The absorption of iron from the gastrointestinal tract depends on the needs of a given animal and on the bioavailability of iron in the food, i.e., the chemical speciation.⁴ Factors involved in maintaining iron homeostasis by modulating the transfer of iron through mucosal cells exist on a luminal, mucosal and systemic level. The transfer from the intestine to the blood may be divided into several steps including binding of iron to the brush border membrane, penetration into the mucosal cell, intracellular processing and transport, and release across the basolateral membrane into the blood plasma (Skikne and Baynes 1994). The interactions between the multiple iron species and the constituents of the brush border membrane are of great importance for the efficiency of uptake. Speciation of nonheme iron depends on the dietary composition as well as on intestinal secretions. The most important enhancing factors in the diet are animal tissue proteins (Lynch et al. 1989) and organic acids, e.g., ascorbic, citric and lactic acids (Gillooly et al. 1983). Potent inhibitors of iron uptake are polyphenols (Gillooly et al. 1983, Siegenberg et al. 1991), phytates (Siegenberg et al. 1991) and calcium (Cook et al. 1991, Hallberg et al. 1991). The solubility of iron is markedly influenced by intestinal secretions such as hydrochloric acid in the gastric juice (Bezwoda et al. 1978), bicarbonate in pancreatic secretions (Zempsky et al. 1989) or an elutable mucosal factor utilized in iron deficiency (Huebers et al. 1974). Several membrane components have been described to keep iron in an available form and to be involved in the transport across the brush border membrane. Mucus and mucins coating the intestinal mucosa bind iron and thus prevent it from hydrolysis and precipitation at higher pH (Conrad et al. 1991), but mucus secretion was observed to be decreased in iron-deficient rats (Wien and Van Campen 1991). Nonesterified fatty acids were found to be a major iron-binding component of rabbit duodenal brush border membrane vesicles (Simpson and Peters 1987). Evidence for iron-binding proteins and a protein-mediated uptake has been shown (Conrad et al. 1993, Teichmann and Stremmel 1990).

From the foregoing, it becomes clear that a certain amount of iron compounds will be adsorbed or bound to the brush border membrane, depending on their charge and thermodynamic properties. In the absence of sufficient chelating ligands, and as a result of the pH rise upon entry to the duodenum, polynuclear iron (hydr)oxide is formed, which may also be partly attached to the surface of the membrane. Ferrous iron will be rapidly oxidized and hydrolyzed. A surface-catalyzed oxidation and/or nucleation similar to that described for ferritin (Bakker and Boyer 1986, Harrison et al. 1994) may not be excluded. Thus, for studying uptake of iron with its complex underlying chemistry, a discrimination between binding to and penetration through the membrane is needed. The initial brush border permeation step was frequently investigated with isolated brush border membrane vesicles (BBMV⁵; Marx and Aisen 1981, Muir et al. 1984, Simpson and Peters 1984, Stremmel et al. 1987). Because the equilibrium of iron uptake is also dependent on binding within the membrane and on the inner surface of these vesicles, this work developed a strategy to determine iron distribution in the following four compartments: outer membrane surface, membrane interior, inner membrane surface and aqueous phase within the vesicles. Further, the interdependence of the uptake of iron and the chelating ligand nitrilotriacetate (NTA) was studied.

MATERIALS AND METHODS

The rabbit brush border membrane vesicles were prepared by a modified method of Hauser et al. (1980). Duodenum and upper jejunum from freshly obtained intestine of one or two rabbits were used. The tissue was rinsed with ice-cold saline. The intestinal segments were turned inside out by means of a perspex stick. The mucosa was scraped off with a glass slide, suspended in 30 mL (per rabbit) 300 mmol/L mannitol, 5 mmol/L EGTA, 12 mmol/L Tris (pH 7.1 with HCl) and then diluted with 120 mL water. The cells were immediately homogenized three times for 1 min at 22,000 rpm with a Warring blender, and contaminating membranes were precipitated with 0.01 mol/L MgCl₂ . A foam containing lipids was removed by suction after 15 min, and the suspension was centrifuged for 15 min at 3000 × g. The supernatant was centrifuged again for 30 min at 28,000 × g. The resulting pellet was suspended in 30 mL buffer (60 mmol/L mannitol and 5 mmol/L EGTA, pH 7.1 with HCl) and homogenized in a glass-teflon-potter (10 times, 2000 rpm). The precipitation and centrifugation procedure was repeated. The pellet of the brush border membranes was suspended in 30 mL of the desired transport buffer (for iron uptake experiments, 50 mmol/L mannitol, 100 mmol/L NaCl, 100 µmol/L MgSO₄ and 100 mmol/L Hepes, pH 7.4 with Tris; for glucose uptake studies, 300 mmol/L mannitol and 20 mmol/L Hepes, pH 7.4 with Tris), homogenized in a glass-teflon-potter, and centrifuged at 28,000 × g (30 min). The final pellet was resuspended in 1-3 mL transport buffer by means of a thin syringe needle (0.8-mm diameter), with a protein content of 15-60 g/L. The vesicles were stored in liquid nitrogen for up to 2 mo.

The osmotic volume of the vesicles was calculated after reaching equilibrium of uptake of D-glucose with an inwardly directed sodium gradient and had a value of 0.5-1.5 µL/mg protein (cf. Kessler and Toggenburger 1979).

Protein concentration was determined according to the method of Bradford (1976) employing an assay kit (BioRad, Glattbrugg, Switzerland) and bovine -globulin as standard.

The timing of the addition of medium components is of crucial importance. A good reproducibility was obtained with the following procedure: NTA was added from a stock solution to the twice concentrated buffer (100 mmol/L mannitol, 200 mmol/L NaCl and 200 mmol/L Hepes, pH 7.4 with Tris). Then iron (stock solution 1 mol/L in 1 mol/L HCl) was slowly added to the well-stirred medium. Finally, the solution was completed with water and ⁵⁹Fe (in 0.1 or 0.5 mol/L HCl).

In the first step, the iron bound to the outer membrane surface was quantified. Loosely bound iron was removed by an incubation in iron-free incubation medium (Fig. 1B). More tightly bound iron was determined by three different methods, namely, (Fig. 1C) exchange of iron by incubation in an iron-containing (182 µmol/L), but tracer-free medium, (Fig. 1D) removal by running through a Sephadex G25 M column (PD-10), and (Fig. 1E) incubation with 10 mmol/L EDTA in the stop solution. All separation procedures were performed on ice (except the column experiments) and were finished by filtration through a nitrocellulose filter (pore size 0.65 µm) and measurement of the tracer on the filter.

In preliminary experiments, it was shown that the Sephadex G25 M column (1.5 × 5 cm) is suitable for the separation of BBMV from the Fe-NTA-incubation medium. The small iron peak appearing with the protein peak was in accordance with the measured amount of iron in the uptake. Average iron recovery was 60-70%, whereas vesicle protein was retrieved almost completely. The columns were prerinsed with 50 mL of transport buffer. After incubation for iron uptake, 100 µL of the vesicle suspension was loaded directly in an undiluted state on a column and eluted with transport buffer. Vesicle proteins appeared after a void volume of 2.5 mL in a volume of 2 mL.

The amount of iron in the inner vesicle compartments became detectable after lysis with 0.1% Triton X-100 in the 2-mL vesicle fraction obtained from the G25 M column (Fig. 1F). Triton X-100 (0.1%) was effective in lysing the vesicles as verified by the release after lysis of the fluorescent dye calcein from vesicles loaded with this dye. Iron in the aqueous phase within the vesicles was discriminated from membrane-bound iron by a second application of a Sephadex column (Fig. 1H) and a subsequent iron analysis of the elution fractions (Fig. 1I). The vesicle peak of this second column was broader than the peak of the first column because of the larger volume of suspension applied (2.2 mL). However, the resolution allowed a good distinction between protein-bound and aqueous iron (cf. Fig. 4). A simultaneous incubation with Triton X-100 and EDTA (Fig. 1G) resulted in the detection of iron bound to the inner membrane surface in the low molecular weight fraction, whereas iron strongly bound in the membrane interior remained in the fraction in which proteins were present. This procedure is based on the assumption that iron in the aqueous phase of the vesicles would be removed by the Sephadex filtration, whereas iron adhering to the inner vesicle surface would resist Sephadex dissociation but be susceptible to EDTA chelation.

RESULTS AND DISCUSSION

The presence of pn-Fe(OH)₃ can be demonstrated by thermodynamic calculations. The percentage of [Fe]_total present as pn-Fe(OH)₃ as a function of given iron and NTA concentrations is shown in Table 1. Iron speciation was calculated considering the formation of the mononuclear and dinuclear complexes of iron with hydroxide Fe_x(OH)^3x-y_y and NTA Fe(NTA)^33z_z as well as the ternary complexes Fe_x(OH)_y(NTA)^3xy3z_z . The calculation was based on stability constants given by Anderegg (1982) for NTA complexes and Baes and Mesmer (1976) for hydroxo complexes at ionic strength of 0.1 mol/L and temperatures of 20-25°C. This ionic strength is comparable to that used in our study. With increasing temperature, the acidity of Fe³⁺_aq and the formation of Fe₂(OH)⁴⁺₂ rise (Baes and Mesmer 1976). The solubility of Fe(III) depends on the solid phase formed in different media. We judge the value of the solubility product used for calculations K_SO = 10³⁸ as a lower limit with respect to our conditions considering reported data on amorphous ferric hydroxide: K_SO = 10^38.6 (25°C, 3 mol/L NaClO₄); K_SO = 10^38.8 (25°C, 0 mol/L) (Smith and Martell 1976). Thus, unless there is a large excess of NTA, the solution is not stable with respect to the formation of polynuclear complexes. However, no precipitation occurs even if pn-Fe(OH)₃ represent the major iron constituents with low NTA concentrations. This is due to colloid-stabilizing effects of other medium components such as Tris, which form surface complexes on pn-Fe(OH)₃ . Thus, the term pn-Fe(OH)₃ has to be read to include not only (oxy)-hydroxides of iron but also all polynuclear mixed complexes with other medium components.

Table 1. Fraction of polynuclear iron complexes [ pn-Fe(OH)3] at given iron and nitrilotriacetate (NTA) concentrations1 [View Table]

There is also experimental evidence for the existence of pn-Fe(OH)₃ . The total iron uptake strongly depended on the ratio of metal to ligand (Fig. 2). Especially with low ligand concentrations, large amounts of iron were associated with the vesicles. The rate of iron transfer through the membrane was low in the presence of an excess of either NTA or iron: this is due in the first case to a complete and stable wrapping of iron, in the second, to the formation of more extended pn-Fe(OH)₃ . Thus, the increase of the total iron content of the vesicles was ascribed mainly to surface-bound pn-Fe(OH)₃ . Adsorption onto the outer surface was indicated by the kinetics of iron uptake (cf. Fig. 7): within the first 1 or 2 min, there was a very rapid increase in the iron content. The following uptake was saturable, whereby equilibrium was not yet entirely reached after 1 h of incubation. The quantity of the first rapid accumulation was verified to be equal to that bound to the outer surface (see the compartmentalization section below).

The results of the several experiments for the determination of the outer surface-bound iron are presented in Figure 3 (steps A-E, corresponding to Fig. 1). The reference value was based on an incubation of 10 min followed by immediate filtration and washing (normal uptake procedure). The dissociation of membrane-bound iron was a slow process, shown by the time dependence of iron release in several washing procedures. Compared with the short-time washing of the reference experiment (Fig. 3A), significantly (P < 0.05) more iron was removed by an incubation at 0°C during 1 h even in an iron-free buffer (Fig. 3B). Because the buffer contained no chelating ligands, this additional portion of iron could not be very tightly associated with the membrane. Efficacy of removal was improved when tracer-free iron was added to buffer (1 h incubation; Fig. 3C).

The total amount of iron located on the outer surface was about 35% of the iron measured in the overall uptake. The findings of exchange of radioactive tracer iron by inactive iron (Fig. 3C), removal by EDTA (Fig. 3D), and also separation by Sephadex columns (Fig. 3E) were not different. We assumed that the retention behavior of the vesicles was not affected by the different treatments to determine the outer surface-bound iron: in the iron exchange and Sephadex column experiments the medium composition was the same as in the reference, and the EDTA treatment gave the same results as the two other procedures.

Although this externally adsorbed iron represented a considerable portion of the vesicle iron, its density on the surface was not high. With an average vesicle diameter of 100 nm (Perewusnyk et al. 1985) and a membrane protein concentration of 67 g/100 g (Pind and Kuksis 1986), the surface of one vesicle was occupied by 140 iron ions. This corresponds to an average iron-iron-distance of at least 30 nm (for this lower limit only mononuclear complexes are assumed to be present), or a protein to iron ratio of about 10 (calculated with an average protein molecular weight of 50 kDa). Furthermore, the adsorbed iron constituted only a low percentage of total iron in the medium. This confirmed the expectation that the small pn-Fe(OH)₃ present are altogether barely adsorbed onto the vesicles as well as onto the filters: they are negatively charged by NTA on their surface and therefore largely protected against aggregation and adsorption to the negatively charged membrane surface.

After removal of medium and externally bound iron and lysis of the vesicles with Triton X-100, no iron was detectable on the nitrocellulose filters, independently of the presence or absence of EDTA (Fig. 3F, G). Therefore, protein-bound iron was further separated from dissolved iron by a second Sephadex column (Fig. 1H). The elution profiles of the lysed vesicles in the presence and absence of the chelator are shown in Figure 4. The first peak was coincident with the protein peak of preliminary experiments, in which the elution profiles of separately loaded medium and BBMV were determined (data not shown). Without EDTA, almost all iron remained in the protein peak. The 8.5% of iron in the second, low molecular weight iron peak represents iron in the aqueous phase within the vesicles. Utilizing a mean vesicle volume determined by measuring the equilibrium value of glucose uptake after 1 h (1 ± 0.5 µL/mg protein), the iron content in the aqueous phase after 10 min incubation corresponded to 41% of the equilibrium with medium iron. As noted above, reaching the equilibrium of iron uptake took at least 1 h.

EDTA bound about two thirds of the remaining iron from lysed membranes. Thus, an astonishingly large amount of iron either was not accessible to EDTA or was very tightly bound to membrane constituents and associated structural proteins. This portion does not include EDTA-resistant pn-Fe(OH)₃ because of the quantitative amount of iron and the incubation time.

The results of the compartmentalization steps are summarized in Figure 5. Binding to both surfaces and also within the membrane represented a predominant part of the overall uptake of iron. However, the relative portions in the four compartments will vary dependently with the medium composition, i.e. the iron speciation. Further, the vesicle volume established by glucose uptake experiments (~1 µL/mg protein) was not in good agreement with that calculated from size (Perewusnyk et al. 1985) and protein content (Pind and Kuksis 1986) determinations of ~4 µL/mg protein (vesicle radius 100 nm, membrane thickness 10 nm, membrane protein concentration 67 g/100 g). This discrepancy is in accordance with the findings of Gains and Hauser (1984), which identified only one of 4-6 vesicles to be closed. Thus, the amount of iron on the outer membrane surface may be overestimated at the expense of the contents of other compartments. In our current work, the integrity of the vesicles is further being tested. However, the present results impressively demonstrate the importance of adsorption and complex formation with proteins in the overall process of iron uptake.

In the presence of iron, NTA uptake was significantly greater than in the absence of iron (P = 0.0001; Fig. 7). However, the additional amount of NTA taken up was smaller than the amount of iron taken up. On the other hand, the NTA concentration was seven times larger than that of iron in the aqueous phase of the vesicles. We explain this experimentally obtained ratio by the formation of mixed complexes membrane donor-iron hydroxide-NTA.

In summary, the present data suggest that iron uptake represents a progression toward equilibrium concentrations of mono- and polynuclear complexes within and on the surfaces of the vesicles. The slow approach to equilibrium may be due in part to the formation of polynuclear complexes within the vesicles. The apparent accumulation of iron with respect to the medium is caused by the high binding capacity of the membrane and the adsorption of polynuclear complexes.