Sorghum ( Sorghum bicolor ) grain is a common feed for ruminant and monogastric animals in Africa, Asia and America because of its high starch and energy content, similar to that of maize ( Streeter et al. , 1990 ; Rowe & Pethick, 1994 ). However, it is seldom used in Europe because of its highly variable feeding quality in terms of protein and starch availability, and digestibility ( Bramel-Cox et al. , 1990 ; Opatpatanakit et al. , 1994 ). Besides, the presence of condensed tannins and other phenolic compounds in some varieties is assumed to limit nutrient availability ( Dykes & Rooney, 2006 ; Barros et al. , 2012 ). Variability among genotypes in nutritive value has been related to the content in amylose and proportion of vitreous starch, the nature of the endosperm protein, and the concentration of phenolic compounds ( Rooney & Pflugfelder, 1986 ). However, improvements in nutritive value are often associated with reductions in agronomic traits such as grain yield, maturity and bird resistance ( O´Brien, 1999 ).

In general, tannins (mostly proanthocyanidins) are assumed to reduce rumen microbial fermentation of substrates by binding to nutrients, which will be available to intestinal digestion after low pH dissociation when passing through the abomasum ( Mueller-Harvey, 2006 ). In contrast, tannins may also have a positive effect, preventing bloat or enhancing efficiency of energy utilisation by reducing methanogenesis ( Mueller-Harvey, 2006 ; Jayanegara et al. , 2009 ). Besides, the low digestible proteins from the sorghum endosperm (kafirins) that are embedded in the starch matrix may reduce access to microbial enzymes ( Rooney & Pflugfelder, 1986 ; McAllister et al. , 1993 ). Because of this restricted rumen fermentation, sorghum probably delivers an important proportion of starch to the lower gut ( Rowe & Pethick, 1994 ), and the non-degraded starch that passes to the intestine provides 42% more energy than that digested in the rumen ( Owens et al. , 1986 ). In this regard, Hibberd et al. (1985) found that up to 25% of variation in the nutritive value of sorghum varieties largely depends on the digestion site. Complexes between tannins and nutrients that are stable at common rumen pH can be dissociated at pH under 3.5 ( Jones & Mangan, 1977 ), thus making nutrients available for digestion at the lower gut.

In a previous study ( Amanzougarene et al. , 2017 ), major differences in in vitro fermentation were observed between three varieties of sorghum grain, but the relative importance of the nutritive factors determining the magnitude of those differences was not defined. Therefore, this in vitro study aimed to evaluate the potential differences in ruminal fermentation of grains from three sorghum varieties (a white sorghum, S1, and two brown sorghums, S2 and S3), and to determine to what extent either the content of tannins and other phenolic compounds or the endosperm protein matrix may affect their microbial fermentation characteristics and starch degradability. In a first incubation trial (Trial 1), the effect of phenolic compounds on sorghum fermentation pattern was monitored for 24 h of incubation; after, a subsequent incubation (Trial 2) for 12 h was carried out to explore the effect of the three sorghum varieties on substrate degradation. Gas production and fermentation parameters were compared between Trials 1 and 2.

Protein of the three substrates was characterised in terms of buffer solubility and enzymatic digestibility, either in presence or absence of PEG, to estimate the possible effect of tannins on those variables (Fig. 1). Inactivation of tannins by the addition of PEG to the medium scarcely affected the proportion of soluble protein of the varieties tested (average differences between means below 0.0024 g/g), and therefore the effect of PEG on protein solubility was not considered; thus, average protein solubility was 0.194 ± 0.0009, 0.190 ± 0.0073 and 0.146 ± 0.0036 g/g for S1, S2 and S3, respectively. Protein enzymatic digestibility was the highest in S1 (0.614 ± 0.0100 g/g), intermediate in S3 (0.454 ± 0.0053 g/g) and lowest in S2 (0.226 ± 0.0104 g/g). The inclusion of PEG only showed an effect on protein digestibility in the case of S2 (0.618 ± 0.0139 g/g when PEG was added).

In a previous study ( Amanzougarene et al. , 2017 ), we compared the fermentation pattern of three varieties of each corn, barley and sorghum grain, and observed the major range of differences in the response occurring in the latter. The aim of this work was to define the influence of nutritional characteristics on the extent of microbial fermentation, rather than to characterise the nutritive value of three varieties of sorghum grain for ruminants.

In general, chemical composition of the experimen­tal sor­ghum substrates agreed with values recorded from 23 samples by Lanzas et al. (2007) , but starch proportion was lower than those reported by Rowe & Pethick (1994) and Offner et al. (2003) . There were no major differences in CP or NDF proportions between the three sorghum varieties (Table 1), despite O'Brien (1999) reported that CP content of sorghum grains greatly varies among genotypes. Besides, only a slightly higher starch proportion was detected in S1 (0.06 and 0.08 higher than in S2 and S3, respectively). The phenolic and tannin contents of the three varieties studied here were within the range cited by Dykes & Rooney (2006) for white and brown sorghums.

The pattern of in vitro gas production of raw sor­ghum varieties ( i.e. , when PEG was not added) showed rates and extent of fermentation in the range of those observed by Streeter et al. (1990) , Opatpatanakit et al. (1994) and Lanzas et al. (2007) . The white sorghum S1 was more extensively fermented than the brown varieties S2 and S3 in both incubation trials (Figs. 2a and 4a, respectively). Throughout both incubation trials, the volume of gas produced from S1 was over 0.10 higher than that from S2 and S3, in a proportion of 0.10 to 0.18 after 10-12 h. These results agree with the greater DM and starch disappearance recorded with S1 (Table 4), although the magnitude of treatment differences in gas production was two-fold higher to that observed among treatments in DMd. It has to be mentioned that incubation conditions in the second trial were adjusted to 12 h for the in vitro study of DM and starch degradation considering that, from the calculations of Offner et al. (2003) , up to 0.73 sorghum degradation occurs at that time and, consequently, differences between substrates in in situ starch degradation are maximised.

Differences in the rate and/or extent of in vitro fermentation of sorghum varieties by rumen microbes have been mainly associated to three factors: the content and nature of starch, the interaction with the protein matrix embedded in the endosperm, and the presence of tannins and phenolic compounds. The nature of starch in terms of either its amylose to amylopectin ratio or the floury or vitreous endosperm was not studied here. Some authors have associated a higher DM degradation and gas production of sorghum or maize grains to the proportion of waxy starch, because its branched nature allows for a better microbial access ( Streeter et al. , 1990 ), or to a lower vitreousness ( Philippeau & Michalet-Doreau, 1997 ). Rooney & Pflugfelder (1986) attributed some effect of chemical composition and physical structure of sorghum starch on grain digestibility; however, differences were minimised when they studied isolate starches. Therefore, as hypothesized by van Barneveld (1999) , minor differences in fermentation between the three studied sorghum varieties can be assumed related to their type or amount of starch. In any case, the higher starch content in S1 might partly explain the higher volume of gas produced, associated to the higher starch disappearance observed (Table 4).

In the endosperm of cereal grains such as maize or sorghum, the nature and structure of the protein matrix in which the starch granules are embedded may restrict access of amylases to starch ( Rooney & Pflugfelder, 1986 ). In maize, McAllister et al. (1993) observed an increase in starch degradation after 4 h of preincubation with protease. Such interaction between endosperm proteins and starch in reducing rate and extent of degradation may partly explain differences between vitreous or floury endosperm ( Philippeau et al. , 2000 ). Sorghum protein fraction includes kafirins, accounting for around 0.80 of total endosperm protein, and non-prolamin proteins ( Belton et al. , 2006 ). The sorghum varieties studied here had similar total CP content (106 to 113 g/kg, Table 1) and proportion of soluble proteins (0.15 to 0.19 of total CP, Fig. 1); however, 4 h protease digestion was the highest in S1, intermediate in S3 and lowest in S2. This should imply a wider access of microbial enzymes to endosperm starch in S1, partly explaining its highest starch disappearance (Table 4) and gas production (Figs. 2a and 4a). However, range of differences in protein digestibility did not apparently affect comparison in microbial fermentation between S2 and S3, since both recorded similar results in gas production and starch disappearance, suggesting that other factors should affect comparison between these two varieties.

The third mentioned factor on sorghum starch utilisation was the presence of phenolic compounds and tannins. Tannins may inhibit amylases but, more importantly, bind to carbohydrates and restrict enzyme access to nutrients, limiting their utilisation in the rumen ( Makkar et al. , 1995 ; Mueller-Harvey, 2006 ). Streeter et al. (1990) observed a 1.4-fold higher gas production from tannin-free than from bird-resistant sorghum varieties. In practical situations, the inclusion of sorghum grain in ruminant diets is commonly linked to its content of condensed tannins, flavonoids, and simple phenolic compounds, with ferulic and p-coumaric acids being dominants ( Dykes & Rooney, 2006 ), that may hamper nutrient utilisation in the rumen. Although flavonoids and simple phenolic acids may form complexes with proteins, there is no conclusive evidence that such interaction would cause a reduction in protein and starch digestibility like condensed tannins ( Duodu et al. , 2003 , Belton et al. , 2006 ), and thus the effect of these compounds on fermentative activity may differ according to their chemical nature. Barros et al. (2012) observed a two-fold increase of in vitro resistant starch with condensed tannins than with phenolic monomers.

The sorghum varieties used in this experiment had variable proportions of phenolic compounds, with negligible contents in S1, minor concentration in S3 and moderate proportions of both phenolic acids and tannins in S2 (Table 1). It is worth mentioning that even being a brown variety, tannins content in S3 were quite low, showing that pericarp colour is not a reliable indicator of tannins in sorghums ( Dykes & Rooney, 2006 ). The ability of PEG to bind to tannins allows for estimating the effect of tannins on fermentative activity of rumen microbial population ( Makkar et al. , 1995 ; Silanikove et al. , 2001 ) and, although some minor changes in microbial community have been reported ( Belenguer et al. , 2011 ), it is consider to be inert to rumen function ( Makkar et al. , 1995 ). To our knowledge there is no information about the binding capacity of PEG to simple phenolic compounds; however, Khazaal et al. (1996) and Getachew et al. (2002) observed almost the same correlation coefficients for the relationship between either TT or TP and the increase in gas production when fibrous feeds were incubated in the presence of PEG. Since in those works tannins were, on average, 0.53 and 0.69, respectively, of TP of the substrates, it could be assumed that other phenolic compounds would react similarly.

The increase in lactate concentration when adding PEG to the media might be attributed to an increase in starch availability when the effect of tannins was inactivated; however, the lack of response differences between the tannin-free S1 and S2 and S3 might partly suggest an artefact in the colorimetric method of analysis. As it can be seen in Figs. 2 and 4, as well as in Table 4, the inclusion of PEG in the incubation solution showed no effect on gas production or starch disappearance when S1 was incubated, indicating the absence of any measurable effect of phenolic compounds on the activity of microbes. A minor response was found with S3, because of its low proportion of TP and TT. In this case, the lower gas production from S3 should thus be mostly a response to the lower digestibility of the protein fraction, interfering in starch utilisation at a higher extent than in S1, as explained above. In contrast, the negative effect of tannins on fermentation of S2 was clearly manifested, since their inactivation increased gas production values (BE) over 1.20 to 1.30 times that of the substrate without PEG in the first incubation trial (Fig. 3). The magnitude of such response is relatively high, considering the amount of TP and TT in S2, since our group previously observed BE of tannins in browse forage legumes ranging from 1.1 to 1.3 in Acacia cornigera , with almost two-fold higher TP and TT concentration, and from 1.3 to 1.6 in Leucaena leucocephala with five and eight-fold higher TP and TT contents ( Rodríguez et al. , 2014 ). Fig. 1 shows that inactivation of tannins when PEG was ad­­ded could also allow for a high digestibility of the protein fraction of S2, and thus protein was not the limiting factor in fermentation of starch with this variety.

From the results obtained it can be concluded that differences in nutritive value between sorghum varieties can be mainly attributed to the digestibility of its protein fraction, which may restrict the access of microbial enzymes to starch, and the biological effect of the phenolic compounds (tannins and simple phenolic compounds), whereas the chemical and physical structure of starch may be less important. Colour of grain does not necessarily imply a high proportion of phenolic compounds, and consequently a lower extent of rumen utilization.