Strain-dependent detoxifying ability is very often the case here, sometimes to a perplexing degree. It is shown by Lb. bulgaricus 259/2 and 171/2: 94 and 28.5% reduced levels of OTA (50 ppb) after 48h incubation in MRS medium. Two strains of Lb. acidophilus (1A and 4A) were also compared but showed a difference of only 13.8% (46.5 vs 32.7) in their ability to reduce OTA levels ^[26]. Another notable illustration of strain specificity concerns four strains Lb. delbrueckii subsp. bulgaricus, J7, P7, 171₂ and SL, which showed great variability in their binding and nothing very impressive even at their best (6 to 34%) ^[27].

If different strains do show similar detoxifying rates – for example, Lb. helveticus 2A (71.9%) and Lb. helveticus 8 (67.1%) ^[26][27] – this must be regarded as sheer coincidence. Note, also, that these studies used very different OTA concentrations, 50 vs 1000 ppb. While the toxin concentration is critical to make sense of the percentage detoxification, the latter is more prominently, and misleadingly, displayed in most studies. The importance of the medium also tends to be neglected. It can be considerable, as demonstrated by Lb. kefiri KFLM3 which achieved 81% decreased OTA (1 mg/mL) in milk, but only 15% in MRS – a startling difference ^[22].

It appears that most LAB strains do not exhibit remarkable mycotoxin detoxifying abilities. Several relatively large-scale screenings persistently point out to that conclusion.

Of 29 LAB strains, Lactobacillus and Lactococcus spp., tested for their ability to remove OTA (1000 ppb) from media, only four were able to reach 60% or more: Lb. rhamnosus GG (87.5%), Lb. acidophilus CH-5 (70.5%), Lb. helveticus 8 (67.1%) and Lc. lactis 202 (59.6%). The process was relatively rapid, reaching maximum values within 15h, but partially reversible: within the next 25h almost 10% of the bound toxin returned to the medium. Based on calculations of the toxin amount in biomass at several different time points, the authors speculated about an alternative mechanism, in addition to binding, of OTA removal, but went no further than this ^[27].

Of 30 LAB strains, Lactobacillus and Bifidobacterium spp., only four proved able to reduce more than 50% OTA (0.5/1.0 mg/mL) for 4h. Only one, Lb. acidophilus VM 20, was truly remarkable: 96–97% decrease with both concentrations. However, as is usually the case, the process was heavily dependent on cell density, pH and toxin concentration. Raising the pH from 5 to 7 or lowering the cell count from 1 x 10⁹ to 1 x 10⁸ nearly halved the efficiency of Lb. acidophilus VM 20 as OTA reducer. The effect proved to have rather a high threshold of saturation. Only at 50 mg/mL OTA did the level of reduction begin to fall, and then only with 10–15% ^[28]. Similar or higher amount of OTA per kilogram is seldom found in contaminated foods; it has been reported only in green coffee beans, raisins and spices ^[6].

One of the few studies so far to concentrate on degradation of OTA found that four Pediococcus parvulus strains isolated from Douro wines achieved 89–98% degradation of OTA (1 mg/mL) in MRS medium (5 mL) after 5 days of incubation at 30°C with 10³ CFU/mL. One strain, P. parvulus UTAD 473, even reached 100% degradation of OTA under these conditions; sixteen other LAB strains (mostly Lb. plantarum and Oenococcus oeni) managed only 10-20%. The rate of the process was dependent on the inoculum size (almost five times faster with 10⁹ CFU/mL) and the incubation temperature (~30% slower at 37°C). The presence of ochratoxin a was confirmed by LC-MS/MS and a putative peptidase was proposed, but the authors unfortunately did not go further in that direction. Fascinatingly, however, they conducted adsorption studies with dead P. parvulus and concluded that only 1.3% of the OTA was actually bound to the cell wall ^[29].

Another study of OTA degradation compared 27 commercial LAB strains cultivated for 24h at 37°C in 10 mL MRS broth contaminated with 0.6 mg/mL OTA. The authors concluded that among the six strains that showed 97-99% total reduction of OTA at pH 6.5, hydrolysis was by far the predominant mechanism; only 2–4% were due to adsorption. Curiously, the hydrolysis was less effective in a more acidic medium (pH 3.5). Degradation products ochratoxin a and phenylalanine were confirmed by mass spectrometry ^[30]. However, no protease was identified or even proposed, much less isolated and characterised.

Interestingly, a study of OTA reduction by Lb. bulgaricus, also tested the ability of these Lactobacilli to neutralize several different trichothecenes such as nivalenol (1 ppm), deoxynivalenol (1 ppm), diacetoxyscirpenol (500 ppb) and T2 toxin (500 ppb), but no effect was observed ^[26]. The same lack of correlation between detoxifying capacities was demonstrated for OTA and patulin by Lactobacillus and Bifidobacterium ^[28].

On the whole, the efficiency of LAB as OTA scavengers is considerable and reinforces their role as probiotics with anti-mycotoxin action. However, as in the case of AFB₁ detoxification, strains must be selected with great care as regards their capacity to neutralise OTA and their optimal conditions.  

2.3. LAB against Patulin

In a pioneering study of patulin removal by LAB in liquid medium, Fuchs and colleagues showed that Bb. animalis VM 12 was capable of 80% reduction of patulin (0.5/0.1 mg/mL) for 4h in MRS. As in the case of OTA mentioned above, the process was rather less effective with lower cell densities and in a medium with higher pH ^[28]. In recent years, perhaps because of its easy availability on mouldy fruits, patulin has attracted some notable attention in the field of LAB detoxification.

An intriguing study with heat-inactivated LAB used methods like Fourier Transform Infrared Spectroscopy (FTIR), Zeta Potential and Contact Angle to confirm the importance of physical and chemical parameters like specific surface area, cell wall volume and N/C ratio for the binding capacity of patulin. Since C–O, OH and NH were the main functional groups involved, probably polysaccharides and/or proteins are the crucial binding molecules. Nothing more specific was established about the patulin-binding mechanism of LAB. Lb. brevis 20023 was found to have the highest specific surface area, greatest cell wall volume and, expectedly, highest capacity (65.02%) to adsorb patulin (4 mg/L) from aqueous solution ^[31].

Central composite design was used to optimise Lb. plantarum ATCC 8014 achieved 96% patulin removal from apple juice during 6 weeks of cold storage. The juice was artificially contaminated with 100 mg/L patulin. However, very high cell density was required (3.6 x 10¹¹ CFU/mL) as well as the addition of prebiotic fructooligosaccharide (2.3%), ascorbic acid (213 mg/L) and citric acid (1.4 g/L). SDS-PAGE was used to confirm that S-layer proteins were involved in the adsorption of patulin. The electrophoresis showed a sharp decline in the amount of a 50-kDa fraction on the first day of incubation, which is in agreement with the kinetics of patulin decrease: almost 70% on the first day, then much slower but steady decrease until the 42^nd day ^[32].

A recent study used LAB from Tibetan kefir grains for detoxification of apple juice and went into some detail about the adsorption mechanism. FTIR was used to establish the most important functional groups, and while the result (C–O, OH, C–H, N–O) was somewhat different than the study mentioned above ^[31], the authors reached the same, admittedly rather general, conclusion: proteins and polysaccharides on the cell surface must be responsible for the patulin adsorption. Of the five strains tested, Lb. kefiranofaciens JKSP109 was the finest patulin scavenger – 93% at 100 mg/L, but only 56%  mg/L – although two other strains did almost as well. The adsorption capacity was found to depend on pH and the °Brix, the higher the better in both cases ^[33].

DON has been a somewhat unpopular research subject in the last few decades, which is surprising considering its prevalence in cereal crops. According to some studies, published and unpublished, 65% of the maize kernels harvested in France during 2004–2006 were contaminated with DON and fumonisins; another study of corn samples from several European countries found that 52 of 67 contaminated samples (78%) contained DON, and while only two of them exceeded the EU recommended values (8 mg/kg in grain and grain products), six others exceeded 1 mg/kg; concentrations from 100 to 1000 mg/kg appeared to be quite common in Europe ^[10][40]. A couple of recent studies have dealt with LAB as DON detractors in a somewhat illuminating way.

Altogether 16 LAB strains, eight commercially available (e.g. Lyofast LPRA, Yo-flex YC-180) and eight isolated from cereals and kefir (mostly Lb. plantarum), were tested for anti-fungal activity and DON reduction. Only six of them significantly inhibited the growth (agar halos bigger than 30 mm in diameter) of Fusarium graminearum JAPAR 2218, a confirmed DON producer and an economic scourge for grain crops worldwide. DON reduction studies were conducted with 1.5 mg/mL toxin in MRS for 4h and volume of 2 mL with average cell densities in the order of 10¹⁰ CFU/mL and three types of bacteria, viable and heat-inactivated by pasteurisation (100°C, 30 min) or sterilization (121°C, 15 min). In all cases, the sterilized cells showed the best ability to reduce DON, usually 20-30% higher than that of the viable cells. There was some correlation between antifungal activity and toxin reduction, but not much. The best of the isolated strains, Lb. plantarum GT III (67% DON reduction), was not among the most potent fungicides (27 mm halo) ^[34].

Lb. paracasei LHZ-1 isolated from yogurt achieved 40.7% reduction of DON (50 mg/mL) by the cell wall fraction in PBS for 24h at 37°C. In contrast, only 10.5 and 8.9% were reduced by culture supernatant or cellular lysate, respectively. Laser scanning confocal microscopy was used to further elucidate the mechanism of DON detoxification. DON was labelled with AMCA-X SE to produce blue fluorescence and thus was obtained visual evidence that DON does form complexes with the bacterial cell wall. As in the study just mentioned (Franco, 2011), pasteurised and sterilised cells removed DON more efficiently than viable cells, only in this case the increase was only 5-6% at most. In addition to heat inactivation, the influence of several chemical agents on the DON removal rate was also tested with the same amount of cells (1 x 10¹⁰ CFU/mL): 10% trichloroacetic acid (15 min) and 4M HCl (2h) increased it with 2-3% compared to viable cells, while 8M urea (6h) and 20 mg/mL lysozyme (18h) decreased it with about 20–25% ^[41].

After DON, fumonisins are the next most prominent contaminants of food and feed ^[40]. Not much has been done so far to combat their toxic effects with LAB, but something has nevertheless been achieved, especially in regard to the elusive mechanism of toxin-binding to the cell wall.

One study found that LAB starter culture (Str. lactis, Lb. delbrueckii) added to a maize meal could reduce the levels of fumonisin B₁ (2 mg/g meal) with almost 75% for 4 days. This fermented meal was comparatively less toxic to SNO human esophageal carcinoma cell line, but the difference was not significant. The authors perceptively note that the reduction of the toxin level may not necessarily result in reduced toxicity because the LAB fermentation does not alter the bioavailability of the toxin. Chronic complications from trace amounts of the toxin remain a potential problem ^[36].

Another study provided some insight into the exact components of the LAB cell wall that bind fumonisin B₁ and B₂ (FB₁, FB₂). The importance of peptidoglycan (PG) was confirmed in two different ways. Mutants with defective PG layer were found to have decreased toxin binding, which affected only FB₂, and only with 20-25%. Purified PG bound fumonisins (5 mg/mL each) in similar, but somewhat lower, degree compared to LAB (20% for FB₁, 60% for FB₂, both at 2 mg/mL PG). Mutants with defective synthesis of lipoteichoic acids showed negligible difference (5-10%) compared to the wild type, indicating that this component of the cell wall is unimportant as far as fumonisin binding is concerned. The tricarballylic acid chains of the fumonisins were confirmed to be essential for the toxin-binding, which decreased when the chains were hydrolysed. The authors also claimed that treatment with lipases and proteases had no effect on the toxin binding, and neither did the use of mutants lacking exopolysaccharides, although in both cases they did not show the data ^[37].

ZEA has been reported in foods and body fluids (animal as well as human) with frequency that is nothing short of alarming ^[7]. As in the cases of AFB₁ and OTA, a great deal of work has been done on LAB detoxification of ZEA, but the accumulated data are often incomplete or inconsistent, if not downright contradictory, the molecular mechanisms remain elusive, and the application of LAB as ZEA-detoxifying probiotics is problematic, at best.

One promising probiotic of the future against ZEA is Lb. plantarum A1. This strain was shown to have potent and rapid ability to bind ZEA (20 mg/mL). The process was partially reversible, dropping from immediate 99% to only 77% after 72h of cultivation in MRS broth, but the relatively small inoculum (10⁸ CFU/mL) is a point in the strain’s favor ^[38]. Similar kinetics were obtained with Lc. Lactis isolated from milk products and 130 mg/mL ZEA, although in this case the process appeared to be virtually irreversible ^[39]. LAB starter culture (Str. lactis, Lb. delbrueckii) added to a maize meal reduced the levels of ZEA (2 mg/g meal) with 68% for 4 days; as in the case of FB₁, the decreased toxicity on the SNO cell line was not significant ^[36]. Lb. kefiri KFLM3 achieved 100% decrease of ZEA (1 mg/mL) in milk, but only 60% in MRS, yet another reminder of the importance of the medium ^[22]. L. plantarum 3QB361, isolated from Brazilian cheese and inactivated in phosphate buffer, managed to reduce ZEA (2 mg/mL) with 70-80% at pH 6.5, but five other strains (from ten tested) hardly managed 20-40% – another timely reminder, this time of species- and strain-specificity ^[24].