This is the first study to characterize a set of CAG expansion-associated phenotypes in neural cells derived from HD iPSC lines. A unique aspect of the current report was that we worked as a consortium, using the same set of lines in a wide range of cellular assays. HD is an ideal disorder for exploring the utility of iPSC for disease modeling as it is caused by a single gene and there is a strong correlation between the length of the expanded of CAG repeat and age of disease onset (Consortium 1993
), Stine et al 1993
). In addition, there is a weaker, though still highly significant correlation between the length of the expanded repeat and rate of progression some clinical phenotypes (e.g. motor and cognitive disorder) though not others (e.g. emotional disorders) (Rosenblatt et al 2012
). Interestingly, we found that in some assays such as cellular aggregation, overall energy metabolism and cumulative risk of death over time in the long differentiation protocol, both the HD60 and HD180 lines showed very similar pathological profiles that were significantly different from control lines. However, in other assays, such as cell survival in a short differentiation protocol, BDNF withdrawal and glutamate toxicity, only the HD180 line (and, where tested, the 109 line) showed a robust phenotype. Finally, in one assay we developed based on calcium homeostasis following repeated glutamate pulsing, we saw a clear repeat dependent phenotype with a graded response across the HD33, HD60 and HD180 lines. The gene array studies also suggested a gradation of expression changes, particularly those that are differentially expressed in both the very long repeat lines (HD180, HD109), and the more moderate length repeat (HD60) compared to control lines, but show differences clearly segregating between the long and medium repeat lines within this group. Together these studies suggest that there are clear phenotypes associated with expanded CAG repeats, and that the choice of assay and exact tissue culture conditions will determine the extent of the phenotypic gradation with length of the CAG expansion that can be detected.
The success of this model reflected the maintenance of the HTT
CAG repeat expansion following reprogramming, stem cell expansion and subsequent differentiation. The HD lines demonstrated only mild CAG repeat instability in culture, and the slight increase in repeat number with passaging for one of the longest CAG lines (HD109) may correspond to the somatic genomic CAG instability seen in tissues from HD patients (Shelbourne et al., 2007
). As expected, we found expression of the mutant protein in these cultures, although no inclusion bodies were found in the cells before or after differentiation, or after the addition of cellular stressors, possibly reflecting the long period of time before inclusions develop in the human disease (Ross and Poirier, 2004
). It is important to note that while HTT inclusions are pathognomonic in post mortem tissue, inclusion formation is not linked to HTT cell toxicity (Arrasate et al., 2004
Alterations in gene transcription and protein expression are prominent in HD mouse models and human HD brain tissue (Ross and Thompson, 2006
). In dividing HD NSCs, which had a very stable number of nestin-positive progenitors, we observed HD-related changes in gene expression including SLC1A3 (Fan and Raymond, 2007
), UCHL1 ( Xu et al., 2009
), EGFR (Lievens et al., 2005
), and TRK receptors (Apostol et al., 2008
; Zuccato et al., 2010
), consistent with changes in human HD brain (Hodges et al., 2006
), and HD transgenic mouse striatum (Luthi-Carter et al., 2000
). Among the genes up-regulated, GLB1, PLSCR4, PTGIS and PLA2 have been implicated in lipid metabolism and membrane fluidity, with possible consequences for cell signaling and receptor function in HD (Karasinska and Hayden, 2011
). Additional novel altered pathways were identified, including a network of G-protein coupled receptors, developmental genes such as PAX6, and matrix metalloproteinases, consistent with the involvement of this family of proteases in mutant HTT toxicity (Miller et al., 2010
). The concordance of pathways and networks at RNA and protein levels further suggests a primary dysfunction of these systems in this model of HD. These results suggest that expanded CAG repeats were having a biological effect even at the neural stem cell stage and further analysis of these novel pathways may uncover factors that contribute to emerging HD pathology (Molero et al., 2009
Following differentiation to a striatal-like phenotype, we found more genes with increased rather than decreased expression, as shown previously in studies of human HD striatum (Hodges et al., 2006
). Genes up-regulated in human tissue and differentiated HD iPSCs included p53, which is also up-regulated in HD mice and may contribute to the cell death seen in this study and by others (Bae et al., 2005
); syndecan4, involved in recycling of lipids and cholesterol from degenerating terminals (Blain et al., 2004
); HMG box protein 1, a tumor suppressor and transcription factor that accumulates in the Alzheimer’s brain and may impair Aβ clearance (Takata et al., 2004
); and SRPX, which contributes to language and cognitive development (Royer et al., 2007
). An advantage of this iPSC model is that gene expression changes in human neurons can be identified overtime during the degeneration process and at specific stages of neuronal differentiation to illuminate pathogenic mechanisms, in contrast to gene expression studies done only in end-stage post-mortem human HD brain tissue.
A number of dominant, CAG length–dependent biochemical and cell biological phenotypes, such as cell adhesion and altered energetics, have been observed in murine Hdh
CAG knock-in ESCs and derived immortalized lines (Gines et al., 2010
; Jacobsen et al., 2011
). Neural progenitors derived from the HD60i and HD180i NSC lines were found to undergo less aggregation upon plating, and showed significantly reduced ATP levels compared to control lines. This suggests disruptions of cell adhesion and energy metabolism, which may alter the cells’ ability to survive and differentiate appropriately. This was reflected by a gradual reduction in the number of neurons that could generate spontaneous and induced action potentials over time in the HD lines, and ultimately cell death by three weeks of differentiation in the HD180i cultures. Thus, our studies showed that under specific culture conditions there was a severe phenotype associated with endogenous levels of mutant HTT expression.
Using a longer differentiation protocol that included addition of multiple growth factors to direct the cells towards a striatal lineage (Aubry et al., 2008
), it was possible to avoid the acute stress-related neural cell death seen in the HD NSC lines, again reinforcing the idea of stress being a key component of these in vitro
models of disease. A very sensitive single cell time-lapse assay (Arrasate et al., 2004
) showed conclusively that cells within the HD60i and HD180i lines had a significantly greater cumulative risk of death than those in control lines. Furthermore, a similar cell death pattern in control lines could be seen by over-expression of mutant but not normal HTT. To uncover further phenotypes in cultures at single time points may require stressors to the system, as shown in iPSC models of Parkinson’s disease and SCA3 (Mattis and Svendsen, 2011
; Seibler et al., 2011
). We found clear repeat expansion-associated differences in the vulnerability of cells to the addition of exogenous stressors, such as H2
or 3-MA, or by repetitive exposure to glutamate. This is of interest in light of studies showing that HD may be associated with increased reactive oxygen species (Tunez et al., 2011
). Withdrawal of BDNF from the medium also revealed a CAG expansion-associated toxic phenotype, based on the time-lapse assay, nuclear condensation assays, and caspase activation. There is a long literature on the role of BDNF in HD pathogenesis in relation to striatal neuron vulnerability (Zuccato and Cattaneo, 2009
), as well as reported toxicity in response to BDNF withdrawal in a single HD iPSC line (Zhang et al., 2010
). Our new data further support a central role of BDNF in HD. It will be of interest to test if there is preferential cell toxicity in neurons expressing striatal markers, as contrasted with other neurons in the culture. This kind of preferential toxicity was seen for dopamine neurons in a recent report on iPSCs derived from a patient with a LRRK2 mutation (Nguyen et al., 2011
). However in HD, unlike Parkinson’s disease, there is widespread neuronal dysfunction and death, especially in cases with long repeats, so cell toxicity may not be limited purely to striatal neurons.
Our study is notable for detecting clear CAG expansion-dependent phenotypes, including cell toxicity, as would be expected for a neurodegenerative disease such as HD. We observed three variations on the relationship between phenotypes and the length of the CAG expansion in cells differentiated from HD iPSCs compared to controls. Some phenotypes were present in the HD lines in a graded fashion that correlated with the length of the CAG expansion. Others were only found in lines with the longest CAG expansion. Still others were present to a similar extent in lines with HD-associated CAG expansions of any length. The ability to detect a relationship between the length of the CAG expansion and dysfunction induced by mHTT will likely depend on a number of factors including the sensitivity and dynamic range of a particular assay and the cell type being studied. Our data would also be consistent with a model of HD pathogenesis in which the number of affected pathways and, in some cases, the extent of their dysfunction varies with the length of the CAG expansion. The CAG-dependence of neurodegeneration could therefore be an emergent property of the cumulative effect of a multifarious network of pathways affected by mHTT.
In conclusion, we have developed and characterized an iPSC model of HD that includes multiple lines, clones and repeat lengths. Future experiments using an allelic series of of cell lines with a range of expanded repeat lengths will help define the expanded- repeat length dependence of different phenotypes. The utility this new model system includes elucidation of HD cellular pathogenesis, development of HD-specific biomarkers, and ultimately screening for small molecule or other therapeutic interventions.