MOLECULAR AND CELLULAR BIOLOGY, Apr. 1989, p. 1576-1586 Vol. 9, No. 4 0270-7306/89/041576-11$02.00/0 Copyright C) 1989, American Society for Microbiology Purification and Activation of the Double-Stranded RNA-Dependent eIF-2 Kinase DAI MATTHEW KOSTURAt AND MICHAEL B. MATHEWS* Cold Spring Harbor Laboratory, P.O. Box 100, Cold Spring Harbor, New York 11724 Received 14 September 1988/Accepted 3 January 1989

The double-stranded RNA (dsRNA)-dependent protein kinase DAI (also termed dsl and P1) possesses two kinase activities; one is an autophosphorylation activity, and the other phosphorylates initiation factor eIF-2. We purified the enzyme, in a latent form, to near homogeneity from interferon-treated human 293 cells. The purified enzyme consisted of a single polypeptide subunit of -70,000 daltons, retained its dependence on dsRNA for activation, and was sensitive to inhibition by adenovirus VA RNA,. Autophosphorylation required a suitable concentration of dsRNA and was second order with respect to DAI concentration, which suggests an intermolecular mechanism in which one DAI molecule phosphorylates a neighboring molecule. Once autophosphorylated, the enzyme could phosphorylate eIF-2 but seemed unable to phosphorylate other DAI molecules, which implies a change in substrate specificity upon activation. VA RNA, blocked autophosphory- lation and activation but permitted the activated enzyme to phosphorylate eIF-2. VA RNA, also blocked the binding of dsRNA to the enzyme. The data are consistent with a model in which activation requires the interaction of two molecules of DAI with dsRNA, followed by intermolecular autophosphorylation of the latent enzyme. VA RNA, would block activation by preventing the interaction between DAI and dsIINA.

Phosphorylation of the a subunit of eucaryotic protein state, and activation is accompanied by p68 phosphoryla- synthesis initiation factor eIF-2 (eIF-2oa) prevents catalytic tion; on the assumption that p68 is a single polypeptide recycling of the protein in polypeptide chain initiation and species and that no other protein is involved, the activation results in the inhibition of protein synthesis (reviewed in event would constitute an autophosphorylation. Detailed reference 30). Control of translational initiation by this study of the enzyme has been hampered by difficulties in mechanism has been demonstrated, or at least inferred, in a purifying it without either eliciting its activation or losing its variety of systems, including hemin-deprived rabbit reticu- dsRNA dependence, and most attempts to isolate the en- locytes, -infected mammalian cells, and cells subjected zyme have met with only moderate success. Nevertheless, to heat shock or changes in redox state (see reference 21 for considerably enriched preparations have been obtained from reviews). Although many effectors and environmental con- the ribosomal salt wash fraction of both rabbit reticulocytes ditions influence the level of eIF-2 phosphorylation, only (32-34) and interferon-induced tissue culture cells (3, 13, 38, two protein kinases that are capable of catalyzing this 41, 44, 45). Highly purified preparations of DAI contain a reaction have been identified (26): the hemin-controlled single major polypeptide subunit of -70 kDa, and the native regulator HCR, found chiefly in reticulocytes, and the dou- enzyme has a molecular mass of approximately 70,000 Da, ble-stranded RNA (dsRNA)-activated inhibitor (DAI, also as determined by sedimentation velocity and gel filtration. known as P1, dsl, and PKdS). DAI, also found in reticulo- These observations suggested that the p68 phosphoprotein cytes, is present at low levels in most mammalian cells and and DAI are identical, but the hypothesis remained un- tissue culture cell lines and at greatly increased levels after proven since complete purity was not achieved. Taking an interferon treatment (reviewed in references 1, 10, and 19). alternative approach, Hovanessian et al. generated a mono- The enzyme plays a role in establishment of the antiviral clonal antibody against the phosphorylated p68 protein (17). state, and several have elaborated defense measures The antibody, coupled to a resin matrix, was able to immo- to counter this form of interference in their replication (12, bilize DAI, which confirmed that the p68 protein was in- 14, 15, 29, 35, 36, 40, 43). Viruses may also exploit the volved in the activity of DAI, but elution of the immunoad- enzyme to achieve selective production oftheir own proteins sorbent yielded a polypeptide of -48 kDa (p48) in addition to (12, 26a). The ubiquity of the enzyme suggests that it may p68 (6). Further study led to the suggestion that DAI is a function in normal, uninfected cells (for example, during dimer of p68 and p48 apd that these two species have growth and differentiation events [31]), but this idea remains dissimilar kinase activities; p48 was thought to phosphory- to be fully substantiated. late p68 in the presence of dsRNA, and phosphorylated p68 Exposure of cell extracts containing DAI to dsRNA re- was held to possess the eIF-2a kinase activity. More recent sults in phosphorylation of a protein of -70 kilodaltons data, however, indicate that DAI is degraded during affinity (kDa) as well as the 36-kDa subunit of eIF-2 (4, 18, 37, 44). purification and that p48 is a proteolytic breakdown product The -70-kDa protein, often referred to as p68, is associated (7), casting doubt on the description of the enzyme as a with DAI during purification and is generally believed to heterodimer. comprise the enzyme. The enzyme exists in a latent, inactive DAI is activated by very low concentrations of dsRNA, but higher concentrations inhibit enzyme activation (9). Short * Corresponding author. (<50 base pairs) or imperfect duplexes are unable to activate t Present address: Department of Biochemical and Molecular DAI (9, 23), and adenovirus VA RNA,, a molecule with Pathology, Merck, Sharp & Dohme Research Laboratory, Rahway, extensive secondary structure (22, 24), inhibits activation NJ 07065. by competent dsRNA (14, 26a, 27, 39, 42). On the basis of 1576 VOL. 9, 1989 INTERFERON-INDUCED PROTEIN KINASE DAI 1577 these observations, we proposed a model for DAI activation monitored by dsRNA-dependent phosphorylation of p68, in which phosphorylation of p68 occurs only when two eIF-2 kinase activity, and sodium dodecyl sulfate (SDS)- molecules interact with one another on a single dsRNA polyacrylamide gel analysis of samples from column frac- molecule (28). The model also proposed that VA RNA acts tions, followed by silver staining. DAI eluted at -180 to 200 by interfering with the binding of DAI to dsRNA. To test this mM KCl. The entire DAI kinase peak was pooled and model, we have purified DAI to apparent homogeneity and concentrated by dialysis against buffer B plus 50 mM potas- studied its activation and interactions with dsRNA and VA sium phosphate (pH 7.2) and 20% sucrose. The dialyzed RNA,. The data largely corroborate the model and suggest S-Sepharose pool was loaded onto a 2.5- by 8.0-cm column that activation is accompanied by a change in the substrate of hydroxylapatite (HPHT grade; Bio-Rad Laboratories, preference of the enzyme. Richmond, Calif.) equilibrated with buffer B plus 50 mM potassium phosphate (pH 7.2). The column was washed with MATERIALS AND METHODS 1 column volume of the same buffer and then eluted with a 300-ml 50 to 500 mM potassium phosphate gradient made up Materials. Human recombinant a-interferon (IFN) was the in the start buffer. The dsRNA-dependent p68 autophospho- kind gift of Paul Trotta (Schering Corp., Bloomfield, N.J.). rylation and eIF-2 kinase activity coeluted from the column Aprotinin, leupeptin, and pepstatin were from Boehringer- with a protein of -70 kDa. DAI activity was pooled and Mannheim Biochemicals (Indianapolis, Ind.). Phenylmethyl- dialyzed against 100 volumes of buffer B (pH 6.8) with 20% sulfonyl fluoride was from Sigma Chemical Co. (St. Louis, sucrose. The dialyzed hydroxylapatite pool was loaded onto Mo.). Human 293 cells, a line of adenovirus type 5-trans- an HR 5/10 Mono S column (Pharmacia) and eluted at 1.0 formed human embryo kidney cells (8), were propagated in in B suspension cultures in F-13 medium (GIBCO Laboratories, m/min with a 40-ml 50 to 500 mM KCl gradient buffer Grand Island, N.Y.) supplemented with 10% calf serum. (pH 6.8). DAI was readily identified in silver-stained gels as Preparation of extracts from IFN-treated cells. Typically, a prominent -70-kDa protein. Individual kinase fractions 40 liters of 293 cells at 3 x 105 to 4 x 105 cells per ml was were stored at -70°C. The more highly enriched fractions treated for 18 h with 800 U of IFN per ml to increase the were subjected to 15 to 35% glycerol gradients for prepara- yields of DAI. All subsequent steps were carried out at 0 to tion of highly purified DAI. Portions (150 to 200 pI) of Mono 4°C. Cells were harvested by centrifugation at 1,000 x g, S-purified DAI were layered onto gradients of 15 to 35% washed with phosphate-buffered saline and then with buffer glycerol in buffer B (pH 7.4) and centrifuged at 49,000 rpm A (10 mM KCl, 20 mM N-2-hydroxyethylpiperazine-N'-2- for 22 h at 4°C in a Beckman SW50.1 rotor. Fractions (150 ethanesulfonic acid [HEPES] [pH 7.4], 1.5 mM MgCl2, 0.1 pI) were taken from the top by using an Auto Densiflow mM EDTA, 1 mM dithiothreitol, and protease inhibitors [1 fraction collector (Buchler Instruments Div., Nuclear-Chi- ,ug of aprotinin, leupeptin, and pepstatin per ml and 1 ,uM cago Corp., Fort Lee, N.J.), and 10 pI was analyzed by gel phenylmethylsulfonyl fluoride]). The final cell pellet was electrophoresis and silver staining. Gradient-purified kinase resuspended in 3 volumes of buffer A. The cells were was pooled and frozen at -70°C in smaller portions. allowed to swell for 20 min and were broken in a tight-fitting Isolation of VA RNA. VA RNA, was purified from adeno- Kontes Dounce homogenizer. Nuclei and cellular debris virus-infected HeLa cells essentially as described previously were removed by centrifugation at 30,000 x g for 20 min. (27). The VA RNA was adsorbed to and eluted from CF-11 The supernatant (S-30) was brought to 100 mM KCl by the cellulose by the method of Franklin (5) after elution from addition of 4 M KCl with gentle stirring and centrifuged at denaturing acrylamide gels. 60,000 rpm for 1.5 h in a type 60 rotor (Beckman Instru- Preparation of dsRNA. Synthetic dsRNA was prepared by ments, Inc., Fullerton, Calif.). The ribosomal pellet was hybridizing sense and antisense transcripts generated in suspended in buffer A plus 0.8 M KCl to 0.25 times the vitro. The template was plasmid pSPT-CAT#2, provided by volume of the original S-30 and vigorously homogenized C. Dery (Cold Spring Harbor Laboratory, Cold Spring with a tight-fitting Dounce homogenizer. The salt-extracted Harbor, N.Y.). It contains a 542-base-pair fragment from the ribosomes were centrifuged again at 60,000 rpm for 1.5 h in bacterial chloramphenicol acetyltransferase cloned into a Beckman type 60 rotor, and the resulting supernatant was the polylinker of pSPT-18 (Pharmacia). Cleavage of this dialyzed for 16 h against 4 liters of buffer B (pH 7.4) (50 mM plasmid with the appropriate restriction enzyme followed by KCl, 20 mM HEPES [pH 7.4], 1.5 mM MgCl2, 0.1 mM transcription with either SP6 or T7 polymerase produced EDTA, 1 mM dithiothreitol, 10% glycerol, and protease complementary single-stranded . Hybridization was inhibitors). The dialysate (ribosomal salt wash) was centri- carried out for 2 h at 50°C in 40 mM piperazine-N,N'- fuged to remove suspended solids and stored at -70°C. bis(2-ethanesulfonic acid) (PIPES) (pH 6.4)-400 mM NaCl-1 Purification of DAI. The ribosomal salt wash from 200 mM EDTA-80% formamide. The hybridization reaction was liters of 293 cells was used for a single preparation of DAI diluted with 10 volumes of 10 mM Tris hydrochloride (pH kinase. Dialyzed salt wash was first passed over a 2.5- by 7.5)-200 mM NaCl-100 mM LiCl-1 mM EDTA and digested 20-cm column of DEAE-cellulose (Sigma) equilibrated in for 1 h at 20°C with 40 ,ug of RNase A per ml. After the buffer B (pH 7.4). DAI was collected in the flowthrough addition of 0. 1% SDS and 0.1 mg of proteinase K per ml, the fraction. The bound material was eluted with buffer B (pH mixture was incubated for 30 min at 20°C and extracted with 7.4) plus 1 M KCl and was a source of eIF-2 as well as a phenol and chloroform; RNA was precipitated with ethanol potent inhibitor of kinase activation. The flowthrough frac- and dried. The dsRNA was further purified over CF-11 tion was titrated to pH 6.8 by dropwise addition of concen- cellulose (5). Typically, this protocol yielded dsRNA that trated HCl with stirring and loaded onto a 2.5- by 20-cm was slightly heterogeneous in size. Labeled dsRNA was column of S-Sepharose Fast Flow (Pharmacia Fine Chemi- made by including [t-32P]UTP (Dupont, NEN Research cals, Piscataway, N.J.) equilibrated with buffer B (pH 6.8). Products, Boston, Mass.) in the transcription reactions The column was washed with 1 column volume of buffer B (calculated final specific activity, -10' dpm/pug). (pH 6.8) and eluted with a 1-liter linear gradient of 50 to 500 Nitrocellulose filter-binding assays. Filter-binding assays mM KCl in buffer B (pH 6.8). Elution of DAI activity was were performed by using glycerol gradient-purified DAI and 1578 KOSTURA AND MATHEWS MOL. CELL. BIOL.

PURIFICATION OF DAI addition of 10 volumes of buffer A (pH 7.4) containing 0.1 mg of bovine serum albumin and 0.1 mg of tRNA per ml, and immediately filtered through alkali-washed nitrocellulose. IFN TREATED 293 CELLS Filtration was performed on a slot blot apparatus (Schleicher & Schuell, Inc., Keene, N.H.) at a slow flow rate. The filters were subsequently washed with 100 p.l of diluent buffer, RIBOSOMAL SALT WASH dried, autoradiographed, and subjected to liquid scintillation counting. Kinase assays. Standard kinase assays consisted of 2 ,ul or DEAE-CELLULOSE less (depending on the purification step) of DAI fraction diluted to 10 pl with buffer B (pH 7.4) containing 0.1 mg of bovine serum albumin and 0.1 mg of tRNA per ml. The eIF-2 and DAI INHIBITOR diluted kinase was added to 20-pI reaction mixtures contain- ing, at final concentrations, 75 mM KCI, 25 mM HEPES (pH 7.4), 10 mM MgCl2, 1.0 mM dithiothreitol, 0.1 mM EDTA, S-SEPHAROSE FAST FLOW 0.1 mM ATP, protease inhibitors, and 5 to 10 puCi of [y-32P]ATP (>3,000 Ci/mmol; Dupont, NEN). Reaction mix- tures were supplemented when needed with reovirus dsRNA HYDROXYLAPATITE (generously provided by A. Shatkin) or synthetic dsRNA as an activator and VA RNA as an inhibitor. When used in the same reaction, dsRNA and VA RNA were added simulta- FPLC: MONO S neously to the enzyme mix, on ice. The reactions were incubated at 30°C and quenched with an equal volume of 2 x -concentrated SDS-gel sample buffer. GLYCEROL GRADIENT SDS-polyacrylamide gels and silver staining. Gels (12.5%) were run by the method of Laemmli (16), fixed in 40% FIG. 1. Scheme for fractionation of DAI from IFN-treated 293 20 washed with cells. FPLC, Fast protein/peptide/polynucleotide liquid chromatog- methanol-10% acetic acid (vol/vol) for min, raphy. water for 20 min, and then dried for autoradiography. Radioactive bands were quantified by excision of the appro- priate region from the gel and counting in a liquid scintilla- labeled dsRNA. Standard kinase reactions (without labeled tion counter. For silver staining, gels were first fixed for 15 ATP) were assembled on ice with 2 pul (20 ng) of DAI and min in 50% methanol and then rinsed in distilled water for 20 appropriate quantities of dsRNA and unlabeled competitors. min. The gel was then soaked for 15 min in a solution The reactions were held on ice for 15 min, quenched by the containing (by volume) 40% methanol, 10% ethanol, and 1%

A B Ft ac t ion RSW RSW DEAE DEAE S-Seph HAP Mcno S : 10 2 5 - CIL Lii U1) -T E-- -- !- ---r--r - (l) + + ++ + + +4+ + + + + + + + VA- Cl) LUJ

94- _ 67- X4 43-s

30-

FIG. 2. Purification of DAI from ribosomal salt wash. (A) Silver-stained gel of column pools up to the penultimate Mono S chromatography step (see Fig. 1). Each lane contained 2 ,ug of total protein. Arrow marks the position of p68 (DAI). (B) Assay of the pooled column fractions for kinase activity. The amounts of protein assayed from the ribosomal salt wash (RSW) and the DEAE-cellulose, S-Sepharose (S-Seph), hydroxylapatite (HAP), and Mono S pools are indicated. The amount of protein assayed from the glycerol gradient (Gly grad.) fraction was estimated to be 0.1 pLg on the basis of the intensity of staining by silver. Reovirus dsRNA (20 ng/ml) and VA RNA (50 ,ug/ml) were present where indicated. Molecular size markers (in kilodaltons) are shown on the left. VOL. 9, 1989 INTERFERON-INDUCED PROTEIN KINASE DAI 1579

A. Froctlon B HAP mono S 10 .5 2 I- extroct l0 5 2 10 .5 .2 .1 H-1 F- - 7 - _ I + + II 11 11 We + + + + + ++ + + + + + reo20ng/ml +4 +4 4-4-+ +4 + + reo dsRNA 20ngImI + + + + VA509qg/m1 + + + + + VA RNA 503g/mI + + + + VA 2001.g/m

.l

100 r HAP mono S 1OCr

c c 0 0 50 - 50 H C 011 0--

0 - 10 0.1 12 5 xl enzyme /.LI enzyme FIG. 3. Effect of DAI concentration on VA RNA inhibition. (A) Various amounts of either the hydroxylapatite (HAP)- or Mono S-purified DAI were added to standard 20-i±l kinase assay reaction mixtures and incubated with the indicated amounts of reovirus dsRNA and VA RNA. The autoradiogram and quantitation of the excised bands are shown. Data are represented as the percent inhibition of phosphorylation of p68 protein. (B) Various amounts of Mono S-purified material were added to standard kinase reactions and incubated in the presence of 50 or 200 p.g of VA RNA per ml. p68 phosphorylation was quantified by excising the gel bands and estimating radioactivity in a liquid scintillation counter. glutaraldehyde, followed by three 10-min washes with water. passed successively through DEAE-cellulose, S-Sepharose Exposure of the gel to ammoniacal silver nitrate and devel- Fast Flow, and hydroxylapatite columns, followed by high- opment using formaldehyde and citric acid were essentially resolution chromatography on Mono S. The final purification according to published protocols (25). step, glycerol gradient sedimentation, was used to prepare highly purified DAI on a relatively small scale. It should be RESULTS noted that a fraction containing eIF-2 and a kinase inhibitor their presence Purification and characterization of DAI. DAI is a ribo- was removed by DEAE-cellulose adsorption; some-bound protein that exists in a latent form, exhibiting no hampered quantitation of the purification. The potent, as yet known kinase activity until activated by dsRNA. Upon uncharacterized inhibitor blocked DAI activation but not the incubation with dsRNA and ATP, phosphorylation of a eIF-2 kinase activity of activated DAI (data not shown). protein of -70 kDa (p68) is accompanied by activation of Results of a typical purification are shown in Fig. 2 and 3. eIF-2a kinase activity (19, 20); these properties serve as Silver staining of the pooled material from each column step convenient assays for the enzyme. To investigate the acti- clearly illustrates the enrichment of the -70-kDa protein vation process, we set out to purify the latent form of the (Fig. 2A). In all cases, the peaks of dsRNA-dependent p68 kinase by standard chromatographic techniques in combina- phosphorylation (Fig. 2B) and of eIF-2 kinase activity (not tion with high-resolution chromatography resins when pos- shown) copurified with this protein. Both activities were sible and by using a broad spectrum of protease inhibitors. strictly dependent on dsRNA at all stages of purification Initial subcellular fractionation of DAI and subsequent (Fig. 2B). They were also both inhibited by VA RNA, column steps are outlined in Fig. 1. As with other cell types although the degree of inhibition was low at some of the later (3, 38), nearly all of the latent DAI kinase activity in stages of purification, where the enzyme concentration was IFN-treated 293 cells was concentrated in the ribosomal salt greatest. Full sensitivity returned after the final step, glyc- wash fraction (data not shown). The ribosomal salt wash was erol gradient centrifugation, during which the enzyme was 1580 KOSTURA AND MATHEWS MOL. CELL. BIOL.

A. 2 4 6 8 10 12 14 16 18 20 22 3 5 7 9 1i 13 15 17 19 21 23

94 --- m 67 -- 4F_

43 --

30-

B Fraction number 5 6 7 8 9 10 11 12 f-Fr ++ ++ *++ ++ ++ ++ ++ reo dsRNA * + + + + + VA RrIA + e. lre _, C, i,:X;-S4 4

O 0 *e

FIG. 4. Glycerol gradient fractionation of purified DAI. (A) Silver-stained SDS-polyacrylamide gel of lO-,ul portions from glycerol gradient fractions. Sedimentation is from left to right; molecular size markers (in kilodaltons) are shown on the left. In parallel gradients, bovine serum albumin sedimented in fractions 7 through 10. (B) Assay of gradient fractions for kinase activity; 2-,ul portions were assayed in the presence of reovirus dsRNA (100 ng/ml) or VA RNA (50 ,ug/ml) as indicated.

once again diluted, suggesting a dependence on enzyme fractions exhibited all of the characteristics of DAI. The concentration. This explanation was tested by using DAI peak of dsRNA-dependent, VA RNA-inhibited p68 phos- from both the hydroxylapatite and Mono S fractions which phorylation, as well as the peaks of eIF-2 kinase activity and exhibited little sensitivity to VA RNA in Fig. 2. The results of dsRNA-binding activity (not shown), cosedimented with of a titration of enzyme, at set concentrations of VA RNA the predominant -70-kDa protein observed in the gradient and reovirus dsRNA, demonstrated that the degree of inhi- (Fig. 4). The activity sedimented at the same position as did bition by VA RNA increased as the enzyme was diluted (Fig. bovine serum albumin, a protein of 67,000 Da, run in a 3A). Likewise, at elevated enzyme levels, the addition of parallel gradient. Therefore, we conclude that DAI is a higher concentrations of VA RNA effectively restored inhi- protein of -70 kDa that exists as a monomer in solution. All bition of p68 phosphorylation (Fig. 3B). of the known characteristics of DAI copurified with this Purification of DAI from highly enriched Mono S fractions protein, and no other polypeptide appeared to be required by sedimentation through glycerol gradients resulted in for their manifestation. In particular, there was no indication fractions containing a protein of -70 kDa at greater than of the 48-kDa protein reported to be present in stoichiomet- 90% purity, as judged by silver staining (Fig. 4A). These ric amounts with p68 (6). VOL. 9, 1989 INTERFERON-INDUCED PROTEIN KINASE DAI 1581

tration of reovirus dsRNA. Control stage 1 reactions were run without kinase to determine the amount of latent kinase that would be activated in stage 2. The stage 2 reaction 2.0 mixtures contained 0.05 or 0.25 volumes of stage 1 material (Fig. 6, lanes headed 1 and 5, respectively) together with [_y-32P]ATP and additional DAI, eIF-2, and VA RNA in different combinations. Figure 6A through C shows the effects of preactivated 0 DAI on p68 phosphorylation. Activation of DAI in stage 1 was virtually complete, since little or no phosphorylation of p68 was detected in stage 2 reaction mixtures that contained no additional DAI (Fig. 6A). When stage 1 reaction mixtures ioF~ were added to stage 2 reaction mixtures containing latent DAI, p68 phosphorylation occurred to an approximately equal extent regardless of the amount of activated DAI l1.- transferred from stage 1 (Fig. 6B). This phosphorylation of 0.2 0.5 1.0 p68 in stage 2 was inhibited almost completely by the Log [protein] addition of VA RNA (Fig. 6C). To verify that activated kinase was in the stage 1 eIF-2 kinase FIG. 5. Kinetics of p68 autophosphorylation in partially purified generated reactions, kinase preparations. DAI equivalent to the S-Sepharose fraction activity was also assayed in stage 2 (Fig. 6D through G). was added in various amounts to standard kinase assays containing eIF-2 kinase activity was indeed present in stage 2 reactions 100 ng of reovirus dsRNA per ml and incubated at 30°C. Equal that received DAI from stage 1 (Fig. 6D). The activity was portions of the reaction mixture were taken at various times of largely, but not completely, resistant to the presence of VA incubation up to 40 min and subjected to SDS-gel electrophoresis RNA in stage 2 (Fig. 6E). When eIF-2 and additional DAI and autoradiography. Phosphorylation of the p68 protein was deter- were added in stage 2, both species became phosphorylated mined as for Fig. 3. The rate of phosphorylation was determined as expected; as in panel B, phosphorylation of p68 was not over the linear range of the reactions for each protein concentration. significantly altered by the presence of preactivated DAI The initial rates, v, were 9.5, 38.5, and 136.5 cpm/min at 2.5, 5, and from stage 1 (Fig. 6F). As in panel C, addition of VA RNA 10 p.g/ml, respectively, giving a slope of 1.92 as determined by linear caused a profound reduction in p68 phosphorylation in stage regression analysis. 2 (Fig. 6G); addition of VA RNA also caused a smaller reduction in eIF-2a phosphorylation. (Note that the eIF-2 Second-order activation of DAI. An important prediction of kinase activity in Fig. 6G represents the sum of the activity the model for DAI activation proposed by O'Malley et al. transferred from stage 1 and the residual activity generated (28) is that p68 phosphorylation should be a bimolecular in stage 2 despite the presence of VA RNA; therefore, eIF-2 event. To determine the kinetic order of the reaction, we phosphorylation was greater in reactions containing DAI measured the rate of p68 phosphorylation at different en- from stage 1.) zyme concentrations and fixed, nonlimiting concentrations Two principal conclusions can be drawn from these ex- of dsRNA and other components. For an intramolecular periments. First, in the presence of VA RNA, the addition of autophosphorylation, the reaction rate would be expected to preactivated DAI makes essentially no difference to the be independent of enzyme concentration. Results showed phosphorylation of p68 in stage 2 reactions (Fig. 6C and G). that the rate of phosphorylation changed as a function of Therefore, autophosphorylation does not occur in the pres- protein concentration; a logarithmic plot of the data (Fig. 5) ence of VA RNA even if some of the DAI is already gave a slope of 1.92, indicating that the change in rate of p68 activated. The activated enzyme can, however, phosphory- phosphorylation was proportional to the square of its con- late eIF-2 under these conditions (Fig. 6E and G). Second, centration. This observation suggested that two protein even in the absence of VA RNA, preactivated DAI exerted species were involved in the phosphorylation of p68. Since no detectable effect on p68 pholsphorylation (Fig. 6B and the kinase appeared to exist as a monomer and no other F), although eIF-2 phosphorylation was greatly affected protein components were required for autophosphorylation, (compare Fig. 6B and D). Since DAI was not fully phosphor- we conclude that kinase activation involves an intermolecu- ylated in any of the stage 2 reactions because of limiting lar reaction between two DAI molecules, possibly one acting concentrations of dsRNA, it seems that activated DAI also as an enzyme and the other acting as a substrate. does not phosphorylate latent DAI under these conditions. It Change in DAI specfficity upon activation. DAI mediates is possible, therefore, that latent DAI is not a substrate for two kinds of phosphorylation events: the primary event, in activated DAI. Finally, these results suggest that VA RNA which a DAI molecule serves as a substrate, resulting in can inhibit the eIF-2 kinase activity of DAI by about activation of latent DAI; and a secondary event, in which the twofold. Nevertheless, comparison of Fig. 6F and G shows substrate is eIF-2. It is well established that the activated that the degree of inhibition was much greater for autophos- form, and only the activated form, of DAI can phosphorylate phorylation than for eIF-2 phosphorylation, consistent with eIF-2, but does it retain the ability to phosphorylate other previous reports (14, 27, 28). It also appears that eIF-2 may DAI molecules? To address this question, we performed reduce p68 phosphorylation (Fig. 6B and F), but it is not yet mixing experiments and also took advantage of the inhibi- clear whether this is a specific effect or a result of contami- tory properties of VA RNA,. This adenoviral transcript nation with the kinase inhibitor mentioned above. blocks dsRNA-dependent phosphorylation of p68 but allows One dsRNA-binding site on DAI. The observation that DAI DAI that is already activated to phosphorylate eIF-2 (14, 27, is activated by low concentrations of dsRNA but not by high 28). The mixing experiments were performed in two stages. concentrations has long been an enigma. O'Malley et al. (28) The first, or activating, stage consisted of a standard kinase proposed that high dsRNA concentrations disfavor the si- reaction containing DAI, unlabeled ATP, and a low concen- multaneous binding of two molecules of DAI to a single 1582 KOSTURA AND MATHEWS MOL. CELL. BIOL.

A B 5 5 51 5 * at,,ge - -+ + _ - - + + DAI

Stage 2 + + + + DAI + + + + VA

* --p68

D. E. F G. 5 5 5 1 5 1 5 1 5 1 5 1 5 -- T- f T- -- rA Stage -- + + + + - + + DAI t-" T + + +++ + + + + DAI Stage 2 + + ++ t + + + + + + + eIF-2 + + + + + + + + VA

_ _ _ -- e6F

-0- 4 w- 44m eIF-2X<

FIG. 6. Results of mixing experiments using activated and latent DAI. Stage 1 reactions contained standard kinase components with 20 ng of reovirus dsRNA per ml and either no DAI (-) or 25 ng of the peak Mono S fraction (+). The reactions were incubated at 30°C for 20 min with unlabeled ATP and then placed on ice. Portions of stage 1 reaction mixtures (1 or 5 ,ul as indicated) were added to stage 2 reaction mixtures (20 ,ul) containing various combinations of Mono S-purified DAI (25 ng), purified eIF-2 (1 ,ug), and VA RNA (50 ,ug/ml). After incubation with [y-32P]ATP (10 ,uCi) for 20 min at 30°C, reactions were analyzed by gel electrophoresis and autoradiography.

dsRNA molecule and that such simultaneous binding is (Fig. 7A). The specificity of binding was demonstrated by required for activation. An alternative explanation might be showing that unlabeled tRNA (routinely included in the that DAI bears two sites for dsRNA binding, an activating binding assay at a high concentration) as well as rRNA, site which has high affinity for dsRNA and an inhibitory site dsDNA, and the single-stranded RNA made in the in vitro of lower affinity. These models can be distinguished by transcription reactions were all unable to inhibit binding of measuring the number of dsRNA-binding sitest the second labeled dsRNA when tested at concentrations of up to 100 model requires the existence of at least two sites, whereas ,ug/ml, whereas heat treatment of the kinase at 100°C for 2 the first model makes no specific prediction about the min abolished dsRNA-binding activity (data not shown). number of sites. These results indicate that the p68 protein contains a specific We examined the binding of dsRNA to DAI by using and high-affinity dsRNA-binding site. Scatchard analysis of synthetic labeled dsRNA as a probe. The dsRNA was the binding data indicated the existence of a single dsRNA- produced by annealing complementary RNA strands tran- binding site on DAI (Fig. 7B), consistent with the first scribed in vitro and was purified by nuclease digestion and model. dsRNA became inhibitory toward DAI activation at cellulose chromatography. It possessed all of the character- concentrations of more than about 100 ng/ml (data not istics of an authentic dsRNA activator and was equivalent to shown); had a second, inhibitory site for dsRNA existed, it reovirus dsRNA in ability to activate DAI (both the degree would have been revealed by a transition in the slope of the of activation and amount of dsRNA required were the same). curve in Fig. 7B at values corresponding to high concentra- The activation was accompanied by phosphorylation of p68 tions of dsRNA (i.e., at high values of B). We conclude that and was inhibited by VA RNA (data not shown). Binding of inhibition by excess dsRNA is not due to second-site bind- labeled dsRNA to DAI was assessed by using a nitrocellu- ing. lose filter-binding assay. When incubated at 0°C under Inhibition by VA RNA, of binding of dsRNA to DAI. standard kinase assay conditions, the DAI-mediated binding Because VA RNA, exhibits considerable secondary struc- of dsRNA to nitrocellulose filters was specific and saturable ture, including duplexed regions, it was suggested that VA VOL. 9, 1989 INTERFERON-INDUCED PROTEIN KINASE DAI 1583

4

3 rI) 0 E

CL cr

(A C-CV crJ 4

LVV*v V [dsRNA] input ng/mi FIG. 7. Binding of labeled dsRNA to nitrocellulose filters. (A) Labeled synthetic dsRNA was incubated at O'C with glycerol gradient-purified DAI, and dsRNA-DAI complexes were separated from free dsRNA by filtration through nitrocellulose. Binding was quantified by liquid scintillation counting of nitrocellulose-bound labeled dsRNA. Symbols represent data from three dilutions of the dsRNA. (B) Scatchard analysis of the data. B, Bound counts per minute; F, free counts per minute.

RNA might act as an inhibitor of dsRNA binding to DAI phosphorylated the p68 subunit; as in more conventional (35). To test this idea, we conducted competition assays to models, the latter moiety was thought to possess the eIF-2 assess the ability of unlabeled dsRNA and VA RNA to block kinase activity. Our data are not compatible with this model, the binding of labeled dsRNA to DAI. Both species were which has now been withdrawn (7), particularly in that we do able to compete for binding, although with greatly different not detect a subunit of 48 kDa. Nevertheless, the demon- efficiencies (Fig. 8). A 50% inhibition of dsRNA binding stration that DAI activation is a second-order reaction with required 50 ng of unlabeled dsRNA per ml but about 500 respect to p68 (Fig. 5) suggests that two molecules of the times more VA RNA,. These concentrations of VA RNA (10 latent enzyme interact during the activation step, consistent to 25 ,ug/ml) are equivalent to those required for inhibition of with an intermolecular reaction in which one molecule acts p68 phosphorylation (14, 27). Thus, VA RNA can specifi- as an enzyme and the other acts as a substrate. Since p48 is cally inhibit the binding of dsRNA to DAI. now thought to be a product of proteolytic breakdown of p68, it is conceivable that the "auto-kinase" catalytic site is DISCUSSION retained by the 48-kDa fragment and that intact p68 func- We have purified to apparent homogeneity a protein of tions as its substrate. -70 kDa (p68) that possesses all known properties of DAI. The phosphorylation of DAI seems to serve as a switch, These include dsRNA-dependent autophosphorylation of changing the substrate preference of the enzyme from other the p68 protein, the accompanying activation of eIF-2 ki- latent DAI molecules to eIF-2. Three lines of evidence nase, inhibition of both of these kinase activities by VA support this conclusion. First, activated DAI does not phos- RNA,, and the ability to bind and retain dsRNA on nitrocel- phorylate latent DAI to a detectable degree, although it does lulose filters in a fashion that is inhibited by VA RNA. An phosphorylate eIF-2 (Fig. 6). Second, VA RNA blocks important physical characteristic of this preparation, shared phosphorylation of p68 by activated DAI but has little effect with the product of most other classical purification proce- on eIF-2 phosphorylation (Fig. 6). In principle, this result dures, is that DAI sediments as a protein of approximately could be interpreted differently, as a direct effect of VA 3.8S in glycerol gradients (Fig. 4), which suggests that it RNA on autophosphorylation independent of its effect on exists as a monomer in solution. At the time this study was dsRNA binding (resulting from a masking of phosphoryla- initiated, however, Galabru and Hovannessian (6) had re- tion sites on the DAI substrate, for example). However, ported that DAI is a heterodimer of -110 kDa containing there is no evidence favoring this alternative and it seems subunits of 68 and 48 kDa. They proposed a model in which unlikely, although it cannot be excluded on the basis of the the p48 subunit constituted the "auto-kinase" activity and data now available. Third, the level of p68 phosphorylation 1584 KOSTURA AND MATHEWS MOL. CELL. BIOL.

[dsRNA]>qi/ml [VA RNA>g/rnl

O 0 L°O0 O- O LO C)O0 0 00 LC) o - CNu Cq ° o

gI I gIIg I 1 111

4.0

o.0 D~< I\\. 20 o 0.L° J^ ILX

.001 10 0 00 Corrmpeti'tor RNA 4g/ml FIG. 8. Binding of labeled dsRNA to nitrocellulose filters in the presence of unlabeled dsRNA and VA RNA. Filter-binding assays using 10,000 dpm of labeled dsRNA and glycerol gradient-purified DAI were performed in the presence of various concentrations of unlabeled synthetic dsRNA or VA RNA. Binding was visualized by autoradiography (top panels) and quantified by liquid scintillation counting of the bands.

depends on the amount of dsRNA present in the reaction, blocked by an excess of the activator. The explanation and longer incubation times do not compensate for lower proposed previously (28) supposes that low concentrations dsRNA concentrations in achieving maximal p68 labeling of dsRNA favor the formation of bimolecular protein-protein (data not shown). Since activated kinase is being generated complexes on a single dsRNA molecule, whereas elevated in these assays yet incomplete phosphorylation of p68 oc- concentrations of dsRNA lessen the probability of these curs, it would appear that activated kinase does not effi- complexes forming. The demonstration that activation is ciently recognize latent DAI as a typical eIF-2-like substrate second order with respect to DAI concentration lends sup- for phosphorylation during the course of these reactions. port to this idea. An alternative explanation would suppose This conclusion is drawn from experiments conducted under that each DAI molecule has two sites for binding dsRNA, a conditions in which the availability of dsRNA is limited high-affinity activating site and a low-affinity inhibitory site. either by its low concentration or by the presence of VA Our failure to observe a second dsRNA-binding site (Fig. 7) RNA and which might preclude the binding of latent DAI to argues against this alternative. On the assumption that the dsRNA. It is possible, therefore, that latent DAI can act as enzyme exists as a monomer in solution and that activation a substrate for phosphorylation by activated DAI only when of DAI requires two enzyme molecules, what is the role of it is complexed with dsRNA. Alternatively, latent DAI may dsRNA in the activation sequence? The nitrocellulose filter- be refractory under all conditions. Unfortunately, we cannot binding data indicate that DAI binds dsRNA quite effi- at present distinguish experimentally between these two ciently; we estimate that the affinity of DAI for dsRNA is in mechanisms; in either case, however, the change in sub- the picomolar range, compatible with the amount of dsRNA strate preference upon activation would serve to prevent the required for maximum activation. Since there is apparently runaway autocatalytic activation of DAI by small quantities only one available site on DAI to which dsRNA can bind, it of dsRNA. seems likely that dsRNA acts as a high-affinity surface to Another feature of DAI, and a peculiarity which compli- which the enzyme can adsorb. Indeed, heparin, a sulfated cates studies of its enzymatic properties, is that activation is polysaccharide, can substitute for dsRNA (6), presumably VOL. 9, 1989 INTERFERON-INDUCED PROTEIN KINASE DAI 1585 by offering a similar array of negatively charged groups. The and substrate specificities of the double-stranded RNA depen- stereochemical specificity of the interaction is underlined by dent protein kinase from untreated and interferon-treated mouse the observation that a variety of related macromolecules, fibroblasts. J. Biol. Chem. 260:11240-11247. including DNA, 2'-O-methylated dsRNA, single-stranded 4. Farrell, P. J., K. Balkow, T. Hunt, R. J. Jackson, and H. Traschel. 1977. Phosphorylation of initiation factor eIF-2 and RNA, and RNA-DNA heteroduplexes (2, 9), are unable to the control of reticulocyte protein synthesis. Cell 11:187-200. activate DAI. Although no specific dsRNA sequences are 5. Franklin, R. M. 1966. Purification and properties of the replica- necessary for DAI activation, there is a length requirement tive intermediate of the RNA bacteriophage R17. Proc. Natl. of at least 50 base pairs of duplex structure (9, 23), consistent Acad. Sci. USA 55:1504-1511. with the notion that DAI activation requires two kinase 6. Galabru, J., and A. Hovanessian. 1985. Two interferon induced molecules to adsorb to a single dsRNA molecule. Therefore, proteins are involved in the protein kinase complex dependent we argue that phosphorylation and activation of DAI involve on double-stranded RNA. Cell 43:685-694. the formation of a complex of at least two proteins in 7. Galabru, J., and A. Hovanessian. 1987. Autophosphorylation of the protein kinase dependent on double-stranded RNA. J. Biol. proximity on the same dsRNA molecule and that at least one Chem. 262:15538-15544. member of this complex becomes phosphorylated. 8. Graham, F. L., J. Smiley, W. C. RusseUl, and R. Nairn. 1977. As noted above, adenovirus VA RNA, blocks the activa- Characteristics of a human cell line transformed by DNA from tion step but has little effect on the eIF-2 phosphorylation human adenovirus type 5. J. Gen. Virol. 36:59-72. activity of activated DAI. VA RNA binds to DAI (11; K. H. 9. Hunter, T., T. Hunt, R. J. Jackson, and H. D. Robertson. 1975. Mellits, M. Kostura, and M. B. Mathews, manuscript in The characteristics of inhibition of protein synthesis by double- preparation) and inhibits the binding of dsRNA to DAI (Fig. stranded ribonucleic acid in reticulocyte lysates. J. Biol. Chem. 8). Although VA RNA has duplex regions, these are not 250:409-417. sufficient for its function (22), and it is not clear whether the 10. Johnston, M. I., and P. F. Torrence. 1984. The role of interferon- two RNAs the same site. induced proteins, double-stranded RNA and 2',5'-oligoadeny- compete for Competition experi- late in the interferon-mediated inhibition of viral , p. ments with dsRNA showed that although dsRNA increased 189-298. In R. M. Friedman (ed.), Interferon 3: mechanisms of the levels of phosphorylation of p68 at all VA RNA concen- production and action. Elsevier/North-Holland Publishing Co., trations tested, it did not fully restore the labeling of p68 Amsterdam. protein to control levels (data not shown). Although it is 11. Katze, M. G., D. DeCorato, B. Safer, J. Galabru, and A. G. difficult to dismiss the possibility that the competitive relief Hovanessian. 1987. Adenovirus VAI RNA complexes with the from inhibition by VA RNA is masked by the inhibitory 68000 Mr protein kinase to regulate its autophosphorylation and effect of high concentrations of dsRNA, this observation activity. EMBO J. 6:689-697. suggests that VA RNA functions as a noncompetitive inhib- 12. Katze, M. G., B. M. Detjen, B. Safer, and R. M. Krug. 1986. a certain fraction of the enzyme from the Translational control by influenza virus: suppression of the itor, removing kinase that phosphorylates the alpha subunit of initiation factor reaction and making it unavailable to act as either a substrate eIF-2 and selective translation of influenza viral mRNAs. Mol. or a kinase. We propose that the initial binding of VA RNA, Cell. Biol. 6:1741-1750. requires one or more of its duplex regions and that a 13. Kimchi, A., A. Zilberstein, A. Schmidt, L. Shulman, and M. secondary interaction between DAI and other structures in Revel. 1979. The interferon-induced protein kinase PK-i from the VA RNA molecule results in inhibition of enzyme mouse L cells. J. Biol. Chem. 254:9846-9853. activation. These structures could lie in the nonduplexed 14. Kitajewski, J., R. J. Schneider, B. Safer, S. M. Munemitsu, C. E. central region of the VA RNA molecule, which has been Samuel, B. Thimmappaya, and T. Shenk. 1986. Adenovirus VAI implicated in the function of the molecule (22). Direct RNA antagonizes the antiviral action of interferon by prevent- analysis of the VA RNA-DAI interactions is in progress. ing activation of the interferon-induced eIF-2a kinase. Cell The data the described 45:195-200. presented here support previously 15. Kitajewski, J., R. J. Schneider, B. Safer, and T. Shenk. 1986. An bimolecular model for DAI activation (28) and suggest that adenovirus mutant unable to express VAI RNA displays dif- the inhibitory action of VA RNA rests on interactions which ferent growth responses and sensitivity to interferon in various include, but are not restricted to, its duplex regions and host cell lines. Mol. Cell. Biol. 6:4493-4498. result in the inhibition of dsRNA binding. Activation of the 16. Laemmli, U. K. 1970. Cleavage of structural proteins during the DAI system is poised to respond sensitively to changes in assembly of the head of bacteriophage T4. Nature (London) dsRNA levels. Coupled with active phosphatases against 227:680-685. p68 (13), such a system could function to limit DAI activa- 17. Laurent, A. G., B. Krust, J. Galabru, J. Svab, and A. G. tion to conditions in which a critical mass of dsRNA accu- Hovanessian. 1985. Monoclonal antibodies to an interferon- mulates in the cell. induced Mr 68,000 protein and their use for the detection of double-stranded dependent protein kinase in human cells. Proc. Natl. Acad. Sci. USA 82:4341-4345. ACKNOWLEDGMENTS 18. Lebleu, B., G. C. Sen, S. Shaila, B. Cabrer, and P. Lengyel. We thank Lisa Manche for excellent technical assistance and 1976. Interferon, double-stranded RNA, and protein phosphor- Claude Dery, Aaron Shatkin, and Paul Trotta for gifts of reagents. ylation. Proc. Natl. Acad. Sci. USA 73:3107-3111. This work was supported by Public Health Service program grant 19. Lengyel, P. 1982. Biochemistry of interferons and their actions. CA 13106 and by Public Health Service training grant 5T32 CA09311 Annu. Rev. Biochem. 51:251-282. to M.K., both from the National Cancer Institute. 20. Levin, D. H., R. Petryshyn, and I. M. London. 1981. Character- ization of purified double-stranded RNA activated eIF-2 kinase LITERATURE CITED from rabbit reticulocytes. J. Biol. Chem. 256:7638-7641. 1. Baglioni, C. 1979. Interferon-induced enzymatic activities and 21. Mathews, M. B. (ed.). 1986. Translational control. Cold Spring their role in the antiviral state. Cell 17:255-264. Harbor Laboratory, Cold Spring Harbor, N.Y. 2. Baglioni, C., S. Benvin, P. A. Maroney, M. A. Minks, T. W. 22. Mellits, K. H., and M. B. Mathews. 1988. Effects of mutations in Nilsen, and D. K. West. 1980. Interferon-induced enzymes: stem and loop regions on the structure and function of adeno- activation and role in the antiviral state. Ann. N.Y. Acad. Sci. virus VA RNA,. EMBO J. 7:2849-2859. 350:497-509. 23. Minks, M. A., D. A. West, S. Benvin, and C. Baglioni. 1979. 3. Berry, M. J., G. S. Knutson, S. R. Lasky, S. M. Munemitsu, and Structural requirements of double-stranded RNA for the activa- C. E. Samuel. 1985. Mechanism of interferon action: purification tion of 2',5'-oligo(A) polymerase and protein kinase of interfer- 1586 KOSTURA AND MATHEWS MOL. CELL. BIOL.

on-treated HeLa cells. J. Biol. Chem. 254:10180-10183. adenovirus virus-associated RNA,. Nature (London) 313:196- 24. Monstein, H.-J., and L. Philipson. 1981. The conformation of 200. adenovirus VAI-RNA in solution. Nucleic Acids Res. 9:4239- 36. Rice, A. P., and I. M. Kerr. 1984. Interferon-mediated, double- 4250. stranded RNA-dependent protein kinase is inhibited in extracts 25. Oakley, B. R., D. R. Kirsch, and N. R. Morris. 1980. A from vaccinia virus-infected cells. J. Virol. 50:229-236. simplified ultrasensitive silver stain for detecting proteins in 37. Roberts, W. K., A. Hovanessian, R. E. Brown, M. J. Clemens, polyacrylamide gels. Anal. Biochem. 105:361-363. and I. M. Kerr. 1976. Interferon-mediated protein kinase and 26. Ochoa, S. 1983. Regulation of protein synthesis initiation in low-molecular-weight inhibitor of protein synthesis. Nature eucaryotes. Arch. Biochem. Biophys. 223:325-349. (London) 264:477-480. 26a.O'Malley, R. P., R. F. Duncan, J. W. B. Hershey, and M. B. 38. Samuel, C. E., G. S. Knutson, M. J. Berry, J. A. Atwater, and Mathews. 1989. Modification of protein synthesis initiation S. E. Lasky. 1986. Purification of double-stranded RNA depen- factors and the shut-off of host protein synthesis in adenovirus- dent protein kinase from mouse fibroblasts. Methods Enzymol. infected cells. Virology 168:112-118. 119:499-515. 27. O'Malley, R. P., T. M. Mariano, J. Siekierka, and M. B. 39. Schneider, R. J., B. Safer, S. M. Munemitsu, C. E. Samuel, and Mathews. 1986. A mechanism for the control of protein synthe- T. Shenk. 1985. Adenovirus VAI RNA prevents phosphoryla- sis by adenovirus VA RNA,. Cell 44:391-400. tion of the eukaryotic initiation factor 2a subunit subsequent to 28. O'Malley, R. P., T. M. Mariano, J. Siekierka, W. C. Merrick, infection. Proc. Natl. Acad. Sci. USA 82:4321-4325. P. A. Reichel, and M. B. Mathews. 1986. The control of protein 40. Schneider, R. J., and T. Shenk. 1987. Impact of virus infection synthesis by adenovirus VA RNA, p. 291-301. In Cancer cells on host cell protein synthesis. Annu. Rev. Biochem. 56:317- 4: DNA tumor viruses. Cold Spring Harbor Laboratory, Cold 332. Spring Harbor, N.Y. 41. Sen, G. C., H. Taira, and P. Lengyel. 1978. Interferon, double 29. Paez, E., and M. Esteban.1984. Resistance of vaccinia virus to stranded RNA, and protein phosphorylation: characteristics of interferon is related to an interference phenomenon between the a double-stranded RNA-activated protein kinase system par- virus and the interferon system. Virology 134:12-28. tially purified from interferon-treated Ehrlich ascites tumor 30. Pain, V. M. 1986. Initiation of protein synthesis in mammalian cells. J. Biol. Chem. 253:5915-5921. cells. Biochem. J. 235:625-637. 42. Siekierka, J., T. M. Mariano, P. A. Reichel, and M. B. Mathews. 31. Petryshyn, R., J.-J. Chen, and I. M. London. 1988. Detection of 1985. Translational control by adenovirus: lack of virus associ- activated double-stranded RNA-dependent protein kinase in ated RNA, during adenovirus infection results in phosphoryla- 3T3-F442A cells. Proc. Natl. Acad. Sci. USA 85:1427-1431. tion of initiation factor eIF-2 and inhibition of protein synthesis. 32. Petryshyn, R., D. H. Levin, and I. M. London. 1980. Purification Proc. Natl. Acad. Sci. USA 82:1959-1963. and characterization of a latent precursor of a double-stranded 43. Whitaker-Dowling, P., and J. S. Youngner. 1984. Characteriza- RNA dependent protein kinase from reticulocyte lysates. Bio- tion of a specific kinase inhibitory factor produced by vaccinia chem. Biophys. Res. Commun. 94:1190-1198. virus which inhibits the interferon-induced protein kinase. Vi- 33. Petryshyn, R., D. H. Levin, and I. M. London. 1983. Double- rology 137:171-181. stranded RNA-dependent eIF-2a protein kinase. Methods En- 44. Zilberstein, A., P. Federman, L. Shulman, and M. Revel. 1976. zymol. 99:346-362. Specific phosphorylation in vitro of a protein associated with 34. Ranu, R. S. 1980. Regulation of protein synthesis in rabbit ribosomes of interferon-treated mouse L cells. FEBS Lett. reticulocyte lysates: purification and initial characterization of 68:119-124. the double-stranded RNA activated protein kinase. Biochem. 45. Zilberstein, A., A. Kimchi, A. Schmidt, and M. Revel. 1978. Biophys. Res. Commun. 97:252-262. Isolation of two interferon-induced translational inhibitors: a 35. Reichel, P. A., W. C. Merrick, J. Siekierka, and M. B. Mathews. protein kinase and an oligo-isoadenylate synthetase. Proc. Natl. 1985. Regulation of a protein synthesis initiation factor by Acad. Sci. USA 75:4734-4738.