Overweight and obesity are identified by a high body mass index (BMI) and carry significant health risks due to associated comorbidities. Although epidemiologic data connect overweight/obesity with 13 cancer types, a better understanding of the molecular mechanisms underlying this correlation is needed to improve prevention and treatment strategies. In this study, we conducted a comprehensive analysis of molecular differences between overweight or obese patients and normal weight patients across 14 different cancer types from The Cancer Genome Atlas. Using the propensity score weighting algorithm to control for confounding factors, obesity-specific mutational features were identified, such as higher mutation burden in rectal cancer and biased mutational signatures in other cancers. Differentially expressed genes (DEG) in tumors from patients with overweight/obesity were predominantly upregulated and enriched in inflammatory and hormone-related pathways. These DEGs were significantly associated with survival rates in various cancer types, highlighting the impact of elevated body fat on gene expression profiles and clinical outcomes in patients with cancer. Interestingly, while high BMI seemed to have a negative impact on most cancer types, the normal weight–biased mutational and gene expression patterns indicated overweight/obesity may be beneficial in endometrial cancer, suggesting the presence of an “obesity paradox” in this context. Body fat also significantly impacted the tumor microenvironment by modulating immune cell infiltration, underscoring the importance of understanding the interplay between weight and immune response in cancer progression. Together, this study systematically elucidates the molecular differences corresponding to body weight in multiple cancer types, offering potentially critical insights for developing precision therapy for patients with cancer.

Significance:

Elucidation of the complex interplay between body weight and the molecular landscape of cancer could potentially guide tailored therapies and improve patient management amid the global obesity crisis.

Overweight or obesity, characterized by an abnormal or excessive accumulation of body fat, is linked to an increased risk of various types of cancer (1). Body mass index (BMI), calculated by dividing the body mass in kilograms by the square of height in meters (kg/m2), is a widely used anthropometric measure for estimating overall body fat and classifying individuals as having a normal weight, being overweight, or obese (2). A meta-analysis of 212 datasets showed strong correlations between BMI and cancers at various sites, including the esophagus, thyroid, colon, kidneys, endometrium, and gallbladder (3). In 2016, a report from the International Agency for Research on Cancer Working Group concluded that excessive body fat is likely causally related to the risk of 13 types of cancer (4). Among these cancer types, endometrial (corpus uteri) and esophageal cancers have the highest risks associated with obesity, with relative risks of 7.1 and 4.8, respectively. Meanwhile, the risks associated with obesity are significant, ranging from 1.5 to 1.8, for colon, gastric cardia, liver, gallbladder, pancreas, and kidney cancers. However, it is noteworthy to mention that other cancers have also been associated with obesity in various studies. For instance, bladder cancer survival has been linked to BMI (5), and obesity has been identified as a risk factor for intrahepatic cholangiocarcinoma (CHOL) progression (6). Obesity has also been associated with non–Hodgkin lymphoma (7) and Hodgkin lymphoma (8). These findings further emphasize the complex relationship between obesity and cancer risk.

Previous studies have proposed various mechanisms to explain the relationship between excess body weight and specific cancers with limited data (9). Adipose tissue, which is not only a reservoir that stores lipids as long-term energy but also an endocrine organ that secretes adipokines, cytokines, and chemokines, has been linked to tumor initiation and progression (10). Three biological pathways have been proposed to link obesity and the increased cancer risks: the sex hormone pathway (11), the insulin and IGF signaling pathway (12), and the adipokine pathway (13). In addition, obesity-induced chronic inflammation and metabolic abnormalities create a tumor microenvironment that is conducive to the growth of multiple cancer types, including gastrointestinal, breast, liver, and pancreatic cancer (13). Insulin resistance and hyperinsulinemia, which are often observed in obesity patients, stimulate tumor cell proliferation and promote the growth of colorectal, pancreatic, liver, breast, and endometrial cancers (14). It is also well established that DNA damage/repair pathways are further altered in obese patients (15). However, previous cohort analyses of overweight/obesity-affected cancers have not considered potentially confounding factors, such as patient age, gender, race, histologic subtype, and tumor stage, which may introduce significant bias to molecular features. A systematic evaluation of these proposed mechanisms in a large cohort of patients with cancer is still lacking.

In this study, we used data from The Cancer Genome Atlas (TCGA) to systematically examine the differences in molecular features between normal weight and overweight/obese cancers (16). TCGA provides a large collection of high-throughput molecular data with corresponding clinical information for multiple cancer types. Our analysis included data from 2,645 patients with BMI information across 14 cancer types. To minimize the effects of confounding factors such as age, gender, tumor stage, etc. on molecular features, we applied a propensity score weighting (PSW) algorithm (17) to explore the differences in mutational patterns, gene expression profiles, and tumor immune microenvironments between normal weight and overweight/obese tumors. Our findings highlight the overweight/obesity-biased features at both genomic, transcriptome, and epigenomic levels, which may provide insight into the development of precision medicine for patients with cancer with overweight or obesity.

Patient samples and data resources

Clinical information for each patient in TCGA was obtained from Genomic Data Commons (GDC) Data Portal (https://portal.gdc.cancer.gov/). A total of 2,645 patients across 14 tumor types containing BMI information were included in this analysis. These patients were categorized into normal weight (BMI = 18.5–24.9 kg/m2), overweight (BMI = 25–29.9 kg/m2), and obese (BMI ≥ 30 kg/m2). Patient clinical information, cancer subtype, and purity data were analyzed by TCGAbiolinks (v2.24.2; ref. 18). The function “TCGAquery_subtype” in the TCGAbiolinks package was applied to retrieve the subtypes for each tumor patient. BMI distributions of 14 cancer types were visualized as boxplots using the ggplot2 package (v3.3.6) in R (v3.6.0).

Confounding factors balancing analysis

As patients had various clinical factors, including age, gender, tumor stage, etc., which could affect molecular characteristics, a PSW algorithm was applied to balance these factors (17). Briefly, the propensity score was based on the “BMI type (normal, overweight, or obese)” using a logistic regression. Calculated scores were then applied to reweight samples using the matching weight scheme. This process included balanced checking steps that continuously revised the calculation until the standardized differences of all covariates were less than 10%, ensuring successful balancing of confounding factors between normal weight and overweight/obese patients. Next, molecular characteristics between the two balanced groups were compared using a linear regression model, which used BMI type as the sole independent variable and quantified relative fold-change and corresponding statistical significance. P values obtained from the linear regression test were adjusted using the Benjamini–Hochberg procedure. Significant features were determined for each cancer type and molecular data type based on P-value threshold of <0.05 and an FDR of <0.2. Overweight or obesity bias was determined according to the relative means in normal weight and overweight or obese patients.

Somatic mutation and copy-number variation analysis

Data on somatic mutations and copy-number variation (CNV) were downloaded from TCGA Pan Cancer Atlas (https://gdc.cancer.gov/about-data/publications/pancanatlas). Somatic mutation data were generated from TCGA exome data by the Multi-Center Mutation-Calling in Multiple Cancers (MC3) network (19) and collected by an ensemble of seven mutation-calling algorithms with scoring and artifact filtering. To reduce biases and increase accuracy, hypermutated samples, defined as samples with over 1,000 somatic mutations, were excluded from downstream analysis (277 samples). Only nonsilent mutations were retained for subsequent analysis. The tumor mutation burden (TMB) was defined as the frequency of nonsilent somatic mutations per Mb genome. Frequencies of mutation subtypes (SNVs, indels) were separately calculated. A PSW algorithm was applied to identify the TMBs and mutation subtypes that showed significant differences between normal weight and overweight/obese tumors at the cutoff of FDR = 0.2 and P value = 0.05.

To detect significantly mutated genes (SMG) enriched in overweight/obese patients, only the nonsilent mutations with ≥5% mutation frequency were analyzed in each cancer type. The PSW algorithm was applied to identify highly mutated genes that showed significant differences in mutation frequencies between normal weight and overweight/obese patients at the cutoff of FDR = 0.2 and permutation test P value = 0.05. Selected genes of interest were collected in lollipop plots using the maftools (v3.15) package (20). In addition, MutSig2CV (21) was applied to identify SMGs in tumors with normal weight and overweight/obesity. SMGs were mutated more often than expected by chance, given the inferred background mutation and mutational processes. Mutated genes with P value <0.001 in overweight/obese patients with tumors, but >0.05 in normal weight patients with tumor were identified as overweight/obesity-enriched genes. Conversely, mutated genes with P value < 0.001 in normal weight tumors but >0.05 in overweight/obese tumor patients were identified as normal weight–enriched genes.

Mutational signatures were analyzed using MutationalPatterns (v3.6.0; ref. 22), an R/Bioconductor package that allowed for easy characterization and visualization of mutational patterns. MutationalPatterns characterized 96 types of base substitutions and quantified the contributions of known mutational signatures based on combinations of base substitutions. Known mutational signatures were summarized as the COSMIC mutational signatures (v2; ref. 23). Differences in mutational signatures contribution between normal weight and overweight/obesity were identified using the PSW algorithm. Significant signatures were defined at the cutoff of FDR = 0.2 and P value = 0.05.

The masked CNV data were downloaded from the GDC Data Portal with the TCGAbiolinks R package (18). Gene-level CNV data, analyzed by GISTIC2 software (24), were downloaded from UCSC (https://xenabrowser.net/datapages/). The number of CNVs was summarized, and the PSW algorithm was applied to compare differences of CNV between normal weight and overweight/obese patients. For focal amplifications/deletions, the PSW algorithm was applied to identify overweight/obesity-biased somatic copy-number alternation (SCNA) with FDR <0.1 and P value <0.05. Overweight or obesity bias was determined according to relative levels in overweight/obesity and normal weight patients and their gain and loss natures. Genes in significant focal regions were extracted for enriched pathways analysis using the Enrichr package (v3.0; ref. 25).

Gene expression and methylation analysis

Normalized mRNA expression data were obtained from TCGA Pan Cancer Atlas (https://gdc.cancer.gov/about-data/publications/pancanatlas). All samples were categorized into normal weight, overweight, and obese based on their BMI values. The PSW algorithm was applied to the log2-transformed RSEM to identify genes [differentially expressed genes (DEG)] that showed significant differences between normal weight and overweight/obese patients with a permutation test of P value < 0.01. To identify BMI-biased pathways, functional enrichment analysis of DEGs was performed using Enrichr (26). Significant Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways with P value <0.05 were selected.

DNA methylation 450K data were obtained from UCSC Xena, which compiled methylation calls from the GDC database (27). If a single gene was measured by multiple probes, the probe that was most negatively correlated with the corresponding gene expression was preserved. The PSW algorithm was applied to selected methylation data to identify the hypermethylated/hypomethylated genes between normal weight and overweight/obese patients at the cutoff of P value = 0.01. Enrichr (26) was applied to detect the significant pathways with P value <0.05.

Analysis of immune features

Thorsson and colleagues (28) had performed an immunogenomic analysis on all cancer types in TCGA and identified six immune subtypes (IS; C1–C6) by scoring and clustering 160 immune expression signatures. The IS data and immune feature scores were obtained from Thorsson and colleagues (https://gdc.cancer.gov/about-data/publications/panimmune) and used for detecting differences between normal weight and overweight/obese tumors. On the basis of their identification of ISs with large-scale data, the distributions of ISs among normal weight, overweight, and obese patients were displayed as the relative percentages. The PSW algorithm was applied to identify significantly different immune cell infiltration and immune scores at the cutoff of FDR = 0.2 and P value = 0.05.

Survival analysis

The “Survival” R package (v3.3-1) was used to build a standard survival object using the “Surv” function. To analyze whether the gene expression levels could impact a patient's survival, a survival curve was generated for the censored data using the Kaplan–Meier method. Patients were then grouped into high and low expression groups based on whether their expression levels were higher or lower than their respective means. Next, survival curves were constructed for the two comparable groups. The log-rank test was used to evaluate whether patients with varying gene expressions or mutation frequencies had significantly different survival durations (P value <0.05).

Statistical analysis

Statistical analyses were conducted using R Studio (v1.2.1335) and R programming (v3.6.1). The R packages and statistical approaches utilized in each step are described in the corresponding sections above.

Data availability

TCGA data were acquired from TCGA Data Portal (https://portal.gdc.cancer.gov/) and the GDC PanImmune Data Portal (https://gdc.cancer.gov/about-data/publications/panimmune). All other data can be found within the article, the Supplementary Data, or are available from the author upon reasonable request.

Code availability

Codes were implemented in R3.6.1 and are deposited at https://github.com/huang1990/ObeseCancer.

Balancing confounding factors to decode the impact of BMI on tumorigenesis

To examine the genomic profiles of patients with cancer with an excessive body weight, we gathered clinical and molecular data from TCGA database. Our study focused on 14 types of cancer, with a total sample size of 2,645 patients with available BMI information (Supplementary Table S1). These cancers include bladder urothelial carcinoma (BLCA), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), CHOL, colon adenocarcinoma (COAD), lymphoid neoplasm diffuse large B-cell lymphoma (DLBC), esophageal carcinoma (ESCA), kidney renal papillary cell carcinoma (KIRP), liver hepatocellular carcinoma (LIHC), rectum adenocarcinoma (READ), skin cutaneous melanoma (SKCM), thymoma (THYM), uterine corpus endometrial carcinoma (UCEC), uterine carcinosarcoma (UCS), and uveal melanoma (UVM; Supplementary Fig. S1A). Unfortunately, several cancers traditionally linked to obesity, including breast and pancreatic cancers, were not part of our analysis due to the lack of BMI data for these types in TCGA.

The patients with cancer were categorized into normal weight (BMI = 18.5–24.9 kg/m2), overweight (BMI = 25–29.9 kg/m2), and obese (BMI ≥ 30 kg/m2) groups based on their BMI values calculated from their weight and height data. The distribution of BMI showed that the median BMI values range from 24.5 to 32.33, indicating that the primary tumors in TCGA data set are slightly skewed toward overweight and obesity. COAD, ESCA, KIRP, LIHC, READ, and UCEC are cancer types that are known to be associated with high BMI, while the remaining eight are not. Of the 14 cancer types, 13 have more than 50% of overweight or obese patients (except ESCA, which has 46.7% overweight or obese patients), with UCEC having the highest mean BMI and the largest fraction of overweight or obese patients (Fig. 1A).

Figure 1.

BMI distribution and PSW algorithm overview in this study. A, Distribution of normal weight, overweight, and obese BMI groups in 14 cancer types. B, Overview of the PSW algorithm used to balance confounding factors and evaluate the body weight–associated molecular features, including somatic mutations, CNVs, gene expression, gene methylation, and immune features. BLCA (n = 353), bladder urothelial carcinoma; CESC (n = 250), cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL (n = 40), cholangiocarcinoma; COAD (n = 232), colon adenocarcinoma; DLBC (n = 46), lymphoid neoplasm diffuse large B-cell lymphoma; ESCA (n = 167), esophageal carcinoma; KIRP (n = 208), kidney renal papillary cell carcinoma; LIHC (n = 320), liver hepatocellular carcinoma, READ (n = 70), rectum adenocarcinoma; SKCM (n = 245), skin cutaneous melanoma; THYM (n = 97), thymoma; UCEC (n = 513), uterine corpus endometrial carcinoma; UCS (n = 52), uterine carcinosarcoma; and UVM (n = 52), uveal melanoma.

Figure 1.

BMI distribution and PSW algorithm overview in this study. A, Distribution of normal weight, overweight, and obese BMI groups in 14 cancer types. B, Overview of the PSW algorithm used to balance confounding factors and evaluate the body weight–associated molecular features, including somatic mutations, CNVs, gene expression, gene methylation, and immune features. BLCA (n = 353), bladder urothelial carcinoma; CESC (n = 250), cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL (n = 40), cholangiocarcinoma; COAD (n = 232), colon adenocarcinoma; DLBC (n = 46), lymphoid neoplasm diffuse large B-cell lymphoma; ESCA (n = 167), esophageal carcinoma; KIRP (n = 208), kidney renal papillary cell carcinoma; LIHC (n = 320), liver hepatocellular carcinoma, READ (n = 70), rectum adenocarcinoma; SKCM (n = 245), skin cutaneous melanoma; THYM (n = 97), thymoma; UCEC (n = 513), uterine corpus endometrial carcinoma; UCS (n = 52), uterine carcinosarcoma; and UVM (n = 52), uveal melanoma.

Close modal

To minimize the impact of other factors such as age, race, and tumor stage (Supplementary Fig. S1B and S1C), we employed the PSW approach (17). This statistical algorithm is crucial in reweighting the confounding effects. The propensity score for each patient is the conditional probability of being exposed to a set of covariates. Samples with the same propensity score have an equal distribution of the measured confounders. Thus, after reweighting the propensity scores, we were able to balance the confounders and evaluate the body weight–associated molecular features (Fig. 1B).

Overweight/obesity-biased mutational landscapes

The accumulation of DNA mutations in somatic cells can play a key role in cancer development (29). In this study, we analyzed the relationship between tumor somatic mutations and BMI. The TMB, an estimate of the number of somatic mutations that occur in a tumor, was calculated as the number of nonsynonymous mutations per Mb. To avoid biases that may result from confounding factors, we employed the PSW approach and conducted weighted regression to identify univariate BMI associations with TMB.

We found that the TMB had a positive and significant correlation with BMI in READ (weighted regression, R = 0.297 and P value = 0.018; Fig. 2A) and UVM (weighted regression, R = 0.57 and P value = 7.89e-06; Supplementary Fig. S2A), implying that in these two cancer types there is a statistical association between excess body fat and an accelerated accumulation of mutations. However, it is important to note that while there is a statistically significant positive correlation between BMI and mutation rate, the effect size is small, and the correlation may be influenced by outliers. When we stratified the mutations into SNVs and small insertion and deletions (indels), we observed that BMI was significantly correlated with SNVs but not with indels in these two cancer types (Fig. 2B and C; Supplementary Fig. S2B and S2C). Upon reevaluation of the association after excluding an outlier, the significance of the correlation for UVM did not persist. It is also important to note that the sample size for UVM in our study was relatively small (n = 52), which could potentially impact the robustness of our statistical analysis. Although no significant correlation between BMI and overall TMB (SNVs plus indels) was observed in other cancer types, compared with normal weight patients, those who were obese had a significantly higher TMB of indels in ESCA (weighted regression, R = 0.15 and P value = 0.036) and deletions in CHOL (weighted regression, R = 0.47 and P value = 0.049; Supplementary Fig. S2D and S2E).

Figure 2.

Association of somatic mutation burden and mutational signatures with BMI in different cancers. A–C, Significant association between BMI and somatic mutation burden in patients with READ. Scatter plots were fitted with weighted regression of BMI and overall somatic mutation burden (TMB; A), SNVs (B), indels (C), respectively. The R value and P value were calculated using the weighted linear regression model. D and E, Biased somatic mutational signatures between normal weight and overweight (D)/obese patients (E). The sizes of circles indicate different P values, with filled-in and open circles indicating statistically significant [weighted linear regression test, FDR (controlled using the Benjamini–Hochberg procedure) <0.2] and no significant differences, respectively. The proposed signature aetiologies are indicated. The P values were calculated by the weighted linear regression model and adjusted by Benjamini and Hochberg correction.

Figure 2.

Association of somatic mutation burden and mutational signatures with BMI in different cancers. A–C, Significant association between BMI and somatic mutation burden in patients with READ. Scatter plots were fitted with weighted regression of BMI and overall somatic mutation burden (TMB; A), SNVs (B), indels (C), respectively. The R value and P value were calculated using the weighted linear regression model. D and E, Biased somatic mutational signatures between normal weight and overweight (D)/obese patients (E). The sizes of circles indicate different P values, with filled-in and open circles indicating statistically significant [weighted linear regression test, FDR (controlled using the Benjamini–Hochberg procedure) <0.2] and no significant differences, respectively. The proposed signature aetiologies are indicated. The P values were calculated by the weighted linear regression model and adjusted by Benjamini and Hochberg correction.

Close modal

Distinct patterns of somatic mutations, arising from both intrinsic cellular processes and external factors, leave unique mutational signatures on the genome that reveal the pathophysiology of cancer (23). These mutational signatures offer valuable insights into the fundamental molecular mechanisms and mutation-causing processes involved. We used MutationalPatterns (22) to characterize the patterns and quantify the contributions of COSMIC mutational signatures. To identify mutational signatures related to overweight/obesity, we compared the contributions of mutational signatures from overweight/obese patients and normal weight patients. In READ, Signature 6 was more prevalent in the overweight group, suggesting a higher number of mutations associated with mismatch repair (MMR) deficiency in this group. Conversely, Signature 20 was more prevalent in the normal weight group, indicating a higher number of mutations associated with MMR alterations in these individuals (Fig. 2D). These two signatures are among the four (the other being Signatures 15 and 26) that are associated with defective DNA MMR and are linked to a high number of small indels. This aligns with the fact that READ is a microsatellite instability-high cancer (30). The observation of defective MMR in both normal weight and overweight groups suggests that indels may be a common mutational driver in READ, not affected by excess body fat, as also indicated by the nonsignificant correlation between indels and BMI in Fig. 2C. READ showed overweight-biased Signature 24 (Fig. 2D), found in cancers associated with aflatoxin B1 exposure. Signature 24 exhibits a strong transcriptional strand bias for C>A mutations, leading to DNA damage and disruption of DNA repair processes. In addition, obese READ showed a higher contribution of Signature13 (Fig. 2E), believed to be caused by the activity of AID/APOBEC family of cytidine deaminases and their contribution to DNA damage. This may indicate that DNA damage and disruption of DNA repair, rather than defective MMR, were associated with excess body fat, and were more pronounced in overweight and obese READ.

Although DLBC has not been reported as an obesity-associated cancer, small cohort studies (31–33) have shown that obesity increased the relative risk of DLBC. We discovered that DLBC displayed three obesity-biased Signatures: Signature 4 (caused by tobacco mutagens and exhibits a transcriptional strand bias for C>A mutations), Signature 9 (attributed to polymerase η and implicated in the activity of AID during somatic hypermutation) and Signature 13 (attributed to the activity of the AID/APOBEC family of cytidine deaminases, causing predominantly C>G mutations). These mutational signatures suggest that there is a statistical association between obesity and an increased occurrence of specific types of mutations in DLBC. In ESCA, in line with the results that indels were positively correlated with BMI (Supplementary Fig. S2D), Signature 6 (defective MMR) was significantly biased in the obese group (Fig. 2E), implying the defective DNA MMR may be common in obese esophagus cancer. Interestingly, UCEC, which has a strong association with obesity, exhibited multiple normal weight–biased (rather than overweight-biased or obesity-biased) mutational signatures (Signature 2, 3, 4, and 13) when compared to both overweight and obese groups. These signatures were linked to APOBEC activity, defective MMR, DNA double-strand break repair, and tobacco mutagens. This finding suggests that normal weight–biased signatures may play a role in UCEC development. In the case of CHOL, normal weight–biased mutational signatures (Signatures 1 and 10) were also observed. However, due to the small sample size (n = 40) of CHOL, it is difficult to draw definitive conclusions about the significance of these signatures in this context.

Genes exhibiting a high mutation frequency are potential candidates for cancer driver genes. We investigated whether gene mutations are biased in overweight or obese cancers in these cancer types. Only the nonsilent mutations with ≥5% mutation frequency were analyzed in each cancer type. The PSW algorithm was applied to identify the frequently mutated genes. The number of selected genes ranged from 4 in THYM to 1,354 in SKCM (see the experimental procedures). Overweight or obesity-biased mutant genes were identified when recalibrated mutation frequencies in overweight or obese patients were higher than those in normal weight patients. Five cancer types (UCEC, READ, ESCA, DLBC, and BLCA) passed the test, with ESCA having the highest number of biased genes (30 normal weight–biased, 1 overweight-biased, and 60 obesity-biased mutated genes; Fig. 3A; Supplementary Fig. S3A–S3C). For overweight-biased genes, we found nine genes (ADAMTS16, BCL9L, FLRT2, FRAS1, HYDIN, NPR1, SOX11, SYNE1, TTN) in READ and one gene (LRRC4C) in ESCA with an FDR (controlled using the Benjamini–Hochberg procedure) <0.2. Interestingly, four of the overweight-biased mutated genes in READ (SOX11, SYNE1, ADAMTS16, and BCL9L) have been reported to have strong associations with tumor development or response to cancer treatment (34–37). For example, one study showed that ADAMTS16 mutations increase sensitivity to platinum-based chemotherapy in ovarian cancer cells (37). At FDR <0.2, there were 60 obesity-biased mutated genes in ESCA, 10 in UCEC, six in DLBC, and one in BLCA (Fig. 3A and B; Supplementary Fig. S3A–S3C). Ten of the top 20 obesity-biased mutated genes in ESCA (Fig. 3B)—SYNE1, FAT4, TENM4, RYR3, CDH11, HYDIN, LAMA1, ABCC9, HUWE1, and NBEA—have been shown to be strongly associated with tumorigenesis, tumor metastasis, and treatment response (38, 39). For instance, ABCC9 is a component of ATP-binding cassette transporters that play critical roles in transporting substances such as sterol, metal ions, proteins, and molecules of chemotherapeutic drugs across membranes (38). We also found that SYNE1 had more mutations in overweight READ and obese ESCA, and that these mutations enhanced the response to immune checkpoint blockade therapy in renal cell carcinoma (39). FAT4 is a gene that encodes a protein that is involved in the Hippo pathway, which regulates cell growth and organ size. Rare variants in FAT4 were associated with obesity in humans (40). Interestingly, UCEC showed that normal weight–biased genes (TAF1, PCLO, DNAH7, DYRK1A, CEP290, HSPG2, AHNAK, REV3L, NFE2L2, and FLNC) were of high frequency (Fig. 3C). The majority of these biased genes have not been reported as SMGs in previous TCGA large cohort studies (41, 42), implying these biased genes may be due to the environment of excessive body fat.

Figure 3.

Identification of BMI-biased mutated genes. A, Distribution of normal weight-, overweight-, and obesity-biased mutated genes across five cancer types. These genes had significantly higher mutation frequencies in normal weight, overweight, or obese patients [weighted linear regression, P value <0.05 and FDR (controlled using the Benjamini–Hochberg procedure) <0.2]. B, Top 20 obesity-biased mutated genes in ESCA (weighted linear regression, P value <0.05 and FDR (controlled using the Benjamini–Hochberg procedure) <0.2). The colors in the oncoplot represent different types of mutations annotated at the bottom. The top bar plot shows the mutation frequency in individual patients. The right bar plot shows the recalibrated mutation frequencies after PSW. C, Recalibrated mutation frequencies of normal weight-, overweight- (none), and obesity-biased mutated genes in UCEC [weighted linear regression, P value <0.05 and FDR (controlled using the Benjamini–Hochberg procedure) <0.1]. D, SMGs detected by MutSig2CV in UCEC. Purple font genes are enriched in overweight or obese patients. Wheat font genes are enriched in normal weight patients. Light-green font genes are enriched in both normal weight and overweight/obese groups. P value <0.001 after correction for multiple hypothesis testing. E, Enriched pathways in UCEC for the SMGs enriched in overweight or obese groups. Fisher exact test, P value <0.05.

Figure 3.

Identification of BMI-biased mutated genes. A, Distribution of normal weight-, overweight-, and obesity-biased mutated genes across five cancer types. These genes had significantly higher mutation frequencies in normal weight, overweight, or obese patients [weighted linear regression, P value <0.05 and FDR (controlled using the Benjamini–Hochberg procedure) <0.2]. B, Top 20 obesity-biased mutated genes in ESCA (weighted linear regression, P value <0.05 and FDR (controlled using the Benjamini–Hochberg procedure) <0.2). The colors in the oncoplot represent different types of mutations annotated at the bottom. The top bar plot shows the mutation frequency in individual patients. The right bar plot shows the recalibrated mutation frequencies after PSW. C, Recalibrated mutation frequencies of normal weight-, overweight- (none), and obesity-biased mutated genes in UCEC [weighted linear regression, P value <0.05 and FDR (controlled using the Benjamini–Hochberg procedure) <0.1]. D, SMGs detected by MutSig2CV in UCEC. Purple font genes are enriched in overweight or obese patients. Wheat font genes are enriched in normal weight patients. Light-green font genes are enriched in both normal weight and overweight/obese groups. P value <0.001 after correction for multiple hypothesis testing. E, Enriched pathways in UCEC for the SMGs enriched in overweight or obese groups. Fisher exact test, P value <0.05.

Close modal

Next, we used MutSig2CV (21) algorithms to detect SMGs for each individual BMI group. MutSig2CV aims to identify genes mutated more often than expected by chance, given inferred background mutation rate, localized hotspots, and conservation across vertebrates. While common cancer driver genes such as TP53, APC, KRAS, ARID1A, and PIK3CA were enriched in most of the cancers regardless of BMI status, the range of SMGs in overweight and obese patients was from 1 in UVM to 137 in UCEC (Supplementary Fig. S3D–S3P). In UCEC, the SMGs in overweight and obese patients (137 genes) were much more numerous than the SMGs in normal weight patients (three genes; Fig. 3D). The SMGs in UCEC were enriched in pathways such as hormone, mTOR, JAK-STAT, insulin signaling pathways (Fig. 3E), which are known mechanisms for cancer due to obesity (43). One variant on FTO has the potential to disrupt ARID5B, which is one of the top SMGs. Once disrupted, ARID5B can activate IRX3 and IRX5, which are involved in the conversion of fat cells to white fat that stores lipids rather than brown fat (44). There are only three genes (CYP4F22, SMC3, TMEM92) enriched in normal-weight UCEC. CYP4F22 is a gene responsible for the metabolism of long-chain fatty acids, such as omega-3 and omega-6 fatty acids, which play important roles in the regulation of inflammation, lipid metabolism, and other physiologic processes (45). SMC3 is a subunit of the cohesion complex required for organizing the three-dimensional structure of chromosomes (46). Alterations in cohesin function can lead to chromosomal instability and genomic instability (47), which are common features of cancer cells. In LIHC, ADAM12 deficiency is involved in obesity induced by a high-fat diet in mice (48) and is significantly enriched in overweight and obese patients (Supplementary Fig. S3K). At a pan-cancer level, we identified the tumor suppressor gene TXNIP, whose expression is negatively regulated by insulin-like growth factor 1 (IGF1; ref. 49) and is involved in glucose metabolism (50). The analysis of SMGs suggests that excessive body fat may contribute to the modulation of functionally important genes across various cancers.

Overweight/obesity-biased CNV

Somatic CNV constitutes another source of mutations. We explored the possibility of CNV bias among different weight groups. To do so, we initially applied a weighted linear regression model to examine the relationship between the number of CNVs and BMI. In THYM, the numbers of total CNVs (weighted regression model, R = 0.28, P value = 0.016), particularly the numbers of CNV loss (weighted regression model, R = 0.34, P value = 0.0036), indicated a positive relationship between CNV numbers and BMI (Supplementary Fig. S4A and S4B). The numbers of CNV gain were also positively correlated with BMI in obese CHOL (weighted regression model, R = 0.53, P value = 0.022; Supplementary Fig. S4C). However, the small sample size in CHOL limits the conclusiveness of this association. Although the total numbers of CNVs between normal weight and overweight/obesity were not significantly correlated with BMI in UCEC, normal weight tumors had a relatively higher number of CNVs (with a mean total of CNVs 128 in normal weight vs. 98 CNVs in overweight/obesity). Consistent with SMC3 being a SMG enriched in normal weight of UCEC, this supports a higher level of genomic instability in normal weight but not in overweight/obese UCEC.

Furthermore, we focused on the SCNAs identified by GISTIC2 (24), a computational tool for identifying recurrent CNVs in cancer genomes and predicting the potential functional significance of these CNVs in tumorigenesis. We characterized the overweight/obesity-biased focal SCNAs in each cancer type. First, we identified overweight-biased focal regions in these cancer types. At FDR (controlled using the Benjamini–Hochberg procedure) <0.1, we found that ESCA and LIHC had 12 and 19 focal CNVs biased in the overweight group. Specifically, ESCA had eight focal amplifications and four deletions, involving 132 amplified and 64 deleted genes, respectively (Supplementary Fig. S4D and S4E). LIHC had 19 focal amplifications, including 482 genes (Supplementary Fig. S4F and S4G). These overweight-biased amplified genes were enriched in pathways such as nucleotide excision, base excision repair, and Glycine metabolism (Supplementary Fig. S4H). Glycine metabolism has a strong link with obesity and insulin resistance (51). Three cancer types (ESCA, UCEC, and CHOL) showed significant obesity-biased amplifications/deletions (Fig. 4AD; Supplementary Fig. S4I and S4J) compared with normal weight tumors. These focal amplifications/deletions were scattered across multiple chromosomes. In ESCA, obesity-biased deletions were primarily located on chromosomes 9, 14, 16, 18, and X (Fig. 4A), while amplifications were concentrated on chromosomes 2, 7, 18, and 20 (Fig. 4B). The deletion of 16q13 contains a series of metallothioneins (MT), which play key roles in metal homeostasis and protection against heavy metal toxicity, DNA damage, and oxidative stress (52). The genes in obesity-biased amplifications were significantly enriched in tricarboxylic acid (TCA) cycle and SNARE interactions in vesicular transport pathways (Fig. 4E). Both pathways are involved in the regulation of cancer cell invasion, chemoresistance, and cancer cell biogenesis (53). The deleted MTs and amplified SNARE genes provide further evidence that obesity may create a microenvironment that can promote tumorigenesis and tumor cell migration.

Figure 4.

Characterization of overweight/obesity-biased somatic CNVs. A–D, Focal CNVs biased in obesity across multiple chromosomes in ESCA and UCEC. The focal deletions (A) and amplifications (B) in ESCA, and the focal deletions (C) and amplifications (D) in UCEC are shown. The hotpink peaks represent obesity-biased focal somatic CNVs, and wheat peaks indicate the normal weight–biased somatic CNVs. FDR = 0.1 is indicated by the vertical green dash lines. The significant somatic CNVs that harbor important genes are annotated. E, Enriched KEGG pathways for obesity-biased amplified genes in ESCA and UCEC. The rich ratios represent the fraction of obesity-biased amplified genes participating in that particular KEGG pathway. Fisher exact test, P value <0.05. F, GO molecular function analysis of obesity-biased and normal weight–biased amplified and deleted genes in UCEC. Nor_am, normal weight–biased amplifications; Nor_del, normal weight–biased deletions; Ob_am, obesity-biased amplifications; Ob_del, obesity-biased deletions.

Figure 4.

Characterization of overweight/obesity-biased somatic CNVs. A–D, Focal CNVs biased in obesity across multiple chromosomes in ESCA and UCEC. The focal deletions (A) and amplifications (B) in ESCA, and the focal deletions (C) and amplifications (D) in UCEC are shown. The hotpink peaks represent obesity-biased focal somatic CNVs, and wheat peaks indicate the normal weight–biased somatic CNVs. FDR = 0.1 is indicated by the vertical green dash lines. The significant somatic CNVs that harbor important genes are annotated. E, Enriched KEGG pathways for obesity-biased amplified genes in ESCA and UCEC. The rich ratios represent the fraction of obesity-biased amplified genes participating in that particular KEGG pathway. Fisher exact test, P value <0.05. F, GO molecular function analysis of obesity-biased and normal weight–biased amplified and deleted genes in UCEC. Nor_am, normal weight–biased amplifications; Nor_del, normal weight–biased deletions; Ob_am, obesity-biased amplifications; Ob_del, obesity-biased deletions.

Close modal

In UCEC, two obesity-biased focal amplifications were identified on chromosomes 11 and 20 (Fig. 4C and D). Both obesity-biased amplified genes and normal weight–biased deleted genes were found to be associated with growth factors and hormone related activities, as revealed by GO molecular function analysis (Fig. 4F). Moreover, both obesity-biased deletions and normal weight–biased amplifications were found to encompass immune-related genes. These results also highlight the known mechanisms of obesity's contribution to cancer (54).

Overweight/obesity-biased expression profiles

To determine the gene expression profile associated with overweight/obesity, we first compared gene expressions between normal weight and overweight/obese tumors after weighting confounding factors. The numbers of DEGs (FDR <0.2) between overweight/obese and normal weight cancers ranged from 17 in THYM to 3,929 in ESCA (Fig. 5AD). Overall, we identified more DEGs between normal weight and obese tumors than between normal weight and overweight, implying overweight's intermediate status. Among all 14 cancer types, the most significant numbers of DEGs were detected in ESCA (853 and 3,929 in overweight vs. normal weight and obese vs. normal weight comparison, respectively) and UCEC (408 and 1,711, respectively). The heat map plot of the top DEGs in ESCA shows different gene expression patterns between normal weight tumors and overweight/obese tumors (Supplementary Fig. S5A). Notably, ACOT11 encodes a member of the acyl-CoA thioesterase family, which catalyzes the conversion of activated fatty acids to the corresponding non-esterified fatty acid and coenzyme A, regulated by ambient temperature in mouse brown adipose tissue (55). It can overproduce free fatty acids, which can provoke insulin resistance that is associated with inflammation and endoplasmic reticulum stress (56). Overall, for both overweight and obese tumors, upregulated DEGs dominated the enriched pathways. Several inflammatory and hormone-related pathways, such as TNF signaling pathway, mTOR signaling pathway, NFκB signaling pathway, insulin signaling pathways, and steroid/thyroid hormones biosynthesis were significantly enriched by the upregulated DEGs (Fig. 5B and D). The relative fewer number of enriched pathways in overweight than obese tumors is consistent with a fewer number of DEGs in overweight tumors. And the enriched pathways in overweight tumors were similar to obese tumors, implying an intermediate status of overweight. Therefore, we focused on obese tumor versus normal weight tumors. In individual cancer types, the DEGs were enriched in several different pathways, indicating that excessive body fat may affect the tumor microenvironment through multiple routes simultaneously.

Figure 5.

Overweight/obesity-biased gene expression profiles. A and C, Number of DEGs in overweight and obese versus normal weight tumors across 14 cancer types. B and D, Enriched KEGG pathways of upregulated DEGs in overweight (B) and obese (D) tumors. Wheat color indicates downregulated genes, while skyblue and hotpink colors represent upregulated genes in overweight and obese tumors, respectively. E and F, Survival analysis of patients with upregulated DEGs in overweight (E) and obese (F) UCEC tumors. Patients with upregulated genes exhibited better survival in UCEC (log-rank test, P value <0.05).

Figure 5.

Overweight/obesity-biased gene expression profiles. A and C, Number of DEGs in overweight and obese versus normal weight tumors across 14 cancer types. B and D, Enriched KEGG pathways of upregulated DEGs in overweight (B) and obese (D) tumors. Wheat color indicates downregulated genes, while skyblue and hotpink colors represent upregulated genes in overweight and obese tumors, respectively. E and F, Survival analysis of patients with upregulated DEGs in overweight (E) and obese (F) UCEC tumors. Patients with upregulated genes exhibited better survival in UCEC (log-rank test, P value <0.05).

Close modal

To investigate the clinical impacts of these DEGs, we tried to identify the genes that were significantly associated with survival time. We found that more than 10% of both overweight and obesity DEGs were significantly associated with survival rate (the log-rank test, P < 0.05) in CESC, KIRP, LIHC, and UCEC (Supplementary Fig. S5B). Although only a small fraction of DEGs were associated with the survival rate in ESCA, multiple obesity-biased genes, which play critical roles in promoting tumor growth and leading to a shorter survival time (Supplementary Fig. S5C–S5E), were discovered. For example, FOS, a gene that encodes the nuclear oncoprotein c-Fos that causes the downregulation of tumor-suppressor genes and leads to the invasion of cancer cells (57), BRCC3, which promotes tumorigenesis by activating the NFκB signaling pathway in bladder cancer (58), and IL13RA, which interacts with IL13 to promote tumor cell growth and suppress tumor-directed immune surveillance (59), and in addition, IL13RA, which interacts with IL13 to promote tumor cell growth and suppress tumor-directed immune surveillance (59), the high expression of which were all associated with a poor survival in ESCA. UCEC had the highest fraction of DEGs that were significantly associated with survival rate. Interestingly, the majority of the high expression in upregulated genes and low expression in downregulated genes predicted better survival for both overweight and obese tumors in UCEC (Supplementary Table S2). We further checked how the DEGs are associated with survival. We found that, in both overweight and obese groups, individuals with upregulated genes have significantly better survival than individuals with downregulated genes (Fig. 5E and F). The upregulated genes in obese and overweight tumors have better survival in UCEC is quite different from other cancer types. We investigated why that was. By re-examining the DEG enrichment pathways (Fig. 5D), we observed that in UCEC, the downregulated DEGs, rather than the upregulated DEGs ones, were enriched in pathways such as cell cycle, MMR, DNA replication, and citrate cycle (TCA cycle). This pattern differed from other cancer types. That suggests that genes involved in cell cycle, MMR, DNA replication pathways were highly active in the normal-weight endometrial tumors, potentially promoting tumor proliferation. BLCA exhibited a similar trend, despite bladder cancer not being one of the 13 obese-associated cancer types.

DNA methylation plays an essential role in the regulation of gene expression, genomic imprinting, X-chromosome inactivation, and the maintenance of genomic stability. In general, methylation of CpG islands in promoter regions is associated with transcriptional repression, leading to gene silencing (60). In this study, we also investigated the differences in DNA methylation levels associated with overweight/obesity. We found that the number of genes with hypermethylation (an increase in the epigenetic methylation of cytosine in DNA) or hypomethylation (a decrease in the epigenetic methylation of cytosine in DNA) were significantly higher in obese tumors compared with overweight tumors, which was consistent with DEGs. For example, in ESCA, 297 hypomethylated and 312 hypermethylated genes were identified in the overweight group, while 2,740 hypomethylated and 1,017 hypermethylated genes were found in the obese group (Supplementary Fig. S5F–S5I). The enriched pathways of these dysmethylated genes revealed similar pathways based on DEG analysis, that is, inflammatory and hormone-related pathways (Supplementary Fig. S5G and S5I). Overall, the methylation patterns were consistent with DEG patterns—hypermethylation repressed gene expression and hypomethylation led to their overexpression. However, in obese LIHC, both hypermethylation and upregulated genes were presented, which contradicted with the general regulatory pattern of methylation and gene expression. This result suggests that besides methylation, there may be other mechanisms involved in the regulation of gene expression in liver cancer.

Tumor microenvironment created by excessive body fat

Excessive body fat can affect tumor microenvironment, represented by an increased immune cell infiltration including macrophages, T lymphocytes, natural killer cells, and others and increased interstitial fibrosis by altering extracellular matrix mechanics (61). On the basis of a study that summarized the immune features of all TCGA samples (28), we were able to detect distinct immune microenvironment in tumors with different BMI. We used CIBERSORTx (62) and applied PSW approach to compare the compositions of infiltrating immune cells based on gene expression values between normal weight and overweight/obese tumors. Although overweight or obesity has been demonstrated to be associated with elevated leukocytes in nontumor tissues (63), the distributions of immune cells in tumors with overweight/obesity had not been explored. In comparison to normal weight tumors, we did not observe a higher proportion of immune cells or immune features in overweight tumors (Supplementary Fig. S5A and S5B). However, for obese tumors, multiple cancer types showed obesity-biased immune features (Supplementary Fig. S6C). In particular, LIHC demonstrated a significantly higher leukocyte fraction compared with normal weight tumors (17% in obesity vs. 6% in normal weight; Fig. 6A). Although overall increased leukocytes were not observed in the other cancer with overweight/obesity, lymphocytes, a type of leukocyte that plays a crucial role in the adaptive immune response, were significantly infiltrated in obese CESC and READ (Fig. 6B and C), a sign of inflammation in cancer (64). Regulatory T cells (Treg) are known to be involved in tumor development and progression by inhibiting antitumor immunity (65) and decline in obesity (66). Follicular helper T cell (Tfh) and resting memory CD4 T cell are supposed to be more abundant in obese than lean individuals (67). However, in UCEC, the expected accumulations of these three types of immune cells did not show in obese tumors. Instead, they were shown in the normal weight tumors. (Fig. 6DF).

Figure 6.

Tumor microenvironment alterations associated with overweight and obesity. A, Comparison of leukocyte fraction between normal weight and obese LIHC tumors. B and C, Lymphocyte infiltration in obese CESC (B) and READ (C) tumors. D–F, Distribution of Tregs, Tfh, and resting memory CD4 T cells between normal weight and obese tumors in UCEC (weighted t test, P value <0.05). G, Composition of ISs (C1–C6) in normal weight, overweight, and obese tumors for various cancer types.

Figure 6.

Tumor microenvironment alterations associated with overweight and obesity. A, Comparison of leukocyte fraction between normal weight and obese LIHC tumors. B and C, Lymphocyte infiltration in obese CESC (B) and READ (C) tumors. D–F, Distribution of Tregs, Tfh, and resting memory CD4 T cells between normal weight and obese tumors in UCEC (weighted t test, P value <0.05). G, Composition of ISs (C1–C6) in normal weight, overweight, and obese tumors for various cancer types.

Close modal

To explore the overweight/obesity-biased immune features, we examined the composition of the ISs represented by the modules of immune signature sets. Thorsson and colleagues (28) summarized six ISs C1–C6, which were characterized by a distribution of scores of 160 immune expression signatures. They were defined as wound healing (C1), IFNγ dominant (C2), inflammatory (C3), lymphocyte depleted (C4), immunologically quiet (C5), and TGFβ dominant (C6), respectively. The composition of IS varied between normal weight tumors and overweight/obese tumors in multiple cancer types (Fig. 6G). For example, in ESCA, C1, C4, and C6 enriched in obesity patients, while C2 and C3 decreased. C4 and C6 were the more mixed signature subtypes and had the least favorable prognosis. This is consistent with our observation that obese ESCA had poor survival. In UCEC, C3 (inflammatory), which was reported to have the best prognosis, increased in both overweight and obese tumors. This is also consistent with our result that overweight and obese had better survival compared with normal weight tumors in UCEC. In KIRP and LIHC, C3 (inflammatory) also increased in obese tumors, while C4 (lymphocyte depleted) decreased. Although DEGs based overall survival analysis did not show better prognosis in overweight or obese tumors (Supplementary Fig. S6D–S6G), the tumor microenvironment indicates that some overweight or obese tumors may have better survival. In COAD, C1 (would healing) decreased and C4 (lymphocyte depleted) increased in obese tumors, implying a worse prognosis in some obese colon cancers. The diverse patterns of ISs among different weight groups in individual cancer highlight the importance of excessive body fat in the formation of the tumor microenvironment.

Accumulated evidence has demonstrated the association between obesity and cancer (9), but the molecular differences between normal weight and overweight/obese patients with cancer are still elusive. Previous cohort studies have attempted to identify the BMI-associated molecular features (3), but these studies have had limitations due to their small sample size or reliance on single-molecular data from animal models. In addition, these studies often failed to account for other confounding factors that can also impact the molecular features (68). Our study is designed to comprehensively examine the molecular differences between patients with normal weight and overweight/obese cancer across a broad range of cancer types, and to systematically characterize the effects of overweight or obesity on the molecular profile of cancer from DNA to RNA levels. By applying the PSW algorithm, we eliminated the potential confounding factors such as age, sex, tumor stage, and purity on all assessments of the features. As a result, the overweight/obesity-associated molecular features we have identified faithfully explain that the molecular differences are due to overweight and/or obesity. To the best of our knowledge, our results provide the most comprehensive understanding of the molecular differences including mutational pattern, gene expression, and immune features between patients with normal weight and overweight/obese cancer.

Previous epidemiologic studies have shown that overweight and obesity are associated with a higher risk of 13 cancer types. In particular, esophageal and endometrial cancers are strongly affected by overweight/obesity. Our results, which combine somatic mutations, CNVs, gene expression, and immune cell composition, have shown that the molecular features in patients with obese ESCA are generally different from those in normal weight patients in almost every aspect. Obese ESCA is more likely to accumulate mutations caused by defect MMR. However, the unique overweight/obesity-biased molecular features in UCEC appear to contradict previous studies (69) that reported obesity as a stronger risk factor for tumorigenesis. This apparent discrepancy in UCEC, known as the “obesity paradox,” has been reported in various studies (70), suggesting that high BMI might confer a protective effect in endometrial cancer, despite being a risk factor for the majority of other cancer types. Our findings lend support to this paradox, as we observed normal weight–biased mutational and gene expression patterns as well as worse prognosis in patients with UCEC. This observation warrants further investigation into the potential mechanisms underlying the obesity paradox in UCEC and its clinical implications. Other obese-associated cancer types, including COAD, KIRP, LIHC, READ, also demonstrated significant differences at molecular levels, suggesting that the association from epidemiologic studies has a molecular foundation. TCGA data provided us with the opportunities to investigate other cancer types with BMI information. DLBC also presented with obese-biased mutational signatures, implying a link between DLBC and obesity.

Among the cancer types associated with obesity, several originate from the digestive system, including ESCA, COAD, READ, and LIHC. These organs are primarily responsible for nutrition intake and energy metabolism, suggesting a potential link between these physiologic processes and tumorigenesis. For instance, excess nutrient intake can lead to an overload of the liver's metabolic capacity, resulting in non-alcoholic fatty liver disease, which is a risk factor for LIHC (71). Similarly, obesity-related changes in gut microbiota composition can influence the development of colorectal cancer (COAD and READ) by affecting gut barrier function and immune response (72). Further research is needed to fully elucidate the mechanisms underlying the preferential effect of obesity on tumorigenesis in digestive organs.

The molecular features we identified, particularly the DEGs, point to a strong involvement of inflammatory, immune, and hormone-related pathways in patients with overweight and obese cancer. These findings are consistent with previous research linking obesity to chronic inflammation and altered immune response (73). Moreover, some DEGs are associated with a significant difference in overall survival, suggesting that these genes may be potential therapeutic markers. These dysregulated pathways or genes in overweight and obese patients may have a significant impact on the development and progression of other cancers associated with obesity, such as breast cancer, and pancreatic cancer, among others. Our results highlight the need to consider the complex interplay between obesity, inflammation, immune response, and hormonal regulation in the development and management of cancer.

In addition to the molecular difference between normal weight and overweight/obese patients, we also observed distinct patterns of immune cell infiltration in various cancer types. This finding indicates that the tumor microenvironment is affected by the patient's body weight status and may have profound implications for cancer immunotherapy. The development of personalized immunotherapies that take into account the patient's BMI and the associated molecular features may lead to improved clinical outcomes.

In conclusion, our comprehensive analysis of molecular differences between patients with normal weight and overweight/obese cancer across 14 cancer types from TCGA provides valuable insights into the complex interplay between body weight and cancer. By uncovering overweight/obesity-specific molecular features and their potential implications in cancer development and progression, our study lays the groundwork for future research aimed at developing tailored therapeutic strategies for patients with overweight and obese cancer. Moreover, our findings call for a better understanding of the molecular basis of the obesity paradox in endometrial cancer, which may have important clinical implications for the prevention and treatment of this particular cancer type.

No disclosures were reported.

F. Huang: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft. P. Xu: Formal analysis, visualization, methodology, writing–original draft. Z. Yue: Methodology. Y. Song: Writing–review and editing. K. Hu: Writing–review and editing. X. Zhao: Investigation, writing–review and editing. M. Gao: Data curation, formal analysis, visualization, writing–review and editing. Z. Chong: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, project administration, writing–review and editing.

This work was supported by a grant from National Institute of General Medical Sciences (1R35GM138212); a grant from the National Institute of Minority Health and Health Disparities (U54MD000502); the BioData Catalyst Fellowship from National Heart, Lung, and Blood Institute (a subaward from 1OT3HL147154) to Z. Chong.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

1.
Hruby
A
,
Hu
FB
.
The epidemiology of obesity: a big picture
.
Pharmacoeconomics
2015
;
33
:
673
89
.
2.
Bandera
EV
,
Fay
SH
,
Giovannucci
E
,
Leitzmann
MF
,
Marklew
R
,
McTiernan
A
, et al
.
The use and interpretation of anthropometric measures in cancer epidemiology: a perspective from the world cancer research fund international continuous update project
.
Int J Cancer
2016
;
139
:
2391
7
.
3.
Hu
C
,
Chen
X
,
Yao
C
,
Liu
Y
,
Xu
H
,
Zhou
G
, et al
.
Body mass index-associated molecular characteristics involved in tumor immune and metabolic pathways
.
Cancer Metab
2020
;
8
:
21
.
4.
Lauby-Secretan
B
,
Scoccianti
C
,
Loomis
D
,
Grosse
Y
,
Bianchini
F
,
Straif
K
, et al
.
Body fatness and cancer–viewpoint of the IARC working group
.
N Engl J Med
2016
;
375
:
794
8
.
5.
Arthuso
FZ
,
Fairey
AS
,
Boule
NG
,
Courneya
KS
.
Associations between body mass index and bladder cancer survival: is the obesity paradox short-lived?
Can Urol Assoc J
2022
;
16
:
E261
E7
.
6.
Yugawa
K
,
Itoh
S
,
Iseda
N
,
Kurihara
T
,
Kitamura
Y
,
Toshima
T
, et al
.
Obesity is a risk factor for intrahepatic cholangiocarcinoma progression associated with alterations of metabolic activity and immune status
.
Sci Rep
2021
;
11
:
5845
.
7.
Skibola
CF
.
Obesity, diet and risk of non-Hodgkin lymphoma
.
Cancer Epidemiol Biomarkers Prev
2007
;
16
:
392
5
.
8.
Strongman
H
,
Brown
A
,
Smeeth
L
,
Bhaskaran
K
.
Body mass index and Hodgkin's lymphoma: UK population-based cohort study of 5.8 million individuals
.
Br J Cancer
2019
;
120
:
768
70
.
9.
Avgerinos
KI
,
Spyrou
N
,
Mantzoros
CS
,
Dalamaga
M
.
Obesity and cancer risk: emerging biological mechanisms and perspectives
.
Metabolism
2019
;
92
:
121
35
.
10.
MacDougald
OA
,
Burant
CF
.
The rapidly expanding family of adipokines
.
Cell Metab
2007
;
6
:
159
61
.
11.
Eliassen
AH
,
Missmer
SA
,
Tworoger
SS
,
Spiegelman
D
,
Barbieri
RL
,
Dowsett
M
, et al
.
Endogenous steroid hormone concentrations and risk of breast cancer among premenopausal women
.
J Natl Cancer Inst
2006
;
98
:
1406
15
.
12.
Wu
Y
,
Cui
K
,
Miyoshi
K
,
Hennighausen
L
,
Green
JE
,
Setser
J
, et al
.
Reduced circulating insulin-like growth factor I levels delay the onset of chemically and genetically induced mammary tumors
.
Cancer Res
2003
;
63
:
4384
8
.
13.
Khandekar
MJ
,
Cohen
P
,
Spiegelman
BM
.
Molecular mechanisms of cancer development in obesity
.
Nat Rev Cancer
2011
;
11
:
886
95
.
14.
Pollak
M
.
Insulin and insulin-like growth factor signalling in neoplasia
.
Nat Rev Cancer
2008
;
8
:
915
28
.
15.
Heilbronn
LK
,
de Jonge
L
,
Frisard
MI
,
DeLany
JP
,
Larson-Meyer
DE
,
Rood
J
, et al
.
Effect of 6-month calorie restriction on biomarkers of longevity, metabolic adaptation, and oxidative stress in overweight individuals: a randomized controlled trial
.
JAMA
2006
;
295
:
1539
48
.
16.
Cancer Genome Atlas Research Network
;
Weinstein
JN
,
Collisson
EA
,
Mills
GB
,
Shaw
KR
,
Ozenberger
BA
, et al
.
The Cancer Genome Atlas Pan-Cancer analysis project
.
Nat Genet
2013
;
45
:
1113
20
.
17.
Li
F
,
Morgan
KL
,
Zaslavsky
AM
.
Balancing covariates via propensity score weighting
.
J Am Stat Assoc
2018
;
113
:
390
400
.
18.
Colaprico
A
,
Silva
TC
,
Olsen
C
,
Garofano
L
,
Cava
C
,
Garolini
D
, et al
.
TCGAbiolinks: an R/Bioconductor package for integrative analysis of TCGA data
.
Nucleic Acids Res
2016
;
44
:
e71
.
19.
Ellrott
K
,
Bailey
MH
,
Saksena
G
,
Covington
KR
,
Kandoth
C
,
Stewart
C
, et al
.
Scalable open science approach for mutation calling of tumor exomes using multiple genomic pipelines
.
Cell Syst
2018
;
6
:
271
81
.
20.
Mayakonda
A
,
Lin
DC
,
Assenov
Y
,
Plass
C
,
Koeffler
HP
.
Maftools: efficient and comprehensive analysis of somatic variants in cancer
.
Genome Res
2018
;
28
:
1747
56
.
21.
Lawrence
MS
,
Stojanov
P
,
Mermel
CH
,
Robinson
JT
,
Garraway
LA
,
Golub
TR
, et al
.
Discovery and saturation analysis of cancer genes across 21 tumour types
.
Nature
2014
;
505
:
495
501
.
22.
Blokzijl
F
,
Janssen
R
,
van Boxtel
R
,
Cuppen
E
.
MutationalPatterns: comprehensive genome-wide analysis of mutational processes
.
Genome Med
2018
;
10
:
33
.
23.
Alexandrov
LB
,
Kim
J
,
Haradhvala
NJ
,
Huang
MN
,
Ng
AWT
,
Wu
Y
, et al
.
The repertoire of mutational signatures in human cancer
.
Nature
2020
;
578
:
94
101
.
24.
Mermel
CH
,
Schumacher
SE
,
Hill
B
,
Meyerson
ML
,
Beroukhim
R
,
Getz
G
.
GISTIC2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers
.
Genome Biol
2011
;
12
:
R41
.
25.
Kuleshov
MV
,
Jones
MR
,
Rouillard
AD
,
Fernandez
NF
,
Duan
Q
,
Wang
Z
, et al
.
Enrichr: a comprehensive gene set enrichment analysis web server 2016 update
.
Nucleic Acids Res
2016
;
44
:
W90
7
.
26.
Chen
EY
,
Tan
CM
,
Kou
Y
,
Duan
QN
,
Wang
ZC
,
Meirelles
GV
, et al
.
Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool
.
BMC Bioinformatics
2013
;
14
:
128
.
27.
Goldman
MJ
,
Craft
B
,
Hastie
M
,
Repecka
K
,
McDade
F
,
Kamath
A
, et al
.
Visualizing and interpreting cancer genomics data via the Xena platform
.
Nat Biotechnol
2020
;
38
:
675
8
.
28.
Thorsson
V
,
Gibbs
DL
,
Brown
SD
,
Wolf
D
,
Bortone
DS
,
Ou Yang
TH
, et al
.
The immune landscape of cancer
.
Immunity
2018
;
48
:
812
30
.
29.
Martincorena
I
,
Campbell
PJ
.
Somatic mutation in cancer and normal cells
.
Science
2015
;
349
:
1483
9
.
30.
Boland
CR
,
Goel
A
.
Microsatellite instability in colorectal cancer
.
Gastroenterology
2010
;
138
:
2073
87
.
31.
Patel
AV
,
Diver
WR
,
Teras
LR
,
Birmann
BM
,
Gapstur
SM
.
Body mass index, height and risk of lymphoid neoplasms in a large United States cohort
.
Leuk Lymphoma
2013
;
54
:
1221
7
.
32.
Castillo
JJ
,
Ingham
RR
,
Reagan
JL
,
Furman
M
,
Dalia
S
,
Mitri
J
.
Obesity is associated with increased relative risk of diffuse large B-cell lymphoma: a meta-analysis of observational studies
.
Clin Lymphoma Myeloma Leuk
2014
;
14
:
122
30
.
33.
Teras
LR
,
Bertrand
KA
,
Deubler
EL
,
Chao
CR
,
Lacey
JV
,
Patel
AV
, et al
.
Body size and risk of non-Hodgkin lymphoma by subtype: a pooled analysis from six prospective cohorts in the United States
.
Brit J Haematol
2022
;
197
:
714
27
.
34.
Huge
N
,
Sandbothe
M
,
Schroder
AK
,
Stalke
A
,
Eilers
M
,
Schaffer
V
, et al
.
Wnt status-dependent oncogenic role of BCL9 and BCL9L in hepatocellular carcinoma
.
Hepatol Int
2020
;
14
:
373
84
.
35.
Tsang
SM
,
Oliemuller
E
,
Howard
BA
.
Regulatory roles for SOX11 in development, stem cells and cancer
.
Semin Cancer Biol
2020
;
67
:
3
11
.
36.
Faraj Shaglouf
LH
,
Ranjpour
M
,
Wajid
S
,
Jain
SK
.
Elevated expression of cellular SYNE1, MMP10, and GTPase1 and their regulatory role in hepatocellular carcinoma progression
.
Protoplasma
2020
;
257
:
157
67
.
37.
Yasukawa
M
,
Liu
Y
,
Hu
L
,
Cogdell
D
,
Gharpure
KM
,
Pradeep
S
, et al
.
ADAMTS16 mutations sensitize ovarian cancer cells to platinum-based chemotherapy
.
Oncotarget
2017
;
8
:
88410
20
.
38.
Ye
Z
,
Lu
Y
,
Wu
T
.
The impact of ATP-binding cassette transporters on metabolic diseases
.
Nutr Metab
2020
;
17
:
61
.
39.
Li
P
,
Xiao
J
,
Zhou
B
,
Wei
J
,
Luo
J
,
Chen
W
.
SYNE1 mutation may enhance the response to immune checkpoint blockade therapy in clear cell renal cell carcinoma patients
.
Aging
2020
;
12
:
19316
24
.
40.
Agrawal
N
,
Lawler
K
,
Davidson
CM
,
Keogh
JM
,
Legg
R
,
Barroso
I
, et al
.
Predicting novel candidate human obesity genes and their site of action by systematic functional screening in Drosophila
.
PLoS Biol
2021
;
19
:
e3001255
.
41.
Cancer Genome Atlas Research Network; Analysis Working Group: Asian University; BC Cancer Agency; Brigham and Women's Hospital; Broad Institute; Brown University
, et al
Integrated genomic characterization of oesophageal carcinoma
.
Nature
2017
;
541
:
169
75
.
42.
Cancer Genome Atlas Network
.
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
43.
Zoncu
R
,
Efeyan
A
,
Sabatini
DM
.
mTOR: from growth signal integration to cancer, diabetes and ageing
.
Nat Rev Mol Cell Biol
2011
;
12
:
21
35
.
44.
Claussnitzer
M
,
Dankel
SN
,
Kim
KH
,
Quon
G
,
Meuleman
W
,
Haugen
C
, et al
.
FTO obesity variant circuitry and adipocyte browning in humans
.
N Engl J Med
2015
;
373
:
895
907
.
45.
Ohno
Y
,
Nakamichi
S
,
Ohkuni
A
,
Kamiyama
N
,
Naoe
A
,
Tsujimura
H
, et al
.
Essential role of the cytochrome P450 CYP4F22 in the production of acylceramide, the key lipid for skin permeability barrier formation
.
Proc Natl Acad Sci U S A
2015
;
112
:
7707
12
.
46.
Ghiselli
G
.
SMC3 knockdown triggers genomic instability and p53-dependent apoptosis in human and zebrafish cells
.
Mol Cancer
2006
;
5
:
52
.
47.
Losada
A
.
Cohesin in cancer: chromosome segregation and beyond
.
Nat Rev Cancer
2014
;
14
:
389
93
.
48.
Masaki
M
,
Kurisaki
T
,
Shirakawa
K
,
Sehara-Fujisawa
A
.
Role of meltrin alpha (ADAM12) in obesity induced by high- fat diet
.
Endocrinology
2005
;
146
:
1752
63
.
49.
Nagaraj
K
,
Lapkina-Gendler
L
,
Sarfstein
R
,
Gurwitz
D
,
Pasmanik-Chor
M
,
Laron
Z
, et al
.
Identification of thioredoxin-interacting protein (TXNIP) as a downstream target for IGF1 action
.
Proc Natl Acad Sci U S A
2018
;
115
:
1045
50
.
50.
Cha-Molstad
H
,
Saxena
G
,
Chen
J
,
Shalev
A
.
Glucose-stimulated expression of Txnip is mediated by carbohydrate response element-binding protein, p300, and histone H4 acetylation in pancreatic beta cells
.
J Biol Chem
2009
;
284
:
16898
905
.
51.
Alves
A
,
Bassot
A
,
Bulteau
AL
,
Pirola
L
,
Morio
B
.
Glycine metabolism and its alterations in obesity and metabolic diseases
.
Nutrients
2019
;
11
:
1356
.
52.
Klaassen
CD
,
Liu
J
,
Diwan
BA
.
Metallothionein protection of cadmium toxicity
.
Toxicol Appl Pharm
2009
;
238
:
215
20
.
53.
Meng
JH
,
Wang
JF
.
Role of SNARE proteins in tumourigenesis and their potential as targets for novel anti-cancer therapeutics
.
Biochim Biophys Acta
2015
;
1856
:
1
12
.
54.
Incio
J
,
Tam
J
,
Rahbari
NN
,
Suboj
P
,
McManus
DT
,
Chin
SM
, et al
.
PlGF/VEGFR-1 signaling promotes macrophage polarization and accelerated tumor progression in obesity
.
Clin Cancer Res
2016
;
22
:
2993
3004
.
55.
Adams
SH
,
Chui
C
,
Schilbach
SL
,
Yu
XX
,
Goddard
AD
,
Grimaldi
JC
, et al
.
BFIT, a unique acyl-CoA thioesterase induced in thermogenic brown adipose tissue: cloning, organization of the human gene and assessment of a potential link to obesity
.
Biochem J
2001
;
360
:
135
42
.
56.
Zhang
Y
,
Li
Y
,
Niepel
MW
,
Kawano
Y
,
Han
S
,
Liu
S
, et al
.
Targeted deletion of thioesterase superfamily member 1 promotes energy expenditure and protects against obesity and insulin resistance
.
Proc Natl Acad Sci U S A
2012
;
109
:
5417
22
.
57.
Muhammad
N
,
Bhattacharya
S
,
Steele
R
,
Phillips
N
,
Ray
RB
.
Involvement of c-Fos in the promotion of cancer stem-like cell properties in head and neck squamous cell carcinoma
.
Clin Cancer Res
2017
;
23
:
3120
8
.
58.
Tao
H
,
Liao
Y
,
Yan
Y
,
He
Z
,
Zhou
J
,
Wang
X
, et al
.
BRCC3 promotes tumorigenesis of bladder cancer by activating the NF-kappa B signaling pathway through targeting TRAF2
.
Front Cell Dev Biol
2021
;
9
:
720349
.
59.
Park
MH
,
Kwon
HJ
,
Kim
JR
,
Lee
B
,
Lee
SJ
,
Bae
YK
.
Elevated interleukin-13 receptor alpha 1 expression in tumor cells is associated with poor prognosis in patients with invasive breast cancer
.
Ann Surg Oncol
2017
;
24
:
3780
7
.
60.
Moore
LD
,
Le
T
,
Fan
G
.
DNA methylation and its basic function
.
Neuropsychopharmacology
2013
;
38
:
23
38
.
61.
Donohoe
CL
,
Lysaght
J
,
O'Sullivan
J
,
Reynolds
JV
.
Emerging concepts linking obesity with the hallmarks of cancer
.
Trends Endocrinol Metab
2017
;
28
:
46
62
.
62.
Newman
AM
,
Steen
CB
,
Liu
CL
,
Gentles
AJ
,
Chaudhuri
AA
,
Scherer
F
, et al
.
Determining cell type abundance and expression from bulk tissues with digital cytometry
.
Nat Biotechnol
2019
;
37
:
773
82
.
63.
Yoshimura
A
,
Ohnishi
S
,
Orito
C
,
Kawahara
Y
,
Takasaki
H
,
Takeda
H
, et al
.
Association of peripheral total and differential leukocyte counts with obesity-related complications in young adults
.
Obesity Facts
2015
;
8
:
1
16
.
64.
Zhao
H
,
Wu
L
,
Yan
G
,
Chen
Y
,
Zhou
M
,
Wu
Y
, et al
.
Inflammation and tumor progression: signaling pathways and targeted intervention
.
Signal Transduct Target Ther
2021
;
6
:
263
.
65.
Ohue
Y
,
Nishikawa
H
.
Regulatory T (Treg) cells in cancer: can Treg cells be a new therapeutic target?
Cancer Sci
2019
;
110
:
2080
9
.
66.
Bradley
D
,
Shantaram
D
,
Smith
A
,
Hsueh
WA
.
Adipose tissue T regulatory cells: implications for health and disease
.
Adv Exp Med Biol
2021
;
1278
:
125
39
.
67.
Cipolletta
D
,
Feuerer
M
,
Li
A
,
Kamei
N
,
Lee
J
,
Shoelson
SE
, et al
.
PPAR-gamma is a major driver of the accumulation and phenotype of adipose tissue Treg cells
.
Nature
2012
;
486
:
549
53
.
68.
Li
CH
,
Haider
S
,
Boutros
PC
.
Age influences on the molecular presentation of tumours
.
Nat Commun
2022
;
13
:
208
.
69.
Onstad
MA
,
Schmandt
RE
,
Lu
KH
.
Addressing the role of obesity in endometrial cancer risk, prevention, and treatment
.
J Clin Oncol
2016
;
34
:
4225
30
.
70.
Renehan
AG
,
Zwahlen
M
,
Egger
M
.
Adiposity and cancer risk: new mechanistic insights from epidemiology
.
Nat Rev Cancer
2015
;
15
:
484
98
.
71.
Hydes
T
,
Alam
U
,
Cuthbertson
DJ
.
The impact of macronutrient intake on non-alcoholic fatty liver disease (NAFLD): too much fat, too much carbohydrate, or just too many calories?
Front Nutr
2021
;
8
:
640557
.
72.
Rebersek
M
.
Gut microbiome and its role in colorectal cancer
.
BMC Cancer
2021
;
21
:
1325
.
73.
Woodall
MJ
,
Neumann
S
,
Campbell
K
,
Pattison
ST
,
Young
SL
.
The effects of obesity on anti-cancer immunity and cancer immunotherapy
.
Cancers
2020
;
12
:
1230
.
This open access article is distributed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.