Adipocytes impact on gastric cancer progression: Prognostic insights and molecular features

Abstract

BACKGROUND

Adipocytes, especially adipocytes within tumor tissue known as cancer-associated adipocytes, have been increasingly recognized for their pivotal role in the tumor microenvironment of gastric cancer (GC). Their influence on tumor progression and patient prognosis has sparked significant interest in recent research. The main objectives of this study were to investigate adipocyte infiltration, assess its correlation with clinical pathological features, develop a prognostic prediction model based on independent prognostic factors, evaluate the impact of adipocytes on immune cell infiltration and tumor invasiveness in GC, and identify and validate genes associated with high adipocyte expression, exploring their potential diagnostic and prognostic value.

AIM

To explore the relationship between increased adipocytes within tumor tissue and prognosis in GC patients as well as the associated mechanisms and potential biomarkers, using public databases and clinical data.

METHODS

Using mRNA microarray datasets from the Gene Expression Omnibus database and clinical samples from Jiangsu Provincial Hospital, survival and regression analyses were conducted to determine the relevant prognostic factors in GC. Feature gene selection was performed using least absolute shrinkage and selection operator and support vector machine recursive feature elimination algorithms, followed by differential gene expression analysis, gene ontology, pathway analysis, and Gene Set Enrichment Analysis. Immune cell infiltration was analyzed using the CIBERSORT algorithm.

RESULTS

Tumor adipocyte infiltration correlated with poor prognosis in GC, leading to the development of a highly accurate and discriminative prognostic prediction model. Key genes, ADH1B, SFRP1, PLAC9, and FABP4, were identified as associated with high adipocyte expression in GC. The diagnostic and prognostic potential of these identified genes was validated using independent datasets. Downregulation of immune cells was observed in GC with high adipocyte expression.

CONCLUSION

GC with high intratumoral adipocyte expression demonstrated aggressive tumor biology and a poorer prognosis. The genes ADH1B, SFRP1, PLAC9, and FABP4 have been identified as holding diagnostic and prognostic significance in GC. These findings strongly support the use of adipocyte expression as a valuable indicator of tumor invasiveness and anticipated patient outcomes in GC.

Key Words: Gastric cancer; Adipocytes; Obesity; Tumor biomarker; Cancer associated adipocytes

Core Tip: Adipocytes within tumor tissue [cancer-associated adipocytes (CAA)] play a crucial role in gastric cancer (GC) progression and patient prognosis. This study confirmed that increased CAA expression correlated with adverse GC outcomes, independent of pathological features but potentially linked to patient age. A prognostic model based on key factors offered high accuracy for clinical decisions. Genes like ADH1B, SFRP1, PLAC9, and FABP4 showed diagnostic and prognostic promise with high CAA expression. Immune analysis revealed reduced immune cells in high CAA GC, suggesting increased tumor aggressiveness. These findings underscore the significance of adipocytes in GC, offering potential biomarkers for future diagnostics and therapeutics.

INTRODUCTION

Gastric cancer (GC), one of the most prevalent malignant tumors in the global digestive system, ranks as the fourth leading cause of cancer-related deaths worldwide and holds the fifth position in terms of global incidence[1]. The distinct features of GC, characterized by high incidence, elevated metastasis rates, increased mortality rates, low early diagnosis rates, limited curative resection rates, and low 5-year survival rates, underscore its significance as a pivotal focus in medical research[2,3]. The pathogenesis of GC is intricate, marked by inconspicuous early symptoms, leading to low rates of early diagnosis and advanced-stage presentations during consultation[4-6].

Obesity is closely associated with an elevated risk of various cancer types, including gastric, colorectal, bladder, liver, kidney, pancreatic, and breast cancers[7-10]. Additionally, obesity is correlated with an increased likelihood of cancer-related fatalities[11]. The accumulation of fat tissue in obesity leads to the release of inflammatory factors, altering the gastric tissue microenvironment and prompting abnormal mucosal cell differentiation, thereby contributing to the development of GC[12].

Adipocytes play a crucial role in the tumor microenvironment, engaging in signal interactions and substance exchange with tumor cells and other stromal cells. They create a conducive growth environment for tumors, actively promoting their malignant biological behaviors[13]. Cancer-associated adipocytes (CAA), a subtype of fat cells with malignant phenotypes, actively participate in the entire spectrum of tumor initiation, development, invasion, metastasis, and treatment response. These cells secrete diverse factors, modify the characteristics and behaviors of inflammatory immune cells, participate in tumor inflammation reactions and immune regulation, reprogram metabolism to meet the demands of tumor cells, and facilitate tumor progression and metastasis by modifying the extracellular matrix[14-17]. Additionally, CAAs are implicated in tumor resistance mechanisms, potentially impeding the efficacy of cancer treatments[18].

Previous studies have demonstrated that co-cultivation of tumor cells with fat cells results in the acquisition of a more invasive phenotype by tumor cells[19]. Notably, GC cells co-cultivated with fat cells exhibited enhanced invasive capabilities[20]. Currently, the infiltration of tumor-associated adipocytes in GC progression and mechanisms remains unclear. Therefore, we conducted a retrospective analysis identifying patients with tumor-associated adipocyte infiltration, analyzing their clinicopathological characteristics and survival outcomes, and exploring the microlevel mechanisms involved. Targeting the interactions between CAA and tumor cells provides a novel perspective for cancer treatment strategies.

MATERIALS AND METHODS

Collection of clinical and transcriptomic data for GC patients

We utilized the terms “gastric cancer” and “survival” to search and filter for mRNA microarray datasets in the Gene Expression Omnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo/) based on the following criteria. First, these datasets had to provide raw data for further analysis. Second, clinical information had to encompass survival and pathological details. One GC-related dataset (GSE84437) was chosen for the next phase of our research. We utilized the xCell computational algorithm to quantify adipocytes within tumor tissues by analyzing public GC transcriptomic data[21]. The algorithm calculated the density of adipocytes within the tumor volume, and these densities were subsequently used to define two groups: Adipocyte high (AH), comprising samples up to the 10^th percentile; and the remaining samples categorized as the adipocyte low (AL) group.

Clinical experimental validation

This study was approved by the Ethics Committee of Nanjing University of Chinese Medicine (No. 2022NL-137-01). This retrospective study was conducted from January 2016 to October 2020 at Jiangsu Provincial Hospital, involving 425 patients diagnosed with GC. All enrolled patients underwent pathological diagnosis confirming adenocarcinoma and received radical tumor resection surgery along with lymph node dissection. Demographic, clinical, and pathological information, including age, sex, tumor staging, as well as neural and vascular infiltration, were meticulously collected. Clinical staging adhered to the guidelines outlined in the 8^th edition of the American Joint Committee on Cancer Staging Manual. Additionally, tumor tissue specimens from the included patients were subjected to hematoxylin and eosin staining. The stained slides were independently assessed by two pathologists. Under high magnification, they carefully observed the presence of small, dispersed adipocytes with tiny lipid droplets within the cancer tissue or intermingling of adipocytes intricately intertwined with the cancerous tissue. Identification of these features constituted the confirmation of adipocyte infiltration. Overall survival (OS) was defined as the time from the date of surgery to the patient’s death from any cause, while disease-free survival (DFS) was defined as the time from the date of surgery to the confirmation of cancer recurrence or patient death based on imaging. These endpoints were achieved through telephone follow-ups or medical records until the cutoff date of the study on December 4, 2023.

Identification and analysis of differentially expressed genes

The “limma” R package was employed for differential expression analysis to identify genes with differential expression between the AH and AL groups in the GSE84437 dataset. A cutoff of absolute log2 (fold change) > 2 and adjusted P value < 0.05 was used. Visualization of these differences was performed using the “ggpubr” and “Heatmap” R packages, generating volcano plots and heatmaps, respectively. In the analysis of differentially expressed gene (DEG), we utilized the RStudio software packages “ggplot2”, “clusterProfiler”, “enrichplot”, and “org.Hs.eg.db” to perform Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes enrichment analyses to gain a deeper understanding of the biological significance of these genes. The final enrichment results were obtained with a filtering criterion of P value < 0.05. To further elucidate the biological pathways involved, we employed Gene Set Enrichment Analysis (GSEA), a recognized and powerful tool for analyzing gene enrichment pathways based on functional categories. This analysis was conducted to compare the biological pathways between the case and control groups. GSEA was executed using the GSEA software with a significance threshold of P value < 0.05.

Feature selection with least absolute shrinkage and selection operator and support vector machine recursive feature elimination algorithms

We employed least absolute shrinkage and selection operator (LASSO) and support vector machine recursive feature elimination (SVM-RFE) algorithms for feature gene selection among DEGs. LASSO, a regression analysis algorithm implemented in the R software package ‘glmnet,’ identifies genes significantly associated with different samples through regularization for variable selection. Another machine learning method, SVM-RFE, a widely used supervised protocol for classification and regression, identifies the best variables by eliminating SVM-generated eigenvectors. The ‘Caret’ package, utilizing a grid search method, selected hyperparameters for all classifiers via 10-fold cross-validation on the training dataset. SVM-RFE, employing the ‘e1071’ package, identifies biomarkers with higher discriminative power, contributing to their diagnostic value. We combined overlapping genes from both LASSO and SVM-RFE algorithms for further analysis. A two-sided P value < 0.05 was considered statistically significant. Subsequently, we validated their expression levels to assess their potential as candidate diagnostic biomarkers in both GSE54129 and TCGA datasets. We identified these overlapping genes online using the Kaplan-Meier Plotter database, obtained prognostic information, and evaluated their potential as prognostic biomarkers for GC.

Evaluation of immune cell infiltration

CIBERSORT is a deconvolution algorithm that applies linear support vector regression to deconvolve the bulk gene expression matrix, treating deconvolution as a machine learning method for denoising. The analysis process was executed using the CIBERSORT package in RStudio. Only samples with a CIBERSORT output P value < 0.05 were considered for further analysis. The correlation between candidate diagnostic biomarkers and significantly altered immune cells was assessed using the ‘reshape2’ and ‘ggExtra’ R packages through the Spearman correlation coefficient.

Statistical analysis

Statistical analyses were conducted using SPSS 25.0 and R software (version 4.2.3). Group comparisons utilized the Wilcoxon signed-rank test or Student’s t-test for variables between groups. Survival analysis employed the Kaplan-Meier method, and univariate analysis utilized Cox regression. Variables with P < 0.05 in univariate analysis were included in subsequent multivariate analysis. A nomogram was constructed based on independent prognostic variables, using a formula derived from the survival curve for individual risk score calculation. In addition, time-dependent receiver operating characteristic (ROC) curves of the nomogram and all independent prognostic variables at 12 mo, 36 mo, and 60 mo were generated, and the corresponding time-dependent area under the curves were applied to show discrimination. Calibration curves at 12 mo, 36 mo, and 60 mo were plotted to evaluate the nomogram. The significance threshold for all analyses was set at P ≤ 0.05.

RESULTS

Baseline characteristics of the study population

We recruited 428 GC patients from the GEO database as the training group, stratifying them into 42 individuals in the AH group and 386 individuals in the AL group based on the adipocyte scoring. The characteristics of GC in the training group are summarized in Table 1. Additionally, we collected 425 GC patients from Affiliated Hospital of Nanjing University of Chinese Medicine as the validation group. According to hematoxylin and eosin staining results, 61 cases (14.35%) in the validation group exhibited infiltration of adipocytes within the tumor (Figure 1). The characteristics of GC in the validation group are summarized in Table 2. The average age of the training and validation groups was 60.70 years (range: 27-86 years) and 62.90 years (range: 25-90 years), respectively. In the training group, tumor infiltration was correlated with age (P = 0.002) but not with sex (P = 0.474), perineural invasion (P = 0.107), vascular invasion (P = 0.519), or TNM staging (P > 0.05). In the validation group, tumor infiltration was not associated with age, sex, perineural invasion, vascular invasion, or TNM staging (P > 0.05).

Figure 1 Typical tumor adipocyte infiltration in hematoxylin and eosin staining.

Table 1 The training group included clinical and pathological features of gastric cancer patients, n = 428, n (%).

Clinical characteristics	Adipose infiltration			P value
	Low, n = 386	High, n = 42	Overall, n = 428
Age in yr				0.002
> 60	221 (57.30)	13 (31.00)	234 (54.70)
≤ 60	165 (42.70)	29 (69.00)	194 (45.30)
Sex				0.474
Male	265 (68.70)	26 (61.90)	291 (68.00)
Female	121 (31.30)	16 (38.10)	137 (32.00)
Perineural invasion				0.107
No	319 (82.60)	35 (83.30)	354 (82.70)
Yes	67 (17.40)	7 (16.70)	74 (17.30)
Vascular invasion				0.519
No	205 (53.10)	20 (47.60)	225 (52.60)
Yes	181 (46.90)	22 (52.40)	203 (47.40)
M stage				1.000
M0	371 (96.10)	41 (97.60)	412 (96.30)
M1	15 (3.90)	1 (2.40)	16 (3.70)
T stage				0.812
T1	11 (2.80)	0 (0.00)	11 (2.60)
T2	35 (9.10)	3 (7.10)	38 (8.90)
T3	83 (21.50)	8 (19.00)	91 (21.30)
T4	257 (66.60)	31 (73.80)	288 (67.30)
N stage				0.318
N0	71 (18.40)	8 (19.00)	79 (18.50)
N1	171 (44.30)	14 (33.30)	185 (43.20)
N2	114 (29.50)	18 (42.90)	132 (30.80)
N3	30 (7.80)	2 (4.80)	32 (7.50)

Table 2 The validation group included clinical and pathological features of gastric cancer patients, n = 425, n (%).

Clinical characteristics	Adipose infiltration			P value
	Negative, n = 364	Positive, n = 61	Overall, n = 425
Age in yr				0.138
> 60	223 (61.30)	44 (72.10)	267 (62.80)
≤ 60	141 (38.70)	17 (27.9)	158 (37.20)
Sex
Male	257 (70.60)	47 (77.00)	304 (71.50)
Female	107 (29.40)	14 (23.00)	121 (28.50)
Perineural invasion				1.000
No	295 (81.00)	49 (80.30)	344 (80.90)
Yes	69 (19.00)	12 (19.70)	81 (19.10)
Vascular invasion				0.396
No	284 (78.00)	44 (72.10)	328 (77.20)
Yes	80 (22.00)	17 (27.90)	97 (22.80)
M stage				0.012
M0	355 (97.50)	55 (90.20)	410 (96.50)
M1	9 (2.50)	6 (9.80)	15 (3.50)
T stage				0.238
T1	206 (56.60)	43 (70.50)	249 (58.60)
T2	85 (23.40)	10 (16.40)	95 (22.40)
T3	65 (17.90)	8 (13.10)	73 (17.20)
T4	8 (2.20)	0 (0.00)	8 (1.90)
N stage				0.444
N0	251 (69.00)	36 (59.00)	287 (67.50)
N1	35 (9.60)	7 (11.50)	42 (9.90)
N2	39 (10.70)	10 (16.40)	49 (11.50)
N3	39 (10.70)	8 (13.10)	47 (11.10)

Risk factors for poor prognosis in GC patients

According to Kaplan-Meier analysis results, significant differences were observed in both DFS and OS between the AH and AL groups, with patients in the AH group exhibiting poorer outcomes (P < 0.05, as shown in Figure 2). Based on the results of univariate and multivariate analyses from the training and validation sets, excluding sex, T stage, lymph node metastasis, vascular invasion, and neural invasion, we experimentally confirmed that distant metastasis, adipocyte infiltration, and age were independent prognostic factors for adverse OS (Tables 3 and 4). Similarly, distant metastasis and adipocyte infiltration independently predicted adverse DFS (Table 5).

Figure 2 Kaplan-Meier survival analysis of patients with adipocyte infiltration. A: Survival analysis in the training group; B: Survival analysis in the validation group; C: Disease-free survival analysis in the validation group.

Table 3 Univariate and multivariate analysis of prognostic factors for overall survival in the training set.

Parameter	Univariate analysis		Multivariate analysis
	HR (95%CI)	P value	HR (95%CI)	P value
Age in yr		0.006
≤ 60	0.675 (0.533- 0.854)	0.006	0.571 (0.446-0.731)	< 0.001
> 60
Sex		0.030
Male	1.396 (1.077-1.809)	0.034	1.305 (1.003-1.697)	0.096
Female
Adipose infiltration		0.009
High	1.854 (1.290-2.664)	0.005	2.143 (1.472-3.121)	< 0.001
Low
Perineural invasion		0.061
No
Yes	1.424 (1.055-1.922)	0.053
Vascular invasion		0.019
No
Yes	1.389 (1.103-1.749)	0.019	1.057 (0.829-1.349)	0.707
M stage		< 0.001
M0
M1	7.094 (4.528-11.114)	< 0.001	5.424 (3.392-8.675)	< 0.001
T stage		< 0.001
T1
T2	1.092 (0.297-4.008)	0.911	0.757 (0.203-2.817)	0.728
T3	3.025 (0.919-9.955)	0.127	1.783 (0.532-5.970)	0.432
T4	3.902 (1.210-12.583)	0.056	2.220 (0.675-7.302)	0.270
N stage		0.845
N0		< 0.001
N1	1.542 (1.057-2.251)	0.060	1.406 (0.956-2.066)	0.146
N2	3.012 (2.064-4.395)	< 0.001	2.230 (1.540-3.430)	< 0.001
N3	3.895 (2.400-6.321)	< 0.001	2.394 (1.438-3.987)	0.005

CI: Confidence interval; HR: Hazard ratio.

Table 4 Univariate and multivariate analysis of prognostic factors for overall survival in the validation set.

Parameter	Univariate analysis		Multivariate analysis
	HR (95%CI)	P value	HR (95%CI)	P value
Age in yr		0.003
≤ 60	0.518 (0.354-0.757)	0.004	0.576 (0.392-0.846)	0.018
> 60
Sex		0.219
Male
Female	0.752 (0.509-1.111)	0.230
Adipose infiltration		< 0.001
High
Low	0.256 (0.180-0.364)	< 0.001	0.283 (0.198-0.404)	< 0.001
Perineural invasion		0.401
No
Yes	1.225 (0.823-1.823)	0.401
Vascular invasion		0.045
No
Yes	1.570 (1.097-2.249)	0.039	1.401 (0.963-2.038)	0.139
M stage		0.001
M0
M1	3.780 (2.124-6.728)	< 0.001	2.355 (1.278-4.338)	0.021
T stage		0.859
T1
T2	0.930 (0.614-1.411)	0.775
T3	0.839 (0.522-1.348)	0.542
T4	1.382 (0.522-3.661)	0.585
N stage		0.845
N0
N1	0.829 (0.447-1.536)	0.616
N2	1.172 (0.711-1.933)	0.601
N3	1.478 (0.910-2.402)	0.185

CI: Confidence interval; HR: Hazard ratio.

Table 5 Univariate and multivariate analysis of prognostic factors for disease-free survival in the validation set.

Parameter	Univariate analysis		Multivariate analysis
	HR (95%CI)	P value	HR (95%CI)	P value
Age in yr		0.025
≤ 60	0.629 (0.443-0.893)	0.029	0.665 (0.468-0.945)	0.056
> 60
Sex		0.302
Male
Female	0.793 (0.547-1.148)	0.293
Adipose infiltration		< 0.001
High
Low	0.347 (0.243-0.494)	< 0.001	0.362 (0.254-0.517)	< 0.001
Perineural invasion		0.732
No
Yes	0.918 (0.605-1.392)	0.735
Vascular invasion		0.239
No
Yes	1.299 (0.908-1.857)	0.229
M stage		0.003
M0
M1	3.315 (1.868-5.883)	< 0.001	3.057 (1.719-5.436)	0.001
T stage		0.937
T1
T2	0.870 (0.580-1.305)	0.572
T3	0.897 (0.579-1.391)	0.684
T4	0.880 (0.270-2.869)	0.859
N stage		0.325
N0
N1	0.982 (0.562-1.715)	0.957
N2	1.030 (0.626-1.695)	0.922
N3	1.691 (1.081-2.645)	0.053

CI: Confidence interval; HR: Hazard ratio.

Prognostic nomogram development and validation

Based on these prognostic factors, a nomogram was established to predict the OS of patients with GC (Figure 3A). ROC analysis showed that the area under the curves of the nomogram in the training set for 1 year, 3 years, and 5 years were 0.721, 0.720, and 0.722 (Figure 3B), and 0.760, 0.743, and 0.732, respectively, in the validation set (Figure 3C), demonstrating good discrimination in predicting the OS of GC patients. Furthermore, we compared the discrimination between the nomogram and each independent prognostic factor, and the results indicated that the discrimination of the nomogram was better than that of all independent prognostic factors at 1 year, 3 years, and 5 years (Figure 3D-I). Moreover, the calibration curves of the nomogram for the probability of 1-year, 3-year, and 5-year OS showed strong agreement between nomogram-predicted OS and the actual outcome in the training set (Figure 3J-L) and validation set (Figure 3M-O).

Figure 3 Construction and validation of the prognostic nomogram. A: A prognostic nomogram for predicting the overall survival (OS) of gastric cancer patients at 1 year, 3 years, and 5 years; B: Time-dependent receiver operating characteristic curve analysis of the prognostic nomogram in the training set at 1 year, 3 years, and 5 years; C: Time-dependent receiver operating characteristic curve analysis of the prognostic nomogram in the validation set at 1 year, 3 years, and 5 years; D: Area under the curve (AUC) comparison of the nomogram with all independent factors in the training set at 1 year; E: AUC comparison of the nomogram with all independent factors in the training set at 3 years; F: AUC comparison of the nomogram with all independent factors in the training set at 5 years; G: AUC comparison of the nomogram with all independent factors in the validation set at 1 year; H: AUC comparison of the nomogram with all independent factors in the validation set at 3 years; I: AUC comparison of the nomogram with all independent factors in the validation set at 5 years; J: Calibration curves of the nomogram for 1 year in the training set; K: Calibration curves of the nomogram for 3 years in the training set; L: Calibration curves of the nomogram for 5 years in the training set; M: Calibration curves of the nomogram for 1 year in the validation set; N: Calibration curves of the nomogram for 3 years in the validation set; O: Calibration curves of the nomogram for 5 years in the validation set. ^aP < 0.01, ^bP < 0.0001. FP: False positive; TP: True positive.

Integrated screening and GSEA analysis of adipose-related differential genes in GC

We conducted an analysis of differential gene expression in adipocytes, comparing the high-expression and low-expression groups using the GSE84437 dataset obtained from the GEO database. This analysis identified a total of 112 DEGs, comprising 98 upregulated genes and 14 downregulated genes. Subsequently, we employed the R programming language to create heatmaps and volcano plots, as depicted in Figure 4. To gain further insights into the biological functions associated with these DEGs, we conducted GO and pathway analyses. The GO analyses revealed that these genes were primarily associated with biological processes such as negative regulation of cell growth, regulation of cell growth, actomyosin structure organization, negative regulation of growth, extracellular matrix organization, extracellular structure organization, and external encapsulating structure organization.

Figure 4 Analysis and functional enrichment of differentially expressed genes in the high and low expression groups of adipocytes. A: Heatmap of differentially expressed genes (DEGs). Different colors represent the expression trends of genes in different tissues; B: Volcano plot of DEGs constructed using fold change values and P-adjust. Red dots represent upregulated DEGs, gray dots represent non-significant genes, and green dots represent downregulated DEGs; C: Bar plot of Gene Ontology enrichment analysis for DEGs; D: Bar plot of (KEGG) Encyclopedia of Genes and Genomes enrichment analysis for DEGs; E: Gene Set Enrichment Analysis enrichment analysis results for the high expression group of adipocytes; F: Gene Set Enrichment Analysis enrichment analysis results for the low expression group of adipocytes. BP: Biological process; CC: Cell compound; MF: Molecular function.

Regarding cellular components, these genes were predominantly enriched in collagen-containing extracellular matrix, contractile fiber, sarcomere, and myofibril. Furthermore, in terms of molecular function, they were mainly associated with collagen binding, extracellular matrix structural constituent, and integrin binding. Kyoto Encyclopedia of Genes and Genomes analysis identified associations with pathways such as vascular smooth muscle contraction, tight junction, leukocyte transendothelial migration, and regulation of actin cytoskeleton. Next, we performed GSEA analysis. The findings indicated that the cell cycle, DNA replication, proteasome, pyrimidine metabolism, and spliceosome were primarily enriched in the AL expression group of GC. On the other hand, the calcium signaling pathway, dilated cardiomyopathy, focal adhesion, hypertrophic cardiomyopathy, and vascular smooth muscle contraction were mainly enriched in the AH expression group of GC.

Identification and validation of GC adipocyte-related genes as diagnostic biomarkers through LASSO and SVM-RFE

Two different algorithms were employed to screen candidate diagnostic biomarkers. Utilizing the LASSO logistic regression algorithm, we identified 34 meaningful feature variables associated with GC from DEGs (Figure 5A). The SVM-RFE algorithm was then employed to classify 16 features from the entire set of DEGs, resulting in the identification of a subset of ten important features (Figure 5B). Ultimately, seven overlapping meaningful features, namely genes ADH1B, SRPX, SFRP1, PLAC9, CDC20, FABP4, and TF, were selected by both algorithms, as illustrated in the Venn diagram (Figure 5C). Subsequently, the accuracy of the aforementioned analysis results and the expression levels of the seven candidate diagnostic biomarkers were validated using the GSE54129 dataset and TCGA GC dataset as validation groups (Figure 5D-Q). The Kaplan-Meier Plotter database was utilized to identify candidate genes and assess their potential as prognostic biomarkers for GC (Figure 5R-X). The results indicated that ADH1B, SFRP1, PLAC9, and FABP4 exhibited significant diagnostic and prognostic value with P < 0.05.

Figure 5 Selection of candidate diagnostic and prognostic biomarkers. A: Logistic regression algorithm screened 34 diagnostic markers through the least absolute shrinkage and selection operator (LASSO); B: Support vector machine (SVM) recursive feature elimination algorithm identified 19 diagnostic markers; C: Venn diagram included seven genes; D-J: Expression levels of candidate diagnostic genes in GSE54129; K-Q: Expression levels of candidate diagnostic genes in TCGA; R-X: Kaplan-Meier survival analysis of candidate genes in gastric cancer. AUC: Area under the curve; CI: Confidence interval; HR: Hazard ratio.

Immunocyte infiltration analysis

Using the CIBERSORT algorithm, we initially calculated the proportions of immune cell infiltration between AL and AH samples (Figure 6A). Furthermore, we computed the correlations among 22 types of infiltrating immune cells (Figure 6B). The results indicated a significant increase in T cell CD4 memory resting, T cell CD4 memory activated, T cell gamma delta, monocytes, macrophages M0, macrophages M2, macrophages M1, mast cells resting, natural killer (NK) cells resting, and neutrophils in AL samples compared to AH samples (P < 0.05) (Figure 6C).

Figure 6 Assessment and visualization of immune cell infiltration between high and low expression of adipocytes in gastric cancer. A: Proportion of infiltrating immune cells in samples with high and low expression of adipocytes in gastric cancer; B: Violin plots showed differential expression of 22 infiltrating immune cells between high and low expression of adipocytes in gastric cancer; C: Heatmap represented the 22 immune cell infiltrations between samples with high and low expression of adipocytes in gastric cancer. Blue and red colors indicate positive and negative correlations, respectively. The darker the color, the stronger the correlation; D: Lollipop plot showed the correlation between ADH1B expression levels and the infiltration of 22 immune cells; E: Lollipop plot illustrated the correlation between SFRP1 expression levels and the infiltration of 22 immune cells; F: Lollipop plot displayed the correlation between PLAC9 expression levels and the infiltration of 22 immune cells; G: Lollipop plot presented the correlation between FABP4 expression levels and the infiltration of 22 immune cells. NK: Natural killer.

Correlation analysis between feature genes and infiltrating immune cells

We conducted a correlation analysis between the expression levels of ADH1B, SFRP1, PLAC9, and FABP4 and the infiltration of immune cells associated with GC adipose tissue. These genes showed significant correlations with various infiltrating immune cells (Figure 6D-G). ADH1B, SFRP1, PLAC9, and FABP4 exhibited positive correlations with B cells memory, T cell CD4 memory resting, mast cells resting, and monocytes (P < 0.001) and negative correlations with T cell CD4 memory activated, macrophages M0, NK cells activated, macrophages M1, and neutrophils (P < 0.001).

DISCUSSION

This study delved into the pivotal role of adipocytes in the tumor microenvironment and their impact on the progression of GC. The research revealed a correlation between high expression of adipocytes within tumors and worse prognosis and outcomes in GC patients. Through the clinical validation phase, we successfully identified GC patients with tumor-associated adipocyte infiltration, underscoring the significance of this factor in clinical outcomes. Remarkably, adipocyte infiltration did not exhibit associations with adverse clinical pathological features, such as tumor staging. Nevertheless, there appears to be a potential correlation with the age of the patients, a connection that necessitates further confirmation. This finding emphasized the unique role of adipocytes in the development of GC, providing a direction for future research.

In order to enhance the accuracy of clinical predictions, we developed a prognosis prediction model based on independent prognostic factors. The calibration curve of the model demonstrated a high consistency between the predicted survival probabilities and actual outcomes, while time-dependent ROC curves confirmed its excellent discriminative ability at 12 mo, 36 mo, and 60 mo. Compared to a single prognostic factor, this model exhibited superior performance, offering a more precise and personalized approach for clinical decision-making in GC patients. It holds promising prospects for improving patient prognosis.

The interaction between adipocytes found near, around the margins, or within tumors and cancer cells is defined as CAAs or tumor-infiltrating adipocytes[22-24]. These CAAs significantly influence the biological characteristics of tumors by promoting angiogenesis and triggering inflammation, playing a crucial role in regulating malignant biology. First, CAAs release various metabolic byproducts such as free fatty acids, lactate, ketones, and glutamine, serving as energy sources for tumor cell metabolism and membrane synthesis, thereby promoting the growth and proliferation of tumor cells[25]. This metabolic support not only provides the foundation for sustaining the energy needs of tumor cells but may also act as a driving force in the process of tumor progression.

Second, CAAs secrete a variety of hormones, cytokines, adipokines, and growth factors, stimulating the proliferation of tumor cells and promoting their survival, growth, and division, thereby driving tumor progression[26]. This interplay between cells generates a complex signaling network within the tumor microenvironment, exerting profound effects on the biological behavior of the tumor.

Furthermore, CAAs play a role in regulating the immune system. Inflammatory factors associated with obesity can affect immune cells such as monocytes, macrophages, and NK cells[27]. Under the influence of CAAs, these immune cells may exhibit altered characteristics, leading to immune suppression in the tumor microenvironment. It has been observed that the immune system changes induced by CAAs upregulate programmed death-ligand 1 on the surface of tumor cells, a key molecule in tumor immune escape[28-30]. This phenomenon emphasizes the far-reaching impact of CAAs on immune regulation in the tumor microenvironment.

Adipocytes themselves also play a significant role in the progression of GC. Experimental results demonstrated that fatty acids released by adipocytes can promote peritoneal metastasis of GC[31,32]. Additionally, adipocytes in omentum (abdominal fat tissue) enhance the migration of GC cells through the secretion of cytokines[33-34]. These findings enrich our understanding of the critical role of CAAs in promoting the malignant biology of GC, providing valuable directions for future research.

We compared the AL and AH groups to elucidate the potential mechanisms by which adipocytes influence GC. At the molecular level, we conducted comprehensive screening and analysis of genes associated with adipocytes. Through differential gene analysis and bioinformatics methods, we identified a set of genes related to adipocytes in GC. After dual validation using LASSO and SVM-RFE algorithms, we ultimately confirmed four genes, namely ADH1B, SFRP1, PLAC9, and FABP4. These genes not only possess significant diagnostic value but also demonstrate good predictive performance in prognosis.

The ADH1B gene is associated with an increased risk of GC, exhibiting the highest expression and a significant fold increase during adipocyte differentiation. It plays a crucial role in the morphology and function of adipocytes[35,36]. FABP4 is a cell-intrinsic lipid transport protein highly expressed and secreted by adipocytes, with significant importance in adipose tissue. It has been shown to regulate peroxisome proliferator-activated receptor gamma, a key transcription factor essential for adipogenesis[37]. Chen et al[38] found that decreased FABP4 expression was associated with poor prognosis in GC patients, and FABP4 inhibited GC metastasis by promoting the translocation of peroxisome proliferator-activated receptor gamma to the nucleus. Ganesan et al[39] reported that increased expression of FABP4 was associated with adverse prognosis in GC and positively correlated with immune infiltration, especially in M2 macrophages. Currently, there are few reports on FABP4 in GC, and its role and mechanism in GC remain incompletely understood, necessitating further exploration in the future. ADH1B, as an upstream regulatory factor of FABP4, may also play a crucial role in the same cellular pathways.

There is conflicting evidence regarding the role of SFRP1 in GC, as it has been reported to both promote and suppress tumor growth[40]. Studies have shown that in most GC tissues, the mRNA and protein expression levels of SFRP1 are relatively low[41]. Furthermore, patients with low expression of SFRP1 in GC tend to have a poorer prognosis compared to those with high expression. The Wnt/β-catenin signaling pathway is believed to be associated with the overexpression of SFRP1 in the invasive subset of human GC, significantly correlated with lymph node metastasis and decreased overall survival rates[42].

SFRP1 overexpression is associated with the activation of the transforming growth factor-beta signaling pathway, inducing cell proliferation, epithelial-mesenchymal transition, and invasion. SFRP1 is expressed in human adipocytes, and its constitutive ectopic expression promotes adipogenesis while inhibiting the Wnt/β-catenin signaling pathway[39]. The crosstalk between the Wnt and transforming growth factor-beta signaling pathways regulated by SFRP1 is essentially interconnected[42]. These data suggest that SFRP1 can either promote or inhibit tumorigenesis.

Furthermore, in our study, analysis of immune infiltration in GC with high expression of adipocytes revealed a downregulation of T cell CD4 memory resting, T cell CD4 memory activated, monocytes, macrophages M0, macrophages M1, mast cells resting, and neutrophils. ADH1B, SFRP1, PLAC9, and FABP4 showed positive correlations with B cells memory, T cell CD4 memory resting, mast cells resting, and monocytes and negative correlations with T cell CD4 memory activated, macrophages M0, NK cells activated, macrophages M1, and neutrophils. These findings boldly and reasonably support the role of ADH1B, SFRP1, PLAC9, and FABP4 in modulating immune cells in GC. Additionally, we observed that immune cells such as macrophages, lymphocytes, neutrophils, T cells, and others were suppressed in the tumor microenvironment of GC, suggesting that GC with high adipocyte expression may exhibit increased aggressiveness[43-46]. These results provide valuable clues for further investigation into the tumor immune microenvironment in GC.

Despite our best efforts, our study has some limitations. We analyzed publicly available cohorts and a single-center clinical cohort, but the staging of GC in the two cohorts was not standardized, introducing certain biases. The publicly available cohorts did not include complete information on patient medical history, limiting the scope of our study. The assessment of adipocytes in different ways in the two cohorts may introduce some bias; however, the evaluated adipocytes in both cases are adipocytes within tumor tissue, excluding those around the tumor. Nevertheless, this does not negate the significance of our findings, which still support the use of adipocyte expression as an indicator of tumor invasiveness and anticipated patient outcomes.

CONCLUSION

In summary, GC with high intratumoral adipocyte expression exhibited a more aggressive tumor biology and poorer patient prognosis. Furthermore, our work identified crucial genes that may play a role in prognosis assessment, disease monitoring, and potential therapeutic targets for GC. Our findings contribute to the initial understanding of the role of adipocytes in GC.