Phosphorylation within the MafA N Terminus Regulates C-terminal Dimerization and DNA Binding*

Phosphorylation regulates transcription factor activity by influencing dimerization, cellular localization, activation potential, and/or DNA binding. Nevertheless, precisely how this post-translation modification mediates these processes is poorly understood. Here, we examined the role of phosphorylation on the DNA-binding properties of MafA and MafB, closely related transcriptional activators of the basic-leucine zipper (b-Zip) family associated with cell differentiation and oncogenesis. Many common phosphorylation sites were identified by mass spectrometry. However, dephosphorylation only precluded the detection of MafA dimers and consequently dramatically reduced DNA-binding ability. Analysis of MafA/B chimeras revealed that sensitivity to the phosphorylation status of MafA was imparted by sequences spanning the C-terminal dimerization region (amino acids (aa) 279–359), whereas the homologous MafB region (aa 257–323) conveyed phosphorylation-independent DNA binding. Mutational analysis showed that formation of MafA dimers capable of DNA binding required phosphorylation within the distinct N-terminal transactivation domain (aa 1–72) and not the C-terminal b-Zip region. These results demonstrate a novel relationship between the phosphoamino acid-rich transactivation and b-Zip domains in controlling MafA DNA-binding activity.

Phosphorylation influences the activity of many transcription factors. However, the precise manner by which this posttranslational modification impacts regulation has been defined mechanistically in only a few cases (1,2). Here, we have focused on understanding how phosphorylation affects the DNA-binding potential of MafA and MafB, whose basic-leucine zipper (b-Zip) 2 region defines their dimerization and DNA-binding properties. There are two subfamilies of mammalian musculoaponeurotic fibrosarcoma (i.e. Maf) proteins, termed large and small Mafs (3). The small Maf proteins (MafF, MafG, MafK, and MafT) lack a transactivation domain and affect transcrip-tion through dimerization with related and distinct proteins (4 -7). The large Maf proteins (MafA, MafB, c-Maf, and NRL) contain an N-terminal transactivation domain (8 -11), which has considerable identity among MafA, MafB, and c-Maf (12,13).
Large Mafs are required in promoting many distinct physiological processes by binding as dimers to Maf-responsive elements and activating transcription (14,15). Among other properties, chicken L-Maf (termed MafA in mammals) is involved in lens development (16), mammalian MafB is required for segmentation of the hindbrain (17), mammalian c-Maf contributes to chondrocyte differentiation (9,18), and mammalian NRL functions in eye rod formation (10). Moreover, MafA and MafB have recently been shown to be essential within the mammalian pancreas, with islet ␣ and ␤ cell production requiring the actions of MafB during development and adult ␤ islet activity uniquely MafA (19 -21). In addition, large Maf proteins mediate cellular transformation in vitro and are overexpressed in human angioimmunoblastic T cell lymphomas and multiple myeloma and contribute directly to cancer progression (22)(23)(24).
Large Maf proteins appear to be heavily phosphorylated (27,28). Here, we first used mass spectrometry to identify the phosphoamino acids in MafA and MafB. A high phosphorylation state was shown to be uniquely important for the production of dimers capable of DNA binding in purified, full-length MafA, as phosphatase treatment induced multimerization and eliminated DNA binding. Exchanging the C-terminal b-Zip-span-ning sequences between MafA and MafB conferred phosphorylation-independent binding to MafA and sensitivity to MafB. Significantly, phosphorylation-dependent DNA binding was controlled by both the phosphorylation-rich N-terminal transactivation domain (amino acids (aa) 1-72) and the leucine zipper region. These results demonstrate that phosphorylation within the N-terminal region is not only critical to transactivation by enabling recruitment of co-activators like P/CAF (27), but also plays a novel role in DNA binding by mediating intramolecular interactions with the b-Zip domain.
Cell Culture Transfection and Nuclear Extract Preparation-HeLa monolayer cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% calf serum. Each expression plasmid (4 g) was transfected using Lipofectamine (Invitrogen) in cells plated on 100-mm dishes. Nuclear extracts were prepared 48 h after transfection in the absence of sodium orthovanadate (Na 3 VO 4 ), as described previously (30). Each experiment was repeated at least three times using independently isolated plasmid preparations.
In Vitro Phosphatase Treatment and the Dimerization/DNA Binding Assays-MafA-or MafB-transfected nuclear extract (25 g, 20-l total reaction volume) was incubated at 37°C in 1ϫ phosphatase buffer (20 mM Tris-HCl (pH 7.4), 1 mM dithio-threitol, 0.1 mM EGTA, 5 mM MgCl 2 , 1ϫ protease inhibitor mixture (Roche Applied Science)) with either 10 mM NaCl or 10 mM Na 3 VO 4 . (The endogenous large Maf phosphatases are inhibited by 10 mM Na 3 VO 4 ) (35). MafA or MafB protein levels were determined from 2 l of the reaction sample by Western blotting, and DNA-binding ability was analyzed from 9 l in the electrophoretic mobility shift assay with 32 P-labeled rat insulin II C1 element (5Ј-GGAAACTGCAGCTTCAGCCCCTC-3Ј) (30). MafA and MafB dimerization capabilities were determined with the remaining 9 l, which was mixed with DNAbinding buffer (30 l) and 1 l of cold 20 fmol/l C1 element. After 10 min at room temperature, the reaction mixture was separated on a 5% native-PAGE, and MafA or MafB protein levels were examined by Western blotting. These same assays were performed under similar conditions with 1.2 g of purified MafA and MafB treated with 6 units of calf intestine alkaline phosphatase (CIAP; Promega) in the presence or absence of phosphatase inhibitors (1ϫ general phosphatase inhibitor (Roche Applied Science), 10 mM sodium pyrophosphate (NaPP i ), or 10 mM Na 3 VO 4 ) for 1 h at 37°C. The following antibodies were used in these studies: rabbit ␣-mouse MafA (Bethyl Laboratories, 1225 (raised against an N-terminal peptide) and 1069 (raised against a C-terminal peptide)), rabbit ␣-mouse MafB (Bethyl Laboratories, 658 (raised against an N-terminal peptide)), mouse ␣-c-myc (9E10; Santa Cruz Biotechnology), ␣-rabbit IgG horseradish peroxidase conjugate (W401B; Promega), and ␣-mouse IgG horseradish peroxidase conjugate (W402B; Promega).

Phosphorylation Is Required for MafA Dimerization, but not
MafB-Because a wide range of MafA and MafB activities is controlled by phosphorylation, the extent of modification with purified MafA and MafB (Fig. 1A) was determined using an unbiased proteomics approach. Mass spectrometry analysis revealed that the phosphorylated serines and threonines within these highly charged proteins were concentrated principally within the N-and C-terminal regions (Fig. 1B and  supplemental Tables 1 and 2). The number of phosphorylated amino acids detected was similar to the estimate obtained from radiolabeled inorganic phosphate incorporation into endogenous MafA (26).
We next examined how phosphorylation influenced the biochemical properties of MafA and MafB, specifically in regard to SDS-PAGE mobility, DNA binding, and homodimerization. Purified MafA and MafB were incubated with CIAP in the presence and absence of Na 3 VO 4 , a general phosphatase inhibitor under the conditions of analysis. (Notably, NaPP i and a commercially available mixture of phosphatase inhibitors behaved similarly to Na 3 VO 4 (supplemental Fig. 1).) CIAP treatment increased the mobility of both MafA and MafB ( Fig. 2A), a result expected from earlier studies (29). Dephosphorylation of MafA in ␤ cell line nuclear extracts profoundly reduced insulin C1 element DNA-binding potency (30), as also observed here with purified MafA (Fig. 2B). In contrast, there was little to no effect on the DNA-binding ability of MafB under these conditions (Fig. 2B). The same DNA-binding properties were obtained with CIAP-treated MafA and MafB using the Maf-responsive element probes from the human ␤-globin and mouse crystalline genes (supplemental Fig. 2). These data demonstrated that phosphorylation only influenced the DNA-binding capabilities of MafA and not closely related and highly phosphorylated MafB.
CIAP-treated MafA and MafB were analyzed by native-PAGE to test whether the loss in MafA dimer formation caused the change in DNA-binding activity. The resulting MafA and MafB products were visualized by Coomassie Blue staining and by dividing the gel lanes into eight fractions and Western blotting the eluted proteins (Fig. 2C). MafA treated with only CIAP did not enter the gel, whereas MafA and MafB treated with Na 3 VO 4 ϩ CIAP migrated similarly to the control, untreated proteins. Dephosphorylation also shifted the mobility of MafB, but to a much lesser extent than the MafA protein.
Because protein charge and size affect mobility in this assay, we further evaluated these proteins by sucrose gradient ultracentrifugation. MafA treated with CIAP alone was found in samples containing high molecular weight proteins, here represented by the 30% sucrose fraction (Fig. 3), suggesting that dephosphorylation induced the multimerization of MafA. However, MafA protected by Na 3 VO 4 co-sedimented with phosphorylated and dephosphorylated MafB in lower sucrose fractions (Fig. 3). Significantly, only MafA dimers (i.e. (MafA) 2 ) and not the high molecular weight form (MafA dP ) n was active in gel shift assays. These data showed that a phosphorylation event(s) only controlled the stable formation of the dimer species required for DNA binding in MafA.
The MafA C-terminal Region Conveys Phosphatase Sensitivity-MafA/MafB chimeras were constructed to localize the region(s) controlling phosphorylation-responsive DNA binding in MafA (Fig. 4A). Nuclear extracts prepared from transfected cells were incubated with or without Na 3 VO 4 , conditions that allow or prevent (plus Na 3 VO 4 ) an endogenous phosphatase(s) in HeLa or islet ␤ cell lines from inhibiting MafA DNA-binding activity (30). Strikingly, phosphorylationindependent dimer formation and DNA-binding activity were obtained upon exchanging the C-terminal region of MafA with that of MafB (compare MafAB and MafAAB with MafA and MafB in Fig. 4B). Significantly, the SDS-PAGE mobility of these dephosphorylated protein chimeras was expectedly faster (also see Fig. 2A), whereas there was no reduction in steady-state protein levels (Fig. 4B). Furthermore, the MafA C terminus also imparted phosphorylation dependence to MafB in these assays (compare MafBA and MafBBA with MafA and MafB in Fig. 4B). These data illustrated the importance of sequences within the leucine zipper-spanning region of MafA and MafB in regulating phosphorylation-dependent DNA binding (i.e. MafA, aa 279 -359; MafB, aa 257-323).   APRIL 23, 2010 • VOLUME 285 • NUMBER 17

JOURNAL OF BIOLOGICAL CHEMISTRY 12657
Phosphatase Sensitivity Is Not Due to Phosphorylation within the MafA C Terminus-To determine whether the unique C-terminal phosphoamino acids at Ser 290 , Ser 297 , and Ser 342 ( Fig. 1) directly influenced DNA binding, a compound alanine mutant was constructed and transfected into HeLa cells (MafA-C3A in Fig. 5A). S290A/S297A/S342A and wild type MafA had similar DNA-binding ability, and both were sensitive to phosphatase treatment (Fig. 5B). Moreover, the cellular phosphatase(s) did not reduce the inherent DNA-binding activity of MafA(233-359), a myc-tagged construct expressing only the C-terminal b-Zip-spanning region of MafA (dephosphorylation did not affect (MafA(233-359)) 2 supershifted levels in lanes 5 and 6 of Fig. 5C as well). These data strongly indicated that the phosphorylated amino acid(s) regulating MafA-binding activity resided outside the MafA C-terminal b-ZIP region.
Phosphoamino acid mutants within the MafA N terminus were made to define their role in (MafA) 2 formation more precisely. It is noteworthy that many more phosphosites were found within this region of MafB by mass spectrometry than MafA (Fig. 1). We believe our inability to detect phosphorylation at conserved Ser 60 , Ser 61 , Ser 64 , Ser 65 , and Ser 67 is due a technical limitation(s) in our analysis, as even the corresponding unphosphorylated peptides were not found (supplemental Tables 1 and 2). Additionally, the phosphorylation-independent-binding properties of MafA(73-359) and MafBBA(77-359) indicated that phosphosite usage within the N-terminal region of MafA and MafB was conserved (Fig. 6B), and mutants were constructed accordingly. Because single and limited combination phosphosite mutants had little or no significant effect on MafA DNA binding (supplemental Fig. 4), comprehensive Ser/Thr to Ala or Glu mutants were constructed to block or mimic the phosphorylation state of the N-terminal region (Fig. 7A). Meaningfully, MafA-N13E migrated to the same position as Na 3 VO 4 -protected  MafA after SDS-PAGE (Fig. 7B). The ability of MafA-N13E to form stable dimers and bind DNA was also unaffected by the endogenous phosphatase(s), with the level and mobility indistinguishable from phosphorylated MafA (Fig. 7B). In contrast, MafA-N13A behaved like unphosphorylated MafA. Thus, MafA-N13A-binding activity was comparatively poor, and it migrated like dephosphorylated MafA after SDS-PAGE (Fig. 7B). These results demonstrate that multiple phosphorylation events within the Nterminal region are fundamental to the detection of (MafA) 2 or (Maf-BBA) 2 dimers and consequently DNA binding.

DISCUSSION
Crystal structure analysis of the fos-jun heterodimer, GCN4 homodimer, and MafG homodimer demonstrated that b-Zip proteins dimerize and bind DNA through a continuous ␣-helix (36 -38), with the basic region principally influencing cis-element binding and the coiled-coil leucine-zipper interface controlling dimerization. This highly structured region can independently mediate dimerization and DNA binding in vitro (39 -42). Although the MafA b-Zip region was capable of DNA binding by itself (Fig. 5), we found that phosphorylation within the distinct N-terminal activation domain was also required in the wild type protein. In contrast, this activity in MafB was unaffected by the state of phosphorylation, although analysis of the MafBBA chimera showed that its highly modified N-terminal region conveyed phosphosensitivity. The C-terminal region of MafA (aa 279 -359) and MafB (aa 257-323) was shown to respond distinctively to N-terminal domain phosphorylation, apparently by their ability to mediate stable dimer formation. These data illustrate a novel mechanism controlling transcription factor activity, involving phosphoregulated interactions between two distinct functional domains of MafA.
Mass spectrometry analysis of purified MafA and MafB identified many common and unique sites within these highly phosphorylated proteins (Fig. 1). Unfortunately, we were unable to detect MafA phosphorylation at Ser 61 , described for glycogen synthase kinase 3 in vitro or at Ser 65 , with the latter required for both glycogen synthase kinase 3 priming and ubiquitin-mediated degradation (26 -28). Our underestimation of the phosphory-  lation events is likely a reflection of poor peptide generation due to the high charge density of the region. The ability to detect Ser 49 , Thr 53 , and Thr 57 phosphorylation in MafA only after limited CIAP treatment supports this idea (supplemental Table 1). Further evidence for phosphorylation within the N-terminal region results from the ability to impart instability to MafA in the S65E phosphomimetic mutant (26). Moreover, MafA-N13E migrated like the heavily phosphorylated wild type protein after SDS-PAGE, whereas the behavior of N13A was similar to the unphosphorylated protein (Fig. 7B). The ability of the MafB N-terminal region to mediate the formation of stable (MafBBA) 2 dimers (Fig. 6) also implies that this area in both proteins is highly phosphorylated, with detection of the MafA Ser 65 equivalent phosphorylation at Ser 70 in MafB indicating that the overall pattern is possibly conserved (Fig. 1).
Because phosphorylation by glycogen synthase kinase 3 at Ser 61 , Thr 57 , Thr 53 , and Ser 49 potentiates MafA transactivation (27), we examined how the N13A and N13E mutants influenced the transcriptional effectiveness of chimeras between MafA aa 1-233 and the Saccharomyces cerevisiae transcription factor Gal4 DNA-binding domain. N13E activity was not significantly different from the wild type, whereas N13A was compromised severely (supplemental Fig. 5). We conclude that a high level of N-terminal region phosphorylation is not only necessary for detection of (MafA) 2 , but also transactivation domain strength.
Although many transcriptional regulators have been shown to be phosphoproteins, in only a few cases has phosphorylation been shown to affect DNA-binding activity directly. For example, phosphorylation within the basic region can be either a positive (e.g. Gal4) (43) or negative (PRH/Hex) (31) influence. It has also been proposed that dimerization of STAT1 is regulated by phosphorylation (32,33), although others suggest that STATs exist as inactive dimers in the absence of phosphorylation (44,45). We believe our data represent the first example of a distinct phosphorylation-rich domain in a transcription factor-stabilizing dimer formation and thereby DNA-binding domain activity. On the other hand, it is unclear whether the dephosphorylated high molecular weight multimers of MafA are due to monomer and/or dimer oligomerization (see Figs. 2 and 3). The phosphoregulatory mechanism described here is not unique to MafA, for at least c-Maf appears to be controlled in a similar manner (supplemental Fig. 6). It was striking that phosphatase(s)-sensitive DNA binding was imparted by multiple phosphorylation events rather than a single one as commonly found and that responsiveness was transferable. The C-terminal control region sequences of MafA and MafB spanned both conserved and nonconserved sequences, and we presently do not know which specific amino acids are essential to this response. We propose that interactions between the phospho-rich activation domain and b-Zip region modulate MafA structure and DNA-binding activity, presumably also creating unique interfaces for co-regulatory proteins.
A key issue is whether the C-terminal phosphoregulatory sequences of MafA defined here in vitro serve in control in vivo. Large Maf proteins play roles in development and tissue differentiation within the brain (MafB) (17), eye (NRL and MafA/L-Maf) (10,16), cartilage (c-Maf) (18), kidney (MafB) (46), and pancreas (MafB) (19,47). However, their significance in these situations may reflect differences in expression rather than fundamental activation properties. Evidence supporting distinctions in activation come from overexpression studies showing that MafA is capable of inducing the normally silent endogenous insulin gene in islet ␣ (glucagon ϩ ) cells (48), liver cells (49,50), and intestinal cells (51,52). Under similar experimental conditions, MafB activates glucagon (and not insulin) in islet ␤ cells. Most pertinent to the present studies, analysis of MafA/B chimeras in the chick in ovo electroporation assay demonstrated that insulin gene activation specifically depended upon MafA C-terminal region sequences (i.e. MafBAA, Fig. 4) (34). Notably, MafB ϩ insulin ϩ cells generated during human embryonic stem cell differentiation were dysfunctional unless becoming MafA ϩ insulin ϩ (53), supporting further efforts to understand how MafA and MafB regulate (at least) islet ␤ cell function.