Reanalysis of Chromosomal Microarray Data Using a Smaller Copy Number Variant Call Threshold Identifies Four Cases with Heterozygous Multiexon Deletions of ARID1B, EHMT1, and FOXP1 Genes

Abstract

Introduction

Chromosomal microarray (CMA) is a highly accurate and established method for detecting copy number variations (CNVs) in clinical genetic testing. CNVs are important etiological factors for disorders such as intellectual disability, developmental delay, and multiple congenital anomalies. Recently developed analytical methods have facilitated the identification of smaller CNVs. Therefore, reanalyzing CMA data using a smaller CNV calling threshold may yield useful information. However, this method was left to the discretion of each institution.

Methods

We reanalyzed the CMA data of 131 patients using a smaller CNV call threshold: 50 kb 50 probes for gain and 25 kb 25 probes for loss. We interpreted the reanalyzed CNVs based on the most recently available information. In the reanalysis, we filtered the data using the Clinical Genome Resource dosage sensitivity gene list as an index to quickly and efficiently check morbid genes.

Results

The number of copy number loss was approximately 20 times greater, and copy number gain was approximately three times greater compared to those in the previous analysis. We detected new likely pathogenic CNVs in four participants: a 236.5 kb loss within ARID1B, a 50.6 kb loss including EHMT1, a 46.5 kb loss including EHMT1, and an 89.1 kb loss within the FOXP1 gene.

Conclusion

The method employed in this study is simple and effective for CMA data reanalysis using a smaller CNV call threshold. Thus, this method is efficient for both ongoing and repeated analyses. This study may stimulate further discussion of reanalysis methodology in clinical laboratories.

Introduction

Chromosomal microarray (CMA) technology is commonly used to accurately detect chromosomal microdeletions and duplications. This technology enables comprehensive analysis of copy number variations (CNVs) to facilitate genetic evaluations. CNV is defined as an increase or decrease in a segment of DNA. It is one of the essential factors in disorders such as intellectual disability (ID), developmental delay (DD), multiple congenital anomalies (MCA), dysmorphic features, and autism spectrum disorder (ASD). CMA is recommended as a first-tier clinical diagnostic test for postnatal patients with these conditions [Miller et al., 2010; Battaglia et al., 2013]. The introduction of CMA for clinical cytogenomic testing has improved diagnostic yield [Beaudet, 2013; Wayhelova et al., 2019; Lee et al., 2021]. Additionally, improvements in CMA resolution have revealed the clinical importance of smaller CNVs [Hollenbeck et al., 2017].

An updated CNV map of the human genome revealed that approximately 4.8–9.5% of the genome contributes to CNV [Zarrei et al., 2015]. Therefore, many CNVs should be considered as individual differences unrelated to diseases. Thus, detected CNVs should be interpreted in relation to the disease. When interpreting CNVs, one can use genomic variant information collected from affected individuals (for example, the Clinical Genome Resource, ClinGen, https://clinicalgenome.org; the DatabasE of genomic variation and Phenotype in Humans using Ensembl Resources, DECIPHER, https://www.deciphergenomics.org; the Human Gene Mutation Database, HGMD, http://www.hgmd.cf.ac.uk), and databases derived from healthy control individuals (e.g., the Database of Genomic Variants, http://dgv.tcag.ca and the Genome Aggregation Database, gnomAD, https://gnomad.broadinstitute.org).

New information regarding the association between CNVs and disease continues to be revealed and validated. Therefore, the interpretation of pathogenicity may change over time. The American College of Medical Genetics and Genomics highlights the need to reanalyze undiagnosed cases [Deignan et al., 2019]. However, periodic re-evaluation requires considerable effort. In addition, setting a smaller call threshold is practical to utilize information on the smaller pathogenic CNVs that have accumulated. In this case, the number of benign/likely benign variants or variants of unknown significance (VUS) among CNVs would significantly increase. Therefore, a re-evaluation should maximize the clinical impact and minimize the effort required.

Here, we re-evaluated previously analyzed CMA cases by lowering the CNV call threshold to the initial set call threshold. We filtered the data using ClinGen dosage sensitivity genes as an indicator to effectively and efficiently check for morbid genes.

Materials and Methods

Samples

A total of 131 patients with DD, ID, MCA, dysmorphic features, or other conditions, suspected of having pathogenic CNVs, who visited Nagano Children’s Hospital were included. Of the 131 patients, 65 were male and 66 were female. Their ages at the initial analysis ranged from 2 days to 19 years, with a mean of 3 years and 6 months and a median of 1 year and 5 months. CMA testing was initially performed in these cases by analyzing 400 kb and 50 probes with copy number gain and loss thresholds from 2013 to 2019. A total of 22 cases were diagnosed as pathogenic or likely pathogenic during the initial analysis, with 109 unresolved cases. The diagnostic rate is 16.8%. Patient profiles are presented in online Supplemental Table 1 (for all online suppl. material, see www.karger.com/doi/10.1159/000530252).

Chromosomal Microarray

CNVs were analyzed using CytoScan HD Arrays (Thermo Fisher Scientific, Grand Island, NY, USA) consisting of approximately 2,700,000 probes or CytoScan 750 K Arrays (Thermo Fisher Scientific) consisting of approximately 750,000 probes. A total of 113 cases were analyzed using CytoScan HD Arrays, and 18 cases were analyzed using CytoScan 750 K Arrays. Samples that passed the median of the absolute pairwise differences (MAPD ≤0.25, SNPQC ≥15, and waviness SD ≤ 0.12) and quality control values were retained for reanalysis.

The CNVs were reanalyzed and annotated using the Chromosome Analysis Suite (ver.4.1, ChAS: reference model file for CytoScan Array, na33.r3.REF_MODEL, Thermo Fisher Scientific) built on GRCh37/hg19 for unsolved cases in the initial study. In the ChAS system, CNV regions are denoted by the minimum region. In this study, we reanalyzed the CNV reporting filter with size and probe thresholds set at 25 kb and 25 probes for loss and 50 kb and 50 probes for gain from the threshold condition of 400 kb and 50 probes for both loss and gain contiguity.

Filtering as a First Step to Selecting Pathogenic CNVs

We created a filtering list that can efficiently narrow down pathogenic CNVs. Dosage sensitivity curves were curated and provided as dose sensitive in ClinGen, and gene symbols suggesting pathogenicity are listed. We created a filtering list by referring to the Dosage Sensitivity Curations file (https://search.clinicalgenome.org/kb/downloads) (2022.6.30), which is a list of genes curated and published as dose-sensitive genes in ClinGen.

The triplosensitivity list (TS list) and haploinsufficiency list (HI list) were prepared for filtering by extracting genes that were evaluated with scores of 3 (sufficient evidence), 2 (some evidence), and 1 (little evidence). Genes with a score of 30 (associated with autosomal recessive phenotypes) were excluded from the analysis. Genes with a score of 40 (dosage sensitivity unlikely), zero (no evidence available), and “not yet evaluated” were excluded. Sixteen genes were included in the TS and 471 in the HI lists. The results are presented in online Supplemental Table 2.

The gene symbols in each filtering list were compared with those in the CNVs detected through reanalysis under a threshold of 25 kb and 25 probes loss calls and 50 kb and 50 probes gain calls. The task was performed using the MATCH function in Microsoft Excel, and the matched gene symbols were selected. CNVs containing the selected gene symbols were candidate pathogenic CNVs.

Interpretations of Pathogenicity

The CNVs selected by filtering were interpreted for clinical significance according to the ACMG guidelines [Riggs et al., 2020]. To interpret clinical significance, we used the ClinGen, DECIPHER, HGMD, DGV, and gnomAD databases. All patients were evaluated by laboratory and clinical geneticists.

Validation of Candidate Pathogenic CNVs through Quantitative Polymerase Chain Reaction

The copy number of the new candidate pathological CNV regions was verified by quantitative polymerase chain reaction (qPCR). qPCR was performed using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific) on a StepOnePlus Real-Time PCR system (Thermo Fisher Scientific). The primer sequences and qPCR conditions are listed in online Supplemental Table 3.

The relative ratios of the copy number of each genomic DNA were calculated by comparing the autosomal internal control loci (ALB) of the tested individuals and standard control samples using the 2^−ΔΔCT method [Livak and Schmittgen, 2001]. The experiments were performed in triplicates. The relative ratio of each tested region to the control locus (ALB) was <0.65, at which point the CNV was confirmed as a copy number loss [Silva et al., 2019]. For CNVs, two PCR primer sets targeting different exons were used, and consistent results were required for validation.

Results

CNV Reanalysis with a Smaller Call Threshold

A total of 1,033 copy number loss regions were detected in 131 samples using reanalysis with 25 kb and 25 probes as the threshold. Similarly, copy number gain was detected in 131 samples from 662 regions using reanalysis with 50 kb and 50 probes as thresholds. In our previous analysis, using a threshold of 400 kb and 50 probes, 52 regions with copy number loss and 192 with copy number gain were detected. The number of copy number losses was approximately 20 times greater compared to the previous analysis, whereas the number of copy number gain was approximately three times greater.

CNVs Selected through Filtering by the HI and TS Lists

Among the unsolved cases in the 400 kb, 50 probe threshold detection system, 19 regions of copy number loss within 12 genes were selected as candidate pathogenic regions for the HI list filtering. Copy number gain regions were not selected for the TS list filtering.

Interpretation of Candidate Pathogenic Regions

Interpretation of the regions selected through filtering as candidate pathogenic CNVs revealed 4 likely pathogenic cases: 1 case of interstitial loss in ARID1B, 2 cases of partial loss of the 3ʹ flank of EHMT1, and 1 case of interstitial loss in FOXP1 (Table 1). The remaining CNVs were interpreted as either benign/likely benign or VUS (data not shown).

Table 1.

Newly confirmed pathogenic CNVs

Patient No.	Sex	Clinical diagnosis	Genome coordinates (GRCh37)	Size, kb	Marker count	Call segment mean weighted log 2 ratio	Gene (ClinGen HI score)	OMIM morbid (OMIM#, inheritance)
A101	Female	MCA/DD	chr6:157,232,924_157,469,440	236.5	256	−0.588	ARID1B (3)	CSS 1 (# 135900, AD)
A076	Female	MCA/DD	chr9:140,694,542_140,745,153	50.6	72	−0.629	EHMT1 (3)	Kleefstra syndrome (# 610253, AD)
							MIR602	NA
A088	Female	MCA/DD	chr9:140,705,085_140,751,594	46.5	72	−0.593	EHMT1 (3)	Kleefstra syndrome (# 610253, AD)
							MIR602	NA
A180	Male	MCA/DD	chr3:71,078,365_71,167,468	89.1	52	−0.602	FOXP1 (3)	Intellectual developmental disorder with language impairment and with or without autistic features (# 613670, AD)

DD, development disability; MCA, multiple congenital anomalies.

Validation of Candidate Pathogenic CNVs by qPCR

Relative value analysis using the 2^−ΔΔCT method validated interstitial loss in ARID1B, 2 cases of partial loss on the 3ʹ flank of EHMT1, and 1 case of interstitial loss in FOXP1. Four cases involving three genes were confirmed to have a copy number loss.

Case Reports of Newly Detected Likely Pathogenic CNVs

A detailed case report presents 4 newly diagnosed likely pathogenic cases.

Case A101

The patient was an 18-month-old girl at the initial analysis. She had DD, short stature (−3.0 SD), hypotonia, distinctive facial features, and hirsutism. The cases were analyzed using a CytoScan HD array. At the time of reanalysis, the patient was 5 years and 3 months old. She had moderate DD, ID, short stature (−2.5 SD), distinctive facial features, and hirsutism. The patient experienced persistent symptoms during the initial analysis. An additional noteworthy finding was the significant delay in speech development.

The patient had a newly detected copy number loss of 236.5 kb at position 6q25.3 (157,232,924_157,469,440), indicating exons 5–8 deletion in the ARID1B gene. A schematic diagram of the deleted region and real-time PCR quantification results are shown in Figure 1. In a previously reported case of Coffin-Siris syndrome (CSS) caused by a partial deletion of ARID1B, deletions of approximately 243.1 kb from exons 4–8 [Halgren et al., 2012], 58.5 kb including exons 6–8 [Wieczorek et al., 2013], and 225.0 kb from exons 5–8 [van der Sluijs et al., 2019] were observed (Fig. 1a). All 3 cases were interpreted as pathogenic CSS. In contrast, DGV Gold Standard Variants has no record of CNVs similar to the areas detected in our patients. The patient showed a phenotype consistent with CSS. Parental DNA samples were not available to confirm whether they were de novo.

Fig. 1.

a A schematic diagram of the ARID1B gene structure (NM_001374820.1, including 20 exons) and the ARID1B protein structure. The ARID1B protein contains conserved domains, a Med15, ARID/BRIGHT DNA-binding domain, and BAF250. The relative positions of the deleted regions in the cases are displayed as solid lines: 6q25.3 (157,232,924_157,469,440) through CMA analysis. Previously published cases with similar deleted regions are also shown. Genomic coordinates are demonstrated according to GRCh37. b Deletion at 6q25.3 (157,232,924_157,469,440) detected through CMA in case A101 was confirmed by qPCR.

Case A076

ptThe girl patient had 47,XX,+mar[13]/46,XX[17] karyotypes detected by G-banding. Marker chromosomes were not detected in the parents. Subsequently, when the patient was 5 years and 3 months old, CMA was performed using a CytoScan HD Array to determine the components of the marker chromosomes. Mosaic copy number gain was not detected by the mosaic search algorithm in the microarray analysis system using both the 400 kb and 50 kb thresholds. The marker chromosome was most likely derived from the vicinity of the centromere, where the array probe was not located. If so, it may not be clinically affected. The patient had moderate to severe DD/ID, hypotonia, distinctive facial features, severe hypermetropia, atrial septal defects, and conductive hearing loss. In addition, the patient had severe language developmental impairments. She was 8 years and 7 months old when she underwent reanalysis. Even reanalysis using a smaller threshold did not detect mosaic CNVs, which was unchanged from the initial analysis results. The phenotype observed during the initial analysis was maintained.

The patient was newly detected with a copy number loss of 50.6 kb at position 9q34.3 (140,694,542_140,745,153). A schematic diagram of the deleted region and the real-time PCR quantification results are shown in Figure 2. This deletion generates a truncated protein that lacks the functionally important component of EHMT1. No registration of a similar region was found in DGV Gold Standard Variants. However, there have been reported cases of EHMT1 deletion in Kleefstra syndrome that lacked a region identical to that detected in this study. Previously, Wang et al. [2016] reported a 39.6 kb copy number loss case (Fig. 2a). The patient’s clinical features were consistent with Kleefstra syndrome. As we did not obtain DNA from the parents, we could not determine whether they were de novo.

Fig. 2.

a A schematic diagram of the EHMT1 gene structure (NM_024757.5, including 27 exons) and the domain structure of the EHMT1 protein. The EHMT1 protein includes three important functional domains, Zink binding domain, Ankyrin repeats, and SET domain (including pre-SET and post-SET domains). The relative positions of the deleted regions in the cases are displayed as solid lines: 9q34.3 (140,694,542–140,745,153) on patient case A076 and 9q34.3 (140,694,542–140,745,153) on patient case A088 by CMA analysis. Both cases are absent in important functional domains. Previously published cases with similar deleted regions are also shown. Genomic coordinates are shown according to GRCh37. b Deletion at 9q34.3 (140,694,542_140,745,153) on patient case A076 and 9q34.3 (140,694,542_140,745,153) on patient case A088 detected through CMA were confirmed by qPCR.

Case A088

At the initial analysis, the girl patient was 14 months old. The analysis platform for this case was a CytoScan HD Array. She had moderate to severe DD/ID, hypotonia, and distinctive facial features. At the time of reanalysis, she was 4 years and 3 months old and continued to have moderate to severe DD/ID. She had severe speech and language development impairments, with no significant evidence of articulated speech. The patient presented with hypermetropia and persistent bilateral otitis media with effusion.

The patient has newly detected a copy number loss of 46.5 kb at position 9q34.3 (140,705,085_140,751,594). A schematic diagram of the deleted region and the real-time PCR quantification results are shown in Figure 2. No registration of a similar region was found in the DGV Gold Standard Variants. However, He et al. [2016] and Goodman et al. [2020] reported a partial deletion of approximately 60.6 kb with the loss of exons 19–27, similar to case A088, which has been interpreted as a pathogenic Kleefstra syndrome (Fig. 2a). Ciaccio et al. [2018] reported a case in which a smaller minor copy number loss, a partial deletion of exon 24–27, was interpreted as pathogenic (Fig. 2a). The patient’s clinical features were consistent with those of Kleefstra syndrome, and the deleted region included functionally essential components of the EHMT1 gene. As we did not obtain DNA from the parents, we could not determine whether they were de novo.

Case A180

The patient was an 18-month-old boy at the initial analysis. The analysis was performed on a CytoScan 750 K Array. He had MCA, DD/ID, language impairment, ASD, arthrogryposis multiplex congenital, and other distinctive facial features. At the time of reanalysis, he was 4 years and 11 months old. His disease symptoms continued, and he presented with attention deficit hyperactivity disorder (ADHD).

The patient was newly detected with a copy number loss of 89 kb at position 3p13 (71,078,365_71,167,468), including exons 7–11. A schematic diagram of the deleted region and the real-time PCR quantification results are shown in Figure 3a, b. In the past, several cases of interstitial deletions that partially overlapped with our patient’s detected deletions have been reported to be pathogenic [Le Fevre et al., 2013; Meerschaut et al., 2017; Trelles et al., 2021] (Fig. 3a). No registration of a similar region was found in DGV Gold Standard Variants. Our patient’s symptoms were consistent with “intellectual developmental disorder with language impairment and with or without autistic features” (MIM #613670; https://www.omim.org/) caused by partial deletion of FOXP1. DNA samples of the parents were not available for analysis. Hence, we could not confirm whether they were de novo.

Fig. 3.

a A schematic diagram of the FOXP1 gene structure (NM_032682.6, including 21 exons) and the domain structure of the FOXP1 protein. The FOXP1 protein contains a DNA-binding site, FH_FOXP2 (Forkhead domain found in Forkhead box protein P2), and FOXP coiled-coil domain. The relative positions of the deleted regions, in this case, are displayed as solid lines: 3p13 (71,078,365–71,167,468) by CMA analysis. Previously published partial deletion cases with similar deleted regions are also shown. Genomic coordinates are shown according to GRCh37. b Deletion at 3p13 (71,078,365_71,167,468) detected through CMA in case A180 was confirmed by qPCR.

Discussion

Published recommendations state that CMA testing performance and analytical sensitivity should achieve a resolution of at least 400 kb for postnatal analysis using microarray platforms [South et al., 2013; Silva et al., 2019]. High-resolution CMA has considerably improved the resolution of cytogenic analysis for detecting small genomic losses and gains relevant to human genetic diseases [Hollenbeck et al., 2017]. The functional resolution of the CMA depends on the probe density, which in turn depends on the platform. Recent studies using high-resolution arrays have set the call threshold to approximately 100 kb as a relatively stringent criterion to ensure reliable CNV calls [Ozyilmaz et al., 2017; Fan et al., 2018; Ünsel Bolat and Bolat, 2021].

We reanalyzed datasets from CMA analyses using CytoScan Arrays conducted at a 400 kb threshold in Nagano Children’s Hospital from 2013 to 2019 with a lower CNV call threshold. The loss-call threshold was set to 25 consecutive probes and 25 kb, whereas the gain-call threshold was set to 50 probes and 50 kb. Based on the results of a study by Uddin et al. [2015], we adopted a call threshold to increase the sensitivity while ensuring specificity. By lowering the call threshold, we increased the number of loss regions by approximately 20-fold and the number of gain regions by approximately 3-fold. More nonpathogenic regions were expected to be included in the newly detected CNV regions. To efficiently extract pathogenic regions from many such CNVs, we created and used a filtering list based on the ClinGen database, using gene symbols as indicators. The ClinGen database is constantly updated with advanced information and acts as a public database in which information is widely accessible. ClinGen has a curated genome-wide dosage sensitivity map that can be used for the clinical interpretation of CNVs. Therefore, the database provided by the Dosage Sensitivity Curation task team in ClinGen is suitable for filtering the CNV data obtained from CMAs.

Of the 19 CNVs detected through filtering in this study, 4 cases with three genes (ARID1B, EHMT1, and FOXP1) were interpreted as likely pathogenic, and 15 cases were diagnosed as benign/likely benign or VUS. Sometimes, even a morbid gene in OMIM or an HI gene in ClinGen could be considered benign. This is because differences occur in the degree of disease relevance of the gene and the structural and functional impact on gene function owing to the CNV position. ARID1B, EHMT1, and FOXP1, which were interpreted as likely pathogenic in this study, had an HI score of 3 (sufficient evidence) in ClinGen. The probability of being loss-of-function intolerant, with a pLI score, is 1.000 (gnomAD v.2.1.1). These genes are highly intolerant to loss-of-function variations and are established virulence genes.

All cases newly identified as likely pathogenic CNVs had partial intragenic deletions. Even in the case of established genes that cause single-gene diseases, partial deletions must be interpreted with caution. The potential for pathogenicity is estimated to be low for intron-only omissions that do not contain functionally significant elements, such as promoters and enhancers, where the deletion is only in the 3ʹUTR or established benign CNV regions. Whether the deletion disrupts the reading frame, the deleted exon is present in a biologically relevant transcript, the truncated region is critical for protein function, or the frequency of loss-of-function variants in pertinent exons of the general population should be considered [Riggs et al., 2020].

There was no precise breakpoint in case A101, which had ARID1B interstitial loss. However, we estimated the breakpoints from the theoretical minimal (chr6:157,232,924_157,469,440) and maximal (chr6:157,226,634_157,470,201) deleted region data obtained by CMA. The breakpoint on the 5ʹ side is between chr6:157,226,634 and chr6:157,232,924, that is, within intron 4. In contrast, the breakpoint on the 3ʹ side would be between chr6:157,469,440 and chr6:157,470,201, that is, between introns 8 and 9, containing exon 9. Exon 5 at the 5ʹ end of the deletion region was an in-frame exon, whereas exons 8 and 9 at the 3ʹ end were both out-frame exons. Recombination of the remaining exons 4, 8, or 9 leads to a frameshift. Although the possibility that the breakpoint was located within exon 9 cannot be ruled out, it is reasonable to assume that loss of function due to frameshifting occurs, given the clinical symptoms. The phenotypic features of those with ARID1B abnormalities are enormously variable, from clearly recognizable CSS to less specific ID [Kosho et al., 2014]. In ARID1B disorders, various types of pathogenic variants (nonsense, frameshift, splice-site changes, CNVs, deletions of different numbers of exons, or entire gene deletions) have been reported, all of which are truncating [Santen and Clayton-Smith, 2014; van der Sluijs et al., 2019]. Interstitial deletions that closely resemble the deletion region detected in our patient have been reported in CSS cases (Fig. 1a) and cases of ID [van der Sluijs et al., 2019]. Van der Sluijs et al. reported 6 cases of ID due to haploinsufficiency caused by partial deletion of some or only one exon between exons 6 and 10 of ARID1B. Interestingly, Lu et al. [2021] pointed out the possibility that the DECIPHER CNVs (266,355), deletion of only a partial intron 5 of ARID1B, contain several candidate enhancers derived from ENCORD (Encyclopedia of DNA Elements) [ENCODE Project Consortium, 2012] chromatin modification data and that loss of these enhancers may affect the transcription efficiency of ARID1B. In addition, several pathogenic single nucleotide variants and small insertions/deletions located in the deletion region containing exon 6–8 or 9 detected in our patient have also been reported [Santen et al., 2013; Wieczorek et al., 2013; Sekiguchi et al., 2019; van der Sluijs et al., 2019; Suzuki et al., 2020]. Based on the reports on these pathogenic variants, it is clear that frameshifts caused by deletions in exons 6–8 or 9 are pathogenic for ARID1B-related disorders. She was reevaluated at the age of 5 years. She consistently had symptoms of ARID1B-CSS at developmental milestones. In particular, severe speech impairment, which becomes more prominent with age, is a signature symptom of ARID1B-CSS. Therefore, this information provides a basis for supposing that her symptoms are due to microdeletion of ARID1B.

EHMT1 is widely expressed and is involved in chromatin modification, leading to the transcriptional silencing of genes. Diverse transcripts are known, but the most conserved and highly expressed transcript, confirmed as NCBI and EMBL-EBI select transcripts [Morales et al., 2022], consists of 27 exons and exhibits a transcript length of 5,095 bp and translation length of 1,298 aa (NM_024757.5). A076 and A088, in which several exons at the 3′ end of EHMT1 were deleted, the deleted regions contained functionally important regions. Various pathogenic variants have been reported with a partial deletion similar to our patient’s (Fig. 2). Furthermore, many cases of Kleefstra syndrome caused by nonsense, frameshift, and splice-site changes within this region have been reported [Willemsen et al., 2012; Bock et al., 2016; Huang et al., 2021; Frisk et al., 2022]. Additionally, Fear et al. [2022] demonstrated that haplozygous cells of the truncated EHMT1 gene, a premature termination codon in the SET domain, confer features of Kleefstra syndrome. These findings indicate that defects in this region are important in the pathogenesis of Kleefstra syndrome. The clinical presentation of our patient was consistent with Kleefstra syndrome due to loss of function caused by the deletion of a functional region of a critical protein.

A180 cases with interstitial deletion of FOXP1, we estimated the breakpoints from the theoretical minimal (chr3:71,078,365_71,167,468) and maximal (chr3:71,077,031_71,167,824) deleted region data obtained by CMA. The 5′ side breakpoint is between chr3:71,167,468 and chr3:71,167,824, that is, within intron 6, whereas the breakpoint on the 3′ end is between chr3:71,077,031 and chr3:71,078,365, that is, in intron 11. As a result, exons 7–11 skip, and in-frame exon 6 and out-frame exon 12 merge, leading to frameshifting. Therefore, a loss of function occurs.

Several patients have been reported to have exon 7–11-containing partial deletions similar to our patient (Fig. 3a). In addition, several patients with similar symptoms have been reported to have nonsense variants causing frameshifts, small deletions, and single nucleotide variants, with effects on splicing [Lozano et al., 2015; Siper et al., 2017; Trelles et al., 2021; Braden et al., 2021]. These cases confirm that this region is necessary for the normal function of FOXP1.

Loss of FOXP1 function results in a variety of clinical phenotypes but consistently reported clinical features, including global DD/ID, speech delay and articulation problems, ASD, and mild dysmorphic features [Lozano et al., 2021]. Language impairment may represent a core feature of FOXP1 dysfunction [Braden et al., 2021]. It has been noted that ADHD symptoms are present in most individuals with a disrupted FOXP1 gene [Trelles et al., 2021; Lozano et al., 2021; Fernàndez-Castillo et al., 2021]. When re-evaluated in this study, our patient continued to present with initial symptoms. In addition, he presented with severe language impairment and ADHD with impulsivity. These results are consistent with the symptoms due to the loss of function of FOXP1. We determined that the copy number loss detected was likely pathogenic.

Uddin et al. [2015] reported 96% reproducibility in detecting copy number loss at a threshold of 25 probes and 25 kb and 77% at a threshold of 8 probes and 1 kb. Additionally, Asadollahi et al. [2014] reported that 39.6% were false positives in a study investigating the clinical significance of all rare, nonpolymorphic, exonic CNVs of sizes 1–500 kb. CNVs that were false positive (3–181 kb, median 19 kb, mean 45.3 kb) were predominantly smaller in size than true CNVs (2–492 kb, median 131 kb, mean 164.7 kb) [Asadollahi et al., 2014]. Therefore, we consider another validation method essential to ensure the accuracy of CNV for small regions obtained by CMA. In this study, we used real-time PCR for validation. It is unrealistic to validate all regions called by the CMA. Still, it is feasible from the perspective of labor and economy to use carried-through regions by the dosage susceptibility gene filter, as in this study. The small number of real-time PCR variations required to ensure accuracy is an acceptable cost and effort to expend as an additional test. As an efficient reanalysis procedure, we propose qPCR validation of CNVs that are considered pathogenic.

Not all dosage-sensitive genes have been incorporated into the ClinGen Dosage Sensitivity Curated Gene List. However, this will eventually be resolved through a database update. Additionally, it would be necessary to establish more efficient validation methods to detect diseases caused by CNVs smaller than the thresholds used in this study.

The increase in analyses conducted through next-generation sequencing has led to an increased understanding of and widespread interest in cases in which very small regions of CNVs or single nucleotide substitutions cause disease. This necessitates the development of new CNV analysis methods in clinical laboratories. Next-generation sequencing analysis remains a problem in establishing analysis methods for CNV analysis in clinical laboratories [Srivastava et al., 2019; Lalonde et al., 2020; Zhao et al., 2020]. Therefore, the CMA has the advantage of being a clinical test for CNV analysis. This study presents a case in which new diagnoses were made by reanalyzing the microarray data with a smaller call threshold.

Using the ClinGen Dosage Sensitivity Curated Gene List to select CNVs to be considered pathogenic among the many benign and VUS CNVs detected at smaller call thresholds leads to reliable and valuable clinical diagnoses with minimal effort and cost. This type of reanalysis should be performed before considering costly and time-consuming methods in an unsolved case. It will be some time before comprehensive genomic analysis by NGS is established as a clinical test and routinely treated as a first-line search method. Therefore, CMA data should be considered worthy of periodic re-evaluation to utilize constantly evolving databases and research findings.

Acknowledgment

We would like to thank Dr. Minatogawa for their advice regarding the interpretation of FOXP1 deletion cases.

Statement of Ethics

Informed consent was obtained from the parents of all patients/study participants and approved by the Ethics Committee of Nagano Children’s Hospital.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This work was supported by a JSPS Grant-in-Aid for the Encouragement of Scientists (Grant No. 19H00394).

Author Contributions

Noriko Kubota designed the study, performed the experiments, analyzed the data, validated the CNVs, interpreted the data, and drafted the manuscript. Jun Kobayashi and Eiko Hidaka performed the experiments, interpreted the data, discussed the results, and critically reviewed the manuscript. Doctors Ryojun Takeda, Kyoko Takano, and Eriko Nishi collected the clinical information, performed the diagnostic evaluation, and critically reviewed the manuscript. Keiko Wakui interpreted the data, discussed the results, conceived and supervised the study, and drafted the manuscript.

Funding Statement

This work was supported by a JSPS Grant-in-Aid for the Encouragement of Scientists (Grant No. 19H00394).

Data Availability Statement

The data supporting the findings of this study are available upon request from the corresponding author.