Phosphorylation of the Transactivation Domain of Pax6 by Extracellular Signal-regulated Kinase and p38 Mitogen-activated Protein Kinase*

The transcription factor Pax6 is required for normal development of the central nervous system, the eyes, nose, and pancreas. Here we show that the transactivation domain (TAD) of zebrafish Pax6 is phosphorylated in vitro by the mitogen-activated protein kinases (MAPKs) extracellular-signal regulated kinase (ERK) and p38 kinase but not by Jun N-terminal kinase (JNK). Three of four putative proline-dependent kinase phosphorylation sites are phosphorylated in vitro. Of these sites, the serine 413 (Ser413) is evolutionary conserved from sea urchin to man. Ser413 is also phosphorylatedin vivo upon activation of ERK or p38 kinase. Substitution of Ser413 with alanine strongly decreased the transactivation potential of the Pax6 TAD whereas substitution with glutamate increased the transactivation. Reporter gene assays with wild-type and mutant Pax6 revealed that transactivation by the full-length Pax6 protein from paired domain-binding sites was strongly enhanced (16-fold) following co-transfection with activated p38 kinase. This enhancement was largely dependent on the Ser413 site. ERK activation, however, produced a 3-fold increase in transactivation which was partly independent of the Ser413 site. These findings provide a starting point for further studies aimed at elucidating a post-translational regulation of Pax6 following activation of MAPK signaling pathways.

Pax6 is a member of the paired box-containing Pax 1 gene family of transcription factors containing nine human (PAX1-PAX9) and murine (Pax1-Pax9) family members (1). The paired box was first discovered in Drosophila as encoding a conserved 125-128 amino acid paired domain unique to this family of developmental control genes (2). Pax6 was initially cloned from human (3), mouse (4), zebrafish (5), and quail (6). Subse-quently, the Drosophila eyeless gene was shown to be a Pax6 homolog and Pax6 homologs have now been described in other invertebrates such as flatworm, ribbonworm, Caenorhabditis elegans, squid, sea urchin, and ascidian (reviewed in Ref. 7) as well as in amphioxus (8). Pax6 is expressed in the developing central nervous system, the eyes, nose, and pancreas in higher vertebrates (4,5,9,10) and plays a pivotal role in the development of these organs (7,11,12). Loss of a functional Pax6 allele results in the Mendelian syndromes aniridia, Peter's anomaly, and congenital cataracts in man (13) and Small eye in rodents (14). Pax6 acts high up in the regulatory hierarchy controlling eye development in both vertebrates and invertebrates (reviewed in Ref. 15). Eyeless, ribbonworm-, squid-, ascidian-, zebrafish-, or mouse Pax6 are all able to induce ectopic eyes in Drosophila upon targeting their expression to different imaginal discs (16 -20). We have recently found that zebrafish contain two Pax6 genes, Pax6.1 and Pax6.2, which are expressed in both overlapping and distinct regions during development of the eyes and the central nervous system. Both these genes are able to induce ectopic eyes in Drosophila (18).
The paired domain is a bipartite DNA-binding domain containing an N-and a C-terminal subdomain each with a helixturn-helix motif (21,22). Pax6, like Pax3, Pax4, and Pax7, also harbors a second DNA-binding domain, the paired-type homeodomain (2,7,23). In Pax6 this domain is separated from the N-terminal located paired domain by a flexible, acidic linker region (5,7). The region C-terminal to the homeodomain is enriched in proline, serine, and threonine residues (PST-rich) and acts as a transactivation domain (TAD) (24 -27).
It has previously been shown that quail Pax6 proteins expressed in the neuroretina are phosphoproteins and phosphoamino acid analysis revealed phosphoserine and a minor proportion of phosphothreonine (28). The activity of many transcription factors is regulated in a rapid and reversible manner by specific phosphorylation events mediated by protein kinases acting in signaling cascades initiated by extracellular stimuli (reviewed in Refs. 29 and 30). Mitogen-activated protein kinases (MAPKs) are proline-directed serine/threonine protein kinases activated by heterogenous extracellular stimuli, including growth factors, hormones, cytokines, antigens, and physical-chemical stimuli such as oxidative stress, heat shock, osmotic imbalance, and UV light. MAPK cascades play essential roles in regulating many critical cellular processes, including cell growth and division, differentiation, apoptosis, and stress-related responses (reviewed in Refs. 31 and 32). These cascades are organized into modules of three protein kinases where a MAPK kinase kinase (e.g. Raf-1) activates a MAPK kinase (e.g. MEK1) which subsequently activates a MAPK (e.g. ERK1) (31,33). Following activation the MAPKs translocate to the nucleus to phosphorylate nuclear substrates. Currently, four distinct MAPK cascades are known in vertebrates. How-ever, more will probably be found since the budding yeast contains five such cascades that orchestrate responses to different physiological stimuli (34). The extracellular-signal regulated kinase (ERK) pathway mainly conveys signals from mitogenic and differentiation stimuli. The same may be true for the most recently discovered ERK5/BMK1 pathway (35) while the Jun N-terminal kinase (JNK) and p38 MAP kinase pathways seems mostly involved in transducing various stress-and cytokine-triggered signals to the nucleus. In each pathway several MAPK isoforms have been found with ERK1 and -2 in the ERK pathway, JNK1-3 in the JNK pathway and p38␣-␦ in the p38 MAPK pathway (reviewed in Ref. 31). A growing number of transcription factors have been identified as nuclear targets for MAPK pathways (reviewed in Ref. 30). Thus, c-Myc (36), Spi-B (37), BCL-6 (38), Microphthalmia (39), and several ETS-domain transcription factors such as vertebrate c-Ets1 and -2 (40), C. elegans LIN-1 (41) and Drosophila Yan and Pointed P2 (42) as well as the C. elegans winged-helix factor LIN-31 (41) are phosphorylated by ERKs. c-Jun (43,44) and NFAT4 (45) are targets of JNK, while CHOP (46) is a nuclear substrate for p38 kinase. This picture is further complicated by the fact that several transcription factors have been shown to be the targets of more than one MAPK pathway. Elk-1, a member of the ternary complex subfamily of ETS proteins, is targeted by all three pathways, while SAP-1 is activated upon phosphorylation by ERK or p38 kinase (Ref. 47 and references therein). Furthermore, MEF2C is phosphorylated by both p38 kinase and ERK5/BMK1 (48) and ATF-2 and ATFa are targets of both the JNK and p38 kinase pathways (49).
Here we report for the first time that a Pax family transcription factor is phosphorylated in vitro and in vivo by MAP kinases. We found that three of the four putative MAPK phosphorylation sites in the Pax6 TAD is phosphorylated by ERK and p38 kinase, but not JNK, in vitro. One of these sites, Ser 413 , is conserved in Pax6 proteins from sea urchin to man and is also phosphorylated in vivo following activation of ERK or p38 kinase. Transient transfection studies with the Pax6 TAD fused to the DNA-binding domain of GAL4 and in the context of the full-length Pax6 protein revealed that mutation of this site to alanine strongly decreased the transactivating potential. Co-transfection of Pax6 with constitutively active MEK1 (which activates ERK1 and -2) or p38 kinase together with constitutively active MKK6b enhanced transcriptional activation by Pax6. However, while the potent p38 kinase-mediated increase in transactivation (16-fold) was largely dependent on the integrity of the Ser 413 phosphorylation site, ERK activation gave a more modest enhancement of transactivation (3-fold) which was only reduced by 30% when this site was mutated to alanine. This suggests that activation of p38 kinase has a more direct positive effect on transactivation due to phosphorylation of the Pax6 TAD at Ser 413 while ERK activation is less efficient and acting more indirectly.
Specific PCR primers (G4-P6(299) and pSG424.3Ј; see Table I) were used to transfer selected Pax6 constructs from pSG424 to pFA-CMV (Stratagene), where the expression of the GAL4-Pax6 fusions is driven by a CMV promoter. All constructs were verified by sequencing and the expression and correct sizes of fusion proteins following transfection of HeLa cells were verified by Western blotting.

5Ј-TGGAATTCAGGAAAAATTTTCACGCTTGAGTTCACAGCTCGAGTA-3Ј
In Vitro Mutagenesis-In vitro mutagenesis was performed according to the instruction manual for the "Quick-Change Site-Directed Mutagenesis Kit" (Stratagene). All mutagenized constructs were checked by sequencing. The specific mutagenesis primers used are shown in Table I.
In Vitro Phosphorylation Assays-GST-Pax6 fusion proteins were purified from Escherichia coli LE392 extracts using glutathione-agarose beads (Pharmacia). The proteins were not eluted, but left on the beads and stored at 4°C in phosphate-buffered saline containing 1% Triton X-100. To equalize the amounts of proteins used in the kinase assays the concentrations of GST fusion proteins were estimated by Coomassie Blue staining after sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The beads were washed twice with the respective kinase buffers before use in vitro kinase assays. GST-Pax6containing agarose beads (5-20 l) were mixed with either: 1) 2 l of 10 ϫ MAPK buffer (500 mM Tris-Cl, pH 7.5, 100 mM MgCl 2 , 10 mM dithiothreitol, 10 mM EGTA), 2 l of 1 mM ATP, 0.5 l of [␥-32 P]ATP (100 Ci/mol), and 1 unit of recombinant active ERK2 kinase (New England Biolabs) in a total volume of 20 l, or with 2) 4 l of 10 ϫ p38 kinase reaction buffer (250 mM Hepes, pH 7.5, 100 mM magnesium acetate), 500 M ATP, 0.5 l of [␥-32 P]ATP (100 Ci/mol), and 1 l (1 g) of recombinant active p38 kinase (Stratagene) in a total volume of 40 l. The reactions were left at 30°C for 30 min, flicking the tubes every fifth min. The agarose beads were subsequently washed with 500 l of phosphate-buffered saline (PBS) before boiling in 25 l of 2 ϫ SDS-PAGE gel loading buffer. The proteins were loaded on a 10% SDS-polyacrylamide gel, and run at 20 mA for 1.5-2 h in a Tris glycine buffer. The gel was dried, and exposed for autoradiography. Assays with c-Jun N-terminal kinase (JNK) was performed as described previously (52).
Analyses of in Vivo Phosphorylation of GST-and GAL4-Pax6 Fusion Proteins-Subconfluent NIH3T3 cells were transfected with 10 g of pCI-GST-Pax6(353-437) or phosphorylation site mutants in 10-cm diameter cell culture dishes using LipofectAMINE (Life Technologies, Inc.) according to the instructions of the manufacturer. Following 4 h of incubation with DNA, the cells were washed 3 times in PBS and left for 8 h in Dulbecco's modified Eagle's medium supplied with 0.1% serum. After 3 washes in Tris-buffered saline, pH 7.5, the cells were left for 30 min in phosphate-free Dulbecco's modified Eagle's medium (ICN) containing 0.1% dialyzed newborn calf serum, 25 mM Hepes, pH 7.5, and 2 mM L-gluthamine and labeled for 8 h by adding 1 mCi/ml [ 32 P]orthophosphate (Amersham). Some of the dishes were stimulated by adding dialyzed serum to 10% for 15 min. The culture dishes were placed on ice, washed twice in ice-cold PBS, and the cells were harvested by scraping in 1 ml of RIPA buffer (PBS containing 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS with added protease and phosphatase inhibitors as described previously (53)). Cleared lysates were mixed with 100 l of 50% slurry of gluthatione-agarose beads (Pharmacia) and incubated at 4°C for 1 h on a rotating wheel. The beads were then washed 8 times in phosphate-buffered saline containing 1% Triton X-100 and the final pellet was resuspended in 40 l of 2 ϫ SDS-PAGE gel loading buffer and boiled for 5 min. The labeled phosphoproteins were separated on a 10% polyacrylamide gel, electrotransferred onto a polyvinylidene difluoride (PVDF) membrane (Millipore), and visualized directly by autoradiography. Quantitation of the signals was performed using a PhosphorImager (Molecular Dynamics). Indirect detection of phosphorylation at serine 413 by monitoring a mobility shift was done by Western blotting using polyclonal Pax6 (P6C) or GST antibodies as described below.
For [ 35 S]methionine/cysteine labeling of proteins the cells were left for 30 min in methionine/cysteine-free Dulbecco's modified Eagle's medium (Sigma) containing 25 mM Hepes, pH 7.5, 2 mM L-glutamine, and 10% dialyzed newborn calf serum before adding 35 S-labeled amino acids (0.1 mCi/ml; Amersham) for 4 h. GST fusion proteins were purified as described above. Following purification the samples were split in two, half of the sample were resuspended in 2 ϫ SDS-PAGE gel loading buffer and the other half was washed twice in -phosphatase assay buffer (New England Biolabs) before addition of 1000 units of -phosphatase (New England Biolabs) in a total volume of 25 l. The dephosphorylation was stopped after a 30-min incubation at 30°C by adding 10 l of 5 ϫ SDS-PAGE gel loading buffer. The samples were boiled and proteins were separated on a 10% SDS-polyacrylamide gel, electrotransferred onto a PVDF membrane, and visualized by autoradiography using Kodak BioMax MR film.
For phosphorylation shift analyses using Western blot detection subconfluent cell cultures of NIH3T3 or HeLa cells were transfected with 10 g of the different expression vectors using LipofectAMINE as described above for NIH3T3 cells or calcium-phosphate coprecipitation for HeLa cells. After 16 h of serum deprivation (0.1% serum) the cells were stimulated with 10% serum or 100 ng/ml phorbol ester (TPA, Sigma) for the indicated times. GST fusion protein purification and SDS-PAGE were performed as described above. Following electrotransfer onto PVDF membranes the proteins were detected using polyclonal anti-GST or anti-Pax6 antibodies (P6C) as described below.
For p38 kinase-mediated in vivo phosphorylation, 10 g of GAL4-Pax6(299 -437) TAD fusion constructs (wild-type and mutants) in the pFA-CMV expression vector (Stratagene) were co-transfected with 5 g of p38 kinase (54) and 5 g of MKK6b(EE) (55) expression vectors or pcDNA3-HA (vector control) into NIH-3T3 cells using either Fugene 6 (Roche Molecular Biochemicals) or LipofectAMINE Plus (Life Technologies, Inc.). For ERK-mediated in vivo phosphorylation, 10 g of GAL4-Pax6 TAD fusion constructs and 5 g of expression vector for an activated mutant of MEK1, MEK1(EE) (56), or the respective vector control was used in co-transfections. In vivo labeling with [ 32 P]orthophosphate was done essentially as described above with the exception that the cells were harvested in HA lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 1 mM EGTA, 1% Triton X-100) for experiments with p38 kinase and JNK lysis buffer (25 mM Hepes, pH 7.7, 300 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.15 Triton X-100) for experiments with activated MEK1. Both lysis buffers contained the following inhibitors: 20 mM p-nitrophenyl phosphate, 50 mM sodium fluoride, 50 M sodium vanadate, 5 mM benzamidine, 0.5 g/ml leupeptin, 2 g/ml aprotinin, 20 mM ␤-glycerophosphate, 0.5 mM phenylmethyl sulfonate, 0.7 g/ml pepstatin A. Polyclonal anti-GAL4 DBD antibodies and protein A/G-Sepharose beads (both from Santa Cruz Biotechnology) were used for immunoprecipitation of the GAL4-Pax6 fusion proteins. The beads were washed 3 times with the lysis buffer before boiling in 30 l of gelloading buffer and electrophoresis on a 10% SDS-polyacrylamide gel. The proteins were blotted onto a PVDF membrane and 32 P-labeled proteins were detected by autoradiography.
Western Blot-HeLa cells were seeded at 4 ϫ 10 5 cells per 10-cm dish the day before transfection. The relevant expression constructs were transfected using calcium phosphate coprecipitation. Ten g of pCI-Pax6 or the vector control pCI-neo was co-transfected with 5 g of expression vector for p38 kinase in combination with 5 g of expression vector for MKK6b(EE) or vector control for the latter or with 5 g of expression vector for MEK1(EE) or its vector control (see Fig. 8). In all transfections 1 g of pCMV-␤gal (Stratagene) was included to allow measurement of ␤-galactosidase activities that were used to normalize for variations in transfection efficiencies. The cells were harvested after 2 days either by directly scraping them into 100 l of 2 ϫ SDS-PAGE gel loading buffer or into the Dual-Light lysis buffer (Tropix Inc.) to set aside an aliquot for ␤-galactosidase activity measurements, before the remaining was mixed with 5 ϫ SDS gel loading buffer and boiled. For some blots nuclear extracts prepared as described (57) were used. When indicated in figure legends the amount of protein loaded on the gel had been adjusted according to measurements of ␤-galactosidase activity. Proteins run on a 10% SDS-PAGE gel were blotted onto a PVDF membrane and blocked overnight at 4°C in a buffer consisting of Tris-buffered saline with 0.1% Tween 20 (TBST) and 5% non-fat dried milk. The following primary antibodies were used: anti-GST, 2 anti-GAL4 DBD (diluted 1:1.000; Santa Cruz Biotechnologies), serum 14 (28) (diluted 1:5.000), and P6C; affinity purified anti-Pax6 C-terminal antibodies (58, 59) (diluted 1:800). The primary antibodies were applied for 1 h at room temperature. The membrane was then washed 5-6 times in the TBST buffer for 30 -60 min. The secondary antibodies (anti-rabbit IgG-AP, Santa Cruz Biotechnology or anti-mouse IgG-AP, Sigma) were diluted 1:2.000 or 1:20.000, respectively, in the blocking buffer and left for 1 h at room temperature. The washing described above was repeated. CDP-Star substrate (Roche Molecular Biochemicals) was used according to the manufacturers instructions to visualize the specific protein bands.
Transient Transfections, CAT, and Luciferase Assays-The conditions used for transfections, extract preparation, and measurements of CAT activity were as described earlier (60). For transfections for luciferase assays, 4 ϫ 10 4 cells were seeded per well in 6-well dishes the day before transfection. Fresh medium was added to the cells 2 h before the DNA on the day of the transfection. Whenever the Pax6 expressing constructs contained a cytomegalovirus (CMV) promoter (pCI-neo and pFA-CMV plasmids), 0.05 g of pCMV-␤gal (Stratagene) was co-transfected for each well as a control of transfection efficiency. When the expression constructs harbored a SV40 early region promoter (pSG424 vector), 0.1 g of pCH110 (Pharmacia) was co-transfected. After trans-fection, the cells were either left in medium with 10% serum for 24 h before being harvested or kept in 10% serum for 24 h and in 0.1% serum for about 20 h before harvesting. The cells were washed twice with PBS in the wells before preparing extracts using the Dual-Light luciferase and ␤-galactosidase reporter gene assay system (Tropix Inc.) and analyzed in a Labsystems Luminoskan RT dual injection luminometer.
Gel Mobility Shift Assay-HeLa cells (4 ϫ 10 5 ) were seeded on 10-cm dishes the day before transfection. A total of 30 g of each Pax6 expression construct was transfected to 3 dishes. When other plasmids were included 15 g of each were added. The cells were left in medium with 10% serum for 2 days before nuclear extracts were made essentially as described (57) omitting the dialysis step leaving nuclear proteins in the high-salt buffer. The protein concentration was determined using the Bio-Rad assay, and 3 g of nuclear protein extract was used for each binding reaction in a buffer containing 10 mM Hepes-potassium hydroxide, pH 7.9, 30 mM potassium chloride, 4 mM spermidine, 0.1 mM EDTA, 0.25 mM dithiothreitol, 1 mM sodium phosphate, pH 7.2, 4% Ficoll 400, and 3 g of poly(dI-dC) (24). The P6CON probe (61) was included (10,000 cpm) and the binding reactions were left on ice for 20 min before they were loaded on a 5% polyacrylamide gel and run in 0.5 ϫ Tris borate-EDTA (TBE) buffer for 2 h at 200 V.

RESULTS
The Transactivation Domain of Pax6 Includes the Entire C-terminal PST Domain-The 139-amino acid long C-terminal transactivating region of zebrafish Pax6.1 (hereafter referred to as Pax6) is enriched in proline (P), serine (S), and threonine (T) residues comprising 45.4% of the total amino acids in this so called PST domain. The over-representation of these residues imposes a hydrophilic nature to this domain (63% hydrophilic amino acids), but it is also noteworthy that there is both a significant under-representation of charged residues and a over-representation of methionine residues. The sequence of this region is extremely conserved among vertebrate Pax6 proteins with zebrafish Pax6 being 93.6% identical (98.6% similar) to human PAX6 (see Fig. 1A). If amphioxus, squid, and sea urchin Pax6 proteins are included in the comparison, the sequence divergence increases but there is a striking conservation of two C-terminal sequence motifs (GLISPGVSVP(V/ I)QVPG and YW(P/S)R(L/I)Q in Fig. 1A). Using a combination of different prediction algorithms we deduced the hypothetical secondary structure of the Pax6 TAD shown in Fig. 1B. This region seems to contain only one ␣-helical segment with most of the domain being occupied by short ␤-sheets and random coil elements as well as four turns. The ␤-sheets are generally hydrophobic suggesting that they pack into the interior of this domain. The prediction of several ␤-turns also supports the notion of a perhaps more globular domain than just a flexible, extended and rather unstructured TAD. Interestingly, all but one of the charged residues are predicted to reside in ␤-turn structures on the surface of the protein. The last charged residue is at the extreme C terminus which is solvent-exposed in most proteins anyway. The predicted isoelectric point is 5.5-6.2 due to a net negative charge. A short repeat module with 8 out of 12 identical residues is also found within the TAD. This module is preceded by a putative ␤-sheet and contains two conserved charged residues and a predicted turn (Fig. 1B).
In order to define the extent of the Pax6 TAD we performed a deletion study by fusing parts of the Pax6 TAD (from amino acids 299 to 437) to the yeast GAL4 DNA-binding domain (DBD) and assayed transactivation by co-transfection of HeLa cells using two different reporters. One reporter,  (19), and sea urchin (Spax6) (25). Putative phosphorylation sites for proline-dependent protein kinases are indicated with asterisks above the alignment. Residues displayed on a black background are identical in all the compared species whereas other residues conserved in most of the proteins aligned are indicated by two shades of gray. Dashes indicate gaps introduced to facilitate optimal alignment. B, predicted secondary structure elements within the Pax6 TAD displayed together with the primary sequence of the zebrafish Pax6.1 TAD (positions 299 to 437). Black bars indicate ␤-sheets and an open bar an ␣-helix. These elements were predicted using both the PHD neural network method with a multialignment as input (72) and the Alexis program of the Seqsee program suite (73). Vertical arrows indicate turns predicted using the ProtScale resource at the Expasy server (http://www.expasy.ch/cgibin/protscale.pl). R1 and R2 denote a short repeat module indicated in bold together with the two highly conserved C-terminal sequence motifs. Plus and minus signs above the sequences demark the basic and acidic residues, respectively. Numbers above the sequence denote deletion end points for GAL4-Pax6 fusion constructs (vertical lines) and putative phosphorylation sites for proline-dependent protein kinases (asterisks). pG 5 E1bTATA-CAT, contained 5 GAL4-binding sites upstream of the TATA box from the minimal promoter of the adenovirus E1b gene (62) allowing activation of a core promoter from a TATA-proximal position to be measured. The other, pTKG 5 CAT (51), contained the same GAL4-binding sites inserted upstream of the herpes simplex virus thymidine kinase promoter facilitating measurements of both activation and repression of a more complex promoter. The CAT activity obtained by the GAL4-Pax6(299 -437) construct was set to 100% and the activities of the truncated constructs are given in percent relative to this value. The expression and correct sizes of all the different fusion proteins were verified by Western blotting of extracts from transfected cells (data not shown). As can be seen from Fig. 2, deletions of the TAD from either the Nor C-terminal end dramatically reduced the transactivation potential. Both reporters produced a similar picture but the reporter containing the minimal E1b promoter was more sensitive and we therefore refer to the results obtained with this reporter in the following. Deletion of 55 amino acids from the N-terminal end of the TAD (retaining positions 353-437) reduced transactivation by 65%. Further deletion of 23 amino acids (retaining positions 375-437) yielded a residual activity of only 5% while no significant activity could be measured when only the C-terminal 42 residues from position 396 to 437 were assayed following fusion to the GAL4 DBD. However, deletion of the C-terminal 43 amino acids showed that, although inactive by itself, this region is very important for full activity since the TAD from position 299 to 395 has lost 80% of the activity of the full-length TAD. Further deletion of the region between positions 395 and 375 reduced only slightly this residual activity while all activity was lost upon removal of an additional 23 amino acids from the C-terminal end (retaining position 299 -352). We also measured the activity of the middle region from position 353 to 395 and the two halves of this region as well as two internal deletions in this area. The middle region showed very low activity while the two halves were devoid of activity. The internal deletion of residues 353 to 374 reduced transac-tivation by more than 50% while deletion of amino acids 375 to 395, removing one of the two repeat modules, had no effect. Taken together, these results clearly show that the entire Cterminal PST-rich region of Pax6 acts as an unusually long TAD with no minimal activation domain. In contrast to what has been found for the Pax2, -5, and -8 subfamily of Pax proteins (63) there does not seem to exist any inhibitory regions within the Pax6 TAD.
To assess the relative potency of the transactivation domain of Pax6 we compared the level of transactivation from the pG 5 E1bTATA-CAT reporter of GAL4-Pax6(299 -437) to that of the weak activator GAL4-AH and the extremely strong activator GAL4-VP16 in HeLa cells. From three independent experiments performed in triplicate we found the Pax6 TAD to be only 2.5-fold weaker than the VP16 TAD and 40-fold stronger than AH. This clearly indicates the strong transactivation potential of the isolated Pax6 TAD upon fusion to the GAL4 DBD. Interestingly, when the region from amino acids 175 to 437, comprising part of the linker region, the homeodomain, and the entire TAD, was fused to the GAL4 DBD a low activity was measured from the minimal promoter reporter compared with that of the isolated Pax6 TAD.
The Transactivation Domain of Pax6 Is Phosphorylated in Vitro by the Mitogen-activated Protein Kinases ERK2 and p38, but Not by JNK-It has previously been demonstrated that the quail Pax6 protein is phosphorylated in neuroretina cells mainly on serine residues but with some phosphorylation of threonine also (28). Inspection of the sequence of the Pax6 TAD revealed four potential phosphorylation sites for mitogen-activated protein kinases at Thr 323 , Ser 376 , Thr 388 , and Ser 413 (indicated with asterisks in Fig. 1). As an initial experiment a GST fusion protein of Pax6 (amino acids 353-437), expressed and purified from E. coli, was used in in vitro kinase assays with the activated MAPKs ERK2, p38 kinase, and JNK (Jun N-terminal kinase). Activated ERK2 phosphorylated the GST-Pax6 fusion protein, whereas GST alone or GST fusions of the C-terminal TADs of Pax2, Pax3, Pax9a, or Pax9b were not phosphorylated at all. The same result was obtained with activated p38 kinase. However, when JNK immunoprecipitated from UV-irradiated NIH 3T3 cells was used only GST-Jun and neither the Pax6, Pax2, Pax9a nor Pax9b GST fusions were phosphorylated (data not shown). To map the phosphorylation site(s) for ERK2 and p38 kinase in the Pax6 TAD, mutations were introduced at the three most C-terminal candidate sites described above by substituting the respective serine and threonine residues with alanine. Upon expression in E. coli the GST-Pax6(353-437) fusion protein gives a band of 35 kDa representing the full-length fusion protein and three other bands of lower molecular masses which are specific proteolytic fragments representing "deletions" from the C-terminal end of the Pax6 TAD (Fig. 3A, lower panel). This specific fragmentation pattern proved beneficially for mapping of the sites that are phosphorylated in vitro. Thus, in vitro phosphorylation of GST-Pax6(353-437) mutant proteins with ERK2 showed that the Thr 388 site is not phosphorylated since the T388A mutant protein gives the same phosphorylation pattern as the wildtype (wt) protein (Fig. 3A, lane 4). Due to the specific band pattern, one can easily see that both the Ser 376 and Ser 413 sites are phosphorylated in vitro. In the S376A mutant protein the lower molecular weight fragments are not phosphorylated (Fig.  3A, lane 3) whereas they are in the T388A and S413A mutants. Phosphorylation at Ser 413 caused a mobility shift that can no longer be seen in the S413A mutant, (Fig. 3A, lane 5). Since the Thr 323 site was not included in this set of experiments we prepared GST-Pax6 fusions harboring the entire C-terminal TAD from amino acid 299 to 437 representing a panel of single and double mutants as well as a triple mutant including all three phosphorylation sites (T322A/T323A, T375A/S376A, and S413A) sites. (For simplicity, we refer to the T322A/T323A and T375A/S376A mutants as single mutants, since only one putative MAPK phosphorylation site is mutated in each of them.) Following in vitro phosphorylation both ERK2 and p38 kinase gave the same phosphorylation pattern (Fig. 3B). The GST-Pax6(299 -437) fusion protein gives two bands, where the slowest migrating band have the expected size for the full-length fusion protein (about 41 kDa), while the faster migrating band is a degradation product containing GST and a short fragment of the Pax6 TAD encompassing the Thr 323 phosphorylation site. As is evident when lanes 3, 8, and 10 are compared with the other lanes in Fig. 3B, the T322A/T323A mutation prevents phosphorylation of the faster migrating band showing that the Thr 323 site is a phosphorylation site for both ERK2 and p38 kinase in vitro. None of the other single site mutations affected the phosphorylation pattern. The two double mutants (T322A/ T323A, S413A and T375A/S376A, S413A) both caused a marked decrease in phosphorylation intensities when ERK2 was used for in vitro phosphorylation (Fig. 3B, lanes 8 and 9). For the p38 kinase, however, only the T322A/T323A, S413A double mutation lead to a decreased phosphorylation indicating that the Thr 323 site is more important for phosphorylation by p38 kinase than the Ser 376 site. For both ERK2 and p38 kinase the triple mutation (Fig. 3B, lane 10) prevented phosphorylation completely. This proves that Thr 323 , Ser 376 , and Ser 413 are the only phosphorylation sites used in vitro by ERK2 and p38 kinases in the C-terminal TAD (amino acids 299 -437) of zebrafish Pax6.
Pax6 Is Phosphorylated in Vivo by ERK and p38 MAPKs-As described above, three of the four potential phosphorylation sites for proline-dependent kinases in the TAD of Pax6 are phosphorylated by ERK2 and p38 kinase in vitro. To see if the Pax6 TAD could be phosphorylated in vivo, NIH 3T3 cells were transfected with a eucaryotic expression vector for GST-Pax6(353-437) and in vivo labeled with [ 32 P]orthophosphate following serum starvation. The cells were harvested after stimulation with 10% serum, to activate ERK1 and -2, and the GST-Pax6 fusion protein was purified from the cell lysate by the use of glutathione-agarose beads. There is some phosphorylation of the fusion protein in serum-starved cells, but addition of serum enhanced this phosphorylation 2.1-fold (Fig. 4A). The membrane was stained with an antibody raised against the Pax6 TAD to control equal loading of the lanes. We then utilized the finding that phosphorylation at the evolutionary conserved Ser 413 site causes a mobility shift of the GST-Pax6(353-437) fusion protein to study the time course of phosphorylation of this site following serum stimulation. As monitored by Western blotting, Ser 413 is phosphorylated after 15 min of serum stimulation with a slight increase at 60 min followed by a decrease 4 h after addition of serum (Fig. 4B). Thus, the kinetics of phosphorylation of the Ser 413 site of the Pax6 TAD is similar to the kinetics of ERK activation. Using in vivo labeling with [ 32 P]orthophosphate followed by serum stimulation of NIH 3T3 cells we confirmed the in vitro data that suggested the Ser 413 site as the cause of the phosphorylationinduced mobility shift. When the Ser 376 site is mutated the lower migrating band disappears, while the shifted band dis-appears for the S413A mutant protein (Fig. 4C, upper panel). The weak phosphorylation of the T388A mutant in this experiment is due to less loading of this protein on the gel as revealed by Western blotting with the anti-Pax6 antibody (Fig. 4C, lower  panel). The results of this in vivo phosphorylation experiment clearly suggest that the Ser 413 site is contributing much more than the Ser 376 site to the total phosphorylation of GST-Pax6(353-437). As a final proof that the mobility shift observed for the GST-Pax6(353-437) fusion protein is indeed due to phosphorylation of Ser 413 we treated 35 S-labeled GST-Pax6 fusion proteins purified from serum-stimulated NIH 3T3 cells with -phosphatase which specifically dephosphorylates phosphoserine and phosphothreonine residues in proteins. As shown in Fig. 4D, phosphatase treatment leads to loss of the shifted band.
To further confirm that the Ser 413 site is phosphorylated in vivo by ERK1 and -2 we stimulated HeLa cells transfected with expression vectors for GST-Pax6(353-437) wt and S413A mutant proteins with the phorbol ester TPA for 15 min. TPA is a rather specific inducer of the MEK-ERK pathway in these cells. We also employed a specific inhibitor of MEK, PD 98059 (64), to determine if MEK-induced activation of ERK is required for phosphorylation of Ser 413 following stimulation with TPA. As seen from the Western blot in Fig. 5A, TPA induced the mobility shift indicative of phosphorylation of Ser 413 while pretreatment with PD 98059 abolished the TPA-induced phosphorylation of Ser 413 . To confirm that TPA leads to ERK activation under these conditions we determined the phosphorylation of a GST-ElkC fusion protein following immunoprecipitation of ERK. Thus, TPA treatment (100 ng/ml) for 15 min induced ERK activation to the same extent as 10% serum (Fig. 5B). These results show that the Ser 413 site is phosphorylated upon activation of ERK MAPKs in serum-and TPA-stimulated cells. Furthermore, the Ser 376 is also phosphorylated following serum stimulation but not as efficiently as the Ser 413 site.
The above mentioned in vivo phosphorylation experiments were conducted with GST fusion proteins that do not contain a nuclear localization signal. Immunostaining of transfected cells revealed that the fusion proteins were located in the cytoplasm (data not shown). To include the entire Pax6 TAD and to study in vivo phosphorylation of nuclear localized fusion proteins we constructed expression vectors where wt and mutant Pax6 TAD (299 -437) were fused in-frame with the GAL4 DBD. We then asked whether activation of ERK and p38 kinase could lead to phosphorylation of the Pax6 TAD in the nucleus of transfected cells. Thus, in one set of experiments the GAL4-Pax6 fusion constructs were co-transfected with expression plasmids for activated MEK1. NIH-3T3 cells were transfected, deprived of serum, and in vivo labeled with [ 32 P]orthophosphate before harvesting and immunoprecipitation with antibodies against the GAL4 DBD. As seen from lane 1 in Fig. 6A, the GAL4 DBD is phosphorylated upon co-transfection with activated MEK1. We found that there are two putative MAP kinase sites in the N-terminal region of the GAL4 DBD. However, the wt GAL4-Pax6 TAD fusion protein is clearly phosphorylated and migrates as a doublet due to the mobility shift induced by phosphorylation of the Ser 413 site (lanes 2 and 3,   Fig. 6A). The background level of phosphorylation in serumstarved NIH 3T3 cells (lane 2) is increased 2.4-fold (following subtraction of the contribution by phosphorylation of the GAL4 DBD) upon co-transfection with an expression vector for activated MEK1 (lane 3). In the S413A mutant the phosphorylation shift is abolished. In another set of experiments, GAL4-Pax6 fusion constructs were co-transfected with expression plasmids for p38 kinase and the constitutively active mutant upstream kinase, MKK6b(EE). The p38 kinase is inactive by FIG. 7. Mutation of the evolutionary conserved Ser 413 site strongly affects the transactivation potential of the isolated Pax6 TAD. The transactivation potential of the wt and mutant Pax6 TAD constructs fused to the DNA-binding domain of GAL4 was determined in HeLa (upper panel) and NIH 3T3 cells (lower panel) growing in 10% serum. Co-transfections were performed in 6-well cell culture dishes with 0.5 g of pSG424 of expression vector for wt GAL4-Pax6(299 -437) or the different mutants together with 0.5 g of pG 5 E1bTATA-LUC reporter plasmid. The cells were harvested for luciferase and ␤-galactosidase activity assays following growth in 10% serum for 24 h. To normalize for variations in transfection efficiencies, 0.1 g of pCH110 was included in each transfection to allow measurement of ␤-galactosidase activities. The ratio between luciferase and ␤-galactosidase activity obtained with the wt GAL4-Pax6(299 -437) expression vector was set to 100%. For HeLa cells the results represent the mean with standard deviations of three independent experiments performed in triplicate. For NIH 3T3 cells the data are the mean with standard deviations of two independent transfections done in triplicate. itself and is activated following co-transfection with an expression vector for MKK6b(EE) (48). As can be seen from Fig. 6B, the wt Pax6 TAD is strongly phosphorylated following coexpression of MKK6b(EE) and p38 kinase with all of the labeled protein displaying the mobility shift indicative of Ser 413 phosphorylation. The S413A mutant is also phosphorylated but does not display the mobility shift seen for the wt fusion protein (compare lanes 3 and 5). In the experiment shown twice as much protein was loaded in lane 5 compared with the others. When this is taken into account and the background due to phosphorylation of the GAL4 DBD is subtracted we measured a 1.8 -2.0-fold reduced phosphorylation in the S413A mutant compared with the wt. A similar experiment including also the triple mutant, where all MAPK phosphorylation sites in the TAD are mutated, showed no increase in the signal above the background due to phosphorylation of the GAL4 DBD (data not shown). Taken together, the results shown in Figs. 4 -6 show that the Pax6 TAD can be phosphorylated following activation of both ERK and p38 MAP kinases in vivo.
Mutation of the Evolutionary Conserved Ser 413 Site Strongly Affects the Transactivating Ability of the Isolated Pax6 TAD-In order to test whether mutation of the MAPK phosphorylation sites would affect the transactivating activity of the isolated Pax6 TAD we transfected HeLa and NIH 3T3 cells with GAL4-Pax6(299 -437) wt and mutant expression constructs using pG 5 E1bTATA-LUC as the reporter. Following transfection, the cells were left in 10% serum for 24 h before harvesting. As shown in Fig. 7, the T322A/T323A and T375A/ S376A mutations had little or no effect on the transcriptional activity. However, when these two mutations are combined the transactivation is reduced by 50% in HeLa cells. In NIH 3T3 cells, this effect is greatly reduced. In both cell lines the S413A mutant, the two double mutants, and the triple mutant (all containing the S413A mutation) displayed markedly decreased transactivation compared with the wt TAD (from 60 to 80% reduction in activity). When the serine in position 413 was mutated to glutamate (S413E) to mimic the effect of a phosphorylation, an increase in transcriptional activation was observed. In some experiments the positive effect of the S413E mutation was even more dramatic than shown here. Such a difference between Ala and Glu mutants were not observed for the 376 and 388 sites (data not shown). Furthermore, we also mutated a putative casein kinase II phosphorylation site at position 425 both to Ala and Glu without observing any differences in transactivation. Taken together, this shows that it is not simply an increase in negative charge that increases transactivation or loss thereof that decreases it. Rather, the evolutionary conserved Ser 413 residue is clearly important for full transactivation by the isolated Pax6 TAD, most likely by serving as an important phosphorylation site.
Phosphorylation of the Ser 413 Site Positively Modulates the Transactivating Activity of Pax6 -To see if phosphorylation of the TAD affected the transactivation potential of the fulllength Pax6 protein, HeLa cells were transfected with pCI-Pax6 wt, the phosphorylation site mutants T322A/T323A, T375A/S376A, S413A, or a triple mutant containing all three mutations. As a reporter pP6CON-LUC (18), containing six consensus Pax6 paired domain-binding sites (65) upstream of the adenovirus E1b minimal promoter, was used. Co-transfection with an expression vector for activated MEK1 caused a 2.9-fold induction of the transcriptional activity of wt Pax6 compared with co-transfection with the vector control (Fig. 8A). This induction was not caused by increased levels of Pax6 proteins, since a Western blot of Pax6 co-transfected with activated MEK1 or the vector control displayed similar amounts of Pax6 protein. The T322A/T323A and T375A/S376A mutants showed elevated transcriptional activation potentials compared with wt Pax6 but where less inducible by activated MEK (2.0-and 1.8-fold, respectively). The S413A mutant demonstrated only half of the transactivation potential of wt Pax6, but was still induced 2.0-fold by MEK1(EE). The triple mutant was less active than the wt protein, but more active than the S413A mutant, and inducible by activated MEK by a factor of 2.2.
We have found that expression of the activated MEK1 mutant alone is sufficient to produce maximal activation of a GAL-Elk-1 fusion protein suggesting that the levels of endogenous ERK1 and -2 are not limiting. To ensure that activated MEK1 gives maximal activation of Pax6 transactivation we also performed experiments where an expression vector for ERK1 was co-transfected with activated MEK1. The results obtained were similar to those obtained with activated MEK1 alone (data not shown) confirming that the levels of endogenous ERK1 and -2 are not limiting.
When the same set of pCI-Pax6 plasmids were co-transfected with expression vectors for p38 kinase and MKK6b(EE), wt Pax6 was induced 16-fold compared with transfection with p38 kinase alone (Fig. 8B). The T322A/T323A and T375A/S376A mutants had nearly the same transcriptional activity as the wt. The S413A mutant showed about 50% of the wt transactivation activity, and was only induced about 6-fold following co-transfection with MKK6b(EE) and p38 kinase. The triple mutant was only stimulated 2.7-fold indicating some contribution from the Thr 323 and Ser 376 sites to the overall transcriptional activity following activation of p38 kinase. The Western blot in Fig.  8B demonstrates that equal amounts of Pax6 protein was expressed in HeLa cells transfected with Pax6 and p38 kinase compared with cells transfected with Pax6, p38 kinase, and MKK6b(EE). The striking increase in transactivation by Pax6 observed following activation of p38 kinase is therefore not caused by an increase in the expression level or stability of the Pax6 protein.
The DNA-binding Affinity of the Paired Domain Is Neither Affected by Phosphorylation Site Mutations in the TAD nor ERK or p38 Kinase Activation-It has previously been shown that mutations in the homeodomain of Pax3 affect the DNA binding of the paired domain and vice versa (74 -76), and that deletion of C-terminal amino acids 332-416 of quail Pax6 aborts DNA binding (24), suggesting that structural constraints arising from other parts of the molecule may modulate the DNA binding by the paired domain. Gel mobility shift assays were performed to determine if the phosphorylation site mutations in the TAD affected DNA binding by the paired domain. Nuclear extracts from HeLa cells transfected with the wt and mutant Pax6 expression vectors were used in gel mobility shift assay with the P6CON probe containing a single consensus Pax6 paired domain-binding site (Fig. 9A). The steady state DNA binding efficiency of wt and mutant Pax6 proteins was determined using a PhosphorImager. The mean result from three independent experiments performed with two different nuclear extracts showed that there were no significant differences in DNA binding between the wt and mutant Pax6 proteins. Next, we wanted to test whether activation of ERK or p38 kinase would affect DNA binding by the paired domain. Thus, wt Pax6 was co-transfected with activated MEK1 or p38 kinase and MKK6b(EE) and the appropriate controls into HeLa cells. The cells were subsequently serum starved before nuclear extracts were made. Phosphorylation of Pax6 caused by co-transfection of activated MEK1 or p38 kinase/MKK6b(EE) (Fig. 9A, lanes 8 and 10) does not display any marked effects on DNA binding compared with the controls (Fig. 9A, lanes 9 and 11). As shown by Western blotting (Fig.  9B), the nuclear extracts used for gel mobility shift assay contained similar amounts of Pax6 proteins.  8 and 10). For experimental details, see "Materials and Methods." Nuclear proteins (3 g) were incubated with the P6CON probe containing a single consensus Pax6 paired domain-binding site (61) on ice for 20 min, and run on a 5% polyacrylamide gel for 2 h at 220 V. B, mutations of the phosphorylation sites in the TAD do not affect the expression level of Pax6 and the Pax6 protein level is not increased upon co-transfection with activated MEK1 or activation of p38 kinase. Three g of nuclear extract was loaded in each lane of a 10% SDS-polyacrylamide gel and Western blot performed using the P6C antibody (1:800 dilution). The differences in the number of detected bands between lanes 1-6 and lanes 7-10 are due to the use of different batches of primary antibody, membrane type, and detection system. The Western blot shown as lanes 1-6 was performed as described under "Materials and Methods," while for the Western blot shown as lanes 7-10 a Hybond membrane (Amersham) and anti-rabbit horseradish peroxidase-conjugated secondary antibodies (Transduction Laboratories) were used with the ECL detection system (Amersham). The location of a Pax6 isoform lacking the paired domain due to expression from an internal start codon (28) is indicated (Pax6⌬PD).

DISCUSSION
In this report we show that the complete region C-terminal to the homeodomain of Pax6 is necessary for maximal transcriptional activation. The Pax6 TAD does not contain any internal sequence elements that may inhibit transactivation as found for the Pax2, -5, and -8 subfamily of Pax proteins (63). While this work was in progress Tang et al. (27) came to a similar conclusion using deletions corresponding to the exons encoding the human PAX6 TAD. Thus, taken together, these two studies using different deletion constructs of human and zebrafish Pax6 proteins confirm that Pax6 contains an unusually long TAD which has been strongly conserved both in primary sequence and function during vertebrate evolution. The large size of the TAD suggests that it may provide interaction surfaces for several cofactors and/or components of the basal transcriptional machinery. Interestingly, both we and Tang et al. (27) found that the presence of the homeodomain in GAL4-Pax6 TAD fusions inhibited the transcriptional activation potential compared with the TAD alone. This has also been observed with the Pax3 homeodomain in a similar GAL4-Pax3 TAD fusion, where inclusion of the homeodomain reduced the activation by approximately 20-fold (66). Different models may account for this behavior. The homeodomain could exert a direct negative effect either by interfering with TAD function or the reduced activity could reflect a titration of the fusion protein away from the reporter promoter due to binding of the Pax6 homeodomain to chromosomal sites. However, Chalepakis et al. (67) have shown that the Pax3 homeodomain itself causes repression of a thymidine kinase promoter when fused to GAL4 making it likely that the homeodomain may recruit a repressor.
Vertebrate Pax6 TADs contain four potential MAPK phosphorylation sites. We found that three of these sites are phosphorylated in vitro by both ERK and p38 kinase. JNK was unable to phosphorylate the Pax6 TAD in vitro and the Cterminal TADs of zebrafish Pax2, Pax9a, and Pax9b were not phosphorylated by any of the three MAP kinases, even though they too contain putative MAPK phosphorylation sites. Furthermore, the murine Pax3 TAD was not phosphorylated by ERK2 or p38 kinase. Of the three sites phosphorylated in the Pax6 TAD the Ser 413 is conserved from sea urchin to man while the other two sites are conserved in the highly similar vertebrate proteins but not in amphioxus, squid, and sea urchin. In vivo phosphorylation experiments using GST-Pax6 TAD wildtype and mutant proteins demonstrated that the Pax6 TAD is an ERK substrate also in vivo. Since the two nuclear localization signals of Pax6 reside in the paired domain and directly N-terminal to the homeodomain, respectively (24), the GST-Pax6 TAD fusions contain no nuclear localization signals and were only expressed in the cytoplasm. Apart from demonstrating that the Pax6 TAD is phosphorylated in vivo following activation of ERKs by serum and phorbol ester, particularly on the evolutionary conserved Ser 413 site, our results with these fusion proteins are also of interest since the paired-less isoform of Pax6 (28) is distributed between the nucleus and cytoplasm with most expression in the cytoplasm (our own observations and Ref. 24). Using Gal4-Pax6 TAD fusions in co-transfection experiments we found that both activation of ERKs and p38 kinase lead to phosphorylation of the Ser 413 site in the nuclear compartment.
Specific phosphorylation events can regulate the activity of transcription factors via several different mechanisms involving changes in protein stability, DNA binding activity, subcellular localization, or protein-protein interactions (reviewed in Refs. 30 and 68). We have shown by gel-mobility shift assays and Western blots that for Pax6 neither the DNA binding activity nor the protein stability are affected by mutations of phosphorylation sites or activation of ERK or p38 kinase. Since the full-length Pax6 protein is localized in the nucleus (data not shown and (24) subcellular localization is not regulated by phosphorylation of the Pax6 TAD. Both in the contexts of fusions to the GAL4 DBD and in the full-length Pax6 protein mutation of the Ser 413 site to an alanine greatly reduced the transactivation potential. This could be due to structural effects not reflecting any direct role for phosphorylation. However, mutation to a glutamate increased the activity of the GAL-Pax6 TAD fusion and activation of p38 kinase strongly enhanced the activity of the wt Pax6 protein whereas the activities of the Ser 413 and triple mutant (where all three MAPK phosphorylation sites in the TAD are mutated to alanines) showed greatly reduced responses to p38 kinase activation. The results shown in Fig. 8 suggest that activation of p38 kinase exerts a much more pronounced positive effect on the transactivation ability of Pax6 than activation of ERKs does (16-fold compared with about 3-fold). The effect mediated by ERK activation is also to some extent independent of the Ser 413 site indicating a significant indirect contribution probably through phosphorylation of other proteins that Pax6 depend on for efficient transactivation. However, the stimulation of Pax6 transactivation following activation of p38 kinase is strongly dependent on the Ser 413 site with some contribution also from the two other phosphorylation sites. In addition, activation of p38 kinase also reveals a smaller but significant, positive modulation of Pax6 transactivation which is independent of the phosphorylation sites in the TAD. This is evident since both the Ser 413 and the triple mutant are stimulated by p38 kinase, although much less so than the wt Pax6 protein. Thus, both ERK and p38 kinase activation display an indirect mechanism whereby the transactivation potential of Pax6 is increased 2-3fold. In addition, p38 kinase, and to a more modest extent ERK, is able to exert a direct effect which is dependent on the integrity of the Ser 413 site. Thus, direct phosphorylation of this site seems necessary for Pax6 to achieve full transactivating activity. Our structure predictions suggest that the Ser 413 site is located in a surface exposed turn between to hydrophobic ␤-sheets making it likely that phosphorylation of this site may induce a conformational change in the TAD. Such a conformational change may affect both inter-and intramolecular protein-protein interactions. Although not proven directly, several studies indicate that there could be an interaction between the TAD of Pax6 and the DNA-binding domains (24,69,70). We also found that the Pax6 TAD acts much more potently when fused to the heterologous GAL4 DBD than in its natural context when assayed on the same promoter containing either five GAL4-binding sites or six paired domain-binding sites, respectively. Hence, phosphorylation may relieve an intramolecular inhibition by inducing a conformational change in the Pax6 TAD. Subsequently, efficient contacts may be established between the TAD and coactivator(s) and/or components of the basal transcriptional machinery to allow full activation of transcription. The indirect mechanism observed suggests activation by ERK and p38 kinase of such cofactors or other proteins involved in unleashing the potential of the Pax6 TAD. To our knowledge, no reports on interactions between cofactors/corepressors or general transcription factors and the TADs of any Pax proteins have been published. The transcription factor Microphthalmia interacts specifically with the cofactor CBP/ p300 following ERK-mediated phosphorylation of Ser 73 (71). However, preliminary experiments have failed to reveal a significant stimulation of Pax6-mediated transactivation by CBP (data not shown) suggesting that other factors are involved.
The different efficiencies of ERK and p38 kinase activation in enhancing the transactivation potential of Pax6 is most readily explained by differences in activity toward the substrate in vivo. This is illustrated in Fig. 6 where only part of the wt GAL-Pax6 TAD fusion proteins displays the mobility shift due to phosphorylation of Ser 413 following co-transfection with activated MEK1 while all of the protein is shifted following activation of p38 kinase. This work represents the first report of induced phosphorylation of a Pax protein. Dörfler and Busslinger (63) stimulated B cell lines with various interleukins, serum growth factors, or phorbol esters in an effort to demonstrate a link between transcriptional activation of Pax5 and signaling pathways, but failed to identify such a link. A challenge for future studies is to determine the role of phosphorylation of the Pax6 TAD in more physiologically relevant settings. Pax6 can potentially receive signals mediated by MAP kinase cascades in all the tissues where it is expressed during development and/or in the adult organism. The nature of the external signal could differ according to the specific tissue, and the final response elicited by phosphorylation of Pax6 will most likely depend on tissue-and cell-type specific parameters such as availability of specific cofactors and other transcription factors acting together with Pax6 on specific regulatory sites.