“fink”. 1.»: 5:? . a; ta 2 . . . 33$“: t? I... ‘ “Wu; . a: v . , .5: 13 > .L r ’9 4 "MP. 13.551” ur- 5 o I u. I I I I 20 30 40 50 60 Preliminary FCMY, kgld 27 CHAPTER 2 Nutrient Demand of Lactating Dairy Cows Affects Feed Intake and Nutrient Utilization Responses to Diets Containing Alfalfa or Orchardgrass ABSTRACT The effect of preliminary feed intake on responses to diets containing alfalfa silage or orchardgrass silage was evaluated using eight ruminally and duodenally cannulated Holstein cows in a crossover design with two 15-d periods. Responses measured were DMI, rates of fiber digestion and passage, and milk production. Cows were 139 a: 83 (mean 2 SD) DIM at the beginning of the preliminary period. During the 14 d preliminary period, milk yield ranged from 24.5 to 46.0 kgld (mean = 37.0 kgld) and preliminary voluntary DMI (pVDMI) ranged from 11.4 to 21.0 kgld (mean = 17.5 kgld). The two treatments were a diet containing alfalfa silage as the sole forage (AL) and a diet containing orchardgrass silage as the sole forage (OG). Alfalfa silage contained 43% neutral detergent fiber (NDF; DM basis) and orchardgrass silage contained 48% NDF; diets contained ~23% forage NDF and 27% total NDF, so forage-to- concentrate ratio was 53:47 for AL and 48:52 for OG. Digestibility of NDF was lower for AL in the rumen and whole tract, and milk fat concentration was greater for 0G than for AL. Mean 3.5% fat-corrected milk yield (FCMY) and DMI were not different between AL and OG, but individual FCMY and DMI responses to AL over 06 were correlated positively with individual pVDMI values. A more positive DMI response to AL over OG among high-pVDMl cows was permitted by 28 a more positive response in ruminal NDF turnover rate for AL over OG as pVDMl increased. This response in NDF turnover rate was because of a differential response in rate of passage rather than digestion; indigestible NDF passage rate response tended to increase with increasing pVDMI, but potentially digestible NDF digestion rate response did not change as pVDMl increased. Therefore, the effects of alfalfa and grass forages on intake, fiber digestion, and milk production depended on the extent to which fill limited intake of an individual cow. INTRODUCTION A meta-analysis of data from experiments using dairy cows demonstrated lower voluntary DMI (VDMI) and milk yield for grass-based diets than for legume- based diets, across maturities, despite greater NDF digestibility for grass (Oba and Allen, 1999b). Although grass NDF usually is more digestible than alfalfa NDF, grass NDF is digested more slowly than alfalfa NDF, and grass cell walls are more resistant to particle breakdown than are alfalfa cell walls (Wilson and Hatfield, 1997). Therefore, we hypothesize that the reduction in VDMI seen for grass-based diets is because of the filling effect caused by slower particle breakdown and slower passage rate in grass forages. However, individual energy balance influences both feed intake responses to forage characteristics and the extent to which physical or metabolic factors limit VDMI (Mertens, 1994; Allen, 1996). The effects on feed intake of diet characteristics (such as forage family) that influence ruminal passage rate of digesta will depend on the extent to which physical filling effects limit feed intake 29 in an individual animal. As a result, testing only overall treatment mean differences may mask important responses in intake, digestion, and production (Allen, 2000). Because cows are now frequently grouped and fed according to milk yield, models that predict the effects of nutrient demand on response to diet are even more necessary. We developed and have successfully used an experimental model to evaluate effects of pVDMl, an index of nutrient demand, on animal responses to dietary treatments (Oba and Allen, 1999a; Burato et al., 2001; Voelker et al., 2002; Harvatine and Allen, 2002; Bradford and Allen, 2004). This model was utilized to test our hypothesis that pVDMI affects individual responses of VDMI and digesta passage rate to diets containing grass silage or alfalfa silage as the sole forage. MATERIALS AND METHODS Cows and Treatments Experimental procedures were approved by the All University Committee on Animal Use and Care at Michigan State University. Eight multiparous Holstein cows from the Michigan State University Dairy Cattle Teaching and Research Center were assigned randomly to treatment sequence in a crossover design experiment with a 14 d preliminary period and two 15 d experimental periods. These eight cows were 138 x 83 (mean :I: SD) DIM at the beginning of the preliminary period (Table 1) and were selected deliberately to provide a wide, uniform distribution of preliminary milk yield and DMI (Figure 1). During the 14 d preliminary period, milk yield ranged from 24.5 to 46.0 kgld (mean = 37.0 kgld) 3O and pVDMl ranged from 11.4 to 21.0 kgld (mean = 18.6 kgld). Cows were cannulated ruminally and duodenally prior to calving. Surgery was performed at the Department of Large Animal Clinical Science, College of Veterinary Medicine, Michigan State University. Cows were housed in tie-stalls and fed once daily (1100 h) at 110% of expected intake. The two treatments were a diet containing alfalfa silage as the sole forage (AL) and a diet containing orchardgrass silage as the sole forage (OG). Alfalfa and orchardgrass forages were raised at the campus farm at Michigan State University (East Lansing) and ensiled in Ag-Bags (Ag-Bag Systems, Inc., St. Nazianz, WI). Alfalfa was harvested at early bud stage, chopped at 3/8" (0.95 cm) theoretical length of cut, and ensiled at 36% DM. Orchardgrass was harvested at early boot stage, chopped at 1/4" (0.64 cm) theoretical length of cut, and ensiled at 37% DM. Cut lengths were selected to yield similar particle size distributions using the Pennsylvania State Particle Size Separator (NASCO, Fort Atkinson, WI). Proportions of fresh-chopped forage retained on the top pan varied greatly by sample batch and averaged 20.1% for alfalfa and 28.1% for orchardgrass. Mean total mass retained on the top and middle pans were similar for alfalfa (62%) and orchardgrass (58%). Nutrient composition for alfalfa silage and orchardgrass silage are shown in Table 2. During the sample collection periods, alfalfa silage contained 43% NDF (DM basis) and orchardgrass silage contained 48% NDF. Diets AL and CG were formulated to contain 23% forage NDF and 27% total NDF, so forage-to- concentrate ratios (DM basis) were 53:47 for AL and 48:52 for 06 (Table 3). 31 The diet fed during the preliminary period was formulated so that alfalfa silage and orchardgrass silage each contributed 50% of forage NDF. Diets also contained dry ground corn, soybean meal (48% CP), expeller-processed soybean meal, a vitamin-mineral premix, and blood meal; limestone, urea, and bloodmeal were used to compensate for lower measured CP and anticipated Ca concentrations in orchardgrass silage than in alfalfa silage. All diets were formulated for 18% dietary CP and fed once daily as totally mixed rations. During the experimental periods, orchardgrass silage CP concentration was similar to alfalfa silage CP concentration, so dietary CP was 0.5% higher on a diet DM basis in OG than in AL. Data and Sample Collection Amounts of feed offered and orts were weighed for each cow daily. Samples of all dietary ingredients (0.5 kg) and orts from each cow (12.5% of orts) were collected daily on d 11 to 13 and combined into one sample per period. Cows were milked twice daily in a milking parlor (0300 and 1500 h); milk yield was measured, and milk was sampled, at each milking on d 11 to 13. Rumen- empty BW was measured after evacuation of ruminal digesta on d 14 of the preliminary period, and on d 15 of each experimental period. Body condition score was determined on the same days by three trained investigators blinded to treatments (WIldman et al., 1982; five-point scale where 1 = thin and 5 = fat). Duodenal samples (1,000 g), fecal samples (500 g), and rumen fluid samples for pH (100 mL) were collected every 9 h from d 11 to d 13 so that eight samples were taken for each cow in each period, representing every 3 h of a 24- 32 hour period to account for diurnal variation. Rumen fluid was obtained by combining digesta from 5 different sites in the rumen and straining it through a layer of nylon mesh (~1 mm pore size). Fluid pH was recorded immediately. All samples were stored at -20°C. Ruminal contents were evacuated manually through the ruminal cannula at 1600 h (5 h after feeding) on d 14 and at 0700 h (4 h before feeding) on d 15 of each period. Total ruminal content mass and volume were determined. During evacuation, 10% aliquots of digesta were separated to allow accurate sampling. Aliquots were squeezed through a nylon screen (1 mm pore size) to separate into primarily solid and liquid phases. Both phases were weighed and sampled (350 mL) for determination of nutrient pool size. All samples were stored at -20°C. Sample and Statistical Analyses Diet ingredients, orts, and feces were dried in a 55°C forced-air oven for 72 h. All samples were ground with a WIley mill (1mm screen; Authur H. Thomas, Philadelphia, PA). Dried, ground fecal samples were combined on an equal DM basis into one sample per cow per period. Frozen duodenal samples for each cow period (n = 8) were chopped into “snow” using a commercial food processor (84142 Food cutter, Hobart Manufacturing 00., Troy, OH) and sub- sampled in the frozen state to obtain representative samples. These duodenal subsamples and the 350 mL ruminal solid and liquid samples were Iyophilized (Tri-Philizer“ MP, FTS Systems, Stone Ridge, NY) and ground as described above. Dried ruminal solid and liquid samples were recombined according to the 33 original ratio of solid and liquid DM. Samples were analyzed for ash, NDF, indigestible NDF (iNDF), CP, and starch. Ash concentration was determined after 5 h oxidation at 500°C in a muffle furnace. Concentrations of NDF were determined according to Van Soest et al. (1991, method A). Indigestible NDF was estimated as NDF residue after 120-h in vitro fermentation (Goering and Van Soest, 1970). Ruminal fluid for the in vitro incubations was collected from a non- pregnant dry cow fed only alfalfa hay. Fraction of potentially digestible NDF (pdNDF) was calculated by difference (1.00 — iNDF). Crude protein was analyzed according to Hach et al. (1987). Starch was measured by an enzymatic method (Karkalas, 1985) after samples were gelatinized with sodium hydroxide. Glucose concentration was measured using a glucose oxidase method (Glucose kit #510; Sigma Chemical Co., St. Louis, MO), and absorbance was determined with a micro-plate reader (SpectraMax 190, Molecular Devices Corp., Sunnyvale, CA). Concentrations of all nutrients except DM were expressed as percentages of DM determined by drying at 105° C in a forced-air oven for more than 8 h. Milk samples were analyzed for fat, true protein, and lactose with infrared spectroscopy by Michigan DHIA (East Lansing). Indigestible NDF was used as an internal marker to estimate nutrient digestibility in the rumen and in the total tract (Cochran et al., 1986), and to estimate rates of passage for iNDF, pdNDF, and starch, and rates of digestion for pdNDF and starch. Nutrient intake was calculated using the composition of feed offered and refused. Ruminal pool sizes (kg) of OM, NDF, iNDF, pdNDF, and starch were determined by multiplying the concentration of each component by 34 the ruminal digesta DM mass (kg). Turnover rate in the rumen, passage rate from the rumen, and ruminal digestion rate of each component (%lh) were calculated by the following equations: Turnover rate in the rumen (%lh) = 100 x (Intake of component! Ruminal pool of component) [24 Passage rate from the rumen (%lh) = 100 x (Duodenal flow of component / Ruminal pool of component) / 24; and Digestion rate in the rumen (%lh) = Turnover rate in the rumen (%lh) - Passage rate from the rumen (%lh). To determine differences between treatments, all data were analyzed using the fit model procedure of JMP® (Version 5.1.2, SAS Institute, Cary, NC) according to the following model: Yijk = M + Ci + Pj 1* Tk 1' PTjk + eijk where I1 = overall mean, C. = random effect of cow ( i = 1 to 8), Pj = fixed effect of period (j = 1 to 2), Tk = fixed effect of treatment (k = 1 to 2), PTjk = interaction of period and treatment, and 9ij = residual, assumed to be normally distributed. To correlate response to treatment with pVDMl, the response (Y) was calculated as follows: 35 Y = YAL — Yoc . where yAL = response for AL diet YoG = response for the OG diet Preliminary VDMI was calculated as the mean of DMI values on d 11 to 14 of the 14-d preliminary period. Relationships between response to treatment and pVDMl were analyzed according to the following model: Yi=u+Si+V+V2+ei Where VI = YAL - Yes u = overall mean, S = effect of sequence (i = 1 to 2), V = pVDMl v2 = pvoMI2 e; = residual, assumed to be normally distributed. Significance was declared at or below P = 0.05, and tendencies were declared at or below P = 0.10. In the pVDMI model, sequence effect (Seq) was removed when P < 0.25 and pVDMI2 effect was removed when P < 0.15. Prediction equations reported are adjusted for Seq (Seq[a] = AL, OG; Seq[b] = OG, AL and was set at 0). The original sample size was 13 cows; data from five cows were excluded from statistical analysis. One cow developed hypocalcemia during the experiment, two were removed from the trial due to duodenal cannula 36 malfunction, one was excluded because feed intake decreased by 50% on d 11 of Period 2 for undetermined reasons (intake slowly returned to normal on the same diet), and one was excluded because several key digestion parameters were outside the 95% confidence interval. None of the causes for removal or exclusion were believed to be associated with either of the two treatments. Among the remaining eight cows, each of the two treatment sequences was represented by four cows. Data in Table 1 and Figure 1 are for the eight cows used in the statistical analysis. RESULTS AND DISCUSSION In Vivo and In Vitro NDF Digestibility Enhanced NDF digestibility of forages usually improves DMI in high- producing dairy cows (Oba and Allen, 1999b). The alfalfa silage used in this experiment was of moderate quality, containing 42.6% NDF and 25.2% iNDF; and the orchardgrass was high-quality, containing 48.0% NDF and 13.1% iNDF. In vitro NDF digestibility (30-h) was much greater for orchardgrass silage (61.1%) than for alfalfa silage (29.4%). Digestibility of NDF was lower for AL than for OG in both the rumen (37.4 vs. 57.1%; P < 0.0001) and the whole gastrointestinal tract (32.8 vs. 55.9%; P < 0.0001; Table 4). This is consistent with previous comparisons of grasses and legumes in lactating cows (Hoffman et al., 1998; Holden et al., 1994; Weiss and Shockey 1991). 37 Intake and Ruminal NDF Kinetics Despite greater NDF digestibility of grass in many studies, VDMI and MY of lactating dairy cows are usually lower for grass-based diets than for legume- based diets (Oba and Allen, 1999b). However, in this experiment, mean DMI (20.4 kgld) and NDF intake (5.4 kgld) were not different between AL and 06 (P > 0.46; Table 4). Feed intake is regulated by a combination of mechanisms, including physical filling effects and metabolic satiety, and the dominant limiting factor varies depending on nutrient demand. These cows represented a wide range in pVDMl, which was used as an estimate of nutrient demand, and their responses to forage family depended on pVDMI. Individual DMI responses to AL over OG (DMIAL - DMIOG) were related positively to individual pVDMI values (Figure 2a; DMIAL - DMIOG = -16.8 + 0.95 x pVDMl; P < 0.01). As pVDMl increased, DMI increased when cows were fed AL (P = 0.05) but not when they were fed OG (P = 0.73). This suggests that the mechanism by which cows are able to increase feed intake was more impaired among cows with higher pVDMI when they were fed OG. In fact, NDF turnover time (TOT) in the rumen decreased more for AL than for OG as pVDMI increased (Figure 2b; TOTAL - TOTOG = 157 + 2.30 x Seq - 1.60 x pVDMl; P < 0.05). The faster disappearance of NDF from AL reduced the physical filling effects for AL more than was possible for NDF from OG. Decreased NDF turnover time may result from an increase in digestion rate and(or) passage rate. In this case, the decreased turnover time was solely because of a differential response in passage rate. Indigestible NDF passage 38 rate response (kpAL — kpoG) tended to increase with increasing pVDMl (Figure 2c; kpAL — kpoc; = -1.44 -0.10 x Seq + 0.10 pVDMI; P = 0.06), but pdNDF digestion rate response (deL - kdoG) did not change as pVDMl increased (P = 0.47). Therefore, among cows with greater drive to eat (greater pVDMI), mechanisms permitting greater passage rate of NDF for AL allowed actual DMI to more closely match demand. Passage rate from the rumen can be increased by increased reticular contractions (Okine and Mathison, 1991), and this does occur with greater ruminal distention (Dado and Allen, 1995). Reticular contractions were not measured, because such measurements may interfere with flow of digesta within and from the reticulorumen (Kaske and Midasch, 1994). Among measures of rumen volume and mass, including NDF pool (kg), rumen pools were similar both between treatments (Table 4) and across the range of pVDMI (data not shown). This suggests that cows were unable to increase rumen pool size in order to allow greater feed intake; that is, physical fill likely was a primary factor limiting feed intake among cows with greater pVDMl. Vlfithout an increase in ruminal pool, the only means to increase feed intake was to increase rates of passage and(or) digestion. As demonstrated, cows with greater pVDMl were able to increase passage rate on AL, but not on OG, in order to allow greater intake. Primary limitations to escape of particles from the rumen are particle size and particle density. Both rate of particle size reduction and rate of increase in particle specific gravity are likely faster in legume forages than in grasses. Particles of legume, and specifically alfalfa, have been demonstrated to be more 39 fragile than particles of grass (Waghorn et al., 1989; Chai et al., 1984). Chewing during eating reduced 61% of alfalfa particulate DM, but only 46% of ryegrass particulate DM, to a size able to pass a 2 mm sieve, and ryegrass particles were cleared more slowly from the rumen (Waghorn et al., 1989). This is likely because of anatomical differences in cell wall structure between temperate grasses and legumes leading to differences in bacterial access to digestible components and the resulting particle shape (Vlfllson and Kennedy, 1996). Although grasses usually contain lower concentrations of lignin and therefore contain more potentially digestible NDF, the lignin generally is dispersed throughout most of the cell wall in both stem and leaf. Furthermore, in grass leaves, veins run the length of the leaf and the tube-shaped cells can be several centimeters long (leon and Kennedy, 1996; Vlfllson and Hatfield, 1997). Therefore, bacterial access to potentially digestible NDF is limited, reducing the rate of NDF digestion and reducing the rate at which particle fragility increases (leon and Kennedy, 1996). Not only does the geometry of Iignification limit bacterial access to the digestible inner cell wall surface, but it also provides fewer natural fracture points and causes grass particles to break into long, narrow particles that are easily trapped within the rumen mat. Although legumes generally contain more lignin than do grasses, that lignin is localized within the xylem and interbundular cells, leaving the mesophyll essentially unlignified. Legume leaves also have shorter, more reticulate veins, so they fracture rapidly into short particles that are less likely to be trapped in the rumen mat (leon and Kennedy, 1996). As a result, 40 the mesophyll of both stem and leaf can be degraded rapidly and completely, increasing particle fragility and quickly producing small particles of indigestible NDF that can escape the rumen more readily (Akin, 1989). Therefore, both rate of NDF digestion and rate of particle size reduction usually are greater, and retention time usually is shorter, in legumes than in grasses (Hoffman et al., 1993; Holden et al., 1994; Waghom et al., 1989). Differences in structure and digestion rate affect not only rate of particle size reduction but also specific gravity of particles. Particle density is decreased by fermentation gasses, so particles with more associated gasses will have longer rumen retention times (Sutherland, 1988). In legumes, once the rapidly- digestible mesophyll is digested, the production rate of fermentation gasses decreases drastically; then the remaining highly lignified tissue, in the form of short, dense particles, will rapidly increase in density and escape the rumen quickly (Allen, 1996; Wilson and Kennedy, 1996). Grass particles, by contrast, have associated fermentation gasses over a longer period of time due to slower digestion and a greater potentially digestible fraction (Allen, 1996). Also, their long, tubular structure might prevent gas from escaping the particle, further reducing the specific gravity and increasing the ruminal retention time of grass particles (\Mlson et al, 1989). Anatomical characteristics that lead to more gas retention, longer particles, and slower NDF digestion combine to explain the greater ruminal filling effects usually observed for grasses compared to legumes. Therefore, grass- based diets have little negative effect on cows that have lower pVDMl and for 41 whom intake is less likely to be limited by fill. Animals with greater pVDMl, however, need to compensate for greater ruminal NDF retention time. These animals could increase chewing when fed grass and thus increase the rate of particle size reduction. Chewing behavior was not measured in this study, and previous comparisons of chewing time for grasses and legumes are rare and have not utilized high-producing dairy cows for whom total chewing time might be a primary limiting factor (Beauchemin and lwaasa, 1993; McLeod et al., 1990). Given the lower fragility of grasses, total chewing time would have to increase greatly in order to bring retention time of grass forage equal to the retention time of legume forage. In the present study, this apparently only occurred among cows with lower pVDMI and not among cows with greater pVDMl. As a result, feed intake of cows with greater pVDMI, for whom ruminal filling effects more often limit feed intake, is much lower for grass-based diets than for alfalfa-based diets. Milk Production Milk yield averaged 29.3 kgld and was similar across treatments (P = 0.77; Table 4). Mean 3.5% fat-corrected milk yield (FCMY) was numerically, but not statistically (P = 0.19) greater when cows were fed OG (33.8 kgld) than when they were fed AL (31.4 kgld; Table 4). This is in contrast with the increase commonly seen in MY or FCMY when legume forage is substituted for grass forage (Oba and Allen, 1999b), and it occurred because milk fat concentration was greater for OG (4.40%) than for AL (3.99%; P = 0.03; Table 4). Milk fat concentration response has varied in previous comparisons of grass- and 42 legume-based diets (Zimmerman et al., 1991; Hoffman et al., 1998; Broderick et al., 2002; Dewhurst et al., 2003a; Al-Mabruk et al., 2004), probably because of differences in forage NDF concentrations, total dietary NDF and diet concentration of forage NDF. Most diets comparing forages are formulated to contain equal forage-to-concentrate ratios, equal total dietary NDF, or equal estimated NEL, or are fed as separate components, all of which eliminate the possibility of directly comparing the specific effects of forage fiber on intake and production parameters. Milk fat concentration is determined by many factors, including the profile of fatty acids (FA) removed from blood by the mammary gland (Bauman and Griinari, 2003). Although the FA concentrations of grass and alfalfa forages are very low, and their FA profiles quite similar (Dewhurst et al., 2003a), faster passage rate for alfalfa-based diets relative to grass-based diets (as discussed above) likely result in greater escape of rumen biohydrogenation intermediates for alfalfa-based diets (Harvatine and Allen, 2006). Milk FA profiles were not measured in this experiment, but Dewhurst et al. (2003a,b) reported greater concentrations of the intermediates of ruminal FA biohydrogenation, such as C132, in milk from cows fed legumes, including alfalfa, compared to milk from cows fed grass. Some biohydrogenation intermediates have inhibited milk fat synthesis (Bauman and Griinari, 2003) and may have caused the reduction in milk fat concentration observed for cows fed alfalfa-based diets. That is, the effect of forage type on milk fat concentration may have been mediated by diet effect on passage rate and ruminal retention time. 43 Just as the effect of diet on passage rate depended on pVDMI, so also the effect of diet on milk fat concentration tended to differ with increasing pVDMI (% FatAL - % Fatoe, = 21.7 - 0.16 x seq - 2.56 x pVDMl + 0.07 x pVDMl2 ; P = 0.07). Previous investigations of the effects of diet on milk fat yield and composition have focused on FA composition of the diet and products of fermentation, and on endocrine responses to diet (Bauman and Griinari, 2003). It is possible that the effects of diet on physical aspects of ruminal digestion and passage may also affect digesta FA profile at the small intestine and therefore influence milk fat synthesis. Because of different responses in passage rate with increasing pVDMI, nutrient demand may also alter the extent to which diet affects the production of milk and its components. The effect of pVDMl on diet utilization was further illustrated by the response of the partitioning of nutrients toward milk production and body tissue accretion. Individual FCMY responses to AL over OG were related to individual pVDMI values (FCMYAL - FCMYOG = 263 — 31.4 pVDMI + 0.90 pVDMIz; P = 0.02; Figure 3a). Similar quadratic relationships with pVDMI were demonstrated for milk yield (P = 0.05) and milk fat percentage (P = 0.07). The quadratic response of F CMY suggests that different factors controlled responses to diet of nutrient partitioning among cows with different pVDMl. Milk yield generally is correlated with DMI (NRC, 2001). Among cows with greater pVDMI, the increase in DMI response to AL with increasing pVDMI resulted in increased 3.5% FCMY on AL compared to OG, but among cows with low to moderate pVDMI, the smaller increase in DMI for AL resulted in similar or lower 3.5% FCMY on AL compared 44 to OG. Cows may have used additional nutrients obtained from slightly greater DMI on AL to replenish body tissue rather than to increase milk production. This is supported by the response of BCS change, which was the opposite of the FCMY response (ABCSAL - ABCSOG = -15.9 + 1.9 pVDMI - 0.05 pVDMIz, P < 0.01; Figure 3b). Blood metabolites and hormones were not measured, so the variation in endocrine response to diet across pVDMl could not be determined. However, it is apparent that the changes in ruminal filling effects and in passage rate from the rumen caused by differences in forage fiber digestion had different effects on nutrient utilization depending on the pVDMI of the individual cow. SUMMARY As hypothesized, DMI on AL became increasingly greater than DMI on OG with greater pVDMl. This occurred because NDF turnover time in the rumen decreased more for AL than for OG as pVDMl increased. The faster disappearance of NDF on diet AL, caused primarily by a greater increase in passage rate of iNDF on AL with increasing pVDMl, reduced the physical filling effects for AL more than was possible for NDF from diet OG. This likely was caused by differences in both rate of particle size reduction and rate of increase in particle specific gravity, which have been demonstrated to be faster in legume forages than in grass forages. Through its effect on passage rate responses, pVDMl also altered the extent to which diet affected the production of milk and its components. Individual milk fat concentration, FCMY and BCS responses to AL over OG were related to individual pVDMl values. 45 CONCLUSIONS AND IMPLICATIONS Cows with the greatest drive to eat, as estimated by pVDMl, responded the most positively in feed intake and milk production to alfalfa versus orchardgrass as the primary dietary fiber source. These results corroborate previous research suggesting that intake is more limited by physical fill effects with increasing nutrient demand and on grass forages compared to legume forages. Many models of feed intake, digestion, and metabolism in dairy cows may be improved by incorporating the quantified effects of nutrient demand and feed sources on feed intake and passage rate, which can be provided by this experiment and future experiments testing other important variations in diet characteristics. Finally, the results of this experiment reinforce the need to provide separate diets for cows with higher and lower nutrient demand, in order to maximize the efficiency of nutrient utilization among the whole herd. 46 Table 1. Status of eight cows during the final 4 d of the preliminary period, when cows were fed a common diet. Parameter Mean Standard Deviation Parity 4.0 2.6 BW, kg 538 17 BCS 2.5 0.4 DIM 139 83 Milk yield, kg/d 40.1 5.5 DMI, kgld 18.6 2.8 Figure 1. Distribution of DMI and 3.5% fat-corrected milk yield of eight cows during the final 4 d of the preliminary period, when cows were fed a common diet. Preliminary VDMI, kgld Preliminary 3.5% FCMY, kgld 22 50 45 20 40 18 35 30 16 25 14 20 . 1 2 3 1 2 3 Number of Cows Number of Cows Table 2. Chemical characterization of alfalfa silage and orchardgrass silage. Alfalfa Silage Orchardgrass Silage DM (% as fed) 30.6 35.3 Nutrient, % DM OM 88.7 89.2 NDF 42.6 48.0 Indigestible NDF 25.2 13.1 Potentially digestible NDF 17.4 34.9 Starch 4.0 2.3 Crude protein 20.5 20.4 30-h in vitro NDF digestibility, % 29.4 61.1 47 Table 3. Ingredient and nutrient composition of treatment diets, one diet (AL) containingalfalfa silage and another diet (OG) containing orchardgrass silage. AL Ingredient % of DM Alfalfa silage 53.0 ----- Orchardgrass silage ----- 47.9 Dry ground corn 36.3 40.3 Soybean meal (48% CP) 6.5 7.0 VItamin mineral mix1 4.2 4.2 Expeller—processed soybean meal2 1.3 1.3 Bloodmeal 0.3 0.9 Limestone ----- 0.4 Urea ------ 0.2 Nutrient DM (% as fed) 43.6 50.6 % of DM OM 91.5 91.5 NDF 26.7 27.5 Forage NDF 22.5 23.0 Indigestible NDF 14.8 7.9 Potentially digestible NDF 11.9 19.7 Starch 30.2 32.1 Crude protein 18.3 18.8 Rumen-undegraded CP3 5.6 6.3 1 VItamin mineral mix contained (DM basis) 11.7% dicalcium phosphate, 11.1% trace-mineral premix, 8.8% sodium bicarbonate, 2.3% magnesium oxide, 134.3 KIU/kg vitamin A, 35.53 KlU/kg vitamin D, 895.5 KIU/kg vitamin E, and 65.2% ground corn grain as a carrier. Nutrient composition: 86% DM, 7% ash, 16% NDF, 5% starch, 51% CP. 3 Estimated using values from NRC (2001). 48 Table 4. Least-squares means of feed intake, digestion, and production responses in response to diets containing alfalfa QAL) or orchardgrass (OG). Treatment LSM1 AL 06 SEM2 P Yield, kgld Milk 29.1 29.4 2.6 0.77 3.5% fat-corrected milk 31.4 33.8 3.3 0.19 Fat 1.17 1.31 0.14 0.13 Milk composition, % Fat 3.99 4.40 0.16 0.03 BW change, kg/15 d -1.2 -17.5 5.6 0.04 BCS change, I15 d -0.04 -0.16 0.08 0.23 Intake, kg DM 21.3 20.4 1.3 0.46 NDF 5.4 5.3 0.3 0.69 iNDF3 3.0 1.6 0.1 < 0.0001 Forage NDF 4.7 4.6 0.3 0.67 Rumen Pool, kg DM 10.3 11.3 0.8 0.31 NDF 5.6 5.1 0.4 0.24 iNDF 4.4 2.7 0.2 < 0.0001 Passage rate from rumen, hr’1 iNDF 2.9 2.4 0.2 0.01 pdNDF4 1.2 1.3 0.4 0.81 Starch 11.4 12.3 1.8 0.70 Ruminal digestion rate, hr’1 pdNDF 6.9 5.2 0.6 0.06 Starch 21.1 18.3 3.2 0.42 NDF digested in the rumen kg 2.0 3.0 0.2 < 0.01 % 37.4 57.1 2.6 < 0.001 NDF digested in the whole tract kg 1.8 3.0 0.2 < 0.001 % 32.8 55.9 1.5 < 0.0001 ‘ Treatment least-squares means. 2 Standard error of the mean. 3 Indigestible NDF. 4 Potentially digestible NDF = NDF — iNDF. 49 Figure 2. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) of (A) DMI (DMIAL — DMIOG = -16.8 + 0.95 pVDMI), (B) ruminal NDF turnover time (TOTAL - TOTOG = 157 + 2.30 seq - 1.60 pVDMI), and (C) iNDF ruminal passage rate (kpAL - kaG = -1.44 -0.10 seq + 0.10 pVDMl). Equations B and C include the effect of treatment sequence (seq). A 6 P < 0.01 o R2 = 0.74 . 2 4 - RMSE = 1.72 at x 6' In C O Q III 0 n: E D ‘4 l l I l l l l I 13 14 15 16 17 18 19 20 21 pVDMI, kgld B 20 I, P = 0.03 : R2 = 0.65 15 1 RMSE = - 4.08 Ruminal NDF Turnover TIme Response, hr 14 15 16 17 1a 19 20 21 22 pVDMI, kgld 50 O iNDF Ruminal Passage Rate Response, hr1 1.0 .P=006 . 0.37 R2 = 0.63 l 14 is 16 i7 is 19 £0 5H 22 pVDMI, kgld 51 Figure 3. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) 3.5% FCM yield (FCMYAL - FCMYOG = 263 - 31.4 pVDMl + 0.90 pVDMl:) and (B) change in BCS (ABCSAL - ABCSOG = -15.9 + 1.9 pVDMI -- 0.05 pVDMI ). P=002 6' R2=070 . RMSE= 4 946 2-I 0", oooooooooooooooooooooooooooooooooooooooooooooo on .2" -4" e .5- ’ 3.5% FCMY Response, kgld > 14 15 16 17 18 19 20 21 22 pVDMI, kgld 52 BCS Change Response, I15d 0.8 0.7" 0.6‘ 0.5' 0.4' 0.3' 0.2 1 0.1 - P<0£1 R2 = 0.82 RMSE = 0.14 0.041 . . . . 14 15 16 17 18 pVDMI, kgld 19 20 53 22 CHAPTER 3 Nutrient Demand of Lactating Dairy Cows Affects Nitrogen Intake and Utilization Responses to Diets Containing Alfalfa or Orchardgrass ABSTRACT The effect of preliminary feed intake on responses to diets containing alfalfa silage or orchardgrass silage was evaluated using eight ruminally and duodenally cannulated Holstein cows in a crossover design with two 15—d periods. Responses measured were N intake, digestion, and utilization. Cows were 139 :I: 83 (mean 1 SD) DIM at the beginning of the preliminary period. During the 14 d preliminary period, milk yield ranged from 24.5 to 46.0 kgld (mean = 37.0 kgld) and preliminary voluntary DMI (pVDMI) ranged from 11.4 to 21.0 kgld (mean = 17.5 kgld). Treatments were a diet with alfalfa silage as the sole forage (AL) and a diet with orchardgrass silage as the sole forage (OG). Alfalfa silage contained 20.5% CF (DM basis) and orchardgrass silage contained 20.4% CF; AL contained 18.3% CP and 5.6 estimated rumen-undegraded CP, and OG contained 18.8% CP and 6.3% estimated rumen-undegraded CP. Mean N intake was similar between treatments (P = 0.95), ruminal N digestibility was greater (P = 0.03) for AL (30.4%) than for OG (17.7%), and whole-tract N digestibility did not differ between treatments (P = 0.50). With increasing pVDMI, intake and duodenal flow of N increased more for AL than for OG because of increasingly greater DMI for AL compared to OG. However, among cows with greater pVDMI, a decreasing proportion of the additional N consumed from AL 54 was digested and used for increased milk production or body tissue gain. Although feeding diets containing alfalfa instead of orchardgrass can increase yields of milk and milk protein among cows with greater pVDMI, increasing N intake at the same rate as DMI likely will lead to less efficient utilization of dietary N for production of microbial protein, muscle, or milk protein. When feeding less- filling diets, such as those containing large proportions of legume forage, to high- producing cows, reducing dietary N concentration could increase the efficiency of N utilization and reduce the extent to which greater DMI leads to greater N excretion. INTRODUCTION Although alfalfa generally is considered to have a higher nutritional value than grass because of its higher crude protein and lower fiber concentrations, the addition of grass to the forage component of a diet for dairy cows can increase the efficiency of alfalfa N use. Using alfalfa alone to meet requirements for forage fiber often results in excess dietary N in a form that is degraded rapidly to ammonia in the rumen. Grasses reduce the ratio of N to fiber in forage (Spandl and Hesterman, 1997) and therefore can reduce fecal and urinary N waste excreted by cows. Grass fiber and alfalfa fiber also have different chemical compositions, physical characteristics, and digestion characteristics that affect both the rate and extent of their digestion (Allen, 1996). Because grass fiber generally contains less lignin than alfalfa at the same maturity, grass fiber is ultimately more 55 digestible. However, grass fiber also is digested more slowly, and its cell walls break down more slowly than alfalfa cell walls. Passage rate increases as density increases and as particle size decreases, so fibrous alfalfa particles might escape the rumen more quickly than fibrous grass particles (Allen, 1996). Faster escape of bacteria from the rumen (shorter residence time) increases efficiency of N and energy utilization (Oba and Allen, 2003c; Voelker and Allen, 2003b) by decreasing bacterial death and breakdown in the rumen (lsaacson et al., 1975; Stouthamer and Bettenhaussen, 1973; Kennedy and Milligan, 1978). Microbial protein flow to the duodenum is limited by the availability of readily fermented feed for growth and by the ability of bacteria to avoid lysis and escape the rumen. Therefore, increasing passage rate of particles and bacteria from the rumen should cause increased microbial protein flow to the duodenum and increased efficiency of microbial protein synthesis. Thus, adding grass to a legume forage would likely reduce passage rate and lower microbial protein efficiency. Finally, the passage rate of fiber from the rumen also depends on nutrient demand (Voelker et al., 2002). We developed and have successfully used an experimental model to evaluate effects of indices of nutrient demand, such as preliminary milk yield, on responses to dietary treatments (Oba and Allen, 1999a; Burato et al., 2001; Voelker et al., 2002; Harvatine and Allen, 2005; Bradford and Allen, 2004). This model was utilized to test our hypothesis that preliminary VDMI (pVDMI) affects individual responses of N intake, digestion, and utilization to diets containing orchardgrass silage or alfalfa silage as the sole forage. 56 MATERIALS AND METHODS Cows and Treatments Experimental procedures were approved by the All University Committee on Animal Use and Care at Michigan State University. Eight multiparous Holstein cows (139 a: 83 DIM; mean :1: SD; Table 1) from the Michigan State University Dairy Cattle Teaching and Research Center were assigned randomly to treatment sequence in a crossover design experiment with a 14 d preliminary period and two 15 d experimental periods. These eight cows were selected. deliberately to provide a wide, uniform distribution of milk yield and DMI (Figure 1). During the 14 d preliminary period, milk yield ranged from 24.5 to 46.0 kgld (mean = 37.0 kgld) and preliminary voluntary DMI (pVDMI) ranged from 11.4 to 21.0 kgld (mean = 18.6 kgld). Cows were cannulated ruminally and duodenally prior to calving. Surgery was performed at the Department of Large Animal Clinical Science, College of Veterinary Medicine, Michigan State University. Cows were housed in tie-stalls and fed once daily (1100 h) at 110% of expected intake. Treatments were a diet with alfalfa silage as the sole forage (AL) and a diet with orchardgrass silage as the sole forage (OG). Alfalfa and orchardgrass forages were grown at the campus farm at Michigan State University (East Lansing) and ensiled in Ag-Bags (Ag-Bag Systems, Inc., St. Nazianz, WI). Alfalfa was harvested at bud stage, chopped at 3/8” (0.95 cm) theoretical cut length, and ensiled at 36% DM. Orchardgrass was harvested at early boot stage, chopped at 1/4" (0.64 cm) theoretical cut length, and ensiled at 37% DM. 57 Theoretical cut lengths were selected to yield similar particle size distributions using the Pennsylvania State Particle Size Separator (NASCO, Fort Atkinson, WI). Proportions of fresh-chopped forage retained on the top pan varied greatly by sample batch and averaged 20.1 % for alfalfa and 28.1% for orchardgrass. Mean total mass retained on the top and middle pans were similar for alfalfa (62%) and orchardgrass (58%). Nutrient composition for alfalfa silage and orchardgrass silage are presented in Table 2. During the sample collection periods, alfalfa silage contained 43% NDF (DM basis) and orchardgrass silage contained 48% NDF. Diets AL and 0G were formulated to contain 23% forage NDF and 27% total NDF, so forage-to-concentrate ratios (DM basis) were 53:47 for AL and 48:52 for 0G (Table 3). The diet fed during the preliminary period was formulated so that alfalfa silage and orchardgrass silage each contributed 50% of total forage NDF. Diets also contained dry ground corn, soybean meal, a vitamin-mineral premix, and blood meal; limestone, urea, and bloodmeal were used to compensate for greater measured CP and anticipated Ca concentrations in alfalfa silage than in orchardgrass silage. All diets were formulated for 18% dietary CP and fed once daily as totally mixed rations. During the experimental periods, orchardgrass silage CP concentration was similar to alfalfa silage CP concentration, so total dietary CP was 0.5% higher in OG than in AL. Data and Sample Collection Amounts of feed offered and orts were weighed for each cow daily. Samples of all dietary ingredients (0.5 kg) and orts from each cow (12.5% of orts) 58 were collected daily on d 11 to 13 and combined into one sample per period. Cows were milked twice daily (0300 and 1500 h) in a milking parlor; milk yield was measured, and milk was sampled, at each milking on d 11 to 13. Rumen- empty BW was measured after evacuation of ruminal digesta on d 14 of the preliminary period, and on d 15 of each experimental period. Body condition score was determined on the same days as BW, by three trained investigators blinded to treatments (WIIdman et al., 1982; five-point scale where 1 = thin and 5 = fat). Duodenal samples (1,000 g), fecal samples (500 g), and rumen fluid samples for microbial isolation (350 mL) were collected every 9 h from d 11 to d 13 so that eight samples were taken for each cow in each period, representing every 3 h of a 24-hour period in order to account for diurnal variation. Rumen fluid for microbial isolation was collected from the reticulum, near the reticular- omasal orifice, and strained. All samples were stored immediately at -20°C. Ruminal contents were evacuated manually through the ruminal cannula at 1600 h (5 h after feeding) on d 14 and at 0700 h (4 h before feeding) on d 15 of each period. Total ruminal content mass and volume were determined. During evacuation, 10% aliquots of digesta were separated to allow accurate sampling. Aliquots were squeezed through a nylon screen (1 mm pore size) to separate into primarily solid and liquid phases. Samples (350 mL) were taken from both phases for determination of nutrient pool size. Samples were stored immediately at —20°C. 59 Sample and Statistical Analyses Diet ingredients, orts, and feces were dried in a 55°C forced-air oven for 72 h and analyzed for DM concentration. All samples were ground with a WIley mill (1mm screen; Authur H. Thomas, Philadelphia, PA). Dried, ground fecal samples were combined on an equal DM basis into one sample per cow per period. Frozen duodenal samples for each cow period (n = 8) were chopped into “snow” using a commercial food processor (84142 Food cutter, Hobart Manufacturing Co., Troy, OH) and sub-sampled in the frozen state to obtain representative samples. These duodenal subsamples and the 350 mL ruminal solid and liquid samples were Iyophilized (T ri-Philizer" MP, FTS Systems, Stone Ridge, NY) and ground with a Wiley mill as above. Dried and ground ruminal solid and liquid samples were recombined according to the original ratio of solid and liquid DM. Samples were analyzed for ash, NDF, indigestible NDF (iNDF), and starch, as described elsewhere (Chapter 2). Crude protein concentrations were determined according to Hach et al. (1987). Concentrations of all nutrients except DM were expressed as percentages of DM determined by drying at 105° C in a forced-air oven for more than 8 h. Duodenal digesta were analyzed for purines and ammonia to estimate microbial N (MN) flow and non-ammonia non-microbial N (NANMN) flow to the duodenum. Purine concentration was used as a microbial marker, and purine to microbial N ratio was estimated by analysis of microbial pellets obtained by differential centrifugation of the rumen fluid collected in the reticulum. Total purines were measured by spectrophotometer (Beckman Instruments, Inc., 60 Fullerton, CA) at 260 nm (Zinn and Owens, 1986). Ammonia concentration was determined for centrifuged duodenal and rumen fluid samples according to Broderick and Kang (1980). Milk samples were analyzed for true protein with infrared spectroscopy by Michigan DHIA (East Lansing). Milk true protein N yield was calculated as milk true protein yield [6.38 (Jenness, 1985), and intake N was calculated as DMI x dietary N concentration. Milk samples from the first experimental period only were analyzed for milk urea N (MUN) with infrared spectroscopy by Michigan DHIA (East Lansing); therefore a t-test was used to determine the difference between treatments. Indigestible NDF was used as an internal marker to estimate duodenal flow of nutrients in order to calculate nutrient digestibility in the rumen and in the total tract (Cochran et al., 1986), and to estimate passage rates of passage for iNDF, pdNDF, and starch, and rates of digestion for pdNDF and starch. Nutrient intake was calculated using the composition of feed offered and refused. Duodenal flow of microbial OM was determined as described by Oba and Allen (2003b), and true ruminally degraded OM (T RDOM) was calculated by subtracting duodenal flow of non-microbial OM from OM intake. Ruminal pool sizes (kg), turnover time in the rumen (h), passage rate from the rumen (h'1), and ruminal digestion rate of each component (h'1) were calculated as described elsewhere (Chapter 2). To determine differences between treatments, all data were analyzed using the fit model procedure of JMP® (Version 5.1.2, SAS Institute, Cary, NC) according to the following model: 61 Yijk = l1l 1' Ci 1' Pj 4' Tk 1' PTjk 'I' eijk where u = overall mean, 0; = random effect of cow ( i = 1 to 8), P] = fixed effect of period (j = 1 to 2), TI. = fixed effect of treatment (k = 1 to 2), PTjk = interaction of period and treatment, and em, = residual, assumed to be normally distributed. To determine the dependence of response to treatment on pVDMI, the response (AY) was calculated as follows: AY = YAI. - Yoc where YAL = response for AL diet, and Y06 = response for the CG diet. Preliminary VDMI was calculated as the mean of DMI values on cl 11 to 14 of the 14-d preliminary period. Relationships between response to treatment and pVDMl were analyzed according to the following model: Yi=u+Sj+V+V2+ei Where Y1: YAL - Yoc ii = overall mean, S, = effect of sequence (j = 1 to 2), v = pVDMl, 62 v2 = pVDMIz, and ei = residual, assumed to be normally distributed. Significance was declared at or below P = 0.05, and trends were declared at or below P = 0.10. In the pVDMI model, sequence effect was removed when P > 0.25 and pVDMl2 effect was removed when P > 0.15. Prediction equations reported are adjusted for Sequence (Seq[a] = AL, OG; Seq[b] = OG, AL and was set at 0) for N intake, rumen N turnover time, N digested in the rumen, ruminal N digestibility, N digested in the whole tract, whole-tract N digestibility, and 9 milk N! g N intake. The original sample size was 13 cows; data from five cows were excluded from statistical analysis. One cow developed hypocalcemia during the experiment, two were removed from the trial due to duodenal cannula malfunction, one was excluded because feed intake decreased by 50% on d 11 of Period 2 for undetermined reasons (intake slowly returned to normal on the same diet), and one was excluded because several key digestion parameters were outside the 95% confidence interval. None of the causes for removal or exclusion were believed to be associated with either of the two treatments. Among the remaining eight cows, each of the two treatment sequences was represented by four cows. Data in Table 1 and Figure 1 are for the eight cows used in the statistical analysis. 63 RESULTS AND DISCUSSION Nitrogen Digestion: AL versus OG Diet CP and estimated rumen-undegraded CP concentrations were similar for OG and AL (Table 3). Because mean DMI was similar for the two treatments (P > 0.45) and dietary N concentrations were similar, N intake was not affected by treatment (P = 0.95, Table 4). Ruminal N pool was greater for OG than for AL (P = 0.01), and ruminal turnover time of N was longer on OG than on AL (P < 0.01; Table 4). This was probably the result of a slower passage rate of digesta from the rumen on OG as indicated by the slower passage rate of iNDF (Chapter 2) and possibly because of greater proteolysis on AL. The two silages contained similar concentrations of N, but protein from legumes usually is degraded more rapidly and extensively in the rumen than is protein from grasses (Kwakkel et al., 1986; Kohn and Allen, 1995). Consistent with greater proteolysis on AL, ruminal ammonia concentration was much greater for AL (29.3 mg/dl) than for OG (18.0 mg/dL; P < 0.001). This helps to explain the greater turnover time of N on OG. The effect of forage type on ammonia production rate likely was even greater than the difference in ruminal ammonia concentration, because rate of absorption likely was greater for AL. Mean pH was higher for AL than for 0G (P < 0.001; Table 4), suggesting that rate of ruminal ammonia absorption was greater for AL than for OG for two reasons. First, decreasing pH causes a slower rate of absorption because more ammonia is converted to ammonium, which is not absorbed. Furthermore, lower 64 pH may inhibit rumen motility (Allen et al., 2006), which also reduces rate of ammonia absorption. Rates of ammonia production, utilization by microbes, absorption from the rumen, and recycling were not measured. However, the higher concentration of ammonia on AL apparently promoted the absorption of more N from the rumen as ammonia, as indicated by a greater digestibility of N in the rumen (P = 0.03) and a greater amount of N disappearing from the rumen (P < 0.01) for AL than for OG (Table 4). Although more N was absorbed in the rumen for AL, compensatory postruminal digestion of N occurred for OG, so that in the whole tract, digestibility of N and the amount of N absorbed were similar on both treatments (P > 0.50; Table 4). Because the cows were not ileally cannulated, N disappearance from the small intestine and in the large intestine cannot be differentiated. Nitrogen Digestion: Effects of pVDMI Forage type was not the only factor that affected the site of N digestion. The nutrient demand of individual cows, as estimated by their pVDMl, also interacted with forage source to affect N digestion and utilization. As pVDMl increased, N intake was increasingly greater for AL over OG (P = 0.01, R2 = 0.83, RMSE = 2.19; Figure 2a). As reported in Chapter 2, when cows were fed AL, DMI increased as pVDMI increased (P = 0.05); but when cows were fed OG, DMI was similar for all cows regardless of pVDMI (P = 0.73). Therefore, because of greater ruminal fill effects of OG, N intake likely was limited on OG relative to AL among cows with greater pVDMl. This was caused primarily by restriction of 65 digesta passage; as pVDMI increased, passage rate of iNDF increased for AL but not for 0G (Chapter 2). Increased passage rate with increasing pVDMl combined with the likely greater ruminal degradation of nitrogenous compounds on AL, so that N turnover time in the rumen was not only shorter on AL than on 0G but also decreased on AL relative to 0G among cows with greater pVDMI (Figure 2b). That is, the increase in passage rate seen for AL with increasing pVDMl permitted shorter turnover time and greater intake of N, along with other nutrients, for cows with greater pVDMI. For OG, the inability to increase passage rate nearly eliminated the ability of cows with greater pVDMl to reduce turnover time and increase intake of N and other nutrients in order to meet nutrient demands. Furthermore, forage source and pVDMl also interacted in their effects on ruminal absorption of N and microbial protein production. The difference in individual responses to diet (yAL - Yos) of N apparently digested and absorbed in the rumen depended on pVDMl in a quadratic relationship (P = 0.05; R2 = 0.91; RMSE = 86.1; Figure 3a). The difference between ruminal N digestibility on AL and OG also demonstrated a tendency for a similar quadratic dependence on pVDMl (P = 0.08; R2 = 0.83; RMSE = 5.69; Figure 3b). Therefore, the linear increase of N intake (Figure 2a) and the quadratic response of ruminal N digestibility (Figure 3b) for AL compared to OG with Increasing pVDMI led to the quadratic response in the quantity of N absorbed in the rumen (Figure 3a). Among cows with lower pVDMI, a greater N absorption rate on AL relative to OG compensated for a lower N intake on AL relative to OG. As pVDMI increased 66 from moderate to high values, the increasing N intake on AL relative to OG and the slightly Increasing ruminal digestibility on AL relative to OG resulted in a sharp increase in the quantity of N absorbed in the rumen on AL relative to OG. Mechanisms could include changes in the rate or extent of protein degradation, in the rate of absorption of ammonia from the rumen, in the rate or extent of incorporation of N into microbial protein, or in the rate of passage of N-containing digesta from the rumen. None of these potential mechanisms were measured directly. The difference in ruminal ammonia concentrations for cows fed the two diets did not depend on pVDMl (P > 0.40), but this does not eliminate the possibility of changes in rate of ammonia production or absorption. As mentioned earlier, lower ruminal pH can reduce absorption rate of ammonia. The difference in response of daily mean ruminal pH to treatment (pHAL - pHOG) tended to depend on pVDMl (P = 0.07; R2 = 0.53; RMSE = 0.31; data not shown). When cows were fed AL, mean pH was between 6.1 to 6.6 regardless of pVDMl (P = 0.55). When cows were fed OG, mean pH tended to demonstrate a quadratic relationship to pVDMI, with a maximum at 6.2 (P = 0.09); pH was particularly lower, around 5.8, among cows with high pVDMI when they were fed OG. This lower ruminal pH may have reduced the rate of ammonia absorption from the rumen among cows with high pVDMI when they were fed OG. Therefore, diet effects and pVDMI effects on ruminal ammonia production and absorption may have contributed to the observed responses of ruminal N digestion. 67 Reflecting N intake, flow to the duodenum of total N for AL relative to OG increased linearly as pVDMl increased (P = 0.01, R2= 0.69; RMSE = 82.2; Figure 4a); flow of NAN responded similarly (P = 0.01; R2 = 0.69; RMSE = 79.6; Figure 4b). Flow of MN to the duodenum also increased for AL relative to OG as pVDMI increased (P = 0.05; R2 = 0.51; RMSE = 82.9; Figure 5a). This resulted from a numerical increase in MN flow with greater pVDMl (P = 0.15) when cows were fed AL and a numerical decrease in MN flow with greater pVDMl (P = 0.14) when cows were fed OG. Greater duodenal flow of MN for AL with increasing pVDMl reflects the greater DM and N intake and passage rate observed for AL compared to OG with greater pVDMl. Response of efficiency of microbial protein production from truly ruminally degraded OM (MNE) demonstrated a quadratic relationship to pVDMI (P = 0.03; R2 = 0.76; RMSE = 0.56; Figure 5b). When cows were fed OG, the production of microbial N from the ruminal digestion of OM decreased linearly with increasing pVDMl (P = 0.05), suggesting that fermentation and microbial growth were increasingly uncoupled on OG as pVDMI increased. Ruminal starch digestion likely played a significant role in determining the extent to which N was incorporated into MN. The relationship between pVDMI and the difference in ruminal starch digestion on the two diets (Figure 6a,b) was the opposite of the relationship between pVDMI and the response in MNE (Figure 5b). The amount of starch truly ruminally digested (TRD starch, kgld) and true ruminal starch digestibility (TRSD, % of intake) demonstrated quadratic relationships between pVDMI and relative response to AL versus OG (P = 0.05 and P = 0.04, for TRD 68 starch (Figure 6a) and TRSD (Figure 6b), respectively). Opposite relationships to pVDMI for MNE and ruminal starch digestion indicate that among cows with low and high pVDMl, for whom ruminal starch digestion was greater on AL than on OG, MNE was lower on AL than on OG; the reverse was true for cows with moderate pVDMI. Across all 16 cow-period values, MNE was correlated negatively with true ruminal starch digestibility (P = 0.02, R2 = 0.32), and MN flow to the duodenum (gld) tended to be correlated negatively with true ruminal starch digestibility (P = 0.07, R2 = 0.22). In general, greater ruminal starch digestion did not result in greater or more efficient production of MN. Rather, it may have reduced efficiency of utilization of N and OM for MN production, probably by uncoupling fermentation and microbial growth (Strobel and Russell, 1986) and(or) by increasing the population of amylolytic bacteria, some of which are also very actively proteolytic (Russell et al., 1981). The negative effect of greater starch digestion on MNE was apparently the greatest among cows with high pVDMI when they consumed diet AL, where the greatest starch intake and ruminal starch digestion were observed. When cows were fed OG, ruminal starch digestion was not affected by pVDMI (P > 0.25) so differences in ruminal starch digestion probably did not cause the decrease in MN production and efficiency observed for OG with increasing pVDMl. The less efficient production of microbial protein on 0G as pVDMl increased was caused, in part, by the increasingly negative effect of 06 on passage rate as pVDMl increased. As demonstrated earlier (Chapter 2), 69 passage rate of iNDF tended to be affected negatively by OG compared to AL as pVDMI increased. Decreasing passage rate can decrease the efficiency of N utilization for microbial protein production. Vlfith greater passage rate, microbes associated with particulate digesta can escape the rumen more rapidly, reducing microbial protein turnover by reducing the extent of autolysis (Wells and Russell, 1996) and protozoal predation (Wallace and McPherson, 1987). The efficiency with which N was incorporated into MN, and the turnover of MN in the rumen, likely contributed to the observed responses of ruminal N digestion for 06 compared to AL. Just as N intake and duodenal N flow increased on AL relative to OG with increasing pVDMI, the amount of N digested postruminally also increased on AL compared to OG with greater pVDMl (P = 0.01; R2 = 0.67; RMSE = 67.7; Figure 7a), as did the amount of NAN digested postruminally (P = 0.01; R2= 0.67; RMSE = 66.2; Figure 7b). The amount of N digested in the whole tract also increased on AL relative to OG with increasing pVDMl (P = 0.02; R2 = 0.89; RMSE = 42.2; Figure 8a). However, whole-tract N digestibility tended to become increasingly lower on AL relative to OG with increasing pVDMl (P = 0.07; R2 = 0.98; RMSE = 1.35; Figure 8b). Vlflth increasing pVDMl, AL permitted increased DMI and N intake, and increased MN production. However, the decreases for AL compared to OG in microbial efficiency (MN, %TRDOM) and whole-tract N digestibility among cows with the highest pVDMl suggest that the efficiency of N utilization did not benefit from the increasingly greater intake and passage rate observed 70 for AL relative to OG with increasing pVDMI, because N supply probably was in excess of requirements, as discussed below. Nitrogen in Milk Production The form in which dietary N was absorbed might have affected its proportion and form in milk. Mean yield and concentration of true protein in milk did not differ across treatments (Table 4). Although yields of milk and true protein, and milk true protein concentration, were similar between treatments, more N was secreted in milk in the form of urea for AL than for OG (Table 4). Milk urea N was measured only during one period (n = 8), but MUN was much greater (P < 0.01) cows fed AL (23.5 mgldL) than for cows fed OG (15.3 mg/dL). This is consistent with the greater disappearance of N from the rumen for AL compared to OG, probably as ammonia which is used to synthesize urea in the liver. Ruminal ammonia concentration and MUN were highly correlated (P < 0.01; R2 = 0.84). The effect of diet on the yield of true protein in milk varied and tended to depend on pVDMI (P < 0.10; R2 = 0.66; RMSE = 0.09; Figure 93). Among the cows with lower and moderate pVDMl, true protein yield was similar or lower on AL compared to OG, but among cows with the highest pVDMI, true protein yield increased on AL relative to OG as pVDMl increased (Figure 9a). Mean efficiency of utilization of N consumed in the diet for synthesis of true protein in milk (9 milk true protein N lg N intake) did not differ between treatments (Table 4). However, with increasing pVDMl, N tended to be used less efficiently for milk protein production on AL compared to 06 (P = 0.10; R2 = 0.49; RMSE = 0.02; Figure 9b) 71 and might have been secreted increasingly as MUN instead of as true protein on AL, with increasing pVDMI. This was likely caused, at least in part, by the increase in N intake and decrease in whole-tract N digestibility observed on AL compared to GR with increasing pVDMl. In addition, this apparent decrease in efficiency of N utilization on AL with increasing pVDMI occurred despite the expected dilution of maintenance N with increasingly greater MY on AL as pVDMl increased. Apparent efficiency of milk protein production from dietary N also can be increased through mobilization of body tissue protein to meet the demand for milk production. The relationship between pVDMl and response in estimated NEL balance (P = 0.03; R2 = 0.52; RMSE = 3.71; Figure 9d) supports tissue mobilization as a mechanism for increased apparent efficiency of milk true protein synthesis, but the relationship between pVDMl and response in BCS change (P < 0.01; R2 = 0.82; RMSE = 0.142; Figure 90) does not. Change in BCS is a direct measurement and NEL balance is an indirect estimate, so it is unlikely that cows actually experienced increasingly positive body tissuegains on AL compared to 0G with increasing pVDMl as is suggested by Figure 9d. Therefore, increased N intake with increasing pVDMl (Figure 2a) and decreasing N digestibility (Figure 8b) likely were primary factors in decreasing the efficiency of N utilization for milk protein production on AL compared to OG as pVDMl increased (Figure 9b). Urinary N output was not measured; it is possible that an increasingly greater amount of digested N was excreted in urine on AL than on OG as pVDMI increased. Although the increasingly greater DMI for AL 72 compared to OG allowed additional N intake among cows with greater pVDMI, that extra feed N apparently was digested and utilized less efficiently. This assertion that utilization of N may be decreasingly efficient with increasing DMI bears important economic and environmental implications for dairy farms. For all the animals in this experiment, fuel availability was likely more limiting than protein for milk production. Diets were formulated to ensure that N and amino acid availability were not limiting to ruminal fermentation or milk production, and the actual dietary concentration of CP was high (approximately 20% of DM) for both diets. Even with excess dietary total and rumen-degradable CP in both diets, the effect of diet on whole-tract N digestibility and efficiency of N utilization for milk true protein depended on pVDMl. This implies that when practices are implemented to permit greater DMI, dietary N concentration might need to be reduced in order to avoid less-efficient digestion and utilization of N. Furthermore, the effects of pVDMI on N digestion and utilization reinforce the need to group and feed animals according to some index of nutrient demand. Reducing the variation in energy and protein demand within the group for which a diet is formulated would reduce the extent to which fuels or N limit ruminal fermentation or milk production in all animals. This would allow diets to be formulated to more accurately meet each individual animal’s demands and thus lead to more efficient utilization of N among all groups of animals on the farm. Increased N digestion and utilization and more accurate diet formulation will reduce the proportion and amount of N excreted in feces and urine. It should also reduce the likelihood of overfeeding N. Thus, adjusting feeding practices for 73 the effects of increasing DMI on efficiency of N utilization can contribute to the reduction of N waste. SUMMARY AND CONCLUSIONS As expected, ruminal N absorption was greater when the dietary forage was alfalfa than when it was orchardgrass. The effect of forage type on N intake, digestion and utilization depended on the pVDMI of individual animals. Because DMI responded increasingly more positively to AL than to OG as pVDMl increased, intake and duodenal flow of N also increased more for AL than for OG with increasing pVDMl. Site of digestion and efficiency of utilization of dietary N for microbial protein and for milk true protein depended not only on intake of N but also on responses of ruminal passage rate and ruminal starch digestion. The reduction of passage rate by OG, particularly among cows with high pVDMl, reduced the total amount of N consumed and utilized for microbial protein and milk true protein production. However, a decreasing proportion of the additional N consumption that was allowed by the increased DMI on AL among cows with greater pVDMI was digested and used for increased milk production or body tissue gain. Increasing passage rate and DMI by feeding a perennial legume forage instead of a perennial grass forage can increase yields of milk and milk protein among cows with greater nutrient demand. However, increasing N intake at the same rate as DMI is increased likely will lead to less efficient utilization of dietary N for production of microbial protein, muscle, or milk true protein. When feeding less-filling diets, such as those containing a legume forage, to high- 74 producing cows, reducing dietary N concentration could increase the efficiency of N utilization and reduce the extent to which greater DMI leads to greater N excretion. A better understanding of the different effects of perennial grass and legume forages on N utilization by cows with different nutrient demands will aid in field management decisions to minimize the turnover and loss of N on the whole farm. 75 Table 1. Status of eight cows during the final 4 d of the preliminary period, when cows were fed a common diet. Parameter Mean SD Parity 4.0 2.6 BW, kg 538 17 BCS 2.5 0.4 DIM 139 83 Milk yield, kgld 40.1 5.5 DMI, kg/d 18.6 2.8 Figure 1. Distribution of DMI and 3.5% fat-corrected milk yield of eight cows during the final 4 d of the preliminary period, when cows were fed a common diet. Preliminary VDMI, kgld Preliminary 3.5% FCMY, kgld 22 50 45 20 40 18 35 30 16 25 14 20 , 1 2 3 1 2 3 Number of Cows Number of Cows Table 2. Chemical characterization of alfalfa silage and orchardgrass silage. Alfalfa Silage Orchardgrass Silage DM (% as fed) 30.6 Nutrient, % DM OM 88.7 89.2 NDF 42.6 48.0 Indigestible NDF 25.2 13.1 Potentially digestible NDF 17.4 34.9 Starch 4.0 2.3 Crude protein 20.5 20.4 30-h in vitro NDF digestibility, % 29.4 61.1 76 Table 3. Ingredient and nutrient composition of treatment diets, one diet (AL) containing alfalfa silage and another diet (OG) containing orchardgrass silage. AL OG _l_n_gredient % of DM Alfalfa silage 53.0 ----- Orchardgrass silage ----- 47.9 Dry ground corn 36.3 40.3 Soybean meal (48% CP) 6.5 7.0 Vitamin mineral mix1 4.2 4.2 Expeller-processed soybean meal2 1.3 1.3 Bloodmeal 0.3 0.9 Limestone ------ 0.4 Urea ------ 0.2 Nutrient DM (% as fed) 43.6 50.6 % of DM OM 91.5 91.5 NDF 26.7 27.5 Forage NDF 22.5 23.0 Indigestible NDF 14.8 7.9 Potentially digestible NDF 11.9 19.7 Starch 30.2 32.1 Crude protein 18.3 18.8 Rumen-undegraded CP3 5.6 6.3 1 \fitamin mineral mix contained (DM basis) 11.7% dicalcium phosphate, 11.1% trace-mineral premix, 8.8% sodium bicarbonate, 2.3% magnesium oxide, 134.3 KlU/kg vitamin A, 35.53 KlU/kg vitamin D, 895.5 KlU/kg vitamin E, and 65.2% ground corn grain as a carrier. Nutrient composition: 86% DM, 7% ash, 16% NDF, 5% starch, 51% CP. 3 Estimated using values from NRC (2001). 77 Table 4. Least-squares means, standard errors, and P-values of effects of forage source on N intake, digestion, and utilization. Treatment LSM1 Variable AL 06 SEM2 P Intake, g/d 620 623 37 0.95 Rumen pool, 9 269 371 20 0.01 Turnover time in rumen, h 10.7 14.4 0.8 <0.01 Rumen ammonia concentration, mg/dl 29.3 18.0 1.1 < 0.001 Mean pH 6.44 6.17 0.04 < 0.001 Ruminally digested g 196 110 42 < 0.01 % 30.4 17.7 4.3 0.03 N flow to duodenum Total, g/d 426 515 35 0.14 Ammonia N, gld 19.9 14.5 1.1 < 0.01 Non-ammonia N (NAN) gld 406 500 42 0.10 Microbial N gld 271 293 26 0.56 % duodenal NAN 66.3 58.3 2.6 0.07 % TRDOM3 Per 1 1.9 2.8 0.3 0.09 Per 2 2.7 2.5 0.3 0.69 NA, non-microbial N (NANMN) gld 135 207 16 0.03 % duodenal NAN 33.7 41.7 2.7 0.09 N digested postruminally g 230 316 30 0.08 % 54.4 60.2 2.8 < 0.01 NAN digested postruminally gld 210 301 29 0.06 % duodenal NAN 52.1 59.0 2.9 < 0.01 N digested in the whole tract 9 425 424 23 0.99 % 68.6 68.1 1.1 0.50 Yield, kgld Milk 29.1 29.4 2.6 0.77 3.5% fat-corrected milk 31.4 33.8 3.3 0.19 True protein 0.88 0.89 0.06 0.77 Milk true protein concentration, % 3.06 3.07 0.12 0.65 9 milk true protein N lfig intake N 0.21 0.22 0.01 0.89 1 Treatment least-squares means. ’2 Standard error of the mean. 3 Per x Tn: P= 0.13. 78 Figure 2. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) of (A) N intake (N intakeAL - N intakeoG = -570 + 30.5 pVDMI), and (B) N turnover time (N TOTAL - N TOTOG = -106 + 12.8 pVDMI — 0.38 pVDMIz). Equation A includes adjustment for Seq. A 150 'O _ O ‘ P-0.01 a 100 ,5- 1 R2=0.63 . 4,5 50' RMSE=56.1 .15 0- z 8 -50‘ g .100 8 -150 a, O I! 400-145...- 1415 16171819 20 2122 pVDMl,kgld B a .5 E a: 3 4f 65 ‘ %o .81P=0.08 . ol- 10 R2=0.70 "‘2 ' iRMSE=2.20 .12 . , , , , f ,°_.. 14 15 16 1718 19 20 21 22 pVDMI,kgld 79 Figure 3. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) N absorbed from the rumen (g N rumen digestedAL - g N rumen digestedog = 2887 = 317 pVDMI + 8.75 pVDMIZ) and (B) ruminal digestibility of N (% of N intake) (ruminal N digestibilityAL - ruminal N digestibilityos = 453 — 47.8 pVDMI + 1.27 pVDMIZ). Equations A and B include adjustment for Seq. A ,, 20° 1' P=0.04 (L), 0.05 (o) 0 fi . R2=o.91 E 15° RMSE= 66.1 3 - 1, 100‘ 2 Ta §§ 50- ° 6 3 9 o. m .9 m 'D 1 14 1516 1716 19 2'0 27 22 pVDMI,kg/d B 50 L P=0.08 _ 4oi R==0.66 g RMSE=5.69 E E e\° z . o 5' 2 § 0 U Q 0) 6 .3 m 'U - o 1 r v v ' Jp—'—i 14 15 16 17 18 19 20 21 22 pVDMl, kgld 80 Figure 4. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) N flow to the duodenum (N flow/(L - N flowos = -823 + 39.5 pVDMI); and (B) NAN flow to the duodenum (NAN flowAL - NANOG = «815 + 38.8 pVDMl). . A 100 r—- 50 . P - 0.01 R2 = 0.69 0 ‘ RMSE = 62.2 -50 1 o -100' -1501 -200‘ .250 -300 . 131 1'5 16 17 16 1'9 2'0 21 22 wow, kgld Response N to duodenum , gld m 1oo _ . 50. P-0.01 R2 = 0.69 ' RMSE = 79.6 -50 ‘ ~100 ‘ ~150 ' -200 ‘ .250 -300 -350-‘- - - - ~ -- 14 15 16 17 16 19 20 21 22 pVDMI,kg/d Response NAN to duodenum, gld 81 Figure 5. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) microbial N flow to the duodenum (MNAL - MNOG = -583 + 30.2 pVDMl) and (B) microbial N (% truly rumen degraded OM (T RDOM)) (MNEAL — MNEOG = -42.1 + 4.64 pVDMI - 0.13 pVDMIZ). A 150 ‘P=0£5 - 10° R2 = 0.51 ' 501 RMSE = 62.9 0.4 .50 .. -1001 .150 -200 e -250- . . - 4 . . . 14 15 16 17 16 19 20 21 22 pVDMI, kgld Response duodenal microbial N flow, gld 03 PP.‘ 001° Response microbial N to duodenum, °/o TRDOM 1': 0| '1'” P = 0.03 45 R2 = 0.76 RMSE=056 .21) . . . . . . . 14 15 16 17 16 19 20 21 22 pVDMI, kgld 82 Figure 6. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) Truly ruminally degraded starch (TRD starch), kgld (T RD starchAL - TRD starches = 66.1 - 7.88 pVDMI + 0.23 pVDMIZ) and (B) True ruminal starch digestibility (TRSD), % (T RSDAL — TRSDOG = 900 — 103 pVDMI + 2.9 pVDMlz). Equations A and B are corrected for Seq. A 1.5 3;. P=0.05 . "‘. 1.01. R2=0.70 {j RMSE=0.86 B 0.51 m D E 0.01 0 2 -0.510 8 . 3 -1.0- n: 0 "14151651619 50 2122 pVDMI,kgId B 30 . g 20- D" . g 101. . '— O 3 c 0‘ O ,9, 10 P=0.04 g ' ' R2=0.75 RMSE=109 . .20 . - . L . . . 14 15 16 17 16 19 20 21 22 pVDMI,kgId 83 Figure 7. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) N digested postruminally (N digested postruminallyAL — N digested postruminallyoc; = -670 + 31 pVDMl) and (B) NAN digested postruminally (gld) (NAN digested postruminallyAL — NAN digested postruminallyoe = -660 + 30 pVDMI). A 0 l .501 .1001 -150l .200 -250 -300 P=091 R2 =0.67 RMSE=677 141516171619202122 pVDMl, kgld Response N digested postruminally, gld 50 P = 0.01 . Q 0‘ R2 "-" 0.67 .501 RMSE = 66.2 -1001 -150-L -200‘ -2501 -3001—'. . . . . . 4 141516171819 20 21 22 pVDMl, kgld Response of NAN digested w postruminally, g 84 Figure 8. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) N digested in the whole tract (9 N digestedAL - g N digestedoe = -375 + 20.2 pVDMl) and (B) whole-tract digestibility of N (% of N intake) (N digestibilityAL — N digestibilityoe = 8.15 - 0.41 pVDMI). Equations A and B include adjustment for Seq. A 300 u P=0.02 , 33.0 200‘ R2=0.89 g3, RMSE=42.2 . ' mg 100- 2 3“: V . :33 0‘ 00 a: . $3.100 ° 12.5 . . o . . . . . - . 14 15 16 17 16 19 20 21 22 pVDMI.kgId B 4 5 . P=0.07 ‘3'; 3- R2=0.96 3,5 2. RMSE=1.35 11g 1 . is“ 7 a,“ g: 0" O 0 %B .11 ', 0.: 0:3 _2 f 1 . _ - . '9— 141516171619 20 2122 pVDMI,kg/d 85 Figure 9. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to the alfalfa diet (AL) over the orchardgrass diet (OG) in (A) milk true protein yield (Milk true protein yieldAL — Milk true protein yieldog = 3.80- 0.47 x pVDMI + 0.014 x pVDMIz); (B) total milk N (% N intake) (Milk True Protein N (% N intake)AL — Milk True Protein N (% N intake)os = 11.6- 0.63 x pVDMI); (C) BCS change (BCS changeAL - BCS changeoe = -15.9 + 1.87 pVDMI - 0.0537 pVDMIZ); and (D) NEL balance (NEL balanceAL - NEL balanceoc; = -26.9 +1.44 pVDMI). Equation B includes adjustment for Sequence effect. A 250 P < 0.10 ‘— 2°° ‘ R2 = 0.60 150 - RMSE = 66 100‘ 50' 0-1 Response milk protein, gld -150 14 15 16 17 16 16 26 21 22 pVDMl,kgld B \ 0.06 z I. .5 0.04- 3 o O a 0.02- e o g 0.004 . . a: g z -0.02-1 g 3 P = 0.10 3% -0-04- R2=0.49 «1 -- RMSE = 0.02 e “tn-0.06 .. . . . .. 14 15 16 17 16 19 20 21 22 pVDMI,kgId 86 O Response BCS change, (1'1 P < 0.01 R2 = 0.82 RMSE = 0.142 '141 U I I I I I I I 516 17 1819 20 2122 pVDMl, kgld Response NEL balance, Mcalld P=093 _ R2 = 0.52 RMSE=3J1 1'4 1'5 1'6 1'7 1'6 119 2'0 2'1 22 pVDMI, kgld CHAPTER 4 Nutrient Demand of Lactating Dairy Cows Affects Ruminal Digestion Responses to a Change in Dietary Forage Concentration ABSTRACT Previous research in our laboratory indicates that physical filling effects of high-forage diets become increasingly dominant in determining feed intake and milk production as nutrient demand increases. This effect was tested further using 14 ruminally and duodenally cannulated Holstein cows in a crossover design experiment with a 14 d preliminary period and two 15 d experimental periods. During the preliminary period, 3.5% fat-corrected milk yield (FCMY) was 15 to 60 kgld (mean = 40 kgld), and preliminary voluntary DMI (pVDMI) was 20.6 to 30.5 kgld (mean = 25.0 kgld). Treatments were a low-forage diet (LF), containing 20% (DM basis) forage neutral detergent fiber (NDF), and a high- forage diet (HF), containing 27% forage NDF. The ability of linear and quadratic factors of pVDMl to predict the difference in responses of individual cows to treatments (YU: - YHF) was tested by analysis of variance, with treatment sequence as a covariate. In contrast to results of previous research, differences in DMI and FCMY responses to LF and HF did not depend on pVDMI. This might be because of combined physical fill and metabolic satiety effects of LF, especially in cows with greatest pVDMI. Digestion and(or) passage of NDF might have been inhibited on LF among high-pVDMI cows. As pVDMI increased, NDF turnover time increased more on LP than on HF. Among high-pVDMI cows, 88 NDF turnover time was unexpectedly greater on LP than on HF. VVlth increasing pVDMI, digestion rate of pdNDF decreased at a similar rate on both diets. Passage rates of potentially digestible NDF and indigestible NDF were not related to pVDMl, regardless of treatment. Because mean and minimum ruminal pH were lower for LF than for HF, a slight numerical reduction in pH with increasing pVDMI observed for both diets likely would inhibit NDF digestion more for LF than for HF. Inhibition of NDF digestion might cause low-forage and high- forage diets to have similar effects on DMI, depending on the VDMI of individual COWS . INTRODUCTION Diet forage NDF concentration affects feeding and digestion in dairy cows through both physical and chemical mechanisms. Physical controls include gut distension (Lehman, 1941) and limitations to time spent eating and ruminating (Allen, 2000). Mechanisms through which diet affects the physical control of feed intake include retention time of digesta fractions (Campling et al., 1961), potential digestibility of fiber (Oba and Allen, 1999), diet particle size and rate of particle size reduction (Poppi et al., 1980), particle specific gravity (Balch and Kelly, 1951), and diet effects on frequency and duration of reticulorumen contractions (Okine and Mathison, 1991; Dado and Allen, 1995). Altered fermentation acid production in the rumen resulting from changes in diet forage NDF concentration may also affect intake and digestion responses to diet (Sheperd and Combs, 1998). Excess production of fermentations acids with low-forage-fiber diets 89 results in lower ruminal pH, which can decrease fiber digestibility (Hoover, 1986). Excess propionate production can result in lower feed intake, independent of pH effects (Allen, 2000). However, energy balance influences both feed intake responses to diet characteristics and the extent to which physical or metabolic factors limit VDMI (Mertens, 1994; Allen, 1996). The effects on feed intake of diet characteristics (such as diet forage NDF concentration) that influence ruminal passage rate of digesta will depend on the extent to which physical filling effects limit feed intake in an individual animal. As a result, testing only overall treatment mean differences may mask important responses in intake, digestion, and production (Allen, 2000). Because cows are now frequently grouped and fed according to milk yield, models that predict the effects of nutrient demand on response to diet are even more necessary. We developed and successfully used an experimental model to evaluate effects of indices of nutrient demand, such as preliminary milk yield, on animal responses to dietary treatments (Oba and Allen, 1999a; Burato et al., 2001; Voelker et al., 2002; Harvatine and Allen, 2002; Bradford and Allen, 2004). This model was utilized to test our hypothesis that preliminary VDMI (pVDMI) affects individual responses of VDMI and digesta passage rate to diets containing high and low concentrations of forage NDF. A previous experiment (Voelker et al., 2002) investigated this hypothesis using intact and ruminally- cannulated cows; based on results from that experiment, we expect passage rates of digesta fractions to become increasingly greater for the low-forage diet compared to the high-forage diet as preliminary VDMI increases. The present 90 experiment was conducted using ruminally- and duodenally-cannulated cows with a wide range of pVDMl to investigate the mechanisms underlying the responses to changes in dietary forage-fiber concentration. MATERIALS AND METHODS Cows and Treatments Experimental procedures were approved by the All University Committee on Animal Use and Care at Michigan State University. Fourteen multiparous Holstein cows from the Michigan State University Dairy Cattle Teaching and Research Center were assigned randomly to treatment sequence in a crossover design experiment with a 14 d preliminary period and two 15 d experimental periods. These fourteen cows were 178 :1: 120 (mean :1: SD) DIM at the beginning of the preliminary period (Table 1) and were selected deliberately to provide a wide distribution of milk yield and DMI (Figure 1). During the 14 d preliminary period, milk yield ranged from 16.1 to 59.1 kgld (mean = 38.7 kgld) and pVDMI ranged from 10.6 to 30.5 kgld (mean = 25.0 kgld). Cows were cannulated ruminally and duodenally prior to calving. Surgery was performed at the Department of Large Animal Clinical Science, College of Veterinary Medicine, Michigan State University. Cows were housed in tie-stalls and fed once daily (1100 h) at 110% of expected intake. Treatments (Table 2) were a low-forage diet (LF) and a high-forage diet (HF) fed once daily as totally mixed rations. Diet LF was formulated to contain 20% of DM as forage neutral detergent fiber (NDF) and 24% as total dietary 91 NDF, and HF was formulated to contain 27% forage NDF and 31% total dietary NDF. Forage-to-concentrate ratios (% of DM) were 45:55 for LF and 61 :39 for HF. Diets also contained dry ground corn, soybean meal, an expeller-processed soybean meal, and a vitamin-mineral premix; urea, and soybean meal were used to achieve similar CP and estimated RUP fractions in the two diets. Diets were formulated for 18% dietary CP but the actual diet CP contents were 16.2 and 16.6% (Table 2). The diet fed during the preliminary period was formulated, using the same ingredients, to contain 24% forage NDF. Sample and Data Collection Amounts of feed offered and orts were weighed for each cow daily. Samples of all dietary ingredients (0.5 kg) and orts from each cow (12.5% of orts) were collected daily on d 11 to 13 and combined into one sample per period. Cows were milked twice daily in a milking parlor (0300 and 1500 h); milk yield was measured, and milk was sampled, at each milking on d 11 to 13. Rumen- empty BW was measured after evacuation of ruminal digesta on d 14 of the preliminary period, and on d 15 of each experimental period. Body condition score was determined on the same days by three trained investigators blinded to treatments (WIIdman et al., 1982; five-point scale where 1 = thin and 5 = fat). Duodenal samples for digestion measurements (700 mL) and for particle— size analysis (700 mL), rumen fluid samples for microbial isolation (350 mL), and rumen fluid samples for pH (100 mL) were collected every 9 h from d 11 to d 13 so that eight samples were taken for each cow in each period, representing every 3 h of a 24-hour period to account for diurnal variation. Rumen fluid for microbial 92 isolation was collected from the reticulum, near the reticular-omasal orifice, and strained. Rumen fluid was obtained by combining digesta from five different sites in the rumen and straining it through a layer of nylon mesh (~1 mm pore size); fluid pH was recorded immediately. Samples were stored immediately at -20°C. Ruminal contents were evacuated manually through the ruminal cannula at 1600 h (5 h after feeding) on d 14 and at 0700 h (4 h before feeding) on d 15 of each period. Total ruminal content mass and volume were determined. During evacuation, 10% aliquots of digesta were separated to allow accurate sampling. Aliquots were squeezed through a nylon screen (1 mm pore size) to separate into primarily solid and liquid phases. Both phases were weighed and sampled (two, 350 mL samples of each phase) for determination of nutrient pool size and particle size analysis. Samples were stored at -20°C. Sample and Statistical Analyses Diet ingredients and orts were dried in a 55°C forced-air oven for 72 h and analyzed for DM concentration. All samples were ground with a Wiley mill (1mm screen; Authur H. Thomas, Philadelphia, PA). One set of frozen duodenal samples for each cow period (n = 8) were chopped into “snow” using a commercial food processor (84142 Food cutter, Hobart Manufacturing 00., Troy, OH) and sub-sampled in the frozen state to obtain representative samples. These duodenal subsamples and one set of 350 mL ruminal solid and liquid samples for each rumen-emptying time were Iyophilized (Tri-Philizer" MP, FTS Systems, Stone Ridge, NY) and ground as above. Dried ruminal solid and liquid , samples were recombined according to the original ratio of solid and liquid DM. 93 Samples were analyzed for ash, NDF, indigestible NDF (iNDF), CP, and starch. Ash concentration was determined after 5 h oxidation at 500°C in a muffle furnace. Concentrations of NDF were determined according to Van Soest et al. (1991, method A). Indigestible NDF was estimated as NDF residue after 120-h in vitro fermentation (Goering and Van Soest, 1970). Rumen fluid for the in vitro incubations was collected from a non-pregnant dry cow fed only alfalfa hay. Fraction of potentially digestible NDF (pdNDF) was calculated by difference (1.00 - iNDF). Crude protein was analyzed according to Hach et al. (1987). Starch was measured by an enzymatic method (Karkalas, 1985) after samples were gelatinized with sodium hydroxide. Glucose concentration was measured using a glucose oxidase method (Glucose kit #510; Sigma Chemical Co., St. Louis, MO), and absorbance was determined with a micro-plate reader (SpectraMax 190, Molecular Devices Corp., Sunnyvale, CA). Concentrations of all nutrients except DM were expressed as percentages of DM determined by drying at 105° C in a forced-air oven for more than 8 h. Milk samples were analyzed for fat, true protein, MUN, and lactose with infrared spectroscopy by Michigan DHIA (East Lansing). Duodenal samples were analyzed for purines and ammonia to estimate microbial N flow and non- ammonia non-microbial N flow to the duodenum. Purine concentration was used as a microbial marker, and purine to microbial N ratio was estimated by analysis of microbial pellets obtained by differential centrifugation of the rumen fluid collected in the reticulum. Total purines were measured by spectrophotometer , (Beckman Instruments, Inc., Fullerton, CA) at 260 nm (Zinn and Owens, 1986). 94 Ammonia concentration was determined for centrifuged duodenal and rumen fluid samples according to Broderick and Kang (1980). Indigestible NDF was used as an internal marker to estimate nutrient digestibility in the rumen (Cochran et al., 1986), and to estimate rates of passage for iNDF, pdNDF, and starch, and rates of digestion for pdNDF and starch. Nutrient intake was calculated using the composition of feed offered and refused. Duodenal flow of microbial OM was determined as described by Oba and Allen (2003b), and true ruminally degraded OM (T RDOM) was calculated by subtracting duodenal flow of non-microbial OM from OM intake. Ruminal pool sizes (kg) of OM, NDF, iNDF, pdNDF, and starch were determined by multiplying the concentration of each component by the ruminal digesta DM mass (kg). Turnover rate in the rumen, passage rate from the rumen, and ruminal digestion rate of each component (%lh) were calculated as reported by Voelker and Allen (2003b). Rates of particle size reduction in and particle passage from the rumen also were determined using iNDF as a marker (Figure 2). Triplicate 20-g feed and orts samples were sieved. Thawed subsamples of ruminal solid and liquid phases (the second set from each of two rumen evacuations per period) were recombined into duplicate 60-g samples based on the original (wet) ratio of solid and liquid phases. The second set of whole duodenal samples were thawed and combined (eight per cow-period), then separated into liquid and solid and stored frozen. The two phases were thawed and recombined in duplicate 200-g samples for sieving. Feed, orts, rumen, and duodenal samples were individually 95 wet-sieved sequentially through 4.75 mm, 2.36 mm and 38 pm screen (W.S. Tyler lnc., Gastonia, NC). Particles retained on each screen were removed and dried at 55° C for 48 h, then weighed. Material retained on each screen from replicate sievings were combined (keeping after-feeding and before-feeding rumen empty samples separate). Because DM in duodenal digesta retained on the 4.75 mm screen was < 5% of total DM on screens, 2.36 mm was selected as the threshold. Residue 2 2.36 mm, including residue on 4.75 and 2.36 mm screens, averaged 13.4% of total DM. Therefore, particles retained on the 2.36 and 4.75 mm screens were combined and the resulting fractions were designated 2 2.36 mm (less likely to escape the rumen) and < 2.36 mm (more likely able to escape the rumen). These two fractions were ground (1 mm, Vlfiley mill). Ground sieving residues were analyzed for DM, iNDF, and NDF concentrations. Indigestible NDF was used to calculate rate of particle size reduction in the rumen (2 2.36 to < 2.36), because (1) kinetics must be calculated for a homogeneous pool, and (2) pdNDF can leave pool by digestion as well as particle-size reduction and passage but iNDF can leave the pool only by breakdown or by passage. Passage rates of iNDF in large (22.36 mm) and small (<2.36 mm) particles, rate of flux of iNDF from the 22.36 mm pool to the <2.36 mm pool, and relative size threshold for escape from the rumen were calculated as follows: Passage rate (kp): iNDka2236 = iNDFDuod22.36 (kg/d) / iNDFRumenPoo|22.36 (kg) 96 iNDka<2_35 = iNDFouod 0.25). The NDF in small particles (<2.36 mm) contained a larger proportion of iNDF than did the NDF in large particles (2 2.36 mm; Table 4). Because a greater lignin concentration results in greater fragility (McLeod and Minson, 1988), particles with greater iNDF concentration likely break down to smaller particles more quickly. Therefore, small particles should contain greater concentrations of iNDF than should larger particles. 100 Furthermore, the proportion of iNDF in total NDF in small particles increased between intake pool (54.5 and 55.4% of NDF for LP and HF, respectively) and rumen pool (61.6 and 62.1% of NDF for LF and HF, respectively). This is expected, because some pdNDF was digested and iNDF was not. However, in large particles (2 2.36 mm), the proportion of iNDF was similar in intake pool (44.1 and 45.7% of NDF for LF and HF, respectively) and rumen pool (44.6 and 44.8% of NDF for LF and HF, respectively). This suggests that the digestion rate of pdNDF in particles 2 2.36 mm was very low, even negligible. Digestion rate of pdNDF in small and large particles could not be calculated, because pdNDF can disappear from the pools by digestion as well as by passage or particle size reduction. Larger forage particles may indeed undergo negligible NDF digestion because of the small surface area available for bacterial digestion relative to the total surface area or volume of the particle (Vlfilson and Hatfield, 1997). The similar proportions of iNDF in feed and rumen particles 2 2.36 mm suggest that almost no NDF digestion takes place in particles until they are broken down (by chewing) to < 2.36 mm. Fractional passage rates of iNDF in particles < 2.36 mm and in particles 2 2.36 mm were greater on LP than on HF (P < 0.03, P < 0.02, respectively), and ranged from 2.10 %/h (particles 2 2.36 mm on HF) to 6.10 %/h (particles < 2.36 mm on LF), spanning the passage rates observed for total iNDF as would be expected. Passage rate of pdNDF tended to be greater on LF than HF in particles < 2.36 mm (P = 0.06), but pdNDF passage rate in particles 2 2.36 mm - was similar between treatments (P 2 0.75) and was numerically much slower 101 than the passage rate of iNDF in particles of similar size (Table 4). The range of passage rate of pdNDF in small and large particles (0.59 to 2.34 %lh) also spanned the passage rates observed for total pdNDF. The proportion of duodenal iNDF or pdNDF flux contained in particles 2 2.36 mm was quite small (13 to 21% of total; Table 4). A slightly greater proportion of iNDF was found in. large duodenal particles on LF than on HF (P = 0.03), and a greater proportion of pdNDF was found in large duodenal particles on HF than on LP (P < 0.05). Passage rate of total pdNDF tended to be greater for LF than for HF (Table 3). A period by treatment interaction existed for passage rate of total iNDF (P = 0.06); iNDF passage rate was similar between treatments during period 1 (mean = 4.8; P 2 0.65) but was greater for LF (4.8 %lh) than for HF (3.9 %lh) during period 2 (P < 0.01). Starch passage rate responded to treatment in a manner similar to iNDF passage rate. The tendency for greater passage rate of digesta fractions for LF than for HF suggests that passage rate could not be increased on HF to permit greater DMI in response to a more physically filling, more slowly digested diet. The slower passage rates of pdNDF and iNDF for HF apparently outweighed both the greater digestion rate of pdNDF and the greater rate of particle size reduction for HF in determining the physical filling effects of the diet. As a result of greater DMI for LF, yields of raw and 3.5% fat-corrected milk also were greater for LP (Table 3). Milk fat concentration was lower for LP (P = 0.04), possibly because of tendencies for faster passage rates of iNDF, pdNDF, , and starch for LP. Faster passage rate of digesta on LP might have resulted in 102 greater escape of rumen biohydrogenation intermediates (Harvatine and Allen, 2006). Some partially biohydrogenated FA may inhibit milk fat synthesis and thus lower milk fat concentration (Bauman and Griinari, 2003). Effect of pVDMl on Response to Diet Many of the treatment effects observed here have been demonstrated previously; the primary hypothesis for this experiment was that the differences in responses of these parameters to treatment would change with increasing pVDMI, used as an index of nutrient demand. Contrary to the hypothesis, individual responses of DMI, digesta passage rates, and 3.5% FCMY did not depend on preliminary intake (data not shown). Only the response to treatment of ruminal NDF turnover time depended on pVDMI (Figure 3). Mean NDF turnover time demonstrated little difference between the two treatments (Table 3), but as pVDMl increased, NDF turnover time increased more greatly for LF than for HF (Figure 3). This is likely why DMI of cows with the greatest pVDMI did not respond as positively to the LF diet as expected; a longer ruminal NDF turnover time suggests that LF may have had more physical filling effects than HF among cows with high pVDMl. Neither digestion rate nor passage rate explain this turnover time effect, because with increasing pVDMI, digestion rate of pdNDF and passage rates of iNDF and pdNDF changed similarly for both diets. Responses of passage rates of iNDF and pdNDF in particles < 2.36 mm and 2 2.36, and response of particle size reduction rate, did not depend on pVDMl (data not shown). It is possible that undetectable interactions between 103 diet and pVDMl in affecting both digestion and passage rates combined to create the detectable NDF turnover time effect. Inhibition of NDF digestion or passage on a low-forage diet at high DMI could be caused by direct and indirect effects of increased starch intake and fermentation. Although DMI response did not depend on pVDMl, the greater starch concentration in the LF diet still led to a greater increase in starch intake for LF than for HF, with increased pVDMl (P = 0.03). An increased rate of starch fermentation can reduce ruminal pH, and lower pH can depress rumination, rumen motility, and NDF fermentation, which would affect NDF turnover time. On both treatments, ruminal pH tended to decrease as pVDMI increased (P = 0.11), and there was no difference in the slopes of the two lines (P 2 0.60). However, mean ruminal pH was lower for LF than for HF (P < 0.0001; Table 3), so any effect of decreasing pH with increasing pVDMl on NDF digestion, rumination, or rumen motility, was likely more severe for LP than for HF. Therefore, it is possible that a lower ruminal pH on LF among cows with high pVDMI caused a longer NDF turnover time on LF with increasing pVDMl. Relationships reported in a previous similar experiment (Voelker et al., 2002) were between response and preliminary milk yield (or milk energy output), but the same relationships also existed with preliminary DMI, so similar responses were expected, but not observed, in the current experiment. Several factors might have contributed to the observation of different responses in the present experiment and the previous experiment. First, the previous experiment utilized 32 cows and the present experiment used only 14 cows. However, the 104 ranges of preliminary DMI and FCMY were similar for the two sample groups, and a 12-animal subgroup of ruminally-cannulated cows in the 32-cow study detected dependencies of DMI and ruminal kinetics responses on pVDMl. Second, although high-forage and low-forage diets in the two experiments contained very similar proportions of NDF (24 and 31% of diet DM for both experiments) and starch (33 and 23% of diet DM for both experiments), differences existed between treatment diets in the two experiments. Particle size distributions of diets might have differed between experiments, but this cannot be determined because diet particle size was not measured for the first experiment. Non-forage fiber sources were included in diets for the previous experiment, so diet forage NDF concentration was lower for both diets in that experiment (16 and 24% of diet DM) than for the diets fed in the present experiment (20 and 27% of diet DM). The NDF in the non-forage fiber sources used in the previous experiment (dried corn distillers grains and whole cottonseed) likely had higher rates of NDF digestion and(or) passage, so those diets may have been less physically filling. The non-forage fiber sources were also sources of fat, and diets in the previous experiment contained a commercial fat supplement. Therefore, the caloric densities of those diets were likely greater compared to diets in the present experiment. Also, the rumen-available fats may have altered ruminal fermentation. Finally, the corn grain used in the previous experiment was rolled high- moisture corn, so its ruminal fermentation characteristics were likely different , from fermentation characteristics of the ground dry corn used in the present 105 experiment. Ruminal starch digestibility and digestion rate are factors involved in determining the effects of dietary starch concentration on NDF fermentation and DMI. Oba and Allen (2003a) reported that increasing dietary starch concentration increased DMI when grain was more slowly fermented (dry corn) but not when it was more rapidly fermented (high-moisture corn). The more rapidly fermented high-moisture corn in the first experiment likely would have contributed to greater intake depression on LF among cows with low pVDMI, but it also would have been more likely to lower ruminal pH and interfere with NDF digestion on LF among cows with higher pVDMl. It is likely that, because of dietary differences, digesta in the previous experiment were more rapidly fermented and(or) escaped more quickly from the rumen compared to digesta in the present experiment. This might have caused cows with lower pVDMl on the previous experiment to respond more negatively to the low-forage diet, which would contribute to an increasingly positive response to that diet as pVDMl increased. SUMMARY AND CONCLUSIONS A longer NDF turnover time on LF with increasing pVDMl led to responses of DMI and milk production to high-forage and low-forage diets that were independent of pVDMl. This response might have been mediated by diet effects on ruminal pH. The results of this experiment suggest that models that predict intake need to account for not only the effects of nutrient demand, but also the 106 effects of the interactions of feed fractions (such as starch and NDF) on the intake responses of individual cows to high- and low-forage diets. 107 Table 1. Status of 14 cows during the final 4 d of the preliminary period, when cows were fed a common diet. Parameter Mean SD Parity 2.9 0.7 BW, kg 597 55 BCS 2.6 0.7 DIM 178 120 Milk yield, kgld 38.7 12.3 DMI, kgld 25.0 2.7 Figure 1. Distribution of voluntary DMI (VDMI) and 3.5% fat-corrected milk yield (FCMY) of 14 cows during the final 4 d of the preliminary period, when cows were fed a common diet. Preliminary VDMI. kgld Preliminary 3.5% FCMY, kgld 32 70 30 60 28 50 26 40 24 30 22 20 20 10 . 1 2 3 4 5 1 2 3 4 5 Number of cows Number of cows 108 Table 2. Ingredient and nutrient composition of treatment diets, a low-forage diet (LF) and a high-forage diet (HF). LF HF figredient % of DM Corn silage1 29.7 40.5 Alfalfa silage2 15.1 20.9 Dry ground corn 33.9 16.1 Soybean meal (48% CP) 11.0 9.1 Vitamin mineral mix3 3.2 4.3 Expeller-processed soybean meal4 5.9 9.1 Urea 0.2 ---- Nutrient DM (% as fed) 47.0 39.6 % of DM OM 93.0 92.1 NDF 24.4 30.7 Forage NDF 19.9 27.3 Indigestible NDF 13.2 15.1 Potentially digestible NDF 11.2 15.6 . Starch 32.8 22.5 Crude protein 16.2 16.6 Rumen-undegraded CP5 7.2 7.3 1 Corn silage contained 46.4% NDF, 16.9% iNDF, 18.6% starch, and 8.1 % CP. 30-h in vitro NDF digestibility was 47.5%. 2 Alfalfa silage contained 40.6% NDF, 26.4% iNDF, 3.5% starch, and 18.3 % CP. 30-h in vitro NDF digestibility was 32.6%. 3 Vitamin mineral mix contained (DM basis) 10.1 % dicalcium phosphate, 4.1% trace-mineral premix, 5.7% sodium bicarbonate, 1.2% magnesium oxide, 124.2 KlU/kg vitamin A, 40.3 KlU/kg vitamin D, 671.6 KlU/kg vitamin E, and 60.1% ground corn grain as a carrier. Nutrient composition: 86% DM, 7% ash, 16% NDF, 5% starch, 51% CP. 5 Estimated using values from NRC (2001). 109 Figure 2. Model of ruminal particle size reduction and passage. Reduction of particle size during eating is included in rate of particle size reduction (kr). Passage rates (k,,) are calculated for indigestible NDF (iNDF) and potentially digestible NDF (pdNDF); kr is calculated for iNDF only. Large particles @236mm) Small particles (< 2.36 mm) 110 Table 3. Least-squares means of responses in feed intake, digestion, and production of 14 Holstein cows to low-forage (LF) and high-forage (HF) diets. Treatment LSM1 Trt x Per LF HF SEM2 P P Yield, kgld Milk 39.6 36.1 3.5 0.0001 N83 3.5% FCM 41.5 39.1 3.8 < 0.01 NS Fat 1.51 1.46 0.15 0.13 NS Milk composition, % Fat 3.79 3.93 0.11 0.04 NS BW change, kg/15 d 8.2 -2.0 2.6 < 0.01 NS BCS change, kg/15d 0.02 -0.03 0.06 0.47 NS Intake, kg DM 27.8 24.7 0.9 < 0.0001 NS NDF 6.8 7.5 0.2 < 0.001 NS iNDF“ 3.6 3.8 0.1 0.66 NS Forage NDF, kg 6.7 5.6 0.2 < 0.0001 NS Rumen Pool, kg DM 10.8 11.1 0.5 0.09 NS NDF 5.7 6.2 0.3 < 0.01 NS INDF 3.1 3.5 0.1 < 0.0001 NS Rumen passage rate, lhr iNDF 0.06 Per 1 4.9 4.8 0.16 0.68 Per 2 4.8 3.9 0.16 < 0.01 pdNDF5 1.57 1.04 0.25 0.06 NS Starch 0.08 Per 1 21.9 23.1 3.1 0.79 Per 2 28.9 14.5 3.1 < 0.01 pdNDF digestion rate, lhr 3.31 4.74 0.29 0.001 NS Starch digestion rate, /hr <0.01 Per 1 23.4 33.1 2.1 < 0.01 Per 2 33.6 26.1 2.0 0.03 Rumen turnover time of NDF, h 0.16 Per 1 20.9 19.0 0.7 0.10 Per 2 20.4 21.6 0.7 0.32 NDF digested in the rumen kg 2.0 3.0 0.1 < 0.0001 NS % 29.3 40.1 2.0 < 0.001 NS Mean pH 5.86 6.00 0.04 < 0.0001 NS ‘ Treatment least-squares means. 2 Standard error of the mean. 3 Not significant (P 2 0.30), Trt x Per removed from model. 4 Indigestible NDF. . 5 Potentially digestible NDF = NDF — iNDF. 111 Table 4. Least-squares means of particle size kinetics responses of 14 Holstein cows to low-forage (LF) and high-forage (HF)diets. Treatment LSM1 LF HF SEM2 P Intake iNDF3 < 2.36 mm, kgld 2.35 1.67 0.04 < 0.0001 iNDF 2 2.36 mm, kgld 1.46 1.96 0.03 < 0.0001 pdNDF‘1 < 2.36 mm, kgld 1.96 1.51 0.06 < 0.0001 pdNDF 2 2.36 mm, kgld 1.85 2.32 0.07 < 0.0001 iNDF < 2.36 mm, % NDF 54.5 55.4 0.2 < 0.01 pdNDF < 2.36 mm, % NDF 45.5 44.6 0.2 < 0.01 iNDF 2 2.36 mm, % NDF 44.1 45.7 0.2 < 0.0001 pdNDF 2 2.36 mm, % NDF 55.9 54.3 0.2 < 0.0001 Rumen pool iNDF < 2.36 mm, kg 2.17 2.52 0.11 < 0.0001 iNDF 2 2.36 mm, kg 0.92 0.96 0.05 0.53 pdNDF < 2.36 mm, kg 1.43 1.50 0.11 0.25 pdNDF 2 2.36 mm, kg 1.16 1.21 0.08 0.48 iNDF <2.36 mm, % NDF 61.6 62.1 1.4 0.80 pdNDF < 2.36 mm, % NDF 38.4 37.9 1.4 0.80 iNDF 2 2.36 mm, % NDF 44.6 44.8 1.8 0.92 pdNDF 22.36 mm, % NDF 55.4 55.2 1.8 0.92 Passage rate, %Ih iNDF < 2.36 mm 6.10 5.66 0.19 0.03 iNDF 2 2.36 mm 2.85 2.10 0.25 0.02 pdNDF < 2.36 mm 2.34 1.70 0.29 0.06 pdNDF 2 2.36 mm 0.64 0.59 0.16 0.78 Rate of reduction (iNDF pool 22.36 to iNDF pool <2.36 mm), lh 3.87 6.86 0.51 < 0.0001 Duodenal flux 2 2.36 /total duodenal flux iNDF 0.15 0.13 0.009 0.03 pdNDF 0.16 0.21 0.02 < 0.05 1 Treatment least-squares means. 2 Standard error of the mean. 3 Indigestible NDF. ‘1 Potentially digestible NDF = NDF — iNDF. 112 Figure 3. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and the response to a low-forage diet (LF) over a high-forage diet (HF) of ruminal NDF turnover time (TOTu: — TOTHF = -5.6 - 0.24 pVDMl; P = 0.05). Equation is adjusted for Seq (Seq LF, HF = -1.08; Sequence HF, LF = 0). i .r: 2.0- % .1 3113 1.0- :: - “3 %g 00- E: - I-u- . -10- C 69 . o z u . . -20 I I ‘l fl I 20 22 24 26 28 30 32 pVDMI, kgld 113 Figure 4. Relationship between mean DMI during the final 4 d of the preliminary period (pVDMI) and mean daily ruminal pH. Across treatments, ruminal pH = 6.4 — 0.02 pVDMl (P = 0.11, R2 = 0.10, RMSE = 0.14). Mean pH was greater for HF than for LF (P < 0.0001), and relative response to treatment did not depend on pVDMI (P 2 0.60). 6.3 o 6.2- o Mean pH pVDMI, kgld +Diet HF -0- -Diet LF 114 CHAPTER 5 Predicting Ruminal Passage Rates of Fiber Fractions in Dairy Cattle ABSTRACT Passage rates of fiber fractions are important factors determining ruminal nutrient digestion, microbial protein production, efficiency, and flow to the duodenum, and the filling effects of a diet. Previous equations predicting passage rate have relied on measurements of ruminal disappearance or fecal appearance of external markers, which leads to inaccurate predictions. Data obtained in our laboratory from experiments utilizing the pool and flux method for estimating passage rates of digesta fractions were compiled and used to develop new regression equations predicting passage rates of indigestible NDF (iNDF) and potentially digestible NDF (pdNDF). Predictors used to develop the regression equations included dietary concentrations of NDF, forage NDF (forNDF), and starch; 30-h in vitro NDF digestibility of forages in the diet (forNDFD); DIM and BW; intake of DM, NDF, starch, and digested OM; MY, milk fat concentration, and 3.5% fat-corrected MY; ruminal pools of DM, NDF, and wet digesta. Equations were developed using both data that can be obtained by commercial dairy farms (e.g., DMI and diet composition) and data obtained in ruminal metabolism experiments (e.g., rumen pools). Predictions using data that can be obtained by dairy farms explained 68% and 53% of variation in passage rates of iNDF and pdNDF, respectively. The equations developed indicate that important predictors of passage rate that can be obtained by commercial dairy 115 farms include proportion of starch in the diet, DMI, forNDFD, proportion of forNDF in the diet, and FCMY. Improving predictions of passage rates will permit more efficient utilization of nitrogen and other nutrients and reduce their excretion as waste. INTRODUCTION To aid in formulating diets for dairy cows, numerous mathematical models of dairy cow digestion have been developed (Baldwin et al., 1987; Russell et al., 1992; NRC, 2001). These models estimate the availability of nutrients for milk production and other needs, given a particular set of feed composition characteristics, cow characteristics, and environmental factors. However, one of the factors most limiting to the accuracy of these models is their inability to account for the effects of dietary characteristics on voluntary feed intake and on the passage rate of digesta fractions from the rumen (lllius and Allen, 1994; Firkins et al., 1998). Wlthout an accurate prediction of passage rate, models cannot account for the effects of particle passage rate on feed intake or true protein flow to the duodenum. Most models overestimate both digestion rate and passage rate, and underestimate rumen pool size, because they rely on in vitro digestion of ground feeds and rare-earth or chromium marker passage data (Allen, 1996). Models of dairy cow digestion can be improved greatly by accurate predictions of passage rate. Therefore, the objective of this study was to develop new equations to predict passage rate of iNDF and pdNDF. Emphasis was placed on predictive 116 parameters that can be obtained by commercial dairy farms. The hypothesis was that important predictors would include DMI and those parameters that describe the potential of a diet to induce physical filling effects or metabolic factors to affect passage rate. MATERIALS AND METHODS Data sets from 11 studies conducted in our laboratory at Michigan State University between 1995 and 2003 were combined and used for estimations of ruminal passage rate of iNDF in dairy cattle (Table 1). Experimental procedures were approved by the Institutional Animal Care and Use Committee at Michigan State University. The data set included 254 animal-periods from multiparous lactating cows (nine studies), 29 animal-periods from primiparous lactating cows (one study), and 32 animal-periods from pregnant heifers (one study). All animals were ruminally and duodenally cannulated (gutter-style T cannulas); surgery was performed at the Department of Large Animal Clinical Science, College of Veterinary Medicine, Michigan State University. Nine studies followed Latin square designs and two followed crossover designs. Studies were designed originally to test hypotheses related to feed intake and ruminal and whole-tract digestion, and they were not designed for development of passage rate equations. Results of five of the 11 studies have been published in previous articles (Oba and Allen, 2000, 2003; Voelker and Allen, 2003; Taylor and Allen, 2005; Harvatine and Allen, 2006). Results of the other studies have been 117 reported in abstract form (Ying and Allen, 1998, 2005; Ying et. al, 1998; Mooney and Allen, 2004; Voelker Linton and Allen, 2005, 2006). Forages fed in the totally mixed rations during the studies were primarily alfalfa silage and(or) corn silage; one study utilized orchardgrass silage. Diets also included ground or cracked dry or high-moisture corn grain and a variety of protein and fat supplements, byproducts, and mineral supplements. All animals were fed ad libitum, at 110% of expected intake. Independent variables included in the data set (Table 2) were cow characteristics, chemical characteristics of diets and forages, parameters of intake and milk production that can be obtained by farms, and parameters of intake and digestion that can be measured in studies of ruminal and whole-tract digestion. All intake values were determined using weights and analysis of feed offered and orts. Rumen-empty BW was measured after evacuation of ruminal digesta on the day immediately before the start of the first period and on the final day of each period. Body condition score was determined on the same days by three trained investigators blinded to treatments (WIIdman et al., 1982; 1=thin and 5 = fat). Changes in BW and BCS are reported per day to correct for different period lengths across studies; the value of cow BW is from the day before the start of period 1. For 9 of the 11 studies, ruminal pH was monitored over a 96-h period using a computerized data acquisition system via an indwelling probe inserted through the ruminal fistula (Dado and Allen, 1993). For the other two studies, ruminal pH was measured in fresh ruminal fluid samples removed through the ruminal fistula every 9 h in a 72-h period. 118 Sample Collection and Analyses Passage rates of iNDF and pdNDF were measured using the pool and flux method (Oba and Allen, 2000). Most published equations predicting passage rate are determined by analysis of fecal excretion curves of external markers applied to intact forages and(or) concentrates and pulse dosed. By contrast, the pool and flux method requires a marker only to determine duodenal flux of digesta, so the marker does not need to flow with a specific digesta phase or fraction. Digesta markers differed among the 11 studies in the database and included a double-marker method (Cr-mordanted wheat straw and Co-EDTA), chromic oxide as an external marker, or iNDF as an internal marker. Dosing of external markers was spaced appropriately throughout the day to account for possible diurnal effects. Duodenal digesta were sampled every 9 h over a 72-h period. Ruminal contents were evacuated manually through the ruminal cannula approximately 4 h after feeding and 2 h before feeding on the second—to-last and last experimental days, respectively, of each period. Total ruminal content mass and volume were determined, and 10% aliquots were separated to allow accurate sampling of liquid and solid phases. Diet ingredients, orts, ruminal digesta, and duodenal digesta were analyzed for NDF according to Van Soest et al. (1991, method A), for iNDF after 120 or 240-h in vitro fermentation (Goering and Van Soest, 1970), and for starch by an enzymatic method (Karkalas, 1985) after samples were gelatinized with sodium hydroxide. Concentration of pdNDF was calculated as NDF (% DM) - iNDF (% DM). Original forage samples from 9 . of the 11 studies were analyzed together for 30—h in vitro digestibility (forNDFD; 119 Goering and Van Soest, 1970). Samples were unavailable for the remaining two studies; one of those studies had published values for forNDFD, so those were included in the data set. When both samples and original (published) forNDF D values were available, the new and old values were compared. They were similar, so the new values were used in the data set. Data were divided into two sets by randomly selecting two-thirds of the animal-periods from each study (a total of 210 records) to be assigned to a database that was used to develop models (BUILD), and then assigning the remaining one-third of the animal-periods from each study (a total of 105 records) to a database that was used to validate the models (VALIDATE). This horizontal division of the database was selected rather than a vertical division (using data from 2/3 of the studies for BUILD and 1/3 of the studies for VALIDATE), because differences in markers and slight differences in methods would have reduced the predictive power of the regressions developed in BUILD when applied to the VALIDATE set. Distributions of iNDF kp, pdNDF k,,, and several predictor variables in BUILD and VALIDATE are reported in Table 4. Ranges of several variables were smaller in VALIDATE than in BUILD. This is expected because VALIDATE is a smaller subsample of the original data set, and it is acceptable because the subsample was selected randomly. Statistical Analyses Regression analyses were performed for iNDF kp and for pdNDF kp using backward stepwise regression of JMP (Version 5.1.2, SAS Institute, Cary, NC) for the BUILD data set. Predictor variables were included in linear, quadratic, and 120 cubic terms; ‘if a higher-order term was significant, then all lower-order terms were also kept, regardless of their significance. Two-way interactions of main effects were included; three-way interactions were not included in order to avoid over-parameterization. Equations were developed by entering all potential predictors and removing predictors with the greatest P—value until all variables had P < 0.10. Overly influential animal-period records were identified by visual analysis of a distribution of Cook’s D influence statistic after initial backward regression. Records were excluded when Cook’s D was greater than 0.05 or when points were determined to be separate from the main cluster of values by visual examination. No more than 5% of records were removed during any regression operation. Backward regression was carried out again after removal of overly influential observations. One set of potential predictors (Model 1) included all of the available parameters that typically could be obtained on or by a dairy farm: DIM, BW, diet % NDF, % forNDF, % starch, forNDFD, DMI, DMI(%BW), MY, FCMY, FCMY/DMI, milk fat %, BW change, and BCS change. A second set of potential predictors (Model 2) reduced the number of redundant parameters (e.g., included FCMY alone instead of FCMY, MY, and % milk fat). That predictor set included diet % NDF, diet % forNDF, forNDFD, diet % starch, DMI, BCS change, and either FCMY (iNDF Model) or both MY and milk % fat (pdNDF Model). These variables were selected not only to avoid over-parameterization, but also to attempt to create prediction equations for iNDF and pdNDF kp that could be 121 incorporated into models of dairy cow digestion used for diet formulation or evaluation. A third set of potential predictors (Model 3) included parameters that would be measured in a study designed to estimate kp using the pool and flux method and that could be expected to be correlated with kp and to have some reasonably causal influence on kp (rather than being determined directly or primarily by k,,). These predictors included Model 1 parameters, plus intake of NDF, forNDF, and starch as a percent of BW, rumen pools of wet digesta, DM and NDF, rumen digesta volume, daily mean, variance and standard deviation of ruminal pH, and OM digested in the whole tract. Initial evaluation of equations was performed in the BUILD data set by visual inspection of plots of residuals (observed - predicted) against predicted values (Neter et al., 1996). Patterns suggesting systematic tendencies for residuals to be positive or negative were considered indicators that the model under consideration did not sufficiently account for variation. Percentage of variation accounted for by an equation (R2) and Akaike’s Information Criterion (AIC; Akaike, 1974) also were used to determine the predictive value of a candidate equation. The AIC is calculated using the equation AIC = 2p + n ln(SSE / n), where p is the number of parameters, n is the number of observations, and SSE is the sum of squares of error (residuals). It examines the complexity of a model together with goodness of fit to the sample data, in order to find the minimal model that correctly explains the data. A lower AIC value indicates a more appropriate model. 122 After equations were developed using the BUILD data set and selected as candidate models, they were evaluated using the VALIDATE data set according to recommendations of Neter et al. (1996) and St—Pierre (2003). Each prediction equation for iNDF kp developed in BUILD as used to calculate a set of predicted values for kp in VALIDATE. Residuals (observed — predicted) were calculated, and predicted values were centered by subtracting the mean of all predicted values from each predicted value (St-Pierre, 2003). Residuals were regressed against centered predicted values and evaluated as recommended by St-Pierre (2003). Centering the predicted values places the intercept of the regression of residuals against predicted at the mean predicted value rather than at zero. This permits a t-test of the regression intercept to determine the statistical significance of mean bias, and a t-test of the regression slope to determine the statistical significance of linear bias. When the linear bias was statistically significant (P < 0.05), the magnitude of the bias was calculated for the maximum and minimum predicted values, using the following equation: e, = bo+ b1(Xi-X) where e is the linear bias at i (maximum or minimum), be is the intercept of the regression of residuals on centered predicted values, b1 is the slope of the regression of residuals on centered predicted values, X. is the maximum or minimum predicted value, and X is the mean of all predicted values. 123 The bias at the maximum and minimum predicted values was then judged relative to the size of the standard error of the regression. If either of the two calculated biases was greater than the standard error of the residual regression, then the linear bias was considered mathematically significant. The effect of study was not included in the regressions performed in this analysis. Modern statistical software allows the Study effect to be included as a random effect in order to account for differences between studies such as experimental design, methods, and physiological status of animals, as well as to account for the mean and linear bias caused by Study (St-Pierre, 2001). However, the experiments from which these data were obtained used similar methods, which reduces the need to account for Study effect. Including Study effect in the iNDF models resulted in higher R2 and lower AIC values for Model 2 and completely eliminated mean and linear bias from Models 2 and 3. However, the proportion of the range in actual predictor values that is accounted for by the range in mean (by study) predictor values was greater for diet-related parameters (e.g., dietary starch concentration) than for response-related parameters (e.g., MY). Thus, including Study as a random effect would remove more variation caused by diet factors than variation caused by other factors. Indeed, including Study resulted in the removal of one diet-related predictor, forNDF D, but no response-related predictor, from Model 2. Therefore, to avoid biasing the predictive power away from diet parameters and toward response parameters, the Study effect was not included in the models. 124 RESULTS AND DISCUSSION Measuring Passage Rates of iNDF and pdNDF This is the only large-scale summary of data known for passage rate of digesta NDF fractions calculated using the pool and flux method. Nearly all passage data available in the literature were measured by analysis of fecal excretion curves of external markers applied to intact forages and(or) concentrates and pulse dosed. Problems with external markers are well documented (F irkins et al., 1998). While these data might be useful to evaluate relative differences among treatments within experiments, they are not useful to predict digestibility of fractions within feeds across a wide range of conditions. Absolute measurements are required for passage rates when they are used with digestion rates to predict digestibility. Other problems with external markers limit their usefulness even as a measure of passage rate of entire feeds. Problems include extensive migration from the labeled feeds (Teeter et al. 1984; Combs et al., 1992), preferential binding to small particles (Erdman and Smith, 1985), and a reduction in digestion rate which can increase density and passage rate (Firkins et al., 1998). Feeds intrinsically labeled with 1‘10 have been used (Holden et al., 1994) but this method is time consuming, expensive, and biased if the 1‘1C is not evenly distributed in the feed (Firkins et al., 1998). An additional problem is that passage rates usually are calculated by analysis of fecal excretion curves, the results of which are difficult to interpret. Two or more significant pools and rates can 'be determined, but it is not clear which rate represents passage from the rumen or 125 even that assignment of the resulting mathematically-defined pools to specific biological pools is valid. Finally, current predictions of ruminal digestibility of digesta fractions (e.g., starch, neutral detergent fiber (NDF), and protein fractions) are calculated using the digestion rates of those fractions and the passage rates of the individual feed ingredients that contain those fractions. Using the passage rates of feed ingredients produces inaccurate predictions for ruminal digestibility of digesta fractions, because the different fractions within a feed ingredient escape the rumen at different rates. Ruminal digestibility is determined for digesta chemical fractions, not for feed ingredients, so both the digestion rate and the passage used to predict digestibility ideally should be for fractions, not for ingredients. Passage rate data for the various chemical fractions have been either completely unavailable or limited until recently, when the development and increasing use of the pool and flux method resulted in the production of more data for passage rate of digesta fractions. To directly measure passage rates of digesta fractions, the fractional passage rates of individual, uniform digesta fractions can be calculated by dividing duodenal flux of the fraction by its ruminal pool size. This pool and flux method was used to obtain the data for this analysis. Mean iNDF kp in the BUILD data set was 3.17 h'1 and ranged from 1.04 to 5.81 (T able 4). In a recent summary, Seo et al. (2006) reported passage rates for dry forages, wet forages, and concentrates, estimated using rare earth markers, of 4.53, 5.17, and 6.69 h'1, respectively, with ranges of 3.42 to 5.70, 3.9 to 6.29, and 3.61 to 9.22 h'1, respectively. All three feed types contain some 126 iNDF; as a proportion of DM, forages contain more iNDF than do concentrates because they contain more total NDF. Mean iNDF kp reported here was lower than kp reported for forages by Seo et al (2006). In addition to iNDF, the NDF fraction of whole forages also contains a significant amount of potentially digestible NDF (pdNDF), which has a slower kp (2.35 51.05 h’1 in BUILD, Table 4) likely due to a greater concentration of fermentation gasses in pdNDF-rich particles (Allen, 1996). However, smaller particles likely contain a greater fraction of iNDF in total NDF than do large particles (see Chapter 3), and the pool and flux method measures passage rate of all iNDF, not only forage iNDF. As a result, iNDF passage rate is likely more representative of smaller, denser particles than are measured in marked forages and should therefore be faster, not slower, than is measured by marked forages. Furthermore, markers frequently increase the density of particles to which they are attached (Ehle et al., 1984), and rare earths migrate into the small particle and liquid pools (Erdman and Smith, 1985), so the actual passage rates of marked forages are artificially inflated compared to the passage rates of unmarked forages. This is why the ruminal passage rate of iNDF estimated using the pool and flux method is lower than the passage rates estimated for whole, marked forages. The overprediction of digesta passage rate with rare earths and other external markers, combined with actual digestion rate, leads to inaccurate estimates of ruminal digestibility and duodenal passage of nutrients in models used to formulate or evaluate dairy cow diets. 127 Passage rate of pdNDF demonstrated a width of range similar to that of iNDF, and the mean and range were approximately one unit lower than the mean . and range of iNDF kp (Table 4). More true and method-associated error are expected with pdNDF than with iNDF, because pdNDF can be removed from the rumen through digestion as well as through passage, and because pdNDF concentration is calculated using measured iNDF concentration. However, the data from these eleven experiments suggest similar variation in measurements of passage for both NDF fractions. Predicting iNDF kp Using Farm Data A large number of parameters that can be obtained by commercial dairy operations demonstrated the potential to predict iNDF kp (Table 2). Backward stepwise regression considering all of these parameters (iNDF Model 1) resulted in very strong predictive power within BUILD (R2 = 0.94, AIC = -156; Table 5) but also in significant over-parameterization and a very weak capability to predict kp in VALIDATE, as evidenced by significant mean (P < 0.0001) and linear (P < 0.0001) biases (Table 5, Figure 1). A description of the predictors included in iNDF Model 1 is presented in Table 6a. Beginning backward regression with a much smaller pool of potential predictors (iNDF Model 2), which were selected for mechanistic importance and to avoid redundancy, resulted in lower predictive power within BUILD (R2 = 0.67) and a less favorable (higher) AIC value (-96; Table 5). However, when the resulting equation was evaluated in VALIDATE, no mean bias existed (P = 0.86; Figure 2). Although linear bias was statistically significant (P < 0.001; Table 5), bias at both the minimum and maximum 128 predicted values was 0.68, which is smaller than standard error of the residual regression (0.87) and therefore biologically insignificant. The model was able to account for a surprisingly large proportion (68%) of variation in iNDF kp (Table 5). Therefore, for applications in models used for diet formulation or evaluation on dairy farms, the equation created in iNDF Model 2 is likely the most appropriate equation for predicting iNDF kp using the parameters available in this data set. A description of the predictors included in iNDF Model 2 is in Table 6b. The iNDF Model 2 accounts for the effects of: (1) proportion of starch in the diet, (2) DMI, (3) forNDFD, (4) proportion of forNDF in the diet, and (5) FCMY. Two- way interactions between diet % starch and forNDFD, DMI, and FCMY, and between DMI and FCMY, also contributed significantly to the prediction of iNDF kp (Table 6b). Direct mechanistic interpretation of the equation is not practical because two quadratic, two cubic terms, and four interactions were included, but biological evidence exists for effects on kp of the parameters selected. Proportion of concentrate in the diet was determined to be the most significant predictor in the NRC (2001) calculations for passage rate of dry forage and concentrate, but not for wet forage, which is the source of the majority of the iNDF in the diets in the present data set. The relationship observed here between diet % starch and iNDF kp was cubic; Seo et al. (2006) reported varying effects of increased diet concentrate on passage rates of concentrate and dry forages. Increasing grain content of the diet can increase passage rate (Grovum, 1986), but starch fermentation might reduce passage rate of digesta in general, or of iNDF in particular, by interfering with fiber digestion (Grant and Mertens, 129 1992) or through effects of ruminal pH on digestion and ruminal motility (Allen et al., 2006). Observed effects of increasing dietary grain concentration on passage rate likely depend on several factors. These include the relative proportions of forage and grain in the diet, the fermentability of the grain (i.e., conservation method, moisture content, vitreousness, and particle size), forage particle size, and the rate of NDF digestion and particle size reduction of the forage. A positive correlation between DMI and kp has often been assumed, but until recently, data were lacking to confirm this assumption (lllius and Allen, 1994). Predictions of passage rate in the 2001 NRC protein model (NRC, 2001), and the recent re-evaluation of those equations (Seo et al., 2006) both included DMI (as a percentage of BW) as a very important factor in predicting k,,; as DMI increases, kp increases (NRC, 2001). In Model 2, DMI was used alone, rather than as DMI (%BVV), because commercial dairy farms may not be equipped to obtain actual BW. Interestingly, effects of FCMY on k.D were not completely accounted for through DMI but needed to be included separately. Mechanisms by which nutrient demand affects passage rate might include increased ruminoreticular contraction rate, strength, or duration. Digestibility of forNDF, estimated by 30-h in vitro fermentation, also contributed to the prediction of iNDF kp (Table 6b). Within forage family (i.e., grasses or legumes), greater in vitro digestibility of forNDF usually results in greater DMI (Oba and Allen, 1999b), which suggests that ruminal passage rate is also increased with greater NDF digestibility. However, passage rate is slower 130 for grass despite greater NDFD (Chapter 2). The relationship between in vitro digestibility and passage rate is complicated not only by forage family but also by the fact that pdNDF usually exhibits a slower kp than does iNDF, as mentioned above. Therefore, NDF digestibility is not a proxy for rate of NDF digestion; a clear example of this is the generally slow rate, but high extent, of NDF digestion in perennial grasses (Wilson and Hatfield, 1997). Rate of NDF digestion is not commonly measured for forages used on commercial dairy farms, but the measurement of in vitro NDF digestibility for forages is becoming increasingly common. Actual NDF concentrations of forages (as opposed to values obtained from tables) also are increasingly available to dairy farms, so dietary forNDF concentration can be calculated. Furthermore, recommendations for NDF and non-forage carbohydrate concentrations in the most recent NRC (2001) include a minimum dietary forNDF concentration, in addition to minimum total dietary NDF and maximum total dietary non-forage carbohydrate. Generally, greater forNDF concentration would be expected to result in slower iNDF kp. Effects of dietary forNDF concentration will depend on the digestibility of that forNDF and also on forage particle size. In the studies from which the present data set was obtained, particle size was seldom measured and so could not be used in prediction equations. Because forage type did not vary widely (diets contained primarily corn silages and(or) alfalfa silages) and chop lengths were similar across experiments, variation in forage particle size was likely much smaller than variation present in diets fed across the US. Quantifying the effects of particle 131 size on feed intake or nutrient digestibility continues to be a significant challenge, and sensitivity to particle size would be low in a data set with relatively small particle size variation. Predicting iNDF kp in Research Studies A third model was tested to predict iNDF kp using data that routinely are obtained in studies designed to estimate kp using the pool and flux method. To the “farm” parameters (iNDF Model 2) were added forNDF intake, rumen pools of wet digesta, DM, and NDF, rumen digesta volume, daily mean and variance of ruminal pH, and digested OM intake. Results of backward regression using these parameters (iNDF Model 3) are presented in Tables 4 and 5c. Surprisingly, although a large number of terms were included in the model, no mean or linear bias existed when the model was evaluated in VALIDATE (Figure 3). The only cow descriptor included was BW; diet characteristics included dietary concentration (% DM) of total NDF, forNDF, and starch, and forNDFD. Instead of DMI, NDF intake, NDF intake (%BVV), starch intake (% BVV), and OM digested in the whole tract were significant predictors of iNDF kp. These parameters likely account for effects of DMI on kp. Intake of NDF likely accounts for some variation in kp due to physical filling effects. No parameters of milk production were included, but BCS change (unit/d) contributed significantly to the model. This suggests that energy balance might affect passage rate, possibly through physiological controls such as rumen motility and reticular contraction frequency or duration. Rumen DM pool was the only significant predictor that could not be measured in intact animals. Creating acceptable passage rate 132 prediction equations that include only parameters that can be measured in intact animals would be preferable to creating equations that require cannulation surgery for continued use in predicting passage rate. Predicting pdNDF kp Using Farm Data Fewer parameters that can be obtained by commercial dairy operations demonstrated the potential to predict pdNDF kp (Table 3) compared to iNDF kp (Table 2). Backward stepwise regression considering all of these parameters (pdNDF Model 1) resulted in strong predictive power within BUILD (R2 = 0.85, AIC = -127; Table 5) but also in significant over-parameterization and a very weak capability to predict RF) in VALIDATE, as evidenced by significant mean (P = 0.05) and linear (P < 0.0001) biases (T able 5, Figure 4). A description of the predictors included in pdNDF Model 1 is presented in Table 7a. Beginning backward regression with a smaller pool of potential predictors (pdNDF Model 2), which were selected for mechanistic importance and to avoid redundancy, resulted in lower predictive power within BUILD (R2 = 0.53) and a less favorable (higher) AIC value (-85; Table 5). However, when the resulting equation was evaluated in VALIDATE, no mean bias existed (P = 0.43; Figure 5). Although linear bias was statistically significant (P = 0.02; Table 5), the maximum bias (0.65) was smaller than standard error of the residual regression (0.95) and therefore mathematically insignificant. Therefore, for applications in models used for diet formulation or evaluation for commercial dairy farms, the equation created in pdNDF Model 2 is likely the most appropriate equation for predicting iNDF kp using the parameters 133 available in this data set. Because pdNDF can be removed from the rumen by digestion as well as by passage, the accuracy of prediction of pdNDF kp might be lower than the accuracy of prediction of iNDF kp, as demonstrated by the lower R2 value for pdNDF Model 2 compared to iNDF Model 2 (Table 5). A description of the predictors included in pdNDF Model 2 is presented in Table 7b. The pdNDF Model 2 accounts for the effects of: (1) proportion of starch in the diet, (2) MY, (3) proportion of forNDF in the diet, (4) proportion of total NDF in the diet, (5) DMI, (6) BCS change, and (7) forNDFD. Two-way interactions between diet % NDF and diet % forNDF, MY, and forNDFD, between diet % forNDF and forNDFD and BCS change, and between diet % starch and BCS change, also contributed significantly to the prediction of iNDF kp (Table 7b). The biological importance of diet % starch, diet % forNDF, diet % NDF, and forNDFD in affecting kp were discussed previously. Two different predictors were included in pdNDF Model 2 compared to iNDF Model 2. The pdNDF Model 2 (Table 7b) included MY instead of FCMY, and BCS change (unit/d), which iNDF Model 2 did not include (Table 6b). However, MY, BCS change, and FCMY are all indicators of nutrient demand and utilization. The mechanisms for these effects are multifactorial and many are yet to be demonstrated empirically; however, in general, greater nutrient demand likely results in greater passage rate in order to permit greater DMI. A negative linear relationship between pdNDF kp and BCS change (data not shown) indicates that more positive energy balance (gaining more body tissue) was associated with slower pdNDF k,,. 134 Predicting pdNDF kp in Research Studies A third model was tested to predict pdNDF kp using the same predictors used for iNDF Model 3 (see above). Results of backward regression using these parameters (pNDF Model 3) are presented in Tables 4 and 6c. As with iNDF Model 3, no mean or linear bias existed when the pdNDF Model 3 was evaluated in VALIDATE (Table 5, Figure 6). However, the predictive power of pdNDF Model 3 was much lower (R2 = 0.53) than the predictive power of iNDF Model 3 (R2 = 0.91; Table 5). Predictors included in pdNDF Model 3 (Table 7c) were different from predictors included in iNDF Model 3 (Table 6c). Model 3 for pdNDF added rumen pools of DM, NDF, and wet digesta, and the standard deviation of ruminal pH, to diet forNDF concentration, DMI, FCMY, and BW change. The iNDF model included more diet and intake descriptors, and the pdNDF model included more rumen pool descriptors. Even though forNDFD is related directly to the concentration of pdNDF in the forages included in the diets, forNDFD did not add to the prediction of pdNDF k,D when ruminal pool and pH data were available. Only the pdNDF model, not the iNDF model, included a measure of ruminal pH; increasing pH standard deviation was related negatively to pdNDF kp (Table 7c). More variation in ruminal pH throughout the day might reduce ruminal motility, which can reduce kp of digesta, and more variation in pH likely slows the rate of pdNDF digestion, which can decrease the specific gravity of particles containing pdNDF and thus reduce the k,, of pdNDF. 135 Limitations to Application of Developed Models The application of any model is limited, in particular, by the range of data from which it was developed and the segment of the population from which that data was obtained. Data for these models were obtained from experiments conducted by a single using animals from a single farm. Therefore, although the database demonstrated wide variation in diet characteristics (e.g., dietary NDF concentration range was 22.6 — 38.0), ingredients (particularly, forages and grains) used in the diets varied less than would be observed in a random sample of US. dairy farms. Furthermore, the animals utilized in the original experiments were samples from one farm over the course of 10 years and therefore represent far less genetic variation than exists across farms throughout the US. Therefore, the application of the current stage of these models to predict passage rate is very limited. Future prediction equations for passage rate utilizing data obtained through the pool and flux method should represent, at least, a wider range in dietary ingredients. Thus, the significance of these models lies not in their ability to predict passage rate across farms but rather in the biological significance of the variables that contributed to the prediction of passage rate. CONCLUSIONS AND IMPLICATIONS Prediction of passage rate of digesta is dependent on diet characteristics and nutrient demand of the individual animal. A model including diet % starch, DMI, forNDFD, diet % forNDF, and 3.5% FCMY accounted for 68% of variation in iNDF kp. A model including diet % starch, MY, diet % forNDF, diet % NDF, DMI, 136 BCS change, and forNDF D accounted for 53% of variation in pdNDF kp. Passage rate of digesta fractions is seldom measured directly, but strategic production of data sets containing more easily-measured parameters along with passage rates can increase the accuracy of the prediction of passage rates in models intended for use on commercial dairy farms. Improving predictions of passage rates will permit more efficient utilization of N and other nutrients and reduce their excretion as waste. 137 Table 1. Studies included in the data set used for development of passage rate equations in dairy cattle. Study Description 1 Treatments were coarsely or finely ground dry or high-moisture corn grain, fed to pregnant heifers (Ying and Allen, 1998) 2 Treatments were coarsely or finely ground dry or high-moisture corn grain, fed to primiparous lactating cows (Ying and Allen, 1998) 3 Treatments were high- or low-NDF diets containing normal or brown- midrib corn silage (Oba and Allen, 2000) 4 Treatments were high- or low-starch diets containing ground high- moisture or dry corn grain (Oba and Allen, 2003) 5 Treatments were dried, pelleted beet pulp substituted for high- moisture corn grain at 0 (control), 6, 12, and 24% of diet DM (Voelker and Allen, 2003) 6 Treatments were sodium bicarbonate, sodium chloride, or no added ions (Mooney and Allen, 2004) 7 Treatments were brown-midrib or control corn silage and floury or vitreous varieties of corn grain (Taylor and Allen, 2005) 8 Treatments were no added fat (control), saturated fat supplement, unsaturated fat supplement, or 50% of each fat supplement (Harvatine and Allen, 2006) 9 Treatments were dry or high-moisture preserved of floury or vitreous varieties of corn (Ying and Allen, 2005) 10 Treatments were alfalfa silage or orchardgrass silage (Voelker Linton and Allen, 2005) 11 Treatments were diets containing high or low concentrations of forage NDF (Voelker and Allen, 2006) 138 Table 2. Independent variables included in the data set used for development of iNDF passage rate equations in daiy cattle. Correlation to iNDF k2 Linear Quadratic Cubic P r P r P r Days in Milk < 0.01 0.22 0.09 -0.26 < 0.01 0.35 BW1, kg < 0.0001 0.32 NS ---— NS ---— Diet % NDF NS ---— < 0.0001 0.32 0.13 -0.34 Diet % forNDF2 NS ---- < 0.01 0.22 0.02 0.28 Diet % starch < 0.0001 0.57 < 0.01 0.59 NS --- ForNDFD3 % < 0.0001 0.29 < 0.0001 -0.45 NS ..-.. DMI, kgld < 0.0001 0.47 NS ---- 0.05 -0.49 DMI,%BW < 0.0001 0.39 NS ----- NS ---- NDF intake, kgld < 0.0001 0.51 NS ---- 0.02 -0.53 ForNDF intake, kgld < 0.0001 0.54 NS ---- 0.02 -0.55 Starch intake, kgld < 0.01 0.24 NS ----- 0.03 0.29 MY, kgld NS ---- NS ---- 0.12 0.15 FCMY“, kgld NS ---- NS ---- 0.11 -013 FCMY/DMI < 0.0001 -0.31 NS ---- NS ---- Milk fat % NS ---- < 0.001 -0.29 NS ---- BW change, kgld NS ---- NS --- NS ---- BCS change, /d NS ----- NS ---- NS ---- Ruminal DM pool, kg NS ----- < 0.0001 -0.29 NS ---- Ruminal NDF pool, kg NS ---— NS ---- 0.03 -0.17 Ruminal digesta, kg NS ----- < 0.01 0.21 NS ---- Ruminal digesta, L NS --- 0.01 -O.20 NS ---- Digested OM, kgld < 0.0001 0.32 < 0.001 -0.40 0.04 -0.43 Ruminal pH mean < 0.001 -0.26 0.08 0.29 0.02 0.30 pH variance < 0.01 0.25 0.01 -0.32 0.01 0.38 pH standard deviation 0.02 0.17 0.04 -0.23 NS ---- 1 Rumen-empty body weight. 2 Forage NDF. 3 30-h in vitro NDF digestibility. ‘1 3.5% fat-corrected milk yield. 139 Table 3. Independent variables included in the data set used for development of pdNDF pasgge rate equations in dairy cattle. Correlation to pdNDF kL Linear Quadratic Cubic P r P r P r Days in Milk < 0.01 -0.22 0.18 0.25 NS ---— BW1, kg NS ---- NS ---- NS Diet % NDF NS ----- NS ----- NS ---- Diet % forNDF 2 NS ---- NS ---- NS ---- Diet % starch NS ---- < 0.01 0.23 NS ---- ForNDFD3 % NS ---- 0.02 -0.19 NS ----. DMI, kgld NS ----- < 0.01 -0.23 NS ---- DMI,%BW NS ---- 0.11 -0.16 0.03 0.16 NDF intake, kgld NS ---- NS ---- < 0.01 0.22 ForNDF intake, kgld NS ---- NS ----- 0.001 0.24 Starch intake, kgld NS ----- NS ---- 0.05 0.17 MY, kg/d 0.01 -0.19 < 0.01 -0.28 0.01 0.11 FCMY“, kg/d 0.04 -0.16 < 0.01 -0.14 0.02 0.33 FCMY/DMI 0.19 -0.10 < 0.01 -0.26 NS --- Milk fat % 0.12 0.12 0.19 0.16 < 0.10 -0.20 BW change, kgld NS ---- NS ---- NS ---- BCS change, d'1 0.17 -010 0.19 0.14 NS ---- Rumen DM pool, kg 0.02 -0.17 < 0.001 -0.30 < 0.01 0.37 Rumen NDF pool, kg 0.11 -0.12 < 0.01 -0.23 0.01 0.29 Ruminal contents, kg 0.05 -0.14 < 0.0001 -0.32 NS ---- Ruminal contents, L NS ---- 0.02 -0.18 0.001 0.29 Digested OM, kgld NS ----- < 0.01 -0.20 NS ---— Ruminal pH mean NS ----- NS ---- 0.07 -0.15 pH variance NS ----- NS ----- NS ---- pH standard deviation < 0.01 -0.23 0.01 -0.30 NS ---- 1 Rumen-empty body weight. 2 Forage NDF. 3 3041 in vitro NDF digestibility. ‘1 3.5% fat-corrected milk yield. 140 Table 4. Comparison of distributions of iNDF and pdNDF kp, descriptors, and potential predictor variables in BUILD and VALIDATE data sets used for development of passage rate equations in dairy cattle. BUILD VALIDATE Variable N Range Mean 8.0. N Range Mean SD. DIM 173 32 - 388 94 69.7 90 32 - 388 91.9 72.5 BW1, kg 193 396 - 760 557 67 100 414 - 753 557 64 Diet % NDF 194 22.9 - 38.0 28.0 3.1 100 22.6 - 36.4 27.6 2.9 Diet % forNDFz 194 16.6 - 29.5 21.1 4.1 100 16.6 - 29.5 20.7 4.1 ForNDFD,3 % 175 29.7 - 65.9 42.9 8.9 99 29.7 - 65.9 43.0 8.7 DMI, kgld 194 3.2 - 34.1 21.1 6.0 99 2.8 - 29.9 21.7 6.0 DMI, %BW 193 0.74 - 5.84 3.79 1.01 99 0.68 - 5.91 3.88 1.00 MY, kgld 173 9.9 - 59.8 36.9 9.4 89 20.5 - 56.3 38.9 8.3 FCMY“, kgld 189 8.5 - 60.7 37.4 8.8 89 21.5 - 62.8 39.2 8.1 Milk fat % 173 1.62 - 6.42 6.66 0.68 89 2.24 - 5.17 3.61 0.58 iNDka,5h'1 173 1.04-5.81 3.17 1.16 69 1.06-5.60 3.05 1.03 pdNDF kp, ‘1 h'1 188 -1.41 - 4.72 2.35 1.05 99 0.07 - 4.84 2.25 1.07 1 Rumen-empty body weight. 2 Forage NDF. 2 30-h in vitro NDF digestibility. ‘1 3.5% fat-corrected milk yield. 5 Indigestible NDF passage rate. ‘1 Potentially digestible NDF passage rate. 141 Table 5. Diagnostic statistics for building and validating models for prediction of passage rate of iNDF and pdNDF. Diagnostics in BUILD Mean and linear bias in VALIDATE Mean Linear Linear Modela n R2 RMSE Alcb bias P° bias Pd sd bias?“ iNDF 1 146 0.943 0.473 -156 <0.0001 <0.0001 1.06 Yes 2 148 0.676 0.687 -96 0.86 <0.001 0.88 No 3 167 0.910 0.382 -296 0.70 0.11 0.55 No pdNDF 1 148 0.854 0.561 -127 0.05 <0.0001 1.06 Yes 2 148 0.527 0.710 -85 . 0.43 0.02 0.95 No 3 143 0.529 0.741 -68 0.72 0.35 0.85 No a See text for description of statistical analyses and see Tables 5(a-c) and 6(a-c) for descriptor statistics. 1’ Akaike’s Information Criterion (AIC) = 2p + n ln(SSE / n), where p is the number of parameters, n is the number of observations, and SS is the sum of squares of the error (residuals). ° Mean bias is the intercept of the regression of residuals against centered predicted values (predicted values - mean predicted value). P < 0.05 indicates significant mean bias (St-Pierre, 2003). ‘1 Linear bias is the slope of the regression of residuals against centered predicted values (predicted values - mean predicted value). P < 0.05 indicates statistically significant linear bias. When the linear bias was statistically significant (P < 0.05), the magnitude of the bias was calculated for the maximum and minimum predicted values, using the equation ei = be + b1 (Xi — X) and then judged relative to the size of the standard error (s) of the regression of residuals on centered predicted values. See text for more details. 142 Figure 1. (A) Plot of residual (observed minus predicted) iNDF kp vs. predicted iNDF kp resulting from Model 1 in BUILD data set. (B) Plot of residual iNDF kp vs. predicted iNDF kp resulting from Model 1 applied to VALIDATE data set. Predicted iNDF kp was centered around the mean predicted value. Both mean bias (0736403) and linear bias (-0.99) were significant (P < 0.0001). Different symbols represent data from individual studies. 3’ . Y . .z @310. Residual mor k, (h‘1) O Predicted INDF k, (ll-1) 143 Residual iNDF k, 01") O I I I I I I I -20 -15 -10 -5 0 5 10 15 20 25 Centered Predicted INDF k" (h") iNDF k, = -0.736 (10.122) - 0.991 (10.027) (x - 3.66) R2 = 0.95, s a 1.06, P < 0.0001 Bias at min. predicted (-13.00) = 16.0 Bias at max. predicted (26.84) = 23.5 144 Figure 2. (A) Plot of residual (observed minus predicted) iNDF kp vs. predicted iNDF kp resulting from Model 2 in BUILD data set. (B) Plot of residual iNDF kp vs. predicted iNDF kp resulting from Model 2 applied to VALIDATE data set. Predicted iNDF kp was centered around the mean predicted value. Mean bias was not significant (P > 0.85). Although linear bias (-0.374) was significant (P < 0.001), maximum bias (0.68) was lower than standard error of residuals (0.88). Different symbols represent data from individual studies. > Residual iNDF kp (h'1) Predicted INDF k" (h") 145 Residual iNDF k, (h’1) -2 l l l I l I l I I -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Centered Predicted INDF kp (h'I) iNDF kp = 0.0183 (£01000) - 0.374 (10.105) (X - 3.15) R2 = 0.15, s = 0.878, P < 0.001 Bias at min. predicted (0.736) = 0.68 Bias at max. predicted (5.01) = 0.68 146 Figure 3. (A) Plot of residual (observed minus predicted) iNDF kp vs. predicted iNDF kp resulting from Model 3 in BUILD data set. (B) Plot of residual iNDF kp vs. predicted iNDF kp resulting from Model 3 applied to VALIDATE data set. Predicted iNDF kp was centered around the mean predicted value. Mean bias was not significant (P > 0.70), nor was linear bias (P > 0.10). Different symbols represent data from individual studies. A 2.0 1.5 " 1.0 ‘ 0.5 " 0.0 ' Residual iNDF kp (h'I) 2.0 1.5 '1 1.0 d 0.5 " 0.0 - -0.5 - -1.0 - -1.5 - -2.0 - CI -2.5 Residual iNDF kp (h'1) I I I I I I l I -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Centered Predicted iNDF kp (h'I) iNDF kp = -o.0227 (10.0597) - 0.0970 (10.0597) (x - 3.06) R2 = 0.030, s = 0.554, P = 0.11 147 Figure 4. (A) Plot of residual (observed minus predicted) pdNDF kp vs. predicted pdNDF kp resulting from Model 1 in BUILD data set. (B) Plot of residual pdNDF kp vs. predicted pdNDF kp resulting from Model 1 applied to VALIDATE data set. Predicted pdNDF kp was centered around the mean predicted value. Both mean bias (0.24) and linear bias (-1.05) were significant (P = 0.05, P < 0.0001, respectively). Different symbols represent data from individual studies. A 1.5 ' Residual pdNDF kp (h’1) .1 0 1 2 3 4 5 Predicted pdNDF k, (h") 148 Residual pdNDF kp (h") A I .3 O I F l I I I I I -8 -6 -4 -2 0 2 4 6 8 Centered Predicted pdNDF kp (h") pdNDF kp = 0.239 (10.122) - 1.05 (20.0496) (X - 2.05) R2 = 0.86, s = 1.06, P < 0.0001 Bias at min. predicted (-4.97) = 7.6 Bias at max. predicted (9.47) =7.5 149 Figure 5. (A) Plot of residual (observed minus predicted) pdNDF kp vs. predicted pdNDF kp resulting from Model 2 in BUILD data set. (B) Plot of residual pdNDF kp vs. predicted pdNDF kp resulting from Model 2 applied to VALIDATE data set. Predicted pdNDF kp was centered around the mean predicted value. Mean bias was not significant (P > 0.43). Although linear bias (-0.36) was significant (P = 0.02), absolute value of the maximum bias (0.65 at maximum predicted value of 3.95) was lower than standard error of residuals (0.95). Different symbols represent data from individual studies. A 2.0 1: Residual pdNDF kp (h'1) Predicted pdNDF kn (h") 150 ‘ N I kp (W1) 0 Residual pdNDF I l N l I T T I I j I -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Centered Predicted pdNDF kp (h'I) pdNDF kp = -0.0865 (10.109) - 0.357 (10.147) (X - 2.36) R2 = 0.07, s = 0.954, P = 0.02 Bias at min. predicted (0.58) = 0.55 Bias at max. predicted (3.95) =-0.65 151 Figure 6. (A) Plot of residual (observed minus predicted) pdNDF kp vs. predicted pdNDF kp resulting from Model 3 in BUILD data set. (B) Plot of residual pdNDF kp vs. predicted pdNDF kp resulting from Model 3 applied to VALIDATE data set. Predicted pdNDF kp was centered around the mean predicted value. Mean bias was not significant (P > 0.72), nor was linear bias (P > 0.35). Different symbols represent data from individual studies. A 2.5 * 2.0" 1.5" 1.0" 0.5" 0.0 -0.5" -1.0" -1.5'1 Residual pdNDF kp (h") Predicted pdNDF k, (h'1) 152 E” or .1) k,(h d” «a 1.0‘ 0.5' 0.0" -0.5" -1.0‘ -1.5" '2.0 I I T I I I I -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Centered Predicted pdNDF kp (h-I) Residual pdNDF pdNDF k9 = -0.0348 (10.0970) -0.137 (10.146) (x - 2.30) R2 = 0.01, s = 0.651, P = 0.35 153 Table 6a. Descriptive statistics of Model 1 for passage rate of indigestible NDF (iNDF kn). created using backward stepwise regression, ranked by total contribution to the prediction (Sum of Squares) of each parameter (sums of linear, quadratic and cubic terms). The regression utilized 146 animal-period observations. Because of the large size and significant mean and linear bias of Model 1, the full model is not included here. Parameter Estimate Sum of Squares P Intercept 6870 --— <0.0001 FCMY“ -61.7 6.09 <0.0001 FCMYZ -0.619 4.30 <0.0001 FCMY3 0.00289 3.97 0.0001 BW” -16.4 5.90 <0.0001 6W2 0.0136 5.51 <0.0001 BW3 0.0000024 2.65 0.001 DMI, %BW -2290 5.69 <0.0001 (DMI, %BW)2 296 6.25 <0.0001 (DMI, %BW)3 4.95 1.63 <0.01 FCMY/DMI 1440 6.24 <0.0001 FCMY/DMIZ 74.9 4.10 <0.0001 FCMY/DMI3 -19.0 2.99 <0.001 DMI 510. 6.41 <0.0001 DMl2 5.93 2.73 0.001 DMl3 -0.0287 2.58 <0.01 DIM 0.0180 2.61 <0.001 DlM2 0.000500 3.74 <0.001 DIM2 -0000001 3.36 <0.001 Diet % NDF 0.704 5.10 <0.0001 Diet % NDF2 0.0943 4.75 <0.0001 BCS change 2.62 0.18 0.38 BCS change2 52.3 4.17 <0.0001 BCS change3 -261 4.37 <0.0001 Diet % forNDF“ 0.163 0.63 <0.10 Diet % forNDFZ -0.256 6.06 <0.0001 Milk fat % -12.0 1.57 0.01 Milk fat %2 -13.0 5.23 <0.0001 BW change 0.663 2.13 <0.01 BW change2 -O.286 2.41 <0.01 BW change3 -0.101 1.33 0.02 ForNDF Digestibility“ 0.115 2.05 <0.01 ForNDF Digestibilityz 0.0114 1.12 0.03 ForNDF Digestibility3 -0000755 1.69 <0.01 Diet % starch 00550 0.32 0.24 MY -1.66 1.11 0.03 MY2 -0.248 2.06 <0.01 M)!3 -000129 1.92 <0.01 ForNDF Digestibility * Diet % starch -0.105 9.47 <0.0001 154 Table 6a, continued Parameter Estimate Sum of Squares P DMI, %BW * FCMY -13.1 8.43 <0.0001 DMI * FCMY 2.94 8.21 <0.0001 Diet % forNDF * FCMY/DMI 13.5 7.46 <0.0001 Diet % forNDF * FCMY -0.538 7.34 <0.0001 DMI, %BW * FCMY/DMI 383 7.10 <0.0001 Diet % forNDF * DMI, %BW 4.96 7.05 <0.0001 BW * FCMY/DMI 2.51 7.02 <0.0001 BW * Diet % forNDF 0.0381 6.87 <0.0001 FCMY/DMI * BCS change 35.7 5.76 <0.0001 BW * F CMY -0.0652 5.43 <0.0001 FCMY * BW change -1.70 5.43 <0.0001 MY * BW change 0.824 5.11 <0.0001 Diet % starch * FCMY -0.388 5.07 <0.0001 Diet % NDF * FCMY -0.528 4.70 <0.0001 DMI * DMI, %BW -86.6 4.54 <0.0001 BW * Milk fat % -0.194 4.27 <0.0001 Diet % NDF * MY 0.486 4.24 <0.0001 Diet % starch * MY 0.278 4.23 <0.0001 Diet % forNDF * Diet % starch -0.125 4.19 <0.0001 BW * BW change 0.0670 4.17 <0.0001 BW * DMI -0.594 4.14 <0.0001 MY * Milk fat % -4.02 3.94 0.0001 Milk fat % * BW change 4.48 3.86 0.0001 FCMY * Milk fat % 4.66 3.73 <0.001 Diet % forNDF * BCS change 3.79 3.66 <0.001 Diet % NDF * DMI, %BW 0.889 3.61 <0.001 DMI, %BW * BW change 8.80 3.58 <0.001 Diet % NDF * BW change -0.352 3.52 <0.001 DIM * Diet % NDF 0.0129 3.23 <0.001 BW * MY -0.0442 3.23 <0.001 Diet % NDF * Milk fat % 2.211 3.13 <0.001 Diet % forNDF * ForNDF Digestibility 0.0594 3.10 <0.001 MY * FCMY 0.721 2.98 <0.001 FCMY/DMI * BW change 17.5 2.92 <0.001 DIM * Diet % starch 0.00363 2.90 <0.001 DIM * ForNDF Digestibility 0.00161 2.70 0.001 BW * ForNDF Digestibility -0.00104 2.69 0.001 Milk fat % * BCS change -8.51 2.66 <0.01 BW * BCS change 0.181 2.52 <0.01 Diet % NDF * BCS change -4.25 2.40 <0.01 DIM * BW change -0.00893 2.35 <0.01 DIM * BCS change -0.299 2.26 <0.01 Diet % starch * DMI, %BW 1.16 2.26 <0.01 155 Table 6a, continued Parameter Estimate Sum of Squares P Diet % starch * Milk fat % 0.916 2.22 <0.01 DIM * BW -0.00156 1.76 <0.01 BW * Diet % starch 0.00817 1.71 <0.01 Diet % starch * FCMY/DMI 2.56 1.68 <0.01 Diet % forNDF * BW change 0.247 1.66 <0.01 ForNDF Digestibility * BCS change 1.36 1.63 <0.01 ForNDF Digestibility * FCMY/DMI 0.138 1.58 0.01 DMI, %BW * MY -3.39 1.56 0.01 DIM * Diet % forNDF -0.00512 1.41 0.02 Diet % NDF * forNDF Digestibility -0.0430 1.17 0.03 DIM * DMI 0.0303 1.08 0.03 DMI, %BW * Milk fat % -7.29 1.01 0.04 DIM * DMI, %BW -0.161 0.90 0.05 Diet % starch * BW change 0.107 0.85 0.06 DMI * MY ‘ 0.247 0.83 0.06 ForNDF Digestibility * Milk fat % 0.0305 0.66 0.09 ‘1 3.5% fat-corrected milk yield. 1’ Rumen-empty body weight. ° Forage NDF. d 30-h in vitro NDF digestibility. 156 Table 6b. Descriptive statistics of prediction Model 2 for iNDF kp, created using backward stepwise regression, ranked by contribution to the prediction (Sum of Squares) of each parameter (sums of linear, quadratic and cubic terms). The rgqression utilized 148 animal-period observations. Parameter Estimate Sum of Squares P Intercept -0.285 -- 0.88 Diet % starch 0.0985 1.73 0.06 Diet % starch2 0.000463 0.0103 0.66 Diet % starch3 -000415 13.4 < 0.0001 Diet % forNDF” -0.000397 0.000021 0.99 Diet % forNDFz -00407 7.47 0.0001 Diet % forNDF3 0.00506 3.39 < 0.01 DMI 0.0642 2.50 0.02 DMI2 0.0156 4.92 < 0.01 ForNDF Digestibility"2 0.0140 0.626 0.19 ForNDF Digestibility2 -000235 4.56 < 0.01 FCMY° -0.0240 2.38 0.03 ForNDF Digestibility * Diet % starch -0.0296 14.6 < 0.0001 DMI * FCMY -0.00703 3.90 < 0.01 Diet % starch * DMI 0.0161 3.52 < 0.01 Diet % starch* FCMY -0.00383 1.45 0.06 ‘1 Forage NDF. 1’ 30- h in vitro NDF digestibility. ° 3. 5% fat-corrected milk yield. Model 2: iNDF kp = -0. 285— 0. 3000397 x %forNDF— 0. 0407 x (%forNDF- 20. 8)2 + 0. 00506 x (%foprNDF- 20. 8)3 + 0.0140 x NDFD (%)- 0. 00125 x (NDFD(%) — 43. 7)2 + 0. 09835 x % starch + 0. 000483 x (% starch- 30. 222) — 0.00415 x (% starch— 30 2)3 + 0. 0642 x DMI + 0.0158 x (DMI- - 0.0296 x [(NDFD(%)- -43. 7) x (% starch— 30.2)] + 0. 0161 x [(% starch- 30. 2) x (DMI [(DMI — 22.7) x (FCMY — 37.2)]. 157 -22. 7)2 — 0.0140 x 3. 5% FCMY — 22.7)1— 0.00363 x [(% starch — 30.2) x (FCMY — 37.2)] - 0.00703 x Table 6c. Descriptive statistics of prediction Model 3 for iNDF kp, created using backward stepwise regression, ranked by contribution to the prediction (Sum of Squares) of each parameter (sum of linear, quadratic, and cubic terms). The regression utilized 167 animal-period observations. Parameter Estimate Sum of Squares P Intercept 18.2 --— <0.0001 Rumen DM pool -0.354 35.5 <0.0001 Rumen DM poolz 0.0333 4.22 <0.0001 Diet % forNDFa -0.153 1.48 <0.01 Diet % forNDF2 -0.0135 0.651 0.02 Diet % forNDF3 0.00626 3.31 <0.0001 ForNDF Digestibilityb -0.0192 0.547 0.05 ForNDF Digestibility2 -0.00323 1.50 <0.01 ForNDF Digestibility3 0.000173 1.74 <0.001 NDF intake 1.45 2.80 <0.0001 NDF intake2 -0.0585 0.720 0.03 OM digested in whole tract -0.153 2.85 <0.0001 Diet % starch -0.0657 0.788 0.02 Diet % starch2 -000321 0.411 I III I Starch kp Residual, h'1 0 I L I i O O 5 1015 20 25 30 35 Predicted starch kn, h'1 182 30 20 10 " Residual, h‘1 0 10" a. 69 lb 9 o I J Starch k -40‘ '50 I I I I I I j I -30 -20 -10 0 10 20 30 40 50 Centered Predicted Starch k,,, h" A Starch kp = -0.83 (1 0.82) - 0.90 (1 0.07) (X - 15.5) R2 = 0.67, s =7.12, P < 0.0001 Bias at min predicted (-14) = 25 Bias at max predicted (63) = -43 183 Figure 2. (A) Plot of residual (observed minus predicted) starch kp vs. predicted starch kp resulting from Model 2 in BUILD data set. (B) Plot of residual starch kp vs. predicted starch kp resulting from Model 2 applied to VALIDATE data set. Predicted starch kp was centered around the mean predicted value. Mean bias was not significant (P = 0.59). Although linear bias was significant (P = 0.01), the absolute value of the maximum bias (5.72) was lower than standard error of residuals (6.59). Different symbols represent data from individual studies. A 15- f. 10- g . .12 5- 8 u o: 0- °- I x .I: '5- 2 . III a 310-. '15 fl IrTfil—fjjl'l ' 0 5101520253035 Predicted starch k,, h'1 184 Starch kp Residual, h‘1 N O O O L. . 0 0| O OI I I I I d or -15 -10 -5 0 Centered Predicted Starch kp, h'1 Starch kp = -0.41 (1 0.76) - 0.40 (1 0.16) (x - 15.2) R2 = 0.08, s = 6.59, P = 0.01 Bias at min predicted (1.85) = 4.91 Bias at max predicted (28.5) = -5.72 185 Figure 3. (A) Plot of residual (observed minus predicted) starch kp vs. predicted starch kp resulting from Model 3 in BUILD data set. (B) Plot of residual starch kp vs. predicted starch kp resulting from Model 3 applied to VALIDATE data set. Predicted starch kp was centered around the mean predicted value. Mean bias was not significant (P = 0.33). Although linear bias tended to be significant (P = 0.06), the absolute value of the maximum bias (5.31) was lower than standard error of residuals (6.96). Different symbols represent data from individual studies. A 15" 10" Starch kp Residual, h'1 0 I .5 -I -10 j d '15 ' V I I v v v I v u - 0 10 20 30 Predicted starch k,, h'1 186 Starch kp Residual, h'1 , . -10 -5 0 5 10 Centered Predicted Starch kp, h'1 Starch kp = -0.84 (1 0.65) - 0.39 (1 0.20) (x - 16.0) R2 = 0.05, s = 6.96, P = 0.06 Bias at min predicted (5.80) = 3.15 Bias at max predicted (27.5) =-5.31 187 CHAPTER 7 Summary and Implications Passage Rate and Forage Family As hypothesized, with greater pVDMl, DMI on AL was increasingly greater than DMI on OG. This occurred because NDF turnover time in the rumen decreased more for AL than for 0G as pVDMI increased. The faster disappearance of NDF on AL, caused primarily by a greater increase in passage rate of iNDF on AL with increasing pVDMI, reduced the physical filling effects for AL more than was possible for NDF from diet OG. These results corroborate previous research suggesting that intake is more limited by physical fill effects with increasing nutrient demand and on grass forages compared to legume forages. Through its effect on passage rate responses, pVDMI also altered the extent to which diet affected the production of milk and its components; milk fat concentration, F CMY and BCS responses to AL over 0G were related to pVDMI values. Cows with the greatest drive to eat, as estimated by pVDMI, responded the most positively in DMI and milk production responses to alfalfa versus orchardgrass as the primary dietary fiber source. As expected, disappearance of N from ruminal digesta was greater when the dietary forage was alfalfa than when it was orchardgrass. The extent to which forage type affected N intake, digestion and utilization depended on the pVDMl of individual animals. Site of digestion and efficiency of utilization of dietary N for microbial protein and for milk true protein depended not only on 188 intake of N but also on responses of ruminal passage rate and ruminal starch digestion. The reduction of passage rate by OG, particularly among cows with high pVDMl, reduced the total amount of N consumed and utilized for microbial protein and milk true protein production. However, despite the expected dilution of maintenance N with increasingly greater MY on AL, a decreasing proportion of the additional N consumption that was allowed by the increased DMI on AL among cows with greater pVDMI was digested and used for increased milk production or body condition gain. Increasing passage rate and DMI by feeding a perennial legume forage instead of a perennial grass forage can increase yields of milk and milk protein among cows with greater nutrient demand. However, increasing N intake at the same rate as DMI is increased likely will lead to less efficient utilization of dietary N for production of microbial protein, body tissue, or milk protein. When feeding less-filling diets, such as those containing large proportions of legume forage, to high-producing cows, reducing dietary N concentration could increase the efficiency of N utilization and reduce the extent to which greater DMI leads to greater N excretion. Furthermore, the effects of pVDMl on N digestion and utilization reinforce the need to group and feed animals according to some index of nutrient demand. Reducing the variation in energy and protein demand within the group for which a diet is formulated would allow diets to be formulated to more accurately meet each individual animal’s demands and thus lead to more efficient utilization of N among all groups of animals on the farm. Increased N digestibility and utilization and more accurate diet formulation will reduce the 189 proportion and amount of N excreted in feces and urine. It should also reduce the likelihood of overfeeding N. Thus, a better understanding of the different effects of perennial grass and legume forages on N utilization, as they are influence by nutrient demand, will aid in field management decisions to minimize the turnover and loss of N on the whole farm. Passage Rate and Dietary Forage NDF Concentration When cows of varying pVDMl were fed a low-forage diet and a high- forage diet, DMI and passage rate responses differed from previously-observed responses (Voelker et al., 2002). Although DMI and passage rate responses previously were dependent on pVDMl, they were not in the present experiment. This apparently occurred because a longer NDF turnover time on LF with increasing pVDMI led to responses of DMI and milk production to HF and LF that were independent of pVDMl. A greater reduction of ruminal pH caused by greater starch intake on LF might have mediated this response. The results of this experiment suggest that models that predict intake need to account for not only the effects of nutrient demand, but also the effects of the interactions of feed fractions (such as starch and NDF) on the intake responses of individual cows to high- and low-forage diets. Predicting Passage Rates of NDF Fractions and Starch Prediction of passage rate of digesta is dependent on diet characteristics and nutrient demand of the individual animal. A model including diet % starch, DMI, forNDFD, diet % forNDF, and 3.5% FCMY accounted for 67% of variation in iNDF kp. A model including diet % starch, MY, diet % forNDF, diet % NDF, DMI, 190 BCS change, and forNDFD accounted for 53% of variation in pdNDF kp. A model including diet concentrations of NDF and starch, forNDF D, intake of DM and of starch, MY, and BCS change accounted for 42% of variation in starch k,,. The ability to account for more variation in iNDF kp than in pdNDF k,,, and to account for more variation in pdNDF kp than in starch kp, suggests that iNDF passage kinetics are more homogeneous than kinetics of pdNDF or of starch. This is likely because of greater variation in the physical and chemical characteristics that affect kp in pdNDF than in iNDF, and in starch than in pdNDF or in iNDF. Accurate prediction of the passage rates of pdNDF and starch from the rumen is needed to accurately predict ruminal NDF, starch, and total OM digestibility. Ruminal digestibility of pdNDF and of starch have important implications for the regulation of feed intake and for the extent and efficiency of microbial protein production. Passage rate of digesta fractions is seldom measured directly, but strategic collection of data sets containing more easily measured parameters along with passage rates can increase the accuracy of the prediction of passage rates in models intended for use on commercial dairy farms. Improving predictions of passage rates will increase the ability to optimize DMI and ruminal diet digestibility, and maximize milk yield. It also will permit more efficient utilization of N and other nutrients, and will reduce the proportion of nutrients that are excreted as waste. Many models of feed intake, digestion, and metabolism in dairy cows may be improved by incorporating the quantified effects of nutrient demand and feed 191 sources on feed intake and passage rate. Data describing these effects can be provided by the experiments reported here and by future experiments testing other important dietary characteristics. Forage-related treatments for which the dependence of response on pVDMl still need to be quantified include: forage particle Size for corn silage, legume forage, and perennial grass forages; forage maturity for legume and perennial grass forages; and forage Iignification (e.g., brown midrib strains). Predictions of passage rate also would be improved by experiments testing the effects on passage rate of treatments such as grain type, conservation method, and physical form, non-forage fiber sources, and supplementary fat amount and composition. Finally, an immediate implication of this research is its clear demonstration of the need to provide separate diets for cows with higher and lower nutrient demand, in order to maximize the efficiency of nutrient utilization within the entire herd. 192 REFERENCES Ahvenjarvi, S., A. Vanhatalo, P. Huhtanen, T. Varvikko. 2000. Determination of reticulo-rumen and whole-stomach digestion in lactating cows by omasal canal or duodenal sampling. Br. J. Nutr. 83:67-77. Ahvenjarvi, 3., B. Skiba, and P. Huhtanen. 2001. Effect of heterogeneous digesta chemical composition on the accuracy of measurements of fiber flow in dairy cows. J. Anim. Sci. 79:1611-1620. Ahvenjarvi, S., A. Vanhatalo, and P. Huhtanen. 2002. Supplementing barley or rapeseed meal to dairy cows fed grass-red clover Silage: l. Rumen degradability and microbial flow. J. Anim. Sci. 80:2176—2187. Ahvenjarvi, S., A. Vanhatalo, K.J. Shingfield, and P. Huhtanen. 2003. Determination of digesta flow entering the omasal canal of dairy cows using different marker systems. Br. J. Nutr. 90:41-52. Ahvenjarvi, S., E. Joki-Tokola, A. Vanhatalo, S. Jaakkola, and P. Huhtanen. 2006. Effects of replacing grass silage with barley silage in dairy cow diets. J. Dairy Sci. 1678-1687. Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic Control 19(6): 716—723. Akin, DE. 1989. Histological and physical factors affecting digestibility of forages. Agron. J. 81:17-25. Alhadhrami, G., and J.T. Huber. 1992. Effects of alfalfa hay of varying fiber fed at 35 or 50% of diet on lactation and nutrient utilization by dairy cows. J. Dairy Sci. 75:3091-3099. Allen, M. S. 1982. Investigation into the use of rare earth elements as gastrointestinal markers. MS. Thesis. Department of Animal Science, Cornell University. Allen, M. S. 1996. Physical constraints on voluntary intake of forages by ruminants. J. Anim. Sci. 74:3063—3075. Allen, M. S. 2000. Effects of diet on short-term regulation of feed intake by lactating dairy cattle. J. Dairy Sci. 83:1598—1624. Allen, MS, and DR. Mertens. 1988. Evaluating constraints on fiber digestion by rumen microbes. J. Nutr. 118:261-270. 193 Allen, M. S., J. G. Coors, and G. W. Roth. 2003. Corn silage. Pp. 547-608 In: Silage Science and Technology, D. Buxton, R. Muck, and J. Harrison, eds. American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America, Madison, WI. Allen, M. S., J. A. Voelker, and M. Oba. 2006. Physically effective fiber and regulation of ruminal pH: it’s more than just chewing. Pp. 270-278 In: Production Diseases in Farm Animals, Wageningen Academic Publishers, Wageningen, Netherlands. Al-Mabruk, R.M., N.F.G. Beck, and R.J. Dewhurst. 2004. Effects of silage species and supplemental Vitamin E on the oxidative stability of milk. J. Dairy Sci. 87:406-412. Anil, M.H., J.N. Mbanya, H.W. Symonds, and J.M. Forbes. 1993. Responses in the voluntary intake of hay or Silage by lactating cows to intraruminal infusions of sodium acetate or sodium propionate, the tonicity of rumen fluid or rumen distension. Br. J. Nutr. 69:699-712. Balch, CC, and A. Kelly. 1951. Factors affecting the utilization of food by dairy cows. 3. Specific gravity of digesta from the reticulo-rumen of cows. Br. J. Nutr. 42395-398. Baldwin, R.L., J.H.M. Thornley, D.E. Beever. 1987. Metabolism of the lactating cow. ll. Digestive elements of a mechanistic model. J. Dairy Research 54: 107-131. Bauman, DE, and J.M. Griinari. 2003. Nutritional regulation of milk fat synthesis. Annu. Rev. Nutr. 23: 203-227. Beauchemin, K.A., and AD. lwaasa. 1993. Eating and ruminating activities of cattle fed alfalfa or orchardgrass harvested at two stages of maturity. Can. J. Anim. Sci. 73:79-88. Bradford, B. J. and M. S. Allen. 2004. Milk fat responses to a change in diet fermentability vary by production level in dairy cattle. J. Dairy Sci. 2004 87: 3800-3807. Broderick, G. A., and J. H. Kang. 1980. Automated simultaneous determination of ammonia and total amino acids in rumen fluid and in vitro media. J. Dairy Sci. 63:64-75. Broderick, G.A., R.G. Koegel, R.P. Walgenbach, and T.J. Kraus. 2002. Ryegrass or alfalfa silage as the dietary forage for lactating dairy cows. J. Dairy Sci. 85:1894—1 901. 194 Burato, G. M, J. A. Voelker, and M. S. Allen, 2001. Effects of pretrial milk yield on feed intake, production, and feeding behavior responses to forage particle size by lactating cows. J. Dairy Sci. 84 (Suppl. 1):199. Campling, R.C., M. Freer, and CC. Balch. 1961. Factors affecting the voluntary intake of food by cows. 2. The relationship of digesta in the reticulo-rumen and the rate of disappearance of digesta from the alimentary tract. Br. J. Nutr. 15:531-540. Chai, K., P.M. Kennedy, and LP. Milligan. 1984. Reduction in particle size during rumination in cattle. Can. J. Anim. Sci. 64 (Suppl): 339-340. Choi, C.W., S. Ahvenjarvi, A. Vanhatalo, V. Toivonen, and P. Huhtanen. 2002. Quantitation of the flow of soluble non-ammonoia nitrogen entering the omasal canal of dairy cows fed grass silage based diets. Anim. Feed Sci. Technol. 96:203-220. Cochran, R. C., D. C. Adams, J. D. Wallace, and M. L. Galyean. 1986. Predicting digestibility of different diets with internal markers: evaluation of four potential markers. J. Anim. Sci. 63:1476—1483. Combs, D.K., R.D. Shaver, and L.D. Satter. 1992. Retention of rare earths by hay particles following incubation in fresh or autoclaved rumen fluid. J. Dairy Sci. 75:132-139. Dado, R. G., and M. S. Allen. 1993. Continuous computer acquisition of feed and water intake, chewing reticular motility, and ruminal pH of cattle. J. Dairy Sci. 76:1589-1600. Dado, R.G., and MS. Allen. 1995. Intake limitations, feeding behavior, and rumen function of cows challenged with rumen fill from dietary fiber or inert bulk. J. Dairy Sci. 78:118-133. Dewhurst, R.J., W.J. Fisher, J.K.S. Tweed, and R.J. Wilkins. 2003a. Comparison of grass and legume silages for milk production. 1. Production responses with different levels of concentrate. J. Dairy Sci. 86:2598-2611. Dewhurst, R.J., R.T. Evans, N.D. Scollan, J.M. Moorby, R.J. Merry, and R. J. Wilkins. 2003b. Comparison of grass and legume silages for milk production. 2. In vivo and in sacco evaluations of rumen function. J. Dairy Sci. 86:2612- 2621. Dixon, RM, and LP. Milligan. 1985. Removal of digesta components from the rumen of streers determined by sieving techniques and fluid, particulate and microbial markers. Br. J. Nutr. 53:347-362. 195 Ehle, F.R., F. Bas, B. Barno, R. Martin, and F. Leone. 1984. Particulate rumen turnover rate as influenced by density of passage marker. J. Dairy Sci. 67:2910—2913. Erdman, RA, and L.W. Smith. 1985. Ytterbium binding among particle size fractions of forage cell walls. J. Dairy Sci. 68:3071-3075. Firkins, J. L., M. S. Allen, B. S. Oldick, N. RHSt-Pierre 1998. Modeling ruminal digestibiiity of carbohydrates and microbial protein flow to the duodenum. J. Dairy Sci. 81: 3350- 3369. Forbes, J. M. 1995. Voluntary Food Intake and Diet Selection in Farm Animals. CAB lnternations, Oxon, UK. Fuentes-Pila, J., M. lbarlez, J.M. DeMiguel, and UK. Beede. 2003. Predicting average feed intake of lactating Holstein cows fed totally mixed rations. J. Dairy Sci. 86:309-323. Goering, H. H., and P. J. Van Soest. 1970. Forage Fiber Analysis (Apparatus, Reagents, Procedures, and Some Applications). Agric. Handbook No. 379. ARS-USDA, Washington, DC. Grant, R.J., and DR. Mertens. 1992. Influence of buffer pH and raw corn starch addition on in vitro fiber digestion kinetics. J. Dairy Sci. 75:2762-2768. Grant, R. J. 1997. Interactions among forages and nonforage fiber sources. J. Dairy Sci. 80:1438—1446. Grovum, W. L. 1986. The control and motility of the Ruminoreticulum. Pages 18- 40 in Control of digestion and metabolism in Ruminants: Proc. Sixth Int. Symp. Rumin. Physiol. L. P. Miligan, W.L. Grovum, and A. Dobson, eds. Prentice-Hall, Englewood Cliffs, NJ. Hach, C. C., B. K. Bowden, A. B. Lopelove, and S. V. Brayton. 1987. More powerful peroxide Kjeldahl digestion method. J. AOAC 70:783-787. Harvatine, K. J., and M. S. Allen. 2002. Saturation effects of rumen-inert fat sources on feed intake, milk production, and feeding behavior in lactating cows varying in milk yield. J. Dairy Sci. 85(Suppl. 1):141-142. Harvatine, K. J., and M. S. Allen. 2005. The effect of production level on feed intake, milk yield, and endocrine responses to two fatty acid supplements in lactating cows. J. Dairy Sci. 88:4018-4027. 196 Harvatine, K.H., and MS. Allen. 2006. Fat supplements affect fractional rates of ruminal fatty acid biohydrogenation and passage in dairy cows. J. Nutr. 136:677-685. Hoffman, P.C., D.K. Combs, and MD. Casler. 1998. Performance of lactating dairy cows fed alfalfa silage or perennial ryegrass silage. J. Dairy Sci. 81: 162-168. Hoffman, P.C., S.J. Sievert, R.D. Shaver, D.A. Welch, and OK. Combs. 1993. In situ dry matter, protein, and fiber digestion of perennial forages. J. Dairy Sci. 76:2632-2643. Holden, L.A., B.P. Glenn, R.Al. Erdman, and WE. Potts. 1994. Effects of alfalfa and orchardgrass on digestion by dairy cows. J. Dairy Sci. 77:2580-2594. Hoover, W. H. 1986. Chemical factors involved in ruminal fiber digestion. J. Dairy Sci. 69:2755—2766. Huhtanen, P., P. G. Brotz, and L.D. Satter. 1997. Omasal sampling technique for assessing ferrnentative digestion in the forestomach of dairy cows. J. Anim. Sci. 75:1380—1392. lllius, AW, and MS. Allen. 1994. Assessing forage quality using integrated models of intake and digestion by ruminants. Pp. 869-890 in Forage Quality, Evaluation, and Utilization, G.C. Fahey, Jr., M. Collins, D.R. Mertens, and LE. Moser, eds. American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America, Madison, WI. Isaacson, H. R., F. C. Hinds, M. P. Bryant, and F. N. Owens. 1975. Efficiency of energy utilization by mixed rumen bacteria in continuous culture. J. Dairy Sci. 58:1645-1659. Jenness, R. 1985. Biochemical and Nutritional Aspects of Milk and Colostrum. Pp. 164-197 in: Lactation. B. L. Larson, ed. Iowa State University Press, Ames. Jung, H. G., and M. S. Allen. 1995. Characteristics of plant cell walls affecting intake and digestibility of forages by ruminants. J. Anim. Sci. 73:2774-2790. Karkalas, J. 1985. An improved enzymatic method for the determination of native and modified starch. J. Sci. Food Agric. 36:1019-1027. Kaske, M., and A. Midasch. 1994. Effects of experimentally inhibited reticular contraction on digesta passage in sheep. In: D. Giesecke (ed.) Proc. Of the Society of Nutritional Physiology, DLG-Verlag, Frankfurt, Germany. 197 Kennedy, P. M., and L. P. Milligan. 1978. Effects of cold exposure on digestion, microbial synthesis and nitrogen transformations in sheep. Br. J. Nutr. 39:105-117. Kohn, RA, and MS. Allen. 1995. Effect of plant maturity and preservation method on in vitro protein degradation of forages. J. Dairy Sci. 78:1544-1551. Kwakkel, R.P., J. Van Bruchem, G. Hof, and H. Boer. 1986. The in sacco degradation of crude protein and cell wall constituents in grass, alfalfa and maize silages. Neth. J. Agric. Sci. 34:116. Lehman, F. 1941. Die Iehre vom ballast. Z. Tierphysiol. Tierernaehr. Futterrnittelkd. 52155-173. Lundy, F.P. III, E. Block, W.C. Bridges Jr., J.A. Bertrand, and TC. Jenkins. 2004. Ruminal biohydrogenation in Holstein cows fed soybean fatty acids as amides or calcium salts. J. Dairy Sci. 87:1038-1046. Martin, C., and B. Michalet-Doreau. 1996. Influence of barley and buffer supplements on quantitative aspects of ruminal fiber digestion of cows. Arch. Tierernahr. 49:203-21 1 . Mathison, G.W., E.K. Okine, A.S. Vaage, M. Kaske, and LP. Milligan. 1995. Current understanding of the contribution of the propulsive activities in the forestomach to the flow of digesta. P. 23 in: Ruminant Physiology: Digestion, metabolism, growth and reproduction, W.V. Engelhardt, S. Leonhard-Marek, G. Breves, and D. Giesecke, eds. Ferdinand Enke Verlag, Stuttgart, Germany. McLeod, M. N., and D. J. Minson. 1988. Large particle breakdown by cattle eating ryegrass and alfalfa. J. Anim. Sci. 66:992-999. McLeod, M.N., P.M. Kennedy, and DJ. Minson. 1990. Resistance of leaf and stem fractions of tropical forage to chewing and passage in cattle. Br. J. Nutr. 63: 1 05-1 19. Mertens, D. R. 1994. Regulation of forage intake. Pp. 450-493 in Forage Quality, Evaluation, and Utilization. G. C. Fahey, Jr., M. Colins, D. R. Mertens, and L. E. Moser, eds., American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America, Madison, WI. Mooney, CS, and MS. Allen. 2004. Effects of dietary sodium bicarbonate on ruminal and total tract digestibility of diet and diet components in dairy cows. J. Dairy Sci. 82 (Suppl. 1): 214. 198 National Research Council. 2001. Nutrient Requirements of Dairy Cattle, seventh revised edition. National Academy Press, Washington, DC. Neter, J., M.H. Kutner, C.J. Nachtsheim, and W. Wasserman. 1996. Applied Linear Statistical Models, Fourth Edition. WCB McGraw-Hill, Boston, Mass. Nicholson, T., and S. A. Omer. 1983. The inhibitory effect of intestinal infusions of unsaturated long-chain fatty acids on forestomach motility of sheep. Br. J. Nutr. 50:141-149. Noftsger, 8., NR. St-Pierre, and J.T. Sylvester. 2005. Determination of rumen degradability and ruminal effects of three sources of methionine in lactating cows. J. Dairy Sci. 88:223-237. O’Connor, J. D., C. J. Sniffen, D. G. Fox, and W. Chalupa. 1993. A net carbohydrate and protein system for evaluating cattle diets: IV. Predicting amino acid adequacy. J. Anim. Sci. 71:1298-1311. Oba, M., and MS. Allen. 1999a. Effects of brown midrib 3 mutation in corn silage on dry matter intake and productivity of high yielding dairy cows. J. Dairy Sci. 82:135-142. Oba, M., and MS. Allen. 1999b. Evaluation of the importance of the digestibility of neutral detergent fiber from forage: effects on dry matter intake and milk yield of dairy cows. J. Dairy Sci. 82:589-596. Oba, M., and M. S. Allen. 2000. Effects of Brown Midrib 3 Mutation in Corn Silage on Productivity of Dairy Cows Fed Two Concentrations of Dietary Neutral Detergent Fiber: 3. Digestibility and Microbial Efficiency. J. Dairy Sci. 83:1350—1358. Oba, M. and M. S. Allen. 2003a. Effects of corn grain conservation method on feeding behavior and productivity of lactating dairy cows at two dietary starch concentrations. J. Dairy Sci. 86:174-183. Oba, M., and M. S. Allen. 2003b. Effects of Com Grain Conservation Method on Ruminal Digestion Kinetics for Lactating Dairy Cows at Two Dietary Starch Concentrations. J. Dairy Sci. 86: 184-194. Oba, M., and MS. Allen. 2003c. Effects of diet fermentability on efficiency of microbial nitrogen production in lactating dairy cows. J. Dairy Sci. 86:195-207. Okeke, G.C., J.G. Buchanan-Smith, D.G. Grieve. 1983. Effect of sodium bicarbonate on rate of passage and degradation of soybean meal in postpartum dairy cows. J. Dairy Sci. 66:1023-1031. 199 Okine, E. K., and G. W. Mathison. 1991. Reticular contraction attributes and passage of digesta from the ruminoreticulum in cattle fed roughage diets. J. Anim. Sci. 69:2177—2186. Onetti, S.G., S.M. Reynal, and RR. Grummer. 2004. Effect of alfalfa forage preservation method and particle length on performance of dairy cows fed corn silage-based diets and tallow. Owens, F.N., and R. Zinn. 1988. Protein metabolism of ruminant animals. Pp. 227-249 In The Ruminant Animal: Digestive Physiology and Nutrition. D.C. Church, ed. Waveland Press, Prospect Heights, Illinois. Poppi, D. P., B. W. Norton, D. J. Minson, and R. E. Hendricksen. 1980. The validity of the critical size theory for particles leaving the rumen. J. Agric. Sci. 94:275-280. Poppi, D. P., D. J. Minson, and J. H. Ternouth. 1981. Studies of cattle and sheep eating leaf and stem fractions of grasses. Ill. The retention time in the rumen of large feed particles. Aust. J. Agric. Res. 32:123. Reynal, SM, and GA. Broderick. 2003. Effects of feeding dairy cows protein supplements of varying ruminal degradability. J. Dairy Sci. 86:835-843. Reynal, S.M., G.A. Broderick, S. Ahvenjarvi, P. Huhtanen. 2003. Effect of feeding protein supplements of differing degradability on omasal flow of microbial and undegraded protein. J. Dairy Sci. 86:1292-1305. Riewe, M. E., and H. Lippke. 1970. Considerations in determining the digestibility of harvested forages. In: Proc. Nat. Conf. on Forage Quality Evaluation and Utilization, p. F1. Univ. Nebraska, Lincoln. Rodrigue, CB, and NM Allen. 1960. The effect of fine grinding of hay on ration digestibility, rate of passage, and fat content of milk. Can. J. Anim. Sci. 40:23- 29. Russell, J.B., W.G. Bottje, and MA. Cotta. 1981. Degradation of protein by mixed cultures of rumen bacteria: Identification of Streptococcus bovis as an actively proteolytic rumen bacterium. J. Anim. Sci. 53:242-252. Russell, J.B., J.D. O’Connor, D.G. Fox, P.J. Van Soest, and C.J. Sniffen. 1992. A net carbohydrate and protein system for evaluating cattle diets: I. Ruminal fermentation. J. Anim. Sci. 70: 3551-3561. St-Pierre, NR. 2001. Integrating Quantitative Findings from Multiple Studies Using Mixed Model Methodology. J. Dairy Sci. 84: 741-755. 200 St-Pierre, NR. 2003. Reassessment of biases in predicted nitrogen flows to the duodenum by NRC 2001. J. Dairy Sci. 86: 344-350. Sec, 8., Lo. Tedeschi, C. G. Schwab, B.D. Garthwaite, and D.G. Fox. 2006. Evaluation of the passage rate equations in the 2001 Dairy NRC model. J. Dairy Sci. 89:2327-2342. Sheperd, A. C., and D. K. Combs. 1998. Long-term effects of acetate and propionate on voluntary feed intake by midlactation cows. J. Dairy Sci. 81 :2240—2250. Smith, L. W., H. K. Goering, and C. H. Gordon. 1972. Relationships of forage compositions with rates of cell wall digestion and indigestibility of cell walls. J. Dairy Sci. 5521140. Sniffen, C.J., J.D. O'Connor, P.J.Van Soest, D.G. Fox, and J.B. Russell. 1992. A net carbohydrate and protein system for evaluating cattle diets. ll. Carbohydrate and protein availability. J. Animal Sci. 70:3562-3577. Spandl, E., and DB. Hesterman. 1997. Forage quality and alfalfa characteristics in binary mixtures of alfalfa and bromegrass or timothy. Crop Sci. 37: 1 581-1 585. Stouthamer, A. H., and C. Bettenhaussen. 1973. Utilization of energy for growth and maintenance in continuous and batch cultures of microorganisms. Biochim. Biophys. Acta, 301253-70. Strobel, H.J., and J.B. Russell. 1986. Effect of pH and energy spilling on bacterial protein synthesis by carbohydrate-limited cultures of mixed rumen bacteria. J. Dairy Sci. 69:2941-2947. Sutherland, TM. 1988. Particle separation in the forestomachs of sheep. In A. Dobson and M. J. Dobson, edS., Aspects of Digestive Physiology in Ruminants, Cornell Univ. Press, Ithaca, New York. Taylor, C. C. and M. S. Allen. 2005. Corn grain endosperm type and brown midrib 3 corn silage: site of digestion and ruminal digestion kinetics in lactating cows. J. Dairy Sci. 88:1413-1424. Teeter, R.G., F .N. Owens, and TL Mader. 1984. Ytterbium chloride as a marker for particulate matter in the rumen [Feeds, digestibility]. J. Animal Sci. 58:464-473. Ulyatt, M. J. 1983. Plant fibre and regulation of digestion in the ruminant. In: G. Wallace and L. Bell (ed) Fibre in Human and Animal Nutrition. p 103. The Royal Soc. of New Zealand, Wellington, N2. 201 Van Soest, P. J., C. J. Sniffen, and M. S. Allen. 1988. Rumen dynamics. In: A. Dobson and M. J. Dobson (ed.) Aspects of Digestive Physiology in Ruminants. Cornell Univ. Press, Ithaca, NY. Van Soest, P. J., J. B. Robertson, and B. A. Lewis. 1991. Methods for dietary fiber, neutral detergent fiber, and nonstarch polysaccharides in relation to animal nutrition. J. Dairy Sci. 74:3583—3597. Van Soest, P. J., M. B. Rymph, and D. G. Fox. 1992. Discounts for net energy and protein - fifth revision. Pp. 40—68 in Proc. Cornell Nutr. Conf., Ithaca, NY. Voelker, J. A., G. M. Burato, and M. S. Allen. 2002. Effects of pretrial milk yield on responses of feed intake, digestion, and production to dietary forage concentration. J. Dairy Sci. 85: 2650—2661. Voelker, J.A., and M. S. Allen. 2003a. Pelleted beet pulp substituted for high- moisture corn: effects on digestion and rumen digestion kinetics in lactating dairy cows. J. Dairy Sci. 86:3553-3561. Voelker, J. A., and M. S. Allen. 2003b. Pelleted beet pulp substituted for high- moisture corn: 3. Efffects on rumen fermentation, pH, and microbial protein efficiency in lactating dairy cows. J. Dairy Sci. 86:3562-3570. Voelker Linton, J. A., and MS. Allen. 2005. Pretrial intake affects relative intake, digestion, and production responses of lactating cows to alfalfa and grass silages. Journal of Dairy Science 88 (Suppl. 1): 252. Voelker Linton, J.A., and M. S. Allen. 2006. Voluntary feed intake affects response to dietary forage content. Journal of Dairy Science 89 (Suppl. 1): 263. Waghorn, G.C., l.D. Shelton, and V.J. Thomas. 1989. Particle breakdown and rumen digestion of fresh ryegrass (Lolium perenne L.) and Iucerne (Medicago sativa L.) fed to cows during a restricted feeding period. Br. J. Nutr. 61 :409- 423. Waldo, D. R., L. W. Smith, and E. L. Cox. 1972. Model of cellulose disappearance from the rumen. J. Dairy Sci. 55:125-129. Wallace, R. J., and C. A. McPherson. 1987. Factors affecting the rate of breakdown of bacterial protein in rumen fluid. Br. J. Nutr. 58:313-323. Weiss, WP, and W.L. Shockey. 1991. Value of orchardgrass and alfalfa silages fed with varying amounts of concentrates to dairy cows. J. Dairy Sci. 74:1933—1943. 202 Wells, J. E., and J. B. Russell. 1996. Why do so many ruminal bacteria die and lyse so quickly? J. Dairy Sci. 79:1487-1495. Wheeler, W.E., C.H. Noller, and CE. Coppock. 1975. Effect of forage-to- concentrate ratio in complete feeds and feed intake on digestion of starch by dairy cows. J. Dairy Sci. 58:1902-1906. V\fildman, E. E., G. M. Jones, P. E. Wagner, R. L. Boman, H. F. Troutt Jr., and T. N. Lesch. 1982. A dairy cow body condition scoring system and its relationship to selected production characteristics. J. Dairy Sci. 65:495-501. VIfllson, J.R., D.E. Akin, M.N. McLeod, and DJ. Minson. 1989. Particle Size reduction of the leaves of a tropical and a temperate grass by cattle. ll. Relation of anatomical structure to the process of leaf breakdown through chewing and digestion. Grass and Forage Sci. 44:65-75. Wilson, JR, and PM. Kennedy. 1996. Plant and animal constraints to voluntary feed intake associated with fibre characteristics and particle breakdown and passage in ruminants. Aust. J. Agric. Res. 47:199-225. Wilson, JR, and RD. Hatfield. 1997. Structural and chemical changes of cell wall types during stem development: consequences for fibre degradation by rumen microflora. Aust. J. Agric. Res. 48:165-180. Woodford, ST, and MR. Murphy. Dietary alteration of particle breakdown and passage from the rumen in lactating dairy cattle. J. Dairy Sci. 71 :687-696. Ying, Y., and MS. Allen. 1998. Effects of fineness of grinding and conservation method of corn grain on ruminal starch digestion kinetics in Holstein heifers before and after calving. J. Dairy Sci. 81 (Suppl. 1): 318. Ying, Y., M.S. Allen, M.J. VandeHaar, and N.K. Ames. 1998. Effects of fineness of grinding and conservation method of corn grain on ruminal and whole tract digestibilities and ruminal microbial production of Holstein cows in early lactation. J. Dairy Sci. 81(Suppl. 1): 339. Ying, Y., and MS. Allen. 2005. Effects of corn grain endosperm type and conservation method on site of digestion, ruminal digestion kinetics and microbial nitrogen production of lactating dairy cows. J. Dairy Sci. 83 (Suppl. 1):393. Zimmerman, C.A., A.H. Rakes, R.D. Jaquette, B.A. Hopkins, and W.J. Croom, Jr. 1991. Effects of protein level and forage source on milk production and composition in early lactation dairy cows. J. Dairy Sci. 74:980-990. 203 Zinn, R. A., and F. N. Owens. 1986. A rapid procedure for purine measurement and its use for estimating net ruminal protein synthesis. Can. J. Anim. Sci. 66:157-166. 204