Abstract

Background

   Diabetes is considered to be a risk factor for colon cancer (CC), and
   CC patients with diabetes tend to have a worse prognosis. However, the
   underlying mechanism of this condition remains unclear. This study aims
   to elucidate the relationship between diabetes and CC further, and to
   find effective therapeutic targets.

Methods

   Transcription and clinical information data were acquired from the Gene
   Expression Omnibus (GEO) database, accessed differentially expressed
   genes (DEGs) between different groups, and enriched function and
   pathway by Gene Ontology (GO) and Kyoto Encyclopedia of Genes and
   Genomes (KEGG) analyses. We conducted a weighted gene co-expression
   network analysis (WGCNA) to obtain significant modules and hub genes of
   diabetic-related CC. A receiver operating characteristic curve (ROC)
   analysis and Kaplan-Meier plotter were performed to diagnosis and
   prognosis prediction. Using the Connectivity Map (CMap) database to
   predict small molecule compounds, and employed molecular docking to
   simulate the binding conformation of the potential agent and key
   targets. Moreover, CIBERSORT was used to depict the immune infiltration
   in diabetic-related CC. The correlations between tumor mutation burden
   score, microsatellite instability score, and hub DEGs expression were
   performed by Spearman’s correlation.

Results

   In this study, 633 DEGs were identified from the tumor (n = 42) and the
   normal colon mucosa samples (n = 42), and 133 DEGs were identified from
   type 2 diabetes mellitus (T2DM) (n = 46) and non-T2DM samples (n = 38).
   We obtained a gene module including 1183 genes significantly related to
   CC patients with diabetes, and finally, the intersection of
   tumor-associated DEGs, diabetes-associated DEGs, and WGCNA identified
   11 hub DEGs. The hub DEGs had great diagnostic and prognostic values
   for CC and diabetes. We found the small-molecule compound NVP-BEZ235
   according to its high binding affinity to the targets and exhibited the
   molecular docking landscape including CDC42BPA, COX6A1, PON2, TM9SF2,
   UBBE2K, UBR2, ZC3H14, and ZNF106. In addition, we found the
   immune-infiltrating differences between CC patients with diabetes and
   those without diabetes. The expression of hub DEGs was significantly
   correlated with tumor mutation burden and microsatellite instability.

Conclusion

   Diabetes plays an important role in CC pathogenesis, and NVP-BEZ235 may
   be a promising therapeutic drug for CC patients with diabetes.

   Keywords: Colon cancer, Diabetes, Molecular docking, Immune
   infiltration, Consensus clustering

Introduction

   Global cancer burden indicated colorectal cancer ranked third as the
   most common tumor and was the second main cause of cancer death, colon
   cancer (CC) accounting for about two-thirds of all these cases [[32]1].
   Recent studies showed that the mean five-year survival rate for CC
   fluctuated around 50–60%, which still deserved people’s attention and
   needed to be improved [[33]2, [34]3].

   Diabetes mellitus has emerged as one of the fastest-growing global
   health challenges. According to the International Diabetes Federation,
   more than 537 million adults were living with diabetes in 2021, a
   figure projected to rise to 783 million by 2045 [[35]4]. The global
   burden of diabetes has also increased significantly, with over
   1.5 million deaths reported in 2019 and a substantial rise in
   disability-adjusted life years [[36]5]. This metabolic disorder not
   only contributes to microvascular and macrovascular complications but
   is also increasingly recognized as a factor influencing cancer
   development and progression. Diabetes intimately binds up with cancer,
   8–18% of cancer patients have diabetes as a comorbid medical condition,
   while a higher cancer-related mortality rate is shown in diabetes
   patients. It has been reported that patients with diabetes are 30% more
   likely to develop colorectal cancer than those without diabetes, and
   70% more likely to develop proximal CC [[37]6]. Many factors make
   diabetes and CC linked, and hyperinsulinemia is considered to be the
   most important [[38]7]. Previous studies have suggested that diabetes
   is a risk factor for CC, and can lead to a worse prognosis [[39]8,
   [40]9].

   The treatment of CC is largely guided by the stage of the disease,
   surgery is the primary treatment for most of them currently, while
   chemotherapy is most commonly used as adjuvant therapy [[41]10]. The
   results of a meta-analysis suggested that insulin therapy for diabetes
   may increase this risk of CC incidence. Recent studies suggest that the
   use of metformin may attenuate the adverse effects of diabetes on the
   prognosis of patients with CC [[42]9]. Some common targets of CC and
   diabetes were found in recent years, including liver X receptors,
   histone deacetylase inhibitors, glucose transporters, peroxisome
   proliferator activator receptors, dipeptidyl peptidase-IV inhibitors,
   cyclin-dependent kinase 4 inhibitors, estrogen receptors, mechanistic
   target of rapamycin, insulin-like growth factor receptors, and Hippo
   pathway [[43]7, [44]11]. However, there is still a shortage of drugs
   that can show positive effects in the treatment of both diabetes and
   CC. This study focuses on the common molecular mechanisms of CC and
   diabetes in an attempt to help find specific targeted therapy for CC
   patients with diabetes.

   We identified the potential targets for CC associated with diabetes. In
   addition, we conducted consensus clustering, immune infiltration, and
   molecular docking analyses to describe the impact of diabetes on CC.
   The underlying mechanisms between diabetes and CC are expected to
   provide a new perspective on therapeutic strategies for these patients.

Materials and methods

Data acquisition

   The RNA-seq transcriptome information was obtained from the dataset
   [45]GSE115313 in the Gene Expression Omnibus (GEO)
   ([46]http://www.ncbi.nlm.nih.gov/geo) database. This dataset contained
   paired tumor and normal colon mucosa samples of 42 CC patients, 23 of
   them with T2DM.

DEGs identification and function enrichment

   The R package “limma” was employed to identify differentially expressed
   genes (DEGs), set p < 0.05 and |log2 fold change (FC)| >1.5 as filter
   condition in the comparison between tumor (n = 42) and normal colon
   mucosa samples (n = 42), while set p < 0.05 and |log2 (FC)| >1.2 as
   filter condition in the comparison between T2DM (n = 46) and non-T2DM
   samples (n = 38). Gene Ontology (GO) functional enrichment analyses and
   Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways enrichment
   analyses were performed by using the R package ‘‘clusterProfiler’’.

Weighted gene co-expression network analysis

   To find the co-expressed gene modules with high biological significance
   and explore the relationship between gene networks and CC patients with
   diabetes, we utilized the “WGCNA” R package to conduct weighted gene
   co-expression network analysis (WGCNA). First, we performed a
   hierarchical cluster analysis to exclude outlier samples, picked
   optimal β by the “pickSoftThreshold” function to meet the scale-free
   distribution, and constructed a weighted adjacency matrix. Then the
   weighted adjacency matrix was transformed into a topological overlap
   matrix (TOM) and 1-TOM matrix, and the genes were divided into
   different modules by the hierarchical clustering method. Finally,
   module eigengene (ME) was calculated to represent the gene expression
   profile of each module, and the correlation between the ME and clinical
   features was evaluated by Pearson correlation coefficients.

Diabetic and CC-related DEGs

   Using Venn diagrams to intersect tumor-associated DEGs,
   diabetes-associated DEGs, and genes from specific modules obtained by
   WGCNA, and obtained diabetic and CC-related hub DEGs. Exhibited the
   expressional differences of hub DEGs between tumor and normal samples
   by violin plot, as well as diabetic and non-diabetic samples. Next, we
   employed the “pROC” R package to perform receiver operating
   characteristic curve (ROC) analysis and evaluated the efficacy of hub
   DEGs.

   In addition, the Kaplan-Meier plotter (KM plotter,
   [47]http://kmplot.com/analysis/) can evaluate the survival condition of
   different genes in 21 cancer types including CC [[48]12]. All of the CC
   patients in this dataset were divided into high and low-expression
   groups according to the median expression of hub DEGs, and we used RFS
   (Recurrence-free survival) to evaluate the prognostic value of related
   hub DEGs in CC.

Potential therapeutic small molecule agents and molecular-ligand docking
analysis

   The connectivity map database (CMap, [49]https://clue.io) is a drug
   prediction database based on DEGs, that can help identify potential
   drug candidates for the treatment of certain diseases [[50]13].
   Selected DEGs of diabetic and non-diabetic samples in CC were imported
   into this platform to acquire potentially small molecules.
   Subsequently, three steps were conducted for molecular docking. First,
   we input candidate drugs into the PubChem database to obtain their
   three-dimensional structures. Second, Selected DEGs were submitted into
   the Protein Data bank databases (PDB, [51]http://www.rcsb.org) to get
   the highest-resolution receptor structures. Third, AutoDock Tools
   software (version 1.5.7) was used to perform molecular linkage,
   illustrating how small molecule compounds bind to their targets.

Immune-infiltrating analysis

   CIBERSORT is a deconvolution algorithm that can estimate the relative
   abundance of 28 immune cell subtypes in each sample. The R package
   “corrplot” was applied to figure out the relationship between hub DEGs
   and immune cells, and the results were visualized using the “pheatmap”
   R package. The Immune-infiltrating patterns of the tumor and the normal
   colon mucosa group in diabetic-related CC were compared by Student’s
   t-test, and presented the differences using violin plot.

Tumor mutation burden and microsatellite instability

   The information about microsatellite instability (MSI) and tumor
   mutation burden (TMB) was obtained from the TCGA database
   ([52]https://cancergenome.nih.gov/), and 527 colorectal cancer patients
   were selected for analysis. We used the cBioPortalData package to
   download clinical data including microsatellite instability (MSI), and
   used the TCGAbiolinks package to extract the TCGA mutation data and
   calculate tumor mutation burden (TMB). The correlations between TMB
   score, MSI score and hub DEGs expression were performed by Spearman’s
   correlation, and the correlation analysis plots were drawn using the
   “ggplot2” R package.

Statistical analysis

   In this study, all data processing and statistical analysis were
   performed using R software (version 4.2.1,
   [53]https://www.r-project.org). A T-test was applied to compare the
   differences between the two groups. Spearman’s correlation was
   conducted for correlation analysis. And p < 0.05 was determined as
   statistically significant.

Results

DEGs and functional annotation and pathway enrichment

   We used the “limma” R package to identify DEGs. When comparing the
   tumor (n = 42) and normal colon mucosa (n = 42) samples of CC patients
   enrolled in this study, 633 DEGs were identified, which included 28
   up-regulated genes and 605 down-regulated genes (Fig. [54]1A). Besides,
   we identified 133 DEGs from the colon mucosa samples of CC patients
   with T2DM (n = 46) and those without T2DM (n = 38), among which 60 were
   upregulated and 73 were downregulated (Fig. [55]1E). Expression pattern
   of DEGs and the relative consistency within the groups exhibited by
   corresponding heatmap (Fig. [56]1B, F).

Fig. 1.

   [57]Fig. 1
   [58]Open in a new tab

   Identification of DEGs and functional annotation and pathway enrichment
   of DEGs. (A) Volcano map of DEGs between the tumor (n = 42) and the
   normal colon mucosal (n = 42) in the CC patients (|log2FC|>1.5,
   p < 0.05). (B) Heatmap of DEGs between the tumor and the normal colon
   mucosal in the CC patients. (C) GO enrichment analysis of DEGs between
   the tumor and the normal groups. (D) KEGG pathway enrichment analysis
   of DEGs between the tumor and the normal groups. (E) Volcano map of
   DEGs between diabetic (n = 46) and non-diabetic (n = 38) groups in the
   colon mucosal of CC patients (|log2FC|>1.2, p < 0.05). (F) Heatmap of
   DEGs between diabetic and non-diabetic groups in the colon mucosal of
   CC patients. (G) GO enrichment analysis of DEGs between the diabetic
   and non-diabetic groups. (H) KEGG pathway enrichment analysis of DEGs
   between the diabetic and non-diabetic groups

   According to the results of GO enrichment analysis, we found that the
   DEGs in tumor and normal groups were significantly enriched in
   “estrogen response late”, “adipogenesis”, and “epithelial mesenchymal
   transition” (Fig. [59]1C), while the DEGs in diabetic and non-diabetic
   groups were significantly enriched in “E2F targets”, “oxidative
   phosphorylation”, and “P53 pathway” (Fig. [60]1G). The KEGG enrichment
   analysis results indicated that the pathways such as “ribosome”,
   “non-alcoholic fatty liver disease”, and “FoxO signaling pathway” might
   be potential mechanisms of CC (Fig. [61]1D). The pathways such as
   “spliceosome”, “parkinson disease”, and “NOD-like receptor signaling
   pathway” might be potential mechanisms of diabetes-related CC
   (Fig. [62]1G).

Identification of key modules in CC patients with diabetes via weighted gene
co-expression network analysis

   WGCNA was performed to construct gene co-expression networks of
   diabetes-related CC, related analysis was based on the data of paired
   tumor and normal colon mucosa samples in CC patients with diabetes. We
   set the soft threshold β to 8 with a scale-free index of 0.85 to
   construct the gene modules (Figs. [63]2A, B). Four gene modules were
   obtained using a dynamic shear tree algorithm (Fig. [64]2C) and module
   feature vector clustering analysis (Fig. [65]2F). In addition,
   Fig. [66]2D displayed the results of sample clustering analysis. Using
   WGCNA, we found that the green and blue modules were the most
   significant gene modules (Fig. [67]2E). Considering there was only a
   small number of genes (n = 29) in the green module, we selected the
   blue module as the key gene module which had the highest correlation
   with diabetes-related CC (R = 0.56, p = 6.1e-5). A total of 1183 genes
   were identified in the blue gene module for subsequent analysis, we
   calculated the correlations between module membership and gene
   significance in the blue module, and found a notable positive
   correlation (R = 0.51, p < 0.001) as well (Fig. [68]2G).

Fig. 2.

   [69]Fig. 2
   [70]Open in a new tab

   WCGNA screens for key modules. (A) Scale-free topological index at
   different soft-thresholding powers. (B) Mean connectivity at different
   soft-thresholding powers. (C-D) Hierarchical clustering under optimal
   soft-thresholding power. (E) Correlations between gene modules and
   diabetic-related CC. (F) The heatmap of the eigengene adjacency. (G)
   Scatter plot and correlation between the blue module memberships and
   the gene significance

Identification of hub DEGs

   We further conducted analysis based on the intersection of genes from
   “Limma” and WGCNA module genes (Fig. [71]3A). A total of 11 hub DEGs
   were identified using Venn diagrams, the correlations among hub DEGs
   were demonstrated in Fig. [72]3B, which suggested significant
   synergistic effects. The MCODE sub-network of the 11 hub DEGs was shown
   in Fig. [73]3C. We draw the expression pattern of 11 hub DEGs among
   different groups by box plots, results showed that the expression of
   all hub DEGs was different between tumor and normal groups
   (Fig. [74]3D), and the expression of most hub DEGs (n = 8) were
   different between diabetic and non-diabetic groups (Fig. [75]3E).
   Furthermore, we performed ROC analysis and used the area under the
   curve (AUC) to evaluate the diagnostic value of hub DEGs for CC
   (Fig. [76]3F) and diabetes (Fig. [77]3G). The gene ZNF106 had the
   largest AUC value (AUC = 0.874) for the diagnosis of CC, while the gene
   CDC42BPA had the largest AUC value (AUC = 0.780) for the diagnosis of
   T2DM.

Fig. 3.

   [78]Fig. 3
   [79]Open in a new tab

   Identification of hub DEGs and ROC analysis. (A) Venn diagram showed
   the overlapping hub genes between tumor-associated DEGs,
   diabetes-associated DEGs, and genes of the blue module. (B) Heatmap
   showing pairwise correlation among the 11 hub DEGs across samples. (C)
   MCODE sub-network analysis visualizing functional interaction among hub
   DEGs. (D) Different expressions of the hub DEGs between the tumor and
   the normal groups. (E) Different expressions of the hub DEGs between
   diabetic and non-diabetic groups. (F) ROC curves for each hub DEG
   evaluating their diagnostic performance in distinguishing tumor from
   normal tissue. (G) ROC curves for hub DEGs assessing their diagnostic
   performance in identifying diabetes status among CC patients.
   *p < 0.05, **p < 0.01, ***p < 0.001

Prognostic value estimation of hub DEGs

   The result of KM-plot analysis revealed that most of the 11 hub DEGs
   had great predictive value for the prognostic of CC patients. We found
   that the higher expression levels of ABI1, CDC42BPA and ZNF106 were
   correlated with worse survival (Fig. [80]4A, B, C), whereas the lower
   expression levels of COX6A1, IP6K2, PON2, TM9SF2, TMED2, UBE2K, UBR2,
   and ZC3H14 were correlated with worse survival (Fig. [81]4D-K).

Fig. 4.

   [82]Fig. 4
   [83]Open in a new tab

   Prognostic value of hub DEGs. The Kaplan-Meier plot analysis of the hub
   DEGs had great predictive value for the prognostic of CC patients. (A)
   ABI1. (B) CDC42BPA. (C) ZNF106. (D) COX6A1. (E) IP6K2. (F) PON2. (G)
   TM9SF2. (H) TMED2. (I) UBE2K. (J) UBR2. (K) ZC3H14

Prediction of potential therapeutic drugs for CC patients

   We selected the 20 top-up-regulated and the 20 top-down-regulated DEGs
   of diabetic and non-diabetic samples in CC patients, and imported them
   into the CMap database to access promising small molecule compounds.
   Potential small-molecule drugs were listed in Fig. [84]5A, which could
   be useful in the treatment of CC patients with diabetes. The
   corresponding three-dimensional structures of the top 6 drug candidates
   were acquired from the PubChem database (Fig. [85]5B). We chose the
   NVP-BEZ235 as the potential therapeutic drug for the next analysis
   based on the correlation scores between drugs and genes.

Fig. 5.

   [86]Fig. 5
   [87]Open in a new tab

   Prediction of potential therapeutic drugs for CC patients. (A)
   Connectivity Map analysis results showing mechanisms of action of top
   small-molecule compounds predicted based on diabetes-related DEGs. (B)
   Three-dimensional molecular structures of the top six candidate
   compounds obtained from the PubChem database

The molecular docking landscape on NVP-BEZ235

   To get the best conformation of small molecule compounds and target
   molecules for interaction, eight molecular targets with their
   highest-resolution receptor structures were downloaded from the PDB
   database including CDC42BPA, COX6A1, PON2, TM9SF2, UBBE2K, UBR2,
   ZC3H14, and ZNF106. We used AutoDock Tools software to perform
   molecular dock. Strong binding interactions were observed between
   NVP-BEZ235 and all 8 target molecules, the binding poses and sites were
   shown in Fig. [88]6A-H.

Fig. 6.

   [89]Fig. 6
   [90]Open in a new tab

   The molecular docking landscape on NVP-BEZ235. Molecular docking
   pattern of NVP-BEZ235 complexed with CDC42BPA (A), COX6A1 (B), PON2
   (C), TM9SF2 (D), UBBE2K (E), UBR2 (F), ZC3H14 (G), and ZNF106 (H)

Immune-infiltrating landscape of diabetic-related CC patients

   The CIBERSORT algorithm was applied to illustrate the infiltration
   abundance of different immune cells between the tumor and the normal
   samples in CC patients with diabetes. The results suggested that
   increased levels of activated CD4 T cells, CD56dim natural killer
   cells, macrophages, mast cells, natural killer T cells, neutrophils,
   regulatory T cells, and effector memory CD4 T cells were displayed in
   the tumor group than that of the normal group. However, the levels of
   eosinophils, natural killer cells, plasmacytoid dendritic cells, and
   Type 2 T helper cells were found to be decreased in the tumor group
   (Fig. [91]7A). We examined the correlations between 11 hub DEGs and 28
   kinds of immune cells (Fig. [92]7B), it showed that the levels of Type
   2 T helper cells, plasmacytoid dendritic cells, natural killer T cells,
   immature B cells, and eosinophils were positively correlated with all
   of the hub DEGs, while the levels of Type 1 T helper cells, regulatory
   T cells, neutrophils, mast cells, macrophages, gamma delta T cells,
   effector memory CD4 T cells, central memory CD8 T cells, central memory
   CD4 T cells, and CD56 bright natural killer cells were negatively
   correlated with all of the hub DEGs. In addition, the correlations
   between 28 types of immune cells are displayed in Fig. [93]7C.

Fig. 7.

   [94]Fig. 7
   [95]Open in a new tab

   Immune-infiltrating landscape of diabetes related CC patients. (A)
   Violin plot showed differences in infiltrated immune cells between the
   tumor group and normal groups. (B) Correlation of 11 hub DEGs and 28
   immune cell types. (C) Heatmap showed the correlation in immune cell
   infiltrations. *p < 0.05, **p < 0.01, ***p < 0.001

The correlations between TMB, MSI, and hub DEGs

   We downloaded clinical data and gene expression of 527 colorectal
   cancer patients from the TCGA database and analyzed the relationships
   between MSI sore (Fig. [96]8), TMB score (Fig. [97]9), and expression
   of hub DEGs. The gene expression of IP6K2, PON2, and TM9SF2 were
   negatively correlated with TMB score and MSI score, while the gene
   expression of ABI1, CDC42BPA, COX6A1, TMED2, UBE2K, UBR2, ZC3H14, and
   ZNF106 were positively correlated with TMB score and MSI score.

Fig. 8.

   [98]Fig. 8
   [99]Open in a new tab

   The correlation between hub DEGs and MSI. The correlation analysis
   analysis of the hub DEGs MSI in colorectal cancer patients. (A) ABI1.
   (B) CDC42BPA. (C) COX6A1. (D) IP6K2. (E) PON2. (F) TM9SF2. (G) TMED2.
   (H) UBE2K. (I) UBR2. (J) ZC3H14. (K) ZNF106

Fig. 9.

   [100]Fig. 9
   [101]Open in a new tab

   The correlation between hub DEGs and TMB. The correlation analysis
   analysis of the hub DEGs TMB in colorectal cancer patients. (A) ABI1.
   (B) CDC42BPA. (C) COX6A1. (D) IP6K2. (E) PON2. (F) TM9SF2. (G) TMED2.
   (H) UBE2K. (I) UBR2. (J) ZC3H14. (K) ZNF106

Discussion

   Diabetes is considered to be a risk factor for CC, which can lead to a
   decreased survival rate and shorter recurrence time [[102]9]. Thus,
   drugs that target both CC and diabetes were needed to offer better
   clinical treatment. In this study, DEGs between CC patients with T2DM
   and without T2DM were identified by comprehensive bio.

   informatics analysis, and it could help predict the diagnosis and
   prognosis of patients. Meanwhile, a potential small molecular compound
   was identified, it might be an effective therapy for diabetic-related
   CC.

   The underlying mechanisms that how diabetes affects CC are complicated
   and remain unclear. Previous studies suggested that the potential
   mechanism may related to the occurrences of hyperinsulinemia,
   hyperglycemia, mitochondrial dysfunction, DNA damage, and high levels
   of insulin-like growth factor-1 induced by diabetes [[103]7, [104]14].
   Recent studies provided new insights into this field. It was indicated
   that endothelial-mesenchymal transition induced by diabetes promoted
   the carcinogenesis and progression of CC [[105]15]. Besides, diabetes
   might lead to colorectal cancer progression by influencing intestinal
   microbiota [[106]16]. As we can see, many different mechanisms were
   involved in this process. Therefore, it is particularly important to
   find out the common mechanism of CC and diabetes, and then find the key
   target. We analyzed the differences between 3 paired groups: the tumor
   and the normal colon mucosa samples in CC patients, the diabetic and
   the non-diabetic colon mucosa samples in CC patients, and the tumor and
   the normal colon mucosa samples in CC patients with T2DM. The
   intersection of 3 groups of DEGs finally output 11 hub genes, which
   were closely related to each other, and might contribute to clarifying
   the relationship between CC and diabetes. The integration of diabetes-
   and tumor-associated DEGs with WGCNA modules was designed to identify
   genes that are co-regulated in both metabolic and oncogenic contexts
   [[107]7]. Diabetes is known to influence colon cancer progression
   through multiple overlapping mechanisms, including
   hyperinsulinemia-induced cell proliferation, mitochondrial dysfunction,
   chronic inflammation, and immune dysregulation [[108]7,
   [109]16–[110]20]. Several of the identified hub genes—such as CDC42BPA,
   COX6A1, and ZC3H14—are involved in these pathways, suggesting they
   could serve as molecular bridges between diabetes and colon cancer
   [[111]21–[112]23]. This integrative strategy thus enabled the
   identification of potential key mediators that link metabolic
   dysregulation to tumor biology in diabetic patients [[113]7, [114]16,
   [115]18].

   The incidence of CC is rising in the young population, recent
   guidelines recommend people should begin screening at early age
   [[116]24]. Most of the hub DEGs we selected were useful to identify
   tumors and diabetes, and provided better screening options. CC is a
   tumor with obvious heterogeneity and aggressiveness, sensitive and
   affordable prognostic methods can reduce its harm greatly [[117]25].
   The expression level of 10 hub genes had significant prognostic value
   for CC patients. The mixed prognostic patterns observed for the hub
   genes suggest that they may have distinct biological functions in
   diabetes-associated colon cancer. Some genes, such as CDC42BPA and
   ABI1, may act as oncogenic drivers by promoting cytoskeletal remodeling
   and metastasis, while others, like PON2 or COX6A1, may play protective
   roles through antioxidative or mitochondrial regulatory mechanisms.
   These findings underscore the complexity of metabolic-oncogenic
   interactions and suggest that therapeutic targeting should consider the
   functional context and expression patterns of individual genes. While
   the the observed correlations between immune cells and hub genes
   provide insight into potential immune remodeling in diabetic-associated
   colon cancer, further mechanistic studies are necessary to confirm
   these interactions and resolve context-dependent immune effects.

   At present, the main treatment therapies for CC are surgery,
   radiotherapy, and chemotherapy. However, these treatments are often
   accompanied by side effects, including decreased gastrointestinal
   function, poor immunity, and worse pain [[118]26]. Targeted drug
   therapy is a new direction that differs from the others. NVP-BEZ235
   (BEZ235, dactolisib) is a dual inhibitor that can suppress the
   PI3K/Akt/mTOR (PAM) pathway strongly [[119]27, [120]28]. The PAM axis
   is a highly conserved signal transduction network in eukaryotic cells,
   and the activation of this pathway frequently occurs in human cancer
   which is considered to implicate drug resistance [[121]29]. The mTORC1
   signaling was enriched by the DEGs of the tumor and the normal colon
   mucosa samples in CC patients. Previous studies suggested that
   suppression of the PI3K/Akt/mTOR (PAM) pathway had a notable anti-tumor
   effect in CC [[122]30–[123]32]. Researchers treated colorectal cancer
   cells with diosmin (a natural NF-κB inhibitor) and BEZ235, and found
   that they inhibited the PI3K/AKT/mTOR/NF-κB signaling collectively,
   leading to apoptosis and cell proliferation inhibition, and altered the
   angiogenesis process [[124]33]. More experiments should be conducted to
   confirm the therapeutic effect of NVP-BEZ235 in CC.

   Eight molecular targets including CDC42BPA, COX6A1, PON2, TM9SF2,
   UBBE2K, UBR2, ZC3H14, and ZNF106 were found to have binding
   interactions with NVP-BEZ235. CDC42BPA, cell division cycle 42 binding
   protein kinase alpha, is a kinase target for several cancers [[125]34].
   It was correlated with metastasis and poor survival of CC, and the
   expression of CDC42BPA was significantly higher in CC tissues
   [[126]21]. On the other hand, a recent study showed that glucose
   restriction could regulate the activity of CDC42BPA [[127]35]. COX6A1,
   cytochrome c oxidase subunit 6A1, was a protein related to
   mitochondrial metabolism and had anti-apoptosis functions in the
   autophagy-lysosome pathway [[128]22, [129]36]. Mitochondrial
   dysfunction and inhibition of apoptosis may contribute to
   diabetes-mediated colonic mucosal carcinogenesis [[130]7, [131]17].
   PON2 belonged to the paraoxonase (PON) gene family, overexpression was
   observed in cancer cells and involved in tumor survival and stress
   resistance [[132]37]. Besides, PON2 seemed to be closely related to
   diabetic retinopathy and diabetic glomerulopathy [[133]38, [134]39].
   The transmembrane 9 superfamily 2 (TM9SF2) was defined as a novel
   colorectal cancer oncogene, with low TM9SF2 expression correlated with
   a better RFS [[135]40]. Moreover, it has a potential regulatory role in
   cell cycle progression, which coincides with the mechanisms of
   diabetes-induced CC [[136]7, [137]40]. UBR2 was one of the N-recognins,
   it was found to be upregulated in cancer to induce muscle atrophy
   [[138]41]. On the other hand, decreased UBR2 was reported to affect
   lipid storage by regulating rate-limiting triglyceride hydrolase levels
   [[139]42]. Notably, recent studies suggested that UBR2 was a positive
   regulator of pro-inflammatory cytokine expression and could effectively
   regulate caspase-independent cell death [[140]20, [141]43], which might
   be beneficial to study the common mechanisms of diabetes and CC. Zinc
   finger proteins were reported to be transcription factors that could
   regulate tumorigenesis and tumor progression in various tumors
   [[142]44]. Zinc finger cys3his protein 14 (ZC3H14) was considered to be
   an important RNA binding protein that plays an important role in
   post-transcriptional regulation, including mRNA stability and transport
   [[143]45]. It also plays a role in regulating cellular energy, and when
   ZC3H14 is knocked out it reduces cellular ATP levels and causes
   mitochondrial fragmentation [[144]23]. Fragmentation of the
   mitochondrial network induces apoptosis and energy reduction, leading
   to reactive oxygen species excess and mitochondrial dysfunction,
   resulting in β cell dysfunction and insulin resistance [[145]19], and
   abnormalities in the mitochondrial quality control system are important
   mechanisms of diabetic cardiomyopathy as well [[146]46]. ZNF106 was a
   member of this family. In combination with other zinc finger proteins,
   it was closely related to immune cell infiltration in esophageal
   cancer, and can be used as a potential predictor of chemotherapy
   sensitivity [[147]44]. And the particular variant of ZNF106 was
   associated with hemoglobin A1c levels in humans [[148]47].

   There were some limitations in our study. First, the sample size of CC
   patients involved was small. Second, the results we acquired were based
   on bioinformatics analysis, lacking validation in vivo and in vitro.

Conclusion

   Our study discovered the different expression patterns between CC
   patients with diabetes and those without diabetes, which might improve
   the diagnosis and prognosis of CC greatly. We identified the potential
   small molecular compound NVP-BEZ235 targeting the DEGs of
   diabetic-related CC, and drew the immune-infiltrating landscape of
   diabetic-related CC patients, providing novel insights into the
   mechanism and treatment of diabetic-related CC.

Author contributions

   All the authors were involved in the study: study design and
   administration: ZC; Extraction of data: MY; Analysis and interpretation
   of the data: MY and RW; writing of the original draft: MY; writing and
   editing: FL and RC. All the authors participated in the discussions of
   the results and contributed to the manuscript.

Funding

   The National Key Clinical Specialties Construction Program supported
   this research.

Data availability

   Data is provided within the manuscript.

Declarations

Ethics approval and consent to participate

   not applicable.

Consent for publication

   All authors are aware of and have consented to publication.

Competing interests

   The authors declare no competing interests.

Footnotes

   Publisher’s Note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

References