Structure and Function of Ovotransferrin I. PRODUCTION OF IRON-BINDING FRAGMENTS FROM IRON-OVOTRANSFERRIN BY THE ACTION OF IMMOBILIZED SUBTILISIN. PURIFICATION AND CHARACTERIZATION OF THE FRAGMENTS*

Immobilized subtilisin Novo was used for the cleavage of iron-saturated ovotransferrin (FezOT) into separate NHzand COzH-terminal iron-binding fragments, designated as FeNF and FeCF, respectively. The M, of each fragment is 39,000. The purified fragments show major differences in the content of histidine, alanine, and methionine. Both apo-NHzand apo-COzH-terminal fragments are able to bind one ferric ion per molecule. FeCF is more resistant than FeNF to dissociation at acid pH and to subtilisin action. FeNF and FeCF are immunochemically distinct. However, equal mixtures of the two show immunochemical reaction indistinguishable from intact Fe20T. The ironbinding sites of FeNF and FeCF are very similar to each other on the basis of visible absorption and CD spectra. The major difference in the backbone conformations between FeNF and FeCF is in the a-helical content of FeCF which is twice that of FeNF. Individually, fragments show quantitative differences in the electron paramagnetic resonance spectra; however, equal mixture of the two fragments produce EPR spectra very similar to that of the intact FezOT. These studies indicate that subtilitic cleavage of FezOT does not produce significant change in the ironbinding capacity or the conformation of the separated iron-binding domains.

1 nonbiological probes indicate that the two iron-binding sites in transferrins are nonequivalent (2).
One direct approach for proving the equivalence or nonequivalence of sites has been to release the individual ironbinding domains from the transferrins by means of chemical and proteolytic procedures, and to investigate them separately. The frst iron-binding fragment' was isolated from apoovotransferrin by cyanogen bromide cleavage, and represents the NH2-terminal domain of the molecule. The iron complex of the fragment, however, showed a significant change in the visible absorption spectrum, indicating a structural alteration in the iron-binding site (7, 8).
The proteolytic approach was initially introduced by Azari and Feeney (9), who showed that the saturated iron complexes of OT2 and serum transferrin are significantly more resistant to proteolysis by chymotrypsin and trypsin than OT. Similar observations were reported by Williams (10).
Proteolysis of partially saturated iron complex has been found to liberate iron-binding fragments; however, the type of fragments liberated depended on the source of transferrin, extent of its saturation with iron, and the nature of the protease.
For example, tryptic and chymotryptic digestion of partially iron-saturated human transferrin and OT have been shown to cause liberation of NF (10, ll), whereas tryptic cleavage of pig and bovine transferrins, either as iron-saturated or apoproteins, produces NF and CF (12, 13). NF and CF have also been obtained from human lactoferrin by pepsin, trypsin, and chymotrypsin action (14). The action of subtilisin on FezOT releases only CF (151, whereas only N F is released from a thermolysin digest of iron-saturated human transferrin (16).
For the investigation of the iron-transferrin activity of N F and CF in a homologous chicken embryo red cell system (17), it was necessary to produce sufficient quantities of pure fragments possessing the structural integrity and iron-binding activity of the native molecule. In our experience, the procedures described by Williams (10,15) systematically gave low yields of NF, and in the case of CF, the subtilisin digest gave multichain iron-binding fragments which were contaminated with residual active subtilisin, even in the presence of PMSF. I In this report, iron-binding fragments NF and CF denote the iron binding domains from the NHs-and C02H-terminal halves of OVOtransferrin.

177
This is an Open Access article under the CC BY license. To avoid contamination of fragments by subtilisin and to increase the yield, we have prepared and employed Sepharoseimmobilized subtilisin Novo. This modification was not only successful in eliminating contamination but also produced a good yield of NF and CF in a single operation. In this report, we present the subtilitic procedure for the production of NF and CF and their purification. The physical, chemical, and immunochemical properties of the fragments are also presented, some of which have not been reported previously. The iron-transferring activity of the fragments, in the chick embryo red cell system, is presented elsewhere (17). both fractions were essentially saturated with specifically bound iron. Peak I11 did not show absorbance at 465 nm and was not studied further. The yields were: peak I, 35%; peak 11, 43% peak 111, 12%.
DE-cellulose chromatography of pooled fractions from peak I1 produced three major peaks Sub I, Sub 11, and Sub 111, as shown in Fig. 1B. The protein distribution of eluted peaks was: Sub I, 35%; Sub 11, 28%; Sub 111, 37%. Fractions under each of these peaks were collected and further purified by rechromatography on the same column. The purified fragments were essentially homogeneous as judged by polyacrylamide gel electrophoresis in the discontinuous Tris-glycinate system ( Fig. 2 . 4 ) . SDS-gel electrophoresis patterns of FezOT, Sub I, Sub 11, and Sub 111, before and after CNBr cleavage, are shown in Fig. 2B. They all show a single band in SDS gel before CNBr cleavage. The molecular weight of FezOT is -78,000 and the molecular weight of each of the three ironbinding fragments -39,000. Reduction and carboxymethylation of Fe20T and the iron-binding fragments produced no additional bands (result not shown), indicating that they all have single-chain structures.
As seen upon CNBr cleavage, Sub I1 yields a major fragment of M, = 35,000, whereas Sub I1 and Sub I11 produce smaller fragments ranging in M, = 13,000 to 23,000. While Sub I1 and Sub I11 have almost identical CNBr cleavage patterns, they share no common CNBr fragment with Sub I. Furthermore, the CNBr cleavage pattern of FezOT is approximately equal to the sum of Sub I and Sub I1 or Sub I and Sub 111. Based on these results and the reported CNBr cleavage pattern of OT (7), it is suggested that Sub I is derived from the NHz-terminal domain of OT and is very similar to NF produced by tryptic action, whereas Sub I1 and Sub I11 are derived from the COZH-terminal domain of OT and represent CF.
Purification of N F from Tryptic Digest of Iron-OT-The Sephacryl and DE-cellulose procedures, described for a subtilitic digest, were also employed for the purification of NF (also designated as TN) from the tryptic digest of FeOT. Sephacryl elution also produces three peaks (similar to Fig.   U). The second peak contains the NF. The NF fraction was further chromatographed on a DE-cellulose column with 10 to 100 mM, pH 8, ammonium acetate gradient. The major peak was rechromatographed under the same conditions and produced a homogeneous NF. The purified NF has an A d A* ratio of -25 which is close to the value for the native Fe20T.
NF shows a single band in the discontinuous Tris-glycinate gel system. SDS-gel electrophoresis of the reduced and carboxymethylated NF also shows a single band of M, = 39,000.
This indicates that NF has a single-chain structure, representing approximately half the Fe20T molecule. Furthermore, CNBr cleavage of NF yields a major fragment with a M, = 35,000, which indicates that NF is derived from the NH2terminal domain of OT (7,8).
Amino Acid Composition, NH2-terminal Residues, and Carbohydrate Content of Iron-binding Fragments-The amino acid compositions of the iron-binding fragments are shown in Table I. The composition of OT is also shown for comparison. The amino acid compositions of the two NFs (produced by tryptic or subtilic digestion) are almost identical to each other and the same is true for the two CFs. In general, all fragments show close similarity in composition except for alanine, histidine, and methionine.
The NH,-tenninal amino acid residues of OT and the ironbinding fragments are also shown in Table I. While   the NH2-terminal amino acid determinations, dansylalanine was the only dansylamino acid detected on the polyamide plates for NFs (TN and Sub I), whereas for Sub I1 and Sub 111, besides threonine and aspartate, several unidentified fainter fluorescent spots were also observed.
Results of carbohydrate analysis of the iron-binding fragments are also included in Table I. It is seen that almost all the carbohydrate of OT is present in CF.
Immunodiffusion Patterns of FeNF, FeCF, and Fe2OT-Immunodiffusion patterns of FeNF, FeCF, and Fe20T against anti-Fe2OT are shown in Fig. 3. Continuous precipitin bands are seen in Fig. 3A when wells 2 and 3 (or 5 and 6) contained Sub I1 and Sub 111, respectively. Similarly, in Fig. 3B continuous precipitin bands are seen when wells 2 and 3 (or 5 and 6) contained Sub I and TN, respectively. These results indicate immunochemical identity between Sub I1 and Sub I11 and Sub I and TN. As is also seen in Fig. 3, A and B, precipitin bands between Fe20T (wells I and 2) and any of the fragments produced a spur which indicates a partial immunochemical identity of fragments with Fe20T. Fig. 3C shows a continuous precipitin band between a mixture of NF and CF (wells 2,3, 5, and 6) and Fe20T (wells I and 4). These results indicate that although NF and CF are immunochemically different and contain only a portion of the antigenic specificity of the OT molecule, their mixture provides a full antigenic specificity of the intact OT molecule.
Rate of Iron Release from the Fragments-Iron release from FeNF, FeCF, and Fe20T was induced by the addition of citric acid and followed by measuring the decrease in Ad%. Since TN and Sub I are essentially identical to each other, as are Sub I1 and Sub 111, results from only one of each group are shown. Fig. 4 shows that the release of iron from FeNF is faster than from FeCF, and the rate of iron release from Fe20T very similar to FeNF. The theoretical sum of the curves of the two fragments is substantially different from that of the intact protein, but very similar to the experimental curve of a 1:l mixture of FeNF and FeCF.
Iron-binding Capacity of NF and CF-Spectroscopic titration of OT, NF, and CF with iron citrate indicated a maximum binding of 1.98, 0.97, and 1.08 mol of iron/mol of protein, respectively.
Trypsin and Subtilisin Hydrolysis of NF and CF and Their Iron Complexes-As shown in Fig. 5 A, FeNF and FeCF are highly resistant to trypsin hydrolysis, as compared to the apofragments. In 20 min, the apo-fragments show a 75 to 80% loss in chromogenic activity, whereas their iron complexes show very little loss in chromogenic activity for the same time period. A further exposure of FeNF and FeCF to trypsin for 2 h shows only 4 and 3% loss in chromogenic activity, respectively. Fig. 5B shows the rate of loss of chromogenic activity of the fragments and their iron complexes in the presence of subtilisin. Once again, the iron complexes are more resistant to subtilisin hydrolysis than the apo-fragments. However, subtilisin destroys the chromogenic activity of both FeNF and FeCF at a faster rate than trypsin. Furthermore, FeNF is hydrolyzed significantly faster than FeCF.
Spectroscopic Studies-The visible absorption spectra (not shown) of Fe20T, FeNF, and FeCF show very simlar charge transfer bands with a maximum at 465 nm and essentially identical absorptivity. Fig. 6A shows the CD spectra of Fe20T, FeNF, and FeCF in the visible region. A strong negative CD band at 450 nm is observed for all three proteins. The presence of the CD band associated with the charge-transfer absorption band indicates that iron-binding is asymmetric. Compared to the corresponding absorption band, however, the CD band has shifted 15 nm to the blue. Two intense positive CD bands are also seen at 305 and 325 nm for all three iron proteins, with FeNF showing a stronger band at 305 and weaker band at 325, as compared to FeCF. The CD spectra in the 250 to 300 nm region are shown in Fig. 6B. The spectra of Fe20T and the two ironbinding fragments are very similar to one another. The negative CD bands at 286 and 277 nm, which could be due to protonated tyrosine and probably also tryptophan residues, are observed in all three spectra. A positive CD at 295 nm, which is very unusual, is also observed for Fe20T and the iron-binding fragments. Fig. 6C shows the CD spectra in the 195 to 245 nm region. For FeCF, the negative band at 218 nm is diminished and becomes a shoulder to the main band at 210 nm. The percentage of various structures in Fe20T and its iron binding fragments are shown in Table 11. Results from the two methods indicate that the percentage of P-sheet structure in Fe20T and in the two iron-binding fragments is about the same but there is a substantial difference in their a-helix content. FeCF shows a two-fold higher a-helical content than FeNF.
The EPR spectra of FeNF and FeCF are shown in Fig. 7A.
The spectrum of FeCF shows a main line around g = 4.2 and a strong shoulder on the low-field side of the main line. A broad signal around g = 8.8 is also observed. The spectra of FeNF also show a sharp signal around g = 4.2. However, its intensity is much lower than that of FeCF. The intensity of the strong shoulder and the broad line around g = 8.8 of FeNF, however, is higher than those of FeCF. Fig. 7B shows the EPR spectra of Fe20T and a 1:l mixture of FeCF and FeNF. The two spectra are closely superimposable.

DISCUSSION
Immobilized subtilisin seems to be the most efficient system to date for the production of native iron-binding domains from FezOT. The yield of fragments is -40%, which is significantly higher than by the tryptic procedure of Williams (10,15). Furthermore, FeNF and FeCF are obtained simultaneously by a single enzymatic digestion. Subtilisin digestion yields less NF (13%) than CF (27%), apparently due to the higher susceptibility of FeNF to further subtilisin digestion than FeCF (Fig. 5B).
Since mechanical force such as stirring and shaking may destroy the solid matrix of the immobilized enzyme and release the attached enzyme into the solution, the digestion was carried out in a column. Usually a small volume of concentrated FezOT solution (10%) was applied to the column. However, the procedure could be adapted to the continuous application of a larger volume of Fe20T. The coupling and hydrolysis were performed at least 4 times, and in each case, the coupled enzyme had a similar activity toward p-nitrophenylacetate (30 to 40% of the soluble enzyme) and produced similar hydrolytic fragments from Fe20T.
The accessibility of the substrate to the matrix-bound enzyme was improved by maintaining a continuous flow of substrate through the column. This procedure eliminated the necessity for the complete removal or inhibition of the enzyme after digestion. The finding that insignificant internal cleavage had occurred in the iron-binding fragments suggests that the reaction catalyzed by the solid state enzyme is more restrictive than in solution. The restrictive action of immobilized enzyme may be due to its decreased activity, as well as a change in its specificity.
It has been shown previously that a 35,000-dalton CNBr fragment is derived from the NHz-terminal domain of OT (7,8). Since both NFs (TN, prepared by tryptic procedure of Williams (IO), and Sub I, isolated in the present study) yielded a 35,000-dalton fragment upon CNBr treatment, it was concluded that they originated from the NHp-terminal domain of the OT molecule. On the basis of the finding that Sub I1 and Sub I11 are complementary to the NFs (TN and Sub I) in immunological properties and CNBr cleavage patterns, it was concluded that they are derived from the COpH-terminal domain of OT.
The production of NF by the tryptic digestion of partly iron-saturated OT is in general similar to the procedure of Williams (10); however, the M, = 39,000 found for NF in our investigation is -4000 higher than reported. Furthermore, the yield of NF is only 4%, and the final digest contains 20% FeZOT. One explanation for the difference may be the use of iron citrate in our investigation instead of iron-nitrilotriacetate (used by Williams) for the preparation of partly saturated FeOT complex. It has been reported by Donovan et al. (35) that iron nitdotriacetate complex promotes indiscriminate and anticooperative binding of iron to OT and in its absence the binding becomes positively cooperative.
A CF has also been isolated by limited tryptic digestion of an OT. Fe complex (15). According to this procedure, at pH 5, FezOT releases iron preferentidy from the NHz-teminal site to give OTFe, which upon digestion by trypsin produces CF. Attempts to produce CF by this procedure were unsuccessful. Instead of CF, we obtained NF (similar to tryptic digestion). While these results are in contrast to the acid-induced iron release data for FeNF and FeCF, which indicate a higher stability for FeCF, they are in accord with the iron release data for intact Fe20T, which indicate a significant increase in the rate of iron release from the COOH-terminal domain. (Fig.  4). The finding that the rate of acid-induced iron release from Fe20T is much faster than from a 1:l mixture of FeNF and FeCF suggests that the release of iron from the two sites of intact Fe20T is cooperative rather than independent.
NF and CF differ in amino acid and carbohydrate composition, CNBr cleavage pattern, rate of acid-induced iron release, and immunological properties. Despite the compositional and stability differences between the NF and CF, their iron-binding environments are very similar to each other as evidenced by the similarity in their absorption and the CD spectra in the visible region. The negative CD band of the charge-transfer transition is not located at 465 nm, as the corresponding absorption band, but is blue-shifted to 450 nm, which may indicate that the charge-transfer band consists of multiple components. Besides the negative CD band a t 450 nm, two positive CD bands are also observed at 305 and 325 nm. Since they are not observed in the apoproteins, they are also induced by iron binding. It has been suggested (36) that they are due to at least one of the disulfide bonds in OT which undergoes changes in dihedral angle upon iron-binding. The unequivocal assignment of CD bands in the 250 to 300-nm region is difficult. On the basis of studies of amino acids and proteins (37), the 295-nm band is most likely due to tryptophanyl residues. The 286 and 277-nm bands may arise from protonated tyrosyl and perhaps also tryptophanyl residues. The 253 nm band is assigned tentatively to disulfide bonds. The positive ellipticity of the 295 nm band is probably due to the superposition of the intense charge-transfer CD band around 305 nm to this region of the spectra.
The CD spectrum in the 190 to 250 nm region depends primarily on the secondary structure of the protein. The CD spectrum of Fe20T obtained in this region is very similar to that observed by Tan (36), except t,hat a more profound band around 218 nm was observed in the present study which is unusual and it may indicate the presence of high content of p structure. The a helix content of Fe20T calculated by the method of Chen et al. (23) is 14%, which is very close to the values of 15 and 16% reported by Tan (36) and Nagy and Lehrer (38), respectively.
The spectroscopic and immunochemical data strongly suggest that the proteolytic cleavage of Fe20T into separate domains has not induced significant changes in the conformation and iron-binding activity of the domains.

Iron-binding Fragments
of Ovotransferrin