The effect of serum iron concentration on iron secretion into mouse milk

Abstract

1. The concentration of iron in mouse milk is approximately 3 times that of the serum. Although there is clear evidence for the presence of the transferrin receptor in the rodent mammary gland, the precise mechanisms of iron transfer into milk are not known.

2. Milk iron was linearly related to the serum iron:transferrin ratio in lactating mice whose serum iron ranged from 8 to 66 μm.

3. Increasing the iron binding capacity of the milk by 340 μm by targeting the lactoferrin transgene to the mammary gland did not alter the relation between milk iron and the serum iron:transferrin ratio.

4. The steady-state distribution ratio of ¹²⁵I-transferrin between plasma and milk was about 0.2, indicating that transcytosed transferrin contributed a maximum of 6% of the milk iron.

5. Fluorescently labelled transferrin incubated with the in situ gland localized mainly near the basal surface of the mammary alveolar cells.

6. These experiments provide evidence that the initial and rate-limiting step in the transfer of iron into milk is binding to a basal transferrin receptor.

7. A theoretical model of the relation between milk and serum iron suggests that the affinity of apotransferrin for the basal recycling system may be higher than observed in many other cell types.

The rodent mammary gland has a remarkable capacity for iron transfer from plasma to milk. Early studies showed that the iron concentrations of both rat and mouse milk were considerably higher than the plasma iron, that the source of milk iron was the plasma and that transfer of iron from rat plasma to milk is a rapid process that does not involve the stable mammary gland iron pool (Loh, 1970, 1971; Loh & Kaldor, 1976). Iron is present in the lipid fraction of the milk complexed with the enzyme xanthine oxidase as well as in the aqueous fraction mainly associated with casein (Casey et al. 1995). The concentration of iron in rat milk (5 μg ml⁻¹ or 90 × 10⁻⁶ M) was found to be related to the iron status of the dams (Keen et al. 1980; Anaokar & Garry, 1981; O'Connor et al. 1988; Casey et al. 1995). The finding that the concentration of iron in mouse milk (12–15 μg ml⁻¹ or 0.22-0.25 mM) is at least 3 times that of the plasma (Loh, 1971) and the observation reported here that the milk:serum iron ratio is approximately constant over a nearly 10-fold variation in serum iron prompted us to examine more closely the mechanism of iron transfer into mouse milk.

Under normal circumstances iron is carried in the plasma by transferrin and enters most cells via a transferrin receptor pathway that internalizes plasma transferrin with its bound iron (Klausner et al. 1983; Lok & Loh, 1998). At the low pH of the endosome, iron is removed from the transferrin and transported into the cytoplasm by mechanisms that are just beginning to be understood (Fleming et al. 1998). Transferrin itself is recycled to the surface of the cell to be re-utilized in the transfer of iron through the blood stream. Transferrin receptors or their mRNA have been found in rodent mammary glands (Grigor et al. 1988; Schulman et al. 1989; Sigman & Lönnerdal, 1990a, b), leading several investigators to suggest that the first step in iron transfer to milk is the binding of plasma transferrin to the basal membrane of the mammary alveolar cell. The observation that the mammary transferrin receptor is developmentally regulated (Schulman et al. 1989; Sigman & Lönnerdal, 1990a) is consistent with this idea. However, this interpretation is called into question by several findings. First, the concentration of transferrin receptors in the rat mammary gland was found to be low compared with reticulocytes and other iron-utilizing cells (Grigor et al. 1988). Since the concentration of iron in rodent milk is higher than the plasma iron and iron transfer is rapid (Loh & Kaldor, 1976), a large number of transferrin receptors might be expected in the lactating mammary epithelium. Second, it has been clearly shown that transferrin is transcytosed from the basal surface of mammary alveolar cells into the milk (Seddiki et al. 1992), raising the possibility that an active transcytosis pathway might be responsible for iron transfer into milk. Finally, it is well known that milk calcium depends in large part on the concentration of calcium binding ligands, both casein and citrate (Neville et al. 1994). However, the role of milk iron binding ligands in milk iron secretion has not been investigated experimentally.

In the present study we measured the steady-state distribution of plasma-derived transferrin between the plasma and milk and examined the effect of excess milk levels of the high affinity iron binding protein lactoferrin on the relation between milk and plasma iron. The results of these studies are consistent neither with a large amount of iron transfer via transcytosis nor with a role for iron binding ligands in the iron secretion pathway. In addition, localization of endocytosed fluorescent transferrin to the basal domain of the mammary alveolar cell is consistent with the presence of a transferrin recycling pathway that transfers iron into the alveolar cell of the lactating mammary gland. A theoretical model of this process raises the question of whether a relatively high affinity of apotransferrin for the transferrin receptor is responsible for the nearly linear relation between milk and serum iron.

METHODS

Animals and dietary treatments

All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Colorado Health Sciences Center. CD1 and B6CBA mice were purchased from Charles River Labs (Wilmington, MA, USA). Transgenic mice carrying the human lactoferrin gene under the control of a mammary gland specific promoter (line 1590) were obtained from Pharming Health Products, Inc. (Leyden, The Netherlands). The milk of the 1590 strain had 13.6 ± 1.0 mg ml⁻¹ (170 ± 13 μM) human lactoferrin. All animals were caged in stainless steel cages and maintained in a temperature controlled room with a 12 h light-dark cycle. Female mice, approximately 10 weeks of age, were bred and pregnancy was confirmed by the presence of vaginal plugs. The day the plug was observed was considered as day 1 of pregnancy. A commercial pelleted rodent diet (Agway PROLAB R-M-H-2000, Brighton Feed and Farm Supply Inc., Brighton, CO, USA) containing 18 % protein, 9 % fat, 4 % fibre and 8 % mineral with an iron concentration of 300 mg kg⁻¹ was fed to all mice ad libitum with tap water until day 14 of pregnancy. At this time the pregnant mice were randomly assigned to one of four dietary treatments. One group continued on the commercial diet. The other three groups were switched to specially formulated diets containing 5, 15 and 50 mg Fe (kg diet)⁻¹, respectively, purchased from Research Diets, Inc. (New Brunswick, NJ, USA) and to distilled and deionized water shown by analysis to be iron free. The basal composition of the special diet is listed in Table 1. On day 1 postpartum all litters born to CD1 mice were standardized to 10 pups. The 1590 transgenic mice and their parent B6CBA strain are less prolific; their litters were standardized to 8 pups. All litters from dams on special diets were removed on the day after birth and replaced with 10 CD1 or 8 B6CBA pups, respectively, from dams maintained on the commercial diet. This procedure ensured that any potentially harmful effects of the low iron diet on fetal development did not affect the sucking ability or growth of the neonates. The litters from all transgenic dams were replaced with 8 pups from B6CBA dams to abrogate any possible effect of the transgene in the pups themselves. All published experiments were conducted on day 10 lactating mice; similar but less extensive results were obtained at days 4, 17, 19 and 21 postpartum.

Table 1.

Composition of low-iron diets

Component	Amount (g)
Casein, vitamin free	220
dl-Methionine	2
Corn starch	276
Maltodextrin 10	35
Sucrose	300
Avicel PH-101	20
Corn oil	100
Salt mixture S18703*	35
Vitamin mixture V10001†	10
Choline bitartrate	2

Ferrous sulphate (7H₂O) was added to the diets at the expense of sucrose to vary the total iron concentration to 5, 15 or 50 mg kg⁻¹.

^*Composition of mineral mixture (g (kg diet)⁻¹): calcium phosphate (dibasic), 500; magnesium carbonate, 50; potassium citrate (1H₂O), 220; potassium sulphate, 52; sodium chloride, 74; chromium potassium sulphate (12 H₂O), 0.55; cupric carbonate, 0.3; potassium iodate, 0.01; manganous carbonate, 3.5; sodium selenite, 0.01; zinc carbonate, 1.6; sucrose, 98.03.

^†Composition of vitamin mixture (g (kg diet)⁻¹): vitamin A palmitate (500 000 i.u. g⁻¹), 0.8; vitamin D₃ (100 000 i.u. g⁻¹), 1; vitamin E acetate (500 i.u. g⁻¹), 10; menadione sodium bisulphite (62.5% menadione), 0.08; biotin (1 %), 2; cyanocobalamin (0.1 %), 1; folic acid, 0.2; nicotinic acid, 3.0; calcium pantothenate, 1.6; pyridoxine-HCl, 0.7; riboflavin, 0.6; thiamin HCl, 0.6; sucrose, 978.42.

Milk

Pups were removed from their lactating dams 3 h before milk collection and killed by CO₂ overdose. Dams were anaesthetized with an intraperitoneal injection of pentobarbital (0.65 mg (10 g body wt)⁻¹). Animals were deeply anaesthetized as determined by lack of reaction to hard toe pinch. A dose of 0.5 i.u. of oxytocin was then injected intraperitoneally to stimulate milk let down and milk was harvested by gentle intermittent suction using a specially drawn glass bell that fitted snugly around the teats. Blood was collected as described below while the animals were still anaesthetized. Animals were not allowed to recover from the anaesthetic. Collected milk was stored at −70°C until analysis. To measure the iron content in whole milk, samples were dried at 40°C in an isotemperature oven and then ashed in a low temperature asher (LFE Corporation, Waltham, MA, USA). The iron concentration was analysed on reconstituted samples using an atomic absorption spectrophotometer (Perkin Elmer 2380, Norwalk, CN, USA). Standard solutions were prepared from a stock iron atomic absorption standard solution containing 1000 μg Fe ml⁻¹ (Sigma). Human lactoferrin (hLf) was measured in the milk of the transgenic dams by an enzyme-linked immunosorbent assay (ELISA), using purified recombinant hLf as a standard and antibodies obtained from Pharming Health Products.

Blood

After milk collection a small volume of whole blood was collected from the tail vein of the anaesthetized mouse to measure the haemoglobin concentration. Additional blood was then collected from the brachial artery of the lactating dam to obtain serum for measuring total serum iron and unsaturated iron binding capacity. Animals were maintained under anaesthesia with additional doses of pentobarbital (0.10 mg (10 g body weight)⁻¹) until killed by cervical dislocation. Samples showing any haemolysis were discarded. Haemoglobin was assayed in a spectrophotometer by the cyanmethemoglobin method (Kit 525A, Sigma). Serum total iron concentration ([Fe]_s) and unsaturated iron binding capacity (UIBC) of the lactating dams were measured with the ferrozine colorimetric method (Kit 565, Sigma). The total iron binding capacity (TIBC) was calculated by summing the serum total iron concentration with the UIBC. As transferrin, which binds two molecules of iron per molecule transferrin, is the major iron binding molecule in the serum, the molar serum concentration of transferrin was estimated as one-half the molar serum total iron binding capacity.

Transcytosis of transferrin

Transcytosis of transferrin across the mammary epithelium was studied with exogenous transferrin either labelled directly with ¹²⁵I or conjugated with iodinated tyramine cellobiose (Pittman et al. 1983). The latter compound does not appear to be de-iodinated in the mammary gland and so served as an alternative labelling method to check on the distribution of the directly iodinated compound.

Carrier free Na¹²⁵I was purchased from ICN Biochemicals (Aurora, OH, USA). Murine apotransferrin was obtained from Sigma. Transferrin was directly labelled with ¹²⁵I by the chemical oxidation method utilizing IODO-BEADS from Pierce (Rockford, IL, USA). ¹²⁵I-Tyramine cellobiose was coupled to transferrin by the method of Pittman et al. (1983). Unincorporated ¹²⁵I or ¹²⁵I-tyramine cellobiose was removed by gel filtration with a Sepharose G-50 column (Pharmacia Biotech, Piscataway, NJ, USA) and dialysed, first against 0.1 M ammonium bicarbonate for at least 1 h and then extensively against three changes of Dulbecco's phosphate buffered saline solution. Labelled transferrin was used within 24 h of labelling.

Each mouse was injected intravenously with 200 μl of radiolabelled transferrin (1 mg total transferrin) through the tail vein under gentle manual restraint. A 20 μl sample of blood was collected from the tail vein at 1–2 h intervals using a simple nick with a razor blade. The tail was subjected to gentle pressure until a clot formed. Radioactivity of the blood sample was monitored with a gamma counter. The plasma radioactivity was calculated assuming a haematocrit of 50 %, as determined earlier in lactating mice (P. Zhang, unpublished data). Five mice were injected with directly labelled transferrin and two of them were milked under anaesthesia as above after 7 h; another three were milked after 9 h. Six mice were injected with ¹²⁵I-tyramine cellobiose-conjugated transferrin, three were milked after 7 h and three after 9 h. The radioactivity in whole milk and whey, obtained by centrifuging the milk at 3000 g for 10 min to remove the fat and then 100 000 g for 1 h to remove the casein pellet, was measured. Trichloroacetic acid (TCA; 50 μl of 20 %) was mixed with an equal volume of whey and the TCA precipitable protein in the whey was also monitored for its radioactivity to determine what proportion of the radioactivity actually represented protein-bound iodine. When run on a gel, the iodinated protein all migrated at the molecular weight of native transferrin, indicating that the protein was not degraded during its passage into milk.

Uptake of fluorescently labelled transferrin into the mammary gland

Mouse apotransferrin (1 mg) was labelled with the red dye rhodamine (Pierce, Rockford, IL, USA) according to directions provided by Pierce and dialysed against Krebs-Ringer phosphate buffer solution. Labelled transferrin (300 μg) was added to 1 ml of Ringer solution containing 0.5 % fetal calf serum. A 10 day lactating female mouse was anaesthetized with sodium pentobarbital (0.65 mg (10 g body weight)⁻¹), a midline incision was made in the abdomenal skin, and the skin attached to the fourth mammary gland on one side was pulled aside and pinned to form a cup holding the gland with the surface normally apposed to the peritoneum exposed. The labelled transferrin solution was layered over the exposed in situ gland and incubated for 2 h. Deep anaesthesia (no reaction to toe pinch) was maintained by additional doses of pentobarbital (0.10 mg (10 g body weight)⁻¹) as needed. After removal of the bath solution, the deeply anaesthetized mouse was killed by perfusion with 60 ml of ice-cold Ringer buffer followed by 30 ml of 4 % paraformaldehyde. Pieces of exposed mammary gland and a control gland were dissected and embedded in JB-4 resin (Polyscience, Inc., Warrington, PA, USA). The embedded tissue was sectioned at 2 μm intervals, stained with fluorescein isothiocyanate (FITC) labelled wheat germ agglutinin (Molecular Probes, Inc.) and DAPI to stain luminal polysaccharides and nuclei, respectively, and viewed with a Nikon inverted microscope equipped for epifluorescence. Images were captured using SlideBook software (Intelligent Imaging Innovation, Inc., Denver, CO, USA). Similar results were obtained with several mice.

Statistical analysis

Results are reported as means ±s.e.m. Student's t test was used to examine the statistical significance between groups of samples when needed. A value of P < 0.05 was accepted as statistically significant unless otherwise indicated. Curves were fitted with appropriate equations using the curve-fitting statistical functions of SLIDEWRITE version 3 (Advanced Graphics Software, Inc., Carlsbad, CA, USA) as described in Results and Appendix.

RESULTS

Serum iron status of lactating dams

We used diets with varying iron contents to produce a series of normal and transgenic mice with varying levels of iron in their serum. Starting at day 15 of pregnancy we fed either the normal laboratory diet (designated Lab) or specially formulated diets with controlled iron contents and achieved the mean serum iron levels shown in Fig. 1 where the shaded bars represent data obtained from 24 normal mice (19 strain CD1 and 5 strain B6CBA). Since there was no difference between the results from the two strains the data were combined. In these studies we also used transgenic mice secreting high levels of the iron binding protein human lactoferrin in their milk; data from 11 transgenic mice are represented by the filled bars. As can be seen, the serum iron levels were significantly reduced by the controlled iron diets particularly those containing 5 and 15 p.p.m. iron. The mean serum iron levels in the mice ingesting the 5 p.p.m. diet (27.0 ± 3.4 μmol l⁻¹ in the control strains and 15.1 ± 1.8 μmol l⁻¹ in the transgenic strain) were 2–3 times lower than the serum iron in the mice ingesting the laboratory diet (62.5 ± 3.2 and 44.4 ± 2.9 μmol l⁻¹ in the control and transgenic lines, respectively). Consistent with reports in the literature, there was little difference between the serum iron of the mice ingesting the 50 p.p.m. diet and those ingesting the laboratory feed.

Figure 1. Effect of dietary iron on the iron status of 10 day lactating mice.

All results are shown as means ±s.e.m.□, normal CD-1 and B6CBA mice combined. ▪, transgenic mice with 170 ± 13 μM hLf in their milk. The difference between the normal and transgenic mice is significant at the P < 0.01 level for all diets.

The transgenic strains had significantly (P < 0.01) lower serum iron values when fed on each of the diets than the control strains. The blood of the transgenic mice contains measureable levels of human lactoferrin (P. Zhang & M. C. Neville, unpublished data), possibly changing the iron balance in the transgenic mice. However, as shown below, there were no statistically significant differences between any of the groups of animals in the relation between serum iron and other serum parameters, namely blood haemoglobin, unsaturated iron binding capacity and total iron binding capacity. For this reason, data from all three strains are presented together with different symbols representing the different strains.

The relationship of changes in serum iron to other serum parameters

We adopted a feeding regimen designed to produce low, but stable levels of serum iron that would not seriously compromise the iron status of the dams. As indicated by maternal haemoglobin levels (Fig. 2A) we achieved this goal. The shallow slope of the curve in Fig. 2A shows that the low serum iron had very little effect on the haemoglobin level at 10 days postpartum when there was a mean haemoglobin level of 12.8 ± 0.3 g ml⁻¹ in dams ingesting the 5 p.p.m. diet compared with 14.4 ± 0.2 g ml⁻¹ in dams on the laboratory diet (n = 12 in both groups). This difference was slightly more pronounced later in lactation; at 17 to 21 days postpartum mean values of 11.2 ± 0.2 and 15.0 ± 0.7 g ml⁻¹ were observed in dams ingesting the 5 p.p.m. and the laboratory diets, respectively (N = 6 in both groups). There was no significant difference between the strains of mice at either time point. The haemoglobin level in the low iron group was significantly lower at the later time point (P < 0.03) suggesting that a steady state had not been achieved at 10 days postpartum. Since erythrocytes live to about 40 to 50 days in the rodent (Bartholmey & Sherman, 1985) this finding is not surprising.

Figure 2. Effect of serum iron on haemoglobin (A), unsaturated iron binding capacity (B), total iron binding capacity (C) and the molar ratio of serum iron to serum transferrin (D) in day 10 lactating mice.

In all graphs the circles are normal mice with • representing CD1 mice and ○ representing B6CBA mice; ▾ are transgenic mice. All lines are the best fitting linear regression lines to all the points in the graph. There was no difference between the strains of mice in the relation of any of these parameters to serum iron.

The plasma unsaturated iron binding capacity (UIBC) increased as the serum iron fell and the relation was approximately linear (Fig. 2B). A similar relation was observed in all strains of mice. The sum of the serum iron and the UIBC gives the plasma total iron binding capacity (TIBC). As each molecule of transferrin binds two molecules of iron, the TIBC is twice the serum transferrin concentration. The data in Fig. 2C, where the TIBC is plotted as a function of serum iron for all mice studied at 10 days postpartum, show that the TIBC, and hence the transferrin level, increased slightly as serum iron fell. The mean transferrin concentration for the mice on the laboratory diet was 38.7 ± 1.0 μM increasing to 49.4 ± 1.4 μM on the 5 p.p.m. diet. The molar ratio of serum iron to serum transferrin (Fig. 2D) varied from 0.14 to 1.87, showing that we have achieved a large variation in the serum iron using this dietary regimen. As expected, there is a linear relation between this ratio, the best measure of transferrin saturation, and the serum iron.

The relation between serum and milk iron

In contrast to the blood haemoglobin, milk iron was significantly reduced in the presence of low serum iron and the milk iron concentration was about 3 times the serum iron at all levels of serum iron in all mice. When the data are plotted as a function of the serum iron:transferrin ratio, linear regression lines calculated individually for the normal and transgenic mice (Fig. 3 and legend) showed that there was no significant difference in the relation between serum iron and milk iron between the two groups. Since the presence of 13.6 mg ml⁻¹ (170 μM) lactoferrin provides 340 μM iron binding sites, more than enough to bind all the iron in milk, these data provide strong evidence against the hypothesis that milk iron depends on the concentration of iron binding ligands, including lactoferrin, in the secretory pathway.

Figure 3. The relation between serum and milk iron in normal (A) and transgenic mice (B).

The molar ratio of serum iron to transferrin is calculated as described in the text. Symbols are defined in the legend to Fig. 2. The parameters for the best fitting linear regression lines are shown in the table below.

Graph	Slope	Intercept (μM)	R
A	115 ± 11	37 ± 11	0.93
B	122 ± 11	40 ± 8	0.96

Role of transcytosis of transferrin in iron transfer from serum to milk

The next question was to what extent the transferrin transcytotic pathway could be responsible for iron transfer into milk. To answer this question we introduced iodinated transferrin into the blood stream of lactating CD1 mice on the laboratory diet and assessed the steady-state distribution of trichloroacetic acid precipitable radiolabelled transferrin in the milk. The specific activity of the transferrin in the plasma was calculated, using the TIBC as a measure of total transferrin, and used to calculate the plasma derived transferrin in the milk 7 and 9 h after injection of the radioactive transferrin. This value, along with the mean percentage saturation of the transferrin in these mice of 77 ± 3 % was used to calculate the maximal milk iron that could have been derived from transcytosis. Note that the transcytosed transferrin does not include all the transferrin in mouse milk because a substantial portion of mouse milk transferrin is made and secreted by the epithelial cells (Lee et al. 1987). However, transferrin synthesized by the mammary cells for secretion into milk would not be seen in these experiments and would not be involved in transcytosis of iron from the serum.

The blood levels of radioactivity following tail vein injection of transferrin conjugated directly with ¹²⁵I are shown in Fig. 4. The concentration of exogenous transferrin reached a steady state 5 h after injection. A similar result was obtained when transferrin was indirectly labelled with ¹²⁵I-tyramine cellobiose. The concentrations of transferrin originating from serum were determined in two experiments with directly labelled transferrin (5 mice total) and two experiments with tyramine cellobiose labelled transferrin (6 mice total); the values are reported in Table 2. As the mammary alveolar cell contains deiodinases (Jack et al. 1994) free iodine is found in the milk after administration of iodinated proteins (J. Monks & M. C. Neville, unpublished observations), so we determined the TCA precipitable radioactivity in the milk. The result suggests that less ¹²⁵I-tyramine cellobiose transferrin is metabolized than the directly labelled transferrin. Although the calculated concentration of directly labelled transferrin in the milk (8 to 10 μM) was slightly higher than the concentration of ¹²⁵I-tyramine cellobiose labelled protein (about 7 μM), the difference did not achieve statistical significance. The similarity of the values observed at 7 and 9 h indicates that a steady-state level of labelled transferrin was achieved in the milk. The overall mean blood derived transferrin in the whey was 8.0 ± 1.0 μmol l⁻¹, about 20 % of its concentration in plasma. Given the plasma transferrin saturation, this amount of transferrin would carry a maximum of 12.3 μmol l⁻¹ of iron into the milk or about 6 % of the total iron in mouse milk. We conclude that the majority of the iron must reach milk by another mechanism.

Figure 4. The concentration of radioactive transferrin in blood after intravenous injection of ¹²⁵I-transferrin.

Results are shown as counts per minute per microlitre of blood per million counts injected in each mouse. These data are the mean results from three mice injected with directly labelled transferrin and followed for 9 h with blood samples taken from the tail vein at the intervals shown on the graph.

Table 2.

Contribution of plasma transferrin to transferrin in mouse milk

	125I-Transferrin		125I-Tyramine cellobiose transferrin
Time after tail vein injection (h)	7	9	7	9
TCA precipitable radioactivity in milk (percentage of total 125I)	76 ± 6	82 ± 2	92 ± 3	99 ± 0.3
TCA precipitable radioactivity (μmol transferrin l−1)	8.0 ± 0.2	10.0 ± 1.1	7.2 ± 0.9	6.8 ± 0.5

Location of exogenous transferrin in the mammary gland

Exogenous transferrin has been localized within a transcytotic compartment of mammary fragments incubated in vitro (Lee et al. 1987). No large pool of basal transferrin was described in these fragments, which were washed extensively prior to analysis. In our hands such fragments were metabolically unstable and showed defects such as swollen endoplasmic reticulum after 1–2 h in vitro incubation (J. Monks & M. C. Neville, unpublished data). We therefore utilized a new technique of in situ incubation of rhodamine labelled transferrin with a surgically exposed mammary gland as described in Methods. After 2 h a large amount of endocytosed transferrin was detected near the basal membrane of the alveolar cells (Fig. 5), supporting the existence of receptor-mediated endocytosis and recycling of transferrin. A small proportion of transferrin was observed in vesicles in the cytoplasm of secretory cells, confirming the observations above that transcytosis of transferrin into milk does occur. The intensity of the fluorescence in the basal compartment of the cells supports the hypothesis that receptor-mediated transferrin endocytosis and cycling is the major pathway for transfer of iron into the mammary alveolar cell as it is for most other cells that have been studied.

Figure 5. The localization of transferrin in mammary epithelial cells.

The fourth mammary gland of a day 10 lactating mouse was incubated in vivo with a bath solution containing rhodamine labelled transferrin for 2 h. The mammary gland was fixed, sectioned, stained with FITC labelled wheat germ agglutinin (green), which stains heparin sulphate proteoglycans at the luminal surface, and DAPI (blue) to stain nuclei and viewed with a digital confocal microscope. The upper view shows transferrin localization. Most transferrin was found in vesicular compartments near the basal membrane of the alveolar cells (arrowheads). A small amount was found in discreet vesicles in the cytoplasm (arrows). The lower figure is the same view with pseudocoloured images added to depict the lumen and nuclei. Bar, 10 μm.

Quantitative analysis of the relation between milk and serum iron

Iron uptake has been studied most completely in reticulocytes where it is used primarily for the synthesis of the large amounts of haem associated with haemoglobin production (Ponka, 1997), a tissue culture cell line (K562 cells), and various types of hepatocytes. In these cell types the affinity of the transferrin receptor for diferric transferrin is 20–25 times its affinity for apotransferrin at 4°C (Young et al. 1984; Núñez et al. 1996) and there is evidence that apotransferrin does not inhibit iron uptake from diferric transferrin at all at 37°C (Young & Aisen, 1981; Huebers et al. 1983; Núñez et al. 1996). If the mammary alveolar cell behaves similarly, the relation between iron saturation in the plasma and milk iron would be highly non-linear as shown by the following theoretical analysis.

We designate the various forms of transferrin present in plasma as apotransferrin, A; transferrin binding a single iron molecule at the N-terminus, T_N; or the C-terminus, T_C; and fully saturated transferrin binding two iron molecules, T₂. The following relation between the concentration of the various forms of transferrin and the milk iron, [Fe]_m, derived in Appendix as eqn (A9), was used to obtain quantitative predictions based on various models of transferrin interaction with the transferrin recycling system:

(1)

where C_T is a proportionality constant equal to [Fe]_m when all plasma transferrin is in the diferric form. F is given by:

(2)

In both equations [A] is the concentration of apotransferrin and $K_{\text{D}}^{\text{A}}$ its dissociation constant from the transferrin receptor. The [T_X] and $K_{\text{D}}^{\text{X}}$ are the plasma concentration and receptor dissociation constant of the respective iron bound forms of transferrin. The coefficients a and b represent the rate of iron extraction from the receptor-transferrin complex relative to the rate of extraction of iron from diferric transferrin. Previous work in reticulocytes suggests that this ratio is 0.5 (Huebers et al. 1981). The various forms of transferrin were shown to bind to the reticulocyte transferrin receptor at 4°C with dissociation constants of 2.2 × 10⁻⁷ M for A, 4 × 10⁻⁸ M for T_N and T_C and 9 × 10⁻⁹ M for T₂ (Young et al. 1984).

The transfer coefficient, C_T, contains terms for the rate of milk secretion and the total concentration of transferrin receptor. That the rate of milk secretion is constant under the conditions of our experiment is suggested by the observation that the weights of the pups sucking dams on the lowest iron diets was decreased by no more than 6 % compared with pups sucking dams on the normal laboratory diet (data not shown). That the total concentration of the transferrin receptor is not up-regulated by low serum iron is suggested by the finding that low serum iron does not up-regulate the transferrin receptor in the lactating rat mammary gland (Sigman & Lönnerdal, 1990b). For these reasons we have treated C_T as a constant in this analysis.

The results of our analysis are presented in Fig. 6 where the milk iron values from Fig. 3 are replotted on a single graph against the measured [Fe]_s/[T]. Curve 1 in this figure shows the expected relation if A, T_C, T_N and T₂ all have equal affinities for the recycling receptor (we used K_D = 25 nM as suggested by Grigor et al. 1988; however, a similar result is obtained with a K_D between 1 and 100 nM). With this assumption most points fall above the line. Curve 2 was obtained assuming that the affinities of the transferrin receptor for mono- and diferric transferrin obtained from rabbit reticulocytes hold for the mammary gland and that at 37°C apotransferrin does not compete for the transferrin receptor. Curve 3 was obtained using the same set of assumptions except that apotransferrin was assumed to bind to the receptor with a 23-fold lower affinity than T₂. For curves 2 and 3 most of the experimental points fall below the line. Using iterative minimization of least square differences, we found the best fit (curve 4) when $K_{\text{D}}^2$ = 12 nM, $K_{\text{D}}^{\text{N}}$ =$K_{\text{D}}^{\text{C}}$ = 18 nM, and $K_{\text{D}}^{\text{A}}$ = 55 nM. A comparison of the errors of the fits (see legend, Fig. 6) suggests the apparent linearity of the relation between transferrin saturation in the plasma and milk iron may be due to the relatively small difference between the affinities of diferric transferrin and apotransferrin for the transferrin receptor in the mammary alveolar cell. This fit depends on the ratio between $K_{\text{D}}^2$ and $K_{\text{D}}^{\text{A}}$, rather than the absolute value adopted for $K_{\text{D}}^2$.

Figure 6. The effect of changes in transferrin affinity on the relation between transferrin saturation and milk iron.

The points are plotted using the symbols defined in the legend to Fig. 2 and data are from all three strains of mice. The lines represent four different sets of assumptions about the interaction of serum transferrin with the transferrin recycling system calculated from eqns (1) and (2). Line 1 assumes that the K_D values are equal for the reaction between all forms of transferrin and their receptor as shown in the table below, where the best-fitting transfer coefficient, C_T, and the fit parameter determined are also shown. Line 2 was calculated assuming that the affinity of apotransferrin for the receptor is very low and that the K_D values of mono- and diferric transferrin were equal to those given for the rabbit reticulocyte (Young et al. 1984). The best-fitting value for C_T, 227 μM, was calculated by minimizing the mean square difference between the theoretical line and the milk iron values for all [Fe]_s/[T] molecular ratios above 1.4. Line 3 was calculated using the K_D values found for the rabbit reticulocyte. Line 4 is the best fit line obtained by iterating eqns (A8) and (A9) with varying values of C_T, $K_{\text{D}}^{\text{A}}$, $K_{\text{D}}^1$ and $K_{\text{D}}^2$ to minimize the mean square difference.

	Line 1 equal KD values	Line 2 A not bound	Line 3 reticulocyte	Line 4 best fit
CT (μM)	240	227	227	240
$K_{\text{D}}^{\text{A}}$ (nM)	25	>1000	240	55
$K_{\text{D}}^1$ (nM)	25	40	40	18
$K_{\text{D}}^2$ (nM)	25	9	9	12
Fit parameter*	36 ± 3	49 ± 4	32 ± 3	18 ± 2

^*

The fit parameter is the square root of the mean square differences between each point and the theoretical line ± s.e.m. The fit of curve 4 to the data is significantly better than that of all other curves (P < 0.001).

DISCUSSION

The major finding of this study is that the level of milk iron is approximately 3 times the serum iron over the entire feasible range of serum iron. The low transfer of iron from iron-poor serum was unexpected because previous data suggest that the transferrin receptor is present in the mouse mammary gland and up-regulated during lactation (Schulman et al. 1989), and the transferrin receptor in many systems discriminates against apotransferrin to maintain the flow of iron for synthesis of iron containing enzymes (Klausner et al. 1983). For this reason we examined other potential mechanisms for the secretion of iron into milk. When these could not account for our findings we went on to develop a model for the iron transfer process that allowed differences in the affinity of the transferrin receptor for apotransferrin to be evaluated. We discuss these matters as well as the adaptive significance of the dependence of milk iron on the serum iron content below. In ‘Conclusion’, we present a general model for iron secretion into milk and set out some questions for future investigation.

Pathways for iron secretion into milk

The amount of calcium in milk is strongly dependent on the concentration of calcium binding ligands in the secretory pathway (Neville et al. 1994). To investigate whether the same could be true for iron we examined the relation between milk iron and transferrin saturation in normal and transgenic mice with 170 μM human lactoferrin, equivalent to an additional 340 μM of iron binding ligand, in their milk. The presence of this ligand had no effect on the relation between milk and serum iron leading to the conclusion that, unlike calcium, the concentration of iron binding ligand in the secretory pathway does not regulate milk iron.

We then analysed the possible role of the transcytosis of transferrin (Seddiki et al. 1992) from the plasma to the milk, finding that, at the steady state, the approximately 13 μM of serum transferrin secreted per litre of milk would provide a maximum of 6 % of the milk iron. If, as in Caco-2 intestinal epithelial cells, the transferrin that moves toward the apical membrane is restricted to apotransferrin (Núñez et al. 1997), this pathway would account for an even smaller fraction of the milk iron. The mRNA for the transferrin receptor has been observed in the mammary epithelium of the rodent (Sigman & Lonnerdal, 1990a,b) and, at least in the rat, is up-regulated many fold during lactation (Schulman et al. 1989). In addition we have localized transferrin to the basal surface of the mammary alveolar cell. The most straightforward interpretation of all these findings is that transferrin receptor mediated recycling of transferrin is the mechanism by which iron is delivered to the mammary epithelial cell to be secreted into milk.

Changes in apotransferrin affinity for the transferrin receptor affect iron transfer

Most cells are protected against the effects of low serum iron by both up-regulation of transferrin receptor (Aisen, 1998; Lok & Loh, 1998) and a low affinity for apotransferrin (Aisen, 1998). However, recent work in the rat mammary gland shows that transferrin receptors are not up-regulated by low iron (Sigman & Lönnerdal, 1990b), possibly because they are already maximally up-regulated by the demands of lactation. To examine the effects of changes in the affinity of the transferrin receptor for apotransferrin we developed the mathematical model derived in the Appendix. The results of this model show that the affinity of apotransferrin for the receptor has a profound effect on the relation between milk iron and the iron saturation of serum transferrin. When apotransferrin had a low affinity for the transferrin receptor as in reticulocytes (Young et al. 1984) and K562 cells (Núñez et al. 1996), the theoretical curve (Fig. 6) fell above most of the experimental points at low serum iron levels. On the other hand, when all forms of transferrin had the same affinity for the receptor, the curve fell below the majority of the data points. A good fit to the data was obtained when the dissociation constant for apotransferrin was about 4 times the dissociation constant for diferric transferrin (curve 4, Fig. 6), rather than 20 times as is the case for the reticulocyte transferrin receptor. This fit suggests that the transferrin receptor in the mammary gland may be somehow modified so that it has a higher affinity for apotransferrin (K_D∼55 μM) than the transferrin receptors in other cells or that the mammary cells may have a different form of the transferrin receptor such as the one recently cloned by Kawabata and colleagues (Kawabata et al. 1999). There is only one piece of evidence in the literature bearing on this point. Grigor and his colleagues (Grigor et al. 1988) found a K_D for human transferrin binding to rat mammary epithelial cells of 21–24 nM. In a single experiment both apo- and diferric transferrin inhibited binding by about 70 % at a concentration of 25 nM suggesting that the two have nearly equal affinities. In an intestinal cell model (Caco-2 cells), Núñez et al. (1996) have shown that the relative affinity of the transferrin receptor for diferric transferrin was about 10 times the affinity for apotransferrin at 4°C and that, unlike K562 cells, apotransferrin significantly inhibited iron uptake from diferric transferrin at 37°C. Thus there is a precedent for high affinity interaction of apotransferrin with the transferrin receptor.

However, other models that can be calculated from eqns (1) and (2) do produce a fit equal to that of curve 4. For example, there is excellent evidence that, in the liver, a second non-specific transfer process involving transferrin endocytosis and recycling exists in parallel with the transferrin receptor (Thorstensen et al. 1995; Trinder et al. 1996). If a non-specific receptor exists in parallel with the conventional transferrin receptor in the mammary gland a relation similar to that observed in curve 4 can be obtained (data not shown). A curve similar to curve 4 can also be obtained if uptake of iron from either the N or the C form of transferrin is less than one-half the rate of uptake from diferric transferrin (i.e. constants a or b or both in eqn (1) are less than 0.2) or, if as suggested by Núñez et al. (1996), binding of apotransferrin were to sequester the transferrin receptor in an internal compartment making less receptor available on the cell surface for binding of iron bearing transferrin. These potential mechanisms require further investigation.

Adaptive value of low milk iron in the face of low serum iron

Carmichael et al. (1977) observed that the amount of maternal iron supplementation necessary to produce maximal haemoglobin and abundant iron stores in suckling neonatal mice (10 mmol Fe l⁻¹ in a liquid milk diet) was far greater than that required to maintain positive iron balance in lactating dams (0.5 mmol l⁻¹). The results of our experiments are consistent with these findings and suggest that iron delivery to the milk is less effective than iron delivery to other cells in the body under iron deficiency conditions. Would such a mechanism have an adaptive significance? The rodent lactational strategy protects the mother rather than the neonates. In species that can produce a large litter as frequently as every 3–6 weeks, an investment in the mother, protecting her iron stores, is more economic than an investment in the pups. That this strategy indeed preserves the dam at the expense of the pups was apparent in high mortality rates after 10 days lactation in litters whose mothers were fed the lowest iron diets (P. Zhang & M. C. Neville, unpublished observations).

Conclusion

The findings of this study together with previous data concerning the transferrin receptor in the mammary gland (Grigor et al. 1988; Schulman et al. 1989; Sigman & Lönnerdal, 1990a,b) lead to the model in Fig. 7 in which the major pathway for iron transfer into milk begins with binding of transferrin to the transferrin receptor at the basal surface of the cell. By analogy to other systems the transferrin is endocytosed into a low pH endosomal compartment where it gives up its iron to an unknown iron transport mechanism and is recycled to the surface. The iron is transferred through the cytoplasm, presumably bound to a transport molecule such as mobiliferrin, and utilized in the synthesis of the iron containing enzyme xanthine oxidase secreted with the milk fat globule (MFG; Mather & Keenan, 1998) or transported, again by an unknown mechanism, into the Golgi and/or secretory compartment where it binds whatever iron binding proteins are present and is secreted by exocytosis. A minor amount of iron (∼6 %) may be transcytosed, possibly by a receptor independent mechanism.

Figure 7. A model for iron transfer into rodent milk.

Transferrin with bound iron is endocytosed into the mammary alveolar cell by a major recycling pathway and a minor transcytotic pathway. A minimum of 94 % of the iron that eventually finds its way to milk is bound to transferrin that enters onto the transferrin receptor and is released in the acidic environment of the endosome to be transferred to the cytoplasm by an unknown transport mechanism. From there it is either incorporated into iron containing molecules like xanthine oxidase (XO) or transported into the secretory vesicle where it binds to proteins such as casein and lactoferrin and is secreted by exocytosis. A minor amount of transferrin, presumably with its iron bound, finds its way across the mammary alveolar cell via transcytosis.

Several major questions remain for investigation. The first is the relative affinity of the mammary transferrin receptor for the various forms of serum transferrin. As described above the approximately linear relation between the serum and milk iron may be explained by a relatively high affinity for apotransferrin. The second question is the nature of the iron transporters in the endosomal membrane and the membranes of the Golgi vesicles and secretory compartment. With the recent identification of iron transporters in yeast (Dix et al. 1997) and eukaryotes (Eide, 1997), it may now be possible to make some progress on this heretofore knotty problem. In particular, it will be of interest to determine whether DMT1, a gene that has been linked to iron transport in intestine and reticulocytes (Fleming et al. 1997, 1998), plays a role in the transfer of iron into milk and whether HFE, a protein that forms a high affinity complex with the transferrin receptor (Waheed et al. 1999), altering interstinal iron absorption, is present in the mammary alveolar cell. If so, it will also be of interest to determine whether either molecule is present in the mammary alveolar cells of species that secrete little iron into their milk.

APPENDIX

Theoretical analysis of the relation between serum and milk iron

The rate of iron transport into milk, V̇_Fe, is equal to the product of the iron concentration in the milk, [Fe]_m, and the rate of milk secretion, V̇_m, assumed to be constant under our experimental conditions. In terms of the model shown in Fig. 7, V̇_Fe is equal to the rate of transfer of iron from the transferrin-receptor complex in the endosome into the cytoplasm. Transferrin can exist in several forms that bind iron at the C-terminal lobe, the N-terminal lobe or both, represented by T_C, T_N and T₂, respectively. Then:

(A1)

where C_X represents the transfer rates of iron from each form of transferrin from each receptor and F_TX is the fraction of each form of transferrin; e.g. F_TX =[RT_X]/[R]_tot where [RT_X] is the concentration of the receptor-transferrin complex for any form of transferrin and [R]_tot is the total transferrin receptor concentration. When diferric transferrin is the only form present in the serum, F_T2 = 1 and:

(A2)

where [Fe]_m^max is the maximum concentration of iron in the milk, obtained when the serum transferrin is completely saturated. If we set C_N =aC₂, C_C =bC₂, then:

(A3)

Under conditions where the rate of milk secretion, V̇_m, and the concentration of transferrin receptors, [R]_tot, do not vary with serum iron, we can define a new constant:

(A4)

and

(A5)

To calculate the ratios F_TX we consider that the transferrin receptor, R, has five configurations: unoccupied, R; bound to apotransferrin, RA; bound to singly ligated transferrin, RT_C and RT_N; and bound to diferric transferrin, RT₂, and:

(A6)

The corresponding dissociation constants are given by:

(A7)

where [A], [T_C], [T_N] and [T₂] are the serum concentrations of apotransferrin, monoferric C and N transferrin and diferric transferrin, respectively. Combining (A6) and (A7):

(A8)

Let [R]/[R]_tot =F. From (A5) and (A7):

(A9)

Finally, it is necessary to calculate [A], [T_N], [T_C], and [T₂] for any molar ratio of serum iron, [Fe]_s, to total serum transferrin, [T]. We assume that the distribution of iron among the various forms of transferrin in mouse serum is equivalent to the distribution found by Makey & Seal (1976) using electrophoresis to measure the proportion of the various forms of transferrin in the presence of different concentrations of iron. Their data are plotted in Fig. 8. We were able to fit these data using the following empirical formulas in which [Fe]_s/[T] is represented by x:

(A10a)

(A10b)

(A10c)

To use eqn (A9), F is calculated from eqn (A8) using values for [A] and the [T_X] from eqns (A10a), (A10b) and (A10c) and any desired combination of K_D^X. In general we make the assumption that extraction of iron from monoferric transferrin is as efficient as extraction from diferric iron. In this case, iron extraction is proportional to the number of iron molecules bound to transferrin and a and b are equal to 0.5. T_C can be calculated from [Fe]_m at high [Fe]_s/[T]. These values are substituted into eqns (A8) and (A9) for any combination of values for the $K_{\text{D}}^X$.

Figure 8. Analysis of the forms of transferrin in the serum.

The points are derived from the data of Makey & Seal (1976). The lines are the best fit to a polynomial using the program SLIDEWRITE. The coefficients of the polynomial are given in the text.