Mechanisms of IRF2BPL-related disorders and identiﬁcation of a potential therapeutic strategy

SUMMARY The recently discovered neurological disorder NEDAMSS is caused by heterozygous truncations in the transcriptional regulator IRF2BPL . Here, we reprogram patient skin ﬁbroblasts to astrocytes and neurons to study mechanisms of this newly described disease. While full-length IRF2BPL primarily localizes to the nucleus, truncated patient variants sequester the wild-type protein to the cytoplasm and cause aggregation . Moreover, patient astrocytes fail to support neuronal survival in coculture and exhibit aberrant mitochondria and respiratory dysfunction. Treatment with the small molecule copper ATSM (CuATSM) rescues neuronal survival and restores mitochondrial function. Importantly, the in vitro ﬁndings are recapitulated in vivo , where co-expression of full-length and truncated IRF2BPL in Drosophila results in cytoplasmic accumulation of full-length IRF2BPL. Moreover, ﬂies harboring heterozygous truncations of the IRF2BPL ortholog (Pits) display progressive motor defects that are ameliorated by CuATSM treatment. Our ﬁndings provide insights into mechanisms involved in NEDAMSS and reveal a promising treatment for this severe disorder.


In brief
Sinha Ray et al. developed a cellular model for IRF2BPL-related neurological disorder NEDAMSS, which led to identification of cytoplasmic mislocalization of full-length IRF2BPL, reduced astrocyte-mediated neuronal support, and aberrant energy household. CuATSM treatment effectively reversed some of the disease phenotypes both in patient cells and a fly model.

INTRODUCTION
IRF2BPL (interferon regulatory factor 2 binding protein-like) is an intron-less gene mapped to 14q24.3 and encodes a ubiquitously expressed transcriptional regulator belonging to the IRF2BP family. 1,2 The protein contains two highly conserved domains, an IRF2BP zinc finger DNA-binding domain and a C3HC4 RING finger domain at the N and C terminus, respectively, both involved in transcriptional modulation. 1,3 The function of the protein is largely undefined; however, several studies have shown a role in the initiation of puberty in female rodents and non-human primates as a transcriptional activator of gonadotropin releasing hormone 1 (GNRH1). 2,[4][5][6] Additionally, one study has shown that IRF2BPL functions as an E3 ubiquitin ligase targeting b-catenin for proteasome degradation in gastric cancer cell lines. 7 Recently, heterozygous truncating mutations in IRF2BPL have been shown to cause variable neurological phenotypes, indicating that the gene might play an important role in both devel-opment and neuronal maintenance. [8][9][10][11][12] Missense variants have also been reported but are mostly associated with milder neurological symptoms such as seizures, developmental delay, and autistic spectrum disorder. 8 Most severely affected individuals have typical initial development until around 3.5 years of age, at which point developmental delay and neurological regression occur, leading to abnormal movements, loss of motor skills and speech, and seizures. 8,9,12 The disorder has since been termed NEDAMSS (neurodevelopmental disorder with regression, abnormal movements, loss of speech, and seizures, MIM: 618088). Since the discovery of this disease in 2018, 25 patients have been reported worldwide 12 with IRF2BPL mutations causing NEDAMSS, but increased genetic testing for this rare disease will likely unravel a higher frequency. 13 Little is known about IRF2BPL function in the nervous system. The first report to demonstrate a key role in the central nervous system showed that neuronal knockdown of Pits, the orthologous gene in Drosophila melanogaster, resulted in neurodegeneration. 8 Moreover, both human IRF2BPL and Pits cause lethality when overexpressed in wild-type flies. However, overexpression of IRF2BPL truncated protein isoforms, traditionally observed in NEDAMSS patients, is not toxic in flies, which potentially implies a loss-of-function disease mechanism. 8 In addition, our recent data indicate a role for IRF2BPL/Pits in downregulating Wnt signaling in the nervous system. 14 While loss-of-function models have been generated in nonmammalian species, the role of IRF2BPL in NEDAMSS has yet to be dissected in patient cells. In addition, the impact of individual patient-specific mutations or cell-type-specific contributions to NEDAMSS is currently unknown. This knowledge is critical to identify and develop a therapeutic strategy. To address these limitations, we developed a human in vitro cell model for NEDAMSS. Using a previously established protocol, we directly reprogrammed human patient fibroblasts into induced neurons (iNs) 15 or induced neuronal progenitor cells (iNPCs). 16 iNs from NEDAMSS patients showed variable disease phenotypes including reduced neurite length. iNPCs were subsequently differentiated into induced astrocytes (iAs), and the impact of IRF2BPL mutations on patient iAs was assessed. Our analyses revealed a mislocalization of IRF2BPL in the cytoplasm of patient iAs, reduced astrocyte-mediated neuronal support, and aberrant mitochondrial activity. Similar mislocalization of IRF2BPL and age-dependent defects were observed in a Drosophila model. Importantly, treatment with a small molecule called copper ATSM (CuATSM; diacetylbis(4-methylthiosemicarbazonato) copperII), which is currently in clinical trials for amyotrophic lateral sclerosis (ALS) (NCT04082832), successfully improved disease phenotypes in vitro and in vivo in the fly model. Hence, we propose an important role of astrocytes in NEDAMSS and describe a dominant-negative disease mechanism. Importantly, we also evaluated a potential treatment that may slow or halt the progression of NEDAMSS disease and might be applicable for other neurological conditions.

RESULTS
Full-length IRF2BPL is mislocalized to the cytoplasm in NEDAMSS patient astrocytes To develop a human in vitro model for NEDAMSS, we obtained four patient fibroblast primary cell lines (Table S1) with de novo heterozygous mutations (confirmed by amplicon sequencing) leading to the truncation of the C3HC4 ring domain ( Figure 1A). Since the disease was recently described to be caused by haploinsufficiency, 8 we first tested if these patient fibroblasts had reduced IRF2BPL protein levels by western blot analysis using a commercially available antibody that binds the nuclear localization signal (NLS) toward the C terminal of the protein ( Figure 1A). This antibody detects only the full-length protein and not the truncated version, with the exception for cell line P4. Surprisingly, most patient fibroblasts showed only a small or no reduction of IRF2BPL protein levels compared with healthy controls (H1, H2, H3, and H4), and only the oldest patient (adult) P3 showed the expected 50% loss of full-length IRF2BPL (Figure S1A). The cellular localization of IRF2BPL was established by immunofluorescence staining, and the protein was found to mainly localize to the nucleus of control fibroblasts with some signal in the cytoplasm of patient fibroblasts ( Figure S1B).
Given that NEDAMSS primarily affects the nervous system and presents with various neurological symptoms in patients, 8,9,12 we explored the cell morphology and expression levels of IRF2BPL in neurons. We utilized an established direct conversion method using small molecules to generate iNs from fibroblasts. 15 After 7 days of exposure to the molecule mix, cells expressed neuronal-specific markers such as Tuj1 and Map2 ( Figure S2A). Similar to fibroblasts, patient iNs mainly exhibited nuclear localization of IRF2BPL with faint signal in the cytoplasm. In contrast, in healthy control iNs, the protein was almost exclusively found in the nucleus ( Figure S2B). Most patient iNs showed none or a slight reduction in IRF2BPL protein, with the exception of adult patient P3, who showed a more significant loss ( Figures S2B and S2C). The mean neuronal conversion efficiency (percentage of Tuj1+ soma/DAPI) ranged between 55.9% and 73.7% for both healthy (H1 child and H3 adult) and patient iNs (P1, P2, and P4). However, the adult patient cell line P3 had a significantly lower rate of conversion (12.5%) ( Figure S2D). Moreover, P3 mostly showed low Tuj1 expression in cells morphologically resembling fibroblasts, indicating incomplete conversion. Additionally, both patient iNs P1 and P3 exhibited shorter neurite length compared with healthy controls and other patient iNs ( Figure S2E).
In the CNS, neurons are supported by a variety of glia. Astrocytes are the most abundant cell type in the CNS and provide a wide range of functions to support neurons. They regulate the extracellular environment of neurons, promote survival, and modulate synaptic transmission and plasticity, which are important for neuronal signaling and development 17 and frequently play an important role in disease. 18,19 To study the cellular phenotypes of NEDAMSS-derived astrocytes, patient and healthy control fibroblasts were directly converted to iNPCs and were further (B and C) (B) Schematic of direct conversion of fibroblasts to iNPCs and further differentiation to astrocytes (iAs) in 5 days in vitro. Illustration was prepared by using biorender.com. (C) Comparison of protein quantification of healthy (H1 child and H3 adult) and NEDAMSS patient (P1, P2, P3, and P4) iAs indicates that adult patient P3 has significant loss of IRF2BPL expression. (D-F) (D) Immunostaining demonstrates mislocalization of full-length IRF2BPL (in white) to the cytoplasm in patient iAs. Representative image was selected from 12 random (63X/Oil) fields captured by Nikon Eclipse Ti2-E. This phenomenon is further confirmed by fractionation studies with lower levels of IRF2BPL seen in the (E) nuclear fraction (normalized to nuclear marker H3) and higher amount of protein observed in the (F) cytoplasmic fraction (normalized to cytoplasmic marker b-tubulin) in patient iAs compared with healthy cell lines. ANOVA followed by Dunnett's multiple comparison test between the mean of the controls and the mean of each line was computed to derive the p value (p), *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Yellow dotted rectangles represent the merged image of DAPI and IRF2BPL stained cell seen within the corner white boxes (dimension of image 40 mm 3 50 mm) for each iAs line. The differentiation from fibroblasts to NPCs was carried out once for each cell line. All experiments were carried out on differentiated iAs with a minimum of four independent biological repeats. Scale bar represents 50 mm. differentiated into iAs, as previously described 16,20,21 (Figure 1B). These iAs express astrocyte-specific markers such as glial fibrillary acid protein (GFAP) and CD44 ( Figure S3). Interestingly, all patient iAs exhibited an activated phenotype with altered morphology and higher GFAP expression when compared with healthy controls ( Figure S3). High expression of GFAP is widely accepted as a marker of activated astrocytes (reviewed in Pekny and Pekna 22 ). Like previously shown in iNs and fibroblasts, western blot quantification of IRF2BPL full-length protein levels demonstrated only the adult patient P3 with significantly lower expression compared with the healthy controls ( Figure 1C). Next, we evaluated the cellular localization of full-length IRF2BPL in patient iAs using immunostaining. Interestingly, full-length IRF2BPL was prominently found to be mislocalized to the cytoplasm in all four patient iAs in the form of smears or aggregated as distinct puncta ( Figure 1D). To confirm these findings, nuclear and cytoplasmic fractionation followed by western blot quantification was performed. Nuclear fractions had lower levels of IRF2BPL, whereas cytoplasmic fractions had higher levels in all patient iAs compared with healthy controls ( Figures 1E and 1F). These data indicate that full-length IRF2BPL is partially mislocalized to the cytoplasm in NEDAMSS patient astrocytes, and this may play a role in the disease pathogenesis.
Truncated IRF2BPL protein dimerizes with and mislocalizes full-length IRF2BPL in vitro A previous study has shown that IRF2BP family proteins (IRF2BP1, IRF2BP2, and IRF2BPL) can mediate homo-or hetero-dimerization/multimerization between different IRF2BP2 family members through their conserved N-terminal IRF2BP zinc finger. 23 Hence, we hypothesized that interactions of fulllength IRF2BPL with truncated forms could underlie sequestration of full-length IRF2BPL in the cytoplasm. Antibodies targeting the N terminus of IRF2BPL are currently not commercially available, and we were unable to confirm the presence of the mutant protein isoforms, but we confirmed the presence of mutated IRF2BPL mRNA in all patient astrocytes ( Figure S4). Hence, our data suggest that the truncated RNA form of IRF2BPL in patients is stable, expressed at similar levels to the wild-type form, and does not undergo degradation. Next, we conducted HA pull-down assays to determine if full-length IRF2BPL and the mutated IRF2BPL forms found in patients can dimerize. We transfected HEK293 cells with constructs either overexpressing N-terminal HA-tagged full-length IRF2BPL (pCMV-HA::FL-IRF2BPL) or the HA-tagged truncated isoforms from each patient cell line (pCMV-HA::P1/P2/P3/P4) (Figures 2A and S5A).
The pull-down assay, using protein lysate from the transfected cells, shows that IRF2BPL protein can dimerize with itself and that the truncated, tagged protein isoforms are stable. Moreover, the truncated protein forms from the patients (P1, P2, P3, and P4) can bind with endogenous full-length protein ( Figures 2B and  S5A). Importantly, aligned with our data from patient cells, fulllength IRF2BPL was partially redistributed to the cytoplasm in the form of aggregates in HEK293 cells expressing the truncated protein forms (Figures 2C and S5B). Co-staining with antibodies against IRF2BPL and the HA tag reveals colocalization of the truncated protein with full-length protein in the cytoplasm ( Figures 2C and S5B). Of note, transfection with HA-tagged truncated protein from P4 showed more protein in the pull-down assay and stronger signal of mislocalization of IRF2BPL ( Figures 2B and 2C) because unlike the other mutations, the antibody can detect both the full-length and the truncated P4 protein. However, we were unable to distinguish HA-P4 and FL-IRF2BPL protein on western blot ( Figure 2B) as their protein sizes (83 and 79 kDA respectively) are very similar.
Cytoplasmic sequestration of full-length IRF2BPL was also confirmed in iAs from a healthy control (H1) after infection with lentivirus overexpressing each truncated IRF2BPL protein isoforms from patients 1-4 via immunofluorescence staining ( Figure 2D). Additionally, there was no overall loss of IRF2BPL upon overexpressing the mutant proteins ( Figure S5D). However, fractionation studies confirmed the mislocalization phenomenon, with lower levels of IRF2BPL protein in the nuclear fraction and higher levels in the cytoplasm upon transduction with N-terminal truncated proteins (P1, P2, and P3) ( Figures S5D and S5E). Interestingly, the mislocalization phenomenon on expression of truncated protein from P4 differed prominently from the other mutant proteins ( Figures 2D, S5D, and S5E). Together, our data show that truncated IRF2BPL protein can dimerize with fulllength IRF2BPL and sequester the protein to the cytoplasm in vitro. This phenomenon is seen in patient cell lines, transfected HEK293 cells, and even in healthy iAs after infection with lentiviral vectors containing the truncated isoforms.
NEDAMSS iAs fail to support neuronal viability that is rescued by CuATSM treatment Astrocytes play a key role in providing structural and metabolic support to neurons and are responsible for the maintenance of brain homeostasis. 18,24 Since NEDAMSS patient astrocytes show aberrant morphology, increased GFAP expression, and strong redistribution of IRF2BPL from the nucleus to the cytoplasm, we assessed if these patient iAs provide proper neuronal support. We conducted coculture assays, as previously described, by co-culturing GFP+ mouse neurons on a monolayer of healthy or patient iAs. 16 After 3 days of coculture, we performed fully automated high-throughput analysis of neuronal survival. We found that neurons cocultured with patient iAs had significantly lower rates of survival than neurons cocultured with healthy iAs (Figures 3A and 3B). In our previous study involving coculture assays with ALS astrocytes, we had observed a similar loss of neuronal viability, and treatment with CuATSM drug significantly improved the survival in certain responding patient lines. 25 CuATSM is a small artificial drug currently being used in clinical trials for ALS 26,27 and Parkinson's disease (PD) (NCT03204929). 26-28 Interestingly, all patientderived iAs treated with 1 mM CuATSM for 4 consecutive days prior to neuron-glia coculture displayed significantly increased neuronal survival ( Figures 3A and 3C). However, treatment with CuATSM did not rescue the mislocalization of IRF2BPL to the cytoplasm in patient iAs ( Figure S6).
Next, we investigated whether loss of IRF2BPL in astrocytes was responsible for reduced neuronal survival in coculture, hypothesizing that the cytoplasmic redistribution seen in patient cell lines leads to reduced availability of the protein in the nucleus, which could mimic a loss of function. We developed a lentivirus-expressing shRNA (LV-sh29) against IRF2BPL that could knock down approximately 50% of the protein in healthy iAs (H1) after 4 days in vitro ( Figure S7A). H1 iAs were transduced with control or shRNA lentivirus 48 h prior to coculture with GFP+ neurons ( Figure S7B). After 3 days of coculture, healthy iAs transduced with LV-sh29 showed a significant loss of GFP+ neuronal survival compared with iAs transduced with LV-RFP. Moreover, overexpression of IRF2BPL (LV-FL-IRF2BPL) in the healthy cell line also showed a slight reduction in neuronal survival ( Figures 3D and 3E). Importantly, overexpressing truncated IRF2BPL proteins (LV-P1, LV-P2, LV-P3, and LV-P4) in healthy iAs decreased neuronal survival in coculture assays in a similar manner as the cells that were transduced with the shRNA construct. This indicates that the sequestration by the mutated protein acts similar to a loss of function and that the nuclear localization of the protein is key for its function ( Figures 3D and 3E). We also evaluated if iAs from severe patients (P1, P2, and P3) release any toxic factors by culturing neurons with their condition media for 48 h. No significant loss of neurons was observed compared with the controls (Figures S8A and S8B). As previously seen in the patient iAs, treating the healthy iAs expressing mutant forms of the protein with 1 mM CuATSM 2 days after transduction with the lentiviral constructs significantly improved neuronal viability ( Figures S7B, 3D, and 3F).
In summary, NEDAMSS iAs and healthy iAs forced to overexpress mutated versions of IRF2BPL lack neuro-supportive function, which can be rescued by treatment with CuATSM. Our data also suggest a loss-of-function mechanism for the disease, resulting from dominant-negative mutations that cause sequestration of the full-length protein from the nucleus into the cytoplasm.
NEDAMSS patient-derived iAs display aberrant mitochondrial respiration that is rescued by CuATSM treatment Since CuATSM showed a beneficial effect without rescuing the mislocalization, we evaluated additional potential disease pathways known to be impacted by CuATSM treatment. Mitochondrial abnormalities and dysfunction are common in many neurodegenerative diseases such as ALS, PD, Alzheimer's disease (AD), and Huntington's disease (HD). 29 Moreover, CuATSM is thought to exert neuroprotective effects via a mitochondrial mechanism. 30,31 To examine if NEDAMSS iAs display alterations in mitochondrial function, we evaluated the expression of cytochrome c oxidase subunit 4 (COX IV), a mitochondrial marker involved in the electron transport chain. 32 In all patient iAs, COX IV showed altered localization, indicating a reduced mitochondrial network with accumulation around the nucleus ( Figure 4A). However, no change in protein expression levels was observed compared with healthy controls ( Figure S9A). Blinded computational analysis of the number of mitochondria per cell compared with healthy controls revealed significantly lower number of mitochondria in patient P1 and significantly higher number in P4 ( Figure 4B). Interestingly, all patient iAs had a significantly higher fractionation index in comparison to healthy controls ( Figure 4C), indicating a more fragmented mitochondrial network that is commonly seen in disease states. 33,34 Based on the above observations, we next evaluated the functionality of mitochondria by conducting Seahorse ATP real-time rate assay. All patient iAs, except P1, exhibited higher levels of base oxygen consumption rate and ATP-linked mitochondrial respiration, which was measured following inhibition of ATP synthase using oligomycin ( Figures 4D and 4E). The non-mitochondrial respiration was subtracted from the readings by using a mixture of rotenone and antimycin C in the assay ( Figure S9B). Remarkably, treating patient P2, P3, and P4 iAs with CuATSM reduced the high levels of respiration to normal control levels ( Figures 4D and 4E). Together, our data show that NEDAMSS patient iAs have aberrant mitochondrial networks and elevated levels of mitochondrial respiration (except in patient P1). Importantly, CuATSM treatment can rescue the abnormal respiration levels.
Genes involved in neuronal development and mitochondrial function are differentially expressed in NEDAMSS astrocytes Given the role of IRF2BPL as a DNA-binding protein in transcription, 1,2 we determined the global changes in gene expression in (D) Representative images of GFP+ neurons following coculture with H1 astrocytes un-transduced or transduced with lentivirus expressing RFP, FL-IRF2BPL, shRNA against IRF2BPL, or truncated patient (P1, P2, P3, and P4) IRF2BPL protein with DMSO or CuATSM treatment. (E) Quantification of neuronal survival shows significant loss in viability upon overexpression, loss, or mislocalization of IRF2BPL. (F) Treatment with CuATSM ameliorates the effect in healthy iAs transduced with truncated proteins. The same DMSO treated cell lines from (E) were used for the comparison. Data were normalized to wild-type un-transduced (UT) H1 astrocytes. ANOVA followed by Dunnett's multiple comparison test between the mean of the controls and the mean of each line was computed to derive the significance (p value). Significance between the treated and the untreated groups was computed using unpaired t tests. A minimum of three different culture repeats was used. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Scale bar represents 200 mm. NEDAMSS-associated iAs. We compared the transcription profiles of patient and healthy iAs and found 37 genes differentially expressed in all patient cell lines compared with the healthy controls ( Figure 5A and Table S2). We performed computational analysis and a non-exhaustive review of the function of these genes and classified them into two functional groups relating to our in vitro experimental findings (Table S3). We show gene assignments to these groups with brief evidence from the literature supporting the assignments. Specifically, 19 genes are involved in neuronal growth, development, function, and support, and seven are associated with mitochondrial function, development, and/or metabolism ( Figure 5A and Table S3). Of these genes, we particularly identified SSTR2, ERG2, AKT3, and NRG2 as functionally important in both capacities and validated their differential expression between patients and controls through qPCR ( Figure 5B). However, CuATSM treatment did not alter the expression of the 37 differentially expressed genes identified between patients and controls (Table S4). Moreover, gene expression changes do occur in a different manner between individual patient cell lines, indicating a patient-specific effect of the IRF2BPL mutations. Importantly, CuATSM improvements in phenotype are associated with mutation-specific gene expression changes that normalize the overall gene expression profile of each patient cell line rather than affecting a global disease-specific gene expression pattern ( Figure 5C).

Full-length IRF2BPL is sequestered to the cytoplasm upon co-expression with truncated IRF2BPL in vivo
To determine if NEDAMSS-associated truncations in IRF2BPL lead to sequestration of the full-length IRF2BPL in the cytoplasm in vivo, we expressed untagged human IRF2BPL and the NEDAMSS-related truncation, IRF2BPL p.E172X ::HA (P1::HA), in the pouch region of the developing wing-disc of the fruit fly using the GAL4-UAS system 35 (Figures 6A and S10A). Expression of full-length IRF2BPL alone leads to the localization of this protein predominantly in the nucleus ( Figure S10A). In contrast, expression of the P1::HA reveals punctate cytoplasmic distribution ( Figures 6B and S10A). Upon co-expression, we observed that full-length IRF2BPL protein partially colocalizes with the truncated form of IRF2BPL in the cytoplasm ( Figure 6B). These data further confirm that the truncated IRF2BPL can sequester some of the full-length protein into the cytoplasm in vivo.
To examine if heterozygous truncations in IRF2BPL can affect full-length function in vivo, we performed genetic interaction studies in flies. The overexpression of IRF2BPL using the glia cell promoter, Repo-GAL4, results in pupal lethality at 29 C ( Figure S10B). Importantly, overexpression of the fly ortholog, Pits, in glia, also leads to pupal lethality, showing conserved function ( Figure S10B). In contrast, glial expression of P1::HA results in viable adult flies. However, co-expression of full-length IRF2BPL and P1::HA causes both third instar larval and pupal lethality. This earlier lethality suggests the interaction of NEDAMSS-related truncated IRF2BPL and full-length protein could be toxic in vivo.
CuATSM treatment is protective in a Drosophila model of loss of IRF2BPL The fly ortholog, Pits, is the single IRF2BP family member in the Drosophila genome and is essential in flies, as Pits TG4 mutants (a GAL4 insertion in Pits) are homozygous lethal. 8 This allele splices in a cassette containing a splice acceptor (SA)-T2A-GAL4-polyA, 36,37 resulting in the premature truncation of the endogenous Pits gene ( Figure 6C). Therefore, to mimic the human condition, we characterized heterozygous Pits TG4 /+ flies that express one full-length copy of Pits as well as a truncated form. As a control line, we used the heterozygous Pits MiMIC /+ flies 38 (Figure 6C). Critically, Pits MiMIC flies act as an ideal control for these experiments as they harbor a non-mutagenetic cassette in the same location as Pits TG4 , and they are homozygous viable. Due to unavailability of antibodies against Pits, we could not assess for mislocalization of the fly protein.
Interestingly, in accordance with our in vitro results, these Pits TG4 /+ flies display increased ATP production in flies aged 35 days post eclosion (d.p.e), while no change in total ATP levels was observed in younger flies aged 15 d.p.e. ( Figure 6D). The partial loss of Pits by neuronal knockdown using RNAi is known to cause progressive climbing defects in flies. 8 In line with this, Pits TG4 /+ flies show increased time to climb in the negative geotaxis assay at 35 d.p.e. Climbing defects were not observed at 25 d.p.e., indicating an age-dependent effect ( Figure 6E). Next, we determined if CuATSM provided in fly food ameliorates the climbing defects observed in 35-day-old flies. 35-day-old Pits TG4 /+ flies treated with CuATSM display decreased time to climb compared with DMSO-treated animals, indicating that CuATSM can suppress the climbing defects in these flies ( Figure 6F). Together, similar to our results in NEDAMSS patient cells, partial loss of Pits in flies leads to functional deficits that can be partially rescued by CuATSM treatment, further underlining the potential of this small molecule for treatment of NEDAMSS.

DISCUSSION
Since the recent discovery of NEDAMSS, several new cases and phenotypes in patients have been documented. [8][9][10][11][12]39,40 Moreover, models in fruit flies and zebrafish have highlighted a role for IRF2BPL in neuronal function and maintenance. 14 These models focus on IRF2BPL loss of function, whereas in attempts to offer insight on the mechanism of action of truncating alleles, we focus on human patient cells carrying NEDAMSS truncations.
(C) All patients exhibit fragmented mitochondrial network compared with healthy iAs. The ''fractionation index'' parameter measured is proportional to how fractionated the mitochondrial network was. The computation was conducted on nine image fields for each line taken by Nikon microscope at 403 magnification. The analysis was conducted for three separate biological repeats, and a minimum of 20 cells were analyzed per n. (D) The base oxygen consumption rate was measured and plotted at three timepoints for healthy, patient, and respective CuATSM-treated iAs. (E) ATP-linked respiration was also measured following ATP synthase shutdown using oligomycin drug and subtracting the oligo OCR from basal OCR. Data represent a minimum of three independent experiments with different cultures. ANOVA followed by Dunnett's multiple comparison test between the mean of the controls and the mean of each line was computed to derive the p value (p). Significance between the treated and the untreated groups was computed using unpaired t tests. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Scale bar represents 50 mm.
Cell Reports 41, 111751, December 6, 2022 9 Article ll OPEN ACCESS Here, we have developed an in vitro disease model for NEDAMSS using direct conversion methods to efficiently reprogram patient fibroblasts to neurons or neuronal progenitor cells, which are further differentiated to iAs as previously described. 15,16 Direct reprogramming to model neurological disorders has several advantages over traditional iPSC method  including speed, the ability to bypass clonal isolation of single colonies or a pluripotent stem cell state, and retention of valuable epigenetic information. 21, [41][42][43] This method has been used successfully in multiple laboratories studying various neurological disorders. [44][45][46][47] We chose to study the role of astrocytes in NEDAMSS as they often represent an overlooked but important source of disease pathogenesis. Astrocytes perform many roles to modulate development and neuronal activity including direct uptake and release of neurotransmitters, metabolism and regulation of metabolites, regulation of inflammation, establishment of brain structure, including the blood-brain barrier, and glial scarring in response to injury. 17,24,[48][49][50] Some or all of these processes may be disrupted in neurological disease/trauma. 19 Astrocyte dysfunction has been shown to contribute to pathology in ALS, AD, PD, HD, tuberous sclerosis, Rett syndrome, and other autism spectrum and seizure disorders. 19,51 Our four subjects with heterozygous nonsense mutations in IRF2BPL gene exhibited a progressive course of neurological regression. Patients P1, P2, and P3 demonstrated severe developmental disability and loss of motor skills, whereas the youngest patient, P4, had a mild form of developmental delay. [8][9][10] Our results show that most patient cell lines, with exception of the adult sample (P3), did not display the expected reduction in IRF2BPL protein expression that would be indicative of insufficient protein production from one allele (haploinsufficiency). Instead, we observed mislocalization of IRF2BPL protein that was highly prominent in NEDAMSS astrocytes leading to smears and aggregation in the cytoplasm. Aggregates of peptides or proteins are a major hallmark of many neurological diseases including ALS, frontotemporal dementia, AD, PD, HD, Creutzfeldt-Jacob's Disease, and spinal and bulbar muscular atrophy. 52,53 Intriguingly, the truncated protein in patient P4 includes the NLS unlike the mutated proteins from other patients. Similar to subject P4, mild autistic syndrome has been reported by another NEDAMSS patient (N701fs) that has the NLS preserved. 54 Thus, it is possible that the preservation of the NLS correlates with a milder form of the disease. However, while we saw differences in the amount of IRF2BPL in the nucleus and cytoplasm of healthy astrocytes transduced with the P4 mutant, we did not observe any difference when using the patients' original cells without overexpressing the mutant protein.
Thus, further studies will be needed to evaluate the impact of an intact NLS on the patient phenotype and potential correlation to mislocalization pattern in the future.
In addition to the altered protein localization, astrocytes from NEDAMSS patients were non-supportive to cocultured neurons and showed an aberrant mitochondrial phenotype with increased network fractionation and mitochondrial respiration. Such an increase in oxidative phosphorylation indicates mitochondrial damage and inefficiency, which has been shown previously in AD neurons. 55 The disease phenotypes shown in vitro were ameliorated by the treatment of NEDAMSS astrocytes with a small molecule, CuATSM. Importantly, the findings from our in vitro assays were recapitulated in the fly model where we confirmed mislocalization of IRF2BPL via co-expression with the truncated form and age-dependent reduction in motor function, which was ameliorated by CuATSM treatment in a dominant model of truncated Pits (IRF2BPL ortholog) in flies.
Our data also show global transcriptional changes in patient iAs compared with healthy cell lines. Somatostatin receptor 2 (SSTR2) was highly downregulated in patient iAs. This G-protein-coupled receptor is involved in mitochondrial-mediated apoptosis and contributes to neuromodulatory effects in the cerebral cortex, 56 and its downregulation may contribute to aberrant phenotypes shown in vitro in patient iAs. Early growth response protein 2 (ERG2) dysregulation could lead to multiple phenotypes displayed by patient iAs as it has been proposed to play a role in dendritic complexity, cognition, and maintenance of mitochondrial membrane potential. 57 Moreover, defects in this gene are associated with Charcot-Marie-Tooth disease. 58,59 Similarly, upregulation of AKT serine/threonine kinase 3 and neuregulin-2 can be detrimental to mitochondrial function 60,61 or neuronal development, 62 respectively. In a recent collaborative report, we discovered that loss of IRF2BPL in NEDAMSS models in Drosophila, zebrafish, and patient iAs causes increased WNT1 transcription. 14 Our RNA-seq data show that some WNT ligands (WNT5A, WNT16, WNT9B, and WNT10B) are also upregulated in NEDAMSS iAs (Table S2). Interestingly, increased Wnt signaling leads to increased mitochondrial biogenesis 63 as well as changes in mitochondrial morphology and dynamics. 64 Thus, altered Wnt signaling could in part underlie the increase in mitochondrial respiration and changes in mitochondrial dynamics in NEDAMSS iAs. Further studies are required to determine if loss of IRF2BPL activity as a transcriptional regulator directly causes differential expression of neurodevelopmental and metabolic genes or if both are interrelated in a different fashion.
CuATSM is a positron emission tomography imaging agent that can readily cross the blood-brain barrier. 26, 65 The drug has a low reduction potential that facilitates selective release of copper in damaged cells. 31, 65 Studies have shown that CuATSM exerts selective, neuroprotective effects in disease-affected regions of the CNS in ALS, PD, and mitochondrial encephalopathy, lactic acidosis, and stroke-like episodes syndrome. 26,30,65 Importantly, results from the phase I clinical trial showed that CuATSM could slow disease progression and improve the respiratory and cognitive function in ALS patients, but the mechanism of action is not fully understood. 66 Interestingly, treatment of NEDAMSS patient iAs with CuATSM significantly improved their supportive role for neurons in coculture assays and reduced the increased mitochondrial respiration in patients to normal rates. Upon investigating the mechanism of action of CuATSM, we found that the drug does not resolve the mislocalization of IRF2BPL protein in patient iAs, which might indicate that the aggregates are not mechanistically responsible for the lack of neuronal support or mitochondrial dysfunction. This is consistent with findings from the literature that protein aggregates might not necessarily be inherently toxic. 67 For example, recent evidence has indicated that in AD, soluble Ab oligomers, rather than the insoluble aggregates, are the main neurotoxic species (reviewed in Zhou and Liu 68 ). Interestingly, previous studies have reported that CuATSM significantly improved the survival and locomotor function in ALS-related SOD1 G37R mouse model, without decreasing misfolded SOD1 aggregates. 26,30,31,65,69 In our study, the transcriptional profile of drug-treated NEDAMSS patient iAs shows upregulation of metal ion homeostasis pathways, which is essential for cellular viability (Table S4). Hence, selective metal homeostasis could result in improved mitochondrial function in patient iAs, which in turn could lead to neuroprotection. Of note, patient P1 iAs did not show elevated mitochondrial respiration, but they showed elevated WNT transcription and still responded beneficially to CuATSM treatment. It is possible that this discrepancy could be due to the overall lower number of mitochondria in these cells, or alternatively, other pathways could be involved or act in combination to improve neuron viability. Furthermore, as the knockdown of IRF2BPL in healthy iAs resulted in neuronal loss in coculture assays similar to that observed with coculture of NEDAMSS patient lines, it is possible that the cellular phenotypes in NEDAMSS iAs likely stem from loss of IRF2BPL nuclear function rather than toxic gain of function from mislocalized or aggregated IRF2BPL. Further studies need to be conducted to shed light upon how loss/reduction of IRF2BPL impacts metal homeostasis or if abnormal Wnt signaling or other factors are responsible for the phenotypes.
Overall, we have generated an in vitro modeling system for NEDAMSS. Using this model, we discovered a cytoplasmic sequestration of full-length IRF2BPL due to its interaction with mutant truncated IRF2BPL in NEDAMSS astrocytes. We report cellular phenotypes including failure of NEDAMSS astrocytes to support neuron growth and survival and aberrant cellular metabolism that could be attributed to the loss of IRF2BPL function. Our data also show that gene expression changes and therapeutic response associated with these phenotypes demonstrate both a disease-specific and mutation-specific pattern, potentially indicating multiple mechanisms of NEDAMSS pathogenicity. Finally, we have identified CuATSM as a promising therapeutic that may be able to slow or halt the progression of NEDAMSS. Future research is required to understand the role of IRF2BPL in causing mitochondrial dysfunction in patients. This would be of particular interest as mitochondrial dysfunction/aberrant cellular metabolism is emerging as an increasingly important factor in many neurological/seizure disorders. 70 Therefore, improvement of mitochondrial function/health could become a potential therapeutic strategy with implications across a wide spectrum of disorders.
Limitations of the study Unlike other patient lines, only the adult subject (P3) showed a significant loss of IRF2BPL protein. Additional adult patient cell lines need to be investigated to determine if age-dependent loss of IRF2BPL takes place in patients. The cytoplasmic sequestration of full-length IRF2BPL was observed in cases of NEDAMSS with nonsense mutations, and it remains unclear whether cytoplasmic sequestration occurs in cells derived from individuals with missense variants or other mutations preserving the NLS.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
The authors thank the California Center for Rare Diseases at UCLA for provision of primary patient cultures. We specially would like to thank the Families of SCN2A Foundation for sponsoring the Seahorse equipment. We are indebted to Dr. Joe Beckman for gifting us the CuATSM drug that was used in our study. We also would like to acknowledge Annalisa Hartlaub for her immense support in laboratory organization, sorting of neurons for coculture assays, and image quantification.  8 UAS-IRF2BPL E172X ::HA flies were generated as previously described. 72 Briefly, site-directed mutagenesis was performed with the Q5 site-directed mutagenesis kit (NEB) on the IRF2BPL clone in the pDONR223 entry vector. 8 Primers listed in Table S5 were used to obtain IRF2BPL E172X ::HA (P1:HA) followed by Sanger verification. Using Gateway cloning (Thermo Fisher Scientific), the P1::HA cDNA entry clone in the pDONR223 vector was shuttled to the pUASg-attB-HA. 73 The construct was inserted into the VK37 (PBac{y[+]-attP}VK00037) docking site by 4C31 mediated transgenesis. 74

METHOD DETAILS
Direct conversion of fibroblasts to neurons Patient and healthy fibroblasts were directly converted to neurons using small molecules as described previously with few modifications. 15 Briefly, 24-well plates with cover slips or 10 cm plates were coated with poly-D-lysine (50 mg/mL, Sigma) in borate buffer for an hour at room temperature. Next the plates were washed with Dulbecco's phosphate buffered saline (DPBS) (Gibco) and coated with laminin (10 mg/mL) in DMEM/F12 (Gibco) at 37 C for 2 h. Fibroblast cells at a density of 25,000 for a 24 well and 800,000 for a 10 cm plate were seeded with fibroblast culture medium for 1 day. The next day the cells were transferred into neuronal induction medium (DMEM/F12: Neurobasal [1:1] (Gibco) with 0.5% N-2 (Gibco), 1% B-27 (Gibco), cAMP (100 mM, Sigma), and bFGF-2 (20 ng/mL, Peprotech) with the following chemicals: VPA (0.5 mM, Sigma), CHIR99021 (3 mM, Axon medchem), repsox (1 mM, Biovision), forskolin (10 mM, Tocris), SP600125 (10 mM, Sigma), GO6983 (5 mM, Sigma) and Y-27632 (5 mM, Sigma). Half of the medium containing the chemicals was changed after 3 days with fresh induction medium. On the fifth day, cells were switched to neuronal maturation medium (DMEM/F12: Neurobasal [1:1] with 0.5% N-2, 1% B-27, cAMP (100 mM), bFGF-2 (20 ng/mL), BDNF (20 ng/mL, Peprotech) and GDNF (20 ng/mL, Peprotech) with the following chemicals: CHIR99021 (3 mM), forskolin (10 mM) and SP600125 (10 mM). The induced neurons were then fixed for immunofluorescence or pelleted for Western blot. Nikon Eclipse Ti2-E motorized inverted microscope and Zeiss LSM 800 confocal microscope were used to image the morphology of the neurons. The neuronal conversion rate (% of Tuj1+ soma/total number of cells stained with DAPI) was calculated by manually analyzing randomly selected 15 (203 magnification) fields for each line from three independent replicates by a blinded investigator. The neurite length was computed from the same set of images by using SNT plugin of image-analyzing tool, Fiji. 75 Reprogramming of fibroblasts to astrocytes Skin fibroblasts were directly converted to induced neuronal progenitor cells (iNPCs) by using a previously described method. 16 The iNPCs were maintained in fibronectin (2.5 mg/mL, Millipore) coated dishes and DMEM/F12 media containing 1% N2 supplement, 1% B27 and 20 ng/mL bFGF-2. To differentiate iNPCs to astrocytes, a small portion (20% of a confluent plate) of the cells was seeded onto fibronectin-coated dishes with DMEM/Glutamax (Gibco) media containing 10% FBS (Gibco) and 0.2% N2. Five days post differentiation induced astrocytes were characterized for astrocyte-specific markers (GFAP and CD44) by immunofluorescence.

Immunofluorescence
Fibroblasts (60,000 cells) and astrocytes (40,000 cells) were seeded on 24 well plates with fibronectin-coated coverslips. The next day, cells were fixed with 4% paraformaldehyde for 15 min and washed 33 with DPBS before the blocking solution consisting of DPBS with 10% goat serum (Gibco), 0.1% Triton X-100 (Sigma-Aldrich), and 0.1% Tween 20 (Fisher Scientific) was applied for 1 h. Chemicallyinduced neuronal cells were fixed with ice-cold 4% paraformaldehyde (Sigma-Aldrich) and 0.1% glutaraldehyde (Sigma-Aldrich) for 20 min instead, and blocked with ice-cold DPBS with 4% goat serum and 0.2% Triton for 1 h. All primary antibodies were diluted in blocking solution and their dilution and provider are listed in the key resources table. Incubation of the primary antibody was performed overnight at 4 C. The following day, cells were washed 33 in DPBS before the secondary antibody (Alexa Fluor) and DAPI (Thermo Fisher Scientific) diluted in blocking solution was applied for 1 h at room temperature. Following three washes with DPBS, the coverslips were mounted in Vectashield (Vector Labs) and sealed. Images were captured either with Nikon Eclipse Ti2-E motorized inverted microscope or the Zeiss LSM 800 confocal microscope and processed with Adobe Photoshop. Anti-IRF2BPL antibody from Novus Biologics (NBP2-14712) which has a different target site was also used to check for mislocalization in patient iAs.