SLC40A1 mediates ferroptosis and cognitive dysfunction in type 1 diabetes

Objective: Cognitive dysfunction often accompanies diabetes. Both hypoglycemia and hyperglycemia cause cognitive dysfunctions. However, the underlying pathophysiology remains unclear. Recent evidence show that ferroptosis primarily triggers nerve cell death, Alzheimer's disease (AD), Huntington’s disease (HD), and Parkinson's disease (PD). We purposed to investigate whether ferroptosis is a vital pathogenic pathway in diabetes-induced cognitive dysfunction. Methods and results: Type 1 diabetic rat model was created by intraperitoneal injection of streptozotocin (STZ). Signicant cognitive dysfunction was observed in the diabetic rats as evidenced by increase in latency period to nd a hidden platform and decreased cumulative time spent in the target quadrant in the Morris water maze test. We detected the amplitude of low-frequency uctuation (ALFF) of the BOLD (Blood Oxygenation Level-Dependent) signal using resting-state functional magnetic resonance imaging (rs-fMRI). Consequently, we found that the ALFF values, as well as the T2 relaxation time of the bilateral hippocampus, were reduced in Type 1 diabetic rats. We detected Fe 2+ level and lipid peroxidation products (MDA and 4-HNE) in the hippocampus. Mitochondria and neuron injury in the STZ-induced diabetic rats were determined using a Transmission Electron Microscope and Nissl body staining. Iron overload and ferroptosis were detected in the hippocampus. Furthermore, mRNA microarray analysis revealed 201 dysregulated mRNAs in STZ-induced type 1 diabetes (T1D). Pathway enrichment analyses indicated that differentially expressed mRNAs associated-coding genes were associated with ferroptosis. Among ferroptosis signaling pathway genes, Slc40a1 gene (ferroportin) was downregulated. Conclusion: We show that ferroptosis is associated with diabetic cognitive dysfunction and Slc40a1 mediates ferroptosis in Type 1 diabetes.


Introduction
Diabetes (diabetes mellitus, DM) is a metabolic disease characterized by hyperglycemia. It has been increasing in recent years and has become a global public health issue. Globally, it is estimated that the number of adult diabetes patients will reach 642 million by 2040 [1]. Diabetes often results in many chronic complications, including cardiovascular disease, kidney disease, retinopathy, peripheral neuropathy, and brain complications [2]. Due to its high morbidity, long course of the disease, and many complications, DM is a threat to human health with a substantial economic burden on families and society [3]. Therefore, for considerable social signi cance, it is necessary to conduct research on the prevention and treatment of diabetes and its complications.
Among the DM complications, there is limited research on cognitive dysfunctions. Some patients with type 1 and 2 DM show cognitive dysfunctions with a remarkable negative impact on their daily living activities. Both hypoglycemia and hyperglycemia have been shown to cause cognitive dysfunctions. The pathophysiology of cognitive dysfunction DM patients include: (1) the role of hyperglycemia, (2) the role of the vascular disease, (3) the role of hypoglycemia, and (4) the role of insulin resistance and amyloid [4]. However, the underlying pathophysiology remains unclear [5]. In 2012, the Cell Death Nomenclature Committee (NCCD) proposed a new mechanism of programmed cell death (Ferroptosis) [6]. This mechanism is iron-dependent fatal lipid peroxidation induced cell death and the process is characterized by an increase in intracellular iron, and lipid peroxidation [7]. Morphologically, the mitochondria shrink, its membrane thickens, and the ridges decrease or disappear [8]. Henceforth, internationally, ferroptosis has attracted much research attention. A signi cant number of animal model studies have shown that the iron overload causes accumulation of iron in the brain, resulting in severe cognitive de cits [9,10] and is a crucial injury mechanism in neurodegenerative diseases [11]. Currently, it is believed that ferroptosis is the main trigger of nerve cell death [12], including Alzheimer's disease (AD) [13], Huntington's disease (HD) [14], and Parkinson's disease (PD) [15]. Moreover, AD has been referred to as diabetes of the brain or type 3 diabetes [16]. We, therefore, hypothesized that iron deposition and ferroptosis constitutes an important pathogenic pathway in Diabetes-induced cognitive dysfunction.
We rst established a Streptozotocin (STZ)-induced T1D rat model. After that, we investigated the behavioral impact and cerebral alterations in the model through behavioral phenotyping and resting-state functional magnetic resonance imaging (r-fMRI). We further examined the iron-ion (Fe 2+ ) level and products of lipid peroxidation (MDA and 4-HNE) in the hippocampus. Mitochondria were also observed through a transmission electron microscope. Finally, the expression pro les of mRNA/lncRNA were assayed using a rat microarray to examine ferroptosis signaling. Herein, we aimed to assess whether ferroptosis contributes to the pathogenesis of diabetes-induced injury to neurons; therefore provides potential therapeutic targets.

Rats and diabetic model
The Institutional Animal Care and Use Committees approved the rats' experimental protocols. All the experimental protocols were strictly performed as per the national guidelines for the care and use of experimental animals. Sixty mature male Sprague-Dawley (SD) rats (weight =250 ~ 270g) were obtained from SPF Biotechnology Co., LTD (Beijing). We housed the rats in cages in groups of three and maintained under standard conditions (20℃ ~ 25℃, 50%~65% relative humidity, 12h light: 12h dark light cycle) with ad libitum access to rodent pellet diet and water. After a 7-day acclimatization period, we randomly assigned the rats into the control group (n = 20) and the model group (n = 40). The rats in the model group were subjected to 12h-fasting and then subjected to the regime of STZ-induced T1D. Brie y, the rats were intraperitoneally (i.p.) injected with 55 mg/kg STZ (diluted in 0.01M sodium citrate, pH 4.5; Sigma-Aldrich, St. Louis, MO, USA. S0130-500MG). The rats in the control arm received an injection of 0.01M sodium citrate. We measured blood glucose levels (by the tail) daily using a glucose meter. The rats in the model arm with blood glucose level ≥ 16.6 mmol/l at three days post-STZ-injection were included in our study. All rats were sacri ced at eight weeks post-STZ injection.

Morris water maze
Body weights and fasting blood glucose levels (by the tail) were monitored weekly. After eight weeks of modeling, we used the Morris water maze for behavioral testing. The water maze experimental device constituted a cylindrical pool with a diameter of 150cm and a height of 60cm, which we arti cially divided into four quadrants. During the experiment, we lled the pool with water to a depth of 30cm. The temperature was controlled at 23 ± 2°C. A circular platform (10cm × 10cm) was placed about 1cm below the level of the target quadrant (TQ). The pool was surrounded by curtains to avoid interference from the external environment. Various striking visual stimulus pictures were posted on the wall of the pool to help rats learn to identify the direction. Brie y, we individually placed the rats in the apparatus in one of the four different quadrants and allowed them up to 60s to reach the submerged platform in the hidden platform test. We guided the rats to the submerged platform if they failed to nd it, and allowed all of them to remain on the platform for an additional 10s. The behavioral data were recorded automatically and analyzed using a video tracking system (#WMT-100S, Techman soft, Chengdu, Sichuan, China). We performed an hidden platform test for ve days, and on the 6th day, the probe trial test was conducted without a hidden platform. During the test, the cumulative time spent in the target quadrant was recorded and analyzed using the same video computer system.

Total RNA preparation and LncRNA microarray analysis
The total RNA was isolated from the hippocampus region. We used 1% agarose gel electrophoresis and NanoDrop ND-1000 (Thermo Fisher Scienti c, Waltham, MA, USA) for sample quanti cation and quali cation. Subsequently, we ampli ed and transcribed the total RNA into uorescent cRNA using the Agilent Array platform, followed by hybridization using the Rat LncRNA Microarray v2.0 (Arraystar, Rockville, USA, including 13,611 lncRNAs and 24,626 protein-coding mRNAs) overnight. After that, the microarray was scanned using an Agilent DNA Microarray ScannerG2505C (Agilent Technologies, Santa Clara, CA, USA). We analyzed and processed the acquired images using the Agilent Feature Extraction (v11) and GeneSpringGXv12.1 (Agilent Technologies). Standard routine analyses, including probe summarization (RMA), quality control, and probe annotation, were performed. The differentially expressed lncRNA (DE lncRNAs) and genes (DEGs) between groups were identi ed using the t-test with the criteria of fold change > 1.5 and p <0 .05.

Gene Ontology (GO) and KEGG analysis
The Gene Ontology (GO) categories of DEGs were derived from the GO database (http://www.geneontology.org). We determined the KEGG pathways associated with DEGs based on the (Kyoto Encyclopedia of Genes and Genomes, http://www.genome.jp/kegg). KEGG pathways and GO terms with p < .05 were considered as categories of interest.
Quantitative real-time PCR (qRT-PCR) The relative expression levels of four target mRNAs (including SLC40A1) and lncRNAs in the rat brain were determined using qRT-PCR using the primers listed in Table 1. Brie y, we extracted total RNA from the rat hippocampus using the TRIzol reagent (TaKaRa, Tokyo, Japan) as per the manufacturer's protocol. After that, we synthesized rst-strand cDNA and ampli ed following the procedure: denaturation at 94°C for 2 min; 40 cycles of 94°C for 20s, 58°C for 20s, 72°C for 20s; and a nal extension at 72°C for 4min. The ampli cation was performed on an Agilent StratageneMx3000PRT-PCR machine (Agilent Technologies) using a Bestar® Sybr Green qPCR master mix kit (DBI Bioscience, Shanghai, China). We used GAPDH as the internal reference gene in our study. The relative expression levels of target mRNAs and lncRNAs were computed using the 2 -∆∆Ct methods.

Western blot analysis
We utilized Western blotting analyses to examine the expression of SLC40A1 protein in the rats' brains. The protein lysates were extracted from the rats' hippocampus using the lysis buffer (Beyotime). We used the Bradford protein assay kit (Thermo Fisher Scienti c Inc.) to quantify the protein samples following the manufacturer's protocol, followed with separation using 10% SDS-PAGE. We performed protein immunoblotting on the PVDF membrane (Merck Millipore, Billerica, MA, USA) with primary antibodies against SLC40A1 (1: 500 ab58695 Abcam, Cambridge, UK) and GAPDH (1: 10000, Abcam). HRP goat anti-rabbit/mouse IgG was used as the secondary antibody (1: 20000, Boster Biotechnology). The membranes were managed with enhanced chemiluminescence (ECL) and then examined using image-Pro Plus 6.0 software. The densitometric data were normalized to the GAPDH level.

Resting state functional magnetic resonance imaging (rs-fMRI) and relaxation time test
We placed the rats in a special anesthesia box and anesthetized them with 2.5% iso urane mixed with oxygen, and then xed the prone position on the rat's special scan bed. The head of the rat was xed with a dental rod and an ear rod to prevent head movement during the scan. The rats' breathing, heart rate, temperature, and other vital signs were closely monitored, and the concentration of the inhaled anesthetics was adjusted in time to prevent accidental anesthesia during the entire data collection process. We utilized the water circulation heating system to maintain the rats' body temperature at about 37°C. According to the depth of anesthesia, the breathing rate was maintained at about 45 times/minute.
We acquired the functional image using the Bruker's 7.0 T magnetic resonance imaging system (Pharmascan). The surface coil with a diameter of 38mm was used to transmit the RF pulse signal and the four-channel surface coil to receive the signal. Repetition time (TR) = 2000ms, echo time (TE) = 20ms, eld of view = 35mm × 35mm, layer thickness (thickness) = 0.55mm, ip angle ( ip angle) = 90°, matrix size = 64 × 64,600 repetition. All original data were converted to DICOM format data using the Paravision 6.0 software that comes with the machine. The transverse relaxation time T2 (T2 relaxation time) uses a multi-layer multi-echo scanning sequence T2map-MSME. The main parameters of the sweep were as follows: (repetition time (TR) = 2855ms, slices (slices) = 12, layer thickness (slice thickness) = hand-drawn interested region (region of interest, ROI), T2 relaxation-time is calculated by workstation software Paravision6.0. The statistical parameter map toolkit (spmratIHEP) and the resting state functional magnetic resonance data processing assistant DPARSF software package were used for data analysis.

Histopathological examinations
We isolated the brain tissues and para n-embedded for Nissl's staining. Serial sections of 5μm were cut onto slides using a rotary microtome (Leica, Wetzlar, Germany) and depara nized. The protocol of staining methods was conducted using cresyl violet (Nissl's staining, Sigma-Aldrich) following the manufacture's instruction. We captured the morphological changes of the hippocampus area and hippocampal neurons using light microscopy (BX51, Olympus, Tokyo, Japan). We utilized ImageJ for digital pathology image analysis.

Transmission electron microscope
We used four specimens of the CA1 zone of the right hippocampus tissue about 1mm 3 , and through xation, rinsing, dehydration, in ltration, embedding, and other steps to make ultra-thin sections with a thickness of approximately 60-80nm. After that, we double-stained the specimens using uranyl acetate and lead citrate for 10min, respectively before overnight drying. We then selected some representative sections for observation using TEM (Hitachi H-7000FA. Wuhan University, China).

Iron assay
We estimated the Ferrous ion (Fe 2+ ) level in the hippocampus using the iron assay kit (MAK025, Sigma-Aldrich, USA) as per the manufacturer's instructions. Brie y, after washing the tissues in ice-cold PBS, we weighed and homogenized them in 6X volumes of iron assay buffer. After that, we added the iron reducer to each supernatant. The mixture was incubated for 30min, and the output was promptly assessed on a colorimetric microplate reader (OD = 593nm). We normalized the Fe 2+ level to the weight of the tissue samples.

MDA assay
We determined the content of MDA in the hippocampus using a Lipid Peroxidation (MDA) Colorimetric/Fluorometric Assay Kit Catalog # K739-100 BioVision USA as per the manufacturer's instructions. Brie y, we weighed the tissue samples and homogenized them in 30X volumes of buffer on ice. Subsequently, we centrifuged the tissue lysates at 13,000g for 10min at room temperature to collect the supernatant. The MDA in the hippocampus sample reacted with thiobarbituric acid (TBA) to generate an MDA-TBA adduct, which we calorimetrically quanti ed (OD = 532nm). We normalized the MDA level to the weight of tissue samples.

4-Hydroxynonenal (4-HNE) ELISA
We used the 4-Hydroxynonenal (4-HNE) ELISA Kit (Catalog # E4645-100, BioVision, USA) to determine the content of 4-HNE in the hippocampus according to the manufacturer's procedure. Brie y, we weighed and homogenized the tissue samples in 9X volumes of PBS. After that, we centrifuged the tissue lysates at 5,000g for 10min at room temperature to collect the supernatant and subsequently added the appropriate microtiter wells. We assayed all samples in triplicates and measured the optical absorbance at 450nm using a Synergy™ HT Multi-Mode Plate Reader (Biotek Instruments, Winooski, VT). A standard curve was constructed (0-2000 pg/ml), and then we determined the concentrations of 4-HNE in the hippocampus from the standard curve.

Statistical analyses
Data are expressed as mean ± standard error (SEM). Utilizing Statistical SPSS 13.0 (GraphPad, USA) software, we performed the statistical analyses. The latency to the platform in the water maze was conducted using the two-way ANOVA. We used the Least Signi cant Difference Test (LSD-test) to compare the two groups. A comparison between the two groups was performed using the T-test, and p < .05 considered as statistically signi cant. We constructed all the statistical graphs in GraphPad Prism 7.0 (GraphPad Software, Inc., Cary, NC).

Results
STZ treatment successfully induced type 1 diabetic rat, and compromised learning and memory ability As shown in Figure 1, the blood glucose of rats increased sharply after intraperitoneal injection of STZ, and remained high for 8 weeks afterwards. Blood glucose levels remained normal in the control group (p =0.00001, Fig. 1A). The weight of rats decreased gradually along the disease time course. As the rats grow, their weight increased, which remained relatively stable with no obvious growth trend. The weight of rats in the control group increases over time. This caused a signi cant difference in weights between the two groups (p =0.012, Fig. 1B). Further, it was observed that it took diabetic rats longer to nd the hidden platform compared to the control rats (p =0.004, Fig. 1C). In addition, the diabetic rats spent lower cumulative time in the target quadrant than rats of control group (Fig. 1D). Thus, we infer that STZ treatment effectively modeled diabetes in rats causing cognitive dysfunctions.
The ALFF value and T2 relaxation time decreased in the brain of diabetic rats The rs-fMRI technology was employed to examine the changes in brain structure and function in diabetic rats. Notably, the ALFF value of the bilateral hippocampus and double cingulate cortex were lower in diabetic rats relative to control rats ( Fig. 2A, blue areas). The transverse relaxation time T2 of the hippocampal region of interest was detected by magnetic resonance imaging using multi-slice multi-echo scanning sequence. Results showed that T2 relaxation time of the hippocampal region of interest in was signi cantly shorter in diabetic rats that in control group rats (P=0.000 Fig. 2B). Thus, cognitive dysfunction occurred in diabetic rats. Particularly, the hippocampus was severely impaired as evidenced by iron deposition in hippocampal cells.
Iron deposition in hippocampus promoted ferroptosis in diabetic rats Measurement of Fe 2+ , MDA and 4-HNE levels in the hippocampus of rats revealed that they were signi cantly higher in diabetic rats than in control group rats (P=0.22,P=0.00, P=0.00, Fig. 3A, B, C). Electron microscopy showed that mitochondria in the hippocampal CA1 area of type 1 diabetic rats was signi cantly smaller while the sputum was shorter and diminished than in rats of control group, (p=0.000, Fig. 3D). Nissl's staining assay con rmed that Nissl stained areas was signi cantly lower in the hippocampal CA1 area of diabetic rats relative to rats of the control group (Fig. 3E). These indicated iron depositing in hippocampus which promoted ferroptosis in diabetic condition.
Changes in mRNA/LncRNA expression pro les of hippocampus of type 1 diabetic rats To identify the key genes related to ferroptosis in hippocampus of diabetic rats, microarray analysis was conducted based on Rat LncRNA Microarray v2.0 comprising 24,626 protein coding mRNAs and 13,611 lncRNAs. A total of 201 dysregulated mRNAs and 101 lncRNAs (Supplementary Table S1 and S2) were found in hippocampal tissues of diabetic rats (Fig. 4A). In the pathway enrichment analyses, we found that mRNAs found to be differentially expressed were enriched in 'AGE-RAGE signaling pathway in diabetic complications', 'glycerophospholipid metabolism', 'Cell cycle', 'Mineral absorption', and 'Ferroptosis', (Fig. 4B). Among ferroptosis signaling pathways, Slc40a1 (Ferroportin) the main gene, was downregulated in diabetic rats (Fig. 4C). These results suggested that ferroptosis was associated with diabetic cognitive dysfunction and Slc40a1 was the key gene that enhanced ferroptosis.

Slc40a1 was signi cantly down-regulated in the hippocampus of type 1 diabetic rats
The qRCR method was used to verify results of mRNA/lncRNA chip screening (The sequences of primers for PCR in Supplementary Table S3). SLC40A1, Atf3, LOC102554280, and LOC102551779 in the diabetic group were markedly lower (Fig. 5A, P =0.03, P =0.000, P =0.018, P =0.034) whereas Sult1a1, Heph, LOC103693580, LOC102549320 were markedly higher than in control group rats (Fig. 5A, P =0.001, P =0.001, P =0.001 P =0.004, P =0.002). Good agreement was identi ed between the test results and the lncRNA/mRNA chip results, indicating that the results of mRNA/lncRNA chip were reliable. Western blotting assay con rmed that the expression of Slc40a1 was lower in hippocampal tissues of diabetic rats relative to control group rats (Fig. 5B, p = 0.002)

Discussion
The development of cognitive dysfunction during diabetes is often latent and not easily recognized. Diabetes accelerates brain aging and promotes the occurrence of dementia [17]. Therefore, it is necessary to develop preventive measures to suppress cognitive impairment in diabetic conditions. Of note, diabetes is linked with cognitive impairment, especially type 2 diabetes, through multiple coexisting diseases (such as obesity, hypertension, hyperlipidemia). These coexisting diseases are independently associated with cognitive de cits [18]. To avoid interference caused by these underlying conditions, type 1 diabetes model was established in this study in rats. The pathogenesis of diabetes cognitive impairment includes impaired insulin signaling pathways, toxic effects of hyperglycemia, oxidative stress, vascular factors, and genetic factors [19]. Currently, its pathogenesis is not fully de ned. Following the recent classi cation of ferroptosis as an independent type of cell death, several studies have been designed to de ne its mechanisms [20][21][22]. Evidence exists con rming that ferroptosis occurs in brain tissues given that these tissues contain the highest proportion of polyunsaturated fatty acids of all body tissues. For this reason, they have high lipid peroxidation levels [23]. Other studies have shown that ferroptosis in uence the development of many neurologic diseases, such as Parkinson's disease, Alzheimer's disease, and stroke [24]. Against this background, we speculate that ferroptosis may drive the cognitive impairment in diabetic conditions. Currently, the link between diabetic cognitive dysfunction and ferroptosis is blurry.
A rat model of type 1 diabetic rat was created using streptozotocin (STZ). Cognitive function of diabetic rats was tested using a water maze experiment. Compared with normal group rats, diabetic rats showed a signi cantly longer escape latency in the positioning and navigation experiment, and spent shorter time in the target quadrant in the space exploration experiment. This indicates that their learning and memory ability was signi cantly reduced. In the subsequent experiment, we employed the rs-fMRI technology to assess the cognitive decline of type 1 diabetic rats by observing the brain structure and function change. Results showed that ALFF signals in bilateral hippocampus and anterior cingulate cortex were signi cantly reduced. The rs-fMRI technology which employs blood oxygenation level-dependent (BOLD) signals to quantify spontaneous brain activity is often exploited in neuroscience research [25]. In the application of this technology, the amplitude of low-frequency uctuation (ALFF) [26] re ects the regional intensity of spontaneous uctuations in the BOLD signal, which reveals the spontaneous neural activity in certain physiological states and brain regions. Thus, ALFF is an ideal measure that can reveal the brain activity [27]. A decrease in ALFF indicates that the spontaneous activity of neurons in the corresponding brain area is weakened and structure function is impaired. In contrast, increase in ALFF indicates enhanced brain activity [28,29]. In our study, ALFF signals in the hippocampus and anterior cingulate cortex of type 1 diabetic rats were reduced, indicating that these central neurons related to learning and memory activity were weakened. Based on this, it was con rmed that type 1 diabetes compromised cognition functions, which is broadly consistent with clinical studies [30]. In addition, the multi-layer multi-echo scanning sequence T2map-MSME was used to detect the T2 relaxation time of the hippocampus. This test showed that the relaxation time of the hippocampus of diabetic rats was shorter compared with rats in the normal group. T2 relaxation time re ects the brain iron content to a certain extent, and the two are negatively correlated. That is, the shorter the relaxation time, the higher the brain iron content [31,32]. In this study, iron content in the hippocampus was measured with colorimetric method. Results revealed that the iron content of diabetic rats was higher relative to that of rats in normal group. Thus, our ndings con rm that cognitive decline in type 1 diabetic rats is accompanied by iron deposition in hippocampal tissues.
Ferroptosis is recently de ned type of cell death, often re ected by iron-dependent lipid peroxidation causing cell death and oxidative stress [33]. Several products are produced during lipid peroxidation. The process of peroxidation is initiated by lipid hydro-peroxides (LOOHs). Other secondary substrates that drive peroxidation include aldehydes. Notably, 4-hydroxynonenal (4-HNE) and malondialdehyde (MDA) have been well-studied [34]. Lipid peroxidation results in production of MDA which has mutagenic effects as it reacts with DNA and primary amines on proteins forming crosslinked adducts [35]. Studies have implicated electrophile 4-HNE in cell proliferation and cell cycle progression where it causes cytotoxicity and suppresses gene expression. In this way, it accelerates disease development [34]. In this study, levels of 4-HNE and MDA in hippocampal tissues of diabetic rats higher than in normal rats. This indicates that diabetes is associated with lipid peroxidation. Further, mitochondrial morphology in the hippocampal CA 1 area of the rat was examined with a transmission electron microscope. Results showed that compared with normal rats, the mitochondria in the hippocampus of diabetic rats were shrunken, lacked ridges, and the mitochondrial membrane thickened. All this mitochondria characteristics were similar to those reported for ferroptosis previously [2]. Nissl body assay was employed to detect the degree of neuronal damage. Damaged neurons reduce the Nissl body [36]. In our study, compared with the control group, a signi cant reduction in Nissl was observed in the hippocampal CA1 area in diabetic rats. Therefore, we inferred that ferroptosis occurred in the hippocampus of diabetic rats.
We further used gene chip technique to identify key genes related to ferroptosis in hippocampal tissues in diabetic rats. The high-throughput-screening revealed 101 lncRNAs and 201 abnormally expressed mRNAs in the hippocampus of diabetic rats. In the pathway enrichment analyses, differentially expressed mRNAs were enriched in 'AGE-RAGE signaling pathway in diabetic complications' 'glycerophospholipid metabolism', 'Cell cycle', 'Mineral absorption', and 'Ferroptosis'. Our results demonstrated that ferroptosis was associated with diabetic cognitive dysfunction. Among ferroptosis signaling pathways, SLC40A1 the key gene, was signi cantly lower in diabetic rat hippocampus relative to control rats. SLC40A1 encodes a ferroportin (FPN). FPN is the only iron export protein on the cell membrane that maintains iron homeostasis [37,38]. Downregulation of SLC40A1 compromises intracellular iron output, causing iron overload. Some scholars have found that knockdown FPN on the membrane of neuroblastoma cells can accelerate ferroptosis in cells. [39] Other reports show that SLC40A1 is associated with Alzheimer's disease [40][41][42]. These ndings suggest that ferroptosis induced by downregulation of SLC40A1 may be an important pathogenic mechanism of cognitive dysfunction. Intracellular iron content is tightly regulated in the body through iron transport-related proteins on the cell membrane, including divalent metal ion transporter proteins (Divalent metal ion transporter 1, DMT1), Ceruloplasmin (CP), Transferrin receptor (TFR), and Ferroportin (FPN). Among these membrane proteins, FPN is only the membrane transferrin that transports iron out of the cell. Therefore, we speculate that SLC40A1 may be a key pathogenic gene driving iron deposition and ferroptosis in the hippocampus of type 1 diabetic rats. In future research, the SLC40A1 adeno-associated virus (AAV) vector should be constructed to facilitate further inquiry into the role of SLC40A1 in cognitive dysfunction in type 1 diabetes.
In summary, our results indicate that down-regulation of SLC40A1 causes iron deposition and ferroptosis in the brain, and the subsequent cognitive dysfunction in type 1 diabetes. Figure 1 Cognitive function, body weight, and blood glucose level in diabetic rats. A, the fasting blood glucose level in the groups. B, body weight in the two groups. C, the time spent to nd the hidden platform in the hidden platform test. D, the cumulative time spent in the target quadrant during the probe trial test. E and F, swim traces of control and diabetes rats in probe trials. *, **, and *** notes p < 0.05, p < 0.01, and p < 0.001 vs. control, respectively. ALFF value and T2 relaxation time of brain in diabetic rats. A, the changes of ALFF value of different areas of brain, red represents ALFF increase areas of the brain, blue represents ALFF lower areas of the brain. B, T2 relaxation time of hippocampus. *, notes p < 0.05, vs. control.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.