Introduction

Liver cancer is the sixth most frequently diagnosed cancer worldwide with an estimated 1 053 619 new cases in 2030 and an age-standardized incidence rate of 8.6 per 100,000 person-years [1]. Liver cancer is one of the deadliest cancers, with an age-standardized annual mortality rate of 7.4 per 100,000 person-years [1]. Liver cancer incidence and mortality continue to rise, despite advances in prevention strategies and new technologies in both diagnosis and treatment [2]. Liver cirrhosis is the most important risk factor for the development of hepatocellular carcinoma (HCC, which comprise the majority of liver cancers), with hepatitis B and C and alcohol consumption, unhealthy dietary patterns, smoking and obesity being among other major risk factors for the development of liver cirrhosis [3].

Mycotoxins are fungal secondary metabolites that contaminate many of the most frequently consumed foods worldwide, such as grains, nuts, fruits and legumes [4–6]. One fungal species may produce many different mycotoxins, and the same mycotoxin may also be produced by several different fungal species [5, 6]. Recent reports revealed that 60% to 80% of the world’s food crops are contaminated by mycotoxins, resulting in a widespread human exposure to one or more mycotoxins, through consumption of cereal-based foods, nuts, fruits, coffee, spices, and oil-based seeds [7–9]. Chronic exposure to mycotoxins, such as aflatoxins (AFs), have been classified as potent human hepatocarcinogens [10–12].

Numerous mycotoxins have been identified but the mycotoxins most commonly related to human health include AFs, ochratoxin A, patulin, fumonisins, zearalenone and nivalenol/deoxynivalenol (see S1 Table) [13]. The International Agency for Research on Cancer (IARC) identifies the individual AFs B1 (AFB1), B2 (AFB2), G1 (AFG1), G2 (AFG2) and M1 (AFM1) as sufficiently evident human carcinogens, while other mycotoxins are designated as possible carcinogens (e.g., ochratoxin A (OTA), fumonisin B1 (FB1) fumonisin B2 (FB2). Recent research indicates that exposure to multiple mycotoxins have a potentially increased carcinogenic effect over single mycotoxins [14–17].

The primary target organ for many mycotoxins is the liver, evidenced especially by their association with HCC [18], where they are metabolized, though not always inactivated [19–21]. There is a strong, established link between AFs exposure and development of HCC [12, 17, 21]. The World Cancer Research Fund (WCRF) has further concluded strong evidence linking AF-contaminated foods with liver cancer risk [22]. The associated mechanism possibly involves metabolization of AFB1 in the liver to a highly reactive species, capable of forming mutagenic DNA-adducts. Subsequently, a synergistic effect promoting tumour growth in the liver has been reported for co-exposure to AFB1 and FB1 [23]. Less pronounced effects were reported in a meta-analysis on co-exposure to OTA, also resulting in increased hepatic lipid levels and increased relative weight of the organ [4].

Recent reports identify various food safety issues such as the presence of mycotoxins, likely to be affected by climate change, particularly in Europe. Fungal shifts towards northern latitudes due to warmer conditions give rise to a subsequent emerging pattern of mycotoxins in different regions in Europe [24–27]. Given the ubiquitous nature of exposures to multiple mycotoxins in many countries, it is imperative to comprehensively investigate their hepatocarcinogenic potency, by means of longitudinal cohort data. Therefore, this manuscript determines the association between mycotoxin exposures and incidence of HCC cancer using a large-scale multinational European cohort study.

Materials and methods

Results and discussion

Descriptive analyses indicated differences between the hepatobiliary cancer cases (N = 255) and the non-cases for age at recruitment, BMI, alcohol consumption, sex, smoking status, energy intake and the percentage classified as physically inactive and with diagnosed diabetes (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of the EPIC study participants.

https://doi.org/10.1371/journal.pone.0315561.t001

A large part of the EPIC population is being exposed to some of the main mycotoxins present in foods, although for most of the mycotoxins, only a small percentage of the population has exposures above the Tolerable Daily Intake (TDI) (Table 2). For mycotoxins like AFs and STC for which a tolerable intake of 0 μg/kg body weight (cf. as low as reasonably achievable (ALARA)-principle) is considered, nearly the whole population showed an intake above this 0-value leading to almost 100% of the population with intakes higher than the TDI in the MB scenario (Table 2B). For all other mycotoxins the percentage of the population that had intakes above the TDI (according to middle bound values) was rather low (ranging from 0% for FB up close to 3% for DAS) in the MB scenario (Table 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2.

Percentage of the population below and above the safety reference values for external mycotoxin exposures assessed based upon dietary questionnaire data for the full EPIC cohort (Table 2A: lower bound values; Table 2B: middle bound values).

https://doi.org/10.1371/journal.pone.0315561.t002

S3–S6 Tables presents a description of the external mycotoxin exposures assessed based upon dietary questionnaire data for the full EPIC cohort for lower bound values and middle bound value for μg/d and μg/kg body weight/d. For some of the mycotoxins such as citrinin, only few insignificant values had been measured/detected in the foods that were analysed by the member states (see lower bound scenario equal to zero for all percentiles). However, low values were assigned when the measurement was below the LOD or LOQ in the middle-bound scenario where half of the LOD or LOQ was considered for the MB scenario.

Associations between mycotoxin exposures and liver cancer risk were investigated and have shown an increased risk for DON with HCC, which remained significant in the fully adjusted model (HR_T3vsT1 (95%CI) = 1.90 (1.18–3.05), p-trend = 0.0079) (Table 3). It is noteworthy that an inverse association was found for MON and HCC (HR_T3vsT1 (95%CI) = 0.65 (0.43–0.99), p-trend = 0.042) and Citrinin and HCC (HR_T3vsT1 (95%CI) = 0.59 (0.39–0.89), p-trend = 0.013. No significant risks were found for the other mycotoxins in the fully adjusted model. Sensitivity analysis in which we additionally adjust for hepatitis infection in a subsample of the cohort for which hepatitis infection status was available confirmed these statistically significant associations for DON (HR_T3vsT1 (95%CI) = 2.25 (1.11–4.57)) while no significant associations were found for the other mycotoxins (although same positive trend was found, results were attenuated due to reduced power and limited number of cases with hepatitis and should therefore be interpreted with caution; S7 Table). Results for the model that does not include coffee consumption as potential confounding factor were similar to those of the fully adjusted model and are included as S8 Table. Sensitivity analysis showed that results obtained in Table 3 did not differ between male and female participants (although same positive trend was found, this was attenuated for women due to reduced power and limited number of cases for women and should therefore be interpreted with caution; S9 Table). Additional analysis was performed to investigate the associations between mycotoxin exposure and intrahepatic biliary tract cancer risk, extra-hepatic biliary tract cancer risk and gall bladder and biliary tract cancer risk. No significant risks were found in this analysis (S10–S12 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Hazard ratios (HR) and their 95% confidence intervals (CI) for the associations between mycotoxin (body weight) exposures and liver cancer risk using a fully adjusted model^*.

P-value of 0.01 was considered statistically significant (after Bonferroni correction).

https://doi.org/10.1371/journal.pone.0315561.t003

Additional analyses identifying the most important food groups contributing to these mycotoxins’ exposures revealed relevant contributors to be cereals and cereal products (39.1%), vegetables (20.3%), and the group of fruits, nuts and seeds (11.7%). Within the other food groups, we found very low or no concentrations of any mycotoxin, except for ZEN and AFM1 that are present in dairy products and patulin present in non-alcoholic beverages (including fruit juices) (S13 Table).

The results presented in this manuscript indicates the potential importance of investigating mycotoxin exposures in Europe. The external exposure analyses using EPIC dietary questionnaire data shows possible important exposures to particular mycotoxins in European countries. Our findings indicate a potentially increased HCC risk with higher exposures to the ubiquitously present Fusarium-toxin DON. Interestingly, no significant associations were found with other mycotoxins including Aspergillus-toxins, especially AFB1, a well-established liver carcinogen designated as a Group 1 human carcinogen by IARC (due to its genotoxic and mutational effects that are considered the main mechanism of action for AFB1) and suggested to induce persistent epigenomic effects (i.e. methyl DNA-mRNA-interactions) in primary human hepatocytes associated with HCC [44]. This lack of association with AFB1 in our study may be because exposures in Europe are very low, in contrast to DON which is ubiquitous in European diets. Indeed, several studies conducted in various European countries indicate that more than 50% of cereals are contaminated with DON [45].

DON targets the ribosome and induces activation of mitogen-activated protein kinases (MAPKs), the key transducers of the ribotoxic stress response [46, 47]. Inflammation, apoptosis as well as cell cycle arrest, endoplasmic reticulum stress, oxidative stress, and autophagy of the chaperone GRP78 are the main cellular effects of DON. Pathological sequelae resulting from chronic low dose exposure include anorexia, impaired weight gain, growth hormone dysregulation, and aberrant IgA production, whereas acute high dose exposure evokes gastroenteritis, emesis, and a shock-like syndrome [48]. It has been concluded that the capacity of DON to evoke ribotoxic stress contributes significantly to its acute and chronic toxic effects in vivo [46]. The potential of DON to induce transcription factors in Human Hepatoma cells has been investigated [50]. DON exposure induces decrease in cell viability in a concentration-dependent manner. Treatment of the Hep-G2 liver cell line with 1 μM DON resulted in at least 75% cell viability, whereas at high-DON concentrations (10 μM) only 15% of the cells showed metabolic activity after 48 h exposure [49]. More recently an inflammatory and apoptotic effect of DON was also observed in mouse precision-cut liver slices [50].

Several in vivo studies also demonstrated that mice orally exposed to low a concentration of DON for 28–30 days display low-grade inflammation as measured by cytokine concentrations in the plasma and the expression of inflammatory mRNA biomarkers in different organs including liver [46, 51]. Histological analysis of DON exposed animals also indicates various types of liver damages such as portal and periportal fibrosis [40] and lymphoid depletion [52]. A study aiming to characterize the chronic effects of DON by exposing cancer-prone transgenic p53 heterozygous (p53+/-) male mice and p53 homozygous (p53+/+) male control mice reported that DON was non-carcinogenic, even in the heterozygous p53 genetic background. The hepatic and renal gene expression analyses further confirmed that chronic exposure to DON was non-inflammatory [53].

Besides DON, the cytotoxic effect on HepG2 cells of acetylated DON, namely 3-ADON and 15-ADON, was evaluated by MTT assays. It has been revealed that the strongest toxic effect was observed at 48 h expressed by an IC50 of 3.6 ±1.2μM, which was the lowest IC50 observed for 3-ADON at any exposure time [54]. The study demonstrated cytotoxic effects on the HepG2 cell line in a dose-dependent manner for 3-ADON and 15-ADON individually [54].

Moreover, a synergy has been observed between DON and other trichotechenes including the acetylated forms of DON, for both cytotoxic and inflammatory effects [55, 56]. Considering that food is co-contaminated by several trichotechenes this synergy is likely to occur in vivo in humans.

Even though DON has not been classified as to its carcinogenicity for humans, it has been shown to exacerbate the DNA damage induced by various genotoxins as measured by the expression of the marker gH2AX. This effect is observed in vitro and in vivo in combination with model genotoxins with different modes of action, including captan, a pesticide with genotoxic potential, and colibactin, a bacterial genotoxin produced by the intestinal microbiota [47, 57, 58]. It would be of interest to determine to which extend individuals of the EPIC cohort exposed to DON and developing HCC were also exposed to low levels of other genotoxic compounds.

Interestingly, our results showed a possible decreased risk for HCC risk with higher exposure to moniliformin and citrinin (both mainly found in cereals). No other epidemiological studies have reported similar inverse associations between cancer risk and higher exposure to moniliformin, an observation that requires further research. As described in the tables, only insignificant values have been detected for citrinin, as such considered as a potentially less reliable exposure assessment. Our results may be prone to residual confounding bias e.g., dietary patterns consisting of various cereal and cereal products in combination with other positive dietary aspects, which may explain the inverse association observed for moniliformin.

Important strengths of this investigation were the access to one of the largest cohort databases currently available for investigating effects of dietary exposures on cancer risk. Strengths of the EPIC study include its large sample size, its prospective design, its long follow-up, and the inclusion of participants from different European countries with harmonised data collection, especially for diet, offering a broad perspective on dietary intakes in Europe. Also, the access to the EFSA mycotoxin occurrence data derived from the European member states and the important support by the EFSA experts were strengths of this study. In addition, the sensitivity analyses performed on the different levels of these analyses underscores the quality of the results obtained in this project.

However, some limitations should be acknowledged. Limitations of this project are the dietary intake assessment methods that were used in EPIC (mostly self-reported dietary questionnaires) which may be prone to reporting bias. Indeed, diet measurement instruments are built to capture the usual dietary intakes of an individual but are still subject to imprecision and inaccuracy. For example, spices are an important source of mycotoxins but are not captured in the EPIC questionnaires which may cause imprecise estimates of the amounts of actual mycotoxins consumed. Also, the mycotoxin levels in foods notably depend on environmental factors like climate and, storage conditions, leading to important variations in mycotoxin concentrations measured in similar foods. However, in epidemiological analysis these variations can unfortunately not be considered. Furthermore, the EFSA food occurrence data has its limitations as food samples have been analysed in different laboratories available all over Europe and disposing of different laboratory infrastructures and methods available and at different timepoints; therefore, not necessarily reflecting the foods consumed in the different EPIC countries and regions. As demonstrated in Table 2, for some of the mycotoxins such as citrinin, only insignificant values had been measured/detected in the foods that were analysed by the EFSA member states, so the values in the Middle-bound scenario were purely based upon the LOD and LOQ values for these mycotoxins. Given the small number of participants exposed (above PMTDI for DON) and the small number of HCC cases, results on DON association with hepatic cancer incidence should be taken cautiously. In addition, caution is needed regarding the extrapolation of these results to the entire European population or to other populations or ethnicities worldwide since this study included volunteers that may be expected to have more health-conscious behaviours (i.e., higher intakes of fruits, vegetables and wholegrains) compared to the general population. Further, in our models, we included all the participants with available dietary intake data, but with potential missing data on other covariates replaced with a missing class or imputation. Although this may have induced some residual confounding bias, a complete case model would lead to a selection towards more compliant participants in an already health-conscious population. In addition, this study used a single assessment of dietary intakes at baseline. Although diet may change over time, it is usually hypothesized that this estimation reflects general eating behaviour throughout middle-aged adult life [59]. Finally, this study was based on an observational cohort. Thus, even though our models included a large range of confounding factors, residual confounding cannot be entirely ruled out.

This indirect approach that aims to assess mycotoxin exposures via food consumption data and mycotoxin occurrences in foods, also referred to as ‘external exposure’ estimation, gives the community a first insight on the global mycotoxin issue at the population level. The ‘internal exposure’ estimation takes into account additional variables such as mycotoxicokinetics and -dynamics [60]. Therefore, additional information derived from biological markers for internal exposure are needed to characterize the physiological processes involved in any potential relationship between mycotoxin exposures and cancer risk. Investigating the internal exposure is essential to understand the human mycotoxicokinetics and to exclude the possible confounding issue of heterogeneous distribution of mycotoxins in foods. Hence, a more suitable and reliable mycotoxin exposure assessment can be achieved by the direct measurement of mycotoxin exposure biomarkers. Calculating intake levels from biomarker levels is still challenging, therefore both, external as well as internal exposure assessments are recommended when investigating and tackling health effects of multiple mycotoxin exposures [61]. In addition, although some of the discussed mechanisms described for DON can be related to carcinogenesis, additional mechanistic studies are also needed for improving our understanding regarding the potential carcinogenic effects of DON exposures.

These results demonstrating potential increases in HCC risk due to chronic mycotoxin exposures can help raise awareness of these high-risk contaminants in the general public as well as product development industries and governing regulatory bodies. Several practical primary and secondary prevention strategies exist for minimizing public mycotoxin exposure, which could be highly beneficial, if the requisite political will and financial investment are applied to what remains a largely ignored global health issue [62]. It should be underlined that prevention strategies should not aim at avoiding foods that are more prone for being contaminated with mycotoxins, but rather at eliminating or reducing the contamination level of the foods by improving storage and transportation conditions and detoxification after contamination.

Conclusions

These analyses showed greater HCC risk associated with long-term dietary exposures to DON. However, validation of these findings in other studies and via biomarkers is necessary. Further research investigating potential mechanisms underlying these putative associations is warranted. Even though DON has not been classified as to its carcinogenicity for humans, this mycotoxin presents a potential threat to human health.

Supporting information

Acknowledgments

We thank the National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands, for their contribution and ongoing support to the EPIC Study.

The coordination of EPIC is financially supported by International Agency for Research on Cancer (IARC) and by the Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London which has additional infrastructure support provided by the NIHR Imperial Biomedical Research Centre (BRC).

The national cohorts are supported by: Danish Cancer Society (Denmark); Ligue Contre le Cancer, Institut Gustave Roussy, Mutuelle Générale de l’Education Nationale, Institut National de la Santé et de la Recherche Médicale (INSERM) (France); German Cancer Aid, German Cancer Research Center (DKFZ), German Institute of Human Nutrition Potsdam-Rehbruecke (DIfE), Federal Ministry of Education and Research (BMBF) (Germany); Associazione Italiana per la Ricerca sul Cancro-AIRC-Italy, Compagnia di SanPaolo and National Research Council (Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), the Netherlands Organisation for Health Research and Development (ZonMW), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), Statistics Netherlands (The Netherlands); Health Research Fund (FIS)—Instituto de Salud Carlos III (ISCIII), Regional Governments of Andalucía, Asturias, Basque Country, Murcia and Navarra, and the Catalan Institute of Oncology—ICO (Spain); Swedish Cancer Society, Swedish Research Council and County Councils of Skåne and Västerbotten (Sweden); Cancer Research UK (14136 to EPIC-Norfolk; C8221/A29017 to EPIC-Oxford), Medical Research Council (1000143 to EPIC-Norfolk; MR/M012190/1 to EPIC-Oxford). (United Kingdom).