Researchers have understood for decades that human breast milk provides optimal postnatal nutrition for infants; however, the molecular mechanisms that enable the transfer of immunity- or metabolic- benefits previously described are incompletely understood. Given emerging evidence that transcellulary acting miRNAs are functionally present in serum, urine, and exosomes 
and are crucial mediators of immune cognition 
, we sought to investigate both their presence and function in human breast milk. We found that not only are known and novel miRNAs present and stable in the milk fat globules, but that their measured expression is altered by a maternal high fat diet. Based on the findings of our target pathway analysis, we speculate that altered expression of miRNA species (from dietary manipulation) bears the potential to modify either mothers’ or infants’ metabolic pathways. This is consistent with the presumptive role of circulating miRNA biogenesis and secretion as pathophysiologic mediators, for which alterations in both concentration and composition of miRNA have been well correlated with cardiovascular morbidity and mortality 
. Zhou 
recently speculated that miRNAs could be transferred through the gut in infants to aid in development of infants’ immune systems; we suggest that breast milk miRNAs could similarly be transferred through the infant gut and modulate gene expression, a mechanism explaining epigenetic maternal effects of offspring phenotype.
We found that the lipid fraction of breast milk, including milk fat globules, contained the highest quantity of miRNA and yielded the purest miRNA fractions. Previous study of mammalian genomes suggested that the most highly conserved proteins in milk were related to secretory processes and were contained within the milk fat globules 
. Conversely, the most evolutionarily divergent proteins were associated with nutritional biogenesis. We found a significant number of conserved and known miRNA species that affect a large number of genes in a focused arena of pathways (Supporting Information S1
and ) in the milk fat globules secreted from mammary cells, suggesting that the process of miRNA excretion in breast milk confers a potential evolutionary advantage.
Since it is not feasible to directly measure nor visualize a human infants’ physiologic responses along these miRNA-targeted pathways, we undertook a robust number of approaches aimed at elucidating and ascribing functional significance to both known and novel miRNAs we identified in breast milk. These included target gene pathway analysis of both known and novel miRNA species, analysis of specific and well-described perturbations in the maternal diet, and ascertainment of the presence of novel miRNA in another tissue unique to reproduction, the placenta.
With respect to the analysis of the mRNA transcripts that are targeted by these miRNAs, the 308 known miRNAs we found in breast milk target genes related to the regulation of gene transcription and metabolism as well as the establishment and refinement of immunity (, Supporting Information S1, Tables S4 through S11
). That the miRNAs in breast milk targeted these functional classes of genes suggests a molecular basis for the benefits experienced by milk–fed infants 
. Additionally, many of the novel miRNAs discovered in our research target genes related to cancer. Target examples such as these suggest that miRNAs in breast milk could contribute to the decreased risk of cancer in humans that were milk-fed as infants 
. Novel-miR-118.2 possibly targets odz, odd Oz/ten-m homolog 2 (Drosophila) (ODZ2
); this family of proteins has been found in high quantities in central nervous system tissue, notably on brain axons, and is thought to be involved in cell-cell communication 
. Based on these target predictions, it is interesting to further speculate that the presence of novel-miR-118.2 in breast milk may suggest a molecular mechanism for neurocognitive benefits derived from breast milk.
To more formally investigate the functional roles of these secretable miRNA species, we interrogated alterations in the relative composition of breast milk miRNAs in response to changes in the maternal diet. We used two previously described, well-characterized dietary manipulations in these experiments, in which breast milk samples from diet-varied women were normalized to rhGH-synchronized lean lactating women (Supporting Information S1, Tables S17 and S18
). With respect to the appropriateness of our normalization cohort, Pearson correlations in pairwise comparisons between pre- and post- rhGH treatment sequencing data for the same subject ranged from 0.96 to 1.0, and correlations across the entirety of the cohort ranged from 0.95 to 1.0 for all known miRNAs identified. Moreover, Limma was applied to the sequencing data after quantile normalization and revealed no differential miRNA expression between the pre- and post- rhGH treatment sequencing data at an FDR <0.01 (Supporting Information S1, Table S18
). These findings are consistent with Maningat’s earlier study of this cohort 
, which demonstrated that although the short-term administration of rhGH induced DNA synthesis, there was no significant difference in milk protein gene expression or milk production between pre- and post-treatment groups.
We have previously demonstrated that high-glucose diets in obese lactating women are associated with higher endogenous fat mobilization and oxidation during meal absorption compared to women fed high-galactose diets, whereas milk production, energy expenditure, and carbohydrate oxidation are no different 
. This is in contrast to earlier findings pertaining to high fat relative to high carbohydrate consumption in lean lactating subjects, whereby a high-fat diet increases milk fat concentration as well as energy output and expenditure 
. In this cohort, a high fat diet (and therefore a low carbohydrate diet, as the two diets were isonitrogenous and isocaloric) resulted in the lactating mothers adapting to low carbohydrate intake by decreasing carbohydrate oxidation. While we did not directly assess the infants’ physiology, our observations pertaining to miRNA excreted in the stable and transferable milk fat globules merit considerable attention.
Consistent with the overall effects of a high-glucose and –galactose diet on the physiology of the mothers and their milk, we did not observe significant variation in any of our novel miRNA species ( and Supporting Information S1, Table S19
). However, the same is not true for women on a high-fat diet. Novel-miR-67 and -27 expression was significantly increased with high fat relative to high carbohydrate consumption (Supporting Information S1, Table S19
) in the cohort as a whole, with as great as 6-fold effect sizes among individuals (). These findings suggest that breast milk has rich epigenetic potential in human infants. Alterations in the quantity and type of miRNA expression in breast milk could thus represent dynamic maternal regulation of infant gene expression based on changing environments and notably with significant maternal diet perturbations. Given that lactation helps mothers and offspring adapt to an ever-changing food supply, this would potentially provide an evolutionary advantage to the offspring 
. Moreover, these findings are consistent with our prior observations in a non-human primate model on the impact of both the maternal and post-natal diet on modification in the offspring’s hepatic epigenome 
. In line with previous research showing that human milk constitution varies with gestational age at delivery as well as through the first six months of life 
, it is logical to speculate that miRNA expression would also change throughout the timeframe of lactation until weaning and future studies will interrogate miRNA profiles by virtue of gestational age at delivery.
Further insights into the potential biologic significance of our identified novel miRNAs can be gleaned by their predicted targeted transcripts. TargetScan was used to predict gene targets for the 12 novel miRNAs that were validated revealing 3554 gene targets for those with a context+ score percentile of ≥90 (Table S14). The top 20 GO terms associated with genes targeted by the novel miRNAs were visualized () and similar to known miRNA targets, a connected network of terms related to regulation of transcription and metabolic processes was observed. With the exception of “integral to plasma membrane” and “intrinsic to plasma membrane”, all GO terms showing enrichment for genes targeted by the novel miRNAs were shared with those enriched for genes targeted by all known miRNAs. Of noted interest, in addition to the transcription and metabolic regulatory network graphs of terms related to the plasma membrane, “regulation of T cell receptor signaling pathway”, and ion binding were uniquely prevalent among targets identified with mapping of our novel miRNA species. The enrichment for “calcium ion binding” is of particular interest given the importance of calcium as a constituent of breast milk 
, and further underscore the likely functional role these novel species heretofore not identified play in both infant and mother.
Finally, we measured the presence and abundance of the novel miRNAs we discovered in placental tissue from both lean and obese subjects to investigate the function of these miRNAs (Fig. S1
and Supporting Information S1, Table S20
). Of our 21 putative miRNAs, none of the 9 species not validated in breast milk fat globules could be validated in placenta. However, of the 12 validated in breast milk, 5 of 12 novel miRNAs were validated in both breast milk and placenta (Novel-miR-62, −114, −118.2, −111, −120; Supporting Information S1, Table S20
). These findings suggest that there may be common expression of miRNA among tissues and fluids which are uniquely produced in women during critical periods of reproduction and development, and may play a regulatory role in both the mother and her infant. This early line of investigation deserves significant attention in future studies.
Our study had several potential limitations. Given the high financial burden of sequencing, we elected to derive limited discovery sets and more robust validation sets. Thus, we performed sequencing on 6 samples from 3 patients in the discovery set, which likely does not completely represent the rich genetic diversity of the human species. However, given that each of our samples came from patients of different races, and that the validation sets were representative of the discovery set with both lean and obese subjects (Supporting Information S1, Tables S17 and S18
), the limit to diversity was theoretically minimized. Another potential limitation of the present study was that potentially important miRNAs might reside in the whey portion of breast milk, which was not represented in our analysis of only the lipid fraction. However, our study is one of few that examines the RNA transcriptome of highly conserved milk fat globules 
, as many other studies of miRNA in breast milk have either discarded the milk fat globules in preparation for study of the whey fraction of milk 
or analyzed whole milk 
. Our findings are thus an important contribution to the existing literature. It is possible that novel or known miRNA may have been degraded in the thaw process prior to purification; however, in previous work by Kosaka 
, the stability of miRNA has been demonstrated in at least 3 freeze-thaw cycles without changes in expression levels and thus we predict that the thaw effect was minimal to negligible. Although we could not biopsy normal infant tissue to more accurately define the functional significance of the novel miRNAs we discovered, their relative abundance in milk from women of various diets may lend merit to their presumptive significance. Finally, small errors in sequencing, though rare in this shotgun sequencing method, may have caused us to misidentify novel miRNAs or fail to identify additional miRNAs. Future studies should consider these limitations and examine the presence of these miRNAs in other breast milk samples from women of varying races and environments, as well as examine profile differences between whey and lipid fractions and fresh versus previously frozen milk. It is imperative that future studies deeply explore the dynamics of miRNA expression due to changing environments. However, despite these limitations, our study, novel in both construct and analysis, makes a pivotal leap in potentially explaining the molecular mechanisms for the benefits milk-fed infants receive.
Our findings indicate that massively parallel sequencing paired with robust analysis in well-described cohorts serves as a valid and powerful means of identifying known and novel miRNAs in human milk fat globules and characterizing their composition and abundance following diet alterations in lactating women. These miRNAs target critical pathways of the regulation of transcription, metabolism, and immunity. Thus, our findings elucidate a novel arena of molecular mechanisms that might explain the well-known benefits of human breast milk for human infants. We speculate that miRNAs function as a previously unrecognized means of dynamic maternal regulation of infant gene pathways and opens an exciting subfield of epigenetics that warrants deeper exploration.