To elucidate the role of individual caseins in the aggregation process and in the secretion of the micelle, we wished to characterise the initial casein aggregate formed in the secretory pathway. In the present study, we focused on the putative aggregation events that occur in the ER for several reasons. First, it is well established that secretory proteins concentrate by up to 100-200-fold between the ER and the cis-Golgi [29–31], the ER export machinery being selective in sorting cargo . Second, caseins that contain cysteine residues dimerise via disulphide bonds  and the ER is the presumptive site of disulphide bond formation. Third, we hypothesized that caseins interact in the ER because, in the absence of αS1-casein, β- and κ-casein accumulate in the ER due to a drastic reduction of caseins exiting from that compartment .
Morphological and biochemical controls demonstrated that the isolated rough ER fraction obtained from rat or goat mammary gland tissues was almost pure and hardly contaminated with casein micelles. In addition, focusing our analysis on the immature ER forms of αS1- and β-caseins (in rat only for the latter) allowed us to specifically monitor the caseins within the lumen of the ER. We would have liked to obtain similar information for κ-casein, but this was hampered by the lack of relevant immunological tools for rat κ-casein. Moreover, the immature form of this protein has not yet been identified. However, preliminary experiments in goat suggest that immature κ-casein behaves similarly to immature αS1-casein (data not shown).
It has been reported that amphiphilic bovine β-casein has the ability to self-associate into micelles in vitro [33, 34]. This aggregation into oligomers is spontaneous, reversible and dependent on various parameters including temperature, ionic strength and protein concentration [2, 35–39]. Moreover, mixed associations of β-casein with αS1-casein in conditions close to that of the ER have been obtained in vitro . The observation that only few chains of immature β-casein remained in the microsomal membrane pellet after saponin permeabilization in conservative conditions suggested that this casein was either weakly associated with primary casein aggregates or was not prone to aggregation in the ER. This, in turn, may suggest that β-casein has no specific role in the initial step of casein micelle formation. In agreement with this, Kumar et al.  showed that formation of casein micelles still occurred after knock-out of the β-casein gene in mouse, micelles lacking β-casein only having a smaller diameter than control ones. Similarly, casein micelles were observed in the secretory pathway of MECs from goats that do not express β-casein (β-Cn0/0 goat, ). In both cases, casein transport in the secretory pathway and secretion were apparently not affected. Moreover, it is also well-known that β-casein easily detaches from casein micelles, e.g. upon cooling of milk , whereas interactions between the other caseins are much stronger. In line with this, it should be noted that our permeabilisation experiments were performed at 4°C to avoid proteolysis during incubation. Notably, it has been demonstrated that the phosphorylation of β-casein occurs in a later Golgi compartment than that of αS1-casein [16–18]. Moreover, since bovine non-phosphorylated β-casein was shown to have a different micellisation process than the phosphorylated form , we can hypothesise that the mature and immature forms of rat β-casein have different association properties. The observation that, after permeabilisation in conservative conditions, the proportion of the mature form of the protein in the membrane pellet derived from a PNS was higher than that of the immature form in the microsomal pellet (data not shown) was consistent with this hypothesis. We concluded that rat β-casein might interact efficiently with the other components of the micelle in a more distal compartment of the secretory pathway, may be after its phosphorylation when it meets high calcium concentration in late Golgi cisternae and/or secretory vesicles.
In contrast to β-casein, the majority of immature αS1-casein was found in the membrane pellet after permeabilisation in conservative conditions, in both rat and goat microsomes. Notably, a substantial proportion of this protein remained associated with microsomal membranes after permeabilisation in non-conservative conditions or extraction with carbonate at pH 11. These results demonstrated the presence of soluble (at most 25% in rat), aggregated and membrane-associated forms of αS1-casein in the rough ER microsomes. Consistent with this, in vitro experiments led to the conclusion that αS1-casein can self-associate [43, 44], the two hydrophobic regions of bovine αS1-casein interacting to form a polymeric chain (for review see ). In addition, bovine αS1-casein was also found to prevent κ-casein aggregation and accumulation of κ-casein fibrils [45–47], as well as to reduce β-casein aggregates in vitro , features that might be relevant for casein micelle formation. Also, Bouguyon et al.  demonstrated that mature αS1-casein from rat milk can form an intermolecular disulphide bond. Finally, in vivo experiments in goat revealed that αS1-casein might interact with the other caseins in the ER to facilitate the export of these proteins to the Golgi apparatus . It is therefore tempting to speculate that αS1-casein acts as an escort protein (or "transport chaperone", ) for efficient packaging of the other caseins into ER-derived transport carriers and transport to the Golgi. In the absence of αS1-casein, however, casein micelles still form. Altogether, these data are consistent with the fact that a large proportion of immature αS1-casein is aggregated in the rat rough ER through covalent and non-covalent interactions with itself or with other caseins. We conclude that, in contrast to β-casein, αS1-casein most likely participates in a primary casein aggregate which forms in the ER prior to its transport to the Golgi apparatus. This differential behaviour of β-casein and αS1-casein might play a key role in the spatio-temporal dimension of casein micelle formation within the secretory pathway, delaying formation of big aggregates to the late steps of transport within secretory vesicles.
We found that a substantial proportion of αS1-casein was strongly interacting with the ER membranes. We also detected a membranous form of αS1-casein in pH 11 extracted membranes prepared from rabbit MECs (data not shown). Moreover, we found αS1-casein in a proteomic analysis of rough ER microsomal membranes prepared from goat (H. Lahouassa, manuscript in preparation). Of note, the proportion of membrane-associated αS1-casein was higher in rat (≈ 15%) than in the other mammals studied so far (≈ 5-10%). However, the latter proportion is similar to those found for other secretory proteins known to interact with membranes (see below). In rat, the structural domain of αS1-casein involved in this interaction might therefore have a higher affinity for its binding site. Consistent with our observation, an in vitro study with liposomes proved interactions between bovine milk caseins (κ-casein oligomers and early globular casein aggregates) and membrane phospholipids .
Membranous forms of secretory proteins have been observed in several instances, including hormones (for review see), prohormone convertases and processing enzymes [51, 52] and members of the granin (chromogranins, secretogranins) family of regulated secretory proteins [53, 54] that are ubiquitously found in secretory granules of neuroendocrine and neuronal cells (for review see ). Concerning the latter proteins, it was proposed that the membranous form of these proteins is a "nucleus" for granin aggregation in the trans-Golgi network (TGN), a process required for targeting of these proteins to secretory granules. The membrane-associated forms of the granins can therefore be considered sorting receptors. In the context of casein micelle formation, we also hypothesize that the membranous form of αS1-casein acts as a "nucleus" for casein association/aggregation in the ER for further targeting of the other caseins to the site of COP II vesicle formation. Casein micelles are often found attached to the membranes of secretory vesicles through electron dense proteinaceous material most likely corresponding to condensates of casein molecules (see figures in [19, 20, 40]). Consistent with this hypothesis, spherical particles are also seen in close apposition to the saccular membranes in the Golgi apparatus.
We observed that dimerisation played a key role in the interaction properties of αS1-casein with rough ER microsomal membranes. In line with this, we showed that rat caseins are actually dimeric in rough ER microsomes, as expected since the ER is the presumptive site for disulphide bond formation. From this, we can conclude that dimerisation is the first step in casein micelle formation. For, chromogranin B, a disulphide-bonded loop in its N-terminal sequence was shown to play a key role at the level of the TGN in its sorting to secretory granules . Subsequently, it has been demonstrated that the disulphide-bonded loop mediates homodimerisation of chromogranin A , and most likely B, as well as their association with membranes . Moreover, their data strongly suggest that when the protein aggregates in the TGN, this cargo with multiple loops on its surface has a high membrane binding capacity, a feature important for efficiency of sorting to secretory granules. Our experiments using total membranes from PNS revealed the existence of a membranous form of mature αS1-casein and confirmed the role of disulphide bonds in the association of the protein with these membranes. This implies that αS1-casein is able to also interact with membranes of downstream compartments of the secretory pathway. Our observation of a higher proportion of membrane-associated αS1-casein in PNS also suggests that phosphorylation of the protein and/or lipid composition of the membranes might be involved in membrane interaction.