Profiling and Cellular Analyses of Obesity-Related circRNAs in Neurons and Glia under Obesity-like In Vitro Conditions

Recent evidence indicates that the pathogenesis of neurodegenerative diseases, including Alzheimer’s disease, is associated with metabolic disorders such as diabetes and obesity. Various circular RNAs (circRNAs) have been found in brain tissues and recent studies have suggested that circRNAs are related to neuropathological mechanisms in the brain. However, there is a lack of interest in the involvement of circRNAs in metabolic imbalance-related neuropathological problems until now. Herein we profiled and analyzed diverse circRNAs in mouse brain cell lines (Neuro-2A neurons, BV-2 microglia, and C8-D1a astrocytes) exposed to obesity-related in vitro conditions (high glucose, high insulin, and high levels of tumor necrosis factor-alpha, interleukin 6, palmitic acid, linoleic acid, and cholesterol). We observed that various circRNAs were differentially expressed according to cell types with many of these circRNAs conserved in humans. After suppressing the expression of these circRNAs using siRNAs, we observed that these circRNAs regulate genes related to inflammatory responses, formation of synaptic vesicles, synaptic density, and fatty acid oxidation in neurons; scavenger receptors in microglia; and fatty acid signaling, inflammatory signaling cyto that may play important roles in metabolic disorders associated with neurodegenerative diseases.


Introduction
Neurodegenerative diseases, including Alzheimer's disease (AD), have been extensively studied and have a variety of risk factors associated with their initiation and progression [1,2]. For decades, the major hypothesis was that the production and accumulation of amyloid-beta peptides and tau hyperphosphorylation were early factors in AD [3,4]. However, targeting those factors did not completely prevent disease progression. Therefore, several studies have attempted to elucidate the wide range of causes of AD [5]. Recent studies have suggested an association between AD progression and metabolic disorders, such as obesity and type 2 diabetes mellitus (T2DM), based on the common major physiopathology of both diseases, such as insulin resistance and inflammation [6][7][8]. Several studies have shown that the serum from the blood of AD mouse models and AD patients commonly have glucose, insulin, and cholesterol (Chol) imbalances and abnormal secretion of inflammatory cytokines, chemokines, and free fatty acids [9][10][11].
Moreover, it has been reported that patients with obesity and T2DM suffer from impaired memory consolidation and cognitive decline, eventually leading to the development of AD [12,13]. In the central nervous system (CNS), neurons and glia play reciprocal regulatory roles in glucose metabolism, insulin signaling, and lipid metabolism to maintain brain metabolic homeostasis [14]. Neuronal and glial dysfunction damage the maintenance of brain metabolic homeostasis, resulting in cognitive impairment through poor synaptic plasticity [14][15][16]. In the brains of obese patients, pro-inflammatory cytokines, including Int. J. Mol. Sci. 2023, 24, 6235 2 of 18 tumor necrosis factor-alpha (TNF-α) and interleukin-6 (IL-6), induce neuronal apoptosis, abnormal microglial autophagy pathway, and astrocytic mitochondrial dysfunction, resulting in neurite shrinkage and memory impairment [17][18][19]. Moreover, fatty acid imbalances, including elevated saturated palmitic acid (PA), polyunsaturated linoleic acid (LA) deficiency, and elevated low-density lipoprotein cholesterol, cause brain insulin resistance, neuronal apoptosis, microglial inflammatory responses, and impaired astrocytic autophagy, leading to synaptic disruptions and memory loss in the brains of obese patients [20][21][22][23][24]. These significant results warrant further investigation into the key factors related to functional mechanisms in neurons and glia under metabolic imbalances.
Until now, many research groups have focused on the modulation of proteins, mRNAs, DNA epigenetics, and microRNAs in neurons and glia to understand and prevent the development of AD [25][26][27]. However, the functional role of non-coding RNAs (ncRNAs), including circular RNAs (circRNAs), has not been fully elucidated in neurons and glia. CircRNAs are a novel class of regulatory RNAs in which the 3 end of the downstream exon is covalently bound to the 5 end of the upstream exon through the back-splicing process [28]. Since most circRNAs are structurally stable, they are likely to play a regulatory role in brain tissue that needs to perform immediate functions; however, in some cases, circRNAs have a fast turnover [29].
Our previous studies have reported the unique expression patterns of specific cir-cRNAs in the brain cortex of obese mice [30], relating to the regulation of the neuronal cell cycle and spatial memory [31]. Similarly, many researchers are trying to elucidate the regulatory functions of circRNAs to understand several diseases, such as cancer and diabetes [32][33][34]. However, studies on the importance of circRNAs in brain metabolic diseases are scarce. CircRNAs are more distributed in the brain than the rest of the body and play a critical role in cellular mechanisms such as synapse formation and neural elongation [29]. Interestingly, a human circRNA, antisense to cerebellar degeneration-related protein 1 (CDR1-AS), is reportedly involved in developmental brain disorders through microRNA (miR)-7 sponging [35]. These studies suggest that elucidating the functions of circRNAs may be an important step in understanding the progression of diseases such as AD.
In this study, we sought to clarify the circRNAs associated with neuronal and glial function under obesity-related in vitro conditions such as neuroinflammation, insulin resistance, and high-saturated fatty acid. Our study provides critical information on various functional roles of candidate circRNAs related to obesity-related neuropathogenesis. Thus, we suggest that these circRNAs may be cardinal regulatory factors in the progression of metabolic-related AD.

Obesity-Related CircRNAs Were Specifically Expressed in Brain Cells
We previously reported a differential expression of circRNAs in the brain cortex of obese mice compared with wild-type mice [30]. Among these circRNAs, we screened 20 that had a log2 fold change greater than 0.5 or lower than −0.5 and with a statistical significance (p < 0.05) ( Figure 1A). To characterize the distribution of the circRNAs in brain cells, we checked the expression of each circRNA in the mouse cell lines: Neuro-2A neuroblastoma cells, BV-2 microglial cells, and C8-D1a astrocytes ( Figure 1B).
We confirmed that the 20 circRNAs were expressed in these cells except for circZbtb16; circKcnq2 was identified as a neuron-dominant circRNA. The identified astrocyte-dominant circRNAs were circZzz3, circNsd2, circUsp3, circKmt2a, circRabgef1, circSnx12, and cir-cAftph. The circRNAs expressed at similar levels in neurons and astrocytes were circMyrip, circStx6, circFaxc, circDennd1b, circAkap6, and circBcl2l13. The circRNAs expressed similarly in the three cell types were circTbc1d14, circFut8, circPan3, circEprs, circMap2k4, and circGnptg. These results show that each obesity-linked circRNA was differentially distributed in brain cells. However, these circRNAs tended to be mainly expressed in neurons and astrocytes, indicating that obesity-linked circRNAs might have regulatory roles in these cells. Through the RNase R treatment and Sanger sequencing analysis, we Figure 1. Differential expression and cell-type specificity of obesity-linked circRNAs in the brain. (A) A histogram displaying the differential expression of circRNAs in the brain cortices of obese mice compared to wild-type mice. The data are expressed as the average Log2 fold change (n = 4). (B) Stacked bars showing the cell-type specific expression of obesity-linked circRNAs in three mouse-brain cell lines. The total circRNA expression in the three cell lines is set at 100%, and the relative distribution of circRNAs in each cell line is expressed as percentages (n = 3). (C) Circular structure confirmation of circRNAs. Cropped bands showing the expression of circRNAs and Gapdh in untreated (−) and RNase R-treated (+) Neuro-2A mouse neuroblastoma cells. A histogram Figure 1. Differential expression and cell-type specificity of obesity-linked circRNAs in the brain. (A) A histogram displaying the differential expression of circRNAs in the brain cortices of obese mice compared to wild-type mice. The data are expressed as the average Log 2 fold change (n = 4). (B) Stacked bars showing the cell-type specific expression of obesity-linked circRNAs in three mousebrain cell lines. The total circRNA expression in the three cell lines is set at 100%, and the relative distribution of circRNAs in each cell line is expressed as percentages (n = 3). (C) Circular structure confirmation of circRNAs. Cropped bands showing the expression of circRNAs and Gapdh in untreated (−) and RNase R-treated (+) Neuro-2A mouse neuroblastoma cells. A histogram displaying changes in expression of circRNAs and Gapdh in RNase R-treated [RNase R (+)] Neuro-2A cells compared to untreated control cells [RNase R (−)]. CircRNAs' structure confirmation results from a comparative analysis of the expression value of Gapdh mRNA (linear) and the expression value of circRNAs (circular) in the RNase R (+) group. The data are expressed as a relative value of the RNase R-treated group when the untreated control value is 1. Neuro-2A: mouse neuroblastoma cells, BV-2: mouse microglial cells, C8-D1a: mouse astrocytes.

Obesity-Related In Vitro Conditions Regulated the Expression of CircRNAs in Brain Cells
To determine how the expressions of the circRNAs were changed by obesity in the brain, we chose seven obesity-related blood serum factors, glucose, insulin, TNF-α, IL-6, PA, LA, and Chol, that mimic obesity in the brain. Various studies have previously studied these factors to examine how obesity impairs brain functions [21]. We introduced each factor to Neuro-2A, BV-2, and C8-D1a cells and established the obesity-like condition models.
We observed that the expressions of various circRNAs were markedly changed in our obesity-like condition models (Figures 2 and S2-S8). We also observed that the expressions of circRNAs matched the unique expression pattern of circRNAs in the obesity-like condition models ( Figure 2) [30]. High glucose and insulin concentrations (HG/HI) significantly regulated the expression of circSnx12 in neurons, circUsp3 in microglia, and circSnx12 in astrocytes (Figures 2 and S3). TNF-α markedly regulated the expression of circGnptg in neurons and circKmt2a, circGnptg, circFaxc, and circZzz3 in astrocytes (Figures 2 and S4), indicating that TNF-α has a significant effect on circRNA expression in astrocytes. Furthermore, we observed that IL-6 significantly regulated circTbc1d14 expression in microglia only, suggesting that IL-6 plays a minimal role in regulating the expression of circRNAs in this assay (Figures 2 and S5). BSA-conjugated PA significantly regulated the expression of circKcnq2 in neurons and circGnptg in astrocytes (Figures 2 and S6). BSA-conjugated LA significantly regulated the expression of circDennd1b, circRabgef1, circFaxc, and cir-cUsp3 in neurons; circGnptg, circEprs, and circStx6 in microglia; and circStx6 in astrocytes (Figures 2 and S7), indicating that contrary to PA, LA has a significant effect on circRNA expression in neurons. Chol significantly regulated the expression of circDennd1b, circPan3, circEprs, circFaxc, and circFut8 in neurons; circPan3, circGnptg, and circKcnq2 in microglia; and circAftph and circUsp3 in astrocytes (Figures 2 and S8), indicating that Chol has a significant effect on circRNA expression in brain cells.
We further confirmed that the obesity-like conditions did not significantly affect the expression of most host genes that corresponded to each of their circRNAs (Supplementary Figure S9), indicating that obesity-like conditions regulate circRNA expression without affecting the transcription of host genes. These results comprehensively suggest that the expressions of circRNAs are significantly changed in our obesity-like condition models. These results also indicate that the unique expression patterns of circRNAs in the brains of obese mice were because of the obesity-related serum factors used in our obesity-like condition models. Moreover, these results ascertain that our obesity-like condition models are proper for the functional analysis of obesity-linked circRNAs. like condition models. Moreover, these results ascertain that our obesity-like condi models are proper for the functional analysis of obesity-linked circRNAs.

Obesity-Related CircRNAs Work as Important Factors for Brain Cell Function
To examine the role of obesity-linked circRNAs in brain cells, we first assessed cir-cRNAs that are well conserved in humans by analyzing the publicly available Gene Expression Omnibus (GEO) dataset and the expression measurement from human SH-SY5Y neuroblastoma cells. We then sorted circRNAs showing unique expression patterns in both obesity-like condition models and obese mouse brains (Figure 2). We found that five circRNAs (circStx6, circDennd1b, circUsp3, circAftph, circZzz3) are well conserved in the human brain cortex (Supplementary Figure S11). Furthermore, we confirmed two additional circRNAs (circSnx12 and circRabgef1) that were not listed in the GEO dataset but stably expressed in human SH-SY5Y cells (Supplementary Figure S11). Thus, we selected these eleven circRNAs for functional analyses in our obesity-like condition models.
We designed two independent siRNAs that bind to the back-splicing junction of each circRNA for the specific depletion of circRNAs without affecting their host mRNA (Supplementary Figure S12). We confirmed that the siRNAs sufficiently depleted the expression of each obesity-linked circRNA (Figures 4 and 5; 60-90% expression depletion). We then observed that obesity-linked circRNAs regulated various functions in brain cells under obesity-like conditions (Figures 4 and 5). CircSnx12 depletion significantly decreased the expression of Bcl2l1 in HG/HI-induced neurons, indicating that circSnx12 has a role in the inflammatory response pathway in this model (Figures 4A and S13A). Downregulation of circRabgef1 significantly decreased the expression of Stx1a in LA-treated neurons, suggesting that circRabgef1 is involved in the cascade of synaptic vesicles and synaptic density in this model ( Figures 4B and S13B). CircDennd1b depletion significantly decreased the expression of its host gene in LA-treated neurons without affecting functional genes ( Figures  4C and S13C). However, circDennd1b knockdown significantly decreased the expression of Stx1a and Syp in Chol-treated neurons without affecting its host gene expression ( Figures  4D and S13D), indicating that circDennd1b has a distinct role depending on obesity-related serum factors and is involved in the formation of synaptic vesicles and density in Choltreated neurons. CircZzz3 downregulation significantly regulated the expression of Il1b, Il6, and Pcx in TNF-α-treated astrocytes, indicating that circZzz3 is involved in inflammatory cytokine production and gluconeogenesis in this model ( Figures 5A and S13E). CircStx6 downregulation significantly regulated the expression of Ffar4 in LA-treated astrocytes, indicating that circStx6 is involved in fatty acid receptor signaling in this model ( Figures 5B  and S13F). However, circStx6 depletion did not show gene expression changes in PA-treated astrocytes ( Figures 5C and S13G), indicating that there might be other roles we did not examine in this assay and that circStx6 may also have distinct roles depending on obesity-like conditions. The downregulation of circAftph significantly increased the expression of Il6 in Chol-treated astrocytes, indicating that these circRNAs involve inflammatory cytokine production in this model ( Figures 5D and S13H). CircUsp3 knockdown significantly increased the expression of Msr1 and Cd36 in HG/HI-induced microglia, indicating that circUsp3 plays a role in scavenger-receptor signaling in this model ( Figures 5E and S13I). CircUsp3 depletion significantly regulated the expression of Tnf and Il6 in Chol-treated astrocytes, indicating that circUsp3 is involved in inflammatory cytokine secretion in this model ( Figures 5E and S13I).
Interestingly, we frequently observed that many circRNAs were involved in regulating Il6 expression in astrocytes exposed to TNF-α and Chol, indicating that obesity-linked circRNAs may be key factors that regulate the IL-6 cytokine production in astrocytes. Together with the results that our selected circRNAs might involve the neuronal inflammatory response, neuronal synapse formation, neuronal fatty acid oxidation, and microglial scavengers, our results suggest that obesity-linked circRNAs have pivotal and multifunctional roles in the obese brain. Our obesity-linked circRNA profiling in brain cells might help to unveil obesity-related brain dysfunction and neurodegenerative diseases.

Analysis of the Protein Interaction for Obesity-Related CircRNAs
Some circRNAs have been reported to encode proteins and peptides [36,37]. We analyzed the protein-coding potential of candidate circRNAs to confirm the possibility of protein translation ( Figure 6A). The protein-coding potential of seven circRNAs was confirmed using two prediction tools: CPC 2.0 and CPAT [38,39]. It was confirmed that some circRNAs, except circSnx12 and circUsp3, are highly likely to code proteins. To identify the transcriptional regulatory mechanism of circRNA candidates, we first identified transcription factors that regulate differentially expressed genes in previously reported RNA-seq data [30]. Using the prediction tool ChEA3, we selected the top ten transcription factors involved in the transcriptional control of 459 genes whose p-values were ≤0.05 in the Cuffnorm results ( Figure 6B) [40]. We confirmed the interaction between each circRNA and transcription factor using the RNA-protein interaction prediction tool RPIseq ( Figure 6C-I) [41]. Among various factors, neuronal PAS domain protein 4 (NPAS4), SRYbox transcription factor 8 (SOX8), and retinoid X receptor gamma (RXRG) were shown to be highly likely to interact with circSnx12 and circUsp3, which have low protein-coding potential ( Figure 6G,I). NPAS4 is known to play an important role in contextual memory formation in the CA3 region of the hippocampus as a transcriptional factor [42]. SOX8 has been reported to enhance the astrogenesis of neural stem and precursor cells by targeting Nfia [43]. RXRG deficiency is known to impair spatial memory in mice and reduce mGluRmediated synaptic plasticity [44,45]. Based on these results, the cellular mechanisms behind the interactions between obesity-related circRNAs and transcriptional factors in neurons and glia in the condition of obesity may be elucidated.

Discussion
Herein we comprehensively analyzed specifically or similarly expressed circRNAs in brain cells, such as neurons, microglia, and astrocytes, exposed to diverse obesity-related in vitro conditions. This study is based on our previous report of distinct expression patterns of circRNAs in the brain cortices of obese mice [30]. We profiled significant obesity-related circRNAs that may have regulatory roles in neuropathological mechanisms in neurodegenerative diseases such as AD. From our analysis, we speculate that obesity-related circRNAs are mainly distributed in neurons and astrocytes, suggesting that circRNAs distributed in these cells play regulatory roles in obesity-related brain dysfunction. Moreover, we profiled unique obesity-related circRNAs in our obesity-like in vitro models. Among these circRNAs, we selected those expressed in patterns similar to those of the transcriptomic analysis data from obese mouse brains for further functional analysis [30]. We consider that each cell type in the brain has different intrinsic functions and that circRNAs are likely to have different cell-specific expressions and functions [46].
We selected 11 human-conserved circRNAs that may have pivotal roles in obesityinduced brain dysfunction. Among these circRNAs, the functions in human diseases of four, including circRabgef1, circUsp3, circZzz3, and circAftph, have not yet been identified. Moreover, for circRNAs whose roles have been previously reported, including circSnx12, circDennd1b, and circStx6, their specific functions have not yet been completely understood in the brain. For example, circDennd1b is involved in atherosclerosis by regulating Chol efflux through a miRNA-17-5p sponge [47], but its role in the brain is largely unknown.
We established obesity-like condition models using brain cell lines to mimic the environment of the brains of obese individuals. These models were based on reports on the characteristics of blood serum profiling in high-fat-fed mice, rats, monkeys, and humans with obesity models [48][49][50][51][52][53]. Blood serum factors in obesity are important mediators in the onset of neurodegenerative diseases since the blood-brain barrier (BBB) is disrupted or loosened in obesity conditions, and consequently, blood serum factors can be delivered into brain tissue resulting in memory loss [54][55][56]. Other studies also showed that the blood serum in obese contains higher pro-inflammatory cytokine levels, such as TNF-α and IL-6, leading to neuroinflammation and spatial memory loss [57,58]. Moreover, a high Chol level in obesity damages the metabolic crosstalk between neurons and glia and aggravates synaptic formation [59,60]. Furthermore, an increased level of PA, a saturated fatty acid, impairs autophagy and insulin signaling in neurons in obesity [61]. An elevated PA level accelerates apoptosis and inflammatory responses in obesity by activating glia [62]. Furthermore, the high glucose level in obesity is accompanied by insulin resistance and, subsequently, increased neuronal cell loss and BBB permeability, leading to memory loss [63,64]. Another study showed that excessive linoleic polyunsaturated fatty acid is related to hypothalamic inflammation, thus boosting weight gain [65] and affecting insulin resistance [66]. In obesity, fatty acid composition is an important factor for predicting the progression of lipid dysregulation and inflammation in obesity. Some studies presented that these characteristics of blood serum in obesity are considerably associated with neurodegenerative diseases such as AD [67][68][69][70]. Thus, we selected obesity-related in vitro conditions, including glucose, insulin, TNF-α, IL-6, PA, LA, and Chol, to mimic obesity status.
In the present study, each obesity-related in vitro condition led to alterations in genes related to synaptic function, inflammatory responses, insulin receptor signaling, fatty acid oxidation, scavenger receptor signaling, gluconeogenesis, gliotransmission, and glutamate transporters in neurons and glia. These obesity-related in vitro condition-induced expression changes indicate that metabolic imbalances influence insulin receptor signaling, synaptic function, and inflammatory responses in neurons and inflammatory cytokine secretion, scavenger ability, and glutamate transporters in glia. Several studies mentioned that chronic metabolic disorders, such as obesity and T2DM, modulate the inflammatory response, synaptic dysfunction, fatty acid oxidation, and microglial scavenger function in the brain [71,72]. Obesity impairs brain function and synapse formation by impair-ing insulin signaling [73]. In neurons, insulin receptor signaling-related genes, such as Irs1 and Slc2a4, could affect the secretion of excitatory glutamate neurotransmitters [74], neuroinflammation [75], and memory formation [76]. Moreover, in neurons, synaptic formation-related genes, such as Map2 and Syp, are related to memory consolidation and synaptic plasticity [77].
Therefore, we assume that our obesity-like in vitro conditions triggered neuronal dysfunction related to insulin signaling and synaptic formation and glial dysfunction related to phagocytosis and Chol uptake, leading to cognitive impairment.
In Figures 4 and 5, we observed changes in the expression of several genes in neurons and glial cells after silencing specific circRNAs using siRNAs. In our data, circDennd1b depletion using siRNA significantly regulated genes related to synaptic vesicles and synaptic formation, such as Stx1a and Syp, in neurons [88] under Chol exposure. This result suggests that the role of circDennd1b may be related to synaptic plasticity in neurons under Cholenriched obesity conditions. In particular, we found a significantly abnormal expression of IL-6 in astrocytes after obesity-related circRNA depletion under obesity-related in vitro conditions. Considering that astrocyte-specific IL-6 knockout mice showed increased body weight, and cerebral IL-6 overexpressing mice resisted diet-induced obesity, IL-6 might be a key cytokine associated with astrocyte function by regulating several circRNAs in obesity [89,90]. This result suggests that the modulation of obesity-related circRNAs might affect the expression of IL-6 in astrocytes to regulate neuronal and glial function in the brain of obese individuals.
Even though the effects of our candidate circRNAs on brain function have not been studied until now, their functions in metabolic disorders in neuronal and glial cells have yet to be identified.
In this study, we investigated circRNAs expressed in CNS cells exposed to an obesitylike environment. We profiled candidate circRNAs that show expression changes in the brain cortex of obese mice. In addition, we selected circRNAs that show expression changes under obesity similar to in vitro conditions and are expressed in patterns similar to transcriptomic data from obese mouse brains. Judging from our findings in this study, we hypothesize that each circRNA expressed in CNS cells exhibits cell-specific expression changes and contributes to brain function in the obese brain.
Since a large part of our data was produced based on immortalized or tumoral cell lines derived from the brain cells, it may less reflect the physiology of the actual nervous system. Therefore, further studies are necessary to verify the expression and functions of each circRNA in in vitro obesity cells obtained by isolating primary cortical and hippocampal neurons, microglia and astrocytes from mice. Moreover, an in vivo study is necessary to determine whether the regulation of circRNAs expression in the mouse brain affects cognitive function in animal models.
Even though there are several limitations, our data show the potential of candidate cirRNAs related to neuropathological issues in the brain with obesity. Thus, we suggest that functional studies on circRNAs in CNS cells of obese brain are necessary for an appropriate therapeutic approach to the neuropathological problems of obesity brain.

siRNA Design and Transfection
The siRNAs to suppress circRNA expression were designed as previously reported [97]. siDESIGN Center (https://horizondiscovery.com/en/products/tools/siDESIGN-Center/ accessed on 4 May 2022) and i-Score Designer (https://www.med.nagoya-u.ac.jp/neurogenetics/ i_Score/i_score.html/ accessed on 4 May 2022) were used to identify the siRNAs that targeted the back-splicing junction of circRNAs. These siRNAs and the AccuTarget negative control siRNA were synthesized by Bioneer (Republic of Korea). The sequences of the siRNAs used are listed in Supplementary Table S1.
The siRNAs were transfected using Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer s instructions. siRNAs with a final concentration of 30 nM were transfected into brain cells. The transfected cells were incubated for six hours. The medium was replaced with a growth medium containing the reagent for obesity-like conditions and incubated for 48 h.

RNA Isolation and Semi-Quantitative Polymerase Chain Reaction
RNA was extracted using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions. The RNA quantification was performed using a NanoPhotometer (IMPLEN, München, Germany), and reverse transcription of RNA to complementary DNA (cDNA) was performed using random hexamers and RevertAid reverse transcriptase (Thermo Fisher Scientific, Waltham, MA, USA). A semi-quantitative polymerase chain reaction (PCR) was conducted using nTaq DNA polymerase (Enzynomics, Daejeon, Republic of Korea) in Master cycler Nexus X2 (Eppendorf, Hamburg, Germany). The results of the semi-quantitative PCR were evaluated by electrophoresis using a 2% agarose gel. The gel was analyzed using Image J (V1.53c) provided by the National Institutes of Health (NIH) [98]. The expression of circRNAs and mRNA was normalized against the expression of Gapdh. The primer sequences are listed in Supplementary Table S1.

Confirmation of the Circular Structure of the CircRNAs
To verify the circular structure of the circRNAs, total RNA was treated with RNase R (Biosearch Technologies, Hoddesdon, UK), which only degrades linear RNAs. The total RNA and RNase R mixture was incubated at 37 • C for 5 min. Then RNase R was inactivated by incubating the mixture at 95 • C for 3 min. The RNA was reverse-transcribed into cDNA using random hexamers and RevertAid reverse transcriptase. Finally, a semiquantitative PCR was performed to amplify the circRNAs of interest. The PCR product was electrophoresed on a 2% agarose gel. The sequences of the PCR product were verified using Sanger sequencing (Solgent, Daejeon, Republic of Korea).

Analysis of CircRNA Function
We used the coding potential calculator 2 (CPC 2.0, http://cpc2.gao-lab.org/ accessed on 25 January 2023) and the coding potential assessment tool (CPAT, http://lilab.research. bcm.edu/ accessed on 25 January 2023) for the prediction of coding potential for the selected circRNAs [38,39]. To predict circRNA-interacting proteins, we performed the analysis as previously described [31]. Using the ChEA3 tools [40], we selected the top ten transcription factors that regulate the 459 differentially expressed genes with p-values less than 0.05 analyzed from the cortex of mice fed with a high-fat diet [30]. For each of these transcription factors, we predicted the probability of their interactions with each circRNA using the RPIseq tool [41].

Statistical Analyses
Data are represented as the mean ± standard error of the mean (SEM). The group sample size was typically set to three for our experiments to optimize the efficiency and power of the statistical tests. The normal distribution and similar variance within each comparison group of data were checked before the statistical tests. An unpaired two-tailed t-test with Welch's correction was used to analyze the comparisons between the control and experimental samples. Statistical significance was established when the p-value was less than 0.5.  Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available within the article. Other data related to this study are available on request from the corresponding author.