Denaturing gradient gel electrophoresis (DGGE) is one of the most powerful methods for mutation detection currently available. For successful application the appropriate selection of PCR fragments and PCR primers is crucial. The sequence of interest should always be within the domain with the lowest melting temperature. When more than one melting domain is present the fragment is generally divided into several smaller ones. This, however, is not always necessary. We found that simple modifications of PCR fragments and primer sequences may substantially reduce the number of amplicons required. Furthermore, by plotting the (natural) melting curves of fragments without a GC-clamp, we could explain why fragments theoretically perfect for DGGE in practice failed to reveal mutations. Alternative fragment selection and the use of modified primers (addition of T/A or G/C tails) result in the detection of mutations that originally remained undetected. Our studies extend the utility of DGGE by using a minimum of PCR fragments and achieving a maximum of mutation detection.
Denaturing gradient gel electrophoresis (DGGE) separates DNA fragments according to their melting behaviour (1). This melting behaviour is highly sequence-dependent. The base pair composition of a fragment and more specifically the order of the bases determines the melting behaviour of a fragment. In practice, melting is achieved by electrophoresis of DNA fragments in a polyacrylamide gel containing a gradient of two denaturants, urea and formamide (UF). When a DNA molecule is (partially) melted, the fragment undergoes a change in conformation and, as a consequence, its electrophoretic mobility is reduced. Doublestranded DNA molecules differing by as little as a single base substitution in their lowest melting domain show different melting behaviour and as a consequence melt at different positions along a denaturing gradient gel (2). Introduction of a GC-rich sequence, through PCR using GC-clamped primers (3), not only prevents the fragment from melting completely, it also alters the melting characteristics of the fragment (4) allowing the detection of mutations in the melted part of the fragment. It has been proposed that the GC-clamping of the PCR-amplified fragment increases the number of the detectable DNA variants from 40 to virtually 100% (3–5). Due to this high accuracy, DGGE is frequently used for the detection of small mutations in hereditary diseases.
Although DGGE has proven to be a powerful technology to identify single base changes as well as small deletions, insertions and other rearrangements (6), successful application of this methodology needs optimized experimental conditions for each DNA fragment.
As mentioned, DGGE can detect mutations in the lowest melting domain only. Therefore, it is important to determine when it becomes partially melted and when it becomes single-stranded while running in a DGGE gel. This melting behaviour can be predicted using special computer programs designed for this purpose (1,3,7). This melting profile can be used to select the positions of PCR primers that will generate fragments suitable for DGGE and for determining the range of denaturant concentrations to be used. To determine whether a DNA fragment is suitable for DGGE two criteria are generally accepted (8). Firstly, complete dissociation of the DNA strands must be prevented, which is usually circumvented by using a GC-clamp primer. Secondly, the sequence of interest must have an optimal melting profile, i.e. one flat melting domain. This may not be possible due to the presence of two or more melting domains within the PCR fragment to be analysed, thus preventing the detection of mutations in the higher melting domains. This is circumvented, although at the cost of effectiveness and time, by using more PCR fragments. To extend the utility of the DGGE system, we have exploited ways to alter the melting behaviour of DNA fragments through slightly modified PCR primers and/or primer positions. Furthermore, we determined that PCR fragments with imperfect melt curves can be used as well. The proposed primer adaptations or alterations result in a minimum number of PCR amplicons, while still achieving maximum mutation detection.
Materials and Methods
DNA amplification and denaturing gradient gel electrophoresis
PCR and time-travel parallel DGGE were carried out as previously described (2,9). The sequences of the primers used for amplification are available on request. From experience we learned that purification of GC-clamp primers is not necessary.
Melting profile construction
The melting profiles for the fragments used in this study were constructed using the MELT87 computer program (7). An updated melting program (MELT94) can be found on the Internet at http://web.mit.edu/osp/www/melt.html
Results and Discussion
To determine whether the GC-clamp should be attached to the 5′- or 3′-end of each PCR fragment, the respective melting maps need to be compared. As an example we show the theoretical melt curves of exon 8 of the RET gene (Fig. 1A). When attaching the GC-clamp to the forward primer several melting domains were predicted (Fig. 1B). When, however, a GC-clamp is attached to the reverse primer an ideal two-domain melting profile consisting of a single flat lower melting domain is created. However, such an optimal two-domain melting profile is not found in all situations. This raises the question whether PCR fragments with imperfect melt plots can be used for the appropriate detection of base pair variations and whether simple modifications of the PCR fragments/primer sequences can alter unsuitable (DGGE) fragments to become suitable for DGGE analysis whilst still maintaining optimal mutation detection. In the following we present examples of a number of such situations.
Multiple melting domains
When a DNA fragment with two or more different melting domains is separated by electrophoresis in a DGGE gel, the fragment will be arrested at the position in the gel where the denaturant concentration dissociates the fragment at its lowest melting domain. As mobility decreases, the fragment may not reach the position in the gel where the second melting domain will melt. Most of these partially melted fragments appear as sharp and focused bands, giving the idea that the fragment is suitable for mutation detection by DGGE. When fragments contain two melting domains (not taking into account the GC-clamp), the most obvious solution to this problem is to divide this fragment into two amplicons. This, however, is not always necessary. To illustrate that fragments with imperfect melt curves can be used for DGGE we present exon 13 of the MSH2 gene (Fig. 2A). When analysing the fragment with a GC-clamp we could only get the ‘ideal’ melt plot when the primer was located within the exon. When choosing the GC-clamped primer 99 bp upstream from the intron-exon boundary, a (4°C) lower melting domain between the GC-clamp and 3′-end higher melting domain of the fragment appeared (Fig. 2B). This fragment melted out as a smear (Fig. 2C) and a known frameshift mutation (del693T; 9) could not be detected (Fig. 2B). When choosing a GC-clamped primer 29 bp upstream from the intron-exon boundary, the interior lowest melting domain at the 5′-end of the fragment had a Tm value 1.0°C lower than the second domain (Fig. 2E). The del693T mutation could be detected easily (Fig. 2F).
From this example and several comparable cases tested, we conclude that a fragment with an interior low melting domain can be used for DGGE, provided that the Tm value of this low melting domain differs by not more than 1°C from the adjacent domains and is not larger than 50–100 bp in length.
Attachment of G/C nucleotides
When a DNA fragment consists of three melting domains (as depicted in Fig. 3A) mutation detection is only possible in the lowest melting domain and is not possible in the higher (non-melted) domain. When this lowest melting domain is short, <50 bp, and is on one of the ends of the fragment (Fig. 3A) a solution for this condition is the addition of a stretch of G/C nucleotides (three to seven) to this end of the fragment through the short primer. This modified primer will increase the Tm value of the small lowest melting domain and thereby promote the establishment of the desired two-domain profile. To illustrate the results of adding G and/or C residues to the short primer, we examined exon 7 of the BRCA2 gene (Fig. 3A). To determine the melting behaviour of the fragment, we introduced a mutation in the forward primer (10). The fragments without an added stretch of GC nucleotides showed a reduction in mobility and appeared as a single sharp focused band after 5 h electrophoresis at 150 V (Fig. 3B). The results can only be interpreted as absence of mutations in this fragment. When three bases (CGC) were attached to the 3′-end of the fragment, through the reverse short primer, the calculated melting map of this fragment predicted that the Tm of the lowest melting domain would differ by ∼1°C from the other domain (Fig. 3C). Although the melt curve is not optimal, the mutation could be detected after 7–8 h electrophoresis at 150 V (Fig. 3D). When five or seven bases were added to the 3′-end of the fragment, the theoretical desired two-domain profile was created and the mutation was resolved after 6–7 h electrophoresis at 150 V (data not shown).
Thus, based on computer calculations and practical experimentation, we propose that the attachment of a small stretch of G/C nucleotides to the short primer may improve mutation detection under these conditions. It should be noted, however, that the stretch of G/C nucleotides should not exceed 10 bases, since that will introduce a small highly GC-rich portion into the fragment, resulting in smears or diffuse bands in a DGGE gel (data not shown).
Attachment of A/T nucleotides
When sequence information is limited it might be impossible to select primers without having a GC-rich domain near the GC-clamp attachment site of the fragment (Fig. 4A). As a consequence, mutations in these higher melting domains will be missed. We hypothesized that insertion of a stretch of 15–20 T nucleotides between the GC-clamp and the specific primer might decrease the Tm value of the GC-rich sequence and thereby make mutation detection possible in the higher melting domain. To test this we examined exon 9 of the MSH2 gene (Fig. 4A). To obtain the (theoretically) desired two-domain melting profile, 20 T nucleotides had to be inserted at the 3′-end or 15 T nucleotides at the 5′-end between the GC-clamp and specific primer (Fig. 4C). To test this we examined a g→t substitution located at the splice acceptor site in intron 8 (9) present in a high melting domain. As expected, given its location within the highly GC-rich portion at nucleotide position −1 (Fig. 4A), the mutation could not be detected when the GC-clamp was directly attached to the forward primer (Fig. 4B). The mutation, however, was detected when 15 T nucleotides were inserted between the GC-clamp and the forward primer (Fig. 4D).
The insertion of T nucleotides between the GC-clamp and specific primer enables mutation detection in small highly GC-rich regions near the GC-clamped primer.
High melting domain at the end of a DNA fragment
We came across several amplicons with perfect melt curves (fragments with GC-clamp) that, however, when electrophoresed in a DGGE gel resulted in smears or diffuse bands. Upon analysis of the melt curve of the native sequences (fragments without GC-clamp), we observed that a high melting domain was present at one end of the fragment. The effect of this high melting domain on mutation detection by DGGE can be illustrated by an analysis of exon 13 of the RET gene (Fig. 5A). The DGGE analysis of a T→G substitution (S767R) in exon 13 of the RET gene located at nucleotide position 18 of the fragment shows that only indistinct bands can be observed (Fig. 5B). This demonstrates a discrepancy between the theoretical melt plot calculated by MELT87 and the practical ‘real’ melting behaviour of the PCR fragment. The easiest way to solve this problem is to remove the GC-rich sequence. In this case, we could remove 44 bp from the 3′-end of the fragment without removing any of the coding sequence. Although the theoretical melting curve is not perfect (Fig. 5C), Figure 5D shows four distinct bands. Similar theoretical data and practical observations were obtained for exons 14 and 15 of the RET gene (data not shown). Since it has been reported that a short stretch of AT-rich sequence could decrease the Tm value of a small GC-rich region (11), we tried to solve the above mentioned problem by introducing a stretch of seven AT nucleotides at the 3′-end of the reverse short primer. This was, however, unsuccessful (data not shown). Use of a longer stretch of AT nucleotides (>20 nt) will introduce a low melting domain resulting in a non-optimal three-domain structure.
If the melting analysis of a short fragment (<200 bp) predicts a high melting domain <40 bp in size located at the end of the fragment and differing by not more than 5°C in Tm value, this fragment most probably is suitable for DGGE analysis. When, however, a high melting domain (>50 bp in size and differing from the rest by ∼5°C in Tm value) is located at one end of a DNA fragment it may significantly affect the melting behaviour of the fragment even after attachment of a GC-clamp.
High melting domain in the middle of a fragment
From the foregoing it is clear that in designing a DGGE mutation detection system, melt curves of both the clamped and the unclamped fragments need to be examined. In several cases we found a GC-rich domain in the middle of a fragment, visible as a peak in the melting profile of the native sequence, but not seen after addition of a GC-clamp (Fig. 6A). To determine whether such a high melting domain has an influence on the detection of mutations in the domain, we examined CFTR exon 2, of which the melt map with and without a 3′-GC-clamp is shown in Figure 6A. We examined three known mutations located in the middle of this fragment: 186-13c→g at nucleotide position −13; 241delAT at nucleotide position 39 of the fragment, located in the peak of the melting domain (79°C); 296+2t→c at nucleotide position +2 (Fig. 6A). Figure 6B shows a DGGE analysis of the three mutations in exon 2 of the CFTR gene. All three mutations were easily detected. Similar theoretical data and practical observations were obtained for exon 5 of the TP53 gene, which has a somewhat broader peak in the middle of the fragment (data not shown). Thus, mutations located in such high melting domains do not pose a problem for detection.
For DGGE, preferentially short PCR fragments (<300 bp) should be chosen, because a GC-clamp has a stronger effect on the melting properties of short fragments. Fragments suitable for DGGE can thus be generated more easily. In long fragments (>400 bp), a large internal high melting domain (100 bp) may have a substantial impact, as the GC-clamp has relatively little effect on the melting behaviour of the central portion, for instance, exon 4 of the TP53 gene (11). Mutations located in the high melting domain (100 bp long, differing by 6°C in Tm value) and in the domain between this high melting domain and GC-clamp could not be detected using a single PCR fragment in DGGE (data not shown).
We conclude that whereas the presence of a high melting domain in the middle of a small fragment (<300 bp) still allows a good mutation analysis, its presence in the middle of a large PCR fragment (>400 bp) makes the fragment unsuitable for DGGE analysis.
Length of GC-clamp
Several studies (3,5) have indicated that a GC-clamp as short as 30 bp would be sufficient in DGGE. This may be true for AT-rich fragments, but for a GC-rich sequence the difference in Tm with a GC-clamp might become too small. Even the standard 40 bp GC-clamp might not be sufficient to prevent total strand dissociation. Here, we demonstrate the effect of the length of the GC-clamp on GC-rich fragments with exon 5 of the RET gene as an example. The melting analysis of the fragment revealed a Tm value of 80°C. Attachment of 40 and 60 bp GC-clamps to the 3′-end of the fragment gave similar theoretical melting profiles (Fig. 7A). The fragment with the 40 bp GC-clamp becomes completely single-stranded and runs off the gel (Fig. 7B and also Fig. 8A and B), whereas the 60 bp GC-clamped fragment melted as a single sharp band (Fig. 7C).
For extremely GC-rich sequences (Tm > 80°C), as is usually the case for the first exon of most genes, the difference in Tm values between GC-clamp and target sequence may be so small that even a 60 bp GC-clamp may not be sufficient to prevent total strand dissociation, although the melt curve might be perfect. If possible, a naturally occurring, high melting domain in combination with a longer GC-clamp may be used (Fig. 8C). As an example we show exon 1 of the MSH2 gene (Fig. 8A–D). When analysing the melting behaviour of a 299 bp fragment, a naturally occurring, high melting domain of ∼60 bp at the 3′-end of the fragment becomes visible (Fig. 8C). By adding a 55 bp GC-clamp to the 3′-end of this molecule, a high melting domain of ∼100 bp was created (Fig. 8C). When running this PCR fragment in a DGGE gel it resulted in a single sharp band melting at an appropriate UF concentration (Fig. 8D).
We conclude that for fragments with a Tm value close to 80°C a GC-clamp with a length of 60 bp will improve mutation detection. DGGE for fragments with a Tm value >80°C might only be possible when making use of a long GC-clamp in combination with a ‘naturally’ occurring GC-rich domain.
We thank Dr Hans Scheffer for providing the CFTR mutations used in this study.