Altered Regulation of ERK1b by MEK1 and PTP-SL and Modified Elk1 Phosphorylation by ERK1b Are Caused by Abrogation of the Regulatory C-terminal Sequence of ERKs*

ERK1b is an alternatively spliced form of ERK1, containing a 26-amino acid insertion between residues 340 and 341 of ERK1. Although under most circumstances the kinetics of ERK1b activation are similar to that of ERK1 and ERK2, we have previously found several conditions under which the activation of ERK1b by extracellular stimuli differs from that of other ERKs. We studied the molecular mechanisms that cause this differential regulation of ERK1b and found that ERK1b is altered in its ability to interact with MEK1 and this influenced its subcellular localization but not its kinetics of activation. ERK1b had a decreased ability to phosphorylate Elk1, but this did not change much the transcriptional activity of the latter. Importantly, the interaction of ERK1b with PTP-SL, which can act as a MAPK phosphatase, shortly after mitogenic stimulation, was significantly affected as well. Using mutants of ERK1b we found that the differential interaction of ERK1b with the three effectors is caused by the site of insertion that abrogates the cytosolic retention sequence/common docking motif of ERKs, and is not dependent on the actual sequence of the insert. Prolonged epidermal growth factor stimulation of Rat1 cells resulted in a differential inactivation and not activation of ERK1b as compared with ERK1 and ERK2. The reduced sensitivity to phosphatases without major differences in the kinetics of activation or activation of substrates, suggests that ERK1b plays a role in the transmission of extracellular signals under conditions of persistent stimulation, where ERK1b and MAPK phosphatases are induced, and the activity of ERK1 and ERK2 is suppressed.

association with MEK1 (7). However, the reduced interaction of ERK1b with MEK1 did not significantly change its kinetics of activation upon EGF stimulation, indicating that additional parameters are involved in the regulation of ERK1b.
We undertook to elucidate the molecular mechanisms that underlie the differential regulation of ERK1b in several cells. Since the insertion of ERK1b is located near the CRS/CD of ERKs we studied the interaction of ERK1b with the upstream activator MEK1, the phosphatase PTP-SL, and the substrate Elk1. We found that the interaction with all three proteins was dramatically reduced as compared with the interaction of ERK1 with the same proteins, and this was due to the site of insertion rather than the actual sequence of the insert. Whereas the reduced interaction with phosphatase resulted in a reduced sensitivity of ERK1b to dephosphorylation, the reduced interaction with MEK1 did not change the kinetics of ERK1b activation. The reduced interaction with Elk1 changed the pattern of Elk1 phosphorylation but did not significantly modify the phosphorylation of the activating Ser 383 or the activity of Elk1 upon EGF stimulation. Similar to the overexpressed protein, the rate of activation of endogenous ERK1b by EGF was not affected, while its inactivation was different from that of ERK1 and ERK2. Our results therefore suggest that the abrogation of CRS/CD in ERK1b results mainly in altered down-regulation and much less in its kinetics of activation or substrate phosphorylation. These results, as well as the elevated expression of ERK1b upon prolonged stimulation, suggest that ERK1b may be required for the transmission of extracellular signals under conditions of persistent stimulation, where induced MAPK phosphatases suppress the activity of ERK1 and ERK2.
ERK Constructs-Rat ERK1 and rat ERK1b constructs (7) were subcloned into EcoRI and XhoI sites of pCDNA1 (Invitrogen, Carlsbad, CA). All the following constructs were prepared by polymerase chain reaction using ERK1 as a template. CT-ERK1b was prepared by fusing the 78-base pair insert of ERK1b to the C terminus of ERK1. The fusion was done by two sequential polymerase chain reaction extending steps. Scrambeld-ERK1b (Sc-ERK1b) was made by random scrambling of the insert of ERK1b to produce the amino acid sequence: RAPVPSVRSGPR-PASLVQCPIPGYCV. The DNA cgggctcccgttccaagcgtacggagtggaccaaggc ctgcgagtcttgtccagtgtcccattcctgggtactgcgtg was inserted back into its original position in ERK1 by sequential polymerase chain reaction (14). D339N-ERK1 was prepared by sequential polymerase chain reaction steps (14). All the constructs obtained were ligated either to the 3Ј end of hemagglutinin (HA) tag and inserted into the EcoRI and XhoI sites of pCDNA3 (Invitrogen) or to the 3Ј end of green fluorescence protein (GFP) followed by insertion into the BspEI and XhoI sites of pEGFP-C1 (CLONTECH, Palo Alto, CA) vector.
Cell Culture and Transfection-Rat1 and COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS). COS-7 cells, grown in 10-cm plates, were transfected using the DEAE-dextran method using 5 g of the desired DNA. Rat1 cells were transfected with polyethylenimine (16). Briefly, the cells were seeded on coverslips in 12-well plates and were grown to 50 -70% confluency. The plasmid (1.5 g) was suspended in 50 l of NaCl (150 mM) and mixed with polyethylenimine solution (50 l of 3 mM polyethylenimine in 120 mM NaCl). The mixture was left at room temperature for 15 min and incubated with the cells for 90 min, after which the cells were washed and placed in Dulbecco's modified Eagle's medium ϩ 10% FCS.
Preparation of Cell Extracts and Western Blotting-COS-7 and Rat1 cells were grown to subconfluency and were then serum-starved for 16 h in Dulbecco's modified Eagle's medium containing 0.1% FCS. The cells were then treated with various stimuli for the indicated times. After stimulation, the medium was removed, and the cells were rinsed twice with ice-cold phosphate-buffered saline and once with ice-cold Buffer A. Cells were scraped into Buffer H (0.5 ml/plate) and disrupted by sonication (2 pulses for 7 s of 50 W). The extracts were centrifuged (20,000 ϫ g, 15 min, 4°C), and the supernatants, which contained the cytosolic and nuclear proteins, were further kept at 4°C. The supernatants were then resolved by a 10% SDS-PAGE, transferred onto a nitrocellulose membrane, and probed with the appropriate Abs. The binding of Abs was detected using alkaline phosphatase or ECL. Anti-C terminus of ERK-Ab (C-16), anti-Elk1-(l-20), anti-pElk1-(Ser 383 )-, and anti-HA-Abs were from Santa Cruz Biotechnology (Santa Cruz, CA), anti-pERK-Ab (DP-ERK) was from Sigma (Rehovot Israel), anti-GFP-Ab from Roche Molecular Diagnostics (Indianapolis, IN), and anti-MEK1-Ab from BD Transduction Laboratories (Lexington, KY).
Immunoprecipitation-One day after transfection, COS-7 cells were serum starved (0.1% FCS) for an additional 16 h and then stimulated with peroxovanadate (Na 3 VO 4 (100 M) and H 2 O 2 (200 M); VOOH; 18 min, 37°C), EGF (50 ng/ml), or treated with phosphate-buffered saline as control. After stimulation, the cells were washed, lysed, and centrifuged as described above. The extracts were then incubated (2 h, 4°C) with a polyclonal anti-HA-Ab or monoclonal anti-GFP-Ab coupled to protein A/G-Sepharose beads. For the determination of ERK activity, and for PTP-SL co-immunoprecipitation the beads were washed once with RIPA buffer, twice with 0.5 M LiCl, and twice with Buffer A as described (17). The immunoprecipitates were subjected either to Western blotting as above or subjected to an in vitro kinase assay as below.
Co-immunoprecipitation of MEK1 with ERKs-COS-7 cells were transfected with each of the ERK constructs or with wt-MEK1 and treated as described above. After starvation, the cells were washed, lysed by sonication, and centrifuged (15,000 ϫ g, 15 min, 4°C). The supernatants from the ERK-overexpressing cells were loaded onto columns of agarose-conjugated anti-ERK-Ab (0.2-ml beads 0.5 mg/ml, Santa Cruz). The columns were washed extensively to remove any impurity and then the supernatants from the wt-MEK1 overexpressing cells were loaded onto the mini-affinity columns. The columns were washed with buffer A (3 ϫ 1 ml) and the attached MEK1 was eluted with buffer A ϩ 0.1 M NaCl (0.5 ml). The eluted fractions were subjected to immunoblotting with anti-MEK1-Ab.
In Vitro Kinase Assay-The immunoprecipitated ERK proteins attached to beads were mixed with either MBP (0.5 g/reaction) or Elk1 fusion protein (0.5 g/reaction; Cell Signaling Inc., Beverly, MA) and Buffer RM that contained 100 M [␥-32 P]ATP at 30°C. After 20 min, the reaction was terminated by adding 10 l of 4 ϫ SB, resolved on SDS-PAGE, and subjected to autoradiography and Western blot analysis as indicated.
Preparation of Recombinant PTP-SL and in Vitro Determination of Phosphatase Activity-COS-7 cells were transfected with a vector containing HA-PTP-SL and maintained in 10% FCS for 48 h, after which the cells were harvested and lysed as described above. HA-PTP-SL was partially purified from the COS-7 extracts by anion exchange chromatography using Mono Q HR 5/5 column (Amersham Pharmacia Biotech) with a linear 0 -0.4 M NaCl gradient in phosphatase buffer (25 mM HEPES, pH 7.5, 5 mM EDTA, 10 mM dithiothreitol (15)). The fractions containing HA-PTP-SL were identified by Western blotting with anti-HA-Ab. In vitro dephosphorylation was performed by mixing the HA-PTP-SL (ϳ10 ng) with immunoprecipitated GFP-ERK1 or GFP-ERK1b in the phosphatase buffer. The dephosphorylation was performed at 30°C with constant shaking. The reactions were stopped by adding 4 ϫ SB, and the mixture was resolved on SDS-PAGE.
Localization of GFP-ERKs-Rat1 cells were co-transfected with each one of the GFP-ERK constructs together with either MEK1 (2:1 with ERK) or vector control. Serum-starved cells were stimulated with EGF (50 ng/ml, 30 min) or left untreated and then washed with phosphatebuffered saline and fixed in 3% paraformaldehyde in phosphate-buffered saline for 20 min (8). The fluorescence was detected by a confocal microscope (Zeiss Axioscopmicroscope, HBO 100 W/2; ϫ40 magnification).
Elk1 Reporter Assay-To examine Elk1 activity, PathDetect Elk1 trans-Reporting System (Stratagene, La Jolla CA) was used. Briefly, COS-7 cells were co-transfected with each one of the examined ERKs or empty vector, together with pFR-Luc containing five GAL4 response element sites and a minimal promoter driving a luciferase gene, plasmid of Elk1 fused to the DNA-binding region of GAL4, and plasmid containing Renilla as a control. The cells were stimulated with different concentrations of EGF for 14 h and harvested. The luciferase and Renilla luminescent was measured using Dual Luciferase Reporter Assay system (Promega) reagents according to the manufacturers instructions and luminometer (TD-20e, Turner). The results were calculated as the ratio between the luminescent of luciferase and Renilla over the luminescent obtained with empty vector.

Construction of Mutants to Study the Role of the Site and
Sequence of the 26-Amino Acid Insert of ERK1b-We have previously shown that ERK1b differs from ERK1 in the ability to be regulated by MEKs (7). Although this differential regulation could have evolved from the disruption of the CRS/CD (8,9), it is also possible that the actual sequence of the 26-amino acid insert of ERK1b plays a role in its differential regulation. To distinguish the role of the amino acid sequence from the role of the insertion site, we constructed several mutants of ERK1 and ERK1b (Fig. 1A). Thus, CT-ERK1b was prepared by fusing the insert sequence to the C terminus of ERK1. Sc-ERK1b was prepared by changing (scrambling) the sequence of the insert in its original position, and D339N-ERK, which contains a mutated amino acid just next to the insertion site, was made in order to prepare an analogous mutation to that of the Drosophila gain of function mutant sevenmaker (18,19). In order to follow the specific activity and the localization of the mutated proteins, we fused either HA or GFP tags to their N terminus.
To confirm the proper expression of the constructs, and to examine whether the modifications inserted interfere with the activity of the mutated proteins, we transfected the GFPtagged constructs into COS-7 cells. Thirty-six hours after transfection the cells were serum-starved (0.1% FCS, 16 h), stimulated with either EGF or peroxyvanadate, and the expressed GFP-ERKs were tested for their phosphorylation and activation. All five constructs induced high, but not identical, level of expression of the various proteins, which was not changed upon stimulation (Fig. 1B). As expected, stimulation of the transfected COS-7 cells (with either EGF or peroxyvanadate (VOOH)) resulted in a dramatic elevation in phosphate incorporation into the regulatory Thr and Tyr residues of ERK1 (Fig. 1B). This phosphorylation was not different from that of ERK1b or of the mutants examined (Fig. 1B), and this was true also for the initial rates of the phosphorylation of the various proteins (data not shown). We then examined whether the activating phosphorylation of the proteins correlates with their kinase activity. To do so, the GFP fused proteins from unstimulated or peroxyvanadate-treated cells were immunoprecipitated and subjected to an in vitro kinase assay toward MBP. Interestingly, elevated basal activity and enhanced phosphorylation were detected in non-stimulated ERK1b, Sc-ERK1b, and in D339N-ERK1 but not in ERK1 or in CT-ERK1b (Fig. 1C), indicating that the first three are regulated differently from ERK1 under basal conditions. However, upon stimulation, all ERKs exhibited a similar activity toward MBP (Fig. 1C, and data not shown), confirming that the stimulated kinase activity of the ERKs was not affected by the various modifications.
The Site and Not the Sequence of the 26-Amino Acid Insert Determines the Reduced Interaction of ERK1b with MEK1-We have previously reported that ERK1b, unlike ERK1 or ERK2, is unable to associate with MEK1, and this causes an altered subcellular localization that may lead to a differential regulation of ERK1b (7). To study the nature of these differences we made use of the ERK constructs described above, which were transfected into COS-7 cells. The expressed ERKs were immunoprecipitated, extensively washed, and then used to precipitate overexpressed MEK1 from extracts of MEK1-transfected COS-7 cells. As reported (7), MEK1 was associated with ERK1 under the conditions of the experiment, whereas the associa- CT-ERK1b, in which the 26-amino acid insertion of ERK1b (f) was fused to the C terminus of ERK1; (iv) Sc-ERK1b, in which the sequence of the ERK1b insertion was scrambled (:::), (v) 339N-ERK1, in which Asp 339 was replaced with Asn, to form an analogous mutation to that of the Drosophila sevenmaker (rl sm ). The amino acid preceding the insertion, the site of the D339N mutation, the kinase domain, the N terminus of the proteins, and the C terminus of the proteins are indicated. B, expression and phosphorylation of the constructs in COS-7 cells. COS-7 cells were transfected with GFP-ERK1, GFP-ERK1b, GFP-CT-ERK1b, GFP-Sc-ERK1b, or GFP-339N-ERK1. After serum starvation, the cells were stimulated with EGF (50 ng/ml, 10 min), VOOH (Na 3 VO 4 (100 M) and H 2 O 2 (200 M), 20 min), or left untreated as control. Cell extracts were separated by SDS-PAGE followed by determination of the phosphorylation and amount of the GFP-ERKs using Western blot analysis with the indicated Abs. The region of the gel presenting GFP-ERKs is presented and the position of GFP-ERKs and phospho-GFP-ERKs (GFP-pERKs) is indicated. C, activity of ERK1, ERK1b, and their mutants. COS-7 cells were transfected with the indicated constructs or with vector control, treated as above, and then either stimulated with VOOH (20 min) or left untreated. The GFP-ERKs were immunoprecipitated with anti-GFP-Ab and subjected to an in vitro kinase assay with MBP as a substrate. The phosphorylation of MBP was detected by autoradiography on x-ray film (AGFA). Phosphorylation and amount of the GFP-ERKs were determined by a Western blot analysis with the indicated Abs. Positions of phospho-MBP (pMBP), GFP-ERKs, phospho-GFP-ERKs (GFP-pERKs), and Abs are indicated. These results were reproduced four times. tion of MEK1 to ERK1b was substantially reduced (Fig. 2). The interaction of CT-ERK1b with MEK1 was similar to that of ERK1, whereas both Sc-ERK1b and D339N-ERK1 exhibited a reduced association, similarly to that of ERK1b. These observations indicate that the reduced association of MEK1 to ERK1b is determined primarily by the disruption of the CRS/CD of ERK1 and not by the actual sequence of the 26 amino acid insert.
Association with MEK1 in non-stimulated fibroblasts causes retention of ERK1 in the cytosol (8,20), and mitogenic stimulation of the cells induces dissociation of ERK1 from MEK1 and its rapid nuclear translocation (21). We have previously shown that the retention of ERKs in the cytosol depends on the integrity of their CRS/CD (8). Since the CRS of ERK1 is altered by the 26-amino acid insert in ERK1b, it was likely that the subcellular localization of ERK1b would be affected. Indeed, as previously shown for HA-ERK1 (7), the ectopically expressed GFP-ERK1 accumulated in the nucleus of Rat1 cells. Co-expression of MEK1 resulted in a cytosolic distribution of the GFP-ERK1 and this became nuclear upon stimulation (Fig. 3). As expected, GFP-ERK1b was not retained by MEK1 in the cytosol of Rat1 cells under similar conditions. We then repeated the same experiment with the three GFP-conjugated ERK1 mutants and found that in the absence of MEK1 there was no change in their nuclear localization. However, coexpression with MEK1 resulted in a stimulation-reversible cytosolic retention of CT-ERK1, but not of Sc-ERK1b that behaved much like the unmodified GFP-ERK1b. Therefore, these results support the notion that the site of insertion, which interferes with the regulatory CRS, rather than the actual sequence of the insert, is responsible for the differential interaction of ERK1b with MEK1. Interestingly, GFP-D339N-ERK1, which is also modified in the CRS, exhibited preferential, cytosolic localization in the cells coexpressing MEK1. These results support previous reports that suggested that Asp 339 contributes only part of the binding energy between MEKs and ERKs, and other residues in the CRS are necessary for this process (8,10). Therefore, the 26-amino acid insertion appears to completely abrogate the CRS and not just the parts close to the site of insertion (Glu 340 (7)).
Modulated Substrate Specificity of ERK1b Is Manifested by Its Altered Phosphorylation and Activation of Elk1-Beside its key role in the interaction with MEKs, CRS/CD has been recently implicated also in the binding of ERKs with their substrates and with phosphatases (9). Because of its crippled CRS/ CD, the interaction of ERK1b with these effectors could be altered, and we first examined whether ERK1b differs from ERK1 in its ability to interact with the transcription factor Elk1, which is a known substrate of ERK1 (13). The main regulatory phosphorylation in Elk1 occurs on Ser 383 that is phosphorylated by ERKs and other MAPKs upon mitogenic stimulation (22). However, in addition to this residue, also residues Thr 353 , Thr 363 , Thr 368 , Ser 389 , Thr 417 and more, can be phosphorylated by ERKs, but the role of their phosphorylation is not fully understood (23). Interestingly, the phosphorylation of each of the sites causes a small retardation of Elk1 migration (upshift) on an SDS-PAGE (23), and this enables the detection of the different phosphorylations using Western blot with antigeneral Elk1 and anti-phospho-Ser 383 Elk1-Abs. To dissect the different phosphorylation by ERK1 and ERK1b, we immunoprecipitated these proteins from EGF-stimulated COS-7 cells that had been transfected with each of the HA-tagged con- (1), GFP-ERK1b (1b), GFP-CT-ERK1b (CT), GFP-Sc-ERK1b (Sc), or GFP-339N-ERK1 (339N) together with either an empty vector (ϪMEK) or a plasmid containing wild-type MEK1 (ϩMEK). The cells were serum starved and then were either left untreated (basal) or treated with EGF (50 ng/ml, 30 min). The cells were fixed and the GFP-ERKs were visualized using confocal microscope (Axiovert 100 TV, Zeiss). The pictures are from a representative experiment that was reproduced three times.
structs. The immunocomplexes were then extensively washed, and comparable amounts of TEY-phosphorylated ERK1 and ERK1b were subjected to an in vitro kinase assay with recombinant Elk1 as a substrate. As we have previously reported (7), the total amount of Elk1 phosphorylation by active ERK1b detected by total 32 P incorporation, was slightly smaller than that induced by active ERK1 (Fig. 4A).
We then examined whether the reduction in the phosphorylation by ERK1b was due to reduction in the phosphorylation of Ser 383 or of other sites. To do so we used increasing amounts of immunoprecipitated ERK1 and ERK1b (comparable in their TEY phosphorylation, Fig. 4B, bottom), to phosphorylate recombinant Elk1. Interestingly, the rate of Ser 383 -Elk1 phosphorylation by the smallest amount (3 l) of ERK1 used was similar to that achieved by similar amounts of ERK1b and this did not significantly change when higher amounts of the active kinases were added to the reaction (Fig. 4B, top). On the other hand, significant differences were seen in the upshift of Elk1 upon phosphorylation (Fig. 4B, center). Thus, incubation with active ERK1 caused a significant upshift of the phosphorylated Elk1, which migrated as a 55-kDa protein upon phosphorylation with 27 l of active ERK1 beads. Phosphorylation by ERK1b caused a smaller upshift of Elk1, which reached a maximal molecular mass of 48 kDa even with higher amounts of ERK1b (data not shown). These results indicate that although the rate of Ser 383 -Elk1 phosphorylation by ERK1 and ERK1b is similar, ERK1b is defective in its ability to phosphorylate other sites in Elk1.
Whether the specific sequence of the 26-amino acid insert of ERK1b has a role in this differential substrate recognition was tested by repeating the above procedure with the three ERK mutants. As above, Elk1 phosphorylation by active ERK1 (30 l of beads, 20 min) caused a significant gel retardation (Fig.  4C), which was lower when Elk1 was phosphorylated by a similar amount of ERK1b. Phosphorylation by 30 l of active CT-ERK1b resulted in a similar upshift to that caused upon phosphorylation by ERK1, while phosphorylation by Sc-ERK1b gave a similar pattern to that caused by ERK1b. These results indicate that the site of insertion, rather than the insert sequence, is important for the differential substrate recognition. The upshift of Elk1 caused by active D339N-ERK1 was similar to that caused by active ERK1, indicating that as with MEKs, alteration of Asp 339 alone cannot modify the specificity of ERK1 toward Elk1.
We next tested whether the differential phosphorylation by ERK1 and ERK1b influences the transcriptional activity of Elk1. This was done using a luciferase reporter gene fused to a promoter containing a GAL4 response element and a vector expressing a chimeric protein composed of the transactivation domain of Elk1-fused GAL4 DNA-binding region. Thus, we co-transfected COS-7 cells with each of the GFP-ERK constructs together with plasmids containing the reporter gene, serum-starved the cells for 16 h, stimulated them with EGF for 14 h, and finally determined their luciferase activity using the Dual Luciferase Reporter Assay system. ERK1b enhanced Elk1 activity 2-fold over the activity induced by ERK1, as detected at all the concentrations of EGF examined (Fig. 5). The enhancement in transcription activity was again due to the site, rather than the actual sequence of the 26-amino acid insert of ERK1b. This was concluded from the similar effect of CT-ERK1b and ERK1, which were smaller than the effects caused by Sc-ERK1b and ERK1b, although their TEY phosphorylation shortly after stimulation were comparable (Fig. 5, inset). Interestingly, the transcriptional activity of Elk1 upon coexpression with D339N-ERK1 was even 2-fold higher than that enhanced by ERK1b (4-fold higher than by ERK1). Thus, there was in The immunocomplexes were then subjected to an in vitro kinase reaction using recombinant Elk1 (0.5 g in each reaction, 20 min phosphorylation) as a substrate. The reactions were terminated with SBx4 and the phosphorylated proteins were subjected to SDS-PAGE and autoradiography. The total phosphorylation was determined by densitometry and the results show the mean and S.E. of three experiments. B, dose response of Elk1 phosphorylation by ERK1 and ERK1b. Active GFP-ERK1 and GFP-ERK1b were each immunoprecipitated with anti-GFP-Ab from EGF-stimulated (50 ng/ml, 5 min) COS-7 cells overexpressing the appropriate proteins. Increasing amounts of ERK-conjugated beads (3, 9, and 27 l) were subjected to an in vitro kinase assay using recombinant Elk1 (0.5 g in 30 l of RM) as a substrate. Phosphorylation of Elk1 was detected with anti-pSer 383 Elk1-Ab, while the amount of Elk1 and its phosphorylation-dependent upshift were detected with anti-Elk1-Ab. The phosphorylation of ERK in each assay was determined using anti-pERK and anti-gERK-Ab. Positions of pSer 383 -Elk, Elk1, and pERK are indicated. C, different phosphorylation pattern of Elk1 by ERK1, ERK1b, and their mutants. COS-7 cells were transfected with GFP-ERK1 (1), GFP-ERK1b (1b), GFP-CT-ERK1b (CT), GFP-Sc-ERK1b (Sc), or GFP-339N-ERK1 (339N). The cells were serum-starved (0.1% FCS, 16 h) and then were either stimulated with EGF (50 ng/ml, 10 min) or left untreated. GFP-ERKs were immunoprecipitated with anti-GFP-Ab and the beads containing GFP-ERKs were subjected to an in vitro kinase assay using recombinant Elk1 (0.5 g in 30 l of RM) as a substrate. Phosphorylation of Elk1 was detected by anti-Ser(P) 383 Elk1-Ab. The amount of Elk1 and its phosphorylation-dependent upshift were detected with anti-Elk1-Ab. ERK phosphorylation was determined using anti-pERK and anti-gERK-Ab. Positions of Ser(P) 383 -Elk1, Elk1, and pERK are shown. These results were reproduced three times. fact no correlation between the amount of phosphorylation on Elk1 with the elevation of Elk1 activity by ERK1b or by D339N-ERK1, indicating that other parameters should be involved in this regulation.

Reduced ERK1b Association with and Dephosphorylation by PTP-SL Is Caused by the Disruption of CRS/CD-Key regulators of ERK activity are protein-tyrosine phosphatases (PTP)
including PTP-SL (15), STEP (15), LC-PTP/HePTP (24), and PTP-ER (25) that have been implicated in the immediate dephosphorylation of ERKs upon mitogenic stimulation. A reduced interaction of such phosphatases with ERK1b, as compared with ERK1, can explain the higher activity of ERK1b in non-stimulated cells and toward Elk1. To study this possibility we used PTP-SL that has been implicated in the inactivation of ERKs and was shown to retain ERK1 in the cytoplasm by association with it through a kinase interaction motif (15,26). Partially purified recombinant PTP-SL was first used in an in vitro dephosphorylation assay using equally activated ERK1 and ERK1b as substrates. ERK1 was very rapidly dephosphorylated by the phosphatase, while the kinetic of ERK1b dephosphorylation under similar condition was much slower (Fig. 6), indicating that ERK1b is less sensitive to dephosphorylation by PTP-SL than ERK1.
We then tested the sensitivity of ERK1 and ERK1b to PTP-SL in vivo by co-transfecting into COS-7 cells either ERK1 or ERK1b together with increasing amounts of a plasmid containing PTP-SL. Similar amounts of ERK1 and ERK1b proteins were produced in these cells, and those were phosphorylated to a comparable level in response to EGF (Fig. 7A, upper  panel). However, the EGF-dependent phosphorylation was differentially prevented by the co-transfected PTP-SL. Thus, ERK1 was dephosphorylated (92%) already in cells containing the lowest amount of PTP-SL (0.1 g, Fig. 7A), while the dephosphorylation ERK1b under these conditions was much lower (ϳ15%, 0.1 g of plasmid) and reached 88% dephosphorylation only when 20-fold higher amounts of PTP-SL was used. It should be noted that the differential dephosphorylation of ERK1 and ERK1b seen in the transfected COS-7 cells (Fig.  7A) was similar to the differences between the dephosphorylation of the two ERKs in vitro (Fig. 6). Therefore, these results indicate that the main effect of the transfected PTP-SL in the COS-7 cells was exerted from its effects on the ERKs themselves and not on any upstream components of the EGF-receptor signaling machinery. Taken together the results in Figs. 6 and 7A, it can be concluded that ERK1b is more resistant to dephosphorylation by PTP-SL than the ERK1.
We then examined the role of the insertion sequence and the insertion site in the resistance of ERK1b to PTP-SL. Again, the effect of ERK1b was achieved by the site of the 26-amino acid insertion and not from the actual sequence of the ERK1b insert. This was judged by the fact that ERK1 and CT-ERK1b were ϳ90% dephosphorylated by wt-PTP-SL (compared with the inactive mu-PTP-SL), while ERK1b and Sc-ERK1b exhibited almost no dephosphorylation when co-transfected with PTP-SL and stimulated with EGF for 5 min (Fig. 7B). As expected by the similarity to other phosphatases (27), also D339N-ERK1 was resistant to PTP-SL as judged by its lack of sensitivity to the coexpressed phosphatase upon EGF stimulation (Fig. 6B, left). As it has been reported that the activity of PTP-SL is dependent on its direct interaction with ERKs (26), we next examined the ability of the PTP-SL to physically associate with the various ERK constructs. Thus, we co-transfected HA-PTP-SL together with the various GFP-ERK constructs and immunoprecipitated the latter with anti-GFP-Ab. Similar amounts of HA-PTP-SL were expressed in all cells examined (Fig. 7C, bottom) but these were co-immunoprecipi-tated only with ERK1 or CT-ERK1b and not with ERK1b, Sc-ERK1b, or D339N-ERK1 (Fig. 7C, top). These results indicate that, in similarity to other phosphatases (27), the resistance to PTP-SL is correlated with reduced binding to the phosphatase, which occurs due to the site and not the sequence of insertion.
Since the above experiments were performed with ectopically expressed ERK1b, we next examined the effect of PTP-SL on the endogenous ERK1b. To do so we transfected plasmids containing either wild type or an inactive form (C293S) of PTP-SL into Rat1 cells (30 -40% efficiency of transfection). After serum starvation, the cells were stimulated with EGF, and examined for the regulatory phosphorylation of the endogenous ERKs using anti-pERK-Ab. The phosphorylation of both ERK1 and ERK2 was reduced (ϳ50%) in the cells transfected with wildtype PTP-SL as compared with the cells transfected with inactive PTP-SL (Fig. 8). However, the phosphorylation of ERK1b not only did not decrease but in fact slightly increased in the cells bearing the active phosphatase. The marked differences between ERK1b and ERK1/2 observed here indicate that the reduced sensitivity of ERK1b to PTP-SL is not confined to the exogenous protein but occurs also with the endogenous ERK1b.
The Dephosphorylation, and Not the Acute Phosphorylation of ERK1b, Is Compromised Upon EGF Stimulation of Rat1 Cells-Since ERK1b seems to be resistant to phosphatases, it was important to study not only its up-regulation in response to extracellular stimulation, but also the kinetics of its downregulation at later stages of stimulation. Therefore, we stimulated Rat1 cells with EGF for up to 48 h, and examined the regulatory phosphorylation of the various ERK isoforms and their expression. Interestingly, shortly after stimulation (15 min) ERK1b phosphorylation was induced to a similar level as ERK1 and ERK2. However, this was changed with time, as the phosphorylation of ERK1 and ERK2 decreased to the basal level 6 h after stimulation, while ERK1b remained about 50% phosphorylated even 42 h later (Fig. 9, A and B). A similar trend was observed with anti-pERK-Ab and also with anti- FIG. 7. Differential regulation of ectopically expressed ERK1, ERK1b, and their mutants by PTP-SL. A, differential dephosphorylation of ERK1 and ERK1b by PTP-SL. COS-7 cells were co-transfected with 1 g of plasmid containing either HA-ERK1 or HA-ERK1b, together with the indicated amounts of plasmid containing wt-PTP-SL or with 2 g of mu-PTP-SL (in 0). After serum starvation, the cells were stimulated with EGF (50 ng/ml, 10 min) and harvested. Cytosolic extracts were subjected to Western blot analysis with anti-pERK-Ab, anti-general ERK-Ab as control for equal amount of ERKs, and anti-HA-Ab to show the increasing amount of HA-PTP-SL. The intensity of the bands was determined by densitometry, and the amount of dephosphorylation is presented as percent staining intensity of ERKs at each amount of wt-PTP-SL per that without overexpression of wt-PTP-SL. The data is from a representative experiment that was reproduced three times. B, dephosphorylation of ERK1, ERK1b, and their mutants by PTP-SL. COS-7 cells were co-transfected with 1 g of plasmids containing HA-ERK1 (1), HA-ERK1b (1b), HA-CT-ERK1b (CT), HA-Sc-ERK1b (Sc), or HA-339N-ERK1 (339N) together with 0.1 g of plasmid containing either an inactive PTP-SL (C293S, mu PTP-SL) or wt PTP-SL (wt PTP-SL) as indicated. After starvation, the cells were stimulated with EGF (50 ng/ml, 7 min) and harvested. Cytosolic extracts were subjected to Western blot analysis using anti-pERK-Ab and anti-HA-Ab as indicated. The position of endogenous ERK1 and ERK2 and the HA-ERKs is indicated. The data is from a representative experiment that was reproduced three times. C, interaction of ERK1, ERK1b, and their mutants with PTP-SL. COS-7 cells were co-transfected with GFP-ERK1 (1), GFP-ERK1b (1b), GFP-CT-ERK1b (CT), GFP-Sc-ERK1b (Sc), GFP-339N-ERK1 (339N), or an empty vector (Vec) together with wt HA-PTP-SL (wt) as indicated. After starvation, the cells were either stimulated with EGF (50 ng/ml, 10 min) or left untreated. GFP-ERKs were immunoprecipitated with anti-GFP-Ab, and washed extensively (1 ml of RIPA buffer, 2 ϫ 1 ml of 0.5 M LiCl PH-8, 2 ϫ 1 ml of buffer A). Anti-GFP-Ab was used to confirm equal immunoprecipitation of the various constructs. Anti-HA-Ab was used to determine the co-precipitated HA-PTP-SL and to confirm equal expression of HA-PTP-SL in the cytosolic extracts. The positions of GFP-ERKs and HA-PTP-SL are indicated. This experiment was reproduced three times.

FIG. 8. PTP-SL activity toward endogenous ERKs.
A, Rat1 cells were transfected with inactive PTP-SL (C293S, mu) or wt PTP-SL (wt). After starvation, the cells were stimulated with EGF (50 ng/ml) and cytosolic extracts were subjected to Western blotting. Similar amounts of ERKs was detected by anti-general ERK-Ab in the times examined. The phosphorylation of endogenous ERKs was analyzed with anti-pERK-Ab (␣-pERK). The position of endogenous ERKs is indicated. The amount of the transfected wt/mu PTP-SL was determined using anti-HA-Ab; the position of the wt/mu PTP-SL is indicated. B, the relative intensity of each ERK was determined by densitometry. The plots represent mean and S.E. of three experiments.
general ERK-Ab in which the phosphorylated protein can be distinguished from the non-phosphorylated form by its retardation on SDS-PAGE. However, while the amount of doubly phosphorylated ERK was reduced to ϳ20% in later stages (6 -48 h), up to 50% of the ERK1b was retarded upon SDS-PAGE. Since monophosphorylated ERK1b is retarded under these conditions, similar to the retardation of doubly phosphorylated (28), but the monophosphorylated forms cannot be recognized by the anti-pERKs-Ab, the difference between the two indicate that part of the retarded ERK1b is monophosphorylated. These results confirm our previous observation (8) that alteration of the regulatory domain of ERK1 does not significantly change the ability of the ERKs to be phosphorylated upon mitogenic stimulation despite its lack of interaction with MEK1. On the other hand, it seems that the main effect of the 26-amino acid insertion stems from differences in the inactivation phase of the ERKs upon stimulation. In addition to the change in the level of phosphorylation, EGF stimulation also induced higher ERK1b expression that peaked (2.5-fold over basal state levels) 40 h after the addition of EGF (Fig. 9, A and  C). These could evolve from modification in the alternative splicing machinery upon long-term stimulation of the Rat1 cells with EGF as reported in other systems (29). This finding also suggests a particular role for the ERK1b after persistent stimulation of cells as will be discussed below.

DISCUSSION
ERK1b is a 46-kDa alternatively spliced form of ERK1 (7), which might be similar to the poorly characterized kinase that had been tentatively termed ERK4 (30,31). ERK1b appears to be expressed only in some of the tissues examined, and this was different from the uniform pattern of expression of ERK1 and ERK2 (7). Although ERK1b showed a similar trend of activation as ERK1 and ERK2 in most cellular systems, there were several systems in which the regulation of this protein was different from the other two ERKs. Thus, in Ras-transformed Rat1 cells, there was a 7-fold higher expression of ERK1b, which was also more responsive than ERK1 and ERK2 to various extracellular treatments (7). Differential kinetics of ERK1b activation was observed also in LH-and FSH-stimulated granulosa cells (11), in H 2 O 2 -stimulated OLN 93 cells (32), and in antigen-stimulated RBL-2H3 cells (33). These findings raised the question as to what might be the molecular mechanism for the differential regulation. In our previous study (7) we showed that the interaction of ERK1b with MEK1, and as a consequence also its subcellular distribution, is different from that of ERK1 and ERK2. However, it was not clear whether this differential interaction is caused by the sequence of the 26-amino acid insert or by the disruption of a regulatory sequence that is responsible for the association. Here we show that a mutant in which the sequence of the 26-amino acid was scrambled behaves similarly to the wt-ERK1b, while the fusion of the 26 amino acids to the C terminus of ERK1 does not have any effect. The behavior of these mutants strongly suggests that the site of insertion, not the actual sequence of the 26amino acid, is responsible for the differential interaction with MEK1. Therefore, these results indicate that the regulatory domain that is adjacent to the insertion site (CRS/CD) (8,9) is disrupted by the 26-amino acid insertion in ERK1b.
The importance of the region which is modified by the 26amino acid insertion of ERK1b for the regulation of ERKs activity is well documented. It was shown that the substitution of Asp 334 in rolled (dERK), which is the Drosophila's homologue of ERKs (same position as the Asp 339 in ERK1) is a gain of function mutation termed sevenmaker (rl sm (18)). The mutation of the homologous Asp in mammalian ERK2 (Asp 319 in ERK2) also resulted in a gain of function mutant probably due to its resistance to phosphatases (19,25,34). Later, our group showed that amino acids 312-320 of ERK2 (332-340 in ERK1) are responsible for its MEK-induced cytosolic retention in resting cells and therefore it was named CRS (8). The important amino acid for the interaction with MEK1 are Glu 336 , Asp 339 , and Glu 340 of ERK1 (8 -10) but the hydrophobic residues in this region are also important (8,10). Although the mutation in this region prevented the binding of ERKs to MEKs, it did not significantly reduce the activation of ERK by MEK (8 -10). Similar results were obtained with ERK1b, which has similar kinetics of activation as ERK1 both in vitro (7) and in tissue culture upon EGF stimulation (Fig. 9). Therefore, the association between MEKs and ERKs by the CRS/CD might be necessary for other functions of ERKs such as determination of subcellular localization, and not for their activation.
Beside the role of the CRS/CD in the interaction with phosphatases and with MEKs it has been shown that this region is also responsible for the interaction of ERKs with components downstream the cascades such as Mnk1 (9). Here we studied another important substrate for ERK1 and ERK2, namely the transcription factor Elk-1 (35). The C-terminal regulatory re- FIG. 9. Induction and activation of ERKs under prolonged EGF stimulation. A, Western blot analysis of EGF-stimulated Rat1 cells with anti-pERK-Ab and anti-general ERK-Ab. Rat1 cells were serum-starved, and stimulated with EGF (10 ng/ml) for the indicated times. Cytosolic extracts were subjected to Western blotting with the indicated Abs. The position of endogenous ERKs is indicated. B, quantitation of the phosphorylated form of ERKs in A. The phosphorylation of ERK2 and ERK1b is presented as percent of the staining intensity of the upper (shifted, phosphorylated) band in the lower panel of A relative to the total intensity of the respective ERK (shifted and not shifted). C, quantitation of the expression of the endogenous ERKs in A. The expression of ERK2 and ERK1b is presented as fold intensity of each ERK per the intensity before treatment. This experiment was reproduced four times. gion of Elk1 contains multiple consensus phosphorylation sites for MAPKs ((S/T)P (23)). Those sites are phosphorylated by ERKs (13,36,37) and also by other kinases of the MAPK family (38 -41) to cause transactivation of Elk1. Detailed investigation of Elk-1 phosphorylation showed that ERKs can phosphorylate Elk-1 on at least six sites (23). Removal of several phosphoacceptor sites in the C-terminal regulatory region resulted in 50 -80% reduction of Elk1 transcription activity (80% when either Ser 383 or Ser 389 were substituted to alanine (36)). The only exception was Thr 336 , in which alanine substitution caused a 20% increase in Elk1 transcription activity (36).
Here we show that ERK1b can phosphorylate Ser 383 to the same extent as ERK1, but the phosphorylation of the other band-shifting sites by ERK1b is altered (Fig. 4). This reduced phosphorylation of Elk1 is due to the abrogation of the CRS/CD in ERK1b and not because of the actual sequence of the 26amino acid insertion (Fig. 4C). As such, this is the first demonstration for the role of CRS/CD in the interaction of ERKs with Elk1. However, despite its reduced phosphorylation of Elk1, ERK1b actually increased the transactivation of Elk1 as compared with that of ERK1 (Fig. 5). One explanation for this discrepancy might be a lack of phosphorylation of an inhibitory site in Elk1 by ERK1b (e.g. Thr 336 ). However, a more tempting explanation is derived from the results with the D339N-ERK1 that phosphorylates Elk1 as much as ERK1 but its expression induces even higher transactivation of Elk1 than that caused by the ectopic ERK1b (4-fold activation; Fig. 5). Thus, the maximal induction of Elk1 activity that can be achieved due to alteration of the CRS/CD of ERKs is the one achieved by D339N-ERK1. Hence, the reduced phosphorylation by ERK1b correlates well with the lower Elk1 activation by EGF in ERK1b expressing as compared with D339N-ERK1 expressing cells. According to this explanation the phosphorylation that is prevented by ERK1b is of activatory residues of Elk1. However, the effect of the phosphorylation by ERK1b is still higher than that caused by ERK1 and this is probably due to the insensitivity of ERK1b to phosphatases. Thus, the higher residual activity of ERK1b over time, which is due to lack of sensitivity to phosphatases (Fig. 9), may cause a prolonged activation of Elk1 in the 14-h experiment (Fig. 5). However, the net of the prolonged activity of ERK1b toward some of the sites on Elk1 is actually not much different from that of ERK1 and might be of physiological relevance since the higher activation of ERK1b does not induce too much activation of Elk1 upon mitogenic stimulation.
Inactivation of ERKs can occur by three types of protein phosphatases (reviewed in Refs. 27 and 42): PTPs, protein-Ser/ Thr phosphatase (PP), and dual specificity phosphatase (MKPs). Recently, it was shown that shortly after mitogenic stimulation, the inactivation of ERKs occurs mainly by PTPs (28), and PTP-ER was found to play a key role in Ras1 signaling in Drosophila (25). We therefore investigated the interaction of ERK1b with PTP-SL that is closely related to PTP-ER (15). We found a reduced sensitivity of the ERK1b to PTP-SL compared with that of ERK1 (Figs. 6 -8), which was dependent on the insertion site rather than on the insert sequence. Due to the similarity in the mode of interaction with ERKs, it is likely that most PTPs and MKPs cannot properly interact with ERK1b. Indeed, we were recently able to show that ERK1b does not interact with, and is less sensitive to, MKP3 (not shown).
The fact that ERK1b is defective in its CRS/CD region that is responsible for both interaction with the activating MEKs and its inactivating phosphatases, raised the question as to what would be the net effect of the insertion on the activity of the endogenous ERK1b. In a previous report (7) we showed that in persistently stimulated cells (e.g. Ras-transformed Rat-1 cells), ERK1b is the major ERK activated by an acute extracellular stimulation. This could have been due to insensitivity of ERK1b to the large number of phosphatases that are induced under such conditions (43). Here we show that the same is true after a prolonged activation of cells with EGF. We found that the kinetics of ERK1b activation shortly after stimulation are similar to that of ERK1 and ERK2 (Fig. 9, A and B), while the inactivation of ERK1b was slower in later time points. Moreover, the activity of ERK1 and ERK2 drops back to basal level within 6 h, while ϳ50% of ERK1b remained phosphorylated up to 48 h after stimulation. Thus, our results indicate that the main function of the CRS/CD region is its interaction with phosphatases while the lack of interaction with MEKs does not significantly modify its kinetics of activation. The reduced phosphorylation of the Elk1 does not significantly modify the activatory phosphorylation of this transcription factor by ERK1b, and therefore the higher activity of Elk1 induced by ERK1b as compared with ERK1 can be attributed to the lack of sensitivity of phosphatases as well.
In addition to the higher activity of ERK1b we observed also elevation in the expression of ERK1b upon prolonged stimulation (Fig. 9, A and C). The EGF-induced expression of ERK1b resembles the higher expression of ERK1b in Ras-transformed Rat1 cells (7). The elevation of ERK1b expression by these two types of persistent stimulation of the cells suggests that Ras signaling, by inducing many phosphorylation events, affects the splicing machinery to induce expression of splicing variants (44). It should be noted, however, that besides the elevation in the amount of ERK1b, stimulation of the Ras pathway induces also expression of MAPK phosphatases such as MKP3 (43), which strongly compromise the ability of ERK1 and ERK2 to be properly activated by exogenous stimuli (7). Therefore, it is tempting to suggest that an elevated amount of ERK1b, which is insensitive to the induced phosphatases, plays a major role in the transmission of exogenous signals in the persistently stimulated cells where the activity of ERK1 and ERK2 is suppressed by induced phosphatases. As such, the existence of appreciable levels of ERK1b may be important for many of the physiological functions and responses in persistently stimulated cells such as cancer cells or T and embryonic cells.
In summary, we showed here that the insertion of 26 amino acids adjacent to the CRS/CD in ERK1b probably causes its differential regulation. Thus, ERK1b is crippled in its ability to interact with MEK1 and this appears to change its subcellular localization but not its kinetics of activation. In addition, ERK1b also has a slightly modified specificity toward the substrate Elk1 and is substantially less sensitive to phosphatases. These properties probably allow ERK1b to transmit exogenous signals under conditions where ERK1 and ERK2 are downregulated upon persistent stimulation.