Hydrolytic fate of deoxynivalenol-3-glucoside during digestion.

Deoxynivalenol-3-β-D-glucoside (D3G), a plant phase II metabolite of the Fusarium mycotoxin deoxynivalenol (DON), occurs in naturally contaminated wheat, maize, oat, barley and products thereof. Although considered as a detoxification product in plants, the toxicity of this substance in mammals is currently unknown. A major concern is the possible hydrolysis of the D3G conjugate back to its toxic precursor mycotoxin DON during mammalian digestion. We used in vitro model systems to investigate the stability of D3G to acidic conditions, hydrolytic enzymes and intestinal bacteria, mimicking different stages of digestion. D3G was found resistant to 0.2 M hydrochloric acid for at least 24 h at 37 °C, suggesting that it will not be hydrolyzed in the stomach of mammals. While human cytosolic β-glucosidase also had no effect, fungal cellulase and cellobiase preparations could cleave a significant portion of D3G. Most importantly, several lactic acid bacteria such as Enterococcus durans, Enterococcus mundtii or Lactobacillus plantarum showed a high capability to hydrolyze D3G. Taken together these data indicate that D3G is of toxicological relevance and should be regarded as a masked mycotoxin.

Introduction

Deoxynivalenol (DON) is a Fusarium mycotoxin, belonging to the class of trichothecenes (e.g. reviewed by JECFA, 2001). A more recent review discusses the mechanisms of action, human exposure and toxicological relevance of this substance (Pestka, 2010). In brief, DON inhibits eucaryotic protein synthesis and alters cell signaling, differentiation and proliferation, which will ultimately result in cellular death. DON can often be found in cereal-based food and feed, and is therefore regulated by several countries. For instance in Europe, maximum levels depend on the foodstuff itself as well as the intended purpose and range from 200 μg/kg for processed cereal-based food or babyfood up to 1750 μg/kg for unprocessed durum wheat, oats and maize (EC Regulation No 1126/2007).

Glucosylation, a reaction occurring in phase II metabolism of plants, represents a major route to detoxify xenobiotics (reviewed in Bowles et al., 2006). Phase II conjugates can either be incorporated into the insoluble fraction of the plant cell wall (phase III metabolism) or converted into a soluble form and transferred into plant cell vacuoles. Experiments with radiolabeled mycotoxins in maize cell suspension cultures indicated that around 10% of the initial radioactivity of 14C-DON was incorporated as insoluble “bound residue” in the plant matrix (Engelhardt et al., 1999). Although the bioavailability rates of mycotoxins from bound residues are largely unexplored, DON bound residues seem to be of limited toxicological relevance.

The situation might be entirely different for the soluble DON-3-β-d-glucoside (D3G, Fig. 1), which is formed from DON in Fusarium infected plants and stored in the vacuole. Such a glucose conjugate of DON was already postulated in the eighties (Miller et al., 1983; Young et al., 1984). Later, it was possible to verify the structure of this conjugate as D3G, which was chemically synthesized (Savard, 1991) and isolated from DON treated maize cell suspension cultures (Sewald et al., 1992). For the first time, we reported the occurrence of D3G in naturally contaminated wheat and maize (Berthiller et al., 2005). Sasanya et al. (2008) showed that the mean concentrations of D3G in selected hard red spring wheat samples exceeded the mean DON concentrations. D3G was also found in naturally contaminated barley as well as in malt (Lancova et al., 2008) and beer (Kostelanska et al., 2009) made thereof. We studied the occurrence of D3G in naturally contaminated cereals (Berthiller et al., 2009a), showing that over 30% of the extractable total DON can be present as D3G in maize. Recently, D3G was also detected in oats to a level similar to that in other cereals (Desmarchelier and Seefelder, 2011). The worldwide occurrence of D3G was confirmed after identification of D3G in Chinese wheat and maize samples in the same concentration range as DON (Li et al., 2011).

Fig. 1

Structure of deoxynivalenol-3-glucoside.

D3G is far less active as protein biosynthesis inhibitor than DON, as demonstrated with wheat ribosomes in vitro (Poppenberger et al., 2003). The glucosylation reaction is therefore considered a detoxification of DON in plants. Wheat lines which are able to more efficiently convert DON to D3G, are more resistant towards the spread of the DON producing fungus Fusarium graminearum inside the plant (Lemmens et al., 2005). A quantitative trait locus responsible for Fusarium spreading resistance, which co-localizes with the DON to D3G conversion capability is incorporated into newly released wheat cultivars worldwide (Buerstmayr et al., 2009). These (whole) plants contain up to 10 fold more D3G than DON when treated with DON (Lemmens et al., 2005). It is thus expected that the incorporation of this detoxification mechanism into highly Fusarium susceptible cultivars will lead to an increase of the D3G/DON ratio also in natural infection. A DON-glucosyltransferase gene from barley has been recently identified (Schweiger et al., 2010), which might be utilized in transgenic approaches to increase Fusarium resistance by overexpression of this gene.

Yet, the fate of D3G after digestion by mammals is largely unknown, and the concern is that this compound may be cleaved to DON and glucose (reviewed by Berthiller et al., 2009b). Another conjugated Fusarium mycotoxin, zearalenone-14-β-d-glucoside (Z-14-G), was shown to produce zearalenone (ZEN) in the digestive tract of swine (Gareis et al., 1990). This reaction was believed to be largely due to the activity of the gut microbiota of animals (Gareis, 1994). In this work, the stability of D3G towards hydrochloric acid, artificial stomach juice, artificial non-microbial gut juice, a variety of enzymes and intestinal bacteria is evaluated and discussed.

Material and methods

Chemicals and reagents

D3G was isolated from wheat plants treated with DON at anthesis, as previously described (Berthiller et al., 2005). The mycotoxin DON was purchased from Romer Labs (Tulln, Austria) as calibrant in acetonitrile. HPLC grade methanol was purchased from J.T. Baker (Deventer, The Netherlands), MS grade ammonium acetate from Sigma–Aldrich (St. Louis, MO, USA). LC grade water was produced with a Millipore Milli-Q plus system (Molsheim, France) after reverse osmosis. Possible hydrolysis of D3G to DON was tested with the following solutions: (1) purified water; (2) 0.02 M HCl (pH approx. 1.7); (3) 0.2 M HCl (pH approx. 0.7); (4) artificial stomach juice (540 mg Helo-acid, Rösch und Handel, Wien, Austria, containing pepsin, in 0.02 M HCl); (5) artificial, non-microbial, gut juice (70 mg Kreon 40,000, Solvay Pharma, Klosterneuburg, Austria, containing 40 mg pancreatin (4000 lipase units, 2500 amylase units, 160 protease units) in 1 g/L NaHCO3, pH 8.0); (6) almond β-glucosidase (EC 3.2.1.21, Sigma–Aldrich G4511, 1 U/mL in 0.1 N sodium acetate buffer, pH-value 5.0); (7) β-glucuronidase (EC 3.2.1.31, isolated from Helix pomatia, Sigma–Aldrich G7396, 10 U/mL in 0.1 N sodium acetate buffer, pH 5.0); (8) cellulase (EC 3.2.1.4, from Trichoderma reesei, Sigma–Aldrich C8546, 1 U/mL in 0.1 N sodium acetate buffer, pH 5.0); (9) cellobiase (EC 3.2.1.21, isolated from Aspergillus niger, Sigma–Aldrich 49291, 130 mU/mL in 50 mM sodium phosphate buffer containing 5 mM EDTA, pH 6.0).

Bacterial strains

The following type strains (acquired from DSMZ, Braunschweig, Germany; ATCC, Manassas, VA, USA; LMG, Gent, Belgium) and isolates from the strain collection of the Department of Food Science and Technology of the University for Natural Resources and Life Sciences Vienna were used in this study: Bifidobacterium adolescentis (DSM 20083, isolate B14A); Bifidobacterium bifidum (DSM 20456, isolates B16A and B43A); Bifidobacterium longum (DSM 20219, isolates B8A and B40A); Citrobacter freundii (isolate IMB-UE5); Enterobacter cloacae (isolate IMB-UE36); Enterococcus avium (DSM 20679); Enterococcus casseliflavus (DSM 20680); Enterococcus durans (DSM 20633, ATCC 6056); Enterococcus faecalis (DSM 2918, DSM 3320, DSM 6177, DSM 20477, ATCC 27270, LMG 11733, LMG 12292, LMG 14407, LMG 15076, LMG 16003, LMG 16196); Enterococcus faecium (LMG 16003, LMG 16196); Enterococcus gallinarum (DSM 20628, DSM 20717); Enterococcus mundtii (DSM 4838); Escherichia coli (DSM 613, DSM 30083); Lactobacillus amylovorus (DSM 20531); Lactobacillus crispatus (DSM 20584, isolate L390); Lactobacillus fermentum (DSM 20052, isolate L381); Lactobacillus gasseri (DSM 20243, isolates L357 and L421); Lactobacillus paracasei (DSM 5622, isolates L365 and L367); Lactobacillus plantarum (DSM 20174); and Lactobacillus rhamnosus (DSM 20021, isolates L359 and L380).

Acidic and enzymatic hydrolysis of D3G

1 mL aliquots of 25 mg/L stock solutions of D3G (according to 25 μg pure substance) or DON (stability control) in methanol as well as of pure methanol (negative control) were transferred into 15 mL polypropylene tubes (Sarstedt, Nümbrecht, Germany) and evaporated to dryness at room temperature under a gentle stream of nitrogen for each experiment. After adding 10 mL of appropriate acidic or enzymatic solution the closed tubes were shaken for 3 h or 18 h at 30 rpm on a overhead shaker (Labor-Brand, Gießen, Germany) in a compartment drier (Heraeus, Wien, Austria) at 37 °C. 1 mL of the incubated solutions were diluted with 1 mL methanol/water (1/1, v/v), filtered through 0.22 μm Millex-GV membrane filters (Millipore, Molsheim, France) and stored at −20 °C until analysis by LC–MS/MS. The molar amount of released DON was used for the calculation of the extent of hydrolysis. All reactions were performed in triplicates.

Hydrolysis of D3G using human cytosolic β-glucosidase

Recombinant human cytosolic β-glucosidase (hCBG; 20 mU/mL final concentration) was combined with 25 μg D3G in a reaction volume of 100 μL in 50 mM sodium phosphate buffer pH 6.0 with 5 mM EDTA. Reactions were set up in triplicate. Reactions set up with DON and enzyme or with D3G without enzyme served as controls. Directly after mixing, as well as after 10, 20, 30, 45, 60, 90, 120, 180 min and 18 h at 37 °C, 10 μl of the incubation were mixed with 90 μl of ethanol. Samples were stored at −20 °C until analysis by LC–MS/MS.

Incubation of D3G with bacterial suspensions

0.375 μg D3G in 15 μL saline magnesium buffer (100 mM NaCl, 8 mM MgSO4, 50 mM Tris–HCl, pH 7.5) were combined with 135 μL of bacterial suspensions (OD600 about 2.0), giving the same concentration of 2.5 mg/L of D3G as with the enzymatic reactions. Bacteria were incubated for 4 h and 8 h at 30 °C or 37 °C according to the optimal growth conditions of the microbes, centrifuged at 13,000 rpm for 5 min and 300 μL of ethanol were added to the supernatant. Before analysis with LC–MS/MS, the solutions were dried under nitrogen and re-suspended in water. The amount of released DON was used for the calculation of the extent of hydrolysis.

LC–MS/MS parameters

An 1100 Series HPLC System (Agilent, Waldbronn, Germany) in conjunction with a QTrap-LC–MS/MS System (Applied Biosystems, Foster City, USA) equipped with a Turbo Ion Spray source were used for analysis. Isocratic separation of the compounds was achieved using methanol/water (25/75, v/v) containing 5 mM ammonium acetate, at 22 °C in a 100 mm × 4.6 mm, 3 μm, RP-18 Aquasil column (Thermo, Bellefonte, PA, USA). 10 μL sample volume was injected into a flow of 0.5 mL/min. The negative ion mode was selected for analyte ionization. ESI parameters were as follows: source temperature 400 °C, curtain gas 20 psi (138 kPa), nebulizer gas 30 psi (207 kPa), auxiliary gas 75 psi (517 kPa), ion spray voltage −4200 V, CAD gas 6 (arbitrary units), MRM dwell time 50 ms, pause between mass ranges 5 ms. The MRM transition of m/z 517.1 to m/z 59.1 (DP −32 V, CE −81 eV) was chosen for D3G, while m/z 355.1 to m/z 59.1 (DP −16 V, CE −30 eV) was chosen for DON. Qualifier transitions were taken from the original LC–MS/MS method (Berthiller et al., 2005).

Results

In order to determine the fate of D3G upon ingestion by mammals, in vitro experiments mimicking the digestion conditions in the gastrointestinal tract were performed. Control experiments proved the stability of the precursor mycotoxin, DON, at all investigated conditions. Furthermore, the sum of the molar amount of DON and D3G remained roughly constant (within 10%) in all experiments, indicating no losses of toxins during the experiments.

Acidic hydrolysis of D3G

Acidic solutions were used to assess the impact of the conditions found in the stomach of mammals on D3G stability. D3G proved to be completely stable towards acid hydrolysis with 0.02 M HCl, at a pH-value of about 1.7, which is at the lower end of the stomach pH range in humans. Even at a 10 times higher concentration of HCl, at a pH-value of about 0.7, no DON could be detected after incubation of D3G at 37 °C for 3 h or 18 h. Artificial stomach juice, containing pepsin at pH 1.7, also had no effect on D3G. The results of the hydrolysis studies under acidic and enzymatic conditions (see below) are summarized in Table 1. In all acid-treated samples 100 ± 2% of D3G were recovered.

Table 1

Extent of hydrolysis of D3G after acidic and enzymatic treatments at 37 °C.

	3 h	18 h
Water	<0.2%	<0.2%
0.02 M HCl	<0.2%	<0.2%
0.2 M HCl	<0.2%	<0.2%
Artificial stomach juice	<0.2%	<0.2%
Artificial gut juice	<0.2%	<0.2%
β-Glucosidase (almond)	<0.2%	<0.2%
β-Glucuronidase (snail)	1%	1%
Cellulase (Trichoderma)	11%	15%
Cellobiase (Aspergillus)	60%	73%
β-Glucosidase (human)	<0.2%	<0.2%

Enzymatic hydrolysis of D3G

A variety of glycosylhydrolases was used to test the enzymatic stability of D3G. Artificial (non-microbial) gut juice, containing amylase, showed no activity at all towards the β-glucoside D3G. Similarly, while testing 1 U/mL of almond β-glucosidase, no activity (<0.01 mg DON/L) was noticed towards D3G. This is in agreement with results obtained previously for D3G (Sewald et al., 1992) while Z-14-G was completely converted to ZEN (although at higher enzyme concentrations) by this enzyme (Gareis et al., 1990). More importantly, also human cytosolic β-glucosidase (hCBG, expressed in Pichia pastoris) did not show any activity for D3G. β-Glucuronidase, commercially purified from snail gut, can cleave β-glucuronides, but also possesses high β-glucosidase and arylsulfatase side activities. At the given conditions using 10 U/mL of β-glucuronidase, just a minimal DON peak (about 1%) was obtained after treating D3G. Finally, experiments were carried out using cellulase, which cleaves β-glucosidic bonds of cellulose. Partial cleavage of D3G to DON (about 11% after 3 h, about 15% after 18 h) was observed after cellulase treatment. We suspect this activity is due to co-occurrence of β-glucosidases (cellobiase) from Trichoderma, rather than to a side activity of an endo- or exo-cellulase. This is in good agreement with the results gained using Aspergillus cellobiase, which yielded the highest conversion of about 60% after 3 h and 73% after 18 h.

Hydrolysis of D3G by bacteria

Forty-seven different bacterial strains, isolated from guts, were examined towards their ability to hydrolyze D3G. B. bifidum, B. longum, C. freundii, E. avium, E. coli, L. amylovorus, L. crispatus, L. fermentum, L. gasseri, L. paracasei and L. rhamnosus showed no activity. E. casseliflavus, E. faecalis, and E. gallinarum liberated minor amounts of DON (1–8% after 8 h) from D3G. However, E. cloacae, E. durans, E. faecium, E. mundtii but also L. plantarum and B. adolescentis efficiently cleaved D3G, releasing up to 62% DON after 8 h (Table 2).

Table 2

Extent of hydrolysis of D3G after bacterial treatment.

Bacterial strain	Number of strains	4 h	8 h
Bifidobacterium adolescentis	2	5–17%	17–25%
Bifidobacterium bifidum	3	<0.2%	<0.2%
Bifidobacterium longum	3	<0.2%	<0.2%
Citrobacter freundii	1	<0.2%	<0.2%
Enterobacter cloacae	1	17%	25%
Enterococcus avium	1	<0.2%	<0.2%
Enterococcus casseliflavus	1	5%	8%
Enterococcus durans	2	8–16%	14–27%
Enterococcus faecalis	2	<0.2–3%	<0.2–3%
Enterococcus faecium	11	4–39%	6–55%
Enterococcus gallinarum	2	<0.2–2%	1–3%
Enterococcus mundtii	1	18%	38%
Escherichia coli	2	<0.2%	<0.2%
Lactobacillus amylovorus	1	<0.2%	<0.2%
Lactobacillus crispatus	2	<0.2%	<0.2%
Lactobacillus fermentum	2	<0.2%	<0.2%
Lactobacillus gasseri	3	<0.2%	<0.2%
Lactobacillus paracasei	3	<0.2%	<0.2%
Lactobacillus plantarum	1	34%	62%
Lactobacillus rhamnosus	3	<0.2%	<0.2%

Discussion

Up to now, data regarding the toxicological relevance of D3G were lacking. Our results indicate that D3G is resistant to acidic conditions. It is, therefore, extremely unlikely that D3G can be hydrolyzed into DON in the stomach of mammals. Pretty much the same results were gained using digestive enzymes in vitro, suggesting that D3G will most likely pass unchanged also through the small intestine. For instance amylase, which is produced in the salivary glands and the pancreas and able to cleave the α-glucosidic bonds of starch, showed no potential to hydrolyze D3G. β-Glucosidase is expressed in human liver, kidney, spleen and gut (Berrin et al., 2002) and plays an important role in the hydrolysis of plant glucosides like flavones, isoflavones, flavanones, flavonoles or cyanogenic glucosides like amygdalin. However, there are several naturally occurring glucosides that cannot be cleaved by hCBG, including D3G. The position of the glucose in the molecule is of importance as, e.g. quercetin-7-glucoside can be cleaved by hCBG in contrast to quercetin-3-glucoside (Berrin et al., 2002). β-Glucuronidase can be found in human plasma and at high levels also in the placenta. The available snail β-glucuronidase showed virtually no hydrolytic activity towards D3G. Therefore, the snail gut β-glucuronidase enzyme mixture, which is frequently used to liberate DON from DON-glucuronic acid conjugates in urine and other tissue samples is unsuitable for hydrolysis of D3G in grain samples for analytical purposes. While enzymes encoded by the human genome seem to be of no relevance, microbial inhabitants of the intestines are providing a rich source of hydrolytic enzymes. Cellulase is produced by a variety of microorganisms found in the gut of ruminants. Since the fungal cellulase and in particular cellobiase preparations used in this study could hydrolyze D3G, it might be speculated that D3G is cleaved and DON is released by plant-based cellulose-foraging ruminants.

Our results clearly show that under the in vitro conditions used in this study, D3G was converted to DON upon incubation with several pure cultures of intestinal bacteria, in particular species of the genera Lactobacillus, Enterococcus, Enterobacter and Bifidobacterium. Only partial hydrolysis was obtained under the semi-aerobic conditions used in this work whereas anaerobic conditions prevail in the mammalian gut. The D3G concentration (corresponding to 2.5 mg/L) used in incubations with bacteria is unrealistically high for food, but not for feed samples, where guideline levels for DON are as high as 12 mg/kg. The bacterial density in the gut is significantly higher than in our in vitro tests; however complex mixtures and matrix influences are occurring. The density of bacteria in faeces is about 1012 cfu/g, while the densities of pure cultures used in our study correspond to about 109 cfu/mL. This suggests that even species that contribute only few percent of the microbiota may release a significant portion of DON from D3G in the lower gastrointestinal tract. Glucoside hydrolases/β-glucosidases are overrepresented in gut metagenome studies (Gill et al., 2006), thus enzymes with specificity for D3G are expected to be abundant. A highly relevant factor seems to be the species composition of the intestinal microbiota. Due to microbial diversity and density, different cleavage rates can be expected in different animals or humans (Abbott, 2004). Metagenome studies (Hattori and Taylor, 2009) indicate that there are also clear trends towards a different composition between adults and infants. For instance, Bifidobacterium and Lactobacillus species are more abundant in infants.

Taken together this in vitro study suggests that DON detoxified by the plant into D3G may become partly bioavailable due to D3G hydrolysis by bacterial β-glucosidases in the colon. Yet, it seems impossible to predict to which extent hydrolysis occurs in a given person. Beside an individual microbiota, D3G hydrolysis may be also highly dependent on other factors, such as the kind of fermented milk products or abundant probiotic bacteria consumed together with D3G contaminated cereal products. If, as our data suggest, most of the present D3G is hydrolyzed to the parental toxin, D3G is of toxicological relevance and should be monitored together with DON in cereals, especially since the portion of the masked toxin might increase in the future due to Fusarium resistance breeding efforts.

Conflict of interest

The authors declare to have no conflict of interests.