Interaction of Fatty Acid Genotype and Diet on Changes in Colonic Fatty Acids in a Mediterranean Diet Intervention Study

Many studies have suggested that a Mediterranean diet, as compared with a typical Western diet, may decrease the risk of various chronic diseases including colorectal cancer (1, 2). Rates of colorectal cancer were very low in Greece and have increased as diet has drifted away from the traditional eating pattern (3). The traditional Greek diet, relative to a Western diet, had lower intakes of n-6 polyunsaturated fatty acids (PUFA) and red meat, but higher intakes of plant-based foods, fish, and monounsaturated fatty acids (MUFA) chiefly from olive oil (2). The fat content of the Mediterranean diet is of particular interest for colon cancer prevention, as in intervention studies, increasing fiber alone does not seem to be preventive, and increased intakes of fruit and vegetables have had modest preventive effects (4–6). In particular, we hypothesized that lower intakes of n-6 linoleic acid and higher intakes of n-3 fatty acids have implications for preventing colon cancer as n-6 fatty acids are metabolized to eicosanoids such as prostaglandin E2 (PGE2) that are proinflammatory in the colon (7). PGE2 is formed from arachidonic acid (20:4 n-6) by COXs in the colonic mucosa, and it plays an important role in colonic crypt cellular expansion and subsequent adenoma formation (8).

In addition to the possible effects of dietary intakes, genetic variation in fatty acid desaturase (FADS) genes has been shown to influence serum and tissue arachidonic acid concentrations (9–15). Delta-5 desaturase (FADS1) and delta-6 desaturase (FADS2) are key desaturase enzymes involved in the synthesis of arachidonic acid and eicosapentaenoic acid (EPA; 20:5, n-3) from 18 carbon precursor fatty acids. Dietary intake of arachidonic acid is low in humans; however, arachidonic acid comprises between 5% and 10% of the phospholipids in cells due to elongation and desaturation of linoleic acid (18:2 n-6) to arachidonic acid (16).

Polymorphisms in the FADS1 and FADS2 genes have been identified, and these significantly affect PUFA concentrations in serum. The minor alleles are associated with lower desaturase activity and lower concentrations of arachidonic acid in blood (9–15). Analogous associations for EPA and docosahexaenoic acid (DHA) have not been consistent across studies, perhaps because certain types of fish can supply high amounts of preformed EPA and DHA. Dietary intakes are important to consider as conversion of dietary linolenic acid to longer chain n-3 fatty acids competes with the analogous process for n-6 fatty acids (17). In addition to diet, desaturase activity seems to be important in cardiovascular health, and presence of the minor allele in FADS1/2 has been associated with improved measures of blood lipids, C-reactive protein, insulin, and fasting glucose (18–21.) This indicates that lower arachidonic acid levels are associated with lower proinflammatory states. The prevalence of minor alleles seems to have evolved in response to Western diets that are plentiful in n-6 fatty acids, and they are more prevalent in persons of European descent than of African descent (11, 22).

Much less research is available on how FADS polymorphisms might affect changes in fatty acids in response to changes in diet, and the available studies have generally focused on n-3 fatty acid supplementation. Flaxseed supplementation, which provides linolenic acid (18:3, n-3), was less effective in increasing EPA concentrations in minor allele carriers of either FADS1 or FADS2, resulting in significant diet by genotype interactions on plasma concentrations of EPA and arachidonic acid (23). Dietary n-3 fatty acids also may interact with FADS genotype in affecting concentrations of blood cholesterol and triglycerides, with significant beneficial effects for carriers of all minor alleles being found in some but not all studies (20, 24–26).

The goal of this present study was to assess potential interactions of polymorphisms in FADS1 and FADS2 with changes in diet on levels of arachidonic acid and EPA in the serum and in the colonic mucosa of persons at increased risk for colon cancer. This was a secondary analysis of a randomized clinical trial that evaluated changes in fatty acids and carotenoids elicited by 6 months of intervention with either a Mediterranean or a standard Healthy Eating diet. In that study, we observed that dietary changes had little effect on colon fatty acids, which led to the hypothesis that metabolic factors may be limiting for changes in fatty acids (27). The randomized study obtained both blood and colon biopsies. Here, the relationships of FADS polymorphisms with serum and colonic fatty acid concentrations were evaluated at baseline and after 6 months of dietary intervention.

Details of recruitment and conduct of the Healthy Eating for Colon Cancer Prevention Study have been published previously (27, 28). The study was approved by the University of Michigan Medical Internal Review Board and was registered at the ClinicalTrials.org (NCT00475722). Briefly, 120 individuals at increased risk of colon cancer gave informed consent and were randomized to follow a modified Mediterranean diet or to Healthy People 2010 diet for 6 months. Blood and colonic mucosal tissue samples were collected at baseline and at 6 months by flexible sigmoidoscopy without prior preparation of the bowels. Blood was drawn after an overnight fast. At baseline, a Health Status Questionnaire was filled out by participants that included health and demographic data. Health information was asked again at 6 months. Dietary data were collected at 0 and 6 months using 2 days of food records and two 24-hour recalls.

The decision to genotype subjects with regard to FADS was made after the study began, and consent for genotyping could not be obtained from 9 individuals, 2 of whom completed 6 months of study and 7 of whom had dropped out after enrolling. Three samples were not genotyped successfully. The present analysis therefore included 108 of 120 subjects enrolled in the study and randomized to 6 months of counseling for either a Mediterranean or a Healthy Eating diet.

The frequency of counseling sessions was the same in both study arms. The Healthy Eating diet had dietary goals based on the Healthy People 2010 diet. The goals were to include two servings/day of fruit, three servings/day of vegetables with at least one of those servings being dark green or orange, six servings/day of grains with at least three from whole grains, less than 10% of calories from saturated fat (SFA) and less than 30% of calories from total fat. The Mediterranean diet had goals for consumption of high n-3 foods, such as fish or flax at least two times a week, consumption of foods in a manner to increase MUFA and decrease n-6 PUFA intakes, six servings/day of grains with at least three from whole grains, and seven to nine fruits and vegetable servings/day in a specified variety.

Fatty acid analysis was conducted by gas chromatography—mass spectroscopy (GC-MS) of fatty acid methyl esters. Total lipids were extracted from serum using a 1:1 mixture of chloroform and methanol, and 17:0 (1,2-diheptadecanoyl-sn-glycero-3-phosphocholine) was used as the internal standard. For colon tissue, one biopsy of about 5 mg was sufficient for analysis of fatty acids. The biopsy was pulverized in liquid nitrogen, sonicated in 150 μL of ice-cold PBS containing 0.1% butylated hydroxytoluene (BHT) and 1 mmol/L EDTA with an Ultrasonic processor (30 seconds twice), and then total lipids were extracted with 1 mL of chloroform and methanol (1:1). The organic layer in either case was used to prepare fatty acid methyl esters with Meth-Prep II derivatization reagent (Alltech). The GC-MS analysis was conducted with A SupelcoSP2330 column, 30 m × 0.32 mm X 0.2 μm film thickness (Sigma-Aldrich), a HP 7673/5971 GC-MS, and helium as the carrier gas with a validated assay (29). The following fatty acids in serum and colon tissue were measured in 12 analytic different batches: 12:0, 14:0, 16:0, 16:1, 18:0, 18:1, 18:2 (n6), 18:3 (n3), 20:0, 20:1, 20:3 (n6), 20:4 (n6), 20:5 (n3), and 22:6 (n3).

Many polymorphisms have been identified in the FADS1/2 gene cluster. Haplotypes have been constructed using three to 18 single-nucleotide polymorphisms (SNP), and arachidonic acid concentrations were typically about 30% higher in carriers of all major alleles (9, 12, 16, 30). This literature has indicated that there was little additional benefit from genotyping more than three SNPs, we therefore chose to genotype the three SNPs used in the study of the Rzehak and colleagues (30). A subsequent genome-wide association study indicated that another polymorphism in the FADS1/2 region explained 18% of the inter-individual variation in arachidonic acid concentrations, we therefore added rs174537 to the present analysis (21).

DNA was extracted from the buffy coat of heparinized blood samples. The buffy coat had been collected for each blood sample and mixed with 1% SDS/1 mmol/L EDTA before freezing at −80°C. After all the samples had been collected, they were treated with RNases A and heat-treated RNase T1 followed by digestion with protease K, solvent extraction, and precipitation of DNA. The DNA was purified using the MinElute Reaction Cleanup Kit (QIAGEN) to ensure high-quality DNA for genotyping. Four SNPs were genotyped: one in the FADS2 gene (rs3834458), two SNPs located in FADS1 (rs174556 and rs174561), and one SNP located in the intragenic region between FADS1 and FADS2 (rs174537).

Three of the SNPs (rs174556, rs174561, and rs174537) were genotyped using TaqMan SNP Genotyping Assays (Applied Biosystems). All assays were conducted both in real time and after read mode for allelic discrimination on an AB7900 system. The rs3834458 polymorphisms were detected by sequencing. For quality control, 10% of all samples were regenotyped. All plates incorporated positive and negative controls.

PCR reactions for rs3834458 included 5 μL (20 pmol/μL) of both forward and reverse primers, 12 μL AmpliTaq Gold Master Mix, 10 μg/μL genomic DNA, and Millipore water for a total volume of 25 μL. Primers used for rs3834458 were 5′-TCCACGATTCCCAAAGAGAC-3′ and 5′-TCTGCAACCTCCCTAGAGACA-3′. Samples were covered in mineral oil, denatured for 10 minutes at 95°C, were passed through 40 cycles of amplification consisting of 1 minute of denaturation at 95°C, 1 minute of primer annealing at 55°C, and 1 minute of elongation at 72°C. The PCR products were checked by running on a 2% agarose gel stained with ethidium bromide before sequencing. Sequencing was done on an ABI 3730 sequencer in the University of Michigan Sequencing Core Facility.

The distributions of fatty acid variables were first checked for normality and transformed to approximate normality as needed before analyses. The transformations applied before analysis are given in the table footnotes, and variables were back transformed to calculate percentage of increases or differences. Untransformed means are shown in the tables for ease of interpretation. Deviations from Hardy–Weinberg equilibrium for the genotypes of each SNP were tested using χ² tests. Differences in baseline parameters between diet arms were assessed using independent t or χ² tests, as appropriate (Tables 1 and 2). Genotype data for the four SNPs were summarized to yield the count of minor alleles (the minimum and maximum counts were zero and eight, respectively). Linear regression was used to evaluate the effect of number of minor alleles on fatty acid concentrations. Subsequently, a binary variable for genotype group was created by the presence/absence of minor alleles, that is, all major alleles versus one or more minor alleles.

A linear mixed model was used to evaluate whether the presence of any FADS variant affects baseline fatty acid concentrations (arachidonic acid and EPA). Each of the baseline fatty acids in both serum and colonic mucosa was regressed on genotype group (Table 2). Batch number was a random effect to account for heterogeneity as fatty acids were measured in different batches. The covariates in the model included age, gender, body mass index (BMI; in kg/m²), and dietary intake measures of n-6 PUFA, n-3 PUFA, and long-chain n-3 PUFA (sum of the n-3 fatty acids 20:5, 22:5, and 22:6) as a percentage of energy using 9 kcal/g.

Next, we used linear mixed models to evaluate the changes in fatty acid concentrations after 6 months of diet intervention: dietary intake, serum, and colon fatty acid concentrations were regressed on time (baseline, 6 month) with a random intercept for each individual. For serum and colon fatty acids, batch number was included in the random effects. Separate analyses were conducted for the two diet groups (Table 3). Finally, analyses were conducted to compare the changes in fatty acid composition over 6 months between the two diet arms and to assess whether the changes were modified by the presence of minor alleles in FADS. For these analyses, each of the outcome variables (arachidonic acid and EPA for both serum and colonic mucosa) at 6-month follow-up was regressed on genotype group, diet arm, and genotype group*diet assignment interaction by a linear mixed model (Table 4). The model was adjusted for age, BMI, and the concentration of each corresponding fatty acid at baseline. In all the models, batch number was incorporated as a random effect when appropriate. All reported P values were two-tailed. The statistical significance was set α = 0.05 level. Analyses were conducted using SAS version 9.1 (SAS Institute).

The overall study consisted of 108 study participants after exclusions for lack of genotyping consent (n = 9) and incomplete genotype data (n = 3). Genotyping success rate of the four SNPs chosen to define the FADS1/2 haplotype as described in Materials and Methods, was between 96.7% and 98.3%. Minor allele frequencies were in the range of 25.0% to 32.9%. The genotype distribution for each SNP did not deviate from Hardy–Weinberg equilibrium (P > 0.05).

Baseline characteristics for the Healthy Eating diet group (n = 54) and the Mediterranean diet group (n = 54) were summarized in Table 1. No significant differences were found in minor allele frequency of any SNP, gender, race, age, or BMI between the two diet groups at baseline. Likewise, baseline measurements of arachidonic acid, EPA, and long-chain n-3 fatty acids (the sum of EPA and DHA) did not differ significantly in the serum or the colonic mucosa between the two diet groups (Table 1).

Linear regression analysis indicated that the number of minor alleles was a significant predictor of baseline serum arachidonic acid concentration (P < 0.001) and almost significant for colonic arachidonic acid concentration (P = 0.058). A greater number of minor alleles was significantly associated with lower arachidonic acid concentration in serum. Dietary arachidonic acid intakes were not a significant predictor of either serum of colon concentrations. For long-chain n-3 fatty acids, however, the situation was the reverse. Dietary intake of long-chain n-3 fatty acids was a significant predictor of baseline serum long-chain n-3 concentration (P < 0.001) and colonic long-chain n-3 concentration (P = 0.044), whereas the number of minor alleles was not a significant predictor of either.

Subsequent analyses were conducted categorizing subjects into two groups by presence or absence of any minor alleles in the FADS gene cluster. The only dietary or demographic factor to differ by genotype at baseline was BMI, which was lower in carriers of any minor alleles (mean of 27.8, SD 3.7, in all major allele carriers and mean of 26.1, SD 3.6, in carriers of any minor alleles; P = 0.02 by the two-sided t test). Age was not significantly different (P = 0.11) but was retained as a covariate. No significant difference was found for other demographic characteristics (race, gender, smoking, and common medication use) between minor allele and all major allele carriers. Results were similar when using any one SNP individually versus all minor SNPS (not shown).

Serum and colon fatty acid concentrations at baseline by genotype group are shown in Table 2. Linear mixed models were used to evaluate differences between genotype groups. The presence of any minor alleles was highly significantly associated with baseline serum 20:4, n-6 concentrations (P < 0.0001) and 18:3, n-3 concentration (P = 0.01), and marginally significant for colonic 20:4, n-6 concentration (P = 0.07), with adjustment for age, BMI, and dietary intakes of n-6 PUFA, n-3 PUFA, and/or long-chain n-3 PUFA as a percentage of energy. Specifically, mean serum 20:4, n-6 concentration (percentage of total fatty acids) for minor allele carriers was estimated to be 2% [95% confidence interval (CI), 1%–3%] lower, whereas mean serum 18:3, n-3 concentration for minor allele carriers was estimated to be 21% (95% CI, 4%–41%) higher, compared with those individuals with all major alleles in the four SNPs in FADS. There was neither any significant association of genotype with EPA nor with long-chain n-3 fatty acids (the sum of EPA and DHA). Genotype group also had no significant effects on total cholesterol, Low density lipoprotein (LDL), high density lipoprotein (HDL), triglycerides, insulin, glucose, and C-reactive protein (CRP) with P > 0.11 in each case (not shown).

We first evaluated changes in fatty acids by diet group assignment alone without considering the genotype groups. Table 3 displays dietary intakes, serum, and colon fatty acid concentrations for the two diet arms at baseline and after 6 months of intervention. On the basis of data from food records and 24-hour recalls, dietary intakes of SFA and MUFA were significantly reduced (P < 0.0001) and long-chain n-3 PUFA was significantly increased (P = 0.004) in the Healthy Eating group after 6 months. The decrease in mean SFA resulted in an increased PUFA:SFA ratio from 0.60 to 0.92 in the Healthy Eating group (P = 0.008 from mixed linear regression models controlling for age). In the Mediterranean group, dietary intakes of SFA and n-6 PUFA both significantly decreased (P < 0.0001), whereas MUFA and long-chain n-3 PUFA significantly increased (P < 0.0001), in accord with the counseling goals. The mean PUFA:SFA ratio increased nonsignificantly from 0.72 to 0.77 in the Mediterranean group.

Serum 18:2 n-6 significantly decreased (P = 0.02), and both MUFA and n-3 PUFA significantly increased (P = 0.0005 and 0.01, respectively) in the Mediterranean arm only (Table 3). There was little change in colon fatty acid concentrations. The only significant change was for long-chain n-3 PUFA that significantly increased in both Healthy Eating (P = 0.01) and Mediterranean groups (P = 0.01).

Figures 1 and 2 show the raw means in each group over time. Table 4 shows the linear mixed model results for the analysis of the genotype by diet interaction. There was a significant interaction of genotype by diet for 20:4, n-6 (arachidonic acid) concentrations in the colon (P = 0.004). Significant genotype-by-diet interactions were neither found for arachidonic acid in serum nor for EPA. Among subjects with no minor alleles, mean colon arachidonic acid concentrations were estimated to be 16% (95% CI, 5%–26%) lower for the Mediterranean arm than the Healthy Eating arm at 6 months. These results indicate that after adjusting for baseline arachidonic acid concentrations, mean colon arachidonic acid concentrations at 6 months were significantly different between diet arms only in persons with no minor alleles in the FADS1/2 gene cluster. This was mainly due to an increase in colon arachidonic acid in the Healthy Eating diet arm, whereas colon arachidonic acid concentrations remained fairly constant in the Mediterranean group.

This randomized, dietary intervention study afforded the opportunity to evaluate the impact of FADS genotype and diet on fatty acid concentrations in both serum and colonic mucosa of individuals at increased risk for colon cancer. The number of minor alleles in the FADS gene cluster, but not diet, predicted serum arachidonic acid concentrations. This agrees well with results of previous studies, namely that carriers of minor alleles have lower arachidonic acid concentrations (9–15). For EPA concentrations in serum, genotype had no effect, whereas diet did have a significant effect, likely because n3 fatty acid intakes were fairly low and limiting in this study population. It should, however, be noted that diet in this study was assessed using self-report on four separate days. In addition to the possibility of mis-reporting of intakes, those 4 days might not represent usual intakes over the last month of study and therefore will weaken any apparent associations with diet.

In epidemiologic studies, relatively higher dietary intakes of both n-3 and n-9 fatty acids are thought to be protective, whereas high intakes of n-6 fatty acids increase risk of several cancers including that of the colon (31). This has been confirmed in experimental models of colon cancer, and low versus high n6 fatty acid diets are associated with decreased tumors and lower production of certain eicosanoids such as PGE2 (32, 33). In the colon, PGE2 has been tightly linked with colon cancer risk (34). Increased n-3 fatty acid intakes also reduce PGE2 production (35–39). Interestingly, a reduction in n-6 fatty acid intakes can augment increases in EPA after n-3 fatty acid supplementation (40). Bartoli and colleagues observed inhibition of aberrant crypt foci, adenocarcinomas, decreased mucosal arachidonate (20:4), and decreased PGE2 in rats fed either n-9 or n-3 diets relative to rats fed diets high in n-6 fatty acids (41). The levels of colon mucosal PGE2 were directly proportional to arachidonate levels in the colon in that study (41). These data make it important to better understand factors that could affect arachidonic acid and EPA levels in the human colon.

Unlike serum fatty acids, genotype had no significant effects on fatty acid concentrations in the colon at baseline (Table 2). It may be the case that serum concentrations of fatty acids are affected by first pass liver metabolism more so than tissues. After absorption of fatty acids, mainly in the small intestine, the liver is the initial site of fatty acid metabolism. The subsequent distribution of fatty acids from the circulation to tissues will be dependent on lipoprotein lipase activity in each tissue site and on tissue-specific metabolic conversions. In a well-controlled study in pigs, increased dietary intakes of linolenic acid and/or linoleic acid significantly affected metabolism of each other to long-chain fatty acids in the liver, but the effect was minimal in brain cortex (42). In a human lipodomic study, FADS activity of blood reflected activity in the liver but not in adipose tissue (43). Serum and colon fatty acid concentrations therefore reflect not only diet and genotype, but any tissue-specific regulation of fatty acid metabolism as well.

Because the present study was a randomized clinical trial, we then evaluated the effects of the two dietary interventions on changes in fatty acid intakes and levels over time. Both dietary interventions decreased SFA intakes and increased n-3 PUFA intakes. Only the Mediterranean intervention resulted in increased MUFA and decreased n-6 PUFA intakes. Serum fatty acids in the Mediterranean arm reflected these changes in diet (Table 3). In the colon, however, the only significant change was an increase in n-3 PUFA. This indicates that tissue-specific processes may limit the impact of dietary changes in colon fatty acids. The increase in colon n-3 PUFA is interesting, however, as the increases in dietary n-3 PUFA were modest in each diet arm.

The effect of FADS genotype on fatty acid concentrations in colon was only evident after intervention (Table 4). Study subjects who were carriers of all major alleles and randomized to the Healthy Eating intervention had higher colon arachidonic acid concentrations after 6 months than subjects with all major alleles in the Mediterranean group. It is not entirely clear why this should be the case, but the Healthy Eating intervention did result in a higher relative amount of n-6 PUFA to other dietary fats. This could have helped increase the percentage of arachidonic acid in the colon fatty acids after the Healthy Eating intervention. In addition to polymorphisms in FADS, other factors could be operative to affect fatty acid desaturation, such as diet-induced changes in the expression and the activity of FADS, and to changes in substrate competition (44). In carriers of all major alleles randomized to the Mediterranean intervention, arachidonic acid levels stayed relatively low at both time points and were estimated to be 16% lower than in the Healthy arm after 6 months of intervention.

Limitations of this study include the small sample size, the relatively short intervention length, and the self-report of diet that is known to be subject to biases. It may take longer for a change in diet to be fully manifest, especially in tissues. In addition, the measurement of fatty acids was done as a percentage of total fatty acids such that increases in one fatty acid on a volume basis would result in decreases in other fatty acids. An additional consideration is that arachidonic acid concentrations are not easily modifiable by changes in n6 fatty acids in the diet, especially if arachidonic acid is not elevated at the outset (45). Strengths of the study include that it was a randomized study, and measures were available before and after diet change in both serum and colonic mucosa of individuals at increased risk for colon cancer.

In conclusion, this study showed that those subjects with no minor alleles in the FADS1/2 cluster had higher concentrations of arachidonic acid in serum. Polymorphism in FADS1/2 had no effect on concentrations of EPA, perhaps because concentrations of this fatty acid are more highly driven by dietary intakes. The trends were similar in colon tissue fatty acids but not significant. After randomization to Mediterranean or Healthy Eating intervention for 6 months, there was a significant gene*diet interaction for colon arachidonic acid concentrations. Subjects who had all major alleles for FADS1/2 had significantly lower arachidonic acid concentrations in the colon after 6 months if they were in the Mediterranean diet arm. Because arachidonic acid is the substrate for PGE2 production, these results indicate that a Mediterranean diet could be especially favorable for reducing colon cancer risk in the subset of subjects with all major alleles in FADS1/2. Future work should evaluate the effects of these FADS polymorphisms on colonic proinflammatory states.

No potential conflicts of interest were disclosed.

Conception and design: S.B. Gruber, B. Mukherjee, Z. Djuric

Development of methodology: Y.-A. Ko, B. Mukherjee, J. Ren

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.R. Porenta, S.B. Gruber

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.R. Porenta, Y.-A. Ko, S.B. Gruber, B. Mukherjee, A. Baylin, J. Ren, Z. Djuric

Writing, review, and/or revision of the manuscript: S.R. Porenta, Y.-A. Ko, S.B. Gruber, B. Mukherjee, A. Baylin, Z. Djuric

Study supervision: S.B. Gruber, Zora Djuric

The authors thank all the individuals who volunteered for the Healthy Eating Study for Colon Cancer Prevention. The parent study was designed and conducted in collaboration with Drs. Dean E. Brenner, Mack T. Ruffin, D. Kim Turgeon, and Ananda Sen. Mary Rapai was the coordinator for the study and Maria Cornellier was the study dietitian.

This study was supported by a grant from the Ronald P. and Joan M. Nordgren Cancer Research Fund, NIH grant RO1 CA120381, and Cancer Center Support Grant P30 CA046592. The study used core resources supported by a Clinical Translational Science Award, NIH grant UL1RR024986 and NIH grant P50 CA130810 (the Michigan Clinical Research Unit), the Michigan Diabetes Research Center NIH grant 5P60 DK020572 (Chemistry Laboratory), and the Michigan Nutrition and Obesity Research Center NIH grant P30 DK089503.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.