Pluripotent stem cell-based screening uncovers sildenafil as a treatment for mitochondrial disease

Abstract

Mitochondrial disease encompasses inherited disorders affecting mitochondrial function. A severe and untreatable form of mitochondrial disease is Leigh syndrome (LS) causing psychomotor regression and metabolic crises. To accelerate drug discovery for LS, we screened a library of 5,632 repurposable compounds in neural cells from LS patient-derived induced pluripotent stem cells (iPSCs). We identified phosphodiesterase 5 inhibitors as leads, and prioritized sildenafil due to its clinical safety. Sildenafil corrected mitochondrial membrane potential defects and restored nervous system development pathways in LS neural cells, and normalized calcium responses in LS brain organoids. In small and large mammalian models of LS, sildenafil extended the lifespan and ameliorated disease phenotypes. Chronic off-label compassionate treatment with sildenafil in six LS patients improved their motor functions and their resistance to metabolic crises. These findings highlight the potential of iPSC-driven drug discovery and position sildenafil as a promising candidate for mitochondrial diseases.

Introduction

Mitochondrial disease refers to a spectrum of rare genetic conditions caused by dysfunction in the cell’s energy-producing organelles.¹ A severe form of mitochondrial disease is Leigh syndrome (LS, OMIM #256000), which is characterized by basal ganglia and brainstem necrosis leading to neurodevelopmental regression and muscle weakness. Early death in these patients is typically due to acute deterioration following metabolic crises triggered by events such as infections that place additional demands on the body's energy production.^2,3 LS can result from pathogenic gene variants in more than 100 genes in the nuclear DNA or mitochondrial DNA (mtDNA) that are involved in oxidative phosphorylation (OXPHOS).⁴ Commonly affected genes include the complex V gene MT-ATP6 (mitochondrially encoded ATP synthase subunit 6),⁵ the complex IV assembly factor gene SURF1 (SURFEIT1),⁶ and complex I genes such as NDUFS4 (NADH dehydrogenase [ubiquinone] iron-sulfur protein 4).⁷

Currently, there are no treatments for LS.⁸ One of the challenges in identifying disease-modifying therapies for LS is the limited availability of appropriate model systems.⁹ The difficulties in editing the mtDNA have hindered the establishment of models for mtDNA defects such as MT-ATP6 variants.¹⁰ Furthermore, although a pig model of LS due to SURF1 knockout (KO) has been developed showing severe neurodevelopmental defects and lethality in the first few days after birth,¹¹ Surf1 KO mice failed to replicate LS phenotypes and instead exhibited increased longevity.^12,13 The most widely used model of LS is the homozygous Ndufs4 KO mouse characterized by progressive encephalopathy, growth retardation, motor regression, cardiomyopathy, and premature death.^14,15 This models has been used to propose potential treatment strategies against LS, including antioxidants,^16-18 rapamycin,¹⁹ interferon-gamma-targeting therapies,²⁰ cannabidiol,²¹ or hypoxia.^22,23 However, only few of these strategies have been tested in clinical application in LS patients, and no successful results have been obtained.

One approach for establishing models of LS suitable for drug discovery studies is the use of patient-derived induced pluripotent stem cells (iPSCs). iPSCs allow the generation of disease-relevant two-dimensional (2D) or three-dimensional (3D) cellular models containing patient-specific nuclear or mitochondrial defects.²⁴ In LS, such models have been instrumental in uncovering pathological mechanisms. Identified phenotypes include neuronal outgrowth defects,²⁵ altered calcium homeostasis,^26,27 and glutamate toxicity in 2D neuronal cultures,²⁸ as well as impaired cortical development and neuromorphogenesis in 3D brain organoids.^25,29,30 Despite these advances, large-scale high-throughput drug screens in iPSC models of mitochondrial disease have yet to be performed.

We previously demonstrated that iPSC-derived neural progenitor cells (NPCs) are an effective drug discover platform for mitochondrial diseases.²⁷ LS NPCs with MT-ATP6 defects exhibited abnormal hyperpolarization of mitochondrial membrane potential (MMP), a feature that can be exploited for high-throughput screens using high-content microscopy.³¹ Here, we leveraged on this MMP phenotype to screen a library of 5,632 repurposable drug candidates in NPCs from LS patients harboring MT-ATP6 variants. We identified phosphodiesterase type 5 (PDE5) inhibitors as lead compounds capable of ameliorating the MMP defect. Among these, we prioritized sildenafil given its known safety profile.³² Sildenafil rescued the neurodevelopmental disease signature, promoted neuronal outgrowth, and normalized calcium homeostasis in 2D and 3D human neuronal models of LS carrying different pathogenic gene variants. In vivo, sildenafil extended the lifespan of Ndufs4 KO mice and of SURF1 KO piglets, leading to functional clinical improvement. Mechanistically, the action of sildenafil appeared to be mediated through modulation of cGMP-dependent protein kinase 1 (PRKG1). Lastly, individualized off-label compassionate treatments with sildenafil in six patients with LS carrying different MT-ATP6 variants resulted in improved muscle strength and endurance and resistance to metabolic crises. The data collectively suggest a potential use of sildenafil as a repurposable drug for individuals with the mitochondrial disease Leigh syndrome.

Results

Small molecule repurposing screen in LS NPCs identifies PDE5 inhibitors

We previously detected abnormally hyperpolarized MMP in LS NPCs carrying the m.9185T>C variant in the gene MT-ATP6, which encodes the “a” subunit of complex V involved in the release of the proton gradient across the inner mitochondrial membrane.²⁷ As MMP levels can be used as a phenotypic readout for compound screening,³¹ we set out to determine whether such a phenotype could be robustly observed across LS NPC lines carrying different MT-ATP6 variants. We generated NPCs from seven LS patient-derived iPSCs carrying different MT-ATP6 variants (three m.9185T>C, one m.8993T>C, two m.8993T>G, and one m.9176T>G) (Figure S1A).^27,33-36 For controls, we used NPCs from eight iPSC lines derived from healthy individuals, including two patient mothers who carried MT-ATP6 variants at low level of heteroplasmy (less than 2 % in iPSCs and NPCs) (Figure S11H). LS NPCs showed disrupted ATP6 protein expression (Figure S1B) and defective bioenergetics (Figure S1C), but expressed NPC markers similarly to control NPCs, demonstrating physiological neural commitment (Figure S1D-E). They also retained the mtDNA variants at the same level of heteroplasmy as the parental iPSCs (Figure S11H). Importantly, in agreement with our previous observations,²⁷ all LS NPCs displayed MMP hyperpolarization when compared to control NPCs (Figure 1A).

Figure 1. Compound screen in LS NPCs leads to the identification of the PDE5i sildenafil.

(A) High-content analysis (HCA) quantification of mitochondrial membrane potential (MMP) in neural progenitor cells (NPCs) from LS patients (ATP6_1, ATP6_2, ATP6_3, ATP6_4, ATP6_5, ATP6_6, ATP6_7) and healthy controls (CTRL_1). Dots represent mean values per well out of n=3-5 independent experiments. ****p<0.0001; ordinary one-way ANOVA with Dunnett’s multiple comparison. (B) Schematic of drug screen approach. (C). MMP screening of 5,632 repurposable compounds in LS NPCs (ATP6_2). Scatter dot plot illustrates the percentage of MMP depolarization (TMRM signal) against the percentage of cell count (Hoechst signal) after treatment for 16 h with 5 μM of each compound in LS NPCs (blue dots) normalized to LS NPCs with either 5 μM FCCP+AA (purple dots) or DMSO (orange dots). Dots represent individual wells. Black dots: untreated control NPCs (CTRL_1); red dots: LS NPCs treated with PDE5 inhibitors (PDE5i) 1 (T-0156) and PDE5i 2 (vardenafil hydrochloride). (D) Dose-dependent effect of PDE5i sildenafil on the MMP of different LS NPCs compared to LS NPCs with DMSO only. Dots represent mean values per well out of n=4 independent experiments; *p<0.05, ***p<0.001, ns (not significant); ordinary one-way ANOVA with Dunnett’s multiple comparison. (E) Sildenafil effect on the MMP of LS NPCs. Dots represent individual cytofluorimetry measurements using 10 μM sildenafil for 16 h out of n=3 independent experiments. **p<0.01, ***p<0.001, ****p<0.0001; two-tailed paired t-test. (F) Complex V (CV) activity normalized to citrate synthase (CS) activity. Dots represent a pool of 10-15 million cells in n=3 independent experiments of control NPCs (CTRL_1, CTRL_3) and LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) using 10 μM sildenafil for 16 h. ***p<0.001, ****p<0.0001, ns; unpaired two-tailed t-test. (G) Luminescence-based ATP content quantification. Dots represent biological replicates out of n=3 independent experiments in control NPCs (CTRL_1, CTRL_3) and LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) grown in glucose-free galactose medium with 10 μM sildenafil for 16 h. **p<0.01, ns; paired t-test. (H) mtDNA copy number in n=3 independent experiments in control NPCs (CTRL_1, CTRL_3) and LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) using 10 μM sildenafil for 16 h. *p<0.05, ***p<0.005, ns; paired t-test. (I) NAD⁺/NADH ratio in control NPCs (CTRL_1, CTRL_2, CTRL_3, CTRL_4) and LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) using 10 μM sildenafil for 16 h. *p<0.05, ns; unpaired two-tailed t-test. (J) Left, machine learning analysis of the MIC60 labeling pattern of stimulated emission depletion (STED) microscopy images in control NPCs (CTRL_1) and LS NPCs (ATP6_2) with 1 μM sildenafil for 24 h. Quantification shows the probability of the pattern to be similar to that of untreated healthy control NPCs (P[healthy]). Right, representative STED images. Scale bar: 500 nm. ****p<0.0001; Kruskal-Wallis with Dunn’s multiple comparison test.

We then applied a well-curated library of 5,632 repurposable molecules³⁷ to possibly modulate the MMP of LS NPCs. We used live-cell high-content analysis (HCA) based on the intensity of the MMP-dependent fluorescent dye tetramethylrhodamine methyl ester (TMRM)³¹ and Hoechst 33342 for counter staining (Figure 1A, Figure S2A-B). All compounds were resuspended in dimethylsulfide (DMSO) at a maximal concentration of 0.05 % DMSO, which was not toxic on LS NPCs (Figure S2C). One control NPC line was included for baseline MMP value, after having confirmed that different control NPC lines displayed comparable MMP (Figure S2D). Using the MMP depolarizing effect of the mitochondrial uncoupler carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) in combination with the mitochondrial complex III inhibitor antimycin A (AA)³¹ (Figure S2E-F), we establish the Z’ factor³⁸ of our assay to be about 0.68 - 0.85 across all screening plates (Figure S2G). We identified 5 μM as the optimal screening concentration, achieving a hit rate of ~5 % with overall low toxicity (Figure S2H) with high data reproducibility (Figure S2I).

We classified the 5,632 compounds based on their depolarizing effect on the hyperpolarized MMP of LS NPCs (MMP inhibition) relative to the standard deviation (SD) of DMSO-treated cells (SD=6.51 %). We identified 767 mild uncoupler molecules that depolarized the MMP by more than two times the SD (13.0 - 19.5 % inhibition), 520 intermediate uncoupler molecules that depolarized the MMP by more than three times the SD (19.5 - 32.5 % inhibition), and 187 strong uncoupler molecules that depolarized the MMP by more than five times the SD (Figure 1C). We excluded toxic compounds based on the cell count of DMSO-treated cells (cut-off: 29.86 %). We focused on mild uncouplers, which were sufficient to depolarize the MMP of LS NPCs to values similar of those of control NPCs (Figure 1C). Among those, we identified two PDE5 inhibitors (PDE5i): T-0156 and vardenafil hydrochloride. Both compounds led to dose-dependent depolarization of MMP in LS NPCs without loss of viability (Figure S3A-D).

Since our previous small-scale proof-of-concept screen in LS NPCs already highlighted the beneficial effect of the PDE5i avanafil,²⁷ we searched the literature for applications of PDE5i in pediatric diseases. The PDE5i sildenafil emerged as a strong candidate, given its safety profile in children treated for pulmonary arterial hypertension (PAH) or lymphatic malformations.^39-42 Sildenafil led to a dose-dependent depolarization of MMP in LS NPCs without inducing toxicity (Figure 1D, Figure S3E-G) with a half-maximal inhibitory concentration (IC₅₀) of approximately 3 μM after 16 hours of exposure (Figure S3H). The MMP normalization occurred similarly in several LS NPC lines carrying distinct MT-ATP6 variants, resulting in an approximate MMP reduction of 16 % (MMP of sildenafil-treated compared to DMSO-treated LS NPCs: mean = 83.98 %, range = 54.52 % - 98.27 %) (Figure 1E). Sildenafil did not affect the MMP of control NPCs, quantified with HCA or cytofluorimetry (Figure S3I).

We next assessed the bioenergetic effects of sildenafil in LS NPCs. Sildenafil treatment did not lead to a measurable improvement of complex V enzymatic activity (Figure 1F) or of complex V assembly structure^43,44 (Figure S4A-C). The protein levels of known mitochondrial modulators were also not significantly impacted by sildenafil (Figure S4D-E). Nonetheless, sildenafil increased the intracellular ATP concentration in LS NPCs grown in glucose-free galactose medium (Figure 1G) and restored the levels of mtDNA copy number (Figure 1H). Furthermore, the alteration of NAD⁺/NADH ratio in LS NPCs was normalized using sildenafil (Figure 1I).

Since MT-ATP6 defects can disrupt mitochondrial cristae morphology,⁴⁵ we analyzed the distribution of mitochondrial crista junctions using the cristae junction regulator MIC60 as a proxy with stimulated emission depletion (STED) microscopy.⁴⁶ We inspected the MIC60 labeling pattern of 4,000 STED images with a machine learning approach by training a neural network classifier to distinguish control NPCs (P[healthy]-score = 1) from LS NPCs (P[healthy]-score = 0). Using images not used for training, we detected alterations of the localization of MIC60 fluorescence signals in LS NPCs that were partially reversed by sildenafil (Figure 1I, Figure S3J-K). Manual quantification of approximately 300 STED images yielded similar results (Figure S3L) and highlighted a peripheral distribution pattern of MIC60 in LS NPCs that was normalized by sildenafil (Figure S3M-N). Collectively, these data suggest sildenafil as a potential treatment strategy for LS capable of improving key mitochondrial phenotypes.

The PDE5i sildenafil ameliorates the LS disease signature by modulating neuronal development pathways

To gain insight into the mechanisms of action of sildenafil in LS, we selected four LS NPC lines (one for each of the MT-ATP6 variants: m.9185T>C, m.8993T>C, m.8993T>G, and m.9176T>G) and four healthy control NPC lines to perform multi-omics analysis. We used 10 μM sildenafil treatment for 16 hours, which was sufficient to elicit changes in mitochondrial function (Figure 1G-J). We compared samples treated with sildenafil dissolved in 0.1 % DMSO to samples treated with only 0.1 % DMSO. The omics design allowed us to identify: i) the disease signature, by comparing DMSO-treated LS NPCs to DMSO-treated control NPCs, ii) the sildenafil signature, by comparing sildenafil-treated LS NPCs to DMSO-treated LS NPCs (Figure 2A).

Figure 2. Neurodevelopmental impact of sildenafil unveiled by multi-omics.

(A) Dot plot highlighting top ten enriched Gene Ontology (GO) biological processes in LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) upon sildenafil treatment (10 μM for 16 h) based on transcriptomics. (B) Dot plot highlighting top ten enriched GO biological processes in LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) upon sildenafil treatment (10 μM for 16 h) based on proteomics. (C). Enrichment analysis for 724 reversed molecules by sildenafil based on multi-omics integration. GO biological processes for reversed molecules with logarithmic fold change (logFC)>0 after sildenafil treatment. (D) 3D multi-omics network of the sildenafil signature in LS NPCs. Size of dots defines their cetntrality in the network; color of dots indicates their logFC value. (E) Expression of putative sildenafil-responsive genes identified in the literature in LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) compared to control NPCs (CTRL_1, CTRL_2, CTRL_3, CTRL_4) (disease signature) and in DMSO-treated LS NPCs (ATP6_2, ATP_4, ATP6_5, ATP6_7) compared to sildenafil treated LS NPCs (sildenafil signature) based on bulk transcriptomics. (F-G) Network analysis for the HOXA5 hub and associated pathways in the disease signature and sildenafil signature.

Global transcriptomics revealed a disease signature affecting biological processes related to nervous system development and axon development (Figure S5A), and cellular components related to mitochondrial inner membrane and neuronal cell body (Figure S5B). The sildenafil signature in LS NPCs affected similar biological processes, including axon development and nervous system development (Figure 2B), and cellular components, such as mitochondrial inner membrane and neuronal cell body (Figure S5C). These data indicated that dysregulation of key genes and pathways resulting from the MT-ATP6 variants could be counteracted with sildenafil. For example, the neurodevelopment-associated gene developing homeobox 1 (DBX1)⁴⁷ was downregulated in LS NPCs compared to control NPCs (Figure S5D) and upregulated in treated LS NPCs (Figure S5E, Table S1). Changes induced by sildenafil appeared specific for LS NPCs, as no significant transcriptional effects were observed in sildenafil-treated control NPCs (Figure S5K).

Proteomic analysis revealed a disease signature associated with dysregulation of biological processes of mitochondrial gene expression and cell projection (Figure S5F) and cellular components of the inner mitochondrial membrane (Figure S5G). The sildenafil signature was associated with pathways involved in development, synaptic activity, and electron transport chain (ETC) including NADH-dependent activities (Figure 2C, Figure S5H). Sildenafil thus restored pathways impaired by MT-ATP6 variants. For example, the protein BAG4 that protects against ATP depletion-induced cell death⁴⁸ was downregulated in LS NPCs compared to control NPCs (Table S2) and upregulated in LS NPCs treated with sildenafil (Figure S5J).

Metabolomics identified key metabolites for brain energy metabolism, including creatine^49,50 and cysteine⁵¹, that were altered by the disease (Figure S6A) and normalized by sildenafil (Figure S6B, Table S3). Creatinine, a non-enzymatic breakdown product of creatine, was however not normalized, possibly reflecting an increased use of creatine. The sildenafil signature mainly affected the glutathione metabolism pathway (Figure S6C-D).

We next performed multi-omics integration as previously described.⁵² The integrated disease signature confirmed the enrichment of dysregulated mRNAs, proteins, and metabolites known to impact development, WNT and Notch signaling, and axon guidance (Figure S6E). The integrated sildenafil signature affected similar pathways (Figure S6F). The integrated disease signature and integrated sildenafil signature shared 796 molecules, with 724 exhibiting opposite regulation upon sildenafil administration. Hence, sildenafil reversed 33.5 % of dysregulated genes (724 reversed out of 2160 dysregulated molecules). Biological processes reversed by sildenafil included one group related to mitochondrial respiration and one group related to morphogenesis and axon development (Figure 2D). The multi-omics map of the sildenafil rescue signature in LS NPCs (including only transcriptomics and metabolomics, as proteomics did not show sufficient significant changes) highlighted a network including metabolites like biotin and pantothenic acid and genes associated with ETC function (e.g. NDUFA6) and signaling and development (e.g. NOTCH1, STAT3) (Figure 2E).

To identify sildenafil-responsive genes, we searched the literature for known targets of sildenafil and identified 20 putative targets whose gene expression was significantly altered in LS NPCs compared to control NPCs and restored in LS NPCs treated with sildenafil (Figure 2F). HOXA5 apperaed as the top upregulated gene by sildenafil (logFC_diseased = −0.02 vs. logFC_treated = +5.2). HOXA5 is a member of the A cluster of HOX genes and is essential for organogenesis and neural development.^53-55 In DMSO-treated LS NPCs, the HOXA5 hub was involved in extracellular matrix organization, but it became associated with differentiation and brain development upon sildenafil treatment (Figure 2G-H). Hence, the HOXA5 network may be dysregulated in LS and its modulation by sildenafil might contribute to promote healthy neural development. Another key sildenafil-responsive gene was cGMP-dependent protein kinase 1 (PRKG1), a known master regulator and downstream target of PDE5,⁵⁶ which was highlighted by the multi-omics map of the sildenafil rescue signature (Figure 2E). We monitored the expression levels of putative sildenafil targets in LS NPCs treated with 1 μM or 10 μM sildenafil for 6 or 24 hours under low-glucose growth conditions. Upregulated genes over time included PRKG1, HOXA5, HOXB8, NOTCH1, BRD4, PRDX4, MMP11, RGS6, and NRG1 (Figure S6E), while downregulated genes included NRG1, SPEN, and HEG1 (Figure S6F), confirming the involvement of these genes in the response of LS neural cells to sildenafil.

LS brain organoids show neurodevelopmental defects and highlight the impact of sildenafil on progenitor populations

As both the LS disease signature and sildenafil signature in NPCs included genes related to nervous system development, we investigated this process using cortical brain organoids. In agreement with previous studies in LS brain organoids,^25,29,30 MT-ATP6 variants impaired neurogenic zone formation (Figure 3A) and altered the ratio of early neurons to neural progenitors (Figure 3B). Two different protocols for generating cortical brain organoids showed defective growth rates in the presence of MT-ATP6 variants (Figure S7A). One protocol⁵⁷ showed size defects in LS brain organoids after 50 days in culture (Figure S7B), whereas another protocol that required additional growth factors and the use of AggreWells plates⁵⁸ allowed an initial homogenous organoid shape, resulting in earlier growth defects in LS brain organoids that became less pronounced over time (Figure S7B). These results collectively suggest that MT-ATP6 variants might affect neural progenitor development.

Figure 3. Sildenafil treatment in LS brain organoids affects progenitor populations.

(A) Representative images of cortical brain organoids from controls (CTRL_2) and LS (ATP6_4, ATP6_7) at day 35 stained with neural progenitor marker PAX6 and neuronal marker TUJ1. Scale bar: 200 μm. (B) Differentiation ratio based on gene expression of CTIP2 over SOX2 in day 70 brain organoids from controls (CTRL_1 and CTRL_2) and LS (ATP6_4, ATP6_7). 2-3 brain organoids per line per experiment, n=3 independent experiments. ***p<0.001; unpaired two-tailed t-test. (C) Experimental setup in brain organoids to assess the acute sildenafil signature with bulk transcriptomics from controls (CTRL_1, CTRL_2) and LS (ATP6_4, ATP6_7) and the chronic sildenafil signature with single-nucleus RNA sequencing (snRNAseq) from controls (CTRL_1) and LS (ATP6_7). (D) Differentiation ratio in day 70 brain organoids from LS (ATP6_4). 10 brain organoids per line per experiment, n=3 independent experiments. ****p<0.0001; unpaired two-tailed t-test. (E) qPCR analysis in LS brain organoids (ATP6_4), 10 brain organoids per line n=3 independent experiments. *p<0.05; unpaired two-tailed t-test. (F) Dot plot highlighting top ten enriched GO biological processes by acute sildenafil (10 μM for 24 h) in day 70 LS brain organoids (ATP6_4, ATP6_7) based on transcriptomics. 5 brain organoids per line per experiment, n=3 independent experiments. (G-H) Uniform manifold approximation and projection (UMAP) plot and bar plot showing distribution of day 70 brain organoids from controls (CTRL_1) and LS (ATP6_7) treated with chronic sildenafil (10 μM for 45 days) or only DMSO. 15 brain organoids per line per experiment, n=3 independent experiments. (I) Volcano plot of pseudo-bulk analysis of snRNAseq dataset depicting the disease signature in the progenitor population. (J) Pseudo-bulk analysis of snRNAseq dataset showing the expression pattern of PRKG1, RGS6, and STMN2 across distinct populations of day 70 brain organoids.

We treated LS brain organoids with sildenafil for either 24 hours (acute paradigm) or 45 days (chronic paradigm) (Figure 3C). Global transcriptomics showed that MT-ATP6 variants in brain organoids elicited responses similar to those seen in NPCs, including effects on biological processes related to synapses and neuronal projections (Figure S7E), as well as effects on cellular components related to neuronal cell bodies and synapses (Figure S7F). Sildenafil treatment rescued the ratio of early neurons to neural progenitors (Figure 3D) and upregulated DBX1 (Figure 3E). The acute sildenafil signature in LS brain organoids affected biological processes related to embryonic development and WNT signaling (Figure 3F) and corrected key gene defects. For example, the WDR45B gene whose variants are associated with neurodevelopmental disorders⁵⁹ was downregulated in LS brain organoids compared to control brain organoids (Table S4) and upregulated in sildenafil-treated LS brain organoids (Figure S7D).

To gain insight into the cell populations affected by MT-ATP6 variants and by chronic sildenafil treatment, we performed single-nucleus RNA sequencing (snRNAseq) (Figure 3C). Unsupervised clustering highlighted 9 clusters within cortical brain organoids at day 70 (Figure S8A-B, Table S5). Based on their gene expression patterns, we assigned clusters 0 and 3 to radial glia, clusters 4 and 8 to progenitors, cluster 6 to proliferating progenitors, cluster 5 to immature neurons, cluster 2 to FOXG1-positive neurons, cluster 1 to FOXG1-negative neurons, and cluster 7 to other cell types (Figure S8A-B, Table S5). This annotation showed that MT-ATP6 variants impaired neuronal commitment, with alterations in the populations of radial glia, progenitors, and FOXG1-positive neurons (Figure 3G-H). The disease signature highlighted the downregulation in the progenitor population of neuroligin 1 (NLGN1), a master regulator for synapse development,⁶⁰ and the PDE5 target PRKG1 (Figure 3I), and the downregulation in the neuronal population of genes regulating neurite outgrowth such as STMN2⁶¹ (Figure S8D, Table S6). Differential gene expression revealed that the effect of sildenafil was mostly evident in radial glia, progenitors, and FOXG1-positive neurons (Figure S8C, Table S6). The sildenafil signature uncovered the upregulation of neuronal outgrowth-associated genes as RGS6⁶² across several populations (Figure S8C-D, Table S6). While STMN2 was mainly present in FOXG1-positive neurons and immature neurons, RGS6 was found mainly in radial glia, and PRKG1 in radial glia and progenitors (Figure 3J, Figure S8D). The glycolytic signature, which has been suggested as an indicator of brain organoid stress,⁶³ was not altered in our samples, suggesting that neither MT-ATP6 variants nor sildenafil treatment posed additional stress on brain organoids (Figure S8E). Taken together, LS variants disrupted brain organoid development by impairing early neuronal organization of radial glia and progenitor populations, and sildenafil treatment specifically affected those populations.

Sildenafil improves calcium homeostasis and neurite outgrowth in human LS models

We next examined the functional consequences of sildenafil treatment. Given the reported dysregulation of calcium responses in LS neural cells^26,27 and the known deterioration of LS patients upon metabolic decompensation,³ we induced acute metabolic stress in LS brain organoids to monitor the resulting intracellular calcium increase. We dissected brain organoids on days 70-74 to prepare cortical brain organoid slices (cBOS)⁶⁴ that we grew until day 129 and then treated with sildenafil for 24 hours before applying acute metabolic stress (2 minutes of glucose deprivation and inhibition of glycolysis and OXPHOS) (Figure 4A). The calcium response to metabolic stress was more pronounced and premature in LS cBOS compared to control cBOS, suggesting an increased susceptibility to metabolic imbalance (Figure 4B-C). Pre-treatment with sildenafil in LS cBOS was sufficient to reduce their calcium load after metabolic stress and their peak calcium amplitude (Figure 4B-C). The findings suggest that sildenafil might prevent excessive decompensation in LS neuronal cells under acute metabolic stress.

Figure 4. Sildenafil ameliorates calcium homeostasis and neurite outgrowth.

(A) Schematic of cortical brain organoid slices (cBOS) for live-cell calcium imaging. (B) Exemplary traces of intracellular calcium in control cBOS (CTRL_2) and LS cBOS (ATP6_7) in response to acute metabolic stress (2 min of: glucose-free medium + 2 mM 2-deoxyglucose and 5 mM sodium azide). Individual cells recorded in a single experiment (grey lines) plus corresponding average traces (colored lines). (C) Area under the curve (AUC) and peak of calcium signals evoked by metabolic stress in control cBOS (CTRL_1, n=10; CTRL_2, n=40) and LS cBOS (ATP6_7, n=58; ATP6_7 +Sil, n=16). Dots represent individual cells within cBOS. *p<0.05, ****p<0.0001, ns (not significant); two-tailed Mann-Whitney U test. (D) HCA-based MMP quantification in SURF1 mutant NPCs (SURF1_1) and respective isogenic control NPCs (CTRL_1). Dots represent mean values per image in n=3 independent experiments per line with 10 μM sildenafil or DMSO for 16 h. **p<0.01, ****p<0.0001, ns; ordinary one-way ANOVA with Holm-Šídák's multiple comparisons test. (E-F) Mean calcium trace and respective AUC quantification of calcium imaging in LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) using ratiometric Fura-2 probe with 10 μM sildenafil or DMSO for 16 h. n=at least 3 independent experiments per line. ****p<0.001; unpaired two-tailed t-test. (G-H) HCA-based MMP quantification in wild-type (WT) and Big-Conductance Calcium-activated potassium (BKCa) channel KO (BKCa KO) human cell lines by averaging the TMRM values following the addition of either DMSO (G) or oligomycin (H) expressed as % of basal values. n=3 independent biological replicates. **p<0.01, ****p<0.001, ns; two-tailed Mann-Whitney U test. (I) Cartoon depicting the PDE5-PRKG1 pathway and its modulation. (J) FACS-based MMP quantification in LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) treated with DMSO, 10 μM sildenafil, or 10 μM 8-Br-cGMP. n=7 independent experiments per line ***p<0.001, ****p<0.0001; one-way ANOVA with Tukey’s multiple comparison. (K) HCA-based MMP quantification in LS NPCs (ATP6_2) treated for 16 h with DMSO, 10 μM sildenafil, or 10 μM sildenafil plus pre-treatment for 1 h with 1 μM of the PRKG1 inhibitor KT5823. n=3 independent experiments. *p<0.05, ns; one-way ANOVA with Dunnett’s T3 multiple comparison. (L) PRKG1 protein expression quantified from n=5 independent experiments in control NPCs (CTRL_1, CTRL_2, CTRL_3, CTRL_4, CTRL_5) and LS NPCs (ATP6_2, ATP6_4, ATP6_5, ATP6_7) treated for 16 h with 10 μM sildenafil or DMSO. *p<0.05, ns; unpaired two-tailed t-test. (M) qPCR-based gene expression of PRKG1 in cortical brain organoids at day 70 from controls (CTRL_1, CTRL_2) and LS (ATP6_4, ATP6_7) treated for 24 h with 10 μM sildenafil or DMSO. 5 brain organoids per sample out of n=3 independent experiments. **p<0.01****, p<0.0001, ns; two-tailed Mann-Whitney U test. (N-O) Neurite outgrowth at day 16 for dopaminergic-enriched neurons from MT-ATP6 mutants (ATP6_2, n=2 and ATP6_4, n=1) and controls (CTRL_1, n=2, CTRL_2, n=2 and CTRL_3, n=1), and from SURF1 mutants (SURF1_1, n=3) and isogenic controls (CTRL_1, n=3) treated for 8 days with 10 μM sildenafil or DMSO. Dots represent the mean neurite length per image calculated as the mean over the median per experiment. *p<0.05, ****p<0.0001, ns; Kruskal–Wallis test with Dunn’s multiple comparison test. (P) Linear regression (sildenafil over DMSO) of the mean neurite length in LS neurons (ATP6_2, n=3; ATP6_4, n=1; ATP6_7, n=3; SURF1_1 n=3; NDUFS4_1, n=2). Black dotted line indicates the hypothetical linear regression in which no effect of the treatment would be seen (y=x or slope=1). Dots represent individual experiments. (Q) Neurite outgrowth at day 16 for dopaminergic-enriched neurons from controls (CTRL_1) exposed to sequential small interfering RNA (siRNA) against PRKG1 (PRKG1_KD) or scrambled siRNA for 4 days. Dots represent mean values per well with 4-9 wells per experiment, n=3 independent experiments. *p<0.05; unpaired two-tailed t-test.

We next fed these calcium traces into our computational model of calcium homeostasis and energy metabolism^65-67 (Figure S9A-C, Table S7). To model MT-ATP6 defects, we computationally reduced the activity of ATP synthase of control cBOS and found that this resulted in an elevated calcium response to stress, slight changes in NADH production, hyperpolarization of MMP, and reduction in cellular and mitochondrial ATP content (Figure S9D). Consistent with our findings with sildenafil, the computational increase in ATP synthase activity of LS cBOS resulted in decreased calcium response to stress, reduced NADH, mitochondrial hyperpolarization, and increased ATP concentration (Figure S9E). The sildenafil profile was not recapitulated by modeling a decrease of the peak flux through the mitochondrial calcium uniporter, a decrease of glycolysis, or an increase of NADH oxidation (Figure S9F). Hence, the amelioration of calcium signaling and ATP synthase by sildenafil could work in tandem to restore mitochondrial function. The model raised the possibility that MMP normalization by sildenafil may not be due to a direct mitochondrial uncoupling effect. If this was the case, sildenafil might also prove beneficial in other forms of LS in which the MMP is depolarized. To test this, we measured MMP in NPCs derived from LS iPSCs carrying a variant in the nuclear gene SURF1²⁵ (Figure S11H). The MMP was depolarized in SURF1 mutant NPCs compared to isogenic control NPCs (Figure 4D). Nevertheless, sildenafil was still able to normalize the MMP of SURF1 mutant NPCs (Figure 4D).

To further dissect the impact of sildenafil on calcium homeostasis, we monitored the levels of cytoplasmic calcium in LS NPCs in the presence of thapsigargin, which blocks the SERCA pumps in the endoplasmic reticulum (ER). Under these conditions, sildenafil led to increased cytoplasmic calcium (Figure 4E-F), suggesting that it could act in a ER-independent manner and possibly through modulation of mitochondrial calcium, as suggested by the mathematical model (Figure S9A). We therefore investigated big-conductance calcium-activated potassium (BKCa) channels, which are localized in the plasma and mitochondrial membrane⁶⁸ and have been implicated in the cardioprotective effect of sildenafil.^69,70 As expected,⁷¹ human BKCa KO cell lines⁷² displayed increased MMP compared to wild-type cells (Figure 4G-H). Sildenafil treatment normalized the abnormally elevated MMP when BKCa KO cells were grown in DMSO (Figure 4G) but not if the cells were acutely exposed to the ATP synthase inhibitor oligomycin (Figure 4H). The findings suggest that BKCa channels may contribute to the action of sildenafil on MMP when the cells lack a functional ATP synthase and that residual ATP synthase activity may be necessary to allow sildenafil to exert its restorative role.

Next, we applied chemical manipulations of the PDE5 pathway to prove its involvement in the MMP rescue of LS NPCs (Figure 4I). The cGMP analogue 8-Br-cGMP recapitulated the MMP amelioration seen for sildenafil (Figure 4J). Moreover, inhibition of the downstream target PRKG1 with KT5823 blunted the effect of sildenafil on MMP normalization (Figure 4K). Indeed, PRKG1 levels were reduced in MT-ATP6 mutant NPCs (Figure 4L, Figure S9G), MT-ATP6 mutant brain organoids (Figure 4M), and in SURF1 mutant neurons and brain organoids (Figure S9H).

We then assessed the effect of sildenafil on neuronal outgrowth, as this process was highlighted by the omics analyses. We generated dopaminergic-enriched neuronal cultures (Figure S9I) and quantified neurite outgrowth using HCA.⁷³ In agreement with previous findings,^25,30 MT-ATP6 mutant neurons and SURF1 mutant neurons exhibited reduced neurite length compared to control neurons (Figure 4N-O, Figure S9J). Sildenafil promoted effective neurite outgrowth in LS neurons (Figure 4E-F, Figure S8F). This improvement was specific to LS neurons (carrying MT-ATP6, SURF1, or NDUFS4 variants) (Figure 4P) and was not observed in control neurons (Figure S9K). LS neurons showed reduced PRKG1 signal that increased with sildenafil (Figure S9L). Accordingly, knock-down of PRKG1 using small interfering RNA (siRNA) in control neurons recapitulated the neurite growth defects seen in LS neurons (Figure 4Q). Thus, PRKG1 might play a key role in the sildenafil response of LS cells (Figure S9M), given that it can act on both calcium homeostasis and neurite outgrowth.^74-78

Sildenafil extends the lifespan of LS animal models and can cross the human blood-brain barrier

To address the therapeutic potential of sildenafil in vivo, we employed the germline Ndufs4 KO mouse.^14,15 Sildenafil was added to the drinking water of the animals after weaning starting on day 25 of life and continued throughout their life. The treatment significantly extended the lifespan of Nduks4 KO mice (Figure 5A). Sildenafil attenuated the neurological decline of these mice by alleviating muscle weakness and ataxia, and by partially correcting the defective energy expenditure (Figure 5B). Upon placing LS mice in metabolic chambers (Figure S10A), we observed that sildenafil improved oxygen consumption and carbon dioxide production of LS mice (Figure 5C), which may indicate enhanced metabolic fitness. Cardiac bradyarrhythmia and dysfunction were also ameliorated by sildenafil, including restoration of heart rate and diastolic function (Figure S10B).

Figure 5. Sildenafil extends the lifespan of small and large mammalian models of LS.

(A) Kaplan–Meier survival curve of untreated Ndufs4 KO mice (10 females, 8 males) compared to Ndufs4 KO mice exposed to sildenafil citrate added in the drinking water (6 females, 6 males). ***p<0.001; Log-rank (Mantel-Cox) test. (B-C) Functional parameters in untreated or sildenafil-treated 42-45-days-old Ndufs4 KO mice (n=15 mice out of five independent experiments) collected by placing mice in a metabolic chamber. **p<0.01, ***p<0.001, ns (not significant); ANOVA posthoc analysis. (D-E) Representative immunohistochemistry and related quantification of positive for cleaved caspase3, Iba1, and Prkg1 in Purkinje cells in untreated or sildenafil-treated 50-days-old Ndufs4 KO mice (n=5-6). **p<0.01, ***p<0.001, ns; ANOVA post-hoc analysis. (F) Kaplan–Meier survival curve of wild-type (WT) piglets (n=10), untreated SURF1 KO piglets (n=16), and SURF1 KO piglets treated with sildenafil (n=3 with 2.1 mg/kg/day, n=4 with 0.5 mg/kg/day). *p<0.05 treated mutants vs. untreated; Log-rank (Mantel-Cox) test. (G) Quantification of body temperature in SURF1 KO piglets. ****p<0.0001, ns; unpaired two-tailed t-test. (H) Neurological score of SURF1 KO piglets. ****p<0.0001, *p<0.05, ns; unpaired two-tailed t-test. (I-J) Representative immunoblot image and related quantification of PRKG1 protein expression in basal ganglia lysates of SURF1 KO piglets. n=3 replicates per sample in two independent experiments. **p<0.01, *p<0.05, ns; unpaired two-tailed t-test.

The brain of Ndufs4 KO mice typically exhibit increased apoptosis and higher levels of neuroinflammation.^14,79,80 Consistently, we observed positive cleaved caspase 3 staining in neuronal cells in the cerebellum and brain stem regions associated with increased Iba1 staining⁸¹ suggestive of microglial activation (Figure 5D-E). The number of cells positive for caspase 3 or Iba1 significantly decreased in LS mice treated with sildenafil (Figure 5D-E). The loss of Purkinje cells of cerebellum in LS mice was also ameliorated by sildenafil treatment (Figure S10C-D). The expression of Prkg1, which is particularly evident in normal cerebellar Purkinje neurons,¹⁵ was significantly reduced in LS mice and rescued by sildenafil treatment (Figure 5D-E).

Next, we employed SURF1 KO piglets to investigate the potential effectiveness of sildenafil in a large animal model. These animals exhibit a severe clinical phenotype with encephalomyopathy, reduced suckling reflexes, hypothermia, motor impairment, and lethargy, leading to death in the first days after birth¹¹ (Video S1). SURF1 KO piglets were treated immediately at birth with sildenafil (n=3 with 2.1 mg/kg/day, n=4 with 0.5 mg/kg/day). Sildenafil-treated piglets exhibited a fast and marked improvement in responsiveness, with increased mobility, improved reflexes, and recovery of suckling behavior (Video S2). This improvement led to lifespan extension, once the highly unstable phase right after birth was overcome (Figure 5F). Some animals survived beyond postnatal day 160 despite reduced body weight (53 kg vs. 90–110 kg in age-matched controls) (Video S3). Sildenafil also increased the body temperature of SURF1 KO piglets, suggesting an effect on overall metabolism (Figure 5G, Figure S10E).

To quantify neurological function of SURF1 KO piglets, we adapted the Newcastle Pediatric Mitochondrial Disease Scale (NPMDS) to piglets (Figure S10F). Mutant animals scored poorly across multiple neuromuscular parameters, while sildenafil led to a significant improvement of multiple domains, confirming a broad improvement in neuromotor integrity (Figure 5H, Figure S10F). In some animals treated with high doses that died in the first few days after birth, we observed the presence of microhemorrhages (Figure S10G). These data are in agreement with the increased mortality seen in patients with PAH treated with high doses of sildenafil⁴⁰ and confirm that high sildenafil dosages should be used with caution especially in very young individuals.

We then monitored PRKG1 protein expression in basal ganglia lysates of SURF1 KO piglets. PRKG1 levels were reduced SURF1 KO and were restored by sildenafil in some animals, although the overall effect did not reach statistical significance (Figure 5I-J). Collectively, the in vivo data highlight the potential clinical effectiveness of sildenafil and underscore PRKG1 as one of the mechanistic targets underlying the beneficial effect of sildenafil in delaying neuropathological deterioration.

In order for systemically administered sildenafil to exert potential beneficial effects in LS patients, the drug must cross the blood-brain-barrier (BBB). To probe the permeability for sildenafil in LS, we applied a human iPSC-based BBB model⁸² (Figure 6A). LS brain capillary endothelial cells (BCECs) showed correct morphology and marker expression similarly to control BCECs (Figure 6B, Figure S11A), as well as acceptable monolayer integrity and transendothelial electrical resistance (TEER) (Figure S11B-C). The permeability of sildenafil and sildenafil citrate (the active pharmacological substance applied orally in clinical application) in LS BCECs was similar to that in control BCECs and was higher than the negative control atenolol (beta-blocker known to not permeate into the brain), but lower than the positive control diazepam (an anxiolytic benzodiazepine) (Figure 6C). These results are consistent with permeability values of sildenafil reported in other clinical settings.⁸³ Thus, sildenafil may be able to effectively cross the BBB of patients with LS.

Figure 6. Sildenafil can cross the BBB and improves clinical features in six LS patients with MT-ATP6 variants.

(A) Schematic of the iPSC-derived blood brain barrier (BBB) model to generate brain capillary endothelial cells (BCECs). (B) Representative immunostaining of BCECs from controls (CTRL_8) and LS (ATP6_2) showing co-localization of occludin (OCLN) and tight junction protein ZO-1 required to form a barrier with high transendothelial electrical resistance (TEER). Scale bar: 100 μm. (C) Permeability coefficiennt (Papp) measured by LC-MS/MS in the apical - basolateral media of control BCECs (CTRL_8) and LS BCECs (ATP6_2) exposed to 10 μM sildenafil or sildenafil citrate for 1 h in n=1-5 independent experiments. Diazepam and atenolol (10 μM each) served as internal positive and negative control, respectively. ns (not significant); two-way ANOVA. (D) Quantification of sildenafil in the plasma using LC-MS/MS in three patients with LS undergoing off-label compassionate treatment with sildenafil. Trough level: before the first morning dosage of sildenafil; peak level: one hour after the first dosage. Depicted time points were separated by more than 3 months for each patient. (E) Cranial magnetic resonance imaging (cMRI) of the six patients with LS undergoing off-label compassionate treatment with sildenafil. Arrows indicate lesions in the basal ganglia and brainstem. (F) Clinical development of patients with LS carrying MT-ATP6 variants rated with the Newcastle mitochondrial disease adult scale (NMDAS) or Newcastle pediatric mitochondrial disease scale (NPMDS) plotted as percentage of the maximum reachable score for the respective age group. Brown lines indicate individuals previously described in a cohort from Newcastle.⁸⁷ Orange lines show individuals in our cohort from Charité Berlin.⁸⁸ (G) Longitudinal clinical evaluation of off-label compassionate treatments with sildenafil in six patients with LS carrying different MT-ATP6 variants based on NMDAS/NPMDS score. (H-J) Clinical development based on different dimensions of the NPMDS scale. Above: individuals in our cohort from Charité Berlin. Below: off-label compassionate treatments with sildenafil.

Lastly, we tested potential cardiotoxic effects of sildenafil using LS iPSC-derived cardiomyocytes (Figure S11A). We found that sildenafil was toxic to LS cardiomyocytes only when used at very high concentrations (Figure S11B-C). Lower sildenafil concentrations did not affect cell viability of LS cardiomyocytes (Figure S11B-C) and did not alter their glycolytic rate (Figure S11D-E).

Compassionate sildenafil treatment improves clinical conditions in six LS patients

The promising preclinical findings of sildenafil in cellular and animal models of LS prompted us to ask whether this drug could be repurposed to treat patients with LS. Sildenafil has been used safely in pediatric conditions including PAH and lymphatic malformations.^39-42 We initiated off-label compassionate chronic oral application of sildenafil citrate (Revatio^®) in six LS patients carrying different MT-ATP6 variants (Table 1, Table S8). All patients had cranial magnetic resonance imaging (cMRI) signs consistent with basal ganglia lesions characteristic of LS (Figure 6E, Figure S10J). Patient 5 was initially diagnosed with NARP (Neuropathy, Ataxia, Retinitis Pigmentosa). NARP and LS caused by MT-ATP6 variants are part of a disease continuum that is also referred to as “Leigh syndrome spectrum” disease.⁸⁴ From patient 1, we also obtained iPSCs (line ATP6_7), which we used to generate NPCs and brain organoids (Figure S11H). When selecting the sildenafil dosage, we referred to the STARTS-1⁴¹ and STARTS-2⁴⁰ trials on children with PAH. Given the potential toxic effects of high dosages, we only considered low (0.5–1.0 mg/kg/day) or medium dosages (1.8–2.1 mg/kg/day). Three parents/patients opted for low dosages and three for medium dosages. Mass spectrometry quantification of total sildenafil concentration in the plasma was performed in three exemplary patients by collecting blood in the morning before sildenafil administration and 1 hour after (Figure 6F). The drug reached a plasma concentration of approximately 1 μM, which is close to the IC₅₀ value of 3 μM that was found to be effective in vitro for normalizing the MMP in LS NPCs (Figure S3H).

Table 1. Clinical characteristics and response to sildenafil therapy in off-label compassionate studies in in six patients with LS carrying MT-ATP6 variants.

Patient [sex]	Variant [mtDNA haplogroup] (heteroplasmy)	mtDNA variant on the protein level	Phenotype [manifestation age]	[HPO number] symptoms	cMRI features [at age in years]	Initiation of sildenafil treatment [years] (dosage)	Effect of sildenafil treatment
1 [male]	m.9176T>G [K1b2a3] (99.4%)	MT-ATP6 p.L217R	LEIGH [2 years]	[HPO:1250] epilepsy [HPO:1260] dysarthria [HPO:1270] motor developmental delay [HPO:1639] cardiomyopathy [HPO:2093] respiratory insufficiency [HPO:3477] demyelinating neuropathy [HPO:6789] encephlopathy [HPO:7240] ataxia	Increased T2-signal intensities at the Putamen, Nucleus caudatus on both sides and at the perisylvian gray matter. [16.5 years]	December 2018 ongoing [16 years] (2.0 mg/kg/day)	Improvements: independent breathing (tracheostoma could be closed), motor abilities (can sit, walk with support, maneuver wheel-chair), seizures stopped, independent eating and swallowing, cardiomyopathy improved. No change: neuropathy, dysarthria
2 [male]	m.8993T>G [U5a1b3] (87.7%)	MT-ATP6 p.L156R	LEIGH [2.5 years]	[HPO:1270] motor developmental delay [HPO:2381] transient aphasia [HPO:4911] metabolic crises [HPO:7021] pain insensitivity [HPO:7240] ataxia	Increased T2- and FLAIR-signal intensities at the Putamen and Nucleus caudatus on both sides. [2.2 years]	October 2019 ongoing [15 years] (1.2 mg/kg/day)	Improvements: no metabolic crises during febrile illness, improvement of exercise tolerance (free walking distance improved from 500 to >5000 meters), no daytime sleepiness, ataxia improved, SARA score improved from 15.5 to 11 points. No change: mild intellectual disability
3 [female]	m.9185T>C [T2b] (100.0%)	MT-ATP6 p.L220P	LEIGH [5 years]	[HPO:0544] horizontal ophthalmoplegia [HPO:1256] mild intellectual disability [HPO:1270] motor developmental delay [HPO:2355] loss of ambulation [HPO:3477] axonal neuropathy [HPO:3752] episodic muscle weakness	Increased T2-signal intensities at the Putamen on both sides [5 years], which had spontaneously disappeared by the age of 14 years.	May 2019 until November 2021 [21 years] (2.9 mg/kg/day)	Improvements: reduction of episodes of muscle weakness, improvement of exercise intolerance. No change: neuropathy. Comments: After stopping sildenafil due to a delayed drug reaction (skin rash), the disease symptoms reappeared.
4 [male]	m.8993T>G [R0a1a] (96.7%)	MT-ATP6 p.L156R	LEIGH [2-3 years]	[HPO:0365] deafness [HPO:1249] intellectual developmental delay [HPO:1270] motor developmental delay [HPO:2123] myoclonic seizures [HPO:2421] loss of head control [HPO:3546] exercise intolerance [HPO:7240] ataxia	Increased FLAIR-signal intensities at the Nuclei caudati on both sides and enlargement of the internal and external CSF spaces due to diffuse brain atrophy. [35 years]	May 2022 ongoing [38 years] (1.5 mg/kg/day)	Improvements: increased muscle strengh (keeping his head position, ability to sit and walk), dysphagia, hyposthenia, slight improvement of in cognitive abilities. No change: deafness, ataxia. Comments: While the patient had to discontinue sildenafil for the time of an operation, the symptoms reoccurred after a short time and disappeared again after reinitiation of sildenafil.
5 [male]	m.8570T>C [H4] (98.5%)	MT-ATP6 p.L15P MT-ATP8 p.X68Q+26AS*	NARP [2 months]	[HPO:1639] hypertrophic cardiomyopathy [HPO:1903] anemia [HPO:2015] dysphagia [HPO:4911] multiple metabolic crises [HPO:8947] floppy infant	Delayed myelination, frontal brain atrophy. Increased T2-signal intensities in the brain stem affecting the Fasciculus longitudinalis and the Substantia nigra on both sides. [2.2 years]	September 2019 ongoing [9 months] (1.5 mg/kg/day)	Improvements: dysphagia, did not have a metabolic crisis after initiation of sildenafil therapy (before nearly one crisis every month with ICU admission), cardiomyopathy stable. No change: intellectual disability, delayed motor and cognitive development
6 [female]	m.8993T>C [T2a1a] (100.0%)	MT-ATP6 p.L156R	LEIGH [7 years]	[HPO:1258] spastic paraplegia [HPO:1268] intellectual decline [HPO:1290] muscular hypotonia [HPO:1332] dystonia [HPO:2015] dysphagia [HPO:2371] loss of speech [HPO:2421] loss of head control [HPO:4911] metabolic crisis [HPO:6789] encephalopathy	Increased T2-signal intensities at the Putamen, Pallidum, and Nucleus caudatus on both sides and within the cortical gray matter. [7.5 years]	June 2024 ongoing (7 years] (1.8 mg/kg/day)	Improvements: sitting with support and standing with assistance, consciousness normalized, normal language perception, slow improvement of muscle tone and movement disorder. No change: still wheel-chair dependent due to profound dystonia

NARP: Neuropathy, Ataxia, Retinitis pigmentosa; HPO: Human Phenotype Ontology; cMRI: cranial Magnetic Resonance Imaging; FLAIR: Fluid Attenuated Inversion Recovery; CSF: cerebrospinal fluid; SARA: Scale for the Assessment and Rating of Ataxia

Upon starting the treatment, we closely monitored the occurrence of potential adverse reactions. We asked patients and guardians to respond to a questionnaire on side effects after six months of sildenafil treatment (Table S9). Sildenafil was generally well tolerated (Table S9) and continued in most patients (Table 1, Table S8). In patient 3, sildenafil had to be discontinued due to a rash, even if the frequent episodes of sudden muscle weakness were resolved during sildenafil medication and returned after the drug was discontinued (Table 1, Table S9). In all other cases, we observed a clear improvement in different clinical parameters, including motor function and resilience against metabolic crises (Table 1, Table S8).

We monitored disease progression using the NPMDS⁸⁵ or the Newcastle mitochondrial disease adult scale (NMDAS).⁸⁶ These rating scales cover all aspects of mitochondrial disease including clinical assessments and quality of life. The total score describes the overall severity of disease burden and typically increases over time in patients with mitochondrial diseases.^85,86 The NPMDS/NMDAS score can rise rapidly in patients with LS, especially following metabolic crises. This increase can be seen in the cohort of LS patients with MT-ATP6 variants previously described in Newcastle,⁸⁷ where the score increased by 8.8 SD 11.4 percent points per year (Figure 6F, brown lines), and in our cohort at Charité in Berlin,⁸⁸ where the score increased by 3.4 SD 2.2 percent points (Figure 6F, orange lines). Remarkably, the NPMDS/NMDAS scores of the six sildenafil-treated patients decreased or showed reduced rates of increase over time (Figure 6G, Figure 6H-J). In patients 1, 5 and 6, sildenafil was started during a metabolic crisis, which resolved rapidly and led to a rapid initial drop in their NMDAS/NPMDS scores. The subsequent NPMDS/NMDAS increase seen in patient 1 may be due to the natural progression of the disease or non-compliance. Patient 5 showed protection from the frequent metabolic crises which had occurred nearly monthly before sildenafil treatment, while mild improvements in cognitive abilities were reported in patients 4 and 6 (Table 1, Table S8). Overall, the results of the individualized compassionate studies suggest that chronic treatment with sildenafil lead to clinical improvement in patients with LS carrying MT-ATP6 variants. A formal randomized controlled clinical trial is now required to evaluate the safety and potential efficacy of sildenafil in patients with LS.

Discussion

Sildenafil is a well-studied molecule in clinic applications.³² In addition to its widespread use in adult men with erectile dysfunction,⁸⁹ sildenafil is used in children to treat PAH and rare lymphatic malformation.^39-42 Recent studies suggest that sildenafil might be a repurposable candidate for treating disorders of the nervous system.⁹⁰ For instance, sildenafil use was associated with a reduced risk of developing Alzheimer’s disease (AD) and promoted neuronal outgrowth in iPSC-derived neurons from AD patients.^91,92 Sildenafil also lowered cognitive decline in mouse models of Huntington’s disease (HD)^93,94 and increased motor and cognitive performances in patients with HD.⁹⁵ It also alleviated neuropathological sings in animal models of multiple sclerosis⁹⁶ and ischemic stroke.⁹⁷ In this study, we found that sildenafil might slow down the neurological decline of the severe mitochondrial brain disease LS. Sildenafil increased the lifespan of two mammalian models of LS. One treated piglet remained alive and clinically stable for over six months, an unprecedented heath span and lifespan extension in this model. In treated individuals, the reduction of the upward sloping trend in NPMDS/NMDAS scores over time possibly implies that sildenafil treatment modified the disease trajectory.

Sildenafil is a PDE5i leading to the release of cGMP, which in turn has a wide range of functions, likely resulting in a poly-pharmacological mode of action.³² In the context of the nervous system, sildenafil can enhance neurogenesis in mice.^98,99 cGMP and PRKG1 can modulate neurotransmitter release, neuronal survival, and neurite formation.^74,75-77 Mice deficient for Prkg1 show impairment in axon guidance and connectivity.⁷⁸ Consistently, we found that the signature of sildenafil primarily involved the regulation of nervous system development and, within brain organoids, acted specifically on progenitors and radial glia populations. LS patient-derived neurons exhibited impaired neuronal outgrowth and reduced PRKG1 levels. Moreover, knock-down of PRKG1 in healthy neurons was sufficient to recapitulate the branching defects. PRKG1 expression was decreased also in cerebellar Purkinje cells of LS mice and in the basal ganglia of LS piglets. In mouse brain, Prkg1 is known to be highly expressed in Purkinje cells, where its selective depletion result in specific motor learning defects.¹⁰⁰ Here, we found that treatment with sildenafil increased Pkgk1 levels in LS mice and LS piglets. These findings, together with the ability of the drug to cross the BBB, suggest that sildenafil may be beneficial in promoting neuronal morphogenesis and connectivity in the brain of patients with LS.

In addition to its neurological impact, sildenafil modulated bioenergetic processes in vitro and improved metabolic fitness of LS mice and LS piglets in vivo. Sildenafil is known to induce mitochondrial biogenesis and oxygen consumption,^101,102 and PRKG1 has been suggested to be a regulator of energy homeostasis.¹⁰³ This is supported by our findings showing that sildenafil failed to normalize the MMP of LS neural cells when PRKG1 was inhibited. Sildenafil also reversed changes in mitochondrial crista junctions distributions that may reflect defects of cristae morphology caused by MT-ATP6 variants.⁴⁵ We postulate that the bioenergetic improvement of sildenafil may be related to the modulation of calcium homeostasis. We have previously reported that PDE5i ameliorated calcium defects in LS NPCs.²⁷ Here, we found that pre-treatment of LS brain organoids with the PDE5i sildenafil reduced their abnormal calcium response to acute metabolic stress. It is tempting to speculate that this effect may also be related to PRKG1, given its known role in mediating neuronal calcium signalling.^104,105 However, since PRKG1 protein levels were not fully normalized in all LS models, it is possible that sildenafil might enhance PRKG1 enzymatic activity rather than its protein abundance.⁵⁶ In fact, even continuous treatment with the cGMP analogue 8-Br-cGMP has been shown to reduce PRKG1 expression.¹⁰⁶ Therefore, further dissection of this pathway is required.

Whether the restorative effects of sildenafil seen for LS are linked to its vasodilatory mechanisms³² remains to be studied. The increase in blood flow and oxygen to muscle and brain tissue may potentially contribute to beneficial metabolic effects. Indeed, sildenafil led to vascular and metabolic improvements in individuals with AD¹⁰⁷ and in subjects with vascular cognitive impairment due to cerebral small vessel disease.¹⁰⁸ The known effect of sildenafil on the vasculature³² and the beneficial consequences of hypoxia seen in animal models of LS^22,23 might indicate that modulation of brain vascularization and associated oxygenation¹⁰⁹ could represent important therapeutic targets in LS. Functional near-infrared spectroscopy might provide longitudinal data on brain oxygenation and development in future clinical trials.

Our findings underscore the power of conducting drug discovery studies driven by cellular reprogramming and iPSC models. Similar strategies could be applied to other rare neurological diseases for which effective model systems are lacking. Since mitochondria play a crucial role in cellular health and brain function and homeostasis,^110-112 understanding the potential effect of PDE5i in mitochondrial diseases may thus have important implications in the context of other more common aging-related neurological disorders.

Limitations of the study

We only report clinical effects observed in a small cohort of mitochondrial disease patients. Therefore, it is imperative that the safety and efficacy of sildenafil in patients with LS should be assessed in interventional phase II/III randomized controlled clinical trials. The fact that high sildenafil dosages were associated with toxicity in LS cardiomyocytes and microhemorrhages in newborn LS piglets underscore the importance of safety data. Although the doses that we used in LS patients have been found safe in neonates with neonatal encephalopathy,^113,114 additional precautions may be taken in children with less than one year of age due to the specific sildenafil clearance in postnatal age leading to longer half-life in neonates compared to adults.¹¹⁵ Sildenafil dosage should therefore be carefully monitored in treated children. Here, we measured total sildenafil and not the pharmacologically active free sildenafil. Based on studies of sildenafil biodistribution in non-human primates,⁸³ we assume the free sildenafil concentration to range between 8.5 nM and 58.4 nM in the plasma and between 6.3 nM and 44.3 nM in the liquor, which could correspond to ~1.7-11.9 fold of the IC₅₀ of sildenafil. Other more potent PDE5 inhibitors such as vardenafil (IC₅₀ = 0.091 ± 0.031 nM) and tadalafil (IC₅₀ = 1.8 ± 0.4 nM)¹¹⁶ could be considered for LS patients as alternatives for sildenafil.

The identified proteomic changes in LS neural cells were not sufficient to be included in the multi-omics integration analysis, likely because of heterogeneity among patient iPSC lines. Future studies with improved mtDNA editing technologies¹¹⁷ might allow the establishment of isogenic iPSC lines carrying different MT-ATP6 variants that could be used to dissect disease-related mechanisms with higher precision.

Lastly, although we identified here putative sildenafil-responsive genes, additional work is needed to dissect their biological role in LS. Despite the poly-pharmacology of sildenafil and its multidimensional effects, gaining additional insights into the mechanisms of action would be instrumental for the development of a treatment strategy for LS and for avoiding unwanted side effects in the treated population.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Alessandro Prigione, MD, PhD (alessandro.prigione@hhu.de).

Materials availability

There are restrictions to the availability of patient-derived iPSCs used in this study due to the nature of our ethical approval that does not support sharing to third parties without a specific amendment and does not allow to perform genomic studies to respect the European privacy protection law.

Data and code availability

The omics datasets supporting the results of this paper have been deposited in international repositories. Raw and processed RNA sequencing data have been deposited in the NCBI Gene Expression Omnibus (GEO). The dataset includes raw FASTQ files, gene expression matrices, and associated metadata. The mass spectrometry data have been deposited in the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository¹¹⁸. The metabolomics dataset has been deposited in the MassIVE repository (https://massive.ucsd.edu/). Codes to reproduce our results are available at: https://github.com/tpentim/Leigh_Sildenafil_Organoids. All accession numbers are listed in the Key Resources Table.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Purified anti-Pax-6 antibody, Poly19013	BioLegend	901301; RRID: AB_2565003
Anti-β-Tubulin III Antibody, mouse monoclonal, 2G10 (TUJ1)	Sigma-Aldrich	T8578; RRID: AB_1841228
PRKG1 polyclonal antibody	Proteintech	21646-1-AP; RRID: AB_2878897
Anti-Phosphorylated Vimentin (Ser55) mAb	MBL Life Science	D076-3; RRID: AB_592963
Anti-Nestin monoclonal antibody, clone 10C2 (mouse)	Millipore	MAB5326; RRID: AB_2251134
Occludin Monoclonal Antibody (OC-3F10)	Invitrogen	33-1500; RRID: AB_2533101
ZO-1 Polyclonal antibody	Proteintech	21773-1-AP; RRID:AB_10733242
Cleaved Caspase-3 (Asp175) Antibody	Cell Signaling Technology	9661
Anti-Iba1 antibody [EPR16589] - Mouse IgG1 (Chimeric)	Abcam	Ab283319; RRID:AB_2924797
ATP6 polyclonal antibody	Immunological Science	AB-83828
Anti-GAPDH antibody [6C5]	Abcam	Ab8245; RRID:AB_2107448
Anti-β-actin Antibody, mouse monoclonal	Abcam	ab8226; RRID: AB_30637
PRKG1 antibody, rabbit polyclonal	Cell Signaling	3248
mTOR-Ser2448 antibody, rabbit polyclonal	Cell Signaling	2971; RRID: AB_330970
p-AMPKa-Thr174 antibody, rabbit polyclonal	Cell Signaling	2535
p-PGC1a-Ser571 antibody, rabbit polyclonal	Biotechne	AF6650; RRID: AB_10890391
p-4EBP-1 antibody, rabbit polyclonal	Cell Signaling	2855
PGC1a antibody, rabbit polyclonal	Novus Bio	NBP1-04676; RRID: AB_1522118
AMPKa antibody, rabbit polyclonal	Cell Signaling	2532; RRID: AB_330331
4EBP-1 antibody, rabbit polyclonal	Cell Signaling	9644
mTOR antibody, rabbit polyclonal	Cell Signaling	2972; RRID: AB_330978
GAPDH antibody, rabbit polyclonal	Abcam	ab181602; RRID: AB_2630358
Donkey anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488	Invitrogen	A-21206; RRID: AB_2535792
Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 633	Invitrogen	A-21050
Goat Anti-Mouse IgG (H + L)-HRP Conjugate	BioRad	1706516; RRID: AB_2921252
Anti-Rabbit IgG (H+L), HRP Conjugate	Promega	W4018
Anti-Mitofilin antibody [EPR8749] (MIC60)	Abcam	Ab137057; RRID: AB_3676556
Anti-ACTN2 Antibody (alpha-actinin)	Sigma-Aldrich	A7811; RRID: AB_476766
ATP5B	Sigma-Aldrich	HPA001528
NDUFA9	Invitrogen	459100
SDH-B	Abcam	ab14714
UQCRSF1	Abcam	ab14746
DCX antibody	Cell signaling	4604
Chemicals, peptides, and recombinant proteins
StemMACS™ iPS-Brew XF, human	Miltenyi Biotec	130-104-368
mTeSR™ Plus	STEMCELL Technologies	100-0276
KnockOut-DMEM	Gibco	10829-018
KnockOut™ serum replacement	Gibco	10828-028
DMEM/F/12	Gibco	31330038
DMEM glucose-free	Gibco	11966025
Glasgow-MEM	Gibco	11710035
Neurobasal A-Medium glucose-free	Gibco	A2477501
Neurobasal	Gibco	21103-049
Non-essential amino acids (MEM-NEAA) 100X	Gibco	11140-050
Sodium Pyruvate	Gibco	11360070
B-27 with Vitamin A (50×)	Gibco	17504044
B-27 without Vitamin A (50×)	Gibco	12587010
N2 Supplement (100×)	Gibco	17502-048
DPBS, no calcium, co magnesium	Gibco	14190144
Accutase	Thermo Fisher Scientific; Sigma-Aldrich	A1110501; A6964-100ML
UltraPure™ 0,5 M EDTA	Invitrogen	11568896
ROCK inhibitor Y-27632	Enzo Life Sciences	ALX-270-333-M005
MycoZap Plus-CL	Lonza	VZA-2012
Pen/Strep	Gibco	15140122
GlutaMAX	Gibco	35050061
L-glutamine	Gibco	25030081
FBS	Gibco	10270106
Matrigel, Growth Factor reduced	Corning	356231
Geltrex™ Reduced-Growth Factor Basement-Membrane Matrix, LDEV-free, stem-cell qualified	Gibco	A1413302
Laminin	Sigma Aldrich	L2020
Chemically Defined Lipid Concentrate	Gibco	11905031
Anti-Adherence Rinsing Solution	STEMCELL Technologies	07010
Purmorphamine	Sigma Aldrich; Miltenyi Biotec	540220; 130-104-465
CHIR99021	Sigma Aldrich	SML1046
Dorsomorphine	Sigma-Aldrich	P5499
SB431542	Miltenyi Biotec	130-105-336
GDNF	R & D System	212-GD-010
BDNF	MACS Miltenyi	130-096-811
EGF	R & D System	236-EG-200
FGF2	R & D System	3718-FB-100
Db-cAMP	StemCell Technologies	73886
Recombinant Human NT-3	Peprotech / Biozol	450-03
FGF8-a	R&D systems	4745-F8-050
TGFbeta 3	StemCell Technologies	78156
(+)-sodium L-ascorbate (Vitamin C)	Sigma Aldrich	A4034
Human Recombinant Activin A	StemCell Technologies	78001.1
cis-4,7,10,13,16,19-Docosahexaensäure (DHA)	Sigma-Aldrich	D2534
WNT antagonist IWR1	EMD Millipore Corp	681669
Heparin	Merck	375095
2-mercaptoethanol	Gibco	31350010
Doxycycline hydrochloride (DOX)	Sigma Aldrich	D3072
DMSO	Sigma-Aldrich	D2650-100ML
SuperSignal West Pico PLUS Chemiluminescent Substrate	Thermo Fisher Scientific
16% Paraformaldehyde (PFA)	Thermo Fisher Scientific	28906
Hoechst 33342	Invitrogen	H3570
Donkey serum	Merck Millipore	S30
Triton-X-100	Sigma-Aldrich	93443; 93426
Tween 20	Sigma-Aldrich	P1504
abberior STAR 635 (STAR RED)	Abberior	ST635-1002
DAPI	Sigma-Aldrich	D9542
MitoProbe™ TMRM Assay Kit for Flow Cytometry	Invitrogen	M20036
Incucyte dye red	Sartorius	4717
Calcein AM Viability Dye®	Invitrogen™	C1430
MitoSpy™ Orange	BioLegend	424803
Sildenafil-d3	CDN isotopes	D-6366
HpaII	NEB	R0171
StuI	NEB	R0187
XbaI	NEB	R0145
FCCP	Biozol	SEL-S8276
Antimycin A	Sigma-Aldrich	A8674
Oligomycin	Sigma-Aldrich	O4876-5MG
CCCP	Sigma-Aldrich	555602
Propidium iodide	Invitrogen	P1304MP
Sildenafil citrate	Sigma	SML3033
KT5823	Biomol	LKT-K7602.100
8-Br-cGMP	Sigma-Aldrich	203820
Sildenafil	Selleckchem	S468402 and S468403
d3-sildenafil	CDN Isotopes
Mowiol with 0.1 % 1,4-Diazabicyclo[2.2.2]octan (DABCO)	Carl Roth	0713.1
LB Agar	Sigma	L2897
SuperFrost Plus glass slides	VWR	631-0447
Pro-Long Glass Antifade Mountant	Invitrogen	P36984
oligo d(T)18 primers	Thermo Fisher Scientific	SO132
dNTP mix	Thermo Fisher Scientific	10319879
Rnase OUT Recombinant Ribonuclease Inhibitor	Invitrogen	10777-019
M-MLV Reverse Transcriptase (200 U/μL)	Invitrogen	28025021
Human Endothelial-SFM	Gibco	11111044
hFGF	Peprotech	100-18B
Retinoic acid	StemCell Technologies	72262
Collagen IV	Sigma-Aldrich	c5533-5MG
Fibronectin	Gibco/ThermoFisher	33016015
Transwell membranes	Greiner	662641
Sodium fluorescein	Sigma-Aldrich	F6377-100G
Diazepam	Sigma-Aldrich	D0899-100MG
Atenolol	Sigma-Aldrich	A7655-1G
LysC	Roche	11420429001
Low melting temperature agar	Gibco	5517UB
Hank’s Balanced Salt Solution (HBSS)	Sigma	H9394
Oregon Green 488 BAPTA-1 AM	Invitrogen	O6807
Deoxy-D-glucose	Apollo Scientific; Sigma-Aldrich	OR3900T; D8375-1g
RNase H	Epicentre/ LGC Biosearch Technologies	E0038-5D1
Acetonitrile	Biosolve, ULC/MS-CC/SFC	001204101BS
Trypan Blue Solution	Sigma-Aldrich	T8154
Medium 199	Life Technologies/ Thermo Fisher Scientific	22340020
Trypsin	Sigma-Aldrich/Merck	T4799
Medium 199 Modified	Sigma-Aldrich/Merck	M3769
Penicillin-G	Sigma-Aldrich/Merck	P7794
Streptomycin sulfate	Sigma-Aldrich/Merck	S1277
Kanamycin	Sigma-Aldrich/Merck	K1377
FBS Premium plus GI	Life Technologies/ Thermo Fisher Scientific	01190048M
Sodium bicarbonate	Sigma-Aldrich/Merck	S5761
ITS premix - Universal Culture Supplement	Corning	354351
Hormones (FSH, LH)	Meriofert	75IU
Long-EGF	Gibco/ ThermoFisher Scientific	AF-100-15
bFGF	Gibco/ ThermoFisher Scientific	100-18B
Hyaluronidase	Sigma-Aldrich/Merck	H3506
Protease from Streptomyces griseus	Sigma-Aldrich/Merck	P8811
Cytochalasin B	Sigma-Aldrich/Merck	C6762
D-Mannitolo	Sigma-Aldrich/Merck	M1902
Phytohemagglutinin (Lectin)	Sigma-Aldrich/Merck	L-1668
Cycloheximide	Sigma-Aldrich/Merck	C7698
CaCl2	Sigma-Aldrich/Merck	C7902
Altrenogest	MSD Animal Health
PGF2-alpha	Fatro S.P.A.
Sodium dithionite	Merck	1.06505
Cytocrome C	Sigma-Aldrich	C2037
n-dodecil β-D-maltoside	Sigma-Aldrich	D4641
Bovine Serum Albumin Fraction V	Roche Diagnostics	10735086001
Acetil coenzima A	Merck	10101907001
Oxaloacetic acid	Sigma-Aldrich	O4126
DTNB (5,5′-dithiobis(2-nitrobenzoic acid):	Sigma-Aldrich	D8130
Critical commercial assays
Lactate-Assay kit	Sigma Aldrich	MAK064
CellTiter-Glo® Luminescent Cell Viability assay	Promega	G7571
Rnase-Free Dnase Set (50) (250)	Qiagen	79254
RNeasy Mini Kit (50)	Qiagen	74104
First Strand cDNA Synthesis Kit	Thermo Scientific	K1612
PDS - Papain Dissociation System	Cell Systems	LK003150
Nucleo-Spin Tissue kit	Macherey-Nagel	740952.50
SYBRTM Green PCR Master Mix	Applied Biosystems	4364344
Bicinchoninic Acid (BCA) protein assay kit	Thermo Scientific	23252
NucleoSpin RNA Plus kit	Macherey-Nagel	740984.50
RNA Cleanup XP beads	Agencourt/Beckman coulter	A66514
TURBO DNase rigorous treatment	Invitrogen	AM1907
TruSeq Stranded Total LT Sample Prep Kit	Illumina	N/A
Chromium Single Cell 3' (vNext) Reagent Kit	10X Genomics	N/A
ViaLightTM Plus Kit	Lonza	LT07-221
CyQUANT Cell Proliferation Assay	Invitrogen	C7026
jetPRIME transfection	Polyplus	101000027
Deposited data
Bulk RNAses NPCs dataset	This paper	https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE292178 GSE292178 Token: kviluouivdmlnmn
Bulk RNAseq organoids dataset	This paper	Reviewer only link: https://dataview.ncbi.nlm.nih.gov/object/PRJNA1248577?reviewer=h0vj30jbkanvobfmhpb90nq1nn SRA BioProject ID: PRJNA1248577 Token: ixobgyempjkrzsr
Codes	This paper	https://github.com/tpentim/Leigh_Sildenafil_Organoids
snRNAseq organoids dataset	This paper	https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE293573 GSE293573 Token: ixobgyempjkrzsr
Proteomics NPCs dataset	This paper	ProteomeXchange http://proteomecentral.proteomexchange.org PXD059519 Token: Edp2AxzzRPDZ
Metabolomics NPCs dataset	This paper	MassIVE: ftp://MSV000097182@massive.ucsd.edu Submission ID: MSV000097182; Account: curemils; Account password: leighpass,
Experimental models: Cell lines
Human iPSC: BIHi043-A	Helmholtz Zentrum München (HMGU)	PMID 29396371
Human iPSC: HHUUKDi009-A	Heinrich-Heine-Universität Düsseldorf (HHUUKD)	PMID 28132834
Human iPSC: CRMi003-A	RUCDR Infinite Biologics	PMID 36459969
Human iPSC: BIHi269-B	Berlin Institute of Health (BIH)	PMID 36669241
Human iPSC: HVRDi004-B	Synthego	PMID 36459969
Human iPSC: IUFi004-A	Cell Applications	PMID 38217996
Human iPSC: BIHi266-A	Berlin Institute of Health (BIH)	PMID 36669241
Human iPSC: WISCi004-B	WiCell	PMID 18029452
Human iPSC: UMGi014-C	University Medical Center Goettingen (UMG)	PMID 33905594
Human iPSC: HHUi001-A	Universitätsklinikum Düsseldorf (HHU)	PMID 28132834
Human iPSC: HHUi002-A	Universitätsklinikum Düsseldorf (HHU)	PMID 28132834
Human iPSC: HHUi003-C	Universitätsklinikum Düsseldorf (HHU)	PMID 36137325
Human iPSC: MDCi008-A	Max Delbrück Center Berlin Buch (MDC)	PMID 35279592
Human iPSC: MDCi009-A	Max Delbrück Center Berlin Buch (MDC)	PMID 35279592
Human iPSC: MDCi010-A	Max Delbrück Center Berlin Buch (MDC)	PMID 35279592
Human iPSC: BIHi267-B	Berlin Institute of Health (BIH)	PMID 36669241
Human iPSC: C1_mut2	Max Delbrück Center Berlin Buch (MDC)	PMID 33771987
Human iPSC: E9	Max Delbrück Center Berlin Buch (MDC)	PMID 36741056
Human iPSC: F3	Max Delbrück Center Berlin Buch (MDC)	PMID 36741056
Experimental models: Organisms/strains
Germline Ndufs4 KO, C57/BL6/J background	Reference	PMID: 20534480
SURF1 KO pig, Large White background	Avantea	PMID: 29601977
Oligonucleotides
ATP6: Forward: CAACCGACTAATCACCACCC, Reverse: GTTGAGCCGTAGATGCCGTC	IDT	N/A
ATP6_2: Forward: AACCAATAGCCCTGGCCGTA, Reverse: AGGGCTCATGGTAGGGGTAAA	IDT	N/A
GAPDH: Forward: CTGGTAAAGTGGATATTGTTGCCAT, Reverse: TGGAATCATATTGGAACATGTAAACC	IDT	N/A
OAZ1: Forward: GGATCCTCAATAGCCACTGC, Reverse: TACAGCAGTGGAGGGAGACC	IDT	N/A
NESTIN: Forward: TTCCCTCAGCTTTCAGGAC, Reverse: GAGCAAAGATCCAAGACGC	IDT	N/A
PAX6: Forward: CCAGGGCAATCGGTGGTAGT Reverse: ACGGGCACTCCCGCTTATAC	IDT	N/A
CTIP: Forward: TGGGTGCCTGCTATGACAAG, Reverse: GATGCCTTTCGTGGGTGAGA	IDT	N/A
SOX2: Forward: GTATCAGGAGTTGTCAAGGCAGAG, Reverse: TCCTAGTCTTAAAGAGGCAGCAAAC	IDT	N/A
PRKG1: Forward: ACAACTGTACCCGGACAGCGA, Reverse: TCCTCTTGCACCCTGCCTGAT	IDT	N/A
OCT4: Forward: GTGGAGGAAGCTGACAACAA, Reverse: ATTCTCCAGGTTGCCTCTCA	IDT	N/A
NANOG: Forward: CCTGTGATTTGTGGGCCTG, Reverse: GACAGTCTCCGTGTGAGGCAT	IDT	N/A
DBX1: Forward: CTGGGCCTGAAAGACTCGC, Reverse: CCGCTAGACAGGAGTTCGC
Myco-f1: Forward: CGCCTGAGTAGTACGTTCGC	IDT	N/A
Myco-f2: Forward: GCGGTGTGTACAAACCCCGA	IDT	N/A
Myco-f3: Forward: TGCCTGAGTAGTCACTTCGC	IDT	N/A
Myco-f4: Forward: CGCCTGGGTAGTACATTCGC	IDT	N/A
Myco-f5: Forward: CGCCTGAGTAGTAGTCTCGC	IDT	N/A
Myco-f6: Forward: TGCCTGGGTAGTACATTCGC	IDT	N/A
Myco-r1: Reverse: GCGGTGTGTACAAGACCCGA	IDT	N/A
Silencer™ select siRNA PRKG1: Forward: GGAUAGAGGUUCGUUUGAATT, Reverse: UUCAAACGAACCUCUAUCCCT	Ambion	4392420; assay ID s11131
Silencer ™ select SiRNA negative Control No. 1	Ambion	4390843
Software and algorithms
CellProfiler	https://cellprofiler.org/ 127	v.4.2.5.
GraphPad Prism	GraphPad Software	v.5.01
Adobe Illustrator	Adobe	v.29.6
Cell Ranger	10x Genomics	v.7.10
Ultivo triple-quadrupole mass spectrometer	Agilent Technologies	https://www.agilent.com/en/product/liquid-chromatography-mass-spectrometry-lc-ms/lc-ms-instruments/ultivo-lcms
MassHunter Software	Agilent Technologies	N/A
MToolBox		v.1 https://github.com/mitoNGS/MToolBox128
ImageJ	Fiji129	v1.54g http://imagej.org
IGV viewer	IGV	v2.163 https://igv.org
CELLCYTE Studio	Cytena	N/A
ZEN Microscopy Software	Zeiss	N/A
ND-1000 software	Thermo Fisher Scientific	v3.8.1
CFX96 software	Bio-Rad	N/A
Image Lab software	Bio-Rad Laboratories	v.6.1
GeneMapper ID	Applied Biosystems	v.3.2.1
ND-1000 software	Thermo Fisher Scientific	v.3.8.1
Xcalibur software	Thermo Fischer Scientific	N/A
TraceFinder 5.1 software	Thermo Fischer Scientific	N/A
Columbus software	Revvity	v.2.9.0
FlowJo software	FlowJo	v.7.6
OriginPro Software	OriginLab Corporation	N/A
nfcore/rnaseq	https://zenodo.org/records/7998767	v.3.12.0
Trim Galore!	Babraham Bioinformatics https://zenodo.org/records/7598955	v.0.6.10
STAR aligner	https://github.com/alexdobin/star/releases	v. 2.7.10a
Salmon	https://github.com/COMBINE-lab/salmon/releases	v.1.10.1
DESeq2 package	https://github.com/thelovelab/DESeq2	v. 1.40.2 and v.7.5.1
ClusterProfiler	DOI: 10.18129/B9.bioc.clusterProfiler	v4.8.3 and v4.10.0
org.Hs.eg.db	DOI: 10.18129/B9.bioc.org.Hs.eg.db	v.3.18.0
ggplot2	https://github.com/tidyverse/ggplot2/releases/tag/v3.5.1	v.3.5.1
R	The R Project for Statistical Computing	v. 4.4.1
Seurat	https://satijalab.org/seurat/ version: https://github.com/satijalab/seurat/releases	v. 5.1.0
AUCell package	DOI: 10.18129/B9.bioc.AUCell	v. 1.26.0
MsigDB package	https://data.broadinstitute.org/gsea-msigdb/msigdb/release/7.5.1/	v 7.5.1
dplyr	https://cran.r-project.org/web/packages/dplyr/index.html	v. 1.1.4
Dia-NN	https://github.com/vdemichev/DiaNN/releases/tag/1.8.1	v1.8.1
OmicsNet platform	https://www.omicsnet.ca/	v.2.0
CytoNCA plugin in Cytoscape	N/A	v.3.10.2
KEGG database	GenomeNet	Release 112.0
IGV viewer	https://igv.org/doc/desktop/	v2.163
Vevo 2100 analysis software	FUJIFILM VisualSonics	v.1.5
NIS-Elements Advanced Research	Nikon	v.3.2
EZ-C1 Silver	Nikon	v.3.91
MATLAB 2023b	MathWorks	https://de.mathworks.com/products/new_products/release2023b.html
IncuCyte Live-Cell Analysis System®	Sartorius	N/A
Other
CytoSmart Cell Counter	Greiner Bio-One	6749
Orbital Shaker Heidolph Unimax 1010	Heidolph	543-12310-00
Operetta® CLS™	Revvity	HH16000020
96-well Cell Culture Microplate, PS, F-Bottom, black TC, μCLEAR, 96-well black, clear bottom	Greiner Bio-One	655090
96-well Cell Culture Microplate, SCREENSTAR	Greiner	655866
BIOFLOAT 96 well plate 4PCS	FaCellitate GmbH	F202003
μ-Plate 96 Well Square	Ibidi	89626
DMS1000 microscope	Leica	N/A
Eclipse Ts2	Nikon	Inverted Microscope
ZEISS Axio Observer Apotome 3	Zeiss	ZEISS Apotome 3: Optical sectioning in widefield fluorescence microscopy
Vibratome Microm HM 650 V	Thermo Fisher Scientific	920120
Ultivo triple-quadrupole mass spectrometer	Agilent Technologies	G6465BA
EnSight multimode plate reader	Revvity	N/A
Cellcyte X	Cytena	CELLCYTE X™ - Live Cell Imager And Analyzer ∣ CYTENA
INFINITY platform	Abberior Instruments	INFINITY - @abberior.rocks
Vevo 2700 VisualSonics System	FUJIFILM VisualSonics	N/A
Promethion	Sable Systems International	Promethion Core Metabolic and Behavioral Phenotyping Systems
Infinite M1000 Pro	TECAN	Tecan ∣ Thermo Fisher Scientific - DE
Confocal laser scanning microscope C1	Nikon Mikroskope Solutions	Nikon's Digital Eclipse C1 Microscope System Delivers High Resolution Confocal Images at Sensible Price ∣ News ∣ Nikon Instruments Inc.
Mastercycler X50s	Eppendorf	6311000010
CFX96™ Real-Time System qPCR machine	Bio-Rad	CFX96 Touch Real-Time PCR Detection System ∣ Bio-Rad
ChemiDoc MP Imaging system	Bio-Rad	ChemiDoc MP Imaging System ∣ Bio-Rad
Low-attachment U-bottom 96-well plates	Corning	3474
AggreWell	STEMCELL Technologies	34815
NovaSeq 6000 system	Illumina	NovaSeq 6000 System ∣ Powerful sequencing with scalable throughput
Dionex Ultimate 3000	Thermo Scientific	UltiMate 3000 HPLC and UHPLC Systems
timsTOF SCP mass spectrometer	Bruker Daltonics	timsTOF SCP ∣ Bruker
SeQuant ZIC-pHILIC	Merck	1.50462
LSR-Fortessa X-20	Becton Dickinson	LSRFortessa™ X-20 ∣ Benchtop Flow Cytometer
384-well black-wall, clear-bottom plates	Revvity	6007460
Multidrop liquid dispenser	Thermo Fisher Scientific	https://www.thermofisher.com/de/de/home/life-science/lab-equipment/microplate-instruments/multidrop-dispensers.html
Rotating incubator Cytomat	Thermo Electron	N/A
Echo 550	LabCyte	N/A
Janus MDT	Revvity	YJLM001
Victor Multiple Plate Reader Spectrophotometer	Revvity	HH35000500
Millicell-CM inserts	Millipore	Millicell® Cell Culture Inserts - Zellkulturplatten-Einsätze
Eclipse FN-1	Nikon	ECLIPSE FN1 ∣ Upright Microscopes ∣ Microscope Products ∣ Nikon Instruments Inc.
Eclipse 90i	Nikon	upright widefield microscope
FLASH 4.0 LT camera	Hamamatsu Photonics	N/A
NanoDrop 2000	Thermo Fisher Scientific	N/A
Nanodrop Spectrophotometer ND1000	peQlab	N/A
Aurora Ultimate column	IonOpticks,	N/A
H-ESI source probe	Thermo Fisher Scientific	N/A
SeQuant ZIC-pHILIC colum	Merck	N/A
MSMLS-1EA library	Merck	N/A
1290 Infinity II HPLC	Agilent Technologies	https://www.agilent.com/en/product/liquid-chromatography/hplc-systems/gpc-sec-solutions/1290-infinity-ii-gpc-sec-system
Poroshell 120 EC-C18 column	Agilent Technologies	N/A
3,500 Series Genetic Analyzer	Applied Biosystems	RRID:SCR_021901
Seahorse XFe24 Analyzer	Agilent Technologies
Incucyte	Sartorius
Thermal camera	Dongguan Xintai Instrument Co., Ltd,	#HT-18
Chemiluminscent Western Blot imager	Azure byosistem	Azure 300

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

We obtained written informed consent to use patient material and health data from the patients and their guardians according to the Declaration of Helsinki. The use of iPSCs was approved by the local ethic committees of Charité-Universitätsmedizin Berlin (EA2/131/13 and EA2/107/14) and Heinrich Heine University Düsseldorf (Study number 2020-967_5). Compassionate treatments were approved by the local ethics committees. Patient data were pseudoanonymized. Age ranges were used instead of actual age to avoid the risk of identifying individuals. Animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Johns Hopkins University.

MT-ATP6 mutant iPSC lines were previously derived from three patients with m.9185T>C variant, one patient with m.8993T>C variant, two patients with m.8993T>G variant, and two patients with m.9176T>G variant (Figure S11H).^27,33-35. SURF1 iPSC isogenic lines and NDUFS4 iPSC isogenic lines were previously derived (Figure S11H).^25,81 Among healthy control iPSCs (Figure S11H), CTRL_1, CTRL_2, CTRL_4, CTRL_7 were previously described^27,33,119, CTRL_3 (CRMi003-A) was purchased from RUCDR Infinite Biologics, CTRL_5 (HVRDi004-B-1, PGP1) was purchased from Synthego, CTRL_6 (IUFi004-A, iPS12) was purchased from Cell Applications, and CTRL_8 (WISCi004-B) was purchased from WiCell. We cultured all iPSC lines on Matrigel (Corning)-coated plates using StemMACS iPS-Brew XF medium (Miltenyi Biotec), supplemented with MycoZap (Lonza). CTRL_8 and ATP6_2 iPSCs were cultured in mTeSR Plus Medium (STEMCELL Technologies) prior to BCEC differentiation. We cultivated all iPSCs in a humidified atmosphere of 5 % CO₂ at 37 °C and 5 % oxygen. iPSCs were passaged at 70-80 % confluence with 0.5 μM EDTA (Invitrogen) in 1 x PBA (Gibco). We added 10 μM ROCK inhibitor (Enzo Biochem Inc) after splitting to promote survival. Cultures were routinely monitored for mycoplasma contamination by PCR using nine primers to amplify the six most common mycoplasma strains (Key Resource Table). The positive control and internal control were kindly provided by Dr. Cord Uphoff (DZMS, Germany). We monitored the identity of iPSCs using STR analysis, which was performed by the Forensic Department of the University Hospital Düsseldorf (Dr. phil. nat. Petra Böhme); 21 microsatellite loci were amplified using PCR and tagged products were analyzed using GeneMapper ID v.3.2.1 (Applied Biosystems).

METHOD DETAILS

Neural progenitor cells (NPCs) and differentiated neurons

We obtained NPCs from iPSCs using our previously published protocol.¹²⁰ Briefly, we detached iPSCs from Matrigel-coated plates using Accutase (Sigma-Aldrich) and transferred the collected cells into low-attachment 6-well plates where they were kept for two days in: KnockOut-DMEM (Thermo Fisher Scientific), KnockOut-SR (Thermo Fisher Scientific), non-essential amino acids (NEAA) (Thermo Fisher Scientific), plus 1 mM pyruvate (Thermo Fisher Scientific) 2 mM L-glutamine (Thermo Fisher Scientific), 1 x MycoZap Plus-CL (Lonza) with the addition of 0.5 μM purmorphamine (PMA) (Merck Millipore), 3 μM CHIR 99021 (Cayman Chemical), 10 μM SB-431542 (Selleckchem), and 1 μM dorsomorphin (Sigma-Aldrich). From day 2 to day 4, the media was switched to: Neurobasal:DMEM/F12 [1:1], 0.5 x N2, 0.5 x B27 without vitamin A, 1 x MycoZap Plus-CL, with the addition of 0.5 μM purmorphamine (PMA) (Merck Millipore), 3 μM CHIR 99021 (Cayman Chemical), 10 μM SB-431542 (Selleckchem), and 1 μM dorsomorphin (Sigma-Aldrich). On day 6, we transferred the suspended cells onto Matrigel-coated well plates using: the same media without SB-431542 and dorsomorphin, but with the addition of 150 μM ascorbic acid (Sigma-Aldrich). We maintained NPCs on this media and used them for experiments between passage 7 and passage 30.

Differentiation into neuronal cultures containing dopaminergic neurons was performed as previously described.²⁵ Starting from NPCs at low confluence (10-30 %), we switched to media containing Neurobasal:DMEM/F12 (1:1), 0.5 x N2, 0.5 x B27 with vitamin A, 1 x MycoZap Plus-CL, with the addition of 200 mM vitamin C, 100 ng/ml FGF8 (R&D Systems) and 1 mM PMA. After 8 days, we used the media Neurobasal:DMEM/F12 (1:1), 0.5 x N2, 0.5 x B27 with vitamin A, 1 x MycoZap Plus-CL, supplemented with 200 mM vitamin C, 0.5 mM PMA, 500 mM cAMP (STEMCELL Technologies), 10 ng/mL BDNF (MACS Miltenyi), 10 ng/ml GDNF (MACS Miltenyi) and 1ng/mL TGFbeta3 (MACS Miltenyi). After ten days, the medium was changed to Neurobasal:DMEM/F12 (1:1), 0.5 x N2, 0.5 x B27 with vitamin A, 1 x MycoZap Plus-CL, supplemented with 200 mM vitamin C, 500 mM cAMP, 10 ng/ml BDNF, 10 ng/ml GDNF, and 1 ng/mL TGFbeta3.

Cortical brain organoids

Cortical brain organoids were generated following two protocols (Figure S7A). The first protocol was based on our modifications⁵⁷ of a previous publication¹²¹. Briefly, on day 0, iPSCs at 80 % confluence were seeded into low-attachment U-bottom 96-well plates (Corning) to induce neurosphere formation. iPSCs were washed with PBS (Gibco) and detached with Accutase (Gibco) for 5 min at 37 °C. Cell pellets were resuspended in cortical differentiation medium I (CDMI) consisting of Glasgow-MEM, 20 % Knockout Serum Replacement, MEM-NEAA, sodium pyruvate, 2-mercaptoethanol and penicillin-streptomycin (all from Gibco). After cell counting, the suspension was diluted with CDMI to a final cell concentration of 90,000 cells per ml and supplemented with 20 μM ROCK inhibitor (Enzo), 5 μM TGF-β inhibitor SB431542 (Cayman Chemical Company) and 3 μM WNT antagonist IWR1 (EMD Millipore Corp). The supplemented seeding suspension was distributed to low attachment U-bottom 96-well plates by adding 100 μl per well and incubated at 37 °C and 5 % CO₂. On day 3, the sides of the 96-well plates were carefully tapped to detach dead cells, and then 100 μl of CDMI supplemented with 20 μM ROCK inhibitor, 3 μM IWR1 and 5 μM SB431542. On day 6, 9, 12 and 15, 80 μl of the supernatant medium were removed from each well and replaced with 100 μl of CDMI supplemented with 3 μM IWR1 and 5 μM SB431542 per well. On day 18, the developing organoids were transferred to 100 mm petri dishes filled with cortical differentiation medium II (CDMII) consisting of DMEM/ F12, 1 % GlutaMAX, 1 % N2 supplement, 1 % chemically defined lipid concentrate and penicillin-streptomycin (Gibco). The petri dishes were placed on an orbital shaker at 70 rpm at 37 °C and 5 % CO₂. CDMII was changed every second day. On day 35, CDMII was replaced with cortical differentiation medium III (CDMIII) that in addition contains 10% FBS (Gibco) and heparin (Merck). CDMIII was changed every two to four days, depending on the colour of the medium. From day 70 on, organoids were kept in cortical differentiation medium IV (CDMIV) that additionally was supplemented with 1 % B27 with vitamin A.

The second protocol was based on a different publication⁵⁸. For this, AggreWell plates (STEMCELL Technologies) were prepared by incubation with Anti-Adherence Rinsing Solution (STEMCELL Technologies) for 15 min. iPSCs were detached from the 6-well plates using Accutase as described above. The cell suspension was pelleted at 200 x g for 4 min and resuspended in iPS Brew (Miltenyi Biotec, Germany) with 10 μM ROCK inhibitor. 2.75 million cells were seeded in 1.5 ml per AggreWell. AggreWell plates were centrifuged at 100 x g for 3 min and left at 37 °C and 5 % CO₂ for 24 h. On day 1, spheroids were transferred to a petri dish using a cut P1000 tip and placed on a shaker. Medium was changed to embryonic stem (ES) medium composed of KO-DMEM, 20% Knockout Serum Replacement, MEM-NEAA, sodium pyruvate, GlutaMAX, penicillin-streptomycin (all from Gibco), and MycoZAP Plus-CL (Lonza), and supplemented with 10 μM SB431542 and 2.5 μM dorsomorphin (Sigma). Until day 6, ES medium was changed daily. Neurospheres were fed with Neural Differentiation Medium (NDM) consisting of Neurobasal A medium (Gibco), GlutaMAX, penicillin-streptomycin, MycoZAP and 2 % B27 without vitamin A (Gibco). From day 6-15, medium was changed daily and supplemented with 20 ng/ml EGF (R&D Systems) and 20 ng/ml FGF2 (R&D Systems). From day 16-21, medium was changed every second day, and from day 22-45, NDM medium was supplemented with 20 ng/ml BDNF (MACS Miltenyi), 20 ng/ml NT-3 (Peprotech), 200 μM (+)-sodium L-ascorbate (Sigma-Aldrich), 50 μM dibutyryl cAMP (STEMCELL Technologies), and 10 μM DHA (Sigma), and changed every other day. For long-term maturation after day 46, NDM medium was prepared with 2 % B27 with vitamin A, and organoids were fed according to their needs. Images of brain organoids were acquired using the Eclipse Ts2 light microscope (Nikon), and at later growth stages, when the organoids were larger, using a DMS1000 microscope (Leica). The size of the organoids was determined by area calculation using ImageJ.

Assessment of mitochondrial membrane potential (MMP)

For MMP quantification, we applied a high-content analysis (HCA)-based live-cell detection assay that we previously established.³¹ Briefly, we seeded NPCs onto black-wall, clear-bottom 96-well plates (SCREENSTAR, Greiner) pre-coated with Geltrex (Thermo Fisher Scientific) at a density of 1.5 x 10⁵ cells/cm², and incubated them in NPC medium overnight at 37 °C, 5 % CO₂. The next day, NPCs were treated with either DMSO or Sildenafil dilution series in 100 μl and incubated for 16 h at 37°C, 5% CO₂. On the day of the assay, we live-stained NPCs with 10 nM TMRM (Molecular Probes, Life Technologies) and 1 μg/μl Hoechst (33342, Thermo Fisher Scientific) in NPC medium for 30 min at 37 °C and 5 % CO₂. After 30 min incubation, cells were washed with 100 μl NPC medium without phenol red. NPCs were kept in 100 μl medium (w/o phenol red) for the duration of the assay. Imaging was performed with an Operetta CLS high content imaging system (Revvity) with 20 x objective in confocal mode. Imaging time did not exceed 30 min. TMRM and Hoechst 33342 intensity within the cells was calculated using Columbus software (Revvity, version 2.9.0). To asses the effect of KT5823 on MMP in presence and absence of sildenafil, NPCs were seeded, incubated as described above, and pre-treated for 1 h with either DMSO alone or 1 μM KT5823 dissolved in DMSO. After the incubation time, 10 μM sildenafil dissolved in DMSO was added on top and incubated for 16 h at 37°C, 5% CO₂. The final DMSO concentration was adjusted to 0.3 % for each condition. On the day of the assay, NPCs were live-stained and imaged. TMRM and Hoechst 33342 intensity within the cells was analyzed with CellProfiler (version 4.2.5).

For MMP assessment using cytofluorimetry, we used LSR-Fortessa X-20 (Becton Dickinson) to monitor the fluorescence of TMRM (Invitrogen) following treatment with either 10 μM sildenafil or DMSO. Cells were harvested and resuspended in culture medium at approximately 1x10⁶ cells/ml. The uncoupling agent carbonyl cyanide 3-chlorophenylhydrazone (CCCP, 50 μM final concentration) was added to the sample and incubated for 5 min at 37°C, 5% CO₂ to depolarize mitochondria. TMRM probe (20 nM final concentration) was added and incubated for 30 min at 37 °C and 5 % CO₂. After incubation, cells were washed once in 1 ml of culture medium, then resuspended in 300 μl of D-PBS and analyzed on the flow cytometer. The TMRM-emitted fluorescence was detected by emission filters appropriate for R- phycoerythrin “PE” (585 nm ± 42 nm). An appropriate gating strategy was applied to select only live and single cells. Data were analyzed with FlowJo software (version 7.6); the median fluorescence intensity (MFI) for each sample was calculated and normalized on the relative unstained sample. TMRM intensity was quantified in at least 20,000 cells per sample; at least three biological replicates were analyzed. Statistical analysis was performed by one-way Anova. p < 0.05 was considered statistically significant.

Compound screening

LS NPCs (ATP6_2) were seeded at a density of 1.5 x 10⁵ cells/cm² onto Geltrex-coated 384-well black-wall, clear-bottom plates (Revvity) in 30 μl NPC medium using a Multidrop liquid dispenser (Thermo Fisher Scientific), which was calibrated prior to each usage. Assay plates with cells were placed into a rotating incubator (Thermo Electron, Cytomat) at 37 °C and 5 % CO₂ for 24 h. Cells were treated with 5 μM (0.05 % DMSO) of 5,632 different compounds from a drug repurposing library.³⁷ Compounds were transferred to 384-well assay plates using an Echo 550 (LabCyte). Columns 23 and 24 contained controls on each plate. In row 23, a combination of 5 μM FCCP (Biozol) and antimycin A (Sigma/Merck) was used as positive control to achieve complete mitochondrial depolarization. Row 24 contained 0.05 % DMSO as negative control. After incubation of 16 h, cells were live-stained for 30 min at 37 °C and 5 % CO₂ with a final concentration of 10 nM TMRM and 1 μg/μl Hoechst in 30 μl/well. Cells were then automated washed twice with 30 μl NPC medium (w/o phenol red) using the Janus MDT (Revvity). Images were directly acquired using an Operetta CLS high content imaging system with 20 x objective in confocal mode. Imaging time did not exceed 30 min. TMRM intensity within the cells and nuclei count was calculated using a customized Columbus software script and downstream data processing was performed with Activity Base (IDBS). To quantify the effects of compounds on depolarizing the hyperpolarized MMP of LS NPCs (ATP6_2), we normalized the percentage of MMP depolarization (TMRM signal) and the percentage of cell count (Hoechst signal) after treatment for 16 h with 5 μM of each compound in LS NPCs with respect to LS NPCs treated for 16 h with either 5 μM FCCP+AA or DMSO only. To achieve this normalization, we used the formula z = (y − min(x)) / (max(x) − min(x)) * 100, where z is the final normalized value, y is the value of interest in the dataset, min(x) is the mean value of DMSO-treated LS NPCs, and max(x) is the mean value of FCCP+AA-treated LS NPCs. We next scored the normalized values (z) based on their ability to depolarize the MMP (inhibition) and classified the compounds in mild, intermediate, and strong uncouplers.

Bioenergetic assessment

For biochemical activity of complex V, approximately 1 x 10⁷ NPCs were suspended in 0.5 ml of 10 mM ice-cold hypotonic Tris buffer (pH 7.6) and homogenized with glass/glass Dounce homogenizer with a tight pestle by 10 strokes. 200 μl of 1.5 M sucrose was added and then samples were centrifuged at 600 g for 10 min at 4 °C. Supernatants were collected and centrifuge at 14,000 g for 10 min at 4 °C. Mitochondrial pellets were suspended in 0.2 ml of 10 mM ice-cold hypotonic Tris buffer (pH 7.6) and subjected to three freeze-thaw cycles. Protein concentration was quantified with the Bicinchoninic Acid (BCA) protein assay Kit (Thermo Fisher Scientific). Complex V activity (ATP hydrolysis) was measured at 340 nm cleared from oligomycin-insensitive ATPase activities as described⁴³ and normalized to citrate synthase activity.¹²²

For ATP content analysis, NPCs were grown in NPC medium and then switched to DMEM glucose-free (Gibco) and Neurobasal A-Medium glucose-free (Gibco) 1:1 medium supplemented with 1 mM sodium pyruvate, 1 % B27, 0.5 % N2, 1 % Pen/Strep, 1 % L-glutamine and 0.01 % MycoZap Plus-CL with 5 mM galactose as carbon source to measure OXPHOS-derived ATP-content for 24 h. Luminescence-based ATP levels were detected in cellular suspension (5 x 10³ cells) with ViaLightTM Plus Kit (Lonza). ATP levels were determined by luminometry using Victor Multiple Plate Reader Spectrophotometer (Revvity). For normalization purposes, we employed the DNA quantification with CyQUANT Cell Proliferation Assay (Invitrogen). Results were then expressed as luminescence A.U. relative to control lines cultured in galactose.

For live-cell assessment of cellular bioenergetics, we employed Seahorse XF96 and XFe24 extracellular flux analyzer, as previously described.²⁵ Briefly, the day before the assay, 100,000 cells (for XF96) or 120,000 (for XFe24) were plated into each Matrigel-coated well of the XF96 well plates and incubated overnight at 37 °C with 5 % CO₂. On the day of the assay, the growth medium was removed and replaced with glucose-based unbuffered media. The cells were incubated at 37 °C for 60 min to allow media temperature and pH to reach equilibrium before starting the measurement of mitochondrial respiration (oxygen consumption rate, OCR). After baseline records, we subsequently injected oligomycin (1 μM), FCCP (1 μM), and rotenone/antimycin A (1 μM each). Normalisation to DNA content in each well of the plate was performed using the CyQUANT Cell Proliferation assay Kit.

Quantification of NAD metabolites

Quantitative analyses of NAD+, NADH, NADP+, and NADPH were carried out as a service in NADMED laboratory (Helsinki, Finland) from same homogenates submitted for metabolomics analysis. NPCs were washed on the plate with PBS buffer to remove protein of the culture media followed by the addition of cold extraction solvent acetonitrile:methanol:MilliQ; 40:40:20 to quench cellular metabolism. Obtained sample homogenates were shipped on dry ice to the measurement facility. Before analysis the homogenates were equilibrated to room temperature and centrifuged at 20,000 x g for 10 min at 4 °C to remove proteins. Next, NAD+, NADH, NADP+, and NADPH were measured individually from every cell extract using modified cyclic enzymatic reactions with colorimetric detection. For normalization of the results, protein content was measured using the pellets obtained after centrifugation of the homogenate.

Bulk RNA sequencing

For bulk transcriptomics of NPCs, we used four MT-ATP6 mutant NPC lines (ATP6_2, ATP6_4, ATP6_5 and ATP6_7.) and four healthy control NPC lines (CTRL_1, CTRL_2, CTRL_3 and CTRL_4) either treated for 16 h with 0.1 % of DMSO alone or with 10 μM sildenafil resuspended in 0.1 % DMSO. We used n=3 biological replicates per condition. Total RNA was isolated from pelleted NPCs using the NucleoSpin RNA Plus kit (Macherey Nagel) and eluted in 30 μl RNase-free H₂O. RNA concentration was measured with NanoDrop 2000. For quality check, 300 ng RNA was run on a 2 %TBE gel.

For bulk transcriptomics of NPCs in low glucose media, we used three MT-ATP6 mutant NPC lines (ATP6_2, ATP6_4, and ATP6_7.) either treated for 6 or 24 h with 0.1 % DMSO alone, 1 μM sildenafil resuspended in 0.1 % DMSO, or 10 μl sildenafil resuspended in 0.1 % DMSO. We used n=3 biological replicates per condition. Total RNA was isolated from pelleted NPCs using the NucleoSpin RNA Plus kit (Macherey Nagel) and eluted in 30 μl RNase-free H2O. RNA concentration was measured with NanoDrop 2000. For quality check, 300 ng RNA was run on a 2 %TBE gel. The values depicted in the boxplot (log2 Fold Change, y axis) in Figure S6G are the log2 Fold Changes between the treated vs untreated (0.1 % DMSO) within each cell line, computed using DESeq2. P-values were computed with a two-tailed t-test using these log2 Fold changes and adjusted with the Benjamini-Hochberg method.

For bulk transcriptomics of brain organoids, total RNA was isolated from day 70 cortical brain organoids generated using the first protocol.^57,121 We used five pelleted organoids per sample with n=3 biological replicates per condition, grown under normal conditions (CTRL_1, CTRL_2, ATP6_4 and ATP6_7), or treated with 10 μM sildenafil for 24 h (ATP6_4 and ATP6_7). Total RNA was isolated using the RNeasy Mini Kit by Qiagen. RNA concentration as well as purity was measured at the Nanodrop Spectrophotometer ND1000 (peQlab) using ND-1000 software (V3.8.1). Total RNA was mixed with 1 μg of a DNA oligonucleotide pool comprising 50-nt long oligonucleotide mix covering the reverse complement of the entire length of each rRNA (28S rRNA, 18S rRNA, 16S rRNA, 5.8S rRNA, 5S rRNA, 12S rRNA), incubated with 1U of RNase H (Hybridase Thermostable RNase H, Epicentre), purified using RNA Cleanup XP beads (Agencourt), DNase treated using TURBO DNase rigorous treatment protocol (Thermo Fisher Scientific) and purified again with RNA Cleanup XP beads. rRNA-depleted RNA samples were further fragmented and processed into strand-specific cDNA libraries using TruSeq Stranded Total LT Sample Prep Kit (Illumina) and sequenced using the NovaSeq 6000 system with stranded technology, generating paired-end reads of 150 bp. Raw RNA sequencing data were processed to generate FASTQ files using Illumina bcl2fastq. Briefly, base calling was performed using Real-Time Analysis (RTA) software, and adapter sequences were trimmed. The resulting BCL files were converted to FASTQ format, followed by quality control assessment using FastQC before downstream analysis. The preprocessing of the FASTQ raw sequencing files was performed with nfcore/rnaseq (version 3.12.0) pipeline. Briefly, reads were trimmed with Trim Galore! (version 0.6.10) and mapped to the human genome (GRCh38 assembly) using STAR (version 2.7.10a) aligner. We performed different comparisons: i) LS with DMSO vs. controls with DMSO (disease signature), ii) LS + sildenafil vs. LS + DMSO (sildenafil signature), iii) controls + sildenafil vs. controls + DMSO (negative control test). Transcript counts were determined using Salmon (version 1.10.1). DESeq2 (version 1.40.2)¹²³ was used to identify differentially expressed genes in the NPC samples (Table S1) or in brain organoid samples (Table S4) (Benjamini-Hochberg-adjusted p-value 0.05). Gene Ontology (GO) enrichment analysis was performed using the enrichGO function within ClusterProfiler (4.8.3), using the annotations of org.Hs.eg.db (version 3.18.0). Visualizations were generated with ggplot2 (version 3.5.1).

Single-nucleus RNA sequencing

For single-nucleus RNA sequencing (snRNAseq), we used day 72 brain organoids generated with the second protocol¹²⁴ from control iPSCs (CTRL1) and LS iPSCs (ATP6_7). Brain organoids were treated with either DMSO or 10 μM sildenafil resuspended in DMSO for 45 days from day 27 to day 72. We used 15 pelleted organoids per sample (n=8 samples) with n=2 biological replicates per condition (n=120 cortical brain organoids in total). snRNA-seq was performed using the 10X Genomics Chromium system, following the manufacturer's protocol. Nuclei were isolated using mechanical dissociation with gradient centrifugation and stained with DAPI to assess nuclear integrity. Barcoding and cDNA amplification were carried out using the Chromium Single Cell 3' (vNext) Reagent Kit. Libraries were prepared and sequenced on the Illumina NovaSeq 6000, generating paired-end reads of 100 bp. Data processing, including demultiplexing, alignment to the reference genome (GRCh38-2020-A), and gene quantification, was performed using Cell Ranger (v7.2.0) from 10x Genomics with default parameters. To generate digital gene expression (DGE) matrices for each sample. DGEs were further processed in R (v. 4.4.1) using Seurat (v. 5.1.0).¹²⁵ Filtered DGEs were imported using the function “Read10x” and only genes detected in at least 5 cells were kept for downstream analyses. Similarly, cells will less than 200 genes and 3,000 unique transcripts or more than 0.5 % mitochondrial transcripts were discarded. After merging all samples in a single Seurat object, we performed gene expression normalization and scaling for each sample independently using SCTransform.¹²⁶ We performed unbiased clustering and dimensionality reduction using the first 20 principal components on all samples together. Clusters were manually annotated in 7 cell types analyzing their marker genes (Table S5). To identify differential genes induced by MT-ATP6 variants and sildenafil treatment in robust and reproducible way, we performed a pseudo-bulk analysis using the DESeq2 package¹²⁶ (v. 7.5.1), as described previously.¹²⁷ Pseudo-bulk analysis was used to determine differentially expressed genes within the individual populations (Table S6). For each population, we generated a pseudo-bulk expression profile for each organoid by summing the expression of 250 randomly selected cells. We filtered genes with less than 50 counts in at least 2 samples, normalized expression values by library size, estimated negative binomial dispersions. We performed different comparisons: i) LS + DMSO vs. controls + DMSO (disease signature), ii) LS + sildenafil vs. LS + DMSO (sildenafil signature), iii) controls + sildenafil vs. controls + DMSO (negative control test). For all comparisons, we identified differentially expressed genes and adjusted p value was calculated using Bonferroni correction for multiple testing correction. To estimate glycolytic pathway activity in single cells, we leveraged the AUCell package (v. 1.26.0)¹²⁸ and the ‘Glycolysis’ gene set in the ‘Hallmark’ database accessed from the MsigDB package (v 7.5.1).¹²⁹ Plots were generated with ggplot2 (v. 3.5.1) and piping using dplyr (v. 1.1.4).

Proteomics analysis

We carried out proteomics using mass spectrometry (MS) for four MT-ATP6 mutant NPC lines (ATP6_2, ATP6_4, ATP6_5, and ATP6_7) and four healthy control NPC lines (CTRL_1, CTRL_2, CTRL_3, and CTRL_4) with n=3 biological replicates per condition that were treated with either DMSO or 10 μM sildenafil in DMSO for 16 h. Cells were washed three times with ice-cold PBS, detached using a cell scraper, transferred to pre-chilled 1.5 ml tubes and centrifuged at 1,000-3,000 rpm for 10 min at 4 °C. Subsequently, the supernatant was removed, and intact cells were snap-frozen. Cells were lysed under denaturing conditions in 300 μl of a buffer containing 3 M guanidinium chloride (GdmCl), 10 mM Tris(2-carboxyethyl)phosphine (TCEP), 40 mM chloroacetamide, and 100 mM Tris-HCl pH 8.5. Lysates were denatured at 95 °C for 10 min shaking at 1,000 rpm in a thermal shaker and sonicated in a water bath for 10 min. The protein concentration of each sample was measured with a BCA protein assay kit (23252, Thermo Scientific). 500 ng protein was used per sample and diluted with a dilution buffer containing 10 % acetonitrile and 25 mM Tris-HCl, pH 8.0, to reach a 1 M GdmCl concentration. Then, proteins were digested with LysC (Roche; enzyme to protein ratio 1:50, MS-grade) shaking at 800 rpm at 37 °C for 3.5 hours. The digestion mixture was diluted again with the same dilution buffer to reach 0.5 M GdmCl, followed by tryptic digestion (Roche, enzyme to protein ratio 1:50, MS-grade) and incubation at 37 °C overnight in a thermal shaker at 800 rpm. Peptides were acidified with formic acid to a final concentration of 2 %. LC-MS/MS was performed by nanoflow reversed-phase liquid chromatography (Dionex Ultimate 3000, Thermo Scientific) coupled online to a timsTOF SCP mass spectrometer (Bruker Daltonics) using the data-independent acquisition (DIA) method with parallel accumulation serial fragmentation (PASEF). Briefly, LC separation was performed using the Aurora Ultimate column (25 cm x 75 μm ID, C18, 1.7 μm beads, IonOpticks, Victoria, Australia). 150 ng of each desalted digest was applied to the column and peptides were eluted using a gradient of 3.8 to 38 % solvent B in solvent A over 60 min (total run time) at a flow rate of 400 nl per minute. Solvent A was 0.1 % formic acid and solvent B was 79.9 % acetonitrile, 20 % H₂O, and 0.1% formic acid. MS data were processed with Dia-NN (v1.8.1) and searched against an in silico predicted human spectral library. The "match between run" feature was used and the mass search range was set to m/z 400 to 1,000.

Targeted metabolomics profiling

We carried out metabolomics using MS of four MT-ATP6 mutant NPC lines (ATP6_2, ATP6_4, ATP6_5 and ATP6_7) and four healthy control N PC lines (CTRL_1, CTRL_2, CTRL_3 and CTRL_4) using n=4 replicates per condition that were treated with either 0.1 % DMSO or 10 μM sildenafil in 0.1 % DMSO for 16 h. NPCs were washed 1 x with PBS, detached in 600 μl ice-cold extraction buffer (methanol, acetonitrile and Milli-Q water at a ratio of 40:40:20) using a cell scraper, transferred into an ice-cold 1.5 ml tube, and stored at −80 °C until use. Metabolites were extracted from cells with 500 μl of cold extraction solvent (Acetonitrile:Methanol:MQ; 40:40:20, Thermo Fischer Scientific). Subsequently, samples were processed with three cycles of sonication (60 s) and vortexing (120 s) followed by centrifugation at 14,000 rpm at 4 °C for 5 min. Next, the samples were centrifuged, the supernatants transferred to evaporation tube and evaporated to dry under nitrogen stream. Samples were reconstituted in 40 μl extraction buffer (40:40:20; acetonitrile:methanol:MilliQ) and transferred to LC-MS vials. 2 μl of the samples were analyzed with Thermo Vanquish UHPLC coupled with Q-Exactive Orbitrap mass spectrometer equipped with a heated electrospray ionization (H-ESI) source probe (Thermo Fischer Scientific). A SeQuant ZIC-pHILIC (2.1 × 100 mm, 5 μm particle) column (Merck) was used for chromatographic separation. The gradient elution was carried out with a flow rate of 0.100 ml/min and mobile phase gradient with 20 mM ammonium hydrogen carbonate, adjusted to pH 9.4 with ammonium solution (25 %) as mobile phase A and acetonitrile as mobile phase B, 0-2 min 80 % B, 2-17 min 80-20 % B, 17-24 min 80 % B. The column oven and auto-sampler temperatures were set to 40 ± 3 °C and 5 ± 3 °C, respectively. Following setting were used for MS: full scan range: 55-825 m/z, polarity switching; resolution of 35,000, the spray voltages: 4250 V for positive and 3250 V for negative mode; the sheath gas: 25 arbitrary units (AU); the auxiliary gas: 15 AU; sweep gas flow 0; capillary temperature: 275 °C; S-lens RF level: 50.0. Instrument control was operated with the Xcalibur software (Thermo Fischer Scientific). The metabolite annotation and integration were done with the TraceFinder 5.1 software (Thermo Fischer Scientific) using confirmed retention times by in-house standard library (MSMLS-1EA, Merck) and their m/z. The data quality was monitored throughout the run using pooled QC sample prepared by pooling 5 μL from each suspended sample and interspersed throughout the run every 10 samples. The data was quality controlled for peak quality (poor chromatography), prefiltered with 20 % RSD cutoff of the pooled QC and noise. We identified metabolites that were differentially present within samples (Table S3).

Multi-omics integration

We performed multi-omics integration as previously described.⁵² We integrated bulk transcriptomics, proteomics, and metabolomics datasets derived from LS NPCs and control NPCs (disease signature), as well as transcriptomics and metabolomics datasets from sildenafil-treated LS NPCs and untreated LS NPCs (sildenafil signature). For both cases, we used all significantly deregulated molecules (adjusted p-value ≤ 0.05) as input lists for the OmicsNet platform (v.2.0) and retrieved their interactions based on STRING (confidence score 0.7) and Recon3D databases. Molecular interactions were filtered with the Prize-Collecting Steiner Forest (PCSF) algorithm, retaining only the most informative nodes regarding network organization and functionality. The final network was analyzed for centrality measures using the CytoNCA plugin in Cytoscape (v.3.10.2) and enrichment analysis using the KEGG database (Release 112.0). For network visualization, we chose the Organic Layout from yFiles plugin in Cytoscape. To dissect the rescue mechanism of sildenafil, we searched for molecules with reversed expression in the disease signature and sildenafil signature datasets. Molecules were annotated in the REACTOME and GeneOntology databases for pathway and biological processes enrichment analyses, respectively, using the clusterProfiler package (v4.10.0).

To explore the regulatory effect of sildenafil in gene expression levels, we retrieved the drug’s signatures from the L1000 FWD database. Genes that had a consistent pattern of re-regulation across all examined cell lines after sildenafil treatment were considered as literature-indicated drug targets. We evaluated these literature-indicated drug targets in our transcriptomics datasets by filtering those genes that had the same pattern of re-regulation in LS NPCs after sildenafil administration (LS + sildenafil vs. LS + DMSO). A final list of literature-indicated drug targets with a similar regulation pattern in treated LS NPCs was considered a cross-validated set of genes involved in the mechanism of action of sildenafil. To describe the mechanistic effect of sildenafil on HOXA5, we performed two parallel clustering analyses in LS + DMSO vs. controls + DMSO and in LS + sildenafil vs. LS + DMSO networks, respectively. Specifically, we implemented the GLay algorithm aiming to identify the densely connected clusters within each network. By isolating the clusters with HOXA5 and comparing them using the Dynet plug-in in Cytoscape, we highlighted their common and/or condition-specific interactions. Each cluster’s interactions were used as inputs for enrichment analysis of biological pathways and terms. Following this pipeline, we identified the modulated pathways of HOXA5 due to sildenafil administration.

Neurite outgrowth quantification

We quantified neuronal outgrowth capacity following our HCA-based protocol.⁷³ Briefly, neurons were split with 500 μl of Accutase at 37 °C for 10-15 min and were centrifuged for 5 min at 120 x g. All pipetting steps to dissociate the cells were performed slowly and gently to minimize damage to the more sensitive neurons. 10 μM ROCK inhibitor was added to the final media after splitting to promote neuron survival. Experiments and quantification were performed as previously described.⁷³ Neurons generated for neurite outgrowth quantification were treated with 10 μM sildenafil in 0.1 % DMSO or 0.1 % DMSO alone following media changes on days 8, 10, and 12. On day 14, the neurons were split and seeded onto black-wall, flat-clear-bottom 96-well plates (Greiner) at a density of 5,000 cells/well, and treated once more with Sildenafil or DMSO. Neurons were then cultured until day 16, when they were fixed with PFA, blocked, and stained for TUJ1 and Hoechst. Neurite images were then acquired using the Operetta CLS high content imaging system (Revvity) the TUJ1 and Hoechst signals at 20 x. Following acquisition, the neurite images were analyzed using CellProfiler (version 4.2.5). Briefly, the Hoechst signal was used to identify the nucleus as a primary object and the TUJ1 signal was used to identify the neurites as secondary objects. The nuclear objects were then expanded to serve as a proxy for the cell body and subtracted from the TUJ1 signal to leaving only the individual neurites. The results were exported as CSV for data analysis.

For PRKG1 knock-down (KD) in dopaminergic-enriched neuronal cultures, we used small interfering RNA (siRNA) against PRKG1 using jetPRIME transfection (Polyplus, Cat#101000027) following the manufacturer’s instruction. On day 12, neurons were transfected with 10 nM of PRKG1 siRNA (Cat#4392420, Ambion) or scramble siRNA (Cat#4390843, Ambion) by adding the transfection mix directly to the well. On day 14, after re-seeding the neurons onto the 96-well plates, neurons were transfected again with 10 nM of PRKG1 siRNA or scramble siRNA. On day 16, PRKG1 knock-down efficiency was evaluated by qPCR and neurite length was quantified with our HCA protocol.⁷³

Calcium imaging in cortical brain organoid slices (cBOS)

cBOS were generated as previously described.^64,130 Briefly, 3 % low melting point agarose (Gibco) was diluted in Hank’s Balanced Salt Solution (HBSS, Sigma), heated to 90 °C and slowly cooled down to 40 °C. Brain organoids were washed in HBSS, embedded in the agarose, and cooled on ice until the agarose was solid. For slicing, a Vibratome Microm HM 650 V (Thermo Fisher Scientific) was adjusted to the following settings: frequency of 60 Hz, amplitude of 1.0 mm, velocity of 33 mm/s, and slice thickness of 300 μm. The slices were directly transferred to a petri dish filled with HBSS, placed in a cell culture hood and washed four times. Finally, the slices were collected on Millicell-CM inserts (Millipore) and placed in a 6-well plate. 750 μl of cortical differentiation medium IV (CDMIV) were added in each well, so that the membrane was soaked but the top side of the slices was still exposed to air. cBOS were cultured at 36 °C and 5 % CO₂ for around 30-59 days before starting calcium imaging experiments.

For calcium imaging experiments, membrane permeable Oregon Green 488 BAPTA-1 AM (OGB-1, Invitrogen) was first solved in 20 % pluronic and 80 % DMSO and diluted to 200 μM in HEPES-buffered saline, which was then injected into cBOS and incubated for 30 minutes. During the incubation and throughout the experiments, cBOS were perfused with artificial cerebrospinal fluid (ACSF), containing 138 mM NaCl, 2.5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 1.25 mM NaH₂PO₄, 18 NaHCO₃, and 10 glucose. The ACSF was bubbled with 95 % O₂ and 5 % CO₂, resulting in a pH of 7.4. OGB-1 was excited at 488 nm and signals were detected using the imaging software NIS-Elements and a variable scan digital imaging system (Nikon) attached to an upright microscope (Eclipse FN-1, Nikon). The microscope was equipped with a Fluor 40x/ 0.8 DIC M/N2 ∞ /0 WD 2.0 water immersion objective (Nikon) and an orca FLASH 4.0 LT camera (Hamamatsu Photonics). Images were routinely obtained at 1 Hz and emission was collected at >500 nm. Regions of interest (ROIs) representing cell somata were identified, and their signals were background-corrected. Additionally, the fluorescence signals were corrected for bleaching and analysed with OriginPro Software (OriginLab Corporation). To probe for the cellular response to acute metabolic stress, ACSF was switched to glucose-free ACSF containing 5 mM sodium azide (Honeywell) and 2 mM 2-Deoxy-D-glucose (Apollo Scientific) for 2 min. Then, it was switched back to ACSF for 20 to 30 min until cellular calcium levels recovered to their original baseline. Analysis of calcium signals was performed as described before.⁶⁴

Calcium imaging in NPCs

One day prior to imaging, NPCs were seeded on FluoroDishes^™ at a density of 1.5 million cells/dish. On the day of recording, the cells were washed twice with Tyrode’s solution (129 mM NaCl, 2 mM CaCl2, 1 mM MgCl2, 5 mM KCl, 30 mM D-glucose, 25 mM HEPES, pH 7.4) and incubated with 3 μM fura-2 acetoxymethyl ester (fura-2 AM, Thermo Fischer) in Tyrode’s solution with 0.01% (v/v) Pluronic F-127 (Sigma-Aldrich) for 20 min at RT in the dark. After washing with Tyrode’s solution, the cells were incubated for 10 min at RT in Tyrode’s solution to allow fura-2 de-esterification and equilibration. Next, the Tyrode’s solution was replaced by a Ca2+-free Tyrode’s solution (129 mM NaCl, 1 mM MgCl2, 5 mM KCl, 30 mM D-glucose, 25 mM HEPES, 0.5 mM EGTA, pH 7.4), and immediately placed in a temperature- and CO₂-controlled (37 °C, 5 % CO₂, and ~20 % O₂) sample holder attached to the stage of an Axio Observer 7 inverted microscope (Zeiss, Jena, Germany) equipped with a x40 objective (F-Fluar M27; NA 1.3; oil immersion; Zeiss) and x1 magnification changer (Tubelens Optovar; Zeiss). Fura-2 was alternatingly excited at 340 nm and 380 nm using a Sutter Lambda DG5 wavelength switcher (Sutter Instruments, Novato, CA, USA), FT 430 dichroic mirror, and BP525/50 emission filter. Fura-2 emission signals were captured in the dark using an Axiocam 702 camera (Zeiss), applying an exposure time of 20 ms and image acquisition interval of 1 s. Microscopy hardware was controlled by Zen Pro software (version 3.2; Zeiss). During time lapse recordings, thapsigargin (TG; 1 μM; T7459; Thermo Fisher) was added to the extracellular medium by gentle pipetting. To achieve instantaneous mixing and a final concentration of 1 μM, 500 μl TG (prepared at 2 μM in Ca2+-free Tyrode’s solution) was added to the FluoroDish^™ containing 500 μl Ca2+-free Tyrode’s solution. Numerical fluorescence intensity data was extracted from the fura-2 images using FIJI software (version 1.53q; https://fiji.sc/). ROIs were manually defined for each time lapse recording in the cytosol of individual cells and extracellular background. Next, in each image fura-2 emission signals were quantified for multiple cells at both excitation wavelengths. All cytosolic ROIs were of identical size and remained inside the cytosol during the full duration of the recording. Cytosolic fura-2 signals were first individually background-corrected for each excitation wavelength, after which the ratio value (340/380 nm) was calculated.

QUANTIFICATION AND STATISTICAL ANALYSIS

We analyzed the data using GraphPad-Prism software (Prism 10, GraphPad Software, Inc.) and employed the R environment for statistical computing. For all datasets, we tested the normality of the distribution using GraphPad-Prism (D'Agostino-Pearson test for large sample sizes and Shapiro-Wilk test for sample sizes under 50). After having performed outlier test analyses, we assessed statistical significance using parametric tests for normally-distributed data (Student's t-test, ANOVA) and non-parametric tests when normal distribution could not be verified (Mann-Whitney U and Kruskal-Wallis test). When possible, data are presented as scatter plots with individual data points showing all individual measurements. Unless otherwise indicated, we expressed the data as mean and standard deviation (mean ± SD). Experiments were repeated in independent biological experiments. Information on statistical details and number of replicates can be found in the respective figure legends.