Transcriptomics and proteomics show that selenium affects inflammation, cytoskeleton, and cancer pathways in human rectal biopsies

Epidemiologic studies highlight the potential role of dietary selenium (Se) in colorectal cancer prevention. Our goal was to elucidate whether expression of factors crucial for colorectal homoeostasis is affected by physiologic differences in Se status. Using transcriptomics and proteomics followed by pathway analysis, we identified pathways affected by Se status in rectal biopsies from 22 healthy adults, including 11 controls with optimal status (mean plasma Se = 1.43 μM) and 11 subjects with suboptimal status (mean plasma Se = 0.86 μM). We observed that 254 genes and 26 proteins implicated in cancer (80%), immune function and inflammatory response (40%), cell growth and proliferation (70%), cellular movement, and cell death (50%) were differentially expressed between the 2 groups. Expression of 69 genes, including selenoproteins W1 and K, which are genes involved in cytoskeleton remodelling and transcription factor NFκB signaling, correlated significantly with Se status. Integrating proteomics and transcriptomics datasets revealed reduced inflammatory and immune responses and cytoskeleton remodelling in the suboptimal Se status group. This is the first study combining omics technologies to describe the impact of differences in Se status on colorectal expression patterns, revealing that suboptimal Se status could alter inflammatory signaling and cytoskeleton in human rectal mucosa and so influence cancer risk.—Méplan, C., Johnson, I. T., Polley, A. C. J., Cockell, S., Bradburn, D. M., Commane, D. M., Arasaradnam, R. P., Mulholland, F., Zupanic, A., Mathers, J. C., Hesketh, J. Transcriptomics and proteomics show that selenium affects inflammation, cytoskeleton, and cancer pathways in human rectal biopsies. FASEB J. 30, 2812‐2825 (2016). www.fasebj.org

Low intake of the dietary antioxidant micronutrient selenium (Se) is associated with increased risk of colorectal adenoma and of colorectal cancer (CRC) mortality (1); for example, in the double-blind, randomized, placebocontrolled Nutrition Prevention Cancer Trial, there was a significant (58%) reduction in CRC incidence in people receiving 200 mg supplementary Se in the form of Seenriched yeast/d (2), and the strongest effect of Se supplementation was observed in subjects within the lowest tertile for plasma Se at the start of the trial (,1.34 mM). In contrast, in the Women'sH e a l t hI n itiative and Selenium and Vitamin E Cancer Prevention Trial (SELECT) trials, no benefit of increased Se intake was observed for individuals with high Se intakes (3,4). Recent data from a European Prospective Investigation into Cancer and Nutrition (EPIC) cohort revealed that, in most European countries, plasma Se status is suboptimal and is associated with increased risk of CRC in women (5). Moreover, evidence from animal models showed that Se supplementation can prevent chemically induced colorectal carcinogenesis (6,7). Taken together, these data support a role for Se in reducing risk of CRC initiation and progression; ABBREVIATIONS: 2-D gel, 2-dimensional polyacrylamide gel electrophoresis; AQPs, aquaporines; CRC, colorectal cancer; EPIC, European Prospective Investigation into Cancer and Nutrition; FC, fold change; GPX, glutathione peroxidase; IEF, isoelectric focusing; IPA, Ingenuity Pathway Analysis; MALDI/ToF, matrix-assisted laser desorption ionization timeof-flight; MS, mass spectrometry; Se, selenium; SELK, selenoprotein K; SEPP, selenoprotein P; SEPW, selenoprotein W however, the intake at which Se has cancer-preventive properties and the mechanisms underlying its anticarcinogenic properties remain unclear.
The biologic actions of Se are thought to be mediated largely by selenoproteins in which Se is incorporated in the form of the amino acid selenocysteine. Functional genetic polymorphisms in selenoprotein genes (GPX4, SEPP1, SELS,andGPX1)modulateriskofCRCorinflammatory responses (8,9), in combination with Se status (9). Selenoproteins are crucial in the biochemical pathways essential for colorectal function, including redox control and responses to inflammatory, oxidative, and endoplasmic reticulum stress (10). For example, GPx4 expression was necessary for mitochondrial function and NFkBresponseto TNFa in a human colorectal adenoma cell model (11,12), and Se intake affected mRNA mTOR, TNFa,a n dN F -kB signaling pathways in the colon of mice fed a marginally Sedeficient diet (13). Thus, the protective effect of Se appears to be mediated through the role of selenoproteins in molecular pathways that help to maintain homeostasis when cells experience oxidative and inflammatory challenges (9). However, the effects of Se status on such pathways within the human colorectal epithelium remain poorly understood.
To identify key factors and pathways affected by Se status, we combined transcriptomics and proteomics in a comprehensive, unbiased analysis of rectal biopsy specimens from healthy participants. For the first time, we have shown in humans that differences in Se status within the physiologic range affect global gene expression patterns in the human rectum. We found that Se status affected factors implicated in inflammatory signaling, immune function, and cytoskeleton remodelling. These findings provide insights into the mechanisms through which Se may influence cancer risk and may help in the development of early biomarkers of disease risk.

Study population and sample collection
A subgroup of 22 adults who took part in the Biomarkers of Risk of Colorectal Cancer (BORICC) study were recruited from patients attending for routine flexible sigmoidoscopy at Wansbeck General Hospital (14), but shown to be free of colorectal disease. Ethics approval was granted by the Northumberland Local Research Ethics Committee, and research governance was obtained from the Northumberland Healthcare Trust (NLREC2/2001). Exclusion criteria included the presence or history of colonic inflammation, colorectal cancers or polyps, or a strong family history of colorectal cancer. Biopsies were performed in individuals who had no evidence of colorectal neoplasia or inflammatory bowel disease at endoscopy. Mucosal pinch biopsies were obtained from anatomically normal mucosa 10 cm from the rectal verge, snap frozen in liquid nitrogen for proteomic analysis or placed immediately in RNAlater (Thermo Fisher Scientific Life Sciences, Paisley, United Kingdom) for transcriptomic analyses, and stored at 280°C. Plasma Se concentration was measured by inductively coupled plasma mass spectrometry (MS). Briefly, RNAlater was drained from the biopsy specimens before they were submerged in 0.5 ml Trizol. After homogenizing the tissues on ice, an additional 0.5 ml Trizol was added, and samples were incubated for 5 min at room temperature. Samples were centrifuged (10 min, 12,000 g,4°C ) , supernatants were extracted for 3 min with 0.2 ml chloroform per 0.2 ml Trizol) and centrifuged (15 min, 12,000 g, 4°C), and the upper phase was precipitated in 70% EtOH prepared in diethylpyrocarbonate-treated milliQ water (EMD-Millipore, Billerica, MA, USA). Samples were purified on a PureLink RNA cartridge (Thermo Fisher Scientific Life Sciences) and eluted in RNase-DNase-free water. RNA integrity was determined by capillary electrophoresis with a Bioanalyzer Nano chip (Agilent Technologies, Stockport, United Kingdom) before microarray analysis, using whole-genome HumanHT-12, v3 single-color bead chips (Illumina, San Diego, CA, USA). Microarray results were confirmed by real-time PCR; however, because the biopsy specimens were very small, the amount of RNA was sufficient to confirm only the expression of 2 genes and a housekeeping control used to normalize the data.

Protein preparation and gel electrophoresis
Proteins were extracted from thawed tissue with a modified ReadyPrep Sequential Extraction Kit (Bio-Rad, Hemel Hempstead, United Kingdom), and the specimens were homogenized with a hand-operated micropestle (Eppendorf, Hamburg, Germany) before sonication in an ice-water bath for 10 min. The samples were centrifuged, and aliquots of supernatants were stored at 280°C. Proteins were separated by 2-D electrophoresis (15). Separation of proteins in the first dimension was performed by isoelectric focusing (IEF), with 24 cm immobilized pH gradient strips (pH 4-7)runonanEttanIPGphorbed(GEHealthcare, Pittsburgh, PA, USA) with a step-n-hold protocol of 500 V for 0.5 kVh; a gradient of 1000 V for 0.8 kVh; a gradient of 8000 V for 13.5 kVh; and a step-n-hold of 8000 V for 30.0 kVh at 20°C, giving a total of 44.8 kVh (8 h 49 min) with a rate-limiting factor of 50 mA. After completion of IEF, the strips were stored at 280°C. The 2dimension protein separation was carried out on 1 mm thick 10% Duracryl homogenous gels [30% acrylamide with 0.65% N,Nmethylene bis-acrylamide cross-linker, 1.5 M Tris (pH 8.7-9.0), 3.6 mM SDS, 0.5 ml/L tetramethylethylenediamine, and 11 mM ammonium persulfate, prepared in-house in 28 3 23 cm gel-plate cassettes]. Electrophoresis conditions were set to give an top voltage of 500 V, power of 20 W/gel, and a total run time of ;4.5 h. Gels were stained with SYPRO Ruby fluorescent stain (Bio-Rad) and imaged with a Pharos FX Plus molecular imager (Bio-Rad; 532 nm excitation laser and 605 nm emission filter at 100 mm resolution). Using Progenesis SameSpot software (Nonlinear Dynamics, Newcastle, United Kingdom), an experienced operator aligned the gel images to a single reference image, chosen for its overall quality and spot clarity, and protein spots were quantified. At this stage, staining artifacts were removed manually, and undetected double spots were corrected. Proteins of interest were extracted and analyzed by using matrix-assisted laser desorption/ionization time-of-flight (MALDI/ToF) MS. In-gel trypsin digestion was performed with a ProGest Protein Digester (Genomic Solutions, Ltd., London, United Kingdom). After preincubation, the digestions were carried out at 37°C for 3 h, using 50 ng sequencing-grade porcine trypsin (5 ml/well; Promega, Southampton, United Kingdom). The digests were analyzed with an Ultraflex MALDI-ToF/ToF mass spectrometer (Bruker Daltonics, Ltd., Coventry, United Kingdom). A 200 Hz nitrogen laser was used to desorb/ionize the matrix/analyte material, and ions were detected in positive ion reflectron mode. Peptide masses obtained from the MALDI-ToF analysis were searched against the Mass Spectrometry Data Base using the Mascot peptide mass-fingerprint program (Matrix Science, Ltd., London, United Kingdom). The results gave a probability-based Mowse score of 210 p log(P ), where P is the probability that the observed match is a random event. Under those parameters, protein scores ,63 were considered statistically significant (P , 0.05).

Transcriptomics and proteomics analysis
Raw gene expression data were analyzed with GeneSpring GX 11 (Agilent Technologies). Raw data were normalized with a quantile algorithm, and the baseline was transformed to the median of all samples. Those probes with a flag value of present or marginal in $80% of any of the experimental categories were selected for further analysis. The R package RankProd (Bioconductor; http://bioconductor.org/) was used to identify differentially expressed genes. Genes were defined as differentially expressed if the percentage of false positives in 1000 permutations of the Rank Products test was ,5% and the change in expression was .1.23 in either direction. Genes meeting these criteria were used to generate functional networks and pathway analyses with Ingenuity Pathways Analysis (IPA; Ingenuity Systems, Redwood City, CA, USA). Spearman's correlation tests were performed on the gene sets with a cutoff of 1.2-fold changes in either direction using SPSS Statistics, version 22.0 software (IBM, Armonk, NY, USA). IPA was used to generate functional networks and pathways for the gene sets for which expression correlated significantly with plasma Se status, and the proteomics dataset and the Comparison function in IPA were used to integrate the transcriptomic and proteomic datasets.

Se status of the participants
Participants were selected on the basis of their plasma Se status and divided into 2 groups that were discordant in plasma Se status, but matched for body mass index (24.7 6 0.88 and 24.1 6 0.72 in suboptimal and optimal Se groups, respectively), age (53.6 6 2.91 and 54.7 6 3.6, in suboptimal and optimal Se groups, respectively), and sex (5 males, 6 females in each group). The Third National Health and Nutrition Examination Survey estimated that the mean serum Se level in the U.S. population is 1.58 mM (16), a level considered to represent Se adequacy (17). In contrast, average plasma Se concentration in healthy British adults was estimated at 1.13 mM (18). The latter concentration is similar to those in other European countries and is considered suboptimal (5). In the present study, the group of participants with plasma Se concentrations similar to that in the U.S. population, having a mean 6 SEM plasma Se status concentration of 1.43 6 0.06 mM(range, 1.25-1.82), a value within the optimal range based on the hazard ratio for mortality (18) was designated as the optimal Se status group. This concentration is well below the Se concentrations (.2.22 mM) that have been associated with adverse health outcomes of high Se doses observed in supplementation trials (4,17,19). The suboptimal Se status group corresponds to individuals with a mean 6 SEM plasma Se status concentration of 0.86 6 0.01 mM(range, 0.79-0.92), a range that is suboptimal as defined by the hazard mortality ratio and that has been associated with increased risk of CRC in women in a large European population in the EPIC study (5).

Effects of Se status on gene expression profiles
Both RNA and proteins were extracted from individual rectal biopsies. After microarray analysis, differential gene expression analysis was determined using the RankProd package and a cutoff of 1.2-fold change showed that 254 genes were differentially expressed between the 2 Se groups, with 128 genes being up-regulated and 126 downregulated in the suboptimal Se group compared with the optimal Se group (for details see Transcript Profiling; http:// www.ncbi.nlm.nih.gov/geo/query/acc.cgi? token=epkzqgighfujhkh&acc=GSE70550).
Two selenoprotein genes (SEPW1 and SELK)w e r e down-regulated in the suboptimal Se group, but they were not among the top 25 down-regulated genes. Table 1 presents the top 25 genes up-and down-regulated in rectal tissue from the suboptimal Se group compared with the optimalSegroup.InthesuboptimalSegroup,waterchannel AQP8,s e r i n ep r o t e a s ei n h i b i t o rSPINK4, antiapoptotic factor OLFM4, chloride channel CLCA4, carbonic anhydrase CA4, and carcinoembryonic antigens CEACAM6 and -7 exhibited the greatest up-regulation. Similarly, other family members (AQP11, SPINK5,andCEACAM1 and -5)werealso upregulated in the suboptimal Se group. In contrast, HLA-A29.1, IL1B, the transcription factor CHURC1, antiapoptotic factors PHDLA1 and IER3, tumor suppressor ERRFI1/Mig6, and microRNA MIR-221 showed the strongest downregulation in the suboptimal Se group compared with the optimal Se group. Real-time PCR confirmed the microarray results, with significant correlation observed between microarray and real-time-PCR expression for MFF (P =0.005)andERRFl1 (P = 0.033) (data not shown).
Ingenuity Pathway Analysis (IPA) was performed on the microarray dataset to identify biologic functions, canonical pathways, and networks modulated by Se status and to predict potential upstream regulators. A large proportion of genes differentially expressed between the 2 Se groups belonged to cancer [204/254 genes (80%)], gastrointestinal diseases (58%), and inflammatory diseases (39%) categories (Tables 2 and 3).
The cell and molecular functions most affected by Se status included cellular growth and proliferation, cellular movement and development, cell death and survival, and cell-to-cell signaling and interaction ( Table 2). The 5 most significantly affected canonical pathways corresponded to pathways involved in inflammatory and immune signaling, suggesting an overall down-regulation of immune response in individuals with suboptimal Se status. IPA generated 25 networks, among which 19 had at least 10 focus molecules (data not shown), with the 5 most significantly affected networks presented in Table 3. As a result of these changes, IPA predicted that PDGFBB (P = 2.46E 236 , regulating 114 genes/254), TNFa (P = 3.58E 230 , regulating 145 genes/254), and IL1b (P = 3.31E 225 , regulating 128 genes/254) are likely to be the most important upstream regulators affected by Se status ( Table 4). In addition, most of the identified upstream regulators (23 of the top 25) are key to immune and inflammatory responses, and the observed changes indicate that this response is inhibited by lower Se status. Crosstalk  IPA was used to identify biologic functions, canonical pathways, and networks in which mRNAs and proteins are differentially expressed between the suboptimal and optimal Se groups. The 5 most significantly enriched molecular and cellular functions, physiological system functions, and canonical pathways are indicated, together with the corresponding P value and number of molecules (n) involved in the datasets. Functions and pathways affected in both datasets are indicated in italics and indented under the 5 most affected canonical pathways belonging to the immune function category.
between NFkB, TNFa, PDGFBB, and IL1B signaling pathways (Fig. 3) and between NFkB and other transcription factors, such as p53 and STAT3 (Table 4), has been described (20). Overall, these results indicate that genes involved in immune and inflammatory responses, cell signaling, cell proliferation, tissue and cell morphology, and cellular movement are differentially expressed in response to Se status.

Correlation of gene expression with Se status
RankProd identified genes that exhibit the highest change in expression between the 2 Se groups, but did not define the degree to which gene expression correlated with Se status. Thus, to further investigate the relationship between Se status and gene expression, Spearman'sc o r r elation analyses were performed on expression data from Cancer, organismal injury and abnormalities, reproductive system disease 33 25 Cellular development, cellular growth and proliferation, cell death and survival 31 24 The most significantly enriched networks are indicated with the z score and number of focus molecules. Networks affected in both microarray and proteomics datasets are indicated in italics. IPA upstream transcriptional regulator analysis was performed to identify the cascade of potential upstream regulators that can explain the observed gene expression changes in the microarray dataset. The most likely regulators are presented, together with the type of molecule, the number of known targets for each upstream regulator present in the dataset, the predicted z score activation state, the P value of overlap between genes targeted by each regulator in the dataset, and the genes known to be a target for each upstream regulator. all participants for the 254 genes modulated by Se status, using plasma Se concentration as a continuous variable. Expression of 69 genes (recognized by 75 different gene probes) correlated significantly with Se status, of which 34 correlated positively and 35 correlated negatively (Supplemental Table S1 and Fig. 1). Expression of the 2 selenoprotein genes SELW and SELK, identified as being down-regulated in the suboptimal Se group, showed a strong positive correlation with Se status (r Selw =0 . 6 3 7 ; P =0 . 0 0 1 ,a n dr Selk = 0.474; P = 0.026). Expression of DNAJC12, a member of the DnaJ/Hsp40 family acting as a cochaperone with Hsp70 and involved in protein folding and export (21); antisense RNA SNHG6;a n dSEPW1 showed the strongest positive correlations with Se concentration. Hsp70 is involved in nonsense-mediated decay (22) and has been involved in the control of selenoprotein mRNA degradation in conditions of low Se status (23). In contrast, expression of acyl-coenzyme A oxidase ACOX1,t h ef i r s t enzyme in peroxisomal fatty acid b-oxidation, strongly negatively correlated with Se status. Because ACOX1 activates NFkB (24), negative correlation of ACOX1 expression with Se status could contribute to the observed NFkBi nhibition. Moreover, NUAK2, which is regulated by NFkB and is involved in actin cytoskeleton remodelling (25), and ACTB, ALDOA, SLC9A3R1, MYH9, DYNLRB1,andTUBAL3, which are implicated in cytoskeleton organization (26,27), had expression that correlated strongly with Se status.
Subsequently, pathway analysis of the 69 genes for which expression correlated strongly with Se status revealed that the most affected diseases and disorders correspond to the cancer and cancer and tumorigenesis subcategories (34 molecules/69; 49%; P = 1.71E 205 -4.93E 202 ) and to inflammatory response (18 molecules/69; 26%; P = 4.88E 205 -4.67E 202 ). Three main networks identified were: 1) cancer, organismal injury, and abnormalities, reproductive system disease (21 focus molecules, z score 45), centered around NFkB; 2) cellular function and maintenance carbohydrate metabolism, small-molecule biochemistry (16 focus molecules, z score 31), centered around TNF; and 3)hematologicdisease, cellular growth and proliferation, cellular development (9 focus molecules, z score 15) centered around TP53 (Supplemental Fig. S1). Supporting these observations, inhibition of sequestosome 1 (SQSTM1), which regulates TNF-driven activation of NFkB, was identified as the most significant upstream regulator (P =2 . 5 0 E 206 ), followed by PLK2 and IL13,al soin vo lv edinNFkB regulation, the second and third most significant (P = 4.09E 205 and P =4 . 6 1 E 205 , respectively). Overall, these results indicate that genes whose expression correlate with Se status play a key role in inflammatory pathways, with a central role for NFkBandTNF.

Effects of Se status on proteomics profiles
Proteomic profiles from whole rectal biopsies were characterized by 2-D gel electrophoresis, followed by in-gel trypsin digestion and MALDI-ToF MS. To limit the bias normally associated with proteomic analysis when the analysis is restricted toward the more abundant proteins, analysis was performed on all 220 spots that were identified by MS. Of these 220, 36 spots, corresponding to 26 individual proteins, were differentially expressed between the suboptimal Se and optimal Se status groups ( Table 5) Table S1. cytoskeletal organization, SS18, S100A9, SFN, and HSP90B, exhibited differential expression between the 2 Se groups. These observations suggest that cytoskeleton remodelling is a major signature of Se effects in rectal tissue. All cytoskeletal proteins and associated factors were up-regulated in the suboptimal Se group compared with the optimal Se group, apart from 1 spot corresponding to cytokeratin 8 and S100A9.
Additional proteins differentially expressed between the 2 Se groups (Table 5) include: 1) proteins involved in immune and inflammatory response (tumor rejection antigen GP96 (HSP90B1), complement component C3, glycyl-tRNA synthetase (28) and calcium-binding protein S100A9); 2) proteins involved in lipid metabolism and disruption of cholesterol homeostasis known to be linked to inflammatory processes (29) and to affect signal transduction and membrane trafficking (29) (FABP5, apolipoprotein A1, StAR-related lipid transfer protein); and 3) proteins involved in antioxidant mechanisms, including glutathione S-transferase v 1 (GSTO1), S100A9, and albumin. The lowered expression of these genes in the suboptimal Se group not only suggests a reduced immunity and capacity to protect against oxidative damage, but is compatible with the links between glutathione and Se metabolism. Moreover, the cytoskeleton, a key mediator of immune and inflammatory processes (30,31), is highly sensitive to oxidative stress (32).
IPA analysis of the proteomic data also predicted activation of cancer pathways (z score activation, 2.575; P = 2.89E 209 ; focus molecules, 21) in individuals with suboptimal Se status. Consistent with observations from RNA expression analysis by microarray, immunologic and inflammatory responses biologic functions and networks, together with cell death and survival, cellular movement and development, cell growth and differentiation, and hematologic system and molecular transport ( Table 2) were among the most affected by Se status. IPA identified 3 networks, with network 1, centered around NFkB (focus molecules, 16; z score 38), representing the most important one(Table3).

Integrated transcriptomics and proteomics
Tables 2 and 3 summarize the pathways and networks altered by Se status in transcriptomic and proteomic datasets. To integrate the 2 approaches, the datasets were combined, and IPA was used to identify biologic processes or pathways in which features from both transcripts and proteins are enriched. Twenty-eight networks were identified, and the interconnection between proteomic and transcriptomic networks was analyzed. Proteomic network 1 is connected to 8 networks from the microarray dataset, and proteomic network 2 is connected to 6 microarray networks. Merging proteomic network 1 (p 1) with individual connected microarray networks (4-7a n d1 0m )r evealed NFkB to be the central node of these merged networks (Fig. 2) and IL1, TNF, and Akt to be the central nodes of merged networks 1-1, 1-18, and 1-12, respectively (data not shown). On the other hand, the proto-oncogenes c-Myc,c -Fos and b-catenin were identified as central nodes from merging proteomic network 2, with connected microarray networks 2 and 8, 4, and 6 (data not shown).

DISCUSSION
This study is the first to integrate transcriptomic and proteomic approaches for assessing the effects of physiologic differences in status of the dietary antioxidant Se on events in the large-bowel mucosa in healthy individuals. Using this approach, we report that the expression of factors implicated in inflammatory signaling, immune function, and cytoskeleton remodelling is altered in individuals with suboptimal Se status in healthy human rectal mucosa,  providing novel insights into the mechanisms through which Se may influence cancer risk and colorectal function. Crosstalk between thesepathwaysisfundamental to the maintenance of gut homeostasis through a partnership among the gut epithelium, immune cells, and responses to bacteria (33). Among the genes and proteins exhibiting the strongest change in expression between the 2 Se groups (Tables 1 and 5), many have been shown to have altered expression in colorectal adenomas.
As expected differences in selenoprotein mRNA expression were observed between the 2 groups, w i t hs i g n i f i c a n tr e d u c t i o no fb o t hs e l e n o p r o t e i nW [fold change (FC) = 21.404] and selenoprotein K (FC = 21.286) in the suboptimal group (Gene Expression Omnibus repository, GSE70550; http://www.ncbi.nlm.nih. gov/geo/). Moreover, expression of both SEPW1 and SELK correlated strongly with Se status (Supplemental Table S1), consistent with the responsiveness of selenoproteins to Se in the colon. It is known that identification of selenoproteins using MS is problematic, and it is therefore not surprising that our proteomic analysis did not identify specific selenoproteins.
We observed significant effects of Se status on immune and inflammatory signaling in both the microarray and proteomics datasets (Table 2), with approximately a third of identified genes (80/254) and proteins (7/26) involved in immune cell trafficking, with most canonical pathways associated with inflammation, and with inflammatory and immune diseases and inflammatory This model illustrates how the observed changes in gene expression and protein levels affect inflammatory and immune signaling, cytoskeletal remodelling and apoptotic/survival pathways, and how these changes and crosstalk between these pathways have the potential to contribute to tissue dysplasia in the rectum of an individual with suboptimal Se status. In the healthy colorectal epithelium, NFkB plays a critical role of coordinator, regulating: 1) cytokine and chemokine production (blue circle) in response to the gut bacteria (blue rectangles), 2) the recruitment of immune cells (red, yellow, and beige) to the epithelium to ensure immuno-surveillance, and 3) the regulation of the balance between survival and apoptotic factors. We propose that NFkB inhibition, as a result of suboptimal Se status, reduces immune and inflammatory signaling and induces cytoskeleton remodelling (changes associated with suboptimal Se are indicated in red and dotted lines). Cytoskeletal alteration and imbalance of apoptotic/survival signals could result in epithelial barrier damage and changes in cell and tissue morphology and cell movement. Moreover, the combined alteration of the epithelial barrier, reduced expression of IL1B and IL8, and reduced immunity could favor bacterial invasion, leading to chronic inflammation.
impact on the colonic epithelium as it undergoes normal cycles of proliferation, migration, and differentiation. Indeed, previous work has found altered expression of cytoskeletal components in CRC; desmoglein 2 and cytokeratins KRT18, KRT6A, and KRT8 were identified as signatures of CRC subtypes (45), high expression of desmin was associated with CRC and cytokeratins 8, 10, 18, and 19 and b-actin were up-regulated in CRC tumors (46). In addition, previous work has indicated a role for the cytoskeleton in mediating the immune and inflammatory responses (30,31).
In the present study, both RNA and proteins were extracted from rectal biopsy tissues, which are complex with multiple cell types and at different stages of differentiation. Therefore the results reflect changes in protein and RNA profiling from diverse cell types of both epithelial and immune origin, and changes in the proportions of these 2 major cell types may account for some of the observed effects. However, we have no reason to believe that there are any systematic differences in cellularity of the biopsies between those with normal and suboptimal Se status, and this complexity takes into account the different cellular actors present in the rectal mucosa.
Overall, the present data indicate that, in the macroscopically normal rectal mucosa, suboptimal Se status is associated with alterations in cytoskeleton remodelling and reduced inflammatory and immune signaling capacity. This finding suggests that the inflammatory response and the capacity to recruit immune cells are reduced in the rectum of individuals with suboptimal Se status. Colorectal cancer is characterized by derangements in immune and inflammatory signaling and cytoskeleton alterations. We hypothesize that suboptimal Se status has the potential to compromise an individual'sr e d o xc a p a c i t y and ability to mount appropriate immune and inflammatory responses to exogenous, physiologic, or microbiological stressors and, as a result, could favor the development of cancer (Fig. 3). In support of this hypothesis, cancer pathways were identified as the most affected pathways in both proteomic and transcriptomic datasets. Recent evidence from genetic association studies has led to the hypothesis of a convergence of hormones, inflammation, and energyrelated factors (CHIEF) pathway in the etiology of CRC (29). We propose that the convergence of reduced immunity and inflammatory response and cytoskeleton remodelling associated with reduced Se status would increase CRC risk.