Human Molecular Genetics, 2015, Vol. 24, No. 5 1420–1431

doi: 10.1093/hmg/ddu556 Advance Access Publication Date: 4 November 2014 Original Article

ORIGINAL ARTICLE Intermediate filament accumulation in motor derived from giant axonal neuropathy iPSCs rescued by restoration of Bethany L. Johnson-Kerner1,2,3,4,†, Faizzan S. Ahmad10,‡,¶, Alejandro Garcia Diaz1, John Palmer Greene1, Steven J. Gray7, Richard Jude Samulski7, Wendy K. Chung6, Rudy Van Coster8, Paul Maertens9, Scott A. Noggle10, Christopher E. Henderson1,2,3,4,5, and Hynek Wichterle1,2,3,4,*

1Project A.L.S./Jenifer Estess Laboratory for Stem Cell Research, New York, NY 10032, USA, 2Center for Motor Biology and Disease, 3Departments of Pathology and Cell Biology, , and Neuroscience, 4Columbia Stem Cell Initiative, 5Department of Rehabilitation and Regenerative Medicine, 6Department of Pediatrics and Medicine, Columbia University Medical Center, New York, NY 10032, USA, 7Gene Therapy Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA, 8Department of Pediatrics, Division of Pediatric Neurology and Metabolism, Ghent University Hospital, Ghent, Belgium, 9Departments of Pediatric Neurology, University of South Alabama, Mobile, AL, USA, and 10New York Stem Cell Foundation, New York, NY 10032, USA

*To whom correspondence should be addressed. Tel: +1 2123423929; Fax: +1 2123421555; Email: [email protected]

Abstract Giant axonal neuropathy (GAN) is a progressive neurodegenerative disease caused by autosomal recessive in the GAN resulting in a loss of a ubiquitously expressed protein, gigaxonin. Gene replacement therapy is a promising strategy for treatment of the disease; however, the effectiveness and safety of gigaxonin reintroduction have not been tested in human GAN cells. Here we report the derivation of induced pluripotent stem cells (iPSCs) from three GAN patients with different GAN mutations. Motor neurons differentiated from GAN iPSCs exhibit accumulation of neurofilament (NF-L) and (PRPH) protein and formation of PRPH aggregates, the key pathological phenotypes observed in patients. Introduction of gigaxonin either using a lentiviral vector or as a stable transgene resulted in normalization of NEFL and PRPH levels in GAN neurons and disappearance of PRPH aggregates. Importantly, overexpression of gigaxonin had no adverse effect on survival of GAN neurons, supporting the feasibility of gene replacement therapy. Our findings demonstrate that GAN iPSCs provide a novel model for studying human GAN neuropathologies and for the development and testing of new therapies in relevant cell types.

† Present address: Department of Pediatrics, University of California, San Francisco, 505 Parnassus Ave, M691, San Francisco, CA 94143, USA. ‡ Present address: Qatar Cardiovascular Research Center, Qatar Foundation, Qatar Science and Technology Park, Doha, Qatar. ¶ Present address: Department of Pharmacology, University of Oxford, Mansfield Road, Oxford OX1 3QT, UK.

Received: September 20, 2014. Accepted: October 27, 2014. © The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]

1420 Human Molecular Genetics, 2015, Vol. 24, No. 5 | 1421

Introduction In this study, we successfully generated GAN iPSCs from three GAN patients and differentiated them in vitro into spinal motor Disorganization of the neurofilament (NF) network is a feature of neurons (iPSC-MNs), a cell type strongly affected in patients. several neurodegenerative disorders, including amyotrophic lat- Strikingly, the GAN iPSC-MNs exhibit the intermediate filament eral sclerosis, Parkinson’s disease and axonal Charcot–Marie– protein accumulation characteristic of patients at early stages Tooth disease (1,2). Patients with giant axonal neuropathy (GAN) of the disease. Importantly, the GAN phenotypes can be rescued provide a particularly striking example: peripheral nerve biopsies by gigaxonin replacement using either a lentiviral vector or stable show accumulations of NFs within enlarged (3), while the transgenesis, demonstrating that the phenotype is caused by intermediate filament is seen to accumulate in muscle fi- GAN . GAN iPSCs therefore faithfully recapitulate key bers, glial fibrillary acidic protein in astrocytes and vimentin in GAN phenotypes and will prove valuable tools for preclinical test- multiple cell types, including primary fibroblast biopsies from pa- ing of gene replacement therapy and biochemical and genetic tients (4,5). Patients with GAN present in their first decade with analysis of the disease process to identify additional therapeutic loss of motor and sensory function. Currently, there is no treat- targets. ment for GAN, and the life expectancy is typically <30 years. Whether NF accumulation in neurons contributes directly to the pronounced sensory and motor neuropathy that is a clinical hall- mark of GAN is not known, but pathological observations of axonal Results loss, enlarged axons with abnormally thin myelin sheaths, atro- GAN is a rare neurodegenerative disease with limited access to phy of the pyramidal tracts, swelling of some axons in the spinal patient samples. We were successful in obtaining skin biopsies roots and anterior horn cell loss in the cervical and lumbar regions from five patients who had either a confirmed mutation in GAN of the spinal cord would be consistent with a causal role (6–9). or a classical GAN phenotype in the absence of an identified mu- GAN is a rare autosomal recessive disease caused by muta- tation. To represent the clinical and genetic heterogeneity of the tions in a single gene, GAN, which encodes gigaxonin, a ubiqui- disease, samples were collected from GAN patients of different tously expressed cytoplasmic protein that has been postulated sex, age and mutation (Fig. 1A). We generated iPSCs from skin fi- to act as an E3 ligase substrate adaptor (10). To model GAN path- broblasts by transduction with retroviral vectors encoding OCT4, ology in the mouse, two loss-of-function approaches were SOX2, KLF4 and c-MYC as described (14). All iPSC lines were as- explored: deletion of either the promoter-exon1 region or exons sayed for formation of embryonic stem cell-like colonies with 3–5. The former did not develop any overt neurological defects well-delineated phase-bright margins (Fig. 1B), expression of nu- over 15 months (11). The latter strategy was adopted for three clear and surface markers characteristic of human stem cells Δ – GAN ex3 5 models. The first, published by Ding et al. (12), was re- (Fig. 1C and D), karyotype (Fig. 1E) and viral transgene expression ported to develop strong motor deficits as early as 6 months of (Fig. 1F and G). Several iPSC lines were generated from each pa- age, but the characterization was performed using a heteroge- tient (Table 1; the line number refers to the GAN patient, and Δ – neous genetic background (11). Subsequently, GAN ex3 5 models the letter refers to a particular iPSC clone). Lines from Patients 1 were made on pure 129/SvJ and C57BL/6 backgrounds, respectively and 4 were excluded because of inadequate morphology of col- (13). Despite the detectable accumulation of intermediate fila- onies, indicative of spontaneous differentiation or high persist- ments and the absence of gigaxonin at 48 weeks of age, the former ent transgene expression (Fig. 1F and G), but those from model shows only a mild motor impairment from 60 weeks on- Patients 2, 3 and 5 satisfied all criteria. One clone per Patient 2, ward and the latter a mild sensory impairment (13). The relatively 3 and 5 (iPSC lines 2E, 3X and 5L) was selected for subsequent de- mild sensorimotor phenotype seen in both cases raises the possi- tailed analysis. bility that gigaxonin engages in species-specific interactions and As a first step toward analyzing the effects of the GAN muta- functions that might more faithfully be modeled using human tions, we differentiated the three selected iPSC lines into motor cells. Mouse and human gigaxonin share 98% similarity, with neurons (iPSC-MNs)—as motor neuron axons are strongly af- amino acid variants in the BR-C, ttk and bab (BTB), BTB and C- fected in patients—using a 28-day embryoid body-based protocol terminal Kelch (BACK) and Kelch domains. It is unknown what (15) that generates iPSC-MNs with ∼20–30% efficiency (Fig. 2A). effect, if any, these amino acid variants have on the structure -associated protein 2 (MAP2)-positive cells generated and function of human and mouse gigaxonin. from GAN iPSCs exhibited typical neuronal morphology (Fig. 2E) Successful derivation of induced pluripotent stem cells and expressed motor neuron markers HB9 and ISL1 at similar le- (iPSCs) (14,15) from patient skin fibroblasts provides an opportun- vels to a control iPSC line 18c (15) (Fig. 2F and data not shown). In ity to model the pathogenesis and treatment of human heritable addition, GAN neurons cultured for 12 days formed dense net- diseases in cell culture (16–21). Given the inaccessibility of pri- works of processes and exhibited abundant spontaneous Ca2+ mary human neurons, iPSCs present an appealing alternative. transients, demonstrating that they are functionally active Not only are patient mutations and genetic background faithfully (Fig. 2G). Patients with GAN have low, sometimes undetectable recapitulated in iPSCs, but neuronal differentiation protocols levels of gigaxonin protein (4). Similarly marked reductions allow for the production of unlimited quantities of cells affected were seen in fibroblasts, iPSCs and iPSC-derived neuronal cul- in the disease, providing a resource for biochemical and mechan- tures for all patients (Fig. 2B, C and D, respectively), including Pa- istic studies, as well as modeling potential therapeutic strategies tient 5, whose mutations remain to be identified. Therefore, (15,22). However, two types of hurdles have been encountered. deficits in gigaxonin levels do not impact the differentiation cap- First, for some diseases, in vitro phenotypes of clear relevance acity of iPSCs to spinal motor neurons. to the human condition have proven hard to uncover. Second, We next asked whether GAN iPSC-derived neurons display even if differences are observed between iPSC derivatives from any of the cytoskeletal defects seen in patients, initially using affected cases and healthy controls, it is essential to correct dis- the mixed spinal neuronal cultures generated by our protocol. ease-causing mutations in iPSCs to demonstrate that the pheno- One study found that a peripheral nerve from a GAN patient type results from the disease gene rather than from other contained 2-fold more NF-L than a control sample, whereas le- differences in genetic background. vels of neurofilament medium chain (NF-M) and neurofilament 1422 | Human Molecular Genetics, 2015, Vol. 24, No. 5

Figure 1. Characterization of GAN patient iPSC lines. (A) Demographics of patient fibroblast samples and summary of identified gigaxonin mutations. Lines generated from the patients shaded in gray met iPSC quality control criteria (see the text for details) and were included in the study. Throughout the text, the line number refers to a unique GAN patient, and the letter following the number refers to a clonal line derived from that patient. (B) Representative normal morphologies of GAN3X iPSC colonies. Representative immunocytochemical analysis of pluripotency marker TRA-1-60 (C)andOCT3/4(D) expression in GAN iPSC lines (GAN3X is shown). (E)Normal representative iPSC line karyotype (GAN3X is shown). (F) Levels of viral reprogramming transgenes (v) in GAN iPSC lines measured by quantitative reverse transcriptase polymerase chain reaction. Viral levels were normalized to HEK293 cells expressing the viruses 3 days post-infection. (G)Endogenouslevelsof pluripotency markers (e) were normalized to an embryonic stem cell line, HUES42.

Table 1 GAN iPSC lines derived levels of NF-M were comparable with control (Fig. 3C, I; P = 0.26, 0.33 and 0.16). Levels of NF-H could not be measured due to Patient Clones derived lack of a reliable signal on western blot (WB). These findings are also in agreement with reports of an increased abundance of the 1 Unable to derive lines NF-L subunits in GAN mouse models at later time points, while 2C,D,E,I,J 3N,S,T,V,X NF-M and NF-H are less affected (11,13). Western blot analysis 4I,J,L further revealed an increase in the neuronal intermediate fi 5I,J,L,K lament protein peripherin (PRPH) in differentiated GAN iPSCs (Fig. 3AandG;P = 0.038, 0.016 and 0.034) as has been reported Δex1/Δex1 For each patient, fibroblasts were infected with reprogramming viruses as in the mouse Gan model (11). As in mouse models (11), described in the ‘Materials and Methods’ section. Several clones (indicated by the elevation of NF-L and PRPH did not reflect accumulation letter) were picked for each patient. Of the clones picked, a subset was able to of mRNA (P > 0.05, Supplementary material, Fig. S1). Human expand. The clone from each patient with the best morphology (indicated in GAN neurons therefore exhibit a selective accumulation of bold) was selected for characterization. Lines were not derived from Patient 1 neural intermediate filament that mimics the situation due to inadequate colony morphology. in vivo. Interestingly, vimentin levels in iPSC-neurons appeared not heavy chain (NF-H) were unchanged (23). In a striking parallel, to differ significantly between patients and control (Fig. 3Dand iPSC-neurons from all GAN patients expressed levels of NF-L J; P = 0.23, 0.45 and 0.49). Although vimentin forms aggregates that were 3- to 4-fold higher than controls (Fig. 3B, H; P =0.049, in fibroblasts derived from GAN patient skin biopsies, its overall 0.026 and 0.015 for GAN2E, 3X and 5L, respectively), whereas levels were not reported to be elevated (24–26). Similarly, GAN Human Molecular Genetics, 2015, Vol. 24, No. 5 | 1423

Figure 2. Generation of functional iPSC-derived MNs from GAN patients. (A) Schematic representation of MN differentiation protocol and characterization experiments. Neuralization was induced using SB and LDN. Cells were caudalized and ventralized, using retinoic acid and SAG/purmorphamine, respectively (see ‘Materials and Methods’ for details). After 28 days of differentiation, embryoid bodies (EBs) were dissociated and MNs were plated. Cells were lysed for WB on Day 0, and fixed for ICC 3 days after dissociation. Ca2+ imaging was performed 12 days after dissociation. (B–D) Reduced gigaxonin levels in patient samples detected by WB of fibroblast (B), iPSC cell line (C) and differentiated iPSC-MNs (D). GIG, gigaxonin; HISH3, histone H3 used as loading control. Control fibroblasts were from male (M) and female (F) donors, respectively. (E) After differentiation for 28 days, EBs were dissociated, plated and, 3 days later, fixed and stained for neuronal (MAP2) and MN (ISL1) antigens. Scale bar represents 50 μ.(F) Quantification of the percentage of all cells (as labeled by DAPI) that express MAP2 or ISL1 using MetaMorph image analysis software (>675 cells were scored in three technical replicates from automatically acquired images, n =4–5 independent experiments). Plot represents the mean ± SEM. ISL was used for automated quantitation due to a higher signal-to-noise ratio compared with HB9. All P values (one-way ANOVA) are non-significant (P = 0.41 for ISL1 and 0.69 for MAP2 comparisons). (G) Twelve days after dissociation, cells were loaded with the calcium indicator Fluo-4, and images were acquired at ∼2 Hz. Ca2+ transients were determined from the soma of individual neurons, using raw pixel intensities. Shown are three representative tracings from GAN3X and GAN2E neurons.

cells showed no accumulation of non-intermediate filament pro- immunoreactivity in GAN (Fig. 3P) but not in control (Fig. 3N) teins, including tau and α-tubulin (Fig. 3E and K and unpublished cultures. This is reminiscent of the perinuclear intermediate fila- data; P = 0.28, 0.27 and 0.15) as well as MAP1B-LC (Fig. 3F and L; P = ment aggregates seen in patient fibroblasts (28) and the neuronal 0.43, 0.11 and 0.24). Overall, a subset of neuronal intermediate intermediate filament accumulations seen in the giant axons of filaments implicated in the human axonal pathology is selective- patients (7). In contrast, no change in MAP2 intensity or localiza- ly upregulated in spinal neurons derived from GAN patients. tion was apparent (Fig. 3M and O). To better characterize these In the developing , PRPH is selectively en- perinuclear accumulations, we took high magnification single riched in spinal motor neurons and sensory neurons, the cell optical plane confocal images (Fig. 3Q–T). Both PRPH and NF-L co- types primarily affected in GAN patients (27). We therefore fo- localize in the perinuclear aggregates and while filamentous cused our analysis on PRPH as a potential source of cell-type- structures were detected by immunocytochemistry at the periph- specific GAN pathology. To determine whether the intermediate ery of the aggregates light microscopy did not provide sufficient filament phenotypes observed in these mixed cultures reflected resolution to determine whether the proteins within the cores changes in iPSC-MNs themselves, we studied the subcellular of the aggregates have filamentous or disorganized structure. localization of PRPH at the level of individual cells. In identified To quantify the observed differences in protein accumulation, ISL1+ iPSC-MNs, we observed perinuclear accumulation of PRPH we measured intensity of MAP2 and PRPH immunostaining on 1424 | Human Molecular Genetics, 2015, Vol. 24, No. 5

Figure 3. GAN MNs exhibit selectively elevated levels of IFs. (A–F) Accumulation of neuronal IF proteins, but not other cytoskeletal proteins, in patient-derived MNs. Levels of indicated intermediate filament proteins were assessed by WB at Day 40 (n = 3). (G–L) Quantification of WB data confirmed that increases were reproducible and selective. Error bars represent the standard error of three independent experiments. GAN patients were compared with control using Student’s t-test (P < 0.05 for NF-L and PRPH comparisons; P > 0.05 for all other comparisons). (M–P)Immunofluorescence analysis was carried out at 7 days post-dissociation in UFdU treated control and GAN2E cultures. Arrowheads indicate MNs (as identified by ISL1 staining, not shown) used in the quantification in (U)and(V). Scale bar, 50 µ.(Q–T)GAN2E Human Molecular Genetics, 2015, Vol. 24, No. 5 | 1425

a cell-by-cell basis. The distribution of MAP2 intensities did at variable levels of transgene expression (Fig. 5D and E, upper not differ between the two genotypes (Fig. 3U). In contrast, band in GIG WB). GAN3X clone 1 expressed slightly higher levels PRPH was increased in GAN iPSC-MNs: only <30% of control of gigaxonin as compared with 18c (1.18-fold, all values normal- neurons have PRPH intensities >2000 units whereas for GAN ized to 18c), and clones 11 and 20 expressed at lower levels (0.89 neurons this figure is >70% (Fig. 3V). Using the non-parametric and 0.63, respectively) but still above the expression level of the Kolmogorov–Smirnov test (two-sample K–S test), the P-value parent GAN3X line (0.60). Interestingly, expression of FLAG-GIG for MAP2 differences between GAN and control is 0.6, whereas seemed to stabilize endogenous gigaxonin (Fig. 5E, lower band for PRPH P < 0.001. Thus the GAN genotype is associated with a in GIG WB). PRPH levels were corrected in the rescue lines as com- selective increase of PRPH levels in iPSC-MNs. pared with the parental line (Fig. 5F and G; two to three independ- Phenotypes reported in iPSC-derived cell types can potential- ent experiments for each clone; P <0.05forallGAN3Xrescue ly reflect inter-individual genetic differences unrelated to dis- clones compared with the parental line). Therefore, accumulated ease. It was therefore critical to determine whether the PRPH in GAN iPSC-MNs is remarkably sensitive to gigaxonin le- phenotypes observed thus far reflected direct effects of the loss vels and its level can be effectively corrected by virally delivered of function mutation in GAN gene. We constructed lentiviral gigaxonin, raising a hope for therapeutic effects of gene therapy vectors expressing myc-GAN or green florescent protein (GFP) in GAN. (as a control) under the control of the strong phosphoglycerate kinase-1 (PGK) promoter and used them to transduce the GAN2E and control iPSC-MN cultures (Fig. 4A and B). To measure overall Discussion levels of PRPH and control cytoskeletal proteins in the culture, iPSC-based models of human familial diseases offer promise for cell extracts were analyzed by densitometry of WBs in three accelerated characterization of pathologic pathways underlying independent experiments (Fig. 4C and E). Despite the high levels the diseases and for the ultimate development and validation of gigaxonin generated by the strong PGK promoter, total num- of new therapeutic strategies. Here, we outline such an approach bers of PRPH-positive cells were indistinguishable in control for GAN, an incurable disease-causing progressive neurodegen- and experimental conditions (Fig. 4D) demonstrating that gigaxo- eration of sensory and motor neural circuits. We report success- nin overexpression is not neurotoxic. Importantly, the elevated ful derivation of iPSCs from the fibroblasts of three GAN patients PRPH levels in GAN cells were significantly reduced (Fig. 4E; with unique mutations in the GAN gene. We provide evidence 2.9 ± 0.5-fold compared with GFP-infected cells; n =3;mean± that GAN iPSCs exhibit normal ability to differentiate into spinal SEM, P = 0.01) whereas levels of tau and α-tubulin were not signifi- motor neurons. Importantly, GAN-specific pathological pheno- cantly affected (1.1 ± 0.2-fold for tau, P = 0.36). types can be observed in GAN iPSC-MNs at early time points in We next examined rescue of the PRPH phenotype on a cell-by- culture. Moreover, replacement of gigaxonin in cultured cells cell basis. Cultures were stained for the virus (either GFP or myc) using lentiviral delivery or stable transfection reversed accumu- as well as for PRPH and MAP2 (Fig. 4F–K). Uninfected cultures dis- lation of PRPH. This suggests that delivery of the gigaxonin pro- played a significant increase in PRPH levels in GAN MNs as com- tein to cells that already exhibit signs of pathology might be pared with controls (Fig. 4L). This was rescued by overexpression sufficient to stop or perhaps even reverse the course of disease of gigaxonin (Fig. 4M; P < 0.0001). Seven days post-infection, PRPH in GAN patients. staining was reduced to control levels in GAN iPSC-MNs infected Many studies using iPSC models from patients and controls with myc-GAN virus (Fig. 4M; P < 0.0001 for GAN2E + GFP as com- have left open the question as to whether the observed differ- pared with the other three conditions). The reduction was ob- ences are true disease-relevant phenotypes or rather reflect served selectively in those neurons that were infected by the inter-individual genetic variability not directly related to disease virus (Fig. 4J, arrowhead). Thus, restoration of gigaxonin at a (29). To directly address this question, we used genetic rescue of time point at which intermediate filament accumulation is al- the disease-causing mutation in iPSC lines to create isogenic ready apparent can revert the pathological accumulation of pairs of lines ideally suited for disease modeling. We found that PRPH without causing iPSC-MN loss. replacement of gigaxonin in cultured human iPSC-MNs com- Because the PGK promoter leads to levels of gigaxonin that are pletely reverses intermediate filament protein accumulation. >15-fold greater than those of the endogenous protein, we sought Therefore, we have defined a gigaxonin-dependent phenotype to confirm these results in a more physiologically relevant con- that likely represents a key early step in the disease process. text by using stable expression of gigaxonin at lower levels. Moreover, our iPSC model is the first cell-based model of this dis- GAN3X iPSCs were nucleofected with a construct for expression ease that recapitulates intermediate filament protein accumula- of EF1α::FLAG-gigaxonin as well as puromycin resistance tion (rather than aggregation alone, as seen in patient fibroblast (Fig. 5A and B). Approximately 1 week after nucleofection, cul- models). One possible explanation is that we studied motor neu- tures were exposed to puromycin, and ∼1 week later resistant rons, one of the principal cell types affected in the disease, rather clones were picked and expanded for test differentiations. Of than undifferentiated skin fibroblasts. the ∼24 clones picked, three derived from GAN3X expressed Therapy for GAN has been focused for some while on replace- FLAG-gigaxonin (Fig. 5C, D and E; individual clones are indicated ment of the missing gigaxonin. Encouragingly, our study is in by their clone number). agreement with in vivo data that this approach may reverse the Positive clones were identified by IF, WB and qPCR and were disease process and clear PRPH aggregates as has been recently shown to generate iPSC-MNs with unaltered efficacy (Fig. 5C shown in a mouse model of GAN (30). Although questions of and data not shown). Rescue clones expressed FLAG-gigaxonin safety and feasibility remain to be addressed, our study provides

neurons were stained for PRPH and NF-L 7 days post-dissociation and imaged using confocal microscopy. Asterisk indicates location of nucleus. Scale bar, 25 µ. (U and V) PRPH and MAP2 intensity levels in the cell body region of individual control and GAN2E MNs (>55 per condition) were measured using MetaMorph and plottedas cumulative frequency of two independent experiments using the Kolmogorov–Smirnov non-parametric test. The P values are 0.598 for the data shown in (S) and <0.0001 for the data shown in (T). 1426 | Human Molecular Genetics, 2015, Vol. 24, No. 5

Figure 4. Viral overexpression of gigaxonin rescues elevated PRPH levels in GAN iPSC-MN cultures. (A) Schematic representation of GFP and GAN viral constructs. (B) Experimental design. Control and GAN iPSC-MN cultures were infected with PGK::myc-GAN or PGK::CD2-GFP lentivirus 12 days after dissociation. Seven to nine days post-infection, cultures were stained for viral markers (myc or GFP, respectively), PRPH and MAP2. (C) Levels of gigaxonin, intermediate filament and non- intermediate filament proteins were assessed using WB (n = 3). (D) GAN overexpression has no effect on survival of PRPH+ cells at 7 days post-infection (mean ± SD, n = 3 independent infections); P value between control conditions is 0.069; between GAN2E conditions P = 0.193 (Kruskal–Wallis test). (E)Quantification of fold reduction of protein levels in the GAN cultures shown in (C). PRPH levels in GAN cells were significantly reduced (2.9 ± 0.5-fold compared with GFP-infected cells; n =3; mean ± SEM, P = 0.01). Levels of tau were not significantly different (1.1 ± 0.2-fold; P = 0.36). (F–K) Gigaxonin rescues PRPH accumulation. Cell bodies were imaged at 7 Human Molecular Genetics, 2015, Vol. 24, No. 5 | 1427

Figure 5. Stable expression of gigaxonin protein reduces elevated PRPH levels in GAN iPSC-MN cultures. (A) Schematic representation of construct used to create GAN iPSC lines stably expressing gigaxonin. PURO, puromycin-resistance cassette. (B) GAN iPSC lines were nucleofected, and positive cells that had taken up the construct were identified by puromycin resistance. Approximately 24 colonies were picked for each parental iPSC line. These lines were expanded and differentiated into MNs and screened for FLAG expression. (C) Rescue clones make neurons and MNs, as identified by co-staining with MAP2 and ISL1, respectively (upper panel), and express FLAG-gigaxonin (lower panel). Rescue clones are identified by a unique number. (D) FLAG expression was detected by WB. (E) Levels of gigaxonin expression in rescue clones as compared with control. (F) Levels of PRPH were measured in the rescue lines. Western blots from two to three independent experiments are quantified in (G). Plots represent the mean ± SEM for three independent experiments (clone 20) or mean ± range for two independent experiments (clones 1 and 11). P < 0.05 for all GAN3X rescue clones compared with the parental line. data relevant to the safety, timing and dosage of gigaxonin at sev- observed that gigaxonin overexpression is still effective when in- eral levels. First, because gigaxonin is believed to activate protea- itiated after the onset of intermediate filament protein accumu- somal degradation, one concern was that overexpression of the lation, suggesting that delivery to cells that already exhibit signs protein might be toxic. However, patient-derived iPSC-MNs toler- of pathology may stop or even reverse disease progression. Lastly, ated levels of gigaxonin ∼15-fold greater than normal. Second, we because gigaxonin is quite effective even at low levels, it may be

days post-infection, and a reduction in the intensity of PRPH staining was noted, as well as an absence of perinuclear accumulations of PRPH. Scale bar,50μ.(L) Levels of PRPH were measured in the cell body of uninfected control and GAN2E MNs, P < 0.0001. Y-axis represents arbitrary units (a.u) of fluorescence intensity. (M) PRPH staining intensity is reduced to control levels by GAN expression. Levels of PRPH were quantified in the cell body of virus and Islet double-positive neurons (>30 per condition in two independent experiments) and fluorescence intensity was measured within the cell body. Y-axis represents arbitrary units (a.u) of fluorescence intensity. P < 0.0001 for GAN2E + GFP as compared with the other three conditions. P = 0.69 for GAN2E + GAN versus Ctrl + GAN. 1428 | Human Molecular Genetics, 2015, Vol. 24, No. 5

possible to attain effective concentrations even in cell types that Quantitative reverse transcriptase polymerase chain are difficult to transduce. reaction In summary, GAN is one of only a few human diseases for Total RNA was isolated using Trizol LS (Invitrogen). One micro- which iPSCs faithfully recapitulate a specific biochemical path- gram was treated with DNase (Invitrogen) and was subsequently ology observed in patients. Using this model, we demonstrate re- used to synthesize cDNA with iScript (Bio-Rad). Quantitative versal of an existing pathological phenotype by gigaxonin RT-PCR was then performed using SYBR green (Bio-Rad) and overexpression, providing proof of principle for gene replacement the iCycler system (Bio-Rad). Quantitative levels for all therapy for this devastating disease. The iPSC-based model of were normalized to endogenous β2 microglobulin (B2M) (for GAN we have developed has the dual potential to significantly ac- iPSC characterization) or glyceraldehyde 3-phosphate dehydro- celerate preclinical studies while serving as a tool for the discov- genase (GAPDH) (for all other experiments). For pluripotency ery and validation of new therapeutic approaches. genes, levels were expressed relative to the levels in human ES line HUES42. For viral genes, levels were expressed relative to the levels in HEK293 cells infected with the reprogramming Materials and Methods viruses for 2–3 days. Standard curves were run to ensure equal ef- Isolation of fibroblasts from patients with GAN ficiency of all primers, and RNA from HEK293 cells transfected with the plasmids encoding the transgenes was used as a posi- fi fi Dermal broblasts from ve patients with GAN were obtained tive control for viral transgene detection. Primer sequences are through skin biopsy. All samples were taken prior to the study, available upon request. Data were analyzed using the compara- ÀΔΔ and written informed consent of the parents of patients was ob- CT ΔΔ Δ – tive quantitation by the 2 equation: CT = CT (sample) tained for the identifiable samples according to the protocol ap- Δ Δ CT (control), where CT is the threshold cycle (CT) value of the proved by the Internal Research Board of Columbia University. housekeeping gene (B2M or GAPDH) subtracted from the CT Three other samples were sent to Columbia University as de- value of the target gene. identified tissue samples. The only demographic data collected was age, date of birth and the mutation, if known. Immunohistochemistry Pluripotency marker stains of iPSCs were performed after fixation Cell culture overnight in 4% paraformaldehyde (PFA) at 4°C as previously de- scribed (31). Neuronal cultures were fixed in 4% PFA for 30 min at Human GAN and control fibroblasts were maintained in Dulbec- 4°C, permeabilized and quenched with 0.1–0.2% Triton-X in co’smodified Eagle’s medium (DMEM) supplemented with 10% phosphate-buffered saline (PBS) (Wash buffer) and 100 m fetal bovine serum and penicillin/streptomycin at 37°C with 5% glycine (Sigma) for 20 min. Cells were blocked in Wash with CO .Allfibroblast samples were tested for mycoplasma, and 2 10% donkey serum for 30 min and then incubated in primary positive GAN fibroblasts (GAN1 and 3) were treated with Plasmo- antibody overnight, followed by incubation in secondary anti- cin (1 : 5000, InvivoGen) for 2 weeks, at which point they were re- bodies for 1 h. Primary antibodies used in this study are TRA1- tested after 1 week in antibiotic-free media and determined to be 60 (1 : 500, Chemicon, MAB4360), NANOG (1 : 500, R&D, AF 1997), cleared of the infection. GAN and control iPSCs were maintained OCT3/4 (1 : 500, Santa Cruz, sc-9081), ISL1 (1 : 200, DSHB, 40.2D6), on irradiated mouse embryonic fibroblasts (MEFs) in Human HB9 (1 : 100, DSHB, MNR2 81.5C10-c), MAP2 (1 : 2000, Neuromics, embryonic stem cell medium (HuESM) [DMEM/F12 supplemented CH22103), NF-H (1 : 2000, Neuromics, CH22104), PRPH (1 : 500, pro- with 20% knockout serum replacement, non-essential amino vided by Dr Ron Liem), anti-MYC (1 : 50, DSHB, 9E10-s) and anti- acids (NEAA), -glutamine, β-mercaptoethanol (BME) and 20 ng/ml GFP (1 : 5000, Invitrogen, A11122). Secondary antibodies used of basic fibroblast growth factor (bFGF)]. Media were changed every were DyLight 488, 549 and 647 conjugated (1 : 500, Jackson 24 h, and lines were passaged using dispase (Gibco, 1 mg/ml in ImmunoResearch). HuESM for 30 min at 37°C).

Microscope image acquisition GAN iPSC derivation Images were acquired on a fully automated Zeiss Observer Z1 fl – Twenty thousand fibroblasts were infected in 1 ml of HuESM in 1- epi- uorescence microscope at standard stage positions (12 25 well of a 6-well plate with a combination of OCT4, SOX2, cMYC and positions per well of a 96-well plate). Objectives used included: KLF4-containing murine leukemia retroviruses as described previ- Plan Apo 20X 0.8 DICII and Plan Apo 10X 0.45 DICII. Confocal ously (31). Cells were incubated with virus in 1 ml of HuESM for images were acquired using a Zeiss LSM 510 META with a Plan 24 h before the medium was supplemented with 1 ml of fibroblast Apo 100X 1.4 Oil objective. Temperature for image acquisition – medium to allow cells to recover. On Day 3, the medium was chan- was 22 24°C. All images were acquired in Wash buffer or ged to HuESM with the small molecule compounds (32) PD0325901 Fluoromount-G (confocal). Camera used was the CoolSNAP 2 (0.5 μ, Stemgent), SB431542 (2 μ, Stemgent) and Thiazovivin HQ . Images were acquired using AxioVision 4.8. Where indi- (0.5 μ, Stemgent). Approximately 1 week later, fluorescence-acti- cated, the Trophos Plate runner was used to acquire whole- vated cell sorting was used to isolate enriched populations of well images from 96-well plates. Single optical plane confocal CD13NEGSSEA4POSTra-1-60POS cells. These were sorted as single images were acquired using a Carl Zeiss LSM 510 confocal cells onto MEFs in 6-well plates, and iPSC colonies were manually microscope. picked based on morphology within 2–4 weeks. Quantitative image analysis Quantitative image analysis of differentiated neuronal cultures Karyotyping was performed using the Multi-Wavelength Cell Scoring module Karyotype analysis was performed by Cell Line Genetics, LLC. in MetaMorph (Molecular Devices) software. In brief, intensity Human Molecular Genetics, 2015, Vol. 24, No. 5 | 1429

thresholds were set (blinded to sample identity) to selectively Where indicated, cell division was inhibited with 5-fluoro-2- identify positive cells that displayed an unambiguous signal in- deoxyuridine (UFdU, 1 μ, Sigma). tensity above local background. These parameters were used on all samples, and only minimally adjusted for different staining batches as necessary. Script and parameter files were available Statistical analysis fi upon request. Where stated, cells were visually identi ed and All quantitative data were analyzed using Microsoft Excel or average cell body intensity measured by drawing a region of GraphPad. The Kolmogorov–Smirnov test was used to compare interest that encompassed the cell body. Brightness and contrast the probability distribution of the PRPH intensities of GAN and of each grayscale channel were adjusted to improve clarity. control MNs using an XLSTAT macro in Excel (Fig. 3S and T). Non-Gaussian sample groups were compared with a Kruskal– ’ Protein isolation and western blot analysis Wallis test using GraphPad Software (Fig. 4H). Student s t-test was used to compare individual GAN patients with control Cells were isolated, suspended in RIPA lysis buffer supplemented (Figs 3G–L, 4L, M and 5F). with protease and phosphatase inhibitor cocktail (Roche), tritu- rated and centrifuged at 10 000g for 15 min at 4°C. Then, 1–10 μg of total protein was separated on 10% sodium dodecyl sulfate– Calcium imaging polyacrylamide gels, transferred to a nitrocellulose membrane For epifluoresence microscopy, cells were loaded with 3 μ Fluo- and probed with various primary antibodies, followed by 4. Images were acquired at ∼2 Hz. Image analysis was performed horseradish-peroxidase-conjugated secondary antibody (1 : using ImageJ software and custom macros. Ca2+ transients were 5000–10 000, for mouse or rabbit primaries, respectively; Invitro- determined from regions of interest encompassing the soma of gen), and visualized using ECL chemiluminescence (33). Primary individual cells, using raw pixel intensities. antibodies used in this study are GIG (1 : 150, a generous gift from Dr Pascale Bomont), HISH3 (1 : 500, Millipore, MAB052), PRPH (1 : 5000, kindly provided by Dr Ron Liem), NF-L (1 : 200, Viral rescue experiments Neuromics, MO22104), NF-M (1 : 200, Sigma, G9670), VIM (1 : 500, Sigma, V2258), TAU (1 : 150, DSHB, 5a6), MAP1B-LC (1 : 4000, Myc-tagged human gigaxonin was cloned from a plasmid cour- Santa Cruz, H-130) and αTUB (1 : 10 000, Abcam, AB4074). For tesy of Dr Steven Gray (UNC). NSC34 cells were infected with quantification, band size and intensities were measured using 10-fold dilutions of viral supernatant. Four days later, cells were fi fi ImageJ software. Values were normalized first to their respective xed and stained for myc or GFP, and quanti ed with Meta- loading control and then to the control line to generate a fold Morph. The transducing units per microliter were determined change. PRPH signal was undetectable in one WB; for a fold from the slope of the linear regression. For subsequent experi- – comparison, the HISH3 loading control was used to estimate ex- ments with MNs, virus was added at 0.22 0.27 TU/cell. For WB fi pression because intermediate filament levels did not vary sig- and intermediate lament data, MNs were infected 12 days fi – nificantly for the control line between experiments. For viral after plating and were collected and xed 7 9 days later. experiments, the membranes were stripped and re-probed for NF-L. Membranes were washed in PBS Tween, incubated for 15 min at room temperature with stripping buffer, followed by Nucleofection of iPSC lines 7 min at 37°C. They were then washed for 1 h in PBS Tween, For rescue lines, the plasmid used was EF1α::FLAG-gigaxonin, blocked and probed with appropriate secondaries to confirm which contains a puromycin-resistance cassette. iPSC lines that the original signal was removed before reprobing. were dissociated into single cells using trypsin and incubated with Y-27632 as single cells for 1 h before nucleofection. Cells were nucleofected using the Amaxa Nucleofector® II (5 million Motor neuron differentiation of GAN and control iPSCs cells and 5 μg linearized plasmid). Cells were then plated on Pluripotent stem cell colonies were treated with dispase (1 mg/ puromycin-resistant MEFs and allowed to form small colonies ml) to separate colonies from feeder cells, washed and then tritu- before beginning antibiotic selection. After 10 or more days of se- rated to five to 10 cell clumps and seeded in low-adherence lection, large colonies were picked and expanded for further dishes in HuESM medium with 20 ng/ml of bFGF, SB431542 characterization. (10 μ, SIGMA), LDN193189 (0.2 μ, Stemgent) and Y-27632 (20 μ Ascent) for the first 2 days. At Day 3, embryoid bodies (EBs) were switched to a neural induction medium (DMEM/F12 Authors’ Contributions with -glutamine, NEAA, penicillin/streptomycin, heparin (2 μg/ B.J.K, C.H. and H.W. conceived and planned the study. B.J.K de- ml), N2 supplement (Invitrogen), brain-derived neurotrophic fac- signed and performed the experiments. F.A. and B.J.K derived tor (BDNF) (10 ng/ml) and ascorbic acid (0.4 μg/ml). At Day 7, the iPSC lines. R.S. and S.G. proposed viral delivery of GAN and retinoic acid (1 µ, Sigma), SAG (1 μ, Calibiochem smoothened provided the GAN plasmid. A.G.D. generated constructs and agonist 1.3), and purmorphamine (1 μ, Stemgent) were added. viruses. J.P.G. assisted with WBs and quantification. W.C., R.V.C At Day 17, the base medium was changed to Neurobasal (Invitro- and P.M. provided fibroblast samples. S.N., C.H. and H.W coordi- gen), with all previous factors and with the addition of 10 ng/ml nated the study and oversaw the results. B.J.K., C.H. and H.W. each of BDNF, glial derived neurotrophic factor and ciliary neuro- wrote the manuscript. All authors discussed the results and com- trophic factor (R&D), plus B-27 supplement (1 : 50, Invitrogen). At mented on the manuscript. Day 28, EBs were then dissociated with 0.05% trypsin (Invitrogen) and plated onto poly-lysine laminin-coated 96-well plates or 6- well plates at 10 000 or 1 million cells/well, respectively. Plated Supplementary Material neuron cultures were cultured in the same medium with the add- ition of 25 μ BME (Millipore) and 25 μ glutamic acid (Sigma). Supplementary Material is available at HMG online. 1430 | Human Molecular Genetics, 2015, Vol. 24, No. 5

Acknowledgements deletion of gigaxonin exon 1. J. Neurochem., 107, 253–264. Epub 2008 Jul 2031. We would like to thank Drs Pascale Bomont (Institut des Neuros- 12. Ding, J., Allen, E., Wang, W., Valle, A., Wu, C., Nardine, T., Cui, ciences de Montpellier), Jean-Pierre Julien (Université Laval), Rob- B., Yi, J., Taylor, A., Jeon, N.L. et al. (2006) Gene targeting of ert Goldman (Northwestern University), Mark Hannink and other GAN in mouse causes a toxic accumulation of microtubule- GAN collaborators for sharing unpublished data, observations associated protein 8 and impaired retrograde axonal trans- and reagents. We are grateful to Drs Pascale Bomont and Ron port. Hum. Mol. Genet., 15, 1451–1463. Epub 2006 Mar 1424. Liem (Columbia University) for generously providing antibodies. 13.Ganay,T.,Boizot,A.,Burrer,R.,Chauvin,J.P.andBomont,P. We also thank Anita Ritz and David Kahler of NYSCF for technical (2011) Sensory-motor deficits and neurofilament dis- assistance and members of the Project A.L.S., Henderson and organization in Gigaxonin-null mice. Mol. Neurodegener., Wichterle labs for helpful suggestions. We thank Lori Sames 6,25. and Hannah’s Hope Fund, as well as Valerie Estess and Project 14. Takahashi, K., Tanabe, K., Ohnuki, M., Narita, M., Ichisaka, T., A.L.S., for their encouragement and support, and the patients Tomoda, K. and Yamanaka, S. (2007) Induction of pluripotent for their generous donation of skin biopsy samples. B.J.K. is grate- stem cells from adult human fibroblasts by defined factors. ful to the Columbia University MD/PhD program for their support. Cell, 131, 861–872. Conflict of Interest statement. None declared. 15. Boulting, G.L., Kiskinis, E., Croft, G.F., Amoroso, M.W., Oak- ley, D.H., Wainger, B.J., Williams, D.J., Kahler, D.J., Yamaki, M., Davidow, L. et al. (2011) A functionally characterized test set of human induced pluripotent stem cells. Nat. Bio- Funding technol., 29,279–286. This work was supported by Hannah’s Hope Fund, Project A.L.S., 16.Dimos,J.T.,Rodolfa,K.T.,Niakan, K.K., Weisenthal, L.M., and NYSTEM (CO24415). B.J.K. is supported by an F31 grant from Mitsumoto, H., Chung, W., Croft, G.F., Saphier, G., Leibel, R., NINDS (NS077544-01A1) and a Hearst Scholarship. J.P.G. was a Goland, R. et al. (2008) Induced pluripotent stem cells gener- NYSTEM summer research fellow. ated from patients with ALS can be differentiated into motor neurons. Sci., 321, 1218–1221. 17. Marchetto, M.C., Carromeu, C., Acab, A., Yu, D., Yeo, G.W., Mu, Y., Chen, G., Gage, F.H. and Muotri, A.R. (2010) A model for References neural development and treatment of Rett syndrome using 1. Lariviere, R.C. and Julien, J.P. (2004) Functions of intermediate human induced pluripotent stem cells. Cell, 143, 527–539. filaments in neuronal development and disease. J. Neurobiol., 18. Brennand, K.J., Simone, A., Jou, J., Gelboin-Burkhart, C., Tran, 58, 131–148. N., Sangar, S., Li, Y., Mu, Y., Chen, G., Yu, D. et al. (2011) Mod- 2. Liem, R.K. and Messing, A. (2009) Dysfunctions of neuronal elling schizophrenia using human induced pluripotent stem and glial intermediate filaments in disease. J. Clin. Invest., cells. Nature, 473, 221–225. Epub 2011 Apr 2013. 119, 1814–1824. doi: 1810.1172/JCI38003. Epub 32009 Jul 38001. 19. Howden, S.E., Gore, A., Li, Z., Fung, H.L., Nisler, B.S., Nie, J., 3. Gordon, N. (2004) Giant axonal neuropathy. Dev. Med. Child Chen,G.,McIntosh,B.E.,Gulbranson, D.R., Diol, N.R. et al. Neurol., 46, 717–719. (2011) Genetic correction and analysis of induced pluripotent 4. Cleveland, D.W., Yamanaka, K. and Bomont, P. (2009) Gigaxo- stem cells from a patient with gyrate atrophy. Proc. Natl. Acad. nin controls vimentin organization through a tubulin chaper- Sci. USA, 108, 6537–6542. one-independent pathway. Hum. Mol. Genet., 18, 1384–1394. 20. Lee, G., Papapetrou, E.P., Kim, H., Chambers, S.M., Tomishi- Epub 2009 Jan 1324. ma, M.J., Fasano, C.A., Ganat, Y.M., Menon, J., Shimizu, F., 5. Peiffer, J., Schlote, W., Bischoff, A., Boltshauser, E. and Muller, Viale, A. et al. (2009) Modelling pathogenesis and treatment G. (1977) Generalized giant axonal neuropathy: a filament- of familial dysautonomia using patient-specific iPSCs. Nature, forming disease of neuronal, endothelial, glial, and schwann 461, 402–406. Epub 2009 Aug 2019. cells in a patient without kinky hair. Acta Neuropathol., 40, 21. Wichterle, H. and Przedborski, S. (2010) What can pluripotent 213–218. stem cells teach us about neurodegenerative diseases? Nat. 6. Berg, B.O., Rosenberg, S.H. and Asbury, A.K. (1972) Giant - Neurosci., 13, 800–804. al neuropathy. Pediatrics, 49, 894–899. 22. Wichterle, H., Lieberam, I., Porter, J.A. and Jessell, T.M. (2002) 7. Asbury, A.K., Gale, M.K., Cox, S.C., Baringer, J.R. and Berg, B.O. Directed differentiation of embryonic stem cells into motor (1972) Giant axonal neuropathy—a unique case with segmen- neurons. Cell, 110, 385–397. tal neurofilamentous masses. Acta Neuropathol., 20, 237–247. 23. Ionasescu, V., Searby, C., Rubenstein, P., Sandra, A., Cancilla, 8. Thomas, C., Love, S., Powell, H.C., Schultz, P. and Lampert, P. P. and Robillard, J. (1983) Giant axonal neuropathy: normal W. (1987) Giant axonal neuropathy: correlation of clinical protein composition of neurofilaments. J. Neurol. Neurosurg. findings with postmortem neuropathology. Ann. Neurol., 22, Psychiatry, 46, 551–554. 79–84. 24. Pena, S.D. (1982) Giant axonal neuropathy: an inborn error of 9. Kumar, K., Barre, P., Nigro, M. and Jones, M.Z. (1990) organization of intermediate filaments. Muscle Nerve, 5, Giant axonal neuropathy: clinical, electrophysiologic, and 166–172. neuropathologic features in two siblings. J. Child Neurol., 5, 25. Mahammad, S., Murthy, S.N., Didonna, A., Grin, B., Israeli, E., 229–234. Perrot, R., Bomont, P., Julien, J.P., Kuczmarski, E., Opal, P. et al. 10. Furukawa, M., He, Y.J., Borchers, C. and Xiong, Y. (2003) Tar- (2013) Giant axonal neuropathy-associated gigaxonin muta- geting of protein ubiquitination by BTB-Cullin 3-Roc1 ubiqui- tions impair intermediate filament protein degradation. tin ligases. Nat. Cell Biol., 5, 1001–1007. Epub 2003 Oct 1005. J. Clin. Invest., 123, 1964–1975. 11. Dequen, F., Bomont, P., Gowing, G., Cleveland, D.W. and Ju- 26. Mussche, S., Devreese, B., Nagabhushan Kalburgi, S., Bacha- lien, J.P. (2008) Modest loss of peripheral axons, muscle atro- boina, L., Fox, J.C., Shih, H.J., Van Coster, R., Samulski, R.J. phy and formation of brain inclusions in mice with targeted and Gray, S.J. (2013) Restoration of homeostasis Human Molecular Genetics, 2015, Vol. 24, No. 5 | 1431

after gigaxonin gene transfer for giant axonal neuropathy. new insights into disease mechanisms. Muscle Nerve, 46, Hum. Gene Ther., 24, 209–219. 246–256. 27. Lariviere, R.C., Nguyen, M.D., Ribeiro-da-Silva, A. and Julien, J. 31. Dimos, J.T., Rodolfa, K.T., Niakan, K.K., Weisenthal, L.M., P. (2002) Reduced number of unmyelinated sensory axons in Mitsumoto,H.,Chung,W.,Croft,G.F.,Saphier,G.,Leibel, peripherin null mice. J. Neurochem., 81, 525–532. R., Goland, R. et al. (2008) Induced pluripotent stem cells 28. Bomont, P. and Koenig, M. (2003) Intermediate filament ag- generated from patients with ALS can be differentiated gregation in fibroblasts of giant axonal neuropathy patients into motor neurons. Science, 321, 1218–1221. Epub 2008 Jul is aggravated in non dividing cells and by microtubule desta- 1231. bilization. Hum. Mol. Genet., 12, 813–822. 32. Lin, T., Ambasudhan, R., Yuan, X., Li, W., Hilcove, S., Abujar- 29. Park, I.-H., Arora, N., Huo, H., Maherali, N., Ahfeldt, T., Shima- our, R., Lin, X., Hahm, H.S., Hao, E., Hayek, A. et al. (2009) A mura, A., Lensch, M.W., Cowan, C., Hochedlinger, K. and chemical platform for improved induction of human iPSCs. Daley, G.Q. (2008) Disease-specific induced pluripotent stem Nat. Methods, 6, 805–808. Epub 2009 Oct 2018. cells. Cell., 134, 877–886. 33. Domanico, S.Z., DeNagel, D.C., Dahlseid, J.N., Green, J.M. and 30. Mussche, S., De Paepe, B., Smet, J., Devreese, K., Lissens, W., Pierce, S.K. (1993) Cloning of the gene encoding peptide-bind- Rasic, V.M., Murnane, M., Devreese, B. and Van Coster, R. ing protein 74 shows that it is a new member of the heat (2012) Proteomic analysis in giant axonal neuropathy: shock protein 70 family. Mol. Cell Biol., 13, 3598–3610.