Downregulation of Enteroendocrine Genes Predicts Survival in Colon Cancer: A Bioinformatics-Based Analysis

Abstract

Colorectal cancer (CRC) is the fourth most common and the third mostly deadly cancer globally. Even with alternative therapies, some patients do not respond to treatment. Identifying modulations in the tumor microenvironment (TME) of CRC is a significant challenge due to the complex and dynamic nature of the TME. The intestinal epithelium comprises different types of secretory lineage cells, including goblet, tuft, Paneth, and enteroendocrine cells (EECs). Yet the relevance of each subtype of secretory intestinal epithelial cell (IEC) within the TME is still debated. This study investigated the involvement of IECs in CRC development through an integrative bioinformatics analysis. We used publicly available datasets from the National Center for Biotechnology Information, the Cancer Genome Atlas Program, and the National Cancer Institute’s Proteomics Tumor Analysis Consortium, encompassing both human and mouse CRC samples. Our findings reveal a CRC microenvironment characterized by elevated expression levels of genes associated with WNT pathway activity. Remarkably, there was increased expression of Paneth cell-associated markers and transcription factors, such as WISP1, LYZ, SOX9, and DEFA1. Conversely, EEC-specific gene markers, such as GCG (encoding glucagon-like peptide-1) and CHGA exhibited significant downregulation in CRC tissue compared with healthy tissue, partially due to Paneth cell activity. Gene ontology analysis showed species-conserved downregulation in hormone/peptide secretion-related pathways in both mouse and human CRC. Of note, lower levels of GCG and CHGA correlated with reduced overall survival and demonstrated a correlation with the cell cycle, apoptosis, and proliferation. These results suggest that the disruption of enteroendocrine cell signaling is a hallmark of CRC development and may hold prognostic and therapeutic value in treating CRC patients.

1. Introduction

According to data from GloboCAN, 20 million new cancer cases occurred worldwide in 2022, resulting in a mortality rate of 9.7 million people. Colorectal cancer (CRC) is the third most diagnosed malignancy globally, surpassed only by sex-specific cancers (prostate cancer in men and breast cancer in women). This makes it the leading non-sex-specific cancer in terms of global incidence [1]. Several factors—including environmental, genetic, and inflammatory factors—are associated with CRC progression [2]. Furthermore, diet and obesity are linked to the development and progression of type II diabetes [3], which is, in turn, associated with the development of certain cancers, like CRC [4] and colitis-associated colorectal cancer (CAC). This suggests that the growth rate of the disease is directly linked to dietary habits [1,5,6]. Despite existing therapeutic alternatives (surgery, chemotherapy, radiation therapy, and immunotherapy), some patients do not respond to treatment. Additionally, modern lifestyle habits are recognized as risk factors for the development and progression of the disease. Importantly, the increasingly frequent appearance of early-onset CRC in young people has led to its ranking as the second most deadly cancer worldwide, significantly impacting the overall mortality rate [7]. This highlights the urgent need for further studies to unravel new molecules and pathways associated with CRC onset or progression that could be targeted for the development of new drugs and therapies.

CRC development is a complex process involving various molecules, pathways, cells, and environmental factors. It is directly linked to mutations in intestinal epithelial cells (IECs) and the breakdown of the epithelial barrier. The intestinal epithelium is composed of cell subtypes originating from intestinal stem cells (leucine-rich repeat-containing G-protein coupled receptor 5, LGR5+). Upon the induction of signaling pathways, including Wnt, β-catenin, and Notch, these stem cells coordinately activate specific transcription factors, promoting their differentiation into various IEC subtypes. IECs differentiate into either absorptive cells, known as enterocytes (differentiated by the transcription factor of the hes family bHLH transcription factor 1, HES1), or secretory cells, which include enteroendocrine cells (EECs), Tuft cells, goblet cells, and Paneth cells [8,9,10]. These secretory subsets are differentiated by the transcription factors Neurogenin 3 (NEUROG3), POU class 2 homeobox 3 (POU2F3), Kruppel-like factor 4 (KLF4), and SRY-box transcription factor 9 (SOX9), respectively.

In addition to their specific transcription factors, these secretory IECs are also characterized by the molecules or compounds that they produce and secrete. Tuft cells function as chemoreceptors, facilitating communication between the intestinal lumen, immune cells, and the neural network. They produce several molecules, such as neurotransmitters (acetylcholine), eicosanoids, and cytokines, (e.g., interleukin-25) [11]. Goblet cells primarily protect the epithelium by producing and secreting mucus, which acts as a protective barrier against pathogens [12]. Paneth cells are typically found in the small intestine and are uncommon in the large intestine under homeostasis. However, their presence in high numbers in the large intestine is an abnormal characteristic, often associated with metaplasia in patients with inflammatory bowel disease [13]. Paneth cells are responsible for producing essential antimicrobial proteins and peptides, such as alpha-defensin, playing a fundamental role in fighting infection and regulating the gut microbiome [14]. The last group of secretory lineage IECs is EECs. EECs represent about 1% of the epithelium, distributed throughout the intestinal epithelium, and are characterized by hormone production, found mainly in the small intestine and, to a lesser extent, in the large intestine. Hormone production is stimulated by diverse compounds, including macronutrients (glucose, amino acids, and fatty acids) and structural components of microorganisms, such as lipopolysaccharides [15]. Among the hormones produced by EECs, the group of incretins stands out—gastric inhibitory polypeptide (GIP) and glucagon-like peptide-1 (GLP-1). They are responsible for systemic functions, such as stimulating insulin secretion, controlling appetite in the central nervous system, and modulating inflammation and the immune system [16]. EECs and their peptide hormones regulate key functions of the gut epithelium, including digestion control, appetite regulation, glucose homeostasis, and immune function. This makes them potential targets for study in intestinal pathologies like CRC, given the role of hormones in mediating intestinal homeostasis [17,18].

It is known that IECs alter their transcriptional profiles and functions in response to the microenvironment, which can be modulated by various perturbations, such as those occurring in neoplastic conditions. However, it remains unknown whether secretory IECs, or their secreted products, are associated with the initiation or progression of CRC. We hypothesize that transcriptional and proteomic alterations affecting EECs and their hormones may destabilize intestinal epithelial homeostasis. To investigate this hypothesis, we performed an integrative analysis combining large-scale transcriptomics and proteomics data from human and experimental CRC samples. Using public gene expression databases, we assessed the abundance and enrichment of pathways related to markers of secretory IECs, with a specific focus on EECs and their hormones. Additionally, proteomics data were analyzed to investigate the relationship between EEC activity and the involved biological pathways. An EEC score was calculated, and its correlation with the proteomic landscape was explored through enrichment analysis, revealing pathways associated with cell proliferation, apoptosis, immune regulation, and inflammation.

Finally, to explore potential regulatory interactions, a mediation analysis was performed to investigate whether WNT pathway activation modulates EEC markers through Paneth cell activity. We aimed not only to characterize the abundance of specific markers of these populations compared to healthy tissue but also to identify potential prognostic markers associated with survival and response to chemotherapy in CRC patients. Taken together, these complementary approaches have allowed us to identify transcriptional and proteomic patterns that connect EEC function to tumor biology and clinical outcomes in CRC.

2. Results

2.1. Transcriptional Profile Analysis Reveals the Conserved Downregulation of EEC Markers in CRC

To gain a clearer understanding of the role that enteroendocrine hormones and secretory IECs play in the complex setting of CRC, we began by examining gene expression to identify potential molecular targets. We explored the presence or absence of secretory IEC markers by analyzing three different mouse CRC microarray datasets from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (GSE86299, GSE31105, and GSE64658). CRC in these models was induced using a well-established inflammation-associated CRC model: azoxymethane (AOM) followed by the administration of dextran sulfate sodium (DSS) [19].

To assess data consistency and explore transcriptional variability, we conducted a principal component analysis (PCA). In mice, PCA clearly separated adenoma samples (red) from control samples (blue) (Figure 1A and Figure S1A,B). We then asked if human CRC data would show the same pattern. Applying the same method to three different human CRC datasets (GSE8671, GSE18105, and GSE110225) again showed a clear split between adenoma and nearby non-tumorous tissue (Figure 1B and Figure S1C,D).

Next, we carried out differential gene expression analysis using a false discovery rate ($\mathrm{FDR}$) threshold of $<0.05$. Genes with a ${\mathrm{l}\mathrm{o}\mathrm{g}}_2$ fold change (${\log }_{2}FC$) $\le -1$ were classified as downregulated (blue), and those with ${\log }_{2}FC\ge 1$ were classified as upregulated (red). In the mouse datasets GSE8671, GSE18105, and GSE110225, we found 257, 537, and 216 genes downregulated and 639, 1151, and 189 genes upregulated, respectively (Figure 1C and Figure S1E,F). As expected, genes related to inflammatory processes (such as chemokines and receptors, cell migration regulators, and growth factors) were upregulated. Strikingly, several key EEC genes—including chromogranin A (Chga), glucagon (Gcg), neurogenic differentiation 1 (Neurod1), peptide YY (Pyy), and glucagon-like peptide-1 receptor (Glp1r)—were consistently reduced in the AOM/DSS-induced CRC model.

The human CRC biopsy datasets revealed a similar pattern, with 634, 415, and 286 genes downregulated and 637, 343, and 240 genes upregulated, respectively. As in mice, genes involved in EEC differentiation and hormone production were consistently decreased in tumor tissue compared with normal tissue (Figure 1D and Figure S1G,H). Taken together, these results show distinct transcriptional profiles between adenoma and normal tissue in both mouse and human CRC, with a consistent decrease in the expression of EEC-associated genes in tumors.

2.2. Antimicrobial Peptide-Related Pathways Are Upregulated While Hormonal Processes Are Downregulated in Colorectal Cancer

After identifying differences in gene expression, we examined which biological pathways were most affected in tumor samples. To analyze unbiased pathways, we first used a Venn diagram to identify genes consistently upregulated or downregulated across all three mouse datasets and all three human datasets. In mice, 39 downregulated genes (Figure 1E, left) and 87 upregulated genes (Figure 1E, right) were common to all studies. In humans, 119 downregulated genes (Figure 1F, left) and 58 upregulated genes (Figure 1F, right) were shared among the datasets.

We then performed Gene Ontology (GO) biological process enrichment analysis on these shared genes. In mice, 295 pathways were enriched in the upregulated gene set, while 66 pathways were associated with the downregulated set (Tables S1 and S2). In humans, we found 157 and 167 pathways associated with up- and downregulated genes, respectively (Tables S3 and S4). As expected, based on existing data [20], for the upregulated genes in the tumor context, the increased pathways were mainly related to the inflammatory response and positive regulation of the cell cycle (Figure 2A,B).

One notable and species-conserved finding was the enrichment of pathways linked to antimicrobial peptide-mediated immune responses among the upregulated genes in tumor samples (Figure 2A,B). Paneth cells, known producers of antimicrobial peptides, interact with the gut microbiota and can influence immune cell activity [21]. In contrast, pathways associated with hormonal processes—many of which depend on products secreted by EECs—were consistently enriched in the downregulated gene set (blue) in tumor tissue. These analyses suggest that tumor samples exhibit conserved stronger activity in antimicrobial peptide-related pathways, potentially driven by Paneth cells, and weaker activity in hormone-related functions of EECs.

2.3. Upregulation of Paneth Cell Markers and Downregulation of EEC Genes Is Conserved Across Colon Tumor Origin

Given the observed pathway alterations, we explored whether markers and transcription factors associated with other secretory IEC subtypes were also dysregulated. To address this, we examined marker genes for secretory IEC subtypes, including critical differentiation transcription factors and their secreted products (

Table 1. Genes related to each intestinal epithelial cell subtype and their descriptions.

Cell Type	Gene	Gene Function/Description	Reference
Enteroendocrine cell	GCG	GLP-1/2 hormone L-cell marker	[24]
	PYY	Hormone secreted by L cells
	CHGA	Classical marker	[15]
	INSM1	Neuroendocrine transcription factor	[25]
	NEUROD1	Neuroendocrine differentiation	[26]
	DPP4	Incretin degradation (GLP-1, GIP)	[27]
	ARX	Regulates the fate of endocrine subtypes	[28]
	NKX2.2	Required for intestinal endocrine differentiation	[24]
	GLP2R	GLP-2 receptor	[29]
	PAX6	Regulates EEC subtypes	[24]
	GIPR	GIP receptor
	NEUROG3	EEC master regulator	[30]
	PAX4	Intestinal endocrine development	[24]
	LMX1A	Regulates serotonin	[31]
	GLP1R	GLP-1 receptor	[24]
	ISL1	Endocrine regulation
	GIP	Hormone produced by K cells
Goblet cell	KLF4	Goblet cell differentiation	[32]
	MUC2	Major secreted mucin	[33]
	SPDEF	Essential for goblet cells	[34]
Tuft cell	POU2F3	Tuft cell master regulator	[35]
	SOX4	Tuft development	[36]
Paneth cell	GFI1	Regulation of Paneth cells	[37]
	DEFA1	Antimicrobial peptide produced by Paneth cells	[38]
	LYZ	Lysozyme classic marker	[39]
	SOX9	Essential transcription factor	[10]
	WISP1	Maintains niche and Paneth differentiation	[40]
Stem cell	LGR5	Intestinal stem cell marker	[41]

). For a marker to be considered significant, it had to be differentially expressed in the same direction (up- or downregulated) across all datasets. We generated a heatmap/bubble plot using the markers listed in Table 1, where the bubble size represents the adjusted p-value ($-{\log }_{10}\left(\mathrm{adj}.\mathrm{p}\right)$) and the color intensity indicates the ${\log }_{2}FC$ (red for $\ge 2$ and blue for $\le -2$). Then, we selected tuft cell markers, namely SRY-box transcription factor 4 (SOX4) and POU2F3, which did not exhibit a consistent pattern of regulation (Figure 3). In contrast, goblet cell markers, such as KLF4, mucin 2 (MUC2), and SAM pointed domain-containing Ets transcription factor (SPDEF), were consistently downregulated in tumor tissue, aligning with the literature describing a reduced goblet cell population in the neoplastic intestinal epithelium [22,23].

Consistent with our enrichment analysis, transcription factors essential for EEC differentiation—such as NEUROD1 and Aristaless-related homeobox (ARX)—were downregulated in both the murine and human tumor microenvironments (TMEs). Canonical EEC markers and hormone genes (e.g., GCG and PYY) were likewise reduced in tumor tissue (Figure 3). In contrast, Paneth cell markers, including defensin alpha 1 (DEFA1), lysozyme (LYZ), and SOX9, were upregulated in tumor tissue. Additionally, WNT1-inducible signaling pathway protein 1 (WISP1)—a transitional marker of Paneth cell differentiation [10]—also showed increased expression in tumor tissue (Figure 3). Together, these data suggest a shift in IEC marker expression in tumor samples, with an increase in Paneth cell markers and a decrease in EEC and goblet cell markers.

We next explored whether this shift also occurs in spontaneously developing CRC, without inflammation induction, by examining the dataset GSE107139, obtained from the APC^min CRC model [42]. Consistent with our observations, APC^min animals exhibited the downregulation of EEC markers and upregulation of Paneth cell markers in tumor tissue (Figure S2A,C). We then assessed whether human CAC exhibits a similar pattern using the GSE37283 dataset. This dataset confirmed the upregulation of LYZ and WISP1, alongside the downregulation of EEC identity genes (GCG, PYY, CHGA, and NEUROD1) (Figure S2B,C). Interestingly, goblet cell markers were not reduced in this CAC dataset, suggesting that this phenomenon may be specific to CAC, further reinforcing the centrality of Paneth cell expansion and EEC loss as conserved features across tumorigenesis (Figure S2A–C).

The STRING platform for protein–protein interaction (PPI) network analysis reveals the largest interaction cluster among the downregulated genes (Figure 1F), mainly involved enteroendocrine-related genes, with GCG as the central node (Figure 4). This is in line with the enrichment analysis highlighting downregulated hormonal processes (Figure 2B), suggesting a key role for EEC genes in human CRC. However, upregulated genes did not form a highly connected network related to Paneth cells, with interactions restricted to only two genes (Figure S4A).

Together with the gene expression data, these results suggest a balance favoring a decrease in EECs and an increase in/maintenance of Paneth cell markers in the TME, with the GCG gene emerging as a potential marker of this phenomenon.

2.4. RNA-Seq Analysis from TCGA Confirms That EEC Markers Are Among Most Downregulated Genes

Given the consistent shift observed, we used the Cancer Genome Atlas Program (TCGA) database, which contains RNA sequencing data from patients with colon adenocarcinoma (COAD), to identify potential therapeutic targets. We validated the downregulation of EEC markers and others using a separate COAD cohort. Remarkably, two EEC markers—GCG and PYY—were among the top 25 most downregulated genes in COAD tissue (Figure 5). In contrast, none of the top 25 upregulated genes were associated with other IEC subtypes (Figure S5A). This indicates that EEC-related genes are disproportionately represented among the downregulated transcripts, reinforcing the observation of consistent EEC marker downregulation across COAD tumor samples.

2.5. Modulation of EEC and Paneth Cell Markers Occurs Independently of Tumor Stage

We investigated whether this IEC marker shift is related to tumor development, which is staged based on the extent of infiltration and metastasis [43]. When comparing COAD stages (I through IV), the expression of most IEC-associated genes did not differ significantly across stages (Figure S6A–X). Nonetheless, an analysis of metastatic versus non-metastatic COAD tissue revealed that EEC-related genes (GCG, PYY, CHGA, NEUROD1) were consistently downregulated in both tumor and normal tissue and between metastatic and normal tissue (Figure 6). This suggests that the reduction in EEC markers occurs in the early stages of tumorigenesis and persists throughout tumor development, even in metastatic progression. Paneth cell markers showed an increase in tumor samples regardless of the tumor stage (Figure 6).

2.6. Protein-Level Evidence Supports EEC/Paneth Cell Marker Balance in CRC

To confirm the mRNA findings at the protein level, we utilized the Human Protein Atlas. As expected, 100% of all normal tissue was positive for EEC markers (GCG, PYY, CHGA, and NK2 homeobox 2 (NKX2.2)) and nearly 100% for Paneth markers (DEFA1, LYZ, SOX9, and WISP1) (Figure S3A). While the percentage of Paneth cell markers remained unchanged in CRC and normal tissue, CRC tissue positive for EEC markers was rare (Figure S3A). Immunohistochemistry imaging visualized the loss of the ECC marker CHGA and the preservation of the Paneth cell marker DEFA1 (Figure S3A–C).

We then analyzed the National Cancer Institute’s Proteomics Tumor Analysis Consortium (CPTAC) dataset to see if the transcriptional profile was maintained at the protein level. Using the PCA plot, we verified the sample quality and group separation (normal tissue (blue) and tumor tissue (red)) (Figure 7A). Differential expression analysis, using the same cut-offs ($\mathrm{FDR}<0.05$; ${\log }_{2}FC<-1>\mathrm{and}>1$), identified eight relevant proteins: CHGA, GCG, PYY, LYZ, SOX9, MUC2, DPP4, and KLF4 (Figure 7B). This analysis confirmed the mRNA findings: a reduction in GCG, CHGA, and PYY was associated with an increase in LYZ and SOX9 (Figure 7C), supporting a balance between these two cell populations in the TME.

2.7. EEC Modulation in TME Is Linked to Apoptosis and Proliferation and Driven by WNT Signaling

WNT activation is known to cause an increase in Paneth cells [44]. Nevertheless, the mechanism underlying the decrease in EECs within the TME remains unclear. To investigate this, we performed a mediation analysis using the CPTAC proteomics dataset. This analysis tested whether the effect of WNT activation on EEC marker expression is mediated indirectly through Paneth cell expansion or directly. Our findings showed that the wnt_score (calculated based on the expression of CTNNB1, MYC, AXIN2) had a direct negative effect on EEC proteins (GCG, CHGA, and PYY), as shown by the average direct effect (ADE) index (

Table 2. Mediation analysis of wnt_score effects on EEC marker expression through SOX9 and LYZ.

Mediator	Outcome	ACME_ Estimate	ACME_p	ADE_ Estimate	ADE_P Value	Summary
SOX9	CHGA	−0.39	0.27	−1.15	0.02	Direct effect of wnt_score on CHGA, independent of SOX9
SOX9	GCG	−0.10	0.82	−1.12	0.03	Direct effect of wnt_score on GCG, independent of SOX9
SOX9	PYY	−0.09	0.81	−1.37	0.01	Direct effect of wnt_score on PYY, independent of SOX9
LYZ	CHGA	−0.18	0.18	−1.36	0.0001	Direct effect of wnt_score on CHGA, independent of LYZ
LYZ	GCG	−0.17	0.17	−1.05	0.0001	Direct effect of wnt_score on GCG, independent of LYZ
LYZ	PYY	−0.19	0.21	−1.27	0.0001	Direct effect of wnt_score on PYY, independent of LYZ

). This indicates that increased Wnt activity is directly associated with reduced EEC marker expression. While the average causal mediation effect (ACME) index showed a trend of a negative indirect effect via SOX9 and LYZ, it was not statistically significant (ACME $<0.05$), suggesting that the Wnt signaling modulation of EEC markers occurs predominantly through a direct mechanism (Table 2). Nonetheless, mediation through LYZ showed a trend toward significance, pointing to a possible partial contribution of this protein expression to the Wnt–EEC regulatory axis.

To correlate EEC downregulation with potential mechanisms, we performed a complementary functional enrichment analysis. We calculated an EEC score for each sample based on the mean normalized abundance of the GCG, CHGA, and PYY proteins and correlated it with all quantified proteins for gene set enrichment analysis (GSEA). The analysis revealed that samples with lower EEC scores showed enrichment in proliferative and apoptotic pathways such as “cell cycle” and “TP53 transcriptional regulation” (Figure 7D). Together, these findings suggest that reduced EEC activity in CRC correlates with the activation of cell cycle and apoptosis pathways, potentially related to the loss or differentiation of EECs.

2.8. High GCG and CHGA Expression Correlates with Improved Overall Survival

We evaluated whether the Paneth cell marker upregulation and EEC marker downregulation could predict patient survival and treatment responses. A survival analysis using the Human Protein Atlas database was performed to evaluate the prognostic relevance of EEC and Paneth cell markers in CRC patients. Patients were stratified into high- and low-expression groups based on transcript levels, and overall survival was assessed using Kaplan–Meier estimates and the log-rank test. Interestingly, only GCG and LYZ demonstrated a significant association with patient outcomes (Figure 8A,B; Table S5). Specifically, high GCG expression was associated with longer 5-year survival in COAD patients (Figure 8A). Similarly, the EEC marker CHGA showed a positive association between elevated expression and prolonged survival (Figure 8B). None of the Paneth cell markers were significantly associated with 5-year patient survival (Table S5). We then investigated whether the expression of these genes correlated with chemotherapy responsiveness. Our analysis showed that chemotherapy-responsive patients tended to have higher GCG, while there was no difference regarding CHGA expression (Figure 8C,D). Collectively, these findings indicate that the downregulation of EEC markers directly impacts patient survival and, in the case of GCG, may also be linked to the chemotherapy response.

3. Discussion

In silico analyses using publicly available sequencing data are a powerful tool for deriving comprehensive overviews and translational conclusions about biological phenomena. Our findings reveal that tumor tissue from CRC patients and experimental models exhibits consistent differences in IEC markers when compared to non-neoplastic tissue. This reinforces the concept that distinct IEC subtypes play key but contrasting roles in the TME. Among the secretory IECs, goblet cells—responsible for mucus production—are particularly affected during tumor development. Their reduced numbers in the tumor context, associated with decreased mucus secretion, are linked to a worse prognosis and more aggressive disease [45]. Our work agrees with the established literature, showing transcriptional alterations and the downregulation of critical differentiation and function genes for goblet cells (MUC2, KLF4, and SPDEF). On the other hand, tuft cells play a dual role in the tumor microenvironment depending on the context. They can promote tumor progression by promoting tumor growth and metastasis through epithelial–mesenchymal transition [46,47,48]. In our analyses, however, the markers of these cells showed heterogeneous expression patterns, suggesting the necessity of additional analyses focusing on other markers and signaling pathways related to this cell subtype.

The role of EECs in tumor biology is controversial, with studies suggesting both promoting and protective influences in different cancers [49,50]. Previous CRC reports, which primarily focused on single EEC markers (e.g., CHGA or PYY), have yielded inconsistent prognostic results [51,52,53,54]. Conversely, NEUROD1 has been implicated in promoting tumor growth and metabolic reprogramming in CRC [55,56]. This duality likely reflects the specificity of the tumor type, the complexity of the hormones’ systemic effects, and their modulation of metabolic and immunological pathways.

Our study advances beyond previous work by providing the comprehensive, cross-species, transcriptome- and proteome-wide characterization of EEC marker expression in CRC. Instead of isolated proteins, we identified a downregulated network of EEC markers (CHGA, GCG, PYY, NEUROD1) associated with tumor processes, reduced survival, and poor chemotherapy responses. This reveals that EEC loss is not merely a histological feature but is part of the functional reprogramming of the intestinal TME, paving the way for mechanistic studies on the EEC–CRC relationship.

While the exact role of EECs in CRC remains unclear, the hormones that they produce may influence neoplastic processes, particularly locally in the intestine. The incretin group (GIP and GLP-1) is of high interest. A recent study reported that patients treated with incretin-based antidiabetic drugs exhibited a lower risk of developing CRC compared with those receiving other medications, independently of weight loss [57]. Since GCG encodes the precursor of GLP-1, the observed reduction in GCG and other key EEC markers (CHGA, NEUROD1, PYY) in tumor tissue, combined with these external clinical data, suggests that EECs and incretin hormones may exert an anti-tumor role in colorectal carcinogenesis.

Interestingly, Paneth cells are producers of antimicrobial peptides that interact with the gut microbiota and modulate immune cells [21], showing a contrasting pattern, being increased in tumor tissue, as indicated by the upregulation of the genes LYZ, DEFA1, SOX9, and WISP1. This, combined with the downregulation of EEC and goblet cell markers, strengthens the hypothesis of a functional balance or shift between secretory IEC populations in the TME. Our GO enrichment analysis supports this shift: upregulated genes were consistently correlated with antimicrobial peptide-mediated immune processes, while downregulated genes were associated with hormonal processes, the peptide hormone response, and hormonal metabolic processes. It is important to note that Paneth cells are typically absent in the large intestine; therefore, the increase in their markers in CRC highlights the complexity of tumor cell development, potentially representing metaplasia. Further studies are warranted to understand the impact of this cellular shift on tumor progression, anti-tumor immunity, and changes in the microbiota composition [39,58].

The WNT pathway is essential for all secretory lineages [59], but WNT hyperactivation specifically promotes Paneth cell differentiation [44]. The upregulation of the WNT signaling marker WISP1 in tumor tissue suggests that the TME upregulates Paneth cells through increased WNT signaling. This is consistent with the finding that Paneth cell accumulation in CRC is linked to WNT/β-catenin signaling activation and a poor prognosis [60]. Crucially, our results show that WNT signaling is directly associated with both an increase in the Paneth cell population and a decrease in EECs. Mediation analysis indicated that this WNT–EEC relationship is primarily direct, with only a limited, non-significant contribution from Paneth cells (LYZ showed a trend). This suggests that WNT hyperactivation in the TME promotes the secretory cell shift, favoring Paneth cells over EECs, via a direct mechanism, potentially affecting intestinal hormone secretion and local immune modulation. Our results point to a possible relevant functional impact of WNT signaling on the cellular composition in the TME. However, experiments are needed to confirm this mechanism and to determine which possible modulations in this pathway and these cell populations could influence tumor formation.

In the TME, several processes and pathways occur that are overactivated, including pathways related to apoptotic processes, DNA repair, p53 signaling, cell proliferation, and the cell cycle, which are characteristic of tumor development. We investigated the functional impact of EEC downregulation. Samples with lower EEC scores were associated with the activation of the cell cycle, apoptotic and proliferative pathways in GSEA. This indicates that reduced EEC activity correlates with greater proliferation and apoptosis susceptibility in the TME, supporting the notion that WNT-mediated EEC downregulation alters the cellular composition and impacts key tumor-related biological processes [61].

Our findings across multiple cohorts consistently demonstrate conserved, dynamic modulation between EECs and Paneth cells in the TME. This cellular interaction appears to have important clinical implications, as it links to the therapeutic response and patient outcomes. Here, we identified GCG and CHGA as key genes associated with survival in CRC patients. High expression of these EEC markers correlated with a better prognosis and significantly improved overall survival in COAD patients. This is consistent with a recent study reporting a positive correlation between GCG expression in immune cells within COAD tumors and favorable patient outcomes [62]. Future studies in patient cohorts will be needed to adjust for clinical variables and validate the findings.

In summary, the molecular signatures of epithelial subpopulations not only reflect tumor biology but may also serve as a prognostic indicator and potential therapeutic target in CRC. A limitation of our study, based exclusively on public datasets, is that some targeted markers were absent from the protein expression datasets, and, in some cases, the antibodies failed to produce detectable staining in any of the tissues examined. These highlight the critical need for follow-up experimental work and clinical approaches to validate these transcriptomic patterns and mechanistically confirm the importance of this WNT–Paneth–EEC balance in CRC development.

4. Materials and Methods

4.1. Database Collection

Datasets for large-scale gene expression profiling were retrieved from the GEO at NCBI, containing intestinal samples from both animal models of CRC and human subjects diagnosed with CAC. For subsequent analyses, we selected the three datasets available on the NCBI platform that had the same cancer induction model as well as the same sequencing and data analysis method. From mice, we included GSE86299, GSE31105, and GSE64658, representing experimental models of CAC induced by azoxymethane and dextran sulfate sodium, as well as GSE107139 from APC^min mice, which spontaneously develop colorectal tumors due to a mutation in the tumor suppressor gene APC. From human samples, we analyzed datasets derived from CRC biopsies (GSE8671, GSE18105, and GSE110225) and from cases of CAC (GSE37283).

4.2. Differential Expression Analysis Datasets from NCBI

Microarray differential expression was analyzed using the Network Analyst platform [63] in conjunction with the limma (version 3.16.2) R package. Statistical cut-offs were set at an $\mathrm{FDR}<0.05$ (Benjamini–Hochberg method) and ${\log }_{2}FC<-1>\mathrm{and}>1$. PCA graph and volcano plots were generated based on these criteria. A bubble-based heatmap was created using R (version 4.5.1) with the ggplot2 (version 4.0.0), readr (version 2.1.5), dplyr (version 1.1.4), and scales (version 1.4.0) packages. Genes passing the established cut-offs were separated into up- and downregulated lists for subsequent enrichment analysis [64].

4.3. RNA-Seq Differential Expression Analysis from TCGA Database

RNA-Seq data from the TCGA-COAD cohort were analyzed for differential expression segmented by tumor stage (I, II, III, and IV, based on the American Cancer Society criteria) using the Oncodb platform [65], applying the same $\mathrm{FDR}<0.05$ and ${\log }_{2}FC<-1>\mathrm{and}>1$ cut-offs as above. The top 25 up- and downregulated genes in COAD were obtained from the integrated cancer data analysis platform Ualcan [66]. PPI network analysis of differentially expressed genes was performed using the STRING platform [67].

4.4. Proteomics Analysis from CPTAC, Differential Expression, and GSEA

Differential protein expression analysis between tumor and normal samples was performed on the CPTAC database (PDC000116). Linear models were used, and adjusted p-values ($\mathrm{FDR}<0.05$) were calculated using the Benjamini–Hochberg method implemented in limma. To investigate pathway-level associations, GSEA was conducted. The EEC score was calculated as the mean expression of EEC markers (GCG, CHGA, PYY). All proteins were ranked by Pearson correlation with the calculated EEC score. The enrichment of Reactome pathways was analyzed using fgseaMultilevel with 1000 bootstrap iterations, and adjusted p-values were corrected by $\mathrm{FDR}$. All analyses were conducted in R (version 4.5.1) using the packages limma (version 3.64.3), fgsea (version 1.34.2), clusterProfiler (version 4.16.0), reactome.db (version 1.92.0), org.Hs.eg.db (version 3.21.0), mediation (version 4.5.1), ppcor (version 1.1), and tidyverse (version 2.0.0).

4.5. Mediator Analyses with Proteomics Data

Mediator analysis was performed using the CPTAC dataset (PDC000116) to examine the indirect effects of WNT pathway activity on EEC markers. The mediate() function was used to estimate the ACME and the ADE, with non-parametric bootstrapping (1000 simulations) to calculate confidence intervals and p-values. Expression data were log-transformed and standardized prior to modeling. The WNT pathway activation score (wnt_score) was calculated as the mean expression of canonical WNT target proteins (CTNNB1, MYC, AXIN2). First, the mediation model was examined, in which the expression of LYZ or SOX9 (Paneth cell markers) was regressed on the wnt_score (LYZ or SOX9~wnt_score). Second, the outcome model was examined, in which the expression of GCG, PYY, or CHGA (enteroendocrine marker) was regressed on either LYZ or SOX9 or the WNT score (GCG/PYY/CHGA~LYZ/SOX9 + wnt_score).

4.6. Survival Analysis and Chemotherapy Responder Correlation

Survival analysis and the association of gene/protein expression with patient outcomes were performed using the Human Protein Atlas platform for the TCGA COAD patient database [68]. Overall survival was evaluated using Kaplan–Meier estimates and assessed with the log-rank test. It is important to note that these analyses did not account for patient variables such as stage, age, or sex. Moreover, the correlation between chemotherapy responsiveness and the expression of GCG and CHGA expression was performed using the rocplot.com platform [69].

5. Conclusions

This integrative study, combining transcriptomics and proteomics analyses across human and experimental models, reveals a conserved and dynamic shift in the gene expression balance between EEC and Paneth cell markers within the CRC TME. This balance is accompanied by the upregulation of biological pathways associated with antimicrobial peptide-mediated immune responses and, conversely, the downregulation of hormone-related processes. Furthermore, we provide robust evidence linking the reduced expression of key enteroendocrine-associated genes—GCG and CHGA—to poorer survival outcomes in patients with CRC. Mediation analysis further suggests that WNT pathway activation may directly suppress enteroendocrine markers, as well as partly through Paneth cell activity. Finally, proteome-based enrichment analysis demonstrated that samples with lower EEC scores were enriched for cell cycle and apoptosis pathways, suggesting that the loss of EEC activity is tightly linked to key proliferative and cell death mechanisms characteristic of cancer progression. Taken together, these results establish that the disruption of EEC signaling and the simultaneous expansion of Paneth cell markers are hallmarks of CRC progression, holding significant prognostic and therapeutic value for the treatment of CRC patients.