Protein Phosphatase-1 Modulates the Function of Pax-6, a Transcription Factor Controlling Brain and Eye Development*

Pax-6 is an evolutionarily conserved transcription factor and acts high up in the regulatory hierarchy controlling eye and brain development in humans, mice, zebrafish, and Drosophila. Previous studies have shown that Pax-6 is a phosphoprotein, and its phosphorylation by ERK, p38, and homeodomain-interacting protein kinase 2 greatly enhances its transactivation activity. However, the protein phosphatases responsible for the dephosphorylation of Pax-6 remain unknown. Here, we present both in vitro and in vivo evidence to show that protein serine/threonine phosphatase-1 is a major phosphatase that directly dephosphorylates Pax-6. First, purified protein phosphatase-1 directly dephosphorylates Pax-6 in vitro. Second, immunoprecipitation-linked Western blot revealed that both protein phosphatase-1α and protein phosphatase-1β interact with Pax-6. Third, overexpression of protein phosphatase-1 in human lens epithelial cells leads to dephosphorylation of Pax-6. Finally, inhibition of protein phosphatase-1 activity by calyculin A or knockdown of protein phosphatase-1α and protein phosphatase-1β by RNA interference leads to enhanced phosphorylation of Pax-6. Moreover, our results also demonstrate that dephosphorylation of Pax-6 by protein phosphatase-1 significantly modulates its function in regulating expression of both exogenous and endogenous genes. These results demonstrate that protein phosphatase 1 acts as a major phosphatase to dephosphorylate Pax-6 and modulate its function.

Pax-6 is an evolutionarily conserved transcription factor that controls eye and brain development in humans, mice, zebrafish, and Drosophila (14 -29). Mutations of the Pax-6 gene result in the absence of eyes in humans (30). On the other hand, targeted expression of Pax-6 induces ectopic eye formation in Drosophila (31). Previous studies have shown that various forms of Pax-6 with different molecular weights exist, and at least four variants of Pax-6 (p46, p48, p43, and p32) were detected in cellular extracts (Fig. 1A) (32,33). All forms of Pax-6 bear a conserved C-terminal transactivation domain, which contains relatively rich proline (P), serine (S) and threonine (T) residues, and is thus named the PST domain. Several phosphorylation sites have been identified in this region of the human and zebrafish Pax-6 ( Fig. 1B). It has been shown that phosphorylation of these sites in Pax-6 is carried out by p38, ERK (34), and homeodomain-interacting protein kinase 2 (35). On the other hand, dephosphorylation of Pax-6 remains largely unknown. Here, we present both in vitro and in vivo evidence to show that PP-1 is involved in the dephosphorylation of Pax-6 to modulate its function. In normal physiological conditions, most cellular Pax-6 protein is in an inactivate status, probably due to its dephosphorylation by PP-1. Dephosphorylation of Pax-6 by PP-1 attenuates its transcriptional activity on the downstream target genes.

EXPERIMENTAL PROCEDURES
Chemicals-Various molecular biology reagents were purchased from Invitrogen, BD Biosciences Clontech, and Promega Biotech (Madison, WI). Various antibodies were obtained from Cell Signaling Inc. (Beverly, MA), Santa Cruz Biotechnology, Inc. (Santa Cruz, CA), and BD Biosciences. The culture medium and most other chemicals and antibiotics were purchased from Sigma, Calbiochem, and Invitrogen.
Preparation of Expression Constructs-The mPax-6 gene was kindly provided by Dr. Melinda K. Duncan (University of Delaware). The complete coding sequence for mPax-6 or for the C-terminal fragment (amino acids 299 -422) was isolated and inserted into the BamHI/SalI sites of pGEX 4T-1 vector to generate the GST fusion protein. A pair of primers (forward, 5Ј-ATGGATCCATTCCTATCAGCAGC-3Ј; reverse, 5Ј-ATGCGGCCGCTCGAGTCGAC-3Ј) were used to amplify the C-terminal fragment above. The human Pax-6 gene was purchased from ATCC. The hPax-6 coding sequence was isolated and inserted into the NotI/SalI sites of pCI-neo vector to construct pCI-neo-hPax-6. The human PP-1␣ catalytic subunit was purchased from ATCC. The coding sequence was isolated and inserted into the EcoRI/XbaI sites of pCI-neo vector. The ␣B-crystallin promoter (Ϫ426 to ϩ44)/chloramphenicol acetyltransferase construct was kindly provided by Dr. Joram Piatigorsky (NEI, National Institutes of Health). A pair of primers with additional restriction enzyme sites were designed as follows: 5Ј-ATCGTCTCGAGTCCAGTCCCTGCCCAGG-C-3Ј (forward) and 5Ј-ATCGTAAGCTTAGAAGGGGCGCC-GGATCC-3Ј (reverse). The two primers were used to amplify the ␣B-crystallin promoter (Ϫ426 to ϩ44). The luciferase assay construct pGL3 Basic vector was purchased from Promega (catalog number E1751). The amplified ␣B-crystallin promoter fragment was isolated and inserted into the XhoI/HindIII sites of pGL3 Basic vector to create p␣B-Luc.
Stable and Transient Transfection-The pCI-neo, pCI-neo-hPax-6, and pCI-neo-PP1␣ constructs were amplified in DH-5␣ and purified by Qiagen kits, according to the instruction manual. After treatment with RNase and extraction by phenol/ chloroform, the plasmids were transfected into HLE using Lipofectamine 2000 from Invitrogen, according to the company instruction manual. The transfected cells were then subjected to G418 (400 g/ml) selection for 4 -6 weeks, and then the individual clones for the stable transfected cell lines were established and confirmed with Western blot and RT-PCR analysis.
Treatment by Calyculin A-Parental or various transfected HLE cells were grown to 100% confluence. Then 10 ml of Dulbecco's modified Eagle's minimal essential medium containing 0.01% Me 2 SO (control) or different concentrations of calyculin A were used to replace the culture medium, and the treatment was continued for 1 h.
RNA Interference to Silence Expression of PP-1␣ or PP-1␤-The PP-1␣, PP-1␤, and PP-2A␣ siRNA oligonucleotides as well as the control oligonucleotides were purchased from Santa Cruz Biotechnology. Transfection was conducted as previously described (11).
Protein Preparation and Western Blot Analysis-Protein preparation from both treated and nontreated cells was conducted as previously described (38). Western blot analysis was performed as described before (38).
Immunoprecipitation-Irradiation of HLECs was conducted as previously described (39). Immunoprecipitation of total proteins from normal or UVA-irradiated cells was conducted as previously described (38,39).
In Vitro Phosphorylation of GST and GST-PST-The fusion proteins of GST-PST and GST protein were prepared as described before (40). To prepare active p38 kinase and parallel control, 10 g of mouse anti-p38 (pT180/pY182) or 10 g of mouse anti-␤-actin antibody (control) was added into 1 mg of HLE total proteins, and after vortexing for 10 s the mixture was incubated on ice for 1 h. Then 100 l of protein A/G plusagarose beads were added into the protein solution, and the protein solution was subjected to rotation at 4°C for overnight. The pellet was collected by centrifugation and washed four times with washing buffer as previously described (11). After washing, the pellet was left in 50 l of 1ϫ kinase buffer (50 mM Tris, pH 7.5, 10 mM MgCl 2 , 1 mM EGTA, 2 mM dithiothreitol, 0.01% Brij 35, 0.02% bovine serum albumin) and used for dephosphorylation substrate.
In Vitro Dephosphorylation Assays-The in vitro dephosphorylation assays were conducted as previously described (11).
In Vivo Dephosphorylation Assays-The in vivo dephosphorylation assays were conducted as described before (11,41). Briefly, HLECs transfected with pCI-neo-hPax-6 were labeled with [ 32 P]orthophosphate (200 Ci/ml) in phosphatase-free Dulbecco's modified Eagle's minimal essential medium. Then total proteins were extracted for immunoprecipitation with an antibody against Pax-6. The precipitated protein pellets were dissolved in 100 l of 1ϫ protein phosphatase assay buffer. Then, 10 l of the precipitated protein solution was mixed with 2 l 0.01% Me 2 SO, or 10 nM specific PP-1 inhibitor, PP1-I2 (from Calbiochem, catalog number 539516, IC 50 ϭ 2 nM), or 90 nM specific PP-2A inhibitor, PP2A-I2 (from Calbiochem, catalog number 539620, IC 50 ϭ 30 nM), or 5 nM calyculin A (Invitrogen). After incubation on ice for 30 min, the immunoprecipitated protein complex was incubated at 30°C for 10 min. After trichloroacetic acid precipitation, the supernatant fraction was recovered for counting the release of free 32 P in a scintillation counter.
RNA Extraction and RT-PCR-Various transfected HLECs were grown to 100% confluence and then harvested for RNA extraction. Briefly, after phosphate-buffered saline washing twice, the pelleted cells were resuspended in RNA extraction buffer, Trizol. The total RNAs were extracted according to the instruction manual. For reverse transcription, 500 ng of total RNAs, 500 ng of oligo(dT) 15 , and diethyl pyrocarbonate-H 2 O were mixed in a total reaction volume of 11 l. After a brief centrifugation, the mixture was heated at 65°C for 5 min followed by immediate incubation on ice for 5 min. After brief centrifugation, the reverse transcription reaction was carried out in a 20-l system with 1 mM dNTP, 0.01 M dithiothreitol, 2 units/l RNase inhibitor, and 15 units of avian myeloblastosis virus at 42°C for 60 min. Then the reaction was stopped at 85°C for 5 min. The following primers were used for PCR analysis: Pax-6, 5Ј-AGCCAAAATAGATCTACCTGAAG-3Ј (forward) and 5Ј-ACACCAGGGGCAATGAGTCCT-3Ј (reverse); ␤-actin, 5Ј-GTGGGGCGCCCCAGGCACCA-3Ј (forward) and 5Ј-CTCCTTAATGTCACGCACGATTTC-3Ј (reverse); and ␣Bcrystallin, 5Ј-TACCTCGAGATGGACATCGCCATCCAC-3Ј (forward) and 5Ј-CAACCCGGGTTCAAGAAAGGGCATCTA-3Ј (reverse). PCR was carried out as previously reported (38).
Analysis of Transient Gene Expression-The Promega dual luciferase reporter assay system was used in our analysis. In brief, 2.5 g of luciferase reporter gene (p␣B-Luc), together with 3.5 g of pCI-neo-Pax-6 plasmid plus 1.5 g of pCI-neo-PP1␣, or 1.5 g of pCI-neo-PP2A␣ and 0.2 g of control plasmid Renilla luciferase pRL-SV40-Luc (catalog number E2231) were co-transfected into mouse lens epithelial cell line ␣TN4-1 or rabbit lens cell line N/N1003A by using the Lipofectamine 2000 kit from Invitrogen. At 24 h after transfection, the cells were harvested for luciferase assay according to the company instruction manual. And the luciferase activity was determined by using the Luminoskan RS microplate reader from Thermo Labsystems Corp.
Quantitation of the Total Phospho-Pax-6 and Nonphospho-Pax-6-The Western blot gel and RT-PCR gel were analyzed with UN-SCAN-IT software from Silk Scientific Corp. Total pixel data were averaged from three or more different groups of samples of each species after normalization against the background.
Statistical Analysis-In the present study, all of the data presented are derived from at least three experiments. During data analysis, statistical analysis was conducted for all sets of data when necessary. Both average and S.D. value were calculated and included in the figures.

PP-1 and PP-2A Can Dephosphorylate Pax-6 in Vitro-To
explore what phosphatases may dephosphorylate Pax-6, we first tested the serine/threonine phosphatase-1 and -2A with the in vitro dephosphorylation assays. Our previous studies have shown that both PP-1 and PP-2A are present in the lens epithelial cells (8). To do so, the full-length Pax-6 protein or its PST domain was fused into the pGEX 4T-1 vector to make fusion proteins ( Fig. 2A). Both Coomassie Blue staining and Western blot analysis were applied to test their expression in Escherichia coli BL-21. Both full-length GST Pax-6 fusion protein (GST-mPax-6) (Fig. 2B, right lane) and the GST PST domain fused protein (GST-PST) (Fig. 2B, middle lane) were obtained. Since the full-length GST-mPax-6 only accounted for a small portion of the expressed GST-mPax-6 (the majority of the GST-mPax-6 full-length fusion proteins were truncated by bacterial enzymes and thus appeared in multiple bands with molecular mass less than 74 kDa), the purified GST-PST protein was used for in vitro dephosphorylation assays, and GST was used for the parallel control (Fig. 2B). Both GST-PST and GST substrates were labeled with [␥-32 P]ATP and the immunoprecipitated p38 kinase or the immunoprecipitated ␤-actin (for mock). The labeled GST or GST-PST was then subjected to dephosphorylation by purified catalytic subunits of PP-1 or PP-2A obtained from Calbiochem. After dephosphorylation, the released free 32 P was determined. As shown in Fig. 2, C and D, respectively, both PP-1 and PP2A were able to dephosphorylate Pax-6. In contrast, the GST did not generate a significant amount of free 32 P, indicating that both PP-1 and PP-2A specifically dephosphorylated Pax-6 but not GST (Fig. 2, C and D). Moreover, the labeling reaction with immunoprecipitated ␤-actin did not yield significant free 32 P from the reactions with either PP-1 or PP-2A (data not shown), further demonstrating that the specificity of dephosphorylation of the serine/threonine residues in Pax-6 by PP-1 and PP-2A. Under the present assay conditions, dephosphorylation of Pax-6 by PP-2C was not detected (data not shown). Together, these results suggest that PP-1 and PP-2A are able to dephosphorylate Pax-6 in vitro.
The Catalytic Subunits for PP-1 but Not for PP-2A Form a Complex with Pax-6 in Vivo-To demonstrate that PP-1 and PP-2A are able to dephosphorylate Pax-6 in vivo, we first examined whether the catalytic subunits for PP-1 or PP-2A could form interacting complexes. To do so, total proteins extracted from HLECs with or without UVA irradiation were subjected to co-immunoprecipitation assays. Two anti-Pax-6 antibodies were used for these studies. The two antibodies made either in mice or in rabbits by two separate companies recognize Pax-6 ( Fig. 3A). Since the rabbit anti-Pax-6 is a polyclonal antibody, it was used for immunoprecipitation, and on the other hand, the mouse anti-Pax-6 was used for Western blot analysis. When the proteins immunoprecipitated by rabbit anti-Pax-6 were probed with the antibody against the ␣ form of the catalytic subunits for PP-1 (PP-1␣), it was found that less than 50% of PP-1␣ was bound to Pax-6. UVA enhanced the interaction (Fig. 3B, panel 1). In the reverse immunoprecipitation-linked Western blot analysis, it was found that more than 90% of 46-kDa Pax-6 and 100% of 32-kDa Pax-6 were found bound to PP-1␣. Less than 10% of 46-kDa Pax-6 was left in the supernatant (Fig. 3B, panel 2). In the same way, we tested the interaction between the ␤ form of the catalytic subunits for PP-1 (PP-1␤) and Pax-6 and also the interaction between the catalytic subunit of PP-2A and Pax-6. As shown in panels 3 and 4 of Fig. 3B, PP-1␤ also interacts with Pax-6, but the affinity of the 46-kDa Pax-6 to PP-1␤ seems to be less than to PP-1␣ (Fig. 3B, panels 2 and 4). Although PP-2A was able to dephosphorylate Pax-6 in the in vitro assay (Fig. 2D), immunoprecipitation-linked Western blot analysis revealed no interactions between the catalytic subunit of PP-2A and Pax-6 ( Fig. 3C), between the scaffold subunit A of PP-2A and Pax-6 ( Fig. 3D), and between the regulatory subunit B of PP-2A and Pax-6 ( Fig. 3E). Thus, it is likely that PP-1 but not PP-2A is involved in direct dephosphorylation of Pax-6 in normal physiological conditions. UVA irradiation enhances the interactions of both PP-1␣ and PP-1␤ with Pax-6 (panels 1-4 of Fig. 3B). To further explore the interactions between PP-1␣ and Pax-6, as well as PP-1␤ and Pax-6, we treated human lens epithelial cells with calyculin A, an inhibitor for both PP-1 and PP-2A. Inhibition of PP-1␣ or PP-1␤ with calyculin A attenuates the interaction between PP-1 and Pax-6 ( Fig. 3F). Thus, both PP-1␣ and PP-1␤ are likely to participate in dephosphorylation of Pax-6.
PP-1␣ Can Dephosphorylate Pax-6 in Vivo, and Overexpression of PP-1␣ Leads to Dephosphorylation of Pax-6-To demonstrate that PP-1 actually dephosphorylates Pax-6 in vivo, we conducted the in vivo dephosphorylation assay as previously described (11,41). HLECs overexpressing hPax-6 were labeled with [ 32 P]orthophosphate, and the labeled proteins were immunoprecipitated with rabbit anti-Pax-6 antibody. The immunoprecipitated proteins were resuspended in dephosphorylation buffer and then incubated at 30°C 10 min for the dephosphorylation reaction. After the reaction, the released free 32 P was determined. As shown in Fig. 4A, a significant amount of free 32 P was released, and this release was mostly blocked when a specific PP-1 inhibitor, PP1-I2 (IC 50 ϭ 2 nM), or 5 nM calyculin A was present in the dephosphorylation reaction (Fig. 4A). However, a specific inhibitor for PP-2A, PP2A-I2 (IC 50 ϭ 30 nM), did not block the dephosphorylation reaction. This result clearly showed that PP-1 binds to Pax-6 and dephosphorylates it in normal human lens epithelial cells. To further explore this in vivo dephosphorylation, we examined the phosphorylation status of the 32-and 46-kDa Pax-6 in normal HLE cells, vectortransfected HLE cells, or PP-1␣ overexpression HLE cells. As shown in Fig. 4B, whereas 46-kDa Pax-6 appears in hypophosphorylation status in all three types of cells, the 32-kDa Pax-6 exists in both hyper-and hypophosphorylation status. Moreover, in the PP-1␣ overexpression cells, the ratio of hyperphospho-Pax-6 to hypophopho-Pax-6 is much lower than that in the vector-transfected or nontransfected HLE cells (Fig. 4C). Together, these results show that two different forms of Pax-6 with different phosphorylation status are present in human lens epithelial cells. Overexpression of PP-1␣ leads to enhanced dephosphorylation of the hyperphosphorylated 32-kDa Pax-6.
Inhibition of PP-1 by Calyculin A Enhances the Phosphorylation of Pax-6-Since overexpression of PP-1␣ did not change the phosphorylation status of the 46-kDa Pax-6, we next

PP-1 Dephosphorylates Pax-6
explored whether inhibition of PP-1 activity would affect the phosphorylation status of the 46-kDa Pax-6 as well as 32-kDa Pax-6. As shown in Fig. 5A, inhibition of PP-1 with 5 nM calyculin A led to the presence of multiple forms of hyperphosphorylated Pax-6, either 32 or 46 kDa in both vector and PP-1␣ transfected human lens epithelial cells. Analysis of the ratio between hyper-and hypophosphorylated Pax-6 through the density scanning of the Western blots revealed that the hyperphosphorylation of the 46-kDa Pax-6 observed in the vectortransfected clone due to inhibition of PP-1 by calyculin A was Rabbit anti-Pax-6 antibody was used to precipitate cellular Pax-6, and then mouse anti-Pax-6 was used to test the immunoprecipitation efficiency. Note that most cellular Pax-6 was precipitated by rabbit anti-Pax-6. B, co-immunoprecipitation between Pax-6 and PP-1␣ or between Pax-6 and PP-1␤. Most cellular Pax-6 was immunoprecipitated by anti-PP-1␣ or anti-PP-1␤ antibody, but only a small amount of PP-1␣ or PP-1␤ was immunoprecipitated by anti-Pax-6 antibody, suggesting that PP-1 is probably the major phosphatase to dephosporylate Pax-6 in HLECs. UVA irradiation enhances the interaction between Pax-6 and PP-1␣ or PP-1␤. C, co-immunoprecipitation between Pax-6 and PP-2A␣. Note that the anti-Pax-6 antibody could not precipitate down any PP-2A␣ protein, and the anti-PP-2A␣ antibody did not precipitate down any Pax-6 protein, suggesting that no interaction could be detected between Pax-6 and PP-2A␣. D, co-immunoprecipitation between Pax-6 and the scaffold subunit, A, of PP-2A. Note that the anti-Pax-6 antibody could not precipitate down any A subunit of PP-2A, and the anti-A subunit of PP-2A antibody did not precipitate down any Pax-6 protein, suggesting that no interaction could be detected between Pax-6 and PP-2A-A. E, co-immunoprecipitation between Pax-6 and the regulatory subunit, B, of PP-2A. Note that the anti-Pax-6 antibody could not precipitate down any B subunit of PP-2A, and the anti-PP-2A-B antibody did not precipitate down any Pax-6 protein, suggesting that no interaction could be detected between Pax-6 and PP-2A-B. F, treatment by calyculin A decreases the interaction between Pax-6 and PP-1␣ or PP-1␤. As the concentration of calyculin A increased, the binding of PP-1␣ (panel 1) or PP-1␤ (panel 2) to Pax-6 was attenuated. IP, immunoprecipitation; WB, Western blot.
almost reversed by overexpression of PP-1␣ in the pCI-PP-1␣transfected clone (Fig. 5B). For the 32-kDa Pax-6, inhibition of PP-1␣ by calyculin A caused more than 30% hyperphosphorylation in vector-transfected cells. In the pCI-PP-1␣ transfected clone, however, much reduced hyperphosphorylation in p32 Pax-6 was observed due to the action of the overexpressing PP-1␣ either without or with treatment by calyculin A (Fig. 5C). Together, these results indicate a counteraction of the overexpressed PP-1␣ to the inhibition function by calyculin A.
A further comparison of the phosphorylation status of the 46-and 32-kDa Pax-6 under treatments of different concentrations of calyculin A revealed that with the increase in its concentration, calyculin A resulted in gradual hyperphosphorylation status of both the 46-kDa Pax-6 (Fig. 5, D and E) and the 32-kDa Pax-6 (Fig. 5, D and F). In the absence of calyculin A treatment, the 46-kDa Pax-6 appeared in a single dephosphorylation band, and the 32-kDa Pax-6 appeared in two bands, both hypoand hyperphosphorylation. Under treatment with 2 nM calyculin A, the single dephosphorylation band of Pax-6 (p46) was reduced, and about 5% hyperphosphorylated Pax-6 (p46) appeared (Fig. 5, D and E); at the same time, about 70% hyperphosphorylated Pax-6 (p32) became visible (Fig. 5, D and F). However, a treatment with 5 nM calyculin A led to further increase in hyperphosphorylation of both forms of Pax-6 ( Fig. 5D). About 60% of p46 Pax-6 ( Fig. 5E) and 90% of p32 Pax-6 became hyperphosphorylated (Fig.  5F). These results further confirm that PP-1 is the protein phosphatase responsible for dephosphorylation of Pax-6 in human lens epithelial cells.
Silence of PP-1␣ and PP-1␤ but Not PP-2A␣ by Specific siRNAs Substantially Enhances the Hyperphosphorylation of Pax-6-To further confirm that PP-1 is responsible for dephosphorylation of Pax-6, we examined the phosphorylation status of Pax-6 with endogenous PP-1 and PP-2A knocked down by specific siRNAs against PP-1␣ or PP-1␤ or PP-2A␣. As shown in Fig. 6A, siRNA for PP-1␣ down-regulated up to 70% of PP-1␣ expression. When the endogenous PP-1␣ was down-regulated, about 20% of the 46-kDa Pax-6 became hyperphosphorylated (Fig. 6B), and 80% of the 32-kDa Pax-6 was found hyperphosphorylated (Fig. 6C). The siRNA for PP-1␣ had little effect on expression of PP-1␤ (Fig.  6D). When the siRNA for PP-1␤ was used for knockdown of PP-1␤ expression, 15% of the 46-kDa Pax-6 and also ϳ80% of the 32-kDa Pax-6 were hyperphosphorylated (Fig. 6, E and F). Thus, both PP-1␣ and PP-1␤ are involved in dephosphorylation of Pax-6. When specific siRNA for PP-2A␣ was used to knock down the expression of the endogenous PP-2A␣, no hyperphosphorylation of Pax-6 was detected (Fig. 7). This result further suggests that PP-2A is not responsible for the dephosphorylation of Pax-6.
Mutation Imitating Constant Phosphorylation or Constant Dephosphorylation of Pax-6 Modulates Its Function-To explore whether dephosphorylation of Pax-6 may have any effects on its function, we have overexpressed the wild type, TS360/361DD, and TS360/361AA mutant Pax-6 in human lens The labeled cells were harvested for protein extraction, and the extracted proteins were subjected to immunoprecipitation with a rabbit antibody against Pax-6. The protein pellet precipitated by anti-Pax-6 was used for dephosphorylation assays as previously described (11) without (no inhibitor column) or with general inhibitor (5 nM calyculin A column) or specific inhibitors for PP-1 (PP1-I2; IC 50 ϭ 2 nM; Calbiochem) or PP-2A (PP2A-I2; IC 50 ϭ 30 nM; Calbiochem). Note that a significant amount of free 32 P was released in the reactions without specific inhibitor against PP-1 or calyculin A. Inhibitor against PP-2A did not block the release of free 32 P. B, overexpression of PP-1␣ leads to dephosphorylation of the 32-kDa Pax-6 (p32). Since the 46-kDa Pax-6 (p46) is already in hypophosphorylated status, overexpression of PP-1␣ did not change its phosphorylation status. C, quantitation of dephosphorylation of the 32-kDa Pax-6 (p32) in B and unpresented Western blots by PP-1␣. Density scanning analysis revealed that more than 60% of the 32-kDa Pax-6 was dephosphorylated by overexpressed PP-1␣. DMSO, Me 2 SO. epithelial cells. As shown in Fig. 8A, the three types of Pax-6 are all expressed in HLE cells with similar mRNA levels (panel 1 of Fig. 8A) and protein levels (panel 3 of Fig. 8A). With the same mRNA extracted from the human lens epithelial cells expressing either mutant or wild type Pax-6, we analyzed expression of one of the downstream target genes, ␣B-crystallin. Previous

. Inhibition of PP-1 by calyculin A enhances phosphorylation of Pax-6 in human lens epithelial cells.
A, overexpression of PP-1␣ partially counteracts the action of calyculin A. In the absence of calyculin A treatment, overexpression of PP-1␣ resulted in dephosphorylation of Pax-6 (p32) but no action on Pax-6 (p46), since it is already dephosphorylated. Under the treatment of 5 nM calyculin A, hyperphosphorylation of both Pax-6 (p46) and Pax-6 (p32) was detected in both pCI-neo-PP-1␣-and pCI-neo vector-transfected clones, whereas the hyperphosphorylation level of Pax-6 (p46) in the PP-1␣-expressing clone was obviously attenuated than that in the pCI-neo clone. B, quantitation of dephosphorylation of the 46-kDa Pax-6 (p46) in A and unpresented Western blots. Density scanning analysis revealed that hyperphosphorylation of the 46-kDa Pax-6 derived from PP-1 inhibition by calyculin A was largely dephosphorylated by overexpressed PP-1␣. C, quantitation of dephosphorylation of the 32-kDa Pax-6 (p32) in A and unpresented Western blots. Density scanning analysis revealed that enhanced hyperphosphorylation of the 32-kDa Pax-6 derived from PP-1 inhibition by calyculin A was also significantly dephosphorylated by overexpressed PP-1␣. D, treatment of normal human lens epithelial cells with different concentrations of calyculin A leads to different phosphorylation status of Pax-6. In the absence of calyculin A (0 nM calyculin A), the 46-kDa Pax-6 (p46) appeared in dephosphorylation status, whereas the 32-kDa Pax-6 (p32) was present in both dephosphorylation status and the intermediate phosphorylation status. Under the treatment of 2 nM calyculin A, a light smear of phospho-Pax-6 (p46) and a much stronger phospho-Pax-6 (p32) could be detected. With the concentration of calyculin A increased to 5 nM, a strong smear of phospho-Pax-6 (p46) and prominent hyperphosphorylation bands of Pax-6 (p32) could be detected. E, quantitation of dephosphorylation of the 46-kDa Pax-6 (p46) in D and unpresented Western blots. Density scanning analysis revealed that hyperphosphorylation of the 46-kDa Pax-6 became dominant when 5 nM calyculin A was used to inhibit PP-1. F, quantitation of dephosphorylation of the 32-kDa Pax-6 (p32) in D and unpresented Western blots. Density scanning analysis revealed that more than 90% of 32-kDa Pax-6 became hyperphosphorylated when 5 nM calyculin A was used to inhibit PP-1. Together, these data support the conclusion that PP-1 dephosphorylates Pax-6 in vivo. FIGURE 6. Silence of PP-1␣ or PP-1␤ by specific siRNA enhances hyperphosphorylation of Pax-6. A, specific siRNA to PP-1␣ reduced 70% of PP-1␣ expression (lane 2) compared with the control siRNA (mutated PP-1␣ siRNA). In contrast, the siRNA to PP-1␤ had little effect on PP-1␣ expression (lane 3) but did cause hyperphosphorylation of Pax-6, because it knocked down PP-1␤ (described in D). Silence of PP-1␣ leads to appearance of a prominent hyperphosphorylation band of the 46-kDa Pax-6 (p46) and enhanced hyperphosphorylation of the 32-kDa Pax-6 (p32). Treatment of the cells with calyculin A resulted in a stronger level of Pax-6 hyperphosphorylation in both p46 and p32. B, quantitation of phosphorylation of the 46-kDa Pax-6 (p46) in A and unpresented Western blots. Density scanning analysis revealed that about 20% of the 46-kDa Pax-6 became hyperphosphorylated due to PP-1 knockdown by siRNA. About 35% of the 46-kDa Pax-6 became hyperphosphorylated due to PP-1 inhibition by calyculin A. C, quantitation of phosphorylation of the 32-kDa Pax-6 (p32) in A and unpresented Western blots. Density scanning analysis revealed that about 80% of the 32-kDa Pax-6 became hyperphosphorylated due to PP-1 knockdown by siRNA. About 90% of the 32-kDa Pax-6 became hyperphosphorylated due to PP-1 inhibition by calyculin A. D, specific siRNA to PP-1␤ reduced about 85% of PP-1␤ expression in comparison with the control siRNA (mutated PP-1␤ siRNA). On the other hand, the siRNA to PP-1␣ had little effect on PP-1␤ expression (lane 1) but did cause hyperphosphorylation of Pax-6, because it knocked down PP-1␣ (described in A). Silence of PP-1␤ led to the hyperphosphorylation of Pax-6 (both 46-kDa Pax-6 and 32-kDa Pax-6). E, quantitation of phosphorylation of the 46-kDa Pax-6 (p46) in D and unpresented Western blots. Density scanning analysis revealed that about 15% of the 46-kDa Pax-6 became hyperphosphorylated due to PP-1 knockdown by siRNA. About 40% of the 46-kDa Pax-6 became hyperphosphorylated due to PP-1 inhibition by calyculin A. F, quantitation of phosphorylation of the 32-kDa Pax-6 (p32) in D and unpresented Western blots. Density scanning analysis revealed that about 80% of the 32-kDa Pax-6 became hyperphosphorylated due to PP-1 knockdown by siRNA. A similar percentage of the 32-kDa Pax-6 became hyperphosphorylated due to PP-1 inhibition by calyculin A. studies have shown that overexpression of Pax-6 causes up-regulation of ␣B-crystallin mRNA (42). We therefore compared the mRNA expression level of ␣B-crystallin in the three types of Pax-6-transfected cells. As shown in Fig. 8B, in the HLE clone expressing mutant Pax-6 imitating the constant phosphorylation, the expression level of the ␣B-crystallin mRNA (Fig. 8B, top, left lane) was much higher than that in either wild type Pax-6-transfected (Fig. 8A, panel 1, right lane) or dephosphorylation mutant-transfected (Fig. 8B, top, middle lane) human lens epithelial cell clones. On the other hand, the mRNA level for ␣B-crystallin in the HLE clone expressing a Pax-6 mutant imitating constant dephosphorylation is clearly lower than that in wild type Pax-6-transfected cells and much lower than that in the clone expressing constant phosphorylated Pax-6 ( Fig. 8B). Thus, dephosphorylation of Pax-6 significantly modulates its function.

PP-1 Dephosphorylates Pax-6
Co-transfection of PP-1 and Pax-6 Can Down-regulate the Function of Pax-6, whereas Knockdown of PP-1 Enhances Pax-6 Transcriptional Activity-To further demonstrate that under physiological conditions, dephosphorylation of Pax-6 by PP-1 negatively modulates Pax-6 functions, we have conducted two lines of experiments. First, luciferase reporter gene assays using co-transfection of pCI-Pax-6 and p␣B-luc with pCI-PP-1␣ or pCI-PP-2A␣ into either mouse lens epithelial cell line, ␣TN4 -1, or rabbit lens epithelial cell line, N/N1003A, revealed that PP-1␣ but not PP-2A␣ was able to down-regulate the transcriptional activity of the promoter of a Pax-6 downstream gene, ␣B-crystallin, in either cell line (Fig. 9A). In both cell lines, the co-transfection of pCI-neo and p␣B-luc gave a basic level of luciferase activity, and this basic level was lower in ␣TN4 -1 than in N/N003A cells (Fig. 9A). The co-transfection of Pax-6 with p␣B-luc led to an increase in the luciferase activity compared with the basic level of luciferase activity (Fig. 9A), and this increase was abolished when pCI-PP-1␣ was included in the co-transfection with pCI-Pax-6 and p␣B-luc. On the contrary, the inclusion of pCI-PP-2A␣ in the co-transfection of pCI-Pax-6 and p␣B-luc did not change the increase (Fig. 9A). These results further demonstrate that PP-1 but not PP-2A display the functional modification of Pax-6 transcriptional activity.
Second, we conducted RNA interference in human lens epithelial cells. The RNA interference with oligonucleotides targeting PP-1␣, PP-1␤, PP-2A␣, or nonspecific control was able to knock down expression of PP-1␣, PP-1␤ (Fig. 6, A and D), and PP-2A (Fig. 7). Knockdown of PP-1␣ and PP-1␤ but not PP-2Aa significantly enhanced the transcription of the pax-6 downstream gene, ␣B-crystallin, further confirming that PP-1 is the phosphatase responsible for the dephosphorylation of Pax-6 under physiological conditions (Fig. 9, B and C).
Together, these results demonstrate that PP-1 is the major phosphatase dephosphorylating Pax-6 and thus actively modulating Pax-6 functions.

DISCUSSION
In the present study, we have demonstrated the following: 1) PP-1 and PP-2A dephosphorylate Pax-6 in the in vitro dephosphorylation assays; 2) the catalytic subunits for PP-1␣ and PP-1␤ (but not the catalytic subunit nor the scaffold subunit, A, and the regulatory subunit, B, of PP-2A) are able to form interactive complex with Pax-6, and more than 90% of Pax-6 are bound to PP-1 within normal human lens epithelial cells; 3) PP-1 directly dephosphorylates Pax-6 in the in vivo dephosphorylation assays; 4) two forms of Pax-6 (p46 and p32) are present in human lens epithelial cells (whereas the p46 exists in hypophosphorylated status, p32 exists in both hyper-and hypophosphorylated status); 5) although overexpression of PP-1␣ promotes dephosphorylation of Pax-6 (p32), inhibition of PP-1 activity with calyculin A or knockdown of PP-1 expression with specific siRNAs enhances hyperphosphorylation of both forms of Pax-6; 6) overexpression of human Pax-6 mutant imitating constant dephosphorylation at residue 360/361 substantially attenuates its ability to regulate its downstream target gene, ␣B-crystallin transcription; 7) co-transfection of PP-1␣ but not PP-2A␣ with Pax-6 could abolish the transcriptional function of a Pax-6 downstream gene promoter, the ␣B-crystallin promoter, whereas knockdown of the transcription of PP-1 could enhance the transcription of the Pax-6 downstream gene, ␣B-crystallin. These results confirm that PP-1 but not PP-2A directly dephosphorylates Pax-6 in human lens epithelial cells and that dephosphorylation by PP-1 modulates Pax-6 function.

PP-1 Is a Major Phosphatase That Dephosphorylates Pax-6 in Human
Lens Epithelial Cells-Pax-6 is an important transcription factor that regulates eye development in a variety of organisms from Drosophila to humans (14 -31). Mutations of the Pax-6 gene result in the absence of eyes in humans (30), and targeted expression of Pax-6 induces ectopic eye formation in Drosophila (31). Pax-6 appears in multiple forms with different molecular weights: p48, p46, p43, and p32 (32,33). Although the N terminus is heterogeneous in different forms, all of the forms have a conserved homeobox domain in the central region and an activating PST in the C terminus (Fig. 1). Sequence alignment of the PST domains revealed very high amino acid sequence homology among different vertebrate species and the relatively rich proline, serine, and threonine residues in this region. These features suggest that Pax-6, as a transcription factor, could be modulated by post-translational modifications, such as phosphorylation and dephosphorylation. Indeed, it has been shown that the function of Pax-6 can be modulated by several kinases, including p38, ERK (34), and homeodomain-interacting protein kinase 2 (35). In the present study, we have demonstrated that although both PP-1 and PP-2A are able to dephosphorylate Pax-6 in the in vitro dephosphorylation assays, in the immunoprecipitation-linked Western blot analysis, PP-2A and Pax-6 failed to form interactive complex, and thus, it is unlikely that PP-2A dephosphorylates Pax-6 in vivo. The fact that in the reciprocal immunoprecipitation-linked Western blot analysis, we observed that more than 90% of 46-kDa Pax-6 and 100% of 32-kDa Pax-6 proteins were FIGURE 9. Demonstration that dephosphorylation of Pax-6 by PP-1 changes its transcriptional activity in vivo. A, overexpression of PP-1␣ but not PP-2A␣ abolishes Pax-6-induced increase in the reporter gene activity. The luciferase reporter gene construct, p␣B-Luc, driven by the promoter of the ␣B-crystallin, a downstream gene of Pax-6, was co-transfected with either a vector, pCI-Neo, or a Pax-6 expression construct without any other construct or together with a PP-1␣ expression construct (pCI-PP-1␣) or together with a PP-2A␣ expression construct (pCI-PP-2A␣) in the presence of a control plasmid pSV40-Luc into mouse lens epithelial cells, ␣TN4 -1, or rabbit lens epithelial cells, N/N1003A. Overexpression of Pax-6 enhanced the luciferase activity from the basic level in both types of cells. Such enhancement was abolished by co-overexpression of PP-1␣ but not by co-overexpression of PP-2A␣. Luciferase activity was assayed as described under "Experimental Procedures." B, knockdown of PP-1␣ and PP-1␤ but not PP-2A␣ enhances expression of the Pax-6 downstream gene encoding ␣B-crystallin in human lens epithelial cells. RNA interference and RT-PCR were conducted as described under "Experimental Procedures." Note that knockdown of PP-1␣ and PP-1␤ barely changed the Pax-6 mRNA level but significantly enhanced expression of the ␣B-crystallin mRNA. C, quantitation of the ␣B-crystallin mRNA levels from three different experiments. The relative mRNA level for ␣B-crystallin was obtained through dividing the total ␣B-crystallin mRNA pixel by the Pax-6 mRNA pixel after normalization against ␤-actin. Note that knockdown of PP-1␣ and PP-1␤ led to more than 50% increase in the ␣B-crystallin mRNA expression. In contrast, knockdown of PP-2A␣ had little effect on the ␣B-crystallin mRNA expression. Thus, dephosphorylation of Pax-6 by PP-1 significantly changes its transcriptional activity. The corresponding number in B and C represents the same conditions. bound to PP-1 suggests that protein phosphatase-1 seems to be the major phosphatase that dephosphorylates Pax-6 ( Fig. 3B). This conclusion is further supported by the fact that knockdown of either PP-1␣ or PP-1␤ by specific siRNAs leads to hyperphosphorylation of a significant portion of the 46-kDa Pax-6 and a majority of the 32-kDa Pax-6 ( Fig. 6). The fact that siRNA only decreased up to 85% PP-1 expression and, moreover, only a small portion of cellular PP-1 was bound to Pax-6 ( Fig. 3B) explains why a large portion of 46-kDa Pax-6 is still in hypophosphorylation status in the PP-1 knockdown cells (Fig. 6).
In the present studies, we also observed that two forms of Pax-6 are present in human lens epithelial cells. Although the p46 Pax-6 only exists as the hypophosphorylated form, the p32 Pax-6 appears in both hyperphosphorylated and hypophosphorylated forms. Since the 32-kDa Pax-6 has slight high affinity to PP-1 (Fig. 3B), its presence of hyperphosphorylation status suggests the possibility that the two forms of Pax-6 proteins are differentially phosphorylated by the cognate kinases. Differential phosphorylation status of the two forms of Pax-6 may indicate a functional difference. First, they might have different target genes. Consistent with this possibility is the fact the 32-kDa Pax-6 contains only the homeodomain for DNA binding; in contrast, the 46-kDa Pax-6 contains two DNA binding regions: the paired domain and the homeodomain. Second, it is also possible that the two forms of Pax-6 may have similar target genes but different activation potential. These possibilities are currently under investigation.
Dephosphorylation Acts as a Molecular Switch-It has been well established that reversible phosphorylation and dephosphorylation at the serine and threonine residues on proteins play important roles in regulating gene expression, cell cycle progression (43), and apoptosis (44). Although protein phosphorylation through activation of various kinases has been the central subject of signal transduction studies and thus the field is well advanced, the study of protein dephosphorylation by protein phosphatases is now becoming established. Results from recent studies in numerous laboratories have shown that protein dephosphorylation has an important impact in regulating different cellular functions.
Dephosphorylation generally inactivates functions of a target molecule. In the present study, we demonstrate that PP-1 but not PP-2A directly dephosphorylates Pax-6. Dephosphorylation of Pax-6 has important functional consequence, as demonstrated from three aspects. First, a mutant imitating constant dephosphorylation at Thr-360/Ser-361 negatively regulates its downstream target gene coding for ␣B-crystallin (Fig. 7). In contrast, a mutant imitating constant phosphorylation at the same residues enhances expression of the same target gene. Second, reporter gene luciferase activity assays through the ␣B-crystallin promoter revealed that co-transfection of pCI-Pax-6 significantly increases the reporter gene activity. This increase derived from Pax-6 overexpression was abolished by PP-1␣ overexpression but not by PP-2A␣ overexpression (Fig.  9A). Finally, knockdown of PP-1␣ or PP-1␤ but not PP-2A␣ by specific siRNAs enhances hyperphosphorylation of Pax-6 ( Fig.  6) and also up-regulates the transcription of the Pax-6 target gene, ␣B-crystallin (Fig. 9, B and C).
Consistent with the fact that dephosphorylation of Pax-6 attenuates its transcriptional activity in human lens epithelial cells, dephosphorylation of p53, as we and others have recently shown, attenuates its transcriptional activity and also its proapoptotic activity in both lens and nonlens cells (11,45). More recently, Lin et al. (46) also demonstrated that dephosphorylation of Smad2/3 by another serine/threonine phosphatase, PPM1A/PP-2C␣, terminates TGF␤ signaling pathway. In addition, PP-2A dephosphorylation of ERK negatively modulates the ERK signaling pathway (47).
On the other hand, dephosphorylation may also activate functions of the target molecules. It has been shown that dephosphorylation of the Na ϩ ,K ϩ -ATPase by calcineurin enhances its enzyme activity (48). Similarly, dephosphorylation of Rb, a tumor suppressor at Thr-821, by protein serine/threonine phosphatase-1 promotes its ability to bind to members of the E2F family (49). Together, these studies suggest that dephosphorylation can act as a molecular switch, which regulates multiple cellular functions.