INTRODUCTION
Protein supplements with a high energy content that are resistant to ruminal degradation, yet available for digestion in the small intestine, may help supply the extra amino acids (AA) and energy needed for high milk production in early lactation. Whole soybeans (SB), rich in protein and fat, are mostly used in feeding of dairy cows, but protein in raw SB is degraded readily by rumen microbes (Stern et al. Reference Stern, Santos and Satter1985) and cannot meet the high demand for rumen undegradable protein (RUP) in early lactation. Heat treatment of SB, especially roasting (Mohamed et al. Reference Mohamed, Satter, Grummer and Ehle1988), reduces ruminal protein degradation and destroys anti-nutritional factors, improving the nutrient balance for dairy cows (Faldet & Satter Reference Faldet and Satter1991). However, there have been conflicting results regarding animal responses to feeding heat-treated SB. Increased milk yield has been obtained from feeding heat-treated SB to lactating cows compared with yields from those fed soybean meal (SBM) and raw SB (Ruegsegger & Schultz Reference Ruegsegger and Schultz1985; Voss et al. Reference Voss, Stehr, Satter and Broderick1988). However, there are also trials reporting no improvement in performance (Mielke & Schingoethe Reference Mielke and Schingoethe1981; Van Dijk et al. Reference Van Dijk, O'Dell, Perry and Grimes1983). In many studies showing improvement, increase in milk yield may be related to type of forage in the diet, as improved performance is usually achieved for diets with lucerne silage as the forage because of the high solubility of its protein (Voss et al. Reference Voss, Stehr, Satter and Broderick1988; Chouinard et al. Reference Chouinard, Girard and Brisson1997). In addition, no improvement was observed when roasted SB compared with SBM was used in mixed lucerne–maize silage diets (Satter et al. Reference Satter, Dihiman and Hsu1994), but mixed lucerne hay–maize silage diets, used widely in tropical areas like Iran, were not tested. Thus, the main objective of the current study was to investigate the performance responses of Iranian Holstein cows in early lactation to the feeding of Iranian roasted SB, raw SB and SBM in combination with lucerne hay–maize silage diets. The other objective was to test the combination of SB (as an oil seed) plus cottonseed meal (CSM), as an available source of low-cost protein supplement, versus cottonseed (CS) plus SBM.
MATERIALS AND METHODS
Cows and treatments
Fourteen multiparous Holstein cows, with similar 305-day milk production in the previous lactation, were housed after calving in a tie stall barn for 2 weeks for barn adjustment and collection of pre-trial data. Data were collected during week 2 of the pre-trial period. During the pre-trial period cows were fed with the same diet which was formulated to meet the early lactational requirements. At the onset of the 45-day trial, cows were grouped according to pre-trial milk yield and then randomly assigned from these subgroups to one of three experimental diets. At the beginning of the trial, cows were at average parity of 2·3 (±0·46), body weight (BW) of 617 (±48·0) kg, DIM of 17 (±3·8), and milk yield of 33 (±3·4) kg/day. Cows in each group were fed individually one of three experimental diets as a totally mixed ration (TMR). The diets, which contained 120 g SBM/kg plus 82 g/kg CS, 120 g raw SB/kg plus 82 g CSM/kg and 120 g roasted SB/kg plus 82 g CSM/kg (Table 1), were formulated to be iso-nitrogenous and iso-caloric, and meet National Research Council (NRC) recommendations (NRC 2001) (Table 1). Roasted SB were obtained by roasting seeds for 2 min in a commercial roaster (exit temperature of seeds was 140–145°C) and immediately placing, without cooling, in covered wooden barrels for 45 min. Both raw and roasted SB were coarsely cracked, giving mostly halves and quarters. Feed was mixed once daily as a TMR but was fed twice daily (09.00 h and 16.00 h) in amounts to provide 100 g refusals/kg. Forages (lucerne hay and maize silage) were sampled twice a week for DM determination in a forced-air oven and other feed ingredients were adjusted accordingly once a week. Feed intakes were recorded daily and cows had free access to water.
Table 1. Ingredients and nutrient composition of diets

* Average 145 g CP/kg, 416 g NDF/kg and 328 g ADF/kg (DM basis).
† Composition of vitamin/trace-mineral mix: Ca, 196·0 g/kg; P, 96·0 g/kg; Mg, 19·0 g/kg; Fe, 3·0 g/kg; Na, 71·0 g/kg; Cu, 0·3 g/kg; Mn, 2·0 g/kg; Zn, 3·0 g/kg; Co, 0·1 g/kg; I, 0·1 g/kg; Se, 0·01 g/kg and Vit A, 500 000 IU/kg; Vit D, 100 000 IU/kg; Vit E, 100 IU/kg.
‡ Estimated from NRC (2001) values.
§ NFC calculated as: 1000− (g CP/kg+g NDF/kg+g EE/kg+g ash/kg).
∥ Estimated RUP (g CP/kg intake)={B×[k p/(k p+k d)]}×1000, where B is potentially degradable protein, k p is rumen passage rate and assumed to be 0·08/h, and k d is degradation rate of protein. For all diet ingredients, k d were determined by the in situ technique (Ørskov & McDonald Reference Ørskov and McDonald1979) and the results were close to values of NRC (2001).
Sampling and routine analysis
Samples of lucerne hay, maize silage, and concentrate mixes were collected weekly, dried at 55°C for 48 h and then ground to pass through a 2 mm screen. The sub-sample was then dried at 105°C for 24 h for DM determination. Contents of crude protein (CP), ether extract (EE), ash, Ca and P of TMR diets and their components were determined by Association of Official Analytical Chemists (AOAC) methods (AOAC 2000) and neutral detergent fibre (NDF) and acid detergent fibre (ADF) were assessed using the method of Van Soest et al. (Reference Van Soest, Robertson and Lewis1991). Because of the importance of the difference in RUP content of raw SB and roasted SB, the RUP of these seeds and all other diet ingredients were determined using the in situ method (Ørskov & McDonald Reference Ørskov and McDonald1979).
Cows were milked three times a day. Milk yield was recorded daily and samples were collected weekly at each milking. Milk samples were combined on an individual cow basis. The 24-h composite of each cow's milk was split into two portions for analysis. One portion was refrigerated at 4°C and then analysed for fat, protein, lactose and solids not fat (SNF) using an infra-red analyser (Milkoscan, 133 B, Foss Electric, Denmark) and the remaining portion was stored at −20°C until analysed for MUN. Milk samples were de-proteinized and analysed for MUN concentration as is described for PUN. Cows were weighed on two consecutive days at the beginning and end of the trial period and once weekly during the period.
During weeks 3, 5 and 7, blood samples (15 ml) were collected in heparinized vacutainers from the coccygeal vein or artery at 2 and 4 h post-feeding. To separate plasma, blood samples were centrifuged immediately after taking at 1500 g at 25°C for 15 min. Blood plasma samples were stored at −20°C for about 1 week for further analysis. Commercial enzymatic kits were used to analyse plasma glucose (Pars Azmoon glucose kit, Tehran, Iran), urea (Pars Azmoon urea kit, Tehran, Iran), and β-hydroxybutyrate (BHB) (D-3-hydroxybutyrate kit, Randox Laboratories Ltd, Antrim, UK). Analysis was performed by an autoanalyser (Biotecnica, Targa 3000, Rome, Italy). Plasma concentration of individual free AA was determined in de-proteinized plasma (Clarke Reference Clarke1993).
Samples of rumen fluid were collected fortnightly from each cow 3 h after feeding by inserting a needle into the ventral rumen and aspirating the rumen fluid directly (rumenocentesis; Duffield et al. Reference Duffield, Plaizier, Fairfield, Bagg, Vessie, Dick, Wilson, Aramini and McBride2004). Rumen pH was determined immediately on a fresh aliquot. A 5-ml aliquot of rumen fluid was acidified with 5 ml of 0·1 M HCl, centrifuged, the supernatant frozen at −20°C, and later analysed for ammonia N as described by Mielke & Schingoethe (Reference Mielke and Schingoethe1981).
Statistical analysis
Data were analysed through repeated measures using the MIXED procedure of SAS (1999). Data are presented as least squares means and standard errors of the means. Variation in the dependent variable, Y, was described as follows: Y ijk=μ+D i+C (i)j+T k+(D×T)ik+βX ijk+e ijk, where Y ijk=dependent variable, μ=overall mean of Y, D i=effect of diet (i=1–3), C (i)j=effect of cow within diet (j=1–5 for raw and roasted diets and j=1–4 for SBM diet), T k=effect of time (k=1–45 for dry matter intake (DMI) (kg/day and proportion of BW), RUP intake, production efficiency, milk yield, milk fat and milk protein yield; k=1–6 for milk composition and MUN; k=1–3 for rumen pH and ammonia N concentration, and plasma glucose, BHB, PUN, and plasma concentrations of AA), (D×T)ik=interaction of diet and time, X ijk=DMI, milk yield, milk fat and milk protein yield and milk composition (data collected during the pre-trial period were included in the model as a covariate and the effects were not statistically significant), β=regression coefficient of Y on X, e ijk=random residual error (within cows).
For changes in BW, the model included the effects of diets. First-order autoregressive covariance was chosen for the structure of the variance–covariance matrix. The following planned orthogonal contrasts were included in the model: (1) roasted SB v. raw SB and (2) SB plus CSM v. SBM plus CS.
RESULTS
Daily intake, milk yield and composition, and BW
DMI expressed as kg/day or proportion of BW was significantly higher for cows fed roasted SB than for cows fed raw SB but there was no significant difference between SB plus CSM diets and SBM plus CS (Table 2). RUP intake also followed the same pattern as DMI with cows fed roasted SB consuming significantly higher RUP (Table 2) than the cows fed raw SB. Mean BW of cows at the beginning and end of the experimental period was not significantly different among diets (Table 2).
Table 2. Daily intake, milk yield, milk composition and BW as affected by diet

* Contrast 1 is roasted SB v. raw SB and contrast 2 is (SB plus CSM) v. (SBM plus CS).
† P>0·05.
‡ Production efficiency=average daily 35 g FCM/kg (kg/day)/average daily DMI (kg/day).
Milk yield and composition are shown in Table 2. Cows fed roasted SB produced 1·3 kg more milk throughout the study than cows fed raw SB. Cows fed SB plus CSM diets produced 1·4 kg/day more milk than cows fed SBM plus CS. This corresponds to an increase in yield of 35 g fat-connected milk (FCM)/kg to 1·2 and 1·3 kg/day, respectively. Milk fat concentration was similar among diets but milk fat yield was higher for cows fed roasted SB compared with cows fed raw SB and cows fed SB plus CSM diets compared with SBM plus CS (40·0 and 40·0 g/day, respectively). Milk protein yield was similar for cows fed different diets. Milk protein concentration was the same for cows fed raw SB and roasted SB, but depressed by 1·8 g/kg for cows fed SB plus CSM diets compared with SBM plus CS.
Blood and ruminal metabolites
A dietary effect on rumen pH values, glucose and BHB concentrations was not detected among cows fed different diets (Table 3). Rumen ammonia N concentration of samples taken 3 h after feeding was significantly lower for the cows fed roasted SB compared with those fed raw SB. There was no difference between rumen ammonia N concentration of cows fed SB plus CSM diets and SBM plus CS.
Table 3. Ruminal and plasma metabolite concentrations (mg/dl) and rumen pH as affected by diet

* Contrast 1 is roasted SB v. raw SB and contrast 2 is (SB plus CSM) v. (SBM plus CS).
† Rumen fluid samples were collected 3 h after feeding.
‡ P>0·05.
§ Mean of blood samples taken 2 and 4 h after feeding.
MUN and PUN concentrations were significantly lower in cows fed roasted SB than in cows fed raw SB but were not different between cows fed SB plus CSM diets and cows fed SBM plus CS (Tables 2 and 3).
Plasma concentrations of five essential amino acids (EAA) were significantly higher in cows fed roasted SB than in cows fed raw SB but Leu and Phe plasma concentrations were higher in cows fed raw SB (Table 4). Plasma concentrations of all EAA, except Phe, were the same in cows fed SB plus CSM diets and SBM plus CS. Plasma concentrations of total EAA were not affected by diet. Furthermore, plasma concentrations of most NEAA (except Glu and Tyr) and total NEAA were not affected by diet.
Table 4. Plasma concentrations (μmol/dl)Footnote * of EAA and non-EAA (NEAA) as affected by diet

* Blood samples were taken 4 h after feeding.
† Contrast 1 is roasted SB v. raw SB and contrast 2 is (SB plus CSM) v. (SBM plus CS).
‡ P>0·05.
DISCUSSION
Daily intake and BW change
The results of the present study are in contrast to previous reports that DMI was not affected when cows were fed raw SB, extruded SB (Stern et al. Reference Stern, Santos and Satter1985) or roasted SB (Faldet & Satter Reference Faldet and Satter1991; Tice et al. Reference Tice, Eastridge and Firkins1993). However, other studies have reported a reduction in DMI as a result of feeding raw SB or heated SB (Anderson et al. Reference Anderson, Obadiah, Boman and Walters1984; Mohamed et al. Reference Mohamed, Satter, Grummer and Ehle1988) or increased DMI when cows were fed heated SB (Voss et al. Reference Voss, Stehr, Satter and Broderick1988) or raw SB (Owen & Edionwe Reference Owen and Edionwe1986). Increased DMI of diets including roasted SB provides indirect evidence that soy oil from roasted SB did not depress fibre digestion. DMI has been affected negatively in other studies when similar amounts of soy oil were fed in the free form (Mohamed et al. Reference Mohamed, Satter, Grummer and Ehle1988). Faverdin et al. (Reference Faverdin, Mohamed and Vérité2003) observed that infusion of soya protein, as a digestible protein source, at a rate of 800 g/day into the duodenum of high producing dairy cows, stimulated appetite and improved DMI and milk production. They suggested that this was due to an increase in the energy demand of the mammary gland and/or a specific action of plasma AA concentration. This theory possibly explains higher DMI in the cows fed roasted SB than raw SB in the current study. Another reason for higher DMI of the roasted-SB-fed cows could be higher palatability of roasted SB compared with raw SB. Cows fed roasted SB had higher RUP intake compared with cows fed raw SB, which is consistent with differences in dietary concentrations (Table 1). The use of CSM in combination with SB in diets instead of using SBM plus CS did not have any significant effect on DMI, which showed the potential of using CSM in diets including SB.
Changes in BW were negative for all cows fed SBM plus CS, raw SB plus CSM and roasted SB plus CSM due to early lactation and no differences occurred among diets. In general, no effect on BW was observed in experiments in which raw or roasted SB was included in the TMR of lactating cows (Tice et al. Reference Tice, Eastridge and Firkins1993). The BW responses to diet are more likely to be statistically and biologically significant when measured over complete lactations to avoid carryover effects from one period to the next.
Milk yield and composition, and blood/ruminal metabolites
Satter et al. (Reference Satter, Faldet, Socha and Jordan1991) presented a summary of nine trials in which heated SB (excluding extruded SB) were compared with raw SB. Average responses for milk and FCM were 1·4 and 1·2 kg/day, respectively, which is close to the current results. Some studies using cows in early lactation and more thoroughly heated SB resulted in about 4·5 kg more milk/day in cows fed heat-treated SB than in cows fed raw SB or SBM (Voss et al. Reference Voss, Stehr, Satter and Broderick1988; Faldet & Satter Reference Faldet and Satter1991). Others have reported no differences in milk production between cows consuming diets with or without roasted SB (Mohamed et al. Reference Mohamed, Satter, Grummer and Ehle1988). Differences in milk production responses to feeding heat-treated SB that have been reported can be attributed to many factors, such as stage of lactation, rate of milk production, type and extent of heat treatment and forage type. Voss et al. (Reference Voss, Stehr, Satter and Broderick1988) suggested that heat-treated SB may offer greater potential for increasing milk yield when diets are based on lucerne haylage because of the greater solubility of its protein. Increased milk yield of cows fed roasted SB compared with raw SB in the current study can be attributed to many factors including higher DMI (primarily), changes in ruminal and blood metabolites (such as the specific action of the concentration of some plasma AA), lower rumen ammonia and lower PUN and MUN, and increased intake of Net Energy for Lactation (NEL) and RUP. Despite the adequate supply of RDP in all diets, because microbial protein supply is insufficient to meet the protein and AA requirements of high-yielding cows both qualitatively and quantitatively, roasted SB with lower ruminal degradation caused a higher supply of RUP and maybe flow of AA to the intestine and hence increased milk production. In addition, higher plasma concentration of Met and Lys, which are the most limiting AA, in cows fed roasted SB could explain the higher milk production of these cows. However, Stern et al. (Reference Stern, Calsamiglia and Endres1994) reported that non-ammonia N flow to the small intestine might not be increased when RUP is increased, because of a decrease in microbial protein synthesis possibly resulting from a reduction in dietary RDP and ammonia formation in the rumen. The response to increased RUP in the current study might be attributed to sufficient supply of RDP across diets. Although the cows fed roasted SB consumed more NEL than raw-SB-fed cows, the energy intake of cows fed different diets was more than the recommended levels of NEL requirement in NRC (2001), so this factor probably has a lower impact on higher milk production of cows fed roasted SB. The higher milk and 35 g FCM/kg in cows fed SB plus CSM diets compared with cows fed SBM plus CS showed that CSM can act as a suitable protein source in SB-included diets.
The same milk fat concentration of cows fed raw SB and roasted SB in the current study was in accordance with the results of other studies (Ruegsegger & Schultz Reference Ruegsegger and Schultz1985; Mohamed et al. Reference Mohamed, Satter, Grummer and Ehle1988). The same milk fat concentration of cows fed SB plus CSM and SBM plus CS showed the similar protection of oil in these two kinds of oil seed (SB and CS) because the oil content of SBM and CSM was the same.
The decrease of milk protein concentration in cows fed whole SB (roasted and raw) confirmed the results of Palmquist (Reference Palmquist and Ørskov1988) who found that the use of dietary SB (raw or roasted) caused milk protein depression.
Rumen pH
Although ammonia concentration, which can act as a buffer in the rumen (Geerts et al. Reference Geerts, De Brabander, Vanacker, De Boeve and Botterman2004), was higher in cows fed raw SB than roasted SB, no diet effects on rumen pH values were detected. These results are in agreement with those of other workers who have observed no differences in rumen pH when comparing diets containing roasted SB or popped SB with diets containing SBM, raw SB, extruded SB or free soy oils (Ruegsegger & Schultz Reference Ruegsegger and Schultz1985; Mohamed et al. Reference Mohamed, Satter, Grummer and Ehle1988). The same milk fat concentration of cows fed different diets could be the result of the same ruminal pH of these cows. According to a review by Garrett (Reference Garrett1996), cows with rumen fluid pH above 5·8 are considered normal, while those between 5·0 and 5·8 may be suffering from subclinical acidosis. The relatively high rumen fluid pH observed in the current study suggests that cows were probably not suffering from subclinical acidosis.
Rumen ammonia N
Concentration of ammonia in the rumen is a function of ruminal N degradation and amount of dietary energy available to ruminal micro-organisms. Because the level of non-fibre carbohydrate (NFC) was the same among diets, lower ruminal ammonia concentration in cows fed roasted SB compared with raw SB possibly arose as a consequence of lower ruminal protein degradability of roasted SB, and so can account for one of the factors affecting higher milk production of these cows compared to the others. These results are in contrast to those of Ruegsegger & Schultz (Reference Ruegsegger and Schultz1985) and Stern et al. (Reference Stern, Santos and Satter1985) who both observed no differences in rumen ammonia N concentration among diets that included raw SB, heat-treated SB or SBM. However, Tice et al. (Reference Tice, Eastridge and Firkins1993) showed that roasting SB lowered ruminal ammonia N concentration.
Glucose and BHB
The results of the present study are in contrast to those of Driver et al. (Reference Driver, Grummer and Schultz1990) who observed that plasma glucose concentrations were significantly decreased in cows fed heat-treated SB compared to heat-treated SBM. Results of other studies that used whole SB as a source of dietary fat have shown increased plasma glucose concentrations and this effect is attributed to insulin resistance due to added fat in the diet. Although plasma glucose concentration was numerically higher for diets containing higher oil (diets containing raw SB or roasted SB), the differences were not significant. Plasma concentration of BHB revealed that all cows were in a healthy condition. Threshold concentrations of 10–14 mg/dl of plasma BHB can be used to discriminate between healthy cows and those with subclinical ketosis (Arieli et al. Reference Arieli, Adin and Bruckental2004).
Plasma and MUN
Although some studies indicate that PUN and MUN concentrations most often are influenced by RUP and RDP intake in the post-peak lactation period, and energy balance and protein mobilization (weight loss) might be important determinants of PUN and MUN in early lactation, in the current study PUN and MUN concentrations were influenced by dietary RUP and RDP. Recent studies (Nousiainen et al. Reference Nousiainen, Shingfield and Huhtanen2004; Wattiaux & Karg Reference Wattiaux and Karg2004) suggest an excess of RDP is expected to raise PUN and MUN concentrations more than an excess of RUP. In the study of Wattiaux & Karg (Reference Wattiaux and Karg2004), 100 g/day excess of RDP relative to NRC (2001) balance increased MUN by 1 mg/dl and an excess RUP of 350 g/day yielded the same rise in MUN. The current results are close to these values, so that for the roasted SB diet with 189 g excess of RUP and 56 g excess of RDP (based on actual amounts of measured DMI and NRC (2001) balance), MUN concentration was 13·3 mg/dl (minimum), whereas for the raw SB diet with no excess of RUP but 230 g excess of RDP, MUN was 14·7 mg/dl (maximum), showing the markedly greater impact of excess RDP on MUN. This may reflect the close association between RUP (AA absorbed from the small intestine) and energy balance. Broderick & Clayton (Reference Broderick and Clayton1997) also reported a marginally closer relationship of MUN with dietary CP content than with the ratio of CP to NEL. Those researchers speculated that this apparent discrepancy for effect of energy intake on MUN and PUN concentrations may reflect energy balance being confounded with dietary CP content in many studies. Similar patterns of ruminal ammonia concentration and urea in milk and plasma confirm the usefulness of PUN and MUN as indices of ruminal ammonia concentrations at similar CP concentrations in the diet (Broderick & Clayton Reference Broderick and Clayton1997).
As in many other studies, because urea equilibrates rapidly between body fluids (De Peters & Ferguson Reference De Peters and Ferguson1992) and enters the mammary gland by diffusion, concentrations of urea in plasma and milk were closely associated in the current study. The correlation between PUN and MUN was 0·90. The scatterplot of PUN versus MUN was linear, and the regression describing the relationship was:

Plasma concentrations of AA
The results of the current study contradict Faverdin et al. (Reference Faverdin, Mohamed and Vérité2003) who found plasma concentrations of all EAA to be higher in cows that received an infusion of soya protein at a rate of 800 g/day into the duodenum as a digestible protein source. Some other infusion studies (Korhonen et al. Reference Korhonen, Vanhatalo and Huhtanen2002) showed plasma concentrations of some AA such as His were very sensitive to changes in its supply, whereas plasma concentrations of Met, Ile, Leu and Phe did not reflect feed or digesta supply, which is in agreement with results of the current study. Post-absorptive metabolism may also explain the sensitivity of plasma concentration of some AA to their supply, as His was found to be the least oxidized EAA in the tissues of dairy cows (Black et al. Reference Black, Anand, Bruss, Brown and Nakagiri1990). Some evidence exists that Leu stimulates protein synthesis and reduces protein breakdown in the tissues (Block Reference Block and Friedman1988). If this is true, then higher milk protein concentration for the SBM plus CS diet could be associated with an increased supply of Leu, although mammary uptake of AA was not determined in this study and, based on the conclusion of Mackle et al. (Reference Mackle, Dwyer and Bauman1999), simply providing the mammary gland with more AA does not guarantee increased yield of milk protein as AA utilization by the mammary gland is controlled by mammary gland metabolism rather than nutrient supply.
CONCLUSIONS
Under the conditions of the current experiment, feeding roasted SB treated to maximize RUP supply to the intestine supported higher milk and FCM production in early lactation cows fed lucerne hay and maize silage as the primary forage source. Furthermore, the higher milk and 35 g FCM/kg in cows fed SB plus CSM diets compared with cows fed SBM plus CS showed that CSM can act as a suitable protein source in SB-supplemented diets. Although it was reduced in some studies where dietary forage comprised solely or partly maize silage, there was no significant difference between diets in milk fat concentration in the current experiment, and heat treatment applied to SB provided additional benefits over raw SB.