TOUSLED is a nuclear serine/threonine protein kinase that requires a coiled-coil region for oligomerization and catalytic activity.

The TOUSLED (TSL) gene is essential for the proper morphogenesis of leaves and flowers in Arabidopsis thaliana. Protein sequence analysis predicts TSL is composed of a carboxyl-terminal protein kinase catalytic domain and a large amino-terminal regulatory domain. TSL fusion proteins, expressed in and purified from yeast, were used to demonstrate TSL protein kinase activity in vitro. TSL trans-autophosphorylates on serine and threonine residues, and phosphorylates exogenous substrates. Using the yeast two-hybrid system, TSL was found to oligomerize via its NH2-terminal domain. A deletion series indicates that a region containing two α-helical segments predicted to participate in a coiled-coil structure is essential for oligomerization. TSL localizes to the nucleus in plant cells through an essential NH2-terminal nuclear localization signal; however, this signal is not necessary for protein kinase activity. Finally, deletion mutants demonstrate a strict correlation between catalytic activity and the ability to oligomerize, arguing that activation of the protein kinase requires interaction between TSL molecules.

During development of many organisms, protein kinases function in signaling pathways important for proper morphogenesis and cell fate determination. Their activity can either be stimulatory, such as that of Sevenless receptor kinase (reviewed in Ref. 1), or inhibitory, as exemplified by the inactivation of cAMP-dependent protein kinase through the Hedgehog cascade (reviewed in Ref. 2). Therefore, to understand the basic cellular processes involved during development requires knowledge of how the relevant kinase(s) is regulated, as well as how its target substrates are affected by phosphorylation.
The TSL gene was identified by mutational analysis in Arabidopsis thaliana as a gene required for proper development of the flower and the margins of the leaf (3). tsl loss of function mutations are recessive and cause a phenotype characterized by two major floral defects. First, there is a stochastic decrease in the number of floral organs, implying that TSL functions at an early stage during flower formation, perhaps regulating the establishment of organ primordia by promoting specific cell divisions within the floral meristem. Second, specific regions of the ovule-housing organ, the gynoecium, fail to develop properly suggesting that TSL also may function to pattern developmental programs within an organ type (3). 1 TSL encodes a 688-amino acid protein (TSL) which is a putative serine/threonine protein kinase with a COOH-terminal catalytic domain (amino acids 409 -688) and an NH 2 -terminal domain (amino acids 1-408) of unknown function (3). Recent data base searches suggest that TSL is a member of an evolutionarily conserved protein kinase subfamily with closelyrelated homologs found in Caenorhabditis elegans, Caenorhabditis briggsae, humans, and maize. 2 The presence of TSL homologs in both plant and animal kingdoms implies the protein performs a fundamental function. Consistent with this hypothesis, TSL is expressed in all organs of the plant (3).
The utilization of the same signaling components in multiple developmental pathways is emerging as a common theme in many organsims (1, 4 -7). The requirement for TSL function at several stages of development suggests that this putative protein kinase is regulated in response to different unidentified developmental cues. The TSL NH 2 -terminal domain contains sequence motifs, such as a coiled-coil region and three consensus nuclear localization signal (NLS) 3 sequences, which together could participate in modulating the activity of the COOH-terminal catalytic domain. The existence of multiple NLS sequences in the NH 2 -terminal domain may direct the subcellular localization of the protein, permitting access to potential regulatory factors and target substrates. The coiledcoil region, including a leucine-zipper motif, may participate in protein-protein interactions that affect kinase activity. Such interactions could include the formation of TSL oligomers. The ligand-binding induced dimerization of receptor protein kinases is known to be critical for kinase activation (8,9). However, oligomerization has only recently been found to provide a possible means of regulation of non-receptor protein kinases (10,11).
To begin to understand the basic properties of the TSL protein kinase and its regulation, we have isolated catalytically active TSL from yeast and characterized that activity in vitro.
In this report, we show that TSL is a nuclear serine/threonine protein kinase and is capable of autophosphorylation in trans. The TSL NH 2 -terminal domain also is shown to mediate oligomerization in the two-hybrid system. Deletion mutants, analyzed for both their enzymatic activity in vitro and ability to oligomerize in the two-hybrid system, indicate that TSL requires self-association for protein kinase activity.

Plasmids
TSL Subclones-Precursor plasmids containing the full TSL coding sequence, and the coding sequence for the NH 2 -terminal domain, and NH 2 -terminal deletions were generated for subcloning into the expression vectors described below. Cloning details will be provided upon request.
K438E Mutation-A polymerase chain reaction product containing the sequence for the catalytic domain with the codon for Lys-438 changed to a codon for Glu was generated by polymerase chain reactiondirected oligomutagenesis using the mutagenic primer 5Ј-GTGC-GAGCTTCATGGTT-3Ј, and the T3 and T7 primers by the method of Bowman et al. (14). The fragment was subcloned into pBluescriptSKϩ at the EcoRI site to generate pTK2E. This plasmid was subsequently used to exchange the mutant fragment with wild-type fragments in the precursor plasmids where appropriate.
GST Fusion Plasmids-YEpLG-GST 4 is a yeast shuttle vector containing a LEU2 and a ␤-lactamase gene for selection in yeast and E. coli, respectively, and a 2-and a ColE1 origin for replication in yeast and E. coli. It also contains the coding sequence for GST under GAL1 control followed by a polylinker. TSL sequences from precursor plasmids were subcloned into the polylinker region of YEpLG-GST such that the TSL coding sequences were in-frame with GST (details will be provided upon request).
Two-hybrid Contructs-TSL sequences from the precursor plasmids described above were subcloned into the two-hybrid vectors pAS1 (12) and pACTII (15) such that the TSL coding sequences were in-frame downstream of the Gal4 DBD or activation domain sequence, respectively (details will be provided upon request).
Myc Epitope Tag Expression Plasmids-The pA6M vector, containing a CaMV35S promoter upstream of the sequence encoding a Myc-epitope tag followed by a polylinker, was constructed by ligating the XhoI-XbaI restriction fragment from pJR1265, 5 containing a translation start followed by a 6x Myc-epitope tag and polylinker, into the plant expression vector, pART7 (16). The sequence for full-length TSL was introduced into pA6M to create pA6M/TSL.2 (for this subcloning only, the sequence at the translational start site of TSL was changed to introduce an NcoI site, resulting in a serine to alanine change at the second residue in TSL to create TSL.2). Deletion mutants were introduced into pA6M by excising appropriate restriction fragments from precursor plasmids and ligating to pA6M.
GUS Expression Plasmids-pRTL2GUS, an expression plasmid containing the GUS reporter gene under the control of the CaMV35S promoter has been described (17). Sequences encoding amino acids 12-438 of TSL were inserted downstream as a BamHI-KpnI fragment replacing the BglII-KpnI Nia fragment from pRTL2SG-Nia/⌬BϩK (17) creating pGUS-NT⌬N12.

Expression and Purification of GST Fusion Proteins
YEpLG-GST constructs were introduced into BJ5460 by chemical transformation (18). For galactose inductions, transformants were grown overnight in 10 ml of SC media lacking leucine and containing 2% raffinose and 0.2% sucrose. 10 ml of fresh media was added and cells grown an additional 4 h. Galactose was added to 2% and cells grown another 4 h. Cells were pelleted by centrifugation and resuspended in lysis 250 buffer (50 mM Tris (pH 7.4), 250 mM NaCl, 5 mM EDTA, 0.1% Nonidet P-40, 50 mM sodium fluoride, supplemented with phenylmethylsulfonyl fluoride (100 g/ml), aprotinin (2 g/ml), leupeptin (2 g/ml), pepstatin A (1 g/ml), dithiothreitol (1 mM), and benzamidine (1 mM)) prior to freezing. Glass beads were added to thawed samples and cells lysed by vortexing. Extracts were cleared by centrifugation (14,000 ϫ g for 20 min at 4°C) and protein concentration of supernatants estimated by OD 280 . Equal protein amounts were mixed with glutathione-agarose beads (Sigma) for 1 h at 4°C. Bound fractions were washed extensively with lysis 250 buffer. For all purifications, a ratio of 1 mg of protein extract/100 l bed volume of glutathione-agarose beads was used.

Western Blot Analysis
Bound GST fusion proteins from equivalent aliquots of glutathioneagarose beads (10-l bed volume) purified as above were eluted by boiling in Laemmli sample buffer and separated by SDS-PAGE (19). For Western blots in Figs. 2A and 5A, the entire sample was loaded, and in Fig. 7A, one-half volume of eluted proteins was loaded. Gels were transferred to Immobilon-P and immunoblots were performed using rabbit anti-GST polyclonal antiserum, horseradish peroxidase-conjugated anti-rabbit IgG (Bio-Rad), and detection with enhanced chemiluminescence (Amersham). Blots were treated with Western blocking buffer (50 mM Tris, pH 7.6, 150 mM NaCl, 0.05% Tween-20, 3% non-fat dry milk) for at least an hour, incubated overnight in primary antibody (1:1000) at 4°C, washed 3 times 20 min each at room temperature, incubated 1 h in secondary antibody (1:3000), washed 3 times 20 min each. All incubations and washes were performed in Western blocking buffer. After the final wash, blots were rinsed in Western blocking buffer without milk and subjected to chemiluminescent detection.

Kinase Assays
Equivalent aliquots of glutathione-agarose beads (10-l bed volume) with bound GST fusion proteins purified as above were washed three times in kinase buffer (50 mM HEPES, pH 7.6, 150 mM NaCl, 10 mM MgCl 2 , 2 mM MnCl 2 ) and resuspended in 37.5 l of kinase buffer. 10 Ci of [␥-32 P]ATP (Amersham, 3000 Ci/mmol) was added and the reaction was incubated at room temperature for 60 min. Laemmli sample buffer was added, samples were boiled for 3 min, centrifuged, and the supernatant was analyzed by SDS-PAGE. Kinase assays in Fig. 7B were washed in kinase buffer after the reaction prior to elution in Laemmli sample buffer. In kinase assays in Fig. 2, B and D, and Fig. 5B, the entire reaction was loaded, whereas in Fig. 7B, one-fourth volume of the reaction was loaded.

Phosphoamino Acid Analysis
Aliquots of purified GST-TSL were allowed to autophosphorylate as described above. Pellets were washed once in kinase buffer, samples were boiled after addition of Laemmli sample buffer and electrophoresed on an SDS-polyacrylamide gel. The portion of the gel containing the phosphorylated protein was transferred to Immobilon-P (Amersham) by electroblotting. The 32 P-containing portion of the membrane was excised, and the sample was hydrolyzed by incubating the membrane in 6 N HCl for 1 h at 100°C (20). The supernatant was dried, redried twice after resuspension in H 2 O, and analyzed by TLC and autoradiography by the method of Cooper et al. (21).

Quantitation of ␤-Galactosidase Activity
␤-Galactosidase activity in yeast was quantified with chlorophenyl red-␤-D-galactopyranoside (CPRG; Boehringer Mannheim). Y153 transformants were grown in 3 ml of SC media lacking tryptophan and leucine to OD 600 1.0 -1.5. Cells were then prepared and permeabilized as described (22), except cell pellets were resuspended in 900 l of H buffer (100 mM HEPES, 150 mM NaCl, 2 mM MgCl 2 , 1% bovine serum albumin, pH 7.0) and 100 l of 50 mM chlorophenyl red-␤-D-galactopyranoside was added following permeabilization. The amount of liberated chlorophenyl red was determined by OD 574 .

TSL Domain Structure and Deletion Constructs-The
COOH-terminal domain of TSL contains the consensus sequences shared by the large family of serine/threonine and tyrosine protein kinases (3) (Fig. 1). In addition to the COOHterminal catalytic domain of TSL, the primary structure of the NH 2 -terminal domain contains several regions which, by homology to other proteins, may act as functional domains. These regions are depicted in Fig. 1, and include a glutamine repeat region, three putative NLS, and a coiled-coil region containing two amphipathic ␣-helical segments, the second of which also harbors a leucine-zipper motif. The site of the tsl-2 point mutation (3) is also indicated. Constructs progressively deleting these putative functional domains ( Fig. 1) were used throughout this study to determine the contribution of each domain to the biochemical properties of TSL. As a control for protein kinase activity, TSL containing a point mutation converting Lys-438 of TSL to Glu (K438E) was also constructed (Fig. 1). Lys-438 corresponds to a similarly located lysine in the Mg 2ϩ -ATP binding site of the catalytic subunit of bovine cAMP-dependent protein kinase-␣ (Lys-72) (25,26), and mutation of this conserved lysine has been shown to abolish catalytic activity in a number of kinases (Ref. 27, and references therein).
TSL Is a Serine/Threonine Protein Kinase-Eukaryotic protein kinases commonly are inactive when isolated from prokaryotic expression systems. To increase the likelihood of isolating active enzyme, full-length TSL was produced in Saccharomyces cerevisiae as a GST fusion protein using a yeast expression vector, YEpLG-GST. 4 A negative control plasmid containing the K438E mutation also was constructed. Fusions are under the transcriptional control of the inducible GAL1 promoter. These plasmids were introduced into the protease-deficient yeast strain BJ5460. Transformants were induced with galactose, and fusion proteins were purified from cell extracts with glutathione-agarose (28). Extracts from uninduced cells show no expression of the TSL fusion proteins ( Fig.  2A, lanes 1 and 2), whereas purified extracts of induced cells expressing either wild-type or mutant TSL each contain a protein of the expected molecular mass (ϳ100 kDa) ( Fig. 2A, lanes 3 and 4, respectively) as detected by anti-GST polyclonal antiserum.
Purified TSL and K438E fusion proteins were used in kinase assays to determine possible catalytic activity. Uninduced cell extracts also were mock-purified and shown to contain no contaminating protein kinase activity (Fig. 2B, lanes 1 and 2). Purified TSL fusion protein autophosphorylates as evidenced by [ 32 P]phosphate incorporation into the 100-kDa protein (Fig.  2B, lane 3) identified as TSL ( Fig. 2A, lane 3). In contrast, the mutant K438E (Fig. 2B, lane 4) or pre-boiled TSL (Fig. 2B, lane  8) show no incorporation. The autophosphorylation of TSL may account for the slight decrease in mobility compared with that of mutant K438E (Fig. 2A, compare lanes 3 and 4). TSL autophosphorylation activity is dependent on either of the divalent cations Mg 2ϩ or Mn 2ϩ (Fig. 2B, lanes 5-7).
Phosphoamino acid analysis was performed on purified TSL fusion protein which had undergone autophosphorylation (Fig.  2C). [ 32 P]Phosphate is incorporated into both phosphothreonine and phosphoserine. Precise mapping of the autophosphorylation sites in TSL will be necessary to determine the number and location of modified residues. Control experiments showed that no [ 32 P]phosphate was incorporated into the GST portion of the autophosphorylated TSL fusion protein (data not shown).
TSL Localizes to the Plant Cell Nucleus-Having established TSL is an authentic protein kinase, next we analyzed the potential functions of the NH 2 -terminal domain. The TSL sequence contains three putative NLS, all in the NH 2 -terminal domain (Fig. 1). To determine the subcellular localization of TSL in plant cells, the TSL coding sequence was fused to sequences encoding a Myc-epitope tag present in a plant expression vector, pA6M. In this vector, expression of tagged proteins is under the control of the constitutive CaMV35S promoter. The A6M.TSL.2 plasmid was electroporated into tobacco protoplasts, and cells incubated for 18 -24 h to allow expression of the fusion protein. Indirect immunofluorescence was performed using the anti-Myc monoclonal antibody, 9E10 (29). The Myc-TSL fusion protein is detected exclusively in the nucleus of tobacco cells (Fig. 3, A and B). To determine if the NH 2 -terminal domain is sufficient to direct nuclear localization in plant cells, sequences encoding amino acids 12-438 of TSL were fused in-frame downstream of the ␤-glucuronidase (GUS) reporter gene to create pGUS-NT⌬N12. When GUS alone is expressed in tobacco protoplasts, enzymatic activity is found only in the cytoplasm (Fig. 3D). When the GUS-NT⌬N12 fusion protein is expressed in protoplasts, GUS activity is nuclear (Fig. 3C), indicating that the NH 2 -terminal domain of TSL is sufficient for nuclear targeting of the protein. Together, these results demonstrate that TSL is a nuclear protein, and that one or more of the NH 2 -terminal NLS motifs likely confers localization.
The Coiled-coil Region Is Required for TSL Oligomerization-The NH 2 -terminal domain of TSL contains two segments predicted to form ␣-helices which each may participate in a coiled-coil structure (3) (cI and cII in Fig. 1). The second of these segments contains a leucine-zipper motif. By analogy to other proteins, this coiled-coil region may be involved in proteinprotein interactions (30,31). To test if TSL can self-associate via either of these ␣-helical segments, the yeast two-hybrid system (32) was used (Fig. 4). Full-length TSL could not be used in these experiments, as a construct containing the NH 2 terminus fused to the Gal4 DNA-binding domain (DBD) weakly activates the GAL1 promoter (data not shown). However, the near full-length deletion ⌬N73-DBD which lacks the glutamine repeat, does not activate the promoter, allowing its use in the assay (Fig. 4A, ⌬N73-DBD plus no insert control). When ⌬N73 is co-expressed as both a DBD and an activation domain fusion, the reporter gene is activated (Fig. 4A), indicating that TSL oligomerizes. To identify the region(s) necessary for this interaction, TSL deletions were inserted into both the activation domain and DBD containing plasmids, and tested for selfassociation as well as for interaction with ⌬N73. ⌬N171, with an intact coiled-coil region, retains the ability to interact both with itself and with ⌬N73 (Fig. 4A). When the first ␣-helical segment and part of the second is removed (⌬N304), however, oligomerizing activity is lost, even though the full leucinezipper is still present. All further deletions, including ⌬N403 which retains the catalytic domain, fail to self-oligomerize or interact with ⌬N73 (Fig. 4A). Thus, the first ␣-helical segment of the coiled-coil region is apparently essential for TSL oligomerization.
Further evidence for the involvement of the coiled-coil region 1 and 3) or GST-K438E (lanes 2 and 4) plasmid. Equivalent amounts of protein extracts from uninduced (lanes 1 and 2) or induced (lanes 3 and 4) cells were affinity purified on glutathione-agarose, and the bound fractions were separated by SDS-PAGE, transferred to a nylon membrane, and probed using an anti-GST antiserum. Immunoblotting was performed using horseradish peroxidase-conjugated secondary antibody and enhanced chemiluminescent detection. B, kinase assays of purified samples. Equivalent aliquots of extracts purified in A were used in each lane. Lanes 1-4 are assayed in kinase buffer (with 10 mM MgCl 2 , 2 mM MnCl 2 ) and loaded as in A. Lanes 5-8, purified GST-TSL from induced cells was tested in different conditions. In lane 5, the kinase buffer contained 1 mM EDTA and no Mg 2ϩ or Mn 2ϩ . In lanes 6 and 7, the kinase buffer contained 10 mM MnCl 2 , or 10 mM MgCl 2 , respectively. In lane 8, the purified GST-TSL was boiled for 3 min before the kinase assay. C, purified GST-TSL was allowed to autophosphorylate using [␥-32 P]ATP and the protein was subjected to phosphoamino acid analysis. Arrows show the positions of phosphoamino acid standards. D, MBP (1 g) (lanes 1 and 2), histone (10 g of Type IIIS) (lanes 3 and 4), and partially dephosphorylated casein (1 g) were added to kinase assays containing purified GST-TSL (lanes 1, 3, and 5) or GST-K438E (lanes 2, 4, and 6). Autoradiograms of the SDS-polyacrylamide gel in B and D, and the thin-layer chromatogram in C are shown. The results presented are representative of at least two independent experiments. in TSL oligomerization was found by using deletion mutants lacking the catalytic domain in the two-hybrid assay (Fig. 4B). When the entire NH 2 -terminal domain (NT) (amino acids 1-406) (activation domain fusion) is co-expressed with the NH 2 -terminal domain deletion NT⌬N73 (amino acids 73-406) (DBD fusion), a strong interaction is observed, indicating that the NH 2 -terminal domain alone can self-associate (Fig. 4B). NT⌬c/c, an internal deletion which removes most of the coiledcoil region (from amino acids 172-327), fails to interact with NT⌬N73 confirming the importance of this region in mediating oligomerization (Fig. 4B). Interestingly, when NT⌬N73 association with ⌬N73 (containing the catalytic domain) was tested, no interaction is observed (Fig. 4B). Thus, the presence of the catalytic domain in only one of the partners in some way precludes the formation of heterotypic oligomers. Alternatively, the NH 2 -terminal domain alone may adopt a different conformation that prevents interaction with full-length TSL. Nevertheless, the results indicate that TSL can oligomerize, and that this ability is dependent on an intact coiled-coil region.

FIG. 2. Purification and characterization of TSL protein kinase activity. A, Western analysis of fusion proteins purified from yeast cells containing either the GST-TSL (lanes
Catalytic Activity of Deletion Mutants Correlates with Ability to Oligomerize-The deletion mutants next were expressed and purified as GST fusions from yeast and tested for their ability to autophosphorylate and to transphosphorylate MBP. Western blot analysis of purified fusion proteins probed with anti-GST antiserum showed that all the deletion mutants are synthesized at similar levels (Fig. 5A). Kinase assays show that the two mutants capable of oligomerization, ⌬N73 and ⌬N171, also are both capable of autophosphorylation and transphosphorylating MBP (Fig. 5B, lanes 2 and 3, respectively) at levels comparable to wild-type. Deletions affecting the coiled-coil region, however, result in catalytically inactive fusion proteins ( Fig. 5B, lanes 4 -7). These results imply that the NH 2 -terminal domain in some way imparts a positive effect on the activity of the catalytic domain. The strict correlation between kinase activity and oligomerization seen with these mutants strongly suggests that oligomerization mediated by the coiled-coil region is required for TSL activity.
Subcellular Localization of TSL-deletion Mutants in Plant Cells-The two catalytically active TSL-deletion mutants, ⌬N73 and ⌬N171, differ in the NLS consensus sequences they retain. ⌬N73 contains all three putative NLS sequences (see Fig. 1). The first and third NLS consensus sequences are SV40 T-antigen-type signals as defined by Chelskey et al. (33), whereas the second is a putative bipartite signal (34). ⌬N171 lacks the first and second putative NLS sequences, but retains the third potential NLS near the catalytic domain. To test if the subcellular distribution of ⌬N73 and ⌬N171 differs, a Mycepitope tag was added at their respective NH 2 termini and fusion proteins were transiently expressed in tobacco protoplasts. Nuclei were visualized by 4,6-diamidino-2-phenylindole staining (Fig. 6, B and E) and bright-field imaging (Fig. 6, C  and F). The epitope-tagged TSL mutants were localized by indirect immunofluorescence using the 9E10 monoclonal antibody (Fig. 6, A and D). As expected, ⌬N73 is nuclear-localized (Fig. 6A), as is the full-length protein (Fig. 3A). However, ⌬N171 is excluded from the nucleus in expressing cells, and is found in the cytoplasm which surrounds the nucleus and the vacuole (Fig. 6D), indicating the third putative NLS is not sufficient to direct the protein to the nucleus. These results also suggest that nuclear localization is not necessary for activation of the catalytic domain as the ⌬N171 protein is an active protein kinase in vitro. Smaller, catalytically inactive deletion mutants, including ⌬N304 and ⌬N403, were found in both the cytoplasm and nucleus presumably because the molecular mass of the mutant proteins is at or below the reported size exclusion limit of the nuclear pore complex (ϳ40 kDa, reviewed in Ref. 35) (data not shown).
Trans-phosphorylation Occurs in TSL Complexes-In many cases autophosphorylation of a protein kinase has been shown to be an intermolecular phosphorylation event (reviewed in Ref. 36). To test if TSL trans-autophosphorylates, extracts containing the kinase-dead mutant K438E were mixed and copurified with ⌬N171, the largest active deletion mutant that clearly resolves from the full-length K438E by SDS-PAGE (Fig.  7A). As expected, purified GST-⌬N171 is active (Fig. 7B, lane  1), whereas GST-K438E is not (Fig. 7B, lane 2). A kinase assay of the copurified samples reveals that ⌬N171 can transphosphorylate K438E (Fig. 7B, lane 3), and K438E is apparently comparable to the wild-type protein as a substrate. This result demonstrates that autophosphorylation of TSL is, at least in part, a trans-phosphorylation event. Precise mapping of the phosphorylation sites on both the active and inactive kinases will be necessary to determine if some sites modified during autophosphorylation also are due to intramolecular events.
To ask if catalytic activity is required for TSL oligomerization, we tested the interaction ability of the K438E mutant in the two-hybrid system (Fig. 7C). ⌬N73-K438E can self-associate (Fig. 7C) and activates the lacZ reporter gene to a similar level as the wild-type homotypic interaction (15.8 versus 23.8 Miller units, see Fig. 4A). This indicates that TSL protein kinase activity is not necessary for self-oligomerization, as earlier suggested by the interaction observed between TSLtruncations lacking the catalytic domain (Fig. 4B). Interestingly, when tested with ⌬N73 and ⌬N171, the ⌬N73-K438E protein showed a much stronger interaction than either of these wild-type proteins with themselves (Fig. 7C, and see Fig.  4A). This could imply that reciprocal autophosphorylation events weaken the stability of the wild-type TSL complex. However, if this were so, it would be expected that the K438E homotypic interaction would also be stronger than wild-type, which is not observed (see above). Also, ⌬N73-K438E displays a weak interaction with NT⌬N73 (Fig. 7C) unlike wild-type ⌬N73 (Fig. 4B). This observation is confirmed by qualitative ␤-galactosidase assays (data not shown). This result suggests that the presence of an inactive (or unphosphorylated) catalytic domain affects the conformation of the NH 2 -terminal domain to a lesser extent than wild-type (see above). In summary, TSL can phosphorylate and form a heterotypic complex with K438E, and this complex is apparently more stable than either homotypic interaction.

DISCUSSION
The loss of TSL gene function results in a mutant phenotype affecting both leaf and flower morphology, and TSL appears to function during both early and late stages of flower development (3). These results suggest that the TSL gene product participates in a commonly used developmental pathway. Sequence analysis predicted TSL to be a serine/threonine protein kinase composed of a carboxyl-terminal catalytic domain and a large amino-terminal regulatory domain. We have shown here that TSL is indeed a protein kinase which autophosphorylates on both serine and threonine residues. That a kinase-dead mutant can serve as a substrate for the wild-type protein demonstrates that some, if not all, sites are phosphorylated in trans. Furthermore, TSL can phosphorylate exogenous substrates such as MBP and casein in vitro.
TSL function presumably is tightly controlled during Arabidopsis development. We have shown that the TSL NH 2 -terminal domain contributes at least two regulatory activities to the protein. First, this domain contains three NLS sequences, at least one of which appears to be functional in targeting TSL to the plant cell nucleus. Deletion of the first two NLS consensus sequences results in a cytoplasmically localized protein, suggesting that one or both of these NLS sequences are essential for nuclear localization. The third NLS is not sufficient for nuclear targeting. Catalytic activity is not required for the nuclear localizing function as the NH 2 -terminal domain alone efficiently targets a heterologous protein to the plant cell nucleus. Reciprocally, the deletion mutant ⌬N171 is not nuclearlocalized, but is catalytically active in vitro, suggesting that nuclear localization is not required for catalytic activity.
The second regulatory function of the NH 2 -terminal domain is to mediate TSL oligomerization. Deletions studies indicate that at least the first ␣-helical segment of the coiled-coil region is critical for TSL oligomerization. The leucine-zipper motif found in the second ␣-helical segment is not sufficient for oligomerization, and its contribution to TSL function is not yet clear. Importantly, deletion mutants revealed a strict correlation between oligomerization in the two-hybrid system and the ability to both autophosphorylate and transphosphorylate exogenous substrates in vitro. That the catalytic domain of TSL alone is inactive indicates that the NH 2 -terminal domain must positively regulate the protein. Together, these data strongly suggest that oligomerization of TSL is required for activation of the catalytic domain. Although this interaction likely is due to direct binding of TSL molecules, it is possible that a cofactor present in the yeast cell mediates this apparent self-interaction.
The kinase-dead K438E mutant can interact with itself and Ligand-mediated dimerization is generally required for activation of receptor protein kinases (reviewed in Refs. 8 and 9). However, oligomerization is not a common mechanism of regulation for a non-receptor kinase, such as TSL. Recent results suggest that oligomerization plays a role in Raf-1 kinase activation (10,11). Several other cytoplasmic protein kinases, including double-stranded RNA-dependent protein kinase (37) and Type I and Type II cGMP-dependent protein kinases (38,39), have been shown to exist as dimers, although the functional relevance of dimerization is not clear. Studies with double-stranded RNA-dependent protein kinase deletion mutants demonstrated that the catalytic domain alone, although unable to dimerize, was fully active in vivo (40). In the case of the cGMP-dependent protein kinases isoforms, it is not known if dimerization is required for catalytic activity. Finally, the crystal structure of the regulatory domain of Lck, a Src family member, revealed that this domain dimerized, but again, it is unknown whether dimerization plays a role in regulating kinase activity (41).
The apparent dependence of TSL catalytic activity on oligomerization argues that TSL self-association via the NH 2terminal domain transmits a conformational change to the catalytic domain, analogous to the effect of dimerization on receptor tyrosine kinases (reviewed in Ref. 9). Dimerization per se is insufficient for activation, as GST can dimerize (42), and therefore, even the catalytically inactive GST-TSL deletion mutants may exist as dimers. TSL self-association mediated by its coiled-coil region, then, may not act merely to bring the catalytic domains into proximity with one another, but instead may cause the protein to adopt a highly specific structure. This conformational change may either allow the catalytic domain to take on a fully active structure, or permit an intermediate that allows one TSL molecule to activate another via transautophosphorylation. This is in contrast to other examples where activation of a protein kinase can be induced by fusion to a heterologous dimerization domain (Refs. 10 and 11, and reviewed in Ref. 8), including fusion to GST (43). However, we cannot exclude the possibility that TSL activation requires complexes larger than dimers, and thus dimerization via GST might be insufficient for activation.
How, then, might TSL oligomerization be controlled? Two obvious possibilities exist. First, the protein may spontaneously form oligomers with a subsequent modification(s) serving as the rate-limiting step for kinase activation. Second, the protein could normally adopt a structure that is unable to oligomerize. An activating event, such as ligand-binding or phosphorylation of a critical residue(s), would then occur to allow TSL self-association. In this scenario, oligomerization could be rate-limiting and/or may require additional modifications for kinase activation. Precedence for the latter mecha-  A). The results in A and B are representative of two independent experiments. C, two-hybrid interaction assay. TSL deletions with or without the K438E mutation as indicated were inserted into the Gal4 DNA-binding domain and activation domain plasmids. Y153 transformants were analyzed for ␤-galactosidase activity (expressed as Miller units). Numbers represent the average of results from three independent transformants. nism comes from the ligand-mediated dimerization of receptor protein kinases known to be critical for their activation (reviewed in Refs. 8 and 9). If such an activation mechanism is used by TSL, analogous activating components must also be present in yeast.
Active TSL proteins have the ability to autophosphorylate, and as mentioned, trans-autophosphorylation may activate recipient TSL molecules. In this regard, protein kinases can be divided into two broad groups, the RD and non-RD kinases (36). RD kinases, including all tyrosine kinases and many serine/threonine kinases contain an arginine immediately preceding the catalytic aspartate (subdomain VIb). A subset of RD kinases require phosphorylation of an amino acid within the "activation segment," a region between the conserved DFG (subdomain VII) and XPE (subdomain VIII), for catalytic activity (reviewed in Ref. 36; subdomains from Ref. 44). TSL is a non-RD kinase, however, as the amino acid preceding the conserved catalytic aspartate is a tyrosine (Tyr-538). Of those studied, non-RD kinases do not require modification in the activation segment. Therefore, we assume that if autophosphorylation is required for activation in TSL, a different region is involved. Precise mapping and mutation of the autophosphorylation sites is currently underway to address this issue.
Loss-of-function tsl mutations affect several aspects of plant development, arguing the protein kinase may play different roles in specific tissues and during different stages of development. Input from the various signaling pathways could modulate TSL function by affecting its oligomerization state and therefore, the catalytic activity of the protein. Identification of signals that regulate TSL and the relevant downstream, presumably nuclear, substrates is essential. Recent genetic studies have shown that TSL may interact with several proteins involved in the control of floral organ number and regional specification during gynoecium development. 1 These include ETTIN (45), PERIANTHIA (46), and LEUNIG (47). The availability of catalytically active TSL protein kinase will allow future studies to determine whether these and other potential interactions are direct and if so, how the proteins may regulate one another.