Bioactive peptides in preterm human milk: Impact of maternal characteristics and their association to neonatal outcomes

Human milk adipokines in term babies seem partially determined by maternal factors and affect infant's development. We aimed to describe bioactive peptide concentration in very preterm human milk and associations to maternal characteristics and postnatal growth. Mothers delivering ≤32 weeks of gestation and their infant/s were recruited. At 4 weeks of lactation, an aliquot of 24‐h‐pooled milk was collected for exclusively breastfeeding dyads. Insulin, leptin, adiponectin, and milk fat globule epidermal growth factor‐8 (MFG‐E8) were measured by enzyme‐linked immunoabsorbent assay in skimmed milk. One hundred mothers (28.8 ± 2.3 weeks at delivery) provided a milk sample. Milk insulin was related to gestational age, pre‐pregnancy body mass index (BMI), and galactagogue treatment (final model: adjusted R2: 0.330, p < 0.0001; adjusted β coefficients: galactagogue treatment: 0.348, p 0.001; pre‐pregnancy BMI: 0.274, p 0.009; gestational age: −0.290, p 0.007). Adiponectin was higher in mothers with gestational diabetes (30.7 ± 6.5 vs. 24.8 ± 8 ng/mL, p 0.044). Leptin was associated with pre‐pregnancy BMI (Spearman's ρ: 0.648, p < 0.0001) and MFG‐E8 to presence of labor and multiple pregnancy (final linear regression model, R2: 0.073, p 0.028, adjusted β coefficients: presence of labor −0.229, p 0.050; twins: −0.192, p 0.099). Milk adiponectin was associated with a greater decrease in length z‐scores from birth to 28 days (Pearson's r: −0.225, p 0.032) and to discharge (Pearson's r: −0.290, p 0.003). Milk MFG‐E8 was lower in milk of mothers whose babies experienced late‐onset sepsis (13.3 ± 5.8 vs. 16.8 ± 6.3 μg/mL, p 0.023). Adipokines levels in preterm human milk are partially related to maternal metabolic status. Milk peptide concentration associates with early neonatal growth trajectories.


K E Y W O R D S
breast milk, growth, milk bioactive peptides, prematurity

| INTRODUCTION
Human milk is a complex biological fluid containing both nutrients and a wide variety of bioactive factors. 1 Downstream functions of these elements span effects like modulation of inflammation and the immune system, metabolic development, 2 growth, eating behavior, 3 and long-term programming. 4he abrupt interruption of transplacental flow triggers a rapid decrease of serum leptin and adiponectin in both term 5,6 and preterm 7 neonates during the first 2 days of life.If breastfed, milk becomes the main source of metabolic hormones for the infant.Human milk composition depends on the characteristics of the mother-infant dyad. 8nsulin, adiponectin, and leptin concentration in human milk seem in part determined by maternal adiposity and lifestyle 9 as well as by racial and ethnic factors. 10,11vidence from term infants indicates that breastfed neonates may be protected against obesity in later life when compared to formula fed counterparts. 12The presence of appetite regulating hormones 11,[13][14][15] may play a role in this by exerting biological effects when ingested.
Several studies also support the impact of hormone concentration in human milk on neonatal growth, gastrointestinal development, and body composition.4][25][26][27][28][29] Regarding adiposity, insulin, leptin, and adiponectin concentrations in human milk show a consistent negative correlation with fat-free mass deposition. 17,27ilk fat globule endothelial growth factor 8 (MFG-E8) is a milk fat globule glycoprotein.It has been shown to protect against gastrointestinal inflammation in children with rotavirus infection 30 and demonstrated trophic effects in a short bowel syndrome model in piglets. 31ll this is especially important in the preterm newborn, in whom adipose tissue accretion is still incomplete and who are at increased risk for adult metabolic syndrome. 32The presence of adipokines and insulin in human preterm milk has already been described, 33 but there is little evidence regarding their effect in the immature infant and in their postnatal growth trajectory.MFG-E8 could play a role in gut maturation and in the prevention of necrotizing enterocolitis. 34e hypothesized that some of the associations mainly described in term cohorts between levels of milk peptides and maternal characteristics and with infant growth could be at least partially replicated in the context of very preterm birth.We aimed to determine whether maternal and neonatal characteristics were associated to leptin, adiponectin, insulin, and MFG-E8 content in human milk after very preterm birth and to analyze the impact of these metabolites on growth and development of very preterm infants (VPI).

| Study design
This was an observational bicentric prospective study on a cohort of mother/infant dyads-triads.We designed a descriptive transversal analysis to estimate hormone concentrations in milk at 28 days of lactation, a casecontrol analysis for maternal determinants of bioactive peptide content in human milk and a prospective longitudinal design looking at the association of maternal milk peptides and in-hospital growth and outcomes of VPI.The local research ethics committee approved the protocol (PIC-147-17) and written consent for mother and infant was obtained before inclusion.

| Inclusion criteria
Mothers delivering one or more liveborns at or before 32 weeks of gestation who were admitted in the neonatal unit of one of the participating centers, who had intention to breastfeed and sufficient milk production to provide full enteral feedings for their child/children.
Hospital Sant Joan de Déu and Hospital Clínic are tertiary hospitals in charge of neonatal care within Barcelona Center for Maternal-Fetal and Neonatal Medicine (BCNatal), ensuring homogeneous clinical care.

| Exclusion criteria
Newborns with congenital malformations, known chromosomal, genetic or metabolic abnormalities, or low chances of short-term survival as assessed by the attending clinical team.

| Milk sampling
A representative sample of milk produced in a 24-h period was obtained at 4 weeks of lactation.Participants were educated on the extraction, collection, and storage of samples.They were asked to manually homogenize the milk after each pumping session by swirling the container immediately after filling it, to extract 1-2 mL with a provided sterile syringe and to transfer it to an independent container, separated from the milk that was going to be used for feeding.This was repeated after each pumping session, intending to fully empty both breasts, with a recommendation for eight or more sessions a day, whether by hand or with the help of a manual or electric pump.Milk samples were aliquoted and stored at À80 C for further analysis.

| Clinical data
Maternal (age, BMI, and ethnicity), obstetric (multiple pregnancy and pregnancy complications), lactation, and neonatal (growth and complications of prematurity) clinical data were obtained from clinical charts and from self-reported information during an interview upon recruitment.The mother-infant/s dyads/triads were followed up until discharge.Mothers whose milk production was insufficient after ensuring an adequate breastpumping technique were suggested to start galactagogue treatment with domperidone 10 mg three times every day until their milk production improved.Days on intensive care were defined according to the criteria of level 1 of the British Association of Perinatal Medicine. 35ll data were collected and managed in a database specifically created for this study, using REDCap electronic data capture tools hosted at Hospital Sant Joan de Déu, within the REDCap secure platform. 36

| Statistical analysis
Statistical methods for estimation of population parameters were used to describe concentration of hormones in very preterm milk.Relationships between quantitative variables were analyzed by correlation (Pearson or Spearman depending on whether variables followed a normal distribution).Differences between groups were assessed by Student's t (parametric) or Mann-Whitney U (nonparametric) tests.The distribution of qualitative variables between groups was analyzed by chi-square tests, with Fisher's exact correction when appropriate for sample size.Results with a p-value <0.05 were considered statistically significant.Evaluation of confounders was undertaken by stepwise backwards multivariate regression analysis, including those factors that were significant in the univariate analysis as covariables.

| Sample characteristics
Over the study period, 227 newborns ≤32 weeks from 195 mothers were recruited.These results are based on samples from 100 mothers (of 120 neonates) who provided a milk aliquot at 4 weeks of lactation and had a valid result for at least one assay.Maternal characteristics are summarized in Table 1.
Out-of-range results and those with coefficient of variation higher than 15% between technical replicates were discarded.Table 2 shows the description of milk bioactive peptide concentrations.
Milk leptin had a strong correlation with maternal pre-pregnancy BMI (Spearman's ρ: 0.648, p < 0.0001) (Figure 1); in the multivariate model it was also independently related to a much lower extent to the use of galactagogues after adjustment for twin lactation (final linear regression model, R 2 : 0.631, p < 0.0001, adjusted β coefficients: maternal pre-pregnancy BMI 0.778, p < 0.0001; galactagogue treatment: 0.146, p 0.033).
Milk levels of MFG-E8 were independently related to presence or absence of labor before delivery and whether the mother was lactating one or two infants (final linear regression model, R 2 : 0.073, p 0.028, adjusted β coefficients: presence of labor À0.229, p 0.050; twin: À0.192, p 0.099) in the multivariate analysis.MFG-E8 was lower in mothers who had a vaginal delivery or a cesarean section (c-section) after labor than in those who underwent an elective c-section (17.8 ± 6.8 vs. 13.6 ± 5.9 μg/mL, p 0.020) and higher in milk of mothers lactating singletons (15.0 ± 6.5 vs. 11.6 ± 3.8 μg/mL, p 0.016).

| Growth
Neonatal anthropometrics at birth and infant growth during hospital admission are shown in Table 3.At birth, all anthropometric measurements were around the mean for gestational age, as shown by z-scores (ZS) around 0. During their stay in the neonatal unit, the babies experience a fall in weight, length, and head circumference (HC) ZS that was progressive until discharge.Neonates fed milk higher in adiponectin, showed a greater decrease in length ZS from birth to 28 days (Pearson's r: À0.225, p 0.032) and from birth to discharge (Pearson's r: À0.290, p 0.003) (Figure 2).We performed a multivariate analysis adjusting for factors associated to either milk levels of adiponectin or growth in length: change of length ZS from birth to 28 days remained significant after adjusting for history of GDM (adjusted R 2 : 0.059, p 0.026; adjusted β coefficients: adiponectin: À0.263, p 0.014; GDM: 0.178, p 0.095) and change of length ZS from birth to discharge was significant after adjusting for days on intensive care (adjusted R 2 : 0.279, p < 0.001; adjusted β coefficients: adiponectin: À0.268, p 0.004; days on intensive care: À0.460,There was a trend for a negative correlation between milk insulin and HC ZS change from birth to 28 days (Spearman's ρ = À0.220,p 0.053) that was significant at discharge (Spearman's ρ = À0.232,p 0.030).This relationship was not significant after adjustment for confounders, particularly severity of illness estimated by days on intensive care (linear regression for change in ZS at 28 days R 2 : 0.261, p < 0.0001; adjusted β coefficient: insulin À0.121, p 0.273; days on intensive care: À0.464, p < 0.0001.Linear regression for change in ZS at discharge; R 2 : 0.099, p < 0.004; adjusted β coefficient: insulin À0.093, p 0.416; days on intensive care: À0.293, p < 0.012).These results were similar irrespective of inclusion or exclusion of insulin outliers.
We did not find any relationship between milk levels of MFG-E8 and postnatal growth for the first 28 days or until discharge and none of the studied hormones showed a relation with neonatal weight during admission.

| Neonatal morbidities
Neonatal condition at birth and main morbidities during admission are shown in Table 3.
We found no significant differences in bioactive peptide concentration between the milk of mothers whose babies went on to develop necrotizing enterocolitis, bronchopulmonary dysplasia, retinopathy of prematurity, or intraventricular hemorrhage, although sample size was small.

| DISCUSSION
In this study, we confirm the presence of adipokines as well as insulin and MFG-E8 in preterm human milk.Also, we show a positive relation between maternal adiposity and leptin and insulin milk concentration at 4 weeks of lactation and a weak negative correlation between milk adiponectin concentration and neonatal length trajectory during admission.
Insulin concentration was negatively correlated with gestational age, although the strength of the correlation was weak.Shehadeh 37 found that human milk insulin levels were stable in mothers delivering between 30 and 41 weeks of gestation, we could speculate that a higher level of insulin could be beneficial in order to exert an effect on the more preterm babies, who seem to be relatively insulin resistant. 38e impact of maternal characteristics on milk hormone concentration is controversial.In our cohort, maternal prepregnancy BMI correlated positively with human milk leptin (strongly) and insulin (weakly), in agreement with the findings for leptin in a systematic review by Andreas et al. 38 The same association with maternal fat mass has been less consistently reported for human milk insulin. 9Even though this correlation has mainly been tested in mothers of moderate preterm and term babies, our results confirm that human milk leptin and insulin are related to maternal adiposity regardless of the duration of gestation.
In line with our results, Brunner et al. 27 did not find a correlation between adiponectin and maternal physical adiposity markers either before or in the first and second trimesters of gestation in a cohort of 150 mother-infant dyads.On the other hand, Martin et al. 11 described a positive correlation between postpregnancy BMI and milk adiponectin in a group of 200 mothers of term babies with milk collected at different lactation points.Those differences suggest that adipose maternal characteristics toward the end of a full term pregnancy may have a more direct effect on milk adiponectin concentration.
The relation between galactagogue treatment and insulin we obtained in this cohort could be explained by the antidopaminergic effect of domperidone enhancing prolactin secretion in these mothers. 39Prolactin has been described to promote insulin resistance and increase plasmatic insulin levels in a mice model. 40ther maternal metabolic conditions such as GDM may be involved in the regulation of hormone concentration in human milk.Breast milk adiponectin concentration in our cohort was higher in mothers with GDM in contrast with some previous publications. 41,42As mentioned by Choi et al., 43 higher levels of adiponectin in the milk of diabetic mothers may counter the inflammatory effect that insulin resistance and hyperglycemia could have in the immature bowel.MFG-E8 has been related to bowel inflammation 44 and carcinogenesis 45 but, as far as we are aware, we are the first to explore the association of MFG-E8 milk concentration to maternal factors.In our cohort, the concentration of MFG-E8 was higher in mothers who delivered without labor, in line with the change in the amount of total milk protein that we recently described in this group of patients. 46everal studies support the impact of hormone concentration in human milk on neonatal growth.Milk adiponectin concentration showed a consistent weak negative correlation with length trajectory during hospital stay.Several authors have previously described poorer growth in term infants that receive milk with higher amount of adiponectin, 41,47 although this was described in terms of less weight gain.Although not significant after adjusting for severity of illness, neonates fed milk with higher insulin content showed a poorer HC growth between birth and discharge similar to the data already reported for adiponectin. 48egarding the lower MFG-E8 milk content in patients who experienced a LOS and taking into consideration its already described anti-inflammatory effect, 44 further research on this area could give an interesting clue in avoiding hospital-acquired infections.
We acknowledge some limitations of this study.In line with previous publications, we measured bioactive peptide concentration in skimmed milk.Both adiponectin and MFG-E8 are lipophilic molecules, so a fraction of the total peptide content could have been discarded with the fat when processing the sample.However, Ley et al. 49 did not find differences in adiponectin concentration on skimmed or whole milk and Asaro et al. had already used skimmed milk to determine MFG-E8 concentration in human milk, 50 which we chose to replicate for comparability.
The analysis of hormone concentration in milk might not reflect the amount of hormone the infant received (related to the volume of the intake).Also, this was a single-point analysis of mother's milk and milk composition is known to be dynamic.Body composition was evaluated with weight and height, as other markers such as skinfolds, fat mass, or previsceral fat were not available.A larger sample size would have been desirable to better address the point of multiple testing, but milk research in the context of hospitalized VPI comes with its own ethical and logistical challenges.Nevertheless, we feel that the nature of the sampling, which was representative of milk production over 24 h, as well as the detailed clinical characterization adds relevant information to the field and gives some basis for the design of further studies.
In conclusion, bioactive peptides are present in human preterm milk, their specific levels associate to certain maternal factors and they may play a role in neonatal growth and clinical outcomes during hospital stay.Further studies including long-term follow-up may be useful to elucidate if hormones in mother's milk may play a role in reducing the later metabolic risk associated to premature birth.

F I G U R E 1
Abbreviations: CI, confidence interval; MFG-E8, milk fat globule endothelial growth factor 8; SD, standard deviation.

F I G U R E 2
Milk adiponectin concentration and change in length z-score.Scatterplots of adiponectin levels in milk and change in newborn length z-score from birth to 28 days (n = 91) and discharge (n = 105), with best fit line for linear regression and 95% confidence interval.dol, days of life; ΔLZS, change in length z-score.F I G U R E 3 Milk MFG-E8 concentration and neonatal late onset sepsis.Boxplots comparing concentration of MFG-E8 in the milk of mothers whose babies had late-onset sepsis and those who did not.Boxplots whiskers extend to the 5th and 95th percentile.Boxplots whiskers extend to the 5th and 95th percentile.LOS, late onset sepsis; MFG-E8, milk fat globule endothelial growth factor 8.