Induced Pluripotent Stem Cells in Huntington’s Disease Research: Progress and Opportunity

The current inventory of HD iPSCs

Early efforts at reprogramming adult cells to iPSCs relied upon lentiviral [3] or retroviral [1] delivery of cDNAs encoding pluripotency factors [4]. iPSCs have been successfully created from a variety of somatic cell types including fibroblasts [1, 3], blood cells [5, 6], renal epithelial cells [7], and keratinocytes [8, 9]. Several reviews discuss the history and recent advances in reprogramming methods used to produce human iPSCs [10–13]. At present, the best characterized HD iPSC lines have been produced from patient fibroblasts using lentivirus [14–16] or retrovirus [2, 17–28] to express a combination of pluripotency factors, including: Oct3/4, Klf-4, Sox2, c-Myc, SSEA4, LIN-28, NANOG, and p53 shRNA (to increase efficiency). However, the potential for off-target effects due to random viral insertions motivated scientists to develop novel non-integrating approaches for delivery including Sendai-virus [29], adenovirus [30] and episomal vectors [6, 31], as well as RNA transfection [32], protein [33], and small molecule [34, 35] based methods [4]. More recently, HD researchers have created and begun to characterize cells produced using non-integrating, episomal vectors for induction of pluripotency [36–42].

Numerous iPSC lines exist with CAG repeats in the range of wild-type to that of HD in the Huntingtin gene (HTT) (from 17 CAGs to 180 CAGs) [2, 36–43]. Many of these cell lines were created by the HD iPSC Consortium and are available through the newly established NINDS Human Cell and Data Repository (NHCPR) including 8 unaffected and 18 HD iPCS lines (the catalog for cell lines can be viewed at the following website: https://stemcells.nindsgenetics.org/). The best described HD iPSC line called HD4 was produced by Park et al. (2008) and contains 72 CAGs [2, 28]. HD4 was used by the Ellerby laboratory to create two cell lines corrected at the HD locus to 21 CAGs by homozygous recombination [17]. HD4 was also used to create an HD allelic series with 21, 72 and 97 CAGs using gene editing employing the CRISPR/Cas9 system along with an antibody screen to confirm the presence of an expanded polyglutamine region in cell lines [44]. These cell lines together are very useful because they offer an isogenic background on which to delineate effects of the HD mutation. Although isogenic lines are a “gold standard” for a well-controlled iPSC experiment, the inherent variability that has been found among control iPSCs substantiates a need for more isogenic lines from additional HD iPSCs.

iPSCs derived from individuals with juvenile onset HD (>60 CAGs) have been used more frequently for genomic and proteomic studies than iPSCs from individuals with adult onset HD (39–60 CAG) [2, 36–43]. However, only about 5% of HD patients have a juvenile onset (prior to age 20) form of the disease, with associated CAG repeat lengths greater than 60 [45–47]. Tables 1 & 2 show summaries of gene expression and protein changes based upon available data sets. The majority of reported work with iPSCs from those with adult onset HD has been with a single cell line derived by the HD iPSC Consortium in 2012 [15, 36]. This line has 60 CAG repeats which is at the high end of the CAG spectrum for adult onset HD [46, 47]. Additional cell lines with repeats ranging from 43–60 CAGs have been described, they are less well characterized [22, 43]. This is likely in part due to the greater ease of detection of mutant huntingtin protein with longer polyglutamine repeats using antibodies that preferentially recognize extended polyglutamine stretches, and the possibility that longer CAG expansions may give more robust changes, as they do in animal models. At present no iPSCs have been produced from patients bearing “intermediate” CAG repeat lengths (27–35 CAGs) or patients with 36–39 CAGs and reduced disease penetrance. Future work with iPSC derived neuronal cells could provide insight into the molecular changes underlying the behavioral changes identified in intermediate repeat length patients [48], and the differences in disease penetrance with low repeat lengths.

Search for phenotypes unveils numerous CAG repeat length dependent and independent properties

Several common phenotypes exist for neural stem cells (NSCs) (also called neural precursor cells (NPCs)) and neuronal cultures differentiated from adult and juvenile HD iPSCs. NSCs from both adult and juvenile-onset patients have decreased metabolic rate (respiratory capacity or ATP/ADP ratio) [15, 19], reduced filamentous actin protein and cell-cell adhesion [15], as well as parallel changes in gene expression compared to NSCs from controls of genes related to nervous system development, cell assembly and organization [15]. HD neuronal cultures derived from both adult and juvenile-onset iPSCs likewise have increased levels of cell death, caspase 3/7 expression in response to BDNF withdrawal, and altered gene expression in pathways related to cell function/signaling and tissue development [15]. One caveat, however, is that differentiated neuronal cultures are by nature composed of a heterogeneous mixture of cell types and the cell death discovered in this study was not demonstrated to be specific to neurons.

Some unique differences were identified when NSCs from adult onset HD iPSCs were compared to those from juvenile onset HD (Tables 1 & 2). Pathway analysis of gene array data showed alterations in axonal guidance in NSCs from juvenile HD (72, 109 and 180 CAGs) that were not present in NSCs from those with adult onset (60 CAGs) [15, 19]. The axon guidance changes were further associated with altered netrin and netrin receptor expression in 72 CAG NSCs [19]. Analysis of gene array data also indicated the presence of alterations in calcium signaling in an HD NSC line with 60 CAGs, but not in ones with 109 or 180 CAGs [15]. These changes in gene expression may have functional relevance in corresponding iPSC induced neuronal cultures: increasing glutamate levels was associated with increased calcium dyshomeostasis and cell death in neuronal cultures with 60 CAGs but not in those with 180 CAGs [15]. As mentioned above, neuronal cultures were heterogeneous and these changes also were not demonstrated to be specific to neurons. But, altered calcium signaling after glutamate treatment was also observed in low repeat (37 and 51 CAGs) embryonic stem cell (ESC)-derived forebrain neuronal cultures [49]. Finally, Mattis et al. (2015) identified a Nestin-positive cell population sensitive to BDNF-withdrawal induced cell death that was present only in neuronal cultures from juvenile-onset cell lines (109 and 180 CAGs) but not adult onset (60 CAGs) [36].

Phenotypes that may be relevant to disease have been identified in naïve and differentiated HD iPSCs. These include conditions related to altered cell growth [18, 24], adhesion [15, 17], differentiation [18, 36], protein clearance (proteasomal, autophagic) [15, 40], survival [15, 38], oxidative stress/antioxidant response [18, 41], metabolism [14, 19], growth factor signaling [15, 41] mitochondrial fragmentation [14], and electrophysiological properties [15]. Tables 3 & 4 list known phenotypes of HD cell lines. Conclusions about HD iPSC phenotypes have been drawn from a maximum of three HD cell lines compared to one or two control lines in a given publication [15, 42], with many groups using a single cell line or pair of cell lines compared to controls [14, 43] (see also Tables 1–4 for number of cell lines used in specific studies with their corresponding references). A major problem therefore in phenotype characterization has been the limited efforts to validate results in multiple iPSCs with similar CAG repeats or a range of CAG repeat lengths.

In future studies, one solution to this problem is to use multiple control and HD cell lines. The number of cell lines used should be determined using a power analysis based on data from individual phenotypes, just as in animal studies. Robust phenotypes may require testing far fewer iPSC lines compared to that necessary for more subtle phenotypes. We recommend comparing at least three control and three HD lines lines established from different individuals as an absolute minimum even for robust phenotypes. Subtle phenotypes may require as many as 10–12 cell lines per group and may require HD lines to have very similar CAG repeats. Ideally at least one genetically corrected iPSC paired with its isogenic HD iPSC should also be included, but two corrected clones are preferred to control for off target effects of genome editing. Alternatively, a huntingtin lowering strategy could be used (discussed later in “Maximizing the potential for research with HD iPSCs”). It would also be helpful if researchers currently growing other iPSCs could test published phenotypes in their cells and report conformational studies. Although these types of studies may not be novel and thus may seem of low importance, reproducible HD phenotypes in human cells validated across laboratories is of high value. Several of the current phenotypes have also been shown to occur in animal HD models, increasing confidence that they are real changes that may be important for HD; however, even for changes already shown in other HD models, demonstrating these changes in just one human HD iPSC line is of limited use.

Varied methods create mixed results: comparing apples to apples or apples to oranges?

Considerable variability exists in the culture methods used to reprogram somatic cells, and to derive NSCs and neurons from HD iPSCs. This limits the ability to draw broad conclusions from the existing published data. Though reprogramming techniques have been consistent: 17 of 24 publications HD cell lines were produced by retroviral or lentiviral reprogramming, the terms ‘NSC/NPC’ and ‘neuron’ were used to describe cell populations that varied in a multitude of ways including: protein markers of differentiation, growth factor dependence, and surface matrix [50–54]. The marked differences that exist in culture protocols can be explained to some extent by the ongoing progress made in the iPSC field to improve differentiation efficiency and specificity. Nonetheless, as with iPSCs from patients with other neurodegenerative diseases, the relative homogeneity of cell lines (i.e. few patient donors or few cell lines used during comparisons), when combined with variability in reprogramming and differentiation methods, may disguise subtle but important phenotypes, and make it challenging to determine which phenotypes are disease-relevant rather than due to culture conditions, clonal variation, or genetic background of the patient [55].

There are several protocols for differentiation of iPSCs favoring certain neuronal subtypes including medium spiny neurons (MSNs), the GABA-ergic projection neurons of the striatum that is particularly vulnerable in HD [15, 56–64]. Table 5 summarizes salient features of MSN differentiation protocols. The source of MSNs in the developing human striatum is the lateral geniculate eminence (LGE) of the ventral telencephalon [65, 66]. LGE precursors are characterized by expression of several transcription factors including: GSX2, DLX2, ISL1, BCL11B(CTIP2), FOXP1, and FOXP2 [67]. Mature MSNs are characterized by the co-expression of several protein markers including: DARPP32, BCL11B, FOXP1 or FOXP2 [68, 69].

Driving the differentiation of MSNs from human pluripotent stem cells requires the manipulation of several signaling pathways, including Wnt/β-catenin, and Sonic Hedghog (Shh), and bone morphogenic protein (BMP) well described in a recent review by Fjodorova et al. (2015) [70]. The inconsistent and frequently limited efficiency of existing MSN differentiation protocols, ranging from (5–80% MSN yield, see Table 5) is in part due to the complexity of factors that combine to specify MSN fate. Differentiation of MSNs from iPSC is a lengthy process, with protocols ranging from 3–16 weeks. Production of MSNs from iPSCs requires the derivation of LGE-like neural precursor cells, rather than those approximating the medial or caudal geniculate eminence, the maturation of the neurons rather than maintenance of multipotent neural precursor cells or development of glia, and finally the designation of MSNs rather than the olfactory bulb interneurons [65].

The general pattern used in the currently available MSN differentiation protocols includes neural induction using stromal cell co-culture [24, 56], exposure to growth factors to induce embryoid body/neural rosette formation [15, 71], or dual-SMAD/BMP inhibition [62]. Neural induction is followed by specification of LGE precuror/MSN fate using a combination of factors including Shh and DKK1 (a WNT-inhibitor) [15, 72], or more recently Activin A, a TGFβ signaling protein [62]. Alternatively, a recent protocol describes the efficient direct induction of adult human fibroblasts to DARPP32+ neurons using lentiviral infection with miRNAs: MiR-9/9* and miR-124 as described by Yoo et al. [73] to encourage neuronal differentiation, and expression transcription factors: CTIP2(BCL11B), DLX1, DLX2, and MYT1L (CDM) to specifically encourage MSN fate [61] (Table 5).

MSN differentiation protocols have largely been described in control cell lines [57–64], with only a few attempted using HD iPSC lines [14, 37]. Further study is necessary to understand how individual cell lines, HD or control, will respond to each differentiation protocol.

Drug discovery and the use of stress induced phenotypes

Drug screens using animal cell models of HD that express endogenous mutant huntingtin (knock-in) or overexpress mutant huntingtin exogenously, as well as those using human cells overexpressing huntingtin, have failed to identify a compound that prevents, slows or reverses disease onset. A well-characterized human neuronal cell model with huntingtin expression from the endogenous gene locus could be a valuable asset in developing a therapeutic. Numerous screens have already been performed using HD iPSC derived cells [14, 43]. Table 6 describes initial efforts using HD iPSCs and derived neuronal cells as models for drug screening. Interestingly, in contrast with the analysis of gene/protein changes and characterization of disease relevant phenotypes which were mostly performed in NSCs bearing juvenile repeat lengths (>60 CAGs), experimental screens have largely focused on neuronal cultures from HD iPSCs, and have included low repeat cell lines (43–47 CAGs) [14, 43]. However for each study, testing is often limited to one or two HD iPSC lines, often without a control or corrected cell line for comparison.

Most phenotypic readouts used to test the effectiveness of novel therapies in HD iPSC derived neuronal cells include a stressor such as growth factor withdrawal, oxidative stress, DNA damage, or glutamate toxicity [14, 43] (Table 6). The effect of a drug on cell response to cytokine treatment [26], induced DNA breakage [37], proteasome inhibition [25], H₂O₂ or Mn^2 + treatment [37, 43] and growth factor withdrawal [27, 39], have been investigated. The TNFα inhibitor XPro-1595 lowered cytokine (TNFα and ILβ) induced apoptosis activation and iNOS levels in astrocytes and neurons, respectively, differentiated from iPSCs with 43 CAGs [26]. Adenosine receptor 2A agonists CGS-21680 and APEC produced a dose dependent reduction of oxidative stress toxicity induced by exposure to H₂O₂ in 43 CAG neuronal cultures, as measured by decreased γH2AX induction and caspase3 cleavage [43]. The microRNA 196a (miR196a) was identified as a candidate that might impact HD using microarray data from transgenic HD monkey tissue; miR196a was found to decrease MG-132 induced EM-48 positive huntingtin aggregates in neuronal cultures differentiated from HD iPSCs with 72 CAGs through an unknown mechanism [25]. An ATM (ataxia-telangiectasia mutated) protein inhibitor KU55933 reversed both neocarzinostatin (a DNA damaging agent) induced increases in phosphorylation of p53, CHK2 and γH2AX, and Mn^2 + decreases in p53 phosphorylation in 70 and 180 CAG “striatal-like” NPCs [37].

Several groups have tested drug effects of growth factor (BDNF or FGF/LIF) withdrawal on cell survival [19, 39]. Although cells were grown in medium containing essential nutrients, acute withdrawal of neurotrophins might also be considered a stress once cells have become dependent upon them. Lu et al. (2014) showed that an alternate ATM protein inhibitor, KU60019, reduced BDNF withdrawal-induced increases in TUNEL-positive nuclei and caspase 3 activation in 109 and 180 CAG mixed neuronal cell cultures [38]. Furthermore, to elucidate mechanisms of BDNF-withdrawal induced cell death in neuronal cultures differentiated from HD iPSCs with 109 and 180 CAG, Mattis et al. (2015) showed reversal of increased TUNEL+nuclei after BDNF withdrawal using several compounds including a calcium chelator (BAPTA), TRKB antibody agonist (αTrkB), MAPK signaling inhibitor (SB2390463), NMDAR antagonist (memantine) and AMPA/Kainate antagonist (CNQX), and allele specific antisense-oligonucleotides of mutant huntingtin (ASO1, ASO2, ASO3) [36]. Importantly, the susceptible population in the neuronal culture was Nestin+ radial glia-like cells, not neurons.

Likewise, Yao et al. (2015) explored the potential role of G-protein coupled receptor 52 (GPR52) in mediating mutant huntingtin protein toxicity. They found that siRNA knockdown of GPR52 both reduced mutant huntingtin protein levels, and BDNF withdrawal induced caspase 3 activation in HD iPSC derived neurons with 47 and 70 CAG repeats [27]. In addition, Dickey et al. (2016) described the ability of a PPARδ activator (KD3010) to reduce numbers of condensed nuclei after BDNF withdrawal from 60 CAG derived neuronal cultures [39]. Finally, Ring et al. (2015) assessed the ability of recombinant TGFβ1/TGFβ2 and netrin-1 to reverse increases in caspase3/7 activity and decreases in maximum respiratory capacity that occur after FGF/LIF withdrawal in NSCs with 72 CAGs; TGFβ1/TGFβ2 treatment was likewise found to be beneficial to HD NSCs under unstressed conditions [19].

Aside from the TGFβ experiment by Ring et al. [19], only two other groups have tested the effects of a compound on iPSC phenotypes without the addition of a stressor. Charbord et al. (2013) demonstrated that treatment of 72 CAG NSCs with a Repressor element-1 silencing transcription factor (REST) inhibitor X5050 produced increased RE1 gene expression and reduced REST activity when compared to control ESC derived NSCs [20]. Guo et al. (2013) showed that a Drp1-selective peptide inhibitor, P110-TAT, reversed: neurite shortening, decreases in mitochondrial membrane potential and ATP level, increases in cell death, mitochondrial fragmentation, and mitochondrial reactive oxygen species (ROS) in mixed neuronal cultures derived from iPSCs derived from juvenile HD patient fibroblasts of unreported CAG length (GM05539, Coriell Cell Repository) [14].

In other disease fields, additional phenotypes have been identified using alternative methods of aging iPSC derived neurons well described in a recent review by Studer et al. (2015) [74]. Recently, exogenous expression of progerin, a mutated form of the nuclear envelope protein lamin A associated with disorder of premature aging, has been used to mimic aging [75]. Overexpression of progerin in iPSC derived dopaminergic neurons resulted in elevated levels of reactive oxygen species (ROS) and DNA damage consistent with the neuronal aging [75]. The same experimental procedure was also used in iPSC derived dopaminergic neurons from Parkinson disease (PD) patients bearing PINK1 and Parkin mutations, and expression of progerin unveiled a decrease in dendrite length specific to PD patient neurons, not identified in progerin-negative cells [75]. The introduction of progerin, along with continued work to understand the molecular mechanisms underlying normal aging could lead to an improved understanding of HD neurodegeneration and future cell models for drug screening. Care should be used with interpretation of results identified using progerin overexpression, however, since cells will be burdened with two mutant proteins, one of which is overexpressed. Furthermore, iPSCs from patients with Hutchinson-Gilford progeria syndrome (HGPS) and atypical Werner syndrome (AWS) which contain mutations in their endogenous LMA gene, have dysmorphic nuclei and premature senescence [76].

Phenotypes relating to disease mechanism

It is possible that aging and stress might compound or enhance secondary effects of mutant huntingtin but not affect primary drivers of HD progression. If the underlying primary problem causing disease in humans HD is not related to stress or aging, then screening for compounds that reverse stress induced cell death or age related morphology changes may be insufficient to identify novel therapeutics for HD that target crucial changes. Assessing the ability of a given therapeutic to reverse additional alterations such as cell morphology, cell signaling activity or gene expression in the absence of stress may allow for more confident conclusions about drug efficacy [77]. Additional phenotypes useful for screening could include any of the number of disease relevant phenotypes already identified in HD iPSC derived NSC/NPCs and neurons, though further work is necessary to determine the extent to which these phenotypes can be generalized using HD iPSC lines derived from a larger patient population (Tables 3 & 4).

Utilizing HD iPSCs as a model for early CNS development in HD patients, and determining phenotypes that precede neuronal death may be an equally important and a more readily attainable goal. Study of rodent and human cell models indicate the influence of mutant huntingtin on neural development in general, and specifically related to: premature onset of neural differentiation [78], altered Notch signaling [78] and aberrant mitotic spindle orientation [79, 80]. Recent work has indicated alterations to HD patient brains years before symptom onset, including: basal ganglia cell death [81], altered striatal/cortical volume and morphology [81, 82–84], cortico-striatal connectivity [85, 86], decreased cortical inhibition [87], and increased oligodendrocyte number [88, 89]. Early cognitive changes have likewise been identified in HD patients related to altered cortico-striatal circuitry [90]. Behavioral changes and altered synaptic connectivity, preceding the onset of motor symptoms, have also been identified in transgenic mouse models of HD [91–94].These findings have encouraged the hypothesis that the huntingtin mutation may alter molecular pathways, cell phenotypes and activity during neural development, and the cumulative effects of these changes over time eventually lead to disease onset well described in several reviews: [50, 96].

Genomic and proteomic analyses of HD iPSC and derived neuronal cells support the importance of continued exploration of the role of mutant huntingtin in disrupting the normal differentiation and maturation of human cells. Differentiation of HD iPSCs to NSCs and neurons has uncovered differences in the efficiency [18] and composition of cultures [36], when compared to iPSCs from healthy controls. Furthermore, comparisons of gene microarray and RNAseq data between HD4 (72 CAGs) HD iPSCs and a cell line corrected to 21 CAGs identified axonal guidance and cadherin, TGF β and SMAD3 signaling [17, 19] as pathways significantly altered in HD cells. Altered expression of genes specifically related to striatal development were also identified in the same cells [19].

Other brain cell types may contribute to HD pathology. Changes to astrocytes, microglia, and oligodendrocytes have been observed in both HD patients [89, 97–101], and transgenic animal models of HD [101–108]. Several protocols have been shown to produce astrocytes [109–114] and oligodendrocytes [115–118] from human iPSCs (45). Two groups have used HD iPSCs to produce GFAP+astrocytes [26, 40] (Tables 3 & 6). Juopperi et al. (2012) reported a CAG-repeat dependent increase in cytoplasmic vacuolization in HD astrocytes derived from 50 CAG and 109 CAG iPSC lines, and related the change to a potential alteration in autophagy as overnight treatment with chloroquine, an autophagy inhibitor, increased LC3+ positive vacuoles in HD cells [40]. Hsiao et al. (2014) showed that a TNF-α inhibitors, XPro-1595, lowers cytokine (TNF-α & IL-β) induced iNOS production in astrocytes derived from iPSCs with 43 CAGs [26]. So far, no work has been published describing HD iPSC derived oligodendrocytes. Further exploration of HD iPSC derived glia, along with co-culture of iPSC derived neurons and glia, may shed light on sources of non-cell autonomous toxicity in HD pathogenesis.

Technical caveats that may undermine phenotype characterization

Despite their advantages, unanswered questions about iPSC as a model remain. One area of particular interest for investigating age-related diseases such as HD is obliteration of age from the patient sample. Several studies indicate that induced pluripotency reverses typical age-related phenotypes, such as telomere shortening [119–121] and mitochondrial dysfunction [74, 122]. A total reversal of age related changes such as alterations in levels of oxidative stress, DNA packaging and damage, nuclear morphology and related gene expression changes occurred in fibroblasts differentiated from iPSCs that had been collected from donor patients in different age ranges [74]. Transdifferentiation or directed differentiation to neurons from aged fibroblasts may be one approach to maintaining age [4, 77]. Another issue is the potential for reprogrammed cells to bear an ‘epigenetic memory’ of their somatic cell source [4, 123–125]. A recent study by Kim et al. (2011) indicated that iPSCs retain some residual DNA methylation from their parental cell type (fibroblast or blood cell), rending them more readily differentiated to their parent cell’s fate [124]. However, recent work also suggests that repeated passaging [126] or chromatin modifying agents [127] can diminish or fully reset this ‘epigenetic memory’ [124].

The extent to which the phenomenon of epigenetic memory and limited pluripotency may affect neuronal cells derived from of HD iPSCs is as yet, uncertain. Reprogramming, continuous passage (cell dissociation and re-culturing) to high passage number, and differentiation of iPSCs have been shown to lead to genetic instability [128–132]. Investigators using iPSCs are thus cautioned when using highly passage cells. Encouraging results showed few changes in human iPSCs due to mutant huntingtin. The HD iPSC Consortium (2013) reported an expansion from 110 to 118 CAGs in one HD iPSC derived NSC line, but no changes in iPSC karyotype [15]. Camnasio et al. (2012) reported occasional changes in karyotype in both control and HD iPSCs but stable CAG repeat lengths [16]. This is in contrast to NSCs established from embryonic brain tissue from homozygous CAG140 knock-in mice which became tetraploid after very few passages in culture [133]. However, a recent finding by Tidball et al. (2016) showed increased somatic instability in HD iPSCs using siRNA targeting p53 to increase the efficiency of reprogramming [42]. Further study is necessary to determine the extent to which additional changes occur in human HD iPSC lines and differentiated cells, but so far the CAG repeat expansion and presence of mutant huntingtin protein do not seem to impose genetic instability.

While progress towards understanding HD mechanisms using iPSC has been unquestionably substantial, there are challenges in drawing broad conclusions about disease pathogenesis from these studies. Limitations on the number of HD and control cell lines used are understandable considering the cost and effort in maintaining human stem cells. However, given the relatively small number of cell lines studied, care should be taken when interpreting results. Although HD is caused by just one gene, the patient donors have diverse genetic backgrounds that could account for many of the changes observed. As with any experiment, sample size increases confidence in results. Efforts should be supported for additional testing of existing cell lines and for continued creation of new cell lines to overcome this limitation. Furthermore, determining the presence of certain phenotypes and susceptibility to specific treatments one day may be possible using iPSC derived neurons allowing for tailored treatments (so called “personalized medicine”) [134]. The availability of cell lines from diverse genetic backgrounds is critical for these efforts.

Maximizing the potential for research with HD iPSCs

Genetic correction to create non-disease CAG repeat lengths in iPSCs with the same genetic background as diseased cells in theory removes at least one confounding variable thereby diminishing barriers to identifying less subtle phenotypes. An et al. (2012) reported the successful use of homologous recombination to correct the 72 CAG allele of an HD4 iPSC line to 21 CAG [17]. Recent advances in gene editing technology should make genetic correction of iPSC easier [50–52, 135–138]. The potential for zinc finger nucleases (ZFNs), transcription activator-like effector nucleases (TALENS), and clustered regularly-interspaced short palindromic repeats/Cas 9 (CRISPR/Cas9) to modify human stem cells has been well described [4, 138]. Each method bears limitations in the targeting of guiding DNA/RNA sequences, as well as differing propensities for off-target effects [55, 138], although off target effects are reduced in pluripotent stem cells in contrast to other cells [139]. Current methods of gene editing rely on endogenous machinery to repair DNA breaks produced by an exogenous nuclease (i.e., CRISPR or Fok1) [4, 138]. Several recent findings indicate that the molecular pathways underlying DNA repair may be altered or aberrantly active in HD iPSCs [18, 43], consistent with somatic CAG repeat expansions in post-mortem brain tissue from HD patients, and oxidative DNA damage in HD animal models [140]. These existing changes in HD cells could have adverse effects on the efficiency of genetic corrections, and susceptibility of cells to off-target effects. There are ongoing efforts to target single nucleotide polymorphisms (SNPs) specific to the mutated HD allele and to contract the expanded CAG repeat in heterozygous cell lines [53] in order to create a series of iPSC lines with different CAG expansions lengths on the same genetic background to determine CAG repeat length dependent phenotypes. An HD allelic series 21, 72 and 97 CAGs, was successfully created by An et al., (2014) using the CRISPR/Cas9 gene editing system [44]. One practical problem is that unfortunately many of the existing iPSC lines are not heterozygous for SNPs that could be targeted by CRISPR/Cas9 making allele-specific changes difficult in these lines (our unpublished observations).

Just as an ‘epigenetic memory’ of the somatic cell source may persist in iPSCs after reprogramming, it is as yet unknown if contracting expanded CAG repeats in HD iPSCs will fully reverse the effects of the HD mutation on derived neuronal cells. Using a genetic correction as the control will obscure phenotypes that cannot be reversed since both the mutant and the isogenic corrected line will continue to share the condition; thus they will not be observed as “different”. Analysis of gene and protein expression and epigenetic modifications, comparing both HD iPSC line and iPSC lines derived from healthy patient controls is necessary to understand the extent to which genetic correction in HD iPSC lines can reverse HD phenotypes; it is possible that some phenotypes can be reversed in one cell line with a particular genetic background, but not in another ([for in-depth coverage of this topic please see [141]). For HD, only one iPSC (HD4) has been subjected to correction. Correction by homologous recombination of the iPSC line HD4 from 72 CAGs to 21 CAGs reversed phenotypes of altered TGF-beta and cadherin gene expression [17] and changes to gene ontology categories including synaptic assembly, axonal guidance and SMAD3 signaling [19]. Far fewer changes were found between HD4 and the corrected line, versus HD4 and an unrelated control iPSC line, supporting the notion that an isogenic background is a good control. Studies comparing iPSC derived neurons from healthy controls to Parkinson patients bearing mutations in LRRK2 gene demonstrated that gene expression cluster profiles did not partition with the mutation, indicating the normal genomic variability had a stronger effect than the mutation [142]; meaningful changes were only found using an isogenic corrected control. Additional studies comparing phenotypes of HD4 and the corrected cell line with numerous control cell lines will be informative. These issues are particularly important in considering the potential of genetically corrected HD iPSCs as a tissue source for cell replacement therapy [45, 123].

As an alternative to genetic editing, levels of mutant protein can be lowered using methods that specifically target mutant huntingtin RNA stability and protein expression including: anti-sense oligonucleotides (ASOs), small interfering RNA (siRNA), short hairpin RNA (shRNA) [53], and microRNAs (miRs). Mattis et al. 2015 reported the ability of allele-specific ASOs to lower levels of mutant huntingtin expression and reverse BDNF-withdrawal induced toxicity in 109 and 180 CAG HD iPSC derived neurons [36]. As mentioned above, many existing iPSC lines are homozygous at SNPs that could be used for allele-specific targeting by known useful siRNAs, shRNAs, and miRs making allele specific knockdown of mutant huntingtin difficult or impossible using these methods. Zinc finger proteins (ZFPs) are also being developed (by Sangamo) to repress transcription of the HTT allele bearing expanded CAGs and thus specifically reduced mutant huntingtin protein levels.