Intrinsic Molecular Activities of the Interferon-induced 2-5A-dependent RNase*

2-5A-dependent RNase (RNase L), a unique endoribo- nuclease that requires 5’-phosphorylated 2‘,5’-linked oligoadenylates (2-5A), functions in the molecular mechanism of interferon action. Because this enzyme is present at very low levels in nature, characterization and analysis have been limited. The molecular cloning of human, 2-5A-dependent RNase cDNA has facilitated its expression to high levels in insect cells by infecting with recombinant baculovirus. To determine the prop- erties of the enzyme in the absence of other proteins, the recombinant 2-5A-dependent RNase was purified to ho- mogeneity. The purified enzyme migrated as a monomer upon gel filtration in the absence of activator and showed highly specific, 24A-dependent RNase activity. The precise activator requirements were determined by stimulating the purified enzyme with a variety of 2’,5‘- linked oligonucleotides. The activated enzyme was capable of cleaving poly(rU) and, to a lesser extent, poly(rA), to sets of discrete products ranging from between 4 and 22 nucleotides in length. Reduced rates of 2-5A-dependent RNA cleavage were observed even after removal of ATP and chelation of divalent cations. How-ever, optimal RNA

The requirement of 2-5A-dependent RNase (RNase L) for an activator makes it unique among the family of known ribonucleases (reviewed in Belasco and Brawerman (1993), Deutscher (19931, and Lengyel (1993)). The RNase activators consist of a series of unusual 5'-triphosphorylated 2',5'-linked oligoadenylates collectively referred to as "2-5A" (Kerr and Brown, 1978). Furthermore, 2-5A-dependent RNase is directly implicated in the molecular mechanism of interferon action, making it one of a relatively few mammalian RNases with a known biological function . The discovery by Clemens and Williams (1978) of ribonuclease activity that was dependent on the addition of 2-5A, followed earlier observations involving extracts from interferon-treated cells. Studies showed that protein synthesis in these extracts was very sensitive to inhibition by double-stranded RNA (Kerr et al., 1974) and that addition of double-stranded RNA stimulated the breakdown of RNA (Brown et al., 1976;Kerr et al., 1976). The Service Grant 5R01 CA44059, awarded by the Department of Health * This investigation was supported by United States Public Health and Human Services, National Cancer Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ology, NN1-06, The Cleveland Clinic Foundation, 9500 Euclid h e . , 7 TO whom correspondence should be addressed: Dept. of Cancer  presence of double-stranded RNA indirectly led to the degradation of RNA by stimulating a group of interferon-inducible, 2-5A synthetases to produce 2-5A from ATP (Hovanessian et al., 1977;Kerr and Brown, 1978). In the final step of the pathway, the 2-5A activates 2-5A-dependent RNase resulting in the cleavage of single-stranded RNA 3' of UpUp and UpAp sequences (Wreschner et al., 1981;Floyd-Smith et al., 1981). Therefore, 2-5A-dependent RNase is relatively nonspecific with respect to substrates. However, a localized activation of 2-5A-dependent RNase has been shown to occur in cell extracts when single-stranded RNA was linked with double-stranded RNA (Nilsen and Baglioni, 1979). In addition, a method which directs 2-5A-dependent RNase to specific RNA sequences has been described . Although there are basal levels of 2-5A-dependent RNase in most mammalian cells, induction occurs during interferon treatment, cell differentiation, and cell growth arrest Jacobsen et al., 1983a, 198313;Krause et al., 1985). Nevertheless, 2-5A-dependent RNase is a low abundance protein comprising only one five-hundred thousandth of the protein in mouse liver (Silverman et a l . , 1988). The human gene for 2-5A-dependent RNase (RNS4) has been recently mapped to chromosome lq25 (Squire et al., 1994).
Despite the fact that the activity of 2-5A-dependent RNase has been previously studied (for instance, in Slattery et al. (1979), Wreschner et al. (19821, Silverman et al. (19881, and Bisbal et al., 1989) much remains to be learned about the structure and intrinsic activities of the enzyme. The recent cloning of the complete human 2-5A-dependent RNase provides the means to directly address questions of its structure, function, and biological significance (Zhou et a l . , 1993). For instance, stable expression of a dominant-negative mutant of 2-5A-dependent RNase in a mouse cell line suppresses both the antiencephalomyocarditis virus and anti-proliferative activities of interferon . Analysis and mutagenesis of murine and human 2-5A-dependent RNases revealed several interesting features Zhou et al., 1993). Notably, the 2-5A-binding domain was localized by deletion analysis and site-directed mutagenesis to a repeated phosphate binding loop motif. In addition, 2-5A-dependent RNase has homology with protein kinases and Escherichia coli RNase E and contains a cysteine-rich region and nine ankyrin-like repeats implicated in mediating protein-protein interactions. To study the molecular basis for the 2-5A-mediated cleavage of RNA, we have expressed high levels of human 2-SA-dependent RNase in insect cells. Analysis of the homogeneous, recombinant enzyme provided novel information on its 2-5Arequirementq substrate specificity, and stimulation by divalent cations and ATP.
Baculouirus Expression of 25A-dependent RNase-The cDNA encoding the entire coding sequence to the human form of 2-5A-dependent RNase, a Hind111 fragment of plasmid ZC5  was cloned into the BamHI site of the transfer plasmid pBacPAKl (Clontech) after filling-in the termini using Klenow fragment. Clones containing the cDNA in the correct orientation were determined by restriction enzyme analysis. The recombinant pBacPAKl/ZC5 DNA (500 ng) was cotransfected into SF21 cells with 200 ng of Bsu36l-digested BacPAKG viral DNA using the Lipofectin reagent (Life Technologies Inc.). Plaques containing recombinant virus were identified by Southem blot analysis of polymerase chain reaction products obtained with Bacl and Bac2 primers (Clontech) probed with a 300-base pair fragment of SacI-digested ZC5 DNA.
To produce recombinant 2-5A-dependent RNase, either monolayer or suspension (for large scale) cultures of SF21 cells were infected at a multiplicity of infection of 10 plaque-forming unitdcell at 27 "C for 3 days before harvesting. The cell pellets obtained aRer washing in phosphatebuffered saline (pH 6.2), were frozen on dry ice and stored at -70 "C.
Preparation of Crude Cell Extracts-To prepare cell extract, four packed cell volumes of buffer A (25 m M Tris-HC1, pH 7.4, 50 m M KCl, 10% glycerol, 1 m M EDTA, 0.1 m M ATP, 5 II~M MgCl,, 14 m M 2-mercaptoethanol, and 1 pg/ml of leupeptin) were added to cell pellets. The cell suspensions were sonicated on ice six times for 15 s at 30-s intervals. Supernatants were collected after centrifuging at 4 "C for three times at 16,700 x g (once for 30 min and then twice more for 10 min each time).
Purification of24A-dependent RNase-Chromatography used in the purification of the 2-5A-dependent RNase was performed with a fast protein liquid chromatography system (Pharmacia LKB Biotechnology Inc.). All purification procedures were at 4 "C.
Blue Sepharose CL-GB Chromatography-Crude cell extract containing about 20 mg of proteidpreparation in 2 ml of buffer A was loaded onto a CL-GB blue Sepharose column (5 x 50 mm, Pharmacia). After washing with 10-column volumes of buffer A at a flow rate of 0.3 mllmin, a linear gradient to 21% buffer B (buffer A supplemented with 1 M KC1) was performed in about 23 min. The ratio of buffer Abuffer B was then held constant while the 2-5A-dependent RNase eluted (in about 8 ml). The column fractions were monitored for 2-5A binding activity and then pooled. The protein was concentrated and desalted with a Centricon filter unit (Amicon).
Mono Q Chromatography-The peak of 2-5A-dependent RNase from the blue Sepharose column (about 2-2.5 mg of protein in 1 ml of buffer Npreparation) was loaded on a Mono Q (HR 515) column (5 x 50 mm, Pharmacia) at a flow rate of 0.4 mllmin. After washing with 10 column volumes of buffer A, the 2-5A-dependent RNase was eluted in a linear gradient to 40% buffer B in 50 min at 0.4 mllmin. The peak of 2-5Adependent RNase (as determined by 2-5A binding assay) was observed after 7.5-15 min, corresponding to a KC1 concentration of about 120-180 mM.
Superose-12 Chromatography-Between 200 and 300 pg of 2-5Adependent RNase per separation, obtained after purification by the previous two steps, in 100 p1 of buffer C (buffer A containing 100 II~M KCl) was loaded on a Superose-12 (HR 10/30) column (10 x 300 mm) at a flow rate of 0.2 mllmin. Calibration of the column was with 100 pg each of the following proteins: p-amylase (200 kDa), human transferrin (75.2 kDa), carbonic anhydrase (29 kDa), and cytochrome c (12.4 kDa) all from Sigma.
Dialysis of 24A-dependent RNase to Remove Divalent Cations-2-5A-dependent RNase (about 100 pgin 100 pl), post-Mono Q column frac-tion, was dialyzed at 4 "C against 25 m M Tris-HC1 (pH 7.4),2.5 nm EDTA, 2.5 m M EGTA, 14 m M 2-mercaptoethanol, and 100 m M KCl; once with 400 ml for 4 h, then with 600 ml for 10 h, and finally with 200 ml of reaction buffer (EDTA and EGTA were reduced to 0.5 m M each) for another 4 h.
Assays for 2-5A Binding Activities-2-5A binding activity was measured by modifications (Silverman and Krause, 1987;Nolan-Sorden et al., 1990) of the filter method of Knight et al. (1980). A 32P-labeled and bromine-substituted 2-5A analog, p(A2'p),(b@A2'p)&3'-[3zPlCp, about 10,000 countatmidassay, at about 3,000 Ciimmol, was incubated with fractions containing 2-SA-dependent RNase (or controls) on ice for 1 h. The reaction mixtures were then transferred to nitrocellulose filters which were washed twice in distilled water and dried and the amount of 2-5A probe bound to the 2-5A-dependent RNase on the filters was measured by scintillation counting (Silverman and Krause, 1987). Covalent, cross-linking of 2-5A-dependent RNase to the same 2-5A probe (about 60,000 counts/midassay) under ultraviolet was according to Nolan-Sorden et al. (1990). Cell extracts or purified 2-5A-dependent RNase was incubated with probe for 1 h on ice and cross-linking was for an additional 1 h on ice under ultraviolet light at 308 nm. Protein separation was by SDS/10% polyacrylamide gel electrophoresis and was followed by autoradiography. Quantitation of 2-5A binding activity was by PhosphorImage analysis (Molecular Dynamics) of the dried gels.
25A-dependent RNase Activity Assays-RNA molecules used were polyfrU), poly(rA), poly(rG), or poly(rC) (Pharmacia) and were labeled at their 3' termini with [32PlpCp (3,000 Ciimmol) and T4 RNA ligase as described previously (Silverman, 1985). Poly(dT) and poly(dA) (Pharmacia) were labeled at the 5' termini with T4 polynucleotide kinase and [y-32P]ATP. Fractions containing 2-5A-dependent RNase were incubated in the presence or absence of 100 m pppA(2'p5'A), (or other oligonucleotides), and 8-60 m 32P-labeled RNAin final volumes of 25 pl at 30 "C. The trichloroacetic acid-insoluble fractions of RNA were determined by filtering on glass-fiber filters in the presence of camer yeast tRNA by scintillation counting as described previously (Silverman, 1985). One unit of 2-5A-dependent RNase is defined as the activity required to degrade 50% of 12 m radiolabeled poly(U) to acid-soluble fragments in the presence of 100 m p3(A2'p),A at 30 "C for 30 min. For gel analysis of RNA cleavage products, 8 p1 of reaction mixtures were boiled in the presence of gel sample buffer and then applied to 8% polyacrylamide, 8 M urea sequencing gels.

Baculovirus Expression of Human 2-5A-dependent RNase-
To further study the structure, function, and properties of 2-5A-dependent RNase it was necessary to obtain greater levels of the enzyme than have been previously possible using mammalian cells and tissues as the sources. Therefore, the cDNA to the human form of the endoribonuclease was subcloned in the baculovirus vector, BacPAK6 (Clontech), under the control of the polyhedrin promoter ("Materials and Methods"). Production of the enzyme in SF21 insect cells infected by the recombinant virus was measured by both 2-5A binding and 2-5A-dependent ribonuclease assays. 2-5Abindingactivity was determined by covalent cross-linking of a bromine-substituted, 32P-labeled 2-5A analog to the RNase under ultraviolet light (Nolan-Sorden et al., 1990). Extract of insect cells infected with the non-recombinant virus showed no detectable 2-5A-dependent RNase by this sensitive assay (Fig. IA, lane 1). In contrast, an intense 80-kDa 2-5A binding activity corresponding to the 2-5A-dependent RNase was detected in extract of insect cells infected with the recombinant virus (Fig. IA, lane 2 ) .
To measure the catalytic activity of the recombinant enzyme, ribonuclease assays were performed in the presence and absence of trimer 2-5A, i.e. pppA(2'pS'A),. The degradation of radiolabeled poly(rU) to acid-soluble fragments was measured in these assays (Silverman, 1985). Crude extract of the insect cells infected with non-recombinant virus had no 2-5A-dependent RNase activity (Fig. 1B). These findings are consistent with a previous study in which 2-SA-dependent RNase was shown to be absent in insect cells . On the other hand, the RNA was extensively degraded (93%) in extract of the Cp for 30 min a t 30 "C in the presence or absence of 100 n~ p,A(2'p5'A), as activator. RNA degradation was determined by measunng the acid-insoluble fractions of RNA as described (Silverman, 1985). enzyme is a fully functional, 2-5A-dependent RNase. It is in its "off-state" in the absence of 2-5A and in its "on-state'' in the presence of 2-5A.
Purification of Recombinant, Human 2SA-dependent RNase-To determine the intrinsic properties of 2-5A-dependent RNase, purification of the recombinant enzyme was performed using a fast protein liquid chromatography system (Pharmacia). Three separation steps were used to obtain apparently pure enzyme (Table I and "Materials and Methods"). The expressed 2-5A-dependent RNase is clearly visible in stained gels as the major protein present in crude extract of the recombinant virus-infected cells (Fig. 2, lane 2). There was no band visible at the corresponding position in the lane containing extract from non-recombinant virus-infected SF21 cells (lane 1). The enzyme was 46 and 85% purified after the first (blue Sepharose) and second (Mono Q ) separation steps, respectively (Table I and Fig. 2, lanes 3 and 4). Only a single band of protein was observed after Superose-12 chromatography (Fig.  2, lane 5). Therefore, the final Superose-12 fraction consisted of apparently homogeneous 2-SA-dependent RNase. The level of 2-5A-dependent RNase in the insect cells is determined to be 6.7% of the total soluble protein fraction on the basis of the purification data (Table I)   ists as a multimer, its migration through the Superose-12 column was compared with the elution volumes of marker proteins (Fig. 3). The elution volume of the 2-5A-dependent RNase, monitored by absorbance, 2-5A binding activity, and 2-5A-dependent RNase activity (Fig. 3), all correspond to that of a single chain of 83.5 kDa molecular mass, determined from the predicted amino acid sequence . From these data we conclude that the 2-5A-dependent RNase less activator (2-5A) exists as a monomer.
To determine the lengths of the cleavage products, the RNA was analyzed on denaturing polyacrylamide gels (Fig. 5 A ) . Interestingly, both the poly(rU) and poly(rA) were cleaved into sets of discrete products. The fragments of the poly(rU) were estimated to be 5, 7, 8, 12, 16, and 22 nucleotides in length. These were apparent after addition of 1 nM or higher concentrations of 2-5A (lanes 5-81, Degradation of poly(rA) was seen only with about 10-100-fold higher levels of 2-5A than were required to degrade poly(rU) (Fig. 5A, compare lanes 5 and 15). Specific cleavage products of poly(rA) were also seen, although to a lesser extent than with poly(rU). The poly(rA) fragments were 13,21, and 22 nucleotides in length. Identical patterns of poly(rU) cleavage were observed with trimer, tetramer, or pentamer 2-5A, whereas the dimer species was without activity (Fig. 5B).
A similar discrete pattern of poly(rU) breakdown products was observed using a crude preparation of naturally occurring mouse L cell 2-5A-dependent RNase (data not shown).
Optimal Activation of 25A-dependent RNase Requires Either Manganese or Magnesium and ATP-The effects of divalent cations and ATP on 2-SA-dependent RNase activity were determined afier extensive dialysis of 2-5A-dependent RNase against buffer lacking ATP and containing 2.5 m~ each of EDTA and EGTA ("Materials and Methods"). Ribonuclease assays were also performed in the presence of EDTA and EGTA (at 0.5 mM each) to ensure the continued chelation of metal ions. Interestingly, 2-SA-dependent RNase cleaved poly(rU) upon addition of 100 nM 2-5A even in the absence of divalent cations and ATP (Fig. SA, lanes 4). The basal activity is not due t o contaminating divalent cation or ATP in the 2-5A because reactions were performed in the presence of a large excess of EDTA and EGTA and analysis of 2-5A by high performance liquid chromatography showed an absence of detectable levels of ATP (data not shown). Furthermore, when added individually, ATP, magnesium, or manganese had no effect on 2-5Adependent RNase activity (Figs. S A , lanes 6, 8, and 12). However, the combinations of either ATP plus magnesium or ATP plus manganese greatly stimulated 2-5A-dependent RNase activity ( Fig. S A , lanes 10 and 14). Calcium added alone inhibited the RNase while calcium plus ATP restored basal activity (lanes 16 and 18). Finally, zinc was inhibitory to the RNase even in the presence of ATP (lanes 20 and 22).
To determine effects of divalent cations and ATP on the affinity of the enzyme for 2-5A, 2-5A-binding assays were per-C. formed (Fig. 6B). Addition of each of the divalent cations clearly enhanced 2-5A binding activity. In this regard, magnesium or zinc enhanced 2-5A binding greater than either manganese or calcium. Therefore, although zinc strongly inhibits ribonuclease activity it nevertheless enhances 2-5A binding. Addition of ATP caused a modest (12-39%) increase in 2-5A binding activity (Fig. 6B). Effects of ATP were generally greater in the presence of divalent cations (19-39%) than in their absence (12%). These findings establish that there is a direct stimulatory effect of ATP on 2-5A-dependent RNase activity that requires either magnesium or manganese (see "Discussion").

DISCUSSION
To provide new insights into the functions and properties of 2-5A-dependent RNase, we have expressed the human form of the endoribonuclease in insect cells. The baculovirus system provided a high level of expression in a cell type containing no endogenous 2-5A-dependent RNase (Fig. 1). In addition, the 2-5A-dependent RNase which was produced is soluble and fully functional with respect to both 2-5A binding and catalytic activities. In the absence of its activator, 2-5A, the recombinant 2-SA-dependent RNase eluted from a gel filtration column as a monomer which had full activity (Fig. 3). Recently it was proposed that 2-5A-dependent RNase is a heterodimer of 2-5A binding and catalytic subunits ; however, the present work shows that both the 2-5A binding and catalytic domains are encoded in a single cDNA expressing just one polypeptide.
Activator Requirements of 24A-dependent RNase-In accord with previous reports, optimal activation of the human form of 2-SA-dependent RNase required one 5'-phosphoryl group linked to a t least three 2',5'-linked adenylyl residues (Figs. 4 and 5) (Torrence et al. (1988), Kitade et al. (1991), Kovacs et al. (1993), and references therein). For instance, absence of a 5'phosphoryl group greatly reduced ribonuclease activity ( Intrinsic Activities of 2-SA-dependent R N a s e 4B). The phosphodiester linkage isomer of 2-5A trimer, 3-5A (ppp5'A(3'p5'A),), was without activity as an activator of the 2-5A-dependent RNase (Fig. 4C), thereby confirming the earlier results (Lesiak et al., 1983;Krause et al., 1986). The enzymes responsible for the biological synthesis of 2-5A, the 2-5A synthetases, can add AMP in 2',5' linkage to important metabolite/regulator entities such as adenosine tetraphosphate (Ap,A)' (Ball, 1980;Ferbus et al., 1981;Cayley and Kerr, 1982). The results of Fig. 4C demonstrated that the "capped" 2-5A derivative, A5'pppp5'A(2'p5'A),, was about equipotent with unmodified 2-5A trimer triphosphate as an RNase activator. Earlier studies Imai and Torrence, 1984;Krause et al., 1986) revealed that this unsymmetrical tetraphosphate could inhibit translation in extracts of mouse L cells and Ehrlich ascites tumor cells, and could activate partially purified 2-5A-dependent RNase from mouse L cells. This result with the cloned and purified human enzyme showed conclusively that the adenosine capped tetraphosphate itself was able to activate the 2-SA-dependent endonuclease, since there was, under these conditions, no possibility of degradation to yield AMP and ppp5'A(2'p5'A),. The critical role of the purine amino groups of the adenine bases of 2-5A has been suggested by earlier studies employing crude extracts of mouse and human cells . The result of Fig. 4C, showing that the inosine-substituted analog ppp5'1(2'p5'A), was more than 10-fold less effective as an activator of the 2-5A-dependent endonuclease was fully consistent with this earlier finding.

Selection of Cleavage Sites by 23A-dependent RNase-
Previous studies on the sequence specificity of 2-5A-dependent RNase indicated that cleavage occurred after UpNp sequences in natural RNAs and within poly(rU) but not in poly(rA), poly(rG), or poly(&) (Wreschner et al., 1981;Floyd-Smith et al., 1981). Here we show that poly(rU) and to a lesser extent, poly(rA) are substrates for 2-5A-dependent RNase. Thus, the 2-5A-dependent RNase is clearly capable of cleaving after adenylyl residues in RNA even when these are not preceded by uridylyl residues. Interestingly, degradation of poly(rU) and poly(rA) produced cleavage products of discrete sizes, ranging from 5 to 22 nucleotides in lengths (Figs. 5 and 6A). Although the reason for this product size distribution is unknown, the discrete cleavage products may be unable to bind to the active site of 2-5A-dependent RNase. Therefore, rejection of the discrete RNA fragments by the enzyme presumably reflects the spatial relationship between the substrate-binding site and the catalytic domain.

Effects of Divalent Cations and ATP on 24A-dependent
RNase-The 2-5A-dependent RNase is capable of cleaving RNA in response to 2-5A even after chelation of divalent cations (Fig. 6A). These findings are consistent with reports that other ribonucleases which leave 3'-phosphoryl groups also do not require divalent cations (reviewed in Deutscher (1993)). However, magnesium and manganese stimulate 2-5A-dependent RNase but only in the presence of ATP. It would appear, therefore, that ATP and 2-5A complexed with magnesium or manganese have enhanced affinities for the enzyme (Fig. 6B).
ity when ATP or the ATP analogs, AMP-CPP or AMP-PCP, were present. These results suggest that hydrolysis of ATP is not required to stimulate ribonuclease activity. Results shown here clearly demonstrate that ATP directly stimulates 2-5A-dependent RNase activity (Fig. 6). An apparent ATP-binding domain in common with protein kinases was previously observed in the predicted amino acid sequence of 2-SA-dependent RNase . Perhaps, therefore, binding of ATP to the enzyme acts to stabilizes its activated form.