Loss of function variants in HPDL impair human cortical development via alterations of mitochondrial function

TL;DR

This study shows that mutations in the HPDL gene disrupt brain development by affecting mitochondrial function and neurogenesis, leading to neurological disorders.

Contribution

The study reveals a novel role for HPDL in cortical development and mitochondrial function, linking genetic mutations to specific neurodevelopmental outcomes.

Findings

HPDL mutations impair respiratory chain supercomplex assembly and redox balance in neuroblastoma cells.

HPDL patient-derived neurons show premature neurogenesis and cortical organoid growth defects resembling microcephaly.

Mitochondrial dysfunction in HPDL mutants is associated with increased ROS and respirasome assembly defects.

Abstract

Human brain development is highly regulated by several spatiotemporal processes, which disruption can result in severe neurological disorders. Emerging evidence highlights the pivotal role of mitochondrial function as one of these fundamental pathways involved in neurodevelopment. Our study investigates the role of 4-hydroxyphenylpyruvate dioxygenase-like (HPDL) protein in cortical neurogenesis and mitochondrial activity, since mutations in the HPDL gene are associated with a childhood-onset form of hereditary spastic paraplegia characterized by corticospinal tract degeneration and cortical abnormalities. Starting from mutant neuroblastoma cells, we demonstrated that HPDL is important to respiratory chain supercomplex assembly and cellular redox balance. Moreover, RNA-seq studies revealed dysregulated pathways related to brain development. Generation of cortical neurons and organoids from HPDL patient-derived induced pluripotent stem cells exhibited premature neurogenesis at early differentiation stages, likely leading to depletion of cortical progenitors, as evidenced by decreased proliferation, slight increase of apoptosis, and unbalanced cortical type composition at later stages. Cortical organoids showed failure to grow at a normal rate, a feature highly reminiscent of the “microcephaly” observed in severe HPDL children. Mitochondrial morpho-functional characterization in mutant neurons confirmed disruption of OxPhos chain functionality in neuroblastoma knock-out model cells and HPDL mutant cortical progenitors also displayed defects in respirasome assembly and increased ROS generation rate. Treatment of mutant cortical cells with antioxidants and CoQ10 intermediates partially rescued premature neurogenesis depending on the mutational context, suggesting potential future personalized therapeutic strategies. Our findings reveal a critical role for HPDL in coordinating cortical progenitor proliferation, neurogenesis, and mitochondrial function, shedding light on a better understanding of the related clinical presentations.

Figures

A, B), with normal levels of complex II. The reduction in complexes I and IV seems to be partly explained by significant decreased levels of NDUFB8 and COXII (CI and CIV subunits, respectively) in KO compared to E.V. cells, while levels of CII (SDHA), CIII (UQCRC2), and CV (ATP5A) subunits remained unchanged, excluding a general decrease of mitochondrial proteins expression (Fig. [1D](#Fig1)). In addition, we measured a reduction in enzymatic activities of complexes I, III, and IV (Fig. [1](#Fig1)) in HPDL KO cells and impaired oxygen consumption with reduced basal and ATP-coupled respiration

Gene Ontology analysis in HPDL mutant cortical progenitors and reduced size of HPDL organoids.A GO analysis of DEGs in HPDL vs Control neural progenitor cells (DIV16) showing the TOP 30 categories in biological processes (BP), and the number of gene counts for each, sorted for False Discovery Rate (FDR), with genes clustering in categories such as *Forebrain Development*, *Neuron Differentiation*, and *Cell-Cell Signaling*. B Volcano plot showing up- and down-regulated genes based on transcriptomics performed in early cortical cultures (DIV16) after comparison of four HPDL mutants versus two c

Premature neurogenesis in HPDL cortical progenitors.A Representative confocal images of Nestin and TUBB3 (beta3-tubulin) immunostaining in DIV16 cortical cultures deriving from controls (CTRL-1 and -2) and different HPDL patients (Patient 1–4). Nuclei are stained with DAPI. Scale bar: 25 μm. B, C Increase of TUBB3 mean fluorescence in all HPDL lines carrying pathological variants and decrease in Nestin fluorescence in Patients 1 and 2, expressed as fold change (FC) after normalization to control lines (CTRLs). Immunofluorescence signals were quantified and represented in a scatter plot. D Quan

A, B). Finally, immunostaining revealed no difference in differentiating progenitors (Supplementary Fig. 6C) together with a slight increase in apoptosis in DIV30 HPDL cortical cultures (Fig. [4F, G](#Fig4)), mostly concentrated in TUBB3-negative cells (Fig. [4H](#Fig4); Supplementary Fig. 6D). Collectively, our results point to premature neurogenesis in HPDL cortical cultures, leading to overall increase of deeper layer neurons combined with a decrease in the number of mitotic cells and slight increase of apoptotic cells. Notably, an increase in apoptosis has been recently reported also in *H

Full text

Introduction

In the growing molecular scenario of the Hereditary Spastic Paraplegias (HSPs) [1, 2] with approximately 90 different genes identified to date [3, 4], pathogenic variants in the HPDL gene, encoding the protein 4-HydroxyPhenylpyruvate Dioxygenase-Like, have been documented in approximately 60 families worldwide [5] and present with a broad spectrum of clinical manifestations and different degrees of severity [6]. Despite the rapid identification of individuals carrying biallelic HPDL mutations, the specific biological function of HPDL remains largely unknown. In vivo studies exploring HPDL-related conditions have faced several limitations. Hpdl^−/−^ mice exhibited early lethality and reduced brain sizes with cortical apoptosis and epilepsy [7], whereas zebrafish knock-down models showed limited motor response impairments [8]. Recent investigations suggested that HPDL could convert the tyrosine catabolite 4-hydroxyphenylpyruvate to 4-hydroxymandelate (4-HMA), to synthesize the coenzyme Q_10_ (CoQ_10_) precursor 4-hydroxybenzoate (4-HB) [9]. Even more recently, administration of 4-HMA and 4-HB to Hpdl^−/−^ mice, showed a better survival and reversion of neurodegenerative processes. Furthermore, 4-HB treatment improved the neurological manifestations of a patient with progressive spasticity due to biallelic HPDL variants [10]. The core hypothesis of this work is the crucial role that HPDL could play in brain development, resulting in a broader spectrum of neurological disorders in case of functional impairment. To investigate this assumption, we conducted in vitro experiments using HPDL knockout (KO) neuroblastoma cells (SH-SY5Y) and bi- and tridimensional cerebral cortical cultures differentiated from HPDL patient-derived induced pluripotent stem cells (iPSCs) [11, 12]. In our hands, HPDL seems critical for regulating the balance between proliferation and neurogenesis of cortical progenitors. Additionally, mutant neuroblastoma and cortical cultures exhibited mitochondrial dysfunction and increase of reactive oxygen species (ROS), highlighting the crucial role of HPDL in human neurodevelopment. Pharmacological treatments with antioxidants and 4-HB partially rescued anticipated neurogenesis at early stages, posing the basis for future personalized preclinical studies.

Results

HPDL KO neuroblastoma cell lines display mitochondrial functional impairment

To delve into the molecular function of HPDL and related etiopathogenesis, we generated a HPDL KO SH-SY5Y line (c.318delC; p.Val107PhefsTer24) via CRISPR/Cas9 technology (Fig. 1A). The mutation obtained resulted in absence of full length HPDL protein as shown by Western Blot (WB) analysis, whereas specific signals were present in cells transfected with empty vector (from now E.V. cells) or wild type (WT) (Fig. 1B). Given the known mitochondrial localization of HPDL, we assessed the expression levels and assembly state of the respiratory chain complex (RC) enzymes by Blue Native-PAGE (BN-PAGE). Formation and/or stability of mitochondrial respiratory chain supercomplexes (RCSs) containing complexes I, III_2_, and IV, dimeric complex V, and individual complexes III, IV, and V, appeared reduced in KO cells (Fig. 1C; Supplementary Fig. 1A, B), with normal levels of complex II. The reduction in complexes I and IV seems to be partly explained by significant decreased levels of NDUFB8 and COXII (CI and CIV subunits, respectively) in KO compared to E.V. cells, while levels of CII (SDHA), CIII (UQCRC2), and CV (ATP5A) subunits remained unchanged, excluding a general decrease of mitochondrial proteins expression (Fig. 1D). In addition, we measured a reduction in enzymatic activities of complexes I, III, and IV (Fig. 1) in HPDL KO cells and impaired oxygen consumption with reduced basal and ATP-coupled respiration rates (Supplementary Fig. 1D). In spite of this, the levels of total CoQ_10_ increased in KO cells (Supplementary Fig. 1E) and no loss of mitochondrial number was detected, as shown by absence of mtDNA copy number reduction in HPDL KO cells (Supplementary Fig. 1F), whereas ROS production appeared increased in stressed cells (Fig. 1E), indicating that HPDL ablation impaired OxPhos metabolism. Bulky RNA-seq analysis in E.V. and mutant cells showed a set of 3052 differentially expressed genes (DEGs) (Supplementary Fig. 2A, B). Interestingly, a profound effect was exerted on metabolism, with significant enrichment in genes related to oxidative metabolism (Fig. 1F). Gene Ontology (GO) analysis defined that most of DEGs belonged to categories related to brain development, structure, and function (Supplementary Fig. 2B). In addition, several genes involved in extracellular matrix organization appeared to be upregulated, whereas genes working in synaptic function were mostly downregulated (Supplementary Fig. 2C).Fig. 1. Generation of HPDL KO SH-SY5Y cell line and biochemical analysis.A Sanger sequencing electropherograms showing predicted CRISPR-Cas9 cut site in wild type HPDL sequence (from E.V. SH-SY5Y cells) and c.318delC mutation in after cutting (in KO cells). B WB analysis shows the presence of full length HPDL protein in WT and E.V. SH-SY5Y cell lines and absence in HPDL KO SH-SY5Y cells. C BN-PAGE blots exhibiting the decrease in the levels of RCSs, individual complexes IV (CIV), and dimeric complex V, while no changes are found in complex II (CII). Bar plots indicate the significant decrease of RCS(CI), RCS(CIII_2_), and RCS(CIII_2_)/CIII_2_ and the proportional increase of CIII_2_ in HPDL KO SH-SY5Y cells compared to E.V. ones. All data in bar plots are represented as mean ± SD (N = 3 for each line). (*), aspecific band. All gels are cropped from the original one. D Gel images of loading control and corresponding bar plots show the significant reduction of NDUFB8 and COXII, while no changes were detected in the levels of SDHA, UQCRC2, and ATP5A. All data in bar plots are normalized to the mitochondrial protein VDAC1 and represented as mean ± SD (N = 3 for each line). E Bar plots show comparison of ROS species in basal and oxidative (OX) stress conditions, demonstrating an increase in OX stress conditions in HPDL KO compared to E.V. lines. All data in bar plots are represented as mean ± SD (N = 12 for each line). F Volcano plot showing dysregulated expression of genes in HPDL KO compared to E.V. cells. In evidence (purple), genes involved in oxidative metabolism, such as response to oxygen levels, reactive oxygen species biosynthetic process, and NAD metabolic processes.

HPDL patient-derived neuronal lines show premature neurogenesis

Leveraging on previous iPS line generation [12], we compared immunocytochemical and transcriptomic profiles of cerebral cortical tissue differentiated from iPS lines deriving from two healthy donors (gold standard ACS1019, from now on CTRL-1, and CTRL-2, reprogrammed from a healthy adult female) and four SPG83 patients carrying different mutations in the HPDL gene (Patients 1–4). Supplementary Table 1 details on the lines used in this study. Fine characterization of new iPS lines is reported in Supplementary Fig. 3.

Control and HPDL cells were differentiated following established protocols [13, 14], producing cortical neural progenitors after 16 days of differentiation (DIV16; Supplementary Fig. 4). Transcriptomic analysis comparing DIV16 control vs mutant cortical cells denoted 785 DEGs and GO analysis implied deregulation of fundamental pathways involved in neurodevelopment, as seen in HPDL KO SH-SY5Y (Fig. 2A, B). Notably, GO categories related to regulation of neurogenesis were the most significant in the analysis with evident upregulation of neurogenesis-related genes such as NEUROG2, NEUROG3, and BCL6 (Fig. 2A, B; Supplementary Fig. 5). Genes related to cortical development and WNT pathway, known to increase expression and operate fundamental regulation in the process of cortical neurogenesis [15–17], also appeared to be upregulated, while genes related to FGF pathway, more linked to proliferative state [18], seemed to be mostly downregulated (Fig. 2A, B; Supplementary Fig. 5).Fig. 2. Gene Ontology analysis in HPDL mutant cortical progenitors and reduced size of HPDL organoids.A GO analysis of DEGs in HPDL vs Control neural progenitor cells (DIV16) showing the TOP 30 categories in biological processes (BP), and the number of gene counts for each, sorted for False Discovery Rate (FDR), with genes clustering in categories such as Forebrain Development, Neuron Differentiation, and Cell-Cell Signaling. B Volcano plot showing up- and down-regulated genes based on transcriptomics performed in early cortical cultures (DIV16) after comparison of four HPDL mutants versus two control counterparts. Highlighted in purple, several DEGs involved in fundamental pathways for brain development (492 up- and 293 downregulated), such as WNT (such as WNT3A, WNT7A, WNT11, LEF1, FZD7, AXIN2), NOTCH (with high expression of master genes like NEUROGENIN-2 and -3, in blue), cortical development-specific genes (in green, like OTX1, FOXP2, and LMX1A/B), and FGF (in red). C Representative confocal images of Control and Patient 1 cortical organoids stained for neural progenitor marker Nestin and neuronal specific marker TUBB3. Severe size reduction is striking. Scale bar: 100 μm. D Organoid growth was followed at different time points and displayed via line plot, showing clear reduction in SPG83-derived cultures. All points are represented as mean ± SEM (N > 3 for each line and time point).

To corroborate these data, we differentiated CTRL and HPDL iPSCs in 3D cortical organoids [19] and monitored cortical organoid growth at DIV5, 10, and 35. HPDL mutant cortical organoids failed to thrive, showing significant reduced growth compared to controls (Fig. 2C). In 3 out of 4 mutant lines, organoids only doubled the size within the analyzed timeframe, whereas CTRL organoids displayed an almost 5-fold increase (Fig. 2D). Notably, this phenotype is highly reminiscent of the most severe NEDSWMA-phenotype found in children carrying biallelic HPDL variants [7, 20], and exhaustion of proliferative cortical progenitor pools by means of premature neurogenesis (possibly associated with an increase in apoptosis) constitutes one of the classical pathological mechanisms in models of “microcephaly” [21–25].

To confirm the hypothesis of increased neurogenesis, we stained control and HPDL mutant adherent (2D) cortical cultures at DIV16 for neural progenitor (Nestin) and neuronal (TUBB3) specific markers. In line with the typical time course of the used protocol, the majority of the cells differentiated from CTRL lines were Nestin-positive neural progenitors at this stage, with scattered isolated foci of TUBB3-positive differentiated neurons (Fig. 3A). Conversely, cortical cultures differentiated from all HPDL lines displayed significant increase of neuronal population (Fig. 3B), together with reduction of Nestin-positive progenitors in two out of four cases (Fig. 3C). In addition, we detected a decreasing trend in the shared area of TUBB3 and Nestin signals (significant for both Patient 1 and 2; Fig. 3D), excluding the possibility that cortical cells remain blocked in a hybrid progenitor-neuron state, in which committed progenitors initiate differentiation without fully exiting from proliferation, ultimately validating occurrence of premature neurogenesis in mutant cells. Collectively, data coming from different experimental approaches support premature neurogenesis in HPDL cortical cultures.Fig. 3. Premature neurogenesis in HPDL cortical progenitors.A Representative confocal images of Nestin and TUBB3 (beta3-tubulin) immunostaining in DIV16 cortical cultures deriving from controls (CTRL-1 and -2) and different HPDL patients (Patient 1–4). Nuclei are stained with DAPI. Scale bar: 25 μm. B, C Increase of TUBB3 mean fluorescence in all HPDL lines carrying pathological variants and decrease in Nestin fluorescence in Patients 1 and 2, expressed as fold change (FC) after normalization to control lines (CTRLs). Immunofluorescence signals were quantified and represented in a scatter plot. D Quantification of shared area between TUBB3 and Nestin signal shows a decrease in neural progenitors derived from Patients 1 and 2 compared to CTRL. B–D All points are represented as mean ± SEM (N > 3 for each line).

Keeping in mind the characteristic “inside-out” pattern typical of neocortical neurogenesis, we performed immunostaining on iPSC-derived cortical neurons (DIV30), expecting enrichment for specific markers of deeper cortical layers (CTIP2 and TBR1, for layers 5 and 6, respectively). We observed an increase of CTIP2-positive layer 5 neurons (Fig. 4A, B) and a correlated drop in TBR1-positive layer 6 neurons (Fig. 4A, C) in cultures derived from Patient 2, Patient 3, and Patient 4. Apparently, the increase in neurogenesis occurred even earlier in cortical cultures from Patient 1, displaying a dramatic increase of layer 6 neurons and a decrease of layer 5 neurons (Fig. 4A–C). Premature neurogenesis was further corroborated by a significant reduction of mitotic cells (Fig. 4D, E) and also confirmed by transcriptomic analyses in the same DIV30 mutant cells (Supplementary Fig. 6A, B). Finally, immunostaining revealed no difference in differentiating progenitors (Supplementary Fig. 6C) together with a slight increase in apoptosis in DIV30 HPDL cortical cultures (Fig. 4F, G), mostly concentrated in TUBB3-negative cells (Fig. 4H; Supplementary Fig. 6D). Collectively, our results point to premature neurogenesis in HPDL cortical cultures, leading to overall increase of deeper layer neurons combined with a decrease in the number of mitotic cells and slight increase of apoptotic cells. Notably, an increase in apoptosis has been recently reported also in Hpdl^−/−^ mouse cerebral cortex [7].Fig. 4. Unbalanced cortical differentiation in HPDL neurons.A Increase in deeper layer neurons in DIV30 HPDL Patient-derived cortical cultures, as shown via immunostaining for layer 5 (CTIP2) and layer 6 (TBR1) cortical markers. B Cell counting on acquired images defined statistically significant increase of CTIP2-positive cells in neurons derived Patient 2 (16.50 ± 2.04%), Patient 3 (13.94 ± 1.37%), and Patient 4 (56.63 ± 3.10%), while decrease was observed in Patient 1-derived cultures (1.60 ± 0.43%), compared to CTRL counterparts (5.69 ± 1.18%). C TBR1-positive neurons were significantly increased in Patient 1 (9.67 ± 0.67%) and decreased in Patients 2 (0.098 ± 0.029%), 3 (1.64 ± 0.093%), and 4 (0.25 ± 0.091%) derived from neurons compared to CTRL counterparts (5.57 ± 1.18%). D, E Reduction of proliferation assessed via positivity for mitotic marker phosphorylated Histone H3 (pHH3) in HPDL neurons compared to CTRLs. F–H Apoptosis is increased in HPDL mutant cultures derived from Patients 3 and 4, as assessed by positivity for total apoptotic marker cleaved Caspase-3 (cCASP3) and by cCASP3^+^ cells negative for TUBB3 marker observed in Patient neurons compared to CTRL ones. B–H All data were represented via scatter plots, showing mean ± SEM (N > 3 for each line). A, D, F Nuclei are stained with DAPI. Scale bars: 20 μm.

Reduction of mitochondrial activity correlates with hyperpolarization in HPDL mutant cultures

To validate OxPhos impairment seen in HPDL KO SH-SY5Y cells, we performed cytochemical staining for cytochrome oxidase (COX) activity in HPDL DIV30 neurons. Cortical tissues derived from control lines showed an intense and homogeneous signal, coherently with the well-known sustained requirement of oxidative phosphorylation in neuronal cells [26]. Conversely, HPDL-derived neurons displayed reduced staining, especially those derived from Patients 1 and 2 (Fig. 5A). No functional abnormalities in mitochondrial complex II activity were seen in cultures derived from all cell lines analyzed upon succinate dehydrogenase (SDH) cytochemistry (Supplementary Fig. 7A). Of note, RNA-Seq using the same DIV30 cultures also showed dysregulation of multiple genes related to mitochondrial function and regulation of redox state in HPDL mutant neurons (Fig. 5B).Fig. 5. Mitochondrial dysregulation, RNA-seq analysis, and ROS increase in HPDL cortical neurons and complex III analysis in HPDL neural progenitors.A SPG83 patient-derived cortical neurons (DIV30) display evident oxidative phosphorylation defects, as shown via Cytochrome-c oxidase histochemistry. Scale bar: 25 μm. B Volcano plot showing the dysregulated expression of FGF pathway, WNT pathway, cortical development, and mitochondria-related genes (such as CAT, SQOR, GSTT2B, PRDX2, and MCUB) in HPDL patient derived neurons (DIV30) compared to CTRL ones. C Representative confocal images showing the increase of oxidative stress in DIV30 Patient-derived vs control neurons, detected via analysis of mean fluorescence of DCFDA probe (in green) and relative quantification, depicted in the relative scatter plot (left side). Enhancement in levels of ROS was significant in all cells carrying HPDL variants. Together, assessment of mitochondrial membrane potential was determined via TMRM probe mean fluorescence (in red, quantified and depicted in the scatter plot at the right side). Decrease of mitochondrial potential was demonstrated in Patient 1-derived neurons, while hyperpolarization was shown in cells differentiated from the other mutant lines (Pt. 2, 3, and 4). All scatter plots show mean ± SEM (in both analysis N > 3 for each line). Scale bar: 20 μm. D Real-time live imaging of TMRM fluorescence in DIV30 control and HPDL mutant neurons for 7 min after treatment with oligomycin. The line plot shows TMRM mean fluorescence decrease in HPDL mutant neurons after acute Oligomycin treatment. All data are represented as intercept lines with a significant difference in slope between HPDL and CTRL conditions (N > 3 for each line). E Representative images of BN-PAGE, loading controls, and corresponding quantifications (in bar plots, represented as mean ± SD; N > 4 for each line) exhibiting the decreased ratio of RCS on CIII_2_ and corresponding increased ratio of CIII_2_CIV on CIII_2_ in HPDL cortical progenitors compared to CTRL counterparts. FC fold change, a.u. arbitrary units. All gels are cropped from the original one; UQCRC2 48 kDa, VDAC1 35 kDa. F Bar plot shows the only increased amount of total CoQ_10_ in Patient 1 neural progenitors compared to CTRLs (mean ± SEM; N > 3 for each line).

Since one of the most likely consequences of mitochondrial dysfunction consists in the increase of oxidative stress, we assessed intracellular ROS levels in DIV30 cortical cultures via the cell-permeant dye DCFDA. At the same time, we also evaluated the mitochondrial membrane potential with TMRM, a cell-permeant dye usually accumulating in active mitochondria. The experiments showed an actual increase in ROS production in all four mutant lines when compared to control neurons (Fig. 5C; Supplementary Fig. 7B), confirming high oxidative stress as a putative common hallmark in HPDL disease. Mitochondrial membrane potential was lower in cells derived from Patient 1 and higher in the other three cell lines (Fig. 5C; Supplementary Fig. 7B). To solve the apparent conundrum between low COX activity and mitochondrial hyperpolarization, we explored the possibility of active proton pumping in intermembrane space by complex V, working in the reverse direction [27, 28]. Following this hypothesis, oligomycin-mediated pharmacological blockage of complex V should dissipate mitochondrial membrane potential in HPDL-patient derived cells, instead of causing hyperpolarization as normally happening in normal cells due to OxPhos-driven protonic accumulation. Time-lapse experiments revealed a drop in TMRM signal in mutant cells upon oligomycin treatment, confirming hyperpolarization in HPDL-derived cultures was dependent on complex V reverse activity (Fig. 5D; Supplementary Movie).

Based on these results, we attempted to characterize mitochondrial morpho-functional properties also in cortical progenitors (DIV16), the first stage where we had observed a distinctive phenotype in HPDL cortical cultures. Ultrastructure analysis revealed that mitochondria from Patient 2 display a more globular shape compared to CTRL, further suggesting mitochondrial disorganization (Supplementary Fig. 8A). Most importantly, patient-derived cortical progenitors showed a significantly reduced assembly of RCS, consequent accumulation of CIII_2_CIV supercomplexes (Fig. 5E), and increased levels of CoQ_10_ in Patient 1 derived cells (Fig. 5F), similarly to HPDL KO neuroblastoma cells. Conversely, changes were not seen in cortical progenitors from other lines (with a not significant decreasing trend in Patient 2 and 3; Fig. 5F).

In addition, mitochondrial potential was depolarized in Patient 1 and hyperpolarized in Patient 4-derived cells already at early stages, while three out of four cell lines showed a significant increase in ROS production, consistently with data observed at DIV30 (Supplementary Fig. 8B). Pharmacological inhibition of F_1_F_0_-ATP synthase still induced a loss of potential in HPDL mutant cells deriving from Patient 2 and 4, while other lines were not affected (Supplementary Fig. 8C). Notably, mitochondrial hyperpolarization was also observed in mutant neuroblastoma cells (Supplementary Fig. 8D). Collectively, these data confirm the link between mitochondrial dysfunction, increase of oxidative stress, and regulation of neurogenesis in HPDL mutant cells.

Premature neurogenesis in HPDL cortical progenitors is rescued by pharmacological treatment in a mutation-dependent manner

Since our findings link early premature neurogenesis with a concomitant increase of ROS in HPDL mutant cortical cultures, we chose to test antioxidant treatment for putative phenotypic rescue. To this end, we treated controls and HPDL mutant lines with GSH-MEE, a blood-brain barrier (BBB)-permeable form of glutathione, or MitoTEMPO, a powerful mitochondrial specific ROS scavenger, for which antioxidant effect was checked in every line for each experiment (Supplementary Fig. 9A). Treatment with 4-HB was also included, as recently proven effective in Hpdl KO mice [10]. When we assessed possible reversion of altered neurogenesis by comparing levels of TUBB3 immunostaining signal, we observed heterogeneous results (Fig. 6A–D). While antioxidant treatment did not perturb the normal levels of neurogenesis in control cells, a significant increase in neuronal staining was instead observed after administration of 4-HB (Supplementary Fig. 9B). Among HPDL mutant lines, cortical cells differentiated from Patient 2 turned out to be the most responsive to antioxidant treatment, as both GSH-MEE and MitoTEMPO reduced anticipated neuronal production. Short-term administration of both MitoTEMPO and 4-HB treatments operated a partial phenotypic reversion in Patient 3-derived cells. Only 4-HB was efficient to decrease neurogenesis in cortical cultures differentiated from Patient 1-derived iPSCs, whereas no antioxidant treatment worked. No treatment appeared effective in Patient 4-derived cells. These results link oxidative stress and premature neurogenesis in some HPDL neuronal cultures, highlighting also the diverse sensitivities of patient-derived cortical cells to different pharmacological interventions, potentially related to subtle differences in the mechanisms of pathogenicity.Fig. 6. Rescue of increased neurogenesis with pharmacological treatment in SPG83-derived cortical progenitor cells.A–D Representative confocal images of neuronal marker TUBB3 in Patient derived neural progenitors (DIV16). HPDL mutant cells were treated for 4 days with antioxidants (GSH-MEE or MitoTEMPO), or 4-HB, and then compared with untreated counterparts (NT). A In Patient 1-derived cells, only administration of 4-HB showed a decrease in TUBB3 positivity compared to NT cells (rescue: 73.59 ± 24.45%). B Antioxidant treatment with GSH-MEE or MitoTEMPO significantly decreased TUBB3 fluorescence in Patient 2-derived progenitors (rescue: 35.95 ± 33.98% and 54.23 ± 10.99%, respectively). C TUBB3 positivity showed a decrease in neural progenitors derived from Patient 3 after treatment with MitoTEMPO or 4-HB (rescue: 29.61 ± 20.77% and 23.85 ± 36.92%, respectively). D In Patient 4-derived cells, none of the pharmacological treatments worked to ameliorate premature neurogenesis. A–D All data are expressed as fold change (FC) normalized to untreated controls and represented as mean ± SEM in a scatter plots (N > 3 for each line). Scale bars: 20 μm. E Schematic representation of the connection between OxPhos impairment and anticipated neurogenesis in CTRL and HPDL conditions. IMS InterMembrane Space, OMM Outer Mitochondrial Membrane, IMM Inner Mitochondrial Membrane, NEC NeuroEpithelial Cell, vRGC ventricular Radial Glial Cell, DN Deeper layer Neurons, CP Cortical Plate, SVZ SubVentricular Zone, VZ Ventricular Zone.

Discussion

Human brain development is a finely tuned spatiotemporal-regulated process that can lead to severe neurological disorders if perturbated. In the past few years, the fundamental role of mitochondrial activity as neurodevelopmental regulator has increasingly emerged [29–32]. In this scenario, SPG83 is a recently identified form of HSP, caused by mutations in the HPDL gene, in which spasticity could be accompanied by severe neurological symptoms. HPDL encodes a mitochondrial protein deemed to be involved in CoQ_10_ biosynthesis. However, its precise role in brain development remains unclear. Our research seeks to address part of this knowledge gap.

Our experiments with HPDL neuroblastoma cells showed reduced mitochondrial oxidative metabolism, with impaired formation of mitochondrial RCS and an increase of ROS levels in conditions of oxidative stress. Gene expression studies showed dysregulated genes involved in brain developmental pathways, strongly suggesting an important role of HPDL in cortical formation and consistent with the neurodevelopmental nature of the disease. From here, we extended our investigation to cortical neurons and organoids differentiated from SPG83 patient-derived iPS lines [12], revealing a strong pattern of premature neurogenesis already at early stages in 2D cultures (DIV16). A slight increase in neurogenesis as the one described here could seem almost negligible at a superficial analysis, but the accumulation of such an effect during the highly proliferative phase of cortical development could lead to an increasing depletion of neuroepithelial cells, limiting heavily the expansion of the cortical tissue. The presence of unbalanced neuronal production was also confirmed by: a) transcriptomics, showing upregulation in neurogenesis-related pathways; b) decrease in proliferation; c) unbalanced production of deeper layer neurons. Depletion of proliferative cortical progenitors due to premature neurogenesis represents a well-documented pathological mechanism in microcephaly. Accordingly, in vitro HPDL 3D-differentiation showed that patient-derived cortical organoids exhibited a reduced growth, reminding the most severe microcephalic cases reported in infants, even if in other cases microcephaly is not primary but associated with postnatal metabolic crisis [7, 20].

Another important modifier in this scenario could be constituted by cell death, considered to be a distinctive hallmark in numerous models of microcephaly [21, 22, 33–35]. In this case, given the limited extent of the phenomenon within mutant cultures, the slight raise in apoptosis at DIV30, although consistent with the murine phenotype [7], most likely reflects a secondary consequence of mitochondrial stress. Therefore, we could cautiously infer that efficacy of pharmacological inhibition of apoptosis for phenotypic reversion is doubtful. All these data suggest a crucial role for the HPDL protein in the maintenance of cortical progenitor population, regulating the balance between proliferation and neurogenesis, strictly required to build the complexity of the human brain. Notably, lack of Spatacsin, another HSP-related protein, leads to a similar phenotype in patient-derived cortical organoids [36].

The known mitochondrial localization of HPDL protein, along with the aforementioned results in KO neuroblastoma cells, prompted further investigations in HPDL cortical cultures, allowing several considerations. First, we confirmed that reduction in the assembly of respirasomes also occurs in HPDL mutant cortical progenitors, together with accumulation of CIII_2_CIV supercomplexes, a configuration for which assembly does not seem to be impaired in HPDL mutant cortical progenitors. Consistently, formation of CIII_2_CIV complexes is reported to be modulated by SCAF1, and independent from complete RCS assembly [37–39]. Notably, patients harboring mutations in the complex IV assembly gene SURF1 present Leigh syndrome [40], and SURF1 mutant cortical organoids display reduced size, mitochondrial dysfunction, and defective RCS assembly, similar to what we observe in HPDL 3D cultures. In spite of this, SURF1 mutations prevented dopaminergic progenitors to shift from glycolytic to the typical OXPHOS-oriented neuronal bioenergetics, thereby promoting a permanent proliferative state [41], eventually highlighting possible diversities that can emerge in the fine tuning of brain development.

CoQ_10_ itself has been described to work in RCS assembly [42], and HPDL has been reported to act in its alternative biosynthesis increasing 4-HB levels via production of 4-HMA [9, 10]. Together with an impairment in respiration, primary CoQ_10_ deficiencies, caused by pathogenic variants in COQ genes (PDSS1/2, COQ2-COQ9), present increased ROS production [43–46] and some of them were associated with HSPs [47, 48]. Notably, fibroblasts harboring HPDL mutations also showed impaired respiration and oxidative stress [5], but no CoQ_10_ reduction [43]. In addition, variability of CoQ_10_ levels in HPDL mutant cell lines (increased in mutant SH-SY5Y and Patient 1 and not significantly in other cultures) do not seem to correlate with the constant impairment of mitochondrial RCSs in the same cells, opening the possibility for a direct role of HPDL in supercomplex assembly.

Second, DIV30 HPDL mutant cortical cells display significant reduction of COX activity. To corroborate these results, we measured mitochondrial membrane potential in live cells via TMRM staining. While Patient 1-derived neurons showed a clear reduction of potential, hyperpolarization was present in other patient-derived HPDL neurons. A possible explanation of this latter finding could be related to the occurrence of a phenomenon called “oligomycin null-point” [49]: in case of fully functional OxPhos chain, oligomycin treatment will tend to hyperpolarize the mitochondrial membrane, while if mitochondrial potential is maintained by the reverse action of ATP synthase, oligomycin will induce depolarization. In our hands, time-lapse experiments showed that HPDL cortical cells followed this last paradigm, as mitochondrial hyperpolarization was dissipated by acute oligomycin-mediated complex V inhibition. Reverse activity of complex V seems to occur also in Patient 1-derived cells, suggesting that in this case the mitochondrial functionality is probably so compromised that the proton-pumping activity of complex V fails to re-establish normal membrane potential. Together, this set of results lend support to the link of OxPhos dysfunction with impaired neural developmental program. Data observed in cell models of Leigh-like syndrome linked to F_1_F_0_ ATPase defects, such as [50–52], display reduction of ATP synthesis rate with preserved ATP hydrolysis capacity, together with mitochondrial dysfunction [53], increased mitochondrial membrane potential [51], and reduced size in iPSC-derived cerebral organoids [54]. A third consideration emerging from our study points to mitochondrial dysfunction and increased ROS production as main hallmarks of HPDL cortical cultures, and major players in brain development. Perturbation to the physiological mitochondrial activity in both senses have been repeatedly reported to influence pro-neurogenic stimuli [29–32], and ROS have been described to accumulate in the newt and mouse brain from neurogenesis stage [55, 56]. For instance, defective glucose and glutamine metabolism leads to ROS increase in mouse FoxO3 KO progenitors, decreasing proliferative potential and leading to neurogenesis [57]. In addition, changes in mitochondrial structure increase ROS in cortical progenitors, leading to neurogenic commitment by NRF2-mediated induction of the gene BOTCH, an inhibitor of the Notch pathway. Notably, these effects can be counteracted by antioxidants, such as NAC and MitoTEMPO [29]. Collectively, we could speculate a direct physiological connection between HPDL function, supercomplex assembly, ROS production, and cortical neurogenesis (see schematic in Fig. 6E).

Manipulation of ROS production could therefore represent a valid therapeutic opportunity not only for SPG83, but also for other genetic forms of HSP, particularly those involving mitochondrial proteins [1, 3]. Compounds with strong antioxidant activity have been used to treat HSP animal models [58–60], but translation in clinical practice has been limited by oral bioavailability and BBB penetration. A known exception is idebenone, an FDA-approved ROS scavenger and CoQ_10_ analogue used for treatment of several neurological conditions [61], but not effective in CoQ_10_ deficiencies [62, 63]. In this view, we tried to rescue HPDL-associated pro-neurogenic phenotype with two water-soluble and cell-permeant compounds, namely GSH-MEE and MitoTEMPO, reported to induce functional recovery in neurological disease models [64–72]. In addition to antioxidants, we tested the CoQ_10_ precursor 4-HB, recently shown to efficiently rescue HPDL-associated phenotypes in null mice and in a single child carrying pathological variants in the gene HPDL [10]. These experiments, run at DIV16 stage, suggested for every patient a peculiar pattern and sensitivity to the three tested compounds, likely reflecting the diverse mutational background of the different donors. In particular, Patient 1-derived cortical progenitors displayed depolarized mitochondria and increased ROS production, clearly indicating OxPhos dysfunction, and exclusive efficacy of 4-HB in reverting the pro-neurogenic phenotype. Considering the uncorrelated increase of CoQ_10_ occurring in these and neuroblastoma cells, we are tempted to speculate a CoQ_10_ redistribution and mislocalization, a paradigm already described in literature [31, 50], for which Patient 1 cells could be the only to follow paradoxically the current HPDL working hypothesis in CoQ_10_ biosynthesis [9, 10]. Our experimental data collectively depict this line as the most affected among the four analyzed, also considering the strongest pro-neurogenic stimulus (with increased production of earlier generated layer 6 rather than layer 5 cortical neurons, as occurring in other lines). It is tempting to correlate the gravity of cellular phenotype with the severity of clinical features seen in donor patients [5], even if this seems too speculative at this stage. Intriguingly, the effect of 4-HB to increase mitochondrial function was present also in wild type fibroblasts [43] and SH-SY5Y parental cells (our lab, unpublished). Accordingly, treatment of wild type cortical progenitors with 4-HB in the present study resulted in increased neurogenesis, an effect previously reported upon increased mitochondrial respiration [29–32], whereas no effect on differentiation was seen in presence of antioxidants (summarized in Supplementary Table 2). Despite OxPhos assembly defects, reversed ATPase activity appears to buffer mitochondrial potential already at DIV16 in Patient 2- and Patient 4-derived cells, partially explaining inefficacy of 4-HB. Patient 2-derived cortical progenitors display the highest increase on ROS production, and the responsiveness to both antioxidants may suggest oxidative stress rather than CoQ_10_ as the main driving force underlying increased pro-neurogenic stimulus. In Patient 4-derived cells, reversion of complex V seems to work even better, generating a slightly hyperpolarized mitochondrial potential and keeping ROS levels comparable to control situation, partly clarifying inefficiency of the treatments. Notably, Patient 4, alike others carrying the p.Gly50Asp variant, presents milder symptomatology [56]. Finally, Patient 3-derived cortical cells do not show modifications in mitochondrial features, maybe reflecting the observed milder OxPhos dysfunction (from COX activity at DIV30). Nonetheless, increase of ROS is already present, possibly explaining the efficacy of treatments. The data obtained in these rescue experiments may enforce the need of characterization for each HPDL variant, posing the base for personalized treatment.

Nonetheless, given the homogeneity of cellular features in later stages, the differential sensitivity to the tested pharmacological interventions do not hinder the possibility that the same treatments could work in preclinical studies. Later aspects of cortical differentiation could follow different mechanisms in physiological and pathological condition compared to early neurogenetic stages, urging the necessity of detailed characterization of HPDL-associated phenotypes in mature cortical tissue. Restricting the phenotypic analysis to DIV30 constitutes an actual limitation of this study. Given that upper cortical neurons are not yet generated at this stage, it was not possible to determine whether the premature neurogenesis occurring in deeper layers is capable to influence generation and distribution of these populations. Future experiment will be required to address these knowledge gaps.

In conclusion, our data collectively point to the critical requirement of HPDL protein in maintaining the mitochondrial morpho-functional stability during neurodevelopment, supporting cortical progenitors in fine-tuning the balance between proliferation and neurogenesis (Fig. 6E). Hence, the increasingly close bond between neurodevelopment and oxidative metabolism could provide new insights into the pathogenesis of HSPs and other related neurological disorders, potentially offering novel therapeutic strategies for treating these diseases.

Materials and methods

Cell reprogramming

Human dermal fibroblasts (HDFs), obtained from patient skin tissue via punch biopsies, were amplified in HDF medium and reprogrammed as previously described [73–75]. From about 14 days onwards, iPS “islands” appeared, and medium was switched to StemFlex medium (A3349401, Thermo Fisher Scientific). iPS clones with uniform flat and round shaped morphology were picked in sterile conditions and propagated as single cell lines. After seven passages, all iPS clones were characterized as described [12, 74].

Cell culture

Neuroblastoma SH-SY5Y cells, iPSCs, neural progenitors, and neurons were cultured in standard conditions at 37 °C and 5% CO_2_.

SH-SY5Y were maintained in DMEM High Glucose (ECM0728L, Euroclone) containing 15% FBS (SIAL-FBS-SA; Sial), 1% MEM-NEAA (ECB3054D, Euroclone) and 1% Pen/Strep (SIAL-PEN/STREP, Sial). To prompt metabolic switching from glycolysis to respiration without affecting cell viability, SH-SY5Y cells were grown for 48 h in DMEM/F12 no glucose (PM150322, Elabscience), supplemented with 2 mM D-Glucose (16325, Riedel-de Haen), 15% FBS, 2mM L-glutamine (25030081, Thermo Fisher Scientific) and 100 µg/ml Pen/Strep (OxPhos medium).

ACS1019 iPSCs were bought from ATCC, Manassas, VA, USA. Human iPSCs were cultured on Geltrex coated 6-well plates in StemFlex medium, refreshing medium every other day. Cells were passed every 5–7 days with ReLeSR™ (100-0484, Stem Cell Technologies), following manufacturer instructions. Cells have been differentiated following the dual-SMAD inhibition protocol [13, 14]. Briefly, pluripotent cells were harvested at high confluence on Geltrex and cultured for 12 days in neural induction medium containing 100 nM LDN193189, 10 µM SB431542, and 2 µM XAV939 (72147, 100-1051, 72674 respectively, Stem Cell Technologies), splitted with Accutase (A1110501, Thermo Fisher Scientific) and replated on poly-D-lysine/laminin coated coverslips (A3890401, Thermo Fisher Scientific and L2020, Sigma, respectively). From day 16 onward, medium was switched to terminal differentiation medium containing 30 ng/ml BDNF (AF-450-02, PeproTech) to improve neuronal differentiation and maturation. Differentiated cells were harvested at day in vitro (DIV) 16 and DIV30 and processed for immunostaining (see below) or RNA extraction via RNeasy Plus kit (74134, Qiagen).

Treatments were performed with 1 mM glutathione monoethyl ester (GSH-MEE; 353905, Merck), 1 µM MitoTEMPO (SML0737, Merck), or 1 mM 4-hydroxy-benzoate (4-HB; 240141, Merck) added on neural progenitors at DIV12 and DIV14 [72, 76, 77]. Cells were fixed in PFA at RT for 12 minutes at DIV16 after a 4-days treatment and then processed for TUBB3 immunostaining.

Cortical progenitors were obtained from iPSCs following dual-SMAD inhibition protocol. At DIV12 cells were seeded as single drop at the concentration of 5 ×10^6^ cells/ml. The day after the medium was replaced with NSC medium (KO-DMEM/F12 (12660012, Thermo Fisher Scientific), 1% Pen/Strep, 2 mM D-Glucose, 2% StemPro Neural Supplement (A1050801, Thermo Fisher Scientific), 20 ng/ml FGF2 (AF-100-18B, Thermo Fisher Scientific), 20 ng/ml EGF (AF-100-15, Thermo Fisher Scientific)) and the cells were replaced at DIV20 with a density of 50.000 cells/cm^2^. The medium was changed every 2-3 days and the cells were splitted when they were almost confluent.

Organoid culture

Cortical organoids were generated with a slight modification of the classical SFEBq-based procedure [19]. Briefly, 9.000 iPSCs for well were plated in 96-well plates (83.3925.400, Sarstedt) in StemFlex medium with Y-27632 10 µM (ab120129, Abcam). The day after, we performed one wash with PBS and added organoid medium I, containing Glasgow’s MEM (11710035, Thermo Fisher Scientific), 1% Pen/Strep, 20% KnockOut Serum Replacement (10828010, Thermo Fisher Scientific), 1% MEM-NEAA, 1 mM Pyruvate (S8636, Sigma), 0.1 mM 2-mercaptoethanol (21985, Gibco), for 18 days, changing the medium twice a week. Cell aggregates were then transferred on ultra-low-adherent 24-well plates (662102, Greiner) in organoid medium II, containing DMEM/F12 – Glutamax (10565018, Thermo Fisher Scientific), 1% N2 (17502001, Thermo Fisher Scientific), 1% Chemically Defined Lipid Concentrate (11905031, Thermo Fisher Scientific), 1% Pen/Strep, until 35th day. Medium was changed once a week. For size quantification, organoids were imaged with Zeiss Axiovert 25 at DIV5 and Leica M205 FA at DIV10 and 35.

Cytochemical staining for cytochrome-c oxidase and succinate dehydrogenase activity

For cytochemical staining for the enzymatic activities of cytochrome-c oxidase (COX) and succinate dehydrogenase (SDH), neurons were processed using standard protocols [78]. After incubation, cells were rinsed twice with phosphate buffer, fixed 15 min with 4% paraformaldehyde in PBS, washed twice with PBS and then imaged with Zeiss Axiovert 25 microscope.

TMRM and DCFDA protocol

TMRM (T5428, Merck) and DCFDA (C6827, Thermo Fisher Scientific) assays were performed on neural progenitors or neurons following a single protocol. After removing the medium, the cells were washed one time with H-HBS solution (HBSS (14175095, Thermo Fisher Scientific) with HEPES (EMR152100, Euroclone) 1 mM pH 7.4). DCFDA 1 µM, Hoechst 33342 5 μg/ml, and TMRM 5 μg/ml were diluted in H-HBS solution and added on cells for 40 min at 37 °C in darkness. Finally, one wash in H-HBS solution was performed and the cells were observed with Zeiss LSM 900 confocal microscope. Then, oligomycin (04876, Sigma) 5 µM was added on cells and TMRM fluorescence was immediately acquired for 7 min.

Fluorescence quantification

Fiji software was used to analyze the mean fluorescence of TMRM, DCFDA, Nestin or TUBB3. In particular, images were thresholded keeping the values constant inside every differentiation and used to select ROIs. Then, the mean fluorescence of the selected marker was measured inside the ROI. Every measure was normalized on the CTRL line of every differentiation.

Mitochondrial purification and Blue Native-PAGE

Prior to the measurements, cells were cultured for 48 h with OxPhos medium. Blue Native-PAGE (BN-PAGE) followed by immunoblotting was performed to identify respiratory chain complexes bands. Mitochondrial enriched fractions were obtained following the protocol outline in Frezza et al. [79]. Mitochondria pellets were resuspended in an appropriate volume of Native Buffer (Invitrogen) with 4% Digitonin (Merck) to be at 10 µg/µl; and membrane proteins were solubilized during 1 hour in ice. After a 20 min centrifugation at 16.000 × g, the supernatant was collected and was added 5% G250 (Invitrogen) at one-fourth of the detergent concentration. About 35 μg of mitochondrial proteins were loaded in a 3–12% Bis-Tris gel (Invitrogen) as described elsewhere (NativePAGE Novex® Bis-Tris Gel System manual) and run first for 30 min at 150 V with Cathode Buffer Dark Blue, and then for 120 min at 150 V with Cathode Buffer light Blue. Mitochondrial complexes were transferred at 100 V for 4 h to a PVDF membrane using the BioRad transfer system. For WB we used the following antibodies: anti-NDUFB8 (1:1000, 4592210, Invitrogen), anti-SDHA (70KDa Fp subunit) (1:1000, 459200, Invitrogen), anti-Cox-II (MTCO2 Monoclonal Antibody 12C4F12) (1:1000, A6404, Invitrogen), anti-UQCRC2 (1:1000, ab14745, Abcam), and anti-ATP5A (1:1000, ab14748, Abcam). The secondary antibody used was peroxidase-conjugated goat anti-mouse (1:2000, 32430, Thermo Scientific). Images were acquired after the chemiluminescence reaction (LiteAblot Turbo, Euroclone) by the iBright FL1500 (Agilent) instrument. All original gels are shown in Supplemental Material. Quantifications of single bands were performed via GelAnalyzer 19.1 software.

Assays of respiratory chain enzyme activities

Prior to the measurements, cells were cultured for 48 h in OxPhos medium. Activities of mitochondrial respiratory chain complexes I, II, III, IV, and II + III were measured as described [80] and normalized to protein and citrate synthase (CS) activity using the Varian Cary 100Bio UV-Visible Spectrophotometer.

Statistical analysis

All data in the manuscript represent three or more independent experiments. We performed the statistical analysis using GraphPad Prism 9.0.0 software. All data were analyzed using either parametric or non-parametric methods, based on the normal distribution analysis. For parametric data, we additionally tested whether standard deviations were significantly different across groups. The significance among groups with 2 independent variables was determined using two-way ANOVA test, among 3 or more groups with 1 independent variable was measured with one-way ANOVA, Welch’s ANOVA, or Kruskal–Wallis tests, and between 2 groups with 1 independent variable was calculated with two-tailed unpaired Student’s t test with or without Welch’s correction. Statistical significance is reported as: *P-value ≤ 0.05, **P-value ≤ 0.01, ***P-value ≤ 0.001, or ****P-value ≤ 0.0001. The number of different acquisition fields, culture wells, and experiments, as well as statistical analysis and significance from each dataset are all comprehensively reported in Supplementary Table 3.

Supplementary information

Supplementary Information Supplementary Fig. 1 Supplementary Fig. 2 Supplementary Fig. 3 Supplementary Fig. 4 Supplementary Fig. 5 Supplementary Fig. 6 Supplementary Fig. 7 Supplementary Fig. 8 Supplementary Fig. 9 Supplementary Table 1 Supplementary Table 2 Supplementary Table 3 Supplementary Movie Original Western blots