Results
We report a female patient who first presented to a syndromic consultation at the age of eight because auf multiple phenotypic abnormalities. The girl had muscular hypotonia since early childhood. During infancy a developmental delay became noticeable and later she scored in the moderate range of intellectual deficiency. Brain MRI showed two heterotopic foci as well as symmetrically clumped hippocampi. Facial dysmorphism, which became more prominent as a teenager, includes a long face, slightly up slanting palpebral fissures, ptosis of the left eye, a prominent, bulbous nasal tip and low-hanging columella (Figure 1). Furthermore, she had pale skin with many moles, thick curly hair, and a missing left upper canine tooth. Her family described her as extremely friendly, but anxious in contact with other children. A chromosome analysis, a chromosomal microarray (CMA) and diagnostics for Fragile X Syndrome, which have been performed after the first consultation at the age of eight years, were unremarkable.
At re-consultation seven years later, the fifteen-year-old female was enrolled into a study protocol that also involved NGP. The computer-assisted assessment of portrait images yielded high gestalt scores for Koolen-de Vries Syndrome (Figure 1). The feature score, which is based on the clinical abnormalities annotated in HPO terminology, was in the lower range reflecting the rather unspecific phenotypic manifestations in the young female (Robinson et al., 2008; Peng et al., 2021). Although some characteristic aspects of the facial gestalt, such as the elongation of the face and the pear-shaped nose, become more prominent with age, the gestalt score for the portrait at the age of eight years was already comparably high (Figure 1).
With facial dysmorphism typical for Koolen-de Vries Syndrome and some matching phenotypic features such as the friendly personality, this diagnosis was suspected despite the inconclusive CMA results. While ~95% of the cases with Koolen-de Vries syndrome are due to 500 to 650 kb deletion in 17q21.31, only ~5% are due to sequence variants in KANSL1 (Koolen et al., 2006, 2012, 2016; Sharp et al., 2006; Shaw-Smith et al., 2006; Zollino et al., 2012, 2015). Around the microdeletion in 17q21.31 large clusters of low complexity repeats at the breakpoints were described, suggesting an underlying mechanism of non-allelic homologous recombination (NAHR) (Stankiewicz and Lupski, 2002; Dubourg et al., 2011). Up to now, these deletions were found by CMA. So far, only few atypical deletions had been reported for individuals affected by Koolen-de Vries Syndrome, the smallest of these still 68 kb in size (Cooper et al., 2011; Dubourg et al., 2011; Koolen et al., 2012; Zollino et al., 2015). All of these deletions were also detected by CMA.
As the recurrent microdeletion in 17q21.31 was not supported by CMA we initiated Sanger sequencing and multiplex ligation-dependent probe amplification (MLPA) of KANSL1 . Both analyses did not show any abnormal findings. Next, a trio exome analysis in the patient and her parents was performed. Data for the patient and her parents was generated using the NovaSeq platform (Illumina) and the SureSelect v6 exome capture kit (Agilent). Initial bioinformatics analysis was focused on relevant single nucleotide variants (SNVs) and indels using a local implementation of GATK best practices pipelines optimized for data from the NovaSeq sequencer. Copy number variants (CNVs) were initially generated using cn.MOPS (Klambauer et al., 2012). No variants inKANSL1 nor any other gene were detected that could explain the phenotype. Following the inconclusive results of the trio exome analysis, a genome sequencing was conducted. The bioinformatics analysis was performed using the NVIDIA Parabricks toolkit. This toolkit enables accelerated genome analysis by utilizing NVIDIA GPU resources. Several algorithms from this toolkit have been used to call SNVs and indels on the genomic data of the patient. In particular, accelerated versions of BWA-mem and the HaplotypeCaller were crucial for fast processing and yielded variant calls of high quality. To determine candidates for structural variants (SVs) and CNVs, we used manta (Chen et al., 2016), delly (Rausch et al., 2012) and lumpy (Layer et al., 2014). Variant calls of all three tools were merged using a vote-based scheme to find candidates supported by all callers. A 4,708 bp deletion affecting the end of intron 6 and only the first 46 bp of exon 7 (NM_015443.4:c.1849-4611_1895del) was detected by all three tools. Furthermore, the deletion was also clearly visible by a drop of coverage and by split reads in the sequence alignment (Figure 2). In a careful re-analysis of the exome data, that was guided by the results from genome sequencing data, the deletion could also be detected using Pindel (Ye et al., 2009) (Figure 2). Changing some alignments preferences in Integrative Genome Viewer enabled the visualization of the deletion inKANSL1 exome sequencing data (Figure 2). The deletion was also subsequently verified by qPCR.