The high-affinity FBP/FR (folate-binding protein/folate receptor) is expressed in three isoforms. FRα and FRβ are attached to cell membranes by hydrophobic GPI (glycosylphosphatidylinositol) anchors, whereas FBPγ is a secretory protein. Mature neutrophil granulocytes contain a non-functional FRβ on the surface, and, in addition, nanomolar concentrations of a secretory functional FBP (29 kDa) can be present in the secondary granules. A statistically significant correlation between the concentrations of functional FBP, probably a γ isoform, in granulocytes and serum supported the hypothesis that serum FBP (29 kDa) mainly originates from neutrophils. The presence of FBP/FRα isoforms were established for the first time in human blood using antibodies specifically directed against human milk FBPα. The α isoforms identified on erythrocyte membranes, and in granulocytes and serum, only constituted an almost undetectable fraction of the functional FBP. The FBPα in neutrophil granulocytes was identified as a cytoplasmic component by indirect immunofluorescence. Gel filtration of serum revealed a peak of FBPα (>120 kDa), which could represent receptor fragments from decomposed erythrocytes and granulocytes. The soluble FBPs may exert bacteriostatic effects and protect folates in plasma from biological degradation, whereas FRs on the surface of blood cells could be involved in intracellular folate uptake or serve as signal proteins. The latter receptors have also been used for therapeutic targeting in malignancy.
A high-affinity FBP (folate-binding protein) was first identified in bovine milk [1–3]. The protein has been purified and sequenced; it consists of 222 amino acids, with a molecular mass of 30 kDa on the basis of amino acid composition and carbohydrate content . FBP purified from human milk has a similar molecular mass, as estimated from its amino acid composition, carbohydrate content and sequence homology with FBP from bovine milk . The purified human milk FBP exists in two forms: one is water-soluble and elutes at 27 kDa on gel filtration, and the other carries a hydrophobic GPI (glycosylphosphatidylinositol) residue and forms micellar aggregates with Triton X-100, and elutes at 100 kDa [5,6]. The micellar FBP is transformed into the water-soluble FBP after enzymatic cleavage of the GPI tail by PI-PLC (phosphatidylinositol-specific phospholipase C) .
The ubiquitous presence of FBPs in mammalian cells, tissues and body fluids is now well established [1,2]. Three isoforms of FBP, α, β and γ, are expressed on chromosome 11q13 in humans [7,8]. Only the former two isoforms can be modified with a GPI tail , which enables them to be anchored to plasma membranes as membrane-bound FRs (folate receptors), which are susceptible to enzymatic cleavage by PI-PLC or proteases. FRα is normally expressed in secretory epithelial cells and is overexpressed in certain epithelial tumour cells, such as ovarian carcinoma ; its amino acid sequence is similar to that of human milk FBP, which represents the soluble form of FRα [5,11].
The presence of soluble and membrane-linked FBPs in human blood cells and plasma have been reported. The outer membrane surface of erythrocytes [12,13] and neutrophil granulocytes contain FRs [14–16]; in addition the secondary cytoplasmic granules of neutrophils contain a high-affinity FBP [17–20]. The membrane receptor of erythrocytes immunoreacts with polyclonal antibodies against human placenta FR , which is a mixture of α and β isoforms , whereas that of neutrophil granulocytes is a β isoform . The water-soluble FBP in neutrophil granules is likely to be a γ isoform without a GPI residue, and is most probably a source of origin of plasma/serum FBP [23,24].
As yet, no α isoforms of FBP/FR have been identified in human blood. A major aim of the present study was therefore to establish whether these isoforms are present in human blood. For this purpose, we reviewed previous data on FBP/FR in human blood, and employed a combination of radioligand ([3H]folate) binding and immunotechniques involving a polyclonal antibody against purified FBPα from human milk . The latter antibody exhibits an immunoreactive profile similar to that of the monoclonal antibody, MOv18. Like MOv18 it only immunoreacts, as shown by immunoprecipitation, immunoblotting, ELISA and immunohistochemistry, with FRα, such as in serous ovarian adenocarcinoma [7,10,11,26,27]; normal ovarian tissue where FRβ prevails  shows no immunoreactive response .
MATERIALS AND METHODS
Proteins and antibodies
FBPs were purified from human and bovine milk as described previously [4,5]. Polyclonal antibodies against human and bovine milk FBPs were raised in rabbits [25,28]. Rabbit antibodies against human lactoferrin (A0186) and a FITC-labelled Ig fraction of pig anti-rabbit serum (F0205) were from DAKO. Rabbit serum was from State Serum Institute.
[3H]Folate, with a specific activity of 26–45 kCi/mol, was supplied by Amersham International Ltd. (Amersham, U.K.).
Preparation of haemolysate from erythrocytes
Haemolysates were prepared as described previously in a study investigating the binding of [3H]folate to haemoglobin . Briefly, samples of EDTA-stabilized venous blood were taken from healthy volunteers with normal haematological parameters [haemoglobin, RBC (red blood cell) count, MCV (mean corpuscular volume), WBC (white blood cell) count and differential count). All participants gave their informed consent and the study had ethical approval. The plasma and buffy-coat layer was discarded after gentle centrifugation (1500 g) for 15 min, and erythrocytes were washed several times in a volume of isotonic NaCl equal to the plasma volume to remove contaminating plasma. The washed erythrocytes were finally diluted with a volume of distilled water equal to the plasma volume and subjected to repeated freeze–thaw procedures. Aliquots of haemolysates were dialysed overnight against 0.17 M Tris/HCl (pH 7.4) before analysis to remove endogenous folate.
Preparation of lysates from leucocytes
Specimens of EDTA-stabilized venous blood were drawn from healthy volunteers with normal haematological parameters (compare with above), the leucocytes were harvested, washed several times in isotonic NaCl to remove contaminating plasma and then subjected to repeated freeze–thaw procedures [19,20]. The cells were homogenized and solubilized in 1 g/l Triton X-100, dialysed against 0.17 M Tris/HCl (pH 7.4) for 24 h to remove endogenous folate, and finally centrifuged at 2000 g for 15 min [19,20]. Aliquots of the supernatants were used for analysis.
Preparation of pooled sera
Pooled sera were prepared as described previously . Briefly, sera were pooled, dialysed overnight against 0.05 M imidazole buffer (pH 6.3)/30 mM NaCl to remove endogenous folate and then centrifuged at 1000 g for 15 min. The supernatant was used for analysis.
An ELISA was carried out with an anti-(human milk FBP) antibody , which specifically immunoreacts with FBP/FRα, as described previously . An ELISA using antibodies against FBP purified from bovine milk was also conducted for further testing of the specificity of the immunoreactive response in ELISA . FBPs purified from human and bovine milk were analysed by ELISA to control the analytical performance of the assays [25,28].
To demonstrate the presence of human FBPα in neutrophil granulocytes, we used ethanol-fixed smears of normal peripheral blood leucocytes as substrate . The smears were incubated with 1.5 mg of protein/ml of rabbit anti-(human FBPα), anti-(human lactoferrin) and rabbit serum (control) for 30 min in a humid chamber. Each smear was individually washed with excess PBS and all smears are submersed in the chamber in PBS for 10 min to remove non-specifically-attached serum. After removal of all of the PBS from the slides, the submersion of slides in PBS was repeated for another 10 min. Excess PBS was removed around the circle of each slide, but a small amount of PBS was left in the circle to avoid drying of the sample. FITC-labelled Ig fraction of pig anti-rabbit serum (1:25 dilution on PBS; F0205, DAKO) was added as a drop on to the encircled cell smear of each slide. The slides were again incubated for 30 min at room temperature (21°C) in a humid chamber. The washing procedures were repeated. A drop of a glycerol/PBS (2:1) was then applied to each smear and a cover glass was applied. The smears were analysed directly with a fluorescence microscope.
To show the specificity, we absorbed the rabbit anti-(human FBP) antibodies with excess human FBP and repeated the immunofluorescence procedures. The lightening of the cytoplasm in the neutrophil granulocytes disappeared.
The specificity was further tested by replacing rabbit anti-(human FBP) antibodies with rabbit anti-(bovine FBP) antibodies or rabbit serum.
FBP/FRα in erythrocytes
Antibodies against FBPα from human milk immunoreacted with serial dilutions of haemolysed erythrocytes (Figure 1). To test the specificity of response, we performed a parallel ELISA analysis with antibodies against bovine milk FBP (Figure 1, inset). The analytical performance of the two ELISA methods were controlled by running serial dilutions of purified human and bovine milk FBP (Figure 1).
FBPα isoform in haemolysates identified by ELISA with antibodies directed against FBPα from human milk
Immunoreactive FBP/FR in haemolysed erythrocytes did not bind [3H]folate in equilibrium dialysis experiments.
FBP in leucocytes
The concentration of functional FBP, in terms of maximum binding of [3H]folate in equilibrium dialysis experiments, was determined in leucocyte lysates obtained from three healthy subjects. The FBP concentration in leucocyte lysate showed a great inter-individual variation with values of: <0.01, 3.8 and 5.0 nmol/l.
Gel filtration of Triton-X-100-solubilized leucocyte lysates containing 3.8 and <0.01 nmol/l FBP were performed after pre-incubation with [3H]folate. The former sample exhibited a single peak of radioactivity at 29 kDa, whereas no peak was observed in the latter sample (Figure 2). No micellar aggregates of GPI-tailed FR appeared at 100 kDa.
FBPα isoform from leucocyte lysates
FBP/FRα in neutrophil granulocytes
ELISA using antibodies against human milk FBPα was performed to identify FR/FBPα in neutrophil granulocytes.
Serial-dilution curves (2-fold initial dilution) of the three leucocyte lysates in Figure 2 showed a weak positive response in ELISA. The response was similar in the three samples, despite widely different concentrations of functional FBP, suggesting that FR/FBPα contributed very little, or most probably not at all, to the total amount of functional FBP. The analytical performance of the ELISA was tested by running serial dilutions of purified human milk FBP (Figure 2).
Indirect immunofluorescence with antibodies against human milk FBPα identified the α isoform of FR/FBP in neutrophil granulocytes that had been subjected to ethanol fixation . The positive immunostaining in Figure 3(A) exhibited a characteristic pattern involving the nuclei or perinuclear zone of neutrophils. This type of pattern is probably due to redistribution of cytoplasmic components to anionic components of the nucleus upon removal of lipids from granule membranes by ethanol fixation . Indirect immunofluorescence with antibodies against another cytoplasmic protein in neutrophils, lactoferrin, showed a very similar pattern of immunostaining (Figure 3B). No immunostaining was observed with rabbit serum (control) (Figure 3C).
Indirect immunofluorescence for identification of cytoplasmic FBPα in neutrophil granulocytes
The specificity of the present anti-FBPα antibody is as follows: immunostaining disappeared when the indirect immunofluorescence procedure was repeated after absorption of the antibody with excess human milk FBP (results not shown); the typical immunostaining pattern disappeared when antibodies against bovine milk FBP were used (results not shown); and there was no immunostaining of FRβ on the surface of neutrophil granulocytes [14–16,22].
Correlation between functional FBP in serum and leucocyte lysates
Matched data for serum and leucocyte concentrations of functional FBP from previous studies [19,20] were re-analysed in order to test the hypothesis that there is a relationship between the FBPs in granulocytes and plasma.
A weak positive correlation (statistically significant) between the concentrations of functional FBP in serum and leucocyte lysates was found in 15 healthy subjects with >0.5 nmol of FBP/109 leucocytes (Figure 4).
Correlation between concentrations of [3H]folate-bound FBP in serum (>0.5 nmol/l FBP) and leucocyte lysates in 15 healthy individuals
Functional FBP and FBPα in serum
The concentration of [3H]folate-bound FBP (functional FBP) was determined by equilibrium dialysis with pooled sera from pregnant women (FBP, 0.73 nmol/l), blood donors (FBP, 1.10 nmol/l) and umbilical cord blood (FBP, 0.71 nmol/l). The pooled sera were analysed by ELISA with antibodies against human milk FBPα to screen for the presence of FBPα (Figure 5).
FBPα isoform in serum
Serial-dilution curves of the three types of pooled sera showed weak positive responses in ELISA, suggesting the presence of FBPα in serum. No response was seen when a parallel ELISA with antibodies against bovine milk FBP was performed to test the specificity (results not shown). The analytical performance of the ELISA was controlled by running serial dilutions of purified human milk FBP (Figure 5).
The gel-filtration experiments in Figure 6 were performed to estimate the molecular mass of functional FBP and FBPα in serum. A Triton-X-100-solubilized pool of sera from blood donors was pre-incubated with [3H]folate to identify functional FBP. The elution profile revealed a major peak (400 c.p.m.) at a position of 29 kDa, which is similar to that in leucocyte lysates, as well as a small (60–80 c.p.m.) broad and shoulder-like peak from 70–120 kDa.
Gel-filtration profile of FBP isoforms in pooled sera from blood donors
The elution profile of FBPα determined by ELISA with antibodies against human milk FBPα contained a small peak at 43 kDa, continuing into a broad peak with a maximum height (A=0.6) at >120 kDa (void volume). No FBPα was found at a position of 29 kDa where the major peak of functional FBP had its maximum height (Figure 6).
The most important finding of the present study was the identification of FBPα in human blood cells and plasma/serum.
A high-affinity FBP (Kd=10−11 M −1), anchored to surface membranes of mature human erythrocytes by a hydrophobic residue, was first identified by Antony and colleagues [12,13]. Incubation of erythrocytes with trypsin resulted in the release of soluble hydrophilic FBP into the external medium [12,13]. Indirect evidence suggested that erythrocyte FR is an α isoform; it immunoreacts with antibodies against human placental FR , which represents a mixture of α and β isoforms , and erythrocyte membranes do not show immunostaining with antibodies against FRβ . Our ELISA results shown in Figure 1, obtained with antibodies specifically directed against FBPα [26,27], provide direct evidence for the presence of FRα in mature erythrocytes.
Despite the removal of endogenous folate, free as well as receptor-bound, by repeated washes at low pH, Antony et al. , who used a radiolabelled folate preparation with a high specific activity, only succeeded in identifying two molecules of functional FBP per cell in reticulocyte-rich young cells. A progressive, up to 7-fold, loss of functional FBP was observed as erythrocytes aged . No folate was released after boiling of the washed erythrocyte membranes, indicating a complete removal of endogenous folate . Hence, the extremely small number of functional receptors explains why we failed to detect any binding of our low-specific activity [3H]folate to haemolysates, despite repeated washes and prolonged dialysis to remove endogenous folate.
The functional role of erythrocyte FR is not clear, but the addition of anti-(placental FR) serum induced a decrease of intracellular folate and megaloblastic changes during erythropoiesis in vitro . Mature erythrocytes contain polyglutamated forms of folate, suggesting that folate is taken up at a more immature stage of erythrocyte development, since the enzyme polyglutamate synthetase is only present in immature cells . Human haemoglobin has been identified as a low-affinity/high-capacity binding protein for folate , and the binding of folate polyglutamates to haemoglobin has also been reported .
Several studies have focussed on the presence of different isoforms of FBP in mature neutrophil granulocytes. A FRβ has thus been identified on the surface of mature neutrophils and subpopulations of CD34+ cells in the myelomonocytic cell line by use of immunostaining and flow cytometry with anti-(human FRβ) antibodies [14–16,22]. No FRα is present on the surface membrane, as shown by flow cytometry with monoclonal antibodies against FRα , as well as by our indirect immunofluorescence results (Figure 3). The FRβ had a variant GPI membrane anchor, which is insensitive to PI-PLC, but cleaved by nitrous acid . The FRβ in mature normal granulocytes remained non-functional, despite acid washing to remove endogenous folate, whereas it was active in acute myeloid leukaemia cells with 0.72 pmol of folate bound per 106 cells . The physiological role of FRβ expressed during neutrophil maturation is obscure, but in CD34+ cells it may serve a function that is distinct from its role in folate uptake, i.e. signal transduction and cell adhesion . The FRβ in myeloid leukemia cells is up-regulated 20-fold at the level of mRNA by all-trans retinoic acid , suggesting that the receptor could be a promising target for therapeutic intervention, for example with folate-conjugated cytotoxins, due to its selective up-regulation in leukaemic cells [14,22].
In addition to the FRβ on the surface, neutrophil granulocytes contain a cytoplasmic functional high-affinity (Kd=<1 nM) FBP only present in detectable amounts in a minority of healthy subjects [17–20]. It was originally found in women who were pregnant or taking oral contraceptives , but the present results (Figure 2), as well as other studies [19,20], reveal great inter-individual variations in the concentration of functional FBP in both males and females, and clustering of persons with high concentration levels in certain families . The protein seems to be located in the matrix of secondary neutrophil granules, and similar to lactoferrin, is released there from during degranulation induced by high lithium concentrations [35–38]. It is thought to be identical with the secretory FBPγ isoform, which is expressed in haematopoietic tissue and lacks a hydrophobic GPI residue [23,24]. The appearance of a single peak of radioactivity at 29 kDa in the gel-filtration profile of the leucocyte lysates (Figure 2) is consistent with a γ isoform possessing no GPI tail.
The ELISA results in Figure 2 showed virtually similar concentrations of FBPα in three samples of leucocyte lysates, despite widely different concentrations of functional FBP. Hence, FBP/FRα seems to contribute very little, most probably not at all, to the total content of functional FBP in the granulocyte. Indirect immunofluorescence results (Figure 3) revealed that FRα, similar to lactoferrin, is localized to a cytoplasmic structure in neutrophils. The exact nature of this structure is difficult to establish, since the ethanol-fixation procedure disrupts cytoplasmic membranes. A putative locus could be secretory vesicles, since their membranes contain several GPI-linked proteins [39,40]. One may argue that subpopulations of haematopoietic stem and progenitor cells do not contain FRα mRNA . This finding, however, does not exclude the presence of quiescent FRα mRNA having been or becoming active at other stages of differentiation.
At least two isoforms of FBP seem to be present in human serum. A functional high-affinity FBP with a molecular mass of 29 kDa which is similar to that of functional FBP in leucocytes (Figures 2 and 6). Figure 4 strongly suggests that this FBP is mainly derived from neutrophil granules [17–20].
An α isoform of FR/FBP that contributes very little to the total concentration of functional FBP in serum (Figures 5 and 6). It appears as a minor peak at 43 kDa that continues into a broad immunoreactive peak, reaching its maximum in void volume (>120 kDa). This isoform is present in serum at a low concentration, as judged from its weak immunoresponse, and could represent fragments of FRα from degraded erythrocytes and neutrophils.
To summarize, the present report and above-mentioned studies have revealed a rather complex expression pattern of FBP/FR isoforms in human blood cells. FRα on reticulocyte/erythrocyte membranes seems to mediate folate uptake and to be of great importance for erythropoiesis. The functional FRβ on membranes of leukaemic granulocytes has been utilized for therapeutic targeting, whereas its non-functional counterpart on membranes of normal neutrophil granulocytes could be a putative signal protein. At this point, it is difficult to assign a biological/functional role to the cytoplasmic FR/FBPα that has been identified in neutrophil granulocytes.
The biological role of the high-affinity FBP in serum is still unresolved. Similar to other soluble nutrient- and vitamin-binding proteins in serum, for example, lactoferrin and vitamin-B12-binding protein which are released during degranulation of neutrophil granulocytes [35–38], FBP may exert bacteriostatic effects by depriving bacteria of endogenous folates . Binding to FBP protected tetrahydrofolate, the endogenous form of folate in pig serum, from biological degradation, whereas low-affinity binding to albumin did not . Therefore serum FBP may ascertain transport of intact folates to target organs/tissues.