Copy-Number Variations Measured by Single-Nucleotide–Polymorphism Oligonucleotide Arrays in Patients with Mental Retardation

Abstract

Whole-genome analysis using high-density single-nucleotide–polymorphism oligonucleotide arrays allows identification of microdeletions, microduplications, and uniparental disomies. We studied 67 children with unexplained mental retardation with normal karyotypes, as assessed by G-banded chromosome analyses. Their DNAs were analyzed with Affymetrix 100K arrays. We detected 11 copy-number variations that most likely are causative of mental retardation, because they either arose de novo (9 cases) and/or overlapped with known microdeletions (2 cases). The eight deletions and three duplications varied in size from 200 kb to 7.5 Mb. Of the 11 copy-number variations, 5 were flanked by low-copy repeats. Two of those, on chromosomes 15q25.2 and Xp22.31, have not been described before and have a high probability of being causative of new deletion and duplication syndromes, respectively. In one patient, we found a deletion affecting only a single gene, MBD5, which codes for the methyl-CpG-binding domain protein 5. In addition to the 67 children, we investigated 4 mentally retarded children with apparent balanced translocations and detected four deletions at breakpoint regions ranging in size from 1.1 to 14 Mb.

Mental retardation (MR) has a prevalence of ∼2%–3%.¹ Whereas the frequencies of mild MR differ among studies, most authors agree that severe MR, defined as an intelligence quotient (IQ) of <50, has a prevalence of 0.3%–0.4%.² Aside from trisomy 21, which accounts for ∼5%–15% of MR cases,² chromosome abnormalities are detected in not more than 5% of cases by cytogenetic analysis of chromosomes prepared from peripheral-blood lymphocytes.^3,4 The resolution of cytogenetic techniques is typically 5–10 Mb. Smaller rearrangements in the subtelomeric regions have been detected in ∼5% of affected children by regional analysis using FISH or muliplex ligation-dependent probe amplification.^5,6 It has further been shown that another 10%–20% of rearrangements can be found by array-based comparative genomic hybridization (array CGH).^7–10 Recently, high-density SNP microarrays have been evaluated for this purpose.^11,12 Special efforts have been undertaken to analyze regions that are flanked by segmental duplication, since these are known to predispose to recurrent rearrangements in genomic disorders.¹³ Array CGH and SNP microarrays have also revealed the presence of several thousand copy-number variations (CNVs) in the general population,¹⁴ complicating the interpretation of the findings in disease cases.

Here, we used Affymetrix GeneChip Human Mapping 100K arrays (100K arrays) to analyze DNA from 67 children with unexplained MR. Gains or losses that are likely to be causative of the disease were present in 11 (16%) of the children. We also show that the resolution of this array is sufficient to detect single gene deletions in regions with low gene content and that the determination of the breakpoints is precise enough to amplify the junction fragments after narrowing the breakpoint with few quantitative PCR (qPCR) products, unless the breakpoints are flanked by repetitive sequences or the rearrangement is complex.¹⁵

Material and Methods

Patients

The 67 children with unexplained MR were ascertained mainly in a single human genetics practice (that of S.S. and B.K.). We classified MR as mild or severe on the basis of clinical criteria and the competence of adaptive behavior. If available, an IQ of <50 on a standardized IQ test was used to classify MR as severe. According to these criteria, 61 cases were classified as severe, and 3 cases as mild. Most of the children had additional but often mild symptoms (table A1 in appendix A). Children with brain malformations were excluded from the study. All children had normal G-banded chromosomes (banding level 500–550). In 29 of the cases, cytogenetic analysis was performed in two different laboratories, with identical results. FISH with subtelomeric probes and metabolic investigations were inconspicuous in 42 and 11 children, respectively. DNA samples were available from the parents of 44 children, which allowed us to investigate whether CNVs originated de novo and to check which parental allele was lost, in cases of de novo deletions. In addition to the children with normal G-banded chromosomes, we analyzed DNA from four mentally retarded children in whom de novo balanced reciprocal translocations had been detected by cytogenetic analysis. DNA samples from 415 children referred for molecular diagnostics of the FMR1 gene but for whom no mutation was found were used for sequence analysis of the MBD5 gene. The study was approved by the Ethics Committee of the Medical Department of the Technical University Munich.

Array CGH

Genomic DNA was isolated from peripheral-blood leukocytes by use of a modified salting-out procedure.¹⁶ DNA concentrations were measured with a NanoDrop spectrophotometer (ND-1000 V.3.1.2). The 100K arrays consist of two arrays, the Xba240 and the Hind240 array, which together include 116,204 SNPs with an average spacing of 23.6 kb. DNA was processed in accordance with the manufacturer′s instructions. In brief, 250 ng of total genomic DNA was digested with XbaI or HindIII and then was ligated to adaptors. A generic primer that recognizes the adaptor sequence was used to amplify the adaptor-ligated DNA fragments in a GeneAmp PCR System 9700 (Applied Biosystems). After purification with the Macherey-Nagel NucleoFast 96 PCR Clean-Up ultrafiltration technology, a total of 40 μg of PCR product was fragmented and labeled with biotin. Hybridization was performed in the Affymetrix GeneChip Hybridization Oven 640. Arrays were washed and stained with the Affymetrix GeneChip Fluidics Station 450 and were scanned with the Affymetrix GeneChip Scanner 3000 7G. Image processing was performed with GCOS 1.4, and genotypes were called with GTYPE 4.0 software by use of the default call threshold of 0.25.

Data Analysis

To account for experimental variations, we hierarchically clustered the euclidean distance matrix generated from the binary logarithm of the sum of the median-normalized intensity values of both alleles. The arrays were then ordered into groups with a similar intensity profile. For each group, copy numbers were calculated as follows. We first determined the raw intensity values at each SNP locus by calculating separately the mean of the perfect-match probes for the A and B alleles. The raw intensity values were then median normalized. The log₂ ratio of the intensity values were calculated separately for the three genotypes (AA, AB, and BB) and the no-calls by dividing the normalized intensity values of the test array by the median values of all arrays with the same genotype for each SNP locus. One can show that the noise is lower when genotype-specific intensity values are used than when the sum of the intensity of both alleles for all genotypes is used. To make intensity values from male and female X chromosomes comparable, the mean dosage of the male SNPs on the X chromosome was adjusted to the mean dosage of the autosomal SNPs.

In many dosage plots, we observed that the mean dosage level depended on the length of the restriction fragments. Thus, we corrected for this dependence by using quadratic regression as described elsewhere.¹⁷ This increased the signal-to-noise ratio (SNR) further.

We used two measures for the assessment of data quality. First, we calculated the SD and the median absolute deviation (MAD) of the final log₂ intensity ratios. Second, we calculated an SNR in male DNA samples by substracting the median log₂ intensity ratio of the X-chromosomal SNPs from the median log₂ ratio of the autosomal SNPs. This difference was then divided by half the sum of the MAD of the log₂ intensity ratio of the autosomal and X-chromosomal SNPs,

Note that the numerator measures the separation between the log₂ intensity levels of autosomal (“nonX”) and X-chromosomal SNPs in males. Because of technological nonlinearities, this difference is <1. We found an average difference of 0.63 in 95 HindIII arrays and of 0.70 in 98 XbaI arrays. The denominator estimates the scale of the variation, putting equal emphasis on the autosomal and X-chromosomal SNPs. All characteristics are estimated robustly, to safeguard against deviations from the standard normal model and against outliers.

The minimal number of consecutive lowly expressed SNPs that significantly indicates a deletion was calculated with the expression

where sep_{1copy,2copies} is the median log₂ intensity ratio of SNPs with 1 copy or 2 copies, respectively; P denotes probabilities for CN, the measured copy number of a SNP with real copy number 2 under a normal model; and Φ is the distribution function of the standard normal distribution. The factor 67×10⁵ indicates the Bonferroni correction for multiple testing.

We also implemented tools to select regions conspicuous for gains and losses and to detect loss of parental alleles. To select CNVs, we took into account the array-specific SD. The number of consecutive SNPs had to be larger when the SD was higher. As a minimum, we used the median of five consecutive SNPs to define a CNV. Conspicuous regions were compared with known CNVs, as provided by the Database of Genomic Variants and the DECIPHER database. All analysis tools were implemented as R or Perl scripts (available at Scripts Web site).

qPCR

We designed 1–3 amplicons for validation of each CNV (table A2 in appendix A). qPCR was performed on a 7900HT real-time PCR system (Applied Biosystems) by use of SYBR Green I for detection. Reaction mixtures contained 0.2 μM of each primer and 10 μl of 2 × Power SYBR Green PCR Master Mix (Applied Biosystems). Each assay included a no-template control, two male and two female control DNAs, and the patient DNA at a final concentration of 2.5 ng/μl in duplicate, in a total volume of 20 μl. We used the same cycling conditions for all reactions: initial step at 50°C for 2 min, denaturation at 95°C for 10 min, then 40 cycles at 95°C for 15 s, and a combined annealing and extension at 58°C for 60 s. To exclude the presence of unspecific products, a melting-curve analysis of the products was performed after completion of the amplification. To control for differences in DNA concentration, reaction efficiency, and threshold cycles (C_t), the C_t values were normalized using the C_t value of a reference gene (BNC1) for each DNA sample. Analysis was performed as relative quantification by use of the comparative C_t method.

Breakpoint Analysis

We used qPCR to narrow the interval of the deletion breakpoints to 2–5 kb and generated junction fragments spanning the breakpoints by long-range PCR (table A2). Junction fragments were cloned into the pGEM-T vector (Promega) and were sequenced using BigDye v3.1 cycle sequencing (Applied Biosystems).

Mutation Analysis

Dye-binding/high-resolution DNA melting analysis was used to screen for single-nucleotide variations in MBD5. Unlabeled primers flanking each coding exon were designed with the ExonPrimer software (table A2). Genomic DNA (∼10 ng) was subjected to PCR amplification performed in 5 μl total volume containing 1× Thermo-Start High Performance Buffer (ABgene), MgCl₂ (1.25 mM), 100 μM of each deoxynucleotide triphosphate, 0.25 U of Thermo-Start DNA polymerase (ABgene), primers (0.4 μM each), and the dye LCGreen PLUS (Idaho Technology) at 1× final concentration. After PCR, the samples were heated to 94°C for 30 s and then were cooled to 20°C before melting. Melting acquisition was performed on a LightScanner HR I 384 instrument (Idaho Technology) in accordance with the manufacturer′s standard procedures. Melting-curve data were analyzed with the standard software provided by Idaho Technology. Abnormal melting profiles were confirmed or excluded by sequencing of independent PCR products.

X-Chromosome Inactivation Analysis

For the investigation of the X-chromosome pattern, we used the trinucleotide repeat in the first exon of the androgen-receptor gene.¹⁸ A total of 2 μg of DNA was digested with 20 units of the methylation-sensitive restriction enzymes HhaI and HpaII (New England Biolabs). After overnight digestion, an aliquot of 2 μl was amplified by PCR by use of the FAM-labeled primer 5′-TCCAGAATCTGTTCCAGAGCGTGC-3′ and the unlabeled primer 5′-GCTGTGAAGGTTGCTGTTCCTCAT-3′. The PCR products were separated on an ABI Prism 3730 DNA sequencer and were analyzed with the GeneMapper software (Applied Biosystems).

Results

Array Data Analysis

DNA from 67 patients with MR and normal G-banded chromosomes and 4 patients with MR and a balanced translocation diagnosis were analyzed with 100K arrays. These arrays use a one-color technique; therefore, the normalized intensity values of a test chip have to be compared with one or more reference chips. We first assessed the data quality by using the SD of the copy-number values of all SNPs. Preliminary analysis showed that using the arrays processed during the study instead of external reference arrays provided by the manufacturer resulted in better data quality. We used hierarchical clustering to group arrays with similar intensity profiles and observed that the SNR could be increased when arrays with similar intensity profiles were analyzed together. The HindIII and XbaI arrays showed at least four and three groups, respectively, with clearly distinctive intensity profiles (fig. 1). Differences in the intensity profiles are most likely caused by DNA and experimental variations. Notably, arrays processed in a single experiment tended to show similar profiles. We eventually determined genotype-specific log₂ intensity ratios for each SNP locus within each group. To increase the SNR further, we corrected for the dependence of the log₂ intensity ratio on the fragment length by using quadratic regression. In total, we analyzed 169 array sets. The data quality was assessed by the SD and the MAD of the log₂ intensity ratio of all autosomal probes; for the HindIII arrays, the mean SD was 0.19 and the mean MAD was 0.16, and, for the XbaI arrays, the mean SD was 0.18 and the mean MAD was 0.16. These values are in accordance with previously published reports.^14,17

Figure 1. .

Heat map of the median normalized intensity values of 169 HindIII arrays. The log₂ intensity values of 1,000 SNPs are displayed after hierarchical clustering. The sum of the intensity values of alleles A and B was used for this calculation. The columns contain the different arrays; the rows contain the different SNPs. The banner across the top of the heat map shows a color code for the four different time points at which the arrays were hybridized. Although the copy-number profile should be almost the same for each chip, it shows at least four clearly separated groups.

Different array types and platforms show different rates of the increase and decrease for the log₂ intensities because of duplications and deletions, respectively. Thus, the MAD or SD alone is not appropriate for the comparison of different array types and platforms. To render our results comparable across different platforms, we estimated the SNR by using the difference of the log₂ intensity ratios of the autosomal and the X-chromosomal SNPs of male samples. The HindIII arrays showed a mean SNR of 4.6, and the XbaI arrays showed a mean SNR of 5.14. These values mean that, on average, any six SNPs for HindIII arrays and five consecutive SNPs for XbaI arrays indicate a deletion at a 95% significance level over all 67 arrays. In our real data, the individual arrays had different SDs, and the log₂ intensity ratios were not normally distributed. We therefore adjusted the number of consecutive SNPs that define a candidate region, depending on the SD, so that we obtained at most eight CNVs per sample.

CNVs

By applying our rules for CNV detection, we obtained 27 candidate regions in 24 patients (table 1). All regions that considerably overlapped with known CNVs provided by the Database of Genomic Variants had previously been excluded. The 27 candidate regions were evaluated by qPCR. The 14 regions that were defined by >20 SNPs could be confirmed by qPCR. Of the 13 regions defined by 5–20 SNPs, 5 were determined to be false-positive findings. In summary, we could confirm 22 CNVs (table 1). They varied in size from 10 kb to 7.5 Mb and contained up to 49 genes. Genotype information revealed that four CNVs originated on the paternal chromosome and two CNVs on the maternal chromosome. From these 22 confirmed CNVs, we further excluded five regions that were inherited from one parent and six regions that did not contain known coding regions or that could not be tested to determine whether they occurred de novo, because we did not have parental DNA available.

Table 1. .

qPCR^[Note]

			SNP								Overlap(%)
Patient ID	Chromosome	Gain or Loss	Start	End	No. of SNPs	Region Length (Mb)	qPCRa	No. of qPCRs	Status of Inheritance	No. of Genes	CNVb	DGVc	Primer(s)
27384	1q31.1-q31.3	Loss	rs10494585	rs10494695	464	7.75	Yes	1	De novo	17	10	100	8920
30437	2p25.3-p25.1	Loss	rs2313466	rs1964092	193	3.79	Yes	1	De novo	4	5	100	8930, 9069, and 9070
29922	8p23.1	Loss	rs2945251	rs2466115	177	3.9	Yes	3	NA	34	6	30	8926, 9061, and 9062
30375	3p25.3-p25.2	Loss	rs10510400	rs10510422	98	2.83	Yes	3	De novo	39	4	100	8928, 9065, and 9066
28735	12p13.33	Loss	rs953385	rs2283285	62	2.01	Yes	3	De novo	18	27	100	8924, 9077, and 9078
30428	17q21.31	Loss	rs436667	rs1918798	38	0.48	Yes	1	De novo	7	6	9	8929, 9067, and 9068
28283	7q21.13	Loss	rs10487988	rs10488004	36	0.27	Yes	3	Paternal	1	0	0	9651, 9652, and 9653
29836	Xp22.31	Gain	rs719632	rs10521669	31	1.42	Yes	4	Maternal	5	47	100	9820, 9821, 9822, and 9060
27737	17p11.2	Gain	rs4073940	rs1373147	29	3.22	Yes	3	NA	49	11	100	8921, 9054, and 9055
28430	15q25.2	Loss	rs17158372	rs10520569	27	1.37	Yes	3	De novo	11	11	100	8923, 9058, and 9059
27581	9p23	Loss	rs10511570	rs1900218	23	0.28	Yes	1	NA	0	85	63	8934
28701	13q12.11	Gain	rs9315234	rs4570685	23	0.53	Yes	3	De novo	5	18	46	8922, 9056, and 9057
29945	3p13	Loss	rs10511008	rs10511014	22	0.44	Yes	5	NA	1	0	0	8939, 9063, and 9064
27581	1q25.2	Gain	rs10494517	rs7555418	20	0.59	No	3	…	…	…	…	8932, 9073, and 9074
27831	3q24	Loss	rs2140300	rs10513274	19	0.19	Yes	1	NA	0	24	100	8935
27526	Xp22.31	Gain	rs10521668	rs5934414	15	0.34	No	4	…	…	…	…	9820, 9821, 9822, and 9060
29608	4q28.3	Gain	rs1495265	rs10518595	14	0.22	Yes	1	NA	0	100	34	8937
27733	11q14.1	Gain	rs870066	rs10501436	13	0.11	No	1	…	…	…	…	8941
29195	2q23.1	Loss	rs2890919	rs10497034	13	0.2	Yes	2	De novo	…	0	0	8936 and 8948
27733	9p23	Loss	rs372412	rs983282	11	0.14	Yes	1	NA	0	37	13	8940
29996	5p15.2	Loss	rs31953	rs26152	10	0.03	Yes	1	NA	1	0	0	8944
30227	16p13.12	Loss	rs190013	rs9452	8	0.05	Yes	1	Maternal	1	2	2	8946
29700	13q31.2	Loss	rs221022	rs452708	7	0.03	Yes	4	Paternal	0	100	7	8938, 9655, 9656, and 9657
29945	3p14.2	Loss	rs9311853	rs10510892	7	0.05	No	…	…	…	…	…	8939
28283	8q22.1	Loss	rs6996243	rs1378125	6	0.66	No	3	…	…	…	…	9648, 9649, and 9650
29199	6q22.33	Loss	rs3778130	rs265353	6	0.01	Yes	1	Maternal	1	0	0	8942
30221	1q23.3	Gain	rs9330294	rs10494355	6	0.06	Yes	2	Paternal	1	46	11	8945 and 9076

Note.— The breakpoint regions were sequenced in individuals 29195 and 29945 (GenBank accession numbers EF504248 and EF504249). NA = parental DNA not available.

^aConfirmation by qPCR.

^bPercentage of the CNV detected in this study that overlaps with one or more CNVs listed in the Database of Genomic Variants.

^cPercentage of CNVs listed in the Database of Genomic Variants that overlaps with the CNV detected in this study.

We finally considered 11 CNVs—8 deletions and 3 duplications—to be causative of MR by criteria as follows (fig. 2 and table 2). Eight of the rearrangements were assumed to be causative because they occurred de novo. Two rearrangements, in 8p23.1 and 17p11.2, were assumed to be causative because they overlapped with known deletion or duplication syndromes (e.g., 17p11.2 duplication syndrome [MIM 610883]), although they could not be proved to have occurred de novo, because of missing paternal DNA. At last, a maternally inherited 1.4-Mb duplication in Xp22.31, including the STS gene, in a male patient was thought to be causative because deletions of this region are known to cause MR and because the chromosome carrying the duplication was nonrandomly inactivated in the mother.

Figure 2. .

Genomic profiles showing the log₂ intensity ratios of CNVs and of the surrounding genomic regions detected in patients with MR. The length of the CNVs is indicated on the X-axis. log₂ Intensity ratios are calculated as described in the “Material and Methods” section. Displayed are the log₂ intensity ratios after median smoothing with a window of 9. The dosage values of homozygous and heterozygous SNPs are depicted in black and gray, respectively. The gray horizontal lines are drawn at log₂(0.75) and log₂(1.25).

Table 2. .

Deletions and Duplications^[Note]

			SNP						LOHa
Patient ID	Gain or Loss	Chromosome	Start	End	Region Length (Mb)	No. of SNPs	No. of Genes	Segmental Duplications	Paternal	Maternal	No. of qPCRs	Confirmation by Second Hybridization
27384	Loss	1q31.1-31.3	rs10494585	rs10494695	7.5	464	17	No	48	0	1	Yes
29922	Loss	8p23.1	rs2945251	rs2466115	3.9	177	34	Yes	NA	0	3	No
30437	Loss	2p25.3-25.1	rs2313466	rs1964092	3.8	193	4	No	0	18	1	No
27737	Gain	17p11.2	rs4073940	rs1373147	3.2	29	49	Yes	NA	…	3	Yes
30375	Loss	3p25.3-25.2	rs10510400	rs10510422	2.7	98	39	No	8	0	3	No
28735	Loss	12p13.33	rs953385	rs2283285	2.0	62	18	No	4	0	3	No
28430	Loss	15q25.2	rs17158372	rs10520569	1.4	27	11	Yes	0	2	3	Yes
29836	Gain	Xp22.31	rs719632	rs10521669	1.4	31	5	Yes	…	…	3	Yes
30428	Loss	17q21.31	rs436667	rs1918798	0.5	38	7	Yes	26	0	1	Yes
28701	Gain	13q12.11	rs9315234	rs4570685	0.5	23	5	No	…	…	3	Yes
29195	Loss	2q23.1	rs2890919	rs10497034	0.2	13	1	No	0	0	2	Yes

Note.— NA = not available.

^aThe number of SNPs for which a paternal or maternal allele is missing. LOH = loss of heterozygosity.

In addition, we investigated DNA of four mentally retarded children with de novo translocations—three terminal and one interstitial translocation—containing a total of eight breakpoints. According to GTG-banding analysis, they appeared to be balanced. By microarray analysis, we detected deletions at two of the six breakpoints in the terminal translocations and at two of the breakpoints in the interstitial translocation containing 3–63 genes (table 3).

Table 3. .

Reciprocal Translocations

	SNP
Patient ID, GTG Banding,a and Chromosomeb	Start	End	Region Length (Mb)	No. of SNPs	No. of Genes
31922, 46,XY,t(12;13)(q22;q32):
13q33.2-33.3	rs2149144	rs9301245	1.8	146	3
28526, 46,XY,t(1;10)(p13.1;p13):
10p14	rs2259442	rs1243963	1.1	57	8
28181, 46,XY,t(4;6)(q28.3q31.1;q23.1q22.2):
4q28.3-31.1	rs10519357	rs6841039	3.9	186	6
6q16.1-21	rs9320518	rs6568702	14.3	625	63

^aGTG banding (with 500–550 bands) contains results of karyotyping.

^bChromosome contains results of microarray experiment.

CNVs Flanked by Low-Copy Repeats

Of the 11 validated CNVs, 5 were flanked by low-copy repeats (LCRs). The first one was a 3.9-Mb deletion in 8p23.1, indicated by 177 SNPs in a child (patient identification [ID] 29922) with mild MR. The mother carried two copies, and the maternal genotypes in the deleted region are consistent with the boy carrying one of the maternal alleles. Thus, it is most likely that the deletion occurred on the paternal chromosome, although we did not have paternal DNA available for investigation. The deletion overlapped with the known 8p23.1 deletion syndrome and included the GATA4 gene, as confirmed by qPCR.¹⁹ The syndromic features of our patient included microcephaly, hypospadia, and an atrial septum defect described to be attributable to haploinsufficiency of GATA4.²⁰ The child did not present with facial dysmorphisms, diaphragmatic hernia, or epilepsy.

The second CNV flanked by LCRs was a duplication of 3.2 Mb in 17p11.2, indicated by 29 SNPs in a boy (ID 27737) with mild MR. His mother carried two copies. DNA from the father was not available. As estimated by the dosage values, the duplication coincides with the common 3.7-Mb interstitial duplication of the 17p11.2 duplication syndrome, a region that harbors deletions in Smith-Magenis syndrome (MIM 182290).^21,22 The patient we investigated had normal birth weight and length, but feeding was poor, and he failed to thrive postnatally. He had a hypospadia grade I. At age 10 mo, his length was in the 10th percentile, his head circumference was <3rd percentile, and he was found to have hypotonia. Facial features, such as telecanthus, triangular face, and broad forehead, were in accordance with the findings recently described in other children who were given a diagnosis of Potocki-Lupski syndrome (fig. 3A).²² Later, the mother reported that the child started speaking single words only at age 27 mo.

Figure 3. .

Facial features of patients. A, Patient 27737, with a common 17p11.2 duplication, at age 10 mo, showing a telecanthus, triangular face, and broad forehead. B, Patient 29195, with an MBD5 deletion, at age 15 mo, showing a normal facies without dysmorphic features.

The third CNV flanked by LCRs was a 500-kb deletion in 17q21.3 in a girl (ID 30428) with moderate MR. The deletion encompasses a known inversion polymorphism.²³ We used the genotype information from 38 SNPs within the deleted region to define the haplotypes in the girl and her parents. The father was homozygous for the H2 haplotype, and the mother was homozygous for the H1 haplotype. The deletion occurred on one of the paternal H2 haplotypes. Deletions of the same region have recently been found in 10 other patients and have defined a new microdeletion syndrome.^13,24,25 The deletions reported therein also occurred on the H2 haplotype and may be facilitated by the direct orientation of the repeats on this haplotype.

The fourth CNV flanked by LCRs was a 1.4-Mb deletion on chromosome 15q25.2, which comprises ∼11 genes, in a patient (ID 28430) with mild MR. Both parents carried two copies of the region. The deletion was indicated by the dosage values of 27 SNPs. Two informative SNPs showed that the maternal allele was deleted. The patient was affected by intrauterine growth retardation. She was born in the 38th wk of gestation, with a weight of 1,950 g and a length of 42 cm. She had only very slight dysmorphic signs. The main symptoms were psychomotor retardation, polysplenia, and a hypoproliferative, macrocytic anemia that developed in the 1st year of life and that required blood transfusions until she was age 4 years. Because of short stature and anemia, the diagnosis Diamond-Blackfan anemia (MIM 105650) was considered. At age 11 years, she was referred to the emergency ward with bleeding from esophageal varices. Ultrasound sonography revealed a severe portal vein stenosis, although liver structure was normal.

The fifth CNV flanked by LCRs was a 1.4-Mb duplication in Xp22.31 in a boy (ID 29836) with severe MR. The duplication was located between the VCX3A and VCX2 genes and was indicated by 18 SNPs (fig. 4A). qPCR revealed a single dose for amplicons ∼6 kb distal to VCX3A and 3 kb proximal to VCX2, whereas amplicons 3 kb proximal to VCX3A and 5 kb distal to VCX2 showed a double dose. Therefore, VCX3A or VCX2 is duplicated or an additional fusion product of both genes was generated, depending on where the recombination occurred. In addition, all genes between VCX3A and VCX2 are duplicated (HDHD1A, STS, VCX, and PNPLA4). Investigations of the parental DNAs showed that the father had a single copy of this region, whereas the healthy mother had three copies and therefore is a carrier of the duplication. Studies of the X-inactivation pattern in peripheral-blood DNA of the mother showed a nonrandom X-inactivation, with the chromosome carrying the duplication being inactivated. We therefore conclude that the duplication is most likely causative of the MR in her son. He had normal motor development but showed, in addition to MR, speech and language deficits. At age 9 years, he used only 2-word sentences. He developed postnatal microcephaly, with a head circumference of 50 cm (3rd percentile) at age 9 years. A recent report describes additional patients with MR carrying duplications of this region.²⁶ The same region contains deletions in 80%–90% of patients with X-linked ichthyosis (MIM 308100) caused by deficiency of the steroid sulfatase enzyme, encoded by the STS gene.²⁷ Most patients with deletions are affected only by ichthyosis, but a few also display MR, although the size of the deletion is often very similar to that of the deletion breakpoints situated near the directly orientated genes VCX3A (VCXA) and VCX2 (VCXB), which are embedded in an LCR of ∼9 kb. Data from two studies suggested that, depending on whether VCX-proximal or VCX-distal homologous sequences are used for recombination, VCX3A is deleted and VCX2 is retained, or vice versa.^28,29 MR seems to be associated with the deletion of VCX3A,²⁹ although the data are not unequivocal, because patients with Xp;Yq translocations leading to a deletion of VCX3A have been reported to have normal intelligence.

Figure 4. .

Schematic representation of the regions affected in patients 29836 and 29195. A, Duplication in Xp22.31 in patient 29836. Genes are indicated by white boxes. The arrows denote the direction of transcription. The primers used for qPCR are shown above the genomic representation as numbered black bars. Primers 9060 and 9821 showed a single dosage; primers 9822 and 9820, a double dosage. The nonallelic homologous recombination must have occurred within the homologous regions flanked by primers 9060 and 9820 and primers 9822 and 9821. B, Deletion of part of the MBD5 gene in patient 29195. Exons are indicated by numbered black boxes (untranslated sequence) or gray boxes (translated sequence). The extent of the deletion is shown by a black bar below the gene scheme. Exons 1A–1E of MBD5 denote the 5′ untranslated exons established in this study.

CNVs Not Flanked by LCRs

In six patients, de novo CNVs could be identified, with a size between 200 kb and 7.5 Mb. The child (ID 27384) with the largest deletion (7.5 Mb) has psychomotor retardation but did not have additional symptoms besides slight dysmorphic facial features. The most prominent syndromic features in these six patients included hydrocephalus and cleft palate (in patient 28701), microcephaly and adipositas (in patient 30437), and an ostium secundum defect (in patient 30375). The 2.7-Mb deletion in the last boy (ID 30437) overlapped with the proximal part of the 3p syndrome, which has been described to cause atrioventricular septal defects in most of the affected individuals.³⁰ The 3.8-Mb deletion on 2p25.1-25.3 contained only four genes, including SOX11. This gene is expressed transiently during embryonic development in many tissues and causes severe defects in several organ systems in homozygous Sox11-deficient mice, leading to death shortly after birth.³¹

The smallest CNV suggestive of being causative of the clinical phenotype observed is a 200-kb deletion on chromosome 2q23.1 in a boy (ID 29195) with severe MR. The deletion was indicated by 13 SNPs, which affected the first 7 of 10 exons of the methyl-CpG–binding domain protein 5 gene (MBD5) (fig. 4B). The deletion region was present in the parental DNA but provided no information with regard to the origin of the deletion. Sequencing of the coding region of the remaining allele did not show any deviation from the reference sequence (GenBank accession number NM_018328). The junction fragment was amplified and sequenced after the breakpoints were narrowed by qPCR (GenBank accession number EF504248). The proximal breakpoint lies within an LTR/ERVL repeat (MLT2B5), and the distal breakpoint lies within single-copy sequence in intron 7 of the MBD5 gene. ESTs suggested that the GenBank entry for MBD5 is 5′ incomplete. We extended the mRNA by five noncoding exons, using 5′ rapid amplification of cDNA ends (RACE) and RT-PCR (GenBank accession number EF542797). The boy had a sandal gap between the first and second toe but no facial dysmorphic features (fig. 3B). In addition to MR, motor development was slightly retarded. At age 8 mo, he had febrile seizures. First seizures without fever started at age 16 mo and proved to be drug resistant. The boy is hypoactive, and social interactions are very limited. To provide further evidence of the pathogenicity, we screened 415 DNAs from children with MR. We found four missense variants that were not present in ∼660 controls (table 4). We did not have parental DNAs available to check whether these variants occurred de novo.

Table 4. .

Variations in MBD5

				No. of Control Individuals with Genotype
Variation Type and Sequence Change	Protein Change	Patient ID(s)	Exon/Intron	11	12	22
Nonsynonymous:
c.431C→T	p.T144I	A12	4	CC: 660	CT: 0	TT: 0
c.1368G→T	p.S456K	A6 and B135	4	GG: 663	GT: 1	TT: 0
c.1382G→A	p.R461H	A47 and B217	4	GG: 649	GA: 0	AA: 0
c.1962C→A	p.D654E	C231	4	CC: 655	CA: 0	AA: 0
c.1963G→A	p.A655T	30224a	4	GG: 653	GA: 0	AA: 0
c.2030G→A	p.S677N	B134, C264, and C281	4	GG: 662	GA: 3	AA: 0
c.2569G→A	p.A857T	31833a	5	GG: 670	GA: 0	AA: 0
c.3143C→T	p.T1048I	C260	7	CC: 640	CT: 0	TT: 0
Synonymous:
c.1638C→T	p.A546A	A109 and B139	4	CC: 658	CT: 3	TT: 0
c.2286C→T	p.H762H	B225	4	CC: 338	CT: 0	TT: 0
c.3279C→T	p.V1094V	A53	7	TT: 648	CT: 0	CC: 0

^aOne of the parents carries the same mutation.

MBD5 was originally identified because of sequence homologies to MBD1–MBD4 and MECP2, which is mutated in Rett syndrome.³² It also contains a PWWP motif, which has been shown to bind to DNA and is found in proteins often containing other chromatin-association domains.³³

Discussion

Hybridization using synthetic oligonucleotide arrays offers great promise for the detection of submicroscopic deletions and duplications. As the technology evolves, driven mainly by the need for dense SNP arrays in genomewide association studies, the resolution to detect relatively small deletions and duplications with strong effects on rare phenotypes will improve. We used arrays that cover the entire genome, with 100,000 SNPs, to test them against 71 patients with MR. Samples from all 71 patients had previously been subjected to cytogenetic analysis, because they had slight or overt symptoms characteristic of chromosomal abnormalities.

Our study led to the identification of 11 deletions or duplications that were interpreted to be causative, because they could be shown to have occurred de novo or because they corresponded to established disease-associated indel mutations. Our detection rate of 16% is in the same range as that of earlier studies that used 100K arrays¹² or whole-genome tiling-path resolution array CGH.⁷ The patient (ID 29922) with 8p23.1 deletion syndrome and the patient (ID 27737) with 17p11.2 duplication syndrome raise the question of why the clinical picture in these patients did not lead to a specific diagnosis and testing. In both cases, the differential diagnosis did include the mentioned syndromes, but its features were by no means unequivocal. In the case of the 17q21.3 deletion, the genotype-phenotype connection was established only during the course of this study.^13,24,25

We identified two CNVs, in 15q25.2 and Xp22.31, with a high probability of being causative of new deletion and duplication syndromes. They were flanked by LCRs, which make it likely that other cases exist, both deletions and duplications. Whereas no other cases have been described so far for 15q25.2, duplications in Xp22.31, which are characterized by MR in combination with autistic behavior, seem to be more frequent.²⁶

The smallest CNV detected and concluded to be causative is a deletion of a single gene, MBD5 (in patient 29195). The deletion measured 200 kb in size. It was detected by 13 consecutive SNPs and deleted one noncoding and seven coding exons of the gene, which codes for the methyl-CpG–binding domain protein 5. Two previous reports and a DECIPHER entry (about patient 1079) have described contiguous gene deletions that also involve the MBD5 gene^7,34 (J. Veltman, personal communication). The clinical features of our case patient and two of the previously described patients include epileptic seizures. A mutation screen involving 415 DNAs of children with unexplained MR revealed four missense variants not present in 660 controls, but no deletions or nonsense mutations were found. We could not check whether the missense variants occurred de novo because we did not have parental DNA available. Thus, a confirmation that the MBD5 mutations are pathogenic awaits the screening of a panel enriched for epileptic seizures for which parental DNA is available.

The CNVs suggest that the genes involved cause the related pathology by dosage differences. The most likely group of genes to be considered for such a mechanism is transcription factors.³⁵ With the exception of the MBD5 deletion, the Xp22.31 duplication, and the 17q21.31 deletion, all regions contained at least one gene involved in the transcription process. From the 189 genes contained in the reported deletions and duplications, 129 have a Gene Ontology annotation, and 16 (12%) of those 129 are annotated as transcription factors.

We showed that 100K arrays allow the detection of CNVs with a size between 10 kb and 7.5 Mb, a range that is not reliably accessible by microscopic cytogenetic techniques. The quality of the intensity data obtained from the arrays was variable and depended on the experimental circumstances, restrictions that must be accounted for in the analysis procedure. The detection of CNVs by at least 20 neighboring SNPs was found to be reliable, whereas the detection of CNVs by 5–20 neighboring SNPs had a false-positive rate of 30%. The resolution might improve with new protocols and new generations of arrays that reduce the experimental error and provide a better SNR. The interpretation of the results in children with MR or other diseases is complicated by the presence of thousands of CNVs within the normal population, especially when their size is small. One usually applies the criterion that a CNV has to have occurred de novo for it to be likely pathogenic. However, the frequency of de novo deletions and duplications in newborns has been estimated to be 1 in 8 and 1 in 50, respectively, and it is obvious that not all of them can be assumed to be pathogenic.³⁶ On the other hand, de novo CNVs may be pathogenic. They may provide a risk that becomes manifest because of sequence variants in the other allele or the genetic background. Databases that collect information about CNVs in healthy individuals and individuals with diseases will become indispensable for their risk estimation. This will become even more important when improved array designs allow detection of intragenic deletions and duplications.

Appendix A

Table A1. .

Clinical Characteristics^[Note]

Patient ID	Sex	Father	Mother	GTG Banding	Subtelomer Screening	EEG	MRI	Metabolism	MR	Delayed Motor Development	Language and/or Speech Retardation	Facial Dysmorphism	Stature	Microcephalus	Miscellaneous
27384	F	27739	27738	2	+	+	+	+	Mod-sev	+		+
27524	M	NA	NA	1					ND
27525	M	NA	NA	1					ND			+	Short
27526	F	NA	NA	1					ND
27581	F	NA	NA	1	+				Mod-sev			+		+
27733	M	NA	NA	1					Mod-sev				Short	+	Clinodactyly
27737	M	NA	31778	1	+				Mild					+	Hypospadia, hypotonia
27831	M	NA	NA	1					Mod-sev	+				+
27877	F	27878	27879	1	+	+		+	Mod-sev	+	+	+		+	Wolff-Parkinson-White syndrome
28037	M	28039	28038	1					Mod-sev			+			Brachydactyly
28142	M	28144	28143	1					Mod-sev				Short		Aortic coarctation, hypospadia, omphalocele
28181	M	28183	28182	1					Mod-sev
28188	F	NA	28189	1					Mod-sev					+	Hypotelorism
28190	M	NA	28191	1	+				Mod-sev	+	+	+			Brachycephalus
28192	M	28194	28193	1					Mod-sev		+				Aggressive behavior
28283	M	28449	28284	1	+				Mod-sev	+				+
28289	M	28290	28291	1	+				Mod-sev	+	+	+			Dolichocephaly, hyperopia
28430	F	30806	30805	1					Mild	+		+	Short	+	Anemia
28448	F	NA	NA	1					ND				Short	+
28499	F	28500	28501	1					ND
28509	F	28510	28511	1	+	+			Mod-sev		+				Brachymetacarpy, hyperopia
28526	M	28527	28528	1					Mod-sev
28529	M	28530	28531	1					Mod-sev	+		+	Short
28631	M	28633	28632	1	+				Mod-sev			+	Short		Brachycephaly, myopia, autoimmune thyreoditis
28634	F	NA	28635	1	+			+	Mild			+	Tall
28651	M	NA	NA	2	+				Mod-sev	+		+	Short		Nystagmus
28701	M	28702	28703	1	+			+	Mod-sev			+			Hydrocephalus, cleft palate, tricuspid dysplasia
28730	M	28731	28734	1		+			Mod-sev	+	+	+			Muscle hypotonia
28735	M	28736	28737	2		+	+		Mod-sev	+	+
29195	M	29197	29196	2	+		+	+	Mod-sev	+		+
29199	F	NA	29198	1					Mod-sev			+	Tall
29200	M	29202	29201	1					Mod-sev
29608	M	NA	29610	2	+				Mod-sev	+	+	+			Epilepsy
29609	M	NA	NA	2	+				Mod-sev	+	+			+
29699	M	NA	NA	2	+	+	+		Mod-sev		+	+		+	Brachycephaly, epilepsy
29700	M	29701	29709	1					Mod-sev		+
29833	F	29835	29834	2	+				Mod-sev	+		+
29836	M	29838	29837	2	+	+			Mod-sev		+	+		+
29901	M	29903	29902	2	+	+			Mod-sev	+		+			Brachycephaly
29922	M	NA	29923	2	+				Mild					+	Hypospadia
29942	M	29944	29943	1					Mod-sev		+
29945	M	NA	33360	1					Mod-sev		+
29949	M	NA	NA	2	+	+			Mod-sev			+	Short		Brachydactyly, attention-deficit disorder
29950	M	29952	29951	2	+	+		+	Mod-sev		+	+		+	Brachycephaly, hyperopia
29962	F	NA	NA	1	+		+		Mod-sev	+	+	+
29993	M	29994	29994	2	+	+	+	+	Mod-sev	+		+			Hypotonia
29996	M	NA	NA	2	+				Mod-sev	+		+			Brachycephaly
30221	F	30223	30222	1	+			+	Mod-sev		+
30224	M	30226	30225	2	+				Mod-sev			+			Hypotonia
30227	M	30229	30228	2	+	+	+	+	Mod-sev		+	+			Dolichocephaly, autistic behavior
30241	F	30243	30242	2	+	+			Mod-sev	+	+	+		+
30303	M	30305	30304	1	+				Mod-sev	+	+	+		+
30372	M	30374	30373	2	+				Mod-sev	+	+	+		+
30375	M	30377	30376	2	+				ND	+		+			Atrial septal defect II
30392	M	30647	30393	2	+	+	+		Mod-sev	+	+
30428	F	30430	30429	1	+				Mod-sev	+	+	+		+
30431	M	30433	30432	2	+				Mod-sev	+	+	+
30434	M	30435	30436	2	+				Mod-sev			+
30437	M	30438	30550	2					Mod-sev			+		+
30508	M	NA	30510	1	+		+	+	Mod-sev	+	+			+
30520	M	30521	30522	1					Mod-sev
30569	F	30571	30570	2	+				Mod-sev	+	+	+			Epilepsy
30704	F	30706	30705	2	+	+	+	+	Mod-sev	+	+
30782	M	30784	30783	2	+				Mod-sev		+	+			Brachycephaly
30830	F	30832	30831	2	+				Mod-sev	+	+			+	Brachycephaly
30872	F	30874	30873	1					Mod-sev	+	+			+
30875	M	30877	30876	1					Mod-sev			+		+	Pulmonary stenosis, esophageal atresia, laryngomalacia
30909	M	30911	30910	1	+				Mod-sev		+	+
30944	F	30945	30946	1					Mod-sev					+
30987	M	30988	30989	2	+				Mod-sev			+
31154	M	NA	NA	1					Mod-sev	+	+
31922	M	NA	NA

Note.— EEG = electroencephalogram; Mod-sev = moderate to severe; MRI = magnetic resonance imaging; NA = not available; ND = not determined; a plus sign (+) = presence of feature.

Table A2. .

Primer Sequences

	Sequence(5′→3′)
Use and Primer	Forward	Reverse	Length (bp)	Annealing Temperature (°C)
qPCR:
8920	TGAAGGCTCTAAATCCCCAG	AGCAACCGCTAAAACCCAG	126	60
8921	GCGAGCCCGATTCTCTG	ACCAATCCTCAGGTCCAGC	126	60
8922	CTTCACTGGACCGAAACACC	TTTATCCCGATTGCTTCTGC	135	60
8923	AGCAGCCTCATGCTCTATGG	TGAAGTGTCCAGTCCAAGGC	128	60
8924	ATTTCCTCTATGGAGCGTGG	AATGGCTCCGATACACTTCC	127	60
8925	AACTGGGAAGGAGGTATCCG	AGGGAGCTCCAGCCAGC	133	60
8926	GGACCACCCTTCGGCTG	GGCTGACGATGTTCGAGG	124	60
8928	TGGTGATTACTTTTGACAGTCTTTG	GGAAAGAGTTGTAGCTCCCG	129	60
8929	CGAGTCGCTGACCAGTTACC	ATCTGCTTGTCCTGGCTGAC	126	60
8930	GACATGTGCAGCCGTGTG	ATGGGCTGTGTTGTCAAGG	124	60
8932	TTTTGGCCCAGTTTAGCC	AATATGATGGTGGCTGGCTC	130	60
8933	CTCATTCAGCCTCATCTCACC	CCTCCATGAAGGGCTAAAAC	131	60
8934	ATGCAAAATAATACCATCCAGAC	GCCACTTGTGGGAAGTGC	137	60
8935	CTTTTCATTAAATGTGCATTAACC	TCTGCAAAGCAACTGAACTG	135	60
8936	CCTCCTACTATATGAAAATTTTGGTCC	GCTGCTTAAGACTGCAAAGC	126	60
8937	TCAACTATGACAGGGTGTCTGC	CCCTTCAGAGGCAAATAATAGG	121	60
8938	AGATTTGGGCCCCTCAAC	AACAACGAGGAAGAAATATTGG	133	60
8939	GCAGCAGCAATATCCCTTTATG	GCTGGTAGTCAAGTAAGGCTTTG	128	60
8940	TGACTTAAATATCAATTGAGGATCAC	TTGAATTAACTGAAACCCAATCTC	135	60
8941	TCTGTCTCTGGATTCCTATTTGC	TCAAACACAAGTCTTGGTCTCC	130	60
8942	GGAAGATGTATTTGTCACTTTTCTTC	GCACAGTCAACCATTATTTTCTC	132	60
8944	AAAAGTGTGTGCAGAGCCAG	GATGAAGCAAGCACATAGCG	130	60
8945	ATTTGTCTTTGTCTCAATATGGG	CCCTTAGGTTCTTCACTTCCC	130	60
8946	TCACACCTTCTCTGTGGCAG	GCAAAACAAACCCCGAAAAC	127	60
8948	AAGGAAGGAGGTCTTCCAGC	CAACTTTGCAGGTACCACAGG	118	60
9054	CTTAACCTGATCACGGGGC	AACAGTGGCCTCAACTCCTC	148	60
9055	CCTTCTAGGCTCTGAAAATTGG	TGATGAATTCGAGTTTGCTG	130	60
9056	GCCCACCCTCATCTACCTG	CGACGAGGGATTGTCCTG	136	60
9057	GTCTGCAACACACTGCAACC	TGGTTTCGTGCCTGTAGTAGG	154	60
9058	TCTCTCAACTCACACGCCC	GAAACGGTGACCGCCTG	133	60
9059	CTGGACATTACACAGGCTCG	CTGCAGAATCTCTGTGGACTG	133	60
9060	GAACATCAGCAGTTGCCTTC	TCCCATGGAGATGCACATAG	152	60
9061	CTCCCGGAAGTGGGAGG	AGAGGAGAGTTGCTGGAGCC	147	60
9062	TAGATCACAGGGTCGGAAGG	CAGGGCAAGGAAGGCAG	131	60
9063	CTACCTGCGGGTCCTGG	GCCCTGAAGCAGTCCCTC	141	60
9064	TTGACTAACCTACGGCCACG	TCTTAAGGAGAAGGGGAGGG	153	60
9065	CCTGATGCTTGACTCTCCTC	GACTTTCCCTCTGCACCAAG	125	60
9066	CTCGGATTGCTCAAGGACC	ACTTTACGCTGGCAGGAGG	136	60
9067	TCACAGAGCAGCTCCCATC	GGCCACCAGGGAGATACAG	130	60
9068	GATTGGCCAGAACCCTGAG	GCGCTGCATGGTCATATTTC	125	60
9069	TGTAGCAAATGGTGGCAAAG	GAGACTATTGGCGATGCGAG	146	60
9070	TCTCCTTGTTAATCTGAGCTCTTG	GGTTATGATGTCCATCCAGG	126	60
9073	GGAAACATCACCAACAGCC	TTCAAACAAGGCAGAATACAGAC	133	60
9074	TGGAGATAGATGCCATTTGC	AGGGATCGATGTGCTAGGAG	133	60
9076	TCAGTGGAGAAGTGTGTGTGG	AATCCAAGACCCAGAAAGCC	141	60
9077	TGGGAGTTCACCTTTCTTGC	ACTTGTTGTAGCTGCCCAGG	120	60
9078	TCAACCTGAACCATTACGCC	GGTGACCAGGGTGGTGTAG	140	60
9648	AATAGCAGCTCTAAATCCCAGTC	CAGGAAGACGGGAACACAAC	148	60
9649	TGTGTTTGTCACAGCCTTGC	CTAGAGATGGACCTCCCACG	141	60
9650	GTTGAAGTACGGCCGCTG	CCTGAGAATGACTCCATGCTTC	149	60
9651	GGACTCAGTGATCAAACCCC	ATTGTTCTCAGAATGCACCC	130	60
9652	CACGATAAATTGTTTCTCTGTCC	GCGTCTCCTTGGTATGGATG	109	60
9653	ATTGCAACCAGAAGCTGGAG	TCCAGGTAAAGCAGGTCCTC	137	60
9655	GAAAGATCTCTGGGTGTCCG	AATCAGTGATGGCCATGGAG	160	60
9656	GTTCACTGCAGAGGAGGTTG	AATCTTGCCTGTTCTTTAGCG	143	60
9657	CCCCTTTGGATTAAGTTGC	ACTTCAGGGTCTGGGATGTC	131	60
9820	CAAGATGTGTCCAGACATTCCC	CATGGTGGCGTATGAACTTGG	162	60
9821	CAACCTTTTAAAACGCTTTGC	TATAGCTGCACGAATCTCCG	119	60
9822	GGTTCTTGAGTGATTAGTGTAGG	CCTCAGTTGCGAATATAAAAAGGC	179	60
MBD5 breakpoint determination:
9265	TGCAAAACTGTGTTTATGATGTTAC	AAGTCTGTGTTAAAAGAAATTTGGTG	131	60
9266	GCTATCCCTTTAAAACTCTCACAGC	TGCTCTTTGGGAGCAATAGG	111	60
9267	TGACTTTGCTTTCCAATAACCC	CAAAACAATAACTGAAGAACTCATGG	113	60
9268	TTTCACACGTGAACCAGGC	GACCTGGGTAGAGCAGCATC	121	60
9269	TTGCATATTTTATAGGCATAGATAGC	TCCCTACTCAACACTAAAGAGCC	134	60
9270	CAGGTTGTCATCACCACTGC	GAATGACCGCACTGACAGC	133	60
9271	AAAAGGTAGGGGTTGATGCC	AATATTTGTCAAAGGTGGCCC	140	60
9272	ACAAGGGGATAACCACAGGG	TTGTCAAGAAAGCTGGGAGG	124	60
9273	CAGCCAAGCCTTAGTATAGCC	TACCTATGGCACCCTTTCCC	124	60
9274	TCCTTGTCTTGCTTTGGTTG	GACTCTACAGGTTATTTGGAGGC	127	60
9348	TCAGGGAAATGTAGGATTGC	TCTCTTCGTCTCCCTCCCTC	133	60
9349	TCATGTGTCTGATGCCTGTTTAC	TGCTGAGAACAAGGATGGC	207	60
9350	AATTTTCTCTTTCTTGACATAAATCTG	GGAAACATGCAGAACTGCTC	132	60
9351	TGCTTAAGCCACAAATCTGAAG	TTGTATGTGCACTGGAACTTG	252	60
9352	TTTCATACACAGAGGAGGAAGG	ACATGGAGGAGATGCCGTC	136	60
9353	CTAACTTTGGGAGGTGCGTC	AGGATTCATTTCCAACAGCG	181	60
9354	AGGTTAAGGCTCCCCTCCC	GTGTCTTCAACTTCTCGGGC	142	60
9355	TTGCCAGTTGAAACATTCTACC	ACAGAAGATTTGCTCCTCGC	264	60
MBD5 mutation screening:
9275	TCATCTTATTGCTGATATCTTTGGAG	TTGCAGGTACCACAGGTAATAATAAG	193	60
9276	AAAATGCTTTTCCCTAGTGGG	GAAAAGGTAGAAAGGTGGTTTTAATG	482	60
9277	TTTTACAGACATATTCTAAACAAAGGC	GGAGAGGAGAATTCCTGAACC	261	60
9278	TCCCTCCCACCACAAAAG	GTGGGTCAGTCCTTGGAGAG	395	60
9279	GAAATCTCCATTCCGTGGC	AACATTGCTCGTGGTATTTCC	354	60
9280	TCTTTCCCCAACCTTGACTAC	GCAATGGGAGATTACTTGGC	335	60
9281	TTAATCCAACCAGTTTCCATTC	TTGGGGACCCTATTGTTGAC	375	60
9282	ATGATGCCACCTGTAGGACC	GATTTGCTAGCTGTGCTTTGG	355	60
9283	TGGGAATGCCTTTAAATCAG	TCCATTTGAGACTGTCTGAGC	355	60
9284	AGACGCATTGCGGAAAAG	GTGTTTGCATTGTGGCAGTG	339	60
9285	GCCTCAAATACTGCTTTGCC	GAAGAGAATTTCACAATGGGG	467	60
9286	TCATGATTAATAACTGGGTTTTGTG	AGTGCTGTTGGATGGAAAATG	243	60
9287	ACCCATTGGGAGTGATTTTC	AGGGTGGAGGTTGATCTCAG	223	60
9288	TCTCGGGTACAAAGAGAGGC	TTGTTTTCTTCTAAAATGACACAGC	320	60
9289	GAGGCCTCAAAATTATTTCCC	AATGGGGCTGATCTGAGTTG	351	60
9290	TTACAAAGCAGTTGTCGATGC	TGTGGCAACTTGCTGACTTG	311	60
9291	GAGGAATTCAAGAGGGGCTC	CATCTTTCCACAGTGTTCTCTAGTC	387	60
9292	CCACCAAGAAACTGTCCAGG	ATCTGAAGGGCTAGGCACAC	315	60
9293	GGAGCAGTCTCCAAGTTCC	GAACAAAGGTTAGTATTTGACAATGG	361	60
9294	TCTGGGTAATGTGGTTTGGTC	TCAGTGAGATTTCATTGTCCC	197	60
9295	TGGAATTGGTACTTTTGTTTTCTG	TTTGACAATACTCAAGAGACTTTACC	258	60
9296	TGTGATTTCATCCTCTGTTGTTG	CACTGCGCATTAGTGGAGTAG	151	60
9658	TCCCCAACCTTGACTACAAAG	AGAGCACAAGAAGGTGGAGG	140	60
9659	TCCACTTGGCATTCTTGACC	CTTGGCAAAGGAACAACAGC	147	60
MBD5 RT-PCR and RACE primer:
9432	…	GACATTCCAAGCCACACTTGC	…	64
9433	…	CATGTTCCATCAGTAAGCAGG	…	62
9434	…	ACACGACGCTGCCAACCCAC	…	66
9468 C1F	AGAGGTACTCCCTTATAGGG	…	…	60
9468 C2F	GGACTCGTAAAGACATAGAGC	…	…	60
9469 C1F	TACTCCCTTATAGGGACTCG	…	…	60
9469 C2F	GGGACTCGTAAAGACATAGAG	…	…	60
9470 C1F	GAAATCAAGAAGAGCACACAC	…	…	60
9470 C1F	ACACTATTTTCCTTCATCAATCC	…	…	60
9725	ATGTCAGTTTCTACATGTGGG	…	…	60
9726	ATGTCAGTTTCTACATGTGGG	…	…	60
9727	GGGAAGTAGACATTGAAGGC	…	…	60
9728	GCCACCCGTCAGAGAGGGAC	…	…	60
9729	GTCAGAGAGGGACATGCGC	…	…	60
10028	…	GTTGACATCCTCTGTTGCCA	…	60
X inactivation:
8774	TCCAGAATCTGTTCCAGAGCGTGC	GCTGTGAAGGTTGCTGTTCCTCAT	287	58