Comparative Analysis of Cobalamin Binding Kinetics and Ligand Protection for Intrinsic Factor, Transcobalamin, and Haptocorrin*

Changes in the absorbance spectrum of aquo-cobalamin (Cbl·OH2) revealed that its binding to transcobalamin (TC) is followed by slow conformational reorganization of the protein-ligand complex (Fedosov, S. N., Fedosova, N. U., Nexø, E., and Petersen, T. E. (2000)J. Biol. Chem. 275, 11791–11798). Two phases were also observed for TC when interacting with a Cbl-analogue cobinamide (Cbi), but not with other cobalamins. The slow phase had no relation to the ligand recognition, since both Cbl and Cbi bound rapidly and in one step to intrinsic factor (IF) and haptocorrin (HC), namely the proteins with different Cbl specificity. Spectral transformations observed for TC in the slow phase were similar to those upon histidine complexation with Cbl·OH2 and Cbi. In contrast to a closed structure of TC·Cbl·OH2, the analogous IF and HC complexes revealed accessibility of Cbl's upper face to the external reagents. The binders decreased sensitivity of adenosyl-Cbl (Cbl·Ado) to light in the range: free ligand, IF·, HC·, TC·Cbl·Ado. The spectrum of TC·Cbl·Ado differed from those of IF and HC and mimicked Cbl·Ado participating in catalysis. The above data suggest presence of a histidine-containing cap shielding the Cbl-binding site in TC. The cap coordinates to certain corrinoids and, possibly, produces an incapsulated Ado-radical when Cbl·Ado is bound.

Intrinsic factor (IF), 1 transcobalamin (TC) and haptocorrin (HC) are three proteins involved in assimilation and transport of cobalamin (Cbl) in an organism (1). They all have extraordinary affinity to the physiologically active forms of Cbl with K d Ͻ 1 pM (1-4) but exhibit different selectivity toward the nonfunctional Cbl analogues. IF and, to some extent, TC are sensitive to variations in the structure of the ligand, which helps these proteins to discriminate between the physiologically active and inactive corrinoids (2,5). On the contrary, HC can successfully bind many defective corrinoids lacking even the whole nucleotide moiety (2,5). Binding to the carriers shields the lower part of the Cbl molecule (also called ␣-site), which contains the nucleotide. On the contrary, the upper surface of Cbl (␤-site) with the active group is thought to be open, as judged from its reactivity with the external compounds in the case of holo-IF and holo-HC (6 -8).
Low amounts of the Cbl-binding proteins available from the natural sources (1,6,9,10) hampered their investigations until several binders were successfully expressed in the recombinant organisms (11)(12)(13)(14)(15). The sufficient amounts of both bovine and human transcobalamin were obtained from the recombinant yeast Pichia pastoris. It allowed to establish the structure of the disulfide bridges in bovine TC (14) and investigate in detail Cbl⅐OH 2 binding to human TC by a stopped-flow technique (15). It was shown that the association between TC and Cbl⅐OH 2 occurred in two steps, when the initial attachment to an open conformation of the protein was followed by transition to a closed conformation with the shielded upper face of Cbl. As a result of this transition, cobalt-coordinated water in Cbl⅐OH 2 was thought to be displaced by a protein residue. The suggestion was supported by the fact that the external compounds coordinated to the ␤-position of TC⅐Cbl⅐OH 2 at exceedingly slow rates. The described features, however, appeared to be characteristic only for Cbl⅐OH 2 interacting with TC, whereas binding of Cbl⅐OH 2 , for instance, to HC occurred in one step (15). Cobalamins with the tightly associated ␤-groups (Cbl⅐CN and Cbl⅐N 3 ) bound both to TC and HC in one step as well (15).
The experiments on Cbl⅐OH 2 interaction with TC and HC suggested a correlation between high specificity of the carrier for Cbl and the biphasic nature of the binding reaction. In this paper we, therefore, investigated the rapid kinetics of Cbl⅐OH 2 binding to the most Cbl-specific protein IF. Interaction of two other ligands (Cbl⅐Ado and an analogue Cbi) with IF, TC, and HC was also characterized. We found no correlation between the ligand specificity and the biphasic kinetics of binding. Slow spectral transformations were observed only for two ligands, Cbi (this paper) and Cbl⅐OH 2 (15), when interacting with TC. The character of these changes was identical to those induced by coordination of external histidine to Cbi or Cbl⅐OH 2 . This fact supports the hypothesis of a cobalt-coordinated histidine residue within the complexes between TC and certain corrinoids. We also addressed the accessibility of bound Cbl to the external cobalt-specific reagents in different protein complexes. The results suggest that the ␤-surface of Cbl associated with IF or HC is moderately open, in contrast to practically closed complex with TC. Binding of Cbl⅐Ado to the proteins protected to some extent this ligand from light-induced decomposition. In addition, the absorbance spectrum of TC⅐Cbl⅐Ado alluded to homolytic cleavage of the carbon-cobalt bond in 10 -20% of the associated ligand molecules.

Materials
All salts and media components were purchased from Merck, Roche Molecular Biochemicals, Sigma, and Beckton Dickinson. The enzymes and kits for DNA handling were obtained from New England Biolabs, Stratagene, and Roche Molecular Biochemicals, the kit for the PCR reaction was from HT Biotechnology Ltd. Oligonucleotides were synthesized by DNA technology. The employed yeast expression system was purchased from Invitrogen. The fermentor Biostat B from B. Braun Biotech International was employed during expression of the recombinant proteins. Sephacryl S-200 and CM Sepharose were obtained from Amersham Biosciences, Inc. The anti-IF serum was raised by DACO.

Methods
Preparation of the DNA Material for Expression of Human IF-IFencoding fragment of DNA was produced from a gastric RNA by the reverse transcriptase and polymerase chain reactions employing IFspecific primers with adaptors for XhoI and NotI endonuclease sites. The obtained product was purified and ligated to the corresponding sites in the expression plasmid pGAPZ␣. The designed sequence of the fusion protein contained the following components counting from the N terminus: a yeast secretion signal (␣-factor), the site for yeast protease Kex2, and the mature human IF. This construction ensured cleavage of the N-terminal peptides from the recombinant protein during its secretion: . . . LEKR2STQTQ . . . , IF residues are underlined.
Expression and Purification of Human IF-The recombinant IF was expressed according to recommendations of the manufacturer (Invitrogen) in yeast P. pastoris (strain SMD 1168). The constitutive promoter of glyceradehyde-3-phosphate dehydrogenase induced the expression. The fermentation of the recombinant yeast was carried out at 30°C for 2 days in 1 liter of YPD medium (containing 0.5 M Cbl⅐OH 2 ) with the constant supply of glucose. The level of oxygen and pH in the medium were maintained at 25% and 6.0, respectively. The cell-free supernatant was saturated with ammonium sulfate (520 g/liter) and centrifuged at 4,000 ϫ g for 40 min. The pellet was dissolved in 50 ml of 0.05 M P i buffer, pH 7.5, whereupon centrifuged once more at 12,000 ϫ g for 10 min. The solution was concentrated by ultrafiltration to the volume 10 -15 ml and applied to a 250-ml Sephacryl S-200 column equilibrated with 0.1 M Tris, 1 M NaCl, pH 7.5. The fractions with red protein were pooled, concentrated to 5-8 ml, and subjected to repeated gel filtration under analogous conditions. The red fractions with IF were collected, concentrated, and stored frozen. SDS electrophoresis, staining of the gel by Coomassie, staining of the glycoproteins by PAS method, and Western blotting were performed according to the standard procedures.
Expression and Purification of Human TC-The recombinant TC was produced as described in our previous publication (15).
Isolation of Human HC-The protein was purified from human plasma as described elsewhere (17).
Preparation of the Apo Forms of Cbl Binders-Holo forms of IF, TC, and HC were dialyzed against 5 M GdnHCl (IF and TC) or 8 M GdnHCl (HC) at 30°C for 2 days with one change. Liberation of Cbl was monitored visually. The proteins were renaturated by overnight dialysis against 0.2 M P i buffer, pH 7.5, 5°C.
Spectral Measurements-The spectra were recorded on M350 Double Beam UV Visible Spectrophotometer (Camspec) or on the stopped-flow equipment, see the next paragraph.
Stopped-flow Experiments on Cbl Binding-Binding of different corrinoids to the specific apo-proteins was followed on DX.17MV stoppedflow spectrofluorometer (Applied Photophysics) using difference in the absorbance spectra of the ligands in their free and bound state, see Ref. 15. The reactions were performed in 0.2 M P i , pH 7.5, at 20°C.
Dissociation of the Protein-Ligand Complexes-When dissociation of Cbl⅐OH 2 from its protein complexes was investigated, the holo-protein (20 M) was mixed with Cbl⅐CN (100 M) and incubated at room temperature for 4 days. The samples of 0.15 ml were collected at different time intervals, suspended for 1 min with charcoal (pellet from 0.3 ml of 1% solution), and centrifuged for 1 min at 15,000 ϫ g. Supernatants were centrifuged once more for 5 min. The loss of the protein due to adsorption on charcoal did not exceed 15%. Spectra of the proteinassociated Cbls were recorded, whereupon displacement of Cbl⅐OH 2 by Cbl⅐CN was measured according to the ratios A 361 /A 330 , A 365 /A 335 , and A 363 /A 330 for IF, TC, and HC, respectively. The transition spectra were compared with those of protein⅐Cbl⅐OH 2 and protein⅐Cbl⅐CN complexes to establish completeness of the reaction.
Dissociation of protein-Cbi complexes (20 M) was initiated by adding 20 M Cbl⅐OH 2 . The measurements were carried out as described above except for the registration wavelengths: A 500 /A 580 , A 515 /A 580 , and A 500 /A 590 for IF, TC, and HC, respectively.
Exchange of the Cobalt-coordinated Groups in the Corrinoids-The displacement of the cobalt-coordinated groups in Cbi, Cbl⅐CN, IF⅐Cbl⅐OH 2 , and HC⅐Cbl⅐OH 2 (20 -25 M) by the external ligands (CN Ϫ , N 3 Ϫ and histidine) was followed spectroscopically in 0.2 M P i buffer, pH 7.5, at 20 or 37°C. Transformation of Cbl⅐Ado to Cbl⅐OH 2 (either free or bound to IF, TC, and HC) was induced by light when the sample in a quartz cuvette was placed in front of a 30 W daylight lamp at the distance of 20 cm. The changes in absorbance were measured at 352 nm (free Cbl) and 359 nm (protein-bound Cbl) with 1-min intervals.
Mathematical Analysis-Fitting of the curves was performed by a computer program for nonlinear regression analysis 2 or a program Gepasi (18). The presented data were obtained from two to four parallel experiments and are shown as the mean values.

RESULTS
Purification of the Cbl-binding Proteins-Details of the isolation procedures for human HC and human recombinant TC were described elsewhere (15,17). Human recombinant IF was expressed in yeast P. pastoris and purified as described under "Methods." The gel filtration profile of the purified recombinant IF contained one protein peak of 70 kDa saturated with Cbl⅐OH 2 . SDS electrophoresis in the presence of a reducing agent revealed the major protein pool of 50 -55 kDa (Fig. 1, lane 2), which was reactive toward IF-specific antibodies (Fig. 1,  lane 4). The determined N-terminal sequence was identical to human gastric IF (STQTQS . . . ). The 50-55 kDa band was not sharp, probably because of variation in the composition of carbohydrates coupled to the protein core of IF. The presence of carbohydrates on recombinant IF was confirmed by PAS staining of the gel (Fig. 1, lane 3).
All three Cbl binders were purified in complex with Cbl⅐OH 2 and preparation of apo-proteins required treatment with GdnHCl followed by a renaturing step. The regained binding capacity corresponded to 80 -90% (TC), 60 -70% (HC), and 30 -40% (IF), when compared with the initial amounts of the bound Cbl.
Changes in Cbl Absorbance Upon Its Binding to the Cblspecific Proteins-Association of Cbls with IF, HC, and TC caused typical changes in the ligand spectrum ( Fig. 2) (6,8,14,15). The extinction coefficients of Cbl⅐OH 2 in complex with the proteins investigated are shown in Table I. These data were obtained on the originally purified holo forms as well as on the GdnHCl-treated, renatured and resaturated proteins. GdnHCl treatment had certain effects on the extinction coefficients. It 2 S. N. Fedosov, unpublished data. was particularly evident for IF where all peaks increased by 15-20% (Table I) mainly due to intensified absorbance of the apo-protein (Fig. 2, B and C). The corresponding changes were insignificant for TC and practically absent for HC (Table I). The spectra of the free ligands are given for comparison in Fig. 2D.
The most significant shifts in the absorbance spectra of all ligands took place after their association with TC ( Fig. 2, solid lines): (i) for Cbl⅐OH 2 one can see a noticeable red shift for all peaks ( Fig. 2A); (ii) for Cbl⅐Ado this was a distortion of the shape at 350 -380 and 480 -550 nm (Fig. 2B); (iii) for Cbi there occurred an unusual redistribution of intensities from A 540 / A 580 Ͻ1 to Ͼ1 (Fig. 2C). Curiously enough, the spectrum of TC⅐Cbl⅐Ado reminded very much of those for Cbl⅐Ado acting as a cofactor in glutamate mutase (19) and methylmalonyl-CoA mutase (20) under steady-state conditions. On the contrary, the spectra of analogous complexes with IF and TC were similar to glutamate mutase⅐Cbl⅐Ado in rest (19).
Development of Slow Spectral Distortions-All spectral changes induced by the binding of Cbl⅐OH 2 , Cbl⅐Ado, and Cbi to the specific proteins were accomplished in less than 1 s, except for the pairs TC ϩ Cbl⅐OH 2 and TC ϩ Cbi. Those cases attracted our special attention.
During the binding of Cbl⅐OH 2 to TC, the initial jump of the ␥-peak was followed by continual spectral changes during the next 3 min (Fig. 3A). These slow perturbations significantly contributed to initially moderate red shift and amplification of the ␥-peak. The process developed exponentially in time with the rate constant of 2.5 ϫ 10 Ϫ2 s Ϫ1 , which did not differ from k ϩ2 obtained earlier in the stopped-flow experiments at a single wavelength (15). Increase of the temperature essentially accelerated the slow phase but had no affect on its amplitude (at least between 20 and 37°C).
The slow decrease of absorbance for the ␥-peak, induced by attachment of Cbi to TC, was not very noticeable due to the originally high absorbance of Cbi at 350 -370 nm (not shown). The spectral transition was more evident for the smaller ␣and ␤-peaks (Fig. 4A), and the effect was expressed better at 37°C than at 20°C. The transition was exponential with the rate constants of 4.3 ϫ 10 Ϫ3 s Ϫ1 (20°C) and 1.1 ϫ 10 Ϫ2 s Ϫ1 (37°C).
Imitation of the Slow Phases by Coordination of Histidine to Cbl⅐OH 2 and Cbi-We have suggested earlier that the unusual spectral behavior of Cbl⅐OH 2 during binding to TC may have been caused by coordination of a protein residue to cobalt (15). The control experiment with several amino acids and free Cbl⅐OH 2 showed that only incubation with histidine caused noticeable spectral response (Fig. 3D, solid line), at least under the shown conditions. This is not surprising since imidazol is a known ligand with intermediate affinity to Cbl (21). The reaction between histidine and Cbl⅐OH 2 was reversible and characterized by the rate constants k ϩHis ϭ 0.92 M Ϫ1 s Ϫ1 and k ϪHis ϭ 2.2 ϫ 10 Ϫ4 s Ϫ1 (K His ϭ 0.24 mM) at pH 7.5 and 20°C. At higher temperature (37°C) the rate coefficients increased 2.0 -2.2-fold without significant change in the equilibrium constant K His .
Addition of 5 mM histidine to either IF or HC complex with Cbl⅐OH 2 caused gradual shift of the ␥-peak (Fig. 3, B and C, respectively) analogous to the reaction between histidine and free Cbl⅐OH 2 (Fig. 3D). All above processes were similar in  their manifestation to the second phase observed for TC ϩ Cbl⅐OH 2 interaction (compare Fig. 3, B and C, dashed lines, with A). The rate coefficients of the forward reaction determined for IF⅐Cbl⅐OH 2 and HC⅐Cbl⅐OH 2 were equal to 0.44 M Ϫ1 s Ϫ1 and 0.05 M Ϫ1 s Ϫ1 , respectively. The complex TC⅐Cbl⅐OH 2 did not react with histidine for at least 2 h (not shown). Addition of histidine to Cbi also evoked spectral changes (Fig. 4B), which testified for coordinatioin of the imidazol group to either ␣ or ␤ surface of the corrinoid (21). The recorded spectra reversibly mirrored those during displacement of the dimethyl-benzimidazol base by cyanide: Cbl⅐CN ␤ ϩ CN Ϫ N CN ␣ ⅐Cbl⅐CN ␤ (Fig. 4C). This may suggest attachment of histidine to the lower axial site of Cbi. The half-maximal optical response was reached at His ϭ 20 mM (not shown). The apparent rate coefficient of the process k ϩapp ϭ 0.021 s Ϫ1 (20°C), k ϩapp ϭ 0.077 s Ϫ1 (37°C), was, however, practically independ-ent on histidine concentrations at His ϭ 5-100 mM. This means that the velocity of conversion CN⅐Cbi⅐CN 3 His⅐Cbi⅐CN is not limited by attachment of histidine to cobalt, although, the details of kinetics are not quite understood. Coordination of 15 mM histidine to Cbi (Fig. 4B) caused the same type of the spectral response as the binding TC ϩ Cbi (Fig. 4A). Protection of the TC-associated Cbi was not as good as for Cbl⅐OH 2 , and addition of 15 mM histidine caused further spectral transition with the velocity 14 times slower, than for free Cbi (not shown).
Binding Kinetics of Different Corrinoids-The change in the absorbance of Cbls and Cbi upon their attachment to the proteins was used to monitor these processes on stopped-flow equipment. The data depicted in Fig. 5, A-C, represent the rapid phase of the binding. The reactions were fitted to a bimolecular mechanism A ϩ B 3 C with the rate constants shown in Table II. Ligand binding to IF was characterized by  (Table II).
As was already mentioned, the initial attachment of Cbi to TC was followed by a slow monomolecular reaction C 7 D. The detected decrease of the ␥-peak was difficult to follow at 12 M TC due to low response on the background of a relatively high absorbance. Therefore, the time course of this second phase was recorded at increased concentrations of TC and Cbi (both 40 M) and at another wavelength corresponding to ␣-peak (Fig. 5D). The rate coefficients, determined from continuous measurements, were 5.6 ϫ 10 Ϫ3 s Ϫ1 (20°C) and 1.4 ϫ 10 Ϫ2 s Ϫ1 (37°C). They did not differ from the data in Fig. 4A. Thorough investigation of Cbl⅐Ado binding to TC at different concentrations and wavelengths did not reveal any additional phase in this process besides the spectral changes during the first 10 ms induced by attachment of the ligand to the protein.

Dissociation of the Ligand-Protein
Complexes-High velocity of association between Cbi and the recombinant IF or TC raised a question about their affinity to this analogue, since Cbi is known to be a poor substrate for IF and TC from the natural sources (1, 2, 5, 16). We have, therefore, characterized dissociation of the protein⅐Cbl⅐OH 2 or protein⅐Cbi complexes by gradual replacement of the original ligand with added Cbl⅐CN or Cbl⅐OH 2 , respectively (Fig. 6, A and B). The process was followed in time by the spectral changes of the protein fraction after charcoal treatment.
The data in Fig. 6A show the reaction, where a 4-fold excess of Cbl⅐CN was added to the holo-proteins saturated with Cbl⅐OH 2 . Computer simulation of the curve obtained for IF allowed to calculate the dissociation rate constants both for Cbl⅐OH 2 (k ϪCblOH ϭ 4.2 ϫ 10 Ϫ6 s Ϫ1 ) and Cbl⅐CN (k ϪCblCN ϭ 9.2 ϫ 10 Ϫ6 s Ϫ1 ), using the known values of k ϩCbl from Table II. The values of k ϪCblOH for dissociation of the corresponding TC and HC complexes were estimated from the initial slopes (v ϭ k ϪCblOH [complex]) as 1 ϫ 10 Ϫ7 s Ϫ1 and 6 ϫ 10 Ϫ7 s Ϫ1 , respectively. Our previous measurements of k ϪCblCN for bovine and human TCs at higher temperature (37°C) were in the range of 1 ϫ 10 Ϫ6 to 3 ϫ 10 Ϫ6 s Ϫ1 (4,15).
When the apo forms of recombinant IF and TC were saturated with Cbi and exposed to equal concentration of external Cbl⅐OH 2 , the complete substitution occurred in less than 1 min (Fig. 6B, upper curves). No detectable dissociation of HC⅐Cbi was found under the same conditions (Fig. 6B, lower curve). The rate constants of Cbi liberation were estimated as k ϪCbi Ͼ 5 ϫ 10 Ϫ2 s Ϫ1 (IF, TC) and k ϪCbi Ͻ 1 ϫ 10 Ϫ5 s Ϫ1 (HC).
Exchange of the ␤-Group in Cbl⅐OH 2 Associated with IF or HC-It has already been shown that accessibility to the upper  face of the ligand in the TC⅐Cbl⅐OH 2 complex is hindered (15). In this assay we exposed IF (Fig. 7, A and B) and HC (Fig. 7, C and D), saturated with Cbl⅐OH 2 , to different concentrations of CN Ϫ or N 3 Ϫ and then followed replacement of the original ␤-group by changes in the absorbance. The observed reactions were practically irreversible in the case of CN Ϫ and reversible for N 3 Ϫ . The calculated rate coefficients are shown in Table III, where the previous results for TC⅐Cbl⅐OH 2 and free Cbl⅐OH 2 (15) are given for comparison. As one can see, neither IF nor HC rendered significant protection against CN Ϫ . At the same time, coordination of N 3 Ϫ to cobalt was somewhat decelerated in both directions when compared with free Cbl⅐OH 2 .
Specific Proteins Protect Cbl⅐Ado against Light-induced Decomposition-When Cbl binders saturated with Cbl⅐Ado were exposed to light, a gradual transformation of Cbl⅐Ado to Cbl⅐OH 2 was observed (Fig. 8). The time course of these photoactivated reactions was monitored spectroscopically and compared with decomposition of free Cbl⅐Ado under analogous conditions. The performed measurements showed a 7-, 15-, and 17-fold deceleration of Cbl⅐Ado decay when the ligand was bound to IF, HC, and TC, respectively.

DISCUSSION
Binding of the Cbl molecule to the specific proteins affects its absorbance spectrum, which turns spectroscopy to an easy and convenient method for monitoring the protein-ligand interactions. The advantages of the method were used for the investigation of Cbl binding to three transporting proteins: IF, TC, and HC. Two first binders were expressed in recombinant yeast, and HC was purified from human plasma. All proteins were isolated as holo forms with bound Cbl⅐OH 2 and their absorbance spectra ( Fig. 2A, Table I) were typical for the binders from other sources. GdnHCl treatment, necessary for production of the apo-proteins, had practically no effect on the spectra of HC and TC. At the same time, the treatment influenced IF, and the increased absorbance of the apo-protein (Fig.  2, B and C) resulted in artificially high extinction coefficients of the newly bound ligand (Table I). The earlier determined extinction parameters of gastric IF (8) were, nonetheless, closer to the overrated absorbance of recombinant holo-IF after GdnHCl than to the coefficients of "fresh" recombinant holo-IF (Table I). This observation stresses importance of IF's history for its spectral features.
Comparison of the data in Fig. 2, A-C, with Fig. 2D showed that the most pronounced alterations in the ligand spectra took place after binding to TC (solid lines). Thus, the record for TC⅐Cbl⅐OH 2 at pH 7.5 (Fig. 2A) demonstrated a remarkable red shift of the ␥-peak (362 nm) and strong expression of the ␣-peak (546 nm). This pattern mimicked better Cbl⅐CN or Cbl⅐imidazol than Cbl⅐OH 2 at neutral pH (21). The spectrum of another complex TC⅐Cbl⅐Ado (Fig. 2B) was characterized by increased optical density at 350 -380 and 400 -450 nm as well as by decreased absorbance at 500 -550 nm accompanied by separation of the individual peaks ␣ and ␤. Similar spectra were observed for enzyme-bound Cbl⅐Ado during catalysis (19,20), which may suggest partial homolysis of the carbon-cobalt bond also in TC⅐Cbl⅐Ado, not trivial for a transporting protein like TC. The complex of TC with the third ligand Cbi (Fig. 2C) likewise revealed some redistribution of intensities between the peaks, i.e. decrease of the ␣-peak (578 nm) and increase of the ␤-peak (544 nm). Analogous spectra can be observed, for instance, for Cbl during transition (base on) ␣ ⅐Cbl⅐CN ␤ N (base off)_CN ␣ ⅐Cbl⅐CN ␤ , when cyanide and the nucleotide base compete for the lower coordination position at cobalt (Fig. 4C).
The peculiar spectra of the above protein-ligand complexes prompted us to thorough investigation of the binding kinetics. Change in the absorbance of the ␥-peak during the ligand binding was followed by a stopped-flow technique (Fig. 5, A-C). All ligands, including the analogue with the missing nucleotide moiety, attached rapidly and in one step to two proteins with widely different Cbl specificity, IF and HC (Fig. 5, A and B). There was no visible indication of any second phase during 100 s of the binding as well, unlike the interaction between TC and Cbl⅐OH 2 examined earlier (15). This fact implies that the slow phase is not an attribute of the selective recognition of Cbl but rather a specific characteristic of TC, when interacting with certain ligands. We doubt that presence of carbohydrates on IF and HC (1,6) and their absence on TC (1, 6, 15) has anything to do with the described effects, because glycosylation does not seem to interfere with the binding of Cbl to IF (22).
There was an interesting observation, concerning the high velocity of association between Cbi and IF or TC. The incomplete ligand bound to these two proteins, known to be Cbl selective (1,2,4,5), almost as quick as the ligand with the correct structure, i.e. Cbl. We compared the association rate constants from Table II with the collision rate constant k coll ϭ 5 ϫ 10 9 M Ϫ1 s Ϫ1 at 20°C estimated for a corrinoid and a binding site of appropriate geometry (23). The calculations showed that the number of the efficient impacts varied from 2 to 20 per 1000 collisions without particular correlation between k ϩCbl and the ligand structure. Similar rate constants found for Cbi and Cbl mean that the Cbl-specific site is not originally tuned to any particular ligand and can accommodate for a time being even some defective molecule. The sensitivity of IF and TC to the substrate's geometry, and its absence in the case of HC, was revealed only in the dissociation experiments (Fig. 6, A and B).
Calculated values of the rate constants allowed us to make the following estimates of K d for Cbl⅐OH 2 : 1 pM (IF), 0.01 pM (HC), and 0.005 pM (TC). The values for Cbi were: K d Ͻ 0.1 pM (HC), K d Ͼ 1 nM (TC and IF). The earlier published K d for Cbl and the specific binders varied in the range 10 Ϫ16 -10 Ϫ9 M (see, for review, Refs. 1-4, 8, 9, 15, and 22), which could hardly be explained by real fluctuations of the affinity. Such a broad dispersion was rather caused by inappropriate mathematic approach to the case when the total concentrations of the binding site E T and the ligand L T are close to each other (complicated by K d Ͻ Ͻ E T , L T ). Under these conditions, half-saturation would be reached at L T(0.5) ϭ K d ϩ 0.5 E T , which may represent rather concentration of the binding site than the dissociation constant. More accurate presentation of the results as EL versus L free may also lead to an erroneous evaluation of K d if the reaction is almost irreversible. Under these circumstances even a small but reproducible overestimate of L free (e.g. L app ϭ L free ϩ 0.05 L T ) inevitably causes great overestimate of K d (e.g. half-saturation at L app(0.5) ϭ 1.05 K d ϩ 0.025 E T Ϸ 0.025 E T ). In such a difficult case, determination of K d from the ratio of the rate constants k ϪL /k ϩL may be advantageous. This statement can be illustrated by comparison of K d measured for chicken HC in an equilibrium assay (10 Ϫ13 M) and from k ϪL /k ϩL (10 Ϫ16 M) by the same authors (3).
Investigation of the rapid kinetics contributed to our understanding of the substrate binding, although, it did not give us a clue to the anomalous appearance of the holo-TC spectra. Therefore, the stopped-flow experiments were repeated at different wavelength and higher concentration of TC. They did not exhibit any second phase for the reaction TC ϩ Cbl⅐Ado but revealed it for TC ϩ Cbi (Fig. 5D). This result demonstrated that Cbl⅐OH 2 was not the only corrinoid characterized by biphasic binding to TC (15). Two atypical ligands (Cbl⅐OH 2 and Cbi) were subjected to thorough analysis. We have recorded deformations of the ␥-peak for Cbl⅐OH 2 (Fig. 3A) and ␣,␤-peaks for Cbi (Fig. 4A) at 20 and 37°C in an attempt to get the best response for each corrinoid. In both cases the initial attachment of the ligands caused slight increase and red shift of the peaks without significant distortions of their shape (see the 1-s records in Figs. 3A and 4A). Continuation of the reactions was, although, accompanied by more pronounced changes in the Ϫ (curves 1-7, respectively).

TABLE III
The rate constants of ␤-exchange in Cbl⅐OH 2 when free or bound to the specific proteins (37°C) spectra, similar to those observed during exchange of the cobalt-coordinated groups in Cbl⅐OH 2 and Cbi. This observation raised, in its turn, a question about the nature of cobalt-coordinated groups in the ligands associated with the transporting proteins.
When Cbl binds to a transporter, its lower part becomes buried inside the protein molecule, whereas the upper part is thought to be open in all carriers under study (7,8,24). At the same time, our analysis of the ␤-group exchange in free and bound Cbl⅐OH 2 confirmed this statement only for IF and HC. The mechanism of the exchange reaction for these two binders was generally the same as for Cbl⅐OH 2 in solution except for somewhat reduced reaction velocities (Fig. 7, Table III). Protection of the upper surface of Cbl⅐OH 2 in holo-TC was much more evident. For instance, coordination of 1 mM CN Ϫ or N 3 Ϫ to holo-TC (15) was 2 orders of magnitude slower than the same reactions with holo-IF/-HC (Fig. 7). The accessibility of Cbl's upper plane was suggested to be hindered due to coordination of a protein residue to cobalt at upper axial position (15). Several amino acids have been tested on Cbl⅐OH 2 for their ability to cobalt coordination (Fig. 3D), but only histidine appeared to be the sufficiently potent compound. This result was quite anticipated, because adsorption of Cbl⅐OH 2 on the proteins with His residues during extraction of Cbl from biological sources is a well known phenomenon (25). We made an attempt to imitate the slow phase by adding histidine to holo-IF and -HC, i.e. the proteins with the open upper surface of Cblc⅐OH 2 . The following spectral changes reflected transition to IF⅐/ HC⅐Cbl⅐His (Fig. 2, B and C) and bore an essential resemblance to the second phase of Cbl⅐OH 2 binding to TC (Fig. 2A).
The interpretation of the spectral changes during association of Cbi and TC seems to be more complex. It is clear that displacement of cyanide from either lower or upper position by a histidine residue of TC is quite feasible (compare Fig. 4, A and  B). We have also suggested that histidine coordinates to the ␣-site of Cbi (Fig. 4, A and B) because of resemblance with the reaction (base on) ␣ ⅐Cbl⅐CN ␤ N (base off)_CN ␣ ⅐Cbl⅐CN ␤ (Fig.  4C). Some aspects remain, however, unclear. Thus, the observed substitution of ␣(?)-cyanide in TC⅐Cbi was incomplete as followed from comparison with the Cbi spectrum at saturating histidine (Fig. 4B). This result did not match the accomplished replacement of ␤-water in the slow phase of the binding reaction TC ϩ Cbl⅐OH 2 (15). The activation energies of the second phases for TC ϩ Cbi (43 kJ/mol) and TC ϩ Cbl⅐OH 2 (120 kJ/mol) also differed quite significantly. In other words, there may be different histidine residues involved in ␤-substitution on Cbl⅐OH 2 and ␣(?)-substitution on Cbi. Still, we cannot exclude that a disc-shaped Cbi molecule binds to TC upside down with the ␣-site exposed to the same His-containing domain as the ␤-site of Cbl⅐OH 2 .
Another still unraveled issue is the binding of Cbl⅐Ado to TC. The spectrum of the produced complex TC⅐Cbl⅐Ado was stable in time and the associated ligand was well protected against light (Fig. 8). At the same time, appearance of the ␣,␤-peaks (Fig. 2B) mimicked Cbl⅐Ado-dependent enzymes under catalysis (19,20) when ϳ20% of Cbl molecules contain detached Ado⅐ radical involved in the substrate transformation. Nonetheless, we did not find any additional phase in the binding reaction, TC ϩ Cbl⅐Ado, which could be potentially ascribed to homolytic cleavage of the carbon-cobalt bond. The apparent absence of the second phase might be caused by high velocity of cleavage estimated, for instance, in the case of methylmalonyl-CoA mutase as Ͼ600 s Ϫ1 (20). The ability of TC to induce formation of the Ado⅐ radical may be not as absurd as it seems to be at first sight. Thus, alignment of the pairs (IF, TC, or HC):(methylma-lonyl-CoA mutase or glutamate mutase) showed 15-19% of homology in all cases, although, at different regions. Anyway, the unusual spectral properties of the TC⅐Cbl⅐Ado complex require additional analysis.
The presented data throw some light upon the structure of the binding sites of Cbl transporting proteins. One can imagine that all three carriers are supplemented with a lid-or a cap-like structure at the entrance to the site cavity. At the same time, development of this structure in the Cbl transporters appears to be different. The cap in IF does not seem to cover any appreciable part of the upper face of Cbl and, therefore, holo-IF demonstrates quite rapid exchange of the ␤-groups as well as a relatively moderate protection of Cbl⅐Ado. The kindred protein HC is rigged somewhat better. Its cap shields to a certain extent the upper plane of Cbl and hinders the inwards-outwards movements of the ␤-coordinated groups, at least for bulky complexons. Analogous cap in TC renders much better protection against all substituents, and it might even produce and stabilize the Ado⅐ radical above the upper plain of Cbl. In addition, the protective shield of TC is, presumably, supplemented by an active His residue, which can coordinate to cobalt and dislodge ␤-water in Cbl⅐OH 2 or ␣(?)-cyanide in Cbi (the latter case requires, although, additional clarification). Coordination to cobalt at the ␤-position locks the lid above the binding site and Cbl (but not Cbi) becomes encapsulated inside holo-TC, with only occasional and short-time openings to occur.
The performed investigation strengthens the view on TC as the best protector of the associated Cbl. It also raises a question about the role of the protective cap in stabilization and destabilization of the cobalt-coordinated groups in TC-bound corrinoids.