Human milk (HM) is dynamic - its composition changes during the day, and over time along with infant needs.
This chronobiology of HM has long been acknowledged to play a role in infant development and programming
but is poorly characterized. While clinical outcomes differ between infants fed at-the-breast (ATB) vs expressed
(ie: pumped) HM, how infant input (via suckling) influences HM biology also remains unknown, even though
“pumping” is an almost universal practice. As such, our over-arching hypotheses are that HM composition and
dynamics, when studied as a biological system, differ between infants fed ATB vs expressed HM, and interrupting
delivery of natural HM chronobiological signals to the infant (via feeding expressed HM) impacts infant sleep
patterns and microbiome development. Our interdisciplinary team will collect daily HM samples, and extensive
metadata from exclusively breastfeeding mothers and infants between 1-4 months postpartum: 60 dyads
exclusively feeding ATB (ATB group) and 60 dyads exclusively expressing HM (Express group). Multi-omics
analytical platforms will be used to characterize HM hormones, macronutrients, microbiota, oligosaccharides,
cytokines, and immunoglobulins to study HM as an ecological system. This novel high-resolution sampling
combined with cutting-edge analytical and modeling techniques power these aims:
1) Detect temporal changes in HM composition diurnally and longitudinally and compare these
trajectories between ATB vs Express groups. Time-series models and machine-learning analytic
approaches will be utilized to infer latent HM dynamics and temporal trajectories, identifying (for the first
time) longitudinal and 24-hour temporal patterns in HM composition. We will then determine if and how
these trajectories differ between ATB vs Express Groups.
2) Identify relationships between HM composition and dynamics with infant sleep patterns. Compare
these relationships and sleep outcomes between ATB vs Express groups. Biometric assessment of
infant sleep will be gathered via actigraphy. Bayesian supervised machine learning algorithms will link HM
“Lactotypes” and components with sleep outcomes. The resulting algorithm will distinguish if relationships
between HM composition and dynamics and infant sleep outcomes differs between ATB vs Express groups.
3) Identify relationships between HM composition and dynamics with the developing infant
microbiome, and compare these differences between groups. Machine learning models will identify
“Milk Lactotypes” and specifics of HM dynamics that predict the time-course of infant microbiome
colonization – a critical marker of future health outcomes. These models will assess how the relationship
between HM and infant microbiome differs between feeding ATB vs Express feeding groups.
Our study will inform precision nutrition guidance to families feeding expressed HM and will also immediately
impact nutrition best-practices for feeding premature infant and milk banking protocols.