Homozygous Pex2^−/− mice were obtained by breeding Pex2 heterozygote mice on a hybrid Swiss Webster-129 (SW/129) background (Faust et al., 2001). Pex2 mice and BALB/c mice had ad libitum access to food and water and were exposed to a 12:12 h light–dark cycle. As their biochemical characteristics were comparable, the control mice consisted of either Pex2^+/+ or Pex2^+/− genotypes (hereafter referred to as Pex2^+/) for the purpose of this study (Kovacs et al., 2004, 2009). A RT-PCR-based method targeting Kdm5c and Kdm5d was used to determine the genetic sex in mouse cDNA samples (Bosso-Lefèvre et al., 2023). Sex was not considered a biological variable in 10-day-old (P10) pups, so both male and female animals were examined in our study. Bile acids (BAs) were administered by orogastric gavage as described previously (Keane et al., 2007). All animal experiments and protocols were approved by the Institutional Animal Care and Use Committee of San Diego State University and Columbia University.

Frozen tissues (1:10 w/v) were homogenized in RIPA buffer (20 mM Tris, pH 7.5; 150 mM NaCl; 1 mM EDTA; 1 mM EGTA; 1% NP-40; 1% sodium deoxycholate) containing the cOmplete protease inhibitor cocktail (Roche; #11697498001) and the PhosSTOP phosphatase inhibitor (Roche; #4906837001) using the Potter S homogenizer (Sartorius, Goettingen, Germany). Homogenates were incubated on ice for 30 min and centrifuged at 20,000 g for 20 min at 4 °C. Protein concentration was determined by the BCA method (Thermo Fisher Scientific; #23225). Equal amounts of proteins were subjected to SDS-polyacrylamide gel electrophoresis and transferred to Amersham Protran Supported 0.2 μm nitrocellulose (Cytvia; GE10600015). After blocking for 1 h in TBST [Tris-buffered saline with 0.05% Tween-20] containing 1% bovine serum albumin (BSA) (PAN Biotech; P06-1391500), membranes were probed with the indicated primary antibodies in TBST containing 1% BSA overnight at 4 °C. The membranes were incubated with horseradish peroxidase-conjugated secondary antibodies and developed using the Clarity Western ECL Substrate (BIO-RAD; Cat#1705061). Blots were either exposed to Kodak X-Omat LS film (Rochester, NY) and scanned using a densitometer (GE Healthcare) or visualized on the Fusion Solo S imaging system. Blots were analyzed semiquantitatively by densitometry using ImageJ 1.52 v (National Institutes of Health). Antibodies are listed in Supplementary Table S1.

Catalase (EC 1.11.1.6) activity was assayed as described previously (Walter et al., 2014). Briefly, protein lysates were diluted with TVBE buffer [0.01% (v/v) Triton X-100 (Sigma-Aldrich; #93426), 1 mM EDTA, 1 mM NaHCO₃, 0.1% (v/v) ethanol; pH 7.5]. Equal amounts of sample and 2% Triton X-100 were added to glass test tubes on ice and incubated for 1 min. The reagent control contained TVBE buffer and 2% Triton X-100. The reaction was started by adding 1 mL cold substrate solution (20 mM imidazole, 0.1% fatty acid-free BSA (PAN Biotec; P06-139450) and 3.1 mM H₂O₂ (Sigma-Aldrich; #216763) at pH 7.0) and the samples were incubated at 0 °C. The enzymatic reaction was stopped by adding 1 mL TiOSO₄ solution. The glass tubes were transferred to room temperature (RT) and incubated for 10 min for full color development. Absorbance was measured at 410 nm. Catalase activity was measured as follows: BU/mL = (1 + x mL Sample + x mL TX-100)/50 * 1/(incubation time in min) * 1/(sample volume in mL) * log (reagent absorbance/sample absorbance) * dilution factor.

The enzyme 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP, EC 3.1.4.37), a marker enzyme for myelin, was assayed as described previously (Kovacs et al., 2001). HMG-CoA reductase (HMGCR, EC 1.1.1.34), farnesyl pyrophosphate synthase (FDPS, EC 2.5.1.10), isopentenylpyrophosphate isomerase (IDI1, EC 5.3.3.2), and squalene synthase (farnesyl-diphosphate farnesyltransferase 1, FDFT1, EC 2.5.1.21) activities were assayed as described previously (Kovacs et al., 2004).

Brain sterols were measured as described previously (Kovacs et al., 2004). Briefly, the mice were euthanized, their tissues harvested and flash frozen in liquid nitrogen before being stored at −85 °C until ready for analysis. After saponification and petroleum ether extraction, the nonsaponifiable lipids were analyzed by reverse-phase radio-HPLC, and the specific activities of the radiolabeled products were determined.

Total RNA was isolated and Northern blot analysis was performed as previously described (Kovacs et al., 2004, 2009). Quantitative RT-PCR assays were performed as described previously (Charles et al., 2020). First-strand cDNA was synthesized with random hexamer primers using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems; #4368813). Quantitative RT-PCR was performed on a Roche LightCycler LC480 instrument in duplicate using 10 ng cDNA, 2 pmol forward and reverse primers, and the 2× KAPA SYBR FAST qPCR Mastermix (Roche; #KK4611). Thermal cycling was performed with a 5 min denaturation step at 95 °C, followed by 45 three-step cycles: 10 s at 95 °C, 10 s at 60 °C, and 10 s at 72 °C. Melt curve analysis was performed to confirm the specific amplification of a target gene and absence of primer dimers. Primer sequences are listed in Supplementary Table S2 or have been previously published (Charles et al., 2020). Expression levels were calculated using the 2 – ΔΔ C T method (Livak and Schmittgen, 2001). 18S rRNA or Cyclophilin were used as the invariant control.

The Cnp cDNA (GenBank accession number NM_009923) was prepared from total mouse brain RNA by a standard reverse transcriptase-PCR procedure using the following forward and reverse primers: 5′-TACCCGCAAAAGCCACACAT-3′ and 5′-TCAGCATCCATCGCCCTTTG-3′. The PCR-generated probe was subcloned into the pCRII-TOPO vector (Invitrogen) and confirmed by sequencing. IMAGE clones for proteolipid protein 1 (Plp1; ID 1617043) and myelin basic protein (Mbp; ID 3982852) were purchased from ATCC.

Indirect immunofluorescence on 2 μm paraffin sections was performed with the indicated antibodies (Supplementary Table S1) as described previously (Kovacs et al., 2012). Mice were cardiac-perfused with 4% paraformaldehyde-phosphate-buffered saline. The brain was post-fixed overnight in 4% paraformaldehyde-phosphate-buffered saline and processed for paraffin embedding. 2 μm thick paraffin-embedded brain sections were mounted on Superfrost Plus slides. For antigen retrieval of paraffin-embedded tissue, deparaffinized and rehydrated sections were digested with 0.01% trypsin (Sigma T7409; Type II-S from porcine pancreas) for 10 min at 37 °C followed by microwaving in 10 mM citrate buffer (pH 6.0) three times for 5 min at 800 watts. Nonspecific binding was blocked with 4% BSA and 0.05% Tween 20 in PBS for 2 h and sections were then incubated overnight at 4 °C with primary antibodies. Secondary antibodies were applied for 2 h. Negative control sections were incubated in parallel by omitting the primary antibody. Images were taken with a Leica SP2-AOBS confocal laser scanning microscope. Fluorescent dyes were imaged sequentially in frame interlace mode to eliminate crosstalk between the channels.

Tissue samples were transferred to 2 mL Eppendorf tubes and 200 μL lysis buffer [4% sodium dodecyl sulfate (SDS) in 50 mM triethylammoniumbicarbonat (TEAB), pH 8.2] was added per sample. Protein extraction was carried out using a tissue homogenizer (TissueLyser II, QIAGEN) by applying 2× 2 min cycles at 30 Hz. Samples were treated with High Intensity Focused Ultrasound for 1 min at an ultrasonic amplitude of 90% before boiling at 95 °C for 10 min while shaking at 800 rpm on a Thermoshaker (Eppendorf). Protein concentration was determined using the Lunatic UV/Vis polychromatic spectrophotometer (Unchained Labs) at a 1:10 dilution.

For each tissue sample, 50 μg of protein was taken and reduced with 5 mM TCEP [tris(2-carboxyethyl)phosphine] and alkylated with 15 mM chloroacetamide at 30 °C for 30 min. Samples were processed using single-pot solid phase enhanced sample preparation (SP3). The SP3 protein purification, digestion and peptide clean-up were performed using a KingFisher Flex System (Thermo Fisher Scientific) and carboxylate-modified magnetic particles (GE Life Sciences; GE65152105050250, GE45152105050250) (Hughes et al., 2014; Leutert et al., 2019). Beads were conditioned following the manufacturer’s instructions, consisting of 3 washes with water at a concentration of 1 μg/μl. Samples were diluted with 100% ethanol to a final concentration of 60% ethanol. Beads, wash solutions and samples were loaded into 96 deep well- or micro-plates and transferred to the KingFisher. The following steps were performed on the robot: collection of beads from the last wash, protein binding to the beads, washing of the beads in wash solutions 1–3 (80% ethanol), protein digestion (overnight at 37 °C with a trypsin:protein ratio of 1:50 in 50 mM TEAB), and peptide elution from the magnetic beads using MilliQ water. The digest solution and water elution were combined, dried to completeness, and re-solubilized in 20 μL of MS sample buffer (3% acetonitrile, 0.1% formic acid). The peptide concentration was determined using the Lunatic UV/Vis polychromatic spectrophotometer.

LC-MS/MS analysis was performed on an Orbitrap Fusion Lumos (Thermo Scientific) equipped with a Digital PicoView source (New Objective) and coupled to an M-Class UPLC (Waters). The solvent composition of the two channels was 0.1% formic acid for channel A and 99.9% acetonitrile in 0.1% formic acid for channel B. The column temperature was 50 °C. For each sample, 300 ng of peptides were loaded onto a commercial ACQUITY UPLC M-Class Symmetry C18 Trap Column (100 Å, 5 μm, 180 μm × 20 mm, Waters) connected to an ACQUITY UPLC M-Class HSS T3 Column (100 Å, 1.8 μm, 75 μm × 250 mm, Waters). Peptides were eluted at a flow rate of 300 nL/min. After an initial hold at 5% B for 3 min, a gradient of 5 to 22% B was applied for 80 min and 22 to 32% B for an additional 10 min. The column was cleaned at the end of the run by increasing to 95% B and holding at 95% B for 10 min before restoring the loading condition. Samples were measured in random order. The mass spectrometer was operated in data-dependent mode (DDA) with a maximum cycle time of 3 s, funnel RF level at 40%, heated capillary temperature at 275 °C, and advanced peak determination (APD) on. Full-scan MS spectra (300–1,500 m/z) were acquired at a resolution of 120,000 at 200 m/z after accumulation to an automated gain control (AGC) target of 500,000 or for a maximum injection time of 40 ms. Precursors with an intensity above 5,000 were selected for MS/MS. Ions were isolated using a quadrupole mass filter with a 0.8 m/z isolation window and fragmented by higher-energy collisional dissociation (HCD) using a normalized collision energy of 35%. Fragments were detected in the linear ion trap with the scan rate set to rapid, the automatic gain control set to 10,000 ions, and the maximum injection time set to 50 ms. Charge state screening was enabled, and singly, unassigned charge states and charge states greater than seven were excluded. Precursor masses previously selected for MS/MS measurement were excluded from further selection for 20 s with a mass tolerance of 10 ppm. Samples were acquired using internal lock mass calibration at m/z 371.1012 and 445.1200.

The mass spectrometry proteomics data were processed using the local laboratory information management system (LIMS) (Türker et al., 2010).

The acquired MS raw data were processed for identification and quantification using FragPipe (version 16.0), MSFragger (version 3.4), IonQuant (version 1.7.17), and Philosopher (version 4.1.1) (Yu et al., 2020). Spectra were searched against a Uniprot Mus musculus reference proteome (taxonomy 10090, canonical version from 2022-05), concatenated to its reversed decoyed fasta database and common protein contaminants. Strict trypsin digestion with a maximum of 2 missed cleavages was set for the closed search settings. Carbamidomethylation of cysteine was set as fixed modification, while methionine oxidation and N-terminal protein acetylation were set as variable. Label-free quantification and match between run option were enabled.

The R package prolfqua (Wolski et al., 2023) was used to analyze the differential expression and to determine group differences, confidence intervals, and false discovery rates for all quantifiable proteins. We started with the combined_protein.tsv file generated by FragPipe, which reports the protein abundances for each raw file. We then employed the Tukeys-Median Polish to estimate protein abundances. Furthermore, before fitting the linear models, we transformed the protein abundances using the variance stabilizing normalization (Huber et al., 2002).

For unsupervised multivariate statistical analysis, principal component analysis and hierarchical clustering was performed using Metaboanalyst 6.0 (Pang et al., 2024). Normalized protein abundance matrices of proteomic data derived from spinal cord, brainstem, cerebellum, and cerebral cortex samples from P10 control and Pex2^−/− mice were filtered for protein species occurring in all tissues, followed by left-censored missing data estimation (1/5th of the minimal positive value) and scaling (mean-centered and divided by the standard deviation of each variable). For hierarchical clustering, the Euclidean distance measurement and Ward clustering algorithm were applied.

All relevant data have been deposited to the ProteomeXchange Consortium via the PRIDE¹ (Deutsch et al., 2023) partner repository with the dataset identifier PXD063585.

Over-representation analysis after proteomics was performed with Enrichr (RRID:SCR_001575) using the Gene ontology, KEGG and reactome databases (Ashburner et al., 2000; Xie et al., 2021; Milacic et al., 2024; Kanehisa et al., 2025). All proteins with an adjusted p-value <0.1 were included. No fold change threshold was set, as some of the proteins may show moderate but biologically relevant changes.

Gene set enrichment analysis (GSEA) was performed using GSEA_4.3.2 software version (RRID:SCR_003199) (Subramanian et al., 2005). The software and all analyzed gene sets were downloaded from the official GSEA website (BROAD Institute).

Data are presented as mean ± SD. Statistical analyses were performed with GraphPad Prism version 9.0. When two groups were compared, statistical significance was assessed by unpaired, two-tailed Student’s t-test or unpaired, two-tailed Student’s t-test with Welch’s correction when variances were significantly different for two-group analyses. To correct for multiple comparisons, the Benjamini–Hochberg procedure was applied in R to control the false discovery rate at 0.05 and calculate the adjusted p-values for all evaluated genes and tissues. For multiple group analyses, one-way ANOVA followed by Tukey’s multiple comparisons test or two-way ANOVA Tukey’s multiple comparisons test was used. Illustrations were created with Adobe Illustrator (v.27.2) and BioRender.

Catalase is the most abundant peroxisomal matrix protein and in the CNS the activity is highest during the first 3 weeks of development and then declines sharply (Kovacs et al., 2001; Ahlemeyer et al., 2007), supporting the hypothesis that the high peroxisomal activity in the CNS during the first weeks of postnatal life may be related to lipid synthesis associated with rapid myelin formation and the production of plasma membranes for growing neurons. First, we divided the CNS of wild-type BALB/c mice into five regions, namely spinal cord (SC), brain stem (BS), cerebellum (CBL), diencephalon (DE), and cerebral cortex (CBR), and measured the activity of catalase at different developmental stages [from 7-days-old (P7) to P56] (Supplementary Figure S1A). Catalase activity was highest at P7 and P14 and decreased significantly thereafter (Supplementary Figure S1A). The activity of the cholesterol biosynthetic enzyme FDPS was also highest at P7 and P14 and decreased significantly thereafter (Supplementary Figure S1B). Likewise, the CNS of P10 SW/129 control and Pex2 knockout mice was divided into the SC, BS, CBL, DE, and CBR. In P10 control mice catalase activity was significantly higher in CNS regions containing a higher proportion of white matter (i.e., SC, BS) as compared to gray matter (i.e., DE, CBR) (Figure 1A). Catalase activity was significantly increased by ~44–133% in all examined regions of the CNS of Pex2^−/− mice at P10 (Figure 1A), with the highest increase in the DE and CBR (~2- and 2.3-fold, respectively). This finding suggests that peroxisome deficiency induces an oxidative imbalance that could activate stress pathways and lead to metabolic dysfunction.

To determine the impact of peroxisome deficiency on the CNS proteome, we characterized the proteome of the SC, BS, CBL, and CBR of P10 control and Pex2^−/− mice. Principal component analysis (PCA) (Figure 1B) and hierarchical clustering analysis (Figure 1C) revealed a clear separation between the proteomes of the different CNS regions. Furthermore, the hierarchical clustering analysis showed that the proteomes of each CNS region could be clearly separated based on genotype (Figure 1C). We detected 4,236, 4,243, 4,241, and 4,356 proteins in the SC, BS, CBL, and CBR, respectively (Supplementary Table S3). Among the detected proteins, 705, 467, 361, and 447 were significantly altered in SC, BS, CBL, and CBR of Pex2^−/− mice when the log₂ fold change (FC) and p-value cut-offs were set at ±0.5 and 0.05, respectively (Figure 1D).

We have compiled a list of bona fide and putative peroxisomal proteins based on published studies and compared them with the proteins detected in our proteomics analysis (Kikuchi et al., 2004; Islinger et al., 2007; Mi et al., 2007; Wiese et al., 2007; Gronemeyer et al., 2013). We detected ~60 peroxisomal proteins in the CNS. As expected and confirming the defect in peroxisome biogenesis, bona fide peroxisomal matrix proteins involved in peroxisomal fatty acid β-oxidation (i.e., ACOX1, HSD17B4, SCP2), etherlipid synthesis (i.e., AGPS), and VLCFA transport (i.e., ABCD3) were strongly downregulated in Pex2^−/− mice (Figure 1E). This is because mislocalized matrix proteins in the cytosol are often unstable and targeted for degradation via the proteasome. Surprisingly, protein levels of the peroxisomal matrix protein phytanoyl-CoA 2-hydroxylase (PHYH), which is involved in peroxisomal α-oxidation and converts phytanoyl-CoA to 2-hydroxyphytanoyl-CoA, were highly increased in all CNS regions examined in Pex2^−/− mice (Figure 1E). Several studies have shown that patients with ZSD have reduced levels of PHYH and increased levels of phytanic acid in their plasma and tissues (Aubourg et al., 1985; Schutgens et al., 1987; Molzer et al., 1989; Jansen et al., 1996; Verhoeven et al., 1997; Verhoeven and Jakobs, 2001). Little is known about how PHYH protein levels are regulated, but it is thought that phytanic acid or its precursor, phytol, could lead to increased levels of PHYH. A study found that PHYH activity is induced by phytanic acid, but this induction was not regulated via PPARα or RXR (Zomer et al., 2000). We determined the mRNA expression of Phyh and found no differences between P10 control and Pex2^−/− mice in any of the regions of the CNS examined (Supplementary Figure S1C), suggesting that PHYH protein levels in the CNS are controlled post-translationally. Peroxins (PEX), which are involved in the biogenesis of peroxisomes, showed a mixed picture, as the protein levels of some peroxins were reduced or increased. For example, the protein levels of PEX14 were increased in Pex2^−/− mice (Figure 1E). Peroxins can localize to peroxisomal membrane ghosts in Pex2^−/− mice or cells (Kovacs et al., 2012; Charles et al., 2020), and PEX14 has been reported to localize to mitochondria in PEX3-deficient mammalian cells (Muntau et al., 2000; South et al., 2000; Sugiura et al., 2017; Jansen et al., 2024). The protein levels of PEX5, which is predominantly cytoplasmic and required for the targeting of peroxisomal matrix proteins to peroxisomes, were significantly increased in all CNS regions of Pex2^−/− mice except for the brainstem.

Next, we performed an overrepresentation analysis (ORA) to investigate which specific biological processes were enriched in differentially expressed proteins in the CNS of control versus Pex2^−/− mice (Figure 2; Supplementary Figure S2 and Supplementary Table S4). As expected, ORA confirmed that peroxisome-associated proteins were significantly altered in Pex2^−/− mice. Importantly, an overrepresentation of processes involved in neurodegenerative diseases (Parkinson’s disease, Alzheimer disease, Amyotrophic lateral sclerosis, prion disease, pathways of neurodegeneration, Huntington disease, Spinocerebellar ataxia), where neuronal death and dysfunction are central themes, was observed in the top pathways in Pex2^−/− mice. The overrepresentation of neurodegenerative disease pathways confirms studies showing severe neurological dysfunction and abnormal neuronal migration in the developing cerebral cortex and cerebellum of Pex2^−/− mice (Faust and Hatten, 1997; Faust et al., 2001, 2005; Faust, 2003). It also suggests that peroxisome dysfunction contributes to the pathophysiology of the above-mentioned neurodegenerative diseases. GSEA of the proteomics data showed that the SC and BS of Pex2^−/− mice had a significant positive enrichment score for “Neuronal system,” whereas the CBR and CBL had a negative enrichment score (Supplementary Figure S3A and Supplementary Table S5). The significantly altered proteins under this term indicate changes in neurotransmission, neuronal excitability, synaptic plasticity, glutamate metabolism, and response to oxidative stress (Supplementary Table S5). Pathways of fatty acid degradation and biosynthesis, such as fatty acid biosynthesis, PPAR signaling, and unsaturated fatty acid biosynthesis, were overrepresented in Pex2^−/− mice, suggesting disruptions in cellular energy regulation and membrane composition that could affect neuronal and oligodendrocyte health.

Interestingly, ORA analysis revealed top terms related to mRNA processing (i.e., mRNA splicing and processing, RNA splicing, mRNA splicing via spliceosome, RNA splicing via transesterification reactions, regulation of mRNA splicing and mRNA processing, regulation of alternative splicing via spliceosome) (Figure 2; Supplementary Figure S2). These terms were particularly prevalent in the CBR. GSEA showed that the top GO terms that were most positively enriched for biological processes in the CBR and CBL of Pex2^−/− mice were related to RNA splicing and processing, as well as mRNA processing, metabolism, and splicing, and the regulation of RNA and mRNA splicing and processing (Supplementary Figure S3B and Supplementary Table S5). In addition, GSEA using gene sets from the Reactome database revealed significant positive enrichment for “Metabolism of RNA”, “mRNA processing and splicing”, and “Processing of capped intron-containing pre-mRNA”, particularly in the CBR and CBL of Pex2^−/− mice (Supplementary Figure S3B). The significantly altered proteins under these terms suggest enhanced RNA processing and splicing activity, enhanced RNA transport and stability, and increased transcriptional activity (Supplementary Table S5).

The top enriched GO term for molecular function was cadherin binding (Supplementary Figure S2). Cadherins are involved in regulating the interactions between cells and the extracellular matrix, which are essential for maintaining proper tissue architecture and cellular communication. More proteins are downregulated than upregulated under this term in the SC, BS, CBL, and CBR, indicating disrupted cytoskeletal dynamics and cell adhesion, altered vesicular trafficking, and impaired stress response (Supplementary Table S6). Cadherins are a class of cell adhesion molecules that play a critical role in neuronal development, synapse formation, and myelination (Payne et al., 1996; Schnädelbach et al., 2000; Laursen and Ffrench-Constant, 2007; Gärtner et al., 2015; Szabó et al., 2015; Chen et al., 2017; Martinez-Garay, 2020). Cadherins are involved in the communication between axons and oligodendrocytes during myelination, ensuring proper axonal growth and myelin formation. Cadherins are also involved in neuronal migration during development to facilitate the proper positioning of neurons. So far, no direct link between dysregulation of cadherins and peroxisomal disorders has been shown.

We previously showed that the rate of cholesterol synthesis in the liver of P9 Pex2^−/− mice was 13-fold higher than in controls, and bile acid (BA) feeding reduced hepatic cholesterol synthesis in Pex2^−/− mice to control levels (Kovacs et al., 2004, 2009). Interestingly, the rate of cholesterol synthesis in the brain of P9 Pex2^−/− mice was significantly decreased ~2-fold compared to controls, and BA feeding increased cholesterol synthesis in Pex2^−/− brain to control levels (Kovacs et al., 2004, 2009). Cholesterol and desmosterol were the only sterols detected in significant amounts in the brains of P8-10 control and Pex2^−/− mice. While cholesterol levels were similar in the brains of control and Pex2^−/− mice, desmosterol levels were significantly elevated in the brains of both untreated and BA-fed Pex2^−/− mice compared to their control littermates (Figure 3A). While BA feeding had no effect on desmosterol levels in control and Pex2^−/− mice compared to untreated mice, total cholesterol levels decreased (Figure 3A).

Since the brain depends on intracerebral de novo synthesis of cholesterol, the rate of cholesterol synthesis is highest during the early phase of postnatal brain development. As SW/129 Pex2^−/− can survive for about 1–2 weeks, we investigated the effects of peroxisome deficiency on the activities of cholesterol biosynthetic enzymes in the CNS of P10 control and Pex2^−/− mice. We determined the activities of HMGCR, IDI1, FDPS and FDFT1 as a measure of cholesterol perturbation (Figures 3B,C). All enzyme activities were significantly increased between 20 and 72% in the cerebral cortex of Pex2^−/− mice. Except for HMGCR the activities were also significantly increased by 40–60% in the diencephalon of Pex2^−/− mice. In contrast, the activities of all four enzymes were significantly decreased by ~11–42% in the spinal cord of Pex2^−/− mice. HMGCR and IDI1 activities were also significantly decreased by ~21% in the brainstem of the knockout mice, whereas FDFT1 was significantly increased by 32% and FDPS was unaltered. The activities of IDI1, FDPS and FDFT1 were slightly increased in the cerebellum of Pex2^−/− mice, while the activity of HMGCR was decreased, although not significantly.

GSEA of the proteomics data showed that the SC of Pex2^−/− mice had a significant negative enrichment score for “Cholesterol metabolism with Bloch and Kandutsch–Russell pathways”, whereas the CBR had a significant positive enrichment score (Figures 3B,D and Supplementary Table S5). There was a trend toward negative and positive enrichment scores in the BS and CBL, respectively (Figure 3D). In addition, GSEA using gene sets from the GOBP database revealed significant negative enrichment scores for “Sterol biosynthetic process”, “Steroid metabolic process”, “Steroid biosynthetic process”, and “Sterol homeostasis” in the SC of Pex2^−/− mice, whereas the CBR has a positive enrichment score for “Isoprenoid biosynthetic process” (Supplementary Figure S4A and Supplementary Table S5). The protein levels of most cholesterol biosynthetic enzymes were decreased in the SC of Pex2^−/− mice (Figure 3E). In particular, the rate-limiting enzyme HMGCR was strongly downregulated in the SC. HMGCR levels were also decreased in the BS and CBL, though this decrease was not significant. These data are consistent with the observed HMGCR activities in these CNS regions (Figure 3C). Despite an upregulation of activity in the Pex2^−/− CBR, HMGCR protein levels remained unchanged, suggesting that posttranscriptional modifications contribute to the increased activity (Figures 3C,E). The protein levels of 24-dehydrocholesterol reductase (DHCR24), which catalyzes the last step in the Bloch pathway, the conversion of desmosterol to cholesterol, were significantly decreased in the SC and BS. In CBR and CBL, DHCR24 could not be detected, but only 7-dehydrocholesterol reductase (DHCR7), which catalyzes the last step of the Kandutsch–Russell pathway. Consistent with the increased enzyme activities of HMGCR, IDI1, FDPS and FDFT1 in the CBR of Pex2^−/− mice, the protein levels of most cholesterol biosynthetic enzymes were also significantly increased in the CBR (Figure 3E). The lipogenic pathways responsible for cholesterol synthesis are the same in oligodendrocytes as in other cells, except that oligodendrocytes preferentially use ketone bodies rather than glucose as a substrate for cholesterol synthesis (Koper et al., 1981). In fact, mRNA and protein levels of the mitochondrial enzyme 3-hydroxy-3-methylglutaryl-CoA synthase 2 (HMGCS2), which is involved in ketone body synthesis (Hegardt, 1999), were significantly increased in the CNS of Pex2^−/− mice (Figures 3E, 4A).

CYP46A1, which converts cholesterol to 24(S)-hydroxycholesterol for excretion from the CNS to the systemic circulation, is expressed exclusively in neurons (Dietschy and Turley, 2004). The protein levels of CYP46A1 were decreased in the CBR, CBL, and BS of Pex2^−/− mice, and the mRNA expression of Cyp46a1 was significantly decreased in the CBR, DE, and BS (Figures 3E, 4C).

While plasma lipoproteins do not cross the blood–brain-barrier, many of the proteins involved in transporting cholesterol in the circulation are also present in the CNS, suggesting that these proteins are involved in cholesterol transport among cells of the brain (Björkhem and Meaney, 2004; Dietschy and Turley, 2004; Vance et al., 2005). The brain contains high-density lipoprotein-like transport vehicles for lipids (Vitali et al., 2014), and astrocytes in particular distribute lipids via lipoproteins (Wahrle et al., 2004; Karten et al., 2006; Kim et al., 2008; Mahley, 2016). The protein levels of apolipoprotein E (ApoE), which is the predominant apolipoprotein in the CNS and mainly expressed by astrocytes and microglia, were significantly decreased in the SC and CBR of Pex2^−/− mice (Figure 3F). However, levels of other lipoproteins such as ApoA1, ApoA4, ApoB, ApoC1, and ApoM, were also decreased in Pex2^−/− mice (Figure 3F).

In conclusion, the activities and protein levels of cholesterol biosynthetic enzymes in the SC, where white matter is more abundant, were decreased in Pex2^−/− mice, whereas they were significantly increased in the cerebral cortex, which is enriched for gray matter. Furthermore, decreased protein levels of several apolipoproteins suggest that intercellular cholesterol and lipid transport is negatively affected in the CNS of Pex2^−/− mice.

We previously reported that the expression of SREBP-2-regulated cholesterol biosynthesis genes was significantly increased in the liver and extrahepatic tissues of Pex2^−/− mice at P10 (Kovacs et al., 2004, 2009; Charles et al., 2020). SREBP-2 regulates the expression of all cholesterol biosynthesis genes, however, the individual genes are regulated to varying degrees (Horton et al., 2002). Next, we examined the expression of Srebf2 and its target genes in the CNS of P10 control and Pex2^−/− mice to further characterize the cholesterol biosynthetic pathway (Figure 4A). While the expression of Srebf2 and Insig2b was comparable in all CNS regions of control and Pex2^−/− mice, the expression of Insig1 was significantly increased in Pex2^−/− mice. Insig1 is an SREBP-2 target gene and blocks the ER-to-Golgi trafficking of SCAP (SREBP cleavage-activating protein)/SREBPs (Brown et al., 2018). The expression of all cholesterol biosynthesis genes and Ldlr (low-density lipoprotein receptor) was significantly upregulated in the CBR of Pex2^−/− mice (Figure 4A). Hmgcr, Fdps, Lss (lanosterol synthase), and Ldlr mRNA levels were significantly increased in both CBL and DE, whereas Sqle (squalene epoxidase) and Msmo1 (methylsterol monooxygenase 1) were significantly upregulated in DE (Figure 4A). The expression of Hmgcs1, Pmvk (phosphomevalonate kinase), Mvd (mevalonate diphosphate decarboxylase), Idi1, Sqle, Msmo1, Lss, Dhcr24, and Ldlr was significantly reduced in the SC of Pex2^−/− mice (Figure 4A). In contrast to genes involved in cholesterol biosynthesis, Hmgcs2 expression was significantly increased in SC, CBL, DE and CBR, while showing a trend toward increased expression in BS (Figure 4A). In general, it can be said that altered enzyme activities and protein levels are reflected in altered transcription of these enzymes.

Next, we examined the spinal cord and cerebral cortex of newborn (P0) control and Pex2^−/− mice to see if the expression of cholesterol biosynthetic genes was already altered (Supplementary Figure S4B). In the spinal cord, only the expression of Dhcr7 was significantly reduced, whereas in the cerebral cortex the expression of Insig1, Hmgcr, Mvd and Dhcr7 was significantly increased.

Cells release lipids through the action of ATP-binding cassette (ABC) transporters, of which ABCA1 and ABCG1 are the most abundant ones in astrocytes, whereas ABCG4 mediates sterol efflux in neurons (Kim et al., 2008; Chen et al., 2013; Wang and Eckel, 2014). The nuclear hormone receptor liver X-activated receptors alpha (LXRα) and beta (LXRβ) are activated by certain oxysterols and desmosterol and positively regulate the expression of Abca1, Abcg1, Abcg4, ApoE, and ApoD (Wang et al., 2002, 2008; Whitney et al., 2002; Yang et al., 2006; Spann et al., 2012; Courtney and Landreth, 2016; Berghoff et al., 2022). Whereas Lxrb expression was similar in control and Pex2^−/− mice, Lxra expression was decreased in the BS, CBL, and CBR of Pex2^−/− mice (Figure 4B). Abca1 expression was significantly decreased in the SC, BS, and DE and tended to be decreased in CBL and CBR (Figure 4B). ABCA1 protein levels were also significantly decreased in the BS of Pex2^−/− mice but were not detected in the other tissues (Figure 3F). Abcg1 expression was very low and could not be reliably determined, and Abcg4 expression was similar in control and Pex2^−/− mice. While ApoE protein levels were decreased in the SC, CBL, and CBR (Figure 3F), Apoe expression was significantly decreased only in the SC (Figure 4B). The expression of Apod was significantly decreased in the SC and BS, but increased in the CBL, DE, and CBR (Figure 4B). Interestingly, an increased expression of Apod was observed in cultured astrocytes and in the brain of Npc1 knockout mice (Suresh et al., 1998), a model for the inherited lysosomal cholesterol storage disorder Niemann-Pick disease type C, which is associated with defects in cellular cholesterol homeostasis. The levels of ApoD were also increased in the hippocampus and cerebrospinal fluid of Alzheimer’s patients (Rassart et al., 2000).

Fatty acid synthesis and myelination in the CNS are intimately connected, as myelination heavily relies on the synthesis of specific lipids, including fatty acids, to build the myelin sheath around neurons (Montani, 2021). It has been shown that endogenous fatty acid synthesis in oligodendrocytes (OLs) and Schwann cells is required for CNS and peripheral nervous system myelination, respectively (Montani et al., 2018; Dimas et al., 2019). Previous studies have shown that VLCFAs accumulate in the brain of Pex2^−/− mice at postnatal days 5, 11 and 13, as both the C26:0/C22:0 ratio and C26:0% were significantly increased (Faust et al., 2001). The increased levels of VLCFAs can be easily explained by the fact that in Pex2^−/− mice, the proteins of peroxisomal fatty acid β-oxidation are among the most reduced proteins (Figure 1E). In the CNS, the primary fatty acids involved in myelination are long-chain fatty acids, such as docosahexaenoic acid (C22:6n3; DHA) and arachidonic acid (C20:4n6), which are critical components of the myelin membrane. DHA is also considered to be essential for normal neurologic development, especially in the brain and retina, and a deficiency in brain DHA is associated with abnormal brain development in humans with ZSD (Faust et al., 2001). The synthesis of DHA requires one cycle of peroxisomal β-oxidation (Wanders et al., 2023), and accordingly it has been shown that the DHA content in the brain of Pex2^−/− mice is reduced compared to controls at birth and at 5, 11 and 13 days after birth (Faust et al., 2001).

GSEA of proteomics data clearly showed a negative enrichment of “fatty acid metabolism” in the SC, BS, and CBR of Pex2^−/− mice (Figure 5A and Supplementary Table S5), and the downregulation of proteins under this term suggests impaired fatty acid activation, degradation, and synthesis. SLC27A1 (FATP1) protein levels were significantly increased in the SC, BS, and CBL of Pex2^−/− mice (Figure 5B). It mediates the import of long-chain fatty acids (LCFAs) into the cell by facilitating their transport at the plasma membrane and also functions as an acyl-CoA ligase, catalyzing the ATP-dependent formation of fatty acyl-CoA using LCFAs and VLCFAs as substrates, which prevents fatty acid efflux from cells and might drive more fatty acid uptake (Hirsch et al., 1998; Watkins, 2008; Murphy, 2017). The fatty acid-binding proteins FABP1, FABP4, and FABP5 were significantly increased in the CBL of Pex2^−/− mice (Figure 5B). FABP4, typically found in adipocytes, has been shown to be significantly overexpressed in cerebellar liponeurocytoma, a very rare CNS tumor, compared to normal adult cerebellum (Anghileri et al., 2012). In the infantile monkey cerebellum, both granule cell progenitors in the external granular layer and oligodendrocyte progenitors in the internal granular layer express FABP5 (Boneva et al., 2010). Protein levels of acyl-CoA synthetase long-chain family member 4 (ACSL4) were also increased in all CNS regions examined (Figure 5B). ACSL4 catalyzes the conversion of long-chain fatty acids to their active form acyl-CoA for both cellular lipid synthesis and degradation via β-oxidation. ACSL4 preferentially activates arachidonate and eicosapentaenoate as substrates and is therefore a likely regulator of lipid mediator production (Golej et al., 2011). Proteins of the mitochondrial beta-oxidation of fatty acids (e.g., ACADS, ACADM, ACADL, HADH, HADHA, HADHB, ACAA2, ECHS1) were generally not significantly altered or moderately upregulated (Figure 5B).

ACACA and FASN, which catalyze the first steps in fatty acid synthesis, were significantly decreased in the SC and BS (Figure 5B). Particularly in SC, proteins involved in fatty acid metabolism were significantly downregulated. Elongases and desaturases play a critical role in regulating the length and degree of unsaturation of fatty acids in mammalian cells (Supplementary Figure S5). Elongation of very long-chain fatty acids 5 (ELOVL5) was decreased in the SC, BS, and CBL, and ELOVL7 and fatty acid desaturase 2 were significantly downregulated in SC (Figure 5B). Another protein that was strongly reduced in SC and BS was fatty acid 2-hydroxylase (FA2H) (Figure 5B). FA2H catalyzes the hydroxylation of free fatty acids at the C-2 position to produce 2-hydroxy fatty acids, which are building blocks of sphingolipids and glycosphingolipids found in neural tissue and major constituents of myelin (Eckhardt, 2023). The levels of other proteins involved in sphingolipid metabolism were also significantly altered (Figure 5B). Levels of ceramide synthase 2 (CERS2) and dihydroceramide desaturase 1 (DEGS1) were significantly decreased in the SC and BS, while CERS1 levels decreased in the CBL (Figure 5B). Serine palmitoyl transferase 1 (SPTLC1) and sphingosine kinase 2 (SPHK2) levels were significantly decreased in the CBR (Figure 5B). In contrast, levels of sphingomyelin phosphodiesterase 3 (SMPD3), which catalyzes the hydrolysis of sphingomyelins, and glucosylceramidase beta 2 (GBA2), which catalyzes the hydrolysis of glucosylceramides, were significantly increased in the SC and BS (Figure 5B).

Next, we determined the mRNA expression of genes involved in fatty acid synthesis and their regulation (Figure 5C). The activities of the enzymes involved in the elongation and desaturation of fatty acids are primarily regulated at the transcriptional level rather than by post-translational modifications (Guillou et al., 2010). While the expression of Srebf1c and fatty acid synthase (Fasn) was comparable in all CNS regions of control and Pex2^−/− mice, the expression of Elovl7 was significantly decreased in Pex2^−/− mice (Figure 5C). In addition, the mRNA levels of Elovl1, Elovl5, and Elovl6 were significantly decreased in the SC and BS of Pex2^−/− mice (Figure 5C). The expression of Elovl1 was also significantly decreased in the CBL and CBR and tended to be decreased in the DE of Pex2^−/− mice, whereas the expression of Elovl2 was significantly increased in the DE and CBR (Figure 5C). Interestingly, ELOVL1 has been shown to catalyze the synthesis of both saturated VLCFA (C26:0) and monounsaturated VLCFA (C26:1) in X-linked adrenoleukodystrophy (X-ALD) skin fibroblasts (Ofman et al., 2010; Schackmann et al., 2015). ELOVL1 was not detected at the protein level, but since Elovl1 expression was significantly reduced while C26:0 levels were increased in the brain, other elongases might be responsible for the synthesis of C26:0 in the CNS of Pex2^−/− mice. Elovl3 expression was very low and could not be reliably determined. Among the desaturases, only the expression of Scd2 was significantly reduced in SC and BS, and the expression of Scd1 was increased in the DE (Figure 5C).

We examined the spinal cord and cerebral cortex of newborn control and Pex2^−/− mice to see if the expression of fatty acid biosynthetic genes was already altered (Supplementary Figure S6A). In the spinal cord, only the expression of Fasn, Scd1, Elovl7 and Fa2h was significantly reduced, whereas in the cerebral cortex the expression of genes encoding fatty acid biosynthetic enzymes was not altered.

We investigated whether the changes in the levels of proteins involved in sphingolipid metabolism were also reflected at the level of gene expression. In general, the expression of genes encoding proteins involved in sphingolipid metabolism was unchanged or only slightly altered (Supplementary Figure S6B). While CERS1 protein levels were reduced in the cerebellum, Cers1 expression was significantly increased. In line with the altered protein levels in the SC, the mRNA levels of Cers2, Degs1, Sphk2, and Sptlc1 decreased significantly, while Smpd3 expression increased. Smpd3 expression increased significantly in the CBL, DE, and CBR, and increased slightly in the BS. Although SPTLC1 and SPHK2 protein levels were significantly decreased in the CBR, their gene expression remained unchanged. The expression of Degs1 was significantly decreased in the BS, which is in line with the reduced protein level.

Myelin formation in the rodent CNS occurs mainly during the first 3 weeks of postnatal development. It has been proposed that myelination proceeds in a strictly rostral-caudal direction in the CNS, with the exception of the dorsal spinal cord, where myelination starts first in the cervical enlargement and continues in both rostral and caudal directions (Foran and Peterson, 1992; Ozarkar et al., 2025). To examine the effect of peroxisome deficiency on the myelination in the CNS, we analyzed activities, expression and protein levels of myelin proteins in P10 control and Pex2^−/− mice (Figure 6). The specific activity of 2′,3′-cyclic nucleotide 3′-phosphodiesterase (CNP), a widely used marker protein of myelin-forming glial cells (Vogel and Thompson, 1988; Sprinkle, 1989; Chandross et al., 1999), in the individual parts of the CNS is a good representation for the stage of myelination, with the highest activities in the spinal cord and brain stem and the lowest in the cerebral cortex (Figure 6A). CNP activity was decreased by 50–60% in all CNS regions of P10 Pex2^−/− mice compared to controls (Figure 6A).

Immunoblot analysis of myelin proteins CNP and myelin basic protein (MBP) was performed to determine whether the decreased CNP activities are a reflection of decreased myelination in Pex2^−/− mice. The protein levels of CNP and MBP were significantly decreased in the SC, BS, CBL, and DE of P10 Pex2^−/− mice (Figure 6B). The heterogeneity of MBPs is generated by alternative splicing of a single gene. CNP and MBP were barely detectable in the cerebral cortex of P10 control mice and were undetectable in Pex2^−/− mice (data not shown).

We performed comparative Northern blot analysis to quantify the mRNA expression of myelin-synthesizing proteins by measuring the expression levels of mRNAs encoding for Mbp, Cnp, and proteolipid protein (Plp1) in control and Pex2^−/− mice (Figures 6C,D). The expression of Mbp, Cnp, and Plp1 was significantly decreased in all CNS regions of P10 Pex2^−/− mice (Figures 6C,D). Thus, myelin gene expression does not reach its normal peak in peroxisome-deficient Pex2^−/− mice.

Representative immunohistochemical staining for MBP in the SC and BS (Figure 7A) and CBL (Figure 7B) demonstrated a prominently reduced level of myelination in P10 Pex2^−/− mice. Spinal cord and medullary axonal myelination were markedly reduced in Pex2^−/− mice (Figure 7A). The reduced area in which MBP is present illustrated the abnormally reduced amount of myelinated deep cerebellar white matter and folial white matter at P10. The previously described cerebellar foliation anomaly is also seen (Figure 7B) (Faust et al., 2001, 2005; Faust, 2003).

Using various proteomic approaches, the proteome of biochemically purified myelin from the brains of C56Bl/6N mice was determined and 1,155 proteins were identified (Jahn et al., 2020). It was shown that PLP1, MBP, CNP, and myelin oligodendrocyte glycoprotein (MOG) represent about 73% of the total myelin protein (38, 30, 5, and 1%, respectively) (Jahn et al., 2020). However, this myelin protein dataset extends well beyond the most abundant myelin components. We showed that myelination is significantly reduced in Pex2^−/− mice at P10 using immunoblots against MBP and CNP, the second and third most abundant myelin proteins (Figure 6B). Next, we compared all proteins detected in SC, BS, CBL and CBR in our study with the mouse CNS myelin proteome determined by Jahn et al. (2020) to investigate how other myelin-related proteins are altered in Pex2^−/− mice at P10. Among the detected proteins, 1,006, 1,002, 936, and 971 proteins in SC, BS, CBL, and CBR, respectively, overlapped with the mouse myelin proteome (Figure 8A and Supplementary Table S7). We then compared all significantly altered proteins (log₂ FC and p-value cut-offs were set at ±0.5 and 0.05, respectively) in SC, BS, CBL and CBR (706, 467, 361, 447, respectively) with the mouse myelin proteome. Among the significantly altered proteins, 145, 67, 61, and 85 proteins in SC, BS, CBL, and CBR, respectively, overlapped with the mouse myelin proteome (Figures 8B,C; Supplementary Figure S7A). The protein levels of the majority of the overlapping and significantly altered proteins were decreased in the SC, BS, CBL, and CBR of Pex2^−/− mice at P10 (Figures 8B,C). For example, the levels of major myelin structure proteins such as PLP1, MBP, CNP, MOG, and MAG were significantly decreased in Pex2^−/− mice (Figure 8C; Supplementary Figure S7B).

In addition, 20 of the overlapping proteins were detected in the SC, BS, and CBL, with the levels of 15 proteins being down-regulated and 5 proteins being up-regulated (Supplementary Figure S7B). The downregulated proteins PLP1, MBP, CNP, TPPP (tubulin polymerization promoting protein), ENPP6 (ectonucleotide pyrophosphatase/phosphodiesterase 6), GSN (gelsolin), MOG, and UGT8 (UDP-glycosyltransferase 8) are well-known myelin proteins (Jahn et al., 2020). TPPP acts as a microtubule nucleation factor in oligodendrocytes and promotes the elongation of the myelin sheath, and Tppp knockout mice have hypomyelination and motor coordination defects (Fu et al., 2019). ENPP6 has been identified as a marker of newly forming oligodendrocytes (Xiao et al., 2016), it is required for the development of the myelin sheath and Enpp6^−/− mice exhibit hypomyelination (Morita et al., 2016). UGT8 catalyzes the transfer of galactose to ceramide, a key enzymatic step in the biosynthesis of galactocerebrosides, which are abundant sphingolipids of the CNS myelin sheath. Hypomyelination has been reported in Ugt8^−/− mice (Dupree et al., 1998). The actin disassembly factor gelsolin is required for normal CNS myelin wrapping (Zuchero et al., 2015). Similarly, MBP is also essential for CNS myelin wrapping and actin disassembly (Zuchero et al., 2015). S100B (S100 calcium binding protein B) is expressed in oligodendrocytes, highly expressed in astrocytes, and one of the most abundant soluble proteins in the brain (Michetti et al., 2019, 2023). It is also used as an astrocyte activation marker. S100B was significantly downregulated in the CNS of Pex2^−/− mice at P10 (Supplementary Figure S7B). The upregulated proteins catalase (CAT), GSTM1 (glutathione S-transferase mu 1), and MGST3 (microsomal glutathione S-transferase 3) are involved in the response to oxidative stress. Increased levels of GLUL (glutamate-ammonia ligase), which regulates the levels of toxic ammonia and converts neurotoxic glutamate to harmless glutamine, may be part of the response of the Pex2^−/− CNS to inflammation or cellular stress. ALDH18A1 plays a role in glutamate metabolism and may also be involved in the response to oxidative stress in the CNS of Pex2^−/− mice.

Next, we investigated whether the changes in the protein levels of myelin-associated proteins are also reflected at the level of gene expression. In accordance with the significantly reduced protein levels of UGT8, MOG, MAG, TPPP, ENPP6 and S100B (Figure 8C), the mRNA levels of these proteins were also significantly reduced (Figure 9). Not only were the protein levels of ERMN (ermin) significantly decreased in the SC, BS, and CBL, but the gene expression was also reduced in all regions of the CNS of Pex2^−/− mice. ERMN is an actin-binding protein found almost exclusively in the CNS and expressed almost exclusively in myelinating oligodendrocytes where it plays a role in the late wrapping and/or compaction phases of myelinogenesis as well as in the maintenance and stability of the myelin sheath (Ziaei et al., 2022). The mRNA levels of Tspan2 (tetraspanin-2) were significantly reduced in the SC, BS, CBL, and CBR and tended to be decreased in the DE (Figure 9). Tspan2 was shown to be expressed predominantly by cells of the oligodendrocyte lineage, but Tspan2^−/− mice showed normal development and biogenesis of myelin (Yaseen et al., 2017). However, TSPAN2 and PLP1 may play a role in regulating neuroinflammation, since they were shown to be involved in suppressing astrocyte and microglia activity (Zeis et al., 2016). The protein levels of EML1 (echinoderm microtubule-associated protein-like 1), a microtubule-associated protein, were significantly decreased in the SC, BS, and CBL, and its expression was also downregulated in the SC, BS, CBL, and DE (Figure 9). EML1 indirectly affects myelination by playing a role in axon guidance and neuronal migration (Collins et al., 2019; Zhang et al., 2024).

The expression of aspartoacylase (Aspa), the enzyme responsible for hydrolyzing N-acetylaspartate (NAA) into acetate and aspartate, was significantly decreased in the SC, BS, and CBL and tended to be decreased in the DE and CBR (Figure 9). Protein levels of ASPA were decreased by ~45 and 30% in the SC and BS, respectively, while ASPA was not detected in the CBL and CBR (Figure 8C). NAA is highly concentrated in the brain and numerous studies have implicated that acetate produced from the hydrolysis of NAA is utilized in the synthesis of lipids and acetylation of nucleosomal histones (Mattan et al., 2010; Appu et al., 2017). Accordingly, it has been shown that Aspa gene expression peaks during myelination, and that a loss of ASPA activity leads to defective myelin synthesis (Namboodiri et al., 2006; Mattan et al., 2010; Moffett et al., 2011).

In the SC, BS, and CBR of Pex2^−/− mice, the mRNA levels of Gjc3 (gap junction protein gamma 3), which is highly expressed in myelinating glial cells of the CNS and peripheral nervous system (Altevogt et al., 2002), were significantly decreased (Figure 9). Sphingolipids containing 2-hydroxylated fatty acids are among the most abundant lipid components of the myelin sheath, and both protein and mRNA levels of Fa2h were decreased in Pex2^−/− mice (Figures 5B, 9). Fa2h expression was already reduced in the SC of newborn Pex2^−/− mice, whereas expression in the P0 cerebral cortex was not altered (Supplementary Figure S6A). Unc-51 Like Kinase 4 (ULK4) is a crucial regulator of myelination and it has been shown that in Ulk4 hypomorphic mice, myelination is reduced by ~50% and the expression of myelination-related genes, major oligodendrogeneic transcription factors, and stage-specific oligodendrocyte factors is significantly decreased (Liu et al., 2018). Interestingly, Ulk4 was similarly expressed in control and Pex2^−/− mice (Figure 9).

Oligodendrocyte (OL) lineage progression and terminal differentiation into mature oligodendrocytes are under tight transcriptional and post-transcriptional control (Elbaz and Popko, 2019; Montani, 2021). The OL development can be divided into the following four stages: oligodendrocyte progenitor cells (OPCs), late OPCs, premyelinating OLs, and myelinating OLs. OPCs proliferate and migrate to their final destination where they differentiate into mature OLs and myelinate axons. All OPCs can be identified by the expression of the proteoglycan Ng2 (Cspg4), the platelet-derived growth factor receptor alpha (Pdgfra), and the transcription factors Olig1/2 and Sox10. These transcription factors are expressed along the entire oligodendrocyte lineage, including mature OLs. NKX2-2 is an important determinant of oligodendrocyte differentiation, and its ablation does not affect OPCs but leads to loss of OLs (Qi et al., 2001). However, the expression of these markers showed a mixed picture. The expression of Sox10 was decreased by ~25% in the SC, DE, and CBR of Pex2^−/− mice, whereas Nkx2-2 was significantly decreased in the SC and CBR (Figure 9). The mRNA levels of Olig1 and Olig2 were similar in control and Pex2^−/− mice, whereas the expression of Sox9 and Pdgfra was significantly increased in the SC, CBL, and CBR of Pex2^−/− mice. Sox9 is another transcription factor that controls gliogenesis and its ablation results in the loss of OPCs (Elbaz and Popko, 2019). Ng2 expression was significantly increased in the SC of Pex2^−/− mice and similar in other CNS regions of control and Pex2^−/− mice. The differentiation of OPCs into oligodendrocytes and regulation of myelin genes depends on the transcription factor MYRF, and its ablation results in the loss of OLs (Emery et al., 2009; Koenning et al., 2012; Elbaz and Popko, 2019). However, the expression of Myrf was similar in control and Pex2^−/− mice. In conclusion, while changes in the expression of markers of lineage progression and terminal differentiation of OL are observed, a dramatic reduction in gene expression that would cause the absence of these proteins and thus prevent the formation of mature OL is not evident.

Next, we investigated whether altered AMPK, AKT/mTOR and MAPK signaling is associated with the hypomyelination phenotype in P10 Pex2^−/− mice. Adenosine monophosphate-activated protein kinase (AMPK) is a metabolic sensor that is activated by falling energy levels to restore energy balance (Steinberg and Hardie, 2023). Studies have shown that proper AMPK signaling is crucial for normal brain development and myelination (Dasgupta and Milbrandt, 2009). Liver kinase B1 activates AMPKα at Thr172 in response to ATP depletion. The SC, BS and CBR of Pex2^−/− mice showed a moderate, but significant, reduction in phosphorylation of AMPKα (Figure 10). Activated AMPK phosphorylates HMGCR at Ser 872 and decreases its activity under conditions of energy stress (Burg and Espenshade, 2011). Since AMPK was not activated, we also did not detect increased phosphorylation of HMGCR in the SC, which could contribute to its reduced activity (Supplementary Figure S4C).

It has been shown that signal transduction through the AKT and mTOR (mechanistic Target Of Rapamycin) pathway regulates myelin biogenesis and myelin maintenance in the CNS and PNS (Bercury et al., 2014; Lebrun-Julien et al., 2014; Norrmén et al., 2014; Wahl et al., 2014; Bercury and Macklin, 2015; Khandker et al., 2022). Knocking out Pten (phosphatase and tensin homologue), an upstream inhibitor of the phosphatidylinositol 3-kinase (PI3K)/AKT pathway, in oligodendrocytes leads to significant hypermyelination throughout the CNS, without affecting OPC differentiation (Gaesser and Fyffe-Maricich, 2016). Myelination mainly depends on mTORC1 (mTOR complex 1) function, with minor contributions from mTORC2. Studies using rapamycin, an inhibitor of mTORC1 and partial inhibitor of mTORC2, suggested that mTOR regulates oligodendrocyte differentiation (Figlia et al., 2018). It has also been shown that ribosomal protein S6 (RPS6) activation, which is a main target of mTORC1, in differentiating oligodendrocytes correlates with the peak of myelination in the brain and spinal cord (Benardais et al., 2022). mTOR also acts as a key regulator of cholesterol biosynthesis (Khandker et al., 2022). mTORC1 has been shown to modulate the activity of SREBPs in mammalian cells through various mechanisms, including p70 ribosomal S6 kinase (p70 S6K) (Peterson et al., 2011; Wang et al., 2011; Eid et al., 2017). p70 S6K is the only kinase known to be directly activated by mTORC1, which then activates RPS6 (Laplante and Sabatini, 2012). We assessed mTORC1 activity by analyzing the phosphorylation of RPS6 in lysates from the SC, BS, CBL, and CBR of P10 control and Pex2^−/− mice. Phosphorylation of RPS6 at Ser240/244 was significantly decreased in all CNS regions of Pex2^−/− mice (Figure 10), suggesting that mTORC1 activity is lower in the peroxisome-deficient CNS.

The mTOR pathway is linked with AKT signaling, both upstream of mTORC1 and downstream of mTORC2 (Laplante and Sabatini, 2012). The PI3K/AKT pathway is also involved in the transport of SCAP/SREBP-2 from the endoplasmic reticulum to the Golgi, thereby contributing to the control of SREBP-2 activation (Du et al., 2006). Unexpectedly, AKT phosphorylation at Ser473 was similar in control and Pex2^−/− mice (Figure 10).

The extracellular signal-regulated kinases (ERK1 and ERK2) are prototypical members of the mitogen-activated protein (MAP) kinase family, which links gene expression to extracellular stimuli (Gaesser and Fyffe-Maricich, 2016). In vitro studies have shown that the mature form of SREBP-2 is a substrate of ERK–MAPK, which affects its trans-activity (Kotzka et al., 2000, 2004). ERK signaling has also been associated with various aspects of oligodendrocyte development, including proliferation, migration, survival, differentiation and myelination (Gaesser and Fyffe-Maricich, 2016). There is also a significant crosstalk between the PI3K/AKT/mTOR and RAS/MAPK pathways, which can influence each other negatively or positively (Aksamitiene et al., 2012). Furthermore, Erk1/2 double knockout mice showed reduced myelination and decreased myelin gene and protein expression, which persisted into adulthood (Ishii et al., 2012). To examine a potential link between the MAPK pathway and the expression of cholesterol biosynthetic genes and myelination in Pex2^−/− mice, we measured total protein levels and phosphorylation of ERK1/2 in P10 Pex2^−/− mice compared to controls. Total ERK levels were similar in control and Pex2^−/− mice, while phosphorylated ERK levels showed a trend toward increase in Pex2^−/− mice (Figure 10).

Our findings suggest that reduced mTOR activity plays a role in the observed hypomyelination in Pex2^−/− mice.

Since the expression of Tspan2, which plays a role in regulating neuroinflammation by suppressing astrocyte and microglia activity, was reduced in the CNS of Pex2^−/− mice at P10, we investigated whether inflammation in the CNS might already play a role in P10 Pex2^−/− mice. Indeed, the expression of the pro-inflammatory cytokines Tnfa, Il6, and Il1b was increased in the CNS of Pex2^−/− mice (Figure 11A). The expression of Il10, which is primarily anti-inflammatory in the brain, was very low and could not be reliably determined. The cytokines CSF1 (colony stimulating factor 1) and IL34 are ligands for the CSF1 receptor, which is essential for microglial development, survival and proliferation. It has been shown that they induce microglial proliferation during neurodegeneration. The expression of Il34 was reduced in the DE and CBR, and Csf1 mRNA levels were only increased in the DE (Figure 11A). Next, we examined whether NF-κB (Nuclear Factor kappa-light-chain-enhancer of activated B cells), which is a central transcription factor that is known to regulate the expression of pro-inflammatory cytokines, was activated. As marker for NF-κB activation we determined the phosphorylation of IκB-α at Ser32 (Vallabhapurapu and Karin, 2009). The phosphorylation of IκB-α was increased in the SC and CBR of Pex2^−/− mice (Supplementary Figures S8A,B).

To gain further insight into neuroinflammation in P10 Pex2^−/− mice, we examined the expression of homeostatic and disease-associated microglial signature genes. Triggering receptor expressed on myeloid cells 2 (TREM2) is predominantly expressed on microglia and is a critical regulator of their proliferation, survival, phagocytosis, polarization, inflammation, and metabolism (Ulland and Colonna, 2018; Deczkowska et al., 2020; Yu et al., 2025). Trem2 and its downstream adaptor Tyrobp/Dap12 (TYRO protein tyrosine kinase-binding protein) belong to the homeostatic microglial genes, and their expression is upregulated in disease-associated microglia (DAM) (Deczkowska et al., 2018). While Trem2 expression was significantly increased in all CNS regions of Pex2^−/− mice examined, Tyrobp mRNA levels were significantly increased in the CBL, DE, and SC, and tended to increase in the BS (Figure 11A). We also determined the mRNA levels of Aif1 (allograft inflammatory factor 1, also known as Iba1), whose expression is upregulated when microglia become activated. Aif1 expression was significantly increased in the BS and CBL of Pex2^−/− mice but decreased in the CBR (Figure 11A). DAM state is characterized by the repression of several homeostatic genes, such as the chemokine receptor Cx3cr1 and Tmem119. However, Cx3cr1 expression was significantly increased in the BS and CBL (Figure 11A). The expression of Tmem119, a protein that marks brain-resident microglia but not CNS-associated macrophages (Ruan and Elyaman, 2022), was increased in the CBL of Pex2^−/− mice, but not in other regions of the CNS (Figure 11A). The expression of Cx3cl1, a ligand for CX3CR1 that is primarily produced in neurons, was similar in control and Pex2^−/− mice (Figure 11A). We also examined the expression of genes involved in the phagocytic activity of microglia which are upregulated in DAM, including Cd68, Clec7a and Cst7. The expression of Cd68 and Clec7a was similar in control and Pex2^−/− mice, but Cst7 mRNA levels were increased in the CBL and CBR and decreased in the SC (Figure 11A). Interestingly, the expression of the pro-inflammatory Spp1 (osteopontin), a gene associated with DAM and the microglial neurodegenerative phenotype (MGnD), was significantly decreased in the SC and BS of Pex2^−/− mice.

Microgliosis is associated with astrogliosis in CNS pathologies, and it has been hypothesized that activated microglia initiate and maintain astrogliosis (Zhang et al., 2010). We examined the expression of astrocytic reactivity markers Gfap (glial fibrillary acidic protein), Osmr, and Serpina3n (Figure 11B). The mRNA levels of Gfap were significantly higher in the BS, CBL, and DE of Pex2^−/− mice, while Serpina3n expression increased in the SC, CBL and DE (Figure 11B). Osmr mRNA levels were similar in control and Pex2^−/− mice. The mRNA and protein levels of S100B, a marker of astrocyte activation, were significantly decreased in the CNS of Pex2^−/− mice at P10 (Figures 9, 11D; Supplementary Figure S7B). Next, we investigated by Western blotting whether the changes in the expression of Gfap were also reflected at the protein level. Western blot analysis showed that GFAP protein levels were similar in the SC, BS and CBR, but decreased in the CBL (Figure 11C). In contrast, proteomics analysis showed that GFAP levels decreased in the SC and increased in the BS (Figure 11D). We also examined the protein levels of additional astrogliosis markers (Zhang et al., 2010; Jurga et al., 2021), including vimentin (VIM), ALDH1L1 (aldehyde dehydrogenase 1 family member L1), aquaporin 4 (AQP4) and synemin (SYNM), in the proteomics data set. However, no pronounced increase in astrogliosis marker protein levels was observed. SERPINA3N was significantly increased in SC, GFAP in BS and ALDH1L1 in CBL (Figure 11D). Conversely, VIM and AQP4 protein levels were significantly decreased in SC and CBL, while the levels of other proteins that did not show significant changes tended to decrease (Figure 11D). SYNM is an intermediate filament protein that associates with GFAP and vimentin in astrocytes, and several studies have shown that SYNM expression increases in reactive astrocytes (Jing et al., 2007; Potokar et al., 2020). SYNM protein levels showed a downward trend in the SC and an upward trend in the BS (Figure 11D).

In summary, the expression of pro-inflammatory cytokines is increased and the expression of some homeostatic and disease-associated microglial genes has changed. However, full DAM activation is not yet observed in P10 Pex2^−/− mice.