Fatty acids in rumen and meat of lambs fed with different condensed and hydrolysable tannin extracts


 Five groups of lambs (n = 9 each) were used to test the effect of plant extracts rich in hydrolysable (HT) or condensed tannin (CT) on animal performance, fatty acid composition of rumen content and meat. The control group (CO) received a concentrate-based diet without tannins supplementation. The other groups received the same diet as the control lambs plus 4% chestnut (CH) and tara (TA) extracts as a source of HT and mimosa (MI) and gambier (GA) extracts as a source of CT. Dietary CH negatively affected animal performance. The rumen content of the different groups showed comparable levels of 18:3 c9c12c15, 18:2 c9c12, 18:2 c9t11, 18:1 t11 and 18:0, whereas 18:1 t10 was greater in CO. Also, 18:1 t10 tended to be lower in the rumen of HT than CT-fed lambs. These data were partially confirmed in meat, where CO showed a greater percentage of individual trans 18:1 fatty acids in comparison with tannins-fed groups. Our findings challenge some accepted generalizations on the use of tannins in ruminant diets as they were ineffective to favour the accumulation of dietary PUFA or healthy fatty acids of biohydrogenation origin in the rumen content and lamb meat, but suggest a general action of tannins on the whole biohydrogenation process.


Introduction
Tannins are water-soluble phenolic compounds that are produced by plants in response to external stressor factors [1]. The interest of the researchers on the inclusion of tannins in the diets for ruminants has progressively risen due to their potential to affect animal performance, rumen metabolism and products quality. This is related to the capability of tannins to create stable complexes with carbohydrates, protein and ions, thus affecting the availability of nutrients [2]. Moreover, they can interact with the bacterial metabolism resulting in antimicrobial activity that may ultimately affect the composition and the functional activity of the rumen microbial community [3,4]. In the literature, the administration of dietary tannins has often led to contradictory results, which does not allow the full exploitation of their potential bene ts. For example, earlier studies described tannins as an antinutritional factor in ruminant diets [5], while, more recently, the administration of limited doses of tannins has been reported to improve the protein utilization along the digestive tract, as well as the growth and animal health [6]. Similarly, dietary tannins could be implemented as a strategy to manipulate the fatty acid composition of ruminant products by modulating the process, known as biohydrogenation (BH), by which ruminal micro ora operates the progressive saturation of the ingested unsaturated fatty acids. However, tannin effect on BH of polyunsaturated fatty acids (PUFA) has been studied in the last decades with different conclusions. Recently, Frutos et al. [7] reviewed the effect of tannins on a broad range of fatty acids in ruminant meat and milk because of their modulatory effect on ruminal micro ora and lipid metabolism. In particular, in vitro and in vivo trials seem to suggest that tannins can interfere with several steps along the BH process and may determine the accumulation of different intermediate products. However, such effects are considerably variable.
The variable effect of tannins could be due to several factors such as the dose and type of tannins, or their association with different basal diets. Chemically, tannins can be broadly classi ed into hydrolysable (HT) and condensed (CT). Hydrolysable tannins are polymers of gallic or ellagic acid associated by the mean of an ester bond with a polyhydroxylated core, whereas CT are represented by avan-3-ol polymers [8]. In the rumen environment, tannins can be differently metabolized and the capacity to interact with feeds, bacteria and microbial molecules may depend not only on their concentration in the diet but also on their chemical characteristics [9]. Differences in the chemistry of dietary tannins could, therefore, exert different effects on ruminant performance and product quality. Also, the basal diet could represent a confounding factor because tannins activity may differ, for example, when a forage-based diet is compared with a concentrate-based diet [10], or when diets differing for the level and composition of polyunsaturated fatty acids (PUFA) are compared [11] Also, when tanniniferous plants are fed to animals, several factors affect the concentration and composition of tannins and plant factors other than tannins may also contribute to the animal's response.
In this context, the inclusion of puri ed tannin extracts in the ruminant diet could help to clary the effects of tannins. So far, in vivo studies assessing the effect of dietary tannins on ruminant meat and milk fatty acids mostly focused on one speci c extract [10,12], whereas the comparison between different extracts has been less investigated [13,14]. Additionally, despite the market availability of several plant extracts providing different kind of tannins, quebracho (Schinopsis lorentzii) and/or chestnut (Castanea sativa) were used in most of the studies to supplement ruminant diets with condensed and hydrolysable tannins, respectively. However, several other plant extracts are commercially available and are characterized by possessing speci c and different chemical groups of tannins.
In a companion paper [15], we characterised the ruminal fermentation and microbiome in lambs supplemented with four different tannin extracts rich in hydrolysable or condensed tannins and we found that in a long-term feeding trial the extracts did not compromise ruminal fermentation, but showed speci c antimicrobial activity against rumen micro ora, which may elicit differences on fatty acids metabolism. Therefore, in the present study, using the same animals of the companion paper, we evaluated and compared the effect of the dietary administration of chestnut (Castanea sativa) and tara (Cesalpinia spinosa) extracts as a source of HT, mimosa (Acacia mearnsii) and gambier (Uncaria gambir) extracts as a source of CT on the animal performance and fatty acid composition of rumen content and meat in lambs.

Animals and diets
This study was carried out indoors in the facilities of the University of Catania. The local veterinary authority certi es that in this farm the animals are reared according to non-experimental practices (N. 015CT325).
According to the Italian regulation (D. Lgs. n. 26/2014, art. 2, point 1 letter a and f) transposing the European Union Directive No.63/2010, the rules on the protection of animals used for scienti c purposes do not apply to this study and an approval of the experimental plan from the ethical committee was not required. Nevertheless, the whole trial was performed under the supervision of the local veterinary authority and animals were handled in accordance with the European Union Directive No.63/2010 on the protection of animals used for scienti c purposes. The study protocol complied to the ARRIVE guidelines (https://arriveguidelines.org/). Full details on the experimental plan are reported by Salami et al. [15]. In brief, forty-four 2-month-old male Sarda × Comisana lambs (initial body weight (BW) 19.6 ± S.D. 2.06 kg) were reared in individual pens where fresh water was always available for the duration of the trial. Lambs were divided into 5 groups (n = 9 each) that were balanced for the initial BW. The control group (CO) exclusively received a barley-based concentrate diet containing on an "as fed" basis: barley (480 g/kg), wheat bran (230 g/kg), dehydrated alfalfa hay (150 g/kg), soybean meal (100 g/kg), molasses (20 g/kg) and mineral-vitamin premix (20 g/kg). The four experimental groups received the same diet as the control lambs with the addition of 4% (as fed) of chestnut (CH) and tara (TA) extracts as a source of hydrolysable tannins and mimosa (MI) and gambier (GA) extracts as a source of condensed tannins. Chestnut, tara and mimosa extracts were purchased from Silvateam S.p.A. (San Michele M.vì, Cuneo, Italy), while gambier extract was purchased from Figli di Guido Lapi S.pA. (Castelfranco di Sotto, Pisa, Italy). All the concentrates were in the form of a pellet and the tannins extracts were added before pelleting at the temperature of 40 °C. Table 1 reports the chemical and fatty acid composition of the experimental basal diet. The experiment lasted a total of 84 days. During a 9-day-adaptation period, the animals received the same commercial starter concentrate used for weaning, with gradual increases in the proportion of the experimental concentrates. During the experimental period (75 days), the lambs were fed ad libitum with their respective diets. Daily, before morning feeding, individual intakes were measured according to the refusal left by each animal. Weekly, lambs were weighed before morning feeding and samples of each experimental diets were collected. Feed e ciency was calculated as the ratio between DMI and ADG during the 75-day-experimental period. At the end of the experimental period, all the lambs were slaughtered on the same day at a commercial abattoir according to the European Union welfare guidelines. All the animals had free access to their respective diet and water until approximately 3 hours before slaughter. Lambs were rst stunned by a captive bolt and then exsanguinated. Each carcass was immediately weighted to record the hot carcass weight. Neutral detergent bre (NDF) was determined according to Van Soest et al. [16]. Furthermore, crude protein, crude fat (ether extract) and ash were analysed according to AOAC [17]. Lipid was extracted and converted to fatty acid methyl esters (FAME) with a 1-step procedure reported by Valenti et al. [18]. Gas chromatographic analysis was carried out as later described for meat. Fatty acids in feed were expressed as g/kg of dry matter.

Rumen samples collection and analyses
Individual ruminal content was collected within 15 min from slaughter. The ruminal wall was cut with a scalpel and the whole rumen content was placed into a 4-L-plastic beaker. After homogenization with a ladle, the ruminal pH was measured using a pH meter (HI-110; Hanna Instruments, Padova, Italy). An aliquot of approximately 120 mL of rumen content was immediately placed in dry ice prior to storage at − 80 °C. Freeze-dried rumen content was trans-esteri ed as described by Natalello et al. [19].
The FAME pro le was obtained by gas-chromatography as later described for the meat samples. Fatty acids were expressed as g/100 g of total fatty acids. The biohydrogenation indexes (%) for 18:2c9c12, 18:3c9c12c15, and the biohydrogenation completeness (%) in rumen content were estimated as reported by Alves et al. [20].

Meat sampling and analysis
After slaughter, carcasses were stored 24 hours at 4 °C and halved. The entire longissimus thoracis et lumborum muscle (LTL) was removed from the right side, packed under vacuum and stored at -80 °C until analysis. Intramuscular fat was extracted in double from 10 g of nely minced LTL samples with a mixture of chloroform and methanol (2:1, v/v) as described by Folch et al. [21] and 30 mg of lipids were converted to FAME by basecatalyzed transesteri cation [22], using 0.5 mL of sodium methoxide in methanol 0.5 N and 1 mL of hexane.
Nonadecanoic acid was used as an internal standard. Gas chromatographic analysis was carried out with a GC  [24,25]. Fatty acids were expressed as g /100 g of total fatty acids.

Statistical analysis
As in the companion paper [15], one-way ANOVA was used to assess the overall effect of dietary treatments, using a stepwise selection of those variables (FA) that better discriminated between the experimental groups, following the procedure described by Valenti et al. [26]. Variables with F < 0.10 were retained at the end of the stepwise procedure and submitted to canonical discriminant analysis. This procedure allows the creation of a linear function (CAN) that includes the quantitative variables that best summarize the variance of the dataset and maximize the distance between the groups. The discriminant capacity was evaluated through the Wilks's lambda test of signi cance and the discriminating e cacy of the model was expressed as the percent correct assignment of each individual to the respective group using the leave-one-out cross-validation. In this procedure, each lamb is treated as an unknown subject and the canonical functions, previously developed with the entire dataset, are used to assign it to a group. The higher the discriminating capacity of the model, the higher the percentage of cases correctly assigned to their respective group.

Results
Animal performance Table 2 reports data on the main animal performance parameters. The supplementation of tannins in the diet of lambs did not affect animal performance (P > 0.05), except for dry matter intake (DMI; P = 0.048), which was greater in the CO group than CH + TA + GA + MI. The multiple comparisons between treatment groups showed a signi cant effect of the diet on the DMI (P < 0.001), nal body weight (BW, P = 0.009), average daily gain (ADG, P = 0.002), carcass weight (P = 0.004) and FE (P = 0.041). These results were principally due to the lower values observed in CH in comparison with the CO and the other groups. In particular, CH lambs showed a lower DMI than the CO group and the lambs fed condensed tannins, the latter explaining the signi cant effect of tannin type (HT vs CT) on this parameter (P = 0.001). The lower DMI intake in the CH group resulted in a lower nal BW, ADG and carcass weight in comparison with CO, TA and GA. However, the FE resulted lower only in CH lambs than in the other group receiving hydrolysable tannins (TA). Dietary treatment means within a row with different superscripts differ at a signi cance of P ≤ 0.05.
Rumen content and meat fatty acids Table 3  Regarding tannin type, the proportion of 17:0 and 17iso was greater in the rumen content of lambs fed CT (P = 0.024 and P = 0.033, respectively). Dietary treatment (P = 0.029) and tannins supplementation (P = 0.003) affected total monounsaturated fatty acids (MUFA). These results were mainly related to the greater proportion of 18:1t6 + t7 + t8 and 18:1t10 found in the rumen content of CO lambs in comparison to the other groups. Among the groups receiving tannins, only GA showed levels of 18:1t6 + t7 + t8 comparable to CO. As for the individual PUFA, TA lambs showed a double proportion of 18:2t9c12 in comparison with all the other groups, which caused that HT fed lambs had a greater proportion of this fatty acid in comparison to CT (P = 0.020). Similarly, 18:2c9c12 tended to be greater in HT lambs than CT (P = 0.075).   18:1 isomers, 18:1t6 + t7 + t8 and 18:1t10 were lower in the HT than CT fed lambs (P = 0.037 and P = 0.010, respectively), moreover, they were lower in CH group as compared to CO (P = 0.004 and P = 0.005, respectively) and MI (P < 0.001 and P < 0.001, respectively). Finally, 18:1t6 + t7 + t8 and 18:1t10 were found at lower proportion in the GA group in comparison with MI (P = 0.019 and P = 0.023, respectively). The 18:1t9 was lower in the GA meat as compared to CO and MI (P = 0.040 and P < 0.001, respectively), and in the CH group in comparison with MI (P = 0.003). Regarding other PUFA in meat, 18:2t8c13 and 18:2t9c12 were affected by tannin supplementation and differed between the ve dietary treatments. In particular, 18:2t8c13 was lower in the meat of CO than tanninfed groups (P = 0.002) which, conversely, had lower proportions of 18:2t9c12 (P = 0.007). The meat of HT-fed groups showed a greater proportion of 18:2t9c12 (P = 0.007) than CT. Moreover, 18:2t8c13 signi cantly differed between CO and CH (P < 0.001) while, 18:2t9c12 was higher in CO group in comparison with GA and MI (P = 0.001 and P = 0.002, respectively) as well as in the CH group as compared to GA and MI (P < 0.001 and P < 0.001, respectively). Dietary treatment means within a row with different superscripts differ at a signi cance of P ≤ 0.05.

Multivariate approach to fatty acid pro le
From the entire dataset of the rumen content, eight FAs were retained at the end of the stepwise selection (Table 5) and were linearly combined to obtain four canonical discriminant functions (CAN), which together explained the total variance of the dataset. The multivariate structure could be appropriately described by the rst two CAN, explaining 70.0% and 22.7% of the variance, respectively. Therefore, the scatterplot (Fig. 1a) represents the distribution of the lambs in the multivariate space, with CAN 1 as the X axis and the CAN 2 as the Y axis. The CAN 1 alone allowed discrimination between the CO and tannins groups. The CAN 2 tended to separate the CH lambs from all the other groups with some overlaps with the MI or GA groups. After cross-validation, 66% of cases were correctly assigned. Regarding wrongly assigned cases: only one CO animal was misclassi ed and assigned to CH; one and four CH cases were assigned to TA and GA, respectively; three TA cases were assigned to the other HT group (CH); two cases of MI were assigned to the other CT group (GA). Table 5 reports the standardized coe cients of the variables describing the total canonical structure for the rumen content. The magnitude (absolute value) of the standardized coe cients assigned to each variable is a measure of the relative contribution of the variable itself to the discriminating capacity of the whole multivariate function: the greater the value of the coe cient, the higher the contribution of the variable to the discrimination. The CAN 1 was mainly in uenced by 17anteiso, 18:1t10, 18:2t8c13; 18:2c9c12 and 22:0; while 17anteiso, 18:1t5, 18:2t9c12, 18:2t10c12 and 18:2c9c12 accounted most in the CAN 2. From the dataset of the meat fatty acids, six were retained at the end of the stepwise selection ( Table 5). The rst two CAN explained 55.6% and 27.1% of the variance, respectively. Figure 1b represents the distribution of the lambs in the multivariate space. The CAN 2 allowed discrimination between the CH and the other groups. Only 56.8% of cases were correctly assigned after cross-validation. Table 5 reports the standardized coe cients of the variables describing the total canonical structure for the meat. The CAN 1 was mainly in uenced by 18:1c14, 18:2c9t11 and 18:2t9c12; while 18:1t9 18:2t8c13 and 18:2t9c12 accounted most in the CAN 2.

Discussion
Literature reports that the inclusion of tannins in the diet for ruminants may depress voluntary feed intake, thus impairing animal performance [27,28]. Several possible mechanisms have been considered, alone or in association, to explain how tannins in uence DMI. Among these, the reduction of the feed palatability would be due to the astringency perception induced by a direct reaction of tannins with the taste receptors or through the interaction with the tannin-binding salivary proteins [6]. The reduction of DMI has been usually observed at tannin concentrations that exceed 50 g per kg of diet, while lower doses would not be unfavourable and may enhance animal health and production [10,29]. In small ruminants, the negative effects of tannins would be mitigated by adaptation mechanisms, such as the production of salivary proteins able to neutralize the astringency of tannins [30]. Also, HT and CT may result in different effects because: i) HT would be more susceptible to be hydrolysed than CT in the rumen and release monomers with toxic effects [2]; ii) the complexes that CT create with proteins and other nutrients are stable in a wider pH range than HT [31,32]. Based on this knowledge and that of previous ndings revealing speci c antimicrobial activity of different tannin extracts [15], we have evaluated the response of animal performance and fatty acids in the rumen content and meat of growing lambs fed commercial sources of HT (chestnut and tara) and CT (mimosa and gambier). Also, within HT and CT classes, we have tested whether different sources of plant extracts, each characterised by the occurrence of chemically different tannins, may elicit different responses.
Our ndings partially contrast with the literature. Indeed, though the level of tannin inclusion in the diet was lower than the threshold of 50 g per kg of diet, it reduced the DMI in comparison with the CO group. The lower DMI observed in lambs fed HT compared to those fed CT, was mainly due to the effect of the CH group, which showed the least DMI and growth performance among the treatments. Conversely, none of the investigated productive parameters differed between CO and TA, GA and MI. These ndings are particularly interesting because they suggest several considerations that challenge some accepted generalizations on the use of tannins in ruminant diets. First, the lack of different performance between TA lambs (supplemented with a source of HT) and CT-fed lambs may indicated that general rules regarding the effects of hydrolysable or condensed tannins on animal performance should be avoided or at least carefully considered. Second, the response to the administration of tannin extracts belonging to the same class of compounds (in our case HT) may differ according to the chemical characteristics of tannins or tannin source. It has been reported that different tannins may have a varying degree of a nity with salivary proteins in sheep [33]. Moreover, beside the botanical difference of the plants, the CH and TA extracts enriched the diet with ellagitannins and gallotannins, respectively [34]. Deschamps et al. [35,36] report that hydrolysable tannins contained in CH and TA extracts are susceptible to biodegradation by bacterial tannase in the rumen with the production of several intermediates along with the enzymatic reactions, to form acetate and butyrate as end products. Though ruminal biodegradation intermediates of tannins were not investigated in the present paper, it could be stressed that the difference observed between CH and TA groups could be in part linked to the different toxicity of intermediate products. The last consideration is related to the ability of tannins to reduce nutrient availability. In our study, CH showed ADG lower than all the other groups. Noteworthy, this result seems to be due to the lower DMI rather than other factors, such as the reduction of protein digestibility. As observed in the companion paper, the level of ammonia in the rumen was not affected by dietary treatment, tannin supplementation nor tannin type [15]. Rumen ammonia is related with the proteolytic bacterial metabolism, which may represent a limiting factor for the rate of growth in young animals.
Similar to animal performance, our results on the effect of dietary tannins on the fatty acid composition of rumen content seem to point a different direction in comparison with that indicated by the literature. Most of the previous in vitro and in vivo studies agree that tannins may impair the progressive saturation of PUFA to 18:0 and, as a consequence, increase the out ow of health-promoting fatty acids that can be absorbed, further metabolized and accumulated in the tissues [37,38]. In the rumen, an increase of PUFA would be observed when the rst steps on the onset of the BH are inhibited [39,40,41]. On the contrary, the inhibition of the BH at different steps along the pathways might result in a lower formation of 18:0 and greater accumulation of intermediates such as 18:2c9t11 and 18:1t11 as well as other trans 18:1 isomers [42,43,44]. It should be mentioned that the characteristics of the basal diet seem to play an important role in determining the effects of tannins on rumen BH. The 18:1t11 accounts up to 70% of total trans 18:1 isomers in forage-fed ruminants [45], while the progressive replacement of forages with concentrates increases 18:1t10 at the expense of 18:1t11 due to a changed BH pathway [46]. Vasta et al. [10] observed that condensed tannins from quebracho hampered this phenomenon when they were administered to concentrate-fed lambs compared to herbage-fed animals. Apparently, in our study, the administration of dietary tannins was ineffective to interfere with the different steps of the biohydrogenation, as no signi cant difference was observed in the rumen for total PUFA, 18:2c9c12, 18:3c9c12c15, 18:2c9t11, 18:1t11 and 18:0. However, we found that the proportion of 18:1t10 was lower in the rumen of tannin-fed lambs. This reduction was not in favour of a greater accumulation of 18:1t11 in the rumen, suggesting that tannin supplementation could have speci cally reduced alternative pathways of BH, with HT tending to be more effective than CT. These results are not easily explainable in relation to what we have found in the companion paper. In that study [15], HT-fed lambs showed a lower ruminal pH than CT-fed lambs, which is a condition that favours the synthesis of 18:1t10 in the rumen [46]. Also, it has been reported that different bacterial species may be responsible for different BH pathways. In the companion paper, Salami et al. [15] observed that the supplementation of tannins did not substantially alter the bacterial community and relative abundance in the rumen of the same lambs used in the present. Only Fibrobacteres, representing a minority group of total bacteria in the rumen, were signi cantly lower in the tannin-receiving groups in comparison with the CO. However, it should be noticed that the relative abundance does not necessarily determine the functional activity of a bacteria group in the rumen. In other studies, changes in Fibrobacter has been linked to the modi cation in BH [47]. Also, Salami et al. [15] reported treatment differences in bacterial genera such as Prevotella and Ruminobacter and these genera have been suggested to play signi cant roles in BH. The lack of a straightforward effect on rumen microbial community seems in agreement with the proportion of OBCFA observed in the rumen. The presence of OBCFA exclusively arises from microbial metabolism and a variation in OBCFA is expected to re ect changes in the relative abundance of the speci c bacterial population [48]. In particular, Vlaemink et al. [49] reported that bre-digesting bacteria are richer in even and/or odd-iso-fatty acids than amylolytic bacteria, which are characterized by the predominance of 15:0 and 15anteiso. In our study, total and individual OBCFA did not follow a speci c trend that could help to unveil possible effects of tannins and tannin type on the microbial community useful to explain the results observed for 18:1t10.
The fatty acid composition of meat partially agrees with the results observed in the rumen content. In particular, the lack of effect of the investigated factors on the main substrates of the BH (18:2c9c12, 18:3c9c12c15), total PUFA and end-product of the process (18:0) would con rm that, in our experimental conditions, the addition of tannins or the type of tannins supplied to growing lambs did not protect dietary PUFA from BH and did not favour the accumulation of healthful fatty acids. Moreover, in accordance with the rumen content, 18:1t10 tended to be greater in the meat from CO group than tannin-receiving groups and in the meat from CT-fed lambs than HT-fed lambs, suggesting that an impairment of the biohydrogenation process occurred, in particular of the alternative pathways of the BH, and that HT could have been more effective than CT. However, some inconsistencies between fatty acids in the rumen content and meat seem to indicate a wider effect of the tannins. Discrepancies between the fatty acid pro le of rumen content and meat could depend on several factors, such as the collection technique or the sampling time. In this study the rumen fatty acids describe the speci c moment of slaughter, while meat fatty acids result from absorption mechanisms and further lipid metabolism in liver and tissues. Therefore, meat composition may add useful information on fatty acid metabolism. Differing from rumen content, 18:1t11 followed the same trend as the 18:1t10. Due to the reasons explained above, usually, these two fatty acids are negatively correlated. In the present study, the contemporary reduction of both 18:2c9t11 and 18:1t10 could indicate that the administration of dietary tannins may have caused a general reduction of BH extent, though with a magnitude lower than reported in the literature. It should be stressed that the characteristics of the basal diet may play an important role in favouring the effects of dietary tannins. For example, Jeronimo et al [50] and Frutos et al. [7] highlighted that the supplementation of PUFA-rich oils may enhance the effect of tannins on ruminal BH. In the present trial we did not add lipid supplements therefore, it cannot be excluded that a greater availability of PUFA in the rumen could have hampered the effect of tannins on BH.
Individual fatty acids signi cantly differed between treatments both in the rumen content and meat. However, the univariate approach did not allow identifying speci c patterns of fatty acid metabolism due to the tannin extracts. Differing from the univariate, the multivariate statistics can be implemented to assess the combined role of the individual variables of a dataset in highlighting overall patterns within the dataset. For example, among the multivariate approaches, the canonical discriminant analysis may allow to discriminate between subjects (samples) belonging to different groups. The canonical discriminant analysis can be coupled with a stepwise procedure that identi es the variables of the dataset having the biggest discriminating ability. This approach has been previously applied to different datasets to discriminate between meat or cheese samples differing for the geographical origin, breed or production system [26,51,52]. In the present study, the same approach was used: i) to assess whether the multivariate structure of the FA pro le was able to reveal overall patterns able to discriminate the dietary treatments, ii) to identify the fatty acids characterized by a greater discriminating power both in the rumen content and meat. After the stepwise procedure, 8 and 6 fatty acids were retained from the respective datasets of rumen content and IMF, and the retained fatty acids were linearly combined into 4 canonical discriminant functions for each biological matrix. All the fatty acids with a high standardized coe cient included in the two rst CAN for each of the two datasets arise from rumen microbial metabolism or BH, except for 22:0 in the rumen content. However, most of them were not signi cantly affected by the dietary treatments and only 66% (rumen content) and 57% (meat) of cases were correctly assigned to the original experimental group after cross-validation. These results seem to con rm that the dietary administration of the four tannin extracts did not produced speci c patterns describing their effects on fatty acid metabolism. A similar nding was reported by Morales et al. [53] after a regression analysis conducted on 12 different experiments to investigate the response of meat fatty acids to tannins supplementation. The authors reported that, tannins were associated only with the reduction of 18:0. Nevertheless, it should be highlighted for both meat and rumen content, the CO group was discriminated from the tannin-fed groups, with only case from the CO group being misassigned to a tannin-supplemented group (data not shown). Together with the results of the univariate ANOVA, the multivariate analysis showed that dietary tannins exerted a general modi cation along the whole BH process, though it was not possible to identify speci c patterns for each of the administered extracts.

Conclusions
In conclusion, our study con rmed that the dietary administration of tannins may have an effect in determining the fatty acid pro le of the rumen content and meat in lambs. However, it was not possible to clearly describe speci c mechanisms of action of the individual supplemented tannin extracts on precise steps of the fatty acid metabolism, as previous studies supposed. Nevertheless, in our study, the multivariate structure of the fatty acid pro le of rumen content and meat allowed the discrimination of the CO treatment from the other groups. Interestingly, the fatty acids showing the greater discriminant power arise from microbial metabolism and the principal or alternative BH pathways. These results may give new perspectives in investigating the effect of tannins on BH. From one side, our study further con rms the impact that dietary tannins may exert on the rumen BH. From the other side, it provides clues that tannins can have a generalized in uence on BH rather than on speci c steps. Declarations