br Results br Discussion We report on

Results
Discussion We report on the derivation and full characterization of two hESC lines (SZ-ALS1 and SZ-ALS3) with a GGGGCC expansion of approximately 270 repeats. Our C9 hESC lines were established from embryos obtained through PGD from a woman with a C9 mutation with >40 repeats in her peripheral blood. Interestingly, despite the sufficiently large expansion, both cell lines were completely unmethylated at the repeats (based on (G4C2)n-methylation assay [Xi et al., 2015a, 2015b]) and upstream of the repeats (based on bisulfite colony sequencing). In addition, we generated iPSCs clones that are haploidentical to the mutant hESCs from skin fibroblasts of the asymptomatic C9-carrier mother (700 repeats, patient H), and from an unrelated 65-year-old ALS-manifesting patient (2,700 repeats, patient M). Unexpectedly, we found a striking difference in methylation levels at the 5′ UTR of C9orf72 between the C9 iPSCs and all other cell types examined. Unlike in the C9 hESCs and parental fibroblasts, methylation was detected at the expanded repeats, and reached almost 100% at the upstream CGI in all iPSC clones. Although the difference in C9orf72 hypermethylation between the C9 hESCs and iPSCs could, in theory, result from a difference in expansion size, this is very unlikely since the number of GGGGCC repeats in the hESCs is well above the threshold for methylation to be triggered in any cell type examined thus far (Xi et al., 2015b). In addition, methylation levels in the iPSCs are much higher than those observed in somatic cells of patients (where it does not exceed 20% [Liu et al., 2014]), and is consistent with the exclusive enrichment of the repressive histone mark H3K9me3 and not H3K27me3. This is somewhat different from the data obtained by Belzil et al. (2013), who demonstrated enrichments for H3K27me3 in addition to H3K9me3, although they analyzed different cell types from those described in this study. While we used undifferentiated cells, they looked at frontal cortex and cerebellum. Indeed, our findings in undifferentiated cells calcium calmodulin dependent protein kinase doubt on the importance of H3K27me3 modification for setting the methylation state and transcriptionally inactive chromatin configuration at the C9 expansion. We argue that the C9 mutation acts as a hotspot for de novo methylation by transcription factor reprogramming considering the dramatic rise in DNA methyltransferase 3B (DNMT3B) expression levels during the reprogramming procedure (Huang et al., 2014). In fragile X syndrome it was also shown that reprogramming of fibroblasts with an unmethylated full CGG expansion results in hypermethylation and complete inactivation of the FMR1 gene in mutant iPSCs (de Esch et al., 2014). Thus, perhaps hypermethylation of the GGGGCC repeats in our C9 iPSCs represents a much wider phenomenon, whereby untranslated repeat expansions that reside within CGIs provide a “sink” for DNMTs during transcription factor reprogramming (rather than first de novo methylate and then failure to demethylate). This may also explain why we and others generally fail to demethylate and reactivate the FMR1 gene by somatic cell reprogramming when producing iPSCs from cells of fragile X-affected patients (Avitzour et al., 2014; Sheridan et al., 2011; Urbach et al., 2010). It should be noted that our results contradict the report of Esanov et al. (2016), who showed demethylation (rather than de novo methylation) of the C9 mutation in C9 iPSCs. The discrepancy could result from the different cell states employed. We used primary skin fibroblasts from two unrelated patients, whereas they used an immortalized cell line from blood cells of a single patient. Finally, to associate hypermethylation with disease pathogenesis, we analyzed the expression of transcript variants 1, 2, and 3 individually. We show that the GGGGCC expansion alters the region to allow the enhancement of V1 transcripts, albeit with much lower levels in iPSCs relative to hESCs. In addition, we demonstrate that the C9 mutation, together with neural differentiation, favors transcription from an upstream promoter (exon 1a-initiating transcripts, V1 and V3) over a downstream promoter (exon 1b-initiating transcript, V2), and that this effect is largely restricted in iPSCs. Furthermore, we found preferential retention of intron 1 in both C9 hESCs and iPSCs by RNA deep sequencing, illustrating the propensity of the mutation to interfere with the proper splicing of this region in both exon 1a- and 1b-initiating transcripts. However, this change between WT and C9 cells demonstrates a greater effect in hESCs (4.5-fold increase) relative to iPSCs (only a 2-fold increase). More importantly, by monitoring for the expression of potentially pathogenic mRNA transcripts (i.e., exon 1a-initiating transcripts that retain intron 1), we find no difference between mutant and WT undifferentiated cells (hESCs and iPSCs). However, their level becomes significantly higher upon differentiation into disease-relevant cell types (NPCs and teratomas) exclusively in C9 hESCs, and not in C9 iPSCs. We propose that methylation counteracts the effect of the expansion by downregulating exon 1a-initiating mRNA species. In line with this idea are previous reports by Liu et al. (2014) and others (Bauer, 2016; Day and Roberson, 2015), which point to a mechanistic link between hypermethylation and reduced accumulation of RNA foci and dipeptide inclusions in patient cell lines, brain samples, and HEK293T transgenic cell lines.