Albumin transcytosis across the epithelium of the lactating mouse mammary gland

Abstract

Murine milk contains 18 mg ml⁻¹ serum albumin, a concentration equal to that in the serum of the lactating mouse. We examined cellular transport using in vivo methods in the mouse. At steady state the specific activity of ¹²⁵I-albumin injected into the blood stream was equal in plasma and whey, confirming that milk albumin is extra-mammary in origin. Fluorescent albumin crossed the gland from basolateral surface to lumen via cytoplasmic vesicles, but was not transported in the apical to basal direction. Albumin was segregated from transferrin at the basal surface of the epithelial cells and did not colocalize with either caveolin-1 or -2. Vesicular transport was not disrupted by filipin providing additional evidence that, unlike the vascular endothelium, caveoli are not involved. Cytoplasmic albumin was localized to vesicles containing IgA and transport was disrupted by agents that interfere with clathrin-mediated endocytosis. Together, these findings provide evidence that albumin is transported across the mammary epithelium by the same pathway as immunoglobulin. The possibility that the massive transfer of albumin into mouse milk is mediated by fluid phase transport is considered.

Albumin is a multifunctional transporter that binds and carries small molecules from one organ of the body to another, probably even transporting them across the capillary endothelium to deliver substances such as fatty acids, phospholipids, steroid hormones, thyroid hormone, divalent cations such as Ca²⁺, Cu²⁺ and Zn²⁺, and many other substances directly to the cells which utilize them. Regulated transcytosis of albumin is now accepted to play a central role in the selective barrier function of the endothelium and the mechanisms are under study in many laboratories (Minshall et al. 2002; Kim & Malik, 2003). Early physiological studies of albumin transport across capillaries had led to the conclusion that a system of small and large pores allowed albumin flux by diffusion and convection (Pappenheimer et al. 1951). However, Ghitescu et al. (1986) demonstrated gold-labelled albumin binding in uncoated pits on the capillary endothelium with subsequent release into the abluminal space and argued for a receptor-mediated transcytotic pathway. Subsequently two small glycoproteins, termed gp18 and gp30, were identified as albumin binding proteins (Ghitescu et al. 1986; Schnitzer et al. 1992; Schnitzer & Bravo, 1993) as well as a larger albumin binding protein, gp60 (Schnitzer & Oh, 1994), that was expressed only in continuous endothelia. It now seems likely that gp18 and gp30 serve as scavenger receptors to mediate albumin catabolism (Tuma & Hubbard, 2003), whereas gp60 is involved in transcytosis.

Although vesicular albumin transport across elements of the mesothelium seems clear (Vogel et al. 2001; Bodega et al. 2002), the transfer of albumin across epithelial barriers was long thought to represent non-specific transfer of protein through the paracellular pathway. However, in recent studies Malik and co-workers (Kim & Malik, 2003) identified a luminal-to-interstitial space pathway in the lung that is abolished both by crosslinked antibodies to gp60 and by filipin, a drug that rapidly disrupts caveoli. In the mammary gland the presence of albumin in milk, when observed, was taken as evidence that mammary tight junctions were open and that the paracellular pathway was freely permeable even to large proteins (Schanbacher & Smith, 1975; Grigor et al. 1991). However, evidence that mouse milk contains a concentration of albumin approximately equal to that of plasma (Halsey et al. 1982) together with data from this and other laboratories indicating that the tight junctions of the mammary epithelium are impermeable during lactation (Linzell & Peaker, 1974; Berga, 1984; Nguyen et al. 2001) prompted us to examine the proposition that albumin is transported by transcytosis across the mammary epithelium into mouse milk. Because no in vitro mammary epithelial system currently exists that transcytoses albumin, it was necessary to devise techniques to study albumin transfer in vivo. We used these techniques to re-evaluate the pathway by which the enormous amount of albumin in mouse milk is transferred across the epithelium and to test the hypothesis that this transport is mediated by caveolar uptake at the basal surface of the cell. Our results show that albumin is indeed transcytosed across the mammary epithelial cells, and the pathway appears to utilize the well-studied IgA transcytotic pathway (Kraehenbuhl & Hunziker, 1998) providing evidence for a novel mechanism for albumin transcytosis across an epithelial layer.

Methods

Materials

¹²⁵I sodium iodide was obtained from ICN; Iodobeads, NHS-fluorescein, NHS-rhodamine, DNP-X-SE, from Pierce; rabbit anti-mouse albumin (for ELISA) and mouse albumin protein standards from Accurate Chemical & Scientific Corp.; fluorescent secondary antibodies and normal donkey serum from Jackson Immunoresearch; JB-4 resin components from Polysciences, Inc.; mice from Charles River laboratories; Cell-Tak from Collaborative Biomedical; slides (Superfrost/Plus) and coverslips (No. 1) from Fisher Scientific; Hybond C, nitrocellulose membrane from Amersham; rabbit anti-mouse albumin (for ELISA capture, and immunostaining) and goat anti-mouse IgA(α-chain) from ICN/Cappell; rabbit anti-caveolin-1(N-20), goat anti-clathrin HC(C-20) and goat anti-caveolin-2(N-20) from Santa Cruz Biotechnology; rabbit anti-dinitrophenyl-KLH, rabbit anti-fluorescein and Alexa 488 conjugate, from Molecular Probes. Rabbit anti-mouse β-casein (AB # 7781) was prepared in this laboratory. All other chemicals were from Sigma Chemical Company.

Mice

CD1 mice were housed in the University of Colorado Health Sciences Center USDA-approved animal quarters. All mice were anaesthetized with sodium pentobarbital (0.1 mg (g body weight)⁻¹) prior to killing by cervical dislocation or perfusion with saline and fixative (see Methods below). Day 1 of lactation was designated the day the pups were born. At birth, litters were standardized to 10 pups. All experiments were performed in mid-lactation (between days 7 and 14) with most performed on day 10. The Institutional Animal Care and Use Committee of the University of Colorado Health Sciences Center approved all procedures which follow the guidelines of the United States Department of Agriculture.

Albumin transport from bloodstream

Mouse albumin (Sigma, fraction V) was labelled with radioactive iodine (Na¹²⁵I, ICN) using the Pierce iodobead kit. ¹²⁵I-albumin was extensively dialysed against PBS and used for experiments within 24 h. Three mice at a time were injected with 100 µl ¹²⁵I-albumin in Ringer solution (1 mg ml⁻¹, 0.18–0.27 mCi mg⁻¹). Timing of the experiment was based on a study of low-density lipoprotein (LDL) transport across the mammary epithelium wherein 7 h were required for steady state levels of labelled LDL, transported from the blood stream, to be achieved (Monks et al. 2001). From t = 5–8 h after injection the mice were anaesthetized and the right mammary gland exposed and incubated in situ with 2 ml of Ringer solution as described for the interstitial space preparation (ISSP) below. Samples of blood, serum, whey and mammary gland were collected at 8 h, proteins were precipitated from these samples, as well as the initial substrate and the ISSP solution with 10% trichloroacetic acid and the precipitates were subjected to gamma counting. Additionally, the concentration of albumin in samples of serum, whey, ISSP, homogenized mammary gland and liver, was determined by enzyme-linked immunosorbent assays (ELISA) for mouse albumin (Igarashi et al. 1993). ELISA data and gamma counts were used to determine specific activities. Additional samples from these tissue fractions were separated by SDS-PAGE, immunoblotted for endogenous mouse albumin, and exposed to X-ray film to visualize the ¹²⁵I-albumin probe.

Interstitial space preparation

The interstitial space preparation (ISSP) was developed to gain experimental access to the interstitial fluid of the lactating mouse mammary gland. A female mouse at mid-lactation was anaesthetized with sodium pentobarbital (0.1 mg (g body weight)⁻¹ with 1 mg boosts as necessary). Upon loss of toe-pinch reflex, the mouse was laid on her back, a midline incision was made on the abdomen, and the abdominal skin was pulled back to reveal the fourth and fifth mammary glands attached to the skin. The skin was cupped and pinned to allow overlay of 1–2 ml of Ringer phosphate buffer with 0.5% fetal calf serum, and addition of probe(s) of interest on the exposed glands as necessary. After 2–4 h of incubation, during which the mouse was kept warm and the open abdomen moist, the bath solution was collected and the exposed tissue rinsed quickly with buffer (+ FCS). The mouse was then subjected to intracardiac perfusion with Ringer solution at a rate of 6 ml min⁻¹ for 5 min, followed by 2%, then 4% paraformaldehyde in Ringer solution, pH 7.4 (4°C, 30 ml each). The exposed mammary tissue (and control gland) were then dissected and processed for embedding in JB-4 resin (for fluorescently labelled, fixable probes) or frozen (for immunofluorescent staining).

For JB-4 embedding, the tissue was further immersion fixed for 4 h in 4% paraformaldehyde, then washed in 50 mm NH₄Cl, 0.1 m phosphate pH 7.0, first with, then without 2% sucrose. The tissue was dehydrated in 75% and 100% acetone, then infused with JB-4 component A, then A + component C, and then embedded (Beckstad, 1985). For frozen sections, the dissected tissue was sandwiched between aluminium foil (to prevent organic extraction) and quickly frozen in isopentane (2-methyl-butane) cooled to freezing with liquid nitrogen. Frozen tissue was then stored at −70°C up to several months, to await sectioning. Freshly sectioned tissue was subsequently vapour-fixed with paraformaldehyde for 15 min, and either mounted or processed further.

Introduction of fluorescent probes into the mammary duct or interstitial space

Fluorescein was conjugated to mouse albumin (fraction V) using the NHS-FITC kit according to manufacturer's instructions. Similarly, mouse apo-transferrin was rhodamine-labelled using NHS-TRITC. Probes at a concentration of about 1 mg ml⁻¹ were extensively dialysed against Ringer solution, sterifiltered and stored at 4°C until use. For intraductal injection one lactating mouse was removed from her pups and anaesthetized with Avertin (0.8 mg (g body weight)⁻¹, with 10 mg boosts as necessary). Using a pulled and fire-polished capillary pipette (o.d. ∼70 µm (Nguyen et al. 2000) 80 µl of FITC-labelled albumin was injected up one teat. The mouse was allowed to recover. After 2 h the mouse was anaesthetized with pentobarbital (0.1 mg (g body weight)⁻¹), perfused as above, and the mammary gland dissected and processed. For the ISSP, a 1 mg probe was added to 2 ml of ISSP bath solution. After a 2 h incubation the mouse was perfused as described above, and the gland processed for embedding in JB-4 and/or freezing.

Immunostaining

Ten-micrometre sections were cut from frozen blocks on a Damon/IEC division minotome set at −18 to −20°C. Sections were collected onto Cell-Tak-coated coverslips and vapour fixed with paraformaldehyde for 15 min. Sections were never allowed to dry. Tissue was permeabilized with 1% Triton X-100 for 15 min, rinsed with PBS and blocked with 10% normal donkey serum. Primary antibodies were diluted in PBS to achieve the following concentrations: 40 µg ml⁻¹ rabbit anti-mouse albumin, 20 µg ml⁻¹ goat anti-mouse IgA, 20 µg ml⁻¹ rabbit anti-caveolin-1, 20 µg ml⁻¹ goat anti-caveolin-2, 10 µg ml⁻¹ goat anti-clathrin, and incubated with frozen sections for 1 h, followed by extensive washes in PBS. Fluorescein-, Cy3- or Cy5-labelled secondary antibodies, all made in donkey and cross-adsorbed against mouse serum, were diluted according to manufacturer's instructions and combined with 0.6 µg ml⁻¹ DAPI. After a 1 h incubation with secondary antibodies, the sections were rinsed briefly and allowed to soak overnight in PBS. Coverslips were mounted with anti-fade mounting medium (20 mg ml⁻¹o-phenylenediamine in TBS, pH 8.5, diluted 1: 10 with glycerol), sealed with clear nail polish and stored at −20°C until imaging.

Because the primary antibodies for albumin and casein were both made in rabbit, they needed to be haptenylated for co-immunostaining. Briefly, rabbit anti-mouse β-casein IgG was precipitated with caprylic acid and labelled with DNP-X-SE according to manufacturer's instructions. Rabbit anti-mouse albumin was labelled with NHS-FITC and rabbit anti-dinitrophenyl-KLH was labelled with NHS-TRITC. After labelling, antibody dilutions were retitred. Rabbit anti-fluorescein, Alexa 488 conjugate was used to enhance the FITC-rabbit anti-mouse albumin stain.

Inhibition of albumin transcytosis

Endocytic inhibitors were used in the ISSP at the following concentrations: filipin (500 µg ml⁻¹), chlorpromazine (500 µg ml⁻¹), N-ethylmaleimide (NEM, 5 mm), or an equivalent volume of dimethylsulfoxide (DMSO) carrier alone. After 15 min pre-incubation with bath solution containing the appropriate inhibitor, FITC-labelled albumin and TRITC-labelled transferrin were added in fresh bath solution with inhibitor. After 1.5 h the gland was rinsed with unlabelled bath solution for 5 min, the mouse was perfused as previously described and the tissue was frozen for sectioning. We chose this time period because it was sufficient to observe labelled probe in the lumen of surface alveoli, implying that vesicles within the cytoplasm were labelled at steady state.

Fluorescence imaging methods

Images were collected, processed and analysed using SlideBook software (Intelligent Imaging Innovations, Inc.) on a Nikon Diaphot TMD microscope equipped for fluorescence with a Xenon lamp and filter wheels (Sutter Instruments), fluorescent filters (Chroma), cooled CCD camera (Cooke) and stepper motor (Intelligent Imaging Innovations, Inc.). Three adjacent z-sections were collected and deconvolved using a Nearest Neighbours algorithm (Agard, 1984). Multi-fluor images were merged and renormalized. Positive staining was defined as twice tissue background. Vesicles were hand-delineated for counting and quantification of colocalization.

Results

Albumin transfer from blood to milk

The albumin in serum from non-pregnant, non-lactating mice, 10 day lactating mice, as well as in their milk was quantified by ELISA. The results (Table 1) confirm earlier data of Halsey et al. (1982) and show, not only that the ratio of the albumin concentration in milk to that in serum is indistinguishable from unity, but also that the concentration of albumin in the serum is reduced by 50% or more during lactation.

Table 1.

Albumin distribution in mouse milk

	Halsey et al. (1982)		Present experiment
Reproductive stage	Serum	Whey	Serum	Whey
NPNLa	47.1 ± 2.8	—	45.2 ± 1.8	—
Lactating
Day 4	23.8 ± 1.4	21.0 ± 2.8	—	—
Day 8	23.9 ± 2.9	17.7 ± 1.9
Day 10	—	—	17.3 ± 1.9	18.0 ± 1.2
	Whey-serum ratio		Whey-serum ratio
Day 4	0.88 ± 0.19	—
Day 8	0.74 ± 0.22	—
Day 10	—	1.04 ± 0.12

Data are albumin concentrations (mg ml⁻¹) presented as means ±S.E.M.

^aNon-pregnant, non-lactating mice.

To be certain that albumin is not synthesized in the mammary gland, we injected ¹²⁵I-albumin via the tail vein and examined total and radioactive albumin in plasma and milk after a period of 7 h. We have previously shown that iodinated LDL requires 7 h to equilibrate with milk, probably because of dilution with residual milk in the gland (Monks et al. 2001), and we based the timing of this experiment on those results. Figure 1A shows that plasma radioactivity falls rapidly for 2 h thereafter declining at a rate of about 10% per hour. The ratio of albumin counts in whey to serum was 1.18 ± 0.42 at the end of the experiment, not significantly different from 1.0 or the ratio of endogenous albumin in Table 1. These results are in accord with the conclusion of Halsey et al. (1982) that all the albumin in murine milk is transported from the blood stream.

Figure 1. Albumin in mouse milk and whey.

A, transport of ¹²⁵I-labelled murine albumin from the bloodstream into the milk of lactating mice. Counts in 10 µl samples of blood, serum or whey were normalized for injected counts. Based on the time course in Monks et al. (2001), pups were removed at the time of injection of radioactivity and replaced 1 h later to nurse for 2 h to empty the glands of as much residual milk as possible. Milk then accumulated for 4 additional hours while hourly blood samples were taken and the ISSP was carried as described in Methods to assess the forms of albumin in the interstitial space. The gland was milked at 7 h and blood taken by heart stick for autoradiography and Western analysis. Data from 6 animals are shown as mean ± s.e.m. Where error bars are not apparent they are within the symbol. B, autoradiograph of ¹²⁵I-serum albumin in serum, overlaid Ringer solution (ISSP), mammary gland (MG) and whey on reducing and non-reducing gels. C, Western blot of similar gel with anti-mouse albumin serum. For B and C fluids were assayed from 3 of the mice used for A.

To examine the forms of albumin present in the various compartments of the mammary gland, we examined samples from injected mice by both autoradiography and Western blot. The autoradiograph in Fig. 1B shows that the injected substrate contained substantial amounts of dimerized and oligomerized albumin as well as the 60 kDa monomeric form. These forms are visible in both serum and whey from the injected mice at the end of the experiment. Additionally, the Western blot shows that the endogenous albumin in both serum and whey contains substantial dimerized albumin. Both endogenous and injected albumin pass through the capillary endothelium to the interstitial fluid as represented in the overlaid fluid (labelled ISSP in the plot) and are present in mammary gland homogenates. The high molecular weight forms of albumin, which appear to be more prevalent in the whey than the serum, disappear when the samples are run on a reducing gel suggesting linkage by disulphide bonds. Breakdown products of endogenous albumin are also seen on the Western blot, most prominently in the mammary gland fraction. Together these observations suggest that albumin may be concentrated in a compartment inside the mammary cells, within which additional dimerization occurs, and that a small amount of albumin is degraded in the epithelium.

Albumin transport across the mammary epithelium in vivo

To verify that a paracellular transport pathway does not account for the albumin in milk, we examined passage of albumin from the lumen to the interstitial space, injecting FITC-labelled murine albumin into the mammary duct. Figure 2A shows the localization of FITC-labelled mouse albumin 2 h later. Stain was confined to the lumen, absent from the interstitial space and from intracellular vesicles. In another experiment, FITC-labelled mouse albumin was added to Ringer solution bathing the in vivo mammary gland using the interstitial space preparation (see Methods; Monks et al. 2001). After 4 h, the lumina were heavily stained (Fig. 2B). Consistent with earlier work in this laboratory (Nguyen et al. 2001), the results of this experiment indicate that albumin is excluded from the paracellular pathway and that albumin is transported preferentially in the basal-to-apical direction via a transcytotic pathway. Similar results were obtained when Texas Red-labelled dextran and horseradish peroxidase were added to the interstitial medium (see Fig. 1, Supplementary Material) suggesting the presence of a fluid phase transport pathway.

Figure 2. Unidirectional transfer of albumin across the epithelium of the lactating mammary gland.

A, FITC-labelled albumin was injected into the mammary lumen via the mammary duct (Nguyen et al. 2000). After 2 h the gland was prepared for sectioning and visualized by fluorescence microscopy as described in Methods. Stain is entirely confined to the lumina (L). Interstitial space (arrows) has no dye. B, the gland was incubated with FITC-labelled albumin in the interstitial space for 4 h. Dye can be visualized in the interstitial space (arrows) surrounding the alveoli as well as in the alveolar lumina. White bar represents 50 µm.

Comparison of albumin and transferrin uptake by the mammary epithelium

In epithelial cells in culture, where the process has been studied in detail, both transferrin and the transcytosed protein IgA are taken up by a clathrin-coated endocytic vesicles and transported to a basal early endosome. There transferrin is sorted for recycling to the basal surface, fluid phase substrates are sorted to the lysosome for degradation, and the IgA is transcytosed to be secreted at the apical surface (Apodaca et al. 1994; Tuma & Hubbard, 2003). Figure 3 shows low and high power views of sections of glands incubated for 2 h with both FITC-labelled murine albumin and rhodamine-labelled murine transferrin using the ISSP technique in vivo. At low magnification (Fig. 3A) it is clear that albumin and transferrin are differentially localized. While substantial amounts of albumin are present in the lumen of each alveolus, transferrin is seen primarily near the basolateral surface of the epithelium. Even at high magnification (Fig. 3B) surprisingly little albumin is colocalized with transferrin, although occasional cytoplasmic vesicles (arrows) containing both probes can be seen in regions away from the basal membranes. At the basal surface of the cell where one would expect the early endosomes, transferrin and albumin show a patchy distribution with very little overlap (asterisk). The putative early endosomes and recycling compartments containing transferrin are visible just inside the basal membrane, but they appear to exclude albumin, which shows its own patchy distribution in the same basal region. When the basal membrane is viewed en face as shown in Fig. 3C, it appears to possess localized microdomains, which bind either albumin or transferrin, indicating sorting right at the basal surface. Whether this sorting occurs at the plasma membrane or just inside it as predicted by results in cultured cells, cannot be told at this magnification. However, most of the transferrin-containing compartments just inside the membrane do not contain albumin as seen in the cross-sectional view at the upper left of this figure. The limited overlap that can be discerned may represent the early endosome prior to sorting of albumin and transferrin or may represent separate vesicles below the resolution of the light microscope, even with digital deconvolution. In either case a large amount of FITC-labelled albumin is arrayed near the basal surface of the cell. The indentation labelled ‘m’ in Fig. 3Bc, and enlarged in the inset, represents the process of a myopithelial cell, a structure which lies within the basal lamina (dotted line in inset). Both transferrin and albumin localize around this structure suggesting that they are bound to or internalized at the basal membrane of the epithelial cell and not simply adsorbed to the basal lamina.

Figure 3. Localization of FITC-labelled mouse albumin and rhodamine (TRITC)-labelled mouse transferrin in the mammary epithelium after in vivo incubation.

The interstitial space preparation was used to expose the in vivo mammary gland to the labelled probes for 2 h. The gland was then fixed, embedded in plastic, sectioned, imaged and deconvolved as described. In all figures the small letter a shows the original black and white images of FITC-labelled albumin, b shows the same area with filters set for TRITC-labelled transferrin, and c shows the two images with pseudocolour and merged with DAPI to demarcate the nuclei. A, low power image of several alveoli. Bar is 50 µm. B, high power images from the same section to show details of dye at the basal surface of the cells and punctate stain within the alveolar cells. Arrowheads, structures with albumin stain only; double arrowheads, transferrin stain only; arrows, structures displaying both stains; asterisk, domain structure at basal surface of the cell; m, myoepithelial cell; this cell binds little albumin or transferrin to its surface. Bar is 50 µm. Inset in Bc is a higher magnification of the region marked m; dotted line depicts the approximate position of the basal lamina. Significant albumin stain can be observed in the walls of the capillary at the left hand side of the image. C, en face view of FITC-labelled albumin and rhodamine–transferrin bound at the basal surface of a mammary epithelial cell. Each of the probes appears to have its own domain at this surface. Bar is 50 µm.

Comparison of albumin and casein localization

To determine whether transcytosis of albumin involves the secretory compartment, sections of a mammary gland from a lactating mouse were stained for endogenous albumin and beta-casein (Fig. 4). Endogenous albumin shows a distribution similar to that of the exogenous albumin shown in Figs 2 and 3, with heavy staining at the basal border, intracellular vesicles and a capillary (C). The discontinuities of stain at the basal surface are less apparent, in part because the frozen sections necessary for antibody interactions with cellular protein are thicker than the plastic sections used to examine fluorescent probes applied from the interstitial space. Dense stain for casein is present in large irregular compartments presumably representing the expanded trans-Golgi sacs of the lactating mammary cell (Clermont et al. 1993). Fainter stain is seen in more diffuse regions throughout the cell presumably representing the rough endoplasmic reticulum. There is insignificant (less than 4% using the masking programs of SlideBook) overlap between casein and albumin, indicating that the transcytotic pathway through which albumin traverses the cell is entirely distinct from the secretory compartment responsible for secretion of casein and other milk proteins into the lumen.

Figure 4. Immunostaining for endogenous albumin and beta-casein.

A frozen section of a gland from a lactating mouse was immunostained as described in Methods for murine albumin and beta-casein. Significant punctate staining for albumin is seen in the cytoplasm (green spots denoted by arrowheads). Regions of heavy stain for casein (arrows, red) presumably represent the swollen Golgi compartment characteristic of the lactating mammary gland. There is no discernable overlap between the albumin and casein compartments. The bar is 50 µm. L, lumen; I, interstitial space; C, capillary showing heavy albumin stain in the endothelial wall. The image was acquired (Leitz-Wetzler 100x Plan; N.A. 1.25) and deconvolved as described in Methods.

Are caveolar vesicles responsible for albumin transport?

We explored the distribution of caveolin in mammary epithelial cells, immunostaining frozen sections of mammary gland from a lactating mouse for caveolins 1 and 2. Anti-caveolin-1 brightly stains myoepithelial cells (arrows, Fig. 5A and B), capillary endothelium and interstitial adipocytes (data not shown), all of which have morphologically distinct membrane caveolae in electron micrographs (Supplementary Material, Fig. 2; Elias et al. 1973). Mammary alveolar cells themselves do not possess morphologically distinct caveolae, but they do express both caveolin-1 and caveolin-2 (Fig. 5A). Both proteins appear to be present near or on basolateral plasma membranes and in punctate structures in the apical cytoplasm with diffuse staining over the apical region of the cell. While both proteins appear to be present on the basolateral membrane, in the merged diagram green and red regions of stain are largely distinct at the membrane as well as throughout the cytoplasm. To obtain a more quantitative picture of colocalization of the two caveolins in vesicular structures, we used the masking function of SlideBook to define the intensity and volume of Cy3- and FITC-labelled regions, selecting regions between 10 and 75 pixels in volume that had intensities greater than twice the background. The mean intensities of the selected regions are plotted in Fig. 6A showing the intensities of structures stained for caveolin-1 and caveolin-2. Most intracellular vesicles contain either caveolin-1 (62%) or caveolin-2 (29%) with only about 9% of the vesicles staining for both proteins. These 9% may represent a real subpopulation of vesicles containing both proteins, or they may have been scored as such because two closely spaced vesicles were not resolved at the resolution of these images (∼200 nm).

Figure 5. Caveolin and albumin distribution in mammary cells from a lactating mouse.

A, frozen sections were immunostained as described in Methods. The deconvolved images are shown separately in black and white for the two caveolins and merged and pseudocoloured in the right hand image which also shows DAPI localization to the nucleus. B and C, the 4th mammary gland was overlaid for 2 h with a solution containing FITC-labelled albumin. Frozen sections were fixed and immunostained for caveolin-1 or caveolin-2. Black and white and merged images as in A. D, the 4th mammary gland was overlaid as described in Methods with a solution containing 500 µg ml⁻¹ filipin, FITC-labelled albumin and TRITC-labelled transferrin. Frozen sections of the gland were stained for caveolin-1 and visualized with Cy5 (middle image). The merged image reflects largely albumin and transferrin as the caveolin-1 stain is masked by the brighter fluors in the merged image. For all treatments alveoli near the surface of the gland, which had access to the largest quantity of overlaid solute, were selected for analysis. Then cells sectioned through the nucleus within these alveoli were selected for counting of caveolin- and/or albumin-containing vesicles. Arrows, myoepithelial cells stained with caveolin-1; arrowhead; capillary showing intense albumin staining; I, interstitial space.

Figure 6. Caveolin overlap and effect of inhibitors on the number of albumin-containing vesicles within alveolar cells.

A, the mean intensities of the caveolin-1 and caveolin-2 spots from 5 sections like those in Fig. 5A were determined using SlideBook software and plotted for each spot. With a cut-off of twice the background, spots that express only caveolin-1 are shown in the lower right quadrant; those expressing only caveolin-2 are shown in the upper left quadrant and those expressing both are shown in the upper right quadrant. B, the number of albumin vesicles in sections like Figs 5D and 8 were counted after incubation for 2 h using the interstitial space preparation with FITC-labelled albumin. Vehicle, filipin (500 µg ml⁻¹), chlorpromazine (500 µg ml⁻¹), or N-ethylmaleimide (NEM, 5 mm) were added to the bath as indicated. *P < 0.05; ***P < 0.001 comparing control and treated samples.

To discern whether albumin transport is associated with caveolin-containing vesicles, a gland that had been incubated with FITC-labelled albumin in the ISSP bath was frozen, sectioned and counterstained for caveolin-1 or caveolin-2 (Fig. 5B and C). Little overlap between either caveolin or albumin is visible in the figure. Quantification of the overlap showed that the FITC-labelled albumin fluorescence colocalized with about 7% of the caveolin-1 immunofluorescence and with 1% of the caveolin-2 immunofluorescence providing strong evidence that albumin is not transported in caveolar vesicles.

To confirm this finding we examined the effect of filipin, an agent that has been demonstrated to remove caveolin-1 from membranes resulting in diffuse cytoplasmic staining and morphological flattening of membrane structures (Schnitzer et al. 1994). Addition of filipin to an interstitial space preparation containing FITC-labelled albumin for a 2 h incubation resulted in subtle changes of staining patterns of both caveolin-1 and caveolin-2 (Fig. 5D and not shown). The number of FITC-labelled vesicles within cells selected from surface alveoli on the basis of a large basally located nucleus and clearly evident basal surface was counted. There was no statistical difference between vehicle-treated controls and filipin-treated cells (Fig. 6B). Thus, filipin treatment resulted in no reduction in the number of intracellular vesicles containing FITC-labelled albumin providing additional evidence that caveolin-containing vesicles are not involved in albumin transfer across the mammary epithelium.

Does albumin utilize the same pathway as IgA to cross the mammary epithelium?

The pathway for transfer of IgA across the mammary epithelium via binding to the polymeric immunoglobulin receptor (pIgR) has been well described (Kraehenbuhl & Hunziker, 1998). IgA binds to the pIgR and is internalized via clathrin-coated vesicles followed by sorting of the IgA–pIgR complex for transcytosis in the endosomes. As shown in Fig. 7, albumin appears to be colocalized with IgA in punctate regions within the cell, although the stains are largely separate at the basal surface where IgA appears to be localized closer to or within the membrane and the albumin localized more distally. The domains of the stain appear to be distinct in at least a part of the membrane region. Within the cell the albumin stain appears to overwhelm the IgA stain, as might be expected since about 30 times as much albumin is transported into mouse milk as IgA. Milk IgA is 0.59 ± 0.32 mg ml⁻¹ (Hendrickson et al. 1996) compared with albumin values of about 20 mg ml⁻¹ (Table 1). Nonetheless, a scattergraph showing the intensities of albumin and IgA stain at each of 140 points within the cells shows that the two are colocalized throughout the cell (Fig. 7D).

Figure 7. Immunostaining for endogenous albumin and IgA in cells from a lactating mammary gland.

Frozen sections were stained for albumin using a FITC-labelled secondary antibody (A), or IgA using a Cy3-labelled secondary antibody (C). The merged image is shown in colour with DAPI stain for the nucleus in B. The inset shows a portion of the basal membrane at higher magnification to demonstrate the apparent lack of overlap of IgA and albumin at or near the basal membrane. Bar is 50 µm. D, scattergraph of intensities of IgA and albumin spots showing that albumin and IgA overlap almost completely in the cytoplasmic region. Intensity of albumin- and IgA-containing spots were determined as in Fig. 5.

Pharmacological agents that have been shown to disrupt clathrin-mediated endocytosis were tested for their effects on FITC-labelled albumin uptake from the ISSP. Figure 8 (lower left) shows that clathrin is most heavily localized at the apical and basal surfaces of the cell, but regions of stain can be observed within the cells as well, presumably corresponding to the Golgi compartment. Both NEM and chlorpromazine severely disrupted this distribution, clathrin being removed from the plasma membrane and, apparently, trapped within an endosomal population. The cells from both treatments showed significantly reduced numbers of albumin-containing intracellular vesicles (Fig. 6B). While NEM has been shown to disturb both clathrin-mediated and caveolin-mediated endocytosis (Lamaze & Schmid, 1995), chlorpromazine is considered to be specific to clathrin-mediated endocytosis. These results are consistent with albumin uptake via a clathrin-dependent endocytic pathway as would be expected from its colocalization with IgA.

Figure 8. Effect of chlorpromazine and NEM-treatment on clathrin localization in mammary alveolar cells from glands that had been incubated with vehicle (control) or with chlorpromazine or NEM in the interstitial space preparation.

FITC-labelled albumin was included in the bath; after 2 h tissues were prepared for frozen sections. Sections were stained with an anti-clathrin antibody using a Cy5-labelled secondary. The upper two rows show the localization of albumin, while the lower row shows clathrin staining of the same or similar sections. Note the paucity of cytoplasmic vesicles containing albumin and the relative disorganization of the clathrin stain in the treated cells. Quantification of these and similar images is shown in Fig. 6B.

Discussion

In the first part of this study we confirmed the earlier observation of Halsey et al. (1982) that the concentration of albumin in mouse milk is equal to that of the plasma during lactation. Measurement of the specific activity of ¹²⁵I-albumin in plasma and milk indicates that the total milk albumin originates in the plasma and none is synthesized in the mammary gland. The mammary glands of a lactating CD-1 mouse secrete about 7.5 g of milk per day, equivalent to about 5 g of whey with about 18 mg ml⁻¹ albumin, meaning that the average CD-1 mouse secretes about 9 mg of albumin, approximately equivalent to its total body content of albumin, per day. This amount of secreted albumin is sufficient to induce a state of hypo-albuminaemia in the dam. Both Western blots and autoradiograms show that a significant fraction of albumin is present at higher molecular weight forms in the plasma and that such forms are transported to the milk.

We then examined the pathway by which albumin traverses the mammary epithelium using two newly developed in vivo techniques that allow us to bring fluorescent tracers directly in contact with the apical and basal surfaces of the epithelium. With intraductal injection we confirmed our earlier finding (Nguyen et al. 2001) that tracers, including FITC-labelled albumin, injected into the lumen of the mammary gland during lactation stay confined to this compartment, ruling out the possibility of flux through the paracellular pathway. On the other hand, FITC-labelled albumin and other ‘fluid phase markers’ placed in the interstitial space found their way to the lumen in large quantities and albumin itself was visible as punctate spots within the cytoplasm. Additionally its localization differed from that of Cy3-labelled transferrin, which was mostly visible in a recycling compartment just beneath the basal membrane of the cell, although we have shown elsewhere (Zhang et al. 2000) that a small amount is transcytosed. A total lack of colocalization of endocytosed albumin with casein indicated that the transcytotic pathway does not intersect the very active protein secretory pathway also present in the lactating epithelium.

Our initial hypothesis, borrowed from transcytosis of albumin in the capillary endothelium (Minshall et al. 2002) was that albumin bound to one or more glycoproteins localized to membrane caveoli and was transcytosed in caveolar vesicles. The finding that both caveolin-1 and -2 are present in mammary epithelial cells, albeit at levels much lower than myoepithelial cells or capillaries (Fig. 5), supported this hypothesis. However, the punctate structures in the cytoplasm of the alveolar cells that contain endocytosed albumin or stain for endogenous albumin do not colocalize with either caveolin-1 or -2, and filipin, an antibiotic that disrupts caveoli, does not affect the localization of these structures. These observations are not compatible with caveolar transport of albumin across the mammary epithelium. It should be noted that albumin did colocalize with caveolin in the capillary endothelium of the mammary gland (data not shown) suggesting that, as in other organs (Schnitzer et al. 1994; Vogel et al. 2001), albumin is transported across the capillary endothelium in caveolar vesicles. Caveolin-1 null mice have been generated, and while the mammary glands show premature initiation of lactation during pregnancy apparently because caveolin down-regulates prolactin signalling, no mention was made of abnormalities during lactation (Park et al. 2002).

We found that albumin does colocalize with another transcytotic marker, IgA, in vesicular structures and, consistent with this finding, albumin endocytosis was sensitive to pharmacological agents that disrupt clathrin-mediated endocytosis. Chlorpromazine is a cationic amphiphilic drug, which inhibits receptor-mediated endocytosis by inhibiting the assembly of clathrin and AP-2 at the coated pit. NEM-sensitive factors, soluble ATPases, are involved in assembly of NSF/SNAP/SNARE complexes necessary for all types of vesicle trafficking studied to date (Lamaze & Schmid, 1995; Mukherjee et al. 1997; Clague, 1998). Both chlorpromazine and NEM inhibited albumin uptake by lactating mammary epithelial cells in situ. We conclude that albumin is transported via a clathrin-mediated endocytic pathway. Further most of the albumin avoids being routed to lysosomes and degraded, although the finding that endogenous albumin fragments in the alveolar cell (Fig. 1) may indicate that there is some intracellular degradation.

To get from the blood stream albumin must pass first through the capillary barrier, then the interstitial space, and finally the basement membrane to come into proximity with the basal surface of the epithelial cell. Here albumin probably gives up some of its transported molecules such as fatty acids, trace elements and hydrophobic hormones. Unless it is transcytosed or modified to become a target for degradation, albumin must be transported back to the blood stream either by re-entering the capillary or entering the lymphatic circulation. Interestingly, the capillaries show extremely heavy uptake of FITC-labelled albumin from the interstitial space (Fig. 4) as well as uptake of gold-labelled albumin into vesicular structures in the endothelium (see Supplementary Material, Fig. 2). While luminal to basal transcytosis of albumin has been demonstrated in many endothelial systems (Minshall et al. 2002), our observation suggests that albumin also returns from the interstitial space to the capillary lumen via a transcytotic process. If this is the case, the lymphatic circulation, although known to be extensive in the mammary gland (Lascelles, 1977; Davis et al. 1992), is probably not responsible for returning all the albumin that enters the interstitial space to the plasma.

Albumin, whether measured by uptake of FITC-labelled protein or by immunohistochemistry, is associated in a heavy layer at the basal surface of the cell clearly segregated from transferrin and possibly segregated from IgA as well. A possible explanation for this unexpected observation is that albumin uptake may be dissociated from transferrin recycling at the plasma membrane rather than in the early endosome as in the tissue culture systems where transcytosis has been most extensively defined (Mostov, 1993). However, other explanations are possible. For example, albumin may associate with regions of the basal surface or basement membrane as it transfers fatty acids to the alveolar cell, whereas smaller amounts may be endocytosed with transferrin in another region of the basal surface. The finding that the heavy binding of albumin is apparently dissociated from IgA binding at the basal surface, while it is uniformly associated with IgA transport within the cytoplasm, is consistent with this notion. Unfortunately, the heavy binding of all of these molecules at the basal surface of the cell and the lack of good anti-mouse transferrin antibodies precluded further analysis of this problem using fluorescence. When we attempted to use gold-labelled albumin at the electron microscope level to resolve the issue, we found probe associated with the capillary endothelium and in macrophages, but insignificant amounts were associated with the basal regions of the mammary epithelial cells (see Supplementary Material, Fig. 2). We suspect the reason to be that gold-labelled albumin associates preferentially with scavenger receptors and is not seen by the endocytotic pathway in the alveolar cells themselves.

Many questions remain, the most important of which is whether albumin transport across the mouse mammary epithelium is receptor-mediated or a function of fluid phase endocytosis. Putative receptor-mediated transport of albumin across epithelia has been described in rat jejunum (Kimm et al. 1997), the lung epithelium (Kim & Malik, 2003), and bile canniliculi (Sztul et al. 1983). The last observation is the most interesting because the vesicle mediators of albumin transcytosis from the space of Disse in rat liver to the bile canniliculi have been shown to contain both albumin and IgA bound to pIgR (Barr et al. 1995; Pol et al. 1997); however, in this system much less albumin is transported than IgA, making fluid phase transport more plausible than in the mammary gland.

Unfortunately, we found that most traditional solutes such as non-metabolizable sugars and dextrans used to study fluid-phase transport were rapidly excreted by the kidney so that steady state ratios could not be measured under our experimental conditions. A potential measure of fluid-phase movement across mammary epithelium is given by the transport of monomeric IgA into the milk of J-chain-deficient mice (Hendrickson et al. 1996) where a steady-state milk/serum ratio of 0.21 can be calculated from the data. (The ratio is 4.2 in wild type mice where the polymeric IgA receptor binds the J-chain.) This is only one/fifth of the steady state plasma/milk ratio for albumin of about unity. Further, if albumin transport occurs by fluid phase transfer and the concentration of albumin in the interstitial space is equal to that of the plasma, the implicaton is that a volume of interstitial fluid equal to that of the secreted milk must be engulfed at the basal surface and the proteins concentrated as fluid and ions are removed from the endocytic vesicles during their transfer across the cell. The volumes involved are not impossible, amounting to less than 6% of the cell volume per hour. On the other hand, if albumin is concentrated in the region of the basal surface of the cell by binding, for example, to a fatty acid transporter, then this adsorptive transport could involve much smaller fluid volumes.

What are the implications of this transcytotic pathway for the lactating dam and her pups? It is unclear whether the lactational hypo-albuminaemia seen in mice is an adaptive mechanism to provide specific beneficial metabolic changes or whether it is a direct result of increased clearance of albumin from the system by the mammary gland. The lowered concentration of albumin is probably balanced by an increase in circulating immunoglobulins and apoB-containing lipoproteins (Monks et al. 2001). A similar compensation is seen in humans with hypo-albuminaemia (Maugeais et al. 1997). Similar changes in serum proteins are not seen in lactating females of other species including rats, ruminants or humans, probably because the concentration of albumin in the milks of these species (5, 0.4 and 0.4 mg ml⁻¹, respectively; (Geursen & Grigor, 1987; Lonnerdal & Atkinson, 1995; Swaisgood, 1995) is much lower than in mice.

Mouse pups are born in a very immature state and unlike tricial mammals are unlikely to possess endogenous stores of trace elements and fat-soluble vitamins sufficient for their rapid growth during the suckling period (less than 3 weeks). Under these circumstances albumin may function to deliver such substances directly to the pups without the need for elaborate transport mechanisms for each of these substances. Although the concentration of albumin in human and cow's milk is very much lower, the existence of a similar transcytotic pathway in the human mammary gland could allow the transport into milk of any substance that finds its way into the interstitial space. Such a mechanism could, for example, account for the presence of cow's milk proteins in certain samples of human milk.

Supplementary material

The online version of this paper can be accessed at:

DOI: 10.1113/jphysiol.2004.068403 http://jp.physoc.org/cgi/content/full/jphysiol.2004.068403/DC1 and contains supplementary material consisting of two figures entitled: Uptake of ‘Fluid phase markers’ into the mammary lumen from the interstitial space; and Electron microscope images of gold-labelled albumin after incubation with the ISSP.

This material can also be found at:

http://www.blackwellpublishing.com/products/journals/suppmat/tjp/tjp480/tjp480sm.htm