Cortical interneuron development is affected in 4H leukodystrophy

Abstract 4H leukodystrophy is a rare genetic disorder classically characterized by hypomyelination, hypodontia and hypogonadotropic hypogonadism. With the discovery that 4H is caused by mutations that affect RNA polymerase III, mainly involved in the transcription of small non-coding RNAs, patients with atypical presentations with mainly a neuronal phenotype were also identified. Pathomechanisms of 4H brain abnormalities are still unknown and research is hampered by a lack of preclinical models. We aimed to identify cells and pathways that are affected by 4H mutations using induced pluripotent stem cell models. RNA sequencing analysis on induced pluripotent stem cell-derived cerebellar cells revealed several differentially expressed genes between 4H patients and control samples, including reduced ARX expression. As ARX is involved in early brain and interneuron development, we studied and confirmed interneuron changes in primary tissue of 4H patients. Subsequently, we studied interneuron changes in more depth and analysed induced pluripotent stem cell-derived cortical neuron cultures for changes in neuronal morphology, synaptic balance, network activity and myelination. We showed a decreased percentage of GABAergic synapses in 4H, which correlated to increased neuronal network activity. Treatment of cultures with GABA antagonists led to a significant increase in neuronal network activity in control cells but not in 4H cells, also pointing to lack of inhibitory activity in 4H. Myelination and oligodendrocyte maturation in cultures with 4H neurons was normal, and treatment with sonic hedgehog agonist SAG did not improve 4H related neuronal phenotypes. Quantitative PCR analysis revealed increased expression of parvalbumin interneuron marker ERBB4, suggesting that the development rather than generation of interneurons may be affected in 4H. Together, these results indicate that interneurons are involved, possibly parvalbumin interneurons, in disease mechanisms of 4H leukodystrophy.


Introduction
Leukodystrophies are genetic disorders characterized by primary brain white matter involvement. In children, leukodystrophies are often progressive and can lead to early death. 1 Until several years ago, up to half of patients with leukodystrophies did not receive a genetic diagnosis. Since, developments in next generation sequencing techniques have led to the rapid identification of gene mutations underlying different childhood leukodystrophies, such as de novo mutations in structural genes, e.g. TUBB4A, in hypomyelination with atrophy of the basal ganglia and cerebellum (OMIM 612438) or mutations in genes encoding proteins essential for translation, e.g. DARS (OMIM 615281) or RARS (OMIM 616140) in hypomyelinating leukodystrophies. [2][3][4][5] Insights into genetic causes allowed incredible progress in diagnostics, but our understanding of the mechanisms responsible for disease pathology is still lacking. 6,7 One of the more prevalent leukodystrophies is 4H syndrome (OMIM 612440), originally characterized by hypomyelination, hypogonadotropic hypogonadism and hypodontia. [8][9][10] Other characteristics are cerebellar atrophy and myopia. Epilepsy has been described in some patients. Whole exome sequencing has revealed that 4H leukodystrophy is caused by abnormal RNA polymerase III (POLR3), and so far variants in genes encoding different POLR3 subunits, POLR3A, POLR3B, POLR1C and POLR3K, have been identified. [11][12][13][14] POLR3 is responsible for the transcription of many different classes of non-protein coding (nc) RNAs, with diverse biological functions, such as transfer (t), ribosomal (r), small nuclear (sn), small nucleolar (sno) and micro (mi) RNAs. 15 Given the variety of 4H presentations and the diverse regulatory functions of POLR3 genes, there are many potential pathways and tissue-types to research. Interestingly, some POLR3 mutations do not lead to classic brain white matter defects associated with 4H but rather show a predominant neuronal phenotype with involvement of the basal ganglia. 10,16,17 Therefore, while 4H was originally described as a typical hypomyelinating disorder, the findings suggest that POLR3 variants could also primarily affect neuronal populations.
The generation of induced pluripotent stem cells (iPSCs) from patient tissue allows for patient-specific disease models starting from early embryonic stage 18 and therefore provide a promising tool to study rare diseases like 4H leukodystrophy. To get insight into disease mechanisms and identify genes that are dysregulated by 4H mutations, we generated patient iPSC-derived cerebellar and cortical cell populations, including neuron-oligodendrocyte cocultures and studied them by transcriptome, cellomics and multielectrode array (MEA) analysis. In patient iPSC-derived cerebellar neurons, we found a significantly lower expression of ARX, which is involved in early brain development and severe infantile-onset encephalopathies. 19 As ARX is associated with cortical interneuron development, we performed immunohistochemistry for GAD65/67 on patient material and confirmed alterations in inhibitory neurons in 4H. So we decided to study the development of cortical neurons in more depth. Indeed, the number of GABAergic synapses was significantly decreased in 4H neuronal cultures, which correlated to an increased neuronal network activity. To further identify the specific subtypes that are affected in 4H, quantitative PCR (qPCR) analysis for the major cortical interneuron populations was performed. Interestingly, the expression of ERBB4, important for the development of parvalbumin (PV) interneurons, was significantly increased in 4H neurons. Together, our results show that cortical interneuron development is affected in 4H patients, possibly due to pathway changes involving ARX that may affect development of PV interneurons.

Patient consent
The Institutional Research Board of the VU University Medical Center approved the study, and written consent was obtained from all participants in accordance with the Declaration of Helsinki.

Induced pluripotent stem cell culture
Fibroblasts from anonymous donors and 4H patients were reprogrammed into iPSC lines as described previously. 20 Briefly, iPSCs were generated using an overexpression of OCT4, SOX2, KLF4 and C-MYC. IPSC lines were confirmed for pluripotency by immunocytochemistry, PCR, alkaline phosphatase, embryoid body formation assay, karyotyping and/or a pluritest. Human embryonic stem cell line H01 was obtained from WiCELL and used as a control line for RNA sequencing experiments. See Supplementary Table 1 for an overview of the iPSC lines. Control and 4H iPSCs were maintained on a vitronectin coating in TeSR E8 medium. Medium was refreshed daily, and cells were passaged once a week using Gentle Cell Dissociation reagent (StemCell Technologies) according to the manufacturer's protocol. Cells were split 1:10-1:50 to a new well for further maintenance.

Differentiation in neural cells
Differentiation into cerebellar granule cell neurons was done using a previously reported protocol, involving minimal factors in an N2 + B27 neural maintenance medium (NMM), modified for a 2D monolayer culture. 20,21 Cortical neuron differentiation was performed as previously described 22 using dual SMAD inhibition and consequent maturation of cells into glutamatergic and GABAergic neurons. For myelinating cultures, human iPSCs were differentiated into glial progenitor cells as previously described. 23 To start a co-culture, Day 18 cortical neurons were plated on a monolayer of rat astrocytes and matured for ∼2 weeks, after which iPSC-derived glial progenitor cells were added to the culture.
See Supplementary material for more details.

RNA sequencing
Total RNA was isolated from cells using TRIzol and chloroformisopropanol extraction. After quality control samples were run on an Illumina HiSeq2500 according to manufacturer's protocol. Sequencing reads were aligned to Humane Genome hg38. Similarity between fibroblast, iPSC and cerebellar cells was analysed, and the data from the iPSC and cerebellar cells were compared to previously published gene expression profiles of two iPSCs and two granule cells. Differentially expressed genes (DEG) between 4H and the control were determined with DESeq2 package in R. See Supplementary material for more details.

Quantitative PCR
For qPCR analysis, RNA was isolated from neurons using TRIzol-chloroform isolation. After production of cDNA, qPCR was performed using SYBR green according to the manufacturer's protocol. For DEG analysis, data were normalized for housekeeping gene EIF4G2. For interneuron subtype analysis, data were normalized for housekeeping gene EIF4G2 and general neuronal marker NEUN. See Supplementary Table 2 for an overview of the primers used. From the results, the fold change 2 −ΔΔCT values were calculated, and the log fold change was used for statistical analysis. Data are shown as relative expression compared with Day 30 control neurons.

Immunostaining
Cells were fixed for immunostaining using either 4% PFA or methanol:acetone fixation. Primary antibodies (Supplementary Table 3

Multi-electrode array
Neuronal activity was measured on 24-well MEA plates (Multi Channel Systems 24W300/30G-288). Cultures were measured weekly for 30 min, after which data were analysed with Multiwell-Analyzer software (Multichannel Systems) to determine spike count, burst count and network burst counts from raw data. See Supplementary material for more details.

Statistical analysis
All experiments were repeated on at least two to three independent cell batches per patient, and the results of those experiments were averaged to obtain one value per iPSC line. MEA data were collected on one batch of neurons per cell line. All analyses were either done automated or blinded. Neuronal morphology, myelination and the number of GAD65/67 + , CTIP2 + , OLIG2 + and MBP + cells were analysed using Columbus software (Perkin Elmer). Synaptic analyses and the percentage of myelinated axons were determined in ImageJ (imagej.nih.gov/ij) using the NeuronJ and SynaptoCount plugins (Supplementary material). Statistical analyses were done in IBM SPSS statistics software version 26 (IBM). Data were tested for normality with a Shapiro-Wilk test. Significance was tested with independent samples t-tests (for parametric data) or a Mann-Whitney U-test (for non-parametric data). To compare the effect of compounds on MEA data or the effect of SAG treatment paired samples t-test or Wilcoxon signed rank tests were used. All statistical tests were two-tailed and tested against alpha-level P < 0.05.

Data availability
The authors confirm that the data supporting the findings of this study are available within the article and its Supplementary material.

In vitro products present with iPSC and cerebellar cell identity
As 4H patients with POLR3B variants often have cerebellar atrophy, we differentiated iPSCs into cerebellar neurons as described previously. 21 We first wanted to establish that reprogramming and differentiation products derived from patient and control fibroblasts were unique cell populations, with relatively little variation due to source material. Further, we tested the validity of the reprogramming protocols by estimating how close each derived cell type resembled published expression profiles. To investigate sample similarity, t-distributed stochastic neighborhood embedding (t-SNE) was performed with perplexity at 5. Cluster analysis showed that samples group together based on cell types (fibroblast, iPSC, product) regardless of cell source (patient or control) (Fig. 1A). The relatively high similarity of samples from the same cell type, regardless of their source, suggests that reprogramming methods were effective at generating a consistent cell type.
To estimate the accuracy of cell reprogramming, pair-wise Spearman's rank correlations were computed for 23 samples from this study and two expression profiles obtained from ENCODE. One was an ENCODE iPSC expression profile (ENCSR722POQ) and the other an ENCODE granule cell expression profile (ENCSR313IUO), both containing two biological replicates (Fig. 1B). Hierarchical clustering showed that samples in both analyses group by cell type. Colour coding indicates that iPSCs and cerebellar cells are more closely related to the relevant ENCODE profiles than the other cell types (iPSCs = P < 0.001; cerebellar cells = P < 0.001). Taken together, fibroblast-derived iPSCs and PSC-derived cerebellar cells can be treated as relevant representatives of those developmental stages/lineages for further analyses.

Transcriptome analysis revealed 4H-associated changes in gene expression
RNA sequencing analysis showed several DEGs between 4H patients and controls. In fibroblasts, 29 genes showed a significantly altered expression (Supplementary Table 4), while in iPSCs 20 genes were differentially regulated ( Fig. 2A and Supplementary Table 5). Notable DEGs that were downregulated in 4H iPSCs include several genes involved in neural development (OTX2, NPTX1, SLITRK6, SEMA6A) and a downregulation of TMEM64 that plays a role in osteoclast differentiation. [24][25][26][27][28] In 4H cerebellar cells, 16 genes showed a significant differential expression ( Fig. 2B and Supplementary Table 6). 4H cerebellar cells showed an increased expression of PDPR, which is associated with intellectual disability 29 and an increased expression of CBLN1, a cerebellum-specific precursor protein. 30 The downregulation of ARX appears to be a potentially significant discovery. ARX mutations are associated with several brain disorders, and ARX loss of function is associated with abnormal cortical interneuron development and migration. [31][32][33][34]

Cortical interneuron changes in 4H
We studied whether the decreased ARX expression point to cortical interneuron involvement in 4H. IPSCs were differentiated into cortical neurons, i.e. a mixture of GABAergic interneurons and glutamatergic cortical projection neurons, according to previously established protocols. 22 At this stage, we were able to include three additional iPSC lines of POLR3A patients (P4-P6; Supplementary Table 1). At a neuronal precursor cell state, cells were harvested for RNA analysis. A selected number of DEGs from the cerebellar cells were tested on immature (Day 18) neurons of 4H and controls. No differences in expression levels of PDPR, ITGA11 or MACIR (not shown) were observed ( Fig. 3A and B). Consistent with the cerebellar cells, ARX expression was decreased in neurons of 4H patients [control 1.08 ± 0.31, 4H 0.26 ± 0.10, t(6) = 12.073, P < 0.001; Fig. 3C], confirming that ARX dysregulation may be a common feature in 4H.
To confirm that cortical interneurons are affected in primary brain tissue of 4H patients, we studied patient tissue for expression of GAD65/67, a pan GABAergic marker widely used to visualize all inhibitory neurons. Patient tissue showed elevated immunoreactivity for GAD65/67 compared to control, strengthening the hypothesis that cortical interneuron populations are affected in 4H (control 6% of 1315 cells; 4H Patient 1 42% of 787 cells; 4H Patient 2 36% of 1656 cells; Fig. 3D and E). As patient tissue is extremely scarce and iPSC-based model systems present differential gene expression in 4H cells, we continued functional analysis in iPSC-derived cortical cultures.

4H neurons show a decreased generation of GABAergic synapses
To investigate how cortical neuron development is affected in 4H, cortical neuronal cultures were matured and analysed for morphology changes (Fig. 4). Morphology of 4H and control neurons was analysed on dendritic (MAP2) and axonal (NF-200) staining with automated morphology analysis using Columbus software (Perkin Elmer). The measured parameters included axonal and dendritic length, branching level, and number of extremities and roots. No significant change in any parameter for neuronal morphology was observed between 4H and control neurons ( Fig. 4A-C).
To investigate whether 4H cultures present changes in the synapses, we analysed synaptic marker expression by immunocytochemistry (Fig. 4A). In our cortical neuronal cultures, cells differentiate into GABAergic and glutamatergic neurons, 22 allowing us to study differentiation efficiency into these neuronal subtypes. At Day 56 of differentiation, we studied the total number of synapses by quantifying the number of synaptophysin1 puncta per MAP2 dendritic length. We observed no significant change between the 4H and control cultures (control 0.078 ± 0.017, 4H 0.061 ± 0.018; Fig. 4D). To study whether the fraction of inhibitory synapses changed in 4H cultures, we measured the number of GABAergic presynaptic terminals by co-localization of pre-synaptic protein synaptophysin1 puncta with vesicular GABA transporter (VGAT) puncta. While 4H neurons showed a normal amount of synapses, the percentage of GABAergic synapses was significantly decreased from 46.4% in control to 26.1% in 4H neurons [t(6) = 2.503, P = 0.046; Fig. 4E]. The decreased generation of GABAergic synapses confirms that interneuron development is affected in 4H.

4H neurons show altered network activity
To assess whether the decreased percentage of GABAergic synapses had consequences for network activity, we recorded spontaneous activity of developing neuronal networks from three controls and four 4H patients plated on MEAs (Fig. 5A). The activity was captured in three parameters namely single spike events, local bursts (spike trains) and network wide bursting events (Fig. 5B). Spontaneous action potential firing in the first weeks mainly consisted of single spike events, while local and network wide bursts occurred as the cultures matured ( Fig. 5C-F). Tetrodotoxin administration (TTX, 1 μM) confirmed that baseline activity was dependent on sodium ion channels (Supplementary Fig. 1A and D). At 14 weeks post-plating, three of four 4H patients showed high levels of local and network bursts, which was mostly absent in control cultures ( Fig. 5C and F). This is an indication of hyperactivity in 4H networks, although the network activity was not significantly different from control (spikes control 1.19 ± 0.40, 4H 2.41 ± 1.88; bursts control 1.65 ± 0.13, 4H 5.61 ± 5.21; network bursts control 0.33 ± 0.60, 4H 5.17 ± 5.16). Interestingly, the single 4H line (P3) that did not show an increased number of bursts and network bursts was also the 4H line with the highest percentage of GABAergic synapses. A correlation analysis indeed confirmed significant correlations between the percentage of VGAT + synapses and the number of bursts (r = −0.774, P = 0.041) and network bursts (r = −0.786, P = 0.036), showing a relation between the amount of GABAergic pre-synapses and network activity.
To establish whether increased network activity in 4H neuronal networks is due to altered synaptic signalling, modulators for glutamate or GABA signalling were added to the cultures at 14 weeks post-plating. Network activity was measured for 10 min before and after addition of modulators. Ionotropic AMPA receptor antagonist DNQX (10 uM) or NMDA antagonist APV (50 µM) were used to measure glutamatergic signalling. The inhibition of either one of these glutamate receptors resulted in a significant decrease of spike count for both 4H and control networks ( Supplementary Fig. 1B-C, E and F). In line with the spike counts, burst and network bursts mostly disappeared (data not shown), showing that coordinated network activity depended on glutamatergic signalling and showed no obvious differences between 4H and control lines. To study GABAergic signalling, GABA A receptor antagonists bicuculline (30 µM) combined with gabazine (20 µM) or the neurotransmitter GABA (10 µM) were administered. After addition, GABA antagonists evoked network bursts in controls that were mostly absent before (burst before 2.85 ± 2.00, after 13.5 ± 5.00, Z = −2.201, P = 0.028; network burst before 0 ± 0, after 10.5 ± 6.99, Z = −2.023, P = 0.043; Fig. 5G-J). In contrast, the number of bursts and network bursts did not significantly change in 4H patients (burst before 14.13 ± 12.57, after 18.06 ± 11.60, Z = −1.886, P = 0.059; network burst before 11.7 ± 10.91, after 15.00 ± 12.45, Z = −1.527, P = 0.127; Fig. 5G These data shows that the decreased generation of GABAergic synapses correlates to an increase in network activity, and there is a decreased response of 4H neurons to treatment with GABA antagonists.

4H neurons show robust myelination in vitro
A subset of human interneurons, mainly the PV neurons, is myelinated 38 and one of the major characteristics of 4H is hypomyelination. Therefore, we decided to analyse oligodendrocyte maturation and myelination on human neuron-oligodendrocyte co-cultures. We altered a previously published protocol for neuron-OPC culture 23 to generate a myelinating co-culture. Indeed, after 40 days of co-culture we observed robust generation of mature oligodendrocytes (defined as MBP + cells) and myelination, shown by overlap between MBP and  A and B). Immunohistochemical analysis for GAD65/67 on primary cortical tissue of a control individual and two 4H patients showed that the pan interneuron marker is increased in 4H brain tissue (D and E). In iPSC-derived cortical neurons, the proportion of glutamatergic and GABAergic cells (based on GAD65/67 and MAP2 staining) was not changed between control and 4H cells (F). A significant increase in the number of CTIP2 and GAD65/67 double positive cells was observed at 14 weeks post-plating (G). ***P < 0.001, *P < 0.05, AU = arbitrary units. Open circles represent average per individual patient/control, filled circles represent the mean per genotype. Error bars represent standard error of the mean (SEM). Bars in F represent average proportion of GABAergic (dark colours) and glutamatergic (light colours) neurons, with circles representing values of individual lines. Control is shown on the left, while data from 4H cultures is shown on the right in each panel. Scale bar = 100 µm.
NF200 staining (Fig. 6A and B). The presence of myelin was confirmed by electron microscopy, which showed compacted myelin sheets around neurites (Fig. 6C).
Next, the myelinating cultures were applied to 4H cell lines. Cortical neurons of 4H or control patients were co-cultured with control glial cells. After 40 days of co-culture cells were fixed and stained for neuron (NF200) and oligodendrocyte markers (OLIG2, MBP). In both 4H and control cultures oligodendrocytes matured into MBP + cells and showed myelination of neurites (Fig. 6D). On average there was no statistically significant change in the generation of oligodendrocyte lineage cells in 4H cultures, although three of six patient lines showed a decrease in the percentage of OLIG2 + cells compared to controls (control 52.2 ± 2.74; 4H 37.9 ± 19.29; Fig. 6E). In 4H cultures, the percentage of mature oligodendrocytes (MBP + /OLIG2 + cells) was about 35% for cultures with control and 4H lines, but the variation between 4H lines was high (36.8 ± 15.97), while maturation was quite stable between control lines (34.5 ± 5.50; Fig. 6F). Myelination was analysed on immunostainings by measuring the percentage of NF200 + neurites that show co-localization with MBP. No changes in the percentage of myelinated neurites were observed between cultures with 4H and control neurons (control 34.7 ± 4.85; 4H 45.1 ± 8.05; Fig. 6G). While this neuron-glia culture set-up provides a novel tool to study myelination defects in brain diseases, we could not identify changes in 4H.

Targeting the SHH pathway does not improve 4H interneuron phenotypes
ARX is an important regulator of SHH gradients during development, 39 suggesting that the decreased expression of ARX in 4H might affect interneuron development through the SHH pathway. We tested the prospects for targeting the SHH pathway in 4H in the cortical neuronal cultures. In a pilot study it was established that twice a week treatment with SHH pathway agonist SAG at 100 nM from Day 18 onwards increased the percentage of GABAergic synapses on a 4H line, without any (negative) effects on a control line (Supplementary Fig. 2A). PCR analysis confirmed upregulation of SHH target GLI1 in 4H cells after treatment with 100 nM SAG for 2 weeks (Supplementary Fig. 2B and C). As such, all 4H and iPSC lines were treated with 100 nM SAG from Day 18 until the end point of the experiment at Day 56. Morphology analysis of neurons showed that SAG treatment affected the dendritic complexity by decreasing the number of dendritic segments, extremities and dendritic length per cell in control cells [segments vehicle 6.34 ± 1.07, SAG 4.29 ± 1.12 t(6) = 2.481, P = 0.024; extremities vehicle 3.34 ± 0.39, SAG 2.36 ± 0.46, t(6) = 3.176, P = 0.010; length vehicle 283 ± 25, SAG 193 ± 17, t(6) = 3.194, P = 0.019; Fig. 7A and B]. In 4H cells, no significant effect of SAG treatment on neuronal morphology was observed. A paired samples test comparing vehicle and SAG treatment showed that SAG treatment significantly increased However, the increase was small and no significant improvement in the 4H lines specifically was observed ( Fig. 7C and D). Although SAG treatment was able to increase expression of SHH target GLI1, it was not able to improve 4H interneuron development in vitro.

4H mutations affect parvalbumin interneuron lineage
Next, we aimed to identify whether specific cortical interneuron subtypes are affected in 4H. Primers targeted the five major human interneuron subtypes: somatostatin (SST), PV, VIP, ID2 and NDNF neurons. 40 The expression of NEUN was decreased in 4H cultures at Day 30, although the difference was not significant [control 1.27 ± 0.82, 4H 0.41 ± 0.25, t(8) = 2.171, P = 0.062; Fig. 8A]. However, since a difference in the number of neurons may affect the relative expression of GABAergic markers when only corrected for a housekeeping gene, all other markers were corrected for expression of both EIF4G2 and NEUN. DLX2 is important for development of cortical interneurons and downregulated upon maturation of neurons. 41 Fig. 8B). Despite the decreased number of GABAergic synapses in 4H cultures, markers for specific interneuron subtypes were not significantly decreased in 4H cultures ( Fig. 8C-H). In contrast, the expression of ERBB4, a marker for PV neurons, was significantly increased in 4H neurons compared to controls [Day 30 control 2.28 ± 3.47, 4H 21.45 ± 19.51, Z = −2.132, P = 0.038; Day 56 control 0.42 ± 0.09, 4H 2.50 ± 2.76, t(6) = −2.710, P = 0.035; Fig. 8H]. This suggests that the generation of specific interneuron subtypes is not impaired in 4H, but rather the development or maturation of interneurons may be affected.

Discussion
Although the genetic defects causing 4H leukodystrophy have been identified, our knowledge of underlying molecular mechanisms and the affected cellular subtypes is lacking. This study aimed to get more insight into the affected brain cell types and pathways. We started with an unbiased RNA sequencing screen on 4H and control fibroblasts, iPSCs and cerebellar cells. Although only a small number of genes were differentially regulated in all cell types, an interesting finding was the decreased expression of ARX in cerebellar cells of 4H patients. Considering the important role of ARX in cortical neuronal development and migration, 32,33,42 we studied GAD65/67 expression in patient tissue and showed an increase in GAD65/67 immunoreactivity confirming interneuron changes in the cortex of 4H patients. In iPSC-derived cortical neuron cultures, 4H neurons also showed decreased expression of ARX and affected interneuron development as measured by a decreased percentage of GABAergic synapses. No changes in the proportion of GABAergic cells were observed, suggesting the change in the number of synapses may be due to affected maturation or synapse formation in 4H neurons. The altered synaptic ratio has functional consequences, as it was correlated to an increased network activity. Decreased GABAergic signalling in 4H also became apparent after treatment with GABA antagonists. 4H neurons did not show changes in activity upon treatment with the GABA antagonists, while control cells showed a significantly increased activity after treatment. As ARX was reported to have important roles in the PV interneurons, 42 we tested whether specific interneuron lineages were affected in 4H. Indeed qPCR analysis identified an increase in ERBB4, a marker for PV neurons, confirming other results that interneuron regulation may be affected in 4H. Together, interneurons are affected in 4H and we show decreased ARX expression in different neuronal subtypes using iPSC models. Cortical neuron cultures identified a decreased number of inhibitory synapses, increased network activity, and the results are specifically pointing to defects in PV neurons.
We focused on the involvement of ARX in 4H leukodystrophy because of its role in cortical development and its association with other disorders. Loss of function mutations in ARX lead to pleiotropic disorders such as X-linked lissencephaly with ambigious genitalia (XLAG, OMIM 300215), agenesis of the corpus callosum with abnormal genitalia (OMIM 30004), and lissencephaly with cerebellar hypoplasia. 19 ARX impacts interneuron generation, development and migration 43 and loss of ARX expression in interneurons alters their excitability and causes epilepsy in mice. 33,42 In animal models it was shown that ARX acts with FOXA2 to regulate expression of SHH. 39 SHH plays important roles in the development of granule cells (and thereby cerebellar volume); oligodendrocytes (and thereby myelination); jaw and teeth; and the eyes. [44][45][46][47][48][49][50] It is possible that POLR3 disorders, and 4H in particular, involve ARX-related pathway defects, where altered ARX expression drives the clinical phenotype through an ARX to SHH pathway.
Although downregulation of ARX may have significant impact on neuronal development, we could not confirm the SHH pathway as a therapeutic target for 4H in our disease model. Neurons were treated with SAG starting from Day 18 to Day 56 of differentiation. At Day 18, the neuronal batches undergo quality control and are frozen for later use. It is possible that treatment starting from Day 18 is too late and some of the developmental alterations following a decreased expression of ARX and a dysregulated SHH pathway have already taken place. However, this would mean that targeting the SHH pathway is not a viable option for patients, who are generally diagnosed postnatally. Another explanation may be that the SAG concentration was not high enough to activate the SHH pathway, although the expression of SHH target GLI1 was increased after SAG treatment. It is possible that the affected development of cortical interneurons is caused by mechanisms that do not involve ARX. In this study we were able to show a correlation between ARX levels and interneuron generation, but additional studies are needed to show a causal effect between ARX and 4H interneuron development. Additionally, ARX works on SHH-independent pathways. For example ARX may affect histone demethylation through KDM5C 51,52 which has been implicated in neurodevelopmental disorders. 53 In conclusion, SAG treatment did not revert the neuronal phenotype in our cortical 4H cultures and following studies should reveal whether a changed treatment protocol or different target would have more beneficial effects.
The percentage of GABAergic synapses was significantly decreased in 4H cultures, suggesting an alteration in the synaptic balance. Indeed, network activity measurements showed a higher activity in most 4H cultures, although results did not reach statistical significance. Interestingly, there was a significant correlation between the percentage of GABAergic synapses and the amount of bursts and network burst, showing that synaptic balance changes also have consequences for functional network behaviour. This was also confirmed by the measurement of network activity after treatment with GABA antagonists. In control cultures, the network activity significantly increased after blocking GABA-A receptors. However, in 4H cultures treatment with GABA antagonists did not significantly change network activity, suggesting that there was less GABAergic signalling that could respond to antagonistic inhibition. Interestingly, the activity in 4H cultures did decrease after addition of GABA, which suggests a normal postsynaptic GABAergic response in 4H cells. The decreased number of inhibitory synapses could be caused by several mechanisms, e.g. reduced network maturation, a reduction in a specific subset of interneurons, less (active) inhibitory synapses per neuron, or an increase in the number excitatory synapses. Follow-up studies should give more insight into this. Also, as patient lines present differential changes, in line with the broad clinical presentation, an increased panel of patient lines would be advised.
In light of our in vitro findings, it seems contradictory that GAD65/67 immunoreactivity is higher in 4H post-mortem tissue. It could possibly be explained if only a subpopulation of interneurons is affected in 4H, as it is reported that interneuron subtypes have different levels of GAD65 54 or it can be indicative of axonal reorganization of remaining interneurons. 55 Alternatively, the increased GAD65/67 expression does not reflect changes in the interneuron populations but is rather the consequence of hyperexcitability. Epilepsy is a feature in 4H and human post-mortem studies on temporal lobe epilepsy have identified increased levels of GAD67. 56 In models of epilepsy, upon neuronal excitotoxicity and increased glutamate release, GAD65/67 is upregulated as compensatory mechanism to convert excess of glutamate into GABA. 57 If the increase in GAD65/67 immunoreactivity is indeed a secondary effect in 4H this would explain why there is no change in the number of GABAergic cells in the iPSC-derived cultures. However, while post-mortem analysis confirmed network changes in 4H, follow-up studies are warranted.
Alterations in GABAergic activity may underlie 4H associated neurological signs like epilepsy, but also hypomyelination. It is currently unknown how POLR3 mutations cause hypomyelination. Coulombe et al. 58 hypothesized that hypomyelination in 4H is either caused by an unknown POLR3 target that plays a key role in myelin biogenesis, or that POLR3 mutations lead to a globally reduced transcription at a crucial milestone in oligodendrocyte development. We investigated an alternative hypothesis, where the altered myelination is secondary to neuronal dysfunction. Cortical (PV) interneurons in the human brain are myelinated, 38,59 and myelination can be regulated by GABA receptor activity on oligodendrocytes. 60 To study whether the decreased interneuron generation in 4H cortical cultures also affected myelination a new co-culture protocol was developed. This co-culture of human neurons and glial cells show robust maturation of oligodendrocytes and myelination of human axons, a challenging phenomenon to study in vitro. No differences in oligodendrocyte maturation or myelination were observed between cultures with control or 4H neurons. This could suggest that the hypomyelination observed in 4H patients is caused by an oligodendrocyte intrinsic effect, rather than mediated through affected neurons. However, it is also possible that our culture set-up was not sensitive enough to pick up changes in myelination for 4H. The neurons used for the co-culture consist of a mix of GABAergic and glutamatergic neurons, so a myelination defect on interneurons may be masked by normal myelination of glutamatergic neurons. Additionally, PV neurons mature during late development and not all neurons in the presented co-culture system showed myelinated processes. Although the novel neuron-glia coculture can measure myelination in vitro, further improvement of these model systems may be necessary to identify mechanisms underlying hypomyelination in 4H leukodystrophy.
We aimed to identify the interneuron subtype that was most affected in 4H cultures. We were not able to identify a significant decrease in any of the five major cortical interneurons populations. 40 Instead, we found a significantly increased expression of ERBB4, an early marker for PV interneurons. There are a few possible explanations for the discrepancy between the decrease in GABAergic synapses and the increase in ERBB4 expression. It is possible that interneuron generation in 4H is normal, but that interneurons are affected during later development or maturation stages. This is consistent with the finding that there were no changes in the number of GAD65/67 + cells in 4H cultures. As ERBB4 is an early marker for PV interneurons, it is also possible that the increase reflects a lack of maturation, rather than an actual increase in PV neurons. A defect in maturation would be consistent with the sustained DLX2 expression, which decreased in control cells between Day 30 and Day 56, but not in 4H cells. Unfortunately, PV interneurons mature late in cortical development (late gestation/postnatal [61][62][63] ), and we were not able to measure PVALB expression in our neuronal cultures. PV interneurons are the largest group of GABAergic neurons in the cortex. They are fast spiking interneurons with a high metabolic demand, which makes them vulnerable for injury. 64 Interestingly, the development of SST and PV interneurons is depending on SHH levels, 65,66 and deletion of Arx in PV neurons led to increased neuronal activity and altered synaptic properties in a mouse model. 42 PV interneurons play a crucial role in spike timing of glutamatergic neurons, and abnormalities in PV signalling have been implicated in epilepsy, occasionally seen in 4H patients. For example, optogenic stimulation of PV neurons can induce ictal events in mice 67 probably due to a depolarization block. 68 Overall, our data suggest that PV interneurons may be affected specifically in 4H and should be further studied.
To conclude, transcriptome analysis of cerebellar cells differentiated from patient iPSCs revealed ARX as a potentially important regulator of 4H brain pathomechanisms. Indeed, a decreased expression of ARX was confirmed in cortical neuron cultures, which showed affected generation of GABAergic synapses consistent with known effects of ARX on interneuron development. The decrease in GABAergic synapses correlated to increased neuronal network activity. QPCR analysis revealed alterations in ERBB4 expression, suggesting the PV interneurons might be of particular interest for 4H. Although more research is needed, this study provides a first insight into specific brain cell types and pathomechanisms affected in 4H leukodystrophy.