HIF1α: A Novel Biomarker with Potential Prognostic and Immunotherapy in Pan-cancer

Cancer is a catastrophic disease that seriously affects human health. HIF1α plays an important role in cancer initiation, progression, and prognosis. However, little is known about the specific role of HIF1α in pan-cancer. Therefore, we systematically and comprehensively analyzed HIF1α using GEPIA, HPA, GeneMANIA, STRING, SMPDB, cBioPortal, UALCAN, and TISDB databases and also 33 cancer and normal tissues in TCGA downloaded from the Genome Data Commons (GDC) data portal. Data and statistical analysis were performed using R software v4.0.3. Our results found that there were differences in the mRNA expression levels of HIF1α in human pan-cancer and its corresponding normal tissues. The expression level of HIF1α correlated with tumor stage in LIHC and also significantly correlated with prognosis in LIHC, LUSC, STAD, OV, PAAD, PRAD, THCA, LUAD, MESO, and READ. The small molecule pathways involved in HIF1α include succinate signaling, fumarate, and succinate carcinogenesis-related pathways. The highest mutation frequency of the HIF1α gene in pan-cancer was head and neck cancer, and the HIF1α methylation level in most tumors is significantly reduced. HIF1α was not only associated with immune cell infiltration but also with immune checkpoint genes and immune regulators TMB and MSI. There were currently 5 small molecule drugs targeting HIF1α.


Introduction
As the socioeconomic status and access to health care improve, the disease burden of the population tends to shift epidemiologically: the population appears to have transitioned from contracting primarily communicable diseases to developing primarily noncommunicable diseases [1]. Cancer is a devastating disease among noncommunicable diseases. With an estimated 19.3 million new cancer cases and nearly 10 million cancer deaths worldwide in 2020, the global cancer burden is expected to reach 28.4 million in 2040, a 47% increase from 2020 [2]. Although new technologies and new drugs for the prevention and treatment of cancer are constantly emerging, the prevention and treatment of cancer still cannot meet the needs of the growing number of cancer patients. Therefore, there is still a long way to actively search for specific and sensitive biomarkers for cancer prevention and to develop new drugs.
The study found that HIF1α is an important part of HIF-1 activity, and it consists of four parts: bHLH domain, PAS domain, ODD domain, and transactivation domain. Binding of the ODD domain to pVHL protein under hypoxic level can prevent HIF1α subunit ubiquitination and degradation, thereby increasing the expression level of HIF1α protein, and tumors adapt to the hypoxic environment by expressing high levels of HIF1α protein [23]. To the best of our knowledge, a comprehensive analysis of HIF1α on human pancancer clinical prognosis, immune microenvironment, and HIF1α-targeted drugs using bioinformatics remains largely unknown. Herein, we use bioinformatics to comprehensively and systematically study the role of HIF1α in human pancancer. Our results suggest that the expression of HIF1α is related to tumor prognosis and immune cell infiltration. In addition, our study also provides information on the involvement of HIF1α in signaling pathways and current drugs targeting HIF1α. In summary, these findings provide insights into the growing interest in HIF1α between the diagnosis and treatment of cancer.

Materials and Methods
2.1. GEPIA Database. GEPIA (http://gepia.cancer-pku.cn/) is a web-based tool that provides fast and customizable functionality based on TCGA and GTEx data [24]. In this study, we used the GEPIA database to analyze the expression of HIF1α in tumor tissues and their corresponding normal tissues and displayed them using BodyMap and dot plot, respectively. Subsequently, we also used this database to explore the correlation between HIF1α expression and tumor pathological stage. All of the above use log2(TPM +1) for log scale. In addition, we used the "Survival Plots" module to explore the relationship between HIF1α expression and pan-cancer prognosis.
2.2. HPA Database. HPA (https://www.proteinatlas.org/) database is a large-scale initiative to map the entire human proteome using the integration of antibody-based proteomics and various other omics techniques [25]. In our study, we explored the mRNA expression levels of HIF1α in human cell lines based on the HPA database; the gene expression levels are represented as log2 TPM values.
2.4. STRING Database. STRING (https://cn.string-db.org/) enables the analysis of sources of protein-protein interaction information [27]. This database was also used in our study to explore the protein-protein interaction network of HIF1α.
2.5. SMPDB. SMPDB (https://smpdb.ca/) is a comprehensive, colorful, fully searchable, and highly interactive database for visualizing human metabolism, drug action, drug metabolism, physiological activity, and metabolic disease pathways [28,29]. In this study, we used this database to explore the small molecule pathways involved in HIF1α.
2.7. UALCAN Database. UALCAN (http://ualcan.path.uab .edu/index.html) database enables genomics, bioinformatics, and integrative approaches to understand the molecular basis of cancer [31]. We, in our present work, investigated HIF1α methylation levels in pan-cancer and its corresponding normal tissues based on the UALCAN database. The significance of differences was evaluated using Student's t-test, and p < 0:05 was considered statistically significant.
2.8. TISDB Database. TISDB (http://cis.hku.hk/TISIDB/) database is a user-friendly portal that integrates multiple types of data resources in tumor immunology [32]. In our study, the TISDB database was used for the analysis of drugs targeting HIF1α.

HIF1α Expression Level and Prognosis of Tumor
Patients. The data of 33 types of cancer and normal tissues in the TCGA dataset (https://portal.gdc.com) were downloaded from the Genomic Data Commons (GDC) data portal website, using the univariate Cox regression analysis, and the forest was used to show the p value, HR, and 95% CI of each variable through "forest plot" R package. All the analysis methods and R package were implemented by R version 4.0.3. Two-group data was performed by the Wilcoxon test. p < 0:05 were considered statistically significant.

HIF1α Expression and Immune Cell Infiltration and
Immune Modulator Genes. The data of 33 types of cancer and normal tissues in TCGA were downloaded from the Genomic Data Commons (GDC) data portal website. We obtained immune scores using an R package "Immunedeconv" that integrates two state-of-the-art algorithms, including TIMER and xCell. The Spearman correlation analysis heatmaps of HIF1α gene expression and genes associated with immune scores or immune checkpoints in different types of cancers were generated, the vertical axis represents different immune scores, and different colors represent correlation coefficients. R software v4.0.3 was used for statistical analysis. p < 0:05 were considered statistically significant.

Pan-cancer Analysis of the Correlation between HIF1α
Expression and Immune Regulators TMB and MSI. We obtained TMB and MSI scores from the dataset downloaded from TCGA using R software v4.0.3 for statistical analysis and Spearman correlation analysis of TMB, MSI, and HIF1α gene expression. The abscissa represents the correlation coefficient between genes and TMB, and the ordinate represents different tumors. The size of the dots represents the size of the correlation coefficient, and different colors represent the significance of the p value. The bluer the color, the smaller the p value.

The mRNA Expression Landscape of HIF1α in Human
Pan-cancer. To explore the mRNA expression landscape of HIF1α in human pan-cancer, we comprehensively analyzed the mRNA expression levels of HIF1α in interactive body maps using the GEPIA dataset. We know that, from the overall level of the interactive body map, the median expression levels of HIF1α in most human tumor tissues and their corresponding normal tissues are different, in particular in the brain, blood, lungs, digestive organs (esophagus, pancreas, stomach, and gallbladder), kidneys, thyroid, and other tissues and organs (Figure 1(a)). Based on the previous findings, we, next, studied the mRNA expression levels of HIF1α in 33 tumors and their corresponding normal tissues using the GEPIA database. Unexpectedly, the median level of mRNA expression of HIF1α was high in only 7 tumor tissues (ESCA, GBM, HNSC, LAML, LGG, PAAD, and STAD) compared to normal tissues (Figure 1(b)). Finally, we further     Oxidative Medicine and Cellular Longevity analyzed the mRNA expression level of HIF1α from the cellular level using the HPA database. As a result, we found higher levels of HIF1α mRNA expression in these tissue organ cell lines, including the brain, liver and gallbladder, gastrointestinal pancreas, male reproductive system, kidneys and bladder, skin, eyes, proximal gastrointestinal, lungs, female reproductive system, endothelial, muscle, and mesenchymal lymphoid myeloid ( Figure 1(c)). Regardless of whether tumor tissue is compared with its corresponding normal tissue, or at the level of organ cell lines, the mRNA expression levels of HIF1α are higher in organs like the brain, digestive organs (esophagus, pancreas, stomach, and gallbladder), and lungs.

Correlations between the HIF1α Expression and Tumor
Pathological Stage. As one of the important indicators of patient prognosis, the pathological stage of the tumor should be looked at closely. Hence, we used the GEPIA dataset to analyze the correlation between HIF1α expression level and tumor pathological stage, including 17 tumors, while other tumors could not be shown in the GEPIA database. As a complete surprise, our results showed that HIF1α expression was only significantly correlated with tumor stage in LIHC (p = 0:0356), and there was no correlation between the expression level of HIF1α and the pathological stage of other tumors (p > 0:05) ( Figure 2). Collectively, these results demonstrate that the expression level of HIF1α is associated with the pathological staging of LIHC, which has a certain guiding significance for guiding the pathological staging of this tumor.

The Relationship between HIF1α Expression Level and
Prognosis of Tumor Patients. Based on the findings of HIF1α expression level and tumor pathological stage, we further evaluated the relationship between HIF1α expression level and survival prognosis of pan-cancer patients by the Cox regression analysis. Of note, the prognostic indicators in this study mainly include OS and DFS. The Cox regression analysis of the results from 33 types of cancer suggests that the expression level of HIF1α was significantly associated with OS in LIHC, LUSC, MESO, and STAD patients (p < 0:05) (Figure 3(a)). In addition, we also found that the expression level of HIF1α was significantly associated with DFS in 4 tumors, including OV, PAAD, PRAD, and THCA (p < 0:05) ( Figure 3

Protein-Protein Interaction Network and Small Molecule
Pathway of HIF1α. To explore the protein-protein interaction network of HIF1α, our analysis using GeneMANIA databases found that there are 20 genes associated with HIF1α ( Figure 4(a)). Furthermore, we explored using the STRING database and found that the number of nodes associated with HIF1α is 11 (Figure 4(b)). We used SMPDB analysis to find that HIF1α in human small molecule pathway includes one protein pathway, succinate signaling ( Figure 4(c)), and two disease pathways, the oncogenic action of fumarate and the oncogenic action of succinate (Figures 4(d) and 4(e)).

Analysis of HIF1α Gene Mutation and Methylation Level
in Pan-cancer. To assess the mutation of HIF1α in pan-cancer, we conducted an in-depth study using the cBioPortal database and found that HIF1α was altered in 5% (132/ 2583) of pan-cancer patients ( Figure 5(a)). In addition, we also analyzed the mutation frequency of the HIF1α gene in different types of tumors, and the results showed that the mutation frequency of head and neck cancer (12.5%), lung cancer (10.53%), and pancreatic cancer (10.06%) ranked the top three, respectively. Notably, amplification is the most common type of HIF1α gene mutation ( Figure 5(b)). To better understand the mutational map of HIF1α in different cancer types across protein domains, our investigation found that a total of 16 mutation sites were detected, located between 0 and 826 ( Figure 5(c)). Aberrant DNA methylation is an important cause of cancer [33]. Hence, we next used the UALCAN database to explore the level of HIF1α methylation in pan-cancer and its corresponding tissues. Our results indicate that, compared with normal tissues, HIF1α methylation levels were significantly decreased in BLCA, BRCA, CHOL, CESC, COAD, ESCA, HNSC, KRIC, KIRP, LIHC, LUAD, LUSC, PAAD, READ, SARC, TGCA, and UCEC tissues ( Figure 5(d)).  Oxidative Medicine and Cellular Longevity 3.6. Pan-cancer Analysis of HIF1α Expression and Immune Cell Infiltration. Since there is a certain relationship between HIF1α and immune response, we performed a pan-cancer analysis of the relationship between HIF1α expression and immune infiltration levels based on the TIMER database. The data presented here implicate that 20 cancers were associated with T cell CD8+, 12 cancers were associated with T cell CD4+, 22 cancers were associated with neutrophils, 20 cancers were associated with Myeloid dendritic cells, 22 cancers were associated with macrophage, and 16 cancers were associated with B cells (Figure 6(a)).

Pan-cancer Analysis of the Correlation between HIF1α
Expression and Immune Checkpoint Genes and Immune Regulators TMB and MSI. To estimate the association between HIF1α expression and TME in the pan-cancer dataset, we further investigated the relationship between HIF1α expression and two major types of immune regulators. The vast majority of tumors include BLCA, BRCA, COAD, DLBC, ESCA, LAML, LIHC, LUAD, MESO, PAAD, PCPG, PRAD, READ, SARC, SKCM, STAD, THCA, THYM, UCEC, UCS, and UVM, and immune checkpoint genes are positively correlated. Only a few tumors including TGCT and HNSC were negatively correlated with immune checkpoint genes (Figure 7(a)).
TMB and MSI are two emerging biomarkers that are promising predictive biomarkers for immunotherapy in cancer treatment [34]. We first investigated the relationship between HIF1α expression and TMB. Our results suggest that the expression levels of HIF1α significantly correlated with TMB in COAD, BRCA, and LIHC (Figure 7(b)). In addition, in this report, we provide evidence that the expression level of HIF1α is significantly associated with MSI in some tumors, including COAD and DLBC (Figure 7(c)).

Drugs Targeting
HIF1α. The development of drugs targeting HIF1α is critical for the treatment of cancer patients. Therefore, we used the TISDB database to analyze the current drugs targeting HIF1α. In the present study, we observed that there are 5 small molecule drugs targeting HIF1α, including carvedilol, 2-methoxyestradiol, ENMD-1198, PX-478, and FG-2216 (Figure 8(a)). Among them, carvedilol is the drug with the most targets, with a total of 17 targets. It downregulates HIF1α in the myocardium of volume-overloaded heart failure [35]. Remarkably, compared with carvedilol, the other 4 small molecule drugs involve relatively few targets. Of course, Table 1 presents more details on drugs targeting HIF1α.

Discussion
HIF is a central regulator for detecting and adapting to cellular oxygen levels, and it regulates oxygen homeostasis and metabolically activated genes through transcriptional activation. In addition to this, HIF affects many other

10
Oxidative Medicine and Cellular Longevity processes including cancer development [9]. Current research suggests that HIF1α, as one of the most important members of the HIF family, is closely related to the occurrence, development, and prognosis of cancers [38][39][40][41][42]. Thus, we explored the role of HIF1α in human tumors by using bioinformatics methods. We first comprehensively analyze the expression of interacting BodyMap HIF1α in human tumors and their corresponding normal tissues. Second, we provide evidence that HIF1α expression levels are associated with the pathological staging of LIHC and that HIF1α expression levels are associated with prognosis in 11 tumors including LIHC, LUSC, MESO, STAD, OV, PAAD, PRAD, THCA, LUAD, MESO, and READ. In addition, we also found that HIF1α in human small molecule pathways includes 3 pathways (succinate signaling pathway, fumarate carcinogenesis, and succinate carcinogenesis), and HIF1α methylation levels are significantly reduced in most tumors. In addition to the above findings, we further discovered the relationship between HIF1α and immune cell infiltration in the cancer microenvironment and small molecule drugs targeting HIF1α.
In this study, our data suggest that the mRNA expression level of HIF1α was higher in organs such as the brain tumors, gastrointestinal tumors (esophagus, pancreas, stomach, and gallbladder), and lung tumors compared with normal tissues, both at the tissue level and the cell line level. Because we all know that the brain, as an advanced nerve center, needs a lot of oxygen and energy to maintain its  CHOL  COAD  DLBC  GBM  HNSC  KIRC  KIRP  LIHC  LGG  LUAD  OV  PCPG  PRAD  SARC  STAD  THCA  TGCT  THYM  UCS  UCEC  UVM  SKCM  READ  PAAD  MESO  LUSC  KICH  ESCA  CESC  BRCA  BLCA

11
Oxidative Medicine and Cellular Longevity normal function, the brain is most sensitive to hypoxia [43]. The stomach is an important digestive tract organ, and it is prone to stress-induced gastric ulcer during stress response [44], which is due to the sensitivity of gastric mucosa to ischemia and hypoxia. In addition to the aforementioned sensitivity to hypoxia, tumors grow faster and require more nutrients. Thus, cells like pancreatic cancer [45], liver cancer [46], and lung cancer [47] can activate the transcription of many genes, including those involved in energy metabolism, angiogenesis, and other protein products, by producing HIF1α to increase oxygen delivery or promote metabolic adaptation to hypoxia.
Next, we, on the one hand, found that the expression level of HIF1α was significantly correlated only with LIHC pathological stage, but not with other tumor stages. This finding may make HIF1α more clinically valuable as a new pathological staging marker for LIHC, but it would be interesting to further clarify the mechanism by which HIF1α is only associated with LIHC pathological staging. On the other hand, we provide evidence that, in pancancer, HIF1α expression levels were associated with LIHC, LUSC, MESO, STAD, OV, PAAD, PRAD, THCA, LUAD, and READ prognosis. Consistent with our findings, HIF1α expression levels are associated with poor prognosis in READ [48], LIHC [49], LUSC [50], MESO [51], STAD [52], OV [53], PAAD [41], PRAD [54], THCA [55], and LUAD [56]. How does HIF1α affect the prognosis of cancer patients? A few reports in the literature have documented that hypoxia is common in tumors. HIFA activity in hypoxic areas within often tumors mediates angiogenesis, epithelial-mesenchymal transition, stem cell maintenance, invasion, metastasis, and resistance to radiotherapy and chemotherapy [20,57,58].
We explored the protein-protein interaction network of HIF1α and found 20 genes associated with HIF1α. These HIF1α-related genes all play a role in cancer. For example, a study has found that HIF1α promotes EPO expression at the transcriptional level under hypoxia [59] and achieves antitumor effects by regulating Epo-activated signaling pathways as interfering with the cell cycle of brain tumors [60]. In addition, the HIF1α-related gene ENO1 can bind and degrade the expression of the hepcidin gene, thereby regulating the metabolic homeostasis of intracellular iron ions, affecting ferroptosis, and promoting the occurrence and development of liver cancer [61]. In conclusion, our discovery of HIF1α-related genes provides more new insights for diagnosis and treatment in pan-cancer. In addition, we also used SMPDB analysis to find that HIF1α in human small molecule pathways includes a protein pathway, the succinate signaling pathway, and two disease pathways, fumarate carcinogenesis and succinate carcinogenesis. Previous studies found that hydroxylase activity was inhibited in the presence of low concentrations of O 2 , high concentrations of   tricarboxylic acid cycle intermediates (isocitrate, oxaloacetate, succinate, or fumarate), or chelating agents for Fe (II). The receptor for activated C-kinase 1 competes with heat shock protein 90 for binding to HIF-1α and mediates O 2dependent ubiquitination and proteasomal degradation [62]. Genetic errors in cancer cells reveal that fundamental biological processes required for cancer to develop and develop go awry. It became clear that cancer genomes contain many additional "passenger" mutations in the process. Driver and passenger DNA mutation patterns derived from cancer genomes provide clues to the different ways cancers manifest as genetically mutated diseases [63]. Interestingly, our data suggest that HIF1α has a mutation rate of 5% in pan-cancer. Therefore, in tumor diagnosis and treatment, some studies have developed HIF1α pharmacogenomic mutation models to study individual changes in the effects of tumor hypoxia drugs [64], which will guide precise treatment. Indeed, DNA methylation analysis is an emerging tool as an aid to improve the accuracy of pathological diagnosis; DNA methylation patterns in circulating tumor DNA hold great promise for minimally invasive cancer detection and classification [65]. We provide preliminary evidence that,  13 Oxidative Medicine and Cellular Longevity according to the UALCAN database, HIF1α DNA methylation levels are reduced in 17 tumors. A study has confirmed that DNA hypomethylation activates gene transcription and increases tumor proliferation, migration, and metastasis [66]. Therefore, DNA hypomethylation predicts that these tumors tend to have a poor prognosis. In addition, a previous study demonstrated the involvement of HIF1α in the carcinogenesis of fumarate and succinate [67].
Enhancing immune cell function in tumors remains a major challenge in cancer immunotherapy. Hypoxia is a common feature of solid tumors, and cells adapt by upregulating the transcription factor HIF1α [40]. Our results indicate that most cancers are not only related to T cell CD8+, T cell CD4+, neutrophils, myeloid dendritic cells, macrophages, and B cells. And there is a certain relationship with the subtype of immune cells. For instance, our results suggest that, among 38 immune cell subtypes, HIF1α expression was significantly negatively correlated with these subtypes of ACC, CESC, HNSC, KIRP, LUSC, TGCT, THYM, and UCEC, whereas HIF1α expression was significantly correlated with these subtypes of COAD, KICH, LAML, and LGG. There was a significant positive correlation between subtypes. Among them, the expressions of T cell CD4+ central memory, T cell CD4+ Th1, and HIF1α have the strongest negative correlation in various cancers. Of course, a set of previous studies have also demonstrated that HIF1α expression is associated with infiltrating T cells and macrophages [68], Treg [69], and B cells [70].
Given the above findings, we next analyzed the relationship between HIF1α and immunomodulators and found that most tumors were positively associated with immunechecking genes. Only a few tumors, such as TGCT and HNSC, were inversely associated with immune check genes. In addition, we also explored the relationship between HIF1α and TMB and MSI and found that the expression level of HIF1α was significantly correlated with TMB in COAD, BRCA, and LIHC, and the expression level of HIF1α was correlated with MSI in COAD and DLBC. Of course, in addition to our study, a previous study confirmed that HIF1α expresses a new marker that separates the MSI-L group from the MSS and MSI-H groups [71]. In short, regardless of our findings, or those of previous studies, HIF1α is implicated in immune checkpoint genes, TMB, and MSI.
Finally, our study also investigated 5 small molecule drugs targeting HIF1α. Of these, carvedilol, a drug commonly used to treat high blood pressure, has recently been shown to protect the body from sunlight-induced cell damage and skin cancer [72]. Although the cancer preventive activity of carvedilol is independent of β-blockers, we envision whether the anticancer activity of carvedilol is related to HIF1α, which is a major topic for further research in the future.
In conclusion, in this study, our findings highly demonstrate the expression landscape of HIF1α in human pancancer and identify the relationship between HIF1α expression levels and tumor immune infiltration and HIF1α-targeting drugs. This may provide a new insight into the use of HIF1α to diagnose and treat human pan-cancer.

Conclusion
HIF1α plays an important role in pan-cancer prognosis and immunotherapy, and it may be a novel biomarker with potential prognostic and immunotherapy roles in pancancer.