Mycotoxins in small grains and maize: Old problems, new challenges

Abstract

This paper reviews the challenges relating to chronic contamination of small grains and maize with deoxynivalenol and related compounds, fumonisin and the use of ensiled cereals in cool dairy areas. Uncertainties in the tolerable daily intakes for deoxynivalenol and fumonisin are discussed as they have the potential to affect current regulatory limits. In addition, climate change is resulting in more extreme rainfall and drought events which favour formation of deoxynivalenol and fumonisin, respectively. The development and refinement of models for predicting mycotoxin accumulation from weather data will become an essential tool for managing these events. Such models are also important for providing timely food aid to developing countries, which experience increased occurrence of acute toxicities, especially in children. Chronic contamination of silage in some areas with some Penicillium toxins deserves more attention in terms of their economic effects and possible implications for the purity of milk.

Keywords:

• Deoxynivalenol
• neurotoxicity
• fumonisin
• neural-tube birth defects
• silage
• roquefortine
• festuclavine
• PR toxin
• climate change

Introduction

It seems timely to review the progress made on mycotoxins research in cereals over the past 15 years and consider the challenges remaining and new problems on the horizon. As noted in an earlier review (Miller 1995), toxigenic fungi in crops have been historically divided into two distinct groups. The first includes those that invade and produce toxins before harvest and the second group, which form toxins after harvest, are known as storage fungi. However, the source of the fungi in both instances is the field (Miller 1995). Four types of toxigenic fungi can be identified: (1) plant pathogens, such as Fusarium graminearum; (2) fungi that produce mycotoxins on senescent or stressed plants, such as F. verticillioides and Aspergillus flavus on maize and A. carbonarious on grapes; (3) fungi that colonise the plant and predispose the commodity to mycotoxin contamination after harvest e.g. A. flavus in subtropical maize and (4) fungi that are found in the soil or decaying plant material that occur on the developing kernels in the field and later proliferate in storage if conditions permit, e.g. Penicillium verrucosum on cereals, P. roqueforti complex on ensiled materials and A. flavus on many commodities. The major toxins that contaminate maize and small grains (wheat, triticale, barley) pre-harvest are deoxynivalenol (replaced in some areas by nivalenol) and zearalenone, fumonisin and aflatoxin on maize. For two of these toxins, namely deoxynivalenol and fumonisin, there are unresolved issues that might affect their hazard assessment. Because they are common in grain, this represents a level of uncertainty that perhaps deserve more attention in this review.

There are other mycotoxins that can cause problems occasionally in small grains and maize, the most important being the Fusarium toxin, T-2, which is normally associated with a derivative, HT-2 toxin. Alimentary toxic aleukia (ATA) disease was described prior to 1900 and was associated with the ingestion of overwintered grain. During World War II, Russians were forced to eat grain left in the field. Thousands of people were affected resulting in the elimination of entire villages (Mirocha 1984; Beardall and Miller 1994). Strains of fungi isolated from the grains at the time were later shown to produce T-2 and related toxins (Joffre and Hagen 1977). These are mainly F. sporotrichioides toxins, a species that grows on wet grain left in the field and to some extent on the glumes of small grains (Miller 1994; Miller et al. 1998). In parts of Europe, F. langsethiae is also an important producer of T-2 toxin on small grains (Thrane et al. 2004; Torp and Nirenberg 2004). However, despite the vast literature on T-2, incidence data show that material concentrations of this toxin are uncommon in most growing areas. This is because most grain is harvested under warm, dry conditions. Modest levels of contamination in grain observed at harvest in parts of western Europe, primarily in cooler, wetter areas, are apparent exceptions to this generalization (Gareis et al. 2001). The Provisional Maximum Tolerable Daily Intake (PMTDI) of the Joint Expert Committee on Food Additives and Contaminants of the World Health Organization/Food and Agricultural Organization (JECFA) for T-2/HT-2 toxin of 0.6 µg kg⁻¹ bw has a larger safety factor than would normally be indicated. This is primarily due to a lack of experimental data (Larsen et al. 2004).

This review and comment will focus on three broad topics. First, a perspective will be offered on research on the Fusarium toxins, deoxynivalenol and fumonisin, in small grains (wheat, barley, oats) and maize. These crops comprise two thirds of cereal supply, which is currently in the order of 350 kg person⁻¹ year⁻¹ (Dyson 2001). The reason for this emphasis is that, for cereals contaminated by aflatoxin, including rice, the guidelines applying to international trade are clear. This is regardless of whether, for example. the difference between WHO and EU guidelines can be defended on a health basis (Wu 2004). In contrast, there are some uncertainties in the PMTDIs for deoxynivalenol and fumonisin relating to aspects of the mechanism and human health effects that might affect current trade limits. Second, factors that resulted in increased exposure to these toxins will be explored with suggestions about actions required to manage this change. Finally, the increased use of ensiled maize in north temperate dairy-producing areas (e.g. Quebec) will be examined in relation to uncertainties about toxins associated with this feed source.

Toxins associated with Fusarium head blight and Gibberella ear rot

Fusarium graminearum, F. culmorum and F. crookwellense are closely related species that produce deoxynivalenol (DON) or nivalenol and zearalenone, depending on the geographic origin of the isolate (Miller et al. 1991). These fungi cause Fusarium head blight in small grains and Gibberella ear rot in maize. These diseases are associated with temperate grain-growing regions. Which of the three species will dominate depends on temperature. These species also vary somewhat in pathogenicity; F. graminearum is regarded as the most virulent, although all three species can cause epidemics. Wheat, maize and barley are most affected by these pathogens (Miller 1994; Mesterhazy 2003; Snijders 2004) and by their toxins. These three crops comprise two thirds of the world's cereal production. Contamination of oats, rye and triticale has also been reported to contain Fusarium mycotoxins (Scott 1989; Gareis et al. 2003). In parts of Europe, F. poae is also an important producer of nivalenol on small grains (Thrane et al. 2004) and nivalenol is commonly reported in European oat samples (Gareis et al. 2003).

F. graminearum is associated with wheat and maize grown in warmer areas (e.g. southern Ontario) and F. culmorum, in cooler areas (e.g. northwestern Europe, but see below). The influence of temperature relates to conditions that allow a sustained period of warm weather (daytime temperatures >30°C) regardless of daily means. The most pathogenic species, F. graminearum and F. culmorum, are generally the most common species found. Since the 1890s, Fusarium head blight has been common in wheat from North America and China (Miller 1994; Wang and Miller 1988; Chen et al. 2000; Goswami and Kistler 2004). In the 1980s and 1990s, F. culmorum was the dominant species in cooler wheat-growing areas, such as Finland, France, Poland and The Netherlands (Snijders and Perkowski 1990; Miller 1994; Toth et al. 2004), but this trend has apparently changed in recent years as European summers have reached record warm temperatures such that F. graminearum largely dominates (Xu et al. 2006).

Morphologically identical isolates of F. graminarum (Gibberalla zeae) can produce either DON and zearalenone or nivalenol and zearalenone as the principal toxic metabolites that accumulate in grain. Within the former group, some strains produce DON by the 3-acetylated precursor and others make the 15 acetylated precursor. DON-producing strains with the 15-acetylated precursor dominate in North and South America. DON-producing strains with the 3-acetylated precursor are common in Europe and Asia (Miller et al. 1991). The Asian and New World strains are genetically distinct (O’Donnell et al. 2000). Nivalenol-producing strains of F. graminearum are common in parts of Europe, Japan and Australasia but very uncommon in the Americas. F. culmorum produces DON and zearalenone (Miller et al. 1991; Jennings et al. 2004; and references cited therein; Toth et al. 2004). The crown rot form of F. graminearum Group 1 is now called F. pseudograminearum (G. coronicola; Aoki and O’Donnell 1999) but also produces deoxynivalenol and nivalenol (Clear et al. 2006).

The use of susceptible wheat cultivars and maize hybrids is largely responsible for incidence of F. graminearum. Under epidemic conditions, agronomic practices have modest impact on disease (Miller 1994; Schaafsma et al. 2001; Hooker et al. 2005; Koch et al. 2006; Miller et al. 1998). As far as can be seen, only countries that enforce clear requirements, such as reductions in Fusarium head blight (including DON measurements) (Wilde et al. 2007), have been able to reduce toxin amounts in the harvested crop (Snijders 2004; see also Larsen et al. 2004).

“Red mold poisoning” was reported in rural Japan coincident with an increase in wheat production from 1800. Major epidemics were recorded in Japan for the 1890, 1901, 1914, 1932, 1946, 1958, 1963 and 1970 crops, with human and animal toxicoses reported throughout (Yozhizawa 1983; Udagawa 1988). Japanese researchers and officials were sensitive to the possibility of toxic chemicals from mold-damaged food. The study of mycotoxins began in 1881 when a Japanese researcher showed that ethanol extracts of rice damaged by Penicillium citreonigrum were fatal to dogs, rabbits and guinea pigs. This led to a commercial ban on the sale of rice damaged by that fungus (Pitt 1991). Well-documented reports of human toxicosis from the consumption of Fusarium head blight-damaged wheat and barley are available. These describe the typical symptoms that consistently include nausea, vomiting and diarrhoea (Yozhizawa 1983; Udagawa 1988). Russian officials reported the same symptoms from humans consuming bread baked from scabby grain in 1923 (Prentice and Dickensen 1968). DON was isolated by Japanese researchers from grain that had made humans ill (Morooka et al. 1972). This toxin was responsible for a large-scale incident of human toxicosis in the Kashmir Valley of India in 1988 (Bhat et al. 1989; Medical Research Council of India, unpublished report). The same symptoms were seen in Indians consuming bread made from highly contaminated wheat. Acute human toxicoses have been reported in China, Japan and Korea, among other countries (Yoshizawa 1983; Beardall and Miller 1994; Kuiper-Goodman 1994; Li et al. 1999).

Fusarium head blight-damaged grain began to be a problem in the Midwest US and Canada coincident with the dominance of Marquis wheat during WW I. By the 1920s, large cultivar-screening programs were underway in Minnesota (Schroeder and Christensen 1963). In 1928, there was a massive epidemic in the mid west, where US scientists showed that damaged barley resulted in emesis in swine (Mundkur 1934). By 1941, a water extract of barley contaminated by a fungus described as G. saubentii [an invalid name that included F. graminearum], induced emesis in swine by gavage (Hoyman 1941). Water extracts and then methanol extracts of maize and barley cultures of F. graminearum given intraperitoneally (i.p.) produced toxic signs in nursing mice and swine, and in swine by gavage, i.p. and intravenously (i.v.) by the mid 1960s (Vesonder and Hesseltine 1981). Using strains isolated from Fusarium head blight-affected cereals provided by W.L. Gordon (Agriculture Canada), Prentice et al. (1959) reported an emetic principle in organic solvent extracts from Fusarium cultures but were unable to determine the chemical structure (Prentice and Dickenson 1968). About the same time, while investigating Fusarium-damaged maize (described as F. culmorum and F. graminearum by Booth) resulting in cattle toxicosis. Australian researchers reported a toxic principle resulting in skin necrosis (Fisher et al. 1967). Finally, US researchers re-reported DON as “vomitoxin” from F. graminearum-contaminated maize in 1973 that had produced emesis in swine (Vesonder et al. 1973).

Humans appear to be quite sensitive to DON (Bhat et al. 1989; Kuiper-Goodman 1994), but the available information does not permit a dose–response to be reliably determined. The domestic animal most affected by DON is swine and, as noted, the use of the second trivial name for DON, vomitoxin, arose from the emetic effect in swine. The minimum oral dose required for emesis is in the order of 100 µg kg⁻¹ bw (Pestka et al. 1987). The emetic response in dogs appears to occur at a similar dose (Ueno 1983). However, DON seldom causes overt toxicity, including emesis, in swine because its presence in feed limits consumption. This anorexic effect typically results in decreased feed consumption and growth in swine at concentrations of more than 1 µg g⁻¹ in diets containing naturally contaminated grains. Trichothecenes in general, including DON, have a variety of immunological effects in laboratory animals at very low exposures. In experimental situations, this leads to increased susceptibility to bacterial, viral and fungal diseases with strong implications for human disease (Bondy and Pestka 2000; Pestka and Smolinski 2005).

There is, therefore, a long and clear historic association between DON and animal disease. After consumption of grains affected by Fusarium head blight, similar symptoms in human have been consistently reported in many populations since the turn of the 19th century. There is no uncertainty that consumption of contaminated wheat results in DON exposure in humans (Turner et al. 2007). For the last 25 years, health authorities have acted to reduce human consumption of this toxin. Considering the available toxicology data, Health Canada established a tentative tolerable daily intake in 1982 of 3 µg kg⁻¹ bw per day and half that for infants (Kuiper-Goodman 1985). Based on a much expanded database, the JECFA established a PMTI that was slightly lower in 2001 (Canady et al. 2001).

These factors (and others discussed in the following section) led to a series of recommendations on future research on DON at an ILSI–EU meeting held in Dublin (Larsen et al. 2004). In relation to either increasing or decreasing the PMTDI, two issues were raised that deserve repeating. There is wide agreement that the mechanism causing neurotoxicity (emesis and feed refusal) needed to be determined.

In the mid-1990s, a great deal of work was done to try and resolve this question for DON in swine. Dosing by a continuous-exposure osmotic pump, implanted intraperitoneally, resolved that the effects could not be due to taste or learned responses (Prelusky 1997). A single dose of 0.25 mg kg⁻¹ bw (i.v.) changed neurotransmitter concentrations in the hypothalamus, frontal cortex and cerebellum up to 8 days post-dosing. Norepinephrine increased in all three tissues, whereas dopamine was decreased. In contrast, serotonin increased and then decreased in the hypothalamus, it was decreased in the frontal cortex and no change was observed in the cerebellum (Prelusky et al. 1992). A lower dose (10 µg kg⁻¹ bw i.v.) resulted in changes in cerebral spinal fluid neurotransmitters (Prelusky 1993). Serotonin-receptor antagonists prevented DON-induced vomiting, while 5HT2-receptor antagonists were moderately effective in high doses. Other anticholinergic actives were also effective but by acting directly at the emetic centre preventing emesis regardless of the cause (Prelusky et al. 1992). This suggested that, although there is no doubt that the emetic centre is responsible for vomiting, the effect seems indirect. In plants, it has long been known that trichothecenes disrupt membranes, causing physical damage to membranes resulting in cell lysis. Red blood cells are a compartment for trichothecene metabolism and these cells will lyse in the presence of excess circulating toxin. The amount of toxin required to lyse red blood cells varies according to animal species. The reason is that the proportions of sphingomyelin, phosphatidylcholine, PPT–ethanolamine and PPT–serine in the inner and outer membranes of erthythrocytes vary between animal species. The concentrations of the trichothecene T-2 toxin that cause membrane deformation in human blood cells are comparable to those that affect protein synthesis in HeLa cells (Cundliffe et al. 1974; Gyonhyossy-Issa et al. 1986; Khachatourians 1990). It is known that, in wheat and maize, some genotypes have greater resistance to the membrane-damaging effects of DON. In maize, this is due to differences in the binding affinity of DON, implying that there are unknown functional changes in the membranes of more resistant types (Snijders and Kreching 1992; Cossette and Miller 1995; Miller and Ewen 1997). It is reasonable to speculate that modest effects on membranes associated with the emesis centre might be responsible for the neurotoxicity as the receptor structure would be altered and, hence, binding affinity.

The ILSI–EU meeting suggested that the establishment of an acute reference dose (ARfD) for DON would be valuable. It was also agreed that the ethical problems in doing a human study would be profound, indicating that perhaps a non-human primate study would be desirable (Larsen et al. 2004), in my opinion, would be less important than determining the mechanism of neurotoxicity. There are human clinical data available from the use of another trichothecene, DAS (also known as anguidine), as a chemotherapy agent in many studies. These studies have demonstrated that nausea and vomiting occurred in ∼50% of the patients at doses of between 200 and 400 µg kg⁻¹ bw i.v. (Bukowski et al. 1982; DeSimone et al. 1997). Considering the relative acute toxicities of DAS to DON, this would translate into an emetic dose for DON in adults of >800 µg kg⁻¹ bw. This suggests that the minimum emetic dose for DON in swine is a reasonable approximation of the human equivalent. It was also suggested that studies of interactions between trichothecenes be performed (Larsen et al. 2004), which, in my opinion, would have little value. It is known that there are interactions between trichothecenes in model systems (Koshinshy and Khachatourians 1992) and in animals (Schiefer et al. 1986; Bhavanishankar et al. 1988), but their dimension is modest (<5×) compared to the safety factors in the PMTI for DON.

Fumonisins from Fusarium verticillioides and related species

F. verticillioides and F. proliferatum are the most common fungi associated with maize. For many years, these fungi have been known to occur systemically in leaves, stems, roots and kernels and can be recovered from virtually all maize kernels worldwide, including healthy kernels. It is important to note that the plating of surface-disinfected kernels is an insensitive method compared to, for example, grinding the sample followed by dilution to extinction on semi-selective media, which largely detects actively growing mycelium in proportion to biomass. Detection is roughly proportional to the number of living cells in kernels and the fungus in the tip cap of the kernel is often not seen. The “diseases” resulting from Fusarium ear rot/Fusarium kernel rot is associated with warm, dry years and insect damage, and is mainly caused by F. verticillioides (=G. fujikuroi) and F. proliferatum. In warmer corn-growing areas, F. verticillioides is one of the most important ear diseases (Miller 2001). F. proliferatum (which also produces moniliformin) becomes dominant under different environmental conditions than F. verticillioides (De La Campa et al. 2005).

Below 25–28°C, F. graminearum grows well, with growth virtually ceasing above that temperature, and in that range, assuming that there is sufficient rain, this fungus out-competes F. verticillioides. Many studies on fumonisin from natural occurrence and experimental infections have demonstrated the importance of drought rather than temperature stress. F. verticillioides grows well at temperatures above 28°C (Reid et al. 1999) and there is evidence that fumonisin can only accumulate in stressed or senescing kernel tissue (Reid et al. 1999; Miller 2001). This is consistent with considerable field data; for example, in a US study, fumonisin concentrations were inversely proportional to June rainfall (Shelby et al. 1994). In the cool corn-growing area of southern Ontario, accumulation was limited to drought-stressed fields. Comparing three counties with similar temperatures, the three with the highest average FB1 concentrations (1.4 µg g⁻¹) had half the rainfall of the counties with the lowest average FB1 (0.4 µg g⁻¹; Miller et al. 1995).

Since drought stress results in greater insect herbivory on maize, it is not possible to totally separate these variables from other complications (Miller 2001). However, there is a strong consistent relationship between insect damage and Fusarium ear rot. Within a year or two of the availability of fumonisin analytical standards, a field survey demonstrated that the incidence of the European corn borer increased Fusarium kernel rot and fumonisin concentrations (Lew et al. 1991). Maize genotypes containing the anti-insectan Bt protein have reduced amounts of fumonisin compared to non-Bt genotypes (Bakan et al. 2002; Hammond et al. 2004; De La Campa et al. 2005).

De La Campa et al. 2005) were able to integrate this information in a study of factors that affected fumonisin accumulation in maize. Insect damage and weather variables in four periods around silking explained most of the variation in fumonisin concentrations at harvest. The first critical period for fumonisin accumulation was 4–10 days before silking when temperatures of <15 and >34°C (permissive temperatures) reduced fumonisin. Within permissive temperatures, some rainfall increased fumonisin. In the week following the silking period, again within permissive temperatures, some rainfall increased fumonisin. Thereafter, fumonisin was increased by slight drought stress.

Since the discovery of fumonisin in 1988, a great deal has been learned about its effects. Consumption of maize contaminated with fumonisin has a number of toxic effects on domestic animals, including equine leucoencephalomalacia in horses (ELEM), and pulmonary edema and immunosuppression in swine. The toxin is carcinogenic in rodents. The mechanism for all these phenomena is directly or indirectly due to the effects of fumonisin on sphingolipid biosynthesis; this work has been reviewed extensively (IPCS 2000; Bolger et al. 2001; SCF 2003). JECFA established a PMTI using the renal toxicity of fumonisin as the endpoint. (Voss et al. 1995; NTP 2001).

Fusarium kernel rot was associated with animal disease in the US midwest in 1904 and there were large epidemics of ELEM in the US during the drought years of the 1930s. In 1971, corn contaminated by the fungus, now called F. verticilliodes, was shown to cause ELEM (IPCS 2000). A South African group studying elevated esophageal cancer in the Transkei and a French group working on ELEM independently described fumonisin as the cause of disease in 1988 (Marasas 2001) and then in 1989 (as macrofusin; Laurent et al. 1989).

There has also been the association of regular consumption of large amounts of maize-based foods, regularly infected with F. verticilliodes, with esophageal cancer in South Africa and northern Italy (IARC 1993, 2002; IPCS 2000). South Africa has been growing maize at least since the 17th century and it is now grown across Africa (Desjardins and McCarthy 2004; McCann 2005). In Latin America, food is prepared primarily from tortilla flours prepared by heating with base which reduces fumonsin concentrations. However, in Africa, fumonisin is not affected by traditional methods of cooking (De La Campa et al. 2004; Shephard et al. 2002; Fandohan et al. 2005), but sorting does effectively reduce fumonisin concentrations (Desjardins et al. 2000; Riley and Miller 2003).

The earliest reports of esophageal cancer in rural black populations (studies from 1955–1969; Rose 1973 and references cited therein) noted the extraordinarily high rates of this cancer in the Transkei. This was striking compared to other parts of the world and other parts of Africa (Day, 1975). Since no biomarkers are available, it has proven impossible, so far, to establish fumonisin as a causative factor in this pattern of esophageal cancer. In the last IARC review (IARC 2002), fumonisin remained in group 2B (possible human carcinogen) and no change in classification was made despite the fact that, since the 1992 review, fumonisin has been shown to be a rodent carcinogen. This would change should positive evidence be found (currently this is sphingolipid ratio change only, concurrent with demonstrated dietary exposure) and, if reliably demonstrated, the IARC classification would change from 2B to 2A (probable human carcinogen), which might require a re-evaluation of the PMTDI.

After the setting of the JECFA TDI, it was found that fumonisin causes neural tube birth defects (NTDs) in mouse somites (Sadler et al. 2002) and a rodent model in vivo (Gelineau-van Waes et al. 2005). These studies arose from a transient increase in NTDs from 10 to 27 per 10,000 live births in Mexican-Americans in Cameron County Texas (Hendricks 1999; Marasas et al. 2004). A follow-up study found that increased NTD risk was associated with fumonisins exposure (Missmer et al. 2006) and, for a number of reasons, animal models had failed to predict this possibility (IPCS 2000). The mechanism relates to material exposure to fumonisin prior to the formation of the placenta. Fumonisin affects folate transport, which results in lowered folate in the embryo (Sadler et al. 2002; Marasas et al. 2004). A study of tortilla production in Cameron County revealed that some preparation methods in local facilities left intact fumonisin in the final product (De La Campa et al. 2004) and NTDs are very high in fumonisin endemic areas (Marasas et al. 2004). At the time of writing, there is no published study of NTDs in a regulatory strain of rodent [The strain used in the Gelineau-van Waes (2005) study is specialized for NTD research]. Another factor that might result in the re-evaluation of the JECFA TDI would be the results of a well-designed study in a regulatory strain of fumonisin.

Exposure to maize and wheat borne toxins is increasing

The existence of a widely accepted JECFA PMTI for DON and fumonisin are major achievements; however, in the recent past, exposure of young children has been close to the PMTDI in the Netherlands (Pieters et al. 2004), Denmark (Rasmussen et al. 2007) and Canada (Kuiper-Goodman et al. 2008). The PMTDI is exceeded in other countries–dramatically so in Africa and in parts of Latin America (JECFA 2001) and there is no doubt that this would be more dramatic if exposures were calculated for wheat-consuming population in endemic areas lacking a diverse source of cereals, as opposed to the standard GEMS diet. The situation for fumonisin exposure is similar, except much worse in parts of Africa (Bolger et al. 2001; Shephard et al. 2005, 2007).

As noted above, large areas of arable land have come under wheat and maize production in China since 1961 (Dyson 2001; Tong et al. 2003). Most arable land is in areas prone to humidity in the summer or is water-limited. Since 1961, China has increased cereal production 5-fold from 100 to 400 kg person⁻¹, with maize and wheat increasing roughly in proportion. Maize has been grown in China since the 16^th century (Desjardins and McCarthy 2004) and currently has a much larger production than wheat. For food, wheat is nearly equal to rice, a well-established food crop since the 6th century (Myer 1978). The ratio of rice to wheat + maize production has changed from 1.2:1 to 0.8:1 (Tong et al. 2003). Fusarium head blight epidemics have been greatly increasing in frequency in recent years (Chen et al. 2001) and, as noted above, there is exposure to DON in China from both wheat and maize, although it is not well documented (Canady et al. 2001; Meky et al. 2003).

The situation in Africa is much different. During the period 1960–2003, cereal production increased 2.5-fold (half that of China) but, approximately over the same period, declined on a per capita basis from 150 to 125 kg person⁻¹ (Dyson 2001). As with China, there has been a modest change in the ratio of maize production to that of the other staple crops (sorghum, millet, rice). However, of these, maize is uniquely susceptible to fumonisin, DON and zearalenone and co-exposures with aflatoxin are certainly common (Doko et al. 1996; Ngoko et al. 2001) Africa has become extremely vulnerable to exposure from mycotoxins found in maize (Riley and Miller 2003; Azziz-Baumgartner et al. 2005). China is producing sufficient food for residents of rural areas to purchase food (Gale et al. 2005), which means that diets are much more diverse in China compared to Africa. In regions where weather conditions result in (more) severe mycotoxin problems, very high exposures are inevitable with the potential for acute toxicoses (Riley and Miller 2003; Azziz-Baumgartner et al. 2006). While this has long been known for aflatoxin in Africa, the PMTDI for fumonisin, as noted, is exceeded in Africa, with the upper 10th percentile of the population being approximately three times that of the PMTDI (JECFA 2001). In areas where the occurrence of fumonisin is chronic, this materially understates the situation. Shephard et al. (2007) estimated fumonisin exposure in some areas of rural South Africa at 2–19 times the PMTDI and exposure in rural Bukino Faso was found to be 12–60 times the PMTDI (Nikiema et al. 2004).

Against this broad background, in both the fully developed market economies and due to the limited diversity of the food supply in developing countries, increased climate variability will produce more frequent epidemics of Fusarium head blight and Gibberella ear rot. As described above, the former requires rain at anthesis or silk emergence and warm conditions while Fusarium kernel rot requires dry conditions but permissive temperatures.

At the end of 2006, the number of weather-related disasters in Canada has increased 10-fold since 1900 and 4-fold since 1960, most of which are related to heavy rainfall (http://www.ec.gc.ca/TKEI/graphs/w_disasters_095_e.xls). These disasters have been associated with the flooding of rivers, caused by prolonged rain during rapid snowmelt, of drainage pathways, primarily caused by short-duration, intensive rainfall from thunderstorms or the residue of hurricanes coming up from the US southeast. These phenomena have also been felt in Europe (Ekström et al. 2005; Lehner et al. 2006; Wilson 2007). In the principal Canadian maize-production area (Ontario), each of the last five summers has been hotter than the previous 30, on average, which creates one of the conditions for increased risk of fumonisin accumulation. The other condition is drought (Miller et al. 1995; Miller 2001), which is also predicted to occur more often in most of the corn regions over the coming decades (Lehner et al. 2006). An increased prevalence of extreme weather events is now anticipated worldwide over the next century (Zhang et al. 2007).

Riley and Miller (2003) argued for increased use of forecasting methods to predict mycotoxins on a countywide-scale. There is a long history of the use of models to predict crop diseases, including Fusarium head blight (De Wolf et al. 2003; Del Ponte et al. 2005; Carranza et al. 2007) however, there are comparatively few reports on models predicting the potential for mycotoxins in field crops–the most useful being developed Schaafsma and colleagues (Hooker et al. 2002; Schaafsma et al. 2006; Schaafsma and Hooker 2008). These models need to be developed against a large background dataset of DON and weather within a particular area as the relationship between disease symptoms and toxin accumulation is cultivar-specific (Miller et al. 1984; Paul et al. 2006). Such work has also been attempted for Gibberella ear rot (Mansfield et al. 2006).

Although there are some recent studies on models for predicting DON in wheat (Eiblmeier 2006; Forrer et al. 2006), only one predictive model for DON, DONcast, has been published and commercialized (Hooker et al. 2002). This model, which adapted to Uruguay (Schaafsma et al. 2006) and French conditions (Schaafsma and Hooker 2008), allows decision-makers to implement changes in agronomic (fungicide application) or harvesting practices, to be aware where emerging DON problems exist and to take the necessary management steps. These might involve diverting the harvest from the affected fields away from human food-use to alternative uses, including use in DON-tolerant domestic animals (cattle) after appropriate dilution. For DON and fumonisin in maize, Hooker and Schaafsma (2005) and, in greater detail for fumonisin in maize, De La Campa et al. (2006) have demonstrated that such models are feasible. While these models are meant for areas with on-line meteorological data and information on emergent insect populations, some modeling might be feasible from remote-sensing information.

Rainfall timing, water stress and permissive temperatures are the key factors for DON and fumonisin accumulation. Modeling of drought and vegetation indices are a component of the Famine Early Warning System, which assesses remotely sensed data, ground-based sources and other factors affecting local food availability (http://www.fews.net/). These data could form the basis of models that might be developed for DON, fumonisin and possibly aflatoxin, adding an important early warning capacity to managing contaminated crops.

Potentially, toxin-predictive modelling is an important research direction for both vulnerable populations and due to increased climate variability in countries with commercial agriculture.

Uncertainties associated with the increased use of short season maize hybrids and silage

In eastern Canada, the use of maize silage in dairy production has increased approximately 5-fold over the past 25 years. This is mainly attributed to the availability of short season maize hybrids suitable for both eastern North America and parts of western Europe. In addition, as long as there is adequate protein available, maize silage has a high starch content and is a useful high-energy feed for cows (e.g. Dawo et al. 2007). Silage production has remained fairly stable in most of Europe in recent decades but there has been a shift towards maize silage in some countries (e.g. Denmark, The Netherlands; Wilkinson and Toivonen 2003).

In recent years, there has been increased recognition that silage is quite frequently contaminated by toxins, mainly from Penicillium roqueforti (Seglar et al. 1997; Auerbach et al. 1998; Seglar 1999). This appears to be due to a combination of changes in feed production technology (e.g. increased reliance on maize fodder instead of transporting grain maize from the Midwest US or southern Ontario) and/or increased farm sizes (making it more likely that small changes in herd health will be noticed). At first, this was, and often still is, attributed to the presence of low levels of Fusarium toxins in maize. Since these do not affect cows or cattle, this is not the explanation. In contrast, a number of toxic phenomena in cows, associated silage contaminated by P. roqueforti group, have been observed. Severe toxicoses in cows, associated with the latter fungi growing on silage, were first reported from Japan and the US in the 1960s (Wei et al. 1973; Omomo et al. 1994). This issue remained unresolved and largely ceased to be a practical problem. More recently, P. roquefori sensu lato has been associated with reports of two syndromes in cows: serious toxicoses associated with P. roqueforti and a general ill-thrift associated with P. paneum (Sumarah et al. 2005). Both of these fungi produce roquefortine but the former produces PR toxins and the later festuclavine, a compound long associated with ill-thrift in cows (Nielsen et al. 2006; O’Brien et al. 2006).

Good silage practice eliminates these fungi (O’Brien et al. 2007); hence, the problem can be managed. However, a careful US study reported in 1997 of dairy herds in Florida, Vermont and Wisconsin suffering from ill-thrift, silage was more likely to be contaminated by PR toxin (Seglar et al. 1997). Similarly, a German study reported widespread contamination by roquefortine (Auerbach et al. 1998), which is proxy for contamination either by P. roqueforti or P. paneum. From a public-health perspective, it remains unresolved whether these compounds occur in milk. This is regarded as needing special care owing to its importance in the diets of infants and children. Tüller et al. (1998) demonstrated that sheep fed ∼1.25 mg kg⁻¹ bw day⁻¹ roquefortine (recalculated from the authors’ data) resulted in the compound being detected in liver (1.15 mg kg⁻¹), bile (0.12 mg kg⁻¹), kidney (0.15 mg kg⁻¹) and trace amounts in muscle and fat. This is a high exposure but emphasizes the point that nothing is apparently known about the fate and distribution of roquefortine, PR toxin or festulcavine in milk–a question that needs resolution.

Summary

All the issues raised in this review regarding toxins have plagued agriculture over the past century and perhaps longer (Matossian 1989). The challenges for the current generation of researchers relates to setting appropriate regulatory limits that improve public health, especially for children. In addition, greater preparedness is needed to manage changes in climate and agricultural technologies which may increase the occurrence of mycotoxins. As some governments adopt increasingly restrictive regulatory limits, officials need to be prepared for supply restrictions leading to price increases and excursions over the TDIs. In developing countries, food sufficiency remains a problem, but risk of acute toxicoses due to certain mycotoxins appears to be increasing.

Acknowledgements

I thank colleagues at Agriculture Canada and Carleton University, Clive James, Maya Pineiro, Art Schaafsma, David Hooker, Kristian Nielsen, Jens Frisvad, as well as the Natural Sciences & Engineering Research Council of Canada and TUBITAK, for financial support. I thank John Gilbert and Hamide Z. Senyuva for inviting me to discuss this topic.