Neurodevelopmental alterations are key drivers of SPG56

Abstract

Biallelic pathological variants in CYP2U1 are associated with SPG56, a complex ultra-rare form of hereditary spastic paraplegia. CYP2U1 encodes a member of the cytochrome P450 family 2, which is highly expressed in the brain and involved in arachidonic acid metabolism and lipid hydroxylation. Little is known about the early stages of neurodevelopment in SPG56. To gain insight into the molecular mechanisms of the disease, we generated and characterized both a zebrafish knock-out model (cyp2u1^−/−) and cortical neurons derived from induced pluripotent stem cells (iPSCs) of two patients affected by SPG56. cyp2u1^−/− zebrafish showed increased mortality and locomotor impairment, and transcriptome analysis revealed defects in the cell cycle, neural development, and oxidative phosphorylation pathways, consistent with observed decreased cell proliferation, increased apoptosis, and bioenergetic impairment. To assess whether these molecular alterations impact network function, we performed local field potential recordings and whole-brain calcium imaging in mutant larvae. Local field potential analyses revealed prolonged high-frequency events despite reduced overall power, and calcium imaging showed increased skewness of neuronal activity distributions, findings indicating altered brain network dynamics in mutants. Transcriptional profiling of cortical cultures performed at neural progenitor, early-born, and mature neuron stages pointed to altered neural network development as one of the main pathogenic mechanisms of the disease and highlighted several similarities with the zebrafish model. Multiple features of the cyp2u1^−/− zebrafish recall the human SPG56 phenotype, and combined transcriptomics profiling in cyp2u1^−/− zebrafish and SPG56 patient iPSC-derived cortical neurons supports impaired neural network development as a key disease mechanism.

Introduction

Biallelic pathological variants in CYP2U1 are associated with SPG56 (MIM #615030), an ultrarare, early-onset, autosomal recessive form of hereditary spastic paraplegia (HSP). To date, 52 patients from 34 families have been described in literature [1–26], exhibiting numerous variants. These patients show heterogeneous clinical manifestations, including early-onset spasticity and cognitive impairment, as well as multiple neuroimaging features such as basal ganglia calcification, hypomyelinating leukoencephalopathy, and different degrees of cerebral and cerebellar atrophy. Macular degeneration develops with disease progression and is invariably present in adult-onset cases. Allelic variants in CYP2U1 also occur in pseudoxanthoma elasticum [16], a complex condition affecting the eye, coronary arteries, and skin, and presenting with soft, ivory-colored papules in a reticular pattern predominantly affecting the neck and large flexor surfaces. Similar features occasionally occur in children with SPG56. Co-occurrence of SPG56 with CLCN2-related leukoencephalopathy (MIM #615651) is reported in a single kindred [20]. Disease severity varies, from severe to mild, and a few reports even describe CYP2U1-related maculopathy in patients without neurological manifestations [15, 19, 23].

CYP2U1 was initially identified as a modulator of the arachidonic acid signaling pathway, and subsequent studies showed it to be involved in the hydroxylation of arachidonic acid to 19- and 20-hydroxyeicosatetraenoic acids and fatty acids [27–31]. However, the significance of alterations of these processes in pathological conditions caused by CYP2U1 mutations remains unclear. CYP2U1 was found to be expressed in the rat thymus and brain, particularly the cortex and limbic structures [28], and throughout the human brain, although its distribution varies across regions [29]. CYP2U1 is reported to show predominantly mitochondrial localization [28]. To date, the pertinent literature contains a limited set of functional studies in this condition. In the first SPG56 case series, described in 2012, in vitro analyses indicated aberrant mitochondrial architecture and bioenergetics with increased oxidative stress [1]. Later, the pathogenicity of several gene variants was determined by spectroscopic and enzymatic characterization in HEK293T cells overexpressing human CYP2U1 missense mutations [11]. The only in vivo study thus far reported—on the Cyp2u1^−/− model— investigated the lipidomic and metabolomic consequences of protein loss in murine brains and in human SPG56 blood samples [32]. Reduced vitamin B2 and folate levels were detected both in SPG56 patients’ samples and in Cyp2u1^−/− mice, together with elevated neopterin levels in the cerebrospinal fluid. Furthermore, the same study showed that progression of the disease was accompanied by enhanced oxidative stress, impaired mitochondrial energy metabolism, and an altered interferon signature. Additionally, folate supplementation appeared to limit cognitive impairment in KO mice [32]. Corroborating these findings, a recent paper describes two siblings with SPG56 showing relative clinical stability after nearly two decades of folinic acid supplementation [21].

Since the aforementioned KO mouse model does not adequately model the early stages of neurodevelopment in SPG56, we generated and characterized a zebrafish cyp2u1 knock-out model (cyp2u1^−/−) and also differentiated human cortical neurons from induced pluripotent stem cells (iPSCs) obtained from two patients with SPG56. We speculated that investigating two models of early neuronal impairment would provide us with further insights into the molecular mechanisms underlying the disease.

Results

Generation of cyp2u1^−/− zebrafish model

The zebrafish cyp2u1 gene (ENSDART00000048281.8) encodes a 533-amino-acid protein (UniProt code B8A5Q1) that shares 58.2% identity with human CYP2U1 (ENST00000332884.11, UniProt code Q7Z449–1) and was recognized as orthologous based on sequence similarity [33]. We analyzed cyp2u1 expression throughout zebrafish development and found it to be strong at the earliest developmental stage (0 hours post fertilization, hpf). Following the midblastula transition at 6 hpf, expression declines (24 hpf) before returning to consistently detectable levels over the following days, as shown by measurements up to 120 hpf (Fig. 1A). We generated a zebrafish knock-out cyp2u1 model (cyp2u1^−/−) that harbored a 2-bp deletion in exon 2 leading to a nonsense mutation (c.688_689del, p.His230Ter) (Fig. 1B). Having crossed heterozygous siblings in the F1 line, the homozygous cyp2u1^−/− F2 line grew to adulthood, and its F3 offspring were used to assess the mutant phenotype. qPCR analysis revealed a severe reduction in cyp2u1 mRNA levels in F3 mutants compared with the control strain at 120 hpf (Fig. 1C).

Figure 1.

(A) cyp2u1 mRNA expression during zebrafish development. (B) Sequencing electropherograms of ctrl, heterozygous, and homozygous individuals indicating successful editing of cyp2u1. (C) qPCR analysis revealed a decrease in the level of cyp2u1 mRNA expression between mutants and Ctrls. T-test was used for statistical analysis. Mean ± SEM is shown. ^***p<0.001. FC = fold change.

cyp2u1^−/− show locomotor impairment and increased mortality

The cyp2u1^−/− model showed significantly increased mortality by 120 hpf (Fig. 2A), although mutant larvae were comparable to controls in terms of body length and eye size (Fig. 2B). Motor behavior assessments at multiple developmental stages revealed an increase in spontaneous tail coiling at 30 hpf in cyp2u1^−/− (Fig. 2C), along with impaired locomotion (reduced swimming distance and velocity, clockwise and counterclockwise rotations) in 120 hpf mutants compared with control larvae (Fig. 2D). Other metrics including heading, meander, turn angle, and angular velocity resulted not significant (Supplementary Fig. S1), thus potentially suggesting a ‘hypokinetic’ rather than ‘ataxic’ pattern in mutant larvae. To explore the nature of this motor deficit, we investigated the morphology of spinal motor neurons using acetylated α-tubulin staining and muscle tissue with a birefringence assay. Neither motor nor sensory neuron abnormalities, nor defects in muscle structure or compaction, were found (Fig. 2E and F).

Figure 2.

(A) Mortality in cyp2u1^−/− mutants was increased vs. Ctrls (n = 300 per group). Fisher’s exact test was used for statistical analysis. ^****p<0.0001 (B) No morphological abnormalities were identified in 120hpf mutants (n = 20 per group). The Mann–Whitney test was used for statistical analysis. Mean ± SEM is shown. (C) Coiling frequency was reduced in cyp2u1^−/− mutants at 30 hpf (n = 255 per group). The Mann–Whitney test was used for statistical analysis. Mean ± SEM is shown. ^****p<0.0001. (D) Analysis of motor activity at 120 hpf showed decreased locomotion in terms of distance moved, velocity, clockwise and counterclockwise rotations (n = 252 per group). The Mann–Whitney test was used for statistical analysis. Mean ± SEM is shown. ^***p<0.001, ^****p<0.0001. (E, F) muscle structure was similar in mutants and controls at 48 and 120 hpf (n = 20 per group).

RNA-seq analysis in cyp2u1^−/− reveals defects in cell cycle progression, neural development, and oxidative phosphorylation

To determine the effects of cyp2u1 loss, we performed RNA-seq analysis on mutant and control whole larvae at 120 hpf. Transcriptomic analysis using a threshold of |log₂ fold change| (lfc)>0.6 identified 3878 differentially expressed genes (DEGs) (Supplementary Table S1), of which 1836 were upregulated and 2042 downregulated. We identified several altered pathways primarily associated with cell cycle regulation and neuronal projection development (Fig. 3A and B). Some of the genes in these pathways are cell cycle regulator genes (Fig. 3C). Our analysis also showed altered expression of genes involved in eye determination, the visual cycle, and retinol metabolism (Fig. 3D). To detect even subtle alterations in the zebrafish model, we then applied less stringent criteria, considering all DEGs with an adjusted p-value<0.05 regardless of fold change, which revealed dysregulation of mitochondrial and oxidative phosphorylation pathways. (Supplementary Fig. S2). Since CYP2U1 is known to be involved in the hydroxylation of arachidonic acid and fatty acids, we searched in transcriptomic data for DEGs associated with these pathways. The cyp2u1^−/− model showed eight DEGs involved in the arachidonic acid cycle (namely, pla2g1b, ptges3a, ptges3b, hpgd, lox2la, lox3la, alox5b.3, and fabp3), and seven in lipid metabolism (pnpla2, abcb4, abcg5, srebf1, srebf2, pparaa, and hmgcra) (Supplementary Table S1). We also identified three DEGs (coq3, coq8ab, and coq9) involved in CoQ10 biosynthesis (Supplementary Table S1), consistent with findings previously reported in biological samples from SPG56 patients [32].

Figure 3.

(A) Pathways identified by gene ontology enrichment. padj<0.05, |lfc|>0.6 (B) pathways identified by KEGG enrichment. padj<0.05, |lfc|>0.6 (C) regulator genes of cell cycle (left) and genes involved in eye determination (right) identified by transcriptomic analysis. ± log fold change SE is shown.

Neural network activity is altered in cyp2u1^−/− mutant zebrafish

To further investigate neural circuit dysfunction in cyp2u1^−/− larvae, we performed local field potential recordings and whole-brain calcium imaging. Electrophysiological analysis revealed prolonged events in the 30–95 Hz frequency band in mutants, despite their reduced overall power, consistent with abnormal temporal organization of network activity (Fig. 4A). In parallel, calcium imaging analysis showed significantly increased skewness in mutant larvae compared to controls, indicating a more asymmetric distribution of neuronal activity and supporting the presence of altered brain network dynamics in the mutant condition (Fig. 4B).

Figure 4.

(A) Representative traces of the local field potential (LFP) recordings analyzed with MATLAB software in the two experimental groups. The graphs represent the duration and power of identified events in controls (n = 11) and cyp2u1^−/− (n = 11) 120 hpf larvae. The Mann–Whitney test was used for statistical analysis. ^*p<0.05, ^****p<0.0001. Mean ± SEM is shown. (B) Representative z-projection images of the brain calcium imaging performed in controls and cyp2u1^−/− in 5 dpf larvae in 16 min of recording, on the right there is the lookup table colour range of calcium fluorescence. Graphs show analysis of calcium imaging performed by calculating Pearson’s coefficient of skewness of events in the two groups. The Mann–Whitney test was used for statistical analysis. ^*p<0.05. Mean ± SEM is shown.

cyp2u1^−/− mutants shows decreased cell proliferation, increased apoptosis, and bioenergetic impairment

To corroborate RNA-seq data, we performed whole-mount immunofluorescence analysis using phospho-histone H3 antibody to determine the rate of proliferative cells and found a decrease in replicating cells at 30 hpf (Fig. 5A). Furthermore, staining with acridine orange highlighted an increase in apoptotic cells in cyp2u1^−/− larvae compared with controls (Fig. 5B). Through qPCR analysis we validated the expression of three respiratory chain complex subunit genes, e.g. ndufb8, cox7c, and atp5a1 (Fig. 5C). In parallel, in vivo micro-oxygraphy analysis revealed differences between mutants and controls in terms of coupling efficiency, ATP production, and non-mitochondrial oxygen consumption (Fig. 5D).

Figure 5.

(A) Detection of PHH3-positive cells in head and tail at 30 hpf. (n = 30 per group). The Mann–Whitney test was used for statistical analysis. ^*p<0.05, ^**p<0.01. Scale bar = 100 μm (B) acridine orange staining for identification of apoptotic cells in head and tail at 30 hpf (n = 30 per group). The Mann–Whitney test was used for statistical analysis. ^*p<0.05, ^**p<0.01. Scale bar = 100 μm (C) qPCR validation of genes involved in oxidative phosphorylation at 120 hpf. T-test was used for statistical analysis. Mean ± SEM is shown. ^*p<0.05. (D) Seahorse analysis for oxygen consumption rate (OCR) measurement of controls and mutant larvae at 120 hpf (n = 18 for control group, n = 27 for cyp2u1^−/− mutants). T-test was used for statistical analysis. FC = fold change. Mean ± SEM is shown. ^*p<0.05, ^**p<0.01, ^***p<0.001.

Folinic acid and 5-methyltetrahydrofolate (5-MTHF) treatment do not rescue motor phenotype in cyp2u1^−/− larvae

Since previous findings showed folate pathway alterations in SPG56 [32] and a single report described long-term benefits of folinic acid supplementation in two siblings [21], we examined our RNA-seq analysis results relative to the folate pathway in cyp2u1^−/−. Our transcriptomic data showed upregulation of a single gene, dhfr (Supplementary Table S1), encoding dihydrofolate reductase, an enzyme that catalyzes the reduction of dihydrofolic acid to tetrahydrofolic acid; although altered pathways did not show enrichment of folate cycle genes, qPCR analysis on these genes confirmed the altered expression of dhfr. We also found dysregulated expressions of mthfr and mthfd1l, which encode methylenetetrahydrofolate reductase and methylenetetrahydrofolate dehydrogenase 1-like, respectively (Fig. 6A). However, neither folinic acid treatment (Fig. 6B) nor treatment with 5-MTHF [34, 35], an active form of folic acid (Fig. 6C), produced any rescue of the phenotype in mutant larvae.

Figure 6.

(A) qPCR analysis of genes related to the folate cycle. T-test was used for statistical analysis. Mean ± SEM is shown. ^*p<0.05, ^**p<0.01, ^***p<0.001. (B) Survival and locomotion analysis after treatment with 20 μM folinic acid. The Kruskal-Wallis test was used for statistical analysis. Mean ± SEM is shown. ^*p<0.05, ^**p<0.01 (C) survival and locomotion analysis after treatment with 30 μM levomefolic acid (MTHF). The Kruskal-Wallis test was used for statistical analysis. Mean ± SEM is shown. ^*p<0.05, ^***p<0.001. FC = fold change, DMSO = dimethyl sulfoxide.

SPG56 iPSC-derived cerebrocortical cells show decreased proliferation and premature neurogenesis

We generated cerebrocortical derivatives from two individuals with a clinical and genetic diagnosis of SPG56 (patients already described in [32]) with the aim of disentangling the molecular mechanisms of the disease. Patient 1 (Pt 1) harbored a maternal splicing mutation (c.1288+1G>A) and a paternal missense mutation (c.343G>A p.Gly115Ser), while Patient 2 (Pt 2) carried c.901dup p.Ile301AsnfsTer11, inherited from the mother, and c.343G>A p.Gly115Ser, transmitted by the father. Human dermal fibroblasts (HDFs) were reprogrammed using previously established methods [36] and pluripotency of the generated iPSC lines was verified by immunofluorescence using OCT4 and TRA-1-60 (Supplementary Fig. S3). Sanger sequencing (Supplementary Fig. S4), short tandem repeat (STR) analysis (Supplementary Fig. S5, Supplementary Table S2), and array comparative genomic hybridization (aCGH) (Supplementary Fig. S6A-D) excluded unexpected alterations during reprogramming.

Neural progenitor identity was confirmed by expression of nestin (marker of intermediate filament in neural progenitors), and neocortical identity by expression of the telencephalic transcription factor FOXG1 (Forkhead box G1) (Fig. 7). Quantification of telencephalic cells, expressed as FOXG1⁺/total nuclei ratio, showed no differences in Pt 1 or Pt 2 versus control, indicating normal corticopoietic capability of SPG56 lines (Fig. 7), a feature maintained at the mature neuron stage (Fig. 8). Indeed, cCASP3 staining at this stage did not reveal any increase in apoptotic rate (Fig. 8). Furthermore, phospho-histone H3 (PHH3) staining at Day30 (D30) in Pt 1 indicated a severe decrease in proliferation (Fig. 8). Also, immunofluorescence analysis of CTIP2 (BCL11B) and TBR1, markers of cortical layers V and VI, respectively, showed at D30 a pronounced increase in CTIP2⁺/nuclei in Pt 2 suggesting premature neurogenesis (Fig. 8).

Figure 7.

Representative images of iPSC-derived neural progenitors at D16 showing the expression of the neural progenitor marker nestin and the telencephalic transcription factor FOXG1. Scale bar 20 μm. Mean ± SD is shown. Ns = not significant.

Figure 8.

Representative images of apoptotic marker cCASP3 and neural marker TUBB3 (left), neocortical identity marker FOXG1, and proliferation marker phospho-histone H3 (PHH3) (middle), and V and VI cortical layer markers TBR1 and CTIP2 (right) in control, Pt 1 and Pt 2 at D30. Pt 1 and Pt 2 did not showed differences in terms of apoptotic rate or FOXG1 expression with respect to ctrl. On the other hand, Pt 1 displayed a marked decrease in proliferation (%PHH3⁺ cells), whereas Pt 2 exhibited an increase in CTIP2⁺ cells. Scale bar 20 μm. The Kruskal-Wallis test was used for statistical analysis. Mean ± SD is shown. Ns = not significant. ^**p<0.01, ^***p<0.001, ^****p<0.0001.

Transcriptomic profiling of cerebrocortical derivatives supports impaired neural network development as a key disease mechanism

To better identify altered gene expression in pathological conditions, we performed RNA-seq analysis on both patient and healthy control iPSC neural derivatives. In particular, we analyzed neural progenitors (D16), early neurons (D30), and late-born neurons (D80). The analysis, performed using the more stringent threshold |lfc|>1, identified 153 DEGs at D16 (117 upregulated and 36 downregulated, Supplementary Table S3), 241 DEGs at D30 (214 upregulated and 27 downregulated, Supplementary Table S4), and 129 DEGs at D80 (105 upregulated and 24 downregulated, Supplementary Table S5). Over-Representation Analysis (ORA) identified 77 DEGs common to the two patients at D16, 74 at D30, and 74 at D80 (Fig. 9A–C, respectively). The main altered pathways identified were Cell–cell adhesion via plasma membrane adhesion molecules, Homophilic cell adhesion via plasma membrane adhesion molecules, Dendrite self-avoidance, Axonogenesis, Neuron recognition, Axon guidance, and Neuron projection guidance; their dysregulation, affecting proper neural network development, thus emerges as a key molecular mechanism of the disease. The pathways found to be altered in SPG56 iPSC-derived neural cells included genes whose aberrant expression also occurred in zebrafish orthologs in most cases, although showing opposite trends of gene expression (Table 1). Among these genes, for qPCR validation we selected two (CHL1 and CNTN4) because of their role in neural adhesion and synapse organization, and SHH and PAX6 because of their relevant biological role in neurogenesis. The human genes CHL1, CNTN4, and PAX6 were strongly upregulated by qPCR in SPG56 neuroderivatives in contrast to zebrafish (Fig. 10A-C), whereas SHH was downregulated in human cells and not in zebrafish (Fig. 10D). Furthermore, we selected and successfully validated by qPCR genes involved in neurodevelopment and glutamatergic function (NEUROG2, HES1, GRIK2, SLC17A7, JAG1, GRM7; Supplementary Fig. S7) and found to be altered in SPG56 neuroderivatives transcriptomic analysis at different stages (Supplementary Tables S3, S4 and S5). In contrast to data observed in the cyp2u1^−/− zebrafish model, we did not find altered genes related to the arachidonic acid cycle, lipid metabolism, and CoQ biosynthesis in hiPSC-derived neural cells.

Figure 9.

GO enrichment analysis of common DEGs in iPSC-derived neural cells from the two SPG56 patients at D16 (A), D30 (B), and D80 (C) obtained using ORA analysis. The analysis indicated that altered pathways are maintained over time. padj<0.05, |lfc|>1.

Table 1.

DEGs identified by over-representation analysis in SPG56-iPSC neural derivatives at all time points, and the trend of their orthologues in zebrafish cyp2u1^−/−. Lfc = log₂ fold change. Ns = not significant. padj = adjusted p-value.

Gene (human)	D16 lfc	D30 lfc	D80 lfc	Gene (zebrafish)	Zebrafish lfc
CDH10	1196	0,811	ns	cdh10a	ns (-0,414; padj 0,078)
CDH12	ns	1987	3392	cdh12a	ns
CHL1	7422	13 123	7654	chl1a   chl1b	ns -0,905
CNTN4	7066	10 112	3357	cntn4	-0,756
CNTN6	7849	5511	ns (2820; padj 0,085)	No orthologue
DSCAM	6725	11 150	9397	dscama   dscamb	-0,750 -0,763
MRPS21	1067	0,981	0,892	mrps21	0,931
PAX6	ns	2258	3304	pax6a   pax6b	-0,493 -1165
PCDHA11	3592	2353	3057	No orthologue
PCDHA6	4115	3155	3462	No orthologue
PCDHA9	-1797	-1956	-2890	No orthologue
PTPRT	6418	8331	ns	ptprt	-0,797
PCDHB5	2441	ns (1224; padj 0,096)	ns	No orthologue
PCDHGB7	1162	0,749	1326	No orthologue
SHH	ns	-3010	ns	shha   shhb	ns ns
SLITRK2	7402	11 038	ns (3256; padj 0,068)	slitrk2	ns
SLITRK4	9107	11 661	6357	slitrk4	ns
TMEM132D	11 624	9691	8788	tmem132d	-0,58
WNK1	1106	0,810	0,714	wnk1a   wnk1b	ns -0,819

Figure 10.

(A) qPCR validation of CHL1, (B) CNTN4, (C) PAX6, and (D) SHH in SPG56 iPSC-derived neural cells and cyp2u1^−/−. T-test was used for statistical analysis. FC = fold change. Mean ± SEM is shown. ^*p<0.05, ^**p<0.01, ^***p<0.001, ^****p<0.0001.

Discussion

Generation of novel disease models allows exploration of transcriptomic signatures in the early stages of neurodevelopment. To further understand the early stages of neurodevelopment in SPG56, we generated and characterized both a zebrafish cyp2u1^−/− model and cerebrocortical cells differentiated from SPG56 patient-derived iPSCs. The knock-out model showed extremely low residual cyp2u1 mRNA expression, as well as locomotor and behavioral features reminiscent of the motor impairment seen in children with SPG56. In the absence of muscle damage, the zebrafish locomotor pathological phenotype is likely attributable to neural impairment.

RNA-seq analysis in the cyp2u1^−/− model pointed to cell cycle, neural processes, and oxidative phosphorylation as the main altered pathways. Accordingly, as already observed in SPG11 and SPG7 models [37–40], functional studies in these mutant embryos revealed increased apoptotic rate and decreased cell proliferation. Moreover, alteration of oxidative phosphorylation confirmed previous findings in SPG56 patient-derived cultured cells and Cyp2u1^−/− mice [1, 32]. Oxidative stress triggering apoptosis and neural defects is a common feature of spastic paraplegias [41] with impaired neurogenesis [42].

To further investigate whether these molecular alterations translate into functional defects at the network level, we performed local field potential recordings and whole-brain calcium imaging in cyp2u1^−/− larvae. Electrophysiological analysis revealed prolonged events frequency band in mutants, though with reduced power, a finding consistent with an abnormal temporal organization of neuronal activity. In parallel, calcium imaging analysis showed significantly increased skewness in mutants compared with controls, indicating a more asymmetric distribution of neuronal activity. Together, our results suggest the presence of altered brain network dynamics in the mutant condition and offer functional evidence to the hypothesis that early defects in neurogenesis and neuronal maturation are associated with impaired circuit activity. Consistently, the locomotor phenotype observed in cyp2u1^−/− larvae occurs in the absence of detectable abnormalities in spinal motor neurons or muscle. This supports the interpretation that the behavioral defect primarily reflects a dysfunction of the central motor apparatus, impaired brain network dynamics, and an abnormal temporal organization of neuronal activity.

Bulk RNA-seq in early models of SPG56 suggested impaired neurogenesis and dysregulated cell proliferation. Transcriptomic profiling of iPSC-derived cortical neurons revealed disturbances of multiple pathways associated with neural cell adhesion and synapse organization, suggesting that impaired neural network formation is a central feature of the disease. A substantial proportion of DEGs were cadherin and protocadherin genes, which are essential for neurite initiation, axon guidance, and synapse formation [43–45]. Their dysregulation, together with their known involvement in neural disorders [46], indicates that defective neuronal connectivity and synapse maturation may contribute to SPG56 pathogenesis. This signature was further supported by altered expression of CHL1, CNTN4, CNTN6, DSCAM, PTPRT, SLITRK2, and SLITRK4, all of which participate in axon guidance, synapse specification, and circuit refinement [47–51]. Notably, several disease-causing genes in HSP encode cell adhesion molecules [41], supporting the idea that cell adhesion dysfunction may represent a shared pathogenic axis across multiple HSP subtypes. The involvement of genes regulating neuronal morphogenesis, such as TMEM132D and WNK1 [52, 53], suggests that cytoskeletal remodeling and neurite outgrowth are also affected, extending the impact of the disease beyond axon guidance and synapse specification. Among the developmental regulators, SHH, one of the most important known morphogens with multiple roles in embryonic development [54], emerged as a particularly relevant node. Its consistent downregulation in human neurons and in the zebrafish ortholog points to possible impairments in neuronal patterning. Furthermore, PAX6, a critical transcription factor that governs cell fate in neural lineages and regulates eye development [55], was upregulated at later stages in human neural cultures but downregulated in its zebrafish orthologues (pax6a and pax6b). Observed differences between the two patients under investigation might be due to factors such as genetic background or other non-genetic circumstances. One potential explanation for the divergent expression patterns observed for most of the shared DEGs in human and zebrafish may lie in the different metabolic mechanisms related to CYP2U1 or species-specific mechanisms. cyp2u1^−/− exhibited extremely low residual mRNA expression whereas both patients carried a truncating mutation in one allele and the same p.Gly115Ser missense which was observed not to alter the protein level despite severely impacting the enzymatic activity [11]. Different biological contexts (in vitro vs. in vivo) or stages of development might be the reason for further explanations. While iPSC-derived neurons represent a neuron-enriched, gradually maturing in vitro model, 120 hpf zebrafish larvae are a whole-organism context in which the nervous system is already anatomically formed and functionally mature. Such differences in time and tissue can alter baseline expression trajectories, resulting in apparent reversals in direction at the single-gene level. Similar discrepancies have already been observed by others [56].

Also of note, transcriptomic analyses highlighted the dysregulation of several genes involved in eye determination, visual cycle, and retinol metabolism, in accordance with visual defects described in late disease stages in SPG56. These data imply that visual alterations occur earlier than expected but often do not produce clinical manifestations. Early monitoring and specialized follow-up of patients could prevent end-stage retinal impairment.

Reduced proliferation and premature neurogenesis may represent two sides of the same coin. Transcriptomic profiling of SPG56 patient-derived iPSC neural differentiation revealed a consistent dysregulation of neuronal development across all stages analyzed. Early differentiation stage (D16) was characterized by premature neurogenic commitment, evidenced by increased proneural transcription factor NEUROG2 expression and reduced progenitor-associated signaling genes such as HES1, GLI1 and PTCH2, together with early activation of neuronal adhesion, axon guidance, synapses, and glutamatergic receptor genes as GRIK2 and GRM7 (Supplementary Table S3). This phenotype became more pronounced at later stages (D30), with persistent upregulation of neurogenic, glutamatergic, and synaptic genes including SLC17A7, JAG1, GRIN2B, and RIMBP2 among others (Supplementary Table S4). Notably, the transcriptional landscape remained remarkably similar even at D80, showing sustained upregulation of proneural and neuronal wiring genes like NEUROG2, GRM7, and DSCAM (Supplementary Table S5). Rather than simply reflecting accelerated maturation, these findings suggest the establishment of a stable intermediate neuronal state marked by persistent developmental signaling and incomplete neuronal maturation. Overall, SPG56 neural cells appear to undergo premature neuronalization, while failing to achieve a fully mature and stabilized neuronal identity.

Our hypothesis that SPG56 pathophysiology involves early defects in neurodevelopmental and connectivity pathways is further supported by independent proteomic evidence derived from studies in Cyp2u1^−/− mice. The overlap between transcriptomic and proteomic datasets across species provides compelling evidence that these neurodevelopmental pathways represent core, evolutionarily conserved components of SPG56 pathophysiology. Indeed, several differentially expressed proteins identified in the hippocampi of Cyp2u1^−/− mice [32] correlate well with DEGs found in our cyp2u1^−/− zebrafish, including epha4, plp1a, rac1b, kcna1b, dnajc11b, rpl24, gabra5, and gnaz. In the human brain, these molecules participate in axon guidance (EPHA4) [57], myelin structure (PLP1) [58], cytoskeletal regulation (RAC1) [59], neuronal excitability (KCNA1, GABRA5) [60, 61], mitochondrial function (DNAJC11) [62], ribosomal activity (RPL24) [63], and GPCR signaling (GNAZ) [64]. Their recurrent dysregulation in both proteomic studies in mice and zebrafish RNA-seq analyses points to highly conserved biological pathways affected by CYP2U1 loss of function. Such cross-species concordance further strengthens the hypothesis that pathological mechanisms in SPG56 coordinate disruptions in cell adhesion, morphogen signaling, neuronal morphogenesis, and synaptic specification, which likely converge to impair the formation and maturation of cortical circuits. However, it is important to note that the two datasets were generated using different methodologies and tissues. Our dataset was generated using RNA-seq analysis of whole larvae at 120 hpf, whereas the mouse dataset derived from proteomics analysis of two-month-old hippocampi [32]. SPG56 remains an incurable condition, although there is evidence of folinic acid use mitigating disease progression, perhaps through metabolic rewiring, in Cyp2U1^−/−mice and in a family [21], as well as growing enthusiasm and expectation surrounding the development of precise gene therapy approaches. Unlike KO mice and two SPG56 siblings [21], our zebrafish KO model did not respond to either 5-MTHF or folinic acid. 5-MTHF (also known as levomefolic acid) is an active form of folic acid and the most efficient means of folate supplementation [34]. This finding in zebrafish might be explained by the downregulation of mthfr in our cyp2u1^−/− model, as folinic acid (also known as leucovorin) requires MTHFR to be converted to 5-MTHF [35]. In addition, we could not treat adult zebrafish with 5-MTHF or folinic acid before mating because of current regulatory restrictions, whereas pregnant KO mice had ad libitum access to folate in water [32]. A further aspect to consider is that different functional domains are assessed in KO mice and in our zebrafish model, namely cognitive and motor functions, respectively, making a direct comparison difficult. Indeed, folate restored cognitive defects in the KO mouse model, but locomotor features were not tested as that model did not exhibit locomotor impairment [32]. Conversely, complex behavioral features in zebrafish can only be reliably assessed at the adult stage, yet this could not be tested in our study due to ethical restrictions.

Thus—and this is a clear limitation of our study- we cannot exclude the possibility that cyp2u1^−/− might benefit from 5-MTHF or folinic acid treatment at later stages, or that mutant larvae might benefit when pregnant females are treated before mating. Moreover, an additional constraint of our results due again to regulatory restrictions is that RNA-seq on whole larvae at 120 hpf inevitably dilutes neuron-specific signals and complicates comparison with the iPSC-derived neuron dataset. Indeed, we used a threshold of |lfc|>1 for iPSCs RNA-seq analysis to set a stringent threshold that we relaxed at |lfc|>0.6 to capture even low differences in gene expression in a more dilute and diverse samples as zebrafish 120 whole larvae.

A further limitation was the low number of patients involved, which made it difficult to interpret differences in the results and may have undermined the accuracy of our analyses. Furthermore, another important constraint is the use of a single clone for each patient, which prevents us from excluding that certain differences in the patients’ cell phenotype may be due to clonal effects. Yet, SPG56 is a highly heterogeneous disease, allowing us to speculate that other genetic factors (already described in this disease) [20] or non-genetic factors likely contribute to inter-individual differences. However, despite results achieved in this study, the link between CYP2U1 function and the neurodevelopmental changes remains mechanistically unresolved, and our findings may reflect downstream consequences of CYP2U1 loss of function.

In conclusion, combined RNA-seq analysis in both cyp2u1^−/− and SPG56 iPSC-derived cortical neurons pointed to impaired neural network development as a main mechanism of the disease. This knowledge could inform early detection and intervention and suggests that perinatal gene therapy may be the most effective strategy to limit disease consequences.

Materials and methods

Zebrafish husbandry

Experiments were carried out using the wild-type AB strain. Adults were housed in tanks at a density of no more than five zebrafish per liter at a constant temperature of 28°C on a 14-h light/10-h dark cycle. Zebrafish eggs and embryos were collected and raised at 28.5°C in E3 medium using established procedures and staged in hours post fertilization (hpf) or days post fertilization. The generation of the CRISPR/cas9 mutant was carried out with the ethical approval (document n° 338/2020-PR) of the Italian Minister of Health, in accordance with the European Union (EU) Directive 2010/63/EU on the protection of animals used for scientific purposes, under the supervision of the Institutional Animal Care and Use Committee of the University of Pisa, and in compliance with the 3R principles [65].

Generation of the cyp2u1^−/− line

The sgRNA was designed—using CHOPCHOP (https://chopchop.cbu.uib.no/) set with NGG PAM sites and zero predicted off targets—against exon 2 of the cyp2u1 transcript (ENSDART00000048281.8) and transcribed using the Megascript T7 Transcription Kit (Thermo Fisher Scientific, Waltham, MA). cyp2u1-sgRNA and TrueCut™ Cas9 Protein v2 (Thermo Fisher Scientific) were co-injected into 1-cell stage embryos. At least three independent injection experiments were performed with spawns from different founder fish to control for batch effect. For mutation screening, sgRNA-injected F0 embryos were raised to adulthood and out-crossed with wild-type adults to obtain F1 heterozygous embryos. To identify adults potentially carrying mutations, genomic DNA from 30 randomly selected F1 embryos was sequenced (primers are listed in Supplementary Table S6). F1 heterozygous fish carrying the selected mutations at the target site were selected and inter-crossed to generate the F2 homozygous cyp2u1^−/− line. All subsequent functional studies were performed in the F3 generation.

cyp2u1^−/− gene expression analysis

Total RNA was extracted from 30 embryos at 120 hpf using the Quick-RNA Miniprep Kit (Zymo Research, Irvine, CA) according to the manufacturer’s instructions. cDNA was synthesized using the PrimeScript RT Reagent Kit (Takara Bio Inc., Kusatsu, Shiga, Japan), and gene expression analysis was performed by qPCR as described elsewhere [39]. Relative expression levels of each gene were normalized to β-actin expression and calculated using the 2^-∆∆Ct method (primers are listed in Supplementary Table S6).

Analysis of larval morphology

Live zebrafish were mounted on glass depression slides with 3% agarose. Images were obtained using a Leica M205FA stereomicroscope (Leica Microsystem, Wetzlar, Germany). The body length and eye size of 120hpf larvae were measured using ImageJ 64 software.

Locomotor behavior

Coiling behavior was measured in 30hpf embryos using Danioscope software (Noldus, Wageningen, The Netherlands). Locomotion was also analyzed in 120hpf larvae using the DanioVision® device (Noldus) and analyzed using EthoVision XT® software (Noldus) as previously described [66].

Birefringence assay

Muscle birefringence was analyzed by placing anesthetized embryos on a glass polarizing filter, covering them with a second polarizing filter, and recording them under a Leica M205FA microscope as described elsewhere [67].

RNA-seq analysis in zebrafish

Transcriptome analysis was performed in 8 zebrafish samples: 4 controls and 4 cyp2u1^−/− specimens. RNA was extracted using the Quick-RNA Miniprep Kit (Zymo Research).

RNA integrity was assessed using the RNA 6000 Nano Kit on a Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Illumina RNA-seq libraries were generated using the TruSeq stranded mRNA ligation kit (Illumina, San Diego, CA, USA) after poly(A) capture and according to manufacturer’s instructions. Quality and size distribution of RNAseq libraries were assessed by capillary electrophoretic analysis with the Agilent 4150 Tape station (Agilent Technologies). Libraries were quantified by real-time PCR against a standard curve with the KAPA Library Quantification Kit (KapaBiosystems, Wilmington, MA, USA). RNAseq libraries were pooled at equimolar concentration and sequenced with Illumina technology in 150PE mode on a Novaseq6000 sequencer (Illumina).

Quality of reads was assessed using FastQC software (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Raw reads were trimmed with fastp (v.0.21.0) and aligned to the zebrafish reference genome assembly (GRCz11) using STAR aligner (v2.7.9a) [68]. Reads distribution on CDS, intronic and intergenic was computed using reads_distribution.py script from RSeQC suite (v4.0.0) [69]. Gene expression quantification has been performed using RSEM and zebrafish GRCz11 annotation. Genes-level abundance, estimated counts and gene length obtained with RSEM [70], were summarized into a matrix using the R package tximport and subsequently the differential expression analysis was performed with DESeq2 [71]. To generate more accurate log2 foldchange estimates for low expressed genes, the shrinkage of the Log2 FoldChange was performed applying the apeglm method [72]. Gene Ontology (GO) enrichment analysis of DEGs with padj<0.05 and |lfc|>0.6 has been performed using ClusterProfiler (v.4.10.1). The enriched biological processes and KEGG pathways obtained from Over-Representation Analysis were graphically represented using the barplot function of the ClusterProfiler R package [73].

Electrophysiological recordings and calcium imaging in zebrafish

Local field potential (LFP) recordings were performed in 120 hpf zebrafish larvae to characterize the epileptic phenotype. Larvae were embedded in 1.2% low melting point agarose, and LFPs were recorded using an AgCl electrode inserted into a glass micropipette backfilled with 2 M NaCl. Electrophysiological signals were amplified 500-fold (EXT-02F, NPI, Tamm, Germany), band-pass filtered (0.3–1300 Hz), and digitized at 5 kHz using an Axon Digidata 1550B acquisition system controlled by Clampex 10.7.0.3 software (Molecular Devices, Berkeley, CA). LFP data were analyzed using custom MATLAB scripts as previously described [74]. For calcium imaging experiments, 120 hpf zebrafish larvae were similarly embedded in low melting point agarose and immobilized for imaging on a Nikon FN1 microscope (Nikon, Tokyo, Japan). Video recordings were acquired with a Prime sCMOS camera (Teledyne Photometrics, Tucson, AZ) controlled by Metafluor software (Molecular Devices, San Jose, CA), using a time-lapse interval of 150 ms and collecting 3860 frames per recording. Fluorescence fluctuations (ΔF/F0) within predefined brain regions of interest (ROIs) were analyzed using ImageJ 64 software, normalized to background fluorescence, and quantified through Pearson’s coefficient of skewness, as previously described [74].

Immunofluorescence analysis in zebrafish

To prevent the development of pigmentation, embryos were treated with 0.005% phenylthiourea from 24 hpf. Whole-mount immunohistochemistry was performed as described elsewhere [39]. The primary antibodies used were rabbit anti-PHH3 1:500 (PA5–85506, Thermo Fisher Scientific) and mouse anti-acetylated-tubulin 1:500 (018M4788V, Thermo Fisher Scientific). Apoptotic cells from 24 hpf larvae were detected by staining with acridine orange (Merck, Darmstadt, Germany). Zebrafish embryos were incubated with 10 μg/mL acridine orange solution for 15 min in the dark, and the larvae were then washed three times with E3 medium. At 48 hpf, we counted acridine orange-positive cells within a pre-defined area and a quantitative analysis was performed as previously described [39].

Mitochondrial respiratory analysis

Mitochondrial respiration was analyzed in control and cyp2u1^−/− larvae at 120 hpf using the XF24 extracellular flux analyzer (Agilent Technologies, Milan, Italy) [39]. Basal respiratory capacity, ATP production, maximal respiration rate, and spare respiratory capacity were measured [75].

Pharmacological treatments

Control and cyp2u1^−/− embryos were selected at 2 hpf and randomly placed in 60 mm × 15 mm petri dishes at a density of 50 per dish, with each dish containing either a drug or egg water as control. Each compound was used as already described in the literature: folinic acid 20 μM (MedChemExpress, Monmouth Junction, NJ) [76], 5-MTHF 30 μM (MedChemExpress) [77].

Patients

This study (registered in ClinicalTrial.gov, code NCT06572046) was approved by the Tuscany Regional Pediatric Ethics Committee. Two patients with a clinical and genetic diagnosis of SPG56 (already described elsewhere [11, 32]) were recruited from the cohort of patients affected by spastic paraplegia attending the research hospital IRCCS Fondazione Stella Maris (Pisa, Italy) and included in the Italian STOP-HSP.net registry.

Human dermal fibroblasts (HDFs) were isolated from skin biopsies of Pt 1 and Pt 2 and a healthy age- and sex-matched control. Cell lines were grown at 37°C with 5% CO₂ in DMEM High Glucose, supplemented with 10% fetal bovine serum, and 1% antibiotics/antimycotics (Euroclone, Milan, Italy). In accordance with the Declaration of Helsinki, all study participants gave their written informed consent to skin biopsy.

Generation of iPSCs

iPSCs were generated from HDFs using microfluidic RNA technology following previously established methods [36]. iPSCs were then cultured at 37°C and 5% CO₂ on Geltrex-coated six-well-plates in Stemflex medium (Thermo Fisher Scientific). Pluripotency was assessed by pluripotency marker expression. A single clone was used for each patient in downstream analyses. Genomic DNA from iPSCs and their corresponding HDFs were extracted using the MagPurix Blood DNA Extraction Kit 200, designed for MagPurix DNA Extract (Zinexts, Zhonghe, Taiwan), and mutations were confirmed by Sanger sequencing.

Genomic analysis by aCGH and STR analysis

aCGH was performed on genomic DNA extracted from Pt 1 and Pt 2 HDFs and iPSCs using the SurePrint G3 Human CGH Microarray (Agilent Technologies). The genomic identity between iPSC clones and parental HDFs was confirmed by STR analysis using DNA typing with a PowerPlex 16 HS system (Promega, Madison, WI), as previously described [78].

iPSC differentiation into cerebrocortical neurons

Neural progenitors and mature neurons were generated from iPSCs through an optimized cerebrocortical differentiation protocol [78, 79] based on a dual SMAD inhibition strategy [80, 81].

Immunofluorescence analysis in iPSCs and cerebrocortical neurons

Immunofluorescence was performed as previously described [78]. The following antibodies were used: OCT4 1:500 (ab19857, Abcam, Cambridge, UK), TRA-1-60 1:1000 (4746, Cell signaling, Danvers, MA), FOXG1 1:500 (ab18259, Abcam), Nestin 1:500 (ab18102, Abcam), cCASP3 1:500 (9661, Cell signaling), TUBB3 1:1000 (ab7751, Abcam), PHH3 1:500 (9706, Cell signaling), TBR1 1:500 (ab31940, Abcam), and CTIP2 1:500 (ab18465, Abcam). All images were acquired with a Zeiss LSM 900 confocal microscope (Zeiss, Oberkochen, Germany) and processed with Fiji software.

RNA-seq analysis in iPSC-derived cerebrocortical neurons

Transcriptome analysis was performed using 27 neuronal samples: 3 control samples, 3 Pt 1 samples, and 3 Pt 2 samples each at D16, D30, and D80 stages. RNA was extracted using the RNeasy Plus Mini Kit (Qiagen, Venlo, The Netherlands). Library preparation, sequencing, and analysis were performed as described in the RNA-seq analysis in zebrafish section, using Homo sapiens GRCh38 as genome reference and |lfc|>1 for Gene Ontology enrichment analysis.

Gene expression analysis on iPSC-derived cerebrocortical neurons

cDNA was synthesized using the PrimeScript RT Reagent Kit (Takara Bio Inc.), and gene expression analysis was performed by qPCR as previously described [82]. Relative expression levels of each gene were normalized to GAPDH expression and calculated using the 2^-∆∆Ct method (primers are listed in Supplementary Table S6). To mitigate variability among controls, cDNA of an additional independent healthy age- and sex-matched control (generated and processed in parallel with the other cell lines) was used exclusively for the gene expression analysis of selected genes (Supplementary Fig. S7).

Statistical analysis

All data in the manuscript represent three or more independent experiments. Statistical analysis was performed using GraphPad software (Boston, MA). All data were analyzed using either parametric or non-parametric methods depending on data distribution. The figure legend reports the specific test and statistical significance applied for each analysis.

Artificial Intelligence use Statement

No generative AI tools were used in the preparation of this manuscript.

Author contributions

Daniele Galatolo (Conceptualization, Data curation, Formal analysis, Investigation, Writing—original draft, Writing—review & editing), Devid Damiani (Data curation, Formal analysis, Investigation, Writing—original draft), Valentina Naef (Data curation, Investigation, Methodology), Matteo Baggiani (Data curation, Formal analysis, Investigation, Methodology), Carla Cona (Data curation, Formal analysis), Giuseppe Neri (Data curation, Formal analysis, Methodology), Claudia Dell’amico (Data curation, Formal analysis, Investigation, Methodology), Ugo Borello (Formal analysis, Funding acquisition, Methodology), Marco Onorati (Formal analysis, Funding acquisition, Investigation, Methodology), Maria Marchese (Funding acquisition, Investigation, Methodology, Project administration, Writing—original draft, Writing—review & editing), and Filippo Maria Santorelli (Funding acquisition, Methodology, Project administration, Supervision, Writing—original draft, Writing—review & editing)

Conflicts of interest

The authors declare no competing interests.

Funding

Research in the F.M.S. clinic is supported in part by the Italian Ministry of Health (Ricerca Corrente 2024) and a project funded under the National Recovery and Resilience Plan PNRR M6C2I2.1 “Valorizzazione e potenziamento della ricerca biomedica del SSN" by the European Union – Next Generation EU - PNRR-MR1-2023-12377843 - CUP C53C24000250006. This work was also partially supported by Telethon (grant GSP23001) and an unconditional donation from CureSPG56, Italy.

M.O. laboratory received support by the European Union Next-Generation EU, National Recovery and Resilience Plan (NRRP), mission 4, component 2, investment 1.4 (“Tuscany Health Ecosystem”, CUP N. B83C22003930001), and investment 1.1, call PRIN 2022 PNRR (D.D. 1409 14-09-2022 – Uncovering spatial and temporal sequence of neurodegeneration in a 3D model of corticospinal-neuromuscular unit of Amyotrophic Lateral Sclerosis – Prot. P2022MSFT3 CUP I53D23006630001). This manuscript reflects only the authors’ views and opinions, neither the European Union nor the European Commission can be considered responsible for them.