Aflatoxins: Occurrence, Exposure, and Binding to Lactobacillus Species from the Gut Microbiota of Rural Ugandan Children

Chronic exposure of children in sub-Saharan Africa to aflatoxins has been associated with low birth weight, stunted growth, immune suppression, and liver function damage. Lactobacillus species have been shown to reduce aflatoxin contamination during the process of food fermentation. Twenty-three Lactobacillus strains were isolated from fecal samples obtained from a cohort of rural Ugandan children at the age of 54 to 60 months, typed by 16S rRNA gene sequencing, and characterized in terms of their ability to bind aflatoxin B1 in vitro. Evidence for chronic exposure of these children to aflatoxin B1 in the study area was obtained by analysis of local foods (maize flour and peanuts), followed by the identification of the breakdown product aflatoxin M1 in their urine samples. Surprisingly, Lactobacillus in the gut microbiota of 140 children from the same cohort at 24 and 36 months showed the highest positive correlation coefficient with stunting among all bacterial genera identified in the stool samples. This correlation was interpreted to be associated with dietary changes from breastfeeding to plant-based solid foods that pose an additional risk for aflatoxin contamination, on one hand, and lead to increased intake of Lactobacillus species on the other.


Introduction
The warm and humid climate conditions of sub-Saharan Africa promote the growth of fungi and associated production of mycotoxins. Approximately 25% of grains harvested annually worldwide contain mycotoxins. Ingestion of these contaminated foods can lead to disease and death [1]. Aflatoxin is their urine for aflatoxin M 1 . The aflatoxin B 1 binding ability was tested of Lactobacillus species isolated from the gut microbiota of these children, and the correlation of gut Lactobacillus species with stunting for 140 children was analyzed from the same cohort at the ages of 20-24 and at 36 months.

Study Design and Sample Collection
In the current study, a small sample of ten children aged 54-60 months was selected on the basis of their previous growth indices at 36 months (5 stunted and 5 non-stunted children). These children took part in a randomized trial in the two districts of Kabale and Kisoro, located in the southwestern part of Uganda ( Figure 1). The trial assessed the effect of an educational intervention (focusing on nutrition, hygiene, and stimulation) on their growth and development as described in detail in previous publications [30,31]. Samples of the stool and urine were taken from every child in this study. The stool was sampled using a sterile disposable stool sampling container. The stool samples from every child were put in two separate containers: one container was filled with mineral oil and kept at room temperature for the purpose of cultivation, and the other container was filled with two milliliters of 15% glycerol. The stool samples with glycerol were immediately kept on ice and transferred to the Uganda Industrial Research Institute for storage at −80 • C. Urine was sampled into a sterile, disposable plastic container and immediately stored at low temperature (0-8 • C). The ten children were widely spread across the districts of Kabale and Kisoro as indicated by subject identification number ( Figure 1).

Study Design and Sample Collection
In the current study, a small sample of ten children aged 54-60 months was selected on the basis of their previous growth indices at 36 months (5 stunted and 5 non-stunted children). These children took part in a randomized trial in the two districts of Kabale and Kisoro, located in the southwestern part of Uganda ( Figure 1). The trial assessed the effect of an educational intervention (focusing on nutrition, hygiene, and stimulation) on their growth and development as described in detail in previous publications [30,31]. Samples of the stool and urine were taken from every child in this study. The stool was sampled using a sterile disposable stool sampling container. The stool samples from every child were put in two separate containers: one container was filled with mineral oil and kept at room temperature for the purpose of cultivation, and the other container was filled with two milliliters of 15% glycerol. The stool samples with glycerol were immediately kept on ice and transferred to the Uganda Industrial Research Institute for storage at −80 °C. Urine was sampled into a sterile, disposable plastic container and immediately stored at low temperature (0-8 °C). The ten children were widely spread across the districts of Kabale and Kisoro as indicated by subject identification number (Figure 1).
Anthropometric measurements were taken as described by Muhoozi et al. [30] and used to compute height-for-age Z-scores (HAZ) [32]. The frequently consumed regional foods (hulled and dehulled maize, and peanuts) at risk of aflatoxin contamination were identified based on a short food frequency questionnaire (Supplementary File S1). Common diet for the age group and last week's diet of the children were evaluated. The foods were sampled and immediately stored at low temperature (0-8 °C). The in vitro measurements were carried out at the Uganda Industrial Research Institute (Kampala, Uganda) and at the Department of Molecular Cell Biology, Vrije Universiteit (Amsterdam, The Netherlands). The Lactobacillus species were isolated, and their ability to bind aflatoxin B1 was assessed. The concentration of aflatoxin B1 was determined in the food samples and aflatoxin M1 in the children's urine. Anthropometric measurements were taken as described by Muhoozi et al. [30] and used to compute height-for-age Z-scores (HAZ) [32]. The frequently consumed regional foods (hulled and dehulled maize, and peanuts) at risk of aflatoxin contamination were identified based on a short food frequency questionnaire (Supplementary File S1). Common diet for the age group and last week's diet of the children were evaluated. The foods were sampled and immediately stored at low temperature (0-8 • C). The in vitro measurements were carried out at the Uganda Industrial Research Institute (Kampala, Uganda) and at the Department of Molecular Cell Biology, Vrije Universiteit (Amsterdam, The Netherlands). The Lactobacillus species were isolated, and their ability to bind aflatoxin B 1 was assessed. The concentration of aflatoxin B 1 was determined in the food samples and aflatoxin M 1 in the children's urine.

Aflatoxin B 1 in Food and Daily Intake
Hulled, dehulled maize flour, and peanut were analyzed using the ELISA Ridascreen ® Aflatoxin B 1 with a reported limit of detection of 1.0 µg kg −1 (R-Biopharm, Darmstadt, Germany). The enzyme immunoassay was first calibrated by the use of an aflatoxin B 1 standard (0, 1, 5, 10, 20, and 50 µg L −1 ) in 10% (v/v) methanol as indicated in Supplementary File S2. A competitive enzyme immunoassay was used for the determination of aflatoxin B 1 in cereals similar to the method described by Wacoo et al. [11]. Briefly, 5 g of homogenized maize flour was weighed into 50 mL centrifuge tubes, followed by addition of 25 mL of 70% (v/v) methanol and thorough mixing using a VWR ADC 3500 Shaker (BioSurplus, Inc, San Diego, CA, USA) for 5 min. In the case of peanuts, 0.4 g of sodium chloride was added to the suspension and thoroughly mixed. The suspension was then centrifuged for 10 min at 3500× g at room temperature. An aliquot of 50 µL of each supernatant was used for aflatoxin B 1 determination using the ELISA kit.
The minimum daily aflatoxin B 1 intake of children in southwestern Uganda (expressed in ng kg −1 day −1 ) was estimated on basis of the measured concentrations of aflatoxin B 1 in maize flour and peanut sampled in each subcounty, the estimated amounts of maize and peanut consumed, and the measured body weight of the child [33]. The frequency of intake of maize flour and peanut in southwestern Uganda was obtained by a food frequency questionnaire (Supplementary File S1). The amount of maize and peanut consumed was based on the previously reported intake estimate for maize flour and products of 60 g per day for children at the age of 24 to 59 months in southwestern Uganda [34].

Aflatoxin M 1 in Urine
Quantitative determination of aflatoxin M 1 in urine samples was carried out using the ELISA Ridascreen ® Aflatoxin M 1 designed with a limit of detection of 5 ng L −1 . The ELISA kit was validated as described by the International Conference on Harmonization (ICH) (1995) [35]. The limit of detection (LOD), precision, and accuracy were determined by using concentrations of aflatoxin M 1 standard (0, 125, 250, 500, 1000, and 2000 ppt) spiked in urine (Supplementary File S2). Briefly, the pH of the urine samples was first adjusted to seven. A volume of 50 µL of either standard or urine samples was pipetted and put into separate wells, followed by addition of 50 µL of enzyme conjugate and 50 µL of anti-aflatoxin M 1 antibody solution. The microwell plate was then mixed by shaking gently and incubated for 10 min at room temperature. The liquid was removed and the wells were washed three times using 250 µL of washing buffer. Then, the wells were filled with 100 µL of substrate and incubated for 5 min at room temperature. The reaction was stopped with 100 µL of stop solution and the absorbance taken at 450 nm wavelength using a microplate reader. The results of the standard solution were used to develop a calibration curve, and the aflatoxin M 1 levels in each urine sample were determined from this curve. Urine creatinine concentrations were determined spectrophotometrically by the modified Jaffe method [36]. The aflatoxin M 1 concentration from each sample was subsequently normalized to creatinine concentration in the urine. The percentage of aflatoxin, which is excreted as aflatoxin M 1 in urine, was calculated on basis on the estimated dietary aflatoxin intake per kg body weight per day (ng), the concentration of aflatoxin M 1 in urine (ng/mg creatinine), and the reported levels of excreted creatinine in urine per day of 15.4 mg per kg body weight per day for boys and 14.3 mg per kg body weight per day for girls [37,38].

Enumeration and Isolation of Lactic Acid Bacteria
Serial dilutions of samples were prepared in physiological saline. Total counts of lactic acid bacteria (LAB) from stool samples were determined by streaking selected serial dilutions on sterile de Man, Rogosa, Sharpe (MRS) agar (Oxoid limited, Hampshire, United Kingdom) containing 0.1% Tween 80. The plates were incubated at 37 • C for 48 h. The experiment was performed four times for every sample. After total counts of lactic acid bacteria (LAB), five colonies with distinct colony morphology were selected from each plate. The colonies were then streaked to freshly prepared MRS agar plates for identification.

Identification of Bacterial Isolates
Isolates were identified as described by Felske et al. [39]. Briefly, 16S rRNA gene fragments were amplified and sequenced using primers 8F (5 -AGAGTTTGATYMTGGCTCAG-3 ) and 1512R (5 -ACGGYTACCTTGTTACGACTT-3 ). The colony PCR reactions were carried out with 1 µL of each primer (10 pmol), 11 µL nuclease-free water (Promega), and 12 µL GoTaq Colorless Master Mix (Promega) in a final volume of 24 µL. To the PCR reaction mix, a small amount of a fresh colony was added using a sterile toothpick. The PCR program was set as follows: initial denaturation was carried out at 94 • C for 5 min, followed by 30 amplification cycles (30 s at 94 • C, 30 s at 55 • C, and 30 s at 72 • C), and a final extension step at 72 • C for 8 min. The PCR products were verified by electrophoresis on 1.5% (w/v) agarose gel, and sequencing was done by the Sanger sequencing method (Macrogen Inc., The Netherlands). Sequences were compared to sequences deposited in GenBank by using the (Basic Local Alignment Search Tool) BLAST algorithm, National Library of Medicine, Bethesda MD, USA [40]. All isolated strains are accessible from the strain collection of the Department of Molecular Cell Biology, Vrije Universiteit Amsterdam, The Netherlands, as well as the Department of Microbiology, School of Biomedical Sciences, College of Health Sciences, Makerere University, Uganda, in order to assure benefit sharing in accordance with the Nagoya protocol [41]. The correlation between the Lactobacillus species isolated at 54-60 months and microbiota at 20-24 and 36 months was assessed by matching the V4 amplicon sequences of our nearly full-length 16S rRNA sequences with our previously collected microbiota data (Sequence Read Archive SUB4476421) [31].

Aflatoxin B 1 Binding to the Isolated Lactobacillus Species
The aflatoxin B 1 binding assay was performed as described by Wacoo et al. [29]. Briefly, the isolated Lactobacillus species were cultured in de Man, Rogosa, Sharpe (MRS) broth with 0.1% (v/v) Tween 80 at 37 • C for 24 h. The cells were pelleted at 3200 g for 10 min at room temperature and subsequently washed twice with physiological saline to remove excess MRS broth. The washed cell pellets were re-suspended in 2 mL of physiological saline solution. Each suspension was then serially diluted with physiological saline to obtain approximately 10 8 cfu mL −1 . These dilutions were centrifuged, and the cell pellets re-suspended in 1.0 mg mL −1 of aflatoxin B 1 followed by incubation at 37 • C for 30 min. After incubation, the aflatoxin B 1 cell suspensions were centrifuged at 3200× g for 10 min at room temperature, and the residual aflatoxin B 1 in the supernatant was determined using the Fluostar Omega microplate reader (BMG Labtech, Ortenberg, Germany) at an excitation of 390 nm and an emission 480 nm. Bound aflatoxin B 1 was calculated by use of the formula below:

Statistical Analysis
For comparison between Lactobacillus species at 20-24 and 36 months, species were identified with BLAST of 16S rRNA amplicon sequences, and the data were presented in a pie chart of Lactobacillus species as the average percentage of the total Lactobacillus per individual. The abundances of Lactobacillus species were calculated by the percentage of all V4 16S rRNA sequence reads from the 23 isolates showing a 100% identity match in the total pool of unique sequence reads at 20-24 months and 36 months. The prevalence of Lactobacillus species was calculated as the percentage of gut microbiota compositions from the 140 children containing at least one exact match to the specific V4 sequence read from one of the 23 Lactobacillus isolates. A permutational multivariate analysis of variance (PERMANOVA) was carried on gut microbiota composition of all rural Ugandan children (n = 140) at 20-24 and 36 months and growth development scores. Analysis was performed using R version 3.3.2, with PERMANOVA as implemented in the 'vegan' package by Oksanen et al. using the Bray-Curtis distance measure [42]. The 16S rRNA gene sequencing data were rescaled and transformed using Wisconsin double transformation and square root transformation. The correlations between Lactobacillus species abundance in stunted (HAZ < −2.0) and nonstunted (HAZ > −2.0) children were displayed in violin plots by the use of OriginPro 2019b 9.6.5.169 (Academic).

Ethical Clearance
The study was approved by the Research Ethics committee of The AIDS Support Organization (no. TASOREC/06/15-UG-REC-009) and by the Uganda National Council for Science and Technology (no. UNCST HS 1809).

Aflatoxin B 1 Contamination in Food
In our survey on the consumption of particular foods for southwestern Uganda, beans were found to be the most frequently consumed food (Table S1). This was followed by posho/porridge (corn bread or porridge), greens (eshiga), and Irish potatoes, which were consumed at least once a day. Sweet potatoes and fermented porridge were consumed five to six times a week. Peanut, millet and sorghum porridge were taken four times a month, and dry maize with beans, cassava, and rice was eaten two to three times a month. The foods with a very high risk of aflatoxin contamination were posho/porridge, which was consumed daily, and peanuts. The intake of these foods may result in accumulation of aflatoxin in the body; thus, the ingredients of posho/porridge (maize) and peanuts were selected for further analysis.
The levels of aflatoxin B 1 in hulled, dehulled maize, and peanut sampled from Kabale and Kisoro district of southwestern Uganda are shown in Figure 2. All of these food ingredients contained mean aflatoxin B 1 levels above the acceptable East African regulatory limit of 5 µg kg −1 for aflatoxin B 1 [7]. The mean aflatoxin B 1 concentration in hulled maize flour was 9.1 µg kg −1 . Only 10% of the hulled maize flour samples contained undetectable levels of aflatoxin B 1 . Dehulled maize flour contained comparable concentrations of aflatoxin B 1 with a mean of 5.3 µg kg −1 , 6% higher than the East African regulatory limit of 5 µg kg −1 [7]. Approximately 50% of dehulled maize flour contained detectable levels of aflatoxin B 1 . Peanut samples contained, on average, higher concentrations of aflatoxin B 1 with 12.8 µg kg −1 . Approximately, 90% of the peanut samples contained aflatoxin B 1 with levels ranging from 1.8 to 20.2 µg kg −1 . More than 50% of the peanut samples contained aflatoxin B 1 levels higher than the East African regulatory limit of 5 µg kg −1 .

Ethical Clearance
The study was approved by the Research Ethics committee of The AIDS Support Organization (no. TASOREC/06/15-UG-REC-009) and by the Uganda National Council for Science and Technology (no. UNCST HS 1809).

Aflatoxin B1 Contamination in Food
In our survey on the consumption of particular foods for southwestern Uganda, beans were found to be the most frequently consumed food (Table S1). This was followed by posho/porridge (corn bread or porridge), greens (eshiga), and Irish potatoes, which were consumed at least once a day. Sweet potatoes and fermented porridge were consumed five to six times a week. Peanut, millet and sorghum porridge were taken four times a month, and dry maize with beans, cassava, and rice was eaten two to three times a month. The foods with a very high risk of aflatoxin contamination were posho/porridge, which was consumed daily, and peanuts. The intake of these foods may result in accumulation of aflatoxin in the body; thus, the ingredients of posho/porridge (maize) and peanuts were selected for further analysis.
The levels of aflatoxin B1 in hulled, dehulled maize, and peanut sampled from Kabale and Kisoro district of southwestern Uganda are shown in Figure 2. All of these food ingredients contained mean aflatoxin B1 levels above the acceptable East African regulatory limit of 5 µ g kg −1 for aflatoxin B1 [7]. The mean aflatoxin B1 concentration in hulled maize flour was 9.1 µ g kg −1 . Only 10% of the hulled maize flour samples contained undetectable levels of aflatoxin B1. Dehulled maize flour contained comparable concentrations of aflatoxin B1 with a mean of 5.3 µ g kg −1 , 6% higher than the East African regulatory limit of 5 µ g kg −1 [7]. Approximately 50% of dehulled maize flour contained detectable levels of aflatoxin B1. Peanut samples contained, on average, higher concentrations of aflatoxin B1 with 12.8 µ g kg −1 . Approximately, 90% of the peanut samples contained aflatoxin B1 with levels ranging from 1.8 to 20.2 µ g kg −1 . More than 50% of the peanut samples contained aflatoxin B1 levels higher than the East African regulatory limit of 5 µ g kg −1 .

Daily Intake of Aflatoxin B 1
Estimates for the daily intake of aflatoxin B 1 by the children of southwestern Uganda are shown in Table 1. The dietary aflatoxin B 1 intake varied from 1.12 to 88.6 ng kg −1 day −1 . The overall mean dietary aflatoxin B 1 exposure was 50.1 ng kg −1 day −1 . The mean dietary aflatoxin B 1 exposures to stunted and nonstunted children were 53.7 and 46.5 ng kg −1 day −1 , respectively. Although there was a notable difference in the mean dietary aflatoxin B 1 exposure to the stunted and nonstunted children, statistically the difference was not significant (p > 0.5). Table 1. Height-for-age Z-scores (HAZ), weight, height, lactic acid bacteria (LAB) in stool and mycotoxin concentration in urine (AFM 1 , aflatoxin M 1 ), and estimated daily aflatoxin intake (AFB 1 , aflatoxin B 1 ) of ten Ugandan children from Kabale and Kisoro, southwestern Uganda, at the age of 54-60 months. Values for lactic acid bacteria (LAB) in stool and AFM 1 are means of three independent experiments. The estimated daily aflatoxin intake levels were based on measured AFB 1 levels in maize and peanut in each subcounty (the means are presented in Figure 2).

Aflatoxin M 1 in Urine
The results of aflatoxin M 1 analyzed from the urine samples of both nonstunted and stunted children are shown in Table 1, which also shows anthropometric parameters for the same children. The aflatoxin M 1 levels in the urine samples varied from 14.8 to 168 pg mg −1 of creatinine. There was no significant difference between aflatoxin M 1 levels in nonstunted children and stunted children (p > 0.05). A detectable amount of aflatoxin B 1 in maize flour and peanut coincided with detectable levels of aflatoxin M 1 in urine samples for all children from the different study regions. No positive correlation was observed between the minimum estimated daily intake of aflatoxin B 1 and the levels of aflatoxin M 1 found in urine samples.

Enumeration of Lactic Acid Bacteria
Lactic acid bacteria from the stool were counted and the results shown in Table 1. The lactic acid bacterial count in the stool samples varied from 8.9 × 10 6 to 1.9 ×10 8 cfu g −1 . There was no significant difference between the bacterial counts between the nonstunted children (on average 7.0 × 10 7 cfu g −1 ) and stunted children (on average 6.9 × 10 7 cfu g −1 ). Approximately 10% of the samples contained lactic acid bacteria at the level of 10 6 cfu g −1 . More than 50% of the samples contained lactic acid bacteria at the level of 10 7 cfu g −1 and 30% contained lactic acid bacterial concentrations of 10 8 cfu g −1 .

Isolation, Identification, and Aflatoxin B 1 Binding Properties of Lactobacillus Species
In this study, 23 Lactobacillus strains were isolated from fecal samples of 10 children aged 54 to 60 months ( Table 2). The identification of the strains was performed based on 16S rRNA gene sequencing [31]. As nearly full-length 16S rRNA sequencing was applied, only one unambiguous identification was found on the species level of either L. casei or L. paracasei with both a percentage identity of 96% (Supplementary File S3). Though the same species of Lactobacillus was repeatedly isolated among subjects, they did not demonstrate equal binding potential to aflatoxin B 1 . Strains of the species of L. casei were most frequently isolated, accounting for 30% of the total Lactobacillus isolates, but it was also the most prevalent appearing in over 50% of the subjects. Both L. plantarum and L. brevis accounted for 14% of the total isolates each. However, L. plantarum was isolated in approximately 40% and L. brevis was isolated from only 20% of the subjects. * Identity value is based on the match of the sequenced nearly full-length 16S rRNA gene of the Lactobacillus isolate with the 16S rRNA sequence from GenBank. ** Lactobacillus isolates with unique APW strain coding were assigned to species with distinct 16S rRNA gene sequences. *** Abundance and prevalence values were based on 100% identity matches of the unique V4 regions of the sequenced 16S rRNA genes of the isolates with the V4 sequences present in the gut microbiota data of the cohort of rural Uganda children. Values cannot be unambiguously assigned to Lactobacillus strains and are only a partial representation of the Lactobacillus species in the gut microbiota.
All isolated Lactobacillus strains were assessed for their ability to bind aflatoxin B 1 [30]. All Lactobacillus isolates demonstrated binding of aflatoxin B 1 in physiological saline as shown in Table 2 at slightly variable levels. The best aflatoxin B 1 binding was registered for L. fermentum APW1317 and L. casei APW2213C of 76.1% and 62.6% at 10 8 cfu mL −1 cell concentration. It should be noted that no systematic differences in binding ability were observed between Lactobacillus species or subjects. The amount of aflatoxin bound to the bacteria increased with a rise in cell density from 4.0 × 10 7 to 8.1 × 10 7 cfu mL −1 . However, some strains indicated higher aflatoxin B 1 binding at relatively low cell densities. This is attributed to the ability of these bacteria to coagulate and form clumps at high cell densities resulting in a smaller cell surface area for binding.
The Lactobacillus species in the gut microbiota of the children at 20-24 and 36 months accounted for 2.2% and 3.4% of the total sequence reads, respectively. All the isolated Lactobacillus species at the age of 54-60 months were found to be present at 20-24 and at 36 months, except for L. rhamnosus and L. pantheris, which could not be detected at 20-24 months. Generally, the abundance of Lactobacillus species isolated at 54-60 months, which were present at 20-24 months, varied from 0 to 0.03%. The food-derived L. fermentum was the most abundant (0.03%) and the most prevalent isolate (22.9%) found amongst the 140 children at 20-24 months (Table 2).

Distribution of Lactobacillus Species in Stool Samples
The distribution of Lactobacillus species in the stool samples of the Ugandan children is shown in Figure 3. At the age of 20-24 months, L. ruminis was the most dominant species of Lactobacillus, accounting for approximately 64.2% of the total Lactobacillus species composition of the gut microbiota. L. ruminis had an abundance of 1.3% of the total Lactobacillus species at 20-24 months. L. salivarius accounted for approximately 30.2% with abundance of 0.6% at 20-24 months. L. delbrueckii and L. fermentum accounted for 4.3% and 1% with corresponding abundance of 0.09% and 0.03%, respectively. The relative abundance of Lactobacillus species to the gut microbiota increased from 2.17% to 3.42% of all 16S rRNA sequence counts in children from 20-24 months to 36 months. Approximately four more dominant species of Lactobacillus emerged at the age of 36 months at the expense of L. salivarius and L. ruminis. There was a notable shift from autochthonous (endogenous) to allochtonous (plant derived) Lactobacillus species, most probably resulting from the change in diet from breast milk to solid food. The allochtonous Lactobacillus species L. brevis emerged at the age of 36 months and became the most dominant species with 34.8%, while the endogenous species L. salivarius and L. ruminis dropped from 64.2% and 30.2% to 6.1% and 0.4%, respectively. L. plantarum, L. delbrueckii, and L. fermentum increased from less than 0.01%, 4.3%, and 1.3% at 20-24 months to 27.8%, 17.9%, and 5.8% at 36 months, respectively. The other prominent species L. kefiri and L. casei were also found at the age of 36 months, accounting for 3.7% and 3.5%, respectively. for 2.2% and 3.4% of the total sequence reads, respectively. All the isolated Lactobacillus species at the age of 54-60 months were found to be present at 20-24 and at 36 months, except for L. rhamnosus and L. pantheris, which could not be detected at 20-24 months. Generally, the abundance of Lactobacillus species isolated at 54-60 months, which were present at 20-24 months, varied from 0 to 0.03%. The food-derived L. fermentum was the most abundant (0.03%) and the most prevalent isolate (22.9%) found amongst the 140 children at 20-24 months (Table 2).

Distribution of Lactobacillus Species in Stool Samples
The distribution of Lactobacillus species in the stool samples of the Ugandan children is shown in Figure 3. At the age of 20-24 months, L. ruminis was the most dominant species of Lactobacillus, accounting for approximately 64.2% of the total Lactobacillus species composition of the gut microbiota. L. ruminis had an abundance of 1.3% of the total Lactobacillus species at 20-24 months. L. salivarius accounted for approximately 30.2% with abundance of 0.6% at 20-24 months. L. delbrueckii and L. fermentum accounted for 4.3% and 1% with corresponding abundance of 0.09% and 0.03%, respectively. The relative abundance of Lactobacillus species to the gut microbiota increased from 2.17% to 3.42% of all 16S rRNA sequence counts in children from 20-24 months to 36 months. Approximately four more dominant species of Lactobacillus emerged at the age of 36 months at the expense of L. salivarius and L. ruminis. There was a notable shift from autochthonous (endogenous) to allochtonous (plant derived) Lactobacillus species, most probably resulting from the change in diet from breast milk to solid food. The allochtonous Lactobacillus species L. brevis emerged at the age of 36 months and became the most dominant species with 34.8%, while the endogenous species L. salivarius and L. ruminis dropped from 64.2% and 30.2% to 6.1% and 0.4%, respectively. L. plantarum, L. delbrueckii, and L. fermentum increased from less than 0.01%, 4.3%, and 1.3% at 20-24 months to 27.8%, 17.9%, and 5.8% at 36 months, respectively. The other prominent species L. kefiri and L. casei were also found at the age of 36 months, accounting for 3.7% and 3.5%, respectively.

Lactobacillus Inversely Correlates with Growth
The permutational multivariate analysis of variance carried out on gut microbiota composition of rural Ugandan children (n = 140) with anthropometric and cognitive development scores indicated a number of significant correlations. The most pronounced variable was age (20-24 and 36 months) explaining 4.03% of the variance in the microbiota composition with p = 0.001 (Supplementary File S4). The anthropometric measures HAZ, and stunting, explained respectively 0.91% and 0.84% of the variance in the gut microbiota (p = 0.001). The genus Lactobacillus appeared as the genus with the highest correlation coefficient (0.014) for stunting among all 256 taxonomic units of bacterial genera identified in the gut microbiota. This was also evident from a representation of Lactobacillus abundance in the gut microbiota of Ugandan children at 20-24 and 36 months for stunted and nonstunted children ( Figure 4A). The average number of Lactobacillus sequence reads was 801 for stunted children and 423 for nonstunted children. However, it should be noted that this difference was not significant (p > 0.05) in non-parametric tests for non-normally distributed data sets. A closer inspection on the Lactobacillus species level ( Figure 4B-D) revealed that the most predominant Lactobacillus species followed this trend, except for L. salivarius at 36 months, which appeared more abundantly present in non-stunted children.
of rural Ugandan children (n = 140) with anthropometric and cognitive development scores indicated a number of significant correlations. The most pronounced variable was age (20-24 and 36 months) explaining 4.03% of the variance in the microbiota composition with p = 0.001 (Supplementary File S4). The anthropometric measures HAZ, and stunting, explained respectively 0.91% and 0.84% of the variance in the gut microbiota (p = 0.001). The genus Lactobacillus appeared as the genus with the highest correlation coefficient (0.014) for stunting among all 256 taxonomic units of bacterial genera identified in the gut microbiota. This was also evident from a representation of Lactobacillus abundance in the gut microbiota of Ugandan children at 20-24 and 36 months for stunted and nonstunted children ( Figure 4A). The average number of Lactobacillus sequence reads was 801 for stunted children and 423 for nonstunted children. However, it should be noted that this difference was not significant (p > 0.05) in non-parametric tests for non-normally distributed data sets. A closer inspection on the Lactobacillus species level ( Figure 4B-D) revealed that the most predominant Lactobacillus species followed this trend, except for L. salivarius at 36 months, which appeared more abundantly present in non-stunted children.

Discussion
A high exposure to aflatoxin-contaminated food negatively correlates with impaired growth in children [4]. The current study provides evidence that Ugandan rural children are exposed to high

Discussion
A high exposure to aflatoxin-contaminated food negatively correlates with impaired growth in children [4]. The current study provides evidence that Ugandan rural children are exposed to high concentrations of aflatoxin B 1 on a daily basis through consumption of contaminated food stuffs. Analysis of a set of regional maize flour samples indicated higher levels for hulled maize compared to dehulled maize. Siwela et al. were able to reduce up to 92% of aflatoxin contamination in maize through the process of dehulling [43]. Notably, the hulled maize with relatively high contamination levels is the most preferred due to the low costs. The dehulled maize is usually purchased by people categorized in the middle income class.
The estimated intake level to dietary aflatoxin in Uganda ranges from 10 to 180 ng per kg body weight per day [44]. In this study, the minimal estimated dietary aflatoxin intake levels for 80% of the children were found to be within this reported range for Uganda. The finding of 20% of the children with a minimal aflatoxin intake of less than 10 ng per kg body weight per day could result from the coincidental regional sampling of maize with a short storage time. However, lower intake values have been reported for other East African countries including Kenya (4−133 ng per kg body weight per day) [45].
The children's exposure to aflatoxin B 1 was confirmed by analysis of their urine samples for the presence of aflatoxin M 1 . As reported previously, approximately 1.2% to 2.2% of the dietary aflatoxin B 1 intake can be excreted in urine as aflatoxin M 1 [46]. If two minimal intake levels below 10 ng per kg body weight are considered outliers, the average percentage of aflatoxins excreted in the urine as aflatoxin M 1 equals 3.3% ± 2.8%. As in this study minimal intake levels for aflatoxin for B 1 were determined, this average percentage will decrease if aflatoxin intake from other food sources such as millet, sorghum, and beans will be included in the analysis.
The ability of lactic acid bacteria including Lactobacillus to protect against food mutagens such as aflatoxins, heterocyclic amines, and phytate among others has been reported in other studies [25,26,47,48]. Physical binding to the bacterial cell wall is reported as one of the mechanisms for the mitigation of aflatoxins from the intestine [49]. In the current study Lactobacillus species were isolated, characterized and their aflatoxin B 1 binding ability was tested. The Lactobacillus strains demonstrated a variable ability to bind aflatoxin B 1. This variation could be attributed to the differences in structure of the proposed binding surfaces such as cell wall polysaccharides, peptidoglycan, teichoic acid, and cell wall proteins, which are known to be variable among bacterial strains of the same species [48,50].
Previous studies demonstrated that gut microbiota of newborns evolves rapidly during the first 12 months of life, remains highly dynamic up to the age of 24 of months, and becomes more stable afterwards [51]. The diet contributes significantly to this modulation [52]. At the age of 20-24 months, most of the children in our cohort of 511 children were still taking breast milk, while at the age of 36 months this was replaced by solid foods [30]. This could explain the change in composition from the autochthonous species L. salivarius and L. ruminis at 20-24 months to the allochthonous species L. plantarum, L. brevis, L. delbrueckii, L. casei, and L. fermentum at 36 months. The latter five species typically originate from plant-based foods.
A substantial part of the solid foods prepared for babies is at risk for contamination with aflatoxins [53,54]. Food contaminated with aflatoxin B 1 was found to affect the gut and injures the stomach and the intestine [55]. Studies carried out in animal models have shown that aflatoxin B 1 promotes intestinal damages through perturbation of the intestinal barrier and activation of cell apoptosis and cell proliferation [55]. Saran et al. hypothesized that stunted children may fail to grow due to injury as a result of recurrent infections to the gut epithelium leading to impaired gut-mediated immunity, poor nutrient absorption, and poor appetite [56]. Thus, exposure to aflatoxin through solid foods from an early age could contribute to the high levels of stunted children observed in our cohort of 511 children [30].
In this study, the allochthonous Lactobacillus species isolated from the gut microbiota of Uganda children, such as L. casei, L. plantarum, L. fermentum, and L. brevis strains were shown to bind aflatoxins effectively, as also observed for these species in previous studies [57]. Therefore, it seems counterintuitive to find a significant, positive correlation for the abundance of the genus Lactobacillus and stunting, as the presence of this bacterial genus in the small intestine could possibly reduce the uptake of aflatoxins through binding. However, the abundance of Lactobacillus originating from plant-based foods may be indicative for the intake of relatively high levels of aflatoxin-contaminated foods. In contrast to the other Lactobacillus species, L. salivarius at 36 months appeared more abundant in nonstunted children. This could be related to the fact that this is a true endogenous or autochthonous Lactobacillus species in the human gut, and it has been shown to negatively correlate to Shigella-induced diarrhea in African children [58].
At this point it is not clear under which conditions binding of aflatoxin to Lactobacillus in the gut is most effective. Our binding assay has been performed in physiological saline with a neutral pH, which is different from the relatively low pH and other environmental conditions in the upper small intestine, where aflatoxins are absorbed [59]. As bacterial growth predominantly takes place in the colon, leading to concentrations up to 10 11 bacteria per mL, concentrations of Lactobacillus in the duodenum may be too low (approximately 10 6 bacteria mL −1 ) to effectively remove aflatoxin through binding [60,61]. Although further studies are needed to warrant their health benefits, we propose that the Lactobacillus species isolated from Ugandan children in this study can be further developed as locally sourced probiotics [62] and are promising candidates for decontaminating of aflatoxins through fermentation of maize-containing foods prior to consumption, as recently shown for L. rhamnosus [29].
Supplementary Materials: The following are available online at http://www.mdpi.com/2076-2607/8/3/347/s1, File S1: The frequency of consumption of a particular food type for southwestern Ugandans, File S2: The limit of detection (LOD), precision, and accuracy of the immunoassays used for aflatoxin determination in urine and food samples, File S3: Collection 16S rRNA gene sequences for Lactobacillus species identification, File S4: PERMANOVA correlation coefficients and p-values.