misassembly or dysfunction of the U4/U6.U5 tri-snRNP complex.

Recent works highlighted the clinical presentation of ReNU syndrome in patients carrying pathogenic variants within the 18-nt “critical region”, with major features including moderate to severe DD, ID, language impairment, microcephaly, abnormal brain morphology, ophthalmological issues, and a recognizable facial gestalt as major hallmark [3, 5, 15]. The detailed clinical description of the two individuals carrying variants within the 5’ Stem-loop (n.43_44insT, Subject 4; n.30 A > T, Subject 5) fitted well within the ReNU clinical spectrum of presentation, including all its core facial features. Overall, the five affected individuals showed severe DD/ID and language impairment. In particular, expressive language was not achieved in Subjects 1 to 4, while Subject 5 showed delayed speech with only a few words spoken at the age of 6 years. MRI analysis revealed an abnormal brain morphology in all individuals, with Subjects 4 and 5 showing corpus callosum hypoplasia and ventricles dilatation with white matter reduction, a finding consistent with previous reports [3, 5, 9, 15]. In line with previous reports, ophthalmological abnormalities including vision issues, strabismus, and/or nystagmus were reported in all subjects.

The first reports on ReNu syndrome generically reported skeletal problems, including osteopenia and bone fractures, though the exact occurrence of these features was not specified [3]. Recently, osteopenia and multiple recurrent fractures were reported in four additional affected individuals and 18/129 affected subjects carrying different pathogenic variants in RNU4-2 [9, 15]. In line with these findings, we observed the occurrence of pathological fractures in Subjects 1 and 2. Our data further evidence that monitoring of osteopenia in individuals with ReNU syndrome is highly recommended addressing them to possible treatment with vitamin D supplementation and surveillance of increased risk for multiple fractures. Similarly, acral cutis marmorata, which had previously been described in three patients by Rosenblum et al. [15], was present in one individual of the present cohort, suggesting that this sign might represent a relatively common finding in ReNU syndrome.

By systematically scoring all the affected individuals for whom clinical pictures were available, we identified hypotonic appearing face, epicanthal folds, broad sparse eyebrows, anteverted nares, open mouth, horizontally oriented philtrum with tented upper lip, flattened cupid’s bow, everted lower lip, and full cheeks in young individuals as core facial gestalt signs. We propose to use these core facial features to clinically screen individuals with severe to moderate DD/ID to identify those who would be eligible for targeted RNU4-2 sequencing, even before cytogenetic/genomic analyses.

Disease-causing RNU4-2 variants have been estimated to account for up to 0.4% of individuals with NDDs, making ReNU one of the most common Mendelian disorders associated with syndromic DD/ID [3]. A higher disease prevalence of causative variants was observed in the present cohort (2.6%), a discrepancy that is likely to be ascribed to the highly selected nature of enrolled individuals for whom ES/WGS analysis was negative. Our findings emphasize the anticipated high diagnostic yield of a targeted sequencing strategy directed to scan the entire RNU4-2 gene when examining NDD patients who have gone undiagnosed despite extensive genomic testing. The present work contributes to the clinical profile of this newly recognized syndromic NDD, and points to the 5’ Stem-loop as a second RNU4-2 mutational hot-spot implicated in ReNU syndrome.

Author contributions

AB and MT conceived the work. AB, CMa, LC and MT wrote the manuscript. AB, CMa, LC, CC, AC, VC, and MF performed the genomic analyses, and analyzed and validated the genomic data. MC performed the structural analyses. MCB, SGC, ADF, LG, CL, CMe, RO, FP, DP, FSa, FSi, EMV and GZ collected the clinical data. MN, MP, and FCR performed the clinical data analyses. All coauthors provided critical feedback on the manuscript.

Funding

This work was supported, in part, by grants from the Italian Ministry of Health (Current Research Funds, to EMV and MT; RCR-2022-23682289 to FMS, EMV, GZ and MT; RF-2021-12374963, to MT).

Data availability

The sequencing data that support the findings of this work are available on request from the corresponding author. The data are not publicly available due to due to privacy/ethical restrictions. The new variants identified in this work and their clinical association have been submitted to ClinVar (n.30 A > T, VCV003390371.1; n.43_44insT, VCV003390370.1).

Competing interests

The authors declare no competing interests.

Ethical approval

The study was approved by the local Institutional Ethical Committee (ref. 1702_OPBG_2018 and 2072_OPBG_2020). Clinical data, pictures, and DNA samples were collected, used, and stored after signed informed consents from the participating subjects/families were secured. Written informed consents were obtained for publication of individual pictures

Supplementary information

The online version contains supplementary material available at 10.1038/s41431-025-01820-1.