The Heme-regulated Eukaryotic Initiation Factor 2 a Kinase A POTENTIAL REGULATORY TARGET FOR CONTROL OF PROTEIN SYNTHESIS BY DIFFUSIBLE GASES*

Nitric oxide (NO) has been reported to inhibit protein synthesis in eukaryotic cells by increasing the phosphorylation of the a -subunit of eukaryotic initiation factor (eIF) 2. However, the mechanism through which this increase occurs has not been characterized. In this re-port, we examined the effect of the diffusible gases nitric oxide (NO) and carbon monoxide (CO) on the activation of the heme-regulated eIF2 a kinase (HRI) in rabbit reticulocyte lysate. Spectral analysis indicated that both NO and CO bind to the N-terminal heme-binding domain of HRI. Although NO was a very potent activator of HRI, CO markedly suppressed NO-induced HRI activation. The NO-induced activation of HRI was transduced through the interaction of NO with the N-terminal heme-binding domain of HRI and not through S -nitrosy-lation of HRI. We postulate that the regulation of HRI activity by diffusible gases may be of wider physiological significance, as we further demonstrate that NO gen-erators increase eIF2 a phosphorylation levels in NT2 neuroepithelial and C2C12 myoblast cells and activate HRI immunoadsorbed from extracts of these non-eryth-roid cell lines. of protein chain reticulocyte lysate (RRL) identified heme-regulated protein kinase that phosphorylated the a -subunit of immu- as Samples analyzed by 10% SDS-polyacrylamide gel electrophoresis, followed by transfer to polyvinylidene difluoride membrane and autoradiography as described 22) . Autophosphorylation of HRI was assayed by the incorporation of [ 32 P]P i into HRI during eIF2 a kinase assays incubated with [ g - 32 P]ATP. 32 P-Labeled HRI and eIF2 a were detected by quantitatively quenching 35 S emissions with three intervening layers of previously developed x-ray film (21, 22). To study the effects of NO or CO on HRI activation in vitro , HRI was synthesized de novo and matured for 60 min in hemin-supplemented or heme-deficient RRL, or in heme-deficient RRL (50 min) followed by the addition of 10 m M hemin (for 10 min) to yield mature-competent, transformed HRI and repressed HRI, respectively. HRI was then affin-ity-purified, and autophosphorylation of HRI was assayed by the incor- poration of [ 32 P]P i into HRI incubated with [ g - 32 P]ATP and treated or untreated with NO or CO as described above. A non-His-tagged [ 35 S]HRI was similarly studied for nonspecific interactions to the resin.

Over 2 decades ago, the heme-regulated inhibitor (HRI) 1 of protein chain initiation in rabbit reticulocyte lysate (RRL) was identified as a heme-regulated protein kinase that phosphorylated the ␣-subunit of eukaryotic initiation factor eIF2 (reviewed in Refs. [1][2][3]. eIF2 delivers Met-tRNA i in a complex with GTP to 43 S ribosomal initiation complexes and is released as a complex with GDP at the completion of the initiation cycle. Recycling of eIF2⅐GDP and the formation of eIF2⅐GTP⅐Met-tRNA i complexes requires the action of the guanine nucleotide exchange factor, eIF2B. Under heme-deficient conditions, HRI is activated and phosphorylates eIF2. Phosphorylated eIF2 avidly binds eIF2B, sequestering eIF2B in a poorly dissociable complex, which subsequently leads to the inhibition of the initiation of translation, as eIF2⅐GDP complexes that are present in excess of eIF2B fail to recycle. Thus, HRI functions to coordinate globin synthesis with heme availability in RRL. The amino acid sequence deduced from HRI cDNA indicates that it is composed of at least five distinct domains (4). The unique N-terminal domain of HRI contains ϳ165 amino acids. The second and fourth domains contain conserved sequence motifs I-V and motifs VI-XI, which comprise the N-terminal and C-terminal catalytic lobes of protein kinases, respectively. The third and fifth domains are also unique, consisting of ϳ140 amino acids that are inserted between the two conserved kinase lobes and ϳ50 amino acids at the C terminus, respectively.
HRI is a hemoprotein that contains two distinct of hemebinding sites (5)(6)(7). Heme binding to the first site is stable, and remains associated with purified HRI, whereas heme-binding to the second site is reversible and appears to be responsible for the rapid heme-induced down-regulation of HRI activity . We have recently demonstrated that the N-terminal domain of HRI contains the stable heme-binding site, whereas the unique third domain of HRI appears to contain the reversible hemebinding site (6,7), thus raising the question as to the function of the N-terminal heme-binding domain (NT-HBD) in the regulation of HRI activation.
Nitric oxide (NO) is now recognized as a major signaling molecule (8 -10). Like NO, carbon monoxide (CO) has also been identified as an endogenous second messenger in the peripheral and central nervous system, and has been demonstrated to play an important role in hemodynamic regulation (10,11). Both NO and CO have high affinity for both protein-bound and free heme-iron (12)(13)(14). Recently, the cytostatic activity of NO was found to correlate with NO-induced increase in eIF2␣ phosphorylation and inhibition of protein synthesis in a number of cell lines (15). However, the mechanism of this NO-induced increase in eIF2␣ phosphorylation was not characterized. In this report, we examined the possible function of the NT-HBD in the regulation of HRI activation, and present evidence that HRI activity is regulated by the binding of the diffusible gases NO and CO to its NT-HBD.

EXPERIMENTAL PROCEDURES
Protein Synthesis and eIF2␣ Phosphorylation in Reticulocyte Lysates--Protein synthesis was determined by measuring the incorporation of [ 14 C]leucine into the acid-precipitable protein in 5-l aliquots taken from standard RRL mixtures containing 20 M hemin at 30°C as described (16,17). The phosphorylation of eIF2␣ in 2 l of protein synthesis mixes was analyzed as previously described by Western blotting of one-dimensional vertical isoelectric focusing (VIEF) slab gels using 1:1000 dilution of anti-eIF2␣ monoclonal ascites fluid (18 -20).
Treatment of Samples with NO and CO-The NO generator NOC-9 (Calbiochem) was prepared as a 100 mM stock solution in 20 mM Tris-HCl (pH 8.0) and used immediately for experiments in RRL or in vitro. NOC-9 breaks down to form NO with a half-life of 3 min at pH 7.4. For CO treatments, samples (RRL or immunoresins) were gassed in a fume hood with CO (99.0ϩ% CO; Sigma) for 3 min on ice by directing a stream of CO through a 21-gauge needle at the surface of the sample with sufficient velocity to cause continuous mixing of the microcentrifuge tube contents. The tubes were then sealed with parafilm prior to any further incubations.
De Novo Synthesis, Maturation, and NO or CO Treatment of HRI in Situ-Coupled transcription/translation reactions to pulse-label [ 35 S]HRI and His 7 -[ 35 S]HRI in nuclease-treated RRL (TnT RRL, Promega) were carried out as described previously (21,22). After radiolabeling, one volume of TnT protein synthesis mix containing either [ 35 S]HRI and His 7 -[ 35 S]HRI was mixed with seven volumes of normal heme-deficient or hemin-supplemented (10 M hemin) protein synthesis mixes containing non-nuclease-treated RRL and the protein synthesis initiation inhibitor aurintricarboxylic acid (60 M). The samples were then incubated for 50 min at 30°C to yield "mature-competent" HRI (50 min in hemin-supplemented RRL), "transformed" HRI (50 min in hemedeficient RRL), or "repressed" HRI (40 min in heme-deficient RRL, followed by a 10-min incubation in the presence of 10 M hemin) (21,22). The in situ effects of NO, CO, N-ethylmaleimide (NEM), or dithiothreitol (DTT) on HRI were then assayed by treating samples with NOC-9, gassing RRL with CO, or treatment of samples with NEM or DTT followed by an additional period of incubation at 30°C as specified in the figure legends. HRI was then affinity-purified and assayed for kinase activity as described below.
Assay of the Kinase Activity of Affinity-purified His 7 -[ 35 S]HRI-Immunoadsorption of His 7 -[ 35 S]HRI by anti-His 5 antibodies (Qiagen) and non-immune control antibodies were done as previously described (22). Assays for the eIF2␣ kinase activity of His 7 -[ 35 S]HRI bound to immunoresin were performed for 4 min at 30°C as described (21,22). Samples were analyzed by 10% SDS-polyacrylamide gel electrophoresis, followed by transfer to polyvinylidene difluoride membrane and autoradiography as described previously (21,22) . Autophosphorylation of HRI was assayed by the incorporation of [ 32 P]P i into HRI during eIF2␣ kinase assays incubated with [␥-32 P]ATP. 32 P-Labeled HRI and eIF2␣ were detected by quantitatively quenching 35 S emissions with three intervening layers of previously developed x-ray film (21,22).
To study the effects of NO or CO on HRI activation in vitro, HRI was synthesized de novo and matured for 60 min in hemin-supplemented or heme-deficient RRL, or in heme-deficient RRL (50 min) followed by the addition of 10 M hemin (for 10 min) to yield mature-competent, transformed HRI and repressed HRI, respectively. HRI was then affinity-purified, and autophosphorylation of HRI was assayed by the incorporation of [ 32 P]P i into HRI incubated with [␥-32 P]ATP and treated or untreated with NO or CO as described above. A non-His-tagged [ 35 S]HRI was similarly studied for nonspecific interactions to the resin.
Spectral Analysis of NT-HBD of HRI-Recombinant NT-HBD of HRI was purified as previously described (7). Spectral analysis was done using a Shimdazu UV-160 spectrophotometer scanning the NT-HBD dissolved in 20 mM Tris-HCl (pH 7.4) containing 150 mM NaCl from 200 to 650 nm. The NT-HBD was then reduced by the addition of several grains of dithionite to the cuvette. Immediately after the spectrum of reduced NT-HBD was obtained, NOC-9 was added to a concentration of ϳ1 mM, or the cuvette was gassed with 99.0ϩ% CO for 3 min to obtain the spectra of NO-bound and CO-bound NT-HBD, respectively.
Immunoadsorption of HRI from Cultured Cell Extracts-Ntera-2/c1.D1 neuroepithelial (NT-2; Ref. 23) and C2C12 myoblast cells were cultured in DMEM in the presence of 10% fetal calf serum, penicillin/ streptomycin and 5% CO 2 to ϳ70% confluence. The cells were then grown overnight (16 h) in DMEM supplemented with 10 M hemin-HCl. Cells were washed two times with Hank's buffered saline, lysed directly by scraping cells from the culture flasks with lysis buffer (ϳ1 ml/2.5 ϫ 10 7 cells) containing 30 mM HEPES-KOH (pH 7.4), 150 mM NaCl, 1 mM EDTA, 2 mM EGTA, 50 mM NaF, 1 mM DTT, 1% Triton X-100 (v/v), and protease-inhibitor mixture (Sigma; 100 l/10 7 cells solubilized). The cell lysates were clarified by centrifugation at 15,000 ϫ g for 10 min, and HRI was then immunoadsorbed with anti-HRI NT-HBD antibody from 10 l of mouse polyclonal ascites fluid or a control non-immune antibody that had been previously adsorbed to mouse anti-IgG cross-linked to agarose (24). Mouse polyclonal anti-HRI NT-HBD antibody was raised against affinity-purified recombinant NT-HBD by the HYCABS core facility (Oklahoma State University, Stillwater, OK). The immune pellets were washed one time with 10 mM Tris-HCl (pH 7.4) containing 150 mM NaCl and 1% (v/v) Tween 20, and two times with 10 mM Tris-HCl (pH 7.4) containing 150 mM NaCl, and assayed for associated kinase activity as described above.
Levels of eIF2␣ phosphorylation present in C2C12 cells were determined by lysis of cells directly in VIEF sample buffer and Western blotting of samples separated on VIEF slab gels as described previously (25). Levels of eIF2␣ phosphorylation present in NT-2 cells were deter-

Effect of NO and CO on Protein Synthesis and eIF2␣
Phosphorylation in RRL-To determine the effects of NO on protein synthesis varying concentrations of the NO generator NOC-9 was added to hemin-supplemented RRL (Fig. 1a). Low concentrations of NOC-9 stimulated protein synthesis slightly, while high concentrations of NOC-9 caused a rapid and complete shut-off of protein synthesis. To examine the mechanism by which NO inhibits translation, the phosphorylation status of eIF2␣ in NOC-9-treated RRL was analyzed (Fig. 1b). NO generation stimulated eIF2␣ phosphorylation in a concentration-dependent manner with 1 mM NOC-9 treatment of hemin-supplemented RRL causing a near quantitative phosphorylation of eIF2␣.
To determine the effect of CO on protein synthesis, RRL was gassed with 100% CO for 3 min on ice. CO stimulated the rate of protein synthesis in hemin-supplemented RRL by ϳ50% (Fig. 2a). Translation in heme-deficient RRL was stimulated 40% by CO during the first 5 min of incubation, but shut-off of translation was not prevented by CO (data not shown). Similarly, pre-gassing of heme-supplemented RRL with CO stimulated translation by ϳ40% during the first 5 min of incubation in NOC-9-treated RRL, but did not prevent translational shutoff (Fig. 2a). These observations suggested that NO and CO have opposing effects on protein synthesis in RRL.
Effect of NO and CO on HRI Activity-To determine whether NO modulated the activation of HRI, His 7 -tagged [ 35  creasing concentrations of NOC-9, and the His 7 -tagged HRI was affinity-purified and assayed for kinase activity (Fig. 3a). His 7 -tagged [ 35 S]HRI that was matured in hemin-supplemented RRL (ϩ, mature-competent HRI) was an inactive kinase. HRI that was matured in heme-deficient RRL was transformed into an active auto-and eIF2␣ kinase (Ϫ, transformed HRI]. The kinase activity of transformed HRI was inhibited by addition of hemin (Ϫ/ϩ, repressed HRI) to the heme-deficient incubations. Addition of 1 mM NOC-9 caused a marked enhancement of the autokinase and eIF2␣ kinase activity of mature-competent, transformed and repressed HRI (Fig. 3a, lanes 12-14). Although 0.01 and 0.1 mM NOC-9 had little stimulatory effect on the eIF2␣ kinase activity of transformed HRI, these concentrations of NOC-9 increased the activity of repressed HRI to that of transformed HRI in untreated heme-deficient RRL. Although 0.1 mM NOC-9 caused maturecompetent HRI to become as active an eIF2␣ kinase as trans-formed HRI in untreated heme-deficient RRL, it also caused a small decrease in the autokinase activity of transformed and repressed HRI. This decreased autokinase activity may explain the slight enhancement of protein synthesis that was observed in the presence of low NOC-9 concentrations in Fig. 1a.
To establish whether the effect of NOC-9 on the kinase activity of HRI was through a direct interaction of NO with HRI, mature-competent, transformed, and repressed His 7 -[ 35 S]HRI were affinity-purified and assayed for autokinase activity in the presence of varying concentrations of NOC-9 in vitro (Fig. 3b). The autokinase activity of all the three forms of HRI were activated upon treatment with NOC-9 in vitro in a concentration-dependent fashion, with repressed HRI being the most sensitive to NO-induced activation (3-fold in the presence of 0.01 mM NOC-9).
Subsequent studies were carried out with repressed HRI, as stress-induced activation of repressed HRI in hemin-supplemented RRL occurs independent of changes in heme concentration (18,22,24,(27)(28)(29). Thus, repressed HRI is the form of HRI that is likely to be the most physiologically relevant target for regulation by diffusible gases in vivo.
Order-of-addition studies were carried out to determine whether CO and NO have competing effects on HRI activation (Fig. 2b). Gassing repressed His 7 -[ 35 S]HRI with 100% CO suppressed its autokinase activity below that of the ungassed control, and dramatically suppressed the autokinase activity of repressed His 7 -[ 35 S]HRI that had been affinity-purified from NOC-9-treated RRL. Furthermore, in vitro treatment with NOC-9 markedly enhanced the autokinase activity of repressed His 7 -[ 35 S]HRI that had been affinity-purified from CO-gassed RRL. These results suggest that CO and NO compete for a common binding site in HRI and have opposing effects on HRI activation.
Mechanism of Regulation HRI Activity by NO and CO-The physiological effects of NO are mediated through a number of mechanisms, including (i) S-nitrosylation of proteins (30), (ii) the generation of free radicals and oxidative stress (31), or (iii) the coordination of NO by protein-bound heme (12,32). The only known biological reactivity of CO is as a ligand for proteinbound heme, and CO is neither sulfhydryl-nor redox-active. Thus, the observations that CO both blocks and reverses NOinduced activation of HRI strongly suggest that NO induces the activation of HRI through its coordination by heme. However, since sulfhydryl-reactive compounds and oxidative stress are well known to cause the activation of HRI in hemin-supplemented RRL (1-3, 33), we carried out experiments designed to test the first two possible mechanisms further.
To examine whether NO activates HRI by modifying sensitive sulfhydryls of HRI, the effect of NO on the kinase activity of repressed His 7 -[ 35 S]HRI that was affinity-purified from RRL treated (or not) with NEM was examined (Fig. 4a). Sulfhydrylreactive compounds, such as NEM, are thought to activate HRI by covalently modifying some sensitive sulfhydryls of HRI (1-3, 33-35). NEM-treatment caused an ϳ3-fold increase in the autokinase activity of repressed HRI. Treatment of repressed His 7 -[ 35 S]HRI with NO in vitro led to an even more marked increase in the autokinase activity of HRI. The autokinase activity of HRI affinity-purified from NEM-treated RRL was also markedly increased upon treatment with NO, indicating that NO can further activate HRI containing previously modified sulfhydryls. However, the possibility remained that NO was modifying a different site on HRI than that which was reactive with NEM.
To further test the hypothesis that the activation of HRI by NO was not mediated through its effect on sensitive sulfhydryls of HRI or through the generation of a generalized oxidative stress, the effect of the reducing agent DTT on NO-induced activation of repressed HRI was examined (Fig. 4b). DTT reverses the effect of NO on proteins that are mediated through S-nitrosylation (36). In addition, DTT protects HRI from activation induced by sulfhydryl-reactive compounds and reverses the activation of HRI induced by generalized oxidative stress and sulfhydryl-reactive heavy metal ions (3,33,37). Addition of DTT prior to NOC-9 treatment had no effect on NO-induced inhibition of protein synthesis (data not shown) or activation of repressed His 7 -[ 35 S]HRI (Fig. 4b).
NO Modulates the Activation of HRI by Binding to HRI's NT-HBD-To determine whether NO and CO might modulate the activation of HRI through their coordination by heme, a spectral analysis of the recombinant NT-HBD of HRI was carried out. Reduction of HRI's NT-HBD with dithionite caused a shift in the absorption maximum of the Soret band from 414 nm to 428 nm, with the absorption maximum in the ␣/␤ region shifting from 534 nm to distinct ␣and ␤-bands with absorption maximum at 560 and 530 nm, respectively (Fig. 5A). Addition of NO caused the Soret peak to shift immediately to 421 nm and decrease in intensity by ϳ50%, with the ␣and ␤-bands shifting to 572 and 542 nm, respectively (Fig. 5B). After 15 min, the Soret band broadened to give an absorption maximum at 402 nm. In contrast, gassing the reduced NT-HBD with CO (Fig. 5C) caused the Soret band to shift from 428 to 422 nm and increase in intensity by ϳ30%, with the absorption maximum of the ␣and ␤-bands shifting to 568 and 536 nm, respectively. These results indicate that NO and CO become coordinated to the heme moiety of the NT-HBD of HRI.
To address the question of whether the activation of HRI by NO is due to the NO binding to NT-HBD, we studied the effect of NO on the activity of HRI from which the NT-HBD has been deleted (HRI/Met-3; Ref. 6). His 7 -[ 35 S]HRI/Met-3 was synthesized de novo, transformed in heme-deficient RRL, and re-pressed by the addition of hemin. Compared with wild type His 7 -[ 35 S]HRI, the kinase activity of His 7 -[ 35 S]HRI/Met-3 was only marginally activated upon NOC-9 treatment (Fig. 6a). In contrast, treatment with NEM markedly increased the kinase activity of both His 7 -[ 35 S]HRI/Met-3 and wild type His 7 -[ 35 S]HRI (Fig. 6a). These results support the hypothesis that the NT-HBD domain mediates NO-induced activation of HRI. However, we cannot currently rule out the possibility that NO may have additional effects mediated through an interaction with heme bound to the second regulatory heme-binding site in HRI, because NOC-9 treatment reproducibly caused a slight stimulation in kinase activity of HRI/Met-3.
To further confirm the involvement of NT-HBD domain of HRI in NO activation, we studied whether [ 35 (Fig. 6b). Furthermore, NOC-9 markedly enhanced Effect of NO on HRI Activity in Non-erythroid Cells-Possible regulatory roles for HRI in nonerythroid cells have yet to be explored in detail. Although HRI expression was previously believed to be specific to erythroid cells (4,38,39), mRNA encoding HRI has been reported to be expressed in several nonerythroid tissues (40,41). A search of the data base of expressed sequence tags, dbEST (Table I) indicated that HRI mRNA is expressed in a wide variety of nonerythroid cells and tissues, as well as a variety of tumors. A 1862-base pair cDNA (GenBank accession no. AA196697) putatively coding for HRI from human NT-2 neuroepithelial cells was purchased from Genome Systems, Inc., and sequenced. The predicted openreading frame, encoding a protein with 80% sequence similarity to rabbit HRI but lacking a start codon, was subcloned into the pS64T plasmid. The properties of His-tagged NT-2 HRI protein that was expressed in and affinity-purified from TnT RRL were indistinguishable from RRL HRI (data not shown).
Subsequent to this work, the predicted amino acid sequences of four full length human HRI cDNAs isolated from libraries prepared from bone marrow (AF147094), dermal hair papilla (AF255050), brain (AB037790) and dermal microvascular endothelial (AF181071) cells have been entered into GenBank. The entries indicate that the predicted amino acid sequence of human HRI is 80.9%, 81.8%, and 81.6% identical to rat, rabbit, and mouse HRI, respectively, and that the NT2 cDNA lacked the first 31 nucleotides of the open reading frame.
To address the question of whether HRI protein was expressed in nonerythroid cells, anti-HRI/NT-HBD antibodies were used to immunoadsorb protein from extracts prepared from NT-2 and C2C12 cells. Assays of the immunoadsorbed protein indicated that an eIF2␣ kinase activity was specifically adsorbed from the NT-2 and C2C12 cells extracts (Fig. 7A). Treatment of immunoadsorbed protein with NOC-9 markedly stimulated phosphorylation of a protein band, which was specifically immunoadsorbed from the NT-2 and C2C12 extracts by the anti-HRI/NT-HBD antibody (Fig. 7B), and co-migrated with RRL HRI (data not shown). A phosphoprotein that migrated near HRI was immunoadsorbed from C2C12 extracts. However, the amount of HRI immunoadsorbed from the C2C12 extracts was below the limit of detection by Western blotting, so we could not determine whether the protein was HRI by Western blotting. In addition, our anti-HRI/NT-HBD antibody failed to adsorb any eIF2␣ kinase or NO-activable autokinase activity from extracts prepared from the blood of HRI-knock  1). Dithionite was then added to reduce the bound hemin and the sample was rescanned (A, spectrum 2). After the spectral analysis of the dithionite-treated NT-HBD, NO was generated by the addition of 1 mM NOC-9, and the sample was rescanned immediately (B, spectrum 1) and after 15 min (B, spectrum 2), or the dithionite-treated NT-HBD was gassed with 100% CO for 3 min followed by spectral analysis (C). Arrows indicate the Soret band, the insets within the panels show an expansion of the region of the ␣/␤ absorption spectrum, and the asterisk (*) indicates an absorption peak due to the presence of dithionite. out mice, 2 indicating that our antibody was not cross-reacting with other eIF2␣ kinases, particularly PKR, which is known to be present in blood cells (42).
To determine whether NO stimulated eIF2␣ phosphorylation in NT-2 (Fig. 8A) and C2C12 (Fig. 8B) cells, cells were cultured in the presence or absence of the NO donor SNAP. Treatment of both NT-2 (Fig. 8A) and C2C12 (Fig. 8B) cells with SNAP caused an increase in eIF2␣ phosphorylation in both cell lines. These results suggest that NO may be an important physiological regulator of HRI activation in non-erythroid cells.

DISCUSSION
Mechanisms that control initiation of protein synthesis play significant roles in regulating cell growth (reviewed in Ref. 43). eIF2␣ phosphorylation performs a critical role in control of cell proliferation (reviewed in Refs. 44 and 45), with increased eIF2␣ phosphorylation (20,46,47) and down-regulation of protein synthesis correlating with cell growth arrest and entrance into G 0 (48). Cell growth arrest is required for the subsequent terminal differentiation of a number of cell types (reviewed in Ref. 44). Increased eIF2␣ phosphorylation also mediates different forms of stress-related apoptosis (44,45,49,50). Furthermore, suppression of eIF2␣ phosphorylation (44) or expression of a nonphosphorylatable mutant of eIF2␣ causes malignant transformation of cells (51) . These observations suggest that eIF2␣ kinases act as tumor suppressors in the regulation of cell growth (reviewed in Ref. 44). Specifically, these studies have focused on the double-stranded RNA-activated eIF2␣ kinase (PKR), as constitutive or inducible expression of this eIF2␣ kinase occurs in most cell types.
Our results suggest that NO may be a physiological regulator of HRI activation. NO was as potent of an activator of HRI as heme deficiency in RRL. Although the 0.25-0.5 mM concentration of NOC-9 required to inhibit protein synthesis in hemin-supplemented RRL likely falls outside of the physiological range of concentrations to which NO accumulates in vivo, the concentration range is reasonable considering the NO-binding capacity of the estimated millimolar levels of hemoglobin present in RRL. However, under physiological conditions, much lower levels of NO might be capable of activating HRI in reticulocytes, as recent work indicates that the reaction of oxyhemoglobin with NO functions to maintain NO bioactivity (52,53).
A more likely physiological target of NO is HRI present in red cell precursors early during hematopoiesis, when significant levels of hemoglobin have yet to accumulate. NO sup-  presses hemoglobin synthesis in K562 cells (54,55). Although it was speculated that the NO-induced suppression of hemoglobin synthesis was a result of NO-induced inhibition of heme bio-synthesis (54), our current results suggest it could be the result of NO directly activating HRI present in K562 cells. In addition, activation of nitric-oxide synthase has been implicated in tamoxifen-induced apoptosis of K562 cells (56), and NO has been reported to (i) influence the growth and differentiation of normal bone marrow cells (57), (ii) induce apoptosis in bone marrow progenitor cells (58), and (iii) be a mediator of cytokineinduced hematopoietic suppression (59). Our results suggest that these effects of NO could, in part, be mediated via NOinduced activation of HRI, considering the well established role that eIF2␣ phosphorylation plays in suppressing cell growth and inducing apoptotic cell death.
The consequences of NO synthesis on the function and fate of cells varies depending on the rate of NO synthesis, cell type, presence of free radicals, and the anti-oxidant status of the cell (reviewed in Refs. 60 -62). NO modulates signaling molecules in apoptotic pathways, contributing to the physiological balance between pro-apoptotic and anti-apoptotic stimuli (63). NO at high levels generally induces cell death, whereas low doses of NO rescues many cells from apoptotic death. NO is cytotoxic to many cells types (Ref. 64, and references therein), and the cytostatic activity of NO has been found to correlate with NO-induced increase in eIF2␣ phosphorylation in a number of cell lines (15).
Opposing effects of NO on protein synthesis and HRI autokinase activity were also observed in RRL. In situ, low levels of NO were observed to stimulate translation and suppress, to a degree, the autophosphorylation of HRI. These observations may reflect the ability of NO to displace CO from hemoglobin (see below), or a requirement for both HBDs to be liganded to NO within the HRI dimer for the NO-induced allosteric activation of HRI to occur.
A search of dbEST indicates that mRNA encoding HRI is present in many non-erythroid cell types. Our results indicate that the HRI mRNA present in NT-2 and C2C12 cells is translated into protein, albeit at levels much lower that that present in erythroid cells. Thus, NO-induced activation of HRI may contribute to NO-induced inhibition of protein synthesis in non-erythroid cells that express HRI. NO-induced activation of HRI, like the effects of PKR activation (44,45), may also play a role in mediating NO-induced apoptosis in these cells. Furthermore, NO may be a principle activator of HRI in non-erythroid cells, as these cells are not likely to experience large fluctuations in their heme content.
Our data also suggest that CO is an important physiological regulator of HRI activation. CO is the product of the first step in heme catabolism catalyzed by heme oxygenase (HO), and has been proposed to be a physiological regulator akin to NO (10,11,65). HO-1 expression is activated in virtually all cell types, not only by the presence of free heme, but also in response to inflammatory cytokines, hypoxia, and many forms of oxidative stress (65,66). Induction of HO-1 plays an important role in protecting cells from the adverse effects of oxidative stress (11,36,(65)(66)(67)(68). Furthermore, low concentrations of CO prevent tumor necrosis factor-␣-induced apoptosis in L929 fibroblasts (69), suggesting that the anti-apoptotic effect of HO-1 expression may be mediated in part via CO. Here we demonstrate that CO suppresses the activity of HRI and stimulates protein synthesis in RRL, and has the capacity to reverse NO-induced activation of HRI. These results suggest that HRI may be regulated through the competition between NO and CO for a common binding site in the NT-HBD of HRI.
The spectral analyses of the binding of NO and CO indicate the potential mechanism through which NO and CO have opposing effects on HRI activation. Whereas CO binds and forms a 6-coordinate heme complex, the spectral changes induced upon the binding of NO to the NT-HBD are similar to FIG. 7. Effect of NO on HRI activity immunoadsorbed from extracts of from NT-2 and C2C12 cells. A, HRI was immunoadsorbed from RRL, NT-2, or C2C12 cells with anti-HRI/NT-HBD (I) or non-immune control (N) antibody and assayed for kinase activity as described under "Experimental Procedures." Autoradiogram of 32 P-phosphorylated eIF2␣. Note that the migration of HRI is indicated, but the exposure time required to observe phosphorylated eIF2␣ in the C2C12 sample overexposes a 32 P-labeled band that migrates near HRI (indicated by * in B). B, NT-2 and C2C12 cells were grown for 16 h in the presence of DMEM supplemented with 10 M hemin. The cells were lysed, and HRI was immunoprecipitated from cell lysates (NT-2 or C2C12) with anti-HRI/NT-HBD (I) or non-immune control antibodies (N). Samples were then treated with 1 mM NOC-9 (ϩ) or buffer (Ϫ), and autokinase activity was assayed as described under "Experimental Procedures." Figure shows autoradiogram of 32 P-phosphorylated HRI (arrow); *, unidentified phosphoprotein that migrates near HRI. those induced upon the binding of NO to the regulatory hemebinding domain of guanylate cyclase (70). This observation suggests that NO activates HRI through the same mechanism by which its activates guanylate cyclase; cleavage of the ironhistidine bond in the HBD to yielded a 5-coordinate ferrousnitrosyl complex (70).
HO-1 expression appears to play a central role in an important regulatory network between NO and CO (11). The ability of CO to antagonize NO-induced activation of HRI leads us to propose a novel feedback mechanism through which protein synthesis may be regulated by these diffusible gases. Increases in cellular NO concentrations cause the release of proteinbound heme (71)(72)(73), and the induction of HO-1 (11). The HO-1-catalyzed breakdown of heme would not only be critical in protecting cells from oxidative damage that would result from the presence of elevated concentrations of free heme, but would also produce elevated concentrations of CO, which could act as a feedback inhibitor of HRI activation induced by NO. The overall effect of these diffusible gases on HRI activity and protein synthesis would ultimately be determined by the relative concentrations to which they accumulate within a cell. High levels of NO might stimulate pro-apoptotic pathways to a degree that could not be reversed by a subsequent elevation in CO. Indeed, NO readily nitrosylates intracellular free heme and prevents its degradation by HO (74). In contrast, low levels of NO might lead to the generation of sustained, elevated levels of CO, which might then promote cell growth. Furthermore, we speculate that the ability of elevated CO levels to inhibit the activity of HRI and suppress eIF2␣ phosphorylation may be a critical event in protecting cells from NO-induced apoptotic cell death.
In summary, we acknowledge that the question of whether HRI is present in cells from tissues of nonerythroid origin is contentious, as it difficult to assure that the source of the HRI mRNA or protein is not from occluded blood. Furthermore, the expression of HRI in cultured nonerythroid cells could be a result of the aberrant phenotype of immortalized cells. In addition, our results indicate that HRI expression in NT2 and C2C12 is an order of magnitude or more lower than its level of expression in reticulocytes. However, with the acknowledgment of these caveats, the results presented here suggest that the following hypotheses are worthy of further investigation: (i) that activation of HRI may contribute to pathophysiologies that result from chronic exposure of normal cells to elevated NO levels and (ii) that HRI may be a potential target for design of tumoricidal agents directed at transformed cell populations expressing HRI.