The multiple molecular facets of fragile X-associated tremor/ataxia syndrome

Fragile X-associated tremor/ataxia syndrome (FXTAS) is an adult-onset inherited neurodegenerative disorder characterized by intentional tremor, gait ataxia, autonomic dysfunction, and cognitive decline. FXTAS is caused by the presence of a long CGG repeat tract in the 5′ UTR of the FMR1 gene. In contrast to Fragile X syndrome, in which the FMR1 gene harbors over 200 CGG repeats but is transcriptionally silent, the clinical features of FXTAS arise from a toxic gain of function of the elevated levels of FMR1 transcript containing the long CGG tract. However, how this RNA leads to neuronal cell dysfunction is unknown. Here, we discuss the latest advances in the current understanding of the possible molecular basis of FXTAS.


Introduction
Fragile X-associated tremor/ataxia syndrome (FXTAS) is a neurodegenerative disorder that affects older adults who have a large CGG-repeat tract in the 5′-untranslated region (UTR) of the Fragile X Mental Retardation 1 (FMR1) gene [1]. Historically, carriers of Fragile X (FX) premutation alleles with 55 to 200 CGG repeats are considered at risk for FXTAS. The prevalence of premutation alleles is approximately 1 in 260 to 1 in 800 for males and 1 in 130to 1 in 250 for females in the general population [2,3]. Given reduced penetrance of FXTAS, it is estimated that 1 in 2,000 men over the age of 50 years in the general population will show symptoms of FXTAS [2,[4][5][6]. Clinical features of FXTAS include progressive intention tremor and gait ataxia, which is frequently accompanied by progressive cognitive decline, parkinsonism, peripheral neuropathy, and autonomic dysfunction [7]. The neuropathology of FXTAS consists of mild brain atrophy and degeneration of the cerebellum, including hyperintensity of the middle cerebellar peduncle (MCP), loss of Purkinje neuronal cells, spongiosis of the deep cerebellar white matter, Bergman gliosis, and swollen axons [8][9][10]. Immunocytochemical staining of post-mortem brain tissue from FXTAS reveals the presence of eosinophilic and ubiquitin-positive intranuclear inclusions that are broadly distributed throughout the brain, including in neurons and astrocytes [9,10], the spinal column, and several non-nervous tissues including thyroid, heart, and the Leydig cells in the testes [11,12].
In contrast to the absence of FMR1 mRNA and protein expression seen in carriers of a full mutation (over 200 CGG repeats), individuals with premutation alleles have markedly increased expression of FMR1 mRNA, but only moderately decreased FMRP levels [13][14][15][16]. FXTAS is not seen in carriers of fully silenced FMR1 alleles, suggesting that a novel mechanism, involving increased expression of the long CGG repeat tract in the FMR1 mRNA, is responsible for FXTAS. In support of this hypothesis, multiple studies have demonstrated adverse consequences of expressing CGG repeats in fly, mouse, and cell models [17][18][19][20][21][22]. Consistent with an RNA-based pathological mechanism, FXTAS has also been reported in individual carriers of intermediate alleles (45 to 55 CGG repeats) [23,24], and in full mutation allele carriers who are mosaics, both for repeat size and methylation, and who still express some FMR1 mRNA [25][26][27]. In addition, there has been a report documenting the presence of intranuclear inclusions in the brains of three older adult males with Fragile X syndrome (FXS) [12]. These results have implications for the spectrum of FX-associated disorders, and suggest that the definition of FXTAS may need to be broadened to include individuals whose FMR1 allele, irrespective of its size, makes sufficient RNA for its deleterious effects to be apparent.
How the RNA containing expanded CGG repeats leads to FXTAS pathogenesis is not yet fully known. This review will cover the recent advances in the understanding of the molecular mechanisms that may contribute to the pathogenesis of FXTAS, including the data presented at the First International Conference on FMR1 Premutation (23 to 26 June, Perugia, Italy). For other aspects of FXTAS, there are a number of excellent reviews available [28][29][30][31], as well the additional articles published in this special issue of JND.
CGG repeats are unstable, and tend to expand over time or with successive generations Increased CGG repeat numbers are associated with an increased risk of FXTAS and with an increased severity and reduced age of onset of FXTAS symptoms [5,6]. The CGG repeat tract responsible for FXTAS is polymorphic in the human population. Normal alleles have between 6 and 45 repeats, and are relatively stable on intergenerational transmission. However, as the repeat number increases, so too does the likelihood that the repeat tract will expand or gain additional repeat units on intergenerational transfer. AGG interruptions, which are commonly seen within FMR1 alleles, typically at 10 to 11 and 20 to 21 triplets from the 5′ end [32], are associated with a reduced risk of expansion [33,34]. Contractions do occur [35], but much less frequently. One of the consequences of the expansion bias is that alleles tend to increase in repeat number with successive generations. In addition to intergenerational expansion, somatic expansion is also seen in certain organs, including the brain in mice, and both lymphocytes and brain tissue in humans [27,36]. Somatic expansion may contribute to the repeat-length mosaicism that is seen in some human premutation carriers [27,[37][38][39][40]. This somatic expansion has the potential to exacerbate FXTAS symptoms, particularly in carriers of alleles with more than 100 CGG repeats, where the repeat may be particularly prone to expansion.
The mechanism responsible for these expansions is unknown. A number of other diseases are known to result from expansions of tracts containing these repeats or other short repeat units. Whether or not these diseases, which are referred to collectively as the repeat expansion diseases [41], share a common expansion mechanism is unknown. However, the unusual nature of these mutations suggests that they might. The expansion bias clearly differentiates the instability in these diseases from the classic microsatellite instability seen in certain cancers, where the repeat is as likely to lose repeats as it is to gain them.
The individual strands of expanded CGG repeats, like other repeats that cause repeat expansion diseases, form secondary structures, including hairpins and quadruplexes [42]. These structures affect DNA processing enzymes such as DNA polymerase, both in vitro [42] and in vivo ( [43,44]. It is generally thought that these structures are the trigger or substrate for expansion [41]. However, expansion in brain and liver, which are organs with a low proliferative capacity, along with expansion in mouse oocytes [45,46], suggest that the expansion mechanism may not involve aberrant DNA replication. Rather, given that oxidative stress exacerbates expansion in mice [47], it may be that expansion results from the aberrant repair of oxidized DNA or of DNA that is damaged in other ways. In contrast to generalized microsatellite instability, expansion in premutation mice actually requires Msh2 [48]. However, whether expansion involves disruption of classic mismatch repair or involves another MSH2-dependent process is unknown. Although Msh2 is required for expansion, it is not required for contractions [45,46,48]. Thus, it seems likely that expansions and contractions occur by different mechanisms, and the expansion bias seen in FX pedigrees may reflect a more efficient operation of the expansion process relative to the process that generates contractions.

Expression of FMR1 mRNA is increased in premutation carriers
Several studies have shown that premutation alleles are characterized by high mRNA expression levels [13,15,16,49]. In our recent analysis, FMR1 gene expression levels were measured in peripheral blood leukocytes from a total of 806 males across the whole range of CGG repeats, including normal individuals (n = 463), and individuals carrying intermediate (n = 60) and premutation (n = 283) alleles [27]. The results showed that FMR1 mRNA levels increased with increased CGG repeat number, and that a significant increase (P < 0.001) was detectable for allele lengths as short as 35 CGG repeats ( Figure 1). In addition, nuclear run-on experiments indicated that this elevated level of FMR1 mRNA in premutation carriers is caused by increased transcription efficiency rather than increased mRNA stability [14,15]. Despite higher levels of FMR1 transcripts, mild deficits of FMRP have been found in premutation carriers, and are probably due to a deficit in translational efficiency, particularly in the upper premutation range [13,16,49]. Thus, an FMRP deficiency is probably not the principal cause of FXTAS. Instead, the crucial observation that RNAs containing expanded CGG repeats accumulate in nuclear RNA aggregates in brain sections of patients with FXTAS [50] supports the notion that elevated levels of FMR1 mRNA trigger neuronal toxicity. In support of this hypothesis, heterologous expression of 90 CGG repeats in Drosophila melanogaster was shown to cause neurodegeneration and formation of ubiquitin inclusions [18]. Similarly, a knock-in (KI) mouse model, in which the endogenous eight CGG repeats of the murine Fmr1 were replaced with an expansion containing around 100 CGG repeats of human origin, showed ubiquitin-positive nuclear inclusions, and mild neuromotor and behavioral disturbances [17,51,52]. Finally, expression of transcripts containing 90 CGG repeats in a transgenic mouse model recapitulated some of the neuropathological and molecular features of FXTAS, despite the presence of a normal Fmr1 allele [19] (see also review on animal models for FXTAS in this issue). These animal models show that the expression of FMR1 mRNA containing expanded CGG repeats is both necessary and sufficient to cause pathological features characteristic of human FXTAS. Several mechanisms have been proposed to explain how increased expression of a RNA containing expanded CGG repeats could be pathogenic.
Is pathology the result of an RNA gain-of-function mechanism?
The first recognized examples of RNA gain-of-function diseases were two other repeat expansion diseases, myotonic dystrophy type 1 and 2 (DM1 and DM2) [53]. DM is the most common muscular dystrophy in adults, and in this condition, RNAs containing hundreds to thousands of CUG (DM1) or CCUG (DM2) repeats accumulate in nuclear RNA aggregates that sequester the Muscleblind-like (MBNL) splicing factors. Depletion of the free pool of MBNL1 leads to specific alternative splicing changes, which ultimately result in the symptoms of DM [53]. Extending this RNA gain-of-function model to FXTAS, the expanded CGG repeats are predicted to sequester specific proteins, resulting in loss of their normal functions, which would ultimately cause the symptoms of FXTAS [54,55]. Consistent with this idea, Iwahashi and collaborators [56] identified more than 20 proteins from inclusions purified from brains of patients with FXTAS. Of these, two RNA binding proteins were of special interest. The first, hnRNP A2/B1 is mutated in families with inherited degeneration affecting muscle, brain, bone, and motor neurons [57], while the second, MBNL1, is the splicing factor that is involved in DM [58]. However, a role for MBNL1 in FXTAS has been excluded, because no genetic interaction between MBNL1 and CGG-mediated neurodegeneration was observed in the fly model of FXTAS [59], and no misregulation of splicing events regulated by MBNL1 was observed in brain samples from patients with FXTAS [60]. By contrast, binding of hnRNP A2/B1 to RNA containing expanded CGG repeats was confirmed by independent proteomic and in vitro analyses [60,61]. Furthermore, overexpression of hnRNP A2/B1 rescued the neurodegeneration in transgenic Drosophila expressing 90 CGG repeats [59,61]. Interaction of hnRNP A2/B1 with RNA containing expanded CGG repeats was evident in cytoplasmic cerebellar lysates. By contrast, nuclear hnRNP A2/B1 presented little binding to CGG RNA, suggesting that some modifications of hnRNP A2/B1, either in the nucleus or in the cytoplasm, may alter the ability of hnRNP A2/B1 to bind to CGG RNA repeats [59]. The importance of titration of the cytoplasmic pool of hnRNP A2/B1 was further demonstrated by expression of expanded CGG repeats in primary cultures of rat sympathetic neurons [62]. RNA containing CGG repeats competed for binding of hnRNP A2/B1 to BC1 RNA, a dendritic regulatory RNA, resulting in impaired dendritic delivery of the BC1 RNA [62]. However, no misregulation of splicing events regulated by nuclear hnRNP A2/B1 was observed in brain samples of patients with FXTAS [60]. Overall, these data suggest that expanded CGG repeats recruit hnRNP A2/B1, resulting in depletion of the cytoplasmic but not the nuclear pool of hnRNP A2/B1. In addition, the ability of hnRNP A proteins to unfold tetraplex RNA structures, formed by expanded CGG repeats [63,64], raises the possibility that hnRNP A2/B1 may also act as a RNA chaperone that destabilizes these RNA structures. Finally, Sofola and collaborators [59] demonstrated that hnRNP A2/B1 recruits, in trans and through protein-protein interactions, other proteins such as CUGBP1, an RNA binding protein, whose expression is increased in heart samples of patients with DM [65]. These data indicated that proteins binding to CGG RNA may recruit other proteins, resulting in dynamic aggregates that expand over time, a model later confirmed in COS7 cells expressing 60 CGG repeats [60]. Overexpression of either hnRNP A2/B1 or CUGBP1 rescued neurodegeneration in a Drosophila model of FXTAS, highlighting the potential importance of hnRNP A2/B1 and CUGBP1 to FXTAS pathology [59].
In addition to hnRNP A2/B1, proteomic analyses performed by Jin and collaborators [61] also showed that purine-rich binding protein α (Purα) binds robustly to RNA containing expanded CGG repeats. Purα is a single-stranded cytoplasmic DNA and RNA binding protein that has been implicated in many biological processes, including RNA transport and translation. Importantly, overexpression of Purα rescued neurodegeneration in a Drosophila model of FXTAS [61]. However, presence of Purα within nuclear aggregates in FXTAS brain samples is inconsistently observed. Jin et al. found Purα in cytoplasmic inclusions in Drosophila expressing 90 CGG repeats, and in inclusions in superior-mid temporal cortex neurons from human FXTAS brain sections [61]. By contrast, Iwashashi et al. did not detect Purα in purified inclusions from cerebral cortex of patients with FXTAS [56]. Furthermore, Purα-positive inclusions have not been observed in mouse models of FXTAS [66], or in hippocampal and cortical brain section of patients with FXTAS [67]. These results suggest that the composition of the inclusions varies from one brain region to the next and from one model organism to the other. Analogous to the recruitment of CUGBP1 by hnRNP A2/B1 to RNA containing expanded CGG repeats, Purα was shown to recruit Rm62, the Drosophila ortholog of the RNA helicase P68/DDX5 [68]. Expression of expanded CGG repeats resulted in the post-transcriptional downregulation of Rm62, ultimately resulting in nuclear accumulation of Hsp70 mRNA and of other mRNAs involved in stress and immune responses [68]. Overexpression of Rm62 rescued neurodegeneration in flies expressing 90 CGG repeats, highlighting the potential importance of P68/DDX5 to FXTAS pathology [68].
SAM68, a splicing regulator encoded by the KHDRBS1 gene, was also found in CGG RNA aggregates [60]. However, overexpression of SAM68 was not sufficient to rescue neuronal cell death induced by expression of expanded CGG repeats [67]. As with CUGBP1 and Rm62, SAM68 did not bind directly to CGG repeats, and recruitment of SAM68 within CGG RNA aggregates occurred in trans through protein-protein interactions [59]. Sellier and collaborators [67] also showed that DROSHA-DGCR8, the enzymatic complex that processes pri-microRNAs into pre-miRNAs, associated specifically with CGG repeats of pathogenic size. Sequestration of DROSHA-DGCR8 within CGG RNA aggregates resulted in reduced processing of pri-miRNAs in cells expressing expanded CGG repeats, and in brain samples from patients with FXTAS. Overexpression of DGCR8 rescued neuronal cell death induced by expression of expanded CGG repeats [67]. These results suggest that titration of DGCR8 by expanded CGG repeats is a leading event to CGG-induced neuronal cell death. However, recent analyses of miRNA expression in blood samples of patients with FXTAS and in Drosophila expressing CGG repeats did not show a global downregulation of miRNA, but rather, the expression of some specific miRNAs was misregulated [69,70]. Whether depletion of DROSHA-DGCR8 varies in blood and brain of patients with FXTAS, and whether the Drosha-Pasha complex is sequestered in cytoplasmic aggregates in Drosophila expressing expanded CGG repeats, remains to be determined. Similarly, whether overexpression of hnRNP A2/B1, P68/DDX5, DROSHA-DGCR8, or CUGBP1 rescues any phenotype in mouse models expressing expanded CGG repeats, would be necessary to determine the importance of these candidate proteins to FXTAS pathology.
These caveats aside, the observations described above suggest that CGG repeats could be pathogenic by sequestering specific RNA binding proteins, resulting in loss of their normal functions, and thus lead to neuronal cell dysfunction ( Figure 2) [56,58,61,67,68]. However, this attractive model has some weaknesses. First, the inclusions observed in FXTAS brain sections differ from those seen in DM, a typical RNA gain-of-function disorder. In FXTAS, inclusions are larger and ubiquitinated, and contain various chaperone proteins such as Hsp27, Hsp70, and αB-crystallin [9,56]. In short, these large inclusions resemble the aggregates seen in protein-mediated disorders, although they are negative for the typical proteins found in tauopathies, synucleinopathies, or polyQ disorders (for example, Huntington's disease). Second, and most disconcerting, although inclusions in brain samples of patients with FXTAS contain the mutant FMR1 RNA with expanded CGG repeats [50], a mouse model, in which the endogenous eight CGG repeats of Fmr1 is replaced with an expansion containing around 100 CGG repeats, shows numerous ubiquitin inclusions but only rare aggregates of SAM68 or DROSHA-DGCR8, associated with rare RNA aggregates of expanded CGG repeats [60,67]. Similarly, overexpression of expanded CGG repeats leads to formation of nuclear RNA aggregate in some cell types, including primary cultures of hippocampal embryonic mouse neurons and PC12, COS7, and SKOV3 immortalized cell lines, but no RNA aggregates have been observed in A172, U-937, THP1, HeLa, HEK293, NG108-15, IMR-32, Neuro-2a, SH-SY5Y, SK-N-MC, or SK-N-SH cells [22,60]. In short, not all cell lines can support CGG repeat aggregate formation, whereas in DM, expression of expanded CUG or CCUG repeats consistently results in formation of RNA foci. Thirdly, a recent and provocative study demonstrates that the toxic effect of CGG repeats depends on their location [71]. Moving expanded CGG repeats from a 5′ UTR to a 3′ UTR position reduced their toxic effect in Drosophila, whereas expanded CUG or CCUG repeats were found to be pathogenic in whatever location tested, provided they were expressed in sufficient amounts to deplete MBNL proteins. These data led Todd and collaborators to reconsider the model of RNA binding protein sequestration, and to explore further the molecular mechanisms of FXTAS.

Non-canonical AUG translation produces a polyglycine-containing protein in FXTAS
An unexpected observation, made by Todd and colleagues in flies transgenic for a construct containing 90 expanded CGG repeats cloned upstream of the green fluorescent protein (GFP) cDNA, was that some of the GFP signal was found in cytoplasmic inclusions. Western blotting analysis showed a band of the expected size for GFP, but also detected a protein 12 kDa larger [71]. Because translation of expanded CAG repeats in the absence of an ATG initiation codon (repeat-associated non-ATG translation or RAN translation) had been previously reported [72], Todd and collaborators tested whether expanded CGG repeats could be translated despite the fact that no ATG codon is present upstream of the repeats. Their analysis revealed that, indeed, translation of CGG repeats occurs in two out of the three frames, giving rise to short proteins containing either a polyalanine or a polyglycine stretch. Expression of the polyglycine protein resulted in the formation of protein inclusions, which were toxic both in neuronal transfected cells and in Drosophila. Further analyses of the polyglycine protein revealed that its translation was probably initiated at non-canonical AUG codons, such as CUG and GUG, which were located upstream of the CGG repeats. A role for non-canonical translation initiation in inclusion formation is consistent with data from two different KI mice mouse models. In a mouse model that showed numerous ubiquitin inclusions, the expanded CGG repeat from a human premutation allele was cloned, along with sequences upstream of the CGG repeats in humans that contained the non-AUG initiations codons [17]. By contrast, in a second mouse model, in which the mouse 5′ flanking sequence was retained, a stop codon was found to be located just upstream of the expanded CGG repeats [21]. These latter mice showed relatively few ubiquitinated aggregates, thus supporting the notion that non-ATG-initiated translation of the CGG tract is required to generate most of the inclusions [71]. That this unusual mode of translation may play a role in FXTAS is evidenced by the fact that, with the aid of specific antibodies, polyglycine protein can be seen in brain sections of patients with FXTAS [71]. Overall, these observations suggest that a protein gain of function may also occur in cells of patients with FXTAS. However, what contribution the polyglycinecontaining or polyalanine-containing proteins make to the etiology of FXTAS is an exciting open question.
Non-coding transcription of the FMR1 locus: a role in FMR1 mRNA toxicity?
The majority of the human genome is transcribed but not translated. Such RNAs are classified as long noncoding RNAs (lncRNAs) when longer than 200 nucleotides [73][74][75]. To date, relatively few lncRNAs have been functionally characterized, but increasing evidence suggests that many may have important functions, including the regulation of transcription, RNA processing and translation, DNA methylation, and chromatin architecture, both locally (cis-acting) and across some genomic distance (trans-acting) [76][77][78].
In addition to the FMR1 transcript, a variety of RNAs are produced from the FMR1 locus. Therefore, it is possible that these lncRNAs produced from the FMR1 locus may modulate certain aspects of FXS/FXTAS, as has been shown in other human diseases [79]. For example, Kumari and Usdin described an abundant antisense transcript of about 5 kb that spans the region upstream of the FMR1 promoter, and whose expression does not change in response to repeat expansion [80]. By contrast Ladd and coworkers described a transcript, Antisense FMR1 (ASFMR1), that spans the expanded CGG repeats and whose expression is elevated in lymphoblastoid cells and peripheral blood leukocytes of individuals with premutation alleles, while it is not expressed in those with full mutation alleles [81]. Multiple splice forms of ASFMR1 have been identified, which show differential expression in carriers of premutation and normal alleles [81]. One of these ASFMR1 splice variants contains a small intron that uses a non-consensus CT-AC splice site that is transcribed in a premutation cell line, but is absent in a normal cell line [81]. We compared the expression levels of this ASFMR1 isoform in blood from individuals with alleles ranging from normal to premutation, and found a significant increase with CGG repeat number (P < 0.001) (Figure 3) [27]. Of interest, both unspliced and spliced ASFMR1 transcripts contain putative open-reading frames encoding polyproline peptides, resulting from antisense-oriented translation of the expanded CGG repeats [81]. Whether ASFMR1 containing expanded CCG repeats is translated and participates in the formation of the pathogenic nuclear inclusions observed in patients with FXTAS remain to be tested.
Another antisense transcript, FMR4, originates upstream of the FMR1 start site, and covers 2.4 kb of sequence [82]. FMR4 is widely expressed in fetal and adult human tissues, and throughout human and macaque brain regions. Expression of FMR4, like that of ASFMR1 and FMR1, is increased in brain tissue of premutation individuals and is silenced in individuals with the full mutation [82]. Importantly, FMR4 overexpression was shown to increase cell proliferation, whereas FMR4 downregulation induced apoptosis in vitro [82]. Additionally, no cis-acting effect was observed upon expression of the FMR1 gene. Therefore, it was hypothesized that FMR4 influences proliferation pathways in trans, by targeting distal genomic loci. Current work is focused on defining a role for this transcript, as it has been found to affect the chromatin state and transcription of several genes involved in neuronal differentiation, axon guidance, and synaptic signaling, as well as cell cycle regulators (Peschansky and Pastori, unpublished data).
Two new transcripts arising from the FMR1 locus, FMR5 and FMR6, were recently identified [83]. FMR5 is a sense-oriented lncRNA transcribed from approximately 1 kb upstream of the FMR1 transcription start site (TSS). FMR5 is not differentially expressed in human brain from unaffected individuals compared with full mutation and premutation patients, suggesting that its transcription is independent of CGG repeat expansion. Furthermore, the TSS of FMR5 appears not to be affected by the chromatin silencing that occurs within full mutation alleles, or by the open chromatin hypothesized to increase transcription of FMR1 premutation alleles. FMR6 is a spliced long antisense transcript, 600 nucleotides i, whose sequence is entirely complementary to the 3′ region of FMR1 [83]. It begins in the 3′UTR, ends in exon 15 of FMR1, and uses the same splice junctions as FMR1. An unexpected finding was that FMR6 is reduced in premutation carriers, suggesting that abnormal transcription and/or chromatin remodeling occurs toward the distal end of the locus. However, the chromatin marks associated with the 3′ end of FMR1 in premutation carriers have yet to be described. The function of FMR6 remains to be identified, but its complementarity to the 3′ region of FMR1 presents several interesting possibilities. FMR6 may bind to FMR1 mRNA, thereby regulating the stability, splicing, subcellular localization, or translational efficiency of FMR1, as has been described for other lncRNAs [77]. Notably, FMR6 overlaps miR-19a and miR-19b binding sites in the FMR1 3′ UTR [84], suggesting that FMR6 may modulate the stability or translational efficiency of FMR1 by interfering with microRNA binding. Overall, these results highlight the importance of non-coding transcription of the FMR1 locus. However, much work remains to fully understand the relevance of these transcripts to the pathology observed in premutation carriers.

Conclusion
The restriction of FXTAS clinical features to unmethylated, transcriptionally active alleles with large CGG repeat numbers suggests that the expression of a mutant RNA is pathogenic to neuronal cells [55]. This hypothesis is supported by data from cell, fly, and mouse models [17][18][19][20][21][22]. However, how these RNAs cause neuronal cell dysfunction and FXTAS symptoms remains unclear. One model proposes that the RNA containing expanded CGG repeats is pathogenic via its sequestration of specific RNA binding proteins. Various proteins, including Purα, Rm62, CUGBP1, hnRNP A2/B1, SAM68, and DROSHA-DGCR8, have been shown to bind directly or through a protein partner to expanded CGG repeats [56,59,61,67,68]. However, it remains to be tested whether overexpression of these candidate proteins rescues any phenotype in mouse models expressing expanded CGG repeats. A second mechanism involves non-canonical translation initiation of expanded CGG repeats. resulting in expression of toxic polyglycinecontaining and polyalanine-containing proteins [71]; however, how these proteins promote neuronal cell dysfunction is an open question. A third model is associated with the expression of antisense FMR1 transcripts. Further investigation is needed to evaluate the pathological consequences of expression of ASFMR1 or other long non-coding RNA mapping within the FMR1 gene. Finally, although decreased expression of FMRP is probably not the principal cause of FXTAS, it cannot be excluded that a reduction in FMRP plays a role in modulating some of FXTAS features. In that context, the level of FMRP depletion in brain samples from a larger cohort of patients with FXTAS needs to be measured.
In conclusion, in addition to increased FMR1 mRNA production, protein titration, non-AUG translation,   Figure 4 Mechanisms that may contribute to fragile X-associated tremor/ataxia (FXTAS) pathology. Expression of FMR1 mRNA and associated antisense transcripts are increased inpremutation carriers. Some transcripts accumulate in the nucleus and recruit various RNA binding proteins, while export and non-AUG translation results in production of polyglycine-containing or polyalanine-containing proteins.
antisense transcription, and decreased expression of FMRP are a number of non-exclusive mechanisms that may all contribute to FXTAS pathology. It is possible that contributions to pathology from more than one mechanism may help to explain the great variability in clinical presentation of premutation individuals, aspects of which have heretofore not been accounted for by CGG expansion size, mosaicism, methylation, alternative spliced isoforms, additional genomic changes, or other known factors. Thus, more work is needed to determine the relative contribution of these processes to disease pathology in this multifaceted disorder (Figure 4).