Exploiting features of adenovirus replication to support mammalian kinase production

Faced with the current wealth of genomic data, it is essential to have robust and reliable methods of converting DNA sequences into their functional gene products. We demonstrate here that when conditions are established that take advantage of the replication-associated virus ampliﬁcation, the virus-induced shutdown of host protein synthesis as well as the activation of signalling pathways that normally occur during virus replication, adenovirus biology can be exploited to generate a potent kinase expression system. Residual virus in the protein production has always been a limitation for adenovirus systems and we describe a DNA intercalator/ ultraviolet light treatment that eliminates residual adenovirus in protein preparations that has no deleterious effect on enzyme activity. The use of mammalian cells in combination with adenovirus generated a variety of active enzymes which could not be produced in Escherichia coli or baculovirus-infected insect cells. Thus, the utility of adenovirus-mediated enzyme expression as a versatile alternative to established protein production technologies is demonstrated.


INTRODUCTION
Amidst the current abundance of new genomic data, it becomes increasingly important to have reliable methods of converting these DNA sequences into their functional gene products. Only a small subset of the mammalian kinome, for example, (1) has been studied at the biochemical level and a full understanding of the biology of these enzymes will require methods to generate functional proteins. Currently, the two most popular methods of generating a protein product from a gene sequence are to introduce the sequence into the appropriate plasmid for production in bacteria or into a baculovirus vector for production in insect cells. However, as increasing numbers of the more exotic mammalian kinases are examined, it becomes clear that expression of these proteins in either prokaryotic or insect cells does not always result in soluble and functional enzymes. Some of these failures can be attributed to the requirement for mammalian speci®c chaperones, co-factors or post-translational modi®cation (e.g. phosphorylation, glycosylation) that are missing in the bacterial or insect cells. Thus, methods to overexpress and purify mammalian enzymes using mammalian cells can be of substantial value.
The use of adenovirus as a protein expression system in mammalian cells is not a recent development. Recombinant forms of adenovirus were some of the ®rst viruses used to direct exogenous protein expression in eukaryotic cells. Unfortunately, the initial experiments were performed in cells that (we now know) have limited amounts of the adenovirus receptor CAR, the expression levels were not strong (2) and early methods of recombinant adenovirus construction were cumbersome. In addition, although replication-defective adenoviruses pose no health risk, many regulatory agencies require that the viruses and any material containing residual virus (e.g. puri®ed proteins derived from the viruses) must be handled with appropriate containment. Elimination of adenovirus is dif®cult because of the high titers of virus produced, and because, unlike enveloped virus such as baculovirus, adenovirus is stable to detergents, chloroform and most conditions used to purify native proteins; thus residual levels of adenovirus can be found in the puri®ed protein. As a consequence, there has been only a limited use of adenovirus systems for recombinant enzyme production (see, for example, 3) and adenovirus use in this ®eld has been largely overshadowed by bacterial and baculovirus systems.
A variety of simpli®ed adenovirus construction techniques have now been developed that accelerate the construction of viruses. Because of the resulting wide availability of these vectors, as well as the problems that were being identi®ed with Escherichia coli and baculovirus systems, it seemed appropriate to re-examine adenovirus function for the production of recombinant enzymes in mammalian cells. We describe here optimized expression conditions combined with a simple onestep af®nity puri®cation to generate 100 mg to 10 mg quantities of enzymatically active mammalian protein kinases that could not be produced by either bacterial or baculovirus systems. An essential problem to solve, if this method is to be generally useful for enzyme production, is the problem of residual adenovirus in the puri®ed protein. Thus, we have identi®ed a DNA intercalator/UV irradiation method that eliminates residual replicating adenovirus in the recombinant enzyme preparations without affecting recombinant kinase activity. This method can serve as a valuable alternative for producing enzymes that require a mammalian system for activity.

Construction and puri®cation of recombinant adenoviruses
Recombinant E1/E3 defective adenoviruses were constructed using homologous recombination in E.coli (4). All viruses contained an expression cassette comprising the CMV immediate early promoter driving the open reading frame of interest followed by a rabbit b globin intron/polyadenylation signal. The expression cassette is inserted into an adenovirus type 5 genome in place of nucleotides 342±3523, which removes the E1 region. In addition the genomes contain a deletion of adenovirus 5 E3 region of nucleotides 28592±30472, a destruction of the SpeI site at nucleotide 27082 and the addition of SpeI linkers at each terminal repeat. Recombinant viral genomes were released from the plasmid backbone by digestion with SpeI, puri®ed by phenol/chloroform extraction and passage over a gel ®ltration column (Pharmacia Nick column). Virus cultures were initiated by transfecting the linearized genomes into 293 cells (5) using polyethylenimine (PEI) (6). After ampli®cation, virus was puri®ed by banding twice on CsCl gradients, transferred into 150 mM NaCl, 20 mM HEPES pH 7.5 (HBS)/40% glycerol by passage over a gel ®ltration column and stored at ±80°C as previously described (7). Viral quantitation was based on protein content using the conversion 1 mg viral protein equals 3.4 Q 10 12 virus particles (8).

Open reading frames
All open reading frames were generated by PCR from commercial cDNA preparations and cloned into derivatives of pPM7 followed by assembly into adenovirus (see above). The following cDNAs were used.
The full-length wild type RICK open reading contained a cDNA identical to GenBank accession number AF027706 with a carboxy terminal haemagglutinin (HA) epitope tag. Kinase essential lysine residue 47 was mutated to arginine (K47R) to generate a kinase inactive form of the protein. RICK deleted of CARD domain (RICKdC) was generated from either the kinase wt or kinase dead (K47R) genes and contained the ®rst 353 amino acid residues of RICK with either a carboxy terminal HA tag (RICKdC, wt or kr) or with a carboxy terminal strep-tag (RICKwtdCst, RICKkrdCst).
The RIP kinase domain construct (RIPkinwtST) contained a cDNA encoding residues 1±300 of the RIP protein identical to the sequence described in GenBank accession number NM_003804 plus a carboxy terminal strep-tag. The kinase dead version (RIPkinKRst) has the essential lysine 45 mutated to arginine (K45R) plus a carboxy terminal strep-tag.
The cdk9 construct contained a cDNA identical to GenBank accession number P50750, and NM_001261; a kinase dead version of the gene was prepared by mutating both kinase essental lysine 48 to arginine and the aspartic acid 167 to asparagine in cdk9KRDN. The cyclin T1 construct contained a cDNA identical to cyclin T1 (GenBank accession number XM_028819). Wild type cdk9, KRDN cdk9 and cyclin T1 were prepared with a carboxy terminal strep-tag.
The p38a construct contained a cDNA identical to GenBank accession number L35253; a kinase dead version of the gene was prepared by mutating lysine 53 to arginine.
The IKKbeta construct contained a cDNA identical to GenBank accession number AF080158 and AF031416.

Extraction of protein
Infected cells were harvested, washed twice with HBS and stored frozen at ±20°C after¯ash freezing in dry ice. The frozen cell pellet was thawed on ice with 2 ml lysis buffer 4 per 3 Q 10 7 cells (equal to one 185 cm 2¯a sk equivalent), pipetting gently to disperse cells. The lysis and extraction were performed on ice for 45 min with vortexing every 5±10 min. The material was sonicated in a bath sonicator for 5 min, an aliquot of the material was saved as the total sample. The material was then centrifuged for 10 min at 1950 g and the supernatant retained as extract. For cdk9/cyclin T1, infected cells were extracted directly after harvest without freezing, using LB7 instead of LB4.

Af®nity puri®cation
Streptavidin±Sepharose (Amersham Pharmacia) or streptactin±macropore (IBA) were washed with the same lysis buffer (at least 2 Q 10 resin vol). Washed resin was mixed with extract (100 ml of packed resin per 185 cm 2¯a sk equivalent of extract) for 60 min on ice with frequent gentle agitation to keep the resin dispersed. The resin plus extract were then centrifuged at 115 g for 3 min and the supernatant was saved as the¯owthrough (FT). The resin was washed with 2 Q 7 resin vol of lysis buffer 4 and 2 Q 7 resin vol NSBII and the resin was then eluted twice with 7 resin vol of NSBII+ 2.5 mM desthiobiotin (Eluate). The eluates were pooled and were stored in aliquots at ±80°C after¯ash freezing in dry ice.

Chromatographic puri®cation of RICKdC
Cell lysis was performed as above using LB5. Extracts were inactivated with psoralen (see below) and then fractionated over Red Sepharose (binding in LB5, elution with 300 mM NaCl). The salt content was diluted to 50 mM and the material was resolved on Q Sepharose using a NaCl gradient from 50 to 500 mM. Fractions were assayed for RICKdC content by western and for RICK autophosphorylation and histone 2b (H2b) or myelin basic protein (MBP) phosphorylation.

Psoralen inactivation
The method used conditions similar to those previously described (9). Lysates were transferred to Nunc 24-well cell culture plates (0.75±1.5 ml per well) and 1/200 vol of 33 mg/ml 8-methoxy psoralen (Sigma, M 3501, dissolved in DMSO) was added per well. The psoralen precipitate that formed was distributed in the sample by pipetting and the plate was irradiated with 365 nm UV light (plate on ice, lid on) for 30 min.

Analysis of virus content
Serial dilutions of test samples were performed in MEMalpha, 2% horse serum and added to 293 cells (40 000 cells/well, 24-well plate) for 2 h at 37°C. The medium was then changed to MEMalpha, 10% newborn calf serum. At 6±8 day post infection, cells were inspected microscopically for the presence of adenovirus (plaque formation, swollen, detached cells etc.) and then medium was removed, the cells were ®xed for 5 min (1% formaldehyde, 150 mM NaCl), washed once with PBS and then stained for 10 min with crystal violet (3% w/v in 150 mM NaCl), washed twice with PBS, once with distilled water and then scanned.

Kinase assays: RICK autophosphorylation and Flashplate assay
Recombinant enzyme was incubated in 40 ml containing 50 mM HEPES, pH 7.5, 1 mM MgCl 2 , 1 mM MnCl 2 , 1.5 mM ATP, 1 mCi [g-33 P]ATP/probe. For autophosphorylation analysis, the samples were then incubated with 0.1% SDS at 99°C and resolved by SDS±PAGE (10% gel), ®xed, stained and visualized by autoradiography. For Flashplate analysis, the samples included 20 ng/ml H2b. Reactions were performed in a Flashplate (basic, Perkin Elmer Life Sciences) well for 1 h at room temperature, the reaction was stopped by the addition of EDTA to 100 mM, washed four times with 100 mM EDTA, 100 mM sodium phosphate, pH 7.2, 0.01% Tween 20. Bound radioactivity was measured.

cdk9 assay
Kinase assay reactions were performed essentially as described in Grana et al. (10). Puri®ed enzyme was prediluted with buffer A (20 mM HEPES, pH 7.4, 4 mM MgCl 2 , 1 mM DTT) the reaction was initiated by adding 25 ml diluted enzyme to an equal volume of buffer A containing substrate [1 mM cold ATP, 1 mCi/well, 1 mg/well GST-CTDII (GST fused to a triple repeat of YSPTSPS)]. After 1 h incubation at room temperature, the assay was stopped by adding 5 ml of 0.5 M EDTA, pH 8.0, half of the reaction volume was applied to a MAIP ®lter plate according to the manufacturer's recommendations (Millipore). The dried plate was soaked with scintillation liquid, sealed with tape and incorporated radioactivity was measured employing a microbeta topcounter (Wallac Inc.).

Western analysis
Proteins were resolved on SDS±PAGE, and transferred to nitrocellulose. A total protein image was obtained after staining with Ponceau S. Blots were then blocked overnight in NET-gelatin, and either probed with the indicated primary antibody and was followed by the appropriate secondary antibody-horseradish peroxidase or blocked with 25 mg/ml avidin (Sigma) and then probed with streptactin±horseradish peroxidase (IBA). Finally, the images were developed using ECL reagents (Amersham).

Optimization of protein expression
Most uses of recombinant adenovirus apply the virus to cells that do not support virus replication. However, for high level protein production, it may be useful to employ conditions that take advantage of both virus ampli®cation and the inhibition of host translation that occurs during a productive adenovirus replication (11±13). Intuitively, if virus ampli®cation occurs, one would use a low virus to cell ratio and indeed, if ampli®cation and growth of the recombinant virus is the goal, then using more than 10±100 virus particles per cell (p/c) is wasteful, because no further ampli®cation of virus occurs at higher multiplicities of infection (m.o.i.). However, for protein production, the parameters are indeed different. The expression of a transgene protein, the kinase domain from RICK [Rip2, CARDIAK, a kinase involved in innate immune response signalling (14±16)] was monitored as a function of various virus to cell ratios (Fig. 1). In 293 cells, in which the defective adenovirus can replicate due to E1 complementation (17), the yield of transgene product was found to increase using 100±10 000 p/c (Fig. 1A). In a replication system, one would normally use 10±100 p/c for maximum yields of virus, higher m.o.i. are not re¯ected in greater virus yields. However for protein production, best yields were obtained at 1000 p/c and higher (Fig. 1A, lanes 4±6). In contrast, transduction of A549 cells that do not support virus replication demonstrated that the best yields were obtained at 3000 p/c and higher (Fig. 1B, lanes 5 and 6), most likely because of the lack of ampli®cation of the viral genome in these cells. A second important factor, host protein shut-off during replication, is apparent when comparing 293 and A549 cells. As an indication of this, prominent biotin-containing host proteins at~70 kDa are seen in both 293 and A549 uninfected cell extracts. In the extracts from 293 cells but not from A549 cells, the abundance of these proteins relative to the desired gene product declines dramatically at higher m.o.i., resulting in a higher product/total protein ratio. Thus the starting material for product puri®cation is already signi®cantly enriched by using the replicating system. A time course of expression demonstrated that maximum yield/cell was obtained at 48 h (results not shown). In conclusion, optimum transgene protein production is obtained using adenoviral replication conditions at relatively high m.o.i. (1000 p/c). Higher levels of transgene product are obtained at 10 000 p/c;

Af®nity puri®cation
Initial results demonstrated that signi®cant quantities of a recombinant protein can be generated. For puri®cation of such a protein, an af®nity puri®cation tag was included in the system. Some years ago, a peptide was selected for its ability to bind streptavidin in a reversible fashion, and these studies led to the development of the strep-tag system (18,19, reviewed in 20). To test this system in the context of recombinant adenovirus, the RICKdC open reading frame was ®tted with a C-terminal, nine amino acid strep-tag and cloned into an E1/E3 defective Ad5 genome. RICKdCst was expressed, extracted as described above, the extract was loaded on streptactin-macropore [an improved form of streptavidin (21,22)] washed extensively and then the strep-tagged protein was eluted with desthiobiotin. Both total protein staining ( Fig. 2A, left panel) and staining with HRP-streptactin ( Fig. 2A, right panel) revealed that a single protein band of the expected molecular mass (41 kDa) was present in the total (tot) and extracted (ext) fractions, depleted from the non-bound FT fraction and eluted from the resin with 2.5 mM desthiobiotin ( Fig. 2A, EL). The eluted fraction demonstrated RICK kinase activity as measured by autophosphosphorylation (Fig. 3B) and a quantitative analysis of phosphorylation of H2b demonstrates that the kinase activity of the strep-tag puri®ed material was substantially more active than the same molecule puri®ed by ion exchange chromatography (Fig. 7B, compare RICKdC MQ puri®ed by Red and MonoQ Sepharose to RICKdcST puri®ed by streptactin±Sepharose).
The broad utility of the method is demonstrated by obtaining similar expression and puri®cation patterns with additional mammalian kinases including the kinase domain of the death signalling kinase RIP (Fig. 2B, but see below); the cyclinT1/cdk9 complex ( Fig. 2C and below); NF-kB activating kinase IKKb (Fig. 2D) and the MAP kinase p38a (Fig. 5A). A Coomassie blue-stained gel demonstrating the purity of a collection of these kinases is shown in Figure 2E.
The generation of CsCl-puri®ed adenovirus to be used as inoculum requires time and effort and as one reviewer pointed out, the procedure would be simpli®ed if crude viral lysates could be used. We compared the yield of puri®ed kinases obtained with pure virus to that obtained using comparable quantities of crude infected cell lysates. Because it is dif®cult to accurately measure viral concentrations in crude lysates, a broad range of inoculum volumes was tested calculated to span the 100±1000 particle/cell window. As shown in Figure 2F, puri®ed RICKdC and p38a protein can be obtained using the unfractionated lysates as inoculum (RICK, lanes 2± 5, p38a lanes 7±10); however, the yield of protein is slightly lower than that obtained with pure virus inoculum. (RICK lane 1; p38a, lane 6; per 185 cm 2¯a sks: RICKdC pure 36 mg; RICKdC crude lysate: 20 mg; p38a pure virus: 28 mg; p38a crude lysate: 20 mg). Furthermore, the resulting puri®ed protein has a slightly different pattern of co-puri®ed contaminants most likely due to the presence of high concentrations of viral capsid proteins added with the crude inoculum. Thus, the use of puri®ed virus inoculum results in slightly higher yields and purity. However, for many proteins, the yields may still be suf®ciently high to justify omitting the virus puri®cation. That no increase in pure protein yield occurs with increasing crude lysate inoculum is likely to be due to limiting toxicity caused by free viral capsid components in the crude lysate.  The question of co-purifying eukaryotic kinases For kinase production, a major consideration with the use of mammalian or insect systems is the issue of co-purifying cellular kinases. Because E.coli lacks classical serine/ threonine and tyrosine kinases, there is little concern about copurifying a host kinase with the target kinase. However, with mammalian kinases overexpressed in eukaryotic cells, there is the possibility that either speci®c or non-speci®c interactions of the recombinant enzyme with cellular factors can lead to the copuri®cation of proteins in addition to the target enyzme. This is especially important when producing kinases or other enzymes for drug screening efforts in which a contaminating enzyme activity can complicate the search for speci®c inhibitors. To demonstrate enzyme purity, a mutated version of the protein is prepared with the kinase activity genetically disrupted by mutating either the highly conserved lysine residue that forms part of the ATP binding pocket, or by mutating the aspartic acid residue in the kinase motif (23). The mutant protein is then expressed and subjected to the same level of puri®cation. A measurement of residual kinase activity in such a preparation indicates the quantity of co-purifying cellular kinases.
Such an analysis was performed with the wild type RICK kinase domain and a kinase inactive RICK K47R mutant. Both proteins were expressed using adenovirus, puri®ed (Fig. 3A) and equal protein quantities were tested for kinase activity (Fig. 3B). Substantial autophosphorylation and phosphorylation of a substrate (MBP) is observed with the wild type RICK kinase domain, no detectable kinase activity was found , lysed, extracted and puri®ed by streptactin±macropore as described in Materials and Methods. Aliquots of the total material before extraction (tot), the extract (ext), unbound material after expsoure to resin (FT) and the material eluted with desthiobiotin (EL) were mixed with SDS application buffer, heated at 99°C for 10 min, sonicated for 5 min in a sonicating water bath and 20 ml was loaded per well on a 10% acrylamide SDS gel. After resolution, the gel was transferred to nitrocellulose, stained with Ponceau S (A±D, left panels) or blocked with NET/gelatin and avidin, probed with streptactin±horseradish peroxidase (IBA) and visualized by ECL (A±D, right panels). (E) Equal aliquots of puri®ed proteins were resolved on a 10% acrylamide SDS gel and visualized by staining with colloidal Coomassie blue (Sigma). (F) Comparison of pure virus to crude lysate inocula. 293 cells were infected with either 1000 particles/cell of CsCl-puri®ed adenovirus encoding RICKdCst or p38aST or with crude infected cell lysates of cells infected with the same adenoviruses, at 30, 100, 300 or 1000 particles/cell. At 48 h post-infection, cells were harvested, lysed and the strep-tagged protein puri®ed as described above. Equal volumes of each of the eluate fractions were resolved on a 10% acrylamide SDS gel, transferred to nitrocellulose, stained for total protein with Ponceau S (upper panels) and then probed with streptactin±horseradish peroxidase (IBA) and visualized by ECL (lower panels).

PAGE 5 OF 11
Nucleic Acids Research, 2003, Vol. 31, No. 21 e128 associated with the RICK K47R preparation (Fig. 3B, compare lanes 1 and 2) demonstrating the absence of non-RICK kinases in the preparations. The kinase domain from RIP was also prepared in both wild type and kinase inactive forms (Figs  2B and 3A). Although the wild type protein domain showed both autophosphorylation and kinase activity against MBP, (Fig. 3B, lane 3) substantial kinase activity was also found associated with the preparation of kinase inactive RIP (RIP K45R Fig. 3B, lane 4), demonstrating the presence of kinase activities in addition to the RIP molecule. Indeed, the total protein-staining pattern of the puri®ed RIP preparations revealed a number of additional protein species (Figs 2E,  lanes 3 and 4, and 3A, lanes 3 and 4) which were examined in more detail. The chaperone protein cdc37 interacts with a number of kinases and facilitates, in complexes with hsp90, the appropriate folding of molecules including Raf (24,25), Akt (26), the IKK complex (27), cdk9/cycT1 (28), cdk4 and cdk6 (29). Puri®ed RIP contained quantities of 40 and 90 kDa proteins, western analysis demonstrated that indeed the puri®ed RIP (but not similarly puri®ed RICK) contained both cdc37 and hsp90 (Fig. 4A). Because both RICK and RIP preparations contained detectable quantities of the inducible (hsp70) and constitutive (hsc70) 70 kDa heat shock protein (Fig. 4A), this family of chaperones could not be the basis for the RIP contamination problem. However, cdc37/hsp90 can form multimeric complexes with a number of additional kinases; we tested for their presence in the RIP preparations. Indeed, cdk4 and raf1, two known cdc37 binding kinases (24,25,29), were present in the puri®ed RIPwt and RIPK45R preparations but not in similarly puri®ed RICK preparations (Fig. 4B). Thus, the presence of the kinases cdk4 and raf1 and perhaps additional cdc37 binding kinases could account for the kinase activity observed with puri®ed, kinase-dead RIP K45R. Fortunately, this phenomenon of co-purifying kinases is rather the exception than the rule. To date, all other kinases that we have puri®ed using these methods have shown negligible kinase activity associated with the kinase dead mutant form (see below for cdk9 and p38a).

The puri®cation of enzyme complexes
For some eukaryotic kinases, it is essential to generate and purify multi-protein complexes. For example, the cellular kinase cdk9 requires the interaction with cyclin T1 for activity as the transcription elongation factor P-TEFb (30,31). Because maturation of cdk9 activity requires interaction of cdk9 with the cyclin T1 during synthesis and folding (28), the two components of the cdk9/cyclinT1 kinase complex were expressed as af®nity tagged proteins by co-infection with adenoviruses and puri®ed as described above (Fig. 2C). Both cdk9 and cyclinT bear af®nity tags, thus it was important to demonstrate that other kinases reported to interact with cyclin T1 are not in the preparation. Immunoblot analysis demonstrated the absence of cdk2 and cdk4 and the puri®ed cdk9/ cycT1 kinase activity was not inhibited by established cdk2 and 4 inhibitors (results not shown). The kinase dead mutant of cdk9 with the essential lysine and aspartic acid residues mutated to arginine and asparagine (K48R, D167N), was expressed with cycT1, puri®ed to the same extent as the wild type kinase/cycT1 complex and assayed for kinase activity. No detectable kinase activity was present in the cdk9K48RD167N mutant protein preparation (Fig. 4C) even when the sample was assayed at 10-fold higher concentration than the wild type cdk9 (compare 0.3 ml of wt with 3 ml of KRDN). Thus, although the potential exists for cyclin T1 to recruit and copurify other cyclin T1 binding kinases, the high level of target protein expression, the shut down of host protein expression by adenovirus and the puri®cation method, results in a cdk9/cyclin T1 preparation that is free of other cyclinT1 binding kinases.

A comparison with baculovirus
cdk9/cyclin T1 can be generated using baculovirus infection of insect cells (30). To compare the adenovirus and baculovirus system, cdk9/cyclinT1 was expressed and puri®ed from SF9 cells and the kinase activity was compared with the adenovirus-produced material. With the adenovirus system, we typically obtain 10 mg of pure cdk9/cyclinT1 per 3 Q 10 7 cells. In our hands, to obtain a similar mass of the kinase with  the baculovirus system required three times more cells, and the resulting puri®ed material had one-®fth to one-tenth the kinase activity as the adenovirus-derived material (compare Fig. 4D,  lanes 1 and 3).

Activation of kinases
For a number of kinases involved in signalling cascades, care must taken to ensure that the kinase is appropriately activated. This is especially well established for components of the MAP kinase cascade where the recombinant enzyme must be either genetically altered with an acidic amino acid residue, mimicking phosphorylation activation, or treated with an activated upstream kinase. For example, the map kinase p38a is normally produced in E.coli and activated using MKK3 or MKK6, which must also be prepared in an active form (32,33). The use of adenovirus in mammalian cells allows one to take advantage of endogenous signalling pathways to generate active forms of the enzyme. The p38 pathway is activated during adenovirus infection (34,35), thus the potential exists that the process of infection and replication may generate active enzyme without additional manipulations. To test this idea, p38a was expressed using this system. Initially, equal quantities of cells were used and one portion was stimulated with anisomycin before harvest in order to deliberately activate the p38 pathway (36,37). The protein was then purifed from the two extracts and examined for p38 kinase activity.
One measure of p38 activation is the hallmark activation phosphorylation on threonine 180 and tyrosine 182, (38,39) which can be readily measured using phospho-speci®c antibodies to these sites. The p38a generated using the adenovirus system was found to be already in a phosphorylated form (Fig. 5A, lane 4). Deliberate activation of p38 with anisomycin lead to no greater level of the p38 phosphorylation (Fig. 5A, lane 8). A more direct assay of p38 activation was performed by measuring the ability of the puri®ed kinase to phosphorylate the transcription factor ATF-2. p38a puri®ed from non-stimulated cells was already active in phosphorylating ATF-2 (Fig. 5B, lane 1±3) and no further increase in activity was obtained by treating the cells with anisomycin before harvest (Fig. 5B, lanes 5±7) consistent with the results obtained by examining the phosphorylation state of p38 (Fig. 5A, bottom panel). These results are consistent with the reports that p38 is activated during the course of adenovirus infection and replication (34,35); thus no further manipulation of this pathway is required to obtain active p38.
The speci®city of the p38a kinase activity observed here is demonstrated by the inhibition of the phosphorylation with the p38 inhibitor SB203580 (Fig. 5B, lane 4 and 8).
As a control to demonstrate the purity of the p38 preparation, a kinase inactive mutant p38a was generated (p38aKR) expressed and puri®ed in the same manner (Fig. 5C,  bottom panel). When tested for activity, the kinase inactive mutant indeed showed no activity [Fig. 5C, compare lanes 5±7 (kinase inactive) to lanes 1±3 (kinase active)] demonstrating that the activity observed with the wild type enzyme preparation can be attributed to p38a and not co-purifying mammalian kinases.

The question of contaminating adenovirus
Although the recombinant adenoviruses used here are replication-defective and pose no signi®cant health threat, in most countries, recombinant adenovirus and all material derived from these viruses must be handled with appropriate containment. One can calculate that a typical lysate produced under the conditions described contains 10 10 to 10 12 virus particles per milliliter and even the extensive puri®cation allowed by af®nity chromatography is not suf®cient to remove virus to <1 in 10 12 /ml. Thus, we sought a method of chemically inactivating the adenovirus present in the protein preparations.
It was previously demonstrated that 8-methoxy psoralen has the ability to enter the adenovirus capsid and upon irradiation with long wavelength UV, the compound form single-and double-stranded crosslinks of the adenoviral genome that block both viral transcription and replication (9,40,41). This provides a method of inactivating adenovirus that is accepted by clinical standards (see for example 42). Psoralen inactivation can be used to inactivate baculovirus in a recombinant protein preparation without altering the antigenic properties of the protein (43) and the treatment does not impair the entry function of the adenovirus capsid (9,40,41). However, a direct analysis of the consequences of psoralen/UV treatment on enzyme activity has not yet been performed.
Cytopathic effect assays were performed to assess the quantity of adenovirus remaining in puri®ed kinase preparations. In this assay, surviving (non-infected) cells stain with crystal violet, the presence of replicating adenovirus results in cell death and detachment from the substrate (referred to as cytopathic effect, CPE) and loss of crystal violet staining; clear areas in the monolayer indicate the presence of replicating adenovirus. An analysis of virus content of a puri®ed kinase preparation shows complete killing of the cells with 50 and 5 ml of the material, (Fig. 6, row B) with substantial CPE observed microscopically at 0.05 and 0.005 ml. In comparison with a pure adenovirus standard (Fig. 6, row A) it can be calculated that the puri®ed kinase preparation contains between 10 4 and 10 5 virus particles/ml. This is not surprising as the starting extract contains~10 9 viral particles/ ml (results not shown). However, if the lysate used as the starting material for kinase preparation is treated with psoralen/UV light before puri®cation, the replicating adenovirus content is eliminated. Treatment with 15 min of UV largely eliminates (row C, cytopathic effect only detectable microscopically) and exposure for 30 min of UV light combined with 8-methoxy psoralen completely eliminates all trace of virus replication (Fig. 4D). More sensitive plaque  assays (not shown) and RT±PCR (9) have demonstrated the elimination of viral replication and transcription with this method.
It is essential to demonstrate that this method does not impair the function of the recombinant enzyme. The 30 min exposure to psoralen/UV had no deleterious effect on the yield and purity of RICKdC with the similar enzyme yield and purity obtained from control extracts (Fig. 7A, lanes 1 and 2) and psoralen/UV treated extracts (Fig. 7A, lanes 3 and 4). The psoralen/UV had no detectable in¯uence on RICK kinase activity (Fig. 7B, compare control, lanes 4±6, and psoralen/ UV treated, lanes 7±9). Similar results were obtained with a variety of other mammalian kinases including cdk9/cyclinT1 and p38a (results not shown, but note that Fig. 4C uses psoralen-treated cdk9/cyclinT1).

DISCUSSION
We report here conditions that take advantage of the biology of a commonly available adenovirus vector for the production of recombinant kinases. The method uses mammalian cells for expression, and this may be important if post-translational modi®cation, host-speci®c folding or activation of the enzyme by upstream kinases is required. A simple af®nity puri®cation method allows the enzyme to be puri®ed in a single step, and an important part of this method is the chemical inactivation of residual adenovirus in the puri®ed protein using a UV activated DNA intercalating agent, an important issue for the use of these proteins in other application. The method provides useful quantities of enzymatically active kinases when other expression systems give poor results (e.g. for RICK and cdk9/cyclin T1). Although the focus here has been on kinases, it is likely that the system will have applications with a variety of other mammalian enzymes.
Although the yields of puri®ed protein will vary from protein to protein, some general comments on the yields obtained with this method can be made. The preparation is Figure 6. Assay of replicating adenovirus content before and after psoralen/UV treatment. 293 cells (40 000 cells/well, 24-well plate) were exposed to the indicated particles/cell m.o.i. of pure adenovirus (row A) or 50, 5, 0.5, 0.05 and 0.005 ml of pure RICKwtdCst kinase (El1 fraction from streptavidin± Sepharose fraction, (row B) or the same quantity of kinase puri®ed from lysates exposed to 8-methoxy psoralen/265 nm UV light for 15 (row C) or 30 min (row D). At 7 days post-infection the cells were ®xed with formaldehyde and stained with crystal violet. See Materials and Methods for details. Figure 7. RICK kinase activity is not impaired by psoralen/UV treatment. RICKwtdCst was puri®ed from control lysate or lysate treated for 30 min with psoralen/UV. Equal protein quantities were resolved on 10% acrylamide SDS gel and stained for total protein (Coomassie blue, A). In parallel, samples were incubated with [g-33 P]ATP under kinase appropriate conditions (see Materials and Methods) and the radioactivity incorporated into a histone H2b substrate was measured (B).

PAGE 9 OF 11
Nucleic Acids Research, 2003, Vol. 31, No. 21 e128 conveniently performed with material derived from 18, 185 cm 2 tissue culture¯asks of 293 cells (~3 Q 10 7 cells/ ask), yielding 36 ml of extract and 12 ml of material (0.04± 0.2 mg/ml) eluted from streptavidin±Sepharose. Thus a typical preparation would yield 0.5±2 mg of puri®ed RICK. Perhaps a more critical parameter is the number of enzymatic assay points provided. In our applications, a 185 cm 2¯a sk yields 0.2 ml (cdk9/cyclin T1) to 2 ml (RICK) of puri®ed protein with 1 ml suf®cient for a high throughput, radioactive assay point. Thus, 200±2000 assay points can be obtained from 185 cm 2¯a sk.
A critical issue is the time involved in the construction and the puri®cation of the adenovirus vectors. It currently takes 6±8 weeks to generate CsCl puri®ed quantities of a novel adenovirus. This is substantially slower than E.coli systems, but comparable to baculovirus. The methods described here used CsCl-puri®ed virus for the infections to generate kinase primarily because in establishing the methods it was essential to use known quantities of the adenovirus. It is also possible to use non-puri®ed lysates for the kinase production infections (Fig. 2F) and this would save several weeks. One should point out that the same adenovirus preparations described here can also be used for biological experiments, thus providing additional justi®cation for the construction and puri®cation of these reagents.
In summary, the technique describe here provides a useful system for obtaining native, active, puri®ed enzymes from mammalian cells.