Identification of the Major Phosphorylation Site of the Hepatitis C Virus H Strain NS5A Protein as Serine 2321*

The hepatitis C virus (HCV) NS5A protein is phosphorylated by a cellular, serine/threonine kinase. To identify the major site(s) of NS5A phosphorylation, radiolabeled HCV-H NS5A phosphopeptides were purified and subjected to phosphoamino acid analysis and Edman degradation. These data identified the major intracellular phosphorylation site in the HCV-H NS5A protein as Ser2321, a result verified by two additional, independent methods: (i) substitution of Ala for Ser2321and the concomitant disappearance of the major in vivophosphorylated peptides and corresponding in vitrophosphorylated peptides; and (ii) comigration of the digestion products of a synthetic peptide phosphorylated on Ser2321 with the major in vivo phosphorylated NS5A peptides. Site-directed mutagenesis of Ser2321 suggested that phosphorylation of NS5A is dispensable for previously described interactions with NS4A and PKR, a cellular, antiviral kinase that does not appear to catalyze NS5A phosphorylation. The proline-rich nature of the amino acid sequence flanking Ser2321 (PLPPPRS2321 PPVPPPR) suggests that a proline-directed kinase is responsible for the majority of HCV NS5A phosphorylation, consistent with previous kinase inhibitor studies.

The hepatitis C virus (HCV) NS5A protein is phosphorylated by a cellular, serine/threonine kinase. To identify the major site(s) of NS5A phosphorylation, radiolabeled HCV-H NS5A phosphopeptides were purified and subjected to phosphoamino acid analysis and Edman degradation. These data identified the major intracellular phosphorylation site in the HCV-H NS5A protein as Ser 2321 , a result verified by two additional, independent methods: (i) substitution of Ala for Ser 2321 and the concomitant disappearance of the major in vivo phosphorylated peptides and corresponding in vitro phosphorylated peptides; and (ii) comigration of the digestion products of a synthetic peptide phosphorylated on Ser 2321 with the major in vivo phosphorylated NS5A peptides. Site-directed mutagenesis of Ser 2321 suggested that phosphorylation of NS5A is dispensable for previously described interactions with NS4A and PKR, a cellular, antiviral kinase that does not appear to catalyze NS5A phosphorylation. The proline-rich nature of the amino acid sequence flanking Ser 2321 (PLPPPRS 2321 PPVPPPR) suggests that a proline-directed kinase is responsible for the majority of HCV NS5A phosphorylation, consistent with previous kinase inhibitor studies.
The hepatitis C virus (HCV) 1 nonstructural protein NS5A is phosphorylated primarily on serine residues (1, 2) by an unidentified cellular kinase. At least two forms of phosphorylated NS5A have been detected: p56 and p58, which are distinguishable by a slight difference in their electrophoretic mobilities. For the HCV-J isolate, it has been shown that production of p58 is stimulated by NS4A (3), apparently through a direct interaction (4). Deletion of the N-terminal region of NS5A necessary for this interaction with NS4A results in a similar increase in the level of p58 (3,5), suggesting that this domain may suppress phosphorylation at certain sites in the absence of NS4A. However, the role of NS4A in p58 production has yet to be confirmed in other HCV isolates.
Although NS5A has also been shown to interact with the cellular, antiviral kinase PKR (6), the identity of the kinase responsible for NS5A phosphorylation remains unclear, since phosphorylation of NS5A is apparently not catalyzed by purified PKR in vitro (6) or stimulated by PKR activators such as double-stranded RNA and heparin (5). Determination of the substrate preferences of the kinase responsible for NS5A phosphorylation could provide a clue to its identity, but information regarding NS5A phosphorylation sites is limited. Serial deletion studies have mapped the sites of NS5A phosphorylation to a central, conserved region between amino acids 2200 and 2250 and a variable region downstream of amino acid 2350 in the HCV-J isolate (3). Site-directed mutagenesis of serines in the region from amino acid 2200 to 2250 has identified Ser 2197 , Ser 2201 , and Ser 2204 as putative sites of p58 phosphorylation; however, these mutations have no noticeable effect on the level of p56 phosphorylation (3), consistent with the idea that NS5A contains additional phosphorylation sites. Moreover, since even point mutations have the potential to disrupt some structural feature of NS5A necessary for its phosphorylation, the negative effects of these mutations on NS5A phosphorylation cannot be conclusively attributed to the ablation of phosphate acceptor residues.
To identify the major NS5A phosphorylation sites unequivocally, total phosphorylated NS5A was digested with trypsin and chymotrypsin, and phosphopeptides that contained the majority of incorporated 32 P when NS5A was phosphorylated in vivo were analyzed by Edman degradation. Ser 2321 was identified as the phosphate acceptor in two of these phosphopeptides, which apparently resulted from partial trypsin cleavage after an arginine in an unfavorable amino acid context. The identification of Ser 2321 as the preferred site for NS5A phosphorylation in vivo was confirmed by the disappearance of these phosphopeptides as a result of an alanine substitution at position 2321 and their comigration with the digestion products of a phosphorylated, 16-amino acid synthetic peptide based on Ser 2321 and its flanking sequence (NH 2 -GCPLPPPRS 2321 PPVPPPR-COOH). This information was subsequently used to examine the effect of phosphorylation at Ser 2321 on the ability of HCV NS5A to interact with NS4A and PKR.
Expression and Isolation of Phosphorylated HCV NS5A or the Glutathione S-transferase (GST)-HCV NS5A Fusion Protein-For small scale expression of HCV NS5A or GST-HCV NS5A, baby hamster kidney (BHK-21) cells were transfected with the corresponding pTM3 (7) construct using the vaccinia virus/T7 hybrid transient expression system (8). HCV NS5A was labeled with [ 32 P]orthophosphate (ICN, Costa Mesa, CA) and immunoprecipitated with the NS5A-specific antibody WU123 (9) and protein A-agarose (Sigma), whereas GST-HCV NS5A was isolated on glutathione-agarose and phosphorylated in vitro as described previously (2), with the following exception: the GST-HCV NS5A/glutathione-agarose complexes were incubated for 1 h at 37°C in 25 l of phosphatase buffer (50 mM Tris-Cl, pH 7.5, 1 mM MgCl 2 , 0.1 mM ZnCl 2 , 1 mM EDTA) with 1 l (20 units) of calf intestinal alkaline phosphatase (Roche Molecular Biochemicals) at 37°C to remove unlabeled phosphate groups that may have been added to the fusion protein inside the cell and washed three times with kinase wash buffer (KWB, 50 mM Tris-Cl, pH 7.5, 5 mM MnCl 2 ) to remove most of the phosphatase prior to the in vitro kinase reaction.
For expression of preparative amounts of GST-HCV NS5A to be used for phosphopeptide sequencing, 24 10-cm dishes of BHK-21 monolayers were coinfected with vTF7-3 (8) and a vaccinia virus recombinant expressing HCV NS5A 2 at a multiplicity of infection of 10 in 1.5 ml/dish of phosphate-buffered saline, 1% fetal bovine serum for 1 h at room temperature. After infection, the cells were washed once with Earles' minimal essential medium (MEM) and incubated in 10 ml of MEM, 5% fetal bovine serum for 10 -12 h at 37°C. For labeling with tritiated proline, the cells were washed once with Earles' MEM and incubated for another 4 h in 4 ml of MEM containing 3% dialyzed fetal bovine serum and 200 Ci/ml L-[2,3,4,5-3 H]proline (Amersham Pharmacia Biotech). After the expression and/or labeling periods, the cells were washed with cold phosphate-buffered saline, lysed in 2 ml/plate of NETN (120 mM NaCl, 1 mM EDTA, 50 mM Tris-Cl, pH 7.5, 0.5% Nonidet P-40) containing 5 mM dithiothreitol and the protease inhibitors aprotinin (1 g/ml), leupeptin (1 g/ml), and phenylmethylsulfonyl fluoride (20 g/ml), clarified by microcentrifugation at top speed for 15 min at 4°C, and isolated on glutathione-agarose (50 l of a 1:1 suspension per 500 l of lysate) as described previously (2). Once the nonspecifically bound proteins had been removed from the GST-HCV NS5A/glutathione-agarose complexes, they were treated with calf intestinal alkaline phosphatase and washed three times with KWB as described above, followed by one wash with KWB containing 5 mM dithiothreitol. Kinase reactions were performed using previously established conditions (2), except that the amount of [␥-32 P]ATP was doubled to 20 Ci, and they were terminated by the addition of 2ϫ protein sample buffer and incubation at 75°C.
Peptide Purification and Sequencing-The radiolabeled HCV NS5A and GST-HCV NS5A proteins were further isolated by SDS-8% PAGE, transfer to Immobilon P (Millipore, Bedford, MA), and excision of the appropriate radiolabeled bands, followed by digestion with chymotrypsin and trypsin and oxidation with performic acid as described previously (2). In vivo labeled phosphopeptides were purified by two-dimensional separation on thin layer cellulose (TLC) plates, followed by recovery of the phosphopeptide from the cellulose by elution with pH 1.9 buffer (2.2% formic acid, 7.8% glacial acetic acid) and lyophilization (10). In vitro labeled phosphopeptides were purified by high performance liquid chromatography (HPLC) on a C18 column in a 0 -60% acetonitrile gradient in 0.1% trifluoroacetic acid with a flow rate of 0.5 ml/min. Fractions were collected every 0.5 min, and the 32 P content of each fraction was determined by Cerenkov counting. Fractions containing the phosphopeptides of interest were lyophilized, dissolved in a small amount of pH 1.9 buffer, pooled, and separated in two dimensions on TLC plates; the purified phosphopeptides were then recovered as described above.
Phosphoamino acid analysis (10) was performed on the purified peptides to verify that they were phosphorylated on serine residues, and the position of the phosphoserine residues was determined by Cerenkov counting of 32 P released in consecutive cycles of manual Edman degradation. The position of tritiated proline residues was determined by automated sequencing using a gas-phase sequenator and liquid scintillation counting (Dave McCourt, Midwest Scientific, St. Louis).
Phosphocellulose Assay of Synthetic Peptide Phosphorylation-Portions of the MAPK reaction mixtures or supernatants of the GST-HCV NS5A-associated kinase reaction mixtures were spotted on 2.5-cm diameter phosphocellulose circles, washed four times for 5 min each in 0.5% H 3 PO 4 , rinsed briefly in ethanol, and air-dried; the quantity of bound 32 P was then determined by scintillation counting.
Anion-exchange Chromatography-Kinase reaction mixtures were diluted to a final volume of 250 l with 30% formic acid, microcentrifuged for 5 min to remove any precipitate, loaded on a 2-ml Dowex-1 column (formate form) equilibrated in 30% formic acid, and separated at a flow rate of 125 l/min. Sixteen 0.5-ml fractions were collected, and their 32 P contents were determined by Cerenkov counting. Peak fractions were pooled and concentrated by lyophilization.

Identification of the Major HCV-H NS5A Phosphorylation
Site as Ser 2321 by Edman Degradation-To obtain direct evidence for the location of HCV NS5A phosphorylation sites, total phosphorylated NS5A was digested with trypsin and chymotrypsin, and the resulting peptides were resolved by a combination of reverse-phase HPLC and two-dimensional separation on thin layer cellulose plates. Phosphorylated peptides were identified by 32 P labeling, and several of the most prominent ones were subjected to consecutive cycles of Edman degradation to determine the position of phosphorylated amino acids. Phosphopeptides a and b were chosen for this analysis because they appeared to contain sites that are preferentially phosphorylated in vivo (Fig. 1A). The region of NS5A corresponding to peptide k is less heavily phosphorylated, but this phosphopeptide was included in the analysis based on the apparent ease with which it was isolated. A few other phosphopeptides, including c, d, and e, were also assayed for 32 P release during Edman degradation, but the results were inconclusive, and they were not further analyzed.
For convenience, GST-HCV NS5A phosphorylated in vitro was used as the source of these phosphopeptides. However, their migration was compared with that of their in vivo phosphorylated counterparts to verify that they were representative of intracellular phosphorylation events (Fig. 1B). Phosphoamino acid analysis was also performed on the purified phosphopeptides to determine whether they were phosphorylated on serine or threonine residues. As shown in Fig. 2, all three of the peptides selected for purification were phosphorylated on serine. Edman degradation of these peptides showed that peptides a, b, and k contained phosphoserine at positions 1, 9, and 4, respectively. However, digestion with trypsin and chymotrypsin was predicted to create multiple peptides with phosphoserine at positions 1, 9, and 4; thus, additional information was required for the unambiguous identification of these peptides. Therefore, the experiment was repeated with GST-HCV NS5A metabolically labeled with [ 3 H]proline prior to its phosphorylation in vitro in the presence of [␥-32 P]ATP. As shown in Fig. 3, a combination of the 32 P peak at position 1 and 3 H peaks at positions 2, 3, 5, 6, and 7 uniquely identified peptide a as NH 2 -SPPVPPPR-COOH, with the phosphoserine at position 1 corresponding to Ser 2321 . Similarly, the 32 P peak at position 9, together with the 3 H peaks at positions 3, 5, 6, 7, 10, and 11, identified peptide b as NH 2 -GCPLPPPRSPPVPPPR-COOH, with the phosphoserine at position 9 also corresponding to Ser 2321 . The isolation of two phosphopeptides of different  (8), and NS5A was labeled in vivo with [ 32 P]orthophosphate, whereas GST-NS5A was labeled in vivo with [ 3 H]proline and in vitro with [␥-32 P]ATP. The radiolabeled NS5A and GST-NS5A proteins were isolated by immunoprecipitation or binding to glutathione-agarose, respectively, SDS-8% PAGE, transfer to Immobilon-P, and excision of the appropriate bands. The purified proteins were digested with trypsin and chymotrypsin and oxidized with performic acid, followed by two-dimensional separation of the resulting peptides on TLC plates and autoradiography in the presence of an intensifying screen. A, a typical phosphopeptide map of NS5A phosphorylated in vivo (modified from Ref. 2). B, three phosphopeptides purified from maps of NS5A phosphorylated in vivo or GST-NS5A phosphorylated in vitro and analyzed separately or as 1:1 mixtures in a second round of two-dimensional separation on TLC plates and autoradiography. In both (A and B), the cathode was on the left side of the TLC plate during the electrophoresis dimension, and the anode was on the right.  2. Phosphoamino acid analysis of purified HCV NS5A phosphopeptides a, b, and k. Portions of the same, purified, in vitro phosphorylated peptides used in the comigration experiment shown in Fig. 1B were hydrolyzed in 6 N HCl for 1 h at 110°C, lyophilized, dissolved in a small volume of pH 1.9 buffer, mixed with 2 g each of non-radiolabeled phosphoserine, phosphothreonine, and phosphotyrosine, spotted on TLC plates, and separated by electrophoresis in pH 1.9 buffer for 20 min at 1.5 kV in the first dimension and pH 3.5 buffer for 16 min at 1.3 kV in the second dimension (10). The plates were air-dried, sprayed with 0.2% ninhydrin in ethanol (Sigma), and heated for 10 min at 65°C to detect the non-radiolabeled phosphoamino acid standards. 32 P-Labeled phosphoamino acids were detected by autoradiography with an intensifying screen.
lengths that contain Ser 2321 may have resulted from partial inhibition of cleavage at Arg 2320 by nearby Pro residues or the negatively charged phosphoserine at P1Ј. The detection of 3 H at position 1 of peptide b was inconsistent with the presence of glycine at this position, but the perfect agreement between the 3 H profile and proposed amino acid sequence over the next 10 residues strongly suggested that the deduced sequence of peptide b was correct. The 3 H peak at position 1 may have been due to the elution of a small amount of undegraded peptide in the first cycle of Edman degradation, among other possibilities.
Effects of Site-directed Mutagenesis on the Presence of Phosphopeptides a, b, and k-The 32 P and 3 H profiles of peptide k suggested that it contained proline at position 3 and phosphoserine at position 4. Since the N terminus of peptide b was apparently produced by cleavage between His 2312 and Gly 2313 , an atypical cleavage site for trypsin and chymotrypsin, all serines immediately preceded by proline in the HCV NS5A amino acid sequence were considered to be potential phosphoacceptor residues, irrespective of their distance from predicted cleavage sites. Two such serines at positions 2197 and 2373 were excluded because they were preceded by additional prolines that should have produced 3 H signals prior to the one observed in cycle 3. To determine whether any of the remaining three serines (Ser 2118 , Ser 2179 , and Ser 2210 ) were phosphorylated on peptide k, they were replaced by alanine, and the resulting mutants were analyzed for the presence of peptide k. A double S2201A/S2204A mutant expected to inhibit p58 production was also analyzed in an attempt to identify the phosphopeptide(s) that might contain these phosphorylation sites.
As shown in Fig. 4, mutation of Ser 2118 , Ser 2179 , and Ser 2210 ,  Fig. 1, divided into two parts, and analyzed by successive cycles of Edman degradation performed manually or in a gas-phase sequenator. Each cycle of manual or automated Edman degradation was analyzed for the presence of 32 P or 3 H, respectively, to determine the positions of phosphoserine and proline residues. The detection of 32 P and 3 H by these methods was mutually exclusive, since (i) little if any 32 P is extracted into the gas-phase sequenator and (ii) the manual Edman degradation products were eluted in approximately 1 ml of trifluoroacetic acid, which quenches 3 H signals in liquid scintillant but allows 32 P detection by Cerenkov counting. as well as Ser 2201 /Ser 2204 had no apparent effect on phosphorylation of peptide k in vivo. Similar results were obtained for NS5A phosphorylated in vitro (data not shown). The identity of peptide k remains unknown, but the available data suggest that it may be a mixture of two or more peptides, one of which contains a phosphoserine at position 4 but no prolines and one which contains proline at position 3 but no phosphoserine. Based on the predicted cleavage sites of trypsin and chymotrypsin, the most likely candidates for the site of phosphorylation in peptide k are Ser 2146 , Ser 2158 , Ser 2356 , and Ser 2409 ; however, the atypical cleavage at the N terminus of peptide b raises the possibility that other peptides with phosphoserine at position 4 were produced by digestion with these enzymes. The phosphopeptide maps of the double S2201A/S2204A mutant looked very similar to those of wild-type HCV NS5A phosphorylated in vivo or the GST-HCV NS5A fusion protein phosphorylated in vitro. However, close inspection of Fig. 4 and a longer exposure of the same autoradiograph suggested that phosphopeptide d may be missing in maps of the double mutant, raising the possibility that it is phosphorylated at one or both of these sites. Also, the disappearance of peptides a and b in the maps of NS5A phosphorylated in vivo (Fig. 4) and in vitro (data not shown) that was observed as a result of the alanine substitution at Ser 2321 was consistent with the radiosequence data used to deduce their amino acid sequences and sites of phosphorylation.
Phosphorylation of a Synthetic Peptide Based on the Ser 2321 Phosphorylation Site-The identification of Ser 2321 as a major HCV NS5A phosphorylation site enabled the synthesis of a peptide corresponding to this site. Unlike GST-HCV NS5A expressed in BHK-21 cells, which is already partially phosphorylated and associated with one or more kinases when isolated on glutathione-agarose, such a peptide could be used to identify biochemical fractions capable of phosphorylating Ser 2321 and develop a scheme for purification of the responsible kinase activity(ies). A phosphorylated synthetic peptide could also be used to verify the identification of Ser 2321 as a major phosphorylation site in comigration experiments. Therefore, NH 2 -GC-PLPPPRSPPVPPPR-COOH, a 16-amino acid peptide corresponding to Ser 2321 and its flanking sequence, was synthesized (Quality Controlled Biochemicals, Hopkinton, MA) and analyzed for phosphorylation in vitro by the HCV NS5A-associated kinase. Reactions were performed in the presence or absence of peptide to determine the background level of 32 P binding and with GST or GST-HCV NS5A to determine the specificity of phosphorylation. Although some elevation in the level of 32 P bound to phosphocellulose was observed in the presence of GST-HCV NS5A peptide relative to GST-HCV NS5A alone or GST plus peptide, the ratio of specific peptide phosphorylation to nonspecific or background phosphorylation was fairly low under a wide variety of conditions (data not shown), suggesting that additional flanking sequences may be necessary for optimal phosphorylation of Ser 2321 by the NS5A-associated kinase.
However, since Ser 2321 and its surrounding residues constitute a consensus MAPK site (PXSP), this synthetic peptide was also tested as a substrate for purified p42 MAPK. Although most of the peptide remained unphosphorylated, an approximately 6 -8-fold increase in the level of phosphocellulose-bound 32 P was observed in reactions that contained peptide relative to those that did not (Table I). Furthermore, when the labeled components of the MAPK reaction were separated from the unincorporated [␥-32 P]ATP by anion-exchange chromatography and analyzed by two-dimensional separation on TLC, a prominent, positively charged spot was observed (Fig. 5). Furthermore, performic acid oxidation and trypsin/chymotrypsin digestion of the MAPK reaction products resulted in their conversion to forms that included two spots that comigrated with peptides a and b (Fig. 6). The comigration of the digestion products of the phosphorylated synthetic peptide, which, other than Ser 2321 , contains no hydroxylated amino acids that could accept a phosphate group, independently corroborates the finding that peptides a and b were derived from amino acids 2313-2328 of HCV NS5A and that phosphorylation of these peptides occurred on Ser 2321 . The identity of the third major spot(s) above the origin in Figs. 5 and 6 is not known but may correspond to autophosphorylated MAPK.
Significance of Phosphorylation at Ser 2321 for Interaction of HCV NS5A with NS4A or PKR-To investigate the possible link between NS5A phosphorylation and function, the effect of a single Ala substitution at Ser 2321 was analyzed on its previously reported interactions with NS4A (4) and the cellular serine/threonine kinase PKR (6). These interactions have been

TABLE I
Phosphorylation of myelin basic protein and synthetic peptides by p42 MAPK Counts/min incorporated in an in vitro kinase assay using purified MAPK without added substrate (none), with myelin basic protein (MBP), or with a synthetic NS5A peptide (NH 2 -GCPLPPPRSPPVP-PPR-COOH) containing the Ser 2321 acceptor site. Reactions (see "Experimental Procedures") contained 100 M ATP with ␥-32 P-labeled ATP to specific activities of either 100 Ci/mol (low 32 P) or 3.3 mCi/mol (high 32 P). speculated to be involved in replication of the HCV genome or the inactivation of host defense pathways, respectively. Therefore, BHK-21 cells were transfected with plasmids expressing GST, GST fused to wild-type NS5A, or GST fused to NS5A containing a serine to alanine mutation at position 2321 in the presence or absence of plasmids expressing HCV NS4A or human PKR. GST and the mutant or wild-type GST-HCV NS5A complexes were then isolated on glutathione-agarose, and the components were separated by SDS-PAGE. GST, GST-HCV NS5A, and HCV NS4A were detected by metabolic label-ing with [ 35 S]cysteine or -methionine, and PKR was detected by Western blotting with a polyclonal PKR antibody (11). Duplicate samples lysed in SDS were also analyzed by immunoprecipitation or Western blotting, respectively, to ensure that the transfected cells contained similar amounts of NS4A and PKR. As shown in Fig. 7A, NS4A interacted specifically with the wild-type and mutant GST-HCV NS5A proteins, with little or no apparent difference in its binding efficiency. Similar observations were made with respect to PKR (Fig. 7B, lanes 2 and 3), although a small amount of nonspecifically bound PKR was observed in the GST complex (lane 1) upon longer exposure. Clearly, phosphorylation at Ser 2321 is not necessary for NS5A binding to NS4A or PKR, nor does it seem to have an adverse effect on these interactions. However, the results do not exclude the possibility that the ability of NS5A to interact with NS4A or PKR is regulated by phosphorylation at one or more other sites.
FIG. 5. Phosphorylation of the synthetic peptide based on the Ser 2321 phosphorylation site by p42 MAPK. MAPK reactions with the synthetic peptide NH 2 -GCPLPPPRSPPVPPPR-COOH were performed as described under "Experimental Procedures," and unincorporated [␥-32 P]ATP was removed by anion-exchange chromatography on Dowex 1 in 30% formic acid. Peak 32 P-containing fractions of the flowthrough were lyophilized, dissolved in pH 1.9 buffer, and analyzed by two-dimensional separation on TLC plates, followed by autoradiography with an intensifying screen.
FIG. 6. Comigration of oxidized digestion products of the MAPK-phosphorylated synthetic peptide and GST-HCV NS5A phosphorylated by its associated kinase. Synthetic peptide that had been phosphorylated by p42 MAPK and separated from unincorporated [␥-32 P]ATP as described in the legend to Fig. 5 was digested with trypsin and chymotrypsin, oxidized with performic acid, lyophilized, and dissolved in pH 1.9 buffer. Equal counts of the resulting products and phosphopeptides a and b, prepared as described in the legend to Fig. 1, were spotted separately or together on TLC plates as indicated and analyzed by two-dimensional separation and autoradiography with an intensifying screen.  1-7, and B, lanes 1-3 and 7-9) or lysis in a buffer containing 0.5% SDS (A, lanes 8 -11, and B, lanes 4 -6). These samples were analyzed by SDS-8% PAGE and autoradiography (A and lanes 7-9 of B) or Western blotting with a polyclonal antibody to human PKR (11) (B, lanes 1-6). The samples in lanes 8 -11 of A and lanes 4 -6 of B were derived from approximately 5 or 500 times fewer cells, respectively, than the rest of the samples.

DISCUSSION
As described above, Ser 2321 was identified as the major phosphorylation site in the consensus HCV-H NS5A protein, based on N-terminal amino acid sequencing of two alternatively digested, purified phosphopeptides; site-directed mutagenesis; and comigration of the digestion products of a phosphorylated, synthetic peptide based on Ser 2321 and its surrounding sequences with peptides derived from a full-length, phosphorylated NS5A protein. The C-terminal region of NS5A varies considerably among HCV isolates, and Ser 2321 , which is located in this region, is not conserved among other HCV genotypes or subtypes (Fig. 8). However, all of the known HCV isolates have an unusually high percentage of serine, threonine, and proline residues in this region (Ͼ33% of the C-terminal 229 amino acids of HCV-H NS5A), suggesting that analogous sites may be present in these isolates at different positions. This suggestion is consistent with the hypothesis that the major HCV-J NS5A phosphorylation site is located downstream of amino acid 2350 (3).
Although Ser 2321 is part of a consensus recognition motif for MAPK, with proline at the Ϫ2 and ϩ1 positions (12), and a synthetic peptide with this sequence was phosphorylated by purified p42 MAPK, definitive evidence for intracellular HCV NS5A phosphorylation by MAPK is lacking. Furthermore, some properties of the NS5A-associated in vitro kinase activity are inconsistent with those of MAPK, such as its weak phosphorylation of myelin basic protein (5) and its unusual and dramatic preference for Mn 2ϩ relative to Mg 2ϩ (2). In addition, a MAPK-specific antibody that recognizes 42-, 44-, 56-, and 85-kDa isoforms of MAPK from human, rat, and mouse cells (Transduction Laboratories, Lexington, KY) failed to detect any endogenous MAPK in phosphorylation-competent GST-HCV NS5A complexes from BHK-21 cells (5). However, MAPK belongs to a large group of kinases for which proline at the ϩ1 position is an important specificity determinant. Other well known kinases in this group include the Jun N-terminal kinases (JNKs) (13) and the cyclin-dependent kinases (CDKs) (14); however, none of these kinases are known to have a preference for Mn 2ϩ . Moreover, histone H1, which is a good substrate for CDKs in vitro (15), is phosphorylated by the NS5A-associated kinase quite poorly in relation to histones H2A, H2B, and H3 (5). Interestingly, both the JNKs (16) and the HCV NS5A-associated kinase (5) exhibit weak myelin basic protein phosphorylation in vitro, but further studies are needed to determine the significance of this observation. Although JNKs and CDKs do not have a marked preference for proline at the Ϫ2 position, as present in the Ser 2321 phosphorylation motif, the presence of proline at this position does not seem to have a deleterious effect on their activity.
Glycogen synthase kinase 3 also appears to phosphorylate serines or threonines based on their proximity to proline residues, but in contrast to MAPK, JNK, and CDK, these prolines do not have to be located at the ϩ1 position (14). However, glycogen synthase kinase 3 is similar to these kinases in its apparent preference for Mg 2ϩ . Although a marked predilection for Mn 2ϩ has been reported for Mst3, a mammalian homologue of the yeast Ste20 kinase (17), its substrate specificity is likely to be inconsistent with Ser 2321 phosphorylation, since the optimal substrate determined for a similar Ste20 homologue was RRFGSLRRF (18), which bears little resemblance to the Ser 2321 phosphorylation site. The possibility that the arginine at the Ϫ1 position of the Ser 2321 phosphorylation site is a more important specificity determinant than the proline at the ϩ1 position cannot be currently excluded, but the effects of multiple kinase inhibitors on NS5A phosphorylation in vitro and in vivo were more characteristic of a proline-directed kinase activity, since inhibitor concentrations reported to decrease cyclic GMP-dependent kinase, ribosomal S6 protein kinase, calcium/ calmodulin-dependent kinase II, myosin light chain kinase, and phosphorylase kinase in vitro or protein kinase C and cyclic AMP-dependent protein kinase in vitro and in vivo had little or no effect on NS5A phosphorylation, whereas olomoucine, an inhibitor of proline-directed kinases, inhibited NS5A phosphorylation in vitro and in intact cells (2).
The observation that p42 MAPK was capable of phosphorylating a synthetic peptide based on Ser 2321 and its flanking sequence, although not necessarily an indication that MAPK is the major effector of HCV NS5A phosphorylation in vivo, provided evidence that Ser 2321 is a viable substrate for prolinedirected kinases. The poor phosphorylation of this peptide by the HCV NS5A-associated kinase was disappointing, since it probably precludes the use of this peptide as a substrate for kinase purification, but it is perhaps not surprising, given the mechanism for recognition of c-Jun as a substrate by JNKs. Although sequences surrounding the Ser 63 and Ser 73 phosphorylation sites in c-Jun are important for recognition by JNKs, phosphorylation of these residues also requires an upstream binding site with an N-terminal boundary ϳ30 -40 amino acids away from the phosphorylation sites (19). If the HCV NS5Aassociated kinase requires similar docking sites outside the immediate vicinity of Ser 2321 , phosphorylation of synthetic peptides corresponding to this site might be improved by the inclusion of additional flanking sequences. However, simply extending the amino acid sequence around Ser 2321 may not be sufficient for its phosphorylation if recognition by the kinase requires a conformation that can only be achieved in the context of a large portion of NS5A or the host folding machinery.
The resistance of the interactions between NS5A and NS4A or PKR to substitution of Ala for Ser 2321 indicates that phosphorylation at this site is not required for NS5A binding to these proteins and raises the possibility that phosphorylation regulates some other aspect of NS5A function. However, since Ser 2321 is clearly not the only site of NS5A phosphorylation, the possibility that phosphorylation at some other site(s) in NS5A influences its ability to interact with NS4A or PKR cannot be excluded.
Despite these uncertainties, the establishment of Ser 2321 as the major site of intracellular HCV-H NS5A phosphorylation provides valuable information about the substrate specificity of the NS5A-associated kinase that might be used in its eventual purification and identification and paves the way for future investigations on the importance of this phosphorylation for NS5A function in the context of virus replication.