Abstract
With the rapid increase in the number of identified human disease genes, the development of accurate and cost-efficient mutation tests has become opportune. Here we present a combination of extensive PCR multiplexing and two-dimensional (2-D) DNA electrophoresis to screen for mutations in 26 exons of the retinoblastoma ( RB1 ) tumor suppressor gene. In 2-D electrophoresis, fragments are separated according to size and base pair sequence in non-denaturing and denaturing gradient gels, respectively. All target fragments, designed to have optimal melting characteristics, were prepared in a two-step PCR (a 6-plex long-PCR pre-amplification and a subsequent 25-plex short-PCR) followed by heteroduplexing. The mixture of PCR amplicons was then subjected to 2-D electrophoresis under a single set of experimental conditions. With this design, 35 previously identified mutations in 18 different exons were detected in 33 bilateral retinoblastoma patients. These results suggest that 2-D electrophoresis in this format provides a generally applicable, practical and fast way to diagnose with high accuracy large genes for a broad spectrum of possible disease-causing mutations.
Introduction
The large size of some human disease genes in combination with a large number of distinct disease-causing mutations essentially constrains the identification of the precise molecular alterations underlying disease phenotypes, in particular in population-based studies. Examples of disease genes difficult to access by current methods of DNA diagnosis are the retinoblastoma ( RB1 ) gene ( 1 ), the cystic fibrosis transmembrane conductance regulator ( CFTR ) gene ( 2 , 3 ), the mismatch repair genes hMSH2, hMLH1, PMS1 and PMS2 involved in colon cancer ( 4 , 5 ) and the breast and ovarian cancer susceptibility ( BRCA1 ) gene ( 6 ). Such genes, alone or in combination, can only be scanned for all possible mutations at high costs or, alternatively, they can be screened for a limited number of mutations. The latter has become common practice in, for example, cystic fibrosis diagnosis ( 7 ).
Before gene mutational scanning on a routine basis becomes practical, there are a number of technical issues to address, involving both template preparation and the actual mutation scanning. Template preparation is performed most conveniently by polymerase chain reaction (PCR) amplification. Using primers encompassing the target sequences (usually exons, splice sites and the promoter region), sufficient DNA template can be obtained for performing most of the currently used mutation assays, including direct sequencing, without the need for using radioactivity. However, when large genes or multiple genes are involved, the numerous different PCR reactions necessary to prepare the mutational target fragments make the test labor intensive and complex. The need to perform multiple PCR reactions per sample also requires much patient material. For these reasons, PCR multiplexing is used to amplify fragments at different loci simultaneously in the same PCR reaction tube. The design of a set of conditions that allows multiplexing of a large number of gene fragments is not trivial. Indeed, many multiplex PCRs have been described, but most involve less than five fragments. Exceptions are the multiplex sets of nine fragments for the dystrophin gene ( 8 , 9 ).
The second major issue in gene mutational scanning of large genes involves the availability of relatively simple and inexpensive technology for detecting all possible mutational changes in a given fragment. Since gene diagnosis by sequencing on a large scale is not yet cost-efficient, other methods such as single strand conformation polymorphism analysis (SSCP), heteroduplex analysis and denaturing gradient gel electrophoretic analysis (DGGE) are presently being employed ( 10 ). Of these systems, DGGE is generally considered the best approach ( 11 – 13 ). In order to apply DGGE in the most optimal way, there are strict limitations regarding the choice of PCR primers. Primers must be selected in such a way that the resulting fragment has a GC-rich ‘clamp’ attached to one end of the PCR product with the target sequence itself as one lower melting domain ( 14–16 ). The design of suitable primers that fulfil criteria for both DGGE and PCR (including multiplexing) is often far from easy, if not impossible.
In previous work, we have demonstrated the feasibility of two-dimensional (2-D) DNA analysis as a comprehensive method for mutation scanning ( 17 ) and presented a new extensive PCR multiplexing design for preparing DGGE-optimized target sequences ( 18 ). Here we describe the combined utilization of extensive PCR multiplexing and 2-D DNA electrophoresis for the efficient and accurate detection of mutations in a large disease gene. As a model gene, we selected the RB1 tumor suppressor gene. The RB1 gene is an example of a large gene (180 388 bp) that has been sequenced in its entirety ( 19 ). However, its sequence integrity is difficult to evaluate in population-based studies because of the broad spectrum of possible mutations and polymorphisms. Primers that satisfy both PCR and DGGE criteria appeared to be difficult to design ( 20 ). Indeed, even PCR multiplexing without the need for optimal melting behavior is not easy. A multiplex group of seven fragments at most has been described for this gene ( 21 ). The multiplex-PCR/2-D DNA electrophoresis system presented here essentially solved these problems for the RB1 gene and can be implemented in routine mutational scanning of other genes for which a comparable design can be made.
Results
Figure 1 shows the design of the 2-D gene scanning test for the RB1 gene. In a situation optimal for mutational scanning of large genes, all relevant regions should be recovered in one and the same PCR reaction, followed by automatic 2-D separation in a pattern that would reveal all possible mutations as positional shifts of the spots representing the target fragments. Our aim was to approach this situation as closely as possible. First, all exon-containing genomic sequences were amplified simultaneously in a multiplex of six amplicons by long-PCR. Primers for the long-PCR were positioned to obtain all target regions in the smallest possible number of fragments that can still be amplified through long-PCR, i.e. up to at least 20 kb (TaKaRa LA PCR Kit. Product Insert; 22 ). Multiplexing of the six different long-PCR reactions into one single reaction was not a problem. Ample margin for adjustment of primer position was available. The overrepresentation of the RB1 exon-containing fragments relative to all other genomic DNA, resulting from this pre-amplification step, greatly increased the flexibility in experimental design of a multiplex system. Moreover, it allows one to focus almost exclusively on optimization of the melting behavior of the eventual DGGE target sequences; PCR conditions are highly permissive with the pre-specified long-PCR products as template ( 18 ). Thus, using the long-PCR fragments as template, primers for short-PCR were selected to yield fragments of between 100 and 600 bp, the optimal size for detecting mutations by DGGE ( 23 ). A total of 25 fragments (exons 15 and 16 were contained in one fragment) were specified, most of which comprised only one melting domain with lower melting temperature than the GC-clamp attached to it ( 15 ). All amplicons designed for the RB1 gene are listed, with their corresponding long-PCR fragments, in Table 1 . Optimal melting behavior was determined for each candidate target sequence by using a computer program (MELT87; 24 ). Of the entire RB1 coding region, only exon 1 was left out in view of its high GC content. This exon can be analyzed separately under another set of conditions (results not shown).
Schematic depiction of the multiplex PCR/2-D DNA electrophoresis system.
Figure 2 A shows the theoretically predicted (on the basis of the known sizes and calculated melting temperatures) fragment positions in the 2-D gel under the electrophoretic conditions specified, as compared with a typical empirical 2-D pattern obtained from a normal control DNA sample ( Fig. 2 B). The melting program was found to predict the spot positions rather accurately. In this pattern, polymorphisms were observed in the fragment representing exons 15/16 and in exon 17. Both polymorphisms are known and have been found also by sequence analysis (P. van der Vlies, unpublished results). (The characteristic 3-or 4-spot patterns of heterozygous mutations are discussed below.) One aspect that should be noted is the variation in spot intensity. Some attempts to obtain spots of equal intensities by modifying experimental conditions suggest that this situation can be improved. However, unequal spot intensities did not pose a problem in accurate mutation detection.
Table 3 lists the 35 mutations detected by 2-D DNA electrophoresis in 33 different patients with bilateral retinoblastoma. Blood samples and retinoblastoma tumor samples were analyzed. All mutations listed had been identified previously or were confirmed later by sequence analysis. Fifteen out of the 35 were detected in a blind study, that is when analyzing these samples plus a number of added negative controls it was not known which sample contained a mutation and where it was located. The location of the 19 other mutations was known before analysis. In no instance was a second 2-D analysis necessary to confirm the presence of a known mutation. Likewise, in no instance could a mutation detected (in the form of a characteristic 4-spot pattern) not be reproduced or, after sequencing, appeared not to be a mutation at all.
Primer pairs for RB1 PCR multiplexing
Figure 3 shows details of a number of 2-D patterns as examples of previously identified mutations in different exons of the RB1 gene. Under the conditions applied, i.e. GC-clamping and heteroduplexing, heterozygous mutations resulted in four spots: the two homoduplex variants and the two heteroduplex variants. For some mutations, the two homoduplexes or the two hetero-duplexes were not well separated, resulting in three instead of four spots (see, for example, the 1 bp deletion mutation in RB-28 shown in Figure 4 B). This has also been observed in CFTR 2-D DNA electrophoresis ( 17 ). Migration of the heteroduplexes was sometimes slightly retarded already during separation in the first (size) dimension. For example, the heteroduplex fragments corresponding to the mutations in exons 3, 12 and 19, shown in Figure 3 , have migrated more slowly than the homoduplex fragments. Also, this phenomenon has been observed with mutations in the CFTR gene, where it was especially prominent for the deltaF508 3 bp deletion ( 17 ). In a few cases, spots representing the heteroduplex fragments were found to (partially) overlap with spots representing other exons (see, for example, the mutation in exon 19 of RB-26 shown in Figure 3 ).
Figure 4 indicates that different mutations can be distinguished on the basis of positional differences in the 2-D pattern. Three different mutations in exon 18 yield characteristic 3-or 4-spot patterns, which can be recognized relative to the known positions of the non-changed spots. Each situation is also schematically depicted. To confirm the positional constancy of different mutations in exon 18, these determinations were also repeated as single exons in 1-D DGGE tests. Figure 5 shows five different mutations in this exon, including the three shown in Figure 4 . In all cases, the conclusion drawn from the 2-D analyses could be confirmed. In fact, even the presumably hemizygous mutation detected in sample RB-19 (a tumor DNA) is clearly recognizable. Also in CFTR 2-D analyses, positional variation allowed us to distinguish between different heterozygous mutations ( 17 ), and results obtained with the p53 gene indicate that even different mutations in the same codon can be clearly distinguished (unpublished results).
Finally, a number of known (and sequence confirmed) polymorphisms were detected in the 33 patients as well as a total of eight mutations (not listed in Table 3 ), which had not been found by SSCP. As yet, these mutations have not been characterized by sequencing. It, nevertheless, suggests that DGGE is more sensitive than SSCP, a conclusion which has also been drawn by others on the basis of direct comparisons ( 11 , 13 ).
Sequence-confirmed mutations in the RB1 gene detected by 2-D DNA electrophoresis
Discussion
The potential of 2-D DNA electrophoresis for comprehensive mutation analysis in single genes has been demonstrated by our earlier results with the cystic fibrosis transmembrane conductance regulator ( CFTR ) gene. Using primers mostly specified by others for use in 1-D DGGE ( 25 ), 2-D DNA electrophoresis enabled us to detect 17 out of 17 previously identified mutations in a format consisting of 29 spots covering all 27 CFTR exons ( 17 ). The CFTR 2-D format, however, was not based on extensive multiplexing. In this present study, 2-D DNA electrophoresis was combined with a recently developed extensive multiplex PCR design to simultaneously detect mutations in all exons, except exon 1, of the RB1 tumor suppressor gene. This method, which is applicable to any other large disease gene, greatly reduces the costs and time of genetic testing and could become the first practical way to screen large genes with a broad spectrum of possible disease-causing mutations. In this respect, the system should permit the simultaneous analysis of more than one gene at the time, which would reduce the costs even further. Based on the current protocol for RB1, the costs were estimated to be ∼$60 per sample, including reagents, depreciation of equipment, personnel and 100% indirect costs (R. Dhanda, manuscript in preparation).
Key issues in mutation analysis of disease genes are the sensitivity (i.e. the likelihood that mutations present are actually detected) and specificity (i.e. the rate of true negatives) of the assay. The sensitivity of 2-D electrophoresis is determined mainly by the physical characteristics of its separation criteria. In DGGE, optimal melting profiles of the target sequences would theoretically allow detection of mutations with very high accuracy, a prediction that has been confirmed experimentally in several instances ( 11 – 13 ). Also, our present results suggest a high accuracy of mutation detection, which warrant further, more extensive validation studies. However, mutations that will be missed by 2-D electrophoresis are large deletions removing one copy of an entire exon or an entire gene. (The same limitation applies to sequence-based diagnosis.) In 2-D electrophoresis, large deletions involving whole exons or the whole gene would not be revealed by any qualitative difference in spot pattern since the second intact copy of the gene would yield the correct unmutated fragment. Only quantitative differences could show the presence of such mutations. As yet, we have no experimental evidence that the current system can be made quantitative.
Theoretical 2-D DNA electrophoresis pattern of the RB1 gene, as predicted from the sizes and melting profiles of the 25 amplicons ( A ), as compared with the empirical pattern ( B ).
Details of typical 2-D DNA electrophoresis patterns of the RB1 gene in DNA samples from retinoblastoma patients RB-5, RB-23, RB-17, RB-31, RB-7 and RB-26 (see also Table 3 ).
With respect to the specificity of the assay, false positives as technical artefacts of the method do not seem to be a major problem. Indeed, the chance that one spot representing a homozygous fragment will turn into four spots based on some artefact seems small. Although shadow spots, possibly due to PCR artefacts, do occur (see also Figs 2–4 ) these can easily be distinguished from real mutations, also by less skilled individuals. A more important potential source of false positives is the lack of immediate sequence information about the mutations detected. This means that neutral mutations (like polymorphisms) cannot be readily distinguished from disease-causing mutations. However, in several instances, known polymorphisms, such as the ones between exons 15 and 16, in exon 17 and in exon 24, could be distinguished from a mutation (C. Eng, unpublished results). Such cases are characterized by more than four spots.
Common disease-causing mutations as well as neutral polymorphisms can be pre-specified with regard to their position in the gel. Indeed, positional pre-specification of known mutations in a diagnostic database of an image analysis program would greatly simplify clinical application of 2-D DNA electrophoresis. Although it cannot be ruled out that different mutant exons will fortuitously co-migrate, our experience with the CFTR, RB1 and p53 genes leads us to predict that most mutations can be identified on the basis of position alone (see also Figs 4 and 5 ). Extensive validation testing should reveal if this prediction is correct. Presently, alterations detected by 2-D DNA electrophoresis in a given exon need further investigation by sequencing. This is certainly true for RB1, in which many mutations are new.
Materials and Methods
Samples
The DNA samples used for establishing this method were obtained from blood and tumor biopsies of retinoblastoma patients in the Dana-Farber Cancer Institute, the Massachusetts Eye and Ear Infirmary (kindly provided by Dr Thaddeus P. Dryja) and the Department of Medical Genetics, University of Groningen. All mutations, originally found by SSCP screening, had been identified by sequence analysis. In a few samples the mutation was originally found by 2-D analysis.
PCR and heteroduplexing
Primers were obtained from Gibco BRL. For long-term storage, primers were kept in a stock solution of 1 mM in ultrapure water, at −20 °C. For short-term use, they were kept at −20 °C as a solution of 100 µM in ultrapure water. PCR reactions were carried out in thermowell tubes (Costar, Cambridge, MA) in a GeneE thermocycler (Techne, Cambridge, UK) fitted with a heated lid, removing the need for an oil overlay on the samples. Multiplex long-PCR reactions were carried out in a 50µl volume with 100 ng of genomic DNA as template and 0.25 µM of each primer, using the LA PCR kit (TaKaRa). PCR reactions were performed according to the manufacturer's instructions. The conditions were as follows. First, one cycle of 94°C, 1 min, followed by 32 cycles of 98 °C, 20 s/68 °C, 12 min (with 15 s incremental increases), and finally one cycle of 72°C, 5 min. The PCR products were stored at −20 °C for further use. Short-PCR reactions were carried out, using the same GeneE thermocycler, in a 50 µl volume with 4 µl of long-PCR product, 0.125–0.5 µM of each primer (0.125 µM for exons 3, 4, 10, 12, 13, 19, 24 and 27, 0.25 µM for exons 6, 7, 9, 11, 14, 15/16, 25 and 26, and 0.5 µM for exons 2, 5, 8, 17, 18, 20, 21, 22 and 23), 0.25 mM dNTPs, 8 mM MgCl 2 , 5 U of Taq polymerase (Gibco BRL or Promega) and 1% dimethylsulfoxide (DMsO). The PCR conditions were as follows. Five cycles of 94°C, 45 s/52°C, 40 s/68°C, 2 min, then five cycles of 94°C, 45 s/47°C, 40 s/68°C, 2.5 min, then 32 cycles of 94°C, 50 s/55°C, 10 s/40 °C, 40 s/ 67 °C, 2.5 min (with 3 s incremental increases per cycle). After the short-PCR, fragments were heteroduplexed by one complete round of denaturation-renaturation. That is, 67°C, 12 min/98°C, 12 min/52°C, 30 min/44°C, 30 min. After PCR and heteroduplexing, 1/10 volume of loading buffer was added. Based on ethidium bromide staining, there was usually enough sample for several runs.
Details of 2-D patterns showing three different mutations in exon 18 (retinoblastoma patients RB-28, RB-8 and RB-12) with accompanying schematic depictions. ( A ) A control sample showing only one spot for exon 18. ( D ) A mutation in both exon 18 and in 13. (The mutation in exon 13 has not yet been confirmed by sequencing.)
Five different mutations in exon 18 of RB-1, RB-8, RB-10, RB-12 and RB-28 as revealed by 1-D DGGE of the single exon.
Two-dimensional electrophoresis
For 2-D electrophoresis, the DGGE instrument from C.B.S. Scientific Co. (Solana Beach, CA) was used. The mixtures of DNA fragments were first subjected to size separation using a 0.75 mm thick 10% polyacrylamide (PAA) gel at 50°C for 5–6 h at 150 V in 0.5×TAE (20× TAE = 0.8 M Tris-HCl, 0.4 M sodium acetate and 0.02 M EDTA (pH 8.0). The separation pattern was visualized by ethidium bromide staining for 10 min and UV transillumination ofthe gel. The 100–600 bp region in the middle part of the lane (so not including the edges) was quickly cut out and applied to a 1 mm thick 10% PAA gel containing a 0–50% urea/formamide (UF) gradient. Gradients were poured using a simple gradient former (Gibco BRL). Electrophoresis was for 7–11 h at 60°C and 150 V After electrophoresis, the gels were stained with 0.5µg/ml ethidium bromide for 15–20 min and destained in water for another 15 min. The patterns were documented under UV illumination using a polaroid camera.
After establishment of experimental conditions, samples were also analyzed by using an automatic 2-D electrophoresis system. For this purpose, an experimental version was used of an instrument made by Ingeny B.V (Leiden, The Netherlands; see also 26 ). Gels were poured, 10 at a time, in the gel-casting device that comes with the automated 2-D electrophoresis instrument according to the manufacturer's instructions. After polymerization, the gels (between glass plates) were removed from the gel-casting box and cleaned with a wet tissue. They were then placed in the instrument according to the manufacturer's instructions, that is in two gel-holding cassettes with silicone-side sealings. The instrument, containing buffer (0.25×TAE) heated to 50°C, was put in the 1-D mode with the power switched off. After adding loading buffer, samples (up to 40 µl) were loaded in the V-shaped wells of the gels in the automated 2-D electrophore-sis instrument. Gels of 10% PAA were used with a gradient of 0–50% urea/formamide. The first dimension was run at 150 V for 6 h at 50µC. The second dimension was run at 150 V for 6 h at 60°C. After electrophoresis, the gels were stained with ethidium bromide and the patterns documented under UV illumination as described for the manual instruments.
Acknowledgements
We thank Dr Thaddeus P. Dryja for making available DNA samples with known mutations in RB1 and his many useful comments on this work, Dr Leonard Lerman, for allowing us to use the program MELT87, and Rahul Dhanda and Traci Ehrhart Kinst for technical assistance. This work was supported by a research grant from Toyobo Co., Ltd. (Osaka, Japan), by NEI grant EY05321, by the Lawrence & Susan Marx Investigatorship in Cancer Genetics and by the Markey Charitable Trust.
