Background

Fungi provide valuable contributions to the diet (mushrooms, cheeses), to medications (penicillins, statins), to food preservation (for example, citric acid production) and to fermentation, and even support complex chemical synthesis of novel compounds; however, some species frequently contaminate cereal crops, and under certain environmental conditions can produce potent toxins (mycotoxins) that can contaminate many dietary staples^(1, Reference Miller²⁾. These mycotoxins contaminate up to 25 % of the world's cereal crops⁽¹⁾. Whilst there are many hundreds of mycotoxins identified, only a few have received significant research activity. Those of major concern to human health include those toxins produced from Aspergillus and Penicillium, the aflatoxins and ochratoxins; and those produced from Fusarium, the fumonisins, trichothecenes (for example, deoxynivalenol (DON), nivalenol, T2-toxin) and zearalenone. Based on their current worldwide frequency of contamination, their established toxicity and our ability to understand exposure based on the use of existing and newly developed biomonitoring tools (also known as exposure biomarkers), the present review will focus on aflatoxins, fumonisins and DON. It is probable that the global distribution of mycotoxin contamination will change alongside anticipated climatic adjustments over the next century⁽Reference Miraglia, Marvin and Kleter³⁾, and that this change in distribution of mycotoxins may ultimately change the focus of this research area. The present review will highlight the known human health effects, the suspected health effects and the hypothesised mechanisms of toxicity, with a particular focus on children and growth, and their possible carcinogenic effects. The complex immune toxicologies of the mycotoxins are not reviewed. Epidemiologists focused on understanding disparate disease aetiologies should remember that exposure to mycotoxins is almost unavoidable in cereal-consuming populations and, as a result, they have the capacity to negatively affect human health and modify human disease susceptibility.

Mycotoxins are typically resistant to processing and are stable in many cooking processes; thus complete avoidance without major dietary restriction is not feasible^(1, Reference Miller²⁾. The specific type of mycotoxin produced is dependent in part on climatic conditions; aflatoxins and fumonisins are more prevalent in tropical and semi-tropical conditions whilst DON occurs in temperate climates^(1, Reference Miller²⁾. The frequency and the amount of mycotoxin exposure are influenced by wealth^(1–Reference Wild and Turner⁶⁾. In developed countries, wealth affords access to greater food choices and the freedom to avoid contaminated foods at the individual level and prevent it from entering commerce at the regulatory level. The poorest regions of the world have neither the infrastructure nor the luxury to allow for such decisions. Additionally, the wealthier countries can restrict imports of contaminated crops, resulting in an additional burden on developing countries⁽Reference Miraglia, Marvin and Kleter^3, Reference Williams, Phillips and Jolly^5, Reference Wild and Turner⁶⁾. Despite the differences in contamination levels, exposure to mycotoxins is apparent on all continents, but the impact on health in most instances remains poorly examined in all regions and requires significant research effort.

Whilst the consequences of individual mycotoxins are more typically examined, in reality typical exposures will include mixtures of mycotoxins, on a background of a wide range of other susceptibility factors that affect human health including additional chemical and biological agents, genetic susceptibility, varied nutrition and immune status. Thus, there is an increasing recognition of the need to examine multiple exposures to understand the health effects from mycotoxin exposures.

Aspergillus and Fusarium mycotoxins

Of the about 200 species of Aspergillus, 10 % of these are harmful to man through a variety of mechanisms^(1, Reference Miller²⁾. Aspergillus flavus and A. parasiticus produce a potent family of liver toxins, known as aflatoxins, whilst A. ochraceus produces the nephrotoxic ochratoxin A.

Aflatoxins predominately occur in hot and humid regions of the world where an estimated 4·5 billion individuals are at risk of exposure⁽Reference Williams, Phillips and Jolly⁵⁾. Aflatoxins contaminate dietary staples including maize and groundnuts; thus, populations highly reliant on these staples, and with limited agricultural capacity and storage facilities, are most frequently exposed though diet⁽Reference Williams, Phillips and Jolly^5, Reference Wild and Turner⁶⁾. Biomarker-derived exposure data (see below) reveal that some of the world's poorest populations experience chronic exposure to aflatoxins throughout life, often at high levels.

Fusarium verticillioides (Sacc.) Nirenberg (formerly known as F. moniliforme Sheldon) and F. proliferatum (Matsushima) Nirenberg are important fungi that contaminate maize and produce fumonisins^(1, Reference Miller^2, Reference Marasas⁷⁾. Fumonisins predominantly contaminate maize in hot and humid climates, and co-contamination of maize with aflatoxins and fumonisins are reported⁽Reference Sun, Wang and Hu^8–Reference Wang, Liang and Nguyen¹⁸⁾. Chronic high levels of exposure are predicted in parts of South Africa, Central America and Asia. Fusarium graminearum (Gibberella zeae) and F. culmorum infection of wheat and maize in more temperate regions (Europe, North and South America and parts of Asia) causes significant economic loss in the form of head blight and contamination by DON and other related trichothecene mycotoxins^(1, Reference Miller^2, Reference Rotter, Prelusky and Pestka¹⁹⁾. A survey of >45 000 food items from countries within the European Union revealed that 57 % of cereals tested were contaminated with DON, demonstrating the frequent contamination of this mycotoxin⁽Reference Brera and Miraglia²⁰⁾.

Biomarkers of exposure

The presence of an accurately quantified amount of a toxin, and/or its metabolite(s), alone, is insufficient for a bio-measure to be classified as a biomarker. In the present review, an exposure biomarker is a biological measure which is correlated with the quantity of the xenobiotic ingested, resulting in improved exposure classification over approaches that are more traditional. The development and validation of aflatoxin exposure biomarkers occurred over 20 years ago⁽Reference Wild and Turner^6, Reference Wild, Turner, Millar, Bartsch, Boffetta, Dragsted and Vainio^21, Reference Kensler, Roebuck and Wogan²²⁾; they have been extensively reviewed elsewhere⁽Reference Wild and Turner^6, Reference Wild, Turner, Millar, Bartsch, Boffetta, Dragsted and Vainio^21–23) and so will not be discussed here in detail. The development of biomarkers for fumonisins and DON is more novel and the present review will highlight this.

Aflatoxins

The aflatoxins are a family of highly substituted coumarins containing a fused dihydrofurofuran moiety⁽²³⁾. Aflatoxins B₁, B₂, G₁ and G₂ all occur naturally, whilst aflatoxin B₁ occurs most frequently and is the most toxic and carcinogenic (Fig. 1). In many developing countries, maize and groundnuts (peanuts) are the predominant contaminated food items, often at high levels. Aflatoxin metabolism (Fig. 2) gives rise to a variety of metabolites^(23, Reference Eaton, Ramsdell, Neal, Eaton and Groopman²⁴⁾ including aflatoxin M₁, a frequent metabolite in milk of exposed lactating animals, including human breast milk following maternal exposure to dietary aflatoxin B₁^(23–Reference Zarba, Wild and Hall²⁵⁾. The parent toxins, aflatoxins B₁, G₁, etc., also occur in breast milk^(23, Reference Polychronaki, West and Turner^26, Reference Polychronaki, Turner and Mykkänen²⁷⁾. The consequences of breast milk exposures to aflatoxins in the developing infant remain poorly examined.

Fig. 1 Structures of the four naturally occurring aflatoxins: (a) aflatoxin B₁; (b) aflatoxin B₂; (c) aflatoxin G₁; (d) aflatoxin G₂. The 8,9 position is where the reactive epoxide can be readily formed across the double bond. Me, methyl.

Fig. 2 Aflatoxin (AF) B₁ metabolism and biomarkers. Me, methyl; GST, glutathione S-transferase; SG, glutathione; MA, mercapturic acid; → , epoxide-related toxicity pathways; ····>, non-epoxide-related toxicity pathways; - - ->, excretion or blood routes. Adapted from Wild & Turner⁽Reference Wild and Turner⁶⁾. (A colour version of this figure can be found online at http://www.journals.cambridge.org/nrr)

Aflatoxin B₁ is metabolised by a number cytochrome P450s⁽Reference Guengerich, Ueng and Kim^28, Reference Gallagher, Kunze and Stapleton²⁹⁾ and generates two reactive epoxide species, an exo-8,9-epoxide and endo-8,9-epoxide, in addition to several hydroxy metabolites, for example, aflatoxins M₁, Q₁ and P₁⁽Reference Wild and Turner^6, Reference Kensler, Roebuck and Wogan^{22, 23)} (Fig. 2). The two epoxides are highly reactive and can cause cellular and macromolecule damage by covalently binding to proteins and nucleic acids⁽Reference Wild and Turner^6, Reference Kensler, Roebuck and Wogan^{22, 23,} Reference Raney, Harris and Stone³⁰⁾. The exo-epoxide is toxic and mutagenic; the specific role of the endo-epoxide is less well examined, but it is predicted to cause a similar level of toxicity as the exo-epoxide. The exo-epoxide additionally forms a stable covalent adduct with the N7 moiety of guanine⁽Reference Wild and Turner^6, Reference Kensler, Roebuck and Wogan^{22, 23,} Reference Raney, Harris and Stone³⁰⁾. Depurination at this site releases 8,9-dihydro-8-(N7guanyl)-9-hydroxy aflatoxin B₁ (AFB₁–N7-Gua), which is observed, in addition to aflatoxin M₁, in the urine of aflatoxin-dosed animals and individuals exposed to aflatoxin through diet⁽Reference Polychronaki, Wild and Mykkänen^31–Reference Zhu, Zhang and Hu³⁶⁾. The urinary concentration of AFB₁–N7-Gua in two separate studies (r 0·80, P < 0·0001; and r 0·82, P < 0·0001) and urinary aflatoxin M₁ (r 0·82; P < 0·0001) strongly correlated with aflatoxin intake in chronically exposed individuals, and both serve as validated exposure biomarkers⁽Reference Groopman, Wild and Hasler^33–Reference Zhu, Zhang and Hu³⁶⁾.

Hydrolysis of both epoxides to aflatoxin B₁-8,9-dihydrodiol leads to a slow base-catalysed ring opening resulting in a resonating dialdehyde phenolate ion, capable of forming adducts with protein amino groups, particularly lysine⁽Reference Sabbioni, Skipper and Büchi^37, Reference Sabbioni, Ambs and Wogan³⁸⁾ (Fig. 2). One such protein adduct known as aflatoxin–albumin is present in the sera of aflatoxin-dosed animals and individuals naturally exposed to aflatoxin through diet⁽Reference Sabbioni, Skipper and Büchi^37–Reference Scussel, Haas, Gong, Njapau, Trujillo, van Egmond and Park⁶¹⁾. The concentration of aflatoxin–albumin in sera was strongly correlated (r 0·69; P < 0·0001) with aflatoxin intake and provides an additional validated exposure biomarker⁽Reference Gan, Skipper and Peng^39, Reference Wild, Hudson and Sabbioni⁴³⁾. The reliability of multiple measuring techniques (including immunoassay, HPLC and LC-MS) on a single set of samples by selected laboratories, and the long-term stability in cryopreserved samples provide additional confidence in its use as a valuable biomarker of exposure⁽Reference Scholl and Groopman^62–Reference Scholl, Turner and Sutcliffe⁶⁴⁾. In high-risk regions of the world, greater than 95 % of those individuals tested are positive for aflatoxin–albumin over a 3 log range, from approximately 3–5 pg/mg albumin to >1000 pg/mg⁽Reference Sabbioni, Skipper and Büchi^37–Reference Kensler, He and Otieno⁵⁹⁾, whilst more developed regions rarely have detectable levels of the biomarker⁽Reference Wild, Jiang and Allen^44, Reference Johnson, Qian and Xu⁶⁵⁾. Given that aflatoxins are genotoxic carcinogens, there is no safe threshold for exposure⁽Reference Williams, Phillips and Jolly^5, Reference Wild and Turner^6, Reference Wild, Turner, Millar, Bartsch, Boffetta, Dragsted and Vainio^21–23). Whilst unmetabolised aflatoxin B₁ occurs in the urine of exposed individuals, there is no significant correlation with intake⁽Reference Groopman, Wild and Hasler³³⁾, perhaps in part due to extensive metabolism of the parent toxin to various metabolites. Thus, urinary aflatoxin B₁ itself is not a useful indicator of the amount of aflatoxin exposure⁽Reference Wild and Turner^6, Reference Kensler, Roebuck and Wogan^22, Reference Groopman, Wild and Hasler^33–Reference Zhu, Zhang and Hu^36, Reference Gan, Skipper and Peng³⁹⁾.

Fumonisins

Fumonsins are secondary metabolites produced by F. verticillioides (Sacc.) Nirenberg and F. proliferatum (Matsushima) Nirenberg, primarily in maize grown in hot and humid climates^(1, Reference Miller^{2, 23)}. Whilst numerous fumonisin analogues are described, the fumonisin B (FB) series predominates, and within this series the occurrence frequency is FB₁>FB₂>FB₃; FB₁ (Fig. 3) is reported to represent on average about 70 % of the fumonisins in naturally contaminated maize⁽Reference Shephard, Marasas and Burger⁶⁶⁾. Fumonisins do not appear to undergo significant metabolism⁽Reference Merrill, Wang and Vales^67–Reference Dilkin, Direito and Simas⁷³⁾; thus biomarker development has not followed the metabolite profile approach used for aflatoxins.

Fig. 3 Structure of fumonisin B₁.

Fumonisins mimic naturally occurring sphingoid bases, and inhibit sphingolipid metabolism via competing with ceramide synthase⁽Reference Merrill, Wang and Vales^67, Reference Van der Westhuizen, Shephard and Van Schalkwyk^74, Reference Riley, An and Showker⁷⁵⁾. Fumonisin modulation of sphingoid base (sphinganine and sphingosine) concentrations in dosed animals has been reviewed⁽Reference Van der Westhuizen, Shephard and Van Schalkwyk^74–Reference Turner, Nikiema and Wild⁷⁷⁾, and this disruption is plausibly linked to their mechanism of toxicity⁽Reference Merrill, Wang and Vales^67, Reference Shephard, Van der Westhuizen and Sewram^76–Reference Merrill, Sullards and Wang⁷⁸⁾ including liver and kidney cancer, and neural tube defects (NTD). The capacity of fumonisins to alter levels of sphingoid bases, as observed in experimental animals, highlighted the possibility that their measurement in human bio-fluids may yield a useful exposure biomarker, though to date no such biomarker has been established. Due to the limited metabolism of fumonisin, measurement of the concentration of the parent compound in bio-fluids provides an alternative route for developing a biomarker. Animal studies indicate that the transfer of fumonisins to urine was about 0·4–2·0 % of that ingested⁽Reference Shephard, Thiel and Sydenham^68–Reference Dilkin, Direito and Simas⁷³⁾, though typically these percentages refer to total transfer over several days, and often at doses higher than would be observed in human subjects.

There have been relatively few studies in human subjects reporting urinary FB₁ measurements. In one study, a subset of Mexican (Morelos County) women were selected based on tortilla consumption (lowest (n 25), medium (n 25) and highest (n 25)) from a larger cohort (n 996), and their urine analysed for FB₁⁽Reference Gong, Torres-Sanchez and Lopez-Carrillo⁷⁹⁾. Overall, fifty-six of seventy-five women (74·6 %) had detectable urinary FB₁ (>20 pg/ml) (range of non-detectable to 9312 pg/ml). Urinary FB₁ was detected more frequently in the high (96 %) compared with the medium (80 %) and the low (45 %) consumption groups, and the geometric means were associated (P < 0·001) with consumption of tortillas (geometric means and 95th percentiles were 147 (88, 248), 63 (37, 108) and 35 (19, 65) pg/ml, respectively). No food samples were collected from this survey; though based on the urinary measures, Gong et al. ⁽Reference Gong, Torres-Sanchez and Lopez-Carrillo⁷⁹⁾ cautiously estimated FB₁ intake in this region to range from non-detectable to 23 μg/kg body weight (BW) per d, using a number of assumptions, including estimated average transfer from animal models. The provisional maximum tolerable daily intake for FB₁, FB₂ and FB₃ combined or individually is 2 μg/kg BW per d⁽⁸⁰⁾ and, consequently, fumonisin exposure is a significant health concern in this region.

Urinary FB₁ was also measured in the urine from Chinese adults from Huaian County, Jiangsu Province (n 43) and Fusui County, Guangxi Zhuang Autonomous Region (n 34). The frequency of urinary FB₁ detection was similar in both regions, with 84 and 83 % of samples positive for FB₁, respectively, though the mean concentrations were significantly higher in Huaian county (13 630 pg/mg creatinine: range non-detectable to 256 000 pg/mg; median 3910 pg/mg) as compared with Fusui county (720 pg/mg: range non-detectable to 3720 pg/mg; median 390 pg/mg). Moreover, the average estimated FB intakes were about three- to four-fold higher in the Huaian region⁽Reference Xu, Cai and Tang⁸¹⁾. These data would suggest that about 1–2 % of the ingested FB was transferred to urine. However, FFQ information and food measures of household items for FB₁ were used to estimate typical FB₁ intake. No significant correlation between urinary FB₁ and estimated intake was found, an observation likely to reflect that the FFQ used measured typical intakes over weeks rather than recent intake (in d) at the time of urine collection. Regardless, in both regions, the mean predicted intake of fumonisin was similar to that in Mexico⁽Reference Gong, Torres-Sanchez and Lopez-Carrillo⁷⁹⁾ and South Africa⁽Reference van der Westhuizen, Shephard and Burger⁸²⁾, and the provisional maximum tolerable daily intake will frequently be exceeded.

A more targeted validation approach to assess the relationship between urinary FB₁ and FB₁ ingestion was attempted in South Africa⁽Reference van der Westhuizen, Shephard and Burger⁸²⁾. Here, urinary FB₁ was measured on two consecutive days whilst FB₁ intake was assessed using plate-ready food (a maize porridge) on the days immediately before urine collection. Here, the resulting urinary FB₁ concentrations were roughly of the same order as those in both Mexico⁽Reference Gong, Torres-Sanchez and Lopez-Carrillo⁷⁹⁾ and Fusui County, China⁽Reference Xu, Cai and Tang⁸¹⁾, but significantly lower than those reported in Huaian County, China⁽Reference Xu, Cai and Tang⁸¹⁾. In the South African study the geometric mean daily estimated FB₁ consumption over 2 d was 4·84 (95th percentiles 2·87, 8·14) μg/kg BW per d, and the geometric mean urinary FB₁ concentration for subsequent mornings' first void was 225 (95 % CI 144, 350) pg/ml. Urinary FB₁ data were also presented after adjustment for creatinine (geometric mean 470 (95 % CI 295, 750) pg/mg). After data were natural log-transformed, a statistically significant, albeit moderate correlation (r ² 0·31) was observed between estimated FB₁ intake/kg BW per d and urinary FB₁ adjusted for creatinine. Based on rough estimates (by authors of the present review) back-transformed data suggest that the highest estimated intake of FB₁ was 90 μg/kg BW per d (forty-five times the recommended tolerable daily intake for FB₁) and the highest urinary FB concentration was approximately 4900 pg/mg; these were not measures for the same individual. The reported correlation used natural log-transformed data, and whilst statistically appropriate, comparisons of the correlations or r ² values in this survey and those used in part to validate other mycotoxin exposure markers are not straightforward. The authors of the South Africa study additionally conducted a successful hand-sorting and kernel-washing intervention on the maize to reduce FB contamination in it⁽Reference van der Westhuizen, Shephard and Burger⁸²⁾. Overall, a significant (P < 0·05) 62 % reduction in estimated FB₁ intake was reported based on FB₁ in plate-ready maize porridge before and following the intervention, whilst a borderline (P = 0·06) 41 % reduction in mean urinary FB₁ (pg FB₁/mg creatinine) was reported. Further, this paper presented an estimate of the transfer of FB₁ to urine to be approximately 0·075 % of that ingested, a value comparable with, but slightly lower than that reported in animals⁽Reference Shephard, Thiel and Sydenham^68–Reference Dilkin, Direito and Simas⁷³⁾ and similar to values in a limited human kinetics study conducted in the USA⁽Reference Riley⁸³⁾. Based on this quantitative data, urinary FB₁ was regarded as an exposure biomarker for fumonisin⁽Reference van der Westhuizen, Shephard and Burger⁸²⁾.

Urinary FB₁ concentrations were more moderate in a recent report from Sri Lanka, though biomarker validation was not discussed⁽Reference Desalegn, Nanayakkara and Harada⁸⁴⁾. Further confirmatory work in using urinary fumonisin as an exposure biomarker in humans is anticipated, including stability studies, with additional interest in measurements of modified sphingoid bases, rather than the unsuccessful alternative putative markers of fumonisin exposure (at least to date) that utilised sphinganine and sphingosine bio-measures⁽Reference Shephard, Van der Westhuizen and Sewram^76, Reference Zitomer and Riley⁸⁵⁾.

Deoxynivalenol

DON is a type B trichothecene mycotoxin (Fig. 4) which is predominantly associated with crop contamination with Fusarium graminearum (Gibberella zeae) and F. culmorum fungi. Both of these fungi are important plant pathogens that cause Fusarium head blight in wheat and Gibberella ear rot in maize⁽¹⁾. Also known as vomitoxin, DON is a common contaminant of cereals such as wheat, maize and barley⁽Reference Rotter, Prelusky and Pestka^19, Reference Pestka and Smolinski⁸⁶⁾. The potent gastrointestinal effects (and hence ‘vomitoxin’) in animals and suspected in humans have been reviewed in detail⁽Reference Rotter, Prelusky and Pestka^19, Reference Pestka and Smolinski⁸⁶⁾ and not discussed further.

Fig. 4 Generic structure of type B trichothecenes, including deoxynivalenol.

In resistant animal species, DON is readily detoxified by gut microbiota to a de-epoxy metabolite known as DOM-1 (de-epoxydeoxynivalerol)⁽Reference He, Young and Forsberg^87–Reference Young, Zhou and Yu⁹⁰⁾, though this pathway was not apparent in a small survey of human faecal metabolism⁽Reference Eriksen and Pettersson⁹¹⁾. DON can also be metabolised in the liver to a glucuronide⁽Reference Rotter, Prelusky and Pestka^19, Reference Pestka and Smolinski⁸⁶⁾. Based on DON metabolism and toxicokinetics in the rat and a pilot survey in human subjects, the combined measure of unmetabolised DON or ‘free’ DON (fD) and DON-glucuronide (DG), subsequently referred to here as urinary fD+DG, was suggested as a putative urinary exposure biomarker by Meky et al. ⁽Reference Meky, Turner and Ashcroft⁹²⁾. Improved DG enzymic digestion conditions to release fD, and the use of [¹³C₁₅]DON as an internal standard have provided significant improvements in precision and accuracy in the urinary assay⁽Reference Turner, Burley and Rothwell⁹³⁾. The process of validating this putative exposure biomarker was subsequently undertaken using this modified assay.

In a survey of UK adults, urinary DON was measured in a subset of 300 individuals selected based on cereal consumption from the 2nd/3rd, the 5th/6th and the 9th deciles (100 individuals from each group) from a larger cohort of 1724 individuals. Urinary fD+DG was detected in 98·7 % of individuals (geometric mean 8·9 (95 % CI 8·2, 9·7; range non-detectable–48·2) ng/mg)⁽Reference Turner, Rothwell and White⁹⁴⁾. A modest, but significant, positive association was observed between the urinary measure and cereal consumption (P < 0·001; r ² 0·23). Of the cereal items consumed, bread predominated in both frequency and mean level of consumption. In order to improve our understanding of the role of cereal in the diet and its relation to the biomarker concentration, a wheat restriction intervention was used to assess the effects of a 4 d avoidance of major sources of dietary DON on subsequent urinary fD+DG concentration. The intervention reduced wheat intake by >90 %, and the urinary concentration of fD+DG was also significantly (P < 0·001) reduced (geometric mean pre-intervention 7·2 (95 % CI 4·9, 10·5) ng/mg) by > 90 % (geometric mean post-intervention 0·6 (95 % CI 0·4, 0·9) ng/mg)⁽Reference Seeling, Danicke and Valenta⁸⁸⁾. This intervention restricted the types of food consumed rather than attempting to reduce the contamination level within the consumed food; thus readers should avoid a direct comparison of this outcome with that of the FB intervention above⁽Reference van der Westhuizen, Shephard and Burger⁸²⁾. The intervention on DON was a more simple study to improve our understanding of the biomarker, whilst the FB intervention was a more practical study aimed at providing a solution to the exposure. These data on DON provided further support for the approach of Meky et al. ⁽Reference Meky, Turner and Ashcroft⁹²⁾ using urinary fD+DG as an exposure biomarker.

Subsequently the concentration of urinary fD+DG was assessed over a longer period (6 d) during the consumption of the normal diet⁽Reference Turner, White and Burley⁹⁵⁾. The frequency of detection and range were similar to those of earlier studies (geometric mean 10·1 ng/mg creatinine; range non-detectable–70·7 ng/mg); again, the concentration of DON was significantly, but moderately, associated with cereal intake (R ² 0·23; P < 0·001). A 4 d dietary restriction was then initiated in which bread provided the only major source of cereals potentially contaminated with DON; consumption of alternative starch-based foods were suggested to maintain energy intake (for example, rice, potatoes). Participants provided a duplicate portion of every bread sample consumed for DON analysis, and they recorded the quantity consumed of each bread sample. The mean bread consumption was 155 (range 0–455) g/d. The mean contamination level of the bread was 74 (range 20–316) μg/kg, and the estimated average daily intake of DON was 10·6 (range 0–42·5) μg/d. On a daily basis, urinary fD+DG was correlated with DON intake (P < 0·001; r ² 0·56, 0·49, 0·54, 0·64, for each day respectively), and a more integrated assessment of the 4 d combined revealed a highly significant correlation (r ² 0·74; P < 0·001). After adjustment for age, sex and BMI, all correlations remained significant (P < 0·001; adjusted r ² 0·63, 0·68. 0·65, 0·63 and 0·83, respectively), providing the first data of a strong quantitative relationship between exposure and the urinary biomarker.

Further studies have been initiated to assess the stability of urinary fD+DG during the collection phase when samples may be at ambient temperature for prolonged periods (hours), and during the cryopreservation period (years). No losses were observed in samples stored at ambient temperature (18°C) for up to 24 h, or at 4°C for 48 h. No reduction in biomarker levels were found in cryopreserved samples kept at − 40°C for up to 3 years⁽Reference Turner, Hopton and Lecluse⁹⁶⁾. In studies to date DG appears to be the major metabolite of DON in human urine, whilst the de-epoxy metabolite has rarely been observed⁽Reference Meky, Turner and Ashcroft^92–Reference Turner, Ji and Shu¹⁰¹⁾. Urinary fD+DG represents best exposure in the previous 24–48 h⁽Reference Turner, White and Burley^95, Reference Turner, Taylor and White¹⁰²⁾, though average cereal intake over 7 d was significantly, but not as strongly, associated with the biomarker. Additionally, colleagues in Austria are investigating a putative direct measure of urinary DG⁽Reference Warth, Sulyok and Berthiller¹⁰³⁾. The combined measure of urinary fD+DG now serves as a validated exposure biomarker for DON intake⁽Reference Turner, Burley and Rothwell^93–Reference Turner, White and Burley^95, Reference Turner, Burley and Rothwell^99, Reference Turner¹⁰⁰⁾, at least in moderately exposed populations. To date, populations predicted to have the highest DON exposures remain unexamined with this assay⁽Reference Turner, Burley and Rothwell⁹³⁾. Nevertheless, for all of the European studies to date, approximately 1–5 % of the population of adults are predicted to exceed the recommended tolerable daily intake of 1 μg/kg BW per d⁽¹⁰⁴⁾. An earlier and perhaps less robust version of this assay was used to analyse urinary fD+DG in a pilot survey in a high-risk region of China⁽Reference Meky, Turner and Ashcroft⁹²⁾. Data from that survey revealed a mean level of urinary fD+DG about three to four times greater than the mean reported in surveys of Europeans⁽Reference Turner¹⁰⁰⁾. Urinary fD+DG has now been observed in studies from the UK, France, Sweden and China⁽Reference Turner¹⁰⁰⁾, and independently in Spain and Italy⁽Reference Rubert, Soriano and Mañes^105, Reference Lattanzio, Solfrizzo and De Girolamo¹⁰⁶⁾, whilst urinary DG was observed in Austrians⁽Reference Warth, Sulyok and Berthiller¹⁰³⁾.

All of the mycotoxin biomarker approaches discussed here are summarised in Table 1; ochratoxin A is additionally included for completeness, as the only other mycotoxin for which biomarker development has been reported⁽Reference Gilbert, Brereton and MacDonald¹⁰⁷⁾. Summaries of aflatoxin biomarker surveys are in numerous reviews; here only aflatoxin–albumin biomarkers are presented from studies of West African infants and children (Table 2), as these relate to latter sections of this review. A summary of surveys on Fusarium biomarkers is included (Tables 3 and 4).

Table 1 Summary of the main studies of mycotoxin biomarker correlations with intake*

* Only the fumonisin study used intake/kg body weight per d. The higher the correlation coefficient the stronger the relationship; thus, for example, urinary OTA would be regarded as the preferred exposure biomarker to assess the level of ochratoxin A exposure compared with serum ochratoxin A, and urinary aflatoxin–N7-guanine would be regarded as strong for aflatoxin exposure.

† Correlation used natural log-transformed data of both intake and the urinary measure; correlations for all other mycotoxin comparisons used unmodified data.

‡ Adjusted for age.

§ Deoxynivalenol and deoxynivalenol-glucuronide combined.

∥ Adjusted for age, body weight and sex.

Table 2 Summary of some aflatoxin–albumin survey data in West African children*

nd, Non-detectable.

* Adapted from Turner et al. ⁽Reference Turner, Van Der Westhuizen, Nogueira Da Costa, Knudsen and Merlo⁵³⁾.

† A total of 200 children measured at three time points within an 8-month period.

Table 3 Exposure biomarker surveys for the Fusarium mycotoxin fumonisin B₁*

(Mean values and 95 % confidence intervals)

* Geometric means are reported except for the study in China⁽Reference Diallo, Sylla and Sidibé¹¹⁶⁾.

† Overall, 56/75 were positive (> 20 pg/ml): mean 70 pg/ml; range non-detectable to 9312 pg/ml.

‡ Urinary data are the mean of 2 d data, overall range of data non-detectable to 3900 pg/ml (estimated by the authors of the present review).

Table 4 Exposure biomarker surveys for the Fusarium mycotoxin deoxynivalenol

nd, Non-detectable.

* Both interventions were voluntary restrictions of major sources of dietary wheat products. The two studies have completely independent data.

It is also prudent to emphasise a significant difference in the urinary assays for aflatoxin, fumonisin and DON. Whilst the analytical sensitivity may be suggestive of greater, lesser or a similar sensitivity in the exposure assessment, the overall sensitivity of the bio-measurement is also related to the transfer kinetics of the toxins. If, for example, we compare urinary biomarkers for aflatoxin M₁, FB₁ and fD+DG, the analytical limits of detection are 5, 20 and 500 pg/ml, respectively, whilst the mean estimated amounts transferred to urine are about 2, 0·075 and 72 %, respectively. Assuming average BW of about 65 kg and 1·5 litres of urine excreted, the limits of detection correspond to a mean intake of 0·006, 0·615 and 0·015 μg/kg BW per d, respectively. Thus, similar levels of these bio-measures in urine do not represent similar levels of exposure, and across a range of typical human exposures, urinary DON will be detected more frequently than FB₁, for example, despite a greater analytical sensitivity for FB₁ compared with DON.

Mycotoxins and human disease

In many parts of the developing world, chronic exposure to aflatoxins at high levels remains a significant health burden⁽Reference Williams, Phillips and Jolly^5–Reference Marasas^7, Reference Wild, Turner, Millar, Bartsch, Boffetta, Dragsted and Vainio^21–23, Reference Liu and Wu^108–Reference Gong, Turner, Hall, Leslie, Bandyopadhyay and Visconti¹¹⁰⁾. About 20 years ago, aflatoxin B₁ was classified as a potent liver carcinogen⁽²³⁾, a conclusion in part defined using the above biomarkers, and mentioned only briefly here; readers are directed to an excellent recent review of the topic⁽Reference Kensler, Roebuck and Wogan²²⁾ for further details. This section of the review will focus more on the recent observation between aflatoxin and growth faltering.

Fumonisin B, deoxynivalenol and growth faltering

To date, there are no strong epidemiological data on links between Fusarium mycotoxins and child growth, though it is hoped that the use of biomarkers discussed above may allow improved exposure assessment to inform such studies. In cultured cells, prolonged exposure to FB₁ has been demonstrated to reduce intestinal barrier integrity⁽Reference Bouhet and Oswald¹³⁷⁾, and to cause an increase in the ease of movement of macromolecules across the monolayer⁽Reference Caloni, Spotti and Pompa¹³⁸⁾. The mechanism(s) for such an effect remains unknown, though it is clearly established that fumonisins disrupt sphingolipid metabolism, key components in the integrity of the cell membranes⁽Reference Merrill, Wang and Vales^67, Reference Merrill, Sullards and Wang⁷⁸⁾. A recent survey from Tanzania suggests an association between estimated FB consumption at age 6 months and the subsequent height and weight of the child at the age of 12 months⁽Reference Kimanya, De Meulenaer and Roberfroid¹³⁹⁾. The study included 215 infants with maize consumption estimated for each child and the level of FB contamination in collected maize samples determined. Maize was consumed frequently (89 % consumers) and 69 % of the maize-consuming infants were exposed to detectable levels of fumonisin; the range of fumonisins in maize was 21–3201 μg/kg, and twenty-six infants were predicted to have exceeded the provisional maximum tolerable daily intake of 2 μg/kg BW⁽⁸⁰⁾. Infant height and weight were significantly negatively (P < 0·05) associated with estimated fumonisin intake. The authors identify a number of limitations in their study. First, there was no use of fumonisin exposure biomarkers. Second, exposure estimates were at 6 months but not 12 months. In that respect, it would also have been interesting to have data on growth velocity. It is notable that the co-occurrence of both aflatoxin and fumonisin contamination had been reported in an earlier survey of maize in this region⁽Reference Fandohan, Zoumenou and Hounhouigan¹⁴⁰⁾, and that aflatoxins and fumonisins also co-contaminate maize in Benin⁽Reference Kimanya, De Meulenaer and Tiisekwa¹⁰⁾, where strong associations between aflatoxin and growth faltering were reported⁽Reference Gong, Cardwell and Hounsa^48, Reference Gong, Hounsa and Egal⁴⁹⁾. This research arena eagerly awaits descriptive epidemiology on multiple mycotoxin biomarker measurements to understand the potential causes and mechanisms of dietary contaminants and growth faltering in high-risk regions.

In contrast to the aflatoxins and fumonisins, the trichothecenes occur most frequently in cooler, moist conditions^(1, Reference Miller²⁾. Their global distribution thus differs from aflatoxin and fumonisins, predominantly in Europe, North America and parts of South America, China, Japan and New Zealand. DON is one of the most frequently observed trichothecenes found globally and the mechanism of toxicity is associated with its ability to inhibit ribosomal protein synthesis by restricting elongation and termination of polypeptide chains⁽Reference Rotter, Prelusky and Pestka¹⁹⁾. This ribosomal binding is responsible for DON's effect on the immune system, termed the ‘ribotoxic stress response’, which causes increased apoptosis in leucocytes and increased cytokine production⁽Reference Pestka¹⁴¹⁾.

Most instances of DON-induced feed refusal occur in experimental animals exposed to dietary DON⁽Reference Arnold, McGuire and Nera^142–Reference Young, McGirr and Valli¹⁴⁴⁾. However, pigs given an intraperitoneal infusion of DON also show reduced food intake of a DON-free diet⁽Reference Prelusky¹⁴⁵⁾, suggesting that DON may affect food intake via a neuroendocrine mechanism. Indeed, pigs given an intravenous infusion of DON show increased levels of the serotonin metabolite 5-hydroxyindoleacetic acid in cerebral spinal fluid⁽Reference Prelusky¹⁴⁶⁾, although similar studies have found no changes in plasma serotonin levels⁽Reference Prelusky¹⁴⁷⁾. DON can also cross into brain tissue⁽Reference Pestka, Islam and Amuzie¹⁴⁸⁾, and cause regional changes in the neurochemistry of pigs, turkeys and rats⁽Reference Fitzpatrick, Boyd and Wilson^149–Reference Prelusky, Yeung and Thompson¹⁵¹⁾.

Recent data from Girardet et al. ⁽Reference Girardet, Bonnet and Jdir¹⁵²⁾ suggest that DON acts centrally to reduce food intake. Here, a 20 μg per mouse intracerebroventricular (ICV) injection of DON caused a significant reduction in food intake over a 12 h period. We point out that 2 μg DON per mouse (weight of 0·020–0·025 kg) is an approximate dose of 0·8–1·0 mg/kg BW and DON has been shown to cause acute anorexia when given via intraperitoneal injection at 1·0 mg/kg BW in female mice⁽Reference Flannery, Wu and Pestka¹⁵³⁾. Furthermore, Girardet et al. ⁽Reference Girardet, Bonnet and Jdir¹⁵²⁾ determined that oral DON administration increased central protein expression of the transcription factor c-fos and mRNA expression of the pro-inflammatory mediators IL-1β, TNF-α, IL-6 and cyclo-oxygenase-2 (COX-2), suggesting that DON does indeed cause physiological changes within the brain.

Given DON's capacity to elicit systemic pro-inflammatory cytokines and cause inflammation in vivo, it has been hypothesised that COX-2 may be important for DON's ability to cause anorexia and subsequent growth suppression in vivo ⁽Reference Zhou, Yan and Pestka¹⁵⁴⁾. COX-2, along with microsomal PGE synthase-1 (mPGES-1), are two enzymes responsible for converting arachidonic acid into PGE₂ during an inflammatory response⁽Reference Ristimaki¹⁵⁵⁾ and COX-2 has been shown to serve a function in anorexia caused by systemic inflammation⁽Reference Johnson, Vogt and Burney^156, Reference Swiergiel and Dunn¹⁵⁷⁾. Indeed, DON induces COX-2 gene expression in vitro (RAW 264·7 murine macrophages) and in vivo ⁽Reference Islam, Moon and Zhou^158, Reference Moon and Pestka¹⁵⁹⁾, though the recent use of mPGES-1 knockout mice suggests that a PGE₂-independent mechanism causes DON-induced anorexia⁽Reference Girardet, Bonnet and Jdir¹⁵²⁾.

Male and female mice fed 5 and 10 parts per million (ppm) DON over a 2-year period exhibit significant weight suppression⁽Reference Iverson, Armstrong and Nera¹⁶⁰⁾; yet, food intake was only significantly reduced in males fed 10 ppm DON, indicating that anorexia may not be the only mechanism by which DON reduces weight. Recently, DON exposure in mice was associated with decreased IGF acid-labile subunit⁽Reference Amuzie and Pestka¹⁶¹⁾, a binding protein of IGF-1 known to be critical for growth⁽Reference Boisclair, Wang and Shi^162, Reference Heath, Argente and Barrios¹⁶³⁾. DON-induced decreases in IGF acid-labile subunit may lead to alterations in the growth hormone system and subsequent growth suppression.

Taken together, growth suppression exhibited by experimental animals exposed to DON probably results from a combination of anorexia caused by alterations in neuroendocrine signalling and modifications of the growth hormone axis.

Fumonisins and neural tube defects

Fumonisins have been demonstrated to cause NTD in animals⁽Reference Marasas, Riley and Hendricks^164–Reference Gelineau-van Waes, Starr and Maddox¹⁶⁶⁾, and given the frequency and high levels of exposure in certain maize-consuming populations in South Africa, Central America and Asia, there are concerns of similar effects in humans. For example, women living in Cameron County, Texas–Mexico border had NTD incidence of 290 per 100 000, whilst Mexican Americans in general have NTD incidence of 90–160 per 100 000; this is a significantly higher frequency compared with white Caucasians (60 per 100 000)⁽Reference Missmer, Suarez and Felkner¹⁶⁷⁾. Sphingolipids are important structural components in cell membranes, and fumonisin disruption of sphingolipid biosynthesis interferes with folate receptors, and thus folate bioavailability⁽Reference Marasas, Riley and Hendricks^164, Reference Gelineau-van Waes, Voss and Stevens¹⁶⁸⁾. In a study conducted in pregnant mice, fumonisin given during early gestation caused NTD in 79 % of exposed fetuses. Alterations in sphingolipid profiles for both maternal and embryonic tissues occurred, as did both the folate concentration and expression of the folate receptor⁽Reference Marasas, Riley and Hendricks¹⁶⁴⁾. Of note in this study was that folate supplementation partially reduced the phenotype of NTD; although the dose level was relatively high, these observations raise some concern for humans chronically exposed to fumonisins.

An increased risk of NTD was significantly associated (OR 2·4; 95 % CI 1·1, 5·3) with moderate compared with low consumption of tortillas in the first trimester of pregnancy for women on the Texas–Mexico border⁽Reference Marasas, Riley and Hendricks¹⁶⁴⁾. Fumonisins are common contaminants of maize in this region, and maize consumption in the form of tortillas forms a major dietary staple. To better assess fumonisin exposure, maternal sphingolipid disruption (measured as the ratio of plasma sphinganine:sphingosine) and estimated FB consumption based on a limited food survey with FB measurements were determined. Both of these measures followed a similar pattern, with an increased risk of NTD between low and moderate groups, but not with the high exposure group. The authors suggest a threshold effect was possible, above which live births may not be occurring. It is important to note that the sphinganine:sphingosine measure was taken some time after birth (weeks), though the authors suggest that diets, and therefore exposure patterns, do not change much. More critical is the fact that sphinganine:sphingosine remains a putative but, to date, unvalidated exposure biomarker. NTD represent a significant health burden in maize-consuming populations where fumonisin exposure is frequent and at high levels. The animal and epidemiological data strongly support a role for this dietary toxin in increasing the risk of NTD through changes in folate receptor activity. Authors of the above studies highlight the desire for better exposure assessment tools to evaluate the temporal variation of fumonisins during critical exposure periods.

Fusarium mycotoxins and cancer

The incidence of squamous-cell carcinomas of the oesophagus (SCCO) is particularly high in parts of China, Iran and South Africa⁽²³⁾. China reports about 50 % of all SCCO cases, with incidence rates in Lixian county as high as 151 and 115 per 100 000 for males and females, respectively. At-risk populations are typically maize consumers, and Fusarium mycotoxins, for example, fumonisins and trichothecenes (DON, nivalenol), frequently contaminate maize in these regions. Ecological surveys examining the differences in the natural occurrence of Fusarium mycotoxins in high-risk regions of both China and South Africa are suggestive of a link with SCCO, and in vitro and animal data provide additional support⁽Reference Luo, Yoshizawa and Katayama^169–Reference Hsia, Tzian and Harris¹⁷⁴⁾, though no biomarker-driven epidemiology is available at this point. Fumonisins cause cancer in the liver and kidneys in rodents⁽Reference Howard, Eppley and Stack¹⁷⁵⁾. Whilst FB₁ has been classified by the International Agency for Research on Cancer (IARC)⁽²³⁾ as class 2B (possibly carcinogenic), data for DON are insufficient. Exposure biomarkers to Fusarium mycotoxins should support an improved understanding of their potential role in SCCO.