Bacterial composition of the human gut

The first part of this review is concerned primarily with the capability of the microflora in the distal gastro-intestinal tract (GIT) to transform PPT provided by the diet. A detailed discussion of the complex nature and composition of this microflora is beyond the scope of this review, but some basic statistics are instructive. The GIT is home to between some 10 and 100 trillion microbes⁽Reference Bäckhed, Ley and Sonnenburg⁹⁾, approximately 10-times as many cells as found in the human body⁽Reference Savage¹⁰⁾. The corresponding microbiome is said to contain more than 100-times as many genes as the human genome⁽Reference Bäckhed, Ley and Sonnenburg⁹⁾. Each portion of the GIT (mouth, oesophagus, stomach, small intestine and large intestine) has a distinctive flora, but the majority of the organisms are located in the distal GIT, where the concentration of some 500 bacterial species may reach 10¹¹–10¹² colony forming units/g, and account for some 35–50 % of the contents⁽Reference Salminen, Isolauri and Salminen¹¹⁾. The flora contains bacteria, archaea and eukarya, and the bacterial component has been studied the most extensively. Three genera, Bacteroides spp., Clostridium spp. and Eubacterium spp., each account for approximately 30 % of the organisms present⁽Reference Bäckhed, Ley and Sonnenburg⁹⁾, but the colon also has a significant content of Fusobacterium spp., Peptostreptococcus spp. and Bifidobacterium spp.⁽Reference Salminen, Isolauri and Salminen^11, Reference Tuohy, Gibson, Crozier, Clifford and Ashihara¹²⁾ These bacteria are accompanied by a single methanogenic archaeon, Methanobrevibacter smithii ⁽Reference Bäckhed, Ley and Sonnenburg^9, Reference Gill, Pop and Deboy¹³⁾. The eukarya include protozoa, but those found in the GIT seem never to have been studied with reference to PPT catabolism.

It is generally accepted that colonisation of the GIT of neonates starts immediately after birth. The type of delivery (passage through the birth canal v. caesarean section), type of diet (breast v. formula feeding) and environment affect the colonisation pattern⁽Reference Gronlund, Lehtonen and Eerola^14, Reference Harmsen, Wildeboer-Veloo and Raangs¹⁵⁾. The composition of the gut flora of a 2-year-old child is essentially the same as that of an adult⁽Reference Conway, Gibson and Macfarlane¹⁶⁾, but the variations in microflora composition associated with age, diet and health status are still not fully determined⁽Reference Eckburg, Bik and Bernstein¹⁷⁾. A study of faecal samples from 230 volunteers, aged 20–50 years, from four European countries has demonstrated marked country–age interactions. The proportions of bifidobacteria were 2- to 3-fold higher in the Italian study population than in any other study group, and this effect was independent of age. Higher proportions of enterobacteria were found in all elderly volunteers independent of the location⁽Reference Mueller, Saunier and Hanisch¹⁸⁾. There is good evidence that comparatively minor changes in the diet of rats can produce significant changes in the catabolites eventually produced from dietary PPT⁽Reference Phipps, Stewart and Wright¹⁹⁾, but the extent to which the human microflora can be modulated by diet is unknown.

Models used to study colonic catabolism

The distal GIT is conveniently viewed as an anaerobic fermentation vessel in which polysaccharides, proteins, lipids and xenobiotics such as PPT can be transformed. Several strategies have been used to investigate these transformations. In vitro studies frequently use either a defined bacterial strain⁽Reference Wang, Hur and Lee^20, Reference Raimondi, Roncaglia and De Lucia²¹⁾, ileostomy fluid⁽Reference Knaup, Kahle and Erk²²⁾ or a flora from freshly voided human⁽Reference Aura, Martin-Lopez and O'Leary^23–Reference Gonthier, Remesy and Scalbert²⁷⁾ or animal faeces⁽Reference Labib, Hummel and Richling^28–Reference Forester and Waterhouse³⁰⁾ cultured in a suitable medium to which a defined substrate has been added, and monitor (i) disappearance of substrate, (ii) formation of catabolites or (iii) changes in the microflora⁽Reference Lee, Jenner and Low^31, Reference Tzounis, Vulevic and Kuhnle³²⁾. Rarely all three parameters are studied at the same time. Typically controls would include (i) incubations containing test substrate but lacking the microbial culture to investigate purely chemical transformation, (ii) incubations containing test substrate and chloroform-inactivated culture to investigate substrate losses associated with cell binding and transformations associated with enzymes released after lysis of the micro-organisms and (iii) incubations containing basal medium and microbial culture in order to monitor catabolites arising from basal metabolism and to distinguish these from catabolites produced from the test substance. The culture may be static⁽Reference Gonthier, Remesy and Scalbert²⁷⁾ or dynamic⁽Reference Gao, Xu and Krul³³⁾. The various types of study used to examine microbial transformation of PPTs are shown in Table 1.

Table 1 Studies concerned with the transformation of phenols, polyphenols and tannins (PPT) by gut flora micro-organisms

It is impossible to know how well these in vitro models reflect the human GIT in vivo but studies using gnotobiotic rats that have been associated with a specific GIT micro-organism do not produce the same catabolites as that of the organism incubated in an artificial medium⁽Reference Peppercorn and Goldman³⁴⁾ and one must assume that the models are imperfect. In particular, micro-organisms that bind to the GIT surface might be very different to those found in the lumen, and it is these ‘unbound’ organisms that might predominate in the voided stool from which the culture is prepared⁽Reference Eckburg, Bik and Bernstein^17, Reference Hooper and Gordon³⁵⁾. The micro-organisms that survive the culturing process will depend on how rigorously oxygen was excluded and the nature of the medium employed. Media for anaerobic micro-organisms are typically rich in protein or polypeptides and these can be incompatible with the recovery of PPT that bind strongly to protein thus preventing proper analysis of the transformations. Studies in the authors’ laboratories suggest that the catabolism of PPT that are not strongly bound to protein is not significantly different from their catabolism in protein-free media, and protein-free media have been used to investigate the catabolism of procyanidins that would bind irreversibly to protein⁽Reference Stoupi, Williamson and Drynan³⁶⁾. One advantage of these models is the requirement for comparatively small amounts of a scarce and costly pure substrate.

A second approach has been the use of laboratory animals, commonly rodents, but occasionally pigs. The rodents might be either (i) normal laboratory animals with their typical flora⁽Reference He, Magnuson and Giusti³⁷⁾, (ii) animals treated with antibiotics (typically neomycin) to destroy the flora, (iii) gnotobiotic animals delivered under sterile conditions, and housed in sterile conditions and fed only on γ-irradiated diets⁽Reference Peppercorn and Goldman^34, Reference Scheline and Midtvedt³⁸⁾ or (iv) gnotobiotic animals associated with a human-type flora⁽Reference Tamura and Saitoh³⁹⁾. The use of such animals has allowed the effects of the rodent flora and human-type flora to be compared, and the effects of the flora and mammalian metabolism to be distinguished. Radio-labelled test substances can be used and full pharmacokinetic studies can be performed with these animals, and the contents (microbial and chemical) can be compared at different sites in the GIT⁽Reference He, Magnuson and Giusti³⁷⁾. However, gnotobiotic animal facilities, are very expensive to maintain. An extensive review of the in vitro and animal models is available⁽Reference Rumney and Rowland⁴⁰⁾. The early studies using these procedures, but focusing on flavonoids, have been reviewed⁽Reference de Eds, Florkin and Stotz^41–Reference Berry, Francis and Bollag⁴³⁾.

A third approach is to use volunteers, either free-living or following prescribed diets, and to collect faeces⁽Reference Knust, Erben and Spiegelhalder⁴⁴⁾, faecal water⁽Reference Jenner, Rafter and Halliwell⁴⁵⁾ and urine for analysis⁽Reference Vitaglione, Donnarumma and Napolitano⁴⁾. Normal volunteers can be compared with volunteers who have undergone an ileostomy but who are otherwise healthy enabling metabolism in the small intestine to be distinguished from that in the large intestine. These are discussed in more detail as follows.

Transformation of polyphenols by colonic microflora

The transformations of which the microflora are capable include O- and C-deglycosylation, the hydrolysis of esters and amides, and deglucuronidation of excreted mammalian metabolites. The aglycones are susceptible to aromatic dehydroxylation, demethoxylation and demethylation, and hydrogenation, α-oxidation and β-oxidation of the aliphatic elements generated following rupture of the aromatic ring⁽Reference Peppercorn and Goldman^34, Reference Scheline and Midtvedt^38, Reference de Eds, Florkin and Stotz^41–Reference Berry, Francis and Bollag^43, Reference Hattori, Shu and El Sedawy^46–Reference Kroon, Faulds and Ryden⁴⁹⁾. Most investigations have focused on phenolic catabolites, i.e. those that retain at least one phenolic hydroxyl, but it is clear that complete dehydroxylation can occur producing aromatic catabolites⁽Reference Phipps, Stewart and Wright^19, Reference Goodwin, Ruthven and Sandler⁵⁰⁾. Benzoic acid, however, can also be produced in human subjects by GIT microflora aromatisation of quinic acid⁽Reference Adamson, Bridges and Evans⁵¹⁾. It should also be noted that there is a significant, but often unquantified, yield of non-aromatic products, including oxaloacetate, CO₂, etc.⁽Reference Walle, Walle and Halushka⁵²⁾. A key step in the catabolism of flavonoids is the fission of the A-ring and loss of carbons C₅ to C₈ as oxaloacetate that is ultimately metabolised to CO₂⁽Reference Das and Griffiths⁵³⁾. Similarly, an oral-dosing volunteer study using [4-¹⁴C]-quercetin demonstrated that approximately 50 % of the radioactivity was eliminated as CO₂⁽Reference Walle, Walle and Halushka⁵²⁾, but the fate of the A-ring and B-ring was not investigated.

Dehydroxylation of ortho-dihydroxy and ortho-trihydroxy substrates can occur at either a meta or para position, but it is accepted that the hydroxyl at the meta position is removed less readily. Accordingly, ortho-dihydroxy and ortho-trihydroxy substrates yield predominantly meta-hydroxy and meta-dihydroxy catabolites, respectively⁽Reference Griffiths and Smith^54, Reference Smith and Griffiths⁵⁵⁾. Absence of a para-hydroxyl in the B-ring of flavonoids is considered to slow the degradation by the GIT microflora⁽Reference Labib, Hummel and Richling^28, Reference Griffiths and Smith⁵⁶⁾. Studies using ²H-labelled substrates and crude human gut flora have established that with 3-(4′-hydroxyphenyl)-propionic acid and 4-hydroxyphenyl-lactic acid, the aliphatic side chain may be moved about the aromatic ring, thus converting a 4-hydroxy substrate to the corresponding 3-hydroxy isomer⁽Reference Curtius, Mettler and Ettlinger^57, Reference Fuchs-Mettler, Curtius and Baerlocher⁵⁸⁾. The enzymes/micro-organisms responsible have not been characterised and their substrate specificity is unknown. Although it is possible that such an isomerisation might explain the apparent preference for removal of the para-hydroxyl from a 3,4-dihydroxy substrate, it would not explain the tendency for 3,4,5-trihydroxy substrates to be converted to 3,5-dihydroxy products.

Aromatic demethylation by rodent, pig and human GIT microflora has been observed in vitro, for example demethylation of 3′-O-methyl-(+)-catechin⁽Reference Gott and Griffiths⁵⁹⁾, 4′-O-methylgenistein (angolensin)⁽Reference Blaut, Schoefer and Braune⁶⁰⁾, 6-methoxyapigenin⁽Reference Labib, Hummel and Richling²⁸⁾ and phenolic acids formed from the B-ring of anthocyanins⁽Reference Keppler and Humpf²⁴⁾. In contrast, the gut microflora of the rat, mouse and marmoset are unable to degrade 3-O-methyl-(+)-catechin⁽Reference Hackett and Griffiths⁶¹⁾, suggesting that the they cannot attack the C-ring aliphatic O-methyl group. Similarly, there is no evidence for demethylation of this compound when given orally to volunteers⁽Reference Hackett, Griffiths and Wermeille⁶²⁾.

Comparatively few of the enzymes associated with these transformations have been characterised in microbes associated with the GIT, and the subtleties of substrate specificity, enzyme inhibition, etc., remain largely unknown. At least with regard to polysaccharide catabolism there appears to be few structures that cannot be degraded⁽Reference Bäckhed, Ley and Sonnenburg⁹⁾, and this situation might well apply also to PPT.

It has been demonstrated that some strains of the GIT micro-organism Eubacterium oxidoreducens can convert pyrogallol to dihydrophloroglucinol and 3-hydroxy-5-oxohexanoate. The phloroglucinol reductase and pyrogallol–phloroglucinol isomerase have been isolated and characterised⁽Reference Krumholz and Bryant^63–Reference Haddock and Ferry⁶⁶⁾. This isomerase can be viewed as creating a new hydroxyl at C₂ in the phloroglucinol (1,3,5-trihydroxybenzene) equivalent to creating a hydroxyl at either C₆ or C₈ in the flavonoid A-ring. Such a hydroxylation of the flavonoid A-ring has been demonstrated with Pseudomonas spp.⁽Reference Jeffrey, Jerina and Self^67, Reference Jeffrey, Knight and Evans⁶⁸⁾ but the general relevance of this observation to GIT transformation of PPT is uncertain because these facultative anaerobes are not typical of the GIT microflora.

Eubacterium ramulus has been reported at 4·4 × 10Reference Crozier, Jaganath and Clifford⁷ to 2·0 × 10⁹ colony forming units/g (n 20) of human faecal dry mass⁽Reference Simmering, Kleessen and Blaut⁶⁹⁾, and its growth is stimulated in vivo by dietary flavonoids⁽Reference Simmering, Pforte and Jacobasch⁷⁰⁾. Eubacterium spp. are well known for their ability to degrade flavonoids anaerobically, variously possessing deglycosylating activity, the ability to split the C-ring, chalcone isomerase and phloretin hydrolase activity⁽Reference Braune, Engst and Blaut^26, Reference Schneider and Blaut^71–Reference Herles, Braune and Blaut⁷⁶⁾. These chalcone isomerase⁽Reference Herles, Braune and Blaut⁷⁶⁾ and phloretin hydrolase⁽Reference Braune, Engst and Blaut^26, Reference Schoefer, Braune and Blaut⁷⁵⁾ enzymes convert flavanones to the isomeric dihydrochalcones and the dihydrochalcones to a C₆–C₃ acid and phloroglucinol, respectively. The phloroglucinol will be rapidly converted to aliphatic products as described earlier. It is not known whether the retro-chalcone associated with anthocyanins in media with pH>7⁽Reference Clifford⁷⁷⁾ are substrates for the phloretin hydrolase.

Eubacterium limosum demethylates isoflavones such as biochanin A, formononetin and glycitein⁽Reference Hur and Rafii⁷⁸⁾, and Eubacterium A-44 has a novel arylsulfotransferase but whether mammalian flavonoid sulphates can serve as donors is not known⁽Reference Kim, Konishi and Kobashi⁷⁹⁾. Eubacterium sp. SDG-2 and Peptostreptococcus sp. SDG-1 convert seco-isolariciresinol diglucoside to mammalian lignans⁽Reference Wang, Meselhy and Li⁸⁰⁾, and this Eubacterium degrades various (3R)- and (3S)-flavanols and their 3-O-gallates to diaryl-propan-2-ols, 3,5-dihydroxyphenylvalerolactone and 3-hydroxyphenyl-valerolactone. Interestingly, this organism can remove the para-hydroxyl from 3R flavanols such as (–)-epicatechin and (–)-catechin but not from the 3S flavanols such as (+)-catechin and (+)-epicatechin⁽Reference Wang, Meselhy and Li⁸¹⁾, and it has been demonstrated in vitro that the conversion of (+)-catechin to C₆–C₅ and C₆–C₃ catabolites required its initial conversion to (+)-epicatechin⁽Reference Tzounis, Vulevic and Kuhnle³²⁾.

Some Clostridium spp. can cleave the C-ring of flavonoids, converting flavonols to a phenylacetic acid and presumably phloroglucinol. Activity with flavanones was much weaker and could not be detected with flavanols⁽Reference Winter, Moore and Dowell^82, Reference Winter, Popoff and Grimont⁸³⁾. Some Fusobacterium spp. and Bacteroides spp. possess α-l-rhamnosidase⁽Reference Jang and Kim^84, Reference Bokkenheuser, Shackleton and Winter⁸⁵⁾, β-d-glucosidase⁽Reference Kim, Sohng and Kobashi^86, Reference Kim, Kim and Park⁸⁷⁾, β-d-galactosidase⁽Reference Bokkenheuser, Shackleton and Winter⁸⁵⁾ and β-d-glucuronidase⁽Reference Sung, Kang and Yoon⁸⁸⁾ activities.

The PPT catabolites most frequently reported are aromatic/phenolic acids having 0, 1, 2 or 3 aromatic hydroxyls, 0, 1 or 2 methoxyls, and a side chain of between one and five carbons, which might be unsaturated and might bear an aliphatic hydroxyl. For most flavonoids, the acids produced retain the intact B-ring, but anthocyanins also yield 2,4,6-trihydroxybenzoic acid and the corresponding benzaldehyde (phloroglucinaldehyde) both derived from the A-ring⁽Reference Keppler and Humpf^24, Reference Schneider, Schwiertz and Collins^72, Reference Justesen, Arrigoni and Larsen^89, Reference Kim, Jung and Sohng⁹⁰⁾. There is evidence that these phloroglucinol derivatives might form from anthocyanins purely by chemical degradation at the prevailing pH value in the GIT⁽Reference Kay, Kroon and Cassidy⁹¹⁾. Whether of chemical or microbial origin, these aromatic/phenolic acids are common to most PPT substrates so far investigated, although some variation of ring substitution and side chain length are substrate specific as summarised in Table 2. Phenolic acids can be decarboxylated to the corresponding hydrocarbons⁽Reference Labib, Hummel and Richling^28, Reference Peppercorn and Goldman^34, Reference Griffiths and Smith^56, Reference Justesen, Arrigoni and Larsen^89, Reference Peppercorn and Goldman^92, Reference Indahl and Scheline⁹³⁾, but the corresponding alcohols seem not to have been reported.

Table 2 The nature of the aromatic and phenolic acids produced by the gut microflora from pure phenols, polyphenols and tannins (PPT) substrates

* C₆–C₃ βOH (phenylhydracrylic acids) probably arise from mammalian metabolism of a microbial catabolite rather than directly by microbial catabolism.

† May occur also as lactones.

‡ In the case of isoflavones the aryl substituent is attached to C₂ of the C₆–C₃ dihydro side chain. For all other PPT, it is attached to C₃.

§ C₆–C₁ acids may form also by a purely chemical mechanism at the prevailing pH value.

Certain larger mass intermediates have been characterised only during the catabolism of flavanols and proanthocyanidins (e.g. diaryl-propan-2-ols)⁽Reference Groenewoud and Hundt^94, Reference Meselhy, Nakamura and Hattori⁹⁵⁾, and catabolites such as S-equol, diarylbutanes and urolithins⁽Reference Doyle and Griffiths^96–Reference Seeram, Aronson and Zhang⁹⁹⁾ are specific to the isoflavones, lignans and ellagitannins, respectively. Faecal water from free-living volunteers also contains some flavonoid aglycones indicating that aglycone degradation is not necessarily complete⁽Reference Jenner, Rafter and Halliwell⁴⁵⁾.

Anthocyanins

The absorption and catabolism of anthocyanins from blackcurrants in rats and human subjects have been studied⁽Reference Matsumoto, Inaba and Kishi¹²⁰⁾. Four anthocyanins, delphinidin-3-O-rutinoside, cyanidin-3-O-rutinoside, delphinidin-3-O-glucoside and cyanidin-3-O-glucoside, were absorbed and excreted as the glycosylated forms in both rats and human subjects. In human subjects, only 0·05 % of the ingested anthocyanin dose was found in urine. Although, unexpectedly, the 3-O-rutinosyl anthocyanins were directly absorbed, the amount found in the urine was very low⁽Reference Matsumoto, Inaba and Kishi¹²⁰⁾. Co-administration of anthocyanins with phytic acid enhances plasma concentration significantly by an unknown mechanism⁽Reference Matsumoto, Ito and Yonekura¹²¹⁾. In all studies on healthy subjects consuming anthocyanin-rich food, it is now well established that < 2 % of the total anthocyanin dose is absorbed intact, as estimated from the amount of anthocyanin and conjugates either in urine⁽Reference Felgines, Talavera and Gonthier¹²²⁾ or in blood⁽Reference Mullen, Edwards and Serafini¹²³⁾. Using ileostomist subjects, a high proportion of the anthocyanins from blueberry (up to 85 %, depending on the attached sugar moiety) traversed the small intestine unchanged and were found in the ileostomy bags; this amount would therefore reach the colon under physiological conditions and be subject to microbial degradation⁽Reference Kahle, Kraus and Scheppach¹⁰⁷⁾. In vitro studies using human microflora in an anaerobic chamber have indicated that protocatechuic acid is one of the most likely major degradation products from anthocyanins having a 3,4-dihydroxy B-ring⁽Reference Aura, Martin-Lopez and O'Leary²³⁾. In a human intervention study on blood orange juice, the C _max in blood for cyanidin-3-O-glucoside was 1·9 nm, whereas the value for protocatechuic acid was approximately 250-fold higher (492 nM). It was calculated that the main product, protocatechuic acid, accounted for up to 70 % of the anthocyanin intake. Only 1·2 % of the anthocyanin dose ultimately appeared in urine, whereas urinary protocatechuic acid represented 28 % of the total anthocyanin dose. The protocatechuic acid appeared to be unconjugated⁽Reference Vitaglione, Donnarumma and Napolitano⁴⁾. Protocatechuic acid was also found in rat plasma after feeding cyanidin-3-O-glucoside⁽Reference Tsuda, Horio and Osawa¹²⁴⁾ and after deglycosylation, cyanidin can breakdown spontaneously to give protocatechuic acid (very pronounced at physiological pH)⁽Reference Kay, Kroon and Cassidy⁹¹⁾. After ingestion of black raspberries by pigs, the profile of compounds in the gastrointestinal tract was analysed. In the entire gut, protocatechuic acid was the major phenolic, followed by 4-hydroxycinnamic, caffeic, ferulic and 3-hydroxybenzoic acids. These accounted for approximately 6 % of the ingested anthocyanin⁽Reference Wu, PittmanIii and Hager¹²⁵⁾ which is considerably less than the human study described previously.

After consumption of a high dose of strawberries by volunteers, the main phenolic acids in urine after 5 h were 4-hydroxybenzoic acid (10·4 mg/l), protocatechuic acid (0·7 mg/l), vanillic acid (1 mg/l) and genistic acid (1 mg/l)⁽Reference Russell, Scobbie and Labat¹²⁶⁾. Strawberry anthocyanins are primarily pelargonidin glycosides with lesser amounts of cyanidin glycosides⁽Reference Felgines, Talavera and Gonthier^122, Reference Hollands, Brett and Dainty¹²⁷⁾ that would be expected to yield 4-hydroxybenzoic acid and protocatechuic acid (Table 3), but these acids are normal preformed constituents of strawberries⁽Reference Stohr and Herrmann¹²⁸⁾.

Consumption of oats added to a purée of bilberries (glycosides of delphinidin, accompanied by lesser amounts of malvidin, peonidin and petunidin glycosides)⁽Reference Ichiyanagi, Kashiwada and Ikeshiro^129–Reference Ichiyanagi, Hatano and Matsugo¹³¹⁾ and lingonberries (cyanidin glycosides)⁽Reference Ek, Kartimo and Mattila¹³²⁾ by volunteers resulted in urinary excretion of 3-methoxy-4-hydroxyphenylacetic (homovanillic) and vanillic acid, low amounts of syringic acid (from malvidin glycosides) but no gallic acid (which would have been expected from delphinidin glycosides). Urinary excretion of these acids was maximal at 4–6 h and intact urinary anthocyanins comprised < 0·01 % of the dose⁽Reference Nurmi, Mursu and Heinonen⁵⁾.

Procyanidins

When rats were fed procyanidin dimer B3, the major urinary catabolites were 3-(3′-hydroxyphenyl)-propionic, 3-hydroxycinnamic, 4-hydroxybenzoic and vanillic acids (total 6·5 % of intake). Feeding of the procyanidin trimer C₃ or a mixture of procyanidin polymers gave in addition 3-hydroxyphenylvaleric acid (C₆–C₅)⁽Reference Gonthier, Donovan and Texier¹³³⁾. When rats were fed ¹⁴C-labelled procyanidin B2⁽Reference Stoupi, Williamson and Viton¹³⁴⁾, approximately 80 % of the ¹⁴C-radiolabel was absorbed and appeared in the urine. The ¹⁴C-radiolabel appeared rapidly in the blood at low levels, but did not reach a maximum until 6 h, indicative of a major contribution of catabolism by colonic microflora. These two studies indicate that the majority (>70 %) of the colonic catabolites of procyanidin dimers are unknown. In human subjects, after consumption of procyanidin and catechin-rich chocolate, the main urinary catabolites were 3-(3′-hydroxyphenyl)-propionic acid, ferulic acid, 3,4-dihydroxyphenylacetic acid, 3-hydroxyphenylacetic acid, vanillic acid and 3-hydroxybenzoic acid⁽Reference Rios, Gonthier and Remesy¹³⁵⁾.

Hydroxycinnamic acids

Hydroxycinnamic acids are most commonly found linked to a quinic acid moiety in fruits and foods. These compounds are called chlorogenic acids and the most common are caffeoyl-quinic acids, found at high levels in coffee. Fibre such as wheat bran is another source, where the hydroxycinnamic acids, particularly ferulic acid, are covalently linked to an arabinose sugar unit. Much of the understanding of the metabolism of hydroxycinnamates is derived from work on coffee. Chlorogenic acids are poorly hydrolysed by conditions in the stomach or small intestine. When coffee is given, there is a relatively small absorption of caffeic and ferulic acids in the small intestine and a low absorption of intact chlorogenic acids. The major absorption occurs in the colon, where dihydroferulic and dihydrocaffeic acids, products of microbial biotransformation, are the major products⁽Reference Stalmach, Mullen and Barron^136, Reference Renouf, Guy and Marmet¹³⁷⁾. The important involvement of the colon is also supported by studies on ileostomists, where approximately 67 % of the dose of chlorogenic acids reaches the colon⁽Reference Olthof, Hollman and Katan¹³⁸⁾. Dihydrocaffeic acid is one of the major phenolic acids in human faecal water⁽Reference Jenner, Rafter and Halliwell⁴⁵⁾ and has been detected in the plasma of coffee drinkers⁽Reference Goldstein, Stull and Markey¹³⁹⁾, in urine as the free form and mainly conjugated in human plasma after ingestion of artichoke leaf extracts⁽Reference Wittemer, Ploch and Windeck¹⁴⁰⁾, in human urine after chocolate intake⁽Reference Rios, Gonthier and Remesy¹³⁵⁾ and in rat urine after ingestion of polyphenol-rich wine extract⁽Reference Gonthier, Cheynier and Donovan¹⁴¹⁾. In summary, the major products from ingestion of chlorogenic acids are dihydrocinnamic acids in the plasma and urine. Studies on rats showed that the major 5-caffeoylquinic acid-derived phenolic acids, 3-hydroxybenzoic, 3-hydroxybenzoylglycine (3-hydroxyhippuric) and 3-hydroxycinnamic acids were from the caffeic acid moiety but that a significant portion of the hippuric acid (benzoylglycine) was produced from the quinic acid moiety⁽Reference Gonthier, Verny and Besson¹⁴²⁾.

Some older work on the tissue distribution of radiolabelled cinnamic acids is worth noting. Injection of ¹⁴C-cinnamic acid to rats gave distribution in organs as shown in Fig. 4, with the remaining radioactivity in urine (48 %), faeces (25 %) and exhaled CO₂ (0·3 %)⁽Reference Teuchy and van Sumere¹⁴³⁾. There was a substantial amount in skin and gonads. A later study showed that 82–90 % of orally administered trans-(3-¹⁴C)-cinnamic acid was absorbed in rats as indicated by the amount present in urine after 24 h⁽Reference Nutley, Farmer and Caldwell¹⁴⁴⁾.

Fig. 4 Distribution of radiolabel in rat tissues after injection of ¹⁴C-cinnamic acid to rats. The remaining radioactivity was in urine (48 %), faeces (25 %) and exhaled CO₂ (0·3 %)⁽Reference Teuchy and van Sumere¹⁴³⁾.

Once formed by microbial biotransformation, compounds must pass the colonic epithelium and enter the bloodstream. Mammalian enzymes will further transform the microbial products, mainly by conjugation, but also by β-oxidation. Conjugation can involve methylation (catalysed by the enzyme catechol-O-methyl transferase), sulfation (sulfotransferases), β-glucuronidation (UDP-glucuronosyl transferases), glycinylation (via a CoA thioester) and glutaminylation (by conjugation with glutamine). Many of the enzymes involved exist as multiple isoforms with different but overlapping specificities, typical of those acting on xenobiotics.

Absorption and metabolism of C₆–C₁

Both protocatechuic acid and phloroglucinaldehyde are transported by Caco-2 cells, and then metabolised to sulphate and glucuronide conjugates by Caco-2 cells⁽Reference Kay, Kroon and Cassidy⁹¹⁾ (Fig. 5). Studies with Caco-2 cells have shown that benzoic acid and the three mono-hydroxybezoic acid isomers are substrates for the monocarboxylate transporter⁽Reference Haughton, Clifford and Sharp¹⁴⁵⁾. 3-Methoxy-4-hydroxyphenylacetic (homovanillic) acid is a substrate for the rat organic anion transporter⁽Reference Mori, Takanaga and Ohtsuki¹⁴⁶⁾. Following absorption, catechol-O-methyl transferase methylates protocatechuic acid to vanillic acid⁽Reference Cao, Zhang and Ma¹⁴⁷⁾. Of the conjugating enzymes, UDP-glucuronosyl transferases 1A6 in human liver microsomes is active on protocatechuic aldehyde⁽Reference Liu, Liu and Zhang¹⁴⁸⁾. In addition, protocatechuic aldehyde is converted to approximately 70 % protocatechuic acid by guinea pig liver slices by the enzymes aldehyde oxidase, xanthine oxidase and aldehyde dehydrogenase⁽Reference Panoutsopoulos and Beedham¹⁴⁹⁾, but approximately 20 % was unidentified polar conjugates of protocatechuic acid. The conversion of caffeic acid to protocatechuic acid did not occur in the absence of a gut microflora in germ-free rats⁽Reference Peppercorn and Goldman³⁴⁾.

Fig. 5 Summary of absorption and metabolic pathways of C₆–C₁ and C₆–C₃ compounds in the gastrointestinal tract.

Biological effects of C₆–C₁

In a rat model of induced carcinogenesis, protocatechuic acid prevented 4-nitroquinoline-1-oxide-induced oral carcinogenesis, N-methyl-N-nitrosourea-induced carcinogenesis in the glandular stomach, azoxymethane-induced carcinogenesis in the colon and diethylnitrosamine-induced carcinogenesis in the liver⁽Reference Tanaka, Kawamori and Ohnishi^150, Reference Tanaka, Kojima and Kawamori¹⁵¹⁾. The doses of protocatechuic acid were expressed only as a percentage of the diet, but can be estimated as approximately 10–40 mg/kg body weight per d assuming that a 200 g rat typically consumes 20 g diet per d.

Several papers have reported in vitro activity of protocatechuic acid on cultured cells, and in some investigations, this microbial catabolite is more potent than cyanidin glycoside, an anthocyanin that gives rise to substantial levels of protocatechuic acid in vivo ⁽Reference Vitaglione, Donnarumma and Napolitano⁴⁾. For example, in human neuronal SH-SY5Y cells, protocatechuic acid (100 μm) is more potent than cyanidin-3-O–glucoside (100 μm) at inhibiting hydrogen peroxide-induced apoptotic events, including mitochondrial function and DNA fragmentation. However, neither was effective below 25 μm⁽Reference Tarozzi, Morroni and Hrelia¹⁵²⁾. Protocatechuic acid induces c-jun N-terminal kinase-dependent hepatocellular carcinoma cell death at concentrations>50 μm⁽Reference Yip, Chan and Pang¹⁵³⁾.

Protocatechuic acid at concentrations above 500 μm promoted time-dependent and concentration-dependent migration of human adipose-tissue derived stromal cells in vitro possibly via effects on matrix metalloproteinase-2⁽Reference Wang, Liu and Guan¹⁵⁴⁾, and promoted cell proliferation of cultured rat neural stem cells, with reduced apoptosis possibly via repression of caspase-3 activation⁽Reference Guan, Ge and Liu¹⁵⁵⁾. Significantly lower production of reactive oxygen species was seen following treatment with protocatechuic acid (6 μm for 7 d or 30 μm for 4 d) but suppression of caspase-3 required 30 μm for both durations. The relative effectiveness of lower doses longer term is encouraging and suggests that dietary exposure over a long period might perhaps promote better health over the subsequent years.

Several investigations have addressed lipid and cholesterol metabolism. Protocatechuic acid, the major component of the Chinese functional medicine, Danshen, has reported anti-angina efficacy⁽Reference Cao, Zhang and Ma¹⁴⁷⁾. Protocatechuic acid is methylated and then diffuses into mitochondria where it is conjugated with CoA. The result is that fatty acid oxidation is decreased, as shown by a lower acyl CoA/CoA ratio in heart, which in turn activates pyruvate dehydrogenase, a key and irreversible step in carbohydrate oxidation. This could switch heart energy substrate preference from fatty acid to glucose that would be beneficial for ischaemic heart conditions. There was no detectable accumulation of protocatechuic acid in tissues⁽Reference Cao, Zhang and Ma¹⁴⁷⁾. Protocatechuic acid at 0·5 g/kg diet given to rats (estimated as approximately 10 mg/kg body weight per d) produced changes in cholesterol and lipid metabolism. The serum total cholesterol, HDL-cholesterol and VLDL-cholesterol were all lower in the protocatechuic acid-treated group, suggested to be partly as a result of induction (estimated by mRNA) of hepatic LDL receptor, apoB, apoE, lecithin-cholesterol acyltransferase and hepatic TAG lipase⁽Reference Tamura, Fukushima and Shimada¹⁵⁶⁾.

Absorption and metabolism of C₆–C₂

Phenylacetate is an endogenous product of phenylalanine metabolism, is present at low levels in the mammalian circulation and is conjugated with glutatmine during metabolism. Its pharmacokinetic parameters have been well studied in a clinical setting in patients owing to its proposed effects on cancer. When phenylacetate was administered as an intravenous infusion as part of a phase I trial in children with refractory cancer, the half life was approximately 1 h, and phenylacetate was conjugated with glutatmine to form phenylacetylglutatmine⁽Reference Thompson, Balis and Serabe¹⁵⁷⁾. After oral consumption of phenylacetate, the concentration of phenylacetate peaked at 2 h and 40 % was excreted in the urine after 40 h⁽Reference Wajngot, Chandramouli and Schumann¹⁵⁸⁾. Intravenous radiolabelled ¹⁴C-phenylacetate was rapidly taken up by rat brain and converted into ¹⁴C-acetate⁽Reference Inoue, Hosoi and Momosaki¹⁵⁹⁾. Phenylacetate was well tolerated when infused into patients twice per d for weeks at a high dose (125 mg/kg body weight), and at these levels, phenylacetate induced its own clearance by 27 % during this period⁽Reference Thibault, Samid and Cooper¹⁶⁰⁾. Both phenylketonuric and normal subjects eliminated an oral dose (80 mg) of [¹⁴C]-phenylacetic acid in the urine almost entirely as phenacetylglutamine, showing that the glutamine conjugation mechanism is not defective in phenylketonuria and that it is able to cope with the large amounts of phenylacetic acid produced in this disorder⁽Reference James and Smith¹⁶¹⁾. The contents of the colon, examined as human faecal water, contained phenylacetic acid, 3-phenylpropionic acid, 3-hydroxyphenylacetic acid, 3,4-dihydroxyphenylacetic acid, 3-(4′-hydroxyphenyl)-propionic acid and 4-hydroxy-3-methoxycinnamic (ferulic) acid at high levels, up to 400 μm for phenylacetic acid and 3-phenylpropionic acid in one individual. The mean levels were 188, 197, 110, 64, 61 and 10 μm, respectively⁽Reference Karlsson, Huss and Jenner¹⁶²⁾.

Biological effects of C₆–C₂

Phenylacetic acid (administered as sodium phenylacetate) is one component of a drug (together with sodium benzoate) which is given for the treatment of acute hyperammonemia⁽Reference MacArthur, Altincatal and Tuchman¹⁶³⁾. As discussed previously, the pharmacokinetic behaviour is well characterised⁽Reference MacArthur, Altincatal and Tuchman¹⁶³⁾, and phenylacetylglutatmine and hippuric acids are products excreted in the urine after an intravenous dose. Phenylbutyrate, a prodrug that is metabolised into its active component, phenylacetate, in vivo, has been reported to extend lifespan in Drosophila ⁽Reference Kang, Benzer and Min¹⁶⁴⁾. Large doses of phenylbutyrate can have some toxic side effects⁽Reference Kasumov, Brunengraber and Comte¹⁶⁵⁾, and it has been shown that phenylacetate and phenylbutyrate modulate medulloblastoma⁽Reference Li, Parikh and Shu¹⁶⁶⁾. Some C₆–C₂ compounds may modulate cholesterol metabolism (see below). When given orally, a single high dose of phenylacetate did not affect plasma glucose concentration nor gluconeogenesis in type II diabetic patients⁽Reference Wajngot, Chandramouli and Schumann¹⁵⁸⁾. Phenylacetate affects cell growth and proliferation⁽Reference Li, Parikh and Shu^166–Reference Shibahara, Onishi and Franco¹⁶⁸⁾, inhibits prenylation and has been tested in both phase I and II clinical trials. 3,4-Dihydroxyphenylacetic acid exhibited antiproliferative activity on prostate and colon cancer cells⁽Reference Gao, Xu and Krul³³⁾. 3-Phenylpropionic acid, 3-hydroxyphenylacetic acid and 3-(4′-hydroxyphenyl)-propionic acid, found at high concentrations in faecal water, decreased the expression of cyclo-oxygenase-2 in HT29 cells at physiologically relevant concentrations⁽Reference Karlsson, Huss and Jenner¹⁶²⁾.

Absorption and metabolism of C₆–C₃

As summarised in Table 2, C₆–C₃ acids can be formed from many PPT substrates including proanthocyanidins. However, information on the absorption and metabolism of C₆–C₃ acids has come largely from studies of chlorogenic acids, conjugates of cinnamic acids with quinic acid. Such conjugates are widespread in foods and beverages, but coffee, artichoke, apples and blueberries are particularly rich⁽Reference Crozier, Jaganath and Clifford^7, Reference Clifford^169, Reference Clifford¹⁷⁰⁾ Approximately, 1 h after the consumption of coffee, ferulic, caffeic and isoferulic acids appear in the plasma at low levels indicating absorption in the small intestine⁽Reference Renouf, Guy and Marmet¹³⁷⁾. This early appearance was characterised in more detail, and the circulating compounds included caffeic acid-3-O-sulphate and ferulic acid-4-O-sulphate, and sulphates of 3- and 4-caffeoylquinic acid lactones⁽Reference Stalmach, Mullen and Barron¹³⁶⁾. These lactones are formed from the chlorogenic acids during coffee roasting⁽Reference Clifford¹⁶⁹⁾ and are rarely found in other foods or beverages. Several hours later (T _max >4 h), greater concentrations of dihydroferulic acid, dihydroferulic acid-4-O-sulphate and dihydrocaffeic acid-3-O-sulphate appear in blood, indicating colonic absorption after microbial transformation⁽Reference Renouf, Guy and Marmet¹³⁷⁾. Dihydroferulic acid, dihydroferulic acid-4-O-sulphate, dihydrocaffeic acid-3-O-sulphate, ferulic acid-4-O-sulphate, dihydroferulic acid-4-O-glucuronide and feruloyl-glycine were major urinary components. These catabolites collectively represented approximately 29 % of the chlorogenic acid dose, indicating good absorption of the microbial catabolites of chlorogenic acids⁽Reference Stalmach, Mullen and Barron¹³⁶⁾.

The colon pH value ranges from pH 5·7 to 6·7⁽Reference Fallingborg¹⁷¹⁾, and unless there are localised pockets of significantly lower pH value, these phenolic acids with pK values in the range pH 4–5⁽Reference Jovanovic, Steenken and Tosic¹⁷²⁾ will be extensively ionised. The absorption of C₆–C₃ catabolites has been examined using Caco-2 cells and rat everted intestinal sacs (Fig. 5). Dihydrocaffeic acid is absorbed mainly by the passive transcellular route and enters the circulation mainly as the free form, although a small proportion is glucuronidated⁽Reference Poquet, Clifford and Williamson¹⁰⁹⁾. The dihydrocaffeic acid is either rapidly sulphated, or methylated then sulphated, by the liver⁽Reference Poquet, Clifford and Williamson¹⁷³⁾, so that the major metabolites in the circulation are free dihydroferulic acid, dihydrocaffeic acid-3-O-sulphate or dihydroferulic acid-4-O-sulphate. Dihydroferulic acid in the gut is absorbed partially by the passive transcellular route, but with some involvement of a monocarboxylic transporter⁽Reference Poquet, Clifford and Williamson^109, Reference Konishi and Kobayashi¹⁷⁴⁾. 3-Hydroxycinnamic acid and its dihydroform, 3-(3′-hydroxyphenyl)-propionic acid, are both absorbed at least partially by the monocarboxylic transporter in Caco-2 cells⁽Reference Konishi and Kobayashi¹⁷⁵⁾.

Once absorbed in the colon, sulphate conjugation occurs primarily in the liver. In human subjects, sulphotransferase 1A1 is the most active isoform for sulfation of caffeic, isoferulic and dihydrocaffeic acids, whereas sulphotransferase 1E1 is most active towards the other methylated forms, i.e. ferulic and dihydroferulic acids⁽Reference Wong, Meinl and Glatt¹⁷⁶⁾. Glucuronidation occurs to a much lesser extent. The levels of dihydrocaffeic acid-3-O-sulphate were approximately 45-fold greater than the corresponding glucuronides, and the amount of dihydroferulic acid-4-O-sulphate in urine was much greater than the amount of dihydroferulic acid-4-O-glucuronide. The only cinnamic acid for which the glucuronides predominated was isoferulic acid⁽Reference Wong, Meinl and Glatt¹⁷⁶⁾.

Cinnamic acids, as distinct from the dihydrocinnamic acids, have rarely been reported as gut flora catabolites of flavonoids. Rats excrete 4-hydroxycinnamic acid in urine after the consumption of apigenin, phloridzin and naringenin⁽Reference Griffiths and Smith⁵⁶⁾.

Biological effects of C₆–C₃

Excess solar UV radiation produces damage and initiates immune response and inflammation in skin, sometimes leading to cancer. Dihydrocaffeic and caffeic acids, but not dihydroferulic and ferulic acids, reduced the cytotoxicity and pro-inflammatory cytokine production (IL-6 and -8) in HaCaT cells, a keratinocyte model, following UV radiation⁽Reference Poquet, Clifford and Williamson¹⁷⁷⁾, and dihydrocaffeic and 4-hydroxycinnamic acids inhibited UV-B damage in human conjunctival cells as assessed using the DNA damage marker, 7,8-dihydro-8-oxo-2′-deoxyguanosine⁽Reference Larrosa, Lodovici and Morbidelli¹⁷⁸⁾. 4-Hydroxycinnamic acid, but not protocatechuic acid, decreases basal oxidative DNA damage in rat colonic mucosa⁽Reference Guglielmi, Luceri and Giovannelli¹⁷⁹⁾. In CCD-18 colon fibroblast cells stimulated with IL-1β, dihydrocaffeic, dihydroferulic and dihydroxyphenylacetic acids attenuated PG E₂ demonstrating an anti-inflammatory effect in this system. In addition, dihydrocaffeic acid diminished the expression of the cytokines IL-1β, IL-8 and TNF-α, reduced malonyldialdehyde levels and reduced oxidative DNA damage (measured as 7,8-dihydro-8-oxo-2′-deoxyguanosine) in distal colon mucosal tissue in the dextran sodium sulphate-induced colitis model in rats⁽Reference Larrosa, Luceri and Vivoli¹⁸⁰⁾. Animal and studies in vitro also suggest that some C₆–C₂⁽Reference Bhat and Ramasarma¹⁸¹⁾ and especially C₆–C₃ ⁽Reference Lee, Choi and Jeon¹⁸²⁾ catabolites interfere with various enzymes in the mevalonate pathway including 3-hydroxy-3-methylglutaryl-CoA reductase, the rate limiting enzyme in hepatic cholesterol biosynthesis, albeit at concentrations unlikely to occur in plasma. However, these observations are of interest since commodities rich in PPT that would yield such catabolites, and the catabolites when given per os, have been shown to inhibit platelet aggregation⁽Reference Yasuda, Takasawa and Nakazawa¹⁸³⁾ or to have cholesterol-lowering activity in animal⁽Reference Lee, Choi and Jeon^182, Reference Bok, Lee and Park^184–Reference Yamakoshi, Kataoka and Koga¹⁸⁸⁾ and human studies⁽Reference Kurowska, Spence and Jordan¹⁸⁹⁾, and such gut flora catabolites may have contributed to the in vivo effect. Interference in the mevalonate pathway, particularly 3-hydroxy-3-methylglutaryl-CoA reductase inhibition, may have broader human significance⁽Reference Mo and Elson¹⁹⁰⁾.