Diet-induced weight loss reduces colorectal inflammation: implications for colorectal carcinogenesis¹²³

Subjects

For this unblinded study, 10 healthy, obese [average (±SD) BMI (in kg/m²): 34.8 ± 3.5], premenopausal women (mean age: 43.2 ± 8.3 y) were recruited from the community through advertisements. Study volunteers included 6 whites, 1 Hispanic, 2 African Americans, and 1 subject with a mixed racial background (Table 1). All subjects underwent a complete medical examination, standard blood and urine tests, and an electrocardiogram. Subjects were weight stable (<2% change) for ≥6 mo before the study. Exclusion criteria included a history of cancer, current treatment of weight loss, previous intestinal surgery, a history suggestive of malabsorption, major medical problems, consumption of medications with antiinflammatory properties (eg, nonsteroidal antiinflammatory drugs, statins, glitazones, and COX-2 inhibitors) or medications that could increase risk of complications associated with consuming a VLCD (eg, lithium), an allergy to soy products or aspartame, or contraindications to undergoing a sigmoidoscopy and biopsy, including any history of excessive bleeding. The study was approved by the Institutional Review Board of The Rockefeller University (New York, NY). Subjects were recruited between September 2007 and May 2008. Written informed consent was obtained from each subject before participation in the study.

Experimental design

Seven subjects chose to complete the weight-loss phase of the study in The Rockefeller University Hospital inpatient unit, and 3 subjects were closely monitored as outpatients. After screening, subjects underwent baseline testing, which included fasting blood determinations, urine analysis, and flexible sigmoidoscopy with multiple mucosal biopsies. Subjects started a VLCD weight-loss program and were monitored weekly by the study physicians. The VLCD was continued until subjects had lost ≥8% of baseline weight, at which time the initial measurements were repeated. This degree of weight loss was chosen because a previous study showed that this would lower rectosigmoid proliferation in obese subjects by ≈40% (21). Subjects were counseled that the study required them to maintain their usual physical activity, and this was emphasized throughout the study.

Subjects consumed a reduced-calorie, Western-style diet that contained 50% of their calculated baseline caloric intake for 3 d before initiation of the VLCD. Subjects were fed a commercially available VLCD product (New Direction Program; Robard Corporation, Mount Laurel, NJ) that provided ≈782 kcal/d with an energy distribution of carbohydrate, fat, and protein of 25%, 25%, and 50%, respectively, and daily requirements of vitamins and minerals (Table 2). This VLCD provided a choice of shakes, soups, bars, and pudding in different flavors; subjects had 4 choices/d, consumed one item every 4 h, and drank ≥2 L water daily. Inpatients were monitored daily by the hospital staff, and outpatients were monitored weekly during clinic visits. Weekly blood tests for electrolytes and renal function tests were performed to monitor for complications that can occur with VLCD consumption.

Procedures and sample collection

Fasting electrolytes, liver function tests, renal function tests, lipid profiles, and high-sensitivity C-reactive protein measurements were performed by the clinical chemistry laboratory at Memorial Sloan-Kettering Cancer Center. Aliquots of serum samples for cytokine measurements were stored at −80°C. Flexible proctosigmoidoscopy was performed between 0800 and 0900 after a 60-mL tap-water enema, and 12–14 rectosigmoid biopsies were obtained from 4 quadrants between 10 and 20 cm from the anal verge. Four biopsies for mucosal cytokine measurements were immediately frozen in liquid nitrogen; 6 biopsies for gene expression analysis were immersed in RNAlater (catalog no. 7024; Ambion) at 4°C for 24 h and frozen in liquid nitrogen. The remaining biopsies were frozen in optimum cutting temperature compound (OTC; International Medical Equipment, Chico, CA) and stored at −80°C for immunohistochemical analysis.

Sample analyses

Immunohistochemistry

Six-micron sections were cut at −20°C onto plus slides (Fisher-Scientific, Waltham, MA) and stored at −80°C until assay. For immunohistochemical analyses, sections were quenched in ice-cold acetone and rehydrated in phosphate buffered saline (PBS). Sections were treated with purified mouse anti-human antibodies to cluster of differentiation (CD) 3 (T cell marker) at a 1:200 dilution (catalog no. 34740; BD Pharminogen, Franklin Lakes, NJ) and to CD163 (macrophage marker) at a 1:500 dilution (catalog no. BM4041; Novus Biologics, Littleton, CO). The secondary antibody, a biotin-labeled horse anti-mouse antibody at a 1:200 dilution was applied and followed by amplification with an avidin-biotin complex (1:100 of avidin and 1:100 of biotin; catalog no. 6102; Vector Laboratories, Burlingame, CA), and the signal was developed with chromogen-3-amino-9-ethylcarbazole (Sigma-Aldrich). The area of the sections and number of positive cells contained were counted manually with the National Institutes of Health Image J software (NIH IMAGE J 1.42; National Institutes of Health, Bethesda, MD) and reported as the number of positive cells per square millimeter. For each subject before and after the weight-loss regimen, all cells were counted in 3 separate sections >100 μm apart at a 10× magnification. Each section was analyzed in a random fashion by a single person blinded to their identity. A subset of sections were analyzed ≥3 times in a similar fashion, and the CV was <10%. Mucosal morphology was assessed by hematoxylin (Fisher Scientific, Waltham, MA) and eosin (Shandon, Pittsburg, PA) staining of sections by using a standard protocol. Data from only 8 subjects are presented for CD163 staining because the staining quality was inadequate for analysis in 2 subjects.

Gene expression

Total RNA was extracted from 2 rectosigmoid biopsy specimens by using the Trizol method (Invitrogen, Carlsbad, CA). RNA purification, quality assessment, and hybridization were performed as previously described (23). Biotin-labeled complementary RNA was hybridized onto Human HT-12v3 Ilumina expression chips (Illumina Inc, San Diego, CA). Samples from baseline and post–weight loss for each subject were analyzed on the same chip. The arrays were scanned by the Illumina BeadStation laser scanning imaging system (Illumina).

For determinations of real-time polymerase chain reactions (PCRs), 2 μg total RNA was used as a template for complementary DNA synthesis with a SuperScript-III First-Strand kit (Invitrogen). Complementary DNA was diluted in water, and amounts corresponding to 50 ng original RNA were used for quantitative gene expression by real-time PCR. SYBR Green PCR master mix (Applied Biosystems, Foster City, CA), primers from Sigma-Aldrich, and the ABI Prism7900 real-time PCR system (Applied Biosystems, Carlsbad, CA) were used for the real-time PCR. Six down-regulated and 4 up-regulated genes were assayed, which were normalized to the housekeeping gene GAPDH to adjust for the sample loading and efficiency of the reaction. Real-time PCR samples were assessed in duplicate, and messenger RNA concentrations were expressed as the log ratio of relative expression concentrations before and after treatment (see supplemental Table 1 under “Supplemental data” in the online issue for a list of real-time PCRs).

Analysis of the gene-expression microarray data

Gene-expression data from the microarrays were analyzed with the R software (version 2.1) and available packages from Bioconductor (http://www.bioconductor.org). Data were normalized by using robust spline normalization functions from the lumi package (Bioconductor). A classical quality control was performed. Expression concentrations with detection P values ≤ 0.05 were considered detectable. Probes with detectable expression in at least one sample and with SD >0.1 were used in further analysis. Expression values (log₂ transformed) were modeled with the package limma (Bioconductor). Moderated paired t tests were used to test the significance of the effect of weight loss. P values were adjusted for multiple comparisons by using the Benjamini-Hochberg procedure. Genes with fold-changes ≥1.2 and P ≤ 0.05 were considered differentially expressed. Heat maps were generated by hierarchical clustering by using one Pearson r as the distance measure and complete linkage as the agglomeration procedure. The gene-expression data were deposited in the public domain (transcription profiling accession no. GSE20931).

Gene set enrichment analysis

Gene set enrichment analysis (GSEA) was conducted with the GSEA desktop application (http://www.broadinstitute.org/gsea) (24). Our analysis of gene expression before and after weight loss in each subject was performed in a paired fashion, and genes were ranked by their effect size (ie, the mean fold change for each subject). This rank-order list was used as the input to the GSEA. For each gene set, the enrichment score (ES) was calculated by using weighted Kolmogorov-Smirnov statistics to measure the proximity of the gene set to the top of the weight-loss effect ranked list. A high and positive ES indicated that the gene set or pathway was collectively up-regulated by the weight-loss intervention, whereas a negative ES indicated down-regulation. The significance (P value) of the observed ES was calculated by using simulations. Normalized ESs were used to compare analysis results across gene sets. The molecular signature database available at the Broad Institute, MIT (http://www.broadinstitute.org/gsea/msigdb), and gene sets created at the Rockefeller University from the data generated by IL-17 stimulation of keratinocytes (25, 26) were used to query the data. We used sets from 3 different collections from the GSEA website [ie, the curated canonical pathway database (C2), the Gene Ontology database (C5), and the transcription factor database (C3)]. Gene sets that were positively or negatively enriched with P < 0.05 and a false discovery rate (FDR) <5% were considered significant, except for the transcription factor gene sets for which an FDR <10% was considered significant.

Statistical analyses

Two-tailed paired t tests were used to compare anthropometric measurements, biochemical variables, serum and mucosal cytokine concentrations, and CD3 and CD163 cell counts at baseline and after weight loss. P ≤ 0.05 was considered significant. S-Plus (S+) version 8.1 for Windows software package (TIBCO Software Inc, Somerville, MA) was used for statistical analyses.

Anthropometric and biochemical measurements

All 10 subjects successfully completed the study without significant adverse events in a mean of 46.5 ± 9.3 d. The VLCD resulted in a mean (±SD) decrease in body weight of 10.1 ± 1% (P < 0.01), decreased BMI of 10.2 ± 1.4% (P < 0.01), and reduction in waist circumference of 8.3 ± 4%, (P < 0.01). The weight-loss regimen also induced improvements in systolic and diastolic blood pressures, fasting blood sugar concentrations, total and LDL serum cholesterol and triglyceride concentrations, and total white blood cell counts (Table 3). Serum electrolyte concentrations, liver function tests, and renal function tests were not significantly altered.

Serum and mucosal cytokines

Diet-induced weight loss resulted in mean reductions in serum TNF-α concentrations of 15% and in IL-8 concentrations of 30% (both P < 0.05). There was also a trend toward decreased serum contents by 9% (P = 0.097) (Figure 1) but no change in concentrations of IL-6 and high-sensitivity C-reactive protein. Baseline serum MCP-1 concentrations were similar to those previously reported in obese individuals (27). In the rectosigmoid mucosa, concentrations of mucosal TNF-α decreased by 57% (P < 0.05), concentrations of IL-8 decreased by 44% (P < 0.05), concentrations of IL-1β decreased by 38% (P < 0.05), and concentrations of MCP-1 decreased by 25% (P < 0.05), and there was a trend toward decreased IL-6 concentrations by 45% (P = 0.06) (Figure 2). Data from one subject were excluded from the analysis because concentrations of all cytokines from her biopsies were >3 SDs from the mean of all subjects.

Immunohistochemistry

Standard hematoxylin and eosin staining of mucosal biopsies at the onset and end of the study showed no identifiable changes in total cellular content or mucosal morphology (data not shown). However, immunohistochemical staining showed a 28% decrease in CD3-positive T lymphocytes (P < 0.01) (n = 10) and a 42% decrease in CD163-positive macrophages (P < 0.05) (n = 8) (Figure 3) at the end of the study.

Microarray analyses

Microarray analyses revealed that diet-induced weight loss was accompanied by 1211 differentially expressed genes (with P ≤ 0.05) of which 70 genes (56 down-regulated and 14 up-regulated genes) had a fold-change ≥1.2. This differential pattern was strikingly shown in the heat map in Figure 4 (see supplemental Table 2 under “Supplemental data” in the online issue for the differential pattern shown in each gene). Genes involved in inflammation and cancer were down-regulated, including many genes associated with cytokines and chemokine signaling such as IL-8, CCL-21, CCL-19, and CCL4L1, as well as the proto-oncogenes JUN and FOS. Genes coding for molecules for which the circulating concentrations decreased with weight loss also were down-regulated, including peptide YY (PYY), vasoactive intestinal peptide (VIP), and somatostatin (SST).Genes that coded for gastrointestinal substrate transport, including ATP-binding cassette, subfamily A members 8 and 9 (ABCA8 and ABCA9), and carnitine palmitoyltransferase-I (CPT-1A) were up-regulated. Six down-regulated and 4 up-regulated genes were identified from the differentially expressed gene list for validation by real-time PCR, and gene-fold changes confirmed the microarray data findings (see supplemental Table 3 under “Supplemental data” in the online issue).

Detecting changes in gene pathways by altered expression of single genes relies on relatively large changes and may be insensitive to many physiologically significant changes of lesser magnitude. A more sensitive approach has been GSEA, which examines many genes in a pathway for changes that may not, by themselves, be significant but are significant in the aggregate (28). In our analyses, we used gene sets curated in the canonical pathway database, in the Gene Ontology database, IL-17 pathway gene sets, and the transcription factor database from the GSEA website. In the canonical pathway database, 74 gene sets were down-regulated, and 24 gene sets were up-regulated (see supplemental Tables 4 and 5 under “Supplemental data” in the online issue); in the Gene Ontology database, 55 gene sets were down-regulated, and none were up-regulated (see supplemental Table 6 under “Supplemental data” in the online issue); and in the IL-17 gene sets, 6 gene sets were down-regulated and 2 gene sets were up-regulated (see supplemental Table 7 under “Supplemental data” in the online issue). Gene sets related to inflammation and cancer that were down-regulated are highlighted in Table 4 and include pathways activated by TNF-α, IL-6, IL-1, IL-17, chemokine and cytokine signaling pathways, prostaglandin synthesis pathways, and pathways involved in oxidative stress, glutathione metabolism, and apoptosis. Up-regulated mucosal gene pathways were involved in carbohydrate metabolism (see supplemental Table 5 under “Supplemental data” in the online issue). The reduction in oxidative stress pathways was supported by down-regulation of downstream gene targets FOS and JUN.

Importantly, the analysis of changes in transcription factor genes identified 38 down-regulated and 9 up-regulated transcription factor gene sets (see supplemental Tables 8 and 9 under “Supplemental data” in the online issue). These gene sets included down-regulated transcription factor gene sets for STAT3, STAT5A, STAT5B, NF-κB, ATF, CREB, and multiple serum response factors, which are recognized as very important in colorectal inflammation, proliferation, apoptosis, and oncogenesis, which are shown in Table 5.