The pattern of peptides released from dairy and egg proteins is highly dependent on the simulated digestion scenario

Dairy and egg proteins either isolated or within the food matrix were subjected to different staticin vitrodigestion models (infant, fed and fasted adult). Proteolysis differed across models and regarding the effect of the matrix/processing.


40
Evaluating the gastrointestinal (GI) fate of proteins is paramount to assess whether they are safe to 41 consume, including their potential to elicit an allergic reaction. The resistance of proteins to 42 digestion may be significant in relation to determining their allergenic potential since incomplete 43 digestion may cause undesired immune responses via sensitisation and/or elicitation in the 44 duodenum. 1 Thus, evaluating the digestibility of proteins as part of a suite of assessments and in 45 vitro protocols seems appropriate when ethical constraints hinder in vivo studies. The in vitro 46 digestion model most commonly used for this purpose in line with the European Food Safety 47 Authority (EFSA) guidelines 2, 3 and Implementing Regulation (EU) No 503/2013 (IR503/2013) is the 48 pepsin resistance test. 4,5 This mimics the gastric phase of digestion with parameters that are not 49 representative of the normal physiological environment that food is exposed to immediately after 50 consumption. Namely, it uses gastric conditions that are highly acidic and enzyme concentration that 51 would simulate the end of gastric emptying (late phase) or fasted state in human adults. A more 52 realistic approach including a subsequent small intestinal phase and other relevant conditions in 53 healthy adults or in infants may provide useful information on how the combined effect of pH and 54 enzyme concentration affects protein digestibility. Therefore, in accordance with the guidance 55 offered by the EFSA Genetically Modified Organisms (GMO) panel 3 this study considers a range of 56 more physiologically relevant conditions that mimic the early phase (or fed state) and late phase (or 57 fasted state) adult and infant GI environments.

58
The aim of this research is to compare the three in vitro digestion models in the context of 59 assessment of the digestibility of proteins. For this purpose, the comparison uses relevant test 60 proteins that are widely consumed and have previously been characterised to some extent under 61 simulated GI conditions. In addition, the effect of the food matrix, in which these proteins are 62 naturally present, on their digestibility will be also evaluated with the three models. Thus, the first 63 part of this study comprises the in vitro digestion of isolated proteins from bovine milk (β-64 lactoglobulin, BLG, and β-casein, BCS) and hen's egg (lysozyme, LYS, and ovalbumin, OVA). These 65 proteins are all major allergens varying in stability within the GI tract. BLG and BCS represent two 66 major proteins in bovine milk (allergens Bos d 5 and Bos d 8, respectively) with contrasting 67 susceptibility to GI digestion, stable versus labile. 6 LYS and OVA (allergens Gal d 4 and Gal d 2, 68 respectively) are major globular proteins in egg white that are moderately resistant to GI digestion. 6-69 8 LYS is an enzyme whereas OVA, the most abundant protein in egg white, is a storage protein.

70
The second part of this study focuses on the in vitro digestion of fresh whole bovine milk and soft-71 boiled hen's egg. Both the food matrix and thermal processing effects are taken into account. 72 Pasteurisation is a standard procedure to ensure microbial stability in milk, and eggs are cooked for 73 safe consumption and improved sensory properties. Thermal processing may affect the structure of 74 the natural food matrix, since proteins are partially denatured and conformations are modified, 75 affecting the stability to digestion. 9 In addition, thermal processing may enhance protein interactions 76 with other components in the food matrix, such as lipids and sugars and can affect the digestion of 77 proteins. 10 In the current study, the commercial fresh whole milk used was previously pasteurised 78 (72 °C for a minimum of 15 s) and homogenised. Homogenisation in milk prior to pasteurisation is 79 also known to affect the ultrastructure by breaking up milk fat globules and changing the interfacial 80 composition from milk fat globule membrane (MFGM) proteins to whey proteins and casein micelles 81 and/or their fragments. 11 The hen's eggs were soft boiled (boiled individually for 3 min 82 approximately), as it is very popular in the British breakfast.

83
The in vitro digestion of milk and egg has been investigated in the literature, although some gaps can 84 be identified. The digestibility of milk protein has been determined as a function of heating and 85 homogenisation processes used commercially by the dairy industry, 12 and even using more 86 sophisticated in vitro semi-dynamic/dynamic gastric models. 13,14 Nevertheless,no combined 87 comparison has been made between the digestibility of proteins in milk and the digestibility of 88 isolated proteins, including other human GI conditions. Menard et al. compared the proteolysis 89 kinetics of an infant formula with the infant and early phase adult models mentioned above. 15 90 However, no direct comparison with the digestion of isolated proteins was performed. Similarly, 91 Egger et al. recently compared the in vitro static (INFOGEST standardised protocol, i.e. early phase 92 adult) and dynamic digestion of skimmed milk powder proteins with in vivo data. 16 Their results 93 showed a good agreement between the gastric and intestinal end points of both in vitro models and 94 in vivo data from pigs. Martos et al. assessed the effect of the whole food matrix on egg protein 95 stability to digestion and compared the results with their previous findings on isolated egg 96 proteins. 17 Nonetheless, they only considered a single adult model of in vitro digestion and the egg 97 was not thermally processed (cooked). The effect of heat treatment on the digestion of egg proteins 98 has been studied on isolated proteins (65 °C for 30 min and 90 °C for 15 min; 80 °C for 6 h), 18-20 egg 99 white (56 and 65 °C for 30 min and 100 °C for 5 min; 60 and 80 °C for 10 min) 21, 22 and in liquid whole 100 egg (pasteurisation at 60-66 °C for 4-10 min) 23 but again only a model of adult digestion was used.

101
The relationship between allergenicity and stability to digestion of a protein is still controversial due 102 to a lack of knowledge on the exact route of exposure and mechanisms behind food sensitisation 103 and food allergy. 1, 24 For this reason, it is of paramount importance to further investigate the 104 behaviour of known allergens during digestion and identify the products of digestion extensively. In 105 the present study, SDS-PAGE was used to identify intact protein or protein fragments larger than 5 106 kDa in digesta samples of isolated proteins and meals. Densitometry analysis of SDS-PAGE allowed 107 the semi-quantification of hydrolysis of intact protein throughout in vitro GI digestion. LC-MS/MS 108 was used to identify protein fragments smaller than 5 kDa in digesta samples of isolated proteins. 109 Only peptides greater than 9 amino acids in length were analysed because of their potential to 110 induce an immune response due to their likelihood of carrying at least two B-cell receptor 111 epitopes. 25, 26 Hydrolysis of total protein in meals was also quantified by measuring the levels of free 112 amine groups with a spectrophotometric assay.

113
To the best of our knowledge, this is the first time that the in vitro digestion of these meals has been 114 compared under physiologically relevant conditions in infants and adults in two different states: fed 115 versus fasted, and the impact of food matrix/processing assessed by comparison with the 116 digestibility of isolated proteins. The current study highlights differences in protein digestibility 117 across the three GI scenarios whether on isolated proteins or within food matrices. In addition, 118 peptides greater than 9 amino acids were present throughout the intestinal phase for all proteins, 119 regardless of the model. This emphasises the difficulty of linking digestibility to potential 120 allergenicity because many other factors are involved that need further investigation. sequence. 15 The only adaptation made in the current study was the replacement of bovine bile 171 extract by either porcine bile extract in the digestion of meals, or an equimolar mixture of NaGC and 172 NaGCDC, which represent the two major forms in human bile, 29 in the digestion of isolated proteins.

173
Briefly, in the gastric phase, 5 mL of isolated protein (5 mg/mL) or 5 g of meal were mixed with infant 174 simulated gastric fluid (SGF) at a ratio protein solution or meal to SGF of 63:37 (v/v). The pH was set 175 to 5.3. The infant SGF comprised NaCl (94 mM) and KCl (13 mM), adjusted to pH 5.3 with 1 M HCl. 176 Pepsin activity was 268 U/mL in the final volume of the gastric chyme. After 60 min of gastric 177 digestion, the pH was raised to 7 with 1 M NaOH in order to inactivate pepsin before intestinal 178 digestion.

179
In the intestinal phase, the gastric chyme was mixed with infant simulated intestinal fluid (SIF) at a 180 ratio of gastric chyme to SIF of 62:38 (v/v) and adjusted to pH 6.6 with 1 M HCl. The infant SIF 181 comprised NaCl (164 mM), KCl (10 mM) and NaHCO 3 (85 mM) adjusted to pH 7. CaCl 2 was added 182 separately before starting the intestinal phase at a concentration of 3 mM within the volume of the 183 SIF. The total concentration of bile salts was 3.1 mM in the final volume of the intestinal content. 184 The trypsin activity was 16 U/mL (also in pancreatin) in the final volume and the chymotrypsin 185 activity was 4 U/mL. This phase lasted for 60 min. Briefly, in the gastric phase, 5 mL of isolated protein (5 mg/mL) or 5 g of meal were mixed with early 193 phase adult SGF at a ratio protein solution or meal to SGF of 50:50 (v/v) and the pH was set to 3. The 194 early phase adult SGF comprised NaCl (47.2 mM), KCl (6.9 mM), KH 2 PO 4 (0.9 mM), NaHCO 3 (25 mM), 195 MgCl 2 (H 2 O) 6 (0.1 mM), and (NH 4 ) 2 CO 3 (0.5 mM) adjusted to pH 3 with 1 M HCl. CaCl 2 was added 196 separately before starting the gastric phase at a concentration of 0.075 mM in the final volume of 197 the gastric chyme. Pepsin activity was 2000 U/mL in the final volume. After 60 min of gastric 198 digestion, the gastric chyme was immediately subjected to the intestinal phase.

199
In the intestinal phase, the gastric chyme was mixed with early phase adult SIF at a ratio gastric 200 chyme to SIF of 50:50 (v/v) and adjusted to pH 7 with 1 M NaOH. The early phase adult SIF 201 comprised NaCl (38.4 mM), KCl (6.8 mM), KH 2 PO 4 (0.8 mM), NaHCO 3 (85 mM), and MgCl 2 (H 2 O) 6 (0.33 202 mM), adjusted to pH 7. CaCl 2 was added separately before starting the intestinal phase at a 203 concentration of 0.3 mM in the final volume. The total concentration of bile salts was 10 mM in the 204 final volume. The trypsin activity was 100 U/mL (also in pancreatin) in the final volume and the 205 chymotrypsin activity was 25 U/mL. This phase lasted for 60 min. was added to all gastric samples. The gastric samples were diluted to the required protein 240 concentration for injection into the spectrometer (37 ng of BLG or BCS and 50 ng of LYS or OVA). All 241 gastric samples were filtered using a 0.45 μm filter before injection of 10 μL. For the intestinal 242 samples, 10 μL were injected, corresponding to 120 ng of protein (unfiltered) for the infant and early 243 phase adult models and 6 ng of protein (filtered) for the late phase adult model. The smaller amount 244 injected for the latter is a limitation from the highly diluted samples of the late phase adult protocol.

Food & Function Accepted Manuscript
Published on 20 May 2020. Downloaded on 5/25/2020 1:59:42 AM. were serine phosphorylation, methionine oxidation, and deamidation of glutamine or aspartic acid. 264 Peptides identified with an e-value < 0.01 were automatically validated, giving an evaluated false 265 discovery rate of less than 1% at the peptide level. Only peptides of minimum 6 amino acids long can 266 be identified with this strategy.

267
Data analyses were performed using the R software, version 3.3.1 (R Core Team, 2014). A statistical 268 analysis of the identified peptides longer than 9 amino acids was performed. Peptides of molecular 269 weight (Mw) higher than 4 kDa were not detected automatically by the technique.
The standard ortho-phthaldialdehyde (OPA) spectrophotometric assay 31 was performed to quantify 272 the amount of free NH 2 groups released during the proteolysis of both meals with the three in vitro 273 models. This is indicative of the hydrolysis of total protein. Prior to the assay, 5% trichloroacetic acid 274 (166 µL) was added to digested sample (100 µL) to cause the precipitation of insoluble protein that 275 could interfere in the analysis, followed by centrifugation at 10,000 g for 30 min at room 276 temperature. OPA reagent was prepared by dissolving 3.81 g of sodium tetraborate in approximately 277 80 mL Milli-Q water under stirring at 50 °C. Then, 0.088 g dithiothreitol and 0.1 g sodium dodecyl 278 sulphate were added after cooling down to room temperature. Finally, 0.080 g OPA dissolved in 2 279 mL of ethanol was added in the solution that was made up to 100 mL with Milli-Q water. L-leucine 280 was used as standard. The calibration curve was obtained with different concentrations (0-10 mM) 281 of the standard solution made in 10 mM phosphate buffer solution. In micro-titre plates, 10 µl of 282 standard/sample were loaded into each well and mixed with 200 µl of OPA reagent, allowing the 283 reaction to proceed for 15 min at room temperature. The absorbance was measured at 340 nm 284 using a microplate photometer (Multiskan FC, ThermoFisher Scientific). Each measurement was 285 conducted in triplicate. Data are presented as mean values ± standard deviation. Comparison 286 between in vitro digestion models over time was done with two-way ANOVA and post hoc 287 Bonferroni multiple comparison test with a threshold for significance p ≤ 0.05.

Results and discussion
290 Three in vitro digestion protocols have been applied to isolated proteins from bovine milk and hen's 291 egg as well as the respective meal. The infant model follows the protocol recently published by 292 Menard and co-workers, 15 which is based on in vivo data available in literature, and represents the 293 mildest digestive conditions tested. In particular, the infant gastric average pH (5.3) is higher than 294 that in adults (3 in the early phase and 1.2 in the late phase) and is out of the optimum range for 295 pepsin activity (pH = 1.6-4). Furthermore, the average enzyme activity in the gastric and intestinal 296 compartment is also lower in the infant model, as is the total concentration of bile acids in the small 297 intestine. The early phase adult simulation is based on the INFOGEST harmonised protocol from 298 Minekus and co-workers 28 that has been validated against in vivo data and its reproducibility has 299 been confirmed by ring trial. 32 The late phase adult gastric model follows the current pepsin 300 resistance test, 5 which uses a low pH and high pepsin activity compared to the early phase adult 301 model. This is followed by the intestinal phase of the INFOGEST harmonised protocol.

303
The digestion of isolated milk proteins, BLG and BCS, was first evaluated with SDS-PAGE and the 304 results for the three in vitro models are shown in Figure 1A and 1B. The lanes "BLG" and "BCS"

311
There are clear differences in the gastric proteolysis of BLG and BCS across the three in vitro models. 312 Namely, the whey protein BLG is pepsin-resistant throughout the gastric phase (Figure 1 A1, A3, A5), 313 whereas BCS is rapidly hydrolysed (Figure 1 B1, B3 Conversely, the band corresponding to intact BCS (23.8 kDa) disappears at different rates depending 318 on the model and is no longer visible after 60 min of infant gastric digestion. This is more accurately 319 shown in the percentage of intact protein determined from densitometry analysis on bands ( Figure  320 1C and 1E For this reason, the intensity of the band corresponding to pepsin (34.6 kDa) is greater in the late 328 phase adult model than in the early phase adult model and essentially invisible in the infant model. 329 Therefore, the limited proteolysis of BCS was observed for the infant protocol, followed by 330 intermediate and extensive hydrolysis for the early and late phase adult models, respectively. In the 331 infant model (Figure 1 B1), BCS was partially hydrolysed by pepsin after 30 s and bands of lower Mw, 332 4-22 kDa peptides, corresponding to hydrolysis products, are already visible and become more 333 intense at 10 min. Their intensity gradually decreases afterwards, although these peptides are still 334 visible at 60 min of the gastric phase. In the early phase adult model (Figure 1 B3), only bands 335 corresponding to protein fragments smaller than 20 kDa are visible after 30 s of gastric digestion and 336 peptides smaller than 5 kDa are detected afterwards, which gradually decrease in intensity until the 337 end of the gastric phase. Figure 1E also shows a faster disappearance (at 30 s) of intact BCS during 338 the gastric phase for the early phase adult model, as compared to the infant one. As discussed by 339 Menard and co-workers, 15 the slower kinetics of BCS gastric digestion in the infant model is largely 340 due to the loss of pepsin activity at pH 5.3, ~ 10% of its activity at pH 2, 34 as compared to the optimal 341 pepsin activity at pH 3 in the early phase adult model, rather than the content of pepsin. Indeed, a 342 previous study reported similar BCS gastric digestibility for an infant and adult model working with 343 similar differences in pepsin content (8-fold lower in the infant) but closer pH values (2.5 and 3 for 344 adult and infant, respectively). 6 The gastric proteolysis of BCS is even faster for the late phase adult 345 model and only a faint band corresponding to protein fragments of 5 kDa approximately is observed 346 after 30 s of gastric digestion, and this is not visually detectable after 5 min (Figure 1 B5). This 347 greater extent of BCS hydrolysis in terms of peptides is not visible in the percentage of intact BCS in 348 the gastric phase of both adult models ( Figure 1E).

349
Page 8  Absence of letters means no significant differences.

359
A different story is observed in the intestinal phase, where BLG was significantly hydrolysed even 360 under the milder conditions of the infant model (Figure 1 A2). This agrees with the rapid intestinal 361 digestion of BLG present in infant formula. 15 A very faint band of ~ 3 kDa appears after 30 s of 362 intestinal digestion, which gradually vanishes afterwards suggesting almost complete hydrolysis by 363 intestinal proteases (see also Figure 1D). BCS seems to be fully digested after 30 s of intestinal 364 digestion ( Figure 1 B2 and F) although it was already largely hydrolysed by the end of the gastric 365 phase. The rest of the bands present in the intestinal phase correspond to the enzymes in the 366 pancreatin extract. In the early phase adult model (Figure 1 A4, B4), the SDS-PAGE suggests that 367 both proteins seem to be completely digested after 30 s of intestinal digestion (see also Figure 1D  368 and 1F). A faster duodenal digestion is expected since the intestinal enzymes to test protein ratio is 369 also larger in the early phase adult model (68.8 trypsin U/mg test protein) as compared to the infant 370 model (6.75 trypsin U/mg test protein). This is reflected in the greater intensity of the bands 371 corresponding to pancreatin enzymes. Dupont and co-workers reported much slower kinetics and 372 lower extent of hydrolysis of BLG with an infant and adult intestinal digestion as compared to our 373 results. 6 This could be partially explained by the use of phospholipids vesicles in their study, which is 374 known to protect BLG against pancreatic proteases degradation. 35 More importantly, the 375 trypsin/chymotrypsin to test protein ratio used in the present protocols are higher. Regarding the 376 intestinal phase in the late phase adult model, Takagi and co-workers also reported the rapid 377 digestion of BLG under intestinal conditions that would correspond to the late phase adult model in 378 the current study, although without previous gastric digestion. 36 The bands shown in the intestinal 379 phase of the late phase adult model correspond to trypsin and chymotrypsin (Figure 1 A6, B6). These 380 individual enzymes replaced the pancreatin extract in the remaining digestion experiments on 381 isolated proteins because the greater intensity of the bands corresponding to the complex mixture 382 of enzymes in pancreatin makes difficult the interpretation of the SDS-PAGE results.

383
The size of the peptides successfully identified by LC-MS/MS in at least two of the three replicates 384 from BLG and BCS digestion at each time point is presented in Figure 2. The number of peptide 385 sequences identified in total for each model of digestion is summarised in Table 1. The data in the 386 box plots indicate the peptides in the gastric phase from BLG ( Figure 2a) were much smaller (< 3 387 kDa) than those observed from BCS (≤ 4 kDa) (Figure 2b). Despite the fact that SDS-PAGE showed 388 that BLG was largely unaffected by pepsin, regardless of the digestion model ( Figure 1A and 1C), the 389 LC-MS/MS data demonstrate that a relatively large number of different peptides (72 in total from 390 the three models) were still detected (Table 1) but were most likely in low abundance. Only a small 391 proportion of BLG was hydrolysed by pepsin during the gastric phase, showing no difference in the 392 band intensity of the intact protein on SDS-PAGE gels for the infant and early phase adult models. 393 However, approximately 20% of intact BLG was hydrolysed in the late phase adult model as seen in 394 Figure 1C and a larger number of gastric peptides (53) was also identified in this model, compared to 395 those in the early phase (42) and infant model (32)  whereas the highest Mw observed with the early phase model was around 2.1 kDa. In the intestinal 400 phase, BLG generated peptides with Mw tending to be higher than in the gastric phase. This can be 401 explained by a much larger proportion of the protein being hydrolysed ( Figure 1A and 1D) leading to 402 more diversity in the peptides released. With the infant and early phase adult models, peptide Mw 403 tended to decrease over digestion time whereas the opposite tendency was seen with the late phase 404 adult model. The low number of peptides identified in the late phase adult intestinal samples (Table  405 1) could be the result of the smaller amount injected compared to the infant and early phase adult 406 models. In summary, a total of 72 unique peptides were identified in the gastric phase after in vitro 407 digestion of BLG with the three models and 138 in the intestinal phase (Table 1). The larger number 408 of peptides identified in the intestinal phase as compared to the gastric phase positively correlates 409 with a greater extent of proteolysis. In contrast to BLG, a total of 472 peptides were unambiguously identified for BCS in the gastric 417 phase, whereas only 296 were found in the intestinal phase (Table 1). The fact that BCS generated a larger number of unique peptides in the gastric phase is certainly related to the greater extent of 419 hydrolysis by pepsin ( Figure 1B and 1E). In the gastric phase, the median values of the peptide Mw 420 observed with the infant model are higher than those in the early phase adult model, which in turn 421 are higher than those in the late phase adult model (Figure 2b). This confirms a more intense 422 proteolysis in the adult models and in particular with the late phase. There is a tendency of the 423 median Mw value to decrease over time regardless of the in vitro digestion model. In the intestinal 424 phase, median Mw values were much lower than those for the gastric phase, indicating further 425 extent of hydrolysis into smaller peptides and in agreement with SDS-PAGE results in terms of 426 protein fragments of Mw smaller than 5 kDa ( Figure 1B). Among the three protocols, the infant 427 model led to peptides with highest Mw as expected. With the infant and early phase adult model, 428 median Mw value decreased consistently over time, which was not the case with the late phase 429 adult model.

430
In general, the total number of unique peptides identified in the gastric phase increases in the order 431 infant < early phase ≤ late phase for BLG and BCS and the opposite is observed in the intestinal 432 phase (Table 1). There may be a positive correlation between the number of unique peptides 433 identified in every digestion product and their in vitro digestibility in the gastric compartment.

440
The proteolysis of isolated egg proteins, LYS and OVA, were first analysed by SDS-PAGE and the 441 results are shown in Figure 3A and 3B for the three models of in vitro digestion. Figure 3C-F displays 442 the calculated percentage of intact protein from densitometry on bands. Figure 3A shows the LYS 443 band with a Mw of around 14.3 kDa that seems unaffected by pepsin throughout the gastric phase 444 in the infant and early phase adult model (Figure 3 A1, A3), which is also reflected in the 445 densitometry analysis in Figure 3C. There is only a statistically significant decrease of the intensity of 446 the band of intact LYS at 20 min of the gastric phase in the late phase adult model ( Figure 3C) and a 447 faint band of much lower Mw, 2-3 kDa, appeared (Figure 3 A5) corresponding to hydrolysis products. 448 The band of intact LYS had completely disappeared after 60 min of the gastric phase in this model. 449 The resistance of LYS to pepsin digestion agrees well with the results reported by Fu and co-workers 450 under conditions of the gastric late phase adult model or pepsin resistance test. 7 However, intact LYS 451 could be seen in their SDS-PAGE until 60 min of the gastric phase, which contrasts with our results, 452 where intact LYS could be seen until 30 min (Figure 3 A5). This difference might be caused by 453 differences in pepsin activity as Fu et al. added the enzyme by weight rather than activity making it 454 difficult to compare.

455
In contrast to LYS, OVA was more susceptible to pepsin digestion, at least under conditions of the 456 early and late phase adult models (Figure 3 B3, B5). The intensity of the band corresponding to intact 457 OVA (Mw ~ 45 kDa) remains constant throughout the infant gastric digestion (Figure 3 B1 and E). In 458 the early phase adult model, the proteolysis starts to be statistically significant after 20 min of the 459 gastric phase as shown by the decrease of band intensity ( Figure 3E) and appearance of protein 460 fragments of slightly lower Mw and much smaller hydrolysis products of Mw of 3-4 kDa (Figure 3  461 B3). A larger extent of OVA gastric digestion was found with an infant model by  workers, 6 which is likely due to the lower gastric pH (pH 3) used in their study that is more optimal 463 for pepsin activity, 70% of the maximum in comparison to 10% at pH 5.3. 34 This also demonstrates 464 the importance of setting a relevant pH because it affects the enzymatic activity. The rate of OVA 465 gastric proteolysis in an adult model was also slightly faster in their study, but the gastric extent is 466 very similar to that obtained here (ca. 20%). In the late phase adult model (Figure 3 B5), the 467 proteolysis occurs at earlier times, 5 min, as seen in Figure 3E, although intact OVA could still be 468 seen at 30 min. This is in complete agreement with the results of pepsin resistance test obtained in 469 the majority of laboratories (6 out of 9) in a ring trial. 5 Furthermore, the densitometry profile shown 470 in Figure 3E  LYS was mostly digested in the intestinal phase ( Figure 3D). This suggests that LYS is more 491 susceptible to pancreatic enzymes while OVA is relatively resistant, which confirms previous findings 492 on OVA intestinal stability without previous gastric phase. 36 Both large protein fragments (36-45 493 kDa) and low molecular weight hydrolysis products (3 kDa) appeared across the intestinal phase of 494 the early phase adult model (Figure 3 B4). Interestingly, the final extent of digestibility is similar for 495 both proteins at the end of the intestinal phase within each model ( Figure 3D and 3F) gastric phase and 20 unique peptides in the intestinal phase in total across the three models (Table  512 1). These low numbers of peptides are in accordance with the high resistance of LYS to GI digestion 513 seen by SDS-PAGE with the infant and early phase adult model, and confirm previously published 514 results. 8 It was shown in that study that the resistance of LYS to pepsin digestion is due to its rigid 515 structure given by the four disulphide bridges and proteolysis only occurred at a highly acidic pH 516 range from 1.2 to 2, likely because of the slightly increased flexibility of LYS at this pH. 38 It is 517 therefore reasonable to identify more peptides in the gastric compartment with the late phase adult 518 model (16) than with the other two models (4 in the infant model and 3 in the early phase adult 519 model) (Table 1). In addition, it has been shown that when cleaved by pepsin, LYS leads to the 520 formation of peptides of Mw less than 4-5 kDa, 8 which are not readily detectable by the LC-MS/MS 521 protocol used in the current study ( Figure 4a). Indeed, SDS-PAGE showed the appearance of 522 hydrolysis products of around 3.5 kDa at 20 min of the gastric late phase adult model (Figure 3 A5). 523 In the intestinal phase, LYS has been shown to precipitate in the presence of bile salts making it quite 524 resistant to proteolysis. 8 In any case, since the number of peptides coming from LYS clearly identified 525 is rather low (Table 1), the statistical analysis of the dataset does not bring much relevant 526 information.
Page 15 of 28 Food & Function

Food & Function Accepted Manuscript
Published on 20 May 2020. Downloaded on 5/25/2020 1:59:42 AM. between the two adult models. This gastric behaviour follows a similar trend to that for BLG and BCS 536 (Figure 2), confirming a more intense proteolysis in the adult models. The median Mw values of the 537 peptides decreased on average slightly with time, regardless of the model. Peptides from OVA of 538 much larger Mw were identified in the intestinal phase, which may be attributable to a larger 539 proportion of the protein being hydrolysed by intestinal enzymes ( Figure 3B). This leads to more 540 diversity in the peptides released and resembles the behaviour found for BLG in Figure 2a, where 541 larger peptides were indentified in the intestinal phase compared to the gastric phase. In this case, 542 peptides of higher Mw were seen in the infant model, followed by the early phase and late phase 543 adult models, which may be explained by the lower extent of proteolysis in the infant model ( Figure  544 3B and 3F). Only in the infant and late phase adult models was evolution of peptide size, median 545 Mw, with time observed. In summary, a total of 91 unique peptides were identified in the gastric 546 phase after in vitro digestion of OVA and 434 in the intestinal phase across the three models (Table  547 1). The larger number of peptides identified in the intestinal phase as compared to the gastric phase 548 positively correlates with a greater extent of proteolysis.

549
As a general trend and as for isolated milk proteins, the number of unique peptides identified in the 550 gastric phase increases in the order infant ≤ early phase ≤ late phase for LYS and OVA and the 551 opposite is observed in the intestinal phase (Table 1). Therefore, this confirms a positive correlation 552 between the number of unique peptides identified in the gastric compartment and the in vitro 553 digestibility. This is inconclusive in the intestinal stage. 554 555

556
The in vitro digestion of proteins in bovine milk with the three models was first monitored with SDS-557 PAGE and results are shown in Figure 5A, with particular focus on BLG and BCS. In this case, the lane 558 labelled as "Milk" corresponds to the meal blank before digestion, showing the band corresponding 559 to intact BCS within the group of bands of caseins, and BLG band at lower Mw. The whey protein α-560 lactalbumin (allergen Bos d 4) of around 14.2 kDa can also be seen below the BLG band, as well as 561 MFGM proteins at higher molecular weight (> 55 kDa) and bovine serum albumin (allergen Bos d 6) 562 (66.5 kDa). 14 The control lane "C" in the intestinal phase corresponds to the digestive enzymes 563 present in pancreatin. This complex pancreatic mixture represents a more realistic environment and 564 was used to digest in vitro the food matrix (milk and eggs). Despite the difficulty of reading SDS-565 PAGE at relatively large concentrations of pancreatin, i.e. in the late phase adult model, some useful 566 information can still be inferred from the comparison with the infant and early phase adult models. 567 The gastric digestibility profile of BLG and BCS in the whole milk matrix ( Figure 5B and 5D) follows a 568 similar trend for the three in vitro models as compared to the isolated proteins. The results from 569 isolated proteins (lines in Figure 5B-E) are also included as a reference. Namely, BLG resisted 570 hydrolysis throughout the gastric phase regardless of the model, whereas BCS was susceptible to 571 pepsin digestion and its hydrolysis was instantaneous upon starting the gastric phase in the late 572 phase adult model, whereas it was slower in the early phase adult and infant models. Although the 573 impact of the food matrix on protein digestion is expected to be less relevant in milk because it is a 574 liquid, the effect of thermal processing and homogenisation needs to be considered as well. It has 575 been reported that the temperature of the pasteurisation process (72 °C) was not sufficient to cause 576 any important changes in the in vitro gastric digestion of milk proteins as compared to raw milk 577 (non-heated). 14 The latter study was carried out with a semi-dynamic gastric model and comparing 578 milk matrices, not with isolated proteins. Nevertheless, a fast comparison in Figure 5D reveals slight 579 but significant differences, at least in the dynamics of the early phase adult model. BCS, as part of 580 the milk matrix, displays a slightly slower rate of gastric proteolysis with the infant and early phase 581 adult models as compared to the isolated BCS in water, but the final extent is similar. It is known 582 that heating above 70 °C induces the denaturation of whey proteins, and that denatured whey 583 proteins bind to κ-casein, both at casein micelle surface and in serum phase. 39 This complexation 584 with the surface of casein micelles may exert a protective effect and slightly delay the hydrolysis of 585 BCS as compared to isolated non-heated BCS in aqueous solution. Tunick and co-workers showed a 586 slightly higher resistance of BCS to pepsinolysis in homogenised pasteurised whole milk as compared 587 to raw milk during the first 15 min of gastric in vitro digestion, as seen by SDS-PAGE. 12 Sanchez-588 Rivera et al. reported a noticeable increased resistance of the casein fraction to pepsin digestion in 589 heated skimmed milk proteins as compare to unheated sample in dynamic in vitro gastric 590 digestion. 40 However, it must be taken into account the higher heating temperature used in the 591 latter study (90 °C), the non-fat nature of the milk, and the dynamic model of the gastric phase. One 592 could also argue that the enzyme/substrate ratio is lower in the in vitro digestion of meals than for 593 isolated proteins due to the larger protein content. For instance, the concentration of total protein 594 in milk is 35 mg/mL approximately as compared to 5 mg/mL in isolated protein samples. 595 Nevertheless, the enzyme/substrate ratio should still be high enough to overcome any inhibitory 596 effect.

597
On the other hand, a rapid BLG hydrolysis was observed in the intestinal phase of milk for both adult 598 models as for isolated BLG ( Figure 5C). However, BLG in the milk matrix exhibits lower extent of 599 intestinal digestion with the infant model as compared to the isolated BLG in water. The most 600 plausible explanation is that the average gastric pH in the infant model ( due to the buffering capacity of the meal. It has been shown that the kinetics of digestion of milk 608 proteins varies according to the in vitro digestion model applied: static versus semi-dynamic. 14 The 609 semi-dynamic model was designed to replicate some realistic behaviour found in the stomach in 610 vivo. 42 The semi-dynamic model considers not only the buffering capacity of the meal, which 611 increases the gastric pH immediately after meal intake, but also the gradual acidification with the 612 progressive secretion of the gastric fluid containing the enzymes, and the gastric emptying. 613 Therefore, in a more realistic scenario, caseins from milk coagulate in the stomach, which affects 614 their digestibility and delays their gastric emptying, as compared to BLG that empties throughout the 615 gastric phase. However, this observation does not apply for UHT-treated milk. 14 Thus, food 616 processing is another factor to be taken into account. Regardless of these potential issues, Egger and 617 co-workers compared the digestion of milk proteins in vivo and in vitro (INFOGEST consensus  618 protocol for static digestion) and observed an agreement between the end points of the gastric and 619 intestinal phases, respectively. 16 (n ≥ 2). Different letters mean significant differences (p ≤ 0.05) between models over time. Absence 629 of letters means no significant differences. Lines are results from isolated proteins in Figure 1. 630 631 Figure 6A shows the proteolysis during the in vitro digestion of soft-boiled hen's egg with the three 632 models. The lane labelled as "Egg" corresponds to the meal blank before digestion, showing the 633 bands corresponding to OVA and LYS. Other egg white proteins can also be identified, such as 634 ovomucins above 95 kDa, ovotransferrin (allergen Gal d 3) of around 76 kDa, and ovomucoid 635 (allergen Gal d 1) of approximately 28 kDa, but with higher apparent Mw (around 36.5 kDa) due to 636 its high degree of glycosylation. 17 Egg yolk proteins can also be detected: α-livetin (allergen Gal d 5) 637 of around 70 kDa. In general, the GI proteolysis of LYS and OVA as part of the soft-boiled egg matrix 638 follows a similar trend ( Figure 6B- and OVA in egg white when heated at 80 °C at a wide range of pH (4-9). 22 However, OVA attained a 656 lower extent of digestion within the egg matrix during the gastric phase for both adult models 657 ( Figure 6D), although a similar extent was reached over the course of the intestinal phase regardless 658 of the model ( Figure 6E), compared to isolated OVA in aqueous solution. Heat-induced (80 °C for 6 h) 659 denaturation and aggregation of isolated OVA has been shown to enhance its in vitro 660 gastrointestinal digestion due to exposure of additional proteolytic cleavage sites that are hidden in 661 the native state. 19,20 The same was observed when heating OVA at higher temperature (100 °C for 5 662 min). 36 However, the lower extent of pepsinolysis of OVA within the egg matrix in the current study 663 as compared to non-heated isolated OVA might be caused by posttranslational modifications, such 664 as glycosylation in the presence of reducing sugars like glucose during cooking. 18 Additionally, it 665 could be due to a limited access of the enzyme cleavage sites in the semi-solid matrix, and once the 666 matrix structure has been broken down at the end of the gastric phase, there is no difference 667 between the intestinal digestion of isolated OVA and within the soft-boiled egg. Martos and co-668 workers observed a lower extent of duodenal digestion of OVA in the whole raw egg matrix as 669 compared to isolated OVA and attributed the effect to the presence of ovomucoid, a trypsin 670 inhibitor that partially retains its inhibitory activity after pepsin digestion. 17 We have not found 671 significant differences in the intestinal digestion of intact OVA when in the soft-boiled egg matrix or 672 isolated. Therefore, soft boiling the egg may have affected the inhibitory activity of ovomucoid. 43 673 However, this inhibitory effect may only be relevant in the infant model where ovomucoid (36.5 674 kDa) seems to resist pepsin digestion. In summary, the effect of the egg matrix and processing, i.e. 675 soft-boiling, had an impact on the kinetics of protein hydrolysis and gastric end point, but not on the 676 intestinal end point.   (n ≥ 2). Different letters mean significant differences (p ≤ 0.05) between models over time. Absence 685 of letters means no significant differences. Lines are results from isolated proteins in Figure 3.

687
The densitometry analysis on SDS-PAGE to determine the digestibility of specific proteins within the 688 food matrix might not be completely accurate since some peptides coming from the hydrolysis of 689 higher Mw proteins can also correspond to the Mw of intact proteins. However, in terms of total 690 protein, one can still appreciate obvious differences in the digestibility of milk and egg proteins 691 across the three in vitro models ( Figures 5A and 6A). Regarding proteolysis in milk, SDS-PAGE shows 692 a faster disappearance of bands corresponding to MFGM proteins (> 55 kDa), bovine serum albumin 693 (66.5 kDa) and α-lactalbumin (14.2 kDa) in the adult models as compared to the infant model ( Figure  694 5A). This can also be generalised for the hydrolysis of other proteins in the egg matrix, such as high 695 Mw ones (>55 kDa) and proteins bands at around 31 and 36.5 kDa ( Figure 6A).

696
To further assess the effect of the different simulated GI conditions on protein digestibility in the 697 food matrix, the OPA assay was performed on digesta samples to quantify the hydrolysis of total 698 protein and results were compared across models. Figure

717
Considering that static models of in vitro digestion can only provide physiologically relevant results 718 at the end points of the gastric and intestinal phases, the conclusions and recommendations in our 719 study are elaborated in terms of the final extent of proteolysis rather than on kinetics.

720
An effect of the digestion scenario was seen based on densitometry analysis of intact protein.

721
Differences were seen in the final extent of gastric and intestinal digestion for isolated proteins and 722 within the food matrix across the three models of in vitro digestion. In this regard, egg proteins LYS 723 and OVA either isolated or as part of the egg matrix showed lower, intermediate and larger extent of 724 gastric and intestinal proteolysis for the infant, early phase and late phase adult model, respectively. 725 Regarding milk proteins BLG and BCS, only BLG showed lower extent of intestinal proteolysis as part 726 of the milk matrix in the infant model. More importantly, the total protein digestion in the milk and 727 egg matrices, quantified by the OPA assay method, followed the trend above for LYS and OVA.

728
When considering the same in vitro digestion model, the food matrix/processing affected the final 729 extent of proteolysis (gastric or intestinal). BLG was digested to a lower extent in the intestinal phase 730 as part of the milk matrix in the infant model. LYS was digested in the gastric phase to a larger extent 731 with the infant and early phase adult model in the egg matrix, whereas the opposite was observed 732 for OVA with the two adult models. Therefore, the interaction of proteins with other components in 733 the food matrix and thermal processing matter even if this is in a liquid/semi-liquid form, which all 734 have an impact on the final extent of proteolysis. Future investigations on the assessment of protein 735 digestibility should consider not only the comparison of different human relevant GI conditions, but also the effect of food matrix and processing relevant to the most likely scenario for the 737 consumption of the protein under investigation.

738
The presence of intact protein throughout the intestinal phase, such as BLG, LYS and OVA as part of 739 the food matrix, may be particularly relevant in infants. Their immature gut is underdeveloped and 740 allows the absorption of appreciable quantities of intact proteins or large peptide fragments, e.g. 741 milk proteins, yet not in nutritionally significant amount but enough to be detected in the circulating 742 blood (4-5 orders of magnitude lower than the oral dose). 44,45 In addition, peptides larger than 9 743 amino acids were present throughout the intestinal phase for all isolated proteins regardless of the 744 digestion model. This increases the possibility that potential immunoactive peptides encounter the 745 immune system through the intestinal route. Work correlating persistent peptide sequences from 746 digesta with binding epitopes positions is still in progress to elucidate differences across in vitro 747 models for improved risk assessment on allergenicity. Further work will also be needed to assess 748 whether this is still the case if brush border enzymes and absorption are included as part of the 749 analysis. In any case, more targeted research will be needed to link these results to immunological 750 outcomes.

753
Authors thank the financial support from the European Food Safety Authority and the School of 754 Food Science and Nutrition of the University of Leeds. K.R. would also like to thank the financial 755 support from Erasmus Plus Programme. Very helpful discussion with EFSA is also gratefully 756 acknowledged.

759
There are no conflicts of interest to declare.

Table of Contents Entry
Dairy and egg proteins either isolated or within the food matrix were subjected to different static in vitro digestion models (infant, fed and fasted adult). Proteolysis differed across models and regarding the effect of the matrix/processing.