Introduction

Floating-Harbor syndrome (FLHS [OMIM #136140]) is an ultra-rare, autosomal dominant, neurodevelopmental disorder caused by truncating variants in exons 33–34 of the SRCAP gene [1,2,3,4,5]. It is characterized by short stature, delayed bone age, characteristic facial appearance, and expressive and receptive language delays [5, 6]. Recently, a novel SRCAP-related neurodevelopmental disorder (NDD) was described, which is caused by SRCAP truncating variants outside of the FLHS locus (exons 33–34) [7]. Despite some phenotypic similarity, these individuals with a NDD lacked key features of FLHS and were more likely to have autism, psychiatric and behavioral problems, and hypotonia.

SRCAP encodes a SNF2-related CREBBP activator protein, which increases gene transcription by incorporating histone variant H2A.Z into nucleosomes [8, 9]. SRCAP contains a N-terminal Helicase/SANT-associated domain, a SNF2-like ATPase, a CPB binding domain, and three AT-hook domains at C-terminal [1]. Interestingly, all reported pathogenic mutations are truncating mutations, and most are located within the last two exons, upstream of the AT-hook motifs [1,2,3,4,5]. Beyond the recent report on DNA methylation signature [7], the consequences of variants located outside of the FLHS locus are still poorly understood.

Advances in sequencing technologies have accelerated clinical discovery of causal variants in various Mendelian disorders. However, a large fraction of such cases remain undiagnosed, in part because conventional analytic methods fail to capture the full spectrum of genomic variations such as transposable elements (TEs). TEs are large DNA elements (several hundreds to kilobases in length) that mobilize in the human genome. TE polymorphisms account for ~30% of structural variation discovered from short-read whole-genome sequencing (WGS) in the human population [10]. Currently, three TE families (LINE-1, Alu, and SVA) can retrotranspose in a copy-and-paste manner via RNA intermediates, with estimated rates in 1/20~200 births [11, 12]. While TE insertions could be disease-causing by altering RNA expression and splicing [13,14,15], they are mostly not evaluated in clinical sequencing analysis, leading to missed diagnoses in some patients.

Here, we identified a de novo exonic SVA insertion in the SRCAP exon 13 in a proband with a NDD. RNA sequencing and qRT-PCR demonstrated significantly decreased abundance of SRCAP in the proband, confirming the pathogenicity of this insertion.

Results

Clinical history and presentation

Following an uncomplicated pregnancy, the proband was born via spontaneous vaginal delivery at 40 weeks gestation with normal Apgar scores. She weighed 2.86 kg and measured 49.5 cm long with normal head circumference. In the first year of life, she was diagnosed with failure to thrive, hypotonia, and developmental delay. From age five to seven, she experienced several afebrile seizures, which were accompanied by a normal brain MRI and EEGs suggestive of a focal seizure disorder. Her seizures recurred at 16 years of age, became intractable, and included myoclonic and generalized tonic-clonic type. The proband was also diagnosed with autism, anxiety, tic, and mood disorders. Additional investigation revealed uterine agenesis and single fused pelvic kidney. The proband had myopia and unique facial features including bitemporal narrowing, large mouth, thick lips, prognathia, high-arched palate with crowded teeth and hypoplastic enamel. Most fingers had fetal fat pads and feet were flat with bilateral 2–3 toe cutaneous syndactyly.

Early metabolic studies were suspicious for short-chain acyl-coA dehydrogenase deficiency due to elevated ethylmalonate; however, no pathogenic variants were found by ACADS gene sequencing beyond a homozygous missense polymorphism (c.625 G > A; p.G209S) classified as benign/likely benign (ClinVar ID: 3831) according to ACMG-AMP guidelines [16, 17]. Additional genetic testing included chromosomal microarray, fragile X, FISH testing for 7q11.23, mitochondrial genome, and 70-gene epilepsy panel; all were considered non-diagnostic. The proband and her family were enrolled, but research trio exome sequencing was unrevealing. Given these negative findings, WGS was then completed and analyzed as described below.

A de novo exonic SVA insertion identified by WGS

Evaluation of single nucleotide variants (SNVs) and indels with the genome analysis toolkit (GATK) failed to identify pathogenic mutations in any disease-causing genes, including SRCAP. However, application of GATK-SV to discover structural variants revealed a putative de novo SVA insertion in SRCAP exon 13 in the proband. Further in silico review at the predicted integration site showed multiple features of target-primed reverse transcription (TPRT)-mediated retrotransposition [18]: (1) soft-clipped reads or junction-spanning reads were fused either to unmapped repeat sequences on one end or unmapped poly-T sequences on the other end, which indicate the beginning and ending (poly-A tails) of the insertion, respectively. (2) discordant read-pairs, in which one of the paired-end reads was mapped to the flanking genomic sequence of the breakpoints, whereas the other was aligned to a TE at unexpected distance or orientation. (3) breakpoints offset by target site duplication (TSD) (Fig. 1A).

Fig. 1: Identification and validation of a de novo SVA insertion in SRCAP exon 13.
figure 1

A The Integrative Genomics Viewer (IGV) plot showed a de novo insertion in exon 13 of the SRCAP gene in the proband and its absence in the parents. This non-reference insertion was supported by soft-clipped reads (colored bases) that span the insertion breakpoints (red dashed lines) with clear hallmarks of TPRT-mediated retrotransposition, including target site duplication (TSD) and poly-A tail. B, C Representative gel images of full-length PCR and 5’ junction PCR validations for the de novo exonic insertion. gDNA samples were extracted from blood samples from the trio and two unrelated individuals. D Confirming the presence/absence of the insertion in the proband and four unaffected siblings using 5’ junction PCR. gDNA from siblings were extracted from saliva samples. Black asterisks indicated bands with target size.

Full size image

To experimentally validate this insertion, we performed full-length PCR, where primers annealed to the flanking sequences on both sides of the insertion. This long-range PCR supported a de novo ~3-kb insertion in the proband (Fig. 1B). We also confirmed the presence of the insertion using both 5′ and 3′ junction PCR assays (Fig. 1C and Fig. S1) and confirmed the absence of the insertion in four unaffected siblings (Fig. 1D).

Variant characterization

To characterize the insertional mechanism, we fully resolved the exonic insertion at single-base resolution using both Sanger sequencing on junction amplicons and next-generation sequencing on sonicated fragments of full-length amplicon (Fig. S2A and Supplementary File 1). This revealed a full-length SVA insertion landed in an antisense orientation in exon 13, accompanied by a 15-bp TSD, a polyA tail longer than 70-bp, and an L1-endonuclease cleavage site at 5′–GC/AAGA–3′ (Fig. 2A). These hallmarks confirmed a L1-mediated retrotransposition via TPRT. This variant was represented as NC_000016.10: g.30712369_30712370ins[SVA;30712355_30712369].

Fig. 2: Characterization and pathogenicity of the exonic SVA insertion.
figure 2

A The schematic diagram showed the fully resolved exonic insertion at single base resolution. It was an ~2.8 kb anti-sense SVA insertion with 5′ transduction. It carried a 15-bp TSD, > 70 bp poly-A tail, and L1 endonuclease cleavage site. B Multiple sequence alignment of the 5′ end of the de novo SVA insertion in the proband (the top row) and other SVA subfamilies (from Repbase consensus sequences). The de novo SVA insertion carried a 20-bp 5′ transduction sequence and an alternative hexamer variant ((CCCTCT)2 CCCGTCT)n. C The source locus traced by the 5′ transduction sequence of the insertion was identified in intron 1 of a lncRNA (RefSeq: NR_135014) region. The IGV plot showed the chimeric reads spanning the 5′ junction of SVA-genome where soft-clipped sequences aligned to exon 13 of the SRCAP gene. The additional SVA insertion led to increased depth of coverage at the source locus. D The exonic SVA insertion caused a significant decrease of SRCAP gene expression in the proband. SRCAP TPM (Transcripts Per Million) levels are shown for the blood RNA-seq of the proband compared to GTEx RNA-seq of fibroblasts, lymphoblastoid cell lines (LCLs), and whole blood from healthy donors. E SYBR green qRT-PCR assay demonstrated reduced SRCAP RNA abundance in the proband compared to two unrelated and unaffected controls. Experiments were performed in triplicate.

Full size image

Unlike the canonical 5′ junction of SVA insertion beginning with hexameric repeat sequences (CCCTCT) [19], this insertion carried a 20-bp 5ʹ transduction, which allowed for tracing back to the source locus (Fig. 2B, C). This 20-bp sequence uniquely mapped to an uncharacterized lncRNA locus (RefSeq: NR_135014 — SNRPF divergent transcript, transcript variant 1) that was specifically expressed in testis. A recent study reported a similar insertion in the exon of MSH6 gene, which was derived from the same locus (chr12: 96233959–96236309 [hg19]) [20]. Both SVA insertions bear the same CCCTCT hexamer variant ((CCCTCT)2 CCCGTCT)n (Fig. S2B)), which suggests that this donor SVA is active in the human genome.

Confirmation of pathogenicity

To understand the consequence of the exonic SVA insertion, we first performed in silico analysis on the triplet reading frame. Although it was hard to precisely determine the length of polyA tail of the anti-sense insertion, all possible open reading frames contained early stop codon, truncating the SRCAP gene and rendering it nonfunctional (Supplementary File 2). We then investigated the effect of this insertion on RNA using blood RNA-seq data. The proband showed a significant depletion of SRCAP expression compared to controls in the Genotype-Tissue Expression (GTEx) dataset (Fig. 2D). The SYBR green qRT-PCR assay further confirmed the decreased SRCAP gene expression in the proband compared to unaffected controls (Fig. 2E).

Discussion

In this study, we identified a de novo full-length SVA insertion in exon 13 of the SRCAP gene in a proband with a NDD. RNA-seq and qPCR assay confirmed the pathogenicity of the insertion. As SRCAP variants typically cause FLHS, this proband had a partial phenotypic overlap (speech delay, genitourinary malformation, seizures, and dental issues) but lacked the short stature and delayed bone age historically observed in FLHS patients (Table 1). While most reported FLHS cases are caused by truncating variants in exons 33–34, a recent publication reported 33 cases with variants outside the FLHS locus causing a related NDD [7]. Like our proband, most patients with variants proximal to the FLHS locus had autism and mood/behavioral problems and also lacked the short stature and delayed bone age typical to FLHS (Table 1). Our case supports a broader phenotypic spectrum and further expands the type and location of SRCAP mutations causing NDDs. To our knowledge, this is the first NDD case caused by a transposon insertion in the SRCAP gene.

Table 1 Black and gray symbols denote phenotypic features that have been reported in ≥40% and <40% of patients, respectively.
Full size table

To test whether this antisense insertion bears a novel splicing site and leads to alternative RNA splicing, we reconstructed a patient customized genome and re-aligned RNA-seq reads to the revised reference. Although splicing analysis predicted several novel splicing signals, no splicing events were observed within this ~3-kb exonic insertion (Supplementary File 3). Notably, low-level exon 13 skipping was observed from RT-PCR (Fig. S3A). As a functional validation at protein level, two commercial SRCAP antibodies were tested for western blot, but neither produced a specific band with expected molecular weight (data not shown).

SVA retrotransposition produces 5′ transduced sequences via alternative transcription start sites at the rates of 9.17% (220 out of 2398) for reference insertions [21] and 8.33% (45 out of 540) for non-reference insertions [22]. Utilizing the unique 20-bp 5′ transduction, this SVA insertion was traced to a source locus in an intron of a lncRNA that is almost exclusively expressed in testis (Fig. S3B). This suggests the possible origin of the de novo insertion in the sperm where the SVA locus was highly active for transcription. Since there are no informative heterozygous SNVs or other types of variants near the insertion, we were not able to phase the SVA insertion onto parental haplotypes. Because both older and younger siblings do not carry the same insertion event, we can speculate that the retrotransposition likely occurred in a very small proportion of sperm cells or even a single sperm that eventually gave rise to the phenotype in the proband. Accurate mosaicism could not be measured due to a lack of the paternal sperm sample.

Genomic sequencing has revolutionized our ability to discover causal genetic variants, however, conventional analysis restricted to SNVs and small indels fail to capture the spectrum of mutational mechanisms accessible to clinical sequencing data. Our case highlights the importance of detailed computational and functional characterization of TE insertions, which represent an important and underexplored source of pathogenic variants in clinical genomic studies. To understand the prevalence and contribution of such pathogenic insertions to genetic diseases, systematic TE profiling needs to be performed on genomic sequencing data from large patient cohorts.