The analytical procedures were described in the previous studies of Hanuš et al. (2007) or Janů et al. (2007). Briefly, composition and physical–chemical and technological properties of milk were carried out according to the following methods:

Fat, lactose and solids-not-fat were measured using MilkoScan 133B (Foss Electric, Denmark).

The following milk indicators were calculated: whey protein, non-protein nitrogen (NPN), fat to crude protein ratio (fat / CP), urea N in NPN ratio, casein number on crude protein basis (Cas-CP), and casein number on true protein basis (Cas-TP).

The feeding rations consisted of roughage (see Table 1) and supplements of concentrate mixture according to milk yield and standard demands. Samples of feedstuffs were collected at the same time as the IMSs. Commercially available quantitative enzyme-linked immunosorbent assay (ELISA) kits (Veratox, Neogen Corp., Lansing, MI, USA) were used for measuring the presence of the mycotoxins deoxynivalenol (DON), fumonisins (FUM), zearalenone (ZEA), aflatoxin (AFL), and T-2 toxin according to the manufacturer's instructions. Based on the obtained results for each analysed feedstuff, a class of mycotoxin load was calculated as a mean from classes assigned to individual mycotoxins (see Table 2). Four classes were created under the correlation of mycotoxin amount with recommendation limits. Class 1 was represented by samples with no mycotoxins, in class 2 there were samples under the concentration limit, and in class 4 there were samples with more than 3 times the concentration limit. Finally, the load of herds was classified as negligible (Load 1: mean class 1), low (Load 2: mean class 2), and medium (Load 3: mean class 3).

Basic statistical characteristics were calculated: arithmetical mean (x), geometrical mean (xg), standard deviation (SD), and t values for significance assessing of the differences (MS Excel – Microsoft, Redmond, Washington, USA). Some values (acetone and somatic cell count) were log-transformed (Hanuš et al., 2007; Janů et al., 2007) as the data were not normally distributed and also for possibility to use the geometrical means.

The GLM procedure of the SAS v. 9 program package was used for the calculation. Multifactor analysis of variance with fixed effects as breed, season, and feeding system was used for statistical evaluation of dataset, according to the following model: yi-q=μ+hi+yj+sk+ll+bm+dn+mo+cp+fq+ei-q, where yi-q is an independent variable (investigated milk indicator), μ is the general mean, hi is the effect of herd (i=8), yj is the effect of year (j=3), sk is the effect of season (k=2), ll is the effect of the number of lactations (l=10), bm is the effect of breed (m=2), dn is the effect of the stage of lactation (n=3, < 100, 100–200, > 200 DIM), mo is the effect of mycotoxin load (o=3), cp is the effect of feeding class (p=9, based on energy: CP ratio), fq is the effect of the type of feeding (q=4, based on the proportion and type of roughage in the diet), and ei-q is the random effect.

In the present experiment cows were fed mainly on various types of silage (maize, lucerne, clover/grass, cereals) as documented in Table 1. Although all samples were positive for at least one mycotoxin, the overall level of mycotoxin contamination of feedstuffs of studied herds was relatively low. The most frequently occurring mycotoxins were FUM, DON, and ZEA, with 82, 73, and 68 % of positive samples, respectively (see Table 3). Relatively high incidence of AFL (56 % positive samples) was also registered in our study. Co-occurrence of two mycotoxins was observed in 98 % of samples; 80 % of samples contained three or more mycotoxins (data not shown). Only 36 IMSs were included in Load 1, with negligible mycotoxin load, while the majority of IMSs belonged to the low (Load 2, 192 IMSs) and medium (Load 3, 708 IMSs) levels of mycotoxin contamination, suggesting that not only occurrence of mycotoxins in silage but also a risk of subclinical exposure of dairy cows to mycotoxins may be much more frequent than expected.

Results of our 3-year study on a total of 936 IMSs that were analysed for a wide scale of chemical, physical, and technological characteristics depending on the mycotoxin load of herds are given in Table 4. The concentration of milk fat differed significantly among groups (P<0.001), while content of lactose, crude, and true protein and casein was not affected by the mycotoxin load (P> 0.05). There was a tendency found in solids-non-fat (P=0.0877).

The SCC tended to differ among groups with the numerically highest value in Load 3 (237 th mL-1) and the lowest value in Load 1 (104 th mL-1, P=0.0918). However, log-transformed SCC did not differ significantly (P> 0.05). Content of urea was the lowest in Load 1 and the highest in Load 2. Although the numerical differences among groups were relatively high, urea was not significantly influenced by the mycotoxin load of herds; however, a strong tendency was observed (P=0.056). Concentrations of Ac, as well as log Ac, were significantly different among groups (P<0.01 and P<0.001, respectively), with highest values found in Load 1, where the lowest milk yield was also observed. Fat / CP differed significantly between Load 1 and Loads 2 and 3 (P<0.001); however, the difference was quite low.

Electric conductivity differed among groups (P<0.001), with the highest value in Load 1 and the lowest in Load 2. The active acidity (pH) was the lowest in Load 1 (6.708) and the highest in Load 3 (6.731, P<0.01).

Titration acidity was not influenced by the mycotoxin load of herds (P> 0.05). Alcohol stability as an indirect marker of milk protein thermostability differed significantly (P<0.001). Although rennet coagulation time in our study was not influenced by the mycotoxin load (P> 0.05), the class of curd quality differed among groups (P<0.001), with the lowest value in Load 2 (2.26), while values in Load 1 and 3 were almost identical (2.69). Cheese curd firmness decreased from 17.61 mm in Load 1 to 16.36 mm in Load 3 (P<0.001). Content of crude protein, casein, and true protein were not affected by the mycotoxin load, while content of whey protein, non-protein nitrogen, and whey volume differed significantly among groups (P<0.05). Similarly, there were differences in the casein numbers (on the basis of CP and TP) among groups depending on the mycotoxin load of herds (P< 0.001).

As well documented in the available literature, animal feeding trials are commonly performed under well-defined conditions with one (DON – e.g. Erickson et al., 2011; ZEA – e.g. Zinedine et al., 2007; AFL – e.g. Xiong et al., 2015; T-2 toxin – e.g. Yoshizawa et al., 1981) or two mycotoxins (DON + ZEA – e.g. Winkler et al., 2014; Korosteleva et al., 2007). However, experimental conditions differ from farm conditions where animals may be exposed to a mycotoxin “cocktail” as documented in recent studies describing natural multi-mycotoxin contamination of feedstuffs (Hofstetter and Rodrigues, 2009; Kiyothong et al., 2012; Zachariášová et al., 2014). Studies focused on the impact of multi-mycotoxin load on animal performance are scarce and they are usually limited in duration and focused on the counteracting effect of various types of mycotoxin sequestering agents (MSAs).

In general, values of milk indicators determined in our study are in accordance with those characterising ordinary population of dairy cows bred in the Czech Republic environment (e.g. Hanuš et al., 2007; Janů et al., 2007) without differences apparent at first glance that could have been caused by the exposure of cows to various levels of dietary mycotoxins.

Our findings concerning basic milk indicators are in accordance with those reported in the literature. Charmley et al. (1993) described quadratic response in milk fat depending on increasing levels of DON in the diet, with no effect of DON on other milk components. According to Korosteleva et al. (2007) milk composition was not affected by a diet containing DON. Similar findings were reported for DON + ZEA (Winkler et al., 2014) or AFL (Xiong et al., 2015). Kiyothong et al. (2012) did not find significant differences in fat, lactose, and solids-non-fat after supplementation of MSAs to multi-mycotoxin (DON, FB1, ZEA, OTA, AFL, and T-2 toxin) contaminated diet, but they reported a positive influence of MSAs on milk protein concentration (P< 0.05) and also noted that milk fat percentage tended to be higher for MSA-supplemented than for non-supplemented cows. Similarly, Hofstetter and Rodrigues (2009) found no differences in milk fat, lactose, solids-not-fat, and total solids after multi-mycotoxin exposure with similar tendency in milk fat percentage.

SCC is a well-known indicator of mammary gland health in lactating mammals and is also used as a marker of milk quality. Increased SCC is generally accepted to be the most significant indication of the onset of subclinical mastitis in dairy cows. The immune system is a target of several important mycotoxins such as T-2, DON, and AFL that may affect the metabolic variables and immunity in dairy cows (Bondy and Pestka, 2000). The SCC in our study tended to differ among groups with the numerically highest value in Load 3 and the lowest value in Load 1 (P=0.0918). This result could indicate the influence of higher mycotoxin load to excite the immune system of cows in the form of increased SCC as a defense against mastitis. Although there is no comparable study evaluating the effect of load of mycotoxin “cocktail“ on SCC, changes in SCC after mycotoxins exposure, either significant or numerical, are often reported in the literature. For example, Applebaum et al. (1982) described a slight increase in numerical values of SCC in cows receiving 13 mg of pure AFL. Takagi et al. (2011) observed a relationship between urinary ZEA concentrations as a marker of ZEA exposure and milk SCC, where a decrease in urinary ZEA concentrations accompanied changes in milk SCC; however, in their study, no intake of ZEA was measured. On the other hand, in a subsequent study, Takagi et al. (2013) did not observe significant correlations between an individual cow's urinary ZEA concentrations and SCC. Similarly, SCC was not affected by a diet containing DON (Korosteleva et al., 2007) or AFL (Xiong et al., 2015). There are some studies focused on the effect of various types of MSAs on SCC assuming that MSA supplementation can ameliorate the adverse effects of mycotoxins on animal immunity system. However, results are not consistent. Recently, Hofstetter and Rodrigues (2009) or Kiyothong et al. (2012) found significantly lower SCC (P< 0.05) in cows supplemented with MSAs compared with non-supplemented cows and SCC decreased linearly with the increasing level of supplementation. Similarly, Takagi et al. (2011) found significant changes in SCC after MSA supplementation. On the other hand, Korosteleva et al. (2007) did not find a positive effect of MSAs supplemented to a diet containing 3.6 mg kg-1 of DON on SCC.

Milk urea can be used as an indicator of the nutritional status of the dairy herd mostly with negative relation to reproductive performance and longevity. In our study, a strong tendency for differences in milk urea was observed (P=0.056). There is no comparable study except for that Kiyothong et al. (2012), who did not find differences in milk urea in cows fed diets containing DON, FB1, ZEA, OTA, AFL, and T-2 with or without MSA supplement. On the other hand, lower urea values were reported with lower milk yield between breeds in many studies (e.g. Hanuš et al., 2007; Janů et al., 2007). A similar relationship was also observed in our study. Thus, it is likely that differences among groups in milk urea were caused rather by differences in milk yield than by mycotoxin load itself.

Milk acetone (Ac) is an indicator of cow energy metabolism. Significant differences in Ac concentration (P<0.01) determined in our study are in accordance with Janů et al. (2007) but in disagreement with Hanuš et al. (2007). However, the opposite results (for the two breeds, Holstein and Czech Fleckvieh: Janů et al., 2007, Hanuš et al., 2007, respectively) could be caused by breed (specific physiology of lactation) in relation to the milk yield with low incidence of subclinical ketosis in the both above-mentioned studies.

The fat / CP ratio is considered a good indicator of energy metabolism of dairy cows and also an indicator of cheese-making properties of milk. In our study, fat / CP differed significantly between Load 1 and Loads 2 and 3 (P<0.001). According to Hanuš et al. (2004), the fat / CP ratio is positively well correlated with milk Ac. Similar findings were also observed in our study.

Electric conductivity as a possible indicator of udder health differed significantly among groups (P<0.001). According to Hanuš et al. (2007) and Janů et al. (2007), lower values of milk electric conductivity are observed in milk with lower SCC. This was not confirmed in our study. As proved by Nájera et al. (2003) changes in milk pH affect the rennet coagulation time. Although significant (P<0.01), differences in pH in our study were too small to influence the rennet coagulation time.

There are no comparable data in the available literature concerning the effect of mycotoxin load on cheese-making properties of milk such as rennet coagulation time, curd quality, curd firmness, and volume of whey separated out during the process of enzymatic coagulation of milk. Furthermore, the technological quality of milk is influenced not only by content of basic components (fat, protein, lactose, and solids-non-fat) but also by parameters such as active acidity (pH) or titration acidity (Janů et al., 2007; Hanuš et al., 2007). Changes in pH were discussed above and titration acidity was not influenced by the mycotoxin load of herds (P> 0.05). The higher value of alcohol stability that was observed in Load 2 and Load 3 (P<0.001) was accompanied by higher milk yields and a lower value of curd firmness. These findings are in agreement with Hanuš et al. (2007).

Rennet coagulation time significantly influences the process of cheese making as well as the yield and quality of formed cheese, and it is an important parameter used in the evaluation of milk suitability for processing and cheese making (e.g. Cassandro et al., 2008). Results concerning the cheese-making indicators such as rennet coagulation time, cheese curd firmness, class of curd quality, whey volume, or casein numbers did not show a clear tendency explaining the possible effect of mycotoxins on cheese-making properties of milk, partly because of relatively low exposure to mycotoxins and partly because of differences in factors that are involved in the cheese-making ability of milk such as titratable acidity, SCC, protein, and casein contents and calcium and phosphorus concentrations (e.g. Jõudu et al., 2008). Further studies are needed to clarify a possible role of dietary mycotoxins on physical and technological properties of milk.

In conclusion, results of our study concerning the mycotoxin contamination of silage confirmed current knowledge on their occurrence and co-occurrence in these kinds of feedstuffs. Furthermore, the presence of aflatoxins in silage in the central European conditions should not be ignored.

The content of basic milk components except for fat was not influenced by the mycotoxin load of herds. From the physical and technological characteristics, acetone, pH, electric conductivity, alcohol stability, curds quality, curd firmness, whey volume, whey protein, non-protein nitrogen (NPN), urea N in NPN, fat / crude protein ratio, and casein numbers on crude and true protein basis were significantly affected by the mycotoxin load of herds. However, these results did not show a clear tendency explaining the possible effect of mycotoxins on cheese-making properties of milk, partly because of relatively low exposure to mycotoxins and partly because of the number of interference interactions among milk indicators.