A Common Molecular Mechanism Underlies Two Phenotypically Distinct 17p13.1 Microdeletion Syndromes

Sample Recruitment

Samples were collected from The Hospital for Sick Children, Stanford University Hospital, The Children's Hospital of Philadelphia, The University Health Network (Toronto), Hospital do Câncer A.C. Camargo, and Emory University. Research was approved by each center's institutional review board.

TP53 Sequencing

All exons and at least 50 bp into exon-intron boundaries of TP53 were sequenced in all available samples (7/8). No base pair sequence mutations were found.

CGH Microarray Design and Hybridization

Six of eight patients' DNA was hybridized to custom arrays. Two samples (from cancer patients) had insufficient DNA for custom array analysis, and their breakpoints were not fine mapped. We instead used multiplex ligation-dependent probe amplification (MLPA) to determine the extent of TP53 exon deletions. Because their exact breakpoints are not known, we excluded them from statistics regarding formation signature (see below). Array CGH was performed with the use of a customized 4×44K microarray platform (Agilent Technologies, Santa Clara, CA, USA), with genomic DNA extracted from peripheral blood via standard methods. A total of 40,577 oligonucleotide probes were placed on the short arm of chromosome 17, in which an 8 Mb target region around TP53 was covered in ultrahigh resolution. Of the 38,061 in the target region, 15,762 (41%) were designed on exons. Exonic probes were overlapping and tiled across all exons, of all alternative splice variants, for every gene. A minimum of one probe per 350 bp was placed in intronic and nongenic regions. An additional set of probes was designed to also extend our coverage to the telomere and centromere of chromosome 17p but at a reduced density (Figure S3, available online). With this array design, we hoped to capture all copy-number changes anywhere within 8 Mb of TP53, from small, single-exon-sized alterations (45–350 bp) to large macroscopic events, and to quickly obtain breakpoint information of copy-number changes, especially those within protein-coding regions. The lab performing array experiments was blinded to all previous results (sequencing, MLPA, quantitative PCR [qPCR], etc.). Patient and male reference DNA were labeled with Cy3-dCTP and Cy5-dCTP (PerkinElmer, Waltham, MA, USA), respectively, with the use of the BioPrime genomic labeling module (Invitrogen, Carlsbad, CA, USA) and were hybridized to the array platform, as recommended by the manufacturer's protocol (Agilent Technologies). The arrays were washed and scanned with the Agilent G2505B microarray scanner. Data analysis was performed with DNA Analytics version 4.0 (Agilent Technologies). Although this approach is capable of ascertaining both forms of copy-number change (namely deletions and duplications), only deletions were found in this cohort.

Gene-Expression Arrays and Analysis

RNA was extracted from blood via standard methods, assessed by Bioanalyzer (Agilent Technologies), and hybridized to Affymetrix Exon 1.0 microarrays (Affymetrix, Santa Clara, CA, USA). High-quality RNA was available for one individual with a large 17p13.1 deletion (encompassing TP53), two individuals with small 17p13.1 deletions (disrupting TP53), and two individuals harboring germline TP53 missense mutations. RNA samples from three noncarrier siblings were used as controls. Gene-expression analysis was performed with the use of the Partek Genomics Suite (Partek, St. Louis, MO, USA).

Breakpoint Simulation

Custom software was developed for the simulation of CNV deletions across all autosomal chromosomes in the human genome. For each size range (from 10 Kb to 2 Mb), 10,000 simulated CNVs were assessed for intersection with Alu elements at both breakpoints.

Quantitative PCR

To obtain better size information on these deletions, we developed a high-throughput quantitative assay by using an automated liquid handling system in a 384-well plate format with 176 qPCR probes (Table S2) to target and detect the copy number of a large region of chromosome 17 (1.3 Mb). qPCR assays were performed on a Roche LightCycler by relative quantification. qPCR plates were set up with the use of a custom script and an automated liquid handling system. Primers were designed with the use of Primer3 and the human genome reference assembly (UCSC Genome Browser, version hg18). Deletion sizes were found to be larger than that reported by array CGH or MLPA and, on average, were improved by 31% with the use of this assay. The developmental delay (DD)-associated deletions were nearly 100 times larger than those involved in early-onset cancer.

Fluorescence In Situ Hybridization

Fluorescence in situ hybridization (FISH) was performed with the use of standard protocols.¹¹

Parent-of-Origin Analysis

SNP genotyping was performed with Affymetrix GeneChip 250 Sty arrays or by direct sequencing. Microsatellite-marker genotyping was performed by The Center for Applied Genomics at The Hospital for Sick Children.

Breakpoint Mapping

After custom-array processing and analysis, breakpoints were mapped by first designing primers flanking the predicted breakpoints and then amplifying junction-specific fragments with the use of long-range PCR (Roche Expand Long Template PCR System). Junction fragments were subjected to sequencing. Putative breakpoints were analyzed by BLAST, BLAT, and manual inspection.

Rare CNVs at TP53 Are Associated with Cancer Predisposition or Developmental Delay

Our six diagnostic labs screened 4524 patients with diverse clinical phenotypes for DNA dosage changes via array CGH or MLPA (Table S1). Eight probands with a microdeletion at TP53 (17p13.1), a tumor-suppressor gene that predisposes one to early-onset cancer when mutated in LFS, were identified.¹⁰ We performed interphase and metaphase FISH, using TP53 and 17ptel probes (Figure 1A). The interstitial deletion could be seen in all cells and was therefore not due to mosaicism.

Individuals with microdeletions at TP53 had cancer (n = 4) or a noncancer phenotype (n = 4) comprising a spectrum of congenital anomalies (Table 1 and Figure S2) that included pervasive DD and mental retardation, speech difficulties, hypotonia, hand and foot abnormalities, and facial dysmorphisms.

Different 17p13.1 Breakpoints Are Related to Two Distinct Phenotypes

Several congenital syndromes are known to also occur in association with cancer predisposition. Such dual phenotypes are frequently caused by gene dosage mutations, either through numerical chromosomal abnormalities or through specific structural changes (e.g., trisomy 21¹²). In contrast, LFS patients do not show increased rates of neurocognitive disability or any phenotype besides cancer. Consistent with this, none of the TP53-deletion patients affected with cancer had DD or congenital anomalies. Similarly, none of the patients with DD exhibited any neoplastic growth that might suggest an underlying susceptibility, nor did they have family histories of cancer consistent with LFS. Therefore, although they share genomic alterations at TP53, their distinct clinical presentations suggest that these patients fall into two nonoverlapping groups.

Next, we determined the genetic basis of this dichotomy. Our initial patient discovery was performed on two platforms, with complementary depth and breadth. Clinical array CGH provides low resolution at TP53 but provides more information on the extent of 17p13.1 CNVs beyond TP53, whereas MLPA provides high resolution across TP53's 11 exons but provides little information for the surrounding regions. To determine whether CNVs defined by MLPA extend beyond TP53, we used qPCR to determine the copy number of the genes immediately flanking TP53 (Figure S1). Both ATP1B2 (telomeric; MIM 182331) and WRAP53 (centromeric; MIM 612661) were disomic in all cancer patients (mean copy numbers = 2.03 and 2.11, respectively). However, all patients with DD were hemizygously deleted for both flanking genes (mean copy numbers = 0.87 [ATP1B2] and 1.13 [WRAP53]), a significant reduction in comparison to the cancer patients (p = 2.90 × 10-4 [ATP1B2] and 2.42 × 10-8 [WRAP53]). We also carried out MLPA experiments on all array-CGH-ascertained samples, and we found that in every DD case all 11 exons of TP53 were contiguously deleted. In contrast, no cancer case harbored a CNV that included all 11 exons. These results demonstrate that our cohort of DD and cancer patients have overlapping but genotypically distinct CNVs at TP53; whereas DD-associated CNVs include all exons of TP53 as well as flanking genes, cancer-associated CNVs are within TP53, causing a change in copy number to some— but not all—of its exons (Figure 1B).

17p13.1 Genomic Deletions Can Be Inherited or Arise De Novo

A review of the eight probands' pedigrees showed that families of DD patients did not have neurocognitive impairment and that the pedigrees of the four cancer patients were consistent with LFS (Figure S2). All DD patients with available parental samples (n = 3) had a de novo deletion, as shown by CGH, MLPA, and FISH analysis (200 nuclei tested, with no evidence of low-level mosaicism in parents). Among the cancer patients with deletions, familial samples were available in two cases. Of these, one family's samples were sufficiently informative to establish inheritance of the deletion. No apparent parent-of-origin bias was observed for deletions in either group of our cohort.

Design of a Custom Ultrahigh-Resolution Tiling Array

Obtaining sequence-level resolution is the most definitive method of validating rearrangements,¹³ because it leads to precise definitions of the CNVs' breakpoints and gene content, provides clues as to the mechanism underlying their formation,^14,15 and reveals their potential architectural complexity.¹⁶

We designed an ultrahigh-resolution array covering 8 Mb of chromosome 17 to get close to sequence-level resolution and to use it as a clinical diagnostic platform for identifying all possible rearrangements in future patients. The array comprises ∼45,000 oligonucleotide probes spanning 4 Mb upstream and 4 Mb downstream of the TP53 locus (7,512,444 to 7,531,588). All exons within this region are tiled, representing the entire coding sequence of 182 genes and all possible alternative transcripts (2,130 exons; Figures S3A and S3B). The precise array design and probe placement are described in Material and Methods and Figure S3.

We tested our custom array on patients whose breakpoints we had already successfully sequenced. These experiments yielded highly precise size and breakpoint information. For example, a patient with DD was found, by the 17p13.1 array, to have a contiguous genomic deletion of 923,492 bp, a difference in size of only 2183 bp (0.2%) from that established by sequencing. The 5′ and 3′ breakpoints of the deletion were 2341 bp and 153 bp away from the true breakpoints, respectively.

Alu Short Interspersed Nuclear Repeats Are Associated with Breakpoints

Using this array, we determined the size and breakpoints of the remaining samples. Using long-range PCR, we amplified junction fragments spanning putative breakpoints. Then, in cases for which high-molecular-weight DNA was available, we sequenced junction fragments and determined the breakpoint and size of the deletions. The average difference between actual CNV sizes and the arrays' predicted sizes was 6.25% (i.e., 93.75% accuracy).

By array, one patient was revealed to harbor another deletion in 17p13.1. The secondary deletion, which is also heterozygous, is 24 Kb in length and is located downstream of the primary deletion's distal breakpoint (Figure S4). In another instance, an identical deletion was found in the proband and the proband's sibling, indicating the inheritance of the same pathogenic CNV (Figure 3, patient 3332). Although asymptomatic, the sibling is now undergoing routine biochemical and radiographic surveillance for cancer.

We looked for repeat elements coinciding with CNV breakpoints. Of the 12 sequenced breakpoints, ten directly intersect with an Alu short interspersed nuclear repeat element (one from the oldest AluJ family, seven from the intermediate AluS family, and two from the young AluY family).

Most 17p13.1 CNVs Arise by Alu-Mediated Nonallelic Homologous Recombination

Analysis of sequenced CNVs revealed the mechanisms by which they arose. Four of six deletions involved nonallelic homologous recombination (NAHR) between Alu elements present at both the proximal and the distal ends (Figure 2, patients 3026, 3148, 3354; Figure 3, patient 3332). The Alu elements flanking these deletions were in the same orientation and shared a moderate degree of homology (81%–84% similarity by BLAST). The remaining two patients did not exhibit extensive homologies spanning their breakpoints. Of these, one breakpoint showed a 4 bp microinsertion (CAAG), an “information scar” that is a hallmark of nonhomologous DNA end joining¹⁷ (NHEJ; Figure 2, patient 2723).

To evaluate the significance of the observed number of Alus at 17p13.1 CNV breakpoints, we performed 10,000 permutation experiments using randomly distributed CNVs of different sizes (10 Kb to 2 Mb). In these simulations, less than 1% of breakpoints coincided with an Alu pair in the same orientation. In contrast, the majority of 17p13.1 CNVs coincide with directly oriented Alus (67%; Figure S5).

A Common Region Implicates Candidate Genes in Developmental Delay

In our study cohort, all CNVs associated with occurrence of childhood cancer were limited to the TP53 locus, deleting between one and ten of 11 exons. Such deletions are predicted to cause protein truncation, thus interfering with the gene's tumor-suppressive activity. Indeed, in a paired tumor specimen we observed an additional copy-number alteration of the same size, thus inactivating the wild-type allele (Figure 3, patient 2760).

In contrast, in the four patients with DD we found, by fine mapping, a common deleted region (Table 2) that includes 24 genes (critical region shown in Figure 2). There are a number of candidate genes for the observed phenotypes. The four patients with DD harbored between 27 and 86 fully deleted genes. Additionally, fine mapping revealed that two DD patients carried partial deletions of genes, disrupting some but not all of their exons (Table 2).

We evaluated the effect of this 17p13.1 CNV on mRNA levels by using expression arrays (see Material and Methods). Among the genes in the minimally deleted region, the expression of TP53 was significantly underexpressed in the patient affected with cancer but not in the patient with DD (p = 6.82 × 10⁻³; fold change = −1.85797; Figures 4A and 4B).

The expression of six other genes (of the 24 candidates in the region) were significantly changed in the patient with DD but not in the cancer-affected patient (Figure 3B; p < 0.01; fold change < −1.5 or > 1.5). Of these DD-specific genes, which are both hemizygously deleted and underexpressed in all patients, the trafficking protein particle complex 1 gene (TRAPPC1 [MIM 610969]) is particularly intriguing. TRAPPC1, involved in vesicular transport from the endoplasmic reticulum to the Golgi apparatus as part of the TRAPP complex, is the most significantly changed at 17p13.1 and, of note, is also the most significantly changed gene genome-wide (Figure S6; p = 2.90 × 10⁻⁵; fold change = −2.34146).

Three additional DD-specific genes are noteworthy: MPDU1 (MIM 604041), mutations of which result in congenital disorder of glycosylation type If involving severe mental and psychomotor retardation;¹⁸ FXR2 (MIM 605339), a homolog of the fragile X mental retardation gene, FMRP (MIM 309550), which itself may play a role in that disease;¹⁹ and EFNB3 (MIM 602297), known to be important in the development of normal locomotor behavior.²⁰ To determine whether the deletion unmasks a recessive mutation, we sequenced these four genes, but we did not find additional mutations.

Having found reduced TP53 expression in the cancer-affected individual, we examined whether other TP53 signaling-pathway members were altered. We first measured the mRNA levels of a proband from an LFS family carrying an established deleterious base pair mutation (Arg273Cys) and conducted pathway analysis. Using the well-annotated Ingenuity Pathway Analysis “Core Pathways,” we noted a subtle but significant difference of genes in the TP53 pathway in individuals with either an established mutation or an internal deletion of TP53, but not in those with complete deletions and DD (p = 4.74 × 10⁻² and 3.01 × 10⁻², respectively). This shows that TP53 is aberrantly expressed in individuals affected with cancer but not in those affected with DD. Furthermore, we find that disregulated genes common to the missense- and internally-deleted-mutation carriers are associated with known molecular mechanisms of cancer (p = 2.59 × 10⁻³). Together, these data highlight gene-expression differences between individuals having large or small CNVs at 17p13.1.