Among the large placental mammals, ursids give birth to the most altricial neonates with the lowest neonatal:maternal body mass ratios. This is particularly exemplified by giant pandas. To examine whether there is compensation for the provision of developmentally important nutrients that other species groups may provide in utero, we examined changes in the lipids of colostrum and milk with time after birth in giant pandas. Lipids that are developmental signals or signal precursors, and those that are fundamental to nervous system construction, such as docosahexaenoic acid (DHA) and phosphatidylserines, appear early and then fall dramatically in concentration to a baseline at 20–30 days. The dynamics of lysophosphatidic acid and eicosanoids display similar patterns, but with progressive differences between mothers. Triglycerides occur at relatively low levels initially and increase in concentration until a plateau is reached at about 30 days. These patterns indicate an early provision of signalling lipids and their precursors, particularly lipids crucial to brain, retinal and central nervous system development, followed by a changeover to lipids for energy metabolism. Thus, in giant pandas, and possibly in all bears, lactation is adapted to provisioning a highly altricial neonate to a degree that suggests equivalence to an extension of gestation.
Milk is the sole source of nutrition for mammalian neonates, and is also an essential conduit of immune support for many species of infant. Mammary secretions change dramatically from colostrum (‘first milk’) to mature, main-phase milk during the immediate post-partum period. Broadly, colostrum tends to be more protein-rich and lipid-poor than later milk, and is particularly rich in immunoglobulins and innate anti-microbial factors1,2,3,4,5,6,7. The change to mature milk represents a switch to more energy-rich nutrition in which sugars and fats may predominate1,2.
The neonates of some species are absolutely dependent for their survival and development on colostrum from their mothers, whilst others are less so. The period of time for this dependence – and hence the duration of the colostrum phase of lactation – also varies among species. The difference is largely a function of the type of placenta involved. For instance, species with epitheliochorial placentae, such as ungulates, cannot transfer immunoglobulins from maternal to foetal circulations. Colostrum is thereby essential to survival of these neonates2. At the other extreme are species with haemochorial placentae (e.g. humans) whose placentae transport immunoglobulins from maternal to foetal blood circulations prior to birth, although this is confined to immunoglobulin G (IgG)2,3,4.
Ursids (bears) give birth to the most altricial (developmentally immature) neonates with the lowest neonate:maternal body mass ratios amongst eutherian (“placental”) mammals8. We hypothesised that the ursine lactation strategy is adapted to meet the developmental needs of highly altricial young in two main ways. First, ursine milks may change more slowly from colostrum to main-phase milk than do milks of other species groups. Second, the mixture of components of ursine colostrum may be unusual if not unique among eutherian mammals. Our examination of proteomics and metabolomics have upheld this hypothesis9,10. The objective for the study reported here was to test whether these predictions hold true for lipids, particularly given the special roles lipids have in development.
We chose giant pandas for this study because they exhibit an extreme even amongst ursids in the altriciality of their neonates11. In addition, the captive breeding program for giant pandas in China and its associated intensive human handling of the animals provided an opportunity for serial milk sampling in the immediate post-partum period. In our previous studies of panda milk, we discovered the transition between colostrum to mature-phase milk to be unusually, perhaps uniquely, prolonged amongst eutherians9. That study revealed a slow maturation in the protein profiles. The changeover and maturation process of these profiles takes approximately thirty days to complete, over which period certain species of oligosaccharide disappeared, whilst others appeared. Further detail was added in a second study, in which a broad-spectrum metabolomics approach discerned three phases in the transition from colostrum to mature lactation10. The milks of different mothers in that study were at first similar in composition, then diverged after about seven days.
There are no precise criteria that define the end of the colostral period and the onset of main-phase lactation for any species. We here take the colostrum phase to end when all the major components of milk reach an approximate steady state, although slight modifications in composition may still occur during the main phase. By this definition, the colostrum period of giant pandas ends at about 30 days9,10, which is considerably longer than is known for any other eutherian, and stands in stark contrast to the less than 24 hours in, for example, Atlantic grey seals (A. Lowe, P. Pomeroy, D.G. Watson, M.W. Kennedy, unpublished).
Lipids in milk could be broadly divided into those needed for metabolic energy support, the construction of membranes and antimicrobial activities, and those that serve as immune and developmental signalling molecules or their precursors. Energy for foetuses is almost exclusively provided in the form of trans-placental delivery of glucose and lactate12. Triglyceride transfer for this purpose is minimal or non-existent, and foetal gluconeogenesis is essentially inactive12. After birth, however, milk offers a switch to lipid- and lactose-based energy provision, the balance of which varies from species to species1,2.
Large-brained mammals exhibit a heavy requirement for polyunsaturated fatty acids (PUFAs) for brain development12,13. This appears to be satisfied in utero by translocases that transport crucial PUFAs such as arachidonic (AA) and docosahexaenoic (DHA) acids en masse from the maternal to the foetal circulation, resulting in higher levels of PUFAs in foetal than maternal circulations12. Trans-placental transportation of PUFAs may be confined to species with haemochorial placentae12. It is minimal in those with epitheliochorial placentae (e.g. sheep14), and may also be restricted in species with endotheliochorial placentae15. The latter placental type is widespread within the Carnivora16,17,18, among whom there may be a need for immediate post-partum provision of PUFAs for continued neonatal development. This requirement may be particularly urgent for highly altricial neonates.
Here we show that there is a slow transition between colostrum and mature milk lipid profiles in giant pandas, and that some species of lipid are at first abundant and then fall away, whilst others increase with time and stabilise in concentration. The main changeover, like that which we demonstrated for the protein, oligosaccharide, and metabolome profiles9,10, occurs between twenty and thirty days. Importantly, lipids such as PUFAs that are considered to be crucial for brain growth and for developmental and immune signalling are present at their highest levels only very early in lactation, while those needed predominately for energy generation and membrane construction appear later.
Lipid profile of giant panda milk
The analysis of milk lipids required sequential rounds of mass spectrometry (MS) and ion fragmentation, particularly for monoacyl and diacyl glycerolipids in which the acyl tails are heterogeneous and MS analysis does not directly identify the positions of carbon:carbon double bonds in a fatty acid chain. A summary of the main types of lipids that were identified is given in Table 1 (a more detailed version of this information is provided in Supplementary Table S1). The principal lipids were phospholipids, sphingomyelins, glycerolipids (e.g., triacylglycerols/triglycerides (TAGs)), free fatty acids (FFAs) and sterols, with a few other classes contributing to a total of 403 distinct species. The lipids are typical of mammalian milks, and ranged as expected from those broadly involved in energy provision (TAGs), membrane construction (phospholipids), precursors of developmental and immune or inflammatory signals, and lipids enriched in the nervous system and retina (PUFAs).
Developmental and signalling lipids and their precursors
Of particular interest, given their abundance in the nervous system and their roles as precursors to pharmacologically and immunologically active compounds, certain PUFAs (e.g. docosahexaenoic acid (DHA), Fig. 1) were relatively abundant very soon after birth but then decayed to a low plateau by about 30–40 days. The same applied, albeit less clearly, to eicosapentaenoic (EPA) and arachidonic acids (AA) (Supplementary Figure S1). Linoleic (LA) and α-linolenic acids (ALA), both of which are absolute essential fatty acids for humans, increased with time after birth (Supplementary Figure S1).
A group of lipids that is increasingly recognised as being important in signalling is the phosphatidylserines (PS). Additional multiple stage mass spectrometry (MSn) analysis of the isoforms in the milk samples showed that the acyl chains of the major PS components comprised 16, 18, 20 or 22 carbons with a diverse degree of unsaturation from 1 to 6 C=C bonds. Unlike other polar lipids (e.g. phosphatidylcholines; PC) some PSs were detected with highly unsaturated acyl chains, e.g. PS (18:0/22:6) and lysoPS (0:0/22:6), which we have not found to have been reported in studies of polar lipids in other milks using similar analytical techniques.
Figure 1 also shows the development of these two PSs during early lactation, showing that their levels are high in the first 7 days then decline rapidly (for one of the animals, Yuan Yuan, in a close to exponential decay) to a low plateau by about 20 days postpartum. Interestingly, a similar time course was obtained with DHA, which is also present as the unsaturated acyl chain of these two PSs (Fig. 1). The neonate is therefore supplied with this conditionally essential fatty acid either as the free fatty acid (FFA) or as one chain of the predominant PSs.
Lysophosphatidic acid, which also has several direct biological activities19, showed a distinct exponential decline with time after birth in one animal (Yuan Yuan; Supplementary Figure S1), but there was only slight evidence of this for Li Li (for whom the sampling period was short), and none for Xiao Yatou.
Given the observed dramatic changes in the concentrations of key lipids in development and energy provision, we embarked on a broader statistical analysis of the changes in all the lipid classes found, and how these changes may vary among mothers.
Sequential changes and progressive divergence of lipid profiles between individual mothers. Figure 2 shows the score plot of the Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) model generated by comparing the samples collected before (red spots) with after (blue spots) 7 days postpartum. Samples collected from all three mothers before day 7 show similar rates of change with time. Their clustering around zero on the vertical axis indicates that their compositions are similar to one another. In contrast, the wider scatter on the vertical axis after day 7 indicates progressive, rapid, and approximately simultaneous divergence among individual mothers, as noted previously for other milk components10.
Figure 3 shows the distribution of the lipid profiles as an S-plot from the OPLS-DA model. Data points located far out on diagonally opposite wings of the “S” indicate lipids that changed significantly and consistently in relative abundance from colostrum to mature milk. The four lipids at the most extreme positions on the “S” distribution (labelled in Fig. 3) were identified as choline phospholipids but with different acyl chain compositions. These lipids exhibited divergent concentration trends during the lactation period such that PC (16:0/18:1) was initially abundant in colostrum, whereas the (18:2/18:0) (18:1/18:1) isoform only became abundant in mature milk. Plots of the relative concentrations of these two lipids with time show that the changeover was complete by about day 20 (Fig. 1).
Other lipid classes, however, showed either no significant change or a weaker differentiation between colostrum and mature milk than did the PCs. For instance, in Fig. 3 the glycerolipids (indicated in blue) distribute along the vertical axis from 0.4 to −0.8 (low in colostrum but high in mature milk), and PSs (in red) had a slightly overlapping but inverse distribution along the y-axis from −0.3 to 0.85 (high in colostrum but low in mature milk). In both cases, the data points cluster close to the 0 on the horizontal axis, indicative of either low abundance, or, more likely, weaker MS signals than other compounds because they ionise poorly. Interestingly, the time course for PSs was also followed by lysoPSs in being high in colostrum and low in mature milk (Fig. 1).
The level of TAGs dramatically increased with time postpartum, reaching a plateau after about 30 days (Fig. 1). The kinetics of this change was similar for all three giant panda mothers. The increase in TAGs mirrors a cruder observation of a change in fat content with time in that the low density fat layer appearing on milk samples that were centrifuged in the cold is small in colostrum, and increases with time (Supplementary Figure S2).
The founding question to this work was whether the bearing of highly altricial neonates by ursids is mirrored by a lactation strategy that is unusual amongst eutherians. Specifically, is there a prolonged transition period between colostrum and main-phase milk production, and are the constituents of the milk modified to compensate for the early developmental state of the neonates?
In the ursid species that exhibits the highest degree of altriciality, the giant panda, we have here found that there is indeed a prolonged maturation phase in lipid profiles that takes about thirty days to complete. This parallels the time course for other components of giant panda milk, namely proteins and oligosaccharides9, and the overall metabolome10. The changes in lipid profiles reveal a dramatic shift in the functional balance of the lipid content from an early predominance of lipids that are precursors of signalling molecules and that are essential for construction of the central nervous system, to those predominately involved in energy metabolism.
Signalling lipids dominate the lipid profile of the prolonged colostrum period
Mammals are unable to introduce C=C double bonds in fatty acids beyond carbons 9 and 10, which is why the polyunsaturated LA (ω-6; 18:2, 9, 12; 18:2n-6) and ALA (ω-3; 18:3, 9, 12, 15; 18:3n-3) are essential in the diet. The so-called conditionally essential fatty acids are derivatives of these and include DHA, AA, and EPA. While some of these PUFAs are major components of neuronal tissues, they also have central roles in developmental and immune activation as precursors to signalling lipids such as prostaglandins and leukotrienes. The lipid classes that predominate early and then diminish in giant panda lactation include DHA as the free fatty acid and as an acyl chain of a PS and a lysoPS, all of which show similar time courses (Fig. 1). DHA is crucial to visual acuity and neural development and is the most abundant ω-3 fatty acid in the brain and retina of mammals13,20,21,22,23,24,25,26. DHA also modulates the carrier-mediated transport of choline, glycine, and taurine27, the latter being crucial to neuronal development28. Taurine is possibly a dietary-essential compound in bears as it is in some other Carnivora29,30,31. The time courses for the decline of two other PUFAs, AA and EPA, in colostrum are similar to that for DHA, albeit less well defined (Supplementary Figure S1).
The decline in DHA in giant panda milk with time, which is also seen in human colostrum32, may have several explanations. First, that lipid desaturation mechanisms in neonates required to convert LA and ALA may initially be inadequate in highly altricial neonates (although this does not apply to preterm humans33). Second, the substantial requirement by neonatal mammals for DHA may not be sufficiently supplied by the relatively low absolute fat content of colostrum, and elevated levels of the FFA are supplied to compensate32. Third, the intestines of neonatal mammals are susceptible to inflammatory conditions that may be ameliorated by anti-inflammatory lipid mediators such as resolvins derived from DHA (and EPA), potentially explaining lower incidence of intestinal inflammation in breast-fed rather than formula-fed human infants32. At this stage our analysis has not discriminated the anti-inflammatory resolvins and lipoxins from other eicosanoid lipids, although we did find diminishing levels of the pro-inflammatory leukotrienes (Supplementary Figure S1) such as are also found in human milk32.
Other important roles of milk fatty acids include direct anti-microbial activities34 that will be relevant to a newborn suddenly exposed to a microbe-rich environment and whose gut microbiome is naïve and must establish a functional balance35,36. Of particular potential importance to a hairless, altricial neonate such as a giant panda cub is the integrity of the epidermal permeability barrier to avoid transcutaneous water loss. There is a particular role for LA in the regulation of epidermal water-permeability, in the form of special O-acylated ceramides37,38,39,40,41.
PSs comprise important membrane structural components in most cells. They are particularly enriched in the inner leaflet of the plasma membrane in neural tissues where they are key to several signalling pathways, and make up 13–15% of the phospholipids in the human cerebral cortex42,43. Moreover, as with DHA, PS-dependent signalling is directly involved in processes of neuronal differentiation and survival42,43. These roles further emphasise the potential significance of maternal delivery of PS, lysoPS and DHA in the immediate neonatal period (Fig. 1). The PS receptor, and therefore also its ligand, is essential during embryogenesis in mice44, which also produce altricial neonates.
PSs also form protein-phospholipid complexes that initiate calcification during the formation of bone45,46, which again could be particularly relevant to a highly altricial neonate. The lyso forms of PSs are particularly notable for involvement in cellular signalling, including neural development, that could be crucial to development of mammalian embryos and survival of neonates19,47,48. Also potentially crucial to a newborn is that lysophospholipids, especially lysoPSs, are increasingly being found to have potent immunoregulatory properties47,49,50,51,52,53,54.
The other major class of phospholipids in giant panda milk, the PCs, are major components of biological membranes and pulmonary surfactant. Our analysis of giant panda colostrum indicated changes in provision of PCs to the neonate with time (Fig. 1). There was, for instance, an intriguing complementarity in time course over the first 20 days, in which one PC isoform (16:0/18:1) was present early and then diminished, whereas the abundance of another (18:2/18:0)(18:1/18:1) was initially low and then increased (Fig. 1).
Triacylglycerides increasingly appear in the transition to mature milk
Lipids that dominate the profile of early colostrum appear to be those involved in a range of developmental functions. In contrast, TAGs, which are likely to be more involved in provision of energy, appear only slowly with time (Fig. 1). An early deficit in energy supply may be associated with the slight loss of body mass seen in panda cubs (as is common amongst mammal neonates) immediately after birth10, but also that neonatal cubs may not be able to digest TAGs efficiently. The persistence of bile-salt-activated lipase (commonly found in milks and thought to aid digestion of triglycerides55) in giant panda milk supports this suggestion, in that their maternal provision may be necessary until a cub’s own pancreatic production of the enzyme becomes sufficient9.
With regard to energy provision, it is also noteworthy that lactose is present in giant panda milk initially and diminishes rapidly, inversely to the appearance of TAGs (refs 9 and 10 and Fig. 1). A similar, though much slower conversion from sugar- to lipid-based energy supply occurs in marsupials56,57.
The imaginative hypothesis8 explaining why ursids produce such altricial neonates states that mobilisation of fat reserves for gluconeogensis to supply embryos trans-placentally with glucose is metabolically wasteful during hibernation. Birthing at an early stage of development allows a switch to lipid-based energy supply via milk that reduces the metabolic demand on a hibernating mother. However, giant pandas, like several other species of ursid (Andean, sloth and sun bears), do not hibernate, and there is no indication yet that they had ancestors that did so.
It is nevertheless clear from this and our previous studies9,10 that the altriciality of giant panda cubs is indeed reflected in a prolonged maturation from colostrum to main-phase lactation. This period takes about twenty days for lipids that are important in signalling and developmental functions (this study), and about thirty days for TAGs (this study), protein, and oligosaccharide components9,10. Within these transitions, there are intriguing subtleties in the changeovers between compounds, such as for the two PC isoforms mentioned above, and oligosaccharides in which early- and late-appearing forms differ merely in their glycosidic linkages9.
Artificial milk replacers cannot mimic the composition and dynamics of neonatal ursine requirements
Whilst our main interest is in lactation biology of ursids as a whole, this study, together with our previous work9,10, highlights the dramatic, time-dependent changes in the lipid content of giant panda milk. The complexity of biochemical composition in these milks emphasises the inadequacy of artificial milk formulae that are used to supplement or replace giant panda milk. As shown in Supplementary Figure S3, none of the three milk replacers used for giant panda cubs (including one recently developed specifically for giant pandas58) provide any of the lipids that predominate for the first twenty days in maternal milk. This deficiency is particularly evident for key PUFAs such as DHA. The analysis presented here and in our proteomic and metabolomic analyses of giant panda milks9,10 indicate how critical the biochemical constitution and dynamic of milks are for optimal neonatal development of organ systems (the central nervous system in particular) and the immune system, and for direct maternal immune protection of the neonate.
These observations draw into question the practice of feeding artificial milk formulae as a matter of routine, and the extent to which such a practice compromises particularly the neurological and immune development of giant panda cubs. Moreover, our previous studies have highlighted the risk that these artificial formulae pose to the health of cubs, such as with their high relative content of lactose9,10. The disparate influence of natural milk and artificial formulae on establishment of an appropriate gut microbiome and avoidance of intestinal disease is also crucial35,36. Taken together, these findings disagree with the simplistic approach for development of panda milk replacers58, and emphasise instead the need to revise husbandry conditions to promote full maternal care of panda cubs, and to eliminate artificial feeding in giant panda breeding centres.
Does ursine colostrum reflect a form of external gestation?
The term “external gestation” was originally applied to marsupials because they produce neonates of extraordinary altriciality following a very short gestation, and exhibit dramatic alterations in milk components with time57,59,60 that may serve to provide nutrients required for early development that eutherians provide in utero.
Specifically with regard to lipids, one of the most dramatic changes in marsupial milks is exhibited by lipocalins, a family of proteins that are widely associated with transport of small lipids61,62,63. Changes in these proteins in marsupial milks may represent a changing requirement for lipid types by the neonate, although there is currently no information on whether these different lipocalins exhibit discrete lipid transportation repertoires. While these marsupial milk-specific lipid carrier proteins do not occur in eutherians (though two forms of a lipid-transporting lipocalin, β-lactoglobulin, do occur in giant panda milk9), the changes in lipids we observed in the milk of giant pandas may indicate a similar role of milk in the provision of developmentally-essential lipids to an altricial neonate that in other species of eutherian are provided pre-partum. In this sense, the term “external gestation” may indeed be applicable to bears.
Materials and Methods
Milk collection and processing
Milk samples were collected from three captive-bred giant pandas at the Chengdu Research Base of Giant Panda Breeding, Chengdu, Sichuan Province, P.R. China, in 2012. Please see Supplementary Table S2 for a complete list of sampling dates, studbook numbers, and reproductive histories. Animals were trained to allow milk sampling during routine health checks, when considered necessary to hand-feed cubs, or for research purposes. Animals were not anaesthetised, sedated, drug-treated, or physically restrained during milk sample collection. Cubs are handled by staff several times a day at the Chengdu facility for monitoring of health and growth, and for supplemental feeding, and were not with their mothers when milk samples were collected. While overt disease was not observed in any of the sampled individuals during the study period, minor ailments and suboptimal blood factor levels were observed in some. Mother pandas normally enter a period of anorexia for 7–10 days postpartum. However, at the Chengdu facility, they are given glucose until normal diet resumes.
Milk samples were stored either in liquid nitrogen or in a −80 °C freezer immediately after collection. They were transferred frozen to Scotland, and stored in the Glasgow laboratories at or below −20 °C until use.
Milk sampling from giant panda mothers and all procedures in this study were carried out under ethical approval from the College of Medical, Veterinary & Life Sciences Ethics Committee of the University of Glasgow, and by the appropriate committee of the Chengdu Research Base of Giant Panda Breeding where the animals were held. All sampling was carried out in strict accordance with relevant guidelines and regulations, and as approved by the above authorities.
Giant pandas are currently classified as “vulnerable” by the International Union for the Conservation of Nature (IUCN Red List 3.1). International transfer of milk samples from captive giant pandas in China to the laboratories in Scotland was covered by the CITES convention through permits issued by both donor and recipient countries. A permit was obtained from the Scottish Executive for the importation of the milk samples into Scotland as veterinary-checked animal products.
Chemicals and standards
High-performance Liquid Chromatography (HPLC) grade acetonitrile and isopropanol were obtained from Fisher Scientific, UK. Ammonium formate was purchased from Sigma-Aldrich, UK. HPLC-grade water was produced by a Direct-Q 3 Ultrapure Water System from Millipore, UK.
Liquid chromatography-high resolution mass spectrometry (LC-HRMS)/multiple tandem HRMS analysis and data processing
The preparation of milk samples and the HRMS settings were described in our previous study10. Briefly, the milk samples were thawed at room temperature, and 50 μl of each were added to 200 μl of methanol/acetonitrile 1:1 (v/v). The solution was mixed and placed in an ultra-sonic bath for 20 seconds. The emulsion was centrifuged for 10 minutes at 15,000 rpm at 4 °C (Eppendorf 5424 R, maximum RCF = 21,130 g). The supernatant was transferred to an HPLC vial for liquid chromatography-mass spectrometry (LC-MS) analysis. The three powdered milk formulae were prepared according to manufacturers’ instructions, and the extraction procedures were as above. In order to improve the separation of lipids, a silica gel column (ACE SIL, 150 × 3 mm, 3 μm, HiChrom, UK) was employed with the mobile phase A as a mixture of water and isopropanol (v/v 8:2) containing 20 mM ammonium formate, and phase B as acetonitrile and isopropanol (v/v 8:2). The liquid chromatographic gradient was programmed as follows: 90% of B from 0 to 5 min, decreasing to 70% at 9 min, 65% at 13 min, 60% at 23 min, 55% 28–30 min, and finally increased back to 90% at 31 min, which was then held until 40 min.
Peak extraction and alignment of the LC-HRMS data were carried out using MZMine 2.10 as previously10. As shown in Table S1, different adduct forms were selected for annotation of different lipid classes and then searched on accurate mass against the database downloaded from the LIPID MAPS Lipidomics Gateway, http://www.lipidmaps.org/. Only putatively identified signals were used for subsequent statistical analysis.
Data-dependent MSn fragmentation scan was carried out with collision-induced dissociation at 35 V using a Surveyor HPLC system combined with a LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific UK) and the chromatographic conditions described above. The MS2 spectra of some of the major lipids are gathered in Supplementary Figure S4, and the interpretations are detailed in Supplementary Panel S1 and summarised in Supplementary Table S1.
SIMCA version 13.0 (Umetrics, Umeå, Sweden) was used for OPLS-DA multivariate analysis. The data were mean centred and Pareto scaled in order to generate an S-plot for visualisation of the components with significant influence in the dataset. The plots of peak areas of individual lipids versus days after birth were generated using Microcal ORIGIN software.
How to cite this article: Zhang, T. et al. Changeover from signalling to energy-provisioning lipids during transition from colostrum to mature milk in the giant panda (Ailuropoda melanoleuca). Sci. Rep. 6, 36141; doi: 10.1038/srep36141 (2016).
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We are grateful to the following at Chengdu Research Base of Giant Panda Breeding: Hairui Wang for supervising the collection and storage of the milk samples, Liang Zhang for administration of export arrangements and collection of veterinary records, and the Director, Professor Zhihe Zhang, for his continued support and encouragement of this project. The analytical work was funded by the University of Glasgow and the University of Strathclyde. The funding of mass spectrometry equipment at the University of Strathclyde was provided by the Scottish Life Sciences Alliance.