METHODS

DNA samples

Initial assay optimisation was undertaken on the following eight lymphoblastoid cell lines: GM07175 (normal female with 30 and 23 CGG repeat alleles), GM06907 (premutation female with 95–140 and 23 CGG repeat alleles), GM06896 (premutation female with 85 and 29 CGG repeat alleles), GM07537 (full mutation female with >200 and 29 CGG repeat alleles), GM06890 (normal male with 30 repeat allele), GM06891 (premutation male with 117 CGG repeat allele), GM06892 (premutation male with 80–85 CGG repeat allele), and GM06852 (full mutation male with >200 repeat allele), which were obtained from the Coriell mutant cell repository in Camden, New Jersey, USA. Assay validation was done on 81 previously genotyped archival patient DNAs and normal DNAs from the DNA diagnostic laboratories of Johns Hopkins Hospital, KK Women’s and Children’s Hospital, and National University Hospital (Singapore). The study protocol was approved by the institutional review board of the National University Hospital.

Sodium bisulphite treatment

Methylated and non-methylated FMR1 alleles were differentially modified in the presence of sodium bisulphite according to Clark et al (1994),¹¹ but with modifications. Briefly, NaOH was added to 1 μg DNA to a final concentration of 0.3 M in a final volume of 5 μl. After a 15 minute incubation at 55°C, 75 μl of fresh and prewarmed sodium bisulphite solution was added. The latter was prepared by dissolving 0.6 g of sodium bisulphite in 45 μl of 10 M NaOH and 960 μl of dH₂O. The deamination mixture was incubated at 55°C for five hours, then purified over a GFX™ column (Amersham Biosciences) and eluted in 50 μl Tris-EDTA, pH 8. The purified deaminated DNA was then desulphonated by the addition of 50 μl of 0.2 M NaOH and incubation at 37°C for 15 minutes. After neutralisation with 50 μl of 0.2 M Tricine, the mixture was purified over a second GFX™ column, eluted in 50 μl Tris-EDTA, pH 8, and stored at −20°C until use. The sodium bisulphite treatment results in methylated and non-methylated FMR1 alleles with distinct differences in nucleotide sequence (fig 1A).

Methylation specific PCR

Three sets of primers were designed to amplify from the antisense strand of bisulphite modified DNA, one set targeting the non-methylated allele, and the other two sets targeting the methylated allele (table 1; fig 1B).

Specific amplification and sizing of the non-methylated FMR1 repeat, designated as “non-Met-PCR”, was accomplished with primers non-Met-F and non-Met-R. Amplification was carried out in a 50 μl volume containing 0.2 μM of each primer, 0.2 mM dNTPs, 2.5 U HotStarTaq™ DNA polymerase (Qiagen), 0.5×Q solution (Qiagen), 1×supplied PCR buffer (including 1.5 mM MgCl₂), and 7 μl (calculated at ~100 ng) of bisulphite modified DNA. An initial denaturation at 95°C for 15 minutes was followed by 40 cycles of 98°C for one minute, 62°C for one minute, and 72°C for two minutes, followed by a final extension at 72°C for 10 minutes. This reaction is expected to detect all non-methylated normal and premutation FMR1 repeats (fig 1C, upper panel).

Two different PCR reactions were carried out to detect the methylated allele. The first reaction, designated as “Met-PCR”, used primers Met-F and Met-R to amplify across the methylated FMR1 repeat. Reaction conditions were similar to that for the non-Met-PCR, except that 1.5×Q solution was used instead. This reaction is expected to determine all normal and premutation methylated allele sizes, as well as full mutation alleles up to~350 repeats (fig 1C, middle panel).

The second methylated allele PCR reaction, designated “mTP-PCR”, is an adaptation of the triplet primed PCR (TP-PCR) strategy first described by Warner et al.¹² In mTP-PCR, primers mTP-F, mTP-R, and Tail-R were used to amplify from within the methylated repeat (table 1; fig 1B). Reaction conditions were similar to that for Met-PCR, except that each reaction contained 0.2 μM each of primers mTP-F and Tail-R, 0.02 μM of primer mTP-R, and 14 μl (~200 ng) of the bisulphite modified DNA. In the presence of either a methylated premutation or full mutation allele, an mTP-PCR product smear extending upwards beyond 300 base pairs (bp) is expected (fig 1C, lower panel). This product is referred to as a pre/full mutation (PFM) smear. In normal female samples, smearing below 300 bp (NL smear) is expected and represents product from the normal FMR1 allele on an inactive X chromosome.

A 15 μl aliquot of each non-Met-PCR, Met-PCR, and mTP-PCR product was analysed by electrophoresis through a 1.5% agarose gel at 6 V/cm, stained with ethidium bromide for 30 minutes, and photographed over an ultraviolet transilluminator. The triple PCR assay will generate distinct banding and smear patterns depending on FMR1 (CGG)_n genotype, and anticipated results for various genotypes are illustrated schematically in fig 1C. Throughout the results and discussion sections of this paper, NL, PM, and FM are used to indicate normal, premutation, and full mutation, respectively.

RESULTS

Initial assay optimisation studies were undertaken using DNA obtained from four male and four female cell lines from the Coriell cell repository. For each cell line, the products of the non-Met-PCR, Met-PCR, and mTP-PCR were easily visualised by ethidium bromide staining (fig 2A), and the FMR1 (CGG)_n genotype classifications obtained were consistent with their known fragile X status (table 2). For example, in NL male 1 (cell line GM06890), the non-Met-PCR reaction produced the expected single non-methylated NL allele, while the Met-PCR and mTP-PCR reactions were negative (fig 2A). The assay easily detected the non-methylated PM alleles from both PM male 1 (GM06891) and male 2 (GM06892), as well as the methylated FM allele from the FM male 1 (GM06852). Using mTP-PCR analysis, a pre/full mutation (PFM) smear was observed only from FM male 1 (GM06852) among the male cell lines. A PFM smear is defined as a smear of DNA amplification product that extends upwards beyond 300 bp and is characteristic of all full mutation alleles and most methylated premutation alleles.

The assay was also able to classify accurately the FMR1 (CGG)_n genotypes of the female cell lines. For example, NL female 1 (GM07175) showed an NL non-methylated allele and an NL methylated allele after non-Met-PCR and Met-PCR analyses, respectively, and a short smear not exceeding 300 bp (NL smear) was observed by mTP-PCR analysis. The NL and PM alleles in PM female 1 (GM06907) and female 2 (GM06896) were also easily detected using this triple PCR assay, as were the NL and FM alleles in FM female 1 (GM07537). Interestingly, the results also revealed evidence of non-random or skewed X inactivation in these cell lines, with the majority of cells carrying the NL allele on the active X and the PM and FM alleles on the inactive X. We confirmed the skewed X inactivation in these samples by Southern blot analysis (data not shown), and have quantified the degree of skewing in all four female cell lines using the HUMARA assay^13,14 (table 2). As expected for the PM female and FM female cell lines, mTP-PCR analysis detected PFM smears in all three samples.

For each of the eight cell lines, the CGG repeats of the non-methylated and methylated alleles were calculated from their respective non-Met-PCR and Met-PCR fragments according to the formula in table 1. The calculated CGG repeats were concordant between the two PCRs and were in good agreement with the approximate CGG repeat sizes inferred from the Southern blot results (table 2). PCR products of FM alleles, however, could not be accurately sized beyond 1 kb on this gel system. Such alleles were labelled as having >295 CGG repeats, which is the literal conversion from a 1 kb methylated fragment rounded off to the nearest five CGG repeats.

We further evaluated the assay on peripheral blood leucocyte DNAs of known FMR1 (CGG)_n genotype classification. These samples were selected to represent the widest spectrum of NL, PM, and FM FMR1 genotypes that we have previously analysed. The triple PCR assay of non-Met-PCR, Met-PCR, and mTP-PCR reactions was able to classify the FMR1 (CGG)_n genotype of each male and female sample accurately (table 3). All non-methylated and methylated NL and PM alleles in males and females could be clearly detected (fig 2B). All samples carrying methylated PM or FM alleles generated a PFM smear by mTP-PCR, while NL males and PM males did not produce a smear, and NL females displayed only an NL smear.

This evaluation allowed us to delineate the upper limit of detection of FM alleles by Met-PCR analysis at ~350 repeats. Taking FM female 4 (FX0025) and FM male 3 (FX0012) as examples, the Met-PCR reaction failed to generate an amplification product from either sample (fig 2B, middle panel), both of whose FM alleles exceed 500 repeats as determined by Southern analysis (table 3). The mTP-PCR reaction, however, produced a PFM smear in both samples, thus ensuring that the FM alleles in both samples were not missed. Although a PFM smear is also generated in the presence of a methylated PM allele, Met-PCR analysis did not detect any methylated PM allele in FM female 4 and FM male 3. Based on our observation that the non-Met-PCR and Met-PCR reactions can reliably detect even the largest PM alleles among our pre-characterised samples, the negative Met-PCR results therefore indicate that the PFM smears of FM female 4 and FM male 3 must have been generated from FM alleles. The combined PCR results thus correctly classified both samples as carrying FM alleles.

To evaluate the specificity of the mTP-PCR reaction in producing a PFM smear only from methylated PM and FM alleles, we carried out mTP-PCR on genomic DNAs from additional NL, PM, and FM males and females. All “normal” samples were first confirmed to contain only an NL FMR1 (CGG)_n allele (for the males), or two NL alleles (for the females) by direct amplification of the CGG repeat¹⁵ and fluorescence detection on an ABI 3100 genetic analyser. The mTP-PCR reaction did not produce any DNA smear in all NL male samples tested (fig 3A), whereas only the short NL smear was evident in all NL female samples (fig 3B). In contrast, the PFM smears generated from all FM males (fig 3C) and all PM and FM females (fig 3D) were clearly more extensive. Although it is conceivable that the NL smear from a large NL allele (for example, 50 CGGs) would be quite similar to a PFM smear from a small PM allele (say 60 CGGs), the non-Met-PCR and/or Met-PCR results will distinguish between most NL and PM alleles by size, confirmed by sequencing or pedigree analysis if necessary. For those samples where the allele sizes fall within the 50–60 CGGs range, further pedigree or sequencing analysis of the CGG repeat structure is necessary for proper mutation classification.

An algorithm for the classification of the various female and male FMR1 (CGG)_n genotypes is shown in fig 4, and is intended to simplify interpretation of the agarose gel results obtained from this methylation specific triple PCR assay.

DISCUSSION

More than a decade after the identification of a trinucleotide repeat expansion in the FMR1 gene as the predominant cause of fragile X mental retardation syndrome, molecular diagnosis of this disorder continues to rely mainly on a combination of PCR and Southern blot analyses. Southern blot analysis is very robust in detecting FM expansions in both males and females. However, it is highly labour intensive and time consuming. Additionally, Southern blots are not ideal for detecting PM alleles at the low end of the size range. PCR amplification across the CGG repeat enables precise sizing, but only of NL and small PM alleles. Larger PM and FM alleles cannot be detected reliably using conventional PCR. Furthermore, PCR does not provide information on the methylation status of the alleles.

More recently, several assays based on ms-PCR have been developed in attempts to address the deficiencies of PCR based methods.^9,10,16 Ms-PCR methods rely on the differential modification of methylated and non-methylated genomic DNA by sodium bisulphite, allowing subsequent specific amplification and detection of methylated and non-methylated alleles.¹¹ This procedure converts the non-methylated FMR1 trinucleotide repeat from 100% G:C base pairing (5′-CCG-3′ on the antisense strand) to one with 33% G:C base pairing (5′-TTG-3′ repeat), while the methylated repeat now has ~67% G:C base pairing (5′-TCG-3′ repeat) (fig 1). Thus this conversion not only allows non-methylated alleles to be discriminated from methylated alleles, but also reduces the difficulty of amplifying across the repeats. Even so, larger PM and FM alleles remain refractory to amplification and detection, and are especially difficult to detect in females owing to preferential amplification of their NL alleles.

Weinhausel and Haas (2001) developed a dual PCR methylation specific PCR assay for detection of fragile X syndrome in both males and females.¹⁰FMR1 genotypes were identified on the basis of several parameters, including a duplex-PCR of the FMR1 and XIST gene promoters coupled with a duplex PCR across the non-methylated and methylated CGG repeats, followed by densitometric ratio analysis of FMR1 promoter methylation status normalised against that of the XIST promoter. These investigators reported that virtually all categories of FMR1 genotypes could be detected, although noting that very large PM alleles in female carriers could be missed, thus resulting in occasional difficulties in distinguishing a PM female from an FM female.

Our assay differs from the method of Weinhausel and Haas¹⁰ in two main respects. First, our assay involves three single PCR amplifications instead of two duplex PCR reactions. We chose not to duplex the non-Met-PCR with the Met-PCR after initial trials indicated a reduced detection sensitivity, especially of the large PM and the FM alleles (data not shown). Separate agarose gel analysis of the non-Met-PCR and Met-PCR reactions allows easy identification of non-methylated and methylated alleles and approximate sizing of NL, PM, and FM bands. With a duplex PCR format, agarose gel analysis of the combined non-methylated and methylated PCR products is less straightforward, especially for certain NL and PM female genotypes, owing to the possibility of overlapping non-methylated and methylated fragment sizes or co-migrating non-methylated and methylated fragments of the same size.

Second, our assay does not rely on densitometry and ratio analysis of non-methylated versus methylated promoter PCR fragments to detect FM alleles. We instead used a protocol modified from the triplet primed PCR (TP-PCR) method of Warner et al,¹² in order to detect affected males and females with very large FM alleles that could not be amplified successfully using Met-PCR. Our mTP-PCR reaction reliably and specifically detects all PM and FM alleles. Although the mTP-PCR PFM smear generated from a PM female is indistinguishable from that of an FM female, the two genotype classes can be distinguished on the basis of their non-Met-PCR and Met-PCR results. A PM female will display a band in the PM size range in both the non-Met-PCR and Met-PCR gels, whereas an FM female with a large expansion of >350 CGG repeats will not display a band in either the PM or the FM size ranges using either non-Met-PCR or Met-PCR; and an FM female with a smaller expansion of ⩽350 will display an FM size band in the Met-PCR gel only. The presence of a methylated smear extending beyond 300 bp is thus diagnostic of all FM males, PM females, and FM females. The PFM smear is not observed in NL males or PM males, while NL females have their own characteristic ethidium bromide staining patterns (fig 3).

In the majority of samples tested, the sizes of the non-methylated and methylated alleles, determined from the non-Met-PCR and Met-PCR analyses, respectively, were concordant and were in good agreement with the approximate CGG repeat sizes inferred from the previous Southern blot or direct PCR results (tables 2 and 3). In PM female 5 (FX0018), however, the non-methylated and methylated PM amplicons migrated as apparently different sizes (table 3). We postulate that the observed CGG repeat discrepancy between the non-Met-PCR and Met-PCR products in PM female 5 most probably reflects the combined effect of her very large PM allele (>140 CGGs) coupled with the difference in nucleotide sequence composition between the modified non-methylated and methylated alleles. These two factors result either in anomalous retarded migration of the bisulphite modified non-methylated PM amplicon compared with its methylated counterpart, or conversely in anomalous accelerated migration of the bisulphite modified methylated PM amplicon compared with its non-methylated counterpart. Whether the methylated or non-methylated PM amplicon represents the true CGG repeat length has not been determined, owing to the difficulty in sequencing across such long CGG repeat stretches. The discordant sizes of these samples should not, however, lead to misdiagnosis or misclassification of their FMR1 genotypes.

It is possible that this assay may misclassify a female mosaic for PM and FM alleles as a constitutive PM female when only the PM allele but not the FM allele is detected by Met-PCR. Such situations may, however, be rare as many mosaic females carry multiple FM alleles including some of less that 350 CGG repeats. If this is the case, the Met-PCR reaction may be able to amplify and detect the smaller FM alleles.

It should be noted that even the gold standard Southern blot analysis is not completely error-proof. For example, some FM affected females harbour highly somatically unstable FM alleles, which sometimes makes them difficult to detect. Detection of such alleles by Southern analysis is obvious only under optimal probe labelling, hybridisation, and wash conditions that maximise signal and minimise background noise. Thus the triple ms-PCR method for fragile X testing demonstrated here provides a suitable alternative to Southern blot analysis which is less time consuming and more amenable to the clinical testing environment.