Allergen Shedding in Human Milk from Mothers of Preterm Infants: Proteomic and Peptidomic Feasibility and Pilot Analysis

population and whether the complex interaction between allergens, proteases, and other human milk components can serve to induce sensitization or tolerance to allergens in infants. We were able to nd several food an environmental allergens in the samples that where run. The samples had good overlapping alignment which meant the results were reliable and there was differential ndings compared to formula which would not be expected to have many allergen peptides aside from those associated with cow milk. We also found there was no difference in variety of allergens between human milk from mothers of term and preterm infants. We were able to subsequently run pilot samples from our Microbiome, Atopy, and Prematurity study through the mass spectrometer and also perform protease studies on these pilot samples. We will now be able expand our analysis to look at the larger cohort of samples we have collected longitudinally over the rst year of life to see how they compare to our feasibility and pilot runs. with the exception that 40% ACN was used to elute peptides. The eluents in speed-vac in mass spectrometry analysis. Trypsin-digested peptides were analyzed by ultra-high pressure liquid chromatography (UPLC) coupled with tandem mass spectroscopy (LC-MS/MS) using nano-spray ionization. The nanospray ionization experiments were performed using a Orbitrap fusion Lumos hybrid mass spectrometer (Thermo) interfaced with nano-scale reversed-phase UPLC (Thermo Dionex UltiMate™ 3000 RSLC nano System) using a 25 cm, 75-micron ID glass capillary packed with 1.7-µm C18 (130) BEH TM beads (Waters corporation). Peptides were eluted from the C18 column into the mass spectrometer using a linear gradient (5–80%) of ACN (Acetonitrile) at a ow rate of 375 µL/min for 2h. The buffers used to create the ACN gradient were: Buffer A (98% H 2 O, 2% ACN, 0.1% formic acid) and Buffer B (100% ACN, 0.1% formic acid). Mass spectrometer parameters are as follows; an MS1 survey scan using the orbitrap detector (mass range (m/z): 400-1500 (using quadrupole isolation), 120,000 resolution setting, spray voltage of 2,200 V, Ion transfer tube temperature of 275°C, AGC target of 400,000, and maximum injection time of 50 ms) was followed by data dependent scans (top speed for most intense ions, with charge state set to only include +2-5 ions, and 5 second exclusion time, while selecting ions with minimal intensities of 50,000 at which the collision event was carried out in the high energy collision cell (HCD Collision Energy of 30%), and the fragment masses were analyzed in the ion trap mass analyzer (With ion trap scan rate of turbo, rst mass m/z was 100, AGC Target 5000 and maximum injection time of 35ms). Protein identication


Introduction
There has been increasing recognition of the protective role of human milk and early food exposure in the development of atopic conditions such as eczema, food allergy, and asthma. [1][2][3][4] Asthma is the most prevalent chronic disease in children, affecting over 300 million people worldwide 5 and disproportionately affects preterm infants. In one study, infants born before 37 weeks (moderate to late preterm) were 50% more likely to develop asthma, and infants born before 32 weeks (extremely to very preterm) were three times as likely to develop asthma 6 . While mechanical ventilation and other causes of direct damage to the lung may account for some of this increased risk for the development of asthma in the preterm population, other factors may also play a role. Lower rates of atopy in term infants were associated with exclusive breastfeeding for the rst four months of life 7 , lack of antibiotic exposure (either maternal intrapartum or early in infancy), 8 vaginal delivery, 9 and furry pets in the home. 10 Preterm infants spend this critical period in the neonatal intensive care unit and some are exposed to antibiotics and c-section delivery but most importantly, almost all experience most of their nutrition via a nasogastric tube or a bottle containing previously frozen maternal or donor breast milk, instead of fresh breastmilk via breastfeeding.
Recommendations regarding the introduction of allergenic foods to infants has shifted from avoiding allergenic foods until 1 year of age to early introduction prior to 6 months of age, as the latter has been shown to be associated with a decreased risk of food allergy during a critical window of the infants immune development. 4 This approach is thought to reduce the induction of Type 2 in ammation that is responsible for allergic conditions. Given that early exposure to food proteins can alter the development of food allergies later in life, early nutrition has become an area of interest in studying the pathogenesis of atopic disease. 11 Over 70% of peanut reactions occur on the rst known exposure, indicating prior sensitization, perhaps transcutaneously, via inhalation, or via human milk. 12,13 Since human milk is often the primary source of nutrition in infancy, it has been postulated that it could be a source for allergen introduction. 14 Although IgEmediated reactions to human milk are rare, they do occur, demonstrating its immunogenic nature. 15 16 Airborne allergens from house dust mite have also been found in human milk at similar quantities to food allergens. 17 One dust mite protein, Der p1 has demonstrated both Toll-like receptor agonist and protease activities, which could potentially initiate allergic immune responses. 18 In addition to food and environmental allergens, human milk also contains many other bioactive substances, including endogenous proteases and protease inhibitors, immunoglobulins, soluble receptors, cytokines, human milk oligosaccharides (HMOs), fatty acids, and microbes. 11,19 20-23 Human milk proteomics studies have utilized different methods for protein identi cation, including Western blotting, ELISA, and mass spectrometry, which may account for heterogeneity between experiment results. [24][25][26][27][28][29][30][31] Mass spectrometry has been utilized to determine the protein content in human milk, however, there are few studies that are primarily focused on allergenic proteins. 11 Studies utilizing mass spectrometry, which allows for a broader untargeted search for proteins, have identi ed 1200-1600 total proteins in human milk, the vast majority of non-human proteins derived from cow's milk, with one study also identifying dog, horse, cat, chicken and rice proteins. 15,32 The degree of protein alteration prior to its appearance in human milk is unknown, and studies are con icting. Some studies have found that peanut Ara h 1 and 2, 14 ovalbumin, 32 and gliadin 25 are not degraded in human milk. This contrasts with other studies that identi ed fragments of ß-Lactoglobin 15,33 and α-S1 casein. 53 Proteases have been identi ed in human milk and appear to play a signi cant role in infant digestion, but the interaction between endogenous human milk proteases and non-human proteins such as house dust mite proteases and human milk forti ers (HMFs) proteases has not been extensively studied. 34 The development of tolerance versus sensitization to allergens is complex and depends on the interaction and often multi-directional relationship between many different factors, such as maternal history, milk composition, gut immunology and microbiome and external environment. 21 Allergen shedding in human milk may be a way to educate the infant's immune system and modulate allergy risk in the infant. 35 Given advancements in medicine that have led to increased survival of preterm infants, we aim to examine longer term outcomes such as allergic sensitization or tolerance in this population by examining human milk protein composition and exposure. Due to newer techniques in proteomic and peptidomic analysis, and the paucity of data regarding the ability of HM to induce tolerance or sensitization to allergens, we developed a feasibility pilot study on a subset of milk samples to investigate how nutrition and environmental exposures may impact allergen shedding in human milk in preterm infants.

Sample Collection
Four human milk samples (2 from mothers of term infants and 2 from mothers of preterm infants) were analyzed from the Mommy's Milk Human Milk Biorepository (HMB) to evaluate the feasibility of analyzing human milk samples by untargeted mass spectrometry (Table 1a). The Mommy's Milk HMB was founded at the University of California, San Diego in 2014 with the goal of building a constant but rotating inventory of 3,000 human milk samples available for future research. 36 Following informed consent, women provide 50 mL up to a full pump of expressed breast milk (convenience sample). Participants are interviewed about their sociodemographic characteristics, pregnancy history, dietary intake using a standard questionnaire (https://www.nutritionquest.com), medication exposure, lifestyle habits, maternal stress, anxiety and depression, breastfeeding behaviors, and signs and symptoms of potential adverse reactions in the offspring. Data on growth of the infant/toddler are captured from medical records, and neurodevelopmental assessments are conducted longitudinally. Sample collections occur at UC San Diego, community sites, or the participant's home. Human milk samples are stored and shipped on ice within 24 hours of collection to the Mommy's Milk lab where the sample is aliquoted and stored at −80°C until requested for study analysis. Sample preparation for proteomic analysis Milk samples were thawed on ice prior to preparation for proteomic analysis. Guanidine-HCl was added to 2 µL of milk sample to achieve a nal concentration of 6 M. The samples were boiled for 10 minutes followed solution was then acidi ed using TFA (0.5% TFA nal concentration) and mixed. Samples were desalted using 100 mg C18-StageTips as described by the manufacturer protocol. The peptide concentration of the samples was measured using BCA after resuspension in sample loading buffer and a total of 0.5 µg was injected for each label free quanti cation run.

Sample preparation for peptidomic analysis
To remove high molecular weight milk proteins, 100 µL of human milk was mixed with 900 µL of methanol and vortexed for 5 seconds. The samples were kept at 25°C for 30 minutes followed by centrifugation at 12,000 rpm for 10 minutes at 25°C. 500 µL of supernatant was transferred to a fresh tube dried in a vacuum centrifuge. The samples were hydrated in 0.5 mL of 0.5% formic acid and 5% acetonitrile (ACN) solution and desalted using a Sep-PAK C18 1 cc Vac (Waters Corporation, Milford MA) according to the manufacturer's protocol with the exception that 40% ACN was used to elute peptides. The eluents were dried in speed-vac in preparation for mass spectrometry analysis.

Liquid Chromatography with Tandem Mass Spectometry (LC-MS-MS)
Trypsin-digested peptides were analyzed by ultra-high pressure liquid chromatography (UPLC) coupled with tandem mass spectroscopy (LC-MS/MS) using nano-spray ionization. HMB samples did not contain forti er and did not undergo protease analysis. (Summary of milk feed content and analysis performed-   (Table 3). We detected peptides from various food, venom/salivary, and airborne sources. There were no appreciable differences between term and preterm samples in terms of total protein content. One sample accounted for over 50% of the non-human peptide variability (R4). In their dietary histories all mothers ingested sh, shell sh, nuts, and wheat. Two out of the four mothers did not consume cow's milk (dairy), although bovine peptides were found in all samples. Two of the four mothers did not ingest egg, yet egg protein was detected in both of those samples. Interestingly, the one mother who drank almond milk was the only one who did not have almond detected.
Proteomic Pilot Study from the MAP samples In total, we identi ed proteins from 23 different species, including aeroallergens, food and contact allergens (Table 3). For quality control purposes peptide alignment maps were made for two common allergens, betalactoglobulin and cat albumin. The maps demonstrated consistent overlap, thereby supporting that these proteins were indeed identi ed and digested similarly between samples. (Supplemental Figure 1) Bovine peptides were the most numerous of the non-human peptides detected in human milk samples without HMF. Speci c allergenic bovine peptides (β-lactoglobulin, α-and β-casein, α-lactalbumin) were in the highest relative quanti cation in regular forti er and formula, intermediate in hydrolyzed forti er and lowest in human milk samples (Supplemental Figure 2).
Peptidomics studies of the MAP samples identi ed peptides that were generally less than 40 amino acids in length ( Figure 2). Most non-bovine peptides were digested (28/33) and found in samples with human milk and forti er which contrasts with only a few (2/28) being found in the formula sample.

Protease activity
Upon analysis of the amino acids at the N-and C-termini of peptides, we discovered that there was a high frequency of proline (P) and glutamine (Q) residues in the MAP samples. This was not surprising because human casein is the major protein in these samples and 17% of residues in this protein are proline while 11% are glutamine. Samples J8-J11 are notable as they have peptides that were cleaved after lysine (K) and arginine (R) at the N-terminal side of peptides. This is not seen for J7 and J12 (Figure 3). To quantify protease activity in these human milk samples, we assayed the samples with the uorogenic substrates, Arg-Arg-AMC. We found that the most abundant protease activity was in human milk sample J9, while the formula only sample J11, had the lowest activity ( Figure 4).

Discussion
To our knowledge, this is the rst combined human milk mass spectrometry and protease analysis using clinically relevant NICU preterm milk samples. While the detection of food peptides in the milk samples is interesting and somewhat expected, the size, breadth, and variety of food and aeroallergens as well as other non-food peptides is fascinating, especially when comparing human milk to cow milk formula. Furthermore, we found differential protease activity between the samples with the highest being in maternal expressed breast milk alone, without forti er (J9) and the lowest in formula (J11).
The presence of allergen peptides in human milk does not appear to be accidental and may be linked to the development of allergy. In one study, there was an increase in atopy in children who were breastfed by atopic mothers and found to have high HM dust mite (Der p 1) levels; this was not noted in the offspring of mothers without atopic history regardless of Der p 1 level in human milk . 18 In food allergy, maternal cow's milk avoidance was associated with increased cow's milk allergy in offspring, mediated by a lower cow's milk speci c IgA and possibly the lack of cow's milk protein exposure. 23 In our analysis, bovine peptides were the most numerous of the non-human peptides detected in the human milk only samples (without formula or forti er) and speci c allergenic bovine peptides (β-lactoglobulin, α-and β-caseins, α-lactalbumin) were found in the highest relative quanti cation in regular forti er and formula samples and lowest in the human milk samples.
Multiple other common food allergens have been identi ed in human breastmilk (HBM) studies. Ovalbumin has been detected in HBM in 8.3%-76% of subjects, 24,30,31,37 while ovomucoid was identi ed in 78% of subjects in one study. 30 There appeared to be a dose-response phenomenon between maternal egg intake and infant serology, whereby for each additional egg ingested, HBM ovalbumin concentration increased by 25% and infant egg-speci c IgG4 increased 22%. 37 Egg protein (β-enolase) was found in our analysis although speci c ovalbumin and ovomucoid peptides were not identi ed. Regarding peanut protein, one study of 23 lactating females found that after a 50 g oral peanut load, 48% of female subjects' HM samples contained peanut. 14 Another small study demonstrated peanut allergen (Ara h 6) in HM that was functional and IgE-reactive as evidenced by in vitro assays and the observation that administration to mice lead to partial oral tolerance. 38 Peanut protein was not identi ed in our analysis. With wheat protein, gliadin was detected in 67.5% -100% breast milk samples in two studies. 25,39 Multiple different wheat peptides were identi ed in our samples, but not gliadin.
In the food diaries associated with the maternal milk samples that were not augmented with forti er or formula, some foods that were reported as consumed did not show up in the samples. Conversely, in other cases, foods that were not reportedly consumed, did show up in the samples. While recall bias and ensuing inaccuracy may partially account for these discrepancies, there is also the issue of timing of food consumption with respect to appearance in human milk. Moreover, the capability of excreting speci c proteins may vary between mothers and further impact the presence of allergenic proteins, which further complicates attempts at correlating dietary ingestion and breast milk peptides.
We know that antigen-presenting cells introduce processed allergens to T-helper lymphocytes and proceed down a TH2 pathway in allergic conditions. 40 How the allergen is processed, the role of proteases, and the exact conformation of different allergenic proteins in human milk is not known, although the size of the original protein was better elucidated in our study. We demonstrated that many bovine peptides are found digested (original protein size> 40 amino acids) and free (original protein< 40 amino acids), indicating that there are a variety of different parent proteins. These proteins were mostly found shared between the human milk plus forti er or formula samples. Conversely, human milk samples without forti er had relatively few digested bovine peptides, supporting that most cow's milk-derived peptides originated from smaller proteins. Interestingly, cow's milk allergy is one of the rst to appear in infants' and the majority of those are sensitized to caseins, which may be able to cross the GI border relatively intact as they coagulate in acidic conditions and may be less susceptible to proteolysis. 32,41 A variety of caseins of different sizes were identi ed in our formula, forti er and human milk samples, although the allergenicity of these speci c caseins are not de nitively established in this current analysis. Additionally, the exact origin of these proteins, although presumably diet-derived, is unknown. As opposed to various sizes, the majority of nonbovine peptides in our human milk samples were digested, thereby originating from peptides over 40 amino acids in length.
Assuming that most of these peptides were generated by proteases, we looked at the amino acid sequence in the protein that ended up being the substate for cleavage. Sample J9 (pure maternal expressed breast milk) from the MAP study showed the highest protease activity. Previous proteomic studies have shown differences in the presence of proteases and protease inhibitors in HM between allergenic and non-allergenic mothers. 42 There is evidence that an imbalance between protease and protease inhibitors in HBM could allow for easier penetration of allergens. 43,44 Speci cally, reduced cystatin, a protein inhibitor that has been detected in HM, secreted by epithelial cells has been linked to easier penetration of Der p1 through skin. 45 Furthermore, protease inhibitors have been detected in the stool of infants who have received HBM indicating that these protease inhibitors may be active in the gastrointestinal tract. 46 It is thought this complex interplay between allergens, proteases, and protease inhibitors is important in the pathogenesis of atopy, and protease inhibitors are being evaluated as a potential therapeutic agent to treat asthma and other atopic conditions.
There are several limitations to this study. We started with a small batch of samples to assess feasibility in this pilot trial. There is inconsistency between dietary documentation between the two sample groups. We will have a more consistent and larger sample size in our future analysis. Other sample-based limitations include the lack of multiple "pure" samples that contain only maternal expressed breastmilk without forti er or the use of pooled donor human milk. Theoretically, a subtractive analysis could be considered, with inference of protein content of breast milk via exclusion of proteins found in forti er, however, this is limited due to the overlap of proteins between forti er and human milk and the variability between the samples, including differences in protein content between the two samples of the same forti er. Moreover, since a large proportion of preterm infants receive supplementation, donor milk, or formula, our results re ect the real world setting in the NICU. Our subject dietary history did not include the temporal relationship of speci c food ingestion and sample collection. Thus, secretion kinetics cannot be concluded, and contamination/inadvertent consumption is an issue with the self-reported dietary histories. Closer analysis of maternal diet and timing of consumption may help to determine the kinetics of human milk peptides and the degree of contamination (dietary or via mass spectrometry) that could account for the detection of proteins that are not found in the diet. Database limitations are also possible. We did not manually blast all proteins against NCBI and Uniprot databases, only those which were positively identi ed, so it is possible that there were false negatives and proteins were not identi ed due to inaccurate database sequences

Future directions and Conclusions
We have taken a large step forward in identifying what a preterm infant immune system may encounter in their milk feeds, however, it was beyond the scope of this study to determine the origin of the human milk peptides identi ed. This is an area we plan to investigate in the future. Peptides may be secreted by lactocytes or enter via the bloodstream. It is also not known where proteolytic cleavage occurs, whether it is locally in the breast or in the GI tract/blood, which could be further investigated by paired blood samples in future studies.
The interaction between allergen, protease, and protease inhibitors also warrants further investigation. Identifying which proteases and protease inhibitors are present in our MAP samples would be of great interest, particularly if their presence or absence augments the development of atopic conditions in infants who have been in the NICU. We do plan to follow subjects out to 5 years and look for the development of allergic outcomes in our MAP cohort. The use of formula, forti ers, and donor milk are important in optimizing the growth and development of preterm infants. However, their use may have unintended longterm consequences, that need further investigation.
In conclusion, the detection of various allergenic peptides and protease activity in our milk samples raises more questions about how modifying feeds in the NICU may impact the development of atopy in preterm infants. Ultimately, whether human milk can serve to induce allergic sensitization or tolerance in an infant is an area of research that needs much further exploration.

Declarations
The data that support the ndings of this study are available from the corresponding author, SL, upon reasonable request.
The authors declare that they have no known competing nancial interests or personal relationships that could have appeared to in uence the work reported in this paper.