Whole grains are associated with health, but are the effects only due to fibre?

The association between greater intake of whole grains and reduced risk of disease is one of the most consistent findings in nutritional epidemiology⁽ Reference Ye, Chacko and Chou ^{1 )}. The diseases where whole grains are associated with reduced risk are diverse, including CVD, diabetes and some types of cancer⁽ Reference Frølich, Åman and Tetens ^{2 )}. These associations point to potentially diverse metabolic and physiological effects of whole grains which might not be fully explain by their macro- and micronutrient composition alone. Much focus on explaining the mechanisms behind the lower risk of disease in people who eat the most whole grain has focused on dietary fibre⁽ Reference Satija and Hu ^{3 ,} Reference Smith and Tucker ^{4 )}. The greater fibre content of whole grains is likely an important factor, but may not explain all observed reduction in disease risk⁽ Reference Ye, Chacko and Chou ^{1 )}. For example, there are well-established mechanisms for cholesterol lowering by cereal-derived β-glucan⁽ Reference Lattimer and Haub ^{5 )}, yet the whole grains that are most commonly consumed in population studies that find associations between whole grains and health are wheat, maize and rice, grains that have a very low content of β-glucan and other soluble fibres. Other grain components are also likely to play a role in explaining the mechanism behind whole grains and health⁽ Reference Andersson, Dimberg and Åman ^{6 )}. This review explores the role of metabolomics in finding new hypotheses that fill the gaps on why eating more whole grains reduces the risk of some diseases.

Proposed mechanisms behind whole-grain health effects

Prior to the application of metabolomics to research on whole grains and health, explanations for why observational studies consistently found associations between eating more whole grains and reduced risk of many diseases centred on the higher fibre, vitamin and mineral content, as well as phenolic phytochemicals for their antioxidant capacity⁽ Reference Slavin ^{7 )}. Fibre in particular was thought to be important not only for reducing blood lipids, but for promoting the growth of beneficial bacteria and increasing the production of SCFA in the large intestine⁽ Reference Slavin ^{7 )}. Although there is little that is controversial about these proposed mechanisms, there was little explanation or evidence about how the compositional differences would impact on health. One major change in thinking between the early 2000s and the present day is the role of phenolic compounds in cereals as antioxidants, which would in turn reduce long-term disease incidence by prevention of oxidative damage. While an attractive explanation at the time, the theory that phenolic compounds reduce disease by acting as free-radical scavenging antioxidants has not been borne out in clinical intervention trials that have consistently found that high intake of antioxidant compounds has little effect on in vivo antioxidant capacity, and no effect on health outcomes⁽ Reference Myung, Ju and Cho ^{8 )}.

Prior to the advent of systems biology tools, most studies looking to understand the effect of whole grains on health have focused on a few clinically relevant endpoints including blood lipids, glucose and markers of inflammation. This has given some insight into possible mechanisms, but may not have been informative about the global impact of whole grains. For the past decade researchers have been encouraging the greater use of newer methodologies, including metabolomics, in nutrition in the search for a better understanding of the link between whole grains and health.

Metabolomics: an ideal tool to study biological mechanisms

Metabolomics is the field of analytical chemistry focused on measuring a broad spectrum of metabolites, normally to understand the global, or systemic, effects of a biological change, such as disease, environment or diet. Metabolites are the end product of the cascade from genes to proteins and then to metabolites, and as such reflect gene and protein modifications such as epigenetic modifications, and protein phosphorylation and glycosylation. In addition to endogenous metabolites, mammalian plasma and urine also contain metabolites that originate from gut microbial metabolism, and can be a window into the status of the gut microbiota.

A number of different methods are used to collect metabolomics data, the most common being proton NMR spectroscopy and GC or liquid chromatography coupled to MS (GC–MS or LC–MS). Several papers have addressed the key advantages of the various techniques⁽ Reference Zulyniak and Mutch ^{9 )}. For the purposes of this review, it should be noted that proton NMR requires less sample preparation and generally has less instrumental drift (instrument response will be very similar for the first or the 100th sample), whereas MS-based techniques are more sensitive and can detect more metabolites, but have higher run to run variation that needs to be corrected for statistically. High resolution MS can detect thousands of metabolites and in recent years have shown promise to greatly enhance our understanding of the breadth of metabolites present in human and food samples. After chemical analysis, the critical steps of post-run data analysis, where the acquired data are aligned and normalised, and statistical analysis are used to determine which of the dozens to thousands of measured metabolic features are predictive of the study question.

Although metabolomics holds much promise for advancing biological understanding, it needs to be emphasised that the goal of metabolomics is to generate hypotheses around metabolic mechanisms, rather than to definitively prove a mechanism or an effect. Because of the statistical and analytical problems in analysing large numbers of metabolites, hypotheses generated based on a metabolomics study need to be followed up with studies specifically powered to test that hypothesis, with analysis carried out using validated quantitative methods.

Metabolomic insights into whole-grain health benefits: animal studies

The first studies on the global impact of a whole-grain diet have been based on samples from animal studies. The first such paper highlighted the potential importance of the methyl donor glycine betaine (betaine) from a whole-grain rye and rye bran enriched bread fed to pigs as a potential mediator of health effects of whole grains and bran-rich cereals⁽ Reference Bertram, Bach Knudsen and Serena ^{10 )}. Urinary urea and creatinine were reduced when pigs were fed rye bread, while urinary hippurate was elevated. Later, a decrease in urinary urea and creatinine due to a whole-grain diet were linked to protein turnover in human subjects⁽ Reference Ross, Pere-Trepat and Montoliu ^{11 )}, while urinary hippurate is often described as a product of microbial metabolism of phenolic compounds, although it may also come from metabolism of aromatic amino acids⁽ Reference Lees, Swann and Wilson ^{12 )}. This work underlined the potential role of whole grains in interacting with gut microbiota, and for the first time linked the high concentration of betaine in rye to improved betaine status in a mammalian model.

Analysis of liver tissue samples from hypercholesterolaemic pigs fed either a high-fibre rye bread diet, or a high-fibre refined wheat diet using magic angle spinning proton NMR indicated differences to liver lipid metabolism, with high-fibre rye leading to an increase in liver free-choline, while the refined wheat diet led to elevated glycerophosphocholine and phosphocholine, suggesting a change in the expression or activity of phospholipases⁽ Reference Bertram, Duarte and Gil ^{13 )}. Notably there was no difference in the signal from trimethylamine oxide/betaine (the NMR signal for both overlaps making differentiation difficult), although plasma betaine was found to be increased in the same pigs⁽ Reference Bertram, Malmendal and Nielsen ^{14 )}. Lipids, possibly TAG, were decreased in plasma ⁽ Reference Bertram, Malmendal and Nielsen ^{14 )}, suggesting a possible link between changes to liver phospholipids and lipid transport and circulation in blood. A secondary study on the same hypercholesterolaemic pig model using untargeted LC–quadrupole time-of-flight MS metabolomics also found evidence for remodelling of lipid metabolism, with reduced plasma concentrations of linoleic acid-derived oxylipins and cholesterol on the rye bread-based diet⁽ Reference Norskov, Hedemann and Laerke ^{15 )}. In the present study, the mammalian lignan enterolactone was detected and highly elevated on the rye diet, a finding backed up by previous studies on similar pig models⁽ Reference Laerke, Mortensen and Hedemann ^{16 )}. In a study designed to compare metabolic responses between pigs and human subjects, both species were fed either whole-grain rye or refined wheat, and post-prandial plasma metabolic profile responses were compared⁽ Reference Nielsen, Hartvigsen and Hedemann ^{17 )}. Pigs and human subjects had similar relative metabolic changes for twenty-one out of twenty-six metabolites that were predictive of the different diet interventions in both species, indicating that pig models give a good reflection of human metabolism of cereals. Among the compounds that were different between the refined and whole-grain rye diets for both pigs and human subjects in this study were phosphatidylcholines 38 : 4 and 36 : 4, and lysophosphocholine 18 : 1, which were all elevated on the rye kernel diet compared with the refined wheat bread diet⁽ Reference Nielsen, Hartvigsen and Hedemann ^{17 )}.

These initial studies had focused on a high-fibre rye bread diet, based on whole-grain rye and rye bran, compared with a fibre (wheat cellulose) matched refined wheat diet. A rat study comparing a whole-grain wheat diet v. a non-fibre matched refined wheat diet identified several changes in urine, plasma and liver-extract related to the whole-grain wheat diet⁽ Reference Fardet, Canlet and Gottardi ^{18 )}. Contrary to the results for pigs, urinary creatine was elevated on the whole-grain wheat diet, along with urinary amino acids (tyrosine, tryptophan and phenylalanine) and hippurate. Excretion of central carbon metabolism metabolites citrate and fumarate were increased, while pyruvate excretion was decreased, hinting at a possible change to energy metabolism. A signal attributed to either trimethylamine oxide or taurine was also decreased on the whole-grain wheat diet. Plasma lipids and lysine were slightly elevated (~ 10 % increase) on the whole-grain wheat diet. Feeding whole-grain wheat also led to a decrease in liver lipids, and an increase in glucose, glutathione and the trimethylamine oxide/betaine signal. The present study also found many NMR signals that were significantly different between diets, but could not be identified⁽ Reference Fardet, Canlet and Gottardi ^{18 )}.

The role of gut microbiota in the release and availability of phytochemcials from cereal fibre was well demonstrated in a metabolomics study on mouse urine after feeding with differently processed wheat aleurone layer fractions. A large number of metabolites, including hippurate, dihydrophenolic acids and benzoxazinoids metabolites were released from the fibre matrix when the aleurone layer was milled or ultramilled. Fermentation of the aleurone layer feed led to a different urinary metabolite response altogether, including many unknown metabolites and small phenolic acids⁽ Reference Pekkinen, Rosa and Savolainen ^{19 )}. Although this work did not link the different responses to metabolic endpoints in the mice, it is a clear demonstration of the ability of metabolomics to detect differences in gut microbiota-related metabolites, as well as to discriminate between fermented and non-fermented diets.

Together, these first animal studies highlighted the link between the high concentration of betaine in wheat and rye to increases in circulating and tissue betaine, as well as the likely increased metabolism of phenolics by the gut microbiota as indicated by hippurate, enterolactone and a diverse range of other phenolic metabolites released from the fibre matrix. Possible effects on lipid metabolism were also detected in porcine liver and plasma, and as some of these involved phospholipids with choline moieties, these changes may be related to improved betaine status via choline sparing. Unfortunately none of these studies were able to link metabolic changes to health-related outcomes, although do give a picture of the potential scope for the metabolic effects of whole grains.

Metabolomic insights into whole-grain health benefits: human studies

Detecting metabolic changes due to diet in human subjects is more complicated than in animal models due to the wide genetic variation in the population (animals are often siblings) and variation in food intake aside from the intervention. Whole grain brings the additional problem that it inherently cannot be condensed into a pill or extract, so proportionally it cannot make up as large a part of the diet as rats or pigs being fed a fortified bread diet. Nevertheless, several studies have found that whole-grain diets do induce some metabolic differences compared with refined grain-based diets.

In a cohort of hypercholesterolaemic post-menopausal women partaking in a crossover study feeding either high-fibre refined wheat bread or high-fibre rye bread, several new metabolites were identified as related to the high-fibre rye diet: concentrations of the ribose metabolites ribitol and ribonic acid were elevated, as was the tryptophan metabolite indole acetic acid⁽ Reference Lankinen, Schwab and Seppänen-Laakso ^{20 )}. The authors correlated ribonic acid with tryptophan, and speculated that this may be because of an increased synthesis of sertotonin, and reduced hunger. Hunger was not measured in the present study, and serotonin was not directly measured, making this link somewhat speculative, although several studies have found that rye is more satiating than other cereals⁽ Reference Forsberg, Åman and Landberg ^{21 ,} Reference Isaksson, Tillander and Andersson ^{22 )}. Serum cholesterol increased on the high-fibre rye bread diet, although the fatty acids palmitic acid, oleic acid and myristoleic acid were decreased in plasma during the rye bread period⁽ Reference Lankinen, Schwab and Seppänen-Laakso ^{20 )} supporting the hypothesis of an effect on fatty acid metabolism beyond cholesterol.

Elevated plasma betaine has also been found in metabolomics studies in human subjects. In an intervention with a high intake of whole-grain rye bread with added rye bran, carried out in males diagnosed with the early stage prostate cancer, elevated plasma betaine and dimethylglycine, the post-methyl donation metabolite of betaine, were measured, backed up by a trend for reduced homocysteine during the intake of rye bread⁽ Reference Moazzami, Zhang and Kamal-Eldin ^{23 )}. Elevated plasma 3-hydroxybutyric acid and acetone was suggested to be related to a shift in energy metabolism from anabolic to catabolic state while fasting. This same study found that the rye bread diet improved glucose metabolism (reduced urinary C-peptide and plasma insulin) and reduced plasma prostate-specific antigen, a marker of prostate cancer risk⁽ Reference Landberg, Andersson and Zhang ^{24 )}. Other studies specifically analysing betaine found that whole-grain-rich or cereal betaine-enriched diets led to an increased amount of circulating betaine⁽ Reference Price, Keaveney and Hamill ^{25 ,} Reference Ross, Bruce and Blondel-Lubrano ^{26 )}, and that with very high doses of cereal betaine from the wheat aleurone layer, downstream effects on reducing homocysteine and increasing dimethylglycine and methionine could be measured⁽ Reference Price, Keaveney and Hamill ^{25 )}.

In work carried out on post-menopausal women, a group of the population at risk of developing type 2 diabetes, both fasting and post-prandial plasma samples had lower leucine and isoleucine concentrations when fed a whole-grain rye bread compared with refined grain bread⁽ Reference Moazzami, Bondia-Pons and Hanhineva ^{27 ,} Reference Moazzami, Shrestha and Morrison ^{28 )}. These results are the first indication that whole grains may have an impact on branched-chain amino acid (BCAA) metabolism, which may in part explain the mechanism how whole grains reduce insulin resistance (see further discussion later).

Although not a whole-grain study per se, Bondia-Pons et al.⁽ Reference Bondia-Pons, Nordlund and Mattila ^{29 )} studied the possible metabolic mechanism behind why sourdough endosperm rye bread gives a lower post-prandial insulin compared with refined wheat bread, using two-dimensional GC–MS based metabolomics. Similar to studies where whole-grain rye/rye bran has been the major intervention, amino acids increased (methionine and phenylalanine), as well as ribitol, a breakdown product of tryptophan, increased after the rye endosperm bread. Organic acids (butenedioic acid, ascorbic acid, picolinic acid and succinic acid) were increased after the refined wheat bread meal. The authors noted that both ribitol and picolinic acid are breakdown products of tryptophan, but with different responses to either rye endosperm bread or refined wheat bread. Their possible biological effects are also quite different, with picolinic acid suggested to induce inflammatory proteins in macrophages⁽ Reference Bondia-Pons, Nordlund and Mattila ^{29 )}. The similar finding of a possible effect on tryptophan metabolism as earlier described⁽ Reference Lankinen, Schwab and Seppänen-Laakso ^{20 )} gives greater weight to the hypothesis that eating cereal-based foods has an effect on aromatic amino acid metabolism.

A study comparing a refined-cereal based diet with either a whole-grain-based diet or a ‘healthy diet’ with whole grains, fish and Nordic berries found that the healthy diet intervention led to both changes to glucose response (2 h glucose and glucose area under the curve), and several changes to the lipidomic response, including n-3 fatty acids that would be expected to increase with increased fish intake, and many changes to TAG, phosphatidylcholine species and phosphatidylethanolamine species, indicating a wide ranging remodelling of circulating lipids due to the healthy diet⁽ Reference Lankinen, Schwab and Kolehmainen ^{30 )}. Conversely, the whole-grain-based diet, which included both wheat- and rye-based products, did not lead to any changes to the lipidome relative to the control diet⁽ Reference Lankinen, Schwab and Kolehmainen ^{30 )}. Although it is difficult to decipher differences in mixed diet interventions, it would appear that in this free-living study design, whole grains were not the main driver of changes in lipids in the healthy diet intervention. No health-related endpoints changed with the whole-grain diet (although a trend for decreased 2 h glucose was observed), which may fit with the observation that there was no major metabolic response to the intervention diet. These data support the idea that analysing the metabolome in samples from interventions where no primary outcomes change may be pointless, as any measured change in primary outcome should be reflected in a change in metabolic homeostasis. Null results are not published as frequently as ‘difference’ results, and it can be interpreted from the results found here that unless the primary goal is to find diet-specific biomarkers, metabolomics analysis of samples where no primary or secondary outcomes have changed will be unlikely to find any further changes to the metabolome. An additional observation is that multivariate models generated from metabolomics analyses whole-grain-based clinical trials is that they are often weak. This indicates a wide inter-individual variation in overall metabolic response, and that other factors such as other types of food intake are stronger drivers of metabolic differences than the whole-grain intervention itself. Care needs to be taken to ensure validation of multivariate models and to confirm results using correction for multiple testing.

Unifying or common mechanisms for whole grains and health benefits

Identification of novel metabolites related to whole-grain cereal exposure

Identifying cereal-related compounds linked to whole-grain intake can be a valuable tool in the search for novel biomarkers of food intake. Biomarkers of food intake can help improve estimates of food intake from questionnaires in observational studies, and be used to check compliance in dietary intervention studies. Using metabolomics, Beckmann et al. identified novel benzoxazinoid metabolites, 2-hydroxy-N-(2-hydroxyphenyl) acetamide and N-(2-hydrxyphenyl) acetamide, in urine as being specific biomarkers of sourdough rye intake⁽ Reference Beckmann, Lloyd and Haldar ^{52 )}. The benzoxazinoid metabolite 2,4-dihydroxy-1,4-benzoxazin-3-one sulphate was also linked to intake of rye bread based on metabolomic analysis of urine samples⁽ Reference Bondia-Pons, Barri and Hanhineva ^{53 )}, whereas 2-hydroxy-N-(2-hydroxyphenyl) acetamide sulphate and N-(2-hydrxyphenyl) acetamide sulphate were identified in post-prandial plasma samples, discriminating between rye sourdough and rye bran bread, and refined wheat bread⁽ Reference Hanhineva, Keski-Rahkonen and Lappi ^{54 )}. While previously known as plant allelochemicals, benzoxazinoids were rarely mentioned among the potentially beneficial compounds present in cereal grains. Metabolomics has helped to rekindle interest in benzoxazinoids and how they may play a role in the health benefits of rye in particular⁽ Reference Andersson, Dimberg and Åman ^{6 )}. Metabolomics has also confirmed that cereal alkylresorcinols are useful markers of whole-grain intake, with two studies confirming that alkylresorcinol metabolites in urine are related to both whole-grain wheat and rye intake⁽ Reference Bondia-Pons, Barri and Hanhineva ^{53 ,} Reference Primrose, Draper and Elsom ^{55 )}.

Research on whole grains tends to focus on health benefits, although eating the outer layers of the grain can also lead to greater exposure to toxic contaminants that may be present in the outer layers. The toxic fungal metabolite deoxynivalenol, when present, is found at highest concentrations in the bran fraction of the grain, and has been suggested to be potentially deleterious for human health⁽ Reference Hopton, Turner and Burley ^{56 )}. A small scale metabolomics study comparing urine from people with high v. low deoxynivalenol exposure found that urinary hippurate was the most discriminating metabolite, with highest concentrations being present in people with greatest deoxynivalenol exposure⁽ Reference Hopton, Turner and Burley ^{56 )}. Hippurate is frequently associated with a greater exposure to plant-based foods, and this case is more likely to be a marker of greater intake of cereal foods, which also increases the chance for deoxynivalenol exposure, rather than a direct metabolite from microbial metabolism of the fungal toxin.

What does the future hold?

Metabolomics is proving to be a useful tool for identifying novel pathways that are impacted by whole-grain diets, although as for the results on disease markers and other clinical chemistry endpoints, the results are mixed. However, there are some fairly consistent changes observed in several of the studies, including changes that can be related to energy metabolism, lipid metabolism and one-carbon metabolism. For these observations to be credible it is necessary to move beyond ‘changes’ and to start quantifying metabolites from the relevant pathways to confirm their role in the observed outcomes. This would require greater advances in the use of semi- and fully-quantitative metabolomics as standard, which would facilitate easier comparison between different datasets. At present few methods are quantitative due to the difficulty in applying standard curves that cover many hundreds of metabolites, even though NMR is inherently a quantitative method, and MS can be, provided a suitable spectrum of internal standards are used.

Notwithstanding future improvements in metabolomics methodology, at the heart of conclusive research on the mechanisms behind whole-grain health benefits lies greater consistency in how pre-clinical and clinical trials are carried out, including what type of grains and grain products are used. With the enormous heterogeneity of populations studied, intervention duration on top of the different products used, we should expect heterogeneous results. Another key gap is knowledge about metabolic response to grains aside from wheat and rye. At present there is a lack of understanding on how commonly consumed grains such as rice and maize may be beneficial for health, and lack of biomarkers for their intake. If it is possible to determine if there is commonality in metabolic responses between different grains then it will aid in understanding if it is necessary to discuss whole grains and health, or study the effect of individual grains.

Future work also needs to include a greater emphasis on molecular mechanisms. Unfortunately intact whole grains are not so amenable to in vitro studies commonly used to test mechanisms, so work may be limited to pure compounds or digests to try and best mimic what is absorbed from whole grains.

The application of metabolomics to studies where the effects of whole-grain or related interventions have been studied has broadened our view of how whole grains may mediate health benefits. Instead of just focusing on established clinical markers often associated with fibre, a researcher in the field can now consider a deeper role for impacts on lipid, glucose and protein metabolism, as well as the emerging field of epigenetics. Greater work is needed to establish if there is any causality between the phytochemicals found in grains and these metabolic effects, but we can now start adding more detail to the previously hypothesised mechanisms behind whole-grain health benefits.