Thioredoxin deficiency increases oxidative stress and causes bilateral symmetrical degeneration in rat midbrain

phenotype-driven

The primary endogenous sources of ROS in mammals, including humans, are the mitochondrial respiratory chain and NADPH oxidases (NOX). (Dan Dunn et al., 2015;Nathan and Cunningham-Bussel, 2013;Sarniak et al., 2016). Living organisms have a defense system that scavenges ROS produced by mitochondria. The enzymatic antioxidant system consists of SOD, CAT, GPx, and thioredoxin. Thioredoxin was first identified as a hydrogen donor for ribonucleotide reductase in Escherichia coli (Laurent et al., 1964). In humans, thioredoxin has been identified in the culture media of adult T-cell leukemia cell lines and Epstein-Barr virus-infected cells (Wakasugi et al., 1990). Subsequently, it has been reported that thioredoxin expression is upregulated in many cancer cells and that it inhibits cancer cell apoptosis (Baker et al., 1997;Lincoln et al., 2003). However, few studies have investigated the role of the thioredoxin system in vivo. Homozygous Txn1 knockout mice are embryonically lethal, whereas heterozygous mice show normal development (Matsui et al., 1996). A previous study reported that transgenic mice overexpressing a dominant-negative mutant of Txn1 show cardiac hypertrophy (Yamamoto et al., 2003). Little is known about the effects of reduced thioredoxin activity in vivo.
We recently discovered a strain, from the Kyoto University archive, that harbors chemically induced Txn1 mutation in rats with running seizures. This study aimed to identify the causative genes for running seizures and elucidate the neuropathology of this rat strain. Here, we present a novel animal model for neurodegeneration designated as Adem "Age-dependent mitochondrial cytopathy" rat.

Animals
We used Txn1-F54L rats (Adem) generated by ENU-mutagenesis and a Txn1-T160C (p. F54L) knock-in rat generated by CRISPR-Cas 9. We used the former for the identification of the causative gene and its characterization. The latter was created to confirm the reproducibility of the phenotype. The Txn1-F54L rat was initially discovered in an ENUinduced mutant archive at Kyoto University (Mashimo et al., 2008;Serikawa et al., 2009) as a rare strain with frequent running seizures only in its juvenile stage. All animal husbandry procedures were performed in accordance with the protocols approved by the institutional experimental animal use committees Okayama University, Osaka University, and Kyoto University. Both sexes were used in the present study, as there was no difference in the frequency of running seizures. Number of rats used in experiments were summarized in Supplementary Table 1.

Genetic mapping and sequencing analysis of the Adem rat generated by ENU-mutagenesis
The rats with running seizures were backcrossed for more than ten generations on the F344/NSlc inbred background (Japan SLC, Hamamatsu, Japan). The repeated backcrosses ensure that the frequency of potential mutations induced by ENU elsewhere in the genome is reduced to approximately 1 in 4 × 10 6 bp of the genome. To identify the causative gene, we produced 117 (Adem×BN/SsNSlc) × BN/SsNSlc backcross progeny. Genetic locus mapping was determined in rats with wild running or generalized tonic-clonic seizure phenotypes. Genomic DNA was extracted from rat tail biopsies by an automatic genomic DNA isolation system Kurabo,Osaka,Japan). To localize the locus to a specific chromosomal region, genome-wide scanning on DNA samples was performed using simple sequence length polymorphism (SSLP) markers covering all autosomal chromosomes (Chrs) as previously reported (Ishida et al., 2012).
To enrich the 1.5-Mb region (Chr 5: 75,159,000 to 76,708,000), SureSelect custom DNA probes were designed by SureDesign under moderately stringent conditions and generated by Agilent Technologies (Santa Clara, CA, USA). The DNA library was prepared using SureSelect reagents and a custom probe kit. Genomic sequence analysis was performed by HiSeq 2500 (2 × 150 bp) in accordance with the standard protocol at Takara Bio (Takara Bio Inc., Shiga, Japan).

Generation of Txn1-T160C (p. F54L) knock-in rat using CRISPR-Cas 9
The possibility of genomic DNA alterations around the Txn1 gene could not be fully excluded in Txn1-F54L rats generated by ENU mutagenesis. As the Txn1 mutant rats showed a unique phenotype never reported previously, we generated a new Txn1-F54L knock-in rat to test the reproducibility of this phenotype. A Txn1 T160C (p. F54L) knock-in rat was generated by CRISPR-Cas9 genome editing with long singlestranded DNA (lssDNA), as previously reported (Yoshimi et al., 2016). A pair of guide RNAs (gRNAs) targeting intron 1 and intron 2 of rat Txn1 were designed. LssDNA for Txn1 was synthesized using a long ssDNA Preparation Kit (Biodynamics Laboratory Inc., Tokyo, Japan) following the manufacturer's protocol. The sequences of gRNA and ssDNA are listed in Supplementary Table 2. Cas9 mRNA (400 ng/μL), gRNA (200 ng/μL), and ssDNA (40 ng/μL) were delivered into 113 F344/Jcl embryos by electroporation as previously reported (Miyasaka et al., 2018). The two-cell stage embryos (n = 110) were transferred into the oviducts of five pseudopregnant females. Genomic DNA from founder rats was extracted from the tail biopsies. The CRISPR target site was amplified using the specific primer pairs listed in Supplementary Table 2.

Genotyping
The last 3 mm of the rat tail was cut at three weeks of age, and DNA was purified with a DNeasy Blood &Tissue kit (Qiagen, Hilden, Germany). A DNA fragment containing a mutation was amplified with a pair of primers as follows: FW 5 ′ -CCA CAT GGG AGA GTC ACA T − 3 ′ and RV 5 ′ -ATA GCC TGG AAG CGG TCA GAT G − 3 ′ (Sigma-Aldrich Japan, Inc., Tokyo, Japan). The PCR products (521 bp) were then treated with the restriction enzyme Bsp1286I (New England Biolabs, Ipswich, MA, USA), and the F54L mutant DNA was digested to 260 bp and 261 bp fragments. Genotyping was determined by the size of the digested fragments on a 3% agarose gel.

Video monitoring of wild running
To find the frequency of wild running and subsequent tonic-clonic seizures, both the heterozygotes (n = 8, including four males and four females) and homozygotes (n = 9, including five males and four females) were placed under a 24-h video monitoring system (Handycam HDR-CX480, SONY, Tokyo, Japan), every week, from three to nine weeks.

Histological analysis
The brains from both sexes were fixed with 10% formalin neutral buffer solution. Tissue sections were cut from paraffin-embedded brain samples taken at 2, 3, 5, and 9 weeks of age and were subjected to hematoxylin and eosin (HE) and Bodian and Klüver-Barrera (KB) staining using standard protocols. We used 31 rats for WT, 37 for heterozygotes, and 33 for homozygotes to confirm the histological changes. The stained sections were examined under a BZX-700 microscope (Keyence, Osaka, Japan). We evaluated brain lesions, including samples from the thalamus and the inferior colliculus, in the sagittal and coronal planes.

Head MRI
To evaluate the extent of brain lesions and their changes over time, an MRI scan was performed in 1-mm slices. Three animals were examined for each genotype at 5, 7, and 9 weeks of age. The same animal was examined twice on days 1 and 14 as the storage period of the MRIattached breeding room was limited to no longer than two weeks. The rats were anesthetized with isoflurane and laid prone on a cradle. Anesthesia was maintained by inhalation of 2% isoflurane in the air at 1.4 L/min through a face mask. Throughout the MRI measurements, respiratory rate and rectal temperature were monitored using a dedicated system (Model 1025, MR-compatible Small Animal Monitoring and Gating System, SA Instruments, Inc., Stony Brook, NY, USA). The body temperature was maintained by a flow of warm air using a heater system (MR-compatible Small Animal Heating System, SA Instruments).

Immunohistochemistry (IHC) analyses
A standard IHC protocol was used to stain the brain tissue samples (n = 5 or 6 rats per genotype). Each phenotype included both sexes. The individual rats used in IHC were the same as those used in histological analysis. In brief, 5-μm-sized paraffin-embedded tissue sections were deparaffinized with xylene. Antigen retrieval was performed by autoclaving at 120 • C for 5 min. Endogenous peroxidases were quenched by treatment with 0.3% hydrogen peroxide (H 2 O 2 ) in methanol for 30 min at 22-26 • C. They were then incubated with blocking serum for 1 h at 22-26 • C. Sections were incubated with the following primary antibodies to determine the cell type in the brain lesion: anti-NeuN rabbit antibody (ab177487; Abcam) and anti-Olig2 rabbit antibody (ab109186; Abcam) for 1 h at room temperature. To determine DNA oxidative injury and lipid peroxidation, sections were stained with anti-8-hydroxy-2 ′ -deoxyguanosine (8-OHdG) mouse antibody (MOG-020P, JaICA, Tokyo, Japan), and anti-4-hydroxynonenal (4-HNE) mouse antibody (MHN-020P, JaICA) overnight at 4 • C, respectively. The sections were then reacted with a biotinylated rabbit polyclonal secondary antibody (PK-6200, Vector Laboratories, Burlingame, CA, USA). To reveal the staining, we used an avidin-biotinylated peroxidase complex (PK-6200, Vector Laboratories). After washing, the slides were incubated with 3,3 ′ -diaminobenzidine (DAB) (SK-4100, Vector Laboratories) and immediately washed with tap water after colour development. The slides were then counterstained with hematoxylin, mounted with dibutyl phthalate xylene (DPX), and observed under a BZX-700 multifunctional microscope (Keyence). IHC was quantified by assessing five randomly selected fields in the inferior colliculus and thalamus. The percentages of positive cells or positive areas were calculated using the BZX Analyzer software (Keyence).

Transmission electron microscopy (TEM)
Six brain samples per genotype were obtained and further immersed in 2% glutaraldehyde and 2% paraformaldehyde in 100 mM phosphate buffer (PB) for 24 h at 4 • C. They were then postfixed with 2% osmium tetroxide in 100 mM PB for 1.5 h at 4 • C, after which they were rinsed with 100 mM PB, and dehydrated through a graded series of ethanol treatments. They were embedded in Spurr resin (Polysciences Inc., Warrington, PA, USA), cut into ultrathin sections, and stained with uranyl acetate and lead citrate. The ultrathin sections were observed using a Hitachi H-7650 TEM (Hitachi High-Tech Corp., Tokyo, Japan).

Plasmids and purification of recombinant proteins
For recombinant Thiredoxin1 proteins, rat Txn1 cDNAs were amplified by PCR using LA-Taq polymerase (Takara Bio) with FWprimer: 5 ′ -GGA TCC ATG GTG AAG CTG ATC GAG AG-3 ′ ; and RVprimer, 5 ′ -GTC GAC TTA GCT GTC CAT GTG CTG GCG TTC GAA TTT AGC GGT TTC TTT GAA TTC GGC AAA CTC CGT AAT AGT GG-3 ′ (Sigma-Aldrich Japan). Txn1 C35S mutation was introduced by PCR with C35S-FW 5 ′ -GTG GAC CTA GCA AAA TGA TCA AG-3 ′ and C35S-RV 5 ′ -TGA TCA TTT TGC TAG GTC CAC AC-3 ′ . The PCR products were cloned into a pCMV-Tag2 plasmid (Stratagene, San Diego, CA, USA) digested with BamHI and SalI containing FLAG-tag and S-tag sequences. The DNA sequence of the plasmids was confirmed by DNA sequencing using a BigDye Terminator FS Ready-Reaction Kit (Applied Biosciences, Little Chalfont, Buckinghamshire, UK) and an ABI 3130× Genetic Analyzer (Applied Biosciences). The plasmids with two clones per genotype were cloned into DH5α competent cells and purified with a Plasmid DNA purification kit (Qiagen 12,145) after being cultured overnight. The plasmids were transfected into HEK293 human embryonic kidney cells using Lipofectamine2000 (Invitrogen, Carlsbad, CA, USA). The recombinant proteins were purified with anti-FLAG agarose affinity gel (A2220; Sigma-Aldrich Japan), eluted with FLAG peptide (A3290; Sigma-Aldrich Japan), and filtered using an ultrafiltration membrane (Amicon Ultra 3 K; Millipore, Burlington, MA, USA).

Insulin-reducing activity assay
Thioredoxin activity was measured thrice using a thioredoxin activity assay kit (Redoxica, Little Rock, AR, USA) according to the manufacturer's instructions. We used two clones per genotype for recombinant assay and three rats, including both sexes, for brain tissue assay. In brief, the purified recombinant thioredoxin1 proteins or protein extracts from rat thalamus and cortex were allowed to react with insulin and NADPH in assay buffer, and thioredoxin activity was determined as the change in the amount of oxidized NADPH by measuring the decrease in absorbance at 340 nm per minute using a spectrophotometer DeNovix DS-11 (DeNovix, Tokyo, Japan).

Assay of oxidative stress in the midbrain lesions
We measured cellular H 2 O 2 concentration of the midbrain in WT and heterozygotes (n = 8, including two males and two females for each hemisphere) at five weeks using a colorimetric H 2 O 2 assay kit (no. STA-844, Cell Biolabs, San Diego, USA) according to the manufacturer's protocol.

Primary fibroblast culture
Newborn rats (n = 4 per genotype) were sacrificed under deep anesthesia with isoflurane on the day of birth. The animals were dissected in the surgical suite and disinfected with 70% alcohol. For primary fibroblast culture, the abdominal skin tissues were cut into small pieces in DMEM (Fujifilm Wako Pure Chemical Corporation, Osaka, Japan) and disrupted by pipetting. The cell suspension was transferred and maintained in DMEM supplemented with 10% fetal bovine serum (Thermo Fisher Scientific), 500 U/mL penicillin, and 500 μg/mL streptomycin (Sigma-Aldrich Japan) in a fully humidified atmosphere of 5% CO 2 . After about two weeks, fibroblast cells were cultured in DMEM supplemented with 10% fetal bovine serum, 100 U/ mL penicillin, and 100 μg/mL streptomycin.

Terminal deoxynucleotidyl transferase (TdT) dUTP Nick-End Labeling (TUNEL) assay
For determining apoptosis, primary fibroblasts derived from a total of four rats were assessed twice using the In Situ Cell Death Detection Kit TMR red (Sigma-Aldrich Japan) following the manufacturer's instructions. We compared cell death during incubation of primary fibroblasts with and without the addition of 300 μM H 2 O 2 for 3 h. In brief, the cells were cultured on an 8-chamber culture slide (BD Falcon, Franklin Lakes, NJ, USA), washed with phosphate-buffered saline (PBS), fixed with 10% formalin neutral buffer solution (Fujifilm Wako Pure Chemical Corporation) for 1 h at room temperature, and treated with permeabilization solution for 2 min on ice. TMR red-labeled dUTP was incorporated into 3 ′ -OH DNA ends by terminal deoxynucleotidyl transferase, and the fluorescence signal was detected using a BZX-700 microscope (Keyence) or a fluorescence microscope (Olympus IX71; Olympus Corporation, Tokyo, Japan).

Metabolite measurements in the brain
We considered the possibility that localization-specific metabolic differences in the midbrain might be related to neurodegeneration. Vacuolar degeneration occurs in the midbrains in both heterozygotes and homozygotes, but the lesion area is narrower in the heterozygotes. While removing brain tissue from gross anatomy, samples from heterozygotes may contain areas that lacked vacuolar degeneration. Thus, we compared metabolite differences between the thalamus and cerebral cortex in WT and homozygous rats (each n = 3). First, 5-week-old rats were anesthetized deeply enough with isoflurane and decapitated after whole blood collection. Then, the removed midbrain samples were immediately frozen with liquid nitrogen and stored in a − 80 • C freezer. The temperature was strictly controlled in a cold box with dry ice while sending the samples to the Human Metabolome Technologies (HMT) company for analysis. The analysis of 116 hydrophilic and ionic metabolites involved in central energy metabolism pathways (glycolysis, TCA cycle, amino acid metabolism, and nucleic acid metabolism) was done. Capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS) and CE-tandem mass spectrometry (CE-MS/MS) analysis were carried out using an Agilent CE capillary electrophoresis system equipped with an Agilent 6210 time-of-flight mass spectrometer (Agilent Technologies) and Agilent 6460 Triple Quadrupole LC/MS (Agilent Technologies), respectively. The systems were controlled by Agilent G2201AA ChemStation software version B.03.01 for CE (Agilent Technologies) and connected by a fused silica capillary (50 μm i.d. × 80 cm total length) with commercial electrophoresis buffer (H3301-1001 and I3302-1023 for cation and anion analyses, respectively, HMT) as the electrolyte. The time-of-flight mass spectrometer was scanned from m/z 50 to 1000 (Ohashi et al., 2008), and the triple quadrupole mass spectrometer was used to detect compounds in dynamic MRM mode. Peaks were extracted using MasterHands, automatic integration software (Keio University, Tsuruoka, Yamagata, Japan) (Sugimoto et al., 2010), and MassHunter Quantitative Analysis B.04.00 (Agilent Technologies) in order to obtain peak information, including m/z peak area, and migration time (MT).

Statistical analysis
Results are expressed as the mean ± SEM. All data were analyzed using the EZR software (Kanda, 2013). Welch's t-test was used for comparisons between two groups. One-way analysis of variance (ANOVA) was used to compare three groups. Statistical significance was set at p < 0.05. Bonferroni adjustment was used to determine the statistical differences between three groups. All significant statistical results are indicated within the figures using the following conventions: * p < 0.05, ** p < 0.01, *** p < 0.001.

Theory/Calculation
For metabolite analysis, hierarchical cluster analysis (HCA) and principal component analysis (PCA) (Yamamoto et al., 2014) were performed using HMT's proprietary MATLAB and R programs, respectively.
We performed next-generation sequencing analysis of genomic DNA from 75,159,000 to 76,708,000 on Chr5 using the SureSelect target enrichment approach (Agilent). In mutant DNA, we found only one heterozygous missense mutation in the exon region, c. T160C, in exon 3 of the Txn1 gene (NM_053800, ENSRNOG00000012081.6) by comparing the parental strain F344/NSlc. This variant caused a substitution of phenylalanine for leucine at residue 54 (Fig. 1b).
Thioredoxin is widespread in prokaryotes and eukaryotes, and its amino acid sequence is well conserved in mammals (Fig. 1d). Amino acid sequences are 90% identical in humans and rats. Since the structure of rat thioredoxin has not been reported, we plotted the position of F54 on the 3D structure of human thioredoxin (UniProt, https://www.uni prot.org/). The positions of amino acids with important functions in thioredoxin were plotted on the structure. Thioredoxin has an active center sequence Cys32-Gly-Pro-Cys35 motif and carries out redox reactions by dithiol/disulfide exchange reactions between Cys32 and Cys35 (Hirota et al., 1997;Kang et al., 2019;Lu and Holmgren, 2014). Both Cys69 and Cys73 are nitrosylated in response to nitric oxide (NO) (Hirota et al., 1997;Kang et al., 2019). Cys-73 can serve as a donor for the nitrosylation of target proteins. The location of the F54L mutation is remote from these active sites ( Fig. 1c and Fig. 1d), suggesting that it is not directly involved in the dithiol/disulfide exchange reaction or Snitrosylation. However, the PolyPhen-2 prediction (http://genetics.bwh .harvard.edu/pph2/index.shtml) found that the F54L mutation is possibly damaging with a score of 0.785, and the Sorting Intolerant From Tolerant (SIFT, http://www.ngrl.org.uk/Manchester/page/ sift-sorting-intolerant-tolerant.html) sequence is deleterious.

Vacuolar degeneration appears during an epileptic period
Video monitoring showed that the frequency of seizures peaks at five weeks of age in both heterozygotes (Fig. 2a) and homozygotes (Fig. 2b). There were no differences in seizure frequencies between heterozygotes and homozygotes.
Next, a histological examination was performed at five weeks to determine whether a structural change was the cause of epilepsy.
Surprisingly, vacuolar degeneration was found in the inferior colliculus and thalamus of the mutants. (Fig. 2c). No changes were observed in the hippocampus, which is prone to epilepsy. Bodian staining, which is used to observe nerve fibers, and Klüver-Barrera (KB) staining, which is used to observe the myelin sheath, exhibited decreased staining in the mutants. These histological tests showed that homozygotes had more severe lesions than heterozygotes. We examined when these lesions had started developing (Fig. 2d). There were no remarkable changes in the inferior colliculus at two weeks of age. However, vacuoles became evident in the inferior colliculus and thalamus at three weeks of age and were widespread at five weeks. They almost disappeared entirely after nine weeks. All rats showed similar changes to the representative findings in Fig. 2c and d, with no exceptions.

MRI showed transient high signals of T2-weighted images in the midbrain
MRI detected hyperintense regions of T2-weighted images (T2WI) spread around the thalamus in mutant rats at five weeks of age (Fig. 3a). The hyperintense regions of the heterozygous and homozygous T2WI were symmetrical. The lesions were localized in the thalamus, inferior colliculus, superior colliculus, and hypothalamus in homozygotes (Fig. 3b). The hyperintensity region of T2WI in homozygous rats was wider than that in heterozygous rats (see also Supplementary Fig. S1, S2, and S3). Temporal lesions were also examined using MRI in the same individual. Serial MRI examination revealed that the lesions at three weeks spread or were similar in size at five weeks. However, the lesions shrank at seven weeks and disappeared at nine weeks (Fig. 3c). The weekly changes observed on MRI were similar to those seen in the histological examination.

Neuronal and oligodendrocytic cell loss occurs in the midbrain
Western blotting of the midbrain using NeuN, Olig2, GFAP, and anti-IbaI antibodies showed a decrease in NeuN and Olig2 and an increase in GFAP and IbaI expression in heterozygotes and homozygotes (Fig. 4a). Immunohistochemical studies showed significantly decreased numbers of NeuN-positive neurons (Fig. 4b and c) and Olig2-positive oligodendrocytes ( Fig. 4d and e). NeuN-positive neurons in heterozygotes and homozygotes had larger cell bodies than wild-type (WT) rats. Electron microscopy of the lesions in the mutant rats revealed that microtubules in the vacuoles and myelin sheath were wrapped around the vacuoles ( Fig. 4h and i). These findings indicate that the vacuoles were dilated axons of the neurons. Examination of the mitochondrial morphology showed sparse cristae in the mutant, even though the neurons retained their structures (Fig. 4j and k). Fig. 5a shows the thioredoxin/peroxiredoxin system. Four central enzyme systems defend against oxidative stress by scavenging ROS generated in the body. These include SOD, which scavenges superoxide, CAT, GPx, and thioredoxin-dependent peroxiredoxin (Prdx). Thioredoxin forms a system with nicotinamide adenine dinucleotide phosphate (NADPH) and thioredoxin reductase (Txnrd) to reduce disulfide bonds in target proteins. Thioredoxin is responsible for converting the target protein from oxidized to the reduced form and scavenging H 2 O 2 via Prdx. H 2 O 2 concentration, an important marker of oxidative stress, in the midbrain was significantly higher (p = 0.01) in the heterozygotes than in the WT (Fig. 5b). To confirm whether the reduced expression of thioredoxin underlies the changes in the brain during 3-5 weeks of age, we examined the expression level of thioredoxin protein in the brain, kidney, heart, and liver by western blotting at four weeks of age when vacuoles appeared and at 16 weeks of age when repair occurred. We found that Txn1 was equally expressed in all organs, and there was no change between 4 and 16 weeks of age (Fig. 5f).

Txn1-F54L shows decreased insulin-reducing activity
Next, we measured thioredoxin activity by using recombinant protein (Fig. 5c), the thalamus at five weeks of age, where vacuolar degeneration was observed (Fig. 5d), and the frontal cortex (Fig. 5e), where vacuolar degeneration did not occur. Our results showed that the insulin-reducing activity of Txn1-F54L was decreased to approximately one-third of that of the WT. Similarly, insulin-reduction activity in the thalamus was low in heterozygotes and even lower further in homozygotes compared to WT rats (Fig. 5d). Interestingly, reduced activity was (b) Seizure frequency of homozygotes each week/rat (n = 9, including five males and four females). (c) Histology at five weeks of age. At least four animals of each genotype were examined. (d) The representative changes of inferior colliculus lesions from 2 to 9 weeks of age for each genotype. also observed in the cortex, where there was no vacuolar degeneration (Fig. 5e).

Oxidative damage in the midbrain and susceptibility to cell death under oxidative stress
Next, we investigated the pathogenesis underlying vacuolar degeneration. Staining with 8-OHdG, which indicates oxidative stress-induced DNA damage, and 4-HNE, which indicates lipid oxidation, showed brain staining in the mutant lesions (Fig. 6a). The area of 8-OHdG (Fig. 6b) and 4-HNE (Fig. 6c) positive cells were significantly increased in heterozygous and homozygous rats compared to WT cells. Next, we examined cellular vulnerability to H 2 O 2 using primary fibroblasts derived from WT and homozygous rats. The TUNEL assay showed that homozygous fibroblasts were slightly stained even under a standard culture medium, and the staining became significantly stronger when H 2 O 2 was added. In contrast, TUNEL-positive cells in WT fibroblasts did not increase with the addition of 0.3 mM H 2 O 2 (Fig. 6d).

WT energy metabolism in the thalamus is higher than that in the cortex
Absolute quantitative values for 116 metabolites of the thalamus and cortex in WT and homozygote rats at five weeks are summarized in Supplementary Table 3. PCA (Fig. 6e) and heat map visualization of HCA (Fig. 6f) showed that energy metabolism was different between the two areas in WT rats and the thalamus of homozygote rats compared with the WT. Compound names in the hierarchical cluster analysis are listed in Supplementary Table 3. Table 1 presents a list of metabolites that are statistically increased in the thalamus compared with those in the cortex of WT rats showing an increase in most of the central energy metabolites. On the other hand, many of the metabolites in the thalamus of homozygotes were statistically decreased compared with those of WT rats ( Table 2). The only metabolite that was increased in the thalamus with a statistically significant difference was alanine (Table 2).

Txn1-F54L rat generated by genome editing replicates symmetrical vacuolar degeneration in the midbrain
Finally, we conducted reproducibility experiments to confirm whether the Txn1-F54L mutation is the sole cause of the F344/NSlc phenotype. Genome editing using CRISPR-Cas9 was used to generate F344/Jcl rats with the Txn1-F54L mutation. We placed 5-week-old F344/Jcl Txn1-F54L heterozygous rats under 24-h video monitoring and recorded running seizures eight times/24 h (a total of n = 5). Running seizures and transition from running to tonic seizures were observed. These seizure symptoms were similar to those observed in Adem rats harboring Txn1-F54L. The lesions were localized in the thalamus, hypothalamus, and superior and inferior colliculi and were more extensive in homozygous than in heterozygous rats ( Fig. 7a and b). The vacuoles in the midbrain started developing at two weeks of age, became apparent at three weeks, and recovered at nine weeks ( Supplementary  Fig. 4). Although the genetic background of F344/NSlc generated by ENU-mutagenesis and F344/Jcl is somewhat different, transient vacuolar degeneration localized in the midbrain was reproduced in F344/Jcl rats carrying the Txn1-F54L mutation.

Discussion
The study of Adem rats, which began with a forward genetic approach, provided significant neurobiological findings. The Txn1 mutation is associated with the development of symmetrical vacuolar degeneration in the midbrain of juveniles. The F54L mutation impaired insulin-reducing activity and susceptibility to cell death under oxidative stress. Vacuolar degeneration in the midbrain seems to be linked to running seizures.

Txn1-F54L mutation causes characteristic symmetrical vacuolar degeneration in juveniles
MRI, histological examinations, and a reproducibility experiment using genome editing techniques revealed that the Txn1 mutation causes bilateral symmetric vacuolar degeneration in the rat midbrain at 3-5 weeks. Western blotting and immunohistochemistry revealed a reduction in the number of neurons and oligodendrocytes in the Adem rats. The number of NeuN-positive cells in Adem rats was significantly lower than that in the WT, and pronounced axon swelling was identified via TEM. These findings suggest that vacuoles reflect neuronal cells swelling and death. However, this study could not identify which degeneration occurs first, neuronal or oligodendrocytic. Under physiological conditions, neuronal activity controls oligodendrocyte development and myelination (Fields, 2015;Foster et al., 2019); therefore, neuronal cell death can hinder myelination. Inflammatory demyelination and axonal damages cooccur in multiple sclerosis, although their relative roles have been debated (Bitsch et al., 2000;DeLuca et al., 2006). Thus, neurons and oligodendrocytes have reciprocal interactions both physiologically and pathologically (Pease-Raissi and Chan, 2021). When changes are observed simultaneously, it is not easy to determine the dominant one.
Western blotting of five-week-old brains showed a dense band reacting to anti-Iba1 antibody, suggesting an increase in microglia.
Microglial activation is a common feature of neurodegenerative diseases. Activated microglia have been reported to kill neurons by releasing TNF-α, glutamate, reactive oxygen, and nitrogen species, which can cause apoptosis, excitotoxicity, and necrotic death of surrounding neurons (Block et al., 2007;Brown and Vilalta, 2015). Therefore, activation of microglia may exacerbate cell death.

Txn1-F54L mutation with decreased function is susceptible to cell death under oxidative stress
Our recombinant protein assay showed that the insulin-reducing activity in the mutant rats decreased to approximately one-third of that of the WT. F54L is located far from the CGPC motif, which plays a central role in thioredoxin activity. The missense mutation may cause a structural change in the region where the dithiol/disulfide exchange reaction occurs.
In primary cultured fibroblasts derived from Adem rats, cell death was induced in the culture medium with H 2 O 2 more readily than in the WT. In addition, we confirmed a higher concentration of H 2 O 2 and stronger staining of IHC with 8-OHdG and 4-HNE in the brain lesions of Adem rats than in those of the WT. These results suggest that excessive ROS production in the midbrain lesions of Adem rats induced oxidative DNA damage and oxidation of lipids.
Thioredoxin has anti-apoptotic functions in various cells (Andoh et al., 2002;Baker et al., 1997;Haendeler et al., 2002). Loss of thioredoxin function can cause cell death (Benhar, 2020; Circu and Aw,   2010; Tonissen and Di Trapani, 2009). Proteins that are associated with apoptosis through the interaction with thioedoxin include apoptosis signal-regulating kinase 1 (ASK1) (Saitoh et al., 1998;Tobiume et al., 2001) and Caspase-3 (CASP3) (Mitchell and Marletta, 2005). Thioredoxin interacts with various proteins, and thioredoxin-interacting protein (TXNIP) is one of the most well-studied proteins with essential functions in vivo. TXNIP is a physiological inhibitor of thioredoxin, and TXNIP activates the NLRP3 inflammasome and leads to apoptosis (Lerner et al., 2012;Oslowski et al., 2012). Moreover, mTOR is an interacting protein of thioredoxin (Oka et al., 2017). mTOR inhibitors have been reported to cause brain tumor cell death via the TXNIP/thioredoxin antioxidant pathway (Malone et al., 2017). Thus, inhibition of thioredoxin leads to cell death, either directly or indirectly, through an interacting protein. In addition to apoptosis, other mechanisms of neuronal death could be involved, such as phagocytosis, pyroptosis, and ferroptosis (Fricker et al., 2018).

Midbrain lesions and running seizures
As a symptom of epileptic seizures, running seizures were reported mainly in the 1980s and 1990s. Electrical stimulation of the inferior colliculus and midbrain reticular formation (Hirsch et al., 1993;Maton et al., 1997;McCown et al., 1984) or intense acoustic stimulation (Marescaux et al., 1987) induced running seizures. In addition, some evoked running seizures are followed by tonic seizures (Hirsch et al., 1993;Marescaux et al., 1987;Maton et al., 1997). The seizures in the Adem rats share similarities with these symptoms. The time of spontaneous appearance and suppression of epileptic seizures and the time of onset and repair of the midbrain lesions were almost the same. This suggests that running seizures in Adem rats are linked to midbrain lesions involving the inferior colliculus and thalamus. Due to the lack of EEG analysis, it has not been possible to determine whether this is an epileptic phenomenon. We performed simultaneous video-EEG monitoring for a long term after this study. However, because of severe body movements during the running seizures, we ultimately could not confirm EEG changes during the seizures. However, we found that absence seizures, tonic seizures, and partial seizures appeared during periods of frequent running seizures. We have reported these findings in another publication (Ohmori et al., 2022).

Dysfunction of the Txn/ Txnrd and Txn/ Prdx systems and related phenotypes
The function of thioredoxin in vivo remains unclear. Herein, we discovered a link between Txn1 mutations and CNS disorders in mammals for the first time. thioredoxin is located in the cytoplasm, nucleus, and extracellular region (Kang et al., 2019;Rubartelli et al., 1992), whereas thioredoxin-2 (Txn2) is located in the mitochondria (Spyrou et al., 1997). A homozygous Txn2 mutation has been linked to an infantile-onset neurodegenerative disorder with severe cerebellar and optic atrophy and peripheral neuropathy (Holzerova et al., 2016). To our knowledge, there are no reports of Txn1 mutations that cause human diseases to date.
Txnrd catalyzes the reduction of oxidized TXN in an NADPHdependent manner. Genetically modified mice with inactivation of Txnrd1 or Txnrd2 show embryonic lethality (Conrad et al., 2004;Jakupoglu et al., 2005), so the Txn/ Txnrd systems are essential for fetal development. Other important interacting proteins include the Prdx family, which scavenges H 2 O 2 . Knockout mice for Prdx1, Prdx2, Prdx3, Prdx4, or Prdx5 exhibited various disorders, such as hemolytic anemia, metabolic abnormalities, inflammation, cancer, and age-related phenotypes (Lee, 2020). The discovery of Adem rats may shed light on new roles of the Txn/ Txnrd and Txn/ Prdx systems.

Why is symmetrical vacuolar degeneration confined to the midbrain?
An interesting phenomenon in Adem rats is the appearance of bilateral symmetrical lesions in the midbrain at 3-5 weeks. Several experiments failed to explain why vacuolar degeneration is localized to the midbrain. First, western blotting showed that thioredoxin was expressed in several organs other than the brain of Adem rats. The thioredoxin is ubiquitously expressed throughout the body (TXN thioredoxin [Homo sapiens (human)] -Gene -NCBI https://www.ncbi.nlm.nih.gov/gene/ 7295). Second, the insulin-reducing activity of thioredoxin in Adem rats was reduced to one-third of that in the WT, both in the thalamus, where vacuolar degeneration appeared and in the cerebral cortex, where vacuolar degeneration was not present. Third, primary cultured fibroblasts derived from Adem rats showed cell death induced by excessive ROS. These results in vitro indicate that cell death can be caused by excessive ROS, regardless of the cell type. The thioredoxin/peroxiredoxin system, not the glutathione system, was reported to significantly contribute to mitochondrial H 2 O 2 removal in the brain (Drechsel and Patel, 2010). This may partially explain why degeneration occurs in the brain. However, it does not explain why degeneration does not occur in the cortex but only in the thalamus. Lastly, we wondered if the locationspecific microenvironment contributed to the limited onset of neuronal death in the midbrain. The results of the metabolome analysis showed that energy metabolism was higher in the thalamus than in the cortex. The neurodevelopmental stage at 3-5 weeks of age might have an increased mitochondrial ROS leakage due to high energy metabolism in the midbrain. Because the number of rats for metabolome analysis was small, we must interpret it cautiously. However, we must consider another possibility, namely that the F54L mutation impairs an essential midbrain protein, the oxidative-reductive state of which is regulated by thioredoxin. Thioredoxin interacts with a wide variety of proteins (Daloso et al., 2015;Fu et al., 2013;Kumar et al., 2004), but we have only identified the impairment of insulin-reducing activity. The specific proteins reduced by TXN1-F54L have not yet been identified, representing a limitation of this study.

The future challenge of elucidating repair mechanisms
Another surprising phenomenon in Adem rats is that repair of the vacuolar lesions can already be seen at seven weeks on MRI, and they are almost entirely repaired in nine weeks. Both astrocytes and microglia have been reported to be involved in cytotoxicity and tissue repair. Astrocytes have long been thought to inhibit neuronal repair, but it has been noted that glial activation induces neurogenesis and acts in repair (Rolls et al., 2009;Silver and Miller, 2004). Microglia phagocytose debris and dead neurons releasing BDNF and IGF to repair damaged neurons (Bohlen et al., 2019;Colonna and Butovsky, 2017;Nayak et al., 2014). In vacuolar lesions, glial cells with opposing effects may coexist, or switch roles from damage to repair. Although this is an exciting point, it is far from the purpose of this study, which was to identify a causative gene of running seizures and clarify the underlying pathology. We will study this in the future.

Application of Adem rats in neuroscience research
Oxidative stress is responsible for many neurodegenerative diseases (Elfawy and Das, 2019;Lin and Beal, 2006;Uttara et al., 2009) and cerebral ischemic-reperfusion injury (Allen and Bayraktutan, 2009;Orellana-Urzúa et al., 2020). New therapeutic agents for oxidative stress-related diseases can be evaluated more easily because neuronal death occurs spontaneously in a short time and does not require surgical treatment. For this purpose, Adem rats could be attractive animal models.

Conclusions
The underlying pathology of Adem rats with Txn1-F54L mutation is the loss of neurons and oligodendrocytes in the midbrain. In comparison to the WT, these mutants have a decreased insulin-reducing activity. Primary cultured fibroblasts derived from Adem rats exhibited cell death under excess H 2 O 2 . The mutation might affect the midbrain microenvironment in a specific manner, leading to the death of neurons and oligodendrocytes in that region. Further studies are required to determine this midbrain-specific function of TXN1.

Funding
This work was supported by Grants-in-Aid for Scientific Research (16H05354 and 22K07914), a Grant-in-Aid for epilepsy research from the Japan Epilepsy Research Foundation, and JSPS KAKENHI Grant Number JP16H06276 (AdAMS).

Declaration of Competing Interest
The authors declare that there are no competing financial interests.

Data availability
No data was used for the research described in the article.