Norepinephrine transporter defects lead to sympathetic hyperactivity in stem cell and mouse models of Familial Dysautonomia


 Familial dysautonomia (FD) is a rare neurodevelopmental and neurodegenerative disorder that affects the sympathetic nervous system. Patients harbor a mutation in ELP1 yet, how loss of Elp1 affects the function of symNs remains unresolved. Such an understanding is critical since the most debilitating hallmarks of the disease include cardiovascular instability, dysautonomic crises and renal failure, which all result from dysregulated sympathetic activity. Here, we employ the human pluripotent stem cell (hPSC) technology as a modeling system to understand human, sympathetic neuron (symN)-specific disease mechanisms and provide a platform for drug testing and discovery. We show that FD symNs are intrinsically hyperactive in vitro, in co-cultures with cardiomyocyte target tissue and in FD animal models. We show that ELP1-rescued isogenic lines remain hyperactive, suggesting a different/additional disease mechanism. Accordingly, we report decreased intracellular norepinephrine (NE) levels, decreased NE re-uptake via NET and excessive extracellular NE in FD symNs. Finally, we performed a mini drug screen showing that current and new candidate drugs were able to lower hyperactivity. These findings may have implications for other peripheral nervous system disorders. Our drug screening platform may allow future drug testing and discovery for such disorders.


Introduction
Familial Dysautonomia (FD) is a complicated, early-onset, genetic disorder that mainly affects the development of the sensory and sympathetic nervous system (SNS) within the peripheral nervous system (PNS), leading to reduced numbers of these neurons. It also leads to neurodegeneration of these cells. 99.5% of all FD patients harbor a nonsense RNA splicing mutation in the gene ELP1 (formerly called IKBKAP 1 , which encodes the protein elongator complex protein 1 (elp1). Elp1 is the structural component of the elongator complex, which is directly involved in tRNA modi cation 2 and is essential for normal translation, and thus can be found in most cells of the body. While FD patients also suffer de cits in the central nervous system (CNS, particularly the eye), due to the critical role of the autonomic nervous system (ANS) in maintaining homeostasis, the altered function of the ANS is the most life-threatening system in FD. It was proposed that ELP1 defects mainly affect particularly long transcripts with an AA nucleotide bias and that many PNS critical genes fall into this category 3 . Interestingly, FD patients vary in the severity of their clinical phenotype, which raises the question of whether modi er genes exist. In fact, it's been shown that FD patients with the most severe symptoms harbor a modi er mutation in the gene LAMB4, encoding lamininb4 4 . FD patients suffer from a variety of symptoms, including di culties swallowing and regulating heart rate and blood pressure 5 . Dysautonomic/adrenergic crises are particularly disabling episodes of tachycardia, arterial hypertension, nausea/vomiting/retching and behavioral changes reminiscent of anxiety attacks 6 . Crisis can be frequent (daily), increase with age 7 , lead to hospitalization and contribute signi cantly to mortality 8 . Interestingly, crises are triggered by emotional or physical stress and associated with anxiety that can escalate into phobias 9 . Uncontrollable sympathetic neuron (symN) over-stimulation and associated prolonged catecholamine secretion into the blood stream caused by afferent barore ex failure has been suggested as the mechanisms mediating FD crisis 10 . This indeed explains the increase in blood pressure during crisis. Nevertheless, the fact that non-FD patients with barore ex failure mainly suffer from unstable blood pressure, but do not experience many of the other aspects of the FD dysautonomic crisis, suggests that the loss of central autonomic control and barore ex defects may not be the only factor that leads to the crisis. For example, vomiting/retching episodes, characteristic for FD dysautonomic crisis, do not occur in barore ex failure patients. In FD, it was suggested that vomiting/retching is a result of excessive dopamine (DA) in the blood stream 11 . Therefore, we hypothesize that intrinsic factors within symNs in FD contribute to triggering dysautonomic crisis.
Reduced symN numbers and overall sympathetic ganglia size reduction has been reported in FD patients 12 and in FD mouse models 13 . Such observations are based on both developmental defects 13 and degeneration 12,13 . While several studies in mouse neurons have revealed that neurons in FD have impairments in mitochondrial function 14 , elevated reactive oxygen species and cytoskeleton issues 15 ; that their proteome and transcriptome is affected 3 , and they have faulty NGF retrograde transport and signaling 16 , no study on neuronal activity has been reported, particularly not in human cells. Loss, hypoor hyperactivity of symNs leads to a variety of human disorders and intrinsic neural hyperactivity has been proposed as the cause of dysfunction and ensuing degeneration in several disorders, including amyotrophic lateral sclerosis (ALS) 17 and Alzheimer's disease 18 . Hyperactivity is further linked with aberrant calcium homeostasis 19 and norepinephrine (NE) levels 20 . Therefore, we aim to investigate symN-speci c phenotypes in FD, with the long-term goal to provide knowledge for drug development for FD as well as other symN disorders.
Assessments of symN function in the clinic are based on indirect measures, such as blood pressure, heart rate and blood glucose level 21 , and the lack of availability of primary sympathetic tissue for research makes it challenging to address neural intrinsic dysfunctions. Thus, it remains to be shown is mechanistic insights into FD gained from mouse studies are present in human cells as well. The human pluripotent stem cell (hPSC) technology is ideal to address such shortcomings, as it allows the generation of near unlimited numbers of human, patient-derived symNs and their observation and manipulation in the dish 22 .
Here, we employ our well-established symN differentiation protocol 23 to study functional phenotypes in FD symNs, using FD patient-and healthy control-derived hPSCs. We recapitulate developmental defects in the sympathetic lineage and reveal spontaneous, intrinsic hyperactivity of FD symNs. We identify a defective norepinephrine autoregulatory pathway that underlies the hyperactivity phenotype. Lastly, we show that this model system is sensitive in con rming clinical drugs and potential drug compounds that can relieve symN hyperactivity. Together, our results reveal symN hyperactivity as a novel pathology in FD, and we provide a novel, human drug testing and screening platform for symN-modulating compounds.

Developmental phenotypes in the sympathetic lineage in Familial Dysautonomia
To model symN defects in FD, we needed an e cient and reproducible differentiation protocol to derive pure symNs from hPSCs. Although there are several published symN differentiation strategies [23][24][25][26][27] , we rst further optimized our previous feeder-free, chemically de ned protocol (Sup. Fig. 1a). Neural crest cells (NCCs) at day 10 are derived at an e ciency of >90% 23 , thus a time-consuming ow cytometry (FACS) puri cation step can be omitted. SymNs from day 14 on express sympathoblast genes including PHOX2B, ASCL1 and HAND2 (Sup. Fig. 1b) and express HOX genes (HOX 5-9) indicating their trunk-like identity (Sup. Fig. 1c). They display typical PNS ganglia-like morphology (Fig. 1d). Mature marker genes are expressed from day 20 on, on the mRNA and protein level (Sup. Fig. 1e-h), including genes important in symN signal transduction CHRNA3, CHRNB4 and VMAT1, autoregulation α 2 /β 2 ARs and NET and NE synthesis/metabolism TH, DBH, AAAD and MOA-A. We also con rmed NE production in symNs (Sup. Fig. 1h). To improve technical ease of the protocol, we show that both human embryonic stem cell (hESC)-and iPSC-derived NCCs can be cryopreserved at day 10, followed by differentiation into symNs, without compromising the neurons quality (Sup. Fig. 1i, j).
Next, we sought to improve the purity of differentiated symNs to prevent possible variations due to contaminating non-neuronal cells during phenotype identi cation. Our original protocol yielded about 45% neurons, of which over 90% were TH + symNs (Sup. Fig. 2a, grey bars). To eliminate non-neuronal lineages, we treated day 20 symNs with aphidicolin (Aphi), a cell cycle inhibitor commonly used in primary peripheral neuron cultures 28,29 , for 10 days. After day 30, Aphi-treated symNs show signi cantly improved purity, while neural speci city is not affected (Sup. Fig. 2a, b, red bars). Proliferating Ki67 + cells and SOX10 + NCCs are dramatically diminished after Aphi treatment (Sup. Fig. 2c). Furthermore, we analyzed electrical activity of symNs after day 20 using multi-electrode array (MEA) (Sup. Fig. 2d). Neural spikes represent functionality and neural bursts represent neuronal maturity, which increases over time (Sup. Fig. 2e, f). Thus, this protocol yields high numbers of pure and functional symNs within a shortened time-period.
With this differentiation protocol in hand, we next aimed to assess phenotypes in the symN lineage in FD. We employed the following hPSC lines: iPSCs from FD patients (iPSC-FD-S2 and iPSC-FD-S3) and healthy control subjects (iPSC-ctrl-C1), and healthy ESCs (hESC-ctrl-H9, Sup. Fig. 3a). These lines were previously established, well characterized and employed in disease modeling 4 . We rst assessed disease phenotypes throughout the developmental stages of NCCs, sympathoblasts and symNs in FD (Fig. 1a).
NCCs can be identi ed in condensed dark "ridge" structures ( Fig. 1b, arrows), which are SOX10 + and correlate with CD49D 23,30 staining on day 10 ( Fig. 1b-e). In accordance with previous ndings 4 , control NCCs (iPSC-ctrl-C1 and hESC-ctrl-H9) show higher differentiation e ciency (~90%) than FD NCCs (iPSC-FD-S2 and iPSC-FD-S3, ~30%) evident by higher SOX10 + NCC ridge coverage (Fig. 1b, c) and higher CD49D expression (by FACS, Fig. 1d, e). Such a ratio of control to FD NCC reduction (90-30%) correlates with the reduction of sympathetic ganglia volume described in FD patients 12 and mice. From here on, we combined data for iPSC-ctrl-C1 and hESC-ctrl-H9 and call it ctrl, as well as iPSC-FD-S2 and iPSC-FD-S3 and call it FD as indicated in the gure legends. To answer the question, whether in FD the remaining 60% of cells die or differentiate into another cell type, we assessed overall cell numbers. We found that at day 10 there is no difference in cell numbers (Sup. Fig. 4a), which is corroborated by similar staining for AP2a (Sup. Fig. 4b), suggesting that FD iPSCs may differentiate into other cell types, likely non-neural ectoderm, rather than die. Using RT-qPCR analysis, we found signi cant expression of SIX1 + /EYA1 + placode contaminant (Sup. Fig. 4c). However, we previously showed that these non-neural cells do not aggregate into neural spheroids or survive 23 , and thus are lost upon further differentiation.
Next, we compared control and FD NCCs during sympathoblast speci cation in spheroids from day 10-14 ( Fig. 1f). Despite replating equal NCC numbers at day 10 in control and FD iPSCs, we observed smaller size as well as lower total cell numbers in FD compared to control spheroids by day 14 (Fig. 1f, g). We also noticed compromised integrity in FD spheroids, forming less compact, smooth and more irregularshaped aggregates (Fig. 1f). Additionally, we compared the survival rate between control and FD cells after dissociation at day 0, 10 and 14 and show that survival in FD drops at the sympathoblast stage, where speci c symNs are speci ed (Fig. 1h). Finally, we assessed symN differentiation upon dissociation and equal cell number replating of day 14 sympathoblasts. We found that, at day 30 both control and FD generate normal looking symNs with similar morphology and distribution (Fig. 1i). RT-qPCR analysis showed no signi cant differences of symN markers (Fig. 1j). SymN morphology was further examined using axonal marker GAP43, dendritic marker MAP2 and synaptic marker synaptophysin, and no difference was found (Fig. 1i, k). Together, our results suggest that in FD both the NCC and sympathoblast numbers are reduced, but the remaining progenitors are still capable of generating symNs. These ndings are supported by our previous work 4 and other's works 12,31,32 . Sympathetic neurons in Familial Dysautonomia are spontaneously hyperactive We next asked if the symNs that do develop in FD are functional. Utilizing MEA, we found that FD symNs are spontaneously hyperactive starting at day 25 (Fig. 2a). To investigate the possibility that the difference found at a single timepoint is because of varying differentiation progression/maturation of each iPSC line, MEA recording was performed in a time course from day 20-60. We found that compared to control, FD symNs are spontaneously hyperactive throughout their maturation and that up to day 60, control neurons never reach the spike frequency of FD (Fig. 2b), excluding the concern of differential maturation speed. Our nding of hyperactivity in FD symNs was further con rmed by increased expression of c-Fos, a gene that represents symN activity 33 (Fig. 2c), neuropeptide Y, a factor that will be upregulated following symN activation 34 (Fig. 2d) and corticotropin releasing hormone receptor 1 and 2 (CRHR1/2), the receptors for stress-induced CRH 35 (Fig. 2e), all highlighting increased neuronal activity and heightened stress responsiveness. Additionally, we con rmed FD symN hyperactivity by monitoring spontaneous calcium (Ca 2+ ) dynamics. Using the Ca 2+ probe Fluo-4, we observed that FD symNs have higher Ca 2+ in ux activity than healthy control symNs during recording 36 (Fig. 2f), re ecting the hyperactive state in FD symNs. In order to understand whether the hyperactivity phenotype is a PNSspeci c feature, we differentiated CNS cortical neurons from control and FD iPSCs using previously described protocols 37,38 (Fig. 2g) and assessed their electric activity in parallel. We did not detect signi cant differences either in morphology (Fig. 2g) or mean ring rate (Fig. 2h) at or before day 45. These observations support that symN hyperactivity is a critical PNS phenotype in FD.
We next assessed the effect of hyperactive symNs on regulating their target tissue in FD. We differentiated cardiomyocytes 39,40 (CMs) from healthy hESC-ctrl-H9 and/or iPSC-ctrl-C1 lines, that started beating at day 7 (Sup. Fig. 5a, Sup. Video 1), red cardiac action potentials from day 10 on (via MEA, Sup. Fig. 5b) and expressed speci c markers cTnT and NKX2.5 at day 15 (Sup. Fig. 5c). To mimic the SNS-cardiac axis, we created a co-culture consisting of day 7 CM precursors and dissociated day 14 symNs in MEA plates (Fig. 2i). Innervation of cardiomyocytes by symNs can be observed one week later (Sup. Fig. 5d). CM beating rate increased when CMs were cultured with conditional medium from symNs, as well as in the co-culture with symNs. It could further be augmented when symNs were stimulated with nicotine (Sup. Fig. 5e), indicating the functionality of the co-culture system. With this powerful tool, we assessed FD symN regulation of its cardiomyocyte target tissue. As shown previously 26 , the co-culture matured and thus increased the beating rate even in control; however, it signi cantly increased CM beating rate in the FD symN co-culture (Fig. 2j). Together, we show that symNs in FD are spontaneously hyperactive which leads to increased target tissue activation, potentially paralleling FD patients heart rate instability 41 .
Sympathetic neuron hyperactivity is conserved in Familial Dysautonomia mouse models Next, we sought to corroborate our ndings of symN hyperactivity in FD in vivo. Previous reports in various FD mouse models have shown symN loss at the embryonic stage, smaller sympathetic ganglion size and defective target innervation 13,15,42 . SymN activity, however has not been investigated previously to our knowledge. Here, we used two FD mouse models, the wnt1-cre; elp1 LoxP/LoxP conditional knock-out (CKO) model 13 and the sox10-cre; elp1 LoxP/LoxP CKO model 43 , both of which delete elp1 expression in the neural crest cell lineage (Sup. Fig. 6a). Most FD CKO mice, including ours, are embryonic lethal 15,42 , and the few surviving pups are of smaller stature compared to control (Fig. 3a). We aimed at focusing on symN activity and thus chose E14.5, the embryonic stage where the superior cervical ganglia (SCG) formed recently [44][45][46][47][48] and their size is not different between control and FD yet. Previous reports have shown that the gross loss of progenitor cell mass in sympathetic ganglia begins before E17.5 13 . Accordingly, we did not observe an embryo body size difference between FD and control (Fig. 3b, c). We dissected the SCG (Sup. Fig. 6b), dissociated the symNs and cultured them for 7 days, before measuring neural activity via MEA (Fig. 3d). The SCG were similar in size (Fig. 3e) and the symNs expressed the appropriate markers (Fig. 3f). SCGs from control and FD mice were plated evenly in MEA dishes (Fig. 3h, top) and mouse elp1 was indeed knocked-out in FD, as no elp1 protein was detectable in the FD cultures ( Fig. 3h, bottom). FD and control symNs showed similar neurite development (Fig. 3h). We then compared symN activity from control and CKO mice for one week, showing that FD symNs re spontaneously at a higher rate compared to WT neurons (Fig. 3i, j), consistent with our ndings in the iPSC-based symN model. Finally, we found that SCG neurons start to degenerate from day 7 in vitro, supporting the notion that hyperactivity is detrimental, as seen in other systems 17 (Sup. Fig. 6c). This data supports the nding that sympathetic neurons are spontaneously hyperactive in FD, both in hPSCderived symNs and as well as in symNs derived from two FD mouse models.

Defects in norepinephrine transporter (NET) underlie sympathetic neuron hyperactivity in Familial Dysautonomia
We next aimed at a better understanding of the molecular mechanism underlying symN hyperactivity in FD. 99.5% of all FD patients, including the patient's cells analyzed here 4 , harbor a homozygous mutation in ELP1. To examine whether hyperactivity is the direct consequence of the ELP1 mutation, we performed symN differentiations using the ELP1 rescued line, iPSC-EPL1 rescued -T6.5, in which the ELP1 mutation is heterozygously corrected by CRISPR-cas9 from the iPSC-FD-S2 line 4 . Surprisingly, MEA recording over time revealed that ELP1-rescued FD symNs are still spontaneously hyperactive (Fig. 4a, b), suggesting a mechanism that is indirectly dependent or independent of ELP1 explaining symN hyperactivity.
We hypothesize that hyperactivity in symNs may be caused by defects in the incoming signal (triggering/regulating action potentials) or at the outgoing signal (neurotransmitter release/re-uptake, (Fig. 4c). To investigate the rst question, we compared a selection of receptors and transporters that regulate neuronal electrophysiology and have been reported in symNs, including nicotinic receptors (CHRNs), sodium channels (Na v ), potassium channels (KCNs and Maxi-K), GABA receptor (GA A R) and calcium channels (P2RX and Ca v ). We assessed differential expression thereof and found no signi cant difference between control and FD (Fig. 4d). We further blocked the incoming signal with the sodium channel blocker tetrodotoxin (TTX) in both control and FD symNs. We found no change in hyperactivity (Fig. 4e), thus con rming that it is caused by a cell-intrinsic mechanism. This result also reduces the potential concern that autapses, which have been shown to form in ex-vivo cultured symNs 49 , lead to FD hyperactivity, since TTX blocks incoming signals from potential self-innervation/stimulation. We next assessed outgoing signals that could trigger hyperactivity. Expression of VMAT, important for NE vesicle transport, α2AR and β2ARs, involved with NE re-uptake were not expressed differently. However, NET, which reuptakes nearly 90% of secreted NE, and thus plays a critical role in extracellular NE clearance was expressed signi cantly lower in FD symN cell bodies and axons (Fig. 4f). NET de ciency has been reported in multiple neural disorders in which SNS dysregulation is involved 50,51 . We con rmed reduced NET protein in FD symNs by IF (Fig. 4g) and showed reduced uptake of uorescently labeled, synthetic NE by NET in FD symNs, using a NET reuptake assay (Fig. 4h, i). Less NETs on the surface of FD symNs is expected to lead to reduced NE levels inside the cell and excessive NE levels in the extracellular space in FD. Indeed, we showed diminished (though not signi cantly) NE levels inside FD symNs (Fig. 4j), using a novel NE tracker NS510, which is a turn-on probe that allows live cell imaging of intracellular NE dynamics 52 . Accordingly, ELISA measurements of NE in symN media revealed that in FD symNs more NE is present in the extracellular space (Fig. 4k). For the NET protein to be tra cked to the cell surface and thus become functional, it has to be phosphorylated and glycosylated 53 . Thus, we assessed those modi cations on NET in FD and control symNs. We did not detect a signi cant difference in glycosylation (Fig. 4l), indicating that NET function may not be affected beyond the lower expression levels.
Finally, to test if NET inhibition is su cient to induce the hyperactive phenotype, we treated healthy control symNs with nomifensine, a norepinephrine-dopamine reuptake inhibitor (NETi) for 24-72 hours to allow the accumulation of extracellular NE and recorded neural activities by MEA. In this analysis, control symNs gradually become hyperactive (Fig. 4m), indicating that indeed NET de ciency in FD symNs may underlie intrinsic hyperactivity. In sum, these results indicate that symNs in FD are intrinsically hyperactive due to defects in auto-regulation of their neurotransmitter (Fig. 4n), which leads to increased activation of their target tissue.
Mini drug screen reveals potential treatments of hyperactive FD symNs With these exciting ndings, we wanted to test whether this hPSC model system could be used as a platform to test drugs and therefore as a future drug discovery tool. To do so, we selected several drugs whose therapeutic potential have been studied at the clinical or experimental level in FD or other PNS diseases. We tested those drugs on day 35 symNs, the timepoint when hyperactivity in FD symNs rst peaks, yet neurodegeneration has not occurred. Dexmedetomidine (Dexmed) is a novel selective α2-AR enhancer, shown to relieve dysautonomic crisis symptoms in some FD patients 54 . Carbidopa is a selective AAAD inhibitor that blocks NE and DA synthesis; its effect of rescuing failed barore ex functions were shown in some FD patients 55 . We found that indeed, both drugs reduce the exaggerated ring activity of FD symNs in our model (Fig. 5a, b and Sup. Fig. 7a), supporting the notion that our hPSCbased model is useful for drug testing. We next sought to treat FD symNs by targeting the NET pathway. Clozapine is an atypical antipsychotic medicine, which has been demonstrated to elevate NET expression in chroma n cells 56 . Again, clozapine reduces FD symN hyperactivity signi cantly in our model (Fig. 5c and Sup. Fig. 7b). Lastly, propranolol, a β2AR inhibitor reduces FD hyperactivity as well (Fig. 5d), in fact when both control and FD neurons are treated with propranolol, FD's ring is signi cantly more reduced (Fig. 5e). Together, these results highlight the usefulness of our modeling platform to assess drugs for FD patients and together with our NETi result on control symNs (Fig. 4m), we show that manipulating NET function is an effective strategy for modulating symN activity.
Next, we aimed to test drugs that have been used in PNS-related neural disorders, but not in FD to our knowledge. Previous research in ALS, a neurodegenerative disease with motor neuron denervation, revealed motor neuron hyperactivity due to potassium channel malfunction and identi ed upiritine as a modulator 17 . We tested this drug on FD symNs, and indeed found that upiritine decreased FD hyperactivity (Fig. 5f). Finally, we examined NSC87877 and BGP-15, two drugs that have been studied to prevent PNS neuron neurodegeneration 14,16 . NSC87877 is a tyrosine phosphatase inhibitor shown to prevent FD symN death from impaired retrograde signaling in FD mice 16 . BGP-15 is a small hydroxylamine compound that improves mitochondrial function 14 . We found that treatment of FD symNs with NSC87877 or BGP-15 does not rescue hyperactivity (Fig. 5g, h and Sup. Fig. 7c, d), likely due to their mechanism of action that targets neuronal survival rather than activity. All drugs used in this study do not affect symN survival (Fig. 5i). Finally, we compared the responses of control and FD symNs after propranolol treatment at the same dosage and found that propranolol inhibits FD symN activity even more compared to its action in control neurons (Fig. 5e). This may imply that β 2 AR signaling in FD symNs is hypersensitized, further driving the loop of hyperactivity.
Our symN platform and FD model are instrumental for testing current drugs and thus promising for future drug discovery attempts. Furthermore, the results from these drug treatments further strengthen our conclusion about the molecular mechanism underlying symN hyperactivity. In healthy symNs (Fig. 6, top) NE is released into the extracellular space and about 10% of it is bound by the target tissue. Another 10% is taken up by α 2 AR on the neuron itself, which signals to inhibit further NE release and 10% is taken up by the β 2 AR, which signals to activate NE release. The nal 70% of NE is taken up via NET. FD symNs have less NET molecules available (Fig. 6, bottom), thus NE is depleted inside the cells and in excess outside the cells. That excess can bind to both α 2 -AR and β 2 AR.
Together, we show that our hPSC-derived symN platform is a powerful tool to select potential drug compounds for future treatment options of SNS hyperactivity in FD.

Discussion
Few in vitro differentiation protocols have been published to date for the generation of postganglionic symNs from hPSCs [24][25][26][27] . Our protocol has the following advantages, which makes it ideal for our FD disease modeling application: 1. It is highly e cient and produces a relatively pure symN population, i.e. >90% NCCs 23 and >75% neurons, of which >90% are peripheral neurons (Sup. Fig. 2). 2. It generates functional and mature symNs and is practical, as we have developed a cryopreservation option (Sup. Fig. 1i, j). 3. Our protocol recapitulates the proper developmental steps of symN development and was the rst protocol to be successfully employed for disease modeling in Zeltner et al. 4 , and here.
FD is a developmental disorder with symptom onset at birth 5 . Accordingly, we show reduced generation of NCCs from FD iPSCs (Fig. 1b-e) and reduced cell numbers and lower cell survival at the sympathoblast stage ( Fig. 1f-h). However, the symNs that passed through early development, do mature comparably to control neurons (Fig. 1i-k). These observations correlate with FD patient autopsies that showed reduced volume in the sympathetic ganglia 12 . FD animal data from conditional knock-out (CKO) and other mouse models 15 are slightly different, in that NCC development ensues normally 13,32 , but neuron numbers are decreased due to maldevelopment 13,32 , defective innervation 16,42 , reduced neuron survival 13,16 and abnormal NGF retrograde transport 16,57,58 . However, the end-result of overall reduced symNs is consistent with FD mouse models.
Using our in vitro disease model for FD, we found that FD symNs are intrinsically hyperactive. We con rmed this nding via electrophysiology, associated gene expression, Ca 2+ -imaging and in co-cultures with target tissue, where cardiomyocytes beat faster when coupled with FD symNs (Fig. 2). We showed that symN hyperactivity is conserved in two CKO FD (elp1 LoxP/LoxP ) mouse models, driven by either WNT1-cre or SOX10-cre (Fig. 3, Sup. Fig. 6). Neural hyperactivity has been shown in a variety of neurological disorders and is associated with neurodegeneration. Clifford Woolf's group has reported the causality of motor neuron hyperactivity and neurodegeneration in ALS using patient-iPSC model 17 . It has been suggested that network activity may lead to neurodegeneration in the brain of Alzheimer's disease (AD) 59,60 , and an hPSC-based AD model that recapitulates neural hyperactivity was proposed recently by Stuart Lipton's group 18 . We indeed found that symNs from FD animals, that are hyperactive, show signs of degeneration quickly in culture even in the presence of NGF (60ng/ml) (Sup. Fig. 6c). Future investigations will show if pharmacological prevention of hyperactivity may prevent neurodegeneration in FD.
The SNS is an important component of the stress response system and is crucial for regulation of body homeostasis. FD patients have di culties regulating arousal, and stress can trigger dysautnomic crisis 6 .
Could it be possible then that in FD symNs are more susceptible to stress signaling from the brain? During a stressful state, the extrahypothalamic areas and hypothalamus in the brain release corticotropin releasing factor (CRF), a stress-related neuropeptide, which has been found to regulate the SNS as well as symN activity 35 . We did nd elevated expression of CRF receptors (CRHR1/2) in FD symNs. NPY is another important neuropeptide involved in stress modulation both in the CNS and PNS. In PNS, NPY is mainly produced by symNs, its levels have been positively correlated to symN activity, and NPY released by symNs has a vasoconstriction effect 34,61 . Elevated plasma NPY levels are also found in stress-related disorders, such as hypertension and heart failure 34 . Indeed, we detected elevated NPY levels in FD symNs, supporting the hyperactivity phenotype in FD symNs, but also suggesting that FD SNS may be more vulnerable to stress stimulation.
We hypothesize that intrinsic defects within FD symNs are the drivers that lead to hyperactivity, which might further trigger other FD phenotypes and symptoms, including dysautonomic crisis and neurodegeneration. Literature from other SNS related disorders supports this notion. For instance, symN neurotransmitter switching from NE to epinephrine in prehypertensive rats 62 indicates that symN abnormalities prior to the onset of cardiovascular symptoms may sensitize the SNS and make it more vulnerable to disease-related stimuli. Another study showed that hypertensive symNs were able to change the metabolism of healthy cardiomyocytes to induce a hypertensive state. Vice versa, when hypertensive cardiomyocytes were co-cultured with healthy symNs, the defective response was suppressed 63 .
Thus, we looked deeper into the molecular mechanism underlying FD symN hyperactivity. We showed that NET expression is reduced in FD, which leads to an accumulation of NE in the extracellular space and to diminished NE inside the cells compared to control (Fig. 4). NET de ciency has been linked to several SNS-related cardiovascular syndromes, such as orthostatic intolerance (OI), postural tachycardia syndrome (POTS) 50,64 , and stress-induced cardiomyopathy 51 , both in patients and animal models 51,65 . Furthermore, a study in patients with hypertension revealed that NE reuptake function was impaired, while the barore ex control remained unaffected 66 . We also show that NET inhibition is su cient to trigger symN hyperactivity in healthy hPSCs (Fig. 4m). We further found that NET protein in FD is functional, i.e. glycosylation is not affected (Fig. 4l), thus the issue is expression of the protein itself. As a consequence, in FD excessive NE accumulates in the extracellular space. This, extra NE is bound by a 2 AR, which inhibits more NE release and b 2 AR, which further activates more NE release 62,67−69 . Theoretically, these effects might cancel each other. However, we found that the reduction in hyperactivity after treatment with propranolol, a b 2 AR inhibitor is more pronounced in FD compared to propranolol treated control symNs (Fig. 5e). Thus, in FD the b 2 AR might be hypersensitized, which might maintain a feed forward loop, which drives hyperactivity in FD symNs. So, in a FD patient, early embryonic hyperactivity of symNs may explain later degeneration of the cells as well as hypersensitization of b 2 AR receptors, that together might be the important players for dysautonomic crisis. In our in vitro system, it remains to be shown when measurable degeneration signs begin and if they can be pharmacologically countered, which is a future goal of our work. In patients, indirect measurements of symN activity at rest, i.e. not in crisis, have not shown symN hyperactivity 31 . This discrepancy might have several explanations: (i) Such measurements are indirect, including skin conductance or heart rate assessments 21 as it is not feasible to directly measure symNs in a patient. Thus, at rest symN hyperactivity might not be strong enough to translate to the downstream target. (ii) In our in vitro, miniature model system, symNs are isolated and thus the released NE is bathing the entire neuron, which provides more opportunity to activate its ring. In comparison, in vivo the excess NE would mainly affect the distal axonal region and thus require more activation, such as during stress-triggered crisis to start hyperactivity in a crisis. (iii) In vivo, in a whole organism, the counteracting parasympathetic nervous system is operating as well.
99.5% of all FD patients, including the FD-iPSCs assessed here carry the identical ELP1 mutation 31 . It has been shown that elp1 de ciency preferentially reduces translation of long and AA-rich mRNAs 3 and increases translation of small AG-biased mRNAs. In a proteomic analysis of the dorsal root ganglia from the Wnt1-cre;Elp1 CKO mouse used here, due to the loss of elongator function, several thousand proteins were expressed at altered levels. Altered pathways included fundamental pathways such as exocytosis, protein transport and chromatin silencing and included changes to histones and transcription factors 3,70 . Our results however also showed that genetic rescue of the FD ELP1 mutation does not reverse hyperactivity (Fig. 4a, b), nor the reduction of NET in symNs (Fig. 4g). One explanation for that observation could be the downstream consequence of altered developmentally-timed genetic programs due to the loss of Elp1 expression and hence they cannot be re-set once the nervous system is mature. Another explanation might be that an additional mechanism is at play. We have shown previously that a modi er mutation in LAMB4 is present in the patients from whom the here assessed iPSCs are derived. Thus, while at the moment unclear how, it is possible that the interplay of ELP1, NET and LAMB4 might lead to the maintenance of symN hyperactivity in the ELP1-rescued line 71,72 .
Finally, we aimed at testing if our symN disease modeling platform is useful for drug screening and drug testing. We found that in this platform, drugs currently given to FD patients for dysautonomic crisis, i.e. dexmedetomidine 54 and carbidopa 55 were able to reduce hyperactivity. Additionally, clozapine and upiritine also could reduce hyperactivity. In contrast, NSC87877 and BGP-15 do not affect FD symN activity in our model. This might be due to the short-term treatment used here and the fact that these drugs have been shown to prevent neurodegeneration. In the future, we aim to establish neurodegeneration in our system and to test if these drugs may prevent it here as well. Together, our drug testing results support our model of the mechanism of defective NE tra cking and NET reuptake, through which hyperactivity in FD is generated and maintained (see model in Sup. Fig. 8). These results also indicate that this platform will be useful for drug screening approaches to identify and test novel compounds that could treat FD or other SNS-related disorders. can be added to the symN medium from day 20-30.
Note: BMP4 quality is highly lot dependent. It is highly recommended to perform titration test and decide the best BMP4 concentration for each batch of BMP4.
On day 3, feed with day 3-5 medium that contains Essential 6 medium, 10 μM SB and 200nM LDN193189. On day 5-9 feed with medium that contains Essential 6 medium and 10 μM SB. On day 10, medium was changed to day 10-20 medium that contains Neurobasal medium, B27 (1:1000) and GlutaMAX Cardiomyocyte differentiation. The differentiation protocol is modi ed from previous publications 39,40 . hPSCs on day -2 were dissociated shortly by EDTA and split at a 1:5 ratio on

NE ELISA
The assay was performed according to manufacturer's instructions (EagleBio, NOU39-K01). Media in two wells of a 24-well of day 35 symNs from each experimental group was changed (500 μl/well) one day before the experiment. 24 hrs later, media from two wells were harvested and pooled together to concentrate the NE. To avoid NE degradation, sample stabilizer included in the kit was added to each sample. Collected media was spun at 300 g for 5 min to remove debris. The samples were ready for NE detection or were stored at -80 °C for long-term storage.     Reduced NE reuptake caused by NET de ciency leads to FD symN hyperactivity. (a) MEA measurements on symNs derived from iPSC-rescued-T6.5 still shows hyperactivity compared to hESC-ctl-H9 (same data Model. In healthy postganglionic sympathetic neurons (symNs) (a), norepinephrine (NE) secretion in the axon terminal is properly regulated by norepinephrine transporter (NET), which reuptakes about 80% of released NE, α adrenergic receptors (αAR), which downregulate NE secretion and β adrenergic receptors (βAR), which upregulate NE secretion. In FD symNs (b), NET expression is decreased, leading to oversecreted NE. In addition, our results indicate that βAR in FD symNs might have stronger response than healthy symNs, which might further strengthen the release of NE and symN activity. Several therapeutic targets are identi ed in this study using a selection of clinical compounds: upirtine activates potassium channels (K channel) and induces neural depolarization; carbidopa blocks NE synthetic pathway and thus reduces NE overspill; dexmedetomidine (Dexmed) activates αAR while propranolol inhibits βAR. All these drugs suppress FD symN hyperactivity in vitro.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.