Trinucleotide repeats associated with human disease
Trinucleotide repeats associated with human disease
Michael Mitas
Department of Biochemistry and Molecular Biology, Oklahoma State University, 246 Noble Research Center, Stillwater, OK 74078, USA
Received December 20, 1996;Revised and Accepted April 21, 1997
ABSTRACT
Triplet repeat expansion diseases (TREDs) are characterized by the coincidence of disease manifestation with amplification of d(CAG·CTG), d(CGG·CCG) or d(GAA·TTC) repeats contained within specific genes. Amplification of triplet repeats continues in offspring of affected individuals, which generally results in progressive severity of the disease and/or an earlier age of onset, phenomena clinically referred to as `anticipation'. Recent biophysical and biochemical studies reveal that five of the six [d(CGG)n, d(CCG)n, (CAG)n, d(CTG)n and d(GAA)n] complementary sequences that are associated with human disease form stable hairpin structures. Although the triplet repeat sequences d(GAC)n and d(GTC)n also form hairpins, repeats of the double-stranded forms of these sequences are conspicuously absent from DNA sequence databases and are not anticipated to be associated with human disease. With the exception of d(GAG)n and d(GTG)n, the remaining triplet repeat sequences are unlikely to form hairpin structures at physiological salt and temperature. The details of hairpin structures containing trinucleotide repeats are summarized and discussed with respect to potential mechanisms of triplet repeat expansion and d(CGG·CCG)n methylation/demethylation.
INTRODUCTION
Recently, 11 human genetic diseases associated with expansion of d(CTG·CAG), d(CCG·CGG) and d(GAA·TTC) triplet repeat sequences have been identified (Fig. 1). In general, expansion of these triplet repeats continues in offspring of affected individuals, resulting in progressive severity of the disease and/or an earlier age of onset, phenomena clinically referred to as `anticipation'. An individual afflicted with one of these diseases typically contains >50 trinucleotide repeats within a specific gene, while unaffected individuals contain <30 repeats (1). Examples of diseases caused by expansion of triplet repeats are Huntington's disease (2), fragile X syndrome (3,4) and myotonic dystrophy (5-7).
Although minor instabilities of microsatellite d(AT·AT)and d(CA·TG)dinucleotide repeat sequences have been observed in cells derived from some colon cancer patients (8-13), the changes in the numbers of these dinucleotide sequences (deletions or increases): (i) are rather small; (ii) occur throughout the genome; (iii) are caused by defects in the mismatch repair system (14-19). In contrast, the changes in the numbers of triplet repeat sequences associated with human disease: (i) are rather large; (ii) are localized to specific genes; (iii) do not involve the mismatch repair system (20). It can therefore be concluded that mechanisms that lead to triplet repeat expansion are inherently different from those that lead to microsatellite instability associated with defective mismatch repair.
To distinguish the diseases that are associated with defects in mismatch repair from those that are associated with trinucleotide repeat expansion, the latter family of diseases will be referred to as triplet repeat expansion diseases(TREDs). An implication of this acronym is that sequence expansion is driven by the ability of certain triplet repeats to form unusual DNA structures (i.e. hairpins). This review will summarize the salient features of hairpin and other structures adopted by trinucleotide repeats, with an emphasis placed on those sequences associated with human disease. Portions of this review will also discuss why some hairpins might be better than others at promoting expansion events and why expansion of CGG repeats leads to gene hypermethylation. For a description of the genetic and clinical features of TREDs, readers are encouraged to consult the excellent reviews by Ashley and Warren (1), Timchenko and Caskey (21) and Sutherland and Richards (22).
Figure 1 Triplet repeat expansions: diseases and fragile sites. Schematic diagram depicting the relative positions of expanded triplet repeats within a fictitious gene containing four exons (boxes) and three introns (dotted lines). Diseases or fragile sites are listed below the corresponding expanded trinucleotide repeat. AUG and TAA indicate initiating and stop codons respectively. Fragile sites XF and 16A are thought not to reside within any genes and, like fragile site 11B, are not known to result in any disease phenotype. The above figure was adapted from Warren (60) and differs from it by the inclusion of synpolydactyly, a disease due to duplication of d(GCG)n (90) and spinocerebellar ataxia type 6 (91). *Location inferred but not proven.
STRUCTURES OF SINGLE-STRANDED TRINUCLEOTIDE REPEATS
Replication of duplex B-DNA requires separation of the two parental strands at the replication fork. It is during this time that stable single-stranded DNA structures, such as hairpins comprised of inverted or triplet repeat sequences, have the opportunity to form. The presence of mismatches in these hairpins provides the ostensibly flexible B-DNA helix an opportunity to sample different H bonding and/or stacking arrangements at the site of the mispair. One goal in understanding mechanisms of triplet repeat expansion is to characterize the most stable base pairing arrangements of hairpins containing triplet repeat sequences and to determine the relevance of these structures to DNA replication and/or DNA repair.
Triplet repeat sequences that adopt hairpin conformations
d(CTG)n: the T-T mismatches in (CTG)n-containing hairpins are well stacked in the helix and contain two hydrogen bonds. CTG-containing sequences have been studied by a variety of techniques, including polyacrylamide gel electrophoresis, KMnO4 modification, P1 nuclease digestion, UV absorbance, 1H NMR and molecular dynamic simulations (23-30). Since the chemical probe KMnO4 oxidizes the 5=6 double bond of thymines that are not involved in extensive stacking interactions (31,32), it has been extremely useful for determining the stability of the T-T mispairs in d(CTG)n hairpins. At 50°C and 50 mM Na+, KMnO4 oxidized the thymines in the loop but not the stem region of d(CTG)15, indicating that the T-T mismatches are well stacked in the helix. In support of this conclusion, 1H NMR studies have shown that the T-T mismatches in the [(CTG)3]2 duplex (28) or (CTG)6 hairpin (33) contain two H bonds formed between O2 and N3 and N3 and N4 (Fig. 2). Due to the asymmetric H bonding arrangement within the T·T pair, the bases (both of which are in the anti conformation) undergo rapid structural transitions (Fig. 2). The C1[prime]-C1[prime] distance in the T·T pair is 8.6 Å (28), much shorter than that observed in canonical B-DNA structures.
Figure 2 Base pair arrangements of hairpin structures containing d(CTG), d(GTC), d(GAC) and d(CGG) triplet repeat sequences. Shown on the left are hairpin alignments for d(CTG)n, d(GTC)n, d(GAC)n and d(CGG)n sequences as determined from studies cited in the text. Ovals denote Watson-Crick C·G base pairs. Lines between mismatched bases denote non-Watson-Crick base pairs, which are shown to the right.
UV absorbance (26,27) and electrophoretic mobility (24) melting profiles of (CTG)n have revealed sharp structural transitions, indicative of cooperative binding/melting of the C·G and T·T base pairs. A cooperative interaction of these base pairs was confirmed by temperature-dependent 1H NMR studies, which showed that the T·T base pairs melt at the same temperature as the C·G base pairs (28).
A thymine-thymine mismatch in the d(CCCTGGG)2 duplex was previously examined by Arnold et al. using high resolution nuclear magnetic resonance (34). The T-T mismatches in the d(CCCTGGG)2 and d(CTG)n structures are similar in the sense that both are flanked by the same nucleotides: a cytosine to the 5[prime]-side and a guanine to the 3[prime]-side. The T-T mismatch in d(CCCTGGG)2 was intercalated into a B-DNA helix but was not H bonded. Replacement of the T·T with an A·A pair resulted in destabilization of the duplex, since the mismatched adenines must tilt and push apart to reduce the overlap of the amino groups. Thus, in the context of a CNG motif, a T·T pair is more stable than an A·A pair. As described below, replacement of T·T pairs in a (CTG)n-containing hairpin with A·A pairs also results in destabilization of the helix.
d(GTC)n: the T-T mismatches in (GTC)n-containing hairpins are well stacked in the helix. For unexplained reasons, the frequency of d(GTC)n·(GAC)n repeats is the lowest among all of the triplet repeats (23,35). The apparent selection against these sequences probably precludes their association with human disease. At 40-50°C and 200 mM monovalent cation, KMnO4 oxidized the thymines in the loop but not the stem region of d(GTC)15 (24), indicating that the T-T mismatches are stacked in the helix, even at temperatures well above physiological conditions. Lowering of the salt concentration to 50 mM monovalent cation resulted in oxidization of the thymines in the loop and stem regions of d(GTC)15, indicating that the hairpin conformation of d(GTC)15 was not as stable as d(CTG)15. In support of this conclusion, electrophoretic mobility (24) and UV absorbance melting profiles (27) have shown that the Tm of d(GTC)n was [sim]10°C lower than d(CTG)n. The decrease in stability of the d(GTC)15 hairpin [relative to d(CTG)15] can be explained on the basis of reduced inter-strand stacking energies (23). 1H NMR studies have shown that the T-T mismatches in the [(GTC)3]2 duplex (30) form an identical H bonding arrangement compared with a (CTG)n-containing hairpin (Fig. 2).
d(GAC)n: the A-A mismatches in (GAC)n-containing hairpins are well stacked in the helix. Previous studies on DNA sequences containing d(GAC)n motifs have focused on the ability of this sequence to adopt a parallel-stranded duplex structure under acidic conditions (36,37). However, at neutral pH oligonucleotides containing repeats of d(GAC)n favor hairpin conformations (26,27,38). Information on the nature of the A-A interactions in d(GAC)n hairpins has been obtained from chemical modification studies of [Delta]AXIV(GAC)15, a DNA sequence in which the adenine in triplet 14 (AXIV) was deleted, and from 1H NMR studies of the duplex [(GAC)3]2 (30). The chemical modifying agent used for analysis of the A-A mismatches was diethylpyrocarbonate (DEPC), which carbethoxylates N7 of adenines that lack extensive base stacking interactions (32,39). In 50 mM Na+, pH 7.5, and [le]40°C, DEPC reacted well with AVIII and AIX (loop region) and AII (the adenine opposite the deleted base), but poorly with the adenines in the hairpin stem (A.Y., M.C. and M.M., unpublished results), providing evidence that the adenine pairs were stacked in the helix. The ratios of DEPC reactivities for the stem/loop were 0.24 at 30°C and 1.02 at 70°C (where a ratio of 0.0 corresponds to a completely H bonded and/or stacked pair and 1.00 corresponds to a random coil structure). The Tm values of the adenine pairs in 50 and 200 mM monovalent cation were 44 and 55°C, respectively. The reduction of chemical modification of the A·A pair under physiological conditions suggests that, like the T·T pair in d(CTG)n-containing hairpins, the A·A pair may be stabilized by H bonds. In support of this, Gao and colleagues have shown that the A-A mismatches in the [(GAC)3]2 duplex contain one H bond formed between N1 and N6 (Fig. 2), similar to that observed in the context of the K-ras gene (40).
d(CAG)n: the A-A mismatches in (CAG)n-containing hairpins are not well stacked in the helix. Electrophoretic mobility and UV absorbance melting profiles have provided evidence that d(CAG)n hairpins are less stable than d(GAC)n hairpins (Table 1). In support of this conclusion, chemical modification experiments have indicated that the mispaired adenines in the stem of d(CAG)15 are only partially resistant to modification by DEPC at 40°C in 200 mM monovalent cation (Fig. 3). Quantification of the modification data revealed that the Tm values of the A-A pairs in d(CAG)15 were <37°C (A.Y. and M.M., unpublished results), providing evidence that the A·A pairs in d(CAG)15 were not well stacked in the helix at physiological temperature. This conclusion is supported by 1H NMR studies of the [(CAG)3]2 duplex, in which the A-A mismatches were rather flexible and contained no H bonds (30).
Figure 3 The mismatched adenines in the d(CAG)15 hairpin are not stacked in the helix at physiological temperature. The sequence shown to the left of the autoradiograph was labeled with 32P, reacted with DEPC in buffer containing 50 mM Na+ and 150 mM K+, pH 7.5, at the indicated temperatures and applied to a DNA sequencing gel. Conventional Arabic numerals depict the position of a nucleotide with respect to the 5[prime]-end. Roman numerals depict triplet repeats. Dotted lines indicate the positions of adenine residues in the hairpin that are extensively modified by DEPC at 30°C. At [ge]40°C the mismatched adenines react highly with DEPC, providing evidence for weak stacking interactions and/or melting of the hairpin structure. Photograph kindly provided by Dr Adong Yu.
. Melting temperatures of hairpins containing 15 triplet repeat sequencesa in [sim]1 mM Na+
d(XXX)15
pH
Tm (oC)
Reference
CGG
8.5
75
42
GAA
8.5
[sim]50
Suen et al., unpublished
GAC
8.5
49
38
CTG
8.5
47
24
CAG
8.5
38
38
GTC
8.5
38
24
CCG
7.5
37
57
CCG
8.5
30
57
aOligonucleotides containing 15 triplet repeats also contained flanking sequences. The sequence of the oligonucleotide containing 15 pyrimidine-rich triplet repeats was GATCC(XXX)15GGTACCAAGCT, where XXX = CCG, CTG or GTC. The sequence of the oligonucleotide containing 15 purine-rich triplet repeats was AGCTTGGTACC(XXX)15GGATC, where XXX = CGG, CAG or GAC.
(CGG)n-containing sequences form hairpins and tetraplexes. A discussion of d(CGG)n structures necessitates nomenclature that allows comparison of studies performed by different laboratories. For example, use of the terms `staggered', `slipped' or `blunt' to describe the base pairing alignments in hairpin or duplex structures containing repeats of d(CGG) (29,30,41,42) is ambiguous. For simplicity, d(CGG)n hairpin or duplex structures that contain the more stable GpC base pair step (43) will be referred to as the a alignment, while those that contain the less stable CpG base pair step will be referred to as the b alignment (see Fig. 2 for the b alignment).
In all studies where the base pairing arrangement of the d(CGG)n-containing structure has been determined, a b alignment is adopted (30,41,42,44). However, there are discrepancies regarding interaction of the G-G mismatches in these studies. For example, in [d(GGC)n]2 (where n = 4-6) duplexes (29,41) and d(GGC)11 (29) or d(CGG)15 (42) hairpins, Gsyn·Ganti base pairs were formed. This arrangement provides a minimal amount of distortion to the DNA helix (Fig. 2). Consequently, of hairpins containing CNG or GNC triplet repeats, those formed from (CGG)n are the most stable (Table 1). In the d[(CGG)3]2 duplex, Zheng et al. reported that the guanines in the mismatches were not in a syn conformation nor a base paired form (30). Rather, the G residues were conformationally mobile, most likely undergoing dynamic exchange among various glycosidic conformational isomers (30). Although this flexible G-G mismatch may be applicable only to short duplexes of d(CGG)n, it nonetheless provides a thermodynamic driving force for the potential folding of d(CGG)n sequences into tetraplexes.
Since d(CGG)n sequences are G-rich, they also have the potential to form inter- or intramolecular tetraplexes that differ markedly in base pairing arrangements. One such theoretical tetraplex contains G4 (45,46) and C42+ (47) quartets in a 2:1 ratio, while the other contains G4 and CGCG quartets in a 1:2 ratio. Patel and colleagues characterized, by 1H NMR, the structure of d(GCGGT3GCGG), a DNA sequence that contains part of the d(CGG)n repeat present in the FMR-1 gene associated with fragile X syndrome. These investigators showed that d(GCGGT3GCGG) formed a hairpin dimer (both hairpins were in the b alignment) that contains two CGCG quartets (Fig. 4) and two G4 quartets (44).
Figure 4 CGCG quartet structure contained within a d(CGG)n tetraplex. The CGCG quartet structure shown above was observed in the hairpin dimer formed by d[(GCGGT3GCGG)]2 (44), which contains part of the d(CGG)n repeats of the FMR-1 gene. Vertical oriented dotted lines are H bonds formed between Watson-Crick C·G within the hairpin. Dotted lines oriented at [sim]45° are H bonds formed between adjacent hairpins.
Usdin and Woodford examined the structure of d(CGG)20 and discovered that the N7 atoms of all guanines in this sequence were protected from modification by DMS in the presence of 20 mM K+ at pH 8.5 (48). Addition of monovalent cations (K+ works best) serves to stabilize a G4 quartet by reducing the electronegativity of the four O6 atoms that point towards its center (45,47,49,50). The K+-dependent d(CGG)20 structure exhibited increased electrophoretic mobility, demonstrating that it formed an intramolecular tetraplex (i.e. a hairpin folded in half) containing G4 quartets.
d(CCG)n-containing sequences adopt two entirely different hairpin structures. Interestingly, of the eight potential mismatches, C-C mismatches are the least effectively repaired in bacterial or mammalian systems (14,17,51,52) and, in addition, have never been crystallized in duplex DNA. On the other hand, G-T mismatches are repaired the most efficiently of the eight mismatches (53) and are stabilized by two H bonds in crystals (54). Smith and colleagues were the first to provide biophysical and enzymatic evidence that d(CCG) repeats form hairpins containing C-C mismatches (55).
Depending upon the value of n, d(CCG)n sequences adopt hairpin alignments that contain either CpG (a alignment) or GpC (b alignment) base pair steps. For example, when the terminal 5[prime] nucleotide in a d(GCC)5-7 sequence is a guanine, hairpins are formed in the a alignment (29,41; Fig. 5). 1H NMR studies have not uncovered any H bonds within the C-C mispair of this alignment, suggesting that the mismatched cytosines are very flexible.
Figure 5 Base pair arrangements of hairpins containing d(CCG) repeats. Shown are schematics of hairpin alignments of d(CCG)n sequences as described in the text. The value n indicates the number of repeats necessary for stable formation of the indicated structure. The 3 nt in the boxes are the recognition motif of the human (cytosine-5) MTase (70). The x and y placed by the cytosines in the extrahelical b alignment are thought to pair together in the minor groove, as shown in the all-atom diagram (kindly provided by I.S. Haworth), where the bold lines represent the mispaired cytosines. For comparison, the same cytosines are marked with an x and y in the b alignment containing intrahelical cytosines. *This conformation has been observed in a d[(CCG)3]2 duplex as a minor species in one NMR study (30).
A completely unexpected and new DNA structure that has emerged from studies of triplet repeat sequences is the e motif (56) formed from the sequence d(CCG)2. This sequence differs from those characterized by Gupta and colleagues in that the 5[prime]-terminal nucleotide is a cytosine instead of a guanine. Gao and colleagues discovered that the duplex structure adopted by d(CCG)2 was one that contained two CpG steps (i.e. a b alignment) rather than a single GpC base pair step (56). Surprisingly, the cytosines within the lone C-C mismatch, which was centrally located in the duplex, was not stacked within the helix. Instead, the cytosines were extrahelical and symmetrically located in the minor groove. Due to the nature of the mispaired cytosines, the d[(CCG)2]2 structure was called the extrahelical, or e motif. In order to accommodate the extrahelical cytosines, the sugar-phosphate backbone of the e motif is highly distorted. We refer to a d(CCG)n duplex (or hairpin) structure containing a CpG base pair step as the b alignment (Fig. 5).
The structure adopted by d(CCG)15 is the most unusual of G+C-rich triplet repeats, hence it has proved difficult to characterize (57). In contrast to hairpins containing small numbers of d(CCG) repeats, the hairpin structure of d(CGG)15 adopts a b alignment at pH values between 6.0 and 8.5. Electrophoretic, circular dichroism (CD) and P1 nuclease digestion studies indicate a structural transition of d(CCG)15 at pH 7.7 ± 0.2, indicating protonation of the mispaired cytosines at physiological pH. However, the results of CD and chemical modification studies suggest that d(CCG)15 contains no C·+C base pairs. The results of a number of studies are consistent with a structure which essentially represents an `extended' e motif such that the extrahelical cytosines interact with one another in the minor groove of a very distorted helix (Fig. 5). The structure is stabilized by creation of pseudo-GpC base pair steps and is not too conceptually different from triad DNA postulated by Kuryavyi and Jovin (58).
d(GAA)n forms a hairpin that appears to contain A·G base pairs. Friedreich's ataxia is an autosomal recessive disease that is caused by expansion of an intronic GAA triplet repeat (59). Recent experimental data indicate that d(GAA)15 adopts an unconventional hairpin that contains A·G base pairs (I.-H.S., M.C., B.M. and M.M., unpublished results). Consistent with the purine-rich nature of the sequence, it is possible that the hairpin is additionally stabilized by stacking interactions of the A-A mismatches.
It was previously suggested that because d(GAA)n sequences were not likely to form hairpins, all triplet repeat sequences might be prone to expansion (60). Others have suggested that there might be multiple mechanisms of triplet repeat expansion, one responsible for C+G-rich sequences and one responsible for GAA sequences (21). The recent finding that d(GAA)15 forms a hairpin suggests that only those triplet repeat sequences that are able to form hairpins are likely to undergo expansion.
Unstructured triplet repeat sequences
d(GAT·ATC)n is a triplet repeat sequence that is not associated with any human disease. d(GAT)15, d(ATC)15 and d(TTC)15 [the complementary sequence of d(GAA)n, associated with Friedreich's ataxia] have been analyzed by native polyacrylamide gel electrophoresis, chemical modification and P1 nuclease digestion (23; unpublished results). These sequences contain no preferred secondary structure.
Uncharacterized triplet repeat sequences
Of the remaining 10 uncharacterized triplet repeat sequences, eight are highly unlikely to adopt hairpin conformations. For example, since d(GAT)15 and d(ATC)15 do not form hairpin structures, it is improbable that the other four triplet repeat sequences containing an AT or TA palindromic dinucleotide [d(TAG)n, d(CTA)n, d(ATA)n and d(TAT)n] will do so either. Likewise, the sequences d(TTG)n, d(CAA)n, d(CTC)n and (CAC)n are predicted not to form hairpins at physiological salt and temperature due to lack of significant base stacking and/or pairing interactions. The remaining two triplet repeat sequences [d(GAG)n and (GTG)n] are rich in guanines, nucleotides that have the potential to stabilize hairpin structures by formation of H bonds through N7. Thus, since d(GAG)n and d(GTG)n can potentially adopt hairpin conformations, it is possible that either of these sequences might undergo expansion in humans and be associated with disease. Indeed, Wells and colleagues have shown that d(GTG·CAC)n sequences are expanded at low levels in Escherichia coli (61).
FLEXIBLE HAIRPIN STRUCTURES AND TRIPLET REPEAT EXPANSION
Flexible hairpin structures: why?
In addition to hairpin structures, some of the sequences described above are capable of forming tetraplexes [d(CGG)n] and triplexes [d(GAA/TTC)n] (62), structures which may play an important role in disease manifestation. For example, there is good evidence that transcription can induce a d(GAA/TTC)n·(GAA)n triplex structure, thus allowing the nascent RNA to bind to the exposed d(TTC)n single strand (E.Grabczyk and K.Usdin, personal communication). Due to the multitude of DNA conformations adopted by triplet repeat sequences, many have questioned whether specific DNA structures are required for triplet repeat expansion. Another question is whether non-triplet repeat minisatellite sequences are capable of undergoing expansion. The recent cloning of the human fragile site FRA16B has apparently provided answers to both questions. This fragile site results from expansion of a minisatellite containing 33 AT-rich base pairs (63). The length of this repeat is not consistent with an expansion mechanism that involves reiteritive synthesis. The bases in the FRA16B repeat are not randomly distributed but instead align to form a hairpin structure (Fig. 6). However, due to the presence of loops and due to a complete lack of C·G base pairs, the hairpin is predicted to be not very heat stable. Since the purine content of the FRA16B repeat is [sim]50%, it cannot form a triplex at physiological salt and pH. Further, since the guanine content of the FRA16B repeat is extremely low, it cannot form a tetraplex. These observations suggest that hairpin structures may facilitate the expansion of non-triplet repeat sequences. In addition, it appears that a further structural requirement of the hairpin may be the presence of `flexible' or `unstable' base pairs.
Figure 6 Proposed hairpin structure of a FRA16B repeat. The first 33 nt of the sequence shown above correspond to one FRA16B repeat unit. Ovals depict Watson-Crick base pairs. Lower case letters represent the positions of degenerate nucleotides, which can be either A or C.
The degree of flexibility may play an important role in determining how well a hairpin structure may participate in an expansion reaction. A hairpin containing flexible base pairs is anticipated to melt rapidly and/or rapidly incorporate additional triplet repeats into the hairpin, relative to one containing only Watson-Crick pairs. In the case of d(CCG)n and d(CGG)n, melting and refolding of the structures formed by these sequences are anticipated to be relatively slow compared with d(CTG)n and d(CAG)n hairpins. For example, incorporation of additional triplet repeats into a d(CCG)n hairpin structure would first require movement of the mismatched cytosines from the minor groove and into the helix, melting of the C·G base pairs, re-orientation of the C·G base pairs and, finally (and perhaps the slowest step), movement of the mismatched cytosines back into the minor groove. Melting and refolding of a d(CGG)n tetraplex would require a similar number of steps, the first of which (unfolding of the tetraplex into a hairpin) might be rate limiting. In contrast, the kinetics of melting and refolding of d(CTG)n and d(CAG)n hairpins would require fewer steps, since the mismatched bases are at least partially stacked in the helix. Thus, on the basis of the known biophysical properties of the d(CCG)n, d(CGG)n, d(CTG)n and d(CAG)n structures, it might be anticipated that d(CTG·CAG)n sequences would be more prone to expansion compared with d(CCG·CGG)n sequences. Evidence to support this hypothesis was obtained by Wells and colleagues, who found that the dominant triplet repeat expansion products in E.coli were d(CTG·CAG)n (61).
Hairpin structures: where?
Although the results of in vitro studies indicate that those triplet repeat sequences that are associated with human disease form hairpin structures in vitro, similar studies have not been performed in vivo. In the absence of such data, one can only speculate where hairpins might form. Triplet repeat expansion is best explained by formation of a hairpin on the lagging (23,26) daughter (64) strand (Fig. 7). Alternatively, the hairpin might be formed during transcription. Although the fate of this hypothetical hairpin is not known, an understanding of how it is processed may provide clues to mechanisms of triplet repeat expansion. One scenario is that due to its double-stranded character, hairpins (or slipped structures; 65) formed from certain triplet repeats are not recognized by proteins involved in the repair of single base mismatches or the repair of large single-stranded loops. Following synthesis of the daughter strands, the hairpin would be present on the leading strand template (Fig. 7) and subject to repair immediately prior to a second round of replication. Recent studies in E.coli have demonstrated that in addition to repair, hairpins that are placed on the leading strand in plasmid DNA also undergo unusual reactions. For instance, when the hairpin contains CTG repeats, expansion events are frequently observed. This result suggests that a component of the expansion process may involve DNA repair. When the hairpin contains CAG repeats, homologous recombination events are observed at a low ([sim]4 × 10-2) frequency (B.W., T.S., B.M. and M.M., unpublished results).
Figure 7 Hairpin structures and triplet repeat expansion.Models of triplet repeat expansion include formation of a hairpin structure on the lagging daughter strand (64,92) (dotted lines; top left). In theory, this hairpin might be present on the leading strand template in the following round of DNA replication (top right). Transformation of plasmids containing loops of CTG repeats on the leading strand template into E.coli cells (B.W., T.S., B.M. and M.M., unpublished results) leads to minor expansion of the triplet repeat region (dotted lines; bottom left), suggesting that expansion may occur during repair of the hairpin structures.
HYPERMETHYLATION OF EXPANDED d(CCG·CGG)n SEQUENCES
Expansion of d(CGG·CCG) repeats are coincident with their hypermethylation and fragility at their loci (66-68). High resolution methylation analysis of the FMR-1 gene has shown that all CpG dinucleotides within and surrounding the trinucleotide repeat were unmethylated in the DNA of normal transmitting male lymphoblasts and were methylated in affected male lymphoblasts (69).
Two theories regarding hymermethylation of expanded d(CGG·CCG) repeats will be discussed. The first is based on active methylation of hairpin structures containing d(CCG)n, while the second is based on lack of demethylation of the expanded sequences.
A theory based on active methylation of d(CCG)n hairpin sequences
To understand the active methylation theory of d(CCG)n, it is first necessary to review the substrate specificity of the human DNA (cytosine-5) CpG methyltransferase (MTase). The substrate for human CpG MTase consists of three bases
5[prime]-C1G2-3[prime]
3[prime]-X4C3-5[prime]
where the C at position 1 is the substrate for methylation and nucleotide X can be any base or can be missing (70). Methylation rates of human CpG MTase are higher when the nucleotide at position 4 does not form Watson-Crick H bonds with C (70-73). Mispair-induced methylation has also been reported for bacterial MTases HhaI and HpaII (74,75), which methylate the internal cytosines of GCGC and CCGG sequences, respectively. The DNA binding affinity of the these MTases correlate inversely with the stability of the target base pairs. Since flipping of the target base is a prerequisite for cytosine methylation (76,77; for a review see 78), structural analogs that mimic a flipped base are predicted to be ideal substrates for methylation.
Figure 8 Model of hypermethylation of expanded d(CCG)n·(CGG)n sequences. Diagram of the 5[prime]-region of the FMR-1 gene containing a (A) non-expanded and (B) expanded triplet repeat region. This figure illustrates anchoring of demethylating activity (depicted as a crane-like structure) to the region 5[prime] of the triplet repeats. Inability of the demethylase to remove 5mC residues from the distal expanded region is thought to result in gene inactivation, which, in turn, leads to remethylation of the entire triplet repeat and surrounding region. For examples of methylation of genes due to lack of expression see Bird (93).
These observations suggest that the methylatable cytosine in the 3 nt motif might be free to flip out of the helix and undergo rapid methylation in a CCG-containing hairpin in the a alignment, which contains a GpC base pair step. In support of this possibility, in vitro methylation studies have shown that human DNA CpG MTase rapidly methylates d(CCG)5-7 hairpin structures that contain GpC base pair steps. It has been proposed that hypermethylation of the triplet repeat region in the FMR-1 gene is a direct result of hairpin formation during DNA replication, followed by the action of human DNA CpG MTase (41,55,73)
A theory based on lack of demethylation
At specific stages during gametogenesis and embryogenesis, virtually the entire mouse and human genomes are demethylated by an active process that does not require DNA replication (79-85). In addition to these genome-wide events, tissue-specific genes are actively demethylated concomitant with the process of cellular differentiation (86-88). Recently, Cedar, Razin and colleagues have shown that demethylation involves replacement of the p5mCp in a p5mCpG dinucleotide with an unmethylated pCp (89). During this process, the methylated cytosine is transformed to an RNase-sensitive form. The authors proposed that demethylation of the promoter regions of housekeeping genes (such as FMR-1) are brought about through interactions anchored at the sites of CpG islands. In their model, demethylation would occur only as long as the demethylating activity was tethered to the specific locus on the DNA and only with the defined region having access to the demethylation activity (89). On the basis of this model, one might predict that, due to spatial or kinetic constraints, expansion of d(CCG·CGG)n sequences would interfere with the ability of the protein to demethylate the cytosines in and around the triplet repeat region, thus leading to gene hypermethylation (Fig. 8).
SUMMARY
Of the six complementary strands associated with TREDs, five form hairpin structures. Although the molecular basis of the expansion process is unknown, it most likely involves formation of flexible hairpin structures, which may promote DNA recombination and/or interfere with the progression of enzymes involved in DNA replication, such as DNA helicases. The correlation between minisatellite amplification and the ability of a sequence to adopt a flexible hairpin conformation likely precludes amplification of unstructured triplet repeat sequences or sequences able to adopt more rigid hairpin conformations [e.g. d(CG)n, d(AT)n or d(GATC)n].
ACKNOWLEDGEMENTS
I thank Drs X.Gao (University of Houston), G.Gupta (Los Alamos), S.Mirkin (University of Illinois), S.Smith (City of Hope) and K.Usdin (NIH) for helpful discussions and for sharing many of their results prior to publication. I thank Dr Ulrich Melcher (Oklahoma State) for his continued guidance and for critical review of this manuscript and Mellisa Christy for her excellent assistance in preparing the figures. I am deeply indebted to my friend and colleague I.S.Haworth (University of Southern California) for providing tutorials on DNA structure. Finally, I dedicate this review to my late father, Henry J.Mitas.
REFERENCES
1. Ashley,J.C.T. and Warren,S.T. (1995) Annu. Rev. Genet., 29, 703-728.
2. Huntingon's Disease Collaborative Research Group. (1993) Cell, 72, 971-983.
M. Brandstrom and H. Ellegren Genome-wide analysis of microsatellite polymorphism in chicken circumventing the ascertainment bias
Genome Res.,
June 1, 2008;
18(6):
881 - 887.
[Abstract][Full Text][PDF]
A.-M. Patch and S. J. Aves Fingerprinting fission yeast: polymorphic markers for molecular genetic analysis of Schizosaccharomyces pombe strains
Microbiology,
March 1, 2007;
153(3):
887 - 897.
[Abstract][Full Text][PDF]
S. J. Gray, J. Gerhardt, W. Doerfler, L. E. Small, and E. Fanning An Origin of DNA Replication in the Promoter Region of the Human Fragile X Mental Retardation (FMR1) Gene
Mol. Cell. Biol.,
January 15, 2007;
27(2):
426 - 437.
[Abstract][Full Text][PDF]
B. T. Farrell and R. S. Lahue CAG{middle dot}CTG repeat instability in cultured human astrocytes
Nucleic Acids Res.,
September 11, 2006;
34(16):
4495 - 4505.
[Abstract][Full Text][PDF]
H. Kim and D. M. Livingston A High Mobility Group Protein Binds to Long CAG Repeat Tracts and Establishes Their Chromatin Organization in Saccharomyces cerevisiae
J. Biol. Chem.,
June 9, 2006;
281(23):
15735 - 15740.
[Abstract][Full Text][PDF]
K. Walter, G. Warnecke, R. Bowater, W. Deppert, and E. Kim Tumor Suppressor p53 Binds with High Affinity to CTG{middle dot}CAG Trinucleotide Repeats and Induces Topological Alterations in Mismatched Duplexes
J. Biol. Chem.,
December 30, 2005;
280(52):
42497 - 42507.
[Abstract][Full Text][PDF]
M. Napierala, A. Bacolla, and R. D. Wells Increased Negative Superhelical Density in Vivo Enhances the Genetic Instability of Triplet Repeat Sequences
J. Biol. Chem.,
November 11, 2005;
280(45):
37366 - 37376.
[Abstract][Full Text][PDF]
J. Subramanian, S. Vijayakumar, A. E. Tomkinson, and N. Arnheim Genetic Instability Induced by Overexpression of DNA Ligase I in Budding Yeast
Genetics,
October 1, 2005;
171(2):
427 - 441.
[Abstract][Full Text][PDF]
S. Bhattacharyya and R. S. Lahue Srs2 Helicase of Saccharomyces cerevisiae Selectively Unwinds Triplet Repeat DNA
J. Biol. Chem.,
September 30, 2005;
280(39):
33311 - 33317.
[Abstract][Full Text][PDF]
V. Handa, H. J. C. Yeh, P. McPhie, and K. Usdin The AUUCU Repeats Responsible for Spinocerebellar Ataxia Type 10 Form Unusual RNA Hairpins
J. Biol. Chem.,
August 12, 2005;
280(32):
29340 - 29345.
[Abstract][Full Text][PDF]
S. Amrane, B. Sacca, M. Mills, M. Chauhan, H. H. Klump, and J.-L. Mergny Length-dependent energetics of (CTG)n and (CAG)n trinucleotide repeats
Nucleic Acids Res.,
July 21, 2005;
33(13):
4065 - 4077.
[Abstract][Full Text][PDF]
R. D. Wells, R. Dere, M. L. Hebert, M. Napierala, and L. S. Son Advances in mechanisms of genetic instability related to hereditary neurological diseases
Nucleic Acids Res.,
July 8, 2005;
33(12):
3785 - 3798.
[Abstract][Full Text][PDF]
Y. Lian and H. R. Garner Evidence for the regulation of alternative splicing via complementary DNA sequence repeats
Bioinformatics,
April 15, 2005;
21(8):
1358 - 1364.
[Abstract][Full Text][PDF]
T. Boby, A.-M. Patch, and S. J. Aves TRbase: a database relating tandem repeats to disease genes for the human genome
Bioinformatics,
March 15, 2005;
21(6):
811 - 816.
[Abstract][Full Text][PDF]
L. M. Chi and S. L. Lam Structural roles of CTG repeats in slippage expansion during DNA replication
Nucleic Acids Res.,
March 14, 2005;
33(5):
1604 - 1617.
[Abstract][Full Text][PDF]
V. I. Hashem, M. J. Pytlos, E. A. Klysik, K. Tsuji, M. Khajav, T. Ashizawa, and R. R. Sinden Chemotherapeutic deletion of CTG repeats in lymphoblast cells from DM1 patients
Nucleic Acids Res.,
December 1, 2004;
32(21):
6334 - 6346.
[Abstract][Full Text][PDF]
R. Dere, M. Napierala, L. P. W. Ranum, and R. D. Wells Hairpin Structure-forming Propensity of the (CCTG{middle dot}CAGG) Tetranucleotide Repeats Contributes to the Genetic Instability Associated with Myotonic Dystrophy Type 2
J. Biol. Chem.,
October 1, 2004;
279(40):
41715 - 41726.
[Abstract][Full Text][PDF]
<
CiteULike 
Connotea 
Del.icio.us What's this?
