Genetic analysis of a potential zinc-binding domain of the adenovirus E4 34k protein.

E4 34k, the product of adenovirus early region 4 (E4) open reading frame 6, modulates viral late gene expression, viral DNA replication, apoptosis, double strand break repair, and transformation through multiple interactions with components in infected and transformed cells. Conservation of several cysteine and histidine residues among E4 34k sequences from a variety of adenovirus serotypes suggests the presence of a zinc binding domain important for function. Consistent with the hypothesis that E4 34k is a zinc metalloprotein, zinc binding by baculovirus-expressed E4 34k protein was demonstrated in a zinc blotting assay. To investigate the relationship between the potential zinc-binding region and E4 34k function, a series of mutant genes containing single amino acid substitutions at each of the conserved cysteine and histidine residues in E4 34k were constructed. The mutant proteins were examined for the ability to complement the late protein synthetic defect of an E4 deletion mutant, to physically interact with the viral E1b 55-kDa protein (E1b 55k) and cellular p53 protein, to relocalize E1b 55k, and to destabilize the p53 protein. These analyses identified a subset of cysteine and histidine residues required for stimulation of late gene expression, physical interaction with E1b 55k, and p53 destabilization. These data suggest that a zinc-binding domain participates in the formation of the E4 34k-E1b 55k physical complex and that the complex is required in late gene expression and for p53 destabilization.

The 34-kDa product of adenovirus early region 4 (E4) 1 open reading frame (ORF) 6 participates in many of the processes that occur in infected cells, including post-transcriptional steps in late gene expression, viral DNA synthesis, and modulation of the activity, level, and intracellular distribution of p53 (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12). The biochemical details of the activities of E4 34k in these processes are not well understood, but it is clear that many of the functions of the protein are mediated by physical interactions with other viral and host cell proteins. Most prominently, E4 34k and the 55-kDa protein of E1b (E1b 55k) form a physical complex that mediates many of the activities of both proteins (13). For example, the E4 34k-E1b 55k complex facilitates the accumulation of cytoplasmic viral late mRNA and concurrently inhibits the export of cellular mRNA (9, 12, 14 -18). Mutations that eliminate either E4 34k or E1b 55k or prevent the formation of the E4 34k-E1b 55k complex substantially reduce late mRNA accumulation and late protein synthesis and decrease the yield of viral infection severalfold (9, 12, 14 -18). It has been shown recently that nuclear localization of E1b 55k at late times in infection is dependent on E4 34k (19,20) and that the E1b 55k-E4 34k complex shuttles continuously between the nucleus and cytoplasm (21). Together, these observations suggest that the complex plays a direct role in exporting viral late mRNA from the nucleus or that it localizes cellular or viral proteins essential to that process to sites of viral mRNA synthesis. The E4 34k-E1b 55k complex also is required for destabilization of p53, which antagonizes E1a-induced increases in p53 levels in infected cells (5). This activity may contribute to the ability of E4 34k to enhance transforming ability of E1a plus E1b (3,4) and, as suggested for its effect on late mRNA metabolism, may be mediated by the ability of the E4 34k-E1b 55k complex to shuttle between the nucleus and cytoplasm. Finally, the E4 34k-E1b 55k complex stimulates viral DNA replication under some conditions (low multiplicity infections and infections done in the presence of the 14-kDa product of E4 ORF 4) by an unknown mechanism (2). E4 34k also binds to p53 independently of E1b 55k (5). E4 34k binding to p53 in the absence of E1b 55k inhibits the ability of p53 both to stimulate and to repress transcription (1,4), which may also contribute to enhancement of transformation and tumorigenicity by E4 34k. Recently, we demonstrated that E4 34k forms a physical complex with the cellular DNA-dependent protein kinase (DNA PK) and have suggested that that interaction is responsible for the inhibition of double strand break repair by E4 34k (22). It is not known whether this interaction requires E1b 55k. Boivin et al. (23) reported that E4 34k binds to six additional unidentified viral and/or cellular proteins. The functional consequences of these interactions are not known; nor is it known whether those interactions require E1b 55k.
The amino-terminal portion of E4 34k has been implicated clearly in the interactions between E4 34k and both E1b 55k and p53 by protein blotting and co-immunoprecipitation experiments (1,4,16,24). A predicted arginine-rich amphipathic ␣-helix at the carboxyl terminus of E4 34k has been shown by genetic studies to be essential for E4 34k-dependent nuclear localization of E1b 55k and for E4 34k function in lytic infection (25) and may also participate in interactions between E4 34k and E1b 55k or cellular proteins. A prominent feature of the central region of the E4 34k amino acid sequence is a large number of conserved cysteine and histidine residues, which suggests the presence of a zinc-based domain. Proteins involved in the regulation of gene expression and oncogenesis are frequently metalloproteins in which zinc-based domains mediate interactions that are essential for protein function, and zinc-based domains have been shown to participate in a wide variety of protein-nucleic acid and protein-protein interactions (26). The multiple physical interactions that involve E4 34k and the large number of cysteine and histidine residues present in its amino acid sequence raise the possibility that interactions important for the function of E4 34k in viral late gene expression and transformation are mediated by a zinc-based domain. To examine this hypothesis, we tested the ability of E4 34k to bind zinc in a ligand blotting assay. We also constructed a series of E4 34k point mutants that lack cysteine and histidine residues that are highly conserved among E4 34k sequences in various adenovirus serotypes and analyzed those mutants for known E4 34k-dependent functions and for the ability to physically interact with E1b 55k and p53.

Cell Lines
The Spodoptera frugiperda cell line, Sf9, was maintained in Grace's Insect Cell Culture Medium (Life Technologies, Inc.) containing 10% fetal bovine serum at 27°C without CO 2 . 293 (human embryonic kidney) cells (27) were maintained as monolayer cultures in Eagle's minimal essential medium (BioWhittaker, Walkersville, MD) supplemented with 10% fetal bovine serum. The p53-deficient SAOS-2 cell line (human osteogenic sarcoma) was maintained as a monolayer culture in McCoy's 5A medium (Life Technologies) supplemented with 15% fetal bovine serum. The SAOS-2 cell line was obtained from the ATCC (Manassas, VA).

Construction of an E4 34k-expressing Recombinant Baculovirus
E4 34k-expressing baculoviruses were constructed using the MaxBac Baculovirus Expression Vector System, essentially as described by the supplier (Invitrogen, Carlsbad, CA). ORF 6 (adenovirus 5 nucleotides 34058 -33207) was amplified from Ad5 genomic DNA using primers that generate a product with an EcoRI restriction site at the amino terminus of the ORF 6 sequence and a BamHI restriction site at its carboxyl terminus. This PCR product was cloned into the EcoRI and BamHI restriction sites of the baculovirus transfer plasmid pVL1392 (Invitrogen) and named pVL1392-ORF 6. Recombinant virus was generated by transfection of pVL1392-ORF 6 and wild-type Autographica californica nuclear polyhedrosis virus genomic DNA into Sf9 cells using the cationic liposome reagent supplied by Invitrogen. To generate wildtype virus, wild-type baculovirus genomic DNA was transfected into Sf9 cells alone. Supernatants were collected 48 h and 4 days post-transfection, and dilutions were used for plaquing. On day 6 postinfection, plaques were visualized by 3-[4, 5-dimethylthiazol-2yl]-2,5-diphenyltetrazolium bromide staining. Plaques were picked as agar plugs and grown into small stocks for screening. DNA was prepared from a fraction of these ministocks and screened for the presence of an insert corresponding in size to ORF 6 by PCR using primers complementary to regions flanking the baculovirus polyhedrin gene. E4 34k-positive ministocks were used for subsequent rounds of plaque purification. Virus used for generation of large scale stocks and for all experiments underwent four rounds of plaque purification and were judged free of wildtype baculovirus by PCR (not shown).
Zinc Blotting 6 ϫ 10 5 Sf9 cells were infected with 0.25 ml of either wild-type baculovirus, ORF 6 recombinant baculovirus (BacORF6), or E4 ORF 4 recombinant baculovirus (BacORF4) stocks. Insoluble protein corresponding to 1 ϫ 10 5 cell equivalents was fractionated by SDS-PAGE and transferred to nitrocellulose. The immobilized proteins were denatured in 25 ml of renaturation buffer (20 mM HEPES, pH 7.9, 60 mM KCl, 6 mM MgCl 2 , 0.6 mM EDTA) supplemented with 6 M Guanidine-HCl, 2 mM dithiothreitol, and 10% glycerol for 30 min at room temperature. The filters were then incubated with two changes (2 h each at room temperature) of 25 ml of renaturation buffer supplemented with 100 mM guanidine HCl, 0.02% polyvinylpyrrolidone, 2 mM dithiothreitol, and 10% glycerol and with a single change (30 min at room temperature) of 25 ml of renaturation buffer. The blots were equilibrated in metal binding buffer (0.1 M Tris, pH 6.8, 50 mM NaCl) for 1.5 h at room temperature and probed with 65 ZnCl 2 at 10 Ci/lane (4 Ci/ml) in metal binding buffer for 80 min at room temperature. After washing the blots three times (7 min each) with 20 ml of metal binding buffer, zinc bound by immobilized protein was visualized by autoradiography.

Plasmids
The Ad2 E4 ORF 6 cDNA plasmid pilE4J has been described (32). For construction of the E4 34k expression plasmid pCMV6.9, a fulllength ORF 6 PCR product was produced from Ad5 genomic DNA. EcoRI and NdeI cleavage sites were added at the 5Ј end, and XbaI and BamHI sites were added at the 3Ј-end of the coding sequences by inclusion in the PCR primers, and the fragment was cloned in pcDNA3 (Invitrogen) after cleavage with EcoRI and XbaI. Both strands of the ORF 6 segment in pCMV6.9 were sequenced to confirm that no mutations were present. The pmycOE6.5 series of E4 34k expression plasmids were prepared by subcloning sequenced wild-type and mutant ORF 6 segments into pmycpl.1, a derivative of pmycRK5 (gift of Dr. Randall Reed, Johns Hopkins University). These plasmids contain the SV40 origin of replication and are amplified in the presence of SV40 T antigen. E4 34k expression is driven by the CMV promoter, and the protein is expressed as a fusion with a c-myc epitope. pC53SN3 was provided by Dr. Thomas Shenk (Princeton University) (33). Dr. David Ornelles (Wake Forest University, Winston-Salem, NC) provided the pCMV55k and pCMVNeoBam plasmids (20,33). Dr. Randall Reed provided pRSV␤gal.

Construction of ORF 6 Point Mutants
Overlap PCR Mutagenesis-Mutants C100S, H115L, and H123L were created using the overlap PCR method (34). For each mutant, two PCR products with a primer length overlap containing the mutant residue were synthesized using pilE4J as the template. One of these was prepared using a mutant E4 sense oligonucleotide (e.g. C100S sense) and a flanking wild-type ORF 6 primer that spans a natural DraIII site near the 5Ј-end of ORF 6; the other was prepared with a mutant E4 antisense oligonucleotide (e.g. C100S antisense) and a wildtype ORF 6 primer that includes a natural KpnI site located near the center of the ORF. The PCR products were purified by the QiaQuick spin system (Qiagen, Inc., Chatsworth, CA), 1 ng each of the purified products were mixed, and a second round of amplification was performed using the wild-type ORF 6 primers. After digestion with DraIII and KpnI, the final PCR product was cloned into the DraIII and KpnI sites of pilE4J. The presence of the mutation and the integrity of the PCR-derived portion of ORF 6 was verified by sequencing. For subsequent experiments, PmlI fragments of ORF 6 containing the mutations were transferred to pCMV6.9. The resulting clones were sequenced across all of ORF 6 to confirm the presence of the desired mutation.
Recombinant PCR (RPCR) Mutagenesis-The remaining mutants (C51S, C67S, C124S, H125L, C126S, C134S, H185L, H196L, C224S, C227S, and the double mutant C237S/C238S) were constructed by the RPCR method (35), using pCMV6.9 as the template. For each mutant, four primers were used in PCR amplifications. Two, used in the construction of all of the mutants, were complementary to the regions 2295-2313 (cDNA3 sense) and 2816 -2796 (cDNA3 antisense) of the pCMV6.9 plasmid. The remaining two primers (the mutant sense and mutant antisense primers) were mutant-specific, complementary, 23-27-base oligonucleotides containing the desired base substitutions. In separate PCRs, using linearized pCMV6.9 as template, two PCR products were produced from the primer combinations cDNA3 sense/mutant antisense, and cDNA3 antisense/mutant sense. These products overlap in the regions containing the mutations and in the region 2295-2816 in the pCMV6.9 backbone. After purification by the QiaQuick system (Qiagen), equimolar amounts of the PCR products were used to transform 50 l of DH5␣ MAX Efficiency competent Escherichia coli cells (Life Technologies), where recombination in the regions of PCR product overlap generated circular plasmids containing the ORF 6 sequence with the desired mutation. Typically, 1 ⁄12 to 1 ⁄25 of the purified PCR product was used for transformations, and transformants were obtained by plating the entire transformation reaction onto ampicillincontaining medium. The sequence of each mutant plasmid was verified by sequencing all of ORF 6 on both strands.
Transfections 293 cells were seeded in 24-well plates at a density of 6.5 ϫ 10 5 cells/well and incubated overnight at 37°C. Monolayers were transfected with 2.0 -2.5 g of DNA and 8 l of LipofectAMINE reagent (Life Technologies) according to the manufacturer's instructions. Formation of DNA-lipid complexes was in a total volume of 200 l of Opti-MEM (Life Technologies) for 30 min at room temperature. Complexes were added to rinsed cell monolayers and incubated for 5 h at 37°C in a volume of 1 ml of Opti-MEM, followed by replacement of the transfection medium with 2 ml of Eagle's minimal essential medium containing 10% fetal bovine serum, penicillin, and streptomycin. For E4 deletion mutant complementation experiments, the transfected DNAs included 0.1 g of pCMV6.9 or mutant plasmid plus 0.9 g of pcDNA3 (or 1 g of pcDNA3 for E4 34k-negative controls) and 1 g of pRSV␤gal. For studies of E4 34k-E1b 55k association, cells received 1 g of pCMV6.9, mutant plasmid, or pcDNA3 plus 1 g of pRSV␤gal. For studies of E4 34k-p53 association, cells were transfected with 1 g of pmycOE6.5, mutant plasmid, or pmycpl.1 plus 1 g of pRSV␤gal and 0.5 g of pRSVTAg.
SAOS-2 cells were seeded onto 60-mm dishes at a density of 1 ϫ 10 6 cells/dish and incubated overnight at 37°C. Monolayers were transfected with 0.125 g of pC53SN3, 0.5 g of pCMV55k, 2 g of pRSV␤gal, and 2.375 g of pCMV6.9 or ORF 6 mutant plasmid, using 10 l of LipofectAMINE reagent. Formation of DNA-lipid complexes was in a volume of 600 l of Opti-MEM, incubation with the cell monolayer was in 3.1 ml of Opti-MEM, and after transfection the medium was replaced with 5 ml of supplemented McCoy's 5A medium.
Transfected cells were harvested 48 -53 h after transfection. Monolayers were rinsed once in phosphate-buffered saline (PBS) and collected by scraping in 1 ml of PBS. Cells were pelleted by low speed centrifugation; resuspended in 100 l of 0.25 M Tris, pH 7.8, containing 1 g/ml each pepstatin A, aprotinin, and leupeptin; and lysed by sonication. Following sonication, the insoluble material in the cell lysate was removed by microcentrifugation at 4°C for 10 min at 14,000 rpm. The supernatants were retained.

␤-Galactosidase Assays
10 l of cell lysate ( 1 ⁄10 total volume) were combined with 221 l of 0.1 M sodium phosphate buffer, pH 7.5, 3 l of 100ϫ magnesium solution (0.1 M MgCl 2 , 4.5 M ␤-mercaptoethanol), and 66 l of 1ϫ o-nitrophenyl ␤-D-galactopyranoside (4 mg/ml in 0.1 M sodium phosphate buffer, pH 7.5) for a final volume of 300 l. These reactions were typically incubated at 37°C for about 8 -12 min (293 cells) or 2 h to overnight (SAOS-2 cells), at which time the reactions were stopped by the addition of 0.5 ml 1 M Na 2 CO 3 . The absorbance was measured at 420 nm, and this value was used to calculate units of ␤-galactosidase activity using the following equation: ((absorbance)/(time at 37°C in min) (volume of lysate in ml)) ϫ 1000).

Complementation Assays
Mutant or wild-type E4 34k plasmids were transfected into 293 cells as described above. At 24 h after transfection, the cells were infected at 50 plaque-forming units/cell with H5dl1004 (9), an E4 deletion mutant that lacks all E4 ORFs and is defective for viral late protein synthesis. The transfected/infected cells were incubated with 20 Ci [ 35 S]cysteine beginning at 24 h postinfection and were harvested 5 h later. Viral late protein synthesis was assessed by immunoprecipitation from cell extracts normalized for ␤-galactosidase activity (30 units each reaction) with a polyclonal antiserum reactive with viral late proteins (␣-late serum). The immunoprecipitates were fractionated by SDS-PAGE and fluorographed. E4 34k synthesis was determined by immunoblotting cell lysates normalized for ␤-galactosidase activity (120 units of each lysate) with anti-E4 34k serum (E4orf6-C). Adenovirus E2 72k expression was determined by immunoblotting with monoclonal antibody B6.
p53 Co-immunoprecipitation-Transfected cells were labeled 48 -53 h post-transfection with 50 Ci of [ 35 S]cysteine. Cell lysates containing equal radioactivity (approximately 6.2 ϫ 10 5 cell equivalents) were used in immunoprecipitation reactions. Immunoprecipitated proteins were fractionated on 12% SDS-PAGE gels and visualized by autoradiography. Equal cell equivalents of cell lysates were also immunoblotted with E4orf6-C antiserum to monitor E4 34k expression.
E1b 55k Co-immunoprecipitation-Cells were harvested 48 h posttransfection. Equal amounts of lysate (approximately 6.2 ϫ 10 5 cell equivalents) were used in each immunoprecipitation reaction. Immunoprecipitated proteins were fractionated on 10% SDS-PAGE gels and immunoblotted with E4orf6-C antiserum to detect co-precipitated E4 34k. Equal cell equivalents of whole cell lysates were also immunoblotted with E4orf6-C antiserum to monitor E4 34k expression.

Measurement of p53 Protein Levels
SAOS-2 cells were transfected with CMV expression plasmids encoding p53, E1b 55k, wild-type or mutant E4 34k proteins, and ␤-galactosidase. 48 h post-transfection, cell lysates were prepared and assayed for ␤-galactosidase activity. Cell lysates were run on 10% SDS-PAGE gels and transferred to nitrocellulose. Levels of E4 34k and p53 were detected by sequential immunoblotting of the same membrane with E4orf6-C antiserum followed by the ␣-p53 DO1 monoclonal antibody. When possible, equal ␤-galactosidase units of cell lysate were immunoblotted. If ␤-galactosidase activity was too low to measure, equal cell equivalents of lysate were confirmed to have equivalent wild-type and mutant E4 34k expression by immunoblotting for E4 34k before determining p53 protein levels. In separate experiments, the ratio of E4 34k to p53 protein was determined to be critical for efficient p53 destabilization. The H123L mutant protein was consistently less abundant than the wild-type E4 34k protein in these experiments, and p53 destabilization by that mutant therefore was determined in experiments where the E4 34k/p53 ratio was maintained by reductions in the amount of p53 DNA transfected. In those experiments, comparisons were made with cells transfected with reduced amounts of both the p53 and wildtype ORF 6 plasmids to maintain both the correct E4 34k/p53 ratio and achieve comparable wild-type and H123L E4 34k protein levels.

Immunofluoresence
Plasmids encoding either wild-type or mutant E4 34k were transfected into 293 cells (1 g of pCMV6.9, mutant plasmid, or pcDNA3 plus 1 g of pcDNA3) seeded onto glass coverslips. 48 h post-transfection, the coverslips were washed three times in PBS, and cells were fixed in 4% paraformaldehyde in PBS for 10 min at room temperature. Coverslips were then washed three times in PBS, and the cells were permeabilized with ice-cold 100% acetone for 30 s and were washed again three times in PBS. Following fixation and permeabilization, each coverslip was incubated with 100 l of 10% fetal bovine serum in PBS for 1 h at room temperature in a humidity box to block nonspecific binding. Primary antibody was diluted in 10% fetal bovine serum in PBS (␣-E1b 55k 2A6, 1:1000; E4orf6-C, 1:750) and 100 l of diluted antibody was added to each coverslip. Coverslips were incubated for 1 h at room temperature in a humidity box and were then washed three times with PBS. Fluorescein isothiocyanate-conjugated goat ␣-(mouse IgG (H ϩ L)) (Roche Molecular Biochemicals) and Texas Red-conjugated goat ␣-(rabbit IgG (H ϩ L)) (Molecular Probes, Eugene, OR) secondary antibodies were diluted 1:200 in 10% fetal bovine serum in PBS, and 100 l of diluted antibody was incubated with each coverslip for 40 min at room temperature in a humidity box. Coverslips were washed three times with PBS and one time with distilled water before mounting onto slides with Permafluor (Lipshaw-Immunon, Pittsburgh, PA). Fluorescence was observed and photographed on color print film with a Nikon Eclipse 800 photomicroscope using epifluorescence illumination and filters for the appropriate fluorophore. Photographs were digitized by scanning, and image quality was adjusted in Adobe Photoshop using the "Auto Levels" function before cropping and assembly to produce Fig. 5.

RESULTS
Alignment of E4 34k Proteins from Mammalian Adenoviruses-Adenovirus E4 34k amino acid sequences were obtained from a periodic automated search of the XREFdb data base (36) using the Ad2 E4 34k protein sequence as the query. In the absence of functional data for most of the proteins encoded by adenoviruses other than Ad2 and Ad5, the criteria for inclusion in the analysis presented here were high homology score and presence of the sequence motif HCHCXXXXSLQC. Sequences in the resulting collection include representatives of human adenovirus subgroups A, C, D, and F (Ad2/5, -9, -12, and -40), as well as canine (CAV-1), ovine (OAV287), and murine (MAV-1) adenoviruses. When these sequences were aligned using MegAlign software, numerous highly conserved cysteine and histidine residues were apparent along the entire length of these proteins (Fig. 1). Conservation of these residues, especially in nonhuman adenoviruses whose protein sequences were otherwise substantially divergent, suggests that they are important for E4 34k function. Further, although no pattern of cysteine and histidine residues common to previously described metal-binding domain sequences was found, their large number suggests that they coordinate a metal ion, most likely zinc. If that is the case, the metal-binding domain may possess a novel structure.
Baculovirus-expressed E4 34k Protein Binds Zinc-To test the hypothesis that the E4 34k protein binds zinc, recombinant E4 34k protein was examined in a ligand blotting assay under conditions that detect zinc binding by authentic metalloproteins (37). Sf9 cells were infected with a recombinant baculovirus that expressed E4 34k (BacORF6), wild-type baculovirus, or BacORF4, a recombinant expressing an unrelated E4 protein. Preliminary characterization of BacORF6 indicated that the majority of the E4 34k produced was insoluble. Therefore, insoluble fractions of infected cells and an uninfected control were prepared 72 h postinfection and were fractionated by SDS-PAGE. After electrophoresis, the fractionated proteins were transferred to a solid support and were examined for E4 34k expression (by immunoblotting) and zinc binding (Fig. 2). For the zinc binding assay, the transferred proteins were first renatured on the membrane and then were probed with 65 ZnCl 2 . Zinc binding by BacORF6-infected cell proteins was compared with zinc binding by proteins synthesized by BacORF4, which contains an identical disruption of the polyhedrin gene. Immunoblotting detected a prominent band at the mobility expected for E4 34k in lysates prepared from FIG. 1. Alignment of E4 34k homologues from mammalian adenoviruses. The sequences collected here are representative of human subgroups A (Ad2/5), C (Ad9), D (Ad12), and F (Ad40) and of canine (CAV-1), ovine (OAV287), and murine (MAV-1) adenoviruses. The criteria for inclusion in the analysis were BLAST homology to the Ad2/Ad5 protein and presence of the sequence motif HCHCXXXSLQC (residues 123-133 in Ad2/Ad5). Black shading indicates cysteine residues that are conserved in a majority (Ͼ4) proteins; gray shading indicates conserved residues other than cysteine or histidine. Each amino acid residue mutagenized in this study is marked with its position in the protein and/or an asterisk.
BacORF6-infected, but not wild-type baculovirus-infected cells. Zinc blotting revealed a zinc-binding protein that comigrated with the E4 34k protein in BacORF6 cell lysates that was absent from BacORF4 cell lysates (Fig. 2). This result suggests strongly that E4 34k is a zinc metalloprotein. A very intense reactive band with a mobility slightly less than that of E4 34k was present in wild-type baculovirus-infected cell lysates (Fig.  2). The identity of this protein is not certain, but it is probably the baculovirus polyhedrin protein (29 kDa), since it is absent from uninfected cell lysates and from BacORF4 cell lysates. Zinc binding by the proteins in the molecular weight standards provides a measure of the specificity of this assay; carbonic anhydrase (30 kDa), the only authentic zinc metalloprotein in the standards, binds zinc under these conditions, while the others do not.
Construction of E4 34k Point Mutants-A total of 15 of cysteine and histidine residues in E4 34k were chosen for mutagenesis (Fig. 1). These included all eight of the residues conserved in every adenovirus serotype examined (Cys 51 , His 123 , Cys 124 , His 125 , Cys 126 , Cys 134 , His 185 , and Cys 227 ; numbers refer to position in the Ad2/Ad5 protein), three residues conserved in all but one serotype (Cys 67 , Cys 100 , and Cys 224 ), three residues conserved among all human adenovirus serotypes (His 196 , and Cys 237 /Cys 238 ; mutagenized together), and one nonconserved residue (His 115 ). In each case, the introduced mutation consisted of the substitution of a serine residue for cysteine or a leucine residue for histidine. Mutations were introduced into ORF 6 by the overlap PCR and RPCR methods. Overlap PCR generated a mutant PCR fragment containing DraIII and KpnI restriction sites that was cloned into the corresponding sites of the wild-type ORF 6 sequence contained in the plasmid pilE4J. The presence of the desired mutation was confirmed by sequencing. A PmlI restriction fragment containing the mutation was then subcloned into the PmlI sites of the plasmid pCMV 6.9, to place ORF 6 expression under the control of the cytomegalovirus promoter. Mutations constructed by RPCR were introduced directly into the E4 34k coding sequence contained in the plasmid pCMV 6.9. Sequencing confirmed the integrity of ORF 6 and the presence of the mutation in each plasmid. For some experiments, the wild-type and sequenced mutant E4 34k genes were subcloned from pCMV 6.9 derivatives into the plasmid pmycpl.1.

Complementation of Viral Late Protein Synthesis by E4 34k
Mutants-E4 is required for viral late gene expression, and E4 deletion mutants are profoundly defective for viral late protein synthesis. The late protein synthetic defect of E4 mutants can be complemented in trans by plasmids that express wild-type E4 34k (30). To test the ability of the E4 34k cysteine and histidine mutant proteins to function in viral late gene expression, the ability of mutant plasmids to complement late protein synthesis in cells infected by the E4 deletion mutant H5dl1004 was examined. 293 cells were transfected with pCMV6.9-based plasmids bearing wild-type or mutant ORF 6, and 24 h later, the transfected cells were infected with H5dl1004. At 24 h postinfection, proteins were labeled with [ 35 S]cysteine. Viral late proteins were immunoprecipitated from extracts of the transfected/infected cells with antiserum against denatured Ad2 capsid proteins, and late protein synthesis was assessed by SDS-PAGE and autoradiography (Fig. 3). To normalize for transfection efficiency, a ␤-galactosidase expression plasmid was included in all transfections, and equal ␤-galactosidase units of cell lysates were used in the immunoprecipitation reactions. Mutants C67S (cysteine to serine at position 67), H196L (histidine to leucine at position 196) and H115L (not shown) functioned as well as wild-type E4 34k in this experiment. All of the remaining mutants were severely defective in their ability to complement the viral late protein synthetic defect of the mutant. The amounts of wild-type or mutant E4 34k proteins synthesized in complementation assays were measured by immunoblotting cell lysates normalized for ␤-galactosidase activity with the E4 34k-specific antiserum, E4orf6-C (Fig. 3). The C51S, H123L, C124S, H125L, C126S, C134S, and H185L mutant proteins accumulated to levels comparable with the wildtype E4 34k level. Therefore, these mutant proteins exhibit a specific defect in the function required for stimulation of viral late gene expression that is not the result of reduced mutant protein accumulation. C67S and H196L, which function normally in the complementation test, also accumulated normal amounts of E4 34k. The C100S (not shown), C224S, C227S, and C237/238S mutant proteins accumulated to significantly lower levels than did wild-type E4 34k, and the defects associated with those mutants in the complementation assay might therefore have been due simply to reduced steady-state levels of mutant protein. To address that possibility, the complementation experiments were repeated in cells transfected with larger amounts of those mutant plasmid DNAs in an attempt to increase mutant protein accumulation. Even when the accumulation of mutant protein was comparable with that of the wild-type protein under the standard conditions, the C224S, C227S, and C237/238S mutant proteins remained unable to stimulate viral late protein synthesis (Fig. 3), indicating that these proteins are inherently nonfunctional. It seems most likely that the reduced accumulation of these mutant proteins is due to reduced stability that results from abnormal folding. Reduced activity in complementation assays might also reflect a nonspecific effect of generally abnormal structure. For that reason, and because of the technical difficulties presented by the low levels of mutant proteins produced by these mutants, they have not been analyzed further. The C100S mutant did not accumulate detectable amounts of protein in these experiments and likewise was not examined further.
Defects in the ability of the E4 34k mutants to stimulate viral late protein synthesis might arise from nonspecific inhibition of protein synthesis by the mutant proteins due, for example, to increased cytotoxicity. If that is the case, expression of viral proteins not dependent upon E4 34k should also be decreased. To examine this possibility, expression of viral E2 72k was measured by immunoblotting in cells transfected with plasmids encoding wild-type E4 34k and each of the stable E4 34k mutant proteins. Equal volumes of cell lysate from the complementation assays were blotted with the E2 72k-specific monoclonal antibody, B6 (Fig. 3). Levels of E2 72k were comparable for all samples, indicating that the failure of the mutants to stimulate viral late gene expression did not result from generalized effects on protein synthesis.
Physical Interactions between E4 34k Mutant Proteins and the Viral E1b 55k Protein-E4 34k and the E1b 55k protein are physically associated in infected cells (13), and the E4 34k-E1b 55k protein complex mediates many of the functions of the two proteins, including those required for viral late gene expression (14,15,17,18). Disruption of the complex by mutations in the conserved cysteine and histidine residues in E4 34k might account for the observed defects in the ability to stimulate viral late gene expression in the transfection assay. Therefore, the ability of the mutant E4 34k proteins to interact physically with E1b 55k was analyzed by co-immunoprecipitation. 293 cells, which constitutively express E1b 55k, were transfected with pCMV6.9-based wild-type or mutant E4 34k expression plasmids, and cell extracts were made 48 h post-transfection under conditions that preserve the wild-type E4 34k-E1b 55k interaction. E1b 55k and associated proteins were immunoprecipitated with the E1b 55k-specific monoclonal antibody, 2A6. Immunoprecipitates were fractionated by SDS-PAGE and were assayed for co-precipitating wild-type or mutant E4 34k by immunoblotting with the E4orf6-C antiserum.
Of the nine stable mutant proteins tested (Fig. 4), only the C67S and H196L mutant proteins co-precipitated with the E1b 55k protein. By contrast, the C51S, H123L, C124S, H125L, C126S, C134S, and H185L proteins do not co-precipitate with E1b 55k. This indicates that the wild-type residues at these positions are required for formation or for normal stability of the E4 34k-E1b 55k complex. The essential cysteine and histidine residues might participate directly in contacts with E1b 55k; alternatively, they may play a role in organizing regions of the E4 34k protein that participate in binding. Since the C67S and H196L mutant proteins were also the only proteins functional in the complementation assay, these results confirm that the ability to form a stable physical complex with the E1b 55k protein is essential for the function of E4 34k in viral late gene expression.

Subcellular Localization of E4 34k Mutant Proteins and Relocalization of the E1b 55k Protein by Mutant E4 34k Proteins-
When expressed alone in transient expression systems, E1b 55k is localized in the cytoplasm. Co-expression of E4 34k redirects a large portion of the cytoplasmic E1b 55k protein to the nucleus (20). Similarly, expression E4 34k increases the amount of E1b 55k found in the nucleus of infected cells (19,20). The ability of the E4 34k mutant proteins to relocalize E1b 55k was examined by an indirect immunofluorescence assay. 293 cells grown on glass coverslips were transfected with wildtype or mutant E4 34k expression plasmids. At 48 h posttransfection, the localization of E1b 55k and E4 34k was determined by dual label immunofluorescence. E1b 55k was visualized with the monoclonal antibody 2A6 and a fluorescein isothiocyanate-conjugated secondary antibody, while E4 34k was visualized with E4orf6-C and a Texas Red-conjugated secondary antibody.
In the absence of E4 34k, E1b 55k was found both diffusely distributed throughout the nucleus and cytoplasm and in very brightly stained flecks also present throughout the cells (not shown). In adenovirus-transformed cells, where E4 34k is also absent, the E1b 55k and p53 proteins co-localize in brightly staining perinuclear bodies (38 -40); the brightly stained flecks observed here resemble this previously published E1b 55k staining pattern. In separate experiments, 293 cells were stained for the p53 protein, which was present in brightly stained flecks that were identical to those observed for E1b 55k in the same cells (data not shown).
As previously reported (16,19,20), transiently expressed wild-type E4 34k was predominantly nuclear, diffusely distributed throughout the nucleoplasm, and excluded from nucleoli (Fig. 5, top left panels). When the distribution of the mutant E4 34k proteins was compared with that of the wild-type E4 34k protein, no obvious differences were observed (Fig. 5, E4 34k  columns). These results indicate that alterations in the subcellular distribution of the mutant proteins are not responsible for defects in their activities.
In individual cells expressing the wild-type E4 34k protein, a large portion of the E1b 55k protein was directed to the nucleus, where it co-localized with E4 34k (Fig. 5, top left panels). A small amount of E1b 55k protein remained in the cytoplasm in these cells diffusely and in flecks. The amount of cytoplasmic E1b 55k staining varied between cells expressing wild-type E4 34k, but in all such cells, the vast majority of E1b 55k protein FIG. 4. Co-immunoprecipitation of E4 34k with the E1b 55k protein. Extracts were prepared from 293 cells transfected with plasmids that express either wild-type (wt) or mutant E4 34k. Top row, extracts were prepared from transfected cells 48 h post-transfection, and E1b 55k was immunoprecipitated with an E1b 55k-specific monoclonal antibody. Immunoprecipitates were fractionated by SDS-PAGE and analyzed for co-precipitating E4 34k by immunoblotting with E4orf6C antiserum. The plasmid used for transfection is indicated at the top of each lane; the position of E4 34k is indicated on the left. Note that in this gel, the E4 34k protein appears just below a nonspecific band (probably an IgG degradation product) of variable intensity. Bottom row, amounts of cell extract equal to those used for immunoprecipitation were fractionated by SDS-PAGE and immunoblotted with the E4orf6C antiserum. Lanes correspond to the immunoprecipitations above. The position of E4 34k is marked on the left. was present in the nucleus. The C67S and H196L mutant proteins, both of which complement the late protein synthetic defect of H5dl1004 and co-immunoprecipitate with E1b 55k, also retained the ability to direct the nuclear localization of the E1b 55k protein. All of the remaining mutant proteins, which are unable to stimulate late protein synthesis and do not associate with E1b 55k in co-immunoprecipitation experiments, failed to alter the distribution of the E1b 55k protein. The perfect correlation between relocalization and co-immunoprecipitation suggests that the physical interaction between E4 34k and E1b 55k detected by co-immunoprecipitation is required for E1b 55k relocalization.
Physical Interaction between E4 34k Mutant Proteins and the Cellular p53 Protein-The wild-type E4 34k protein physically associates with the cellular p53 protein both in vitro and in vivo (1,5) and has consequent effects on the activity of p53 as a transcription factor (1,4,11) and on its stability (3,4,11). To determine whether the conserved cysteine and histidine residues in E4 34k are important for the interaction between E4 34k and p53, physical associations between the cellular p53 protein and the mutant E4 34k proteins were assessed by co-immunoprecipitation.
For these experiments, wild-type or mutant E4 34k proteins were expressed in 293 cells by transfection with pmycOE6.5based expression plasmids. 48 h post-transfection, the transfected cells were labeled with [ 35 S]cysteine, and cell lysates were prepared for co-immunoprecipitation. Endogenous p53 and associated proteins were immunoprecipitated with the p53-specific antibody D01, and the immunoprecipitates were analyzed by SDS-PAGE for co-precipitating E4 34k (Fig. 6).
Expression levels of wild-type and mutant E4 34k proteins in the cell lysates used for co-immunoprecipitation were determined to be equivalent by immunoblotting equal volumes of cell lysate with E4orf6-C antiserum.
All of the mutant proteins retained the ability to physically interact with p53. These results suggest that the conserved cysteine and histidine residues in E4 34k are not important for the physical interaction between E4 34k and p53. Additionally, they show that the ability of E4 34k to interact with p53 is not sufficient for E4 34k function in viral late gene expression; nor does it confer the ability to associate physically with E1b 55k.
Destabilization of the Cellular p53 Protein by E4 34k Mutant Proteins-Several laboratories have reported that E4 34k reduces the stability and levels of p53 in infected, transformed, and transfected cells (3)(4)(5)11). However, conflicting data concerning the requirement for E1b 55k in these activities have been presented. Because the mutant collection described here includes both proteins that associate normally with E1b 55k and proteins that do not, the abilities of the mutant proteins to destabilize p53 might shed light on the E1b 55k requirement. Therefore, we tested the effects of mutant E4 34k proteins on p53 protein levels in cells co-transfected with E4 34k, p53, and E1b 55k expression plasmids.
SAOS-2 cells were co-transfected with expression plasmids encoding E1b 55k, p53, and wild-type or mutant E4 34k. 48 h post-transfection, cell lysates were prepared and analyzed for p53 and E4 34k protein levels by sequential immunoblotting of the same membrane with the D01 and E4orf6-C antibodies. The presence of the wild-type E4 34k protein substantially reduced p53 levels (Fig. 7), as reported earlier (11). Of the 10 stable E4 34k cysteine and histidine mutants, only C67S and H196L were as effective as wild-type E4 34k in reducing levels of p53. C67S and H196L are also the only mutants capable of forming complexes with E1b 55k. Therefore, under the conditions used here, the ability to form a physical association with E1b 55k correlates perfectly with the ability to reduce p53 levels, consistent with an obligatory role for the E4 34k-E1b 55k complex in p53 destabilization. The magnitude the reduction of p53 levels by E4 34k expression is less dramatic in our experiments than in earlier reports. However, the ratio of E4 34k to p53 expression strongly influences the magnitude of the effect of E4 34k on p53 levels (data not shown), and differences in that ratio might account for the discrepancy.
Summary of E4 34k Mutant Protein Phenotypes-A summary of the phenotypes of the E4 34k cysteine and histidine mutant protein collection is presented in Fig. 8. These phenotypes suggest a link between the ability to form a physical FIG. 5. Relocalization of E1b 55k by E4 34k mutant proteins. Wild-type or mutant E4 34k plasmids were transfected into 293 cells growing on coverslips. At 48 h post-transfection, the coverslips were processed and the distribution of E1b 55k and E4 34k was visualized by dual-label indirect immunofluorescence. Pairs of photographs of the same field, showing fluorescence due either to E4 34k (left) or E1b 55k (right) are presented for each mutant and for wild-type E4 34k.
FIG. 6. Co-immunoprecipitation of E4 34k with p53. Plasmids encoding wild-type (wt) or mutant E4 34k proteins were transfected into 293 cells. 48 h post-transfection, the transfected cells were labeled with [ 35 S]cysteine. Top row, p53 was immunoprecipitated from cell extracts with p53-specific monoclonal antibody, the immunoprecipitates were fractionated by SDS-PAGE, and p53 and co-precipitating E4 34k were visualized by autoradiography. Lanes are labeled at the top with the E4 34k mutant used for transfection, and the positions of p53 and E4 34k are indicated on the right. The modest reduction in co-precipitation of C134S and H125L seen in this autoradiogram was not reproducible. Bottom row, amounts of cell extract equal to those used for immunoprecipitation were fractionated by SDS-PAGE and immunoblotted with the E4orf6C antiserum. Lanes correspond to those in the upper row. association with E1b 55k and the functions of E4 34k in viral late gene expression, relocalization of the E1b 55k protein, and p53 destabilization. The results additionally identify a subset of cysteine and histidine residues important for some functions of E4 34k that may participate in the formation of a metalbinding domain.

DISCUSSION
The adenovirus type 5 E4 34k protein contains a large number of cysteine and histidine residues that are conserved among the corresponding proteins found in several mammalian adenovirus serotypes (Fig. 1). E4 34k participates in multiple physical associations with other components in infected cells and lacks any recognized interaction domains (for example, a leucine zipper). These combined features raise the possibility that E4 34k physically interacts with other proteins through a novel zinc-based structure that involves the conserved cysteine and histidine residues. In the experiments described here, the ability of the wild-type E4 34k protein to bind zinc was assessed, and the involvement of the conserved cysteine and histidine residues in E4 34k function was addressed by examination of mutants lacking those residues individually.
Zinc binding by E4 34k was analyzed by a solid phase zinc blotting assay. This assay accurately detects zinc binding by authentic zinc metalloproteins and has a low nonspecific background (37). In addition, zinc blotting requires neither soluble nor purified material, which is critical for spectroscopic methods of zinc determination. In these experiments, proteins from crude extracts are fractionated by SDS-PAGE and immobilized on nitrocellulose. The filters are then probed with the radioactive zinc isotope, 65 Zn, and proteins that bind to the isotope are visualized by autoradiography. Zinc binding by E4 34k was evaluated using insoluble protein produced by recombinant baculovirus. E4 34k binds zinc under zinc blotting conditions, consistent with the hypothesis that E4 34k is a zinc metalloprotein (Fig. 2).
The arrangement of the cysteine and histidine residues iden-tified as essential for function by the E4 34k mutants described here does not correspond closely to any of the well described families of zinc binding motifs, suggesting that zinc is coordinated in E4 34k by a novel arrangement of residues. Zinccontaining proteins described so far exhibit a very wide variety of structures that arise from alternate arrangements of zinccoordinating residues. Even within single families of zinc-binding proteins, such as the functionally diverse RING finger proteins, substantial variations in the relative spacing of zinccoordinating cysteine and histidine residues are possible (41), and these variations are reflected in the structural differences noted for those proteins whose structure has been solved (42,43). Therefore, it is not unreasonable to imagine that there are zinc-binding structures not yet described. The number of essential zinc residues present in E4 34k is consistent with the coordination of two or more zinc ions, and E4 34k may contain multiple small zinc-based domains or a single domain formed by a large number of cysteine and histidine residues. If zinc-dependent structures mediate important interactions between E4 34k and other components of infected cells and if the conserved cysteine and histidine residues of E4 34k participate in zinc binding, mutant proteins that lack those amino acids may display functional defects. Therefore, a series of E4 34k mutants in which individual cysteine and histidine residues were replaced with serine and leucine, respectively, were constructed and were analyzed in transfection assays for the ability to function in viral late gene expression, to interact physically with the viral E1b 55k protein, to relocalize E1b 55k, and to bind to and destabilize the cellular p53 protein. Mutation of several of the conserved cysteine and histidine residues resulted in loss of function. Out of 15 mutants constructed and analyzed, seven (C51S, H123L, C124S, H125L, C126S, C134S, and H185L) accumulated wild-type levels of protein but were defective in the ability to complement the late protein synthetic defect of an E4 deletion mutant. The same seven mutants were also defective for physical interaction with the E1b 55k protein, relocalization of the E1b 55k protein, and destabilization of p53. Three mutants (C67S, H115L, and H196L) functioned like the wild-type E4 34k protein in all of these assays. Four additional mutants (C100S, C224S, C227S, and C237/238S) produced low steady-state levels of mutant proteins compared with wild type and were not analyzed further.
In infected cells, the E4 34k and E1b 55k proteins function to stimulate export of viral late mRNA from the nucleus, and that activity is dependent on the formation of the E4 34k-E1b 55k physical complex (9, 12, 14 -18). The exact correlation between the ability of mutant E4 34k proteins to stimulate viral late gene expression and their ability to bind E1b 55k therefore suggests that, among the mutants described here, disruption of the E4 34k-E1b 55k interaction is the primary defect. While the mechanism of action of the E4 34k-E1b 55k protein complex is not certain, it has recently been shown that the complex shuttles between the nucleus and the cytoplasm and that that activity is dependent upon the physical interaction between the E4 34k and E1b 55k proteins (21). Shuttling of the E4 34k-E1b 55k protein complex presents the possibility that E4 34k, in combination with the E1b 55k protein, may directly control the movement of viral RNA out of the nucleus of an infected cell. Alternatively, the shuttling of the complex might regulate the activity of other proteins required for efficient viral replication through alterations in their subcellular localization (e.g. by directing them to the regions surrounding viral replication centers). In addition to stimulating export of viral mRNA, the E1b-55k complex inhibits export of newly synthesized cellular mRNA (14,18). Recruitment to viral replication centers of cellular factors required for efficient RNA transport would FIG. 7. Effects of mutant E4 34k proteins on accumulation of p53. SAOS-2 cells, which lack endogenous p53, were transfected with plasmids that express p53, E1b 55k, and either wild-type (wt) or mutant E4 34k proteins. At 48 h post-transfection, extracts were prepared, fractionated by SDS-PAGE, and immunoblotted either with antibodies specific for p53 (top row) or E4 34k (bottom row). The positions of the p53 and E4 34k signals are indicated. restrict their activity to viral RNAs and thus could account for the observed alterations of host cell RNA metabolism (14,17,18,44). Alterations in subcellular localization are known to affect the activity of a number of proteins; members of the E2F family of transcription factors function when nuclear, but not cytoplasmic (45), and the ability of the Cdc25 protein to regulate cell cycle progression has recently been linked to its cell cycle-dependent nuclear localization (46,47). It has previously been proposed that the E4 34k-E1b 55k complex relocalizes a primate cell-specific factor to viral replication domains to allow preferential access of viral late mRNAs to the export machinery (19,20).
Reductions in p53 levels by E4 34k mutant proteins also correlated with the abilities of the proteins to bind E1b 55k, and nucleo-cytoplasmic shuttling of the E4 34k-E1b 55k protein complex may additionally contribute to the regulation of p53 levels. It has been shown recently that both the Mdm2 and Hdm2 proteins regulate p53 protein levels by destabilizing the protein (48 -50) and that destabilization is dependent on those proteins' nucleo-cytoplasmic shuttling activities (51)(52)(53). It was proposed that Hdm2 and Mdm2 can induce degradation of p53 by directing p53 from the nucleus to a cytoplasmic proteasome, and a parallel may exist in the activity of the E4 34k-E1b55k complex in reducing p53 levels. All of the E4 34k mutant proteins retained the ability to physically interact with p53, confirming that physical interaction between E4 34k and p53 is not dependent on E1b 55k (5).
The data presented here demonstrate that several of the conserved cysteine and histidine residues found in E4 34k are essential for the formation of the E4 34k-E1b 55k complex. This is consistent with the hypothesis that the cysteine-and histidine-rich central region of E4 34k forms a zinc-based domain that participates in the physical interaction between E4 34k and E1b55k. Additional regions of E4 34k also have been implicated in interaction with E1b 55k. Most conclusively, both in in vitro protein blotting experiments and in co-immunoprecipitation experiments with extracts from transfected cells, E4 34k mutants that lack the amino-terminal 55 amino acids were unable to associate with E1b 55k, indicating that an element present in that region is essential for complex formation (24). Supporting that conclusion, the E4 ORF 6/7 protein, which contains only the amino-terminal 58 amino acids of E4 34k, forms a complex with E1b 55k in both blotting experiments and in co-immunoprecipitation experiments with transfected cell extracts. This suggests that the amino-terminal portion of E4 34k mediates E1b 55k binding independently of the remainder of the protein, although the interpretation of these data is complicated by the presence of ORF 7 sequences and by the apparent absence of the interaction in infected cells (16,24). Additionally, Cutt et al. (16) have described an E4 34k amino terminus-specific monoclonal antibody that recognizes only the noncomplexed form of E4 34k, as if the amino terminus of E4 34k is masked by participation in the physical interaction with the E1b 55k protein. The carboxyl terminus of E4 34k has also been implicated in formation of a functional E4 34k-E1b 55k complex by the recent observations that carboxyl-terminal deletion and amino acid substitution mutations abolish E4 34k function in lytic infection and prevent E4 34k-dependent localization of E1b 55k to the nucleus of cells. Experiments to test directly the effects of these mutations on the E4 34k-E1b 55k complex have not yet been reported. The carboxyl-terminal mutations were constructed specifically to disrupt a hypothesized arginine-faced amphipathic helix involving E4 34k residues 239 -255 and thus support the suggestion that the amphipathic helix is required for function or formation of the complex (25). The effects of these mutations on the E4 34k-E1b 55k complex might be due to a general effect on E4 34k structure. Alternatively, they may reflect direct participation of amino acids altered by the mutations, or of specific structures dependent on those residues, in complex formation. In the case of the potential E4 34k zinc-binding region, a correlation between zinc binding and E4 34k function would provide support for the involvement of a specific zinc-binding structure. The functional characterization of the E4 34k mutant proteins presented here provides a basis for studies that might establish such a correlation.