Methylation Drives SLC2A1 Transcription and Ferroptosis Process Decreasing Autophagy Pressure in Colon Cancer

Colon cancer is a common malignant tumor in the digestive tract, with relatively high rates of morbidity and mortality. It is the third most common type of tumor in the world. The effective treatment of advanced colon cancer is limited, so it is particularly important to study the new pathogenesis of colon cancer. Ferroptosis is a nonapoptotic regulated cell death mode driven by iron-dependent lipid peroxidation, a process which has been discovered in recent years. Autophagy involves lysosomal degradation pathways that promote or prevent cell death. High levels of autophagy are associated with ferroptosis, but a clear association has not yet been made between ferroptosis and autophagy in colon cancer. Through the analysis of transcriptome expression profiling data in colon cancer, we obtained the common upregulated genes and downregulated genes by recording the intersection of the differentially expressed genes in each dataset. Solute Carrier Family 2 Member 1 (SLC2A1) was identified by combining autophagy genes obtained from GeneCards and ferroptosis genes obtained from FerrDb. In order to explore the clinical significance and prognostic value of SLC2A1, we utilized massive databases to conduct an in-depth exploration of the methylation of SLC2A1, and we also investigated the differences in immune infiltration between tumor and normal tissues. We found that there are abundant methylation sites in SLC2A1 and that the methylation of SLC2A1 is correlated with the immunosuppression of tumor tissue. We discovered that during the induction of environmental factors, the transcription and methylation levels of SLC2A1 were greatly increased, autophagy and ferroptosis were inhibited, and the immune system was defective, resulting in a poor prognosis for patients. These results suggest that the autophagy and ferroptosis-related gene SLC2A1 is involved in the tumor immune regulation of colon cancer, and SLC2A1 may become a new therapeutic target for colon cancer.


Introduction
Cancer is a chronic disease characterized by abnormal cell growth that gradually spreads and destroys normal body tissues as the cancer progresses [1]. It is one of the most prominent life-threatening diseases worldwide. According to the latest estimates from the International Agency for Research on Cancer (IARC), new cancer cases will continue to increase. Colon cancer is the third most common malignant tumor in the world, the second most common cancer among females, and the third most common cancer among males, with 1.2 million new cases and more than 600,000 deaths. Survival rates have improved over the past three decades due to early detection, but the 5-year survival rate for patients diagnosed at a later stage of the disease is still only 14% [2][3][4][5]. Colon cancer remains a relatively difficult disease to treat, although advanced treatments are available which can improve survival rates for the disease. erefore, it is particularly important to conduct an in-depth exploration of the pathogenesis of colon cancer and new therapeutic targets.
Ferroptosis is a nonapoptotic regulated cell death mode driven by iron-dependent lipid peroxidation, a process which has been discovered in recent years. Ferroptosis has been reported to be involved in the pathogenesis of colon cancer [6,7]. e induction of ferroptosis is accompanied by an increase in intracellular lipids and reactive oxygen species (ROS), leading to cell death, which is why lipid antioxidants can inhibit ferroptosis. Since iron is an essential ion in the mitochondrial oxidative respiratory chain, mitochondria, as organelles that are rich in iron ions and which generate ROS, are considered to be an important site for ferroptosis. Not only are they an important site for intracellular ROS generation, but fatty acid metabolism provides specific lipid precursors for cellular ferroptosis. Autophagy, which refers to any intracellular process involving lysosome degradation of cytosolic components, also occurs during tumor progression [8]. Autophagy is essential for cell survival, differentiation and development, and homeostasis [9]. Autophagy is involved in a variety of diseases, including infection, cancer, neurodegeneration, and ageing, as well as heart, liver, and kidney disease [10]. When ferroptosis occurs, cells will face additional pressures such as hypoxia, oxidative stress, and oncogene activation, which may induce autophagy. In mitochondria, ferroptosis results in decreased mitochondrial volume and cristae and increased membrane density. is can also cause cells to erroneously and selectively phagocytose mitochondria. In conclusion, there may be an association between ferroptosis and autophagy in colon cancer, and the mechanism of their interaction has not yet been explored.
In this study, we attempted to explore the mechanism of the interaction between ferroptosis and autophagy in colon cancer. By analyzing three different transcriptome datasets of colon cancer patients, we obtained the common upregulated genes and downregulated genes by taking the intersection of differentially expressed genes. e gene SLC2A1 was identified by combining autophagy genes obtained from GeneCards and ferroptosis genes obtained from FerrDb. To explore the clinical significance and prognostic value of SLC2A1, we used massive databases to explore in depth the methylation of SLC2A1, and we investigated the differences in immune infiltration between tumor and normal tissues using the gene expression omnibus (GEO) dataset. We found that there are a lot of methylation sites in SLC2A1 and that the methylation of SLC2A1 is correlated with the immunosuppression of tumor tissue. Our results suggest that SLC2A1 further modulates the immune microenvironment by regulating ferroptosis and autophagy processes in colon cancer, thereby affecting disease prognosis.

Data Acquisition.
GEO is the world's largest public database containing genetic data on various types of diseases [11]. We downloaded the expression profile data of GSE10972, GSE23878, and GSE113513 for colon cancer. Data processing was performed using R software to obtain differential genes.

Information of Patients.
We downloaded the data from the Cancer Genome Atlas (TCGA) for various analyses, including the differential expression analysis of unpaired samples, the differential expression analysis of paired samples, construction of the survival curve and receiver operating characteristic (ROC) curve , and the correlation analysis of the DNA methylation probe and expression level. e differential expression analysis of unpaired samples and the data of the ROC curve were all derived from the RNAseq data in Transcripts Per Kilobase of exon model per million mapped reads (TPM) format from the TCGA and the Genotype-Tissue Expression (GTEx) databases, processed by UCSC Xena (https://xenabrowser.net/datapages/) through the Toil process. We extracted the corresponding normal tissue data from colon adenocarcinoma (COAD) and GTEx of TCGA, and then the log2 normalization was used to transform the RNAseq data in TPM format. e paired samples and survival curve data were derived from RNAseq data in the level 3 HTSeq-Fragments Per Kilobase per Million (FPKM) format in the TCGA (https://portal.gdc.cancer.gov/) COAD portal. We converted the RNAseq data in FPKM format into TPM format and performed a log2 normalization. ] (for statistical analysis of survival data) R packages. e types of prognosis included overall survival (OS) and progressionfree interval (PFI). Significance mark: ns, p ≥ 0.05; * p < 0.05; * * p < 0.01; * * * p < 0.001.

PPI Analysis.
We used the STRING (https://cn.stringdb.org/) website, which contains data of many species [12] relating to protein-protein interaction networks and functional enrichment analysis. We entered the details of the intersection genes, obtained the protein interaction network, and downloaded the protein interaction data for subsequent analysis. We used the GeneMANIA (https://genemania.org/) website to help us predict the biological functions of specific genes and gene sets [13]. Cytoscape is a software application that visualizes molecular interaction networks and biological pathways and integrates these networks with annotations, gene expression profiles, and other status data [14]. In Cytoscape (https://cytoscape.org/), we imported the protein-protein interaction networks of 186 genes obtained from STRING to remove the unconnected proteins and then used the cytoHubba plug-in to calculate and sequence the MCCs.

Harvesting of Functional Gene Lists: FerrDb and
GeneCards. GeneCards (https://www.genecards.org/) is a comprehensive database comprising information about genes, including details of genomes, transcriptomes, and proteins, as well as genetic, clinical, and functional aspects of gene-centric data. We downloaded 6866 genes related to autophagy using "autophagy" as the keyword. e autophagy gene was intersected with the differential gene in GEO.
FerrDb (https://www.zhounan.org/ferrdb/legacy/index. html) was the first comprehensive database containing the genes proven to be related to ferroptosis; it also includes a number of monomers that can regulate ferroptosis [15]. We obtained 260 genes related to ferroptosis, including driver, marker, and suppressor genes. e EWAS (https://ngdc.cncb.ac.cn/ewas/ datahub) database contains a large amount of data of disease types and was therefore convenient for helping us to visualize methylation levels [16]. We were able to determine the relationship between the methylation status of SLC2A1 in COAD and the methylation level and prognosis.
2.6.2. UALCAN. UALCAN (https://ualcan.path.uab.edu/ index.html) is a comprehensive, user-friendly tumor database that provides users with a convenient way to explore TCGA, the National Cancer Institute's Clinical Proteomic Tumor Analysis Consortium (CPTAC), and other datasets [17]. From this database, we obtained methylation data (from TCGA) and protein data (from CPTAC) for SLC2A1.

e Human Protein Atlas (HPA).
is is a unique website dedicated to providing immunohistochemical images, including those of many kinds of tumors and corresponding normal tissues; it proved useful to us for exploring the protein content of genes (https://www.proteinatlas.org/) [18]. We obtained high-resolution immunohistochemical images of normal tissues and tumor tissues from the website to determine the abundance of protein.

CIBERSORTx.
is was used to estimate the gene expression profile and provide the estimated abundance of member cell types in the mixed cell population by using the gene expression data (https://cibersortx.stanford.edu/) [19]. We cleaned the original expression data from GEO, with the requirements that the expression values do not undergo log conversion and cannot have null values.
2.6.5. EcoTyper. EcoTyper (https://ecotyper.stanford.edu/) is a new machine learning framework for the identification of cell states and ecosystems from bulk, single-cell, and spatially-resolved expression data [20]. EcoTyper extends CIBERSORTx for large-scale profiling of cellular ecosystems.
2.6.6. xCell. xCell (https://xcell.ucsf.edu/) is a webtool that performs cell type enrichment analysis from gene expression data for 64 immune and stroma cell types [21]. xCell is a gene signature-based method learned from thousands of pure cell types from various sources. In order to explore the tumor heterogeneity of colon cancer, we inferred the types of sample cells from the expression profile in xCell.

MethSurv.
is is a web tool for performing multivariate survival analysis using DNA methylation data (https://biit.cs.ut.ee/methsurv/) [23]. We used this website to explore the relationship between methylation levels of different methylation probes and prognosis.

JASPAR.
is database contains a large amount of transcription factor data, which can help researchers to predict the binding sites of transcription factors and their sequences (https://jaspar.genereg.net/) [25].

Excellent Sample Data Quality with High Confidence.
After downloading the expression profile data, it is essential to assess the quality of the information [26]. Principal component analysis (PCA) maps show the features of highlatitude data by extracting their feature vectors, which are then converted into low-dimensional data, and using twodimensional maps. e distance between samples indicates the differences between them. e figure showed that the characteristics of samples in the same group are similar, and the differences in samples between different groups are significant (Figures 1(a

Screening Genes to Obtain the Key Gene Associated with
Ferroptosis and Autophagy. We crossed the differential genes analyzed from the three expression profiles with the autophagy genes obtained from GeneCards to obtain a gene set containing 186 genes. en we input the gene set into STRING, and a complex and comprehensive protein-protein interaction network diagram was produced (Figure 2(a)). After the unconjugated proteins were removed, the protein interaction data were imported into Cytoscape to calculate the Maximal Clique Centrality (MCC) value, and 156 genes were found to be involved in this interaction network. We selected 55 genes with an MCC value greater than 10 and then used the clinical data from TCGA to explore their survival significance; from this, we obtained 19 genes with clinical significance. Next, we intersected the ferroptosis gene set obtained from FerrDb with the 19 genes and finally obtained the SLC2A1 gene (Figures 2(b)-2(c)). In previous studies, researchers used all-trans retinoic acid (ATRA) to treat different lung cancer cells and found that the most significant reduction in the transcription level of SLC2A1 was found in SLC2A1 and the genes in the same family, thus excluding other gene families [5]. erefore, we determined that the SLC2A1 gene not only plays an important role in ferroptosis but also plays an indispensable role in autophagy in colon cancer. In order to explore the clinical relevance of SLC2A1 expression in colon cancer, clinical data from TCGA were downloaded for analysis and SLC2A1 was found to have high predictive power for the variable SLC2A1 in predicting the outcomes of Normal and Tumor (AUC � 0.968, CI � 0.957-0.979, Figure 2(d)). In the main treatment outcome, PD, PR, and CR were all highly correlated with the expression level of SLC2A1 and all of them were statistically significant (Figure 2(e)). Regarding pathological stages, the expression of SLC2A1 increased as the stages progressed (Figure 2(f )). In stages M and N of the tumor, the expression level of SLC2A1 increased slowly as the tumor progressed, and it could be seen that the high expression of SLC2A1 seriously affected the deterioration of the tumor (Figures 2(i)-2(j)). Moreover, SLC2A1 expression levels in patients with lymphoid invasion were higher than those in patients without lymphoid invasion, and SLC2A1 expression levels in patients with perineural invasion were higher than those in patients without perineural invasion (Figures 2(k)-2(l)). More importantly, in terms of PFI, SLC2A1 expression levels were higher in deceased patients than in surviving ones (Figures 2(m)-2(n)). e KM prognostic curve also illustrates the excellent prognostic value of SLC2A1 in both PFI and OS prognosis types (Figures 2(o)-2(p)). In addition, both high SLC2A1 expression groups had poor prognostic outcomes in the presence of residual tumor and lymphoid invasion (Figures 2(q)-2(r)). We also performed univariate and multivariate COX regression using TCGA data and plotted the baseline data table (Supplementary-Tables 1-2, Supplementary Figure 1C).  (Figures 3(a)-3(b)). In addition, we explored the   Journal of Oncology protein expression of SLC2A1 in the HPA and CPTAC databases, and we found that the protein abundance of SLC2A1 in tumors was much higher than that in normal tissues (Figure 3(c)-3(g)).

Multidatabase Joint Exploration of the Methylation
Level of SLC2A1. N6-methyladenosine (m6A) is a post-transcriptional methylation modification that is widely distributed on the adenosine bases of RNA transcripts. is modifier is involved in the degradation of RNA transcripts, subcellular localization, and the regulation of splicing and local variations. In mammalian transcriptomes, only a small fraction of this motif is indeed modified, determining the other sequential and structural features of the m6A modification site. Results from both the UALCAN and EWAS databases revealed higher methylation levels of SLC2A1 in tumors and worse case outcomes in the hypermethylated group (Figures 4(a)-4(c)). After exploring the correlation between the methylation levels of SLC2A1 methylation probes and the transcription level of SLC2A1 in TCGA, we found that a total of five probes were correlated with the transcription level, with probes cg09824328, cg12656391, and cg00102166 being negatively correlated, and probe  Journal of Oncology cg22176566 being positively correlated (Figures 4(d)-4(l)). After the sequence of SLC2A1 was downloaded from NCBI and the methylation sites of m6a were predicted using the SRAMP website, we found that there were a large number of methylation sites on the RNA of SLC2A1, and a total of 117 sites were obtained, including 16 sites with very high reliability. We also showed the RNA secondary structure of these 16 sites (Supplementary Table 3, Supplementary  Figure 2). In MEXPRESS, we found that the previously mentioned probes exhibiting a negative correlation were validated against SLC2A1 expression levels. In addition, the expression level of SLC2A1 has a certain correlation with sample type, tumor stage, venous invasion, and lymphatic invasion. e heatmap also indicated that cg00102166 and cg12656391 showed obvious hypomethylation. At the same time, the methylation status of four methylation probes had considerable prognostic significance for survival. e group with low methylation of cg04287330 and cg13790786 had significantly poor prognostic outcomes, while the group with high methylation of cg20294984 and cg22025263 had even worse prognostic outcomes (Supplementary Figure 3).

Transcription Level of SLC2A1 Is Associated with Immune Infiltration in Tumors.
In order to explore the relationship between the transcription level of SLC2A1 and immune infiltration, we downloaded the RNAseq data from TCGA to study the infiltration level of multiple immune

cells, including aDCs [activated dendritic cells], B cells, CD8 T cells, cytotoxic cells, DCs, eosinophils, iDCs [immature DCs], macrophages, mast cells, neutrophils, NK CD56 bright cells, NK CD56dim cells, NK cells, pDCs [plasmacytoid DCs], T cells, T helper cells, Tcm [T central memory] cells, Tem [T effector memory] cells, Tfh [T follicular helper] cells, Tgd [T gamma delta] cells, 1 cells, 17 cells, 2 cells, and Treg cells.
In the data of TCGA, we found that the expression level of SLC2A1 was correlated with some of these immune cells, and there was a positive correlation with NK CD56 bright cells, eosinophils, mast cells, NK cells, neutrophils, DCs, iDCs, Tem cells, and NK CD56dim cells. However, it was negatively correlated with Tcm, 2, and helper cells (Figure 5(a)). e association of the high transcription level of SLC2A1 with the infiltration of a number of immune cells includes DCs, eosinophils, iDCs, mast cells, neutrophils, NK CD56bright cells, NK cells, and T helper cells (Figure 5(b)). At the same time, the correlation between the infiltration fraction of different immune cells and the transcription level of SLC2A1 can also be observed from the scatter diagram (Figures 5(c)-5(h)). We also performed a correlation analysis (Supplementary Table 4) for the key marker of each immune cell.
In order to systematically identify the cell states and cellular communities (ecotypes) in colon cancer, we used the EcoTyper tool to perform the analysis, inputting data from GSE10972, GSE23878, and GSE113513, which meet the stringent data requirements of CIBERSORTx, entered for analysis (Supplementary Figure 4). Tissue is a complex environment composed of numerous cell types.
Understanding cellular heterogeneity in the tumor microenvironment is an emerging field of research in cancer. We used xCell to explore the tumor microenvironment in COAD using three datasets from GEO. e three heat maps reveal some interesting observations. Among the 64 types of immune cells, a small number of them were found to have a low infiltration fraction in normal tissue but a significantly increased infiltration fraction in tumor tissue. However, most types of immune cells revealed a high infiltration fraction in normal tissue but a significantly decreased infiltration fraction in tumor tissue, a phenomenon similar to "depletion" (Figures 5(i)-5(k)), such as typical T-cell depletion. Various T-cell subtypes are such as CD4+ Tcm; CD4+ T cells; and CD4+ naïve T cells such as CD8+Tcm. In addition, we found that M1 and M2 macrophages, various B cell subtypes, and different subtypes of endothelial cells all share the same cell state as T cells. After initial T cells are activated by antigens, costimulation and inflammation, they proliferate exponentially and differentiate into effector T cells and memory T cells. At the same time, we can also observe that the total immune infiltration of the tumor tissue and the sum of the scores of all types of immune cells are much lower than those of the normal tissue, indicating that the patient's immune response was decreased, and it even promoted the immune escape of the tumor cells (Supplementary Figure 5).

Discussion
Colon cancer is the third most common malignant tumor in the world. Both ferroptosis and autophagy are thought to be involved in the development of colon cancer, but the relationship between them has not been revealed. To explore the mechanisms underlying the association between ferroptosis and autophagy in colon cancer, we downloaded three high-quality colon cancer datasets from the GEO database and analyzed differentially expressed genes. We then intersected the differentially expressed genes with autophagy-related genes to obtain a gene set containing 186 genes. A protein-protein interaction network was constructed for this gene set, and genes with low correlation were eliminated. At the same time, survival analysis was performed using TCGA data, with 19 genes with prognostic significance being considered as key genes. Subsequently, we found the intersection between ferroptosis-related genes and key genes to obtain SLC2A1 as a potential marker gene. e expression level of SLC2A1 itself is associated with the prognosis of colon cancer patients. SLC2A1 has a large number of methylation sites, and the methylation of SLC2A1 is related to T cell exhaustion and immunosuppression in tumor tissue. SLC2A1 encodes the major glucose transporter in the mammalian blood-brain barrier. e encoded protein is mainly present in the cell membrane and cell surface and promotes the glucose transporter responsible for constitutive or basal glucose uptake [27][28][29][30]. SLC2A1 can affect functions such as carbohydrate homeostasis and carbohydrate kinase activity. e increase in SLC2A1 expression may stimulate cellular glycolysis, and our analysis also  suggests that the expression of SLC2A1 may contribute to the Warburg effect and promote tumor cell metastasis (Supplementary- Figure 6A). Previous studies have shown that high transcription levels of SLC2A1 may lead to lymphoid and perineural invasion [31], suggesting that SLC2A1 plays a conserved function in a variety of malignancies.
By further investigating the mechanism by which SLC2A1 affects prognosis, we learned that SLC2A1 is closely related to HK family proteins [32]; therefore, we explored the epigenetic modification of SLC2A1, which has been shown in past studies to be hypermethylated with high levels of transcription, and the protein level is high in colon cancer  Figure 5: e immune infiltration of COAD and SLC2A1 was investigated according to the combined data of TCGA and GEO. (a-h) Correlation between the expression level of SLC2A1 in the TCGA database and the infiltration of different immune cells. (i-k) Immunoinfiltration analysis in xCell using expression profiles from three GEO datasets. [33]. We predicted the methylation sites of SLC2A1 and found that the SLC2A1 sequence contained a large number of m6a modification sites (with 16 very high-confidence sites); methylation sites were also revealed in the RNA secondary structure. e correlation between SLC2A1 and the expression levels of several methylase and autophagyrelated proteins was also explored, and SLC2A1 was found to be associated with these proteins (Supplementary Figure 6B). At the same time, we studied the correlation between the expression levels of several probes, which also proved that the overall DNA methylation level of SLC2A1 and the methylation levels of these probes were significantly associated with prognosis.
According to the literature, when SLC2A1 is significantly activated and transcribed by the DNA methylation modifier, lymphoid-specific helicase will significantly inhibit ferroptosis, which also leads to significantly reduced lipid peroxidation and ROS [34]. However, after the deacetylase Sirtuin 2 (SIRT2) is inactivated by ROS, acetylated Forkhead Box O1 (FoxO1-Ac) accumulates in the cytoplasm and complexes with Autophagy Related 7 (ATG7), which can stimulate autophagy. is autophagyinducing mechanism keeps FoxO1 inactive as a transcription factor in the cytoplasm but still inhibits the growth of tumor xenografts in nude mice [35,36]. Our results suggest that methylation of SLC2A1 is associated with immunosuppression in colon cancer. Chronically high expression of SLC2A1 may lead to prolonged antigen exposure of immune cells and inflammation. Effector T cells are continuously stimulated, transforming them into exhausted T cells; T cell exhaustion is one of the major factors in immune dysfunction in cancer patients.
is leads to immune escape and tumor progression. ese results suggest that the expression and hypermethylation of SLC2A1 affects the processes of ferroptosis and autophagy, which in turn affects tumor immunity.
Our study substantiates that the ferroptosis and autophagy-co-associated gene SLC2A1 is involved in tumor immune regulation and tumor prognosis in colon cancer. Hypermethylation of SLC2A1 is associated with T cell exhaustion and suppression of the immune microenvironment in colon cancer, a process which is most likely mediated by ferroptosis and autophagy. However, further research is required to determine the specific biological mechanism for this. e limitation of this study is that the data were obtained using exclusively bioinformatic methods; thus, there is a lack of experimental data. In conclusion, our study demonstrates that SLC2A1 is involved in the regulation of ferroptosis and autophagy in colon cancer, which in turn regulates tumor immunity and affects tumor progression. Our results also indicate that SLC2A1 may become an effective therapeutic target for colon cancer in the future.
Data Availability e data of this study are available from the public databases.

Ethical Approval
e Research Ethics Committee has confirmed that no ethical approval is required.

Conflicts of Interest
e authors have no conflicts of interest.

Supplementary Materials
Supplementary Figure 1. Intersection of differential genes and autophagy genes. A: there were 113 intersecting genes of common downregulating genes and autophagy genes. B: there were 73 intersecting genes of coupregulated genes and autophagy genes. C: forest map is shown. Supplementary  Figure 2. e m6a site on the SLC2A1 sequence and its secondary RNA structure are predicted. A: distribution map of m6a prediction sites. B-Q: secondary RNA structure with very high-confidence methylation sites. Supplementary