https://orcid.org/0000-0002-0805-8149
Abstract
The majority (99%) of individuals with 22q11.2 deletion syndrome (22q11.2DS) have a deletion that is caused by non-allelic homologous recombination between two of four low copy repeat clusters on chromosome 22q11.2 (LCR22s). However, in a small subset of patients, atypical deletions are observed with at least one deletion breakpoint within unique sequence between the LCR22s. The position of the chromosome breakpoints and the mechanisms driving those atypical deletions remain poorly studied. Our large-scale, whole genome sequencing study of >1500 subjects with 22q11.2DS identified six unrelated individuals with atypical deletions of different types. Using a combination of whole genome sequencing data and fiber-fluorescence in situ hybridization, we mapped the rearranged alleles in these subjects. In four of them, the distal breakpoints mapped within one of the LCR22s and we found that the deletions likely occurred by replication-based mechanisms. Interestingly, in two of them, an inversion probably preceded inter-chromosomal ‘allelic’ homologous recombination between differently oriented LCR22-D alleles. Inversion associated allelic homologous recombination (AHR) may well be a common mechanism driving (atypical) deletions on 22q11.2.
Introduction
With an estimated incidence of 1 in 4000 live births, the 22q11.2 deletion syndrome (22q11.2DS, MIM# 188400/192430) is the most common microdeletion disorder in humans (1,2). The phenotypic spectrum is broad, with incomplete penetrance for each associated feature. The syndrome is primarily characterized by a characteristic facies, palatal anomalies and congenital heart defects. A subset of patients have immunodeficiency, and neonatal hypoparathyroidism with resultant hypocalcaemia (2). Individuals with 22q11.2DS also have developmental delay in infancy and/or learning disabilities. In addition, mild to severe behavioral problems are common, and prevalence of psychiatric disorders including autism spectrum disorder, attention deficit and hyperactivity disorder, anxiety disorders, and schizophrenia is high in comparison to the general population (3–6).
The presence of low copy repeats on chromosome 22 (LCR22s) entails that the 22q11.2 locus is structurally one of the most complex areas of the human genome. Chromosome 22q11.2 harbors eight LCR22s, commonly termed LCR22-A for the most centromeric, followed by LCR22-B, -C, -D, -E, -F, -G and –H. These LCR22 blocks (A–H) are built of segmental duplications, defined as genomic duplications longer than 1 kb sharing at least 90% sequence similarity (7,8). Over 200 segmental duplications can be distinguished in the human genome, which can be decomposed into over 500 non-overlapping subunits (9).
This structural heterogeneity significantly increases the complexity of accurate genome assembly of these regions. Consequently, even the most recent reference genome (Human December 2013; GRCh38/hg38) still includes three unresolved sequence gaps in LCR22-A. Current standard techniques for whole genome sequencing are based on short reads of 100-150bp and therefore, due to the repetitive nature of the >200kb LCR22s, they are unable to span them. Because of this, mapping or assembly pipelines fail to distinguish reads from paralogous loci.
The LCR22s form an ideal substrate for non-allelic homologous recombination (NAHR) events because of their large size and high sequence homology (10). The non-allelic LCR22s can serve as mediators of misalignment during meiosis. Subsequent crossover between the homologous chromosomes results in deletions and reciprocal duplications, whereas intrachromosomal crossovers can only cause deletions in the genome (11). In 90% of the individuals with 22q11.2DS, a 3-million base pair (Mb) deletion occurs between the two largest LCR22s, LCR22-A and LCR22-D (12, 13). This LCR22-A/D deletion is present as a de novo event in 90% of the diagnosed individuals (2). The reciprocal 22q11.2 microduplication syndrome (MIM# 608363) was described by Portnoï (14).
In addition to the most common 3 Mb LCR22-A/D deletion, several other rearrangements between LCR22s exist. Approximately 9% of individuals have nested LCR22-A/B (1.5 Mb) or LCR22-A/C (2 Mb) deletions, with similar phenotypes as those with the common LCR22-A/D deletion (2, 13). Deletions involving the telomeric LCR22s (i.e. LCR22-E, -F, -G and -H) are less frequently observed and are associated with a heterogeneous phenotype, including developmental delay, congenital heart defects, and a higher risk of preterm birth (12). Furthermore, a newly recognized recurrent deletion just telomeric to LCR22-A was recently described in 2.3% of 22q11.2DS subjects and it was termed LCR22-A+ (13). All the breakpoints were localized to a 12kb low copy repeat, thus acting as a hotspot for meiotic rearrangements. Recombination with LCR22-B or LCR22-D results in a 1.3 Mb LCR22-A+/B or a 2.8 Mb LCR22-A+/D deletion, respectively (13).
Aside from the 22q11.2 deletions with endpoints within the LCR22s, atypical 22q11.2 deletions have been described with at least one breakpoint in the unique sequence between the LCR22s (15). In the majority of published atypical deletions, one of the breakpoints resides in an LCR22, while the second breakpoint is located in unique sequence between the LCR22s. In only two of the reported cases were both breakpoints nested in unique sequences (16,17). Thus far, those atypical deletions have been detected with standard techniques including short tandem repeat marker analysis, single nucleotide polymorphism (SNP) arrays, and fluorescence in situ hybridization (FISH). Probes covering the unique 22q11.2 region were used in the FISH assays, complementary to the commercial probes (TUPLE1, ARSA), which only detect 22q11.2 deletions involving the HIRA gene. Consequently, the breakpoints were never cloned and sequenced, and the mechanisms underlying those rearrangements remained unclear (15–24).
In this study, we leveraged a large-scale, whole genome sequencing study on >1500 subjects with 22q11.2DS (25) to map the atypical deletion breakpoints. These subjects were part of a large international consortium referred to as the International 22q11.2 Deletion Syndrome Brain Behavior Consortium (IBBC) (25). Six individuals were found to have atypical deletions with a proximal breakpoint between LCR22-A and LCR22-B (Fig. 1). To improve our understanding of the mechanisms causing those atypical deletions and possibly the driving forces of NAHR in the common deletions, we charted the rearrangements at sequence resolution by cloning and sequencing the breakpoints. In addition, fiber-FISH was applied to resolve the complex architecture of the rearranged alleles. The deletions could be divided into two groups, where the first one provides signatures of replication-based mechanisms at the breakpoints. The second group is characterized by an allelic homologous recombination (AHR) preceded by an inversion, which is, to our knowledge, not described yet.
Overview of atypical deletion lengths and genes in the 22q11.2 locus. UCSC Genome Browser screenshot with tracks for BACs and Genes (GENCODE v29) in the 22q11.2 region. To visualize atypical breakpoints in the unique region between LCR22-A and LCR22-B, the locus is covered with labeled BAC probes: CH17-203M7 (red), CH17-6O11 (cyan) and CH17-395B16 (yellow). Distal from LCR22-D, the BAC probe RP11-354K13 (magenta) was added to visualize the inversion rearrangement in LEUV-2 and TOR-2. Deletion and duplication sizes of the patients are visualized using black and red lines, respectively.
Results
Non-recurrent, atypical 22q11.2DS breakpoint regions detected by coverage plotting
Affymetrix 6.0 SNP microarray and whole genome sequencing analyses of the IBBC cohort were used to identify deletion sizes. Six individuals (LEUV-1, UTRE-1, TOR-1, DUKE-1, LEUV-2 and TOR-2) harbored atypical deletions with a proximal breakpoint between LCR22-A and LCR22-B (Fig. 1). Analysis of the coverage plots of these six individuals, based on a BWA-MEM alignment against hg38, confirmed the presence of atypical deletions (Fig. 2). Phenotypic information of the probands is provided in Supplementary Material,Table S2 and in the Supplementary Information.
Coverage plots of 22q11.2DS patients with an atypical deletion (hg38). Overview of coverage plots with log2 ratio depicted in the x-axis and chr22 position showed in the y-axis. Blue regions indicate LCR22s, with LCR22-A for the largest proximal box, followed by smaller boxes LCR22-B and LCR22-C, to end with the distal LCR22-D. Coverage in the blue regions is not representative, since the high inter- and intra-chromosomal duplication events hamper correct mapping of these repeats. (A) Coverage plot of a control individual, not carrying a 22q11.2 deletion. (B) Coverage plot of an individual harboring a common 3Mb LCR22-A/D deletion. (C–H) Coverage plots of the six individuals with an atypical 22q11.2 deletion.
Coverage plots were generated for a control individual without deletion (Fig. 2A), an individual carrying the common 3 Mb 22q11.2 deletion (Fig. 2B) and the six individuals carrying atypical 22q11.2 deletions (Fig. 2C–H). The log2 ratios of the reads within the LCR22 blocks are not representative, due to the presence of paralogous sequences. Different deletion types were compared based on the coverage in the unique sequence surrounding the LCR22s. In a typical 3 Mb 22q11.2 deletion between LCR22-A and LCR22-D, the log2 ratio drops from 0 to -1 over LCR22-A and increases over LCR22-D to the diploid state of 0 (Fig. 2B). In these atypical subjects, centromeric coverage drops are observed in the unique, non-LCR22 region between LCR22-A and LCR22-B, rather than being embedded in LCR22-A (Fig. 2C–H). The log2 ratio of subjects LEUV-1, UTRE-1, TOR-1 and DUKE-1 increases to the chromosome’s average distally from an LCR22, indicative of a deletion with a typical telomeric breakpoint embedded in the LCR22. In individuals LEUV-1, UTRE-1 and TOR-1 log2 ratio is –1 up to LCR22-D (Fig. 2C–E). Similarly, log2 ratio is –1 up to LCR22-B for individual DUKE-1 (Fig. 2F). However, in subjects LEUV-2 and TOR-2 the plots represent an increase of the log2 ratio to 0.58 in part of the unique sequence distal from LCR22-D (Fig. 2G-H), before dropping to the normal diploid state. This suggests the presence of a duplication of this distal part. Hence, coverage plots of the whole genome sequencing data uncovered two subtypes of atypical deletions and defined the global rearrangement regions.
Sequence resolution mapping of breakpoints
Proximal breakpoints were then refined based on whole genome sequencing data (Table 1). The nucleotide position of the breakpoint is corresponding to the coverage drop from diploid to haploid by visual inspection of the data in Integrative Genomics Viewer (IGV) (26). In individuals LEUV-1, UTRE-1, TOR-1and DUKE-1, this proximal breakpoint was precisely mapped to chr22:19,251,190; chr22:19,184,629; chr22:19,627,753; and chr22:19,244,529, respectively (hg38). Hence, all proximal breakpoints differ in coordinates.
Rearrangement-spanning read pair analysis of whole genome sequencing data
| Patient | Coverage drop | Nucleotide position (hg38) | Gene | Fiber-FISH probe/BAC |
|---|---|---|---|---|
| LEUV-1 | Proximal | Chr22:19,251,190 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| UTRE-1 | Proximal | Chr22:19,184,629 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| TOR-1 | Proximal | Chr22:19,627,753 | / | CH17-395B16 |
| Distal | LCR22-D (2) | / | Proximal in D3 | |
| DUKE-1 | Proximal | Chr22:19,244,529 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-B (1) | / | Proximal from B2 | |
| LEUV-2 | Proximal | Chr22:19,427,384 | HIRA | CH17-6O11 |
| Distal | Chr22:21,625,347 | / | RP11-354K13 | |
| TOR-2 | Proximal | Chr22:19,140,370 | ESS2 | CH17-203M7 |
| Distal | Chr22:21,674,345 | PPIL2 | RP11-354K13 |
| Patient | Coverage drop | Nucleotide position (hg38) | Gene | Fiber-FISH probe/BAC |
|---|---|---|---|---|
| LEUV-1 | Proximal | Chr22:19,251,190 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| UTRE-1 | Proximal | Chr22:19,184,629 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| TOR-1 | Proximal | Chr22:19,627,753 | / | CH17-395B16 |
| Distal | LCR22-D (2) | / | Proximal in D3 | |
| DUKE-1 | Proximal | Chr22:19,244,529 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-B (1) | / | Proximal from B2 | |
| LEUV-2 | Proximal | Chr22:19,427,384 | HIRA | CH17-6O11 |
| Distal | Chr22:21,625,347 | / | RP11-354K13 | |
| TOR-2 | Proximal | Chr22:19,140,370 | ESS2 | CH17-203M7 |
| Distal | Chr22:21,674,345 | PPIL2 | RP11-354K13 |
Overview of the exact coverage drop positions observed in Figure 2. Exact nucleotide positions in hg38 are presented, with additional information of annotated genes, BACs and fiber-FISH probes in the locus. For breakpoints in the LCR22s, the specific LCR22 is given, together with the number of mapping locations in that LCR22.
| Patient | Coverage drop | Nucleotide position (hg38) | Gene | Fiber-FISH probe/BAC |
|---|---|---|---|---|
| LEUV-1 | Proximal | Chr22:19,251,190 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| UTRE-1 | Proximal | Chr22:19,184,629 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| TOR-1 | Proximal | Chr22:19,627,753 | / | CH17-395B16 |
| Distal | LCR22-D (2) | / | Proximal in D3 | |
| DUKE-1 | Proximal | Chr22:19,244,529 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-B (1) | / | Proximal from B2 | |
| LEUV-2 | Proximal | Chr22:19,427,384 | HIRA | CH17-6O11 |
| Distal | Chr22:21,625,347 | / | RP11-354K13 | |
| TOR-2 | Proximal | Chr22:19,140,370 | ESS2 | CH17-203M7 |
| Distal | Chr22:21,674,345 | PPIL2 | RP11-354K13 |
| Patient | Coverage drop | Nucleotide position (hg38) | Gene | Fiber-FISH probe/BAC |
|---|---|---|---|---|
| LEUV-1 | Proximal | Chr22:19,251,190 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| UTRE-1 | Proximal | Chr22:19,184,629 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-D (1) | GGT2 | Proximal from D7 | |
| TOR-1 | Proximal | Chr22:19,627,753 | / | CH17-395B16 |
| Distal | LCR22-D (2) | / | Proximal in D3 | |
| DUKE-1 | Proximal | Chr22:19,244,529 | CLTCL1 | CH17-203M7 |
| Distal | LCR22-B (1) | / | Proximal from B2 | |
| LEUV-2 | Proximal | Chr22:19,427,384 | HIRA | CH17-6O11 |
| Distal | Chr22:21,625,347 | / | RP11-354K13 | |
| TOR-2 | Proximal | Chr22:19,140,370 | ESS2 | CH17-203M7 |
| Distal | Chr22:21,674,345 | PPIL2 | RP11-354K13 |
Overview of the exact coverage drop positions observed in Figure 2. Exact nucleotide positions in hg38 are presented, with additional information of annotated genes, BACs and fiber-FISH probes in the locus. For breakpoints in the LCR22s, the specific LCR22 is given, together with the number of mapping locations in that LCR22.
The genomes of the subjects were paired-end sequenced, allowing for an accurate detection of insertions, deletions and inversions based on the general fragment length of the library and orientation of the reads. Read pairs can be mapped with the BLAST-like alignment tool (BLAT) in UCSC (27). Primers were developed to clone the breakpoint (Supplementary Material, Table S3 and Fig. S1). Since the distal breakpoints of subjects LEUV-1, TOR-1, UTRE-1 and DUKE-1, are in LCR22 repeat sequence, we expected several BLAT results for the breakpoint-spanning read pair sequences. For subject TOR-1, these sequences match to four regions in LCR22-A and two regions in LCR22-D in hg38, all within probe D3 of the fiber-FISH pattern. Therefore, the forward primer was designed in the unique sequence proximal from the breakpoint, the reverse primer in this LCR22 sequence. The generated PCR product, specific for this atypical patient, was Sanger sequenced (Supplementary Material, Fig. S1B). The first part mapped to the unique sequence predicted by the whole genome sequencing data with an 18bp deletion compared to the reference genome, the last part of the sequence can be mapped to LCR22 probe D3. This 18 bp deletion is a known polymorphism in the population (rs530634277), inherited from the father, who is the parent-of-origin in whom the rearrangement occurred. Both sides of the breakpoint locus share a homologous region of 132 bp. Therefore, we were not able to exactly pinpoint the nucleotide position of the breakpoint junction. For subject UTRE-1, there was only one BLAT result in LCR22-D, proximal from the yellow D7 fiber-FISH probe. The PCR generated a patient-specific product (Supplementary Material, Fig. S1C). Consecutive Sanger sequencing unraveled the presence of unique 22q11.2 sequence, followed by a fragment mapping on the negative strand in LCR22-D (with an internal deletion of 130bp), ended by sequence mapping to LCR22-D on the positive strand. No DNA nor cell line was available for subject DUKE-1 to validate and clone the breakpoint.
BLAT mapping of the breakpoint-spanning reads of proband LEUV-1 matched one position in LCR22-D. Notably, both sequenced ends of the read mapped in the same orientation on hg38. This observation is indicative for the inversion of a DNA segment, creating a read pair with two sequences in forward orientation with respect to the reference genome. Hence, the reverse breakpoint-cloning primer had to be in the same orientation as the forward primer, allowing cloning of the patient-specific breakpoint with the inversion as a prerequisite. The first part of the Sanger sequenced fragment mapped to the unique 22q11.2 region, the last part to LCR22 sequence. In addition, breakpoint-spanning reads feature a polyA insertion. Illumina and Sanger sequencing encounter problems to sequence this long stretch of adenine nucleotides, since the polymerase is making mistakes in this repetitive nature. Therefore, PacBio singe molecule real-time sequencing was performed to exactly calculate the length of the polyA insertion. Long read sequencing of the breakpoint-spanning amplicon validated the presence of a 22 bp polyA at the rearrangement breakpoint, which is neither present at the proximal, nor the distal breakpoint coordinates in the reference genome (Supplementary Material, Fig. S2A). Sanger sequencing of the parental control products showed the absence of this polyA segment in the parental chromosomes. Hence, the insertion can be considered as a rearrangement-related event.
The coverage plots of subjects LEUV-2 and TOR-2 present an extra duplication distal from LCR22-D. Whole genome sequencing data analysis revealed proximal coverage drops uniquely mapping to chr22:19,427,384 and chr22:19,140,370, respectively. The distal trisomic to disomic coverage drop observed in the coverage plots is uniquely mapped to chr22:21,625,347 and chr22:21,674,345, respectively. In the analysis of the whole genome sequencing data, read pairs were observed where the first fragment mapped to this proximal coverage drop and the second fragment to the distal coverage drop. In addition, both fragments of the read pair mapped in a forward orientation with respect to the reference genome. This link can be explained by the presence of an inversion between both loci. Primers were designed to validate this inversion junction in subjects TOR-2 and LEUV-2 (Supplementary Material, Fig. S1D and E). Both sequences were blunt-end ligated without the presence of insertions or deletions.
Detailed read pair analysis of the whole genome sequencing data allowed us to pinpoint exact deletion breakpoints in the first subgroup, and to identify an inversion rearrangement in the second subgroup. Nevertheless, overall architecture of the region remained unclear.
Fiber-FISH assemblies uncover the structural composition of the rearranged 22q11.2 allele
To overcome the biased mapping of sequencing data and resolve the structure of these atypical rearranged alleles in five subjects (LEUV-1, LEUV-2, TOR-1, TOR-2 and UTRE-1), a de novo assembly was performed by using fiber-FISH (28). In this technique, long DNA molecules (>200 kb) were extracted from cells and stretched onto coverslips. These fibers were subsequently hybridized with labeled probes targeting the LCR22 subunits (Fig. 3A and B). In contrast to current sequencing technologies, fiber-FISH was shown to be capable of spanning the LCR22s (29). To visualize the proximal breakpoints in the region between LCR22-A and LCR22-B in the atypical subjects, labeled bacterial artificial chromosomes (BACs) were added to the standard probe composition to visualize the unique sequence in the 22q11.2 locus (Fig. 1).
Fiber-FISH analysis of the rearranged allele in atypical 22q11.2DS patients. (A) UCSC Genome Browser screenshot with track for fiber-FISH probe composition of LCR22-D. (B) Fiber-FISH pattern of a normal, non-rearranged LCR22-D allele. (C) Fiber-FISH results for the rearranged allele in patient TOR-1. (D) The de novo assembly of the rearranged allele in individual UTRE-1. (E) In individual LEUV-1, the probe set was supplemented with green probe D8 to visualize a supposed inversion. (F) Allele de novo assembly of individual TOR-2 uncovered the juxtaposition of BACs CH17-203M7 (red) and RP11-354K13 (magenta). (G) The same observation was made for individual LEUV-2, with a fusion of BACs CH17-6O11 (cyan) and RP11-354K13 (magenta).
In the de novo assembled patterns, one represents that of a normal LCR22-B, -C and –D allele on the remaining, non-deleted allele. Other patterns are indicative for LCR22-A heterozygosity (data not shown). The proximal breakpoint (chr22:19,627,753) of patient TOR-1 is embedded in the yellow-labeled BAC CH17-395B16, which is directly fused to the probe pattern of LCR22-D (Fig. 3C). A magenta BAC RP11-354K13 was added distally from LCR22-D. The rearranged allele features CH17-395B16 (yellow), interrupted by the probe composition D3 (cyan), B2 (red), A3 (cyan), D7 (yellow), A1 (blue), A2 (green), D2 (red), A3 (cyan), B2 (red), D3 (cyan), D6 (green), D5 (magenta), D6 (green) and magenta BAC RP11-354K13 (Fig. 3C). The breakpoint locus observed within this fiber-FISH pattern is concordant with breakpoint-spanning reads mapping to LCR22-D. For individual UTRE-1, a similar pattern was assembled, for which the red BAC CH17-203M7 was directly fused to a pattern suggestive for LCR22-D (Fig. 3D). The LCR22-D breakpoint observed by fiber-FISH was concordant with the breakpoint suggested by the sequencing results.
To allow the detection of an internal LCR22-D inversion in proband LEUV-1, probe D8 was designed within LCR22-D to locally increase the pattern density. In the fiber-FISH pattern of the rearranged allele, red BAC CH17-203M7 is directly proceeded with probes D8 (green), A3 (cyan), B2 (red), D3 (cyan), D6 (green), D5 (magenta) and D6 (green) (Fig. 3E). This pattern suggests an indirect orientation of SD22-4 (probe order D2, A2, A1 and D8 from centromere to telomere) in LCR22-D (29), prior to the deleterious rearrangement that fused CH17-203M7 to LCR22-D (Supplementary Material, Fig. S2C and D). The presence of this inversion embedded in the rearranged LCR22-D is validated by the orientations of the read pairs spanning the rearrangement in the whole genome sequencing data of proband LEUV-1. This SD22-4 orientation in LCR22-D has been observed in 6% of the population (29). However, no fused BAC signals were present, nor was SD22-4 found to be inverted in the LCR22-D alleles of the parent-of-origin, suggesting that the inversion and deletion event occurred de novo (Supplementary Material, Fig. S2A and B).
To accommodate the inversion breakpoints in individual TOR-2, BAC probes CH17-203M7 (red) and RP11-354K13 (magenta) were added to the LCR22 probe set. The rearranged allele displays a fusion of BACs CH17-203M7 and a fragment of RP11-354K13 immediately continued with probes D6 (green), D5 (magenta), D6 (green) and a full-length magenta signal of RP11-354K13 (Fig. 3F). This probe composition suggests AHR to have occurred between two LCR22-D alleles. However, prior to or concomitant with the deleterious rearrangement, an inversion between chr22:19,140,370 and chr22:21,674,354 occurred on one allele (Fig. 4A and B), resulting in segmental duplications oriented in the same direction. Consequently, the rearranged LCR22-D harbors RP11-354K13 on both sides. A similar pattern was observed for patient LEUV-2 with a fusion of the cyan (CH17-6O11) and magenta (RP11-354K13) BAC (Fig. 3G).
Mechanisms to create the rearranged allele in proband TOR-2. To obtain the observed fiber-FISH pattern of TOR-2 (Fig. 3F), a two-step mechanism is supposed. (A) First, an inversion of the normal allele results in the juxtaposition of the red and magenta BAC probe, distal from LCR22-A. (B) As a second event, the AHR between two LCR22-D repeats of a different allele causes a deletion. Hence, partial presence of the red BAC probe, deletion of the locus, and duplication of the magenta BAC probe is explained. (C) Alternatively, FoSTeS/MMBIR replication-dependent mechanisms generate such complex architectures.
We reasoned that the inversion would have preceded AHR and might be present in one of the parents, driving the 22q11.2 rearrangement. Alternatively, the inversion could have arisen de novo. To investigate this, PCR with primers spanning the inversion breakpoint was performed on DNA derived from peripheral blood lymphocytes from both parents of TOR-2 and LEUV-2. In none, these inversion-specific PCRs produced a positive amplicon (Supplementary Material, Fig. S1D and E). Similarly, fiber-FISH mapping of Epstein Barr virus transformed lymphoblastoid cell lines using both LCR22 subunit probes and BAC probes that were found to display rearranged patterns in their offspring were normal for all parental LCR22 alleles (data not shown). LCR22 patterns of the parents were concordant with previous observations (29). No tissues other than EBV cell lines were tested.
Discussion
Six out of >1500 individuals from the IBBC cohort carried an atypical nested 22q11.2 deletion. These cases are incidental findings, since the presence of a nested and/or atypical deletion was one of the exclusion criteria of the IBBC project. However, samples were included accidentally or due to the limited resolution of the FISH assays. The overall incidence of these atypical deletions in the general 22q11.2DS population was estimated at 2% (30). The phenotypic spectrum of this cohort with atypical 22q11.2 deletions is not significantly different of that observed in patients with the common 3 Mb 22q11.2DS.
We mapped the rearranged chromosome 22 in these six IBBC individuals to characterize those rearrangements and their causal mechanism. Analyses of read pairs spanning the rearrangements identified the unique deletion breakpoint regions in four (LEUV-1, TOR-1, UTRE-1 and DUKE-1) subjects. Unexpectedly, in two individuals (LEUV-2 and TOR-2) an inversion of a fragment including LCR22-B, LCR22-C and LCR22-D was present. Since no sequencing technology is capable of spanning the LCR22s, the overarching structure of those rearranged alleles remained elusive. Using fiber-FISH, the deletion breakpoint regions were mapped at subunit resolution within the LCR22s for three of these individuals. In the two probands with an inversion spanning LCR22-B, LCR22-C and LCR22-D, we hypothesize the final rearrangement is a consequence of two separate events: first an inversion followed by an inter-chromosomal AHR between differently oriented LCR22-D alleles (Fig. 4A and B). Additionally, an internal LCR22-D inversion of SD22-4 (29) was present in proband LEUV-1. Surprisingly, these inversions were not observed in EBV cell lines of any of the parents. Hence, they occurred in the germline precursor or they coincide with the HR.
For AHR or NAHR to occur, at least 300 bp of perfect sequence identity are required (11). Since LCR22s contain several shared subunits they are common substrates for NAHR. Absence of LCR22 sequence at one side of the atypical rearrangement suggests that other mechanisms drive these atypical deletions. Non-homologous end joining (NHEJ) or microhomology-mediated end joining occasionally introduce complex rearrangements (31). Both repair mechanisms are invoked if spontaneous double stranded breaks occur in a cell. Hence, to explain the observed rearrangements, two or more double stranded breaks must have occurred simultaneously prior to erroneous repair. Alternatively, replication-based mechanisms as fork stalling and template switching (FoSTeS) or microhomology-mediated break-induced replication (MMBIR) could have led to these rearrangements (32). Resulting breakpoint junctions of these events are characterized by signatures as insertions, deletions, inversions and microhomology traces (33). Several consecutive fork stalls do occur, and generate complex rearrangement patterns (32). Microhomology of up to 132 bp was detected in TOR-1 and 17 nucleotides in LEUV-1 between the two breakpoint regions. Additionally, 30 adenine base pairs were inserted at the breakpoint of LEUV-1, consistent with polymerase slippage events (34). PolyA insertions were previously linked to LINE1 endonuclease-dependent de novo insertions (35). However, there is no evidence for a de novo insertion of LINE1 sequence at these breakpoints. The complex architecture of the rearranged fragment of UTRE-1 (Supplementary Material, Fig. S1C) can be explained by the FoSTeS mechanism, where template switches occurred during DNA replication (33). Although the LCR22s are not directly implicated in the (proximal) breakpoints of these non-recurrent atypical deletions, Carvalho et al. (36) suggests a mediating role for LCRs in general as a destabilizing factor making the locus sensitive to rearrangements.
In two out of six individuals, rearranged allele patterns suggest an inversion preceded the deletion (Fig. 4A). Subsequent to these inversions, AHR between the inverted and a normal LCR22-D produced the observed rearranged allele (Fig. 4B). Although an inversion was present, parts of the LCR22-D locus still have the same orientation and can be considered as substrates for AHR. Fiber-FISH assemblies suggest that the recombination has taken place distally in LCR22-D (Fig. 3F and G). Since these breakpoints are embedded within LCR22-D, the breakpoint region could not be delineated at sequence resolution with short-read data only. Read pairs in TOR-2 and LEUV-2 only explain the observed inversion, since the mapping of the AHR reads within the LCR22s is biased. In the other atypical nested deletions, reads do map to the deleterious event, which occurred between the unique sequence and an LCR22. In LEUV-1, however, the breakpoint-spanning mates feature the same orientation compared to the reference genome hg38. This is indicative for the inversion of the SD22-4 duplicon in LCR22-D (29).
Since the inversions are hypothesized to have occurred prior to the deleterious rearrangements, these could be present in the parent-of-origin as well. Inversion polymorphisms between LCRs are frequently observed in the parent-of-origin of patients with genomic disorders, predisposing these alleles to NAHR (10, 37–39). However, none of the inversion breakpoint PCRs, nor fiber-FISH did detect the inversions in the parents. This does not rule out the possibility that germline mosaicism for the inversions could be present and subsequently, NAHR between a normal and an inverted chromosome 22q11.2 produced these alleles. Alternatively, FoSTeS/MMBIR could have generated these complex rearrangements during replication, which could explain why the parents are not carriers of the inversions observed in the probands (Fig. 4C). In this model, it would be coincidental that the template switching occurred at the homologous region within LCR22-D.
In summary, fiber-FISH allowed us to validate six atypical deletions detected by whole genome sequencing, and map the rearrangements within the LCR22s. In two cases, the rearrangements are not merely deletions but are complex rearrangements characterized by the presence of a deletion, duplication and an inversion. Scrutinizing these breakpoint regions paves the way to enhance our understanding of the LCR22 architecture and to a better correlation of the phenotype with the genotype.
Materials and Methods
Patient resource
IBBC, a 22 clinical and 5 genomic site worldwide collaboration focusing on the association of 22q11.2DS and schizophrenia, performed multiplex ligation-dependent probe amplification (MLPA), Affymetrix 6.0 SNP microarrays, and whole genome sequencing on >1500 patients with a 22q11.2 deletion (25). Patients were diagnosed with 22q11.2DS using the FISH assay with TUPLE1/ARSA probes (Abbot Molecular, Abbot Park, Illinois, USA), the MLPA SALSA P250 DiGeorge diagnostic probe kit (MRC-Holland) or the CytoSure Constitutional v3 (4x180k) (OGT, Oxfordshire, UK).
Six patients (LEUV-1, LEUV-2, TOR-1, TOR-2, UTRE-1 and DUKE-1) were identified with an atypical deletion in the IBBC cohort. For this study, two proband-parent trios were recruited from Leuven (probands LEUV-1 and LEUV-2), one duo from Utrecht (proband UTRE-1 with her father) and two trios from Toronto (probands TOR-1 and TOR-2). All experiments were performed on EBV cell lines. Cell line transformation was carried out in Leuven for the families from Leuven and Utrecht, and in Toronto for the Toronto families. Genomic DNA was extracted from the cell lines with the DNeasy Blood and Tissue kit (Qiagen). For patient DUKE-1 (recruited from Duke), no cell line nor DNA was available for experiments. An informed consent was signed by all participants of the study, regarding the use of their EBV cell lines and DNA for sequencing and genotyping purposes. The study was approved by the Medical Ethics Committee of the University Hospital/KU Leuven (S52418) and of the University Medical Centre of Utrecht (08/354). The Institutional Review Board approved the research protocol for the study of the Clinical Genetics Research Program at the Centre for Addiction and Mental Health (REB# 114/2001-02).
Refined whole genome sequencing analysis
The patients included in the IBBC cohort were whole genome sequenced at the HudsonAlpha Genome Sequencing Center (Huntsville, AL) on an Illumina HiSeq2500 platform for the first 100 samples (including sample TOR-1) and on Illumina HiSeq X Ten for the remaining samples (including samples LEUV-1, LEUV-2, TOR-2, UTRE-1, DUKE-1). HiSeq 2500 runs produced 100 bp paired-end reads and HiSeq X Ten runs produced 151 bp paired-end reads. Reads were aligned to genome build hg38 with BWA-MEM (40) to enable manual inspection of breakpoints using read pair analysis, visualized in the IGV (26). The average coverage depth of diploid loci on chromosome 22 is 62×, 47×, 35×, 36×, 46× and 37× for patient LEUV-1, UTRE-1, TOR-1, DUKE-1, LEUV-2 and TOR-2, respectively.
CNVs of chromosome 22 were detected using the Control-FREEC tool (41). Default settings were applied, except for windows size which was set to 10 kb (Fig. 2). Obtained copy number ratio values and called segments of CNVs were then used to generate the plots with the R package ggplot2 (42). Coverage plots should therefore show a 50% reduction of the unique sequence coverage depth for deletions, present as a drop from 0 to –1 on the log2 ratio scale. Reciprocally, duplicated sequence is observed as a 50% coverage depth increase, concordant with a log2 ratio increase to 0.58. Within the LCR22s, significant sequence paralogy leads to collapsing read mapping on the (incomplete) reference genome. In addition, interindividual read depth variability hampers the identification of a narrow breakpoint region within the repeats.
Fiber-FISH
DNA fibers were stretched onto coverslips using the Genomic Vision extraction kit and combing system (Genomic Vision, Paris, France). Coverslip probe hybridization and de novo allele assembly was performed as previously described (29). The standard LCR22 probe pattern consists of fourteen fluorescent probes, designed using the characterized subunit sequences library. This probe set was supplemented with BAC probes to visualize unique sequence amid the LCR22s (Fig. 1). BAC DNA was extracted from BAC clones (BacPac Resources, CHORI, Oakland) using the Nucleobond Xtra BAC kit (Macherey-Nagel). Subsequent labeling of the BAC probes was performed with the Bioprime DNA labeling system (Invitrogen). Labeled BAC probes are CH17-203M7 (red), CH17-6O11 (cyan), CH17-395B16 (yellow) and RP11-354K13 (magenta). An additional probe D8 was developed to confirm the presence of an internal LCR22-D inversion of SD22-4 (29) in LEUV-1. The forward primer (5′-GTCTTGTCAAGGTGGAATGA-3′) and reverse primer (5′-TCTGTCTCTGTGCCTCAGTT-3′) produced an amplicon of 7516 bp, using the TAKARA LA v2 kit (Takara Bio Inc.). The probe was subsequently labeled with fluorescein-dUTP, creating a pseudocolored green signal on the slides (Fig. 3E).
PCR validation of patient-specific (inversion) breakpoints
To validate the positions of rearrangement breakpoints, primer pairs were generated for PCR amplification based on sequencing reads spanning the breakpoint (Supplementary Material, Table S3). To determine recurrence of breakpoints, the reaction was performed in the patient, one or two parents, and additional patients with a different breakpoint location (Supplementary Material, Fig. S1). An additional primer pair was developed to generate a control product on the non-rearranged allele of the patient or on both alleles of individuals without the specific deletion. A reaction mixture of 50 μL was prepared according to the Taq DNA polymerase protocol (Invitrogen). The amplification reaction started with an initial denaturation at 94°C for 3 min, followed by 25/30 cycles of 45 s at 94°C (denaturation), 30 s at 57°C (annealing) and 70 s at 72°C (extension). A final elongation step of 10 min at 72°C was included. For the LEUV-1 PCR, the extension time was 90 s. Product presence or absence was then examined on a 2% agarose gel.
Long-read sequencing of breakpoint amplicon of LEUV-1
The breakpoint amplicon generated by primer pair ‘Breakpoint LEUV-1’ (Supplementary Material, Table S3) was prepared for long-read sequencing according to the Template Preparation and Sequencing protocol (Template Prep kit 3.0, Pacific Biosciences, Menlo Park, CA). This library was spiked in on a single SMRT cell on a PacBio RSII using a DNA/polymerase binding kit P6 v2 (Pacific Biosciences, Menlo Park, CA) and DNA Sequencing Reagent kit 4.0 v2 (Pacific Biosciences, Menlo Park, CA). The analysis was performed using the RS_Long_Amplicon_Analysis.1 pipeline (software: Smrtanalysis_2.3.0) with the following settings: minimum sub-read length 950bp, default barcode score 22, and default amount of sub-reads 2000. On this SMRT cell 27909 reads were assigned to the barcode of this amplicon.
Funding
Fondation Jérôme Lejeune (1665); Fonds Wetenschappelijk Onderzoek (GOE1117N); KU Leuven (C14/18/092); National Institute of Mental Health (5U01MH101723-02).
Conflict of Interest Statement: None declared
References
Author notes
Members of the International 22q11.2 Brain and Behavior Consortium (IBBC) are provided in Supplementary Material, Table S1.
The authors wish it to be known that, in their opinion, the first two authors should be regarded as joint First Authors.
