Two Heme-binding Domains of Heme-regulated Eukaryotic Initiation Factor-2 a Kinase N TERMINUS AND KINASE INSERTION*

, In heme deficiency, protein synthesis in reticulocytes is inhibited by activation of heme-regulated a -subunit of eukaryotic initiation factor-2 a (eIF-2 a ) kinase (HRI). Previous studies indicate that HRI contains two distinct heme-binding sites per HRI monomer. To study the role of the N terminus in the heme regulation of HRI, two N-terminally truncated mutants, Met2 and Met3 (dele-tion of the first 103 and 130 amino acids, respectively), were prepared. Met2 and Met3 underwent autophosphorylation and phosphorylated eIF-2 a with a specific activity of approximately 50% of that of the wild type HRI. These mutants were significantly less sensitive to heme regulation both in vivo and in vitro . In addition, the heme contents of purified Met2 and Met3 HRI were less than 5% of that of the wild type HRI. These results indicated that the N terminus was important but was not the only domain involved in the heme-binding and heme regulation of HRI. Heme binding of the individual HRI domains showed that both N terminus and kinase insertion were able to bind hemin, whereas the C terminus and the catalytic domains were not. Thus, both the N terminus and the kinase insertion, which are unique to HRI, are involved in the heme binding and the heme regulation of HRI. m M ) at 30 °C for 5 min. Protein kinase assays were initiated by the addition of 5 m Ci of [ g - 32 P]ATP (3000 Ci/mmol), 50 m M ATP, and 100 ng of eIF-2, and the reactions were proceeded for 5 min at 30 °C. Reactions were terminated by the addition of SDS-sample buffer containing 1.25% b -mercaptoethanol. The phosphorylated polypeptides were separated on 10% SDS-polyacrylamide gels, stained with Coomassie Brilliant Blue G-250. The extents of eIF-2 a phosphorylation were visualized by autoradiography and quantitated by scintillation counting of the gel slices corresponding to eIF-2 a . Protein kinase assays of the S10 crude extract of each HRI protein in the absence of purified eIF-2 was also performed and was used for background subtraction. Western blot analysis of HRI and various domains was performed, using anti-HRI mABF monoclonal antibody (38) or anti-His 6 monoclonal antibody (Am-ersham Pharmacia Biotech), as described previously (39). Hemin-dependent Transformation of HRI in Rabbit Reticulocyte Ly-sates— Wt and Met2 HRI were synthesized in TNT rabbit reticulocyte lysate (RRL) at 30 °C for 15 min followed by 4 min pulse of [ 35 S]Met. The extents of transformation of HRI were analyzed by 10% SDS-PAGE of the 35 S-labeled HRI bound to Ni-NTA resins. The transformation of 35 S-labeled HRI at different concentrations of hemin in RRL was car- ried out fo r 1 h asdescribed previously (40). Spectral Analysis— The Wt and mutant HRI cells m M and were immunoaffinity The UV and visible spectra of heme-en- Wt and HRI were using spectrometer The salt buffer was blank. The UV and visible difference spectra of

In heme deficiency, protein synthesis in reticulocytes is inhibited by activation of heme-regulated ␣-subunit of eukaryotic initiation factor-2␣ (eIF-2␣) kinase (HRI). Previous studies indicate that HRI contains two distinct heme-binding sites per HRI monomer. To study the role of the N terminus in the heme regulation of HRI, two N-terminally truncated mutants, Met2 and Met3 (deletion of the first 103 and 130 amino acids, respectively), were prepared. Met2 and Met3 underwent autophosphorylation and phosphorylated eIF-2␣ with a specific activity of approximately 50% of that of the wild type HRI. These mutants were significantly less sensitive to heme regulation both in vivo and in vitro. In addition, the heme contents of purified Met2 and Met3 HRI were less than 5% of that of the wild type HRI. These results indicated that the N terminus was important but was not the only domain involved in the heme-binding and heme regulation of HRI. Heme binding of the individual HRI domains showed that both N terminus and kinase insertion were able to bind hemin, whereas the C terminus and the catalytic domains were not. Thus, both the N terminus and the kinase insertion, which are unique to HRI, are involved in the heme binding and the heme regulation of HRI.
Initiation of protein synthesis in reticulocytes is inhibited during heme deficiency as a result of the activation of hemeregulated eIF-2␣ kinase (HRI). 1 The activation of HRI is ac-companied by its autophosphorylation and the phosphorylation of eIF-2␣ at serine 51 (reviewed in Refs. 1 and 2). eIF-2 is the eukaryotic translational initiation factor that binds GTP and Met-tRNA f and is necessary for the formation of the 43 S preinitiation complex. The recycling of eIF-2 requires the exchange of GDP for GTP. Under physiological conditions, the affinity of eIF-2 for GDP is much higher than for GTP. This exchange reaction is catalyzed by eIF-2B, which is rate-limiting and which is present at 15-20% of the amount of eIF-2 in reticulocytes. The phosphorylation of eIF-2␣ by HRI renders eIF-2B nonfunctional, resulting in the shut off of protein synthesis (reviewed in Ref. 3).
The two dsRNA binding domains of PKR are located in the first 171 amino acids (2,(15)(16)(17)(18). Recently, it has been shown that the KI region of PKR is required for its autokinase, eIF-2␣ kinase activities (19,20), and substrate binding (20). Taylor et al. (21) have identified a cluster of amino acids important for autophosphorylation between the dsRNA binding motif and the first kinase domain of the PKR. Activation of PKR by dsRNA is achieved by its autophosphorylation (21). Furthermore, mutation of Thr-446 and Thr-451 in the PKR activation loop to Ala impaired the kinase activity of this protein (22). The exact mechanism of activation of PKR is not known, but it may be brought about by dimerization or conformational changes (2). The involvement of the N-terminal 184 residues in the dimerization of PKR has been demonstrated (23)(24)(25)(26)(27). Patel and Sen (28) have shown recently that the hydrophobic residues present in the N-terminal domain of PKR are required for dimerization and protein-protein interaction between PKR monomers.
The GCN2 protein contains a domain in the C terminus that is closely related in sequence to histidyl-tRNA synthase. This domain is essential for the activation of the protein kinase domain when uncharged tRNA accumulates due to amino acid starvation (14,30). Romano et al. (22) have demonstrated that autophosphorylation of Thr-882 and Thr-887, located in the activation loop of GCN2, is important for the kinase activity of GCN2. The C terminus of GCN2 expressed in Escherichia coli has been shown to bind yeast tRNA (31). The importance of dimerization in the activation and catalytic function of GCN2 has been demonstrated (30,32). Qiu et al. (33) have provided evidence indicating that the protein kinase domain and the C terminus are required for dimerization of GCN2.
HRI has been expressed in insect Sf9 cells using the baculovirus expression system. This recombinant HRI is an active autokinase and eIF-2␣ kinase and is regulated by heme. HRI expressed in insect cells can inhibit the protein synthesis in Sf9 cells (34). Baculovirus-expressed wild type (Wt) HRI and the inactive K199R mutant, in which the conserved Lys-199 in the catalytic domain II is changed to Arg, have been purified to homogeneity. Purified HRI is a homodimer and a hemoprotein with two distinct heme-binding sites. Binding of heme to the first site is stable, whereas binding to the second site is dynamic and is responsible for the rapid regulation of HRI by heme (35).
To further understand the mechanism of the regulation of HRI by heme, we have localized the heme-binding domains of HRI in this study. We found that N-terminal deletion mutants, Met2 and Met3 HRI, were active kinases. Met2 and Met3 HRI were less heme-responsive and had very low heme contents as compared with Wt HRI. Furthermore, both the N terminus and KI domains expressed in E. coli were able to bind heme, whereas the kinase catalytic domains and the C terminus were not. These results show that the N terminus and the KI region of HRI are involved in the heme binding and heme regulation of HRI.

EXPERIMENTAL PROCEDURES
Preparation of the Constructs-The Wt HRI was previously cloned in the TMV-SP64 plasmid in which a unique NcoI site at the first ATG of HRI cDNA was engineered (13). Similarly, a NcoI site was engineered in Met2 and Met3 HRI constructs at the initiation codon. The introduction of the NcoI site resulted in the change of the second amino acid from Arg to Gly for Met3 construct, whereas no change was made in Met2 construct. The Met2 and Met3 HRI were prepared by site-directed mutagenesis and recombinant polymerase chain reaction (PCR) technique (36). PCR products were digested with NcoI and KpnI and subcloned to TMV-HRI plasmid.
Expression and Purification of Recombinant Wild Type and N-Terminally Truncated Mutant HRI-The plasmid containing the Met2 or Met3 HRI was digested with BglII and EcoRI and ligated into the polylinker region of the pV1392 baculovirus recombination vector. The recombinant HRI baculovirus were generated as described previously (34,35). The postribosomal supernatants was subjected to 50% ammonium sulfate precipitation. The pellet was dissolved in a low salt buffer containing 50 mM KCl, 20 mM Tris-HCl, pH 7.8, 0.1 mM EDTA, and 10% glycerol and was dialyzed against 2 liters of this buffer overnight (13). Wt and mutant HRI proteins were purified by immunoaffinity column chromatography, which was prepared by using anti-HRI monoclonal antibody mABF as described previously (35). The antigenic determinants of m ABF is localized in the KI region; 2 therefore, the Wt and the N-terminally truncated mutants would purify similarly.
Protein Kinase Assays and Western Blot Analysis-The autokinase and eIF-2␣ kinase activities of Wt, Met2 and Met3 HRI were determined by protein kinase assays as described previously (37). HRI proteins were incubated with various concentrations of hemin (0 -5 M) at 30°C for 5 min. Protein kinase assays were initiated by the addition of 5 Ci of [␥-32 P]ATP (3000 Ci/mmol), 50 M ATP, and 100 ng of eIF-2, and the reactions were proceeded for 5 min at 30°C. Reactions were terminated by the addition of SDS-sample buffer containing 1.25% ␤-mercaptoethanol. The phosphorylated polypeptides were separated on 10% SDS-polyacrylamide gels, stained with Coomassie Brilliant Blue G-250. The extents of eIF-2␣ phosphorylation were visualized by autoradiography and quantitated by scintillation counting of the gel slices corresponding to eIF-2␣. Protein kinase assays of the S10 crude extract of each HRI protein in the absence of purified eIF-2 was also performed and was used for background subtraction. Western blot analysis of HRI and various domains was performed, using anti-HRI mABF monoclonal antibody (38) or anti-His 6 monoclonal antibody (Amersham Pharmacia Biotech), as described previously (39).
Hemin-dependent Transformation of HRI in Rabbit Reticulocyte Lysates-Wt and Met2 HRI were synthesized in TNT rabbit reticulocyte lysate (RRL) at 30°C for 15 min followed by 4 min pulse of [ 35 S]Met. The extents of transformation of HRI were analyzed by 10% SDS-PAGE of the 35 S-labeled HRI bound to Ni-NTA resins. The transformation of 35 S-labeled HRI at different concentrations of hemin in RRL was carried out for 1 h as described previously (40).
Spectral Analysis-The Wt and mutant HRI were expressed in Sf9 cells supplemented with 5 M hemin for 48 h and were immunoaffinity purified as described above. The UV and visible spectra of heme-enriched Wt and mutant HRI were performed using a Hewlett-Packard spectrometer as described previously (35). The low salt buffer described above was used as the blank. The UV and visible difference spectra of purified domains of HRI were determined similarly using a Shimidzu 1601 BioSpec. However, phosphate-buffered saline (PBS) containing 2 mM DTT and hemin was used as the blank.
Expression and Purification of HRI Domains-The first 154 amino acids (aa) of HRI were subcloned into the modified pET28a vector, which contained a His 6 tag and TEV protease cleavage site, by blunt end ligation into NdeI/EcoRI site. The first kinase lobe (aa 144 -231), the kinase insertion region (aa 219 -420), the mouse kinase insertion region (aa 241-406), the second kinase lobe (aa 421-540), and the C terminus (aa 541-626) were prepared by PCR. For subcloning into the pET28a vector, an NdeI site and an EcoRI site were engineered at the 5Ј and 3Ј ends of each PCR fragment, respectively. Bacterial cells (BL21 strain) containing the first kinase lobe and second kinase lobe plasmids were induced at A 0.6 with 100 M isopropylthiogalactoside at 25°C overnight. For the N terminus, cells were grown and induced at 37°C for 3 h at A 0.6. The rabbit KI (aa 219 -420) and C terminus were induced with isopropylthiogalactoside at A 0.8 and 0.95, respectively, and cells were grown at 37°C for 3 h. The mouse KI (aa 241-406) was induced at A 0.5 with 100 M isopropylthiogalactoside at 25°C for 6 h. Bacterial cells were lysed according to the Qiagen manual in a buffer containing 50 mM NaH 2 PO 4 (pH 8.0), 300 mM NaCl, and 10 mM imidazole and sonicated. Triton X-100 was added to the sonicated cell extracts to a final concentration of 1%. Cell lysates were incubated on ice for 15 min. The 10,000 ϫ g supernatants were loaded on Ni-NTA agarose column. The column was washed with 50 mM NaH 2 PO 4 , 300 mM NaCl, and 20 mM followed by a 60 mM imidazole wash for kinase I and kinase II. For the N terminus, rabbit KI, mouse KI, and C terminus, a 60 mM imidazole wash was employed. Proteins were then eluted in the elution buffer containing 50 mM NaPO 4 , 300 mM NaCl, and 250 mM imidazole. The imidazole was removed by dialysis against 2 liters of PBS with 2 mM DTT overnight. Purified proteins were concentrated by Slide-A-lyzer dialysis and concentration solution (Pierce) and then stored at Ϫ80°C prior to spectral analysis. The concentration of each domain was determined by the absorbance at 280 nm.
Digestion with rTEV Protease-To remove the His tag from each domain expressed as His tag fusion protein, 20 g of each fusion protein were incubated with 10 units of recombinant tobacco etching virus (rTEV) protease in the presence of rTEV buffer (50 mM Tris-HCl, 0.5 mM EDTA) and 1 mM DTT at 30°for up to 3 h. The efficiency of His tag cleavage was analyzed by 15% SDS-PAGE of aliquots of the expressed proteins before and after rTEV protease treatment. The cleaved His tag was removed by dialysis against 2 liters of PBS with 2 mM DTT overnight.

RESULTS
Expression and Characterization of the N-Terminally Truncated HRI-To determine the role of the N terminus in the heme regulation of HRI, two N-terminal deletion mutants were prepared by site directed mutagenesis and recombinant PCR as described under "Experimental Procedures." These two mutants are schematically illustrated in Fig. 1. Met2 HRI began at the second methionine, and the first 103 amino acids were deleted. Met3 HRI began at the third methionine, and the first 130 amino acids were deleted. Wild type and mutant HRI were expressed in Sf9 cells using the baculovirus expression system. The expressions of the Wt, Met2, and Met3 HRI proteins were examined by Western blot analysis using anti-HRI monoclonal antibody ( Fig. 2A). Wt HRI migrated at about 90 kDa in the SDS-polyacrylamide gels ( Fig. 2A, lane 1). As expected, Met2 and Met3 HRI were smaller than the Wt HRI and migrated at about 67 kDa in SDS-polyacrylamide gels ( Fig. 2A, lanes 2 and  3). The levels of expression of Met2 and Met3 HRI were comparable to that of the Wt HRI ( Fig. 2A).
The autokinase and eIF-2␣ kinase activities of these mutants in Sf9 cell extracts and in purified proteins were examined. Both Met2 and Met3 HRI were active eIF-2␣ kinases and autokinases (Fig. 2, B and D). The autophosphorylations of Wt, Met2 and Met3 HRI were more evident when larger amounts of HRI proteins were used (Fig. 2D). To determine the specific activities of these mutants for eIF-2␣, the protein kinase assays were done quantitatively by using limiting amounts of Wt, Met2, and Met3 HRI. Thus, the autophosphorylation was not as obvious in these experiments (Fig. 2B). Under these conditions, the rate of eIF-2␣ phosphorylation was linear for the first 10 min, as shown in Fig. 2C, and is also linear with the amount of HRI used (data not shown). The specific activities of Wt, Met2, and Met3 HRI, were determined from Fig. 2B to be 5.64, 3.49, and 2.72 pmol of eIF-2␣ phosphorylated per ng of HRI per 5 min at 30°C, respectively. We have consistently observed that the specific activities of Met2 and Met3 HRI were about 50% of that of the Wt HRI. These results suggest that the N-terminal 103 amino acids may be important for achieving the high specific eIF-2␣ kinase activity but may not be essential for kinase activity of HRI.
The Heme Responsiveness of Wt, Met2, and Met3 HRI-To determine the role of the N terminus in the heme regulation of HRI, the inhibitions of eIF-2␣ kinase activities of the Wt, Met2, and Met3 HRI by heme were compared. Incubation of HRI proteins in S10 extracts with hemin (0 -5 M) reduced the eIF-2␣ kinase activities of the Wt, Met2, and Met3 HRI. However, the eIF-2␣ kinase activity of Wt HRI was always more sensitive to inhibition by heme than that of Met2 or Met3 HRI (Fig. 3). Furthermore, the extent of the inhibition of Met2 and Met3 HRI by heme was less than that of Wt HRI at all hemin concentrations tested.
Previously, it was shown that baculovirus expressed HRI could act as an inhibitor of protein synthesis and shut off its own synthesis. The addition of hemin to culture medium has been shown to increase the expression of Wt HRI, presumably by inhibiting its activity in insect cells and thus allowing a higher level of HRI protein expression (34). Therefore, the increase in the expression level of Wt or mutant HRI in response to exogenous heme was used as a criterion for the heme responsiveness of HRI in vivo. In the presence of 5 M hemin, the expression level of Wt HRI increased to 3 times the control value. In contrast, no significant increase in the expression level of Met2 or Met3 HRI was observed when heme was added to the culture media (Table I). These results indicated that Met2 and Met3 HRI were less heme-responsive than the Wt HRI in vivo. It is to be noted that the Sf9 cells are very sensitive to hemin and that hemin concentrations higher than 5 M result in cell death (34). Therefore, it is not possible to examine hemin response at concentrations greater than 5 M in Sf9 cells.
Using synchronized pulse-chase translations, we have shown previously that newly synthesized HRI can be transformed into a stable and an active heme-regulated eIF-2␣ kinase by autophosphorylation in RRL (40). The transformed HRI exhibits a slower electrophoretic mobility on SDS-PAGE. This transformation of HRI is inhibited by heme and can be used as another criterion for heme regulation of this protein under erythroid environment. As shown in Fig. 4 Heme Contents in Purified Wt, Met2, and Met3 HRI-Recently, we have reported that HRI purifies as a hemoprotein with a characteristic absorption peak of the Soret band at 424 nm (35). To determine whether Met2 and Met3 HRIs were hemoproteins, Wt and mutant HRI were expressed in the Sf9 cells supplemented with 5 M hemin and were immunoaffinity purified as described under "Experimental Procedures." The UV-visible spectra of equal amounts (80 g/ml) of Wt, Met2, and Met3 HRI and nonhemoprotein controls were obtained by scanning from 250 to 500 nm (Fig. 5). Both Wt and K199R HRI displayed the characteristic spectra of the hemoproteins, with the absorption peak of the Soret band at 424 nm. In contrast, Met2 and Met3 displayed very little absorption at 424 nm. The nonhemoproteins, such as glyceraldehyde dehydrogenase and pyruvate kinase, had no detectable Soret band. These results demonstrated that the N-terminal 103 amino acids of HRI were required for stable heme binding to HRI.
When a higher concentration of Met2 HRI (250 g/ml) was used, a small absorption peak of Soret band at 424 nm was observed (Fig. 6). Scanning of Met3 HRI (80 g/ml) from 300 to 500 nm also indicated the presence of a small peak of Soret band at 424 nm (Fig. 6). However, the absorption peaks of these two mutants were less than 5% of the Wt and K199R HRI absorptions at 424 nm. When exogenous hemin was added to purified Met2 and Met3 HRI (60 g/ml), these mutants, like Wt HRI, could bind heme and showed the characteristic Soret band at 420 nm (Fig. 7). These results suggested that other regions of HRI might also be responsible for heme binding, because Met2 and Met3 HRI were not devoid of heme and could bind heme in vitro. This interpretation is consistent with the results of Figs. 3 and 4 that Met2 and Met3 HRI were still regulated by heme, although less heme-responsive than Wt HRI.
Localization of the Heme-binding Domains of HRI-To localize the heme-binding domains of HRI, HRI was divided into five domains: the N terminus, the first kinase lobe, the kinase insertion, the second kinase lobe, and the C terminus (Fig. 1). All five domains were expressed as His-tagged fusion proteins in E. coli and were purified by Ni 2ϩ column chromatography. The heme binding of each domain was then determined by difference UV-visible spectroscopy as described under "Experimental Procedures." The first domain consisted of the N-terminal 154 amino acids. The UV-visible spectrum of this polypeptide in the presence of exogenous heme showed the   (34). Lysate proteins were prepared, separated by SDS-PAGE, and then reacted with anti-HRI monoclonal antibody as described under "Experimental Procedure." The expression levels of Wt, Met2, and Met3 HRI in the absence and presence of hemin were quantitated by Western blot analysis as described previously (34 characteristic Soret band at 414 nm (Fig. 8), indicating that the N terminus by itself could bind heme. In addition, when this region was expressed in the presence of 5 M hemin, it purified as a hemoprotein with a Soret band at 415 nm and visible pink color (data not shown). We have also expressed the mouse HRI N-terminal domain (aa 1-138), which is highly conserved among HRI from different species, except for the first 5-15 amino acids. Similar to rabbit N-terminal domain, the mouse N-terminal domain was also capable of binding heme and displayed the characteristic Soret band (data not shown).
The second domain consisted of amino acids 141-231 and included mostly the first lobe of kinase domain; it is referred to as the kinase I domain. When hemin was added to this polypeptide, it did not bind heme, and the characteristic Soret band was not observed (Fig. 8). This result indicated that the kinase I domain could not bind heme.
The third domain consisted of aa 219 -420, which contained the entire rabbit KI sequence unique to HRI and included one putative heme regulatory motif (42). Unfortunately, rabbit KI, expressed as a His-tagged fusion protein, was insoluble. Therefore, the UV-visible spectrum of rabbit KI could not be obtained. However, when part of the rabbit KI domain (aa 301-420) was expressed, it was soluble. It bound hemin and displayed the characteristic Soret band of hemoproteins at 414 nm (Fig. 8). To determine the heme binding of the KI domain, the entire KI domain of mouse HRI (aa 241-406) was expressed as a His-tagged fusion protein. This mouse KI was soluble and was purified as described under "Experimental Procedures." When hemin was added to this polypeptide, it could bind heme and displayed the characteristic Soret band at 415 nm (Fig. 8). These results demonstrated that the KI domain could bind heme. In contrast to the N-terminal domain, when mouse KI was expressed in the presence of 5 M hemin, it did not purify as a hemoprotein (data not shown). This observation suggested that KI might be the heme-binding site responsible for the dynamic heme regulation of HRI.
The fourth domain consisted of amino acids 421-540 and included the second kinase lobe, referred to herein as kinase II. The fifth domain was composed of amino acids 541-626, which are unique to HRI and well conserved among HRI from different species. Neither the kinase II domain nor the C terminus could bind heme, because the characteristic Soret band of hemoproteins was not observed in either case (Fig. 8).
As shown in Fig. 8, heme could bind only to certain domains of HRI. The presence of the His tag did not affect the heme binding of these domains. Similar results were obtained when the His tags of domains of HRI were cleaved by TEV protease (data not shown). In addition, when KI was purified from inclusion body under denatured conditions with subsequent attempts of renaturation by the stepwise removal of denaturant, it did not bind heme. Thus, heme did not bind nonspecifically to proteins under our assay conditions. These observation further strengthen the validity of the heme binding assay using difference spectroscopy. The results from the heme binding study of the domains of HRI demonstrated that the N terminus FIG. 5. The absorption spectra of purified heme-enriched Wt and mutant HRI. Wt and mutant HRI were expressed in Sf9 cells supplemented with 5 M hemin and were immunoaffinity purified as described under "Experimental Procedures." The UV-visible spectra of equal amounts of Wt, mutant HRI, and nonhemoproteins (80 g/ml) were scanned from 250 -500 nm with the low salt buffer with 2 mM DTT as the blank. GAD, glyceraldehyde dehydrogenase; PK, pyruvate kinase.
FIG. 6. The visible spectra of purified heme-enriched Wt, Met2, and Met3 HRI. Wt and mutant HRI were expressed and purified as described in the legend of Fig. 5. Immunoaffinity purified Wt (128 g/ml), Met2 (250 g/ml), and Met3 (80 g/ml) HRI were used for scanning the visible spectra from 300 to 500 nm. Scanning from 300 nm avoided the large protein peak at 280 nm and allowed a more sensitive scale for the Soret bands. Low salt buffer containing 2 mM DTT was used as the blank. and KI could bind heme, whereas kinase I, kinase II, and the C terminus could not.

DISCUSSION
In this report, we have investigated the role of individual domains of HRI in its heme binding and heme regulation. We showed that the N-terminal deletion mutants of HRI, Met2 and Met3, were less sensitive to heme regulation than the Wt HRI (Table I and Figs. 3 and 4) and that the N-terminal domain of HRI could bind heme (Fig. 8). In addition, the heme contents of the purified Met2 and Met3 HRI were much less than that of the Wt HRI (Fig. 6). However, both Met2 and Met3 HRI could still bind exogenous hemin (Fig. 7). These results indicated that the N-terminal domain was required but was not the only region involved in the heme binding and heme regulation of HRI. These observations are in agreement with our recent finding that HRI contains two heme-binding sites per monomer (35). The binding of heme to the first site is stable and copurifies with HRI to homogeneity. In contrast, the heme binding to the second site is dynamic and may be involved in rapid downregulation of HRI by heme (35). Because deletion of the Nterminal domain of HRI resulted in nearly complete loss of the heme content of purified Met2 and Met3 HRI (Fig. 6), the N terminus is most likely the first and the stable heme-binding site.
Recently, we have found that there is a significant structural similarity between N-terminal amino acids 11-118 of HRI and amino acids 16 -120 of mammalian ␣-globin with His-83 in the HRI N terminus as the predicted proximal heme ligand. Furthermore, secondary structure prediction of HRI indicates that the N terminus of HRI has mainly a helical structure and can be modeled to fit the structure of ␣-globin (41). In hemoglobin, heme is held by hydrophobic interactions of amino acid residues brought together by the tertiary structure of the protein.
The stable heme binding to the N terminus of HRI, like that of hemoglobin, may also be facilitated by the helical structure and hydrophobic interactions.
The importance of heme binding in the folding of ␣-globin has been demonstrated by Komar et al. (43). The binding of heme to the N-terminal region of HRI may be required for proper folding and stability of full-length protein in reticulocytes. Furthermore, the N terminus may facilitate the heme regulation of the second site because, despite the presence of a second heme-binding site, Met2 and Met3 HRI were less responsive to heme regulation in vivo and in vitro ( Fig. 3 and Table I).
Rabbit KI domain (aa 219 -420) expressed as His-tagged fusion protein was not soluble, and therefore, the heme binding of this domain could not be determined. To overcome this problem, we expressed the KI region of mouse HRI, which has about 60% overall homology to the KI domain of rabbit HRI. The mouse KI was soluble, could bind heme, and displayed the characteristic Soret band of hemoproteins (Fig. 8). In addition, a portion of the KI (aa 301-420) of rabbit HRI is sufficient for heme binding (Fig. 8). Unlike the N terminus, the KI did not purify as a hemoprotein when it was expressed in E. coli in the presence of heme. These results supported our hypothesis that the KI is the second heme-binding site and that the heme binding to KI is dynamic and reversible. We have shown previously that heme inhibits ATP binding to HRI and kinase activities of HRI (44). Binding of hemin to the KI region might block the ATP binding site and thereby inhibit the kinase activities of HRI.
Little is known about the proteins or the motifs that bind heme reversibly, as is the case in the second heme-binding site  Each of the domains of HRI was expressed as a His-tagged fusion protein in E. coli and was purified by Ni ϩ2 column as described under "Experimental Procedures." The difference UV-visible spectra of purified domains of HRI were scanned from 250 -600 nm. The protein concentrations of each domain used were approximately as follows: N terminus, 300 g/ml; kinase I, 300 g/ml; mouse KI, 800 g/ml, rabbit KI, 150 g/ml; kinase II, 350 g/ml, and C terminus, 600 g/ml. PBS containing 2 mM DTT and 5 M hemin was used as the blank. Rabbit KI* represents only part of the KI domain (aa 301-420) of rabbit HRI. of HRI. We found three examples of proteins involved in the reversible heme binding in the literature. The first is a histidine-proline-rich serum glycoprotein, which has been shown to bind heme (45)(46)(47). The exact physiological function of this protein is still unknown. A recent study suggests that the histidine-proline-rich serum glycoprotein functions as a plasma pH sensor that binds to glycosaminoglycans only when the pH decreases or when the local free metal concentration increases (48). The second example is hemopexin a plasma protein that transports heme to liver. Mutations at His-127 or at His-56 and His-127 have been shown to reduce the affinity of this protein for heme, indicating that these His residues were the hemeiron coordinating residues of hemapexin (49). The third example is histidine-rich protein HRP II, III, and IV from the malaria parasite Plasmodium falciparum (50). HRPs are found in the digestive vacuoles of this parasite, where hemoglobin degradation and heme polymerization to crystalline hemozoin occur. HRPs have been shown to bind heme and catalyze its polymerization. Among the 23 His residues in rabbit HRI, 10 are in the KI sequence and are clustered together (aa 226 -243 and 366 -376). All of these His residues except His-243 and His-366 are conserved among human, rat, rabbit, and mouse HRI. The importance of His residues in the reversible heme binding of HRI is currently under investigation.
In addition to hemoglobin, myoglobin, and cytochromes, the structures of several hemoproteins, such as FixL (51,52) and nitric-oxide synthase (53,54), have recently been solved. Of particular interest is the HasA of Serratia marcescens, a pathogenic bacterium. HasA takes up heme from hemoglobin and transfers it to a receptor called HasR, which in turn, releases heme into the bacterium (55). HRI and HasA are similar in that both proteins bind heme reversibly, and heme does not serve as an oxygen carrier in either protein. In HasA, heme is held by hydrophobic and stacking interactions of the amino acid residues present in the two loops at the interface of ␣ and ␤ structures (55). Interestingly, the secondary structure predictions have indicated that the KI has a loop structure and the reversible heme binding in HRI might resemble that of HasA.
HRI is among six hemoproteins that have a putative heme regulatory motif (HRM). The other five proteins that have this core motif are HAP1 (42), erythroid ␦-aminolevulinate synthase precursors (56), heme lyase (57), heme oxygenase-2 (58), and E. coli catalase (29). The HRM contains preferentially a basic amino acid followed by conserved Cys-Pro dipeptide often flanked downstream by a hydrophobic residue (42,56). HRI contains two of these motifs that are not present in the other eIF-2␣ kinases: Cys-406/Pro-407, located in the kinase insertion region, and Cys-548/Pro-549, located in the C terminus (1). The C terminus (amino acids 541-626), which contained the second HRM of HRI, did not bind heme, indicating that this HRM by itself may not be sufficient to bind heme. It is to be noted that HRM is defined as heme-regulatory motif, and it may not be the heme-binding motif. The role of HRMs in the heme regulation of full-length HRI is currently under investigation.
It has been reported that the PfPK4 eIF-2␣ kinase from human malarial parasite is regulated by heme. PfPK4 contains a single CP motif (Cys-983/Pro-984) that may correspond to the HRM (8). Comparison of the amino acid sequence of the malaria parasite eIF-2␣ kinase with that of HRI indicated that there was no obvious homology between the unique regions of HRI and PfPK4. In contrast to the homology observed among mammalian HRI, the overall homology of PfPK4 to HRI is about 46%, which is similar to the homology of HRI to other eIF-2␣ kinases (PKR, GCN2, PERK, and PfPK4). The homology of mammalian HRI is greater than 82%.
The results presented in this study demonstrate that there are two heme-binding domains in HRI. The heme binding to the N-terminal domain is stable and copurifies with the protein. The heme binding to the KI region is reversible and is responsible for rapid regulation of HRI by heme. In addition, it appears that there is a cooperation of two heme-binding sites for the optimal eIF-2␣ kinase activity and heme regulation of HRI.