Aberrant mitochondrial function in patient-derived neural cells from CDKL5 deficiency disorder and Rett syndrome

Abstract

The X-linked neurodevelopmental diseases CDKL5 deficiency disorder (CDD) and Rett syndrome (RTT) are associated with intellectual disability, infantile spasms and seizures. Although mitochondrial dysfunction has been suggested in RTT, less is understood about mitochondrial function in CDD. A comparison of bioenergetics and mitochondrial function between isogenic wild-type and mutant neural progenitor cell (NPC) lines revealed increased oxygen consumption in CDD mutant lines, which is associated with altered mitochondrial function and structure. Transcriptomic analysis revealed differential expression of genes related to mitochondrial and REDOX function in NPCs expressing the mutant CDKL5. Furthermore, a similar increase in oxygen consumption specific to RTT patient–derived isogenic mutant NPCs was observed, though the pattern of mitochondrial functional alterations was distinct from CDKL5 mutant–expressing NPCs. We propose that aberrant neural bioenergetics is a common feature between CDD and RTT disorders. The observed changes in oxidative stress and mitochondrial function may facilitate the development of therapeutic agents for CDD and related disorders.

Introduction

Mutations in the X-linked genes methyl-CpG–binding protein 2 (MeCP2) and cyclin-dependent kinase-like 5 (CDKL5/STK9) are associated with Rett syndrome (RTT) and CDKL5 deficiency disorder (CDD), respectively, predominantly affecting females and resulting in intellectual disability, infantile spasms and seizures. The CDKL5 gene product is a member of a shared neurodevelopmental pathway including MeCP2 and FOXG1 (1, 2), and mutations in genes coding for any of these products/proteins cause very similar and overlapping clinical manifestations (3). Biochemical and functional relationships exist between MeCP2 and CDKL5 proteins: temporal and expression patterns overlap in embryonic and postnatal murine brains, CDKL5 binds to and mediates MeCP2 phosphorylation (1) and CDKL5 has been demonstrated to be a MeCP2-repressed target gene (4). While mutations in MeCP2 have been well-documented in the etiology of RTT, the impact of mutations in CDKL5 is still less understood due to a lack of understanding about the cellular processes controlled by CDKL5 and the substrates it may target (5). Thus, despite overlap in clinical phenomenology, the potential functional overlaps in cellular processes between CDKL5 and MeCP2 have yet to be determined.

Even prior to the recognition that the genetic causes of RTT are mutations in MeCP2, clinical findings including elevated lactate and pyruvate levels in serum and cerebral spinal fluid, as well as increased glucose metabolism and upregulation of the glucose transporter SLC2A4, suggested the presence of metabolic dysfunction (6). Further evidence for metabolic abnormalities came from the observation of markers of increased oxidative stress in RTT patients, including decreased superoxide dismutase, glutathione peroxidase and thioredoxin reductase activities (7), and increased plasma malondialdehyde, an indicator of lipid oxidation (8). More recently, RTT has been linked to an increase in the oxidative stress marker serum 4-hydroxynonenal plasma protein adducts (4HNE-PAs) (9). Neurons may be particularly susceptible to damage by reactive oxygen species: lipid peroxidation from patients with RTT was demonstrated as early as 1988 (10).

Additional evidence implicates mitochondrial dysfunction in RTT as a contributor to observed oxidative stress. Specifically, abnormal mitochondrial ultrastructure by electron microscopy was independently observed in muscle and frontal lobe biopsies from RTT patients (11). Additionally, data from Mecp2-knockout mice, as well as patient postmortem brain biopsy, indicate that expression of mitochondria-related genes required for mitochondrial structure, function and oxidative stress homeostasis are abnormal in MeCP2-deficient systems (6).

Several lines of evidence also implicate aberrant mitochondrial function in CDD. Notably, the oxidative stress marker 4HNE-PAs (9) are also present in CDD patient serum, and one study showed that CDKL5-deficient fibroblasts exhibited reduced levels of NFE2L2, an apparent protective mechanism against oxidative stress (12). Additionally, studies in a Cdkl5-null mouse model of CDD identified brain mitochondrial functional abnormalities (13). We hypothesized that common mitochondrial dysfunctional pathways may be shared between these related, though genetically distinct, neurodevelopmental disorders. In this study, we demonstrate alterations in normal cellular bioenergetics in CDD patient- and RTT patient-derived neural models, raising the possibility that interventions that can slow or reverse these metabolic aberrancies may represent novel treatments for these and related disorders.

To investigate mitochondrial function in relation to CDD at a cellular level, we developed isogenic induced pluripotent stem cell (iPSC) clonal lines from a CDD female patient heterozygous for the CDKL5 mutation due to X-inactivation, such that only one copy of CDKL5 (truncated null mutant or wild-type) is expressed in each clonal line and further derived neural progenitor cells (NPCs). We tested the hypothesis that mutations in CDKL5 may be associated with aberrant mitochondrial function using bioenergetics and mitochondrial functional assays.

Results

Generation of CDD and RTT isogenic iPSCs and NPCs by stable X-inactivation clonal selection

We generated iPSCs from fibroblasts obtained from a skin biopsy of a pediatric female CDD patient using a non-viral, non-integrating synthetic mRNA-based reprogramming method (14). We generated stable isogenic clones from each patient via X-chromosome inactivation (Fig. 1A) as previously described (15) and confirmed expression of either wild-type or mutant CDKL5 by cDNA sequencing (Fig. 1B). All iPSC clones had typical characteristics of human pluripotent stem cells indicating successful reprogramming, exhibiting normal karyotypes (Supplementary Material, Fig. S1A) and expressing markers for pluripotency (Supplementary Material, Fig. S1B). Purified and expandable NPCs were derived from the iPSC clones using a neural induction supplement based on defined small-molecule induction (16, 17), followed by magnetic-activated cell sorting (PSA-NCAM⁺, CD271⁻, CD133⁺) and verified by immunocytochemical analyses for the standard NPC markers SOX1, SOX2, PAX6 and NESTIN (Supplementary Material, Fig. S1C).

Figure 1.

Generation and confirmation of isogenic CDD iPSC clones by stable X-inactivation. (A) Location schematics of CDKL5 and MeCP2 frameshift (truncation) mutations (NLS-nuclear localization domains, MBD-methyl DNA binding domain, TRD-transcription regulatory domain) from patients used in this study. Overview of isogenic clonal isolation methodology of iPSCs. (B) cDNA sequencing results confirms the presence of iPSC isogenic clonal pairs derived from CDD patients.

CDD patient–derived NPCs show greater mitochondrial respiration in comparison to isogenic wild-type

Aberrant mitochondrial function in neurodegenerative diseases can lead to changes in energy production, generation of reactive oxygen species (ROS) and induction of stress-induced apoptosis, contributing to the complex pathology of these diseases (18). To uncover potential differences in bioenergetics in CDD, we analyzed mitochondrial bioenergetics between paired isogenic patient-derived NPCs. Using the Agilent Seahorse XFe96 (SXf) extracellular flux analyzer, we measured oxygen consumption rate (OCR) and extracellular acidification rate (ECAR). Measuring both OCR and ECAR simultaneously enables a more comprehensive assessment of cellular energetics and the ability to determine the relative contribution of these two dominant energy-yielding pathways. We found that CDD NPC clones expressing the truncated mutant CDKL5 had a significantly higher basal OCR and mitochondrial respiratory capacity (Fig. 2A and B; Supplementary Material, S2A) compared to wild-type CDKL5-expressing cells. The basal respiration and oligomycin-linked ATP production in the mutant clones exhibited an increase in OCR as compared with wild-type. Unlike mitochondrial respiration, ECAR plot profiles did not appear to be associated with an appreciable amount of variation in glycolytic activity between CDKL5-mutant and wild-type pairs, suggesting that isogenic pairs have similar rates of glycolysis (Supplementary Material, Fig. S2C). Further analyses revealed increases in basal respiration, oligomycin-sensitive ATP production and spare respiratory capacity (Fig. 2B). These differences were not evident when measured in clonal isogenic immortalized patient fibroblast clones expressing mutant versus wild-type CDKL5 (Fig. 2C), supporting that the bioenergetic differences observed are more pronounced in neural cells and not a generalized phenomenon across cell types.

Figure 2.

An increase in oxygen consumption rate (OCR) in CDD NPCs. Seahorse Cell MitoStress test was performed to measure OCR in CDKL5-mutant and wild-type-expressing isogenic NPCs clones (n = 4; 2 wild-type, 2 mutant) following a sequential addition of inhibitors of mitochondrial function: oligomycin, carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and a combination of antimycin A and rotenone; (A) OCR Seahorse assay profile plot, (B) basal respiration, maximal respiration, ATP production and spare respiratory capacity. Basal respiration was calculated after subtraction of non-mitochondrial respiration. ATP-linked respiration was calculated following the addition of oligomycin. Maximal respiration was measured following the addition of FCCP. Spare respiratory capacity was calculated based on the difference between the basal respiration and maximal respiration. (C) Metabolic potential in immortalized CDD isogenic mutant and wild-type fibroblasts. Asterisks represent significance comparing OCR of mutant to wild-type; n = three well replicates per condition in 96-well assay plates, P < 0.05.

Mitochondrial characterization in CDD NPCs implicates mitochondrial functional anomalies

In light of previous clinical studies suggesting an increase in oxidative stress in CDD and RTT patients (7–9, 12), we first analyzed the production of ROS in our CDD isogenic NPCs. Mitochondrial ROS are generated as a byproduct of oxidative phosphorylation. The incomplete reduction of approximately 1% to 3% of consumed mitochondrial oxygen produces superoxide, the predominant ROS in mitochondria, which is quickly converted by mitochondrial superoxide dismutase to H₂O₂ (19). Elevated levels of superoxide results in oxidative stress that can lead to abnormal damage to nucleic acids, proteins and lipids, resulting in neuronal functional deficiencies as well as cellular death by apoptosis, which has been linked to a number of neurodegenerative and neurodevelopmental disorders (18). To detect intracellular superoxide levels in mitochondria, we treated live NPC cultures with MitoSOX™ Red mitochondrial superoxide indicator, a fluorescent lipophilic cation that partitions into mitochondria and has fluorescence intensity that increases with superoxide concentration (20). We did not observe a significant difference in MitoSOX™ fluorescence in CDKL5-mutant compared to wild-type NPCs (Fig. 3A and B), demonstrating that mitochondrial ROS levels were unaffected by functional mutations in CDKL5 and similarly extracellular ROS levels were not affected in CDKL5-mutant NPCs (Fig. 3C). Additionally, by challenging the NPCs in galactose-containing media, shifting towards more reliance on mitochondrial oxidative phosphorylation (OXPHOS) (21, 22) and decreasing the activity of glycolytic ATP production, we observed no ROS differences associated with the CDKL5 truncation mutation.

Figure 3.

Mitochondrial membrane depolarization is significant in CDKL5-mutant NPCs. (A) CDKL5-mutant and isogenic wild-type NPCs were grown in 2 mM glucose and 8 mM galactose and incubated with MitoSOX™ Red for 1 h. Mean ROS levels were determined by red fluorescence (SSC: side scatter; MitoSOX™: red channel) in flow cytometry. (B) ROS values from three independent experiments (n = 30 000 cells) were normalized to a baseline value of 100 and averaged across indicated media conditions. (C) Extracellular ROS values were normalized to a baseline value of 100 for wild-type and averaged across indicated conditions. (D) CDKL5-mutant and isogenic wild-type NPCs were grown in 2 mM glucose and 8 mM galactose for 24 h and stained with JC-10. The mean JC-10 fluorescence green/red intensity ratio was detected using flow cytometry. (E) Results of three independent membrane potential experiments (n = 25 000 cells) from JC-10 green/red ratios in flow cytometry in indicated culture media conditions normalized to wild-type = 100. Asterisks represent significance compared to wild-type; n = 3, P < 0.05. (F) Representative images of JC-10 staining in indicated culture media conditions.

Decreased mitochondrial membrane potential detected in CDD patient–derived NPCs

Since increased OCR is one indicator of potential aberrant mitochondrial function that could be caused by changes in mitochondrial membrane potential (ΔΨm) (23), we analyzed ΔΨm in mutant and wild-type isogenic CDD NPCs. NPC clones were treated with JC-10 dye that forms red-fluorescent J-aggregates within polarized mitochondria and green-fluorescent cytosolic monomers in cells upon lowered mitochondrial integrity; treatment with FCCP was used as a positive control for functionality of the JC-10 dye. The plots indicate CDKL5 mutant–expressing NPCs shifted towards green fluorescence when compared to wild-type (Fig. 3D–F), suggesting modest mitochondrial depolarization (increased green/red fluorescence ratio) in these mutant NPCs. Furthermore, we observed an augmented difference in green/red fluorescence ratio between isogenic NPC clones in galactose-containing media, suggesting greater depolarization in mutant cells when shifted towards OXPHOS cellular metabolism (Fig. 3E).

Altered mitochondrial structural features in mutant CDKL5 NPCs

We considered whether a structural defect might underlie and account for these functional perturbations. To determine mitochondrial structural organization, cells grown in varying glucose media, 10 mM or 2 mM glucose plus 8 mM galactose, were subjected to transmission electron microscopy (TEM) (Fig. 4; Supplementary Material, Fig. S3). Most mitochondria in CDKL5 wild-type NPCs presented as bean-shaped structures with transversely orientated cristae enveloped by an intact outer membrane (Fig. 4A–F). In contrast, we observed a high frequency of mitochondrial morphological alterations in CDKL5-mutant NPCs (Fig. 4G–L). CDKL5-mutant mitochondria displayed small spherical structures with fewer and disarrayed cristae, and decreased electron density of the matrix (Fig. 4J–L, arrows). In addition, these mitochondria also displayed concentric rings and bleb-like structures (Fig. 4G–I, arrows) within the outer mitochondrial membrane, as well as membrane ruptures and vacuoles (Fig. 4G, arrow).

Figure 4.

Altered mitochondrial morphology in CDKL5 NPC isogenic lines.

iPSC-derived NPCs were subjected to transmission electron microscopy analysis and mitochondrial structure was compared. (A–F) Representative electron micrograph of mitochondria from CDKL5 wild-type NPCs, showing bean-shaped structures with numerous transversely orientated cristae enveloped by an intact outer membrane. (G–L) Representative electron micrographs show alterations in mitochondrial morphology of CDKL5-mutant NPCs: small spherical mitochondria with fewer and disarrayed cristae and a decreased electron density of the matrix (arrow), and mitochondria with membrane rupture or large vacuoles (arrow). Scale bar = 100 nm.

These ultrastructural defects have been correlated with mitochondrial fission, metabolic disorders and cell death (24). This finding supports our bioenergetics profiling describing a possible alteration of mitochondrial function in CDKL5-mutant NPC lines.

Transcriptome analyses identify dysregulation of mitochondrial and REDOX-related genes in CDD patient–derived NPCs

To investigate genome-wide transcriptome changes in CDKL5 mutant cells, we performed preliminary transcriptomic analyses between a pair of CDD patient–derived mutant and wild-type isogenic NPCs. Expression analysis revealed 18 differentially-expressed genes (FDR < 0.05), suggesting a role in transcription regulation in addition to the known kinase role of CDKL5 (Fig. 5A and B; Supplementary Material, Table S1). Interestingly, these upregulated genes include IRAK1 and POU3F4, which have been associated with intellectual disability (25) and developmental delay (26). The most significantly downregulated gene is FZD10, which functions as a receptor for Wnt proteins in the canonical Wnt/beta-catenin signaling pathway. Growing evidence implicates Wnt signaling in a variety of neuropsychiatric disorders (27, 28). Additionally, in agreement with the gene expression data, we have further validated increased protein expression in mutant CDKL5-expressing clones (Supplementary Material, Fig. S4) for the X-linked neurodevelopmental associated transcription factor BRN4 (encoded by the POU3F4 gene), implicated in transcriptional regulation of MeCP2 (29).

Figure 5.

Downregulation of mitochondrial and REDOX related genes in CDD patient–derived NPCs. (A) Volcano plot of differential expression. Labels are show for significantly differential genes (FDR < 0.05). (B) Heatmap of 18 differentially expressed genes (FDR < 0.05). (C) GSEA plot for 720 Human MitoCarta Genes, black vertical lines indicating overlap genes with our dataset and MitoCarta, red-upregulated, blue-downregulated genes in common. Overlap is strongly biased toward the left (blue) downregulated genes common in these gene sets. (D) Gene-set module of significantly downregulated (P < 0.01 and FDR < 0.1) gene sets related to mitochondria and REDOX. Node size is proportional to the number of genes in each set. Edge thickness is proportional to the fraction of shared genes between the sets. Other unrelated modules that emerged are not shown. Image was generated using the EnrichmentMap plugin in cytoscape (56).

To understand transcriptomic changes in a more specific mitochondrial context, we performed a gene set enrichment analysis (GSEA) on 720 genes from the Human MitoCarta 2.0 inventory (30) (see Methods section). Figure 5C demonstrates a significant enrichment of downregulated genes within this set. An unbiased GSEA of our differential expression results on all curated gene sets in MSigDB (31) likewise highlights downregulation of mitochondrial and REDOX-related pathways (Fig. 5D; Supplementary Material, Table S2, providing GSEA results).

RTT patient–derived NPCs demonstrate alterations in bioenergetics not attributed to mitochondrial membrane potential

Besides evidence that CDD and RTT share similar clinical metabolic features suggesting mitochondrial and/or oxidative stress abnormalities (7–9, 12), biochemical and functional relationships have been described between MeCP2 and CDKL5 proteins (1, 4). Studies have also shown that absence of MeCP2 affects mitochondrial function and gene expression (6, 32). Thus, we hypothesized that these proteins not only share similar molecular pathways but also similar bioenergetic profiles. In order to compare bioenergetics between CDD and RTT disorders, we evaluated mitochondrial bioenergetics in MeCP2 truncation mutant and wild-type isogenic NPC clones derived from a female RTT patient (Fig. 6A) (15). Profiling bioenergetics with RTT NPC clones, we measured OCR and ECAR, demonstrating mutant MeCP2-specific increases in basal respiration, maximum respiration and oligomycin-dependent ATP production, with no changes in spare respiratory capacity (Supplementary Material, Fig. S2B), similar to that observed between CDD NPC clones. Further, to identify if ΔΨm was different in these RTT clones as we observed between CDD clones, membrane polarization was assessed by JC-10 green/red fluorescence ratio. There were no differences between RTT NPC clones, indicating no alterations in ΔΨm (Fig. 6E and F) upon expression of mutant MeCP2. These findings indicate a distinct difference of mitochondrial dysfunction between CDD patient and RTT patient–derived NPCs.

Figure 6.

Mitochondrial bioenergetics in RTT-derived isogenic NPCs. Higher ROS generation is detected in MeCP2 mutant NPCs compared to wild-type: (A) Seahorse Cell MitoStress test was performed to measure OCR in RTT mutant and wild-type isogenic NPCs. (B) Isogenic clones were grown in and incubated with MitoSOX™ Red for 1 h and ROS levels were determined as red fluorescence by flow cytometry. (C) ROS values from three independent experiments (n = 15 000 cells) were normalized to a baseline value of 100 and averaged across indicated media conditions. Asterisks represent significance compared to wild-type; n = three well replicates per condition in 96-well assay plates, P < 0.05. (D) Extracellular ROS values were normalized to a baseline value of 100 to wild-type and averaged across indicated conditions. (E) Cells were grown in 2 mM glucose and 8 mM galactose and stained with JC-10. The mean JC-10 green/red fluorescence intensity ratio was detected using flow cytometric analysis. (F) Results of three independent membrane potential experiments (n = 25 000 cells) in indicated culture media conditions normalized to wild-type = 100.

Increased ROS in RTT patient–derived NPCs harboring MeCP2 mutation

The MeCP2 protein is responsible for binding to methylated and unmethylated DNA (33), and increased ROS plays an important role disrupting MeCP2 regulation of epigenetic gene expression (34). Additionally, there is evidence of increased oxygen consumption and ROS production in the brain mitochondria of Mecp2-null mice (35). To determine if ROS production is altered in RTT NPCs, we compared superoxide levels in our isogenic RTT NPC pairs. Mutant MeCP2-expressing NPCs showed a significant increase in MitoSOX™ Red intensity consistent with increased intracellular (Fig. 6B and C) and extracellular (Fig. 6D) ROS generation in these cells as a marker of increased oxidative stress. Additionally, we evaluated ROS levels in other media conditions more appropriate to brain physiological glucose concentrations (10 mM) (data not shown) as well as conditions shifted towards more dependence on mitochondrial OXPHOS by the addition of galactose (21, 22) (2 mM glucose and 8 mM galactose) to augment differences. The galactose media further augmented the ROS difference between RTT clones (Fig. 6C). Conversely, CDKL5 mutant–expressing NPCs did not demonstrate any significant increase in ROS generation (Supplementary Material, Fig. 3A–C), suggesting that the observed ROS phenotype is restricted to MeCP2 mutations in RTT patient–derived lines.

Discussion

An important pathological consequence of MeCP2 mutations in RTT is likely to be mitochondrial dysfunction (6, 10, 11, 36–39). Mitochondrial abnormalities were identified in RTT patients (36, 39) including enzymatic defects in mitochondrial respiratory chain (10, 37) and increased oxidative stress (8, 39). However, similar dysfunction in CDD has not been investigated in depth, despite the recognized interactions between MeCP2 and CDKL5 and the similarity of the clinical presentations of these syndromes (1, 4, 40).

Here we show that iPSC-derived neural cellular models generated from a pediatric patient with CDD exhibit abnormalities within mitochondrial respiration as determined by increased OCR in CDKL5-mutant expressing NPC clones but not in clonal CDKL5-mutant fibroblasts. Our results in NPCs suggest pathophysiology during early development are supported by previous studies that have concluded that NPCs exhibit oxidative metabolism in vitro similar to normal neuronal development and preserve patient-specific mitochondrial abnormalities (22). This makes them important tools for phenotypic screening of compounds for developing novel therapies towards neurological disorders such as CDD and RTT (22). Furthermore, we demonstrate similar increased OCR in RTT patient–derived MeCP2 mutant–expressing NPC clones. A modest change in spare respiratory capacity was observed between mutant and wild-type NPCs from either CDD or RTT patients, consistent with impaired mitochondrial respiration. Previous studies in human postmortem tissues (11), Mecp2- KO mice brain (41) and peripheral blood lymphomonocytes (42) reveal similar aberrant mitochondrial function. Although the increase in OCR was similar between CDD and RTT mutant and wild-type NPCs, further evaluation indicated that mitochondrial functional differences vary between these diseases; modest mitochondrial depolarization was observed only in CDKL5 mutant–expressing NPCs while increased ROS was specific to MeCP2 mutant–expressing NPCs. An increase in expression of IRAK1 in Mecp2-null mice has been reported (43) similar to our observation in CDD. Similarly, an increase in expression of the early neuronal development gene POU3F4, similar to that observed in the X-linked fragile X syndrome (44), was associated in CDKL5-mutant NPCs along with a downregulation of FZD10. Mitochondrial respiration assay results were further supported by ultrastructural analyses that revealed morphologic changes in mitochondria of CDD patient–derived mutant NPCs. These cells appear to have distended, sparsely packed mitochondria with abnormal or absent cristae. These observations are consistent with shifts in mitochondrial dynamics and also a possible disruption of normal mitophagy processes.

The brain is a highly metabolically active organ with a high oxygen requirement but low antioxidative activity, making neurons particularly susceptible to damage by mitochondrial dysfunction-derived oxidative stress. Thus, a potential pivotal role for mitochondrial dysfunction in some neurodegenerative and neurodevelopmental diseases is gaining increasing attention (44). Mitochondrial dysfunction resulting from genetic variation in both the nuclear and mitochondrial genomes has been studied in a range of psychiatric disorders, among them RTT syndrome, bipolar disorder, major depression and schizophrenia (45, 46), but the aggregate effects of genetic variation remain difficult to model directly.

A number of mechanisms may be relevant to the impaired bioenergetics described. Our bioenergetics findings correlate with the membrane potential assay, implying loss of mitochondrial integrity essential for preserving normal OXPHOS function. Damage to the mitochondrial membrane is a proven central player in the initiation of intrinsic apoptotic events, with cytochrome c release from the inner mitochondrial membrane being a factor that can be modulated by ROS (47).

In summary, the results of this study provide evidence of parallel-impaired neural bioenergetics between both CDD and RTT disorders, but with some divergent aspects. Features such as greater OCR and corresponding changes in mitochondrial morphology are tied to cellular health and are observed in both disorders. Neural progenitors derived from a CDD patient expressing mutant CDKL5 exhibit decreased membrane potential via modest lower polarization of mitochondria, while our study indicates that the observed increased in ROS in RTT patient–derived NPCs is not observed in our CDD NPC counterparts. Further, transcriptomic comparison of downregulated genes in mutant CDKL5-expressing NPCs to the mitochondrial database Mitocarta2.0 (30) identified enrichment of mitochondrial and OXPHOS genes. Future studies using models designed to be closer to in vivo neuronal conditions, such as 3D organoids, and studies with well-characterized neuronal-glial connections may help to further elucidate relevant pathophysiology. Earlier studies have verified that free-radical scavengers and antioxidants can improve oxidative markers in RTT (48, 49). For example, curcumin, an anti-inflammatory compound, was found to decrease intravascular ROS production in RTT mice (50), and studies have shown oxidative damage and oxidative stress could in part be ameliorated by omega-3 polyunsaturated fatty acids (40, 51). A clinical phase 2 trial in RTT involving the vitamin E derivative (EPI-743, vatiquinone) demonstrates antioxidant capacity and improves mitochondrial electron transport within the respiratory chain, which has also been reported (https://clinicaltrials.gov/ct2/show/NCT01822249).

Taken together, our results using new cellular models of these diseases suggest an impact of a loss-of-function mutation on mitochondrial function and bioenergetics in CDD. Such models facilitate the identification of interventions that may ameliorate these phenotypes and yield high-priority candidate therapeutics for these rare, devastating neurodevelopmental disorders.

Materials and Methods

Fibroblasts reprogramming and iPSCs culture

Fibroblasts from a 2-year-old female CDD patient harboring a truncating mutation 1412delT (p.Asp471Ala) were collected following parental informed consent via the Manton Center (Children’s Hospital, Boston, Massachusetts). Skin punch biopsy (3 mm) was performed from the non-dominant upper arm by a physician investigator. Primary human fibroblasts were reprogrammed, stabilized and expanded under xeno-free conditions by Cellular Reprogramming, Inc. (www.cellular-reprogramming.com) as previously described (48, 49). Primary fibroblasts were reprogrammed using mRNA reprogramming in a feeder-free culture system. Derived iPSCs were expanded in Nutristem XF media (Corning) on rLaminin-521 (BioLamina) coated plates to at least passage 3 before cryopreservation. Previously derived iPSCs from fibroblasts obtained from a 25-year-old female RTT patient (Coriell Institute for Medical Research #GM07982) who harbored a single-nucleotide deletion (frameshift 705delG) were reprogrammed and characterized as described (15).

Generation of iPSC clones

Previously, RTT isogenic iPSC clone derivation (07982–22 and 07982–23) by stable X-inactivation was performed as described (15). Similarly, isogenic CDD iPSC clones were generated by selecting clones that express either wild-type or mutant CDKL5. CDD iPSCs were harvested with Accutase (Sigma) and single cells were plated at low density in the presence of ROCK-inhibitor thiozovivin (StemCell Technologies). Emerging iPSC colonies were manually picked and expanded on mouse embryonic fibroblast (MEF) feeders for several passages. Established clones that maintained a good hESCs-like morphology were moved to feeder-free culture conditions on Geltrex (ThermoFisher) coated dishes in E8 medium (Stem Cell Technologies). From this point, cells were routinely passaged by Collagenase IV treatment (Invitrogen) in the presence of thiozovivin. iPSC quality was assessed by typical colony morphology as well as positive staining of the pluripotency markers Oct4, Nanog, Tra-1-60 and SSEA4 (Supplementary Material, Fig. S1B). Clones were sequenced as per methods described (15) (genotype forward: 5′-GAAGGCCCAGGGACAAAGTAC-3′, genotype reverse: 5′-TACCTACTGGTACTGGGCTCC-3′). Isogenic pairs of iPSC clones differing for CDKL5 allelic expression were used: clone #12 expressing the wild-type CDKL5 allele and clone #5 expression of the mutated CDKL5 allele as confirmed by cDNA sequencing. RTT isogenic iPSC clones were generated and confirmed as described (15).

Clone maintenance and neuronal differentiation and characterization

iPSCs were maintained and NPCs were differentiated as described (16, 17). Briefly, purified and expandable NPCs were derived from the iPSC clones using a neural induction supplement (ThermoFisher; PSC Neural Induction Media) based on defined small molecule induction (16, 52), followed by magnetic-activated cell sorting (PSA-NCAM⁺, CD271⁻, CD133⁺; Miltenyi Biotec) according to the manufacturer’s protocol and verified by immunocytochemical analyses for standard NPC markers, PAX6, SOX1, SOX2, NESTIN (Supplementary Material, Fig. S1C) and analyzed by image analysis using IN Cell Analyzer 6000 (GE Healthcare Life Sciences) to assess purity of the population.

Generation and confirmation of isogenic immortalized CDD patient fibroblasts

Cells were transiently transfected with 4.5 μg of pBABE-neo-hTERT and 4.5 μg of pAmpho vectors using Lipofectamine 3000 (Invitrogen). The viral supernatant was collected 24 h after transfection, filtered, supplemented with polybrene (final concentration of 4 μg/mL) and used to immortalize fibroblasts. Forty-eight hours after retroviral infection, cells were split with media containing 0.5 mg/ml Geneticin (Invitrogen). Following drug selection, immortalized fibroblasts were serially diluted in 96-well plates to obtain clonal cultures. After expansion, RNA was isolated from potential clones by TRIzol extraction (Invitrogen) and converted into cDNA using SuperScript III Reverse Transcriptase (Invitrogen). Allele-specific CDKL5 primers were used to verify by qPCR whether clones expressed mutant or wild-type CDKL5. Allele-specific CDKL5 qPCR primers were designed to exclude (wild-type primer) or include (mutant primer) the CDKL5 deletion. Allelic specificity of transcript amplification following qPCR was validated using cDNA from clonal wild-type or mutant CDKL5 NPCs [CDKL5 forward primer: 5′-CCACACCTTCTTAGCCCAAA-3′, CDKL5 reverse primer (wild-type): 5′-CTAGAGGACTGGGGAATTGTATC-3′, CDKL5 reverse primer (mutant): 5′-CCTAGAGGACTGGGGAATTGTAC-3′].

Whole transcriptome RNA sequencing

Total RNA was extracted from cell pellets using the Qiagen miRNeasy mini kit and eluted in 30ul RNase-free water. RNA QC was performed on a 1:100 dilution using Agilent Bioanalyzer with the Pico RNA chip. RNA samples with RIN scores greater than 7 were diluted to 40 ng/μl using concentrations measured with the Qubit HS RNA assay and prepared for sequencing using KAPA Stranded RNA-Seq Kits with RiboErase and sequenced on Illumina Hiseq4000 to achieve ~ 80 M paired end 100 bp reads for each sample.

Differential expression analysis

RNA-Seq samples were processed using a standard pipeline. Sequence alignment was performed with STAR (53) and transcript quantification with RSEM (54), both using default parameters. Differential expression was evaluated with DESeq2 (55) using default parameters. MitoCarta genes were extracted from the Human MitoCarta inventory (30) using only genes with FDR < 0.01. GSEA was performed using the preranked option where genes were ranked based on their Log2 fold change computed in DeESeq2. The MSigDB sets for the GSEA analysis included the “H” and “C2” groups.

Mitochondrial respiration assay

Mitochondrial respiration and aerobic glycolysis were monitored in real-time with the Seahorse Bioscience Extracellular Flux Analyzer (XFe96; Seahorse Bioscience) by measuring the OCR (indicative of respiration) and ECAR (indicative of glycolysis). Cells were seeded at 500 000 per well of a Seahorse 96-well microplate. All assays were done using Seahorse XFe Cell Energy Phenotype Test Kit (Agilent) and Cell Mito Stress Test Kit (Agilent).

Transmission electron microscopy

NPCs were grown in 6-well plates and fixed with 2% glutaraldehyde in 0.1 M cacodylate buffer and rinsed with 0.1 M cacodylate buffer. A 1 ml solution containing l mM ADP, 1 mM AMP-PCP, 2 mM MgCl2, 5 mM EDTA, 5% ethylene glycol and 1% CHAPS, pH 7.5 was centrifuged at 14 000 rpm in a desktop centrifuge for 5 min. The resultant supernatant was then adsorbed onto glow-discharged carbon/parlodion-coated copper grids and stained with 1% uranyl acetate containing 0.02% Tylose (methylcellulose). After removing the excess negative stain by blotting and then air drying, samples on the grid were visualized under a Philips CM120 electron microscope in a low-dose mode at an accelerating voltage of 80 kV, a magnification of 80 000, and underfocus settings ranging from 0.5 to 2 m.

Immunocytochemistry and immunoblotting

NPCs were seeded on a 24-well plate at 120 000 cells/well on glass cover slips and fixed with 4% PFA after 24 h. For verification of NPC identity, cells were stained for PAX6 (Developmental Studies Hybridoma Bank; #PAX6-s), SOX1 (Cell Signaling Technology; #4194S), SOX2 (Cell Signaling Technology; #3579S), Nestin (EMD Millipore; #ABD69) and counterstained with Hoechst 33342 (Invitrogen; #H3570). For representative images indicating membrane potential, live cells were incubated with JC-10 fluorogenic dye (Abcam; #ab112134), fixed and stained with Hoechst. For representative images of BRN4 expression, fixed cells were stained for BRN4 (Abcam; ab251968) and counterstained with Alexa Fluor 594 Phalloidin (Invitrogen; A12381) and Hoechst. Stained cells were imaged on IN Cell Analyzer 6000 (GE Healthcare Life Sciences).

For immunoblotting, NPC pellets were resuspended in RIPA buffer (Boston BioProducts) with protease and phosphatase inhibitors (Roche) and centrifuged at 15 000g for 15 min at 4°C. Total protein amount was determined by BCA protein assay, and 20 μg of protein was loaded onto a 10% Mini-PROTEAN TGX Precast Protein Gel (Bio-Rad) and run at 120 V for 1 h. The gel was transferred onto an Immobilon-FL PVDF membrane (EMD Millipore) at 120 V for 1 h, and the membrane was then blocked using Odyssey Blocking Buffer in PBS (LI-COR Biosciences) overnight at 4°C. The membrane was incubated with BRN4 antibody (Abcam; ab251968) for 1 h, washed three times with 0.1% PBS-Tween for 5 min each, incubated with IRDye 800CW anti–rabbit antibody (LI-COR Biosciences) for 1 h, then washed three more times (all steps performed at room temperature). β-Actin (Abcam; ab8226) was used as a loading control, and imaging was performed on the Odyssey CLx system (LI-COR).

Mitochondrial ROS determination

To measure total cellular ROS, cells were incubated with 10 μM FCCP and 1 μM antimycin A for 1 h at 37°C before being returned to prewarmed media without phenol red. Mitochondrial ROS (superoxide) was assessed using the MitoSOX™ reagent (Life Technologies) and following manufacturer’s instructions. Cells were fixed with 4% PFA and were analyzed by flow cytometry (BD FACSAria™ III cell sorter) using the PE (red) channel. Mean MitoSOX™ intensity values were calculated by averaging single cell fluorescence values in the PE red channel, and normalized by dividing by the mean intensity from wild-type. Differences between mutant and wild-type cells were calculated across three experiments, and t tests were performed with significance indicated for P < 0.05.

Extracellular ROS was recorded in culture media using ROS-Glo™ H2O2 Assay system. Cells were plated at 50 000 cells/well, allowed to attach overnight. The media was collected 24 h later and assay was performed as per manufacturers protocol. The luminescence was measured for all the plates using a microplate reader (SpectraMAX M5e; Molecular devices). The results were expressed as relative luminescence units (RLUs) and normalized to total cell number.

Mitochondrial membrane potential assay

Mitochondrial membrane potential was assessed by flow cytometry (BD FACSAria™ III cell sorter) using a fluorogenic dye, JC-10 (Abcam). Treated cells were loaded with JC-10 dye according to the manufacturer’s instructions with modifications: spent medium was aspirated and complete medium added to adherent cells. JC-10 solution was added at equal volume and incubated in the dark at 37°C for 15 min prior to analysis. For the positive control, cells were incubated with FCCP [carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone], preceding the JC-10 solution. Monomeric (green) and J-aggregate (red) fluorescence were measured using the GFP (ch.1) and PE far-red (ch.3) channels, respectively, and analyzed following compensation for spectral overlap. On a cell-by-cell basis, JC-10 monomer value, measured in the GFP (green) channel, was divided by JC-10 aggregate value, measured in the PE (red) channel, and normalized to the wild-type ratio. This normalized green/red ratio represents membrane potential status, with larger values indicating increased depolarization of the mitochondrial membrane potential compared to wild-type.

Statistical analysis

All values are expressed as the mean of at least three different experiments. The significance of results was evaluated by Student’s t test, and statistical significance was set as P ≤ 0.05. Statistical analysis was performed using GraphPad Prism 7.0.

Study approval

De-identified cell lines were accessed via a tissue bank as approved by the Partners HealthCare Institutional Review Board (Protocol: 2009P000238); the de-identified CDD patient line was accessed via the Manton Center (Children’s Hospital, Boston, Massachusetts).