Colon cancer combined with obesity indicates improved survival- research on relevant mechanism

Obesity is an independent prognostic factor for improved overall survival

To investigate the relationship between BMI and Overall Survival (OS), we divided colon cancer patients in TCGA database into 3 groups according to the WHO standards (normal weight: 18.5≤BMI<25, overweight: 25≤BMI<30 and obesity: BMI≥30 kg/m². There are only 2 cases of underweight patients, BMI<18.5 kg/m², so they were eliminated) and performed with Kaplan-Meier analysis [12]. Both Compared to normal weight and overweight patients, obesity patients (BMI >=30 kg/m²) showed improved OS. However, no significant difference in OS was demonstrated between normal weight and overweight patients (Figure 1A). Subsequently, we combined normal weight and overweight patients as normal group and redistributed all patients into two groups (normal group: 18.5≤BMI<30kg/m²vs obesity group: BMI≥30kg/m²), and performed Kaplan-Meier analysis repeatedly. The obesity group showed significantly better OS compared with normal group (P value=0.012, Figure 1B). Additionally, we also investigated the correlations of obesity with Disease-Free Survival (DFS) and Progression-Free Survival (PFS) respectively (Figure 1C, 1D). Though improved RFS and PFS were demonstrated in obesity group, p values were more than 0.05 (DFS: 0.093, PFS: 0.012). This might be due to the limited sample size. Furthermore, we investigated risk factors of OS via univariate and multivariate Cox regression model respectively. In univariate Cox regression model, Primary tumor (T), Regional lymph node (N), Distant metastasis (M) as well as BMI group (obesity group vs normal group, Hazard Ratio, HR=0.323, P value=0.018) were associated with OS (Figure 1E). In multivariate Cox regression model, only Primary tumor (T) and BMI group (HR=0.3, P value=0.049) were independent prognostic factors (Figure 1F). We also investigated the association of obesity and other characteristics of colon cancer patients, including age, gender, pathological stage, microsatellite status, and so on. Whereas BMI showed no association with all other characteristics (Table 1). In summary, we demonstrated obesity group ((BMI >=30 kg/m²) had improved OS compared with normal group (BMI<30kg/m²).

Figure 1. Obesity is an independent prognostic factor for improved overall survival. (A) The association of BMI and Overall Survival (OS) performed by Kaplan-Meier analysis. Patients were divided into 3 groups according to WHO standard (normal weight: BMI<25, overweight: 25≤BMI<30 and obesity: BMI≥30 kg/m²). (B) All patients redistributed into 2 groups (combined normal weight and overweight patients as normal group: BMI<30 vs obesity group: BMI≥30 kg/m²) and performed by Kaplan-Meier analysis. (C) Disease-Free Survival of normal group and obesity group performed by Kaplan-Meier analysis. (D) Progression-Free Survival of normal group and obesity group performed by Kaplan-Meier analysis. (E) Risk factors of OS analyzed by univariate Cox regression model. (F) Risk factors of OS analyzed by multivariate Cox regression model. (^*p<0.05, ^**p<0.01, ^***p<0.001).

Differentially expressed genes in obesity patients were closely related to tumor regulation

R software (“Limma” Package) was used to screen differentially expressed genes (DEGs) between normal and obesity groups based on transcriptome data. A total of 225 DEGs were confirmed and further performed with GO, KEGG, and PPI analysis respectively (Figure 2A). In both GO and KEGG analysis, nutrient digestion annotations and pathways were the most enriched, including vitamins, proteins, and so on (Figure 2B, 2C). In PPI analysis, we found 4 main modules, including small proline-rich protein 2 family (SPRR2s), Carboxypeptidase associated genes (CPA1, CPB1), Keratin family (KRTs) and Apolipoprotein family (APOA1, APOB, APOC3) (Figure 2D). All these families were reported to regulate the development of tumor.

Figure 2. Differentially expressed genes in obesity patients are closely related to tumor regulation. (A) Differentially expressed genes (DEGs) between obesity and normal groups. (B) Gene Ontology (GO) analysis of DEGs. (C) KEGG pathway analysis of DEGs. (D) Protein-protein interaction (PPI) analysis of DEGs.

Obesity was associated with repressed hypoxia in colon cancer via inhibiting miR-210

Similarly, we also screened differentially expressed miRNA (DEMs) between normal and obesity groups. 11 DEMs were finally confirmed, including miR-26a-1, miR-210, and so on (Figure 3A). Subsequently, we investigated the relationship between DEMs and OS. We found miR-210 was decreased in obesity group and associated with better OS (Figure 3B). Possible targeted genes of miR-210 were predicted by the TargetScan database (http://www.targetscan.org/vert_72/) and performed with GO and KEGG analysis (Figure 3C, 3D). Interestingly, target genes of miR-210 were enriched in hypoxia associated pathways and annotations (VEGF signaling pathway, HIF-1 signaling pathway, Response to hypoxia). Hypoxia plays a crucial role in cancer metabolism triggering cancer aggressiveness, invasiveness, and treatment resistance. Therefore, we further assessed the hypoxia condition in tumors of normal and obesity groups. Three different hypoxia scoring systems were adopted, including Buffa system [13], Ragnum system [14] and Winter system [15]. We found the hypoxia score of obesity group was less than normal group in all three scoring systems (Figure 3E). Additionally, the expression of miR-210 also showed co-expression with all hypoxia scores derived from different systems (Figure 3F). Finally, we also investigated the Reverse Phase Protein Arrays (RPPA) data and confirmed several reported tumor regulating proteins were differentially expressed in obesity group, including 14-3-3β, GTP-binding protein Di-Ras3 (ARHI), Fatty Acid Synthase (FASN) and ribosomal protein S6 (RPS6, Figure 3G). In summary, downregulated miR-210 in obesity patients repressed hypoxia of tumors, thus prevented the progression of colon cancer.

Figure 3. Obesity was associated with repressed hypoxia in colon cancer via inhibiting miR-210. (A) Differentially expressed miRNAs (DEMs) between obesity and normal groups. (B) The relationship of OS and miR-210 (divided by median of expression) performed by Kaplan-Meier analysis. (C) KEGG pathway analysis of miR-210-targeted genes (predicted by TargetScan database). (D) GO analysis of miR-210-targeted genes. (E) Hypoxia score of obesity and normal groups, derived from three different hypoxia scoring systems (Ragnum, Winter, Buffa hypoxia system). (F) The correlation of miR-210 and 3 different hypoxia scores analyzed by Pearson Correlation Coefficient. (G) Differentially expressed proteins between obesity and normal groups derived from Reverse Phase Protein Arrays (RPPA) data. (H) The invasion and migration ability of HCT-116 cells under hypoxia and normoxia. (I) The proliferation ability of HCT-116 cells under hypoxia and normoxia. (J) The expression of hypoxia associated genes of HCT-116 cells under hypoxia and normoxia. (^*p<0.05, ^**p<0.01, ^***p<0.001).

Subsequently, we researched the role of miR-210 in vitro under both normoxia and hypoxia. Overexpression of miR-210 significantly promoted the invasion and migration ability of HCT-116 under both normoxia and hypoxia. Additionally, overexpression of miR-210 promoted the proliferation ability of HCT-116 only under hypoxia, but not under normoxia. Finally, we investigated hypoxia associated genes (glycolysis associated genes: HK2, HIF1A, PKM, Glut1 and angiogenesis associated genes: VEGFA). Interestingly, overexpression of miR-210 significantly promoted the expression of HK2, HIF1A and VEGFA only under hypoxia. Under normoxia, miR-210 showed no effect at these genes.

Obesity was associated with the alteration of gene mutations

In colon cancer, various oncogene and tumor suppressor gene mutations are reported to be crucial in the development of the disease [16]. Therefore, we further investigated the somatic mutations of colon cancer via R software (“maftools” Package) based on MAF data. We demonstrated an obvious difference between gene mutations frequency in compare of obesity group to normal group. Obesity group showed a higher mutation frequency of LDHAL6B, CPXM2, HAPLN3, and so on. Obesity group also showed a lower mutation frequency of PLCG2, STK31, ADGB, and so on (Figure 4A). Furthermore, we selected 10 most common mutations in colon cancer patients. In the top 10 mutations assessment, the frequency of KRAS and PIK3CA mutations were significantly decreased, whereas the frequency of TP53 mutation were significantly increased in obesity group compared with normal group (Figure 4B). KRAS, TP53 and PIK3CA were important gene mutations in colon cancer that associated with progression and prognosis [17, 18]. We further showed specific mutation sites, types and frequency of them in schematic representations of protein domain information (Figure 4C). Mutation sites were shown on the x-axis, and the frequency of a particular mutation was represented by the height (y-axis). We demonstrated most mutation type of PIK3AC and KRAS was missense mutation in both normal and obesity groups. Obesity group not only have a lower frequency of mutations but also have fewer mutation sites. Compared with PIK3AC and KRAS, TP53 (p53) had various mutation types, including missense mutation, frameshift mutation, in-frame mutation nonsense mutation, and splice-site mutations. Though obesity group showed increased mutation frequency of TP53 compared with normal patients. There was significantly decreased mutation sites in obesity group patients. Especial in TP53 tetramer domain, obesity group showed no mutation site, whereas normal group showed 4 mutation sites, including 2 missense mutations, 1 splice site mutation, and 1 nonsense mutation. Subsequently, we investigated the correlation between each top 10 mutations in two groups (Figure 4D). Interestingly, more co-occurrences between mutations were demonstrated in obesity group compared with normal group (APC and KRAS, PIK3CA and KRAS, etc.) Finally, we investigated the copy number variation (CNV) in two group. Obesity group showed reduced CNV compared with normal group, and the reduction mainly came from chr3, 10, 11, 22 (Figure 4E). In summary, we demonstrated the gene mutation frequency (especial in PIK3CA and KRAS), as well as CNV of obesity groups, were decreased. We inferred the genome of obesity patients was more stable compared with normal patients.

Figure 4. Obesity was associated with the alteration of gene mutations. (A) Forest plot for different mutations frequency compared between normal and obesity groups, analyzed by Chi-square test. (B) Top 10 mutations in normal and obesity groups. (C) “Lollipop” graph for specific mutation frequency, types and sites in PIK3CA, KRAS and TP53 (p53) protein domain. Mutation sites were shown on the x-axis, and the frequency of a particular mutation was represented by the height (y-axis). (D) Interaction of each top 10 mutation in normal and obesity group analyzed by Fisher exact test. (E) The distribution of copy number variation (CNV) in normal and obesity groups analyzed by Chi-square test. (^*p<0.05, ^**p<0.01, ^***p<0.001).

Obesity was associated with repression of immune checkpoints

Immune checkpoints are receptor-based signal cascades that negatively regulate T cells and cause immune tolerance, which allows tumors to evade and escape immune surveillance. The advent of immune checkpoint inhibitors (ICIs) have changed the landscape of cancer treatment. Therefore, we investigated the correlation between immune checkpoints and obesity. Compared with normal group, obesity group showed decreased expression of LAG3 and PD-1 (Figure 5A). Furthermore, the expression of LAG3 and PD-1 in healthy bowel tissues, colon cancer derived from normal (non-obesity) patients and obesity patients was confirmed by IHC (Figure 5B). These results inferred immune surveillance of T cells was activated in obesity patients, thus repressed the growth of tumors. However, decreased expression of immune checkpoints was also associated with the resistance of ICIs. Therefore, we investigated possible replacement through Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/). Low expression of LAG3 indicated sensitiveness of PLX4720, SB590885, and PHA-665752. Similarly, low expression of PD-1 indicated sensitivity of MG-132, Z-LLNle-CHO, TGX221, and so on (Figure 5C). Subsequently, we investigated co-expression between each immune checkpoint in obesity and normal groups. In general, all immune checkpoints were significantly co-expressed. It is noteworthy that TIGIT and IDO1 showed stronger co-expression with other immune checkpoints in normal group compared with obesity group (Figure 5D).

Figure 5. Obesity was associated with repression of immune checkpoints. (A) The expression of various immune checkpoints in obesity and normal groups. (B) The expression of LAG3 and PD-1 in healthy bowel tissues, colon cancer derived from normal (non-obesity) patients and obesity patients was confirmed by immunohistochemistry. (C) The association of drug sensitiveness and gene expression (LAG3 and PD-1) analyzed by Genomics of Drug Sensitivity in Cancer (GDSC) database. (D) The correlation matrix of each immune checkpoint in obesity and normal group. (^*p<0.05, ^**p<0.01).

Additionally, we also investigated the proportions of different immune cells (22 types) in colon cancer via CIBERSORT algorithm. In all 22 types of immune cells, only gamma delta T cell (γδT) showed significant inhibition in obesity group (Supplementary Figure 1A. 1B). As Interleukins play important roles in immune response, we also investigated the mRNA expression of all Interleukin families. IL11RA and IL17RB were decreased in obesity group, whereas IL19 and IL25 were increased (Supplementary Figure 1C). Among them, low expression of IL11RA also indicated improved OS (Supplementary Figure 1D).

Data acquisition and processing

All data were derived from the TCGA database. There were 227 cases of colon cancer patients with full Body Mass Index (BMI) data. We downloaded the transcriptome data, copy number data, Reverse Phase Protein Arrays (RPPR) data and mutation annotation format (MAF) from cBioPortal (http://cbioportal.org). Transcriptome data were normalized to gene expression data through the R software package “Limma”. MAF files were summarized, analyzed and visualized through R software package “maftools”. Copy number data and RPPR data can be analyzed directly. Additionally, relevant clinical characteristics were obtained (Table 2).

Gene ontology, KEGG pathway, and Protein–protein interaction analysis

Gene list was uploaded to Database for Annotation Visualization and Integrated Discovery (DAVID, david.ncifcrf.gov/) online tool for Gene ontology (GO) and KEGG pathway analysis. Concreate pathways and P-value were acquired and visualized by R software. For Protein–protein interaction (PPI) analysis, gene list was uploaded to STRING online tool (https://string-db.org/), genes whose combined score >0.4 were selected and imported into Cytoscape software (Version 3.5.1, The Cytoscape Consoritum, New York, NY, USA) to get visualization.

Assessment of immune cells

CIBERSORT is a deconvolution algorithm using the expression values of 547 genes to characterize the composition of immune cells in tissues. In this study, we used this algorithm to estimate the relative proportion of 22 infiltrating immune cell types based on gene expression. We uploaded the normalized gene expression data to the CIBERSORT website (http://cibersort.stanford.edu/) and set the algorithm to 1,000 rows. P< 0.05 was considered statistically significant [43].

Patients and specimens

Colon cancer specimens and healthy bowel tissues were collected from July 2017 to July 2019. Patients with the following criteria were excluded from participation: had received adjuvant chemotherapy or radiotherapy prior to surgery; had additional cancers diagnoses. All patients were classified according to the 7^th edition of the TNM staging system 23. Postoperative adjuvant therapies were performed, according to standard schedules and doses. All participants gave their written informed consent. This study was approved by the Ethical Committee of Shanghai Pudong Hospital.

Immunohistochemical (IHC) staining

IHC staining was carried out according to the manufacturer’s instructions. Briefly, formalin-fixed and paraffin-embedded tissue sections were deparaffinized in xylene and hydrated with decreasing concentrations of ethanol (100, 95, 80 and 75%). The slices were then soaked in 10% BSA to inhibit endogenous peroxidase activity and incubated with anti-PD-1 or anti-LAG3 rabbit polyclonal antibody (1:100; Cat: 86163&15372, Cell Signaling Tech, Boston, MA USA) at 4°C overnight. A horseradish peroxidase (HRP)-conjugated rabbit secondary antibody (1:400; Cat: 8114, Cell Signaling Tech) was added for 60 min at room temperature; then, 3,3’-diaminobenzidine development (DAB Substrate Chromogen System; Dako) and hematoxylin staining were performed according to standard protocols. Slides were fixed and images were obtained with an Olympus IX71 inverted microscope and a DP2-BSW Olympus image acquisition software system.

Cell culture, hypoxia induce and miRNA mimics transfection

Human colon cancer cell lines HCT116 was purchased from the University of Colorado Cancer Center Cell Bank. The cells were cultured in RPMI 1640 medium supplemented with 10% FBS (Invitrogen, Carlsbad, CA, USA) at 37°C in a 5% CO₂ atmosphere. For hypoxia induce, cells were treated with cobalt chloride (CoCl₂) which prevents HIF-1α from degeneration by replacing the prolyl hydroxylase (PHD) cofactor Fe²⁺ [44].

The NC mimics and miR-210 mimics were purchased from RiboBio (Shanghai, China). In accordance with the instructions of the product manual, Lipofectamine 3000 (Invitrogen, Inc.) was used to transfect the mimics.

Cell proliferation assay

3×10³ cells suspended in 100ul RPMI-1640 medium were seed into 96-well plate. The cell proliferation was assessed by the CCK8 (Dojindo Molecular Technologies, Japan). 10ul CCK8 solution was given to each well of the plate after different incubation times: 0h, 24h, 48h and 72h. Finally, we measured the absorbance at 450nm wavelength after 2h incubation.

Cell migration and invasion assays

Cell migration and invasion were analyzed with transwell plates (24-well insert, 8 μm pore size; BD Biosciences, Bedford, MA, USA). The filters (Corning Inc., USA) were coated with (invasion assay) or without (migration assay) 55 μL Matrigel (1:8 dilution; BD Biosciences). The 10⁴ cells were suspended in 100μl RPMI-1640 medium without serum and seeded in the upper chamber. Next, 600μl 90% RPMI-1640 supplement with 10% FBS was added to the bottom chamber. After incubation for 24h, the chambers were fixed by 4% paraformaldehyde for 30 min and then stained by 0.1% crystal violet for 30 minutes. At last, we used a magnification microscope to count the amount of the invasion cells in the bottom of the chamber.

RNA extraction, reverse transcription and quantitative PCR (RT-qPCR)

Total RNA was extracted by Trizol Regent (Invitrogen) from CRC cells. cDNA was obtained from total RNA with PrimeScript™ RT reagent kit (Takara Bio, Inc., Otsu, Japan). The mRNA expression was assessed by Real-time quantitative PCR, which was carried out in triplicate by a SYBR Premix Ex Taq™ kit (Takara Bio) and ABI 7900HT Real-Time PCR system (Applied Biosystems Life Technologies, Foster City, CA, USA). The primers used are shown in Supplementary Table 1. The comparative cycle threshold values (2-ΔΔCt) were adopted to analyze the final results.

Statistical analysis

All analyses were performed using SPSS 23.0 and R 3.5.3. All statistical tests were two-sided, and a P value <0.05 was considered statistically significant. Continuous variables that conformed to the normal distribution were compared with the use of an independent t test for comparison between groups, while continuous variables with skewed distribution were compared with the Mann-Whitney U test. Categorical data were tested using the Chi-square test. The correlation matrix was constructed by R software based on Pearson Correlation Coefficient (immune checkpoints) or Fisher exact test (gene mutations). The analysis of survival was performed through the Kaplan-Meier method, which was evaluated by the log-rank test. The univariate Cox regression model was used to analyze the effects of individual variables on survival, and the multivariate Cox regression model was used to discover the independent factors associated with survival.

Ethics approval

All procedures involving human participants were performed in accordance with Shanghai Pudong Hospital ethical committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. All patients provided their written informed consent. The study protocol was approved by the Pudong Hospital Committee on human research.