UCR1 and UCR2 Domains Unique to the cAMP-specific Phosphodiesterase Family Form a Discrete Module via Electrostatic Interactions*

The cAMP-specific phosphodiesterases (PDE4) enzymes contain unique “signature” regions of amino acid sequence, called upstream conserved regions 1 and 2 (UCR1 and UCR2). UCR1 and UCR2 are located between the extreme amino-terminal region and the catalytic region of the PDE4 enzymes. The UCR1 of the PDE4D3 isoform was used as a “bait” in a two-hybrid screen, which identified a PDE4D cDNA clone containing UCR2 and the catalytic region but not UCR1. Two-hybrid and “pull down” analysis of constructs incorporating various regions of the PDE4D3 cDNA demonstrated that the carboxyl-terminal region of UCR1 interacted specifically with the amino-terminal region of UCR2. The interaction was blocked by mutations of two positively charged amino acids (Arg-98 and Arg-101 to alanine) located within an otherwise largely hydrophobic region of UCR1. Mutation of three negatively charged amino acids in UCR2 (Glu-146, Glu-147, and Asp-149, all to alanine) also blocked the interaction. The phosphorylation of UCR1 by cAMP-dependent protein kinase (PKA) in vitro attenuated the ability of UCR1 to interact

Modulation of the levels of the second messengers cAMP and cGMP by cyclic nucleotide PDEs 1 plays an important role in the regulation of numerous physiologic processes, including those in the immune/inflammatory systems, vascular smooth muscle, and the brain. The cAMP-specific phosphodiesterases (PDE4s) are members of a large family of cyclic nucleotide phosphodiesterases (1). PDE4 enzymes can be differentiated from other PDEs by sequence homology in the catalytic regions of the enzymes (2) and by their ability to be specifically inhibited by the drug rolipram. Rolipram and other specific PDE4 inhibitors have been shown to have anti-depressant, anti-inflammatory, and smooth muscle relaxant activity in humans (2). The mammalian PDE4s show strong evolutionary conservation to the dunce gene of Drosophila melanogaster, which was first isolated as a mutation affecting learning and memory in that organism (3,4). In mammals, the PDE4s are comprised of a large family of isoforms, encoded by four different genes (PDE4A, PDE4B, PDE4C, and PDE4D), with additional diversity being generated by alternative mRNA splicing (2).
We have demonstrated previously (2,5) that the PDE4 enzymes are uniquely characterized by two regions of amino acid sequence, called upstream conserved regions 1 and 2 (UCR1 and UCR2, respectively). UCR1 and UCR2 are located between the extreme amino-terminal regions of the proteins and their catalytic regions (Fig. 1). UCR1 and UCR2 appear to be distinct, in that they lack homology to each other and are separated by a region of relatively low homology (2,5). Significantly, UCR1 and UCR2 show strong evolutionary conservation throughout mammalian PDE4s (2) and are also conserved in PDE4 homologs in organisms as distantly related as D. melanogaster (4) and Caenorhabditis elegans (6). This strong evolutionary sequence conservation suggests that UCR1 and UCR2 are functionally important, as sequence motifs that are strongly conserved in evolution are often of functional significance. The function(s) of UCR1 and UCR2 are not known. However, a phosphorylation site for the cAMP-dependent protein kinase (PKA) is located at the beginning of UCR1 (Ser-54 in PDE4D3, Fig. 1) and phosphorylation of the PDE4D3 isoform activates the enzyme and changes its ability to be inhibited by rolipram (7)(8)(9)(10).
In this study, we demonstrate, by two independent approaches, that UCR1 and UCR2 interact directly with each other. We also show that the interaction involves discrete regions of UCR1 and of UCR2 and that it is dependent on specific charged amino acids within these regions. We also show that this interaction is disrupted by the phosphorylation of UCR1 by PKA and by mutation of Ser-54 in UCR1 to aspartic acid, which mimics the stoichiometrically PKA-phosphorylated form of PDE4D3 (9,10). We propose a model in which UCR1 and UCR2 form an interacting module that serves to regulate the catalytic region of the enzyme.

EXPERIMENTAL PROCEDURES
Materials-A HeLa cell (HeLa S3 cells; American Type Culture Collection) two-hybrid library cloned into the EcoRI and XhoI sites of the pGADGH vector (11) was obtained from David Beach (Cold Spring Harbor Laboratory). This vector expresses proteins as fusions with the activation domain of the Saccharomyces cerevisiae GAL4 protein. A monoclonal antibody to human PDE4D proteins, which does not crossreact with other PDE4 species and which we have described previously (12), was a gift from Sharon Wolda, ICOS Corp. This antibody was generated against the extreme carboxyl-terminal region of the PDE4D3 protein and can detect PDE4D3 isoforms truncated at the amino terminus, such as those lacking UCR1 and UCR2 (12).
Two-hybrid Screens-These were performed using methods we have described previously (13,14). In brief, various regions of the pPDE43 cDNA encoding human PDE4D3 ( Fig. 1; GenBank TM accession number L20970 (5)) were cloned into the NotI site of pLEXAN, to produce constructs encoding fusions between the PDE and the DNA-binding domain of the Escherichia coli LexA protein. These constructs were prepared by the addition of NotI sites to the cDNA regions by the use of PCR, as described previously (5). Screens were performed with the HeLa two-hybrid cDNA library in the S. cerevisiae strain L40 (15). To screen the library, positive clones were initially selected for growth in the absence of histidine (in the presence of 1 mM 3-aminotriazole) and then patched to plates selecting only for the two plasmids and assayed for lacZ activity using a filter ␤-galactosidase assay, as described (13). Library plasmid DNA was then isolated from the positives and reassayed for interaction with the LexA fusion, using methods described previously (13). For additional two-hybrid experiments, various pieces of the PDE4D3 cDNA were cloned into the NotI site of pGADN, using PCR as described for the LexA fusions.
Quantitative ␤-Galactosidase Assays of Two-hybrid Interactions-These were performed using the whole-cell assay of Guarente (16). The activity units were defined as follows: ((OD 420(reaction) Ϫ OD 660(reaction) ) (10,000)) Ϭ ((time in min)(OD 660(culture) )), where OD 420(reaction) and OD 660(reaction) are the optical densities (at the indicated wavelengths, in nm) of the final ␤-galactosidase reaction and OD 660(culture) is the optical density at 660 nm of the cultures at the time of harvest.
Generation of Bacterial Expression Constructs-Various pieces of PDE4D3 cDNA were cloned into the NotI site of pMALN, using PCR as described above for the LexA fusions. pMALN is a derivative of pMALC2 (New England Biolabs (17)), with a NotI site inserted into the polylinker. These clones generate fusions between maltose-binding protein (MBP) and the amino terminus of the protein encoded by the insert. Specifically, pMALU1DTR encodes the fusion protein MBP-UCR1-C, consisting of amino acids 80 -116 of PDE4D3. pMAL43U1 encodes MBP-UCR1, consisting of amino acids 17-136 of PDE4D3.
Generation of COS7 Cell Expression Constructs-Various pieces of the PDE4D3 cDNA were cloned into the NotI site of the vector pcDNA3 (Invitrogen). In these constructs, the insert is placed under the control of the cytomegalovirus intermediate early gene promoter. pcDNA43VSVF encodes the full open reading frame of PDE4D3. pcDNAFN43N2VSVF encodes UCR2 and the catalytic regions of PDE4D3 (i.e. UCR2ϩCat). pcDNAFN43N3VSVF encodes only the catalytic region of PDE4D3 (i.e. Cat). In all cases, a sequence corresponding to the vesicular stomatitis virus (VSV) glycoprotein epitope (18) was added immediately downstream from the last native codon of the PDE to encode a carboxyl-terminal fusion. The native PDE4D stop codon was removed in this process, but a synthetic stop codon was placed immediately downstream from the epitope sequence, as described (10,14).
A portion of the PDE4D3 cDNA corresponding to UCR2 (amino acids 134 -212) was cloned into pEBGN, to produce pEBGNUCR2. pEBGN is a derivative of pEBG2 (19), but with a NotI site in the polylinker. It encodes a fusion between glutathione S-transferase (GST) and the amino terminus of the protein encoded by the insert. The expression of the fusion protein in mammalian cells is driven by the polypeptide chain elongation factor 1␣ (EF-1␣) promoter.
Generation of cDNAs Encoding Mutant Forms of PDE4D3-Portions of the PDE4D3 cDNA insert cloned into the various vectors described above were subjected to site-directed mutagenesis with the QuikChange site-directed mutagenesis kit (Stratagene).
Verification of Two-hybrid, Expression, and Mutagenesis Constructs-All mutant or PCR-generated constructs were verified by sequencing prior to use.
Expression and Purification of MBP Fusions in E. coli-E. coli JM109 transformed with pMALU1DTR, pMAL43U1, or pMALC2 (described above) were grown overnight in LB medium (20) supplemented with 100 g/ml ampicillin and 2 mg/ml glucose. These cultures were FIG. 1. Structure of UCR1 and UCR2. A, a schematic diagram of the five known human PDE4D isoforms (12). The numbers 1-5 represent isoforms PDE4D1 through PDE4D5, respectively. The heavy bar indicates sequences homologous to those in other PDE4 isoforms, with the strongest regions of conservation (the catalytic region and UCR1 and UCR2) indicated by the cross-hatched areas. The thin, branched lines adjacent to the numbers indicate the extreme amino-terminal sequence regions that are unique to each isoform. The thin lines merge where the sequences of the various isoforms join the shared sequence. The PDE4D2 isoform begins in the middle of UCR2, at a methionine that is internal to the other four isoforms. B, an alignment of the amino acid sequences of the common amino-terminal regions of the human PDE4 isoforms (2,5,30). The GenBank TM accession numbers for the sequences are as follows: PDE4A, L20965; PDE4B, L20966; PDE4C, U66346; and PDE4D, L20970. The alignment begins at the point where the PDE4D3, PDE4D4, and PDE4D5 isoforms converge and extends to the beginning of the catalytic region. Also included in the alignment is the corresponding region of the proteins encoded by the D. melanogaster dunce gene ((4) GenBank TM accession number X55247). The numbering of the alignment corresponds to the amino acid sequence of PDE4D3 (5). Amino acid sequences included in UCR1, LR1, UCR2, and LR2 are underlined. LR1 and LR2 refer to linker regions 1 and 2, which connect UCR1 with UCR2, and UCR2 with the catalytic region, respectively. Amino acids subjected to site-directed mutagenesis are indicated in bold type and also with small arrows. The heavy arrows correspond to U1E (i.e. the region encoded by pLEXAU1E) and UCR2-N (i.e. the region encoded by pGADNU2J and other UCR2-N constructs). then used to inoculate 400-ml cultures of the same medium, which were grown at 37°C with agitation until the OD 600 was 0.6 to 1.0. Isopropyl-␤-D-thiogalactopyranoside was added to a final concentration of 0.3 mM, and the cultures were grown for an additional 4 -6 h at 30°C. The bacteria were collected by centrifugation at 2500 ϫ g for 5 min at 4°C and then resuspended in 20 ml of KHEM buffer (50 mM KCl, 50 mM HEPES-KOH, pH 7.2, 10 mM EDTA, 1.92 mM MgCl 2 ) containing 1 mM dithiothreitol (DTT) and complete protease inhibitor mixture (Roche Molecular Biochemicals). The resuspended bacteria were stored in 10-ml aliquots at Ϫ20°C until needed.
The aliquots were thawed at room temperature and held on ice. They were sonicated in 20-s pulses, separated by 30-s intervals, until cell lysis was complete. The sonicate was centrifuged at 9,000 ϫ g for 30 min at 4°C. The supernatant was then incubated end-over-end for 2-4 h at 4°C with 100 l (bed volume) amylose resin (New England Biolabs) equilibrated in KHEM buffer/DTT/protease inhibitor mixture. The beads were collected by centrifugation at 2500 ϫ g for 5 s at room temperature, held on ice, and then washed three times, each time with 1 ml of KHEM/DTT/protease inhibitor mixture. The purified protein was then eluted from the beads by three incubations, end-over-end, for 10 min at 4°C with 250 l of KHEM buffer plus 10 mM maltose. The eluted fractions were pooled and then dialyzed for at least 6 h at 4°C against three changes of dialysis buffer (20 mM Tris-Cl, pH 7.6, 50 mM NaCl; 2 liters total volume). The fusion protein was assayed for protein concentration and stored at Ϫ80°C.
MBP Fusion Pull Down Assays-COS7 cells were transfected with 10 g/78-cm 2 dish area of pcDNA43VSVF, pcDNAFN43N2VSVF, pcDNAFN43N3VSVF. Growth of COS7 cells, transient transfections, and harvesting, disruption, and fractionation of COS7 cells were performed as described in detail by us previously (10,12,14,21). 50 g of MBP or MBP-UCR1 was incubated with sufficient crude COS7 cytosolic extract to contain 18 -20 units of PDE enzyme activity (pmol of cAMP hydrolyzed/min, with 1 M cAMP as substrate) in a total volume of 350 l for 1 h, end-over-end, at 4°C. 50 l (bed volume) of amylose resin, equilibrated in KHEM/DTT/protease inhibitor was then added, and the incubation was continued for 1 h more. The resin was then collected by centrifugation 2500 ϫ g for 5 s at room temperature, and the supernatant was retained as the unbound fraction. The pelleted resin was held on ice and washed three times over 15 min, each time with 350 l of KHEM/DTT/protease inhibitor mixture. These washes were pooled along with the unbound fraction, and aliquots were taken for PDE assay (see the appropriate figure legends) and immunoblotting (described below). The bound PDE was eluted from the beads by three incubations, end-over-end, for 10 min at 4°C with 100 l of KHEM buffer plus 10 mM maltose. The three eluted fractions were pooled, and aliquots were taken for PDE assay and immunoblotting.
GST Fusion Pull Down Assays-COS7 cells were transfected with 10 g of pEBGN or pEBGNUCR2 (to express GST or GST-UCR2, respectively). Pull down assays were performed to test whether immobilized MBP fusions could pull down free GST-UCR2 and also whether immobilized GST-UCR2 could pull down free MBP-UCR1. Assays using immobilized MBP or MBP-UCR1 were performed as described above for the PDE4D3 truncations, with the following changes: (i) 30 g of crude cytosolic extract from transfected COS7 cells was incubated with 50 g of MBP fusion and (ii) the detection of the bound GST fusion was by immunoblotting with an antibody to GST (Amersham Pharmacia Biotech). For assays using immobilized GST or GST-UCR2, pull down assays were performed as described above for the PDE4D3 truncations but with the following changes: (i) 200 g of crude cytosolic extract from transfected COS7 was incubated with 0.25 g of MBP or MBP-UCR1; (ii) the resin used to precipitate the GST fusion was glutathione-Sepharose 4B (Amersham Pharmacia Biotech); (iii) the elution buffer was 10 mM glutathione, 50 mM Tris-HCl, pH 8.0; and (iv) the detection of the PDE was by immunoblotting with an antibody to MBP (New England Biolabs).
SDS-Polyacrylamide Gel Electrophoresis and Immunoblotting-These were performed using standard techniques (22), modified as we have described previously (14). All quantitative immunoblotting was performed using the PDE4D monoclonal antibody (see "Materials"). The open boxes refer to UCR1, UCR2, or the catalytic regions of the protein, as indicated. The wavy lines refer to regions of amino acids that separate UCR1, UCR2, and the catalytic region. The cross-hatched box (Tag) represents the GAL4 activation domain or, in the case of U1D and U1E, the LexA DNAbinding domain in the two-hybrid constructs pLEXAU1D and pLEXAU1E, respectively. b, various regions of PDE4D3, as described in a and listed to the right of the figure, were cloned into pGADN to encode fusions with the trans-activation domain of the GAL4 protein. They were then tested for interaction with amino acids 80 -109 of UCR1, expressed as the LexA DNA-binding domain fusion pLEXAU1E (indicated as UCR1-C for this figure only). pLEXAN or pGADN without inserts (vector) were used as controls. F.L. indicates full-length PDE4D3. A filter ␤-galactosidase two-hybrid assay was used as described (13). The two patches at the right serve as internal positive and negative controls, respectively (the oncoproteins Ras Val12 and Raf1 (15) and the vectors without inserts). The ␤-galactosidase reactions were deliberately overdeveloped to demonstrate the lack of interaction of isolated UCR1 with full-length PDE4D3. This produced a few weak false-positive signals (i.e. the negative controls are weakly blue). figures. The analyses were set up so that approximately 35% of the UCR2ϩCat protein was bound to MBP-UCR1. This allowed us to detect readily any potential increase or decrease in binding. In these experiments multiple lanes, each with different amounts of protein, were run to ensure that quantitation was performed over a linear range, as we have described in detail previously (12,23).
The quantitation of immunoreactive protein was verified by three different procedures. For two of these, SDS-PAGE gels were run with a range of different protein amounts/lane and then subjected to immunoblotting. For routine analysis, immunodetection of the protein was performed by the ECL method (Amersham Pharmacia Biotech). The films were scanned and optical density plotted against sample protein content. Reference lanes were run with various amounts of PDE4D species to calibrate the linear portion of the range. As a confirmation, immunodetection was also performed using an 125 I-labeled secondary antibody. As a third method, the protein was also quantitated by an enzyme-linked immunosorbent assay procedure (14).
Measurement of Protein Concentrations-Protein concentrations were measured by the method of Bradford (24), using bovine serum albumin as a standard.
PDE Assays-These were performed as described previously (14,25). Phosphorylation of UCR1 by PKA-Purified MBP or MBP-UCR1 fusion protein was incubated for 30 min at 30°C in PKA phosphorylation buffer (100 mM Tris-HCl, pH 7.5, 0.2 mM ATP, 10 mM MgCl 2 , 30 mM 2-mercaptoethanol) supplemented with 1 unit of PKA catalytic subunit per 10 g of fusion protein. Mock phosphorylations were performed in an identical manner but without the PKA catalytic subunit. Amylose resin (50-l bed volume, equilibrated in KHEM/DTT/protease inhibitor plus 22 mM NaF, 11 mM Na 4 P 2 O 7 ⅐10H 2 O) was then added, and the reaction was incubated for 1 h, end-over-end at 4°C. The resin was then collected by centrifugation (2500 ϫ g for 5 s at room temperature), and the supernatant was discarded. The pelleted resin was held on ice and washed 3 times, each 15 min, with 330 l of KHEM/DTT/protease inhibitor/NaF/Na 4 In experiments where the incorporation of phosphate was monitored, 5 g of fusion protein was treated as above, but in a final volume of 20 l, and the reaction mixture was supplemented with 20 Ci of [␥-32 P]ATP. In experiments where the treated fusion protein was used in pull down assays, 50 g of protein was treated as above in a final volume of 350 l. The treated protein was then mixed with sufficient crude cytosolic extract from COS7 cells expressing UCR2ϩCat so as to contain 18 -20 units of PDE enzyme activity (pmol of cAMP hydrolyzed per min, with 1 M cAMP as substrate), in a final volume of 350 l of KHEM/DTT/protease inhibitor/NaF/Na 4 P 2 O 7 ⅐10H 2 O plus 10 M okadaic acid. Centrifugation and washing were then performed as described above.

Isolation of UCR2 as an Interacting Partner for UCR1 in a
Two-hybrid Screen-The two-hybrid screen is a powerful ge-netic approach for the isolation of cDNAs encoding proteins that interact with a protein of interest (26). We have used the two-hybrid system previously to identify the scaffold protein RACK1 as an interacting "partner" for the PDE4D5 isoform and to demonstrate that RACK1 interacts with the PDE4D5 unique extreme amino-terminal region (14). To isolate cDNAs encoding proteins that might interact with UCR1 or UCR2, we cloned various portions of the PDE4D3 cDNA into the twohybrid vector pLEXAN and used them as "baits" in two-hybrid screens (see "Experimental Procedures" for details). All the constructs that included the amino-terminal half of UCR1 produced very high background in two-hybrid assays and thus could not be used as baits in a screen. The reason for the high background associated with constructs containing this region of UCR1 is not clear. However, it is possible that the large number of charged/acidic residues in the amino-terminal half could produce transcriptional trans-activation activity when coupled to the LexA DNA-binding domain (i.e. even in the absence of the normal activation domain). Therefore, we prepared a bait, pLEXAU1D, incorporating only the carboxylterminal half of UCR1 (amino acids 80 -116 of PDE4D3, Fig.  1B), which produced sufficiently low background. A two-hybrid screen was performed with a HeLa cell cDNA library, as we have shown previously that several PDE4D species are present in HeLa cells (12). Of a total of 6 ϫ 10 6 recombinants that were screened, approximately 220 were HISϩ, and of these, 82 were also lacZϩ. Only 9 of the 82 lacZϩ recombinants were still positive after retrieval of the library plasmid from the yeast cells and subsequent re-testing for interaction with pLEXAU1D. Five of these nine remaining positives encoded non-physiologic fusions (i.e. the open reading frame incorporated in the GAL4 fusion was different from that of the protein encoded naturally by the cDNA), and all of these were discarded. All four remaining plasmids encoded an identical insert, consisting of a truncated PDE4D cDNA, encoding all of UCR2 and the catalytic region of the protein but none of UCR1. This suggested to us that UCR1, through its carboxyl-terminal region, potentially interacted with another region(s) of the PDE4 protein, located in UCR2 and/or the catalytic region. Furthermore, this interaction appeared to be highly specific, as no other cDNA in the two-hybrid library was isolated as a partner for pLEXAU1D.
To obtain supporting data that the interaction detected in the two-hybrid screen was specific, we used yeast two-hybrid ␤-galactosidase assays to test the interaction of the two-hybrid positive with a variety of baits expressed as LexA fusions. These included lamin (15), casein kinase II, Ras, Raf, several transcription factors, and the DNA-binding region of LexA itself (i.e. not as a fusion). In similar fashion, we tested pLEXAU1D for its ability to bind to these proteins expressed as GAL4 fusions and also to the GAL4 activation domain itself (i.e. not as a fusion). No interaction was detected under conditions that demonstrated an interaction between pLEXAU1D and the positive from the screen. Therefore, we felt it was unlikely that our results could be explained by nonspecific interactions between UCR1 and other proteins.
The Carboxyl-terminal Half of UCR1 Interacts with the Amino-terminal Third of UCR2-We wished to confirm the results of the two-hybrid screen and to determine precisely which region(s) of PDE4D3 interacted with the carboxyl-terminal half of UCR1. Therefore, we created constructs encoding various portions of the PDE4D3 cDNA as GAL4 activation domain fusions (Fig. 2a) and used yeast two-hybrid ␤-galactosidase assays to test them for interaction with the carboxyl-terminal half of UCR1. For these assays, we prepared a construct, pLEXAU1E, which encoded a LexA fusion with amino acids 80 -109 of PDE4D3, corresponding precisely to the UCR1 carboxyl-terminal half (Figs. 1B and 2a). The results of these assays (Fig. 2b) demonstrated that the UCR1 carboxyl-termi- FIG. 4. Specific amino acids (Arg-98 and Arg-101) in UCR1 mediate its interaction with UCR2. A, site-directed mutagenesis was used to mutate various codons in pLEXAU1D to alanine. The mutagenized constructs were then tested in a two-hybrid ␤-galactosidase filter assay (13) for their ability to interact with various regions of PDE4D3, expressed as GAL4 fusions (Fig. 2). nal half interacts with a small portion of the PDE4D protein, located within the amino-terminal half of UCR2 (amino acids 134 -180 of PDE4D3). Also of interest is that pLEXAU1E did not interact with any portion of PDE4D3 that contained UCR1, even if that portion also included UCR2 (Fig. 2b). We interpret these results as competition for LexA-UCR1 binding by the UCR1 already present in the UCR1-containing constructs. All these results were reproduced with pLEXAU1D (data not shown). We also showed that the isolated PDE4D UCR1 interacted with the UCR2 of PDE4B and other PDE4 isoforms, provided that UCR1 was not present in these constructs (data not shown).
UCR1 and UCR2 Interact in Vitro-We wished to provide an independent demonstration of the UCR1-UCR2 interaction, using biochemical approaches. Therefore, we generated fusion proteins of UCR1 and demonstrated that they interacted with UCR2 by "pull down" assays. We have previously exploited this approach to demonstrate the interaction of the PDE4A4/5 isoform with proteins containing SH3 domains, particularly SRC family tyrosyl kinases, and to demonstrate the interaction of PDE4D5 and RACK1 (14,21,27). For these studies, we generated MBP-UCR1, a fusion between MBP and all of UCR1 (amino acids 17-136 of PDE4D3) and also MBP-UCR1-C, a fusion between MBP and the carboxyl-terminal half of UCR1 (amino acids 80 -116 of PDE4D3, corresponding to the region encoded in pLEXAU1D). These fusion proteins were expressed in E. coli and then purified on affinity columns. They exhibited apparent molecular masses under denaturing conditions of 77 Ϯ 1.61 kDa for MBP-UCR1 and 75.5 Ϯ 2.8 kDa for MBP-UCR1-C, compared with 67 Ϯ 1.0 kDa for MBP.
We then used these MBP fusions in pull down experiments with constructs encoding various portions of UCR2. For this purpose, we expressed truncated forms of PDE4D3 by transient transfection in COS7 cells. We have used this system extensively in the past to study the properties of PDE4D isoforms (10,12,14). We expressed full-length PDE4D3 and various amino-terminal truncations of PDE4D3 (Fig. 2a) in COS7 cells and tested them for the ability to interact with MBP-UCR1 or MBP-UCR1-C in pull down assays (14,21,27). The ability of MBP-UCR1 or MBP-UCR1-C to pull down the PDE4D3 truncations was measured by both immunoblotting (Fig. 3, a and b) and quantitative PDE activity assays (Fig. 3c). The data demonstrate that MBP-UCR1 and MBP-UCR1-C were able to bind to UCR2ϩCat, a PDE4D3 truncation that lacked UCR1 but still contained UCR2. In contrast, they were unable to bind to full-length PDE4D3 or the isolated catalytic region (Cat) of PDE4D3 (Fig. 3, a-c). In separate experiments, a construct expressing only UCR2 as a GST fusion was expressed and demonstrated to be able to bind to both MBP-UCR1 and MBP-UCR1-C (Fig. 3d). These data confirm the two-hybrid data described above. In particular, they confirm that UCR1 can only interact with UCR2 when UCR1 is not present in the UCR2-containing construct. In addition, they demonstrate that full-length UCR1 can interact with UCR2.
Two Charged Amino Acids within UCR1 Are Necessary for Its Interaction with UCR2-We wished to determine which specific amino acids in the carboxyl-terminal region of UCR1 were necessary for its interaction with UCR2. The carboxylterminal half of UCR1 (amino acids 80 -109 of PDE4D3; Fig.  1B) is highly hydrophobic. A few highly charged amino acids (e.g. Asp-83, Asp-84, Arg-98, and Arg-101) are embedded in this region, however, and these are highly conserved among all the mammalian PDE4 isoforms and in D. melanogaster (Fig. 1B). Helical wheel analysis suggested that the carboxyl-terminal region of UCR1 may form an amphipathic helix, with the charged groups on a discrete side (2). The amino-terminal region of UCR2 (amino acids 134 -180 of PDE4D3; Fig. 1B) is very hydrophilic and contains numerous conserved negatively charged amino acids. It is likely that this region of UCR2 would interact with the charged amino acids in UCR1. Therefore, site-directed mutagenesis was used to mutate the codons for Asp-83, Asp-84, Arg-98, and Arg-101 in pLEXAU1D to alanine, separately and in various combinations, and two-hybrid assays were used to test the mutants for their ability to interact with UCR2 (Fig. 4a). The results demonstrate that the R98A/R101A double mutant completely blocked the interaction of pLEXAU1D with UCR2 and UCR2-N (Fig. 4a). In contrast, the double mutant D83A/D84A produced no detectable change in the interaction (Fig. 4a). The single R101A mutant also appeared to block the signal produced with UCR2ϩCat but not with UCR2 or UCR2-N. The results obtained with the R101A mutant may reflect a partial attenuation of the interaction by this mutant, combined with lower expression of the UCR2ϩCat construct (compared with the smaller UCR2 or UCR-N constructs) in cells.
These results were confirmed by pull down assays (Fig. 4, b  and c). The interaction was demonstrated by immunoblotting for the bound PDE4D3 truncation (Fig. 4, b and c) or with quantitative PDE assays (Fig. 4c). The R98A and R101A mutants each reduced the ability of MBP-UCR1 to interact with UCR2ϩCat, whereas the double R98A/R101A mutant almost totally ablated its interaction with UCR2ϩCat (Fig. 4, b and c). These data confirm the model that the bulky, positively charged Arg-98 and Arg-101 project out from a generally hydrophobic region in UCR1 and are necessary for its interaction with UCR2.
Specifically Charged Amino Acids in UCR2 Are Required for Its Interaction with UCR1-We then attempted to determine   in UCR2 attenuate its interaction with UCR1 Site-directed mutagenesis was used to mutate various codons in pGADNU2J (UCR2-N (W.T.)) to alanine. The mutagenized constructs (listed in the left-hand column) were then tested by two-hybrid quantitative ␤-galactosidase assay for their ability to interact with pLEXAU1D (UCR1-C (W.T.)) or pLEXAU1D containing the mutations R98A or R101A. Vector, pLEXAN or pGADN. Values are the mean Ϯ S.D. for assays performed in triplicate. which amino acids in UCR2 are necessary for it to interact with UCR1. Site-directed mutagenesis was used to mutate various combinations of glutamic acids or aspartic acids within the amino terminus of UCR2 to alanine (Fig. 1B). The resulting mutants were then tested for their ability to interact with pLEXAU1D, using two-hybrid quantitative ␤-galactosidase assays (Table I). The results demonstrate that the simultaneous mutation of three amino acids (Glu-146, Glu-147, and Asp-149) in UCR2 significantly attenuated its interaction with UCR1. Mutation of the pair Asp-153/Glu-153 to alanine also appeared to attenuate the interaction. In contrast, mutation of the pair Glu-143/Glu-146 had no detectable effect on the interaction.
We then wished to determine if the positively charged aspartic acids (Arg-98 and Arg-101) in UCR1-C interacted with specific negatively charged glutamic/aspartic acid residues (Glu-146, Glu-147, and Asp-149) in UCR2. Therefore, we tested the ability of the single R98A and R101A mutants to block the interaction, when paired with various combinations of E146A, E147A, and D149A mutants. The results (Table I)  D149A completely blocked the interaction, whereas the corresponding single mutants did not. As the R101A mutant interacted less well with UCR2-N than did R98A (see also Fig. 4a), the magnitude of the interaction obtained with the R101A/ D149A mutant must be interpreted cautiously. However, the data are consistent with Arg-98 and/or Arg-101 interacting directly with Asp-146 and/or Asp-47 and/or Asp-149 and that this interaction is necessary for the interaction of UCR1 with UCR2. In conclusion, our data strongly suggest that UCR1 and UCR2 interact by electrostatic interactions between the positively charged arginines in the carboxyl-half of UCR1 and the negatively charged glutamic acid/aspartic acid residues in the amino-terminal end of UCR2.
Phosphorylation of UCR1 by PKA Attenuates Its Ability to Interact with UCR2-PDE4D3 is activated upon phosphorylation by PKA (7)(8)(9)(10). Phosphorylation of PDE4D3 occurs at two serines, both within the consensus RRXS: Ser-13, located in its unique amino-terminal region, and Ser-54, which is located at the beginning of UCR1 (Fig. 1B). The phosphorylation at Ser-54 leads to both enzyme activation and a change in sensitivity to inhibition by rolipram (9,10). We now demonstrate that phosphorylation of UCR1 by PKA attenuates its ability to interact with UCR2. First, we demonstrated that MBP-UCR1 was a PKA substrate in vitro (Fig. 5a). This phosphorylation was ablated if the PKA consensus site (RRES 54 ; Fig. 1B) was disrupted by mutation to AAES (i.e. MBP-UCR1R51A/R52A). This mutation had been shown by us previously (10) to prevent PKA from phosphorylating and activating PDE4D3 and not to change either PDE activity or sensitivity to inhibition by rolipram. We also demonstrated that PKA was unable to phosphorylate UCR1 with a S54A mutation (i.e. MBP-UCR1S54A; data not shown). We then tested the ability of PKA-phosphorylated MBP-UCR1 to interact with UCR2ϩCat. Treatment of MBP-UCR1 with PKA produced a marked attenuation in its ability to interact with UCR2ϩCat in pull down assays (Fig. 5, b and  c). In contrast, no attenuation was seen with PKA-treated MBP-UCR1R51A/R52A.
Mutation of Ser-54 to aspartic acid (S54D) of PDE4D3 completely mimics both enzyme activation and the change in rolipram sensitivity produced by PKA phosphorylation (10). We wished to determine if this mutation affected the UCR1-UCR2 interaction. We created the S54D mutant in MBP-UCR1 and demonstrated in a pull down assay that it profoundly reduced the ability of UCR1 to interact with UCR2ϩCat (Fig. 5, d and  e). This confirms that phosphorylation of UCR1 by PKA could attenuate its ability to interact with UCR1.
We also tested the effects of several other mutations in the PKA consensus site (Fig. 5, c-e). We have shown previously that the S54A mutation in PDE4D3 had no effect on enzyme activity (9, 10). However, it caused a conformational change in the PDE4D3 catalytic region, as determined by increased sensitivity of the enzyme to inhibition by rolipram (10). The conservative S54T mutation did not change the sensitivity of the enzyme to rolipram (10). Conversely, mutation of Glu-53 to alanine mimicked the activation of PDE4D3 by PKA phosphorylation but did not alter sensitivity to inhibition by rolipram (10). We created these three mutations in MBP-UCR1 and showed by pull down assay that the S54A mutation prevented the interaction of UCR1 with UCR2, whereas the S54T and E53A mutations did not affect the interaction (Fig. 5, d and e). These data suggest that the UCR1-UCR2 interaction could produce conformational changes in the catalytic region that would change the sensitivity of the enzyme to rolipram (see "Discussion").
Disruption of the PKA consensus phosphorylation site by mutation of RRES to AAES (MBP-UCR1R51A/R52A) had no effect on the UCR1-UCR2 interaction, as measured by pull down assays (Fig. 5, d and e). For this reason, we used the MBP-UCR1R51A/R52A construct as a control for examining the effect of direct PKA phosphorylation of UCR1 (see above; Fig. 5, a-c).

DISCUSSION
UCR1 and UCR2 are regions of sequence that are unique "signatures" of the PDE4 cAMP-specific phosphodiesterase family. They are located in the amino-terminal regions of these enzymes and are clearly separate from their catalytic regions. They are highly conserved in evolution, as they are present in isoforms encoded by all four human and rat PDE4 genes and also in PDE genes in organisms as distantly related as D. melanogaster (4) and C. elegans (6). Studies of PDE4 deletion constructs have demonstrated that UCR1 and UCR2 have no direct role in catalysis (reviewed in Ref. 2). However, regions of amino acid sequence that are highly conserved in evolution are usually of functional significance. To determine more about the properties of UCR1, it was used in a two-hybrid screen. Intriguingly, the screen demonstrated that UCR1 interacted with UCR2. Additional two-hybrid tests demonstrated that the carboxyl-terminal region of UCR1 interacted with the amino-terminal end of UCR2 and that this interaction was dependent on specifically charged amino acids within these regions. The interaction was confirmed and further characterized by pull down assays using UCR1 and UCR2 expressed as fusion proteins.
Our data are consistent with the concept that both UCR1 and UCR2 act as authentic protein domains. In other words, UCR1 and UCR2 can each fold into independent modules that appear to retain their structure in a variety of contexts. This is demonstrated by our data in which UCR1 interacts with UCR2 when both are expressed as one of several different fusions (e.g. the LexA DNA-binding domain or MBP for UCR1, and the GAL4 activation domain or GST for UCR2). This behavior of UCR1 and UCR2 is reminiscent of SH2 and SH3 regions, which also form self-folding domains able to participate in proteinprotein interactions (21,(27)(28)(29).
Our data provide additional insight into the mechanism of regulation of PDE4D3 by PKA phosphorylation. Conti and coworkers (7)(8)(9) have demonstrated that the PDE4D3 isoform can be phosphorylated by PKA in vivo and that the critical phosphorylation site is Ser-54, which is located in the RRES motif at the start of UCR1 (7,9). Phosphorylation of Ser-54 in PDE4D3 changes the enzymatic activity (relative V max ) and shifts the dose-response curve for rolipram (7-10). However, analysis of PDE4D3 mutants in the region of the Ser-54 PKA phosphorylation site led us to propose a model that implied FIG. 6. A model for the UCR1-UCR2 interaction. Various regions of the PDE4D3 isoform are shown, with the interactions between various domains depicted schematically. NT refers to the extreme aminoterminal region. UCR1 is divided into two domains, with the aminoterminal domain containing the PKA phosphorylation site, and the carboxyl-terminal domain interacting with UCR2. LR1 and LR2 refer to linker regions 1 and 2, which connect UCR1 with UCR2, and UCR2 with the catalytic region, respectively. UCR2 is divided into several potential domains (2), of which only the amino-terminal domain is necessary for interaction with UCR1. Cross-hatched areas indicate the positions of amino acids essential for the UCR1-UCR2 interaction.
that Ser-54 and Glu-53 were involved in forming bonds that influenced the structure of the catalytic unit of the enzyme (10). In this model, Glu-53 was involved in forming an ion pair which kept the enzyme in a low activity state and whose disruption (e.g. the E53A mutant) changed the activity of the enzyme but did not alter the sensitivity of the enzyme to inhibition by the PDE4-selective inhibitor rolipram. The model also suggested that the side chain hydroxyl group of Ser-54 was involved in forming a hydrogen bond that did not alter enzyme activity but influenced the conformation of the enzyme catalytic region, as measured by an increase in susceptibility of the enzyme to inhibition by rolipram. Phosphorylation of Ser-54 (or the S54D mutation) disrupted both the Glu-53 and Ser-54 bonds, leading to enzyme activation and enhanced sensitivity to inhibition by rolipram. However, the S54A mutation blocked only the Ser-54 bond, producing a change only in rolipram inhibition. In the present paper, we demonstrate that the S54D and S54A mutants, but not the E53A or S54T mutants, block the UCR1-UCR2 interaction. Therefore, it is likely that the Ser-54 hydrogen bond, but not the Glu-53 ion pair, is essential for the UCR1-UCR2 interaction. In turn, this suggests that disruption of the UCR1-UCR2 interaction leads to an alteration in the conformation of the catalytic unit that is detected by altered rolipram inhibition but that is insufficient for enzyme activation.
We present a model for the structure of the PDE4 enzyme, based on our data (Fig. 6). The model proposes that UCR1 and UCR2 each contain one or more self-folding domains that interact with other regions of the PDE4 enzymes. One important interaction, as demonstrated by our data, is between the carboxyl-terminal region of UCR1 and the amino-terminal region of UCR2. This interaction is mediated by electrostatic interactions between positively charged amino acids in UCR1 and negatively charged amino acids in UCR2, as demonstrated by our mutational analysis. The interaction may be modulated by PKA phosphorylation of a serine in the extreme amino-terminal region of UCR1 (i.e. a region not interacting directly with UCR2). This is compatible with the UCR1 amino-terminal domain affecting the UCR1 carboxyl-terminal region and preventing it from interacting with UCR2. UCR1 and UCR2 thus appear to form a regulatory module that in turn regulates the PDE4 catalytic unit. Structural analysis of the PDE4 enzyme will be needed for further insight into the molecular mechanisms of this regulation.