INTRODUCTION

NF-E2 is an obligate heterodimer of basic leucine zipper polypeptides consisting of a larger p45 polypeptide and a smaller subunit belonging to the p18/Maf family (1-5). Of the two subunits of NF-E2, expression of p18/Maf is ubiquitous, while that of p45 is restricted to the erythroid and megakaryocytic lineages (1, 6). Indeed, intact p45 gene and its expression are required for transcriptional regulation of globin genes (Refs. 1 and 7-9 and references therein) as well as for normal differentiation of the megakaryocytes (10).

As a transcriptional activator, NF-E2 exerts its function by binding to enhancers or promoters at a consensus sequence that are also binding sites of AP1 (Refs. 1 and 11-13 and references therein). The activation domain of p45 is located between amino acids 1-206 (9). Most likely, NF-E2 functions through DNA binding as well as interactions with other cell type-specific and ubiquitous co-activators. For example, binding of p45 with hormone receptor(s) and CREB-binding protein (14) potentiates transcriptional activation mediated by hormones. NF-E2 binding to its cognate sequence also induces local nucleosome disruption (15, 16).

The NH₂-terminal region of p45 has been implicated as the transcriptional activation domain of NF-E2 (9), and it contains 18% of proline residues. Interestingly, during data base search and visual comparison of the p45 sequence, we have noticed the sequence PPPPY located at amino acids 79-83 of human p45. In mouse, it is PPPSY. This sequence fits the consensus of the so-called "PY" motifs, XPPXY, in which P is a proline, Y is a tyrosine, and Xs are any amino acid (17, 18; Fig. 1). The PY motifs are ligands capable of binding to the WW domains, which in turn were first identified in the proto-oncogene Yes-associated protein (YAP),¹ dystrophin, transcriptional regulator FE65, and others (19). These domains are of the length 38 amino acids, and they contain  strands grouped around four aromatic positions (17, 20). Two of these positions are most frequently occupied by tryptophans, hence the name "WW" domain was given. The WW domains are found in a number of unrelated proteins, including human and mouse YAP (hYAP, mYAP), human dystrophin (hDys), human ORF1, yeast Rsp5, yeast Ess1 (yEss1), fission yeast Pub1, mouse NEDD4 (mNEDD4), human RPF1/NEDD4 (hRPF1), and FE65 (reviewed in Ref. 17). The functions of these proteins range from cell cycle control of yEss1 (21), cell cycle control and ubiquitin ligase activity of Pub1 (22), to the ubiquitin ligase activity of Rsp5 (23, 24) and mNEDD4 (25), and transcriptional co-activator properties of the hRPF1 (26).

As demonstrated below, p45 indeed could interact with several of these proteins, in vivo and in vitro, through specific interaction between the PY motif at amino acids 79-83 and the different WW domains. We also show that RNA pol II-CTD and p45 bind similar WW domains, including the mNEDD4-WW2. The study has suggested new pathways through which the processes of transcriptional activation by the NF-E2 molecule could be regulated in erythroid and megakaryote cells.

EXPERIMENTAL PROCEDURES

p45 cDNA was amplified from K562 RNA by RT-PCR (27) with primers A and B and blunt-end-cloned into the AvaI site of pGEX-2T. To create mutations in the PY motif of p45, two fragments were first generated by PCR amplification of p45 cDNA with primer pairs B/C and A/D, respectively. These two fragments were then used as the templates for PCR with primers A and B. The final product, p45(P81A,Y83G), was blunt-end-cloned into the AvaI site of pGEX-2T. The p45 protein expressed from this plasmid contains two amino acid substitutions in the PY motif: Pro  Ala at 81 and Tyr Gly at 83. Cloning of GST-WBP1-PY was described previously (18). hYAP-WW, mYAP-WW1, mYAP-WW2, mNEDD4-WW1, mNEDD4-WW2, mNEDD4-WW3, yEss1-WW, and hDys-WW were individually PCR-amplified from the appropriate cDNAs with primer pairs E/F, G/H, I/J, K/L, M/N, O/P, Q/R, and S/T, respectively, and ligated at the BamHI-EcoRI sites of pGEX-2TK. Protein kinase A-phosphorylatable GST-CTD fusion protein was kindly provided by Michael E. Dahmus (University of California, Davis, CA).

Sequences of all the primers (A-Z) used in this study are available upon request. Authenticity of the clones was confirmed by restriction enzyme digestion and DNA sequencing. p45 and p45(P81A,Y83G), respectively, were purified with GSH-agarose beads (Sigma) from isopropyl-1-thio--D-galactopyranoside-induced extracts of Escherichia coli cells harboring the individual plasmids (28). Different WW domains and WBP1-PY motif fusions were purified in a similar way.

For preparation of ³²P-labeled probes, equal amounts of GST-p45- and GST-p45(P81A,Y83G)-bound glutathione-agarose were used in a 4 °C phosphorylation reaction with bovine heart muscle kinase (Sigma) and [-³²P]ATP (29). After phosphorylation, the beads were washed, and proteins were eluted with a buffer containing glutathione. To prepare the blots, approximately 4 µg each of GST, or GST-WW fusion proteins, were electrophoresed on a 10% SDS-polyacrylamide gel and blotted onto nitrocellulose membranes. The membranes were first blocked with GST-expressing E. coli extract in Hyb75 (29), followed by hybridization with the ³²P-labeled GST-p45 or GST-p45(P81A,Y83G) for 4 h at 4 °C in the same blocking solution (~50,000 cpm/ml). After hybridization, the membranes were washed four times with Hyb75 containing 1% non-fat dry milk, air-dried, and exposed to x-ray films.

p45-(38-115) and p45-(38-115: P81A, Y83G) fragments were individually PCR-amplified with primers U/V, using the p45 and p45(P81A,Y83G) cDNAs, respectively, as the templates. PCR products were digested with EcoRI and PstI and cloned into the same sites in Gal4-DNA binding domain (GBD) vector pGBT9, resulting in the plasmids pGBD-p45-(38-115) and pGBD-p45-(38-115:P81A,Y83G). The NH₂-terminal hRPF1 coding for amino acids 1-553 (N-hRPF1) was amplified by RT-PCR from KG-1 cell RNA using the primers W and X, digested with HindIII and BamHI, and cloned into yeast Gal4 activation domain (GAD) vector pACT2-I at NcoI and BamHI sites. Ligation between the HindIII site of N-hRPF1 and the NcoI site of pACT2-I was facilitated by an annealed oligo consisting of the primers Y and Z. This cloning resulted in the generation of plasmid pGAD-N-hRPF1. pGAD-hYAP-(1-345) was constructed by ligation of a NcoI-XhoI fragment from hYAP cDNA at the same sites in pACT2-I.

The yeast two-hybrid assay was carried out as described (30, 31). Different combinations of the above GBD- and GAD-based plasmids were co-transformed into yeast HF7c and tested for their ability to grow on -Leu-Trp-His plates.

RESULTS

The interaction between p45 and different WW domains is analyzed by Far-Western blot analysis. As a positive control, we first studied the binding of the PY motif of WBP-1 to the WW domains by this approach (Fig. 2). The WBP-1 PY motif, of the sequence GTPPPPYTVG, is known to interact with the YAP WW domains (17-19). As shown, the WBP-1 PY motif binds the three YAP-WW domains with approximately equal affinities. It also binds mNEDD4-WW2 and hDys-WW with lower and varying affinities (Fig. 2B).

Far-Western analysis then demonstrates that p45 indeed binds to a subset of the WW domains tested (Fig. 3). As shown in Fig. 3B, binding between p45 and mNEDD4-WW2 is the strongest (lane 5, Fig. 3B). p45 also binds efficiently to hYAP-WW and mYAP-WW1 (lanes 1 and 2, Fig. 3B), but less efficiently to mYAP-WW2 (lane 3, Fig. 3B). However, p45 does not bind to the other WW domains tested (lanes 4, 6, and 7, Fig. 3B) as well as GST-WBP1-PY or GST (lanes 8 and 9, Fig. 3B).

To test whether PY motif at amino acids 79-83 of p45 is essential for the observed interaction between p45 and the WW domains, we then carried out Far-Western blot analysis using the GST-p45 and GST-p45(P81A,Y83G) mutant as the hybridization probes. As shown in Fig. 3C, the two amino acid substitutions at positions Pro⁸¹ and Tyr⁸³ of p45 abolish its interaction with all of the tested WW domains (Fig. 3C). Far-Western blot hybridization with cold p45 gave the same results.² Note that in the blot of Fig. 3C, a NIPP1-(1-203) polypeptide fused to GST was included as a positive control. NIPP1 is a novel p45-binding protein,³ and its interaction with p45 is not modulated through the PY motif of p45 (compare lanes 10 of Fig. 3, B and C).

The yeast two-hybrid system was used to test whether p45 could interact in vivo with two of the WW domain-containing proteins, hRPF1 and hYAP. hRPF1 is the human homolog of mNEDD4 (19, 24, 25). Plasmids expressing GBD only, GBD fusion of p45-(38-115), and GBD fusion of p45-(38-115:P81A,Y83G), respectively, are individually co-transformed into yeast HF7C strain with plamids expressing GAD only, GAD fusion of hRPF1-(1-553), and GAD fusion of hYAP-(1-345). As indicated by the data of Fig. 4, both hRPF1 and hYAP interact with p45-(38-115) in yeast cells, but not with p45-(38-115:P81A,Y83G) containing the mutated PY motif.

Recently it was shown that yeast RNA pol II is the substrate of the ubiquitin ligase Rsp5. This reaction is mediated through binding between the CTD domain of RNA pol II and the NH₂-terminal part of Rsp5 (24). RNA pol II-CTD contains several py motifs with a consensus "XSPXY." We have tested whether the CTD of mammalian RNA pol II could also bind mNEDD4, the mammalian homolog of Rsp5, through one of the three WW domains of the latter. Indeed, as shown in Fig. 5, CTD of mouse RNA pol II binds efficiently to the YAP WW domains as well as to mNEDD4-WW2, but not to the WW1 and WW3 of mNEDD4, yEss1-WW, hDys-WW, or GST (Fig. 5B). This clearly indicates that mammalian RNA pol II-CTD could bind to the mammalian ubiquitin ligase. Furthermore, this binding, in mouse or humans, is most likely mediated through the WW2, but not with the other WW domains of the ubiquitin ligase(s) (also see "Discussion"). It is interesting to note that RNA pol II-CTD binds the same set of WW domains as p45, although the affinities toward different WW domains are different between the two probes (compare Figs. 3B and 5B).

DISCUSSION

NF-E2 activates transcription of the globin genes through binding to the locus control regions of the globin loci (reviewed in Refs. 32 and 33). Alternatively, it can activate transcription of genes such as porphobilinogen deaminase through binding to their upstream promoters (11, 34). Similar to the other sequence-specific, DNA-binding transcription factors (reviewed in Refs. 35 and 36), these activation processes by NF-E2 must involve the interaction of NF-E2 with other transcription activators or co-activators, as exemplified in Ref. 14. The present identification of the PY motif of p45 as a ligand for specific WW domains has revealed a new class of interacting proteins through which the trans-activation function of NF-E2 could be regulated.

Far-Western blot analysis demonstrates that the p45 subunit of NF-E2, as well as the CTD of RNA pol II, binds specific sets of the WW domains (Figs. 3 and 5). The various WW domains of several WW domain-containing proteins and the relative strength of their interaction with p45 and RNA pol II-CTD are listed in Table I.

Table I. Different WW domains binding with p45/NF-E2 and RNA pol II-CTD Protein/species Position Sequence of WW domains Accession no. Binding to p45 NF-E2 Binding to RNA pol II-CTD 11                 30      38       48a mYAP-1 151 VPLPAGWEMAKTSS-GQRYFLNHNDQTTTWQDPRKAMLS X80508 +++ ++++ mYAP-2 218 GPLPDGWEQAMIQD-GEVYYINHKNKTTSWLDPRLDPRF X80508 + ++++ hYAP 171 VPLPAGWEMAKTSS-GQRYFLNHIDQTTTWQDPRKAMLS X80507 +++ ++++ mNEDD4-1 139 SPLPPGWEERQDVL-GRTYYVNHESRRTQWKRPSPDDDL U96635     mNEDD4-2 295 SGLPPGWEEKQDDR-GRSYYVDHNSKTTTWSKPTMQDDP U96635 ++++++ +++ mNEDD4-3 350 GPLPPGWEERTHTD GRVFFINHNIKKTQWEDPRLQNVA U96635 NDb   yEss1 29 TGLPTPWTARYSKSKKREYFFNPETKHSQWEEPEGINKD P22696     hDys 3052 TSVQGPWERAISPN-KVPYYINHETQTTCWDHPKMTELY P11532     hRPF1-1 218 SPLPPGWEERQDIL-GRTYYVNHESRRTQWKRPTPQDNL D42055 hRPF1-2 375 SGLPPGWEEKQDER-GRSYYVDHNSRTTTWTKPTVQATV D42055 +++c ND hRPF1-3 448 GFLPKGWEVRHAPN-GRPFFIDHNTKTTTWEDPRLKIPA D42055 hRPF1-4 500 GPLPPGWEERTHTD-GRIFYINHNIKRTQWEDPRLENVA D42055 yRsp5-1 228 GRLPPGWERRTDNF-GRTYYVDHNTRTTTWKRPTLDQTE L11119 yRsp5-2 331 GELPSGWEQRFTPE-GRAYFVDHNTRTTTWVDPRRQQYI L11119 ND  +++ d yRsp5-3 387 GPLPSGWEMRLTNT-ARVYFVDHNTKTTTWDDPRLPSSL L11119 a Numbers indicate the amino acid position in WW domain as shown for hYAP-WW in Ref. 20. b ND, not determined. c p45-(38-115) interacted with hRPF1-(1-553) in yeast two-hybrid system. d The NH2 terminus of RSP5 binding with RNA pol II-CTD contains all of the three WW domains listed (see Ref. 24).

Sequence specificities affecting the interaction between different PY motifs and WW domains have been studied by biochemical and structural approaches (18, 20, 37, 38). In particular, NMR analysis of the hYAP-WW domain-bound WBP1-PY motif indicated that Trp³⁹ in the WW domain contacts the 2nd and 3rd prolines of PPPPY. In addition, Leu³⁰ and His³² contact the tyrosine (20). Binding site for the peptides is a large hydrophobic patch on the WW domain surface formed by side chains of Tyr²⁸, Leu³⁰, Trp³⁹ and by the methyl groups of one or more threonines at positions 36-38. Consistent with the above, and the extensive data on site-directed mutagenesis of the PY motif of WBP1 (37), mutation at the 3rd proline and the tyrosine of the p45 PY motif abolished its binding with the WW domains (Fig. 3C).

A close examination of the sequences listed in Table I suggests that the presence of three clustered threonines at 36-38 is one of the major determinants for a strong interaction between a WW domain and p45. The binding affinity is drastically reduced when one of the threonine's is substituted with different amino acids. Furthermore, a change of Leu³⁰ to Val³⁰ in mNEDD4-WW2 appears to result in a relative higher affinity toward p45, in comparison with those of mYAP-WW1 and hYAP-WW (Table I). Following the above reasoning, we anticipate that it is the WW2 domain, but not WW1, WW3, or WW4 of hRPF1, that is responsible for the interaction in vivo between p45-(38-115) and N-hRPF1 (Fig. 4).

Although the exact functions of the motif-specific interaction between p45 and WW domains await further investigations, it is possible that the WW domain-containing proteins are involved in the regulation of the NF-E2 activity via one or more of the following mechanisms. First, p45 interacts efficiently with YAP (Figs. 3 and 4). Since YAP can associate with the SH3 domain of the Yes tyrosine kinase, it is possible that YAP-like protein(s) may target some of its WW domain-binding proteins, such as p45 and RNA pol II, for phosphorylation by kinases. Thus, activity of NF-E2 could be regulated by phosphorylation at specific residues. Second, hRPF1 has been shown to function as a co-activator/potentiator for hormone receptor-mediated transcriptional activation (26). Thus, hRPF1 or its family members may act, through direct binding to p45, as co-activators or even repressors for NF-E2 regulated transcription in erythroid and/or megakaryotic cells.

Third, like its homologs, Rsp5 (23, 24) and mNEDD4 (25), hRPF1 may also possess ubiquitin ligase activity. As demonstrated (24), Rsp5 and its homologs, when bound to RNA pol II-CTD, may ubiquitinate the RNA polymerase II. As shown in Figs. 3 and 5, p45 and RNA pol II are capable of binding to a similar subset of WW domains. Furthermore, p45 interacts with hRPF1 in vivo (Fig. 4). Thus, it is likely that mNEDD4 or hRPF1 negatively regulates NF-E2 activity by ubiquitinating p45 and targeting it to the proteasome degradation pathway. It is also possible that through utilization of different WW domains, Rsp5 homologs such as the mNEDD4/hRPF1 or others could serve as a binding bridge between NF-E2 and RNA polymerase II. Such a tertiary complex may play a role during transcriptional activation of erythroid-specific genes by NF-E2. Alternatively, DNA-bound p45 may recruit mNEDD4/hRPF1-like proteins to ubiquitinate nearby histones, a reaction that may lead to an open chromatin structure (39, 40). p45 and the bound WW domain proteins may also form tertiary complexes with other nuclear proteins such as the protein kinases of splicing factors, as identified recently (41).

Finally, it should be noted that there is an increasing number of WW domain-containing proteins found in cells carrying out diverse regulatory functions (Ref. 42). Indeed, three more such proteins belonging to the Rsp5/NEDD4 family have recently been cloned by a ligand screening procedure (38), and by the yeast two-hybrid system.⁴