Combination of Specific Antibodies with the Human Vitamin A-transporting Protein Complex*

SUMMARY The interactions of specific antibodies with the human retinol-binding protein (RBP) and the thyroxine-binding prealbumin have been investigated. When an antibody combines with either of these proteins, the tryptophyl fluorescence of the antibody is partially quenched. Polarization of fluorescence of RBP-retinol increased rapidly on formation of antibody-RBP complexes. With fluorescence-quenching and polarization measurements as indicators, the stoichiometries and equilibrium constants of the antigen-antibody reactions have been studied. The data show that RBP, whether free or in complex with prealbumin, exhibits identical reactivity with anti-RBP Fab’-fragments. The conclusion is reached that the prealbumin binding site of RBP is not a major antigenic structure. The data indicate that the number of antigenic sites on RBP is limited and is the factor which controls the number of antibodies bound per RBP molecule. quantum yield measurements of retinol any measureable conformational change on complex formation with

of Uppsala, Uppsala, Sweden SUMMARY The interactions of specific antibodies with the human retinol-binding protein (RBP) and the thyroxine-binding prealbumin have been investigated. When an antibody combines with either of these proteins, the tryptophyl fluorescence of the antibody is partially quenched.
Polarization of fluorescence of RBP-retinol increased rapidly on formation of antibody-RBP complexes.
With fluorescencequenching and polarization measurements as indicators, the stoichiometries and equilibrium constants of the antigenantibody reactions have been studied.
The data show that RBP, whether free or in complex with prealbumin, exhibits identical reactivity with anti-RBP Fab'fragments.
The conclusion is reached that the prealbumin binding site of RBP is not a major antigenic structure.
The data indicate that the number of antigenic sites on RBP is limited and is the factor which controls the number of antibodies bound per RBP molecule.
By quantum yield measurements of retinol it was found that RBP does not undergo any measureable conformational change on complex formation with Fab'-fragments.
Free prealbumin could simultaneously interact with a maximum number of 12 anti-prealbumin Fab'-fragments. This suggests that there are three antigenic sites on each of the four identical prealbumin subunits. RBP, on forming a complex with prealbumin, competed with four anti-prealbumin Fab'-fragments.
These Fab'-fragments showed a limited heterogeneity indicating that they may be directed toward a single antigenic site. It is proposed that the competition between RBP and the anti-prealbumin Fab'-fragments may be interpreted as a consequence of a negative cooperativity on RBP forming a complex with prealbumin. The equilibrium constant for Fab'-fragments forming a complex with prealbumin was lowered on thyroxine binding to prealbumin.
This result supports the earlier suggestion that thyroxine binding to prealbumin is an example of a negative homotropic interaction.
Antibodies are known to react in a highly specific manner with their antigen.
According to current concepts the antibody com-* This work was supported by grants from the Swedish Medical Research Council  and Stiftelsen Svensk Naringsforskning.
1 To whom to address all correspondence.
bining sites are complementary to the patterns of the antigenic determinants and conformational changes of an antigenic site should thus give rise to observable perturbations in the antigenantibody reaction.
On the other hand, accumulating evidence (l-5) suggests that antibodies may also induce conformational changes in the antigen.
The use of immunological techniques thus provides a powerful tool to establish conformational identity and to detect changes in conformation.
We report here the application of immunological techniques to the study of complexes of antibodies and the human vitamin A-transporting protein complex (6-8).
Since the constituents of the protein complex, the retinol-binding protein and prealbumin, are able to bind ligands, retinol and thyroxine, respectively, which quench protein tryptophyl fluorescence, we employed fluorescence measurements to determine the effects of the interaction of specific monovalent antibodies with the antigens. Furthermore, retinol is a natural fluorescent reporter group bound to a single specific site in RBP,' the fluorescence of which may give information about the structure of its environment in the protein.
The data obtained in these studies indicate that the structure of RBP is not altered on complex formation either with prealbumin or with specific Fab'-fragments. The interaction between prealbumin and thyroxine, on the other hand, perturbed the prealbumin complex forming with specific monovalent antibodies.

Materials
Proteins-Prealbumin was isolated as described elsewhere (10). The prealbumin-RBP complex was isolated according to a published procedure (8). RBP was isolated from the prealbumin-RBP complex (10) or from urine of patients with tubular proteinuria by means of affinity chromatography on a prealbumincoupled Sepharose column (11).
Other Materials-Sephadex G-100 and G-200 and Sepharose 4B, products of Pharmacia Fine Chemicals, were used according to the instructions supplied by the manufacturer. All other chemicals were the best available grade from commercial sources.

Methods
Preparation of Fab'-fragments--The antisera were prepared in rabbits by a schedule of injections previously described (12). Specific antibodies were prepared in the following way. The 4041 antisera were mixed with agarose immunoadsorbents. These immunoadsorbents were prepared by binding of prealbumin or RBP to Sepharose 4B (13) according to the method of Cuatrecasas (14). In a typical preparation, antigen-Sepharose immunoadsorbent (10 mg of antigen per g of Sepharose) was dispersed in the appropriate antiserum.
The adsorbed antibodies were eluted with 0.2 M glycine HCl buffer, pH 2.9, and the eluate was immediately titrated to pH 8.0 with 1.0 M Tris.
By this procedure between 1 to 2 mg of specific antibodies were obtained per ml of antiserum.
To avoid the complication in the fluorescence experiments of turbidity resulting from precipitate formation, the purified antibodies were digested with 1% by weight of pepsin at pH 4.5 as described by Nisonoff et al. (15). The digestion was stopped by raising the pH to 8 with 1 M Tris. The digest was then applied on a Sephades G-200 column (2 x 130 cm), equilibrated with 0.02 M Tris-HCl buffer, pH 8.0, containing 0.15 M NaCl, to isolate (Fab')z-fragments free from undigested immunoglobulin. Monovalent Fab-fragments were prepared from the (Fab'lzfragments by reduction with 0.01 M dithiothreitol for 1 hour and alkylation with 0.022 M iodoacetamide in the dark for 30 min. The Fab'fragments were freed from excess reagents and small amounts of aggregated protein by means of gel chromatography on Sephadex G-100 equilibrated with 0.02 M Tris-HCl buffer, pH 8.0, containing 0.15 M NaCl.
Part of the highly purified Fab'-fragments was then applied to an immunoadsorbent column to test their antigen-binding activity.
The adsorbed fragments were eluted with 0.2 M glycine HCl buffer, pH 2.9. Approximately 95% of both the anti-RBP and the anti-prealbumin Fab'-fragments were capable of binding specific antigen as judged by this method.
Fluorescence illeasurements-Most measurements were carried out with an Aminco-Bowman spectrophotofluorometer. Polarized fluorescence was estimated in a Zeiss ZFM4C spectrophotofluorometer equipped with double monochromators. Fluorescence was maximal when the excitation wave length was 285 nm and the fluorescence at 340 nm was measured.
Therefore, this combination of wave lengths was used for protein fluorescence measurements. No correction was made for the wave length dependence of either the light source or the photomultiplier output.
Protein concentrations were low enough so that there was little absorption of either the exciting or fluorescence radiation.
All measurements were carried out at room temperature (23 f 2"). Quantum yields of retinol fluorescence were determined with quinine sulfate in 0.1 M HzS04 as a reference substance and taking 0.55 as its quantum yield (16). For these measurements the spectra recorded were corrected for instrumental variations (17).
Calculations-The stoichiometry of binding of antibodies to the protein antigens was determined by titrating a solution of monovalent antibodies with antigen.
Protein tryptophyl fluorescence decreases progressively with antigen addition until a maximum is reached.
This quenching is probably due to radiative transfer to retinol and thyroxine (17). The intersection of the initial slope of the plot of percentage of maximal quenching against added antigen concentration and the line of maximum quenching defines the reaction stoichiometry, i.e. the number of antibodies that bind to a single antigen molecule.
The equilibria of macromolecular reactions can be quantified by fluorescence techniques (18). If the binding sites are heterogeneous and can be characterized by a Sips' distribution of binding free energies (19), the mass law can be expressed (20) Log + q a log c + a log K, The total concentration of combining sites of the antibodies is n, and r is the concentration of the average number of sites on the antibodies occupied at the concentration c of the free antigen. A plot of log r/(n -T) versus log c over a sufficient range of c will test the adequacy of the Sips' distribution.
If the plot is linear the values of a, the index of heterogeneity, and Ko, the average association constant, can be directly established (20).
From fluorescence polarization data /3, the fraction of fluorescence reactant bound, may be evaluated from the following equation (21) where 2, Aa, and A,, the average observed anisotropy, the allbound, and all-free anisotropies are direct experimental values. The fluorescence anisotropy, A, is defined by and the average observed anisotropy x is where f and 6 refer to the all-free and all-bound forms of the fluorescence reactant and I is the fluorescence intensity in the polarization measurements. R, the ratio of the fluorescence yields per exciting quanta, can be directly determined by measuring separately the polarized intensities I ,, and I for the free and the all-bound fluorescent reactant R q (I,,+ 2i&/(i II On rearrangement of the mass law derived from Sips' equation (18,19) where Fb max is the maximum value of the molar concentration of the bound fluorescent ligand, and X is the total concentration of the fluorescent ligand. When prealbumin, Fab'-fragments against prealbumin, and RBP are simultaneously present, there will be a competition between Fab' and RBP for the sites of prealbumin.
The mass law for the competition-type experiment is hence where d is the effective concentration of Fab'-fragments, i.e. the concentration divided by the stoichiometric reaction number, and KR the apparent association constant of prealbumin and RBP.
Ultracentrifugations-Molecular weights were determined at 20" in a Spinco model E analytical ultracentrifuge equipped with an RTIC temperature control unit and an electronic speed control. All samples were dissolved in the appropriate buffers and dialyzed in the cold against two changes of the solvent.
Densities were determined by pycnometry.
Six-channel Epon-filled Yphantis centerpieces and sapphire windows were used throughout. Recordings were made with the photoelectric scanning system set at 280 or 330 nm. The sedimentation equilibrium experiments were performed by means of the low speed method of Richards and Schachman (22). Speed settings and equilibrium times were estimated as described by Teller et al. (23). The ex-4042 periments were discontinued when no redistribution of material could be observed over a period of several hours.
Calculations of apparent weight average molecular weights were computed from the following equation (22).
where the symbols have their usual meaning.
The value 0.72 was used for the partial specific volume of RBP and Fab-fragments (7). Local weight average molecular weights were obtained as described by Yphantis (24) over five equally spaced 2 coordinates.
Portions (200 ~1) of these antisera, diluted 1: 5, were incubated with various amounts of the antigens. The mixtures were allowed to react for 2 hours at 25" followed by 48 hours at 4". The precipitates were collected by centrifugation at 15,000 X g for 30 min. The supernatants were discarded whereas the precipitates were washed repeatedly with ice-cold 0.15 M NaCl.
The washed and dried precipitates were dissolved in 0.5 M NaOH and the protein content was estimated by measuring the absorbance at 280 nm.
Other d8ethods-Affinity chromatography of RBP on prealbumin-coupled Sepharose was accomplished as described elsewhere (11). Estimations of the affinity between prealbumin and thyroxine were performed by use of the Colowick and Womack method of rate of dialysis (26). The details have been given elsewhere (27).
Protein and retinol concentrations were determined by relating the absorbance at 280 and 330 nm, respectively, to the relevant extinction coefficients (8). Diluted protein solutions were concentrated by ultrafiltration (28) with use of the Visking dialysis tubing, 23 X 32 inches (Union Carbide Corp., Chicago, Ill.), as the ultrafiltration membrane (29). The negative pressure was not allowed to exceed 400 mm Hg in order to obtain protein recoveries of 857, or more.

RESULTS
Antigenic Reactivity of Prealbumin, RBP, and Prealbumin-RBP Protein Complex-With use of specific antisera directed against RBP and prealbumin, respectively, the immunological reactivity of the individual proteins and of the prealbumin-RBP complex was investigated.
The quantitative precipitin technique was employed and the results are summarized in Fig. 1. It is evident from the figure that RBP when free exhibited the same antigenic characteristics as when bound to prealbumin.
However, a clear difference was noted between free and complex-bound prealbumin. The precipitin curves suggest that the prealbuminRBP complex reacts with a smaller number of antibodies than free prealbumin.
The results from the quantitative precipitin experiments were corroborated by binding of the antibodies to insolubilized antigen. Table I shows that prealbumin could bind a greater number of monovalent prealbumin antibodies than could prealbumin-RBP. Virtually identical binding behavior was noted when monovalent RBP antibodies were reacted with either immobilized RBP or prealbuminRB1' (Table I). It is, however, difficult to investigate the reaction stoichiometry from these experiments, since it has been showu that only a fraction of the immobilized antigen is reactive (11). The amount of bound monovalent antibodies was estimated by measuring the absorbance at 280 nm of the dissociating eluate (0.2 M glycine-HCl buffer, pH 2.9).  RBP (or prealbumin) was added to the prealbumin (or RBP)coupled Sepharose to form a noncovalent interaction with the matrix-bound protein.
The amount of protein added wassufficient to saturate all accessible binding sites on the protein coupled to Sepharose.
It is obvious from the above results that RBP and anti-prealbumin Fab-fragments compete for similar binding sites on prealbumin.
To isolate the specifically competing Fab'-fraction a column of prealbumincoupled Sepharose was saturated with anti-prealbumin Fab'-fragments. After having washed the column with the equilibrating buffer until the eluate was virtually devoid of protein, RBP was applied.
It can be seen from Fig. 2 that Fab'-fragments were displaced from the column by RBP. On lowering the ionic strength of the eluting buffer the prealbuminbound RBP was eluted from the column, and by lowering the pH, the remaining Fab'-fragments were released (Fig. 2) Fro. 2. Purification of anti-prealbumin Fab' I and II fragments by immuuoadsorption.
The column was washed with the buffer and when the eluate was virtually devoid of protein a solution of RBP (0.8 mg per ml) in the equilibrating buffer was applied. When a constant protein to retinol ratio (estimated by the absorbance at 280 and 330 nm, respectively) was obtained the eluting buffer was applied.
To detach prealbumin-bound RBP, 0.002 M Tris-HCl buffer, pH 8.0, was applied to the column (solid arrow). Fab' II fragments were obtained by eluting the column with 0.2 M glycine HCI buffer, pH 2.9 (stippled arrow).
Fractions of 3 ml were collected at S-min intervals. effect the quantum yield of the retinol fluorescence was measured for mixtures of Fab'-fragments and RBP in varying molar ratios, The values obtained from the quantum yield measurements were identical (0.04) within experimental error irrespective of the presence or absence of Fab'-fragments.
It is well known that the fluorescence of retinol decays with time, probably due to molecular alterations under the influence of ultraviolet light. Fig. 3 shows the effect of prolonged irradiation on the retinol fluorescence.
It is evident from the figure that prealbumin, when bound to RBP, shelters the vitamin moiety in part from the irradiative effects. Fab'-fragments directed against RBP, on the other hand, do not seem to exert any effect of similar kind (Fig. 3). Within the limit of the techniques employed, it may thus be concluded that Fab'-fragments bound to RBP neither affect the prealbumin-RBP interaction nor influence the microenvironment of retinol. The percentage of the maximally quenched fluorescence is plotted against the ratio of total RBP concentration to total Fab'-fragment concentration.
Since prealbumin binds thyroid hormones, its affinity for thyroxin was investigated in the presence of specific Fab'-fragments. Bv rate of dialysis it was shown that the apparent association constant of thyroxine and the high affinity site of prealbumin (27) appeared identical in the absence of Fab'-fragments (1.9 x lo7 M-I) and the presence of saturating amounts of the monovalent antibodies (1.7 x 10' M-I).
This result suggests that the Fab'-fragments do not induce any conformational change of prealbumin affecting its interaction with thyroxine.
Stoichiometric Estimations-Aliquots of a solution of anti-RBP Fab'-fragments were titrated with RBP or with the prealbumin-RBP complex.
The results are shown in Fig. 4. The percentage of the maximum of the quenched fluorescence is plotted as a function of the ratio of the antigen concentration to the total antibody concentration.
The line of the final fluorescence quenching at complete reaction intersects the initial slope of the plot at an antigen to antibody ratio of approximately 0.25 both for RBP and prealbumin-RBP.
This implies an average total binding of 4 Fab' molecules per RBP molecule.
Due to the specific absorbance of retinol at 330 nm it was possible to measure the reaction stoichiometry of Fab'-fragments and RBP also by sedimentation equilibrium ultracentrifugation.
In separate experiments it was found that both RBP and Fab'-fragments appeared homogeneous with weight average molecular weights of 21,000 and 46,000, respectively.
Mixtures of Fab'-fragments and RBP (within the molar range of 8 to 1 for the initial concentrations of the individual components) showed, as expected, a heterogeneous behavior in the ultracentrifuge.
At high ratios of Fab'-fragments to RBP, the limiting local weight average molecular weight approached a value of 200,000, close to the espected theoretical value of 205,000 for a reaction stoichiometry of 4:l. Both fluorescence measurements and sedimentation ultracentrifugation thus gave similar results.
The titrations of anti-prealbumin Fab'-fragments with thyroxine-containing prealbumin gave plots of qualitatively similar appearance as that shown for RBP (cf. Fig. 4). The calculated reaction stoichiometry is presented in Table II. As can be seen in the table an average total binding of 3 Fab' molecules per prealbumin subunit (12 per prealbumin molecule) was obtained. Fab' I and Fab' II fragments exhibited an average maximal binding of 4 and 8 molecules, respectively, per prealbumin molecule. These data then suggest that RBP compete with Fab'-fragments directed against one-third of the prealbumin antigenic determinants.
Equilibrium Constants jw Fab'-fragments and RBP, Prealbumin, 4044 and Prealbumin-RBP-The equilibrium constants for the various reactions were calculated from the fluorescence-quenching experiments described above, or, for reactions involving RBP, by polarization of the retinol fluorescence. The polarization of the fluorescence of RBP-retinol increased rapidly as specific Fab'fragments were added to the system (Fig. 5) and the polarization approached a limiting value of 0.385. The corresponding value for the prealbumin-RBP retinol complex, when saturat,ed with Fab'-fragments, was 0.40. The fluorescence spectra of RBPretinol and prealbumin-RBP-retinol and of their corresponding complexes with Fab'-fragments were identical in both polarization and fluorescence yield (see above). The fraction bound of the fluorescent antigen, p, was determined by the use of Equation 2. From the plots of /3 against molarity of FaWfragments added, at a constant concentration of RBP-retinol and prealbumin-RBP-retinol, respectively, the number of moles bound per mole of antigen were obtained (30) (see Table II and Fig. 6). It is also evident from Fig. 6 that RBP-retinol and prealbumin- RBP-retinol exhibited identical react.ions with the anti-RBP Fab'-fragments.
This was further corroborated by determinations of the association constant for the reactions.
In both cases the best fit to the data was obtained by values for K. of approximately 3 x lo7 M-I and for a of 0.63. In conclusion, these data suggest that the prealbumin binding site of RBP is not one of the major antigenic sites. Furthermore, no conformational change of RBP is evident on complex forming with either Fab'-fragments or prealbumin (31) and both interactions appear to be quite independent.
The association constant for prealbumin and specific Fab'fragments was evaluated from the fluorescence-quenching experiments since prealbumin does not contain any fluorescent moiety suitable for polarization measurements.
A logarithmic Sips' plot (Equation 11 for Fab' II fragments and prealbumin and prealbumin-RBP, respectively, is shown in Fig. 7. Different values for K. and a were obtained.
It is obvious from Fig. 7 and Table  II that the association constant for Fab' II and prealbuminthyroxine is markedly less than for Fab' II and prealbumin-RBP.
The binding of Fab' I to prealbumin was estimated both by fluorescence quenching, with use of prealbumin-thyroxine, and by RBP competition experiments evaluated by fluorescence polarization (Equation 7).  . Table II).

DISCUSSION
The human vitamin A-transporting protein complex is intriguing in view of the many molecular interactions pertaining to this system. Since antibodies against prealbumin and RBP are not cross-reacting and are directed against various sites on the two proteins, they seemed to be suitable tools for evaluation of some of the characteristics of the binding processes. An inherent limitation to this approach is, however, that the antibodies, formed in rabbits, are produced only against antigenic structures of the injected proteins differing from the endogenous counterparts of the animal.
This means that there are two possible factors controlling the stoichiometry of the binding of antibodies to the antigen (32). The first is a limitation imposed by there being a finite number of antigenic sites. The second is a steric restriction; if enough antibodies bind to the antigen, they will completely cover its surface, making it impossible for other antibodies to approach.
In the latter case the number of antibody binding sites would exceed the binding stoichiometry.
The former alternative gains support from the finding that antibodies against RBP react identically with the free antigen and the pre-albuminRBP complex. Prealbumin, which is somewhat larger than the FalLfragments, binds to RBP irrespective of the presence of anti-RBP Fab'-fragments and it may thus be concluded that the prealbumin binding site of RBP is not antigenic.
The lack of antigenicity of the prealbumin binding site of RBP may be explained assuming that the prealbumin binding site of RBP has been conserved during evolution (33). Amino acid substitutions in this region of RBP may require complementary mutations in the gene for prealbumin to yield proteins with sustained ability to interact.
The great specificity in the interaction of RBP and prealbumin is strikingly shown by the fluorescence experiments reported here. The average of four simultaneously bound Fab'-fragments to RBP do not appreciably change the fluorescence characteristics of retinol compared to free RBP-retinol whereas the interaction of prealbumin and RBP causes profound effects. The equilibrium constants for RBP and each of the two types of proteins are of similar magnitude indicating that the differences encountered for the binding of prealbumin and Fab'-fragments, respectively, are not merely related to binding strength.
Prealbumin could simultaneously bind an average of 12 Fab'-fragments.
Since prealbumin is composed of four identical subunits (10) it may be inferred that there are three independent antigenic sites per polypeptide subunit. These numbers are identical with those found for hemoglobin (32) which has about the same molecular size as prealbumin.
In view of the molecular symmetry of the arrangement of the prealbumin subunits (34) it is conspicuous that RBP compete with four of the Fab'-fragments for binding to prealbumin.
A possible interpretation for this observation is that the RBP binding site of prealbumin corresponds to one or two of the antigenic sites and that RBP binds to prealbumin in a negatively cooperative manner. Accordingly, this means that the prealbumin molecule contains multiple (two or four) RBP binding sites. On forming a complex with RBP, a conformational change in prealbumin could alter the corresponding antigenic sites so that they no longer are recognized by the antibodies.
Recent results obtained in this laboratory give support to this view since it has been found that the free prealbumin subunit can interact with RBP.2 Furthermore, the index of heterogeneity (a) is significantly higher for the monovalent antibodies competing with RBP than for the rest of the anti-prealbumin Fab'-fragments.
This limited heterogeneity may point to the fact, that the anti-prealbumin Fab' I fragment are directed against a single antigenic determinant.
The anti-prealbumin Fab' II fragments reacted with eight sites on prealbumin.
It is interesting to note that the equilibrium constant for the reaction was dependent on the presence of thyroxine.
We had suggested earlier (27) and shown recently that thyroxine binds to prealbumin with a negative cooperativity.2 This invokes that after 1 molecule of thyroxine has formed a complex with a prealbumin subunit, the thyroxine binding sites of the other subunits should change their structure so that further complex formation of thyroxine with prealbumin should be impeded.
The lowered affinity of the Fab' II fragments for prealbumin may thus be interpreted as a result of a slight conformational change in part of the prealbumin molecule.
It may be inferred that the suggested conformational change is not a result of thyroxine binding but RBP binding to prealbumin. This was, however, excluded by performing the titration in the presence of both thyroxine and RBP. The affinity constant obtained was similar to that found when prealbumin-thyroxine was used as the antigen.
There is no direct competition between thyroxine and the Fab' II fragments since the high affinity thyroxine binding site of prealbumin exhibits the same equilibrium constant for thyroxine whether a large excess of anti-prealbumin Fab-fragments are present or not. Assuming that a conformational change occur in prealbumin on forming a complex with thyroxine, this means that only part of the molecule is altered.
This can be deduced from the fact that RBP complexes with prealbumin irrespective of its thyroxine binding status (27,35). Furthermore, the binding characteristics for the Fab' I fragments are obviously not perturbed by the presence of thyroxine bound to prealbumin.
The generality of the differences encountered has not been established.
The antibodies were obtained from pools of three animals.
The different reactivity of Fab' II fragments toward prealbumin with and without thyroxine may not be a general phenomenon.
Similarly, other details of our results such as equilibrium constants and the magnitudes of the differences found for the binding of a given antibody preparation to different antigens may well vary from one preparation to the next. However, the ability of fluorescence methods to detect small changes in 4046 antibody binding and differences in ecmilibrium constants that 18.
result from small conformational changes in the structure of an antigen has been amply shown. 19.