METHODS

DNA samples

DNA from a cohort of unrelated subjects, mostly Nottinghamshire and Derbyshire residents, and anonymous males selected from those referred to the Nottingham Clinical Molecular Genetics Laboratory for Fragile X DNA testing, was prepared using standard techniques. DNA samples with known subtelomeric imbalance were obtained from several sources: from University of Nijmegen Department of Medical Genetics at Nijmegen, The Netherlands (DMGN), as part of their clinical screening programme, from the Coriell Cell Repository (CCR) at Camden, New Jersey, USA, from the European Collection of Animal Cell Cultures (ECACC) at Salisbury, Wiltshire, UK, and from Sheffield Children’s Hospital, Sheffield, UK. All samples were collected under the appropriate ethical guidelines and ethics committee approval of each country.

MAPH probes

Probes for subtelomeric MAPH were prepared as previously described¹² by subcloning fragments of PACs or cosmids mapped to subtelomeric regions, or direct cloning of specific amplicons. After cloning, each probe was sequenced using the BigDye terminator cycle sequencing kit and an ABI 377 Genetic Analyser, and any sequences not represented in the Genbank database were submitted (accession numbers AY061937-AY061940, see laboratory website for all probe sequences, http://www.nottingham.ac.uk/~pdzjala/maph/maph.html). Maximum distances of each probe from the telomere were estimated by noting the size of the half YACs from which each probe was ultimately derived,⁹ and linking the half YAC sequence to a physical map which included the BAC containing the probe sequence, as identified by BLAST. For nine chromosome ends (1p, 2p, 9q, 7p, 15q, 16q, 18p, 18q, and 19q), the distance between the probe and the telomere is represented by partially sequenced BACs, which may lead to an incomplete genome assembly and hence inaccurate estimates of distance.

MAPH

MAPH was carried out using 1 μg genomic DNA, with fluorescent detection of amplification products using ABI Prism technology with an ABI 377 DNA analyser, or with Amersham Pharmacia ALFexpress technology. Hybridisation and generation of a first round amplification product were performed as described previously.¹²

The first round amplification product was used to seed a further 20 μl amplification using 5‘ FAM labelled PZA primer or 5‘ Cy5 labelled PZA primer for detection by the ABI 377 and Pharmacia ALF respectively. After ethanol precipitation (for the ABI machine only) and addition of a formamide based loading buffer, amplification products were separated using PAGE LongRanger gels on the ABI 377, and bands quantified using GeneScan™. An arbitrary threshold of 1000 relative fluorescent units was set, so that if the average of the top two and bottom two bands was above that threshold, the result for that sample was deemed technically satisfactory. On the Amersham Pharmacia ALFexpress equipment, the amplification products were run on a ReproGel High resolution gel, and bands quantified using the packaged software.

Data processing and univariate analysis

Normalised ratios of signal strength, setting diploid dosage at 1.0, were calculated as previously described.¹² Briefly, each band intensity was quantified by measuring the area under the peak, and that value was divided by the sum of the four nearest band intensities within that lane. This value represents band intensity relative to neighbouring bands, and, in the absence of copy number changes, was reproducible across samples. Dividing each relative band intensity value by the mean for that probe allowed a normalised ratio to be calculated with a mean of 1.0, equivalent to a dosage of two copies per diploid cell. Calculation of the standard deviation of the normalised ratios allowed each sample/probe combination to be represented by a Z value, corresponding to the number of standard deviations away from the mean of 1.0.

Bivariate analysis

For observations in duplicate (x,y), the correlation coeffiecient r_xy , was calculated and a test statistic T(x,y) calculated for each value and compared with C_α , a value, unique to the data set from each probe, which describes the confidence ellipse of that data at a significance level of α.¹⁸ Details on calculating T(x,y) and C_α are available from the laboratory website. Where performed, all corrections to significance values to allow for multiple tests used the Dunn-idák method: α’ = 1 − ( 1 – α )^k where α is the p value for a given result, and α’ is the p value for the given result, corrected for k tests.

All sample/probe combinations giving significant deviations using this analysis were examined in detail. Firstly, the gel tracking was checked and the peak allocation value was also checked to confirm that the software had recognised the equivalent peak to other lanes. The value was renormalised against a different set of bands within the same lane to ensure that variation in a neighbouring band, usually used for normalisation, was not causing the positive result. Finally, if the chromosome end had two representative probes in the probe set, the results from these were compared. For the analysis of patient samples, sample/probe combinations reported as retests were screened as above as well.

Dosage PCR

Fluorescent multiplex dosage PCR primers were designed to amplify the same subtelomeric segments that hybridise with the MAPH probes. One primer of each pair was labelled with 6-FAM fluorescent dye; 20 ng of genomic DNA was amplified in 20 μl reactions with 10 pmol of each of the appropriate primers. The cycling conditions were 95°C for one minute, 56°C for one minute, and 72°C for one minute, for 25 cycles using a Perkin Elmer 480 Thermal Cycler. A total of 1 μl of the amplification was loaded on an ABI 377 DNA sequencer with formamide based loading buffer and ROX500 size marker, and the bands quantified using GeneScan™ software. The sequences of all primers used are available from the laboratory website (http://www.nottingham.ac.uk/~pdzjala/maph/maph.html). Normalised band areas which differed from the mean with a one tailed significance of greater than 90% (p<0.1) were reported as positive.

RESULTS

We have applied multiplex amplifiable probe hybridisation (MAPH) to analyse DNA for subtelomeric copy number changes in one gel lane per sample. MAPH is a method that detects copy number changes by hybridising specific probes corresponding to the sequence of interest to immobilised genomic DNA, followed by quantitative recovery and amplification of these probes.¹² All probes have shared primer binding sites for amplification, but the different sizes (between 100 and 600 bp) allow each probe, and hence each locus, to be resolved, identified, and quantified on a polyacrylamide gel. Normalisation of a probe of interest against probes with an expected dosage of two (in a diploid cell) allows relative dosage to be determined.

A probe set consisting of 47 probes was constructed with all chromosome ends represented by at least one probe, except the p ends of acrocentric chromosomes. Six ends are covered by more than one probe: XpYp, because each probe was known to detect independent but infrequent polymorphic deletions (unpublished work), 22q 19p, 12q, 19q, and 1p. Each probe was initially characterised on a small cohort of normal subjects using MAPH with radioisotopic detection, and sequenced before being incorporated in the final probe set. A table of the properties of each probe in the set is available at the laboratory website.

The probable distances of the probes from their respective telomeres was deduced by previous physical mapping (Human telomere mapping and sequencing project, University of Pennsylvania, http://www.wistar.upenn.edu/Riethman/, Joint Genome Institute chromosome 19 physical map, http://www.jgi.doe.gov), and from the human genome draft assembly (University of California, Santa Cruz Working Draft Human Genome Assembly http://genome.cse.ucsc.edu). Inconsistencies between the physical map and the assembly, and indeed between different assemblies, indicate that the human genome working draft contains extensive misassemblies at the ends of chromosomes. As the physical mapping and sequencing of the telomeres proceeds, replacement of these probes with sequences of known physical distance will be a priority for construction of the next subtelomeric probe set.

To characterise the probe set, DNA from 83 normal subjects was analysed in duplicate using the ABI 377 DNA Analyser. We tested each sample in duplicate to reduce the number of false positives and false negatives, and hence give more statistical power to the results. Our initial approach for analysing the data was to average the two normalised ratios for each sample/probe combination and plot these values as a histogram (fig 1). The mean and standard deviation of the distribution shown in the histogram were calculated, and the expected normal distribution using these calculated parameters is also shown in fig 1. It is clear that the data fit the normal curve reasonably well, but there is a slight skew resulting in conservative estimates of significance for deletions but more accurate estimates of significance for duplications. Several outliers were identified (fig 1), and dosage PCR confirmed the presence of an XqYq duplication (p=5.4 × 10^-6) and a 1q duplication (p=3.2 × 10^-9). The XpYp deletion is one of a range of XpYp polymorphisms found using MAPH (data not shown) and is only detected with one XpYp probe, ST17H1C, so was not confirmed by dosage PCR. Dosage PCR was used to verify these changes in preference to FISH as copy number changes involving small (<10 kb) segments, such as found in alleles at XpYp (unpublished work), would not be detected by large FISH probes.

Control samples with known subtelomeric copy number changes were used to verify the dosage response of 14 probes (table 1). These positive controls had been phenotyped by FISH using the subtelomeric probe set available (Cytocell), except the DNA from CCR and ECACC where only G banded karyotypes were available. Table 1 shows the probes for which positive controls were available, together with the p values of each observation. Fig 2 shows analysis of an unbalanced 3p−;16q+ translocation carrier as a positive control, together with a normal negative control.

The analysis method described above is only suitable when the probe/sample values are independent of each other every time they are tested in independent experiments. To determine whether duplicate values for a specific sample were independent of each other, the correlation statistic r was calculated for each probe from the 83 pairs of observations of the normal subjects. This statistic was significantly different from 0 (H₀, no correlation, p<0.05) for 34 of the 47 probes. Because duplicate results are not completely independent, hypothesis testing using the average of two values, as shown in fig 1, is not a valid method for further data analysis.

We therefore applied a bivariate statistical analysis approach to the data from the 83 normal subjects to correct for this observed correlation effect. The duplicate results for each probe were plotted on a scatter graph, and an ellipse calculated for a specific confidence limit around (1.0,1.0) (for the normal values), (0.5,0.5) (for any deleted values), and (1.5,1.5) (for any duplicated values). If the value was outside all ellipses, or in the intersection of two ellipses, it was reported as a “retest”. If the value is inside the duplication or deletion ellipse only, then it is reported as a “positive”. If the value is within the normal ellipse only, it is reported as “normal”. Further statistical methods are available from the laboratory web site. An appropriate confidence limit for the normal ellipse (which is equivalent to the false positive rate) was calculated by iteration to minimise the number of retests reported by the analysis. This value was found to be 1.7 × 10^-5 per probe per sample, equivalent to 0.0008 (0.08%) when corrected for multiple probes, and resulted in a total of 6.37% of observations being reported as positives or retests. Any “positive” samples were then tested by dosage PCR. A false negative rate was set at 1% for deletions and 1% for duplications. The outliers identified in the univariate analysis were classed as retests by the bivariate analysis.

We used the methods developed above to screen patients with learning disability to examine whether MAPH could detect copy number changes in subjects whose subtelomeric DNA would not normally be analysed. Thirty-seven DNA samples from patients referred to the Nottingham Clinical Molecular Genetics Laboratory for diagnostic testing for the fragile X syndrome, but who had no FRAXA repeat expansion, were tested in duplicate using the ABI 377 and in duplicate by ALFexpress technology. The results were assembled to give 74 results in duplicate. These were then analysed using the same false negative and false positive parameters used for the analysis of the 83 normal controls, with the results classified into normal, retest, and positive results. If one analysis was reported as a positive, and the other as a retest, the sample was classified as a positive. Two samples had very high results for the 9q probe owing to a gel artefact, which caused all samples to be reported as retests. Therefore, the 9q probe produced no reliable data and was excluded from further analysis. All positives and retests were then tested by dosage PCR. To conform to local ethical guidelines, these DNA samples were collected and tested anonymously, making it impossible to collect cells for cytogenetic diagnosis or to identify the proband’s parents.

There were 65 sample/probe combinations (out of a possible 1480) that were classed as a retest in both the ALF and ABI data sets, and four loci that were classed as a positive in the ALF/ABI and a retest in the other. All “positive” results and retests were tested by dosage PCR (table 2) and six (three deletions and three duplications) were confirmed as true positives. The corresponding overall false positive rate was 3.99%. Among the confirmed positives, the 7q− and 16q− deletions not only gave the lowest p values in both MAPH tests, but were also identified as a genuine positive instead of a retest in one of those MAPH experiments. Terminal 7q deletion has been reported to be associated with holoprosencephaly 3, which has learning disability as a symptom (MIM 142945).¹⁹ This was later shown to be the result of deletion of the SHH gene,²⁰ but we have not determined the extent of the 7q deletion in this person. Similarly, 4p terminal deletions (as identified in sample A16) have been associated with the closely related Pitt and Wolf-Hirschhorn syndromes (MIM 262350, 194190), both conditions that have learning disability as a clinical phenotype. The 16q− deletion in sample B1 is accompanied by 3p duplication, suggesting that this may be an unbalanced translocation. However, for all of these observations, only FISH analysis could act as definitive confirmation of these subtelomeric changes.

DISCUSSION

We have shown that a new technique, MAPH, can be used to identify copy number changes in subtelomeric DNA. It has several advantages over FISH based methods, in that it does not require metaphase chromosomes, can analyse all unique chromosome ends simultaneously, and one polyacrylamide gel can analyse 45 samples. These combine to give the technique a very high throughput with a much lower cost per sample compared to FISH. In addition, repeat testing is straightforward and strengthens statistical analysis of the results. MAPH compares favourably with other techniques used for subtelomeric DNA screening. Inheritance at short tandem repeat polymorphisms (STRPs) has been used to detect copy number changes at subtelomeric regions,^21–23 but this technique depends on the informativeness of the STRPs used, needs DNA samples from parents, may miss duplications, and requires multiple PCRs to screen each karyotype. A recent paper applied array comparative genomic hybridisation (array-CGH) to screen for subtelomeric copy number changes.²⁴ Using the published data, we calculate that triplicate testing of each sample should result in a false positive rate of 4.6%, assuming independence between results. However, the authors routinely removed noisy data, so the unselected false positive rate may be closer to that using MAPH. Array-CGH also requires expensive equipment and consumables, an important fact when high throughput screening is proposed. Multiplex ligation dependent probe amplification (MLPA) is a method similar to MAPH that could be applied to detect subtelomeric DNA copy number changes, but there are no published data on subtelomeric screening.²⁵

We have applied MAPH to study a cohort of patients, and have identified six subtelomeric copy number changes in five patients that were confirmed by fluorescent dosage PCR (13.5% of the total number of patients screened). In a diagnostic setting, the technique is best suited as a screening method for the many samples for which current analyses are uneconomical, and using FISH analysis of reported samples as a “gold standard” of diagnosis. Because of the many loci that are analysed simultaneously, the fairly high false positive rates can cause a significant number of loci to be reported, especially if many different samples are analysed together. Nevertheless, the method can still be very useful in reducing the FISH workload by 96% and focusing on the 4% of loci that are likely to show a copy number change. The false negative rate is important for such a situation. By using appropriate confidence thresholds for calling deletions and duplications in bivariate or univariate analysis, we can adjust the expected false negative rate. In this study we have set this false negative rate (the probability of a genuine copy number change not being detected) at 2%. The impact of this value depends on the frequency of genuine copy number changes expected in the cohort that is analysed. The frequency of subtelomeric rearrangements in subjects with idiopathic learning disability has been reported to be around 6%,³ which would result in about 0.012% of a cohort of mentally retarded patients being wrongly reported as having no subtelomeric copy changes. This assumes that subtelomeric rearrangements are copy number changes; since MAPH can only detect net copy number changes it cannot detect balanced translocations or inversions where no net loss or gain of DNA has occurred. More recent studies find different frequencies of subtelomeric rearrangements in patients with learning disability. Using a combination of G band cytogenetics and subtelomeric FISH, Vorsanova et al²⁶ reported that 3.8% of patients with learning disability have subtelomeric rearrangements, although a more recent study suggests that the value may be as high as 9%.²⁷ Knight et al²⁸ emphasised the importance of patient selection when conducting studies of this type. By dividing patients into two classes, mild learning disability and moderate to severe learning disability, the frequency of subtelomeric rearrangements in the two groups was 0.5% and 7.4%, respectively. Another report has suggested that subtelomeric rearrangement is not a significant cause of learning disability, and may be found at polymorphic frequencies in the general population.²⁹ A recent review of published data³⁰ suggests that the initial study by Flint et al³ reporting an overall frequency of 6% may be about right. To obtain an accurate assessment of the frequency of subtelomeric rearrangements, a very large sample of patients selected using the same clinical criteria must be assayed. Using FISH alone, this would be extremely laborious and time consuming, but the ability of MAPH to act as a screen before FISH makes this approach practical. It is also likely that, by using MAPH, smaller subtelomeric copy number changes, previously undetected by cytogenetic methods, will be reported.

As the human genome project generates a more accurate assembly of sequence near the telomere of each chromosome, it will be possible to design single copy probes that are as near the telomere as possible. The position of these will be a compromise between proximity to the telomere and ability to recognise single copy sequences, since on most chromosome ends the subtelomeric DNA consists of multicopy sequences shared between chromosomes immediately after the telomeric repeat, followed by the gene rich single copy DNA.¹ A third generation set of probes will ideally have several probes at each chromosome end so an estimate of the extent and position of each copy number change could be determined.