Beruflich Dokumente
Kultur Dokumente
Precision nutrition aims to prevent and manage chronic diseases by tailoring dietary interventions or recommendations Lancet Diabetes Endocrinol 2018
to one or a combination of an individual’s genetic background, metabolic profile, and environmental exposures. Published Online
Recent advances in genomics, metabolomics, and gut microbiome technologies have offered opportunities as well as February 9, 2018
http://dx.doi.org/10.1016/
challenges in the use of precision nutrition to prevent and manage type 2 diabetes. Nutrigenomics studies have
S2213-8587(18)30037-8
identified genetic variants that influence intake and metabolism of specific nutrients and predict individuals’
Department of Nutrition
variability in response to dietary interventions. Metabolomics has revealed metabolomic fingerprints of food and (D D Wang ScD, Prof F B Hu MD),
nutrient consumption and uncovered new metabolic pathways that are potentially modified by diet. Dietary and Department of
interventions have been successful in altering abundance, composition, and activity of gut microbiota that are relevant Epidemiology (Prof F B Hu),
Harvard T H Chan School of
for food metabolism and glycaemic control. In addition, mobile apps and wearable devices facilitate real-time
Public Health, and Channing
assessment of dietary intake and provide feedback which can improve glycaemic control and diabetes management. Division of Network Medicine
By integrating these technologies with big data analytics, precision nutrition has the potential to provide personalised (Prof F B Hu), Department of
nutrition guidance for more effective prevention and management of type 2 diabetes. Despite these technological Medicine, Brigham and
Women’s Hospital and Harvard
advances, much research is needed before precision nutrition can be widely used in clinical and public health settings.
Medical School, Boston, MA,
Currently, the field of precision nutrition faces challenges including a lack of robust and reproducible results, the USA
high cost of omics technologies, and methodological issues in study design as well as high-dimensional data analyses Correspondence to:
and interpretation. Evidence is needed to support the efficacy, cost-effectiveness, and additional benefits of precision Prof Frank B Hu, Department of
nutrition beyond traditional nutrition intervention approaches. Therefore, we should manage unrealistically high Nutrition, Harvard T H Chan
School of Public Health, Boston,
expectations and balance the emerging field of precision nutrition with public health nutrition strategies to improve
MA 02115, USA
diet quality and prevent type 2 diabetes and its complications. nhbfh@channing.harvard.edu
can also enable more precise dietary assessment and technologies and wearable devices can facilitate real-time
disease risk stratification, both of which are fundamental data collection of diet, lifestyle, and biochemical variables
for more effective targeting of dietary approaches for and encourage individuals to actively participate in their
prevention and management of type 2 diabetes. own behaviour change and disease management.
Recent advances in omics technologies and wearable Precision nutrition can integrate data from emerging
devices have improved the prospect of applying precision technologies and traditional nutritional assessments into
nutrition to prevention and management of type 2 diabetes epidemiological or dietary intervention studies of
(figure 1). The huge increase in genome-wide association type 2 diabetes to achieve several goals: to better under
studies (GWAS) in the past decade has generated extensive stand the mechanisms underlying different responses by
new knowledge about the genetic architecture of the individuals to a dietary exposure or intervention, in terms
disease.15 High-throughput next-generation sequencing of the risk of type 2 diabetes and glycaemic outcomes; to
technology has made whole-exome and whole-genome achieve better assessment of dietary intakes and
sequencing studies feasible and has underpinned studies nutritional status in free-living populations; to identify
of biological axes beyond the genome, such as the novel biomarkers that are more effective than traditional
transcriptome, epigenome, and microbiome. Mass biomarkers at predicting risk of disease and its
spectrometry and nuclear magnetic resonance have been complications; to identify new targets for lifestyle and
applied to analyse small-molecule metabolites in pharmacological interventions; and to provide person
biospecimens, allowing for a comprehensive assessment alised dietary and lifestyle guidance for more effective
of an individual’s metabolic status. Meanwhile, mobile prevention and management.
Diet and type 2 diabetes Measurement and mechanism Data integration and analysis Translation and intervention
9
Diet Questionnaires
5
Healthy
eating
plate Precision nutrition
6
7 Big data analytics
Omics
13
2
Physical activity
4 10
3
Medication
8 11
Wearable devices 14
22·0 17·0
Type 2 diabetes
12
15
Figure 1: Conceptual framework for precision nutrition in prevention and management of type 2 diabetes
(1) General recommendations for healthy diets. (2) Dietary intake interacts with level of physical activity. (3) Dietary approaches interact with antidiabetic
medications in management of type 2 diabetes. (4) Various omics technologies, such as genomics, metabolomics, metagenomic and metatrascriptomic analysis of
the gut microbiota, and epigenomics, provide powerful tools for deep phenotyping individual characteristics and understanding mechanisms underlying diet and
type 2 diabetes. (5) Validated questionnaires such as food frequency questionnaires are the most important and feasible tools for measuring long-term usual diet in
nutritional epidemiological research. (6) Wearable devices and mobile apps provide objective and real-time diet and physical activity measurements. (7) Application
of omics technologies to improve dietary assessment in free-living populations. (8) Wearable devices provide continuous measures of blood glucose and other
physiological variables. Data input from (9) self-reported questionnaire-based dietary assessment tools in epidemiological studies, (10) omics studies, (11) wearable
devices, and (12) conventional clinical measures, such as fasting glucose and lipids. (13) Results derived by use of big-data analytics inform development and
application of precision nutrition. (14) Wearable devices and mobile apps provide useful tools for monitoring and implementing precision nutrition. (15) Precision
nutrition aims to provide personalised nutrition guidance to achieve more effective dietary prevention and management of type 2 diabetes.
Overview of recent advances in type 2 diabetes metabolism and therefore could be nutritional inter
omics studies vention targets for prevention.
GWAS, and more recently exome sequencing studies, The application of next-generation sequencing
have identified more than 100 loci reproducibly associated technologies (eg, shotgun metagenomic and meta
with risk of type 2 diabetes and glycaemic traits.16 Most transcriptomic sequencing) for comprehensive mapping
identified loci have a small effect size and are com of microbes has offered new insights into the role of gut
mon across populations, with a few low-frequency microbiota in glycaemic control and pathobiology of
exceptions.15,17,18 Causal variants have only been identified type 2 diabetes.34 Accumulating evidence suggests an
for a few loci. These GWAS loci collectively explain less association between the overall composition and diversity
than 10% of the heritability of the disease.19 Overall, of the gut microbiota and type 2 diabetes,35 although
GWAS have not yet led to meaningful clinical advances causality remains uncertain. In addition, the relative
in prevention or management of type 2 diabetes. abundance of specific gut microbes is linked to risk of
Researchers have also discovered genetic variants directly the disease.36 For example, increased abundance of
related to dietary intake and nutrient metabolism, such Akkermansia muciniphila has been associated with
as polyunsaturated fat intake,20,21 macronutrient intake,22 improvement in glucose control and lower risk of type 2
alcohol drinking,23 coffee consumption,24 and zinc diabetes.37,38
transport.25 Interestingly, many BMI loci identified In summary, recent genomics studies have identified a
through GWAS are linked to hypothalamic function and large number of type 2 diabetes loci that can be used to
energy homoeostasis, and therefore closely related to characterise an individual’s genetic predisposition to
food intake.26 These findings suggest that taking into the disease. Metabolomics studies have identified
account individual genotypes when formulating dietary metabolites associated with type 2 diabetes that can be
recommendations may be more effective in improving potentially modified by diet. There is also increasing
nutritional status. evidence that the gut microbiome plays a role in
Studies on diet and changes in transcription and glycaemic control and pathophysiology of the disease.
methylation patterns have elucidated potential causal These findings provide a scientific basis for the use of
variants of type 2 diabetes and the biological basis personalised risk characterisation and stratification in
underlying the effect of dietary exposure on predisposition dietary interventions.
to the disorder. Most transcriptomic studies using
unbiased microarray and RNASeq approaches have Dietary assessment in free-living populations
focused on changes in the transcriptome induced by and causal inference
dietary interventions during a brief time period, such as Dietary assessment in free-living populations is
several days. A few studies have evaluated transcriptomic challenging because of random and systematic errors in
response to long-term dietary interventions, such as the commonly used self-report tools. Omics technologies
Mediterranean diet,27 or they have examined differentially offer the potential for more accurate assessment of
expressed transcripts due to different habitual dietary dietary intake and nutritional status. Improvements in
exposures.28 Genome-wide DNA methylation studies the quantification of dietary intake, characterisation of
comparing patients who have type 2 diabetes with healthy dietary patterns, and assessment of adherence to dietary
controls found varying levels of DNA methylation in interventions are fundamental in building an evidence
pancreatic islets for thousands of CpG sites, corresponding base to support precision nutrition.
to a large number of genes, including many known Metabolomics analysis has the potential to capture the
type 2 diabetes loci such as TCF7L2, FTO, and PPARG.29–31 complexity of habitual diet better than do traditional
Some genes implicated in these identified CpG sites also single biomarkers. New metabolomics technologies are
displayed altered mRNA expression. These findings shed able to profile a large number of metabolites derived from
light on the mechanisms underlying epigenetic regulation ingestion and absorption of foods, measure endogenous
of transcriptional activity contributing to β-cell function, biotransformation of nutrients by both the host and gut
insulin secretion, and the development of type 2 diabetes.32 microbiota, and uncover diet-induced metabolic response.
In recent years, high-throughput metabolomics has Several studies have identified new biomarkers for acute
been widely applied in population-based research on food exposure, short-term food or nutrient intake, and
type 2 diabetes. In a meta-analysis33 of eight prospective long-term dietary intake.39,40 Urinary proline betaine and
studies including 8000 individuals (1940 cases of 4-hydroxy-proline betaine have been identified as
type 2 diabetes), our group found statistically significant biomarkers for citrus foods. Plasma ether-linked
positive associations of plasma concentrations of phospholipids and plasmalogens, dihexosylceramide,
branched-chain aminoacids (BCAAs), including leucine and GM3 ganglioside are biomarkers that can distinguish
and valine, and aromatic aminoacids (AAAs), including between major dietary fat sources. Plasma and urinary
tyrosine and phenylalanine, but inverse associations anserine is a marker for chicken whereas plasma and
with glycine and glutamine, with type 2 diabetes. These urinary trimethylamine-N-oxide and carnitines are
metabolites are influenced by both dietary intake and markers for red meat and fish.41,42 More recently,
Garcia-Perez and colleagues43 identified a combination of Genetically predicted milk intake was not significantly
multiple metabolites, including those showing higher associated with risk of type 2 diabetes or multiple glycaemic
intake of fruits (eg, hippurate, tartrate, and glycolate), traits. Genetic markers may also be indicators of variation
vegetables (eg, N-acetyl-S-methyl-cysteine sulphoxide and in nutrient metabolism. Several research groups have used
S-methylcysteine sulphoxide), fish (dimethylamine), and genetic variants in the fatty acid desaturases gene family as
lean white meat (1-methylhistidine and 3-methylhistidine). surrogates of the activity of Δ5 desaturase (FADS1) and Δ6
These metabolite combinations were indicative of overall desaturase (FADS2), key enzymes in the metabolism of
dietary pattern in a controlled crossover feeding study polyunsaturated fatty acids. Genotype-predicted
and were validated in a subpopulation of cohort studies.43 Δ5 desaturase activity was inversely associated with risk of
Andersen and colleagues44 identified combinations of type 2 diabetes whereas genotype-predicted Δ6 desaturase
urinary metabolites that quantified compliance to either a activity was positively associated,52 suggesting a causal
new Nordic diet or an average Danish diet. effect of the activity of these enzymes on development of
Several studies have used metabolomics to uncover the disease and a mediating role in the effect of
novel dietary patterns and understand the role of diet in polyunsaturated fatty acid intake on risk.
prevention of type 2 diabetes.45–49 Floegel and co- It is noteworthy that only a few genetic variants have
workers46 made use of serum metabolites previously been found to strongly predict and thus serve as good
linked to risk of type 2 diabetes to derive dietary patterns proxies for dietary intake and nutritional status. In
in the EPIC-Potsdam study.46 A dietary pattern with high addition, Mendelian randomisation analyses can be
intake of red meat and low intake of whole-grain bread subject to confounding by population stratification,
and tea was associated with a metabolic profile high in pleiotropy of genetic effects, and inadequate statistical
BCAAs and AAAs. A second dietary pattern with high power.53
intake of coffee, cake, cookies, and canned fruit and fish
was linked to a metabolic profile (including many Identification of susceptible or responsive
phosphatidylcholines with a high degree of unsatu individuals through gene–diet interactions
ration) associated with low risk of type 2 diabetes. Type 2 diabetes is a complex phenotype that is influenced
Metabolomics analysis has also been applied to assess by both genetic and environmental factors as well as
long-term adherence to the dietary intervention in the their interactions.54 Studies that apply gene–diet inter
PREDIMED trial.50 In this trial, Vazquez-Fresno and action findings to precision nutrition could provide a
colleagues found multiple urinary metabolites, scientific basis for genotype-based precision nutrition.
including BCAAs, creatine, creatinine, oleic acid, and Some classic examples of translating genetic information
metabolites related to carbohydrate and gut microbiota into personalised nutrition are dietary modifications for
metabolism, that could reflect adherence to the individuals with mutations in the phenylalanine
Mediterranean diet during the 3 year intervention. hydroxylase gene that lead to phenylketonuria55 and
Overall, current metabolomics platforms show only mutations in the lactase gene affecting lactase
modest ability to distinguish different dietary patterns persistence.56 However, the genetic basis of these dietary
and have not yielded biomarkers with high sensitivity and modifications usually involves single-gene mutations
specificity for food or nutrient intake. This failure is partly with large effect sizes, whereas such clear examples do
because metabolites reflect not only dietary intakes, but not exist for type 2 diabetes.
also metabolic activity, the microbiome, and genetic Earlier studies of gene–diet interactions for type 2
background. Therefore, these technologies are unlikely to diabetes focused on candidate genetic variants.57 This
replace traditional assessment tools such as validated approach did not generate robust evidence,58 as shown by
dietary questionnaires and well established nutritional Li and colleagues,59 who were unable to replicate eight
biomarkers. However, they can serve as complementary published statistically significant inter actions between
tools for measuring dietary intakes in observational single-gene variants and macronutrient intake in relation
studies and assessing compliance to dietary interventions. to type 2 diabetes in a large cohort from the EPIC-InterAct
Recently, the Mendelian randomisation method has study. Since 2008, investigators have been using polygenic
become popular in nutritional epidemiological studies to scores, which are combinations of multiple susceptibility
determine the causal effect of dietary intake as predicted loci identified by GWAS. That these scores reflect the
by genotypes on health outcomes. Mendelian random polygenic nature of type 2 diabetes is crucial when
isation mimics a natural randomised controlled trial exploring gene–diet interactions for this disease. Some
because genotypes are assigned to individuals randomly at evidence supports an interaction between a polygenic
birth. Therefore, this method can largely eliminate score of ten type 2 diabetes risk alleles and a western
unmeasured confounding and reverse causation, which dietary pattern, characterised by high intakes of red and
are two major biases in epidemiological studies. A recent processed meat, butter, high-fat dairy products, and refined
study used a single nucleotide polymorphism (rs6754311) grains.60 In a recent study, Langenberg and co-workers61
as a proxy for milk intake; each allele increment was tested the interaction between a polygenic score of
associated with a 66 g increase in daily milk consumption.51 49 type 2 diabetes loci and a Mediterranean dietary pattern
in EPIC-InterAct but found no statistically significant gene–diet interactions have been difficult to replicate,
interaction effect. However, healthy dietary patterns have probably because of heterogeneity in diet and lifestyle
been found to mitigate the effects of genetic variants on across populations, insufficient statistical power, and
obesity62 and coronary heart disease.63 On the other hand, lack of standardisation in dietary assessment and
unhealthy diet habits, such as regular consumption of analytical approaches.
sugar-sweetened beverages, fried food, and saturated fat
intake, were found to exacerbate the effects of genetic Identification of nutritional intervention
variants on obesity.64–67 Evidence on gene–diet interactions targets for prevention of type 2 diabetes
for obesity is crucial for prevention of type 2 diabetes given In addition to assessing individual metabolic state,
that a small reduction in bodyweight will lead to clinically metabolomics can capture metabolic changes in response
meaningful improvement in glycaemic control.68 In to extrinsic exposures such as dietary intake and can
addition, robust evidence indicates that the FTO variant uncover metabolic pathways through which dietary
can modify the association between physical activity, exposures influence risk of type 2 diabetes. Floegel and
another major lifestyle risk factor of type 2 diabetes, and colleagues78 identified several metabolic networks that
body adiposity.69 These findings imply that individuals with correlated with both dietary intake and metabolic disease
greater genetic predispositions might benefit more from outcomes. For example, metabolic networks consisting of
interventions to improve diet quality or physical activity, particular metabolites within the subclasses of phos
although this hypothesis has not been explicitly tested in pholipid, sphingomyelin, lyso-phosphatidylcholines, and
randomised trials. acyl alkyl-phosphatidylcholines were positively associated
In the Diabetes Prevention Program (DPP), Florez and with coffee intake, but inversely associated with risk of
colleagues70 found that lifestyle intervention and obesity and type 2 diabetes,78,79 delineating potential
metformin treatment attenuated the effect of the TCF7L2 pathways through which coffee may be protective.
genotype on the progression from prediabetes to Recent studies within randomised controlled trials have
type 2 diabetes, although the interaction between the used metabolomics to provide novel evidence to explain
genotype and interventions was not statistically significant heterogeneity in individual responses to a specific dietary
(table). The POUNDS LOST trial investigated interactions intervention. In a study based on two weight-loss trials—
between intervention diets varying in macronutrient the POUND LOST trial and the DIRECT trial—weight-
composition and type 2 diabetes genetic risk scores on loss diets decreased metabolites that were associated with
markers of glucose homoeostasis.71,72 Among white risk of type 2 diabetes, including BCAAs and AAAs.74
participants, those with a lower genetic risk score showed More importantly, the observed decreases in the
more favourable responses to low-protein diets, including circulating aminoacid metabolites were predictive of
greater decreases in fasting insulin, HbA1c, and improvement in insulin resistance independent of weight
homoeostatic model assessment (HOMA) for insulin loss. In the DPP, Walford and colleagues75 found that a
resistance, and a lesser increase in HOMA for β-cell lifestyle intervention that incorporated dietary
function, than did those with a higher risk score modification was effective in in creasing betaine
within 2 years of follow-up.71 In response to low-fat diets, concentration from baseline to 2 year follow-up, which
participants with a lower genetic risk score showed more predicted lower risk of type 2 diabetes. In a subpopulation
positive responses in fasting glucose, insulin resistance, from the PREDIMED trial, a Mediterranean diet rich in
and insulin sensitivity during 6 months of follow-up.72 extra-virgin olive oil decreased plasma concentrations of
Only a few observational studies and randomised BCAA (Ruiz-Canela M, University of Navarra, personal
controlled trials have tested gene–diet interactions in communication). Moreover, a decrease in BCAA
relation to type 2 diabetes and glycaemic traits. Although concentrations from baseline to year 1 predicted lower risk
there is some indication that individuals with varying of disease in the remainder of the follow-up period,
genotypes respond differently to dietary interventions in whereas baseline BCAA concentrations modified the
terms of glycaemic markers, few independent effect of the Mediterranean diet intervention on risk of
replications exist. Overall, current evidence is not type 2 diabetes.
sufficient to make personalised dietary recommendations Metabolomics techniques are powerful tools for
for diabetes prevention or management based on genetic uncovering novel biomarkers linking dietary components
information. Several limitations of current studies are and risk of type 2 diabetes, which can be useful for
worth noting. First, some studies were cross-sectional distinguishing responders from non-responders to a
and therefore were susceptible to reverse causation and dietary intervention. However, current studies have
had insufficient control of confounding factors. Second, generally focused on several candidate metabolites or a
measurement errors in self-reported dietary intake, limited number of targeted metabolites, and there are
pathological heterogeneity of type 2 diabetes, and modest few independent replications of findings. Whether the
effect sizes of genetic variants collectively lead to limited identified metabolites represent causal biological
statistical power that may obscure true interaction pathways linking dietary intake to risk of the disease is
effects, leading to false negative findings. Third, largely unknown. Metabolomics studies that cover all
GWAS=genome-wide association study. SNPs=single nucleotide polymorphisms. HOMA-B=homoeostasis model assessment of β-cell function. HOMA-IR=homoeostasis model assessment of insulin resistance.
BCCAs=branched-chain aminoacids. AAAs=aromatic aminoacids.
Table: Summary of key randomised controlled trials on precision nutrition and prevention and management of type 2 diabetes
detectable, untargeted metabolites are needed to provide Recent studies have demonstrated that both short-term
a less biased understanding of potential metabolic dietary changes and long-term habitual diet could
pathways. Additionally, replication of statistically signifi influence the composition of gut microbiota.76,80–83 In a
cant findings in independent populations, and functional 5 day feeding study, David and coworkers80 showed that
studies of identified metabolites are needed to provide high intakes of animal fat and protein increased
causal evidence supporting the role of metabolites as bile-resistant bacteria including genera of Alistipes,
intervention targets. Bilophila, and Bacteroides, and decreased types of bacteria
with an enhanced capacity to ferment polysaccharides, Personalised nutrition for the management of
including Roseburia spp, Eubacterium rectale, and type 2 diabetes
Ruminococcus bromii. Two recent studies found that The use of omics tools to elucidate differing individual
long-term intakes of various dietary factors were responses to dietary exposures, in terms of the progression
associated with the diversity and composition of gut of type 2 diabetes or the onset of associated complications,
microbiota.81,83 Hallmarks of a western-style dietary seems highly relevant for personalised nutrition for the
pattern, including sugar-sweetened beverages, higher management of the disease. However, such evidence is
energy intake, snacking, and whole milk were associated very sparse. The Look AHEAD randomised trial compared
with lower microbiota diversity, whereas consumption of the effect of an intensive lifestyle intervention with
coffee, tea, and red wine was associated with high diabetes support and an education programme on
microbiota diversity. cardiovascular disease in patients with type 2 diabetes. A
Recent evidence suggests that dietary interventions recent study investigated whether the effect of the
may affect glycaemic control through modification of the intensive lifestyle intervention varied across different
composition of gut microbiota. After a 6 week calorie- genotypes of a genetic variant (s10911021) related to
restricted diet enriched with fibre and protein,38 the coronary heart disease, but failed to identify statistically
abundance of A muciniphila, a mucin-degrading significant interaction effects.73 Numerous antidiabetic
bacterium, was inversely associated with measures of medications with different mechanisms of action add
fasting glucose and insulin sensitivity in 49 participants another layer of complexity to studies on personalised
in France. Baseline abundance of this organism appeared nutrition in patients with type 2 diabetes.86 For example,
to modify the effect of diet on insulin sensitivity markers. metabolomics studies have found that metformin and
In a study of 39 healthy participants in Sweden, sulphonylurea affect the circulating metabolome pro
Kovatcheva-Datchary and colleagues76 found that the ratio foundly.75,87,88 Some of the identified metabolites, such as
of Prevotella spp to Bacteroides spp was statistically sphingomyelins, overlap with metabolomic signatures of
significantly higher in participants who showed improved dietary intakes, suggesting that antidiabetic medications
glucose metabolism after the 3 day intervention of barley may potentially interfere with the effects of dietary intakes
kernel-based bread than in those who did not show on multiple metabolic pathways. Additionally, recent
improved glucose metabolism. studies suggest that a large proportion of the previously
By combining metabolomics and the gut microbiome, identified gut microbial signatures associated with
Wu and colleagues84 provided some early mechanistic type 2 diabetes can be explained by metformin use89 and
evidence to explain heterogeneous responses to dietary that the effects of dietary interventions on gut microbiota
pattern in a cross-sectional study of 15 vegans and are masked by metformin.90 Beyond these glucose-lowering
six omnivores. The investigators found that a plant-based agents, some specific characteristics of type 2 diabetes,
diet was necessary but not sufficient for the bacterial such as duration, severity, comorbidities (eg, renal or
production of short-chain fatty acids, a group of metabolites hepatic impairment), and pathophysiology of the disease
implicated in host metabolic regulation and risk of (eg, the relative contributions of the defect in insulin
type 2 diabetes.85 A high yield of short-chain fatty acids secretion and insulin resistance), are also important
required high enrichment of genera such as Prevotella, but considerations for studying and designing personalised
low abundance of Ruminococcus bromii. nutrition approaches for the management of type 2
Thus, there is increasing evidence that diet has both diabetes. However, to our knowledge, no studies have
short-term and long-term influences on composition of comprehensively explored how these factors modify the
the gut microbiota. Certain dietary interventions, responses of the human body to dietary interventions to
particularly plant-based high-fibre diets, may be able influence the progression of the disease.
to modify specific types of gut microbes implicated
in regulation of glucose metabolism. Integrating meta Integrated omics technologies and big-data
bolomics and gut microbiome technologies can reveal analytics
the functional activity of gut microbiota and the role of Precision nutrition requires integration of multiple kinds
gut microbes in modifying individual responses to a of omics data as well as information from both traditional
dietary exposure. However, research on gut microbiota is sources, such as questionnaire interviews and standard
still at a nascent stage and is not ready for translation clinical tests, and modern sources such as electronic
into clinical practice. Existing studies are limited by medical records, mobile apps, and wearable devices.91,92
insufficient statistical power, cross-sectional study New bioinformatics tools for data analysis and
design, and lack of information on functional activity and visualisation, such as big-data analytics, are imperative
strain-level variation of the gut microbiota. Future large given the recent increase in the volume and complexity of
studies within prospective cohorts and long-term available data. However, integration of data from multiple
randomised controlled trials are needed to examine the disparate sources and the application of big-data analytics
modifiable capacity of the gut microbiome and its role in have not yet provided valuable physiological insights that
the prevention and management of type 2 diabetes. can be used in clinical practice. Researchers using these
A
Per person profiling Computational analysis
Gut microbiome Diary (food, sleep, physical activity)
16S rRNA Using smartphone-adjusted website Main cohort PPGR prediction
Metagenomics 5435 days, 46 898 meals, 9·8 million calories, 2532 exercises
100 Continuous glucose monitoring
Blood tests Using a subcutaneous sensor (iPro2)
130 000 hours, 1·56 million glucose measurements 800 participants
Questionnaires
Standardised meals (50 g available carbohydrates)
Food frequency Validation cohort Dietary intervention
Lifestyle Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7
Medical
G G F
B C
* † ‡ ‡ ‡ * ‡ * † ‡ ns ns * ‡ * * ‡ ns † † § § § ns ‡ ns ‡ ‡
140 100
Bad diet week
120 Good diet week
80
PPGR (iAUC, mg/dL.h)
80 60
60 40
40
20
20
0 0
P6 P10 P3 P8 P2 P5 P9 P4 P1 P11 P7 P12 E8 E7 E9 E4 E14 E11 E10 E12 E5 E3 E2 E1 E6 E13
rt
or
pe
ict
Ex
ed
Predictor-based arm Expert-based arm
Pr
Figure 2: Personalised nutrition in reducing postprandial blood glucose
(A) Illustration of the experimental design of Zeevi and colleagues.77 (B) Left, mean postprandial glycaemic responses to personalised dietary intervention (good diet)
vs control diet (bad diet), and right, traditional dietary advice (good diet) vs control diet (bad diet) at each timepoint of intervention period (by weeks). (C) Left, mean
postprandial glycaemic responses to personalised dietary intervention vs control diet, and right, traditional dietary advice vs control diet. PPGR=postprandial glucose
response. Good diet=meals predicted to have low postprandial glycaemic responses. Bad diet=meals predicted to have high postprandial glycaemic responses.
*p<0·001, †p<0·01, ‡p<0·05, §p<0·1, ns=not significant (Mann-Whitney U test). iAUC=incremental area under the curve. Adapted from Zeevi and colleagues,77 by
permission of Elsevier.
new methods have faced challenges including issues with continuous glucose monitor, physical examinations, blood
incomplete and unreliable input data, as well as tests, and 16S rRNA metagenomics profiling. The
misleading interpretations of findings owing to a lack of investigators reported that nutrition interventions based
expert knowledge. In addition, big data usually require on this algorithm were more effective in reducing
many statistical tests which create an increased chance of postprandial blood glucose than was traditional dietary
false-positive results. Therefore, independent replication advice. Although this proof-of-concept study is an
of findings has become even more important.93 important first step, several methodological issues
Several recent studies combined multiple data sources complicated the interpretation and translation of the
and applied big-data analytics to inform personalised findings. One of the key limitations was that the authors
nutrition interventions. In one example, Price and were unable to demonstrate whether the observed high
colleagues94 showed the benefits of combining information variability in postprandial glycaemic responses was due to
collected by genomics, metabolomics, proteomics, intra-individual or interindividual variability.95 The clinical
microbiome analysis, and wearable devices in an integrated applications of this approach are challenging because of
framework to develop personalised dietary interventions. the huge amounts of data that need to be collected,
In another study, Zeevi and co-workers77 examined the analysed, and interpreted. In addition, although wearable
utility of personalised nutrition in reducing postprandial devices showed the potential for scalable use in
blood glucose (figure 2). They developed a machine- continuously monitoring health behaviours and motivating
learning algorithm for predicting postprandial glycaemic behavioural changes of participants in the short term, little
responses based on integrated data on dietary intake, is known about their sustainability and effectiveness in the
biomarkers, physical activity, sleep, anthropometric long term.96 Long-term studies of this type that have a
variables, and gut microbiota that were collected through more refined study design as well as simple and
instruments including questionnaires, smartphones, a cost-efficient data collection tools are clearly needed.
7 Ley SH, Hamdy O, Mohan V, Hu FB. Prevention and management 32 Ronn T, Ling C. DNA methylation as a diagnostic and therapeutic
of type 2 diabetes: dietary components and nutritional strategies. target in the battle against type 2 diabetes. Epigenomics 2015;
Lancet 2014; 383: 1999–2007. 7: 451–60.
8 Wang DD, Li Y, Chiuve SE, Hu FB, Willett WC. Improvements in 33 Guasch-Ferre M, Hruby A, Toledo E, et al. Metabolomics in
US diet helped reduce disease burden and lower premature deaths, prediabetes and diabetes: a systematic review and meta-analysis.
1999–2012; overall diet remains poor. Health Aff 2015; 34: 1916–22. Diabetes Care 2016; 39: 833–46.
9 Scheen AJ. Precision medicine: the future in diabetes care? 34 Arora T, Backhed F. The gut microbiota and metabolic disease:
Diabetes Res Clin Pract 2016; 117: 12–21. current understanding and future perspectives. J Intern Med 2016;
10 Reddy SS. Evolving to personalized medicine for type 2 diabetes. 280: 339–49.
Endocrinol Metab Clin North Am 2016; 45: 1011–20. 35 Brunkwall L, Orho-Melander M. The gut microbiome as a target for
11 Florez JC. Precision medicine in diabetes: is it time? prevention and treatment of hyperglycaemia in type 2 diabetes: from
Diabetes Care 2016; 39: 1085–88. current human evidence to future possibilities. Diabetologia 2017;
12 Ferguson LR, De Caterina R, Gorman U, et al. Guide and position 60: 943–51.
of the International Society of Nutrigenetics/Nutrigenomics on 36 Qin J, Li Y, Cai Z, et al. A metagenome-wide association study of gut
personalised nutrition: Part 1 - Fields of precision nutrition. microbiota in type 2 diabetes. Nature 2012; 490: 55–60.
J Nutrigenet Nutrigenomics 2016; 9: 12–27. 37 Yassour M, Lim MY, Yun HS, et al. Sub-clinical detection of gut
13 Collins FS, Varmus H. A new initiative on precision medicine. microbial biomarkers of obesity and type 2 diabetes. Genome Med
N Engl J Med 2015; 372: 793–95. 2016; 8: 17.
14 The Precision Medicine Initiative. 2016. https://obamawhitehouse. 38 Dao MC, Everard A, Aron-Wisnewsky J, et al. Akkermansia
archives.gov/node/333101 (accessed May 21, 2017). muciniphila and improved metabolic health during a dietary
15 Fuchsberger C, Flannick J, Teslovich TM, et al. The genetic intervention in obesity: relationship with gut microbiome richness
architecture of type 2 diabetes. Nature 2016; 536: 41–47. and ecology. Gut 2016; 65: 426–36.
16 Florez JC, Udler MS, Hanson RL. Genetics of type 2 diabetes. In: 39 Guertin KA, Moore SC, Sampson JN, et al. Metabolomics in
Cowie CC, Casagrande SS, Menke A, eds. Diabetes in America, nutritional epidemiology: identifying metabolites associated with diet
3rd edn. Bethesda: National Institutes of Health, 2016. and quantifying their potential to uncover diet-disease relations in
populations. Am J Clin Nutr 2014; 100: 208–17.
17 Estrada K, Aukrust I, Bjorkhaug L, et al. Association of a
low-frequency variant in HNF1A with type 2 diabetes in a Latino 40 Jones DP, Park Y, Ziegler TR. Nutritional metabolomics: progress in
population. JAMA 2014; 311: 2305–14. addressing complexity in diet and health. Annu Rev Nutr 2012;
32: 183–202.
18 Steinthorsdottir V, Thorleifsson G, Sulem P, et al. Identification of
low-frequency and rare sequence variants associated with elevated 41 Cheung W, Keski-Rahkonen P, Assi N, et al. A metabolomic study
or reduced risk of type 2 diabetes. Nat Genet 2014; 46: 294–98. of biomarkers of meat and fish intake. Am J Clin Nutr 2017;
105: 600–08.
19 Manolio TA, Collins FS, Cox NJ, et al. Finding the missing
heritability of complex diseases. Nature 2009; 461: 747–53. 42 Stella C, Beckwith-Hall B, Cloarec O, et al. Susceptibility of human
metabolic phenotypes to dietary modulation. J Proteome Res 2006;
20 Fumagalli M, Moltke I, Grarup N, et al. Greenlandic Inuit show
5: 2780–88.
genetic signatures of diet and climate adaptation. Science 2015;
349: 1343–47. 43 Garcia-Perez I, Posma JM, Gibson R, et al. Objective assessment of
dietary patterns by use of metabolic phenotyping: a randomised,
21 Buckley MT, Racimo F, Allentoft ME, et al. Selection in Europeans
controlled, crossover trial. Lancet Diabetes Endocrinol 2017; 5: 184–95.
on fatty acid desaturases associated with dietary changes.
Mol Biol Evol 2017; 34: 1307–18. 44 Andersen MB, Rinnan A, Manach C, et al. Untargeted metabolomics
as a screening tool for estimating compliance to a dietary pattern.
22 Tanaka T, Ngwa JS, van Rooij FJ, et al. Genome-wide meta-analysis
J Proteome Res 2014; 13: 1405–18.
of observational studies shows common genetic variants associated
with macronutrient intake. Am J Clin Nutr 2013; 97: 1395–402. 45 Bouchard-Mercier A, Rudkowska I, Lemieux S, Couture P, Vohl MC.
The metabolic signature associated with the Western dietary pattern:
23 Schumann G, Liu C, O’Reilly P, et al. KLB is associated with alcohol
a cross-sectional study. Nutr J 2013; 12: 158.
drinking, and its gene product beta-Klotho is necessary for FGF21
regulation of alcohol preference. Proc Natl Acad Sci USA 2016; 46 Floegel A, von Ruesten A, Drogan D, et al. Variation of serum
113: 14372–77. metabolites related to habitual diet: a targeted metabolomic approach
in EPIC-Potsdam. Eur J Clin Nutr 2013; 67: 1100–08.
24 Cornelis MC, Kacprowski T, Menni C, et al. Genome-wide
association study of caffeine metabolites provides new insights to 47 Zheng Y, Yu B, Alexander D, Steffen LM, Boerwinkle E. Human
caffeine metabolism and dietary caffeine-consumption behavior. metabolome associates with dietary intake habits among African
Hum Mol Genet 2016; 25: 5472–82. Americans in the atherosclerosis risk in communities study.
Am J Epidemiol 2014; 179: 1424–33.
25 Cauchi S, Del Guerra S, Choquet H, et al. Meta-analysis and
functional effects of the SLC30A8 rs13266634 polymorphism on 48 Schmidt JA, Rinaldi S, Ferrari P, et al. Metabolic profiles of male meat
isolated human pancreatic islets. Mol Genet Metab 2010; eaters, fish eaters, vegetarians, and vegans from the EPIC-Oxford
100: 77–82. cohort. Am J Clin Nutr 2015; 102: 1518–26.
26 Locke AE, Kahali B, Berndt SI, et al. Genetic studies of body mass 49 Lloyd AJ, Beckmann M, Haldar S, Seal C, Brandt K, Draper J.
index yield new insights for obesity biology. Nature 2015; 518: 197–206. Data-driven strategy for the discovery of potential urinary biomarkers
of habitual dietary exposure. Am J Clin Nutr 2013; 97: 377–89.
27 Castaner O, Corella D, Covas MI, et al. In vivo transcriptomic profile
after a Mediterranean diet in high-cardiovascular risk patients: 50 Vazquez-Fresno R, Llorach R, Urpi-Sarda M, et al. Metabolomic
a randomized controlled trial. Am J Clin Nutr 2013; 98: 845–53. pattern analysis after mediterranean diet intervention in a nondiabetic
population: a 1- and 3-year follow-up in the PREDIMED study.
28 Bouchard-Mercier A, Paradis AM, Rudkowska I, Lemieux S,
J Proteome Res 2015; 14: 531–40.
Couture P, Vohl MC. Associations between dietary patterns and
gene expression profiles of healthy men and women: 51 Yang Q, Lin SL, Au Yeung SL, Kwok MK, Leung GM, Schooling CM.
a cross-sectional study. Nutr J 2013; 12: 24. Genetically predicted milk consumption and bone health, ischemic
heart disease and type 2 diabetes: a Mendelian randomization study.
29 Volkov P, Bacos K, Ofori JK, et al. Whole-genome bisulfite sequencing
Eur J Clin Nutr 2017; 71: 1008–12.
of human pancreatic islets reveals novel differentially methylated
regions in type 2 diabetes pathogenesis. Diabetes 2017; 66: 1074–85. 52 Kroger J, Schulze MB. Recent insights into the relation of
Δ5 desaturase and Δ6 desaturaseactivity to the development of type 2
30 Dayeh T, Volkov P, Salö S, et al. Genome-wide DNA methylation
diabetes. Curr Opin Lipidol 2012; 23: 4–10.
analysis of human pancreatic islets from type 2 diabetic and
non-diabetic donors identifies candidate genes that influence 53 VanderWeele TJ, Tchetgen Tchetgen EJ, Cornelis M, Kraft P.
insulin secretion. PLoS Genet 2014; 10: e1004160. Methodological challenges in mendelian randomization. Epidemiology
2014; 25: 427–35.
31 Volkmar M, Dedeurwaerder S, Cunha DA, et al. DNA methylation
profiling identifies epigenetic dysregulation in pancreatic islets from 54 Franks PW, McCarthy MI. Exposing the exposures responsible for
type 2 diabetic patients. EMBO J 2012; 31: 1405–26. type 2 diabetes and obesity. Science 2016; 354: 69–73.
55 Feillet F, Agostoni C. Nutritional issues in treating phenylketonuria. 77 Zeevi D, Korem T, Zmora N, et al. Personalized nutrition by
J Inherit Metab Dis 2010; 33: 659–64. prediction of glycemic responses. Cell 2015; 163: 1079–94.
56 Järvelä I, Torniainen S, Kolho K-L. Molecular genetics of human 78 Floegel A, Wientzek A, Bachlechner U, et al. Linking diet, physical
lactase deficiencies. Ann Med 2009; 41: 568–75. activity, cardiorespiratory fitness and obesity to serum metabolite
57 Parnell LD, Blokker BA, Dashti HS, et al. CardioGxE, a catalog of networks: findings from a population-based study.
gene-environment interactions for cardiometabolic traits. Int J Obes (Lond) 2014; 38: 1388–96.
BioData Min 2014; 7: 21. 79 Floegel A, Stefan N, Yu Z, et al. Identification of serum metabolites
58 Franks PW, Mesa JL, Harding AH, Wareham NJ. Gene-lifestyle associated with risk of type 2 diabetes using a targeted metabolomic
interaction on risk of type 2 diabetes. Nutr Metab Cardiovasc Dis 2007; approach. Diabetes 2013; 62: 639–48.
17: 104–24. 80 David LA, Maurice CF, Carmody RN, et al. Diet rapidly and
59 Li SX, Imamura F, Ye Z, et al. Interaction between genes and reproducibly alters the human gut microbiome. Nature 2014;
macronutrient intake on the risk of developing type 2 diabetes: 505: 559–63.
systematic review and findings from European Prospective 81 Falony G, Joossens M, Vieira-Silva S, et al. Population-level analysis of
Investigation into Cancer (EPIC)-InterAct. Am J Clin Nutr 2017; gut microbiome variation. Science 2016; 352: 560–64.
106: 263–75. 82 Wu GD, Chen J, Hoffmann C, et al. Linking long-term dietary
60 Qi L, Cornelis MC, Zhang C, van Dam RM, Hu FB. Genetic patterns with gut microbial enterotypes. Science 2011; 334: 105–08.
predisposition, Western dietary pattern, and the risk of type 2 diabetes 83 Zhernakova A, Kurilshikov A, Bonder MJ, et al.
in men. Am J Clin Nutr 2009; 89: 1453–58. Population-based metagenomics analysis reveals markers for gut
61 Langenberg C, Sharp SJ, Franks PW, et al. Gene-lifestyle interaction microbiome composition and diversity. Science 2016; 352: 565–59.
and type 2 diabetes: the EPIC interact case-cohort study. 84 Wu GD, Compher C, Chen EZ, et al. Comparative metabolomics in
PLoS Med 2014; 11: e1001647. vegans and omnivores reveal constraints on diet-dependent gut
62 Wang T, Heianza Y, Sun D, et al. Improving adherence to healthy microbiota metabolite production. Gut 2014; 65: 63–72.
dietary patterns, genetic risk, and long term weight gain: gene-diet 85 McLoughlin RF, Berthon BS, Jensen ME, Baines KJ, Wood LG.
interaction analysis in two prospective cohort studies. BMJ 2018; 360: Short-chain fatty acids, prebiotics, synbiotics, and systemic
j5644. inflammation: a systematic review and meta-analysis. Am J Clin Nutr
63 Do R, Xie C, Zhang X, et al. The effect of chromosome 9p21 variants 2017; 106: 930–45.
on cardiovascular disease may be modified by dietary intake: evidence 86 Dawed AY, Ali A, Zhou K, Pearson ER, Franks PW.
from a case/control and a prospective study. PLoS Med 2011; Evidence-based prioritisation and enrichment of genes interacting
8: e1001106. with metformin in type 2 diabetes. Diabetologia 2017; 60: 2231–39.
64 Qi Q, Chu AY, Kang JH, et al. Fried food consumption, genetic risk, 87 Bao Y, Zhao T, Wang X, et al. Metabonomic variations in the
and body mass index: gene-diet interaction analysis in three US drug-treated type 2 diabetes mellitus patients and healthy volunteers.
cohort studies. BMJ 2014; 348: g1610. J Proteome Res 2009; 8: 1623–30.
65 Qi Q, Chu AY, Kang JH, et al. Sugar-sweetened beverages and genetic 88 den Ouden H, Pellis L, Rutten G, et al. Metabolomic biomarkers for
risk of obesity. N Engl J Med 2012; 367: 1387–96. personalised glucose lowering drugs treatment in type 2 diabetes.
66 Brunkwall L, Chen Y, Hindy G, et al. Sugar-sweetened beverage Metabolomics 2016; 12: 27.
consumption and genetic predisposition to obesity in 2 Swedish 89 Forslund K, Hildebrand F, Nielsen T, et al. Disentangling
cohorts. Am J Clin Nutr 2016; 104: 809–15. type 2 diabetes and metformin treatment signatures in the human
67 Olsen NJ, Angquist L, Larsen SC, et al. Interactions between genetic gut microbiota. Nature 2015; 528: 262–66.
variants associated with adiposity traits and soft drinks in relation to 90 Pedersen C, Gallagher E, Horton F, et al. Host-microbiome
longitudinal changes in body weight and waist circumference. interactions in human type 2 diabetes following prebiotic fibre
Am J Clin Nutr 2016; 104: 816–26. (galacto-oligosaccharide) intake. Br J Nutr 2016; 116: 1869–77.
68 Bray GA, Fruhbeck G, Ryan DH, Wilding JP. Management of obesity. 91 Hou C, Carter B, Hewitt J, Francisa T, Mayor S. Do mobile phone
Lancet 2016; 387: 1947–56. applications improve glycemic control (HbA1c) in the
69 Franks PW, Poveda A. Lifestyle and precision diabetes medicine: self-management of diabetes? A systematic review, meta-analysis,
will genomics help optimise the prediction, prevention and treatment and GRADE of 14 Randomized Trials. Diabetes Care 2016; 39: 2089–95.
of type 2 diabetes through lifestyle therapy? Diabetologia 2017; 92 McGloin AF, Eslami S. Digital and social media opportunities for
60: 784–92. dietary behaviour change. Proc Nutr Soc 2015; 74: 139–48.
70 Florez JC, Jablonski KA, Bayley N, et al. TCF7L2 polymorphisms and 93 Obermeyer Z, Emanuel EJ. Predicting the future - big data, machine
progression to diabetes in the Diabetes Prevention Program. learning, and clinical medicine. N Engl J Med 2016; 375: 1216–19.
N Engl J Med 2006; 355: 241–50. 94 Price ND, Magis AT, Earls JC, et al. A wellness study of
71 Huang T, Ley SH, Zheng Y, et al. Genetic susceptibility to diabetes 108 individuals using personal, dense, dynamic data clouds.
and long-term improvement of insulin resistance and beta cell Nat Biotechnol 2017; 35: 747–56.
function during weight loss: the Preventing Overweight Using Novel 95 Wolever TM. Personalized nutrition by prediction of glycaemic
Dietary Strategies (POUNDS LOST) trial. Am J Clin Nutr 2016; responses: fact or fantasy? Eur J Clin Nutr 2016; 70: 411–13.
104: 198–204.
96 Afshin A, Babalola D, McLean M, et al. Information technology and
72 Wang T, Huang T, Zheng Y, et al. Genetic variation of fasting glucose lifestyle: a systematic evaluation ofiInternet and mobile interventions
and changes in glycemia in response to 2-year weight-loss diet for improving diet, physical activity, obesity, tobacco, and alcohol use.
intervention: the POUNDS LOST trial. Int J Obes (Lond) 2016; J Am Heart Assoc 2016; 5: e003058.
40: 1164–69.
97 Walford GA, Porneala BC, Dauriz M, et al. Metabolite traits and
73 The Look AHEAD Research Group. Prospective association of GLUL genetic risk provide complementary information for the prediction of
rs10911021 with cardiovascular morbidity and mortality among future type 2 diabetes. Diabetes Care 2014; 37: 2508–14.
Individuals with type 2 diabetes: The Look AHEAD Study. Diabetes
2016; 65: 297–302. 98 Aronson N. Making personalized medicine more affordable.
Ann N Y Acad Sci 2015; 1346: 81–89.
74 Zheng Y, Ceglarek U, Huang T, et al. Weight-loss diets and
2-y changes in circulating amino acids in 2 randomized intervention 99 Mozaffarian D. Dietary and policy priorities for cardiovascular
trials. Am J Clin Nutr 2016; 103: 505–11. disease, diabetes, and obesity: a comprehensive review.
Circulation 2016; 133: 187–225.
75 Walford GA, Ma Y, Clish C, Florez JC, Wang TJ, Gerszten RE.
Metabolite profiles of diabetes Incidence and intervention response in
the Diabetes Prevention Program. Diabetes 2016; 65: 1424–33.
76 Kovatcheva-Datchary P, Nilsson A, Akrami R, et al.
Dietary fiber-induced improvement in glucose metabolism is
associated with Increased abundance of Prevotella. Cell Metab 2015;
22: 971–82.