Blood samples were obtained from 808 patients with schizophrenia (mean age: 49.0 years [standard deviation [SD]: 14.2 years]) and 636 healthy control subjects (mean age: 40.6 years [SD: 13.0 years]) for genome resequencing of AKR1A1. Patients were randomly recruited inpatients and outpatients. The cases included 437 men (mean age: 48.5 years [SD: 14.0 years]) and 417 women (mean age: 49.5 years [SD: 14.5 years]). Control subjects comprised 283 men (mean age: 41.5 years [SD: 13.6 years]) and 424 women (mean age: 40.1 years [SD: 12.5 years]). We did not assess the association between common variants and schizophrenia as the aim of this study was to focus on rare variations to reveal large biological effects, thus enabling clarification of pathophysiology in rare cases of schizophrenia. Therefore, these samples were not matched for age or sex. Schizophrenia was diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders, fourth edition (4th ed., text rev.; DSM–IV–TR; American Psychiatric Association, 2000) to obtain a best-estimate lifetime psychiatric diagnosis, with consensus from at least two experienced psychiatrists. No structured interviews were conducted. The available medical records and family informant reports were also taken into consideration. Peripheral blood cells were obtained from patients with schizophrenia with/without the c.753G > A variant (n = 6) and two healthy controls among the subjects included in the genetic study. All participants provided written informed consent, and the study protocols were approved by the ethics committees of the Tokyo Metropolitan Institute of Medical Science (approval no. 20-17) and Tokyo Metropolitan Matsuzawa Hospital (approval no. 2018-8).

Genome was purified from whole blood by SRL (Tokyo, Japan), which is a private clinical laboratory test company. Polymerase chain reaction (PCR) amplification was performed using the primer sets and PCR kits (Supplementary Table S1) as per the manufacturer’s instructions. All coding regions, exon–intron boundaries, 5′ untranslated region (UTR), and 3′ UTR of AKR1A1 were examined by directly sequencing the PCR products using BigDye Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, United States) and ABI PRISM 3130/3500 Genetic Analyzer (Applied Biosystems). We read both strands when an inserted or deleted nucleotide yielded dual signals derived from the wild-type (WT) and mutant strands. AKR1A1 sequence analysis was performed based on RefSeq NM_006066. Variant data we found are deposited in the NCBI dbSNP: NCBI_subsnp# are 2137544288 and 5316170415–5316170441.

The exons 7–9 and the internal introns in AKR1A1 gene were amplified from the human genome using the AKR1A1_ex7-9_fw and AKR1A1_ex7-9_rv primers (Supplementary Table S2) and inserted into the pTA2 cloning vector. Then, the AKR1A1 minigene in the pTA2 vector was cut using HindIII and PstI enzymes and transferred into the pAcGFP1-C1 vector at the same site.

A c.753G > A mutant at the first position of exon 8 was produced using the KOD-plus- Mutagenesis Kit (Toyobo, Tokyo, Japan). For mutagenesis, AKR1A1_ex7-9_c.753_fw and AKR1A1_ex7-9_c.753_rv primers were used (Supplementary Table S2).

HEK293, SH-SY5Y, and 1321N1 cells were maintained in Dulbecco’s modified Eagle’s medium (Gibco, Waltham, MA, United States) with 10% fetal bovine serum (Life Technologies, Carlsbad, CA, United States), 100 U/ml penicillin, and 100 μg/ml streptomycin (Gibco). At ∼80% confluency in a 10 cm dish, 5 μg of DNA constructs were transfected using FuGENE6 transfection reagent (Promega, Madison, WI, United States) according to the manufacturer’s protocol. The cells were collected 48 h after transfection.

Total RNA was extracted from HEK293, SH-SY5Y, and 1321N1 cells using the SV Total RNA Isolation System (Promega) and the extracted RNA was purified using the RNeasy MinElute Cleanup Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s protocol. The obtained RNA was quantified using NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States), and 500 ng of the purified RNA was used as a template for complementary DNA (cDNA) synthesis using ReverTra Ace qPCR RT Kit (Toyobo). To investigate exon skipping, 10% of the synthesized cDNA was amplified using the pAcGFP1-C1_fw and AKex9_PstI_rv primers (Supplementary Table S2).

Total RNA derived from a human postmortem brain tissues was prepared using the SV Total RNA Isolation System (Promega), cleaned with an RNeasy MinElute Cleanup Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s protocol. A High-Capacity cDNA Archive Kit (Applied Biosystems, Foster City, California) was used for reverse transcription of total RNA to first-strand cDNA synthesis according to the supplier’s protocols.

Full-length AKR1A1 was cloned from the cDNA library using hAKR1A1_fw and hAKR1A1_rv primers (Supplementary Table S2). The amplified products were inserted into the pTA2 vector using TArget Clone (Toyobo). AKR1A1 mutants, c.753G > A and c.264delC, were produced using pTA2_AKR1A1 as a template with the KOD-plus-Mutagenesis Kit (Toyobo). AKR1A1_c.753_fw1 and rv1 primers were used for the c.753G > A mutant, and AKR1A1_c.264_fw1 and rv1 primers were used for the c.264delC mutant (Supplementary Table S2). Moreover, to delete the extra sequence below the new termination codon caused by the frameshift mutation, the c.753G > A and c.264delC mutants were amplified using the primer sets AKR1A1_c.753/c.264_fw2 and AKR1A1_c.753_rv2 and AKR1A1_c.264_rv2, respectively, and then self-ligated.

pTA2_AKR1A1_WT, pTA2_AKR1A1_c.753, and pTA2_AKR1A1_c.264 were used as templates for amplification of AKR1A1 with EcoRI and SalI restriction enzyme cut sites at the 5′ and 3′ termini, respectively. The EcoRI_AKR1A1_fw and AKR1A1_SalI_WT_rv primers for the amplification of AKR1A1 WT, AKR1A1_SalI_c.753_rv primer for the amplification of AKR1A1 c.753, and AKR1A1_SalI_c.264_rv primer were employed for the amplification of AKR1A1 c.264 (Supplementary Table S2). After amplification of AKR1A1 with EcoRI and SalI restriction enzyme cut sites, the products were cut with EcoRI and SalI enzymes. The pGEX-4T-2 plasmid was cut with EcoRI and XhoI enzymes and ligated together to make the plasmid pGEX-4T-2_AKR1A1. The sequence of pGEX-4T-2_AKR1A1 was confirmed using the pGEX_fw and pGEX_rv primers (Supplementary Table S2).

BL21 Escherichia coli transformed with pGEX-4T-2_AKR1A1 was added to 3 ml of Luria Bertani medium containing 100 μg/ml ampicillin and incubated at 37°C with shaking. After 12–24 h, the optical density at 600 nm (OD₆₀₀) was measured with GloMax Explorer Multimode Microplate Reader (Promega). The culture was diluted to OD = 0.4–0.5 with Luria Bertani medium containing 100 μg/ml ampicillin and incubated at 37°C with shaking. After 1 h, the culture was induced with 30 µl of 100 mM isopropyl β-D-1-thiogalactopyranoside and incubated at 37°C with shaking for 3 h. Cells were harvested by centrifugation at 20,400 × g for 1 min.

Glutathione-S-transferase (GST) fusion AKR1A1 from BL21 E. coli transformed with pGEX-4T-2_AKR1A1 was purified using the MagneGST Protein Purification System (Promega) according to the manufacturer’s protocols. Protein concentrations were determined using the bicinchoninic acid protein assay with GloMax Explorer Multimode Microplate Reader (Promega). Finally, we verified the purification of GST-AKR1A1 proteins using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE).

A total of 150 µl of cOmplete Lysis-M (Merck, Darmstadt, Germany) was added to the pellet of HEK293, SH-SY5Y, and 1321N1 cells or human red blood cells. The mixture was sonicated, followed by centrifugation at 20,400 × g for 15 min. The supernatant was then collected from the cell lysate.

The activity of AKR1A1 was measured by monitoring NADPH consumption. The reaction mixture contained 100 mM HEPES (pH 7.4), 0.1 mM NADPH, and 10 mM GlucA. Reactions were monitored by assessing the decrease in absorbance at 340 nm using an Epoch Microplate Spectrophotometer (BioTek Instruments, Winooski, VT, United States) at 37°C. The enzyme activity was defined as the amount of enzyme that catalyzes the oxidation of 1 µmol of NADPH per min.

Total RNA was extracted from whole blood cell in using the NucleoSpin RNA Blood kit (NucleoSpin, MACHEREY-NAGEL, Duren, Germany) and the extracted RNA was purified using the RNeasy MinElute Cleanup Kit (QIAGEN, Hilden, Germany) according to the manufacturer’s protocol. The obtained RNA was quantified using NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States), and 500 ng of the purified RNA was used as a template for cDNA synthesis using ReverTra Ace qPCR RT Kit (Toyobo).

The qPCR assay was performed using Power SYBR Green PCR Master Mix (ThermoFisher Scientific). AKR1A1_qPCR_fw primer, AKR1A1_qPCR_rv primer, GAPDH_qPCR_fw primer and GAPDH_qPCR_rv primer were used (Supplementary Table S2). The qPCR protocol was as below: at 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 56°C for 30 s and 72°C for 30 s. For relative expression, each expression was normalized to average of two subjects without c.753G > A variant using the ΔΔC_T calculation.

The statistical tests were performed using GraphPad Prism version 9.0. Results are expressed as mean ± standard deviation (SD) or standard error of mean (SEM). Comparisons between groups were performed using two-way analysis of variance (ANOVA) or one-way ANOVA followed by Bonferroni’s multiple comparison test. Statistical significance was set at p < 0.05.

The AKR1A1 sequence was analyzed in patients with schizophrenia (n = 808) and control subjects (n = 636), and 28 variants were identified as a result (Supplementary Table S3). Among them, four were novel variants: g.-403_-397 del AAGTTC, g.753-62_65 delC, g.913-114 C > A, and *85 G > C. The other variants have already been registered with the single nucleotide polymorphism database dbSNP (https://www.ncbi.nlm.nih.gov/snp/). In addition, four variants were found in the coding region: c.264 delC (rs755292778, p.Glu89ArgfsTer23), c.474 G > A (rs147059021, p.Ala158Ala), c.753 G > A (rs745484618, p.Arg251Arg), and c.911 C > T (rs150392728, p.Thr304Met), although the frequencies of the genotypes and alleles in the four variants did not noticeably differ between patients with schizophrenia and healthy controls (Table 1; Figure 1).

The c.753G > A variant is a silent mutation. This variant was identified in 14 cases in the schizophrenia patient group (including 13 heterozygous and one homozygous) and five subjects in the healthy group, but no remarkable difference in genotype (p = 0.191) or allele frequency (p = 0.074) was observed. Next, we applied the frequency of the variants to “East Asian” and “Total groups (including African/African-American, Amish, Ashkenazi Jewish, East Asian, Finnish, non-Finnish European, Latino/Admixed American, Middle Eastern, Other, and South Asian)” obtained from Genome Aggregation Database (gnomAD) v3.1.1, a public database containing human genome data, as the frequency of control. The results showed that the frequency of genotypes GA and AA was significantly higher in patients than that in “East Asian” and “Total groups.” In addition, the frequency of allele A in patients was significantly higher than that in “East Asian” and “Total groups.”

The c.264delC variant was identified as a deletion-type mutation, which can have a significant impact on the expression product. This was observed in only one patient with schizophrenia in our sequence analysis. There was no significant difference in the genotype and allele frequencies between patients and controls in our dataset. However, further analysis using the database gnomAD showed that the frequency of genotype ID and allele D in patients was significantly higher than that in “Total groups,” but not “East Asian.”

The c.474G > A variant is a silent mutation without amino acid substitution. It was only found in one healthy person. In addition, the c.911C > T variant was identified in two patients in the schizophrenia patient group (heterozygous), and it was not identified in the healthy group. Regarding these variants, there was no significant difference in the frequency of the genotype and allele between patients and controls both in our dataset and gnomAD.

A mutation at the first position of an exon has been reported to cause exon skipping by alternative splicing (Fu et al., 2011). To address whether the variant c.753G > A located at the first position of exon 8 in the AKR1A1 gene results in exon skipping, a minigene assay was performed. We constructed minigenes by inserting AKR1A1 exons 7, 8, and 9 and the internal introns into the pAc-GFP1-C1 vector and by introducing the G > A mutation at the first position of exon 8 (Figure 2A). The minigenes were transfected into HEK293, SH-SY5Y, and 1321N1 cells, and their products were confirmed by PCR. The results showed that exon 8 skipping occurred in the A allele minigene (mutant [MT]) but not in the G allele (WT) in any cell type (Figure 2B). In HEK293 and SH-SY5Y cells, MT considerably increased the frequency of exon skipping compared to WT (Figure 2C). These findings suggest that the variant c.753G > A in the AKR1A1 gene induced the skipping of exon 8.

The skipping of exon 8 causes a frameshift mutation and may result in the decreased enzymatic activity of AKR1A1 due to the c.753G > A variant, even though it is a silent mutation. To confirm this, we purified the AKR1A1 mutant recombinant protein produced by exon skipping and evaluated its activity (Figure 2D). We found that AKR1A1 produced by the c.753G > A variant exhibited no activity, which was significantly lower than that of the WT AKR1A1 (Figure 2E). Furthermore, c.264 delC mutation, which cause a frameshift mutation, also resulted in a complete loss of AKR1A1 enzymatic activity.

To investigate the effect of the c.753G > A variant on AKR1A1 enzymatic activity in patients with schizophrenia, AKR enzymatic activity in red blood cells from six patients with schizophrenia and two control subjects was measured (Figure 3A). Two out of four patients who were heterozygous at the c.753G > A variant exhibited lower enzymatic activity (patients with schizophrenia [SCZ]#1: 1.50 mU/mg protein, SCZ#3: 1.11 mU/mg protein), and a slight decrease was observed in patients with the variant compared to those without it (SCZ/GA: 1.62 ± 0.41 mU/mg protein; SCZ/GG: 1.70 ± 0.04 mU/mg protein). In addition, the enzymatic activity in four patients with GA was slightly lower than that in four subjects with GG (GA: 1.62 ± 0.41 mU/mg, GG: 1.85 ± 0.25 mU/mg). Furthermore, the enzymatic activity in patients with schizophrenia was slightly lower than that in control subjects (SCZ: 1.65 ± 0.32 mU/mg protein; control: 1.99 ± 0.32 mU/mg protein). Although there was no significant difference between any of the groups due to the small number of subjects, these results suggest that the c.753G > A variant of AKR1A1 may reduce the enzymatic activity of AKR1A1.

Finally, to investigate the effect of the c.753G > A variant on AKR1A1 expression, we quantified mRNA level in the whole blood cells obtained from five subjects (CON#2 and SCZ#6 with c.753 GG alleles, and SCZ#1, SCZ#2, and SCZ#4 with c.753 GA alleles), since unfortunately we were unable to collect peripheral blood from three individuals whose AKR activity we have measured. We found that mRNA expression of AKR1A1 with c.753 GA alleles decreased to approximately 50% compared to that with c.753 GG (Figure 3B). These results suggest that the reduction of AKR activity in subjects with the variant may be caused from the reduction of gene expression.