Toward systems epidemiology of coffee and health.

REVIEW URRENT C OPINION

Toward systems epidemiology of coffee and health Marilyn C. Cornelis a,b,c

Purpose of review Coffee is one of the most widely consumed beverages in the world and has been associated with many health conditions. This review examines the limitations of the classic epidemiological approach to studies of coffee and health, and describes the progress in systems epidemiology of coffee and its correlated constituent, caffeine. Implications and applications of this growing body of knowledge are also discussed. Recent findings Population-based metabolomic studies of coffee replicate coffee–metabolite correlations observed in clinical settings but have also identified novel metabolites of coffee response, such as specific sphingomyelin derivatives and acylcarnitines. Genome-wide analyses of self-reported coffee and caffeine intake and serum levels of caffeine support an overwhelming role for caffeine in modulating the coffee consumption behavior. Interindividual variation in the physiological exposure or response to any of the many chemicals present in coffee may alter the persistence and magnitude of their effects. It is thus imperative that future studies of coffee and health account for this variation. Summary Systems epidemiological approaches promise to inform causality, parse the constituents of coffee responsible for health effects, and identify the subgroups most likely to benefit from increasing or decreasing coffee consumption. Keywords coffee, gene–diet interaction, health, omics, personalized nutrition

INTRODUCTION Coffee is one of the most widely consumed beverages in the world [1]. North American coffee consumers typically drink about two cups per day; nearly twice this amount is the norm among European coffee consumers [1]. Although coffee is traditionally a beverage of Western countries, the demand for coffee is rising in other parts of the world [1,2]. With widespread popularity and availability of coffee, there is increasing public and scientific interest in the potential health consequences of its regular consumption. Coffee consumption has previously been viewed as an unhealthy habit, but meta-analyses of epidemiological studies of coffee for over 30 health outcomes confirm few risk associations and even point to beneficial associations with certain conditions [3]. These mixed findings likely explain the near absence of guidelines on the consumption of coffee for health, with the exception of special populations, such as children and pregnant women [4,5]. This review discusses the challenges in the traditional epidemiological approach to studies of coffee, new insights into coffee exposures and responses gleaned by systems epidemiology, and www.co-lipidology.com

how this knowledge might be used to optimize the future investigations of coffee and health.

CHALLENGES OF EPIDEMIOLOGICAL STUDIES OF COFFEE AND HEALTH Epidemiological studies are highly efficient and relevant approaches to investigate the role that habitual coffee intake plays in population health but have been criticized for shortcomings in study design, including exposure misclassification and confounding [6,7]. Furthermore, they do not

a

Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois, bDepartment of Nutrition, Harvard School of Public Health and cChanning Division of Network Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, USA Correspondence to Dr Marilyn C. Cornelis, Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, 680 N Lake Shore, Suite 1400, Chicago IL 60611, USA. Tel: +1 312 503 4548; fax: +1 312 908 9588; e-mail: marilyn.cornelis@northwestern. edu Curr Opin Lipidol 2015, 26:20–29 DOI:10.1097/MOL.0000000000000143 Volume 26 Number 1 February 2015

Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.

Toward systems epidemiology of coffee and health Cornelis

KEY POINTS The complex composition of coffee and the interindividual variation in the physiological exposure or response to caffeine or other coffee constituents challenge the traditional epidemiological approach to studies of coffee and health. Habitual coffee consumption correlates with blood or urine levels of caffeine, caffeine metabolites, trigonelline, quinate, and specific sphingomyelin derivatives and acylcarnitines. Genetic variants in or near AHR, CYP1A2, POR, ABCG2, GCKR, MLXIPL, BDNF and SLC6A4 are associated with habitual coffee or caffeine consumption, and likely function in the metabolism of or response to caffeine. Systems epidemiological approaches promise to inform causality, parse the constituents of coffee responsible for health effects, and identify the subgroups most likely to benefit from increasing or decreasing coffee consumption.

provide causal or mechanistic insights into the relationship between coffee and health [8].

coffee has been to compare the risk of a disease among regular coffee consumers with that among decaffeinated coffee consumers. However, regular and decaffeinated coffee consumers appear to differ in more ways than just coffee preference [19,20]. Moreover, one of several different methods can be used in the decaffeination process, and all potentially remove more than caffeine and introduce other compounds [21]. Chang et al. [22 ] recently profiled the metabolites of regular and decaffeinated coffee samples, and detected 37 metabolites at higher levels in regular than in decaf and 32 metabolites at higher levels in decaf than in regular. Among these discriminant metabolites were several benzoate-derived and cinnamate-derived phenolic compounds, organic acids, sugar, fatty acids, and amino acids. In light of its complex composition, coffee may potentially elicit a multitude of physiological effects, which may impact a number of health conditions in different ways. For example, caffeine reportedly enhances the activation and carcinogenic potential of environmental mutagens [23], stimulates and suppresses tumors [24], and reduces neurodegeneration and amyloid-b production [25]. Polyphenols have antioxidant activities and favorable roles in fatty acid oxidation and glucose homeostasis [26–28]. &&

Coffee is a complex exposure The Food Frequency Questionnaire (FFQ) is the most common dietary assessment tool used in the large epidemiologic studies of coffee and provides a relatively accurate assessment of habitual coffee intake [9–11]. However, the precise chemical composition of different coffee preparations is not captured by standard FFQs, and is likely to vary within and between populations. For most populations, coffee is the main source of caffeine, which is by far the bestcharacterized naturally occurring component of the beverage. Owing to the strong collinearity between caffeine and coffee in many populations, it has been difficult to assess whether caffeine or other compounds found in regular coffee may be responsible for associations with disease or related traits. Indeed, caffeine is just one of the hundreds of biologically active chemicals present in coffee [12–18]. For example, brewed coffee contains melanoidins and lipid-soluble heterocyclic compounds such as furans, pyrroles, and maltol, and in many countries coffee is the richest dietary source of natural phenolics [12– 16]. Boiled or unfiltered coffee contains diterpenoid alcohols, including cafestol and kahweol [18]. The precise chemical composition of a brewed cup of coffee will depend on a series of factors, from bean species to brewing method [12]. A common epidemiological strategy to detangle the effects of caffeine from other substances in

Individuals vary in their physiological exposure and response to coffee Variation in the metabolism of or physiological response to any of the many chemicals present in coffee may alter the persistence and magnitude of their individual effects. Caffeine metabolism varies two-fold to 12-fold between individuals, and this is largely attributable to the variation in the activity of cytochrome P450 (CYP)1A2, which accounts for more than 95% of caffeine metabolism [29–31]. The main, but variable, psychological and physiological effects of caffeine in humans are reportedly due to competitive inhibition of central and peripheral adenosine receptors; this inhibition indirectly modulates the release of dopamine, norepinephrine, serotonin, acetylcholine, g-aminobutyric acid, and glutamate [32,33]. Environmental factors such as smoking, oral contraceptive use, pregnancy, and habitual alcohol and caffeine intake alter CYP1A2 activity [30,31,34]. Genetic factors also influence caffeine metabolism, as well as the pathways mediating its effects (discussed below). Variation in the metabolism of or response to other coffee constituents is less understood [15,35], but will nevertheless at least be modulated by the variable duration and magnitude of effects exerted by caffeine. Not accounting for these factors will magnify the variability of response to a given amount of coffee

0957-9672 Copyright ß 2015 Wolters Kluwer Health, Inc. All rights reserved.

www.co-lipidology.com


21

Nutrition and metabolism

consumed by a population, resulting in potentially different associations across the studies.

Advancements in high-throughput measurement and analyses of genomic, transcriptomic, proteomic, and metabolomic traits present epidemiologists with an unprecedented opportunity to optimize their research approach while acquiring mechanistic and causal insight to their observed associations. ‘Systems epidemiology’ couples traditional epidemiologic methods with ‘omic’ advancements to enhance the biological understanding of metabolic pathways in humans [36 – 38,39 ]. Coffee research is ideally suited for the application of systems approaches, and this has been demonstrated by a series of recent metabolomic and genomic studies.

correlation with a medium-chain acylcarnitines (C10:1) was later confirmed by Menni et al. [53] in a twin study of women. In larger analyses of the EPICPostdam cohort [51,52], dietary patterns of coffee and other foods correlated with specific lysophosphatidylcholines (LPCs) and sphingomyelin derivatives [51]. An exploratory analysis specific to coffee, however, did not yield significant results [52]. Sphingomyelin is a major phospholipid in lipid rafts, which are active centers for lipid and glucose metabolism [56,57]. Acylcarnitine esters facilitate the transport of fatty acids across the mitochondrial membrane for b-oxidation [58,59]. The physiological functions of specific sphingomyelin, acylcarnitines, and LPCs and their role in health are active areas of investigation. Potential mechanisms linking coffee to these specific metabolites is therefore unclear, and the cross-sectional design of these reports limits the causal inferences.

Metabolite markers of coffee exposure and response

Genetic markers of coffee exposure and response

Over 41 000 metabolites that have been identified in human specimens to date, as listed in the Human Metabolome Database [40]. Each metabolite represents a substrate (‘exposure’) for or a product (‘effect’, ‘response’) of a cellular process, and the comprehensive analysis of these metabolites under a given set of conditions defines ‘metabolomics’ [41,42]. Earlier applications to coffee research targeted an a priori list of coffee-derived metabolites and were performed in small controlled trial settings. Blood or urine levels of methylxanthines (i.e., caffeine and its metabolites), N-methylpyrimidine, trigonelline, and derivatives of chlorogenic acid and hydroxycinnamate increase after the recent intake of a specified quantity of coffee [43–48]. Populationbased metabolomic studies of coffee intake have been largely ‘untargeted’ and cross-sectional by design (Table 1). Untargeted studies are optimal for broad coverage of the metabolome to enhance the opportunities of discovering discriminatory metabolites or metabolic ‘features’ of habitual coffee intake. Two population-based studies and a third study that specifically compared nonhabitual coffee drinkers and high coffee drinkers each observed positive correlations between coffee intake and serum methylxanthines, as well as other coffee derivatives [53,54,55 ]. Studies exploring various other classes of metabolites have identified novel features correlated with habitual intake [49–52]. In an earlier study of men, specific sphingomyelin derivatives were positively correlated and longchain and medium-chain acylcarnitines were negatively correlated with coffee intake [49]. The

Genomics may also be used to parse the constituents of coffee with biological effects while further accounting for individual variation in these effects. Advantages in using genomic data over other omic data are that they are easily measured, are completely stable, and can be used to inform causality. Disadvantages pertain to the identification of causal single nucleotide polymorphisms (SNPs) and their generally small effect sizes. Prior to genome-wide analyses, genetic epidemiological studies of coffee focused on a limited number of SNPs in known candidate genes related to caffeine metabolism (CYP1A2) or its target of action (ADORA2A, ADORA1, and DRD2) [3,60 ]. Candidate variants in genes reportedly involved in metabolism of other coffee constituent classes, such as diterpenes (UGTs and SULTs), mutagens (NAT2, CYP1A2, CYP1A1, and UGTs), flavonoids (CYP1A2, CYP2D6, CYP2C09, CYP3A4, CYP1A1, CYP2E1, UGTs, and SULTs), and polyphenols (UGTs, GSTs, and SULTs), have also been considered [6]. In 2010, three independent genome-wide association studies (GWAS) of selfreported habitual coffee and caffeine intake behavior were conducted among European-populationbased studies [61–63]. Two studies identified significant associations upstream of AHR [61,62], and all three identified associations mapping to the CYP1A1–CYP1A2 bidirectional promoter. The aryl hydrocarbon receptor (encoded by AHR) plays a key role in regulating the expression of a number of genes, including CYP1A1 and CYP1A2 [64]. CYP1A2 is an established candidate gene in caffeine metabolism, as already discussed. Despite the intrinsic and

SYSTEMS EPIDEMIOLOGY OF COFFEE

&

&&

22


&

Volume 26 Number 1 February 2015


Toward systems epidemiology of coffee and health Cornelis Table 1. Population-based metabolomic studies of habitual coffee consumptiona Metabolite measurementc

Significant findings

FFQ

Overnight fasting serum, morning draws, 2006

Coffee intake was positively associated with sphingomyelin species:

2004–2005 cups/day

ESI-MS/MS (Biocrates): 363 metabolites

Reference

Study sample

Coffee measurementb

[49]

Germany, EUd n ¼ 284, male; Age: 55–79 years

Sphingomyelin (OH,COOH) x:y– 20:2, 16:2, 18:2, 24:0, 18:1 Sphingomyelin (OH) x:y – 20:3, 22:1, 28:0 Coffee intake was negatively associated with long-chain and mediumchain acylcarnitines: C16:1, C10:1, C12:1, C14:1, C6.

[50]

[51]

[52]

[53]

UK, EU

FFQ

Overnight fasting serum

Coffee intake was negatively associated with acylcarnitine C10:1.

n ¼ 1003, Female; Twins; Age: 58.5 10.45 years

Categories (0 to 6þ cups/day)

Absolute-IDQTM Kit p150 (Biocrates): 126 metabolites

Germany, EU

FFQ

Serum, fasting status unknown, 1994–1998

Diet pattern with high intake of margarine, nonwhole-grain bread, meat, and coffee and low intake of butter, pasta or rice, and tea was positively associated with LPCs: C20:4, C18:2.

n ¼ 2380, male/female; Age: 49.8 8.9 years

1994–1998 categories (0 to 5þ cups/day)

Absolute-IDQTM Kit p150 (Biocrates): 127 metabolites

Diet pattern with high intake of butter, garlic, and coffee and low intake of margarine, fresh fruit, and soup was positively associated with sphingomyelins, notably OH-C16:1, OH-C14:1, OHC24:1, OH-C22:2, OH-C22:1

Germany, EU

FFQ

Hypothesis testing:

n ¼ 1610, male/female; Age: 35–64 years

1994–1998 Categories (0 to 5þ cups/day)

Serum, fasting status unknown, 1994–1998 Absolute-IDQTM Kit p150 (Biocrates)

Coffee was inversely associated with diacyl-phosphatidylcholine C32:1 in male/female.

Hypothesis testing: 13 metabolites previously associated with type 2 diabetes

Coffee was positively associated with acyl-alkyl-phosphatidylcholines C34:3, C40:6, and C42:5 in female.

Exploratory: 113 metabolites

Exploratory: none significantly associated with coffee. Coffee intake was positively associated with quinate, paraxanthine, 5-acetylamino-6-amino-3-methyluracil, 1,7-dimethylurate, 1-methylurate, 1-methylxanthine, caffeine, 1,3,7-trimethylurate, and 7methylxanthine.

USA, AA

FFQ

8-h and fasting serum, 1987– 1989

Discovery: n ¼ 1500, male/female; Age: 52.9 5.8 years

1987–1989 categories (0 to 6þ cups/day)

GC/LC-MS (Metabolon): 118 metabolites

FFQ

Serum, fasting status unknown

Replication: n ¼ 477, male/female; Age: 52.7 5.7 years [54]

USA, EU


Coffee intake was positively associated with trigonelline, quinate, 1methylxanthine, paraxanthine, N2-furoyl-glycine, catechol sulfate, and nine other ‘unknown’ metabolites.



23

Nutrition and metabolism Table 1 (Continued) Reference

[55 ] &&

Metabolite measurementc

Study sample

Coffee measurementb

n ¼ 502, male/female (cancer cases and matched controls); Age: 64 5 years

1993–2001 cups/day

GC/LC-MS (Metabolon): 412 ‘knowns’ and 231 ‘unknowns’ detected

France, EU

Multiple 24-h dietary recalls

Morning spot urine

132 Total ions (60 metabolites) significantly different.

UPLC–QTOF–MS: 1111 total ions detected

Strongest contributors to discrimination of low and high coffee consumers (all increase with coffee intake): atractyligenin glucuronide, cyclo(isoleucyl-prolyl), 1methylxanthine, 1,7-dimethyluric acid, kahweol oxide glucuronide, 1-methyluric acid, trigonelline, dimethylxanthine (paraxanthine or theophylline) glucuronide, AFMU, kahweol oxide glucuronide analog, hippuric acid, trimethyluric acid, 3hydroxyhippuric acid, 1,3dimethyluric acid or 3,7-dimethyluric acid, and caffeine.

Male/female; 19 Coffee nonconsumers; 20 High coffee consumers (>180 ml/day)

Significant findings

FFQ, Food Frequency Questionnaire. AFMU, 5-acetylamino-6-formylamino-3-methyluracil; ESI-MS, electrospray ionisation - mass spectrometry; GC/LC-MS, gas chromatography/liquid chromatography - mass spectrometry; QTOF, quadrupole-time-of-flight; UPLC, ultra performance liquid chromatography. a Studies examined predominately regular or total (regular and decaffeinated) coffee. b Includes method of dietary assessment, year of collection, and type of data collected (if reported). c Includes fasting status at time of specimen collection, year of blood and urine collection, and platform used for metabolomic profiling (if reported). d Population ancestry: EU – European, AA – African American; sex: male, female.

extrinsic challenges in ascertaining habitual coffee and caffeine consumption, these successes served as proof-of-principle for an agnostic approach to gene discovery. In 2014, a greater effort to discover additional loci linked to coffee and caffeine behavior was pursued by the Coffee and Caffeine Genetics Consortium [65] (Table 2). Eight loci, including six novel loci, met genome-wide significance, with per allele effect sizes of 0.03–0.14 cups per day of predominately regular coffee. Taken together, these loci explain approximately 1.3% of the phenotypic variance of coffee intake among Westernized populations [65]. An overview of each locus and their postulated roles in coffee intake behavior are provided in Table 3. Loci near AHR, CYP1A2, POR, and ABCG2 potentially impact consumption behavior by altering the metabolism of caffeine and thus the physiological levels of this stimulant. Other loci near BDNF and SLC6A4 likely act directly by modulating the acute psychological rewarding properties of caffeine. The consortium presented the first links between a behavioral trait and variation in GCKR and MLXIPL [65]. Whether these associations are direct or mediated by the established roles of these genes in glucose and lipid metabolism is unknown but is an avenue of future study. 24


Findings from the population-based GWAS of serum caffeine partly align with the results from GWAS of self-reported caffeine and coffee intake: variants near AHR and CYP1A2 linked to increased intake are also linked to lower caffeine serum levels [66 ]. Additionally, SNPs near NAT2 and NAT1 are associated with 1-methylxanthine and 1-methyluric acid serum levels, a result that is plausible but somewhat unexpected in light of the previous literature pointing to other enzymes in this pathway [66 ]. &&

&&

IMPLICATIONS AND APPLICATIONS Marked variation in caffeine metabolism is well established, and much progress has been made in identifying the factors contributing to this variation. Although variation in the metabolism of other coffee constituents is unclear, their overall physiological effects will be modulated by those of caffeine (which we know are variable). Genetic evidence showing that individuals naturally modulate their coffee intake for optimal effects exerted by caffeine only further magnifies the variability in physiological exposure to noncaffeine constituents of coffee. Genetic variants in Table 3 linked to higher coffee and caffeine intake have also been Volume 26 Number 1 February 2015


Toward systems epidemiology of coffee and health Cornelis Table 2. Genome-wide association studies of habitual coffee and caffeine consumption Reference

Study design and samplesa

Primary phenotype

Significant locib

[61]

One-stage meta-analysis

Total dietary caffeine, mg/day

7p21

Five studies, EUc

FFQ: categories of regular coffee, tea, soda, chocolate

15q24

Mean age range: 48–68 years

Transformation: cubic root

Total n 47 341

Mean phenotype range: 231–491 mg/day

Two-stage joint meta-analysis

Regular coffee, cups/day

7p21

Stage 1:

FFQ: categories of intake

15q24

Adjusted for smoking: yes [62]

Four studies, EU

Transformation: none


Mean phenotype range: 1.8–4.2 cups/day

Total n 6611

Adjusted for smoking: yes

Stage 2: Two studies, EU Mean age range: 30–50 years Total n 4050 [63]

Two-stage joint meta-analysis

Regular coffee, cups/day

Stage 1:

FFQ: categories of intake or cups/day

Eight studies, EU

Transformation: categories of intake



Total n 18 176

Adjusted for smoking: no

15q24

Stage 2: One study, EU Mean age: 47 years Total n ¼ 7929 [65]

Two-stage joint meta-analysis:

Regular coffee, cups/day among coffee consumers

7p21

Stage 1:

FFQ and diet records: categories of intake or cups/day

15q24

28 Studies, EU

Transformation: none

7q11.23 (POR)



7q11.23 (MLXIPL)

Total n 91 462

Adjusted for smoking: yes

2p24

Stage 2:

4q22

13 studies, EU

11p13

Mean age: 32–75 years

17q11.2

Total n 30 062 7 Studies, AA Mean age: 50–75 years Total n 7964 FFQ, Food Frequency Questionnaire. a All contributing studies were population-based studies, and the final meta-analysis included men and women. b Ranked according to P-value. c Population ancestry: EU – European, AA – African American.

associated with smoking initiation and higher adiposity and fasting insulin and glucose, but with lower blood pressure (BP) and favorable lipid, inflammatory, and liver enzyme profiles in GWAS [67]. Whether these relationships reflect pleiotropy or confounding or offer insights into the potential causal role coffee plays in these traits merits further investigation. Future Mendelian randomization and gene–coffee interaction studies

will need to consider the direct and indirect roles that each SNP has in altering coffee drinking behavior as well as the potential for pleiotropic effects and interactions between loci that may violate assumptions underlying these statistical approaches. These data nevertheless substantiate the broader implications of knowledge acquired through system approaches to coffee and health in populations.




25

Nutrition and metabolism Table 3. Genome-wide significant loci associated with habitual coffee or caffeine consumption Locus (index SNPa)

Closest gene(s)

Encoded protein(s): function (UniProtKb)

Hypothesized link to caffeine or coffee consumption

2p24 (rs1260326)

GCKR

Glucokinase regulatory protein (GKRP): Inhibits glucokinase by forming an inactive complex with this enzyme.

Response to caffeine or coffee: may function in the glucose-sensing process of the brain that may influence central pathways responding to caffeine/coffee.

4q22 (rs1481012)

ABCG2

ATP-binding cassette sub-family G member 2: high-capacity urate exporter. Plays a role in porphyrin homeostasis and cellular export of hemin and heme. May play an important role in the exclusion of xenobiotics from the brain. Implicated in the efflux of numerous drugs and xenobiotics.

Metabolism of caffeine: caffeine/metabolite efflux transporter.

7p21 (rs4410790, rs6968554)

AHR

Aryl hydrocarbon receptor: ligand-activated transcriptional activator. Activates the expression of multiple phase I and II xenobiotic metabolizing enzymes. Involved in cell-cycle regulation and likely plays a role in the development and maturation of many tissues.

Metabolism of caffeine: regulates CYP1A2 expression.

7q11.23 (rs7800944)

MLXIPL

Carbohydrate-responsive element-binding protein: transcri ptional repressor.

Response to caffeine or coffee: may regulate transcription of genes (e.g. GCKR) implicated in the response to caffeine.

7q11.23 (rs17685)

POR

NADPH-cytochrome P450 reductase: required for electron transfer from NADP to cytochrome P450 in microsomes and can also facilitate electron transfer to heme oxygenase and cytochrome B5.

Metabolism of caffeine: required for CYP1A2 catalytic activity.

11p13 (rs6265)

BDNF

Brain-derived neurotrophin factor: during development, promotes survival and differentiation of selected neuronal populations of the PNS and CNS. Major regulator of synaptic transmission and plasticity at adult synapses in many regions of the CNS.

Response to caffeine: modulates neurotransmitters potentially mediating the rewarding response to caffeine.

15q24 (rs2470893, rs2472297)

CYP1A1, CYP1A2

Cytochrome P450 1A1/2: cytochromes P450 are a group of enzymes involved in NADPHdependent electron transport pathways. They oxidize a variety of compounds, including steroids, fatty acids, and xenobiotics.

Metabolism of caffeine: CYP1A2 metabolizes >95% of caffeine.

17q11.2 (rs9902453)

EFCAB5; SLC6A4

EF-hand calcium-binding domain-containing protein 5: unknown

Response to caffeine or coffee: serotonin may mediate the rewarding response to caffeine.

Sodium-dependent serotonin transporter: in CNS, regulates serotonergic signaling via transport of serotonin molecules from the synaptic cleft back into the presynaptic terminal for reuse. a Index SNPs reported by the Coffee and Caffeine Genetics Consortium [65]. Genic SNPs are in boldface. CNS, central nervous system; PNS, peripheral nervous system; SNP, single nucleotide polymorphism.

Metabolomic studies of coffee are recent, and the cross-sectional nature of their design warrants additional follow-up of novel coffee metabolites (Table 1). In contrast, a set of confirmed environmental and genetic factors contributing to variation in coffee metabolism exist, and an increasing number of studies are now accounting for these factors in the population studies of coffee and health. As reviewed in detail elsewhere [3], over 30 gene–coffee interaction studies have been published to date. Studies 26


of cancers, cardiovascular disease, Parkinson’s disease, and pregnancy outcomes yield promising but preliminary results. Most have targeted the caffeine component of coffee but have examined a limited number of SNPs. None have considered the SNPs recently identified in GWAS. The intronic CYP1A21F (C734A, rs762551) polymorphism has been considered in the most genetic epidemiological studies of coffee, caffeine, and other CYP1A2 substrates. Studies suggest that the caffeine component Volume 26 Number 1 February 2015


Toward systems epidemiology of coffee and health Cornelis

of coffee may have adverse cardiovascular effects, but that these effects are limited to individuals with the genotype corresponding to impaired or slower caffeine metabolism [3]. Further, individuals who can rapidly metabolize caffeine may benefit from other compounds present in coffee. Motivated by GWAS reporting associations between CYP1A2 variants and both caffeine intake [61] and BP [68,69], Guesssous et al. [70] designed a thoughtful study to explore, in part, the effects of CYP1A21F on hypertension, focusing on caffeine as the potential mediator of CYP1A2 effects. The CYP1A2 genotype corresponding to rapid caffeine metabolism was associated with a lower risk of hypertension among nonsmokers, and this was mediated by caffeine intake, which was also inversely associated with hypertension. These associations were not observed among smokers, possibly because of the strong CYP1A2-inducing effect of smoking.

CONCLUSION Given the widespread popularity and availability of coffee, even small effects this beverage might have on a disease will translate to a substantial proportion of that disease related to coffee (i.e., attributable risk). Further, coffee consumption is a modifiable behavior (for most) and thus a highly relevant target for disease prevention strategies. Traditional epidemiology studies report ‘marginal’ associations and are challenged by the variable and complex nature of coffee itself, as well as individual responses to this beverage. Indeed, the strength and direction of associations with health conditions could be attributable to any one or more components of coffee and might vary in different subgroups of populations. Meta-analyzing associations across populations magnify this variability which contributes to several known pitfalls of this common epidemiological practice. Systems epidemiological approaches promise to study the impact of coffee on health more comprehensively. The recent integration of metabolomics into population-based studies replicates coffee– metabolite correlations observed in clinical settings but also sheds new light on the molecular pathophysiological responses to coffee. GWAS of self-reported coffee and caffeine intake and serum levels of caffeine support an overwhelming role for caffeine in modulating the coffee consumption behavior. Genetic control of this behavior would incidentally govern exposure to other potentially bioactive constituents of coffee that may relate to the health effects of coffee or other sources of caffeine. By combining the measures of habitual coffee intake behavior with measures of environmental and genetic factors modulating caffeine metabolism,

we can now devise a more refined measure of biological internal dose of the caffeine component of coffee for use in epidemiological studies. More attention will be needed to extend this framework to other constituents of coffee, as well as markers of coffee response. With its affordability, stability, and ease of measurement and analysis, genomics has thus far attracted more attention than other omics in the epidemiology field and will likely remain a fundamental systems tool in this research setting. No human transcriptomic, proteomic, or other emerging omic studies of coffee have been conducted to date. As technologies for capturing other omic information advances and becomes affordable, these should naturally be integrated with genomics. ‘Integrative personal omics profiling’ [71 ] seeks to model all systems simultaneously and, if applied in the context of a coffee trial, would be a novel and efficient approach to identify the comprehensive profiles of exposure and response to coffee. These profiles may subsequently be targeted in a populationbased setting for disease outcomes. Systems epidemiology has the potential to provide new insight into the underlying mechanisms and ways to study the health effects of coffee more comprehensively. Findings thus far provide compelling evidence in support of important interindividual variation in caffeine metabolism that needs to be accounted for in any population study of coffee. Continued advances in high-throughput omic technology, bioinformatics and analytical tools, and reduced costs will enable more widespread application of system techniques for coffee and health research. &&

Acknowledgements None. Financial support and sponsorship M.C. is supported, in part, by the American Diabetes Association grant 7-13-JF-15. Conflicts of interest None.

REFERENCES AND RECOMMENDED READING Papers of particular interest, published within the annual period of review, have been highlighted as: & of special interest && of outstanding interest 1. International Coffee Organization. www.ico.org. [Accessed 1 July 2014]. 2. United States Department of Agriculture. Coffee: world markets and trade. 2012.




27

Nutrition and metabolism 3. Cornelis M. Gene–coffee interactions and health. Curr Nutr Rep 2014; 3:178–195. 4. March of Dimes Foundation. Available at http://www.marchofdimes.com/ pregnancy/print/caffeine-in-pregnancy.html. 5. Health Canada. Available at http://www.hc-sc.gc.ca/fn-an/securit/addit/caf/ food-caf-aliments-eng.php. 6. Cornelis MC, El-Sohemy A. Coffee, caffeine, and coronary heart disease. Curr Opin Clin Nutr Metab Care 2007; 10:745–751. 7. Schreiber GB, Robins M, Maffeo CE, et al. Confounders contributing to the reported association of coffee or caffeine with disease. Prev Med 1988; 17:295–309. 8. Institute of Medicine Committee for the Study of the Future of Public Health. The future of public health. Washington, DC: National Academy Press; 1988. 9. Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a Semiquantitative Food Frequency Questionnaire. Am J Epidemiol 1985; 122:51–65. 10. Stevens J, Metcalf P, Dennis B, et al. Reliability of a Food Frequency Questionnaire by ethnicity, gender, age and education. Nutr Res 1996; 16:735–745. 11. Paalanen L, Mannisto S, Virtanen MJ, et al. Validity of a Food Frequency Questionnaire varied by age and body mass index. J Clin Epidemiol 2006; 59:994–1001. 12. Spiller MA. The chemical components of coffee. In: Spiller GA, editor. Caffeine. Boca Raton: CRC; 1998. pp. 97–161. 13. Borrelli RC, Visconti A, Mennela C, et al. Chemical characterization and antioxidant properties of coffee melanoidins. J Agric Food Chem 2002; 50:6527–6533. 14. Yanagimoto K, Ochi H, Lee KG, Shibamoto T. Antioxidative activities of fractions obtained from brewed coffee. J Agric Food Chem 2004; 52:592–596. 15. Scalbert A, Williamson G. Dietary intake and bioavailability of polyphenols. J Nutr 2000; 130:2073S–2085S. 16. Clifford MN. Chlorogenic acids and other cinnamates – nature, occurance, dietary burden, absorption and metabolism. J Sci Food Agric 2000; 80:1033–1043. 17. Coffee, tea, mate, methylxanthines and methylglyoxal. IARC Working Group on the evaluation of carcinogenic risks to humans. Lyon, 27 February to 6 March 1990. IARC Monogr Eval Carcinog Risks Hum 1991; 51:1–513. 18. Urgert R. Levels of the cholesterol-elevating diterpenes cafestol and kahweol in various coffee brews. J Agric Food Chem 1995; 43:2167–2172. 19. Shlonsky AK, Klatsky AL, Armstrong MA. Traits of persons who drink decaffeinated coffee. Ann Epidemiol 2003; 13:273–279. 20. Cornelis MC, El-Sohemy A. Coffee, caffeine, and coronary heart disease. Curr Opin Lipidol 2007; 18:13–19. 21. Katz S. Decaffeination of coffee. In: Clarke R, Macrae R, editors. Coffee. Springer: the Netherlands; 1987. pp. 59–71. 22. Chang KL, Ho PC. Gas chromatography time-of-flight mass spectrometry && (GC-TOF-MS)-based metabolomics for comparison of caffeinated and decaffeinated coffee and its implications for Alzheimer’s disease. PLoS One 2014; 9:e104621. The first metabolomic study comparing regular and decaffeinated coffee. Differences between these beverages extend beyond their caffeine content, which has implications for interpreting traditional epidemiological studies of coffee intake that compare the risks of disease among regular coffee consumers vs. decaffeinated coffee consumers. 23. Zhou T, Chen Y, Huang C, Chen G. Caffeine induction of sulfotransferases in rat liver and intestine. J Appl Toxicol 2012; 32:804–809. 24. World Cancer Research Fund/American Institute for Cancer Research. Food, nutrition, physical activity and the prevention of cancer: a global perspective. Washington, DC: American Institute for Cancer Research; 2007. 25. Arendash GW, Cao C. Caffeine and coffee as therapeutics against Alzheimer’s disease. J Alzheimers Dis 2010; 20 (Suppl. 1):S117–S126. 26. Natella F, Nardini M, Giannetti I, et al. Coffee drinking influences plasma antioxidant capacity in humans. J Agric Food Chem 2002; 50:6211–6216. 27. Rodriguez de Sotillo DV, Hadley M, Sotillo JE. Insulin receptor exon 11þ/ is expressed in Zucker (fa/fa) rats, and chlorogenic acid modifies their plasma insulin and liver protein and DNA. J Nutr Biochem 2006; 17:63–71. 28. Johnston KL, Clifford MN, Morgan LM. Coffee acutely modifies gastrointestinal hormone secretion and glucose tolerance in humans: glycemic effects of chlorogenic acid and caffeine. Am J Clin Nutr 2003; 78:728–733. 29. Benowitz NL. Clinical pharmacology of caffeine. Annu Rev Med 1990; 41:277–288. 30. Gunes A, Dahl ML. Variation in CYP1A2 activity and its clinical implications: influence of environmental factors and genetic polymorphisms. Pharmacogenomics 2008; 9:625–637. 31. Zhou SF, Wang B, Yang LP, Liu JP. Structure, function, regulation and polymorphism and the clinical significance of human cytochrome P450 1A2. Drug Metab Rev 2010; 42:268–354. 32. Fredholm BB, Battig K, Holmen J, et al. Actions of caffeine in the brain with special reference to factors that contribute to its widespread use. Pharmacol Rev 1999; 51:83–133. 33. Ferre S. Role of the central ascending neurotransmitter systems in the psychostimulant effects of caffeine. J Alzheimers Dis 2010; 20 (Suppl. 1):S35–S49.

28


34. Thorn CF, Aklillu E, Klein TE, Altman RB. PharmGKB summary: very important pharmacogene information for CYP1A2. Pharmacogenet Genomics 2012; 22:73–77. 35. Dorne JL, Walton K, Renwick AG. Human variability in xenobiotic metabolism and pathway-related uncertainty factors for chemical risk assessment: a review. Food Chem Toxicol 2005; 43:203–216. 36. Haring R, Wallaschofski H. Diving through the ‘-Omics’: the case for deep phenotyping and systems epidemiology. Omics 2012; 16:231–234. 37. Lund E, Dumeaux V. Systems epidemiology in cancer. Cancer Epidemiol Biomarkers Prev 2008; 17:2954–2957. 38. Bictash M, Ebbels TM, Chan Q, et al. Opening up the ‘Black Box’: metabolic phenotyping and metabolome-wide association studies in epidemiology. J Clin Epidemiol 2010; 63:970–979. 39. Cornelis MC, Hu FB. Systems epidemiology: a new direction in nutrition and & metabolic disease research. Curr Nutr Rep 2013; 2:225–235. One of a series of the recent reviews describing the concept of ‘Systems Epidemiology’. This particular article discusses the progress in applying system-level tools to the study of both type 2 diabetes (T2D) and nutrition. 40. Wishart DS, Jewison T, Guo AC, et al. HMDB 3.0 – the Human Metabolome Database in 2013. Nucleic Acids Res 2013; 41:D801–D807. 41. Norheim F, Gjelstad IM, Hjorth M, et al. Molecular nutrition research: the modern way of performing nutritional science. Nutrients 2012; 4:1898 – 1944. 42. Rezzi S, Ramadan Z, Fay LB, Kochhar S. Nutritional metabonomics: applications and perspectives. J Proteome Res 2007; 6:513–525. 43. Allard E, Backstrom D, Danielsson R, et al. Comparing capillary electrophoresis-mass spectrometry fingerprints of urine samples obtained after intake of coffee, tea, or water. Anal Chem 2008; 80:8946–8955. 44. Redeuil K, Smarrito-Menozzi C, Guy P, et al. Identification of novel circulating coffee metabolites in human plasma by liquid chromatography-mass spectrometry. J Chromatogr A 2011; 1218:4678–4688. 45. Nagy K, Redeuil K, Williamson G, et al. First identification of dimethoxycinnamic acids in human plasma after coffee intake by liquid chromatographymass spectrometry. J Chromatogr A 2011; 1218:491–497. 46. Stalmach A, Mullen W, Barron D, et al. Metabolite profiling of hydroxycinnamate derivatives in plasma and urine after the ingestion of coffee by humans: identification of biomarkers of coffee consumption. Drug Metab Dispos 2009; 37:1749–1758. 47. Fumeaux R, Menozzi-Smarrito C, Stalmach A, et al. First synthesis, characterization, and evidence for the presence of hydroxycinnamic acid sulfate and glucuronide conjugates in human biological fluids as a result of coffee consumption. Org Biomol Chem 2010; 8:5199–5211. 48. Lang R, Wahl A, Skurk T, et al. Development of a hydrophilic liquid interaction chromatography–high-performance liquid chromatography–tandem mass spectrometry based stable isotope dilution analysis and pharmacokinetic studies on bioactive pyridines in human plasma and urine after coffee consumption. Anal Chem 2010; 82:1486–1497. 49. Zheng Y, Yu B, Alexander D, et al. Human metabolome associates with dietary intake habits among african americans in the atherosclerosis risk in communities study. Am J Epidemiol 2014; 179:1424–1433. 50. Guertin KA, Moore SC, Sampson JN, et al. Metabolomics in nutritional epidemiology: identifying metabolites associated with diet and quantifying their potential to uncover diet-disease relations in populations. Am J Clin Nutr 2014; 100:208–217. 51. Rothwell JA, Fillatre Y, Martin JF, et al. New biomarkers of coffee consumption identified by the nontargeted metabolomic profiling of cohort study subjects. PLoS One 2014; 9:e93474. 52. Altmaier E, Kastenmuller G, Romisch-Margl W, et al. Variation in the human lipidome associated with coffee consumption as revealed by quantitative targeted metabolomics. Mol Nutr Food Res 2009; 53:1357 – 1365. 53. Menni C, Zhai G, Macgregor A, et al. Targeted metabolomics profiles are strongly correlated with nutritional patterns in women. Metabolomics 2013; 9:506–514. 54. Floegel A, von Ruesten A, Drogan D, et al. Variation of serum metabolites related to habitual diet: a targeted metabolomic approach in EPIC-Potsdam. Eur J Clin Nutr 2013; 67:1100–1108. 55. Jacobs S, Kroger J, Floegel A, et al. Evaluation of various biomarkers as && potential mediators of the association between coffee consumption and incident type 2 diabetes in the EPIC-Potsdam Study. Am J Clin Nutr 2014; 100:891–900. In addition to an exploratory metabolomic study of habitual coffee intake (Table 1), these authors investigated the association between coffee consumption and diabetes-related biomarkers and their potential role as mediators of the association between coffee consumption and T2D. Overall, the biomarkers explained only a small extent of the inverse association between long-term coffee consumption and T2D. 56. Hla T, Dannenberg AJ. Sphingolipid signaling in metabolic disorders. Cell Metab 2012; 16:420–434. 57. Chakraborty M, Jiang XC. Sphingomyelin and its role in cellular signaling. Adv Exp Med Biol 2013; 991:1–14. 58. Mingrone G. Carnitine in type 2 diabetes. Ann N Y Acad Sci 2004; 1033:99–107.

Volume 26 Number 1 February 2015


Toward systems epidemiology of coffee and health Cornelis 59. Mihalik SJ, Goodpaster BH, Kelley DE, et al. Increased levels of plasma acylcarnitines in obesity and type 2 diabetes and identification of a marker of glucolipotoxicity. Obesity (Silver Spring) 2010; 18:1695–1700. 60. Cornelis MC. Coffee intake. Prog Mol Biol Transl Sci 2012; 108:293–322. &

This book chapter reviews the genetics of coffee intake including heritability, candidate and genome-wide studies, and the application of the findings to studies of gene–coffee interactions. 61. Cornelis MC, Monda KL, Yu K, et al. Genome-wide meta-analysis identifies regions on 7p21 (AHR) and 15q24 (CYP1A2) as determinants of habitual caffeine consumption. PLoS Genet 2011; 7:e1002033. 62. Sulem P, Gudbjartsson DF, Geller F, et al. Sequence variants at CYP1A1– CYP1A2 and AHR associate with coffee consumption. Hum Mol Genet 2011; 20:2071–2077. 63. Amin N, Byrne E, Johnson J, et al. Genome-wide association analysis of coffee drinking suggests association with CYP1A1/CYP1A2 and NRCAM. Mol Psychiatry 2011; 17:1116–1129. 64. Nukaya M, Moran S, Bradfield CA. The role of the dioxin-responsive element cluster between the Cyp1a1 and Cyp1a2 loci in aryl hydrocarbon receptor biology. Proc Natl Acad Sci USA 2009; 106:4923–4928. 65. Cornelis M, Byrne E, Esko T, et al. Genome-wide meta-analysis identifies six novel loci associated with habitual coffee consumption. Mol Psychiatry 2014. [Epub ahead of print]

66. Shin SY, Fauman EB, Petersen AK, et al. An atlas of genetic influences on human blood metabolites. Nat Genet 2014; 46:543–550. This study provides the most comprehensive exploration of genetic loci influencing human metabolism thus far and will be an important resource for future metabolomic and genetic research. The investigators report genome-wide significant associations at 145 metabolic loci and their biochemical connectivity with more than 400 metabolites in human blood. Among these are the associations between AHR and CYP1A2 with plasma caffeine. 67. Hindorf L, MacArthur J, Morales J, et al. Catalogue of published genome-wide association studies. http://www.genome.gov/gwastudies/ [Accessed 1 January 2013]. 68. Levy D, Ehret GB, Rice K, et al. Genome-wide association study of blood pressure and hypertension. Nat Genet 2009; 41:677–687. 69. Newton-Cheh C, Johnson T, Gateva V, et al. Genome-wide association study identifies eight loci associated with blood pressure. Nat Genet 2009; 41:666–676. 70. Guessous I, Dobrinas M, Kutalik Z, et al. Caffeine intake and CYP1A2 variants associated with high caffeine intake protect non-smokers from hypertension. Hum Mol Genet 2012; 21:3283–3292. 71. Chen R, Mias GI, Li-Pook-Than J, et al. Personal omics profiling reveals && dynamic molecular and medical phenotypes. Cell 2012; 148:1293–1307. These authors present the first ‘integrative personal omics profile’ (iPOP) by combining several omic profiles from a single individual over a 14-month period. &&




29

Coffee and health.

Coffee and liver health.

[Coffee and health].

Coffee Consumption and Cardiovascular Health.

Coffee: biochemistry and potential impact on health.

Toward standardardized VCG systems.

Toward a definition of asthma for epidemiology.

Coffee and tea breaks for liver health.

Epidemiology and health policy.

Systems oncology: toward the clinical application of cancer systems biology.

Impact of caffeine and coffee on our health.

Epidemiology, population health, and health impact assessment.

Toward a science of learning systems: a research agenda for the high-functioning Learning Health System.

Use of syndromic surveillance at local health departments: movement toward more effective systems.

Epidemiology and evaluation: steps toward hypertension treatment in the 1990s.

Epidemiology of adult asthma in Asia: toward a better understanding.

Engineering biological systems toward a sustainable bioeconomy.

Thallium antagonism toward potassium dependent systems.

Epidemiology and public health aspects of malocclusion.

Impact of Organic and Conventional Systems of Coffee Farming on Soil Properties and Culturable Microbial Diversity.

[Investigation on coffee and coffee-substitutes. XVII. Behaviour of polysaccharide-complexes of robusta-coffee during roasting].

Toward mutuality and kinship in health care.

Chinese health care system and clinical epidemiology.

Social justice, epidemiology and health inequalities.