Fragile X Syndrome (FXS) is one of a group of genetic diseases that are caused by pathogenic expansion of a trinucleotide repeat. The diseases, including Huntington's disease and Friedreich's ataxia, are characterized by neurological dysfunction, often in specific regions of the CNS. The trinucleotide repeats disrupt specific genes in each disease, but very little is known about how the expansion reaches pathological levels and how the dysfunction of specific genes leads to neurological disorder. In some cases, such as Huntington's disease, the expanded repeat is exonic and causes the expression of a pathogenic protein 
. In FXS and Friedreich's ataxia 
, the repeat is in a non-coding region of the gene and the pathogenic expansion results in silencing of the gene through epigenetic mechanisms. In FXS, the disease is triggered when there are 200 or more copies of a CGG trinucleotide repeat in the 5′UTR of the FMR1
gene. The number of repeats predicts the pathology: unaffected individuals have about 30 repeats, the FMR1
gene promoter is unmethylated, and the FMRP protein is expressed; full mutation individuals (>200 repeats) have fully methylated FMR1
and produce no protein 
. Interestingly, the intermediate premutation (ca. 100–150 copies) has elevated transcription but translation of the FMRP protein is inefficient 
. Premutation carriers, while having normal intelligence, demonstrate a range of psychiatric and behavioral symptoms and are associated with a number of medical conditions such as Fragile X-associated tremor/ataxia (FXTAS) and Premature Ovarian Insuffuciency (POI) in response to the elevated transcription levels of FMR1 
Trinucleotide repeat diseases have been difficult to study because of limitations in availability of cells and tissues from the brains of affected individuals. Mouse models of these diseases are suboptimal because of differences in neural development in mouse and human. The development of iPSC technology has enabled in vitro studies of central nervous system cells derived from patients with genetic neurological disease. However, the value of iPSC modeling of human disease relies on the assumption that the resulting iPSC lines contain the same causative elements of the disease that the input patient cells contained. The data we present here draw into question this assumption, and show that the iPSCs derived from FXS individuals do not necessarily faithfully reproduce the CGG-repeat lengths, CpG methylation status, and silencing of the FMR1 gene in the fibroblasts of origin. We also show that differences in neuronal differentiation among FXS iPSC lines are attributable at least in part by the epigenetic status of the FMR1 gene promoter.
Existing mouse models with a knock-out of Fmr1
are not appropriate for investigating questions of repeat stability or the epigenetic mechanisms of FMR1
silencing as they lack the expanded trinucleotide repeat. Knock-in mouse models in which the murine CGG repeat has been replaced with a premutation-sized CGG repeat from humans were reported to exhibit moderate repeat instability with both paternal and material transmission 
. However, in addition to CGG-repeat length, since the nature of the flanking sequences in combination with the patterns of interruption of CGG repeats can influence nucleosomal structure and alter CGG repeat instability 
, the use of genetically accurate, human neuronal models will be advantageous to investigate the molecular mechanisms of trinucleotide repeat instability and epigenetic regulation.
In one case, we discovered that a patient fibroblast cell line, GM05131, is a heterogeneous mixture of normal and full FXS mutation cells. Reprogramming of these fibroblasts resulted in two iPSC clones, one with the full FXS mutation (ca. 700 CGG repeats) and the other with premutation repeat length (ca. 142 repeats). As expected, the full mutation cells produced no FMR1 transcript and the premutation clone had above-normal FMR1 transcription levels but very low translation of the FMRP protein. These two clones are presumably otherwise genetically matched, which will be valuable for comparison of the effects of the full- and premutation in the absence of potentially confounding background genetics. Interestingly, the CGG-repeat lengths in the iPSCs appeared to be slightly shorter than those of the fibroblasts (700 vs. 800; 142 vs. 166); in light of the similar changes that appeared upon reprogramming of the other fibroblast lines (see below), it seems possible that the reprogramming process may lead to instability of trinucleotide repeat lengths.
A second FXS line, GM05848 (ca. 700 CGG repeats), gave rise to an iPSC clone that apparently possessed multiple trinucleotide repeat lengths ranging from 400 to 900. And the third FXS line reprogrammed (GM05185) had a predominant trinucleotide repeat of 800, but produced a heterogeneous iPSC line (185-iPS1) with discrete repeat lengths (200 and 700). There was no apparent heterogeneity in the input fibroblasts, as evidenced by lack of FMR1 transcript detected, even after extensive qRT-PCR (40–45 rounds). The full mutation iPSC clones (848-iPS1 and -iPS3) showed no detectable FMR1 expression, but the 185-iPS1 clone, with a de novo premutation subpopulation present, expressed the high levels of FMR1 transcript typical of premutation cells.
These data suggest that reprogramming of full mutation FXS fibroblasts results in changes, generally shortening, of the repeat length in the resulting iPSC clones. A varied population of repeat lengths in one isolated iPSC clone (848-iPS3) suggests that change in CGG-repeat length is a dynamic process that occurs because of instability of the CGG repeat initiated during reprogramming. An earlier study reported that full mutation human FMR1
alleles stably maintained in patient fibroblasts and murine A9 somatic hybrid cells contracted upon transfer to pluripotent embryocarcinoma (PC13) cells due to instability upon passage 
. Repeat instability is also suggested by the variable repeat lengths observed in a human embryonic stem cell line derived from a FXS-affected embryo, which showed repeat length heterogeneity from 200 to more than 1,000 triplet CGG repeats in the same isolated clone 
This is the first report of trinucleotide repeat length change in FXS iPSCs. A previous analysis of FXS iPSCs did not report trinucleotide repeat length changes 
. However, changes in repeat length with reprogramming has been reported for another trinucleotide repeat disease, Friedrich's ataxia 
; in that case, in iPSCs there was an expansion of an intronic GAA repeat that silences the FXN
gene on chromosome 9. That report and the current study suggest that reprogramming may destabilize repeats in certain trinucleotide repeat diseases. Further investigation of this phenomenon may help in understanding the basis of transgenerational instability of pathological trinucleotide repeat sequences in many neurodevelopmental diseases.
The impact of repeat instability on iPSC in vitro models of FXS could be considerable if the iPSC repeat length is not determined. We found that in general the actual repeat length in the iPSCs predicted the methylation status and expression levels of FMRP transcripts and proteins, and therefore the disease state, regardless of the status of the input fibroblasts. If the changes in repeat length are truly dynamic, researchers may find unexpected phenotypes in iPSC derivatives if they do not monitor the repeat length in the cells.
Two previous reports have investigated FMR1
expression in human pluripotent cells, with conflicting results: one study used FXS human embryonic stem cells (hESCs) 
and the second studied FXS iPSCs 
. The first report indicated that the FMR1
gene was expressed in the FXS-hESCs, despite the cells having full mutation status, and was repressed only after differentiation 
. The second study reported that FMR1
expression was repressed in both full mutation undifferentiated FXS-hESCs and FXS patient-derived iPSCs (from the GM05848 line) 
. Our results support the report on FXS iPSCs; we observed promoter CpG methylation and FMR1
repression in GM05848-derived iPSCs as well as in all other iPSC clones that contained only full mutation alleles. We also characterized neuronal differentiation in several FXS iPSC lines, showing for the first time that the CpG methylation state of the FMR1
gene in iPSCs persists during neuronal differentiation, an observation that is critical for efforts to use iPSC-derived cells to model FXS.
We observed FXS-associated morphological differences in iPSC-derived neurons, with FXS cells having fewer and shorter neurites than controls. Similar neuronal morphology has been reported in FMR1
knock-out mouse models 
and post-mortem fetal FXS brain tissue 
. The morphological differences correlated with FMR1
promoter CpG methylation status and expression of FMR1,
and occurred in multiple iPSC lines from different source fibroblasts. We also observed variations in glial differentiation as assessed by GFAP immunostaining, although these phenotypes were not strictly linked to FMR1
methylation status. There have been previous reports of differences in glial/neuronal ratios in FXS-derived cell cultures. Adult neural stem cells from the dentate gyrus of Fmr1
knockout mice showed increased glial differentiation as compared to controls 
. Observations using human neural tissue differ and are possibly brain region-specific; neurospheres derived from FXS hippocampal tissue showed reduced glial differentiation 
, whereas cortex-derived cells were unaffected 
Overall, our results suggest an important role for FMRP early in human neurodevelopment. In this context, future studies will be aimed toward understanding the molecular basis of the observed phenotypes and exploring the consequence of a loss of FMRP on signaling and synaptic function in FXS–derived neuronal cells. Having identified a robust, morphological phenotype upon neural differentiation of FXS iPSCs provides an opportunity for the characterization of existing pharmacological agents and to potentially discover novel therapeutics that can reverse disease-associated phenotypes in FXS and other ASDs sharing common pathophysiology.