Regulation of Heme-controlled Eukaryotic Polypeptide Chain Initiation Factor 2 a-Subunit Kinase of Reticulocyte Lysates*

We have obtained highly purified preparations of the heme-controlled eukaryotic initiation factor 2 a-sub-unit (eIF-2a) kinase (HCI) from rabbit reticulocyte lysates containing five different polypeptides. One of these is a 87-kDa (p87) phosphopeptide which appears to show an autokinase activity. The controlled digestion with trypsin of HCI preparations leads to the suggestion that phosphorylation of p87 is not needed for kinase activity and, furthermore, that another 89-kDa polypeptide could be the kinase catalytic subunit. In agreement with this, monoclonal antibodies directed against p87 do not interfere with eIF-2a kinase activity. Moreover, the anti-p87 antibodies and those directed against the mammalian 90-kDa heat shock protein recognize the same p87 polypeptide from rabbit reticulocyte lysates. Upon incubation of the HCI preparation with hemin (6-10 PM), the eIF-2a kinase is converted into an inactive form and appears to become associated with related peptides forming high molecular weight complexes which can be reversibly activated by 2-mercaptoethanol. The maintenance of the integrity of the porphyrin ring is absolutely required for kinase inactivation and although the presence of metal ion is not essential, the iron and cobalt metalloporphyrins

Regulation of Heme-controlled Eukaryotic Polypeptide Chain Initiation Factor 2 a-Subunit Kinase of Reticulocyte Lysates* (Received for publication, May 29,1991) Raul MendezS, Aurora Morenot, and Char de Haroll From the Centro de BwlogM Molecular, Consejo Superior de Investigacwnes Cienttficas, Universidad Automma de Madrid, Canto Blanco, 28049 Madiid, Spain " We have obtained highly purified preparations of the heme-controlled eukaryotic initiation factor 2 a-subunit (eIF-2a) kinase (HCI) from rabbit reticulocyte lysates containing five different polypeptides. One of these is a 87-kDa (p87) phosphopeptide which appears to show an autokinase activity. The controlled digestion with trypsin of HCI preparations leads to the suggestion that phosphorylation of p87 is not needed for kinase activity and, furthermore, that another 89-kDa polypeptide could be the kinase catalytic subunit. In agreement with this, monoclonal antibodies directed against p87 do not interfere with eIF-2a kinase activity. Moreover, the anti-p87 antibodies and those directed against the mammalian 90-kDa heat shock protein recognize the same p87 polypeptide from rabbit reticulocyte lysates. Upon incubation of the HCI preparation with hemin (6-10 PM), the eIF-2a kinase is converted into an inactive form and appears to become associated with related peptides forming high molecular weight complexes which can be reversibly activated by 2-mercaptoethanol. The maintenance of the integrity of the porphyrin ring is absolutely required for kinase inactivation and although the presence of metal ion is not essential, the iron and cobalt metalloporphyrins are more effective than protoporphyrin IX. The formation of the inactive form of HCI by hemin is prevented by either N-ethylmaleimide, monoclonal antibodies directed against p87, or phosphorylation of p87. The data strongly suggest that hemin regulates eIF-2a kinase activity by promoting formation of the inactive dimer HCI*p87 via disulfide bonds and direct binding of hemin. A model of HCI regulation is discussed.
Protein synthesis in reticulocyte lysates starts at a high rate but declines sharply within a few minutes unless the system is supplemented with hemin. Heme deficiency activates an inhibitor of protein synthesis initiation designated * This work was supported by Direcci6n General de Investigaci6n Cientifica y Tbnica Grant PB-0366 and by an institutional grant from the Fundaci6n Ram6n Areces to the Centro de Biologii Molecular. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in  as heme-controlled inhibitor (HCI).' HCI is a cyclic nucleotide-independent protein kinase which specifically phosphorylates the small, or a, subunit of the initiation factor eIF-2 (reviewed in Refs. 1 and 2). The phosphorylated eIF-2 is defective neither in binding eukaryotic initiator methionyl tRNA and GTP to form a ternary complex, nor in catalyzing the binding of this complex to 40 S ribosomal subunits. However, phosphorylation of the eIF-2 a-subunit results in the binding and sequestration of the guanine nucleotide exchange factor GEF, also designated as eIF-2B, in a GEF . eIF-2 (aP) complex; the unavailability of GEF results in the cessation of the initiation of protein synthesis (2).
HCI is present in heme-supplemented reticulocyte lysates in an inactive form (proinhibitor) (3). The activation of HCI by the lack of heme in reticulocytes seems to be a relevant physiological mechanism for translational control in those cells. There are two other kinds of HCI activation which might have physiological significance in reticulocytes and other cells; (i) by oxidized glutathione (GSSG) (4) and (ii) by calcium ions and phospholipid (Ca'+/phospholipid) (5).
Although the mechanisms of the regulation of protein synthesis by HCI have been extensively studied, little is known about the structure and the regulation of HCI itself. Hunt (6) assigns a molecular mass of about 170 kDa to the native kinase and suggest that HCI is a homodimer of the 85-kDa polypeptide which is phosphorylated in a process which appears to be an autokinase reaction. Recently, Chen et al. (7) have reported that HCI is a single 92-kDa polypeptide; its functions of binding ATP and of autophosphorylation and eIF-Sa phosphorylation are inhibited by hemin. Other studies (8) indicate, on the other hand, that the eIF-2a kinase activity copurifies with a 95-kDa polypeptide which appears to be the catalytic subunit of HCI that binds ATP and has eIF-2a kinase activity but is not phosphorylated. Finally, Matts and Hurst (9) have suggested that HCI is associated with hsp90 in an inactive form in hemin-supplemented lysates and dissociated from hsp90 upon activation. In summary, there are obvious discrepancies with respect to the isolation, the physical characteristics, and the mechanism by which the CAMPindependent protein kinase of the HCI system is activated. Moreover, autophosphorylation of HCI has been suggested (1, 2, 6, 7), but the role of phosphorylation, if any, in the activation process and the mechanism by which heme suppresses kinase activity remain unclear.
I n an attempt to elucidate some of the questions on the The abbreviations used are: HCI, heme-controlled translational inhibitor (an eIF-Pa kinase); eIF-2, eukaryotic polypeptide chaininitiation factor 2; eIF-2a, the a subunit (38 kDa) of eIF-2; hsp, heat shock protein; NEM, N-ethylmaleimide; GSSG, oxidized glutathione; PBS, phosphate-bufferedsaline; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis; TEMED, N,N,N',N'-tetramethylethylenediamine. structure and the regulation of HCI itself, we have further purified pre-activated HCI from reticulocyte lysates. In line with a previous report (8), our results suggest that the catalytic subunit of HCI is not a phosphopeptide and that the 87-kDa peptide phosphorylation is not related to eIF-Sa kinase activity. In this paper we present evidence that the 87-kDa peptide associated with HCI is related to the 83-90-kDa eukaryotic heat shock protein family. In addition, to examine the role of hemin in the regulation of HCI, we have purified hemereversible eIF-2a kinase. The data indicate that hemin promotes formation of an inactive complex between HCI and a peptide of approximately 87 kDa, via disulfide bonds and direct binding of hemin. In the studies described herein, we have utilized other metalloporphyrins, metal-deficient porphyrins, non-ringed structures, and free metal ions to examine further the structural requirements for porphyrins to inhibit the protein kinase activity of HCI. We have shown that the maintenance of the integrity of the porphyrin ring is absolutely required for kinase inactivation. The mechanism of activation of HCI by sulfhydryl reagents such as N-ethylmaleimide (NEM) and the effect of phosphorylation in HCI activation are now better understood.

Preparations
Reticulocyte lysates were prepared from phenylhydrazine-treated rabbits by the method of Hunt et al. (10) with slight modifications (11). To obtain postribosomal supernatants, lysates were centrifuged in a Ti-65 rotor (Beckman) for 120 min at 60,000 rpm (230,000 X g). Highly purified eIF-2 was prepared as described earlier (12). AC88 monoclonal antibody to hsp90 was provided by Dr. David Toft (Mayo Clinic, Rochester, MN). Control and heat shock lysates from Drosophila embryos were the gift of Dr. Zapata of this laboratory. Porphyrins were prepared daily as previously described (13) and maintained in the dark at 4 "C. Concentrations were determined from molar extinction coefficients and absorption maxima (14). Labeled ATP was prepared as described by Schendel and Wells (15). Assays HCI was assayed by its inhibitory effect on translation in hemincontaining reticulocyte lysates and by its ability to phosphorylate eIF-2a. (i) The translation assay was as described (10,11). The standard 30-pl assay system contained 14-16 pl of lysate and ["C] leucine as the labeled amino acid, with or without 20 p~ hemin and HCI as indicated. Incubation was for 60 min at 30 "C. Radioactivity in acid-precipitable protein was determined as described previously (11). One unit of eIF-2a kinase activity was defined as the amount of protein causing 50% inhibition of hemin-dependent translation, and specific activity is expressed as units/mg of protein. (ii) eIF-2a phosphorylation assay was carried out as reported previously (16) with modifications as described. In a total volume of 30 pl, samples containing 25 mM Tris-HC1 (pH 7.6) and 0.5 mM Mg(OAc),, enzyme fractions as specified in the figure legends, and 0.3 pg of purified eIF-2 were incubated with 0.1 mM [T-~'P]ATP (7,000 cpm/pmol) for 10 min at 30 "C. After incubation, the reactions were terminated by the addition of SDS-denaturing buffer. Labeled polypeptides were analyzed by SDS-PAGE (17) as indicated below and autoradiography using Kodak X-Omat S film and Ilford intensifying screen. The areas corresponding to the phosphorylated a-subunit of eIF-2 and 87-kDa polypeptide were scanned at 633 nm in a Computing 300 A (Molecular Dynamics) densitomer. Protein concentrations were determined by the Lowry (18) or the Bradford (19) procedure with bovine serum albumin as the standard.

Purification of HCZ
All operations were performed at 0 to 4 'C. Preactivated eZF-2a Kinase-Preactivated HCI was purified from postribosomal supernatant of rabbit reticulocyte lysates. Postribo-soma1 supernatant was initially processed through the following steps: ammonium sulfate fractionation ( t o 50% saturation), DEAE-cellulose chromatography, CM-Sephadex chromatography and DEAE-Sephadex chromatography. The translation inhibitory activity was retained in both DEAE columns and eluted at 250 mM KC1, and it was not retained by the CM-Sephadex column. The specific activity of the resulting partially purified eIF-2a kinase preparation (translation assay) was 300 units/mg of protein.
Activation of HCZ-This material was diluted with buffer A (20 mM Tris-HC1 (pH 7.6), 1 mM EDTA, 7 mM 2-mercaptoethanol) to give a final KC1 concentration of 80 mM, and was then activated by two consecutive incubations with N-ethylmaleimide and ATP. This strategy for the purification of rabbit reticulocyte HCI has been described (21).
The activated material was made 3 mM in EDTA and subjected to chromatography on DEAE-Sepharose CL-GB as described (21). The main peak of inhibitory activity was eluted at KC1 concentrations of 280 mM, pooled, and loaded onto a polylysine-Sepharose column previously equilibrated with buffer B (20 mM Tris-HC1 (pH 7.6), 0.1 mM EDTA, 1 mM dithiothreitol) containing 280 mM KC1. The column was washed in the same buffer, then modified in the loading solution to contain 0.5 M KCl. Finally, the eIF-2a kinase was eluted with a 20-ml linear gradient of 0.5-1.5 M KC1 in buffer B. During the final steps of isolation, activity was followed by eIF-Pa phosphorylation assay as described under "Assays." The peak of kinase activity was eluted at a KC1 concentration of 0.8 M but was somewhat broad, as previously described (8). The pooled material was dialyzed exhaustively against buffer B containing 90 mM KC1 and applied to a Heparin-Sepharose CL-GB (Pharmacia) column equilibrated with the same buffer. HCI was not retained by the column, but the casein kinases which phosphorylate the @-subunit of eIF-2 were retained and, therefore, largely eliminated. The protein was then dialyzed and concentrated in 15% glycerol in buffer B containing 100 mM KCl. Portions (0.5 ml) of this fraction were layered onto 15-50% glycerol linear density gradients (12 ml) made in the same buffer and centrifuged in an SW 40 rotor (Beckman) for 67 h at 39,000 rpm (191,000 X g). Gradients were monitored at 280 nm in an ISCO fractionator, and 0.3-ml fractions were collected. A gradient containing protein standards bovine serum albumin (67 kDa), aldolase (158 kDa), and catalase (232 kDa) was run in parallel. The main peak of inhibitory activity sedimented in the molecular weight range of 90,000 was pooled and kept frozen at -70 "C until used.
Heme-reversible eZF-2a Kinuse-The procedure used was essentially as described above for the preactivated HCI with the following modifications. (i) Two purification steps (polylysine-Sepharose chromatography and heparin-Sepharose chromatography) were omitted. (ii) Obviously, the NEM treatment and the prephosphorylation reaction were also omitted. (iii) All buffers contained 1 mM dithiothreitol instead of 7 mM 2-mercaptoethanol. (iv) HCI eluted from the DEAE-Sephadex column was diluted, made 3 mM in EDTA as before, and directly loaded onto the same DEAE-Sepharose CL-GB column. Now, the peak of inhibitory activity was eluted at much lower salt concentration (200 mM). This material was further purified by glycerol gradient centrifugation, and fractions with HCI activity sedimenting in the molecular weight range of 90,000 were pooled and kept frozen at -70 "C until used as before.
Mon~clonul Antibodies against 87-kDa Polypeptide-The antigen used for mouse immunization was a preactivated HCI preparation in which two purification steps, polylysine-Sepharose chromatography and heparin-Sepharose chromatography, were omitted. Monoclonal antibodies were obtained as described (22) by hybridation with the mouse myeloma X63/Ag 8.653 (23) and tested by Western blot analysis (24).
Immunoprecipitation-The hybridoma culture supernatant was added to protein A-Sepharose suspended in phosphate-buffered saline (PRS: 10 mM phosphate buffer, p H 7.2, containing 150 mM NaCI). After incubation for 2 h a t 4 "C, the Sepharose was washed three times with PBS. [-y-J2P]ATP-labeled purified HCI was added to the protein A-Sepharose-bound anti-p87 antibodies. After incubation a t room temperature for 2 h (with continuous mixing), the supernatant was removed and the protein A-Sepharose pellet was washed 10 times with PBS containing 1% Nonidet P-40. Washed pellet and the supernatant were treated with Laemmli sample buffer and were analyzed by SDS-PAGE and autoradiography.

RESULTS
Purification and Characterization of HCZ-Several laboratories have reported the purification and partial characterization of heme-controlled eIF-2a kinase from rabbit reticulocytes (7,25). However, in no case has it been possible to isolate its catalytic subunit free of related polypeptides. Activation of the kinase by autophosphorylation has been suggested (6,7,21,25,26) but is not fully accepted (8). In order to clarify this apparent discrepancy and further investigate the real structure of HCI, we have obtained a preparation of the enzyme, from postribosomal supernatant of rabbit reticulocyte lysates, with a higher purity than previously reported (7,8). During glycerol density gradient centrifugation, the bulk of inhibitory activity sedimented in the molecular weight range of 90,000; however, we also noticed the presence of a minor eIF-2a kinase activity in fractions of higher sedimentation coefficients, usually in the range of 180 kDa and occasionally in the range of 360 kDa (data not shown). Previously, a size heterogeneity of the kinase had been suggested (8).
The protein composition of' the preactivated HCI was analyzed by SDS-PAGE in a 5-15% acrylamide linear gradient, in order to effect a better separation and identification of the polypeptides, and visualized by silver staining (Fig. lA, lane 2). Five major bands are seen corresponding to polypeptides with masses of 120, 110, 93, 89, and 87 kDa. In some HCI preparations, however, either the 110-or the 93-kDa polypeptide was absent (data not shown). Although the isolation procedure involves incubation with unlabeled ATP, the labeling of the kinase preparations in a subsequent assay using [y-"'PIATP detect a phosphorylated band corresponding to the p87 phosphopeptide (Fig. lB, lane 2). The preactivated HCI was subjected to controlled digestion with trypsin, and the tryptic digests were analyzed for their ability to phosphorylate e1F-2~1 (Fig. 1). This result demonstrated that eIF-2a kinase activity appears to be independent of p87 phosphorylation (Fig. 123, tracks 6 and 7). Interestingly, there was a good correlation between the rate of disappearance of the p89 polypeptide measured from the silver staining gel and the drop in eIF-2a phosphorylation estimated by quantitating the 38-kDa band density of the autoradiogram, and furthermore, a similar correlation between the rate of disappearance of the p87 polypeptide and the drop in p87 phosphorylation was also observed (Fig. IC). In addition, our purified HCI appears to be free of other protein kinases, particularly the casein kinases which phosphorylate the &subunit of eIF-2. Consistent with this result the p87 phosphorylation shown with preactivated HCI could be due to an autokinase reaction, since the rate is insensitive to dilution (data not shown).
We have obtained monoclonal antibodies directed against p87 that recognized a single polypeptide band of 87 kDa in a Western blot of preactivated HCI (Fig. 2, panel B, lane 1 , and  panel D, lane 3). The same 87-kDa polypeptide was also recognized in reticulocyte lysates as well as in mouse liver lysates (Fig. 223, lanes 2 and 3 ) . Interestingly, both ACRO monoclonal antibody, which reacts with uncomplexed 90-kDa heat shock protein (hsp9O) in cytosols prepared from a variety of mammalian and avian cells (27), and our anti-p87 antibody recognized the same 87-kDa polypeptide in reticulocyte lysates (Fig. 2C). In addition, AC88 antibody recognized an additional band of molecular mass = 93 kDa. Furthermore, our antibodies have also recognized, by Western blot (Fig.  2 0 ) a n d immunoprecipitation (not shown), the 83-kDa heat shock protein (hsp83) from Drosophila embryo lysates. It is known that the amino acid sequence of the mammalian hsp9O (mouse, human, chicken) has a high degree of identity (around 79%) with the known amino acid sequence of the Drosophila hsp83.
After incubation of preactivated HCI with these monoclonal antibodies, the phosphorylation of p87 was completely suppressed whereas the eIF-2n kinase activity was slightly increased (Fig. 2E). As expected, the anti-p87 antibodies immunoprecipitated phosphorylated p87 (Fig. 2F).
In contrast to previous reports (7,21,26), the above results  analyzed for their ability to phosphorylate eIF-2n followed by SDS-PAGE in 10% gels and autoradiogaphy. Tracks 1-3 2,5, and 10 pI of culture medium, respectively; tracks 4-6: 2,5, and 10 pl of hybridoma culture supernatant, respectively. F, immunoprecipitation of 9'labeled p87 polypeptide from [y-"PIATP-labeled HCI preparation by our anti-p87 antibodies. Track 2, supernatant; track 2, immunoprecipitate. support the notion that the self-phosphorylated protein present in most HCI preparations (p87 in our case) does not have to be part of the catalytic subunit of the enzyme, and, moreover, that the p87 peptide is closely related with the 83-90-kDa eukaryotic heat shock protein family.
Role of Hemin on Regulation of HCZ-Previously, it has been proposed that binding of hemin to HCI promoted covalent homodimer formation of HCI (26), and recently, the same research group have indicated that hemin inhibits both the autokinase and the eIF-2a kinase activities of HCI by preventing the binding of ATP to HCI (7). On the other hand, Matts and Hurst (9) have suggested that HCI is associated with hsp9O in an inactive form of hemin-supplemented Iysates. Therefore, the molecular mechanism by which hemin regulates HCI still remains to be clarified.
In an attempt to investigate the mechanism by which the HCI kinase is controlled by hemin, we have purified hemereversible HCI as described under "Experimental Procedures." The main differences with respect to the preactivated HCI were the following. (i) Although most eIF-Pa kinase activity remained in the slowly sedimenting fractions, in the molecular weight range of 90,000 (Fig. 3A), the proportion of HCI activity present in the fast-sedimenting fraction was higher. (ii) Although our heme-reversible HCI is a preparation of fairly high purity, it showed more contaminating polypeptides. The heme-reversible eIF-Pa kinase preparation contains a low amount of casein kinase I1 which phosphorylates eIF-28 (Fig. 3B), only detectable in long time exposures of the gels. (iii) Since our isolation procedure omits the incubation with unlabeled ATP, labeling of the p87 polypeptide in the presence of [y:'2P]ATP was significantly higher.
With our heme-reversible HCI, we have studied the effect of hemin on the eIF-2a kinase activity. To determine any change in mobility of HCI by this treatment, both the untreated HCI and the HCI pretreated with hemin were fractionated again by glycerol density gradient centrifugation and the eIF-20 kinase activity of each fraction was tested with or without a previous incubation with 2-mercaptoethanol. The results are shown in Fig. 3. The untreated HCI migrates in a molecular weight range of 90,000 (Fig. 3A 1, as expected, and the presence of 2-mercaptoethanol appears to have no significant effect on the kinase activity (Fig. E ) . However, after hemin treatment the HCI was converted to an inactive form (Fig. 3 B ) that can be reactivated in the presence of 2-mercaptoethanol (Fig. 3 0 ) . Now the eIF-2cr kinase activity was only present in the faster sedimenting fractions. Thus, there was a significant shift in mobility of HCI from a range of 90 kDa to a region of 180 kDa. Note that the addition of hemin also inhibited phosphorylation of the pR7 polypeptide. Furthermore, the presence of 2-mercaptoethanol restored the ability of p87 to be phosphorylated and, interestingly, showed that p87 had comigrated with the eIF-2n kinase to the fast-sedimenting fractions. This result indicates that hemin promotes the formation of an inactive high molecular weight complex(es) between HCI and related pol-ypeptides, probably p87, via disulfide bonds.
The unexpected hemin induction of an eIF-2P kinase observed in this particular experiment (in Fig. 3 Samples (30 p l ) containing 25 mM Tris-HCI (pH 7.6), 20 mM KCI, with purified hemereversible HCI (2.2 pg) were preincubated either with 0.1 mM (y"'P] ATP (7,000 cpm/pmol) and 0.5 mM Mg(OAc)? for 10 min at 30 "C in B, or with 2 mM NEM as described under "Experimental Procedures" in C or with monoclonal antibodies directed against p87 for 5 min at 30 "C in D. After incubation, these samples and the corresponding untreated samples in A were further incubated for 5 min a t 30 "C with increasing concentrations of hemin specified below. Finally, all samples were analyzed for their ability to phosphorylate eIF-Pa followed by SDS-PAGE in 10% gels and autoradiography. Panels A and C: track 1, without hemin; tracks 2-7, with 6, 12, 18, 24, 30, and 36 p~ hemin, respectively. Panel R: tract I , without hemin; tracks 2-6, with 4,6, 10, 14, and 20 p~ hemin, respectively. Panel D track 1, without hemin, tracks 2-6, with 2, 4, 6, 10, and 16 p~ hemin, respectively. Arrows show the positions of the 87-kDa polypeptide and of the a-and &subunits of eIF-2.
tivity occurred a t hemin concentrations of 6-10 pM (Fig. 4A), lower than the optimal concentration needed to maintain protein synthesis in lysates. This is in agreement with the results obtained by Hronis and Traugh (13). Note that phosphorylation of p87 was also inhibited by hemin, whereas phosphorylation of eIF-28 by casein kinase was not affected. When the heme-reversible HCI was briefly incubated with [y-:''PP]ATP (Fig. 4B) or pretreated with NEM (Fig. 4C), the subsequent addition of hemin cannot prevent the phosphorylation of eIF-2n. Hemin inhibits the eIF-20 kinase activity of NEM-treated HCI only at higher concentrations (Fig. 4C). Of greater interest was the finding that similar results were obtained with heme-reversible HCI preincubated with monoclonal antibodies directed against p87 (Fig. 4 0 ) . Under these conditions, HCI was also converted to the irreversible active state and the addition of hemin could no longer inhibit eIF-2a kinase activity. In this case, in contrast to the results obtained with preactivated HCI (Fig. 2E), the specific binding of monoclonal antibodies only caused a slight reduction of p87 phosphorylation. This difference may be explained by the existence of multiple phosphorylation sites in the p87 polypeptide.
The above results support the notion that hemin promotes formation of an inactive dimer (HCI .p87) which is prevented by treatment with sulfhydryl reagents such as NEM or when p87 polypeptide is previously phosphorylated. In both cases, a conformational change of p87 could explain the irreversible activation of HCI through blocking dimer formation.
Effect of Other Porphyrias and Some Metal Ions on Regulation of HCZ-To provide a better understanding of the molecular mechanism by which hemin promotes formation of the inactive HCI .p87 dimer, we investigated the structural requirements of the porphyrins for inhibition of eIF-2n kinase activity and of p87 phosphorylation. The two metalloporphyrins tested as well as a metal-deficient porphyrin were capable of inhibiting eIF-2a kinase activity, whereas a nonringed porphyrin structure such as bilirubin had no significant action (Fig. 5 ) . Furthermore, the presence of a metal atom in the porphyrin ring was essential for inhibition of p87 phosphorylation. Thus, addition of high concentrations of either protoporphyrin XI or bilirubin produced only a slight decrease of p87 phosphorylation (Fig. 5 E ) .
Both metalloporphyrins inhibited eIF-2n and p87 phosphorylation to the same extent (Fig. 5, D and E ) . The concentration of cobalt protoporphyrin IX required to achieve 50% inhibition of phosphorylation of eIF-2n by HCI was lower, e.g. 0.7 pM for cobalt protoporphyrin Ix and 1.3 p M for hemin (Fig. 5 0 ) . In early studies, Hronis and Traugh (28) obtained slight different values in the concentration required for half-maximal inhibition of HCI by the same metalloporphyrins. The effectiveness of protoporphyrin IX in inhibiting eIF-2n kinase activity was significantly lower (2.6 p M for 50'6 inhibition), but, interestingly, this effect was not accompanied by a parallel inhibition of p87 phosphorylation. These results may indicate that the presence of a metal atom in the porphyrin ring could play an important role in the control mechanism of HCI. Furthermore, the observed differences in the behavior of specific porphyrins may mean that a particular porphyrin may interact directly with HCI itself, or with both HCI and p87 polypeptide with distinct affinities. T o further investigate the possible porphyrin-eIF-2n kinase interaction and the role of the metal atom present in the metalloporphyrins, a number of metal ions were analyzed for their effectiveness in preventing the hemin-promoted HCI inactivation. The data presented in Fig. 6A show a interesting specificity among the free metal ions tested. The presence of either iron or cobalt ions brought about a significant suppression of the hemin effect on heme-reversible eIF-2n kinase (Fig. 6A, tracks 4 and 8 ) , probably as a result of competition between the free metal ion and the metal-containing porphyrin for a binding site on HCI itself or the p87 polypeptide. Manganese ion was a worse competitor than free iron (Fig.   6A, track IO), and the free copper ion was completely ineffective in preventing the hemin effect (Fig. 6A, track 6). As shown in Fig. 5A, protoporphyrin IX, a metal-deficient porphyrin, also promotes eIF-2n kinase inactivation but does not significantly interfere with the phosphorylation of the p87 polypeptide (Fig. 6R, track 1 ). Consistent with our hypothesis that the free metal ions must only compete with the metalloporphyrins, we found that none of these free metal ions can modify the protoporphyrin IX effect on eIF-2n kinase activity but all of them inhibited phosphorylation of p87 pol-ypeptide at the level seen for metalloporphyrins (Fig. 6R, tracks 5-7).
As expected, the addition of each free metal ion solely to heme-reversible HCI had no effect on eIF-2n kinase activity and p87 phosphorylation (Fig. 6R, tracks 2-4). The results obtained in these studies support that metalloporphyrins and metal-deficient porphyrins are capable of inhibiting phosphorylation of e1F-2~1 by HCI, whereas only metalloporhyrins have the capacity of preventing p8'i phosphorylation. On the basis of our results it seems likely that eIF-2n phosphorylation by HCI does not correlate with p87 phosphorylation, and even more, it is possible to inactivate HCI with protoporphyrin IX but not with metalloporphyrins, by promoting dimer formation irrespective of the phosphorylation state of p87 polypeptide.

DISCUSSION
One of the most striking features of the heme-regulated eIF-2a kinase is the apparent high difficulty to obtain preparations that contained a single polypeptide with HCI activity. In this respect, Chen et al. (7) have recently shown a purified HCI preparation containing several polypeptides in a wide range of molecular masses.
In the present report, we describe the purification and further characterization of the preactivated eIF-2a kinase. Our purified HCI preparation contains mainly five polypeptides (in a 5-15% SDS-PAGE gel). Three of them (p120, p110, p93) are probably related to the @-subunit of spectrin (29); the other two could play an important role in the eIF-2cr kinase activity: a 89-kDa species that binds ATP' but is not phosphorylated and that, on the basis of its proteolytic hehavior (Fig. l), appears to be the catalytic subunit of the enzyme; and a 87-kDa species which also binds ATP (not shown) and becomes phosphorylated during activation of the kinase, probably has autokinase activity, and on the basis of our results must modulate the activity of the enzyme. Therefore, the old idea that activation of the eIF-2a kinase is closely coupled with the self-phosphorylation reacticn appears to he incorrect.
In addition, several observations strongly suggest that the 87-kDa eIF-2a kinase associated peptide and the 83-90-kDa eukaryotic heat shock protein (hsp90) are closely related. ( a ) Both AC88 and our anti-p87 monoclonal antibodies recognized a polypeptide band of 87-kDa in a Western blot of rabbit reticulocyte lysates. ( b ) Our monoclonal antibody also recognized the Drosophila 83-kDa heat shock protein by Western blot and immunoprecipitation. (c) The hsp90 proteins are phosphorylated by casein kinase I1 (30) as well as our p87 polypeptide (data not shown). Furthermore, a recent report indicates that hsp9O also possesses an ATP binding site and autophosphorylating activity (31). ( d ) The association of the heme-regulated eIF-2a kinase with hsp90 in hemin-supplemented reticulocyte lysates has been suggested (9).
The molecular mechanism(s) by which hemin regulates HCI is also not understood. Some reports have suggested that R. MBndez and C. de Hero, unpublished observations. the binding of hemin promotes a homodimer HCI. HCI by disulfide bond formation and thus inhibits autokinase and eIF-Sa kinase activity of HCI (7,26). However, Matts and Hurst (9) have presented data supporting the view that HCI exists as an inactive complex with hsp90 in hemin-supplemented lysates and dissociates from hsp90 upon activation. Accordingly, we undertook the purification of heme-reversible HCI and with this purified HCI preparation we were able to demonstrate that hemin promotes, indeed, the formation of an inactive dimer via disulfide bonds, but this is an heterodimer between HCI itself and a 87-kDa polypeptide which appears to be the known hsp90, as mentioned above. Our results may be summarized in the scheme shown in Fig. 7. Many data previously reported from several laboratories may now be understood. ( a ) HCI has been shown to exist in various stages of activation (32). ( b ) Alkylation of partially purified HCI with the sulfhydryl agent NEM followed by treatment with dithiotreitol leads to the irreversible activation of HCI (32). ( c ) Phosphorylation of partially purified HCI has also been correlated with HCI activation (7). ( d ) The phosphorylation state of the reticulocyte hsp90 affects its ability to increase phosphorylation of eIF-2a by HCI (33).
It seems likely that under physiological conditions HCI is in an equilibrium between the active free sulfhydryl state and the inactive disulfide state. Hemin and, to an even greater extent, cobalt protoporphyrin IX promote disulfide formation in an HCIep87 complex and inactivation of the enzyme. A deficiency of heme shifts the equilibrium to the free sulfhydryl form of the enzyme that is active. HCI is active when sulfhydryl groups are maintained in a reduced state or when disulfide formation is prevented by alkylation of the sulfhydryl groups with NEM. Nevertheless, at very high concentrations, hemin can promote the inactivation of NEM-treated HCI. However, in the presence of low concentrations of hemin, the activation of HCI and its probable dissociation from p87 could only be mediated through 8-mercaptoethanol and not through NEM (data not shown). These results strongly suggest the existence of disulfide bonds in the HCIep87 complex promoted by hemin, but, at least in vitro, an excess of hemin could promote formation of HCJ.p87 complex even when sulfhydryl groups are blocked.
In addition, conditions affecting the phosphorylation state of p87, also lead to the irreversible activation of HCI. Although the phosphorylation site involved in the mechanism of HCI activation remains to be determined, based on preliminary data, the casein kinase I1 appears to be the enzyme implicated in the regulation of HCI. 2 On the basis of several results presented here, it seems likely that the maintenance of the integrity of the porphyrin ring is essential for promoting the inactive complex formation and that the presence of a metal atom in the porphyrin ring is essential for preventing phosphorylation of p87. Furthermore, the presence of free metal ions significantly interferes with the action of the metalloporphyrins, but not with that of protoporphyrin I X moreover, only when both free metal ion and protoporphyrin IX are present, the phosphorylation of p87 is clearly diminished. These results support the notion that free and porphyrin-bound metal ions compete for a metal binding site present in the 87-kDa polypeptide and the porphyrin-bound metal ion interaction with p87 severely affects its phosphorylation state.
The presence of a metal atom in the porphyrin ring may stabilize, even more, the HCI.p87 complex. However the predictable binding site for the porphyrin ring in HCI and/or p87 remains to be clarified. As previously suggested (9). there are striking similarities between steroid hormone receptor transformation and HCI activation.
Finally, we have noticed that a low proportion of eIF-2a kinase activity appears to be associated with much higher molecular weight complexes (data not shown). That such a large cytosolic HCI complex plays an important role in the regulation of eIF-Sa kinase is a matter of speculation. In this respect, the applicability of our current view to the regulation of HCI by heme under physiologic conditions in uiuo remains to be determined.