Dystrophin deficiency affects human astrocyte properties and response to damage

In addition to progressive muscular degeneration due to dystrophin mutations, 1/3 of Duchenne muscular dystrophy (DMD) patients present cognitive deficits. However, there is currently an incomplete understanding about the function of the multiple dystrophin isoforms in human brains. Here, we tested the hypothesis that dystrophin deficiency affects glial function in DMD and could therefore contribute to neural impairment. We investigated human dystrophin isoform expression with development and differentiation and response to damage in human astrocytes from control and induced pluripotent stem cells from DMD patients. In control cells, short dystrophin isoforms were up‐regulated with development and their expression levels changed differently upon neuronal and astrocytic differentiation, as well as in 2‐dimensional versus 3‐dimensional astrocyte cultures. All DMD‐astrocytes tested displayed altered morphology, proliferative activity and AQP4 expression. Furthermore, they did not show any morphological change in response to inflammatory stimuli and their number was significantly lower as compared to stimulated healthy astrocytes. Finally, DMD‐astrocytes appeared to be more sensitive than controls to oxidative damage as shown by their increased cell death. Behavioral and metabolic defects in DMD‐astrocytes were consistent with gene pathway dysregulation shared by lines with different mutations as demonstrated by bulk RNA‐seq analysis. Together, our DMD model provides evidence for altered astrocyte function in DMD suggesting that defective astrocyte responses may contribute to neural impairment and might provide additional potential therapeutic targets.


| INTRODUCTION
Duchenne Muscular dystrophy (DMD), a progressive neuromuscular disease that affects 1 in 3000-5000 male children, leads to loss of ambulation by the early teens and death by the 3-4th decade of life (Blake et al., 2002;Muntoni et al., 2003;Ricotti et al., 2016). In addition to muscle wasting, and much less well understood, are the central nervous system comorbidities occurring in at least a third of DMD patients. These patients exhibit an IQ on average one standard deviation lower than the general population and suffer from a range of neuropsychiatric comorbidities including attention deficit hyperactivity disorder (ADHD) and autism (Pane et al., 2012;Ricotti et al., 2016;Wingeier et al., 2011). This research suggests that the incidence of the brain co-morbidities may be underreported, as borderline deficits occur in a larger proportion of DMD patients.
DMD is caused by mutations in the dystrophin gene, DMD, the largest gene in the human genome. DMD is composed of 79 exons and its expression is regulated by different promoters with distinct tissue specificity. It contains 3 promoter regions encoding full-length isoforms (Dp427) as well as at least 4 internal promoter regions encoding the shorter isoforms (Dp260, Dp140, Dp116 and Dp71/ Dp40). All these shorter isoforms with possibly the exception of Dp116 are transcribed in the central nervous system (CNS) but not in muscle (Waite et al., 2012). DMD individuals typically have intragenic loss of function mutations of the dystrophin gene, with the most frequent mutations being out-of-frame deletions (Aartsma-Rus et al., 2016;Juan-Mateu et al., 2015). All DMD mutations affect the expression of the full length dystrophin isoform, Dp427, produced both in muscle and in the brain. Genetic studies have clearly implicated a role for Dp427, Dp140 an Dp71 in the severity of the brain co-morbidities (Chaussenot et al., 2018;Hoogland et al., 2018;Pane et al., 2012;Ricotti et al., 2016).
The role of dystrophin has been extensively studied in muscle, where the full-length isoforms play an important structural role by connecting the cell cytoskeleton to the extracellular matrix (Blake et al., 2002). In contrast, relatively little is known about the role of the different isoforms in the brain, particularly in humans (Hoogland et al., 2018). It is still debated whether the neural deficits observed in DMD children are entirely acquired during prenatal development, hence "fixed," or there is some progression of the neural phenotype at least in the early years when the CNS is still highly plastic (Bagdatlioglu et al., 2020). Recent research has indicated differential expression of DMD transcripts between fetal and adult brain (Doorenweerd, Mahfouz, et al., 2017a); however, information on different dystrophin isoform expression in the developing human brain and in different neural cell types is still limited. More basic information on the expression of dystrophin isoforms in different human neural cells is required to develop a better understanding of how DMD mutations may affect human brain function. Dystrophin protein isoform analysis is challenging due to the complexity of this protein, with "new" dystrophin splice isoforms still being discovered, and to the low abundance of these proteins and transcripts (Aragon et al., 2018;Rani et al., 2019;Tadayoni et al., 2012).
Studies in a DMD mouse model lacking the full-length isoform have suggested a reduction in functional receptors at the GABAergic synapses, which is consistent with increased anxiety and fear responses observed in behavioral studies in these mice (Sekiguchi et al., 2009;Vaillend & Chaussenot, 2017). Altered distribution and clustering of synaptic proteins was also observed. In addition, mice lacking Dp71, which is expressed in various brain regions including neurons in the hippocampus, cortex and olfactory bulb, as well as in retinal and perivascular astrocytes, present several abnormalities (Aragon et al., 2018).
Astrocytes have been shown to play a role in the development of synaptic connections, and plasticity and homeostasis of neuronal circuits, for instance by modulating glutamate uptake (Pekny et al., 2016;Sofroniew & Vinters, 2010). Increasing evidence points towards a role for astrocytes in neuropsychiatric disorders including depression, in neurodegenerative disorders such as Alzheimer's disease and neurological disorders like epilepsy, which has high occurrence in DMD patients (Akther & Hirase, 2021;Hendriksen et al., 2018;Jones et al., 2017;Patel et al., 2019;Tsao & Mendell, 2006;van den Bergen et al., 2014). We therefore hypothesized that defects in astrocyte function may also contribute to the neural pathology of DMD.
Further knowledge of DMD expression in the developing human brain and of which cell types may be affected by dystrophin mutations is required to better understand the brain co-morbidities in DMD patients and eventually develop therapeutic strategies. Here we have addressed some of these issues, firstly by analyzing changes in dystrophin isoforms in the human developing brain and upon differentiation of human neural stem cells (hNSCs) into neurons and astrocytes, and secondly by testing the hypothesis that astrocytes derived from DMD patient iPSCs (induced pluripotent stem cells) with mutations affecting several brain isoforms have functional deficits. We show that all shorter dystrophin isoforms are up-regulated with normal development, and that whilst neurons express most isoforms, Dp71 is the main isoform expressed in astrocytes. Importantly, we show for the first time that DMD astrocytes not only display morphological and molecular changes under basal conditions as compared to healthy controls, but also demonstrate altered response to inflammatory stimuli and oxidative stress. Our results highlight the importance of studying dystrophin expression in human neural cells, and point at a potential involvement of astrocytes in the observed brain comorbidities in DMD patients.

| METHODS
All cell culture reagents were obtained from Life Technologies unless stated otherwise.

| Human tissues
All procedures involving human tissue were carried out in compli-

| Cell maintenance and differentiation
All cell lines and primary cultures were maintained in a humidified incubator at 37 C with 5% CO 2 .
They were grown in serum-free medium as previously described and passaged using Accutase (Kin Pong et al., 2014;Vagaska et al., 2016).
All experiments were carried out using hNSCs between passages 10 and 20.
For astrocyte differentiation, hNSCs from all three Carnegie stages were passaged when at 80% confluency and plated directly into astrocyte differentiation medium consisting of DMEM/F12 medium with Glutamax supplemented with 1% P/S and 10% fetal bovine serum (FBS, Gibco). Astrocytes were cultured for a minimum of 3 weeks before commencing experiments with medium changes every 2-3 days. Astrocytes were plated at 6 Â 10 5 cells/6 well plate for analysis by immunocytochemistry or western blotting. Neuronal differentiation was induced by a growth factor removal protocol previously described with minor modifications (Sun et al., 2008;Vagaska et al., 2016). Briefly, hNSCs from the three different Carnegie stages (CS17, CS21 and CS23), were plated onto laminin coated 6-well plates (Corning) at 6 Â 10 5 cells/plate. Cells were maintained in EGF free hNSC medium for 7 days, and then FGF2 and heparin were also withdrawn and hNSCs grown for a further 5 days.
For 3D cultures, cells were mixed in hydrogels consisting of 1 mg/ml collagen I (Rat Tail High Concentration, Corning) and 2 mg/ml Matrigel (Corning) as described elsewhere (Vagaska et al., 2020) and allowed to polymerize for 40 min. In brief, a cold collagen I solution at pH to 7.4 was added to the Matrigel at 4 C before mixing the hydrogel with the cell suspension and inducing polymerization at 37 C for 30 min. In some experiments, hydrogels made of 100% collagen or 100% Matrigel were tested and different cell seeding protocols compared (Supplementary Figure 1). Astrocytes were differentiated for 3 weeks before being seeded into hydrogels, neurons were predifferentiated for 14 days. Hydrogels were set in either 8 well chamber slides (50 μl with 1.5 Â 10 4 cells per gel) or 6 well plates (500 μl with 1.5 Â 10 5 cells per gel) and maintained for 5 days (hNSCs) or 14 days (astrocytes and neurons) before being fixed in 4% PFA for 30 min.

| Induced pluripotent stem cells (iPSCs) growth and astrocyte differentiation
iPSCs were grown on Matrigel-coated 6 well plates, maintained in mTESR1 medium (Stemcell Technologies) with 1% penicillin/ streptomycin (P/S) (Life Technologies) and passaged using ReLESR (Stemcell Technologies). iPSC lines were generated using episomal reprogramming either in our laboratories (healthy controls 1 and 2, according to (Hawkins et al., 2016), and DMD68 (c.9851G > A [p. Trp3284X] in exon 68) by mRNA technology (Ferrari et al., 2020) To induce a neural progenitor fate, we used a protocol previously described (FitzPatrick et al., 2018). iPSCs were plated at 1 Â 10 4 cells per well in u-bottom 96 well plates in neural induction medium supplemented with ROCKi. After 24 h, neuroepithelial clusters were visible and ROCKi was withdrawn. Daily medium changes were performed for 5 days after which the clusters were replated onto laminin coated 6 well plates. Neural rosettes were visible after 4 days, manually picked, centrifuged at 200g for 5 min and replated onto laminin coated plates in neural maintenance medium. These neural progenitor-like cells (NPCs) were expanded for a minimum of five passages before astrocyte differentiation was induced by directly plating NPCs into astrocyte differentiation medium as for hNSCs. These were maintained for a minimum of 3 weeks before further experiments.
Astrocytes were plated at 1 Â 10 4 cells/cm 2 for immunocytochemistry, cell viability and other assays.

| Protein extraction
Proteins were extracted from embryonic brains at 10, 15 and 19 pcw (n = 3 for each stage) from human fetal muscle at 20 pcw, and from wild type mouse brains (mouse protein extracts were kindly provided by S. Torelli). In brief, brain tissues mechanically dissociated in lysis buffer (4 M Urea, 125 mM Tris, 4% Sodium Dodecyl Sulfate) were shaken for 24 h at 4 C, before centrifuging the suspension for 15 min at 10,000g. To extract proteins from cell cultures, cells were rinsed with PBS and then lysis buffer added into the well. In some experiments, proteins were extracted with RIPA buffer (Sigma) for comparison. The cell lysate was kept at 4 C for 30 min and then centrifuged for 15 min at 10,000g. Protein concentrations were measured using the BCA assay kit (Thermofisher) according to the manufacturer's instructions; protein extracts were stored at À20 C until needed for western blot. Cells in 3D cultures were not analyzed by western blot due to interference of hydrogel proteins.

| Western blotting
After denaturing the protein extracts in sample buffer (Invitrogen) for Tween (TBS-T) and 10% semi-skimmed milk powder) for 1 h. Incubation with primary antibodies was overnight at 4 C and with secondary antibodies for 1 h at room temperature. All antibodies were diluted in in TBS-T containing 5% semi-skimmed milk powder. The antibodies used are listed in Table 1.
Bound antibodies were at first visualized using the Amersham ECL Western Blotting Detection system (GE Healthcare) kit and imaged on X-ray films (GE Healthcare) due to greater sensitivity in detecting dystrophin. For semi-quantitative analysis, Li-Cor Intercept Blocking Buffer and IRDye secondary antibodies were used and all Western blots were developed using an Odissey CLx (Li-Cor), band intensity was measured Li-Cor Image studio. Band intensity was normalized to house-keeping proteins (actin or GAPDH). Independent experiments were carried out at least 2 times, and most commonly, whenever sufficient material was available, in triplicates.

| Immunocytochemistry
All antibodies used are listed in Table S1. 2D and 3D cultures were fixed in paraformaldehyde (PFA) in phosphate-buffered saline (PBS, pH 7.4) at room temperature. After 1 hour in blocking buffer (10% Normal Donkey Serum/Normal Goat Serum and 0.1% Triton-X100 in PBS), cells were incubated overnight with the primary antibodies at 4 C and then with the secondary antibodies, together with Hoechst dye 33,258 (2 μg/ml) to counterstain the nuclei, for 1 h at room temperature. Negative controls were incubated with the secondary antibodies only.
Images were acquired using either an Olympus IX71 inverted microscope with an ORCA-R2 digital camera (Hamamatsu Corporation, Bridgewater, NJ) or a confocal microscope (LSM710, Carl Zeiss, Jena, Germany). All images from immunostainings carried out at the same time and collected under the same conditions for comparison were analyzed with Fiji/Image J and/or Imaris software. Fluorescent intensity values were obtained by measuring intensity in 10 cells per image.
Corrected total cell fluorescence intensity was obtained by multiplying the area of each selected cell by the mean fluorescence background reading and deducting this value for the integrated density. Analysis of cell morphology was performed as detailed below ( Figure S2).

| Reverse transcription semi-quantitative PCR (RT-qPCR)
RNA was isolated using a RNeasy Mini kit (Qiagen) according to manufacturer's instructions. cDNA conversion was performed using RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific) also per manufacturer's instructions. cDNA was amplified either using the Taqman or SYBR-green method. Taqman Fast Advanced master mix (Thermo Fisher Scientific was used with Taqman probes for IL-6 (Hs00174131_m1) and VEGFA (Hs00900055_m1), alongside three housekeeping genes ATP5B (Hs00756996), EIF4A2 (Hs00756996_g1) and UBC (Hs00824723). The ΔΔCt method was applied to obtain the geometric mean of relative fold-change of data normalized to each reference gene. The QuantiTect SYBR Green master mix (QIAGEN) was used to amplify PADI2 (Forward primer: TGAAGCACTCGGAACACGT, Reverse primer: TTGTCACTGCTGGCCTCG) and the housekeeping gene RPL19 (L19) (Forward primer: GCGGAAGGGTACAGCCAAT, Reverse primer: CAGGCTGTGATACATGTGGCG). Three independent samples, each with technical triplicates, were amplified. The ΔΔC t method was used to calculate the relative fold-change of data normalized to each reference.

| Astrocyte growth assay
Astrocytes derived from either healthy or DMD NPCs were plated at a density of 1.5 Â 10 4 cells/cm 2 and 5 0 -bromo-2 0 -deoxyuridine (BrdU) (10 μM, Sigma) was added 2 days after plating for 24 h. Cells were fixed in 4% PFA for 10 min, washed 3 times in PBS and incubated at room temperature in 2 N hydrochloric acid for 20 min. The pH was then neutralized with 0.1 M borate buffer and cells were stained as described above with Ki67 and BrdU antibodies.

| Astrocyte functional assays
For all functional assays, astrocytes derived from either normal or DMD NPCs were plated at a density of 1.5 Â 10 4 cells/cm 2 and experiments started by 3 days after plating to minimize possible bias due to differential growth. The CellTiter-Glo 3D ATP assay (Promega) was used to measure production of ATP. Equal amounts of CellTiter-Glo reagent was added to the astrocyte medium and the cultures shaken for 5 min. After 25 min at room temperature, the luminescence was measured using a FLUOstar OPTIMA (BMG Labtech).
Briefly, MTT (2.5 mg/ml) was added to astrocytes grown in 96 well plates and incubated for 2 h. After washing the cells with PBS, 200 μl dimethyl sulfoxide (DMSO) was added to each well and 100 μl of the supernatant was used to measure absorbance at 595 nm in a Thermo Scientific MultiSkan spectrophotometer.

| Glucose oxidation assay
Astrocyte glucose metabolism was measured using gas chromatography isotope-ratio mass spectrometry (GC-IRMS). Astrocytes were kept in experimental medium (glucose free DMEM with 15 mM HEPES, 2.9 mM sodium bicarbonate, 2 mM L-glutamine, 0.5 mM sodium pyruvate and 21.5 μM phenol red) supplemented with 10% FBS and 3 mM D-glucose for 20 h. Cells were washed with PBS and 3 ml experimental DMEM containing 3 mM [U-13 C] Glucose was added per well. To prevent the loss of 13 CO 2 wells were sealed with a 3 ml layer of heavy mineral oil, and 100 μl samples were taken every hour for 6 h. Samples were stored in rubber-sealed Exetainer vials (Labco Ltd, Ceredigion, UK) and stored at À20 C until analysis. Once thawed, 100 μl 1 M hydrochloric acid was injected through the rubber lid to release 13 CO 2 . After centrifugation for 30 s at 500g, samples were analyzed on a GasBenchII linked to a Thermo Delta-XP isotoperatio mass spectrometer (Thermo-Finnigan, Bremen, Germany).

| Astrocyte stimulation
For all stimulation assays, astrocytes derived from either normal or DMD NPCs were plated at a density of 1.5 Â 10 4 cells/cm 2 and experiments started by 2 days after plating to minimize possible bias due to differential growth. Astrocytes were stimulated for 24 h.

LPS/IFNy treatment
Astrocyte activation was induced by addition of lipopolysaccharides (LPS, 1 μg/ml, Sigma) and interferon γ (IFNγ, 90 ng/ml, Peprotech) to the astrocyte medium. Untreated and treated astrocytes were kept for 24 h following stimulation, after which they were fixed as previously described and stained for Vimentin and GFAP (glial fibrillary acidic protein). Astrocyte morphology (soma size) and cell number were measured as indicators of astrocyte response. Soma size was measured in sub-confluent cultures using ImageJ by manually outlining the soma of individual astrocytes excluding any processes ( Figure S2); dividing cells were not measured. A minimum of 100 cells were measured per condition for soma size, and a minimum of 100 cells were counted per biological repeat in cell response experiments.  (Brown, 1975). The analyses were performed on three cell line comparisons, each DMD astrocyte line versus healthy 2A, 12pcw and Hipsc astrocytes; p values were Bonferroni corrected for three tests. This is conservative given that the disease groups are not independent. For comparison, p values were also combined using the Fisher's method (Dai et al., 2014) and the harmonic mean (Wilson, 2019). Log2fold change was combined by fixed effects meta-analysis to yield a standardized mean difference. These analyses were performed using a custom R script.

| Statistical analysis
For western blotting and experiments using hNSCs, 3 different cell lines with a minimum of 2 replicates each were assessed. In all metabolic experiments 6 repeats per cell line (3 DMD lines, 2 control lines) were included. Significant changes in protein levels, astrocyte growth, metabolic activity and response to damage were assessed either by Welch's t-test or one-way analysis of variance (ANOVA) with Bonferroni correction where appropriate using Graphpad Prism 7 software. Data are presented as mean ± SEM. A p value <.05 was considered significant. *: p < .05, **: p < .01, ***: p < .001, ****: p < .0001.

| Dystrophin expression in human embryonic brain and neural cells
Expression of dystrophin isoforms in human neural tissue and cells was investigated using a number of antibodies binding to different regions of dystrophin, some of which have been widely used, though mainly in studies on muscle tissue ( Figure S3). These included polyclonal antibodies, referred to here as Dp(A) and Dp(P) for simplicity, that were raised against epitopes in the C-terminus of dystrophin and in a sequence at the N-terminus of Dp71 shared by all isoforms, respectively, and a range of a monoclonal antibodies from Prof. Morris (http://www.glennmorris.org.uk/mabs/Dystrophin.htm). These had not been previously tested on developing human brain. MANEX1A, shown to recognize an epitope within the first 68 amino acids of dystrophin (Morris et al., 1995), hence referred to here as Dp1-68, appeared to recognize a high molecular weight band both in brain and muscle. However, in depth analysis of this band showed it could be resolved into two bands, with the larger one detected only in muscle, consistent with Dp427, and the smaller one in developing brain and muscle as well as neurones and astrocytes ( Figure S4a). Based on analysis of expression in DMD68 cells, which lack all dystrophin isoforms ( Figure S4b), this band appeared to represent a cross-reactivity of the antibody as it was present in both normal and DMD68 cells. Among other monoclonal antibodies tested, neither MANDYS1, which reacts with dystrophin rod domain and is expected to detect DP427 and Dp260, nor DYS2, which recognizes 17 amino acids at the C-terminus ( Figure S4c), or MANDRA1 ( Figure S5a) (Lam et al., 2014), which recognizes three amino acids within the C-terminus peptide used to raise Dp(A), displayed high sensitivity on brain proteins in western blots.
Hence, the widely used Dp (A), as well as Dp (P), were used to investigate dystrophin isoform expression during human embryonic development by western blotting in protein extracts from 9 human embryonic brain samples from 10, 15 and 19 pcw (  (Figure 2c). This distribution was not altered either by changing the stiffness of the gel using either 100% collagen or 100% Matrigel (data not shown) or by modifying the seeding protocol to exclude higher affinity of the astrocytes for the plastic rather than the hydrogel ( Figure S1).
All three cell types were stained with Dp(A) antibodies both in 2D and 3D cultures (Figure 2). hNSCs grown in 2D or in 3D hydrogels were immunostained 5 days after seeding to avoid possible occurrence of spontaneous differentiation due to increased culture density; astrocytes were immunostained 21 days after seeding, and neurons were immunostained 14 days after growth factor withdrawal either in

| Differentiation of human iPSCs into astrocytes
The three DMD patient derived iPSC lines, DMD52, DMD67 and DMD68 and two healthy control iPSC lines were expanded and differentiated into neural progenitor cells (NPCs) and then into astrocytes as summarized in Figure 3a. Similar reactivity for a range of neu- detected in a small subset of NPCs (<3%), and so was β3-tubulin (<6%). In contrast, O4, an oligodendrocyte lineage marker, was undetectable (not shown). It was also observed that SOX2 expression was reduced at later passages ($7 upwards) indicating a shift towards a more differentiated phenotype. No staining was detected when primary antibodies were omitted ( Figure S8).
Astrocyte differentiation was assessed by immunofluorescence staining (Figure 3c and Figures S8 and S9). Again, the three DMD lines displayed a comparable behavior as compared to controls. Following (c) Control and DMD52 astrocytes. Note the change in morphology and loss of SOX2 upon differentiation. Expression of glutamine synthetase (GS), and of the excitatory amino acid transporter 1 (EAAT1) are detected in addition to GFAP and S100β. All scale bars = 100 μm F I G U R E 4 Glial Fibrilliary acidic protein (GFAP) and aquaporin 4 (AQP4) expression is altered in Duchenne muscular dystrophy (DMD) astrocytes. (a) GFAP detection by western blot. Note increased GFAP expression in DMD astrocytes as compared to control astrocytes (normalized to GAPDH). (b) AQP4 detection by western blot. Note that there is no difference in the expression of the 37 kDA isoform, whereas expression of the 30 kDa and of the large AQP4 (80-100 kDa, ubiquinated and tetrameric AQP4) isoforms is significantly reduced or increased, respectively, in DMD astrocytes as compared to controls (normalized to GAPDH). N = 3 DMD lines, N = 2 ctrl lines, 2 technical repeats. Data are presented as mean ± SEM and analyzed by Welch's t-test. *: p < .05 neuronal (MAP2, Figure 3b,c and Figure S7) and oligodendrocyte (O4, not shown) markers was detected. Together the staining pattern confirmed that healthy and DMD NPCs had undergone astrocytic differentiation, though a reduction in soma size was observed in the three DMD cell lines as exemplified in Figure S2 (see also Figure 6 and Figure S9). Furthermore, a significant reduction in EAAT1 staining was observed in all DMD astrocytes as compared to control astrocytes ( Figure S10).

| Characterization of DMD astrocytes
DMD astrocytes were further characterized in the three DMD lines by assessing expression of a component of the astrocyte cytoskeleton, GFAP, and of Aquaporin 4 (AQP4), a protein expressed by astrocytes and known to be important in blood brain barrier function, by western blot. As dystrophin contains actin binding sites, we used expression of GAPDH to normalize protein expression in addition to expression of β-actin. Normalization to either house-keeping gene yielded similar results. While in the DMD astrocytes β-actin levels were not affected, GFAP was significantly higher than in controls ( Figure 4a). Analysis of AQP4 showed that high molecular weight species (87 and 100 kDa), likely to be ubiquitinated AQP4 or AQP4 tetramers (Goodyear et al., 2008), were significantly increased in DMD astrocytes, whereas expression of the low molecular weight isoform (30 kDA) was significantly reduced compared to control astrocytes ( Figure 4b) suggesting functional changes in AQP4 channels.

| Metabolic activity in healthy and DMD astrocytes
As astrocyte metabolic activity may have a substantial impact on neuronal health and homeostasis, we examined glucose oxidation in control and the three DMD astrocyte lines as an indicator of metabolic activity. The amount of glucose oxidized to CO 2 by astrocytes varied between cell lines, with slightly higher values in DMD than control astrocytes observed under basal conditions (Figure 5a). We then assessed astrocyte metabolism under basal condition by using an MTT assay and measuring ATP levels. Both assays showed a significantly increased metabolic activity in DMD astrocytes compared to control astrocytes (Figure 5b,c). To establish whether this increase primarily reflected increased cell number due to increased proliferation, expression of Ki67, a nuclear protein expressed at all stages of cell replication but not in quiescent cells, and BrdU incorporation were assessed (Figure 5d,e). Both were found to be higher in all three DMD astrocyte lines than in controls, but the fold change in BrdU incorporation (5.5%) and in Ki67-positive cells (29.2%) was lower than the over two-fold increase in MTT and ATP activity (Figure 5e). Furthermore, when the total cell number was counted 3 days after plating, there was no statistically significant difference between healthy and DMD astrocytes, though a trend towards a higher cell number in DMD was observed ( Figure S10). This suggests that increased proliferation does not entirely account for the increase in metabolic activity detected with these assays under basal conditions.

| DMD astrocyte response to noxious stimuli
Given the morphological, molecular and growth differences observed in DMD astrocytes (Figures 5b-e and 6a, Figure S6), we investigated their function by exposing them to two types of stressors. One was exposure to LPS and IFNγ, which is commonly used to mimic inflammation and induces activation of primary astrocytes (Lange et al., 2018;Sheng et al., 2011), and the other exposure to H 2 O 2 to induce oxidative stress ( Figure 6). in DMD but not in healthy astrocyte cultures. Furthermore, at 500 μm H 2 O 2, cell death was significantly higher in DMD astrocytes than in controls. Together, the response to H 2 O 2 observed suggests that DMD astrocytes are significantly more sensitive to oxidative stress than control astrocytes.

| Pathways affected in DMD astrocytes
In order to assess which are the pathways affected by dystrophin mutations that underpin the increased sensitivity to damage in DMD astrocytes, we compared astrocyte from three healthy lines, and the  (Table S1). Here we focus on terms that were significantly enriched in both DMD52 and DMD68 astrocytes Note that DMD astrocytes are significantly smaller than their healthy counterparts under basal conditions and following stimulation; significant reduction in soma size upon LPS/IFNγ stimulation is observed in healthy but not in DMD astrocytes. LPS/IFNγ stimulation increases the percentage of healthy but not DMD astrocytes. >100 cells were counted per condition, N = 3 DMD lines, N = 2 healthy lines. (c) Analysis of cell survival and death following induction of oxidative stress with hydrogen peroxide (H 2 O 2 ) for 24 h in DMD and healthy astrocytes. Note that the percentage of cells present declines more rapidly with increasing concentration of H 2 O 2 in DMD than in healthy astrocytes and a significant increase in dead cells, measured as percentage of nuclei (blue) that are propidium iodide (PI, red)-positive is already observed at 100 μM H 2 O 2 in DMD astrocytes. >100 cells were counted per technical repeat (technical repeats: N = 3). Scalebar =100 μm. Data are presented as mean ± SEM and analyzed by one way ANOVA with Bonferroni's correction. *: p < .05; ***: p < .001 ***, ****: p < .0001 data by assessing expression of some genes by RT-qPCR. Consistent with the RNA-seq changes, interleukin 6 (IL-6), vascular endothelia growth factor A (VEGFA) and the calcium-dependent enzyme, peptidyl-arginine deiminase 2 (PADI2) were found to be upregulated in both DMD52 and DMD68 astrocytes (Figure 10b). Furthermore, according to RNA-seq analysis, EAAT1 expression is down-regulated in DMD52 and DMD68 astrocytes, consistent with reduced EAAT1 protein expression assessed by immunofluorescence intensity data ( Figure S10).
As we observed increased metabolic activity in DMD astrocytes, we performed metabolic reaction enrichment analysis (MAREA) (Damiani et al., 2020) focusing on carbon metabolism ( Figure 11).
Metabolic reactions involved in ornithine/arginine/citrulline mitochondrial transport were upregulated across both DMD astrocyte lines, as well as phosphoenolpyruvate carboxykinase, a key enzyme of gluconeogenesis, folate metabolism and H 2 O metabolism. Metabolic reactions involved in fatty acid metabolism were upregulated in both lines, although this was more extensive in DMD68 F I G U R E 7 Gene ontology (GO) enrichment analysis reveals common pathways are disrupted in DMD52 and DMD68 astrocytes. (a) Venn diagram derived from analysis of bulk RNA sequencing of astrocytes from 3 control lines and two DMD lines (52 and 68): 5657 genes are found to be up-regulated and 6527 genes down-regulated in both DMD lines as compared to controls. (b-d) GO enrichment analysis shows upregulation of genes involved in transport and localization, as well as binding, in astrocytes from both DMD lines. In contrast genes involved in neurogenesis, nervous system development and synapses are significantly down-regulated (p < .01) astrocytes. MAREA analysis confirms that metabolic reactions are upregulated in DMD astrocytes, though to a greater extent in DMD68 astrocytes.

| DISCUSSION
Here we have reported changes in dystrophin protein expression in the human developing brain and in human neural progenitors and their progeny, and importantly shown that the phenotype, metabolism and injury response of astrocytes from DMD patients are altered, consistent with very recent observations (Patel et al., 2019).

| Expression of the smaller isoforms of dystrophin increases with development
Low levels of Dp427 protein were observed at early and mid-fetal stages of human brain development and are consistent with a recent analysis of dystrophin expression at the transcriptional level F I G U R E 8 Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis highlights up-regulation of pro-inflammatory pathways in DMD astrocytes. List of down-regulated and up-regulated pathways with Log2Fold changes >2 and FDR < 0.05 that have been extracted from KEGG pathway analysis for each line and are enriched in both DMD52 (52) and DMD68 (68) astrocytes. Note enrichment of several proinflammatory pathways in both DMD astrocytes; glutathione metabolism is also significantly up-regulated; pathways involved in astrocyte function, such as calcium signaling, are highly dysregulated and pathways involved in synapses are significantly down-regulated (Doorenweerd, Mahfouz, et al., 2017b). In contrast, we found expres- showing full western blots and/or focusing on human neural tissues, calls for future in-depth characterization of the putative small dystrophin isoforms and their alternative splicing in the brain. Research on DMD has so far focused mainly on the muscle, but as it is increasingly extending to other human tissues, including brain, retina and heart, and to dystrophin role in development. Hence, antibody characterization will need to be further broadened to these tissues with a greater emphasis on the complexity of dystrophin isoforms, newly emerging ones and potential roles of dystrophin products, as well as careful analysis of potential cross-reactivities, as exemplified by our unexpected finding of MANEX1A reactivity with "non-dystrophin" protein(s) both in developing muscle and brain (Barnabei et al., 2015;Fujimoto et al., 2017;Kawaguchi et al., 2018;Morris et al., 1995;van den Bergen et al., 2014).
Together, the predominance and developmental regulation of small dystrophin isoforms in early fetal brains, and published studies showing association of mutations in Dp140 and Dp71 with brain comorbidities, is consistent with the hypothesis that these deficits are largely due to developmental abnormalities, and not purely progressive as in the muscle (Bagdatlioglu et al., 2020;Chaussenot et al., 2018;Hoogland et al., 2018;Muntoni et al., 2003). Given the plasticity of the young brain, it is conceivable they could be ameliorated by early intervention. Significantly, in the mdx mouse an antisense oligonucleotide treatment that induced expression of a truncated dystrophin form with some functional activity improved their emotional/cognitive responses (Relizani et al., 2017).
F I G U R E 1 0 Heatmap of genes involved in PI3K-AKT signaling and validation of differentially expressed genes by RT-qPCR. (a) Hierarchical clustering of differentially expressed genes (DEGs) shows dysregulation of key genes involved in PI3K-AKT signaling in DMD astrocytes, with a greater number of significantly up-regulated than down-regulated genes in both DMD52 and DMD68 astrocytes. (b-d) analysis of differentially-expressed genes by RT-qPCR is consistent with RNA-seq data, as exemplified by up-regulation of interleukin 6 (IL-6), vascular endothelia growth factor A (VEGFA) and the calcium-dependent enzyme, peptidyl-arginine deiminase 2 (PADI2) in DMD astrocytes  Imamura & Ozawa, 1998). The distinct dystrophin profiles between neural cell types is consistent with the findings that DMD mutations disrupting distinct isoforms may differentially impact brain function in DMD patients (Blake et al., 2002;Muntoni et al., 2003;Ricotti et al., 2016).
Another parameter that affected dystrophin expression was culture dimensionality, as differences in dystrophin staining were observed when hNSCs, neurons and astrocytes were grown in 3D as compared to 2D cultures. Nonetheless, the antibodies used allowed comparison of full-length dystrophin against all isoforms and further highlighted different pattern of expression among neural cell types.
Culturing astrocytes in 3D appeared to greatly increase dystrophin staining. Based on the western blot results, this most likely reflects expression of Dp71 in astrocytes. Given the known structural role of dystrophin in muscle, the changes observed here between 2D and 3D support an important structural role for dystrophin also in the nervous system.

| DMD mutations affect astrocyte phenotype and function
The DMD patients from whose cells the iPSCs were generated carried different mutations, with two of them anticipated to affect all isoforms, including Dp427, Dp140, Dp71 and Dp40 (DM67 and DM68), and one (DMD52) expected to affect the larger isoforms, but not Dp71 and Dp40 (Ferrari et al., 2020). Defects in Dp71 have been proposed to contribute to neural impairment in patients as well as in animal models (Chaussenot et al., 2018;Daoud et al., 2009). Most recently Patel et al. (2019) have shown that mutations affecting the full-length isoform are sufficient to disrupt homeostatic activity of astrocytes and affect neuronal health. This is consistent with our findings that there are not notable differences in astrocytic responses between the patient lines studied, all lacking a functional full-length isoform, but with short isoforms differently affected.
Although expression of classical astrocytic markers indicated that DMD-iPSCs could generate astrocytes, they presented morphological abnormalities that were independent of the dystrophin mutation carried. Furthermore, GFAP protein expression was significantly increased in all our DMD astrocytes, which is consistent with the upregulation of GFAP transcripts in DMD astrocytes also recently reported (Patel et al., 2019). Importantly, GFAP up-regulation is a feature of reactive astrocytes. Several studies in mouse (Annese et al., 2016;Nico et al., 2004) also suggested alterations in blood brain barrier astrocytes and reported a decrease in the expression of the 30 kDa form of AQP4, a water-selective membrane channels that plays a key role in maintaining water homeostasis in the brain (Frigeri et al., 2001;Nico et al., 2004). In our human DMD astrocytes we observed a shift from the short AQP4 isoforms, where expression was significantly lower than in control astrocytes, to a significant increase in ubiquitinated AQP4 (87 and 100 kb). Changes in astrocytic GFAP and AQP4 are likely to disrupt astrocyte function and responses in humans, hence impacting upon blood-brain barrier function as proposed in the mdx mouse brain and in the Dp71 null mouse retina (Frigeri et al., 2001;Giocanti-Auregan et al., 2016;Li et al., 2020). Interestingly, astrocyte abnormalities and a reduction in AQP4 and Dp71 levels have been reported in post-mortem studies of patients with idiopathic normal pressure hydrocephalus (Eide & Hansson, 2017). While defective cerebral perfusion in DMD patients has recently been reported, studies on blood-brain barrier function in these patients are still missing and will be valuable to further understanding of the DMD neural pathology (Doorenweerd, Dumas, et al., 2017).
As astrocytes play a crucial role in brain function, changes in their properties may have a substantial impact on neuronal homeostasis and response to damage (Belanger et al., 2011;Pekny & Pekna, 2014).
Recently, defects in glutamate clearance and changes in Ca 2+ homeostasis in DMD astrocytes carrying different mutations have been shown to contribute to neurotoxicity, which was alleviated through restoration of dystrophin function and defects in glutamate handling were mainly attributed to lack of Dp427 (Patel et al., 2019). The DMD astrocytes investigated in that study did not show differences in population doublings. In contrast, under basal conditions, astrocytes from the DMD iPSCs studied here displayed significantly higher BrdU incorporation and Ki67 expression than controls, though the total cell number at the time point investigated was not significantly different.
Hence the increase in metabolic activity detected by the MTT and ATP assays in DMD astrocytes under basal conditions may not be fully due to higher proliferative activity. Together, increased DNA synthesis and metabolic activity in DMD astrocytes as well as their raised GFAP expression and morphological changes are consistent with patient-derived astrocytes being in a more activated state than healthy astrocytes.
While analysis of glucose oxidation under basal conditions revealed an upward trend in DMD astrocytes as compared to control healthy ones, more variability between both healthy and DMD lines was observed in this parameter than in any other studied. In contrast, upon exposure to stressors, differences in cellular responses between DMD and healthy astrocytes were clearly evident in all lines. Stimulation with LPS/IFNy, commonly used to activate astrocytes, induced the expected decrease in soma size and an increase in proliferative activity in healthy astrocytes indicative of inflammatory activation (Anderson et al., 2014;Pekny & Pekna, 2014). This did not occur in DMD astrocytes, consistent with them being constitutively in a more reactive state, hence less responsive to LPS/IFNy stimulation than healthy ones. A systemic increase in mediators of inflammation has been reported in DMD patients and defects in blood-brain barrier function have been found in the mdx mouse (Nico et al., 2004;Pelosi et al., 2017). Astrocytes can play an important role in reducing neuroinflammation and its potential deleterious effects (Liddelow & Barres, 2017;Pekny & Pekna, 2014). Therefore, it is tempting to speculate that a chronic inability of DMD astrocytes to respond appropriately to inflammatory stimuli, which are possibly abnormally high due to astrocyte impaired control of blood-brain barrier tightness, may contribute to functional impairment in DMD brains (Frigeri et al., 2001;Giocanti-Auregan et al., 2016).
Furthermore, given increasing evidence of astrocyte heterogeneity (Anderson et al., 2014;Liddelow & Barres, 2017), it is also possible, and not mutually exclusive with the above proposition, that due to DMD mutations different types of astrocytes preferentially differentiate from DMD and healthy iPSCs. Further support to the hypothesis that mutations in dystrophin affect astrocyte function was lent by their response to oxidative stress induced by H 2 O 2 treatment. Lower concentrations of H 2 O 2 were required to induce a loss of viability in DMD astrocytes than in healthy astrocytes. This is consistent with increased vulnerability to oxidative stress reported in muscle biopsies and iPSCs from DMD patients (Jelinkova et al., 2019;Petrillo et al., 2017).
The fact that altered morphology and responses to damage reported in our study were shared by all the DMD astrocyte lines tested, but Dp71/Dp40 expression is maintained in DMD52 astrocytes, suggests that the larger isoforms affected in all lines, Dp427 and Dp140, and possibly Dp116, though not expressed at high levels in astrocytes, are responsible for these defects. In depth analysis of each of these isoforms will be required to dissect their specific role(s).
Together, our results suggest that normalization of astrocyte function should be considered one of the therapeutic targets in DMD patients. Interestingly, both responses of DMD astrocytes to noxious stimuli and the RNA-seq data suggest a role for dystrophin in modulating astrocyte metabolic activity. The altered responses in DMD astrocytes observed in our functional experiments are consistent with the significant up-regulation of GO-terms involved in cell metabolism, such as protein, peptide and amide transport. In addition, defects in metabolic pathways common to both DMD lines are highlighted by the MaREA analysis. Another evidence of altered homeostatic functions in both DMD52 and DMD68 is provided by the evidence of dysregulation of calcium and cAMP signaling pathways detected by KEGG pathway enrichment analysis. Furthermore, this analysis also highlights several pathways that could contribute to the pro-inflammatory phenotype in DMD astrocytes, as indicated by the GFAP protein up-regulation and EAAT1 down-regulation we have observed. This is consistent with reports that mimicking inflammation by exposing astrocytes to activated microglia or TNFalpha, an inflammatory cytokine released from activated microglia, results in EAAT1 down-regulation (Dumont et al., 2014;Takaki et al., 2012). Also PADI2, a calcium-dependent enzyme known to be expressed in the developing human brain, which has been implicated in a range of inflammatory and neurodegenerative diseases, is up-regulated in DMD astrocytes (Jang et al., 2013;Kin Pong et al., 2014;Wu et al., 2020). Finally, in agreement with results reported by Patel et al. (2019), we show here that GO terms involved in neuronal differentiation are down-regulated, suggesting that DMD astrocytes provide a less supportive environment to neurons and could be implicated in neurodevelopmental deficits. Altogether, our transcriptional analysis has shown that loss of dystrophin in astrocytes significantly impairs several key pathways, which could potentially lead to abnormal interactions of astrocytes with different cell types in the brain, such as neurones and endothelial cells of blood vessels, hence affect brain homeostasis and function. This will deserve in depth investigation.

| DMD astrocytes exhibit major transcriptional changes
While here we have focused on transcriptional and protein expression changes common to all DMD lines studied, which could underpin the functional defects observed, some significant differences in gene expression between DMD52 and DMD68 astrocytes were also noted. Future analysis of these differences will be crucial to shed light on the functions of different dystrophin isoform in the nervous system and in patients carrying different dystrophin mutations.

| Conclusions
Together, the predominance of the shorter dystrophin isoforms observed in the human developing brain is consistent with studies suggesting an association between mutations in the shorter isoforms and brain comorbidities that characterize a significant proportion of DMD patients.
However, the abnormal responses to damage in DMD astrocytes identified here appear to be mainly due to the larger isoforms. Importantly, this study has further highlighted that astrocyte function is affected in DMD astrocytes, hence this may contribute to the aforementioned comorbidities. Therefore, to ameliorate neural impairments in DMD patients, therapies should also target these cells. Significantly, given the complexity of dystrophin isoforms, precisely mapping their preferential expression in different human neural cell types and assessing their intracellular distribution and interacting proteins will be crucial to better understanding the role of dystrophin in the human brain.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are stored in local repositories and available from the corresponding author upon reasonable request.