Gene expression profiling and protein–protein network analysis revealed prognostic hub biomarkers linking cancer risk in type 2 diabetic patients

Type 2 diabetes mellitus (T2DM) and cancer are highly prevalent diseases imposing major health burden globally. Several epidemiological studies indicate increased susceptibility to cancer in T2DM patients. However, genetic factors linking T2DM with cancer have been poorly studied. In this study, we followed computational approaches using the raw gene expression data of peripheral blood mononuclear cells of T2DM and cancer patients available in the gene expression omnibus (GEO) database. Our analysis identified shared differentially expressed genes (DEGs) in T2DM and three common cancer types, namely, pancreatic cancer (PC), liver cancer (LC), and breast cancer (BC). The functional and pathway enrichment analysis of identified common DEGs highlighted the involvement of critical biological pathways, including cell cycle events, immune system processes, cell morphogenesis, gene expression, and metabolism. We retrieved the protein–protein interaction network for the top DEGs to deduce molecular-level interactions. The network analysis found 7, 6, and 5 common hub genes in T2DM vs. PC, T2DM vs. LC, and T2DM vs. BC comparisons, respectively. Overall, our analysis identified important genetic markers potentially able to predict the chances of PC, LC, and BC onset in T2DM patients.

differentially expressed genetic markers that could confer T2DM susceptibility were identified 4 .Subsequent bioinformatics analysis of these differentially expressed genes has revealed the genetic association of T2DM with these co-morbidities 5 .These findings have advanced our understanding of complications arising due to T2DM and have prospective applications in designing personalized prognostic and diagnostic tools for such heterogenic human diseases.
Cancer is another heterogenic disease that is also the second leading cause of human death 6 .It is characterized by unrestricted growth of abnormal cells.In some cases, these abnormal cells could metastasize to other parts of the human body.Liver, pancreatic, and breast cancers are among the most common cancer types 7 .It is wellknown that T2DM and many common cancers share several risk factors like aging, obesity, and an unhealthy lifestyle 8 .Different epidemiological studies in the past suggest that T2DM condition increases the risk of several cancers, including liver, pancreatic 9 , breast 10,11 , and endometrial 12 .They report standardized incidence ratios to indicate an increased risk of cancers in T2DM patients.Pancreatic and liver cancers showed the highest standardized incidence ratios in different populations of T2DM patients from Denmark, Tyrol/Austria, Taiwan, Sweden, Australia, the Chinese mainland 13 , Finland 14 , and Lithuania.In addition, a few meta-analysis studies reported an increased risk of breast cancer in diabetic women 10,11 .There is no clear molecular understanding of T2DM link to specific cancer types yet.However, the state of insulin resistance, hyperinsulinemia, hyperglycemia, chronic inflammation, and increased oxidative stress in T2DM could probably elicit mitogenic pathways and cause these cancers 15 .Moreover, a few Mendelian randomization studies indicate a positive association between T2DM and the risk of pancreatic, breast, lung, liver, and kidney cancer 16,17 .Despite the availability of extensive evidence from epidemiological and meta-analysis that links cancer risk to T2DM, a systematic study of the shared genetic markers possibly predisposing this risk in T2DM patients is lacking for the common cancer types, namely pancreatic (PC), liver (LC), and breast (BC) cancer.
In this work, we performed gene expression analysis to identify predominant differential expressed genes (DEGs) from the peripheral blood mononuclear cell (PBMC) samples of T2DM patients, posing a risk towards three common cancer types (PC, LC, and BC).Their functional enrichment analysis indicated the involvement of gene expression, cell transport, and oxidation pathways.The protein-protein interaction (PPI) network provided common hub genes between T2DM and the three cancer types.We identified TGFB1 as a common hub gene between T2DM and PC/LC, significantly affecting survival in cancer patients.Therefore, the identified hub genes have a potential prognostic and therapeutic value in patients with T2DM patients and high cancer risks.

Shared DEGs in T2DM and three common cancer types
Gene expression data of Homo sapiens in different diseased conditions were obtained from gene expression omnibus (GEO) database.Table 1 provides the summary of three different datasets used in our study.We employed a three-tiered filtering criterion to identify the shared DEGs between T2DM and three cancer types (Fig. 1).The raw gene expression datasets were normalized for each GEO study using the GEO2R tool (Supplementary Fig. S1).We identified 94 DEGs shared between T2DM and LC using our filtering criterion (Supplementary Fig. S2).Of these, 59 DEGs were upregulated, while 35 were downregulated (Supplementary Fig. S3).Likewise, for T2DM vs. BC comparison, we identified 16 shared DEGs (Supplementary Fig. S2), including 8 upregulated and 8 downregulated (Supplementary Fig. S3).For T2DM vs. PC, the three stringent filtering criteria resulted in an insignificant number of shared DEGs.However, using FDR ≤ 0.1 at filter 1, we identified 66 shared DEGs (Supplementary Fig. S2).Interestingly, the GSE15932 dataset also has data from 8 patients suffering from T2DM and PC disease.We applied the first two filtering criteria on T2DM & PC dataset, which identified 1203 DEGs (Supplementary Fig. S2).Surprisingly, we observed 69% (46 DEGs) shared DEGs of T2DM vs. PC overlapping with the DEG identified from T2DM & PC.For further functional analysis, we proceeded with the above 46 DEGs in T2DM vs. PC, where 26 were upregulated, and 20 were downregulated (Supplementary Fig S3).Similar validation could not be performed for T2DM vs. LC and T2DM vs. BC due to the unavailability of such a dataset.We observed a strong Pearson's correlation (0.98 for T2DM vs. PC, 0.90 for T2DM vs. LC, and 0.87 for T2DM vs. BC) for the identified shared DEGs between T2DM and three common cancer types (i.e., PC, LC, and BC).Table 2 summarizes the number of shared DEGs narrowed down after applying our filtering criteria to each dataset.www.nature.com/scientificreports/

Functional enrichment and pathway analysis of shared DEGs
To understand the functional relevance of identified common genes between T2DM and three cancers (PC, LC and BC), we utilized GO biological process and KEGG signaling pathway analysis tools from GENECODIS software.The GO biological process of the shared DEGs in T2DM vs. PC comparison showed enrichment of vesicle-mediated transport, protein export from the nucleus, engulfment of target by autophagosome, positive regulation of cellular protein metabolic process, positive regulation of protein localization to nucleus, etc. (Table 3).The KEGG pathway analysis results were insignificant for the T2DM vs. PC comparison.In the T2DM vs. LC comparison, the shared DEGs revealed enrichment in several critical biological processes such as ATP biosynthetic process, peptide modification, regulation of RNA splicing, regulated exocytosis, and neutrophil degranulation (Table 4).We also found enrichment of KEGG pathways shown in Supplementary Table S1, (section T2DM vs. LC), which includes synaptic vesicle cycle (hsa04721, p = 5.63 × 10 −2 ), rheumatoid arthritis (hsa05323, p = 5.63 × 10 −2 ), collecting duct acid secretion (hsa04966, p = 5.63 × 10 −2 ), phagosome (hsa04145, p = 5.63 × 10 −2 ), and oxidative phosphorylation (hsa00190, 9.05 × 10 −2 ).Similarly, analysis of common DEGs in T2DM vs. BC comparison led to the enrichment of hydrogen peroxide catabolic process, cellular oxidant detoxification, autophagic cell death, regulation of DNA recombination, hemoglobin biosynthetic process, and plasminogen activation from GO biological process (Table 5).The KEGG pathway analysis results were insignificant for the T2DM vs. BC comparison.Overall, cellular transport, gene expression, and cellular oxidation pathways were affected in this analysis, which motivated us further to perform the interaction analysis at the molecular level.

Deducing molecular level interactions using PPI network
The above analysis linked identified common genes to specific biological pathways, which intrigued us to investigate their relationship at the molecular level.We constructed the protein-protein interaction (PPI) networks of identified common DEGs between T2DM and three cancer types (PC, LC, and BC).Our analysis yielded 16 nodes (genes) in T2DM vs. PC, 25 in T2DM vs. LC, and 5 in T2DM vs. BC in the main connected PPI networks (Supplementary Table S2, Fig. 2a-c).
In the PPI network analysis, the node size reflects their degree, and node color indicates the expression pattern, thus making it possible to deduce certain hub genes for respective disease conditions.Moreover, network analysis provides information on the important hub genes and enriched processes and pathways involving Table 3. Shows top significantly enriched GO:BP involving the identified DEGs for T2DM vs. PC patients.FDR: false discovery rate, DEGs: differentially expressed genes, Count: number of DEGs.

Description Annotation ID Count FDR DEGs
Vesicle-mediated transport GO:0016192  www.nature.com/scientificreports/major hub nodes as represented by a colored ring around each node (Fig. 2a-d).Our network analysis revealed important genes associated with more than one biological pathway, thus suggesting their evident involvement in the respective disease conditions.The important genes from the T2DM vs. PC network included, HIST2H2AA3, HIST2H2AA4, NFKBIA, SESN2, SMURF1, TGFB1, TNRC6A (Fig. 2a).The T2DM vs. BC network could not reveal the genes associated with the biological pathways due to insufficient DEGs (Fig. 2b).For the T2DM vs. LC network, ALDOA, ATP6V0D1, ATP6V0C, ATP6V0E1, TGFB1, CYBA, CTSD, and DBNL genes were considered significant (Fig. 2c).Interestingly, a common gene, TGFB1, between T2DM vs. PC and T2DM vs. LC, is upregulated in both conditions.This gene codes for a growth factor in cell proliferation, differentiation, and death.We used these findings as a basis for further investigation of hub genes.

Hub genes identification and survival analysis
We ranked the nodes in our PPI network analysis based on eleven different topological features using the cyto-Hubba plugin in the Cytoscape tool (see details in the materials and methods section).Thus, we identified the top 15 genes for T2DM vs. PC and T2DM vs. LC (Supplementary Table S3).We noticed several hub genes such as HIST2H2AA4, SESN2, and TNRC6A for T2DM vs. PC and ATP6V0D1, ATP6V0C and TGFB1 for T2DM vs. LC that were top-ranked in almost all the computed topological features.This analysis was not performed for T2DM vs. BC due to the insufficient number of identified common DEGs.We further narrowed common hub genes based on their top ranking and commonness across the computed topological features.Accordingly, we could identify seven genes as the hub genes for the T2DM vs. PC comparison.The relative log 2 fold expression of these genes in two diseased conditions is represented in Fig. 3a.For the T2DM vs. LC comparison, we identified six hub genes.We found similar expression levels of these genes in T2DM and LC disease conditions, as indicated in Fig. 3b.For the T2DM vs. BC comparison, five genes identified from PPI network analysis were considered the hub genes, as represented in Fig. 3c.We identified 17 hub genes combined for the three comparisons, which could serve as potential biomarkers.Further, survival analysis was also performed for all the common hub genes using a web resource UALCAN, which analyzes publicly available cancer OMICS data 18,19 .The survival analysis revealed 4 hub genes (ATP6V0C, p < 0.051; ATP6V0D1, p < 0.02; ATP6V0E1, p < 0.0002 and TGFB1, p = 0.042) with significant p-value ≤ 0.05 to be linked with poor survival (Supplementary Fig. S4).The expression levels of these four hub genes were similar to our analysis.Interestingly, we found a common hub gene TGFB1 between T2DM vs. pancreatic cancer and T2DM vs. liver cancer patients, which significantly affects survival.In the case of breast cancer samples, the diagnosis was benign, so, likely, the survival is not dependent on the identified common hub signatures for these samples.
We further validated the identified prognostic 17 hub markers from our study using three additional publicly available datasets for T2DM patients (Supplementary Fig. S5a).Our validation analysis revealed overlapping functionally-enriched gene ontology biological processes (Supplementary Fig. S5b-e).Out of the 17 hub genes, we could validate 12 genes (p-value ≤ 0.05) based on their expression profile (Supplementary Fig. S5f), despite differences in their sample sources (whole blood sample vs. PBMC) and variability in detection ranges due to different platforms (Supplementary Fig. S6).

Discussion
T2DM and cancer have burdened the health sector throughout the world.Recently, several epidemiological studies have indicated a causal link between T2DM and common cancer types.However, the genetic association of T2DM to these cancers remains largely unknown.Our work identifies a genetic association between T2DM and three common cancer types, i.e., PC, LC, and BC.Our analysis identified 7, 6, and 5 hub genes showing a correlation between T2DM and the three cancers, i.e., PC, LC, and BC, respectively.
The KEGG pathway and GO biological process analyses showed enrichment of vesicle-mediated transport, vital in tumor microenvironment remodeling and transport of secretory insulin or other circulating mediators in diabetes [20][21][22][23] .Neutrophil deregulation is known to be associated with diabetes 24 and cancer cell progression, metastasis, and activating dormant cancer cell 25 .Moreover, we noticed positive regulation of metabolic and www.nature.com/scientificreports/www.nature.com/scientificreports/cellular protein metabolic processes in T2DM vs. PC conditions.Our analysis also revealed important processes contributing to gene expression, including regulation of RNA splicing and protein degradation, which are required to maintain homeostasis.In nutshell, we identified seven hub genes (HIST2H2AA3, HIST2H2AA4, NFKBIA, SESN2, SMURF1, TGFβ1, TNRC6A) in T2DM vs. PC, six common hub genes (ATP6V0D1, ATP6V0C, ATP6V0E1, CTSD, CYBA, TGFβ1) in T2DM vs. LC and five common hub genes (ALYREF, CDKN2D, NGDN, THRAP3, UBE2M) in T2DM vs. BC, respectively (Fig. 3).Supplementary Table S4 provides the detailed function and description of these hub genes.Noteworthy, The NFKBIA gene involved in the NFKB pathway has a crucial role in the initiation and progression of PC 25 , and its gene polymorphism has an implicit role in the prognosis of T2DM 26 .TGFB1 is involved at an early and advanced stage in liver tumorigenesis 27 , and TGF-β signaling plays diverse roles in β cell development and functioning that has an impactful role in diabetic condition 28 .The V-ATPase H+ transporting genes (ATP6V0D1, ATP6V0C, ATP6V0E1) maintain the intracellular pH and thus are critical for the Warburg effect observed in the cancer cells 29 .The CTSD gene has a role in decreasing the expression of IGFBP3, contributing to mitogenesis in hepatoma cells 30 , and it also has increased plasma activity in T2DM male patients 31 .The ALYREF gene plays a significant role in cellular growth, apoptosis, and mitochondrial energy metabolism in BC 32 , and as a 5mC-related gene that could have a functional role in T2DM 33 .THRAP3 deficiency sensitizes BC cells, suggesting a probable involvement in DDR 34 .Also, THRAP3 plays a direct role in controlling diabetic gene programming by interacting with PPARγ 35 .Thus, our analysis provides insight into biological and molecular events that could link T2DM with three common cancer types (PC, LC, and BC).Further, the identified genetic markers hold the potential to predict the chances of cancer onset in T2DM patients.However, systematic approaches for data collection, which consider variations in genetic profiling based on ethnicity, sex, and age, could further expand our understanding.Notably, such markers in T2DM patient PBMC samples predisposing to increased cancer risk could help diagnosis at an early stage and provide benefits for developing personalized therapeutic strategies.

Figure 1 .
Figure 1.Schematic representation of the workflow used to identify differentially expressed genes (DEGs) and narrow down to the common hub genes in T2DM vs. respective cancers datasets.

Figure 2 .
Figure 2. Protein-protein interaction network for (a) T2DM vs. PC, (b) T2DM vs. BC, and (c) T2DM vs. LC.The colored circle around the nodes represents different enriched pathways to which these nodes are linked.The small nodes (in white color) indicate additional interactors.The node color indicates overexpressed genes (red) and under-expressed genes (blue).The node size represents the node's degree, and the node's color intensity represents the log 2 fold change (FC) value of differentially expressed genes (DEGs).The color chart (d) illustrates different enriched pathways and their annotation identity.

Figure 3 .
Figure3.The barplot (a-c) shows log 2 fold change expression of identified hub genes in T2DM vs. PC, T2DM vs. LC, and T2DM vs. BC comparison, respectively.The volcano plots (d-f) highlight the common hub genes for T2DM vs. PC, T2DM vs. LC, and T2DM vs. BC, respectively.The symbols represent whether the p-value and log 2 FC value of that particular DEG is for T2DM (triangle) or three cancers (circle).The shared hub genes are labeled in black, while other DEGs are shown in lighter colors.

Table 1 .
Overview of GEO datasets used in this study.The dash (-) symbol indicates that the information is not available.T2DM: type 2 diabetes mellitus, PC: pancreatic cancer, LC: liver cancer, BC: breast cancer.

Table 2 .
Summary of the number of genes filtered with three-tiered filtering criteria (statistical significance (filter 1; adjusted p-value ≤ 0.05 unless specified), biological significance (filter 2; log 2 FC ≥ |0.5|) and 10% relative distance around linear regression line of the correlated gene (filter 3).

Table 5 .
Shows top significantly enriched GO:BP involving the identified DEGs for T2DM vs. BC comparison.