The Novel WD-repeat Protein Morg1 Acts as a Molecular Scaffold for Hypoxia-inducible Factor Prolyl Hydroxylase 3 (PHD3)*

Hypoxia-inducible factor-1 (HIF-1), a transcriptional complex composed of an oxygen-sensitive α- and a β-subunit, plays a pivotal role in cellular adaptation to low oxygen availability. Under normoxia, the α-subunit of HIF-1 is hydroxylated by a family of prolyl hydroxylases (PHDs) and consequently targeted for proteasomal degradation. Three different PHDs have been identified, but the difference among their in vivo roles remain unclear. PHD3 is strikingly expressed by hypoxia, displays high substrate specificity, and has been identified in other signaling pathways. PHD3 may therefore hydroxylate divergent substrates and/or connect divergent cellular responses with HIF. We identified a novel WD-repeat protein, recently designated Morg1 (MAPK organizer 1), by screening a cDNA library with yeast two-hybrid assays. The interaction between PHD3 and Morg1 was confirmed in vitro and in vivo. We found seven WD-repeat domains by cloning the full-length cDNA of Morg1. By confocal microscopy both proteins co-localize within the cytoplasm and the nucleus and display a similar tissue expression pattern in Northern blots. Binding occurs at a conserved region predicted to the top surface of one propeller blade. Finally, HIF-mediated reporter gene activity is decreased by Morg1 and reduced to basal levels when Morg1 is co-expressed with PHD3. Suppression of Morg1 or PHD3 by stealth RNA leads to a marked increase of HIF-1 activity. These results indicate that Morg1 specifically interacts with PHD3 most likely by acting as a molecular scaffold. This interaction may provide a molecular framework between HIF regulation and other signaling pathways.

Mammalian cells respond to reduced oxygen tensions by the expression of several dozens of genes, most of them are directly or indirectly induced by the transcriptional complex hypoxia-inducible factor (HIF) 2 (1,2). HIFs are heterodimeric basic helix-loop-helix proteins belonging to the family of PAS (PER-ARNT-SIM) domain transcription factors (3,4). They consist of an ␣ subunit of which three isoforms have been identified in human (HIF-1␣, -2␣, and -3␣) and a ␤ subunit (HIF-1␤/ ARNT) (5)(6)(7). Under normoxia, HIF-1␣ is continuously expressed in the cell but immediately degraded via the proteasomal pathway after ubiquitination (8). The von Hippel-Lindau protein acts as a particle recognition protein of the responsible E3 ubiquitin-ligase complex if two distinct prolyl residues within a region, referred to as the oxygendependent degradation domain of HIF-1␣, Pro 302 and/or Pro 564 , are hydroxylated (9 -13). The site-specific hydroxylation of HIF prolyl residues is catalyzed by a conserved class of 2-oxoglutarate-and Fe(II)-dependent dioxygenases, designated prolyl hydroxylase domain-containing enzymes PHD1 (also named HPH3, EGLN2, and Falkor), PHD2 (HPH2 and EGLN1), and PHD3 (HPH1, EGLN3, and SM-20) (13)(14)(15)(16). The K m value of the three PHDs for O 2 is slightly above its atmospheric concentration, indicating that they are effective oxygen sensors and therefore represent a link between oxygen tension and HIF stability (17). PHD1, PHD2, and PHD3 share highly conserved COOH-terminal regions responsible for hydroxylase activity but differ greatly at the N terminus (16,18,19). Each isoform displays its own tissue and cell linespecific expression pattern as well as its particular subcellular distribution (15, 17, 20 -25).
PHD3 is distinct in many ways because it can mediate diverse cellular outcomes depending on the cell type and the extracellular cues. In the hypoxic pathway it displays high substrate specificity because it exclusively hydroxylates Pro 564 , contributes to the regulation of HIF-2␣ and HIF-1␣, and is strikingly induced by hypoxia so that it may serve as a feedback loop by limiting physiological activation of HIF in hypoxia (17, 20, 21, 26 -29). Alternative splicing in various adult and fetal human tissues suggests a complex regulation, in particular because major transcripts encode catalytically inactive polypeptides on HIF-1␣ substrates. Interestingly, PHD3 is significantly shorter than either PHD1 or PHD2, and a possible additional binding partner required for full activity was discussed (30). Moreover, no rigid consensus sequence for hydroxylation of HIF-1␣ was found and a long peptide binding site with multiple interactions is likely (8,17,27). Finally, PHD3 was found as a growth factor responsive gene involved in growth regulation and/or apoptosis in smooth muscle, neuronal, and PC12 cells (23,(31)(32)(33)(34)(35)(36)(37).
These results suggest that additional cofactors, other prolyl hydroxylation substrates, and/or molecular scaffolds are required to allow multiple interactions. In the present work we identified Morg1, a WDrepeat protein, as a molecular scaffold that directly binds PHD3 in vitro and in vivo and may link PHD3 to different pathways. PHD3 and Morg1 perfectly co-localize in the cytoplasm of mammalian cells and display a similar tissue expression pattern in different tissues. Binding of Morg1 occurs within the ␤-sheet of strand b, involving residues predicted to the top of one WD-propeller blade. Finally, we show that Morg1 decreases HIF-mediated reporter gene activity and that this effect is additive by co-expression of PHD3. Vice versa, suppression of Morg1 or PHD3 by stealth RNA results in a marked increase of HIF-1 activity. Morg1 may therefore serve as a scaffold for protein interactions and represent a link between PHD3 and diverse cellular outcomes.

EXPERIMENTAL PROCEDURES
All chemicals other than indicated were purchased from Sigma, restriction and modifying enzymes from New England Biolabs, the yeast two-hybrid system vectors, primers, yeast strain, yeast plasmid purification products, cDNA library, and co-immunoprecipitation products were from Clontech. Cell culture reagents were purchased from Invitrogen.

Yeast Two-hybrid Screening
pGBKT7 was expressed as a fusion protein to a GAL4 DNA binding domain and a c-Myc epitope tag using tryptophan as selection marker. A rat brain cDNA library cloned into pACT2 was expressed as a fusion protein to the GAL4 activation domain and a HA epitope tag using leucine as selection marker. All yeast transformations were performed according to the high transformation efficiency protocol of yeast transformation system 2 (Clontech). Expression of PHD3 in yeast was confirmed by immunoblotting yeast protein extracts using monoclonal antibodies against the c-Myc epitope. Protein extracts were obtained by the urea/SDS methods according to the yeast protocol handbook (Clontech). To check for autonomous activation, PHD3-transformed yeasts were plated onto SD-agar plates lacking tryptophan and containing X-␣-galactosidase. To increase transformation efficiency both plasmids were sequentially transformed into yeast strain AH109. AH109 transformants were selected using three different reporters: adenine (ADE2), histidine (HIS3), and ␣-galactosidase (MEL1), under the control of distinct GAL4 upstream activating sequences and TATA boxes. Transformants were screened on SD-agar plates lacking adenine, histidine, leucine, tryptophan, and containing X-␣-galactosidase (high stringency selection). SD-agar plates lacking tryptophan and leucine were used to calculate the cotransformation efficiency and the number of clones screened. To eliminate false positive clones yeast plasmids were isolated using a yeast plasmid isolation KIT (Clontech), transformed into Escherichia coli, purified, retransformed into the yeast strain expressing PHD3, and streaked out twice. Positive clones from this screening were subjected to direct sequencing. Deletion constructs were screened accordingly.

Plasmid Construction
Generation of full-length rat PHD3, Morg1, deletion constructs, and the 6ϫ HIF responsive element (HRE) is described as supplemental data.

Northern Blot, RNA Extraction, 5 RACE, and 3 RACE
Northern Blots-RNA extraction and Northern blotting were performed as described (38). Morg1 and PHD3 cDNAs were labeled with [ 32 P]dATP (3000 Ci/mmol; Amersham Biosciences) using random hexamer primers. Probes were hybridized overnight at 42°C in rapid-hybridization buffer (Amersham Biosciences), washed several times in 2ϫ SSC, 0.1% SDS and 0.1ϫ SSC, 0.1 SDS and exposed to a Hyperfilm ECL (Amersham Biosciences). Morg1 cDNA was used as a probe to label a rat multiple tissue Northern blot (Clontech).
5Ј RACE-Total rat brain and testis RNA were extracted using TRI Reagent and DNase I treated (DNA-free; Ambion, Austin, TX). First strand cDNA was synthesized using primer CCCTGTGTCTT-TATCTAGAA and Superscript II reverse transcriptase (Invitrogen). The original mRNA template was removed by treatment with RNase H/T1 (Invitrogen), and unincorporated dNTPs, primer, and proteins were separated from the cDNA by Qiaquick PCR purification (Qiagen). A polymeric tail was then added to the 3Ј end of the cDNA using terminal deoxynucleotidyl transferase (Invitrogen) and dCTP. One nested round of PCR was sufficient to reveal the full 5Ј end (5Ј-GGCC-ACGCGTCGACTAGTACGGGIIGGGIIGGGIIG-3Ј (Invitrogen) and 5Ј-GGTGATGGGGCTGCCCACAT-3Ј). PCR products were cloned into pGEM T-easy and subjected to direct sequencing.

Morg1 Interacts with PHD3
ribonuclease inhibitor, T7 polymerase (Promega), [ 35 S]methionine (1,000 Ci/mmol; Amersham Biosciences), and incubated at 30°C for 90 min. 10 l of in vitro translated Morg1 and 10 l of in vitro translated PHD3 were incubated 1 h at room temperature, before 1 g of monoclonal anti-c-Myc or 1 g of polyclonal anti-HA antibody was added and incubated for 1 h at room temperature. Protein A beads were washed, added, and incubated with the reaction for another 1 h at room temperature. The reaction was washed five times, denatured, and separated by 12% SDS-PAGE. The gel was rinsed in amplify fluorographic reagent (Amersham Biosciences), dried, and exposed to a Hyperfilm ECL (Amersham Biosciences).

Co-immunoprecipitations
HEK293 cells were transfected with Morg1-HA pCEP4 and PHD3pcDNA3-T7 fusion constructs (kindly provided by Robert S. Freeman). After 48 h cells were lysed in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 140 mM KCl, 3 mM EDTA, 0.5% Triton X-100, Complete Mini Protein Inhibitor Mixture (Roche)) and clarified by centrif-ugation at 10,000 ϫ g for 10 min at 4°C. For immunoprecipitations of PHD3, 2 l of anti-T7 antibody (Molecular Probes) or 2 l of anti-PHD3 antibody (36) were incubated with 1,000 g of protein lysate overnight at 4°C. 40 l of Protein G plus-agarose (Santa Cruz Biotechnologies) was added for another 1 h at 4°C. Subsequently, the agarose-protein complexes were washed three times with lysis buffer at 10,000 ϫ g, at 4°C for 20 s and dissociated in SDS running buffer at 95°C for 5 min. Immunoprecipitations of Morg1 were performed with either 5 g of anti-HA (clone 12CA5; Roche) or 2 l of purified anti-Morg1 antibody. Immunoprecipitations with either normal rabbit serum or just Protein G plus-agarose served as controls.

Transient Transfection, Confocal Laser Scanning Microscopy, and Tryptic Cleavage
For co-localization experiments, PC12 and HEK293 cells were cultured in Lab-TekII chamber slides (Nalge Nunc Co.) and transiently transfected with the chimeric pEGFP-PHD3 (gift of Robert S. Freeman) and Morg1-HA pCEP4. After 48 h PC12 cells were attached to Cell-TAK (BD Biosciences)-treated chamber slides for 1 h. Fixation and immunostaining were performed as described (39). In brief, cells were washed with PBS and fixed with methanol for 10 min on ice. Immunostaining was performed at room temperature using the mouse monoclonal anti-HA antibody 12CA5 (Roche) with an isotype-matched control mouse (DAKO) and rabbit anti-EGFP antibodies (Molecular Probes) with nonimmune rabbit IgG as a control. Secondary antibodies were Texas Red anti-mouse IgG and fluorescein anti-rabbit IgG (DAKO). The subcellular distribution of fluorescent activity was examined by conventional fluorescence (Axioscope, Zeiss) and confocal laser scanning microscopy (LSM 510 META, Zeiss). For tryptic cleavage, HEK293 cells were transfected with Morg1-pCEP-4 using Superfect reagent. Cells were lysed in trypsin reaction buffer (50 mM Tris-HCl, 20 mM CaCl 2 , pH 8.0) and incubated with L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin (New England Biolabs) for 10 min at 30°C as described (40). The reaction was stopped by adding Laemmli sample buffer, immediately heated at 95°C for 5 min, and analyzed by 12% SDS-PAGE.

Luciferase Assays
PHD3, Morg1, 6ϫ HRE, and a normalizing pCMV-␤-galactosidase plasmid were transfected into HEK293 cells using Superfect transfection reagent (Qiagen) according to the manufacturer's protocol. An empty pCEP4 vector was transfected along with all transfections to achieve equimolar amounts of plasmids. The PGL3-promotor plasmid (Promega) was used as a control. After 8 h, 800 M L-mimosine was added as indicated. 24 h post-transfection cells were washed with PBS, lysed (25 mM glycylglycin, pH 8, 15 mM MgSO 4 , 4 mM EGTA, 1% Triton X-100, 1 mM dithiothreitol), and luciferase assays were performed as described (41,42). All transfections were performed in triplicate for each construct, and all transfection sets were repeated at least three times. Transfection results were averaged, normalized with the ␤-galactosidase results, and expressed as the means.

Production of Anti-Morg1 Antisera and Enzyme-linked Immunosorbent Assay
Two polyclonal antisera to the synthetic peptide, CRSRKPEP-VQTLDEA of Morg1 (comprising amino acids 138 -152), coupled to hemocyanin of limulus polyphemus were raised in New Zealand White rabbits. Specific IgGs were purified by affinity chromatography (Biotrend). Antibodies were characterized by immunoblotting and indirect enzyme-linked immunosorbent assay as described (43). In brief, Transfection of PHD3 with an empty prey vector revealed no intrinsic DNA binding or transcriptional activation properties, indicated by a lack of growth on high stringency agar (G and H). Co-transfection efficiencies were tested by growth on agar lacking leucine (selection marker of the cDNA library) and tryptophan (selection marker of PHD3) (B, D, F, and H). PHD3 was properly expressed by yeast (I1 and I3) and showed no autonomous activation as indicated by white colonies on galactosidase agar lacking tryptophan. MARCH 31, 2006 • VOLUME 281 • NUMBER 13 96-well plates were coated with 100 ng/well peptide in 0.1 M NaHCO 3 , pH 8.2, overnight, washed with PBS-T, blocked with 5% goat serum in PBS-T, 2 h at room temperature, washed with PBS-T, and incubated with the appropriate concentration of sera (diluted in 5% goat serum) for 2 h at 37°C. After four additional wash steps plates were incubated with an alkaline phosphatase-labeled antibody (Southern Biotechnology, dilution 1:2500, 1 h at room temperature). Detection was performed with p-nitrophenyl phosphate for 30 min at 37°C.

Stealth RNA Gene-silencing Assay and Real-time PCR Analyses
A stealth RNA targeting the 25-nucleotide sequence, ACUCUACU-CUGAGGCUUCUAGAUAA of Morg1, the Morg1 control stealth RNA, ACUCUCAGAGUUUCGGAUCAUCUAA, a stealth RNA targeting the PHD3 gene, TGGAGCGAGTCAAGCAGCTGCATTA, and the corresponding PHD3 control RNA, TGGGAGCAACTGACG-CGTCAGATTA, were obtained from Invitrogen. 100 nM stealth RNA was transfected with Lipofectamine (Invitrogen), after 48 h RNA was isolated using the RNeasy Mini Kit (Qiagen). cDNAs were synthesized using 200 ng of RNA, random hexamers, and Moloney murine leukemia virus reverse transcriptase (Invitrogen). Real-time PCRs were performed as described (44)

RESULTS
Identification of Morg1 as PHD3-associated Protein-We used the yeast two-hybrid technique to identify proteins that interact with PHD3. We took advantage of three different reporter genes to eliminate false positive clones: adenine, histidine, and ␣-galactosidase. A PHD3 cDNA was taken as bait to screen a rat brain cDNA library. From 2.97 ϫ 10 7 library transformants, 54 clones were obtained that restored growth of AH109 yeast on high stringency selection agar in a PHD3-dependent manner. Three positive clones encoded the carboxyl-terminal half of Morg1 (Fig. 1, A and B), a WD-repeat protein recently described as a component of the ERK cascade (45). To confirm this interaction, the full-length Morg1 coding sequence was cloned by PCR and used as prey taking PHD3 as bait. Full-length Morg1 strongly interacted with PHD3 ( Fig. 1, C and D), indicating the specificity of the PHD3-Morg1 interaction. Interaction of p53 with SV40 large T-antigen served as positive controls (Fig. 1, E and F). Transfection of PHD3 with an empty prey vector revealed no intrinsic DNA binding or transcriptional activation properties, indicated by a lack of growth on high stringency agar (Fig. 1,  G and H). Co-transfection efficiencies were tested by growth of AH109 yeast on agar lacking leucine (selection marker of the cDNA library) and tryptophan (selection marker of PHD3) (Fig. 1, B, D, F, and H). PHD3 was properly expressed by yeast (Fig. 1I) and showed no autonomous activation as indicated by white colonies on galactosidase agar lacking tryptophan.
Sequence Analysis, Identification of the 3Ј and 5Ј Ends of Morg1, and Tissue Expression-Sequence alignments revealed that the identified clone of 715 bp comprised four WD-repeat motives of Morg1 in contrast to seven repeats in mouse and human homologues. 3 A rat multiple tissue Northern blot was next hybridized with the cDNA as a probe revealing that rat Morg1 consists of an ϳ1200 bp cDNA. Expression was strong in testis and brain, moderate in heart, liver, and kidney, weak in spleen and lung, and absent in muscle ( Fig. 2A). Rat brain and testis RNA was probed for the 5Ј sequence of rat Morg1. 5Ј RACE yielded a 612-bp product that was highly homologous to the 5Ј sequence of murine Morg1 (Fig. 2B). 3Ј RACE using brain and testis RNA as templates revealed a 274-bp product (Fig. 2C). Taken together, rat Morg1 comprises a cDNA of 1137 bp (Fig. 3). The full-length rat clone was constructed by ligating a 5Ј PCR product to the 3Ј end of rat Morg1. Full-length rat Morg1 is 92/87% identical at the nucleotide level and 97/93% identical at the amino acid level to murine Morg1 and human MORG1, respectively. The ATG (nucleotides 37-39) is likely to be the start site for translation for several reasons: (a) the presence of three TAA upstream terminator codons, (b) the first ATG codon lies in a strong Kozak context (A Ϫ3 and G ϩ4 ), and (c) translation usually initiates uniquely at the first ATG codon in an adequate context. The second ATG (nucleotides 580 -583) (with an equally good context) unlikely serves as the initiation site simply because it is internal and thus inaccessible to the scanning 40 S ribosomal subunit, which advances from the 5Ј end (46). Therefore, the rat Morg1 cDNA predicts a protein of 316 amino acids and a molecular mass of 34,380 daltons (GenBank accession number AY940050). Seven WD-repeat domains span the entire length of the sequence with neither a putative signal peptide nor a predicted mitochondrial targeting signal (Fig. 3).
PHD3 and Morg1 Co-localize in the Cytoplasm-cDNAs of PHD3 and Morg1 were next utilized to probe RNAs of different cell lines to identify cells expressing both proteins. Northern blots revealed a strong expression of PHD3 and Morg1 in PC12 cells and a weak expression of 3 U. Hopfer and G. Wolf, unpublished observation. Expression is strong in testis and brain, moderate in heart, liver, and kidney, weak in spleen and lung, and absent in muscle. Rat brain and testis RNA was probed for the 5Ј and 3Ј sequence of Morg1. B, 5Ј RACE yielded a 612-bp product that was highly homologous to the 5Ј sequence of murine Morg1. C, 3Ј RACE revealed a 274-bp product. both proteins in HEK293 cells (Fig. 4A). Therefore, PC12 and HEK293 cells were used for in vivo interaction assays. Additionally, HEK293 cells were chosen because preparations of PHD3 in HEK293 cells were proven to be catalytically active (47).
To examine the subcellular distribution of PHD3 and Morg1, HEK293 and PC12 cells were transiently transfected with HA-tagged Morg1 and a PHD3-GFP fusion protein, stained with anti-HA/antimouse Texas Red and anti-GFP/anti-rabbit fluorescein antibodies, and examined by confocal microscopy. PHD3 occurred in a punctuate pattern throughout the cytoplasm and within the nucleus of PC12 cells (Fig. 4B) consistent with recent observations (33,36,48). Immunostaining of Morg1 showed a punctuate staining pattern predominantly within the cytoplasm and to a lesser extend within the nucleus of PC12 cells. Neighboring untransfected cells or cells transfected with the simple expression vector did not show any immunofluorescence. In double staining experiments Morg1 co-localized with PHD3. In HEK293 cells both PHD3-GFP and Morg1 localized primarily in the cytoplasm and less in the nucleus. Morg1 immunofluorescence co-localized perfectly with PHD3 (Fig. 4B). It has been proposed that properly folded WDrepeat proteins should be largely resistant to trypsin cleavage (40). To check whether high expression of Morg1 leads to proper folding in HEK293 cells, cell lysates were incubated with 2 M trypsin as described (40). Despite 32 potential trypsin cleavage sites Morg1 showed resistance to proteolysis (Fig. 4C).
Morg1 Interacts with PHD3 in Vitro and in Vivo-To further establish the interaction between PHD3 and Morg1 we carried out immunoprecipitation experiments with in vitro transcribed and translated cMyc-PHD3 and HA-Morg1 fusion proteins. As shown in Fig. 5A, the presence of ϳ40and ϳ34-kDa proteins indicate that anti-cMyc-PHD3 specifically precipitated Morg1, which is consistent with an immunoprecipitation of anti-HA-Morg1 precipitating PHD3. Anti-cMyc antibodies did not bind to Morg1 nor bind anti-HA to PHD3. HA-SV-40 large T and cMyc-p53 fusion proteins served as co-immunoprecipitation controls (Fig. 5B).
To explore whether the interaction between PHD3 and Morg1 detected in yeast and in in vitro translated products occurs in intact mammalian cells, immunoprecipitation experiments were performed in HEK293 cell lysates co-transfected with PHD3 and Morg1. Immunoprecipitation of either anti-Morg1 (Fig. 6, A and B), anti-PHD3 (Fig. 6C), or anti-T7 (protein tag of PHD3, Fig. 6D) led to recovery of both proteins in HEK293 cells, supporting an in vivo interaction between PHD3 and Morg1. Immunoprecipitations of co-transfected cell lysates with either preimmune rabbit serum or Protein G plus-agarose did not lead to recovery of either protein. Specific Morg1 antisera used in immunoprecipitation experiments were tested by immunoblotting lysates from HEK293 cells transfected with full-length Morg1 cDNA. Both antisera recognized Morg1 at a mobility consistent with the predicted molecular mass of 34 kDa (Fig. 6, E and F). Immunoblotting of lysates using an anti-HA antibody served as control. An indirect enzyme-linked immunosorbent assay against the immunogen was used to estimate the sensitivity of both antibodies and revealed specific binding of the crude sera and purified IgGs above the 128-fold dilution and no binding of preimmune sera (Fig. 6G).    (F2 and G2). G, enzyme-linked immunosorbent assay against the immunogen was used to estimate the sensitivity of both antibodies and revealed specific binding of crude sera and purified IgGs above dilutions of 1:128.
Characterization of the Binding Region of Morg1-The Morg1/PHD3 interaction was explored further using truncated Morg1 and full-length PHD3 constructs in yeast two-hybrid assays. PHD3 and all truncated constructs were sufficiently transfected in yeast as represented by growth on agar plates lacking leucine and tryptophan (Fig. 7, AA-JJ). Strong interaction was accomplished by expression of all seven WDrepeats (Fig. 7A). Truncation of WD5 to WD7 did not change the interaction (Fig. 7, B-E). However, yeast failed to grow by truncation of ⌬aa 167-318 (Fig. 7F), indicating that this variant was not able to interact efficiently with PHD3 anymore. Remarkably, truncations of ⌬aa 172-318 (Fig. 7G), ⌬aa 177-318 (Fig. 7H), ⌬aa 181-318 (Fig. 7I), or the truncation of ⌬aa 1-128 (Fig. 7J) led to growth of yeast, suggesting that the interaction at least requires aa 166 and 167. Given the structure of one blade of the propeller of a typical WD-repeat protein, Leu 167 and Ala 168 represent highly conserved residues of WD-repeats within the ␤-sheet of strand b. In a three-dimensional structure of a typical WDrepeat protein they would be predicted on the top surface of the protein.
Taken together, given a supramolecular structure of Morg1 as in typical WD-repeat proteins, e.g. G␤ subunits of heterotrimeric G proteins, binding of Morg1 to PHD3 occurs at the beginning of a ␤-sheet of strand b and appears to be closely coordinated over the center of the top tunnel opening. Additionally, the binding region within PHD3 was further confined. Deletion of the conserved jelly roll core of PHD3 did not hamper the interaction with Morg1 (Fig. 7, K and L), indicating that binding most likely occurs within the unique NH 2 -terminal tail.
Morg1 Decreases HIF-mediated Reporter Gene Activity-PHD3 was reported to suppress HRE-mediated reporter gene activity when overexpressed in cells (8,28). To test whether Morg1 influences HIF-mediated reporter gene expression we fused six copies of the HRE of the EPO gene to a SV40 promotor and a luciferase gene. Transfection of this construct into PC12 (Fig. 8A) or HEK293 cells (Fig. 8B) significantly induced luciferase activity compared with the empty SV40 promotor/ luciferase gene (n ϭ 6, p Ͻ 0.05) indicating sufficient HIF activity in these cells. Inhibition of PHDs by L-mimosine as described (49) superinduced the reporter gene activity in both cell lines (p Ͻ 0.01). In conformity to a high PHD3 expression (Fig. 4) this further implicates PHD3 activity in these cells. As reported previously, overexpression of PHD3 decreased HRE-mediated luciferase activity. Luciferase activity was also decreased when Morg1 was co-expressed with the reporter gene, suggesting an interaction of internal PHD3 and Morg1. Co-expression of PHD3 and Morg1 reduced reporter gene response to almost basal levels (p Ͻ 0.05). These results suggest that Morg1 promotes PHD3 action on HIF-mediated reporter gene activity.
To assess whether Morg1 indeed affects HIF-mediated reporter gene activity and to show that PHD3 is truly active under this condition we knocked down both mRNAs with stealth RNA oligonucleotides. We either targeted mock plasmid-transfected PC12 cells (for endogenous protein) or Morg1-or PHD3-transfected PC12 cells (for exogenous protein) with stealth RNAs and determined the amount of Morg1 and PHD3 mRNAs by quantitative real-time PCR. Morg1 was silenced by Ϸ85% in mock-transfected cells and Ϸ90% in cells overexpressing Morg1. A control stealth RNA did not change Morg1 gene expression (expression Ϯ10%). PHD3 was silenced by Ϸ50% in mock and Ϸ60% in PHD3-transfected cells. No significant difference was detected by transfection of the control stealth RNA (expression Ϯ10%).
Overexpression of PHD3 Has No Impact on the Status of pERK1 or pERK2 in HEK293 Cells-Morg1 facilitates ERK activation after stimulation with fetal bovine serum (45). To determine the impact of PHD3 upon ERK phosphorylation, 293 cells were transfected with Morg1, PHD3, or both and stimulated with 10% fetal bovine serum. Western blots probed with anti-ppERK1/2 (Fig. 8E) showed that overexpression of neither PHD3 (lane 3) nor PHD3 and Morg1 (lane 5) notably changed the status of pERK1/2 phosphorylation when compared with mock transfected cells (lane 7). Anti-p-ERK1/2 was used as loading control. To control for transfection efficiency a second blot was probed against anti-T7-PHD3, anti-HA-Morg1, and ␤-actin.

DISCUSSION
Regulated protein degradation is a key event in many cellular processes. Ubiquitin-dependent degradation uses covalent attachment of a polyubiquitin chain on lysine residues of the substrate, which mediates its recognition and subsequent degradation by the 26 S proteasome.
How does PHD3 act in these divergent pathways? Substrates other than HIF-1␣ for hydroxylation in various pathways may be involved, however, a WD-repeat protein as a molecular scaffold for different interaction partners or other substrates provide a good explanation. We identified Morg1 as an interaction partner of PHD3 by yeast two-hybrid assays and described seven WD-repeat domains by cloning the fulllength cDNA. Morg1 displays in part a similar tissue expression pattern to PHD3 that is expressed in brain, heart, skeletal muscle, kidney, liver, spleen, and lung (56).
The defining feature of proteins in the WD-repeat family is the presence of 4 -8 repeating units containing a conserved core of 27-45 amino acids that are bracketed by two characteristic dipeptide sequences, GH and WD, but neither the GH nor the WD dipeptide is absolutely conserved (57)(58)(59). The crystal structure of one WD-repeat protein, the signal transducing G protein ␤ subunit of heterotrimeric G proteins, shows that the repeating units form a circular, propeller-like structure with each blade made up of four ␤ strands (60,61). If there is a common functional theme, it appears to be that the WD-repeat propeller structures create a stable platform that can reversibly form complexes with several proteins, thus coordinating sequential and/or simultaneous interactions involving several sets of proteins. G␤, for example, interacts tightly with G␥ and, simultaneously, interacts with one of Ͼ15 different proteins (59,62). Given that Morg1 possess the three-dimensional structure seen with other members of the WD40 protein family, it is likely that PHD3 interacts with multiple partners. It was suggested that, as with non-WD-repeat propellers, the top surface of WD-repeat proteins, including the central tunnel opening, coordinates interaction with other proteins and/or small ligands (60,61,63,64). We were able to define short sequences within Morg1, which were sufficient for recognition by PHD3. Interestingly, these sequences correspond to a region of striking conservation and localize to the top surface of the protein near the center of the top tunnel opening. WD-repeat ␤-propeller domains may be fused to domains that are predicted to have enzymatic activity, e.g. kinase activity (65,66), but no catalytic activity has been attributed to the propeller portion of a WD-repeat protein (58). As in G␤ the repeat of Morg1 spans the entire length of the sequence, to this end catalytic activities seem unlikely. PHD1, -2, and -3 share a conserved COOH-terminal region responsible for hydroxylase activity but differ greatly at the NH 2 terminus (18). Interestingly, deletion of the conserved COOH terminus of PHD3 did significantly induced luciferase activity (n ϭ 6, p Ͻ 0.05) compared with the empty SV40 promotor/luciferase gene (designated 100%), indicating sufficient HIF activity in these cells. Inhibition of PHDs by L-mimosine (L-Mim) superinduced reporter gene activity in both cell lines (p Ͻ 0.01), which implicates PHD activity in these cells. Expression of PHD3 decreased HRE-mediated luciferase activity. Luciferase activity was also decreased when Morg1 was co-expressed with the reporter gene, suggesting an interaction of internal PHD3 and Morg1. Co-expression of PHD3 and Morg1 reduced reporter gene response to almost basal levels (p Ͻ 0.05). C, knock-down of endogenous Morg1 noticeably induced reporter gene activity in PC12 cells. Luciferase activity was also superinduced by silencing exogenous Morg1 alone or co-expressed with PHD3. D, by silencing either endogenous PHD3, exogenous PHD3 or PHD3 cotransfected with Morg1, luciferase activity increased markedly in PC12 cells. E, to determine the impact of PHD3 upon ERK phosphorylation, cells were transfected with Morg1 (lane 1 and 2), PHD3 (lanes 3 and 4), PHD3 and Morg1 (lanes 5 and 6), mock plasmid (lanes 7 and 8), or left untreated (lanes 9 and 10). Cells were stimulated with 10% serum (lanes 1, 3, 5, 7, and 9). Western blots probed with anti-ppERK1/2 showed that overexpression of neither PHD3 nor PHD3 and Morg1 notably changed the status of pERK1/2 phosphorylation when compared with mock transfected cells. Anti-pERK1/2 was used as loading control. To control for transfection efficiency a second blot was probed against anti-T7-PHD3, anti-HA-Morg1, and ␤-actin. not reverse the interaction with Morg1. Therefore, an interaction between PHD1 or PHD2 and Morg1 seems unlikely.
A mitochondrial targeting sequence of rat PHD3 has been reported (33,36). Sequence predictions and subcellular fractions enriched for mitochondria could not localize Morg1 to the mitochondria. Indeed, the mitochondrial targeting sequence of rat PHD3 is not conserved among other mammalian species, and mutants lacking the mitochondrial targeting sequence retain the ability to induce caspase-mediated apoptosis (18,33). These results indicate that mitochondrial targeting is not a general feature of PHD3. Furthermore, the current data provide evidence that Morg1 co-localizes with PHD3 within the cytoplasm and the nucleus (8,48). Nuclear-cytoplasmic trafficking of HIF-1␣ is required for efficient oxygen-dependent degradation of HIF-1␣. Nuclear proxylation and interaction with von Hippel-Lindau protein is not only required for ubiquitination but also for efficient nuclear export of HIF-1␣ in reoxygenated cells (67). To what extend Morg1 contributes to this nuclear-cytoplasmic shuttling remains to be elucidated.
It is well established that expression of PHD3 leads to decreased HIFdependent reporter gene activity. Our results indicate that expression of Morg1 supports the specific activity of PHD3 in vivo. PHD3 may require Morg1 for maximal activity under certain circumstances. For example, PHD3 enzyme assays performed from different sources revealed contrary catalytic activities of PHD3 compared with PHD2 (8,47). Docking and/or recognition sites distant from the hydroxyl acceptor residue and from the catalytic site of the hydroxylase were considered to govern in vivo substrate specificity within the PHD family members (8,27). Our findings unveil a potential adapter protein for substrate recognition.
Finally, PHD3 itself is regulated by proteasomal degradation through Siah2, which belongs to the RING domain class of E3 ubiquitin ligases (68 -70). Whereas the precise architecture of Siah complexes remains to be worked out, it is well established that RING domain E3 ubiquitin ligase complexes interact with a set of adapter proteins that recruit different binding partners through specific protein-protein interaction domains such as WD40-repeats and leucine-rich repeats (71,72).
During the course of this work a role of Morg1 as molecular scaffold for components of the ERK cascade was described (45). Whether ERK and PHD3 interactions represent connected or unconnected functions of Morg1 is not yet known. Our initial experiments using overexpression of PHD3 and Morg1 in HEK293 cells did not show differences in the status of ERK activation, but clearly more detailed studies are needed. HIF-1␣ is strongly phosphorylated by ERK1/2 and this activation is sufficient to promote the transcriptional activity of HIF-1 (73,74). However, another WD-repeat protein (CPF/Set1) was found as a component of very different complexes: CPF is a cleavage and polyadenylation factor, Set1 methylase modifies lysine 4 of histone H3 (75).
In conclusion, this study introduces a novel WD-repeat protein that interacts with PHD3 and most likely promotes degradation of HIF-1␣ through PHD3. Considering that WD-repeat proteins act as molecular scaffolds for multiple protein complexes, the next challenge will be to identify further interaction partners of Morg1 in the HIF-1␣ degradation pathway, PHD3-dependent apoptosis, or growth control.