Several effects of dietary fiber (DF ),³ nonstarch polysaccharides (NSP) plus lignin, are influenced by the bacterial breakdown of NSP in the large bowel. Fibers that are extensively degraded lose their structure and water-holding capacity, which reduces their effect on stool weight (Stephen and Cummings 1980). The main end products of bacterial polysaccharide degradation, the short-chain fatty acids (SCFA) acetate, propionate and butyrate, seem to have several functions in humans. They are rapidly absorbed by the colonic mucosa and stimulate water and electrolyte absorption (Ruppin et al. 1980). Butyrate is the preferred energy source of the colonic epithelial cells (Roediger 1982) and has been shown in cultures of carcinoma cells to increase the doubling time and to stimulate the differentiation of cells (Kim et al. 1982). Whether propionate plays a specific role in liver lipid metabolism, as has been discussed in earlier work (Chen et al. 1984), is still an open question (Topping and Pant 1995). The SCFA not utilized in the colonocytes are used as fuels at different sites of the organism (Remesy et al. 1992). At high intakes, fermentable NSP can contribute substantially to human energy supply. Hence, there are proposals to attribute energy values to NSP (Livesey 1990).

Despite the physiological importance of bacterial NSP degradation, there are only limited data in humans on the fermentation of NSP derived from foods and diets. This may be because the determination of NSP fermentation in conventional balance experiments in humans is rather time consuming and costly and requires highly motivated subjects. Fermentation studies in rats are easier to perform and less expensive, but the fermentation of several types of fiber present in mixed diets was significantly lower in rats than in humans (Bach Knudsen et al. 1994, Wisker et al. 1997). The main end products of fermentation, the SCFA, cannot be measured easily in humans, because it is very difficult to take samples from the colon, especially the proximal colon, in healthy subjects. Because about 95% of the SCFA produced are absorbed during transit of digesta through the gut (Cummings 1981), fecal SCFA are not necessarily related to the events occuring in the upper parts of the large bowel.

Various in vitro batch systems utilizing fecal bacteria have been developed for the determination of NSP degradation and SCFA production (e.g., Ehle et al. 1982, McBurney and Thompson 1990). In vitro fermentation systems may be an alternative to human and animal studies. At present, there are efforts to investigate whether the energy values of NSP supplements, e.g., for food labeling purposes, can be predicted from NSP fermentation in vitro (Barry et al. 1995). In some in vitro studies (Barry et al. 1995, Bourquin et al. 1992 and 1993), fermentation of NSP derived from single fiber sources was lower than would have been expected from published results of human and rat balance studies. This could be due to insufficient fermentation conditions in vitro (e.g., composition of the incubation medium, concentrations of inocula or substrates, incubation times), differences in the NSP sources studied or a limited fermentation capacity of the fecal inocula used in these studies. However, there are no experimental data in humans on the fermentation of a given NSP source in vitro compared with fermentation in the colon, where the situation is more complex. Monsma and Marlett (1996) suggested that inocula from the rat cecum, but not from rat feces, should be used for in vitro fermentations, because fecal inocula would lead to an underestimation of both the initial rate and maximum fermentation of dietary fiber.

If quantitative information on the fermentation of NSP is required, e.g., for the estimation of energy values of NSP (Barry et al. 1995, Livesey 1990), the reliability of in vitro systems should be validated by studies of in vivo fermentation. In addition, previous in vitro fermentations were performed with single, sometimes highly purified NSP sources (Barry et al. 1995, Bourquin et al. 1992, Englyst et al. 1987, McBurney and Thompson 1990), but there is no information on whether in vitro systems are also suitable for NSP in mixed diets containing mixtures of different fibers at variable intake levels, which is the way humans eat.

The objective of this study was to measure whether an established in vitro batch system (Goering and Van Soest 1970) using human fecal bacteria could predict the extent of fermentation in humans of NSP in mixed diets low or high in fiber and of NSP in a barley fiber concentrate containing insoluble fiber. For this purpose, fermentation data measured in vitro were compared with those obtained when the fermentation of the same NSP was determined in human balance experiments. There are significant differences in the capacity of the individual gut flora to ferment a given substrate (Bourquin et al. 1992). Therefore, to validate the in vitro system, comparisons were performed for those subjects who participated in the balance study and served also as donors of the inocula. Amount and pattern of SCFA resulting from the degradation of NSP in the diets and the fiber concentrate were also determined.

MATERIALS AND METHODS

Human balance studies

Table 1. Subject characteristics1 [View Table]

Table 2. Daily intake of fiber-containing foods during the balance studies1 [View Table]

In vitro fermentation studies

Fermentations were conducted in 125-mL Erlenmeyer flasks using the technique and the solutions as described by Goering and Van Soest (1970). The basic components of this system include substrate, culture medium, reducing solution and fecal inoculum. Culture medium (32 mL) was added to 400 mg of substrate (DF residue) 12 to 24 h before the start of the incubation to ensure complete hydration of the samples. Flasks were sealed with parafilm and stored in the refrigerator to limit the possibility of microbial growth. At 1-2 h before inoculation, the bottles were placed in a 37°C shaking water bath, reducing agent (1.6 mL) was added, and then flasks were sealed with rubber stoppers. Stoppers were fitted with three openings (an inlet tube, a bunsen valve, and a gassing tube connected to a common manifold as described by Goering and Van Soest 1970). The manifold was connected to supply of CO₂ , and flasks were bubbled with CO₂ . Into each flask 8 mL of fecal suspension was injected through the inlet tube, and incubations were performed under a steady stream of CO₂ (3-4 mL/min). Fermentation was stopped by adding 1 mL of ethylmercurithiosalicylate (10 g/L).

From each incubation flask, a 0.5-mL portion was removed for SCFA analysis. The remaining contents of the flasks were quantitatively transferred to pretared beakers, freeze-dried and used for the determination of residual NSP.

Chemical analyses

Neutral NSP monomers in foods, feces and in the in vitro fermentation substrates and residues were determined by gas-liquid chromatography as alditol acetates by the Uppsala method C (Theander and Westerlund 1986) using 1-methylimidazole as a catalyst for the derivatization of NSP monomers. Corrections for hydrolytic losses and detector response were made by performing the analyses with known sugar standards. Uronic acids were measured in the acidic hydrolysate according to the method of Englyst et al. (1982). Total NSP was calculated as the sum of neutral sugars and uronic acids. The NSP constituents and total NSP were expressed as polysaccharides (weight of monomers × 0.9).

Samples to be analyzed for SCFA were mixed with 2.5 mL of oxalic acid (0.03 mol/L) and 0.1 mL of internal standard (pivalic acid, 30 mL/L) and centrifuged (4000 × g) for 10 min. Concentrations of SCFA were measured by gas-liquid chromatography. Glass column packing material was 4% Carbowax 20M/80/120 Carbopack B-DA (Supelco, Bellefonte, PA). For quantification, analyses were performed with known SCFA standard solutions.

Calculations

Statistical analyses

The comparison in vitro vs. in vivo was performed as a paired comparison by analyzing the differences between fermentation of NSP in vitro and in vivo for the same subject, DF source and adaptation form. Least squares means for these differences were calculated by mixed model analysis to allow for heterogenity of error variance for diets (Littell et al. 1996). The statistical analysis was performed using the procedure MIXED of SAS, release 6.11 (SAS Institute 1996), modeling variance heterogenity with the option GROUP = Diet in the RANDOM statement. The correlation between in vitro and in vivo fermentation was examined by linear regression (Box et al. 1978).

RESULTS

Table 3. Chemical composition of the dietary fiber residues used as substrates for the in vitro fermentations [View Table]

Table 4. Contribution of nonstarch polysaccharide (NSP) constituent sugars to total NSP present in the experimental diets (in vivo) and in the dietary fiber residues used as substrates for the in vitro fermentations [View Table]

The fermentation of NSP contained in the experimental diets and in the barley fiber concentrate as measured in the balance experiments (in vivo) for all subjects (n = 12) and for the subgroup (n = 6) participating in the in vitro experiments is presented in Table 5. All NSP monomers and thus total NSP present in the low fiber diet were highly fermentable (>80%). When the high fiber diets were eaten, NSP fermentation was 65.4 and 61.3%, respectively, when measured in 12 subjects, and 61.8 and 59.2%, respectively, when measured in the subgroup. Glucose was the most resistant NSP monomer, followed by xylose. Fermentation of total NSP from the barley fiber concentrate was calculated to be 40.3% in 12 subjects and 31.2% in the six subjects participating in the in vitro experiments. On average, barley fiber-derived glucose was very resistant to fermentation, but there were large individual differences.

Table 5. Percentage of fermentation of nonstarch polysaccharides (NSP) in the mixed diets and the barley fiber concentrate as measured in the human balance experiments [View Table]

Correlations between in vitro and in vivo fermentation of total NSP are shown in Figure 1. The plot shows that as the fermentability of NSP rises, the differences between in vitro and in vivo data decreased. The relationship between in vitro and in vivo data was y = 0.43 ± 0.04x + 40.8 ± 2.5, r² = 0.72. 

Adaptation of the subjects to the barley fiber concentrate had no influence on the in vitro fermentation of NSP and the production of SCFA (data not shown). Therefore, for each subject results obtained from the two in vitro fermentations ("adapted" and "not adapted" to barley fiber) were pooled for analysis of variance.

The differences between the degradation of NSP in the balance experiments (in vivo) and in vitro for six subjects are given in Table 6. Total NSP in the low fiber diet was slightly, but significantly less fermented (4%; P = 0.022) in vitro than in the balance trials. Fermentation of total NSP contained in the high fiber diets and in the barley fiber concentrate was higher in vitro than in vivo, the differences being 4.9% (P = 0.193), 8.8% (P = 0.008) and 19.7% (P = 0.023) for the high fiber diets high or low in protein and the barley fiber concentrate, respectively. With respect to the main NSP constituent sugars, differences between in vivo and in vitro fermentation were most pronounced for glucose and xylose.

Table 6. Differences between the fermentation of nonstarch polysaccharides (NSP) in human balance experiments (in vivo) and in vitro1,2 [View Table]

Amounts and molar ratios of the main SCFA produced during in vitro fermentation are shown in Table 7. Acetate was the most abundant SCFA, followed by propionate and butyrate. Traces of isobutyrate, valerate and isovalerate (data not shown) were also found, but the sum of these constituted less than 5% of total SCFA and hence they were not taken into account with the calculations.

Table 7. Short-chain fatty acids (SCFA) produced during the fermentation in vitro1 [View Table]

The production of SCFA in the incubation flasks corresponded to the fermentability of NSP. Significantly more SCFA were produced during the degradation of NSP in the low fiber diet as compared with the other NSP sources, whereas fermentation of NSP in the barley fiber concentrate led to significantly less SCFA than the other substrates, with SCFA produced from NSP in the high fiber diets being intermediate. The yield of SCFA per gram of fermented NSP was similar for the experimental diets but significantly lower for the barley fiber concentrate. Concerning the single SCFA, there were some differences in the fermentation pattern. Although molar ratios of SCFA were similar for the three experimental diets, the barley fiber concentrate led to a significantly higher proportion of propionate than the other substrates, to less acetate and more butyrate than NSP in the low fiber diet, and to less acetate than NSP in the high fiber diet at low protein intake.

DISCUSSION

Our results also show that under the conditions used, human fecal inocula had a high fermentation capacity in contrast to what may be the case when rat fecal inocula are used (Monsma and Marlett 1996). The results obtained in this study in vitro agreed better with those in the human subjects than the results obtained for the same NSP sources in rats, in which fermentation was 7.7 to 22.3% lower than in humans (Bach Knudsen et al. 1994, Wisker et al. 1997). Hence, in vitro systems may be an alternative at least to rat studies and can contribute to a reduction of animal experiments.

For quantitative purposes, the results of in vitro fermentations should come close to those in vivo, but there may be several reasons for discrepancies. Theoretically, differences between in vivo and in vitro fermentation could arise from the fact that fermentability in vivo is usually obtained by the difference between dietary intake and fecal output, i.e., apparent digestibility of NSP. Feces contain bacterial and host-derived material in addition to undigested dietary residues. Resistant starch, bacteria and endogenous mucins and mucopolysaccharides contain saccharides also present in dietary NSP, which theoretically can decrease the apparent digestibility of dietary NSP. However, unabsorbed dietary carbohydrates of non-NSP origin (e.g., starch) and carbohydrates present in mucins and mucopolysaccharides are utilized by the colonic microflora (Englyst and Macfarlane 1986, Salyers and McCarthy 1989) and usually will not contribute significantly to fecal carbohydrates. In the present study, feces contained only very small amounts of starch (0-0.15 g/d). Rat and pig experiments indicate that the contribution of bacteria to fecal NSP is small and will have no important influence on the apparent digestibility of dietary NSP (Longland and Low 1988 and 1990, Nyman and Asp 1985). The pentoses, arabinose and xylose, were not detected in fecal bacteria (Longland and Low 1988 and 1990), but were important NSP constituents in our study. The only occasion when a nondietary NSP constituent sugar could cause a major error in its apparent digestibility value can be when it is a very minor component in the diet, as was the case with galactose in the present study. However, the digestibility of galactose contributed only to a small extent to total NSP digestibility. In vitro, residual NSP was corrected for (dietary and bacterial) carbohydrates present in the inocula by subtracting a substrate-free blank. However, it is likely that in the substrate-containing flasks, bacterial mass increased due to the energy provided by the breakdown of NSP (Monsma and Marlett 1996), leading to more bacterial polysaccharides in these flasks than present in the substrate-free blanks. However, it can be assumed that the content of NSP constituent sugars in these bacteria was small, similar to what occurs in vivo, and that any effect of bacterial saccharides on NSP fermentability in vitro would point in the same direction as in vivo, and hence should not have contributed significantly to the differences found.

Although it is easy to determine the fermentability of a NSP supplement in vitro, it is not easy to measure in humans the fermentation of a NSP supplement added to mixed diets because these diets already contain NSP, the fermentation of which cannot be distinguished from the fermentation of the additional fiber. As in previous studies in rats and humans (Nyman et al. 1986, Wisker et al. 1994, 1996 and 1997), the fermentation of NSP in the barley fiber concentrate was calculated from differences in NSP intake and excretion during the high fiber and the low fiber diet periods, i.e., basal fermentation was subtracted. Because animal studies indicate that NSP supplements do not interfere with the fermentation of NSP in the basal diet (Goodlad and Mathers 1991, Key and Mathers 1993), this procedure seems to be valid. However, it has some shortcomings, which are most obvious in the case of fiber constituents that are nearly completely fermented or very resistant and for those constituents present in very small amounts. Small changes in basal fiber excretions (due to real changes and/or analytical errors) can lead to negative fermentation data, e.g., in the case of poorly fermentable fibers, or to fermentabilities above 100% in the case of highly fermentable ones. The same can be the case for fermentations in vitro, where a blank is subtracted. Negative fermentations and fermentations above 100%, respectively, have been found for poorly (maize bran, cellulose) and highly (pectin) fermentable NSP in vivo and in vitro (Barry et al. 1995, Fredstrom et al. 1994, Southgate and Durnin 1970). In vitro, we observed no negative fermentations and no data above 100%. But in vivo we found in two subjects a negative fermentation of cellulose in the barley fiber concentrate, which for these subjects led to great differences between in vivo and in vitro data, because the fecal bacteria of these subjects could degrade barley cellulose in vitro. In some subjects, an increase in cereal fiber intake may shorten colonic transit time (Cummings 1978) and thereby decrease the fermentation of those NSP that are difficult to degrade, such as cereal cellulose (Van Soest et al. 1982). These potential effects of fiber in humans were not taken into account in vitro, where we used unique incubation times for all substrates.

For the in vitro fermentation of NSP in foods, diets and various fiber supplements, available nutrients such as starch should be removed prior to incubation, because these nutrients affect the estimation of NSP fermentation and contribute to SCFA production. Removal of available nutrients requires additional treatments that the fiber sources ingested in vivo are not subjected to and that may influence the extent of NSP degradation in vitro. During the preparation of the DF residues, NSP experienced additional heating steps (removal of starch and drying of residues). Heat drying, especially of wet forages at temperatures above 50°C, resulted in significant increases in the yield of acid detergent fiber because of the formation of apparent lignin, i.e., maillard products (Van Soest 1965), which may affect estimations of fiber digestibility when fiber is determined gravimetrically. In our study, a high formation of maillard products during drying of the residues is not very likely, because residues were washed free of reducing sugars, whereas such sugars are present in natural feeds and foods. Moreover, we calculated fermentability on the basis of NSP constituent sugars, the percentage of which was very similar in the in vitro substrates and in the diets. However, heating even under mild conditions can solubilize some insoluble NSP components (Wisker et al. 1994). This could have occurred during drying, but also during starch solubilization. If there was an influence of the preparation of the DF residues in the present work, only the less fermentable NSP sources (high fiber diets and barley fiber concentrate)  and within these, especially cellulose  must have been affected. The more efficient fermentation in vitro could also have been due to smaller particles of the in vitro substrates compared with the DF sources ingested in vivo. However, when NSP in coarse or fine whole meal bread was fermented in vitro, particle size did not affect NSP degradation, but these NSP were much better fermented than NSP in the barley fiber concentrate (Wisker, E., Daniel, M., Rave, G. and Feldheim, W., unpublished observations).

Probably, in the case of NSP being rather resistant to bacterial attacks, fermentation in vitro may be more sensitive to physical properties of a NSP source than fermentation in vivo, where heat treatment and particle size of natural NSP seem to have no great effect. Extrusion at 155°C of a barley-based diet increased the proportion of soluble NSP but had no effect on the fecal digestibility of total NSP and its constituents in pigs (Fadel et al. 1989). Heat treatment of carrots (blanching or canning) also increased the proportion of soluble NSP but had no effect on the fermentability of NSP in humans (Wisker et al. 1994). Particle size of cereal fiber sources had also no significant influence on the extent of NSP breakdown in humans (Van Dokkum et al. 1983, Wisker et al. 1996).

As derived from equations developed for the fermentation in the human colon, between 9.4 mmol (Cummings 1994) and 10.3 mmol of SCFA (Miller and Wolin 1979) were produced for each gram of fermented hexose. In the present study, the mean yield of SCFA was 7.4-9.4 mmol (corresponding to 0.5-0.6 g SCFA) per gram of fermented NSP. Thus, for the high fiber diets and the barley fiber concentrate, SCFA yields were lower than theoretically possible, but they were higher than those found in other studies for isolated arabinogalactan and xylan (Englyst et al. 1987).

Theoretically, energy values of NSP can be estimated on the basis of NSP fermentability (Livesey 1990) or the production of SCFA (Barry et al. 1995). If energy values of NSP are estimated by the coefficient of fermentability, the heat of combustion of NSP (17.6 kJ) and the efficiency of the conversion of fermented energy to digestible energy (0.7), as suggested by Livesey (1990), NSP in the low fiber diet, the high fiber diets high or low in protein and the barley fiber concentrate would have provided 10.3, 7.6, 7.3 and 3.8 kJ/g of ingested NSP, respectively, when fermentability in vivo was taken into account, and 9.8, 8.2, 8.3 and 6.3 kJ, respectively, when the calculations were performed on the basis of in vitro fermentability. Thus, in the case of NSP in the mixed diets, for quantitative purposes the differences observed between in vivo and in vitro fermentation can be neglected. In the case of the barley fiber concentrate, it is likely that a shorter incubation time would have reduced the differences between in vivo and in vitro fermentation. However, it is not obvious, from our study, whether similar results can be expected for other resistant types of NSP.

In vitro production of SCFA was highly correlated with energy values of fiber supplements in rats (Barry et al. 1995). Whether it is possible to estimate from the amount and pattern of SCFA produced in vitro the energy provided to humans by NSP remains to be investigated. The estimation of energy values of NSP on the basis of in vitro SCFA production may have advantages, because analyses of SCFA in the incubation medium may be easier to perform than NSP analyses.

Although differences between the fermentation measured in humans and in vitro were significant for two diets, the magnitude of the differences was such that fermentation of NSP in mixed diets could be predicted with sufficient accuracy in vitro. These results indicate that for NSP in mixed diets, in vitro fermentation systems may replace expensive and laborious human fermentation studies or animal experiments. However, the fermentation in humans of NSP in the barley fiber concentrate was considerably overestimated in vitro. Further knowledge of the effect of the physical properties of resistant NSP types on their in vitro fermentability may contribute to a better agreement between the fermentation in vivo and in vitro.