In vitro unfermented fiber is a good predictor of the digestible and metabolizable energy content of corn distillers dried grains with solubles in growing pigs

Abstract

Characterizing fiber into fermentable and unfermentable fractions may enhance the accuracy of estimating DE and ME energy content in fiber-rich ingredients. Therefore, the objective of this study was to analyze the concentrations of NDF, representing both the fermentable (fNDFom) and unfermentable (uNDFom) portions among sources of corn distillers dried grains with solubles (DDGS), and determine their relative contributions to DE and ME content. The concentrations of DE and ME, as well as apparent total tract digestibility (ATTD) of GE, were measured in a previous experiment. Samples of DDGS (0.5 g) were mixed with fecal inoculum and incubated for 8, 12, and 72 h. The ash corrected NDF (NDFom) content of DDGS residues at each time point was determined. The fNDFom increased with fermentation time of 8 h (21.6%), 12 h (29.0%), and 72 h (68.6%). The ATTD of GE increased as the uNDFom decreased at 8 h (uNDFom8; R² = 0.83; P < 0.01) and 72 h (uNDFom72; R² = 0.83; P < 0.01). Likewise, ME content of DDGS increased as uNDFom72 decreased (R² = 0.59; P < 0.01). The best-fit DE equation was DE (kcal/kg DM) = 2,175 – 3.07 × uNDFom8 (g/kg, DM) – 1.50 × uNDFom72 (g/kg, DM) + 0.55 × GE (kcal/kg DM) (R² = 0.94, SE = 36.21). The best-fit ME equation was ME (kcal/kg DM) = 1,643 – 2.31 × uNDFom8 (g/kg, DM) – 2.54 × uNDFom72 (g/kg, DM) + 0.65 × GE (kcal/kg DM) – 1.42 × crude protein (g/kg DM) (R² = 0.94, SE = 39.21). These results indicate that in vitro unfermented fiber is negatively associated with GE and NDF digestibility, and therefore, is a good predictor of DE and ME content in corn-DDGS.

INTRODUCTION

Corn distillers dried grains with solubles (DDGS) has been widely used in swine diets because it is a cost-competitive source of ME and digestible amino acids but the ME content in DDGS is quite variable (Stein and Shurson, 2009). Because energy is the most expensive nutritional component in animal feeds, DE and ME prediction equations have been developed using chemical composition data, to dynamically estimate the DE and ME content among various DDGS sources (Anderson et al., 2012; Kerr et al., 2013; Li et al., 2015). Dietary fiber from total dietary fiber (TDF) or NDF have been shown to be necessary components to predict DE and ME content among sources of DDGS (Kerr et al., 2013). However, these equations may not provide the most precise inputs in energy prediction equations because they depend on the measurement of TDF, NDF, or ADF as a homogenous portion of the diet.

In fact, dietary fiber in corn DDGS is composed of multiple carbohydrates with distinct properties (Jaworski et al. 2015) and rate of disappearance along the gastrointestinal tract (Urriola et al., 2010). Therefore, using more specific characteristics of fiber may improve the prediction of DE and ME content. For example, fiber fermentation produces short-chain fatty acids, which contribute to DE and ME, while unfermented fiber reduces apparent total tract digestibility of lipids (Kim et al., 2013) and CP (Urriola and Stein, 2010), which subsequently decrease digestibility of energy (Gutierrez et al., 2013; Iyayi and Adeola, 2015). Therefore, analyzing the unfermented fiber fraction of corn DDGS may enhance the accuracy and consistency of DE and ME estimates derived from energy prediction equations. We hypothesized that uNDF is a useful predictor of the DE, ME, and nutritive value of DDGS. The objective of this study was to determine the in vitro fNDF and uNDF content of various DDGS sources using fecal inoculum to determine their relative contributions for predicting DE and ME content of DDGS.

MATERIALS AND METHODS

Sample Collection

Fifteen sources of corn DDGS were obtained from different ethanol plants and diverse geographical locations of U.S. corn production, as well as different co-product processing technologies, to represent the variability in chemical composition among sources in the U.S. market. These 15 corn DDGS samples were used in a previous study (Kerr et al., 2013) to determine the nitrogen (CP), ether extract (EE), and dietary fiber (TDF, NDF, or ADF) content and were kept at −20 °C. We used nutrient composition from the previous manuscript, calculated other nutrients such as insoluble hemicellulose (NDF minus ADF), and used the published values for in vivo apparent total tract digestibility (ATTD) of nutrients, DE, and ME content for growing pigs. The digestible nutrient content of DDGS was calculated by multiplying total concentrations by the corresponding ATTD values, and the indigestible portion was subsequently calculated by the difference between total and digestible nutrient content.

In Vitro Fermentation

Feces were obtained from 3 finishing pigs (90 kg BW) from Cargill Animal Nutrition (Elk River, MN), which had been fed a corn-wheat-soybean meal diet with no antibiotics. Fecal samples were collected directly from the rectum, immediately placed in zipper plastic bags without air, and kept in a water bath at 39 °C until used as inoculum for incubation. The time from fecal collection until incubation was less than 1 h.

For the in vitro fermentation assay, 500 mg of DDGS samples were added to 125 mL serum bottles with rubber-stoppers containing 40 mL buffer solution containing macro- and micro-minerals (Menke and Steingass, 1988). The inoculum was prepared by diluting blended feces in an inoculation solution composed of distilled water (474 mL/L), trace mineral solution (0.12 mL/L containing CaCl₂ 132 g/L, MnCl₃·4H₂O 100 g/L, CoCl₂•6H₂O 10 g/L, and FeCl₃•6H₂O 80 g/L), in vitro buffer solution (237 mL/L containing NH₄HCO₃ 4.0 g/L and NaHCO₃ 35 g/L), macro-mineral solution (237 mL/L composed of Na₂HPO₄ 5.7 g/L, KH₂PO₄ 6.2 g/L, MgSO4·7H₂O 0.583 g/L, and NaCl 2.22 g/L), and resazurin (blue dye, 0.1% w/v solution; 1.22 mL/L), and was filtered through 4 layers of cheesecloth. The final inoculum concentration was adjusted to 0.094 g feces per mL of buffer, which represented the same feces to substrate ratio used in previous studies (Jha et al., 2011; Huang et al., 2017a, 2017b). Forty milliliters of inocula were transferred into bottles containing DDGS samples, and bottles were sealed with rubber stoppers before placing in a 39 °C water bath for incubation. Anaerobiosis was maintained in the inoculation solution by the addition of a reducing solution (distilled water 47.5 mL/L, 1 M NaOH 2 mL/L, Na₂S 335 mg/L) and CO₂. Bottles were sealed with a rubber stopper and placed in the water bath for incubation. The fermentation was terminated after 8, 12, or 72 h of incubation by placing bottles in ice, and gas production was recorded at 2, 5, 8 12, 16, 20, 24, 30, 36, 48, and 72 h.

After the termination of fermentation, the inoculum (40 mL) was directly mixed with NDF washing detergents (60 mL) and loaded on a reflux apparatus for NDF analyses as described by Mertens (2002). After reflux solubilization, residues were filtered with the use of a glass microfiber filter (934-AH by Whatman, Whatman Limited-GE Healthcare, Maidstone, UK) with a porosity of 1.5 μm in Pyrex Gooch crucibles (40 to 60 μm; Corning, Inc., Corning, NY). Subsequently, residues were ashed at 550 °C to determine ash-free NDF (NDFom). Blanks were created by inoculating the bottles with buffer, fecal inoculum, and were subjected to the same process as the test feed ingredients to adjust for any particles introduced into the in vitro fermentation system. The NDF content of DDGS samples was analyzed in duplicate, and all unfermented NDF residues after 8, 12, and 72 h fecal incubation (uNDFom8, uNDFom12, and uNDFom72, respectively) were analyzed in triplicate. The in vitro fermented NDFom (fNDFom8, fNDFom12, and fNDFom72) was calculated by subtracting total NDFom and uNDFom after 8, 12, and 72 h of incubation. The digestibility coefficients of NDFom (DigNDFom8, DigNDFom12, and DigNDFom72) were calculated by the ratio of digestible NDFom and total NDFom content.

Gas accumulation curves were recorded during the 72 h of fermentation and were modified according to France et al. (1993):

$$\begin{array}{l}\text{G}\left(\mathrm{mL}\text{g}/\text{DM}\right)=0,\text{\hspace{0.17em}if\hspace{0.17em}}0<\text{t}<\text{L},\\ \text{G}\left(\mathrm{mL}\text{g}/\text{DM}\right)={\text{G}}_{\mathrm{f}}(1-\exp (-[\text{b}\left(\text{t}-\text{L}\right)+\text{c}(\sqrt{\text{t}}-\sqrt{\text{L}})])),\text{if}\mathrm{t}\ge \text{L},\end{array}$$

where G denotes the gas accumulation at a specific time (t), G_f (ml g⁻¹ DM) was the maximum gas volume for t = ∞, and L (h) represents the lag time before the fermentation began. Gas accumulation rapidly reached one-fourth of the maximum accumulation in 2 h, and the parameter L was very close to 0, which resulted in the model failing to converge. Therefore, L(h) data were removed from the final model. The constants b (h⁻¹) and c (h^−1/2) were used to determine the fractional rate of degradation of the substrate µ (h−1), which is postulated to vary with time as follows:

$\mu =\mathrm{b}+\mathrm{c}/(2\sqrt{\mathrm{t}}),\mathrm{t}=\mathrm{T}/2$, representing the time to half-asymptote when G = G_f/2.

Statistical Analyses

The PROC CORR of SAS (Version 9.3; SAS Inst. Inc., Cary, NC) was used to determine if there was a correlation among DE, ME, fNDFom, and uNDFom fractions, and the chemical composition of corn DDGS samples. Correlations with a value of P < 0.05 were considered significant. The PROC REG STEPWISE of SAS was used to select input variables for the equations to predict DE and ME content from chemical composition, and in vitro fNDFom and uNDFom of the corn DDGS samples. Variance Inflation Factor (VIF) was used to determine multicollinearity, and variables with VIF > 10 were considered as multicollinear and were removed from the prediction equations. The P value, R², and the root of SEM were used as parameters to determine the accuracy of the prediction equations.

RESULTS AND DISCUSSION

Variability of Energy and Fiber-Related Composition

The sources of DDGS obtained in this study represent a wide range in GE, DE, and ME content, with differences between maximum and minimum values of were 387 kcal/kg for GE, 396 kcal/kg for DE, and 430 kcal/kg for ME (Table 1). The fiber-related composition was more variable (CV > 5.6) compared to ME and DE (CV < 4.1). The average NDFom value was slightly greater (388 g/kg NDF) than the average value reported by Kerr et al. (2013), with a difference of 122 g/kg NDF between the minimum and maximum values of the 15 DDGS sources. In our study, the NDFom analyses involved using glass microfiber filters with pore sizes that were in the submicron range, resulting in retention of very fine particles (Barbosa et al., 2015), which may partially explain the slightly greater average NDF values in our study compared with those reported by Kerr et al. (2013).

Table 1.

Fitted kinetics parameters of gas accumulation and fermented and unfermented NDF of corn distillers dried grains with solubles, DM basis¹

Item	Mean	Minimum	Maximum	Interval	SD	CV
Chemical composition (Kerr et al., 2013)
GE, kcal/kg	4,996	4,780	5,167	387	111	2.2
DE, kcal/kg	3,650	3,474	3,870	396	130	3.6
ME, kcal/kg	3,435	3,266	3,696	430	140	4.1
Crude protein, g/kg	305	277	329	52	14	4.5
Ether extract, g/kg	97	49	132	84	23	23.3
TDF, g/kg	342	308	378	69	19	5.6
NDF, g/kg	354	288	440	152	40	11.3
ADF, g/kg	117	90	140	50	18	15.3
Fiber fractions characterized by fecal incubation
NDFom, g/kg	388	335	457	122	28	7.2
uNDFom8, g/kg	304	247	350	103	30	9.8
uNDFom12, g/kg	276	234	329	95	28	10.3
uNDFom72, g/kg	123	84	165	81	28	22.6
fNDFom8, g/kg	84	43	121	78	19	23.2
fNDFom12, g/kg	112	73	146	73	18	16.4
fNDFom72, g/kg	265	232	297	65	21	7.7
DigNDFom8, %	21.6	11.0	30.4	19.4	4.8	22.4
DigNDFom12, %	29.0	18.7	36.6	17.9	4.6	15.8
DigNDFom72, %	68.6	59.1	75.9	16.8	5.7	8.4
Gas accumulation kinetics
Gas8, mL/g	97	81	117	36	8.5	8.8
Gas12, mL/g	120	105	139	33	9.2	7.7
Gas72, mL/g	199	174	214	39	13	6.5
Gf, mL/g	240	211	271	60	22.4	9.3
T/2, hour	17.2	10.9	27.1	16.2	4.8	27.6
μ T/2, h−1	0.036	0.024	0.048	0.024	0.007	18.8

¹SD = standard deviation; CV = coefficient of variation; GE = gross energy; DE = digestible energy; ME = metabolizable energy; TDF = total dietary fiber; ADF = acid detergent fiber; NDFom = ash-free NDF; DigNDFom = percentage of fermented NDFom after 8, 12, or 72 h of fecal incubation; fNDFom = fermented NDF after 8, 12, or 72 h of fecal incubation; uNDFom = unfermented NDF after 8, 12, or 72 h of fecal incubation; Gas = cumulative gas production after 8, 12, or 72 h of fecal incubation; T/2 = half-time to asymptote (hour); μ _T/2 = fractional rate of degradation (h⁻¹) at t = T/2; Gf = maximal gas production.

A 21.6% of the NDFom (average value of 15 sources DDGS) disappeared after 8 h of incubation (fNDFom8), which was similar to the average apparent ileal digestibility (AID, 21.5 %) value of TDF for 10 sources of corn DDGS reported by Urriola and Stein (2010). This portion of NDF had the greatest CV among the 3-time tested suggesting that it may provide great differentiation among sources of DDGS, the disappearance of NDF in this rapid portion (fNDFom8) was quite variable 43 to 121 g/kg. Furthermore, the percentage of NDFom disappearing after 8 h fecal incubation (DigNDFom8) ranged from 11.0 to 30.4% among DDGS sources. The fNDFom8 (assumed to be readily degradable fiber in the small intestine of pigs) of corn DDGS was highly variable with covariation coefficients greater than 20%. The use types and amounts of enzymes in the ethanol and co-product production process varies among ethanol plants and may have contributed to high variance of readily degradable fiber in the sources of corn DDGS evaluated in this study. Cellulolytic enzymes have been shown to improve ethanol yield and oil recovery during bioethanol production (Luangthongkam et al., 2015). Therefore, it is possible that greater amounts of readily degradable fiber in corn DDGS were removed in some sources due to the type of enzyme pretreatment during the ethanol production process (Pedersen et al., 2015).

A 68.6% of the NDFom disappeared after 72 h of fecal incubation (fNDFom72), which was greater than the ATTD of NDF (59.3%) in corn DDGS reported by Urriola et al. (2010). This difference may have been a result of using the 72-h in vitro incubation time compared with a typical 30 to 51 h in vivo digesta transit time (Latymer et al., 1990). In addition, nondietary materials with fiber-like properties in the gastrointestinal tract may have contributed to the measurement of TDF or NDF, which are likely to be detected as endogenous fiber loss and consequently result in a relatively low apparent fiber digestibility (Montoya et al., 2015; Montoya et al., 2016).

Unfermented Fiber and Apparent Total Tract Digestibility

The ATTD of GE, DM, EE, and carbon linearly decreased (P < 0.05) in response to increased uNDFom8, uNDFom72, and NDFom of corn DDGS (Fig. 1a–c, respectively). The ATTD of CP decreased (P < 0.01) with increasing NDFom and tended to decrease linearly (P = 0.067) with increasing uNDFom72, but ATTD of P was not affected by any fiber fraction (NDFom, fNDFom or uNDFom) of corn DDGS (data not shown). It has been well documented that there is a linear decrease in ileal and total tract apparent digestibility of GE, DM, and CP with increased levels of dietary fiber (Nortey et al., 2007; Gutierrez et al., 2013). In a meta-analysis review, Zeng et al. (2018) also reported that standardized ileal digestibility of CP and amino acids linear decreased with increasing NDF or ADF content of different sources of DDGS. The ATTD of EE has been reported to be greater in extracted corn oil compared with intact oil from corn germ meal (Kil et al., 2010). This implies that intact oil is encased within the fiber matrix and is more resistant to the formation of emulsions and enzymatic digestion than extracted corn oil (Knudsen et al., 1993; Grundy et al., 2016). In addition, slow fermentable fiber (the difference between uNDFom8 and uNDFom72) can ultimately reach the hindgut for microbial digestion and synthesis of microbial lipids (Drackley, 2000). Therefore, dietary components that increase microbial activity and microbial synthesis of lipids in the hindgut will increase endogenous EE loss and reduce the ATTD of EE. This is consistent with our results where DDGS sources that had greater unfermented fiber (uNDFom8 or uNDFom12), had less ATTD of EE, but uNDFom72 had no effect on ATTD of EE.

Figure 1.

Association between apparent total tract digestibility of gross energy (GE), carbon (C), dry matter (DM), and ether extract (EE) and unfermented NDFom8 (A), unfermented NDFom72 (B) and NDFom (C) in 15 sources of corn distillers dried grains with solubles. uNDFom = unfermented NDF after 8, 12, or 72 h of fecal incubation; NDFom = ash-free neutral detergent fiber.

Correlation Among Energy Components, Gas Production, and Chemical Composition

Gross energy was positively (P < 0.01) associated with total EE and digestible EE content in corn DDGS (Fig. 2). The DE content was negatively correlated with uNDFom8 (r = −0.86, P < 0.01), uNDFom12 (r = −0.86, P < 0.01), uNDFom72 (r = −0.86, P < 0.01), and NDFom (r = −0.84, P < 0.01), and positively correlated with digestible DM (r = 0.73, P < 0.01; Fig. 3). Similarly, ME was negatively (P < 0.05) associated with fiber-related components (NDF, fNDFom, and uNDFom), and tended to be positively correlated (r = 0.48, P = 0.07) with digestible EE (Fig. 4). However, there were no significant correlations between the digestible content of CP, ADF, NDF, and NDFom with DE or ME content (data not shown). In contrast, Noblet and Perez (1993) observed good correlations between DE and digestible CP, EE, and NDF content, and developed DE, ME, and NE prediction equations using these measures. A possible explanation for these inconsistent results can be attributed to differences in approach for determining the digestibility of nutrients (i.e., protein, starch, and lipid) that contribute to DE. Noblet and Perez (1993) directly measured the ATTD of diets and used these values for developing DE and ME prediction equations. Low molecular weight sugars were recently proposed as a necessary component of carbohydrates that contribute to energy (Navarro et al., 2018). However, the overall relevance of low molecular sugars for development of prediction equations depend on the overall contribution of this component to energy and the variability among sources and needs to be investigated.

Figure 2.

Correlation coefficients (r) between chemical composition (digestible and unfermented) and GE of 15 corn distillers dried grains with solubles samples. ADF = acid detergent fiber; EE = ether extract; NDF = neutral detergent fiber; NDFom = ash-free NDF; TDF = total dietary fiber; dDM, dEE, dADF, and dNDF = digestible EE, ADF, and NDF on a dry matter basis calculated by multiplying total concentration by the corresponding apparent total tract digestibility coefficients; dNDFom = fermented NDFom after 8, 12, or 72 h of fecal incubation; Gas12 = gas production after 12 h of fecal incubation; uNDF and uCP = undigested NDF and CP (total – digestible); uNDFom = unfermented NDFom after 8, 12, or 72 h of fecal incubation. #Means are different (P < 0.10), *Means are different (P < 0.05).

Figure 3.

Correlation coefficients (r) between chemical composition, digestible and unfermented content, and DE of sources of 15 corn distillers dried grains with solubles. NDFom = ash-free NDF; TDF = total dietary fiber; dDM, dEE, dADF, and dNDF = digestible ether extract (EE), ADF, and NDF by multiplying total concentrations by the corresponding apparent total tract digestibility coefficients); dNDFom = fermented NDFom after 8, 12, or 72 h of fecal incubation; Gas12 = gas production after 12 h of fecal incubation; uNDF and uCP = undigested NDF and crude protein (total – digestible); uNDFom = unfermented NDFom after 8, 12, or 72 h of fecal incubation. #Means are different (P < 0.10), *Means are different (P < 0.05).

Figure 4.

Correlation coefficients (r) between chemical composition (digestible and unfermented) and ME of 15 corn distillers dried grains with solubles samples. NDFom = ash-free NDF; TDF = total dietary fiber; dDM, dEE, dADF, and dNDF = digestible ether extract (EE), ADF, and NDF on a dry matter basis by multiplying by the total concentration by the corresponding apparent total tract digestibility coefficients); dNDFom = fermented NDFom after 8, 12, or 72 h of fecal incubation; Gas12 = gas production after 12 h of fecal incubation; uNDF and uCP = undigested NDF and crude protein (total – digestible); uNDFom = unfermented NDFom after 8, 12, or 72 h of fecal incubation. #Means are different (P < 0.10), *Means are different (P < 0.05).

In the current study, the ATTD values of CP and EE for corn DDGS (Kerr et al., 2013) were obtained by calculating the difference between the corn-basal diet and the corn + 30% DDGS diets. The assumption of using the difference method is that there is linear additivity of ATTD in corn + 30% DDGS diets. However, the low dietary CP and EE in the corn-basal diet increase the relative contribution of endogenous loss of CP and EE to total CP and EE in the fecal outflow compared with the corn + 30% DDGS diets. As a result, ATTD of CP and EE of DDGS may be overestimated due to the lack of additivity of ATTD values in mixed diets (Stein et al., 2005; Stein et al., 2007).

The ATTD of NDF and ADF digestibility may also vary with increasing dietary inclusion rates. Gutierrez et al. (2013) fed diets containing increasing levels (10, 20, 30, and 40%) of corn bran in corn-soybean meal diets to growing pigs and showed that the ATTD of NDF slightly decreased from 42.6% to 41.9% as corn bran was added at 10% or 20% of the diet, respectively, and was sharply reduced to 29.3% and 30.5 % when corn bran was added at 30% and 40% of the diet, respectively. This implies that the digestibility of corn fiber may be quite different when pigs are fed a corn basal diet or a corn + 30% DDGS diet. Therefore, it may be less accurate to estimate the ATTD of NDF and ADF in corn DDGS by subtracting the differences in digestibility between a corn basal diet and corn + 30% DDGS diets, as reported by Kerr et al. (2013). The in vitro system allows evaluating the nutritive value of single ingredients without the confounding effects of diet inclusion rate of the ingredient in in vivo determinations. Therefore, the in vitro unfermented fiber (uNDFom8, 12 or 72) had greater Pearson correlation coefficients with DE and ME compared with the in vivo unfermented NDF or ADF correlations.

Gas production at 8, 12, and 72 h were positively (P < 0.05) associated with fermented NDFom at the corresponding time points (Table 2). The disappearance of NDFom at 8 and 12 h was negatively associated (P < 0.05) with a time of achieving half maximal gas production (T/2). These results are consistent with previous data reported by our research group, where an increase in ATTD of TDF was observed as the maximal gas production increased among sources of DDGS, wheat straw, and soybean hulls (Huang et al. 2017a). In vitro gas accumulation measurements can be used to estimate substrate degradation and yield valuable information about feed ingredient fermentation kinetics of feed ingredients (France et al. 1993). However, the disappearance of NDF may not precisely match with gas accumulation from fermentation because gas is generated from fermenting a wide range of substrates, including both soluble and insoluble fiber components (Schofield et al. 1994).

Table 2.

Correlation coefficients (r) between energy values, total dietary fiber, and fermented neutral detergent fiber among 15 sources of corn distillers dried grains with solubles¹

Item	TDF	NDF om	uNDF om8	uNDF om12	uNDF om72	fNDF om8	fNDF om12	fNDF om72	DigNDF om8	DigNDF om12	DigNDF om72	Gas8	Gas12	Gas72	Gf	T/2	μ T/2
TDF	1.00
NDFom	0.30	1.00
uNDFom8	0.43	0.78**	1.00
uNDFom12	0.51*	0.79**	0.94**	1.00
uNDFom72	0.60*	0.73**	0.81**	0.88**	1.00
fNDFom8	−0.24	0.24	−0.43	−0.31	−0.21	1.00
fNDFom12	−0.33	0.30	−0.27	−0.35	−0.25	0.84**	1.00
fNDFom72	−0.40	0.38	−0.04	−0.12	−0.36	0.60*	0.75**	1.00
DigNDFom8	−0.34	−0.11	−0.71**	−0.60*	−0.47	0.94**	0.76**	0.49	1.00
DigNDFom12	−0.47	−0.14	−0.63*	−0.72**	−0.59*	0.76**	0.90**	0.61*	0.83**	1.00
DigNDFom72	−0.60	−0.52*	−0.72**	−0.79**	−0.96**	0.36	0.43	0.60*	0.56*	0.68**	1.00
Gas8	−0.62*	−0.15	−0.51	−0.58*	−0.45	0.57*	0.67**	0.40	0.62*	0.75**	0.48	1.00
Gas12g	−0.18	−0.07	−0.41	−0.53*	−0.38	0.54*	0.72**	0.43	0.57*	0.77**	0.44	0.77**	1.00
Gas72	−0.49	−0.06	−0.42	−0.43	−0.64*	0.56*	0.58*	0.78**	0.59*	0.62*	0.78**	0.62*	0.59*	1.00
Gf	−0.25	0.06	−0.02	−0.07	−0.46	0.12	0.20	0.70**	0.11	0.18	0.60*	0.13	0.18	0.73**	1.00
T/2	0.27	0.10	0.42	0.41	0.12	−0.51	−0.49	−0.03	−0.54*	−0.55*	−0.09	−0.58*	−0.48	−0.15	0.54*	1.00
μ T/2	0.10	−0.44	−0.40	−0.44	−0.12	−0.03	0.01	−0.44	0.13	0.22	−0.04	0.21	0.25	−0.30	−0.70**	−0.66**	1.00

¹GE = gross energy; DE = digestible energy; ME = metabolizable energy; TDF = total dietary fiber; NDFom = ash-free neutral detergent fiber; DigNDFom = percentage of fermented NDF after 8, 12, or 72 h of fecal incubation; fNDFom = fermented NDFom after 8, 12, or 72 h of fecal incubation; uNDFom = unfermented NDFom after 8, 12, or 72 h of fecal incubation. Gas = cumulative gas production (mL/g) after 8, 12, or 72 h of fecal incubation; T/2 = half-time to asymptote (hour); μ _T/2 = fractional rate of degradation (h⁻¹) at t = T/2; Gf = maximal gas production (mL/g).

*Means are different (P < 0.05).

**Means are different (P < 0.01).

Prediction Equations for DE and ME

Stepwise regression analysis of fiber-related measurements was used to generate a series of prediction equations for DE (Table 3). The initial regression equation (Eq. 1) included uNDFom8 as the most important component to predict DE followed by Eq. 2, which included both uNDFom8 and uNDFom72, and ultimately resulted in the best-fit equation (Eq. 3), where DE (kcal/kg DM) = 2,175 – 3.07 × uNDFom8 (g/kg, DM) – 1.50 × uNDFom72 (g/kg, DM) + 0.55 × GE (kcal/kg DM) (R² = 0.94, SE = 36.21).

Table 3.

Stepwise regression equation for estimating the DE content among 15 sources of corn distillers dried grains with solubles

	Regression coefficient1				Statistics2
Item	Intercept	uNDFom8	uNDFom72	GE	SE	R 2	Adjust R2
Eq. 1	4,783	−37.28			69.96	0.73	0.71
SE3	191	6.24
P-value3	<0.01	<0.01
Eq. 2	2,388	−4.19		0.51	42.89	0.91	0.89
SE3	517	0.395		0.11
P-value3	<0.01	<0.01		<0.01
Eq. 3	2,175	−1.5	−23.59	0.55	35.41	0.94	0.93
SE3	435	0.54	0.58	0.09
P-value3	<0.01	<0.01	0.026	<0.01

¹Equations were based on analyzed nutrient content expressed on a DM basis. GE = gross energy. Units are kcal/kg DM for GE and DE and g/kg DM for unfermented NDFom after 8 and 72 h fecal incubation (uNDFom8 and uNDFom72).

²SE = SE of the regression estimate defined as the root of the mean square error.

³SE and P-values of the corresponding regression coefficient.

A series of prediction equations were also developed for ME content of corn DDGS (Table 4). The initial regression (Eq. 4) included uNDFom72 as the most important component to predict ME, followed by Eq. 5, which included both uNDFom12 and GE, and Eq. 6, which included uNDFom8 (Eq. 6), and ultimately resulted in the best-fit equation (Eq. 7), where ME (kcal/kg DM) = 1,643 – 2.31 × uNDFom8 (g/kg, DM) – 2.54 × uNDFom72 (g/kg, DM) + 0.65 × GE (kcal/kg DM) – 1.42 × CP (g/kg DM) (R² = 0.94, SE = 39.21).

Table 4.

Stepwise regression equation for estimating the ME content among 15 sources of corn distillers dried grains with solubles

	Regression coefficient1					Statistics2
Item	Intercept	uNDFom8	uNDFom72	GE	CP	SE	R 2	Adjust R2
Eq. 4	3,911		−38.78			93.53	0.59	0.55
SE3	113		9.02
P-value3	<0.01		<0.01
Eq. 5	648.7		−4.49	0.67		58.49	0.85	0.83
SE3	711		0.58	0.14
P-value3	0.380	<0.01	<0.01
Eq. 6	899	−2.23	−2.57	0.71		42.64	0.93	0.91
SE3	524	0.65	0.71	0.11
P-value3	0.114	<0.01	<0.01	<0.01
Eq. 7	1,643	−2.31	−2.54	0.65	−1.42	39.21	0.94	0.92
SE3	645	0.60	0.65	0.10	0.82
P-value3	0.029	<0.01	<0.01	<0.01	0.013

¹Equations were based on analyzed nutrient content expressed on a DM basis. CP = crude protein; GE = gross energy. Units are kcal/kg DM for GE and ME and g/kg DM for CP and unfermented NDFom after 8 and 72 h fecal incubation (uNDFom8 and uNDFom72).

²SE = SE of the regression estimate defined as the root of the mean square error.

³SE and P-values of the corresponding regression coefficient.

Kerr et al. (2013) suggested that TDF is a better variable to use in DE and ME prediction models than NDF because TDF provides a more complete estimate of the fiber content in corn-DDGS. However, it is more expensive and time consuming to analyze TDF than NDF and ADF, even though most of the fiber (95 to 100%) in corn DDGS is insoluble fiber (Urriola et al., 2010). Therefore, NDF or ADF content (insoluble fiber) may serve as alternative variables to predict DE and ME in corn DDGS when TDF content is difficult to obtain in practice. Li et al. (2015) reported that NDF or ADF are useful predictors of DE and ME in both conventional high-oil and reduced-oil DDGS.

In the current study, the in vitro uNDFom8 and uNDFom72 were selected as the initial variables in the DE and ME regression model, which indicates that unfermented fiber is a better predictor of energy utilization of DDGS in pigs than analytical fiber measures (e.g., TDF, NDF, and ADF). Although these analytical fiber measures represent both fermentable and unfermentable fiber portions, they have been reported to be negatively associated with DE or ME of fibrous ingredients in pigs in several studies (Anderson et al., 2012; Kerr et al., 2013; Li et al., 2014, 2015). However, the fermentable portion of fiber can result in the production of a considerable amount of short-chain fatty acids in the hindgut and contribute 10–20% to DE in the diet (Anguita et al., 2006; Iyayi and Adeola, 2015). The energy obtained from the fermentable fiber may decrease the negative correlations between analytical fiber measures (e.g., TDF, NDF, and ADF) and DE or ME, especially because the fermentability of fiber is highly variable among high fiber ingredients. Therefore, it may be more appropriate to use in vitro unfermented fiber (as measured by TDF) as a predictor for energy utilization of DDGS. Urriola et al. (2010) reported a strong positive association between the ATTD of TDF and NDF in diets that contained 30% DDGS as the exclusive fiber source. Therefore, uNDFom may show good correlations with unfermented TDF and thus, serve as a good alternative predictor for DE and ME in corn DDGS when in vitro disappearance of TDF is not available.

It is worth mentioning that the procedure as proposed in this manuscript has some limitations. The ingredients in the fermentation process were incubated with feces prior simulation of small intestine digestion, which would remove amino acids, lipids, and starch. There is biological relevance in adding a predigestion step with pepsin and pancreatin to the model, but the objective is to predict with the simplest process possible (Bohn et al., 2018). We decided that simplicity of the model was necessary for the intended use because the pepsin and pancreatin steps would add complexity to the model. This research also provided equations that were not validated as has been done in previous work using DDGS from another data set (Urriola et al., 2014). Therefore, the proposed equations require a new experiment for validation.

In conclusion, the use of the in vitro fermentation assay is an effective method to estimate fermented and unfermented fiber content in corn DDGS. The in vitro unfermented fiber measures of uNDFom8 and nNDFom72 are the best predictors for DE and ME content of corn DDGS fed to growing pigs, compared with in vivo apparent total tract digestible nutrients, which were not good predictors for DE and ME content in corn DDGS. Further investigations are encouraged to develop energy prediction equations based on in vitro digestible nutrients (CP, EE, starch, and carbohydrates) and unfermented residues (ash and TDF).