Mammalian Milk Glycomes: Connecting the Dots between Evolutionary Conservation and Biosynthetic Pathways

Abstract

Milk oligosaccharides (MOs) are among the most abundant constituents of breast milk and are essential for health and development. Biosynthesized from monosaccharides into complex sequences, MOs differ considerably between taxonomic groups. Even human MO biosynthesis is insufficiently understood, hampering evolutionary and functional analyses. Using a comprehensive resource of all published MOs from greater than 100 mammals, we develop a nonparametric pipeline for generating and analyzing MO biosynthetic networks, which readily generalizes to other glycan classes. We then use evolutionary relationships and inferred intermediates of these networks to discover (i) distributional glycome biases, (ii) biosynthetic restrictions, such as reaction path dependence, and (iii) conserved biosynthetic modules. This allows us to prune and pinpoint biosynthetic pathways despite missing information. Machine learning and network analysis cluster species by their milk glycome, identifying characteristic sequence relationships and evolutionary gains/losses of motifs, MOs, and biosynthetic modules. These resources and analyses will advance our understanding of glycan biosynthesis and the evolution of breast milk.

Predicting glycan structure from tandem mass spectrometry via deep learning

“…15, Supplementary Table 4), such as the reducing end GalNAc (likely due to the mass range of the used mass spectrometer), while others, such as Gal3Sβ1-3GalNAc, were nearly always reliably measured whenever larger structures that included this building block as a substructure were present in a sample. We believe that this approach might shed light on subsets of the O-glycome that are currently hard to measure, as we here, again 7 , noted the peculiar absence of GlcNAc-terminated structures from measured glycans as a trend. Myeloid Leukemia (AML) cell lines (GPST000214) and differentiated colorectal cancer cell lines (CaCo-2, GPST000256), via a CandyCrunch model not trained on these datasets, are shown via t-SNE (n = 103), using glycan abundance as features.…”

“…Further analyses across our networks then allowed us to compare the reaction order of glycosyltransferases, reinforcing the highly dominant nature of galactosyltransferases 7 (Fig. 4B).…”

“…Reasoning that glycans in a biological sample tend to be biosynthetically related, i.e., contain precursors/intermediates of larger biosynthetic pathways, we turned to our recently developed method of constructing glycan biosynthetic networks 7 . Applying this method to a typical CandyCrunch output (Supplementary Fig.…”

“…For this, we re-used the total 250,000 O-glycan spectra mentioned in the context of Fig. 2, to construct biosynthetic networks 7 . As described above, this process filled in the gaps of unobserved intermediates in the biosynthesis of observed structures.…”

“…For a given sample, all retained top1 predictions were used to construct a biosynthetic network as described previously 7 , using the implementation within glycowork. For milk oligosaccharides, this also included evolutionary pruning, as pre-calculated species networks were available.…”