Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Nat Med. Author manuscript; available in PMC 2017 August 14.
Published in final edited form as:
Published online 2016 November 14. doi:  10.1038/nm.4220
PMCID: PMC5555047

Capturing the biology of mild versus severe disease in a pluripotent stem cell-based model of Familial Dysautonomia


Familial Dysautonomia (FD) is a debilitating disorder that affects derivatives of the neural crest (NC). For unknown reasons, FD patients show marked differences in disease severity despite carrying the identical, homozygous IKBKAP point mutation. Here, we present disease related phenotypes in pluripotent stem cells (PSCs) that capture FD severity. Cells from severe but not mild patients exhibit impaired specification of NC derivatives including autonomic and sensory neurons. In contrast, both severe and mild FD cells show defects in peripheral neuron survival, indicating neurodegeneration as culprit for mild FD. While genetic repair of the FD mutation reversed early developmental NC defects, sensory neuron specification was not restored indicating that other factors may contribute to disease severity. Whole exome sequencing identified candidate modifier genes for severe FD. Our study demonstrates that PSC-based modeling is sensitive in recapitulating disease severity, which presents an important step on the road towards personalized medicine.


In vitro disease modeling using embryonic or induced pluripotent stem cells (PSCs) has shown considerable potential for applications in drug discovery as well as for the elucidation of disease mechanisms1,2. Human PSC-based models can be particularly powerful where suitable animal models are not available or where access to affected primary tissues is not feasible. Furthermore, the technology promises a personalized approach to the biology and treatment of each individual patient by using cells that carry an exact copy of the patient’s own DNA. While there has been considerable success in capturing disease phenotypes in vitro using the human PSC technology3, the ability to recapitulate disease severity in a defined patient population has not been demonstrated yet, in particular for individuals carrying an identical genetic defect such as in the case of Familial Dysautonomia (FD).

FD was one of the first genetic disorders to be faithfully modeled in PSC-derived lineages in vitro4. FD, also known as Riley-Day Syndrome or hereditary sensory and autonomic neuropathy III (HSAN-III), is a genetic disease affecting the development, function and survival of various neural crest (NC)-derived lineages such as autonomic (AN) and peripheral sensory neurons (SN)57. FD is an autosomal recessive disease and occurs almost exclusively in the Ashkenazi-Jewish population with a carrier frequency of 1 in 328. From birth on, FD patients display a broad range of symptoms that include difficulties in swallowing/suckling, absence of fungiform papillae on the tongue, kyphoscoliosis, lack of tears, difficulties in regulating body temperature and blood pressure, reduced sensitivity to pain9 and recurring vomiting crisis (dysautonomic crisis). Autopsy data show evidence of severely reduced cell numbers in the dorsal root ganglion (DRG)9 and superior cervical ganglion (SCG)10. The clinical progression of FD patients indicates a degenerative component to the disease11 in addition to any developmental defect. Nearly all (99.5%) FD patients carry the identical homozygous T>C point mutation in the splice donor (in intron 20) of the I-κ-B kinase complex-associated protein (IKBKAP) gene, leading to partial skipping of exon 20 in the mutant IKAP protein and a reduction of wild type protein. IKAP (ELP1) is a component of the transcriptional elongator complex and is ubiquitously expressed. It is not known why its impairment leads to a NC tissue-specific defect. The plant hormone kinetin12, a food supplement, was shown to increase splicing of wild type IKBKAP transcript in primary and PSC-derived NC cells4 and is currently being tested in human FD patients13. FD was also the target of the first large-scale chemical screen in a PSC-based disease model that led to the identification of additional FD drug candidates including the α2-adrenergic receptor antagonist SKF-8646614. SKF-86466 enhances IKBKAP mRNA levels and IKAP protein expression in PSC-derived NC cells independent of splicing14.

While nearly all FD patients carry the same homozygous IKBKAP mutation, the severity of disease symptoms can vary. Here, we use FD as a platform to address whether PSC-technology is sufficiently sensitive to predict disease severity among individual patients carrying the identical mutation in IKBKAP.


Modeling mild and severe FD in neural crest progenitor cells in vitro

To model disease severity, we obtained fibroblasts from three patients each with severe (S) or mild (M) FD symptoms, from one heterozygous carrier with no obvious symptoms (A) and from one healthy control (C) subject. Supplementary Fig. 1a and Supplementary table 1 summarize available information used to classify severe versus mild patients, including decreased pain perception and frequency of dysautonomic crisis. We did not match sex and age as there are no known gender-specific defects and age-related cellular features are thought to be largely reset after reprogramming15.

We first characterized genotype (Supplementary Fig. 1b) and growth characteristics (Supplementary Fig. 1c) in the fibroblasts, but did not find any differences that discriminate between disease severity at the primary fibroblast stage. We next reprogrammed fibroblasts from two patients per severity group (S2, S3, M1, M2, A1, C1) into PSCs using non-integrating, cytoplasmic RNA Sendai-virus vectors expressing OCT4, SOX2, KLF4 and cMYC16 (Supplementary Fig. 2a). At least two PSC clones per patient were reprogrammed and characterized. However, due to low inter-patient variability (data not shown), the majority of subsequent phenotypes were carried out with one clone per patient. PSC lines were characterized for pluripotency (Supplementary Fig. 2b, c) and splicing of IKBKAP (Supplementary Fig. 2d), as well as in vitro differentiation potential towards cell types representing the three germ layers that are unaffected in FD (Supplementary Fig. 3a, b).

To address whether the FD PSC-model can capture differences between mild and severe FD, we first revisited three previously described phenotypes (IKBKAP missplicing, deficit in ASCL1 expression levels and migration defects)4 at the NC stage. We initially followed the protocol used in our past FD studies based on delamination and fluorescence activated cell sorting (FACS)-based enrichment of NC precursors from early neural rosette structures17,18 (termed here rosette-NC (rNC)). Supplementary Fig. 4a–c shows that all PSC lines generate neural rosettes and rNC cells at comparable efficiencies, with no differences in division rates as previously reported4,14.

To address whether cells at the NC progenitor stage can capture aspects of disease severity we first measured IKBKAP splicing using qRT-PCR. We found that both severe and mild rNC progenitors express a low ratio of wild type to mutant IKBKAP transcripts in contrast to control cells (Fig. 1a). Thus, neither the nature of the IKBKAP mutation nor the extent of alternative splicing can explain mild versus severe forms of the disease. We next investigated the functionality of rNC cells. We observed that rNC cells derived from severe FD patients did not migrate efficiently in a scratch assay4. In contrast, cells from mild patients migrated as efficiently as control cells (Fig. 1b). To control for potentially differential proliferation, cells that migrated into the scratch were normalized to the number of total cells in the picture. Furthermore, while rNC cells from severe FD patients expressed low levels of ASCL1, a marker for autonomic neuron progenitors, rNC cells from mild FD showed ASCL1 levels comparable to those in unaffected controls both at the mRNA (Fig. 1c) and the protein (Fig. 1d) levels.

Figure 1
Phenotypes in rNC cells derived from FD-PSCs recapitulate differences between mild and severe FD patients. a. Quantification of wild type to mutant IKBKAP transcript expression ratios in severe (S2 and S3) and mild (M1 and M2) FD PSC-derived rNC cells ...

To further assess global changes in gene expression between mild versus severe FD we performed two RNA sequencing studies. First, we compared PSCs and FACS-purified rNC from one biological replicate of each severe, mild and control. Unsupervised global clustering as well as GO term analysis showed that at the rNC stage mild FD samples were more similar to control and distinct from severe FD samples in contrast to cells at the PSC stage where mild and severe FD were distinct from control cells (Supplementary Fig. 5a,b). These results were confirmed in the second RNA sequencing study conducted with 2 or 3 biological replicates (Supplementary Fig. 6a). GO term analysis revealed changes in transcripts associated with extracellular region and cell adhesion terms as enriched in severe versus mild FD rNC and neurogenesis terms enriched in mild samples (Supplementary Fig. 6b, c). However, none of the top genes differentially regulated between mild and severe rNC pointed towards an obvious molecular explanation of disease severity (Supplementary Fig. 6d).

Modeling mild and severe FD in neural crest progeny in vitro

Next, we asked if the differentiation disparity between mild and severe FD persist in NC derivatives, i.e. in cells relevant for patients post embryonic development. FD autopsy data revealed that sensory neurons (SNs) in the DRG and autonomic neurons (ANs) in sympathetic ganglia are reduced to 20% and 34% of normal subjects, respectively9,10 and thus are key targets for therapeutic interventions. In fact, severe FD patients studied here have a pronounced decrease in pain sensation and report frequent vomiting episodes, likely a result of a deregulation of the autonomic nervous system. These symptoms are less commonly observed in mild FD patients (Supplementary table 1). Thus, we investigated the differentiation capacity of severe and mild FD lines into those NC derivative lineages.

We derived FD SNs employing a differentiation protocol19 that yielded 33% BRN3A+ SNs (Supplementary Fig. 7a). By day 12 of differentiation it was obvious that lines from severe FD patients were deficient in making SNs. In contrast, both mild FD and control cultures yielded a dense network of neuronal processes (Supplementary Fig. 7b). Immuno-staining and quantification of SNs, marked by BRN3A+/TUJ1+ at day 14, confirmed these results (Fig. 2a and b). A temporal expression analysis of SOX10 established the NC origin of SNs (Supplementary Fig. 7c). One potential reason for the inability of severe lines to generate SNs would be a lower splicing ratio of IKBKAP20. However, IKBKAP ratios were low in both severe and mild SNs (Fig. 2c). Thus, IKBKAP expression alone did not explain the SN phenotype or the differences between mild and severe FD.

Figure 2
FD phenotypes in peripheral sensory and autonomic neurons. a. SNs were differentiated from FD PSCs in vitro using an established protocol19. At day 12 of differentiation the cells were replated at equal densities and immuno-stained for BRN3A, TUJ1 and ...

To address FD phenotypes in autonomic-like neurons (ANs) we first established a differentiation protocol (Supplementary Fig. 7d), as no reliable published reports were available. NC cells are induced by activating WNT signaling at day 2 of differentiation using a modified dual-SMAD inhibition protocol19,21. At day 11 the cells were FACS isolated based on the expression of CD49D (α4-integrin receptor), which marks SOX10+ NC cells22 (Supplementary Fig. 7e). Gene expression analysis by qRT-PCR at day 30, 40, 50 and 60 and immuno-staining showed strong up-regulation of sympathetic AN markers, such as ASCL1, PHOX2A, SCG10, TH and DBH,23 while markers of related cell types, such as parasympathetic ANs (CHAT23), enteric neurons (EDNRB24) and sensory placode lineages (Six125) were only moderately induced (Supplementary Fig. 8a,b). However, analysis of mature markers such as autonomic ganglia-specific acetylcholine receptor CHRNA3, CHRNB4 and vesicular monoamine transporter VMAT1, VMAT2 was largely negative (Supplementary Fig. 8d), and we did not detect noradrenaline release by HPLC (data not shown), suggesting an immature AN-like phenotype. A recent study showed that only upon co-culture with mouse cardiomyocytes, HES-derived sympathetic neurons express such late markers, further supporting our immaturity hypothesis26. When we applied the AN-like protocol to FD-PSC lines, we found that cells from severe FD patients could not effectively generate the CD49D+ NC, in contrast to control and mild FD cells (Fig. 2d). In fact, severe FD SOX10+ NC cells survive poorly while mild FD and control SOX10+ NC cells generated AN-like cells (PHOX2A+ and TUJ1+) by day 25 (Supplementary Fig. 9a) at efficiencies of 10–25% of total cells (TH+ and ASCL1+, Supplementary Fig. 9b). The non-AN-like cell fraction was identified as mostly non-proliferating (Ki67), myofibroblast-like cells (αSMA+) with a fraction of SN contaminants (BRN3A+, ISL1+, Supplementary Fig. 9c, d).

Neurodegeneration in mild and severe FD cells accounts for disease symptoms and offers a therapeutic target

We next aimed to define cellular phenotypes that can discriminate mild from controls and explain why mild FD patients are sick at all. FD is a progressive disease as illustrated by the deterioration in pain perception with increasing age6,11. Therefore, we tested whether mild FD patients exhibit primarily neurodegenerative rather than developmental phenotypes. We replated differentiated FD SNs on day 12 at equal densities (Fig. 3a, first panel) and measured survival as the number of BRN3A+/TUJ1+ SNs every 3 days. Control SNs remained healthy. In contrast SN numbers from two mild FD patients significantly diminished. Similarly, the few severe SNs progressively died (Fig. 3a and b). We further found that cleaved Caspase3 staining increased, paralleled by loss of BRN3A+ cells in FD but not control cells (Supplementary Fig. 10a). To investigate potentially contaminating cells in the cultures, we stained for progenitor (SOX10, SOX2) and proliferation (Ki67) markers, which were largely negative. Most non-SNs in the cultures were α-SMA+ non-proliferating, myofibroblast-like cells (Supplementary Fig. 10b). Finally, we assessed the survival of mild FD AN-like cells over time. Surprisingly, mild FD AN-like cells did not shown any defect in survival for up to 50 days based on TH+ and ASCL1+ cells. Even when the cells were stressed by growth factor removal or adding toxin (carbonyl cyanide 3-chlorophenylhydrazone CCCP), neither mild FD nor control AN-like cells died (Supplementary Fig. 11a and b). Future studies may determine whether the failure to detect degenerative phenotypes is due to the immaturity of AN-like cells (Supplementary Fig. 8d and 11a). Together, this data suggests degeneration as the primary culprit for mild FD while both developmental and degenerative defects characterize severe FD.

Figure 3
Mild and severe FD SNs have impaired survival capacity in vitro, which can be rescued pharmacologically. a. PSC-derived SNs were replated at equal density (DAPI staining (first panel) and maintained for at least 14 more days in optimal survival conditions. ...

We have previously identified positive effects of kinetin12 and SKF-86466 (α2-adrenergic receptors antagonist) on IKBKAP splicing in FD rNC cells14. However, rNC phenotypes represent a neurodevelopmental paradigm that may not translate to FD patients who no longer harbor significant numbers of NC-equivalent progenitors. Thus, we asked if neurodegeneration can be halted in mild FD SNs. After verification of expression of α2-adrenergic receptors (Supplementary Fig. 12a), SNs were differentiated and treated with kinetin, SKF-86466 or vehicle (DMSO), starting from day 12 and survival was assessed. Fig. 3c shows that SNs from mild FD, treated with either drug survived, while DMSO-treated cells progressively died. This rescue was correlated with an increase in IKBKAP splicing transcripts in kinetin but not SKF-86466 treated cells (Supplementary Fig. 12b). These results indicate potential treatment options for patients after birth and support efforts to bring kinetin into the clinic27.

Genetic repair of the FD mutation yields insight into the molecular mechanism underlying disease severity

Recent evidence in FD animal models point to the importance of ikbkap expression levels. Conditional knockout mouse models suggest that phenotype severity is a direct result of ikap levels28,29. To determine the role of IKBKAP expression on disease severity in human FD patient samples, we first looked at the effect on IKBKAP splicing. We observed that the ratio of wild type to mutant IKBKAP expression is equally low in severe and mild PSCs (Supplementary Fig. 2d), rNC cells (Fig. 1a), SNs (Fig. 2c), primary fibroblasts and PSC-derived endoderm, cardiomyocytes and cortical neurons (Supplementary Fig. 13a).

At the protein level in PSCs both severe and mild FD showed low IKAP levels compared to controls (Supplementary Fig. 13b, left). However, at the rNC stage, mild and control lines expressed comparable IKAP protein levels, while severe patients showed lower levels (Supplementary Fig. 13b right; note, wild type IKAP is expressed in two non-disease associated splice variants (150 kDa and 111 kDa)). These results match previous IKBKAP splicing (Supplementary Fig. 2d, Fig. 1a) and RNA sequencing data (Supplementary Fig. 5 and 6), where at the PSC stage mild and severe cells show comparable expression levels while at the rNC stage mild cells show levels comparable to control cells distinct from severe cells. Thus, it is conceivable that increased IKAP protein levels in rNC could contribute to protecting mild FD cells from the early developmental phenotypes characteristic for severe FD cells.

To confirm that the inability of severe FD cells to generate NC precursors and derivatives is indeed specific to the IKBKAP mutation, we used the CRISPR/Cas9 system30,31 to repair of the FD-mutation. We identified 3 clones that showed heterozygous repair (targeting frequency of 10% (Supplementary Fig. 14a–c)). No Cas9 off-target cutting was detected at the top 10 predicted sites (Supplementary Fig. 15a, b). To compare those gene-corrected clones to an FD heterozygous line, we established a PSC line from the mother of S2, a heterozygous FD carrier (A1). Analysis of IKBKAP splicing in gene-targeted clones revealed elevated splicing ratios compared to the parental severe (S2) line and were comparable to splicing in the heterozygous FD carrier (A1, Fig. 4a,b). Analysis of IKAP protein levels yielded similar results (Supplementary Fig. 14d, e). Heterozygous FD carriers do not show FD disease symptoms5, so we tested whether our in vitro FD phenotypes reverted in gene corrected cells. Importantly, both the ASCL1 phenotype in rNC cells (Fig. 4c) and the CD49D phenotype during early NC induction (Fig. 4d) were rescued in gene-corrected clones either to wild type or FD carrier (A1) levels. These results confirm that the IKBKAP point mutation is responsible for the early developmental NC disease phenotypes in severe FD. Next, we investigated if genetic rescue in IKBKAP reverts differentiation defects of NC derivatives, i.e. SNs. Interestingly however, SN differentiation was not restored in any of the three genetic rescue lines as compared to control or the A1 heterozygous carrier lines (Fig. 4e). This data suggests that FD phenotypes downstream of the NC are not solely caused by the IKBKAP mutation.

Figure 4
Genetic rescue of severe FD PSCs and in vitro FD phenotypes. a, b. Ratio of wild type to mutant IKBKAP splicing in PSCs (N=3–5 biological replicates, one-way ANOVA, ***p<0.0001, F=9.4, Dunnett’s multiple comparison test. S versus ...

Candidate modifier mutations present in severe but not mild FD may contribute to disease severity

To identify candidate modifier genes in severe FD patients, we performed whole exome sequencing in three severe (S1, S2, S3) and three mild (M1, M2, M3) FD patients. We searched for genes that display non-synonymous coding variants (not necessarily the same variants) in all 3 severe patients following a model of complete penetrance given the small sample size. We focused on rare variants, with a minor allele frequency (MAF) of less than 1% in public databases (<1% KG in and We further ranked the candidate genes based on the predicted impact of each variant by variant-level metric CADD (Combined Annotation Dependent Depletion: CADD32,33) and on the number of damaging mutations found in healthy individuals for each gene by gene-level metric GDI (Gene Damaging Index: GDI34). We also took into account the mutation significance cutoff (MSC), a gene-level specific cutoff for CADD scores35. First, testing an autosomal dominant mode of inheritance, we identified 7 candidate modifier genes harboring homozygous variations common in all severe FD patients. However, the same variations can also be found in the mild patients (Fig. 5a, Supplementary Fig. 16, Supplementary table 2, sheet 1). Second, we tested an autosomal recessive mode of inheritance and identified 23 candidate genes. 20 of these genes were also mutated in mild patients and thus filtered out. 3 candidate genes remained: LAMB4, KIAA1211, and FAT2. Within the severe patients, these 3 genes carried variations predicted to be damaging (CADD score > MSC score35). With the lowest GDI score LAMB4 was our top candidate (Fig. 5a, Supplementary Fig. 16 and Supplementary table 2, sheet 2). In a separate analysis, based on incomplete penetrance but taking into account close functional relatedness to IKBKAP, as revealed by the gene’s connectome36, few additional candidates were identified from the initial list of affected genes including URI1, LCA10, MSH3, YEATS2.

Figure 5
Mechanistic insight into the molecular difference between mild and severe FD. a. Representation of whole exome sequencing of three mild and three severe FD patients. Red triangles represent mutations that passed all filtering criteria, i.e. exonic mutations, ...

Laminin β4 (LAMB4) harbors one missense variant in patient S1 and another shared by S2 and S3 (Supplementary table 2, sheet 2, Supplementary Fig. 17). Laminins are secreted heterotrimeric (α, β, γ-chains) proteins of the extracellular matrix. They greatly influence the differentiation, migration and adhesion of cells within tissues and have important roles in neuronal development, such as axonal growth. To date no laminin trimer, containing LAMB4 has been identified and its function is largely unknown. This gene is expressed highest in the SCG and DRG (, thus it may play a role in the development of tissues particularly affected in FD patients and that show clear phenotypic differences between mild and severe FD in our model. We used our RNA sequencing data conducted in rNC cells to investigate whether the three mutations lead to differential gene expression comparing severe, mild and control rNC. Only FAT2 was significantly higher expressed in severe compared to mild cells, particularly in exon 9, where severe patient 2 has a mutation (Supplementary Fig. 17), indicating that the candidate mechanisms of action will involve in-depth molecular and functional studies. Experiments from this study are summarized in Supplementary table 3.


FD has been at the forefront of PSC-based disease modeling4 and drug discovery14 efforts. In the current study we use FD as a model to address a next major challenge for PSC research, the ability to capture disease severity, as an important step towards implementing it in personalized medicine. FD is particularly suitable to probe this question since it is caused by the identical point mutation in nearly all FD patients; yet it can present with mild or severe symptoms and a variable slope of progression.

Our study demonstrates that it is possible to recapitulate disease severity in vitro and that a strong neurodevelopmental phenotype is restricted to patients with severe FD. In contrast, both severe and mild FD patient cells display neurodegenerative phenotypes as illustrated in the progressive SN loss. We postulate that severe FD constitutes a genetic sub-disease within FD, as some of the secondary mutations we discovered, unique to severe patients, could be the cause of additional and/or more severe symptoms (Fig. 5b). Our work raises the question whether in human FD genetic background differences represent distinct disease subtypes or the contribution of more minor disease modifiers. Future studies will be required to address in more detail the molecular mechanism explaining the difference between mild and severe forms of FD. Such studies may also address whether it is possible to identify phenotypes representing intermediate disease severity at the cellular level, though there are currently no obvious criteria to clinically define FD patients with intermediate disease severity.

Independent of defining a detailed molecular mechanism, it is particularly exciting to report that kinetin and SKF-86466 are capable of inhibiting degeneration in neurons already born. Those results provide an additional impetus to pursue the clinical translation of those compounds in FD.

Finally, our results may catalyze similar efforts of modeling severity in other genetic disorders. One important strategy for future studies is the analysis of much larger cohorts of patient-derived PSC lines for a given disease, without “a priori” clinical information on disease severity and carried out under fully blinded conditions. Such a study may confirm stratification along existing clinical subgroups or yield additional subgroups not captured in clinical datasets. The ability to model severity in a genetically defined disease is an important step in moving PSC research forward towards a new era of personalized medicine that captures the biology of each patient individually.

Online Methods

FD fibroblasts

Fibroblasts were purchased from Coriell, under the catalog numbers S1=GM04959, S2=GM04899, S3=GM04589, M1=GM02341, M2=GM02342, M3=GM02343, C1=AG02602, C2=GM00316, A1=GM04895, A2=GM04665, A3=GM04897 and cultured in fibroblast medium (DMEM (Gibco, 11965-092)/10% FBS (Atlanta Biologicals, S11150)/0.5% L-Glutamine (Gibco, 25030-081)/0.5% Pen Strep (Gibco, 15140-122)). For genotyping the FD mutation, gDNA was extracted using the DNeasy Blood&Tissue kit (Qiagen, 69506) and a ~500 bp long PCR product (PCR primers: 5′ CCATAAGGCTCAAAGCGAAA and 5′ TGAGTGTCACGATTCTTTATGC) was sequenced (primers: 5′ GAAAGTCAAGATAAATATAGAGAAC and 5′ TCTGCTAGGAATCTCCACTA). To establish fibroblast growth, each line was seeded at 100,000 cells/6-well and cell numbers were counted every day. For IKBKAP splicing, fibroblast’s RNA was harvested in TRIzol Reagent (Ambion, 15596018). 1 μg RNA was used to generate cDNA (QuantiTect Reverse Transcription kit, Qiagen, 205313) and SYBR green (Applied Biosystems, 4309155) was used for qRT PCR on an Eppendorf Mastercycler according to the manufacturer’s instructions with primers: for wild type IKBKAP 5′ GCAGCAATCATGTGTCCCA and 5′ ACCAGGGCTCGATGATGAA for mutant IKBKAP 5′ CACAAAGCTTGTATTACAGACT and 5′ GAAGGTTTCCACATTTCCAAG. All results were normalized to 18S control, primers: 5′ GGCCCGTAATTGGAATGAG and 5′ GCTATTGGAGCTGGAATTAC. Cycling parameters correspond to those previously used6,14.

Reprogramming, PSC culture and characterization

Fibroblasts were seeded at 150,000 cells per 12-well in 5 wells per line (resulting in 5 PSC sub-clones) and transduced with 4 CytoTune Sendai viruses (Life Technologies, A1378002) encoding Oct-4, Sox2, Klf4 and c-Myc each at an MOI of 316. Cells were fed daily with fibroblast medium. At day 7, the cells were dissociated using trypsin-EDTA (Gibco-Life Technologies, 25300-054) and replated onto 10-cm MEF (GlobalStem, GSC-6105M)-coated plates in HES medium supplemented with 0.5 mM VPA (Tocris Bioscience, 2815), as described in Zeltner et al,18. Cells were fed daily with HES medium/VPA until day 14, then with HES medium only. Between day 20 and 40 PSC colonies were manually picked, expanded and frozen. PSCs were fed daily with HES medium and passaged manually or with trypsin-EDTA18 approximately once per week onto MEFs. PSC cultures were routinely tested for mycoplasma contamination every other week. For the characterization of PSC lines, the cells were immuno-stained with Oct-4 (mIgG2b, Santa Cruz, sc-5279), Nanog (goat, R&D, AF1997) and Sendai virus (rabbit, MBL International, PD029). IKBKAP splicing in PSCs was conducted as described for fibroblasts. The karyotypes of the PSC clones were normal, except at high passages in S3 and M2.

In vitro differentiation of PSCs

HES media, KSR differentiation media, N2 differentiation media, matrigel dish coating and PO/LM/FN dish coating, plating of PSC for differentiation and day 0 to day 11 dual-Smad inhibition differentiation were all done as described in Zeltner et al.18.

CNS neuron differentiation

For the in vitro differentiation to CNS neurons, PSCs were detached using Accutase (Innovative Cell Technologies, AT104) for 20 min. at 37°C, washed twice with 1X PBS and plated onto matrigel (BD, 354234)-coated 6-wells in HES medium supplemented with 10 μM Y-27632 dihydrochloride (Tocris, 1254). When 100% confluency was reached the cells were fed daily with KSR- or N2-medium supplemented with LDN193189 (Stemgent, 04-0074) and SB431542 (Tocris, 1614)18. On day 11 they were accutased, washed twice and replated onto Poly-L Ornithine hydrobromide (Sigma, P3655)/Laminin-1 (R&D Systems, 3400-010-01)/Fibronectin (BD Biosciences, 356008)-coated plates in droplets of approximately 100–150,000 cells per 10 μl as described in Zeltner et al.18, in N2 medium supplemented with 0.02 μg/ml BDNF (R&D Systems, 248-BD), 0.2 mM Ascorbic Acid (Sigma, A4034) and 0.1 μM Purmorphamine (Stemgent, 04-0009). On day 13 the cells were fed with N2 medium/AA/BDNF/Purmorphamine, on day 15, 17 and 19 they were fed with N2 medium/AA/BDNF/ 0.01 mM DAPT (R&D, 2634/50) and on day 20 they were fixed with 4% PFA (Affymetrix, 19943) and immuno-stained with Tbr1 (rabbit, Millipore, AB10554) and Tuj1 (mIgG2a, Covance, MMS-435P-250).

Endoderm differentiation

PSCs were plated as described previously18. The cells were fed daily with RPMI medium containing Glutamax (Invitrogen, 61870-036) supplemented with 0.5% FBS and 100 μg/ml ActivinA (R&D, 338-AC), fixed on day 5 and immuno-stained for Sox17 (goat, R&D, af1924) and DAPI.

Mesoderm/cardiac differentiation

Cardiac differentiations were done as described previously37,38 with the following modifications: RPMI Medium 1640 (Life Technologies, 11875-093)/B27 (Life Technologies, 17504044) media was used instead of StemPro34. Embryoid bodies (EBs) were treated with 10 ng/ml ActivinA (R&D Systems, 338-AC) and 10 ng/ml of BMP4 (R&D Systems, 314-BP) from day 1–3, followed by 10 μM of XAV939 (Stemgent, 04-0046) from day 3–5. Differentiation cultures were assessed for mesoderm formation by flow cytometry staining for KDR/VEGFR2 (Alexa Fluor 647 anti-human CD309, BioLegend, 7D4-6) and PDGFRA (PE anti-human CD140a, BioLegend, 16A-1). The differentiations were carried on until day 20 in RPMI Medium 1640 (Life Technologies, 11875-093)/B27 (Life Technologies, 17504044) media and stained for the anti-cardiac isoform of TroponinT (cTnT) (Thermo Scientific Lab Vision, clone 13–11) by flow cytometry as well as plated onto plastic and stained for cTnT via IF on day 30 of differentiation.

For the quantification of neurons, endoderm and cardiac cells, images from two wells of each of 3–5 differentiations were counted for Tbr1+/Tuj1+ cells or Sox17+ or TroponinT+, respectively and the percentage to total DAPI+ cells was reported.

rNC differentiation

rNC differentiations, FACS isolation, HNK1, Pax6, αAP2a, ASCL1 and Tuj1 staining were carried out as described in Zeltner et al.18. To measure the proliferation rate, FACS isolated rNC cells were plated at 30,000 cells/96-well18. 24 hrs. later they were fed with N2/FGF2/EGF media containing EdU using the Click-iT Plus EdU Alexa Fluor 555 Imaging kit (Life technologies, C10638). 48 hrs. later the cells were fixed and stained according to the manufacturer’s instructions and EdU+/DAPI+ cells were counted for the quantification. The IKBKAP splicing PCR was conducted as described for fibroblasts. The migration (scratch assay) was conducted as described in Zeltner et al.18. In short, FACS isolated rNC cells were plated in PO/LM/FN-coated 96-wells at 20,000–60,000 cells/well and scratched 24 hrs. later using a 200 μl pipet tip. 6–8 wells were fed with media containing 10 ng/ml Hoechst33342 trihydrochloride, trihydrate (Invitrogen, H3570) and reference pictures (0 hrs. = before migration) were taken immediately. The media in the other wells was changed to remove dead cells. The cells were allowed to migrate for 48 hrs. and stained with Hoechst33342 before pictures were taken. The number of migrated cells was quantified automatically by the Konstanz information miner (KNIME) workflow39. Results are expressed as percentage of cells migrated into a ROI (Region of interest) of the total number of cells per picture. Ascl1 qRT PCR was conducted on RNA isolated from FACS sorted rNC cells on day 18 of differentiation. Taqman gene expression assay (ABI, 4304437) for ASCL1 (primers: ABI, Hs00269932_m1 (ASCL1) and 4326321E (HPRT)) was conducted on an Eppendorf Mastercycler machine. Fold-changes were calculated using the comparative Ct method (ΔΔ Ct method) and then normalized to the expression in undifferentiated HES cells. Immuno-staining of rNC cells was conducted one week post FACS isolation as described in Zeltner et al.18.

Peripheral sensory neuron differentiation

The SN differentiations were carried out as described in Chambers et al.,19 with few modifications. In short, PSC were plated on matrigel coated dishes and differentiations were induced once 100% confluency was reached, using a gradient of KSR and N2 differentiation media, supplemented with LDN193189 and SB43154218. From day 2 to 11, 1.5 μM CHIR99021 (GSK3β-inhibitor, activation of Wnt signaling; Stemgent, 04-0004), 10 μM SU5402 (FGF inhibition; BioVision, 1645-1) and 10 μM DAPT (Notch inhibition; Tocris, 2634) were added to the media. On day 12, the cells were detached using Accutase, washed twice and replated into PO/LM/FN coated 96-wells at 280,000 cells per well in N2 differentiation media supplemented with 50X B-27 supplement (Life technologies, 12587-010), 20 ng/ml BDNF (R&D Systems, 248-BD), 20 ng/ml GDNF (Peprotech 450-10), 25 ng/ml β-NGF (PeproTech, 450-01), 100 μg/ml Primocin (InvivoGen, ant-pm-2), 0.6 μg/ml Laminin-1 (R&D Systems, 3400-010-1), 0.6 μg/ml Fibronectin (Fisher Scientific, CB40008A-BD) and 10 μM Y-27632 (Tocris, 1254). For maintenance, the cells were fed with the same media without Y-27632 every 2–3 days. On day 14, 17, 20, 23 and 26 the cells were fixed in 4% PFA and stained with Brn3a (mIgG1, 1:100, Millipore, MAB1585), Tuj1 (rabbit, 1:1000, Covance, MRB-435P-100) and DAPI (1:1000, Sigma-Aldrich, D9542-5M), cleaved Caspase3 (rabbit, 1:100, Cell Signaling, 9661S), Sox10 (goat, 1:100, Santa Cruz, sc-17342), Ki67 (rabbit, 1:500, SP6, RM9106-S1), Sox2 (rabbit, 1:400, Cell Signaling, 3579P) and αSMA (mIgG2a, 1:1000, Sigma, A5228-200UL). For the quantification, Brn3a+/Tuj1+ cells in pictures taken from 2–3 wells from 3–9 independent differentiations were counted (Figure 2b) or the percentage of DAPI+ of Brn3a+ cells were quantified using the Meta Express software (Figure 5b). IKBKAP splicing was done was described for fibroblasts. For the Sox10 expression time course, cells were harvested for RNA extraction at day 4, 8, 12 and 16 without replating. The real-time RT PCR was conducted using SYBR SsoFast EvaGreen Supermix (BioRad, 172–5202) and the QuantiTect Primer Assay for Sox10 (Qiagen, QT00055405). The alpha2adrenergic receptor PCR was previously described40. Drug treatment of SNs for the rescue of FD phenotypes were done either from day 0 throughout the entire differentiation or from replating at day 12. Kinetin (Sigma-Aldrich, K0753-5G) and SKF86466 hydrochloride (Tocris, 3866) were used at 10 μM.

Autonomic neuron differentiation

PSCs were plated and dual-Smad inhibition differentiation until day 11 was conducted as reported in Zeltner et al.18. For Wnt-based NC induction the culture was supplemented with 3 μM CHIR99021 from day 2 to 14 and the cells were sorted for CD49d (1:800, PE-Cy7 anti-human CD49d, BioLegend, 304314) on day 11. The cells were aggregated in Ultra Low Attachment 24-well culture plates (Fisher Scientific, 3471) as 3D spheroids (500,000 cells/well) in Neurobasal (NB, Life Technologies, 21103-049) media supplemented with N2 supplement (Stem Cell Technologies, 07156) and B27, 3 μM CHIR99021 and 10 ng FGF2 (R&D Systems, 233-FB-001MG/CF). On day 14, the spheroids were plated onto PO/LM/FN coated 24-wells in NB+N2 supplement+B27 containing GDNF and Ascorbic Acid (Sigma, A8960-5G). On day 25, cells were fixed in 4% PFA and stained for Phox2a (rabbit, 1:800, kindly gifted from Dr. J-F Brunet, France), Tuj1 and DAPI. For the AN marker expression, RNA was extracted at day 30, 40, 50 and 60 and real-time RT-PCR was conducted either using SYBR SsoFast EvaGreen Supermix (QuantiTect primer assays for ASCL1: QT0023775, Phox2a: QT00215467, ChAT: QT00029624, EDNRB: QT00014343, Six1: QT00010584, CRNA3: QT00015386, CRNB4: QT00010570, VMAT1: QT00028420, VMAT2: QT00059857) or Taqman gene expression assays (primers for Scg10: Hs00199796_m1, TH: Hs00165941_m1, DBH: Hs01089840_m1). For the stress conditions in the AN survival assay, the cells were maintained either in the above mentioned media or in NB only or in NB only supplemented with 0.1 μM Carbonyl cyanide 3-chlorophenylhydrazone (CCCP, Sigma, C2759-100MG). For intracellular FACS analysis, the cells were dissociated using Accutase, then fixed and permeabilized using BD Cytofix/Cytoperm (BD Bioscience, 554723) according to the manufacturer’s instructions. Primary antibody staining for ASCL1 (mIgG1, 1:200, BD Pharmingen, 556604) and TH (rabbit, 1:700, Pel-Freez, P40101-0) was done at 4 °C over night and secondary antibody staining was done at RT for 30 min. using AlexaFluor 647 goat anti mouse IgG1 (Invitrogen, A21240) for ASCL1 and AlexaFluor 647 goat anti rabbit (Invitrogen, A21245) for TH. The analysis was done on a flow Cytometer and using the FlowJo software. Intracellular FACS and immunohistochemistry analysis was done as described above. ISL1 (mouse IgG2b, 1:200, DSHB, 39.4D5-c).

Genetic rescue


The 3′ homology arm spans from intron18 to intron 20 of the IKBKAP gene. The FD point mutation is located in intron 20, 5bps downstream of exon 20. The gRNA binding site is 38bps upstream of the FD mutation. The selection cassette is located in intron 20, thus upon correction of the splicing site (FD mutation) it will be removed from the protein. The 3′ homology arm spans intron 20.

Cloning of IKBKAP donor plasmid

The 3′ and 5′ homology arms are 697 bps and 682 bps long, respectively and were amplified from C1.2 PSC genomic DNA using primers 5′CCCAGGGAGATTGTACACAC and 5′CGACTCTAGAGGATCCCAATGCCAAGCT or 5′CGGTACCCGGGGATCAGTGGGAGAGCTAAGGCACA and 5′TATGGATCCGCTAGCTTCACAGGCCACCTAAAACC, respectively. The selection cassette consists of the human PGK promoter followed by the puromycin resistance gene. The three fragments were ligated into a linearized pUC19 vector backbone using the Gibson Assembly Master Mix (New England Biolabs, E2611S) and the plasmid insert was sequenced entirely.

gRNA design

The CRISPR/Cas9 technology31 was used to design the gRNA and induce the double strand break to enhance homologous recombination to repair the FD mutation in IKBKAP. Option A from the hCRISPR gRNA synthesis protocol (deposited was used to generate the gRNA. The website was used to generate the gRNA target sequence as well as to check for potential off-target effects. The gRNA with the best score was chosen (shown in bold below), synthesized as a gBlock from IDT and cloned into a TOPO vector (Invitrogen, 450245). The target sequence is located 39 bps upstream of the FD mutation.


Homologous recombination in severe FD PSCs

10 million S2 FD PSCs were nucleofected with 10 μg IKBKAP donor plasmid, 5 μg gRNA-TOPO vector and 5 μg Cas9-GFP plasmid (Addgene, plasmid #44719) using solution I and program A23 from the Amaxa Human Stem Cell Nucleofector Kit 1 (Lonza, VPH-5012). The cells were plated on MEFs in half fresh HES media and half HES media conditioned on S2 FD PSCs supplemented with 10 μM Y-27632 dihydrochloride. 48 hours later the cells were detached with Accutase and GFP+ cells were isolated by FACS using low pressure/μm nozzle settings. The 122,000 GFP+ cells were plated on fresh puromycin-resistant MEFs in a 10 cm dish. 72 hours post sort puromycin selection was initiated (1 μg/ml Millipore, 540411-100MG) and the cells were fed daily until colonies appeared, which were manually picked, expanded and frozen.


Genomic DNA was extracted from 21 clones (DNeasy Blood & Tissue kit (Qiagen, 69506) and screened using a 3 primer PCR strategy. Primers 5′GCTAGCGGATCCATAACT and 5′TGCCATTTGATTTCAGTCTCC bind on either side of the homology cassette and primer 5′AGGGGGCTGGAAGAGCTA binds inside the selection cassette. The expected amplicon sizes are shown in Fig. 4a. To sequence the FD mutation, a 483 bp amplicon surrounding the FD mutation was PCR amplified and sequenced.

CRISPR/Cas9 off target site sequencing


Western blot analysis

For protein analysis of neural crest cells, HNK1/p75 double positive cells were isolated by cell sorting and cells were pelleted and snap frozen. Pluripotent stem cells were harvested by scrapping after removal of the MEF layer. To extract protein, the cell pellet was resuspended in 1x RIPA buffer (Sigma # R 0278) containing 1x Halt Protease and Phosphatase inhibitor cocktail (Thermo #87786). In few samples to obtain sufficient protein for analysis, cell pellets from two separate neural crest inductions were combined or FACS isolation was conducted on single positive cells. For western blotting, a total of 10 μg of protein was mixed with 4x Novex sample buffer (Thermo #NP0007) and 10x NuPAGE reducing agent (Thermo #NP0004) to a final concentration of 1x. Lysates were run on 4–12% pre-cast Bis-Tris gels (Novex, NP0322BOX), transferred onto PVDF membranes (Thermo, 88518) and blocked with 5% milk. Antibodies used for western blotting analysis were: β-Actin (Sigma, A2228) and IKAP (GeneTex, GTX110274). Band intensity was measured using NIH ImageJ and normalized to β-actin. For the quantification of WBs in Supplementary Fig. 13b and Supplementary Fig. 14d, e, the control lines (S2 and HES) represent the same data points, run on the same gels, for easier comparison.

RNA sequencing

Global gene clustering

First round

Paired-end RNA sequencing was performed on H9-ESCs, mild (M)-FD PSCs, severe (S)-FD PSCs and subsequent NC progenitors with no replicates. Reads were mapped to the human genome build hg19 using the TopHat software41. Reads that aligned with no more than two base pair mismatches were accepted for downstream analysis. Second round: The FASTQ files were mapped to the target genome (UCSC HG19) using the rnaStar aligner42 that maps reads genomically and resolves reads across splice junctions. We used the 2 pass mapping method43, in which reads are passes twice. The first mapping pass uses a list of known annotated junctions from Ensemble. Novel junctions found in the first pass are then added to the known junctions and a second mapping pass is done. After mapping we compute the expression count matrix from the mapped reads using HTSeq ( and one of several possible gene model databases. The raw count matrix generated by HTSeq are then be processed using the R/Bioconductor package DESeq (, which is used to both normalize the full dataset and analyze differential expression between sample groups. A heatmap was generated using the heatmap.2 function from the gplots R package. The data plot was the mean centered normalized log2 expression of the top 100 significant genes. For simple hierarchical clustering the correlation metric was used (Dij = 1 - cor(Xi,Xj)) with the Pearson correlation on the normalized log2 expression values.

GO term analysis

Genes that were expressed significantly different when compared in the rNC state to the PSC state were subjected to GO term analysis using the DAVID functional annotation tool ( The occurrence of GO terms with a Bonferroni score ≤0.05 were quantified. Certain GO terms were combined in order to score 10 GO terms per line, the numbers were displayed as percentages of the total, in pie charts. The analysis was done independently for each of the FD severity groups (S, M, C).

Whole-exome sequencing

500ng of DNA was captured by hybridization using the SureSelect XT HumanAllExon V4 (Agilent). Samples were prepared according to the manufacturer instructions. PCR amplification of the libraries was carried out for 6 cycles in the pre-capture step and for 7 cycles post capture. Samples were barcoded and run on a Hiseq 2500 in a 50bp/50bp Paired end run, using the TruSeq SBS Kit v3 (Illumina). The average number of read pairs per sample was 77 million, the average duplication rate was 2.1%, and the mean target coverage was over 100x. Intergenic, UTRs, intronic and synonymous variations were removed in our initial search for variants in the severe patients. In an autosomal recessive approach, 26 genes carried homozygous variations in all 3 severe patients with a MAF of less than 1% in the 1000 genomes database. 63 variations were dismissed due to their high frequency (MAF>1%) in the Exac database, leaving 7 genes carrying 14 variations as candidates. All of these 7 genes carry the same variations in the mild patients and therefore were discarded as candidate modifier genes. In an autosomal dominant approach, 77 genes carried heterozygous variations in all 3 severe patients with a MAF of less than 1% in the 1000 genomes database. 446 variations were removed based on their high frequency (MAF>1%) in the Exac database, leaving 23 genes carrying 51 variations as candidates genes commonly mutated between all severe patients. Of these, 15 carry the same variations in the mild patients and therefore were discarded as candidate modifier genes. 5 of the remaining 8 genes harbor variations with the frequency lower than 1% in the 1000 genome database predicted to be damaging as therefore not prioritized as candidates under a complete penetrance hypothesis. FAT2, KIA1211 and LAMB4 thus are the primary hit candidates. A connectome analysis using IKBKAP as the core gene on the website ( and including all of the initial candidate genes identified URI1, LCA10, MSH3, YEATS2 as being closely related to IKBKAP and could therefore be of interest for further studies.

Statistical analysis

All statistical comparisons between PSC lines were done using one-way ANOVA. Multiple comparisons were done among samples in each subfigure. The test specifics described in the figure legends were compiled from pooling data from the two severe, two mild or two control lines, respectively. Normal distribution of the raw data was not assumed, since the sample sizes were too small (N<5), instead individual data points were plotted in the figures. Similar variance among patient lines was calculated using the Brown-Forsythe test and was only assumed in Figure 1b, ,2d,2d, 3d, 3e, Supplementary Figure 3b, 4b, 4d, 12a, 12d, 13b (left and right), 14d.

Reproducibility definitions

Biological replicates (N) are defined as differentiations performed at least one cell split apart, which is generally one week. Number of replicates are indicated in each figure legend as well as represented by individual plotted data points in the graphs. Sample sizes were not predefined, but generally at least 3 biological replicates were performed; for more subtle phenotypes the sample size was subsequently increased. Exclusion criteria: In each differentiation experiment the HES control line was included and experiments were excluded from analysis if the HES control line did not perform as expected, probably due to technical difficulties. Researchers were never blinded. None of the cell lines used here are currently listed in the database of commonly misidentified cell lines maintained by ICLA.

Supplementary Material

Supplementary Information

Supplementary Figure 1. Characterization of FD fibroblasts. a. List of obtained FD and control fibroblasts indicating clinical parameters. FD symptoms that differ between severe and mild patients are depicted (complete symptom list in Table 1). Samples in bold were subsequently reprogrammed to PSCs and used further in this study. n/a=data not available, HES=H9 (WA09). b. The FD point mutation in intron 20 of the IKBKAP gene was sequenced from a 500 bp PCR product to confirm the patient’s fibroblast genotype. c. Growth properties were analyzed in the FD fibroblast lines to exclude differential growth or passage number effects on later analysis of the cells. N=3 biological replicates. Two-way ANOVA, n.s., p=0.41, F=1.1. Data are presented as mean ± S.E.M. and circles indicate biological replicates from individual differentiations, throughout.

Supplementary Figure 2. Generation of FD PSC lines. a. Scheme summarizing PSC reprogramming and quality control (QC). b. Classic PSC morphology of reprogrammed lines. c. PSC lines were analyzed for expression of pluripotency markers OCT4 and NANOG and immuno-staining to show that the lines do not contain the pluripotency factor delivery vehicle Sendai-virus (SeV) anymore. This was achieved after approximately 10 passages as PSCs, which is also thought to be enough time to reset the age-related epigenetic memory1. Scale bar in b, c: 500 μm. d. Analysis of wild type and mutant IKBKAP transcripts in the FD-PSC lines. N=2–3 biological replicates, one-way ANOVA, ***p<0.0002, F=11.86, Tukey’s multiple comparison test. S versus C/HES: ***p<0.0005, q=7.1. S versus M: ns, q=0.01. M versus C/HES: ***p<0.0005, q=6.9. M/S versus A: ns, q=0.2/0.2. C/HES versus A: ** p<0.005, q=5.6.

Supplementary Figure 3. Differentiation capacity of PSC lines derived from FD patients and controls into all three germ layers in vitro. a. FD cells were subjected to in vitro differentiation protocols for CNS forebrain neurons (ectoderm), endoderm and cardiac mesoderm. Scale bar: 100 μm. b. Quantification of the differentiation potential across PSC lines. Left: The percentage of TBR1+/TUJ1+ forebrain neurons of total cells do not differ between PSC lines. N=3 biological replicates, one-way ANOVA, n.s. p=0.72, F=0.45, Tukey’s multiple comparison test. S versus M: q=0.97. S versus C/HES: q=0.54/0.4. M versus C/HES: q=1.51/1.36. C versus HES: q=0.12. Middle: The percentage of SOX17+ endoderm cells does not differ between PSC lines. N=3 biological replicates, one-way ANOVA, n.s. p=0.3, F=1.4, Tukey’s multiple comparison test. S versus M: q=0.38. S versus C/HES: q=0.99/2.25. M versus C/HES: q=1.37/2.66. C versus HES: q=1.28. Right: The percentage of TroponinT+ mesoderm-derived cells does not differ between PSC lines. N=5–6 biological replicates, one-way ANOVA, n.s. p=0.23, F=1.58, Tukey’s multiple comparison test. S versus M: q=0.3. S versus C/HES: q=1.01/1.98. M versus C/HES: q=0.67/2.2. C versus HES: q=2.8.

Supplementary Figure 4. In vitro generation of rNC cells from FD PSCs. a. During the differentiation protocol for the generation of rNC cells, all FD and control PSCs pass through a neural rosette stage captured at day 18 (stained in DAPI), where HNK1+/αAP2+ cells migrate away from PAX6+ neural rosette cells. Scale bar: 100 μm. b. Representative FACS blot of the purification of rNC cells at day 18 of differentiation. c. FACS analysis reveals no significant difference between PSC lines in their capacity of generating HNK1+/p75+ rNC cells at day 18. N=7–17 biological replicates, one-way ANOVA, n.s. p=0.56, F=0.59, Tukey’s multiple comparison test. S versus M: q=0.78. S versus C/HES: q=0.71. M versus C/HES: q=1.4. Mean ± S.E.M. d. HNK1+/p75+ sorted rNC cells were labeled with EdU one day after FACS isolation for 48 hrs. EdU+ cells were quantified, showing no significant difference in cell divisions. N=3–4 biological replicates, one-way ANOVA, n.s., p=0.52, F=0.77, Tukey’s multiple comparison test. S versus M: q=2.08. S versus C/H9: q=0.74/0.67, M versus C/HES: q=1.34/1.56. C versus HES: q=0.13.

Supplementary Figure 5. First round of whole genome RNA sequencing analysis of FD PSCs compared to rNCs. a. Dendrogram showing the relationship between PSC and rNC cells in FD lines. Comparison is based on all genes. b. GO term analysis was performed with all genes that were up- or down-regulated from the rNC compared to PSC state. The pie charts were produced by counting all GO terms with a Bonferroni score <0.05 and an FDR score <0.5 and plotting them as the percentage of all GO terms. N=1 biological replicate each (M1, S2, C1) was analyzed in each condition. Arrows highlight how mild cells’ expression signature is more similar to control rather than severe FD cells.

Supplementary Figure 6. Second round of whole RNA sequencing of FD rNC cells. a. Global gene analysis confirms the clustering of mild with control instead of severe FD at the rNC stage. b. Heat map of the top 100 differentially expressed genes between mild and severe FD rNC. Groups lined in pink or green were analyzed for GO terms in c. c. Unbiased top 10% of all GO terms from the two clusters in b are shown. d. Top 10 genes significantly upregulated in mild compared to severe rNC. N=2 (S2) or N=3 (M1 and C1) biological replicates were analyzed.

Supplementary Figure 7. SN phenotypes in FD lines and establishment of an autonomic neuron (AN)-like differentiation protocol. a. Generation of SNs2 in HES cells yields on average 33% BRN3A+ cells over DAPI+ cells. Scale bar 200 μm. b. FD PSC lines were differentiated into SNs and bright field images were taken at day 12 to show neuronal cultures with extending axons or the lack thereof in S2. Scale bar: 100 μm. c. Time course analysis of SOX10 expression by qRT-PCR in the SN protocol in control and FD lines. SOX10 expression is plotted relative to the day 4 unpatterned (LSB) control. d. Schematic of the AN-like differentiation protocol. NC=neural crest, AN=autonomic neuron, KSR/N2=knock-out serum replacement to N2 differentiation media gradient. NB=Neurobasal media, N2=supplement, B27=supplement, LDN193189=BMP inhibitor, SB431542=TGF-β inhibitor, FGF2=fibroblast growth factor 2, AA=Ascorbic Acid, GDNF=Glial-derived neurotrophic factor, CHIR9902=GSK3β-inhibitor, spheroid=culture in suspension. e. Correlation of SOX10 and CD49D expression in NC. At day 11 of the AN-like differentiation protocol, about 62% of NC cells are CD49D+/SOX10+ (immuno-staining on the left and FACS plot, on the right). Post CD49D-FACS isolation all cells are SOX10+ 3. Scale bar: 200 μm.

Supplementary Figure 8. Validation of the AN-like differentiation protocol. a. Time course analysis of AN marker expression in AN-like cells derived from HES cells. qRT-PCR analysis of sympathetic neuron marker genes, ASCL1, PHOX2A, SCG10, TH and DBH. CHAT, a marker for the parasympathetic neuron lineage, EDNRB, a marker for the enteric neuron lineage and SIX1, a marker for the placode lineage were used to exclude major contamination with related cell types during the protocol. N=3–4 biological replicates. b. Immuno-staining of day 35 AN-like cells. Scale bar: 100 μm. c. Immunohistochemistry of day 50 AN-like cells stained for DBH and TUJ1. d. qRT-PCR analysis of mature sympathetic neuron marker genes nicotinic acetylcholine receptor (CHRNA3, CHRNB4), and vesicular monoamine transporter (VMAT1 and VMAT2) at day 30, 40, 50, 60 and 70 of differentiation. N=3–4 biological replicates.

Supplementary Figure 9. FD phenotypes in AN-like cells. a. Both mild and control FD PSC lines can generate AN-like cells. The cells were stained for PHOX2A or TUJ1 and DAPI at day 25 and bright field images were taken at day 45. Scale bar: 500 μm. b. Efficiency of generation of AN-like cells. At day 30 of differentiation, 10 to 25% of cells in FD or HES lines, respectively are TH+ and ASCL1+ measured by intracellular FACS. N= 3 biological replicates is shown. This data is also represented as part of Supplementary Figure 11a. c–d. The identity of the non-AN-like cells is myofibroblast-like with few SNs. N= 3 biological replicates (c). Pictures were taken in specific areas of the well to highlight the particular cell types (d). Scale bar: 200 μm.

Supplementary Figure 10. Validation of SN survival assays in FD PSC-derived SNs. a. PSC-derived SNs in mild and severe FD die by apoptosis. SNs from HES control, mild or severe FD PSC lines were immuno-stained for activated Caspase3 and BRN3A at day 14, 17 and 20. Scale bar: 200 μm. b. Non-SNs in the SN cultures are non-proliferating cells of mesenchymal origin. Day 14 SNs derived form control and FD lines were immuno-stained for the NC progenitor marker SOX10, the proliferation marker Ki67, the neural stem cell marker SOX2 and the mesenchymal marker α-SMA in red. All pictures were counter-stained with TUJ1 in green and DAPI in blue. Scale bar: 200 μm.

Supplementary Figure 11. Survival assay in FD PSC-derived AN-like cells. a. AN-like cells were differentiated, FACS purified for CD49D at day 11 and plated in equal numbers in neural spheroids for 4 days. They were plated down as attachment cultures on day 14 and maintained in optimal growth media (blue line), growth factor restricted media (green line) or growth factor restricted media complemented with carbonyl cyanide 3-chlorophenylhydrazone (CCCP 0.1 μM, red line) for up to 60 days. Every 10 days cells were harvested and the percentage of ASCL1+ or PHOX2A+ AN-like cells were quantified by intracellular FACS analysis. N=3 biological replicates. b. Bright field pictures of day 50 AN-like cells derived from control or mild FD cells don’t show distress in the neuron cultures even upon growth factor removal and toxin addition. Scale bar 100 μm.

Supplementary Figure 12. Targets of drug compounds that rescue degeneration in mild FD SNs. a. α2-adrenergic receptors are expressed on SNs. PCR analysis of subunits A, B and C of the α2-adrenergic receptor shows expression in SNs from mild FD patients and control HES line. b. Kinetin rescues SN degeneration via increased expression of IKBKAP. Wild type to mutant ratio of IKBKAP splicing was measured by qRT-PCR during the survival rescue drug treatment. N=3 biological replicates.

Supplementary Figure 13. IKBKAP splicing and protein levels in multiple tissues. a. The ratio of wild type to mutant IKBKAP splicing in multiple cell types derived from PSCs. Transcripts are low in both severe and mild in all cell types. Data for the two patients of each group were combined, N=3 biological replicates each, one-way ANOVA, Tukey’s multiple comparison tests. Fibroblasts: ***p<0.0003, F=16.77. S versus M: n.s, q=0.36, S/M versus C: ***p<0.0005, q=7.5/7.2. Endoderm: ****p<0.0001, F=36.2. S versus M: n.s, q=0.2, S/M versus C: ****p<0.00005, q=10.6/10.1. Cardiomyocytes: **p<0.0011, F=12.7. S versus M: n.s, q=0.5, S/M versus C: **p<0.005, q=5.3/6.7. Cortical neurons: ****p<0.0001, F=36. S versus M: n.s, q=0.04, S/M versus C: ****p<0.00005, q=10.2/10.9. b. Wild type IKAP protein levels in PSCs and rNC of FD patients. Non-disease associated isoforms of wild type IKAP are shown at 150 kDa and 110 kDa. Actin is used as the loading control. Quantification in PSCs: N=4–5 biological replicates, one-way ANOVA, Tukey’s multiple comparison tests. ***p<0.0002, F=12.4. S verus M: n.s, q=0.8, S versus C/HES: **p<0.005, q=5.7, M versus C/HES: ***p<0.0005, q=6.2. Quantification in rNC: N=3 biological replicates, one-way ANOVA, Tukey’s multiple comparison tests. **p<0.003, F=8. S versus M: *p<0.05 q=3.6, S versus C/HES: **p<0.005, q=5.6, M versus C/HES: n.s. q=1.9.

Supplementary Figure 14. Quality control of gene targeted clones. a. S2-PSCs were gene targeted to revert the FD mutation in IKBKAP to a heterozygous genotype using the CRISPR/Cas9 technology4, 5. The table in a shows expected results from the PCR screening and sequencing of targeted clones. b. PCR screen of representative candidate rescued clones and parental mutant (S2) or control wild type (HES) lines. c. Roughly 500 bps surrounding the FD mutation in IKBKAP were PCR amplified and sequenced. We found three clones to be heterozygous targeted, S2-rescued6 remained mutant. d, e. Western blot analysis of wild type IKAP relative to actin in PSCs and rNC. Statistics in PSCs: N=5–6 biological replicates, one-way ANOVA, Dunnett’s multiple comparison test. *p<0.05, F=3.1. S2 versus S2-rescue/A1/HES: n.s. q=0.8/0.03/2.2. Statistics in rNCs: N=3 biological replicates, one-way ANOVA, Dunnett’s multiple comparison test. n.s. F=2. S2 versus S2-rescue/A1/HES: n.s. q=0.5/1.2/2.4.

Supplementary Figure 15. Assessment of CRISPR/Cas9 off-target sites in gene corrected lines. a. The top 24 predicted off-target sites ( are listed for the IKBKAP gRNA used in the genetic rescue. The top row shows the used gRNA for the IKBKAP rescue studies. b. The top 5 sites for both non-coding and coding regions were Sanger sequenced, each with 2 PCR primer pairs and 2 sequencing primers in the S2 parent line and the three gene targeted clones. Representative sequence traces of one non-coding (ID#1) and one coding (ID#6) sites are shown. Blue rectangle=gRNA target sequence, red arrowhead=predicted cut site, red line=PAM sequence. We did not find any of the 10 sites mutated in any of the assayed clones. ID#22 did not yield a PCR product in all of the clones tested and thus its sequence could not be assessed.

Supplementary Figure 16. Schematic representation of whole exome sequencing analysis. Variants occurring in all three severe FD patients (locus homogeneity) were separated into homozygous or heterozygous mutations and analyzed separately. The path of analysis with the stepwise exclusion and ranking parameters are shown leading to the three top candidates LAMB4, FAT2 and KIAA1211. MAF=Minor Allele Frequency; 1KG=1000 genomes phase 3 ( and Exac ( CADD=Combined Annotation Dependent Depletion6, 7; GDI= Gene Damaging Index: GDI8.

Supplementary Figure 17. Effects of LAMB4, FAT2 and KIAA1211 mutations on mRNA expression in rNC of those genes. BAM files from the RNA sequencing data in severe, mild and control rNC cells from Supplementary Figure 6 were used to investigate weather the secondary mutation found in severe but not mild FD patients lead to a change in mRNA expression. The exons of each gene are represented in blue, filled rectangles; arrows indicate strand direction; red triangles indicate the mutation position. Mutated exons are enlarged and the RNA sequencing traces are depicted for each of the patient groups. Maximum scale in LAMB4 and FAT2 is 10 and in KIAA1211 it is 85.

Supplementary Table 2

Supplementary Table 1. Complete listing of clinical data available for FD patients in this study. Patients in bold were reprogrammed to PSC lines and used in various assays throughout this study. Listed symptoms complement Supplementary Figure 1a. #assumed symptoms, based on general FD diagnostics (, *symptom likely originating from the additional ASPA mutation.

Supplementary Table 2. List of candidate gene analysis from whole exome sequencing. Genes identified as having a close connection to IKBKAP, i.e. via the ‘connectome9’ analysis are highlighted by blue background. Final hit modifier gene is highlighted by green background.

Supplementary Table 3. Listing of all the cell lines and phenotype that were used in this study. *=data not shown, na=not applicable, nd=not done, blue=phenotype shown in Lee et al., 2009 and was repeated here in an newly reprogrammed severe PSC line originating from the same patient.


We would like to thank J-F. Brunet, Ecole Normale Supérieure, Paris, France for providing us with the PHOX2A antibody. S. Irion and M. Tomishima for critical reviewing of the manuscript. A. Hudon for general help. H. Ralph from the Weill Cornell Cell Screening Core for image quantification. Bioinformatics, Integrated Genomics Operations (funded by the NCI Cancer Center Support Grant (CCSG, P30 CA08748), Cycle for Survival and the Marie-Josée and Henry R. Kravis Center for Molecular Oncology) and Flow Cytometry Core Facilities at Sloan Kettering Institute. This work was supported by a fellowship for advanced researchers from the Swiss National Science Foundation (N.Z) and through contracts from NYSTEM (C026446; C026447) and the Tri-institutional stem cell initiative (Starr Foundation) (L.S.). Robertson Investigator Award from New York Stem Cell Foundation (G.L.), Maryland Stem Cell Research Funding (MSCRF/TEDCO) (G.L.) and the Adrienne Helis Malvin Medical Research Foundation (G.L.).


Accession codes

FASTQ files for the first round of RNA sequencing are deposited at the NCBI Gene Expression Omnibus (with the accession number GSE67073 for first round and GSE77513 for second round of RNA sequencing). Whole exome sequencing files are deposited under BioProject accession number: SRP070100.

Author Contributions

N.Z.: Design and conception of the study, writing of manuscript, cell maintenance, reprogramming, GO term analysis, directed differentiation and survival assays, protocol optimization, gene targeting of PSCs, cellular/molecular assays. F.F.: AN differentiation protocol establishment, execution and survival assays in ANs. J.T.: RNAseq data analysis. N.D.: Design and execution of cardiac mesoderm differentiations. B.Z.: Data quantification of scratch assay. N.S/M.A.S.: Western blot. G.L.: Reprogramming of S3 FD PSCs, mentoring. J-L.C., L.S. and F.L.: Advice and analysis of whole exome sequencing. L.S.: Design and conception of the study, data interpretation, writing of manuscript, mentoring.

Competing Financial Interest Statement

The authors have no conflicting interests to disclose.


1. Inoue H, Nagata N, Kurokawa H, Yamanaka S. iPS cells: a game changer for future medicine. The EMBO journal. 2014;33:409–417. [PubMed]
2. Zeltner N, Studer L. Pluripotent stem cell-based disease modeling: current hurdles and future promise. Current opinion in cell biology. 2015;37:102–110. [PubMed]
3. Bellin M, Marchetto MC, Gage FH, Mummery CL. Induced pluripotent stem cells: the new patient? Nat Rev Mol Cell Biol. 2012;13:713–726. [PubMed]
4. Lee G, et al. Modelling pathogenesis and treatment of familial dysautonomia using patient-specific iPSCs. Nature. 2009;461:402–406. [PMC free article] [PubMed]
5. Axelrod FB. Familial dysautonomia. Muscle & nerve. 2004;29:352–363. [PubMed]
6. Slaugenhaupt SA, et al. Tissue-specific expression of a splicing mutation in the IKBKAP gene causes familial dysautonomia. Am J Hum Genet. 2001;68:598–605. [PubMed]
7. Axelrod FB, Goldberg JD, Ye XY, Maayan C. Survival in familial dysautonomia: Impact of early intervention. J Pediatr. 2002;141:518–523. [PubMed]
8. Dong J, Edelmann L, Bajwa AM, Kornreich R, Desnick RJ. Familial dysautonomia: detection of the IKBKAP IVS20(+6T --> C) and R696P mutations and frequencies among Ashkenazi Jews. American journal of medical genetics. 2002;110:253–257. [PubMed]
9. Pearson J, Pytel BA, Grover-Johnson N, Axelrod F, Dancis J. Quantitative studies of dorsal root ganglia and neuropathologic observations on spinal cords in familial dysautonomia. Journal of the neurological sciences. 1978;35:77–92. [PubMed]
10. Pearson J, Pytel BA. Quantitative studies of sympathetic ganglia and spinal cord intermedio-lateral gray columns in familial dysautonomia. Journal of the neurological sciences. 1978;39:47–59. [PubMed]
11. Axelrod FB, et al. Progressive sensory loss in familial dysautonomia. Pediatrics. 1981;67:517–522. [PubMed]
12. Slaugenhaupt SA, et al. Rescue of a human mRNA splicing defect by the plant cytokinin kinetin. Hum Mol Genet. 2004;13:429–436. [PubMed]
13. Axelrod FB, et al. Kinetin improves IKBKAP mRNA splicing in patients with familial dysautonomia. Pediatr Res. 2011;70:480–483. [PMC free article] [PubMed]
14. Lee G, et al. Large-scale screening using familial dysautonomia induced pluripotent stem cells identifies compounds that rescue IKBKAP expression. Nat Biotechnol. 2012;30:1244–1248. [PMC free article] [PubMed]
15. Miller JD, et al. Human iPSC-Based Modeling of Late-Onset Disease via Progerin-Induced Aging. Cell Stem Cell. 2013;13:691–705. [PMC free article] [PubMed]
16. Fusaki N, Ban H, Nishiyama A, Saeki K, Hasegawa M. Efficient induction of transgene-free human pluripotent stem cells using a vector based on Sendai virus, an RNA virus that does not integrate into the host genome. Proc Jpn Acad Ser B Phys Biol Sci. 2009;85:348–362. [PMC free article] [PubMed]
17. Lee G, et al. Isolation and directed differentiation of neural crest stem cells derived from human embryonic stem cells. Nature biotechnology. 2007;25:1468–1475. [PubMed]
18. Zeltner N, Lafaille FG, Fattahi F, Studer L. Feeder-free derivation of neural crest progenitor cells from human pluripotent stem cells. Journal of visualized experiments: JoVE. 2014 [PubMed]
19. Chambers SM, et al. Combined small-molecule inhibition accelerates developmental timing and converts human pluripotent stem cells into nociceptors. Nature biotechnology. 2012;30:715–720. [PMC free article] [PubMed]
20. Wainger BJ, et al. Modeling pain in vitro using nociceptor neurons reprogrammed from fibroblasts. Nature neuroscience. 2015;18:17–24. [PMC free article] [PubMed]
21. Mica Y, Lee G, Chambers SM, Tomishima MJ, Studer L. Modeling neural crest induction, melanocyte specification, and disease-related pigmentation defects in hESCs and patient-specific iPSCs. Cell reports. 2013;3:1140–1152. [PMC free article] [PubMed]
22. Fattahi F, et al. Deriving human ENS lineages for cell therapy and drug discovery in Hirschsprung disease. Nature. 2016;531:105–109. [PMC free article] [PubMed]
23. Howard MJ. Mechanisms and perspectives on differentiation of autonomic neurons. Dev Biol. 2005;277:271–286. [PubMed]
24. Grundy D, Schemann M. Enteric nervous system. Current opinion in gastroenterology. 2007;23:121–126. [PubMed]
25. Schlosser G. Induction and specification of cranial placodes. Dev Biol. 2006;294:303–351. [PubMed]
26. Oh Y, et al. Functional Coupling with Cardiac Muscle Promotes Maturation of hPSC-Derived Sympathetic Neurons. Cell Stem Cell. 2016 [PMC free article] [PubMed]
27. Axelrod FB, et al. Kinetin improves IKBKAP mRNA splicing in patients with familial dysautonomia. Pediatr Res. 2011;70:480–483. [PMC free article] [PubMed]
28. Dietrich P, Alli S, Shanmugasundaram R, Dragatsis I. IKAP expression levels modulate disease severity in a mouse model of familial dysautonomia. Hum Mol Genet. 2012;21:5078–5090. [PMC free article] [PubMed]
29. Morini E, et al. Sensory and Autonomic Deficits in a New Humanized Mouse Model of Familial Dysautonomia. Hum Mol Genet. 2016 [PMC free article] [PubMed]
30. Ding Q, et al. Enhanced efficiency of human pluripotent stem cell genome editing through replacing TALENs with CRISPRs. Cell stem cell. 2013;12:393–394. [PMC free article] [PubMed]
31. Mali P, et al. RNA-guided human genome engineering via Cas9. Science. 2013;339:823–826. [PMC free article] [PubMed]
32. Adzhubei IA, et al. A method and server for predicting damaging missense mutations. Nat Methods. 2010;7:248–249. [PMC free article] [PubMed]
33. Kumar P, Henikoff S, Ng PC. Predicting the effects of coding non-synonymous variants on protein function using the SIFT algorithm. Nat Protoc. 2009;4:1073–1081. [PubMed]
34. Itan Y, et al. The human gene damage index as a gene-level approach to prioritizing exome variants. Proc Natl Acad Sci U S A. 2015;112:13615–13620. [PubMed]
35. Itan Y, et al. The mutation significance cutoff: gene-level thresholds for variant predictions. Nat Methods. 2016;13:109–110. [PMC free article] [PubMed]
36. Itan Y, et al. HGCS: an online tool for prioritizing disease-causing gene variants by biological distance. BMC Genomics. 2014;15:256. [PMC free article] [PubMed]