Difference in postprandial GLP-1 response despite similar glucose kinetics after consumption of wheat breads with different particle size in healthy men

Ten healthy men [age 24 ± 0.6 years, BMI 22 ± 0.2 kg/m² (mean ± SEM)] were recruited. The main criteria for exclusion were use of medication, blood donation, or use of antibiotics in the past three months, gastrointestinal surgery or dysfunction, inflammatory diseases, and diabetes mellitus. Approval was obtained from the Medical Ethics Committee of the ‘Beoordeling Ethiek Biomedisch Onderzoek’ foundation, Assen, The Netherlands. Each subject gave written informed consent for the study. This trial was registered at trialregister.nl as NTR3020.

In the overall study design, addressing two different research questions, four different wheat-based products were tested. To increase clarity and be able to focus on the results of one of these questions, only two meals are described in this paper. A description of the other products and the results can be found elsewhere [19]. The study was performed in a randomized, crossover manner, with at least 1 week between each study day. The subjects were asked to refrain from consuming foods naturally high in ¹³C, like cane sugar, corn products, and pineapple, for 3 days preceding the experiments and from alcohol consumption and strenuous exercise for 24 h before each study day. Food intake on the day before each study day was individually standardized. To minimize variation, a standard evening meal was provided at the commercial research facility (QPS Netherlands B.V.), where the participants stayed overnight. In the evening, a venous catheter was inserted in each forearm for blood collection and for infusion of the tracer D-[6,6-²H₂]glucose (98 % ²H atom percent excess, Isotec) [20, 21]. Subjects fasted overnight, but were allowed to drink water. In the morning (t = −122 min), a bolus of 26.7 mL D-[6,6-²H₂]glucose solution (80 × 0.07 mg/kg body weight) was injected within 2 min, and a continuous infusion of 0.07 mg/kg body weight D-[6,6-²H₂]glucose per min was started (t = −120 min) and maintained for 8 h (until t = 360 min). The bolus amount was 80× the infusion rate over 1 min, according to the paper of Tissot et al [20]. The infusion rate over 1 min was 0.07 mg/kg BW [21], to reach a good steady state. So the continuous infusion contained 4.2 mg/kg BW per 60 min, administered at a desired infusion rate of 20 mL/h, resulting in the concentration of 0.21 mg/mL/kg BW. The bolus had the same concentration, and therefore, to administer 80 × 0.07 mg/kg BW an amount of 26.7 mL was administered.

Two hours after the start of the infusion the test meal was ingested (t = 0 min). Figure 1 shows a simplified time line of each study day. Water (150 mL) was provided hourly, starting at t = 120 min. During the study period physical activity was limited.

Two types of wheat bread were prepared (TNO, Zeist, The Netherlands), with the same composition, but a different structure, due to different milling and bread making processes; a kernel bread (KB) prepared with flour and 85 % broken wheat kernels, and a control bread (CB) made from wheat flour, combined with wheat bran, to obtain the same dietary fiber content as in the KB.

To provide the necessary ingredients, unlabeled (1.085 at.% ¹³C) wheat grains [T. aestivum var Capo, grown in Austria] were milled in different ways; by conventional roller-milling the wheat kernels were milled to obtain refined white flour and wheat bran, and broken wheat kernels were obtained using a breaker mill. For ¹³C-enrichment of the products, ¹³C-labeled wheat [T. aestivum var Paragon (1.359 at.% ¹³C)], cultured in a ¹³CO₂ enriched atmosphere, was used. A small part of the ¹³C-labeled wheat kernels was milled to obtain a fine wholemeal wheat flour, and the other part was broken to obtain broken kernels that were similar to the unlabeled broken kernels.

CB was prepared with 1446 g unlabeled white wheat flour, 240 g ¹³C-labeled wholemeal wheat flour, 314 g wheat bran, 1300 g water, 33.4 g yeast, 36 g salt, 3 g malt, and 70 ppm ascorbic acid. After kneading, the dough was left to rise for 30 min, molded, and left to rise for 60 min. KB was prepared with 264 g unlabeled white wheat flour, 36 g ¹³C-labeled wholemeal wheat flour, 1496 g unlabeled broken wheat kernels, and 204 g ¹³C-labeled broken wheat kernels. All broken kernels were soaked in 1000 g water overnight at 10 °C. The following day, the soaked kernels were mixed with the flour, 230 g water, 33.4 g yeast, 36 g salt, 3 g malt, and 70 ppm ascorbic acid. The resulting dough was left to rise for 30 min, molded, and left to rise for 50 min. Subsequently, the breads were baked for 30 min at 240 °C. Bread slices were stored at −20 °C until use.

All test meals provided 50 g available carbohydrates; for the CB, the portion size was 138 g and for the KB 137 g. The breads were consumed together with 10 g light margarine (4 g fat), 2 slices lean ham (5 g fat, 6 g protein), and 250 mL tap water within 20 min. We assume that the digestibility of starch from both test products is similarly affected by this addition.

Starch, dietary fiber, and moisture content were determined at Eurofins Analytico Food, The Netherlands. To quantify starch fractions (G_T, G_RA, G_SA, G_TA, and RS) of the test meals in vitro, an adapted version of the Englyst method [22] was used [23].

Particle size distribution of the broken kernels was determined by sieve analysis in duplicate.

Bread products volume (mL) was determined in triplicate by rapeseed displacement (AACCI Method 10-05.01 [24]) and, together with the bread weight (g), the product overall specific volume (mL/g) and density (g/mL) were obtained.

Breads were further characterized using microscopy and X-ray microtomography (XRT, porosity). A detailed description of both techniques can be found in Online Resource 1.

Blood was collected into several blood collection tubes (BD Diagnostics): 2 mL fluoride tubes (NaF) for glucose and bile acid measurements, 3 mL EDTA tubes [+30 µL DPP-4 inhibitor (Millipore)] for GIP, GLP-1, glucagon, and insulin determination, and 3 mL lithium/heparin tubes (CCK). Three basal blood samples were collected (t = −60, −30, −15 min), and postprandial samples were drawn every 15 min for 2 h, every 30 min for an additional 3 h, and once after 6 h. To obtain heparin plasma, blood was collected at t = −60, t = −15 min and then every 30 min for the first 3 h, and hourly for the last 3 h. After centrifugation (1300×g for 10 min at 4 °C), plasma aliquots were stored at −20 °C (NaF and EDTA plasma) or −80 °C (heparin plasma) until analysis.

Breath samples were collected by breathing through a straw into 10 mL Exetainer^® vials (Labco Limited). Two basal breath samples were collected (t = −30, t = −5 min), and after the test meal, a sample was taken every 30 min until t = 360 min.

To get an impression about the impact of the breads on appetite, subjects were asked to rate their feeling of appetite (hunger) using a visual analog scale (VAS) at 15 min before and hourly after the test meal, although the number of subjects (n = 10) is insufficient to detect differences in these subjective measures [25]. At the same time points, their feeling and extent of discomfort (abdominal pain, flatulence, other complaints) was recorded (0 = no complaints, 3 = severe complaints). Shortly after consumption of the meal, the subjects scored (VAS) how tasty the meal was.

Plasma glucose concentrations were measured on a Roche/Hitachi Modular automatic analyzer (Roche Diagnostics, Hitachi) using a glucose hexokinase method. The within- and between-run CV were ≤2 %. The ARCHITECT ^® insulin assay (Abbott Laboratories) was used to determine insulin concentrations in plasma. The total CV of this chemiluminescent microparticle immunoassay was ≤7 %. The glucagon assay was directed against the C-terminal of the glucagon molecule (antibody code no. 4305) and therefore measures glucagon of mainly pancreatic origin [26]. Total GIP was measured using the C-terminally directed antiserum (no. 80867) [27], which reacts fully with intact GIP (1-42) and the N-terminally truncated metabolite GIP (3-42). Total GLP-1 concentrations were determined as previously described [28], using a radioimmunoassay (antiserum no. 89390) specific for the C-terminal of the GLP-1 molecule and reacting equally with intact GLP-1 and the primary (N-terminally truncated) metabolite. The glucagon and incretin assays have detection limits of <2 pmol/L, and an intra-assay coefficient of variation of approximately 6 %. CCK was measured using antiserum no. 92128, which binds the bioactive forms of CCK with equal potency without cross-reactivity with any gastrin [29]. The detection limit of the CCK assay is 0.1 pmol/L, and the intra-assay coefficient of variation approximately 5 %.

Fasting and postprandial concentrations of 15 individual BAs were determined using LC/MS: the primary BAs [cholic acid (CA) and chenodeoxycholic acid (CDCA)], the secondary BAs [deoxycholic acid (DCA), ursodeoxycholic acid (UDCA), and lithocholic acid (LCA)], as well as their glycine (G, glyco-) and taurine (T, tauro-) conjugates. The lower limit of quantitation (LOQ) was 0.05 µM. Concentrations of TCA, LCA, GLCA, TLCA, TDCA, UDCA, and TUDCA were below the LOQ. The intra- and inter-assay CV ranged from 1.6 to 11.3 and 4.4 to 13.3 %, respectively. For sample preparation, 250 µl of internal standard solution was mixed with 25 µl plasma and centrifuged at 15.900×g for 10 min. The supernatant was transferred into a new vial, evaporated under nitrogen at 40 °C, and reconstituted in 100 µl of 50 % methanol. The solution was filtered with a 0.2-µm centrifugal filter at 2000×g for 10 min. After this step, the samples (injection volume 10 µl) were ready for analysis, using two different LC/MS systems. A detailed description of the systems and settings can be found in Online Resource 2.

Analysis of ¹³C abundance in breath CO₂ was performed using GC/IRMS (Delta Plus XL; Thermo Fisher Scientific) measuring the ¹³C/¹²C ratio versus the international standard Pee Dee Belemnite (δ¹³C_PDB, in ‰).

Plasma sample preparation required for analysis of isotopic enrichment by GC (derivatization) is described in detail elsewhere [30, 31]. ²H enrichment was measured by GC/MS as previously described [30], and ¹³C/¹²C isotope ratio was measured using GC/C/IRMS as previously described [32], both with some modifications [33].

Calculation of glucose kinetics was performed as previously reported [34].

To determine differences in glucose kinetics and plasma glucose, insulin, glucagon, incretin, CCK, and bile acid concentrations, the 0–2 and 0–6 h iAUCs were calculated as previously described [34].

Human data are presented as mean ± SEM, n = 10. Baseline-subtracted values are shown in the figures and were used in the analyses as well, for clarity, consistency, and comparison of variables. The overall study involved testing of four test meals. From the statistical analysis of the full experiment, we highlight only the comparisons involving the products of current interest. For parameters that result in individual points of a time curve, we fitted a model that accounts for the differences between the 10 subjects and the four occasions, the main effects of treatment and time, respectively, and the time × treatment interaction. Residual plots based on such a model were used to assess approximate normality and homogeneity of variances. Parameters that were not compatible with these assumptions were transformed before definitive analysis. Most variables were log-transformed, except for glucose (reciprocal values), VAS scores (angular transformation), RaE, and ¹³CO₂ values (Poisson regression model on log scale).

We performed F tests on time × treatment interaction effects to assess whether the four test meals gave rise to curves of a different shape. If the test resulted in a P value <0.05, we tested differences between the meals for each time point. If there were no statistically significant differences in shapes, we tested overall differences between the meals based on the above model without the time × treatment interaction.

To assess summary measures such as iAUC, we fitted a model accounting for the differences in the means of the 10 subjects and the four occasions, and the main effects of treatment. Transformed data were used whenever residual plots showed incompatibility with normality and homogeneity assumptions. We performed an F test to see whether there were differences between the meals. If this test resulted in a P value <0.05, pairwise comparisons among the meals were conducted.

A Benjamini–Hochberg correction was applied on P values from all partial tests to correct for multiple comparisons, resulting in a set of differences in which at most 5 % were falsely selected (P < 0.0052 for partial tests was considered significant). The analyses were performed with the software package GenStat, release 13. The within-subject relationship (correlation) between variables was tested by regression analysis according to the method of Bland and Altman [35]. Test meal differences (e.g., density, porosity) were assessed using a Student’s t test; a P value <0.05 was considered significant. These analyses were performed using SPSS 20.0 for Windows (SPSS Inc., Chicago).

The breads (CB and KB, respectively) were comparable in amount of starch (36.2 and 36.4 %), dietary fiber (7.6 and 8.3 %), and moisture content (38.6 and 37.8 %). In vitro quantification of starch fractions showed similar digestive characteristics for CB and KB when a sample from the total bread was analyzed (Table 1).

A main difference between the two breads (Fig. 2a) was their density; CB had a density of 0.29 ± 0.007 g/mL compared to 0.41 ± 0.003 g/mL for KB (P < 0.05). The difference in density was clearly visible in the product structure (Fig. 2b). In the case of KB, 85 % of the wheat material consisted of broken kernels. The majority (72.6 %) of the broken kernels had a particle size between 1680 and 2800 µm as measured by pan sieving. Both visual evaluation and microscopy confirmed that the KB structure consisted of broken wheat kernels with a particle size of 1 to several mm (Fig. 2b). Light microscopy with starch (lugol; blue) and protein staining (Ponceau 2R; red) (Fig. 2c) showed that CB had a porous structure and the thin cell walls were dominated by a continuous phase of starch granules in blue. In this matrix bran particles were clearly visible as clusters of red aleurone cells aligned on the outer side by pericarp layers. An apparent difference was the presence of broken wheat kernels in which no starch granules (blue) were observed (Fig. 2c). Apparently, the matrix was governed by proteins (in red) embedding the starch granules.

A description of the porosity measurements of both breads using XRT can be found in Online Resource 1.

Postprandial glucose concentrations were similar after the consumption of CB and KB (Fig. 3a; Table 2). The insulin response was lower after KB consumption compared to CB at t = 60 min (P = 0.002) (Fig. 3b), which resulted in a 31 % smaller iAUC (0–2 h) compared with CB intake (NS, P = 0.011).

The RaE was similar after CB and KB consumption (Fig. 4a), except for t = 120 min, where RaE was higher after CB intake (P = 0.005). EGP was not significantly different after consumption of CB and KB. The rate at which glucose was cleared from the circulation (GCR) was also similar (Fig. 4b; Table 2).

The average postprandial GIP response was somewhat higher after CB compared with KB intake (Fig. 5a; Table 2), but a significant difference was only found at t = 120 min (P < 0.005). The average GLP-1 response was higher after CB compared to KB consumption at each time point (Fig. 5b), resulting in a difference in the 0–2 and 0–6 h iAUC (Table 2, P < 0.005). Differences between time points could, however, not be determined, as time × treatment interaction failed to reach significance (P = 0.073).

There was a slight increase in CCK concentrations in response to the test meals, which seemed somewhat lower after the KB from 90 min postprandial (Fig. 6a; Table 2), although differences were not significant (iAUC 0–6 h, P = 0.03).

The postprandial pattern of glycine and taurine conjugated BAs in plasma was intra-individually identical, with the highest concentrations for GCDCA, followed by either GDCA or GCA. The unconjugated BAs (primary and secondary) also responded in the same way within each person per test period. Therefore, by summing up concentrations, individual BAs were grouped as conjugated and unconjugated BAs, and together they formed the total BA response (Fig. 6b). Unconjugated BAs in plasma showed a decrease after test meal consumption (Fig. 6c), whereas the conjugated BAs increased postprandial and peaked around t = 30 min (Fig. 6d). After CB consumption, conjugated BAs showed a second peak around t = 120 min. Due to the great inter-individual variation in postprandial responses, F tests did not result in significant P values, so differences between CB and KB could not be tested.

¹³CO₂ excretion in breath, reflecting the rate of oxidation of the ¹³C-labeled substrate, was higher after CB consumption at t = 210 min (P = 0.0014). The time to peak was not different after the meals, but the peak value was significantly higher after CB consumption (P = 0.005) (Table 2).

Correlations between several variables (all time points, 0–6 h; P < 0.05) were determined (Table 3). RaE and GIP were very well correlated, especially for CB (CB, r = 0.84; KB, r = 0.71). RaE and GLP-1 showed a moderate correlation (CB, r = 0.70; KB, r = 0.58). The correlation between GLP-1 and conjugated BAs was moderate for CB (r = 0.60) and lower for KB (r = 0.39). Total BAs with GLP-1 showed similar correlation coefficients (CB, r = 0.53; KB, r = 0.43).

The subjective sensation of appetite (hunger), as determined hourly using a VAS, was similar after consumption of CB and KB (data not shown). Occasional mild complaints about flatulence were not meal type related. The liking of the test meal was rated by using a VAS (0 = not tasty, 100 = very tasty). The resulting scores (CB = 57, KB = 33) indicate that the KB was not well appreciated by the subjects.