1. Introduction
The behaviour of almonds in the gastrointestinal tract (GIT) may explain why almonds have potential health benefits and reduce risk factors associated with type 2 diabetes, cardiovascular disease, cancer and obesity [
1,
2,
3]. Previous studies have established that almond cell walls play a crucial role in regulating nutrient bioaccessibility in the GIT [
4,
5]. The term ‘bioaccessibility’ is defined as the proportion of a nutrient or phytochemical compound ‘released’ from a complex food matrix during digestion and, therefore, potentially available for absorption in the GIT. Using an in vitro and an in vivo study, we have recently demonstrated that test meals containing almonds of different particle sizes behaved differently: the degree of lipid encapsulation affected the rate and extent of bioaccessibility in the upper GIT [
6]. We have also demonstrated that mastication of natural raw almonds released only a small proportion (7.9%) of the total lipid and was only slightly higher for roasted almonds (11.1%) [
7]. The lipid release from masticated almonds was in close agreement with that predicted by a theoretical model for almond lipid bioaccessibility [
7,
8]. Using an in vitro model of duodenal digestion [
9], it was observed that a decrease in almond particle size resulted in an increased rate and extent of lipolysis.
Novotny et al. [
10] conducted a feeding study in healthy adults to determine the energy value of almonds as a representative food from a group for which the Atwater factors may overestimate the energy value. They showed that only 76% of the energy contained within almonds (based on the Atwater factors) was actually metabolised [
10]. Furthermore, when calculating the metabolisable energy (ME) of whole natural almonds, whole roasted almonds, chopped almonds and almond butter, it was demonstrated that the number of calories absorbed was dependent on the form in which almonds were consumed.
Based on these findings, the aims of the present work were: (a) to investigate the mechanisms responsible for the loss in observed in vivo metabolisable energy compared to that calculated from nutrient composition using the Atwater general factors; this was performed by microscopy in post-GIT faecal samples; (b) to carry out microstructural investigations on freshly artificially “masticated” almond samples, to determine the particle size distribution of the “masticated” samples and the extent of lipid release after oral digestion; and (c) to further validate the mathematical model previously developed to predict lipid release from masticated almonds [
8].
2. Materials and Methods
2.1. Almond and Faecal Samples
Four almond types all with brown testa present (natural raw almonds, NA, roasted almonds, RA, roasted diced almonds, DA and almond butter, AB) were provided by the Almond Board of California. Smooth unsalted AB was industrially produced by grinding unskinned roasted almonds. It contained (as per label): fat (50%, of which 4.7% were saturated), total carbohydrates (25%, of which 12.5% were dietary fibre and 6.2% were sugars), protein (15.6%). Faecal samples were collected from humans who were participants in a study to measure the metabolisable energy of the almonds [
11]. This feeding study was a crossover, randomized control trial. Volunteers (10 men and 8 women) were fed the 5 distinct feeding regimes (control, NA, RA, DA and AB) as part of a highly controlled diet. During the feeding periods, all meals (using a 7-day menu cycle) for the volunteers were prepared at the Beltsville Human Nutrition Research Centre (the Centre) and Monday through Friday breakfast and dinner were consumed at the Centre under supervision of the research investigators. Lunch and weekend meals were prepared at the Centre and packaged for consumption off-site. Foods for all meals and snacks were identical (except for the form of nut). Food for all meals was prepared by weight, to the nearest 1 g, to produce daily menus providing a range in energy from 1600 kcal to 4000 kcal. Volunteers were fed the energy needed to maintain their body weight (body weight was measured each morning, Monday through Friday) and adjustments to the amount of food consumed was made by increasing or decreasing the amount of all food, proportionately, such that the composition of the diet was identical for all volunteers, independent of the energy they required to maintain body weight. A total of 42 g/day of each form of nut (NA, RA, DA and AB) was consumed daily with half the amount consumed at breakfast and the other half consumed at dinner. For the control diet, the amount of all foods was increased proportionately such that the energy content of the control diet was designed to be equal to the 4 four diets that contained the control foods plus the nuts. Volunteers were recruited from the area around the Centre and were screened to insure they met the study criteria. Briefly, subjects were healthy individuals (not taking any medications or supplements that might interfere with study outcomes) without dental or digestive conditions. At the beginning of the study, the mean (±SEM (Standard error of the mean)) age of the volunteers was 56.7 ± 2.4 year, their mean height was 170.2 ± 2.1 cm and mean weight 88.6 ± 5.6 kg.
From each of the 18 volunteers, faecal samples were collected from the beginning and end of each of 5 distinct feeding regimes (control, NA, RA, DA and AB) (each treatment period lasting 3 weeks). Following a 14-day adaptation to each of the 5 feeding regimes, the study subjects received blue dye capsules to mark the beginning of a one week excreta collection period and a second blue dye capsule to mark the end of the collection period.
The study protocol and informed consent form were reviewed and approved by the MedStar Health Research Institute and the associated faecal samples were registered with the QIB Human Research Governance Committee in December 2016.
2.2. Simulated Oral Digestion
The aim of this procedure was to simulate the chewing of the almond meals in the mouth. Chewing is the initial step in the digestion process and this procedure was designed to simulate both the salivary amylase activity and the mechanical breakdown of the food. Four almond samples, NA, RA, DA and AB (25 g), were minced 3 times using a mincer (Lexen mincer, Windermere, UK) to simulate the mechanical oral breakdown of the meal. Thereafter, 12.5 mL of Simulated Salivary Fluid (SSF) at pH 6.9 (0.15 M sodium chloride, 3 mM urea) and 900 U Human Salivary Amylase (HSA) dissolved in 1 mL SSF were added to the minced almonds or the almond butter [
12] and mixed. This process produced a paste of equal ratio of solid to water as calculated from human chewing [
13].
2.3. Particle Size Distribution (PSD)
The particle size of the samples before (AB) and after simulated oral digestion (NA, RA, DA and AB) was measured using mechanical sieving. Briefly, 22.5 g of each sample mixed with SSF was loaded on a stack of sieves with 9 aperture sizes: 3350, 2000, 1000, 850, 500, 250, 125, 63 and 32 µm (Endecott test sieve shaker, Endecotts Ltd., London, UK). The samples were washed with deionized water, shaken for 15 min and washed again, thus ensuring separation of the particles. The sieves were then dried in a forced-air oven at 56 °C for 6 h. The bases were left to dry at 100 °C overnight (about 15 h), which permitted complete evaporation of the water. The sieves were weighed before loading the sample and then again after having been oven dried. The dried fractions retained on each sieve and the base were expressed as a percentage of the weight of almonds before simulated oral processing.
2.4. Lipid Release after Oral Processing and Mathematical Model
Lipid extraction for total fat determination on all samples, before and after oral processing, was performed with a Soxhlet automatic Soxtec 2050 extraction (FOSS Analytical, Hilleroed, Denmark) using
n-hexane as a solvent [
5]. For AB, it was assumed that the continuous oil phase was bioaccessible and therefore the aim was to measure the additional lipid released from the almond particles present in AB. The almond particles were separated from the continuous oil phase of AB by centrifugation (REMI Elektrotechnik LTD., Vasai (East), India (13,000×
g, 15 min) before and after chewing and the lipid content of the particles was determined as described below. The fat present in the pellets (almond particles) was also determined: the pellet was washed 5 times with warm (37 °C) distilled water to remove any free fat released from the cells, then separated by centrifugation and quantified by
n-hexane extraction. The lipid present within the remaining wet pellet (theoretically inside the almond cells) was extracted by the Bligh & Dyer [
14] method. Briefly, almond particles were extracted in chloroform/methanol/water in the proportions 1:2:0.8 and the lipid was then quantified in the chloroform layer.
2.5. Results of Lipid Content Were Expressed as a Percentage of Dry Weight
Lipid release was predicted from the measured particle size distribution by sieving and the previously measured average cell diameter, 36 µm using the method of Grassby et al. [
8]. The threshold diameter below which 100% release would be achieved was 54 µm, particles below the threshold were not included in the calculations of lipid release. The spreadsheet provided as supplementary information was modified to accept particle size data from sieving alone and to account for the particles above the threshold diameter that were recovered on the 63 µm sieve (50 µm <
p < 100 µm).
2.6. Microstructural Analysis
Portions of each almond sample (NA, RA, DA and AB), before and after simulated oral digestion, were incubated in the chelating agent CDTA (1,2-Cyclohexylenedinitrilotetraacetic acid) (50 mM CDTA, pH = 7) at 4 °C for a minimum of 4 weeks [
15]. This treatment weakens the pectin layer in the middle lamella between cells so that individual cells can be separated from lumps of tissue by gentle pressure. These individual cells were then observed, unstained, by microscopy (bright-field or polarizing optics), or after staining with Sudan IV (0.1% Sudan IV in 1:1 acetone and 70% ethanol) to visualize their oil content. The CDTA treatment has previously been found to prevent microbial growth and to retain the oil fraction in the form it exists in the cells of the kernel, either as individual oil bodies in NA or as large coalesced oil globules in RA which are characteristic of the roasting process [
15]. Fresh sections of raw and roasted samples were not used because the sectioning process was found to release the oil from the damaged cells so that the spatial information was lost.
Initial observations were made on faecal samples from 3 randomly-chosen volunteers on material stored in CDTA [
15]. It was found that this method was not optimal as some of the free oil content rose to the surface and larger lumps of tissue sank in the CDTA making it difficult to obtain a representative sample. However, it was a useful preliminary step to help identify the range of plant structures that survived passage through the bowel. These included wheat bran layers (primarily aleurone, the cells of which are similar in size to almond tissue) and brush hairs, vascular tissue, xylem and tannin body inclusions.
Subsequent investigations on samples from volunteers with measured high faecal fat content were made by mixing a small sample of the frozen faecal matter directly on a microscope slide with the oil stain Sudan IV to minimise loss of components.
For microscopy, samples were examined and photographed using an Olympus BX60 (Olympus, Southend-on-Sea, UK) microscope and ProgRes® Capture Pro 2.1 software (Jenoptik, Jena, Germany).
2.7. Statistical Analysis
Results of lipid release from mastication were expressed as mean ± standard deviation (SD) of four independent experiments and analysed by one-way analysis of variance (ANOVA). The significance was assayed by using the Student-Newman-Keuls test using the SigmaPlot 12.0 software (Systat Software Inc., San Jose, CA, USA). Statistical significance was considered at p < 0.001.
4. Discussion
This work further validates the findings of loss in metabolisable energy compared to that calculated by the Atwater general factors [
10,
11]. Further, these results provide insights into the mechanisms by which the almond structure affects the release of lipid and energy. Given the structural diversity of the almond meals analysed in the present study, it is evident that the almond cellular structure plays a crucial role in determining lipid bioaccessibility in the gut. Using in vitro and in vivo techniques, we have recently demonstrated that almond lipid bioaccessibility is significantly affected by the particle size within a food matrix [
6]. The importance of nutrient encapsulation within intact cell walls has been previously studied using in vitro models of digestion [
5,
7]. This present study builds on our previous work [
5,
6,
7,
15] by measuring lipid release during simulated mastication from 4 almond meals with different structure and particle size, while simultaneously observing the microstructural changes. It is interesting to note that there was limited lipid digestibility for NA and RA during mastication. In AB, all the intracellular lipids made available by cell-wall rupture, as well as the lipid molecules present at the interface and within the continuous lipid phase are readily available for absorption. The exposure of the remaining intact almond particles to mastication resulted in a further small release of lipid. The coalescence of lipid in RA and DA baseline samples did not limit the digestibility in the mouth, with very similar values of lipid release obtained from the two matrices, slightly above NA. It is worth noting that although roasting did not appear to cause any damage to the cell walls prior to mastication, it could affect the structure of proteins surrounding the coalesced oil bodies typical of roasted samples. If the hydrophilic component of the lipid body stabilizing proteins is reduced or hydrolysed following roasting, the protein could become more lipophilic and thus structurally better suited to stabilising the reversed curvature of a water-in-oil emulsion. However, the stability is likely to be weak and could be compromised through extended storage.
This work further validates our mathematical model of lipid release, showing considerable potential for predicting nutrient bioaccessibility from plant foods which satisfy the model’s assumptions (broadly spherical cells, cell fracture the main mode of failure) [
8]. The model was also shown to work without having to combine PSDs from laser diffraction and sieving. This simplifies the experimental work and calculations required and suggests that the model could be used successfully with any method that covers the whole particle size range (see supplementary file).
With the exception of AB, particle size of almonds decreased with mastication and the distributions obtained with NA and RA samples were comparable with our previous investigation [
15]. It is known that processing of nuts, such as roasting, chopping and grinding, impacts mastication, particle size and lipid bioaccessibility [
7,
18,
19]. The decrease in size of almond particles, with consequent reduction of intact cell walls, determines the rate and extent of lipid bioaccessibility during digestion. Cassady et al. [
20] reported important differences in appetitive and hormone responses after mastication of almonds. Gebauer et al. [
11] have recently reported that the number of calories absorbed from almonds in the GIT is strictly dependent on the form in which they are consumed: Atwater factors overestimate the ME of natural raw, roasted and chopped almonds. Overestimation of the ME of nuts could explain data from epidemiological and clinical studies indicating lower body weight of individuals consuming nuts [
21,
22], who are also less likely to gain weight over time [
23,
24]. As a result of incomplete macronutrient loss in the upper GIT, it is believed that a large proportion of nutrients from almonds reaches the large bowel, where it is fermented by the microbiota [
4,
5]. Incomplete rupturing of the cell walls during mastication results in macronutrient encapsulation, which remain inaccessible to digestive enzymes and, if not fermented in the colon, are excreted in faeces. Here, we have further validated the ME results of Novotny et al. [
10] and Gebauer et al. [
11]. Our microstructural evidence showed recognisable multicellular particles of almond tissue in the same size range as that seen in the chewed sample of the volunteer consuming NA, whereas small multicellular particles of almond tissue were present in the faecal sample of the volunteer consuming AB. Our micrographs provide further evidence of increased energy absorption when cells are broken down during grinding (AB), thus exposing lipid to lipolytic enzymes during digestion. The extensive cell-wall disruption and small particle size of almond butter played a crucial role in the extent of lipid absorption in the gut.
While a relatively small percentage of lipid was released from the almond tissue during mastication, we believe the additional lipid released after mastication [
10] is due to a number of mechanisms, including the breakdown of small particles, lipid becoming accessible within intact cells followed by erosion of the particles, or microbial degradation.
In conclusion, we report evidence on the low digestibility of lipids from almonds, which is strictly related to the structure of the meal and the particle size distribution. We believe that food structure influences health through nutrient bioaccessibility in the gut. A greater understanding of the relationship between food structure and nutrient bioaccessibility may be useful to improving bioaccessibility of nutrients that may be poorly available, especially in plant-based diets.