Abstract
Huntington disease is a monogenic, autosomal dominant, progressive neurodegenerative disorder caused by a trinucleotide CAG repeat expansion in exon 1 of the huntingtin (HTT) gene; age of onset of clinical symptoms inversely correlates with expanded CAG repeat length. HD leads to extensive degeneration of the basal ganglia, hypothalamic nuclei, and selected cortical areas, and a wide range of molecular mechanisms have been implicated in disease pathology in animal or cellular models expressing mutated HTT (mHTT) proteins, either full-length or amino-terminal fragments. However, HD cellular models that recapitulate the slow progression of the disease have not been available due to the toxicity of overexpressed exogenous mHTT or to limitations with using primary cells for long-term studies. Most investigations of the effects of mHTT relied on cytotoxicity or aggregation end points in heterologous systems or in primary embryonic neuroglial cultures derived from HD mouse models. More innovative approaches are currently under active investigation, including screening using electrophysiological endpoints, as well as the recent use of primary blood mononuclear cells and of human embryonic stem cells derived from a variety of HD research participants. Here we describe how these cellular systems are being used to investigate HD biology as well as to identify mechanisms with therapeutic potential.
Introduction
Huntington disease (HD) is a dominantly inherited neurodegenerative disorder caused by the expansion of a CAG repeat in the HTT gene, which encodes an amino-terminal stretch of polyglutamines. 1 HD is characterized by motor, cognitive, and psychiatric disturbances.2 –6 Despite its ubiquitous expression, the exact mechanisms whereby expanded HTT protein causes neuronal degeneration, particularly early in the disease progression within the medium spiny neurons (MSNs) of the striatum, are still unclear. Evidence suggests that the pathophysiology of HD arises from both cell-autonomous (directly within the vulnerable neurons) and non–cell-autonomous processes, most evident at the level of the cortical-striatal projections.7,8
The identification of the HTT mutation enabled the development of animal models of HD. Some of these models express either truncated or full-length human or mouse mHTT, and while they display some phenotypic differences, there remains a significant overlap in the mechanisms and pathways thought to be dysfunctional in human disease.9–14 As in the human disease, many of these rodent models display inclusion formation, widespread white matter atrophy, ventricular enlargement, some but limited striatal and cortical neuronal death, and transcriptional dysregulation.11,15–20 Recent advances in engineering large HD genetic models (e.g., mini-pigs, sheep, and nonhuman primates21–27) offer some advantages over rodent models due to their brain size, physiology, and brain circuitry that make them better suited for preclinical testing and safety evaluation, particularly for the development of molecular and viral-mediated therapies.28,29
While the presence of aggregates of mHTT aggregates in the brain of R6/2 mice 30 and human postmortem samples 31 has been described, their exact role in HD pathogenesis remains unknown. The processes governing oligomerization,32–40 the mechanisms by which they cause cellular dysfunction,8,15,41–57 and strategies to modify the aggregates or aggregation kinetics and manipulate the cellular mechanisms that would either ensure correct protein folding or eliminate misfolded HTT protein are areas of intensive research.58–63 Beyond aggregation or toxicity triggered by mHTT, other mechanisms are being explored in a variety of cell systems, as described below. The most common biological mechanisms that seem to be affected by mHTT expression include autophagy, the heat shock/stress response, transcriptional dysregulation, immune system alterations, energetic disturbances, and synaptic deficits, and each of these mechanisms requires defined (and mechanistically relevant) end points, both in cellular systems, as well as in vivo correlates, which need to be established. Significant progress has been made in all of these areas, some of which will be discussed below in the context of novel cellular models. This review addresses past and current screening efforts using cellular and invertebrate systems as a way to identify pathways or small molecules of potential therapeutic relevance to HD.
Previously described small-molecule or inhibitory RNA screens to identify molecules that reduce mHTT levels or pathogenic activity were mainly discovered using aggregation or cellular toxicity assays. In addition, most studies used an overexpression system, in lieu of the endogenous protein, typically expressing a toxic fragment of HTT with large CAG expansions, in the range known to cause juvenile HD (see Table 1 ).
Primary Chemical or Molecular Screens in Cellular or Invertebrate Models of HD Since 2005.
CFP, cyan fluorescent protein; CNS, central nervous system; EGFP, enhanced green fluorescent protein; FRAP, ; FRET, fluorescence resonance energy transfer; GFP, green fluorescent protein; HCS, high-content screening; HD, Huntington disease; HTT, huntingtin; LDH, ; MSN, medium spiny neuron; mTT, mutated huntingtin; NF-κB, nuclear factor–κB; NINDS, National Institute of Neurological Disorders and Stroke; TR-FRET, time-resolved fluorescence resonance energy transfer; YFP, yellow fluorescent protein; LDH, lactate dehydrogenase; FRAP, fluorescence recovery after photobleaching.
An ideal approach to identify mHTT relevant pathways would use a native system, expressing physiologically relevant levels of the full-length (FL) protein (and messenger RNA [mRNA]) derived from patient cells—fibroblasts, lymphoblasts, neuroglial cells derived from human embryonic stem cells (ESCs), or induced pluripotent stem cells (iPSCs), for instance. Unfortunately, many of the FL mHTT systems do not present robust cellular toxicity or mHTT aggregation end points for screening purposes. However, the endogenous FL context enables the discovery of small molecules or factors with the ability to alter mHtt at the transcriptional, translational, or posttranslational level. Until recently, technologies measuring quantitative changes in mHTT protein level or activity have not been available. These technical advances for mHTT quantitation will be discussed in later sections.
Small-Molecule Screens for HTT Aggregation or Cytotoxicity
The end points traditionally used to detect mHTT-driven effects in primary screens are described in Table 1 . These end points range from the quantification of aggregates formed by overexpressing mHTT proteins, either by direct detection of HTT using antibodies or via quantification of green fluorescent protein (GFP) aggregates if experiments employ fusion proteins with GFP,64–68 the utilization of reporter systems,69,70 or measurements of cellular toxicity. 64 More recently, a quantitative method for measuring mHTT by time-resolved fluorescence resonance energy transfer (TR-FRET) has been reported, which enables the direct measurement of mHTT levels in cellular systems.71,72
In many studies, only a fragment of mHTT (sometimes with a mixture of CAA/CAG repeats) was overexpressed to replicate the toxic effects seen in HD (see Table 1 and references within). Often, the mHTT fragments used have large polyglutamine repeats that are required to have a sufficiently large and robust assay window. This is in contrast to the average pathogenic lengths seen in most clinical cases of adult-onset HD. 73 As overexpression of mHTT is often toxic, some screens relied on inducible systems to prevent nonspecific toxicity and ability to propagate cells to compounds. The hits identified from such screens are often difficult to replicate in orthologous assays, due to artifacts derived from compound interference with the induction system or because of the use of heterologous systems. In some screens, end points were multiplexed to rule out compound-mediated cytotoxicity or assay interference, and nonexpanded HTT counterscreens were used to ensure hits were specific to expanded HTT expression and did not affect the nonexpanded HTT control. Oftentimes, cells that form aggregates do not die after expression of mHTT. It is our experience that the hits detected in toxicity and aggregation screens do not overlap, and most of the hits identified in such screens display intriguing cellular selectivity. However, this appears to be relevant only to the HD model employed. This cell-dependent context is important, as it points to the hits being relevant only in the context of the assay employed and not directly linked to conserved toxicity mechanisms directly triggered by mHTT activity, independent of cellular context.
Some large screening efforts have allowed assay development in highly efficient multiwell formats (384- or 1536-well plates), enabling the screening of large libraries (up to 2 × 106 molecules; see Table 1 ) for their effects on mHTT-dependent phenotypes. Published hit rates range between 1% and 5% of compounds screened (see Table 1 ), acceptable hit rates for cellular assays, enabling a quick progression to the more labor-intensive process of orthogonal assay hit confirmation and the assessment of the mode of action for such hits. In this latter process, a much higher rate of compound rejection is expected, due to a failure to reproduce activity in untagged or full-length mHTT expressing cell lines under a native promoter, undesired modes of action, poor physicochemical properties, or difficulties in identifying a precise mode of action. This has meant that little progress has been made in validating the output of these small-molecule screens and developing good proof-of-concept pharmacological tools for testing in HD animal models. In Table 1 , we describe the main publications reporting on small molecules identified via a variety of screens in yeast, invertebrate systems, or mammalian cells. Most of these include aggregation or toxicity end points. In most cases, few follow-up studies describing further development of the identified “hits” have been reported. An exception is the molecule C2-8 identified in Stott et al. 74 This molecule, originally identified as an inhibitor of aggregation in yeast, was shown to exert antiaggregation and protective effects in Drosophila, cultured neurons, and the R6/2 model of HD. 75 The current status of this molecule in terms of its potential development for HD, as well as any structure-activity relationship (SAR) associated with making this molecule more suitable for long-term chronic dosing studies, is unknown. A recent study showed that C2-8 had weak Sirt2 inhibitory activity, and several additional molecules were engineered to enhance potency against Sirt2. 76 The development status of these molecules is unknown to us. From a focused small chemical library screen in a Drosophila climbing assay, Wu et al. 77 identified a lead molecule (EVP4593) with inhibitory activity in the nuclear factor–κB (NF-κB) signaling pathway, presumably through the modulation of store-operated calcium entry. No further work has been described for additional testing or development of this small molecule.
The recent development of more sensitive and quantitative assays to measure wild-type and mutant HTT species,71,72 capable of accurately determining mHTT levels in native systems (fibroblasts, lymphoblasts, peripheral blood mononuclear cells [PBMCs], brain, and other tissues), suggests that renewed efforts to identify modulators of native mHTT expression will be more successful. Despite several reports of small molecules that can decrease aggregate formation in model systems, no promising candidates are currently in development. The lack of understanding of the molecular target of such small molecules strongly hinders drug optimization. We have conducted a small screen monitoring both aggregation effects in the frequently used inducible PC12 model of mHTT expression 78 ( Fig. 1 ), as well as in cytotoxicity screens (not published); we found no overlap between the small-molecule hits that affected the aggregation versus toxicity end points. In general, both for our hits as well as for all other reported molecules, we found that this approach was not fruitful. None of the molecules we identified or synthesized based on the existing publications advanced in a therapeutic trajectory due to their low tractability after SAR studies or because of the lack of robust, quantitative assays that translated between cellular systems. In general, despite the enthusiasm of these initial findings, we found a lack of validation for these molecules among different cellular backgrounds. One notable exception might be the ROCK inhibitors, which appear to reduce HTT aggregation in a variety of systems, including neuronal systems.69,79,80 Therefore, using primary cells or cells not overexpressing HTT might be a better strategy to conduct these types of phenotypic screens.
Aggregation assay in inducible PC12 cells, based on Apostol et al.
78
(
Molecular Screens in HD Cellular or Organismal Models
To identify proteins and pathways that modify mHTT toxicity, Miller et al. 81 carried out an siRNA screen in cells expressing the N-terminal 558 amino acids of mutant Htt (141Qs) fused to GFP; the screen identified multiple components of the RRAS signaling pathway as loss-of-function suppressors of mutant huntingtin toxicity in human and mouse cell models. Doumanis et al. 68 carried out a high-throughput RNAi screen for modifiers of aggregate formation in Drosophila melanogaster larval central nervous system (CNS)–derived cells expressing mutant human HTT exon 1 fused to enhanced green fluorescent protein (EGFP) with an expanded polyglutamine repeat (62Q), using ArrayScan (Thermo Scientific, USA) technology; newly identified modifiers included genes related to nuclear transport, nucleotide processes, and signaling. A high-content RNAi suppressor screen was also conducted by Schulte et al. 82 using a Drosophila primary neural culture HTT model. Using live-imaging technology tracking the subcellular distribution of fluorescently tagged pathogenic HTT, as well as assaying neurite branch morphology changes, the authors identified suppressors that could reduce HTT aggregation and/or prevent the formation of dystrophic neurites.
An analogous large-scale RNAi screen was conducted in Caenorhabditis elegans expressing N-terminal HTT in touch receptor neurons. 83 Network-based analysis revealed a subset of modifier genes in pathways of interest in HD, including metabolic, neurodevelopmental, and pro-survival pathways. Significantly, 49 modifiers of 128Q-neuron dysfunction were dysregulated in the striatum of either R6/2 or Q150 knock-in homozygous (CHL2) HD mice, or both. Teuling et al. 67 evaluated the effects of knocking down 186 genes on aggregation of an exon1-Q74-GFP fragment in HEK293 cells as a follow-up study on an initial genome-wide RNAi screen conducted previously in a C. elegans model for polyglutamine aggregation. 66 The authors show that 26 human orthologs of these genes suppressed aggregation of mutant HTT. Among these were genes that encode eukaryotic translation initiation, elongation, and translation factors and genes that had been previously associated with other neurodegenerative diseases, like the ATPase family gene 3–like 2 (AFG3L2) and ubiquitin-like modifier activating enzyme 1 (UBA1). Overall, the data obtained using molecular approaches appear more fruitful in identifying mechanisms of relevance across cellular systems and organisms than the chemical approach. This is perhaps not surprising but highlights the critical importance of choosing the right cellular context and of monitoring endogenous HTT levels and functional outcomes beyond cell death. Similarly, focusing on end points that are mechanistically relevant to the function of HTT or mHTT in native systems is likely to reveal mechanisms that will be shared across organisms and cellular contexts. The ideal scenario would be one where phenotypic small-molecule screens would be coupled with molecular screens to identify druggable and nondruggable (but informative) mechanisms, which might help in identifying the target(s) of the small-molecule hits.
Novel HTT Binding and Aggregation End Points Developed at the CHDI Foundation
The CHDI Foundation is working on the clinical development of HTT-directed molecular therapeutics with several biotechnology companies, using different technologies: antisense oligonucleotides (ASOs) with ISIS Pharmaceuticals (Carlsbad, CA, USA) Roche, siRNAs with Alnylam Pharmaceuticals (Cambridge, MA, USA)/Medtronic (Minneapolis, MN, USA), and zinc-finger protein (ZFP) repressors with Sangamo Biosciences (Richmond, CA, USA). 4 Some of the challenges of these programs include measuring HTT levels in the human brain or finding biomarkers for HTT expression lowering in patients using either cerebrospinal fluid (CSF) sampling or noninvasive technologies. CHDI has begun a program to develop a positron emission tomography (PET)–ligand to monitor HTT levels in the brain as a pharmacodynamic measure to aid in clinical development of these programs.
To monitor the effects of these therapies on HTT expression (e.g., by visualizing HTT aggregates in the brain via imaging ligands), we set out to develop quantitative methods to assess whether compounds were able to bind HTT directly. We adapted biophysical approaches to screen for compounds that bind directly and selectively to expanded polyQ (42 repeats) as part of a greater effort to develop a PET ligand directed to mHTT. Such binding assays interrogate direct binding or interactions with HTT aggregates (or monomers) and are independent of mechanisms that may require disruption, clearance, or reduction of aggregate, which have been associated with the previously reported cell-based type of assays. While seeking a PET ligand for mutant HTT that uses this “direct binding” approach does not necessarily invoke mechanisms that can directly reflect therapeutic intervention, further exploration of a “mutant HTT binder” may be reconfigured toward a reagent that modifies HTT aggregation. One such direct-binding approach, which was adapted from Okamoto et al., 84 uses surface plasmon resonance (SPR) technology to identify small molecules that interact and bind to polyQ peptides of normal (Q19) and expanded (Q46 and Q62) polyQ repeats; additional efforts extended these non-HTT proteins to exon1-containing HTT with normal (Q16) and mutant (Q46) polyQs with similar results. 85 To enable SPR on polyQ repeats, fusion protein constructs were produced that allowed for efficient purification of recombinant protein and immobilization on SPR chips. 86 Using a QBP-1 peptide as a positive reference molecule for SPR analysis, 86 we were able to identify high-affinity, Q46-selective small-molecule binders from a focused library screen containing chemotypes previously associated with interacting with amyloid proteins (e.g., β-amyloid protein, α-synuclein, prion proteins, tau protein). Details of this work will be published elsewhere.
To further characterize the small-molecules/HTT aggregate interactions, we established a solution nuclear magnetic resonance (NMR) assay, adapted from Klein et al., 87 to assess whether screening hits identified by SPR can modulate polyQ aggregation dynamics in solution. We also applied a conventional radioligand binding type of assay to interrogate the binding potential of small molecules to mHTT exon-1 aggregates. These cell-free assays allow us to identify real HTT or mHTT interactors, which can then be interrogated in a cellular context for effects on aggregation. One of the big hurdles in the aggregation screen field is the lack of understanding of the molecular target for the molecules and, typically, the inability to demonstrate that these molecules bind to HTT itself. By developing HTT-dependent biophysical assays that are highly sensitive, it is possible to identify nM or pM binders of mHTT. The likelihood of these HTT binders to progress through traditional SAR cascades is much more likely, since the molecular target is known from the outset. Traditional aggregation end points in cellular systems can then be included to assess the cellular activity of these small molecules.
Primary Culture Systems and Medium-Throughput Methods Using Slices Derived from Rodent Models of HD
Neurodegenerative disease drug screening using clinically relevant cells is hampered by either the limited availability of CNS tissue or the fact that primary neuroglial cultures are short-lived and obtained from late embryonic or early postnatal periods. HD brain tissues for molecular or biochemical studies are generally obtained postmortem, from very late-stage disease, where the regions of interest have severely atrophied.88,89 Alternatively, primary neuronal cultures from rodent models can be used for screening purposes. For HD, such cultures can be derived from wild-type rodent models and cells can be transfected or transduced to express wild type or mHTT; alternatively, these cells can be derived directly from various HD rodent models. In the former approach, Kaltenbach et al. 90 used primary rat cortical and striatal neurons transfected with Htt exon1 Q73 to monitor cell toxicity. The readouts observable in such a short period (days) are constrained by the experimental system and typically have been limited to mechanisms involving cell death or aggregation of overexpressed mHTT. However, some laboratories have developed sophisticated methods to increase the throughput of such systems, and other end points are being developed, such as cytoskeletal changes or DNA repair end points that query other biological consequences of mHTT expression. These methods, initially developed for primary rodent cells, will undoubtedly be applied to human embryonic stem cell–derived neuroglial populations, to enable studying effects of molecules or genes on human mHTT. 91 We continue to support the development of additional end points in co-culture neuronal systems that might be more indicative of HTT-dependent alterations, rather than monitoring cell death as a phenotypic end point. 90
Due to the immediate toxicity caused by overexpression of mHTT, an inducible model has recently been developed that allows for the maturation of the neurons and formation of synaptically active networks, therefore enabling development of more physiologically relevant and slower developing end points. These include neurite extension, calcium handling capacity, HTT expression/clearance, autophagy induction, and other readouts based on hypotheses about the mechanism of mHTT-dependent pathogenesis.4,6 While the use of inducible systems allows investigators to induce mHTT expression after synaptic networks are established in culture, the end points are likely to be affected by the level of overexpression. Therefore, culture systems derived from genetically engineered models would be more desirable to avoid confounds associated with overexpression of mHTT. In addition, acute or organotypic slice preparations can be incorporated to study the effects of mHTT in hippocampal or corticostriatal slices. These preparations offer several advantages: the relevant network architecture is preserved, the cells monitored by imaging technologies are adult as opposed to embryonic, and the electrophysiological properties typically correlate better with the functional properties of those cells in the intact brain. Although these techniques do not possess the relevant throughput for large-scale screens, they can be used for evaluating a small set of well-characterized small molecules to understand the involvement of their targets in a hypothesis-driven manner.
For instance, Reinhart et al. 92 conducted a small-molecule screen to identify mechanisms that could affect MSN death due to expression of mHTT or the appearance of mHTT-containing aggregates in wild-type rat corticostriatal slices where mHTT together with reporters had been delivered via a biolistic device. Slices were treated with compounds, and end points measured with an imaging reader. In this work, Reinhart et al explored the effects of several signaling pathways implicated in the immune and neural alterations observed in HD. These included the NF-κB, JNK, and adenosine pathways, among others. We have taken a different approach with acute slice preparations. Synaptic disturbances constitute some of the earliest detectable alterations in all HD models.93–95 Therefore, in collaboration with several academic laboratories, we characterized the electrophysiological properties of neurons in the corticostriatal and hippocampal systems derived from a variety of rodent models of HD, including fragment, full-length, and knock-in models. Generally, there is remarkable consistency in the types of alterations present in HD models, which enabled the evaluation of mechanisms that could restore these deficits toward a wild-type profile. To date, we have evaluated over 100 small molecules with known selectivity profiles in the hippocampal system, using a multielectrode-array system (MEA; see Johnstone et al. 96 and Society for Neuroscience abstract 148, Kleiman et al. 2012 97 ), which increases the throughput of electrophysiological readouts. MEAs have also been used to monitor cytotoxicity, and their application to a variety of neuronal networks makes this a useful platform to evaluate effects of pathway-modulating small molecules in a defined, electrically coupled network in culture.
While this approach is useful for the HD hippocampal system, the complexity of the basal ganglia circuitry makes this approach more challenging. We therefore used conventional patch-clamp techniques to ascertain which mechanisms could restore the progressive alterations determined in MSNs or layer V pyramidal cortical neurons in acute slice preparations. Through this approach, we were able to identify a selective class of phosphodiesterase (PDE) inhibitors that modulate cGMP signaling and restore the HD model system toward a wild-type profile. These studies culminated in the designation of a PDE10 inhibitor as a clinical development molecule, which, in collaboration with Pfizer, is scheduled to begin clinical trials in patients with HD in 2013. For a description of this work, see Society for Neuroscience abstract 148 97 or watch the following presentation from Pfizer (New York, NY) on the role of PDE10 inhibition in HD at the following link: https://vimeo.com/64569651. Additional mechanisms identified in this manner are currently under active investigation at CHDI.
In addition to using neuroglial cells, other primary cells could be used to investigate the molecular mechanisms triggered by mHTT expression. Notable examples include muscle and blood cells, which show mHTT-dependent changes. The ability to isolate human peripheral blood mononuclear cells (hPBMCs) from different stages of the disease and expanding them in culture enables investigation of mechanisms thought to affect disease progression, well beyond their traditional use in immune system investigation. An important study was recently conducted using HD hPBMCs that evaluated the role of the NF-κB pathway in the cytokine alterations observed in HD-derived macrophages/monocytes (Weiss et al. 72 and Tabrizi, personal communication, 2013). Importantly, in knockdown studies, these alterations in cytokine production were reversible upon decreasing HTT expression. The use of patient PBMCs has other important implications. Assuming the targeted mechanisms are operative in these cells, patient PBMCs could be used to monitor pharmacodynamic effects of drugs in clinical stages. Such mechanisms might include, for instance, energetic deficits, autophagy, or epigenetic, chromatin changes in patient cells derived at various stages of the disease. We think the incorporation of these patient-derived cells, applied in a more systematic manner, will provide new opportunities to explore such mechanisms in the context of the disease state.
Embryonic Stem Cell–Derived Cellular Models
Stem cells have the capacity to differentiate into any tissue type, including cells from tissues that are difficult to biopsy or access from living subjects. Murine ESCs (mESCs) can be generated from existing HD mouse models. Although their culture conditions and differentiation requirements may be more complicated and costly than traditional primary neuronal cultures, Baldo et al. 71 reported using HD mESCs and differentiated neurons to validate hits from a screen aimed at reducing HTT cellular levels. For the most part, the primary use of HD mESCs has been to study the role of wild-type Htt, as well as the impact of Htt CAG expansion, during development. We have developed mESCs derived from several mouse models, including knock-in mouse model lines, which can be used to generate rodent cellular models of disease. These lines are available for the HD community through CHDI (www.chdifoundation.org).
For example, various studies using Htt knock-out mESC have revealed that wild-type HTT modulates adenosine triphosphate (ATP) levels in cells and affects gene expression relative to mHTT, regulates neural rosette formation by modulating ADAM10 metalloprotease and N-cadherin levels and protects cells from apoptosis, is involved in cellular motility, and promotes neural induction and gliogenesis.98–103 CAG length-dependent or gene dosage studies in mESCs have helped to identify potential mHTT pathogenic mechanisms. For example, Ritch et al 103 compared neural stem cells (NSCs) derived from homozygous knock-in Q140 mice to address a variety of end points and reported slower growth rates of the Q140 NSCs, increased reactive oxygen species (ROS) production, decreased total cholesterol production, and decreased cell motility. Jacobsen et al. 100 describe CAG length-dependent alterations in gene expression pathways relevant to nucleotide, energy and lipid/sterol metabolism, cell cytoskeletal structure, and adhesion. Studies by Conforti et al. 98 reported mHTT-associated caspase 3/7 activation during neuronal differentiation. More recently, Nguyen et al. 102 reported a role for mHTT in promoting neural lineage commitment. These types of studies will undoubtedly allow the identification of potentially tractable intervention points from a therapeutic perspective, based on the role of wild-type and mHTT in more physiologically relevant systems. However, the challenges that remain include turning these end points into drug discovery efforts and comparing these findings across species, preferably in systems expressing human disease–causing alleles.
hESCs as a Model to Study HD
Given the exclusively monogenic nature of HD, this is perhaps the best disorder to realize the potential of hESCs in drug discovery applications. hESCs derived from HD embryos following preimplantation genetic diagnosis (PGD) have been reported by several groups.104 –106 Although PGD-derived HD hESCs provide a bona fide cell model system that recapitulates the unaltered genomic context and endogenous HTT expression, access to hESCs, as well as funding for hESC-related research, is often restricted and subject to fluctuating political, ethical, and regulatory dispositions. In addition, the existing WARF patents require commercial entities to purchase a license to use hESCs, which limits the broad applicability of these systems in a private setting in many countries. To promote hESC-based research, CHDI has negotiated a license covering for-profit entities working in HD research and development.
Using non-HD hESC, Lu and Palacino 107 recently established a stable line expressing an HTT exon1 fragment. Similar to the transgenic fragment rodent models with accelerated disease-related phenotypes, neurons generated from this line exhibited HTT aggregates and progressive cell death in culture. These phenotypes could be reversed by targeting HTT with siRNAs. Furthermore, knockdown of another gene target implicated in HD pathology, Rhes, 108 was also shown to improve cell viability. After comparison of gene expression profiles from a series of HD hESCs and NSCs, Feyeux et al. 99 identified a limited number of conserved dysregulated genes in the HD cells. To validate these findings, the investigators transiently transfected an mHTT fragment in control hESCs and showed a consistent reduction in PGC1α levels. Reduced PGC1α levels had previously been reported in HD rodent models and a contributing factor to the altered bioenergetics measurements observed in HD.11,57 Despite the genetic heterogeneity of hESC lines, these studies in mHTT fragment-expressing lines, as well as PGD-derived (preimplantation genetic diagnosis) HD ESCs, were able to recapitulate some of the phenotypes seen in rodent models, allowing future studies and screens to be carried out in more a relevant human cell-based context.
Human HD iPSC Phenotypic Studies
With the advent of iPSC technologies, it is now possible to generate human stem cells while minimizing the ethical confounds associated with traditional hESC sources.109–112 The first HD iPSCs were reported by Park et al. 112 in 2008, which are being made available for academic research through the Harvard Stem Cell Institute (http://www.hsci.harvard.edu/ipscore/). The National Institute of Neurological Disorders and Stroke (NINDS)–funded HD iPSC Consortium generated a series of HD iPSCS, including juvenile HD iPSC with 180Q, 113 which are being banked for broad distribution through the Coriell Institute (http://ccr.coriell.org/Sections/Collections/NINDS/ipsc_list.aspx?PgId=711&coll=ND). Although many groups have since generated HD iPSCs, Camnasio et al. 114 were the first group to generate iPSCs from homozygous HD subjects, while An et al. 115 were able to correct the HD mutation by homologous recombination, effectively creating the first isogenic human HD and control line.
Given the variety of mechanisms implicated in HD pathology, a logical starting point for using patient-derived stem cells would be to compare global gene expression and proteomic profiles between HD and control lines. Gene expression profiles at the pluripotent and NSC stages have been carried out by several groups,113,116,117 and alterations in various signaling pathways have been implicated. Of note was the study carried out by Feyeux et al. 99 using six independent HD hESCs, in which very few genes were consistently differentially regulated between the HD and control lines. Genetic heterogeneity of the starting material was likely a major contributing factor. In the study carried out by An et al. 115 using isogenic iPSC lines, gene expression differences should be solely attributable to the HD mutation. With the advent of zinc finger nucleases (ZFN) and TALEN technologies, it should be feasible to efficiently correct the mutation in HD cell lines and generate suitable isogenic controls for each HD line, thereby minimizing loss of signal due to genetic heterogeneity.
Proteomic analyses have been more limited, with Chae et al. 116 reporting alterations in oxidative stress and cytoskeleton in HD iPSCs, while the HD iPSC consortium reported changes in the axonal guidance signaling of NSCs. 113 Although preliminary, these global analyses should help identify HD relevant pathological pathways, druggable targets, and phenotypic readouts amenable to screening of larger collections. Given the widespread neuronal loss during the course of HD, signs of cell death continue to be a prominent phenotype for HD researchers. Zhang et al. 118 compared HD iPSC-derived neurons with control lines and reported enhanced caspase activation after growth factor withdrawal in the HD line. The HD iPSC consortium also reported various cell toxicity readouts, such as condensed nuclei, caspase activation, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) staining, and overt cell death following various stressors, such as growth factor withdrawal, exposure to H2O2, 3-MA (an autophagy inhibitor), and repeated glutamate exposure. In addition, more subtle phenotypes, reminiscent of cellular dysfunction, were also explored. HD iPSC-derived NSCs had altered ATP/adenosine diphosphate (ADP) ratios; showed deficits in cellular adhesion and cytoskeletal integrity; and had reduced intracellular calcium handling capacity following glutamate pulsing. 113 More recently, using HD hESC-derived forebrain neurons, Niclis and colleagues 105 noted enhanced intracellular calcium levels following glutamate exposure in the HD neurons. Overall, these early studies argue for the importance of the hESC systems to understand pathogenic mechanisms. The challenges include studying the biology of HTT in clinically relevant cells derived from hESCs and in how best to develop quantitative measurements that can truly enable a drug discovery campaign.
Directed Differentiation of hESCs
One limitation of the approaches taken to date has been the lack of robust differentiation protocols to differentiate stem cells into relevant neuronal subtypes. HD is characterized by overt neurodegeneration of the MSNs in the basal ganglia, which are DARPP32+ expressing GABAergic projection neurons. The selective vulnerability of MSNs makes these neuronal cells an attractive cell model for studying HD disease pathology. To that end, the search for a reproducible and robust differentiation protocol is ongoing, and each version is aimed at increasing MSN yields, refining the neuronal characterization, and decreasing the differentiation time. In 2008, Aubry and colleagues 119 published the first such protocol that yielded approximately 5% to 10% DARPP32 neurons but required more than 60 days in culture. Most recently, Delli Carri et al. 120 published a method for generating a higher yield of extensively characterized MSNs over a shorter period. The ability of hESC-derived neuronal (and glial) cells to survive in culture for many months makes it possible to study age-dependent effects in a human context, something not possible with rodent cells. The pluripotent nature of hESCs will in time also facilitate the study of mHTT effects in other cell populations of interest, such as corticostriatal projection pyramidal cells, skeletal muscle cells, or other nonneuronal cells relevant to disease pathology. For instance, astrocytes derived from HD iPSCs have been shown to harbor extensive cytoplasmic vacuoles, possibly affecting their function. 121 If robust cellular phenotypes are identified and sufficiently characterized to be “reversible” when mHTT expression is halted, and the methods to expand committed progenitors are optimized to enable drug discovery campaigns, the field of HD (and neurodegeneration in general) will experience a significant expansion in our understanding of the molecular mechanisms driving human phenotypes and of uncovering novel targets with treatment potential. To realize this potential, collaborative projects are needed between academic centers to allow for independent validation of findings in multiple hESC lines and with industrial partners with access to high-throughput methodologies and medicinal chemistry expertise. Although the field of chemical and molecular screens using mESC- or hESC-derived neurons is still emerging, recent work suggests that the ability to scale up these cells, engineer them for reporter expression, and automate screening methodologies will enable the incorporation of these cells in the drug discovery process.121–124 In particular, the use of pluripotent human cells or progenitors derived from HD stem cells, such as glial or medium spiny neuron progenitor cells, allows the rapid expansion and homogeneity needed for some of these applications. Other populations require further optimization, but the use of high-content imaging technologies will enable heterogeneous populations to be used. Perhaps more challenging than the expansion issues with hES-derived cells is the fact that the neuronal differentiation protocols can take between 60 and 120 days for many of these hES-derived cells to mature to the desired cell type, and that poses a challenge in their incorporation for high-throughput applications. It is likely that in this regard, a close collaboration between industry and academic scientists is needed to adapt differentiation protocols to an industrial platform to enable high-throughput applications using hESC or iPSC derivatives.
In conclusion, the field of HD has undergone tremendous progress in understanding pathogenic mechanisms and the development of novel therapeutics, many of which will be entering clinical trials in the near term. However, despite these advances, identifying critical intervention points in the biological pathways implicated in HD pathogenesis is still a challenge. The development of novel cellular systems that can recapitulate aspects of the disease from a pathway or biological mechanism perspective will be key in ensuring that the pipeline of novel therapeutic molecules for HD is robust and, it is hoped, more predictable of their in vivo biological effects. The maturity in research involving primary patient-derived cells and human embryonic stem cells, coupled with novel assay methodologies, will change how drug discovery in neurodegenerative disorders will be conducted. We think the HD field will be instrumental in leading this change.
Footnotes
Acknowledgements
The authors thank Simon Noble for his comments on the manuscript.
Declaration of Conflicting Interests
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: BioFocus conducted the research described through a fee-for-service agreement from CHDI Foundation. As employees of CHDI Management and advisers to CHDI Foundation, JB, JA, and IM-S were all intimately involved in the study design, data collection and analysis, decision to publish, and preparation of the manuscript.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.