Abstract

Background

   Colonic adenocarcinoma (COAD) is the most common pathological type of
   colon cancer. Tumor microenvironment (TME) plays an important role in
   the occurrence and development of COAD. There are currently no specific
   studies indicating the mechanism of action of TME in COPD patients.

Methods

   The percentage of tumor-infiltrating immune cells (TIC) in 512 COAD
   cases from The Cancer Genome Atlas (TCGA) database was calculated using
   CIBERSORT and ESTIMATE. Weighted gene coexpression network analysis
   (WGCNA) was performed to find modules of differentially expressed genes
   (DEGs) with high correlations followed by Gene Ontology (GO) and Kyoto
   Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses to
   determine the function of distant metastasis (M)-stage-related modules.
   Pathway enrichment analysis, protein–protein interaction (PPI) network,
   Cox regression analysis, and Kaplan–Meier survival analysis were
   performed on DEGs to select the most critical genes. The correlation
   between SIGLEC1 expression in COAD and TME status and between immune
   checkpoints and SIGLEC1 was examined using gene set enrichment analysis
   (GSEA) and Pearson correlation coefficients.

Results

   A WGCNA screen was performed to obtain 12,342 DEGs and 209 key genes
   associated with M stage between tumor and normal samples. GO and KEGG
   analysis revealed that the DEGs primarily engaged in pathways such as
   Th1 and Th2 cell differentiation and cell adhesion molecules. SIGELEC1
   gene was identified by univariate Cox regression, PPI network
   construction, and survival analysis. GSEA showed that the genes in the
   high-expression SIGLEC1 group were mainly enriched in immune-related
   activities. In the low-expression SIGLEC1 group, the genes were
   enriched in MYC targets. CIBERSORT analysis of the proportion of TICs
   showed that SIGLEC1 was positively correlated with macrophages (M0,
   M2), T-cell CD8 and immune checkpoint-related genes, suggesting that
   SIGLEC1 may be responsible for maintaining the immune dominance of TME.
   Immunohistochemical and prognostic analysis showed that the group with
   higher SIGLEC1 expression had more severe lesions and a worse prognosis
   than the group with lower SIGLEC1 expression.

Conclusions

   SIGLEC1 gene is a distant metastasis-related gene that affects the
   survival prognosis of COAD patients and provides additional insight
   into the treatment of COAD.

Supplementary Information

   The online version contains supplementary material available at
   10.1007/s12672-025-02093-2.

   Keywords: Colon cancer, Tumor microenvironment (TME), Distant
   metastasis, ESTIMATE, Algorithm, WGCNA analysis

Introduction

   Colon adenocarcinoma (COAD) is a malignant neoplasm occurring in
   adeno-squamous epithelial cells and is one of the most common
   pathological types of colon cancer due to its extremely high incidence
   in many regions and countries, with approximately millions of COAD
   patients worldwide [[32]1, [33]2], and the mortality rate increases
   with age [[34]3]. However, the current surgical and adjuvant treatments
   for COAD are very limited in improving the prognosis of patients and no
   longer meet the treatment needs of patients [[35]4, [36]5]. Therefore,
   it is urgent to explore effective immunotherapies and accurate
   biomarkers for the treatment of this disease.

   The tumor microenvironment (TME) is a highly complex ecosystem [[37]6]
   that includes fibroblasts, immune cells, adipocytes, vascular
   endothelial cells, and extracellular matrixes. During tumor
   development, the TME interacts with tumor cells and jointly mediates
   immune tolerance to tumors, which has an impact on the clinical outcome
   of immunotherapy [[38]7]. Furthermore, it has been shown that TME can
   largely enhance the immune effect of tumors [[39]8–[40]10], and a large
   number of immune cells can lead to tumor infiltration by secreting
   different factors, which are closely related to the prognosis of
   patients [[41]11]. A recent study elucidated the important role of
   CSF-1R in the TME and the immunotherapeutic target for COAD, suggesting
   that the immune status of the TME is greatly beneficial for the
   treatment and prognosis of COAD [[42]12]. Estimate, an approach to
   determining tumor purity by gene expression characteristics, can be
   used to infer the ratio of tumor samples’ mesenchymal and immune cells
   that play a significant role in tumor tissues. In addition, Weighted
   Gene Co-expression Network Analysis (WGCNA), a systems biology approach
   that can describe gene association patterns and identify biomarker
   genes or therapeutic targets based on highly synergistic sets of
   variation, has been widely used in immunotherapy studies [[43]13,
   [44]14]. Although the impact of TME on the efficacy of immunotherapy
   has now become a hot topic of research, the specific mechanism of
   action regarding precise genetic analysis remains to be investigated.
   Therefore, the present study attempted to analyze the tumor
   microenvironment of patients with colon adenocarcinoma in the TCGA
   database by these two methods and to find differential genes possessing
   prognostic value.

Materials and methods

Data sources

   A total of 512 samples were obtained from the TCGA‐COAD database
   combined with transcriptome data and clinical information, including
   471 tumor samples and 41 normal samples from a public database UCSC
   Cancer Genomics Browser ([45]https://xenabrowser.net/datapages/).

Generation of ESTIMATE algorithms and survival analysis

   Based on the ESTIMATE algorithm, the scores associated with the
   immune-stromal component of TME were estimated for each sample [[46]10]
   and presented as three subsets of scores: ImmuneScore, StromalScore,
   and ESTIMATEScore. According to the median values of these subsets, the
   COAD cases were divided into the high group and the low group,
   respectively. Additionally, the R package survival and survminer were
   applied for survival analysis.

Analysis of differentially expressed genes and WGCNA analysis

   Differentially expressed genes (DEGs) with p-value < 0.05 were
   generated using the R “limma” package and visualized using the R
   “pheatmap” package [[47]15, [48]16]. Based on the DEGs, co-expression
   networks were constructed using the WGCNA. Then, DEGs and clinical data
   were analyzed according to the appropriate soft-threshold power.
   Moreover, to further analyze the modules, the merged cutoff threshold
   was set to 0.2, the dissimilarity result of the modules was calculated,
   and some of them were merged. Finally, the significant module was
   chosen for PPI network using the STRING database [[49]17], followed by
   reconstruction via Cytoscape. To better identify the hub genes, the
   loaded R package survival was applied for univariate COX regression
   find hub gene.

Functional analysis

   To obtain the functions of the genes, enrichment analyses were
   performed on the Kyoto Encyclopedia of Genes and Genomes (KEGG) and
   Gene Ontology (GO) databases by R “ggplot2,” “org.Hs.eg.db,”
   “enrichplot,” and “clusterProfiler” packages [[50]18].

GSEA analysis

   Hallmark and C7 gene collections were downloaded from the Molecular
   Signatures Database ([51]http://www.gsea-msigdb.org/gsea/index.jsp).
   Using SIGLEC1 expression level as the phenotype annotation, data from
   COAD patients in the TCGA cohort were divided into the low and high
   expression for analysis, and only gene sets with NOM p < 0.05 and FDR
   q < 0.25 were considered significant.

Tumor-infiltrating immune cells (TICs) profile

   CIBERSORT algorithm was carried out to estimate the contents of
   immune-infiltrating cells in all tumor samples, followed by quality
   filtering to select samples with p < 0.05 for subsequent analysis.
   GEPIA2 ([52]http://gepia2.cancer-pku.cn/) [[53]19], an updated version
   of gene expression profiling interactive analysis, was performed to
   analyze the association of SIGLEC1 with immune checkpoints.

Immunohistochemistry (IHC) staining

   Tumor and paracancer tissue microarrays from the Affiliated Tumor
   Hospital of Nantong University were used for the validation cohort.
   Immunohistochemical staining was conducted using primary anti-SIGLEC1
   antibody (1:200; 55427-1-AP, proteintech, China). The density of
   positive staining was measured, and the H score was evaluated by two
   independent pathologists blinded to the clinicopathological
   information. A microscopy system (Nikon, Japan) was used to scan
   immunohistochemistry sections.

Statistical analysis

   R 4.2.3 was used for the statistical analysis in this study. Statistics
   were deemed significant at p < 0.05.

Results

Immune scores revealed more prognostic value of COAD patients

   The overall workflow of this study was shown in Fig. [54]1. To
   establish the correlation between the ESTIMATE algorithms and the
   survival rate, we calculate the immune score, stromal score, and
   ESTIMATE score based on the ESTIMATE algorithm and performed a
   Kaplan–Meier survival analysis separately. The results showed that
   there was no significant correlation between stromal score and ESTIMATE
   score and overall survival (Fig. [55]2A and C). However, there was a
   positive correlation between immune scores and overall survival, where
   colonized patients with low immune scores were associated with poor
   survival outcomes (Fig. [56]2B). These results implied that the immune
   components were more suitable for indicating the prognosis of COAD
   patients.

Fig. 1.

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

   The overall workflow of this study

Fig. 2.

   [59]Fig. 2
   [60]Open in a new tab

   ESTIMATE algorithms are associated with the survival of the COAD. A
   Kaplan–Meier curves of the high and low stromal scores of COAD
   patients. B Kaplan–Meier curves of the high and low immune scores of
   COAD patients. C Kaplan–Meier curves of the high and low estimate
   scores of COAD patients

   Additionally, we investigated the relationship between immune scores
   and stromal scores and clinical features, including pathological stage
   and AJCC‐TNM stage. As shown in the immune scores, TMN stages and
   pathological stages were negatively correlated with M—distant
   metastasis classification. However, no significant difference was
   observed in the stromal scores and ESTIMATE scores with any other
   clinical features according to the results from the above survival
   analysis. These results suggested that the immune components were
   linked with the progress of COAD, such as metastasis and pathological
   stage (Fig. [61]3).

Fig. 3.

   [62]Fig. 3
   [63]Open in a new tab

   Correlation of ImmuneScore, StromalScore and EstimateScore with
   clinicopathological features. A–D Distribution of StromalScore in
   pathological stage and TMN stages. E–H Distribution of ImmuneScore in
   pathological stage and TMN stages. I–L Distribution of EstimateScore in
   pathological stage and TMN stages

   To confirm the exact gene alteration in the immune components of COAD
   patients regarding, we classified the 471 COAD patients into high‐level
   and low‐level groups based on the median immune scores. As shown in the
   heatmap, we identified 12,342 differentially expressed genes (DEGs)
   according to immune scores. (Fig. [64]4A)

Fig. 4.

   [65]Fig. 4
   [66]Open in a new tab

   Screening of DEGs related by heatmap and WGCNA analysis. A Heatmap of
   DEGs by comparison of the high score and the low score groups in
   ImmuneScore. B Analysis of the scale-free index for best soft threshold
   powers (β). When β = 5, it satisfies the scale-free topology threshold
   of 0.89. C Analysis of the mean connectivity for best soft-threshold
   powers. D The DEGs clustered dendrogram and nodule color of WGCNA. E
   Correlation analysis of the modules and clinical features P‐values are
   shown. Scatter plot analysis in the F brown modules, G green modules, H
   greenyellow modules and I midnightblue

   The DEGs were comprehensively analyzed and classified into modules by
   WGCNA. In addition, the modules were further analyzed by clustering
   analysis to remove abnormal data. A soft threshold (β = 5, scale‐free R
   2 = 0.89) was used to construct a scale‐free network that determined
   seventeen modules (Fig. [67]4B and C). Next, the whole gene expression
   levels of related modules were analyzed to find correlations between
   the corresponding modules and clinical data including the M
   classification of TMN stages (Fig. [68]4D–E). Then, the related modules
   (darked, green, drak-green, green-yellow, and midnight-blue) exhibited
   a significant correlation with the M stage as the M stage-related
   module (Fig. [69]4F–I). Therefore, these modules were chosen for
   further analyses, and 209 critical genes associated with M stage were
   screened using the threshold correlation.

Functional analysis of genes

   GO analysis was performed to analyze the function of the M
   stage-related module (Fig. [70]5A and Additional file [71]1: Table S1).
   We found that these genes were functionally almost associated with
   responding to immune-related GO terms, such as receptor activity,
   cytokine receptor activity, MHC protein complex binding, and T cell
   receptor binding.

Fig. 5.

   [72]Fig. 5
   [73]Open in a new tab

   A Gene ontology enrichment analysis and B Genes and Genomes enrichment
   analysis

   Genes and Genomes (KEGG) enrichment analysis also displayed the
   enrichment of cell adhesion molecules, Th1 and Th2 cell
   differentiation, Th17 cell differentiation, and hematopoietic cell
   lineage (Fig. [74]5B and Additional file [75]2: Table S2). Therefore,
   the overall function of DEG seems to be linked to immune-related
   activities, suggesting that the involvement of immune factors was a
   major feature of TME in COAD.

Intersection analysis of PPI network and univariate COX regression

   To further explore the underlying association of the module genes, a
   PPI network based on the STRING database was constructed using
   Cytoscape software, as shown in Fig. [76]6A. The prognosis of COAD
   patients was analyzed by univariate Cox regression and only two
   factors, CYHR1 and SIGLEC1 were determined as the hub genes
   (Fig. [77]6B). Then, the survival analysis further screened out only
   SIGLEC1 (Fig. [78]6C and D).

Fig. 6.

   [79]Fig. 6
   [80]Open in a new tab

   Protein–protein interaction network and univariate COX. A Gene networks
   from the module gene based on the WGCNA method. B Univariate Cox
   regression analysis, listing to prognosis. C, D Overall survival
   analysis of the CYHR1 and SIGLEC1 genes in colon cancer based on the
   Kaplan–Meier plotter

SIGLEC1 had potential to be an indicator of TME modulation

   To identify SIGLEC1-associated signaling pathways, we implemented the
   Gene Set Enrichment Analysis (GSEA) between high and low SIGLEC1
   expression data sets. Genes in SIGLEC1 high-expression group were
   mainly enriched in hallmark gene sets, such as inflammatory response,
   allograft rejection, epithelial mesenchymal transition, and IL6-STAT3
   signaling (Fig. [81]7A). However, the low-expression SIGLEC1 gene set
   was enriched in the MYC targets (Fig. [82]7B). Regarding the C7
   collection defined by the MSigDB, the immunologic gene sets and genes
   enrichment of the high SIGLEC1 expression group were shown in
   Fig. [83]7C. As to SIGLEC1 low-expression group, the enrichment of
   genes was shown in Fig. [84]7D. These results suggested that SIGLEC1
   might be a potential indicator for the TME in COAD. The ROC curve also
   confirmed the diagnostic value of SIGLEC1 for M1 stage (AUC = 0.904)
   and COAD (AUC = 0.891) (Figure S1). Therefore, the GSEA findings
   implied that SIGLEC1 expression was correlated with immune-related
   signals in COAD.

Fig. 7.

   [85]Fig. 7
   [86]Open in a new tab

   GSEA delineates high SIGLEC1 expression and low expression. A HALLMARK
   collection enriched in the high SIGLEC1 expression sample. B HALLMARK
   collection enriched in the low SIGLEC1 expression sample. C Enriched
   immunologic gene sets in C7 collection of the high SIGLEC1 expression
   group. D Enriched immunologic gene sets in C7 collection of the low
   SIGLEC1 expression group

Correlation of SIGLEC1 with the TICs

   To further confirm the correlation of SIGLEC1 expression with the
   immune microenvironment, we detected the part of the entire 22 subtypes
   of immunocytes in each sample, as shown in Fig. [87]8A, and as revealed
   by the hierarchical clustering, immune genes presented distinct
   distribution differences in COAD tumor samples and adjacent samples. It
   was easily found that Neutrophils and activated mast cells exerted a
   remarkable positive association; however, Fig. [88]8B showed an obvious
   negative correlation between resting mast cells and activated ones.

Fig. 8.

   [89]Fig. 8
   [90]Open in a new tab

   The profile of tumor-infiltrated immune cells in tumor samples and
   correlation analysis. A Barplot showing the fractions of 21 types of
   tumor-infiltrating immune cells in COAD samples. B Heatmap displaying
   the correlation between the 22 types of tumor-infiltrating immune cells

   Notably, a comparison of the TIICs levels between the high and low
   expression of SIGLEC1 confirmed an elevated level of macrophage M0 and
   decreased B cells naive in the high SIGLEC1 expression (Fig. [91]9A).
   Through correlation analysis, SIGLEC1 was positively related to
   macrophages(M0, M2), resting mast cell, monocytes, neutrophils,
   activated NK cells, and T cells CD8, while negatively associated with
   activated mast cells, resting T cells CD4 memory, and activated T cells
   CD4 memory (Fig. [92]9B). These results further supported that the
   SIGLEC1 levels affected the immune activity of TME. Consequently, the
   study further determined whether SIGLEC1 was correlated with immune
   checkpoints such as PD-L1, which was also crucial for predicting the
   efficacy of immunotherapy in COAD. Immune-checkpoint-related genes such
   as CTLA-4, LGALS9 (GAL9), LAG-3, PDCD1 (PD-1), PDCD1LG2 (PD-L2), CD274
   (PD-L1), TIGIT, and HAVCR2 (TIM3) were selected for further analysis
   [[93]20]. Interestingly, all the genes (CTLA-4, LAG-3, PD1, PDL2, PDL1,
   TIGIT and TIM-3) were positively associated with the SIGLEC1 expression
   in COAD (Fig. [94]9C).

Fig. 9.

   [95]Fig. 9
   [96]Open in a new tab

   Significant correlation between the expression of SIGLEC1 and the
   proportion of tumor-infiltrating immune cells. A Scatter plot showing
   the correlation between 10 types of tumor-infiltrating immune cell
   ratio with the SIGLEC1 expression (all p < 0.05). B Mendelian
   randomization analysis. C Correlation between the immune checkpoint
   genes between SIGLEC1 expression

Clinical COAD samples validate SIGLEC1 signature

   A summary of the clinicopathological characteristics of SIGLEC1
   expression levels based on tumor tissue microarrays from Nantong
   University Cancer Hospital showed that N stage (p < 0.001), M stage
   (p = 0.042), and Pathologic stage (p = 0.018) were significantly
   associated with SIGLEC1 expression levels in COAD (Table [97]1). We
   conducted immunohistochemical (IHC) staining on tissue sections
   obtained from surgical specimens to perform prognostic analysis of the
   patients. The IHC staining of SIGLEC1 exhibited heterogeneity but was
   able to distinguish between high and low expression samples
   (Fig. [98]10A). Subsequently, we evaluated the patient prognosis based
   on varying levels of SIGLEC1 expression and observed that patients with
   high SIGLEC1 expression had a significantly lower survival rate than
   those with low expression (p = 0.003) (Fig. [99]10B).

Table 1.

   Patients’ characteristics and clinicopathological parameters of the
   study group
   Characteristic Low expression of SIGLEC1 High expression of SIGLEC1 p
   value
   n 96 119
   T stage, n (%) 0.983
    T1 11 (5.1%) 13 (6%)
    T2 10 (4.7%) 14 (6.5%)
    T3 43 (20%) 51 (23.7%)
    T4 32 (14.9%) 41 (19.1%)
   N stage, n (%) 0.001
    N0 66 (30.7%) 55 (25.6%)
    N1 13 (6%) 46 (21.4%)
    N2 16 (7.4%) 18 (8.4%)
    N3 1 (0.5%) 0 (0%)
   M stage, n (%) 0.042
    M0 87 (40.5%) 96 (44.7%)
    M1 9 (4.2%) 23 (10.7%)
   Pathologic stage, n (%) 0.018
    I 20 (9.3%) 22 (10.2%)
    II 44 (20.5%) 34 (15.8%)
    III 26 (12.1%) 56 (26%)
    IV 6 (2.8%) 7 (3.3%)
   Age, median (IQR) 64 (55, 73) 66 (56, 73) 0.537
   [100]Open in a new tab

Fig. 10.

   [101]Fig. 10
   [102]Open in a new tab

   Clinical COAD samples validate SIGLEC1 signature. A The
   immunohistochemical (IHC) staining on clinical sample tissue
   microarrays indicates high and low expression of SIGLEC1. B
   Kaplan–Meier analysis of SIGLEC1 expression levels

Discussion

   COAD is one of the common malignancies occurring in adeno-squamous
   epithelial cells. In the present experiment, the association with
   immune checkpoint-related genes in COAD samples was validated, starting
   with the comparison of DEGs generated by immune components, suggesting
   that SIGLEC1 expression may be an effective predictor of prognosis in
   COAD patients and a clue to altering TME status.

   By WGCNA analysis, we identified genes associated with the M stage,
   which a large number of studies have shown to be linked to cancer
   development [[103]21–[104]23]. In addition, univariate Cox regression
   and Kaplan–Meier analysis indicated that SIGLEC1 is a key gene
   associated with the M phase. SIGLEC1, also known as CD169, can encode a
   protein that is expressed as a type I transmembrane protein on
   macrophages. Cassetta et al. found that SIGLEC1 plays an important role
   in cancer regulatory pathways [[105]24] and some investigators have
   also determined SIGLEC1 as a potential prognostic biomarker and a basis
   for immunotherapy through the ESTIMATE algorithm [[106]25].

   Numerous studies confirm that TME remodeling is a potential regulator
   of cancer progression and a source of therapeutic targets
   [[107]26–[108]29]. However, whether dynamic regulation of TME
   contributes to the prognostic indicators of COAD remains unknown, while
   studies related to immunotherapy targeting TME have become a hot topic
   in cancer treatment [[109]30]. Additionally, a previous study found
   that macrophages, one of the major immune cell types in TMB, are
   associated with poor prognosis among patients with breast cancer and
   head and neck squamous cell carcinoma (SCC) [[110]31, [111]32].
   Nevertheless, the prognostic value of SIGLEC1 in TME in patients with
   COAD has not been verified.

   To validate the prognostic value of SIGLEC1 in TME of COAD patients, we
   explored the potential mechanisms of SIGLEC1 by GSEA analysis. Earlier
   studies found immune-related pathways with high SIGLEC1 expression such
   as inflammatory response, allograft rejection, epithelial mesenchymal
   transition, and IL6-STAT3 signaling, and confirmed the role of SIGLEC1
   in inflammatory responses [[112]33–[113]35]. And Jing et al. observed
   that SIGLEC1 on monocytes was associated with tumor progression and
   inhibited the STAT3 signaling pathway [[114]31].

   To further confirm the correlation between SIGLEC1 expression and
   immune microenvironment, we analyzed the levels of TIICs with high and
   low SIGLEC1 expression and found that high and low expression coincided
   with an increase in macrophage M0 levels and a decrease in B cell
   naive, while it has been shown that SIGLEC1 induces homologous
   activation of B cells and stimulates the expression of the molecule
   [[115]36], which is highly expressed in M2 macrophages [[116]37]. It
   has been shown that SIGLEC1 expression levels in monocytes correlate
   with disease severity [[117]38, [118]39] and that SIGLEC can regulate T
   cells, macrophages, and neutrophils in the immune response [[119]40,
   [120]41], while our results found that SIGLEC1 was positively
   correlated with macrophages (M0, M2), monocytes, neutrophils, and T
   cells and negatively correlated with resting T cell CD4 memory and
   activated T cell CD4 memory. Mendelian randomization provides an
   alternative method to randomized clinical trials of investigating
   causal associations between an exposure and an outcome [[121]42,
   [122]43]. Through Mendelian randomization analysis, SIGLEC1 was
   positively related to macrophages (M0, M2), resting mast cell,
   monocytes, neutrophils.

   Immune checkpoints are closely linked to tumor proliferation, invasion,
   metastasis, and patient prognosis assessment, and play a role in
   stabilizing the immune function and are good targets for tumor therapy.
   At the same time, SIGLEC1 and SIGLEC2 are potential biomarkers that are
   considered sentinels of disease activity or drug response in autoimmune
   diseases [[123]44]. In the present study, we found a significant
   association between SIGLEC1 expression and immune checkpoints, such as
   CTLA-4, LAG-3, PD1, PDL2, PDL1, TIGIT, and TIM-3, while SIGLEC has been
   shown to serve as an immune checkpoint target for enhancing anti-tumor
   immune responses [[124]45]. Furthermore, numerous studies indicated
   that SIGLEC1 is associated with tumor metastasis [[125]24, [126]46,
   [127]47], which is consistent with our analysis of metastasis from
   distant sites to TME.

   In this study, we identified a predictive biomarker, SIGLEC1, and used
   TCGA’s COAD transcriptome sequencing model to calculate immune,
   stromal, and estimated scores in COAD patients and to investigate the
   correlation of these scores with the clinical characteristics and
   overall survival of the patients. Subsequently, the role of the SIGLEC1
   gene in TME and its association with immune checkpoint genes were
   explored, and we found that SIGLEC1 may be involved in the progression
   of COAD and could be used to estimate prognosis and be a new
   therapeutic target. However, the specific role of SIGLEC1 in regulating
   the development and progression of COAD requires more exploration and
   research, and further experiments using in vivo and in vitro models to
   confirm TME in colon cancer.

Limitation

   This study was mainly completed through bioinformatics, and the size of
   the sample size will directly affect our results. In addition, our
   study still needs more experiments to verify.

Supplementary Information

   [128]Supplementary Material 1.^ (11.9KB, docx)
   [129]Supplementary Material 2.^ (14.1KB, docx)
   [130]Supplementary Material 3.^ (588.3KB, tif)

Acknowledgements