Instability and Chromatin Structure of Expanded Trinucleotide Repeats

Abstract

Trinucleotide repeat expansion underlies at least 17 neurologic diseases. In affected individuals, the expanded locus is characterized by dramatic changes in chromatin structure and in repeat tract length. Interestingly, recent studies show that several chromatin modifiers, including a histone acetyltransferase, a DNA methyltransferase, and the transcription factor CTCF can modulate repeat instability. Here, we propose that the unusual chromatin structure of expanded repeats directly impacts their instability. We discuss several potential models for how this might occur, including a role for DNA repair-dependent epigenetic reprogramming in increasing repeat instability, and the capacity of epigenetic marks to alter sense and antisense transcription, thereby affecting repeat instability.

Long Triplet Repeats are Unstable and Cause a Number of Degenerative Disorders

In 1991, a novel type of mutation—the expansion of trinucleotide repeats—was shown to cause two human neurological diseases: fragile X syndrome (FRAXA) and spinal and bulbar muscular atrophy (SBMA)^{1, 2}. To date, 17 such diseases have been identified³. Normal individuals typically harbour fewer than 30 repeats, whereas patients can carry from 35 to several thousand repeats. The disease incidence can be as common as 1 in 4000 males for FRAXA and as rare as 1 in 50 000 for Friedreich ataxia (FRDA)³. Large tracts of trinucleotides can cause disease in several ways: by affecting gene expression, producing a toxic RNA species, or altering the function of the resultant protein³ (Table 1). Changes that impair gene expression or protein function are recessive (e.g. FRDA), whereas those that produce toxic RNAs or proteins are dominant (e.g. myotonic dystrophy (DM1) and Huntington disease (HD)).

Table 1.

Disorders caused by expansion of trinucleotide repeats^a

Disease	Gene Product	Repeat type	Pathogenic category	Somatic instability
Dentatorubro-pallidoluysian atrophy (DRPLA)	Atrophin 1	CAG	GOFb	moderate
Fragile X syndrome (FXS)	Fragile site mental retardation 1	CGG	LOFc	Low
Fragile X syndrome (FRAXE)	Fragile site mental retardation 2	CGG	LOF	NDd
fragile-X–associated tremor/ataxia syndrome (FXTAS)	Fragile site mental retardation 1	CGG	Toxic RNA	Low
Friedreich Ataxia (FRDA)	Frataxin	GAA	LOF	moderate
Huntington disease (HD)	Huntingtin	CAG	GOF	moderate
Huntington Disease Like 2 (HDL2)	junctophilin-3	CAG	Unknown	ND
Myotonic dystrophy type 1 (DM1)	DMPK	CTG	Toxic RNA	high
Spinal and bulbar muscular atrophy (SBMA)	Androgen receptor	CAG	GOF	low
Spinocerebellar ataxia type 1 (SCA1)	Ataxin-1	CAG	GOF	moderate
Spinocerebellar ataxia type 2 (SCA2)	Ataxin-2	CAG	GOF	moderate
Spinocerebellar ataxia type 3 (SCA3)	Ataxin-3	CAG	GOF	moderate
Spinocerebellar ataxia type 6 (SCA6)	CACNAIA	CAG	GOF	ND
Spinocerebellar ataxia type 7 (SCA7)	Ataxin-7	CAG	GOF	moderate
Spinocerebellar ataxia type 8 (SCA8)	C10orf2	CAG	Unknown	ND
Spinocerebellar ataxia type 12 (SCA12)	PPP2R2B	CAG	Unknown	ND
Spinocerebellar ataxia type 17 (SCA17)	TBP	CAG	GOF	limited

^ahe data were obtained from^{3, 4, 83}

^bgain of function

^closs of function

^dND: not determine

Analyses of repeat tracts in patients and animal models have revealed that trinucleotide repeats often show increases in repeat number (expansions) as well as decreases (contractions) in both germline and somatic tissues. In addition, the rate of instability varies during development and between tissues within a single organism. This variability implies multiple mechanisms of instability, an idea that is supported by the identification of a number of stage specific modifiers of repeat instability in mouse models (Figure 1). The bias of instability towards expansion in the germline and the embryo accounts for the increased severity of disease symptoms in subsequent generations. This phenomenon was coined ‘anticipation’ long before its molecular basis was discovered. In affected individuals, ongoing, expansion-biased repeat instability in somatic tissues, especially those associated with the disease phenotype, might decrease the age of onset of symptoms and accelerate disease progression⁴.

Figure 1.

Modulators of trinucleotide repeat instability. A number of gene products have been shown to modify trinucleotide repeat instability throughout mouse development^{42, 76-81}. Msh2 and Msh3 are required for instability in all stages investigated for CAG/CTG repeats^{46, 76-79}. Postmeiotic segregation increased 2 (Pms2)⁸⁰ as well as Ogg1⁴² are specifically involved in somatic repeat instability. Dnmt1, on the other hand, protects against expansions in an expanded CAG model only in the germlines ⁹. Ataxia telangiectasia and Rad3 related (Atr), a signaling kinase, prevents expansion in the female germline and age-dependent instability in somatic tissues of CGG repeats⁸¹. It should be noted that Msh6 null mice have mild triplet repeat instability phenotypes in mice, but this observation has been generally dismissed due to its effect on the stability of Msh2 and Msh3 ^{77, 79, 82}.

Instability of disease-causing trinucleotide repeats appears to be triggered by the folding of repeat sequences into abnormal secondary structures, including hairpins, slipped-strand structures, triplexes, and quadruplexes⁴. The mishandling of these aberrant structures by the DNA repair machinery is thought to cause changes in tract length.

Genetic studies in bacteria, yeast, flies, mammalian cells, and mice have revealed an astounding assortment of proteins whose mutation or knockdown significantly influences repeat stability⁴. Paradoxically, this wealth of possibilities has slowed the discovery of the major mechanisms that alter repeat stability.

To add to this complexity, the instability of repeats depends on internal influences, including the type of the repeat, its purity, and its length, as well as on the surrounding genomic context. Indeed, repeat tracts are typically much more unstable in mouse models generated through the insertion of large genomic fragments than in models that carry cDNAs, suggesting that the surrounding sequence influences repeat stability⁵. Examples of potential cis-acting elements include origins of replication⁴, transcription factor binding sites⁶, sense and antisense promoters⁷, neighboring GC-rich sequences⁸, and the local chromatin environment^{9, 10}.

DNA is embedded in chromatin, which influences their metabolism (Box 1). Chromatin packing is influenced by histone tail modifications and DNA methylation at CpG dinucleotides. Recently, several studies have carefully catalogued the epigenetic marks associated with expanded GAA repeat tracts, showing that they are buried in heterochromatin^{6, 11, 12}. In addition, several publications within the last two years have unearthed evidence that chromatin modifying enzymes and epigenetic marks contribute to repeat instability^{9, 10, 13}. No cause-effect experiment has been done, however, to directly measure the effect of altering local chromatin structure on repeat instability. Therefore, we acknowledge at the outset that it is not yet clear whether the epigenetic status of repeats is a principal driver of repeat instability, or merely a by-product of repeat expansion.

Box 1. Chromatin structure and DNA metabolism.

Approximately 146bp of DNA is wrapped around a histone octamer to form the basic subunit of chromatin: the nucleosome. Chromosomes, however, are packaged into higher order structures, a process influenced by DNA methylation and covalent modification of histones. These epigenetic marks tend to be dynamic and vary between cells, between loci within a cell, and between different developmental stages at same locus³². Chromatin structure is broadly categorized into euchromatin and heterochromatin. Euchromatin is a looser structure that contains a high proportion of histone H3 and H4 molecules with acetylated lysine residues in their N-terminal tails. Heterochromatin is more densely packaged and is enriched in H3K9me and methylation of cytosine residues at CpG dinucleotides. Whether a given region is euchromatic or heterochromatic does not depend on any individual histone modification, but rather on collective influence of many different marks.

Chromatin regulates all DNA transactions including transcription. Transcription initiation and elongation require a complex interplay of nucleosome remodelling and chromatin modifying enzymes⁷⁰. NER and BER proteins are often unable to bind nucleosomes in the absence of accessory proteins that expose damaged DNA for recognition^{71, 72}. Origins of replication in higher eukaryotes are chiefly defined by their chromatin context, which needs to be maintained after the passage of the replication fork⁷³. DNA damage signalling in heterochromatin has specific requirements and kinetics that differ from euchromatic regions⁷⁴. Finally, the repair of double-strand breaks is accompanied by an intricate interplay of chromatin remodelling enzymes, nucleosomes eviction and modification, which allow accurate signalling, repair, and recovery from damage⁷⁵. Clearly, these DNA transactions are heavily dependent on chromatin structure, and each of these processes affects triplet repeat instability.

Here, we summarize what is known about the chromatin environment of expanded repeat tracts and highlight the correlation between the epigenetic changes that occur during development and the timing of triplet repeat instability. To spark discussion, we speculate about how local chromatin structure might drive repeat instability throughout development.

The Chromatin Environment of Expanded Triplet Repeats

Several studies have identified alterations in DNA methylation, histone modification, and chromatin structure around expanded repeat tracts that are not present at the corresponding wild type alleles (Table 2). Expanded CGG repeats at the fragile X mental retardation 1 (FMR1) locus in FRAXA patients provided the first indication that such repeats are associated with heterochromatin marks. The absence of FMR1 expression, which is responsible for the mental retardation phenotype in FRAXA, is caused by extensive DNA methylation within the CGG repeat, as well as in the flanking DNA^14-16. More recently, chromatin immunoprecipitation analyses of specific modified histones in the FMR1 5′ region near the expanded repeat revealed high levels of other repressive marks, including methylated histone 3 at lysine 9 (H3K9me)¹⁷, and lower levels of two marks typically encountered in euchromatic regions: acetylated (Ac) H3 and H4¹⁸. These data indicate that expanded CGG repeats, unlike their normal-length counterparts, carry epigenetic marks typical of heterochromatin, and display diminished levels of euchromatic marks.

Table 2.

Epigenetic changes at expanded repeats compared to wild type alleles

Disease	Repeat	Length	Species	Tissue	Positiond	Heterochromatic markse	Euchromatic markse	References
FRA12A	CGG	270- 300	Human	Transformed lymphoblast	U, P	mCpG ↑	ND	84
FRAXA	CGG	50-900	Human	Various tissuesc	U	mCpG ↑	ND	14, 15, 24, 25, 50, 52, 53, 55, 56, 85
		50- 200+a		Blood	U, W	mCpG ↑	ND	15, 24
		NRa,b		Blood	U, W, D	mCpG ↑	ND	86
		NRb		Transformed lymphoblast	P	ND	H3K4me2 ↓ H3K9Ac ↓ H4K16Ac –	69
		NR		Transformed lymphoblast	D	ND	H3K4me2 ↓ H3K9Ac ↓ H4K16Ac ↓	69
		230- 530a		Transformed lymphoblast	U, W, D	mCpG ↑ H3K9me2 ↑	H3K4me2 ↓ H3Ac ↓ H4Ac ↓	16-18, 86
FRAXE	CGG	100- 1500	Human	Blood	U,D	mCpG ↑	ND	87-90
		130- 1500		Fibroblast	D	mCpG ↑	ND	89, 91
		180- 733		Transformed lymphoblast	D	mCpG ↑	ND	88, 92
DM1	CTG	NR	Human	Fibroblast	U	mCpG ↑	ND	93
				Fibroblast	U, W, D	H3K9me3 ↑	H3K4me3 ↓	21
		1000- 1800		Blood	U, W, D	mCpG ↑	ND	20
				Fetal dura matter	U, W, D	mCpG ↑	ND	20
				Fetal skeletal muscles	U, W, D	mCpG ↑	ND	20
SCA1	CAG	143	Mouse	Testis	P	H3 K9me2 -	ND	9
					U	mCpG - H3K9me2 -	H3Ac -	9, unpublished data
					D	mCpG ↑ H3K9me2 -	H3Ac -	9, unpublished data
FRDA	GAA	650- 750	Human	Brain	P	mCpG – H3K9me2 ↑ H3K9me3 ↑	H3K9Ac ↓ H3K14Ac – H4K5Ac – H4K8Ac ↑ H4K12Ac – H4K16Ac –	12
					U	mCpG ↑ H3K9me2 ↑ H3K9me3 ↑	H3K9Ac ↓ H3K14Ac – H4K5Ac – H4K8Ac – H4K12Ac – H4K16Ac –	12
					D	mCpG ↓ H3K9me2 ↑ H3K9me3 ↑	H3K9Ac ↓ H3K14Ac – H4K5Ac ↓ H4K8Ac ↓ H4K12Ac ↓ H4K16Ac ↓	12
				Cerebellum	U	mCpG ↑	ND	12
				Heart	P, U	mCpG ↑	ND	12
					D	mCpG ↓	ND	12
		650- 1030a		Transformed lymphoblasts	P	ND	H3K9Ac – H3K14Ac – H4K5Ac – H4K8Ac ↓ H4K12Ac ↓ H4K16Ac –	94
					U	mCpG ↑ H3K9me1 ↓ H3K9me2 ↑ H3K9me3 ↑	H3K9Ac ↓ H3K14Ac – H4K5Ac ↓ H4K8Ac ↓ H4K12Ac ↓ H4K16Ac ↓	6, 12, 94
					D	ND	H3K9Ac ↓ H3K14Ac – H4K5Ac ↓ H4K8Ac ↓ H4K12Ac ↓ 4K16Ac ↓	94
		580		HEK293	U, D	H3K9me3 ↑	H3K9Ac ↓ H3K14Ac – H4K5Ac ↓ H4K8Ac –	11
		90-120	Mouse	Brain	P	mCpG ↑ H3K9me2 ↑ H3K9me3 ↑	H3K9Ac – H3K14Ac – H4K5Ac – H4K8Ac – H4K12Ac – H4K16Ac –	14
					U	mCpG ↑ H3K9me2 ↑ H3K9me3 ↑	H3K9Ac – H3K14Ac – H4K5Ac – H4K8Ac – H4K12Ac – H4K16Ac –	12
		230			U	H3 K9me3 ↑	H3K14Ac ↓ H4K5Ac – H4K8Ac ↓ H4K16Ac ↓	95
		90-120			D	mCpG – H3K9me2 ↑ H3K9me3 ↑	H3K9Ac ↓ H3K14Ac – H4K5Ac – H4K8Ac – H4K12Ac – H4K16Ac ↑	12
		90-120		Cerebellum	U	mCpG ↑	ND	12
				Heart	P	mCpG ↓	ND	12
					U	mCpG ↑	ND	12
					D	mCpG –	ND	12

^aIncludes unknown repeat length alleles obtained from affected individuals or premutation

^bNR, repeat length not reported

^cA variety of somatic tissues from affected individuals was analyzed.

^dP, promoter; U, upstream; D, downstream; W, within the repeat tract.

^e↑, enriched compared to wild type levels; ↓, depleted; –, not changed; ND, not determined

Expanded CTG repeats elicit similar effects on surrounding chromatin structure in cell lines derived from patients with a severe form of myotonic dystrophy (DM1). These cells harbor over 1000 CTGs in the 3′ UTR of the dystrophia myotonica protein kinase (DMPK) gene that are largely resistant to digestion with DNaseI and methylation sensitive restriction enzymes^{19, 20}, properties typical of heterochromatin with high levels of DNA methylation. In addition, the expanded locus contains histones enriched in H3K9me and depleted in H3Ac²¹. Thus, at least in the case of congenital DM1 cells, highly expanded CTG repeats, like expanded CGG repeats, appear to assume a heterochromatic conformation.

Expansion-induced changes in chromatin structure have been most extensively studied in the first intron of frataxin gene, where the expansion of a GAA repeat in the first intron gives rise to FRDA. Similar to CGG and CTG repeats, the region upstream of the expanded GAA repeat in several somatic tissues, is more highly methylated at CpG sites than it is at wild type alleles^{6, 12}. Moreover, decreased H3Ac and H4Ac levels and increased H3K9me levels surround expanded frataxin GAA repeats ^{6, 11, 12}.

It is not known what triggers or maintains heterochromatin at expanded triplet repeats²². In all cases, the repeat tract length must first reach a certain threshold. When the repeat nears this threshold, heterochromatin forms and spreads in a manner reminiscent of position effect variegation, the stochastic spreading of heterochromatin to adjacent euchromatic regions²³. For instance, some individuals carrying FRAXA premutation alleles (harboring ~60-200 CGG repeats) display mosaic patterns of DNA methylation in different tissues²⁴, whereas the full expansion (>200 CGGs) is associated with complete FMR1 promoter methylation²⁵. The idea that repeats can induce heterochromatin spreading was directly tested in mice²⁶. Whereas arrays of reporter genes were silenced only when they integrated near a heterochromatic region²⁶, similar arrays, harboring 192 CTG or 200 GAA repeats, were silenced regardless of the transgene’s genomic location²⁶. Thus, CGG, CTG, and GAA repeats, independent of their genomic location, seem to direct formation and spreading of heterochromatinn.

Repeat Instability during Early Embryogenesis

Several studies in FRAXA individuals indicate that instability can occur during early embryogenesis. Some individuals harbor two major FMR1 CGG repeat lengths—derivatives of the same allele—at high frequency in every examined tissue, suggesting that repeat instability arose during the first zygotic cell division^{27, 28}. This possibility is supported by the existence of monozygotic twins who carry different allele lengths^{27, 29}.

In support of these conclusions from patient studies, Savouret et al provided convincing evidence for instability of DM1-associated CTG repeats during early embryogenesis³⁰. They examined the effects of knocking out the mismatch repair (MMR) gene Msh2 on intergenerational repeat instability in mice that carried >300 DMPK CTG repeats. Msh2 is central to the recognition of mismatches in the DNA³¹ and binds hairpins formed by CTG repeats⁴. Savouret et al observed ongoing instability in the germline of the transmitting parent, pointing to germline development as a period prone to repeat instability. However, if the effect of Msh2 deletion were solely confined to the parent’s germline, the genotype of the embryo should make no difference. This was not the case: whereas Msh2^+/− offspring of Msh2^+/− parents harbored mostly CTG expansions, Msh2^−/− littermates were strongly biased toward contractions. Similarly, in a cross between Msh2^−/− mice carrying the expanded allele and Msh2^+/− mice, the Msh2^−/− offspring had mostly contractions, but their Msh2^+/− littermates displayed about equal proportions of contractions and expansions. This dependence on genotype indicates that some repeat instability must occur in the embryo. In addition, the lack of multiple repeat lengths in individual newborn mice indicates that embryonic instability must occur very early during development, probably just after fertilization³⁰.

Vertebrate embryogenesis entails extensive epigenetic changes that are essential for resetting the developmental potential of the zygote³². This process is called reprogramming. In the mouse, the paternal chromatin is remodeled in the egg by histone exchange and acetylation and by elimination of DNA methylation via an active process that remains poorly understood^{32, 33}. By contrast, the maternal genome is passively demethylated through replication during the first 3.5 days post fertilization. The methylation pattern is subsequently restored by de novo methyltransferases and maintained by Dnmt1 (DNA (cytosine-5-) methyltransferase 1). The temporal overlap of chromatin structure reprogramming and trinucleotide repeat instability raises the possibility of a cause-and-effect relationship between the two processes.

To probe the effects of reprogramming on repeat instability, Gorbunova et al attempted to model, in cultured cells, the demethylation that occurs during embroygenesis³⁴. Treatments that reduced genome-wide DNA methylation levels and Dnmt1 activity stimulated CAG repeat contractions up to 1000-fold in a selection assay in mamalian cells and promoted expansions in unselected DM1 cell lines³⁵. These data suggest that DNA demethylation provokes repeat instability, but the mechanism is unclear. Several possibilities have been eliminated, however, including homologous recombination, nonhomologous end joining, and transcription through the repeat tract³⁶. Similar effects of CpG methylation have also been noted in bacteria: methylated CGG repeats and CAG repeats surrounded with methylated sequences are more stable than their unmethylated counterparts³⁷. These studies provide a strong experimental link between DNA demethylation and repeat instability.

DNA demethylation during reprogramming could trigger repeat instability by engaging DNA repair pathways. Active demethylation in the paternal genome during early embryogenesis has been proposed to involve DNA repair via cytosine deaminases of the activation induced cytosine deaminase (Aid)/apolipoprotein B editing complex (Apobec1) family^{38, 39} in combination with the glycosylase Mbd4 (methyl-CpG binding domain protein 4) and the nucleotide excision repair (NER) component Gadd45a (growth arrest and DNA-damage-inducible alpha)^{39, 40}. Aid and Apobec1, which are expressed in pluripotent cells, can deaminate 5-methylcytosine to thymine, creating T-G mismatches^{38, 39}. Removal of T and its replacement by C could occur via MMR, base excision repair (BER), or even NER. Because these repair pathways introduce nicks in the DNA, and because nicks are associated with instability⁴¹, we speculate that reprogramming might cause destabilization of repeats by generating nicks at nearby methylated CpGs. Thus, methylated CpG sites might serve as signals to recruit DNA repair factors to expanded repeat tracts, leading to further instability (Figure 2i). This scenario could account for the MMR dependence of repeat instability in the early embryo³⁰. Although no glycosylase is known to cause trinucleotide repeat instability in the embryo, there is precedence for glycosylase-induced instability in somatic tissue: removing 7,8-dihydro-8-oxo-guanine-DNA glycosylase 1 (Ogg1) greatly diminishes somatic instability in an HD mouse model⁴².

Figure 2.

Models for chromatin-dependent trinucleotide repeat instability. Expanded trinucleotide repeats (orange) are embedded into heterochromatin, which is enriched in H3K9me and DNA methylation at CpG sites (blue circles). Here, we present three speculative mechanisms by which chromatin environment might impact repeat instability. We imagine that these models might act in parallel at different loci, tissues, or developmental stages. i) Heterochromatin as a recruitment signal during reprogramming. During germline development and early embryogenesis, DNA is demethylated, probably through a DNA repair mechanism. DNA methylation at CpGs leads to recruitment of DNA repair factors (green circles), creating nicks that eventually lead to repeat instability. ii) Chromatin structure as a cis-acting factor in triplet repeat instability. Loose euchromatin could facilitate the access of DNA repair factors to aberrantly structured repeat tract. Promoting access would enhance gratuitous repair of the repeat tract and therefore instability (open circles are unmethylated CpG sites). iii) Transcription in triplet repeat instability. Sense, antisense, or convergent transcription through an expanded repeat could promote secondary DNA structure formation and therefore instability.

Repeat Instability during Germline Development

Instability in the germline and during gametogenesis is well documented in humans and mice. FRAXA patients, for example, show a striking parent-of-origin effect. Expanded alleles are often observed in children whose asymptomatic mothers harbor premutation alleles. Male fetuses from these mothers carrythe full mutation in all tested tissues except fetal testes, in which contractions were observed⁴³. Similar analyses of female fetuses support the conclusion that CGG expansions occur in the female germline⁴³.

Several studies have addressed the timing of germline instability. In HD patients, a substantial fraction of all instability of the CAG tract—mostly expansions—is generated pre-meiotically⁴⁴. Similarly, in a DM1 mouse model, repeat instability increased with age, and all instability was already present in spermatogonia, a cell type found at the earliest stage of spermatogenesis, and no further increase was apparent at later stages of gametogenesis⁴⁵. In contrast, using a mouse model for HD, Kovtun and McMurray⁴⁶ detected instability in males only after the development of spermatozoa. The basis for the difference is unclear.

Analogous studies in the female germline of spinocerebellar ataxia type 1(SCA1) mouse models found that moderate to large contractions dominated transmissions from the maternal parent and the size of the contractions increased with age^{9, 47, 48}. Thus, CAG repeat instability arises during both spermatogenesis and oogenesis..

As is the case during embryogenesis, instability of repeat tracts in the male and female germlines coincides with epigenetic changes to the genome. In the mouse, primordial germ cells are demethylated between days 8.5 and 13.5 of embryonic development, with the bulk of DNA demethylation occurring between E11.5 and E12.5, concurrent with substantial changes in histone modification and content³². DNA remethylation in the male germline starts at E15³². In the female germline, the genome is remethylated only after birth, during oocyte growth ³². Thus, epigenetic changes occur throughout germline development, at all stages where repeat instability has been identified.

The rapid loss of methylation in the developing germline suggests that demethylation is active and possibly dependent on DNA repair, as proposed for embryogenesis⁴⁹. But germline reprogramming has not been studied as extensively as early embryogenesis, where evidence for a role of DNA repair in reprogramming is rapidly accumulating³³. Nonetheless, we suggest that a spike in DNA repair activity near the repeat tract during reprogramming might account for the increased instability (Figure 2i). Alternatively, as reprogramming decreases the density of heterochromatic regions, DNA repair enzymes might gain freer access to the repeat tract, leading to enhanced recognition of secondary structures, gratuitous repair, and subsequent instability (Figure 2ii).

To probe the effects of DNA methylation on repeat stability, we removed one copy of Dnmt1 in a SCA1 mouse model carrying 143 CAG repeats at the endogenous locus⁹. Both male and female Dnmt1^+/− SCA1 mice passed an expansion to their progeny 3 to 4 times more frequently than did Dnmt1^+/+ SCA1 mice. In addition, the effects in Dnmt1^+/− mice were confined to the germline; repeat instability was not observed in the early embryo or in somatic tissues⁹.

Analyses of testes and ovaries from Dnmt1^+/− SCA1 mice did not reveal any global differences in DNA methylation of repetitive elements relative to Dnmt1^+/+ SCA1 mice⁹, suggesting that Dnmt1 deficiency had little effect on genome-wide methylation. At CpG sites 17 and 20bp upstream of the repeat tract, however, ovaries displayed significantly reduced methylation levels, whereas testes showed significantly elevated levels. In the same region, testes of Dnmt1^+/− mice also displayed variegated H3K9me2 levels, suggesting that the local chromatin environment underwent alterations. Collectively, these observations suggest that Dnmt1 deficiency enhances an expansion-biased process early on in the development of both the male and female germlines, at a time overlapping, and perhaps perturbing, the period of hypomethylation associated with reprogramming.

Studies of intergenerational instability in a fly model of CAG repeat instability suggest that histone acetylation is also involved¹³. Haploinsufficiency of CREB-binding protein (CBP), which encodes a histone acetyltransferase, increases triplet repeat instability¹³. By contrast, treatment with the histone deacetylase inhibitor, trichostatin A (TSA), decrease triplet repeat instability¹³. Although genome-wide levels of acetylated histones differed between the CBP-mutant flies or TSA-treated flies and their wild type counterparts, changes in histone acetylation were not observed near the repeat tract, nor was transcription through the repeat affected¹³. These results implicate histone acetylation as a modifier of triplet repeat instability but the effect, in this case, appears to be indirect.

Repeat Instability in Somatic Tissues

Somatic instability of CGG repeats, which, unlike CAGs and GAAs, can be methylated, is low in patients harboring the full mutation (>200 CGGs)^{50, 51}. By contrast, FRAXA males, who express FMR1 although they harbor a large expansion, have unmethylated repeats with a high degree of instability in somatic tissues^{52, 53}. One such individual displayed instability in some tissues and not in others, a pattern that correlated perfectly with methylation status of the repeat⁵². Cultured cells from FRAXA patients show the same absolute correlation: methylated CGG repeats are stable, whereas unmethylated repeats are unstable^54-56. Importantly, in cultured primate cells, DNA methylation of plasmids containing CGG repeats stabilized the repeat tract, specifically implicating DNA methylation within or near repeat tracts as a key modifier of triplet repeat instability³⁷.

In addition to epigenetic modifiers, transcription has emerged as an important player in triplet repeat instability. Transcription through a repeat tract can stimulate repeat instability in human cells and in fly models of CAG repeat diseases⁵⁷. Lin et al ^{7, 58} showed that transcription through a repeat of 95 CAGs caused substantial contractions of the repeat tract⁷. Moreover, transcription-induced destabilization required MMR and transcription-coupled NER (TC-NER)⁵⁷. Thus, heterochromatin might reduce repeat instability by lowering the rate of transcription through the repeat tract, thereby decreasing the potential for TC-NER (Figure 2iii). The absence of transcription through the highly expanded CGG repeats at the FMR1 locus could account for their surprising stability in somatic tissues.

Many repeat loci, however, are more unstable than their shorter counterparts even though they are transcribed at reduced levels⁵⁷. For these loci, another mechanism must be at work. We suggest that this mechanism might be antisense transcription. Indeed, a surprisingly high fraction of human genes are associated with antisense transcripts, including most wild type alleles of triplet repeat-associated genes⁵⁹ (Table 3). At the FMR1 locus, heterochromatin appears to shut-down both sense and antisense transcription⁶⁰. By contrast, at the SCA8 and DM1 loci, the levels of antisense transcription are higher on the expanded alleles^{21, 61}. At least at the DMPK locus, high levels of antisense transcription are associated with heterochromatin marks²¹.

Table 3.

Sense and antisense transcripts at disease loci with normal length repeats^a

Disease	Repeat	Gene	Cell Lineb
			PBMC		HCT116		Jurkat		MaPaC		MRC5
			Sc	ASd	S	AS	S	AS	S	AS	S	AS
DM1	CTG	DMPK	34	0	35	2	72	13	35	1	307	11
DRPLA	CAG	ATN1	31	5	67	2	113	19	75	1	455	3
FRAXA	CGG	FMR1	69	0	16	1	98	0	9	0	35	0
FRAXE	CCG	AFF2	10	8	2	4	1242	9	2	4	9	7
FRDA	GAA	FXN	16	0	16	1	25	0	13	0	32	2
HD	CAG	HTT	569	6	223	10	879	15	211	2	672	10
HDL2	CTG	JPH3	5	41	2	15	16	92	0	9	8	31
SBMA	CAG	AR	7	1	3	0	1	0	1	0	91	2
SCA1	CAG	ATXN1	758	15	107	8	212	12	6	2	1458	22
SCA2	CAG	ATXN2	97	5	60	1	235	1	42	1	325	5
SCA3	CAG	ATXN3	36	2	24	4	77	6	3	2	37	6
SCA6	CAG	CACNA1A	9	11	4	8	11	14	11	2	207	13
SCA7	CAG	ATXN7	380	9	107	2	473	11	26	3	327	9
SCA8	CTG	C10orf2	11	0	41	2	19	1	16	0	8	1
SCA12	CAG	PPP2R2B	232	13	8	4	11	9	2	2	11	7
SCA17	CAG	TBP	14	1	5	0	30	1	4	0	13	0

^aData compiled from⁵⁹. In these experiments, RNA was isolated from 6 different cell lines and subjected to deamination to differentiate between sense and antisense strands. cDNAs generated from the deaminated transcripts were deep sequenced⁵⁹. The numbers reported in the table refer to the number of different sequencing reads that mapped to a given transcript. We highlighted the cases where the number of antisense sequences found for a gene amounts to at least 10% of the number of sense sequencing reads for the same gene. Data for the Jurkat and MRC5 cell lines have been summed from duplicate experiments. Note that in the case of HDL2 antisense transcripts were found to be more abundant than the sense transcript in all cell lines tested.

^bPBMCs are primary peripheral blood mononuclear cells; the HCT116 is a colorectal cancer cell line; Jurkat 1 is T cell leukemia cell line; MaPac is a pancreatic cancer line; MRC51 is a lung fibroblast cell line.

^cS is the number of sense sequences that map to a single gene.

^dAS is the number of antisense sequences that map to a single gene.

At the DM1 locus, the expanded CTG repeat tract is associated with methylation of a CCCTC-binding factor (CTCF) binding site, which appears to insulate the repeat from antisense transcription¹⁹. When the CTCF site is methylated, however, CTCF cannot bind, allowing enhanced antisense transcription through the expanded repeat¹⁹. Thus, paradoxically, heterochromatinization allows for higher levels of antisense transcription through the repeat, which could promote instability via TC-NER (Figure 2iii).

Libby et al investigated the role of the CTCF binding site in triplet repeat instability in a mouse model for SCA7¹⁰. In these mice, CTCF binding sites flank the 92 CAG repeats in the transgene. Mutation of the downstream CTCF site dramatically increased instability in the germline. Furthermore, the kidney, cortex, brainstem, and liver of mice mutant for the CTCF binding site also displayed a higher level of CAG instability compared to mice with a wild type binding site¹⁰. Intriguingly, one mouse that carried a wild type CTCF binding site showed elevated levels of instability only in its kidney. Further investigation showed an aberrantly methylated CTCF binding site in this tissue. DNA methyaltion prevented CTCF binding its target sequence in vitro. The authors concluded that CTCF binding, which is influenced by DNA methylation, contributes directly to preventing repeat instability. These results are akin to those for the expanded CTG repeat in DM1 cells, where methylation of the CTCF binding sites correlate with loss of CTCF binding and increased antisense transcription^{19, 21}. In the SCA7 mice, however, the role of antisense transcription has not been analyzed. Taken together, these data provide a tantalizing link between CAG repeat instability, CG-rich sequences, DNA methylation, and perhaps antisense transcription through the locus Antisense transcription could promote repeat instability in other ways. Simultaneous sense and antisense transcription might lead to head-on collisions between converging RNA polymerases, which could amplify the problems that arise when a repeat is transcribed in a single direction (e.g. enhance formation of secondary structure, persistence of RNA/DNA hybrids, and increased gratuitous TC-NER) thereby elevating instability.

Antisense transcription could also generate double-stranded RNA, which might trigger the RNA interference (RNAi). RNAi is a specialized pathway that allows post transcriptional repression of specific mRNAs⁶². Double stranded RNAs, such as microRNAs, are processed by the ribonuclease Dicer into 21nucleotide fragments. Short RNAs are subsequently loaded in the RNA-induced silencing complex (RISC), which targets mRNAs for degradation or inhibits their translation⁶². In fission yeast, the RNAi pathway prevents the appearance of centromeric transcripts by facilitating formation and maintenance of centromeric heterochromatin⁶². Whether an analogous process operates in mammalian cells is not yet clear. Nonetheless, repeat-containing transcripts can be processed by Dicer, and enter the RNAi pathway^{22, 63, 64}. Thus, the RNAi pathway could promote heterochromatin formation at disease loci²², which could impact instability.

RNAs perform a variety of other functions that might be relevant to repeat instability: they can serve as a template for DNA synthesis during DNA repair⁶⁵, guide programmed chromosome rearrangements⁶⁶, and help remove T-G mismatches generated during reprogramming⁶⁷. One of these functions might ultimately be found to play a role in repeat instability.

Concluding Remarks

The instability of trinucleotide repeats presents a surprisingly complex puzzle. One key point of agreement is that instability follows from the capacity of unstable repeats to form secondary structures, which in turn engage a variety of DNA repair activities in an attempt to regenerate a normal Watson-Crick duplex. In the past decade, most of the effort geared towards elucidating the mechanisms of triplet repeat instability has focused on knocking out or knocking down individual candidate genes and observing the resultant instability phenotype. A high fraction of tested genes can modulate triplet repeat instability, including elements of the replication machinery, DNA damage checkpoint components, and proteins involved in MMR, NER, BER, recombination, and transcription^{4, 57}.

The chromatin environment in which repeat tracts are embedded adds yet another layer of complexity. The distinctive heterochromatic structure at expanded repeats might prevent or promote instability depending on the developmental stage and nearby cis elements. The well characterized links between chromatin structure and DNA metabolism require an expansion of the discussion of the mechanisms of triplet repeat instability to include the possible roles of epigenetics and chromatin.

In the future, it will be important to catalogue the changes in epigenetic marks that occur at expanded repeat tracts during development. To move beyond correlation to causation will undoubtedly prove challenging. In addition, a better understanding of the relationship between chromatin structure at repeat tracts and repeat instability might guide the design of new therapeutic approaches. For example, histone deacetylase inhibitors can reactivate gene expression at the FMR1 and FRDA loci^{68, 69}. It will be interesting to test whether these molecules also alter the stability of repeat tracts. Drugs that can preferentially induce repeat contractions could lead to permanent improvements to patient health.