High-Resolution Mapping of Genotype-Phenotype Relationships in Cri du Chat Syndrome Using Array Comparative Genomic Hybridization

Patients with Cri du Chat Syndrome

Subjects were selected from a group of >150 patients with 5p deletions who have been extensively studied by one of us (E.N.) over the past several decades. Subjects were included if detailed clinical data and genomic DNA were available. To facilitate additional studies, we analyzed only those subjects for whom previous studies had provided knowledge of the parental origin of the aberrant chromosome and somatic cell hybrids with the aberrant chromosome 5 and for whom familial genomic DNA was available. For most subjects, results of conventional cytogenetic analysis were available. Table 1 lists the 94 qualifying subjects and includes the phenotypic assessment, conventional cytogenetic assessment, and the array results. (Table A1 in appendix A [online only] contains an electronic version of table 1; it can be opened in Microsoft Excel for data manipulation and plotting.) Patient numbers 1–69 refer to consecutive (time at diagnosis) Danish families. Subjects with numbers ⩾100 are from other countries—patients 100–117 from Norway, patients 201–256 from England, patient 300 from Czechoslovakia, and patient 402 from Australia—and were referred for a more-detailed clinical and cytogenetic evaluation. Thus, these cases represent a quasirandom sampling of individuals with 5p deletions. The array CGH measurements on these specimens were performed after appropriate approval from the institutional review board of the University of California San Francisco was obtained.

Phenotypic features were evaluated by a single observer (E.N.) so that classification criteria were uniformly applied. Since aspects of the phenotype may change during development, most patients were evaluated when aged <5 years and again when aged >5 years. The presence or absence of a phenotypic characteristic is labeled by a “Y” or “N,” respectively, in table 1. Lack of data is indicated by “ND.” Because phenotypic characteristics may be age dependent, the presence of a “Y” for either age group classified a subject as positive for that phenotypic characteristic. For example, the cry and facial dysmorphology are most distinct when the patients are young, so that the loss of those phenotypes in the older group was not deemed significant. Several patients did not show a particular phenotype at one age range (absence indicated by “N”), and data were not available at the other, which resulted in uncertainty in their phenotypic status. Such cases were excluded from the determinations of the relationship of that characteristic to genotype. Thus, patients 18, 45, 56, 112, and 117 were not used to define the cry region, and patients 56 and 112 were not used to determine the genomic region associated with facial dysmorphology. Patients with the most-severe mental defects were not used in establishing the speech-delay region, since their ability to speak may have been impaired for other reasons. Thus, patient 45 was not used to localize the speech component of the phenotype.

Psychomotor-development assessments defining the severity of MR were based on personal observations (by E.N.), written information, psychological tests, school performance information, etc. MR is difficult to assess in young children, so only patients with data at age >5 years were used for correlations of MR with the deletion. Thus, patients 22, 225, and 231 were excluded from this analysis. The degree of MR was indicated by a numerical scale that ranged from 0 (unaffected) to 7 (profoundly affected). The general criteria for each classification are listed in table 2. The uncertainty of these assessments is estimated to range from ±0.5 at the lower end of the scale to ±1.0 at the upper end.

Arrays.—Arrays for high-resolution analysis of 5p contained elements produced from BAC, P1, and PAC clones that were selected using genetically mapped STS markers (Hudson et al. 1995; Peterson et al. ¹⁹⁹⁹), as well as variable numbers of clones located on other chromosomes that were used for normalization and for detection of other aberrations. Locations of all clones are based on the July 2003 University of California–Santa Cruz (UCSC) freeze of the human genome. The density of coverage on 5p was highest in regions previously defined as important in cri du chat syndrome. The 55 clones that encompassed all of the 5p aberrations found in these patients are listed in table 3. (Table A2 in appendix A [online only] contains an electronic version of this table; it can be opened in Microsoft Excel for data manipulation and plotting.) With completion of the genome sequence, we were able to accurately map most clones. However, the sequence positions of several had to be interpolated on the basis of genetic-map information. All patients were analyzed on arrays that contained the 55 clones on 5p, plus an additional 100–300 clones at other locations. Thirty-seven of the patients were also analyzed on various versions of arrays that contained ∼1,750–2,000 clones distributed over the entire genome (Snijders et al. 2001), to determine whether there were aberrations involving other chromosomes. Some of these larger arrays did not include the full set of 5p clones.

DNA for the array elements was isolated from the clones and was amplified by linker-adapter PCR. The PCR products were suspended in 20% dimethyl sulfoxide in water at a concentration of ∼0.8 μg/μl (Snijders et al. 2001) and were spotted onto chromium-coated slides by use of a custom-built printer employing capillary-tube printing pins (authors' unpublished material). Neighboring triplicate spots were printed for each clone. For the smaller arrays, two sets of triplicates at widely separated locations on the array were printed.

Array CGH hybridization and analysis.—All genomic DNA was isolated from peripheral blood of patients and control subjects by use of standard techniques. DNA was labeled using either nick translation (1 μg) (Pinkel et al. 1998) or random primer extension (0.5 μg) (Snijders et al. 2001). Fluorochromes directly coupled to dCTP were used for labeling. Some measurements employed fluorescein and Texas-red (or Alexa 594), whereas others used Cy3 and Cy5 to label the test and reference genomes, respectively.

Labeled test and reference genomic DNAs along with 50 μg of Cot-1 DNA, included to suppress the hybridization of the repetitive sequences, were ethanol precipitated and resuspended in hybridization mix to a final composition of 50% formamide/10% dextran sulfate/2× SSC/1%–4 % (v/v) SDS to a total volume of ∼50 μl. The hybridization mix was heated to 70°C to denature the DNA, and the temperature was lowered to 37°C for ∼1 h to allow reassociation of the repetitive sequences. For hybridization, a low barrier made from rubber cement was formed around the array. The hybridization mix was then applied to the array, and the array was placed in a sealed chamber that left the upper surface of the fluid free. Hybridization proceeded for 16–48 h on a slowly rocking table at 37°C (Pinkel et al. 1998). After hybridization, arrays were rapidly washed with a stream of PN buffer (0.1 M sodium phosphate; 0.1% nonidet P40; pH 8) to remove most of the hybridization solution; they were then immersed in 50% formamide/2× SSC at 45°C for 15 min, followed by a final wash in room-temperature PN buffer for 15 min. After arrays were washed, a solution of 90% glycerol and 10% phosphate buffer, pH 9, containing 1 μM of the DNA stain DAPI (4′,6-diamidino-2-phenylindole) was applied to each array, and a cover slip was added. The DAPI stained the array spots, rendering them visible independent of the hybridization signals.

Arrays were imaged in a custom-built charge-coupled device (CCD) imaging system (Pinkel et al. 1998; authors' unpublished data). We acquired DAPI plus Cy3 and Cy5 or fluorescein and Texas-red images, depending on which dye pair was used to label the test and reference genomes. The entire array was contained within a single CCD image. The image-analysis software UCSF SPOT (Jain et al. 2002) was used for most analyses. This program determined the location of the array spots by use of the DAPI image and calculated the test and reference intensities of each pixel in each spot, after subtraction of local background. We use the ratio of the total (background subtracted) test intensity to total (background subtracted) reference intensity as the measure of relative copy number for each spot, and we averaged the ratios of the replicates. No computational adjustments of any type (e.g., shading corrections, lowess normalization, spatially dependent normalizations, and clone-specific normalizations) were made to either the images or the ratio data. Quality criteria, including the correlation of test and reference signals within a spot, were applied to recognize problematic signals. Generally, spots with a correlation r<0.8 were rejected. Clones with an SD >0.2 for the replicates were removed from the analysis. In most cases, the SD of replicate spots was <0.02. Measurements of clones with only one of the replicates surviving quality checks were also rejected. An overall normalization factor was applied, so that the median of the log₂ratios, or median linear ratio, of array elements at two copies per cell was set equal to 0 or 1, respectively.

Array CGH Measurements

Representative copy-number profiles for chromosome 5 and a whole-genome scan are shown in figure 1. Single-copy deletions and gains in a homogeneous sample would ideally have log₂ratios of −1.0 and +0.58, respectively. In practice, the ratios for the deletions had a range of −1 to −0.7, with deviations from ideal presumably due to incomplete suppression of signals from highly repeated sequences, cross hybridization, imperfect definition of background levels, etc. With this nearly quantitative relationship of ratio to copy number and with the noise levels (indicated in fig. 1), deletions could be detected with very high reliability by use of a simple threshold set at a log₂ratio equal to −0.4. In most cases, sufficient precision to establish one or both deletion boundaries of the interval between two array clones was achieved in a single hybridization. If the initial measurement contained too much noise to identify clones flanking a deletion boundary, the measurement was repeated. Measurements with dye reversal were not performed, since they provided no more information than a rehybridization. Occasionally, it was found that repurification of the specimen DNA improved data quality, indicating that unknown contaminants in some specimens affected the ability to consistently label or hybridize the genomic DNA.

Table 1 summarizes the deletion data, conventional cytogenetic analysis, and phenotypic characteristics for the 94 patients. The patients are listed in numerical order, corresponding to previous numbering (Niebuhr ^1978a; authors' unpublished material). The deletions in 82 subjects were terminal, whereas, in 12, the deletions were interstitial, for a total of 106 deletion boundaries on 5p. As discussed more fully below, 15 of the subjects were found to have copy-number aberrations in addition to 5p deletions. Three of those aberrations were duplications of 5p sequences adjacent to the deletion boundary, whereas the other aberrations involved genomic regions other than 5p, as shown in table 4.

Two indications of interesting DNA-sequence features were found on the basis of our measurements. First, a significant breakpoint cluster was detected. All of the 106 deletion boundaries found in these patients occurred within the distal ∼33 Mb of the chromosome, resulting in an average density of about three boundaries per megabase. Of the 94 cases, 9 had distal or proximal boundaries in the ∼0.5 Mb region of clones 26, 27, and 28 in 5p15.2, yielding a density of ∼18 boundaries per megabase. Moreover, if the clones in this region are ordered as indicated in the July 2003 sequence freeze, then the deletion boundaries appear complex, having oscillations in the ratio of neighboring clones at the boundaries. Reordering the clones simplifies the apparent structure of the deletion boundary in all cases. In table 3, we list the clones in our revised order, but we indicate sequence positions based on the genome-sequence assembly. These results suggest that there may be some sequence motifs in this region that facilitate aberration formation and that complicate proper assembly of the genome sequence. Alternatively, if the sequence assembly is correct, then the cluster of deletion boundaries in this region has a recurrent complex copy-number structure that reflects local sequence features. Higher density clone coverage would potentially narrow the size of this cluster region and would potentially reveal other breakpoint clusters in 5p.

The second sequence feature we found was a region of sequence duplication on 5p in the normal human genome. Data for 58 cases (e.g., fig. 1A and 1B,) indicated that clone 39 had a log₂ratio of ∼−0.4, intermediate between that expected for a normal region and that for a deletion, when the surrounding clones were clearly deleted. However, if the deletion extended proximally to include the region of clones 51 and 52, then the ratio on clone 39 decreased to the expected value. This suggests that a substantial fraction of the sequence of clone 39 was duplicated on 5p between clones 51 and 52, so that deletions including only clone 39 would represent a change from four copies to three (log₂[3/4] ∼−0.4). FISH analysis with this BAC as a probe confirmed this supposition, with the discovery of two closely spaced signals on 5p. Some freezes (June 2002, November 2002, and April 2003) of the human genome sequence have indicated the presence of this duplication, locating the marker in this BAC at both 20.791 Mb and 34.070 Mb (April 2003). The July 2003 freeze we have used does not indicate the duplication. Large duplications of this type occur frequently in the genome (Lupski et al. 1996; Eichler ²⁰⁰¹). No polymorphism in DNA copy number (Albertson and Pinkel 2003; Iafrate et al. ²⁰⁰⁴; Sebat et al. ²⁰⁰⁴) that involved clone 39 was found in this set of samples. Further indication of the likely sequence complexity in this region is the fact that the sequence assembly disagrees with the clone order indicated by the simplest interpretation of our deletion boundaries near clone 51, as noted in table 3.

Dependence of MR on Deletions

Figure 2A shows the relationship of deletions on 5p to the MR status of the patients, with the patients ordered by increasing level of retardation. The relationship appears complex and somewhat perplexing. There is a clear increase in the severity of retardation with an increasing extent of deletion, but there are some significant exceptions. For example, there are some patients with interstitial deletions on 5p who are either unaffected or minimally retarded, whereas others with smaller deletions in the same region are profoundly retarded.

Conventional cytogenetic and FISH analyses had previously been performed on these subjects and had shown that the chromosomal aberrations in some of the patients were complex, involving rearrangements of chromosomes in addition to chromosome 5. Therefore, we analyzed 37 of the patients, using arrays that provided genomewide data. These cases included all those whose 5p deletions seemed too small to explain their retardation level, when compared with the general trend (shown in fig. 2A), plus an approximately equal number that fit the trend. All subjects with severe retardation and small 5p deletions were found to have additional gains or losses (e.g., patient 45 [shown in fig. 1C]). No differential phenotypic effects could be specifically attributed to any of these additional aberrations. Details of the subjects with the additional aberrations are listed in table 4. Because arrays with evolving clone compositions were used for these measurements, the precision with which the boundaries of the additional aberrations are defined is variable. However, it is clear that these aberrations typically involve at least several contiguous array elements, making them considerably larger in extent than the copy-number polymorphisms in unaffected individuals (Albertson and Pinkel 2003; Iafrate et al. ²⁰⁰⁴; Sebat et al. ²⁰⁰⁴). Some of these complex cases involve additional aberrations affecting chromosome 5, including several with duplications of material proximal to the deletion boundary. Figure 2B shows the relationship between deletion and MR level for all patients with 5p deletions and no other detected aberrations. The relationship between severity of MR and deletion size and location is now much more consistent.

Chromosomal Regions Affecting Cry, Speech Delay, and Facial Dysmorphology

The chromosomal regions affecting cry, speech, and facial features have previously been mapped to the distal portion of 5p (Overhauser et al. 1994; Gersh et al. ^{1995, 1997}). Figure 3 shows our data for the subjects with deletion boundaries that provide information on these genotype-phenotype relationships. The clones at the boundaries are indicated for each case. The region that affects the characteristic cry is most narrowly defined by the difference in patients 49 and 252, both of whom have interstitial deletions. For patient 49, who has the cry, the deletion begins proximal to clone 12, whereas for patient 252, who does not have the cry, the deletion may retain sequences up to clone 14. Thus, the chromosomal region responsible for this phenotype is located in the 1.5-Mb region between these clones.

The region affecting the characteristic facial features is most narrowly defined by patients 117, 202, and 114. Patient 117 has a terminal deletion beginning between clones 17 and 18 and has no facial phenotype. Thus, the important region must lie proximal to clone 17, since material may be retained until this position. Patient 202 has a larger terminal deletion, with a boundary between clones 26 and 27, and has the phenotype. This defines a 2.4-Mb region between clones 17 and 27 as the region responsible for the facial features. Patient 114 supports the proximal limit of this region. Patient 114 has an interstitial deletion, with the distal boundary between clones 26 and 27, and has no phenotype, which places the critical location for this phenotype distal to clone 27. We note that patient 252, who has a deletion that includes this proposed critical region, is discordant, because the patient does not have the phenotype.

The chromosomal region responsible for speech delay is best defined by the data from patients 102, 225, and 49, who have interstitial deletions. In patients 102 and 225, who have the phenotype, the deletion is proximal to clone 4. In patient 49, who does not have the phenotype, the distal portion of the chromosome may be retained up to clone 13. Thus, speech delay is due to deletion within a 3.2-Mb region between clones 4 and 13.