Impact of food processing on rye product properties and their in vitro digestion

Of the six products included in the study, three were commercially available: refined wheat bread (WB) (Pågen AB, Sweden), yeast-fermented whole grain rye crispbread (RCB) and unfermented whole grain rye crispbread (uRCB) (Barilla Sweden AB, Sweden). Flour of the same origin and composition, but milled to different particle sizes, was used for production of RCB and uRCB. According to data provided by the manufacturer, the uRCB flour had 30–42% particles <125 µm and 20–28% > 1040 µm, while the corresponding proportions in the RCB flour were 51–57% and 3–6%, respectively. RCB was fermented in two steps; first 120 min at 29 °C, followed by 35 min with an increase from 30 to 38 °C. For uRCB, flour was mixed with water at 12 °C and then whipped at 6 °C to incorporate air into the batter.

The other three products were produced in-house, using the same flour as for RCB. Sourdough-fermented whole grain rye bread (sRB) was prepared by mixing 5150 g whole grain rye flour, 35.9 g NaCl, 1300 g commercial sourdough (Jästbolaget AB, Sweden), 125 g fresh yeast and 4405 g H₂O. The dough was mixed for 4 min at low speed and 4 min at high speed in a dough mixer (Varimixer, Charlotte, NC, USA), proofed for 30 min at room temperature, divided into 900 g portions and placed in baking tins, and then put in a proofing chamber at 38 °C and 80% relative humidity for 40 min. Baking (50 min) was initiated at 230 °C with 8 s of steam, and then immediately lowered to 190 °C.

Extruded whole grain rye (extR) was produced at VTT Technical Research Centre of Finland (Espoo, Finland). Whole grain flour, with addition of 0.8% NaCl, was extruded using a co-rotating twin-screw extruder APV MPF 19/25 with conveying and mixing elements (Baker Perkins Group Ltd, Peterborough, UK). The flour feed rate was 50 g/min, water addition 3.0 ml/min, screw speed 400 rpm and in-barrel residence time around 2 min. Die pressure was measured using a pressure transducer (Dynisco Ltd., UK) in the die plate. The temperature profile of the four heating blocks was: 140–100–80–60 °C (from die exit to feeding section). The parameters measured were: torque 76–86% (of maximum 100%), pressure at die 3.9–4.4 bar and temperature at die 124 °C. The expanded products were cut by the cutter blade (in front of the die exit) into breakfast cereal spheres (5–7 mm in diameter) and oven-dried at 80 °C for 15 min.

Whole grain rye porridge (RP) was prepared by mixing 64 g whole grain rye flour and 0.58 g NaCl with 200 g boiling water and stirring with a spoon for 2 min to ensure good mixing. The porridge was allowed to rest for 3 min before use.

Salt content in the rye products was equal and based on the content in the commercially produced rye crispbreads. WB had a higher content of salt (1.5% db.).

Samples were milled with a cyclone sample mill (Retsch, Haan, Germany). Samples with high water content (WB, sRB, RP) were freeze-dried prior to milling. RP was prepared as described above and immediately frozen in liquid nitrogen before freeze-drying. Extractable and unextractable dietary fibre content and composition were analysed according to the Uppsala method [18]. Content of arabinoxylan and arabinogalactan was calculated assuming an arabinose/galactose ratio of 0.69 in extractable arabinogalactan [19]. β-glucan, fructan and resistant starch content was analysed using a K-BGLU kit [20], a K-FRUC kit [21], and a K-RSTAR kit [22], respectively (Megazyme, Bray, Ireland). Calcofluor average molecular weight of β-glucan (M_cf) was analysed using size exclusion chromatography with fluorescence detection [23]. Crude fat was determined according to the method described in the Official Journal of the European Communities [24] and protein according to the Kjeldahl method (N × 6.25) [25]. Dry content was determined by drying the samples at 105 °C for 16 h according to AACC method 44-15A.

Samples for microstructural characterisation were collected from the products, after mastication, after static simulated gastric digestion and in the TIM model from the duodenum (I in Fig. 1) at 30, 60, 90, 120 and 180 min, and from the ileal delivery (III in Fig. 1) at 60, 90, 120, 180, 240 and 360 min. All samples, except product samples to be embedded in plastic, were frozen in liquid nitrogen on collection and stored at −80 °C until analysis. Image J (fiji.sc/Fiji) was used for processing of micrographs and figures.

Statistical analysis was conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). Differences between glucose profiles were evaluated with a mixed effect model, PROC mixed, suitable for repeated measurements. For glucose profiles, separate models were used for concentration–time profiles and area under the curve (AUC). AUC was calculated using the trapezoid rule for the intervals between 0–90 and 0–180 min. For concentration models, time, product and a time × product interaction term were used as fixed effects, with time as a repeated variable. Run order was included as a random factor. When significant time × product interactions were found, pair-wise comparisons were performed at these time points. For AUC models, only product and run order were included in the model and pair-wise comparisons made. Differences in final viscosity between samples were evaluated using one-way ANOVA. All pair-wise comparisons were made using Tukey’s honest significance test.

Water content (% wb) was 35.7 and 53.0% for the soft breads WB and sRB, 7.5, 8.6 and 8.0% for the dry products RCB, uRCB and extR, and 77.0% for the semi-solid RP. The protein, fat and starch content were higher in WB than in the rye products, while the rye products contained more fibre (Table 1). Total fibre content was comparable between the rye products, but fibre composition varied. β-glucan and arabinoxylan differed only slightly in content, while average molecular weight (M_cf) of β-glucan was lower in the fermented sRB and RCB than in the unfermented uRCB, extR and RP. β-glucan extractability was higher in the extruded rye product than in the other unfermented products. The extractable arabinoxylan content followed a similar pattern, being lowest in uRCB and RP. Resistant starch was highest in sRB and lowest in extR, but relatively similar in the other products.

Differences between the products in terms of structure and distribution of β-glucans and arabinoxylans were observed in the CLSM micrographs (Fig. 2). β-Glucan was distributed as smaller fragments throughout the matrix in the fermented WB, sRB and RCB, compared with the unfermented uRCB, extR and RP. Arabinoxylan followed a similar pattern, but was also distributed throughout the starch matrix in RP. In extR, β-glucan and arabinoxylan were clearly separated and distributed as smaller fragments throughout the matrix. The cell walls in RP appeared more swollen than in the other products.

In WB, protein formed a continuous network, encapsulating starch granules, while all rye products had a continuous starch network encapsulating protein (Fig. 3). For RCB and uRCB, the starch phase consisted of highly swollen starch granules. In uRCB the granules were often indistinguishable, indicating a higher degree of starch gelatinisation than in RCB. The lamella also appeared thinner in uRCB than in RCB. Furthermore, uRCB contained larger pieces of bran and intact cell structures compared with the other rye products. In sRB, starch granules were less swollen and surrounded by a layer of leaked amylose. Leaked amylose was also observed in uRCB and, to a lesser extent, in RCB. In extR all starch granules were completely disrupted, resulting in a homogeneous starch phase, which encapsulated small fragments of cell walls and aleurone layers. The thickness of the lamella in extR was comparable to that in uRCB. In RP, larger fragments, consisting of aleurone layers and starch granules encapsulated in intact cells, were separated by a continuous phase consisting of loose starch granules in a dilute phase of leaked starch.

The number of chewing cycles before expectoration varied between the products: 3–5 for RP, 10–15 for extR, 15–20 for WB, 20–25 for sRB and 25–30 for RCB and uRCB. Bolus water content (%wb, standard deviation in brackets) was 51.6% (2.0), 49.2% (2.2), 48.9% (1.9) and 47.2% (1.5) for WB, RCB, uRCB and extR, respectively, 60.5% (1.3) for sRB and 75.1% (0.1) for RP. During mastication, the protein/starch matrix of WB appeared to be compacted, forming a bolus consisting of aggregates of starch granules and protein (Fig. 4). For sRB and RCB, the protein/starch matrix instead appeared to have fractured during mastication, resulting in fragments retaining structural features of the original product matrix. In sRB, the fragments appeared more fractured than in RCB. The uRCB bolus also contained fragments, but these mainly consisted of aleurone layers or endosperm cells connected by a weakly stained starch phase, most likely due to hydration. For extR, a similar hydrated starch phase mainly characterised the bolus and, as in WB, the structure appeared to have been compacted into aggregates. RP appeared unaffected by the masticatory process.

After completed gastric digestion, the protein network in WB was completely digested, leaving only free starch granules. For sRB the fragments from the bolus had been mostly disintegrated, while for RCB fragments similar to those observed in the bolus remained. In both uRCB and extR the hydrated starch phase had been disrupted, but the fragments with intact cells in uRCB remained. For RP the effect appeared to be mainly one of dilution.

The glucose concentrations and AUC values differed between the products and there was a statistically significant product × time interaction (Fig. 6). The AUC value for the whole curve, AUC_{0–180 min}, was lower for WB, sRB and RCB than for RP. For AUC_{0–90 min}, only sRB was significantly lower than WB (p < 0.05). For specific time points, the glucose concentration at 120 min was higher for RP than for all other products and for uRCB and extR compared with WB, sRB and RCB. At 180 min, glucose concentration was higher for sRB, extR and RP compared with WB. For WB and RCB, the peak value occurred at 90 min, while for all other products it occurred at 120 min.

Progressive degradation of the products was observed in samples collected from the duodenal compartment at different time points (Fig. 7). In the WB digesta, the protein/starch aggregates found in the bolus could be observed initially. Over time, these aggregates decreased in size and after 120 min were almost absent. In all rye products except extR, the digestion of starch was observed to proceed from the outside towards the centre of larger fragments. The amount of starch remaining also appeared to decrease with time. Fragments consisting of intact endosperm cells and aleurone layers were observed in sRB, RCB, uRCB and RP, while in RCB there were also fragments consisting of only starch granules, as observed after simulated gastric digestion by RVA. In extR, the digesta consisted of notably smaller starch particles than those seen after simulated gastric digestion. In general, the observations from the TIM model were in line with the particle size distribution after 2-h simulated gastric digestion. The main discrepancies were observed for WB, where larger aggregates were initially present, and extR, where the particles appeared smaller.

The development of digesta viscosity for the different products will depend on the progressive disintegration of the bolus and hydration, swelling and solubilisation of different components [31]. The rapid initial decline in viscosity and the progressive decrease in frequency and amplitude of peaks was most likely the result of disintegration of the boluses. Although this does not give a direct measure of size or amount of larger fragments the longer duration of occurring peaks could indicate a slower disintegration of the boluses of uRCB and extR (Fig. 4). This could be related to more cohesive boluses resulting from the formation a connective starch phase during mastication, as observed for uRCB and extR (Fig. 4). The thinner lamella and more disrupted and gelatinised starch granules in uRCB and extR compared with sRB and RCB (Fig. 3) may have resulted in structures which were more easily hydrated by saliva during mastication. Plasticisation of the starch phase could give more flexible structures which were compacted during mastication, rather than fractured as appeared to be the case for sRB and RCB (Fig. 4). The continuous protein phase in WB may in a similar way have contributed to a more cohesive bolus compared with sRB and RCB. Refined wheat bread has also been reported to form larger particles after mastication than wholegrain and endosperm rye sourdough bread [32]. The viscosity curve for sRB seemed to stabilise rapidly compared with the other products, including RCB. This could be related to a weaker structure which was more easily disintegrated. Fractures were also observed in the bolus fragments (Fig. 4). The formation of fractures may have been promoted by less swollen, amylose-surrounded starch granules in sRB. sRB also appeared more disintegrated than RCB after completed gastric digestion (Fig. 4), as reflected in the particle size distribution (Fig. 5). Peaks in the viscosity curve were also observed for the semi-solid RP, indicating presence of larger bolus fragments. As RP was used in a heated state, submersion of the bolus in the colder simulated gastric fluid most likely resulted in gelling of the continuous starch phase and solidification of the bolus which then disintegrated.

In the present work only one individual was used to chew the products and expectorate prior to swallowing (and transfer into the stomach of the dynamic model), since previous investigations have reported that the inter-individual variability of food bolus particle size is very limited, as is the effect of salivary α-amylase in relation to the action of pancreatic α-amylase [26, 33, 34]. A limitation with mastication is that smaller particles may be lost by “intermediary swallowing” following dispersion in the saliva, and thereby not included in in vitro digestion [35]. Further studies are needed to evaluate and identify key parameters involved in cereal starch digestion and to confirm that the use of one individual to chew is representative.

The observed differences in final viscosity between the products (Fig. 5) may partially relate to fibre composition. Lower molecular weight of β-glucan (Table 1) and disrupted cell walls (Fig. 2), as observed in sRB and RCB compared with uRCB, extR and RP, is an indication of fibre degradation. β-glucan is known to be degraded by endogenous enzymes, which become active with the increased moisture content during fermentation [36]. Comparing uRCB with extR and RP, the time needed for mixing and baking of uRCB appeared sufficient for some degradation to occur. Although not analysed, molecular weight of arabinoxylans has been shown to be degraded similarly, but not to the same extent, as β-glucan [37, 39]. Arabinoxylan-degrading enzymes are mainly active at higher temperatures and lower pH, around 4.5, than β-glucanases [37, 39], and in uRCB, extR and RP arabinoxylan was probably relatively unaffected. In sRB and RCB; however, some degradation likely occurred. An increase in polydispersity and slight decrease in molecular weight has also been reported for fermented rye crispbreads compared with unfermented [38]. Furthermore, due to lower dough pH, the use of sourdough in sRB might have promoted more extensive degradation of arabinoxylan.

The digesta viscosity is also influenced by the characteristics, e.g., size and shape, of the particles present, although the relationship is not well known [31]. This would explain the relatively low digesta viscosity of extR, despite high extractability and retained molecular weight of β-glucan and high extractability of arabinoxylan, as it contained the smallest particles of the rye products (Fig. 5).

Viscosity and the rate of disintegration during gastric digestion can influence how rapidly starch becomes available for digestion in the small intestine. Solid foods are considered to be emptied from the gastric compartment first when reaching particle size < 1–2 mm [40]. Furthermore, in mixed meals, liquids are preferentially emptied first [41]. As boluses of cereal foods typically form more cohesive masses than e.g., vegetables [42], the rate of disintegration is likely to be of importance for gastric emptying and consequently the postprandial responses in humans [43]. High viscosity can also contribute to reduced gastric emptying [44]. Furthermore, both content of soluble fibres and viscosity can influence the diffusion rates of enzymes and glucose [8, 9, 45]. During digestion in vivo or in dynamic in vitro models, digesta viscosity will decrease due to continuous dilution by gastric and intestinal juices. However, differences in both viscosity and diffusion rates between products resulting from variations in particle characteristics and fibre composition may still persist.

The differences in glucose release between products were not clearly reflected in the progression of starch digestion as observed in the digesta from the TIM model at different time points (Fig. 7). The model has previously been used to visualise differences in starch digestion [17], but in that study a larger difference in total release of maltose was reported and both products tested, oat and barley tempeh, were relatively similar in structure. In our study a range of products with varying structures and properties were used, so different factors may have contributed to the concentration profiles of each product.

It is possible that slower gastric disintegration of uRCB and extR (Fig. 4) contributed to a later peak in glucose concentration, but similar AUC, compared with WB and RCB by decreasing the rate at which starch became available for digestion. No large bolus fragments (>2 mm) were observed in the intestinal compartments, and gastric sieving appears to occur in the TIM model too, indicating that gastric disintegration will be a factor to consider. Furthermore, lower diffusion rates due to less degraded fibres could decrease the rate of starch hydrolysis and, in the TIM model, reduce the rate of filtration/removal of digested compounds through the dialysis filters. This might also have contributed to the shift in glucose curves observed for uRCB and extR compared with WB and RCB. The highly disrupted starch in extR could have been expected to result in faster starch hydrolysis, but the gastric disintegration and diffusion rates may be more important factors. Despite similar glucose profiles in the TIM model, the responses to uRCB and extR may differ in humans, as the lower digesta viscosity of extR may result in faster gastric emptying rate [39]. The difference in glucose profiles between RCB and uRCB is in line with results of a recent comparison of RCB and uRCB with a refined wheat crispbread, where RCB produced an insulin response more similar to that of the wheat reference than to uRCB [5].

Differences in the glucose profiles for RCB and WB could have been expected considering the large differences in fibre content and composition, viscosity and particle size distribution. However, at equal viscosities, lower diffusion rates have been demonstrated for solutions of arabinoxylan from wheat compared with arabinoxylan from rye [9]. Moreover, as observed in the TIM model, particles were initially of comparable sizes for WB and RCB (Fig. 7). While starch hydrolysis rate is often reported to be higher for refined wheat breads than rye breads, measurements are usually made after simulated gastric digestion [4, 6] and the progressive changes in particle sizes occurring in vivo are not accounted for.

For RP, the large product volume, due to its high water content, may have contributed to its high glucose AUC. Compared with the other products, RP occupied a larger fraction of the volume in the gastric compartment and less liquid could be emptied before only the product remained, which should result in faster emptying of product. In humans this may compensate for slower gastric emptying due to high digesta viscosity [44]. Moreover, due to the high water content during preparation and the temperature of the product when initiating in vitro digestion, starch may have been more gelatinised and easily hydrolysed. The high AUC for RP compared with sRB is also in line with findings in a human study by Rosén et al. [7], where endosperm rye porridge and whole grain rye porridge induced higher glucose and insulin responses during the first 30 min than corresponding rye breads.

Neither of the explanations above can account for the low AUC_0–90min and late peak for sRB, with degraded fibres, low digesta viscosity and what appears to be the most rapidly disintegrated bolus of the rye products. However, the results are consistent with the lower insulin or lower insulin and glucose responses compared with refined wheat bread that have been repeatedly shown for soft rye breads, both sourdough-fermented and yeast-fermented [4, 6]. Although lactic acid produced during fermentation has been suggested to inhibit starch hydrolysis, with increased interaction between starch and gluten as the mechanism [46], the absence of a gluten network in sRB indicates other mechanisms. Rather, sRB was the only product where the presence of an amylose layer was observed, which has been suggested to inhibit starch hydrolysis in certain rye products [4]. The high content of resistant starch in sRB compared with the other products (Table 1) is likly related to the observed amylose layer (Fig. 3). To our knowledge, the factors contributing to the formation of an amylose layer in certain rye products are not known. However, organic acids produced during sourdough fermentation have been suggested to increase the content of resistant starch and may be a contributing factor [47]. Presence of an amylose layer may also be related to the high water content, compared with RCB, uRCB and extR, which could promote starch retrogradation during storage [48]. Whether the amylose layer is a result of sourdough fermentation or not warrants further investigation.