INTRODUCTION

Many extracellular stimuli transmit signals into eukaryotic cells in part through a conserved signal transduction mechanism, a mitogen-activated protein (MAP)¹ kinase, or extracellular signal-regulated kinase (ERK) pathway. MAP kinase pathways are important pleiotropic signaling enzymes that regulate cytoplasmic enzymes, cytoskeletal processes, membrane proteins, gene activity, and other intracellular events. Although upstream regulatory mechanisms may vary, the overall organization of a MAP kinase pathway, which includes a three-kinase cascade, is conserved in evolution from yeast to metazoans; a MAP/ERK kinase kinase (MEKK) activates a MAP/ERK kinase (MEK), which activates an ERK or MAP kinase.

Currently, several different mammalian MAP kinase cascades have been defined, and three are relatively well characterized. One of them, the ERK module, is activated in response to stimuli that induce proliferation and differentiation. The c-Jun N-terminal kinase/stress-activated protein kinase (JNK/SAPK) and p38 MAP kinase modules are activated in response to environmental stresses such as osmotic shock or UV irradiation (reviewed in Refs. 1 and 2). Distinct MEK isoforms lie upstream in each cascade. MEK1 and -2 activate ERK1 and -2; MEK3 and -6 activate p38; MEK4 (also known as JNKK or SEK1) activates JNK/SAPK. MEKKs are a diverse group that include Raf isoforms, which are believed to act exclusively on MEK1 and MEK2, and MEKK1, which can activate MEK isoforms in all three modules.

One of the obstacles in studying the biochemical properties of individual MAP kinase modules and their constituents is the difficulty in obtaining sufficient quantities of pure active components. While most MAP kinases can be easily expressed in bacteria, they must be activated by corresponding MEKs, which must also be expressed and activated. Mutants with increased kinase activity have been described for some but not all MEK isoforms. Much less success has been achieved in activating MAP kinases, themselves, by mutagenesis (1, 3).

In this study, we coexpressed different components of the three best characterized MAP kinase modules in bacteria. We show that, in each case, if the upstream component is present in constitutively active form, coexpression results in the activation of the downstream component. Expression of a C-terminal, constitutively active fragment of MEKK1 allows us to activate MEK1, MEK2, MEK3, and MEK4 from all three modules and subsequently to activate the downstream MAP kinases. Coexpression of kinase and substrate provides a convenient method to generate highly purified milligram amounts of activated MAP kinases as well as other phosphoproteins for biochemical and structural studies.

MATERIALS AND METHODS

cDNAs encoding constitutively active MAP kinase cascade members were subcloned under the control of an isopropyl--D-thiogalactopyranoside-inducible Tac promoter into the plasmid pBB131 (Fig. 1B). pBB131, a low copy plasmid with a p15A origin of replication and a kanamycin resistance marker (4), was kindly provided by Dr. J. Gordon (Washington University, St. Louis, MO).

To subclone the constitutively active mutant of human MEK1 (kindly provided by Dr. Natalie Ahn, University of Colorado at Boulder) into pBB131, an NcoI restriction site was introduced at the start ATG and an MluI site after the stop codon by polymerase chain reaction (PCR) with specific primers (5-GGGCCATGGCCAAGAAGAAGC-3 and 5-CAAACGCGTCATGATCAACCACCGG-3). The PCR product was digested with MluI and partially with NcoI and ligated into pBB131 digested with the same restriction enzymes. To subclone the catalytic C-terminal fragment of rat MEKK1 (MEKK-C) (5) into pBB131, an NcoI-EcoRI fragment of MEKK containing its catalytic domain was ligated into pBB131 digested with the same enzymes.

To express MEKs and ERKs in one plasmid, MEK1 was modified as follows. Using PCR with the specific primers (MEK1, 5-CCCGCTAGCCATATGCCCAAGAAGAAGCCG-3 and 5-CTCTTTGCATATGGGTACCTCCTTAGACGCCAGCAGCATGGGT-3), an NdeI restriction site was introduced at the MEK starting ATG, and a ribosomal binding site was introduced after the MEK stop codon followed by a second NdeI site. A similar approach was used for MEK2. The resulting PCR products were inserted into pET His₆ERK2, which contains rat ERK2 (6) at the NdeI site. pET His₆ERK2 is a derivative of pET-His₆TEV (Life Technologies, Inc.) from which the Tev protease cleavage site has been deleted. The resulting plasmids direct expression of untagged MEK1 or MEK2 and ERK2 with a His₆ tag (Fig. 1B).

For the expression of human MEK4 (also known as JNKK or SEK1; Refs. 7 and 8) and rat -SAPK (9) from the same plasmid, a ribosomal binding site and then an NcoI site were introduced immediately after the -SAPK stop codon using PCR with the T7 primer and a specific primer, 5-CCAGCCATGGTCTCCTTTCACAGACAAGTGCGCCATCTGCGAGGTTT-3. The resulting PCR product was digested with NcoI and ligated into NpT7-5 containing a full-length MEK4 cDNA (kindly provided by M. Karin, University of California San Diego) digested with the same enzyme. The resulting plasmid directed expression of untagged MEK4 and -SAPK with a His₆ tag (Fig. 1B).

For the expression of MEK4 and murine p38 from the same plasmid, MEK4 was excised from pGEX-KG with NcoI and HindIII and cloned into pT7-5 (10). A ribosomal binding site and an NcoI site were introduced after the p38 stop codon (11) by PCR with the T7 primer and a specific primer, 5-CCAGCCATGGTCTCCTTTCAGGACTCCATTTCTTCTTGGTCAAGGGG-3, using pET14b-p38 as template. The PCR product was cleaved with NcoI and ligated into pT7-5 containing MEK4 that had been digested with NcoI.

The resulting plasmids, one encoding kanamycin resistance with an upstream constitutively active kinase and the other encoding ampicillin resistance with downstream effector kinase(s), were transformed into Escherichia coli strain BL21(DE3). Transformants were plated on Luria broth plates containing 0.1 mg/ml of both carbenicillin and kanamycin. From each plate, a minimum of five colonies were selected for analysis. Expression was induced with 0.25 mM isopropyl--D-thiogalactopyranoside (final concentration) for 7-12 h at 30 °C before harvesting. Cells were lysed by sonication in 0.5 ml of buffer C (see below) and clarified by sedimentation for 15 min at 15,000 × g at 4 °C. Protein kinase activity in the lysates was assayed using the following substrates: myelin basic protein for ERKs, GST-c-Jun1-223 for -SAPK, GST-ATF21-254 for p38 MAP kinase, K52R ERK2 for MEK1 and MEK2, p38 for MEK3, and GST-SAPK K55A for MEK4.

His₆-tagged proteins were expressed and purified essentially as described previously (5) with the following modifications. Expression of recombinant protein was induced with 0.25 mM isopropyl--D-thiogalactopyranoside at 30 °C for 12-14 h. Cells were harvested and then either frozen in liquid nitrogen and stored at 80 °C or lysed in buffer A (50 mM sodium phosphate, pH 8.0, 0.3 M NaCl, 0.5 mM phenylmethylsulfonyl fluoride, 1 µg/ml leupeptin, 1 µM pepstatin, 1 mM benzamidine). The lysate was clarified at 27,000 rpm for 30 min at 4 °C in a Ti35 rotor (Beckman, Palo Alto, CA). The resulting supernatant was applied by gravity flow to a 1-2-ml column of Ni²⁺-NTA resin (Qiagen Inc., Chatsworth, CA) equilibrated in buffer A. The resin was washed with 20 volumes of buffer B (20 mM Tris-HCl, pH 8.0, 0.3 M NaCl, 10 mM imidazole) and eluted with 0.25 M imidazole, pH 7.0, in buffer B. 1-ml fractions were collected, and aliquots were analyzed by SDS-PAGE. Fractions containing recombinant protein were pooled and dialyzed overnight against buffer C (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM dithiothreitol, 1 mM EGTA, 10 mM benzamidine, 0.2 µM pepstatin, 0.5 mM phenylmethylsulfonyl fluoride, and 20% glycerol). For fast protein liquid chromatography purification, proteins were diluted 2-fold with buffer C without glycerol, applied to a Mono Q HR5/5 column equilibrated in the same buffer, and eluted with 500 mM NaCl in buffer C. Fractions containing purified MAP kinases were pooled, dialyzed against buffer C, frozen in liquid nitrogen, and stored at 80 °C in small aliquots. GST-tagged proteins were expressed as above and purified on reduced glutathione-Sepharose (Sigma) as described previously (12).

Protein concentration was determined using the Bio-Rad dye reagent diluted according to the manufacturer's recommendations with bovine serum albumin as a standard. Protein kinases and substrates were incubated in 30 µl of 20 mM HEPES, pH 7.3, 10 mM MgCl₂, 1 mM benzamidine, 1 mM dithiothreitol, and 100-200 µM ATP ([-³²P]ATP to 10-30 cpm/fmol) for 15 or 30 min at 30 °C. Serial dilutions of active enzymes were tested to guarantee that assays were in the linear range. Reactions were stopped by either 1) the addition of 10 mg of bovine serum albumin and 1 ml of 10% trichloroacetic acid followed by collection of precipitates on glass filters or 2) the addition of 7.5 µl of 5 × Laemmli sample buffer followed by heating for 5 min at 100 °C for analysis by SDS-PAGE.

Antisera raised against a C-terminal peptide of MEKK1, MEK1 (A2227), and MEK2 (A2228) were described previously (5). Antibodies to MEK3 were raised against a GST-MEK3 fusion protein (cDNA kindly provided by Dr. K. Guan, University of Michigan). Antibodies against MEK4 proteins were elicited with the synthetic peptide KRKALKLNFANPPVKSTART (residues 35-54 of MEK4), coupled to hemocyanin using methods described earlier (13). Antisera specific to phosphorylated forms of ERKs and JNK/SAPK were from Promega Corp.

The catalytic subunit of protein phosphatase 2A (PP2A) was purified as described (14). A plasmid that encoded a truncated form of a protein-tyrosine phosphatase (PTP1) fused with GST was kindly provided by Dr. Guan. GST-PTP1 was expressed in E. coli, purified, and cleaved with thrombin, and PTP1 was separated from GST as described (15).

For inactivation with phosphatases, purified active ERK2 (1.7 µg/ml), p38 (1.9 µg/ml), or -SAPK (1.65 µg/ml) was treated with 3 µg/ml PP2A in 20 mM Tris-HCl, pH 7.4, 2 mM EGTA, 2 mM dithiothreitol, 2 mM MgCl₂, 0.5 mg/ml bovine serum albumin, 0.05% Brij-35, 5 µg/ml each of aprotinin and leupeptin, and 100 ng/ml phenylmethylsulfonyl fluoride or with 7.4 µg/ml PTP1 in 50 mM Hepes, pH 7.5, and 0.1% -mercaptoethanol for the indicated times at 30 °C. For kinase assays, dephosphorylation reactions were terminated by the addition of okadaic acid or vanadate to final concentrations of 1 µM and 5 mM, respectively. Alternatively, for immunoblotting (ECL kit, Amersham Life Science, Inc.), reactions were terminated with electrophoresis sample buffer.

RESULTS

It was shown previously that in eukaryotic cells ERK1 and ERK2 may be activated by both of the two closely related kinases, MEK1 and MEK2 (16, 17). This activation results in phosphorylation of two closely positioned residues between subdomains VII and VIII of ERKs, Thr²⁰² and Tyr²⁰⁴ in ERK1 and Thr¹⁸³ and Tyr¹⁸⁵ in ERK2 (13, 18-20). MEKs also require two phosphorylations, both on serine or threonine, for high activity; thus, to activate ERKs efficiently, MEKs must be activated by phosphorylation by an upstream kinase (Refs. 18 and 19; Fig. 1A).

Constitutively active mutants of MEK1 and MEK2 have been described that have several hundred-fold elevated activity toward ERKs (3). In the absence of their own phosphorylation, these mutants can activate ERK1 and ERK2 in vitro or in intact cells (3, 21). We coexpressed one of these highly activated MEK1 mutants, MEK1 N3 S218D S222D, (referred to hereafter as MEK1-R4F) with wild-type ERK1 or ERK2 in E. coli to see if we could produce activated ERKs. To accomplish this, we initially employed the two-plasmid system used for expression of N-myristoylated forms of G-protein -subunits in E. coli (Ref. 4; Fig. 1B). Such coexpression of MEK1-R4F with either ERK1 or ERK2 results in their activation (data not shown).

We tested methods to increase the amount of active ERK produced by our coexpression system. A single-plasmid system expressing both MEK1-R4F and ERK2 from a single bicistronic mRNA under the control of the bacteriophage T7 promoter yielded nearly complete conversion of ERK2 to the reduced mobility, highly phosphorylated form. The proportion of active ERK based on shifted mobility varied from 10% to close to 100%, depending on growth medium, conditions, and time of induction (data not shown). The specific activity toward myelin basic protein of ERK2 obtained from this single plasmid system and purified on Ni²⁺-NTA-agarose and then on Mono Q varied from 1.6 to 2.3 µmol/min/mg protein in three ERK2 preparations. To demonstrate that ERK2 expressed in this system is doubly phosphorylated on Tyr and Thr, we treated the purified active kinase with protein phosphatases selective for either the phosphotyrosine (PTP1) or the phosphothreonine (PP2A) residue. As shown in Fig. 2A, treatment with either phosphatase, but not with phosphatases pretreated with the corresponding inhibitors (vanadate and okadaic acid), results in loss of ERK2 activity.

To characterize the coexpressed ERK proteins further, we used an antiserum raised against a doubly phosphorylated peptide from the ERK phosphorylation lip. The epitope was a 16-amino acid sequence encompassing phosphothreonine and phosphotyrosine residues that are required for ERK activation. This antibody does not recognize 10-20 ng of ERK2 expressed in the absence of an activating MEK (Fig. 2B). In contrast, if ERK2 has been coexpressed with MEK1-R4F, as little as 1-2 ng are recognized by this antibody (Fig. 2B). Treatment with PTP1 or PP2A greatly reduced ERK2 recognition in a time-dependent manner (Fig. 2C). Treatment with both phosphatases together nearly abolished recognition of 25 ng of ERK2 by the anti-active ERK antiserum (Fig. 2B). In some experiments, three bands corresponding to ERK2 with different electrophoretic mobilities could be resolved after phosphatase treatment: one corresponding to inactive, one to active, and the intermediate to inactive but singly phosphorylated forms. To generate the fastest migrating band, treatment with both phosphatases was required.

We next employed the coexpression system to activate the other components of the ERK module, MEK1 and MEK2 (1). To achieve this, we used the C-terminal constitutively active fragment of rat MEKK1, MEKK-C, shown previously to activate MEKs in vitro and in transfected eukaryotic cells (5, 22, 23). MEKK-C was subcloned into pBB131 and coexpressed together with histidine-tagged wild-type human MEK1 and MEK2. As shown in Fig. 3A, neither MEK1 nor MEK2 expressed in the absence of an activator in bacteria phosphorylated ERK2 in vitro. However, coexpression of MEK1 or MEK2 with MEKK-C greatly increased the capacity of bacterial lysates containing them to phosphorylate ERK2. Activated forms of MEK1 and MEK2 with reduced electrophoretic mobility were detected on Western blots probed with specific anti-MEK1 and anti-MEK2 antisera (Fig. 3A, lower panel). Treatment with PP2A alone, but not with PP2A pretreated with okadaic acid, resulted in the disappearance of the upper band (Fig. 3A, lower panel).

The specific activities of coexpressed, histidine-tagged MEK1 and MEK2 after purification on Ni²⁺-NTA-agarose were in the range of 3-10 nmol/min/mg in three different MEK preparations (Fig. 4); however, MEKs purified by this single step are relatively impure (20% or less, data not shown). Nevertheless, we were able to achieve an average of 20-fold activation of MEK1 and 150-fold activation of MEK2 using our coexpression system (Fig. 4).

MEK3 and MEK4 belong to different MAP kinase modules (Fig. 1A) and are activated by different stimuli (7, 8). Because MEKK1 can activate these other modules after cotransfection into eukaryotic cells (5, 23, 24), we wished to determine if MEK3 and MEK4 can be activated in E. coli by coexpression with MEKK-C. Histidine-tagged MEK3 and GST-MEK4 were purified from bacterial cells expressing these kinases together with MEKK-C and assayed for their ability to phosphorylate SAPK and p38 MAP kinases. Coexpression of MEKK-C with MEK3 results in the appearance of an activity capable of phosphorylating p38 MAP kinase, a MEK3 substrate. Coexpression of MEK4 with MEKK-C results in the appearance of an activity that phosphorylates both p38 MAP kinase and SAPK (Fig. 3B, upper panel). MEK3 and MEK4 expressed alone had little ability to phosphorylate p38 or SAPK.

After purification of coexpressed MEK3 or MEK4 on Ni²⁺-NTA-agarose and glutathione-Sepharose, respectively, Western blotting with specific antibodies revealed forms with multiple electrophoretic mobilities (Fig. 3B, lower panel). Multiple forms, corresponding to low activity and activated MEK3, were also detected by Coomassie Blue staining of gels (data not shown). The expression of MEK3 and MEK4 with MEKK-C results in activation of approximately half of each kinase as deduced from their distribution between faster and slower migrating bands.

The specific activity of MEK3 after coexpression was lower than that of MEK4 using p38 as a substrate, despite the fact that in our protein preparations the low electrophoretic mobility, active form of MEK3 represents approximately 25% of the total protein, while active MEK4 is much less pure. We have observed equivalent differences in activity of MEK3 and MEK4 activated in vitro by a novel MEKK.²

The coexpression studies above suggested that each MAP kinase module component can be activated in bacteria by coexpression with a constitutively active upstream module component. Thus, we wished to determine if three components could be expressed simultaneously, leading to activation of the downstream enzymes. The wild-type forms of MAP kinases from three different modules were subcloned with an appropriate MEK on the same high copy plasmid in "head to tail" orientation under the control of the T7 promoter (Fig. 1B). As an activator for all three modules, we used MEKK-C expressed from the low-copy pBB131 plasmid, used previously for activation of MEK family members.

The expression of three kinases together resulted in activation of the downstream kinase in the corresponding MAP kinase module, namely -SAPK (Fig. 5A) and p38 (Fig. 5B). Expression of ERK2 with wild-type MEK1 or MEK2 and MEKK-C also resulted in ERK2 activation (data not shown). The specific activity of -SAPK toward GST-c-Jun1-223 from this coexpression system and purified on Ni²⁺-NTA-agarose and then Mono Q varied from 1 to 1.2 µmol/min/mg protein (Table I). The activity of p38 with GST-ATF2(1-254) as substrate varied from 0.6 to 0.9 µmol/min/mg protein. Expression of wild-type MEK1 or MEK2 with ERK2, MEK4 with -SAPK, or MEK4 with p38 but without MEKK-C resulted in little or no activation of downstream MAP kinases (data not shown).

A, specific activity of -SAPK expressed alone or coexpressed with MEK4 and MEKK-C, before and after treatment with protein phosphatases. Purified -SAPK (1.6 µg/ml) expressed in E. coli alone or coexpressed with MEK4 and MEKK-C was treated with the indicated protein phosphatases with or without pretreatment with corresponding phosphatase inhibitors. 3 ng of SAPK were assayed with GST-c-Jun1-223 (40 µg/ml). Control experiments showed that both protein phosphatases were completely inhibited and had no activity in the GST-c-Jun phosphorylation assay (data not shown). A representative experiment is shown. B, specific activity of p38 MAP kinase expressed alone or coexpressed with MEK4 and MEKK-C after protein phosphatase treatment. Purified p38 MAPK (1.9 µg/ml) expressed in E. coli alone or coexpressed with MEK4 and MEKK-C was treated with protein phosphatases as indicated. ~3 ng were assayed with GST-ATF21-254 (0.18 mg/ml). A representative experiment of two is shown. C, immunoreactivity of coexpressed -SAPK with anti-phospho-SAPK antibody after protein phosphatase treatment. Purified -SAPK (1.6 µg/ml) expressed in E. coli alone or coexpressed with MEK4 and MEKK-C was treated with protein phosphatases as indicated. An equivalent amount of -SAPK (20 ng/lane) was probed with an antibody specific to phosphorylated SAPK isoforms. Bottom, the same blot was stripped and reprobed with anti-SAPK antibody O977. Shown here is a representative of two experiments.

Table I. Specific activities of MAP kinases expressed alone or coexpressed with activators MAP kinases were expressed in E. coli alone or coexpressed with constitutively active MEK1 (ERK2) or with MEKK-C and MEK4 (-SAPK and p38). After purification, the specific activities were measured with MBP for ERK2, GST-c-Jun1-223 for -SAPK, and GST-ATF21-254 for p38. Specific activities in nmol/min/mg, determined as under "Materials and Methods," are the average of two or three determinations ± range/n. ERK2  -SAPK p38 Coexpressed 2100  ± 500 1500  ± 310 770  ± 130 Expressed alone 1.3  ± 0.7 0.003a 29  ± 8 a Too low to be calculated in other experiments.

To demonstrate that -SAPK and p38 expressed in this system are doubly phosphorylated on Tyr and Thr, we treated the purified active kinases with PTP1 or PP2A. Treatment with either phosphatase results in a time-dependent loss of -SAPK (Fig. 5A) and p38 activity (Fig. 5B) toward their substrates. Inactivation was blocked by the phosphatase inhibitors (Fig. 5, A and B).

To characterize the coexpressed -SAPK protein further, we used an antiserum raised against a doubly phosphorylated peptide from the JNK/SAPK phosphorylation lip sequence encompassing phosphothreonine and phosphotyrosine residues that are required for their activation (7, 8). This antibody did not recognize 20 ng of -SAPK expressed in the absence of activators (Fig. 5C, last lane). In contrast, if -SAPK was coexpressed with MEK4 and MEKK-C, 7 ng of it were recognized easily by this antibody (Fig. 5C, lanes 1, 8, 16, and data not shown). Treatment with PTP1 or PP2A greatly reduced -SAPK recognition in a time-dependent manner (Fig. 5C), although treatment with PP2A did not eliminate -SAPK immunoreactivity. Treatment with both phosphatases together completely abolished recognition of 20 ng of -SAPK by the anti-active JNK/SAPK antiserum (Fig. 5C). Upon SDS-PAGE, the coexpressed kinase preparations each contained major silver-stained bands that correspond to their activated forms (not shown).

DISCUSSION

We describe the reconstitution of three different MAP kinase cascades in bacteria. The MEK and MAP kinase family members were activated by coexpression with MEKK-C or MEK1 R4F as the upstream activator. In both the two- and three-component systems, we observed a high degree of ERK2, JNK/SAPK, or p38 activation, concomitant with the conversion of the expressed kinases into the doubly phosphorylated forms. Thus, we demonstrate for the first time that bacteria can efficiently phosphorylate and activate expressed proteins if suitable modifying enzymes are present. This method has wide applicability in the production of significant quantities of selectively modified phosphoproteins.

One of the applications of the coexpression system described above is the production of milligram amounts of active MAP kinases and MEKs for biochemical and physical studies. As one example, we have recently optimized this system for production of large amounts of phosphorylated ERK2 that has led to the determination of its crystal structure. The capacity to activate MAP kinase family members by expressing three components of the kinase cascade simultaneously offers a significant advance in efforts to obtain large quantities of downstream MAP kinases even when a constitutively active MEK is not available. Using this approach, we found that the specific activities of activated ERK2, -SAPK, and p38 are similar, ~1-2 µmol/min/mg protein.

MEKs 1-4, which belong to the most extensively characterized MAP kinase modules, are activated by MEKK-C despite its low expression in E. coli. We were able to detect expressed MEKK-C after purification of GST-tagged or His₆-tagged protein only on Western blots or by assay, not by protein staining. Although these findings suggest little selectivity of the catalytic domain of MEKK1 among MEK family members, we found that MEK5, a MEK from a different kinase module, cannot be activated by MEKK-C either in vitro or by coexpression in E. coli (12).³ Thus, MEKK1 has a restricted ability to phosphorylate MEK family members.

When we engineered bacteria to express three kinase cascade members simultaneously, we observed strong activation of the downstream enzyme in each MAP kinase module. This finding indicates that activation of downstream components of MAP kinase cascades does not require other components. Because MEKK1 phosphorylates several MEKs, the function of MEKK1 in a given context may not be due to its inherent enzymatic specificity. Perhaps components that exist only in mammalian cells, such as scaffolding proteins analogous to Ste5p found in the budding yeast S. cerevisiae, may confer specificity within a cellular context (25, 26). Specificity might also be achieved through other means of cascade compartmentalization. Indeed, we have found alternatively spliced forms of MEK5, which have different intracellular localizations that may lead to regulatory differences (12).