The difficulty to model Huntington’s disease in vitro using striatal medium spiny neurons differentiated from human induced pluripotent stem cells

Huntington’s disease (HD) is an autosomal dominant neurodegenerative disorder caused by an expanded polyglutamine repeat in the huntingtin gene. The neuropathology of HD is characterized by the decline of a specific neuronal population within the brain, the striatal medium spiny neurons (MSNs). The origins of this extreme vulnerability remain unknown. Human induced pluripotent stem cell (hiPS cell)-derived MSNs represent a powerful tool to study this genetic disease. However, the differentiation protocols published so far show a high heterogeneity of neuronal populations in vitro. Here, we compared two previously published protocols to obtain hiPS cell-derived striatal neurons from both healthy donors and HD patients. Patch-clamp experiments, immunostaining and RT-qPCR were performed to characterize the neurons in culture. While the neurons were mature enough to fire action potentials, a majority failed to express markers typical for MSNs. Voltage-clamp experiments on voltage-gated sodium (Nav) channels revealed a large variability between the two differentiation protocols. Action potential analysis did not reveal changes induced by the HD mutation. This study attempts to demonstrate the current challenges in reproducing data of previously published differentiation protocols and in generating hiPS cell-derived striatal MSNs to model a genetic neurodegenerative disorder in vitro.

and there is no treatment to slow the progression of HD, which is always lethal within 15 to 20 years following symptom onset 4 .
Several transgenic and knock-in rodent models have been generated to investigate HD pathological features 5 . They revealed some important mechanisms such as a decreased dendritic spine density 6 or a down-regulation of the voltage-gated sodium channel (Nav) β4 auxiliary subunit 7 . The Navβ4 subunit is highly expressed in the striatum and in striatal projection fibers. It modulates sodium channel activity and regulates neurite outgrowth 8 . Its expression level is significantly reduced in post mortem tissue of HD patients and in three rodent models of HD 7,9,10 . A Navβ4 down regulation in WT mice was shown to affect neurite outgrowth and to decrease repetitive firing frequency in mouse MSNs 7,11 . However, significant species differences between rodent and human cells limit the use of HD rodent models to accurately represent the disease and to predict the electrical activity in human striatal MSNs.
As an alternative, patient-derived human induced pluripotent stem (hiPS) cells have emerged as a powerful tool to decipher mechanisms underlying MSN degeneration and to investigate their firing properties 12 . The reprogramming of skin fibroblasts or mesenchymal stromal cells (MSCs) into human iPS cells allows the generation and the in vitro study of human neurons carrying the huntingtin mutation 4,13 . hiPS cells retain their genetic background and can generate striatal neurons. These essential properties gave rise to the establishment of many hiPS cell-derived MSN differentiation protocols.
To investigate the reliability and reproducibility of these already-existing protocols, we selected and established two previously published protocols 14,15 to generate striatal MSNs and to investigate their firing activity. We used hiPS cell lines from non-HD donors as well as from different juvenile-onset HD patients and the differentiated neurons were functionally characterized using electrophysiology, immunostaining and quantitative PCR. Our findings highlight the challenges that are innate to the study of hiPS cell-derived MSNs and the difficulties in interpreting data derived from a single differentiation protocol.

Methods
Subjects. All patients involved in this study signed an informed consent and the study was approved by the local ethics committees as detailed below (see also Supplementary Table S1). All experiments were conducted in accordance with the Declaration of Helsinki and approved by the local ethics committee of the RWTH Aachen University, Germany.
Healthy Caucasian non-HD subject fibroblasts from Ctrl1 and Ctrl3 were obtained from the healthy donors under their consent 16 in accordance with the EK 206_09 ethical application approved by local ethics committee of the RWTH Aachen University, Germany 17 .
HD72 human iPS cell line (Cat. Nr. GM23225) was obtained from Coriell Institute for Medical Research (New Jersey, USA) under their consent and privacy guidelines (https:// www. corie ll. org/).
Healthy human iPS cell Ctrl2 and the HD iPS cell HD109 and HD180 lines were provided by the Cedars-Sinai Medical Center (Los Angeles, California) and informed consent was obtained in accordance with the local ethics committee at the Cedars-Sinai Medical Center (Los Angeles, California, IRB/SCRO protocols Pro00021505 and Pro00024899).
One clone per cell line (six clones in total), corresponding to the biological replicate, with a passage number ranging from 18 to 41, was investigated in this study. One to three technical replicates (the number of differentiations) were performed for each clone (Supplementary Table S4). To be able to perform thorough patch-clamp experiments for all clones and conditions, the differentiations needed to be started at slightly different time points. Table S2). Human iPS cells were maintained at 37 °C with 95% O 2 -5% CO 2 . They were cultivated either on 0.1% Matrigel hESC-qualified matrix (Becton Dicksinson, 354277) or on 0.1% Geltrex LDEV-Free Reduced Growth Factor Basement Membrane Matrix (Life Technologies, A14133-02). They were supplied daily with either mTeSR 1 supplemented with 5 × mTeSR 1 Supplement or E8 medium supplemented with TeSR-E8 Basal Medium-E8 Supplement (both Stemcell Technologies, Cologne, Germany).

Generation and maintenance of hiPS cells. All cell lines were obtained as hiPS cells and reprogramming of donor cells into hiPS cells was not part of this study (Supplementary
HD mouse line BACHD. Both heterozygous C57Bl/6 BACHD and homozygous WT C57Bl/6 mice of either sex were used in this study. All methods were carried out in accordance with relevant guidelines and regulations. All experiments were performed according to the approval number 54-2532.1-49/12 by local ethical boards of the District Government of Middle Franconia, Bavaria, Germany. Differentiation into MSNs according to the Stanslowsky protocol. The Stanslowsky protocol 14 was used to differentiate Ctrl1, Ctrl2, Ctrl3, HD72, HD109 and HD180 hiPS cell lines (Fig. 1).
Differentiation into MSNs according to the Fjodorova protocol 15 . The Fjodorova protocol 15 was used to differentiate Ctrl1, Ctrl2, HD72, HD109 and HD180 hiPS cell lines (Fig. 2). hiPS cells were cultured daily with 1 mL/well mTeSR1 or E8 medium until they formed colonies 80% to 90% confluent (Fig. 2B). Cells were split using 0.5 mM EDTA and transferred into wells coated with Geltrex. They were cultivated for 10 days in N2B27 medium (64% DMEM/F-12 + GlutaMAX-1 (Life Technologies), 30% Neurobasal medium (Life technologies), 1% Penicillin/Streptomycin, 1:150 N2 supplement, 1:150 B27 supplement without vitamin A and 0.1% 50 mM β-ME) supplemented with 10 µM SB, 100 nM LDN-193189 (Miltenyi Biotec), 200 nM DM and 10 µM ROCK. Half the medium was replaced every two days. Between DIV5 and DIV9, N2B27 medium was supplemented with 100 nM LDN-193189 (Miltenyi Biotech) and 200 nM DM (Fig. 2C). On DIV9, cells were incubated one hour with N2B27 medium supplemented with 10 µM ROCK and 25 ng/mL Recombinant Human Activin A (Biolegend, San Diego, United States) before being split with 0.02% 0.5 M EDTA and transferred at a 1:10 ratio into wells coated with Geltrex. The N2B27 medium with Activin A was changed every other day until DIV21 (Fig. 2D). At DIV21, cells were split again and 10,000 to 20,000 cells were distributed onto each cover slip of a 24 well plate coated overnight with Geltrex. Cells were maintained in N2B27 medium supplemented with 10 ng/mL BDNF, 10 ng/mL GDNF (all PeproTech) and 100 nM retinoic acid (Sigma-Aldrich). The medium was replaced every four days until DIV35 ± 3 (Fig. 2E). The maturation time was not extended compared to the original publication of this protocol 15 (see Results).
Huntingtin gene sequencing. Genomic DNA of hiPS cells was isolated using the NucleoSpin Tissue Kit (Macherey-Nagel, Düren, Germany) according to the manufacturer's instructions. The primers used to identify the number of CAG repeats within the huntingtin gene are available upon request. The number of CAG repeats is indicated in Supplementary Table S3. RNA extraction and RT-qPCR. RNA was isolated with the NucleoSpin RNA Kit (Macherey-Nagel) according to the manufacturer's instructions. RNA was reverse transcribed into cDNA using SensiFAST cDNA Synthesis Kit (Bioline, Luckenwalde, Germany). Quantitative PCR was performed on a RotorGeneQ Real-Time cycler (Qiagen, Hilden, Germany) with SensiMix SYBR No-ROX Kit (Bioline). Data are shown as the relative DM, SB and IWP2 induce the neural pathway driving the formation of embryoid bodies. IWP2 associated with PMA induces the regional patterning towards the subpallium that comprises large parts of the basal ganglia including striatum and globus pallidus. Y27632 inhibits apoptosis to prevent cell decline at the beginning of the differentiation. The neuronal growth factors BDNF, GDNF and TGF-β3 terminate the differentiation into MSNs and drive their maturation, while dbcAMP increases the intracellular Ca 2+ levels and contributes to the maturation process. (B) hiPS cells grow in basal medium until DIV0. DIV0 to DIV12-14 initiates the formation of embryoid bodies. (C) From DIV12-14 to DIV55 ± 3, the neurons are exposed to growth factors to induce neurite outgrowth (D) and to establish neuronal networks (E). Immunofluorescence imaging and quantification. At the end of the differentiation process, the neurons were fixed with 4% paraformaldehyde and blocked and permeabilized with 0.2% bovine serum albumin, 0.1% Triton X-100 in PBS and 3% normal goat serum (all Sigma-Aldrich). Neurons were stained with mouse anti-βIII tubulin (or clone Tuj1, MAB1195, Abcam, R&D Systems, 1:400), rabbit anti-βIII tubulin (Sigma-Aldrich, 1:1000), mouse anti-GAD67 (clone 1G10.2, Merck Millipore, Darmstadt, Germany, 1:200), rabbit anti-DARPP32 (Abcam, 1:100), rabbit anti-SCN4B (Abcam, 1:500), rabbit anti-DRD1 (Abcam, 1:200), rabbit anti-TH (ab152Merck Millipore, Darmstadt, Germany, 1:500), mouse anti-peripherin (Santa Cruz, Heidelberg, Germany, 1:100). The presence of striatal interneurons was investigated using the following primary antibodies: rabbit anti-neuropeptide Y (Immunostar 22,940, Wisconsin, United States, 1:500), rat anti-somatostatin (Merck Millipore MAB354, 1:500), rabbit anti-somatostatin (T4103Dianova, Hamburg, Germany,, 1:500), mouse anticalretinin (Swant 6B3, Switzerland, 1:500), rabbit anti-neuropeptide VIP (Immunostar 20077, 1:500), rabbit anti-parvalbumin (Swant, 1:1,000). The secondary antibodies used were goat anti-mouse IgG Alexa Fluor 488, goat anti-rabbit IgG Alexa Fluor 594, goat anti-rabbit IgG Alexa Fluor 488 (both Life technologies, 1:1,000), goat anti-rabbit Cy3 (Jackson, Cambridge, United Kingdom, 1:2,000), goat anti-rat Cy3 (Invitrogen, Schwerte, Germany, 1:2,000), goat anti-mouse Cy3 (Jackson, 1:2,000). Nuclei were counterstained with DAPI (Thermo Fisher Scientific). Three to five random regions of interest (ROI) were investigated for each cover slip using an LSM700 confocal microscope (Carl Zeiss, Oberkochen, Germany) with 40 × oil immersion objective or a DMi8 inverted microscope with instant computational clearing (Leica, Germany) with 20 × objective. Counting was performed manually using the ImageJ-win64 software and the observer was blinded to the cell type and to the protocols. To detect nucleus staining in dense culture condition, the contrast brightness/background was increased. We counted as neurons only cells with a nucleus fully embedded by a fluorescence signal that suggested a close interaction with the cell membrane. Results of ROI counting were averaged and presented as mean.
Electrophysiology. Whole-cell patch-clamp experiments were performed with an EPC-10 USB amplifier (HEKA Elektronik, Lambrecht, Germany) at room temperature (20-22 °C). Glass pipettes of 2-4 MΩ (Biomedical Instruments) were manufactured with a DMZ puller (Zeitz Instruments GmbH, Martinsried, Germany). Sampling rate was set to 50 kHz for the recordings while using a 10 kHz low-pass filter. Series resistance (< 5.5MΩ for voltage-clamp mode) was compensated by at least 70%. Leak currents were subtracted online using the P/4 procedure. The liquid junction potential was not corrected.  For all voltage-clamp protocols, the holding potential was set to − 120 mV. Voltage dependence of activation was assessed from holding potential using 40 ms pulses to a range of test potentials from − 100 mV to + 60 mV in 10 mV increment (test-pulse) with an interval of 5 s. To isolate somatic currents and avoid space clamp artefacts, a voltage pre-pulse protocol was used to inactivate distant sodium channels 16,20 (Fig. 6A). The pre-pulse (− 60 mV to − 20 mV for 4 or 5 ms) was followed by a repolarizing inter-pulse (− 120 to − 70 mV for 1 ms) which preceded the regular test-pulse. In few cells no pre-pulse was applied due to small current amplitude and because sodium currents were well clamped without pre-pulse. Pre-pulse and inter-pulse voltage and duration were adjusted to each cell individually to obtain optimal voltage-clamp conditions. Subsequently, 500 nM TTX was applied in a limited number of cells to measure the TTX-sensitive current. Conductance (G) was calculated at each voltage (V) using the equation G = I V−Vrev with V rev being the reversal potential and I being the inward current at the respective voltage. Conductance-voltage curves were fit using a Boltzmann equation where G min and G max are the minimum and maximum of sodium conductance respectively, V half is the potential of half-maximal channel activation and k is the slope factor.
Steady-state fast inactivation was measured with or without 500 nM intracellular Ca 2+ concentration using a two-step protocol. A 500 ms pre-pulse with potentials ranging from − 120 mV to + 10 mV in 10 mV incremental steps was used to inactivate Nav channels. This pre-pulse was immediately followed by a 40 ms test-pulse to 0 mV, allowing the determination of the remaining fraction of available channels. Relative current was calculated as test-pulse current at each voltage (Vm) divided by the maximum test-pulse current and plotted against the pre-pulse voltage. Relative currents were fitted with the above Boltzmann equation.
The resting membrane potential (RMP) was measured immediately after establishing the whole-cell configuration. Holding current was then adjusted to achieve a membrane voltage of − 70 mV ± 10 mV. Action potentials (APs) were induced by incremental current injections (from 1 to 50 pA in 200 ms) to determine the rheobase, i.e. the amount of current required to trigger one AP. A subsequent protocol with 0.5 × rheobase current injection was made during 200 ms to generate more APs (up to 4.5 × rheobase stimulus). APs were only considered if they exceeded 0 mV. The first AP evoked by the square pulse protocol was used to calculate AP properties. The AP threshold was defined as the minimum of the first derivative of the AP (i.e. the point of inflection during the depolarization). The amplitude was measured between holding voltage and AP peak. The afterhyperpolarization is the minimum voltage following the AP peak. The time-to-peak is the duration between current pulse onset and AP peak. Current-clamp recordings were performed at DIV55 ± 3 (Stanslowsky protocol) and at DIV35 ± 3 (Fjodorova protocol). Neurons that fired two or more APs to a given stimulus were considered tonically firing in our experiments. Neurons firing a single AP were considered as phasic firing.
Statistical analysis. Data were analyzed using GraphPad Prism version 5 or 6 (GraphPad Software, Inc) and SPSS (IBM SPSS Statistics Version 25). Two groups were compared using a 2-tailed Student's t test or a Mann-Whitney test, depending on normal distribution. Comparison between 3 or more groups was performed using a 1-way ANOVA followed by a Bonferroni, a Tukey's multiple comparison test or a Kruskall-Wallis test. The exact value of n (number of cells) is indicated in the Fig. legends or in the Tables. Data are presented as mean and error bars denote 95% confidence interval (CI). Outliers were identified and addressed.

Results
A longer time in culture increases the percentage of active neurons in the Stanslowsky protocol. To compare some of the previously published protocols for MSN differentiation from hiPS cells, we selected two well established protocols 14,15 in order to generate striatal neurons and to characterize their neuronal identity and functional activity. For the latter, it is important to work with neurons that express sufficient amounts of voltage-gated ion channels to be electrically mature. To address this question and prior to any experiment, we aimed to determine the differentiation time in vitro required to reach this functional maturation. Using the Stanslowsky protocol, we found in two control cell lines (Ctrl1 and Ctrl3) and in one HD cell line (HD72) a higher percentage of electrically active neurons (i.e. = neurons firing at least one action potential) at DIV55 than at DIV40 (50%, 53.8% and 58.8% for Ctrl1, Ctrl3 and HD72 respectively at DIV40 vs 65.2%, 71.8% and 70.4% respectively at DIV55) (Fig. 3A). However, this observation was not verified for Ctrl2, with a higher percentage of electrically active neurons at DIV40 (66.7%) compared to DIV55 (53.1%). In addition, the Ctrl1, Ctrl3 and HD72 neurons showed a tendency to fire a higher number of APs following two additional weeks in culture ( Both differentiation protocols produce low amount of GABAergic positive neurons and MSNs. Immunostaining was performed on hiPS cell-derived neurons of both the Stanslowsky protocol at DIV55 and Fjodorova protocol at DIV35 14,15 to characterize their neuronal identity (Fig. 4). We used the neuronal marker Tuj1 to distinguish neuronal and non-neuronal cells (Fig. 4A Fig. S1). We are confident that those Tuj1 + cells are central neurons.
Striatal MSNs in vivo are all GABAergic 3 and express the specific marker DARPP32 19 . These neurons provided a large population of GABAergic neurons, with a range of 75% to 81% Tuj1 + neurons expressing the enzyme GAD67 (precursor of GABA) in the two cell lines Ctrl2 and HD72 investigated in both protocols (Fig. 4A,B and Supplementary Table S6). Except for Ctrl1 neurons in the Fjodorova protocol and HD180 neurons in both protocols, all hiPS cell-derived neurons expressed GAD67 (Fig. 4A,B,E). In contrast, very few GABAergic neurons   Table S6). In summary, both protocols seem to produce only very few striatal MSNs.
As MSNs seemed to be the minority of the cells derived by the differentiation protocols in our hands, we investigated which other neuronal populations were present in our preparations. We stained the neurons with DRD1, a dopaminergic marker expressed by MSNs and other neuronal populations in the brain 20 . 50% to 77% of the Tuj1 + neurons investigated expressed DRD1 (Supplementary Fig. S1). We also investigated the expression of tyrosine hydroxylase (TH), the rate-limiting enzyme in the synthesis of catecholamines, among which dopamine. Previous findings revealed that TH is not only found in dopaminergic neurons, but also in striatal interneurons of humans 21,22 and rodents 23,24 , although not accompanied by dopamine release 24 . In the striatum, these interneurons make GABAergic synapses onto MSNs 24 . We confirmed first the expression of TH in two Tuj1 + cell lines where it was investigated: Ctrl2 and HD109 neurons (Fig. 4C, top and Supplementary Table S6). In addition, we revealed the presence of striatal interneurons in our culture, confirmed by a positive staining for somatostatin and calretinin ( Supplementary Fig. S1D-F, Table S6). We also tested for the interneuron markers parvalbumin, neuropeptide Y and neuropeptide VIP, but could not identify any stained cells ( Supplementary  Fig. S1D-F, Table S6). Finally, we investigated whether some of our TH + neurons may also be striatal interneurons. Three out of four striatal interneuron populations are GABAergic 24 , therefore we stained our neurons with TH and GAD67. The two clones Ctrl2 and HD72 revealed a range of 28.5% to 81.5% of TH/GAD67 co-expression (Fig. 4C, Supplementary Table S6).
All together, these findings suggest that the GABAergic neurons in our cultures are characterized by higher striatal interneuron populations than MSNs. Thus, both protocols designed to result in differentiation of MSNs mainly produced central neurons with a mixed identity. Importantly, we found a range of 4.5% to 12.4% of GABAergic calretinin interneurons while a considerable fraction of differentiated neurons are not striatal MSNs,  The Navβ4 subunit is very little expressed in the differentiated neurons of both genotypes. Nav channels are crucial for neuronal function and AP generation. In search for a biomarker of MSN degeneration, the Navβ4 subunit has been reported to be down-regulated in neurons of HD patients and rodent models 7,9 . Therefore, we performed RT-qPCR to investigate expression of Nav α and β subunits in our hiPS cellderived neurons (Fig. 5). In general, almost every cell line revealed a higher expression level of all Nav α and β subunits in the neurons differentiated using the Stanslowsky protocol compared to those differentiated using the Fjodorova protocol. We found that all hiPS cell-derived MSN lines expressed Nav1.1, Nav1.2, Nav1.3, Nav1.6 and Nav1.7 as well as the Navβ3 subunit, although expression varied between cell lines and genotype (Fig. 5), probably due to the various number of differentiations performed for each cell line (n = 1-3 differentiations). Navβ1, Navβ2 and Navβ4 are expressed to a lower extent compared to Navβ3 (Fig. 5B,D). These results are in accordance with data from the original Stanslowsky protocol 14 . With regards to the expression of Navα subunits, we found a similar expression pattern in WT mouse striatum (Supplementary Fig. S2). However, we found considerably higher amounts of β-subunit mRNA expression in mouse tissue compared to our human iPS cell-derived neurons. This is interesting in the context that an HD-related down-regulation of the Navβ4 subunit has mostly been based on rodent tissue 7,9 . In fact, we confirmed the Navβ4 subunit down-regulation in BACHD mice, one of the most physiological mouse models 5,6 , using in situ hybridization and RT-qPCR ( Supplementary Fig. S2). It is not clear, however, whether this down-regulation also affects human MSNs of HD patients. We were therefore particularly interested in the expression of the Navβ4 subunit in our iPS cell-derived neurons. Importantly, expression of Navβ4 mRNA was found to be very low in all clones, regardless of genotype or differentiation protocol (Fig. 5A-E). These findings question the relevance of the Navβ4 subunit as a biomarker for HD.
Sodium channel gating is reversely affected by HD genotype in the two differentiation paradigms. Whole-cell voltage-clamp recordings were performed in differentiated neurons to assess whether two different protocols may influence Nav gating in control and HD neurons. A pre-pulse voltage protocol was used to measure whole-cell current in control and HD neurons (Fig. 6A). This protocol allows for good voltage clamp even in cells with long neurites and larger currents 16,25 . 500 nM TTX was applied in a limited number of cells to test whether the TTX-resistant cardiac Nav1.5 channel or the peripheral sensory nerve isoforms Nav1.8 and Nav1.9 are expressed in the differentiated neurons. No TTX-resistant current could be measured in the hiPS  Table S8). Concerning the voltage dependence of activation, a large variability in the half maximal voltage was observed between the neurons of each protocol, and no significant difference was found. This suggests that the protocols do not exert any obvious effect on the activation of the Navs expressed in the differentiated neurons (Fig. 6C-D and Table S8). We found a more depolarized half maximal voltage in all four differentiated neurons from the  Table S8).
The plots displaying the differences between V half values of activation (Fig. 6F) and V half of fast inactivation (Fig. 6G) sum up the discrepancy within the two protocols observed in the previous panels of Fig. 6, while they do not suggest any obvious effect of the HD genotype on the Nav function of the neurons differentiated in both protocols (Fig. 6H).
All together, these results again highlight the considerable variability induced by the differentiation paradigm. Surprisingly, the largest effects on Nav channel gating do not come from the genotype but from the protocols used to generate hiPS cell-derived striatal neurons. This suggests the presence of neurons with various functions and point out a striking lack of homogeneity within the differentiated neurons derived from the two differentiation protocols. Finally, we did not observe a clear effect of the HD mutation on Nav channel activation and fast inactivation.  Table S8  www.nature.com/scientificreports/ Action potential features are unaffected by HD genotype. We previously showed a variability between protocols concerning the fast inactivation (Fig. 6E,F). These significant effects may lead to an alteration of AP features in hiPS cell-derived neurons. We therefore performed current-clamp experiments to determine whether the differentiation protocol may have an effect on their electrical activity (Fig. 7A). The resting membrane potential (RMP) was around − 40 mV (Fig. 7B), which is within ± 5 mV to what has previously been reported 14,26 Fig. 7D). In addition, the AP time-topeak feature shows opposite results for the Ctrl2 and HD72 neurons of the two different protocols (Fig. 7E). Most of the neurons recorded in current-clamp mode presented an AP time-to-peak between 0 and 200 ms. Previous findings revealed a long delay to first spike ranging from 320 to 370 ms characteristic of the striatal MSNS 27,28 . Unfortunately, we found in our preparations 24 out of 309 neurons with an AP time-to-peak longer than 200 ms, so a total of 7.8%, and just one neuron with a time-to-peak longer than 300 ms (dotted lines, Fig. 7E). These results suggest that most of the recorded neurons are non-striatal MSNs but most likely central neurons of other subtypes (Supplementary table S11). Finally, most neurons generated only a single AP following a depolarization and were therefore characterized as phasic neurons. It appears that neither the differentiation paradigm nor the genotype has a clear effect on the electrical activity of the differentiated central neurons (Fig. 7F and Tables S9 and S10). Overall, these data show here again differences in the excitability and the AP features of the neurons generated by the two protocols, without any obvious effect of the genotype itself.   www.nature.com/scientificreports/ tion where the cells are exposed to stress [29][30][31] . It is possible that we did not see HD related changes in excitability of the neurons as we did not accurately mimic the pathophysiological environment, such as an elevated  35 . We therefore wanted to investigate whether increased [Ca 2+ ] i would alter Nav gating and AP firing in hiPS cell-derived HD neurons. First, we tested in transiently transfected HEK293 cells whether gating of hNav1.3, subtype highly expressed in our hiPS cell-derived neurons (Fig. 5), is affected by a high [Ca 2+ ] i , here 500 nM. Indeed, we found a − 8 mV hyperpolarized shift in hNav1.3 steady-state fast inactivation ( Supplementary Fig. S3). To test whether this hyperpolarized shift in steady-state fast inactivation may also be observed in human MSNs expressing multiple Nav isoforms, we performed voltage-clamp recordings with 0 or 500 nM [Ca 2+ ] i in Ctr2 and HD72 neurons of the Stanslowsky protocol. However, varying the intracellular [Ca 2+ ] i did not affect their gating and steady-state fast inactivation occurred over the same voltage range in both genotypes (ANOVA, p > 0.99) (Fig. 8A-B). Again in neurons from the Stanslowsky protocol, we did not find changes in AP firing properties of both Ctrl3 and HD72 neurons under high [Ca 2+ ] i (ANOVA, p > 0.99) (Fig. 8C-F). In summary, both Nav gating properties and AP firing of the differentiated neurons remain unchanged under elevated [Ca 2+ ] i.

Discussion.
In this study, we used two protocols for differentiating MSNs from human iPS cells reported in the literature to compare the impact of the HD related CAG-repeats on the neurons' electrical properties. Both differentiation protocols showed variability and we did not observe an influence of the HD genotype on the electrical functionality of the neurons. Differentiating hiPS cells into MSNs represents a suitable way to study the neurodegenerative features of HD 13 . We were able to work with three HD cell lines carrying either 72, 109 or 180 CAG repeats (where the HD72 hiPS cell line is the most commonly studied line in the literature 36 ). The literature reports a CAG repeat mean of 41 to 45 in patients suffering from HD 37,38 whereas a CAG repeat length higher than 50 to 75 is associated with juvenile onset 2,39 . However, a study using PCR performed on a small pool of human post mortem HD brain tissue revealed 700 to 1,000 CAG repeats in some striatal neurons prior to pathological cell loss 40 . According to the authors, it seems that the CAG repeat expansion occurs earlier in striatal cells than in other regions of the brain. Therefore, we decided to also include a cell line carrying an extreme CAG repeat length to mimic the pathological genetic features of human striatal neurons.
We chose to work with two different previously published differentiation protocols which aim at driving hiPS cells into striatal neurons 14,15 . These two protocols were published to produce similar amounts of GABAergic (around 80%) and DARPP32 + (around 40%) neurons 14,26 , percentages higher than in other protocols 4,12,41 .
The two protocols share common small molecules throughout the whole differentiation but vary in specific application of small molecules and in the use of further factors and media. They present three phases of MSN generation: neural induction, regionalization and terminal differentiation 42 . While the Fjodorova protocol works with a monolayer of cells throughout the whole differentiation process, the Stanslowsky protocol mimics the three-germ-layer differentiation (ectoderm, mesoderm, endoderm) by inducing the formation of embryoid bodies (EBs) from hiPS cells. EBs are defined as small aggregates of cells in suspension 14,15 . To our knowledge, the Stanslowsky protocol is the only one reported in the literature using the formation of EBs already during the first day of differentiation.
Six different cell lines have been used throughout this study: Ctrl1, Ctrl2, Ctrl3, HD72, HD109 and HD180. Except for Ctrl3, all cell lines were used for both protocols. HD180 worked well with the Fjodorova protocol but was not successful with the Stanslowsky protocol, which is 20 DIV longer. This could be due to the large CAG expansion which may affect its differentiation potential in this specific paradigm. The HD Consortium showed that the cumulative risk of cell death is significantly higher for the HD180 than for a control cell line (33 CAG repeats) and an overexpression of 134 CAG repeats in this control cell line led to an increase risk of cell death 2,4 . In this context, it is interesting that in our hands, HD180 neurons survived the shorter Fjodorova protocol, suggesting that the critical duration for cell survival lies between DIV35 and 55. Along these lines, the HD109 iPSC-derived neurons were also not affected in our experiments, suggesting that the critical CAG length for accelerated cell death in vitro lies somewhere between 109 and 180 CAG repeats.
Although a generation of around 40% of GABAergic DARPP32 + neurons was reported in the protocols 14,15,26 , the amount of DARPP32 + neurons was much more variable in our hands, with a range of 5 to 64%, and no DARPP32 expression for the HD109 neurons of the Stanslowsky protocol. Surprisingly, no DARPP32 + signal was detected in any hiPS cell-derived neurons of the Fjodorova protocol. This discrepancy is in line with some other previously published protocols that also struggle with the homogeneity and purity of their differentiated neurons, reaching no more than 21% 41 , 14% 12 or even 5% 4 DARPP32 + MSNs in vitro. The heterogeneity of neuronal populations observed in culture and the low percentage of DARPP32 + neurons may be explained by the lack of specificity of some small molecules used in these two protocols, as well as in other published protocols 12,26,42,43 . It is also possible that the tissue of origin and the donor of the cells play an important role for a successful differentiation [44][45][46] .
We found GABAergic neurons expressing TH and hypothesized that our preparations also contained interneurons. It was previously shown that some TH + neurons do not present the dopaminergic molecular pattern but instead release GABA 47 . In agreement with this study, experiments with mice revealed that TH is found in some striatal GABAergic interneurons which exert a GABAergic inhibition onto striatal MSNs 24 . In addition to this TH/GAD67 + neurons, the expression of GABAergic interneuron markers calretinin and somatostatin in two clones of different genotypes reinforced the finding that our preparations contained striatal interneurons as  Fig. S1 and Table S6). The expression of this GABAergic interneuron marker is relatively high, thus offering the possibility to potentially study GABAergic interneurons with these differentiation protocols (20.3% and 40.9% of the HD72 and Ctrl2 neurons were positive, respectively). Recent findings revealed that striatal interneurons are involved in HD pathophysiology 48 . However the reported cells were parvalbumin positive, a cell type which we did not detect in our differentiations (Supplementary Table S6).
The RT-qPCR revealed the RNA expression of five sodium channel α isoforms, Nav1.1-Nav1.3, Nav1.6 and Nav1.7 in the iPSC-derived neurons. These results match those obtained from the striatum of WT mice (Supplementary Fig. S3). Surprisingly, except for β3 subunit, none or almost no expression of the β subunits was detected, although their level of expression in mouse striatum is higher than this of the α subunits ( Supplementary Fig. S3). Expression of the sodium channel α and β subunits in human central neurons is in accordance with previously published data from the literature. Especially, Nav1.3 is mainly found in the hippocampus and in the striatum 49   50,51 . It may be surprising to detect Nav1.7 expression in central neurons, since this sodium channel is mainly expressed in nociceptors 52 . But Nav1.7 mRNA was also expressed and regulated in some regions of the CNS, such as the hypothalamus and the olfactory bulb 53 . We were particularly interested in the level of expression of Navβ4 mRNA between control and HD cell lines. Our in situ hybridization results showed a much higher Scn4b expression in the striatum than in the cortex of WT mice ( Supplementary Fig. S1). In addition, several studies attested a significant reduction of the striatal β4 subunit expression level in human post mortem tissue as well as in rodent models of HD 7,9 . We wondered whether a β4 mRNA down-regulation may be linked to MSN degeneration and may represent a neurodegenerative marker. Unfortunately, the level of β4 subunit was extremely low in both genotypes and did not allow any statistical comparison test. Accordingly, it was not possible to use the β4 subunit expression as a marker of down-regulation in our study and no conclusion on the role of β4 subunit in the HD pathology can be made from our experiments. Additionally, it might be that hiPS cell-derived central neurons are not old enough to mimic this phenotype. As extending the maturation time of hiPS cell-derived MSNs is not trivial and promotes cell death in vitro (see above), this issue is not easily resolved. Another possibility is that hiPS cell-derived MSNs do present this β4 mRNA down-regulation. But since our hiPS cell-derived cell cultures are poor in DARPP32 + neurons, we may have missed a potential difference in β4 expression in the MSNs of each genotype. The absence of any observed effect in our study does not exclude a possible effect in human HD pathogenesis.
The use of hiPS cells is a very valuable technique since it comprises human material carrying the genetic and epigenetic features of both healthy subjects and HD patients 13,26,54 . Unfortunately, our study also points to the limits of such an approach in its current form, which is mainly a low reproducibility of published protocols and a high clone-to-clone variability. This leads to low amounts of neurons of interest in addition to various other neuronal populations in vitro.
Thus, in these conditions, a regulation e.g. of Nav β4 expression, may not be observable, even if it may occur in human HD. This study highlights the gap that still needs to be crossed to recapitulate the pathogenic features of a disease to investigate them properly using hiPS cell derived neurons.
Voltage-clamp data from our hiPS cell-derived neurons revealed an obvious lack of consistency in the Nav function of the neurons generated by the two different protocols. If no change in V half of activation was found either in genotypes or protocols despite a large variability, this was not true concerning the fast inactivation. Surprisingly, we found a more depolarized inactivation in every single clone of the Stanslowsky protocol compared to those of the Fjodorova protocol. Such strong deviations of results from the two differentiation protocols complicate any interpretation about the potential impact of the mutation on ion channel activity or neuronal firing. Moreover, the largest differences were not observed between genotypes but between protocols. This is surprising since the same cell lines were compared in both protocols and we therefore would have expected Nav channels to show the same gating properties. It is possible that the heterogeneous neuronal populations obtained from the two differentiation protocols support different Nav gating behavior, potentially explained by varying Nav channel expression or levels of posttranslational modification. Current-clamp data also revealed counterintuitive results with differences between cell lines found not only in the RMP but also in the AP features. Here again, these data emphasize the large variability among iPSC-derived MSN differentiation protocols established so far 36 and highlight the fact that data obtained from a single protocol need to be considered with care.
We focused on the AP time-to-peak criterion to assess the potential percentage of MSNs in our cultures, since a long-lasting delay to first spike is characteristic of striatal MSNs. This delay was shown to be around 370 ± 20 ms in mouse striatal MSNs 28 and as high as 325 ± 20 ms in rat MSNs 27 . The same observation was made in human iPS cell-derived MSNs, with a delay of 200 to 300 ms 26,55 . According to the low number of neurons with a latency above 200 ms to first spike (Fig. 7E), we assumed that most of our recorded neurons were no MSNs, thus confirming our immunostaining results (Supplementary Table S11). In these conditions, it was not possible to determine any potential damaging effect of the mutated huntingtin on the Nav function or the electrical excitability on MSNs, as we most likely did not have many, if any MSNs in our cultures. Thus, additional techniques such as e.g. patch-seq 56 , would only reveal reasonable, interpretable results following an optimization of the existing MSN differentiation protocols to increase the percentage of striatal MSNs generated in vitro.
Previous experiments showed a link between elevated intracellular cytosolic Ca 2+ levels and cellular apoptosis in dysregulated physiological conditions 29,30 . Particularly, experiments performed with the yeast-artificial chromosome (YAC128) HD mouse model showed a degeneration of MSNs in HD condition following disturbed Ca 2+ signaling 31,57 . The toxicity of a high intracellular Ca 2+ concentration on MSNs has not been associated with a potentially impaired neuronal activity but it is known that intracellular Ca 2+ modulates gating properties or kinetics of voltage-gated sodium channels [58][59][60] . We decided to use 500 nM as an elevated and potential pathological concentration in our own patch-clamp recordings, as this concentration is reached after repetitive dopamine applications in YAC128 mouse MSNs 57 . However, although we found a hyperpolarized shift of fast inactivation in Nav1.3 expressed in HEK cells ( Supplementary Fig. S2), 500 nM [Ca 2+ ] i did not influence either Nav channel gating of the hiPS cell-derived neurons or their excitability. Thus, we cannot conclude that pathological [Ca 2+ ] i affect HD neurons more strongly than control neurons and it remains unclear whether [Ca 2+ ] i regulation really is a hallmark of HD pathophysiology or we just cannot observe it in our model. It should be noted, that again the low amount of DARPP32 + neurons in our culture as well as the reduced age of hiPS cell-derived neurons makes it difficult to reliably compare our findings with those of animal models of HD and to make clear-cut conclusions. An optimization of the existing differentiation protocols is essential to reach larger amounts of MSNs in vitro in order to better study the pathogenicity of the mutated huntingtin on the function of human MSNs.
Conclusion. The hiPS cell-derived striatal MSNs represent a powerful tool to investigate genetic disorders such as HD. However, in this study, we have highlighted the challenges in reproducing published data from iPS