Decreased DIO3OS Expression Predicts Poor Prognosis in Hepatocellular Carcinoma and is Associated with Immune Infiltration

Hepatocellular carcinoma has become one of the most shared cancers in the whole world because of its high morbidity, poor survival rate, and low recovery rate. LncRNA DIO3 opposite strand upstream RNA (DIO3OS) has been reported to be obviously important in several human cancers, while its biological function in hepatocellular carcinoma (HCC) remains unclear. Here, DIO3OS gene expression data and clinical information of HCC patients were extracted from the Cancer Genome Atlas (TCGA) database and the university of California Santa Cruz (UCSC) Xena database. In our study, the Wilcoxon rank sum test was used to compare DIO3OS expression between healthy individuals and HCC patients. It was found that patients with HCC had significantly lower DIO3OS expression than healthy individuals. Furthermore, Kaplan–Meier curves and Cox regression analysis showed that high DIO3OS expression tended to predict better prognosis and higher survival rate in HCC patients. In addition, the gene set enrichment analysis (GSEA) assay was used to annotate the biological function of DIO3OS. It was found that DIO3OS was significantly correlated with immune invasion in HCC. This was also aided by the subsequent ESTIMATE assay. Our study provides a novel biomarker and therapeutic strategy for patients with hepatocellular carcinoma.


Introduction
Malignant neoplasms have become the main cause of death in twenty-first centuries. Despite diagnosis and strategies have been improved, cancer mobility and mortality continue to increase over the past decades. Liver hepatocellular carcinoma (LIHC) is the third most shared cause of cancer-related death in the world (El-Serag 2011). Over the last few years, global clinical data has showed that hepatocellular carcinoma (HCC) patients are the most common in LIHC patients (El-Serag and Rudolph 2007). Currently, the main diagnostic methods for HCC included serum marker detection, CT, MRI, and pathological biopsy, but these methods lack sensitivity and accuracy. Due to the lack of effective biomarker screening and early prevention, most patients are already in advanced stage at the time of diagnosis. According to the above reasons, investigating more sensitive and specific biomarkers for HCC patients is a crucial task.
Previous reports suggested that lncRNA was defined as transcripts with more than 200 nucleotides and had no obvious protein coding function (Sanchez Calle et al. 2018). Eukaryotic transcriptome transcripts produce complex transcriptional networks, including tens of thousands of lncRNAs, which have little protein coding function. Previously, scholars took the attitude that lncRNA was only the "noise" generated during transcription, and had no practical effect on diseases (Struhl 2007). However, more and more examines have proven that lncRNA can play a role in a variety of ways, and affect nearby or distant genes expression, which was a crucial regulatory molecule in cells. Previous studies have pointed out that the expression level or mutation of lncRNA gene is closely related to tumor proliferation, migration, and invasion. LncRNA can promote chromatin remodeling, regulate protein activity, regulate epigenetics, chromatin modification, and so on (Wilusz et al. 2009). Thus, lncRNA is a crucial therapeutic target for diseases.
DIO3OS located on chromosome c14q32.31 (Heilig et al. 2003). DIO3 gene encodes type 3 deiodinase (D3) (Hernandez 2005). Human D3 gene participates in maintaining the appropriate level of thyroid hormone (TH) in fetuses and adults (Hernandez 2005). Hernandez et al. analyzed the expression of DIO3OS in human tissues by northern blot using human DIO3OS cDNA as a probe (Hernandez et al. 2004). His result showed that DIO3OS was widely expressed in human tissues. And their observation results indicated that the predicted DIO3OS polypeptide sequence had no significant homology with any entry in the protein database (Reik and Walter 2001), which proved that DIO3OS was a noncoding gene. Previous examines indicated that DIO3OS was highly expressed in thyroid cancer, and the DIO3OS/let-7d/NF-κB2 axis regulated the activity of thyroid cancer cells (Wang et al. 2021). DIO3OS can promote the growth and invasion of pancreatic cancer cells by competing miR-122 to regulate the expression of ALDOA (Cui et al. 2019). However, there is little exploration on DIO3OS, and its specific role in HCC is not clear. The purpose of this article was to investigate whether lncRNA DIO3OS plays a role in the early diagnosis and prognosis of HCC as a potential biomarker.

Data Acquisition and Processing
In this study, all data were included from the UCSC xena (https:// xenab rowser. net/ datap ages/) database to determine the expression levels and prognostic significance of DIO3OS in 32 types of cancers. Besides, normal and tumor samples were extracted from the LIHC project RNA-seq data of the TCGA database (https:// portal. gdc. cancer. gov/) to compare the changes of lncRNA DIO3OS expression between tumor and normal tissues. The characteristic markers of tumor-infiltrating lymphocytes in the study were downloaded from the CELLMARKER database (http:// biobigda ta. hrbmu. edu. cn/ CellM arker/) to analyze the correlation between DIO3OS and immune cell signature genes. In all the above tests, the median of DIO3OS was used for separating high and low expression groups online. The limma package of the R (version 3.6.3) software was used to convert the level3 high-throughput sequencing fragments per kilobase per million (HTSeq-FPKM) data into transcripts per million reads (TPM) data, and then convert the data undergoes log2 conversion. The Wilcoxon rank sum test was used to assess the significance of differences. The purpose of our work is to explain the prognostic diagnosis and impact of lncRNA DIO3OS on HCC. Therefore, patients who died of other causes are not included in the scope of this study.

Survival Analysis
The Kaplan-Meier plotter was conducted to detect the prognostic significance of DIO3OS expression in HCC. The mainly prognostic indicators in our study were overall survival rate (OS), progression free disease (DSS), and progression free interval (PFI). The clinical data was gained from the TCGA database. The "survival" package (version 3.2-10) of R software was used for statistical analysis, and "survminer" package (version 0.4.9) was used for visualization. The significance of the difference was calculated using the cox regression method.

Univariate and Multivariate Cox Regression Analysis
We implemented proportional risk regression model to detect the effects of clinicopathological factors and DIO3OS expression on the OS of HCC patients. The relevant data came from the TCGA database (Liu et al. 2018). p value < 0.1 was taken as the index included in multivariate analysis. The "survival" package (version 3.2-10) was used for statistical analysis and "ggplot2" package (version 3.3.3) was used for visualization. The significance of the differences was assessed using the Wald test. The clinical variables in our job included the following: T stage, N stage, tumor status, pathological stage, historical grade, vascular invasion, residual tumor, AFP level, albumin level, and DIO3OS expression level.

Enrichment Analysis
The principle of GSEA analysis is to compare the target gene with the predetermined gene set to analyze the whole genome expression profile microarray data. Through this method, we can understand the expression trend of genes of interest in specific functional gene sets and whether the expression trend is statistically significant (Subramanian et al. 2005). We applied GSEA assay to explore the special biological function of DIO3OS in the progression of HCC. Possible molecules associated with DIO3OS were inferred by single gene differential analysis. According to the intermediate value of DIO3OS expression, it was divided into low expression group and high expression group. The DESeq2 package (version 1.26.0) was used for analysis, and the differential gene expression list associated with DIO3OS was obtained. The sorted gene list was imported into clusterProfiler package (version 3.14.3) for enrichment analysis. In this process, the c2.cp.v7.2. symbols. gmt gene collection from the human molecular signatures database (MsigDB) (https:// www. gsea-msigdb. org/ gsea/ msigdb/ colle ctions. jsp# C2) is referred to, the species is set as homo sapiens, the computation times of each gene set is set as 1000, and the p.value is corrected by BH method. On the condition that |NES| (normalized enrichment score) > 1, false discovery rate (FDR) ≤ 0.25, and p val ≤ 0.05, the results were sorted according to the enrichment score, and the significantly enriched pathways were screened.

Tumor Immunoassay
In this study, RNA-seq data from the TCGA database LIHC project were downloaded and processed, and the "ESTIMATE" package (version 1.0.13) was used to evaluate the content level scores of immune cells and stromal cells expressed by DIO3OS in each sample. The significance of the differences was assessed using the Wilcoxon rank sum test. The association between DIO3OS expression level and tumor immune infiltration of HCC was compared by score estimation and data visualization.
Referring to Bindea et al., we included 24 types of immune cells (Bindea et al. 2013). In our study, ssGSEA algorithm in "GSVA" package (version 1.34.0) and spearman correlation tests were used to study the association between DIO3OS and immune cells. In addition, we also downloaded the markers of all immune cells in liver tissue from the CELLMARKER database, and compared the degree of correlation between DIO3OS and different immune cell markers through the spearman algorithm.

Statistical Analysis
The packages used for statistical analysis and data visualization in the above experiments are derived from the R software (version 3.6.3) (http:// www. Rproj ect. org). The significance indicator setting is as follows, "ns" is no statistical significance, *p < 0.05, **p < 0.01, ***p < 0.001. In this study, experimental results of p < 0.05 were considered statistically significant.

The Expression of lncRNA DIO3OS was Significantly Lower in Hepatocellular Carcinoma
We used Wilcoxon rank sum test to investigate the expression differences of DIO3OS in a variety of tumors (Fig. 1a). Then, we compared the DIO3OS expression in tumor and normal samples. As the result demonstrated that DIO3OS expression was notably declined in HCC patients (p < 0.001) (Fig. 1b). Furthermore, we investigated the DIO3OS expression in 50 pairs of tumor tissues and normal adjacent tissues by paired sample t test (Fig. 1c). The results showed that DIO3OS levels were significantly lower in various types of tumors than in normal tissue. Not only that, the differential analysis of the expression of this lncRNA between tumor samples and normal samples of HCC also proved that a significant decrease in the level of DIO3OS (p < 0.001).

The Low Expression Level of DIO3OS Predictor of HCC Patients with Poor Prognosis
Kaplan-Meier plotter can describe patients' survival. As shown in Fig. 2a-c, with the decrease of DIO3OS expression, the OS (HR = 0.63, p = 0.024), DFS (HR = 0.56, p = 0.015), and PFI (HR = 0.70, p = 0.016) of patients also presented a trend of decreased survival probability and shortened survival time. When the hazard ratio (HR) < 0 in the Kaplan-Meier curves analysis, the study subject was a protective factor. Thus, high levels of DIO3OS are favorable indicators of higher patient survival.
The results of the Cox regression analysis are shown in Fig. 2e, the size and extent of tumor invasion (HR = 2.598, p < 0.001), tumor status (HR = 0.432, Fig. 2 Low expression of lncRNA DIO3OS is associated with poor OS, DSS, and PFI in patients with hepatocellular carcinoma. a Overall survival (n = 373), b disease specific survival (n = 365), c progress free interval (n = 373). d Multivariate Cox regression analysis is used to determine the effect of selected variables on OS. e Univariate Cox regression analysis is used to determine the effect of selected variables on OS p < 0.001), pathological stage (HR = 2.504, p < 0.001), and DIO3OS expression (HR = 0.777, p = 0.017) were significantly correlated with poor OS in HCC patients. Among them, high level of DIO3OS is a protective factor for HCC patients. And multivariate analysis revealed that tumor status (HR = 0.561, p = 0.005) was an independent predictor for negative prognosis (Fig. 2e). This finding suggests that lowlevel DIO3OS is closely related to the process and variation of HCC.

GSEA Assay Identifies Biological Functions Related to DIO3OS
The Fig. 3 and Table 1 showed that DIO3OS was closely related to multiple tumor immune-related signaling pathways. It turns out that as DIO3OS goes down, PI3K/ AKT signaling pathway, Notch pathway, TGFβ signaling pathway, WNT pathway, JAK/STAT pathway, neutrophil degranulation process, immunoregulatory interactions between a lymphoid and a non-lymphoid cell, HIPPOMERLIN signaling dysregulation, ECM receptor interaction process, Leukocyte trans leukocyte migration process showed a significant up-regulation trend. However, regulates the stability and active function of PTEN showed a significant downward trend.

Relationship Between DIO3OS Expression and Immune Infiltration in Hepatocellular Carcinoma
By using the "ESTIMATE" package, we can draw the following conclusions. As shown in Fig. 4a, both immune scores and matrix scores were significantly higher in the high DIO3OS expression group than in the low DIO3OS expression group. The immune score represents the content of immune cells in the tumor, and the stromal score represents the content of stromal cells. This suggests that the number of immune cells in the tumor is relatively reduced when DIO3OS levels are reduced.
According to the document of Bindea et al. (2013), we selected 24 kinds of immune cells to investigate the correlation between DIO3OS expression and immune cells. Figure 4b and Table 2 illustrate that in HCC, DIO3OS is associated with NK cells, mast cells, two types of DC cells (iDC, pDC), five types of T cells (Tem, TFH, CD8 T cells, Th1 cells, Tgd), eosinophils, macrophages, B cells, NK CD56 bright cells were significantly positively correlated, and Treg cells were significantly negatively correlated. This result is consistent with that of the ESTIMATE analysis.
As Table 3

Discussion
Statistical studies in the past few decades have proved that LIHC is the third highest cause of cancer-related death around the world (Global Burden of Disease Cancer et al. 2019), in which about 90% of cases were diagnosed as HCC (Forner et al. 2018). Compared with western countries, the cases of HCC in China account for about half of the total cases worldwide (Parkin et al. 2001). Statistical analysis revealed that the average five-year survival rate of HCC patients was less than 6% (Chen et al. 2016). It has only been in recent years that harnessing the lncRNA in the fight against cancer has become well appreciated. Previous outcomes had shown that lncRNA DIO3OS expression was closely related to a multitude of cancers, including LIHC (Song et al. 2019), but the specific role of DIO3OS in HCC is still unclear. Regulatory T (Treg) cells are a subset of CD4 T cells. Treg cells have immunosuppressive ability, which can avoid the occurrence of autoimmune diseases, but there is also the possibility that tumors can escape the immune defense of the body (Nishikawa and Sakaguchi 2010). Numerous reports have proved that Treg cells easily infiltrate the tumor microenvironment (TME), which often predicts a poor prognosis for patients (Ohue and Nishikawa 2019). FOXP3, a member of the forked transcription factor family, is a determinant of differentiation and development of Treg cell and plays a key role in regulating immune homeostasis (Marson et al. 2007). In the TME, Treg cells can secrete TGF-β, IL-10, IL-35, and other inhibitory cytokines to play an immunosuppressive role (Zhou et al. 2015). In addition, the IL-2α chain receptor (CD25) highly expressed on the surface of Treg cells can competitively bind with IL-2, giving Treg cells a competitive advantage in utilizing the limited IL-2 in TME, thus inhibiting the proliferation of CD4 + T cells and CD8 + T cells and inhibiting anti-tumor immunity (Sakaguchi et al. 1950). In addition, the surface of Treg cells constitutionally express cytotoxic T lymphocyte-4 (CTLA-4) and lymphocyte-activating gene-3 (LAG-3), and these inhibitory receptors can interact with DCs surface ligand CD80/CD86 to induce indoleamine 2,3-dioxygenase (IDO) enzyme to further inhibit the ability of T cells (Echarti et al. 2019). Previous studies also demonstrated that Treg cells can express granzyme in TME. Granzyme and perforin-like molecules inhibit antitumor immunity by dissolving NK cells and CTL cells (Cao et al. 2007). Our analysis results showed that DIO3OS was significantly negatively correlated with Treg cells in HCC, as well as the expression of FOXP3, a specific marker of Treg cells. Natural killer (NK) cells are important congenital cytotoxic lymphocytes, which have the ability to recognize and kill tumor cells quickly and efficiently (Fregni et al. 2012). One of the major factors limiting NK cell activity is TME. Schleypen et al. found that, in renal cell carcinoma, a reduction of tumor-infiltrating NK cells and down-regulation of CD16 receptor expression indicated attenuated NK cell cytotoxic activity (Schleypen et al. 2006). The TME contains many cells with immunosuppressive functions, such as Treg cells and tumor-associated macrophages (TAM) (Arneth 2019). These cells produce and secrete factors that directly or indirectly inhibit the activation of NK cells, including IL-6, IL-10, TGFβ, and IDO (Konjević et al. 2019). TGF-β downregulates the expression of activating receptors and effector functions in NK cells, and reduces IL-2-mediated mitochondrial metabolism, including oxidative phosphorylation (OXPHOS) and respiratory function in human NK cells (Wang et al. 2020a, b). In addition, tumor derived metabolites such as adenosine and lactate inhibit the antitumor activity of NK cells (Scott and Cleveland 2016). Adenosine stimulates NK cells through the adenosine A 2A receptor (A 2A R) and inhibits its function (Young et al. 2018). However, high concentrations of lactate and low pH decreased the cytotoxic activity of NK cells. In advanced liver disease and hepatocellular carcinoma, accumulation of lactate in the liver microenvironment inactivates hepatotoxic lymphocytes such as NK cells and CD8 T cells (Piñeiro Fernández et al. 2019;Sun et al. 2015). NK cell activity is inhibited by TME, and tumor cells disrupt the balance between NK cell inhibitory activation and inhibitory signaling, resulting in NK cell dysfunction (Habif et al. 2019).
In the tumor immunity, antigen presenting cells (APC) are required to present tumor associated antigen (TAA) to T cells in order to initiate an effective adaptive immune response (Peterson and Barry 2020). Dendritic cells (DC) are considered as the most important APCs. DCs are part of the hematopoietic cell lineage, and three subsets of DCs can be roughly identified from blood and lymphoid organs, including plasmacytoid DCs (pDCs), conventional type 1 DCs (cDC1s), and conventional type 2 DCs (cDC2s) (Amon et al. 2019). It has been proved in many studies that the interaction between NK cells and DCs, especially cDC1s, is of great significance in regulating the immune response of cancer (Bödder et al. 2020). CCL5, XCL1, and XCL2 produced by NK cells can promote the entry of cDC1s into the TME (Böttcher et al. 2018). IL-12 produced by activated DCs can stimulate NK cell activation and anti-tumor T cell immunity (Sarhan et al. 2015). The immunosuppressive factors in the TME can not only inhibit the aggregation of DC cells, but also inhibit their stimulatory activity. Studies have shown that tumor derived prostaglandin 2 (PGE2) inhibits the production of IL12 by cDC1s, while downregulating the expression of costimulatory molecules and inhibiting the induction of NK cells by cDC1s, resulting in impaired NK cell function (Böttcher et al. 2018). In addition, TGF-β and vascular endothelial growth factor A (VEGF-A) secreted by tumors can affect the differentiation and maturation of DCs, while TGF-β also regulates the activity of T cells, B cells, and macrophages, and supports the growth of Treg cells (Farhood et al. 2020;Suuring and Moreau 2021). To sum up, DCs interact with a variety of immune cells in tumor immunity and are regarded as key targets in the body's immune response. As far as the current status of immunotherapy is concerned, DCs have gradually become a research hotspot that has attracted much attention.
Carcinoma is a genetic and immune-mediated disease. Tumor immunotherapy is the most attractive anticancer method in the field of medicine. However, due to the balance between immune cells, the differences in immune cell infiltration between different tumor tissues, and the complexity of individual immune systems, immunotherapy dose not achieved the desired effect, and many patients develop many immune system problems during treatment. Therefore, there are still many obstacles to targeting immune cells within tumors.
Our analysis showed that DIO3OS in HCC often predicted negative prognosis of patients, and this lncRNA has certain accuracy and sensitivity as a diagnostic indicator. Many studies have proved that lncRNAs cannot translate proteins, but have diverse biological functions. Biological functional studies have found that decreased DIO3OS expression is significantly associated with a variety of tumor immunerelated pathways. From this, we hypothesized that the expression of DIO3OS in HCC would affect the abundance of immune cells in the tumor immune microenvironment (TIME) thereby causing tumor immune escape. In order to verify this hypothesis, the "ESTIMATE" algorithm was used to calculate the immune infiltration score, matrix score, and estimated score of tumor samples. The results showed that the immune infiltration score of the low DIO3OS expression group was significantly lower than that in the high expression group, which can prove that in the tumor sample, the low DIO3OS expression is indeed associated with the lower level of immune cell content. We further explored the relationship between low DIO3OS expression and the abundance of different types of immune cells in HCC. The results of ssGSEA algorithm obviously showed that the infiltration degree or expression intensity of a variety of immune cells with anti-tumor function, such as T cells, B cells, NK cells, and DCs, also showed a trend of significantly decreasing with the decrease of DIO3OS expression. However, the infiltration degree and expression intensity of Treg cells were significantly negatively correlated with the expression of DIO3OS. After that, we collected various markers of liver immune cells and analyzed their correlation with DIO3OS. The results were consistent with the conclusions of the immune cell correlation analysis. The above experiments confirm our hypothesis that decreased DIO3OS expression may directly or causally affect the abundance of various immune cells in the TIME. This is just a preliminary conjecture, and the specific mechanism by which DIO3OS expression changes affect the abundance of immune cells needs to be further investigated. The results were consistent with the conclusions of the immune cell correlation analysis. Studies have shown that DIO3OS can disrupt the Hedgehog (Hh) pathway by binding to miR-328 mediated Hh interacting protein (Hhip), thereby inhibiting the occurrence of HCC (Wang et al. 2020a, b). This further proves that DIO3OS is an effective target for the treatment of HCC, but few studies on DIO3OS in tumor immunity can be found. Therefore, our study can provide a new idea for the treatment of HCC, by upregulating the expression of DIO3OS to activate the anti-tumor activity of NK cells or DC cells, or inhibit the activity of Treg cells to prevent their immunosuppressive effects. Our research is based on big data analysis of multiple databases. Despite our comprehensive collection of relevant data, the study is still flawed, demonstrating a lack of experimental and clinical validation. Therefore, the preliminary theoretical basis we currently propose requires more studies to testify.

Conclusion
In summary, decreased DIO3OS expression in HCC indicates poor prognosis in patients with increased Treg cells infiltration and decreased other immune cell infiltration. Besides, low DIO3OS expression is associated with multiple inflammatory pathways, and is positively or negatively correlated with multiple immunomarker genes. These evidences suggest that DIO3OS may play an important role in tumor immune cell infiltration and may serve as a prognostic biomarker or therapeutic target for HCC patients.