Prevalence of Aflatoxin- and Fumonisin-Producing Fungi Associated with Cereal Crops Grown in Zimbabwe and Their Associated Risks in a Climate Change Scenario

In most sub-Saharan African countries, staple cereal grains harbor many fungi and some produce mycotoxins that negatively impact health and trade. Maize and three small grain cereals (sorghum, pearl millet, and finger millet) produced by smallholder farmers in Zimbabwe during 2016 and 2017 were examined for fungal community structure, and total aflatoxin (AF) and fumonisin (FM) content. A total of 800 maize and 180 small grain samples were collected at harvest and during storage from four agroecological zones. Fusarium spp. dominated the fungi associated with maize. Across crops, Aspergillus flavus constituted the main Aspergillus spp. Small grain cereals were less susceptible to both AF and FM. AF (52%) and FM (89%) prevalence was higher in maize than in small grains (13–25% for AF and 0–32% for FM). Less than 2% of small grain samples exceeded the EU regulatory limit for AF (4 µg/kg), while <10% exceeded the EU regulatory limit for FM (1000 µg/kg). For maize, 28% and 54% of samples exceeded AF and FM Codex guidance limits, respectively. Higher AF contamination occurred in the drier and hotter areas while more FM occurred in the wetter year. AF exposure risk assessment revealed that small grain consumption posed low health risks (≤0.02 liver cancer cases/100,000 persons/year) while maize consumption potentially caused higher liver cancer rates of up to 9.2 cases/100,000 persons/year depending on the locality. Additionally, FM hazard quotients from maize consumption among children and adults were high in both years, but more so in a wet year than a dry year. Adoption of AF and FM management practices throughout the maize value chain coupled with policies supporting dietary diversification are needed to protect maize consumers in Zimbabwe from AF- and FM-associated health effects. The higher risk of health burden from diseases associated with elevated concentration of mycotoxins in preferred maize during climate change events can be relieved by increased consumption of small grains.


Introduction
In sub-Saharan Africa (SSA), maize (Zea mays) and small grains, particularly finger millet (Eleusine coracana), pearl millet (Pennisetum glaucum), and sorghum (Sorghum bicolor), are important food security crops for millions of smallholder farmers [1][2][3]. In Zimbabwe, maize is the most important cereal crop and is cultivated throughout the country. Often, small grains are popular in the semiarid and drought-prone areas, including those of

Samples
A total of 800 maize samples and 180 small grain samples (60 each of sorghum, finger millet, and pearl millet) were collected during the study period to determine both the prevalence of AF-and FM-producing fungi and levels of the two groups of toxins. In each cropping season, 200 maize samples were collected at harvest while the crop was still in the field or from drying sheds, and another 200 samples of shelled maize were collected during storage (4-5 months after harvest). At harvest, 30 maize cobs were randomly picked from the field or drying shed, shelled by hand and mixed to form a 5 kg composite sample. For stored maize, a 5 kg composite sample was sourced randomly from three to five bags kept within the homesteads of farmers. Small grain samples of 2 kg each were collected from farmers' stores, local markets, or grain marketing boards. All samples were geo-referenced and transported to the laboratory of the Department of Research and Specialist Services in Harare, Zimbabwe.

Sample Processing
Maize samples collected at harvest were dried in a hot air oven at 45 °C for 48 h. For each maize sample, a 2.5 kg subsample was obtained and shipped along with small grain samples to the IITA Pathology laboratory in Lusaka, Zambia, for further processing and analysis. Samples were stored at −20 °C until analysis. For microbiological analyses using whole grains, 200 g of grain was subsampled from each sample. The remaining sample

Samples
A total of 800 maize samples and 180 small grain samples (60 each of sorghum, finger millet, and pearl millet) were collected during the study period to determine both the prevalence of AF-and FM-producing fungi and levels of the two groups of toxins. In each cropping season, 200 maize samples were collected at harvest while the crop was still in the field or from drying sheds, and another 200 samples of shelled maize were collected during storage (4-5 months after harvest). At harvest, 30 maize cobs were randomly picked from the field or drying shed, shelled by hand and mixed to form a 5 kg composite sample. For stored maize, a 5 kg composite sample was sourced randomly from three to five bags kept within the homesteads of farmers. Small grain samples of 2 kg each were collected from farmers' stores, local markets, or grain marketing boards. All samples were geo-referenced and transported to the laboratory of the Department of Research and Specialist Services in Harare, Zimbabwe.

Sample Processing
Maize samples collected at harvest were dried in a hot air oven at 45 • C for 48 h. For each maize sample, a 2.5 kg subsample was obtained and shipped along with small grain samples to the IITA Pathology laboratory in Lusaka, Zambia, for further processing and analysis. Samples were stored at −20 • C until analysis. For microbiological analyses using whole grains, 200 g of grain was subsampled from each sample. The remaining sample (≈2.3 kg for maize and ≈1.8 kg for small grains) was milled using a coffee mill grinder (Bunn-o-Matic Corporation; Springfield, IL, USA), mixed thoroughly, and a 250 g subsample obtained for mycotoxin and microbiological analyses for Aspergillus characterization (see below). The grinder was thoroughly washed with 70% ethanol between samples to prevent cross contamination of both toxins and microorganisms.

Fungal Characterization
Initially, fungi from the grains were characterized on low strength potato dextrose agar (PDA, HiMedia Laboratories GmbH, Einhausen, Germany; 10 g/L) amended with 5 mL/L chloramphenicol (10 mg/L). After autoclaving (121 • C, 20 min) and cooling, 10 mg/L dichloran and 50 mg/L streptomycin sulphate were added before dispensing into Petri plates. For each 200 g whole grain subsample, 50 g of maize or 5 g of small grains were washed in 1% sodium hypochlorite-for surface-disinfection purposes-for 5 min and rinsed three times with sterile distilled water. Six grains of maize and 10 of the other cereals were placed onto the medium (20 plates per sample). Plates were incubated at 25 • C (7 d, dark), and then fungal colonies were enumerated. Based on macroscopic and microscopic characteristics, the fungi were classified as Fusarium [43], Aspergillus [44], Penicillium [45], or other genera.
Further classification of Aspergillus fungi to the species level was conducted on modified Rose Bengal agar (MRBA; 2% Bacto agar, 3 g sucrose, 3 g NaNO 3 , 0.75 g KH 2 PO 4 , 0.25 g K 2 HPO 4 , 0.5 g MgSO 4 7H 2 O, 0.5 g KCl, 1 mL micronutrients, 5 mL Rose Bengal, pH 6.0) amended with dichloran and streptomycin sulphate [44]. In total, 1-5 g of milled sample was suspended in 10 mL sterile distilled water and known volumes of the suspension plated. Plates were incubated at 31 • C (3 d, dark). Thereafter, Aspergillus colonies were enumerated, and 15 colonies transferred to 5-2 agar media (5% V8 Juice (Campbell Soup Company, Camden, NJ, USA); 2% agar, pH 6.0). Plates were incubated at 31 • C (5 d, dark). Aspergillus fungi were assigned to their corresponding species based on growth on specialized media [44], morphological characteristics, and spore ornamentation [46]. For microscopic assessment, wet mounts were prepared, and the slides examined using a light microscope at a magnification of 100×. Isolates with abundant small (<400 µm) sclerotia and scarce olive green conidiation were classified as S morphotype, while isolates with copious olive green conidiation and few, large sclerotia (>400 µm) were classified as the A. flavus L morphotype. S morphotype could be any of the following species: A. flavus, A. aflatoxiformans, A. minisclerotigenes, or an unknown species. Isolates with dark green, rough spores were classified as A. parasiticus. Isolates producing abundant tan conidia were classified as A. ochraceous and isolates producing abundant black conidia were classified as A. niger.

Mycotoxin Quantification
In each of the 180 small grain and 800 maize samples, AF and FM were extracted using procedures outlined in the Neogen protocol handbook (Reveal Q+ for aflatoxin and Reveal Q+ for fumonisin, Neogen Corporation, Lansing, MI, USA). Briefly, for each sample, 50 g of milled sample was weighed and transferred to a 500 mL Pyrex bottle. Then, 250 mL of 65% ethanol was added to the bottle, capped, and the content mixed by swirling. Bottles were placed on a MaxQ 2000 orbital shaker (Thermo Fisher Scientific Inc., Bartlesville, OK, USA), and shaken at 500 rpm for 3 min. Extracts were filtered through a Whatman No. 1 filter paper (Whatman International Ltd., Maidstone, UK). The toxins were quantified using the Neogen AccuScan Lateral Flow Device. The AccuScan total AF lower limit of detection (LOD) is 2 µg/kg while the upper detection limit (UDL) is 150 µg/kg. For FM, the LOD is 300 µg/kg and UDL is 6000 µg/kg. Samples exceeding UDLs were diluted with 65% ethanol and the analysis was repeated. AF-and FM-contaminated reference materials from Neogen were repeatedly analyzed after every 20 samples for quality control checks. Additionally, maize, millet, and sorghum samples were spiked with pure AF or FM and recovery rates of over 90% were obtained. To validate the AccuScan results, 7.5% of maize samples were analyzed with high-performance liquid chromatography (HPLC) in an accredited analytical laboratory, the Perishable Produce Export Certification Board (PPECB) in South Africa.

Risk Assessment: Dietary Exposure and Potential Health and Trade Impact
AF and FM levels quantified in all samples were used to assess dietary exposure. The average per capita consumption rates used in the analysis was 248 g per person per day (g/p/d) for maize, 8.71 g/p/d for both finger millet and pearl millet, and 18.04 g/p/d for sorghum [47]. The average body weight (bw) for adults and children were set at 60.7 and 25.0 kg, respectively [47]. Aflatoxin B1 (AFB1) data was unavailable for the simulation of liver cancer potency attributed to this toxin; therefore, a best available estimate approach, employing the common ratio and maximum permissible limit settings used worldwide of 2:1 for total aflatoxin (AF) to AFB1, was used. Often, AFB1 is estimated to be at least 50% of the quantity of AF as demonstrated in some studies [48]. Thus, for each sample, AFB1 value was obtained by halving the total AF value. Based on probable daily intake (PDI), dietary exposure to total AF and AFB1 was calculated by multiplying mean concentration in each crop by the consumption rates of each commodity and the obtained value was divided by the bw value as shown below (Equation (1)). To determine mean AF values that were used to study risk assessment, AF concentration in all samples with below the instrument LOD were replaced with instrument LOD value of 2 µg/kg. The margin of exposure (MOE) was computed using a benchmark dose lower limit (BMDL) of 170 ng/kg bw/day [49] for AF (Equation (2)).
PDI (µg/kg bw/day) = (µ × C)/bw (1) where µ = mean AF or AFB1 or FM concentration; C = quantity of maize or small grains consumed in Zimbabwe. MOE = BMDL/PDI (2) To estimate the potential impact of maize AFB1 and FM exposure on health and trade, storage datasets for AF (n = 400) and both harvest and storage datasets for FM (n = 800) were used. Estimation of the posterior risk exposure for AFB1 was carried out using a Bayesian model [50,51], based on an acute exposure estimation on hepatocellular carcinoma (HCC) effect of the toxin on humans [52]. AF-induced HCC rates (defined as Liver Cancer (LC) rates in Equation (3) below) were computed by multiplying AF potency as shown in Equation (3) by exposure risk (PDI) described in Equation (1). The average potency (AP) was derived from Equation (4) on the basis of prior values of proportions of hepatitis B virus (HBV) prevalence rates for individuals positive for the HBV surface antigen (HBsAg; a biomarker of chronic HBV infection) and HBsAg-negative individuals [27]. The proportions are estimated at 0.144 and 0.856, respectively, for Zimbabwe [52]. Therefore, in the model, AFB1 carcinogen potency prior value of 0.05176 cases per adult population in ng/kg bw/day for Zimbabwe was used based on the national HBV prevalence of 14.4%, and a national population of 13,076,978 [52]. Briefly, studies have shown that the risk of LC in individuals exposed to HBV infection and aflatoxin is 30 times greater as compared to individuals exposed to aflatoxin only [53] due to a synergistic effect of the two factors.
where AP = estimated national carcinogen potency to AFB1 exposure in ng/kg bw/day (based on cases/100,000 persons/yr (CPY)). AP = (0.3 × proportion of HBsAg-positive prevalence rate) + (0.01 × proportion of HBsAg-negative prevalence) The cut-off points for simulated AFB1 to assess the impacts for maximum limits (MLs) were set at 0 (no regulation), 4 µg/kg (EU) [54], a provisional value of 10 µg/kg (simulation using the cutoff value for AF in Zimbabwe) [28], all of which would translate into 0, 8, and 20 µg/kg for AF, respectively. ALARA (as low as reasonably achievable) values [55,56] were also set for each province. ALARA value was derived as the mean of samples up to the 95th percentile in the range of samples from lowest to highest AF concentration. For FM, the MLs cutoff points were set at 0 (no regulation), 2000 µg/kg (Codex) [28], and ALARA values of 885.6 and 2873 µg/kg for 2015/2016 and 2016/2017 seasons, respectively. Risk exposure to FM was estimated on the basis of a hazard quotient (HQ), defined as the ratio of estimated exposure to FM (Equation (1)) divided by the Provisional Maximum Tolerable Daily Intake (PMTDI) for FM that was set at 2 µg/kg bw/d [57]. Similarly, as with AF, the LOD value was used to determine mean FM concentration such that a sample whose value was below the instrument LOD of 300 µg/kg was scored as equivalent to 300 µg/kg.

Data Analysis
Incidences of Fusarium, Aspergillus, Penicillium, or other fungal species on grain were calculated as the number of kernels with visible growth of the corresponding genera divided by total number of plated kernels × 100. For mycotoxin data, all samples with AF and FM below 2 and 300 µg/kg, respectively, were given a value of zero. Thereafter, AF and FM data were subjected to a mixed model analysis of variance with location (districts) considered as a random effect. Since there were neither seasonal nor sampling time effects on AF levels, the AF data sets for the two seasons were pooled and used in descriptive analysis to determine the mean values ± SD. Data for FM were not pooled. To determine the impact of mycotoxins on health, the data was computed as described under Equation (1) for FM and Equations (1)-(4) for AF. Unlike for the mixed model analysis, datasets used for computing exposure risks were modified by replacing data points whose values were below limit of detection by 2 µg/kg (AF) and 300 µg/kg (FM). All statistical analyses were conducted using SAS statistical software version 9.4 (SAS Institute, Cary, NC, USA).

Fungal Characterization
Incidences of fungi on grains were relatively moderate, with 27-60% of samples not yielding any fungus. Fungal incidence was highest in maize, followed by finger millet, sorghum, and lowest in pearl millet ( Table 1). The predominant genera encountered were Fusarium, Aspergillus, Penicillium, Paecilomyces, Trichoderma, Phoma, Cercospora, Rhizopus, and yeasts ( Table 1). Distributions of fungal species varied across crops. In small grains, over 70% of the fungi belonged to species of Phoma and Cercospora, while for maize 84% belonged to Fusarium, Aspergillus, or Penicillium spp. Subsequent Aspergillus characterization revealed several species associated with maize and small grains: A. flavus, A. parasiticus, A. ochraceous, A. niger, and S morphotype fungi. In general, maize contained higher densities of Aspergillus spp. than small grains (data not shown). Irrespective of the crop, A. flavus L morphotype occurred at higher frequencies (>67%) than A. parasiticus or the S morphotype fungi, which were isolated in relatively  Table 2). The least frequently isolated Aspergillus species were A. niger (occurring in all crops) and A. ochraceous (occurring only in maize). The L morphotype, found in all districts, was the predominant species in all locations. On the other hand, A. parasiticus was isolated in all districts except in Muzarabani and Rushinga. The S morphotype fungi were found in eight of the 12 districts. In Guruve, L morphotype and A. parasiticus occurred in equal proportions, while in Rushinga the proportions of the L morphotype exceeded that of the S morphotype fungi by about 10% (Figure 2). frequencies (> 67%) than A. parasiticus or the S morphotype fungi, which were isolated in relatively equal proportions across crops ( Table 2). The least frequently isolated Aspergillus species were A. niger (occurring in all crops) and A. ochraceous (occurring only in maize). The L morphotype, found in all districts, was the predominant species in all locations. On the other hand, A. parasiticus was isolated in all districts except in Muzarabani and Rushinga. The S morphotype fungi were found in eight of the 12 districts. In Guruve, L morphotype and A. parasiticus occurred in equal proportions, while in Rushinga the proportions of the L morphotype exceeded that of the S morphotype fungi by about 10% (Figure 2).

Mycotoxin Levels in Grains
Overall, the incidence of AF contamination was relatively low in small grains (13% to 25%), but moderate (≈ 50%) for maize (Table 3). Maximum AF levels were 1369 µg/kg for maize, 58.0 µg/kg for pearl millet, 4.3 µg/kg for sorghum, and 3.6 µg/kg for finger millet. The average AF levels were 16.8 µg/kg for maize, 2.5 µg/kg for pearl millet, and < 2 µg/kg for both finger millet and sorghum. Over 80% of the maize had AF levels within the EU (4 µg/kg) and 89% within Zimbabwe (10 µg/kg) regulatory limits (Table 3). On the other hand, all small grain samples except for one and two pearl millet and sorghum sample(s), respectively, contained safe AF levels.

Mycotoxin Levels in Grains
Overall, the incidence of AF contamination was relatively low in small grains (13% to 25%), but moderate (≈50%) for maize (Table 3). Maximum AF levels were 1369 µg/kg for maize, 58.0 µg/kg for pearl millet, 4.3 µg/kg for sorghum, and 3.6 µg/kg for finger millet. The average AF levels were 16.8 µg/kg for maize, 2.5 µg/kg for pearl millet, and <2 µg/kg for both finger millet and sorghum. Over 80% of the maize had AF levels within the EU (4 µg/kg) and 89% within Zimbabwe (10 µg/kg) regulatory limits (Table 3). On the other hand, all small grain samples except for one and two pearl millet and sorghum sample(s), respectively, contained safe AF levels. Table 3. Aflatoxin and fumonisin prevalence in maize, sorghum, pearl millet, and finger millet grain samples obtained from Zimbabwe during two cropping seasons along with the proportion of samples exceeding regional and international regulatory limits for human consumption.

Toxin
Crop detection. c Samples with toxin concentration exceeding regulatory limit of the European Union (4 µg/kg for aflatoxin (AF) and 1000 µg/kg for fumonisin (FM)). d Samples with toxin concentration exceeding regulatory limit of Zimbabwe (10 µg/kg for AF). e Samples with toxin concentration exceeding regulatory limit of the United States Food and Drug Administration (20 µg/kg for AF and 4000 µg/kg for FM). f Not detected; limit of detection is 2 µg/kg (AF) and 300 µg/kg (FM). g Zimbabwe regulatory limit for FM is not available, therefore it was not computed.
Analysis of maize AF data both at harvest and after storage across the two seasons revealed no effect of season (P = 0.2895), sampling time (P = 0.7461), nor AEZ (P = 0.7869). Equally, no interaction occurred between these variables (P ≥ 0.6631 ≤ 0.9377). Though nonsignificant, maize from Muzarabani had the highest mean AF level. Nyanga district, with less than 60% AF positive maize samples, showed the lowest mean AF level. Rushinga, Buhera, and Muzarabani, had 12-28% of maize samples with AF levels above the US regulatory limit (20 µg/kg; Table 4). FM was either undetected or detected at very low levels in small grain samples. However, most maize samples (89%) had FM. For positive samples, FM ranged from 300 to 40,000 µg/kg in maize, 300 to 1500 µg/kg in finger millet, and 300 to 2800 µg/kg in sorghum. Mean FM levels were 2383, 412, and 127 µg/kg, for maize, sorghum, and finger millet, respectively. Less than 10% of the millets and sorghum had FM above the EU regulatory limit, while 54% of the maize exceeded that limit (Table 3).  (Table 5). Muzarabani and Nyanga districts harbored the largest proportion of samples containing FM above the regulatory limit. 40,000 µg/kg in maize, 300 to 1500 µg/kg in finger millet, and 300 to 2800 µg/kg in sorghum. Mean FM levels were 2383, 412, and 127 µg/kg, for maize, sorghum, and finger millet, respectively. Less than 10% of the millets and sorghum had FM above the EU regulatory limit, while 54% of the maize exceeded that limit (Table 3). Season and sampling time significantly influenced maize FM content (season: F = 46.48, df = 1, P < 0.0001; sampling time: F = 22.12, df = 1, P < 0.0001). However, FM levels were not influenced by AEZ (F = 0.65, df = 3, P = 0.5854). Compared to the 2015/2016 season, FM level was significantly higher in the 2016/2017 season. Maize FM level, however, depended on the interaction between season and sampling time (F = 5.91, df = 1, P = 0.0153). In both cropping seasons, FM content in samples collected at harvest was higher than in samples collected after storage (Figure 3). Additionally, FM levels in maize were influenced by the interaction between season and AEZ (F = 3.13, df = 3, P = 0.0252). At harvest, maize from the 2015/2016 season contained 3-6-fold significantly lower FM levels in the hotter and drier region IV AEZ when compared to the other three AEZs. For the 2016/2017 season, there was no statistical difference in FM levels across AEZs (Figure 3). Across districts, maize samples from Makoni, Mutasa, Chirumanzi, and Kwekwe had > 90% of samples containing FM levels within regulatory limits (Table 5). Muzarabani and Nyanga districts harbored the largest proportion of samples containing FM above the regulatory limit.

Exposure Risk Assessment: Probable Daily Intake and Mycotoxin Health and Trade Impacts
In all seasons and sampling time, dietary exposure to AF was highest among maize consumers with PDI reaching a maximum of 5593 ng/kg bw/d for total AF and 2797 ng/kg bw/d for AFB1. On the other hand, very low dietary exposure was observed among millet and sorghum consumers with PDI not exceeding 10 ng/kg bw/d (Table 6). Consequently, maize with a relatively small MOE (<5) posed higher hazardous health risks to the public when compared to the other cereal grains. Over the two cropping seasons, consumption of the sampled maize had an estimated potential LC rate of 1.8 CPY. Maximum PDI was highest in region IV AEZ and lowest in region IIA AEZ. Consumption of maize from region IV AEZ had an estimated LC rate of 4 CPY and posed the greatest dietary exposure risk to consumers (Table 7). Though exposure risks across districts followed a similar pattern to AF prevalence, the MOE values in Buhera, Chirumanzi, and Muzarabani were relatively small (1-5), highlighting a potentially higher exposure risk for consumers in these districts. Overall, maize from Muzarabani in region IV posed higher health risks to the public, potentially leading to at least 9 LC CPY (Table 8).      Table 9). The highest risk of developing LC was observed in Mashonaland followed by Manicaland (Table 9). Enforcing stringent measures/regulations of 4 µg/kg (EU) would reduce LC risks throughout the provinces to 0.2 CPY from the maximum value of 2.1 CPY (no regulation) for the most affected province. However, the potential for grain rejection in trade increased from 0% to up to 13.9% (for the most affected province). Employing the ALARA limit showed a 5-fold risk reduction for the most affected province from 2.1 CPY (no regulation) to 0.4 CPY. The potential for grain rejection in this case increased from 0% to 5.6%. Using the provisional value of 10 µg/kg as a limit, reduced the risk exposure to 0.3 CPY across the provinces. However, this increased the potential for grain rejection from 0% to 8.6% (for the most affected province).
For FM, the estimated HQ in both seasons was 2.4-17.9 times the PMTDI value across all age groups with season 2016/2017 exhibiting higher risks. Based on mean FM levels, the mean HQ amongst children (5.

Discussion
Ear and panicle rot fungi, predominantly species of Fusarium, Aspergillus, and Penicillium were recovered from maize, sorghum, pearl millet, and finger millet grown in Zimbabwe between 2015 and 2017 in relatively high frequencies. Studies conducted in Zimbabwe and elsewhere, show that F. verticillioides, A. flavus, and A. parasiticus are major mycotoxin producers contaminating cereals [32,37,[58][59][60][61][62][63]. Results from our study revealed that the maize mycobiome was dominated by Fusarium and Aspergillus spp., while in small grains other filamentous fungi dominated, mainly Phoma and Cercospora spp. These findings are similar to studies in West Africa and elsewhere in which sorghum, pearl millet, and finger millet were reported as less susceptible to AF producers compared to maize [12,31,32]. Among the aflatoxin-producing fungi, the A. flavus L. morphotype was the predominant species on both maize and small grains. Studies conducted nearly a decade ago similarly reported A. flavus as the predominant Aspergillus spp. infecting maize from northern Zimbabwe [36]. In neighboring Zambia and Mozambique, A. flavus L morphotype was also reported as the most frequently isolated species in maize, followed by A. parasiticus [59,63].
For the first time, we report the prevalence of FM and AF in finger millet that were locally produced and traded in Zimbabwean informal markets. AF and FM co-occurred in all crops except for pearl millet, which had no FM. Detection of both AF and FM in maize at harvest indicates that contamination occurs at the preharvest stage. For small grains, it is unclear if the contamination occurred at preharvest. In the current study, FM

Discussion
Ear and panicle rot fungi, predominantly species of Fusarium, Aspergillus, and Penicillium were recovered from maize, sorghum, pearl millet, and finger millet grown in Zimbabwe between 2015 and 2017 in relatively high frequencies. Studies conducted in Zimbabwe and elsewhere, show that F. verticillioides, A. flavus, and A. parasiticus are major mycotoxin producers contaminating cereals [32,37,[58][59][60][61][62][63]. Results from our study revealed that the maize mycobiome was dominated by Fusarium and Aspergillus spp., while in small grains other filamentous fungi dominated, mainly Phoma and Cercospora spp. These findings are similar to studies in West Africa and elsewhere in which sorghum, pearl millet, and finger millet were reported as less susceptible to AF producers compared to maize [12,31,32]. Among the aflatoxin-producing fungi, the A. flavus L. morphotype was the predominant species on both maize and small grains. Studies conducted nearly a decade ago similarly reported A. flavus as the predominant Aspergillus spp. infecting maize from northern Zimbabwe [36]. In neighboring Zambia and Mozambique, A. flavus L morphotype was also reported as the most frequently isolated species in maize, followed by A. parasiticus [59,63].
For the first time, we report the prevalence of FM and AF in finger millet that were locally produced and traded in Zimbabwean informal markets. AF and FM co-occurred in all crops except for pearl millet, which had no FM. Detection of both AF and FM in maize at harvest indicates that contamination occurs at the preharvest stage. For small grains, it is unclear if the contamination occurred at preharvest. In the current study, FM incidence and concentration were higher than that of AF in maize and sorghum but not for the millets. High FM content compared to AF in maize was previously reported in Zimbabwe [38,39]. In the millets, AF incidence of 13-25% was recorded while FM in these crops ranged from 0% to 3.3% (Table 3).
Though low mycotoxin contamination of small grains was recorded, we observed moderate AF (≈50%) and high FM (≈90%) incidence in maize (Table 3). AF content in maize was considerably higher than in the other crops (7-19-fold higher) and most maize had FM. Thus, consumption of maize in Zimbabwe significantly exposes people and/or livestock to both AF and FM. This has been previously documented [17]. Consumption of the less AF-and FM-susceptible small grains to diversify diets and reduce exposures to both groups of toxins should be encouraged [12,31]. Surveys conducted in various regions indicate that millet and sorghum are less prone to AF and FM than maize [12,31,40]. In West Africa, Bandyopadhyay et al. [31] found maize collected in the field to harbor 4-8 times higher AF levels than sorghum or pearl millet. Similarly, in Ethiopia, variable levels of several mycotoxins contaminating small grains were reported [32]. In Nigeria, high FM content in maize was related to higher frequencies of FM-producing fungi compared to sorghum or pearl millet [12]. Contrary to our findings, a mycotoxin study in Zimbabwe [61] found FM in 61% of sorghum samples but detected no AF. The contrast may be due to differences in seasonal mycotoxin fluctuations, sampling time, and location, which often influence mycotoxin levels.
Though AF levels were not influenced by sampling time, AEZ, or season, FM levels in maize differed across AEZ and cropping season. Generally, environmental factors (especially drought stress) dictate the prevalence and severity of mycotoxin contamination in food and feed crops [62,64]. In our study, the highest AF content was observed in the hotter and drier parts of Zimbabwe, which receive 450-600 mm rainfall annually [65]. Similarly, though not statistically significant, we observed that maize collected after the 2015/2016 cropping season and that received low rainfall (368 mm) had higher AF content than maize collected after the 2016/2017 cropping season (rainfall: 536 mm), which concurs with studies conducted elsewhere [23,64,66]. For instance, in neighboring Zambia, AF content in maize and groundnut was higher in a low rainfall and drier region compared to a cool and wet region [64]. Similarly, maize and sorghum grown in semiarid tropical areas of Kenya were more vulnerable to AF contamination than in temperate areas [67]. Likewise, Ding et al. [23] reported that groundnut cultivated in areas with low rainfall and relatively high daily mean temperature (≈25 • C), or exposed to heat stress (26-31 • C) during the last month of the growing season resulted in high AF levels [66]. Our study provides additional evidence on the influence of weather and environmental stress in favoring the proliferation of AF-producing fungi and AF severity.
Millets were remarkably less contaminated with AF and FM (Table 3). A hotter, drier season as well as a more humid season contributed to higher AF and FM, respectively, in maize and/or sorghum. Climate change is known to cause higher mycotoxin-producing fungi build up and mycotoxin contamination in susceptible crops [8,9]. Higher temperatures, greater CO 2 concentrations, erratic rainfall patterns, all are known to increase crop susceptibility to mycotoxin contamination. However, regardless of the weather fluctuations that occurred in Zimbabwe during the study, these did not readily impact mycotoxin levels in the millets. It is likely that mycotoxin contamination of prone crops, especially maize and, to some extent sorghum, will increase in the future due to climate change. Cultivation of millets is a valuable option to palliate the effects of climate change in a sustainable manner [68]. Research to understand the potential and limits of the millets in resisting mycotoxin contamination as a result of current and future environmental fluctuations caused by climate change should be conducted.
Generally, daily exposure to mycotoxin levels exceeding 1 µg/kg bw/d (AF) and 2 mg/kg bw/d (FM) are known to have adverse effects on health [17,20,22]. Furthermore, for carcinogenic substances like AF, MOE values ≥ 10,000 are considered protective. In the present study, all the cereal grains had mycotoxin levels and MOE values less than 10,000 signifying that all posed health hazard to consumers. However, maize's high estimated PDI (up to 2797 ng/kg bw/d) along with a relatively small MOE value (≤5) for AF, and high FM HQ (>1) suggested that maize consumers compared to consumers of small grains faced a potentially higher risk of developing illnesses related to AF (e.g., liver cancer, child stunting, immunosuppression) and FM (e.g., esophageal cancer, neural tube defects, immunosuppression).
Data on FM exposure revealed a high risk of chronic exposure to this toxin for all age groups over the two seasons.  [17] indicated a high risk of exposure to FM among populations in Zimbabwe. However, Hove and coworkers reported a relatively lower exposure to FM contamination (HQ up to 2.7) compared to our findings (HQ up to 7.4). This variability could be attributed to seasonal differences as well as the locations from where the samples were obtained, all of which are factors known to influence mycotoxin build up.
Overall, the most AF-affected region was AEZ IV followed by III. Those AEZ of Zimbabwe are relatively hotter and drier; thus, maize from those AEZ may pose AFexposure risks to consumers in the region. According to Wild et al. [69], all SSA countries have favorable conditions supporting high proliferation of mycotoxin-producing fungi and mycotoxin production, and as such inhabitants are highly exposed to mycotoxin health risks. Depending on the crop, the AEZ, location (district), and any possible intervention (e.g., enforcement of regulations), the estimated cancer risks of 0.2-9.2 CPY were recorded over the 2-year period. We report that enforcing stringent regulations along with dietary diversification to small grains would significantly reduce exposure risks and the negative health impacts associated with the two toxins [24]. Wu et al. [30] reported that enforcement of current regulatory AF standards of 4-20 µg/kg can adequately protect human health worldwide, except in Africa, if the protection level of 1 in 10,000-lifetime HCC cases in the population are desired. Nonetheless, application of ALARA or the adopted FM Codex limit of 2000 µg/kg for both seasons may not sufficiently protect consumer groups (Figure 4a).
From a trade perspective, the observed mycotoxin levels would have resulted in significant rejection of maize grain lots. Such rejections can lead to economic losses for the nation. For instance, we observed that the current Codex ML of 2000 µg/kg for FM, despite not translating into sufficient protection for the different consumer groups, translated into potential grain rejections of 14.3% and 50.3% for the two successive cropping seasons, respectively (Figure 4b). Similarly, up to 28% maize grains did not meet the Codex standard for AF and were liable for rejection (Table 4). Thus, it is imperative that Zimbabwe establishes and enforces appropriate interventions [24] including strict regulations and policies that support dietary diversification if consumer exposure risks to mycotoxins and associated negative health and trade impacts are to be minimized. However, the practicality of enforcing regulation in SSA settings has been questioned [67]. It is necessary to make farmers aware of the dangers of mycotoxins and enable them to implement proven technologies for mycotoxin mitigation. Greater mycotoxin reductions and exposure would be achieved with policy and institutional support.

Conclusions
Our study revealed that AF-and FM-producing fungi along with their secondary metabolites, i.e., AF and FM, commonly co-occur in four cereal grains grown in Zimbabwe. AF and FM content depended on the crop, with AF levels in maize being 7-fold, 15-fold, and 19-fold higher than that detected in pearl millet, sorghum, and finger millet, respectively. This confirms a higher susceptibility of maize to these contaminants relative to small grains, while at the same time stresses the need for dietary diversification to minimize mycotoxin exposure under climate change scenario. Mycotoxin incidence and prevalence was influenced by seasonal and agroecological factors. Crops from drier and hotter localities exhibited relatively higher AF levels, while a relatively wetter season/AEZ supported higher FM levels in maize. These results point to the need for routine surveillance and monitoring of these commodities by food regulators. The MOE for all the crops were below the established protective values of 10,000, signifying that all the crops posed detrimental health hazards to consumers. However, maize with exposure estimates of 5.5-177.7 ng/kg bw/d and relatively low MOE values showed that consumers were at risk of developing AF-related illnesses with potential cancer burdens of 0.3-9.2 CPY against small grains values of less than 0.1 CPY. Enforcement of regulations would have reduced AF cancer burdens by at least 9%. For FM, the estimated HQ exceeded the Codex PMTDI of 2 µg/kg bw/d across all provinces in both seasons. It may appear that the use of 300 µg/kg LOD for FM would result in higher HQ value compared to HQ value calculated using zero, the two HQ values were very similar to each other (data not shown). The FM level of exposure and risk posed was higher in children than adults signifying the vulnerability of this age group to mycotoxin's detrimental health effects (immunity impairment and growth retardation) attributed to chronic exposure.
Maize contamination with the two mycotoxins had potential of hindering international/regional trade by up to 13.9% for AF-related rejections and 100% for FM-related rejections. Thus, to reduce potential health hazards and boost regional or international trade, inclusion of less susceptible food crops like sorghum and millets in the diets should be promoted along with other mitigation measures like enforcement of regulations, good agricultural and storage practices, as well as advocacy and awareness campaigns. In general, millets (and to some extent sorghum) appeared to be less affected by environmental fluctuations and accumulated less mycotoxins. This may be important to consider for the frequently expected disruptive weather patterns caused by climate change. The current research presented the need for a multifaceted approach in minimizing mycotoxin exposure, most especially FM exposure, other than use of regulatory limits. Future studies should address the occurrence of other potentially harmful mycotoxins in these commodities.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy issues.