A Pan-cancer Analysis of Thioredoxin-interacting Protein as an Immunological and Prognostic Biomarker

doi:10.21203/rs.3.rs-1243997/v1

Download PDF

Research Article

A Pan-cancer Analysis of Thioredoxin-interacting Protein as an Immunological and Prognostic Biomarker

https://doi.org/10.21203/rs.3.rs-1243997/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Background

The critical role of thioredoxin-interacting protein (TXNIP) in cellular sulfhydryl redox homeostasis and inflammasome activation is already widely known, however, no pan-cancer analysis is currently available.

Methods

We thus first explored the potential roles of TXNIP across thirty-three tumors mainly based on The Cancer Genome Atlas and Gene Expression Omnibus datasets.

Results

TXNIP is lowly expressed in most cancers, and distinct associations exist between TXNIP expression and the prognosis of tumor patients. TXNIP expression was associated with tumor mutational burden, microsatellite instability, mismatch repair genes, tumor infiltrating immune cell abundance as well as cancer-associated fibroblasts. Moreover, ubiquitin mediated proteolysis, protein post-translational modification and other related pathways were involved in the functional mechanisms of TXNIP.

Conclusions

Our first pan-cancer study offers a relatively comprehensive understanding of the carcinostatic roles of TXNIP across different tumors. And this molecule may be considered as a potential immunological and prognostic biomarker.

Cancer Biology

TXNIP

pan-cancer

prognosis

genetic alteration

immune infiltration

ubiquitination

Given the complexity of tumorigenesis and development, it is necessary to carry out a multi-dimensional pan-cancer expression analysis of interest gene and further assess its potential molecular mechanisms. The publicly funded TCGA (The Cancer Genome Atlas) project and the available GEO (Gene Expression Omnibus) database contain functional genomics data sets of different tumors¹^,² and thus allow us to conduct pan-cancer analysis.

Thioredoxin-interacting protein (TXNIP), also known as Thioredoxin-binding protein-2 or Vitamin D3 up-regulated protein-1, was originally identified as an endogenous antagonist of thioredoxin (Trx), a key regulator in cellular redox equilibrium³. The TXNIP gene belongs to the arrestin family and structure analyses suggests that it is highly conserved among species³^,⁴. With respect to the human TXNIP protein, two alternatively spliced isoforms have been reported, and we have focused on TXNIP1 (NM_006472.6) in this study. TXNIP protein may act as an oxidative stress mediator by inhibiting Trx activity or limiting its bioavailability, however this protein also exerts several other physiological and pathological effects, including regulating cell proliferation, cell division, apoptotic process, and cellular response to estradiol/progesterone or mechanical/chemical stimulus⁵^-⁷.Importantly, the role of TXNIP in carcinogenesis and modulating tumor progression has been attracting increasing interest. Fig. 1A provides an overview of specific role of TXNIP in cancer. In spite of the current reports indicating the significant contribution of TXNIP in carcinogenesis, our knowledge of the specific implication concerning TXNIP still remains limited. Considering the increasing essential role of TXNIP in cancer, it is of great interest to unravel the whole landscape of expression, mutation, immune response as well as their prognostic potential.

Here we first identified the expression profiles of TXNIP in various normal tissues and certain types of cancer utilizing data from the Genotype-Tissue Expression (GTEx) and the Cancer Genome Atlas (TCGA) data portals. Then, we also included a group of factors, such as survival status, genetic alteration, immune infiltration, protein ubiquitination, protein phosphorylation, and relevant cellular pathway, to investigate the potential molecular mechanism of TXNIP in the pathogenesis or clinical prognosis of different cancers.

2.1 Gene expression analysis

The systematic analysis of TXNIP expression levels between tumor and adjacent normal tissues was conducted according to RNA sequencing data from the TCGA project via the “Gene_DE” functional module of the Tumor Immune Estimation Resource 2 (TIMER2) web (http://timer.cistrome.org)⁸. Gene Expression Profiling Interactive Analysis 2 (GEPIA2) (http://gepia2.cancer-pku.cn/#analysis)⁹ was applied to profile the tissue-wise expression of TXNIP gene in several cancer types without normal or with highly limited normal tissues [e.g., adrenocortical carcinoma (ACC), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), ovarian serous cystadenocarcinoma (OV), etc.]. The threshold settings were as follows: | log₂FC (fold change) | cutoff = 1, P-value cutoff = 0.01, and “Match TCGA normal and GTEx data”. Additionally, the differential expression between tumor tissues and the matched normal tissues was also identified in the Oncomine database (https://www.oncomine.org)¹⁰. We identified the expression levels of the total protein of TXNIP (NP_006463.3) in UALCAN portal (http://ualcan.path.uab.edu/analysis-prot.html)¹¹.

2.2 Clinical specimens

A total of 50 patients who underwent radical operation for CRC at Nanfang Hospital of Southern Medical University were included in the study after obtaining informed consent. A diagnosis of CRC was histopathologically confirmed for each patient sample. Fifty pairs cancer and matched normal tissues were prepared for qRT-PCR analysis, twenty-two pairs of which were randomly selected for immunohistochemistry (IHC) analysis, and eighteen pairs for western blotting analysis. Additionally, none of the patients had received radiotherapy or chemotherapy before surgery. The protocols used in this paper were approved by Nanfang Hospital’s Protection of Human Subjects Committee (NFEC-201809-K3).

2.3 Survival prognosis analysis

“Survival” module of GEPIA2 was utilized to assess the correlation between TXNIP expression and prognosis of cancers. The overall survival (OS) and disease-free survival (DFS) significance map data of TXNIP was obtained using Mantel-Cox test, under the settings of cutoff-high = 50%, cutoff-low = 50%, group cutoff = Median, and significance level = 0.05.

The Kaplan Meier plotter database (http://kmplot.com/analysis/)¹² is capable to assess the effect of proposed genes symbols on survival in 21 cancer types including breast (n = 6234), ovarian (n = 2190), lung (n = 3452), and gastric (n = 1440) cancer. To analyze the prognostic value of TXNIP, the patient samples are split into two groups according to quantile expressions of TXNIP. Then, the two patient cohorts are compared by a Kaplan-Meier survival plot, and the hazard ratio (HR) with 95% confidence intervals (CI) and logrank P-value are calculated. A P-value < 0.05 was considered statistically significant.

2.4 Genetic alteration analysis

Data concerning the alteration frequency of TXNIP gene and the schematic diagram of the mutated site information of TXNIP were acquired in the “Cancer Types Summary” and “Mutations” module in cBioPortal (http://cbioportal.org)¹³.

2.5 Correlation analysis of TXNIP and TMB/MSI/MMR/DNMT/PTM

The potential correlation between TXNIP expression and the tumor mutational burden (TMB) or microsatellite instability (MSI) across tumors in TCGA was investigated using the Sangerbox tools (http://sangerbox.com/tool), a free online platform for data analysis. Spearman’s rank correlation test was performed, and the P-value and partial correlation (cor) value were obtained. Data revealing the relationship of TXNIP expression with 5 Mismatch repair (MMR) genes (MLH1, MSH2, MSH6, PMS2, and EPCAM) and 4 DNA methyltransferases (DNMTs) in all TCGA cancer cases was generated through the same public platform.

Protein post-translational modification (PTM) regulates a broad range of cellular biological processes, many of which are closely related to multiple cancer types¹⁴^-¹⁶. We analyzed the protein modification sites of TXNIP using the open-access SMART web (http://smart.embl.de/), and obtained the predicted ubiquitination and phosphorylation features of TXNIP.

2.6 Immune infiltration analysis

The TXNIP expression in different TCGA tumors and its correlation with the abundances of eight tumor-infiltrating immune cells (TIICs), together with cancer-associated fibroblasts (CAFs) was explored by the corresponding gene modules of TIMER2 web server. The P-values and partial correlation (cor) values were obtained via the purity-adjusted Spearman’s rank correlation test. Additionally, the correlation module was applied to infer the relationships between TXNIP expression and gene markers of TIICs in three given cancer types, and a total of 62 immune-related gene markers were selected for analysis via the CellMarker database (http://bio-bigdata.hrbmu.edu.cn/CellMarker/)¹⁷.

For reliable evaluation of tumor immune microenvironment, we used an R software immuneeconv to analyze the immune scores of each tumor sample. Immune checkpoint inhibitor is a promising therapy for tumor immunotherapy¹⁸. The expression values of the 47 transcripts related to immune checkpoints was extracted, and its association with TXNIP expression in different tumor tissues was organized in a heat map. The expression levels of immune checkpoints genes in the same tumor tissue was averaged first, and then performed zscore standardization. Tumor neoantigen is a new antigen encoded by mutated genes of tumor cells, which has immunological activity¹⁹. Therefore, it is expected to design and synthesize therapeutic vaccines for specific mutations of tumor cells. Herein we collected the numbers of neoantigen in multiple cancers and further analyzed the relationship between TXNIP expression and the numbers of neoantigen of each type of tumor.

2.7 TXNIP-related gene enrichment analysis

The interaction network of the experimentally verified TXNIP-binding proteins were obtained via the STRING website (https://string-db.org/). The top 100 TXNIP-correlated targeting genes was acquired using the GEPIA2 web server based on the datasets of all TCGA tumor and normal tissues. Further, the pairwise gene Pearson correlation analysis of TXNIP and selected genes was conducted by GEPIA2, and the log₂ TPM was applied for the dot plot. The heatmap data of the abovementioned genes was supplied by the “Gene_Corr” module of TIMER2, which contained the partial correlation (Cor) and P-value in the purity-adjusted Spearman’s rank correlation test. Moreover, the Jvenn (http://bioinformatics.psb.ugent.be/webtools/Venn/), a practical tool for drawing custom Venn diagrams of list of elements, was used to calculate the intersection of the TXNIP-binding and interacted genes.

Moreover, the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis was performed with the combined gene list. The functional annotation chart of the TXNIP related genes was obtained from the Database for Annotation, Visualization, and Integrated Discovery (DAVID) database (https://david.ncifcrf.gov/home.jsp)²⁰. To conduct the Gene Ontology (GO) enrichment analysis, data for biological process (BP), cellular component (CC), and molecular function (MF) were visualized as circular plots through the RCircos R package. The R software v4.0.3 (https://www.r-project.org/) was used in this analysis.

2.8 RNA isolation and quantitative real-time PCR (qRT-PCR)

Total RNA was isolated from tissues using the TRIzol reagent (TaKaRa, China). RT-qPCR was performed using the PrimeScript RT Reagent Kit (#RR035A, TaKaRa, China) and SYBR Premix Ex Taq (#RR820A, TaKaRa, Dalian, China) according to the manufacturer’s instructions. The specific primers used are as follows: TXNIP-forward：ATTGGAGAGCCCAACCACTC, TXNIP-reverse：TTCCACATGCTCACTGCACA, GAPDH-forward：AAATCCCATCACCATCTTCC, GAPDH-reverse：TCACACCCATGACGAACA. The RT-qPCR results were analyzed to obtain the Ct values of the amplified products, and data were analyzed by the 2-ΔΔCt method.

2.9 Western blotting

Western blotting was conducted as previously described²¹. Protein lysates were prepared, subjected to 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), transferred onto polyvinylidene difluoride (PVDF) membranes and blotted according to standard methods using the following antibodies: TXNIP (1:1000, 18243-1-AP, Proteintech), GAPDH (1:50000, 60004-1-Ig, Proteintech).

2.10 TXNIP IHC analysis

IHC staining was performed following the manufacturer’s instructions (PV-6001, ZSGB-BIO, Beijing, China) using TXNIP (1:200, 18243-1-AP, Proteintech). Two independent pathologists used software ImageJ to calculate the proportion of positive area. Data were expressed as the average of three randomly selected microscopic fields.

2.11 Statistical Analysis

Gene expression data from the TCGA and GTEx databases were analyzed using Student’s t-test. In PrognoScan, the univariate Cox regression model was used to calculate the HR and P value. In GEPIA2 and Kaplan-Meier Plotter, log rank test was used to calculate the HR and its P value in order to compare survival curves. The correlations between TXNIP expression and abundance scores of immune cells/ TMB/MSI/MMR/DNMT evaluated by Spearman’s correlation. Results with P < 0.05 were considered as statistically significant, providing credibility for the data analysis.

3.1 TXNIP expression analysis data

To conduct a comprehensive analysis of human TXNIP (NM_006472.6 for mRNA, NP_006463.3 for protein, data relating with its genome location, conserved functional domain and the phylogenetic tree were obtained(Fig. S1A-S1C). As shown in Fig. S1D, the levels of mRNA expression of TXNIP shows a low tissue specificity with all normalized expression values in detected tissues > 1. Then, a low cell type specificity of TXNIP expression in different blood cells is also observed (Fig. S1E).

To figure out whether TXNIP expression correlates with cancer, we firstly evaluated TXNIP expression in different tumors and adjacent normal tissues using the online Oncomine database. The results showed that TXNIP expression was lower in several cancer groups than in normal tissues, including breast cancer, colorectal cancer (CRC), head and neck cancer, lung cancer, ovarian cancer, sarcoma (SARC), lymphoma, liver cancer or bladder cancer (Fig. 1B, all P < 0.05), and more detailed results were summarized in Table S1. Moreover, we further used TIMER2 and GEPIA2 web servers to evaluate the RNA sequencing data of TXNIP in TCGA and GTEx. Data confirmed that significantly more tumor tissues expressed lower levels of TXNIP mRNA than the corresponding control tissues (Fig. 1C, 1D), of which included BLCA, UCEC, breast invasive carcinoma (BRCA), LUAD, lung squamous cell carcinoma (LUSC), colon adenocarcinoma (COAD), rectum adenocarcinoma (READ), stomach adenocarcinoma (STAD), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), prostate adenocarcinoma (PRAD), thyroid carcinoma (THCA) (P < 0.001), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), esophageal carcinoma (ESCA) (P < 0.01), OV, DLBC, ACC, skin cutaneous melanoma (SKCM), uterine carcinosarcoma (UCS), and liver hepatocellular carcinoma (LIHC) (P < 0.05). Results of the TXNIP proteomic expression profile analysis showed absolutely lower expression of TXNIP total protein in the primary tissues of breast cancer, colon cancer, ovarian cancer, LUAD and UCEC (Fig. 1E, all P < 0.001) than in normal tissues. Taken together, the data confirmed that the TXNIP gene was down-regulated in multiple cancers compared to normal samples.

To detect the transcriptional level of TXNIP in CRC, we collected 50 pairs of fresh tumor specimens with matched normal tissues from CRC patients in our hospital. The qPCR analysis (n = 50) revealed that the expression level of TXNIP mRNA in CRC was significantly lower than that in adjacent tissues (Fig. 1F, P < 0.0001). Further, western blotting analysis of 18 pairs of randomly selected CRC samples showed that tumor tissues exhibited markedly lower level of TXNIP than the adjacent tissues (Fig. 1G, P = 0.0133). TXNIP IHC staining of 22 pairs of CRC samples drew consistent conclusion. TXNIP was lowly expressed in 12 out of 22 tumor tissues and highly expressed in 19 out of 22 normal tissues (Fig. 1H). Immunohistochemical score of TXNIP showed that the expression level of TXNIP in tumor tissues was significantly lower than that in paired normal tissues (Fig. 1I, P = 0.0003).

3.2 Survival analysis data

Next, we analyzed the prognostic value of TXNIP expression across cancers in GEPIA2. As shown in Fig. 2A, OS analysis data showed a correlation between low TXNIP expression and poor prognosis for cancers of KIRC (P = 2.0e-04) within the TCGA project. Meanwhile, lowly expressed TXNIP was linked to poor prognosis of DFS for the TCGA cases of CHOL (P = 0.010), KIRC (P = 0.002), SARC (P = 0.019), and uveal melanoma (UVM, P = 0.026) (Fig. 2B). Interestingly, high expression of the TXNIP gene was related to poor OS prognosis for OV and STAD (Fig. 2A, all P < 0.05), suggesting the prognostic value of TXNIP may has tissue or tumor specificity.

Moreover, the TXNIP prognostic value was evaluated by the Kaplan-Meier plotter database using Affymetrix microarrays data (Fig. 2C–O, Fig. S2A–H). And results demonstrated that low levels of TXNIP was closely linked to poor OS, relapse-free survival (RFS), post-progression survival (PPS), and distant metastasis-free survival (DMFS) prognosis for breast cancer (Fig. 2C–F, all P < 0.001); poor OS (P = 0.000), RFS (P = 0.017), progression-free survival (PFS, P = 0.027), and disease specific survival (DSS, P = 0.001) prognosis for liver cancer (Fig. 2G–J); and poor OS (P = 2.7e-10), PPS (P = 0.000), and first progression (FP, P = 0.001) prognosis for lung cancer (Fig. 2K–M). However, we failed to detect a correlation between TXNIP expression and the OS, PFS and PPS prognosis of ovarian cancer (data not shown, all P > 0.05). Based on the datasets of the Kaplan-Meier plotter, we observed that high expression of TXNIP was associated with poor clinical outcomes of OS and PPS for gastric cancer, especially in subgroup analysis of “stage 2 to 4”, “lauren classification/intestinal/diffuse” and “treatment/surgery alone” (Table S2). We also analyzed the correlation of TXNIP expression and the selected clinicopathological factors in liver (Table S3), lung (Table S4), and breast cancer (Table S5), and finally observed distinct conclusions concerning these tumors. These findings suggest that low TXNIP expression implies reduced survival in breast, lung and liver cancer.

In addition to TXNIP microarray analysis in the Kaplan-Meier plotter database, the impact of TXNIP expression to survival rates of patients was evaluated using the PrognoScan(Table S6 and Fig. S2I–T). Notably, decreased expression of TXNIP indicated poorer survival prognosis in seven out of eight cancers, including breast, bladder, skin, lung, brain, eye and soft tissue cancers (Fig. S2I–S, all P < 0.05). Unexpectedly, analysis of two cohorts (GSE14333, GSE17536) showed that high TXNIP expression was linked to poorer DFS prognosis of CRC patients (Table S6 and Fig. S2T, all P < 0.05). Accordingly, it is conceivable that low TXNIP expression is an independent risk factor and leads to a poor prognosis in breast, bladder, and lung cancer patients.

3.3 Genetic alteration analysis data

The genetic alteration status of TXNIP in different tumor samples of the TCGA cohorts was examined by the cBioPortal website. As shown in Fig. 3A, bladder (n = 411), liver (n = 372) and lung (n = 566) cancer contribute to the top 3 tumors with the highest alteration frequencies of TXNIP, with an alteration frequency of 12.17%, 9.95%, and 9.89%, respectively. It is worth noting that the “amplification” type of copy number aberration (CNA) is the primary genetic alteration type in almost all tumors. The three-dimensional structure of TXNIP protein as well as the sites and types of the TXNIP genetic alteration are further presented in Fig. 3B and 3C. The missense mutation of TXNIP appears to be the main type of genetic alteration in PRAD (Fig. 3C). N389del/L386F alteration, detected in ACC, SKCM, UCEC and PRAD (Fig. 3C), is able to induce inframe and missense mutation of the TXNIP gene, leading to the subsequent deletion of N (asparagine) at the 389 site of TXNIP protein, the substitution of L (Leucine) with F (Phenylalanine) at the 386 site of TXNIP protein, respectively.

The tumor mutational burden (TMB) is a quantitative biomarker, which can reflect the number of mutations in tumor cells²². Microsatellite instability (MSI), a mode of genomic instability, is closely related with carcinogenesis, prognosis and response to immunotherapy²³^,²⁴. To assess the role of TXNIP in tumorigenesis, we analyzed the correlation between TXNIP expression and TMB/MSI across all tumors of TCGA. As shown in Fig. 3D, we observed a negative correlation between TXNIP expression and TMB for BRCA (P = 1.6e-12), LIHC (P = 0.013), LUSC (P = 0.007), PRAD (P = 3.2e-05), BLCA (P = 0.036), DLBC (P = 0.021), KIRC (P = 0.011), LUAD (P = 2e-09), STAD (P = 0.000), and THCA (P = 4.5e-07). TXNIP expression is also negatively correlated with MSI of BRCA (P = 0.044), LIHC (P = 0.021), LUSC (P = 0.000), PRAD (P = 0.002), ESCA (P = 0.019), SKCM (P = 2.8e-06), STAD (P = 2e-05), head and neck aquamous cell carcinoma (HNSC) (P = 2.6e-05), brain lower grade glioma (LGG) (P = 0.039) but is positively correlated with that of COAD (P = 1.3e-08) and READ (P = 0.012) (Fig. 3E). This suggests that TXNIP may be associated with tumorigenesis and treatment.

DNA methylation is an epigenetic modification that can alter gene expression. The change of DNA methylation status is an important factor of tumorigenesis²⁵^. Therefore, we next investigated the correlation between TXNIP expression and that of four DNA methyltransferases (DNMT1, DNMT2, DNMT3A, and DNMT3B) expression via the Sangerbox platform. The expression of TXNIP is negatively correlated with that of DNMTs in KIRC, KIRP, PRAD and PCPG (all P < 0.05). The deficiency in mismatch repair (MMR), a common DNA repair pathway, may increase mutations and result in MSI and carcinogenesis²⁶. Furthermore, TXNIP expression and the levels of five MMR genes, especially MLH1, MSH2, MSH6 and PMS2, are positively correlated in KIRC, kidney renal papillary cell carcinoma (KIRP), THCA, HNSC, PRAD, pancreatic adenocarcinoma (PAAD) and pheochromocytoma and paraganglioma (PCPG) (Fig. 3G, all P < 0.01). These results conformably demonstrate the important role of TXNIP as a tumor suppressor in certain types of cancer, suggesting that its deficiency may play a crucial role in oncogenesis.

Ubiquitination and phosphorylation are the enzymatic post-translational protein modification that regulate cellular biological processes in various ways and have been implicated in a number of cancer types¹⁴^-¹⁶. It is of note that dysregulation of specific protein ubiquitination may contribute substantially to cancer development and metastasis¹⁶. Here we outlined the possible ubiquitination and phosphorylation sites of TXNIP protein, as shown in Fig. 3H. This observation merits further molecular assays for further exploration, which may explain the mechanism of low expression of TXNIP protein.

3.4 Immune infiltration analysis data

Tumor-infiltrating immune cells, an indication of the host immune reaction to tumor antigens, play a critical role in tumor progression and antitumor activity²⁷^-²⁹. As the major TIIC population, tumor infiltrating lymphocytes are independent predictors of sentinel lymph node status and cancer survival²⁸^-³⁰. As shown in the heatmap of Figure 4A, TXNIP expression was significantly correlated with TIICs in various types of cancers. Specifically, in the tumors of BRCA-LumA (n = 568), CESC (n = 306), ESCA (n = 185), HNSC (n = 522), HNSC-HPV- (n = 422), HNSC-HPV+ (n = 98), KIRC (n = 533), LIHC (n = 371), SARC (n = 260), SKCM (n = 471), and STAD (n = 415), TXNIP expression is negatively correlated with non-polarized M0 macrophages, but positively correlated with polarized M1 and M2 macrophages, especially the M2. We thus hypothesized that TXNIP may activate the tumor-associated macrophages (TAMs) along with shifting macrophage phenotype to a more anti-inflammatory state. Among these studied cancers, TXNIP expression has significant correlations with infiltrating levels of CD4⁺ T cells in 22 types of cancer, CD8⁺ T cells in 19 types of cancer, macrophages in 22 types of cancer, neutrophils in 25 types of cancer, and dendritic cells in 23 types of cancer (Fig. 4B-4D and Fig. S3A-AC, all P<0.05).

We then selected the specific cancers in which TXNIP was correlated with oncologic outcomes and infiltrating immune cells. Using the prognostic results related to TXNIP from the GEPIA2, Kaplan-Meier-plotter and PrognoScan analyses, we eventually selected BRCA, LUAD and KIRC for further research on immune infiltration via TIMER2. The TXNIP expression had significant positive correlations with six types of infiltrating immune cells in BRCA (Fig. 4B, all P<0.01), LUAD (Fig. 4C, all P<0.001), and KIRC (Fig. 4D, all P<0.001). Tumor purity, defined as the proportion of tumor cells in the tumor sample, was reported to influence the genomic sequencing analysis of immune infiltration in clinical tumor samples³¹. After purity adjustment, we found that the correlation of TAMs genes (CCL2, CCL5 and CD68) obviously changed, especially the values in BRCA (n = 1100), which were the most significant changes (Table S7). These findings strongly demonstrate that TXNIP could recruit immune cells in the TME in BRCA, LUAD and KIRC.

CAFs in the stroma of the TME were reported to participate in modulating the function of various TIICs³¹. As shown in Fig. 4E, we observed a statistical positive correlation of TXNIP expression and the estimated infiltration value of CAFs for the TCGA tumors of BRCA, BRCA-Basal, COAD, LIHC, LUSC, READ, STAD, and testicular germ cell tumors (TGCT) (all P < 0.05). The scatterplot data of the above tumors are presented in Fig. 5F–5M.

To further assess the stromal content (StromalScore), overall immune infiltration (Est_ImmuneScore), and the combined score (ESTIMATEScore) of the tumor tissues, the ESTIMATE algorithm was employed³²^,³³. As shown in Fig. 5A–5C, the TXNIP expression was significantly positively correlated with the immune and matrix scores in the tumors of LUAD, LUSC, and STAD (all P<0.001). Mounting preclinical and clinical evidence discloses that immune checkpoint blockade therapy is emerging as one of the most promising strategy in oncology¹⁸. Next, in order to identify the potential therapeutic targets and further interpret the immune microenvironment, we analyzed the expression of a set of immune checkpoints in 33 types of studied tumors. Many types of tumors, especially CHOL, LUSC, and SKCM (Fig. 5D), highly express multiple immune checkpoint molecules, indicating that combinatorial treatment with different checkpoint inhibitors may be required for optimal tumor control. Increasing evidence suggests that tumor neoantigen plays a pivotal role in antitumor immune response and cancer immunotherapies¹⁹. Our data showed that the numbers of neoantigen in tumors like CESC (Fig. 5E), KIRP (Fig. 5G), and glioblastoma multiforme (GBM, Fig. 5F) was significantly positively associated with the expression of TXNIP (all P < 0.05). However, tumors such as STAD (Fig. 5H), LUAD (Fig. 5I), BRCA (Fig. 5J), and PRAD (Fig. 5K) displayed a negative association between the numbers of neoantigen and TXNIP expression (all P < 0.05).

3.5 Enrichment analysis of TXNIP-related partners

To further investigate the molecular mechanism of the TXNIP gene in tumorigenesis and development, we attempted to screen out the targeting TXNIP-binding proteins and the TXNIP expression-related genes for a series of pathway enrichment analyses. Based on the STRING tool, we obtained a total of 100 TXNIP-binding proteins, which were supported by experimental evidence. Fig. 6A shows the protein-protein interaction (PPI) network of these proteins. We used the GEPIA2 tool to combine all tumor expression data of TCGA and obtained the top 100 genes that correlated with TXNIP expression. As shown in Fig. 6B, the TXNIP expression level was positively correlated with that of CALCOCO1 (calcium binding and coiled-coil domain 1) (R = 0.63), KLF9 (Kruppel like factor 9) (R = 0.62), TGFBR3 (transforming growth factor beta receptor 3) (R = 0.64), TNS2 (tensin 2) (R = 0.59), TSC22D3 (TSC22 domain family member 3) (R = 0.59) and ZBTB16 (zinc finger and BTB domain-containing protein 16) (R = 0.56) genes (all P = 0.000). The corresponding heatmap data also showed a positive correlation between TXNIP and the above six genes in the majority of detailed cancer types (Fig. 6C). An intersection analysis of the above two groups showed one common member, namely, ZBTB16 (Fig. 6D).

We combined the two datasets to perform KEGG and GO enrichment analyses. The KEGG data of Fig. 6E suggest that “ubiquitin mediated proteolysis” might be involved in the effect of TXNIP on tumor pathogenesis. The GO enrichment analysis data further demonstrated that most of these genes are linked to the cellular biology of PTMs, such as protein dephosphorylation or phosphorylation, polyubiquitination, phosphoric ester hydrolase activity, protein phosphorylated amino acid binding, receptor tyrosine kinase binding, and are related to pathways including cell-cell signaling by Wnt, stress-activated protein kinase signaling cascade, insulin receptor signaling pathway, and vascular endothelial growth factor receptor signaling pathway, etc (Fig. 6F and Fig. S4).

It has been reported that the multifunctional TXNIP protein participates in an ocean of cellular biological processes across different species, such as the cell cycle, cell division, apoptotic process, and gene transcription regulation³^-⁷. In this study, our “HomoloGene” and phylogenetic tree analysis data also displayed conservation of the TXNIP protein structure across different species, suggesting that similar mechanisms might exist for the normal physiological role of TXNIP. Emerging evidences have proclaimed the functional link between TXNIP and clinical disorders, especially tumors. Whether TXNIP can play a role in the pathogenesis of different tumors through certain common molecular mechanisms remains to be explored. Unfortunately, we failed to retrieve any publication with a pan-cancer analysis of TXNIP from the perspective of overall tumors. Thus, we comprehensively examined the TXNIP gene in a total of 33 different tumors based on the data of TCGA, CPTAC and GEO databases through the molecular features of gene expression, genetic alteration, DNA methylation, protein ubiquitination or phosphorylation, and related pathways.

In most tumors of our study, the expression of TXNIP was lower than that in matched normal tissues. Nevertheless, for the gene of TXNIP, data on the survival prognosis analysis conducted between different databases suggested inconsistent conclusions for different tumors. For example, high expression of TXNIP was associated with a more favorable prognosis for patients with KIRC in GEPIA2, while there was no significant effect on prognosis in Kaplan-Meier-plotter and the PrognoScan database. Similarly, elevated TXNIP and a good prognosis of breast cancer and lung cancer were only correlated in the data sets of Kaplan-Meier-plotter and PrognoScan, but not of the GEPIA2. However, in these databases, we found consistent results regarding prognosis in most tumor types (breast cancer, lung cancer, liver cancer, bladder cancer, and UVM). STAD and CRC were exceptions where low levels of TXNIP expression showed a better prognosis. (Fig. 2A and 2O, Fig. S2T). The discrepancies in the prognostic role of TXNIP in various tumors in different databases may be a reflection in data collection and processing mechanisms and the underlying mechanisms pertinent to different biological properties. It’s worth noting that the prognostic values of TXNIP in gastric cancer were controversial (Fig. 2N, 2O). Interestingly, when we confined the data of gastric cancer patients with stage 2 to 4 disease, patients with lymph node metastasis, and patients treated with surgery alone, a correlation between decreased TXNIP and good survival outcomes was observed (Table S2). Thus, necessary clinical features should be fully taken into consideration as well. Together these findings powerfully draw a systematic prognostic landscape for TXNIP and demonstrate that TXNIP is a prognostic biomarker for breast, lung, liver, and bladder cancer.

MSI has been demonstrated to be associated with poor differentiation, proximal localization and chemotherapy efficacy of tumor²³^,²⁴. In this study, we first presented evidence of the potential correlation between TXNIP expression and MSI/TMB across all TCGA tumors. An MMR deficiency may result In MSI and carcinogenesis²⁶^,³⁴. Consistently, our findings showed that five genes related to the MMR of DNA were positively linked to TXNIP expression in many types of tumor (Fig. 3G, all P<0.01). DNA methylation, catalyzed by the DNMT family, is one of the most important epigenetic modifications, and its role in carcinogenesis has become a topic of considerable interest in recent years²⁵. Moreover, MSI has been reported to frequently occur in the hypermethylated cases²⁴. For KIRC, KIRP, PRAD and PCPG patients of TCGA, we observed a close correlation between up-regulated expression of TXNIP and down-regulated DNMTs using the Sangerbox platform (Fig. 3G, all P<0.05). A recent publication notes that TXNIP promoter methylation may mediate the carcinogenesis of human papillomavirus (HPV) -induced cervical cancer³⁵. All in all, these results congruously reveal the alteration characteristics of anti-oncogene TXNIP in different tumors and its potential association with tumorigenesis.

We applied multiple immune deconvolution approaches to obtain positive correlations between TXNIP expression and the immune infiltration levels of diverse TIICs in various types of cancers (Fig. 4A). We found that resting state macrophages (M0) displayed negative correlations with TXNIP in some specific TCGA tumors, but polarized macrophages, especially M2 macrophages, showed significant positive correlations. Given the fact that macrophage polarization status are intimately linked to their biological functions, our findings suggest a possible activating effect of TXNIP in the polarization of TAMs. Furthermore, BRCA, LUAD and KIRC were recognised as the three tumors with the most significant correlation both with oncologic outcomes and infiltrating immune cells score. CAFs were reported to regulate the function of various TIICs in TME³¹, and our findings first suggested the association of TXNIP expression and the estimated infiltration value of CAFs in tumors of BRCA, BRCA-Basal, COAD, LIHC, LUSC, READ, STAD, and TGCT (Fig. 4E–4M, all P < 0.05). In addition, based on the analyses of the immune and matrix scores, immune checkpoints, and tumor neoantigen, we obtained the specific immunologic landscape of TXNIP in different tumors. Altogether, our study provides insights in understanding the potential role of TXNIP in tumor immune microenvironment and its possible application in tumor immunotherapy.

ZBTB16, a member of the Krueppel (C2H2-type) zinc-finger transcription factor family, is a probable substrate-recognition component of an E3 ubiquitin-protein ligase complex which mediates the ubiquitination and subsequent proteasomal degradation of target proteins³⁶. Strikingly, an intersection analysis of the TXNIP-binding and expression-correlated genes indicated that the expression of ZBTB16 was closely related to that of TXNIP (Fig. 6D). Additionally, our protein modification sites analysis predicted that TXNIP has multiple modification sites, especially sites for ubiquitination and phosphorylation (Fig. 3H). To further conduct the KEGG and GO enrichment analyses, we integrated the information of TXNIP expression-correlated genes and TXNIP-binding components across all tumors, and identified the potential impact of “ubiquitin mediated proteolysis”, “protein dephosphorylation or phosphorylation”, “polyubiquitination”, “MAPK signaling pathway”, and “Wnt signaling pathway” in cancer pathogenesis.

In summary, our first pan-cancer analyses of TXNIP indicated statistical correlations of TXNIP expression with clinical prognosis, TMB/MSI, MMR genes, DNA methylation, immune cell infiltration, protein ubiquitination, and protein phosphorylation across multiple kinds of tumors, which aids in understanding the role of TXNIP in tumorigenesis from the perspective of clinical tumor samples. However, these results were based on data analysis. Further experimental verification will be carried out in our next study. Therefore, playing an important role in tumorigenesis and tumor immunity, TXNIP may be a potential prognosis and therapy biomarker.

The protocols used in this paper were approved by Nanfang Hospital’s Protection of Human Subjects Committee (NFEC-201809-K3). The undersigned authors warrant that the manuscript is original, and has neither been published nor is currently under consideration for publication by any other journal. The undersigned authors have seen and approved the final version of the manuscript, and transfer all copyright ownership of this manuscript to the publisher in event the paper is published.

Data availability statement

Publicly available data sets were analyzed in this study. All of the data in this article were used the TCGA datasets (https://www.cancer.gov/about-nci/organiza-tion/ccg/research/structural-genomics/tcga), GTEx datasets(https://www.genome.go-v/Funded-Programs-Projects/Genotype-Tissue-Expression-Project) and GEO databases(https://www.ncbi.nlm.nih.gov/geo/).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

This work was supported by grants from the National Natural Science Foundation of China (NO.12026605), the Guangdong Basic and Applied Basic Research Fund (2021A1515010992, 2020A1515110916), the Guangdong Medical Science and Technology Research Fund Project (A2020143), Guangzhou Basic and Applied Basic Research Fund (202102020163), the Foundation of Nanfang Hospital, Southern Medical University (2018C027), and the National Program on Key Research Project (2016YFC0901402).

Author contributions

Xuxue Guo: Formal analysis (Equal); Investigation (Equal); Methodology (Equal); Writing-original draft (Equal). Mei Huang, Haonan Zhang, Qianhui Chen: Data curation (Equal); Formal analysis (Equal); Visualization (Equal). Yan Meng, Ying Hu: Data curation (Equal); Investigation (Equal). Changjie Wu, Chenge Tu: Data curation (Equal); Visualization (Equal). Aimin Li, Yongfeng Liu: Conceptualization (Equal); Data curation (Equal); Methodology (Equal). Qingyuan Li: Project administration (Equal), Supervision (Equal) and Writing-review & editing (Equal). Peirong Zhou: Methodology (Lead), Supervision (Equal), Writing-review & editing (Equal). Side Liu: Conceptualization (Lead), Project administration (Equal) and Supervision (Lead).

Acknowledgements

This study was supported by Guangdong Provincial Bio-engineering Research Center for Gastroenterology Diseases.

Blum A, Wang P, Zenklusen JC. SnapShot: TCGA-Analyzed Tumors. Cell. 2018;173(2):530.
Clough E, Barrett T. The Gene Expression Omnibus Database. Methods Mol Biol. 2016;1418:93-110.
Hwang J, Suh HW, Jeon YH, et al. The structural basis for the negative regulation of thioredoxin by thioredoxin-interacting protein. Nat Commun. 2014;5:2958.
Mohamed IN, Hafez SS, Fairaq A, et al. Thioredoxin-interacting protein is required for endothelial NLRP3 inflammasome activation and cell death in a rat model of high-fat diet. Diabetologia. 2014;57(2):413-423.
Spindel ON, Burke RM, Yan C, et al. Thioredoxin-interacting protein is a biomechanical regulator of Src activity: key role in endothelial cell stress fiber formation. Circ Res.2014;114(7):1125-1132.
Hu J, Yu Y. The Function of Thioredoxin-Binding Protein-2 (TBP-2) in Different Diseases. Oxid Med Cell Longev. 2018;2018:4582130. Published 2018 May 2.
Alhawiti NM, Al Mahri S, Aziz MA, et al. TXNIP in Metabolic Regulation: Physiological Role and Therapeutic Outlook. Curr Drug Targets. 2017;18(9):1095-1103.
Li T, Fu J, Zeng Z, et al. TIMER2.0 for analysis of tumor-infiltrating immune cells. Nucleic Acids Res. 2020;48(W1):W509-W514. doi:10.1093/nar/gkaa407.
Tang Z, Kang B, Li C, et al. GEPIA2: an enhanced web server for large-scale expression profiling and interactive analysis. Nucleic Acids Res. 2019;47(W1):W556-W560.
Rhodes DR, Yu J, Shanker K, et al. ONCOMINE: a cancer microarray database and integrated data-mining platform. Neoplasia. 2004;6(1):1-6.
Chandrashekar DS, Bashel B, Balasubramanya SAH, et al. UALCAN: A Portal for Facilitating Tumor Subgroup Gene Expression and Survival Analyses. Neoplasia. 2017;19(8):649-658.
Nagy Á, Lánczky A, Menyhárt O, et al. Validation of miRNA prognostic power in hepatocellular carcinoma using expression data of independent datasets [published correction appears in Sci Rep. 2018 Jul 26;8(1):11515]. Sci Rep. 2018;8(1):9227. Published 2018 Jun 15.
Buechner P, Hinderer M, Unberath P, et al. Requirements Analysis and Specification for a Molecular Tumor Board Platform Based on cBioPortal. Diagnostics (Basel). 2020;10(2):93. Published 2020 Feb 10.
Le Clorennec C, Lazrek Y, Dubreuil O, et al. ITCH-dependent proteasomal degradation of c-FLIP induced by the anti-HER3 antibody 9F7-F11 promotes DR5/caspase 8-mediated apoptosis of tumor cells. Cell Commun Signal. 2019;17(1):106. Published 2019 Aug 23.
Ashton TM, McKenna WG, Kunz-Schughart LA, et al. Oxidative Phosphorylation as an Emerging Target in Cancer Therapy. Clin Cancer Res. 2018;24(11):2482-2490.
Liu X, Lu Y, Chen Z, et al. The Ubiquitin-Specific Peptidase USP18 Promotes Lipolysis, Fatty Acid Oxidation, and Lung Cancer Growth. Mol Cancer Res. 2021;19(4):667-677.
Zhang X, Lan Y, Xu J, et al. CellMarker: a manually curated resource of cell markers in human and mouse. Nucleic Acids Res. 2019;47(D1):D721-D728.
Jacobs J, Smits E, Lardon F, et al. Immune Checkpoint Modulation in Colorectal Cancer: What's New and What to Expect. J Immunol Res. 2015;2015:158038.
Gupta RG, Li F, Roszik J, et al. Exploiting Tumor Neoantigens to Target Cancer Evolution: Current Challenges and Promising Therapeutic Approaches. Cancer Discov. 2021;11(5):1024-1039.
Huang da W, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 2009;37(1):1-13.
Li Q, Lai Q, He C, et al. RUNX1 promotes tumour metastasis by activating the Wnt/β-catenin signalling pathway and EMT in colorectal cancer. J Exp Clin Cancer Res. 2019;38:334.
Samstein RM, Lee CH, Shoushtari AN, et al. Tumor mutational load predicts survival after immunotherapy across multiple cancer types. Nat Genet. 2019;51(2):202-206.
Hause RJ, Pritchard CC, Shendure J, et al. Classification and characterization of microsatellite instability across 18 cancer types. Nat Med. 2016;22(11):1342-1350.
Yang G, Zheng RY, Jin ZS. Correlations between microsatellite instability and the biological behaviour of tumours. J Cancer Res Clin Oncol. 2019;145(12):2891-2899.
Lyko F. The DNA methyltransferase family: a versatile toolkit for epigenetic regulation. Nat Rev Genet. 2018;19(2):81-92.
Buechner P, Hinderer M, Unberath P, et al. Requirements Analysis and Specification for a Molecular Tumor Board Platform Based on cBioPortal. Diagnostics (Basel). 2020;10(2):93. Published 2020 Feb 10.
Zhang X, Quan F, Xu J, et al. Combination of multiple tumor-infiltrating immune cells predicts clinical outcome in colon cancer. Clin Immunol. 2020;215:108412.
Robbins PF. Tumor-Infiltrating Lymphocyte Therapy and Neoantigens. Cancer J. 2017;23(2):138-143.
Ghatalia P, Gordetsky J, Kuo F, et al. Correction to: Prognostic impact of immune gene expression signature and tumor infiltrating immune cells in localized clear cell renal cell carcinoma. J Immunother Cancer. 2019;7(1):273. Published 2019 Oct 22.
Azimi F, Scolyer RA, Rumcheva P, et al. Tumor-infiltrating lymphocyte grade is an independent predictor of sentinel lymph node status and survival in patients with cutaneous melanoma. J Clin Oncol. 2012;30(21):2678-2683.
Sahai E, Astsaturov I, Cukierman E, et al. A framework for advancing our understanding of cancer-associated fibroblasts. Nat Rev Cancer. 2020;20(3):174-186.
Aran D, Sirota M, Butte AJ. Corrigendum: Systematic pan-cancer analysis of tumour purity. Nat Commun. 2016;7:10707. Published 2016 Feb 5.
Yoshihara K, Shahmoradgoli M, Martínez E, et al. Inferring tumour purity and stromal and immune cell admixture from expression data. Nat Commun. 2013;4:2612.
Cilona M, Locatello LG, Novelli L, et al. The Mismatch Repair System (MMR) in Head and Neck Carcinogenesis and Its Role in Modulating the Response to Immunotherapy: A Critical Review. Cancers (Basel). 2020;12(10):3006. Published 2020 Oct 16.
Kim MJ, Lee HJ, Choi MY, et al. UHRF1 Induces Methylation of the TXNIP Promoter and Down-Regulates Gene Expression in Cervical Cancer. Mol Cells. 2021;44(3):146-159.
Furukawa M, He YJ, Borchers C, et al. Targeting of protein ubiquitination by BTB-Cullin 3-Roc1 ubiquitin ligases. Nat Cell Biol. 2003;5(11):1001-1007.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

A Pan-cancer Analysis of Thioredoxin-interacting Protein as an Immunological and Prognostic Biomarker

Archived Versions:

Version 1

Abstract

Figures

1. Introduction

2. Material And Methods

3. Results

4. Discussion

5. Conclusion

Declarations

References

Additional Declarations

Supplementary Files

Archived Versions:

Version 1